A detailed atlas of the tumor-immune microenvironment in murine and human MM

To investigate the dynamic changes within the MM-TME, we used the immunocompetent 5T33MM mouse model. Tumor progression was monitored in the BM, spleen and serum at 7, 14, and 20 DPI, the latter corresponding to end-stage of disease with mice showing signs of paralysis, and compared to naïve control mice (Fig. 1A, Supplementary Figure S.1A–B). In both BM and spleen, the increase in tumor load, demonstrated by an increase in plasmacytosis and M-protein (Fig. 1A), was associated with a significant decrease in the percentage of CD45+ immune cells as from 14 DPI (Supplementary Figure S.1C). Notably, in the spleen, the absolute number of CD45+ immune cells increased until 14 DPI when compared to naïve mice, suggesting immune infiltration in the first phases of tumor progression (Fig. 1B).

Fig. 1

A detailed atlas of the tumor-immune microenvironment in murine and human MM. (A) Bar graphs show the tumor load in the 5T33MM bone marrow (BM), spleen and serum at 7, 14, and 20 days post-tumor inoculation (DPI). This is demonstrated by the percentage of plasmacytosis in BM and spleen, and the presence of the M-protein (g/L) in serum. (B) Bar graphs show the absolute number of CD45+ cells in BM and spleen of naïve and 5T33MM mice at 7, 14, and 20 DPI. n = 5 per group. (C) Schematic outline of the experimental procedures, detailing the sorting of 7-AAD−CD45+idiotype− immune cells from naïve and 5T33MM mice (at 14 DPI and 20 DPI) for single-cell RNA-sequencing (scRNA-seq; 10 × Genomics). n = 4 samples pooled. (D) UMAP plot shows high-resolution clustering of the immune cell compartment in BM and spleen, using the scRNA-seq dataset of naïve mice and the 5T33MM model. (E–F) UMAP embedding as shown in panel D, but colored according to (E) tissue of origin; including BM and spleen, and (F) stage of disease; including naïve mice and 5T33MM mice at 14 DPI and 20 DPI. (G-H) Bar graphs show the frequency of immune cells within the CD45+ cells across different stages of disease in murine BM and spleen, analyzed using (G) scRNA-seq and (H) flow cytometry. (I) Schematic outline of the consulted publicly available datasets (scRNA-seq; 10 × Genomics), including a CD138−CD45+ sort on mononuclear cells from healthy individuals (n = 9), precursor stages such as Monoclonal Gammopathy of Undetermined Significance (MGUS; n = 5) and Smoldering Multiple Myeloma (SMM; n = 11), as well as Newly Diagnosed Multiple Myeloma (NDMM; n = 7) and Relapsed/Refractory Multiple Myeloma (RRMM; n = 20). (J) UMAP plot shows high-resolution subclustering of the immune cell compartment within the human BM scRNA-seq dataset. (K) Bar graph shows the frequency of immune cells within the CD45+ cells of the human BM scRNA-seq dataset across different stages of disease. Error bars represent mean values ± SD. Statistical analysis was performed by Ordinary One-way ANOVA. ns: p ≥ 0.05, *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001

To dissect how the immune TME changes upon MM disease progression, we performed scRNA-seq on fresh CD45+idiotype− immune cells using the 10 × Genomics platform on BM and spleen samples from naïve, 14 DPI and 20 DPI 5T33MM-bearing C57BL/KaLwRij mice (Fig. 1C). Unsupervised clustering, dimensionality reduction and UMAP were performed on 47.129 cells (Fig. 1D). Individual clusters were identified based on the expression of known marker genes (Supplementary Figure S.1D). Our scRNA-seq data revealed the heterogenous myeloid and lymphoid compartment within the BM and spleen, thereby identifying at least 17 cell types and progenitor cells (Fig. 1D). Among the myeloid cells, which mainly derived from the BM samples, precursors, DC-precursors, monocytes, and neutrophils were the most abundant populations. The lymphoid cells included NK-/T-cells and B cells, and were more abundant in the spleen (Fig. 1D-E). Importantly, our scRNA-seq dataset unveiled the dynamic changes in the composition of various cell types at different stages of MM disease progression (naïve, 14 DPI and 20 DPI) (Fig. 1F-G). In the BM, the B-cell compartment was strongly reduced in MM-bearing mice as compared to naïve mice, while precursors (including DC precursors and Neu precursors), conventional dendritic cells (cDCs), natural killer T (NKT) and T cells increased. Similar but less pronounced trends were observed in the spleen, however, in contrast to the BM, neutrophils increased in the spleen upon tumor progression (Fig. 1G). Of note, within the B-cell compartment, we identified a small fraction of plasma cells, which poorly expressed the MM marker Tnfrsf17, indicating that our dataset is only minimally contaminated with MM cells (Supplementary Figure S.1E and F). We further validated these kinetic changes observed in scRNA-seq by flow cytometry, using the gating strategy illustrated in Supplementary Figure S.1G, in naïve and 5T33MM-bearing mice at 7, 14 and 20 DPI. Overall, similar alterations were observed in the distinct immune cell types (Fig. 1H).

To address the current need for a comprehensive understanding of dynamic changes within the immune cell compartment during MM disease progression, and to assess the translatability of our pre-clinical mouse model, we correlated the results obtained from the murine 5T33MM model with publicly available scRNA-seq datasets [6, 7] of healthy donors (n = 9), and MM patients at different stages of the disease, being MGUS (n = 5), SMM (n = 11; including low-risk/high-risk SMM), NDMM (n = 7), and RRMM (n = 20) patients (Fig. 1I). We reanalyzed the CD45+CD138− immune compartment, devoid of cancer cells, and identified 15 broad cell types; ranging from precursor cells to myeloid cells (monocytes, macrophages, DCs), lymphoid cells (B-/NK-/NKT-/T-cells), but also erythroid cells and mast cells (Fig. 1J, Supplementary Figure S2A–C). Similar to our murine data, we observed an increase in cDCs, CD8+ T cells, NK cells and NKT cells upon MM disease progression (Fig. 1K). Intriguingly, BM-precursor cells increased in mouse, while a decrease could be observed in human samples at all stages of disease progression. Of note, the kinetic human scRNA-seq dataset does not contain neutrophils. This could be due to the low transcript counts in neutrophils, due to the use of density gradients during sample processing or due to the fact that the sequencing was performed on frozen patient samples [17].

In conclusion, temporal analysis of the dataset we generated allows us to track the evolution of the immune TME and identify key changes associated with disease progression in both murine and human MM.

The MM microenvironment is characterized by the presence of T cells with an exhausted phenotype

T cells are major players in adaptive immunity and are essential for mounting an effective anti-tumor immune response [18, 19]. Therefore, we delved deeper into the NK-/T-cell compartment and upon subclustering, we could distinguish at least 14 distinct cell types based on signature and activation genes (Fig. 2A, Supplementary Figure S.3A). The annotation of the distinct clusters was validated by the alignment of our murine scRNA-seq data with the tumor-infiltrating T-cell atlas of Andreatta et al. [20] (Supplementary Figure S.3B).

Fig. 2

The MM microenvironment is characterized by the presence of T cells with an exhausted phenotype. (A) UMAP plot shows high-resolution subclustering of the NK-/T-cell compartment within the in-house murine scRNA-seq dataset. (B) Bar graph shows the frequency of the different NK-/T-cell subsets within the murine BM and spleen, and across different disease stages; including naïve mice and 5T33MM mice at 14 DPI and 20 DPI, analyzed in the murine scRNA-seq dataset. (C and D) Bar graphs show (C) the frequencies of CD8+ T cells, CD4+ T cells, Foxp3+ T cells, γδT cells, NKT cells and NK cells as a percentage of the NK and T cells (CD11b− CD19− cells), and (D) the CD8+ T cells/Treg ratio in BM (top) and spleen (bottom) from naïve and 5T33MM mice at 7, 14, and 20 DPI. Analyzed using flow cytometry. n = 5 per group. (E) Bar graph shows the frequency of CD44+CD62L+ Central memory T cells, CD44−CD62L− T cells, CD62L+CD44− Naïve T cells and CD62L−CD44+ Effector T cells within the CD8+ T cells in murine BM and spleen, and across different stages of disease, analyzed using flow cytometry. n = 5 per group. (F) Bar graphs show ΔMFI (median fluorescence intensity) of IFN-γ in CD4+ T cells and CD8+ T cells present in the BM and spleen across different stages of disease, analyzed using flow cytometry. n = 5 per group. (G) UMAP plot shows high-resolution subclustering of the CD8_Effector memory cluster and the CD4_T cluster, identifying the CD4_Exhausted cluster and the CD8_Exhausted cluster as well as their evolution across different stages of disease, analyzed in the murine scRNA-seq dataset. (H) Dot plot shows the expression of marker genes associated with exhaustion within the murine scRNA-seq dataset. Dot size represents the percentage of cells expressing the gene and color gradient represents the average scaled expression within a cell cluster. (I) UMAP plot shows high-resolution subclustering of the NK-/T-cell compartment within the human BM scRNA-seq dataset. (J) Bar graph shows the frequency of the different NK-/T-cell subsets within the human BM scRNA-seq dataset across different stages of disease. Error bars represent mean values ± SD. Statistical analysis was performed by Ordinary One-way ANOVA. ns: p ≥ 0.05, *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001

Unique to our scRNA-seq data is the ability to track the evolution of the different T-cell subsets within the MM-TME (Fig. 2B, Supplementary Figure S.3C and D). Within the T cells, a clear shift from naïve-like to activated/effector T cells was observed in MM-bearing mice when compared to naïve mice (Fig. 2B). As such, in the 5T33MM model, we observed a decrease in CD4_Naïve-like and CD8_Naïve-like clusters, which were marked by a high expression of Ccr7, Sell (encoding for CD62L), Tcf7 and Lef1, and corresponded to naïve and/or central memory T-cells [21] (Supplementary Figure S.3A). In addition, the CD8_Effector memory, CD8_IFN and T_Prolif cluster were defined based on the expression of genes associated with effector CD8+ T-cells (Gzmb, Gzma, Gzmk, Prf1, Ifng), IFN-associated genes (Ifit3, Isg15, Ifit1, Stat1, Rsad2) and cell-cycle genes (Mki67, Top2a, Stmn1) respectively, and drastically increased upon disease progression (Fig. 2B, Supplementary Figure S.3A). Other changes induced by the disease were the radical decrease in γδT cells, a slight decrease in Tregs and NK cells, and an important increase in NKT cells in the BM (Fig. 2B).

Next, we aimed to validate the changes in the NK-/T-cell compartment using flow cytometry and confirmed that, although the absolute numbers of each lymphoid cell type was significantly decreased upon disease progression (Supplementary Figure S.4A and B), the proportion of CD4+ T, CD8+ T, as well as Tregs significantly increased within the NK-/T-cell compartment of MM-bearing mice (Fig. 2C), though overall the CD8+ T/Treg ratio was decreased in MM-bearing mice compared to naïve mice (Fig. 2D). Similar to what was observed in the scRNA-seq dataset, we observed a significant increase in CD62L−CD44+ effector memory CD4+ T cells in the spleen, and CD8+ T cells in the BM and spleen, upon MM disease progression using flow cytometry (Fig. 2E, Supplementary Figure S.4C–F). However, when further looking into the functional state of several effector cells (NK-, NKT-, and CD4+ and CD8+ T-cells), we found that their potency to produce IFN-γ was significantly decreased in end-stage 5T33MM mice (Fig. 2F, Supplementary Figure S.5A and B). These findings suggest an impaired T-cell function, especially at end-stage of disease. Indeed, at the transcriptome level, we observed a higher expression of exhaustion/dysfunction markers such as Tox, Havcr2, Vsir, Lag3, Ctla4, Pdcd1 in the 5T33MM model (Supplementary Figure S.5C). Moreover, within the effector cluster, the exhausted CD4+ T cells (CD4_Exhausted) and CD8+ T cells (CD8_Exhausted) were exclusively present in MM-bearing mice and increased upon MM disease progression from 14 to 20 DPI, and were absent in naïve mice (Fig. 2G-H, Supplementary Figure S.5D). Accordingly, a significant increase of PD1 in CD4+ and CD8+ T cells was observed at the protein level in end-stage 5T33MM mice (Supplementary Figure S.5E and F).

Within the human scRNA-seq dataset, the CD4+ T cells showed a slight increase in precursor stages of the disease (MGUS and SMM), however, once progressing to MM (NDMM), CD4+ T cells were strongly decreased and remained low in RRMM (Fig. 1K, Supplementary Figure S.2C). While CD8+ T cells showed minimal variation throughout disease progression, NKT cells were enriched in NDMM and RRMM, when compared to healthy individuals and precursor stages (Fig. 1K, Supplementary Figure S.2C). Upon subclustering the NK-/T-cells in the human scRNA-seq dataset, we were able to identify similar activation states based on shared DE genes between mouse and human (Fig. 2I, Supplementary Figure S.6A–C). Interestingly, an increase in the CD8_IFN cluster could be observed in NDMM and RRMM patients compared to healthy controls (Fig. 2J, Supplementary Figure S.6A). Similar to our murine dataset, T cells from NDMM and RRMM patients also showed an increase in the expression of exhaustion markers, thus indicative of an exhausted phenotype (Supplementary Figure S.6D and E). Hence, despite the presence of effector cells, their function may be impaired in NDMM and RRMM, potentially leading to unsuccessful tumor control.

In conclusion, our data showed strong alignments within the NK-/T-cell compartment between both species, including an increase in effector cells upon MM disease progression, accompanied by an increase in T cells with an exhausted phenotype.

Neutrophils acquired a more pronounced pro-tumor phenotype upon MM disease progression

T cells can be directly suppressed by myeloid cells, which hence contribute to the immunosuppressive microenvironment in MM, thereby impeding the success of current immunotherapeutic approaches [14, 22]. Subclustering of the murine neutrophil and macrophage lineage, revealed 8 distinct clusters, based on specific marker genes, including granulocyte-monocyte progenitors (GMP) that can give rise to monocyte precursors (MP), monocytes and macrophages, and granulocyte precursors (GP) that give rise to neutrophils (Fig. 3A, Supplementary Figure S.7A).

Fig. 3

Neutrophils acquired a more pronounced pro-tumor phenotype upon MM disease progression. (A) UMAP plot shows high-resolution subclustering of the myeloid compartment within the murine scRNA-seq dataset. (B) Bar graph shows the frequency of the different myeloid subsets within the murine BM and spleen across different disease stages, including naïve mice and 5T33MM mice at 14 DPI and 20 DPI, analyzed using the murine scRNA-seq dataset. (C) Bar graphs show the frequency of neutrophils within the CD11b+ cells in the murine BM (top) and spleen (bottom) across different stages of disease, analyzed using flow cytometry. n = 5 per group. Statistical analysis was performed by an Ordinary One-way ANOVA. (D) Schematic outline of the experimental procedures. 5T33MM-inoculated mice received alternating anti-Ly6G and anti-Rat (MAR18.5), starting at 4 DPI. Mice were sacrificed at end-stage. (E) Bar graphs show the tumor load, assessed by the percentage plasmacytosis in BM and spleen, and via the M-protein in serum. n = 3–4 per group. Statistical analysis was performed by Mann-Whitney U-test. (F–L) Reclustering of the neutrophils using (F–I) the in-house murine scRNA-seq dataset, and (J–L) the human scRNA-seq dataset, based on de Jong et al. [9] (F, J) UMAP plot shows high-resolution subclustering of neutrophils and neutrophil precursors, analyzed in (F) the murine and (J) the human scRNA-seq dataset. (G) Violin plot shows the maturation status of the different neutrophil clusters. The maturation status was determined by the top 50 differentially expressed genes, based on Xie et al. [16] (H, K) Bar graph shows the frequency of the different neutrophil clusters across different stages of disease, analyzed using (H) the in-house murine scRNA-seq dataset, and (K) the human scRNA-seq dataset. (I, L) Heatmap plot shows the expression of IFN-associated genes by the different neutrophil clusters across different disease stages, analyzed using (I) the in-house murine scRNA-seq dataset and (L) the human scRNA-seq dataset. Error bars represent mean values ± SD. ns: p ≥ 0.05, *p < 0.05, **p < 0.01, ***p < 0.001, and *****p < 0.0001

In mice, based on prototypical signature genes, we identified a classical monocyte cluster (Mono_CM; Ccr2high and Cx3cr1low), and a non-classical monocyte cluster (Mono_NCM cluster; Ccr2low and Cx3cr1high), the latter being mainly restricted to the naïve condition (Fig. 3A-B, Supplementary Figure S.7A–C). In human, monocytes were subdivided in three clusters: classical monocytes (Mono_CM; CD14++CD16−), intermediate monocytes (Mono_IM; CD14++CD16+), and non-classical monocytes (Mono_NCM; CD14+CD16++) (Supplementary Figure S.8A–D). The Mono_CM cluster and Mono_IM cluster aligned with the murine Mono_CM cluster (expressing Ccr2). An additional monocyte cluster, the Mono_IFN cluster, which highly expresses IFN-associated genes, was found in both mouse and human datasets, and was slightly increased in NDMM and RRMM patients (Fig. 3B, Supplementary Figure S.8B and C). Furthermore, macrophages, expressing high levels of C1qa, Maf, Ctsd, and Mertk, were strongly reduced in MM-bearing mice, when compared to naïve mice, whereas, in human, macrophages were increased in NDMM and RRMM patients when compared to healthy individuals and precursor stages of MM (Fig. 3B, Supplementary Figure S.7D, Supplementary Figure S.8B and C). Finally, neutrophils, expressing high levels of Csf3r, Cxcr2, Mmp8, Mmp9 and Ly6g, represented the most dominant myeloid cell type within the TME of MM-bearing mice. In addition, neutrophil precursors showed a transient expansion during tumor progression (Fig. 3A-B). Although neutrophils were slightly decreased in BM, we observed a major increase in spleen (Fig. 3A-B), which was also confirmed by flow cytometry at end-stage of disease (Fig. 3C).

Given that neutrophils are such a predominant cell population in the BM and have been shown to suppress T cells in tumors [22, 23], we aimed to assess whether neutrophils influenced T-cell numbers and activation status in MM, by use of neutrophil-depleting agents. A daily injection of alternating anti-Ly6G/Isotype and anti-Rat (MAR18.5) in the 5T33MM mice from 4 DPI till end-stage resulted in successful depletion of mature neutrophils in blood, BM and spleen of MM-bearing mice (Fig. 3D, Supplementary Figure S.9A and B). Neutrophil depletion within the 5T33MM model resulted in a significant reduction in tumor burden solely in the BM (Fig. 3E), which is the organ that contains the highest number of neutrophils (Fig. 1G-H). While neutrophil depletion did not affect the frequencies of CD8+ or CD4+Foxp3− T cells, it did lead to a significant reduction in CD4+Foxp3+ Tregs and a higher activation of CD8+ T cells in the BM (Supplementary Figure S.9C and D). These results suggest that neutrophils might have an immunosuppressive or tumor-promoting phenotype.

To assess how the neutrophil phenotype changes with tumor progression, we further subclustered the neutrophils from the murine scRNA-seq dataset and annotated four neutrophil subclusters (Neu1, Neu2, Neu3, Neu4), which aligned with an increasing maturation status (based on Xie et al. [16]), as well as a pronounced IFN-associated signature (Ifit1, Ifit2, Ifit3, Ifitm3, Rsad2, Isg15) (Fig. 3F-I, Supplementary Figure S.9E–G). Interestingly, less mature or immature neutrophils (e.g. Neu2 and Neu3) were increased in MM-bearing mice compared to naïve mice (Fig. 3H). In addition, tumor progression was accompanied with a decrease in the most mature neutrophils (Neu4), which showed a pronounced IFN-signature at 14 DPI, but decreased expression of IFN-associated genes at end-stage of the disease (Fig. 3H-I). We assessed several functionally relevant genes in neutrophils, and found that in the different neutrophil clusters, most genes associated with neutrophil function and maturation including phagocytosis, chemotaxis, NETosis and granule formation, were expressed to a lower level in MM-bearing mice compared to naïve mice (Supplementary Figure S.9H). Overall, our data suggests that less mature neutrophils infiltrate the TME upon tumor-progression, which might lead to a more immunosuppressive environment.

To assess the translational value of our findings, we reanalyzed the scRNA-seq dataset generated by de Jong et al. [9] on human BM neutrophils from healthy individuals, NDMM and treated MM patients. We successfully aligned the neutrophil substates identified in our murine dataset with those found in the human dataset, which included precursors cells (myelocytes prolif, myelocytes, PreNeu1, PreNeu2), immature neutrophils (ImmNeu) and mature neutrophils (MatNeu1 to 5) (Fig. 3J-K). Consistent with our mouse data, human mature neutrophils in NDMM patients showed an upregulated IFN-associated signature compared to healthy individuals (Fig. 3L). Furthermore, the MatNeu4-IFN cluster (RSAD2, ISG15, IFIT1) was significantly increased in NDMM compared to healthy individuals. However, in contrast to our murine data, the percentage of immature neutrophils in human BM remained unchanged, while mature neutrophils increased in both NDMM and treated MM patients compared to healthy individuals (Fig. 3K, Supplementary Figure S.10A and B).

Conventional DCs (cDCs) show a less activated phenotype in MM-bearing mice

Having demonstrated an increase in the proportion of effector T cells within the lymphoid compartment in MM-bearing mice, we questioned which antigen presenting cells (APCs) within the 5T33MM-TME could potentially induce this T-cell phenotype. Hereto, we performed a differential NicheNet analysis to predict how MM modulates the interactions between the different APCs and (effector) T cells in the spleen, which harbors the highest numbers of T cells. Interestingly, while in naïve mice, neutrophils, monocytes, B cells and cDCs all displayed multiple predicted ligand-receptor interactions with T cells, in MM-bearing mice, the interactions with effector T cells were strongly skewed towards Ccr7+ activated cDCs (Supplementary Figure S.11A). Hence, cDCs might be important players in the generation of anti-tumor responses in the MM-TME. Remarkably, the role and function of various DC subsets remain poorly investigated in MM. Therefore, we explored our scRNA-seq data to unravel the heterogeneity and activation states of the DC compartment within the MM-TME and healthy tissue. Using the publicly available DC atlas provided by Liu et al. [24], we were able to annotate 16 clusters in the mouse MM DC compartment, including six progenitor clusters and 10 DC (or precursor-DC) clusters (which included precursor plasmacytoid DCs (pre-pDCs)), pDCs and cDCs (Fig. 4A, Supplementary Figure S.11B-C). The cDCs could be further subdivided into Ccr7+cDCs (also termed migDCs or mregDCs [25]), pre-cDC1s, cDC1s, pre-cDC2s, cDC2s, pro-DC3s and DC3s. Individual clusters were annotated based on the expression of known marker genes (Fig. 4A, Supplementary Figure S.11C).

Fig. 4

Conventional DCs (cDCs) show a less activated phenotype in MM-bearing mice. (A) UMAP plot shows high-resolution subclustering of the dendritic cell (DC) compartment within the murine scRNA-seq dataset. (B) Bar graph shows the frequency of the different DC subsets with the murine BM and spleen across different stages of disease, including naïve mice and 5T33MM mice at 14 DPI and 20 DPI, analyzed using the murine scRNA-seq dataset. (C-D) UMAP embedding as shown in panel A but colored according to (C) stage of disease; including naïve mice and 5T33MM mice at 14 DPI and 20 DPI, and (D) tissue of origin; including BM and spleen. (E–F) Bar graphs show the frequency of (E) pDCs and cDCs and (F) cDC subsets within the CD45+ cells in murine BM (top) and spleen (bottom) across different stages of disease, analyzed using flow cytometry. n = 5 per group. (G) Bar graphs show the ΔMFI (median fluorescence intensity) of CD80 in total cDCs derived from murine BM (top) and spleen (bottom) across different stages of disease, analyzed using flow cytometry. n = 5 per group. (H) UMAP plot shows high-resolution subclustering of the DC compartment within the human BM scRNA-seq dataset. (I) Bar graph shows the frequency of the different DC subsets within the human BM scRNA-seq dataset across different stages of disease. Error bars represent mean values ± SD. Statistical analysis was performed by Ordinary One-way ANOVA. ns: p ≥ 0.05, *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001

Within the DC compartment of the 5T33MM model, pDCs seemed to decrease, whereas the pDC_IFN cluster, which expressed additional IFN-associated genes (Ifit2, Ifi211, Rsad2, Irf7) and was enriched for gene ontology (GO) terms associated with “response to virus”, “innate immune response” and “regulation of type I interferon production”, was exclusively present in MM-bearing mice, and was most abundant at 14 DPI (Fig. 4C, Supplementary Figure S.11C–E). This increase in pDCs was also confirmed via flow cytometry at 14 DPI (Fig. 4E).

Among the cDCs, the pre-cDC1s, cDC1s, cDC2as, DC3s, and Ccr7+cDCs were more prominent in splenic tissue when compared to BM tissue, as the BM mainly included progenitor cells and precursor-cDCs (pre-cDC2 and pro-DC3) (Fig. 4B-D), which aligns with the fact that DCs arise in the BM. Only pDCs fully differentiate in the BM, whereas cDCs leave the BM in a precursor stage and differentiate into cDC subsets in peripheral tissues [26]. Strikingly, MM induced the recruitment of progenitors in the spleen (Fig. 4B), which is consistent with observations in other tumor models [27]. Flow cytometry analysis on the TME of end-stage 5T33MM mice revealed an increase (in BM) of all cDC subsets, which was very pronounced for cDC2s. More specifically, in the spleen, cDC1s decreased at 14 and 20 DPI, cDC2s percentages remained unaltered, and CCR7+cDCs, which contain both cDC1s and cDC2s, significantly increased at end-stage of disease (Fig. 4E-F).

Next, we assessed the functional state of the cDC subsets by flow cytometry and observed a significant increase in cDC-activation (CD80, CD86, CD40) at 7 DPI in most subsets (Fig. 4G, Supplementary Figure S.12A–D). At this critical time point, cancer cells homed and began to proliferate in hematopoietic sites (BM and spleen in 5T33MM-bearing mice) suggesting recognition and uptake by cDCs, inducing the activation of the latter. However, at 14 DPI and 20 DPI, the expression of activation markers decreased, indicating a suppressed/less activated phenotype across all cDC subsets upon tumor progression (Fig. 4G, Supplementary Figure S.12A–D). This was also observed, though to a lesser extent, at the transcriptome level (Supplementary Figure S.12E).

When correlating our murine dataset with the patient-derived dataset, based on the expression of canonical genes, we found that the abundance of precursor cells fluctuated between the different stages of disease (Fig. 4H and I, Supplementary Figure S.13A and B). Although mature cDC subsets were in a minority in murine BM, we could observe distinct well represented mature cDC subsets in human BM. In the diseased BM, we observed a massive increase in cDC1, cDC2s and DC3s, with DC3s being the most abundant population in RRMM patients (Fig. 4H-I). CCR7+cDCs, which can encompass mature cDC1s and cDC2s [25], were absent at precursor and early disease stages, but similarly to the mouse dataset, were found to be increased at NDMM and were also present at RRMM (Fig. 4I, Supplementary Figure S.13B). Very similar to the murine data, the pDC_IFN cluster was exclusively present in the BM of NDMM patients and also increased in RRMM patients. GO terms related to immune activation such as “response to virus”, ‘‘response to IFNα” and “response to IFNβ”, were highlighted in this subset, suggesting the induction of an immune response (Fig. 4H-I, Supplementary Figure S.13C and D).

In conclusion, our data highlights a parallel evolution of the DC heterogeneity within the human and murine MM-TME, with a reduced activation state of the cDCs in later stages of the disease.

Anti-CD40 agonist (αCD40) therapy induces DC activation and T-cell activation ex vivo

To address the reduced activation of cDCs at end-stage of disease, the use of DC-boosting therapies could be a promising solution. CD40 ligation was shown to activate DCs and to be required for cross-priming of CTL-responses by DCs [28, 29]. CD40 is therefore an emerging target for cancer immunotherapy and preclinical results using anti-CD40 (αCD40) agonist antibodies have indeed shown its potential in enhancing DC-activation and promoting sustained anti-tumor immune responses [30], thus offering a potential avenue for improving patients’ outcomes in several solid cancers. However, the effect and mechanism of αCD40 agonist therapy in MM remains to be elucidated.

To assess the effect of αCD40 on MM-derived cDCs, we collected BM and spleens from 5T33MM-bearing mice, which were sacrificed at 14 DPI, and incubated total BM and spleen samples for 24 or 72 h with αCD40 therapy (clone FGK4.5) ex vivo (Fig. 5A). αCD40 therapy significantly reduced tumor load in the spleen, as measured by the amount of viable idiotype+ plasma cells, and inhibited tumor cell proliferation at 72h (Fig. 5B-C, Supplementary Figure