Single‐cell RNA sequencing infers the role of malignant cells in drug‐resistant multiple myeloma

摘要

Dear Editor, Using single‐cell RNA‐sequencing (scRNA‐seq), we identified special populations that might be involved in the progression of drug resistance and various poor prognostic biomarkers in multiple myeloma (MM). Although the overall treatment outcomes of MM have been improved, ~(1) challenges still exist in relapsed MM due to the lack of effective drugs and predictive biomarkers. ~(2) , ~(3) ScRNA‐seq has been applied to unbiasedly identify the cellular heterogeneity and novel biomarkers. ~(4) , ~(5) To unravel the tumour microenvironment dynamics associated with carcinogenesis and survey cellular heterogeneity in MM, we performed scRNA‐seq on bone barrow cells from 3 primary (newly diagnosed), 1 recurrent, 3 drug‐resistant MM and 1 healthy donor (Figure S1A and C ; Table S1 ; Supplementary materials). After rigorous quality control (Figure S1B ; Supplementary materials), 52?793 cells were obtained for further analysis based on known marker genes and differentially expressed genes (DEGs) (Table S2 ). Notably, we retrieved B cells, T cells, NK cells, myeloid cells, DCs, erythrocytes, haematopoietic progenitor cells (HPCs) (Figure? 1A ) and key marker genes in these clusters, that is NKG7 for NK cells, CD34 for HPCs and MZB1 for B cells (Figure? 1B ). Further, we found B, T and myeloid cells were highly enriched in drug‐resistant MM (Figure? 1C and E ), suggesting that B cells, T cells and myeloid cells might be involved in the progression of drug resistance in MM. Our results showed substantial variation in B cells proportion in MM compared with control group (Figure? 1D ). FIGURE 1 Single‐cell transcriptional profiles in multiple myeloma (MM). (A) Identification of cellular types using single cells of human multiple myeloma. (B) Dot plot showing the canonical marker genes in 7 cellular types. (C) UMAP plot showing the distribution of cells in different conditions including control group, primary MM, recurrent MM and drug‐resistant MM. (D) Cell type proportion of all patients with or without MM. (E) Cell type proportion of cellular types in different conditions including control group, primary MM, recurrent MM and drug‐resistant MM. UMAP: Uniform Manifold Approximation and Projection; Pt: patients We then focus on B cells, T cells and myeloid cells using unsupervised clustering and Uniform Manifold Approximation and Projection (UMAP). ~(6) First, B cells were clustered into 12 clusters or four subgroups according to gene expression of MS4A1, CD19, SDC1 and MKI67 (Figure? 2A and B ; Table S3 ). Clusters 1, 3, 5 and 7 were predominantly enriched in drug‐resistant group, and cluster 1 and 7 are proliferating/cycling cells with high expression of MKI67 (Figure? 2B and C ), suggesting that B cells proliferation may contribute to the progression of drug resistance of MM. Besides, we calculated the large‐scale chromosomal copy number variations (CNV) to distinguish the malignant B cells from normal cells. ~(7) Drug‐resistant MM showed remarkably highest CNV levels among groups (Figure? 2D ). Interestingly, cluster 5 malignant B cells that is significantly enriched in drug‐resistant MM ( p ?&?.05, χ ~(2) test) (Figure? 2C ) exhibited high CNV levels (Figure? 2D and E ), suggesting that cluster 5 B malignant cells were the major source of malignant cells in drug resistance. Further, we focused on cluster 5 and identified DEGs through the comparison of drug‐resistant MM versus primary MM (Figure? 2F ; Table S4 ) and found several marker genes in this subpopulation including CD27. By ordering these cells to reconstruct pseudo‐time trajectories, ~(8) we observed cluster 5 cells bifurcated to 2 branches, the drug‐resistant MM and the relapsed MM, suggesting distinct cellular differentiation paths of these two MM stages (Figure? 2G ). Importantly, we identified novel genes of CCL4, TNFRSF17, LMAN2 and MZB1 that were positively correlated with drug resistance (Figure? 2H ). Concordantly, the patients with higher expression of these genes [i.e., CCL4 ( p ?=?.017), TNFRSF17 ( p ?=?.0082), LMAN2 ( p ?=?.025) and MZB1 ( p ?=?.016)] had poorer prognosis than those with low expression (Figure? 2I ). This data further supports that cluster 5 cells expressing CCL4, TNFRSF17, LMAN2 and MZB1 contribute to drug resistance. FIGURE 2 Transcriptional landscape of B cells in MM. (A) UMAP plot suggesting identification of cellular subtypes in B cells. (B) Scatterplot showing expression of marker genes in B cells; (C) Cell type proportion of cellular subtypes in B cells in different conditions including control group, primary MM, recurrent MM and drug‐resistant MM. (D) Heat map showing large‐scale CNVs of B cells from 8 patients. The red colour represents high CNV level and blue represents low CNV level. (E) The distribution of malignant cells in different conditions including control group, primary MM, recurrent MM and drug‐resistant MM. (F) Differentially expressed gene profiles along malignant progression. (G) Pseudo‐time