Yolk sac, but not hematopoietic stem cell–derived progenitors, sustain erythropoiesis throughout murine embryonic life

Yolk sac erythropoiesis is not replaced by hematopoietic stem cell–derived progeny up until birth. Yolk sac–derived erythroid progenitors require 10-fold lower concentrations of erythropoietin for differentiation, which confers a competitive advantage over progenitors of hematopoietic stem cell origin.


Introduction
Erythrocytes are the most abundant cells in circulation. They transport oxygen and have a half-life of around 22 d in mice and 120 d in humans. Therefore, constant production in the bone marrow (BM) is required to maintain the numbers of circulating RBCs.
Erythropoiesis is the process whereby hematopoietic stem cells (HSCs) progressively differentiate into megakaryocyte/ erythrocyte progenitors (MEPs) and later into lineage-committed erythroid progenitors, immature burst-forming unit-erythroid cells (BFU-Es), and the more mature CFU-erythroid cells (CFU-Es). CFU-Es successively progress in differentiation through nucleated proerythroblast, basophilic, polychromatophilic, and orthochromatic stages, enucleation, and formation of RBCs. The distinct stages of erythroid differentiation are characterized by changes in surface expression of the progenitor marker Kit, of the transferrin receptor CD71, of the adhesion molecule CD44, and of the mature erythroid marker Ter119 (Kina et al., 2000;Aisen, 2004;Chen et al., 2009).
In c-Myb mutants, where primitive hematopoiesis is preserved but HSC-derived hematopoiesis is missing, YS-derived primitive erythrocytes suffice to maintain living embryos up until E15.5 (Mucenski et al., 1991;Tober et al., 2008;Schulz et al., 2012). In the absence of HSC activity, EMP-derived hematopoietic cells maintain viable embryos throughout development up until birth (Chen et al., 2011). YS hematopoiesis has long been considered a transient wave devoted to the production of erythrocytes, megakaryocytes, and a few myeloid cells that ensure oxygenation and tissue hemostasis. HSC-derived hematopoiesis was thought to replace YSderived cells after HSCs migrate to the FL at E10.5 (Palis, 2016). Recently, growing evidence endows the YS with the capacity to contribute to tissue-resident cells such as macrophages that persist throughout life (Gomez Perdiguero et al., 2015) and mast cells (Gentek et al., 2018) maintained up until birth. Primitive erythrocytes were also shown to persist throughout gestation (Fraser et al., 2007), and EMP-derived cells contribute to the erythrocyte compartment for more than 20 d upon transplantation (McGrath et al., 2015a). Nonetheless, it has been difficult to establish the temporal relative contribution of EMP-or HSCderived progenitors to erythropoiesis because they share surface markers and transcriptional regulators and are therefore indistinguishable.
Here we report a large population of Kit + CD45 − Ter119 − erythroid progenitors unique to FL, comprising >70% of E14.5 Ter119 − CD45 − cells (>10% of FL cells). These are the most actively proliferating progenitors at early stages and rapidly progress in erythroid differentiation. These cells, which require c-Myb expression, originate from YS EMPs as they are colabeled with microglia in the Cdh5 CreERT2 Rosa26 YFP and labeled in the Csf1r MeriCreMer Rosa26 YFP lineage-tracer models induced at E7.5 and at E8.5, respectively. They persist through fetal life and are the major contributors to the RBC compartment.
We show that HSCs do not contribute significantly to embryonic erythropoiesis by tracing Flt3 + progenitors or YFP-expressing cells in Cdh5 CreERT2 Rosa26 YFP induced in E10.5 embryos. HSC-derived erythroid progenitors require >10-fold higher concentrations of erythropoietin (Epo) than their YS-derived counterparts for erythrocyte differentiation. The limiting amounts of Epo available in the embryo (Suzuki et al., 2011) result in a selective advantage of YS-derived over HSC-derived erythropoiesis.

Results
A unique population of Kit + cells represents the majority of FL Ter119 − CD45 − cells We detected in FL, by flow cytometry, a large fraction of Kit + cells (>50%) expressing neither Ter119 (specific for erythrocytes) nor CD45 (a pan-hematopoietic marker; Fig. 1, A and B;and Fig. S1). Single-cell surface marker expression data from E14.5 FL cells was projected as tSNE1 vs. tSNE2 (t-distributed stochastic neighbor embedding; Fig. 1 A), and three major clusters were defined by the expression of epithelial cadherin (CD324) on epithelial cells, platelet/endothelial cell adhesion protein (PECAM-1/CD31) on endothelial cells, and Kit. Combined analysis of Kit expression together with CD24 further defined three populations in the Ter119 − CD45 − CD31 − CD324 − compartment: Kit + CD24 − (hereafter called P1), Kit + CD24 + (P2), and CD24 + Kit − (P3) cells ( Fig. 1 B and Fig. S1 A). Numbers of P2 cells reached a maximum (around 10 6 cells per FL) at E14-E15 and decreased thereafter, although they were still detected around birth (E18.5; Fig. 1 C). Kit + CD45 − Ter119 − Lin − (P1 and P2) cells were also negative for the expression of Sca-1, which marks multipotent progenitors, for CD16/32, which marks granulocyte/macrophage progenitors (GMPs), and for CD34, marking common myeloid progenitors (CMPs), and therefore they fall in a gate that typically defines MEPs in the FL and in the BM (Fig. S1 B). Unlike their BM counterparts, however, where all Kit + cells coexpressed CD45, most FL Kit + (P1 and P2) cells within the Lin − Kit + Sca1 − (LK) compartment did not express CD45 ( Fig. 1 D), raising the possibility that they did not belong to the hematopoietic lineage. We identified here a major population of Kit + CD45 − Ter119 − cells unique to FL of undefined lineage affiliation and origin.
P1 and P2 cells in the FL have an erythroid progenitor signature RNA sequencing (RNA-seq) of the three major populations P2, CD324 + , and CD31 + cells indicated that the highest expressed transcripts in P2 cells were Myb, Bcl11a, Klf1, Gata1, and Epor, which are associated with erythrocyte differentiation (Fig. 1 E). The 122 genes upregulated more than twofold in P2 vs. CD324 + cells were subjected to gene ontology analysis using Enrichr (Chen et al., 2013; list of submitted genes in Table S1). The top biological processes and tissue-associated genes revealed an erythrocyte/erythroblast profile ( Fig. 1 F). These results were validated by quantitative RT-PCR indicating that Gata1, Lmo2, Klf1, and Epor expressions gradually increased from P1 to P3 cells, with the latter showing comparable expression levels of these transcripts to Ter119 + erythroblasts ( Fig. 1 G). Hemoglobin transcripts for Hbb-y, Hbb-bh1, and Hbb-b1 were detected in P3 cells but only significantly expressed in Ter119 + cells. Multipotent hematopoiesis associated transcription factors such as c-Myb, Runx3, and Bmi1 decreased as the erythroidspecific transcripts increased. The results above indicated that the CD45 − Kit + subsets (P1 and P2) are erythroid progenitors and suggested a hierarchy where P1 cells further differentiate into P2 and later lose Kit expression (P3) before acquiring Ter119 expression, the definitive marker of erythroid identity.
P1, P2, and P3 cells represent increasingly mature stages within the erythroid lineage Erythroid differentiation has been characterized by the expression of CD71 (transferrin receptor) and CD44 in Ter119 + cells (McGrath et al., 2017). Imaging flow cytometry (Fig. 2 A) showed that the pro-erythroblast marker CD71 was low in P1, increased in P2, and was highly expressed in all P3 cells (Fig. 2 B), indicating they correspond to consecutive stages of erythrocyte development.
In BM and FL, CFU-Es are characterized as Kit + CD71 + Ter119 − , whereas low levels of Ter119 expression mark pro-erythroblasts that lost proliferative capacity (McGrath et al., 2017). P2 FL cells express CD71 but not Ter119, indicating that they correspond to CFU-Es. P3 cells express low levels of Ter119 ( Fig. 2 B), visible in 8% of them, and express four out of the five key erythroid genes analyzed at levels similar to Ter119 + erythroblasts ( Fig. 1 G), suggesting they correspond to proerythroblasts. CD71 expression is limited to P2 and P3 cells, indicating that CD71 and CD24 are redundant markers in this context, further confirmed by conventional flow cytometry, as previously described (Fig. S1 C;Flygare et al., 2011).
Administration of three doses of the nucleotide analogue 5ethynyl-29-deoxyuridine (EdU), which labels newly synthesized DNA, in E12.5 or E13.5 pregnant females (Fig. 2 C) indicated that P1 E13.5 FL cells are the highest proliferating cells (80% EdU + ) compared with P2 and P3 cells (∼50% EdU + ). E14.5 FL cells show the same level of EdU incorporation (40-50%) in all three subsets (Fig. 2, D and E). Cell proliferation was further assessed by analyzing the expression of the nuclear protein Ki-67 that in association with DAPI allows distinguishing cells in G0, G1, and G2M phases of the cell cycle. Consistent with the EdU-labeling experiments, P1 showed the lowest frequency of cells in G0 (∼10%) and the highest frequency of cells expressing Ki-67, from which around 20% were actively synthesizing DNA (DAPI + ; Fig. 2 F). By contrast, P3 cells showed the lowest frequency of proliferating cells (Fig. 2 F). Taken together, these results indicated that P1 are the most proliferating cells and, as they transit onto the P2 and further into the P3 subset, lose proliferative activity.
To show a lineage relationship between P1, P2, and P3 cells, we cultured purified P1 and P2 cells. 6 h in culture was sufficient to upregulate CD24 expression in P1 cells, and after 12 h, 40% of cells expressed CD24 (Fig. 3 C). After 24 h, all P1 cells had differentiated to P2 cells, upregulating CD24 and losing c-Kit expression ( Fig. 3 C). Ter119 expression was detected in P1 cells after 48 h in culture, whereas P2 cells already expressed this marker after 18 h of culture, demonstrating that they represent a more differentiated progenitor (Fig. 3, D and F).
To probe the differentiation potential in vivo, we injected P2 cells purified from E13.5 ubiquitin C (UBC)-GFP embryos into E13.5 C57/BL6 recipient embryos in utero (Fig. 3 G). FL and blood collected 3 d later indicated that GFP + P2 originated exclusively GFP + Ter119 + cells whereas LKs generated a majority of myeloid cells (Fig. 3, H and I), while none gave rise to lymphocytes. The low numbers of P1 cells precluded similar in vivo differentiation analysis with these cells.
These results demonstrated that P2 FL cells are committed erythroid progenitors while P1 cells retain residual in vitro myeloid differentiation potential.

P1/P2 progenitors require c-Myb expression
The transcription factor c-Myb is essential for definitive hematopoiesis, and Myb −/− embryos are not viable after E15. Only primitive YS-derived erythropoiesis, primitive megakaryocytes (Tober et al., 2008), and tissue-resident macrophages (Schulz et al., 2012) are found in c-Myb mutants. Ter119 + cells were drastically decreased in frequency and numbers in Myb −/− FLs when compared with heterozygous littermates (Fig. 4, A and B). P1 cells were undetectable in Myb −/− whereas P3 cells were present, albeit in reduced numbers (Fig. 4 B). CD24 + cells in Myb −/− FL expressed high levels of Afp and Alb, indicating their hepatic cell affiliation (Fig. S1 D). c-Myb was reported to regulate c-Kit expression (Ratajczak et al., 1998). To assess whether P1 and P2 cells, although unable to express c-Kit, were present in the Myb −/− FL, we analyzed the expression of erythroid genes in CD24 − , CD24 + , and Ter119 + cells from Myb +/− and Myb −/− FLs. Epor, Tal1, and Klf1 were not detected in CD24 − and CD24 + Myb −/− compared with Myb +/− cells (Fig. 4 C). Only Ter119 + cells Figure 1. A population of Kit + cells unique to FL represents the majority of Ter119 − CD45 − cells and has an erythroid progenitor transcriptional signature. (A) tSNE analysis and hierarchical clustering of flow cytometry data of Ter119 − CD45 − cells from E14.5 FLs stained with the surface markers CD31 (endothelial cells), CD324 (epithelial cells, hepatoblasts), Kit, and CD24. (B) Representative FACS plots of the clusters identified in A using the same color code. (C) TER119 − CD45 − cell numbers from E12.5 and E18.5. (D) Histogram of CD45 expression in LK and LSK cells from E12.5 FL and adult BM (E11.5, n = 6; E12.5, n = 17; E14.5, n = 4; E16.5, n = 8; and E18.5, n = 2). See also Fig. S1. (E) CD31 + , P2, and CD324 + cells from E14.5 FL were sorted and subjected to RNA-seq. Differentially expressed genes are represented as a heatmap, and most expressed genes are listed (n = 3 independent litters). (F) Gene set enrichment analysis on genes with more than twofold difference in expression level between P2 and CD324 + cells using Enrichr web application; top 10 significantly associated Gene Ontology Biological Process and ARCHS4 Tissues are shown. The top biological process is erythrocyte differentiation (q-value <1.0e −6 ; Gene Ontology term GO:0030218), and the top tissue associated is erythroblast (q-value <1.0e −14 ; ARCHS4 Tissues gene database). (G) E14.5 FL P1, P2, P3, and Ter119 + cells were sorted and gene expression of key erythroid genes (Gata1, Lmo2, Klf1, Epor, and Bmi1), progenitor-associated genes (c-Myb and Runx3), and hemoglobins (Hbb-y, Hbb-bh1, and Hbb-b1) was analyzed. qPCR data were analyzed using the ΔCt method and were normalized with β-actin. Statistical significance was assessed using one-way ANOVA followed by Tukey's multiple comparison test. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. Data are represented as mean ± SD from three independent experiments. NK, natural killer. Figure 2. P1, P2, and P3 cells represent increasingly mature stages within the erythroid lineage. (A) E13.5 Ter119 − CD45 − cells were analyzed by imaging flow cytometry using CD71, Ter119, Kit, CD24, and CD45 as surface markers, DRAQ5 to label nuclei, and Thiazole Orange to label RNA. Representative images of expressed detectable levels of Epor and Tal1 together with high levels of Hbb-y, indicating they represent primitive erythrocytes. Of note, Klf1, a transcription activator of the β-globin promoter, was not expressed in primitive Ter119 + Myb −/− cells (Fig. 4 C). These results demonstrated that differentiation and/ or survival of CD45 − Kit + erythroid progenitors required the transcription factor c-Myb. In subsequent days, the frequency of YFP + cells in more differentiated P3 and erythroblasts (Lin + CD71 + ) increased, whereas that of more immature P1 cells decreased. In line with previous reports, YFP + Lin − Kit + Sca1 + cells (LSKs) were undetectable ( Fig. 5 B and Fig. S2 A). The dynamic of YFP-labeled erythroid progenitors is consistent with a progression in erythroid differentiation and indicates a lineage relationship between the three subsets. Moreover, the frequency of YFP-labeled erythroblasts decreases in FL from E12.5 until E14.5 while it increases in blood ( Fig. 5 C). The frequency of YFP-labeled P1 and P2 decreases between E11.5 and E13.5, a dynamic that reflects a fast differentiation progression, culminating with the exit of erythroblasts from FL into circulation.
To test whether YFP + and YFP − cells represented two distinct populations, we performed single-cell multiplex gene expression analysis in FL P1, P2, and P3 cells of Csf1r MeriCreMer Rosa26 YFP embryos pulsed at E8.5. Unsupervised hierarchical clustering did not segregate YFP + from YFP − cells for the expression of progenitor, erythroid, and myeloid transcripts, indicating that they have a similar profile and therefore do not represent two divergent progenitor populations (Fig. 5 D). Clusters I and IV contained P1 cells characterized by the expression of Gata1, Lmo2, and c-Myb. Cluster I differed from cluster IV by a high frequency of cells expressing Epor, Tal1, Klf1, and Ki-67. Interestingly, some cells in this cluster also coexpressed the myeloid factors Runx1, Gata2, Zfpm1, and Mpl, suggesting a broad myeloid transcriptional priming, consistent with data from in vitro differentiation assays (Fig. 4 B). High expression of Ki-67 also indicates that cluster I in contrast to cluster IV cells are proliferating. Few cells segregated from all others in cluster II, defined by expression of Csf3r, Ly6c, and Runx2 in the absence of erythroid-associated transcripts. Cluster III comprises a majority of P2 cells, expressing high levels of erythroid genes and low levels of hemoglobin, thus defining a transitional erythroid population. Cluster V contains P3 cells that express high levels of hemoglobin in the absence of c-Kit or c-Myb.
To analyze the differentiation trajectory between the three populations, we generated a diffusion map and obtained a trajectory in which P1 cells progress through a P2 and subsequently a P3 stage (Fig. 5 E). This differentiation trajectory is in line with the gene expression data ( Fig. 1 G), with the imaging flow cytometry results (Fig. 2 A) and with the clonal differentiation assays (Fig. 3, B-F). YFP + and YFP − cells have similar trajectories, indicating they do not represent different progenitor subsets (Fig. 5 E).
The Csf1r MeriCreMer Rosa26 YFP lineage-tracing model induced at E8.5 labels Csf1r-expressing EMPs and macrophages. To complement the results above and to minimize the possibility of disregarding the contribution of Csf1r − YS progenitors, we used the Cdh5 CreERT2 Rosa26 YFP model that, with a single dose of OH-TAM at E7.5, efficiently labels YS hematopoiesis (Fig. 5 F;and Fig. S3, A and B;Zovein et al., 2008;Gentek et al., 2018). P2, P3 and erythroblasts are labeled at the same frequency as microglia at E12.5 and E14.5 and decrease thereafter (Fig. 5 G), similar to the results obtained with Csf1r MeriCreMer Rosa26 YFP embryos. While YFP + FL erythroblasts decrease from E12.5 to E18.5, blood erythroblasts increase (Fig. 5 H;and Fig. S3,C and D), indicating that they exit the FL and enter circulation. Moreover, the percentage of YFP + blood erythrocytes is similar to that found in microglia ( Fig. 5 I), establishing that circulating erythrocytes of EMP YS origin dominate the compartment up until E18.5.

HSCs do not contribute to erythropoiesis up until birth
To evaluate the HSC contribution to fetal erythropoiesis, we analyzed Flt3 Cre Rosa26 YFP embryos where HSC-derived multipotent progenitors (MPP Flt3 + ) and their progeny are YFP + (Benz et al., 2008;Buza-Vidas et al., 2011). In addition, it labels to a transient population of Flt3 + progenitors recently identified (Fig. 6 A; Beaudin et al., 2016). Less than 2% of microglial cells are YFP + at all time points analyzed. LSKs were increasingly labeled with >30% of YFP + cells at E16.5 and reaching >40% at E18.5 (Fig. S4, A-D). Both CMP and GMP compartments exhibited similar levels of YFP expression to those in LSKs (Fig. 6  B). By contrast, P1 + P2 or P3 cells were minimally labeled with YFP by E18.5. YFP + erythroblasts were virtually undetectable at E16.5 and were <20% of those found in LSKs by E18.5, indicating that HSCs are not contributing to mature erythrocytes up until 1 d before birth. MEPs showed a delayed profile, reaching 50% of LSK labeling only around birth. Purified YFP − and YFP + CMPs P1, P2, and P3 cells. BF, Brightfield. (B) Expression of CD71 and Ter119 was assessed in P1, P2, and P3 cells and plotted as a histogram. (C and D) Experimental design of cell cycle analysis using EdU labeling. E12.5 or E13.5 pregnant mice were injected intraperitoneally (IP) with 100 μg EdU at 0 h, 8 h, and 16 h. FLs were collected 2 h after the last injection, and EdU expression was analyzed on Ter119 − CD45 − CD54 − CD31 − cells using the indicated gates (D). (E) Percentages of EdU incorporation in P1, P2, and P3 cells at E13.5 and E14.5 (n = 3). (F) Cell cycle analysis of E14.5 FL cells using Ki-67 and DAPI. Statistical significance was assessed using one-way ANOVA followed by Tukey's multiple comparison test. *, P < 0.05; **, P < 0.01. Data are represented as mean ± SD from three independent experiments. (G) Schematic representation of transplantation experiment. E13.5 C57/BL6 pregnant females were anesthetized, the peritoneal cavity was opened, and the uterus was exposed. Embryos were injected intraperitoneally with 20,000 E13.5 GFP + purified cells from UBC-GFP embryos. FL and blood were collected 3 d after injection. (H) Representative FACS plots showing erythroid contribution of GFP + cells in FL and blood after injection of P2 or LK cells and quantification (I; P2 n = 3, LK n = 4, four independent experiments). Statistical significance was assessed using one-way ANOVA followed by Tukey's multiple comparison test. ***, P < 0.001; ****, P < 0.0001. Data are represented as mean ± SD. yielded similar frequencies of erythroid and myeloid cells and generated similar numbers of hematopoietic colonies, indicating that YFP expression did not impair hematopoietic differentiation (Fig. S5, C and D).
Most erythrocytes are derived from MEPs that differentiate from Flt3-expressing MPPs. However, recent evidence indicated that a fraction of megakaryocytes might bypass the MPP stage and differentiate directly from a progenitor phenotypically and Hbb-y) and key erythroid genes (Epor, Tal1, and Klf1) in CD24 − , CD24 + , and Ter119 + cells was analyzed by qPCR. Statistical significance was assessed using one-way ANOVA followed by Tukey's multiple comparison test. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. Data are represented as mean ± SD from two independent experiments. indistinguishable from HSCs (Carrelha et al., 2018). These results raised the possibility that erythrocytes may bypass the MPP stage because they originate from a common MEP. Using the Flt3 Cre Rosa26 YFP mouse model, we could therefore underestimate the contribution of HSCs to the erythrocyte compartment.
FL stroma produces Epo, essential for erythrocyte production, albeit at lower concentrations than the adult kidney (Suzuki et al., 2011). Embryonic progenitors have been reported to react to lower concentrations of Epo than their adult counterparts (Rich and Kubanek, 1980). We compared erythrocyte production from P2 of YS origin with that from CD45 + MEPs of HSC origin from the same FL (Fig. 7 A). HSC-derived MEPs were about twofold less efficient in generating erythrocytes than YSderived P2 at limiting levels of Epo and required >10-fold higher concentrations to reach 50% of erythrocyte colony formation. P2 cells exhibited higher sensitivity to Epo, and both P1 and P2 cells express higher levels of Epo receptor (EpoR) than CD45 + CMPs or MEPs of HSC origin from the same FL (Fig. 7, B and C). Gata1, Sp1, and Tal1 transcripts that regulate Epor expression (Zon et al., 1991;Feng and Kan, 2005;Rogers et al., 2012) were increased in P1 and P2 when compared with CD45 + CMPs and CD45 + MEPs, respectively (Fig. 7, B and C). These results provide experimental evidence for the mechanism controlling the selection of YSderived over HSC-derived progenitors in fetal erythropoiesis.
Gene expression analysis and in vitro assays indicated a developmental progression where P1 cells further develop into P2 and later into P3 cells before acquisition of Ter119 expression. P3 cells did not generate colonies in vitro and expressed the erythroid transcripts at levels similar to Ter119 + cells, a stage at which they also express Hbb-b1, low levels of Hbb-bh1, and undetectable Hbb-y transcripts, indicating they are within the definitive erythroid lineage. Of note, P1 cells are among the most actively proliferative FL progenitors, indicating that they can considerably expand before terminal differentiation.
YS-derived primitive erythrocytes and megakaryocytes (Tober et al., 2008) and tissue-resident macrophages are c-Myb independent, and so is EMP generation (Schulz et al., 2012). Accordingly, CD45 + cells present in the FL of c-Myb mutants representing tissue resident macrophages were not affected (Schulz et al., 2012). Ter119 + cells, however, were drastically decreased in the FL of c-Myb mutants, and P1 cells were undetectable (Schulz et al., 2012). Ter119 + cells in the c-Myb −/− FL expressed embryonic globins (εy and βh1), consistent with their primitive origin, and did not express Klf1, a transcription activator of the β-globin promoter essential for the transition from embryonic to adult hemoglobin expression (Perkins et al., 1995;Perkins et al., 2016). c-Myb induces proliferation of erythroid progenitors and c-Kit expression, and therefore HSC-independent erythroid cells are affected by c-Myb mutations (Vegiopoulos et al., 2006). Recently c-Myb was shown to control the expression of Klf1 and Lmo2 required for erythropoiesis, offering an explanation for the differential impact of c-Myb inactivation in the progeny of YS-derived EMPs (Bianchi et al., 2010).
Single injections of OH-TAM at E8.5 in Csf1r MeriCreMer Rosa26 YFP mark exclusively YS-derived cells and their progeny, among which is the microglia (Gomez Perdiguero et al., 2015). In these mice, P1 and P2 cells are marked at levels similar to the microglia 3 d after OH-TAM, indicating they are the progeny of YS EMPs. Consistent with the lineage relationship previously established, the frequency of labeled P1/P2 cells decreased with time after injection, whereas the frequency of labeled erythroblasts in FL and later in blood increased. EMPs emerge in the YS between E8.5 and E10.5, and a single injection of OH-TAM at E8.5 will reach the highest circulating levels of the drug 6 h later, rapidly decreasing thereafter (Zovein et al., 2008). Only a fraction of EMPs or already differentiated myeloid progeny that maintain Csf1r expression will be labeled with YFP. By contrast, differentiation into erythroid progenitors results in the loss of Csf1r expression, thus explaining the decreasing frequency of labeled immature erythroid progenitors with time. P1, P2, and P3 and erythroblasts are labeled at similar levels to those found Figure 7. P1/P2 cells have higher sensitivity to Epo than HSC derived progenitors. (A) Frequency of erythroid colonies in P2 and MEP (CD45 + ) cells from E13.5 FL using serial dilutions of Epo. Representative plot of three independent experiments. Curves represent the linear regression of the data. (B) Quantitative RT-PCR analysis of the expression levels of Epor and its regulators Gata1, Sp1, and Tal1 in P1 and CD45 + CMPs (n = 3). (C) Quantitative RT-PCR analysis of the expression levels of Epor and its regulators Gata1, Sp1, and Tal1 in P2 and CD45 + MEPs (n = 3). In each experiment, cells were sorted from the same embryos. Statistical significance was assessed using paired t tests. *, P < 0.05; **, P < 0.01; ***, P < 0.001. Data are represented as mean ± SD. in microglia up until E14.5, a time point after which labeling of erythroblasts and erythrocytes in blood increases. Similar results were obtained by analyzing Cdh5 CreERT2 Rosa26 YFP embryos, in which a pulse with OH-TAM at E7.5 labels the hemogenic endothelium, which gives rise to YS progenitors but does not generate HSCs (Gentek et al., 2018).
An opposite kinetic is found in Flt3 Cre Rosa26 YFP mice, where Flt3-expressing cells and their progeny are permanently labeled with YFP. By E16.5, where equivalent frequencies of LSKs, CMPs, and GMPs are YFP + , only a small frequency of erythroid progenitors including MEPs and virtually undetectable frequencies of P3 or erythroblasts are labeled. By E18.5, MEPs were labeled at similar levels to those of CMPs, although the erythroblast compartment still shows a modest contribution of Flt3-expressing progenitors. Because lymphoid progenitors persistently express Flt3 after commitment, they are the highest labeled population in this model and were excluded from the analysis. It has been recently described that MEPs can bypass the stage of Flt3-expressing MPPs and would therefore be undetected in this model (Carrelha et al., 2018). Cdh5 CreERT2 Rosa26 YFP embryos pulsed with OH-TAM at E10.5 where HSC-derived hematopoiesis is labeled (Zovein et al., 2008) exhibited a minor fraction of YFP + erythroblasts or erythrocytes at any time point. By contrast, nonerythroid CD45 + Kit − mature populations in FL do not appear to be preferentially originated from one of the sources. We thus observed consistent results in both models and excluded the possibility of underestimating the contribution of HSCs to fetal erythrocytes due to the bypass of Flt3 + progenitors by MEPs.
HSCs in FL expand but also differentiate, giving rise to multilineage progeny that comprise CMPs, GMPs, and lymphoid progenitors. However, our data show that despite a rapid differentiation of FL HSCs, they do not contribute significantly to the erythroid compartment before birth, and therefore in vivo embryonic HSC differentiation is skewed (Fig. 8). The FL stromal microenvironment can sustain erythropoiesis, and FL HSCs can differentiate into erythrocytes in vitro. We show that the low levels of Epo available in FL, before the kidney is competent to produce adult levels of this hormone, modulate the differentiation pattern of HSCs that do not produce MEPs and do not contribute to erythropoiesis (Zanjani et al., 1981). Consistently, low levels of the EpoR regulator Gata1 result in embryonic anemia but normal adult erythropoiesis (McDevitt et al., 1997). Our observation that YS-derived erythrocyte progenitors express higher levels of Gata1, Tal1, and consequently Epor is also consistent with an advantage of embryonic vs. adult erythroid progenitors. Large numbers of expanding YS-derived erythrocyte progenitors efficiently outcompete HSC progeny in an environment where resources for erythroid differentiation are limiting.
These results reinforce the notion that in contrast to what has been accepted, YS hematopoiesis is not only devoted to providing oxygen to the embryo before HSCs differentiate in FL but also to sustaining embryonic survival until birth (Fig. 8). FL was shown to be a less efficient environment for cytokine production than BM. Recently, low levels of interleukin-7 in FL were identified as the selective mechanism that resulted in autoreactive properties of B1 B cells (Wong et al., 2019). It is therefore tempting to speculate that low cytokine production is a general characteristic of FL cells with important implications for fetal hematopoiesis. In humans, kidney-derived Epo is also only produced after birth, offering an explanation for why premature neonates develop severe anemia (Widness, 2008).
In this line, a recent report analyzing human FL hematopoiesis indicates that all cells in the erythrocyte lineage, similar to the observation in the mouse reported here, do not express CD45 at stages ranging from 7-17 wk after conception (Popescu et al., 2019).
Our observations may also help clarify why patients with mutations in Epo that alter the kinetics of receptor binding only show postnatal anemia (Kim et al., 2017). These observations suggest that fetal erythropoiesis originates in the YS also in humans and will impact our understanding of embryonic hematopoiesis in general and in the pathogenesis of infant erythrocyte abnormalities.

RNA-seq and analysis
Total RNA from sorted E14.5 FL cells was extracted using the RNeasy Micro kit (Qiagen) following the manufacturer's instructions, and ribosomal RNA sequences were eliminated by enzymatic treatment (Zap R, Clontech). cDNA libraries were generated using the SMARTer Stranded Total RNA-Seq Kit-Pico Input Mammalian (Clontech). The single-read RNA-seq reads were aligned to the mouse reference genome GRCm38 using STAR. The numbers of reads aligned to genes were counted using FeatureCounts (Liao et al., 2014). The R package DESeq2 (Love et al., 2014) was used to normalize reads and identify differentially expressed genes with statistical significance using the negative binomial test (P < 0.05; Benjamini-Hochberg correction).
Enrichr was used to perform gene set enrichment analysis of the highly differentially expressed genes in P2 vs. CD324 + cells (greater than twofold differential expression; the gene list is available in Table S1; Chen et al., 2013). The top 10 terms from the Gene Ontology Biological Process 2018 and ARCHS4 Tissues were retrieved. Expression datasets are available in GEO under accession no. GSE138960.  Primitive erythroid progenitors (EryP) are first generated in the YS (E7.5) and give rise to primitive nucleated erythrocytes (P-Ery), still found in low frequencies at birth. A second wave of progenitors emerges in the YS (E8.5) as EMPs. EMPs migrate to the FL, where they differentiate into highly proliferative erythroid (CD45 − Kit + ) progenitors that sustain erythropoiesis during embryonic life. HSCs generated in the aorta-gonadsmesonephros (AGM; E9.5-E11.5) migrate to the FL, where they expand and differentiate into myeloid and lymphoid lineages. Contribution of HSCs to the erythroid lineage is only detected after birth. YS-derived progenitors respond to lower levels of Epo than their HSC counterparts and have a selective advantage in FL, where Epo levels are lower than in adult BM. D-Ery, definitive erythrocytes; Ery, erythroid cell; Ly, lymphoid lineage; Mk, megakaryocyte; My, myeloid lineage. β-actin, and relative expression was calculated using the 2 −ΔCt method.

EdU incorporation and cell cycle analysis
EdU detection was done using the Click-iT EdU Pacific blue flow cytometry assay kit (Invitrogen; Cat# C10418). The cell cycle was analyzed after fixation with the Fixation/Permeabilization kit (Invitrogen; Cat#72-5775-40) and staining with Ki67 (SolA15). DAPI was added 7 min before analysis.

In vitro liquid and semi-solid cultures
For limiting dilution analysis, sorted cells were plated in 1:3 diluting densities starting at 27 cells/well until 1 cell/well was reached in complete medium OPTI-MEM with 10% FCS, penicillin (50 U/ml), streptomycin (50 µg/ml), and β-mercaptoethanol (50 µM) supplemented with a saturating amount of the following cytokines: macrophage CSF, GM-CSF, c-Kit ligand, Epo (R&D Systems; 959-ME), and thrombopoietin (R&D Systems; 488-TO) for myeloid and erythroid differentiation. Except if stated otherwise, cytokines were obtained from the supernatant of myeloma cell lines (provided by F. Melchers, Deutsches Rheuma-Forschungszentrum, Berlin, Germany) transfected with cDNA encoding those cytokines. After 5-7 d, wells were assessed for the presence of hematopoietic colonies. Cell frequencies, determined with extreme limiting-dilution analysis software from the Walter and Eliza Hall Institute Bioinformatics Division, are presented as the number of positive wells and the number of total tested wells (Hu and Smyth, 2009).
For colony-forming assays, sorted cells were plated at 100 cells/35 mm culture dishes in duplicate in Methocult M3434 (StemCell Technologies; Cat# 03434) as described by the manufacturer. CFU-Es were assessed at 3 d and remaining colonies at 7 d.
In vivo analysis of lineage potential Pregnant WT females were anesthetized by intraperitoneal injection of a solution of ketamine 10 mg/ml plus xylazine 1 mg/ml diluted in PBS (50-100 µl per 10 g body weight), using a 1-ml syringe. E13.5 GFP + Kit + CD24 + or CD45 + LK FL cells from UBC-GFP embryos were purified and injected intraperitoneally into recipient E13.5 WT embryos (20,000 cells/embryo) of anesthetized females. Sham controls were performed. The FL and the fetal blood of injected and control embryos were analyzed at E16.5.
Multiplex single-cell qPCR Single cells were sorted directly into 96-well plates loaded with RT-STA Reaction mix (CellsDirect One-Step qRTPCR Kit; Invitrogen; Cat# 11753-500; according to the manufacturer's procedures) and 0.2× specific TaqMan Assay mix and were kept at −80°C at least overnight. TaqMan probes used were as follows: For each subset analyzed, a control well with 20 cells was also sorted. Pre-amplified cDNA (20 cycles) was obtained according to manufacturer's instructions and was diluted 1:5 in Tris-EDTA (TE) buffer for qPCR. Multiplex qPCR was performed using the microfluidics Biomark HD system for 40 cycles (Fluidigm) as previously described (Chea et al., 2016). The same TaqMan probes were used for both RT/preamp and qPCR. Only single cells for which at least two housekeeping genes could be detected before 20 cycles were included in the analysis.

Quantification and statistical analysis
All results are shown as mean ± SD. Statistical significance was determined using one-way ANOVA followed by Tukey's multiple comparison test where a P value of <0.05 was considered significant and a P value >0.05 was considered not significant.