Early TCR diversity and recipient stem-cell persistence predict relapse and reveal therapeutic targets after allogeneic transplantation

Abstract

Abstract Introduction Standard treatment for high and intermediate risk acute myeloid leukemia (AML) includes induction chemotherapy followed by allogeneic hematopoietic stem cell transplantation (HSCT), which relies on the graft-versus-leukemia (GVL) effect to prevent relapse. However, relapse remains common, particularly in patients with high-risk mutations such as TP53. Declining donor chimerism often signals impending relapse, yet at this point, therapeutic options are limited. In this study, we characterized immune reconstitution dynamics and malignant stem cell characteristics to identify earlier determinants of relapse and uncover therapeutic targets for preemptive intervention. Methods We analyzed 49 bone marrow samples from 33 AML/MDS patients who underwent HSCT with T-cell–replete grafts from HLA 8/8 (n=30) or 7/8 (n=3) matched donors, following reduced-intensity (n=20) or myeloablative (n=13) conditioning and tacrolimus-based GVHD prophylaxis (one received post-transplant cyclophosphamide). We flow-sorted mononuclear or CD3+ T cells and generated paired single-cell RNA+T cell receptor (scRNA+TCR) sequencing libraries using 10x Genomics 5′ chemistry. We used Seurat to filter ~500,000 high-quality cells, BoneMarrowMap to annotate cell types, and DESeq2 and GSEA for gene expression analysis. We reconstructed TCR clonotypes with scRepertoire and calculated diversity by inverse Simpson index after downsampling. We used Souporcell to identify 227,388 donor and 115,392 recipient cells and Numbat to detect copy number variations (CNVs) in recipient cells and track malignant clones. Results We compared patients in long-term remission (n=19, >3.5 years) to those who relapsed (n=14, median relapse=9 months), focusing on 3–6 months after HSCT, a critical window for immune reconstitution. While T cell proportions were similar between the complete cohorts, patients with TP53 mutations who later relapsed had fewer naïve CD4+ T cells, more CD8+ effector memory T cells, and a lower CD4/CD8 ratio (median 0.54 vs. 1.62, p=0.032) compared with those who stayed in long-term remission. TCR diversity was similar between the complete long-term remission and relapse cohorts; however, when restricted to patients with TP53 mutations at 3–6 months after HSCT, those who later relapsed had significantly lower T cell diversity compared with patients who stayed in long-term remission, despite both groups being in clinical remission at the time of sampling. Extending our analysis to the hematopoietic compartment, we quantified donor and recipient contributions. While whole blood chimerism did not show a difference between long-term remission and relapse cohorts (median 99% vs. 98%) at 3–6 months, single-cell analysis showed significantly lower chimerism in the hematopoietic stem and progenitor cell (HSPC) compartment of patients who later relapsed (100% vs. 63%, p=0.0039). All patients with >5% residual recipient HSPCs at this timepoint relapsed within 18 months. In contrast, patients with near-complete HSPC chimerism in remission remained relapse-free. Using Numbat, we identified somatic CNVs in 36,837 recipient-derived cells from six patients. In remission samples, persistent recipient HSPCs harbor CNVs, nominating these cells as relapse precursors. Compared with donor HSPCs, these cells showed reduced proliferative capacity and upregulation of potential immunotherapeutic targets, PRAME and CALCRL. Together, these findings suggest that recipient-derived HSPCs are not only markers of relapse risk, but likely active drivers of recurrence, and our comparison of recipient vs. donor HSPCs highlights potential progenitor cell markers that might targeted to prevent relapse. Conclusions Our analysis of immune reconstitution dynamics following HSCT has important translational implications. First, in patients with TP53-mutated AML or MDS, reduced TCR diversity at 3–6 months after transplant was associated with relapse risk. Second, using a genotype-based pipeline, we showed that early persistence of recipient HSPCs predicts relapse across risk groups. Notably, whole blood-based chimerism failed to identify relapse risk, whereas early HSPC chimerism may serve as a more reliable biomarker. Together, these findings lay the groundwork for improved relapse prediction and support the development of personalized therapies, such as strategies to increase TCR diversity or preemptive adoptive targeted immunotherapies, to improve outcomes in AML and MDS patients.