Cytokine score for prognostication of newly diagnosed multiple myeloma Free

Introduction Multiple myeloma (MM) is clinically heterogeneous, and therapeutic response is profoundly influenced by the complex cytokine milieu of the bone marrow microenvironment. Unlike most cancers, MM staging incorporates non–tumor-burden markers such as serum albumin, β₂-microglobulin, and LDH, which are prognostically useful but biologically nonspecific. Although individual cytokines (e.g., IL-6, IL-17, TGF-β) have been linked to disease activity, a comprehensive, systems-level assessment comparing an exhaustive list of cytokines, chemokines, and growth factors across large patient cohorts is lacking. Here, we present evidence for an actionable cytokine panel, using gene expression–derived cytokine bioactivity as a surrogate measure.

Methods CD138 enriched bone marrow plasma cell RNA-seq, Seq-FISH, and clinical data were obtained from the MMRF CoMMpass study. Data were TMM-normalized, log₂(x + 1) transformed, and mean-centered. Bioactivity scores for 64 cytokines, chemokines, and growth factors were inferred using CytoSig (National Cancer Institute). CytoSig leverages transcriptomic data from over 20,000 samples treated with individual molecules to model gene expression changes of cytokine regulators, targets, and response elements to provide z-scores.

Reverse validation of bioactivity between samples scoring top and bottom quartiles of each cytokine was done using pathway analysis via Gene Set Enrichemnent analysis platform that verifies bioactivity across 7561 well-studied gene sets.(MSigDB C2 2025). Cytokines with upregulation of associated pathways were considered for downstream analyses (Criteria: Normalized enrichment score > 1). Genes were ranked by sign(logFC) × –log₁₀(p), identifying 44 significantly enriched pathways.

Univariate Cox regression identified prognostic cytokines. To account for variable correlation and improve prediction, Elastic Net Cox regression (α = 0.5) was used, combining LASSO (L1) and Ridge (L2) penalties. For biological validation, pathway analysis was done between samples with high vs low serum C-Reactive Protein(CRP). Survival kinetics and clinical characteristics were analyzed for the derived model.

Results A total of 767 NDMM patients were included (median age 63 years, range 27–93); median overall survival (OS) was 44.4 months (IQR 70.9). Revised International Staging System (R2-ISS) distribution was: Stage I 14.7%, II 24.0%, III 34.7%, and IV 6.4%. Based on the International Myeloma Society–International Myeloma Working Group (IMS–IMWG) definition, 32.2% of patients were classified as high-risk.

Supervised analysis identified 30 univariate prognostic cytokines. Elastic Net retained BMP-7, IGF-1, IL-3, NRG1, IL-17A, IL-2, IL-21, and TGF-A (positive coefficients), and IL-4, IL-18, BMP-4, palmitic acid, DHT, and BMP-2 (negative coefficients). In sensitivity analysis, after removing cytokines not passing GSEA validation, the cytokine score (CS) was established using BMP-7, IL-3, TGF-A, IL-17A, BMP-2, IL-18, BMP-4, and IL-4. High serum CRP–associated gene expression was linked to upregulation of IL-17 and BMP pathways and downregulation of IL-4 signaling. Patients with CS ≥ median had higher β2M (p < 0.001) and LDH (p = 0.005); creatinine (p = 0.07) and albumin (p = 0.6) were not significant. CS was higher in patients with del(17p), 1q21 gain, high-risk IGH, t(8;14), and del(1p), and lower in those with hyperdiploidy (all p < 0.01). CS achieved a C-index of 0.67. Median OS was 58.7 months in the high-CS group vs. 102.5 months in the low-CS group (p < 0.001). Adding continuous CS to a model with continuous β2M + continous albumin improved C-index from 0.65 to 0.72. ISS improved from 0.62 to 0.68; R2-ISS from 0.65 to 0.69. A multivariate model with β2M ≥ 5.5, del(1p) + 1q gain, t(4;14), deep del(1p), and del(17p) improved from 0.63 to 0.68 with CS. Discussion These findings require functional validation in patient samples through direct measurement of cytokine levels. BMP-7 remains understudied in MM. Elevated TGF-α, IL-3, and IL-17 promote angiogenesis, bone resorption, and immune dysfunction, while BMP-2 and BMP-4 act as tumor suppressors. IL-18 is being utilised to boost T-cell responses, and IL-4 is consistently reduced in MM.

Conclusion This study provides evidence for a robust, clinically translatable cytokine, chemokine, and growth factor panel. IL-17, IL-4, and BMPs emerged as consistently regulated across all analyses. CS is associated with high-risk disease and inferior survival.