PDGF Receptor Alpha+ Mesoderm Contributes to Endothelial and Hematopoietic Cells in Mice

Background: Early mesoderm can be classified into Flk-1+ or PDGF receptor alpha (PDGFRα)+ population, grossly representing lateral and paraxial mesoderm, respectively. It has been demonstrated that all endothelial (EC) and hematopoietic (HPC) cells are derived from Flk-1+ cells. Although PDGFRα+ cells give rise to ECs/HPCs in in vitro ES differentiation, whether PDGFRα+ population can become hemato-endothelial lineages has not been proved in mouse embryos. Results: Using PDGFRαMerCreMer mice, PDGFRα+ early mesoderm was shown to contribute to endothelial cells including hemogenic ECs, fetal liver B lymphocytes, and Lin-Kit+Sca-1+ (KSL) cells. Contribution of PDGFRα+ mesoderm into ECs and HPCs was limited until E8.5, indicating that PDGFRα+/Flk-1+ population that exists until E8.5 may be the source for hemato-endothelial lineages from PDGFRα+ population. The functional significance of PDGFRα+ mesoderm in vascular development and hematopoiesis was confirmed by genetic deletion of Etv2 or restoration of Runx1 in PDGFRα+ cells. Etv2 deletion and Runx1 restoration in PDGFRα+ cells resulted in abnormal vascular remodeling and rescue of fetal liver CD45+ and Lin-Kit+Sca-1+ (KSL) cells, respectively. Conclusions: Endothelial and hematopoietic cells can be derived from PDGFRα+ early mesoderm in mice. PDGFRα+ mesoderm is functionally significant in vascular development and hematopoiesis from phenotype analysis of genetically modified embryos. Developmental Dynamics 242:254–268, 2013. © 2013 Wiley Periodicals, Inc. Key Findings PDGF receptor alpha–positive mesoderm contributes to endothelial and hematopoietic cells in physiological mouse embryogenesis. PDGF receptor alpha–positive mesoderm in early embryo is distinct from yolk sac blood island mesoderm, representing a source of hematopoietic cells on the embryo proper side. Genetic manipulation of Etv2 or Runx1 in PDGF receptor alpha–positive mesoderm demonstrates the functional significance of this mesoderm subset in vascular development and hematopoiesis.


INTRODUCTION
Mesoderm is an intermediate state for epiblasts or ES cells to develop into endothelial (EC) and hematopietic cells (HPC). Early mesoderm can be classified grossly into Flk-1þ lateral-extraembryonic or PDGFRaþ paraxial mesoderm. While Flk-1þ mesoderm has been shown to contribute to virtually all HPCs and ECs (Lugus et al., 2009), several reports indicate that HPCs and ECs originate from somite, intraembryonic mesenchyme, or allantois containing PDGFRaþ mesoderm (Jukkola et al., 2005;Bertrand et al., 2005;Zeigler et al., 2006;Corbel et al., 2007). Among candidate origins for HPCs and hematopoietic stem cells (HSCs), yolk sac (YS) Yoder et al., 1997;Samokhvalov et al., 2007) and lateral mesoderm (Wasteson et al., 2008) consist mainly of Flk-1þ cells with few or not so many PDGFRaþ cells. In contrast, more PDGFRaþ cells are present in intraembryonic para-aortic splanchnopleura (p-Sp) (Cumano et al., 1996), placenta (Gekas et al., 2005), and allantois, raising the possibility that PDGFRaþ cells can generate HPCs from these sites . It has been demonstrated that PDGFRaþ mesoderm generates ECs and HPCs in ES differentiation culture (Sakurai et al., 2006). However, in vitro culture system defines each cell population by marker staining but lacks anatomical information that is critical to understand the physiological differentiation process. In mouse embryos, we have demonstrated that Flk-1þ/PDGFRaþ cells accumulate in the absence of Etv2, failing to differentiate into Flk-1þ/ PDGFRa-cells (Kataoka et al., 2011). This suggests that PDGFRaþ cells can contribute to ECs and HPCs in mouse embryogenesis. In mouse development, however, how PDGFRaþ population including Flk-1þ/PDGFRaþ cells contribute to various cell types has not been thoroughly evaluated. It is also important to confirm if the differentiation pathway in in vitro ES cell differentiation can be recapitulated in the real animal. In ES differentiation, it is expected that PDGFRaþ/Flk-1þ cells are multipotential for hemato-endothelial, muscle, or mesenchymal lineages partly due to the greater plasticity of differentiating ES cells. Since Flk-1þ cells have been shown to differentiate into skeletal muscle and cardiomyocytes in mouse embryos (Motoike et al., 2003), it is possible that PDGFRa induction in Flk-1þ cells might enforce the differentiation of Flk-1þ cells preferentially into muscle or mesenchymal lineages in the in vivo context. Therefore, we examined if PDGFRaþ cells contribute to ECs and HPCs in mouse embryos where differentiation is controlled in a more physiological manner. For this purpose, PDGFRa-MerCreMer (PRa-MCM) knock-in mice, expressing tamoxifen (Tmx) inducible MerCreMer (MCM) under control of the PDGFRa locus (Fig. 1A), was crossed with ROSA26-LacZ or YFP reporter strains (PRa-MCM-LacZ or PRa-MCM-YFP mice) to trace labeled PDGFRaþ cells in mouse embryos. We focused on ECs and HPCs derived from PDGFRaþ cells, as this may help to clarify the origin of HSCs that are one of the most important cell types to be created for therapeutic purposes.

PDGFRa Mesoderm Is Distinct From Extraembryonic Runx11 Mesoderm in Early Embryos
To locate the PDGFRaþ mesoderm, E7.5 neural plate (Fig. 1B), E8.0 head fold (Fig. 1C), or E8.5 somite stage (Fig. 1D), embryos were immunostained by PDGFRa, Flk-1, and Runx1 antibodies. As we reported, PDGFRa and Flk-1 stained almost distinct subset of mesoderm with some overlap in lateral mesoderm closer to the paraxial region (Kataoka et al., , 2011. Runx1 was used to stain HPC precursors including erythroid progenitors and part of HSCs (Tanaka et al., 2012). No clear overlap was observed between PDGFRaþ and Runx1þ mesoderm, indicating that PDGFRa or Runx1 specifies distinct mesoderm population. This result was also confirmed by FACS analysis of NP-and HF-stage Runx1-Venus Knock-in embryos, in which almost no PDGFRaþ/Runx1þ cells were detected (Fig. 1E). In situ hybridization for Runx1 also revealed that its expression is limited in the proximal region of the extraembryonic yolk sac, namely the blood island, validating that our immunostaining by Runx1 antibody for multi-color detection correctly reflects Runx1 in situ hybridization (data not shown). These findings suggest that any HPCs coming from PDGFRaþ cells develop from those cells that do not express Runx1 in these stages. At E7.5-8.5, we were able to detect an area stained by both PDGFRa and Flk-1. This double-positive population almost disappeared at E9.5 (see Fig. 6C), indicating that vasculogenic capacity in PDGFRaþ cells is dependent on Flk-1 and is limited in early time frame during embryogenesis.
As we aimed to analyze the contribution of PDGFRaþ cells into ECs and HPCs, LacZ-stained embryos were examined if there were any ECs labeled, especially those regarded as hemogenic ECs. We found that some ECs were LacZ positive in PRa-MCM-LacZ embryos exposed to Tmx at E7.5-8.0 (Fig. 2B). Notably, ECs on the ventral side of aorta in the AGM region were LacZ positive, including HPCs budding from the EC layer (Fig.  2B). This finding demonstrates that PDGFRaþ cells labeled at E7.5-8.0 can contribute to hemogenic ECs that potentially give rise to definitive HPCs. Tmx injection at E6.5 failed to label PDGFRaþ mesoderm derivatives such as somites or paraxial mesoderm (Fig. 2C), which were extensively labeled by E7.5 injection. Rather, E6.5 Tmx injection labeled mainly extraembryonic part including YS HPCs, demonstrating the time frame specificity of labeling in PRa-MCM line (Fig. 2C), as 24 hr delay of Tmx injection resulted in the labeling of distinct populations.
These findings suggest that PRa-MCM transgene has labeled in relevant cell types that are expected to be the descendants of PDGFRaþ cells including cardiomyocytes and head mesenchyme. Additionally, we observed LacZ-positive ECs including those with potential hemogenic capacity. Tmx injections at E6.5 and E7.5 also support our assumption that Tmx will be almost ineffective 24 hr after injection (Zovein et al., 2008), ensuring the time frame specificity of the PRa-MCM transgene.
PDGFRa1 Cells Labeled at E7.5-8.0 Contribute to Fetal Liver HPCs and ECs at E15.5 To analyze the contribution of PDGFRaþ early mesoderm into HPCs, especially for fetal liver HPCs, ECs/ HPCs in PRa-MCM-YFP embryos were analyzed by FACS at E15.5 after E7.5 or E8.0 Tmx injection into pregnant females. After Tmx injection at E7.5, 6-7% of B lymphocytes and KSL cells were YFPþ, indicating the contribution of PDGFRaþ mesoderm to de-finitive type HPCs (Fig. 3A,a). Tmx injection at E8.0 labeled over 20% of B cells or KSL cells at E15.5 (Fig. 3A,b). Since most of PDGFRaþ cells reside in the intraembryonic area at E8.0, this finding is consistent with the report that intraembryonic caudal mesenchymal cells develop into aorta-gonad-mesonephron (AGM) HSCs (Bertrand et al., 2005). PDGFRaþ/Etv2-YFPþ cells, potentially developing into ECs/ HPCs, are also demonstrated to exist on the embryo proper side at E7.5 (Koyano-Nakagawa et al., 2012).  , 79,307-85,194; 3 0 side, 85,253-89,284 from RP23-55P22. Correct integration was confirmed by Southern blotting. PCR genotyping can be performed by primers, PRa WT Rev1; ggagaacaaggacgcatgtgtgg, PRa5 0 F2; gcttccctcttatctatctgactg, MCM Rev2; aaggtggacctgatcatggagattc (WT, 260 bp; Targeted, 700 bp). SA, splice acceptor; pA, polyadenylation signal. B: Immunostaining of PDGFRa, Flk-1, and Runx1 in NP stage (E7.5) mouse embryos. NP-stage wild type embryos were stained by PDGFRa, Flk-1, and Runx1 antibodies. There was a narrow overlap of PDGFRa and Flk-1 domains. Note, however, that PDGFRa and Runx1 expression was seen in distinct mesoderm subsets. Scale bar¼100 mm. C: Immunostaining of PDGFRa, Flk-1, and Runx1 in HF stage (E8.0) mouse embryos. In HF-stage embryos, similar staining was observed as in NP-stage samples. Note also in this stage that there is no overlap between PDGFRa and Runx1þ populations. Scale bar¼100 mm. D: Immunostaining of PDGFRa, Flk-1, and Runx1 in E8.5 somite stage embryos. As seen in E7.5 or E8.0 embryos, there is some overlap between PDGFRa and Flk-1 areas, however no overlap between PDGFRa and Runx1. In this stage, PDGFRa or Flk-1 antibody stains form somite structure or aortas, respectively. Scale bar¼100 mm. E: FACS analysis of NP-and HF-stage embryos. NP-(left) or HF-(right) stage embryos were stained by indicated antibodies to examine the relationship between Runx1, PDGFRa, and Flk-1. In both stages, almost no Runx1þ/PDGFRaþ cells were detected, while more that 70% of Runx1-GFPþ cells were Flk-1þ. To activate MerCreMer, Tmx was injected into pregnant females at E7.5 (neural plate stage, top panels) or E8.0 (head fold stage, middle panels), when PDGFRa starts to be clearly detected in mesoderm with initiation of vasculogenesis and hematopoiesis. Embryos were LacZ stained at E10.5. LacZ-stained cells were distributed mainly in the intraembryonic area including the head, heart, and umbilical cord (arrows in lower panel). Note that only a few labeled cells were seen in the yolk sac except the umbilical cord (b, d). Scale bar¼1 mm. B: Histological analysis of LacZ-stained PRa-MCM-LacZ embryos with Tmx injection at E7.5 (a-c) or E8.0 (d-f) into pregnant females. Embryos were LacZ stained and sectioned at E10.5. LacZ-positive cells were distributed in cranial mesenchyme, atrial and ventricular cardiomyocytes. Higher magnification is shown on the right side of a-f. Note the presence of labeled ECs in cranial vessels (a, d) and ventral wall of the descending aorta (arrows on right in c, f). Presence of LacZþ ECs and HPCs in the aorta-gonad-mesonephron region is consistent with the finding that fetal liver HSCs exist as E7.5-8.5 labeled cells in the fetal liver (see Figs. 3A and 6A,a). Scale bar¼100 mm. C, D: Shortterm tracing of PDGFRaþ cells from E7.5 to E8.0-8.5. C: PRa-MCM-LacZ embryos were analyzed at E8.0 and E8.5 after Tmx injection into pregnant females at E6.5 or E7.5. Note that labeled cells are present in the lateral mesoderm area that overlaps Flk-1 staining (see Fig. 1C) at E8.0 ("Posterior" panels). Existence of some labeled cells in this lateral area where Flk-1þ cells exist at E8.0 supports that part of the labeled cells might become Flk-1þ. At E8.5,labeled cells become more concentrated around the paraxial zone (bottom panels). Tmx injection at E6.5 labeled YS HPCs that are different from those labeled by E7.5 injection (top panels). Scale¼100 mm. D: Sections of LacZ-stained E8.5 embryos after E6.5 injection demonstrate the labeling in YS blood islands (arrowheads) and cardiac tissue (arrows). Note that part of YS HPCs (top panel) and cardiac tissue (bottom panel) contains LacZþ cells. Scale¼100 mm.

Fig. 3. A:
PDGFRaþ cells labeled at E7.5 or E8.0 contribute to fetal liver HPCs. a: Pregnant females with PRa-MCM-YFP embryos were Tmx injected at E7.5. Contribution of YFPþ cells to fetal liver HPCs was analyzed at E15.5. To evaluate the contribution of labeled cells to definitive HPCs, B cell and KSL populations were gated. YFPþ cells constituted 6-7% of B or KSL population gated. (YFP expression from ROSA locus is the indicator of Cre recombination by PRa-MCM transgene.) b: PRa-MCM-YFP embryos were analyzed at E15.5 after Tmx injection at E8.0 into pregnant females. To evaluate the contribution of labeled cells to definitive HPCs, B cell and KSL populations were gated. YFPþ cells constituted 22 or 25% of B and KSL population, respectively. B: PDGFRaþ cells from embryo proper develop into B cells in explant culture. After Tmx injection into pregnant females at E7.5, caudal part of E8.25 PRaMCM-YFP embryos was explanted and cultured on OP9 feeder cells with IL-7 (20 ng/ml) and Flt-3 ligand (10 ng/ml). After 2 days, explants were dissociated and further cultured in the same condition for another 2 weeks. Floating cells were harvested and stained by anti-CD19 antibody to demonstrate the B cell generation. Note that significant proportion of CD19þ B cells exist as YFPþ in cultured cells from the PRaMCM-YFP embryo (right panel).

Contribution of embryo proper
PDGFRaþ cells into B lymphocytes, reflecting definitive hematopoietic potential, was further confirmed by E8.25 p-Sp explant culture following E7.5 Tmx injection (Fig. 3B). As explant was started from early somite stage before blood circulation, this finding supports that at least part of the definitive hematopoiesis can be established from precursors present in or close to the embryo proper. Fate of YFP-labeled cells by E7.5 Tmx injection was further traced to adult HPCs. In adult PRa-MCM-YFP (Tmx injected at E7.5) mice, we found that labeled cells contributed to multiple lineage BM cells including KSL population as well as lymphocytes in spleen and thymus (Fig. 4). This finding supports the notion that E7.5 PDGFRaþ mesoderm can contribute to hematopoietic progenitors in adult mice. We have observed contribution of PDGFRaþ mesoderm to ECs in PRa-MCM-LacZ embryos at E10.5. Consistent with this finding at E10.5, we confirmed EC contribution also at E15.5 in PRa-MCM-YFP embryos. Around 7 or 10% of VE-cadherinþ/Flk-1þ ECs in the cranial region became YFPþ at E15.5 by E7.5 or E8.0 Tmx injection, respectively ( Fig. 5A, B), demonstrating the contribution of PDGFRaþ mesoderm into cranial ECs at this stage. As cranial ECs arise from the Etrp/Etv2þ clusters in zebrafish (Proulx et al., 2010) and Etv2 induction is colocalized with Flk-1 (Kataoka et al., 2011), these ECs may be from PDGFRaþ/Flk-1þ or PDGFRaþ/Flk-1-cells primed to express Flk-1 to acquire vasculogenic potential.

PDGFRa1 Mesoderm Contribution to ECs and HPCs Is Limited Until E8.5 Embryos
Contribution of PDGFRaþ cells to E15.5 fetal liver HPCs and cranial ECs declined when Tmx was injected at E8.5 (Fig. 6A a, B,a) compared to E7.5-8.0 Tmx injection, becoming almost undetectable by E9.5 injection (Fig. 6A,b, B,b). Since FACS analysis shows that PDGFRaþ/Flk-1þ cells profoundly reduced at E9.5 from E7.5-8.5 (Fig. 6C), these findings suggest that PDGFRaþ cells contributing to ECs/HPCs may be mainly from PDGFRaþ/Flk-1þ population. This limited time frame for the ECs/HPCs generation from PDGFRaþ cells also suggest that ECs and HPCs develop from PDGFRaþ cells rather immediately while those cells keep Flk-1 expression. To support this idea, we observed the existence of labeled YFPþ/VE-cadherinþ cells in E8.0 embryos after E7.5 Tmx injection in 12 hr. It is also likely that these E7.5 PDGFRaþ cells can contribute to hemato-endothelial lineages mainly on the embryo proper side, because we observed labeled cells in PRa-MCM-LacZ embryos mostly in the embryo proper and lateral mesoderm, failing to detect LacZ-labeled cells in the extraembryonic yolk sac region by E7.5 Tmx injection (Fig. 2C, middle and bottom panels). This result was further supported by FACS analysis of PRa-MCM-YFP embryos exposed to Tmx at E7.5. In these embryos, YFPþ-labeled cells including YFPþ/VE-cadherinþ cells are mainly present in the embryo proper. Absence of YFPþ/CD41þ cells also indicates that primitive HPCs in the extraembryonic yolk sac region were not labeled by PRa-MCM transgene in these embryos (Fig. 6D,a, b).

Functional Significance of PDGFRa1 Mesoderm in Vascular Development and Hematoipoiesis
Results presented in the preceding parts indicate that PDGFRaþ cells contribute to ECs and HPCs. However, the functional importance of ECs and HPCs derived from PDGFRaþ mesoderm remains unclear. To examine the functional importance of ECs/HPCs derived from PDGFRaþ mesoderm, we generated PRaBAC-CreERþ Etv2fl/-mice (PRaCreER-Etv2KO) to delete Etv2, which is absolutely required for ECs/HPCs commitment from primitive mesoderm. Etv2 inactivation in PDGFRaþ mesoderm by E8.0 Tmx injection resulted in changes in vitelline plexus or intersomatic vessels (Fig. 7A), indicating the importance of PDGFRaþ mesoderm-derived ECs in vascular development. However, the defects observed were variable among analyzed embryos for several reasons. First, Etv2 expressed in Flk-1þ/PDGFRaþ mesoderm may be deleted after Etv2 has activated EC differentiation genes. Due to its transient requirement, PDGFRaþ mesoderm can generate ECs if Etv2 is deleted after EC commitment (Kataoka et al., 2011;Wareing et al., 2012;Lammerts van Bueren and Black, 2012). Second, Etv2 deletion in PDGFRaþ cells can be compensated by other EC sources such as Flk-1þ PDGFRa-population. Nevertheless, the observation that the embryonic vasculature is partly impaired supports the functional significance of ECs derived from PDGFRaþ mesoderm.
Etv2þ mesoderm has been shown to give rise to both ECs and HPCs. Thus, it is possible that Etv2 ablation in primitive mesoderm can affect HPCs as well. Analysis of caudal half of E10.5 embryo proper revealed reduced CD45þ and CD45 low /Kit high populations in $50% of PRaCreER-Etv2KO embryos compared to control littermates. This HPC defect was seen in embryos showing concomitant reduction of CD34þ cells (Fig. 7B), indicating that HPC defect is accompanying EC loss. Since E10.5 CD45 low /Kit high population is enriched for hemogenic potential (Nobuhisa et al., 2012), this result indicates that EC loss derived from PDGFRaþ mesoderm can include hemogenic ECs. As another functional test for HPC contribution, fetal liver HPCs were analyzed after Runx1 restoration in E7.5 PDGFRaþ cells in Runx1-LacZ homozygous embryos, which are functionally deficient for Runx1. As reported previously, this Runx1-LacZ allele can restore Runx1 expression after Cre recombination (Samokhvalov et al., 2006;Tanaka et al., 2012). Runx1 restoration by PDGFRa-MCM transgene partly rescued CD45þ or KSL HPCs in PRa-MCM;Runx1-LacZþ/þ fetal liver while these populations were totally absent from Runx1-LacZþ/þ embryos, indicating that there is some rescue by HPCs derived from E7.5 PDGFRaþ cells (Fig. 8A). To further examine the rescue of Runx1 null phenotype in PRa-MCM;Runx1-LacZþ/þ embryos, colony-forming assay was performed using fetal liver cells. While Runx1-LacZþ/þ liver cells failed to generate colonies in methylcellulose, we observed partial recovery of colonyforming units in PRa-MCM;Runx1-LacZþ/þ fetal liver cells (Fig. 8B). It should be noted that Runx1-LacZþ/þ embryos with rescue in PDGFRaþ cells looked pale and modestly anemic and unable to survive beyond E12.5. This is presumably due to the lack of Runx1 rescue in YS, which is reported to be important for primitive and definitive erythropoiesis (Yokomizo et al., 2008). Since PRa-MCM transgene will not induce Cre recombination in YS by E7.5 Tmx injection, erythropoiesis in YS will not be rescued in PRa-MCM;Runx1-LacZþ/þ (E7.5 Tmx) embryos. These results suggest that embryo proper and/or lateral mesoderm Runx1 can restore HPCs to some extent without Runx1 restoration in YS.
To further characterize the rescued cells in the in vivo context, PRa-MCM;Runx1-LacZþ/þ (E7.5 Tmx) fetal liver cells were injected into irradiated recipient mice to evaluate the repopulation capacity. Two months after transplantation into NOD-Scid recipients, we found the significant contribution of PRa-MCM;Runx1-LacZþ/þ (E7.5 Tmx) fetal liver cells into circulating HPCs including monocyte/macrophage, B, and T cells (Fig.  8C). These results suggest that E7.5 PDGFRaþ mesoderm have the potential to give rise to multi-lineage HPCs, possibly HSCs.

PDGFRa Is Dispensable for the Development of Fetal Liver HPCs
Although PDGFRaþ mesoderm was demonstrated to give rise to hemogenic ECs and HPCs, whether PDGFRa signaling is required for hematopoiesis has not been determined. In PDGFRa null embryos, PDGFRa seems to be dispensable for HPCs development since the embryos die apparently from cephalic closure defect or skeletal abnormalities (Soriano, 1997). However, fetal liver seems to be normally formed in null mutants. To confirm that PDGFRa signaling is dispensable for HPC development, FACS analysis of fetal liver cells from homozygous PRa-MCM embryos was performed. PRa-MCM homozygous embryos, demonstrated to be functionally null for PDGFRa, showed indistinguishable HPCs profile from control embryos (Fig.9A, B). This result suggests that  PDGFRaþ cells labeled at E7.5 or E8.0 contribute to cranial ECs. A: Head regions from PRa-MCM-YFP embryos were analyzed at E15.5 after Tmx injection into pregnant females at E7.5. By FACS analysis, $20% of VE-cadherinþ/Flk-1þ cells or $4.5% of CD45þ/Kitþ cells were YFPþ, indicating that E7.5 PDGFRaþ mesoderm contributes to cranial ECs and HPCs. B: Pregnant females with PRa-MCM-YFP embryos were Tmx injected at E8.0. Contribution of YFPþ cells to cranial ECs analyzed at E15.5. By FACS analysis, $10% of VE-cadherinþ/Flk-1þ cells were labeled as YFPþ, indicating that a significant proportion of cranial ECs are derived from E8.0 PDGFRaþ mesoderm. Note also that $5% of CD45þ/ Kitþ cells were YFPþ, indicating the contribution of E8.0 PDGFRaþ mesoderm into HPCs.  6. A: PDGFRaþ cells labeled at E8.5 contribute to fetal liver HPCs, but with lower efficiency. By E9.5 labeling, almost no PDGFRaþ cells contribute to fetal liver HPCs. a: After E8.5 Tmx injection, fetal liver HPCs were analyzed. YFPþ-labeled cells contributed to $2% of KSL cells but are almost negligible in B cell population, indicating that the contribution of PDGFRaþ mesoderm significantly declines around E8.5. b: In fetal liver HPCs, no YFPþ cells were observed after E9.5 Tmx injection into pregnant females. B: Contribution of YFPþ cells to cranial ECs/HPCs was analyzed in E15.5 PRa-MCM-YFP embryos after Tmx injection at E8.5 or E9.5 into pregnant females. a: In cranial samples, almost no E8.5-labeled YFPþ cells contribute to either Flk-1þ (VE-cadherinþ) ECs or CD45þ HPCs, indicating that only early stage PDGFRaþ mesoderm can give rise to cranial ECs and HPCs. Note that the majority of YFPþ cells existed as PDGFRaþ population. b: In E15.5 cranial samples, YFPþ cells were present not in Flk-1þ (VE-cadherinþ) or CD45þ population, but in the PDGFRaþ population after E9.5 Tmx injection. C: PDGFRaþ/Flk-1þ cells that exist from E7.5-8.5 decreased profoundly at E9.5. Wild type embryos were stained by anti-PDGFRa and Flk-1 antibodies and analyzed by FACS. Note that the proportion of cells co-expressing PDGFRa and Flk-1 has significantly declined at E9.5 from earlier embryos. D: FACS analysis of the PRa-MCM-YFP embryos demonstrating that VE-cadherinþ ECs were labeled within 24 hr after E7.5 Tmx injection. To trace the short-term fate of PDGFRaþ cells, Tmx was injected into pregnant females at E7.5. YS (a), and embryo proper (b) from PRa-MCM-YFP embryos were analyzed at E8.0 after 12 hr. YFPþ cells were present mainly in the embryo proper (b), but very few in the YS (a). We could not detect any labeled CD41þ primitive HPCs either in yolk sac or embryo proper. Note, however, that part of VE-cadherinþ cells were labeled in the embryo proper (red box), suggesting that VE-cadherinþ/PDGFRa-ECs can differentiate from PDGFRaþ mesoderm within 12 hr (b, right panel).
PDGFRa itself is dispensable for hematopoiesis to the fetal liver stage. Despite this dispensability in early hematopoiesis, our finding that PDGFRaþ cells give rise to HPCs suggests that PDGFRa is useful for the tracing of particular mesoderm subsets contributing to HPCs.

DISCUSSION
In this study, we have traced the fate of PDGFRaþ mesoderm cells from early embryos and showed that (1) PDGFRaþ mesoderm in early embryos (possibly, PDGFRaþ/Flk-1þ cells) contributes to ECs and HPCs including B lymphocytes and KSL cells, (2) PDGFRaþ mesoderm at E7.5-8.0 giving rise to ECs and HPCs represents a distinct population from extraembryonic Runx1þ population, and that (3) ECs and HPCs from PDGFRaþ mesoderm have functional significance revealed by Etv2 deletion or Runx1 restoration, respectively. It has been demonstrated that all HPCs and ECs are derived from Flk-1þ cells in mice.
in early mouse embryos, we demonstrated that ECs and HPCs derived from PDGFRaþ mesoderm exist as a physiological population in developing embryos and postnatal mice. PDGFRaþ mesoderm first emerge at E6.5-E7.0 around the proximal part of the primitive streak . Thus, Tmx injection at E6.5 labeled cells that migrate to blood islands corresponding to the proximal part of the extraembryonic mesoderm. As PDGFRaþ cells exist mostly on the embryo proper side at E7.5, Tmx injection at E7.5 labeled cells giving rise to embryo proper structures including somites, head mesenchyme, and dorsal aorta that hardly overlap with those labeled by E6.5 injection (Fig. 2C). This result demonstrates that at least in these developmental stages Tmx activity does not persist beyond 24 hr (Zovein et al., 2008) and validates the time frame specificity of the PRaMCM transgene. This difference of labeling along the primitive streak is also consistent with the injecteddye tracing studies that almost distinguished the descendants migrating into extra or intra embryonic regions from spatiotemporally different parts of the primitive streak (Wilson and Beddington, 1996;Kinder et al., 1999). At E7.5-8.0, immunostaining revealed that PDGFRaþ mesoderm population exists in embryo proper and lateral mesoderm close to the embryo proper side, which is distinct from Runx1þ population in the extraembryonic yolk sac. Several places in early embryo are proposed as candidate sites generating HSCs, including yolk sac, p-Sp, intraembryonic caudal mesenchyme, allantois, and placenta. Our finding that PDGFRaþ cells labeled at E7.5 contribute to ECs and HPCs including B lymphocytes and Fig. 9. PDGFRa is dispensable for fetal liver HPCs development. PRa-MCM heterozygous mice were used to generate PDGFRa-deficient embryos. Fetal liver HPCs from PRa-MCM heterozygous or homozygous embryos were analyzed at E12.5. A: Parts of embryos were stained by anti-PDGFRa antibody confirming that the PRa-MCM homozygous embryo lacked the population expressing PDGFRa B: Fetal liver HPCs from PRa-MCM heterozygous or homozygous embryos were analyzed. Homozygous PRa-MCM embryos were assumed to be functionally deficient for PDGFRa, which was confirmed by antibody staining. Ter119þ, CD45þ, and KSL populations were similar between control heterozygous and PDGFRa deficient embryos, indicating that PDGFRa is dispensable for the fetal liver HPCs development at least until E12.5. Numbers of colonyforming units using heterozygous and homozygous cells were also indistinguishable (data not shown).
fetal liver and BM KSL cells suggests that at least part of the definitive hematopoietic cells originate in locations corresponding to p-Sp including part of the lateral mesoderm or allantois distinct from extraembryonic YS. This notion was also supported by p-Sp culture experiment from PRa-MCM-YFP embryos exposed to Tmx at E7.5. Furthermore, we showed that PDGFRaþ mesoderm labeled at E7.5 can give rise to multi-lineage adult BM HPCs. Repopulation of Runx1 restored cells derived from E7.5 PDGFRaþ mesoderm also supports that PDGFRaþ cells can be a source for HSCs.
It is noteworthy that p-Sp region contains PDGFRaþ/Flk-1þ cells that are likely to be the source for ECs and HPCs. Previous reports including ours suggest that Flk-1 (and Etv2 coexpressed in Flk-1þ population) is necessary for EC/HPC development in mouse embryos. Whether PDGFRaþ mesoderm-derived ECs/ HPCs are generated directly from PDGFRaþ/Flk-1þ population or there are PDGFRaþ/Flk-1-cells primed to express Flk-1 needs to be determined.
We also noted that PDGFRaþ cells contribute to the ventral side of the aorta where hemogenic ECs are present. In short-term tracing experiment, Tmx injection at E7.5 labeled cells that were distributed in the caudal-lateral mesoderm at E8.0 (Fig.  2C, middle panels). Although the molecular mechanism as to how labeled cells first distribute laterally before settling in the paraxial region at E8.5 (Fig. 2C, bottom panels) is unclear, mesoderm cells in this region at E8.0 overlap cells labeled by Hoxb6Cre transgene, which is known to label cells that contribute the hemogenic ECs on the ventral side of the aorta (Wasteson et al., 2008). Indeed, descendants of Hoxb6Cre-labeled cells have been demonstrated to contribute to adult bone marrow HPCs (Zovein et al., 2010), supporting our result that part of early PDGFRaþ cells can contribute to KSL population in fetal liver and adult HPCs. In chicken, it is postulated that ECs derived from splanchnopleura with hemogenic potential distribute on the ventral side of aorta, which are replaced by non-hemogenic somite-derived ECs (Pouget et al., 2006). In mouse, PDGFRaþ cells are assumed to give rise mainly to paraxial mesoderm including somites. However, we found that early PDGFRaþ mesoderm contributes to hemogenic ECs possibly through lateral mesoderm, which corresponds to the Hoxb6-positive area. The molecular mechanism as to how part of the PDGFRaþ cells can give rise to lateral mesoderm in addition to somites is unknown at present. Nevertheless, we should note that at E7.5-E8.0 VEGF is broadly expressed in the visceral endoderm over extraembryonic-lateral mesoderm (Miquerol et al., 1999), suggesting that PDGFRaþ cells responding to VEGF expressed in that area obtain the hemato-endothelial fate.
This study also showed the ECs and HPCs derived from PDGFRaþ mesoderm are functionally important in vascular development and hematopoiesis by genetic manipulation of key factors in PDGFRaþ mesoderm. Previous investigation showed that ECs can be derived from early somites where PDGFRa is highly expressed. Finding that Etv2 deletion in PDGFRaþ mesoderm causes vascular defect also suggests that part of the paraxial mesoderm may be a significant source of ECs for embryonic vascular development. Despite the defective vascular patterning in vitelline or intersomatic vessels, we were unable to detect significant changes in the anterior part of the vasculature including cranial region of PRaCreER-Etv2KO embryos. It has been reported that anterior part of the mouse embryonic vasculature forms earlier than the posterior part (Drake and Fleming, 2000). Thus, we speculate that compared to the posterior region, more ECs in the anterior part already might have passed the phase of transient Etv2 requirement at E8.0 when Tmx was injected. Additionally, the finding that PDGFRaþ cells hardly contribute to HPCs or ECs if Tmx was injected after E9.0-9.5 suggests that the vascular phenotype may be coming from Etv2 deletion in PDGFRaþ cells until E8.5. It is also noteworthy that caudal-lateral mesoderm, labeled by Hoxb6 Cre, that gives rise to vitelline plexus overlaps PDGFRaþ area, which is consistent with the change we observed in vitelline vasculature in PRaCreER-Etv2KO embryos.
It has been controversial whether HSCs are mainly coming from extraembryonic yolk sac or intraembryonic part. We could observe some restoration of HPCs in Runx1-deficient fetal liver by Runx1 restoration in PDGFRaþ early mesoderm. PRa-MCM;Runx1-LacZþ/þ (E7.5 Tmx) embryos were apparently pale, partly due to the bleeding as seen in Runx1 null embryos. Runx1 may be dispensable for primitive erythropoiesis but required for the proper erythroid development, making Runx1 null embryos modestly anemic (Yokomizo et al., 2008). Based on the labeling experiments, we can predict that activation of PRaMCM transgene at E7.5 will fail to rescue Runx1 in extraembryonic YS blood island. Runx1 deficiency in YS blood island will create modest defects in erythropoiesis in PRaMCM; Runx1-LacZþ/þ embryos even after E7.5 Tmx injection. However, some rescue in Runx1 dependent CD45þ and KSL populations, colonyforming units, and transplantable cells in PRa-MCM;Runx1-LacZþ/þ fetal liver after E7.5 Tmx injection demonstrate that embryo proper side mesoderm has the potential to contribute to hemogenic ECs and possibly HSCs independently of YS extraembryonic mesoderm. So far, the origin of HSCs in early mouse embryos has been mainly investigated by explant culture of specific parts without any cell tracing from specific areas from early embryogenesis. Runx1-MCM-based cell tracing was used to follow the E7.5 extraembryonic yolk sac mesoderm. PRa-MCM transgene gave us an opportunity to perform labeling or gene manipulation specifically on the embryo proper side including lateral mesoderm to address where HPCs can originate. Indeed, PDGFRaþ cells labeled at E7.5 distribute exclusively on the embryo proper side at E8.0-8.5, indicating that labeling (Cre activity) is limited to the embryo proper and lateral mesoderm close to embryo proper. HPC rescue to achieve repopulating fetal liver cells by Runx1 restoration in E7.5 PDGFRaþ cells suggests that these cells can be the origin of HPCs and HSCs. On the other hand, several reports claim that extraembryonic YS can be a source for HSCs (Yoder et al., 1997;Tanaka et al., 2012). How YS-derived HPCs/ HSCs co-exist with PDGFRaþ mesoderm-derived HPCs or how they interact needs further investigation.
We have shown that PDGFRaþ mesoderm generates ECs/HPCs in embryos as well as in ES differentiation culture. While Flk-1 has been used as a primary marker to identify HPC precursors in ES differentiation, recent reports suggest that Flk-1þ cells are heterogenous. Irion et al. (2010) reported that two distinct Flk-1þ populations with different hematopoietic potential are generated from differentiated ES cells, showing that the later day-5 Flk-1þ cells have more definitive hematopoietic potential than day-3 population. It is notable that E7.5-8.0 Flk-1þ/PDGFRaþ cells are present in lateral mesoderm, which is labeled by Hoxb6 Cre transgenic line and shown to become hemogenic ECs. This raises the possibility that introducing PDGFRa may be useful for marking "hemogenic" mesoderm if used at appropriate time points during differentiation. In human ES differentiation, CD34þ cells exist as PDGFRaþ (Davis et al., 2008) supporting the usefulness of PDGFRa to monitor HPC differentiation from ES cells. Some PRaCreER-Etv2KO embryos showed vascular patterning defect in vitelline plexus that is tightly linked to definitive HSC development (Zovein et al., 2010). These results together support the idea that PDGFRa may serve as an additional surface marker to subdivide Flk-1þ cells and enrich useful populations if used in appropriate ES differentiation stages in HPC induction.

Mice
PRaMCM mice were generated by knocking Tmx-inducible MCM cDNA (gift from M. Reth) into the PDGFRa locus. Adenoviral splice acceptor sequence (SA) was added to 5 0 side of the MCM cDNA to prevent cryptic splicing. Homology arms used for targeting were 5 0 side, 79,194; 3 0 side, 85,253-89,284 from RP23--55P22. Frt sites flanked Neo cassette was placed 3 0 to the SA-MCM-polyA sequence. The targeting vector constructed was linearized by AscI and electroporated into TT2 ES cells to obtain targeted ES cells, which was used to generate mice by standard procedure. MerCreMer was activated in embryos by injecting Tmx into pregnant females (approximately 100 mg/g [body weight] 4-OH tamoxifen [Tmx]; Sigma, St. Louis, MO, H7904). Detail for the Etv2 conditional allele will be reported elsewhere. In brief, two loxP sites were placed in the Etv2 locus to flank the DNA-binding domain. Hetrozygosity of the DNA-binding domain deletion caused no obvious phenotype while homozygosity recapitulated the Etv2 null phenotype as reported. PDGFRaBAC-CreER (PRaBAC-CreER) line was from B. Richardson (University College London) (Rivers et al., 2008). Runx1-IRES-GFP knockin mice were a gift from Dr. Lorsbach (St. Jude Children's Research Hospital, Memphis, TN) (Lorsbach et al., 2004). For Runx1 recue experiments, mice were crossed to get PRa-MCM;Runx1-LacZþ/þ alleles that harbor PRaMCM transgene over restorable Runx1 null allele. For colony-forming assays, 10,000 fetal liver cells/35 mm dish were plated in methylcellulose medium (Stem Cell Technology, Vancouver, Canada, M3434). Sublethally radiated (249.9cGy) NOD-Scid mice (Charles River, Wilmington, MA) received a half million fetal liver cells from PRa-MCM;Runx1-LacZþ/þ embryos. After 2 months, peripheral blood samples were analyzed by indicated antibodies to evaluate the contribution of engrafted cells.

Tissue Culture
Explant culture to generate B cells and mice for Runx1 restoration and labeling have been described (Tanaka et al., 2012). In brief, the caudal part of embryo proper was separated from E8.25 early stage embryos and explanted on OP9 cells in the presence of IL-7 (10 ng/ml) and Flt-3L (10 ng/ml). After 2 days, explant was dissociated and continued to be cultured for an additional 12-14 days in the same condition. Wells positive for growing HPCs were analyzed by FACS using indicated antibodies.