A 12 Gene Signature Accurately Predicts Multiple Myeloma Progression from Monoclonal Gammopathy of Undetermined Significance

Abstract

Introduction: More than 176,000 individuals are diagnosed with multiple myeloma (MM) worldwide per year. MM is always preceded by an asymptomatic phase, monoclonal gammopathy of undetermined significance (MGUS) or smoldering multiple myeloma (SMM). MGUS is found in 2.4% of individuals ≥ 50 years. The results of long-term retrospective follow-up studies have shown a 1.0% annual risk of MGUS progressing to MM. At present, using conventional serum markers, MGUS patients can be stratified into clinical risk groups. However, these models were based on retrospective analyses of cohorts tested for a limited set of clinical variables which did not incorporate genomic properties and are inconsistent as suggested by recent studies. Our previous work has established that the GEP70 signature can identify high-risk MM patients with median survival is about 2.5 years. A robust high-risk molecular signature predicting MGUS progression is still lacking. Methods: A total of 268 MGUS patients with gene expression profiles (GEPs) were included in this study. Twenty-six progressed to MM within 10 years and were considered the high-risk group, while 242 MGUS patients not progressing to MM within 10 years were referred to as the stable group. Plasma cells were enriched from bone marrow aspirates to >85% purity using anti-CD138 immunomagnetic beads. GEP was performed with the Affymetrix U133 Plus 2.0 microarray platform, using 54,613 gene probes. Affymetrix signals MAS5.0 normalized and transformed by log-base 2 for each sample. We used 3-fold cross-validation to identify the number of genes at the top of this list, probes were ranked by q-values, which collectively maximized the concordance between risk score and MGUS progression. Gene scores were computed by subtracting the average of the expression of down-regulated genes from the average of the expression of up-regulated genes. The overall accuracy of using a gene score to predict the outcome of progression was analyzed by a receiver operating characteristic curve (ROC) and its area under the curve (AUC). MGUS progression was assessed by Kaplan-Meier curves comparing the different groups. Cox proportional hazards regression analyses were applied to evaluate the significance of the gene signature for MGUS progression. Results: Of the 54,613 probes, using q statistics at <0.01, 81 probes were significantly associated with MGUS progression. After the 3-fold cross-validation analysis, the top twelve probes that appeared in each validation and maximized the concordance between risk score and MGUS progression were included in the gene signature 12 (GS12). The up-regulated genes were MTRR (5p15.31) and BAG1 (9p13.3). The down-regulated genes were ECHDC2(1p32.3), NLRP7(19q13.42), BIRC3(11q22.2), PLEKHF1(19q12), PTPN1(20q13.13), CKAP2(13q14.3), RNF5(6p21.32), LIMA1(12q13.12), CEACAM3(19q13.2) and LOC100128079(16p12.1). The ROC analysis showed that the GS12 could accurately predict the MGUS progression in all 268 MGUS patients (AUC = 0.920). An optimal cut-point for risk of progression by the GS12 score was found to be 2.73, which identified a subset of 23 (8.6%) patients with a 10-year progression probability of 73.9%. The remainder of the 245 patients (91.4%) had a probability of progression of only 3.7%. The sensitivity and specificity were 65.4% and 97.5% respectively. Multivariate analysis showed the GS12 to be an independent risk factor for the progression of MGUS to MM. Comparing GS12 with GEP70 showed that GS12 is better at predicting progression to MM than GEP70. GEP70-defined high-risk MGUS predicted risk of progression to MM of 18% while GS12-defined high-risk MGUS raised that estimate to 73.9%. Conclusion: A gene score based upon twelve highly differentiating genes showed excellent accuracy in predicting 10-year MM progression from MGUS. Integration of gene signatures into the routine management of MGUS patients may improve the application of risk-adapted approaches to prevent progression to overt MM. We are in the process of collecting more samples and incorporating RNA-seq data to determine whether this model can be adapted to predict MGUS progression on a larger scale with current state-of-the-art technologies.