The high frequency of somatic mutations in myelodysplastic syndromes (MDS) provides an additional biomarker modality for MDS diagnosis. This has significant potential in the investigation of cytopenias particularly in cases with <5% blasts when the diagnosis of MDS relies heavily on morphological assessment, which is limited by interobserver bias. However, consideration of mutations in the diagnostic workup has been complicated by reports of frequent somatic mutations and clonal expansion in aging healthy individuals involving genes recurrently mutated in myeloid malignancies. As a result, the presence of a somatic mutation is not a core criterion in the diagnosis of MDS. There is a clear and unmet need to understand the clinical and diagnostic implications of acquired mutations along the axis of clinical presentation from cytopenias to confirmed myeloid disease. This study aimed to investigate the clinical significance of detecting somatic mutations in an unselected cohort of cytopenic patients by correlating baseline mutation analysis with clinical phenotype and outcome.
All patient samples referred for unexplained cytopenia or suspected MDS to the Haematological Malignancy Diagnostic Service in the UK over a 2-year period (July 2014-2016) were included. Samples were analysed using established diagnostic assays (morphology, flow cytometry, cytogenetics) and in parallel subjected to targeted sequencing of 27 genes commonly mutated in myeloid malignancies. Outcome data was captured including serial blood count data, subsequent diagnoses and overall survival. A subgroup of 37 mutation negative samples underwent extended genotyping of 126 genes.
A total of 2089 samples passed QC for targeted sequencing. Of these, 538 had a confirmed diagnosis (26%), the majority of which were diagnosed with a myeloid malignancy (449/538; 83%). Mutations were commonly identified in the latter with ≥1 somatic mutation +/or karyotypic abnormality detected in 91% of patients. Importantly, mutations were also detected in 28% of non-diagnostic (ND) samples (412/1496). The spectrum of mutations was similar between ND and MDS samples, though varied in frequency. In ND samples TET2, SRSF2 and DNMT3A mutations predominated (47%, 22%, 19% respectively) while SF3B1, TET2 and ASXL1 were most frequently mutated in MDS (25%, 24%, 24% respectively). The number of mutations detected per sample and the variant allele fraction (VAF) was also significantly lower in ND samples (median no. of mutations 1 vs 2; median VAF 17.7% vs 35.1%; p<0.0001) suggesting that clonal expansion and/or additional mutations are required for overt dysplasia.
To determine the clinical significance of mutations, correlation was made with subsequent diagnoses and clinical outcome. Follow-up bone marrow samples were received in 205 ND samples and 82 of these had a confirmed diagnosis of which 61 had a myeloid malignancy. The presence of a mutation at baseline was strongly associated with a subsequent myeloid diagnosis (NDmut vs NDunmut ;13% vs 0.5%; p<0.0001) though was rarely seen in those with an isolated mutation (1 vs >1 mutation; p<0.0001). BCOR, EZH2, RUNX1 and SRSF2 mutations were found to be most predictive. The mutations in the NDmut samples also had a VAF comparable to MDS (median 31.2% vs 35.1%; p=0.14). Extended genotyping of NDunmut patients who underwent follow-up sampling identified additional rare mutations involving DDX41, SMC1A and PHF6.
Importantly the presence of a mutation impacted significantly on survival in the ND group (odds ratio=1.59; p<0.0001). The mutations with the greatest impact on survival differed from those that predicted a subsequent diagnosis and involved ASXL1, BCOR, IDH2 and TP53. Certain mutations can therefore predict increased mortality which is unrelated to having a subsequent diagnosis.
By analysing a large unselected cohort of cytopenic patients, this study has demonstrated the power of mutation analysis to predict both a subsequent myeloid diagnosis and clinical outcome in those without a confirmed diagnosis. While standard techniques will identify some of these patients over time, crucially, those at highest risk can remain undetected. Identifying these patients would ensure close clinical follow-up and provide an opportunity for early intervention. Ultimately this will facilitate the development of predictive models and refine diagnostic algorithms and this work is ongoing.
Cargo:Celgene: Research Funding. Evans:Diaceutics: Honoraria; Novartis: Honoraria. Papaemmanuil:Celgene: Research Funding.
Asterisk with author names denotes non-ASH members.