Cellular Differentiation of Human Monocytes Is Regulated by Time-Dependent Interleukin-4 Signaling and the Transcriptional Regulator NCOR2

Abstract

Human in vitro generated monocyte-derived dendritic cells (moDCs) and macrophages are used clinically, e.g., to induce immunity against cancer. However, their physiological counterparts, ontogeny, transcriptional regulation, and heterogeneity remains largely unknown, hampering their clinical use. High-dimensional techniques were used to elucidate transcriptional, phenotypic, and functional differences between human in vivo and in vitro generated mononuclear phagocytes to facilitate their full potential in the clinic. We demonstrate that monocytes differentiated by macrophage colony-stimulating factor (M-CSF) or granulocyte macrophage colony-stimulating factor (GM-CSF) resembled in vivo inflammatory macrophages, while moDCs resembled in vivo inflammatory DCs. Moreover, differentiated monocytes presented with profound transcriptomic, phenotypic, and functional differences. Monocytes integrated GM-CSF and IL-4 stimulation combinatorically and temporally, resulting in a mode- and time-dependent differentiation relying on NCOR2. Finally, moDCs are phenotypically heterogeneous and therefore necessitate the use of high-dimensional phenotyping to open new possibilities for better clinical tailoring of these cellular therapies.

Keywords: IL-4; IL-4 activated macrophages; M(IL-4); NCOR2; activation; human; inflammatory dendritic cells; inflammatory macrophages; macrophages; monocyte-derived dendritic cells; monocytes.

Figure 1

Relationship of In Vitro Activated Monocyte-Derived Cells (A and B) PCA (21,250 present probes); displayed principal components (PCs): (A) 1 versus 2 and (B) 1 versus 3. (C) Heatmap of 1,000 most variable genes in dataset. Log₂-expression values, z-transformed, scaled (−2 [blue] to 2 [red]). (D) Heatmap, Pearson correlation values (PCV) calculated pairwise between all cell types (top 1,000 most variable genes). (E) PCA (23,952 present probes). (F) Relative fractions of MO, BDCA1⁺ DC, infM, and infDC gene signatures in CD14⁺ MOs, different MO-derived cells, DCs. (G and H) Heatmaps of genes specifically expressed in (G) infDCs compared to infMs, BDCA1⁺ DCs, MOs (dataset 1), and MOs-GM-CSF^IL-4(0-72h) versus MO-derived cells, CD14⁺ MOs, and DCs (dataset 2), or in (H) infMs compared to infDCs, BDCA1⁺ DCs, and MOs (dataset 1), and in MOs-GM-CSF and MOs-M-CSF versus MOs-GM-CSF^IL-4(0-72h), CD14⁺ MOs, and DCs (dataset 2). PCVs between indicated group patterns of dataset 1 versus dataset 2, barplot next to heatmaps (correlation cutoff > 0.4). Genes analyzed in (J) and (K) highlighted in red. Log₂-expression values, z-transformed, scaled (−2 [blue] to 2 [red]). (I) Histograms (flow cytometry analysis), relative expression CD226, MARCO, VSIG4, and CCR7 (representative data, n = 4). (J) Analysis of cell culture supernatants of MOs-M-CSF, MOs-GM-CSF, and MOs-GM-CSF^IL-4(0-72h) for CCL22 and CCL2 using ELISA (n = 3, 2 technical replicates each, mean + SEM, one-way RM [repeated-measures] ANOVA, Tukey’s method for multiple test correction, with ^∗p < 0.05, ^∗∗p < 0.01, and ^∗∗∗p < 0.001; n.d., not detected). (K) MMP12 quantification (relative) in CD14⁺ MOs, MOs-M-CSF, MOs-GM-CSF, and MOs-GM-CSF^IL-4(0-72h) by immunoblot (n = 3, mean + SEM, one-way ANOVA and Tukey’s method for multiple test correction, with ^∗p < 0.05). Please also see Figure S1.

Figure 2

MOs-GM-CSF^IL-4 Are Most Distinct from MOs-M-CSF and MOs-GM-CSF (A) Schema describing the questions addressed here and in Figure S2. (B) PCA (18,318 present probes). (C) Heatmap of 1,000 most variable genes in dataset. Log₂-expression values, z-transformed, scaled (−2 [blue] to 2 [red]). (D) Heatmap of specifically expressed genes in a single out of the three MO-derived cells versus CD14⁺ MOs and the other two MO-derived cell types. Log₂-expression values, z-transformed, scaled (−1.5 [blue] to 1.5 [red]). (E) Co-expression networks (union of 2,086 DEG [fold-change > 2 or < −2 and FDR-adjusted p value < 0.05]) between each of three MO-derived cells types versus CD14⁺ MOs. Fold-change of respective cell type versus overall mean mapped onto networks and displayed blue (negative fold-change) over white to red (positive fold-change). Based on fold-change patterns, networks were divided into four clusters, each cluster representing one of four cell types. (F) Co-expression network (411 TRs expressed in dataset). For each cell type, fold-change of respective cell type versus overall mean mapped onto network. Cell type-specific clusters of upregulated regulators were generated, indicated by color-coded shadings behind network. TRs highlighted in red were predicted as unique master regulators of corresponding cell type. Prediction performed on all genes highlighted in red (fold-change > 1.5 over overall mean) in corresponding cell type-specific cluster in (E). (G) t-SNE display of CD14⁺ MOs, MOs-M-CSF, MOs-GM-CSF, and MOs-GM-CSF^IL-4(0-144h) analyzed by MC (n = 3). (H) Heatmap and HC of mean surface marker expression analyzed using MC. Normalized intensity values, z-transformed, scaled (−6 [blue] to 6 [red]). Color code depicts cluster assignment according to culture condition. Color code as in (G). Please also see Figure S2.

Figure 3

Prediction of Differentiated MO Functionality (A and B) Flow cytometry analysis of MOs-M-CSF, MOs-GM-CSF, and MOs-GM-CSF^IL-4(0-72h) after incubation with (A) GFP-expressing yeast (1 hr, histogram: representative result [n = 3], bar plot [n = 3], mean + SEM, one-way RM ANOVA, with p > 0.05) or (B) YG beads (4 hr, n = 5–6, mean + SEM, one-way RM ANOVA and Tukey’s method for multiple test correction, with ^∗∗∗p < 0.001). (C) Migration tracks (3 hr) of MOs-M-CSF, MOs-GM-CSF, and MOs-GM-CSF^IL-4(0-72h) (representative result, n = 3). (D and E) OCR (D) and ECAR (E) of MOs-M-CSF, MOs-GM-CSF, and MOs-GM-CSF^IL-4(0-72h). (F) Heatmap of mean secreted cytokine concentrations (n = 4). Expression values, z-transformed, scaled (–3 [blue] to 3 [red], raw data, Table S3). Please also see Figure S3.

Figure 4

MOs-GM-CSF^IL-4(0-144h) Differ from MOs-GM-CSF^{IL-4(72-144h)} MO-Derived Cells (A) Schema describing questions addressed herein and in Figure S4. (B) PCA (18,857 present probes). (C) Co-expression network (13,691 present genes) describing relationships between CD14⁺ MOs and four types of MO-derived cells. (D) Heatmap of 1,000 most variable genes in dataset. Log₂-expression values, z-transformed, scaled (−2 [blue] to 2 [red]). Highly expressed genes grouped together (black boxes) according to cell type. Corresponding group-related cell types highlighted, left side of heatmap. Important genes of each cluster depicted, right side of heatmap. (E) Flow cytometry analysis of MOs-GM-CSF^{IL-4(72-144h)} and MOs-GM-CSF^IL-4(0-144h) after incubation with GFP-expressing yeast (1 hr, n = 4–6, mean + SEM, Student’s t test with ^∗p < 0.05). (F) Migration tracks (3 hr) of MOs-GM-CSF^{IL-4(72-144h)} and MOs-GM-CSF^IL-4(0-144h) (representative result, n = 3). Please also see Figure S4.

Figure 5

Timing of IL-4 Determines Transcriptional Regulation in Differentiated MOs (A) Schema describing the IL-4 time kinetic experiment. (B and C) Histograms, relative expression of CD14 and CD209 analyzed by flow cytometry. (D) PCA (12,794 present genes). (E) Heatmap of 1,000 most variable genes across dataset. Log₂-expression values, z-transformed, scaled (−2 [blue] to 2 [red]). Below, SOM-clustering (12,794 present genes across cell types). (F) Co-expression networks (union of 2,775 DEGs, fold-change > 1.5 or < −1.5 and FDR-corrected p value < 0.05) between MOs-GM-CSF^IL-4 and MOs-GM-CSF. For each cell type, fold-change of respective cell type versus overall mean mapped onto networks, displayed in blue (fold-change ≤ 1.5) or red (fold-change ≥ 1.5). (G) Example of genes located in condition-related clusters depicted in (F) and in first column (fold-change ≥ 1.5). First column: condition-specific genes; following columns: genes shared between clusters of two consecutive time points. Please also see Figure S5.

Figure 6

NCOR2 Is a Transcriptional Regulator of MOs-GM-CSF^{IL-4(0-72/144h)} (A and B) Co-expression networks (267 TRs) for (A) MOs-GM-CSF, (B) MOs-GM-CSF^IL-4(0-72h), fold-change versus CD14⁺ MOs mapped onto network. MOs-GM-CSF^IL-4(0-72h)-specific cluster of elevated regulators (dark blue). (C) TR Heatmap, specifically upregulated in MOs-GM-CSF^IL-4(0-72h), MOs-GM-CSF^IL-4(0-144h) versus CD14⁺ MOs, MOs-GM-CSF. Log₂-expression values, z-transformed, scaled (−1.15 [blue] to 1.15 [red]). (D) Immunoblot of NCOR2, Lamin A/C, and β-tubulin in cytoplasm and nucleus of MOs-GM-CSF and MOs-GM-CSF^IL-4(0-144h) (representative result, n = 4). (E) Quantification of relative enrichment of NCOR2 in nuclear fractions isolated from MOs-GM-CSF and MOs-GM-CSF^IL-4(0-144h) (n = 4, ^∗p < 0.05, mean ± SEM). (F) Representative confocal microscopy images of medial nuclear region (MNR) of MOs-GM-CSF and MOs-GM-CSF^IL-4(0-144h) (n = 3, green: NCOR2; blue: DAPI; red: tubulin). (G) Quantification of mean fluorescence intensity of NCOR2 in MNR identified by confocal microscopy of GM-CSF and MOs-GM-CSF^IL-4(0-144h) (n = 3, green: NCOR2; blue: DAPI; red: tubulin, ^∗p < 0.05, mean ± SEM). (H) Scatterplot (1,834 variable genes in dataset) containing cells treated with αNCOR2 (y axis), scrambled siRNA (x axis, log₂-mean expression values). Highlighted genes determined (see Figure S6F) induced (red) or repressed (blue) by IL-4. (I and J) GSEA of genes upregulated (I), downregulated (J) in IL-4 signature in MOs-GM-CSF^IL-4(0-72h) treated with scrambled or αNCOR2 siRNA (nominal p value, empirical phenotype-based permutation test [p < 0.05, FDR < 0.25], 1,000 samples permutations). Please also see Figure S6.

Figure 7

Mass Cytometry Analysis Identifies Unappreciated Phenotypic Heterogeneity in Clinically Relevant Mo-GM-CSF^IL-4(0-144h) Cultures (A) Phenograph of CD14⁺ MOs, MOs-M-CSF, MOs-GM-CSF, and MOs-GM-CSF^IL-4(0-144h) based on MC expression data (n = 3, 36 myeloid-related surface markers). Affiliation of cells to the 11 identified clusters indicated by color coding and visualized in t-SNE plot. (B) Heatmap and HC of mean surface marker expression of 11 individual clusters. Right side: Differentiating conditions according to (A) and Figure 2G. (C) Phenograph of MOs-GM-CSF^IL4(0-144h) and visualized in t-SNE plot (representative donor; 3,500 cells). (D) Heatmap and HC of mean surface marker expression of 11 individual clusters. (E) Expression feature plot of the depicted surface markers in MOs-GM-CSF^IL-4(0-144h).