Microfluidic Preparation of 89Zr-Radiolabelled Proteins by Flow Photochemistry

89Zr-radiolabelled proteins functionalised with desferrioxamine B are a cornerstone of diagnostic positron emission tomography. In the clinical setting, 89Zr-labelled proteins are produced manually. Here, we explore the potential of using a microfluidic photochemical flow reactor to prepare 89Zr-radiolabelled proteins. The light-induced functionalisation and 89Zr-radiolabelling of human serum albumin ([89Zr]ZrDFO-PEG3-Et-azepin-HSA) was achieved by flow photochemistry with a decay-corrected radiochemical yield (RCY) of 31.2 ± 1.3% (n = 3) and radiochemical purity >90%. In comparison, a manual batch photoreactor synthesis produced the same radiotracer in a decay-corrected RCY of 59.6 ± 3.6% (n = 3) with an equivalent RCP > 90%. The results indicate that photoradiolabelling in flow is a feasible platform for the automated production of protein-based 89Zr-radiotracers, but further refinement of the apparatus and optimisation of the method are required before the flow process is competitive with manual reactions.


Introduction
Due to their unique structures and functions, high affinity and target specificity, protein-based drug-conjugates have fast become essential tools in medical imaging. Proteinderived immunoglobulin fragments and monoclonal antibodies (mAbs) are frequently utilised for the development of both therapeutic and diagnostic agents [1]. For example, in the clinic, mAbs functionalised with metal-binding chelates such as desferrioxamine B (DFO) and radiolabelled with zirconium-89 ( 89 Zr-mAbs) provide sophisticated, rationally designed, radiopharmaceuticals for use in positron emission tomography (PET) [2].
The preparation of radiolabelled protein-conjugates requires the formation of a new covalent bond between the protein and the ligand, which must be achieved without disrupting the structural and biological properties of the target protein. Standard conjugation methods require the use of pre-installed reactive groups such as activated esters or benzylisothiocyanates, prior chemical modification and/or pre-activation of the protein [3], and rely on thermochemically-driven reactions (at room temperature to 37 • C) with amino-acid side chains or glycans [4][5][6][7]. These processes often include time consuming multi-step syntheses that are difficult to automate, and can be incompatible with protein formulation buffers which mandate pre-purification of the protein vector.
In the last decade, flow chemistry has been combined with radiochemistry to prepare a variety of radiotracers with different radionuclides, including 11 C [8,9] and 18 F [10][11][12][13]. Small reaction volumes, efficient mixing and reproducible control over all essential reaction parameters are some of the features that make microfluidic reactions attractive for radiochemistry. In 2016, Wright et al. [14] demonstrated the microfluidic radiolabelling of 89 Zr-mAbs. In 2019, Poot et al. [15] reported proof-of-concept studies demonstrating that automated radiolabelling of 89 Zr-mAbs in a batch reactor can also be combined with automated purification. However, both approaches relied on the use of pre-functionalised proteins bearing the DFO chelate. In contrast, we provide a proof-of-concept showing that flow-based photochemistry can be combined with radiochemistry to produce 89 Zrradiolabelled proteins direct from the unfunctionalised (native) protein source.
Light-induced functionalisation of proteins with metal-binding chelates bearing photochemically active groups presents an alternative to traditional protein conjugation chemistries [4,16]. The aryl azide (ArN 3 ) group absorbs various wavelengths of light (302-400 nm) to generate highly reactive nitrenes [17], which can be utilised to form new covalent bonds to a target protein [16]. This light-induced process is compatible with biologically relevant media, and occurs extremely rapidly (lifetimes of reactive intermediates are in the nanosecond to microsecond range) compared with traditional bioconjugation methods that react directly with native functional groups on the protein [18]. The proposed reaction mechanism favours the formation of a seven-membered azepine ring species kinetically which can then react rapidly with nucleophiles like primary amines to form a new covalent bond [16].
We recently reported the photoradiosynthesis of several viable 68 Ga 3+ and 89 Zr 4+ protein-conjugate PET radiotracers, from photoactivatable metal-binding chelates functionalised with an ArN 3 group [19][20][21][22]. Importantly, this photochemical conjugation process occurs at wavelengths that do not disrupt protein structure or function. Photoradiolabelling to produce 89 Zr-mAbs is compatible with several different antibody formulations (with mixtures containing large quantities of histidine; ascorbic acid; sugars such as α,α-trehalose; and surfactants such as polysorbate 20) which allows for direct protein-conjugation without the need for pre-purification of the protein before performing the bioconjugation step.
To enhance the water-solubility of the photoactivatable DFO derivatives, we recently introduced two new compounds (including DFO-PEG 3 -Et-ArN 3 1; Scheme 1) that link the metal binding chelate to the ArN 3 group via a polar tris-polyethylene glycol (PEG 3 ) linker [22]. Compound 1 is an excellent ligand for exploring the potential for automated radiosynthesis of 89 Zr-radiolabelled protein conjugates via photochemistry in flow ( Figure 1).  Here, we present the synthesis of a radiolabelled protein conjugate prepared by lightinduced photoconjugation using a microfluidic photochemical reactor in continuous flow.

Chip Design and Instrumentation
Flow photoradiochemistry was performed by using a FutureChemistry FlowStart B-222 photochemistry module ( Figure 2A) equipped with a twin light-emitting diode (LED) light source (365 nM; LedEngin, Inc.) connected in series ( Figure 2B). Light intensity was set to 100% power and was controlled by using a prototype FutureChemistry B-271 photochemistry module. The emission profile was measured experimentally with an emission maximum observed at 366.5 nm ( Figure 2C; full-width at half-maximum ≈14 nm). The photochemical flow reaction was performed using a mounted FutureChemistry borosilicate glass microfluidic chip with approximately 700 wide and 500 µm deep channels and a total internal volume of 112 µL ( Figure 2D). The chip is comprised of two segments including a shorter split and recombine mixing section (which splits the flow and recombines it multiple times to ensure homogeneity of solutions injected), and a longer linear reaction channel. Due to the small channels, the surface area of the microfluidic chip is much larger compared to standard batch reactors, which ensures efficient heat transfer and exposure of the reagents to light for sufficient time to complete the photochemical activation of the ArN 3 group [25,26].

Flow Radiochemistry
The [ 89 Zr]ZrDFO-PEG 3 -Et-azepin-HSA protein conjugate was prepared by flow photoradiochemistry on a microfluidic chip by reaction of a pre-labelled solution of [ 89 Zr]ZrDFO-PEG 3 -Et-ArN 3 ( 89 Zr-1 + ; solution A) at a pH of 8.0 to 8.5, and a solution of native (unfunctionalised) HSA in Chelex-treated water (solution B; 45 mg mL -1 protein concentration; Scheme 2). As an example, solution A was prepared by incubating ligand 1 (20 µL; 2 mM stock solution <1% DMSO/H 2 O) with aliquots of [ 89 Zr][Zr(C 2 O 4 ) 4 ] 4-(40 µL; 5.532 MBq) in H 2 O (40 µL; pre-treated with Chelex-100 resin) at room temperature and at a pH of 8.0 to 8.5 (the optimal range for ArN 3 photoconjugation) [19,20]. Quantitative 89 Zr-radiolabelling yields were obtained in <5 min and 89 Zr-1 + was characterised by radio-instant thin layer chromatography (radio-iTLC; Figure 3A) and radio-HPLC methods ( Figure 3B). The chemical identity and radiochemical purity (RCP) of 89 Zr-1 + ( Figure 3B; blue trace) was confirmed by comparison of the elution profile of the corresponding [ nat Zr]ZrDFO-PEG 3 -Et-ArN 3 ( nat Zr-1 + ) complex ( Figure 3B; green trace). The 89 Zr-1 + complex was then used in the microfluidic photoconjugation reaction without further purification. Irradiation of a solution of nat Zr-1 + with a powerful LED confirmed the complex was photochemically active ( Figure 3B; green trace). Then, by using a pair of syringe pumps operated by independent drive units, 100 µL of solution A and 100 µL of solution B were injected simultaneously onto the microfluidic chip at a flow rate of 5 µL/min. The reaction vessel was then irradiated at a wavelength of 365 nm for 20 min (100 µL total volume of chip). After this time, 150 µL of H 2 O (for each syringe) was injected on to the chip (to flush the system) with irradiation continuing for a further 20 min at the same flow rate. The crude product was collected in an Eppendorf tube, reaching a final volume of approximately 500 µL. Aliquots of the crude [ 89 Zr]ZrDFO-PEG 3 -Et-azepin-HSA protein conjugate mixtures were retained for analysis and fractions purified by size-exclusion gel filtration (PD-10) chromatography. Crude and purified samples were then characterised by radio-iTLC, analytical PD-10 size-exclusion chromatography (SEC), and automated radio-HPLC equipped with a SEC gel-filtration column (Figure 4). After optimisation of the reaction conditions, the decay-corrected radiochemical yield (RCY) for the isolated [ 89 Zr]ZrDFO-PEG 3 -Etazepin-HSA product was 31.2 ± 1.3% (n = 3 independent experiments; with final protein concentration in the reaction mixture of 135 µM; errors reported as 1 standard deviation). For each reaction the radiochemical purity (RCP) of the isolated product was >90% (determined by HPLC). The fraction of protein aggregation (indicated by an asterisk in Figure 4C) was <10%. Experimental data confirm that the 89 Zr-radiolabelled proteins can be produced by photochemical methods in an automated, microfluidic system. For comparison, manual reactions were also performed by using direct, top-down irradiation a stirred reaction mixture in a ≈1 mL glass vial. In this manual approach previously reported by Guillou et al., [22] [ 89 Zr]ZrDFO-PEG 3 -Et-azepin-HSA was produced with a decay-corrected RCY of 59.6 ± 3.6% (n = 3) and with a RCP > 90%. Further refinement of the microfluidic apparatus and optimisation of the chemical methods is required before the process is competitive with manual reactions.

Conclusions
The synthesis of [ 89 Zr]ZrDFO-PEG 3 -Et-azepin-HSA, was achieved by light-induced flow photoconjugation by using commercially available photochemistry modules. Using a microfluidic chip comprised of a split and recombine mixing section and a linear reaction section, coupled to a twin LED light source, [ 89 Zr]ZrDFO-PEG 3 -Et-azepin-has was prepared in a radiochemical yield of 31.2 ± 1.3% (n = 3) in high radiochemical purity (>90%). The results provide an encouraging proof-of-concept that continuous flow procedures can be developed to produce protein-based radiotracers by automated instrumentation.

General
All reagents and anhydrous solvents were purchased from commercial sources (Sigma-Aldrich (St. Louis, MO, USA), Merck (Darmstadt, Germany), Tokyo Chemical Industry (Eschborn, Germany) or abcr (Karlsruhe, Germany)) and were used without any further purification unless otherwise stated. All aqueous reactions were carried out using MilliQ H 2 O (>18.2 M MΩ·cm at 25 • C, Merck, Darmstadt, Germany). All anhydrous reactions were carried out in oven-dried glassware under an inert atmosphere. Reactions (where possible) were monitored using thin layer chromatography (TLC) analysis on aluminium plates coated with Silica Gel 60 F 254 (E. Merck), and were visualised by short-wave ultraviolet irradiation (254 nm; where applicable), stained with ninhydrin in EtOH or propargyl alcohol and Cu(I)Br in EtOH, followed by charring at ≈200 • C. Purification was carried out by flash chromatography on a column of silica gel 60 (0.040-0.063 mm) or by reversedphase C18 column chromatography using a Teledyne Isco CombiFlash ® Rf+ Lumen flash chromatography system fitted with RediSep Rf Gold ® reversed-phase C18 columns (5 to 50 g), eluting in a gradient of 0 to 100% of solvent B (MeOH with 0.1% TFA added). Solvent A: MilliQ H 2 O with 0.1% TFA added. Evaporation of solvents was performed under reduced pressure by using a rotary evaporator (Rotavapor R-300, Büchi Labortechnik AG, Flawil, Switzerland). 1 H NMR and 13 C{ 1 H} NMR experiments were performed using deuterated solvents (Sigma-Aldrich, St. Louis, MO) on a Bruker AV-400 ( 1 H: 400 MHz, 13 C: 100.6 MHz) or a Bruker AV-500 ( 1 H: 500 MHz, 13 C: 125.8 MHz) spectrometer. Chemical shifts (δ) for 1 H and 13 C spectra are reported in parts per million (ppm) and are relative to the residual solvent peak. Coupling constants (J) are reported in Hz. Peak multiplicities are abbreviated as follows: s (singlet), d (doublet), dd (doublet of doublets), t (triplet), q (quartet), quint (quintet), m (multiplet), and br s (broaden singlet). Two-dimensional 1 H-1 H correlation spectroscopy (COSY) and 13 C heteronuclear single quantum coherence (HSQC) NMR experiments were also performed to aid in the assignment of the 1 H and 13 C spectra, respectively. High-resolution electrospray ionisation mass spectrometry (HR-ESI-MS) was performed in either positive or negative ionisation mode (as indicated) using a Bruker MaXis QTOF-MS instrument (Bruker Daltronics GmbH, Bremen, Germany) and was measured by the mass spectrometry service at the Department of Chemistry, University of Zurich.
Purities of synthetic intermediates after chromatographic purification were judged to be >90% by analysis of 1 H and 13 C NMR spectra. Purities of final compounds were ≥95% (NMR or HPLC analysis), after reverse-phase C18 chromatography.

Synthesis of DFO-PEG 3 -Et-ArN 3 (1)
Synthesis of compound 2. 3-(4-Aminophenyl)propanoic acid (1.00 g, 6.05 mmoL) was taken up in MeOH (30 mL) to which imidazole-1-sulphonyl azide HCl (1.52 g, 7.26 mmol), K 2 CO 3 (2.26 g, 16.3 mmoL) and CuSO 4 ·5H 2 O were added and the reaction mixture was stirred for 16 h at rt. The reaction was monitored by TLC and on complete conversion of the starting materials, the mixture was concentrated under reduced pressure. The resulting crude residue was dissolved in H 2 O (60 mL), acidified with concentrated HCl and extracted with EtOAc (3 × 50 mL). The organic layers were combined, dried over NaSO 4 and concentrated under reduced pressure. The residue was then co-evaporated with cyclohexane to give compound 3 (875 mg, 76% yield) as a pale-yellow solid. Synthesis of compound 5. Compound 4 (638 mg, 1.29 mmoL) was dissolved in CH 2 Cl 2 (10 mL) and cooled to 0 • C, and then TFA (2 mL) was added drop-wise. The reaction mixture was then allowed to warm slowly to rt and then stirred for 1 h. At this time, TLC (10% MeOH in EtOAc) showed complete consumption of the starting material. Synthesis of DFO-PEG 3 -Et-ArN 3 (1). Compound 6 (222 mg, 0.45 mmoL) and HATU (232 mg, 0.61 mmoL) were dissolved in dry DMF (6 mL) and stirred for 20 min under N 2 (g). DFO mesylate (267 mg, 0.41 mmoL) was then dissolved in DMF (4 mL) and added to the reaction mixture with stirring for a further 10 min. At this time, DIPEA (0.29 mL, 1.63 mmoL) was added and the was reaction stirred under N 2 (g) for 16 h at rt. The reaction was monitored by TLC, and on completion the mixture was concentrated under reduced pressure and the crude residue was purified by flash column chromatography (C18, H 2 O/MeOH 0% MeOH to 100%) followed by washing with ice-cold acetone (6 × 5 mL; separated by centrifugation between each wash) to give compound 1 (DFO-PEG 3 -Et-ArN 3 ; 244 mg, 58% yield) as an off-white solid.

Flow Photochemistry
Flow photoradiochemistry was performed using a FlowStart B-222 (Future Chemistry, Nijmegen, The Netherlands) photochemistry module equipped with a twin light-emitting diode (LED; LedEngin Inc., San Jose, CA, USA) light source (365 nM), connected in series. Light intensity was set to 100% power and controlled using a prototype Future Chemistry B-271 (Future Chemistry, Nijmegen, The Netherlands) photochemistry module. LED intensity was measured by using a S470C Thermal Power Sensor Head Volume Absorber, 0.25-10.6 µm, 0.1 mW-5 W, Ø15 mm. Light intensity for each LED was 366.5 nm (FWHM of ≈10 nm). The photochemical flow reactions were performed using a mounted Micronit microfluidics E3 custom borosilicate glass chip (Future Chemistry, Nijmegen, The Netherlands) with an internal diameter width of less than 700 µm, depth of 500 µm and a 112 µL total volume. The temperature of all photochemical conjugation reactions was typically 23 ± 2 • C (ambient conditions).

Radioactivity and Radioactive Measurements
All instruments for measuring radioactivity were calibrated and maintained in accordance with previously reported routine quality control procedures.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author. All relevant data are presented in the manuscript and supporting information.