Multiple Mesoderm Subsets Give Rise to Endothelial Cells, Whereas Hematopoietic Cells Are Differentiated Only from a Restricted Subset in Embryonic Stem Cell Differentiation Culture



In the developing mouse, vascular endothelial cell (EC) and hematopoietic cell (HPC) lineages are two initial cell lineages that diverge from mesodermal cells, which have been roughly subdivided into three subtypes according to their geographical location: the organizer, embryonic mesoderm in the primitive streak, and extraembryonic mesoderm during gastrulation. Although the initial progenitors that become the two lineages appear in both vascular endothelial growth factor receptor 2+ (VEGFR2+) lateral and extraembryonic mesoderm, little is known about the underlying molecular events that regulate the derivation of ECs and HPCs. Here, we describe an experimental system consisting of two types of embryonic stem cell lines capable of distinguishing between organizer and the middle section of the primitive streak region. Using this system, we were able to establish a defined culture condition that can separately induce distinct types of mesoderm. Although we were able to differentiate ECs from all mesoderm subsets, however, the potential of HPCs was restricted to the VEGFR2+ cells derived from primitive streak-type mesodermal cells. We also show that the culture condition for the progenitors of primitive erythrocytes is separated from that for the progenitors of definitive erythrocytes. These results suggest the dominant role of extrinsic regulation during diversification of mesoderm.

Disclosure of potential conflicts of interest is found at the end of this article.


During the process of mouse gastrulation, mesoderm initially diverges into three subpopulations that can be distinguished by their position along the primitive streak: the organizer, which subsequently gives rise to axial mesoderm; embryonic mesoderm in the primitive streak (streak type), which contributes to paraxial and lateral mesoderm; and extraembryonic mesoderm, which continues to form blood and endothelial cells (ECs) in the yolk sac [1, [2]–3]. Although the majority of these mesoderm subpopulations can preserve their multipotency, their eventual fates are determined at an early developmental stage depending on the spatiotemporal localization of progenitor cells [4, 5].

One of the most extensively studied fate determination processes to date is the divergence of vascular ECs and hematopoietic cells (HPCs), an event that is thought to take place in yolk sac blood islands [6, 7]. These cells initially appear following gastrulation and originate from both the extraembryonic mesoderm and the lateral mesoderm generated from streak-type mesodermal progenitors [8]. Although these two lineages are thought to originate from a common mesoderm progenitor [9, 10], the notion of bipotential hemangioblast for ECs and HPCs has been challenged by recent studies [9, 11, [12]–13].

Unlike Xenopus or zebrafish, the embryonic mouse may not be a convenient experimental system to observe the process by which ECs and HPCs are specified, because of the lack of accessibility and difficulty in obtaining sufficient cell numbers for analysis from the animal. To overcome these problems, we have been exploiting an in vitro embryonic stem (ES) cell differentiation culture. To overcome the problem of ES cell differentiation cultures in which multiple events are occurring simultaneously, we have attempted to establish molecular markers for defining the cell lineages generated in cultures [14, [15]–16]. Using such markers for identifying the point of lineage divergence, we were able to show that vascular endothelial growth factor receptor 2 (VEGFR2; FLK1) and platelet-derived growth factor receptor-α (PDGFRα) are useful surface markers that prospectively distinguish lateral and paraxial mesoderm, respectively [17, 18]. These markers enabled us to further dissect the course of mesoderm differentiation to HPCs, revealing the presence of two distinct pathways for HPC differentiation, one leading directly from VEGFR2+ mesoderm and the other from VEGFR2+ vascular endothelial cell cadherin (VE-cadherin+) ECs derived from VEGFR2+ mesoderm [14]. More recently, we showed that another set of mesoderm cells expressing goosecoid+ (Gsc+), PDGFRα+ (αR+), and E-cadherin (ECD) was generated from Gsc+, αR+, ECD+ mesendoderm and was able to be induced from ES cells under an Activin-containing defined culture condition [19, 20]. Our study previously found that the mesoderm generated under this culture condition was able to give rise to ECs at a low frequency, although the ECs continued to express VEGFR2 [19]. These observations suggest the presence of multiple differentiation pathways to ECs.

The aim of this study is to determine the divergence process between the hematopoietic and endothelial lineages during mesodermal diversification in ES cell culture. To do so, we have developed two ES cell lines that allow us to distinguish three subsets of mesoderm in the primitive streak and have used this system to determine conditions for inducing distinct sets of mesoderm, particularly those fated to ECs and HPCs. The results show that ECs are generated from a broader set of ES cell-derived mesoderm subpopulations, whereas HPCs are derived only through a restricted pathway. Interestingly, our results show that differentiation of each subset of mesoderm is largely determined by extrinsic signals added in ES cell cultures.

Materials and Methods

Constructions and Cell Lines

The green fluorescent protein (GFP) and DsRed T4 cDNAs were inserted into the downstream of the 500-base pair (bp) Brachyury (T500) promoter reported previously [21]. The T500 promoter was amplified by polymerase chain reaction (PCR) using the bacterial artificial chromosome clone containing the mouse Brachyury genome as a template and confirmed by DNA sequencing. The GFP construct was transfected into CCE ES cells with the plasmid carrying the Neo resistance gene by electroporation. Stably transduced ES cells (T500-GFP ES cells) were established as G418-resistant colonies. The ES cell lines (Gsc ESgfp/+cell) bearing a Gsc-GFP knock-in allele were established by homologous recombination as described previously [19].

In some experiments, such as the EC and HPC colony formation assay, we selected and used two independent ES cell lines that are marked with the T500-GFP transgene or the insertion of GFP cDNA into Gsc allele. These ES cell clones are genetically not identical because they were established from different culture plates. In other experiments, we used the same ES cell lines, which express a high level of GFP.

Maintenance of ES and OP9 Cells

T500-GFP ES cells, Gsc ESgfp/+cells, and OP9 stromal cells were maintained as described previously [14, 15, 22].

Hematopoietic and Endothelial Colony Formation

The various VEGFR2+ cells purified by fluorescence-activated cell sorting (FACS) were recultured on an OP9 cell layer. Three days later, cells were gently harvested and analyzed for the hematopoietic colony-forming activity as described previously [15, 18, 23]. Endothelial colony formation was performed as described previously [18, 23]. All experiments were performed in triplicate and were done twice, with two independent ES cell clones. The statistical significance was calculated by the Student t test.

Reverse Transcription-PCR

Reverse transcription (RT)-PCR analysis was performed as described previously [18].

In Vitro ES Cell Differentiation

Under the serum-containing condition, induction of ES cell differentiation was performed in either α-minimal essential medium (αMEM) or SF-O3 supplemented with 10% fetal calf serum (FCS) and 50 μM 2-mercaptoethanol (2ME) as described previously [18, 19].

For induction under serum-free condition, T500-GFP ES cells and Gsc ESgfp/+ cells were seeded onto type IV collagen-coated dishes at densities of 1.2 × 106 and 1 × 105 cells per 10-cm dish, respectively, in SF-O3 medium (Sanko Junyaku Co, Tokyo, supplemented with 0.1% bovine serum albumin and 50 μM 2ME. Recombinant human Activin A, bone morphogenetic protein 4 (BMP4), Wnt3a, and fibroblast growth factor 2 (FGF2) (R&D Systems Inc., Minneapolis, were added at the start of the cultures at various concentrations. In some experiments, retinoic acid (RA; 10−7 M) was added to the ES cell differentiation culture from day 2 to day 5 of induction of ES cell differentiation. For erythropoiesis, 2.0 × 105 of either T500+VEGFR2+, T500VEGFR2+, or Gsc+VEGFR2+ECD cells purified by FACS were recultured on a confluent OP9 cell layer in one well of a six-well plate with the condition containing αMEM, 10% FCS, 50 μM 2ME, 2 U/ml Epo (R&D Systems) and 200 U/ml interleukin-3 (R&D Systems). On day 8 (for primitive erythrocytes) and day 15 (for definitive erythrocytes), the cultured cells were harvested and collected by gentle pipetting to investigate cell morphology with May-Giemsa staining.

For induction of VE-cadherin+ ECs and CD45+ hematopoietic cells, 2 × 105 FACS-purified cells were recultured in one well of six-well type IV collagen-coated plates with αMEM containing 10% FCS, 50 μM 2ME, 3 ng/ml vascular endothelial growth factor, and 100 ng/ml stem cell factor (Kit ligands) [14, 18]. Three days later, the cells were harvested and collected for the examination of expression of the surface markers VE-cadherin and CD45.

Antibodies and Cell Staining for FACS

Rat monoclonal antibodies (MoAbs) APA5 (anti-PDGFRα), ECCD2 (anti-E-cadherin), AVAS12 (anti-VEGFR2), APB5 (anti-PDGFRβ), and VECD1 (anti-VE-cadherin) were prepared as reported previously [14, 18, 19]. Anti-Cadherin 11 MoAb was kindly provided by Dr. A. Kudo (Tokyo Institute of Technology, Yokohama, Japan). APA5, APB5, anti-Cad11 antibody, and VECD1 were biotinylated by standard methods as described previously [18]. Phycoerythrin (PE)-conjugated anti-CD45 rat MoAbs was purchased from BD Pharmingen (San Diego, PE-conjugated streptavidin (BD Pharmingen) or allophycocyanin (APC)-conjugated streptavidin (Molecular Probes, Eugene, OR, was used for detecting biotinylated antibodies. For some experiments, ECCD2 and AVAS12 MoAbs were directly conjugated with APC by the standard method. Cultured cells were harvested and stained with antibodies as described previously [14, 18].

Immunohistochemical and In Situ Hybridization Analyses

These procedures are described in the supplemental online Materials and Methods.

Transgenic Mice

Transgenic mice carrying GFP or DsRed T4 cDNAs driven by the 500-bp Brachyury promoter were generated as described in a previous report [24].


500-bp 5′ Flanking Region of the Brachyury Gene Drives GFP Genes in the Middle Part of the Primitive Streak

Mesodermal cells can be roughly divided into three subsets based on their origin: those derived from the organizer region (OR), the primitive streak outside the OR (PR), and the extraembryonic region [3, 19]. Brachyury (T) is the most popular marker for marking mesodermal cells in the primitive streak [25]. However, as T is expressed in all mesoderm cells [25, 26], expression of T by itself is not sufficient to distinguish OR from PR. On the other hand, previous studies had demonstrated that the 500-bp (T500) Brachyury promoter that was used in this study was active in PR but not OR, raising the possibility that it could serve as an appropriate marker to distinguish between the two mesoderm subsets [21]. To confirm this result, we generated transgenic mice carrying enhanced green fluorescent protein (eGFP) cDNA or DsREDT4, both of which are driven by the T500 promoter. As expected, we were able to detect fluorescence activity in the middle part of the primitive streak (Fig. 1Aa, 1Ab), but not in OR or in the extraembryonic region. Among the primitive streak cells, we found the highest GFP expression in the mid-streak. Following gastrulation, GFP expression gradually spread over the somites and presomitic mesoderm, although we found that the expression level gradually diminished (Fig. 1Ac–1Af). As this expression pattern suggested that T500 is useful in distinguishing PR from other mesodermal cells, we introduced the same constructs into ES cells (T500-GFP ES cell).

Figure Figure 1..

Establishment of T500-GFP ES cell lines. (A): Activity of the 500-base pair (bp) Brachyury promoter in transgenic mice. Transgenic mice bearing either DsReds or GFP cDNA driven by the 500-bp Brachyury promoter were established as described previously [24]. The expression of DsReds was observed in a posterior part of the embryo, particularly in the primitive streak at embryonic day (E) 7.5 (in bright field [Aa] and under fluorescence microscopy [Ab]). The area expressing GFP spread over somites and presomitic mesoderm at E8.5 (in bright fields [Ac, Ae] and under fluorescence microscopy [Ad, Af]). (B): GFP expression in transgenic ES cell lines carrying GFP driven by Brachyury promoter (T500). Peak GFP expression was observed on day 4 under the serum-containing condition, and the expression dramatically decreased on day 5. (C): Suppression of Brachyury expression by RA treatment. Endogenous Brachyury expression (upper panel, reverse transcription-polymerase chain reaction analysis) and transgenic GFP expression (lower panel, fluorescence-activated cell sorting [FACS] analysis) were suppressed by RA treatment during in vitro ES cell differentiation. (D, E): Coexpression of Brachyury and GFP. Almost all FACS-purified GFP+ cells on day 4 (B) expressed endogenous Brachyury transcript ([D], right panel; in situ hybridization for Brachyury) and the Brachyury protein (immunostaining for Brachyury [E]), whereas few positive cells were observed in the GFP population ([D], left panel). Abbreviations: GFP, green fluorescent protein; RA, retinoic acid.

After inducing mesodermal cells from these cell lines in serum-containing medium, the GFP signal initially appeared on day 3, with peak expression occurring on day 4 (Fig. 1B). After day 5, GFP expression decreased, consistent with results from the mouse embryo. We were also unable to detect any GFP expression in the culture using RA (Fig. 1C). Evaluating Brachyury mRNA by RT-PCR revealed that the decrease in the GFP+ population by RA treatment corresponded with the reduction of endogenous Brachyury mRNA (Fig. 1C). Subsequent in situ hybridization and immunostaining of Brachyury showed that almost all of the GFP+ cells purified by FACS expressed endogenous Brachyury, with a few GFP cells also exhibiting the signal (Fig. 1D, 1E). We were able to validate the specificity of both probe and antibody by whole mount in situ hybridization and immunostaining, respectively (supplemental online Fig. 1). On the basis of these results, we conclude that these transgenic ES cell lines enable us to monitor differentiation of Brachyury+ cells that correspond to PR type, which are identifiable by eGFP expression.

Search for the Defined Condition to Induce T500-GFP+ Cells

We next attempted to determine defined conditions for inducing T500-GFP+ cells. For this purpose, we used the monolayer culture on collagen IV rather than embryoid body culture, and we tested different concentrations of molecules that have been shown to affect mesoderm differentiation, such as BMP4, Activin, BMP2, FGF2, and Wnt3a. T500-GFP+ cells were detected only in the culture with BMP4. The optimal concentration of BMP4 for inducing GFP was 1 ng/ml, and the effective range of concentration appeared to be narrow (Fig. 2A; data not shown). It should be emphasized, however, that this condition is less efficient than the serum-containing medium in supporting a T500high population as well as in inducing T500+ cells (Figs. 1B, 2A). This result suggests a requirement of additional unknown factors for the generation of T500-GFPhigh cells.

Figure Figure 2..

Induction of streak-type Brachyury expression by BMP4. All cultures of T500-GFP embryonic stem (ES) cells were performed on type IV collagen-coated dishes under serum-free condition with various factors. The expressions of GFP and other markers were examined on day 4-differentiation day. (A): GFP induction under serum-free condition with various concentrations of BMP4. A small amount of BMP4 (1 ng/ml) induced T500-GFP+ cells. It should be noted that the intensity of GFP expression under the serum-free condition with BMP4 is lower than that under the serum-containing condition. GFP+ cells were undetected in the culture under serum-free condition without BMP4. (B): Addition of Activin A inhibits generation of T500-GFP+ cells. A large amount (10 ng/ml) of Activin A could inhibit the induction of T500-GFP expression even in the presence of 1 ng/ml BMP4. (C): Effects of the order of Activin A addition on the generation of T500-GFP+ cells. T500-GFP ES cell differentiation was induced under serum-free condition with 1 ng/ml BMP4. From day 1, day 2, and day 3 to day 4, 10 ng/ml Activin A was added into the cultures. The addition of Activin A on the initial two days, day 1 and day 2, could suppress the induction of T500-GFP+ cells. (D): Expression pattern of PDGFRα (αR) and VEGFR2 (VR2) in T500-GFP+ population. Expression pattern of αR and VR2 in T500-GFP+ population derived from serum-containing culture ([D], left panel) and serum-free culture with 1 ng/ml BMP4 ([D], center panel). Half of the T500-GFP+ cells expressed both αR and VR2. The induction of T500+αR+VR2+ and T500+αR+VR2 cells was inhibited by the addition of 10 ng/ml Activin A even in the presence of BMP4 ([D], right panel). Abbreviations: BMP, bone morphogenetic protein; GFP, green fluorescent protein; PDGFR, platelet-derived growth factor receptor; VEGFR, vascular endothelial growth factor receptor.

To further investigate the effects of BMP4 on PR differentiation, we next examined whether the combination of extrinsic factors actually enhanced the generation of T500-GFP+ cells. Combining BMP4 with Wnt3a or FGF2 did not show any synergic effect on the induction of T500-GFP+ cells (data not shown). Moreover, consistent with previous studies in Xenopus [27, 28], Activin antagonizes the effect of BMP, as the generation of T500-GFP+ cells in a culture with 1 ng/ml BMP4 was suppressed threefold by the addition of 10 ng/ml Activin (Fig. 2B). To detect this inhibitory effect, Activin needs to be added into the culture no later than day 1 of the culture (Fig. 2C).

Expression of VEGFR2 and PDGFRα in T500+ Cells

Our previous study demonstrated that the differentiation of ES to mesoderm could be monitored by the expression of three markers, ECD, PDGFRα (αR), and VEGFR2 (VR2) [14, 17, 18]. More recently, we showed that paraxial and lateral mesoderm cells could be defined as αR+VR2ECDlow and αRVR2+ECD, respectively, which were differentiated from the αR+VR2+ECD common precursor in an in vitro ES cell culture [18]. In an attempt to further characterize the T500-GFP+ cells corresponding to PR-type mesoderm, we investigated the expression of these mesodermal markers in T500-GFP+ cells derived from either a serum-containing culture (SCC), a serum-free culture (SFC) with 1 ng/ml BMP4, or an SFC with both 1 ng/ml BMP4 and 10 ng/ml Activin (Fig. 2D). The T500-GFP+ population consists of four distinctive fractions, αR+VR2+, αR+VR2, αRVR2+, and αRVR2. All four fractions were detected in the cultures with serum or BMP4. Although the proportion of αR+VR2 cells in SFC with BMP4 was lower than that found in SCC, the proportions of the other populations in SFC with BMP4 were equivalent to those in SCC (Fig. 2D, left and center panels).

We next investigated which subfraction of T500-GFP+ cells is affected by the addition of Activin. As shown in Figure 2D, Activin suppresses the generation of the αR+VR2+ and αR+VR2 populations more preferentially than the αRVR2+ population. This suggests that Activin suppresses the differentiation of T500+ cells, which have the potential to give rise to paraxial mesoderm [18]. It should be noted that in the SFC with 3 ng/ml BMP4, addition of Activin enhances the generation of αR+VR2+ and αRVR2+ mesodermal cells in the day 4 whole culture (supplemental online Fig. 2A).

BMP4 Antagonizes the Effect of Activin on Induction of OR-Derived Mesoderm

We showed that a relatively high dose of Activin selectively induced goosecoid+ (Gsc+) αR+ECD+ mesendoderm, eventually giving rise to both Gsc+αRECD+ definitive endoderm and Gsc+αR+ECD mesoderm [19, 20]. Under these conditions, we found that virtually all cultured cells had become Gsc+ cells by day 6 (supplemental online Fig. 2B). In contrast to the biphasic dose response of T500-GFP ES cells to BMP4, their response to various concentrations of Activin revealed a simple upward curve, suggesting that Activin enhances Gsc+ cell generation in a dose-dependent manner (Fig. 3A). Interestingly, FGF2 showed a marked enhancing effect on Activin-induced generation of Gsc+ cells, particularly on that of Gsc+ECD population (Fig. 3B). On the other hand, BMP4 alone had no activity to induce Gsc+ cells (data not shown). Moreover, BMP4 suppressed the Activin-induced generation of Gsc+ cells in a dose-dependent manner (Fig. 3C). This inhibitory effect of BMP4 decreased when BMP4 was added at the later point during in vitro ES cell differentiation (Fig. 3D). Taken together, the direction of ES cell differentiation in SFC, at least in the monolayer culture, is strictly related to the molecular cue present in the culture.

Figure Figure 3..

Induction of organizer region (OR)-type mesoderm by Activin A under serum-free conditions. (A): Dose-dependent induction of goosecoid (Gsc)-GFP+ cells corresponding to OR by Activin A. Gsc embryonic stem (ES)gfp/+ cell differentiation was induced under serum-free conditions with various concentrations of Activin A, and GFP expression was examined on day 4. The generation was enhanced in a dose-dependent manner. (B): Fibroblast growth factor 2 (FGF2) synergistically enhances the generation of mesendoderm and mesendoderm-derived mesoderm induced by Activin. The Gsc+E-cadherin (ECD)+ mesendoderm population was increased by adding FGF2 into the serum-free culture with a high dose of Activin. FGF2 also stimulated the generation of the mesendoderm-derived mesoderm progenitors that were defined as Gsc+ECD [19, 20]. (C): Inhibition of Gsc induction by BMP4. BMP4 suppressed the generation of Gsc+ cells in a dose-dependent manner. GFP expression was examined at day 4 of differentiation culture. (D): Effects of the BMP4 added at different times on the generation of Gsc-GFP+ cells. Gsc ESgfp/+ cell differentiation was induced under serum-free condition with 10 ng/ml Activin A. At day 1, 2, 3, 4, 5, or 6, 10 ng/ml BMP4 was added into the culture, and GFP expression was examined on day 6. Inhibition of Gsc-GFP+ cell generation was observed except for BMP addition later than day 5. The inhibitory effect was reduced gradually as addition timing was delayed. Abbreviations: BMP, bone morphogenetic protein; GFP, green fluorescent protein.

VEGFR2 Expression in the Goosecoid+ Population

Our previous study showed that VR2+ mesoderm cells were generated from mesendodermal cells defined as day 4 Gsc+αR+ECD+ cells [19]. When we induced Gsc+ cells by incubating Gsc ESgfp/+ cells for 6 days with 10 ng/ml Activin, we found that half of the Gsc+ECD mesoderm derived from mesendoderm expressed VR2 (Fig. 4A). Gsc+ECD cells, but not Gsc+ECD+ cells, also expressed other mesodermal markers, such as PDGFRβ and Cadherin11.

Figure Figure 4..

Expression of mesoderm markers in the serum-free culture with a high dose of Activin or BMP4. (A): Expression of a set of mesodermal markers in day 6 Gsc+ cells derived from the serum-free culture with Activin A. Upper panels show marker expression in the day 6 Gsc+ECD+ population, and lower panels show the expression pattern of mesoderm markers in day 6 Gsc+ECD population. A set of molecular markers expressed in mesoderm, such as PDGFRβ, VEGFR2 (VR2), and Cad11, was expressed in the day 6 Gsc+ECD population. The horizontal axis indicates the FSC, showing cell size. (B): Suppression of the generation of Gsc+VR2+ECD cells by BMP4. BMP4 suppressed the induction of Gsc+ECD+ cells even in the presence of Activin A (left panel). Although Gsc+ECD cells were induced under serum-free conditions with Activin and BMP4 up to day 6, expression of VR2 in this population dramatically decreased (right panel). (C): VR2 and PDGFRα (αR) expressions of unclassified population induced by a high dose of BMP4. The differentiation of T500-GFP ES cells was induced under a serum-free condition with 10 ng/ml BMP4. Expression pattern of VR2 and αR on day 4 was examined using fluorescence-activated cell sorting. Both T500VR2+ and T500αR+ cells were induced by serum-free culture with a high dose (10 ng/ml) of BMP4. (D, E): Domination of T500αR+ECD cells under serum-free conditions with 10 ng/ml BMP4. Analysis was performed at day 6 of culture. T500VR2+ cells decreased, and T500αR+ cells became the dominant population (D). Almost all of the T500αR+ cells do not express ECD, a marker for both ectoderm and endoderm in early mouse development (E). Abbreviations: Cad, Cadherin; ECD, E-Cadherin; FSC, forward scatter; GFP, green fluorescent protein; Gsc, goosecoid; PDGFR, platelet-derived growth factor receptor; VEGFR, vascular endothelial growth factor receptor.

As BMP4 acts as an antagonizing molecule to Activin in the induction of Gsc+ mesendoderm lineages, we next investigated the effect of BMP4 on the induction of Gsc+VR2+ECD cells. Unlike the positive effect of BMP4 on T500+VR2+ population, BMP4 showed a marked suppressive effect on the differentiation of Gsc+VR2+ECD and Gsc+ECD+ cells than on the differentiation of Gsc+αR+ECD cells, compared with the results shown in Figure 4A (Fig. 4B).

Induction of VEGFR2+ Cells That Are Neither PR nor OR

A high dose of BMP4 blocks the expression of T500-GFP under a serum-free condition (Fig. 2A). One way of interpreting this result is to consider that extraembryonic mesoderm that is T500Gsc is induced under this culture condition. We next measured the expression of αR and VR2 of cultured cells after day 4 and day 6 culture in SFC with 10 ng/ml BMP4 (Fig. 4C, 4D). Although both αR+ and VR2+ cells were induced under this condition, the proportion of VR2+ cells was substantially lower than those in SFC with 1 ng/ml BMP4 (Fig. 4C; supplemental online Fig. 2C). On day 6 of the culture, almost no VR2+ cells were detected, and the culture was occupied by unclassified ECD cells and ECDαR+ cells (Fig. 4D, 4E). These αR+ cells failed to express the E-cadherin that is a marker for ectoderm and endoderm in early mouse development (Fig. 4E), suggesting that they actually belong to a mesoderm cell lineage rather than OR and PR.

Endothelial and Hematopoietic Potential of VEGFR2+ Mesodermal Cells

The above results indicated that different combinations of BMP4 and Activin result in induction of distinct subsets of mesoderm, but all populations contained VR2+ cells. As VR2 was initially considered a marker for cells with hematoendothelial fate [14], we next investigated whether or not each of the VR2+ populations induced by a distinctive condition has the potential to give rise to endothelial cells. VR2+ cells were generated under distinctive culture conditions as described above, purified by FACS, and cultured on OP9 feeder cells to measure the clonogenic progenitors that give rise to ECs (Fig. 5A). Although a small fraction of Gsc+VR2+ECD cells induced by Activin or those of T500VR2+ cells induced by a high dose (10 ng/ml) of BMP4 consistently gave rise to EC colonies, T500+VR2+ cells induced by a low dose (1 ng/ml) of BMP4 contained much higher numbers of EC progenitors. Nonetheless, these results suggest that the potential to give rise to ECs is present in most mesoderm subsets, although the major source of ECs is the T500+ population.

Figure Figure 5..

Endothelial potential of VEGFR2 (VR2)+ cells derived from distinct mesoderm subpopulations. (A): Generation of endothelial cell (EC) colonies from VR2+ mesoderms induced from embryonic stem (ES) cells under distinct culture conditions. Shown is the morphology of ES cell-derived endothelial colonies (left panel, CD31 immunostaining). T500+VR2+ cells derived from serum-free culture generated round endothelial colonies on the OP9 feeder layer. Graph shows the frequency of endothelial colonies in VR2+ cells generated from various mesoderm cells induced under distinct conditions. VR2+ cells were purified by fluorescence-activated cell sorting (FACS) and recultured on the OP9 stromal cell layer for 3 days. The colony number per 1,000 VR2+ cells was calculated after CD31 immunostaining. The endothelial potential of Gsc+VR2+E-cadherin (ECD) cells induced under serum-free condition by Activin A and T500VR2+ cells induced under serum-free condition with 10 ng/ml BMP4 (T500VR2+SerumBMP4high) was far less than those in T500+VR2+ populations induced by serum (T500+VR2+Serum+) or by 1 ng/ml BMP4 (T500+VR2+SerumBMP4low). (B): Schematic illustration of hemogenic EC differentiation in in vitro ES cell culture. ES cells are able to generate VR2+ mesoderm under serum-containing conditions, which can give rise to VE-Cad+ hemogenic ECs, some of which can differentiate into both CD45+ definitive HPCs and ECs. (C): Potential to generate VE-Cad+ cells from primitive streak outside the organizer region-type or organizer region-type ES cell-derived mesoderms. T500+VR2+ induced under serum-containing culture (SCC), T500+VR2+ induced under serum-free culture (SFC) with 1 ng/ml BMP4, or Gsc+VR2+ECD induced by SFC with 10 ng/ml Activin A were cultured with growth factors including SCF and vascular endothelial growth factor. Three days later, the cells were stained with anti-VE-Cad and CD45 antibodies and analyzed by FACS. Gsc+VR2+ECD could not give rise to CD45+ cells. T500+VR2+ cells induced by SFC with 1 ng/ml BMP4 could generate far fewer CD45+ cells than T500+VR2+ induced by SCC. (D): Hematopoietic potentials of distinct types of VR2+ mesoderm. Gsc+VR2+ECD cells and T500VR2+ cells induced by 10 ng/ml BMP4 were unable to give rise to hematopoietic progenitors that were detectable by the colony assay. The hematopoietic potential of T500+VR2+ cells induced under the serum+ condition was significantly higher than that of T500+VR2+ cells induced under serum condition. Error bars represent SD. The statistical significance was calculated by the Student t test. All experiments were performed twice with each of two independent cell lines, and the representative result is shown here. Abbreviations: BMP, bone morphogenetic protein; ES, embryonic stem; Gsc, goosecoid; HPC, hematopoietic cell; VE-Cad, VE-Cadherin; VEGFR, vascular endothelial growth factor receptor; VR, vascular endothelial growth factor receptor.

To confirm the functional assay of EC progenitors, we next investigated the differentiation of VR2+ cells to VE-Cadherin+ (VE-Cad+) ECs. In parallel, we also analyzed the differentiation of CD45+ HPCs and enucleated erythrocytes. Our previous study demonstrated that VR2+ cells induced under serum-containing condition could differentiate into the hemogenic VE-Cad+ ECs that gave rise to CD45+ definitive HPCs and enucleated erythrocytes (Fig. 5B) [14, 29].

Consistent with the functional assay, only a small fraction of Gsc+VR2+ cells could give rise to VE-Cad+ cells. Furthermore, no CD45+ cells were detected in the same culture stimulated by Activin (Fig. 5C), whereas T500+VR2+ cells could give rise to VE-Cad+ ECs and CD45+ cells. Our results further show that the SFC with BMP4 is less efficient for inducing EC progenitors than SCC and almost unable to induce CD45+ cells (Fig. 5C) or enucleated erythrocytes (supplemental online Fig. 3). According to our previous study, CD45+ cells and enucleated erythrocytes are derived mainly from hemogenic VE-Cad+ cells [14]. Thus, it is likely that SFC containing 1 ng/ml BMP4 can induce VE-Cad+ ECs but is insufficient for inducing hemogenic VE-Cad+ cells. This observation is consistent with our findings that the frequency of HPC colonies in T500+VR2+ cells induced by BMP4 is much lower than those induced under the serum-containing condition (Fig. 5D).

Hematopoietic potential was also analyzed for the Gsc+VR2+ECD cells induced by Activin and VR2+ cells induced by a high dose of BMP4 (Fig. 5D). FACS-purified VR2+ cells were recultured on an OP9 cell layer for 3 days, and the hematopoietic colony-forming activity was examined. In contrast to T500+VR2+ cells induced by BMP4, as well as those induced by serum that give rise to primitive erythrocytes under the same condition, neither Gsc+VR2+ECD cells induced by Activin nor VR2+ cells induced under a high dosage of BMP4 could give rise to primitive erythrocytes (data not shown). Similarly, we could not detect any erythroid colony-forming activity in these two VR2+ populations (Fig. 5D). These results indicate that although all distinct VR2+ populations have potential to give rise to ECs, the hematopoietic potential is restricted to T500+ cells representing the PR-type mesoderm.

Diversity Among VEGFR2+ Progenitors for ECs and HPCs

The serum-free conditions developed in this study failed to induce any definitive type of hematopoietic cells. Likewise, the SFCs tested here were unable to induce a T500high population. Hence, we investigated whether or not HPC potential is correlated with the level of T500 activity. ES cells were cultured in serum-containing medium, and VR2+ cells were dissected into three populations, T500high, T500medium, and T500low, and assayed for hematopoietic colony (Fig. 6A). As shown in Figure 6A, hemogenic activity in the VR2+ population is gradually lost following an increase in T500-GFP expression. Of particular interest is that myeloid colony activity is present almost exclusively in T500low population. The SFC-induced VR2+ cells that have a potential to give rise to hematopoietic cells expressed a low level of T500 (Fig. 6B). Indeed, a higher number of hematopoietic colony-forming cells was found in a T500-negative population (Fig. 6B). Hence, hematopoietic potential has no correlation with a high level of T500 expression. On the other hand, more CD31+ endothelial colonies were found in the T500lowVR2+ population than in T500VR2+ (Fig. 6C).

Figure Figure 6..

Hematopoietic potential of VEGFR2 (VR2)+ cells generated under distinct conditions. (A): The hematopoietic potential of T500high, T500medium, and T500low populations in VR2+ mesoderm. T500highVR2+, T500mediumVR2+, and T500lowVR2+ cells induced under serum-containing condition were recultured on an OP9 stromal layer. Three days later, the floating cells were gently harvested by pipetting, and the frequency of hematopoietic cell (HPC) progenitors was examined by colony assay. Day 4 colonies correspond to colony forming-unit erythroid, whereas day 7 colonies consisted of burst forming unit-erythroid. The highest colony-forming activity was found in the T500lowVR2+ population. This activity in the VR2+ population was gradually lost with an increase of T500-GFP expression. (B): Hematopoietic potential of T500lowVR2+ and T500VR2+ mesoderm. T500lowVR2+ and T500VR2+ cells induced under serum-free condition with 1 ng/ml BMP4 were purified by fluorescence-activated cell sorting (left panel) and recultured on an OP9 stromal layer under serum-containing condition. Three days later, the floating cells were gently harvested by pipetting, and the frequency of HPC progenitors was examined by colony assay. We counted the colony number at days 4 and 7 as described in (A). Hematopoietic potential of T500VR2+ population was significantly higher than that of T500lowVR2+ population. T500VR2+ cells have a higher potential to form mix, erythroid, and myeloid colonies than T500lowVR2+ cells. Error bars represent SD. The statistical significance was calculated by the Student t test. #, p < .05; *, p < .01 (C): Endothelial potential of T500lowVR2+ and T500VR2+ mesoderm derived from the serum-free culture with 1 ng/ml BMP4. Endothelial cell potential of T500lowVR2+ mesoderm was five times higher than that of T500VR2+ mesoderm. Error bars represent SD. The statistical significance was calculated by the Student t test. All experiments were performed twice with each of two independent cell lines, and the representative result is shown here. Abbreviations: GFP, green fluorescent protein; VEGFR, vascular endothelial growth factor receptor.


Recent cell labeling studies in zebrafish have demonstrated that ECs are generated from most areas that develop mesodermal lineages, although HPC potential is restricted to the posterior region in ventral mesoderm [30]. Moreover, studies based on animal cap assays in amphibians have demonstrated that mesoderm retains fate flexibility even after lineage commitment [31], and some ECs in the chick are even generated from committed somites [32]. In this study, we investigated the potential of various mesoderm subsets in terms of potential to give rise to ECs and HPCs. We prepared distinct types of mesoderm either by FACS sorting in terms of expression of various mesoderm markers or by culturing ES cells under different culture conditions. We wanted to prepare mesoderm subsets that are fractionated geographically into three populations corresponding to the most anterior OR, the primitive streak outside the OR (PR), and extraembryonic region. As shown in our previous study, Gsc is used as marker for mesoderm derived from node area [19]. In addition, we developed ES cell lines transduced with GFP markers driven by the promoter region in the 500 bp upstream of T gene [21]. As shown in the transgenic embryos, this construct expresses GFP at the highest level in the anterior half of the primitive streak, except for the organizer region, and at a medium level in the posterior half of the primitive streak. Neither of the two markers is expressed in extraembryonic regions. Taken together, this experimental system can divide ES cell-derived mesoderm into three populations, T500Gsc+, T500+ Gsc, and T500Gsc.

Although we failed to determine the definite condition to induce the T500high population, we could show that a set of other mesoderm subsets with distinct phenotypes was induced strictly dependent on extrinsic signals in the culture. Consistent with previous studies showing that the differentiation of Xenopus animal cap cells is able to be controlled by the balance of Activin and BMP4 [33, 34], we show that the two factors indeed comprise the fundamental signals for the in vitro differentiation of ES cells. Moreover, our results in ES cell culture are consistent with the phenotype of BMP4-null mice, which exhibit a defect in formation of the posterior part of the embryo that is required for generation of mesoderm at the early gastrula stage [35, 36]. On the other hand, our results on the role of Activin only partially corroborate previous results on the null mutation of Nodal, a molecule that is known to provide the similar signal to Activin. For instance, a null mutation of Nodal displayed a defect not only in the formation of node but also the elongation of the primitive streak [37]. As loss of a specific cell type should have secondary effects on the subsequent morphogenetic processes in early embryogenesis, it is likely that the same treatment causes a more complex phenotype in vivo than in vitro. In fact, as the node emanates molecules involved in the formation of correct primitive-streak [38], it is likely that it affects multiple processes that are not regulated directly by Nodal.

Nonetheless, T500+ cells corresponding to the middle part of the primitive streak are induced by 1–3 ng/ml BMP but not by any concentrations of Activin or a high dose of BMP4, whereas Gsc+ cells corresponding to the node region were induced only in the presence of Activin. In this study, we tested almost 10 different conditions, and none of those conditions gave the same results, even though molecular markers that can be used for specifying mesoderm subsets are limited. Thus, if we are able to specify mesoderm subsets in a detailed manner, we could observe differential induction of a wide variety of different mesoderms. To this end, developing a high-throughput screening platform combined with ES cell culture would be necessary.

Despite the fact that ES cell culture in SFC has enormous future potential, we selected, essentially, only three conditions for this study: 1 ng/ml BMP4 alone, 10 ng/ml Activin alone, and 10 ng/ml BMP4; these conditions, according to the molecular markers used in this study, induce typical three distinct mesoderm subsets: T500lowGsc, T500Gsc+, and T500Gsc populations. VR2+ cells were present in all three populations. VR2, therefore, cannot be a marker for a particular mesoderm population. It should be noted that VR2+ cells in all three mesoderm populations gave rise to endothelial cell colonies on OP9 stromal cell layer, although the frequencies of colony-forming progenitors varied significantly. These data suggest that the EC potential is distributed broadly to various mesoderm subsets in the mouse, as well as other species (Fig. 7) [30, 32]. However, it is clear that T500+ populations contain the highest EC activity compared with other populations, which is consistent with the notion that the primitive streak is the major source of EC progenitors in the actual embryo [11]. Whether or not other mesoderm subsets recruit ECs in the actual embryo and, if so, which vascular systems consist of such subsets generated outside the primitive streak are interesting questions for the future.

Figure Figure 7..

Pathways for endothelial cells and hematopoietic cells in ES cell differentiation cultures. Primitive streak outside the organizer region (PR)-type (T500+ cells) and organizer region (OR)-type (Gsc+ cells) of mesoderms were selectively induced from ES cells under serum-free conditions with a low dose of BMP4 and a high dose of Activin, respectively. Although both types of mesoderm can differentiate into VR2+ cells, endothelial potential in the PR-type mesoderm was significantly higher than that in the OR-type mesoderm. VR2+ cells in OR-type mesoderm have no hematopoietic potential. Although BMP4 alone—as well as the serum—can promote the generation of T500+VR2+cells, only T500low population was induced under serum-free condition with BMP4. T500+VR2+ cells generated under serum-free condition with BMP4 show much lower potential to definitive erythrocytes than those induced under the serum-containing condition, suggesting that some unknown factor(s) is required for differentiation into definitive erythrocytes. A high dose of BMP4 induced another T500VEGFR2+ mesoderm, showing a low endothelial and no hematopoietic potential, closely similar to those of the OR-type mesoderm. The identity of this population is not clear. Abbreviations: BMP, bone morphogenetic protein; ES, embryonic stem; Gsc, goosecoid; VEGFR, vascular endothelial growth factor receptor.

Previous studies by us and other groups have shown that ECs are divided into hemogenic and nonhemogenic populations and that most CD45+ HPCs in ES cell differentiation cultures are derived from VE-cadherin+ ECs [14, 39]. To our disappointment, all defined conditions tested in this study turned out to be insufficient for inducing hemogenic ECs, whereas they are induced in serum-containing medium. Likewise, all defined conditions failed to induce the definitive erythrocytes that are enucleated. Identification of molecules that are required for inducing hemogenic ECs that can give rise to CD45+ cells and the definitive erythrocytes is an important task for future. SFC with 1 ng/ml BMP, on the other hand, is able to support the induction of primitive erythrocytes. Thus, mesoderm cells with hemogenic potential, which are present in T500low population, are induced only by a narrow range of BMP4 concentrations, whereas acquisition of the potential to give rise to definitive hematopoietic cells requires additional signal. Nonetheless, cultures under any defined conditions with serum-free media contain only a restricted range of lineages.

Those results are summarized in Figure 7, which describes four distinct types of VR2+ cells induced by different SFCs. The first one is derived from the Gsc+ cells corresponding to the node region that is induced by Activin. A low dose of BMP under SFC induces a T500low population which have a potential to give rise to the primitive erythrocytes.

The intensity of T500-GFP expression may reflect the spatial localization of mesodermal progenitors in the primitive streak because the highest activity of the T500 promoter is observed in the middle of primitive streak of the transgenic embryos. Further increase of BMP4 concentration results in the T500 population that does not have hematopoietic potential. In conclusion, our study has clearly shown that a narrow range of mesoderm subsets are differentially induced using the monolayer culture of ES cells in SFC. Finally, the potential to give rise to ECs is found in most subsets of mesoderm, whereas hematopoietic potential is induced only in the narrow range of BMP4 concentration.

Disclosure of Potential Conflicts of Interest

The authors indicate no potential conflicts of interest.


We thank M. Royle and Dr. H. Sakurai for critical reading and technical support, respectively. We also thank Dr. K. Nakao for providing transgenic mice. This work was supported by grants from the Leading Project for Realization of Regenerative Medicine (to S.N.), the Knowledge Cluster Initiative (to T.E.), and the Ministry of Education and Science (Grants 17045039 and 18052023) (to T.E.).