AUTHORS
Mehul Jani, PhD, Maria Cardenas2, Bharath Muralikrishnan3, Victor Pastor Loyola, PhD4, Tamara Westover, BS5, Anna Huskey6, Xuefei Ma7, Becky Wallace8, Masayuki Umeda, MD, PhD9, Jing Ma, PhD10, Pandurang Kolekar, PhD11, Xiaotu Ma, PhD12, Raul Ribeiro, MD13, Hiroto Inaba, MD. PHD14, Yenchun Liu, MD, PhD15, Jeffery M. Klco, Member; Director, Division of Hematopathology and Molecular Pathology16, Vidya Balagopal17, Lu Wang18
AFFILIATIONS
1 St Jude Children's Research Hospital, Memphis, Tennessee, United States
2 St Jude Children's Research Hospital, Memphis, Tennessee, United States
3 St Jude Children's Research Hospital, Memphis, Tennessee, United States
4 St Jude Children's Research Hospital, Memphis, Tennessee, United States
5 St Jude Children's Research Hospital, Memphis, Tennessee, United States
6 St Jude Children's Research Hospital, Memphis, Tennessee, United States
7 St Jude Children's Research Hospital, Memphis, Tennessee, United States
8 St Jude Children's Research Hospital, Memphis, Tennessee, United States
9 St Jude Children's Research Hospital, Memphis, Tennessee, United States
10 St Jude Children's Research Hospital, Memphis, Tennessee, United States
11 St Jude Children's Research Hospital, Memphis, Tennessee, United States
12 St Jude Children's Research Hospital, Memphis, Tennessee, United States
13 St Jude Children's Research Hospital, Memphis, Tennessee, United States
14 St Jude Children's Research Hospital, Memphis, Tennessee, United States
15 St Jude Children's Research Hospital, Memphis, Tennessee, United States
16 St Jude Children's Research Hospital, Memphis, Tennessee, United States
17 St Jude Children's Research Hospital, Memphis, Tennessee, United States
18 St Jude Children's Research Hospital, Memphis, Tennessee, United States
ABSTRACT
Most pediatric AML cases are defined by characteristic gene fusions or structural variants (SVs). Although measurable residual disease (MRD) monitoring is increasingly important, standard methods have limitations. Multiparameter flow cytometry (MFC) lacks full standardization, and ~10% of patients have no detectable leukemia-associated immunophenotype. Quantitative RT-PCR (qRT-PCR) is sensitive but low throughput and requires transcript-specific assays, limiting detection of rare or noncanonical fusions and enhancer-hijacking SVs. Next-generation sequencing (NGS) improves detection of SNVs and indels, but detecting SVs at low variant allele frequencies (VAFs) remains challenging. We developed AML-MRD, a hybrid-capture NGS panel optimized for pediatric AML that uses advanced algorithms to sensitively detect genomic alterations, particularly SVs. The panel targets coding exons of 92 genes and selected introns of 24 genes.
Six AML cell lines harboring AML-defining fusions and cooperating mutations were used to develop the assay. Serial dilutions of genomic DNA (gDNA) and culture supernatant-derived cell-free DNA (cfDNA) underwent AML-MRD NGS testing. The optimized workflow using dual-UMI NGS library preparation and ultra-deep sequencing detected all driver SVs and selected cooperating SNVs/indels at VAF as low as ~0.1% in gDNA and 0.5% in cfDNA. The clinical validation cohort included 22 patients with diverse AML-defining driver fusions: RUNX1::RUNX1T1 (2), CBFB::MYH11 (2), KMT2A-r (fusion, n=11; PTD, n=1), NUP98-r (3), ETV6::MNX1 (1), PICALM::MLLT10 (1), and ETV6::FLT3 (1). Diagnostic NGS comprised whole-genome sequencing (WGS) for 20 patients and targeted NGS for two. Fourteen patients had paired bone marrow aspirate (BMA) and peripheral blood (PB) samples collected after induction cycles 1 and 2 (EOI-1 or 2); the remaining 7 had 1-3 longitudinal follow-up BMA samples. BMA gDNA (660 ng) and PB cfDNA (2.5–5 ng) underwent AML-MRD NGS test.
NGS data processing and SNV/indel calling were performed by DRAGEN, and SVs were detected with Aperture for unguided analysis and Fuzzion2 (Fz2) for tumor-informed screening. For Fz2, a junction-specific pattern was derived from each patient’s diagnostic sample and screened across all diagnostic samples in the cohort and 40 normal BMAs. Each of the 22 patterns was detected only in its matched sample, with no false positives, demonstrating 100% analytical sensitivity and specificity for personalized MRD monitoring using Fz2 approach.
AML-MRD NGS detected patient-specific driver fusions in 28 of 47 BMA MRD samples from 22 patients. In six samples (21%), Fz2 was critical as Aperture provided weak or no fusion evidence. Among the 14 patients with available PB specimens, the same bioinformatics approaches detected driver fusions in 85% of cfDNA samples with corresponding AML-MRD NGS-positive BMA gDNA, supporting the applicability of using cfDNA when needed.
AML-MRD NGS results were compared with MFC and, when available, qRT-PCR at matched time points. Driver-fusion detection by AML-MRD NGS and leukemic-clone detection by MFC were 77% concordant, including 16 positives and 20 negatives by both methods. AML-MRD NGS alone detected the driver gene fusion in the remaining 11 samples from eight patients: six from four patients with core-binding factor AML; one sample each from three patients with high-risk fusions (NUP98::NSD1, KMT2A::MLLT10, NUP98::HOXA9); and two samples from one diagnostically challenging myeloid neoplasm with ETV6::FLT3.
Common clinical qRT-PCR assays were unavailable for 12 patients: 10 had rare fusions (e.g. NUP98-r) and two had noncanonical transcripts (one KMT2A::MLLT3 and one KMT2A::ELL). In these patients, AML-MRD NGS enabled fusion-based monitoring that was not feasible with qRT-PCR in clinical setting. Among 20 samples of the remaining 10 patients, the two methods were 90% concordant; the two discordant samples were positive only by qRT-PCR, showing 3~4-log reductions in fusion transcript levels.
In conclusion, our AML-MRD NGS assay combines UMI-based error correction with dual-engine SV analysis to directly detect DNA fusion junctions. It enables unbiased monitoring of common and rare SVs along with cooperating SNVs/indels to ~0.1% VAF. Its sensitivity, specificity, quantitative performance, and compatibility with minimally invasive cfDNA provide a robust and scalable solution for pediatric AML MRD monitoring.
DISCLOSURES
Lu Wang: No relevant conflicts of interest to declare.