Culture expansion of CAR T cells evokes aberrant DNA methylation that is associated with adverse clinical outcome (RNA-Seq)

Cellular immunotherapy using autologous, genetically modified T cells has delivered remarkable initial response rates in various hematological malignancies, including acute lymphoblastic leukemia (ALL) or lymphoma. Upon treatment with these chimeric antigen receptor (CAR) T cells, approximately 30–40% of patients show long-term remission. There is a growing perception that not only CAR constructs, but also in vitro manufacturing processes may determine therapeutic efficacy. However, little is known about how duration of culture expansion affects molecular and functional features of CAR T cells. A better understanding of molecular signatures, reflecting therapeutic activity of CAR T cells, may help to further optimize the manufacturing process. In this study, we therefore investigated DNAm changes during culture expansion of T cells and CAR T cells during a period of up to three weeks. The results of our current study indicate that culture expansion of CAR T cells leads to DNAm changes that are continuously acquired during culture expansion, which is also reflected on gene expression level. Further analysis of culture-associated methylation profiles in CAR T cells from clinical trials support the notion that these chnages are disadvantageous with regards to therapeutic outcome. A shortened cultivation period to avoid dysfunctional methylation programs might be a promising manufacturing strategy to improve therapeutic efficacy of adoptive cell therapies. Our epigenetic signature for culture-associated DNAm may provide a biomarker for quality control of CAR T cell productions and to identify patients at risk for adverse events. Overall design: Total RNA was isolated from freshly isolated T cells (day 0) of four healthy donors, as well as from CD2/CD3/CD28-microbead activated T cells of the same donors upon three weeks of culture (day 22) in IL-2 supplemented medium, with the NucleoSpin RNA Plus kit (Macherey-Nagel). Library preparation with the QuantSeq 3' mRNA-Seq Library Prep Kit FWD (Lexogen, Vienna, Austria) and sequencing on the HiSeq 2500v4 platform (Illumina; 10 M 50 bp single-end reads) was performed at Life and Brain GmbH. RNA sequencing (RNA-seq) reads were aligned with the R package QuasR and Rhisat2 as aligner to the human genome (hg19) with default parameters. Reads that aligned with less than 10 reads per gene were excluded from further analysis and counts per million (CPM) were calculated and TMMwsp normalized with standard parameters using the R package edgeR. Count data was voom-transformed by applying the R package limma. For differential expression analysis the limma model was then fitted and t-statistics were computed using the Empirical Bayes method with the adjusted p-value threshold of 0.05.