Signatures of dysfunctional CAR T cells

In this issue of Blood, Selli et al¹ have discovered that whereas CD28-bearing chimeric antigen receptor (CAR) T cells become dysfunctional and exhausted via the classical events occurring with chronic stimulation from tumors or chronic infections, 4-1BB–bearing CAR T cells become dysfunctional in a different way that is driven by the transcription factor FOXO3.

A little over a decade ago, CAR T cells were shown to have remarkable efficacy in B-cell malignancies.^2-4 All of them targeted the B-cell molecule CD19 and bore the intracellular signaling domain of 1 of 2 costimulatory molecules, either CD28 or 4-1BB. The incorporation of a costimulatory domain into the CAR was hypothesized to improve persistence and function and, therefore, antitumor effects. Owing to the remarkable initial success of these initial products, all 6 constructs that are approved by the US Food and Drug Administration, including those targeting the B-cell maturation antigen expressed by myeloma plasma cells, and most other CARs in clinical development across a variety of indications, incorporate 1 of these 2 signaling domains.

Because CARs recognize antigen with high affinity and are expressed by constitutive promoters, investigators noted early on that continuous stimulation with antigen, and sometimes even just high-level CAR expression, could drive T-cell dysfunction. These dysfunctional T cells had reduced cytotoxicity, cytokine production, and proliferation after antigen stimulation; this could be overcome by intermitting rest⁵ from signal transduction. In patients, high tumor burden has been correlated with reduced response and increased T-cell dysfunction.⁶ Until now, this dysfunction was thought to be essentially the same phenomenon as T-cell “exhaustion,” which has been extensively described in the setting of chronic viral infection and cancers treated with checkpoint blockade.⁷ 

Here, Selli et al modeled chronic activation by repetitive in vitro stimulation of CD19-directed CAR T cells bearing either CD28 or 4-1BB intracellular signaling domains to generate dysfunction, and then performed a comprehensive analysis of the transcriptional, epigenetic, and phenotypic programs. The CAR T cells were stimulated with new tumor cells at a low effector-to-target ratio every 48 hours. After 2 weeks, the CAR T cells lost their ability to kill antigen-positive targets, could no longer make cytokines, and could no longer proliferate. Next, they interrogated these dysfunctional CAR T cells phenotypically, using multiparameter cytometry by time-of-flight. They found that CAR T cells bearing CD28-based CARs bore the hallmarks of classic T-cell exhaustion, including resurgent expression of PD-1, TIGIT, LAG3, TIM3, and CTLA4 (see figure). In contrast, dysfunctional T cells bearing 4-1BB CARs occupied different t-distributed stochastic neighbor embedding space by nearest neighbor analysis and expressed higher CD62L and CD25 on their surface. Interrogation of the transcriptional programs of these dysfunctional CAR T cells using RNA sequencing revealed that CD28-based CAR T cells were highly enriched for exhaustion-associated genes, whereas 4-1BB–based dysfunctional CAR T cells did not have the classic exhaustion signatures. Recent studies of T-cell exhaustion have emphasized that the state is defined by specific epigenetic alterations.⁸ The investigators also confirmed these differences using assay for transposase-accessible chromatin with sequencing (ATAC-seq), particularly of the chromatin accessibility of the gene encoding PD1 (PDCD1), which looked like classic exhaustion in the CD28-based CAR T cells but not the 4-1BB–based CAR T cells. The authors then used longitudinal single-cell RNA sequencing and identified that most of the dysfunctional 4-1BB CAR T cells were defined by a distinct set of genes involved in cytotoxicity (GNLY, CCL5, PRF1, GZMA, GZMK, CTSW), natural killer cell identity (KLRK1, KLRC2), and T-cell differentiation (ID2), which they termed the T_BBD signature (see figure). Intriguingly, their T_BBD signature was confirmed in a single patient who had received tisagenlecleucel (an anti-CD19/4-1BB CAR construct) and whose lymphoma had progressed despite persisting CAR T cells in their blood.

To understand the origin of this novel molecular program of dysfunction, they interrogated transcription factor binding accessibility over time. As expected, CD28-CAR dysfunction had increased accessibility for Jun:Fos (AP1) binding, but in contrast, 4-1BB-CAR dysfunction opened homeobox (HOX) and forkhead box (FOX) sites, and the small conditional RNA-sequencing data revealed high activity of FOXO3. Furthermore, in mechanistic studies, disruption of FOXO3 by CRISPR/Cas resulted in resistance to dysfunction after repetitive antigen stimulation, and, conversely, FOXO3 overexpression dramatically reduced the expansion of 4-1BB (but not CD28) CAR T cells in the setting of repetitive antigen stimulation. In murine xenograft stress models, in which very low numbers of CAR T cells are injected to treat higher tumor burdens, FOXO3 knockout CAR T cells bearing a 4-1BB intracellular signaling domain improved survival.

Remarkably, the authors discovered a new molecular mechanism by deep technological interrogation, spanning from high-dimensional cytometry to longitudinal single-cell RNA sequencing, and while still using a relatively simple in vitro model (repetitive antigen stimulation). The molecular program itself, and reactivation of FOXO3, is interesting and perhaps not typically found in nature. In natural biology, 4-1BB stimulation is temporally dissociated from T-cell receptor (TCR) stimulation, and TCR expression itself cycles with antigen exposure, thus reducing the possibility of tonic signaling. Future studies could include further validation of the T_BBD transcriptional signature and FOXO3 activity in more patients who received tisagenlecleucel and did not respond. Likewise, a deeper understanding of whether tisagenlecleucel gets exhausted in patients based on chronic exposure to nascent CD19⁺ B cells, regardless of disease status, should be explored. Interestingly, chronic exposure to nascent B cells was thought to have a role in the decade-long persistence of tisagenlecleucel in the first 2 patients treated, both of whom had prolonged remissions.⁹ In addition, it will be interesting for the field to see whether 4-1BB–bearing CAR T cells that are specific to other antigens, such as both of the B-cell maturation antigen (BCMA)–targeted CAR T cells that are approved for multiple myeloma, also exhibit the T_BBD signature and FOXO3 reactivation when patients with BCMA-positive disease recur despite persistent BCMA-directed CAR T cells. Finally, at the mechanistic level, it will be interesting to see whether FOXO3 knockout or c-Jun overexpression¹⁰ is more potent in reducing CAR T cell exhaustion.

Conflict-of-interest disclosure: M.V.M. is an inventor on patents related to adoptive cell therapies, held by Massachusetts General Hospital (some licensed to Promab) and University of Pennsylvania (some licensed to Novartis); receives grant/research support from Kite Pharma; has served as a consultant for multiple companies involved in cell therapies; holds equity in 2SeventyBio, Century Therapeutics, Neximmune, Oncternal, and TCR2; and serves on the board of directors of 2Seventy Bio.