Early Detection and Monitoring of Multiple Myeloma By a Blood-Based Multi-Transcript Molecular Signature

Background:

Multiple myeloma (MM) is a hematologic plasma cell disorder with heterogeneous dissemination throughout the bone marrow (BM). Current therapeutic monitoring of MM relies on imaging, serum biomarkers and serial single-site BM biopsies, which however fail to capture the intrinsic spatial and temporal heterogeneity of the disease in a fraction of patients. Over the past years, liquid biopsy technologies, i.e. the interrogation of circulating MM cells (CMMCs) or nucleic acids from the blood or body fluids, have emerged as a promising minimally invasive strategy for the diagnostic management of MM. Here, we aimed to develop and validate the diagnostic capability of a whole blood mRNA-based MM-specific gene panel (termed MMest) to detect MM.

Methods:

MMest is derived from a transcriptional signature reflective of MM biology, which captures the expression of 25 genes. This signature was generated from peripheral blood mononuclear cells (PBMCs) acquired from MM patients and healthy controls followed by an innovative extreme gradient boosting (XGB) machine-learning algorithm after normalization of the target gene expressions to the housekeeping gene TPT1. The algorithm results in a MMest score of 0.0-100.0 with 0.0 identifying controls and 100.0 representing a perfect categorization for MM, and 20.0 is a cutoff for negative or positive detection. Sensitivity and specificity of the test were validated by spike-in experiments using common MM cell lines (MMCL: IM-9, MM1R, RPMI-8226) and solid tumor cell line controls (prostate: LnCAP, breast: BT-20, skin: A375/CRL-1619). Clinical validation was conducted in a total of n=329 samples, collected from 94 MM patients with heterogeneous disease activity, 22 MGUS patients, 68 patients with solid tumors (prostate, breast, skin) and 145 apparently healthy controls.

Results:

First, we explored the diagnostic capacity of our assay in 25 MM patients and 25 age-and gender-matched healthy controls. Diagnostic accuracy in this test cohort was excellent with an AUC of 0.96 (p<0.0001). We next defined a score value of ≥20.0 (scale: 0.0-100.0) as a clinically robust threshold to define positive scores with high sensitivity and specificity of 88% and 100%, respectively. In parallel, mRNA expression levels were found significantly elevated in MMCL vs. controls (18-38,000 fold) for all but one gene (NR4A1). To evaluate the sensitivity of our approach, we performed spike-in experiments by adding 1, 10 and 100 MM cells into 1 mL of normal blood. A clear correlation between the number of cells and the calculated score was observed. Our assay was sufficiently robust to even detect spike-ins of 1 cell/mL (MMest scores 22.9-37.2) illustrating its exquisite sensitivity. In contrast, spike-ins of solid tumor cell lines and samples from patients with solid tumors yielded negative score results defined as MMest<20.0. Specificity of MMest was additionally confirmed by correlating scores from paired whole blood samples with scores from flow-sorted CD138+ plasma cells isolated from the BM at the time of analysis (Pearson r=0.73, p<0.0001). In addition, a statistically significant correlation was noted between MMest and the percentage of plasma cells detected in the BM (r=0.29, p=0.0059) at the time of blood testing. For the BM infiltration cut-off of ±10%, MMest uncovered significantly higher scores for those patients with high vs. low plasma cell count (66.5±3.4 vs. 38.6±4.9, p<0.0001). Scores were equally increased in patients with newly diagnosed MM (n=29, 55.0±2.5) and relapsed/refractory MM (n=41, 55.3±2.0) as compared to patients with complete remission (n=24, 31.3±1.9). Healthy controls (n=145, 10.2±0.3), and MGUS patients (n=22, 42.0±2.7) showed an even lower MMest score if compared to patients with newly diagnosed MM (Mann-Whitney test p<0.0001 and 0.0003, respectively). Separately, intra- and inter-assay metrics for the MMest were 2.5±2.1% and 2.8±2.2% respectively.

Conclusions:

In this study, we developed and validated MMest as a blood-based gene expression profiling tool to monitor MM disease burden over time. Our test differentiates active MM from controlled disease and precursor states and may thereby help to overcome temporospatial sampling biases to facilitate early detection and risk stratification of monoclonal gammopathies without the need for CD138 enrichment or invasive tumor sampling.