Single-cell transcriptional profiling identifies ADGRG1+ CD8 T cells as critical mediators of graft-versus-leukemia effect in AML relapse after haploidentical hematopoietic stem cell transplantation

Abstract

Abstract Purpose: Relapse remains a major obstacle in acute myeloid leukemia (AML) patients after haploidentical hematopoietic stem cell transplantation (haplo-HSCT). This study aimed to (1) establish a single-cell transcriptional atlas of the bone marrow microenvironment (BMME) in relapsed AML patients post-haplo-HSCT; (2) identify key cellular subsets mediating graft-versus-leukemia (GVL) effect; (3) map intercellular communication networks between GVL effectors and AML cells within BMME; and (4) characterize dynamic evolution of T cell subsets via TCR repertoire analysis. Methods: Bone marrow specimens were collected and cryopreserved from AML patients with post-haplo-HSCT recurrence during complete remission (CR) and relapse phases (3–12 months post-transplantation). Single-cell RNA sequencing (scRNA-seq) libraries were generated from thawed samples followed by high-throughput sequencing. Comprehensive bioinformatics analyses were performed for transcriptional profiling, functional annotation, and TCR repertoire characterization. Results: We successfully constructed a single-cell atlas of BMME in relapsed post-haplo-HSCT patients. UMAP visualization revealed patient-specific clustering of AML cells, whereas non-malignant BMME components clustered by cellular lineage. Systematic T cell subset analysis identified 11 distinct populations, among which ADGRG1+ CD8 T cells (expressing GPR56 encoded by ADGRG1) exhibited robust cytotoxic properties and enrichment in TCR signaling pathways, implicating their central role in GVL effect. Pre-relapse functional conversion of ADGRG1+ CD8 T cells was observed, shifting from cytotoxic dominance to enhanced antigen presentation (upregulated HLA-DRA). Post-relapse, these cells displayed either attenuated cytotoxicity (4/6 patients) or hyperactivation (2/6 patients) with concurrent fluctuating exhaustion scores, indicating their dysfunction as a critical driver of GVL failure. Cell-cell communication analysis demonstrated that AML cells suppressed ADGRG1+ CD8 T cell activation via LGALS9-CD45 interactions (patients P1, P2, P4, P6) or promoted exhaustion through SIRPA-TIGIT pathway (patients P3, P5), with downstream impacts on T cell proliferation, oxidative stress, and metabolic functions—identifying these pathways as potential therapeutic targets. ADGRG1+ CD8 T cells exhibited high TCR clonal heterogeneity with relative stability across CR and relapse phases, while serving as major contributors to clonal expansion and de novo clone generation. Post-relapse, significant phenotypic transition toward GZMH+ CD8 T cells was observed, with clonotype dynamics reflecting GVL evolutionary changes. Clinical correlative analysis showed significantly higher frequencies of GPR56+ CD8 T cells in CR versus relapse cohorts, with inverse correlation to GZMK+ CD8 T cells. Xenograft models confirmed that GPR56+ CD8 T cell infusion significantly delayed tumor progression, supporting their therapeutic potential for relapse prevention. Conclusion: ADGRG1 + CD8 T cells represent critical effectors of GVL effect in post-haplo-HSCT AML. Their early post-transplant functional conversion and subsequent dysfunction create a permissive microenvironment for AML immune escape. Reciprocally, AML cells induce ADGRG1+ CD8 T cell dysfunction through patient-specific mechanisms. The interplay between AML cell type specificity and ADGRG1+ CD8 T cell functional abnormalities drive post-transplant relapse. These findings provide novel mechanistic insights into GVL biology and identify potential therapeutic strategies for preventing AML relapse after haplo-HSCT.