Scalable droplet-based radiosynthesis of [¹⁸F] fluorobenzyltriphenylphosphonium cation ([¹⁸F] FBnTP) via a “numbering up” approach

Abstract

The [¹⁸F]fluorobenzyltriphenylphosphonium cation ([¹⁸F]FBnTP) has emerged as a highly promising positron emission tomography (PET) tracer for myocardial perfusion imaging (MPI) due to its uniform distribution in the myocardium and favorable organ biodistribution demonstrated in preclinical studies. However, a complex and low-efficiency radiosynthesis procedure has significantly hindered its broader preclinical and clinical explorations. Recently, Zhang et al. developed a pinacolyl arylboronate precursor, enabling a one-step synthesis process that greatly streamlines the production of [¹⁸F]FBnTP. Building upon this progress, our group successfully adapted the approach to a microdroplet reaction format and demonstrated improved radiosynthesis performance in a preliminary optimization study. However, scaling up to clinical dose amounts was not explored. In this work, we demonstrate that scale-up can be performed in a straightforward manner using a “numbering up” strategy (i.e. performing multiple droplet reactions in parallel and pooling the crude products). The resulting radiochemical yield after purification and formulation was high, up to 66 ± 1% (n = 4) for a set of experiments involving pooling of 4 droplet reactions, accompanied by excellent radiochemical purity (>99%) and molar activity (339–710 GBq μmol⁻¹). Notably, we efficiently achieved sufficient activity yield (0.76–1.84 GBq) for multiple clinical doses from 1.6 to 3.7 GBq of [¹⁸F]fluoride in just 37–47 min.

Concept graphic

Rapid, efficient and straightforward radiosynthesis scale-up strategy for droplet-based reactions via a numbering up technique.

1. Introduction

The recent World Health Organization report highlights ischemic heart disease as the leading cause of global mortality, causing 8.9 million deaths in 2019.¹ Early and precise detection of cardiac ischemia is crucial, enabling timely consideration of appropriate therapy and reducing the risk of disease progression. Utilizing molecular imaging modalities like positron emission tomography (PET) and single-photon emission computed tomography (SPECT), myocardial perfusion imaging (MPI) emerges as a powerful non-invasive tool for early detection and disease monitoring of cardiac ischemia.^2–6

In the United States, the Food and Drug Administration (FDA) has approved six MPI radiotracers, including four SPECT tracers ([^99mTc]Tc-teboroxime, [^99mTc]Tc-sestamibi, and [^99mTc]Tc-tetrofosmin all with half-life t_1/2 = 6.04 h, and [²⁰¹Tl]thallium chloride with t_1/2 = 73.1 h) and two PET tracers ([¹³N]NH₃ (t_1/2 = 10 min) and [⁸²Rb]Rb-chloride (t_1/2 = 1.27 min)). Despite the numerous advantages of PET over SPECT, such as high spatial resolution, attenuation correction, sensitivity, and quantitation, SPECT tracers continue to play a central role in clinical use, mainly due to the limited accessibility of MPI PET tracers.^2,3,5–7 Challenges of using [⁸²Rb]Rb-chloride include its ultrashort half-life, low first-pass extraction (∼65% at rest), high positron range (2.6 mm), and the high cost of monthly generator replacement,^3,4,6 and [¹³N] NH₃ is restricted by the requirement for an on-site cyclotron for production, significantly limiting its availability. PET imaging with F-18 presents an alternative that can overcome these limitations and provide several benefits, like longer half-life (t_1/2 = 109.8 min) enabling greater flexibility in study design, lower injected activity requirement due to low positron energy and high positron yield, and the feasibility of using exercise for stress imaging (in contrast to short half-life tracers that only permit pharmacological stress).⁴

Two types of ¹⁸F-labeled radiotracers for MPI are currently gaining significant attention: [¹⁸F]flurpiridaz and ¹⁸F-labeled phosphonium cations.^4,7 [¹⁸F]flurpiridaz, a pyridaben analogue, exhibits high binding affinity towards mitochondrial complex I, and recent clinical phase III trials have revealed its improved diagnostic performance characteristics in patients with suspected coronary artery disease (CAD) when compared to the ^99mTc-labeled SPECT MPI agents.⁸ For several ¹⁸F-labeled phosphonium cations, including 4-[¹⁸F]fluorobenzyltriphenylphosphonium cation ([¹⁸F]FBnTP), (5-¹⁸F-fluoropentyl)triphenylphosphonium cation ([¹⁸F]FPTP), (6-¹⁸F-fluorohexyl)triphenylphosphonium cation ([¹⁸F]FHTP), and (2-(2-¹⁸F-fluoroethoxy)ethyl) triphenylphosphonium cation ([¹⁸F]FETP), promising potential for MPI is envisioned by monitoring myocardial mitochondria potential.^9,10 Particularly noteworthy, [¹⁸F] FBnTP has demonstrated uniform distribution in the myocardium and favorable organ biodistribution, and superior diagnostic performance in comparison to [^99mTc]Tc-tetrofosmin for assessing the severity of coronary artery stenosis,^11–14 rendering it a promising candidate for diverse MPI applications. However, further clinical trials are imperative to confirm the suitability of [¹⁸F]FBnTP for routine human use, but such studies are in part hindered by the complex and low-efficiency radiosynthesis procedure.

The original method, reported by Ravert et al. in 2004,⁹ involved a demanding 4-step manual process that required large amounts of precursor (20 μmol for fluorination) and reagents (20–7930 μmol per step), involved high corrosive reagents (HBr), and had low activity yield (6%) and long preparation time (82 min). Using microwave activation, Ravert et al. later showed the synthesis could be performed more quickly (52 min) and the activity yield increased to 8.3%;¹⁵ however the requirement for a custom synthesis module was limiting. Further improvements were made by Waldmann et al. in 2018, including automation on a commercially-available synthesis module (ELIXYS FLEX/ CHEM, Sofie Inc., Dulles, VA, USA), as well as an improvement in activity yield (16%),¹⁶ but the complex synthesis route remained a challenge for routine preclinical and clinical studies in most radiochemistry labs. Tominaga et al. later reduced the reaction steps from four to three, though critical information such as activity yield, molar activity, and synthesis time were not disclosed.¹¹ More recently, Zhang et al. introduced a vastly-simplified one-step preparation of [¹⁸F]FBnTP through Cu-mediated radiofluorination of a pinacolyl arylboronate precursor.¹⁷ This method substantially streamlined the radiosynthesis and exhibited high radiochemical conversion (RCC) of 62 ± 1.4% (n = 2), though the overall synthesis performance was not disclosed. Recently, Lin et al. reported an automated version of this simplified synthesis route using a commercial module (GE Tracerlab FXN),¹⁸ however the overall synthesis exhibited a relatively low activity yield of 6.2 ± 1.0% (n = 7) and molar activity of 6.4 ± 1.0 GBq μmol⁻¹ (n = 7).

Recently, we showed that droplet-based radiochemistry approaches could be leveraged to substantially improve Cu-mediated radiosynthesis of [¹⁸F]FDOPA (ref. 19) and a novel monoacylglycerol lipase (MAGL) ligand, [¹⁸F]YH-149.²⁰ Droplet radiochemistry offers advantages of minimal reagent cost, rapid synthesis time, high yield, high molar activity, and low space and infrastructure requirements. A further advantage of droplet radiochemistry is the ability to perform high-throughput optimization via arrays of droplet reactions performed in parallel.^21,22 Using a newly developed robotic platform, we used this technique to develop a preliminary droplet-based radiosynthesis of [¹⁸F]FBnTP,²³ resulting in substantial reduction in reagent usage and enhancement in radiosynthesis performance (RCC: 89 ± 1%, n = 4). Following purification and formulation, [¹⁸F]FBnTP was produced with high isolated radiochemical yield (RCY, 66 ± 6%, n = 3) within 42 min, corresponding to an activity yield of 49 ± 3% (n = 3).

In this work, we aimed to establish the clinical relevance of the previous result by scaling up the droplet-based production of [¹⁸F]FBnTP. Previously, we have shown two different approaches for scale-up of droplet reactions: (1) accumulating [¹⁸F]fluoride at a reaction site by depositing a small aliquot of the [¹⁸F]fluoride solution, evaporating the liquid, and repeating (Fig. 1A)^24,25 and (2) pre-concentrating the [¹⁸F]fluoride using a trap-elute process on a miniature cartridge, enabling a greater amount of activity to be loaded to a reaction site (Fig. 1B).²⁶ While the first approach is straightforward and suitable for moderate scale-up, handling very large activity amounts and volumes becomes impractical due to extended evaporation times at a single reaction site. In addition, a modest drop in RCY was observed, potentially due to the increased amount of impurities present when using large volumes of radioisotope source solution.²² Conversely, the second approach effectively worked with much larger volumes and avoids the build-up of impurities, but required a more complex setup. Furthermore, this approach requires optimization of the [¹⁸F]fluoride elution protocol for each radiotracer because the type and amount of phase transfer catalyst (PTC) and base needed for efficient elution can impact the subsequent radiotracer synthesis. Moreover, reductions in RCY were observed at higher activity levels, potentially attributable to radiolysis and/or other factors.²⁶

Fig. 1.

Approaches for scale-up of radiopharmaceutical product amount in droplet-based radiosynthesis. (A) Starting activity for a single droplet reaction is increased by repeated loading and evaporation of [¹⁸F]fluoride aliquots on the droplet reaction chip prior to the fluorination reaction. (B) [¹⁸F]fluoride is pre-concentrated using a miniature cartridge into a final volume that is compatible with a single reaction site. (C) Multiple reaction sites are loaded with 20–30 μL of (unconcentrated) [¹⁸F]fluoride and multiple droplet reactions are conducted in parallel. The crude reaction products are pooled prior to purification to increase the total product activity.

To address these challenges, we developed an alternative scale-up method based on the concept of “numbering up”, in which multiple droplet reactions are conducted in parallel and pooled together to increase the product quantity (Fig. 1C). This novel approach is faster than the other approaches because it eliminates the need to process the [¹⁸F]fluoride ahead of the reactions, and because each individual reaction is performed at smaller scale, issues due to radiolysis or impurities in the radioisotope source are eliminated. Numbering up provides a rapid path to scale-up, minimizing the effort and cost spend to transition from optimization of droplet-based reactions (at low activity scales) to larger scale production. We demonstrate that this approach can be used to conduct production of [¹⁸F]FBnTP at clinically-relevant levels in a rapid and high-yield manner.

2. Experimental

2.1. Materials

Cesium carbonate (Cs₂CO₃, 99%), potassium carbonate (K₂CO₃, >99%), potassium trifluoromethanesulfonate (KOTf, 98%), anhydrous pyridine (Py, 99.8%), anhydrous methanol (MeOH, 99.8%), dichloromethane (DCM, >99.8%), anhydrous N,N-dimethylformamide (DMF, 99.8%), 1,3-dimethyl-2-imidazolidinone (DMI, >99.5%), trifluoroacetic acid (TFA, 99%), copper(II) trifluoromethanesulfonate (Cu(OTf)₂, 98%), and tetrakis(pyridine)copper(II) triflate (Cu(OTf)₂(Py)₄, 95%) were purchased from Sigma-Aldrich (St. Louis, MO, USA). Tetraethylammonium trifluoromethanesulfonate (TEAOTf, >99%) was purchased from TCI America (Portland, Oregon, USA). Precursor and reference standard were prepared as described previously.¹⁷ Deionized (DI) water was obtained from a Milli-Q water purification system (EMD Millipore Corporation, Berlin, Germany). Reagent and collection vials (0.5 mL, PCR clean) were purchased from Eppendorf (Hamburg, Germany). Acetonitrile for high performance liquid chromatography (HPLC) was purchased from Fisher Scientific (Pittsburgh, PA, USA). C18 Plus Short cartridges (WAT020515) were purchased from Waters Corporation (Milford, MA, USA). 50 mL polypropylene centrifuge tubes were purchased from Corning Inc. (430 304, Corning, NY, USA). [¹⁸F]fluoride (typically 32–44 GBq) in ∼1.2 mL of [¹⁸O] H₂O was obtained from the UCLA Crump Cyclotron and Radiochemistry Center. Portions of this activity were used directly without further purification for all microscale radiosynthesis.

2.2. Droplet-based [¹⁸F]FBnTP synthesis

Droplet-based reactions were conducted on Teflon-coated silicon chips, featuring 2 × 2 or 3 × 3 arrays of circular (4 mm diameter) hydrophilic reaction sites (Fig. S1^†). These chips were operated on a temperature-controlled heating platform, as previously described.²¹ The general synthesis process (Fig. 1C) involved the following steps: first, 10–47 μL of a [¹⁸F]fluoride stock solution containing 25–1510 MBq of activity mixed with a desired amount of phase-transfer catalyst (PTC) and base, was added via micropipette onto a reaction site of the chip. The droplet was then dried at 105 °C for 1–2 min. Next, 10 μL of a precursor/Cu(OTf)₂(py)₄ stock solution was added and heated at 110 °C for 5 min to facilitate fluorination. After completion of the reaction, the crude product was extracted from the reaction site by adding a collection solution (20 μL) and transferring it to a 0.5 mL Eppendorf tube for further analysis. To ensure minimal activity residue on the chip, the collection step was repeated a total of 4 times.

Several stock solutions were prepared just prior to each set of experiments. The stock solution of PTC and base was prepared in DI water, with a 5 μL aliquot containing 0.3 μmol of TEAOTf and 0.01 μmol of Cs₂CO₃ for a single droplet reaction unless otherwise indicated. [¹⁸F]Fluoride stock solution was prepared by mixing a desired volume (5–42 μL) of [¹⁸F]fluoride/[¹⁸O]H₂O (containing 25.3–1510 MBq of activity) with a 5 μL aliquot of PTC/base stock solution. Individual stock solutions of the precursor (with varied concentration based on precursor amount studies) and Cu(OTf)₂(Py)₄ (136 mM) were prepared in the desired reaction solvent mixture, and then these stock solutions were mixed in a 1 : 1 (v/v) ratio just before synthesis, such that each 10 μL portion of the mixed solution contained the desired amount of precursor and 0.68 μmol of Cu(OTf)₂(Py)₄. The collection solution was prepared by mixing MeCN and DI water (35 : 65, v/v) with 0.1% TFA (v/v), matching the mobile phase used for HPLC purification.

When performing scaled-up synthesis, the single droplet process was repeated at multiple reaction sites on the same chip. For these reactions, the crude product was collected with smaller aliquots of collection solution (i.e. 10 μL × 4 instead of 20 μL × 4). For example, when performing two reactions in parallel, the total volume of the pooled crude products was ∼80 μL.

2.3. Analytical methods

Radioactivity measurements were performed using a calibrated dose calibrator (CRC-25R, Capintec, Florham Park, NJ, USA). To assess RCC, we employed multi-lane radio-thin layer chromatography (radio-TLC) methods.²⁷ Briefly, 0.5 μL of samples were spotted on TLC plates (6 cm × 5 cm pieces cut from 20 cm × 5 cm sheets, silica gel 60 F₂₅₄, Merck KGaA, Darmstadt, Germany). These plates were then developed for 4 cm using a mobile phase of DCM and MeOH (9 : 2, v/v), dried, and then covered with a glass microscope slide (75 × 50 × 1 mm³, Fisher Scientific, Hampton, NH, USA) for readout via Cerenkov luminescence imaging (CLI) with a 5 min exposure time. The RCC of each sample (lane) was determined via region of interest (ROI) analysis as previously described.²⁷ The collection efficiency was obtained by dividing the activity of the product mixture collected from the microdroplet reactor by the starting activity (corrected for decay). The crude RCY was computed as RCC multiplied by the collection efficiency. To determine RCY, radio-HPLC purification was performed using an analytical column (ZORBAX RP Eclipse Plus C18, 100 × 4.6 mm, 3.5 μm, Agilent Technologies, Santa Clara, CA, USA) using an isocratic mobile phase of DI water and MeCN (65 : 35, v/v) with 0.1% TFA (v/v) at a flow rate of 1.2 mL min⁻¹. For some experiments, purification was performed on an semi-prep column (C18 Gemini-NX, 250 × 10 mm, 5 μm, Phenomenex, Torrance, CA, USA) using isocratic mobile phase of DI water and MeCN (60 : 40, v/v) with 0.1% TFA (v/v) at a flow rate of 5 mL min⁻¹. This was followed by formulation via a C18 plus short cartridge (preconditioned with 3 mL of EtOH and then 20 mL of DI water). The radio-HPLC system (Smartline, Knauer, Berlin, Germany) was equipped with a degasser (Model 5050), pump (Model 1000), UV detector (254 nm; Eckert & Ziegler, Berlin, Germany), and a gamma-radiation detector and counter (BFC-4100 and BFC-1000, Bioscan, Inc., Poway, CA, USA). To confirm the radiochemical purity (RCP), we analyzed the formulated [¹⁸F]FBnTP on the same analytical radio-HPLC system using a mobile phase of DI water and MeCN (60 : 40, v/v) with 0.1% TFA (v/v) at a flow rate of 1.2 mL min⁻¹. Co-injection of the formulated [¹⁸F]FBnTP with reference standard was performed to validate product identity.

3. Results and discussion

3.1. Preliminary development of droplet-based conditions

To enable high-throughput exploration of reactions on 4 mm diameter reaction sites, we developed a new chip with a 3 × 3 array of reaction sites (Fig. S1^†). As a starting point for the droplet-based synthesis of [¹⁸F]FBnTP, we conducted experiments using four sets of conditions. First, we scaled down the macroscale synthesis method described by Zhang et al.,¹⁷ from 850 μL to 10 μL, reducing reagents by ∼27× (condition 1, ESI† section 2). Second and third, we employed our previously reported droplet-based conditions for the Cu-mediated synthesis of [¹⁸F]FDOPA (ref. 19) but substituted the [¹⁸F]FBnTP precursor (conditions 2 and 3, ESI† Section 2). For condition 2, the precursor amount (0.15 μmol) was set to match condition 1. For condition 3, the precursor amount (0.45 μmol) matched our prior work with [¹⁸F]FDOPA.¹⁹ Additionally, we used our previous preliminary droplet conditions for [¹⁸F]FBnTP, but performed reactions on 4 mm reaction sites instead of 3 mm sites²³ (condition 4). Comprehensive details of reaction conditions and summary of performance can be found in Table S1.^†

Surprisingly, our attempts to produce [¹⁸F]FBnTP using condition 1 did not yield any product (n = 3). This could potentially be attributed to the fast degradation of Cu(OTf)₂ due to exposure to atmosphere in the open droplet reaction format, or low effectiveness of Cu(OTf)₂ to promote the fluorination in a droplet reaction. In our previous report,²⁰ the preparation of [¹⁸F]YH149 via a similar Cu-mediated route but using the copper reagent Cu(Py)₄(OTf)₂ resulted in a RCC of 0% (n = 2) in the absence of pyridine. Interestingly, in the current study, the synthesis of [¹⁸F]FBnTP using the copper reagent Cu(OTf)₂ even with the addition of pyridine led to a similar outcome (no conversion). This observation might suggest that Cu(Py)₄(OTf)₂ and pyridine could be a critical pair of reagents necessary for forming the radiofluorinated product.

After switching the PTC/base to Cu(Py)₄(OTf)₂/Cs₂CO₃ (condition 2), we observed a small amount of product formation, but with a poor crude RCY of only 8 ± 0% (n = 3), due to both low RCC (25 ± 1%, n = 3) and low collection efficiency (32 ± 1%, n = 3). Increasing the precursor amount from 0.15 to 0.45 μmol (i.e. condition 3) resulted in a significant improvement in RCC (53 ± 8%, n = 3), but the collection efficiency remained low (33 ± 1%, n = 3), resulting in only a moderate improvement in crude RCY (17 ± 3%, n = 3).

In contrast, when taking conditions from our previous high-throughput optimization study (condition 4), but performing the reaction on a 4 mm reaction site, the performance was significantly improved, with high RCC (92 ± 1%, n = 3) and collection efficiency (90 ± 1%, n = 3), corresponding to a high crude RCY of 83 ± 2% (n = 3), similar to the performance observed when using 3 mm diameter reaction sizes previously (Table S2^†).

3.2. Influence of precursor amount

According to the initial experiments with condition 2 and 3, the precursor amount showed a significant effect on the synthesis performance of [¹⁸F]FBnTP. Therefore, we performed further optimization to investigate the influence of the precursor amount. Details of measurements and calculations can be found in Table S3^† and results are summarized in Fig. 2. Excellent performance was achieved even with a small amount of precursor (even at the lowest amount tested, i.e., 0.15 μmol). We observed that an increased amount of precursor led to a slight increase in RCC and no significant change in collection efficiency, and thus a slight increase in crude RCY. The highest performance was observed for 0.45 and 0.60 μmol of precursor, giving crude RCY of 88 ± 3% (n = 3) and 92 ± 2% (n = 3), respectively, and these amounts were used in further studies.

Fig. 2.

(A) [¹⁸F]FBnTP synthesis scheme through one-step Cu-mediated fluorination of a pinacolyl arylboronate precursor. (B) Influence of precursor amount on the performance of the droplet radiosynthesis of [¹⁸F]FBnTP. Each experiment was repeated n = 3 times.

3.3. Influence of volume (and activity) of [¹⁸F]fluoride

In our previous report on scaling up the droplet synthesis of [¹⁸F]FET and [¹⁸F]FBB,²⁵ we observed a reduction in the performance of droplet reactions when higher starting activity was used. The decrease was attributed to multiple potential factors: (i) impurities in the [¹⁸F]fluoride source (e.g. anionic impurities and metal contaminants); (ii) radiolysis; (iii) stoichiometric differences in the reaction or degradation processes due to lower precursor : fluoride ratio. Though we have not completed our studies, there is significant evidence that cationic impurities present in the [¹⁸F]fluoride source (e.g. contaminants originally in the [¹⁸O] water, or metal ions released from the target body or foil) play a significant part in this phenomenon.²⁸

In the present work, we first explored the possibility of scaling up the synthesis of [¹⁸F]FBnTP by loading larger amounts of [¹⁸F]fluoride. To eliminate the potential impact of radiolysis, we first performed a study where the volume of [¹⁸F]fluoride used for a reaction was varied (5 to 40 μL), but activity level were kept low (11.7 to 69.6 MBq) where there is no impact of radiolysis. To maintain a relatively low activity level for higher volumes, the activity was allowed to decay for different amounts of time prior to use. This study was performed using 0.45 μmol of precursor. Detailed measurements and calculations are tabulated in Table S4^† and the performance is summarized in Fig. 3 (with blue markers). When using 5–25 μL of aqueous [¹⁸F]fluoride, we achieved high RCC and collection efficiency with excellent consistency, resulting in similar crude RCY among these conditions (RCC of 91–95%, collection efficiency of 90–92%, and crude RCY of 81–87%; n = 9). However, when increasing the isotope volume to 40 μL, we observed a significant drop and lower consistency of RCC (32 ± 27%, n = 2), collection efficiency (81 ± 13%, n = 2), and corresponding crude RCY (24 ± 18%, n = 2). Since we can rule out radiolysis, these results suggest that increased amount of impurities from the isotope solution could be responsible for the reduced performance.

Fig. 3.

Droplet synthesis performance of [¹⁸F]FBnTP as a function of [¹⁸F]fluoride volume (μL) loaded. Impact on (A) fluorination conversion, (B) collection efficiency, and (C) crude RCY are shown (each conditions was repeated n = 2 times unless otherwise indicated).

We performed a small study of precursor quantity during experiments involving a higher volume of [¹⁸F]fluoride (i.e., 40 μL) to assess whether performance could be improved using increased amounts of precursor (i.e., 0.45–1.05 μmol). Detailed measurements and calculations are tabulated in Table S5.^† Increasing the precursor quantity from 0.45 to 0.6 μmol did not restore the high performance, but significantly increased the crude RCY (from 24 ± 18% (n = 2) to 44 ± 2% (n = 2)). Subsequent increments did not exhibit significant changes in synthesis performance, and therefore we selected 0.6 μmol as the precursor amount for later scale-up synthesis.

Next, we performed additional tests where the activity level (25–1510 MBq) of the initial [¹⁸F]fluoride was varied over nearly 2 orders of magnitude. For practical reasons, it was not possible to maintain a consistent volume of [¹⁸F]fluoride, which varied from 5–42 μL in these studies. Detailed measurements and calculations can be found in Table S6^† and the performance is summarized in Fig. 3 (with red markers). These data exhibited a similar trend as the prior isotope volume study. Regardless of activity level, volumes up to 20 μL exhibited high performance (crude RCY 80–90%), volumes of 25 and 30 μL exhibited moderate performance (crude RCY 60–80%) and higher volumes gave much lower and variable crude RCY.

Since the product yield drops significantly if using >30 μL of [¹⁸F]fluoride, we used this as a maximum volume of isotope to load in each reaction site, and to perform further scale-up we performed multiple syntheses in parallel (numbering up).

3.4. Synthesis scale-up

Assuming a [¹⁸F]fluoride concentration of ∼30 MBq μL⁻¹, each 30 μL portion of fluoride contains ∼900 MBq. Thus, given an estimated crude RCY (∼60%) and estimated synthesis time of ∼40 min, we estimated that combining two droplet reactions would be sufficient to prepare a batch (>740 MBq) sufficient for two or more clinical doses (estimated to be 92.5–315.5 MBq each, based on doses used for [¹⁸F]flurpiridaz (ref. 29)), or one dose, if significant transport is required prior to use. The results are summarized in Table 1. This study, where two reactions with combined starting activity of 1.6–2.1 GBq, resulted in RCY of 54 ± 6% (n = 3), activity yield of 43 ± 5% (n = 3) and radiochemical purity (RCP) of 100% (Fig. S2–S5^†). A comparison with the results from a 30 μL reaction in a previous optimization study within this work (i.e., crude RCY of 62 ± 2% (n = 2), exhibited a slightly lower RCY the in the scaled-up synthesis. This discrepancy can likely be attributed to minor activity loss during HPLC purification and the product formulation process. With this scale-up strategy, we successfully provided a clinically-relevant dose of [¹⁸F]FBnTP (0.76–0.80 GBq, n = 3) with excellent molar activity of 665–877 GBq μmol⁻¹ at the end of synthesis.

Table 1.

Comparison of [¹⁸F]FBnTP synthesis performance under microscale and macroscale conditions. Where applicable, values are given as averages ± standard deviations for the indicated number of replicates. RCY = Radiochemical yield. RCP = Radiochemical purity. N.R. = Not reported. EOS = end of synthesis. K₂₂₂ = 4,7,13,16,21,24-hexaoxa-1,10-diazabicyclo[8.8.8]hexa-cosane

Conditions	This work (2 droplets)	This work (4 droplets)	Ravert et al. 2004 (ref. 9)	Ravert et al. 2014 (ref. 15)	Tominaga et al. 2016 (ref. 11)	Zhang et al. 2016 (ref. 17)	Waldmann et al. 2018 (ref. 16)	Lin et al. 2023 (ref. 18)	Jones et al. 2023 (ref. 23)
METHOD
Radio- synthesis platform	Droplet-based synthesizer	Droplet-based synthesizer	5 mL v-vial	5 mL v-vial in a microwave cavity	Glass vial	Glass vial	ELIXYS FLEX/CHEM	GE Tracerlab FXN	Droplet-based synthesizer
Manual or automated?	Manual	Manual	Manual	Remote control	Manual	Manual	Automated	Automated	Manual
Synthesis steps	1	1	4	4	3	1	4	1	1
Synthesis route	Cu-mediated fluorination	Cu-mediated fluorination	1) Fluorination 2) Reduction 3) Bromination 4) Alkylation	1) Fluorination 2) Reduction 3) Bromination 4) Alkylation	1) Fluorination 2) Reduction 3) Alkylation	Cu-mediated fluorination	1) Fluorination 2) Reduction 3) Bromination 4) Alkylation	Cu-mediated fluorination	Cu-mediated fluorination
Precursor consumed (μmol)	0.6 × 2	0.6 × 4	20	12.8	42.9	4	14.3	15.9	0.45
Major reagent(s) consumed (μmol)	Cu(Py)4(OTf)2 (0.68 × 2)	Cu(Py)4(OTf)2 (0.68 × 4)	1) N(Me)3Bz·OTf (20), K2CO3 (20), K222 (30) 2) NaBH4 (7930) 3) PBr2Ph3 (200) 4) PPh3 (80)	1) N(Me)3Bz·OTf (12.8), K2CO3 (82.6), K222 (111) 2) NaBH4 (31.7) 3) HBr (48% aq; 0.8 mL) 4) PPh3 (11.4)	1) N(Me)3Bz·OTf (42.9), K2CO3 (N.R.), K222 (37) 2) NaBH4 (529.) 3) PPh3·HBr (367)	Cu(OTf)2 (20)	1) N(Me)3Bz·OTf (14.3), K2CO3 (7), K222 (27) 2) NaBH4·(Al2O3)x (10 wt%, 350 mg) 3) PBr2Ph3 (200) 4) PPh3 (11.4)	Cu(Py)4(OTf)2 (50)	Cu(Py)4(OTf)2 (0.68)
Solvent volume (μL)	10 × 2	10 × 4	200–3000	400–1400	600–6000	850	800–4100	N.R.	10
PERFORMANCE
Number of repeats (n)	3	4	20	27	5	2	3	7	3
Starting activity (GBq)	1.6–2.1	0.9–3.7	N.R.	N.R.	3.7–5.6	N.R.	9.4–12.0	11.1–55.5	0.12
RCY (%)a	54 ± 6	66 ± 1	10d	11.5 ± 3.3d	12–14	N.R.e	28.6 ± 5.1	9.1 ± 1.4d	66 ± 6
RCP (%)b	100	>99	>99	>99	>99	>97	>99	>99	100
Activity yield (GBq)	0.76–0.80	0.38–1.84	N.R.	N.R.	N.R.	N.R.	1.4–2.2	0.7–3.7	0.056–0.063
Activity yield (%)	43 ± 5	49 ± 1	6	8.3 ± 2.4	N.R.	N.R.	14.9–18.3d	6.2 ± 1.0	49 ± 3
Molar activity (GBq μmol−1) at EOS	665–877	339–710c	16.7	530 ± 370	N.R.	N.R.	80–99	6.4 ± 1.0	N.R.
Total synthesis time (min)	37 ± 1	47 ± 1	82	52 ± 14	N.R.	N.R.	90–92	60 ± 5	42 ± 1

^aRCY includes purification and formulation.

^bRCP was determined by radio-HPLC.

^cThe molar activity was calculated from the synthesis starting with 2.7–3.7 GBq of activity.

^dThese valves were calculated based on other information in the literature report.

^eOverall performance was not reported, but radiochemical conversion of product was 62 ± 1.4 (n = 2), determined by radio-HPLC using an aliquot of diluted crude product.

Since all reactions ran in parallel, the preparation time remained similar to performing a single droplet reaction, the only difference being that additional time is required for the collection step (since multiple droplets need to be sequentially collected). The total synthesis time was 37 ± 1 min. Note that this synthesis time is shorter than reported in our prior high-throughput optimization study (42 ± 1 min),²³ due to using a different mobile phase for radio-HPLC purification with a slightly higher proportion of MeCN which shortened the retention time (14.0 vs. 15.3 min), and reducing the amount of dilution of purified product (15 vs. 20 mL) which shortened the formulation process (10 min vs. 13 min).

Noting that the use of 20 μL aliquots of [¹⁸F]fluoride performed better than 30 μL in our isotope volume study, we performed an additional set of experiments, in which we performed pooling of four droplet reactions, each starting with 20 μL of activity (Table 1). For purification, we used a semi-prep column instead of analytical to ensure that the mass and volume of injection material did not exceed the column capacity (Fig. S6^†). Starting with 0.9–3.7 GBq, the resulting performance exhibited slightly higher RCY (66 ± 1%, n = 4) and activity yield (49 ± 1%, n = 4) than the 30 μL study, while maintaining high RCP (>99%) and molar activity (339–710 GBq μmol⁻¹ for the synthesis starting with 2.7–3.7 GBq of activity) (Fig. S7 and S8^†). Following purification and formulation, up to 1.26 GBq of [¹⁸F]FBnTP was produced. Due to the use of the semi-prep column, the HPLC purified fraction had larger volume, and required more dilution for formulation, than the two-droplet experiments, increasing the formulation time (from 10 min to 19 min), and thus extending the overall synthesis time by ∼9 min.

3.5. Comparison of droplet and conventional methods

In comparison to the previously reported macroscale conditions by Zhang et al.,¹⁷ this scaled-up droplet synthesis through the same Cu-mediated route offered significant advantages. The droplet format reduced the reaction volume from 850 μL to 10 μL, allowing for higher reagent concentration while consuming much less reagents (i.e., 2–13× less precursor and 7–37× less copper reagent, depending on whether 2 or 4 droplets were pooled). Moreover, our droplet synthesis demonstrated 6–7× increase in RCY and 7– 8× improvement in activity yield when compared to the recently reported automated synthesis by Lin et al.,¹⁸ along with superior radiochemical purity of 100% (vs. 97% in ref. 17 and 99% in ref. 18). Despite commencing with significantly lower starting activity than Lin et al.¹⁸ (i.e., 0.9–3.7 GBq vs. 11.1–55.5 GBq) the droplet synthesis could still provide a product amount sufficient for multiple patient doses, and exhibited significantly higher molar activity (339– 877 GBq μmol⁻¹ vs. 6.42 ± 0.98 GBq μmol⁻¹). In addition, the radio-HPLC chromatogram of the crude [¹⁸F]FBnTP injection displayed excellent separation resolution on both analytical and semi-prep columns (Fig. S2 and S6^†). Only two major radio-peaks, corresponding to unreacted [¹⁸F]fluoride and [¹⁸F]FBnTP, were observed in the HPLC chromatogram from the droplet reaction (Fig. S2 and S6^†), whereas multiple peaks were seen in the initial macroscale reactions.¹⁷ This suggests that the microscale synthesis had fewer side reactions which may give opportunities for further optimization and shortening of the purification process. Though the numbering up method required more precursor consumption compared to the single-reaction paired with concentration method (Fig. 1A and B), the quantity is still lower than that for macroscale approach (1.2–2.4 μmol for 2–4 droplet reactions vs. 4–15.9 μmol for the macroscale reaction).

In comparison to other macroscale conditions involving multiple reaction steps, the one-step radiosynthesis approach significantly simplifies the preparation of [¹⁸F]FBnTP, shortens the synthesis time and purification, and eliminates the need for handling corrosive reagents, making it more practical for both preclinical and clinical studies. Moreover, the droplet scale-up method dramatically reduced precursor consumption by 2–35× while providing 2–6× higher RCY (compared to reported data by Ravert et al.,^9,15 Tominage et al.,³⁰ and Waldmann et al.¹⁶) and 3–8x higher activity yield (compared to reported values by Ravert et al. 2014 (ref. 15) and Waldmann et al.¹⁶). Even with 3–4× lower starting activity compared to Waldmann et al.’s method (2.7–3.7 GBq in this work vs. 9.4–12.0 GBq in ref. 16), the droplet scale-up technique achieves comparable quantity of [¹⁸F]FBnTP (1.26–1.84 GBq from 4 droplet reactions vs. 1.4–2.2 GBq (ref. 16)), and over 4–8× higher molar activity. This enables efficient production of small tracer batches through Cu-mediated radiofluorination, especially suitable for preclinical imaging scenarios where high molar activity is needed. Additionally, the total preparation time is 5–55 min shorter than all reported macroscale approaches.

Building on the successful flexible scale-up of radiotracer product amount by parallel droplet reactions presented in this study, further investigations will explore the feasibility of combining more droplet reactions to provide multiple patient doses in an automated format, using a previously-described automated system for droplet reactions.³¹

4. Conclusions

In this work, we successfully developed a droplet-based one-step Cu-mediated fluorination synthesis for [¹⁸F]FBnTP using a pinacolyl arylboronate precursor. After performing optimization of droplet reactions at low activity scale using high-throughput techniques, a short study enabled determination of the maximum practical volume of [¹⁸F] fluoride per reaction (i.e. that did adversely impact performance). Subsequently, within this constraint, the synthesis was scaled by performing multiple reactions in parallel to achieve the desired amount of product. The resulting radiochemical yield after purification and formulation was high for both a two-reaction approach (30 μL per reaction; RCY = 54 ± 6%, n = 3) and four-reaction approaches (20 μL per reaction; RCY = 66 ± 1%, n = 4), with excellent radiochemical purity (>99%) and high molar activity (339–877 GBq μmol⁻¹). Sufficient product for multiple clinical doses, 0.76–1.84 GBq, was efficiently achieved from 1.6 to 3.7 GBq of [¹⁸F]fluoride in a synthesis time of just 37– 47 min. The simplicity and speed of this synthesis method, along with improved yield and reduced precursor amount, will greatly facilitate further preclinical and clinical evaluation of [¹⁸F]FBnTP for MPI or other applications, like lung cancer studies.^32,33 Moreover, this efficient droplet-based scale-up technique can readily be applied to prepare other radiotracers on demand, enabling quick and cost-effective production of various radiotracers for diverse applications. This work represents the first successful trial of scaling up the synthesis in a droplet microreactor by the numbering up technique. Automation of this approach is ongoing and provides a promising route to reliably supply multiple patient doses per batch using droplet radiochemistry methods.