BMP-mediated specification of the erythroid lineage suppresses endothelial development in blood island precursors

The large majority of endothelial cells (ECs) within the embryo arise as individual angioblasts throughout mesodermal tissues. If hemangioblasts, bipotential precursors of blood and ECs (B/ECs), do indeed exist, they are located in those regions where blood and endothelial precursors develop in close proximity; for example, in the primitive streak and early blood island of the avian¹  and mammalian embryos, and presumably within the primitive blood-forming regions of fish and amphibian embryos. Because blood and endothelium are both mesoderm derivatives, it is important to define the putative hemangioblast as an immediate precursor of these 2 lineages and not simply as an early mesodermal cell.

The hemangioblast theory was bolstered by molecular evidence that blood and endothelial precursors cells express an overlapping set of marker genes. For example, the cell surface receptor, Kdr, and the transcription factors, Erg, Etv2, Fli1, Gata1/2, Tal1, and Lmo2, are expressed in both blood and endothelial precursors.^2-6  In vitro studies have demonstrated that cells isolated from mouse embryoid bodies and the primitive streak can generate blood and ECs in clonal culture,^2,7-9  and similar results have been obtained using human embryonic stem cells.³  Although these studies show that cultured single cells can give rise to multiple cell lineages, including blood and endothelium, it has been much more challenging to demonstrate the existence of such cells within the embryo itself.

Several studies searching for embryonic hemangioblasts reached different conclusions about the existence, or at least the frequency, of such cells in vivo.^10-13  Photoactivation of lineage tracers in the zebrafish embryo showed that single cells within the ventral marginal region of the gastrula could generate both blood and endothelial derivatives. It was estimated that as many as 12% of cells in this region had properties consistent with the hemangioblast.¹⁰  In contrast, lineage tracing of cells in the mouse blood islands failed to reveal any cells that generated both blood and endothelial progeny from a single clone.¹¹  Although it is not possible to demonstrate the nonexistence of a cell type, the authors concluded that if hemangioblasts exist, they must be exceedingly rare.¹¹  Numerous subsequent studies have failed to generate consensus on the existence of hemangioblasts in vivo, primarily due to questions concerning experimental design or interpretation of results.

In this study, we tested for the presence of hemangioblasts in the ventral blood island (VBI) of Xenopus, which provides a uniquely accessible and localized precursor population. Although markers of both endothelial and blood precursor cells are coexpressed in the VBI, our data demonstrate that these cells become blood and do not normally give rise to ECs, effectively precluding the existence of a classic bipotential precursor. However, experimental inhibition of bone morphogenetic protein (BMP) signaling blocks erythroid development and reveals a latent capacity for early VBI cells to differentiate into ECs. Furthermore, blocking the erythroid transcriptional program resulted in endothelial differentiation. These results indicate that precursor cells in the Xenopus VBI possess both endothelial and erythroid potential, but that in normal development, these cells are not hemangioblasts as such and give rise only to blood. These results may help to explain contradictory results in the literature and aid in understanding the endothelial/blood lineage decision in higher vertebrates.

Embryos were cultured in 0.2× Marc's Modified Ringer (MMR), and at stage (st)12 or 17, embryos were moved to 0.6× MMR for surgical manipulation and healing for 6 to 8 hours, after which they were moved back to 0.2× MMR for subsequent development. Hand-made tungsten wire tools were used to excise the VBI or ventral posterior (VP) portion of the embryo. Inhibitors were dissolved in dimethylsulfoxide, and dose curves were performed to find the optimal potency without developmental defects. Embryos were treated with BMP inhibitors LDN-193189 (Cellagen) at 50 μM and DMH1 (Calbiochem) at 20 μM, NOTCH inhibitor RO4929097 (Cellagen) at 50 μM, WNT inhibitor XAV939 (Tocris) at 100 μM, fibroblast growth factor inhibitor SU50492 (Calbiochem) at 10 μM plus 0.1 mM ATP,¹⁴  and vascular endothelial growth factor (VEGF) inhibitor Ki8751 (Calbiochem) at 10 μM beginning at st13. This study received Institutional Animal Care and Use Committee approval from the University of Arizona (approval number 08-044).

Coding regions of Xenopus laevis gata2 (BC108544, ID# 7978680), lmo2 (BC097502, ID# 4174203), and tal1 (BC072130, ID# 4175038) were inserted into pT7TS expression plasmid for mRNA synthesis. Plasmids were linearized with SalI or XbaI, and mRNA was synthesized using T7 RNA polymerase (Ambion). Histone H2B–red fluorescent protein (RFP) plasmid, VEGF,¹⁵  and Noggin (EXRC clone KL-150) were linearized with NotI, and mRNA was synthesized with SP6 RNA polymerase (Ambion). Antisense digoxigenin-labeled probes were prepared and visualized using standard protocols with further detail available on request.

The tal1 antisense morpholino (MO) (60 ng) (tal1 MO, 5′ ACAGGGACTTCTACTACCTCTCCAA 3′) and lmo2 antisense MO (60 ng) (lmo2 MO, 5′ TACAGTAGTCGATATCTCTCTTTCT 3′) are complementary to the X. laevis translational start site of each gene. The gata2 antisense MOs (25 ng total, 12.5 ng each MO; gata2a MO 5′ CTTCCATCGCAGGAGCAAAGTTCTC 3′, gata2b MO 5′ GGTCAGTAGCCACTTCCATTGCAGG 3′) and etv2 MO have been reported previously.^5,16,17  The etv2, tal1, and lmo2 MOs were all injected at the 1-cell stage. The gata2 MOs were targeted ventral injections at the 4-cell stage. Noggin mRNA injections (total, 25 pg) were targeted to ventral regions of the embryo.

Donor embryos were injected at the 1-cell stage with 500 pg of mRNA encoding H2B-RFP, and labeled VBIs were transplanted to wild-type hosts at st17. Fluorescent imaging was performed on a Leica M205 FA. Migratory cells were manually tracked using the ImageJ plug-in MTrackJ.¹⁸  Embryos of the Flk1::eGFP transgenic line¹⁹  were injected with 500 pg of mRNA encoding H2B-RFP, and VBIs were transplanted to wild-type recipients at st17. Mouse anti-green fluorescent protein (GFP; Invitrogen) and rabbit anti-RFP (Invitrogen) were used at 1:250 following standard procedures.²⁰ 

Similar to the blood island precursor populations of other organisms,^2,3  markers of both the endothelial and blood lineages are coexpressed in the early VBI of Xenopus^4,5,17,21  (supplemental Figure 1 available on the Blood Web site). To our knowledge, however, no study has investigated whether ECs ultimately differentiate within, or arise from, this region. To explore the contribution of the VBI precursor population to the developing B/EC lineages, we surgically excised the entire blood island tissue at stage 17 (Figure 1A), when markers of both lineages are highly expressed⁴  and prior to any migration of cells from the region²²  (and see below). Excision of the VBI resulted in complete elimination of hematopoietic marker expression from the tailbud stage embryo (Figure 1B-K). These results show that the excisions were accurate, and more importantly, the absence of blood markers from the embryo indicates that any putative bipotential B/EC precursors in the VBI were removed. Next we examined endothelial marker expression in embryos from which the VBI had been excised. Multiple markers were examined in the event that the VBI contributed a specific subpopulation of ECs to the embryonic vasculature. Endothelial marker expression appeared unchanged relative to controls for every marker assayed (Figure 1L-W). These experiments demonstrate that the Xenopus VBI is required for development of blood but contributes at most a small proportion of cells to the embryonic endothelium. Assay of the explant tissue at tailbud stage showed strong expression of blood differentiation markers as expected (Figure 1X), but we were surprised to find that expression of EC differentiation markers was undetectable (Figure 1Y). To control for the possibility that EC precursors migrated from the VBI earlier during development, we carried out excisions at st12, soon after B/EC markers are first expressed,^4,21  and observed identical results to the st17 studies (supplemental Figure 2). All subsequent manipulations were carried out at st17, which is the time reported to have the greatest overlap of B/EC markers.^4,5 

If B/EC precursors within the VBI are indeed a source of both hematopoietic cells and ECs, then culture of explanted VBIs in isolation should result in detection of cells expressing EC markers. Because expression of EC markers was undetectable by in situ hybridization (Figure 1Y; supplemental Figure 1), we used quantitative real-time polymerase chain reaction (qPCR) as a more sensitive assay. As a baseline control for these VBI explant studies, tissue was isolated from the VP region of the embryo, which normally produces very few ECs.^14,19  VBI or VP regions were excised at st17 (supplemental Figure 3A). A fraction of the explants were immediately processed for qPCR analysis of B/EC precursor markers, whereas the remainder continued to develop until the tailbud stage. As expected, st17 VBI explants showed strong enrichment for the B/EC precursor markers etv2, erg, fli1, kdr, and tal1 compared with VP explants (supplemental Figure 3B).

Explants were cultured until the tailbud stage, with a proportion of the samples grown in the presence of VEGF, to stimulate/maintain the development of any EC precursors.^15,23  Activity of the VEGF construction was demonstrated experimentally (supplemental Figure 4). VBI explants expressed high levels of erythroid differentiation markers relative to VP explants (supplemental Figure 3C), consistent with their known lineage potential, and the presence of VEGF down-regulated erythroid marker expression, in agreement with previous studies.^24,25  In contrast, analysis of a range of EC differentiation markers showed no appreciable enrichment in the VBI explants relative to the negative control VP explants (supplemental Figure 3D). The presence of active VEGF did not result in up-regulation of endothelial marker expression. Consistent with our previous results (Figure 1Y), this experiment suggests that few ECs are produced in the VBI.

It is possible that excision of VBI tissue interfered with normal endothelial differentiation, and therefore we examined development of possible VBI-derived ECs in whole embryos. Because embryonic angioblasts are reported to be migratory,²⁶  we reasoned that precursors might migrate from the VBI and differentiate elsewhere. To examine this possibility, fluorescent labeling was used to lineage trace cells of the VBI in the embryo. At st17, the VBI of host embryos was removed and replaced with the VBI of donors expressing fluorescent tracer (H2B-RFP) (Figure 2A). When transplanted embryos were observed by time-lapse imaging, migration of cells from the VBI was first observed at st20 to st21 and then in increasing numbers during subsequent development (Figure 2B; supplemental Video 1). Long-distance migration of cells is not a general property of embryonic tissue, because implantation of labeled VP tissue into the VBI region resulted in almost no migratory activity (Figure 2C-D). Both the time of emigration and the morphology of the migratory cells is consistent with embryonic myeloid cells²²  (Figure 2E-L), and labeled cells migrated toward sites of injury on the epidermis, consistent with a myeloid phenotype (data not shown). When embryos were allowed to develop past the initiation of the heartbeat, labeled erythroid cells were observed circulating within embryonic vessels (data not shown).

To identify myeloid cells using molecular markers, we carried out VBI transplants using red Dextran labeled donor tissue, because this fluorescent dye can withstand fixation. When transplanted embryos were assayed by double in situ hybridization using the myeloid markers spib and runx1 (Figure 2N-N′), we observed that ∼88% of Dextran-labeled cells colocalized with myeloid markers (290/328 cells, n = 4 embryos; Figure 2M-O′). However, a significant number of cells were not detected using the myeloid probes (white ring in Fig. 2O′). These cells are candidates for VBI-derived ECs but alternatively could be hematopoietic cells that are not expressing spib or runx1.

When angioblasts differentiate into ECs, they express cell adhesion molecules, cease migration, and assemble into tubular vessels.^26,27  To determine whether any of the labeled VBI-derived cells were postmigratory (consistent with a differentiated endothelial phenotype), embryos were observed by time-lapse microscopy for ∼7 hours after initiation of heartbeat. Fluorescently labeled cells were then manually tracked to assess migration (Figure 2P; supplemental Video 2). Of >200 cells examined, all displayed migratory behavior, consistent with a myeloid, but not an endothelial, phenotype. These results lead us to conclude that any contribution of VBI-derived cells to the embryonic vasculature must be rather small. It also suggests that most of the cells that were fluorescently labeled, but not positive for spib or runx1, were immune system cells and not ECs.

Our results show that if any VBI derived ECs exist, they are not abundant. Therefore, we carried out 2 additional studies in an attempt to detect these rare cells. First, we asked whether transplantation of wild-type VBI (containing B/EC precursors) could replenish embryos in which the molecular pathway controlling EC development was blocked by depletion of the critical transcription factor Etv2. Previous studies have shown that loss of Etv2 eliminates EC specification and marker expression.^17,28,29  Therefore, we transplanted wild-type red Dextran-labeled VBI regions into host embryos injected with etv2 MO (Figure 3A). After developing to the tailbud stage, experimental embryos were assayed by in situ hybridization for expression of the EC differentiation marker cdh5. Because Etv2 activity is depleted, the only ECs detected in the host embryo should be red Dextran-labeled cells derived from the wild-type VBI. First, we confirmed that injection of the etv2 MO almost completely eliminated detectable cdh5 expression (Figure 3B-C). In VBI implant embryos (Figure 3D-G′), red-labeled cells from the VBI were present in normal numbers distributed throughout the recipient embryos (Figure 3F′,G′). However, the rare cdh5-positive cells in recipient embryos (circled regions in Figure 3F-G) were never Dextran labeled (n = 8 embryos), indicating that they were not VBI derived.

Second, to rule out the possibility that the Etv2 depletion might somehow affect migration or survival of wild-type derived EC within the host embryo, we carried out related studies in a wild-type host environment (Figure 3H). Briefly, Flk1::EGFP transgenic donor embryos¹⁹  were injected with lineage tracer (H2B-RFP), and at st17, the VBI region was transplanted into wild-type host embryos. If cells from the VBI contribute to the embryonic vasculature, then RFP-labeled, GFP-positive cells will be present in the endothelium of the host embryo during later development. As expected, numerous RFP-labeled cells were visible in host embryos (Figure 3J); however, unlike the transgenic donor embryo, which showed GFP expression throughout the developing vasculature (Figure 3L), no GFP/RFP double-labeled cells were detected in any of the recipient embryos (Figure 3K; n = 6). Because ∼400 red-labeled VBI-derived cells were counted in each of the host embryos, these results show that VBI-derived ECs are extremely rare and possibly nonexistent in the Xenopus embryo.

Previous studies have implicated the fibroblast growth factor, VEGF, BMP, WNT, and NOTCH signaling pathways in regulating the relative abundance of hematopoietic vs ECs in the embryo.^4,30-35  We tested whether these signaling pathways might regulate the blood/endothelial fate decision in the Xenopus VBI (Figure 4). In all cases, embryos were treated continuously from postgastrulation until tailbud using small molecule inhibitors specific for individual pathways. The effectiveness of the inhibitors was demonstrated experimentally (supplemental Figure 5). Treatment was commenced after gastrulation to allow completion of primary axial patterning of the embryo before signaling pathways were inhibited. Examination of the globin (hba1) in situ pattern showed that expression was similar to controls for all inhibitors except those of the BMP and WNT pathways, which caused a marked reduction in globin staining (Figure 4D-E″). When endothelial markers were analyzed, only BMP inhibition consistently resulted in ectopic endothelial expression in the VBI region (Figure 4J-J″). Previous studies have shown that WNT modulates BMP in the Xenopus VBI,³⁶  suggesting that the observed effects are mediated through the BMP pathway.

Because small molecule inhibitors can exhibit off-target effects or nonspecific toxicity, we tested whether Noggin, a protein inhibitor of BMP signaling,³⁷  and another small molecule BMP inhibitor, DMH1, exhibited similar effects on blood and endothelial marker expression. The results confirm that inhibition of BMP signaling increased endothelial marker expression at the expense of blood markers (Figure 5; supplemental Figure 6). Targeted inhibition of BMP signaling in the VBI using Noggin resulted in complete elimination of globin expression (Figure 5D-D′) and ectopic EC marker expression (Figure 5F-F′). Noggin was probably more effective than small molecule inhibitors because it was present throughout early development while the small molecules were added only after gastrulation. In addition to erythroid and endothelial markers, we examined the effect of inhibition of BMP signaling on the myeloid markers mpo, runx1, and spib. In situ hybridization analysis and quantitation of cell numbers showed that myeloid development was unaltered by inhibition of BMP (Figures 5G-J′ and 6H-M′′; supplemental Figure 7), indicating that the precursors of the erythroid and myeloid lineages respond independently to BMP signaling. Furthermore, the observation that myeloid development was normal indicates that patterning of extreme ventral tissues was not disrupted by inhibition of BMP signaling.

Examination of additional EC differentiation markers in BMP-inhibited embryos confirmed increased expression of EC genes in the VBI region (Figure 5M-T). These results were validated by qPCR when VBI explants treated with BMP inhibitor were assayed for blood and EC marker expression (Figure 5U).

The appearance of ECs in the VBI domain can be explained in at least 2 different ways. First, multipotent precursors may have changed fate from the erythroid to the endothelial lineage. Second, inhibition of BMP signaling may have eliminated the blood lineage and, perhaps independently, allowed lateral ECs to migrate into the ventral domain. To distinguish between these possibilities, we first made VBI explants from unmanipulated postgastrula embryos, treated the explants with BMP inhibitor or control medium, and assayed for EC markers at the tailbud stage (supplemental Figure 8). Because the explanted tissues were isolated from the embryo, it would not be possible for angioblasts from more dorsal lateral regions to migrate into the VBI region. As in intact embryos, BMP inhibition resulted in reduction or elimination of erythroid marker expression (supplemental Figure 8A-B), whereas robust expression of endothelial markers was observed within the explanted tissue (supplemental Figure 8C-D). We conclude that ECs are derived from within the explanted VBI tissue and not from the lateral plexus.

Second, we confirmed that erythroid and EC precursor markers were indeed expressed in the same cells. Single-cell PCR analysis of cells isolated from the mesodermal layer of the VBI confirmed that transcripts of the EC marker, kdr, were coexpressed with hematopoietic precursor markers (Figure 6A). Third, we carried out a time course analysis of B/EC precursor marker expression in response to inhibition of BMP signaling. In control embryos, VBI expression of the erythroid precursor markers gata2, lmo2, and tal1 increased as blood development progressed (Figure 6B,D,F), whereas expression of angioblast markers decreased (Figure 6N,P,R). In contrast, when embryos were treated with BMP inhibitor, expression of the erythroid markers rapidly decreased (Figure 6C,E,G), and expression of angioblast markers persisted (Figure 6O,Q,S). Expression of myeloid markers was unchanged by BMP inhibition (Figure 6H-M). These results indicate that expression of EC precursor markers was maintained in the precursor cell population and did not arise in a different cell population in response to inhibited BMP signaling.

BMP signaling may have multiple different roles in modulation of cell fate or behavior in the blood island, but perhaps the most critical function is as an activator of erythroid differentiation.^38,39  To test this proposition, we cell autonomously disrupted erythroid development by knockdown of critical blood lineage transcription factors in the context of normal BMP signaling. Previous studies have demonstrated that the transcription factors Gata2, Lmo2, and Tal1 are essential for erythroid development.^16,40-42  MO-mediated depletion of each of these transcription factors resulted in reduced erythroid differentiation as expected (Figure 7B-D). Significantly we found that blocking blood development by knockdown of Gata2, Lmo2, or Tal1 was sufficient to induce robust expression of angioblast markers in the VBI region (Figure 7I-K,M-O). These results suggest that, during normal development, commitment to the erythroid lineage blocks EC fate. However, when development of the erythroid lineage is disrupted, differentiation of precursors to mature ECs is able to proceed.

Despite elegant experiments in mice, chicks, and zebrafish, in vivo existence of the hemangioblast remains controversial.^10-13  Starting at the late gastrula stage, markers of both hematopoietic and endothelial precursors are expressed in the VBI of Xenopus.^4,5,17,21  Coexpression of these markers raises the possibility that the blood island might contain cells with hemangioblast properties. Although hematopoiesis within the VBI is well established, development of ECs from VBI precursors has not previously been investigated. In the studies reported here, no VBI-derived ECs were detected (Figures 1-3; supplemental Figures 2-3). Because the classically defined hemangioblast gives rise to both endothelial and hematopoietic progeny, our results infer that hemangioblasts are not present in the Xenopus blood island during normal development.

During early postgastrula development, angioblast markers are expressed at high levels in the VBI, but expression is lost ∼12 hours later by early tailbud stages, and these cells never form blood vessels. Over this same period, expression of hematopoietic markers expands. When we examined the role of different growth factor signaling pathways in this process, we found that BMP was essential for maintenance of the erythroid program (Figures 4-6; supplemental Figure 6). When BMP signaling was compromised using either small molecule or protein reagents, markers of the erythroid lineage disappeared while angioblast markers persisted and EC differentiation markers appeared. Nonerythroid blood lineages appeared to be unaffected by inhibition of BMP signaling (Figures 5G-J′ and 6H-M′′; supplemental Figures 7 and 8G-H). The critical role of BMP signaling in embryonic hematopoietic development is well established,^16,33,38,39  but the reciprocal relationship between erythroid and endothelial differentiation in the blood islands has not been reported previously. Similar to the original hemangioblast hypothesis, our work using BMP inhibitors indicates a remarkably close relationship between the endothelial and blood lineages, particularly the erythroid lineage. However, although VBI cells double positive for B/EC precursor markers have a latent potential to develop toward either blood or endothelial lineages, we find that only blood cells are derived from the VBI precursor population during normal development, indicating that these cells are not classical hemangioblasts.

Although growth factor signaling could play multiple instructional or inhibitory roles in the VBI, we focused on the recognized role of BMP in promoting blood development.⁴³  We reasoned that activation of the blood program by BMP might prevent further development of the precursor population toward endothelial differentiation. A very similar model has been proposed by Nishikawa, who hypothesized that, within a precursor population, specification of the blood lineage might obviate the endothelial program.⁴⁴  We directly tested this proposal by reducing the activity of 3 individual components of the erythroid transcriptional machinery: Gata2, Lmo2, and Tal1 (Figure 6). In each case, erythroid differentiation was severely reduced, with a concomitant appearance of endothelial differentiation markers. The lineage fate switch occurred in the context of normal BMP signaling, because it has previously been shown that disruption of blood development does not alter BMP expression.¹⁶  In agreement with the Nishikawa prediction,⁴⁴  these results strongly suggest that commitment to the erythroid lineage eliminates the potential for endothelial differentiation in the precursor population. These results also raise the possibility that transcription factors required for activation of the erythroid transcriptional pathway might simultaneously suppress the endothelial transcriptional program in the precursor population.⁴⁵ 

What population of cells in the amniote embryo is the functional equivalent of the early postgastrula blood island of Xenopus? Lineage tracing studies in mice have shown that cells normally giving rise to extra-embryonic blood and endothelium occupy distinct locations in the primitive streak.⁴⁶  However, Flk⁺ cells isolated from regions of the early and mid-streak can develop into endothelial and hematopoietic cells in clonal culture.^7,9  We suggest that cells in the early Xenopus blood island are equivalent to the previously described Flk⁺ precursor population in the mouse streak. This raises the possibility that cells ingressing through the amniote streak at slightly different times encounter different BMP signaling activity. Cells of the mouse early streak population may encounter BMP ligand³⁸  and differentiate into extra-embryonic blood. If cells that ingress at mid-streak do not receive BMP signals, they will develop as extra-embryonic endothelium.⁴⁶ 

The authors thank Parker Antin for the use of the time-lapse imaging microscope, Cristy Lewis for dorsal aorta images, Aaron Zorn and Ondine Cleaver for comments on the manuscript, and Sarah Appleby for help with in situ hybridizations.

C.T.M. is supported by National Institutes of Health, National Institute of General Medical Sciences training grant T32 GM08659 and the Philanthropic Educational Outreach Scholar award. P.A.K. is the Allan C. Hudson and Helen Lovaas Endowed Professor of the Sarver Heart Center at the University of Arizona College of Medicine. This work was supported by the Sarver Heart Center and by Heart, Lung, and Blood Institute grant HL093694.

Contribution: C.T.M. designed the study, performed experiments, analyzed data, and wrote the manuscript; and P.A.K. designed the study, analyzed data, and wrote the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Paul A. Krieg, Department of Cellular and Molecular Medicine, University of Arizona College of Medicine, 1656 E Mabel St, MRB311, Tucson, AZ 85724; e-mail: pkrieg@e-mail.arizona.edu.