NK-like Skewing As the Key Driver of T Cell Dysfunction in AML: Implications for Enhancing TCR-T Cell Therapy Efficacy

Understanding acute myeloid leukemia (AML)-specific mechanisms of T cell dysfunction is key to improving T cell therapies.

In solid tumors, reduced T cell persistence commonly links to the expression of exhaustion markers (PD1, CTLA4, TIM3, LAG3, TIGIT) targetable with checkpoint blockade. In AML, the presence of T cell exhaustion has been contested (Penter et al., Blood 2023), highlighting alternative T cell dysfunction mechanisms based on terminal differentiation of CD8⁺ T cells expressing NK-like (NKL) markers (Abbas et al., Nat Commun. 2021; Mazziotta et al., Blood 2024). Whether AML induces this skewing of antigen-specific T cells remains an open question.

Wilms’ Tumor 1 (WT1) protein is overexpressed and correlates with poor prognosis in high-risk AML. In a phase I/II clinical trial, we employed T cells engineered to express a high-affinity WT1-specific T cell receptor (TCR_C4) to treat 15 HLA-A2⁺ patients with relapsed/refractory AML after allogeneic transplant. To mitigate graft-versus-host disease (GVHD) from the endogenous TCR, we used EBV or CMV-specific CD8⁺ T cells obtained from HLA-matched donors. This framework enabled us to longitudinally monitor post-transfer AML-specific T cells via high-dimensional multimodal approaches.

We studied the impact of virus-specificity on persistence, finding higher persistence (p < 0.05) for EBV vs. CMV-specific T_TCR-C4. To directly compare CMV and EBV-specific T cells, we reanalyzed mass cytometry data (Schmidt et al, Cell Rep 2023) from healthy individuals and others with non-AML cancers (n = 143). Differential abundance analysis showed increased (FDR < 0.05) terminally differentiated (Temra) CMV vs. EBV-specific CD8⁺ T cells suggesting that CMV-specific T_TCR-C4 cells were affected by the substrate cell’s terminal differentiation, compromising post-infusion persistence. Disease status at T_TCR-C4 infusion also affected persistence, with lower persistence in patients with minimal residual disease (MRD)⁺ or overt AML vs. MRD^- patients.

To study T_TCR-C4 post-transfer, we built a 24-color spectral flow-cytometry panel for peripheral blood samples at ~1, ~7, ~28 days, and 4 months post-transfer. Unsupervised clustering and differential abundance analysis showed T_TCR-C4’s progressive differentiation from proliferative (Ki67⁺) effector to non-proliferative (Ki67^-) Temra cells with cytotoxic/NKL markers (CD45RA, CD57, KLRG1, GNLY) (p <0.05). This shift correlated with a drop in IFNg and TNF⍺ producing T_TCR-C4, associating the NKL shift with T cell dysfunction.

Single-cell RNA sequencing (scRNAseq) corroborated these results at the transcriptional level. Principal component analysis, trajectory, and velocity inference methods positioned T_TCR-C4 between effector memory/activated and NKL/Temra cells in the CD8⁺ T cell differentiation spectrum. However, when comparing timepoints with detectable (⁺) (n = 9) and non-detectable (^-) (n = 6) AML, we found that T_TCR-C4 overexpressed NKL markers in AML⁺, suggesting that AML accelerates NKL/terminal differentiation in antigen-specific cells. To verify, we rechallenged CD8⁺ WT1-specific T cells in an in-vitro model with WT1⁺ HLA-A2⁺-transduced K562 cells every 3-4 days. By day 13, T cells lost control of K562 cells and skewed to an NKL transcriptional profile (bulk-RNAseq). To further prove that T cells do not undergo T cell exhaustion in AML vs. solid tumors, we reanalyzed AML, melanoma, pancreatic and lung cancer scRNAseq independent datasets documenting exhaustion as a typical cell state in solid tumors but not in AML.

Data from one AML patient suggested that azacitidine treatment can prevent T_TCR-C4 terminal differentiation, likely increasing persistence. Another patient exhibited blasts with a myeloid derived suppressor cells transcriptional profile which became treatment-resistant, consistent with previous findings (Van Galen et al., Cell 2019). Killing assays showed that using a higher affinity TCR and co-targeting WT-1 with CD4⁺ and CD8⁺ T cells, can overcome this immune escape.

In conclusion, we demonstrate that AML-specific T cell dysfunction is driven by NKL skewing rather than T cell exhaustion. Our findings inform strategies that can be developed to target T cell dysfunction and overcome immune escape in TCR-T cell therapies for AML patients.