Regulation of Endothelial Cell Differentiation and Specification

Signaling Pathways

Fibroblast Growth Factor 2 (FGF2 or bFGF) and Bone Morphogenetic Protein 4 (BMP4) are two key signaling components that are not only important for specification of mesoderm,^3–5 but also for its differentiation toward endothelial and hematopoietic cell fates.^6–8 BMP4 is sufficient to induce mesodermal differentiation in vitro, whereas its ablation results in a failure to generate mesoderm and leads to early embryonic lethality.^9–11 Embryos deficient for downstream effectors of BMP4 signaling, such as Smad5, lack an organized yolk sac vasculature similar to Bmp4^−/− mutant mice.⁶ Smad4 null mice display similar phenotypes and are also remarkably smaller in size, owing to severe cell proliferation defects.^{7, 12} Meanwhile, gene deletion experiments demonstrate FGF2 signals via FGFR1 to induce and pattern the mesoderm.^{5, 8, 13} The hierarchy of these signals has not been clearly defined in vivo, and ES cell systems have yielded conflicting data. Experiments utilizing mouse ES (mES) cells suggest that BMP4 resides at the top of this signaling hierarchy, promoting mesoderm formation and initiating a FGF2-dependent program to regulate the specification of angioblasts, which are thought to function as endothelial progenitors.^{9, 10, 14} In contrast, differentiation studies in human ES (hES) cells, which require bFGF for survival and growth even in an undifferentiated state, do not support a role for this factor in the commitment of mesodermal cells to the endothelial lineage.¹⁵ Rather, in hES cells, BMP4 is a critical regulator which functions downstream of Indian Hedgehog (IHH) to promote endothelial cell differentiation.¹⁵

IHH has also been shown to be a critical regulator of endothelial cell development during murine embryogenesis, although its role as a key inductive signal is not entirely clear. IHH is secreted by the yolk sac visceral endoderm during vasculogenesis; its receptors are expressed within the underlying mesoderm. IHH has been shown to be sufficient to promote the formation of endothelial and hematopoietic cells in mouse embryo explant cultures that lack endoderm (Figure 1), and it can also re-specify prospective neural ectoderm cells to adapt hematovascular fates. However, embryos deficient for IHH or its receptor Smoothened¹⁶ do exhibit some degree of endothelial cell differentiation, although less than normal. Thus, IHH may not be the only inductive signal in vivo, and the visceral endoderm may produce other signals required for endothelial cell development within the mesoderm.

Another visceral endoderm-derived factor that is a key regulator of vasculogenesis is Vascular Endothelial Growth Factor (VEGF); however, its cellular role in endothelial cell development in vivo is also not entirely clear. VEGF-A is the most extensively studied member of the VEGF family and is expressed by the extraembryonic visceral endoderm in the mouse as early as embryonic day (E)7.5, coincident with blood island formation in the yolk sac.¹⁷ The requirement for VEGF-A is established early during vasculogenesis, as heterozygous mutants are embryonic lethal due to failed development of the vasculature.^{18, 19} Overexpression of VEGF-A also impairs cardiac development, and causes embryonic lethality at midgestation.²⁰ These data reveal a precise dosage requirement for this growth factor for proper cardiovascular development.

VEGF-A signals through its main receptors, VEGFR1 (Flt-1) and VEGFR2 (Flk-1 or Kdr), and also interacts with the co-receptors Neuropilin 1 and 2 (Nrp-1/2). Although Flk-1 has a lower affinity for VEGF-A than Flt-1, it has stronger tyrosine kinase activity, and VEGF-A responses in endothelial cells and their precursors are usually attributed to Flk-1 activation. Mice lacking Flk-1 are embryonic lethal at E8.5–9.5 and lack blood island and vascular plexus development, despite normal formation of angioblasts.²¹ Consistent with this, Flk-1^−/− mES cells can generate endothelial cells, however they fail to propagate in vitro.²² Similarly, VEGF-A treatment of undifferentiated hES cells does not promote their differentiation toward an endothelial cell phenotype.¹⁵

Collective in vivo and in vitro data from murine and human systems suggest that VEGF-A likely regulates the survival and/or propagation of endothelial cells, and not necessarily their differentiation. However, given that there are observed differences in the hierarchy of signals that controls endothelial cell differentiation/propagation in murine and human systems, it is critical to conduct more studies of vasculogenesis in human pluripotent stem cells, which may provide novel insights more directly applicable to the treatment of human pathologies.

Transcriptional Regulators

ETS transcription factors are also known to play important roles in endothelial cell development. ETS family members share a highly conserved winged helix-turn-helix DNA binding domain that binds to the ETS DNA binding site 5′GGA(A/T)3′. There are approximately 30 known mammalian ETS factors that generally serve as transcriptional activators, although some have been associated with repressor activity.²³ Several members of this family, such as Ets1, Erg, Fli-1 and Etv2 (discussed below) have been demonstrated to play important roles in vascular, as well as hematopoietic, development.²³ Regulatory regions for almost all endothelial genes contain ETS binding sites, implying that ETS factors could potentially regulate most, if not all, endothelial-specific gene expression.^{23, 24}

Although many ETS factors are thought to share redundant functions, recent studies have demonstrated a specific and critical role for Etv2(ER71/etsrp), or ETS variant 2, in the differentiation of mesodermal progenitors toward an endothelial cell fate. Etv2 expression is initially observed to be more widespread within the primitive streak mesoderm, however it is soon restricted to developing vascular endothelial cells^25–28 and its critical function in their development is evolutionarily conserved.^{25–27, 29–31} Morpholino knockdown²⁵ of, or mutations³⁰ within, the etsrp gene in zebrafish results in complete abrogation of circulation, together with loss of expression of vascular endothelial cell markers, including VE-cadherin/Cdh5 and VEGFR2/Flk-1, in developing embryos.^{25, 30} Likewise, Etv2^−/− mice are embryonic lethal, and despite having intact mesoderm, they fail to generate angioblasts,²⁵ a yolk sac primitive vascular plexus, dorsal aortae, or endocardium.^{26, 32} Overexpression of Etv2 leads to ectopic expression of endothelial markers in multiple cell types in vivo,^{25, 27, 29, 31} suggesting an expansion of the differentiation of endothelial cells. Thus, Etv2 appears to be unique among ETS domain transcription factors in that it is both necessary and sufficient for in vivo vasculogenesis.^{25–27, 30, 32}

Importantly, maintaining Etv2 expression within a narrow developmental window is critical for proper vasculogenesis. Transient Etv2 activation in the mesoderm is sufficient to initiate endothelial cell development^{28, 32, 33} and Etv2 is rapidly downregulated once endothelial cells begin to mature.²⁸ Constitutive expression of Etv2 in endothelial cells under control of the Tie2 promoter results in dilated yolk sac vessels and impaired remodeling of the extraembryonic vasculature, eventually leading to embryonic lethality.³⁴ Interestingly, this also results in perturbation of normal hematopoiesis due to “endothelialization” of hematopoietic progenitor cells that do not form properly.³⁶

Molecular interactions between Etv2 and other signaling factors that regulate endothelial cell development are not yet clearly defined; however, Etv2 is closely associated with and known to precede Flk-1 expression during vasculogenesis.^{25, 26, 30} As previously mentioned, in vivo disruption of Etv2 or its orthologues results in loss of Flk-1-expressing cells,^{25, 26, 30, 32} whereas forced expression of this ETS factor leads to ectopic Flk-1 expression.^{25, 27, 29, 31} Flk-1:GFP transgenic zebrafish embryos injected with etsrp-directed morpholinos demonstrate loss of GFP+ cells in blood vessels.²⁵ Etv2 expression is observed exclusively within Flk-1+ mesodermal cells in mES cell differentiation studies, and its induction within embryoid bodies results in de novo production of Flk-1+ cells.²⁶ Thus, Etv2 appears to regulate expression of VEGF receptor Flk-1 within newly forming endothelial cells. Consistent with this, Etv2 also appears to be required for the generation of Flk-1+PDGFRα-vascular progenitors from Flk-1+PDGFRα+ primitive mesoderm.³⁵

Combining what is known about the role of ETS transcription factors and the VEGF-A/Flk-1 pathway, we can surmise that ETS transcriptional activity in mesodermal progenitors is necessary, if not sufficient, to induce Flk-1 expression, as well as promote commitment to an endothelial cell fate. VEGF-A signaling via Flk-1 likely regulates the survival and propagation of newly forming endothelial cells, although it does not likely directly regulate their differentiation.

VEGF-A

As mentioned above, VEGF-A is required to stimulate endothelial cell proliferation, and does so largely via signaling through Flk-1. Binding of VEGF-A to Flk-1 triggers its autophosphorylation, leading to complex formation with integrin alpha v beta 3 (αVβ3),³⁶ an endothelial cell adhesion receptor, and αVβ3 auto-phosphorylation.³⁷ Acting in a feedback loop, integrin αvβ3 is also capable of phosphorylating Flk-1 once it is bound by its ligand vitronectin (Vn).³⁸ The cross-activation of these two receptors leads to the recruitment and activation of Src kinases,³⁹ which mediate the mitogenic effects of VEGF-A^{40, 41} by activating the MAPK intracellular signaling pathway.³⁸

Flt-1/VEGFR1 also indirectly controls endothelial cell proliferation, by competing with Flk-1 for binding of VEGF-A. Flt-1^−/− mutants^{42, 43} and zebrafish flt1 morphants⁴⁴ exhibit endothelial cell hyper-proliferation that leads to abnormal angiogenic development, suggesting that unlike Flk-1, Flt-1 plays a negative role in endothelial cell proliferation. Although VEGF-A has a higher affinity for Flt-1 than Flk-1, Flt-1 has a weaker kinase activity, and is thought to act in part as a VEGF-A decoy receptor.⁴⁵ In support of this idea, genetic studies have shown that the Flt-1 tyrosine kinase domain is dispensable for vasculogenesis,⁴⁶ whereas the extracellular ligand-binding domain is essential to its ability to regulate proper VEGF-A signaling.⁴⁵ Moreover, alternative splicing results in a soluble, catalytically inactive Flt-1 isoform that acts as a molecular “sink” for VEGF-A, and thereby limits Flk-1 ligation and activation.^{47, 48}

Retinoic Acid

Retinoic acid (RA) signaling also controls endothelial cell proliferation, as a negative regulator of cell cycle progression. RA is the biologically active derivative of dietary vitamin A or retinol. In circulation, retinol is complexed to retinoid-binding protein (RBP) and can enter the cell via a specific receptor, STRA6. Retinol is initially oxidized into retinaldehyde by either alcohol dehydrogenases (ADH) or retinol dehydrogenases (RDH). Retinaldehyde dehydrogenases (Raldh1–4) then catalyze the conversion of retinaldehyde to all-trans RA, which is released and can be taken up by surrounding tissues. Cellular RA-binding proteins (CRABP) facilitate the uptake of RA in target cells, as well as its transport to the nucleus. RA then binds to nuclear RA receptors (RARα, β, and γ), which heterodimerize with retinoid X receptors (RXRα, β, and γ) to form active transcription factor complexes. These complexes occupy RA-response elements (RARE) within target genes, altering the binding of transcriptional repressors and co-activators to modulate gene expression. In cells with inactive RA signaling, RA is oxidized for degradation by the cytochrome P450 enzyme, CYP26 (reviewed in Duester ⁴⁹).

In the murine and rat yolk sac, Raldh2 is expressed specifically by the visceral endoderm at the onset of vasculogenesis;^{50, 51} whereas yolk sac endothelial cells within the adjacent mesoderm exclusively express RA receptors RARα1 and 2.^{51, 52} Mice lacking the enzyme Raldh2 are RA-deficient and die in utero at approximately E10. One of the first morpohological defects observed in these mutants is lack of remodeling and blood vessel dilation within the yolk sac and embryo proper, as well as a marked decrease in blood cell production.^51–53 Raldh2^−/− mutant endothelial cells are hyper-proliferative and exhibit reduced levels of p21(Cdkn1a) and p27(Cdkn1b), members of the Cip/Waf family of cyclin-dependent kinase (Cdk) inhibitors (CKIs). In vivo and in vitro studies collectively suggest that RA regulates endothelial cell cycle progression by upregulating the expression of p21 and p27, resulting in G1 arrest of endothelial cells⁵² (Figure 2).

TGF-β and Extracellular Matrix

Retinoic acid signaling also suppresses endothelial cell proliferation by upregulating Transforming Growth Factor beta-1 (TGF-β1). Signal transduction via TGF-β family members is facilitated by heteromeric type I and type II serine/threonine kinase receptors. TGF-β first binds to the constitutively active type II receptor dimer, leading to recruitment, phosphorylation and activation of type I receptor dimers. Type I receptors in turn phosphorylate receptor-regulated Smad proteins, which function as effectors of TGF-β signaling by binding to common partner Smad proteins (Co-Smads), such as Smad4. These TGF-β “pre-complexes” then shuttle to the nucleus, where they associate with other transcription factors to regulate target gene expression (reviewed in Lee and Bae⁵⁴).

TGF-β signaling contributes to the control of endothelial cell proliferation in two important ways; firstly, by inducing the production and deposition of fibronectin (Fn), an extracellular matrix (ECM) protein. Endothelial-derived Fn not only promotes visceral endoderm survival and function, which is necessary for VEGF-A production, but also binds to integrins alpha 5 beta 1 (α5β1) and αVβ3 expressed on endothelial cells to elicit opposing effects on cell cycle progression. Fn signaling via α5β1 inhibits proliferation, whereas binding to αVβ3 induces endothelial cell proliferation via the Flk-1/MAPK pathway⁵¹ (Figure 2). TGF-β signaling also controls endothelial cell cycle progression via the regulation of key cell cycle control genes. As previously discussed, progression through G1 into S phase of the cell cycle is regulated by cyclin-cdk complexes such as cyclin D-Cdk4/6 and cyclin E-Cdk2. Conversely, interaction between CKIs (such as RA-induced p21 and p27) and Cdks promotes cell cycle arrest. TGF-β signaling increases expression and stabilization of p15, a member of the INK4 family of CKIs.^{55, 56} Increased p15 protein levels in the cytoplasm allows it to bind cyclin D-Cdk4/6 complexes more readily than p27, and active cyclin D-Cdk4/6-p27 complexes are replaced by inactive cyclin D-Cdk4-p15 complexes. Displaced p27 molecules are shuttled towards and interact with cyclin E-Cdk2 complexes, rendering them inactive; since p27 binding prevents Cdk2 activation which leads to G1 arrest.⁵⁷ TGF-β signaling also contributes to stabilization of p27 by blocking its phosphorylation⁵⁸, a pre-requisite for its export from the nucleus and subsequent degradation.⁵⁹ Thus, regulation of endothelial cell cycle progression and proliferation during vascular plexus remodeling requires a balance between multiple signaling effectors and downstream intracellular pathways.

Notch and VEGF Signaling

Interplay between VEGF-A/Flk-1 and Notch pathways is primarily responsible for the specification of tip versus stalk cell identity during sprouting angiogenesis. There are multiple vertebrate Notch genes; humans and mice express at least four isoforms (Notch–-4), and Notch1 and Notch4 are expressed by vascular endothelial cells. Evolutionarily conserved across diverse species, Notch-mediated signaling occurs over short distances and is commonly implicated in cell fate determination. Single-pass transmembrane Notch receptors are activated by engagement of ligands presented by neighboring cells, and are cleaved by metalloprotease enzymes which remove most of the receptor extracellular domain. An enzymatic complex known as γ-secretase, then cleaves the Notch transmembrane domain, releasing a Notch intracellular domain (NICD) which is capable of translocating to the nucleus and forming complexes with other DNA-binding proteins (reviewed in Bigas and Espinosa ⁶²).

Expression of the Notch ligand Dll4 by tip cells of growing vascular sprouts is VEGF-dependent.^63–65 Thus, VEGF-A provides an angiogenic stimulus that guides sprouting behavior and subsequent initiation of endothelial cell proliferation.^65–68 Dll4 activates Notch signaling in adjacent endothelial cells, reducing their ability to respond to VEGF-A stimulation by downregulating Flk-1 expression^{65, 67} Dll4 thus functions to suppress tip cell fate by effectively conferring stalk cell identity on these neighboring cells.^{67, 69, 70} Impairment of Dll4-Notch signaling results in dysregulation of tip versus stalk cell specification. Dll4 heterozygous mutant mice and dll4 zebrafish morphants display increased angiogenic sprouting due to increased tip cell formation.^{63, 67–69, 71} Loss of Dll4 in mice also results in increased proliferation of endothelial cells.^{65, 68, 69} Similarly, inhibition of Notch signaling results in increased sprouting, increased expression of tip cell-associated genes,⁶⁷ hyper-proliferative endothelial cells, and a denser and more highly-interconnected vasculature.^{67, 69, 70} Endothelial-specific deletion of Notch1 is sufficient to increase tip cell numbers.⁷¹ Interestingly, when both VEGF-A and Notch signaling are impaired, excessive sprouting and tip cell formation are not observed.^{63, 67, 69} VEGF and Notch regulation of angiogenic sprouting is therefore clearly complex; VEGF-A provides a pro-angiogenic stimulus and induces Flk-1 expressing cells to sprout and form new vessels, whilst simultaneously activating Dll4-Notch pathway signaling to maintain the necessary balance between tip and stalk cell specification.

Interactions between the VEGF and Notch pathways are further complicated by the involvement of other members of these signaling pathways. For example, VEGFR3 (or Flt-4) becomes ectopically expressed throughout remodeling vessels when Notch signaling is blocked.⁶⁹ Endothelial-cell specific deletion of Flt-4 in mice leads to excessive angiogenic sprouting and branching in retinal vessels, resembling loss of Notch signaling. VEGF-C binds to Flt-4 in tip cells, promoting stalk cell specification.⁷² Flt-1 also acts as a negative regulator of tip cell differentiation in a Notch-dependent manner. Loss of flt-1 function in zebrafish results in a phenotype consistent with loss of Notch signaling, while overexpression of the notch1a intracellular domain in flt-1 morphants restores proper vessel branching and patterning.⁴⁴

Other Notch ligands, including Dll1 and Jag1, are also involved in angiogenic sprouting, with reported pro-angiogenic effects.^{73, 74} Heterozygous deletion of Dll1 leads to reduced vessel branching, although it is not expressed by endothelial cells themselves. Dll1 is expressed by cells peripheral to the vasculature, and acts as an extrinsic signal to activate Notch signaling in presumptive tip cells, which reinforces VEGF-A action by further inducing Dll4 expression to promote tip cell specification.⁷³ Endothelial-specific deletion of Jag1 also results in reduced vessel sprouting. Stalk cells are enriched for Jag1, which antagonizes Dll4 signaling by competing for Notch receptor binding, and reduces Dll4-Notch activation in adjacent tip cells.⁷⁴

Although Notch ligands can exert opposing effects on endothelial cells, proteasomal degradation of Notch signaling components positively regulates angiogenesis and tip cell specification. Endothelial cell-specific deletion or morpholino knockdown of Fbxw7, an E3 ubiquitin ligase, results in increased stability of Notch signaling components and impaired tip cell formation.⁷⁵ BAZF (B-cell chronic lymphocytic leukemia/lymphoma6-associated zinc finger protein), a component of the cullin3-based E3 ligase complex, exerts its pro-angiogenic effect by promoting degradation of Notch signaling factor CBF1 (C-promoter binding factor 1), thereby curtailing Notch signaling and lateral inhibition of tip cell phenotype.⁷⁶

The specification of tip and stalk cell identities among endothelial cells is dynamic in both in vivo and in vitro systems.^{77, 78} Arima and coworkers⁷⁷ speculated that in addition to other key functions, tip cells create a localized microenvironment which facilitates further elongation by the successive tip cell. Interestingly, among the genes enriched in tip cells are ECM-degrading enzymes and secreted proteins nidogen-1 and -2, which interact with ECM proteins such as laminin and collagen to stabilize basement membranes.⁷⁹ Laminin and laminin-binding integrins have also been found to stimulate Dll4-Notch signaling during tip versus stalk cell selection.^{80, 81}

MicroRNA

MicroRNAs (miRs) are emerging as new players in the regulation of tip versus stalk cell specification of endothelial cells. Primary microRNA transcripts are transcribed from intergenic or intronic genomic sequences, and then cleaved by the RNase III endonuclease Drosha, allowing export to the cytoplasm. Once there, miR transcripts are processed through the endonuclease Dicer to yield mature miRs, small RNA fragments 18–25 nucleotides long. These bind to the 3′ UTRs of their respective target mRNAs to mediate post-transcriptional gene silencing.⁸² In vivo silencing of either Dicer or Drosha results in reduced sprouting of endothelial cells, demonstrating an essential role for miRs in angiogenesis.⁸³

miR-126 is known to elicit pro-angiogenic effects by repressing endothelial expression of SPRED1 (Sprouty-related protein 1), a known negative regulator of VEGF-A signaling, thereby enhancing the angiogenic response of endothelial cells to VEGF-A signaling.^{84, 85} Another pro-angiogenic microRNA is miR-221, which is also expressed by embryonic endothelial cells. Among its targets is p27,⁸⁶ previously implicated in regulating endothelial cell proliferation in response to RA signaling.⁵² Notch signaling represses expression of miR-221, thereby allowing p27 expression and subsequent regulation of stalk cell proliferation.⁸⁶ miR-27b also promotes endothelial cell tip specification, as well as increased branching of the vascular plexus by downregulating Dll4 and Sprouty homologue 2 (Spry2) expression.⁸⁷

Prox1

Prospero homeodomain transcription factor (Prox1) is a homeobox transcription factor expressed by the subpopulation of venous endothelial cells that sprout and migrate to form the lymphatic system.¹⁰² A lineage-specific marker for LEC, Prox1 is essential for embryonic lymphatic development^{103, 104} (Figure 1). Venous endothelial cells begin expressing Prox1 at E9.75,¹⁰⁵ and by E11.5 clusters of Prox1-positive lymphatic progenitor cells emerge along the anterioposterior axis of the cardinal vein.^{102, 103, 106} These clusters undergo ballooning to generate intermediate structures known as lymph sacs, from which lymphatic vessels emerge by a combination of sprouting^{102, 106} and migration of streams of Nrp-2-positive LEC from the cardinal vein.¹⁰⁶ Ablation of Prox1 leads to a general failure of lymphatic endothelial cell specification and ultimately, a failure of lymphatic vascular development.^103–105 Prox1-null endothelial cells are devoid of expression of LEC markers such as LYVE-1 and VEGFR-3, and instead express blood vessel endothelial cell markers.¹⁰⁴ Conditional deletion of Prox1 in venous endothelium results in a severe reduction in LEC numbers.¹⁰² Moreover, LEC were shown to dedifferentiate into blood endothelial cells upon post-natal Prox1 inactivation.¹⁰⁵ Prox1 is therefore not only necessary for the specification of lymphatic lineage, but also for maintenance of lymphatic endothelial cell identity.

Sox18

Although Prox1 expression in venous endothelium is necessary for LEC specification, not all venous endothelial cells express Prox1; instead, its initial expression is clustered and polarized along the A-P axis of the cardinal vein,^{103, 106} suggesting tight regulation of its expression. Both the SRY-related gene Sox18 and COUP-TFII, which is necessary for specifying venous endothelial cell fate, were shown to regulate Prox1 (Figure 1). Sox18 is expressed by cardinal vein endothelial cells that eventually express Prox1 and form the lymphatic vasculature. Ectopic Sox18 expression in blood endothelial cells induces Prox1 and LEC marker expression; conversely, Sox18-null embryos lack LEC specification from cardinal vein endothelium.¹⁰⁷ Although Sox18 was demonstrated to induce Prox1 expression by direct binding to its promoter,¹⁰⁷ it is not sufficient to control the distinct expression pattern exhibited by Prox1, since arteries also express Sox18.¹⁰⁸

COUP-TFII

COUP-TFII potentially cooperates with Sox18 to temporally and spatially regulate Prox1 expression. In addition to being expressed by venous endothelium, COUP-TFII is also expressed by LEC¹⁰⁹, and its conditional ablation either in venous endothelial cells alone¹⁰² or ubiquitously during early development¹¹⁰ results in impaired lymphatic specification^{102, 110}. Along with Sox18, COUP-TFII binds to the regulatory regions of Prox1¹¹¹ to cooperatively induce its expression in LEC progenitors. COUP-TFII also forms a stable complex with Prox1 itself^{109, 111} to maintain Prox1 expression in LEC and regulate lineage-specific expression of markers such as Flt4/VEGFR3¹⁰⁹ and Nrp-2,¹¹⁰. Accordingly, siRNA-mediated knockdown experiments demonstrated a cell-autonomous requirement for COUP-TFII during maintenance of lymphatic identity and VEGFC-induced lymphangiogenesis.¹¹⁰

MicroRNA

MicroRNAs have also been reported to regulate Prox1 expression. For example, miR-181a binds to the 3′ untranslated region of Prox1 mRNA to inhibit its expression, thereby acting as a negative regulator of lymphatic specification and development.¹¹² Similarly, miR-31 is expressed in venous endothelial cells, and inhibits lymphatic specification by modulating Prox1 transcript and protein levels.¹¹³

VEGF-C

Following specification of LEC via induction of Prox1 expression in a subset of venous endothelium, LEC migrate out of the veins to form lymph sacs and then lymphatic vessels.¹⁰⁶ VEGF-C is necessary for the initial steps of lymphangiogenic development and, as discussed above, expression of its receptors Flt4 and Nrp-2 by LEC enables VEGF-C responsiveness.^{109, 110} Homozygous and heterozygous deletion of VEGF-C in mouse embryos results in a failure of Prox1-positive LEC progenitors to sprout, suggesting that specification of lymphatic lineage can proceed without VEGF-C, but that subsequent morphogenesis requires precise dosage of this secreted factor. VEGF-C, but not VEGF-A, can rescue the sprouting defect observed in these mutants, indicating specificity of Flt4 function during lymphangiogenesis.¹¹⁴ In support of this, several studies have demonstrated a direct association between inhibition of Flt4 function and impairment of lymphatic system development.^115–118 Several mechanisms for regulation of Flt4 expression in LEC have been proposed. As previously mentioned, COUP-TFII induces expression of both Flt4 and Nrp-2.^{109, 110} Tbx1 binds to an enhancer element in Flt4 regulatory regions to induce VEGF-C receptor expression in LEC.¹¹⁹ Tbx1 itself is required for lymphatic vessel development.¹¹⁹ The ETS factor Ets-2 interacts with Prox-1 to induce Flt4 gene expression and effectively enhance the lymphangiogenic response to VEGF-C.¹²⁰ Finally, Notch has been shown to induce Flt4 expression in endothelial cells by binding to its promoter in a complex consisting of CBF-1/Suppressor of Hairless/Lag1, allowing these endothelial cells to increase their responsiveness to VEGF-C.¹²¹ Interestingly, the Notch signaling pathway may also be active during the specification phase of lymphangiogenic development. Knockdown of Dll4, or its receptors Notch1b or Notch6, in zebrafish severely impairs LEC sprouting and lymphatic system development.¹²²

Hemangioblast vs. Hemogenic Endothelium in Blood Production

The initial but transient phase of blood development known as primitive hematopoiesis involves the emergence of aggregates of mesoderm-derived cells at E7.0–7.5 within regions of the murine yolk sac referred to as blood islands (Figure 3A). Cells within these structures develop into nucleated primitive erythroblasts and megakaryocytes, while those along the border flatten and become endothelial cells.^{123, 124} The close spatial and temporal development of blood and endothelial cells during this early phase of hematopoietic development led to the suggestion that a common bipotent progenitor known as the “hemangioblast” gives rise to both endothelial and hematopoietic cells. To date however, the existence of hemangioblasts in any species and their relative contribution to hematopoietic development remain controversial.

A second wave of blood development known as definitive hematopoiesis involves the generation of all blood cell lineages. The extraembryonic yolk sac is the first site of definitive hematopoietic development (~E8.25; Figure 3A); thereafter, multilineage hematopoietic stem/progenitor cells (HSPC) develop successively within the extraembryonic placenta (~E9.5), as well as the embryonic aorta/gonad/mesonephros (AGM) region (~E9.5–10.0), the fetal liver (E11.0–12.0), and the fetal bone marrow, where hematopoiesis persists postnatally. It is during definitive hematopoiesis that a subset of the vascular endothelium, referred to as “hemogenic endothelium”, acquires blood-forming potential and gives rise to the first multi-lineage HSPC.

The first observations suggesting that endothelial cells give rise to blood cells came from studies of developing chick embryos, wherein endothelial cells of the aorta and yolk sac were found to reside in close proximity to hemoglobin-filled progeny, which appeared to project into the blood vessel lumen and bud off into the blood plasma.^{125, 126} Many years later, in an attempt to prove this experimentally, the chick endothelium was labeled with Di-AcLDL prior to the onset of definitive hematopoiesis; dye-retaining hematopoietic clusters were demonstrated in direct physical association with the vascular endothelium of the aorta and within the blood circulation.¹²⁷ Aortic endothelial-attached hematopoietic clusters were also noted by others in both mouse and human embryos, as well as in the lumens of murine yolk sac blood vessels.^128–131 Phenotypic analysis of blood and endothelial cells revealed common expression of cell-surface markers during embryogenesis,^{127, 128, 132} providing further evidence of their coordinated development in vivo.

More recent lineage-tracing and single cell-imaging studies have provided direct evidence that multipotent hematopoietic progenitors arise developmentally from vascular endothelium. Cells initially expressing the endothelial cell marker CD31 and then blood markers CD41, Sca-1, and c-Kit were shown to emerge from the ventral wall of explanted murine dorsal aortas.¹³³ Endothelial lineage tracing also revealed hematopoietic cell clusters budding from reporter-labeled endothelium of the murine dorsal aorta, which were then found in the fetal liver and subsequently within the adult hematopoietic system.¹³⁴ Similar studies in zebrafish demonstrated early hematopoietic cells emerging from the ventral wall of the dorsal aorta, expressing definitive hematopoietic genes, and repopulating the adult hematopoietic organ.^{135, 136} These studies provide direct evidence that specialized hemogenic endothelial cells generate definitive blood cells, however we still know relatively little regarding mechanisms of their specification.

Phenotype of Hemogenic Endothelial Cells

To gain insight into the specialized functions of hemogenic endothelial cells and the molecular pathways that regulate their development, it is necessary to define their phenotype relative to non-blood-forming vascular endothelium. In vitro assays performed on sorted cell populations expressing both endothelial and hematopoietic markers have been used to try to define the phenotype of endothelial cells with hematopoietic potential. VE-cadherin⁺Ter119⁻CD45⁻ endothelial cells which co-express CD31, CD34, and Flk-1 were isolated from murine yolk sac and AGM and shown to exhibit lymphopoietic potential in colony-formation assays.¹³⁷ CD34⁺CD31⁺CD45⁻ endothelial cells from human yolk sac and AGM also display lymphomyelopoietic potential when cultured with supporting stromal cells.¹³⁰ In other studies, Flk-1⁺VE-cad⁺CD45⁻ endothelial cells were shown to generate Flk-1⁻CD45⁺ hematopoietic colonies more readily than those that were VE-cad^−.138 Although these studies suggest that expression of VE-cadherin is necessary for hematopoietic potential, it was not found to be required for hemogenic endothelial cell function in the murine yolk sac.¹³⁹ However, Hoechst dye-efflux capacity, characteristic of “side population” (SP) cells,^{139, 140} was found to be an essential characteristic of all blood-forming cells in the murine yolk sac and AGM. Furthermore, it was found that among SP cells within these tissues, Flk-1⁺c-Kit⁺CD45⁻ SP cells specifically represent hemogenic endothelial cells that exhibit multi-lineage hematopoietic potential on a clonal level¹³⁹ (Figure 3B). These cells are distinguishable from non-blood-forming vascular endothelial cells, which do not exhibit an SP phenotype and do not express c-Kit. It was further established that Flk-1⁺c-Kit⁺CD45⁻ SP cells give rise to Flk-1⁻c-Kit⁺CD45⁻ SP cells which represent multi-lineage HSPC within these tissues.¹³⁹ HSPC are distinguishable from mature blood cell types, which do not exhibit an SP phenotype and do not express Flk-1 or c-Kit, although they retain expression of CD45. Thus, flow cytometric analysis enables identification and isolation of these cell types for further investigation of function and molecular regulation.

Specification of Hemogenic Endothelial Cells

Only a small subset of the primordial endothelium within the yolk sac and AGM acquires hematopoietic potential and gives rise to HSPC. Thus, they are present in small numbers in these tissues, which limits the study of their development and function. Consequently, there is little known about the molecular mechanisms that regulate hemogenic endothelial cell specification and their generation of HSPC; nonetheless, we review what is known thus far about factors implicated in these processes, and summarize the impact of their deletion on murine hematopoietic development in Table 1.

Notch signaling

Several studies demonstrate the involvement of Notch signaling in multiple stages of blood cell development and/or function. For example, Notch1−/− mutant explants exhibit significantly impaired hematopoietic colony formation, and transplantation studies indicate that Notch1 is required for HSC activity and definitive hematopoiesis in both mouse yolk sac and AGM. Thus, Notch1 appears to be required for the generation of HSC from hemogenic endothelial cells,¹⁴⁸ as well as for the establishment of long-term, self-renewing HSC¹⁴⁹. Ablation of RBPjκ, a co-factor for Notch signaling, also leads to loss of HSPC, and loss of expression of the GATA2 transcription factor,^{150, 151} which is required for maintenance of HSC in an undifferentiated state. There may be earlier functions of Notch signaling in the specification of hemogenic endothelium, as Notch1^−/−Notch4^−/− mutants die midgestation with defects in vascular remodeling and lack of establishment of specialized endothelial cell phenotypes.¹⁵²

Runx

Also known as AML1, Runx1 has been shown to function downstream of Notch during developmental and adult hematopoiesis.^{148, 153, 154} Runx1 encodes a transcription factor that heterodimerizes with core binding factor β (Cbfβ) via its evolutionarily conserved Runt homology domain. Ablation of Runx1 results in embryonic lethality at E11.0-E12.0 due to defects in definitive hematopoiesis.¹⁵⁵ Runx1 expression by yolk sac and AGM endothelial cells is observed prior to the emergence of hematopoietic clusters^{131, 156} and is necessary for HSPC formation.^{131, 157, 158} Once the endothelial-to-hematopoietic progenitor transition is completed, Runx1 function is no longer required.¹⁵⁸ A novel role for Runx1 as a transcriptional repressor in hemogenic endothelial cells was suggested by mES cell differentiation studies, in which Flk-1-expressing cells were shown to initiate Runx1 expression, leading to silencing of the Flk-1 promoter and downregulation of receptor expression.¹⁵⁹ This is consistent with a previous study demonstrating suppression of the Flk-1 regulatory region in Runx1-expressing hemogenic endothelial cells.¹⁶⁰ Flk-1 gene silencing has been independently associated with the generation of HSPC from hemogenic endothelial cells.

Scl/Lmo2/Gata1/2 transcription complex

The regulatory regions of hematopoietic genes, including the early hematopoietic marker CD41, are simultaneously bound by Runx1 and a transcription factor complex comprised of Scl/Tal1, GATA1 and GATA2.^{161, 162} Mice deficient in Scl/Tal-1, Lmo2, or GATA1/2 all display mid-gestation lethality due to severe hematopoietic defects,^163–166 suggesting that these transcription factors have closely related roles during embryonic hematopoiesis. Morpholino knock-down of scl and lmo2 in zebrafish results in similar phenotypes, including loss of intra-aortic hematopoietic clusters and impaired expression of hematopoietic genes.^{167, 168} Embryonic stem cells deficient in Lmo2 also fail to contribute to blood cell production, or to the endothelium of large vessels in adult chimeric mice.^{169, 170} Both Scl/Tal-1 and Lmo2 are expressed by hematopoietic and vascular endothelial cells in midgestation embryos.^{168, 169, 171} Scl is necessary for the initial specification of hematopoietic lineage,^{172, 173} but dispensable for later fetal liver and adult HSC functions.^{172, 174} Expression of GATA2, which promotes the expansion of hematopoietic progenitors, is enriched within blood islands during primitive hematopoiesis.^{165, 175} This is followed by induction of GATA1 expression,¹⁷⁵ which in turn represses GATA2, leading to a developmental switch from proliferation to differentiation.¹⁷⁶ Lmo2 acts as a bridging molecule between Scl/Tal-1 and GATA-binding proteins in a transcriptional transactivating complex which has been shown to drive vertebrate hematopoietic commitment.^{167, 177} Although this complex is known to play a critical role in primitive hematopoiesis, it also likely regulates HSPC production and function during definitive hematopoiesis.

c-Myb

Another transcriptional regulator thought to play a role in the regulation of HSPC self-renewal and propagation is c-Myb. Ablation of c-Myb results in embryonic lethality at E15.5 due to failure of erythroid and myeloid development;¹⁷⁸ whereas, conditional deletion in adult bone marrow results in depletion and impaired self-renewal of the HSC pool.¹⁷⁹ Loss of function studies in zebrafish also demonstrate impaired definitive hematopoietic development, despite intact primitive hematopoeisis.¹⁸⁰ AGM explants from c-Myb^−/− mutant mice also yield fewer HSPC,¹⁸¹ which usually express high c-Myb levels that are down-regulated during differentiation. Consistent with this finding, overexpression of c-Myb results in inhibition of erythromyeloid differentiation.¹⁸

Blood Flow

In addition to soluble factors and transcriptional regulators, blood flow and hemodynamic forces also play a role in the regulation of HSC development. Endothelial cells lining blood vessels are known to respond to fluid forces through a complex of receptors capable of transducing biomechanical signals. Such signals initiate a cascade of downstream signaling events which precipitate changes in gene expression and cell behavior (reviewed in Li, Haga and Chien¹⁸³). For example, shear stress regulates endothelium-dependent nitric oxide (NO) production, which is a known regulator of hematopoiesis. A decrease in hematopoietic potential of the murine AGM is observed when the NO pathway is inhibited.¹⁸⁴ In zebrafish, inhibition of NO synthase, the enzyme responsible for NO production, leads to a decrease in transplantable HSC; Nos3^−/−knockout mice exhibit a similar phenotype.¹⁸⁵ Recapitulating the hemodynamic environment of the murine AGM in an in vitro ES cell system significantly increases the generation of CD31-expressing endothelial cells, upregulates expression of HSC-associated genes such as Runx1 and c-Myb, and increases the hematopoietic potential of differentiated CD41⁺c-Kit⁺ HSC.¹⁸⁴ Zebrafish sih (silent heart) mutants, which lack a heartbeat and an established circulation, display impaired HSC generation.¹⁸⁵ Ncx1^−/−-null embryos, which also lack a heartbeat, properly initiate HSC development within the placenta, yet also display fewer CD41-expressing hematopoietic cells than their wildtype counterparts.¹⁸⁶ Collectively, these studies suggest that blood flow forces may not be required for hemogenic endothelial cell specification, but may be necessary to promote HSC emergence and/or maintenance.