AUTHORS
Liang Zhao, PhD, Julie Geyer, PhD2, Dona Kanavy, PhD3, Shannon Gray, PhD4, Rubina Islam5, Dipali Akolekar6, Srijita Kar7, Mayur Parihar, MD,PDF8, Nickhill Bhakta9, Thomas B Alexander, MD10, Jeremy Wang, PhD11
AFFILIATIONS
1 University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, United States
2 University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, United States
3 University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, United States
4 University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, United States
5 Tata Translational Cancer Research Center, Kolkata, India
6 Tata Medical Center, Kolkata, India
7 Tata Translational Cancer Research Center, Kolkata, India
8 Tata Medical Center, Kolkata, India, Tata Translational Cancer Research Center, Kolkata, India
9 St. Jude Children's Research Hospital, Memphis, Tennessee, United States
10 University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, United States
11 University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, United States
ABSTRACT
Introduction
Molecular and cytogenetic classification determines risk stratification and treatment options in pediatric acute leukemia, and typically requires a combination of flow cytometry, fluorescence in situ hybridization (FISH), karyotyping, and often targeted sequencing-based assays. Whole genome nanopore sequencing with adaptive sampling can consolidate these into a single assay. Adaptive sampling is a nanopore-specific enrichment method that selectively sequences DNA from selected loci, while off-target coverage is correspondingly sparse, making chromosome and arm-level copy number alterations difficult to resolve despite their frequency and prognostic weight. The malignant fraction is sometimes unknown or unreliable, and low purity compresses the coverage separation between copy number states. To address these challenges, we implemented a novel digital karyotyping method in NASVar – our automated leukemia variant caller – to estimate chromosome copy number and blast percentage by combining binned coverage and minor allele frequency (MAF) at enriched polymorphic sites.
Methods
Diagnostic bone marrow or peripheral blood samples were sequenced individually or multiplexed using nanopore adaptive sampling targeting a curated pediatric acute leukemia panel of 259 genes. Digital karyotyping and variant analysis were performed using NASVar with reads aligned to the T2T-CHM13v2.0 reference genome. To estimate chromosomal copy number, we first identify distinct coverage levels by constructing a histogram of aligned read counts in 1 Mbp bins, with bin width set by the Freedman-Diaconis rule, and calling candidate coverage levels at the histogram peaks. MAF was computed at 227,475 common variant sites derived from the 1000 Genomes Project and modeled per coverage level as a beta mixture. The MAF peak distinguished copy number states, concentrating near 0.5 for diploid, near 0 for haploid, and near 0.33 for triploid. Coverage and MAF together therefore separated 1n from 2n or 2n from 3n at blast percentages where coverage alone could not. The resolved baseline was used to assign copy number genome wide. When the autosomes showed more than one coverage level, blast percentage was estimated from the distance between levels and used to constrain sex chromosome copy number in cases with X or Y aneuploidy.
Results
137 samples met QC criteria (≥20x on-target coverage) with on-target coverage ranging from 20x to 127x (median 38x). Gross concordance between NASVar and clinical karyotyping, defined as agreement on aneuploidy category (low hypodiploid, <40 chromosomes; near-diploid, including low hyperdiploid, 40 to 49; high hyperdiploid, >=50), was achieved in 130/134 (97.0%) evaluable samples. Three cases were excluded from this analysis that showed diploid karyotype, but discordant gains of multiple genes by FISH, consistent with gross hyperdiploidy as indicated by our sequencing-based analysis. Whole-genome concordance was found in 108/134 (80.6%) samples, and was higher for cytogenetically normal samples (72/76, 94.7%) than for aneuploid samples (36/58, 62.1%). Twenty-two samples were gross-concordant but whole-chromosome discordant (19 aneuploid, 3 normal). NASVar called 1007/1033 chromosomes (97.5%) correctly in these samples, indicating that whole-chromosome discordance reflects a small number of specific chromosome-level disagreements.
Discussion and Conclusions
Concordance with the clinically relevant aneuploidy class was demonstrated in 97% of cases. Apparent discordance between traditional G-band karyotyping and sequencing-based copy number analysis may stem from several known phenomenon, including selection or failure of the malignant clone to grow out for G-banding, and low blast count resulting in diluted aneuploidy signature in sequenced DNA. Our three excluded cases that showed discordance between G-banding karyotype and FISH support this phenomenon. In all three, digital karyotyping showed chromosomal gains consistent with the single-gene copy numbers indicated by FISH and abnormal DNA index by flow cytometry. Some whole-chromosome and arm-level CNV discordances can also be attributed to inconclusive identification by karyotyping (marker chromosomes). Nanopore whole-genome sequencing with adaptive sampling has the potential to replace conventional karyotyping for pediatric acute leukemia diagnostics, making leukemia diagnostics faster, more affordable, and more accessible in resource-limited settings.
DISCLOSURES
Liang Zhao, PhD: No relevant conflicts of interest to declare.