Abstract
Introduction Multiple Myeloma (MM) precursors Monoclonal Gammopathy of Undetermined Significance (MGUS) and Smoldering Multiple Myeloma (SMM) have variable risk of progression to MM and identifying which patients may progress is challenging. Bone marrow (BM) biopsies are used for staging and identifying high-risk events associated with progression. However, they are invasive and cannot be repeated often for monitoring tumor burden. Proteome profiling of peripheral blood (PB) plasma may advance non-invasive precursor disease staging, monitoring and characterization. Here, we performed comprehensive plasma proteomic profiling across the MM disease continuum, including progressive and stable disease, to provide biological insights and identify protein-based markers of high-risk disease for improved prognostication.
Methods We executed high-throughput plasma proteomic profiling for ~3000 proteins using the Olink® Explore 3072 library and Proximity Extension Assay (PEA) technology. We profiled 462 PB plasma samples from 351 individuals, including MGUS (n=66), SMM (n=174), MM (n=49), and healthy donors (n=98). Samples from patients with progressive disease (n=32) and stable disease (n=32) with matched clinical follow-up time were also profiled; 17/32 patients with progressive disease had sequential samples from both precursor and active disease, while 15/32 patients had a precursor stage sample only. Precursor PB samples ranged 1.04-6.91 years (median of 2.33 years) prior to MM progression. T-tests, ANOVAs, and a linear mixed effect model were used to identify significant proteins across disease stages and progression status. Results were adjusted for multiple testing using the Benjamini-Hochberg Method. A subset of individuals also underwent single-cell RNA sequencing (scRNA-seq) of tumor and immune cells from paired PB/BM from the same proteomics timepoint to enable cellular mapping of signals detected in the plasma.
Results We captured high levels of plasma cell surface proteins, including BCMA, SLAMF7, CD38 and FCRL5, highlighting the utility of PEA technology to monitor soluble levels of clinically relevant targets. We analyzed functional protein networks showing stepwise dysregulation over disease progression and identified enrichment of proteins involved in immune evasion, cell motility, inflammation and cell adhesion. Correlation analysis of proteins with clinical features demonstrated BCMA and TACI levels had a strong combined positive correlation with BM plasma cell infiltration, M-protein and FLC ratio. Additional proteins, FCRL5, CD79B, MZB1, CD48, FCRLB, LY9 and QPCT, showed moderate positive correlations specifically with BM infiltration, suggesting their potential as surrogate markers for BM tumor burden and/or for improving risk prediction in routine blood-based assessments of precursor patients. We next aimed to improve the discrimination of disease states by training a machine learning-based classifier using plasma proteomic features and assigning samples to disease stages. We demonstrated 97% SMM/MM samples could be identified from healthy samples, indicating our classifier could confidently screen disease-related cases. Moreover, 85% of SMM samples were correctly classified as SMM, while misclassified cases labelled as MM exhibited early signs of progression, suggesting that the plasma proteome may provide earlier indications of evolving disease. Next, by evaluating protein levels in patients with progressive and stable SMM disease, we identified a prognostic five-protein signature that was significantly elevated at the precursor stage timepoint of patients who progressed to active MM. Validation of the signature in an external international cohort collected from three institutions confirmed that 4 of the 5 proteins were indeed significantly elevated in patients with progressive disease. Finally, integrative analysis of scRNA-seq of tumor/immune cells and plasma proteomics was used to elucidate cell-type level information of the signature proteins. Four of the proteins were predominately expressed in malignant vs. non-malignant plasma cells and/or memory B-cells, suggesting the signature partially provides a readout of malignant plasma cell biology.
Conclusion Overall, we characterized dysregulated protein networks across disease stages, developed a plasma-based classifier for accurate stage classification, and identified and validated a prognostic protein signature associated with progressive disease.
