In Vivo Persistence and Function of CAR19 T-Cell Products for Pediatric B-Cell Acute Lymphoblastic Leukemia Are Impacted By Media, Cytokine Supplementation and Metabolite Bioavailability during Ex Vivo Manufacturing

Chimeric antigen receptor (CAR) T-cell therapy targeting CD19 has been highly effective in treating pediatric B-cell acute lymphoblastic leukemia (B-ALL), reporting complete responses one month after infusion in close to 90% of patients. However, these responses are not durable in all patients, and up to 50% will relapse with CD19+ or CD19- disease within two years of infusion. Relapse of CD19+ disease is likely due to loss of functional CAR T-cells. There is interest in developing strategies to enhance persistence of CAR T-cells, but the exact mechanisms that drive persistence in a patient are poorly understood. The initial quality of T-cells isolated, the design of the CAR, and conditions during ex vivo manufacturing can all impact T-cell differentiation and exhaustion. As T-cells lack endogenous nutrient stores, they are dependent on their local milieu to fuel their proliferation, cytotoxicity and persistence. We previously demonstrated how the unique metabolic fate of CD19/41BBz CAR T-cells supports their longevity following adoptive transfer. Still, how T-cells respond to the metabolic environment during manufacturing and what impact this has on long-term functional persistence in patients is poorly understood. In two recent clinical trials performed at the Children’s Hospital of Philadelphia evaluating a murine (NCT04276870) and humanized (NCT03792633) CD19/41BBz CAR for pediatric patients with B-ALL, we observed that two modifications made to the manufacturing process appeared to have impacted long-term functional persistence of CAR T-cells: base media formulation and cytokine supplementation. This led to four manufacturing conditions across these trials: X-VIVO IL-2, X-VIVO IL-7/IL-15, OpTmizer IL-2, and OpTmizer IL-7/IL-15. Initial clinical response and expansion of CAR T-cells appear similar across manufacturing conditions when compared to historical X-VIVO IL-2 manufacturing conditions used in prior murine (NCT02906371) and humanized (NCT02374333) pediatric CD19/41BBz CAR T-cell studies; however, cells produced in X-VIVO IL-2 demonstrate a longer duration of functional persistence in patients as defined by B-cell aplasia. Additionally, there were significantly increased numbers of CAR T-cells in the X-VIVO IL-2 group, measured by qPCR, in peripheral blood of patients 3-months post-infusion compared to other manufacturing groups. Switching manufacturing back to original X-VIVO IL-2 condition rescues the loss of CAR T-cells at month 3 to pre-manufacturing change levels. Bulk and single-cell RNA-sequencing reveals that media formulation, but not cytokine stimulation, drives major transcriptional differences in patient CAR T-cells both pre-infusion and at peak expansion post-infusion, including differences in master transcriptional regulators. T-cells grown in X-VIVO demonstrate an increased effector/effector memory signature, increased cell volume at end of manufacturing harvest, upregulation of cholesterol biosynthesis pathway genes, differential granzyme profile, and increased CD8+ T-cells. Healthy donor CAR T-cells grown in X-VIVO demonstrate superior expansion and tumor control in a xenograft model of leukemia compared to donor-matched OpTmizer-grown cells, culminating in increased survival. In a targeted metabolomics screen, we identified several differences in metabolite abundance across formulations. We note that X-VIVO was enriched with metabolites which lie at the intersection of TCA cycle activity, one carbon metabolism, as well as arginine, ornithine, and inosine metabolism, several of which have been implicated in T-cell stemness and mitochondrial function. Based on this data, we propose that metabolite bioavailability during manufacturing of CD19/41BBz CAR T-cells has a large impact on the phenotype, transcriptional signature, and overall in vivo functional competence of infused cells, which can determine long-term durability of pediatric B-ALL patient responses.