Deep immune signature of immune-mediated aplastic anemia patients shows distinct subsets of regulatory T cells associated with response to treatment: Results from the phase 3, randomized EBMT race clinical trial

Abstract Introduction: Immune-mediated aplastic anaemia (AA) is bone marrow failure syndrome, where T-cell mediated destruction of hematopoietic stem and progenitor cells (HSPCs) results in pancytopenia. Overall, two third of patient respond to immune suppressive therapy (IST) with antithymocyte globulin (ATG) and cyclosporine A (CsA), adding eltrombopag (EPAG) further improves response, but the immune signatures that predict outcome remain unclear. Methods: We studied at diagnosis 130 previously untreated severe/very-severe AA patients enrolled in a randomised phase-3 trial of hATG + CsA with (Arm B, n = 65) or without EPAG (Arm A, n = 65). Number of patients analysed at both baseline and 6 months was 95. The frequency of 11 CD8+ and 22 CD4+ T-cell subsets, as well as B cell, Myeloid and NK cells were defined by mass cytometry (CyTOF) using panel of 38 antibodies. The immune architecture was associated with patients' age, disease severity, treatment arms, treatment outcomes at 6 months and other patient specific features using non-parametric tests. Predictive features were identified using multinomial and multivariable logistic regression. Results: At baseline, across all the cohort exhibited broad immune dysregulation. Most patients presented reduced myeloid and NK cell frequencies, while activated effector T-cells, and particularly non-naïve CD8⁺ Tc1 and effector memory CD4⁺ T cells, were expanded. Patients with very severe AA (vsAA) had significantly lower myeloid cells and higher levels of effector Th1 CCR4⁺ and Th1 CCR4⁺ CD4⁺ subsets compared to those with severe AA (sAA). Immune environment skewed towards pro-inflammatory cells in vsAA patients, exhibiting elevated levels of activated cytotoxic and effector T-cell subsets, including CD38+PD1+ Tc1, Tc2, and CD103+ CD8+ cells. This shift was more pronounced in vsAA patients younger than 40 years compared to vsAA patients older than 60 years, while vsAA patients between 40 and 60 years showed intermediate findings. When stratified by 6-month treatment response, patients achieving complete response (CR) had a distinct baseline immune profile. Naïve Tregs CD45RA⁺CCR4⁻ were significantly enriched in CR patients, while memory Tregs CD45RA⁻CCR4⁺ were elevated in non-responders (NR). Effector Th1 CCR4⁺ cells and central memory CD45RA-CD27+ cells were also decreased in CR, and partial responders (PR) displayed intermediate frequencies across these subsets. Multinomial logistic regressions identified two regulatory-T-cell (Treg) signatures. Naïve Treg CD45RA+ CCR4- independently predicted better hematological response, while memory Treg CD45RA- CCR4+ was inversely associated. HLA-DRB1*15:01 (present in 47 %) co-segregated with older age and higher memory-Treg levels but did not affect response. At 6 months, both treatment groups shared immunological shifts; Myeloid cell, NK cell and Effector memory CD45RA-CD27- frequencies increased, indicating improved hematopoiesis, restored innate immunity and accumulation of differentiated T cells. Activated Tc CD8+ cells and effector Th1 CCR4+ were decreased however terminally differentiated effector memory Tcells, which can be identified by markers CD45RA+ CD27-, were increased in both treatment arms but significantly in arm B. Effector Th1 CCR4+ cells declined in CR at 6 month, this declined more pronounced in arm B; however, this may be influenced by the number of CR difference between arms. Responder profiles were marked by maintenance or enrichment of naïve Tregs and reduced memory Tregs post-treatment, reinforcing the prognostic value of baseline Treg subset balance. Conclusions: In summary, AA is characterized by a broadly shared immune activation state marked by reduced innate immunity and expanded effector T-cell subsets. Within this framework, two regulatory T-cell subsets, naïve and memory Tregs, emerge as independent predictors of treatment response, regardless of age. After treatment they remained stable in CR and PR; however, memory Treg slightly decreased, and Naïve Treg slightly increased in NR. These findings highlight immune profiling as a powerful tool for predicting outcome and tailoring therapy in AA.