Early Diagnosis of Myelodysplastic Syndromes Can be Improved By Deep Sequencing and Array Based Cytogenetics

Background

Identifying myelodysplasia (MDS) by morphological assessment, particularly in patients with <5% blasts, is problematic. There is poor inter-observer concordance in recognising dysplasia and difficulty distinguishing MDS from non-malignant conditions. This method is also limited by sample quality. To date cytogenetics has provided the only clonal marker of disease, however, recent reports of frequent driver mutations and/or structural variants detected by single nucleotide polymorphism (SNP) arrays have provided potential targets for assessment. To determine whether deep sequencing or SNP arrays can improve early diagnosis of MDS, we investigated a series of patients who developed progressive dysplasia or acute myeloid leukaemia (AML) following a non-diagnostic bone marrow biopsy taken at an earlier date for investigation of cytopenia. The objective of the study was to determine if pathogenetic abnormalities could be detected in the initial samples that were not diagnostic using conventional techniques.

Methods

A retrospective search identified 82 patients with paired non-diagnostic and diagnostic samples of whom 69 had adequate molecular material at both time points for assessment. We developed a targeted gene panel which is both time efficient and economically viable as a diagnostic tool incorporating 26 of the most commonly mutated genes in myeloid malignancies. Primers were designed using the Access Array^TMTarget-Specific Primers design service (Fluidigm). DNA either stored at the point of referral or extracted from archived slides was subjected to SNP array analysis (Illumina CytoSNP-12) and targeted deep sequencing (Illumina MiSeq). Initial read alignment and variant calling were performed using MiSeq reporter (Illumina) and variants were annotated using the Ensembl variant effect predictor. Following exclusion of known polymorphisms and correlation with the COSMIC database, mutations were confirmed using Sanger sequencing. CytoSNP data was visualised using KaryoStudio software (Illumina).

Results

A driver mutation, with or without a structural variant, was identified in 91% (63/69) of non-diagnostic samples and this increased to 94% (65/69) at the point of diagnosis.

Crucially the majority of mutations identified in the non-diagnostic sample were likely oncogenic being either truncating variants (nonsense mutations, indels) or a mutation previously reported in the literature. There was a predominance of mutations involving epigenetic regulators and spliceosome genes at the earlier time point with mutations in TET2, SRSF2 and ASXL1 being detected in 38%, 26% and 20% of patients respectively. The spectrum of mutations identified in the non-diagnostic sample mirrored that reported in large cohorts of MDS patients with the exception of SF3B1 which was identified in only 3 patients (4%). This likely reflects the relative ease in identifying ring sideroblasts morphologically.

The median time between samples was 476 days (range 19-2484). A total of 26 patients (38%) acquired mutations between samples. This occurred most commonly in those patients who developed AML (15/21 patients; 71%) with a resultant mutation rate of 100% in this disease subgroup. Transcription factor and cell signalling genes were the most commonly acquired mutations between samples (26/35 mutations; 74%). Importantly, with the exception of 1 patient, all those who progressed to AML had an abnormality detected on the non-diagnostic sample.

Structural variants (copy number variation or loss of heterozygosity) were identified in 23% (16/69) of non-diagnostic samples increasing to 48% (33/69) at the point of diagnosis. However, these consistently occurred in the presence of a driver mutation.

Conclusion

This study has shown that MDS patients with ambiguous morphology at the time of presentation possess very frequent driver mutations that can be readily demonstrated using a limited, cost-effective sequencing panel. Mutational analysis may represent an objective and more reliable method of identifying patients with early stage disease. The paucity of structural variants along with the frequent demonstration of epigenetic regulators and spliceosome mutations in the non-diagnostic samples provides further insight into disease pathogenesis.