Clustering Analysis of Myeloma Clone Phenotype Is Informative for Disease Heterogeneity and Prognosis at Relapse

• Published in: Blood 2018-11, Vol.132 (Supplement 1), p.4492-4492
• Main Authors: Ahsan, Syeda Maheen, Nadal-Melsio, Elisabet, Robinson, Mark E, Syzdlo, Richard, Yebra-Fernandez, Eva, Auner, Holger W, Atta, Maria, Bua, Marco, O'Boyle, Farah, Naresh, Kikkeri, Apperley, Jane, Karadimitris, Anastasios, Chaidos, Aristeidis
• Format: Article
• Language: English
• ISSN: 0006-4971; 1528-0020
• DOI: 10.1182/blood-2018-99-117460

Background: Multiple myeloma (MM) is a remitting-relapsing malignancy with variable clinical outcome. Certain cytogenetic abnormalities, either present at diagnosis or emerged at later stages, predict for poor outcome and highlight the clinical importance of MM genetic heterogeneity. Whole genome and exome sequencing studies reveal a complex intraclonal genetic landscape, organised in linear and branching Darwinian patterns, which evolves in space and time. Clones with a more complex genetic architecture may be more fit to escape treatment and those patients are likely to have a worse clinical outcome. Clinical multicolour flow cytometry (MFC) is routinely used in MM diagnosis and detection of minimal residual disease. Previous studies have shown that MM cell subpopulations with discrete phenotypic features correspond to genetic subclones, therefore it is plausible that MFC data captures clonal heterogeneity. On that basis, we propose that clustering analysis of MM phenotypic subpopulations could be clinically relevant. Methods: We retrospectively analysed clinical MCF data at diagnosis from 44 patients eligible for autologous stem cell transplantation (AutoSCT) and 14 ineligible patients and data from 52 relapsed patients after first AutoSCT. All patients were treated between 2012 - 2018. The 8-colour MCF marker panel included CD138, CD38, CD56, CD45, CD20, CD19, cytoplasmic kappa and lambda light chains (cytLC). Data was analysed in FlowJo software and MM plasma cells were identified as CD38high, CD19-, cytLC+, within their FSC-A/SSC-A physical gate. The gated events were exported in a new fcs file. Clustering analysis was performed in Cytofkit, a R-based Bioconductor package, using the Rphenograph, Cluster-X and FlowSOM algorithms. All fcs files were subjected in the same clustering analysis, but CD56 positive and CD56 negative cases were analysed separately to offset bias from differential CD56 expression. Parameters inserted in the algorithms were FSC-A, CD138, CD38, CD45, CD20 and CD56. The number of clusters was produced by FlowSOM (k=4) and only clusters with size >1% of the total events were accepted. Results: At diagnosis, FlowSOM identified 1 (n=32, 56.1%) or 2 clusters (n=19, 33.3%) in most cases. Three clusters were found only in 5 patients (8.8%) and 4 clusters in 1 patient (1.8%). The number of clusters at diagnosis did not correlate with cytogenetic risk group or ISS. Also, the number of clusters did not predict for depth of response or rel