Requirement of Runx1/AML1/PEBP2αB for the generation of haematopoietic cells from endothelial cells


Communicated by: Kohei Miyazono*Correspondence: E-mail:


Recent studies revealing that endothelial cells derived from E8.5-E10.5 mouse embryos give rise to haematopoietic cells appear to correspond to previous histological observations that haematopoietic cell clusters are attached to the ventral aspect of dorsal aorta in such a way as if they were budding from the endothelial cell layer. Gene disruption studies have revealed that Runx1/AML1 is required for definitive haematopoiesis but not for primitive haematopoiesis, but the precise stage of gene function is not yet known.

We found that mice deficient in Runx1/AML1 (an α subunit of the transcription factor PEBP2/CBF) lack c-Kit+ haematopoietic cell clusters in the dorsal aorta, omphalomesenteric and umbilical arteries, as well as yolk sac vessels. Moreover, endothelial cells sorted from the embryo proper and the yolk sac of AML1–/– embryos are unable to differentiate into haematopoietic cells on OP9 stromal cells, whereas colonies of AML1–/– endothelial cells can be formed in culture.

These results strongly suggest that the emergence of haematopoietic cells from endothelial cells represents a major pathway of definitive haematopoiesis and is an event that also occurs in the yolk sac in vivo, as suggested by earlier in vitro experiments.


Gene-targeting techniques in mice have revealed a set of transcription factors involved in haematopoiesis (Shivdasani & Orkin 1996; Orkin & Zon 1997), one of which is PEBP2/CBF. RUNX1/AML1/PEBP2αB/CBFA2 encodes the DNA binding α subunit of PEBP2/CBF (Ito 1999) and its deficiency in mice results in a lack of foetal liver haematopoiesis, suggesting that AML1 is required for definitive haematopoiesis (Okuda et al. 1996; Okada et al. 1998; Wang et al. 1996). However, these studies are not sufficient for concluding that AML1 is essential for the generation of definitive haematopoietic cells (HPC), as the foetal liver is not the site where definitive haematopoietic activity is first detected.

In recent years, a mesodermal component of the para-aortic splanchnopleura (P-Sp) of day 7–8.5 (E7-8.5) mouse embryos has been identified as a site containing haematopoietic precursors prior to liver colonization (Cumano et al. 1996; Godin et al. 1993). The aorta-gonad-mesonephros (AGM) region that develops from the P-Sp at E10 contains multipotent stem cells with long-term repopulating activity (Medvinsky & Dzierzak 1996). Histological analyses of the AGM region have revealed clusters of undifferentiated HPC that express CD34 and c-Kit on the ventral luminal wall of the dorsal aorta in E10 mouse embryos (Garcia-Porrero et al. 1995; Wood et al. 1997; Yoshida et al. 1998). These aortic clusters were first described in chick embryos and since then have been noted in several other species, including mouse and human (Dieterlen-Lièvre 1975; Tavian et al. 1996, 1999). By using a cell-tracing technique, Jaffredo et al. (1998) recently demonstrated that haematopoietic cell clusters in the dorsal aorta of chick embryos are directly derived from the aortic endothelium. More direct experiments to demonstrate the haematogenic activity of endothelial cells (EC) have been performed by Nishikawa et al. (1998) who took advantage of the fact that EC expresses a unique type of cell adhesion molecule, VE-cadherin (Breier et al. 1996). They isolated cells displaying VE-cadherin but not HPC markers (VE-cadherin+ CD45 Ter119 EC) from E9.5–10.5 mouse embryos by fluorescence activated cell sorting (FACS) and showed that these cells give rise to HPC including lymphocytes in in vitro culture, firmly establishing the presence of haematogenic EC in the mouse embryo (Nishikawa et al. 1998).

The purpose of this study was to determine the precise stage affected in AML1-deficient mice. Processes likely to be disrupted included formation of EC from mesoderm, generation of haematopoietic stem cells from EC, migration of haematopoietic stem cells from the AGM region to foetal liver, and self-renewal of haematopoietic stem cells. While this work was in progress, North et al. (1999) reported that AML1 is expressed in the haematopoietic cell clusters and endothelium in the ventral wall of the dorsal aorta, and that AML1-deficient embryos are devoid of haematopoietic cell clusters in the vitelline artery. Although these histological observations suggested the requirement of AML1 for formation of haematopoietic cell clusters, there was no evidence to indicate whether the gene is required for the generation of HPC from EC. Mukouyama et al. (2000) analysed cultures of cells from the P-Sp/AGM region prepared from AML1–/– embryos and reported that no haematopoietic activity could be detected. However, the precise pathway affected in AML1–/– mice could not be identified as the P-Sp/AGM region used in their culture system contains several different cell types, including EC and mesenchymal cells.

In this study we examined intact dorsal aorta, omphalomesenteric and umbilical arteries, and yolk sac tissues of AML1–/– mice, and found a complete lack of haematopoietic cell clusters in these regions. Furthermore, we showed that EC (VE-cadherin+CD45 Ter119) fractions sorted from either the yolk sac or the caudal half of embryo proper of AML1–/– mice could not give rise to HPC in OP9 stromal culture, while AML1–/– EC colonies were able to form under similar conditions. These results clearly demonstrate that in the AML1–/– mouse, HPC cannot be generated from EC, and strongly suggest that the emergence of HPC from EC represents a definitive haematopoiesis.


Absence of c-Kit+ haematopoietic cell clusters in the dorsal aorta, omphalomesenteric artery, umbilical artery of AML1–/– embryos

The haematopoietic cell clusters associated with the endothelium of dorsal aorta within the AGM region are positive for CD34 and c-Kit antigens (Wood et al. 1997; Yoshida et al. 1998), suggesting that these intra-aortic clusters are definitive HPC (Dzierzak et al. 1998). To investigate whether such clusters are present in the dorsal aorta of AML1–/– mice, E10.5 embryos were stained with anti-c-Kit antibody and cleared with a BABB solution to increase the transparency of tissues (see Experimental procedures). This procedure allows us to visualize all c-Kit+ cells present within an intact dorsal aorta. At E10.5, AML1–/– mice were indistinguishable from AML1+/+ mice by gross examination. On the ventral side of the dorsal aorta, however, c-Kit+ clusters were seen in AML1+/+ mice (Fig. 1A,C), but completely absent from AML1–/– mice (Fig. 1B,D).

Figure 1.

Absence of haematopoietic cell clusters in the dorsal aorta, omphalomesenteric artery, and umbilical artery of an AML1−/– embryo (E10.5). Whole-mount immunostaining with anti-c-Kit antibody and subsequent transfer into BABB (benzyl alcohol and benzyl benzoate) solution revealed the haematopoietic cell clusters present within the dorsal aorta, omphalomesenteric artery and umbilical artery. (A–D) The side view of the embryo proper. Arrowheads indicate the ventral wall of the dorsal aorta. In an AML1+/+ embryo, many c-Kit+ cells are attached to the ventral wall of the dorsal aorta (A), whereas no c-Kit+ cells are present in the corresponding region of the AML1−/– embryo (B). (C, D) Higher magnifications of (A, B), respectively. Note that the budding-like, c-Kit+ haematopoietic cell clusters seen in the AML1+/+ embryo (C) are absent in the AML1−/– embryo (D). Similar results were obtained in the omphalomesenteric (E, F) and umbilical (G, H) arteries. The size of haematopoietic cell clusters in the omphalomesenteric and umbilical arteries is larger than those seen in the dorsal aorta. NT, neural tube. Scale bars: A, B = 40 µm; C, D = 10 µm; E, F = 20 µm; G, H = 15 µm.

c-Kit+ clusters are also present in both the omphalomesenteric and umbilical arteries at this stage in wild-type embryos (Yoshida et al. 1998). We confirmed that a large number of c-Kit+ clusters, larger than those in the dorsal aorta, were present in both the omphalomesenteric and umbilical arteries of AML1+/+ embryos (Fig. 1E,G). Such clusters could not be detected within the corresponding arteries of AML1–/– mice (Fig. 1F,H). In some AML1–/– embryos, however, a small number of c-Kit+ cells attached to the external wall of the omphalomesenteric artery were observed (data not shown), but the exact nature of these cells requires further characterization.

The possibility exists that the expression of c-Kit may be regulated by AML1 and thus the absence of AML1 hinders the detection of haematopoietic cell clusters in the AML1–/– embryos. However, this appears unlikely since primordial germ cells, which express c-Kit (Bernex et al. 1996), could be detected with anti-c-Kit antibody in AML1–/– embryo (data not shown). Furthermore, the absence of haematopoietic clusters in the dorsal aorta of AML1−/– embryos was also confirmed histologically by serial sectioning (data not shown).

AML1–/– EC does not differentiate into HPC on OP9 stromal cells

The histological results shown in the previous section appear to indicate that the transition from EC to HPC is blocked in AML1−/– mice. To verify this possibility, we investigated the haematopoietic potential of EC purified from AML1−/– embryos.

Since VE-cadherin expression localizes exclusively in EC (Breier et al. 1996; Nishikawa et al. 1998) and sorted VE-cadherin+ cells co-express CD31, Flk1 and CD34, and take up acetylated low-density lipoproteins (LDL), we previously defined VE-cadherin+CD45 Ter119 cells as EC and showed that this fraction, sorted from either the yolk sac or the embryo proper (the caudal half of the embryo below the level of the heart), gave rise to HPC when it was cultured on an OP9 stromal cell layer (Nishikawa et al. 1998; Ogawa et al. 1999). Therefore, this in vitro assay system appears suitable for studying the haematogenic activity of EC derived from AML1–/– embryos.

As shown in Fig. 2B, the fraction of VE-cadherin+ cells from dissociated AML1–/– embryos was similar to that obtained from AML1+/+ embryos. VE-cadherin+CD45Ter119 cells sorted from the yolk sac and the embryo proper of E10.5 derived from AML1+/+, AML1+/– and AML1–/– mice were cultured on OP9 stromal cells in the presence of SCF, Epo, G-CSF and IL-3 (Fig. 2A). EC sorted from both yolk sac and embryo proper of AML1+/+ embryos gave rise to HPC under such conditions (Fig. 3A,C, respectively), whereas no HPC were generated from EC derived from yolk sac or embryo proper of AML1–/– mice (Fig. 3B,D, respectively). The morphology of HPC derived from EC from embryo proper shown in Fig. 3E) indicates the presence of mainly erythroid and myeloid cells. Similar results were obtained by a colony assay in methyl cellulose containing the same mixture of cytokines. We could not detect any colonies in cultures of EC derived from either yolk sac or embryo proper of AML1–/– mice (Fig. 3F). These results indicated that EC derived from AML1–/– embryos lacked the potential to give rise to HPC. An important point to note is that this defect was observed not only in the embryo proper but also the yolk sac, suggesting that the generation of HPC from yolk sac EC also requires AML1. In addition, the number of colonies in AML1+/– EC culture was lower than that of AML1+/+ EC (Fig. 3F), indicating that AML1 functions in a dose-dependent manner.

Figure 2.

In vitro assay system for evaluating haematogenic EC sorted from AML1−/– embryo. (A) Scheme of the in vitro culture system. The yolk sac and the caudal half of an E10.5 embryo were digested to single-cell suspensions by dispase and EDTA. The cells were stained with APC-anti-VE-cadherin, FITC-anti-CD45 and OrGreen-anti-Ter119 MoAbs. Simultaneously, genotyping was done by PCR using the rostral half of the embryo. Cells having the same genotype were pooled and sorted by FACS Vantage. Sorted cells were cultured on an OP9 stromal cell layer in the presence of IL-3, Epo, SCF and G-CSF. (B) Surface expression of VE-cadherin, CD45 and Ter119 in AML1+/+ and AML1−/– embryos. Percentages of cells that fell within the sorting parameters (boxes) are indicated.

Figure 3.

Complete lack of generation of HPC from EC (VE-cadherin+CD45 Ter119) sorted from AML1−/– embryos. (A–D) VE-cadherin+CD45Ter119 cells sorted from E10.5 embryos were cultured on to an OP9 stromal cell layer (shown in Fig. 2). Photographs were taken by phase-contrast microscopy after 4 days of culture. HPC are generated from both yolk sac EC (A) and embryo proper EC (C) of AML1+/+ mice. No haematopoietic cells were detected in the culture containing AML1−/– EC during 7 days of culture (B, D). (E) Erythroid and myeloid lineage cells are observed in the culture of AML1+/+ cells. (F) Frequency of haematopoietic colony-forming cells in the EC (VE-cadherin+CD45Ter119) fraction sorted from E10.5 embryos. Error bars indicate standard deviations for at least two independent determinations. Scale bars: A–D = 50 µm; E = 10 µm.

It has been estimated that approximately 10% of the EC fraction sorted from the caudal half of embryos as VE-cadherin+CD45Ter119 is haematogenic using colony assay in type I collagen gel (Nishikawa et al. 1998) and that the majority of these sorted cells retain EC character, that is, the ability to form EC colonies on a stroma cell layer (Hirashima et al. 1999; Ogawa et al. 1999). Two morphologically distinct types of EC colonies were observed: one type presents as sheet-like colonies formed by EC that strongly adhere to each other, whereas the other type consists of EC that have migrated into the OP9 cell layer forming cord-like colonies. The difference in physiological activities between the two, however, remained obscure. To evaluate the properties of EC derived from AML1–/– mice, cells were stained with anti-VE-cadherin or anti-PECAM-1 antibody after 7 days of culture. No significant differences between AML1–/– and control EC were detectable in size and morphology of the colonies (Fig. 4). The number of AML1–/– EC colonies formed on OP9 cells was comparable and even higher than that of the AML1+/+ EC colonies (Table 1), demonstrating that the ability of AML1–/– EC to survive and proliferate is not impaired compared with AML1+/+ EC in this culture system. The reason for the slightly larger number of EC colonies of AML1–/– compared with that of AML1+/+ is still not clear at present (see Discussion).

Figure 4.

Formation of EC colonies from VE-cadherin+CD45Ter119 cells sorted from E10.5 embryos. VE-cadherin+CD45Ter119 cells were cultured on OP9 stromal layer in the presence of IL-3, Epo, SCF and G-CSF and stained with antibody against VE-cadherin (VECD1) after 7 days of culture. Morphology and frequency of EC colonies are shown. Two types of colonies, cord-like structure (arrowhead) and sheet-like structure (arrow) (Hirashima et al. 1999), were detected in both AML1+/+ and AML1−/– cultures.

The level of AML1 expression in the sorted VE-cadherin+CD45Ter119 fraction derived from AML1+/– mice was approximately 50% of that detectable in the wild-type by RT-PCR. No AML1 expression was detected in EC from AML1/– mouse as expected (Fig. 5). These data are consistent with the interpretation that AML1 functions in haematogenic EC.

Figure 5.

Detection of AML1 mRNA in EC (VE-cadherin+ CD45Ter119) fraction sorted from E10.5 embryos using RT-PCR. mRNA from AML1−/– embryos were used as a negative control.

Lack of c-Kithi clusters in AML1–/– yolk sac

It has been shown that EC sorted from the yolk sac gives rise to HPC in the in vitro culture system (Nishikawa et al. 1998; Ogawa et al. 1999; Fig. 3A). However, the view that HPC are generated from yolk sac EC in vivo still struggles to be widely accepted, because a clear histological picture supporting this in vitro data has been difficult to obtain.

To overcome this difficulty, we took advantage of AML1–/– mice. As shown above, EC from wild-type yolk sacs gave rise to HPC, whereas EC from AML1–/– yolk sacs lacked haematogenic potential (Fig. 3A,B,F), strongly suggesting that the pathway leading from EC to HPC also exists in the yolk sac EC in vivo, but is blocked in the AML1–/– yolk sac. To confirm this, we investigated whether the c-Kit+ haematopoietic cell clusters present in AML1+/+ yolk sacs can be found in the vessels of AML1–/– yolk sacs. In E9.5 yolk sacs, clusters expressing high levels of c-Kit (c-Kithi) were detected in AML1+/+ yolk sacs (Fig. 6A,C) but not in AML1–/– yolk sacs (Fig. 6B,D). Although no c-Kithi clusters were seen in the yolk sac of AML1–/– embryos, c-Kit HPC and those with low levels of c-Kit (c-Kitlo) were present and scattered within the vessels (Fig. 6B,D).

Figure 6.

Lack of c-Kithi clusters in AML1−/– yolk sacE9.5 embryos (A–D) and E8.5 (5–9 somite stages) embryos (E–H) were stained with anti-c-Kit antibody. (A, B) Whole-mount and (C–H) flat mount views of yolk sac. Note that c-Kithi clusters in the yolk sac (A, C) are absent in AML1−/– embryo (B, D). C and D are larger magnifications of A and B, respectively. AML1+/+ yolk sac at the 5 somite (E), 8 somite (F) and 9 somite (G) stage. AML1−/– yolk sac at the 9 somite stage (H). Some c-Kithi cells form a part of the internal surface of arteries (arrows) and some others present as clusters inside the vessel (arrowheads). (H). YS, yolk sac; EP, embryo proper; S, somite. Scale bars: C, D = 30 µm; E, F = 200 µm.

Although these results strongly support that the c-Kithi clusters in the yolk sac have arisen in situ, the possibility remains that these cells may have migrated from other regions such as the dorsal aorta or omphalomesenteric artery via the circulation. To exclude this possibility, we examined embryos before the onset of circulation which takes place at the 9 somite stage (Cumano et al. 1996). In E8.5 (5 somite stage) embryos, c-Kithi clusters were already present within wild-type yolk sac vessels (Fig. 6E, arrowhead), and the number gradually increased with progression of stage (Fig. 6F,G). Some c-Kithi cells on the internal surface of vessels were seen as half-spheres (Fig. 6E–G, arrows). In AML1–/– yolk sacs, c-Kithi cells could not be detected, with most cells displaying a c-Kitlo phenotype (Fig. 6H). It thus appears conclusive that c-Kithi clusters originate from the yolk sac EC in vivo.


Our data indicate that, in the AML1–/– mouse, the emergence of HPC from EC does not occur. These results together with earlier observations which showed that definitive, but not primitive, haematopoiesis is blocked in the AML1–/– mouse strongly suggest that emergence of HPC from EC represents definitive haematopoiesis. Furthermore, the results re-emphasize the notion that generation of a definitive HPC is specific to cell type rather than a specific region of the embryo.

AML1, an essential factor for generating definitive HPC from EC?

Hematopoietic cell clusters attached to the intra-aortic region were reported in several vertebrate species (Dieterlen-Lièvre 1975; Garcia-Porrero et al. 1995; Tavian et al. 1996, 1999). In the chicken aorta, EC marked by DiI-conjugated acetylated LDL has been shown to become haematopoietic cell clusters (CD45+ cells) (Jaffredo et al. 1998). Moreover, in mice, both the cell clusters and the EC layer in the dorsal aorta express CD34 and c-Kit (Wood et al. 1997; Yoshida et al. 1998), markers for haematopoietic stem cells and progenitors. The dorsal aorta is a part of the AGM region where the first definitive HPC including long-term repopulating haematopoietic stem cells (LTR-HSC) originate (Medvinsky & Dzierzak 1996).

Nishikawa et al. (1998) previously reported that EC (VE-cadherin+CD45Ter119) have the potential to give rise to lymphohaematopoietic cells in the in vitro assay system. By using this in vitro assay system, in this study we showed that EC purified from the embryo proper of AML1–/– mice lack haematogenic potential. Disruption of AML1 also abolished formation of haematopoietic cell clusters in the dorsal aorta, omphalomesenteric artery, and umbilical artery, suggesting strongly that HPC generated from haematogenic EC in the in vitro assay system correspond to the budding cells seen in the vessels. Unlike earlier histological observations by sectioning (North et al. 1999), we checked for the presence or absence of haematopoietic cell clusters by whole-mount staining with anti-c-Kit antibody which allowed a more thorough examination. Our results strongly support the idea that haematopoietic cell clusters generated from aortic EC are definitive HPC (Dzierzak et al. 1998).

AML1 was shown to be expressed in some EC of the ventral wall of the dorsal aorta (North et al. 1999) and our RT-PCR analyses also indicated the presence of AML1 transcripts in the sorted VE-cadherin+CD45 Ter119 EC. These observations support a role for AML1 function at the EC stage and suggest that a low incidence of generation of HPC from EC reflects the low incidence of AML1 expression in EC. Consistent with such a presumed role of AML1, we detected a decreased number of HPC colonies in AML1+/– EC culture, indicating that half the amount of AML1 in EC influences EC-HPC transition. The Runt-domain family transcription factors are known to show a gene dosage effect at the stage where these factors actually play a critical role: Drosophila runt in sex determination, RUNX2 in bone formation, and RUNX1 in leukaemia (Duffy & Gergen 1991; Otto et al. 1997; Song et al. 1999). The gene dosage effect observed in this study may support the interpretation that AML1 is required at the exact stage of EC-HPC transition. The absence of haematogenic activity in AML1–/– EC suggests either that EC requires AML1 to acquire haematogenic potential, or haematogenic EC requires AML1 for transition to the HPC lineage. We are still unable to distinguish between these two possibilities. North et al. (1999) demonstrated that cells expressing a lacZ reporter gene inserted into the AML1 allele are still present in the dorsal aorta of AML1-deficient embryos. This suggests that the developmental processes at the EC stage can proceed in the absence of AML1 function. Therefore, the role of AML1 might be at the step of transition of haematogenic EC to HPC, but further studies are required to substantiate this point.

It might be argued that our results could be attributed to a possible defect of EC in proliferation or survival in the in vitro culture system. However, our conclusion that AML1–/– EC lack haematogenic activity and are not defective in either proliferation or survival was based on the following rationale. The ratio of EC fraction (VE-cadherin+CD45Ter119) sorted from AML1–/– mice is indistinguishable from that of AML+/+ mice (Fig. 2B). In addition, no significant differences were detectable between AML1–/– and AML1+/+ in the size and morphology of EC colonies in the in vitro assay. The number of EC colonies from AML1–/– mice was somewhat higher, not lower, than that from AML1+/+ (Fig. 4, Table 1). It has been estimated that approximately 5–10% of sorted EC from the embryo proper is haematogenic (Nishikawa et al. 1998; Ogawa et al. 1999). It would be interesting to determine whether EC gives rise to HPC by asymmetric cell division or whether EC itself is transformed into HPC. If the latter is the case, about 5–10% of EC in cell sorting preparations would transform into nonadherent, floating HPC and hence the number of EC would be correspondingly decreased. If the former is true, the number of EC should not change even after generation of HPC. As we observed that a significantly larger number of HPC were generated from AML1–/– EC than from AML1+/+ EC in this system, our data seem to support the latter pathway. However, a further careful analysis of the process of transition from EC to HPC in our in vitro system is necessary and it is still too premature to draw conclusions one way or the other on the properties of AML1+/+and AML1–/– EC from the number of EC colonies formed.

Analyses of AML1–/– mice revealed that they haemorrhage mainly in the central nervous system at E11.5–12.5 (Okada et al. 1998; Okuda et al. 1996; Wang et al. 1996), suggesting some defects in their EC. To resolve the cause of the haemorrhage, it would also be helpful to examine other types of mutant mice that have abnormalities in vessel organization. The first step of vessel formation, termed vasculogenesis, involves the association of EC precursors (angioblasts) to form primitive tubes. During subsequent angiogenesis, these tubes undergo remodeling to form a mature network of capillaries (Risau 1997). Mice lacking FLK1, angiopoietin-1 or TIE2/TEK have defects in vasculogenesis or an early stage of angiogenesis and display embryonic lethality at E8.5–10.5 (Gale & Yancopoulos 1999). In contrast, blood vessel abnormalities are not apparent at such early stages in AML1–/– mice, indicating that AML1–/– EC is likely to be normal at E8.5–10.5 when the clusters can be seen. The haemorrhage observed in AML1–/– mice may be due to a defect in a late stage of angiogenesis. Observations by Takakura et al. (2000) are consistent with this notion. They suggest that abnormal vascular network formation in AML1–/– embryos is not due to abnormality in EC but rather to a lack of soluble factors influencing angiogenesis (Takakura et al. 2000). These results support the notion that EC derived from AML1–/– embryos (E10.5) are not intrinsically defective.

EC-HPC transition also exists in the yolk sac

Although it has been shown that EC sorted from the yolk sac gives rise to HPC in vitro (Nishikawa et al. 1998 and this study), it still remains to be elucidated whether this pathway, yolk sac EC to HPC, also exists in vivo. Histological analysis demonstrated the almost concurrent emergence of EC and HPC in the yolk sac blood islands (Moore & Metcalf 1970). Furthermore, a clear picture showing that HPC is generated from yolk sac EC has been difficult to obtain, since the vessels of the yolk sac are smaller and structurally more complicated than the dorsal aorta.

One way to overcome this difficulty would be to specifically mark HPC derived from EC. By introducing the lacZ gene into the AML1 gene locus, North et al. (1999) showed that a fraction of EC and HPC were stained with X-gal, and that in AML1-deficient mice, no X-gal+ HPC were seen within the vessels of the yolk sac nor within the vitelline artery. Based on this histological analysis, they speculated that at least some HPC in the yolk sac, like intra-aortic haematopoietic cell clusters, might be generated from EC. However, it still could not be proven that the X-gal+ HPC detected in the yolk sac were truly derived from EC.

Our studies have demonstrated that the haematopoietic cell clusters which highly express c-Kit are present in the yolk sac blood vessels (Yoshida et al. 1998). c-Kit is a marker for haematopoietic progenitor cells, and it was reported that the c-Kithi cell population sorted from E9.5 yolk sacs is highly enriched for burst forming unit-erythroid (BFU-E) and colony forming unit-granulocyte/macrophage (CFU-GM) (Ogawa et al. 1993). Therefore, those c-Kithi clusters are thought to contain immature haematopoietic progenitor cells. In this study, we showed that these c-Kithi haematopoietic cell clusters in the yolk sac are detectable at embryonic stages prior to the establishment of circulation, indicating that these clusters are generated in situ rather than as a result of migration from other regions. Most importantly, these c-Kithi clusters are not detectable in the yolk sacs of AML1–/– embryos at any stage. As described above, we also showed that EC sorted from yolk sacs of AML1–/– embryos lacked the potential to give rise to HPC when they were cultured on a stromal cell layer. These observations strongly suggest that HPC are generated from EC in the yolk sac as well as in the AGM region in vivo, and that AML1 is essential for this process.

It is still unclear whether or not HPC derived from EC in the yolk sac contributes to the definitive HPC lineage in vivo. Studies by Medvinsky & Dzierzak 1996) and Cumano et al. (1996) strongly argue that the origin of definitive HPC is restricted to the intraembryonic region. In contrast, Yoder et al. (1997) showed that CD34+ cells from the yolk sac can differentiate into lymphocytes when these are transplanted to neonatal mice. Furthermore, our previous report demonstrated that EC sorted from a E9.5 yolk sac gave rise to HPC, including T and B lymphocytes, in an in vitro culture (Nishikawa et al. 1998). These findings, together with the results obtained in this study, suggest that EC in the yolk sac have the potential to give rise to definitive HPC, and that it is the microenvironment in the yolk sac that is defective or deleterious to the maintenance of the sustained proliferation of progenitors of definitive HPC lineage. We propose that the main precursor for the definitive HPC lineage is EC, irrespective of whether EC is derived from the embryo proper or the yolk sac.

Experimental procedures


AML1+/– mutant mice were established by Okada et al. (1998) and maintained in a C57BL/6 background. C57BL/6 mice were purchased from Shimizu Co. Ltd (Kyoto, Japan).

Monoclonal antibodies

Anti-VE-cadherin (VECD1) (Matsuyoshi et al. 1997), anti-Ter119 (Ikuta et al. 1990) and anti-c-Kit (ACK2) (Nishikawa et al. 1991) monoclonal antibodies were prepared and conjugated to either allophycocyanin (APC) or Oregon Green as previously described (Nishikawa et al. 1991; Ogawa et al. 1991). Unconjugated, fluorescein isothiocyanate (FITC) conjugated, or R-phycoerythrin (PE) conjugated anti-CD45 (LCA, Ly-5), anti-CD31(PECAM-1), and anti-CD11b(Mac-1) monoclonal antibodies were purchased from PharMingen (San Diego, CA).


Whole-mount immunostaining was performed as previously described (Takakura et al. 1997). Stained yolk sacs were dissected free of the embryo propers and flat-mounted. Dehydration and treatment with BABB, a 1 : 2 mix of benzyl alcohol and benzyl benzoate was used to increase the transparency of tissues (Davis 1993).

Cell preparation and sorting

Dissections were performed as follows: yolk sacs were removed and the remaining embryo was bisected at a level just below the heart. The caudal half was used as the embryo proper. The rostral halves used for genotyping were digested at 55 °C for 30 min in 50 µL of lysis buffer containing proteinase K (Nacalai Tesque, Kyoto, Japan) and then heated at 95 °C for 5 min. One microlitre was used for polymerase chain reaction (PCR) amplification. Primers for AML1 detection have been described previously (Okada et al. 1998). PCR was performed for 35 cycles (94 °C for 1 second, 60 °C for 5 s and 72 °C for 10 s) using Z-Taq polymerase (Takara Shuzo, Osaka, Japan).

Dispase (Gibco BRL, Grand Island, NY) and Cell Dissociation Buffer (Gibco BRL) were used to prepare the cells from yolk sacs and embryo propers as described (Nishikawa et al. 1998). Dissociated cells from embryos with similar genotypes were pooled. Procedures for surface staining and cell sorting were as previously described (Nishikawa et al. 1998).

Culture with OP9 stromal cells

For generating HPC from EC, 1 × 103 sorted EC were cultured per well in a six-well plate for 7 days on OP9 stromal cell layer in αMEM medium (Gibco BRL) containing 10% foetal calf serum and 5 × 10−5 m 2-mercaptoethanol, and supplemented with 100 ng/mL of stem cell factor (SCF), 200 U/mL of interleukin-3 (IL-3), 100 ng/mL of granulocyte-colony stimulating factor (G-CSF), and 2 U/mL of erythropoietin (EPO). Recombinant mouse IL-3 and SCF were prepared as described (Ogawa et al. 1991). The recombinant human G-CSF and EPO were kindly provided by Sankyo Co. Ltd and the Kirin Brewery Co. Ltd, respectively. Haematopoietic cells were harvested carefully to avoid damage to EC colonies at the bottom and analysed for expression of Ter119 or Mac-1 using flow cytometry in addition to morphological analysis with May–Grünwald Giemsa staining.

To detect the formation of EC colonies on the OP9 cell layer, cultured cells were fixed with 4% paraformaldehyde and stained with anti-VE-cadherin or anti-PECAM-1 monoclonal antibodies. The subsequent colorization step with alkaline phosphatase-conjugated anti-rat IgG(H+L) was performed as previously described (Hirashima et al. 1999).

Colony formation of haematopoietic cells

3 × 103 sorted EC were plated in triplicate in 1.0% methylcellulose with αMEM medium containing 30% FCS, 10−4 mβ-ME, 2 mm l-glutamine, and 1% bovine serum albumin (Stemcell Technology Inc., Vancouver, Canada) and supplemented with 100 ng/mL of SCF, 200 U/mL of IL-3, 100 ng/mL of G-CSF and 2 U/mL of EPO. Hematopoietic cell colonies were scored on day 7 under a microscope.


Total RNA was isolated from sorted cells using Trizol (Gibco BRL). RNA derived from 3 × 103 cells was reverse-transcribed and amplified using an RT-PCR kit (Gibco BRL). Random hexamers were used as primers for cDNA synthesis, and rTaq polymerase (Toyobo, Osaka, Japan) was used for amplification. Primers for AML1 were as previously described (Okuda et al. 1996). Sequences of primers specific to hypoxanthine phosphoribosyl transferase (HPRT) were: 5′-GTTGGATACAGGC CAGACTTTGTTG-3′, 5′-GATTCAACTTGCGCTCATCT TAGGC. The expected size of the amplified fragment was 164 bp.


We thank Dr M. Hirashima for advice regarding immunohistochemistry and Dr Ruth T. Yu for a critical reading of this manuscript.


  1. Communicated by: Kohei Miyazono