Integrated multidisciplinary approaches to improve prognosis prediction and preclinical in vitro modelling of high-risk multiple myeloma

Despite therapeutic advances, a subset of multiple myeloma (MM) patients with high-risk (HR) disease continues to experience poor outcomes. This prospective monocentric study aimed to (i) improve MM patients’ stratification through patient’s omics data collection and analysis; (ii) generate a preclinical model of HR MM to study primary and tumour microenvironment (TME)-mediated resistance to treatment. Patients with smoldering (SMM), newly diagnosed (NDMM), and relapsed/refractory MM (RRMM) referring to IRST were enrolled. Molecular profiling included SKY92 gene expression signature, circulating soluble BCMA (sBCMA), cell-free DNA methylation (cfDNAmeth), and low-pass whole genome sequencing for copy number analysis. Apparent Diffusion Coefficient (ADC) and Relative Fat Fraction radiomic features were extracted from Whole-body MRI (WB-MRI) profiling. Bone marrow–mimicking 3D scaffolds (BM3Ds) were engineered by a crosslinking and a freeze-drying process combining collagen type I and hydroxyapatite. SKY92-defined HR (SKY92 HR) patients increased from SMM to NDMM and RRMM and were characterized by a poor outcome compared to SKY92 standard risk patients (SKY92 SR) in NDMM. SKY92-based virtual FISH outlined high concordance with standard FISH for t(4;14) and gain(1q). High levels of plasmatic sBCMA and cfDNAmeth were associated with worse prognosis in NDMM. ADC and rFF varied across disease stages and showed trends of association with molecular stratification. Multi-omics integration identified latent factors strongly linked to outcome. BM3Ds reproduced bone marrow architecture and supported the growth of endothelial and stromal cell lines beyond MM cells. MM cells growth in BM3Ds showed higher IC50 for bortezomib and lenalidomide compared to standard cell culture. A model of primary resistant MM cells to the clinically relevant compound lenalidomide was generated by continuous exposure of MM cells to this molecule to further implement our preclinical model of HR MM. This study supports integrated multi-omics and biomimetic 3D modelling as promising strategies to improve HR identification and preclinical characterization.