Improving predictive accuracy in multiple myeloma using a plasma cell profile derived from single-cell RNA sequencing

Multiple myeloma (MM) shows inherent clinical and biological heterogeneity, leading to variable treatment responses and outcomes. The complex molecular landscape of MM makes precise risk stratification through clinical genetic testing difficult. Thus, identifying better biomarkers is essential to enhance existing stratification methods and guide personalized therapy decisions. Here, we systematically analyzed the intratumor heterogeneity of tumor cells from 12 newly diagnosed MM patients with different outcomes at single-cell resolution, especially those with an overall survival of less than 2 years, considered extremely high-risk in the real world. Among the eight heterogeneous tumor cell subclusters in these patients' myeloma cells, a particularly aggressive subset was discovered, characterized by severe chromosomal instability, high-level drug resistance, and high-risk genes. Survival analysis indicated that a high rate of this aggressive cell subset was associated with poor outcomes of the patients. We identified seven genes (LILRB4, CD74, TUBA1B, CCND2, HIST1H4C, ITGB7, and CRIP1) with extremely high expression within this subset of aggressive myeloma cells. Multivariate Cox analysis showed that the seven-gene signature score was the worst factor for patients' outcome independently of aberrant cytogenetics and International Staging System stage. We then established an integrated risk stratification model combined with the seven- gene signature score. This model significantly improved the risk discrimination capabilities, especially in distinguishing the ultra-high-risk myeloma patients with the worst outcome in our cohort, and was validated in five independent datasets of MM patients. We further devised a simple digital polymerase chain reaction method for feasible quantification of the seven-gene signature, which still significantly differentiated the survival of MM patients and has considerable value for clinical application. Overall, this integrated risk-scoring model derived from single-cell RNA-sequencing data was significantly associated with a more advanced stage of myeloma, facilitating guided risk-adapted treatment strategies for such ultra-high-risk patients.