Malignant clonal evolution drives multiple myeloma cellular ecological diversity and microenvironment reprogramming

Single-cell landscape of control BM

From scRNA-seq, 64,718 cells were obtained, including 5238 from the control donor BM aspirate and 59,480 from MM patients after quality control (Fig. 1A, Supplementary Table 3). The t-SNE approach captured 16 cell types, including hematopoietic lineages, myeloid lineage cells, and lymphoid cells, all of which belong to common cell groups in the BM (Fig. 1B), consistent with the established phenotypic characteristics of immune cells (Fig. 1C). The pseudotime trajectory of cell development revealed the continuous process of myeloid development and hematopoiesis, that hematopoietic stem cells (HSCs) differentiate into promonocytes to monocytes to finally dendritic cells (DCs), and that HSCs differentiate into erythroblasts to post erythroblasts (Fig. 1D). The pseudo-sequential differentiation trajectory was negatively correlated with the stemness and cycle scores of BM cells, which is in line with the physiological changes of cell differentiation (Fig. 1E).

Fig. 1figure 1

Cell population of control donor BM samples. A Overview of the study workflow. BM aspirates were collected and processed from control donors and MM patients for scRNA-seq and Nanopore sequencing to characterize the global single-cell ecological landscape and clonal evolution model of MM. B Single-cell profiles of the control donor BM based on t-SNE approach. Each color represents a cell identity, including hematopoietic lineages such as hematopoietic stem cells and juvenile red blood cell lineages, myeloid cells such as pro-monocytes and monocyte dendritic cells, and lytic cells such as T cells, B cells, and NK cells, for 16 cell types. C Tracks plots showing known marker genes specific to the identity of control donor BM cells. The cluster modules in the columns indicate the cell identity of the control donor BM, while the rows indicate the expression of the marker genes, along with cell abundance and cell ratio. D Single-cell atlas based on t-SNE showing the cell cycle score, stemness score, and pseudotime score of control donor BM cells. E Correlation between stemness score, cell cycle score, and pseudotime score (p < 0.001). MM, multiple myeloma; t-SNE, t-distributed stochastic neighbor embedding; TF, transcription factor; BM, bone marrow; GRN, gene regulation network; RSS, regulon specificity score

The single-cell profile of the control donor was constructed to reveal the ecological composition of different hematopoietic cell types in the normal BM, suggesting differentiation trajectory and fate choice, which is consistent with the current hematopoietic concept. Thus, this concept can be applied as a training cohort and reference system for the cellular ecological landscape of MM.

Cellular ecosystem landscape in MM

The malignant plasma cells extracted from NDMM and RRMM patients (CD38+CD56+CD138+CD19−CD20−; Supplementary Fig. 1A, B and Fig. 2A-C) were further clustered, with 11 malignant plasma cell clusters obtained, whereas six malignant plasma cell subsets were obtained for CD20+ RRMM patients (CD38+CD56+CD138+CD19−CD20+; Supplementary Fig. 1C and Fig. 2D, E). Malignant cell clusters shared cancer characteristics and widely significantly overexpressed HLA-A, HLA-B, MCL1, HDAC1, LCK, HSPB1, and IL6R (Fig. 2F), indicating a common tumor origin. Corresponding specific markers were also identified between different malignant cell clusters (Fig. 2G, Supplementary Table 4), providing direct evidence for the formation of different subpopulations.

Fig. 2figure 2

Cellular ecosystem of the BM of MM patients. A Clinical data and single-cell profiles of MM patients. Cell counts indicate the single-cell transcriptome that passed the quality threshold for each patient. The proportions of malignant plasma cells detected in smears, flow cytometry, and scRNA-seq in BM aspirates for each patient are demonstrated in pie charts. The two pie charts for scRNA-seq in CD20+ RRMM patients represent the proportion of CD20+ malignant plasma cells and the proportion of all malignant plasma cells, respectively. B-D Identification of malignant plasma cells in patients with NDMM (B), RRMM (C), and CD20+ RRMM (D). Left: Single-cell atlas of t-SNE-visualized malignant cells of the microenvironment in patients. Top right: Identification of malignant plasma cells with clinical and laboratory markers of MM malignancy such as CD38, CD56 (NCAM1), CD138 (SDC1), and CD20. Bottom right: Flow cytometry plot for validation of malignant plasma cell identification. E Malignant plasma cell subpopulations in patients with MM. Left: t-SNE single-cell atlas visualizing malignant plasma cell subpopulations in the BM of patients with NDMM and RRMM. The top panel shows the cell cycle score and stemness score of these malignant plasma cell subpopulations. Right: t-SNE single-cell atlas visualizing malignant plasma cell subpopulations in CD20+ RRMM patients. F Malignant marker genes shared by malignant plasma cell subpopulations in patients with NDMM and RRMM. Cluster modules in columns indicate malignant plasma cell subpopulations, while rows indicate expression of marker genes, cell abundance, and cell ratio are also shown. G Specific expressed marker genes for malignant plasma cell sub-clusters. H Microenvironment cells in the BM of patients with MM. A single-cell atlas based on t-SNE showing cells in the microenvironment in NDMM, RRMM (left), and CD20+ RRMM (right). I Expression patterns of known cell identity-specific markers in MM patient myeloid tumor microenvironment cells. Relevant marker genes were specifically expressed in the corresponding cell identity. J High-confidence communication network between malignant plasma cells and cells of the microenvironment in the BM of patients with MM. The Circos plot demonstrates each high-confidence ligand-receptor interaction pair of malignant plasma cell subpopulations and microenvironmental cells. The arrowheads are oriented from the ligand of the source cell toward the receptor of the target cell, while the thickness of the arrowheads represents the mean value of expression of the ligand-receptor interaction pair. NDMM, newly diagnosed multiple myeloma; RRMM, refractory or recurrent multiple myeloma; t-SNE, t-distributed stochastic neighbor embedding; BM, bone marrow

We identified and annotated the major cell types within the microenvironment in MM patients by investigating the expression patterns of known marker genes, which enabled to accurately define the specific identity of the microenvironment cells in MM patients (Fig. 2H, I, Supplementary Table 5). There was complex and active communication between different BM cells in MM patients, regardless of newly diagnosed and relapsed patients (Supplementary Fig. 1D, E and Fig. 2J). Consequently, the global cell ecological landscape of BM from MM patients can thus be characterized and clarified.

Variation events related to cell clonal evolution

Chromosomal instability is a hallmark of human cancer and tumor heterogeneity. The structural variations catalog (Supplementary Fig. 2A) in the present study showed significant genomic instability in the majority of chromosomes, except for chr4, including deletion (DEL), duplication, inversion, and insertion (INS). Genes related to DEL and INS events were expressed and played a major role in driving malignant cell clustering (Fig. 3A). Furthermore, malignant CNV events (Fig. 3B) and the corresponding transcripts (Fig. 3C), such as TNFSF13B, CD79A, TNFRSF13B, PARP1, IMPDH2, and MYC, were identified, expressed either alone or in combination in different malignant cell subpopulations and contributing to polyclonality. IFITM2 has been proven as an effector gene of the type I interferon response that protects cells against invading viral pathogens [44]; the shared mutation of IFITM2 (chr11:309127:A > G) in all malignant subpopulations was detected at both the bulk and the single-cell levels, with high expression activity in most malignant cells (Fig. 3D), suggesting its potential as a candidate mutation of malignant origin. Similarly, the shared mutation of ANK1 (chr8:41510767:T > G) was detected in all malignant subpopulations, but only expressed in type I malignant subpopulation. These mutations were concentrated in the patterns of C > T and T > C (Supplementary Fig. 2B), which is consistent with the basic genetic concept that C in CpG dinucleotides tends to mutate to T after methylation [45, 46]. Further analysis showed that other patients had higher TMB levels than that of the CD20+ RRMM patient and had enriched type IX malignant cell clusters (Supplementary Fig. 2C). TMB was consistent at the bulk and the single-cell levels (Supplementary Fig. 2D). In addition to genomic variation, the gene regulatory network with TFs as pivots was organized into six modules (Fig. 3E), such as BCL6, FOXO1, E2F7, and FOXP2, to regulate the specific gene expression (Fig. 3F) and RNA transcription rate (Fig. 3G) of MM malignant cell subpopulations to guide cellular fate choice. This promotes the transformation and differentiation of the core state (Supplementary Fig. 3A), ultimately mediating the formation of a series of clonal phenotypes [30].

Fig. 3figure 3

Relationship between the malignant progression of MM and the evolution of cell clones. A Gene transcriptional activity of SV event genes in MM malignant subclones. Left: SV spectrum of patients with MM. These SVs occur only in patients with MM, but not in control donors. Right: Expression pattern of SV genes in malignant plasma cell clusters. B CNV atlas of MM patients at the large-volume BM tissue level and at the single-cell level. C Transcriptional activity of CNV event genes in MM malignant subclones. Genes identified to be associated with the development of MM in previous studies have been highlighted. D Transcriptional activity of SNV event genes in MM malignant subclones. The SNV events detected simultaneously by the single-molecule long-read genome sequencing and single-cell transcriptome are demonstrated. Left: SNV mapping of patients with MM at the level of large-volume BM tissue. Middle: SNV atlas of MM patients at the level of BM monocytes. Right: Corresponding SNV gene expression patterns in the malignant plasma cell clusters. E Co-expression modules of transcription factors in malignant subclones of BM from patients with MM. Left: Identification of regulator modules based on the regulator’s CSI matrix. Middle: Representative transcription factors and their binding motifs in the module. Right: association of modules with malignant subclones. F t-SNE single-cell atlas mapping of MM malignant subclone-specific GRN. G RNA rate flow of MM malignant subclones mapped in t-SNE single-cell profiles. H Proposed chronological clonal evolutionary trajectory of MM malignant plasma cells mapped on the t-SNE single-cell atlas. The proportion of NDMM and RRMM cells characterizing the drug sensitivity of malignant cell subclones demonstrated using pie charts. The clonal evolution landscape characterizes the core state of MM malignant subclones in the malignant process with phenotypic transition differentiation. I Expression patterns of genes associated with the proposed chronological clonal evolution of MM malignant plasma cells and their biological signaling and cascade activation. J Structural changes in MM malignant plasma cells. Fish plots demonstrate the structural changes in MM malignant plasma cells from their origin through natural development, drug selection, and eventual relapse. K Global clonal evolutionary patterns of MM malignant plasma cells. Pie chart showing the cell proportion of NDMM and RRMM cells characterizing the drug sensitivity of malignant cell subclones. SV, structural variation; CNV, copy number variations; MM, multiple myeloma; BM, bone marrow; SNV, single nucleotide variation; CSI, connection specificity index; t-SNE, t-distributed stochastic neighbor embedding; GRN, gene regulation network

The pseudo-sequential clonal evolution atlas of MM malignant cells was finally constructed (Fig. 3H), which is consistent with the trend of the cell stemness index and cell cycle score (Supplementary Fig. 3B, C), whose corresponding gene expression, signaling pathways, and biological functions are shown in Fig. 3I. According to the cell stemness index score and pseudo-sequential clonal evolution analysis, the malignant origins of the cells were divided into type I and type IX malignant progenitor cell origins with the highest level of cell stemness index score (Fig. 3J). During the natural development of MM cells (Fig. 3K), type I and type IX malignant progenitor cells evolved into type II, III, and IV and type VIII, X, VII, and V malignant cells, respectively, with different drug sensitivity profiles; thus, type IV, VIII, VII, and V subpopulations explosively grew to an occupied advantage. However, with the selection of drugs and the occurrence of chemical carcinogenic variants, new cloned type VI malignant plasma cells were formed and became a superior subpopulation.

Multi-omics abnormal program identification of primitive MM malignant progenitor cells

Subsequently, we focused on primitive MM cell subpopulations, including type I and IX malignant progenitor cells, which promote the occurrence and growth of tumors with high stemness and cell cycle activity (Fig. 4A). Type IX malignant progenitor cells showed higher stem cell characteristics, resulting in a higher degree of malignancy in patients. Their cell markers with logFC > 1 were extracted to verify the expression and relative abundance in a large-scale clinical cohort of 9574 patients with 24 independent datasets (Supplementary Table 6) and the MM pathological classification of malignant origin was performed (Fig. 4B, Supplementary Table 7). A higher relative abundance of malignant progenitor cells was associated with a worse prognosis (P < 0.0001; Fig. 4C) for both types I and IX, especially for the latter (P < 0.0001). We observed similar results at the single-cell level with high robustness. Patients with a double-positive origin of type I and IX had the worst prognosis, followed by those with a double-negative origin (P < 0.0001). Consistent with the relative abundance analysis, the prognosis of patients positive only for type IX origin was worse than that for only type I origin.

Fig. 4figure 4

Multi-omics anomaly procedure for MM malignant origin. A Expression pattern characteristics of MM malignant origin. The bottom heatmap shows the expression pattern of type I and IX malignant progenitor marker genes, while the top annotations represent the GSVA score, cell cycle score, and tumor stemness score of type I and IX malignant origin. B Verification and typing of malignant origins in a large-scale clinical MM patient cohort. Top: malignant origin advantage score of the MM clinical patient cohort. The yellow bar represents the type I origin advantage and the blue bar represents the type IX origin advantage. Middle: abundance score of type I and type IX malignant progenitor cells in the clinical patient cohort of MM. The advantage score of malignant origin at the top = type I malignant progenitor cell abundance score - type IX malignant progenitor cell abundance score. Bottom: expression patterns of type I and type IX malignant progenitor marker genes in the MM clinical patient cohort. C Clinical prognostic value of MM malignant origin. Survival curves demonstrating the survival prognostic potential (OS and RFS) of type I and IX malignant progenitor abundance score, malignant origin predominance score, and malignant origin predominance typing in a cohort of patients with MM. D Variation in the malignant origin of MM drives a global gene expression regulatory network. The associated genes were regulated by GRN, SNV, and CNV, and clustered into four modules based on expression correlation to activate or inhibit seven biological signaling pathways. E Molecular mechanism of malignant origin mediated by early carcinogenic drivers. F Expression patterns of immune checkpoints PD-1, PD-L1, and CTLA4 in malignant plasma cells and tumor microenvironment cells in patients with MM. The negative expression of immune checkpoint-related genes in malignant plasma cells and tumor microenvironment cells provides molecular insight at the single-cell level for the poor efficacy of immune checkpoint blocker therapy. G Expression pattern of antitumor immune response cascade-related genes in microenvironment cells. H Communication network with high confidence between early malignant progenitor cells and tumor microenvironment cells in patients with MM. The Circos diagram illustrates each high-confidence ligand-receptor interaction of type I and type IX malignant progenitor cells and microenvironment cells. The direction arrow is from the ligand of the source cell to the receptor of the target cell, and the thickness of the arrow represents the average expression level of the ligand-receptor interaction. I Immune escape mechanism of early malignant progenitor cells in MM. MM, multiple myeloma; GRN, gene regulation network; CNV, copy number variation; SNV, single nucleotide variation; OS, overall survival; RFs, relapse-free survival

We evaluated the malignant forerunner events that were preferentially highly expressed by two malignant origin progenitor cells, including DNA damage repair, cell cycle, proliferation, migration, invasion, and stemness (Supplementary Fig. 3D). In this regard, they showed extensive similarities, with many previously reported hematological tumor-related genes, such as HMGB1, CCND2, CDK1, CDKN2A, and MYC [47,48,49,50,51]. These genes were more significantly dysregulated in type IX malignant cells, which explains the higher stem cell characteristics and malignant degree in the type IX origin. The similarity of malignant origin plasma cells also lies in their significantly dysregulated signaling pathways (Supplementary Fig. 3E). It is noteworthy that the type I malignant origin was significantly enriched in both base excision repair and DNA damage repair to reduce random variation events in the process of malignant proliferation, while type IX only activated the base excision repair, causing more random variation events in the process of clonal proliferation with a richer clonal pattern. Related dysregulated genes were clustered into co-expression modules according to expression similarity, which are mainly regulated by CNV events, involving SNVs and TFs (Fig. 4D), resulting in a shared carcinogenic initiation program and its own specific carcinogenic regulation mechanisms in two malignant origins of MM (Fig. 4E). Homogeneity and specificity are reflected not only in the carcinogenic processes of these origin types but also in the change in the expression pattern of their evolution (Supplementary Fig. 3F) with joint high expression of HBG2, MYC, CD79B, and MCL1 (Supplementary Fig. 3G). By contrast, the type I origin will reduce the expression of CDK6, ITGB1, BCL2L1, and STAT1 during evolution, whereas the type IX origin will promote their expression. Additionally, the type IX origin shows precursory potential for further evolution, in which the local cell population takes the lead in showing the gene expression pattern of subclone type VIII, with higher evolution efficiency. In contrast, the type I origin maintained a similar expression pattern in some cell populations cloned by the offspring, resulting in relatively slower evolution. These different evolutionary transition forms and efficiency indicate that the type IX origin can evolve more cell subclonal clusters than the type I origin within the same timeframe.

We further explored early immune escape mechanisms, as the foundation for immune resistance and evasion, throughout the malignant process of MM. Immune checkpoints PD-1, PD-L1, and CTLA4 were negatively expressed in malignant plasma and microenvironment cells (Fig. 4F). The expression pattern of genes related to tumor immune killing in the microenvironment cells (Fig. 4G) was used to build a high-confidence immune escape intercellular communication network in the early stage of MM malignant transformation (Fig. 4H), providing insight into the potential mechanism (Fig. 4I) from innate to specific immunity. Among them, HLA-DMA and HLA-DRA were generally missing in antigen-presenting cells (B cells, classical dendritic cells (cDCs), and monocytes), which is the first weakness in the immune response cascade against MM (P.adj < 0.0001). CCL5 expression was also inhibited in CTLs and NK cells (P.adj < 0.0001), which significantly reduces the efficiency of recruiting DCs and more CTLs into the nidus [52]. Additionally, defective expression of the perforin PRF1 and the granzyme GZMB in NK cells is the main reason for the inherent antitumor immune inactivation in MM patients (P.adj < 0.0001). Moreover, the expression of CD38 and CD27 in the B cell lineage (B, ProgB-I, ProgB-II, and plasma) was inhibited (P.adj < 0.0001), indicating deficiency of humoral immunity, while low levels of CCL3L1 also suggest low immune activity of monocytes (P.adj < 0.0001) [53].

Relapse and drug resistance in MM patients at the single-cell level

We evaluated the inhibitory effect of drugs on the expression of the established target genes at the patient level (Fig. 5A) to infer the part of the drug action that failed. The expression levels of target genes for thalidomide, melphalan, lenalidomide, and cyclophosphamide in the RRMM group did not show significant differences from those of the NDMM group, whereas those for dexamethasone, bortezomib, and doxorubicin decreased in sectional cell subpopulations. It is clear that the drug resistance of RRMM1 mainly occurs prior to the action on target genes, whereas the drug resistance of RRMM2 mainly occurs after interaction with the target gene. RRMM3 shares these two drug resistance mechanisms due to the extensive application of drugs. The characteristics of drug sensitivity at the patient level are also reflected at the cell subpopulation level, as demonstrated by systematic comparison between the cell subpopulations of the RRMM and NDMM groups (Fig. 1A, Fig. 5B). In addition, we observed the increased expression of genes such as IGHG3, IGLC2, and IGHG2 and the decreased expression of genes such as IGHG4, ITM2C, and IGHA1, which mediates the formation and development of malignant clones, causing heterogeneity in patients. The signal score of malignant cells was applied to estimate the pathway activity (Fig. 5C), in which apoptosis and the FOXO/p53 signaling pathways were inhibited in RRMM, while the calcium/Rap1/JAK-STAT/VEGF/mTOR signaling pathways related to survival, proliferation, migration, and stem cell characteristics were activated, which are mainly driven by CNV events supplemented by SNVs and TFs (Fig. 5D). This is similar to the pattern of the malignant origin. During clonal evolution of malignant cells to acquire drug resistance, the transformation of gene expression (Fig. 5E) was clustered into two gene expression modules, involving four evolutionary patterns. The acquisition and maintenance of the drug resistance phenotype of malignant cells require high expression of gene expression module 1 (related to drug sensitivity) and relatively low expression of gene expression module 2 (related to the development advantage in the natural state). The genes of both modules are related to the above-mentioned drug resistance signals, which are significantly regulated by CNV events as the main driving force for the clonal evolution of drug resistance.

Fig. 5figure 5

Adverse pathological features of drug resistance recurrence in MM patients observed at the single-cell level. A Expression patterns of drug targets in malignant subclones and tumor microenvironment cells. Left: Bubble size represents the number of drugs with resistance in patients with RRMM. The connecting lines are colored according to the drug. Middle: Bubbles of different colors represent the tolerance of RRMM patients to different drugs, and the size represents the number of targets. Right: Expression of drug targets in malignant subclones and tumor microenvironment cells. B Single-cell contribution of DEGs in RRMM patients compared with NDMM patients. DEGs are expressed at the single-cell level (left), the average expression pattern at the cell cluster level (middle), and the differential expression (logFC) at the cell cluster level (right) in the malignant subclones and tumor microenvironment of NDMM and RRMM patients. C Biological signals related to drug resistance. These biological signals were significantly activated and inhibited in RRMM patients compared with NDMM patients in the six shared malignant subclones. Inhibitory signals included apoptosis, FoxO, and p53. The activated signals involved signals related to survival, proliferation, migration, and stem cell characteristics, such as calcium/Rap1/JAK-STAT/VEGF/mTOR signaling. Each edge represents a comparison of the NDMM and RRMM groups for any of the six shared malignant subclones with correlation signal scores in types I, II, III, VIII, IX, X, and XI. D Drug resistance-related genes significantly involved in signaling pathways. The DEGs of shared malignant subclones in RRMM patients compared with NDMM patients are identified as drug resistance-related genes. Top: DEGs and their significant signaling pathways. Middle: malignant driver of DEGs involving GRN, SNV, and CNV. Bottom: DEGs expression changes in six shared malignant subclones. E Clonal evolution mediates the signaling pathways significantly involved in drug resistance-related genes. Top: MM malignant clonal evolution mediates drug resistance-related genes and their biological signals. Middle: Malignant drivers of malignant clonal evolution of MM involving GRN, SNV, and CNV. Bottom: Drug resistance-related genes clustered into two gene expression modules during the evolution of MM malignant clones. F Comprehensive regulatory network of drug resistance in patients with RRMM. MM malignant subclones not only inhibit their own apoptosis-related signals and activate their own survival-promoting, proliferation, migration, and stem cell characteristics related signals but also reprogram microenvironment immune cells through intercellular communication, driving the latter to activate drug metabolism signals, thereby improving the drug microenvironment and survival probability. G Prognostic potential of drug resistance-related genes in a large-scale MM clinical patient training cohort. H Prognostic value of drug resistance-related genes in a large-scale MM clinical patient cohort based on the multivariate Cox model. I Time-independent ROC curves for evaluating the prediction performance of the prognostic model in MM patients’ OS and RFS. NDMM, newly diagnosed multiple myeloma; RRMM, refractory or recurrent multiple myeloma; GRN, gene regulation network; CNV, copy number variations; SNV, single nucleotide variation; DEG, differentially expressed gene; OS, overall survival; RFS, recurrence-free survival

The global regulatory network of intercellular communication (Fig. 5F) showed that activation of the drug metabolism signal pathway in B cells and cDCs may be reprogrammed by malignant cells, whereas the drug resistance evolution of malignant cells may involve microenvironment cells such as B cells, plasma cells, T cells, and cDCs. Genes in the drug resistance global regulatory network also showed significant prognostic potential in the training cohort (Fig. 5G). Among them, IL5RA, KRAS, and PPP2R5C were independently linked to prognosis (Fig. 5G); thus, a drug resistance-related prognostic model for MM based on multivariate Cox regression analysis was established for OS and RFS, which was verified in an independent cohort (Fig. 5H). Time-independent ROC curves demonstrated that the prognostic model enabled to accurately predict MM patients’ OS and RFS (Fig. 5I).

Intratumoral cellular heterogeneity in MM patients

We applied the tumor cell-specific transcriptional diversity score to measure the intratumoral heterogeneity of MM (Supplementary Fig. 4A), detailed in the single-cell profile map of malignant cells from each patient (Supplementary Fig. 4B), which demonstrates diverse and specific ecological components with vague traces of two malignant origins. Most of the malignant cells in MM patients exhibited the activated antigen processing and presentation (Fig. 6A), although the genes involved were different, indicating that malignant plasma cells retained incomplete antigen presentation ability to a certain extent. In particular, compared with other patients, for patient RRMM3, more distinctive malignant cell ecological components were found, regulated by TFs such as E2F7, E2F8, and SAP30 (Supplementary Fig. 4C) and leading to greater heterogeneity. SOM analysis showed the contribution (Fig. 6B) and concrete manifestations (Supplementary Fig. 4D) of malignant cell subpopulations to patient heterogeneity, of which NDMM and RRMM patients showed distinct SOM neural units, suggesting that different gene clusters respond differently to the drug sensitivity of MM. Compared with malignant cells, patient heterogeneity of microenvironment cells was relatively obscure (Supplementary Fig. 4E), which was also observed at the molecular level (Fig. 6C).

Fig. 6figure 6

The MM malignant clonal evolution reprogramming tumor microenvironment mediates tumor heterogeneity in patients. A Patient-specific transcriptional patterns of MM malignant subclones. B Malignant subclonal self-organization contribution of specific marker genes in MM patients. C Patient-specific transcriptional patterns of MM BM microenvironment cells. D Lymphocyte subsets of the BM microenvironment of control donors and MM patients. The atlas involves B cells, plasma cells, CTLs, T cells, and NK cells. Left: Single-cell subpopulation atlas of each cell type. Middle: Proportion of each cell subgroup in the control donor and different MM patients. Right: Expression pattern of cell subgroup-specific markers mapped in the single-cell atlas. E Atlas of myeloid cell subsets in the BM microenvironment of control donors and MM patients. The atlas involves classical dendritic cells (cDCs) and monocyte cells. Left: Single-cell subpopulation atlas of the cell type. Middle: Proportion of each cell subgroup in the control donor and different MM patients. Right: Expression pattern of cell subgroup-specific markers mapped in the single-cell atlas. F Expression patterns of lymphocyte subset-specific markers and the biological signals involved. G Expression patterns of myeloid cell subset-specific markers and the biological signals involved. CTL, cytotoxic T lymphocytes; MM, multiple myeloma; BM, bone marrow

Immune cell subpopulation atlas of MM

The single-cell profiles (Fig. 6D) of B, plasma, CTLs, T, and NK cells were obtained by clustering the immune cell subpopulations, so that the heterogeneity of the immune microenvironment in MM patients was amplified at the subpopulation level. Interestingly, we found that plasma, CTLs, T, and NK cells highly expressed markers of MM malignant clonal evolution as subpopulation-specific markers, reflecting the immune cell heterogeneity of patients. Among them, IGHG3 was mainly expressed in lymphocytes of RRMM1, involving plasma-1, CTL-0, T-2, and NK-0, whereas IGLC2 was specifically expressed only in RRMM2 lymphocytes, related to CTL-1, T-5, and NK-2. Alternatively, IGHG2 was specifically expressed in the lymphocytes of CTL-2 and NK-4 in RRMM3. IGHG4 was specifically expressed in the lymphocytes of NDMM1 and NDMM3, touching upon CTL-3, T-4, and NK-4. The molecular specific markers of these microenvironment lymphocyte subpopulations are consistent with the distinctive expression of patient heterogeneous malignant single cells (Supplementary Fig. 4D). This provides strong evidence that clonal evolution of MM can reprogram lymphocyte-mediated patient heterogeneity, inspiring the novel strategy of individualized treatment in clinics. As this is an unprecedented discovery, we ruled out the possibility of wrong cell type identification for the sake of caution. The CTLs, T, and NK cells showed positive expression of their corresponding specific marker genes, which have significant cell co-localization with the corresponding patient heterogeneous malignant marker immunoglobulins (Supplementary Fig. 5A-C). Importantly, the same phenomenon was observed in cDCs, monocytes (Fig. 6E), and HSCs (Supplementary Fig. 5D). The transformation of these cell phenotypes from normal to depleted was accompanied with transformation of a series of gene expression patterns (Supplementary Fig. 5E, F) with various degrees of cell death signal activation, involving ferroptosis, necroptosis, and apoptosis (Fig. 6F, G). This raises the question of how malignant plasma cell markers appear in microenvironment cells to reprogram the microenvironment. By reviewing the biological signals activated by these depleted cells, we found that they not only activated various cell death signals, but also activated phagocytosis-related pathways, such as phagosome, Fc gamma, R-mediated phagocytosis (Fig. 6F, G). Correspondingly, during the clonal evolution of MM malignant cells, the key gene of exosome synthesis CD63 was expressed and a series of biological pathways related to vesicle synthesis and secretion were activated to varying degrees (Supplementary Fig. 5G, Fig. 3I). It is axiomatic that the MM malignant cells reprogramming microenvironment to augment immune escape is involved in vesicle synthesis and secretion of malignant cells as well as phagocytosis of microenvironment cells during the clonal evolution process, resulting in the emergence of malignant marker mRNA and mediating apoptosis and depletion.

In contrast, the positive expression of GNLY in T-1 emphasizes the potential malignant events that effectively activate the immune response of T cells. T-1 mainly existed in NDMM, concentrated in NDMM2, which may be related to the negative expression of malignant clonal evolution markers in the lymphocytes of NDMM2 patients, further indicating that chemoresistance of MM not only resists drugs but also inhibits immune killing.

Importantly, we found the negative expression of malignant clonal evolution markers in B cells, implying certain defensive means of B cells to resist the reprogramming and transformation during malignant cell clonal evolution, suggesting that a cellular immunotherapy strategy may benefit MM patients. CCL5 and CCL4 were specifically expressed in B-0 (Fig. 6D, F), which activate antigen presentation pathways and PD-L1/PD-1 immune checkpoints in cancer to recover the adverse effect of low CCL5 expression reducing the recruitment efficiency for cells in immune killing cascades (e.g., cDCs and CTLs). Surprisingly, B-0 were widely distributed in MM patients, mostly in the NDMM group (concentrated in NDMM2 patients), but not in the control, suggesting that an immunotherapy strategy of injecting CCL5-positive B cells derived from autologous expansion in vitro may be beneficial. Likewise, T-1 with potential immune killing efficacy highly expressed CCL5 and CCL4, activating antigen processing/presentation and chemokine signaling pathways to augment leukocyte transdermal migration. There are similar patterns of molecular characteristics and gene expression of T-1 in T-3 that were specific to controls, further highlighting the important role of CCL5 and CCL4 in the antitumor immunity of MM.

留言 (0)

沒有登入
gif