Clonal architecture and dynamics of somatic evolution in aplastic anemia and paroxysmal nocturnal hemoglobinuria

Abstract Aplastic anemia (AA) is a life-threatening bone marrow failure driven by T-cell attack on hematopoietic stem and progenitor cells (HSPCs). Pancytopenia results in infection, bleeding, and anemia. Diagnosis remains difficult due to the lack of biomarkers, and current classification, based solely on cytopenia severity, is outdated. Most patients receive immunosuppressive therapy (IST), but 30-40% do not respond, and 10-15% of responders later develop myeloid malignancies. PIGA-mutant clones may also emerge with or without marrow failure, resulting in paroxysmal nocturnal hemoglobinuria (PNH). HSPC clones with a survival advantage emerge in AA's hostile environment. These often carry somatic mutations in genes related to myeloid malignancy, antigen presentation, or the GPI-anchor pathway. Bulk targeted sequencing – used in most studies to date – is limited to the targeted regions, and cannot identify mutation timing, clonal trajectories, or stem cell dynamics. As a result, although somatic mutations correlate with outcomes and likely reflect disease biology, they are not routinely integrated into clinical decision-making. Building somatic single-cell phylogenies from individual patients allows clonal reconstruction, mutation timing estimation, and tracking of clonal dynamics. We hypothesized that applying this approach to AA and PNH would uncover key aspects of disease pathogenesis: What is the latency between an autoimmune trigger and clinical disease? Why do specific mutations arise in some but not all patients? Can these data help predict disease course or treatment response? We used spontaneous genome-wide somatic mutations in single HSPCs to build high-resolution hematopoietic lineage trees from 10 patients with AA (n=7) or PNH (n=3). CD3-depleted peripheral blood HSPC-derived colonies were cultured in methylcellulose and underwent whole-genome sequencing (WGS) at 10-15X coverage (16-78 colonies per patient). Final dataset: 340 colonies (73 PNH, 267 AA) passed QC. Ultradeep whole-exome duplex sequencing was also performed on granulocytes from 13 patients. Using dN/dS, we identified genes under positive selection - those with more non-synonymous mutations than expected by chance. Additionally, single-cell WGS using primary template amplification (PTA) of HSPCs and lymphoid cells avoids any in vitro culture bias and enables sequencing of multiple different cell populations. Key findings:AA patients showed two distinct clonal landscapes: myeloid driver predominant (MDP) and immune escape predominant (IEP). MDP patients had parallel clonal expansions with mutations in genes such as ASXL1, U2AF1, or BCOR. IEP patients had multiple clones with inactivated HLA genes-either by 6p loss-of-heterozygosity (LOH) or nonsense mutations.Previously unreported recurrent mutations were identified in ERAP1, a gene involved in antigen processing, suggesting a novel immune escape mechanism.Immune escape variants often arose early in life. In one patient presenting with AA at age 8, the dominant 6p LOH clone was estimated to originate in utero (<15 weeks post-conception).Among PNH patients, both monoclonal (2/3) and polyclonal (1/3 with 4 independent PIGA mutations) patterns were observed. PIGA mutations were acquired prior to age 15, though clinical onset occurred decades later.Unlike in malignancy, where driver mutations and clonal expansion occur closely together, PIGA clones expanded years after mutation acquisition. Our phylogenies separate the timing of mutation acquisition from the onset of the selection pressure of autoimmunity.Ongoing analysis of PTA-based single-cell WGS and methylomes will further delineate early clonal events and epigenetic changes.Our data represent the largest WGS dataset in AA. Despite the small cohort size, our in-depth analysis of clonal architecture and dynamics provide unprecedented insights into AA evolution. Patients exhibit distinct evolutionary patterns: either immune escape or myeloid driver predominance. We also identify a novel gene under selection in AA – ERAP1. Immune escape variants typically arise early and expand gradually, whereas myeloid drivers often emerge later but expand more rapidly. These differences likely reflect underlying disease mechanisms. A revised AA classification that integrates clonal architecture could guide treatment selection, avoid ineffective IST, and identify those at increased risk of malignant transformation.