CAR T_reg cells: prime suspects in therapeutic resistance

Abstract

Using comprehensive single-cell profiling, two studies reveal the molecular phenotypes of CAR T cells associated with durable response in patients with lymphoma, and highlight the role of CAR regulatory T cells in mediating resistance.

Chimeric antigen receptor (CAR) T cell therapy products that target CD19 have shown unprecedented efficacy in patients with relapsed or refractory B cell malignancies^1,2,3. However, more than half of treated patients eventually relapse, prompting investigation of the mechanisms of resistance to CAR T cell therapy. In this issue of Nature Medicine, two complementary studies by Haradhvala et al.⁴ and Good et al.⁵ used single-cell transcriptomic and single-cell proteomic profiling, respectively, to identify cellular phenotypes (in the infusion products or in post-infusion peripheral blood mononuclear cells) associated with responses and toxicity in patients with large B cell lymphoma treated with either axicabtagene ciloleucel (axi-cel) or tisagenlecleucel (tisa-cel). Other findings also subtly delineate the effect of manufacturing processes and design of CAR constructs and their association with clinical outcome.

Although both axi-cel and tisa-cel are derived from autologous T cells and target CD19 using the same antibody clone, they differ in several ways. Key differences include the design of the CAR molecule (CD28 transmembrane and CD28 costimulatory domains in axi-cel versus CD8α transmembrane and 4-1BB costimulatory domains in tisa-cel), the retrovirus (axi-cel) versus lentivirus (tisa-cel) used for CAR transduction, fresh (axi-cel) versus cryopreserved (tisa-cel) leukapheresis products as starting material, and other manufacturing conditions such as the T cell activation process and cytokines used in the culture medium. Moreover, the phenotype and function of the T cells in the leukapheresis products also varies markedly between patients, and is probably influenced by disease characteristics, previous therapies and host factors. Together, these differences give rise to heterogeneity across different CAR T cell products and in different patients treated with the same products — which have been identified as important determinants of response and toxicity to these therapies^6,7. The comprehensive profiling performed by Haradhvala et al.⁴ and Good et al.⁵ at the single-cell level provide additional insights into the distinct molecular phenotypes of CAR T cells associated with clinical outcome.

Using single-cell transcriptomic sequencing of infusion products and peripheral blood mononuclear cells before and after infusion, Haradhvala et al.⁴ showed that modest increases in the numbers of CAR regulatory T cells (T_reg cells) in infusion products of axi-cel were associated with relapse (Fig. 1). The CAR T_reg cells are inadvertently generated during manufacturing as there is no selection process to exclude them from a heterogeneous T cell population in the leukapheresis products. Using elegant in vitro and in vivo models, the authors showed that the CAR T_reg cells generated with either CAR construct suppressed the antitumor activity of conventional CAR T cells. Furthermore, they show in humans that differences in CAR design, in particular the costimulatory domain of the CAR between tisa-cel and axi-cel, is associated with differences in cellular dynamics and cell fate in the immediate post-infusion period. Although both types of CAR product showed evidence of activation and cellular proliferation after infusion, CD8⁺ cells in axi-cel showed a stronger upregulation of the activation marker PDCD1 and checkpoint regulator SLAMF6 at day 7, consistent with rapid activation associated with CD28 signaling. The authors also found that expansion and differentiation of infusion product-derived rare central memory-like CD8⁺ T cells into IL7R⁺ effector memory T cells were associated with durable responses to tisa-cel — whereas responses to axi-cel were associated with a heterogeneous cell population.

Fig. 1: CAR T_reg cells are associated with poor responses to anti-CD19 CAR T cell therapy in patients with lymphoma.

Higher numbers of CAR T_reg cells in CAR products and their subsequent proliferation after infusion suppresses the proliferation and function of conventional cytotoxic CAR T cells, leading to increased tumor growth.

Good et al.⁵ used single-cell proteomic profiling of peripheral blood mononuclear cells after infusion in patients with large B cell lymphoma treated with axi-cel. They showed that within populations of CD4⁺ and CD8⁺ CAR T cells, higher frequencies of CD57⁺ T-bet⁺ cells with high cytotoxic potential (on day 7 after infusion) were associated with durable responses, whereas higher numbers of CD4⁺ HELIOS⁺ CAR T cells were associated with disease progression and decreased neurotoxicity. Using deep phenotyping and single-cell sequencing studies, the authors confirmed that the CD4⁺ HELIOS⁺ cells were T_reg-like cells. Furthermore, they show that a model that combines the percentage of CAR T_reg cells with baseline levels of lactate dehydrogenase as a surrogate measure of tumor burden showed a stronger association with progression-free and overall survival compared with either feature alone.

T_reg cells are crucial regulators of immune responses and are involved in several physiological and pathological processes, including autoimmunity, chronic infections, allergy, organ transplantation and tumor immunity. The above findings from patient samples corroborate a previous study in a mouse xenograft model that showed that natural T_reg cells modified to express CD19 CAR trafficked to the sites of tumors and created a hostile microenvironment for CAR T effectors cells, limiting their functionality⁸. Also, as noted by Haradhvala et al.⁴, a trend toward higher numbers of T_reg cells was observed in non-responders in an external single-cell dataset of axi-cel infusion products⁶, providing further validation of the importance of CAR T_reg cells in affecting clinical efficacy.

Together, these findings have important implications for the development of CAR T cell therapies, by informing optimization of manufacturing processes and/or CAR design. For example, one could hypothesize that depleting T_reg cells in the leukapheresis products before CAR transduction or selecting specific T cell subsets for the generation of CAR T cells may enhance the efficacy of CAR T cell therapy. Haradhvala et al.⁴ also showed that there were significantly fewer T_reg cells in tisa-cel than in axi-cel infusion products. The authors speculate that the cryopreserved leukapheresis products used for tisa-cel generation may lead to lower numbers of T_reg cells compared with the fresh leukapheresis products used for axi-cel generation⁹. However, alterations in the manufacturing processes of CAR T cells need to be studied prospectively in well-designed clinical trials, as Good et al.⁵ found that CAR T_reg cells may also be associated with some beneficial effects by lowering the incidence of neurotoxicity. Whether depletion of T_reg cells may increase the risk of hemophagocytic lymphohistiocytosis (a severe inflammatory reaction) associated with CAR T cell therapy is also a potential concern^10,11.

The differences in the post-infusion temporal dynamics of expansion and differentiation — depending on CD28 versus 4-1BB costimulatory domain — revealed by Haradhvala et al.⁴ could have implications for CAR design for both autologous and allogeneic cell products. Because allogeneic CAR T cell products generally do not persist in the long term, a more rapid expansion and differentiation induced by the CD28 costimulatory domain may have potential advantages in terms of efficacy. By contrast, the lower peak expansion and longer persistence offered by a 4-1BB co-stimulatory domain may reduce the incidence of acute toxicity with autologous CAR T cells.

A recent study noted that circulating levels of CAR T cells were associated with changes in the tumor microenvironment after treatment — with increases in IL-15 (a key T cell growth factor), IFNγ-regulated genes and cytotoxic T cell activity in responders compared with non-responders¹². Future studies should therefore incorporate concurrent analysis of dynamic changes within the tumor microenvironment after infusion of CAR T cells. Ultimately, understanding how post-infusion CAR T cells dynamically differentiate and alter their phenotypes in patients will allow us to develop more effective and safer CAR T cell therapies. With this in mind, the comprehensive single-cell profiling used in the two studies by Haradhvala et al.⁴ and Good et al.⁵ serve as a model to evaluate and compare future CAR T cell products with new designs.