Abstract Background. Allogeneic Hematopoietic Stem Cell Transplantation (HSCT) is the only cure for high-risk acute myeloid leukemia (AML); nonetheless, relapse remains the major cause of death after such therapeutic option. Patients and Methods . We investigated the expression of Inhibitory Receptors (IR; i.e. PD-1, CTLA-4, TIM-3, LAG-3 and KLRG1) on different T-cell subsets infiltrating the bone marrow (BM) of 8 healthy donors (HD) and 32 allogeneic HSCT recipients diagnosed with Acute Myeloid Leukemia, collected at relapse (median 251 days) or at complete remission (CR) 1 year after HSCT. Inclusion criteria were: a diagnosis of acute myeloid leukemia or myelodysplastic syndrome, a relapse-free survival of at least 4 months after allogenein HSCT, absence of active GvHD, CMV infections or other complications at the time of sampling. Samples were analysed by multi-parametric flow cytometry for the expression of inhibitory receptors on T-cell subsets and the results were validated with BH-SNE, an unbiased dimensionality reduction algorithm. We exploited HLA-mimicking fluorescent molecules loaded with a specific epitope to screen anti-tumour and anti-viral T cells whereas the T-cell receptor repertoire was assessed by TRAC and TRBC RNA sequencing and the relative frequency of each T-cell receptor calculated. To evaluate T-cell function and specificity, CD107a expression, cytokine profiles and killing of autologous blasts were quantified. Results. After Haploidentical-HSCT PD-1, CTLA-4, 2B4 and Tim-3 were expressed at higher percentage when compared to HD, independently from the clinical outcome. In contrast, after HLA-matched HSCT, patients who relapsed displayed a higher frequency of BM-infiltrating T cells expressing PD-1, CTLA-4 and Tim-3 than CR pts (p<0.05) or HD samples (p<0.01). These data were validated by BH-SNE, who retrieved 32 clusters associated with relapse and composed of T cells co-expressing multiple IRs at high fluorescence, indicating a dominant inhibitory T-cell profile in the BM of relapsing patients. To further corroborate the biological relevance of immune checkpoints in the context of post-transplant relapse we typed AML cells, detecting high levels of the IRs ligands PD-L1, CD48, Galectin-9 and CD80/86 on blasts. To gain insights on the inhibited T-cell subpopulation identified in the BM of relapsing patients, we separately profiled the different T-cell memory subsets: in both HD and CR patients the IR expression was confined to effector memory and effectors whereas at relapse PD-1, 2B4, KLRG1 and Tim-3 were also expressed in BM-infiltrating central memory (TCM) and memory stem T cells (TSCM, p<0.01), thus indicating a pervasive and profound immune suppression that specifically involved the memory T-cell compartment. In accordance to this exhausted phenotype, we observed that BM CD8 T cells at relapse displayed lower degranulation ability and IL-2 production compared to CR (p<0.05). Notably, this functional impairment could be reversed by in vitro culture with high doses of IL-2. Interestingly, the TCR repertoire of BM-infiltrating T cells at relapse displayed a restricted clonality, suggesting that immune inhibitory signals are active on discrete and specific T-cell clones. To gain further insights on such clones, we assessed the IR expression profile on CD8 T cells specific for viral (CMV) and tumor-associated antigens (including peptides from WT1, EZH2 and PRAME). We observed a higher IR expression and co-expression on tumor-specific T cells when compared to viral-specific CD8 cells, particularly in case of patients who experienced post-transplant relapse. In accordance, IRpos sorted T cells harvested from relapsing patients showed a restricted TCR repertoire and, when challenged with autologous leukemic blasts, proved enriched in leukemic specificities as shown by higher expression of the activation marker HLA-DR (p<0.05), higher granzyme A and B production (p<0.001) and higher blast lysis in cytotoxicity assays (p<0.05) when compared to IRneg T cells. Conclusion. These results highlight a wide, yet reversible, immunological dysfunction likely mediated by AML blasts in the BM of patients relapsing after allogeneic HSCT, that is particularly evident on memory T cells specific for tumor antigens. This suggest and open new therapeutic opportunities for AML. Figure. Figure. Disclosures Bondanza: Novartis: Employment. Vago:GENDX: Research Funding; Moderna TX: Research Funding. Bonini:Intellia Therapeutics: Research Funding.