Patient characteristics
We included a total of 56 patients (12 MGUS, 11 SMM, 15 MM within 6 months of diagnosis, 18 RMM) and 12 normal controls. Their baseline characteristics are shown in the table. There was no significant age difference between the median age of cases compared to that of controls (66 years vs 61 years, respectively, p = 0.18). High risk patients were overrepresented in this study (53%), which was reflected in the short time to progression for newly diagnosed and relapsed patients (median of 15 and 5 months, respectively).
The BM bone proteome is a diverse ecosystem of proteins that varies significantly across patients.
We identified a total of 1951 distinct proteins across all samples (raw protein MS1 counts.xlsx, supplementary material). There was no difference in the total proteomic content (total number of normalized spectral counts for all proteins in a sample) according to diagnosis type (data not shown). WebGestalt analyses of the top pathways associated with the entire proteome and the top 5% most abundant proteins are shown in Fig. 1. Proteins involved in structural support (“ECM organization” pathway) were, not surprisingly, overrepresented. Notably, proteins involved in protein translation machinery and immunity (the majority of which were immunoglobulins) were also overrepresented, suggesting that the bone tissue is a metabolically and immunologically active tissue. To visualize major patterns in the data across patients we performed a UMAP analysis of all patients using their entire proteome and normalized spectral counts as semi-quantitative measures of abundance. No significant grouping of patients was apparent (supplemental Fig. 2), suggesting significant variability in the BM bone proteome even within the same diagnostic categories (e.g., MGUS or SMM or MM). To identify which proteins were responsible for most of the variability across patients we performed PCA. The first 2 principal components explained 15.1% and 6.4% of the variability in the data, respectively. We then performed WebGestalt pathway analyses in the top 5% of proteins with the highest and lowest loadings, respectively for each principal component (supplemental Figs. 3,4). This demonstrated that proteins with high loadings within principal component 1, were mostly proteins involved protein translation whereas though with low loadings were proteins involved in the activation of the complement cascade, keratinization, and ECM formation. Similarly, proteins with high loadings within principal component 2, were involved in ECM formation and those with low loadings were involved in protein translation.
Normal controls and MGUS patients have a distinct proteome compared to those with higher plasma cell burden states
Given the high level of variability observed and the limited number of cases within each group, we combined normal controls and MGUS cases together (i.e., low plasma cell burden) and compared them to SMM and MM cases (i.e., high plasma cell burden) to increase statistical power. The results of a WebGestalt analysis are shown in Fig. 2. The abundance of proteins involved in ECM formation pathways were decreased whereas proteins involved in gamma-carboxylation pathways were increased in SMM/MM. Amongst the proteins (supplemental material) most decreased in SMM/MM were immunoglobulin genes, collagen 4 and tissue inhibitor of metalloproteinase 3 (TIMP3). Among the proteins most increased in SMM/active MM were amyloid precursor protein (APP), ectonucleotide pyrophosphatase/phosphodiesterase 1 (ENPP1), Lectin, Mannose Binding 2 (LMAN2), Marginal Zone B And B1 Cell Specific Protein (MZB1) and X-Prolyl Aminopeptidase 1 (XPNPEP1). A GSEA analysis largely confirmed the above findings (Figs. 3 and 4, full list of proteins within each pathway is provided as supplemental data).
Within the most upregulated pathways in the MGUS/control groups were immunoglobulin proteins (dominating all pathways except those relating to keratinization or ECM formation). Protein pathways upregulated in SMM/MM cases again included those involved in gamma-carboxylation, endoplasmic reticulum (ER) stress response, and protein trafficking within cells. All differentially expressed proteins and a ranked list used for WebGestalt and GSEA analyses, respectively, are included as supplemental data. No significant pathway differences were identified between patients with newly diagnosed MM and relapsed or smoldering MM. We also compared the following groups: males versus females, patients receiving osteoclast inhibitors (bisphosphonates or denosumab) versus not, high risk versus non-high risk active MM patients (per IMWG criteria) and patients with a bone fracture within 6 months from sample collection but found no differences.
We have previously published CyTOF data from immune cells, Luminex data from BM plasma proteins, and transcriptomic data from malignant plasma cells collected from the same 15 patients with MM diagnosed within 6 months and all but one RMM patient (n = 17) (12). We hypothesized that bone proteins having a high correlation with any of these other BM components (immune cells, BM plasma proteins or malignant cell genes) would be more likely to coregulate or influence each other. We interrogated a correlation matrix of bone proteomic data with the previously reported CyTOF, Luminex and transcriptomic data from BM immune cells, plasma proteins and malignant plasma cells respectively, and, using a cutoff of Pearson’s R > 0.9 (R2 > 0.8), however we identified no significant correlations.