EKLF/KLF1 expression defines a unique macrophage subset during mouse erythropoiesis

Abstract

Erythroblastic islands are a specialized niche that contain a central macrophage surrounded by erythroid cells at various stages of maturation. However, identifying the precise genetic and transcriptional control mechanisms in the island macrophage remains difficult due to macrophage heterogeneity. Using unbiased global sequencing and directed genetic approaches focused on early mammalian development, we find that fetal liver macrophages exhibit a unique expression signature that differentiates them from erythroid and adult macrophage cells. The importance of erythroid Krüppel-like factor (EKLF)/KLF1 in this identity is shown by expression analyses in EKLF-/- and in EKLF-marked macrophage cells. Single-cell sequence analysis simplifies heterogeneity and identifies clusters of genes important for EKLF-dependent macrophage function and novel cell surface biomarkers. Remarkably, this singular set of macrophage island cells appears transiently during embryogenesis. Together, these studies provide a detailed perspective on the importance of EKLF in the establishment of the dynamic gene expression network within erythroblastic islands in the developing embryo and provide the means for their efficient isolation.

Keywords: EKLF/Klf1; developmental biology; erythropoiesis; fetal liver island macrophage; mouse; single cell sequencing; transcription factors.

Figure 1.. Gene expression comparison of fetal liver (FL) F4/80+ macrophages with extensively self-renewing erythroblasts (ESREs; Erythr) and adult spleen (Spl) F4/80+ macrophages showing unique gene expression in F4/80+ FL macrophages.

(A) Hierarchical clustering dendrogram using scaled Z-scores based on the expression of the top 10,000 highly expressed genes is shown for individual RNA-Seq biological replicates from each cell type (source data: Figure 1—source data 1). (B) Principal component analysis of the cell types is plotted showing principal components 1 and 2 for each biological replicate (source data: Figure 1—source data 1). (C) Macrophage-specific or erythroid-specific marker expression in the cell types is shown, with replicates averaged together (source data: Figure 1—source data 3). (D) k-means clustering of individual RNA-Seq biological replicates of the different cell types (ESREs, Erythr; fetal liver, FL; spleen, Spl) by log2 FPKM displayed as a heatmap (source data: Figure 1—source data 4). Flower bracket indicates the gene cluster with enriched expression in F4/80+ FL macrophages. (E) Heatmap of only the uniquely expressed genes in F4/80+ FL macrophages that define the signature genes of this cell type (source data: Figure 1—source data 2). A few representative signature gene names are displayed.

Figure 1—figure supplement 1.. Isolation of a pure population of F4/80+ E13.5 fetal liver cells by FACS.

(A) Gating strategy used to stringently sort singlets that are F4/80+ is shown from unstained (no antibody [Ab]) or treated (anti-F4/80 Ab) fetal liver cells from EKLF+/+ E13.5 fetal livers. (The same approach was used for EKLF-/- cell sorting in Figure 3A.) (B) FACS-sorted cells using the strategy above were cytospun on a slide and observed after May-Grunwald-Giemsa staining for singlets, doublets, or >3 cells. 25 µm scale bars are indicated. (C) Quantification of data from (B) (source data: Figure 1—figure supplement 1—source data 1 ).

Figure 2.. EKLF/Klf1 is expressed in fetal liver macrophages during development.

(A) Immunofluorescence tests with anti-EKLF (white), 4′,6-diamidino-2-phenylindole (DAPI) (blue), and anti-F4/80 (red) antibodies in E13.5 fetal liver cells. (A) White arrowheads show coexpression of EKLF and F4/80 proteins in single cells (representative of over 20 EKLF+/F4/80+ cells in this field of 300 cells); red arrow shows that not all F4/80+ cells are EKLF+. (B) White arrow shows that not all EKLF+ cells are F4/80+ , as expected from the FACS data (cytoplasmic EKLF signal is expected [Quadrini et al., 2008; Schoenfelder et al., 2010]). (C) Collated RNA-Seq data (Mass et al., 2016) of sorted macrophage cells from multiple staged embryonic (E) day 10.25–16.5 fetal livers or postnatal (P) day 2–21 livers (ckit-/CD45+/F480+/AA4.1-/CD11b+; n = 24 samples) show transient and abundant Klf1 reads (UCSC Genome Browser). (D) Same analysis as (C) showing RNA-Seq reads of the gene encoding F4/80 (Adgre1) as a positive control across all samples.

Figure 3.. EKLF-dependent gene expression in fetal liver (FL) macrophages.

(A) A representative yield of cells from EKLF-/- FL sorted by F4/80 expression, used for RNA-Seq analysis, is shown (compare to WT yield in Figure 1—figure supplement 1A). (B) k-means clustering of absolute log2 FPKM of F4/80+ EKLF+/+ and F4/80+ EKLF-/-, and log2 FKPM ratio EKLF-/-(KO)/WT is displayed as a heatmap. Flower bracket indicates downregulated genes. (C) Differentially expressed genes in EKLF-/- (KO) compared to WT shown as a volcano plot (source data: Figure 3—source data 1).

Figure 4.. Comparison of gene expression in F4/80+ EKLF/GFP+ and F4/80+ EKLF/GFP- fetal liver macrophages.

(A) Principal component analysis using scaled Z-score based on the expression level of the top 10,000 highly expressed genes from RNA-Seq replicates of F4/80+ EKLF/GFP+ and F4/80+ EKLF/GFP- is plotted with each axis depicting the two major principal components (source data: Figure 4—source data 1). (B) Scatterplot showing the significantly enriched genes in the F4/80+ EKLF/GFP+ population compared to F4/80+ EKLF/GFP-. Vcam1, Klf1, and Epor are highlighted in blue (source data: Figure 4—source data 2). Fragments per kilobase million (FPKM) values of (C) EKLF/Klf1 and Vcam1 and (D) Epor in the two populations.

Figure 4—figure supplement 1.. Comparison of gene expression in F4/80+ EKLF/GFP+ and F4/80+ EKLF/GFP- fetal liver macrophages.

(A) Correlation analysis of Z-score transformed gene expression data for each replicate in RNA-Seq showing high correlation between EKLF/GFP+ and EKLF/GFP- replicates, respectively (source data: Figure 4—figure supplement 1—source data 1). (B) Volcano plot showing genes enriched in EKLF/GFP+ and EKLF/GFP- (source data: Figure 4—source data 2).

Figure 5.. EKLF specifies lineage and cell-cycle transcription factors in F4/80+ fetal liver (FL) island macrophages.

(A) Scatterplot of log2-fold changes in EKLF/GFP+ plotted against EKLF-/-. Red box shows the genes that are common and of interest from both datasets, that is, enriched in EKLF/GFP+ and downregulated in EKLF-/- F4/80+ FL macrophages (source data: Figure 5—source data 1). (B) Venn diagram showing the number of genes in each category from (A). Centrimo analysis of promoters of EKLF-dependent genes showing differential motif enrichment of (C) EKLF/Klf1 and (D) Klf3, Sp4, E2f1, and E2f4 motifs (source data: Figure 5—source data 2, 3). Dotted line depicts the expected probability of occurrence of the respective motif in the background dataset (see 'Materials and methods'). (E) Heatmap showing log2-fold change of expression in EKLF-/- and EKLF/GFP+ of the above EKLF-dependent transcription factors in F4/80+ FL macrophages.

Figure 5—figure supplement 1.. EKLF-dependent genes expressed in F4/80+ fetal liver (FL) macrophages.

(A) Heatmap showing log2-fold change expression in EKLF-/- over WT and EKLF/GFP+ over EKLF/GFP- of all the 504 EKLF-dependent genes in F4/80+ FL macrophages (source data: Figure 5—figure supplement 1—source data 1). One in every six genes’ name is displayed for visibility. (B) Centrimo analysis of E2f2 promoter of EKLF-dependent genes showing differential motif enrichment of E2f2 motifs (source data: Figure 5—source data 2, 3). Dotted line depicts the expected probability of occurrence of the E2f2 motif in the background dataset (see 'Materials and methods'). (C) Heatmap showing log2-fold change expression of potential EKLF-dependent transcription factors (source data: Figure 5—figure supplement 1—source data 1). Red boxes highlight the lineage Klf transcription factors and the cell-cycle E2f transcription factors.

Figure 5—figure supplement 2.. EKLF-dependent signature genes in F4/80+ fetal liver (FL) macrophages.

(A) Heatmap of log2 FPKM values of F4/80+ FL signature genes (Figure 1E) that are also enriched in F4/80+ EKLF/GFP+ cells (source data: Figure 5—figure supplement 2—source data 1). (B) Heatmap of log2 FPKM and fold changes of FL signature genes that are significantly downregulated in EKLF-/- (source data: Figure 5—figure supplement 2—source data 2). Red boxes depict the EKLF-dependent signature genes that are common to both (A) and (B). (C) Density plot of flow cytometry analysis of E13.5 FLs stained with F4/80 and Adra2b antibodies. Gating scheme for F4/80-hi and Adra2b+ cells is shown in blue. The percentage of double-positive Adra2b+ and F4/80-hi cells, compared to total Adra2b+ cells, is indicated. (D) Representative pictures of erythroblastic islands from E13.5 FLs are shown after immunostaining with anti-F4/80 (red) or anti-Adra2b (green) antibodies. DNA stain was with DAPI (blue).

Figure 6.. Resolving the cellular heterogeneity of E13.5 fetal liver (FL) macrophages using single-cell RNA-Seq.

(A) Unsupervised clustering using principal component analysis and subsequent U-MAP projections computed and plotted using the R Seurat package for single-cell RNA-Seq of purified E13.5 FL F4/80+ cells. Cluster numbers are indicated on the clusters. (B) Violin plot showing the distribution of F4/80 (Adgre1) mRNA expression in the clusters identified in (A). (C) Feature plots (left panel) showing individual cellular expression superimposed on the cluster, and Violin plots (right) showing the distribution of expression in each cluster of macrophage markers Vcam1 and Marco, and the macrophage-specific transcription factor PU.1 (Spic). (D) Differential mRNA enrichment in each cluster plotted as a heatmap, showing putative unique markers of each cluster (source data: Figure 6—source data 1). Relative expression levels are indicated by color: yellow=high, black=mid, and purple=low.

Figure 6—figure supplement 1.. F4/80 purity check.

F4/80-PE cells isolated from E13.5 fetal livers using the EZSep (Cell Signaling Technologies) magnetic bead method in the presence of Icam4/αv inhibitor peptide (Xue et al., 2014) and analyzed by flow cytometry to determine purity of the F4/80+ population for single-cell sequencing.

Figure 6—figure supplement 2.. Markers for each gene expression-based cluster of cells identified from single-cell sequencing of F4/80+ fetal liver macrophages.

Cluster number and marker names are indicated.

Figure 6—figure supplement 3.. Markers of F4/80+ cell clusters with various cell identities.

Violin plot showing the distribution (left) and feature plot showing individual cell expression (right) of (A) Csf1r, Dnase2a, and Il4rα genes associated with activated macrophages; (B) Gypa, Snca, and Spta1 genes, which are markers for erythro-myeloid characteristics; (C) Gapdh; and (D) Tfrc (CD71), which are uniformly expressed in most clusters.

Figure 7.. EKLF/Klf1-expressing clusters in F4/80+ fetal liver macrophages.

Violin plots showing distribution (left) and feature plots (right) showing individual cellular mRNA expression of (A) Klf1, (B) Epor, and (C) Adra2b superimposed on the clusters.

Figure 8.. Identification of novel markers for F4/80+/EKLF+ fetal liver macrophages from single-cell sequencing.

Using differential enrichment analysis of EKLF clusters 4, 5, and 7 compared with the rest of the cells, putative markers for F4/80+ EKLF+ cells were identified. (A) Violin and feature plots for the identified markers Add2 (adducinβ), Hemgn (hemogen), Nxpe2 (neurexophilin and PC-esterase domain family, member2), and Sptb (spectrinβ). (B) Data (as in Figure 2C, Mass et al., 2016) showing RNA-Seq reads of F4/80+ EKLF+ cell markers from staged and sorted fetal or postnatal liver macrophages. (C) FPKM expression levels of EKLF markers in F4/80+ EKLF/GFP+ and F4/80+ EKLF/GFP- fetal liver macrophage. (D) FPKM expression levels of EKLF markers Add2 and Hemgn in F4/80+ EKLF+/+ and F4/80+ EKLF-/- fetal liver macrophage.

Figure 9.. An improved strategy for antibody-based isolation of F4/80+/EKLF+ cells using novel markers identified from single-cell sequencing.

(A) Flow cytometry analysis of E13.5 fetal liver cells stained with anti-F4/80-PE and anti-adducinβ (top) or anti-spectrinβ (below) antibodies conjugated to AlexaFluor 647. Gates are drawn based on unstained and single-color compensation controls for PE and AlexaFluor 647. Population percentages within each gate are indicated. (B) F4/80+ cells purified from E13.5 fetal livers using magnetic bead selection stained for anti-adducinβ (top) or anti-spectrinβ (below). Gates are the same as (A) and population percentages are indicated. (C) Imaging flow cytometry analysis of E13.5 fetal liver cells from the pEKLF/GFP mouse stained for F4/80-PE and Add2-TxRed. Single cells positive for F4/80, Add2, and GFP are shown. (D, E) Isolated erythroblast islands stained for DAPI, F4/80-PE, and (D) Sptb-Alexa647 or (E) Add2-Alexa647 and examined by fluorescent microscopy. Scale bars are indicated.