Automated radiosynthesis of [¹⁸F]fluoromannitol on the Sofie Biosciences ELIXYS FLEX/CHEM system

Abstract

Bacterial infections remain a significant global health concern, exacerbated by rising antimicrobial resistance and a growing immunocompromised population. Improved diagnostic tools are essential to accurately detect infections, distinguish them from sterile inflammation, and reduce unnecessary antibiotic use that drives resistance. Positron Emission Tomography (PET) has emerged as a powerful non-invasive modality for detecting and monitoring infections in vivo, especially when conventional diagnostics are inconclusive.

[¹⁸F]Fluoromannitol is a novel PET radiopharmaceutical that selectively targets both Gram-positive and Gram-negative bacteria, enabling broad-spectrum imaging and differentiation from sterile inflammation. To support preclinical studies and facilitate broader research use, we report the automated synthesis of [¹⁸F]fluoromannitol using the ELIXYS FLEX/CHEM radiosynthesizer.

The synthesis was adapted from a manual protocol into a fully automated, two-pot, three-step process involving a nucleophilic fluorination, acid deprotection, and sodium borohydride reduction. Reaction conditions were optimized on the ELIXYS FLEX/CHEM platform. The fluorination was optimized for temperature and time to a yield of 78 ± 1.5 %. Acid-catalyzed deprotection yielded an impurity when manual conditions were applied. A large increase in temperature was necessary in order to efficiently produced [¹⁸F]fluoromannose with >99 % radiochemical purity, which was then reduced to [¹⁸F]fluoromannitol. The complete synthesis required approximately 136 min, yielding a 15 ± 0.9 % activity yield, a 35 ± 2.0 % radiochemical yield, >99 % radiochemical purity, and a final pH of 5.5 ± 0.5.

This automated synthesis protocol supports reliable production of [¹⁸F]fluoromannitol and its broader adoption in bacterial infection imaging, with the potential to enhance early diagnosis and improve clinical management.

1. Introduction

Infections are a leading contributor to global morbidity and the third most common cause of death worldwide (Roth et al., 2018). Healthcare-associated infections such as surgical site infections, ventilator associated pneumonia, and Methicillin-resistant Staphylococcus aureus (MRSA) are frequently caused by pathogens like S. aureus, A. baumannii, P. aeruginosa, and Enterobacteriaceae species such as E. coli and Salmonella. The increase in antimicrobial resistance, along with a growing number of immunocompromised individuals due to conditions such as HIV/AIDS, chemotherapy, organ transplants, and diabetes, is placing significant strain on healthcare systems, with annual costs in the U.S. estimated between $10 and $33 billion and in Europe, exceeding €7 billion (Voidazan et al., 2020; Weinstein et al., 2014). By 2050, drug-resistant infections are projected to overtake cancer as the leading cause of death (Ordonez et al., 2019; Programme, U. E., 2023). The Centers for Disease Control and Prevention (CDC) has classified several antimicrobial resistant pathogens such as carbapenem-resistant Acinetobacter and Enterobacteriaceae, ESBL-producing Enterobacteriaceae, multidrug-resistant P. aeruginosa, and methicillin-resistant S. aureus, as urgent or serious threats to public health (CDC, 2019). Carbapenem-resistant A. baumannii (CrAB) infections are of particular concern, especially in individuals with underlying health conditions and compromised immune systems (Benjamin et al., 2013). Recently, CrAB infections alone caused 8500 hospitalizations, 700 deaths, and $281 million in U.S. healthcare costs (CDC, 2019). This highlights the urgent need for better methods to diagnose and treat bacterial infections.

Biopsy sampling, followed by pathogen culturing, is widely regarded as the gold standard for confirming infections and identifying causative organisms of infection. However, this method is invasive and often challenging for diagnosing deep-seated infections (Źródłowski et al., 2020). Positron Emission Tomography (PET) imaging offers a non-invasive approach to visualize bacterial infections and monitor their progression in real-time, enabling accurate diagnosis and ongoing assessment. PET imaging is particularly advantageous for detecting infections that may be difficult to identify using conventional modalities such as Magnetic Resonance imaging (MRI) or Computed Tomography (CT), which rely on morphologic changes that are relatively nonspecific to infection. To enhance diagnostic accuracy, newly developed radiopharmaceuticals have been designed to target specific bacterial signatures, including metabolic activity (Gowrishankar et al., 2017; Ning et al., 2014; Weinstein et al., 2014), cofactor biosynthesis (Schulte et al., 2024; Zhang et al., 2018, Neumann et al., 2017), and labeled antibiotics (Sellmyer et al., 2017; van Oosten et al., 2013).

Recently, [¹⁸F]fluoromannitol was reported as a novel radiopharmaceutical to image diverse pathogenic bacteria in vivo. [¹⁸F]Fluoromannitol is able to image a broad range of pathogenic organisms including S. aureus (gram positive) and E. coli (gram negative), while clearly delineating from sterile inflammation (Simpson et al., 2022). To enable reliable, routine production and facilitate implementation in infectious disease research, we sought to develop an automated synthesis method. To this end, the synthesis of [¹⁸F]fluoromannitol was automated using the Sofie Bioscience ELIXYS FLEX/CHEM system. This advancement facilitates widespread access to [¹⁸F]fluoromannitol in research settings within the scientific community.

2. Experimental

General: All chemicals were purchased from Sigma-Aldrich without further purification, unless otherwise noted. All aqueous solutions were prepared using ultra-pure water from a Milli-Q Integral Water Purification System (Millipore Corp.; 18.2 MΩ cm resistivity). The radiochemical precursor, 4,6-O-Benzylidene-3-O-ethoxymethyl-2-O-trifluoromethanesulfonyl-1-O-methyl-β-D-glucopyranoside, was synthesized in-house using a previously reported method (Simpson et al., 2022) or purchased from ABX (Product Code: MBETG, 1118) and used as supplied. Radiochemical optimizations and product analyses were performed using Radioactive Thin Layer Chromatography (RadTLC), [AR-2000 TLC Scanner, Eckert & Ziegler], or High-Performance Liquid Chromatography (HPLC), [Agilent 1200 series].

Radiochemistry: All radiochemical experiments were performed using the ELIXYS FLEX/CHEM automated synthesizer (Sofie Biosciences). ~100 mCi of aqueous [¹⁸F]fluoride was produced via the ¹⁸O (p,n)¹⁸F nuclear reaction in an IBA Cyclone^® 18/9 cyclotron, utilizing ¹⁸O-enriched water. Briefly, [¹⁸F]fluoride was captured on a commercially available anion-exchange resin cartridge (Myja Scientific, chloride anion, 10–15 mg) pre-conditioned with 15 mL of ultra-pure water. Fluoride was subsequently eluted into reactor 1 using a solution consisting of 0.9 mL of anhydrous acetonitrile (MeCN) and 0.1 mL of ultra-pure water containing 6 mg of Kryptofix (K₂₂₂) and 1 mg of K₂CO₃.The solvent was evaporated at 110 °C under negative pressure. Anhydrous acetonitrile (1 mL × 2) was added to the residue to azeotropically remove any remaining water at 110 °C. Next, 3 mg of 4,6-O-Benzylidene-3-O-ethoxymethyl-2-O-trifluoromethanesulfonyl-1-O-methyl-β-D-glucopyranoside in 1 mL of anhydrous acetonitrile was added to the dried K₂₂₂/[¹⁸F]FK complex and heated at 140 °C for 10 min. The solution was then allowed to cool to 100 °C. The solvent was subsequently removed under negative pressure at 110 °C. After solvent removal, 1 mL of 2.5 M hydrochloric acid (HCl) was added to the reactor vial, and the mixture was heated at 180 °C for 10 min. After cooling to 40 °C, the HCl solution was diluted with 3 mL of ultra-pure water and passed through a cartridge purification system to waste. This system consisted of an Accell Plus CM short cartridge (Waters), followed by 3 g of BT AG 11 A8 resin (Bio-Rad; custom-packed in a Phenomenex Strata C18-U (55 μm, 70 Å) cartridge), and a Sep-Pak Alumina N Plus long cartridge (Waters) all connected in series. The system was pre-conditioned with 18 mL of water prior to use. Two 2-mL aliquots of ultra-pure water were passed through the cartridge system into reactor 2 containing 5 mg of NaBH₄. The aqueous solution was then heated to 60 °C for 30 min. After cooling to 40 °C, the aqueous solution was passed through a Chromabond Set V (ABX) and an Accell Plus QMA Light sep-pak (Waters) to waste. The Chromabond SET V was previously conditioned with 10 mL of water and the Accell Plus QMA Light was previously conditioned with 3 mL 1 M sodium bicarbonate followed by 5 mL of water. Two 3-mL aliquots of ultra-pure water were then passed through the Chromabond Set V and the Accell Plus QMA light into a collection vial containing 54 mg of NaCl to isolate the [¹⁸F]fluoromannitol. The final product was sterilized using a 0.22 μm filter. No HPLC purification was required.

HPLC for Identification and Radiochemical Purity: The HPLC system consisted of the following Agilent 1200 Series modules: G1379B Degasser, G1312B Binary Pump SL, G1316B Thermostat Column Compartment (TCC) SL, G1367C High-Pressure Auto Sampler (HiP-ALS) SL, and G1315C Diode Array Detector (DAD) SL. The system was coupled in series with a Bioscan Flow-Count 106 radio flow detector for radioactive measurements. Control and data processing were managed using ChemStation software (Agilent). HPLC analysis was performed on a Rezex RPM-Monosaccharide PB²⁺ column (300 × 7.8 mm. 80 °C) at a flow rate of 1 mL/min using 100 % ultra-pure water as the eluent. The UV chromatogram was recorded at 254 nm along with signal from the activity detector.

Crystallography: A single crystal of 2A was coated with Paratone oil and mounted on a MiTeGen Micromount. The X-ray intensity data was measured on a Bruker D8 Venture PhotonIII Kappa four-circle diffractometer system using either an Incoatec IμS 3.0 micro-focus sealed X-ray tube (Cu Kα, λ = 1.54178 Å) and a HELIOS double bounce multilayer mirror monochromator. Reflections were merged by the Bruker SHELXTL Software Package according to the crystal class for the calculation of statistics and refinement. All frames were integrated with the Bruker SAINT software V8.41 using a narrow-frame algorithm. Data were corrected for absorption effects using the Multi-Scan method (SADABS). The structure was solved and refined using the Bruker SHELXTL Software Package within APEX3 and OLEX2. All esds (except the esd in the dihedral angle between 2 l s planes) are estimated using the full covariance matrix. The cell esds are considered individually in the estimation of esds in distances, angles and torsion angles; correlations between esds in cell parameters are only used when they are defined by crystal symmetry. An approximate (isotropic) treatment of cell esds is used for estimating esds involving l.s. planes. Non-hydrogen atoms were refined anisotropically. Hydrogen atoms were placed in geometrically calculated positions with Uiso = 1.2 (C(H) and C(H,H)) or Uiso = 1.5 (H,H,H)).

3. Results and discussion

The synthesis of compound 3 was performed in-house under previously reported conditions outlined in Scheme 1A (Simpson et al., 2022). Initially, compound 1 was dissolved in THF with sodium hydride to deprotonate the alcohol at the C3 position. The chloromethyl ethyl ether protecting group was then introduced, yielding compound 2. Depending on the site of protection, two regioisomers of compound 2 can form (Scheme 1B), which were observed in a 60:40 proportion (2A:2B). Purification was carried out using a CombiFlash^® Next Gen 300 with deactivated silica, where the desired isomer (2A) was confirmed via X-ray crystallography (Scheme 1C). Finally, a triflate group was then introduced to form compound 3, as previously reported (Simpson et al., 2022).

Scheme 1. Synthesis of the radiochemical precursor of [¹⁸F]fluoromannitol.

A) Synthetic scheme for the production of compound 3. B) Regioisomers of compound 2. Isolation of 2A is needed to successfully synthesize compound 3. C) Crystal structure (ellipsoid plot) of compound 2A grown to confirm the isolation of desired intermediate.

With compound 3 in hand, we turned to the main objective of this work: adapting the previously reported manual synthesis of [¹⁸F]fluoromannitol to an automated radiosynthesizer, facilitating easier availability of the imaging agent for preclinical studies. We selected the ELIXYS FLEX/CHEM primarily for its flexibility in executing multi-step syntheses using multiple reaction vials within a single sequence. The ELIXYS FLEX/CHEM features three identical and independent reactors capable of heating, cooling, and stirring. Each reactor can interface with a disposable reagent cassette that accommodates up to 12 reagents (36 total with three cassettes), each in 3 mL septum-sealed vials. The cassette includes all reagent pathways, tubing, and ports for external addition and [¹⁸F]fluoride delivery. It enables multiple operations, including reagent addition (2 positions), evaporation (1), sealed high-pressure reactions (2, up to 150 psi), and product transfer (1). Reactor activity is monitored via a live camera feed and system-wide radiation sensors.

The radiochemical synthesis of [¹⁸F]fluoromannitol is a straightforward two-pot, three-step reaction that includes an S_N2 reaction with [¹⁸F]fluoride, followed by acid deprotection, and a final sodium borohydride reduction, as shown in Scheme 2. The synthetic scheme was implemented on the ELIXYS FLEX/CHEM utilizing two cassettes. The configuration of materials and consumables on the FLEX/CHEM is detailed in Fig. 1.

Scheme 2.

Radiochemical synthesis of [¹⁸F]fluoromannitol on the ELIXYS FLEX/CHEM.

Fig. 1.

Materials and consumables required for the radiosynthesis of [¹⁸F]fluoromannitol on the ELIXYS FLEX/CHEM.

Initial optimization efforts focused on the fluorination of compound 3 to form compound 4. The manual conditions specified heating the fluorination reaction to 120 °C for 20 min. To determine the optimal temperature on the ELIXYS FLEX/CHEM system, a range of temperatures were tested (Fig. 2), as prior work with the ELIXYS FLEX/CHEM system suggested that slightly higher temperatures might be necessary (McCauley et al., 2022). While manual fluorination of compound 3 achieved yields approaching 80 % at 120 °C, the automated synthesizer at the same temperature only yielded 61 ± 2 %. The highest fluorination yields were obtained by heating the reaction to 140 °C (73 ± 2 %). Upon identifying the optimal temperature, we next optimized reaction time to produce compound 4. While manual synthesis required 20 min of heating, automated fluorination using the ELIXYS FLEX/CHEM reached optimal yields in just 10 min (78 ± 1.5 %), as summarized in Fig. 3. Increasing reaction time from 5 to 10 min resulted in significant increases in production of compound 4, however, further heating beyond 10 min did not deliver meaningful changes in yields. Radiochemical reactions to produce compound 4 resulted in satisfactory yields, thus, changes to the precursor weight and reaction volume were not investigated.

Fig. 2. Influence of temperature on fluorination reaction to produce compound 4.

All reactions were conducted for 20 min using the ELIXYS FLEX/CHEM module. Percent fluorination was determined via radio-TLC. All fluorination yields are presented as means (n = 3, ± S.D.).

Fig. 3. Influence of time on fluorination reaction to produce compound 4.

All reactions were carried out at 140 °C using the ELIXYS FLEX/CHEM module. Percent fluorination was determined via radio-TLC. All fluorination yields are presented as means (n = 3, ± S.D.).

Once the fluorination conditions were optimized, deprotection of compound 4 was performed using 1 mL of 2.5 M HCl. Previously reported manual deprotection conditions were performed at 140 °C for 10 min. On the FLEX/CHEM system, we observed the formation of two radiochemical products, not previously observed under manual conditions. By HPLC, we identified, one product as the desired intermediate [¹⁸F]fluoromannose (5). However, the other product was an unidentified impurity. Attempts to carry the impure reaction mixture through to the next step, sodium borohydride-mediated reduction, resulted in no substantial changes to the impurity identity by HPLC, while [¹⁸F]fluoromannose (5) was completely converted to [¹⁸F]fluoromannitol (6). As this impurity did not change when exposed to sodium borohydride, but disappeared with prolonged heating in acid, we suspected the unidentified impurity to be the partially deprotected methyl acetal of compound 4 (Scheme 2). Extending the deprotection reaction time at 140 °C did not affect the ratio of [¹⁸F]fluoromannose to impurity. Therefore, the focus shifted to adjusting the reaction temperature for deprotection. The results, summarized in Fig. 4, showed that heating the deprotection reaction for 10 min at 180 °C was required to exclusively form the desired [¹⁸F]fluoromannose (5).

Fig. 4. Influence of temperature on the ratio of product formation during acid-catalyzed deprotection.

All reactions were conducted for 10 min on the ELIXYS FLEX/CHEM module. Percent conversions of deprotected products were determined via HPLC on a Rezex RPM-Monosaccharide PB²⁺ column (300 × 7.8 mm. 80 °C) at a flow rate of 1 mL/min using 100 % ultra-pure water as the eluent (n = 3, ± S.D).

After successfully producing [¹⁸F]fluoromannose, we aimed to optimize its purification using a cartridge-based system, absent of HPLC. The manual purification process included an Accell Plus CM Plus Short cartridge, 3 g of BT AG 11 A8 resin (custom packed in a Phenomenex Strata C18-U cartridge, 55 μm, 70 Å), a Sep-Pak Alumina N Plus Long cartridge, and an Uniflo 25 mm 0.45 μm PVDF filter (Simpson et al., 2022). The Accell Plus CM Plus Short cartridge is used to capture residual Kryptofix, the BT AG 11 A8 resin is used to remove any excess acid, and the Sep-Pak Alumina N Plus Long cartridge was used to remove [¹⁸F]fluoride.

When applying the manual method to the ELIXYS FLEX/CHEM, the Uniflo filter introduced significant backpressure, pushing the ELIXYS FLEX/CHEM system to its transfer limits and prolonging purification. Removing the Uniflo filter resolved the backpressure issue, enabling low-pressure transfer through the cartridge stack. This adjustment had no adverse effect on the chemistry and reduced overall production time.

The volume of water required to elute [¹⁸F]fluoromannose was evaluated due to the ELIXYS FLEX/CHEM’s 5 mL reactor volume limitation. The manual method used a total of 9 mL of water—3 mL for washing and 6 mL for elution (Simpson et al., 2022)—which exceeded the reactor’s capacity. To reduce the elution volume, we evaluated the activity eluted from the cartridge system in 1 mL increments over 9 mL on the ELIXYS FLEX/CHEM. Minimal activity was observed in the first 3 mL, with activity beginning to elute in the fourth. The highest activity was collected in the fifth and sixth milliliters, while the seventh contained the remaining product. The eighth and ninth milliliters showed only background activity. These results guided the revised elution protocol described in the Methods.

Additionally, the ELIXYS FLEX/CHEM accommodates only 3 mL reagent vials, limiting the volume delivered per addition. To enable a 4 mL elution, we tested two approaches: (1) combining two 2 mL aliquots in the reactor before passing them through the cartridge stack, and (2) passing two separate 2 mL aliquots directly through the cartridge stack. The latter approach yielded higher activity recovery and was selected for the final protocol.

After optimizing the two reaction steps in the production of [¹⁸F] fluoromannose (5), full-scale productions of [¹⁸F]fluoromannose (5) were carried out. The automated synthesis of the [¹⁸F]fluoromannose (5) intermediate afforded an activity yield of 35.8 ± 3.5 % and a radiochemical yield of 56.0 ± 2.4 %, with a radiochemical purity of >99 % (n = 7); thus, we proceeded with the reduction reaction. [¹⁸F] Fluoromannose was transferred into a vial containing NaBH₄ in reactor 2, then heated for 30 min at 60 °C. Complete reduction to desired [¹⁸F] fluoromannitol was achieved by adapting the reported manual conditions to the ELIXYS FLEX/CHEM sequence. The product was subsequently passed through a Chromabond SET V (ABX) cartridge. This cartridge was chosen for the manual synthesis as it was known to help stabilize pH along with removing residual NaBH₄ (Mota et al., 2021). Although the Chromabond Set V purification delivered a final product that was >99 % radiochemically pure, the pH of the final solution was too low for biological studies (3 ± 0.5, n = 3), which was not observed in the manual synthesis (Simpson et al., 2022). To ensure optimal pH, we added a QMA cartridge preconditioned with sodium bicarbonate and ultra-pure water at the end of the Chromabond SET V, which resulted in an optimal pH (5.5 ± 0.5, Table 1) of our final product solution.

Table 1. Metrics for the production of [¹⁸F]fluoromannitol on the ELIXYS FLEX/CHEM.

Radiochemical purity was >99 %. Kryptofix was below the FDA limit. pH was in a suitable range as sterile water (5–7 pH).

Descriptive Data	Run 1	Run 2	Run 3	Run 4	Run 5	Run 6
End of Bombardment (mCi)	103	100	104	105	100	101
End of Synthesis (mCi)	16.5	13.6	16	16.7	13.7	15.6
Activity Yield (mCi)	16	13.6	15.4	15.9	13.7	15.4
Run Time (min)	135	137	132	134	136	131
Radiochemical Purity (%)	>99	>99	>99	>99	>99	>99
Kryptofix (ug/mL)	<50	<50	<50	<50	<50	<50
pH	5	5.5	6	5	6	5.5

[¹⁸F]Fluoromannitol was successfully synthesized with an activity yield of 15 ± 1 % (15.3 ± 1.2 mCi) and a radiochemical yield of 35 ± 2 % in 134 ± 2 min (n = 6, Table 1, Table 2). Radio HPLC indicated the radiochemical purity was >99 % (Table 1, Fig. 5). Residual Kryptofix was tested using a previously published method (Mock et al., 1997). Residual K222 was <50 μg/mL, which is within acceptable limits set for 2-[¹⁸F]FDG by the United States FDA (Table 1). No chemical impurities were observed at the 254 nm wavelength (Fig. 5) by HPLC. Gas chromatography indicated no residual solvent above acceptable limits (Supplemental Fig. 1). Due to the lack of a chromophore on [^18/19F] Fluoromannitol, confirmation of tracer identify via coinjection with a non-radioactive standard on HPLC using a UV detector is not feasible. To address this limitation, the final product solution of [¹⁸F]fluoromannitol was subjected to high resolution liquid chromatography mass spectrometry analysis. The resulting data confirmed the identity of the product (Supplementary Fig. 2).

Table 2. Accounting of activity during the production of [¹⁸F]fluoromannitol on the ELIXYS FLEX/CHEM.

Data listed as percentage of activity decay corrected to End of Bombardment.

Descriptive Data	Run 1	Run 2	Run 3	Run 4	Run 5	Run 6
Final Product Vial (%)	37.5	32.3	35.4	37.1	32.4	35.2
Myja Exchange Cartridge (%)	<1	<1	<1	<1	<1	<1
Reactor Vial 1	1	2	2	1	2	1
Acell Plus CM short (%)	3	4	5	4	4	5
BTAG 11 A8 Resin (%)	7	6	8	7	9	6
Alumina N Plus Long (%)	28	43	30	34	33	32
Reactor vial 2	2	1	2	2	1	1
Chromabond SET V (%)	8	6	8	7	8	9
Waste (%)	6	2	4	4	2	5
Total (%)	93.5	97.3	95.4	97.1	92.4	95.2

Fig. 5.

HPLC chromatograms of [¹⁸F]fluoromannitol. [¹⁸F]fluoromannitol was eluted through a Rezex RPM-Monosaccharide PB²⁺ column (300 × 7.8 mm. 80 °C) at a flow rate of 1 mL/min using 100 % ultra-pure water as the eluent. Top: ¹⁸F Radio Chromatogram. Bottom: 254 nm detection.

Activity readings of all consumables used in the trapping, purification, or collection of activity during the six test runs of [¹⁸F]fluoromannitol were taken to determine where loss of yield took place (Table 2). Minimal losses of radioactivity are observed at every point in the sequence, however, the majority of loss occurred while during the Alumina N Plus Long cartridge purification. As this cartridge was used to retain unreactive [¹⁸F]fluoride, this indicating the fluorination step to produce 4 is our yield-limiting step. Future optimizations will be focused on increasing the fluorination yield to be as quantitative as possible.

4. Conclusion

In summary, we have described the development and implementation of an automated radiochemical synthesis of [¹⁸F]fluoromannitol on the ELIXYS FLEX/CHEM system. The process reliably yields sufficient quantities to support an afternoon of preclinical studies; however, under the current scale, the yield is adequate for only a single human dose. The availability of this method will facilitate access for the scientific community to this tracer for further research studies.

Supplementary Material

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.apradiso.2025.112076.

Data availability

A check cif file has been provided. Crystal data was submitted to CCDC. CCDC number and instructions on how to obtain data provided in manuscript. All other data available upon request.