Dynamic mutational evolution and transcriptomic remodeling on 3rd-generation TKI therapy in TKI-resistant patients with chronic myeloid leukemia

Abstract BackgroundThird-generation tyrosine kinase inhibitors (3G-TKIs), olverembatinib and ponatinib, improved outcomes for patients with resistant chronic myeloid leukemia (CML). However, nearly half still developed resistance or progression. The longitudinal molecular evolution during therapy, at both genomic and transcriptomic features, remains incompletely understood. We investigated the dynamic clonal architecture and transcriptional reprogramming during 3G-TKI therapy to elucidate mechanisms of therapeutic response and resistance. MethodsWe analyzed 202 CML patients treated with 3G-TKIs. Baseline and serial bone marrow or blood samples were collected for targeted DNA sequencing and RNA sequencing. Variant allele frequency (VAF) trajectories were used to classify mutational dynamics. Gene expression profiles (GEPs) were analyzed by multi-machine learning and unsupervised clustering. Pathway enrichment, immune deconvolution (xCell algorithm) and T cell state analysis (TcellSI algorithm) characterized biological transitions during therapy. ResultsDynamic mutational profiling revealed divergent clonal evolution. In responders (n = 49), the frequency and number of ABL1and other somatic mutations significantly declined (p < 0.05). In contrast, primary resistance cases (n = 62) showed persistent or newly emergent high-VAF mutation clones, particularly ASXL1G646Wfs*12, RUNX1, and PHF6. Unsupervised clustering identified four prognostically distinct mutational patterns: (1) persistent clones; (2) emergent/expanding clones; and (3 - 4) contracting or eradicated clones. Patterns 1 and 2 were significantly correlated with therapy resistance, whereas Patterns 3 and 4 were correlated with favorable responses. Transcriptomic remodeling during 3G-TKI therapy provided novel insights into the molecular basis of response and resistance. In responders, leukemic stem cells (LSCs), E2F, MYC and DNA replication related pathways were markedly down-regulated (FDR < 0.001) during therapy, while immune and apoptotic pathways (TNF-α and IFN-α) were significantly activated (FDR < 0.001 - 0.04). Immune infiltration, including T cells, NK cells and dendritic cells, significantly increased (p < 0.001). In contrast, resistant patients retained or more-activated LSCs and metabolic related pathways with minimal immune induction, reflecting an immune-cold, transcriptionally static state. Using multi-machine learning and unsupervised clustering, 3 molecular subtypes were defined: GEPC1 (immune-active, n = 42), GEPC2 (stromal-remodeling, n = 86) and GEPC3 (stemness-high and immune-cold, n = 21) with distinct therapy responses and outcomes (p < 0.001 - 0.006). During therapy, patients in GEPC1 showed transcriptomic features characterized by down-regulation of LSCs and cell cycle related pathways (FDR < 0.001), along with strong up-regulation of immune-related pathways, such as TNF-ɑ, interferon-ɑ pathways (FDR < 0.001). Immune infiltration, such as T cells, B cells and NK cells, increased significantly during therapy. GEPC2 exhibited a moderate but partial transition, with suppression of proliferation signatures and activation of extracellular matrix and stromal pathways, and enrichment of fibroblasts and M2 macrophages. Immune activation was present but attenuated. In contrast, GEPC3 showed only limited transcriptional remodeling during therapy. Although LSCs and proliferative signatures declined slightly, they remained markedly higher compared to GEPC1/2, and immune activation was absent. xCell and TcellSI showed no increase in effector T cells, dendritic cells or cytotoxic function, consistent with an immune-cold micro-environment. Importantly, these patients often experienced disease progression without significant new mutations, suggesting transcription mediated resistance independent of clonal evolution. Conclusions Our integrative longitudinal analyses revealed that 3G-TKI therapy induced divergent molecular trajectories in CML: durable responses were defined by mutation clonal clearance, LSCs suppression and immune activation, whereas resistance was driven by persistent or expanding mutation clones (e.g., ASXL1G646Wfs*12, RUNX1and PHF6), sustained stemness, and immune-cold states. Transcriptomic remodeling provided a powerful early indicator of therapeutic failure and supports dynamic molecular monitoring to guide personalized therapy.