Historically diagnosis and prognosis of myeloid disorders including acute myeloid leukemia (AML) have been determined using a combination of morphology, immunophenotype, cytogenetic and more recently single gene, if not single mutation, analysis. The introduction of NGS technology has resulted in an explosion in the quantity of mutation data available. However, the feasibility and utility of NGS technology with regards to decision-making in routine clinical practice of myeloid disorders is currently unknown. We therefore developed an advanced NGS tool for simultaneous assessment of multiple myeloid candidate genes from low amounts of input DNA and present clinical utility analysis below.
We designed a targeted resequencing assay using a TruSeq Custom Amplicon panel with the MiSeq platform (both Illumina) consisting of 341 amplicons (~56 kb) designed around exons of genes frequently mutated in myeloid malignancies (ASXL1, ATRX, CBL, CBLB, CBLC, CEBPA, CSF3R, DNMT3a, ETV6, EZH2, FLT3, HRAS, IDH1, IDH2, JAK2, KIT, KRAS, MPL, NPM1, NRAS, PDGFRA, PHF6, PTEN, RUNX1, SETBP1, SF3B1, SRSF2, TET2, TP53, U2AF1, WT1 & ZRSR2). Filtering, variant calling and annotation were performed using Basespace and Variant Studio (Illumina) with additional indel detection achieved using Pindel. A cohort of samples previously characterised with conventional techniques was used for validation and the lower limit of detection established using qPCR. Post-validation, DNA from 152 diagnostic blood or bone marrow samples from patients with confirmed or suspected myeloid disorders; both AML (n=46) and disorders with the potential to transform to AML i.e. myelodysplasia (confirmed n=54, suspected n=10) and myeloproliferative neoplasms (n=42), were analysed using this assay. To gather clinical utility data we developed a reporting algorithm to feed back information to clinicians; only those variants with a variant allele frequency (VAF) of >10% and described as acquired in publically available databases were reported with the exception of novel mutations predicted to result in a truncated protein. Further utility data was obtained using published mutation algorithms to determine the proportion of patients in whom mutation data altered prognosis.
In the validation cohort, initial concordance for detection of clinically significant mutations was 88% rising to 100% once Pindel was used to identify FLT3 ITDs. The lower limit of detection was 3% VAF, and mean amplicon coverage was 390 reads. Using our reporting algorithm 66% of patients in the post-validation cohort had a suspected pathogenic mutation relevant to a myeloid disorder, rising to 74% in patients with confirmed diagnoses. The median number of reported variants per sample for all diagnoses was one (range 0-6). When mutation data for patients with AML with intermediate risk cytogenetics was analysed using the algorithm of Patel et al (N Engl J Med. 2012;366:1079-1089), 4/22 (18%) moved into another risk category. A further two patients had double CEBPA mutations, improving their prognosis. Identification of complex mutations in KIT exon 8 in 2/6 patients with core binding factor AML resulted in more intensive MRD monitoring due to the increased risk of relapse. Interpretation of mutation data for patients with confirmed myelodysplasia using the work of Bejar et al (N Engl J Med. 2011;364:2496-2506) revealed 13/54 (24%) had a high risk mutation independently associated with poor overall survival. 2/8 (25%) patients with chronic myelomonocytic leukemia and 1/12 (8.3%) patients with primary myelofibrosis had high risk ASXL1 exon 12 mutations, independently associated with a poor prognosis. Among suspected diagnoses confirmatory mutations were found in 2/19 (11%), while the absence of mutations reduced the probability of myeloid disease in 11/19 (58%), in some cases sparing elderly patients invasive bone marrow sampling. A further 20 patients had clinically relevant mutations.
The NGS Myeloid Gene Panel provided extra information to clinicians in 57/152 patients (38%) helping inform diagnosis, individualize disease monitoring schedules and support treatment decisions. The targeted panel approach requires rigorous validation and standardisation in particular of bio-informatics pipelines, but can be adapted to incorporate new genes as their relevance is described and will become central to treatment decisions.
No relevant conflicts of interest to declare.
Asterisk with author names denotes non-ASH members.