All of the clinical trials that led to initial FDA CAR T-cell approval used bulk T-cells derived from the peripheral blood of patients, without intentionally excluding or enriching a particular subset of T-cells for CAR modification or re-infusion. Subsequently, a CAR T-cell product has been approved by the FDA that consists of a defined 50:50 mixture of CD4:CD8 T-cells (lisocabtagene maraleucel). T-cells exist in a number of phenotypes, which depends on such things as whether they are in the throes of an antigen-specific response, have previously been strongly activated but have had some time to rest, or have never been activated at all. All T-cells start in a naïve state, and upon antigen encounter acquire an effector (TE) or a memory phenotype (TM), which can further be divided into central memory (TCM) and effector memory (TEM) compartments. An additional group of stem cell-like memory T cells (TSCM) have been identified as a long-lived human memory T cell population with an enhanced capacity for self-renewal. In addition to these phenotypes, T-cells are either CD8 or CD4 positive. CD8+ T-cells are traditionally thought of as the cytotoxic T-cells which are responsible for killing, while CD4+ T-cells perform a helper function, by secreting cytokines to improve CD8+ T-cell proliferation and function. However, after modification with a CAR, CD4 T-cells can effectively kill similarly to CD8+ T-cells. Amidst all of these complexities, it is not inherently obvious whether particular subsets, or combinations of subsets, may provide greater efficacy against tumor.
To begin to answer this question, mouse studies were performed in which purified individual CD8+ and CD4+ T-cell subsets were CAR modified and then injected into tumor-bearing mice. First, pure CD4+ CAR T-cells generated from naïve or TCM precursors significantly improved survival compared with TEM CD4+ CAR T cells. Among the purified CD8+ CAR T-cell subsets, TCM cells were slightly better than naïve, which were slightly better than TEM. Additionally, combining CD8+ TCM CAR T-cells with CD4+ naïve or CD4+ TCM CAR T-cells significantly improved survival relative to pure CD8+ TCM cells alone.
These findings inspired clinical trials utilizing defined T-cell subsets. A phase I-II study of CAR T-cells targeting CD19 using a 1:1 ratio of CD4:CD8 was per formed in ALL patients, in which the CD8 T-cells were enriched for a TCM phenotype. Patients achieved a 93% remission rate by flow cytometry and 86% MRD-negative CR rate, which compared favorably with other trials using bulk T-cells. CD19 CAR-T cells have also been administered in a 1:1 CD4+:CD8+ ratio, in which the CD8 T-cells were enriched for a TCM phenotype, to relapsed and/ or refractory B cell non-Hodgkin lymphoma patients. The CR in patients treated with Cy/Flu at the maximally tolerated dose was 64%, with a low incidence of serious toxicity.
Since patients exhibit a wide range of dominant T-cell phenotypes at baseline, which is reflective of their antigen-exposure history, treatment history, and numerous other factors, informative correlations have been drawn from those who respond well to CAR T-cells versus those who do not. In single cell RNAseq analysis of CD19 CAR T-cells isolated from the infusion bag of large cell lymphoma patients, the CD8 T-cell phenotype most associated with achieving a CR was a TCM phenotype. The frequency of these cells were also low in patients with high-stage disease (III and IV) and high international prognostic index (IPI 3–4), suggesting clinical factors prior to leukapheresis may influence the transcriptional state of CAR T cells. Similarly, transcriptomic profiling revealed that CAR T cells from complete- responding patients with CLL were enriched in memory-related genes prior to infusion.
It is still unclear whether enriching for a particular phenotype prior to infusion, or whether attempting to induce a favorable phenotype during the CAR modification process such as through culture with IL-7 and IL-15, will lead to improved result in patients. However, the preclinical and phase I human data suggest these approaches may confer advantages, especially for more difficult to treat tumors, and warrant further investigation.
