Single-cell chromatin profiling reveals early epigenetic reprogramming, subtype-specific regulatory networks, and JAK/STAT–IRF vulnerabilities in high-risk chronic myeloid leukemia

Abstract Myeloproliferative neoplasms (MPNs), including chronic myeloid leukemia (CML), are oncogene-driven blood cancers arising from hematopoietic stem cells (HSCs) that often progress from a ‘smouldering’ chronic phase to a drug-resistant, pre-terminal blast phase. In CML, recent studies have shown that stem and progenitor cells (SPCs) from chronic phase (CP) patients at high risk of blast crisis (BC) transformation exhibit a convergent gene expression signature that mirrors that of overt BC, independent of underlying somatic mutations (Ko et al., Blood 2020; Krishnan et al., Blood 2023). Collectively, these findings suggest the existence of a common drug-resistant SPC state. Identifying the therapeutic vulnerabilities of such states holds promise for transforming the management of high-risk MPNs. To identify these resistant states, we performed single-cell (SC) multi-omic profiling (scRNA-seq, scATAC-seq, and targeted mutational analysis) on pre-treatment bone marrow (BM) mononuclear cells from (n=24, CML) three clinical groups (optimal imatinib [IM] responders, suboptimal IM responders, IM failure with BC progression; n=5/5/6), normal BM (n=4), and BC samples (myeloid n=2, lymphoid n=2). scATAC-seq data annotated 28 cell types from 240,456 cells; we identified DARs with DESeq2, inferred TF activity with ChromVAR, integrated gene-to-peak and enrichment analyses, mapped epigenetic remodeller occupancy with HOMER, and built Gene Regulatory Networks (GRNs) using SCENIC+ algorithm. Across all prognostic groups, analysis of LSCs revealed a shared, CML-specific epigenetic state distinct from normal BM HSCs, characterized by 29,317 differentially accessible regions (DARs) among 421,985 peaks (6.95%), with ~75% located in intronic and intergenic regions. Consistent with the drug-resistant phenotype of LSCs, functional annotation of these DARs highlighted hallmark features of resistance, including increased inflammatory and proliferative signaling and diminished stemness. Importantly, these LSC states were closely linked to canonical BCR::ABL1 signaling pathways, with enrichment of AP-1, STAT1/3/5A/5B, and NF-κB TF binding activity, implicating these pathways in the establishment of the LSC state. Strikingly, among the top 10 CML-specific transcription factors driving the LSC program, only STAT binding activity correlated with increasing clinical resistance, specifically through enhanced IL2-STAT5 and TNFα signaling. Given the relative stability of the LSC state across prognostic groups, we next examined the progenitor population for cell state changes. Unlike LSCs, nearly 32% of the progenitor epigenome showed altered chromatin accessibility with BC progression more extensive in lymphoid BC (LBC) (94,245 DARs) than myeloid BC (MBC) (43,378 DARs). In CP patients who developed LBC, 29,167 LBC-specific DARs were enriched in lymphoid progenitors at diagnosis. Whereas MBC-associated DARs (1,278–5,569) were weakly detected across myeloid subsets in CP patients who developed MBC. These findings support the notion of a common epigenetically-primed state in diverse progenitor compartments, predisposing them to subsequent transformation. Subtype-specific TF analysis showed AP-1/CEBP enrichment in MBC, and IRF8/9 in LBC. GRN reconstruction from scRNA-seq showed these TFs drive resistance associated programs characteristic of full-blown BC (Ko et al. Blood 2020). Ongoing GWAS-linked analysis seeks germline or acquired variants tied to these primed states. Lastly, we identified shared upregulation of STAT and IRF pathways in both BC subtypes, TFs downstream of JAK1/2/3, indicating convergent regulatory programs. This led us to testt, an FDA-approved JAK inhibitor, to dismantle these cell states. Although tofacitinib alone showed minimal cytotoxicity, it markedly sensitized BC cells (n=4) to TKIs (~90% killing) while sparing normal progenitors (~25% killing). Thus, mapping the CML epigenetic landscape identified core transformation regulators and revealed actionable therapeutic opportunities. This study presents the first scATAC atlas of CML resistance and progression, revealing that: (1) Epigenetic reprogramming in CP precedes BC transformation; (2) Distinct TF networks drive BC subtypes; and (3) JAK inhibition is a novel strategy to dismantle JAK/STAT-IRF activation in high-risk SPCs. Our findings highlight the power of SC- epigenomics to identify actionable vulnerabilities and guide pre-emptive therapies for high-risk MPN patients.