OBJECTIVES: 89Zr-labelled proteins are gaining importance in clinical research in a variety of diseases. To date, no clinical study has been reported that utilizes an automated approach for radiosynthesis of 89Zr-labelled radiopharmaceuticals. We aim to develop an automated method for the clinical production of 89Zr-labelled proteins and apply this method to Durvalumab, a monoclonal antibody targeting PD-L1 immune-checkpoint protein. PD-L1 expression is poorly understood and can be up-regulated over the course of chemo- and radiotherapy treatment. The ImmunoPET multicentre study aims to examine the dynamics of PD-L1 expression via 89Zr-Durvalumab PET imaging before, during, and after chemoradiotherapy. The developed automated technique will enable reproducible clinical production of [89Zr]Zr-DFOSq-Durvalumab for this study at three different sites. METHODS: Conjugation of Durvalumab to H3DFOSqOEt was optimized for optimal chelator-to-antibody ratio. Automated radiolabelling of H3DFOSq-Durvalumab with zirconium-89 was optimized on the disposable cassette based iPHASE technologies MultiSyn radiosynthesizer using a modified cassette. Activity losses were tracked using a dose calibrator and minimized by optimizing fluid transfers, reaction buffer, antibody formulation additives and pH. The biological profile of the radiolabelled antibody was confirmed in vivo in PD-L1+ (HCC827) and PD-L1- (A549) murine xenografts. Clinical process validation and quality control were performed at three separate study sites to satisfy clinical release criteria. RESULTS: H3DFOSq-Durvalumab with an average CAR of 3.02 was obtained. Radiolabelling kinetics in succinate (20 mM, pH 6) were significantly faster when compared to HEPES (0.5 M, pH 7.2) with >90 % conversion observed after 15 min. Residual radioactivity in the 89Zr isotope vial was reduced from 24 % to 0.44 % ± 0.18 % (n = 7) and losses in the reactor vial were reduced from 36 % ± 6 % (n = 4) to 0.82 % ± 0.75 % (n = 4) by including a surfactant in the reaction and formulation buffers. Overall process yield was 75 % ± 6 % (n = 5) and process time was 40 min. Typically, 165 MBq of [89Zr]Zr-DFOSq-Durvalumab with an apparent specific activity of 315 MBq/mg ± 34 MBq/mg (EOS) was obtained in a volume of 3.0 mL. At end-of-synthesis (EOS), radiochemical purity and protein integrity were always >99 % and >96 %, respectively, and dropped to 98 % and 65 % after incubation in human serum for 7 days at 37 °C. Immunoreactive fraction in HEK293/PD-L1 cells was 83.3 ± 9.0 (EOS). Preclinical in vivo data at 144 h p.i. showed excellent SUVmax in PD-L1+ tumour (8.32 ± 0.59) with a tumour-background ratio of 17.17 ± 3.96. [89Zr]Zr-DFOSq-Durvalumab passed all clinical release criteria at each study site and was deemed suitable for administration in a multicentre imaging trial. CONCLUSION: Fully automated production of [89Zr]Zr-DFOSq-Durvalumab for clinical use was achieved with minimal exposure to the operator. The cassette-based approach allows for consecutive productions on the same day and offers an alternative to currently used manual protocols. The method should be broadly applicable to other proteins and has the potential for clinical impact considering the growing number of clinical trials investigating 89Zr-labelled antibodies.