Long-read profiling of structural variants reveals mechanisms of chemo-resistance and prognostic heterogeneity in acute lymphoblastic leukemia Free

Background Acute lymphoblastic leukemia (ALL) is the most prevalent malignant tumor among children and adolescents, with poor prognosis in adults, as evidenced by a complete remission rate of <40%. Chemoresistance remains a major challenge in ALL treatments, with survival rates for resistant patients falling below 30%. While molecular subtyping based on next generation sequencing (NGS) is a key prognostic indicator of B-cell precursor ALL (BCP-ALL), chemotherapy responses often do not align with expectations based on molecular classification. This discrepancy arises, in part, because certain cases of ALL are driven by structural variants (SVs), which NGS, with its short-read sequencing approach, struggles to detect particularly large insertions, deletions, and complex rearrangements. In this study, we performed long-read sequencing (third-generation sequencing) to identify functional SVs that were missed by NGS, aiming to elucidate their roles in chemo-resistance and prognostic heterogeneity within defined BCP-ALL subtypes.

Methods A panel of patient-derived xenografts (PDXs) were established, representing major BCP-ALL subtypes, including MEF2D-rearranged (n=3), BCR::ABL1/-like (n=3), TCF3::PBX1 (n=2), KMT2A-rearranged (n=2), PAX5::ETV6 (n=1), TCF3::ZNF384 (n=1) and B-other (n=2). Long-read whole-genome sequencing was performed using ONT PromethION platform. SV calling was carried out using an in-house bioinformatics pipeline that integrated 5 publicly available SV callers (cuteSV, Severus, Sniffles, SVIM, and SVision). Germline SVs were removed by comparison with the PGG.SV and DGV databases. In addition, HiPore-C (high-throughput Pore-C) was employed to explore alterations in topologically associated domains (TADs) related to SVs. Candidate SVs were annotated and subjected to functional validation in molecular biology studies.

Results 5,468 to 8,097 SVs per sample (average 6,557) were identified across 14 BCP-ALL PDXs. Consistent with prior reports, deletions and insertions accounted for the majority of identified SVs, with 89.7% and 91.2% being <1 kb, respectively. In contrast, inversions and duplications exhibited a broader size distribution with a substantial proportion beyond 10 kb (38.5% and 11.7% respectively). Deletions and duplications were enriched in transcriptionally active and GC-rich regions, with breakpoints closer to TAD boundaries. Translocations and insertions, while also enriched at TAD boundaries, exhibited an opposite coverage pattern of GC content and more frequently associated with transcriptional repression. Inversions, however, showed limited association with transcriptional activity.

Next, we interrogated intra-subtype heterogeneity of SVs in ALL. Two MEF2D::BCL9 PDXs, although similar in transcriptional profiles, exhibited distinct clinical prognoses and glucocorticoid (a standard of care drug for ALL) sensitivities in vitro and in vivo. Long-read sequencing pinpointed SVs associated with distinct drug response of the two PDXs. The glucocorticoid-resistant PDX demonstrated a more complex MEF2D::BCL9fusion, with MEF2D duplication and an insertion of a 1.6 kb downstream non-coding sequence into the fusion gene. HiPore-C captured aberrant chromatin interactions in these complex SVs. In addition, intragenic amplificationofPAX5 was found in a previously unclassified PDX (B-other subtype), which diminished PAX5 binding at glucocorticoid response elements (GREs) conferring glucocorticoid resistance. In contrast, the glucocorticoid-sensitive PDX exhibited PAX5 and glucocorticoid receptor (GR) colocalization at GREs. Moreover, beyond standard chemotherapy, PDXs harboring deletion at MTAP gene locus showed increased sensitivity to MTA-cooperative PRMT5 inhibitors via a synthetic lethal effect, providing a novel therapeutic strategy for MTAP-deficient ALL.

Conclusion Overall, this study identified SVs missed by short-read NGS in BCP-ALL and demonstrated the critical role of these SVs in intra-subtype heterogeneity of drug response across various ALL subtypes. Our findings suggest that SV-based stratification could enhance prognostic accuracy and guide therapeutic decisions, offering a novel precision-medicine framework for ALL that transcends traditional subtype definitions.