Relapse is the principal cause of mortality in acute myeloid leukaemia (AML), with ~50% of patients who achieve remission ultimately developing recurrent disease. Relapse is driven by rare therapy-resistant minimal residual disease (MRD) clones that persist after treatment and later re-expand. Despite their clinical importance, these clones remain poorly understood: they are rare, heterogeneous, and frequently rely on dynamic, non-genetic resistance mechanisms not captured by genomic profiling. Progress has been limited by the lack of technologies enabling prospective identification, isolation, and functional interrogation of these resistant populations.

To address this, we developed Clone Claw, a programmable clone-retrieval platform integrating single-cell profiling, lineage tracing, and targeted clone isolation. Building on SPLINTR (Single-cell Profiling and Lineage Tracing), it incorporates CRISPR-mediated transcriptional activation, split fluorescent reporters, and barcode-specific guide RNAs to selectively label and enrich predefined clones. Validation across diverse barcodes assessed specificity, reproducibility, and enrichment efficiency. Adapter sequences were incorporated to enable downstream multi-omic and spatial profiling, facilitating integrated transcriptional and epigenetic characterisation. To model clinically relevant MRD, barcoded human AML cells were engrafted into immunocompromised mice and treated with frontline therapies to capture remission, MRD, and relapse in vivo.

Clone Claw achieved robust activation across diverse barcodes (>60% reporter activation) and detected clones down to 0.01% frequency in spike-in assays. In vivo treatment induced remission followed by relapse, recapitulating clinical trajectories. Lineage tracing revealed heterogeneous clonal responses and identified discrete clones that persist through therapy and drive recurrence. Importantly, the platform enables prospective isolation of these clones before treatment, during therapy, and at relapse, allowing direct investigation of adaptive state transitions. Ongoing transcriptional and epigenetic analyses are defining persistence-associated programmes and nominating candidate therapeutic vulnerabilities. Comparative studies with matched patient samples are underway.

Clone Claw provides a sensitive and scalable framework for longitudinal interrogation of rare therapy-resistant clones in AML. By enabling dynamic analysis of MRD, this approach addresses a central barrier in relapse biology and establishes a broadly applicable strategy to dissect and target therapeutic resistance across cancers.

I thank my supervisors, laboratory members, collaborators, and core facilities at the Peter MacCallum Cancer Centre for their support.