Detection of KMT2A Partial Tandem Duplication in Myeloid Samples By Pacbio® Long Read Sequencing

Short read next-generation sequencing (NGS) targeted panel testing is standard of care for myeloid diseases such as myelodysplastic syndromes (MDS) and acute myeloid leukemia (AML). However, several clinically important variants are challenging to detect using NGS panels including KMT2A partial tandem duplication (KMT2A-PTD), TP53 biallelic mutations (TP53bi), and FLT3 internal tandem duplications (FLT3-ITD). In a classification scheme for MDS utilizing variants in 31 genes, the presence or absence of these 3 variants had the strongest negative impact upon MDS prognosis. KMT2A-PTD is an in-frame duplication and requires breakpoint amplification using RT-PCR for accurate confirmation. Variant allele frequency (VAF) cutoffs have been suggested for the detection of TP53bi but this inference can be confounded by tumor clonality and purity. FLT3-ITD requires complex detection algorithms in short read sequencing data. In this study, we assess whether long read sequencing can detect these three variant classes in a single reaction.

Proof-of-concept data were generated using one AML and two acute lymphoblastic leukemia (ALL) cell lines. These cell lines were also diluted with healthy cell lines to create mixtures at 20%, 10%, and 5% tumor. One eosinophilic leukemia cell line (EOL-1) with known KMT2A-PTD, and 5 clinical samples with FLT3-ITDs were additionally used for proof-of-concept.

Using AML or MDS clinical samples, 12 candidate KMT2A-PTD, 7 TP53bi and 6 additional FLT3-ITD samples were identified using Illumina® short read sequencing. KMT2A-PTD and TP53bi samples were analyzed by digital multiplexed ligation-dependent amplification (dMLPA) to assess copy number aberrations (CNAs) in KMT2A and TP53. Candidate samples were then subjected to library preparation and hybridization enrichment before long read sequencing on a PacBio Sequel II using one of two custom probe designs - a 2x tiled 141 gene panel with an identical region of interest to the panel used for short read sequencing, or a 0.1x tiled, 59 gene long read panel custom for structural and other complex variants common in hematologic malignancies. Single nucleotide variations (SNVs) and insertion/deletions (indels) were detected using a combination of Mutect2 and Integrated Genome Viewer (IGV). KMT2A-PTD was detected using a combination of Sniffles2 and a custom algorithm to detect breakpoint-supporting reads. KMT2A-PTD breakpoints and the inferred genomic copy numbers (CNs) were compared to dMLPA. Reads containing TP53bi and FLT3-ITD were detected using IGV. All TP53bi samples had sufficient read lengths to confirm without the requirement for haplotyping.

For the proof-of-concept study, long read sequencing confirmed KMT2A-PTD in the expected region of EOL-1 and one TP53bi in an ALL cell line. 43 out of 44 expected SNV/indels were detected in the cell lines dilution series. The one SNV not detected had a low expected VAF of 0.63%. 5 of 5 FLT3-ITDs were detected in the proof-of-concept clinical samples with VAFs of 31.9%-80.2% and lengths of 21-69 bp. For the long read panel design study, 12 of 12 KMT2A-PTDs were detected with at least 10 breakpoint supporting reads and dMLPA CNs ranging from 1.38 - 2.78. 6 of 7 TP53bi were confirmed by visual inspection of overlapping reads and showed VAFs from 24.0% - 52.5%. The 1 TP53bi not detected was a result of low read depth with each mutation only found in 1 CCS read. 6 of 6 FLT3-ITD were detected with a VAFs from 4.0% - 73.7% and lengths from 21 - 78 bp.

This study shows that these 3 challenging mutations can be detected using long read sequencing in a single reaction using minimal alterations to established clinical NGS workflows, allowing this method to potentially be used as a reflex confirmation test or as a stand-alone NGS panel. KMT2A-PTD regions detected agreed with dMLPA and TP53bi mutations up to 2303 bp apart were identified. Coverage uniformity and read lengths observed suggest that one could additionally utilize phasing to identify TP53bi throughout the length of TP53 and to detect the presence of TP53 CNAs by relative read coverage. FLT3-ITDs were observed in IGV without a dedicated algorithm using both short and long read panel designs. In conclusion, this study shows the potential of long read sequencing for comprehensive detection of multiple challenging variant classes.