CAR T cells converge on a distinct transcriptional state over time in patients with diverse hematologic malignancies Free

Chimeric antigen receptor (CAR) T cells directed to cancer cells act as a living drug in patients and have the potential to induce long-term remissions following a single treatment. They have transformed the treatment of hematologic malignancies, leading to FDA approval in relapsed or refractory acute lymphoblastic leukemia (ALL), chronic lymphocytic leukemia (CLL), large B-cell lymphoma (LBCL), and multiple myeloma. Infusion products are typically created from unselected or CD4/CD8-selected T cells from patients and constitute a heterogeneous population. Current observations about post-infusion CAR T cell dynamics and phenotypic evolution appear to be patient-, disease-, or product- specific. It is unknown whether there are shared properties among CAR T cells with prolonged function in patients.

Here we identified 10 patients (7 with LBCL, 3 with CLL) who were treated with different CD19-directed CAR T cell therapies on clinical trials and remained in complete remissions (CR >19-55 months). The CAR T cells used in patients with LBCL have a modified 1XX CD3ζ signaling module and a CD28 costimulatory domain, while the CAR T cells used in patients with CLL have a humanized anti-CD19 scFv and a 4-1BB costimulatory domain. Of note, CAR T cells with CD28 and 4-1BB co-stimulation differ significantly in their activation kinetics and clinical courses based on experiences with commercial products. We sorted rare (<1%) long-lived CAR T cells from the peripheral blood of patients at 10-40 months post-infusion by flow cytometry and performed multimodal single-cell sequencing to simultaneously capture transcriptomes, surface proteins, and T-cell receptors (TCRs). These rare CAR T cells were likely functional, since the patients had ongoing clinical responses and B cell aplasia at the time of the sample collections.

We found that long-lived CAR T cells in patients with LBCL and CLL clustered independently from endogenous immune cells and showed remarkably similar transcriptional states despite different cell engineering designs and underlying diseases. The transcriptional states are marked by differential expression of genes including GZMK, GZMA, CD27, FXYD2, HMOX1, GYPC and enrichment of gene sets associated with interferon-γ response, oxidative phosphorylation, and cell cycle regulation. Furthermore, we compared our findings to published transcriptional signatures of persisting CAR T cells in children with ALL treated with the recently approved obe-cel design and showed that this signature is highly reproducible across multiple contexts.

Next, we inferred the gene regulatory networks that govern the transcriptional states of long-lived CAR T cells using our single-cell RNA sequencing data. We identified specific overexpression of the MYB transcription factor and high activity of the MYB regulon in long-lived CAR T cells in comparison to various endogenous T cell subsets in our LBCL and CLL datasets. While MYB was previously shown to downregulate the effector function and preserve the long-term self-renewal capacity of T cells in a mouse model of chronic viral infection, its role in CAR T cell biology is unknown and warrants further investigation.

Notably, this distinct transcriptional state resembles that of naturally-occurring CD4/CD8 double-negative (DN) T cells with αβ TCRs found in healthy individuals based on published single-cell atlases of human blood. Consistent with this finding, the long-lived CAR T cells in patients with CLL demonstrate absence of CD4 and CD8 surface proteins. The long-lived CAR T cells in patients with LBCL are CD8 negative but show variable expression of CD4. To trace the cell lineage of long-lived CAR T cells in patients with CLL, we compared their endogenous TCRs to those of CAR T cells at peak expansion 10 days post-infusion and found no overlap, suggesting that the clones that eradicate disease in the initial phase are not the same as the clones that perform long-term immunosurveillance.

In summary, we showed that long-lived CAR T cells with various engineering designs in patients with CLL, LBCL, and ALL converge on a distinct transcriptional state that mimics that of peripheral DN T cells with αβ TCRs in healthy individuals. The MYB transcription factor and its gene regulatory network appear to be specific to this cell state. Lineage tracing using TCRs suggests that CAR T cells at peak initial expansion in patients are not clonally related to the CAR T cells that persist for months to years.