Perfusion-powered production of over 100 CAR-T doses in a 2-litre stirred-tank bioreactor with automated product harvest, wash and concentration

Background & Aim

Allogeneic CAR-T therapies offer the potential to mass produce "off-the-shelf" doses, addressing the limitations of lengthy manufacturing processes in approved autologous products. To realise this potential, industrially scalable upstream and downstream workflows are required. To this end, we characterised a CAR-T cell perfusion manufacturing process in a 2-litre stirred-tank bioreactor (STR), with a solution to automate the bioreactor harvest, final product wash, and concentration steps.

Methodology

Healthy donor-derived anti-CD19 CAR-T cells were cultured in parallel in the Ambr® 250 High Throughput Perfusion and Univessel® 2-litre single-use STR, the latter coupled with the XCell ATF® 2 cell retention device. CAR-T cells were expanded in 4Cell® Nutri-T serum-free medium in perfusion for seven days, and performance was benchmarked against a G-Rex® 24 well plate process. The Ksep® 400 was evaluated as a closed, automated solution for the harvesting, concentration, and washing of the final CAR-T product from the 2L STR. Comparability between the 250mL and 2L systems was examined, alongside the impact of automated harvesting on cell yield, viability, activation, maturation, exhaustion, cytotoxicity, and cytokine release.

Results

CAR-T cells were successfully expanded ∼125-fold in perfusion in both 250mL and 2L STRs. At the 2L scale, final yields of ∼50 billion total viable cells were consistently generated, with a CAR transduction efficiency of 34%, corresponding to over 100 representative doses of 150 × 10⁶ CAR+ T cells. Harvested cells were predominantly in the naive and central memory subset (>95%) and expressed negligible levels of exhaustion markers (<1%). No statistical differences were observed in CAR-T cell yield, viability, surface marker expression, or cytotoxicity between the 250mL and 2L bioreactors. Automated harvesting recovered >90% of the total yields, concentrated cells 8-fold, and had no impact on CAR-T cell phenotype, cytotoxicity, or growth kinetics post-harvest.

Conclusion

These findings demonstrate the feasibility of multi-litre expansion of functional CAR-T cells to generate hundreds of doses in perfusion-based stirred-tank bioreactors. The Ambr® 250 effectively served as a scaled-down model and facilitated seamless scale-up to 2L. Finally, automated harvesting and concentration streamlined the downstream workflow while maintaining CAR-T cell quality. This scalable upstream and downstream workflow is ideally suited for mass production of future allogeneic therapies.