Prediction of Therapeutic Potential of CD19 CAR-T Cells for LBCL By Histone Mark Analyses of Core Epigenetic Programming

Chimeric antigen receptor-T cell (CAR-T) therapy has produced durable responses in a subset of patients with otherwise incurable B-lineage malignancies. Efforts are underway to identify genetic and epigenetic programs in CAR-T cells that influence product quality, thereby enabling development of next generation CAR-T products. The subset-of-origin (SOO) for CAR-T cell manufacturing and the effects of serial stimulation are key factors that might influence CAR-T quality. We hypothesized that serial stimulation might impart different effects on the epigenetic programming of CAR-T cells manufactured from distinct subsets; and that these epigenetic programs might associate with outcomes of clinical CAR-T therapy. We previously showed that sequencing of DNA bound to transcriptionally permissive (H3K4me2) and repressive (H3K27me3) histone methylation marks identifies numerous important differences in CAR-T cells from healthy donors (HD) and patients that were not apparent by RNA-seq. Here, we applied novel histone mark analyses to identify epigenetic programs that impact outcomes of large B cell lymphoma (LBCL) patients receiving CAR-T cell therapy. We first validated a novel in vitro model generating a roadmap of epigenetic and transcriptional programs during longitudinal serial stimulation of CAR T-cells manufactured from distinct SOO. H3K27me3 and H3K4me2 analyses and RNA-seq were performed on ex vivo isolated CD8+ Naïve (N), Central Memory (CM) and Effector Memory (EM) subsets from HD, on each SOO-derived CAR-T cell product after manufacturing, and prior to cycles of serial stimulation with a CD19+ lymphoma cell line. We first derived principal component analysis (PCA) plots from the 3 datasets (H3K4me2, H3K27me3, RNA-seq), each comprising 15 samples (3 SOO, 5 timepoints) from 7 HD. In PCA plots of histone mark data, we identified two epigenetic programmatic axes. One axis distinguished the SOO of CAR-T manufacturing. A second axis recapitulated the stimulation history. Importantly, clustering of samples showed histone mark analyses clearly distinguished not only each ex vivo isolated CD8+ T cell subset and CAR-T cell products manufactured from each of those subsets, but also identified each of the cycles of stimulation of each subset-derived CAR-T cell product. In contrast, when PCA plots were generated from RNA-seq data, SOO and stimulation history was indistinguishable after first stimulation. Thus, histone mark analyses enabled dissection of epigenetic programs to identify the SOO and stimulation history of CAR-T cells. We considered that treatment of LBCL patients with CAR-T infusion products that harbor histone marks similar to multiply stimulated HD CAR-T cells might be less effective than treatment with their counterparts that resemble HD CAR-T cells with limited stimulation history. We identified histone mark signatures corresponding to each cluster in the HD reference matrix from our model. We then demonstrated that an epigenetic imprint of stimulation history harbored by infused CD8+ CAR-T cells was associated with failure of LBCL patients treated in a phase 1/2 clinical trial (NCT01865617) to achieve complete response (P=0.024). We further interrogated histone mark data to identify genetic programs associated with robust in vivo CAR-T cell proliferation. Histone mark analyses demonstrated that marks associated with less repression of the transcription factor, KLF7, were associated with higher peak in vivo CAR-T counts (H3K27me3 log-scaled count: 7.846 in high-expanders vs 14.6 in low-expanders, P=0.049). Transduction of KLF7 into T cells and CAR-T cells enhanced proliferation (P=0.014 and 0.012 for T cells and CAR-T cells, respectively) and IL-2 production (P=0.004), and increased the fraction of CAR-T cells expressing a less-differentiated CM immunophenotype (P=0.043). Our data show that epigenetic programs identified by histone mark analyses distinguishes both the SOO for CAR-T cell manufacturing and the number of cycles of serial CAR-T stimulation in multiply stimulated CAR-T cell products. These epigenetic programs associated with efficacy of CAR-T cells for LBCL and enabled identification of a key transcription factor that governs in vivo CAR-T proliferation. Imprinted signatures that govern product competence will be presented, along with implications for the development of the next generation of highly effective CAR-T cell products.