Kỹ Thuật - Công Nghệ - Khoa học xã hội - Cơ khí - Vật liệu Article Single-cell RNA sequencing-guided fate-mapping toolkit delineates the contribution of yolk sac erythro-myeloid progenitors Graphical abstract Highlights d scRNA-seq profiles of early yolk sac identify primitive and definitive subsets of EMPs d Csf1r pEMPs generate Csf1r+ pEMPs d Only Csf1r pEMPs contribute to ECs transiently during early embryogenesis d pEMPs and dEMPs give rise to different tissue-resident macrophages Authors Y.X. Zhao, J.Y. Song, X.W. Bao, ..., X.L. Bai, T.B. Liang, J.P. Sheng Correspondence shaowei20022005nuaa.edu.cn (W.S.), shirleybaizju.edu.cn (X.L.B.), liangtingbozju.edu.cn (T.B.L.), shengjpzju.edu.cn (J.P.S.) In brief Zhao et al. found that Csf1r pEMPs can differentiate into Csf1r + pEMPs, and only Csf1r pEMPs are responsible for transiently contributing to endothelial cells during early embryogenesis, suggesting that pEMPs and dEMPs give rise to distinct populations of tissue- resident macrophages. Zhao et al., 2023, Cell Reports 42 , 113364 November 28, 2023 ª 2023 The Authors. https:doi.org10.1016j.celrep.2023.113364 ll Article Single-cell RNA sequencing-guided fate-mapping toolkit delineates the contribution of yolk sac erythro-myeloid progenitors Y.X. Zhao, 1,2,3,8 J.Y. Song, 1,2,3,8 X.W. Bao,4,8 J.L. Zhang,1,2,3,8 J.C. Wu, 1,2,3,8 L.Y. Wang, 5 C. He,6 W. Shao,7, X.L. Bai,1,2,3, T.B. Liang,1,2,3, and J.P. Sheng1,2,3,9, 1 Department of Hepatobiliary and Pancreatic Surgery, The First Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou 310002, China 2 Zhejiang Provincial Key Laboratory of Pancreatic Disease, The First Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou 310002, China 3 Zhejiang University Cancer Center, Zhejiang University, Hangzhou 310002, China 4 Department of Medical Oncology, The First Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou 310002, China 5 Eye Center, The Second Affiliated Hospital, School of Medicine, Zhejiang University, Hangzhou 310003, Zhejiang, China 6 Infinity Scope Biotechnology Co., Ltd., Hangzhou 311200, China 7 College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics, Nanjing 210000, China 8 These authors contributed equally 9 Lead contact Correspondence: shaowei20022005nuaa.edu.cn (W.S.), shirleybaizju.edu.cn (X.L.B.), liangtingbozju.edu.cn (T.B.L.), shengjpzju.edu. cn (J.P.S.) https:doi.org10.1016j.celrep.2023.113364 SUMMARY Erythro-myeloid progenitors of the yolk sac that originates during early embryo development has been sug- gested to generate tissue-resident macrophage, mast cell, and even endothelial cell populations from fetal to adult stages. However, the heterogeneity of erythro-myeloid progenitors (EMPs) is not well characterized. Here, we adapt single-cell RNA sequencing to dissect the heterogeneity of EMPs and establish several fate-mapping tools for each EMP subset to trace the contributions of different EMP subsets. We identify two primitive and one definitive EMP subsets from the yolk sac. In addition, we find that primitive EMPs are decoupled from definitive EMPs. Furthermore, we confirm that primitive and definitive EMPs give rise to microglia and other tissue-resident macrophages, respectively. In contrast, only Kit + Csf1r primitive EMPs generate endothelial cells transiently during early embryo development. Overall, our results delineate the contribution of yolk sac EMPs more clearly based on the single-cell RNA sequencing (scRNA-seq)-guided fate-mapping toolkit. INTRODUCTION Erythro-myeloid progenitors emerge in the yolk sac of the mouse embryo at embryonic day 7.25 (E7.25) as the first detectable he- matopoietic progenitors. 1 The predominant hematopoiesis output of this E7.25 erythro-myeloid progenitor (EMP) is a large-size nucleated red blood cell expressing embryonic globin genes that is very distinct compared to the adult form of small- size enucleated red blood cells (RBCs). Thus, E7.25 EMPs were also called primitive EMPs. 1,2 Primitive EMPs could also give rise to macrophage and megakaryocyte lineages. 1,3,4 Shortly after the onset of primitive EMPs, the definitive EMP emerges in the yolk sac of the mouse conceptus at E8.25. 1 Definitive EMPs produce definitive erythroid and myeloid cell types such as neutrophils, mast cells, and macrophages. 5 Both primitive and definitive EMPs are major sources of hemato- poiesis in the early conceptus to cover the developing require- ments before forming a permanent blood system. It is widely accepted that EMPs generate tissue-resident mac- rophages in the early embryonic development stages. However, the detailed contribution of EMP subsets is debated. Ginhoux et al. suggested that brain-resident macrophages, microglia, originated from primitive EMPs and that other tissue-resident macrophages mainly originated from Myb + definitive EMP- derived fetal monocytes. 2,4,6 On the other hand, Schulz et al. suggested that a single wave of Myb-independent EMPs contributed to all tissue-resident macrophages, including microglia. 7 The above findings were mainly based on the Crelox fate- mapping system. In such a system, a defined cell population at a selected time was labeled by irreversible activation of the expression of a Cre-responsive reporter transgene driven by a carefully chosen promoter specific to a progenitor. For example, EMPs were often defined as CD45low Kit + Csf1r + , and the gen- eration of EMPs depends on the Runx1 gene. Thus, Ginhoux et al. mainly utilized Runx1 MerCreMer to label EMPs, while Schulz Cell Reports 42, 113364, November 28, 2023 ª 2023 The Authors. 1 This is an open access article under the CC BY-NC-ND license (http:creativecommons.orglicensesby-nc-nd4.0). ll OPEN ACCESS et al. mainly utilized Csf1rMerCreMer to label EMPs. After inducible or constitutive activation of Cre recombinase, marked cells are detected later to determine how the originally labeled progeni- tors contribute to specific structures and cell types during pre- and postnatal development. 8 Similarly, Plein et al. utilized Csf1r Cre , Csf1r MerCreMer , and Kit CreERT2 lineage-tracing tools to label EMPs and analyze the contribution of EMPs to endothelial cells (ECs). They observed that ECs were tagged by fluorescent reporters in many organs from early embryonic development to the adult stage after tamoxifen induction in the Csf1r Cre , Csf1r MerCreMer , and Kit CreERT2 fate-mapping system, including the heart, lung, and, especially, liver. 9 Thus, they claimed that the hematopoietic sys- tem’s EMPs served as a complementary source of embryonic vascular endothelium, which lasted into the adult stage. 9 How- ever, in an independent study by Feng et al., the authors utilized the same Csf1r MerCreMer lineage-tracing tool and redeveloped Csf1r Cre and CD45 Cre lineage-tracing systems based on different genetic designs. No EC was tagged in any lineage-tracing sys- tems throughout embryonic development. Based on their obser- vation, Feng et al. suggested that EMPs were not the origin of in- traembryonic ECs, and it was unlikely that the contribution of EMPs to ECs could be detected in the adult stage. 10 The controversial findings regarding the contribution of EMPs were likely due to two reasons. First, the heterogeneity within the EMP population was not clearly defined. Second, the single pro- moter chosen to drive Cre recombinase expression might not be specific enough. Controversial findings and misinterpretations of the fate-mapping results are often due to the fate-mapping model utilized. To increase the specificity of a fate-mapping model, a DreRox recombination system was introduced in the fate-mapping studies. 11 Two different promoters, respectively, drove Cre and Dre recombinases, and combinatory usage of Cre and Dre recombinases increases the labeling specificity of progenitors to be studied. 12 To solve this discrepancy and determine the contributions from EMPs, we first performed single-cell RNA sequencing (scRNA-seq) for E7.5 and E8.5 YSs (yolk sacs) to dissect EMP heterogeneity comprehensively. Based on the EMP subsets identified, we developed three fate-mapping systems for each subset using Cre and Dre systems. With the scRNA-seq-guided fate-mapping system design, we could delineate the contribu- tions of EMPs more accurately. RESULTS Single-cell atlas of early mouse YS The heterogeneity of EMPs may cause a discrepancy in the EC ontogeny study. However, the heterogeneity of EMPs was not addressed clearly in the previous study. 9,10 Thus, we first per- formed scRNA-seq for E7.5 and E8.5 YSs for comprehensive characterization of EMPs’ heterogeneity (Figures 1A and S1) with scRNA-seq technology. 13 38,099 cells were sequenced with high quality, and 14 cell populations were found (Figures 1B and 1C). The cell count, frequency, and top 3 feature genes are shown (Figures 1C and 1D). The cell population’s iden- tity was annotated by SingleR with manual assistance based on their feature gene expression since the reference for early mouse YS cells was inadequate (Figures 1B–1D; Table S1). Cluster 0 (15,507 cells, 40.70) was defined as erythroid cells due to the high expression level of hemoglobin (hemoglobin beta adult t chain Hbb-bt) (Figure 1D). Cluster 1 (9,321 cells, 24.47) carried epithelial marker Ttr 14 and Epcam and was defined as Ttr + epithelial cells. 15 Cluster 2 (4,296 cells, 11.28) was defined as ECs due to the expression of endothelial markers such as Vwf, Sparc, and Col4a2 16–18 (Figure 1D). Cluster 3 (3,416 cells, 8.97) was defined as definitive EMPs due to high expression levels of Kit and Myb and a low level of CD45 (Kit + CD93 CD45low Myb + ), consistent with the previous description of definitive EMPs (Figures 1D and 1E). 2 In addition, Pf4, the marker found in early hematopoietic progenitors, was also identified in part of the cluster 3 definitive EMP cells 19–22 (Figures 1D and 1E). Cluster 4 (1,650 cells, 4.33) expressed high levels of Kit and CD93 and low levels of CD45 and Myb (Figures 1D and 1E). Thus, cellular cluster 4 was defined as prim- itive EMPs (Kit + CD93 + CD45low Myb). 7 Please note that although both primitive EMPs and definitive EMPs were CD45low and Kit + , primitive EMPs were CD93 + Myb , while definitive EMPs were CD93 Myb + . In addition, cluster 3 defini- tive EMPs emerged at E8.5, and cluster 4 primitive EMPs ap- peared at E7.5 (Figure 1F). The emergence timing also supported that cluster 3 was the definitive EMP and cluster 4 was the prim- itive EMP. Cluster 5 (530 cells, 1.39) cells were defined as hemogenic ECs since they expressed Mdk, Hmga2, and Meis2, all of which were detected in hemogenic ECs. 23–25 Cluster 6 (675 cells, 1.77) cells were defined as fibroblasts since a high amount of collagen was detected in cluster 6 (Figure 1D). Cluster 7 (1,136 cells, 2.98) cells were defined as megakaryocyte pro- genitors due to their specific expression of Kit, Rap1b, and Tmsb4x. 26,27 CD41 is a widely recognized marker for EMPs and megakaryocytes. 28,29 As expected, our findings demon- strate that CD41 is expressed in definitive EMPs, primitive EMPs, and megakaryocyte progenitors (cluster 3, 4, and 7) in Figures 1B and 1E. Cluster 8 (668 cells, 1.75) cells were defined as epithelial stem cells since they expressed epithelial marker Krt18 30 and Bex2, an important transcription factor for stemness maintenance 31,32 (Figure 1D). Cluster 9 (457 cells, 1.20) ex- hibited significantly elevated expression of macrophage markers, including Lyz2, Cd74, and Cd68, 33 as shown in Figures 1D and 1E, confirming its categorization as a macro- phage population. Additionally, the presence of S100A8, S100A9, and major histocompatibility complex (MHC) class II expression further supported the identification of cluster 9 as maternal macrophages, as these markers were not expressed in fetal macrophages. These findings raise the possibility of maternal cell contamination. 34 Cluster 10 cells (453 cells, 1.19) were cytotoxic cells, and they expressed cytotoxic mol- ecules granzymes D and G (Figure 1D). Cluster 11 (211 cells, 0.55) cells were identified as smooth muscle cells due to Tac2 and Pdgfrb expression 35 (Figure 1D). Cluster 12 (182 cells, 0.48) was defined as tissue-resident macrophages since they expressed high levels of complement C1q B chain and Cd68 (Figure 1D).36 To elucidate the differences between the two 2 Cell Reports 42, 113364, November 28, 2023 Article ll OPEN ACCESS Figure 1. scRNA-seq profiling of YSs (A) Experimental scheme of scRNA-seq profiling of yolk sacs (YSs). YSs at E7.5 and E8.5 were collected, and scRNA-seq was performed for comprehensive characterization of EMP heterogeneity. (B) 14 major populations are shown by t-distributed stochastic neighbor embedding (tSNE). Cell populations 0–13 were annotated based on their feature gene expression. (C) Cell count and proportion of 14 major populations. (D) Heatmap showing the scaled expression level of top 3 feature genes for each cell population. The dot size indicates the percentage of cells that express the feature gene, and the color indicates the average expression level. (E) tSNE plots showing the expression of Kit, CD93, Ptprc (CD45), Myb, Pf4, Itga2b, Hbb-bh1, Cx3cr1, and Mrc1 at E7.5 (top) and E8.5 (bottom). (F) Split tSNE plots showing the cell distribution of the 14 major populations at E7.5 and E8.5. Cell Reports 42, 113364, November 28, 2023 3 Article ll OPEN ACCESS (legend on next page) 4 Cell Reports 42, 113364, November 28, 2023 Article ll OPEN ACCESS macrophage clusters in Figure 1, we compared their gene expression profiles. Our analysis revealed that these clusters exhibit distinct gene expression patterns, which may imply diver- gent functional roles or developmental stages for these macro- phages. The corresponding top 20 feature genes for each subset are provided in Table S1. Cluster 13 (127 cells, 0.33) was identified as a minor contamination of placenta cells, as evidenced by the expression of Ctla2a and Cryab37,38 (Figure 1D). Given the developmental stage at which our study was conducted, it is currently imprac- tical to distinguish between YSs and other embryonic tissues that eventually contribute to placenta formation, such as allan- tois and chorion. Overall, we profiled YS cells at E7.5 and E8.5 and discovered both primitive and definitive EMPs, which mainly differed in CD93 and Myb expressions. Dissection of the heterogeneity of primitive and definitive EMPs EMPs were our focus, and we wanted to dissect the EMP hetero- geneity further. In order to comprehensively analyze the primitive and definitive EMP populations, we conducted re-clustering and included all of the different definitive EMP (dEMP) clusters iden- tified in Figure 1B in our downstream analysis. Four subsets within primitive EMPs could be found (Figure 2AI). Feature genes for each subset were defined, and the top 3 feature genes were shown (Figure 2AII). These four subsets could be generally divided into two groups. Subsets 0, 2, and 3 formed a Kit + Csf1r group, while subset 1 formed a Kit + Csf1r + group (Fig- ure 2B, feature plot). In addition, we noticed that 2 out of the top 3 feature genes of Csf1r + CD11b + F480 + EMPs (subset 1) were macrophage markers, including Lyve1 39–41 and Mrc1 (CD206)42 (Figure 2AII), implying that Csf1r + EMPs were progen- itors for primitive macrophages. In addition, we have highlighted Cx3cr1 and Mrc1 in the t-distributed stochastic neighbor embedding (tSNE) plot presented in Figure 1E, which demon- strates that Cx3cr1 is predominantly enriched in tissue-resident macrophages (cluster 12), while Mrc1 is mainly enriched in prim- itive EMPs. Furthermore, we noticed that subset 0 of primitive EMPs ex- pressed a high level of hemoglobin chains, including the embry- onic form of the hemoglobin beta chain, Hbb-bh1 (Figures 1E and 2AII). We also noticed that macrophage-oriented primitive EMPs (subset 1) quickly shrunk from E7.5 to E8.5 (Figure 2C), indicating a transient wave of primitive macrophages. Then, we performed a pseudo-time analysis of primitive EMPs, and subset 0 primitive EMPs (pEMPs) were in the earliest node of differentiation (Fig- ure 2D). Subset 1 EMPs were in the latest node of differentiation. Furthermore, we performed RNA velocity 43 analysis for pEMPs, which showed clearly that subset 0 gave rise to other subsets (Figure 2E). Both pseudo-time and RNA velocity results sug- gested that Csf1r pEMPs (Hbb-bH1 + ) were the progenitors for Csf1r + pEMPs, which were further differentiated by group skew- ing to the macrophage lineage compared with Csf1r pEMPs, consistent with the previous report that Hbb-bh1 was devoid from ECs and started to be expressed at E7 in YS EMPs. 44 Similarly, four dEMP subsets (Figure 2AIII–2AIV) could be iden- tified and broadly separated into three categories. CD31 + Myb CD45 subset 2 (ECs), CD31 Myb + CD45 subset 0 (dEMPs), and CD31 Myb CD45 + differentiated immune cells, including subset 1 (macrophage biased) and subset 3 (neutrophil biased) (Figure 2B). The macrophage-biased subset was found to ex- press C1qb and CD74, both of which are detectable on macro- phages. In contrast, the neutrophil-biased subset expressed the typical neutrophil marker MPO, as shown in Figure 2AIV, and it was clear that dEMP subsets appeared around E8.5 (Figure 2C). Pseudo-time analysis of dEMPs showed that subset 2 EMPs were in the earliest node of differentiation (Figure 2D), consistent with their CD31 + Myb CD45 EC identity. CD31 Myb CD45 + subsets 1 and 3 were in the latest node of differentiation, consis- tent with the Myb CD45 + Csf1r + macrophage phenotype (subset 1). However, RNA velocity analysis did not show a clear trend of development (Figure 2E). Regarding the relationship between the tissue-resident mac- rophages (clusters 12) in Figure 1 and the Kit + Csf1r + pEMPs (Figure 2), we included the tissue-resident macrophages from Figure 1 in the RNA velocity analysis in Figure S2. Considering the potential contamination of maternal cells in the maternal macrophage (cluster 9) in Figure 1, we did not include the maternal macrophage in RNA velocity analysis in order to ensure a more accuracy representation of the developmental trajectory of fetal macrophage. The updated analysis provides further in- sights into the dynamic transcriptional changes and potential developmental trajectories of these macrophage populations. We noticed that resident macrophages were derived from Kit + Csf1r + pEMPs. Figure 2. Primitive and definitive EMP subsets (A) Primitive EMPs were further re-clustered into four subsets, shown in the uniform manifold approximation and projection (UMAP) plot in (AI). Subsets 0, 2, and 3 formed the Kit + Csf1r CD11b F480 group (red circle), and subset 1 formed the Kit + Csf1r + CD11b + F480 + group (blue circle). Heatmap showing the scaled expression levels of the top 3 feature genes for each subset of primitive EMPs is in (AII). Definitive EMPs were re-clustered into five subsets, shown in (AIII). Cluster 2 formed CD93 Kit CD45 Pecam + endothelial cells (ECs) (red circle). Cluster 0 formed CD93 Kit + CD45 Myb + definitive EMPs. Clusters 1 and 3 formed CD93 Kit CD45 + Myb mature immune cells, including Csf1r + monocytes and macrophages (yellow circle). Heatmap showing the scaled expression level of the top 3 feature genes for each subset of definitive EMPs is in (AIV). (B) Expression levels of feature markers of Kit, CD93, Ptprc (CD45), Myb, Pecam1, Csf1r, Adgre1, and Itgam are shown in the UMAP plots, separated by primitive (top) and definitive (bottom) EMP populations. (C) UMAP plots showing the primitive EMP subsets and definitive EMP subsets at E7.5 and E8.5. (D) Pseudo-time analysis of primitive and definitive EMP subsets. Potential trajectory of primitive (top) and definitive (bottom) EMP subsets identified distinct cell fates colored by cluster. The branches show the potential evolutionary direction in the trajectory. (E) RNA velocity analysis was performed to infer developmental lineages and cellular dynamics of primitive (top) and definitive (bottom) EMP subsets. (F) Heatmap showing the transcriptional factor activities inferred by SCENIC analysis in primitive (top) and definitive (bottom) EMP subsets. The color indicates the intensity of regulon activity for each transcription factor (TF) in the primitive EMP (pEMP) and definitive EMP (dEMP) subsets. Cell Reports 42, 113364, November 28, 2023 5 Article ll OPEN ACCESS (legend on next page) 6 Cell Reports 42, 113364, November 28, 2023 Article ll OPEN ACCESS Additionally, we have compared the Csf1r + pEMP and macro- phage-biased dEMP populations and noted that Csf1r + pEMPs were more specialized in morphogenesis, while macrophage- biased dEMPs were specialized in immune response (Figures S3A and S3B). We also provided a better explanation for the relationship between the EC cluster in Figure 2AIII and the main ECs in Figure 1. We have performed differential gene expression analysis, which suggests that the EC cluster in Fig- ure 2AIII represents a distinct subset of ECs that have undergone further differentiation. Our analysis has revealed that the main EC population was more specialized in development and angiogen- esis, while the ECs in Figure 2AIII were more specialized in im- mune functions (Figures S3C–S3E). Moreover, we compared cluster 5 (hemogenic ECs) in Figure 1B with cluster 0 in the dEMPs in Figure 2AIII and observed that hemogenic ECs were more specialized in development and morphogenesis, while cluster 0 dEMPs were more specialized in coagulation pro- cesses compared to hemogenic ECs (Figures S3F and S3G). These findings provide further insights into the heterogeneity and functional diversity of ECs in different developmental stages. Next, we performed SCENIC analysis for both pEMPs and dEMPs (Figure 2F). We noticed distinct transcription require- ments for different EMP subsets. Ets1 and Elk3 seemed impor- tant for the macrophage-oriented pEMP subset 1, and Klf1 seemed involved in developing pEMP subset 0. In terms of dEMPs, Maf seemed to regulate the CD45 + Csf1r + monocytesmacrophages (dEMP subset 1), consistent with the previous report about the critical role played by the Maf family within macrophages, 45,46 and Mef2c was found to be involved in the regulation of CD31 + subset 2, consistent with the previous report that Mef2c was pivotal for the hematopoietic progenitor generation from hemogenic ECs. 47 Overall, we identified two CD93 + Myb pEMP subsets (Csf1r and Csf1r + pEMPs), one CD93 Myb + dEMP subset, and the CD45 + Csf1r + monocytemacrophage population. In addition, trajectory analysis and RNA velocity results suggested that Csf1r pEMPs gave rise to Csf1r + pEMPs. Independence of pEMPs and dEMPs Next, we wanted to determine the relationship between pEMPs and dEMPs. Although progeny of pEMPs and dEMPs were quite distinct, 2,5 the independence between pEMPs and dEMPs has not been validated in a proper fate-mapping model. The previous model utilized was not specific enough. For example, E8.5 tamoxifen treatment in Csf1r MerCreMer tagged Kit + Csf1r + pEMPs and CD45 + Csf1r + monocytesmacrophages, and tamoxifen treatment in Kit MerCreMer and Runx1 MerCreMer mice could tag both pEMPs and dEMPs (Figure 3A). Three models were developed for each EMP subset (Fig- ure 3B). The fetal form of the hemoglobin Hbb-bH1 gene was shown to be explicitly expressed in Kit + Csf1r pEMPs via scRNA-seq (Figure 2AI). The expression specificity of Hbb-bH1 was also validated at the protein level by fluorescence staining (Figure S4A). Hbb-bH1 was not detected in CD31 + ECs (precur- sor for EMP generation) or Csf1r + cells (Kit + Csf1r + pEMPs or monocytesmacrophages) (Figure S4B). An Hbb-bH1 Cre mouse (model I) was constructed, and the Cre transgene was inserted into the 30 UTR of Hbb-bH1 to avoid the disruption of the original gene (Figure S4C). To specifically tag Kit + Csf1r + pEMPs, the Kit MerCreMer mouse was crossed to the Csf1r DreERT2 mouse, fol- lowed by mating with the LSL-RSR-tandemTomato reporter mouse (model II; Figures 3B and S4C). Similarly, Kit MerCreMer mice were crossed to Myb DreERT2 mice to tag Myb + Kit + Csf1r dEMPs (model III; Figures 3B and S4C). We established a flow cytometry gate strategy for different EMP subsets based on scRNA-seq analysis. First, pEMPs and dEMPs were separated by CD93. CD93 + pEMPs consisted of Kit + Csf1r pEMPs and Kit + Csf1r+ pEMPs. CD93 CD45 low Kit + Csf1r cells were defined as dEMPs, and CD93 Kit CD45 + Csf1r + cells were identified as monocytesmacrophages (Figure 3C). In the model I fate-mapping system, Kit + Csf1r and Kit + Csf1r + pEMPs were labeled, indicating that Kit + Csf1r + pEMPs were derived from Kit + Csf1r progenitors, and neither Myb + Kit + Csf1r dEMPs nor CD45 + Csf1r+ monocytesmacrophages were labeled, proving the independence of pEMPs and dEMPs (Figures 3D and 3E). In the model II fate-mapping system, only Kit + Csf1r + pEMPs were tagged, showing that Kit + Csf1r + pEMPs could not generate Kit+ Csf1r pEMPs in a reverse way, and dEMPs were not labeled either, further proving the independence of pEMPs and dEMPs (Figures 3D and 3E). In the model III fate-mapping system, both Myb + Kit + Csf1r dEMPs and CD45 + Csf1r + monocytesmacrophage were labeled, indicating that CD45 + Csf1r + monocytesmacrophages were derived from Myb + Kit + Csf1r dEMPs (Figures 3D and 3E) Figure 3. Independence of pEMPs and dEMPs revealed by fate-mapping toolkits (A) Summary of three previously established fate-mapping models and their associated tagging populations, including Csf1r MerCreMer , Runx1 MerCreMer , and KitMerCreMer . These models have been used to trace the developmental fate of EMPs in previous studies. (B) Mating and experimental strategy of three fate-mapping tools used in this study to trace the developmental fate of EMPs. Model I involves the use of Hbb- bH1Cre ::LSL-YFP, while models II and III utilize Kit MerCreMer ::Csf1r DreERT2 ::LSL-RSR-tandemTomato and Kit MerCreMer ::Myb DreERT2 ::LSL-RSR-tandemTomato, respectively. Tamoxifen was administered to models II and III at E8.5, and YS tissue was collected at E10.5 from all three models for fate-mapping analysis. (C) Gating strategies used to isolate different EMP subsets for scRNA-seq analysis. CD93 + CD45low cells were identified as pEMPs and further separated into Kit+ Csf1r and Kit + Csf1r + subsets based on their expression of these markers. Kit+ CD93 CD45low cells were identified as dEMPs and were separated from CD45+ immune cells. (D) Representative histogram showing the expression levels of fate-mapping reporters in EMPs from the different fate-mapping models used in this study. Model I utilizes Hbb-bH1Cre ::LSL-YFP, while models II and III utilize Kit MerCreMer ::Csf1r DreERT2 ::LSL-RSR-tandemTomato and KitMerCreMer ::Myb DreERT2 LSL-RSR-tan- demTomato (n = 6, 8, and 5), respectively. The histograms show the expression levels of YFP (model I) and tandemTomato (models II and III) in EMPs. (E) Statistical results of fate-mapping reporters in EMPs from the different fate-mapping models used in this study. Model I utilizes Hbb-bH1 Cre ::LSL-YFP, while models II and III utilize KitMerCreMer ::Csf1r DreERT2 ::LSL-RSR-tandemTomato and KitMerCreMer ::Myb DreERT2 LSL-RSR-tandemTomato, respectively. Data are pre- sented as means ± SEM of 6, 8, and 5 embryos in each group from 2 independent experiments. Cell Reports 42, 113364, November 28, 2023 7 Article ll OPEN ACCESS (legend on next page) 8 Cell Reports 42, 113364, November 28, 2023 Article ll OPEN ACCESS since CD45 + Csf1r + monocytesmacrophages did not express Kit or Myb (Figure 2B). In addition, pEMPs were not tagged in the model III fate-mapping system, indicating that the reversal generation from dEMPs to pEMPs was not likely. Overall, we developed three specific fate-mapping systems for each EMP subset based on the scRNA-seq results, and we showed that pEMPs and dEMPs were independent of each other. Kit+ myb+ hematopoietic progenitors contribute to most tissue-resident macrophages The ontogeny of macrophage origin has been hotly debated 48 ; thus, we decided to trace the contribution of early hematopoietic progenitors through the newly developed fate-mapping tools. We have included the gating strategy for each tissue-resident macrophage in Figure S5. Microglia were efficiently labeled in the Hbb-bH1 Cre constant fate-mapping system and the Kit MerCreMer ::Csf1r DreERT2 inducible fate-mapping system, which corresponded to the Kit + Csf1r pEMP and Kit + Csf1r + pEMP subsets, indicating microglia were derived pEMPs (Figures 4A and 4B). In contrast, other resident macrophages in the colon, kidney, lung, spleen, and peritoneum were mainly tagged in Kit MerCreMer ::Myb DreERT2 (Figures 4A and 4B). In addition, epithe- lium Langerhans cells were tagged in Hbb-bH1 Cre , Kit MerCreMer :Csf1r DreERT2 , and Kit MerCreMer ::Myb DreERT2 fate-map- ping systems, indicating that both pEMPs and dEMPs contrib- uted to Langerhans cells (LCs) (Figures 4A and 4B). We have conducted additional experiments to investigate the contribution of pEMPs and dEMPs to tissue-resident macro- phages during fetal development. Our results demonstrate that, in both models I and II, the contribution from pEMPs to liver tissue-resident macrophages decreases over time, while the contribution from dEMPs increases during fetal development. Interestingly, in model III, we observed a different trend, with the contribution from dEMPs increasing throughout fetal devel- opment (Figure S6). Our results were in good accordance with Ginhoux et al. 2,4 that primitive macrophages gave rise to micro- glia and that Myb + dEMPs contributed to most tissue-resident macrophages. Only Kit + Csf1r pEMPs contributed to ECs After the dissection of heterogeneity within the EMP population, we wanted to analyze the contribution of the Kit + Csf1r and Kit + Csf1r + EMP subsets to ECs. Since EMPs are known to be precursors of microglia, 49,50 microglia-labeling efficiency was used as an internal reference. Like the previous tamoxifen-induc- tion schedule in Figure 3B, we treated Kit MerCreMer ::Csfr1 DreERT2 and Kit MerCreMer ::MybDreERT2 fate-mapping mice with tamox- ifen at E8.5, and the Hbb-bh1 Cre mice allowed consecutive label- ing of Csf1r+ pEMPs. Brain and YS from the E10.5 embryos were harvested and analyzed. Brain microglia were gated as CD45 int CD11b + F480 + , and ECs were gated as CD45 CD11b CD31 + (Figure 5A), and we noticed that brain ECs and YS ECs were negative for Kit expres- sion (Figures 5B and S4B). After E8.5 tamoxifen treatment in Kit MerCreMer ::Csf1r DreERT2 fate-mapping mice, about 0 of ECs were tagged by tandemTomato, and 20 of microglia were tagged. In contrast, after E8.5 tamoxifen treatment in Kit MerCreMer Myb DreERT2 fate- mapping mice, about 0 of ECs and microglia were tagged (Figures 5C and 5D). In the Hbb-bh1 Cre fate-mapping mouse model, brain microglia were labeled at very high efficiency at 80 due to constitutive expression of Cre driven by Hbb-bh1, while ECs were labeled at about 20 (Figures 5C and 5D). This result suggested that ECs were partially derived from Csf1r pEMPs, as Csf1r + pEMPs did not generate any ECs (Figures 5C and 5D). Fate-mapping results based on three different models sug- gested that EMPs could contribute to ECs partially and that only the Kit + Csf1r subset of pEMPs could generate ECs. Kit + Csf1r EMPs served as a transient source of vascular endothelium at the early embryonic stage Next, we wanted to see if Csf1r pEMP-derived ECs at the early embryonic stage would persist during fetal development. 9 Since we already showed that ECs could only be tagged in the Hbb- bh1 Cre constitutive labeling mouse, we analyzed the later time points only in this fate-mapping model. ECs from organs, including the brain, liver, heart, and spleen, were analyzed at different time points (Figure 6A). We found that the proportion of YFP + ECs in the brain was about 20 at E10.5 (Figures 5C and 5D), and less than 2 of YFP-tagged ECs were detected in brain and fetal liver at E13.5, with the microglia-labeling efficiency around 80 at E13.5 (Figures 6B and 6C) in Hbb-bh1 Cre mice. Next, we selected another time point during mouse develop- ment, pre-birth (E18.5–E19), to investigate the contribution to ECs by EMPs throughout mouse fetal development. Different or- gans were collected and analyzed, including the brain, heart, liver, and spleen. In the Hbb-bh1Cre fate-mapping mice, we showed that very few labeled ECs were found at the pre-natal stage, while the YFP-labeling efficiency of microglia reached about 80 (Figures 6B and 6C). Therefore, fate-mapping results based on the Hbb-bh1 Cre strain further confirmed that Kit + Csf1r pEMPs contributed ECs to blood vessels only transiently at the early embryonic stage and that vascular ECs derived from EMPs were barely found at the pre-birth stage. Figure 4. Ontogeny of adult tissue-resident macrophages (A) Representative histogram showing the expression levels of fate-mapping reporters in various adult tissue-resident macrophages following the experimental settings shown in Figure 3B. Fate mapping was performed using Hbb-bH1Cre ::LSL-YFP (model I) and KitMerCreMer ::Csf1r DreERT2 ::LSL-RSR-tan- demTomatoKit MerCreMer ::Myb DreERT2 LSL-RSR-tandemTomato (models II and III). The histograms show the expression levels of YFP (model I) and tandemTomato (models II and III) in different tissue-resident macrophages. The gating strategies are shown in Figure S5. (B) Statistical results of fate-mapping reporters in various adult tissue-resident macrophages following the experimental settings shown in Figure 3B. Fate mapping was performed using Hbb-bH1Cre ::LSL-YFP (model I) and Kit MerCreMer ::Csf1r DreERT2 ::LSL-RSR-tandemTomatoKit MerCreMer ::MybDreERT2 LSL-RSR- tandemTomato (models II and III). Data are presented as means ± SEM of 6, 5, and 7 tissues in each group from 2 independent experiments. Cell Reports 42, 113364, November 28, 2023 9 Article ll OPEN ACCESS Figure 5. Partial contribution of Csfr1 pEMPs to ECs in early embryo development (A) Gating strategy used to identify microglia and ECs in the brain. Microglia were identified as CD45 int CD11b + F480 + cells, while ECs were identified as CD45 CD11b CD31 + cells using a gating strategy. (B) Histogram showing the expression levels of Kit in ECs. The histogram includes staining of Kit and isotype control in ECs. (C) Fate mapping of microglia and ECs in the brain following the same tamoxifen treatment plan as shown in Figure 3B. Representative flow cytometry plots are shown for each of the three fate-mapping models (models I–III), depicting the expression of YFP in microglia and ECs. (D) Ratio of YFP+ microglia and YFP + ECs in the fetal brain at E10.5. Data are presented as means ± SEM of at least three embryos in each group and represent the proportion of YFP+ microglia and YFP + ECs in the brain following fate mapping using the three different models (models I–III). n = 7, 10, and 9 tissues in each model from 2 independent experiments. 10 Cell Reports 42, 113364, November 28, 2023 Article ll OPEN ACCESS Figure 6. The Csf1r pEMP subset transiently contributes to vascular endothelium during early embryonic development (A) Collection of YSs and various organs, including the brain, liver, and spleen, from Hbb-bH1 Cre fate-mapping mice (model I) at different developmental time points. Tissues were collected at E13.5, pre-natal stages, and adulthood for fate-mapping analysis. (B) Representative histogram plots showing the expression of YFP in microglia (left square) and ECs (right square) from fate-mapping experiments using Hbb- bH1Cre mice (model I) at different developmental time points. The histograms provide a visual representation of the proportion of YFP + microglia and YFP+ ECs in each tissue at the specified time points. (legend continued on next page) Cell Reports 42, 113364, November 28, 2023 11 Article ll OPEN ACCESS Finally, we wanted to see if EMPs contribute to ECs in the adult stage. The data collected 2 months after birth (adult) basically conformed with the data collected at the pre-birth stage. Micro- glia-labeling efficiency was high in the Hbb-bh1 Cre fate-mapping strains, while YFP-tagged ECs were hardly detected across different organs, including the brain, heart, liver, and spleen (Figures 6B and 6C). Overall, we found that only Kit + Csf1r pEMPs served as a tran- sient origin of early embryonic ECs, which could explain why EMP-derived ECs were hardly found in Csf1r MerCreMer mice, and we further showed that a proportion of ECs were EMP derived, and the proportion declined along with the development of mouse embryos, probably diluted by angioblast-derived ECs. The results are summarized in the schematic plot in Figure 6D. DISCUSSION With the help of the scRNA-seq, we could dissect the heteroge- neity of EMP heterogeneity from the early YS, and we found that EMPs consisted of Csf1r (CD93 + Myb Kit + Csf1r) and Csf1r + (CD93 + Myb Kit + Csf1r + ) pEMP subsets and a dEMP subset (CD93 Myb + Kit + ). Csf1r pEMPs expressed the embryonic form of hemoglobin beta chains like Hbb-bh1, while Csf1r + pEMPs expressed macrophage markers like CD206. Thus, we developed three fate-mapping tools to tag three EMP subsets specifically. Next, we employed three different fate-mapping mice for the lineage tracing of EMPs. Our results showed that the Csf1r EMP subset gave rise to the Csf1r + subset and contributed to ECs. However, EMP-derived ECs only existed in the vascular endothelium at the early embryonic stage, and ECs derived from EMPs were hardly detected at the pre-birth and adult stages. Therefore, our results suggested that only the Csf1r EMP subset served as a transient origin of intraem- bryonic ECs. We also traced the ontogeny origin of tissue-resi- dent macrophages, and we found that pEMPs generated only microglia and a minor portion of LCs. In contrast, dEMPs contrib- uted to most tissue-resident macrophages based on fate mapping. Two major hypotheses existed for the ontogeny origin of tis- sue-resident macrophages. Gomez et al. believed that only one wave of EMPs generated all tissue-resident macrophages. 49 Ginhoux et al. proposed that pEMPs generated only microglia and a minor portion of LCs. Meanwhile, dEMPs contributed to most tissue-resident macrophages. 6 However, the previous model utilized was not specific enough. For example, E8.5 tamoxifen treatment in Csf1r MerCreMer tagged Kit + Csf1r + pEMPs and CD45 + Csf1r + monocytesmacrophages, leading to the conclusion that one wave of EMPs generated tissue-resident macrophages and that tamoxifen treatment in Kit MerCreMer and Runx1 MerCreMer mice could tag both pEMPs, dEMPs, and fetal hematopoietic stem cells (HSCs). In addition, the design of Runx1 MerCreMer causes haploinsufficiency of the Runx1 gene, leading to temporal and spatial interruption of hematopoie- sis. 51,52 Development of Hbb-bH1 Cre , Kit MerCreMer ::Csf1r DreERT2 , and Kit MerCreMer ::Myb CreERT2 fate-mapping systems allowed specific tagging of three subsets of EMPs, and fate-mapping re- sults clear the controversial findings of EMP contributions. EMPs were suggested to be a complementary source of em- bryonic vascular endothelium by Plein et al.,9 while the study per- formed by Feng et al. came to a contrary conclusion based on a similar fate-mapping system. 10 To solve this discrepancy and refine the origins of ECs, we performed scRNA-seq with early YSs and employed three lineage-tracing tools to see whether EMPs contribute to ECs of blood vessels. We found that the Csf1r + pEMP subset was the progeny of the Csf1r pEMP subset and that only Csf1r pEMPs served as a tran...
Trang 1Article Single-cell RNA sequencing-guided fate-mapping toolkit delineates the contribution of yolk sac
erythro-myeloid progenitors Graphical abstract
d scRNA-seq profiles of early yolk sac identify primitive and definitive subsets of EMPs
d Csf1r pEMPs generate Csf1r+pEMPs
d Only Csf1r pEMPs contribute to ECs transiently during early embryogenesis
d pEMPs and dEMPs give rise to different tissue-resident macrophages
Y.X Zhao, J.Y Song, X.W Bao, , X.L Bai, T.B Liang, J.P Sheng
Zhao et al found that Csf1r pEMPs can differentiate into Csf1r+pEMPs, and only Csf1r pEMPs are responsible for transiently contributing to endothelial cells during early embryogenesis, suggesting that pEMPs and dEMPs give rise to distinct populations of tissue-resident macrophages.
Zhao et al., 2023, Cell Reports42, 113364 November 28, 2023ª 2023 The Authors.
Trang 2Single-cell RNA sequencing-guided
fate-mapping toolkit delineates the contribution of yolk sac erythro-myeloid progenitors
Y.X Zhao,1,2,3,8J.Y Song,1,2,3,8X.W Bao,4,8J.L Zhang,1,2,3,8J.C Wu,1,2,3,8L.Y Wang,5C He,6W Shao,7,*X.L Bai,1,2,3,*
T.B Liang,1,2,3,*and J.P Sheng1,2,3,9,*
1Department of Hepatobiliary and Pancreatic Surgery, The First Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou 310002, China
2Zhejiang Provincial Key Laboratory of Pancreatic Disease, The First Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou 310002, China
3Zhejiang University Cancer Center, Zhejiang University, Hangzhou 310002, China
4Department of Medical Oncology, The First Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou 310002, China
5Eye Center, The Second Affiliated Hospital, School of Medicine, Zhejiang University, Hangzhou 310003, Zhejiang, China
6Infinity Scope Biotechnology Co., Ltd., Hangzhou 311200, China
7College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics, Nanjing 210000, China
8These authors contributed equally
Erythro-myeloid progenitors of the yolk sac that originates during early embryo development has been sug-gested to generate tissue-resident macrophage, mast cell, and even endothelial cell populations from fetal to adult stages However, the heterogeneity of erythro-myeloid progenitors (EMPs) is not well characterized Here, we adapt single-cell RNA sequencing to dissect the heterogeneity of EMPs and establish several fate-mapping tools for each EMP subset to trace the contributions of different EMP subsets We identify two primitive and one definitive EMP subsets from the yolk sac In addition, we find that primitive EMPs are decoupled from definitive EMPs Furthermore, we confirm that primitive and definitive EMPs give rise to microglia and other tissue-resident macrophages, respectively In contrast, only Kit+ Csf1rprimitive EMPs generate endothelial cells transiently during early embryo development Overall, our results delineate the contribution of yolk sac EMPs more clearly based on the single-cell RNA sequencing (scRNA-seq)-guided fate-mapping toolkit.
Erythro-myeloid progenitors emerge in the yolk sac of the mouse embryo at embryonic day 7.25 (E7.25) as the first detectable he-matopoietic progenitors.1 The predominant hematopoiesis output of this E7.25 erythro-myeloid progenitor (EMP) is a large-size nucleated red blood cell expressing embryonic globin genes that is very distinct compared to the adult form of small-size enucleated red blood cells (RBCs) Thus, E7.25 EMPs were also called primitive EMPs.1,2Primitive EMPs could also give rise to macrophage and megakaryocyte lineages.1,3,4
Shortly after the onset of primitive EMPs, the definitive EMP emerges in the yolk sac of the mouse conceptus at E8.25.1 Definitive EMPs produce definitive erythroid and myeloid cell types such as neutrophils, mast cells, and macrophages.5 Both primitive and definitive EMPs are major sources of hemato-poiesis in the early conceptus to cover the developing require-ments before forming a permanent blood system.
It is widely accepted that EMPs generate tissue-resident mac-rophages in the early embryonic development stages However, the detailed contribution of EMP subsets is debated Ginhoux et al suggested that brain-resident macrophages, microglia, originated from primitive EMPs and that other tissue-resident macrophages mainly originated from Myb+ definitive EMP-derived fetal monocytes.2,4,6 On the other hand, Schulz et al suggested that a single wave of Myb-independent EMPs contributed to all tissue-resident macrophages, including microglia.7
The above findings were mainly based on the Cre/lox fate-mapping system In such a system, a defined cell population at a selected time was labeled by irreversible activation of the expression of a Cre-responsive reporter transgene driven by a carefully chosen promoter specific to a progenitor For example, EMPs were often defined as CD45/lowKit+Csf1r+, and the gen-eration of EMPs depends on the Runx1 gene Thus, Ginhoux et al mainly utilized Runx1MerCreMerto label EMPs, while Schulz
Trang 3et al mainly utilized Csf1rMerCreMerto label EMPs After inducible or constitutive activation of Cre recombinase, marked cells are detected later to determine how the originally labeled progeni-tors contribute to specific structures and cell types during pre-and postnatal development.8
Similarly, Plein et al utilized Csf1rCre, Csf1rMerCreMer, and KitCreERT2lineage-tracing tools to label EMPs and analyze the contribution of EMPs to endothelial cells (ECs) They observed that ECs were tagged by fluorescent reporters in many organs from early embryonic development to the adult stage after tamoxifen induction in the Csf1rCre, Csf1rMerCreMer, and KitCreERT2fate-mapping system, including the heart, lung, and, especially, liver.9Thus, they claimed that the hematopoietic sys-tem’s EMPs served as a complementary source of embryonic vascular endothelium, which lasted into the adult stage.9 How-ever, in an independent study by Feng et al., the authors utilized the same Csf1rMerCreMer lineage-tracing tool and redeveloped Csf1rCreand CD45Crelineage-tracing systems based on different genetic designs No EC was tagged in any lineage-tracing sys-tems throughout embryonic development Based on their obser-vation, Feng et al suggested that EMPs were not the origin of in-traembryonic ECs, and it was unlikely that the contribution of EMPs to ECs could be detected in the adult stage.10
The controversial findings regarding the contribution of EMPs were likely due to two reasons First, the heterogeneity within the EMP population was not clearly defined Second, the single pro-moter chosen to drive Cre recombinase expression might not be specific enough Controversial findings and misinterpretations of the fate-mapping results are often due to the fate-mapping model utilized To increase the specificity of a fate-mapping model, a Dre/Rox recombination system was introduced in the fate-mapping studies.11Two different promoters, respectively, drove Cre and Dre recombinases, and combinatory usage of Cre and Dre recombinases increases the labeling specificity of progenitors to be studied.12
To solve this discrepancy and determine the contributions from EMPs, we first performed single-cell RNA sequencing (scRNA-seq) for E7.5 and E8.5 YSs (yolk sacs) to dissect EMP heterogeneity comprehensively Based on the EMP subsets identified, we developed three fate-mapping systems for each subset using Cre and Dre systems With the scRNA-seq-guided fate-mapping system design, we could delineate the contribu-tions of EMPs more accurately.
Single-cell atlas of early mouse YS
The heterogeneity of EMPs may cause a discrepancy in the EC ontogeny study However, the heterogeneity of EMPs was not addressed clearly in the previous study.9,10Thus, we first per-formed scRNA-seq for E7.5 and E8.5 YSs for comprehensive characterization of EMPs’ heterogeneity (Figures 1A and S1) with scRNA-seq technology.13 38,099 cells were sequenced with high quality, and 14 cell populations were found (Figures 1B and 1C) The cell count, frequency, and top 3 feature genes are shown (Figures 1C and 1D) The cell population’s iden-tity was annotated by SingleR with manual assistance based on
their feature gene expression since the reference for early mouse YS cells was inadequate (Figures 1B–1D;Table S1).
Cluster 0 (15,507 cells, 40.70%) was defined as erythroid cells due to the high expression level of hemoglobin (hemoglobin beta adult t chain [Hbb-bt]) (Figure 1D) Cluster 1 (9,321 cells, 24.47%) carried epithelial marker Ttr14and Epcam and was defined as Ttr+epithelial cells.15Cluster 2 (4,296 cells, 11.28%) was defined as ECs due to the expression of endothelial markers such as Vwf, Sparc, and Col4a216–18(Figure 1D).
Cluster 3 (3,416 cells, 8.97%) was defined as definitive EMPs due to high expression levels of Kit and Myb and a low level of CD45 (Kit+CD93CD45/lowMyb+), consistent with the previous description of definitive EMPs (Figures 1D and 1E).2In addition, Pf4, the marker found in early hematopoietic progenitors, was also identified in part of the cluster 3 definitive EMP cells19–22 (Figures 1D and 1E) Cluster 4 (1,650 cells, 4.33%) expressed high levels of Kit and CD93 and low levels of CD45 and Myb (Figures 1D and 1E) Thus, cellular cluster 4 was defined as prim-itive EMPs (Kit+ CD93+ CD45/low Myb).7 Please note that although both primitive EMPs and definitive EMPs were CD45/low and Kit+, primitive EMPs were CD93+Myb, while definitive EMPs were CD93Myb+ In addition, cluster 3 defini-tive EMPs emerged at E8.5, and cluster 4 primidefini-tive EMPs ap-peared at E7.5 (Figure 1F) The emergence timing also supported that cluster 3 was the definitive EMP and cluster 4 was the prim-itive EMP.
Cluster 5 (530 cells, 1.39%) cells were defined as hemogenic ECs since they expressed Mdk, Hmga2, and Meis2, all of which were detected in hemogenic ECs.23–25 Cluster 6 (675 cells, 1.77%) cells were defined as fibroblasts since a high amount of collagen was detected in cluster 6 (Figure 1D) Cluster 7 (1,136 cells, 2.98%) cells were defined as megakaryocyte pro-genitors due to their specific expression of Kit, Rap1b, and Tmsb4x.26,27 CD41 is a widely recognized marker for EMPs and megakaryocytes.28,29 As expected, our findings demon-strate that CD41 is expressed in definitive EMPs, primitive EMPs, and megakaryocyte progenitors (cluster 3, 4, and 7) in
Figures 1B and 1E.
Cluster 8 (668 cells, 1.75%) cells were defined as epithelial stem cells since they expressed epithelial marker Krt1830and Bex2, an important transcription factor for stemness maintenance31,32(Figure 1D) Cluster 9 (457 cells, 1.20%) ex-hibited significantly elevated expression of macrophage markers, including Lyz2, Cd74, and Cd68,33 as shown in
Figures 1D and 1E, confirming its categorization as a macro-phage population Additionally, the presence of S100A8, S100A9, and major histocompatibility complex (MHC) class II expression further supported the identification of cluster 9 as maternal macrophages, as these markers were not expressed in fetal macrophages These findings raise the possibility of maternal cell contamination.34 Cluster 10 cells (453 cells, 1.19%) were cytotoxic cells, and they expressed cytotoxic mol-ecules granzymes D and G (Figure 1D) Cluster 11 (211 cells, 0.55%) cells were identified as smooth muscle cells due to Tac2 and Pdgfrb expression35(Figure 1D) Cluster 12 (182 cells, 0.48%) was defined as tissue-resident macrophages since they expressed high levels of complement C1q B chain and Cd68 (Figure 1D).36 To elucidate the differences between the two
2 Cell Reports 42, 113364, November 28, 2023
OPEN ACCESS
Trang 4Figure 1 scRNA-seq profiling of YSs
(A) Experimental scheme of scRNA-seq profiling of yolk sacs (YSs) YSs at E7.5 and E8.5 were collected, and scRNA-seq was performed for comprehensivecharacterization of EMP heterogeneity.
(B) 14 major populations are shown by t-distributed stochastic neighbor embedding (tSNE) Cell populations 0–13 were annotated based on their feature geneexpression.
(C) Cell count and proportion of 14 major populations.
(D) Heatmap showing the scaled expression level of top 3 feature genes for each cell population The dot size indicates the percentage of cells that express thefeature gene, and the color indicates the average expression level.
(E) tSNE plots showing the expression of Kit, CD93, Ptprc (CD45), Myb, Pf4, Itga2b, Hbb-bh1, Cx3cr1, and Mrc1 at E7.5 (top) and E8.5 (bottom).(F) Split tSNE plots showing the cell distribution of the 14 major populations at E7.5 and E8.5.
Trang 5(legend on next page)
4 Cell Reports 42, 113364, November 28, 2023
OPEN ACCESS
Trang 6macrophage clusters in Figure 1, we compared their gene expression profiles Our analysis revealed that these clusters exhibit distinct gene expression patterns, which may imply diver-gent functional roles or developmental stages for these macro-phages The corresponding top 20 feature genes for each subset are provided inTable S1.
Cluster 13 (127 cells, 0.33%) was identified as a minor contamination of placenta cells, as evidenced by the expression of Ctla2a and Cryab37,38(Figure 1D) Given the developmental stage at which our study was conducted, it is currently imprac-tical to distinguish between YSs and other embryonic tissues that eventually contribute to placenta formation, such as allan-tois and chorion.
Overall, we profiled YS cells at E7.5 and E8.5 and discovered both primitive and definitive EMPs, which mainly differed in CD93 and Myb expressions.
Dissection of the heterogeneity of primitive and definitive EMPs
EMPs were our focus, and we wanted to dissect the EMP hetero-geneity further In order to comprehensively analyze the primitive and definitive EMP populations, we conducted re-clustering and included all of the different definitive EMP (dEMP) clusters iden-tified in Figure 1B in our downstream analysis Four subsets within primitive EMPs could be found (Figure 2AI) Feature genes for each subset were defined, and the top 3 feature genes were shown (Figure 2AII) These four subsets could be generally divided into two groups Subsets 0, 2, and 3 formed a Kit+ Csf1rgroup, while subset 1 formed a Kit+Csf1r+group ( Fig-ure 2B, feature plot) In addition, we noticed that 2 out of the top 3 feature genes of Csf1r+CD11b+F4/80+EMPs (subset 1) were macrophage markers, including Lyve139–41 and Mrc1 (CD206)42(Figure 2AII), implying that Csf1r+EMPs were progen-itors for primitive macrophages In addition, we have highlighted Cx3cr1 and Mrc1 in the t-distributed stochastic neighbor embedding (tSNE) plot presented inFigure 1E, which demon-strates that Cx3cr1 is predominantly enriched in tissue-resident macrophages (cluster 12), while Mrc1 is mainly enriched in prim-itive EMPs.
Furthermore, we noticed that subset 0 of primitive EMPs ex-pressed a high level of hemoglobin chains, including the embry-onic form of the hemoglobin beta chain, Hbb-bh1 (Figures 1E and
2AII) We also noticed that macrophage-oriented primitive EMPs
(subset 1) quickly shrunk from E7.5 to E8.5 (Figure 2C), indicating a transient wave of primitive macrophages Then, we performed a pseudo-time analysis of primitive EMPs, and subset 0 primitive EMPs (pEMPs) were in the earliest node of differentiation ( Fig-ure 2D) Subset 1 EMPs were in the latest node of differentiation Furthermore, we performed RNA velocity43analysis for pEMPs, which showed clearly that subset 0 gave rise to other subsets (Figure 2E) Both pseudo-time and RNA velocity results sug-gested that Csf1rpEMPs (Hbb-bH1+) were the progenitors for Csf1r+pEMPs, which were further differentiated by group skew-ing to the macrophage lineage compared with Csf1rpEMPs, consistent with the previous report that Hbb-bh1 was devoid from ECs and started to be expressed at E7 in YS EMPs.44
Similarly, four dEMP subsets (Figure 2AIII–2AIV) could be iden-tified and broadly separated into three categories CD31+Myb CD45subset 2 (ECs), CD31Myb+CD45subset 0 (dEMPs), and CD31MybCD45+differentiated immune cells, including subset 1 (macrophage biased) and subset 3 (neutrophil biased) (Figure 2B) The macrophage-biased subset was found to ex-press C1qb and CD74, both of which are detectable on macro-phages In contrast, the neutrophil-biased subset expressed the typical neutrophil marker MPO, as shown inFigure 2AIV, and it was clear that dEMP subsets appeared around E8.5 (Figure 2C) Pseudo-time analysis of dEMPs showed that subset 2 EMPs were in the earliest node of differentiation (Figure 2D), consistent with their CD31+MybCD45EC identity CD31MybCD45+ subsets 1 and 3 were in the latest node of differentiation, consis-tent with the Myb CD45+ Csf1r+ macrophage phenotype (subset 1) However, RNA velocity analysis did not show a clear trend of development (Figure 2E).
Regarding the relationship between the tissue-resident mac-rophages (clusters 12) in Figure 1and the Kit+Csf1r+pEMPs (Figure 2), we included the tissue-resident macrophages from
Figure 1in the RNA velocity analysis inFigure S2 Considering the potential contamination of maternal cells in the maternal macrophage (cluster 9) in Figure 1, we did not include the maternal macrophage in RNA velocity analysis in order to ensure a more accuracy representation of the developmental trajectory of fetal macrophage The updated analysis provides further in-sights into the dynamic transcriptional changes and potential developmental trajectories of these macrophage populations We noticed that resident macrophages were derived from Kit+ Csf1r+pEMPs.
Figure 2 Primitive and definitive EMP subsets
(A) Primitive EMPs were further re-clustered into four subsets, shown in the uniform manifold approximation and projection (UMAP) plot in (AI) Subsets 0, 2, and 3formed the Kit+
Csf1rCD11bF4/80group (red circle), and subset 1 formed the Kit+
group (blue circle) Heatmap showing the scaledexpression levels of the top 3 feature genes for each subset of primitive EMPs is in (AII) Definitive EMPs were re-clustered into five subsets, shown in (AIII) Cluster2 formed CD93KitCD45Pecam+
endothelial cells (ECs) (red circle) Cluster 0 formed CD93Kit+
definitive EMPs Clusters 1 and 3 formedCD93KitCD45+
Mybmature immune cells, including Csf1r+
monocytes and macrophages (yellow circle) Heatmap showing the scaled expression level ofthe top 3 feature genes for each subset of definitive EMPs is in (AIV).
(B) Expression levels of feature markers of Kit, CD93, Ptprc (CD45), Myb, Pecam1, Csf1r, Adgre1, and Itgam are shown in the UMAP plots, separated by primitive(top) and definitive (bottom) EMP populations.
(C) UMAP plots showing the primitive EMP subsets and definitive EMP subsets at E7.5 and E8.5.
(D) Pseudo-time analysis of primitive and definitive EMP subsets Potential trajectory of primitive (top) and definitive (bottom) EMP subsets identified distinct cellfates colored by cluster The branches show the potential evolutionary direction in the trajectory.
(E) RNA velocity analysis was performed to infer developmental lineages and cellular dynamics of primitive (top) and definitive (bottom) EMP subsets.(F) Heatmap showing the transcriptional factor activities inferred by SCENIC analysis in primitive (top) and definitive (bottom) EMP subsets The color indicates theintensity of regulon activity for each transcription factor (TF) in the primitive EMP (pEMP) and definitive EMP (dEMP) subsets.
Trang 7(legend on next page)
6 Cell Reports 42, 113364, November 28, 2023
OPEN ACCESS
Trang 8Additionally, we have compared the Csf1r+pEMP and macro-phage-biased dEMP populations and noted that Csf1r+pEMPs were more specialized in morphogenesis, while macrophage-biased dEMPs were specialized in immune response (Figures S3A and S3B) We also provided a better explanation for the relationship between the EC cluster inFigure 2AIII and the main ECs inFigure 1 We have performed differential gene expression analysis, which suggests that the EC cluster in Fig-ure 2AIII represents a distinct subset of ECs that have undergone further differentiation Our analysis has revealed that the main EC population was more specialized in development and angiogen-esis, while the ECs inFigure 2AIII were more specialized in im-mune functions (Figures S3C–S3E) Moreover, we compared cluster 5 (hemogenic ECs) inFigure 1B with cluster 0 in the dEMPs inFigure 2AIII and observed that hemogenic ECs were more specialized in development and morphogenesis, while cluster 0 dEMPs were more specialized in coagulation pro-cesses compared to hemogenic ECs (Figures S3F and S3G) These findings provide further insights into the heterogeneity and functional diversity of ECs in different developmental stages Next, we performed SCENIC analysis for both pEMPs and dEMPs (Figure 2F) We noticed distinct transcription require-ments for different EMP subsets Ets1 and Elk3 seemed impor-tant for the macrophage-oriented pEMP subset 1, and Klf1 seemed involved in developing pEMP subset 0.
In terms of dEMPs, Maf seemed to regulate the CD45+Csf1r+ monocytes/macrophages (dEMP subset 1), consistent with the previous report about the critical role played by the Maf family within macrophages,45,46and Mef2c was found to be involved in the regulation of CD31+subset 2, consistent with the previous report that Mef2c was pivotal for the hematopoietic progenitor generation from hemogenic ECs.47
Overall, we identified two CD93+MybpEMP subsets (Csf1r and Csf1r+pEMPs), one CD93Myb+dEMP subset, and the CD45+Csf1r+monocyte/macrophage population In addition, trajectory analysis and RNA velocity results suggested that Csf1rpEMPs gave rise to Csf1r+pEMPs.
Independence of pEMPs and dEMPs
Next, we wanted to determine the relationship between pEMPs and dEMPs Although progeny of pEMPs and dEMPs were quite distinct,2,5the independence between pEMPs and dEMPs has not been validated in a proper fate-mapping model The previous
model utilized was not specific enough For example, E8.5 tamoxifen treatment in Csf1rMerCreMertagged Kit+Csf1r+pEMPs and CD45+ Csf1r+ monocytes/macrophages, and tamoxifen treatment in KitMerCreMer and Runx1MerCreMer mice could tag both pEMPs and dEMPs (Figure 3A).
Three models were developed for each EMP subset ( Fig-ure 3B) The fetal form of the hemoglobin Hbb-bH1 gene was shown to be explicitly expressed in Kit+ Csf1r pEMPs via scRNA-seq (Figure 2AI) The expression specificity of Hbb-bH1 was also validated at the protein level by fluorescence staining (Figure S4A) Hbb-bH1 was not detected in CD31+ECs (precur-sor for EMP generation) or Csf1r+cells (Kit+Csf1r+pEMPs or monocytes/macrophages) (Figure S4B) An Hbb-bH1Cremouse (model I) was constructed, and the Cre transgene was inserted into the 30UTR of Hbb-bH1 to avoid the disruption of the original gene (Figure S4C) To specifically tag Kit+Csf1r+pEMPs, the KitMerCreMermouse was crossed to the Csf1rDreERT2mouse, fol-lowed by mating with the LSL-RSR-tandemTomato reporter mouse (model II; Figures 3B and S4C) Similarly, KitMerCreMer mice were crossed to MybDreERT2mice to tag Myb+Kit+Csf1r dEMPs (model III;Figures 3B andS4C).
We established a flow cytometry gate strategy for different EMP subsets based on scRNA-seq analysis First, pEMPs and dEMPs were separated by CD93 CD93+pEMPs consisted of Kit+Csf1r pEMPs and Kit+Csf1r+pEMPs CD93CD45/low Kit+ Csf1r cells were defined as dEMPs, and CD93 Kit CD45+Csf1r+cells were identified as monocytes/macrophages (Figure 3C).
In the model I fate-mapping system, Kit+ Csf1r and Kit+ Csf1r+pEMPs were labeled, indicating that Kit+Csf1r+pEMPs were derived from Kit+Csf1r progenitors, and neither Myb+ Kit+Csf1rdEMPs nor CD45+Csf1r+monocytes/macrophages were labeled, proving the independence of pEMPs and dEMPs (Figures 3D and 3E).
In the model II fate-mapping system, only Kit+Csf1r+pEMPs were tagged, showing that Kit+ Csf1r+ pEMPs could not generate Kit+ Csf1r pEMPs in a reverse way, and dEMPs were not labeled either, further proving the independence of pEMPs and dEMPs (Figures 3D and 3E).
In the model III fate-mapping system, both Myb+Kit+Csf1r dEMPs and CD45+ Csf1r+ monocytes/macrophage were labeled, indicating that CD45+Csf1r+monocytes/macrophages were derived from Myb+Kit+Csf1rdEMPs (Figures 3D and 3E)
Figure 3 Independence of pEMPs and dEMPs revealed by fate-mapping toolkits
(A) Summary of three previously established fate-mapping models and their associated tagging populations, including Csf1rMerCreMer, Runx1MerCreMer, andKitMerCreMer
These models have been used to trace the developmental fate of EMPs in previous studies.
(B) Mating and experimental strategy of three fate-mapping tools used in this study to trace the developmental fate of EMPs Model I involves the use of respectively Tamoxifen was administered to models II and III at E8.5, and YS tissue was collected at E10.5 from all three models for fate-mapping analysis.(C) Gating strategies used to isolate different EMP subsets for scRNA-seq analysis CD93+CD45/lowcells were identified as pEMPs and further separated intoKit+
Csf1rand Kit+
subsets based on their expression of these markers Kit+
CD93CD45/lowcells were identified as dEMPs and were separated fromLSL-RSR-tan-demTomato (n = 6, 8, and 5), respectively The histograms show the expression levels of YFP (model I) and tanLSL-RSR-tan-demTomato (models II and III) in EMPs.(E) Statistical results of fate-mapping reporters in EMPs from the different fate-mapping models used in this study Model I utilizes Hbb-bH1Cre
LSL-RSR-tandemTomato, respectively Data are pre-sented as means± SEM of 6, 8, and 5 embryos in each group from 2 independent experiments.
Trang 9(legend on next page)
8 Cell Reports 42, 113364, November 28, 2023
OPEN ACCESS
Trang 10since CD45+Csf1r+monocytes/macrophages did not express Kit or Myb (Figure 2B) In addition, pEMPs were not tagged in the model III fate-mapping system, indicating that the reversal generation from dEMPs to pEMPs was not likely.
Overall, we developed three specific fate-mapping systems for each EMP subset based on the scRNA-seq results, and we showed that pEMPs and dEMPs were independent of each other.
Kit+myb+hematopoietic progenitors contribute to most tissue-resident macrophages
The ontogeny of macrophage origin has been hotly debated48; thus, we decided to trace the contribution of early hematopoietic progenitors through the newly developed fate-mapping tools We have included the gating strategy for each tissue-resident macrophage inFigure S5 Microglia were efficiently labeled in the Hbb-bH1Cre constant fate-mapping system and the KitMerCreMer::Csf1rDreERT2inducible fate-mapping system, which corresponded to the Kit+Csf1rpEMP and Kit+Csf1r+pEMP subsets, indicating microglia were derived pEMPs (Figures 4A and 4B) In contrast, other resident macrophages in the colon, kidney, lung, spleen, and peritoneum were mainly tagged in KitMerCreMer::MybDreERT2(Figures 4A and 4B) In addition, epithe-lium Langerhans cells were tagged in Hbb-bH1Cre, KitMerCreMer:Csf1rDreERT2, and KitMerCreMer::MybDreERT2 fate-map-ping systems, indicating that both pEMPs and dEMPs contrib-uted to Langerhans cells (LCs) (Figures 4A and 4B).
We have conducted additional experiments to investigate the contribution of pEMPs and dEMPs to tissue-resident macro-phages during fetal development Our results demonstrate that, in both models I and II, the contribution from pEMPs to liver tissue-resident macrophages decreases over time, while the contribution from dEMPs increases during fetal development Interestingly, in model III, we observed a different trend, with the contribution from dEMPs increasing throughout fetal devel-opment (Figure S6) Our results were in good accordance with Ginhoux et al.2,4that primitive macrophages gave rise to micro-glia and that Myb+dEMPs contributed to most tissue-resident macrophages.
Only Kit+Csf1rpEMPs contributed to ECs
After the dissection of heterogeneity within the EMP population, we wanted to analyze the contribution of the Kit+Csf1r and Kit+Csf1r+EMP subsets to ECs Since EMPs are known to be precursors of microglia,49,50 microglia-labeling efficiency was used as an internal reference Like the previous tamoxifen-induc-tion schedule inFigure 3B, we treated KitMerCreMer::Csfr1DreERT2 and KitMerCreMer::MybDreERT2 fate-mapping mice with tamox-ifen at E8.5, and the Hbb-bh1Cremice allowed consecutive
label-ing of Csf1r+/pEMPs Brain and YS from the E10.5 embryos were harvested and analyzed.
Brain microglia were gated as CD45intCD11b+F4/80+, and ECs were gated as CD45CD11bCD31+(Figure 5A), and we noticed that brain ECs and YS ECs were negative for Kit expres-sion (Figures 5B andS4B).
After E8.5 tamoxifen treatment in KitMerCreMer::Csf1rDreERT2 fate-mapping mice, about 0% of ECs were tagged by tandemTomato, and 20% of microglia were tagged In contrast, after E8.5 tamoxifen treatment in KitMerCreMer MybDreERT2 fate-mapping mice, about 0% of ECs and microglia were tagged (Figures 5C and 5D).
In the Hbb-bh1Crefate-mapping mouse model, brain microglia were labeled at very high efficiency at 80% due to constitutive expression of Cre driven by Hbb-bh1, while ECs were labeled at about 20% (Figures 5C and 5D) This result suggested that ECs were partially derived from Csf1rpEMPs, as Csf1r+pEMPs did not generate any ECs (Figures 5C and 5D).
Fate-mapping results based on three different models sug-gested that EMPs could contribute to ECs partially and that only the Kit+Csf1rsubset of pEMPs could generate ECs Kit+Csf1rEMPs served as a transient source of vascular endothelium at the early embryonic stage Next, we wanted to see if Csf1rpEMP-derived ECs at the early embryonic stage would persist during fetal development.9Since we already showed that ECs could only be tagged in the Hbb-bh1Creconstitutive labeling mouse, we analyzed the later time points only in this fate-mapping model ECs from organs, including the brain, liver, heart, and spleen, were analyzed at different time points (Figure 6A).
We found that the proportion of YFP+ECs in the brain was about 20% at E10.5 (Figures 5C and 5D), and less than 2% of YFP-tagged ECs were detected in brain and fetal liver at E13.5, with the microglia-labeling efficiency around 80% at E13.5 (Figures 6B and 6C) in Hbb-bh1Cremice.
Next, we selected another time point during mouse develop-ment, pre-birth (E18.5–E19), to investigate the contribution to ECs by EMPs throughout mouse fetal development Different or-gans were collected and analyzed, including the brain, heart, liver, and spleen In the Hbb-bh1Cre fate-mapping mice, we showed that very few labeled ECs were found at the pre-natal stage, while the YFP-labeling efficiency of microglia reached about 80% (Figures 6B and 6C).
Therefore, fate-mapping results based on the Hbb-bh1Cre strain further confirmed that Kit+Csf1r pEMPs contributed ECs to blood vessels only transiently at the early embryonic stage and that vascular ECs derived from EMPs were barely found at the pre-birth stage.
Figure 4 Ontogeny of adult tissue-resident macrophages
(A) Representative histogram showing the expression levels of fate-mapping reporters in various adult tissue-resident macrophages following theexperimental settings shown inFigure 3B Fate mapping was performed using Hbb-bH1Cre
::LSL-YFP (model I) and KitMerCreMer
::LSL-RSR-tan-demTomato/KitMerCreMer
LSL-RSR-tandemTomato (models II and III) The histograms show the expression levels of YFP (model I) andtandemTomato (models II and III) in different tissue-resident macrophages The gating strategies are shown inFigure S5.
(B) Statistical results of fate-mapping reporters in various adult tissue-resident macrophages following the experimental settings shown inFigure 3B Fatemapping was performed using Hbb-bH1Cre
::LSL-YFP (model I) and KitMerCreMer