Clonal architecture in human adult B cell acute lymphoblastic leukemia

Abstract

Leukemia represents a malignant disease of the hematopoietic system that is classified into various subtypes based on genetic and phenotypical characteristics. The present work focuses on adult B cell precursor acute lymphoblastic leukemia (BCP-ALL), which is characterized by an accumulation of undifferentiated blasts in the bone marrow (BM). Approximately 5 % of ALL patients display central nervous system (CNS) infiltration at diagnosis. Advanced first- line chemotherapy, including intrathecal CNS prophylaxis, achieves initial remission rates of over 90 %, yet the prognosis for elderly patients remains poor, with a five-year survival rate below 5 % for patients over 70 years. The causes of this dismal outcome are disease relapse and therapy resistance, presumably due to persisting leukemia-initiating cells (LICs). However, the identification and molecular characterization of LICs in adult BCP-ALL is challenging due to the considerable heterogeneity of the disease. Currently, two non-mutually exclusive models are used to explain this heterogeneity. The cancer stem cell model describes a hierarchical tumor organization, comparable to normal hematopoiesis, which has been established for acute myeloid leukemia (AML). Thereby, leukemia stem cells (LSCs) at the top of a differentiation hierarchy comprise self-renewal potential and the ability to produce subordinate progeny. In contrast, the model of clonal evolution assumes a continuous accumulation of mutations, resulting in a survival advantage for certain clones. Delineating the heterogeneous clonal composition resulting from ALL LICs poses a significant medical need to develop strategies to increase survival rates and prevent relapse. By developing a novel transcriptomic barcoding system, this study investigated the subclonal architecture of adult BCP-ALL and examined the molecular phenotype and local distribution of clones across various bones, the spleen, and the brain after in vivo leukemogenesis. First, the high-complexity ClonTracer library, comprising up to 4×10⁷ unique nucleotide sequences (barcodes), was introduced into the 3’ untranslated region of the fluorescent reporter mCherry-coding open reading frame. Subsequently, the lentiviral barcode construct was successfully introduced into five B-ALL patient-derived long-term cultures (PDLTCs) via viral transduction. The PDLTC cell model system reflects the heterogeneity and polyclonal composition of primary ALL samples while being modifiable and expandable. Upon xenotransplantation, cells from different patients homed to the BM and spleen in patient- specific patterns. Interestingly, while hundreds of clones persisted in these organs without notable expansion, most of the leukemic burden originated from the localized expansion of a small number of dominant clones. To identify key molecular features of these clones, single-cell Cellular Indexing of Transcriptomes and Epitopes sequencing (scCITE-seq) of barcoded PDLTC-CR cells was performed on the BD Rhapsody platform. This approach combined the information about clonal identity with the comprehensive mRNA and surface epitope profile at single-cell resolution, enabling the projection of differentiation trajectories. As a result, three distinct subpopulations (clusters) of leukemic cells were identified that emerged exclusively in vivo. Comparing their expression profiles to healthy B cell developmental stages revealed that Cluster 1 (MHC-I-high) was enriched for large Pre-B stage genes. Meanwhile, Clusters 2 (CD44+) and 3 (MHC-I-low, CD44-) were associated with later stages of B cell development. Furthermore, Cluster 1 displayed the highest clonal diversity, whereas Clusters 2 and 3 contained fewer clones but encompassed larger cell numbers per clone. Partition-based graph abstraction (PAGA) discovered that Cluster 1 serves as the sole connection hub between all clusters, suggesting that Cluster 1 cells sit at the apex of a hierarchical cell structure. Thereby, each clone displayed a specific differentiation behavior, but cells within the same clone exhibited similar molecular signatures across different animals, indicating that intrinsic differentiation properties are a key determinant of clonal behavior. Importantly, principal component analysis (PCA) enabled the classification of clonal behavior into three classes with unique, predictable differentiation patterns. Class A clones remained specifically in the immature cluster 1 and thus showed LSC character and self-renewal. In contrast, class B & C clones differentiated into cluster 2 (CD44+) and cluster 3 population, respectively. The identified subpopulations were further functionally examined by prospectively isolating and testing them in serial transplantation experiments in immunocompromised NSG mice. The results strikingly revealed that the more differentiated, CD44+ subpopulation exhibited a lower leukemogenic potential. Mice receiving CD44+ cells showed delayed engraftment of leukemic cells and reduced BM leukemic burden. Importantly, they demonstrated very much prolonged survival, and a subset of mice did not develop leukemia at all.