Automated light-induced synthesis of 89Zr-radiolabeled antibodies for immuno-positron emission tomography

Clinical production of 89Zr-radiolabeled antibodies (89Zr-mAbs) for positron emission tomography imaging relies on the pre-conjugation of desferrioxamine B (DFO) to the purified protein, followed by isolation and characterization of the functionalized intermediate, and then manual radiosynthesis. Although highly successful, this route exposes radiochemists to a potentially large radiation dose and entails several technological and economic hurdles that limit access of 89Zr-mAbs to just a specialist few Nuclear Medicine facilities worldwide. Here, we introduce a fully automated synthesis box that can produce individual doses of 89Zr-mAbs formulated in sterile solution in < 25 min starting from [89Zr(C2O4)4]4– (89Zr-oxalate), our good laboratory practice-compliant photoactivatable desferrioxamine-based chelate (DFO-PEG3-ArN3), and clinical-grade antibodies without the need for pre-purification of protein. The automated steps include neutralization of the 89Zr-oxalate stock, chelate radiolabeling, and light-induced protein conjugation, followed by 89Zr-mAb purification, formulation, and sterile filtration. As proof-of-principle, 89ZrDFO-PEG3-azepin-trastuzumab was synthesized directly from Herceptin in < 25 min with an overall decay-corrected radiochemical yield of 20.1 ± 2.4% (n = 3), a radiochemical purity > 99%, and chemical purity > 99%. The synthesis unit can also produce 89Zr-mAbs via the conventional radiolabeling routes from pre-functionalized DFO-mAbs that are currently used in the clinic. This automated method will improve access to state-of-the-art 89Zr-mAbs at the many Nuclear Medicine and research institutions that require automated devices for radiotracer production.

. In both reports, the radiolabeling reaction relies on the use of a pre-functionalized desferrioxamineantibody conjugate (DFO-mAb). In 2016, Wright et al. performed the 89 Zr-radiolabeling of DFO-Bz-NCStrastuzumab on a microfluidic chip followed by manual purification. The procedure was performed in 45-60 min and gave an isolated product with sufficient activity yield for multiple patient doses 15 . More recently, Poot et al. reported the fully automated radiolabeling and purification of 89 Zr-mAbs in 2019 3 . In this example, single patient doses of purified 89 ZrDFO-N-Succ-rituximab and 89 ZrDFO-N-Succ-cetuximab were produced in 77 min starting from 89 Zr-oxalate and the respective pre-functionalized DFO-mAb 3 . The DFO-mAb radiolabeling precursors are produced by DFO-conjugation to lysine residues on the protein using reagents bearing the activated ester N-hydroxysuccinimide (NHS) or a benzylisothiocyanate (Bz-NCS) group which form amide and thiourea bonds, respectively 16,17 . DFO-mAb conjugates are typically prepared in advance, characterized, stored, and radiolabeled on demand, and so far, this holds true for all automated and manual production of 89 Zr-mAbs. In general, this chemistry is extremely successful and has facilitated clinical translation of many 89 Zr-mAbs [18][19][20][21][22] . However, the necessity to produce, characterize and store the DFO-mAb intermediate presents several technological and financial limitations-synthesizing sufficient material for toxicological studies is both difficult and expensive-that mean accessing 89 Zr-mAbs is not always feasible in smaller nuclear medicine facilities. In addition, the long-term storage of an intermediate can lead to questions over the stability and shelf-life of the material.
Recently, we developed a one-pot route that combines the bioconjugation and radiolabeling steps using new DFO-chelates bearing a photoactivatable aryl azide group (ArN 3 ) [23][24][25][26] . Three key features of this alternative photoradiochemical method for making 89 Zr-mAbs are: i) rapid reaction times that depend on the rate-limiting photoactivation step; ii) high chemical tolerance of the light-induced bioconjugation process with many components of mAb formulation buffers, and iii) avoidance of the need to use pre-functionalized DFO-mAbs. The compatibility of the photolabeling process with common mAb formulation buffers, including high concentrations amino acids (e.g. histidine), surfactants (polysorbate-80), sugars (α,α-trehalose), antioxidants (ascorbate), and various salts such as phosphate buffered saline (PBS) mean that the chemistry often works without the need to pre-purify the protein from clinical-grade stocks. This critical difference between our photoradiosynthesis approach and the classic multiple step routes for making 89 Zr-mAbs led us to postulate that it could be feasible to make an automated radiosynthesizer that combines all critical steps required in the manufacture of 89 Zr-mAbs, namely, oxalic acid neutralization, bioconjugation, radiolabeling, purification and sterile formulation (Fig. 1).
Here, we report the design, manufacture, and proof-of-concept pre-clinical evaluation of ALISI -a prototype radiosynthesizer for the fully automated light-induced synthesis of 89 Zr-mAbs for immuno-positron emission Zr-stock solution, chelate radiolabeling, light-induced protein conjguation, in-line purification, sterile filtration and product formulation in < 25 min at the push of a button. www.nature.com/scientificreports/ tomography (PET). An overview of the design and plumbing diagram of the ALISI system is presented in Fig. 2 with full descriptions given in the Methods section.

Results
Prior to developing the fully automated synthesis of 89 Zr-mAbs, we first tested different aspects of the ALISI system by using model reactions involving the bioconjugation between human serum albumin (HSA) and three photoactivatable compounds. The compounds used were RhodB-PEG 3 -ArN 3 , pre-radiolabeled 68 GaDFO-PEG 3 -ArN 3 , and the free ligand (DFO-PEG 3 -ArN 3 ) together with 89 Zr-oxalate which forms 89 ZrDFO-PEG 3 -ArN 3 in situ. Chemical structures are shown in Figs. 1 and 3a, and reaction details are presented in Supplemental  Tables 1-3. Photochemical activation kinetics were measured to assess the efficiency of the newly designed photoreactor by irradiation of a stock solution of DFO-PEG 3 -ArN 3 . Samples were analyzed by reverse-phase high-performance liquid chromatography (HPLC). The irreversible photoactivation of DFO-PEG 3 -ArN 3 was complete within 90 s, as indicated by full consumption of the starting material in HPLC (Fig. 3b).
The three model reactions were used to assess different aspects of the automated system. First, the photoactivatable fluorophore RhodB-PEG 3 -ArN 3 was used to optimize the sequence and timing of the different steps in the automation protocol. The use of a highly colored (pink) RhodB-PEG 3 -ArN 3 dye also helped to evaluate the completeness of all reagent transfers by visual tracking and by quantitative measurements of dilution factors using electronic absorption spectroscopy. We discovered that the lipophilic nature of the small-molecule RhodB-PEG 3 -ArN 3 compound, and the related photolyzed products, means that these species stick to Sephadex columns which are typically used for protein purification. Consequently, the fluorescently labeled RhodB-PEG 3azepin-HSA product can be purified easily from the small-molecule byproducts on the ALISI system by using standard PD-10 columns but this model fluorophore was not ideal for optimizing the purification procedure for radiolabeled proteins.
Next, test reactions between HSA and pre-radiolabeled 68 GaDFO-PEG 3 -ArN 3 were used to optimize the photochemical conjugation step. Optimization was performed by maximizing the decay-corrected radiochemical www.nature.com/scientificreports/ conversion (RCC, measure by size-exclusion chromatography high-performance liquid chromatography, SEC-HPLC) to the labeled protein, 68 GaDFO-PEG 3 -azepin-HSA. We found that pH is a critical factor in determining the yield of photo-induced bioconjugation reactions using ArN 3 species 27 . Therefore, we investigated the use of different buffers with the aim of controlling pH throughout all synthetic steps. 2-[4-(2-hydroxyethyl)piperazin-1-yl]ethanesulfonic acid (HEPES) is a zwitterionic buffer commonly used in the synthesis or radiolabeled mAbs 17 . In our hands, the use of HEPES at concentrations ≥ 0.5 M had a detrimental impact on the RCC of 68 GaDFO-PEG 3 -azepin-HSA (data not shown). In contrast, reactions in sodium borate buffer (0.25 M, pH8) did not quench the light-induced conjugation reaction. Experimental data showing the measured RCC versus reaction pH are presented in Fig. 3c. These data show a trend toward increased RCC in the pH window from 8.0-to-9.5, with RCC reaching 47% between pH8.0-8.7. Below pH8.0, conjugation yields decrease sharply due to progressive protonation of lysine residues. Based on these data, further reaction parameters were optimized at pH8.0. Our goal was to minimize the manual handling of radioactive components by producing 89 Zr-mAbs directly from 89 Zr-oxalate stocks and the unfunctionalized protein. Therefore, we examined the complexation reaction to form 89 ZrDFO-PEG 3 -ArN 3 on ALISI (Fig. 3d) 24,25 . 89 Zr-oxalate was transferred into the reaction vial, and then neutralized by automatic transfer of the Na 2 CO 3 and sodium borate solutions. Then, the DFO-PEG 3 -ArN 3 solution was transferred into the magnetically stirred vial, and 10 s after addition, an aliquot of the reaction mixture was removed manually with a pipette and spotted immediately onto a silica gel instant-thin-layer chromatography (ITLC) strip (developed with 50 mM DTPA, pH7.4). The radio-ITLC data showed quantitative radiochemical conversion to give 89 ZrDFO-PEG 3 -ArN 3 (which is retained at the baseline, R f = 0.0) in < 10 s. In contrast to radiolabeling of pre-functionalized DFO-mAbs which requires > 45 min, the high diffusion coefficient of the small-molecule components and gentle stirring of the reaction mixture facilitate rapid 89 Zr-complexation by DFO-PEG 3 -ArN 3 . www.nature.com/scientificreports/ Data in Fig. 3e show the effect of changing the protein concentration on the measured photochemical conversion yields (PCY) for the synthesis of RhodB-PEG 3 -azepin-HSA (purple), or RCC yields for 68 GaDFO-PEG 3azepin-HSA (blue) and 89 ZrDFO-PEG 3 -azepin-HSA (black). As with many bioconjugation reactions, photoinduced labeling using ArN 3 compounds show a steep dependence on the initial concentration of protein.
The purification of 89 Zr-mAbs typically involves using PD-10 desalting columns (Sephadex G-25, 5 kDa exclusion limit) which only work well when very high RCCs are obtained, or when the small-molecule byproducts are trapped on the Sephadex (as with the RhodB-PEG 3 -azepin-HSA synthesis, vide supra). However, when RCCs drop below ~ 80%, the separation efficiency of the PD-10 matrix is sub-optimal. To enhance the radiochemical purity (RCP) of the purified 89 Zr-labeled proteins, we constructed custom-made SK-10 columns that have the same geometry as PD-10 columns but are loaded with Sephadex® G-100 (150 kDa exclusion limit) that increases the resolving power. SK-10 columns had a dead volume of ~ 2 mL followed by a protein collection volume of 2.5 mL (Fig. 4a). Small-molecules were effectively retained on SK-10 columns, where a peak-to-peak separation between the desired 89 ZrDFO-PEG 3 -azepin-trastuzumab and the byproducts was > 3 mL, giving a final RCP of the purified protein fraction of > 99% starting from a crude reaction mixture that had an RCC of 42%.
Next, we investigated the fully automated production of 89 ZrDFO-PEG 3 -azepin-trastuzumab using ALISI (Supplemental Table 4). Again, changing the protein concentration had an impact on both decay-corrected RCC for the protein conjugation step and on the isolated decay-corrected RCY of 89 ZrDFO-PEG 3 -azepin-trastuzumab where a plateau was observed in the protein concentration range of 10-15 mg mL -1 (Fig. 4b). The same trend observed in reactions with HSA was seen with Herceptin. Automation successfully gave purified 89 ZrDFO-PEG 3azepin-trastuzumab in < 25 min starting directly from non-purified Herceptin (14.9 mg mL -1 of protein) with an isolated decay-corrected RCY of 20.1 ± 2.4% (n = 3), and an RCP > 99% (Fig. 4c). Notably, SEC analysis of the isolated 89 ZrDFO-PEG 3 -azepin-trastuzumab product revealed less than 5% of the activity was associated with an aggregated protein fraction (which elutes at a slightly shorter retention time). The activity recovered in the purified product was lower than the RCC by a factor of ~ 2. By tracking the activity along the fluid pathway, we found that 89 Zr-oxalate was loaded into the reaction vial with 82.7 ± 3.1% efficiency. The photochemical conjugation efficiency is given by the RCC (~ 45%) and loading of the crude reaction mixture onto the SK-10 column had a transfer efficiency > 99%. The purification step was, on average, 75.0 ± 4.3% efficient, while sterile filtration led to a small loss in protein-bound activity with a recovered activity efficiency of 87.7 ± 2.8%. Transfer losses account for the difference between the measured RCC and RCY values. www.nature.com/scientificreports/ Qualitative testing of the 89 ZrDFO-PEG 3 -azepin-trastuzumab products for residual borate was performed by adding a few drops of an ethanolic solution of curcumin, which produces a red/orange solution due to the formation of rosocyanine dye in the presence of borate species 28 . Visual detection is possible at borate concentrations down to ~ 2.5 mM and tests confirmed that none of the isolated samples of 89 ZrDFO-PEG 3 -azepin-trastuzumab produced by ALISI contained residual sodium borate.
In the clinic, patients undergoing PET scans with 89 Zr-mAbs typically receive a dose of 37 MBq of activity administered with between 3 -100 mg of total protein [29][30][31] . In scaled-up syntheses (Supplemental Table S4), we demonstrated the potential of ALISI to produce individual patient doses of 89 ZrDFO-PEG 3 -azepin-trastuzumab. Starting from 152 MBq of 89 Zr-oxalate and 9.7 mg of trastuzumab (formulated as Herceptin), photoradiosynthesis gave an activity yield of 27.9 MBq of isolated 89 ZrDFO-PEG 3 -azepin-trastuzumab with a decay corrected RCY of 18.3%, an RCP > 99%, a chemical purity > 99%, and a molar activity A m of 0.43 MBq nmol -1 of protein. If the ALISI system was qualified for use in a clinical radiopharmacy, this reaction product would be sufficient to image a patient.
Finally, to illustrate the flexibility of the ALISI radiosynthesizer, we adapted the system to automate the radiolabeling and purification of 89 Zr-mAbs via the conventional two-step approach (Fig. 5a). First, trastuzumab was recovered as a purified protein from a clinical-grade sample of formulated Herceptin by using PD-10 gel filtration. Next, the functionalized DFO-Bz-NCS-trastuzumab radiolabeling precursor was produced in accordance with the methods of Vosjan et al. 16,17 . Finally, 89 Zr-radiolabeling and automated PD-10 purification using ALISI gave 89 ZrDFO-Bz-NCS-trastuzumab with a decay-corrected isolated RCY of 48.3 ± 8.4% (n = 3), an RCP > 99%, and chemical purity > 99% (Supplemental Table S5) in 90 min. Corresponding radio-ITLC and SEC-HPLC characterization data for 89 ZrDFO-Bz-NCS-trastuzumab are shown in Fig. 5b and c, respectively. Importantly, the purified sample of 89 ZrDFO-Bz-NCS-trastuzumab contained ~ 15% of 89 Zr-activity associated with an aggregated protein peak. This is common feature of the conventional two-step radiolabeling method used in the clinic, which requires removal of the mAb from the stabilizing formulation components prior to functionalization with DFO-Bz-NCS 23 . The presence of 89 Zr-labeled protein aggregate is associated with accumulation of activity in the liver www.nature.com/scientificreports/ and spleen. In comparison, the photoradiochemical approach reduces this aggregate fraction by a factor of ~ 3 which would likely improve image contrast and reduce the radiation burden to the patient.

Discussion
The ALISI radiosynthesizer was constructed by using a combination of open-source microcontrollers (Arduino) and computer-aided design (CAD), coupled with additive manufacturing 32,33 . Arduino electronic components are modular and expandable. This allows for rapid and facile integration of new features during prototype development. Liquid handling components and all fluidic pathways employ commercially available, single-use, sterile tubing and three-way switching valves operated by digital servomotors. These components were chosen to facilitate future translation of ALISI to a clinical environment. The custom-made, electropolished photoreactor houses three high-powered LEDs with peak emission at 365 nm (Fig. 2c). Light-induced activation of our ArN 3 reagents is rate-limiting in the photoradiosynthesis of 89 Zr-mAbs, and in comparison with manual reactions, the mirrored photoreactor of ALISI gave a 5-to-10-fold decrease in reaction time 26 .
It is important to note that the bioconjugation efficiency (as measured from the PCY or RCC values) shows a strong dependence on the reaction geometry and the conditions used. For instance, the nature of the photoactive compound and protein substrate, the choice of radionuclide, the concentrations of all reagents, the buffer composition, and the pH influence conjugation efficiency. Under identical conditions, RhodB-PEG 3 -ArN 3 gave consistently lower yields for HSA labeling than pre-radiolabeled 68 GaDFO-PEG 3 -ArN 3 , which in-turn, was lower than radiolabeling with 89 ZrDFO-PEG 3 -ArN 3 generated in situ. Under our optimized conditions, the observed bioconjugation yields using an initial 1:1 stoichiometric ratio between HSA and the photoactive compound were in the range of 35%-65%. Standard bioconjugation reactions used to make the DFO-mAbs, including activated ester 16 or thiourea 17 chemistry, have efficiencies in the range of ~ 20%-75% (see Supplementary Methods). These data confirm that photo-induced bioconjugation is equally successful compared to current state-of-the-art methods used to make radiolabeled mAbs in clinical practice. For the photoradiosynthesis of 89 ZrDFO-PEG 3azepin-trastuzumab from Herceptin, the RCC was ~ 45%. This is remarkable considering that the radiolabeling and bioconjugation steps are complete in < 90 s, and the reaction uses non-purified mAb where the mixture contains all formulation components of clinical-grade Herceptin. We note that radiochemists performing manual syntheses of 89 Zr-mAbs are familiar with obtaining near quantitative RCYs, but in most cases, the reaction only involves the radiolabeling step (not the conjugation). Since test reactions are usually employed to predetermine the maximum molar activity of a sample, quantitative labeling is expected. In a recent study, Poot et al. automated the radiolabeling and purification of 89 Zr-mAbs from pre-functionalized DFO-mAbs with RCYs in the range 60%-75% 3 . With the present photochemistry, formation of 89 ZrDFO-PEG 3 -ArN 3 in situ is quantitative and the decay-corrected isolated RCY of 89 ZrDFO-PEG 3 -azepin-trasutzumab (20.1 ± 2.4%) encompasses all chemical reactions, transfers, and processing steps. Transfer losses are one of the main limitations of adapting manual chemistry to automated platforms. Considering the multi-step nature of the ALISI protocol, which includes all reagent transfers, acid neutralization, buffer control, radiolabeling and bioconjugation, in-line purification, sterile filtration, and product formulation, the observed RCYs are an excellent benchmark.
Purification using SK-10 columns filled with Sephadex® G-100 media provided greatly enhanced separation and improved RCP (> 99%) of the isolated 89 ZrDFO-PEG 3 -azepin-trasutzumab when compared with PD-10 columns. This is expected since the Sephadex G-25 media used in PD-10 columns is primarily intended for desalting and is sub-optimal for separating large proteins from small-molecule components.
Finally, automated radiosynthesis of 89 Zr-mAbs on ALISI improved the reproducibility of the chemistry by standardizing the reaction geometry and conditions, and by decreasing the potential for user-related, irreproducible errors. Collectively, the successful synthesis and isolation of 89 ZrDFO-PEG 3 -azepin-trastuzumab using our automated radiosynthesizer suggest that ALISI is potentially useful for preparing individual patient doses or small batches of 89 Zr-mAbs (and other radiolabeled proteins) on demand. With further development, we anticipate that the system can be adapted for use with other radionuclides and photoactivatable chelates to access radiopharmaceuticals for applications in radioimmunotherapy.

Conclusion
We developed a radiosynthesizer unit that performs the fully automated light-induced synthesis of 89 Zr-mAbs for immunoPET. The ALISI system can produce 89 Zr-mAbs in < 25 min in high radiochemical purity (RCP > 99%) starting from stock solutions of 89 Zr-oxalate, a photoactivatable DFO-PEG 3 -ArN 3 chelate, and a protein of interest. Features of the automated protocol include neutralization of the oxalic acid, buffer exchange and pH correction, quantitative formation of the photoactivatable 89 ZrDFO-PEG 3 -ArN 3 complex, rapid (< 90 s) lightinduced protein-ligation, in-line purification using novel SK-10 size-exclusion chromatography columns, and finally sterile filtration and product formulation in a biocompatible medium. After loading the reagent reservoirs, all steps are performed at the touch of a single button. The system is also highly flexible, producing purified 89 Zr-mAbs via conventional labeling of a pre-functionalized DFO-mAb conjugate. With further development, we anticipate that ALISI could facilitate on-demand access to individual patient doses of 89 Zr-mAbs in Nuclear Medicine facilities that rely on automated devices for radiotracer production.

Methods
Further details are presented in the Supplemental Materials. Schematics and sample code for operating the microcontrollers are available on request. Additive manufacturing. The ALISI radiosynthesizer was constructed by using the computer assisted design (CAD) software Solidworks2020 (Dassault Systèmes, Vélizy-Villacoublay, France). Components were prepared through additive manufacturing, laser cutting or purchased directly from commercial vendors. Additive manufacturing components were produced by selective laser sintering and made from PA2200, a fine powder based on polyamide-12. The synthesizer case was manufactured from laser cut, 5 mm plates of high tensile strength, black polyoxymethylene (POM) and an aluminum profile modular assembly system (Kanya AG, Rüti, Switzerland).
Control and electronics. Electronic  Setup for automated radiolabeling. Reactions with ALISI were performed on new liquid handling equipment, freshly assembled from sterile packaging. Briefly, individual reservoirs were assigned to stock solutions of Na 2 CO 3 (aq.), DFO-PEG 3 -ArN 3 , 89 Zr-oxalate, protein, reaction buffer, and sterile PBS for purification by size-exclusion chromatography (SEC) (see Supplemental Tables 1-4). A minimum volume of ~ 50μL was necessary to achieve adequate liquid transfer. For test reactions where reagent volumes were below this threshold, the total volume was adjusted with an appropriate volume of water.
Description of the automated 89 Zr-photoradiolabeling procedure. A schematic of the plumbing diagram and a photograph of the ALISI system set-up for the synthesis of 89 Zr-mAbs is shown in Fig. 2. The radiosynthesizer unit is initialized by pushing the power button. After initialization, the system provides a digital prompt on the built-in LCD-display to indicate that the device is ready to start. Automated radiosynthesis and purification is initiated by pressing the start button. Thereafter, the radiosynthesizer transfers all reagents and components to the reaction vial, located inside the photoreactor, in the following sequence: i) 89 Zr-oxalate in ~ 1 M oxalic acid is transferred. ii) An equal volume of 1 M Na 2 CO 3 (aq.) is transferred. iii) DFO-PEG 3 -ArN 3 in a solution of ~ 10% DMSO and aqueous sodium borate buffer (0.25 M, pH8.0) is removed from the reservoir and first used to wash the 89 Zr-stock solution reservoir before being delivered to the reaction vial. iv) The protein solution is transferred. v) Additional sodium borate buffer (0.25 M, pH8.0) is used to wash the protein reservoir and then transferred. After the reagent transfer sequence, the reaction mixture is irradiated with 365 nm light for 90 s. Tests indicated that the temperature of the reaction mixture does not change during this time, but the high-powered LEDs require cooling with an aluminum heat-sink attached to a fan. After irradiation, the crude mixture is transferred to the SK-10 size-exclusion column (SEC) for purification. After automatic separation, the product fraction containing the high molecular weight protein is filtered through a standard 0.22 μm sterile filter and collected in a sterile vial. The product is formulated in sterile PBS (pH7.4), and after quality control, is ready for use in radiochemical, cellular, or in vivo assays. Data analysis. Data were plotted by using the GraphPad Prism 9.0 software (GraphPad Software Inc., San Diego, California USA).