Generating Human Hematopoietic Stem Cells In Vitro: Exploring Endothelial To Hematopoietic Transition As A Portal For Stemness Acquisition

Abstract

Advances in cellular reprogramming technologies have created alternative platforms for the production of blood cells, either through inducing pluripotency in somatic cells or by way of direct conversion of non-hematopoietic cells into blood cells. However, de novo generation of hematopoietic stem cells (HSCs) with robust and sustained multilineage engraftment potential remains a significant challenge. Hemogenic endothelium (HE) has been recognized as a unique transitional stage of blood development from mesoderm at which HSCs arise in certain embryonic locations. The major aim of this review is to summarize historical perspectives and recent advances in the investigation of endothelial-hematopoietic transition (EHT) and HSC formation in the context of aiding in vitro approaches to instruct HSC fate from human pluripotent stem cells. In addition, direct conversion of somatic cells to blood and HSCs and progression of this conversion through HE stage are discussed. A thorough understanding of the intrinsic and microenvironmental regulators of EHT that lead to the acquisition of self-renewal potential by emerging blood cells, is essential to advance the technologies for HSC production and expansion.

Introduction

Advances in cellular reprogramming technologies have created alternative platforms for the production of blood cells through inducing pluripotency in somatic cells or by way of direct conversion of non-hematopoietic cells into blood cells [1–6]. Generating autologous hematopoietic stem cells (HSCs) from in vitro expandable cells that can be clonally selected, represents a promising approach for novel patient-specific gene therapies. Induced pluripotent stem cells (iPSCs) can be precisely genetically modified with designer endonucleases, including the CRISPR/Cas9 system, and subsequently single cell sorted to ensure homogeneity of genomic editing and to eliminate potential clones with deleterious off-target effects [7,8]. Alternatively, the generation of gene-targeted iPSC lines with only a single clonal event can be achieved using simultaneous reprogramming and CRISPR/Cas9 gene editing in somatic cells [9]. In addition, reprogramming non-hematopoietic somatic and mature blood cells to pluripotency and their subsequent differentiation into hematopoietic stem cells/progenitors offers a promising tool for modeling blood diseases [10], studying primitive leukemia cells [11], and drug discovery [12]. However, de novo generation of blood cells with robust multilineage engraftment potential from human PSCs (iPSCs and embryonic stem cells; ESCs) remains a significant challenge [1–3]. Overcoming the inherent limitations in this process requires a complete understanding and recapitulation in a culture dish of the process of HSC formation in the embryo. Since blood formation from endothelium represents the key event during HSC emergence in vivo, this review summarizes the historical perspectives and recent advances in the investigation of HSC formation through endothelial-hematopoietic transition (EHT), and understanding how HE and blood progenitors develop from human PSCs. Furthermore, I discuss the current challenges in de novo production of HSCs from human PSCs and emerging new concepts in blood and HSC generation by direct conversion from somatic cells.

Specification of adult HSCs during embryonic development

Elegant experiments by Dieterlen-Lie`vre and coworkers using homotypic grafting of quail embryos on chicken yolk sacs revealed embryo proper as a major source of hematopoietic stem cells (HSCs) [13,14], and eventually led to the identification of the para-aortic splanchnopeure/aorta-gonad-mesonephros (P-Sp/AGM) region as a major site of definitive hematopoiesis and HSC origin in vertebrates, including humans [15–19]. In aorta, HSCs mature in intra-aortic hematopoietic clusters (IAHC) budding from the endothelial lining of the dorsal aorta. IAHC was originally described more than a century ago by Dantschakoff [20]. Further studies by early embryologists revealed that IAHCs are restricted to the ventral portion of caudal aortic wall and could only be found at certain stages of embryonic development, but not in adults [21–25]. Based on careful microscopic analysis of serial tissue sections of IAHC, these studies provided compelling support for the idea that hematopoietic cells in aorta arise through transition of flat aortic endothelial cells into round hematopoietic cells. They also introduced the terminology hemogenic endothelium (HE) to define the specialized subset of endothelium with blood forming potential. Direct confirmation of blood formation from HE came relatively recently from lineage tracing in chicken embryo using Di-LDL labeling [26], and in mouse embryo using a VE-cadherin-driven Cre-recombinase [27]. In addition, real-time in vivo observations documented the gradual acquisition of hematopoietic morphology and phenotype by aortic endothelial cells [28–30]. Although the concept of HE was initially developed based on AGM studies, it became clear that endothelium in other embryonic and extraembryonic sites also possess hemogenic potential. Among them are the vitelline and umbilical arteries [31,32], placenta [33], head vasculature [34], endocardium [35], and yolk sac vessels [36–39].

Transient waves of hematopoiesis preceding HSC stage

While HSC function is required for establishing life-long hematopoiesis after birth, the metabolic and growth-promoting processes in the embryo are supported by transient hematopoietic progenitors that are formed prior to the emergence of HSCs (embryonic day (E) 10.5–11 in mouse). The first wave of hematopoiesis takes place in yolk sac (E7.5, in mouse) with the generation of large nucleated primitive red blood cells, macrophages and megakaryocytes [40–42]. These primitive hematopoietic progenitors are derived directly from the mesodermal precursor, hemangioblast, through endothelial intermediates [43–45]. Shortly thereafter (E8.25 in mouse), transient definitive erythro-myeloid progenitors (EMPs) are generated in yolk sac [40,46]. Cells with T and B lymphoid potentials also arise prior to HSC emergence in different hematopoietic sites including yolk sac, para-aortic splanchnopelura, viteline and umbilical arteries, and placenta (reviewed in [47,48]). Interestingly, similar to the AGM region, EMPs and lymphoid cells in yolk sac originate through EHT from endothelial cells lining nascent capillaries, arterial and venous vessels [36–39,49], thereby indicating that blood formation through endothelial intermediates is a central process during development of the entire hematopoietic system.

Hematopoietic differentiation and engraftment of human pluripotent stem cells

Over the past decade multiple studies have developed various protocols for blood induction from human pluripotent stem cells (iPSCs and embryonic stem cells; ESCs) and tested their hematopoietic engraftment in immunocompromised mice (reviewed in detail in [1–3,50,51]). Overall, these studies have revealed a relatively infrequent and low level of hematopoietic engraftment that was mostly restricted to the myeloid lineage. In addition, engraftment was usually observed following intra-bone marrow injection and was restricted to bone marrow, without a significant presence of engrafted cells in peripheral circulation. Two studies have demonstrated that hematopoietic differentiation could be accomplished in vivo during teratoma formation from human iPSCs [52,53]. The generation of blood cells within teratomata was accomplished by injecting hiPSCs together with OP9 stromal cells ectopically expressing DLL1 and WNT3a [52], or by administering human SCF and TPO to teratoma site via micro-osmotic pump [53]. Hematopoietic cells isolated from teratomata demonstrated engraftment in NSG mice with a pattern closely resembling that of hematopoietic engraftment of PSC-derived cells generated in vitro; i.e. a limited 0.1–2% hematopoietic chimerism, with predominantly myeloid engraftment.

Gaining a better understanding of why differentiated human PSCs fail to engraft requires the precise identification of the stages of hematopoietic differentiation and the phenotype of emerging hematopoietic progenitors. Although CD34 specifically identifies hematopoietic progenitors and stem cells within bone marrow and peripheral blood, it is also expressed on other cell types, including endothelial and mesenchymal cells, that are generated in hPSC cultures simultaneously with blood cells [54,55]. Our studies have revealed that hematopoietic progenitors within the CD34⁺ cell population in human PSC cultures could be precisely identified based on the expression of leukosialin (CD43), that reliably separates them from CD43⁻CD31⁺ endothelial cells and CD43⁻CD31⁻ cells with mesenchymal characteristics [54,56]. CD34⁺CD43⁺ PSC-derived progenitors are composed of two major populations: i) erythromegakaryocytic progenitors that coexpress CD235a and CD41a and ii) CD235a⁻CD41a⁻CD43⁺CD45^+/− multilineage progenitors that are lacking lineage markers and display CD90⁺CD117(c-kit)⁺CD38⁻ CD45RA⁻ phenotype [11,54,56] typical for human HSCs emerging in AGM [57] (Fig. 1). The PCS-derived lin⁻CD34⁺CD43⁺CD45^+/−CD117⁺CD90⁺CD38⁻CD45RA⁻ population possesses CFC potential similar to cord blood lin⁻CD34⁺CD38⁻ cells [58], LTC-IC potential, a high ALDH activity, the ability to efflux rhodamine-123, and is able to differentiate into all types of myeloid cells and lymphoid cells [11,54,59-61]. Recent studies also demonstrated that hESC-derived CD34⁺CD38^−/lowCD90⁺CD45⁺ hematopoietic cells express GPI-80, a marker of human fetal liver HSCs [62,63]. Despite the phenotypic similarity with HSCs, we and others failed to detect any significant hematopoietic repopulation activity following intravenous injection of these cells into NSG mice [60,62]. The one important implication of these findings is that the lin⁻CD34⁺CD43⁺CD45^+/CD117⁺CD90⁺CD38⁻ phenotype is not sufficient to detect HSCs and there is a strong need to define more specific HSC markers to aid in the identification of the conditions required for HSC induction in vitro.

Figure 1. Hematopoietic development from human PSCs.

Stages of hematopoietic development from hPSCs. The mesodermal stage of development is defined by the expression of the mesodermal markers, APLNR and KDR, with a lack of expression of typical endothelial (CD31, VE-cadeherin), endothelial/mesenchymal (CD73, CD105), and hematopoietic (CD43, CD45) markers [75,86,87]. The most primitive APLNR⁺PDGFRα⁺ mesodermal precursors have features of posterior primitive streak and capable of forming mesenchymoangiobalst (MB) and hemangioblast (HB) colonies in the presence of FGF2 [72,75,86,87]. The formation of MB and BL colonies in clonogenic medium proceeds through VE-cadherin⁺ endothelial intermediates. These intermediates undergo endothelial-mesenchymal transition (MB colonies) or EHT (HB colonies). Following EHT transition, HB colonies generate primitive hematopoietic cells with erythroid, megakaryocytic, and macrophage potentials [72,75]. Progressive mesodermal commitment to endothelial and hematopoietic cells is associated with downregulation of PDGFRα [72,137] and primitive streak genes along with upregulation of TAL1, GATA2, and ETV2 genes which lead to the formation of KDR^brightPDGFRα low/− hematovascular mesodermal precursors (HVMPs) lacking the expression of CD31, VE-cadherin, CD105, CD73 and CD43 endothelial, mesenchymal and hematopoietic markers (^EMHlin- phenotype) [72]. HVMPs lack BL-CFC potential, but are highly enriched in cells that form hematoendothelial clusters on OP9 [72]. After gaining VE-cadherin expression, cells gradually acquire an endothelial or hematopoietic cell morphology and respective gene expression profile. HE population is discriminated from non-HE and early hematopoietic cells based on the lack of CD43 and CD73 expression [72]. VE-cadherin⁺CD43⁻CD73⁻ HE express high levels of GFI1 and RUNX1 [72], similar to HE isolated from mouse embryos. The low DLL4 expression could be used to further enrich HE progenitors [73,74]. Endothelial cells with arterial and venous properties segregate into CXCR4⁺ and CXCR4⁻ fractions within CD73⁺ non-HE population [74]. The earliest hematopoietic progenitors that emerge within the VE-cadherin⁺ population display a CD43^low phenotype and retain endothelial potential and are therefore designated as angiogenic hematopoietic progenitors (AHPs) [72]. Advanced hematopoietic development is associated with the upregulation of CD43 expression and establishment of distinct subsets of CD43⁺ hematopoietic cells, including CD41a⁺CD235a⁺ erythro-megakaryocytic progenitors and lin⁻CD34⁺CD43⁺CD45^+/− multipotent myelolymphoid progenitors [54,56]. The days of differentiation are shown based on the protocols described in [54,60].

In the ongoing search for the intrinsic regulators required for HSC specification from human PSCs, several groups have performed molecular profiling studies of ESC-derived and in vivo produced primitive hematopoietic cells. These studies have revealed that the most primitive hematopoietic cells, including lin⁻CD34⁺CD45⁺ “phenotypical HSCs” generated from human ESCs, expressed a significantly lower level of genes from the HOXA cluster, but had a higher expression levels of genes from HOXB cluster [58,62,64]. Based on these findings, Ramos-Mejia et al., [64] tested the effect of HOXA9 on hematopoiesis from human ESCs. Although they found that HOXA9 overexpression increased hematopoiesis from hESCs, it failed to bestow hematopoietic engraftment. Similar results were observed by Dou et al., who found no effect of HOX5, HOX7 and HOX9 overexpression, alone or simultaneously, on the expansion and engraftment potential of hPSC-derived CD34⁺ cells [62]. Recently, Doulatov et al., revealed that a combination of HOXA9, RORA, ERG, MYB, and SOX4 conferred reproducible erythroid and myeloid engraftment of PSC-derived CD34⁺CD45⁺CD38⁻ cells with an average 2.1% hematopoietic chimerism [65]. Interestingly, despite the well established effect of HoxB4 in the induction of definitive repopulating cells from mouse ESC-derived hematopoietic progenitors [66], human studies have failed to detect a significant effect of HOXB4 on hematopoietic engraftment from differentiated ESCs [67]. This can be explained by an already high level of endogenous HOXB4 expression in human PSC-derived lin-CD34⁺CD45⁺ hematopoietic cells, in contrast to reduced HoxB4 expression in mouse PSC-derived hematopoietic progenitors. Overall, these studies indicate that a self-renewal hematopoietic program can be potentially induced from PSCs, however derivation of true HSCs with long-term and robust multilineage engraftment potential from human PSCs remains elusive.

Hemogenic endothelium in human PSC cultures

Identification of hemogenic endothelium in human PSC cultures

Since EHT represents a key mechanism of blood and HSC formation in vivo, access to HE in human PSC cultures is essential to characterize PSC hematopoiesis and interrogate the microenvironmental and intrinsic factors required for the acquisition of self-renewal potential by blood cells emerging from endothelium. Initial evidence that blood is formed in PSC cultures through EHT come from studies with mouse ESCs that demonstrated blood emergence from endothelium using continuous single-cell imaging of Flk1⁺ mesodermal cells [45,68]. These studies also found that only a portion of endothelial cells undergo EHT and revealed the need to identify HE-specific markers to facilitate HE studies in vitro. In embryo, HE can be identified morphologically as the cells lining vascular structures underneath budding blood cells. AGM versus yolk sac location could distinguish between HE associated with HSC development and HE associated with pre-HSC waves of hematopoiesis. The expression of Runx1 and Gfi1 and Runx1+23 enhancer-reporter aids in the identification of HE and emerging HSCs in mouse embryo in conjunction with stage and site-specific isolation of endothelial cells [69–71]. Since anatomical and morphologic criteria are difficult to apply to human PSC differentiation cultures, defining markers that precisely separate blood cells, non-HE from HE, and HE with primitive versus HE with definitive potential in vitro is very critical. The identification of CD43 as a marker for all hematopoietic progenitors in hPSC cultures, allowed us to separate blood cells from endothelium very accurately [54,56]. In addition, kinetic analysis of the phenotype of VE-cadherin⁺ endothelial cells following PSC differentiation, revealed that HE within the VE-cadherin⁺CD43⁻ endothelial compartment could be separated from non-HE based on a lack of CD73 expression [72]. We also found that VE-cadherin⁺CD43⁻CD73⁻ hemogenic progenitors expressed intermediate levels of CD117 (c-kit) and displayed gene expression profiles similar to HE isolated from mouse embryos, emphasized by high levels of GFI1 and RUNX1 expression in comparison to non-HE. Based on these studies in human PSC cultures, we defined HE as VE-cadherin⁺CD43⁻CD73⁻ epithelioid cells with primary endothelial characteristics that are lacking hematopoietic colony-forming potential and surface markers, but are capable of generating blood cells upon secondary culture on a feeder. We also showed that a vast array of endothelial markers, including CD34, CD31, CD146, CD201, KDR, TEK, ESAM, are shared by HE and non-HE. However, VE-cadherin⁺CD43⁻CD73⁻ HE cells lacked the expression of CXCR4 (CD184) and expressed lower levels of CLDN5 and CAV1 endothelial genes as compared to VE-cadherin⁺CD73⁺ non-HE [72]. Other recent studies have also demonstrated that a lack of/low DLL4 expression could be used to further enrich HE progenitors [73,74].

Specification and diversification of hemogenic endothelium

Identifying HE in human PSC cultures has significantly advanced our understanding of in vitro hematopoietic development. However, the process of hematopoietic specification and diversification at HE and post-HE stages remains unknown. One commonly accepted model assumes that HE possesses hemangioblastic properties, i.e. is capable of producing both blood and endothelial cells (Fig. 2A; Model 1). However, recent studies by Swiers, et al., [69] using Runx1 +23-marked HE, demonstrated at the clonal level that specification to the hematopoietic lineage in the AGM region in mouse occurs at the HE stage, and that HE cells are lacking of bipotential progenitors (Fig. 2A; Model 2). Similarly, studies with human and mouse PSCs revealed the paucity of bipotential hematoendothelial progenitors following EHT analysis at single cell level [68,72,74]. In both human and mouse clonogenic cultures, hemangioblast colonies develop from mesoderm through HE endothelial intermediates that subsequently give rise to blood cells following EHT [45,75]. Although these HE intermediates possess the ability to produce endothelial cells in vitro when collected from hemangioblast clonogenic cultures at early stages of colony formation, they are eventually replaced by blood cells when hemangioblast colony completely developed [72,75]. This observation is also consistent with the proposed model that HE in the embryo represents a transitional stage of blood specification from mesoderm and suggests that bipotentiality and the capacity of HE to form endothelial cells may solely be an in vitro phenomenon caused by placing HE in artificial conditions that do not support hematopoiesis and promote cell growth in adherent conditions. The concept of HE as a transitional stage of hematopoietic commitment in vivo, rather than a bipotential progenitor, can also explain why lineage tracing experiments have failed to reveal any clonal contribution to endothelial and hematopoietic cells in early mouse embryo [76,77].

Figure 2. Specification and diversification of HE.

(A) Models of HE specification. According to the conventional Model 1, HE cells possess hemangioblastic properties and are capable of generating non-HE and blood cells. Alternative Model 2 assumes that HE represents a transitional stage of development that is already committed to hematopoietic fate and lacking endothelial potential. This model is supported by recent functional studies in the mouse embryo. (B) Hypothetical models of blood specification and diversification from HE. Scenario 1, nascent HE produces HSCs, which eventually diversifies into major blood lineages. Scenario 2, specification to particular blood lineage occurs at HE stage. Scenario 3, HE produces nascent hematopoietic cells with a potential to form HSCs and other types of blood progenitors.

Mouse studies have revealed that in the embryo proper, HSC potential is mostly restricted to arterial vasculature [70,78,79]. The lack of venous contribution to HSCs when considered along with the shared requirements for Notch, VEGF, and Hedgehog signaling in both arterial fate acquisition and HSC formation, provided strong support for the hypothesis that arterial specification is a prerequisite to HSC formation (reviewed in [80]). Evidence also suggests that a commitment to arterial fate creates a permissive environment for HSC specification through the initiation of Notch ligands and receptors expression [80]. In human PSC cultures, Ditadi et al., [74] revealed that venous and arterial endothelial cells segregate within a CD73⁺ non-HE population and can be discriminated based on CXCR4 expression. Subcutaneous transplantation of CXCR4⁺CD73⁺ in mice led to the formation of large blood vessels lacking expression of the venous EPHB4 marker that attracted smooth muscle cells. In contrast, CXCR4⁻CD73⁺ cells produced smaller EPHB4⁺ venous vessels following transplantation. In addition, this study reported that formation of CD73⁻ HE was independent of NOTCH signaling, however NOTCH signaling was required for EHT. These findings suggested that HE cells generated from PSCs were distinct from arterial and venous lineages. However, it remains to be investigated whether the lack of certain arterial markers in PSC-derived HE reflects a distinct pathway for HE development during embryogenesis as an arterial fate-independent cell lineage or if it suggests that HE in PSC cultures is not properly specified to arterial pre-HSC stage and more closely resembles yolk sac HE. Blood formation in yolk sac does not rely on arterial signaling and was observed from arterial, venous and capillary endothelium [37]. Unfortunately, HE cells from AGM and yolk sac share the CD73⁻CXCR4⁻ phenotype [74] and any potential markers that would allow for the discrimination of yolk sac and AGM HE are currently unknown. Since the activation of Myb expression following EHT is an essential prerequisite for HSC formation from arterial endothelium [28], a MYB-reporter human ESC line was generated to track MYB-dependent hematopoiesis [81]. This study revealed weak or absent MYB promoter activation and subsequent low MYB expression at the EHT stage, thereby leading to the conclusion that HE generated from PSCs might not undergo proper HSC pre-specification. Thus, pre-specification to arterial fate and hemogenesis from arterial-fated cells remains to be explored as an alternative pathway to achieve production of HSCs from PSCs.

Heterogeneity of hemogenic endothelium in human PSC cultures

There are at least three possible scenarios for hematopoietic development from HE (Fig. 2B). In the first, HE produces HSCs that eventually diversify into main erythro-megakaryocytic (E-Mk), lymphoid and myeloid blood lineages. This scenario assumes that all HE is equal and may also align with the hypothesis that all young endothelium is hemogenic. In the second scenario, the choice to make a particular hematopoietic lineage is already made at the HE stage. This scenario assumes that there are different types of HE. Finally, in the third scenario, HE give rise to nascent blood cell which eventually acquire self-renewal program and become HSCs or specify to transient E-Mk, myeloid and lymphoid progenitors.

The assumption that all early endothelium has hematopoietic potential and all HE is equal has little support from both in vivo and in vitro studies. In human PSC cultures, the onset of endotheliogenesis is associated with the emergence of mesenchymoangioblast, a mesodermal precursor with endothelial and mesenchymal stem cell potentials [75]. Mesenchymoangioblast-derived endothelial cells in contrast to endothelial cells emerging later from hemangioblast, lack blood forming potential. In addition, we demonstrated that KDR^dim in contrast to KDR^bright mesodermal cells from human PSC cultures produce endothelium lacking blood forming potential [72]. Demonstration in an avian model of the distinct origin of endothelium with hemogenic capacity from splanchnopleura with apparent lack of hemogenic activity of somite-derived endothelium [82,83], is another strong argument supporting the concept that HE represent a cell lineage distinct from non-HE. In addition, accumulating evidence supports the concept that not all HE is equal. There are clear differences in the blood-forming activity of HE in yolk sac versus the AGM region. Even in AGM, HE is not functionally uniform. It has been demonstrated that only the ventral domain of dorsal aorta produces HSCs, while the dorsal domain produces multipotential blood cells lacking HSC potential [84]. Thus, it is very likely that HE is a heterogeneous population and that hematopoietic specification has already occurred at HE stage of development. The strongest support for this hypothesis comes from studies by Chen et al., [85] in mice, who re-introduced a Runx1 partner GFP-CBFβ on a CBFβ deficient background under control of a Tek and Ly6a regulatory cassette. These studies demonstrated that pan-endothelial expression of CBFβin Tek-expressing cells rescued erythro-myeloid progenitors in yolk sac, but not HSCs. In contrast, Ly6a-driven CBFβ expression rescued HSC formation, but not erythro-myeloid progenitors. In human PSC cultures, cells with primitive hematopoietic potential arise from hemangioblasts on day 3 of differentiation [75,86,87]. Hemangioblasts originate from APLNR⁺PDGFRα⁺ primitive posterior mesoderm expressing T, MIXL1, and FOXF1 primitive streak genes while lacking TAL1 and GATA2 expression. APLNR⁺PDGFRα⁺ mesoderm possesses a unique ability to form colonies of primitive hematopoietic cells in response to FGF2 [72,75]. Formation of blood cells from APLNR⁺PDGFRα⁺ mesoderm in a clonogenic medium proceeds through endothelial intermediates [75], which undergo EHT to generate red blood cells expressing embryonic (ε) hemoglobin, but not adult (β) hemoglobin, as assessed by flow cytometry [72]. In contrast to hemangioblasts, VE-cadherin⁺CD43⁻CD73⁻HE formed in differentiation cultures originates from hematovascular mesoderm expressing TAL1, ETV2 and high levels of GATA2, but lacks the expression of primitive streak genes [72]. VE-cadherin⁺CD43⁻CD73⁻ HE possesses broad myelolymphoid differentiation potential, including the capacity to produce T cells and CD19⁺ B cells expressing surface IgM with rearranged B cells receptor and a gene expression profile that is very similar to B cells generated from cord blood CD34⁺ cells [60,72,74,88]. At least two distinct waves of EHT were also observed by Rafii et al., [89] using continuous imaging of VE-cadherin and CD41a dual-reporting transgenic human ESCs undergoing hematopoietic differentiation. Our recent studies have also discovered that HE with distinct hematopoietic programs can be induced from hPSCs by overexpression of a specific set of transcription factors and that different groups of transcription factors are required for the induction of HE with a pan-myeloid or restricted erythro-megakaryocytic-macrophage program [90]. Overall, accumulating evidences supports the assumption that there are different types of HE. The functional heterogeneity model of HE is also compatible with the hypothesis that hematopoietic specification occurs at the HE stage [69].

It should also be mentioned that the earliest hematopoietic cells emerging in human PSC cultures have a VE-cadherin⁺CD235a⁺CD43^lowCD41a⁻ phenotype [72]. These cells have the unique property to form FGF2-dependent CFCs in the presence of hematopoietic cytokines. They also retain endothelial potential, which is why we designated them as angiohematopoietic progenitors. However, it remains unclear whether these VE-cadherin⁺CD235a⁺CD43^lowCD41a⁻ cells represent distinct subset of HE, the earliest stage of EHT, or a distinct hematopoietic lineage.

How different types of hemogenic endothelium can be discriminated

Although major progress has been made in the delineation of the HE phenotype vs non-HE in human PSC cultures, little is known about the markers that separate functionally distinct subsets of HE. The Keller group used T lymphoid potential as a marker to distinguish primitive versus definitive waves of HE. They found that inhibition of activin-nodal signaling through the addition of SB-431542 antagonist between days 2 and 3 of human PSCs differentiation, promotes the formation of VE-cadherin⁺CD34⁺ endothelial cells with T lymphoid potential [91]. Interestingly, expression of the erythroid marker CD235a (glycophorin A) in hPSC cultures is initiated on day 3 of differentiation in KDR⁺ mesodermal cells, prior to the induction of typical endothelial or hematopoietic markers [54,92]. Functional analysis of KDR⁺CD235a⁺ mesoderm by Sturgeon et al., [92] revealed that this population is enriched in hemangioblasts with primitive hematopoietic potential. Suppression of the KDR⁺CD235a⁺ population, by activin-nodal signaling inhibition or Wnt signaling activation in PSC cultures, allowed for the selection of HE with T lymphoid potential. However, further studies at the single cell level are required to determine whether T lymphoid potential in HE is linked with multilineage potential, including the capacity to produce all myeloid and T and B lymphoid lineages. Although the ability to separate primitive and definitive program based on lymphoid potential has provided an important tool to interrogate hematopoietic development in PSC cultures and advance technologies for immunotherapies, it is less clear how well lymphoid potential can predict the HSC potential of in vitro generated cells. In vivo studies have conclusively demonstrated the formation of lymphomyeloid progenitors prior to HSC emergence (reviewed in [47]). The use of lymphoid potential as a predictor of HSC potential is based on the assumption that the activation of the multipotential program is a prerequisite for establishing the self-renewal program. However, recent studies have demonstrated the functional heterogeneity of HSCs and the contribution of lineage-restricted progenitors to long-term hematopoiesis [93–96], thereby indicating that multipotentiality and self-renewal programs may not be closely connected. Thus, assessment of self-renewal and HSC potential in tedious in vivo engraftment studies remains the only reliable tool to determine whether in vitro generated cells acquire blood reconstituting potential.

Regulation of HSC formation from endothelium

Microenvironmental factors

Formation of HSCs in AGM is tightly regulated by highly coordinated local environmental signals originating from subaortic mesenchyme and neighboring endothelial and hematopoietic cells. The dynamic of interaction between HE and surrounding cells is poorly understood, and the EHT stage at which blood cells acquire self-renewal potential is unknown. Is HSC fate pre-determined at the HE stage, or does the acquisition of self-renewal potential require an interaction of emerging blood cells with their neighbors (Fig. 2B, scenario 3)? According to one hypothesis, budding of blood cells in AGM first occurs into the abluminal space followed by intravasation through a vessel in the vicinity [30]. This allows for direct step-wise exposure of emerging blood cells to signals from the subaortic mesenchyme, neighboring endothelial cells and eventually IAHCs. Elegant experiments by Richard et al., [97] in an avian model have demonstrated that prevention of subaortic mesenchyme formation through mechanical interruption of splanchnopleural mesoderm ingression underneath the aorta, blocked initiation of Runx1 expression and formation of IAHC in the AGM region without affecting vessel formation or arterial identity. In addition, these studies have revealed an important role of subaortic mesenchyme in controlling Notch signaling at the subaortic mesenchyme-HE interface. Notch signaling is required for arterial specification and HSC, but not yolk sac erythro-myeloid cell, formation (reviewed in details [80]). However, contribution of Notch signaling to HSC development is very complex. Although Notch signaling is required at the early stages of HSC specification, it must eventually be downregulated to permit HSC formation. The Daley group performed comprehensive molecular profiling studies to compare in vitro generated hematopoietic cells from mouse ESCs with cells from embryonic hematopoietic sites. These studies revealed that in vitro generated blood cells lack a Notch signaling signature which may explain their limited lymphoid and self-renewal potential [98]. Studies in human PSC cultures revealed that Notch signaling promotes blood formation from endothelium [74,99], but no engraftment was reported following Notch activation, suggesting that Notch signaling alone is not sufficient to induce HSC fate from PSC-derived HE.

Molecular profiling studies have revealed the robust proinflammatory signature of hematopoietic stem and progenitor cells in mouse AGM, and led to the identification of the critical role of interferons produced by primitive myeloid populations in HSC development [100]. Additionally, signaling through TNF-mediated NF-κB and Notch activation has been shown to be essential for HSC specification in mouse and zebra fish [101,102]. The specification of HE in AGM and yolk sac is also dependent upon retinoic acid signaling. Inactivation of the retinoic acid metabolizing enzyme retinaldehyde dehydrogenase 2 (RALDH2) abrogated development of HE and HSCs in mice [103,104]. In human PSC cultures, retinoic acid signaling promotes definitive HE and hematopoietic development and induces the expression of HOXA cluster genes, which are essential for HSC expansion during human development [62,105]. Nevertheless, modulation of retinoic acid signaling had no effect on engraftment of PSC-derived cells.

The emergence of HSCs requires SCF [106,107], Wnt [108] and BMP4 signaling [109]. SCF is highly expressed in the ventral floor of the dorsal aorta. Addition of SCF to explant cultures of the mouse dorsal aorta enhances HSC generation in the ventral compartment and induces HSC generation in the dorsal compartment [110]. BMP4 is highly expressed in the ventral mesenchyme underneath of the dorsal aorta and is regulated by FGF signaling [111,112]. In zebrafish, inhibition of FGF signaling enhances BMP4 expression and represses the expression of BMP antagonists noggin2 and gremlin1a in subaortic mesenchyme to consequently trigger the definitive hematopoietic program [112]. However, the requirement for BMP/TGFβ and Wnt signaling pathways is restricted to a narrow temporal window. Following EHT, BMP/TGFβ pathways are inhibited in IAHC [110,113] and Wnt signaling becomes dispensable for HSC maintenance [108]. Inactivation of BMP4 signaling in developing HSCs is achieved through upregulation of BMP antagonist, Noggin [110]. Expression of Noggin is potentiated by sonic hedgehog (Shh). Noggin in turn enhances Shh expression, thus creating a feed-forward loop supporting HSC maturation [110].

Recently Pearson et al., [114] explored the role of BMP4, FGF, Activin A, and VEGF in the specification of HE from Flk1⁺ mesodermal cells generated from mouse ESCs. Following culture in serum-free conditions on gelatinized plates in presence of these factors, Flk1⁺ mesodermal cells generated c-kit⁺ hemogenic cells expressing endothelial markers. When c-kit⁺ cells generated in these conditions were injected into the bone marrow of sublethally (1.25 Gy) irradiated immunodeficient NSG mice, they produced hematopoietic engraftment that was detected in peripheral blood, bone marrow and spleen at 22 weeks post-injection. Importantly, engrafted B cells had a phenotype of B2 cells, but not embryonic B1 cells. Engraftment was observed in up to 53% of mice with the highest frequency of engraftment being 21%, although the majority of mice displayed engraftment at less than 5%. Engraftment was the highest when mice injected with the earliest c-kit⁺ cells formed after the first day of culture of Flk1⁺ cells in presence of all four factors (BMP4, FGF, Activin A, and VEGF). Withdrawal of FGF2 had minimal effect on engraftment, while withdrawal of activin A significantly diminished engraftment of c-kit⁺ cells. The authors claim, that their conditions prevent differentiation and allow for the accumulation of engraftable cells due to a lack of SCF, IL3 and IL6 hematopoietic cytokines in culture. Whether similar conditions permit for the generation of engraftable cells from human PSCs remains to be determined. It is also important to elucidate the additional factors that are required for HSC maturation in vitro. For instance, HSCs from human AGM possess an unprecedented self-renewal capacity. Following transplantation in mice, a single HSC from AGM produces at least 300 daughter HSCs [19]. However, the hematopoietic repopulating potential of in vitro generated cells remains far below those observed levels.

Tissue-specific endothelial cells establish a vascular niche that participates in organ regeneration by supplying angiocrine factors (reviewed in [115]). Endothelial cells are an important component of the HSC niche. They also provide factors that aid in the specification of HE in embryo [116] and from human PSCs [89]. Overexpression of adenoviral protein E4ORF1 activates AKT, which promotes endothelial cell survival in vitro, and maintains a state that mimics niche [117]. AKT-activated endothelium provides a niche that allows for in vitro reprogramming of human umbilical vein endothelial cells (HUVECs) into self-renewing HSCs [6] and generation of engraftable cells from monkey iPSCs [118]. AKT-activated AGM-derived endothelial cells also demonstrated a capacity to amplify embryonic HSCs from mouse AGM [119]. Among multiple angiocrine factors that promote the HSC program, Notch ligands expressed by AKT-activated endothelium play the most critical role [118,119].

Extracellular matrices within AGM contribute to establishing a niche that is required for HSC specification. Interestingly, subaortic mesenchyme directly underneath of IAHC in the AGM region is highly enriched in extracellular matrix protein tenascin C [120]. We also found that tenascin C is highly expressed in overgrown OP9 stromal cells with superior hemato-inducing activity, and demonstrated that tenascin C promotes the development of hematoendothelial progenitors with broad myelolymphoid potential from human PSCs in chemically defined conditions [60]. However, tenascin C was not sufficient to endow differentiated cells with engraftment potential.

In conclusion, HSC specification from endothelium is regulated by an array of microenvironmental factors including Notch ligands, growth factors, inflammatory cytokines and extracellular matrices, which orchestrate the acquisition of self-renewal potential by blood cells following EHT. Demonstration that E9-9.5 yolk sac CD34⁺c-kit⁺ or CD34⁺c-kit⁺CD41⁺ cells lack adult repopulation potential can reconstitute multilineage long term hematopoiesis when transplanted into neonates [121–123], and that VE-cadherin⁺CD45⁻CD41⁺ cells from the E9.5 AGM can mature into HSCs via coaggregation with OP9 stromal cells [124], strongly indicate that exposure of immature hemogenic cells to the appropriate microenvironmental signaling could induce their transition into adult-repopulating HSCs. Thus, understanding the complexity and key components of signaling in the embryonic HSC niche will be essential to benefit in vitro approaches to instruct HSC fate from PSCs.

Intrinsic factors

EHT and HSC emergence in AGM is controlled by combinatorial transcription factor interactions. Runx1 is recognized as the master regulator of EHT in AGM [125]. Runx1 activates its downstream target Gfi1, which subsequently recruits the chromatin-modifying protein LSD1 to epigenetically silence the endothelial program in aortic HE, thereby allowing for the activation of the hematopoietic program [71]. Runx1 acts in combination with Gata2 to promote EHT [126]. The HE stage depends on the balance of Runx1 and HoxA3 expression. Downregulation of HoxA3 is associated with an increase in Runx1 expression in aortic endothelium, while re-expression of HoxA3 in emerging hematopoietic progenitors reactivates the endothelial program [127]. HoxA3 upregulates the transcriptional factor Sox17 [127], which plays a critical role in governing arterial and HE specification and fetal HSC emergence [128–130]. Following arterial specification, Sox17 prevents EHT and supports endothelial identity and its downregulation is required for hematopoiesis to proceed [131]. In human PSC cultures, overexpression of SOX17 in CD34⁺CD43⁻ cells promoted the formation and expansion of VE-cadherin⁺CD43⁺CD45⁻ semiadherent cells that produced blood cells following SOX17 downregulation [132]. However, this study did not report whether SOX17-induced differentiated PSCs acquired the capability to engraft in mice.

To identify the key transcriptional regulators of HE formation from human PSCs, we developed a gain-of-function screening system that employs PSCs growing as a monolayer on matrigel in mTeSR1 medium supplemented with basic fibroblast growth factor (bFGF) and stem cell factor (SCF) and thrombopoietin (TPO). In these conditions, the control hPSCs remain undifferentiated while hPSCs transduced with transcription factors obtain their differentiation phenotypes [90]. Following selection of 27 transcription factors that we found were upregulated following HE specification and EHT in PSC cultures, and testing their overexpression in our system, we revealed that none of the transcription factors alone were enough to induce blood formation. However, overexpression of ETV2 or ERG in hPSCs was sufficient to induce cells with a typical endothelial morphology, phenotype and gene expression profile. Interestingly, the addition of GATA1 or GATA2 to ETV2 led to the induction of CD43⁺ blood cells with robust pan-myeloid potential, including granulocytic, macrophage/monocytic, erythroid and megakaryocytic potentials. In addition, we were able to induce blood formation from PSCs following the overexpression of GATA2 and TAL1. However, hematopoiesis in GATA2/TAL1 induced cultures was restricted to erythroid and megakaryocytic cells with the exception of few macrophages. In both cases, ETV2/GATA2 and GATA2/TAL1, transcription factors directly converted hPSCs to endothelium, which subsequently transformed into blood cells with pan-myeloid or erythro-megakaryocytic potential, respectively (Fig. 3). These studies indicate that only a few transcription factors are sufficient to activate the hematoendothelial program from PSCs, and that specification to hematopoietic fate is initiated at the endothelial stage and is regulated by a distinct hematopoietic program. Uncovering the most critical molecular factors leading to the development of distinct HE lineages, established a novel platform to further explore the transcriptional programs required for the induction of HE with a lymphoid program and eventual HSC formation from human PSCs.

Figure 3. Direct conversion of PSCs and somatic cells to blood cells using transcription factors.

A summary of the published strategies and transcription factors used for the direct induction of blood cells from somatic cells and PSCs, self-renewing hematopoietic progenitors from PSC-derived hematopoietic cells, and conversion of committed bone marrow hematopoietic progenitors to HSCs. Abbreviations: PSC, pluripotent stem cells; PMP, pan-myeloid progenitors; EMkP, erythro-megakaryocytic progenitors; HUVECs, human umbilical cord endothelial cells; DMEC, dermal microvascular endothelial cells; E4EC, ACT-activated endothelial cells transduced with the adenoviral E4ORF1 gene; Hb, hemoglobin; CFC, colony-forming cell.

Direct induction of blood and HSC formation from PSC and somatic cells using transcription factors

The discovery of successful reprogramming fibroblasts to iPSCs [133] led to the exploration of alternative strategies for inducing blood cells and HSCs by direct conversion of somatic cells, without proceeding through a pluripotency stage (Fig. 3). The feasibility of such approach was demonstrated by several recent studies. Riddell et al., [134] showed that transient expression of six transcription factors: Runx1t1, Hlf, Lmo2, Prdm5, Pbx1, and Zfp37, endowed HSC potential onto committed lymphoid and myeloid progenitors in a mouse system. The efficiency of HSC induction was markedly increased by adding Meis1 and Mycn to the combination of transcription factors. In the human system, Sandler et al., [6] demonstrated that overexpression FOSB, GFI1, RUNX1 and SPI1 transcription factors in HUVECs or adult dermal microvascular endothelial cells followed by co-culture with AKT-activated endothelial cells, induced the HSC program. To deploy more accessible cell populations, several studies have investigated reprogramming fibroblasts to blood cells and HSCs. Szabo et al., [135] reported the generation of blood cells from human fibroblasts following overexpression of a single transcription factor OCT4. These cells produced myeloid engraftment compatible with cord blood CD34⁺ cells following transplantation into NSG recipients, and erythroid colonies expressing adult β-hemoglobin and lacking embryonic ε-hemoglobin. In mouse system, induction of the transcriptional hematopoietic program in fibroblasts was achieved with Gata2, Gfi1b, Etv6, and c-Fos [5]. Batta et al. [4] found that Erg, Gata2, Lmo2, Runx1c, and Scl reprogram mouse fibroblasts to hematopoietic colony-forming cells with erythroid, megakaryocitic and myeloid potentials. Using fibroblasts with a p53^−/− background, enabled the induction of hematopoietic cells with a broader lineage range, including T and B lymphoid potentials. Following culture on OP9-DL1, fibroblast-derived hematopoietic cells exhibited a short-term erythroid engraftment capacity in NSG mice. Interestingly, the conversion of fibroblasts to hematopoietic cells proceeded through HE and was associated with the expression of β-H1 embryonic hemoglobin in reprogrammed blood cells. These findings indicate that direct conversion of fibroblasts to blood cells even without the aid of pluripotency factors recapitulates embryonic hematopoiesis. When considered together, these studies provide evidence that direct conversion of somatic cells into HSCs is a feasible option for future clinical applications. At the same time however, they raised multiple questions. Overexpression of pluripotency factors in fibroblasts establishes an autoregulatory loop that sustains their own expression and maintains pluripotency stage [136]. The same set of reprogramming factors is effective in establishing pluripotency across different species and somatic cell types. However, there is little overlap among the transcription factors that were shown to be effective in blood and HSC induction in different systems (see Fig. 3). Are the described sets of factors for direct HSC conversion still suboptimal? Is there a unique autoregulatory loop that exists in HSCs, or is HSC maintenance strongly dependent on an undefined epigenetic program that is sustained through niche signaling? It is not entirely clear whether maintenance of exogenous factor expression is needed to sustain HSC and blood cell function following direct reprogramming of fibroblasts and endothelial cells. Since many of these factors are associated with developing leukemia, the risk of cancerous transformation remains. If direct conversion of somatic cells requires a transition through an embryonic hematopoiesis stage, how much of an advantage does it offer over blood induction from PSCs?

Conclusions and perspectives

During the last decade significant progress has been made in hematopoietic differentiation from human PSCs, including identifying the cellular pathway leading to blood development. Yet, generation of HSCs with robust and multilineage potential from human PSCs remain unaccomplished. Cell phenotypic transformation from endothelium to blood represents a unique form of morphogenesis during development. It is associated with segregation of precursors with epithelioid morphology in hematopoietic versus endothelial fate, and subsequent specification and diversification of hematopoietic lineages. Importantly, EHT establishes HSCs with self-renewal potential that sustain hematopoiesis throughout adult life. The recent precise identification of HE in human PSC cultures aids in the translation of embryonic studies to PSCs and establishes the platform for the exploration of EHT as a portal for stemness acquisition. The efficient utilization of this platform requires the identification of new HSC-specific markers. It has become increasingly clear that the commonly used lin⁻CD34⁺CD117⁺CD90⁺CD38⁻CD45RA⁻ HSC phenotype, along with ALDH expression or Rhodamine efflux, does not discriminate between iPSC-derived primitive hematopoietic cells lacking engraftment potential and true HSCs.

Over the past few years, several sets of critical transcription factors that are required for induction of the blood program from PSCs and non-hematopoietic somatic cells have been identified, and the feasibility of HSC induction from somatic cells has been demonstrated. However, a direct conversion approach remains to be more fully characterized in preclinical models and adopted to genomic footprint-free technologies and easily accessible and expandable somatic cell types. This will require improving our knowledge of the gene regulatory networks and epigenetic mechanisms that control HSC identity. Already established single cell RNAseq technologies, in conjunction with emerging methodologies for single cell epigenomics are expected to produce much needed insights into the dynamics of epigenetic and the associated transcriptional changes following EHT that trigger HSC emergence. Collectively, the efforts in the fields of developmental hematopoiesis and cellular reprogramming should provide a better understanding of HSC development and ultimately enable de novo generation of clinical grade HSCs.