Establishing a scalable perfusion strategy for the manufacture of CAR-T cells in stirred-tank bioreactors using a quality-by-design approach

Abstract

Chimeric antigen receptor T cell (CAR-T) therapies show high remission rates for relapsed and refractory leukemia and lymphoma. However, manufacturing challenges hinder their commercial viability and patient accessibility. This study applied quality-by-design principles to identify perfusion critical process parameters for CAR-T expansion in stirred tank bioreactors to maximize yields. A design of experiments in the Ambr® 250 High Throughput Perfusion small-scale bioreactor revealed that earlier perfusion starts (48 h vs. 96 h post-inoculation) and higher perfusion rates (1.0 VVD vs. 0.25 VVD) significantly increased cytotoxic CAR-T cell yields without compromising critical quality attributes. Optimizing perfusion improved growth kinetics and yields across donor samples, achieving densities >21 × 10⁶ cells/mL in 7 days, outperforming traditional fed-batch and static flask cultures. This study underscores the importance of optimizing perfusion parameters to maximize CAR-T yields and quality and highlights the utility of scale-down models in reducing time, costs and risks associated with process development.

Keywords: CAR‐T; perfusion, immunotherapy; process control; process intensification; quality‐by‐design; stirred‐tank bioreactor.

FIGURE 1

Impact of perfusion start time and perfusion rate on cell growth and viability in the Ambr® 250 DOE. (a) Viable cell density by day. (b) Cell fold by day. (c) DOE contour plots representing the effects of perfusion parameters on cell fold per donor. (d) Impact of perfusion start time on average cell viability by day and (e) on day 3 post‐inoculation. Data represents n = 15 perfusion DOE conditions. Error bars represent standard deviation; *p < 0.05. VVD, vessel volumes per day.

FIGURE 2

Culture trends including pH, DO, and metabolite concentrations impacted by DOE perfusion parameters. (a) Dissolved oxygen (DO) and (b) pH over time in the Ambr® 250 perfusion cultures. Black dashed lines represent parameter setpoints. (c) Impact of perfusion parameters on daily glucose and lactate, and (d) daily glutamine and ammonia concentrations (mmol/L). Arrows represent onset of perfusion. Data represents n = 15 perfusion DOE conditions. VVD, vessel volumes per day.

FIGURE 3

Cell quality markers impacted by donor in the perfusion DOE. (a) CD8+ T cell differentiation marker expression at the end of the 7‐day bioreactor perfusion cultures was broken into four populations: %CD8+CD45RO‐CCR7+ (naïve); %CD8+CD45RO+CCR7+ (central memory (CM)); %CD8+CD45RO+CCR7‐ (effector memory (EM)); and %CD8+CD45RO‐CCR7‐ (effector). (b) PD1+ and LAG3+ exhaustion marker expression at the end of the 7‐day bioreactor on CD3+ T cells. (c) CAR expression percentage in harvested CD3+ T cells. (d) Changes in CD4:CD8 ratio from day 0 – 7. Data represents n=15 perfusion DOE conditions. Error bars represent standard deviation; *p < 0.05, ***p < 0.001. ****p < 0.0001 VVD=vessel volumes per day.

FIGURE 4

Optimizing perfusion significantly improved cell fold in the Ambr® 250 High Throughput. (a) Viable final cell density by day and (b) by day 7 post‐inoculation. (c) Daily glucose and lactate, and (d) daily glutamine and ammonia concentrations (mmol/L). Media additions were performed on days 3, 4, 5 in the fed‐batch bioreactor, perfusion was started 48 h post‐inoculation at 1 volume vessels per day (VVD) and cell passaging was completed on days 2, 4, 6 in flasks. (e) CD8+ T cell differentiation marker expression at the end of the 7‐day cultures was broken into four populations: %CD8 + CD45RO‐CCR7+ (naïve); %CD8 + CD45RO + CCR7+ (central memory (CM)); %CD8 + CD45RO + CCR7− (effector memory (EM)); and %CD8 + CD45RO‐CCR7‐ (effector). (f) PD1+ and LAG3+ exhaustion marker expression on CD3+ T cells by day 7. (g) Changes in CD4:CD8 ratio from day 0 to 7. (h) CAR expression percentage in harvested CD3+ T cells. Data shown as the mean of n = 3 replicates. Error bars represent standard deviation; *p < 0.05, **p < 0.01, ***p < 0.001.

FIGURE 5

CAR‐T cells remain cytotoxic following perfusion bioreactor expansion. Following expansion in flasks and bioreactor, CAR‐T cells were co‐cultured 1:1 with target Nuclight Green+ NALM6 cells for 2 days. (a) Relative number of NALM6 cells over time and (b) final relative number of NALM6 cells after 2 days, (c) Representative 20X Incucyte® images of CAR‐T cells and controls after 2 days, (d) IFN‐γ and (e) TNF‐α concentration in the medium after 2 days. Data shown as the mean of n = 4 replicates. Error bars represent standard deviation; ***p < 0.001. NT, non‐transduced.