Anti-TIM3 with hypomethylating agent revives NK and cytotoxic CD4+ T cell activity in patients with AML or MDS

Abstract Recent clinical trials in AML combining immune checkpoint inhibitors (anti-PD1, anti-CTLA4) with hypomethylating agents have yielded only modest response rates. However, there has been exceptional responders achieving durable complete responses in all these trials. TIM3 is a checkpoint molecule expressed both on immune and leukemic cells, making it an interesting target in AML. Here, we conducted a comprehensive immunomonitoring of a phase Ib trial (NCT03066648) evaluating decitabine in combination with anti-TIM3 antibody sabatolimab (MBG453). We studied paired bone marrow (BM) and peripheral blood samples from 11 unfit newly diagnosed (ND) or relapsed/refractory (R/R) AML patients and 1 MDS patient with single-cell RNA and T cell receptor sequencing (scRNA+TCRαβ-seq) and flow cytometry. We also performed co-culture assays of primary immune and leukemic cells with scRNA+TCRαβ-seq readout in a patient with a durable complete response. In scRNA+TCRαβ-seq data, HAVCR2 (encoding TIM3) is highly expressed in NK cells, myeloid cells, and unconventional T cells, with some expression in CD8+ T cells. This contrasts with PDCD1 (encoding PD1) and CTLA4, which are primarily expressed in CD8+ and CD4+ T cells, respectively, albeit at low levels in AML. At baseline, responders had higher proportions of CD4+ T cells and B lymphocytes, whereas non-responders had more cytotoxic CD8+ T cells, which was confirmed by flow cytometry. After starting anti-TIM3+HMA therapy, both groups showed increased numbers of CD8+ T cells and NK cells. Responders to anti-TIM3+HMA therapy generally had more mature (CD56dim and adaptive) NK cells, with key transcriptional changes in type I/II interferon (IFN) and NF-κB pathways. Functional co-culture assayswith primary blasts from a complete responder revealed three NK activation states: cell-contact activated, cytokine-secreting, and type I IFN-responsive. NK cells showed enhanced type I IFN responses after therapy correlating with clinical remission, but this response was reduced when NK cells were co-cultured with blasts from a relapse time point, suggesting potential immune evasion. Cytotoxic CD4+ T cells were more abundant and highly clonal in pre-treatment samples from responders. In an exceptional responder with concomitant CD4+ T-LGLL diagnosis, the dominant CD4+ T-LGLL clone comprised 21% of the TCR repertoire at baseline, persisted at remission, but fell to 7% at relapse. Transcriptomic analysis showed upregulation of IFNG and TNF post-therapy, with a subset peaking soon after treatment. To assess whether the T-LGLL clone targets patient's leukemic cells, we engineered Jurkat reporter cells with the TCRαβ from the CD4+ T-LGLL clone and performed co-culture assays with scRNA-seq readout. In comparison to mock transduced cells, T-LGLL reporter cells showed upregulation of type I IFN genes when co-cultured with patient's leukemia cells, especially in presence of antigen-presenting cells. In co-culture of patient's own T-LGLL cells, pre-therapy cells responded strongly to blasts from screening but lost reactivity to relapse blasts, while post-therapy T-LGLL cells retained activity to both. Exhausted CD8+ T cells were rare in AML BM (<0.5% pre-treatment), most CD8+ T cells exhibited effector or memory phenotypes. Functional co-culture assays showed that CD8+ T cells retained robust activation capacity, marked by TNF and IFNG expression, when co-cultured with autologous blasts. Anti-TIM3+HMA therapy preferentially expanded small CD8+ T cell clones in responders, whereas non-responders showed expansion of larger, more cytotoxic clones, some of which were recognized to target viral epitopes. ScRNA-seq analysis of myeloid cells revealed that responders had a higher proportion of differentiated myeloid cells, such as classical monocytes. In the exceptional responder, relapse blasts in co-culture upregulated immune evasion genes (e.g., CD274 encoding PDL1) and showed the highest HLA class I expression, indicating putative immune escape mechanisms from T and NK cells, respectively.Our study provides a comprehensive analysis of anti-TIM3 in combination with decitabine in AML/MDS. We demonstrate that TIM3 blockade modulates the immune landscape by activating mature and adaptive NK cells, promotes cytotoxic CD4+ T cells, and primes small CD8+ T cell clones for expansion. Our results suggest that cytotoxic CD4+ T-LGLL cells may boost responses to immune checkpoint therapy in AML.