Endothelial Cell Development and Its Application to Regenerative Medicine

Abstract

The formation and remodeling of a functional circulatory system is critical for sustaining pre- and post-natal life. During embryogenesis, newly differentiated endothelial cells require further specification to create the unique features of distinct vessel subtypes needed to support tissue morphogenesis. In this review, we explore signaling pathways and transcriptional regulators that modulate endothelial cell differentiation and specification, as well as applications of these processes to stem cell biology and regenerative medicine. We also summarize recent technical advances, including the growing utilization of single-cell sequencing to study vascular heterogeneity and development.

Introduction

The formation, maintenance and remodeling of a functioning vascular network is essential for embryonic development and postnatal life. During embryogenesis, vessel formation is initiated during vasculogenesis where endothelial cells emerge de novo from mesodermal precursors. This process is particularly important during vertebrate development as it lays the foundation for development of other organ systems. Ongoing studies in multiple model systems have revealed signaling molecules and transcriptional regulators critical for activating or maintaining endothelial cell gene expression, which will be discussed herein. Following de novo differentiation of endothelial cells and formation of primitive vascular plexi, there is rapid expansion and remodeling of circulatory networks to meet the growing needs of the organism undergoing morphogenesis. During this process, primordial endothelial cells are specified into arterial, venous, hemogenic and lymphatic subtypes, which are required for distinct functions throughout development and postnatally. Mechanisms that control endothelial cell fate decisions are under intense investigation, and we will summarize progress to date on the specialization of endothelial cells, per se, and their acquisition of tissue-specific phenotypes. Such studies have not only provided insights into normal development of the vascular system in vivo, but have also provided guidance for the generation of various endothelial cell subtypes from human pluripotent stem cells in vitro. Recent advances in this field will also be discussed, along with their application to tissue engineering and regenerative medicine.

Vasculogenesis: the de novo formation of endothelial cells and primitive vascular networks

Vasculogenesis is defined as the process by which vascular endothelial cells emerge de novo from the mesodermal progenitors to form primitive vascular networks that undergo rapid expansion and pattern formation. Significant progress has been made toward understanding the molecular regulation of vasculogenesis during embryonic development. And more recently, investigators have expanded studies of vasculogenesis into postnatal life and throughout adulthood, where neovascularization plays a critical role in injury repair, as well as the progression of many diseases including endometriosis, rheumatoid arthritis and Ewing’s Sarcoma [1–3]. In this review, we will predominantly focus on the regulation of vasculogenesis during embryogenesis.

Overview of vertebrate vasculogenesis.

Over the years, the regulation of vasculogenesis has been investigated in a number of model organisms. This section highlights work from two widely used vertebrate models, zebrafish and mouse.

The small size, optical clarity, short gestation, ex utero development and regenerative capacity of zebrafish have made them an increasingly powerful lower-vertebrate model organism in which to study vascular development. Zebrafish are particularly well-suited to genetic manipulation and can be used to recapitulate a number of human cardiovascular diseases, including congenital heart defects, arrhythmia and cardiomyopathy (reviewed in [4]).

In zebrafish, cell lineage tracing reveals that endothelial cells first emerge from the lateral plate mesoderm located on the ventral side of the embryo. Upon gastrulation, the lateral plate mesoderm migrates and rearranges into bilateral stripes that reside adjacent to the medial paraxial mesoderm, from which somites develop (reviewed in [5]). Around 14 hours postfertilization (hpf), cells that express endothelial markers begin to migrate medially and ventrally to the somites and notochord [6]. By 16 hpf, the first wave of endothelial cells reaches the midline and initiates the formation of the dorsal aorta. The formation of a single vascular cord can be observed by 19.5 hpf. Time-lapse, two-photo imaging revealed that the formation and fusion of intracellular vacuoles within endothelial cells drives subsequent vascular lumenization [7]. A subpopulation of endothelial cells within the developing dorsal aorta contributes to the formation of posterior cardinal vein, indicating both share a common vessel primordium [8]. Development of the dorsal aorta is completed by 22 hpf, prior to the onset of circulation. Intersegmental vessels then begin to sprout from the dorsal aorta through a process referred to as angiogenesis, and contribute to the establishment of the primary circulatory loop in zebrafish embryo (reviewed in [5]).

Over the past decades, mouse has become one of the most commonly used animal models for studying human development and disease due to the relatively high level of genetic and physiological similarity shared by mouse and human. Researchers have been able to model a number of human cardiovascular diseases in mouse including atherosclerosis, abdominal aortic aneurysms and heart failure (reviewed in [9]). Thus, despite its limitations, mouse is a popular higher-vertebrate model system for studying the process of vascular formation and remodeling.

In the mouse embryo, upon the formation of early mesoderm, a primitive vascular network begins to form in the proximal extraembryonic yolk sac around embryonic day (E) 6.5 (reviewed in [10]). Mesodermal cells give rise to endothelial precursors that form blood islands at or near the site of their origin around E7–7.5 (reviewed in [11, 12]). Endothelial cells make up the periphery of blood islands, whereas primitive hematopoietic cells form in the center. A simple capillary network, or primitive vascular plexus, is then formed via fusion and extension of blood islands [13]. All of these processes take place in the absence of hemodynamic force, although blood flow is known to be essential for vascular remodeling later in the yolk sac and embryo proper [14]. During the formation of the extraembryonic vascular network, the formation of its intraembryonic counterpart also takes place [13]. Initiated in the cranial region, endothelial precursors aggregate to form the dorsal aorta and cardinal vein, as well as vitelline arteries and veins. A rudimentary circulatory system connecting extra- and intra-embryonic vessels is established around E8.0 [15].

Transcriptional control of vasculogenesis.

Vasculogenesis is tightly regulated on the transcriptional level, and the E26 transformation-specific (ETS) transcription factors have been recognized as essential regulators of this process. To date, 19 ETS transcription factors are known to be expressed in human endothelial cells. Studies in zebrafish and mouse suggest potential redundancy in functions among different ETS factors, as germline deletion or mutation of many individual ETS genes results in little or no impairment of vasculogenesis (reviewed in [16]). However, one family member, Etv2, has emerged as a master regulator of vascular endothelial cell development [17]; disrupting its function in mouse or zebrafish leads to profound impairment of vascular development [18, 19]. Interestingly, Etv2 expression is restricted to early stages of vascular development, beginning in the blood islands of the yolk sac at around E7.0, and its expression is downregulated within the endothelial lineage by E10.5 [18, 19]. This further supports the central regulatory role of Etv2 during early vasculogenesis, and a number of recent studies have shed light on mechanisms by which Etv2 regulates vascular development (reviewed in [20]). Etv2 knockout in mice is known to be embryonic lethal and lack endothelial lineage [21]. Forced expression of Flk1 failed to rescue such phenotype, indicating that Etv2 acts downstream of Flk1. One study shows that Etv2 functions downstream of BMP, Notch and Wnt signaling to induce vascular endothelial growth factor (VEGF) receptor 2 (or Flk1) expression in mesoderm that contributes to early vasculogenesis [18]. Another study maps Etv2 global binding sites through chromatin immunoprecipitation-sequencing (ChIP-seq) analysis, revealing a number of potential downstream target genes thought to be involved in blood vessel formation including Scl, Gata2, Dll4, Notch1, Flt4, Ets1, Ets2 and Etv2 itself [22]. Current research is focusing on elucidating the roles of these newly identified target genes. For instance, one recent study showed that interaction of Etv2 and Gata2 promotes endothelial and hematopoietic lineage differentiation [23]. Another recent study demonstrated that sequential activation of Etv2 and Notch1 is required to differentiate functional arterial endothelial cells from embryonic stem cells [24].

Another group of transcription factors known to play an important role in vasculogenesis is Forkhead box (FOX) transcription factors that are recognized by their conserved ‘forkhead’ or ‘winged-helix’ DNA-binding domain. FOX factors are often involved in fine-turning the spatial and temporal expression pattern of target genes and past studies have revealed their essential roles in adult tissue homeostasis and embryonic development, including early stages of vasculogenesis [25]. More specifically, Foxf1 has been shown to be essential for the differentiation of extra-embryonic and lateral plate mesoderm. Disruption of Foxf1 in murine embryos results in embryonic lethality at midgestation due to the complete absence of vasculogenesis in yolk sac and allantois [26]. Another transcription factor from this superfamily, FoxO1, has been shown to play an essential role during subsequent steps of vessel development. The lack of FoxO1 expression in murine embryos leads to embryonic lethality beyond E10.5 due to severely underdeveloped embryonic and extraembryonic vascular structures. In vitro studies further revealed that disruption of FoxO1 results in impaired endothelial cell elongation in response to VEGF stimulation, which has been associated with arterial gene expression [27, 28]. Expression of both Gja4 and Gja5 are also dramatically decreased in FoxO1-deficient embryos; these genes encode for gap junction proteins Cx37 and Cx40, respectively, that are critical for normal arterial development and function ([29], reviewed in [30, 31]). Thus, FoxO1 may be a key regulator of endothelial cell development and arterialization.

Specification of various endothelial cell subtypes is essential for normal embryogenesis

Subsequent to vasculogenesis, a number of events take place on the molecular, cellular and tissue level to facilitate the remodeling of the primitive vasculature into a functional circulatory network. During this process, primordial endothelial cells are further specified into arterial, venous, lymphatic and hemogenic subtypes. Understanding the molecular regulation of these specification events is critical for deciphering normal vascular development, as well gaining insights that can be applied to tissue engineering and regenerative medicine.

Arteriovenous endothelial cell specification.

Arterial and venous vessels represent distinct components of the elegant hierarchical structure of the blood circulatory system. The establishment of arterial vs. venous identities takes place early during development, perhaps initiated prior to the onset of circulation. A number of signaling pathways and transcriptional regulators have been implicated in this process, including SRY-Box (Sox) genes, and VEGF/VEGFR2, Notch, Bone Morphogenetic Protein (BMP) and Transforming Growth Factor Beta (TGF-β) signaling pathways.

Three members of the SoxF subgroup of genes, Sox7, Sox17 and Sox18, have been shown to be critical for arteriovenous specification during vascular development. In vivo studies in zebrafish revealed that impaired Sox7 expression leads to shunt formation and a block in arterial development. In particular, a short circulatory loop is found around the heart due to aberrant connections between arteries and veins [32]. Other studies in mouse have shown that Sox17 is essential for the acquisition of arterial identity [33]. Sox17 is found to be selectively expressed in arteries, and endothelial-specific deletion of Sox17 leads to embryonic lethality in utero, due to impaired vascular remodeling and absence of large arteries [33]. Sox7 and Sox18 have been shown to function synergistically to control the initial establishment of arteriovenous identities and the positioning of major trunk vessels [34]. Other work demonstrated that Sox17 and Sox18 function cooperatively during early cardiovascular development in mouse embryos [35]. The Sox17/Sox18 double-null embryos display more severe defects in development of both the anterior dorsal aorta and the head/cervical vasculature, compared to Sox17 single-null embryos [35]. Sox7, Sox17 and Sox18 share conserved intron regions, suggesting that these genes have a common ancestor [36]. In addition, they are co-expressed in developing vascular endothelial cells [37–39]. Interestingly, recent studies have revealed interactions between SOX proteins and the Notch signaling pathway, which will be discussed below [32, 40, 41].

The VEGF/VEGFR2 signaling pathway has long been recognized as essential for normal vascular development, as well as vascular disease progression (reviewed in [42–44]). In addition to its roles in regulating endothelial cell survival, proliferation and migration [45, 46], VEGF/VEGFR2 signaling pathway is important during arteriovenous specification. Early studies revealed that disruption of VEGF/VEGFR2 leads to embryonic lethality around E8.5 to E12.0 [47, 48]. Other studies demonstrated involvement of this pathway in regulating early arterial differentiation. It is thought that cells exposed to high VEGF concentrations tend to differentiate toward an arterial fate, whereas others tend to differentiate toward venous fate [49]. In fact, overexpression of VEGF, or its upstream regulator Shh, leads to ectopic expression of arterial markers, and loss of either results in loss of arterial identity [49]. It is thought that during early development, endothelial cells exhibit a high level of plasticity that enables them to switch between arterial vs. venous identities, depending on their signaling environment including local VEGF concentrations (reviewed in [50]).

One common feature shared by the Sox and VEGF pathways is their downstream target, Notch signaling, which is required for arteriovenous specification during embryonic development [51]. Loss of Notch signaling in zebrafish embryos leads to loss of arterial identity, as well as ectopic expression of venous markers. In contrast, ectopic activation of Notch signaling results in suppression of venous markers [51]. One study showed that direct binding of both the RBPJ/Notch intracellular domain (NICD) and the SoxF transcription factors (Sox7 and Sox18) are required for expression of the arterial-specific Notch ligand Dll4 [40]. Another study demonstrated that loss of Sox7 function leads to aberrant vascular development, which can be rescued by overexpressing the NICD in arterial cells [32]. In addition to Sox7 and Sox18, Sox17 has also been previously linked to the Notch signaling pathway. That is, Notch signaling inhibitor DAPT suppresses the expression of Notch-dependent transcription factors (Hes, Hey1 and Hey2) without affecting Sox17 expression. However, overexpression of Sox17 in endothelial cells leads to upregulation of Notch signaling components (Hey1, Dll4, Dll1 and Notch4) and arterial markers (CXCR4 and ephrinB2), as well as downregulation of venous markers (EphB4 and COUP-TFII) [33].

Notch signaling has also been linked to the VEGF/VEGFR signaling pathway. Both Shh and VEGF were shown to act upstream of the Notch signaling; loss of Shh signaling leads to impaired arterial differentiation, which is rescued by VEGF expression [49]. Conversely, VEGF expression fails to induce arterial marker expression in the absence of Notch signaling. Another study showed that VEGF signaling induces expression of Dll4 and Notch4 through activation of MAPK-dependent ETS factors in arterial endothelium [52]. Wythe and coworkers identified a VEGF/MAPK-dependent arterial-specific enhancer of Dll4 and showed that its activities are regulated by ETS transcription factors [52]. These studies suggest a signaling hierarchy in which Shh is upstream of VEGF/MAPK, which is upstream of Notch signaling. However, whether crosstalk exists in between the Shh/VEGF/MAPK/Notch cascade and the SoxF/Notch signaling cascade remains unclear and requires further study.

One recent study has shown that fluid shear stress can maximally activate Notch signaling specifically in response to arterial shear stress magnitudes [29]. This triggers a number of downstream events including upregulation of gap junction protein GJA4 (Cx37) and cell cycle inhibitor CDKN1B (p21), which promote endothelial cell cycle arrest to enable arterial gene expression. The interdependence of endothelial cell cycle state and fate decision has been shown for both arterial and hemogenic endothelial specification [29, 53], as well as the conversion from venous to arterial endothelial cells during coronary vessel formation [54]. Although the molecular connections that link endothelial cell cycle state and phenotype need further investigation, work in stem cell models may provide clues [55, 56](50). For example, embryonic stem cells in different cell cycle phases respond differently to inductive signals (49). That is, early G1 arrest enables responses to factors that promote mesodermal or endodermal fates; whereas, ectodermal commitment occurs in late G1 (49). Arteriovenous specification may be similarly sensitive to endothelial cell cycle state, and studies on both the gene expression level and chromatin structural level will shed more light on the regulation of this process.

A number of signaling pathways have been implicated in venous specification, including BMP and Chicken Ovalbumin Upstream Promoter Transcription Factor 2 (COUP-TFII) signaling [57–60]. Wiley an coworkers showed that BMP2 signaling is essential for sprouting angiogenesis from the axial vein during zebrafish development [57]. Furthermore, the proangiogenic function of BMP2 is context dependent, which is mediated by a cargo-specific adaptor protein for Clathrin, Disabled homolog 2 (Dab2) [58]. BMP signaling is also essential for the establishment of venous identity in both zebrafish and mice, through the ALK3/BMPR1A receptor and SMAD1/SMAD5 [59]. Another signaling pathway that is important for venous specification is COUP-TFII, via inhibition of the Notch signaling pathway [60]. And Tie2 signaling has been found to regulate COUP-TFII via its downstream PI3K/Akt pathway [61].

During venous development, the formation of valves is another important step. Genes involved in lymphangiogenesis have been shown to regulate the formation and maintenance of venous valves in mice [62]. These genes include prospero-related homeobox 1 (Prox1), Vegfr3 and integrin-α9. In addition, venous endothelial cells demonstrate expression of ephrin-B2, previously thought to be a marker of arterial endothelial cells, during early stages of valve formation. Other genes implicated in venous valve formation include the gap junction genes Gja4 (Cx37), Gja1 (Cx43) and Gjc2 (Cx47) [63, 64]. These genes demonstrate polarized expression patterns around the valves, and deletion of Cx37 results in the absence of venous valves. Cx37 has also been previously shown to regulate lymphatic valve formation [65]. In addition, the histone-modifying enzyme histone deacetylase 3 (Hdac3) has also been shown to regulate the formation of both venous and lymphatic valves [66]. Thus, evidence to date suggests molecular similarities between venous and lymphatic valve formation. The latter will be further discussed in the next section.

Lymphatic endothelial cell specification.

Lymphatic vessels were initially thought to arise from endothelial cells within the cardinal vein. However, a recent study suggests that lymphatic endothelial cells arise from specialized angioblasts located in a previously uncharacterized niche within the cardinal vein, and have potential to give rise to arterial and venous endothelial cells, as well [67]. Other studies suggest that lymphatic endothelial cells arise from mesenchymal precursors in various tissues of the body, not necessarily only from cells within the cardinal vein (reviewed in [68]).

A number of factors are known to be indispensable for lymphatic endothelial cell specification, including transcriptional regulators Sox factors and Prox1. As discussed above, three members of the SoxF subgroup of genes, Sox7, Sox17 and Sox18, are important for arteriovenous specification during vascular development, and other studies have revealed their essential roles during lymphatic specification. Sox18 mutations in human can cause both dominant and recessive hypotrichosis-lymphedema-telangiectasia syndrome (HLTS) due to lymphatic dysfunction [69]. However, when investigating the role of SoxF subgroup of genes in mice, researchers should note the differences among mouse strains. For instance, Sox18 mutations in mouse result in a wide range of phenotypes, from no phenotype to lethality and complete absence of lymphatic vasculature (reviewed in [70]). This degree of phenotypic variability is attributed to strain-specific modifiers of Sox7 and Sox17 [71]. In fact, neither Sox7 nor Sox17 is expressed during normal lymphatic development, but are activated in the absence of Sox18 function in certain mouse strains.

Sox18 is considered a molecular switch that induces lymphatic specification as it directly activates the hallmark lymphatic gene Prox1 by binding to its proximal promoter [72]. Prox1 expression is required not only for the specification, but also the maintenance, of lymphatic fate [73, 74]. In fact, lymphatic endothelial cells exhibit a level of plasticity and can revert back to blood endothelial cells when Prox1 is downregulated. Tamoxifen-mediated Prox1 deletion at embryonic or postnatal stages in mice leads to reduced or absent of lymphatic endothelial markers (e.g. Slc and Pdpn) and ectopically expressed blood endothelial markers (e.g. Eng and CD34). In addition, si-RNA mediated Prox1 deletion in vitro leads to dedifferentiation of primary human lymphatic endothelial cells into blood endothelial cells [74].

One recent study showed that ERK is near the top of a signaling cascade (ERK/Sox18/Prox1) that determines lymphatic endothelial cell specification [75]. Induced activation of ERK in mice leads to upregulation of Sox18 and Prox1, which leads to enlarged lymphatic vessels, edema and Prox1 expression in blood endothelial cells. In contrast, inhibiting ERK signaling leads to reduced Sox18 and Prox1 expression, as well as reduced lymphatic endothelial cell specification. A number of additional transcription factors have also been found to play a role in lymphatic endothelial cell specification, including Coup-TFII, Protein kinase B (AKT), Notch1, as well as Wingless-Type MMTV Integration Site Family, Member 5B (Wnt5b) [67]; reviewed in [76].

As mentioned in the previous section, the formation of venous and lymphatic valves share significant similarities on the molecular level. The gap junction protein Cx37, which has been implicated in venous valve formation, is also critical for lymphatic valve formation [65]. The transcription factors Prox1 and Foxc2, together with blood flow forces have been found to coordinate the expression of Cx37 and activation of calcineurin/NFAT signaling, which are required for the formation and maintenance of lymphatic valves. Another gap junction protein, Cx43, has also been implicated in both venous and lymphatic valve formation [63, 64, 77]. Lymphatic endothelial-specific ablation of Cx43 leads to delays in the formation and maturation of lymphatic valves, as well as reduction in their total number [77]. Additional signaling pathways have also been implicated in lymphatic valve formation in recent years, including EMILIN1/α9β1 integrin interaction, Polydom/Svep1, and the mechanosensitive ion channel PIEZO1 [78–80].

Hemogenic endothelial cell specification.

Along with the specification of arterial, venous and lymphatic endothelial cells, specification of blood-forming, or hemogenic, endothelial cells is also critical for normal embryogenesis. Hemogenic endothelial cells are specified in multiple tissues throughout the course of gestation (described below and in [81]), and give rise to hematopoietic stem and progenitor cells (HSPC) that serve as the foundation of the embryonic and postnatal hematopoietic system.

Compared to our current understanding of arteriovenous and lymphatic specification, the process of hemogenic endothelial cell specification remains less clear, largely due to the inherent complexity of mammalian hematopoiesis. Mammalian hematopoiesis can be roughly divided into primitive hematopoiesis, which takes place in the extraembryonic yolk sac, and definitive hematopoiesis, which takes place in the extraembryonic yolk sac and placenta, as well as the embryonic aorta-gonad-mesonephros (AGM), fetal liver and fetal bone marrow (reviewed in [81]). Currently, the de novo generation of HSPC from hemogenic endothelial cells is thought to occur in the yolk sac, placenta and AGM; HSPC generated therein are thought to migrate to, and propagate within, the fetal liver and fetal bone marrow as those tissues develop. Thus, our understanding of the regulation of hemogenic endothelial cell specification comes largely from studies of murine yolk sac, and murine and zebrafish AGM.

In mouse yolk sac and AGM, a retinoic acid (RA)/c-Kit/Notch/p27 signaling cascade is thought to be essential for hemogenic endothelial cell specification [53, 82, 83]. RA is synthesized via the retinaldehyde dehydrogenase-2 (RALDH2)-mediated conversion of dietary retinol (Vitamin A) into its biologically active metabolite, RA [84]. Embryos lacking RALDH2 activity exhibit RA deficiency and abnormal vascular remodeling [85, 86]. In addition, loss of RA signaling impairs development of hemogenic endothelium in the yolk sac and AGM, resulting in the loss of multipotent HSPC production [53, 82, 83]. In contrast, exogenous provision of active RA to RALDH2−/− embryos in utero via maternal circulation, or directly in embryo culture, rescues the specification of hemogenic endothelium and production of multipotent HSPC therefrom.

Follow-up studies further defined the molecular signals downstream of RA, which include c-Kit, Notch and p27 [53]. Re-expression of c-Kit in RALDH2-deficient embryos not only induces Notch signaling and p27 expression, but rescues hemogenic endothelial cell development, as well. Similarly, re-expression of p27 in RA-deficient and Notch-inactivated embryos rescues the specification of hemogenic endothelial cells and the generation of HSPC therefrom. Studies in mouse and zebrafish embryos have demonstrated that Notch1 acts via Foxc2 to promote the proper development of definitive hematopoiesis [87]. In addition, the loss of Notch1 leads to increased production of hematopoietic cells via upregulation of hematopoietic transcription factors such as Runx1 and Gata2 [88]. Other genes have also been implicated in hemogenic specification including scl-β, cAMP, Wnt, TGFβ and EphrinB2 [89–93]. Despite these studies that unveil key players during hemogenic specification, a number of questions remain to be answered. For instance, how does c-Kit activate Notch signaling? What is the role of p27 and cell cycle control? In addition, following hemogenic endothelial cell specification, the generation of HSPC therefrom is a complicated process that involves multiple signaling pathways, including m6A, the hypothalamic-pituitary-adrenal/interrenal (HPA/I) stress response axis, angiopoietin-like proteins (ANGPTL2), and RUNX1 ([94–96], reviewed in [97]). Continued intense investigation, using novel in vivo and in vitro tools, will be needed to further dissect the underlying mechanisms of hemogenic endothelial cell specification and HSPC production therefrom.

Use of novel tools such as scRNA-seq can significantly add to our understanding of the developmental transition from endothelial cells to hemogenic endothelial cells to HSPC [98, 99]. For example, Baron and coworkers performed scRNA-seq on hemogenic endothelial cells, non-hemogenic endothelial cells and whole intra-aortic hematopoietic clusters isolated from mouse embryo aortas [98]. These studies revealed transcriptional regulators that are activated during the endothelial-to-hematopoietic transition (EHT). More specifically, three distinct clusters of transcription factors were found to be upregulated during early, intermediate and late stages of EHT, respectively. These three clusters also represent distinct endothelial, endothelial-to-hematopoietic and hematopoietic programs, which can be used to study the role of particular signaling pathway during EHT. Understanding the underlying mechanisms of EHT will pave the way for producing HSC ex vivo for therapeutic purposes.

Using scRNA-seq to study specialization of tissue-specific vasculature.

In addition to the specification of major endothelial cell subtypes, the phenotypic specialization of tissue-specific vasculature has attracted much interest lately. Understanding the unique functional properties of tissue-specific endothelial cells has tremendous implication for drug development. Global sequencing and microarray studies have revealed a multitude of molecule signatures of these cells, while highlighting the high level of heterogeneity [100–103].

Recent advances in scRNA-seq technology have generated unprecedented insights into the heterogeneity of a number of vascular tissues [98, 99, 104–106]. For example, in the brain, blood-brain barrier (BBB) is established in endothelial cells to create a physical barrier between circulating blood components and cells within brain tissue. Using a combination of bulk and scRNA-seq, a number of transcription factors have been identified to act downstream of the Wnt–β-catenin signaling pathway during BBB formation and maturation [104]. Among these factors are Foxf2 and Zinc Finger Protein 203 (Zic3), whose expression in human umbilical vein endothelial cells can induce the expression of BBB markers. These findings have potential implications for developing in vitro BBB models that can be used to test and optimize the uptake of therapeutics into brain tissue for the treatment of a number of neurodegenerative diseases such as Parkinson’s or Alzheimer’s disease.

In addition, scRNA-seq can be combined with spatially mapped landmark genes to enable spatial gene expression mapping. For example, Halpern and coworkers took advantage of the spatial information of hepatocytes to investigate the spatial information of liver endothelial cells (LEC), using paired-cell scRNA-seq [105]. The zonation of LEC allows the extraction of a landmark gene panel. Spatial reconstruction of single LEC reveals the spatial heterogeneity of these endothelial cells, as well as several zonated transcription factors and surface markers. Spatial reconstruction also reveals the molecular signature of pericentral LEC, which make up an important liver niche. Both positive and negative regulators of liver Wnt signaling have been identified in different subpopulations of the LEC, indicating a potential balancing effect that has been previously demonstrated in niche cells of other tissues [107–109].

Lastly, scRNA-seq has also allowed a better understanding of the endothelial cell fate decision process. For instance, Su et al. has recently uncovered the mechanism of how arteries arise from pre-existing veins during heart development in mice [54]. This process of cell fate conversion was once poorly understood due to the lack of resolution in traditional approaches. Using computational pipeline, this group was able to plot cell fate trajectory of endothelial cell subtypes. They showed that pre-artery cells differentiate from pre-existing venous cells before the onset of blood flow. These cells demonstrated not only the expression of mature arterial marker such as Cx40, but also decreased cell cycling. The vein specific transcription factor COUP-TFII, which was previously discussed in this review, has been shown to prevent this cell fate conversion by inducing cell cycle progression. ScRNA-seq has enabled the discovery of unknow cell subtypes and has contributed tremendously to our understanding of the cell fate decision process, which is critical for the development of regenerative medicine.

Regenerative medicine: recapitulation of vascular morphogenesis

Cardiovascular diseases such as coronary artery disease and heart failure are among the leading causes of mortality in the US [110], and many are associated with vascular malfunction and chronic ischemia. For the treatment of such disorders, investigators have been intensively focused on devising strategies to promote neovascularization in vivo and to engineer blood vessel replacements ex vivo. Many of these applications would benefit from the use of autologous vascular cells. Thus, numerous protocols have been developed to efficiently differentiate pluripotent human stem cells, including the potentially autologous induced pluripotent stem (iPS) cells, into endothelial cells and specialized subtypes. The recent discovery of tissue-resident endothelial stem/progenitor cells has also sparked new hope for novel adult cell-based therapies. We will review the current understanding of tissue-resident endothelial stem/progenitor cells, as well as differentiation protocols for human stem cell-derived endothelial cells and their potential therapeutic applications.

Discovery of tissue-resident endothelial stem/progenitor cells shines new light on novel cell-based therapies.

It was recently discovered that a subset of tissue-resident endothelial cells has hidden regenerative capacity in response to injury. These endothelial cells are referred to as endothelial stem/progenitor cells and are believed to retain phenotypic flexibility to adapt to normal and pathological conditions.

Recent studies describe the regenerative properties of these endothelial stem/progenitor cells [111–113]. More specifically, using an aortic endothelial injury model, McDonald and coworkers were able to characterize the cellular dynamics during the regenerative process [112]. They identified a hierarchy of endothelial cells with greater or lesser proliferative capacity within the aortic endothelium. They also discovered changes in these endothelial cells’ transcriptome profiles during the regenerative process, including upregulation of a number of stress response genes. Using DNA microarray analyses, Wakabayashi and colleagues identified CD157 as a potential marker for endothelial stem/progenitor cells [113], and single CD157+ endothelial cell are capable of forming functional arterial, venous and capillary blood vessels following experimental liver injury. Transplantation of endothelial stem/progenitor cells can represent a novel cell-based therapy for various vasculopathies. In addition, studying the transcriptome profile of these cells may potentiate their generation from human PS cells (hPSC) in vitro.

Directed differentiation and characterization of human pluripotent stem cell-derived endothelial cells.

There are a number of methods that can be used to induce differentiation of hPSC into endothelial cells, including via embryoid body formation and co-culture with feeder cells. As our understanding of vascular development has advanced, additional protocols were recently developed to eliminate the need for feeder cells [114–116].

These studies not only provide means to efficiently generate endothelial cells for tissue engineering and regenerative medicine applications, but also shed more light on our understanding of vascular development. Lian and coworkers discovered that the canonical Wnt signaling pathway is critical for the directed differentiation of endothelial cells from hPSC [114]. Using an iPSC line expressing β-catenin shRNA, they showed that the emergence of endothelial cells from hPSC is β-catenin dependent. In addition to signaling pathways, other studies focus on the microenvironment that can potentially provide cues for differentiation. Nguyen and colleagues discovered that the endothelial substratum, a specialized basement membrane, is critical for differentiation and maintenance of their phenotype [115]. More specifically, they have found that using a combination of three human recombinant laminins (LN521, LN511, and LN421) can mimic the vascular microenvironment in vivo and allow maintenance of PECAM1 and CDH5 expression in long-term culture.

The hPSC-derived endothelial cells have also been characterized in a number of ways. The expression of cell surface proteins, assessed using flow cytometry, is often used to quantify the extent of their differentiation. Cell surface proteins thought to be indicative of endothelial cell phenotype include, but is not limited to, PECAM1, VEGFR2, CDH5 and ICAM1. Functional assays can also be performed to evaluate the ability of stem cell-derived endothelial cells to form tube structures, take up acetylated low-density lipoprotein (AcLDL), establish barrier function, and undergo angiogenesis. In addition to these basic analyses, Wang and coworkers recently characterized the phenotype of hPSC-derived endothelial cells through gene and microRNA (miRNA) profiling [117]. A number of miRNAs previously associated with angiogenesis are found to be upregulated in either human embryonic stem cell (hESC)- or human iPSC-derived endothelial cells (miR-296–5p and miR-27b); however, other miRNAs are downregulated. Despite the high degree of similarity exhibited by the hESC- and hiPSC-derived endothelial cells in terms of endothelial cell marker gene expression (e.g. KDR, PDGFRA, CX3CL1), they show distinct miRNA expression profiles, indicating the inherent differences between the two pluripotent cell types and their derivatives. In addition, scRNA-seq revealed transcriptional heterogeneity among human iPSC-derived endothelial cells, marked by enrichment of CLDN5, APLNR, GJA5 and ESM1 [118]. Pathway enrichment analysis further revealed potential distinguishing characteristics of these subpopulations, such as differences in mitochondrial integrity and metabolic function, innate immune response and activation states. Knowledge of human PSC-derived endothelial cells on the transcription level can be used to improve the overall efficiency of differentiation protocols, as well as direct differentiation of specific endothelial cell subtypes.

Directed differentiation of human pluripotent stem cells into subtypes of endothelial cells.

In addition to protocols for differentiating hPSC into endothelial cells, other protocols have been developed to differentiate hPSC into various subtypes of endothelial cells. We will briefly overview the current protocols for differentiating endothelial cell subtypes, and point out their potential future applications for tissue engineering, as well as vascular grafts/co-implants.

One group recently developed a protocol to efficiently differentiate hESC into arterial and venous endothelial cells using feeder- and serum-free conditions [119]. Under chemically defined conditions, hESC are sequentially differentiated into primitive streak cells, early mesodermal cells, and endothelial progenitor cells marked by VEGFR, PDGFRβ, CD34 and CD31. At the same time, a decrease in the expression of pluripotent markers is observed (e.g. OCT4, SOX2, and NANOG). The resulting hESC-derived endothelial progenitor cells are then subjected to distinct chemically defined conditions to induce arterial and venous endothelial cell specification, as indicated by expression of arterial (NRP1, DLL4, CXCR4) vs. venous (NRP2, EPH-B4) enriched genes, respectively. In addition, transplantation of hESC-derived arterial and venous endothelial cells into immunodeficient mice can lead to formation of microvessels that are connected to the host circulation. In contrast to the immunodeficient mice injected with H1-hESCs, the mice injected with hESC-derived arterial and venous endothelial cells showed no sign of teratoma within 3–4 months of transplantation, demonstrating the stability of the phenotypes.

Other groups have developed protocols to differentiate hPSC into lymphatic endothelial cells that may potentially be used for therapeutic purposes [120, 121]. It was recently reported that when induced with VEGF-C, endothelial cells expressing CD34 and VEGFR3 express lymphatic endothelial-specific markers such as LYVE-1 and Prox-1, and form lymphatic capillary structures in 3D collagen gel or Matrigel [120]. Another group discovered that, in response to OP9 co-culture and addition of various growth factors, hPSC can be differentiated into lymphatic endothelial cells (LYVE-1+PODOPLANIN+) [121] that can enhance would healing through lymphangiogenesis and lymphvasculogenesis. Thus, use of such in vitro models of human lymphatic development can yield insights needed to optimize therapeutic potential in vivo.

Despite recent delineation of signaling pathways that promote hemogenic endothelial cell specification in vivo, and markers that characterize their specialized subtype, investigators have yet to agree upon the phenotype of human stem cell-derived hemogenic endothelial cells. A number of protocols have been developed to differentiate hPSC into blood cells via an endothelial cell intermediate; however, different cell surface markers were used to identify the blood-forming endothelial cells [122–125]. This work has nonetheless shed light on our understanding of the differentiation of hPSC toward blood cell lineages. For example, Kennedy and colleagues discovered that by differentiating hESC into embryoid bodies in chemically defined media (BMP-4, Activin A, bFGF, VEGF and hematopoietic cytokines), they can generate blood colony-forming cells [123]. They also discovered that Activin/Nodal signaling acts as a switch between human primitive and definitive hematopoiesis in vitro, where the former is dependent on Activin/Nodal signaling and the latter is not. French and coworkers obtained similar results using an OP9 co-culture system [125], and further characterized B lymphocytes derived therein from endothelial cell intermediates and found that they exhibit a high level of similarity compared to the those derived from human umbilical cord blood, suggesting they may have therapeutic potential.

Derivation of stem cell-derived, tissue-specific endothelial cells has also been the focus of many investigators, due to the potential application of such cells in tissue engineering and regenerative medicine. For example, corneal endothelial cells are essential for the maintenance of corneal dehydration and transparency. The successful generation of specialized corneal endothelial cells could potentially eliminate the need for donor cells to treat human corneal endothelial dysfunction. Zhang and coworkers co-cultured hESC with human corneal stromal cells, which promoted differentiation of hESC into periocular mesenchymal precursors [126]. Using a combination of corneal fibroblast differentiation medium and lens epithelial cell–derived medium, they were then able to differentiate periocular mesenchymal precursors into hexagonal/pentagonal-shaped corneal endothelial cells. These hESC-derived corneal endothelial cells demonstrate a function-related protein profile very similar to that of native human corneal endothelial cells and transport/pump activities in vitro and in vivo. Such results provide support for the possibility of using hPSC-derived endothelial cells for therapeutic purposes in the near future.

Summary

Delineating the molecular regulation of endothelial cell differentiation and specialization during embryogenesis provides needed insights that can be used to enhance postnatal tissue repair and treat vascular pathologies. Studies in multiple model systems have enabled the discovery of important regulators of this highly coordinated process. These studies reveal the high level of complexity of endothelial cell development. For example, there are multiple pathways commonly employed during the specification of different endothelial subtypes, yet it is unclear how they lead to distinct endpoints. The use of emerging high-throughput sequencing technologies will continue to shed light on the molecular heterogeneity of endothelial cells, even within known subtypes. Studies in human stem cell culture models will continue to work toward recapitulating these processes in vitro. Collective studies and progress, gained via in vivo and ex vivo systems, will be needed to further dissect molecular mechanisms and enable continued application to tissue engineering and regenerative medicine.

Fig1: Transcriptional Control During Endothelial Cell Specification:

Growth factors that are involved in specification of different endothelial cell subtypes (in alphabetical order).

Fig2: Differentiating hPSCs into Subtypes of Endothelial Cells:

Recent published protocols to differentiate hPSCs into different subtypes of endothelial cells.

Non-standard Abbreviations and Acronyms

AcLDL

acetylated low-density lipoprotein

AGM

aorta-gonad-mesonephros

BBB

blood-brain barrier

BMP

bone morphogenetic protein

COUP-TFII

chicken ovalbumin upstream promoter transcription factor 2

ChIP-seq

chromatin immunoprecipitation-sequencing

Dab2

disabled homolog 2

ETS

E26 transformation-specific

E

embryonic day

EHT

endothelial-to-hematopoietic transition

FOX

forkhead box

HSPC

hematopoietic stem and progenitor cells

Hdac3

histone deacetylase 3

Hpf

hours postfertilization

hESC

human embryonic stem cell

hPSC

human pluripotent stem cell

HPA/I

hypothalamic-pituitary-adrenal/interrenal

HLTS

hypotrichosis-lymphedema-telangiectasia syndrome

iPSC

induced pluripotent stem cell

LEC

liver endothelial cells

miRNA

microRNA

NICD

notch intracellular domain

Prox1

prospero-related homeobox 1

AKT

protein kinase B

RALDH2

retinaldehyde dehydrogenase-2

RA

retinoic acid

Sox

SRY-Box

TGF-b

transforming growth factor beta

VEGF

vascular endothelial growth factor

Flk1

vascular endothelial growth factor receptor 2

Wnt5b

wingless-type MMTV integration site family, member 5b

Zic3

zinc finger protein 203