AUTHORS
Oyeronke Ayansola, PhD, Yael Morgenstern2, Jae-Sook Ahn, MD, PhD3, Georgios Nteliopoulos4, Jane Felicity Apperley, FRCP, FRCPath, MB5, Alexandra Bazeos6, Sin Tiong Ong, MD, MA, MBBChir, MRCP7, Tomasz Stoklosa, MD, PhD8, Danielle Pyne, BSc9, Azin Keshavarz, MHSc.10, Amirthagowri Ambalavanan, PhD11, Maria Perusini, MD, Msc12, Hassan Sibai, MD, Msc, BSc13, Hyeoung-Joon Kim, MD14, Dennis Dong Hwan Kim, MD, PhD15
AFFILIATIONS
1 Princess Margaret Cancer Centre, University Health Network, Toronto, Ontario, Canada
2 Princess Margaret Cancer Centre, University Health Network, Toronto, Ontario, Canada
3 Chonnam National University Hwasun Hospital, , South Korea, Hwasun, Korea, Rep. of South
4 Imperial College London, London, United Kingdom
5 Imperial College London, London, United Kingdom
6 Imperial College London, London, United Kingdom
7 DUKE-NUS Medical School, Singapore, Singapore
8 Department of Tumor Biology and Genetics, University Clinical Center, Medical University of Warsaw, Warsaw, Poland
9 Princess Margaret Cancer Centre, University Health Network, Toronto, Ontario, Canada
10 Princess Margaret Cancer Centre, University Health Network, Toronto, Ontario, Canada
11 Princess Margaret Cancer Centre, University Health Network, Toronto, Ontario, Canada
12 Faculty of Medicine, University of Toronto, Toronto, Canada, Princess Margaret Cancer Centre, University Health Network, Toronto, Ontario, Canada
13 Faculty of Medicine, University of Toronto, Toronto, Canada, Princess Margaret Cancer Centre, University Health Network, Toronto, Ontario, Canada
14 Chonnam National University Hwasun Hospital, , South Korea, Hwasun, Korea, Rep. of South
15 Faculty of Medicine, University of Toronto, Toronto, Canada, Princess Margaret Cancer Centre, University Health Network, Toronto, Ontario, Canada
ABSTRACT
Introduction
DNA-methylation instability (DMI) is the per-sample dispersion of beta-values across tightly controlled CpGs, complementary to genetic mutation. Epigenetically stable loci (ESL)-gain DMI captures aberrant methylation gain at loci held unmethylated in healthy hematopoiesis, scored against an external panel (5% recurrence threshold; Basrai et al., Nat Commun 2026). DMI is measurable from a single specimen without a matched normal, serving as a clonality biomarker in chronic myeloid leukemia (CML).
ASXL1 encodes a chromatin regulator acting through Polycomb-repressive complex 2 (PRC2) and the BAP1 deubiquitinase module; loss-of-function mutations reduce repressive histone marks (notably H3K27me3) and derepress leukemogenic programs. ASXL1 is frequently mutated in CML and linked to inferior tyrosine kinase inhibitor (TKI) response and progression, plausibly connecting this lesion to methylation instability and treatment failure.
We hypothesized that 1) high baseline DMI marks patients who subsequently fail TKI therapy, 2) DMI rises with progression, and 3) DMI associates with ASXL1 mutation status.
Patients & Method
Three independent cohorts (263 arrays from 185 cases, 178 CML patients and 7 healthy controls): 1) discovery cohort from UK (CD34+ PB/BM; Nteliopoulos et al., Haematologica 2019; n=62; 450K), 2) replication cohort 1 from Canada (unselected BM; n=78, 156 paired samples; preprocessNoob; EPIC), and 3) replication cohort 2 from Singapore (CD34+ PB/BM; Ko et al., Blood 2020; n=45; 450K), spanning chronic phase (CP) and blast phase (BP). DMI was the external ESL-gain metric (defined above). Endpoints were TKI response, time to TKI failure, and disease stage (CP vs BP). ASXL1 mutation status was available in discovery cohort and replication cohort 1, while ongoing for replication cohort 2. Diagnosis and follow-up were co-normalized (Noob).
Result
DMI showed a clear trend, highest in BP, intermediate in CP and lowest in normal healthy controls, discriminating BP from CP almost perfectly (AUC 0.997; trend rho 0.92, p=1e-18).
Baseline ESL-gain DMI separated responders from non-responders in the discovery (AUC 0.80, p=5.5e-4) and validated at diagnosis in replication cohort 1 (AUC 0.62; age/sex-adjusted OR 2.02 per SD, p=0.037). DMI was uncorrelated with BCR::ABL1 transcript level (rho=-0.02, p=0.85), implying it captures epigenetic rather than tumour cell burden. In paired samples (replication 1), DMI rose from diagnosis to follow-up in non-responders (p=0.016) but plateaued in optimal responders (p=0.42; between-group change p=0.21), indicating instability accrues under inadequate disease control.
Higher baseline DMI predicted shorter time to TKI failure in the discovery (Cox HR 3.05, 95% CI 1.53-6.08, p=1.5e-3). Combining discovery and replication 1, the high-DMI group had 3.62 times higher risk of failure than the low-DMI group (HR 3.62, 1.58-8.26, p=0.002). After Sokal score adjustment, DMI retained its prognostic value (adjusted HR 3.29, p=2.5e-4), adding value beyond established CML risk scores.
Mutation data were available for 120 patients (discovery & replication 1): ASXL1 was mutated in 6 (5%), 16 had other mutations, and 98 had none. ASXL1-mutant patients had only marginally higher DMI (p=0.16) and a modest differential-methylation signature (264 CpGs; FDR<0.05, Δβ>5%). Critically, the DMI-response signal was independent of ASXL1 mutation status: odds of TKI failure per SD of DMI were 2.17 (p=0.008), remaining significant after adjusting for ASXL1 mutation (OR 2.03) or restricting to ASXL1-wild-type patients (OR 1.93, p=0.020). Of note, the patients carrying any somatic mutation had higher DMI than mutation-negative patients (p=0.002), reflecting enrichment among resistant patients: the difference vanished in optimal responders (p=0.53) and persisted only in resistant patients (p=0.019), where the other-mutation group contains ABL1 kinase-domain mutations conferring to TKI resistance.
Conclusion
Baseline ESL-gain DMI is a robust biomarker of TKI response, treatment failure, and CML progression, independent of ASXL1 mutation status, and uncorrelated with BCR::ABL1 transcript burden. Measureable from single pre-treatment sample without a matched normal, ESL-gain DMI can be a potential tool for baseline CML risk stratification, warranting prospective validation.
DISCLOSURES
Dennis Dong Hwan Kim, MD, PhD: Novartis - Research Funding; Enliven - Research Funding; Abbvie - Research Funding; Incyte - Research Funding; Sanofi - Research Funding; Ascentage - Research Funding; JAZZ - Consultancy (Includes expert testimony); Abbvie - Consultancy (Includes expert testimony); Ascentage - Consultancy (Includes expert testimony); Daiichi-Sankyo - Consultancy (Includes expert testimony); Novartis - Consultancy (Includes expert testimony); Sanofi - Consultancy (Includes expert testimony); J&J - Consultancy (Includes expert testimony); Astellas - Consultancy (Includes expert testimony)