Automated radiofluorination of HER2 single domain antibody: the road towards the clinical translation of [¹⁸F]FB-HER2 sdAb

Abstract

Background

With the next generation of Human Epidermal Growth Factor Receptor 2 (HER2) -targeting therapies, such as antibody–drug conjugates, showing benefit in “HER2 low” and even “HER2 ultralow” patients, the need for novel methods to quantify HER2 expression accurately becomes even more important for clinical decision making. A HER2 PET/CT imaging assessment could evaluate HER2 positive disease locations while improving patient care, reducing the need for invasive biopsies. A single-domain antibody (sdAb)-based PET tracer could combine the high specificity of sdAbs with short-lived radionuclides such as fluorine-18 (¹⁸F) and gallium-68 (⁶⁸Ga). SdAb-based PET tracers have clinically been used via a ⁶⁸Ga-chelator approach. However, the distribution of ⁶⁸Ga-labelled pharmaceuticals to peripheral PET centres is more challenging to organize due to the short half-life of ⁶⁸Ga, most certainly when the available activity is limited by a generator. Cyclotron produced ⁶⁸Ga has removed this limitation. Distribution of ¹⁸F-labelled pharmaceuticals remains less challenging due to its slightly longer half-life, and radiofluorination of sdAbs via N-succinimidyl-4-[¹⁸F]fluorobenzoate ([¹⁸F]SFB) has shown to be a promising strategy for developing sdAb-based PET tracers. Although [¹⁸F]SFB automation has been reported, automating protein conjugation proves challenging. Herein we report the fully automated, cartridge-based production of [¹⁸F]FB-HER2 sdAb on a single synthesis module.

Results

[¹⁸F]FB-HER2 sdAb (> 6 GBq) was obtained after a fully automated production (95 min), with a RCP > 95%, apparent molar activity > 20 GBq/µmol and decay-corrected radiochemical yield (RCY d.c.) of 14 ± 2% (n = 4). Further upscaling amounted to production batches of 16 GBq with an apparent molar activity > 40 GBq/µmol and RCY d.c. of 8 ± 1% (n = 4). Ex vivo biodistribution and PET imaging showed specific HER2-positive tumour targeting and low kidney retention.

Conclusion

The [¹⁸F]FB-HER2 sdAb tracer was produced with clinically relevant activities using a fully automated production method. The automated production method was designed to ease the translation to the clinic and has the potential to be used not only in mono-centre but also multi-centre clinical trials with one central production site. [¹⁸F]FB-HER2 sdAb showed a favourable biodistribution pattern and could be a valuable alternative to ⁶⁸Ga-labelled sdAb-based PET tracers in the clinic.

Introduction

Breast cancer is still one of the most common cancers in women worldwide. In 15 – 20% of breast cancers, overexpression of the Human Epidermal Growth Factor Receptor 2 (HER2) can be observed, making it a potent tumour marker for breast cancer. HER2 expression is predictive for response to HER2-specific therapies, such as monoclonal antibodies (mAbs), tyrosine kinase inhibitors, as well as the newer generation HER2-targeting therapies, such as antibody–drug conjugates (Swain et al. 2022; Venetis et al. 2022). These next generation therapies are showing benefits in “HER2 low” and even “HER2 ultralow” patients, making the need for novel methods to quantify HER2 expression accurately even more important in clinical decision-making and patient management (Venetis et al. 2022). HER2 positivity is currently assessed using a combination of fluorescence in situ hybridization (FISH)/ISH and immunohistochemistry (IHC) (Swain et al. 2022), both of which require biopsy or resection specimens—an invasive approach necessary for obtaining tissue samples. In contrast, positron emission tomography (PET) is a non-invasive, highly sensitive and specific molecular imaging technology that could assess HER2 positivity at all disease locations throughout the patient’s body. This non-invasive alternative not only improves patient care by reducing the need for invasive biopsies (Xavier et al. 2016) but also overcomes limitations related to spatial and temporal heterogeneity of HER2 expression, which can be more comprehensively addressed through HER2 imaging.

In accordance with this rationale, a single-domain antibody (sdAb) targeting HER2 (Keyaerts et al. 2016; Keyaerts et al. 2019) has already been successfully translated to the clinic as a diagnostic tracer. SdAbs, also called nanobodies or VHHs, have attracted considerable interest as targeting moieties for PET imaging because of their key characteristics, such as small size (approximately 15 kDa), high affinity, high specificity, low off-target accumulation, high (thermo)stability and solubility. Compared to mAb-based diagnostics, their most notable advantages are that their short biological half-life and fast tumour penetration allow for labelling with short-lived radionuclides such as gallium-68 (⁶⁸Ga) and fluorine-18 (¹⁸F) (Pauw et al. 2023). SdAb-based PET tracers using a ⁶⁸Ga-chelator approach, targeting HER2 (Keyaerts et al. 2019, 2016) and CD206 (Gondry et al. 2023) have successfully been translated. The advantages of the generic ⁶⁸Ga-chelator approach, when compared to a generic radiofluorination approach, is the straightforward chemistry, higher RCYs, and lower initial financial investment, as no cyclotron nor automation modules are necessary. These same advantages also ease the translation of a research and development (R&D) to Good Manufacturing Practices (GMP) compliant radiolabelling process. However, there are still some significant advantages of a generic radiofluorination strategy, such as the superior imaging quality and the longer half-life of ¹⁸F, which also simplifies the radiopharmaceutical distribution. The ease of producing a large amount of radionuclide with a cyclotron can be exploited fully when using an automated radiofluorination approach. Centralised production of multi-patient radiopharmaceutical batches will significantly reduce the cost per patient dose and allow multi-centre clinical trials with one central production site. A characteristic of automated production methods, besides the reduced radiation exposure to the technician, is their robustness and reproducibility, which facilitates the translation to a GMP compliant process. An additional advantage comes from using a commercially available automation module, such as the Trasis AllinOne (AiO) synthesizer, which facilitates the technical transfer to a GMP accredited radiopharmacy as the production setup and dispensing of the radiopharmaceutical can be aligned with already established production processes.

In this study, we utilised the prosthetic group (PG) N-succinimidyl-4-[¹⁸F]fluorobenzoate ([¹⁸F]SFB). Automation of this PG group has already been successfully achieved on various commercial and in-house developed automation modules (Xavier et al. 2016; Fujimoto et al. 2021; Al-Qahtani 2024; Ackermann et al. 2011; Blykers et al. 2015; Scott and Shao 2010; Tang et al. 2010), which has been previously discussed in detail (Dierick et al. 2024). However, to our knowledge, the full automation of [¹⁸F]FB-sdAb has not been reported on a Trasis AiO module nor without the use of high-performance liquid chromatography (HPLC) for purification purposes (Nagachinta et al. 2022). The advantage of this automation module is its ability to accommodate both the PG production and the sdAb-conjugation reaction. This prevents the need for multiple modules and hotcells for a single production, as described by Xavier et al. (Xavier et al. 2016) and Veen et al. (Veen et al. 2019). While the Neptis (Ora) module (Nagachinta et al. 2022) also can accommodate the production of [¹⁸F]FB-sdAb, it requires a time-consuming cleaning step.

The current study aims to develop an automated production method for the [¹⁸F]FB-HER2 sdAb tracer and upscale it to clinically relevant final activities. Together with its preclinical validation, these are the first steps towards the clinical translation of a radiofluorinated sdAb based PET tracer.

Materials and methods

The sdAb proteins

The HER2 sdAb was kindly provided by Abscint NV. The non-targeting control sdAb (R3B23) was produced and characterised as previously described (Lemaire et al. 2014).

Chromatographic analysis

Radio-size-exclusion (SE)-HPLC analysis was performed using a Superdex 30 Increase 3.2/300 column (Cytiva) and 0.01 M phosphate buffer with 0.14 M NaCl (PBS), pH 7.4, at a flow rate of 0.15 mL/min. The radio-SE-HPLC analyses were performed on a Shimadzu Prominence LC-40AT system connected to a UV–VIS detector and γ-detector (Elysia- Raytest).

Automated production of [¹⁸F]FB-sdAbs

Both [¹⁸F]FB-sdAbs tracers, [¹⁸F]FB-HER2 sdAb and [¹⁸F]FB-R3B23 sdAb, were produced on an AllinOne® (AiO) synthesizer (Trasis) with 36 rotors, 5 syringe actuators and 1 heater, using disposable single-use cassettes. [¹⁸F]F⁻ was produced by irradiation of enriched [¹⁸O]water (Rotem medical or Campro) with an IBA Cyclone® KIUBE cyclotron via a ¹⁸O(p.n)¹⁸F nuclear reaction. N-succinimidyl-4-[¹⁸F]fluorobenzoate ([¹⁸F]SFB) was synthesized using a three-step, one-pot reaction (Fig. 1a), purified via HLB cartridge and eluted with ethanol, as we previously described (Dierick et al. 2024). The implementation of the subsequent step in the synthesis process, i.e. conjugation of [¹⁸F]SFB to the sdAb on the AiO, necessitates customisation of the cassette, as described in the results section.

Fig. 1.

Automated production of [¹⁸F]FB-sdAb:a Synthesis of [¹⁸F]SFB in a three-step, one-pot reaction; b Conjugation of [¹⁸F]SFB to sdAb. TPAOH = tetrapropylammonium hydroxide, DMSO = dimethyl sulfoxide, HSTU = N,N,N′,N′-tetramethyl-O-(N-succinimidyl)uronium hexafluorophosphate, CHES = N-cyclohexyl-2-aminoethanesulfonic acid, RT = room temperature

For the conjugation of [¹⁸F]SFB to sdAb (Fig. 1b.), the ethanolic [¹⁸F]SFB was incubated with the protein (5.5 mg; HER2 sdAb: 0.44 µmol; R3B23 sdAb: 0.40 µmol) for 30 min at room temperature in 0.1 M N-cyclohexyl-2-aminoethanesulfonic acid (CHES) buffer pH = 8.6–9. The radiolabelled sdAb was purified using three disposable desalting HiTrap cartridges in series and collected in a final formulation buffer containing 5 mg/mL Na-ascorbate in NaCl 0.9%. The results section describes optimisations of this step to allow for successful automation of the conjugation reaction compared to the manual conjugation method previously described (Dierick et al. 2024). Activity was measured by the activity detectors available within the AiO, unless otherwise specified.

Shelf-life determination

The shelf life of the final product [¹⁸F]FB-HER2 (final activity: > 16 GBq; 804 MBq/mL) was determined by evaluating the stability of the tracer at room temperature for 12 h after the end of production. At different time points, samples were analyzed using radio-SE-HPLC.

Residual solvent analysis

The determination of residual solvent content, specifically dimethyl sulfoxide (DMSO), acetonitrile and ethanol content, was done according to the European Pharmacopeia 2.4.24. (4.24. IDENTIFICATION AND CONTROL OF RESIDUAL SOLVENTS - European Pharmacopoeia 11.5 n.d. 2024). Gas chromatography analyses were acquired on a Shimadzu GC-2010 Plus series with an autoinjector AOC-20i and Flame Ionisation Detector FID-2010 plus (Elysia- Raytest).

Cell line and culture conditions

The human ovarian cancer cell line SK-OV-3, bearing HER2 cell surface proteins, was obtained from the American Type Culture Collection (ATCC) and cultured as previously described (Xavier et al. 2016).

Animal model

Swiss nude Crl: Nu(Ico) Foxn1nu mice (6 weeks old) were purchased from Charles River to evaluate the biodistribution and tumour uptake of [¹⁸F]FB-HER2 compared to its negative control, [¹⁸F]FB-R3B23. They were subcutaneously inoculated with SK-OV-3 cells (10 × 10⁶) in sterile PBS in the right hind flank. The tumours were allowed to grow for up to 2–3 weeks, and their size was measured using an electronic calliper and calculated following the formula (Length x Width²)/2. The animals were anesthetized with 2.5% isoflurane in oxygen (Abbott) for injections. The mice were anesthetized, for imaging and euthanasia, with a mixture of 75 mg/kg ketamine hydrochloride and 1 mg/kg medetomidine hydrochloride 50 min after tracer administration. All of the experiments were performed in accordance with the European guidelines for animal experimentation under the license LA1230272. The ethical committee for animal experiments at the Vrije Universiteit Brussel approved the animal study protocols (22–272-3).

Biodistribution and PET/CT imaging

Tumour bearing mice (n = 4 per group) were i.v. injected with [¹⁸F]FB-HER2 (10 µg; 15 MBq) or [¹⁸F]FB-R3B23 (25 µg; 15 MBq). Micro-PET/CT images were acquired one hour post injection, followed by dissections 80 min post injection. Organs and tissues were collected and counted against a standard of known activity with a gamma-imaging counter (Wizard 2® 2480 Automatic Gamma Counter inspector, PerkinElmer) and expressed as a percentage of injected activity per gram (%IA/g), corrected for decay.

The acquisition of the PET and CT images were respectively carried out with a β-CUBE PET system and a X-CUBE CT system (both from Molecubes), 1 h p.i. The total PET/CT scanning time was 6 min. The PET images were acquired over 5 min and reconstructed into a matrix of 193 × 192 × 384 voxels with 400 μm voxel size. The CT images were iteratively reconstructed using the ISRA reconstruction algorithms into 200 μm voxels (matrix 200 × 200 × 393).

Statistical analysis

Data were expressed as average ± SD. The statistical analysis was performed in GraphPad Prism 10. One-way ANOVA, two-way ANOVA with multiple comparisons tests, or unpaired t-test were used to evaluate statistical significance. Statistical significance was set at p < 0.05 (ns, not significant, * p < 0.05; ** p < 0.01; *** p < 0.001; **** p < 0.0001).

Results

Radiolabelling

Automated production of [¹⁸F]FB-sdAbs

A schematic representation of the automated radiosynthesis procedure is shown in Fig. 2. To reduce the overall production time of the automated synthesis, the module performs some preparative steps before the [¹⁸F]F⁻ is received by the AiO (starting time of synthesis). These “preliminary steps” include transferring the mixture of sdAb and coupling buffer (vial Position (P)31) to the second reactor, and conditioning of the three Hitraps in series (P34) for which 60 mL of saline solution (saline bag P22) is used. The saline solution is applied to the Hitraps in three fractions of 20 mL at a flowrate of 5 mL/min, using the syringe on P9.

Fig. 2.

Layout of the automated radiosynthesis of [¹⁸F]FB-sdAb on a Trasis AiO module. The prosthetic group production (upper row, rotors 1- 18), the purification of the prosthetic group (rotors 25–29 and vial position 16 and syringe position 15), as well as the conjugation reaction and its purification (rotor 20–24; 30–36) are included on the module

[¹⁸F]SFB was synthesized using a three-step, one-pot reaction, and this process was fully automated, as previously described (Dierick et al. 2024). However, for the conjugation of the PG to the sdAb in the disposable cassette to produce [¹⁸F]FB-HER2 sdAb, three changes had to be implemented in the PG synthesis. The first change concerns the acetic acid and saline solution mixture used to dilute the crude reaction mixture (RM). Previously, the mixing of the two solvents was part of a pre-programmed sequence that controls the preparation of the mixture on the module, while we now opted to use a ready-made mixture of acetic acid and saline (provided by vial P13). The 20 mL syringe (P9) is still used to dilute the RM and to apply the diluted RM to the HLB cartridge. The second change concerns the concentration of the acetic acid/saline mixture, which changed from 1.6 V/V% previously used (Dierick et al. 2024) to 1.3 V/V% to take into account the differences in death volumes with the new mixing approach. This modification was necessary to ensure the same level of trapping of [¹⁸F]SFB onto the HLB compared to the previously described process (Dierick et al. 2024).

Thirdly, the purification step of the PG on the HLB was moved to P27 (Fig. 2). This change only affected the fluidics of the purification. The washing step and reverse elution were kept as previously reported. The ethanolic [¹⁸F]SFB was then directly eluted in the second reactor containing the sdAb/coupling buffer mixture. The total time of this PG production was 54 min and allowed to obtain [¹⁸F]SFB (33.78 ± 6.75 GBq, n = 10) with an estimated radiochemical yield (RCY) decay corrected (d.c.) of 45 ± 7% (n = 10).

The automated conjugation reaction starts at this timepoint. The manual conjugation reaction described previously (Dierick et al. 2024) was used as the starting point to optimise the fully automated radiosynthesis. The most notable change is the increase of the conjugation reaction volume with a factor of four, while keeping the concentrations of sdAb and conjugation buffer constant. This is necessary to allow for a higher volume of ethanolic [¹⁸F]SFB (800 µL) and to keep the ethanol concentration at 20 V/V% within the conjugation mixture, as higher concentrations of ethanol can negatively impact the sdAb. The increased reaction volume also impacts the purification step on the HiTrap cartridges. The new conjugation reaction volume (4 mL) exceeds the sample volume of two HiTraps in series. For this reason, three HiTraps were used with an air-eliminating filter placed on top, to ensure that no air bubbles were introduced onto the column resin during purification.

The last modification made, to keep the RCY of the conjugation reaction in line with the previously described results (Dierick et al. 2024), was raising the sdAb concentration from 1.0 mg/mL to 1.37 mg/mL.

Throughout the duration of the conjugation reaction (30 min), a low flow of N₂-gas (right lower pressure inlet connected to P36) was used to ensure adequate mixing by bubbling through the solution. During the purification step on the HiTrap columns, the crude mixture was removed from the reactor (Fig. 2, second reactor) by the 12 mL syringe P12, via the line connecting P32 and P18 and applied onto the HiTraps (flow rate: 5 mL/min) using the same line to reduce the loss of radiolabelled product. Subsequently, the second reactor was rinsed with 2 mL PBS (vial P23), and the rinse mixture was applied to the HiTraps via the same lines as the conjugation mixture. To start the elution of the [¹⁸F]FB-sdAb, rotor P35 is positioned to send the purified [¹⁸F]FB-sdAb towards the “final product vial”. The elution buffer containing 6.3 mL of PBS is located in vial P24 (labelled “eluent”). The eluent is collected in the syringe P12, via the line connecting P32 and P18 and applied thereafter to the HiTraps, using syringe P12, with a flow rate of 5 mL.min⁻¹). Six mL of purified [¹⁸F]FB-sdAb is collected in the “final product vial” which was prefilled with a Na-ascorbate solution to ensure that the final formulation of the tracer consists of 5 mg/mL of Na-ascorbate in saline at pH 5.4 ± 0.1 (final volume 20 mL).

Overall, for the medium activity (starting activity: 100 GBq trapped on QMA) productions, the conjugation reaction produced [¹⁸F]FB-HER2 sdAb with a RCY of 32 ± 3% (n = 4), starting from the activity measured in the second reactor at the start of the conjugation reaction. The purified [¹⁸F]FB-HER2 sdAb was obtained with a RCP of more than 95%, and the end of synthesis activity (measured with activimeter (ISOMED 2010)) amounted to 7.87 ± 0.95 GBq (n = 4). Apparent molar activity was 21.91 ± 2.90 GBq/µmol (n = 4). The automated conjugation reaction and its purification took 41 min. The whole automated tracer production requires 95 min and has an overall RCY d.c. of 14 ± 2% (n = 4).

Upscaling automated radiolabelling

Further upscaling of the radiolabelling was done using the same automated process as described above. The production’s starting activity was increased to approximately 350 GBq instead of 100 GBq trapped on the QMA (P5). In these conditions, the complete automated production of [¹⁸F]FB-HER2 sdAb has an overall RCY d.c. of 8 ± 1% (n = 4).

The first production step, the synthesis of the prosthetic group (PG), allowed to obtain 83.95 ± 7.43 GBq (n = 4) of [¹⁸F]SFB with a slightly decreased estimated RCY d.c. of 33 ± 2% (n = 4).

The second production step, the conjugation reaction, produced [¹⁸F]FB-HER2 sdAb with a RCY d.c. of 25 ± 3% (n = 4) respectively, starting from the activity measured in the second reactor at the start of the conjugation reaction. The purified [¹⁸F]FB-HER2 sdAb was obtained with a RCP of more than 95%, and the end of synthesis activity (measured with activimeter (ISOMED 2010)) amounted to 15.63 ± 1.14 GBq (n = 4) and the apparent molar activity was 43.29 ± 3.88 GBq/µmol (n = 4).

Shelf-life determination

The stability of the final product (final activity: > 16 GBq; 804 MBq/mL) was investigated by measuring the RCP directly after labelling and at different time points up to 12 h after production. Table 1 shows that within 12 h, the RCP did not go below the specification of ≥ 95%. Therefore, the expiration time is set at 12 h after production. As expected, the decrease in RCP is caused by the formation of radiolysis products (Fig. 3).

Table 1.

Shelf-life determination of [¹⁸F]FB-HER2: RCP determined by SE-HPLC of tracer determined at different time points after production

Timepoint (hours)	RCP [18F]FB-HER2 (%)
0	98.4
3	98.3
6	98.2
8	98.1
10	97.7
12	97.5

Fig. 3.

Chromatogram of [¹⁸F]FB-HER2 sdAb 12 h after production: Upper chromatogram: radioactive signal: RCP determined by SE-HPLC of tracer 12 h after production: ¹⁸F-labelled oligomers (Rt = 7.07 min): 0,2%: High Mass Molecular species; “.¹⁸F-FB-sdAb”: (Rt = 8.86 min): 97.5%: sdAb; “radiolysis product 1” (Rt = 12.75 min): 1.2%; “radiolysis product 2” (Rt = 14.03 min): 0.5%; “radiolysis product 3” (Rt = 16.63 min): 0.6%. Lower chromatogram: UV profile at 280 nm: sdab (Rt = 8.20 min); Na-ascorbate (Rt: 14.08 min)

Residual solvent analysis

Three possible residual solvents could be present in the final formulation based on the production process, namely acetonitrile and DMSO from the PG production and ethanol from the PG production and the conjugation reaction. Therefore the final product should be analysed for their presence. Specifications (see Table 2) were set based on the class of solvent as described in Ph. Eur. 5.4 (4. Residual solvents - European Pharmacopoeia 11.5 n.d. 2024). No trace of acetonitrile and DMSO could be detected in the final formulation of [¹⁸F]FB-HER2, while ethanol was well below the limit of a class 3 solvent (Table 2).

Table 2.

Overview of residual solvent determination (LOD = limit of detection; ppm: parts per million)

Solvent	Expected retention time	Class of solvent	Result
Acetonitrile	5.9 min	Class 2	< LOD
Dimethyl sulfoxide	5.5 min	Class 3	< LOD
Ethanol	5.2 min	Class 3	< 5000 ppm

Biodistribution studies and PET/CT imaging

Tumour-bearing mice (n = 4 per group) were i.v. injected with [¹⁸F]FB-HER2 sdAb (13 µg; 13 ± 2 MBq, 12.3 ± 1.8 GBq/μmol) and [¹⁸F]FB-R3B23 sdAb (26 µg; 13.3 ± 2.0 MBq, 7.1 ± 1.1 GBq/μmol) as negative control sdAb. The injected and apparent molar-specific activities are reported at the time of injection.

Tumour uptake of [¹⁸F]FB-HER2 sdAb was visible on the PET image (1 h p.i Fig. 4). It was confirmed by quantification of dissection data (80 min p.i.; Fig. 5), showing statistically significant (p < 0.01) higher tumour uptake (4.43 ± 1.24 IA/g) for the HER2- targeting sdAb compared to the non-targeting sdAb (0.16 ± 0.08 IA/g). For both tracers, fast excretion of the unbound tracer was observed via the kidneys ([¹⁸F]FB-HER2 sdAb: 8.25 ± 2.03% IA/g; [¹⁸F]FB-R3B23: 9.23 ± 3.55% IA/g).

Fig. 4.

Maximum intensity projection PET/CT imaging of [¹⁸F]FB-HER2 sdAb: mouse bearing SKOV-3 tumour 1 h p.i

Fig. 5.

Ex vivo biodistribution results: [¹⁸F]FB-HER2 sdAb compared to [¹⁸F]FB-R3B23 sdAb, 80 min post injection; Unpaired student t-test was used to calculate statistical significance. Statistical significance was set at p < 0.05 (ns, not significant, * p < 0.05; ** p < 0.01; *** p < 0.001; **** p < 0.0001)

Discussion

An important step in the translation of a radiofluorinated tracer to the clinic is the development of a fully GMP-compliant automated production process, which can be challenging and time-consuming. The starting point of this development was the semi-automated production process previously described (Dierick et al. 2024) in which the first step — the synthesis of the [¹⁸F]SFB prosthetic group — was already implemented on the Trasis automation module, but the second step in the process — the conjugation of HER2 sdAb to [¹⁸F]SFB — was still carried out manually. The first hurdle was the automation of the conjugation reaction and more specifically, the final purification. The purification of the radiolabelled sdAbs was performed using size-exclusion resin HiTrap desalting cartridges, three in series, instead of the PD-10 desalting column, with the latter being the most described option in literature (Xavier et al. 2016, 2019; Blykers et al. 2015; Nagachinta et al. 2022; Bala et al. 2016). Both the gravity-based PD-10 columns and the HiTrap desalting columns are packed with Sephadex G25 resin, allowing for a similar purification profile. The main advantage of these cartridges compared to gravity-based SE purification is their compatibility with the manifolds of our automation module, making it a plug-and-play approach. This approach was deemed more efficient than using a module-specific auxiliary device allowing the use of these gravity-based columns as reported by Nagachinta et al. (Nagachinta et al. 2022). Another option would have been to use a SE-HPLC to perform the purification, however this option was considered more time consuming and would require the Trasis AiO to include an HPLC module. With the complete production process implemented on the AiO, the upscaling of the process could start. Initial upscaling (starting with ± 100 GBq) did not show any negative impact on the PG’s production yield, i.e. 45 ± 7% compared to the 44 ± 4% as reported previously (Dierick et al. 2024). This shows that the changes required to the PG’s production to accommodate the subsequent conjugation reaction and its purification on the AiO, did not impact the PG’s yield. These results are in line with other automated production methods of [¹⁸F]SFB described in literature (Xavier et al. 2016; Al-Qahtani 2024; Ackermann et al. 2011; Blykers et al. 2015; Tang et al. 2010, 2008; Vaidyanathan and Zalutsky 2006; Thonon et al. 2011). To the authors knowledge, the automated PG production utilised, is unique in delivering the PG in ethanol instead of organic solvents such as acetonitrile or diethylether. This is a significant advantage when using the PG with biomolecules sensitive to organic solvents, such as small antibody fragments and sdAbs. With the automation of the conjugation reaction, upscaling of this second part of the production process to clinically relevant activities became feasible unlike for the previously described processes (Xavier et al. 2016; Scott and Shao 2010; Dierick et al. 2024; Bala et al. 2016). At the first upscaling level (starting activity of the PG’s production: 100 GBq) the conjugation’s RCY d.c. as well as the overall RCY d.c. reported in this study, are in line with or higher than the conjugation/ overall RCY previously reported for sdAbs and other proteins (Xavier et al. 2016; Scott and Shao 2010; Dierick et al. 2024; Nagachinta et al. 2022; Bala et al. 2016; Thonon et al. 2011; Davis et al. 2019). However, even at this upscaling level, the main advantage of a fully automated production process could already be observed, namely the significant increase in final activity of the purified [¹⁸F]FB-sdAb, compared to previously reported final activities of radiofluorinated sdAbs (Xavier et al. 2016; Dierick et al. 2024; Nagachinta et al. 2022) or even ⁶⁸Ga-labelled sdAbs (Xavier et al. 2019). The final activity was around 6 GBq, which would theoretically allow for four patient doses per production, assuming a patient dose of 150 MBq and 30 min between two patient injections. Further upscaling led to the production of [¹⁸F]FB-HER2 batches of almost 16 GBq at end of synthesis, which moreover would allow for shipment of the radiopharmaceutical to other hospital centres. The shelf life was determined to be 12 h after production for batches in this activity range, further strengthening the statement that central production and distribution to peripheral locations is possible. While the further upscaling did increase the final activity, the RCY of the overall production was reduced to 8%. This decrease in yield is mostly due to the PG synthesis yield that lowered to 33%. This reduction is most likely caused by radiolysis occurring during the production process of the PG. Including a radiolytic scavenger in the PG synthesis could be a strategy to increase its yield as well as the overall yield. For the coupling reaction, a similar decline in RCY, however less pronounced, could be observed. The addition of an extra radiolytic scavenger to the conjugation reaction, besides the ethanol already present, could have a positive impact on the overall yield. However, further increasing the ethanol content most likely will have a negative impact as denaturation of proteins caused by alcohols occurs at concentrations above 20%V/V (Nikolaidis and Moschakis 2018; Nikolaidis et al. 2017), thus other radiolytic scavengers should be considered. Further upscaling of the overall process was not explored within this study, as production batches of > 15 GBq are deemed sufficient for clinical production and distribution of the tracer.

The implementation of this automated production process as a generic radiofluorination strategy for sdAbs lays the groundwork for a reliable method that can be extended to a whole array of sdAbs. This development enables the possibility to ‘easily’ radiolabel other targeting sdAbs as new PET imaging tracers.

The preclinical in vivo and ex vivo biodistribution studies show specific tumour targeting of the tracer. These preclinical findings are in line with the findings previously published (Xavier et al. 2016) with an [¹⁸F]FB-HER2 sdAb showing that the upscaled and optimised production process did not negatively impact the specificity of the tracer. Furthermore, it confirms the kidney retention reduced with a factor of 4 with radiofluorinated HER2 sdAb compared to ⁶⁸Ga-labelled HER2 sdAb (Xavier et al. 2013). This reduction was expected as radiohalogenated compounds exhibit significantly less kidney retention (Chigoho et al. 2021). It is well known that the radiometabolites of tracers with metallic radionuclides, such as ⁶⁸Ga, are typically trapped in the lysosomal compartment of the kidneys, while radiometabolites of tracers with radiohalogens, such as ¹⁸F, are often rapidly eliminated from the kidneys into the urine (Chigoho et al. 2021; Maschauer et al. 2016; Bala et al. 2019; Läppchen et al. 2017).

The next steps to bring the [¹⁸F]FB-HER2 sdAb tracer to early phase clinical trials are dosimetry and toxicological studies, the tech transfer and validation of the automated production method in cGMP, the extension of the quality control (QC) of the final product and preparation of the Investigational Medicinal Product Dossier (IMPD) and Investigation brochure (IB). The QC of the final product needs to be extended from RCP and residual solvent determination to a complete QC as required for radiopharmaceuticals, including radionuclide specific QC tests (radionuclide identity and purity, half-life determination and Kryptofix determination) and parenteral drug specific QC tests (appearance, free from particles, osmolarity, sterility and endotoxins determination). Most of these QC methods are compendial methods described in the Ph. Eur. Assessment of the human biodistribution and dosimetry of [¹⁸F]FB-HER2 sdAb tracer will also allow for a head-to-head comparison with the [⁶⁸Ga]Ga-NOTA-anti-HER2 sdAb for clinical use (Keyaerts et al. 2019, 2016).

Conclusion

The [¹⁸F]FB-HER2 sdAb production method was successfully automated and upscaled to the point of yielding clinically relevant activities of pure radiofluorinated sdAb. The fully automated production is fast (95 min), robust and reproducible. The automated production is performed on a cGMP compliant automation module and does not require the use of an HPLC module. With one single production, up to 16 GBq of [¹⁸F]FB-HER2 sdAb was obtained. This opens opportunities to extend the use of the tracer from single to multi- centre clinical trials with one central production site. The [¹⁸F]FB-HER2 sdAb showed specificity towards the HER2-antigen in vivo, which combined with the significantly lower kidney retention, makes it a valuable alternative to ⁶⁸Ga-labelled sdAbs targeting the same antigen.

Abbreviations

[¹⁸F]SFB

N-Succinimidyl 4-[¹⁸F]Fluorobenzoate

¹⁸F

Fluorine-18

⁶⁸Ga

Gallium-68

AiO

AllInOne

ATCC

American Type Culture Collection

CHES

2-(Cyclohexylamino)ethane-1-sulfonic acid

d.c.

Decay corrected

DMSO

Dimethyl sulfoxide

EtOH

Ethanol

FISH/ISH

Fluorescence in situ hybridization

GMP

Good Manufacturing Practices

HER2

Human Epidermal Growth factor Receptor 2

HPLC

High Performance Liquid Chromatography

HSTU

N,N,N′,N′-Tetramethyl-O-(N-succinimidyl)uronium hexafluorophosphate

IB

Investigation brochure

IHC

Immunohistochemistry

IMPD

Investigational Medicinal Product Dossier

LOD

Limit of detection

mAbs

Monoclonal antibodies

P

Position

PBS

Phosphate buffered saline

PET

Positron emission tomography

PG

Prosthetic group

Ph. Eur.

European Pharmacopoeia

Ppm

Parts per million

QC

Quality Control

R&D

Research and development

RCP

Radiochemical purity

RCY

Radiochemical yield

RM

Reaction mixture

SD

Standard deviation

SdAbs

Single-domain antibodies

SE

Size-Exclusion

SPE

Solid-phase extraction

TPAOH

Tetrapropylammonium hydroxide

Author contributions

All authors, T.L., V.C., L.N., J.B and H.D., contributed to the study conception and design. Material preparation, data collection and analysis were performed by H.D. S.V.d.B. and J.S carried out automated productions and assisted in data collection. The manuscript was written by H.D. All authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.

Funding

This research was performed with the financial support of Strategic Research Programs (SRP50, SRP 95) and the Industrial Research Fund (IOF3018 and IOF3009) of the VUB Research Council. J.B. is a postdoctoral fellow (1230824N) of the Research Foundation Flanders (FWO-V) during the execution of this work. This research was partly performed at ICMI Core facility, a core facility financially supported by the University Medical Center Onderzoeksraad (UMCOR). The Molecubes β-CUBE PET/CT system were funded via an FWO-Hercules grant (I005622N).

Availability of data and materials

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Ethics approval and consent to participate

The ethical committee approved the animal study protocols for animal experiments at the Vrije Universiteit Brussel (22–272-3). All mice experiments were executed in accordance with the European guidelines for animal experimentation. Written informed consent was not required for this study.

Consent for publications

Not applicable.

Competing interests

J.B. and T.L. have patents on using sdAbs for imaging and therapy. T.L. has ownership in AbScint and holds ownership in Precirix. L.N. is an employee of Precirix. J.B is an unpaid board member of eSRR.