GM-CSF treatment leads to impaired human EBI formation by decreasing adhesion molecule expression of CD163.
To begin to study the potential roles of GM-CSF on human EBI formation, we performed human EBI formation assay using “EBI-like” macrophages and late-stage erythroblasts derived from human cord blood CD34+ cells as previously described [17]. “EBI-like” macrophages were pretreated with EPO (as a control), or GM-CSF plus EPO, respectively. Then the two groups of “EBI-like” macrophages were co-cultured with day11 erythroblasts at a ratio of 1:20. Representative images of EBIs indicate that GM-CSF pretreated “EBI-like” macrophages surround fewer erythroblasts than control “EBI-like” macrophages (Figure 1A). Further analyses showed that the percentages of EBIs with 3,4, or 5 or more erythroid cells significantly decrease, while the percentage of EBIs with 0 or 1-2 erythroid cells significantly increases (Figure 1B). Previous studies have indicated that GM-CSF works by binding to its receptor for signal transduction. We checked the mRNA expression pattern of G-CSFR, GM-CSFR and M-CSFR from our published studies on EBI macrophages and the distinct stages of erythroblasts [12, 14, 17]. Interestingly, all of these three receptors are expressed by EBI macrophages (Figure 1C). In contrast, the expression levels of G-CSFR, GM-CSFR, and M-CSFR were very low or undetectable in late-stage erythroblasts (Figure 1C). This expression pattern was confirmed by QRT-PCR (Figure 1D). This data suggests that GM-CSF impairs EBI formation by affecting the function of EBI macrophages but not late-stage erythroblasts. Adhesion molecules of CD163, CD169, VCAM1, EMP, and αV-integrin are significant for EBI formation [24, 27, 36–38]. To define the mechanisms of the impaired EBI formation, we examined the effects of GM-CSF on adhesion molecules expression of CD163, CD169, EMP, and αV-integrin. Figure 1E shows that GM-CSF significantly decreases the mRNA expression of CD163 but not of CD169, EMP, and αV-integrin. Taken together, this data suggests that GM-CSF impairs EBI macrophages in supporting erythropoiesis at least in part by inhibiting interaction between EBI macrophages and erythroid cells via decreased adhesion molecule expression of CD163.
RNA-seq analysis suggests that GM-CSF treatment impairs the supporting function of human EBI macrophages during erythropoiesis.
To define transcriptional changes following GM-CSF treatment, we performed RNA-seq analysis. As expected, GM-CSF-treated human “EBI-like” macrophages clustered distinctly from control human “EBI-like” macrophages in principal component analysis (PCA; Figure 2A). Interestingly, the analysis revealed that human “EBI-like” macrophages exposed to GM-CSF for only 24 hours are already drastically distinct from control human “EBI-like” macrophages. We then performed pairwise comparison of differentially expressed genes. A heatmap of the differential expression of genes is shown in Figure 2B. A total of 1,722 genes are differentially expressed, of which 859 are up-regulated and 863 are down-regulated in GM-CSF-treated human “EBI-like” macrophages versus control human “EBI-like” macrophages (Figure 2C). Differentially expressed genes are listed in Supplementary Table 4. GSEA analysis of the differentially expressed genes revealed that the top upregulated pathways in GM-CSF-treated human “EBI-like” macrophages include Graft-versus-host disease (GvHD), inflammatory-mediator regulation of TRP channels, antigen processing and presentation, AMPK, PI3K-AKT, and cytokine-cytokine receptor interaction (Figure 2E). In contrast, the top down-regulated pathways in GM-CSF-treated human “EBI-like” macrophages are mostly related to Fc-gamma R-mediated phagocytosis, the VEGF signaling pathway, phagosomes, ECM receptor interaction, signaling pathway regulating pluripotent stem cells, and the chemokine signaling pathway. (Figure 2D).
Our previous research indicated that the expression levels of genes encoding proteins known to be important for the EBI macrophage function of supporting erythropoiesis include adhesion molecules, molecules for nucleus engulfment and digestion, iron recycling molecules, and growth factors [17]. GM-CSF significantly decreases adhesion molecule expression of CD163 but not of CD169, while the expression of MAEA and αV-integrin (enriched in ECM receptor interaction) slightly decreases upon GM-CSF treatment (Figure 3A). MERTK and MARCO are significant for the engulfment of EBI macrophages, while GM-CSF treatment decreases the expression of MERTK and MARCO (Figure 3B). HMOX1 is important for iron recycling, while GM-CSF treatment decreases the expression of HMOX1(Figure 3B). IGF1, IL-18, and VEGF-B are the main growth factors secreted by EBI macrophages for erythroblast proliferation, while there are no significant differences between control and GM-CSF-treated EBI macrophages (data not shown). Overall, GM-CSF impairs EBI macrophage functioning by decreasing adhesion between erythroblasts and EBI macrophages, decreasing engulfment, and decreasing iron recycling.
RNA-seq analysis suggests that GM-CSF treatment upregulates the immune regulatory function of EBI macrophages.
Our and other groups’ findings indicate that EBI macrophages are M2-like macrophages that support erythropoiesis via multiple mechanisms [17, 39]. In contrast, GM-CSF mainly stimulates the diverse functions of macrophages, including induction of MHC-class II and pattern recognition receptor (PRR) expression, antigen processing and presentation, cell adhesion and chemotaxis for leukocytes, migration for leukocytes, and so on[40, 41]. In the present study, we described that the GvHD pathway, antigen processing and presentation, and inflammatory-mediator regulation of TRP channels are enriched with GM-CSF treatment. Interestingly, of the enriched gene sets, expression of HLA-DRA, HLA-DRB1,3,4,5, HLA-DQB1, HLA-DPB1, and CD83 significantly increased upon GM-CSF treatment (Figure 3C). HLA-DRA, HLA-DRB, HLA-DQB, and HLA-DPB are the families of the HLA class, which plays a central role in the immune system and immune response by presenting peptides derived from extracellular proteins, in particular, pathogen-derived peptides to T cells. Importantly, the GvHD pathway, antigen processing and presentation, and inflammatory-mediator regulation of TRP channels enriched in GM-CSF treatment all include the HLA-DRA, HLA-DRB1,3,4,5, HLA-DQB1, and HLA-DPB genes. Additionally, C-C motif chemokine ligand 1(CCL1), CCL8 and CCL13 are involved in immunoregulatory and inflammatory processes that also increase after GM-CSF treatment, suggesting the immune regulatory function of GM-CSF treated “EBI-like” macrophages (Figure 3D). Thus, GM-CSF treatment upregulates the immune regulatory function of EBI macrophages.
GM-CSF treatment decreases the key transcription factors MAF and NR1H3.
Gene expression is regulated by transcription factors. MAF and NR1H3 are the two main selective transcription factors (TFs) expressed by EBI macrophages. ChIP-X Enrichment Analysis Version 3 (ChEA3) indicates that several key molecules significant for the function of EBI macrophages such as CD163, VCAM1, HMOX1,MERTK,AXL, and IGF1 may be regulated by NR1H3 or MAF (Supplementary Fig. 1; Keenan et al., 2019). We then analyzed the differentially expressed TFs between control and GM-CSF-treated EBI macrophages. Interestingly, MAF and NR1H3 dramatically decreased upon GM-CSF treatment (Figure 3E). Collectively, GM-CSF induced many gene changes that may be at least partially correlated with down-regulation of MAF and NR1H3.
GM-CSF treatment leads to decreased erythroblasts and EBI formation in mouse BM.
Having shown that GM-CSF significantly impairs human EBI formation in vitro, we then examined how GM-CSF affect erythropoiesis in vivo. Erythropoiesis is a process by which HSCs proliferate and differentiate via multiple distinct developmental stages, to eventually generate mature RBCs. BFU-E and CFU-E are well defined as lin−CD16−CD32CD41−CD34−Scal−CD117+CD71low and lin−CD16−CD32CD41−CD34−Scal−CD117+CD71high, respectively, by Harvey F. Lodish’s group [43]. We first stained control and GM-CSF-treated mouse BM cells and quantified BFU-E and CFU-E numbers. The gating strategy is shown in Supplementary Figure 2. Within the lin−CD16−CD32CD41−CD34−Scal− population, three populations are gated as I (CD117+CD71−), II (CD117+CD71low/medi), and III (CD117+CD71high, Figure 4A). Quantitative analysis indicated that GM-CSF treatment does not affect the numbers of either BFU-E (I) and CFU-E cells (III) in mouse BM (Figure 4B). Terminal erythroid differentiation was defined by Xiuli An’s group using Ter119 as the erythroid lineage marker in conjunction with CD44 and FSC [44]. The gating strategy is shown in Supplementary Figure 3. Using this method, erythroblast populations are clearly separated and named as Pro, Baso, Poly, and Ortho (Figure 4C). Figure 4C depicts four populations of control and GM-CSF-treated mouse BM. Quantitative analysis indicated that GM-CSF treatment significantly decreases erythroblast numbers (Figure 4D). EBI formation is significant for erythropoiesis [13,17]. To further examine EBI formation under GM-CSF treatment in vivo, we enriched EBI in both control and GM-CSF-treated mouse BM. Figure 4E showed representative EBI images of control and GM-CSF treatment. Quantitative analysis demonstrated that GM-CSF significantly decreases both EBI numbers (Figure 4F) as well as erythroblast numbers associated with EBI macrophages (Figure 4G). Collectively, GM-CSF treatment leads to decreased erythroblast numbers and EBI formation in mouse BM. Despite a reduction in the number of BM erythroblasts, EBI macrophages, and EBI, GM-CSF treatment did not contribute to development of an apparent peripheral blood anemia (Supplementary Figure 4A-D).
GM-CSF does not induce stress erythropoiesis in SP.
Stress erythropoiesis is characterized by increased numbers of erythroblasts in mouse SP. We therefore analyzed the erythropoiesis in mouse SP upon GM-CSF treatment. Firstly, we examined the SP index in control and GM-CSF-treated mice. Interestingly, the SP index does not change significantly following GM-CSF treatment (Supplementary Figure 5A). We then stained total SP cells to analyze terminal erythroid differentiation. Supplementary Figure 5B presents the representative flowcytometry images, which suggest that SP is still a non-erythropoietic organ upon GM-CSF treatment. The HE staining image using frozen control and GM-CSF-treated mouse SP slices reveal no significant difference, which confirms the flowcytometry data (Supplementary Figure 5C and D). Hence, GM-CSF does not induce stress erythropoiesis in SP.
GM-CSF treatment leads to decreased mouse BM EBI formation via decreased EBI macrophage numbers and CD163 and Vcam1 adhesion molecule expression
To gain further insight into inhibited erythropoiesis in mouse BM upon GM-CSF administration, we analyzed EBI formation in mouse BM. EBI formation is dependent on the interaction of EBI macrophages and erythroblasts. We then examined the numbers of EBI macrophages. We stained control and GM-CSF-treated mouse BM cells with Gr1, CD11b, and F4/80. The gating strategy is shown in Supplementary Figure 6. Figures 5A and B indicate that GM-CSF treatment significantly decreases the numbers of EBI macrophages and the expression of F4/80. GM-CSF binds to GM-CSFR on macrophages to induce STAT5 phosphorylation for signaling transduction and then affects macrophage functioning. We then examined GM-CSF/GM-CSFR signaling in BM macrophages in both the control and GM-CSF-treated samples. Flowcytometry analysis revealed that GM-CSF significantly increases Stat5 phosphorylation of BM macrophages (Figure 5C). Adhesion molecules CD163, CD169, and Vcam1 are significant for EBI formation in mouse BM. We then analyzed the CD163, CD169 and Vcam1 adhesion molecules expressed by EBI macrophages from control and GM-CSF-treated groups. Interestingly, the expression of CD163 and Vcam1 but not of CD169 significantly decreases upon GM-CSF treatment (Figures 5D, E, F, G, H, and I). Our findings demonstrate that GM-CSF impairs EBI formation at least in part by decreasing the interaction between EBI macrophages and erythroid cells via decreased EBI macrophage numbers and CD163 and Vcam1 surface expression.
Macrophage depletion with clodronate-loaded liposomes leads to decreased erythroblast numbers in BM
Having shown that GM-CSF treatment leads to decreased erythroblast numbers in BM as well as decreased the numbers of EBI macrophages. We then examined whether erythroblast numbers in BM can be affected following macrophage depletion with clodronate-loaded liposomes. Consistent with previous studies, clodronate-loaded liposomes induced total macrophages depletion in mouse[27, 45]. Supplementary Figure 7A and B showed that erythroblast numbers significantly decrease in clodronate-loaded liposomes treated mouse BM. Collectively, decreased erythroblast numbers in mouse BM under GM-CSF treatment may be due to reduction in the numbers of EBI macrophages at least in part.
GM-CSF treatment leads to decreased expression of Mertk, Axl, and Timd4 on mouse BM EBI macrophages as well as phagocytosis of senescent RBCs.
Having shown that GM-CSF treatment significantly decreases the expression of MERTK on human EBI-like macrophages in vitro, we also examined the expression of Mertk, Axl, and Timd4 on mouse BM EBI macrophages in vivo. Consistent with the in vitro study, GM-CSF treatment significantly decreased the expression of Mertk, Timd4 and Axl on mouse BM EBI macrophages (Figures 6A-F). These data indicated that GM-CSF may impair the phagocytosis of EBI macrophages during mouse BM erythropoiesis. To examine whether GM-CSF decreases phagocytosis of senescent RBCs by EBI macrophages in vivo, we performed a phagocytosis assay. Phagocytosis of senescent RBCs was defined by Jessica A. Hamerman’s group [28]. Among Gr1−CD11b− F4/80+ macrophages that had internalized RBCs (as determined via intracellular anti-Ter119 staining, Figure 6G), compared with control mice, GM-CSF-treated mice showed decreased Ter119+ percentages (Figure 6H), suggesting decreased phagocytosis in senescent RBCs among GM-CSF-treated mice versus control mice.
GM-CSF treatment leads to the polarization of BM EBI macrophages from M2-like to M1-like phenotype.
Having shown that GM-CSF treatment induces immune regulatory functions in human EBI macrophages in vitro, to examine whether GM-CSF can affect the phenotype of EBI macrophages in vivo, we performed an experiment to detect the expression of M1 (MHC-II, CD14 and CD80) and M2 (CD86, CD206, CD163) surface markers on mouse BM EBI macrophages using flowcytometry. Notably, GM-CSF significantly increased the expression of MHC-II (Figures 7A and F) but decreased the expression of CD206 (Figures 7B and F). In addition, GM-CSF did not affect CD86 (Figures 7C and F), CD14 (Figures 7D and F), or CD80 expression. (Figures 7E and F). GM-CSF significantly decreased the expression of CD163. Previous studies have indicated that GM-CSF can induce myeloid cells to secrete several inflammatory cytokines, such as IL-1 and TNF-α. We therefore examined the expression of the M1 and M2 cytokines iNOS and Arg1 in control and GM-CSF-treated BM F4/80+ macrophages using QRT-PCR. Supplementary Figure 8 reflects that the expression of iNOS, IL-1β, and TNF-α significantly increased. In contrast, the expression of Arg1, IL-10, and TGF-β significantly decreased. Accordingly, GM-CSF leads EBI macrophages to assume a more M1-like phenotype compared with control EBI macrophages.