Patient characteristics
We used BM samples from 44 patients with WM (n = 35) or IgM MGUS (n = 9), and 5 healthy control subjects. Patient characteristics are summarized in Supplementary Table 1. WM patients were further divided into 3 groups – MGUS/smoldering WM (n = 8), symptomatic WM (n = 31) and previously treated WM patients in remission (n = 5), based on their clinicopathological diagnosis that included serum IgM and ß2 microglobulin levels, treatment, and the International Prognostic Scoring System for Waldenstrom Macroglobulinemia (IPSSWM). Samples were collected at diagnosis in 36 patients and after one line of therapy in 5 patients with WM (median time from last therapy to sample collection 44 months, range 6–91 months).
PMN-MDSCs are significantly enriched in the WM bone marrow microenvironment.
To determine the prevalence of MDSCs in the BM microenvironment of WM patients, we first performed flow cytometry on negatively sorted CD19− CD138− BM cells from patients with WM (n = 18) and NBM (n = 12). We analyzed the percentage of MDSCs (Lin− HLA-DRlowCD11b+CD33+) and found that MDSCs were significantly expanded in the BM of WM patients as compared to NBM (p = 0.0009; Fig. 1A and B). CD11b+CD33+ cells were further classified by CD14 and CD15 expression to define M-MDSCs (CD14+) and PMN-MDSCs (CD15+) (Fig. 1A). We observed that both PMN-MDSCs (p = 0.019) and M-MDSCs (p = 0.063) were increased in WM-BM samples as opposed to NBM samples. The median percentage of PMN-MDSCs, relative to the total MDSCs population (i.e. CD11b + CD33+), in WM patients was 48.5% (range:0.66–93.3%) as compared to NBM patients which was 9.66% (range:0.69- 84%). Further, M-MDSCs, the median percentage and range in WM and NBM was 21.4% (range:11.22 to 68.1%) and 15% (range:1.31 to 28.9%) respectively (Fig. 1C).
Three phenotypically distinct subsets of MDSCs are identified in the WM bone marrow.
Next, we explored the phenotype of BM MDSCs using mass cytometry (CyTOF) in NBM (n = 3), MGUS/smoldering WM(n = 4), symptomatic WM (n = 7), as well as WM patients in remission post treatment (n = 4). Interestingly, CD11b+CD33+ MDSCs were phenotypically distinct in three subsets (Fig. 2A and 2B). One subset expressed high levels of CD66b, were CD15+ and CD16+, but lacked expression of CD68. These CD66b+ cells had a phenotype consistent with PMN-MDSCs and were more expanded in symptomatic WM patients compared to NBM, MGUS/smoldering WM and WM patients in remission (Fig. 2A and 2B). A second MDSC subset had bright expression of CD15 and CD68, with some expression of CD45RO, but had low expression of CD66b. This subset seemed to remain quantitatively constant in NBM (13.37%), MGUS/Smoldering WM (10.26%) and symptomatic WM samples (22.37%), even though was less frequent in previously treated patients (1.25%). A third subset had moderate expression of CD68 but did not express CD66b, CD15 and CD16. This subset highly expressed in NBM (68.42%), MGUS/smoldering WM (77.42%) and patients in remission (52.36%) while decreased in symptomatic WM patients (5.50%) (Fig. 2B). We then isolated CD66b+ and CD66b− cells to further define their phenotypes. The CD66b+ subset demonstrated characteristics of PMN-MDSCs with elevated CD15 expression (Fig. 2C). Conversely, the CD66b− population exhibited traits consistent with M-MDSCs exhibiting increased expression of CD14 (Fig. 2D).
MDSCs exhibited upregulated inflammation and metabolic signatures.
To further characterize MDSCs, we performed CITE-seq analysis on BM cells from patients with WM (n = 3) or NBM (n = 2). Unsupervised clustering and uniform manifold approximation and projection (UMAP)(19) identified 15 distinct clusters (Fig. 3A-B). Each cluster underwent meticulous annotation based on their discernible phenotypes. Notably, Clusters 2, 3, and 9 displayed CD11b and CD33 expression, characteristic of the MDSCs phenotype. Further exploration highlighted a shared gene signature abundant in S100 Calcium Binding Protein A8 (S100A8), S100A9, Transforming Growth Factor Beta 1 (TGFB1), Colony Stimulating Factor 3 Receptor (CSF3R), Lysozyme (LYZ), Fos Proto-Oncogene, AP-1 Transcription Factor Subunit (FOS), and C-C Motif Chemokine Ligand 2 (CCL2) across these clusters (Fig. 3B and Supplemental Fig. 21). The detailed gene signatures within these clusters further elucidate their distinctive molecular profiles: Cluster 2 exhibited prominent expression of IL13RA1, IL17RA, IL1B, IL10RB, STAT1, STAT6, CSF1R, CSF3R. Cluster 3 displayed a distinct gene signature, including STAT2, STAT6, CSF1R, CSF3R, SOD2, and CD84. Cluster 9 expressed AZU1, MPO, CD163, FOS, CD36, LYZ, and CCL2 as part of its gene signature (Supplemental Fig. 2).
Differential gene expression analysis of WM MDSCs as compared to NBM MDSCs identified 146 upregulated and 94 downregulated genes (p ≤ 0.05). A detailed list of genes is provided in Supplemental data 3. Briefly, THBS1, FOS, HIF1A, JunB, CD86, IL13RA1 and TGFBI, were significantly upregulated in WM-MDSCs. These genes have been associated with the development of MDSCs (20–24). On the other hand, glycolysis related genes including fructose-bisphosphatase 1(FBP1) and enolase 1(ENO1), and genes regulating the citric acid cycle namely, ATP synthase membrane subunit c locus 2(ATP5MC2), ATP synthase membrane subunit e (ATP5ME) and f (ATP5MF) were downregulated, indicating an inhibition of metabolic pathways (25). Additionally, gene set enrichment analysis (GSEA) showed positive enrichment for interferon gamma signaling, TNFA signaling, and an inflammatory response in WM (Fig. 3D-E). All these pathways may promote MDSC generation and expansion (26). On the other hand, we observed downregulation of cell metabolism signature, namely oxidative phosphorylation, and fatty acid metabolism in WM patients, which may promote tumor growth.
PMN- MDSCs exhibit distinct immune signature associated with poor outcome.
Given that PMN-MDSCs (CD66b+/CD15+) are expanded in symptomatic WM patients, we focused our analysis to better characterize the CD66b+ PMN-MDSCs cluster. Differential gene expression between CD66b+ PMN-MDSCs and CD66b− M-MDSCs identified 249 upregulated and 100 downregulated genes. CD66b+ MDSCs were characterized by genes involved in inflammation and growth, such as proteinase 3 (PRTN3), CXCR4, SOX4, IL2RG, and insulin-like growth factor-binding protein 7 (IGFBP7) (Fig. 4A). We also found upregulation of nuclear proliferative markers including MKI67, cell cycle and antiapoptotic genes including survivinHMGB1, HMGB2, and immune suppressive markers such as myeloperoxidase (MPO), Azurocidin 1 (AZU1) (27), Elastase, Neutrophil Expressed (ELANE) and centromere Protein F (CENPF)]. Additionally, CD59 (an inhibitory gene for the complement system) and LDHB were also significantly upregulated (Fig. 4A). Notably, these genes have been previously associated with MDSC growth and expansion (28–30). GSEA showed a positive enrichment for gene sets related to cell cycle, mitotic spindle, and DNA replication, suggesting a growth pressure. In contrast, there was a negative correlation for antigen presentation, genes involved in cytoskeleton structure and oxidative phosphorylation, supporting putative immune escape mechanisms. Altogether, this data shows that CD66b+ PMN-MDSCs have a distinct transcriptomic profile with significant enriched for genes associated with cell growth and proliferation, suppression of antigen presentation and oxidative phosphorylation, which may be collectively part of the immune suppressive mechanisms of PMN-MDSCs.
CD66b + PMN-MDSCs mediated T-cell suppression inhibits TNFα and IFN-γ production.
To investigate the immunosuppressive role of CD66b+ PMN-MDSCs compared to CD66b− M-MDSCs in WM patients, we enriched the MDSCs extracted from 3 representative WM patients and separated them into CD66b+ PMN and CD66b− M-MDSCs. These cells were then co-cultured with sorted T-cells from healthy donors. After 24h of co-culture, we found a substantial decrease in mean fluorescence intensity (MFI) of CD69 expression on T cells (74.34% median decrease, range 73.02% − 78.1% n = 3). While co-culture of T-cells with CD66b− MDSCs also downregulated CD69 expression, the extent of immune suppression (median: 42.9% was inferior to CD66b+ MDSC (p < 0.0001 - Fig. 5A). We also analyzed the cytokine (IFN-γ and TNF-α) production by CD4+ and CD8+ T cells when cocultured with CD66b+ or CD66b− MDSCs. As shown in Fig. 5F, we found that CD66b+ MDSCs downregulated IFN-γ expression in CD4+ (Fig. 5B) and CD8+ (Fig. 5D) T cells with a 49.8% (p < 0.0001) and 44.4% (p < 0.0001) decrease in MFI, respectively. CD66b− MDSCs showed a lower suppressive effect on cytokine production than CD66b+ MDSCs and downregulated IFN-γ in CD4+ and CD8+ T-cells by only 7.67% and 1.9%, respectively. A similar pattern of TNFα downregulation in CD4+ (Fig. 5C) and CD8+ (Fig. 5E) T cells was observed with the co-culture of CD66b+ MDSCs. (Fig. 5G). Altogether, these results suggest that CD66b + PMN-MDSCs possess profound immunosuppressive properties in WM patients.
G-CSF and TNF -alpha promotes growth of CD66b + PMN-MDSCs.
We then evaluated potential cytokines that may account for the expansion of the CD66b+ PMN-MDSC population. We had previously shown that granulocyte colony-stimulating factor (G-CSF) is increased in WM BM serum compared to that in healthy individuals (31). Based on this finding and on our identified upregulation of TNF signaling in PMN-MDSCs, we investigated the potential role of G-CSF and TNFα in expanding CD66b+ MDSCs in the WM BM microenvironment. To do this, we enriched CD66b + cells from WM patients (n = 3), treated them with G-CSF and TNFα for 24 and 48 hours and then measured the numbers of CD15+/CD66b+ (Lin− CD11b+ CD33+) PMN-MDSCs by flow cytometry. After treatment with G-CSF (10ng/ml) for 24 h, we found that the PMN-MDSCs population expanded by 35.8% (p < 0.001) (Fig. 6A and B). This population further increased by 45.97% after 48 h (p < 0.001). Similarly, TNFα treatment also increased the CD66b + PMN-MDSC population by 41.29% (p < 0.001) after 24 h, but this percentage remained about the same after 48 h of treatment (42.34%, p < 0.001; Fig. 6A and B). These findings suggests that G-CSF and TNFα promote PMN-MDSCs expansion and survival.
Lymphoplasmacytic cells of WM patients recruit PMN-MDSCs in the bone marrow.
While we found that GCSF and TNFα promote the expansion of PMN-MDSCs, we also evaluated whether WM B-cells play a role in attracting PMN-MDSCs to the BM. Therefore, to investigate whether WM cells recruit MDSCs to the BM, we conducted a transwell migration assay of MDSCs in response to the presence of malignant BCWM.1 cells. Our results showed that BCWM.1 cells exhibited a greater capacity to attract MDSCs when compared to normal B-cells or control chemoattractants including 10% FBS and MCP-1 (p = 0.04; Fig. 6D). Additionally, we sought to determine which subtype of MDSCs was predominantly attracted by malignant B-cells. To address this, we assessed the expression of CD14 and CD15 on migrating MDSCs to delineate PMN-MDSCs and M-MDSCs. We found that BCWM.1 cells attracted PMN-MDSCs to a greater degree than M-MDSCs (Fig. 6E). These findings indicate that WM cells exhibit robust chemotactic properties for MDSCs, with a significant preference for attracting PMN-MDSCs. These observations suggest that the high prevalence of PMN-MDSCs within the bone marrow of WM patients is in part due to recruitment of these cells to the BM by malignant B-cells.
Increased PMN-MDSCs are associated poor prognostic factors.
In addition to the biological effect, the increased prevalence of CD66 + PMN-MDSCs showed clinical relevance. Clinical data from 14 WM patients was analyzed to assess the correlation of CD66b+ PMN-MDSCs with clinical features. The average percentage of CD66b+ PMN-MDSCs was employed to establish parameters for low and high levels. As shown in Fig. 6F, both a shorter time to next treatment (TTNT) (p = 0.03) and higher clinical risk as determined by a higher ISSWM score (p = 0.026) were significantly correlated with an increased prevalence of PMN-MDSCs in the BM.