Abstract B035: Cell state transitions drive the evolution of disease progression in B-lymphoblastic leukemia

Abstract Cancer stem cells (CSCs), first identified in acute myeloid leukemia with a CD34+/CD38- phenotype, are hypothesized to sustain clonal survival through self-renewal and resistance to chemotherapy. In B-cell acute lymphoblastic leukemia (B-ALL), stem-like properties can be reacquired through plastic transitions among stem-like, hematogone-like, and naïve B-cell–like states. This plasticity provides an evolutionary route by which chemotherapy-resistant subpopulations re-establish stem-like compartments, contributing to disease relapse under therapy. Such fluidity complicates efforts to target the stem-like compartment directly and underscores the need to understand how chemotherapy reshapes these transitions to drive resistance. We analyzed flow cytometry data from bone marrow (N=63) and peripheral blood (N=46) samples of B-ALL patients at diagnosis, post-induction, and in remission (N=38). Four major cell states are defined from stem-like, progenitor, and differentiated: 1) CD34+/CD38-, 2) CD34+/CD38+, 3) CD34-/CD38+, and 4) CD34-/CD38-. Patient-specific, continuous-time Markov chain modeling quantified transition rates, Qij between state i and j. In remission, differentiation (Q31) consistently exceeded dedifferentiation (Q13), reflecting normal lineage progression. In relapse, however, Q31 – Q13 was often negative with greater variance, indicating that resistant clones rebuilt stem-like compartments through dedifferentiation. This heterogeneity underscores distinct evolutionary mechanisms of resistance across patients. We next explore evolution-based interventions based on our model. Promoting differentiation (increasing Q31) transiently reduced stem-like fractions but failed to prevent rebound due to ongoing dedifferentiation. Inhibiting dedifferentiation (reducing Q13) produced sustained depletion of CD34+/CD38- cells and more effectively shifted relapse dynamics toward remission-like equilibria. Thus, targeting the dedifferentiation fraction emerges as a promising therapeutic strategy. To link transition dynamics with clinical outcomes, we applied mechanistic learning by integrating Markov-derived features with machine learning. Linear discriminant analysis (LDA) achieved high predictive accuracy for BCR: : ABL1 status (80%), disease status (84% marrow, 89% blood), and minimal residual disease (71%). Principal component analysis highlighted the balance of differentiation and dedifferentiation as the most informative predictor. These results show that resistance evolves not only from persistence of stem-like cells but also from reciprocal transitions that replenish this compartment. Our findings suggest that therapy pressure influences cell transition landscapes in B-ALL, with relapse often associated with dedifferentiation-driven restoration of stem-like states. Mechanistic learning highlights stem-state reciprocity (in/out rate balance) as a potential predictive biomarker, while modeling indicates that inhibiting dedifferentiation may be effective to reduce resistant stem-cell like clones. Citation Format: Sadegh Marzban, Curtis Gravenmier, Ling Zhang, Jeffrey West. Cell state transitions drive the evolution of disease progression in B-lymphoblastic leukemia abstract. In: Proceedings of the AACR Special Conference in Cancer Research: Cancer Evolution: The Dynamics of Progression and Persistence; 2025 Dec 4-6; Albuquerque, NM. Philadelphia (PA): AACR; Cancer Res 2025;85 (23Suppl): Abstract nr B035.