A common gene expression signature across multiple myeloma disease stages to identify high-risk of disease progression.

Abstract

7560 Background: An accurate and timely identification of high-risk (HR) disease is crucial to maximize therapeutic opportunities nowadays available for multiple myeloma patients. Methods: In this study, 105 consecutive patients with smoldering (SMM = 33), newly diagnosed (NDMM = 42) and relapse/refractory (RRMM = 30) myeloma were enrolled. Bone marrow samples were processed by SKY92 assay (Skyline Dx) to detect HR gene expression (GE) signature and cytogenetics according with manufacturer guidelines. Statistics were performed using GraphPad Prism 8 and R (v4.4.2). Two-sided statistical tests were used, with significance of P < 0.05 and Benjamini–Hochberg method for false discovery rate (FDR) correction. Results: Based on SKY92 quality requirements, this test was feasible in the 81.1% of SMM, 92.8% of NDMM and 80.0% of RRMM cases, showing an increased detection of HR SKY92 signature frequency from SMM (18%) toward NDMM (43.5%) and RRMM (67%, P = 0.0018). By comparing SKY92 virtual FISH with available standard FISH data we observed a high performance of t(4;14), t(14;16/20) and gain1q detection with an accuracy [95%(CI)] of 97.9 (93.9-100), 85.2 (71.8-98.6) and 87.2 (77.7-96.8) respectively, decreasing to 66.7 (49.8-83.5) for del17p suggesting that this chromosomal aberrancy can be less likely recapitulated by GE-based projections. NDMM HR SKY92 patients showed a worst overall survival compared to patients with SKY92 standard risk (SR, P = 0.0079), a trend maintained in RRMM with a nearly significant statistics due to the cohort sample size ( P = 0.093). These results suggested that HR patients could share a transcriptional trademark across disease stages. This hypothesis was tested by comparing SKY92 HR NDMM and RRMM versus SR GE data with the following observations: (i) relative risk-related clusterization by a principal component analysis (PCA) and (ii) modulation of canonical pathways frequently altered in cancer by a gene set enrichment analysis (GSEA). To test the predictivity of GE-based approaches in the asymptomatic phase of myeloma we projected GE data of SMM on the PCA bi-dimensional space previously generated with NDMM and RRMM data and we observed a partial overlap of samples based on SKY92 risk profiles across MM stages. Interestingly, a GSEA comparing SKY92 HR vs SR SMM outlined deregulated gene pathways associated with inflammation and aggressive disease. Consistently, available follow up data show a progression rate of 60% for HR SKY92 SMM compared with 0% of the SR group (P = 0.173). Conclusions: Our findings support the use of GE-based tools for MM patient’s risk profiling and HR-related cytogenetic detection if FISH is not available. Furthermore, it provides evidence that HR patients share transcriptional traits across asymptomatic and symptomatic phases consistent with recent findings underlying common genomic features shared by a portion of SMM with MM.