Chimeric antigen receptor (CAR) T cell therapy has led to impressive remission rates in hematological malignancies. Current protocols have been designed to generate large numbers of cells that can be reinfused into the patient. T cell expansion, however, comes at the cost of differentiation and the associated loss of stemness, which are desirable properties with regards to functional persistence of CAR T cells in cancer patients. Multiple preclinical as well as clinical studies suggest that less-differentiated precursor T cell subset are the ideal population for cancer immunotherapies due to greater proliferation, enhanced survival, and improved functionality after infusion into patients¹. However, little is known with regards to cellular and molecular mechanisms driving CAR T cell (dys-)function and persistence in patients. Manufacturing of cell products by means of in vitro expansion has been generally associated with a continuous functional decline that is reflected in DNA methylation (DNAm) changes². With this in mind, we wanted to map DNAm changes in CAR T cells that are acquired during the in vitro manufacturing process and potentially correlate these with patient outcome.

First, we investigated DNAm changes during culture expansion of T cells and CAR T cells in culture conditions commonly used for manufacturing. We were able to show that several CpGs become increasingly hypermethylated, particularly within genes that are related to T cell function and that have previously been associated with T cell dysfunction in vivo (Fig. 1)³.

Next, we aimed to generate a linear predictor for time in culture that would also associate with patient survival post CAR T cell infusion. For this purpose, predictor training and validation datasets were compiled to pre-select only CpGs with very high linear correlation between DNAm and time in culture. This resulted in 336 hypermethylated and only 3 hypomethylated CpGs (Pearson correlation r  >  0.9 or r  <  -0.9, respectively) (Fig. 2).

Then, an additional dataset, that included metadata on patient survival from three clinical trials⁴, was used for multivariate Cox regression analysis. This dataset was also randomly divided into a target identification cohort of 82 patients and a target validation cohort of 32 patients. In the target identification cohort, 14 CpG sites (out of the 336 culture-associated CpGs) were associated with a higher patient death rate (HR > 1, p < 0.01) (Fig. 3a). To derive an epigenetic culture-time predictor that is indicative for loss of therapeutic potential, we went back to the initial predictor training dataset and generated a model based these 14 CpGs. This model showed again high correlation between predicted and real time in culture (r2 = 0.94 in training and r2 = 0.70 in validation dataset). When we tested the association between the culture-time-predictions from the 14 CpG model with overall survival in the clinical dataset, we found that longer culture time predictions were associated with increased patient death rates (HR = 1.37 with p < 0.0001 in identification cohort and HR = 1.34 with p <0.01 in validation cohort) (Fig. 3b and c).

Taken together, we found that DNAm changes are continuously acquired during culture expansion in CAR T cells which negatively relate to patient outcome post infusion. Shortened cultivation period to avoid dysfunctional methylation programs might be a promising future direction to improve CAR T cell therapy. Our findings are in line with the growing perception that a cell’s phenotype is more relevant than absolute cell numbers that are infused back into the patient to generate a product with sustained antitumor abilities⁵. Currently, bulk T cells are isolated from patient’s peripheral blood and then activated and expanded in vitro for at least a week. Even though these protocols can generate large amounts of cells, their phenotype is dominated by terminally differentiated effector cells rather than multipotent and long-lived progenitors⁶. Thus, multicenter phase I/II studies have recently been initiated to evaluate the safety and preliminary efficacy of CAR T cells that have been manufactured in less than two days, in patients with diffuse large B-cell lymphoma (NCT03960840) or multiple myeloma (NCT04318327). Thus, it will be interesting to analyze if shortened manufacturing will improve the therapeutic efficiency.

In the future, we aim to further validate our epigenetic signature for loss of T cell potential and develop a biomarker that can identify patients at risk for adverse events.

Link to the paper: https://rdcu.be/dgPeT

References

1. Biasco, L. et al. Clonal expansion of T memory stem cells determines early anti-leukemic responses and long-term CAR T cell persistence in patients. Nature Cancer 2, 629-642 (2021).
2. Zebley, C.C. et al. CD19-CAR T cells undergo exhaustion DNA methylation programming in patients with acute lymphoblastic leukemia. Cell Rep 37, 110079 (2021).
3. Prinzing, B. et al. Deleting DNMT3A in CAR T cells prevents exhaustion and enhances antitumor activity. Sci Transl Med 13, eabh0272 (2021).
4. Garcia-Prieto, C.A. et al. Epigenetic Profiling and Response to CD19 Chimeric Antigen Receptor T-Cell Therapy in B-Cell Malignancies. J Natl Cancer Inst 114, 436-445 (2022).
5. Zebley, C.C., Gottschalk, S. & Youngblood, B. Rewriting History: Epigenetic Reprogramming of CD8(+) T Cell Differentiation to Enhance Immunotherapy. Trends Immunol 41, 665-675 (2020).
6. Ghassemi, S. et al. Rapid manufacturing of non-activated potent CAR T cells. Nat Biomed Eng 6, 118-128 (2022).