|Home | About | Journals | Submit | Contact Us | Français|
Hemogenic endothelium (HE) has been recognized as a source of hematopoietic stem cells (HSCs) in the embryo. Access to human HE progenitors (HEPs) is essential to enable the investigation of the molecular determinants of HSC specification. Here we show that HEPs capable of generating definitive hematopoietic cells can be obtained from human pluripotent stem cells (hPSCs) and identified precisely by VE-cadherin+CD73−CD235a/CD43− phenotype. This phenotype discriminates true HEPs from VE-cadherin+CD73+ non-HEPs, and VE-cadherin+CD235a+CD41a− early hematopoietic cells with endothelial and FGF2-dependent hematopoietic colony-forming potential. We found that HEPs arise at the post primitive streak stage of differentiation directly from VE-cadherin-negative KDRbrightAPLNR+PDGFRαlow/− hematovascular mesodermal precursors (HVMPs). In contrast, hemangioblasts, which are capable of forming endothelium and primitive blood cells, originate from more immature APLNR+PDGFRα+ mesoderm. The demarcation of HEPs and HVMPs provides a platform for modeling blood development from endothelium with a goal to facilitate generation of HSCs from hPSCs.
Establishing a system for de novo generation of hematopoietic stem cells (HSCs) from human pluripotent stem cells (hPSCs) would open a unique opportunity to study human HSC development and provide a novel source of therapeutic cells for blood disease. Achieving this goal requires a detailed understanding of cellular and molecular pathways that lead to blood formation from hPSCs and identification of the immediate precursors of multipotential hematopoietic cells.
Avian, mouse and human embryonic studies demonstrated that definitive HSCs which give rise to all lineages of an adult hematopoietic system, are generated in the aorta-gonad-mesonephros (AGM) region and are located at the ventral aspect of the dorsal aorta (de Bruijn et al., 2002; Ivanovs et al., 2011; Pardanaud et al., 1996; Taoudi and Medvinsky, 2007). In this area, hematopoietic cells arise from a unique population of endothelial cells known as hemogenic endothelium (HE) through an endothelial-hematopoietic transition (EHT) (Boisset et al., 2010; Jaffredo et al., 2000; Zovein et al., 2008). Dynamic tracing and imaging studies conducted in vivo demonstrated that EHT represents a continuous process in which cells with endothelial characteristics gradually acquire hematopoietic morphology and phenotype (Bertrand et al., 2010; Boisset et al., 2010; Kissa and Herbomel, 2010). Definitive hematopoiesis in the AGM region is preceded by primitive hematopoiesis in the yolk sac, which initially generates primitive erythrocytes, megakaryocytes, and macrophages (Palis et al., 1999; Xu et al., 2001). The second wave of yolk sac hematopoiesis, defined as erythromyeloid hematopoiesis, is associated with expansion of erythroid precursors producing adult β hemoglobin and unilineage and multilineage myeloid precursors (Palis et al., 1999). Although the concept of HE was coined based on observations of blood formation within the aorta, it is also known that endothelium lining nascent capillaries in the yolk sac (Ferkowicz et al., 2003) and possibly vitelline and umbilical arteries (Yokomizo and Dzierzak, 2010) have the capacity to generate blood as well.
The demonstrations of HSC formation from endothelium emphasized the need for access to well-defined populations of HE cells in hPSC cultures in order to develop technologies for de novo generation of HSCs from human induced pluripotent (hiPSCs) or embryonic stem cells (hESCs). In the embryo definitive HE can be identified based on anatomical location, morphology, and expression of Runx1 (Jaffredo et al., 2010; North et al., 1999; North et al., 2002). Because these criteria cannot be entirely applied to cells differentiated in vitro, the precise identification of HE in hPSC cultures remained as a significant challenge. Although VE-cadherin+CD41a− and/or CD45− phenotype is commonly used for detection and isolation of HE, it has very limited utility in human PSC cultures since it covers the entire population of endothelial cells, does not fully exclude hematopoietic cells, and does not discriminate between endothelial lineages with primitive and definitive hematopoietic potentials. In addition, the direct mesodermal precursor of HE with definitive hematopoietic potential remains largely unknown.
In this study, we showed that HE progenitors (HEPs) can be generated from hPSCs and be identified precisely based on VE-cadherin (CD144) expression but the lack of CD73 and CD235a/CD43 expression. We demonstrated that HEPs represent a transient population of cells with the stroma-dependent capacity to generate the entire spectrum of myeloid progenitors including β-hemoglobin producing erythroid cells and pan-myeloid CFC-GEMM. In addition, we found that the earliest VE-cadherin+CD73−CD43lowCD235a+CD41a− blood cells retain endothelial potential and possess a unique FGF2-dependent hematopoietic colony-forming activity. A novel population of endothelial progenitors lacking hematopoietic potential (non-HEPs) was distinctively recognized by the expression of CD73 and a high level of CD117, i.e. VE-cadherin+CD73+CD235a/CD43−CD117high phenotype. VE-cadherin+CD73−CD235a/CD43− HE cells originated from VE-cadherin-negative KDRbrightAPLNR+PDGFRαlow/− hematovascular mesodermal precursors (HVMP), which were highly enriched in cells forming hematoendothelial clusters on OP9 stromal cells. These progenitors were distinct from the more primitive APLNR+PDGFRα+ mesoderm that contained a population of hemangioblasts (HBs), which have the capacity to form colonies composed of primitive type blood cells through endothelial intermediates in serum-free semisolid medium.
To characterize the development of various mesodermal lineages, we employed the hPSC differentiation system in coculture with OP9 (Choi et al., 2009a; Vodyanik et al., 2006; Vodyanik et al., 2010). In these cultures, we have previously identified CD43 as a marker for hPSC-derived progenitors that have the potential to form hematopoietic cytokine-dependent colonies in semisolid medium and demonstrated that CD43 expression separates hematopoietic cells from endothelial cells (Choi et al., 2009a; Vodyanik et al., 2006). To investigate the developmental steps immediately preceding the formation of CD43+ blood cells and map the diverging point of hematopoietic and endothelial cell lineages, we analyzed the kinetic expression of various endothelial markers following H1 hESC differentiation in OP9 coculture. The first cells expressing VE-cadherin endothelial marker (Breier et al., 1996) were detected by day 4 of differentiation (Figures 1A and S1A). Upregulation of VE-cadherin expression on differentiated hESCs in OP9 coculture coincided with the expression of another endothelial marker CD31 (PECAM) (Figure 1A). Interestingly, cells expressing CD235a (Glycophorin A), a hematopoietic marker of erythroid lineage, could be detected within the first emerging VE-cadherin+ cells (Figures S1B). On the next day (day 5) of differentiation, the number of VE-cadherin+ cells and the proportion of CD235a+ cells within this population substantially increased. All of the VE-cadherin+CD235a+ cells were negative for CD41a (abbreviated as V+235+41− cells) on day 4 of differentiation. However, on day 5 of differentiation, a small proportion of CD235a+ cells coexpressing CD41a (V+235+41+ cells) could be detected (Figures 1B and S1B). Although V+235+41+ cells expressed a high level of CD43 which defines hematopoietic commitment (Vodyanik et al., 2006), expression of CD43 in V+235+41− cells was relatively low and was best detectable with antibodies conjugated with APC or PE (Figure S1C). Thus, we combined CD235a and CD43 antibodies in our studies to achieve optimal pan-hematopoietic detection at all stages of differentiation.
Phenotypic analysis of day 5 VE-cadherin+ cells revealed almost uniform expression of CD31, KDR, CD34, CD201, ESAM, and CD146 endothelial markers by these cells. However, we noticed that another typical endothelial marker, CD73 or 5′-nucleotidase (Thomson et al., 1990), was expressed only in 20-60% of total VE-cadherin+ cells almost exclusively within the 235a/CD43− population (Figures 1B, S1B and S1D). This observation led us to identify three distinct major subsets within emerging VE-cadherin+ cells: V+235+41−, V+73+, and V+73−235− (Figure 1B and Table 1). Kinetic analysis revealed that V+73−235−cells represent a transient population that develops during the earliest stages of endothelial commitment, but is mostly lost within the next 3 days of differentiation. The V+73+ population was minor at onset of endotheliogenesis, but gradually increased with advanced differentiation. The proportion of VE-cadherin+ cells expressing 235a and/or CD43 hematopoietic markers peaked on day 5 of differentiation and then decreased (Figure S1B).
As demonstrated in Figure 1B, all three major VE-cadherin+ cell subsets had very similar endothelial phenotype and were capable of acetylated low-density lipoprotein (AcLDL) uptake, indicative of endothelial function. However, we noticed that expression of CD117 (c-Kit), a marker for early stage angiohematopoietic progenitors, was highest in V+73+ cells, while its expression was almost undetectable in V+235+41− cells. V+73−235− cells expressed an intermediate level of CD117 (Figure 1B). We also found that in contrast to other day 5 VE-cadherin+ subsets, V+73+ cells lacked the expression of CD226 (DNAM-1), a cell surface marker typically found on the hematopoietic cells (Kojima et al., 2003; Shibuya et al., 1996). Morphologically, the V+235+41− population consisted predominantly of cells with a high nuclear-cytoplasmic ratio typical for immature hematopoietic cells. In contrast, almost all V+73+ cells had characteristic endothelial morphology. V+73−235− cells had an intermediate morphology that resembled both V+235+41− and V+73+ cells, i.e. pale blue cytoplasm similar to endothelial cells, but higher nuclear-cytoplasmic ratio similar to immature hematopoietic cells (Figure 1B).
To fully analyze the differentiation potential of each newly discovered VE-cadherin+ cell subsets, they were isolated using FACS and cultured in endothelial conditions and assayed for hematopoietic colony-forming activity (Figure S2). As shown in Figure 1B, the three major day 5 VE-cadherin+ subsets (V+73−235−, V+235+41− and V+73+ cells), but not the minor V+235+CD41+ subset, formed a monolayer of adherent cells with endothelial morphology when cultured on fibronectin in endothelial growth medium. Consistent with their endothelial nature, these cells expressed VE-cadherin, took up AcLDL, and formed vascular tubes in Matrigel matrix. In contrast, hematopoietic colony-forming (CFC) potential was detected almost exclusively within V+235+41− and V+235+41+ cells. Although the hematopoietic CFC potential of V+235+41− cells in standard serum-based CFC medium was low and mostly restricted to small CFC-E, we found that the number and spectrum of hematopoietic CFCs was markedly increased in serum-free medium containing FGF2, SCF, EPO, IL-3, and IL-6. In the serum-free conditions, day 5 V+235+41− cells formed large erythroid, megakaryocyte, myeloid, and mixed colonies composed of erythroid cells, macrophages and megakaryocytes indicating that emerging blood cells expressing CD235a erythroid marker had multilineage potential (Figure S3A and S3B). To define which growth factors are required for V+235+41− cells to form hematopoietic colonies, we eliminated each cytokine individually from clonogenic cultures. These experiments demonstrated that both FGF2 and EPO were essential for the development of large CFC-E and CFC-Mix (Figure S3C). The removal of SCF almost entirely abrogated CFC-Mix, but had little effect on large CFC-E. Myeloid colonies required IL-3 and FGF2 for optimal development. The day 5 V+235+41+ cells formed predominately CFC-E and -Mix, however they had downregulated expression of APLNR and TEK and failed to grow into endothelial cells in endothelial conditions indicating that the acquisition of the CD41a expression was associated with the complete loss of endothelial potential (Figure 1B and S1D).
Hemogenic potential of embryonic endothelial cells can be identified in culture with bone marrow stromal cells (Nishikawa et al., 1998; Oberlin et al., 2002), thus we cultured day 5 VE-cadherin+ subsets on OP9 separately. In these conditions, both V+73−235− and V+235+41− cells generated CD31+CD43/45− endothelial cells and a significant amount of CD43+ blood cells (Figure 2A). The CD43+ cells consisted of CD235a/CD41a+ erythromegakaryocytic cells and CD235a/CD41a−CD45+/− multipotent hematopoietic progenitors (MHP) which we typically observe from hESCs differentiated on OP9 for 8-9 days (Vodyanik et al., 2006). These day 8-9 CD235a/CD41a−CD45+/− MHP express CD34 but lack CD38 and other hematopoietic lineage markers, i.e. they have a lin−CD34+CD43+CD45+/− CD38− phenotype (Vodyanik et al., 2006). Although both V+73−235− and V+235+41− cells generated a broad range of hematopoietic colonies in standard serum-containing CFC medium after culture on OP9, V+73−235− cells formed a higher numbers of myeloid colonies, including large multicentric pan-myeloid GEMM colonies. In contrast, V+73+ cells formed mostly CD31+CD43/45− endothelial cells with very few hematopoietic cells (Figure 2A).
To determine the frequency of progenitors with hematopoietic and endothelial potential of each subset of day 5 VE-cadherin+ subset, we performed a single cell deposition assay. As shown in Figure 2B, V+73+ single cells generated only endothelial clusters on OP9 with a frequency of about 1/5, while V+235+41− cells formed predominantly hematopoietic cell clusters. Although the majority of V+73− 235− cells gave rise to either hematopoietic or endothelial clusters, 2.5% of them had the potential to form hematoendothelial clusters, indicating the presence of bipotential progenitors within this population.
Based on the functional and phenotypical properties of each VE-cadherin+ subset, we defined them as the following: 1) V+73−235− are HEPs that have primary endothelial characteristics lacking hematopoietic CFC potential and surface markers, but expressing an intermediate level of CD117 and are capable of generating blood and endothelial cells upon coculture with stromal cells; 2) V+235+41− are angiogenic hematopoietic progenitors (AHPs) that possess primary hematopoietic characteristics but are capable of generating endothelial cells; 3) V+73+ are non-HEPs that have all functional and molecular features of endothelial cells, form endothelial colonies on OP9, and express high level of early progenitor marker CD117 (see also Table 1).
Molecular profiling studies revealed a high similarity between the day 5 VE-cadherin+ population subsets. All subpopulations of these cells expressed the typical endothelial genes, TFP, HIF1A, AAMP, F2R, EDF1, and PROCR, and the genes associated with angiohematopoietic and HSC development, FLI1, TEK, LMO2, TAL1, RUNX1, CBFB, PBX1, PTEN, and TCEA1. However, V+235+41− AHPs expressed higher levels of hematopoietic-specific genes and lower levels of the typical endothelial (CAV1, CTGF, APOLD1, and AMOT) and endothelial junction (CDH5, CDH2, and CLDN5) genes (Figure 3A). In contrast, V+73−235− HEPs expressed higher levels of the endothelial genes, CLDN5, CAV1, and MMRN1N, and lacked the expression of hematopoietic genes. In comparison with the HEPs, the V+73+ non-HEPs expressed higher levels of the endothelial genes EMCN, CAV1, CXCR4, CLDN5, and COL15A1 (Figure S4A). Genes found to be more highly expressed in HEPs versus non-HEPs included NTS neurotensin, BMPER an endothelial regulator that controls BMP4-dependent angiogenesis (Heinke et al., 2008), and SMAD6, a negative regulator of BMP signaling (Ishida et al., 2000) and RUNX1 activity (Knezevic et al., 2011).
To identify the direct mesodermal precursor of HE cells, we analyzed the expression of mesodermal markers APLNR (D’Aniello et al., 2009; Vodyanik et al., 2010), KDR (Shalaby et al., 1997), and PDGFRα (CD140a) (Kataoka et al., 1997) in differentiated hESCs before the first VE-cadherin+ cells could be detected. This analysis revealed the population of KDRbrightAPLNR+ cells which was initially detected on day 3.5 of differentiation (Figure 4A) immediately preceding the formation of the first VE-cadherin+ cells in hESC/OP9 coculture (Figure 1A). Emerging day 3.5 KDRbrightAPLNR+ cells essentially lacked the typical CD31, VE-cadherin endothelial, CD73, CD105 mesenchymal/endothelial, and CD43, CD45 hematopoietic cell markers (here on referred to as EMHlin− cells), however the early VE-cadherin+ cells became clearly detectable within this population from day 4 of differentiation (Figure 4B). Flow cytometric analysis of day 4 VE-cadherin− KDRbrightAPLNR+ cells revealed that they maintain EMHlin− phenotype (EMHlin− KDRbrightAPLNR+). However, in contrast to the more primitive day 2 and day 3 APLNR+ (Vodyanik et al., 2010) and day 4 KDRdim mesodermal cells, day 4 EMHlin− KDRbrightAPLNR+ cells had downregulated expression of PDGFRα (Figure 4B). Although these day 4 EMHlin−KDRbrightAPLNR+PDGFRαlow/− (KbrA+P−) cells lacked the most specific endothelial markers VE-cadherin and CD31, they expressed other markers typically found on endothelial cells including TEK, CD34, CD201, and CD146 (Figures 4B and S5A) suggesting that these mesodermal cells could be direct precursors of endothelial progenitors in hESC cultures. To confirm our hypothesis, we isolated day 4 KbrA+P−, KDRdim and KDR− cells using flow cytometry (Figure 4C) and cultured them on OP9.
As shown in Figure 4C, after 5-6 days culture on OP9, only KbrA+P− cells generated both CD31+CD43/45− endothelial cells and CD43/CD45+ hematopoietic cells, while KDRdim cells predominantly generated CD146+CD31− mesenchymal cells, few endothelial cells, and almost no blood cells. KDR− cells lacked hematovascular potential completely. Importantly, day 4 KbrA+P− cells generated V+73−235−, V+235+41− and V+73+ subsets we observed on day 5 of primary hESC/OP9 coculture (Figure 4C). It should be also noted that KbrA+P− cells were multipotential and capable of differentiating into CD146+CD31− mesenchymal cells in addition to blood and endothelial cells (Figure 4C). Double staining of KbrA+P− cells grown on OP9 with VE-cadherin and CD43 antibodies revealed that they form HE clusters, i.e. sheets of endothelial cells generating non-adherent blood cells (Figure 4D). Morphological examination of HE clusters at different stages of development revealed that endothelial cells within these clusters gradually transitioned into hematopoietic cells. During the early stages of transition, VE-cadherin+ cells had upregulated CD43 expression and transformed from a cuboidal to a round cell morphology (Figure 4D, Movie S1). Single cell deposition experiments demonstrated that KbrA+P− cells formed HE clusters at a high frequency (about 1 in 10 cells), strongly indicating that these cells represent the direct precursors of HE (Figure 4E).
To confirm that KbrA+P− cells were direct precursors of the HEPs, we isolated these cells from day 3.5 of hESC/OP9 cocultures, before VE-cadherin+ cells became detectable (see Figure 1A), and recultured them on OP9 for 2 days. Flow cytometric analysis of these KbrA+P− secondary cultures revealed that they had upregulated VE-cadherin expression and differentiated into V+73−235−, V+235+41− and V+73+ subsets we observed on day 5 of primary hESC/OP9 coculture (Figure S5B). When these subsets were isolated from the secondary cocultures by FACS and analyzed for endothelial and hematopoietic potentials, we found that the V+73−235−, V+235+41− and V+73+ cells generated from isolated day 3.5 KbrA+P− cells had the same hematopoietic and endothelial differentiation potentials as the primary day 5 HEP, AHP, and non-HEP subsets, respectively (Figure S5C).
Morphologic analysis revealed that KbrA+P− cells were large blast-like cells, noticeably different from KDRdim and KDR− cells which had a more abundant and vacuolated cytoplasm (Figure 4C). Molecular profiling studies revealed that in KbrA+P− cells, expression of transcriptional regulators of hematopoietic and endothelial development, LMO2, TAL1, CBFB, GATA2, and FLI1, was upregulated, while expression of the primitive streak genes, MIXL1, EOMES, T, and MESP1, was downregulated (Figure 3A). However, these cells retained high expression levels of genes representing lateral plate/extraembryonic mesoderm (FOXF1, BMP4, and WNT5A). Based on their phenotypic features, gene expression profile, morphology, and functional properties we designated KbrA+P− mesodermal cells as hematovascular mesodermal precursors, (HVMP). These precursors may resemble embryonic angioblasts which are defined as cells that have not yet formed a lumen and express certain, but not all endothelial markers. They are committed to differentiate into endothelial cells and give rise to vascular primordia (Risau and Flamme, 1995).
Blast CFCs (BL-CFCs) were identified by the Keller group as progenitors that generate blast colonies composed of cells with hematopoietic and endothelial potential (Choi et al., 1998). Widely referred to as hemangioblasts (HBs), BL-CFCs represent the earliest cells with detectable hematopoietic potential in mouse and human ESC differentiation systems (Choi et al., 1998; Kennedy et al., 2007). However, the exact position of HBs (BL-CFCs) within the hierarchy of human angiohematopoietic cells and their developmental potential remains unclear. BL colonies were reported from differentiated hESCs at early stages of mesodermal development (Davis et al., 2008; Kennedy et al., 2007) as well as from cells at more advanced stages of differentiation including cells already expressing endothelial markers (Lu et al., 2008; Zambidis et al., 2008). Moreover, two types of HB colonies have been recently described, one with and one without myeloid potential (Kennedy et al., 2007).
Previously we demonstrated that BL-CFCs arise from day 2-3 EMHlin− APLNR+PDGFRα+ (A+P+) mesodermal population which expresses genes associated with primitive streak and lateral plate/extraembryonic mesoderm development reminiscent of primitive posterior mesoderm (PM) in the embryo (Vodyanik et al., 2010). We showed that BL-CFCs could be detected using serum-free FGF2-containing clonogenic medium (Vodyanik et al., 2010). Here, we also found that the number of BL-CFCs could be increased by adding APLNR ligand apelin-12 to the clonogenic medium (Figure 5B). Although hematopoietic cytokines are commonly added to BL-CFC clonogenic medium, we avoided their use in our assay to increase its specificity by eliminating false-positive results due to the detection of hematopoietic progenitors. Using optimized BL-CFC-specific assay with FGF2 and apelin-12 we detected BL-CFC activity almost exclusively in day 3 A+P+ cells (Figure 5F), indicating that HBs (BL-CFCs) are distinct from day 4 HVMPs and day 5 HEPs and AHPs.
To characterize the developmental potential of BL-CFCs, we analyzed the mature BL colonies using flow cytometry and hematopoietic CFC assay. As shown in Figure 5A, HB (BL) colonies collected on day 12 of clonogenic culture consisted almost entirely of CD235a and/or CD41a expressing cells with erythroblast morphology. In contrast to erythroid colonies generated from V+73−235− HEPs, BL-CFCs expressed no adult β hemoglobin (Figure 5C). The replating of 20 individual blast colonies into serum-free hematopoietic clonogenic medium demonstrated that they could give rise to erythroid, megakaryocyte, macrophage colonies and mixed colonies composed of all three cell types (Figures 5D and 5E). When a pool of 200 blast colonies was collected, we were able to detect the same spectrum of hematopoietic CFCs. The spectrum of hematopoietic CFCs was similar when BL-CFCs were replated into standard serum-containing hematopoietic CFC medium, although we observed a reduction in number of erythroid colonies and a slight increase in macrophage colonies (not shown). These results indicate that BL-CFC hematopoietic potential is mostly restricted to primitive cells of erythromegakaryocytic and macrophage lineages.
As previously demonstrated, BL-CFCs represent single cell-derived clonogenic progenitors, which generate hematopoietic cells through the formation of an endothelial intermediate (Lancrin et al., 2009; Vodyanik et al., 2010). This transitional intermediate appears as a core-like structure that forms during the first 3 days of clonogenic culture (Figure 5A) and distinguishes HB colonies from FGF2-dependent hematopoietic colonies formed from day 5 AHPs (Figure S6A). HB cores were formed by epithelioid cells, which stained positively for VE-cadherin. However, in contrast to membranous VE-cadherin expression typically seen in mature endothelial cells, HB cores cells showed predominantly cytoplasmic with limited membranous VE-cadherin expression. When HB cores were cultured in endothelial conditions, up to 95% of them generated typical VE-cadherin+ endothelial clusters that were capable of efficient AcLDL incorporation. When HB cores were collected and cultured on OP9 with hematopoietic cytokines, they generated hematoendothelial clusters (Figure 5A).
Molecular profiling studies demonstrated that HB cores had a gene expression profile very similar to day 5 HEPs, although the HB cores had much lower expression of RUNX1 gene associated with definitive hematopoiesis, compared to day 5 HEPs (Figure 3A and S4B). Together, these studies indicate that BL-CFCs originate from the more primitive A+P+ PM and reflect the first wave of yolk sac hematopoiesis which proceeds through the endothelial intermediate stage with restricted erythroid, megakaryocytic and macrophage potential.
To find out whether the hematopoietic potential of A+P+ PM is restricted to primitive HB-derived hematopoiesis or whether these cells contain precursors of definitive angiohematopoietic progenitors, we isolated day 2.5 A+P+ cells and recultured them on OP9. At this stage, KbrA+P− cells were not detected. As shown in Figure S6B, A+P+ cells rapidly expanded on OP9 and generated the entire spectrum of angiohematopoietic and hematopoietic progenitors, which we typically observed in primary hESC/OP9 coculture. These data indicate that A+P+ PM contains precursors for both primitive and definitive hematopoiesis. Although maturation of primitive angiohematopoietic progenitors was achieved in serum-free semisolid medium in presence of FGF2, stromal factors were essential for the maturation of definitive type angiohematopoietic progenitors from A+P+ PM.
To determine whether other hPSC lines follow similar patterns of hematoendothelial differentiation as we observed with H1 hESC, we analyzed development of newly identified subsets of angiohematopoietic progenitors from transgene-free fibroblast-derived hiPSCs (Yu et al., 2009) and H9 hESCs. As shown in Figure S7, all examined hPSC lines formed phenotypically and functionally similar subsets of progenitors, including day 4 HVMPs, day 5 HEPs, AHPs, and non-HEPs.
During the last decade, significant progress has been made in identifying the major stages of hematopoietic development from hESCs/iPSCs (Kennedy et al., 2007; Vodyanik et al., 2006; Zambidis et al., 2005) and their differentiation toward particular blood lineages (Choi et al., 2009b; Lu et al., 2008; Olivier et al., 2006; Timmermans et al., 2009; Woll et al., 2005). However, the development of cells with hematopoietic reconstitution potential from ESC/iPSCs remains a challenge. Although several studies have shown the bone marrow engraftment of differentiated human ESCs and iPSCs, the engraftment rates were low and mostly restricted to myeloid cells (Ledran et al., 2008; Lu et al., 2009; Narayan et al., 2006; Risueno et al., 2012; Tian et al., 2006; Wang et al., 2005). The most likely explanation for these findings is that in vitro conditions do not support HSC formation from its direct HE precursor. Thus, access to well-defined population of HE cells is essential for developing an in vitro system for the identification of the critical factors that control maturation of engraftable hematopoietic cells from endothelium.
Previous studies demonstrated that cells expressing endothelial molecules differentiated from mouse and human ESCs can generate blood cells (Eilken et al., 2009; Hashimoto et al., 2007; Nishikawa et al., 1998; Vodyanik et al., 2006; Wang et al., 2004). It has been also shown that HE can be prospectively separated from non-HE in mouse ESC cultures based on the activity of Flk1 promoter/enhancer (Hirai et al., 2003). Here we demonstrated for the first time that the CD73 phenotypic marker can be used to separate prospectively HE cells from non-HE progenitors. Importantly, we also found that the VE-cadherin+CD41a−CD45− population in hPSC cultures includes CD235a+ (Glycophorin A+) hematopoietic progenitors, which retain angiogenic potential. Based on these findings we were able to further specify the phenotype of HEPs as VE-cadherin+CD73− CD235a/CD43− and demonstrate that HEP represented a transient population of endothelial cells emerged immediately after the beginning of endotheliogenesis in hPSC cultures and rapidly declining within next 3 days of differentiation. These HEPs had the potential to generate β-hemoglobin-producing red blood cells and the entire spectrum of myeloid progenitors including pan-myeloid GEMM progenitors, which has been identified within human embryonic tissues, but not yolk sac (Hann et al., 1983; Huyhn et al., 1995). Although we and others already demonstrated that CD34+CD43+ progenitors generated in hPSC/OP9 coculture have T and B lymphoid potential (Carpenter et al., 2011; Timmermans et al., 2009; Vodyanik et al., 2005; Vodyanik et al., 2006), further studies will be required to prove that CD34+CD43+ cells with lymphoid potential arise directly from HEPs.
By analyzing the expression of mesodermal markers at stages preceding endotheliogenesis, we identified EMHlin−KDRbrightAPLNR+PDGFRαlow/− HVMPs as the direct precursors of definitive type of HEPs. The HVMPs and HEPs required stromal factors for hematopoietic development and were distinct from HBs which arise from day 3 A+P+ PM cells and can be specifically detected in serum-free semisolid medium in the presence of FGF2 and apelin-12. The hematopoietic potential of HB colonies detected using these conditions was mostly restricted to cells of erythromegakaryocytic lineage reflecting the first wave of hematopoiesis observed in the yolk sac. These results are also consistent with mouse studies which demonstrated that Flk1-positive hemangioblastic cells are mainly primitive hematopoietic cells (Fehling et al., 2003). Hematopoietic cells within HB colonies arise through core-forming VE-cadherin+ cells, which in contrast to definitive angiohematopoietic progenitors develop in serum-free medium without stromal support. HB cores have endothelial gene expression profile and potential. However, our finding that HB cores express intracellular rather than membranous VE-cadherin indicates that they are different from definitive HEPs and may represent distinct type of immature cells of endothelial lineage which are more similar to angioblastic mesodermal cells than to more mature endothelial cells which line already established blood vessels. VE-cadherin+ cells that co-express CD31, CD34, CD105, and TEK endothelial markers were identified within subset of Flk1-positive cells in the extraembryonic mesoderm region during gastrulation and yolk sac blood islands of the E7.0-7.5 mouse embryos (Ema et al., 2006; Yokomizo et al., 2007). These cells are capable to generate endothelial and primitive blood cells. Because HB cores have similar phenotypic and functional characteristics they could be equivalent to the VE-cadherin+ cells detected within Flk1+ extraembryonic compartment.
The present study revealed a unique population of AHPs expressing VE-cadherin and glycophorin A (CD235a) erythroid marker but lacking CD41a. These AHPs represent multipotential hematopoietic progenitors, which similar to BL-CFCs (HBs) require serum-free conditions and FGF2 for colony formation. However, in contrast to BL-CFCs, the development of colonies from AHPs depends on hematopoietic cytokines and does not proceed through an endothelial core stage. Another unique feature of AHPs is their angiogenic capability, which is completely lost in CD235a+CD41a+ cells that arise at later stages of differentiation. FGF2- and hematopoietic cytokine-dependent colonies with and without endothelial potential have been described in mouse yolk sac, fetal liver and AGM (He et al., 2010; Yao et al., 2007), indicating that AHPs may have in vivo counterparts. Given the fact that AHP cells express a definitive hematopoiesis transcriptional factor RUNX1 (Figure 4 and S4B), and possess erythroid, uni- and multilineage myeloid differentiation potential, they may represent precursors for a transient wave of definitive erythromyeloid hematopoiesis similar to the one described in mouse yolk sac (Palis et al., 1999).
In addition our studies provided important insight on endothelial development from hESCs. Although one commonly held view implies that all endothelial cells in PSC cultures originate from HBs, our current and prior studies (Vodyanik et al., 2010) are in agreement with other studies (Era et al., 2008) that indicate that PSCs give rise to multiple types of endothelial progenitors. Importantly, we demonstrated that emerging endothelial progenitors are multipotent and are able to differentiate into cells of other mesodermal lineages. The first progenitors with endothelial potential, the mesenchymoangioblasts arise from PSCs on day 2 of differentiation and are capable of forming mesenchymal colonies (Vodyanik et al., 2010). HBs capable of generating primitive blood cells through endothelial intermediates in semisolid medium arise one day later. Endothelial intermediates that form HB colonies most likely resemble the yolk sac HE. HEPs that develop by day 5 in hPSC/OP9 coculture express RUNX1, and have the potential to generate multipotential myeloid cells and β-globin producing eryhroid cells, and thus resemble definitive-type endothelial progenitors. Non-HEPs were distinctively recognized by the expression of CD73 (Figure 6).
CD73, also known as 5′-ectonucleotidase, is a glycosylphosphatidylinositol (GPI) linked 70-kDa glycoprotein that produces extracellular adenosine and is abundantly expressed by endothelial cells, MSCs, subsets of peripheral blood lymphocytes and a variety of other tissues (Delorme et al., 2008; Thomson et al., 1990). CD73 is involved in the regulation of vascular permeability and maintenance of the barrier function, adaptation to hypoxia, ion and fluid transport, and regulation of inflammatory responses in the extracellular milieu (Colgan et al., 2006). Given the physiological significance of CD73, it is likely that expression of this molecule reflects not only differences in developmental potential, but also in the functional properties of HE and non-HE subsets. Other distinctive features of CD73+ non-HEP were the lack of CD226 hematopoietic marker expression and the strong expression of CD117 (c-KIT). CD117 is known to mark HSCs arising from AGM, and is also found in CD45−CD31+ circulating endothelial progenitors and cardiac endothelial progenitors (Peichev et al., 2000; Sandstedt et al., 2010; Tallini et al., 2009). The strong expression of CD117 and the lack of hematopoietic potential in CD73+ endothelial cells, indicate that these cells represent a population of endothelial progenitors distinct from blood-forming endothelial progenitors and may resemble circulating or tissue-specific endothelial progenitors. Whether these newly identified subsets of endothelial cells possess distinct functional properties and endothelial differentiation potential remains to be explored.
In conclusion, the identification of distinct subsets of cells with angiohematopoietic potential in our studies provides a hPSC-based platform for identification of molecular determinants of HSC development with a goal to facilitate generation of HSCs from hPSCs.
The experimental procedures used in this manuscript are briefly described as below. Supplemental Information contains a complete description.
H1 and H9 hESC lines were obtained from WiCell Research Institute (Madison, WI). H9-EGFP (Xia et al., 2008) was kindly provided by Su-Chun Zhang (UW-Madison). Transgene-free DF4-3-7T and DF19-9-7T human iPSC cell lines produced using episomal vectors (Yu et al., 2009). All hESC/iPSC lines were maintained in undifferentiated state on irradiated mouse embryonic fibroblasts as described previously (Yu et al., 2007).
hESC/iPSCs were differentiated in coculture with OP9 stromal cells provided by Dr. Toru Nakano (Osaka University, Japan) and depleted of OP9 cells using anti-mouse CD29 antibodies (Serotec) as described (Vodyanik et al., 2010).
The approach and antibodies used for isolation of distinct subsets of progenitors with angiohematopoietic potential is depicted in Figure S2 and Table S2. VE-cadherin+ or CD31+ cells were isolated from day 5 hESC/OP9 cocultures by positive MACS selection using corresponding FITC-conjugated antibodies and anti-FITC magnetic beads (Miltenyi). MACS separated cells were stained with CD73-PE, CD235a-APC and CD43-APC, and CD41a-PECy7 antibodies and sorted using a FACSAria™ cell sorter (BD Biosciences) to select subsets as depicted in Figure 1B and S2. To isolate day 4 subsets from hPSC/OP9 cocultures, KDR positive cells were selected by positive MACS selection using KDR-PE antibody (R&D Systems) and anti-PE magnetic beads (Miltenyi). After labeling with VE-cadherin-APC antibody, KDRbrightVE-cadherin−, KDRdimVE-cadherin− and KDR− cells were further separated using the FACSAria sorter. APLNR+ cells were isolated from day 3 hESC/OP9 cocultures by MACS sorting with APLNR-APC antibodies and APC-magentic beads or FACSAria sorter after depletion of OP9 with anti-mouse CD29 antibodies as described (Vodyanik et al., 2010). Hematopoietic and endothelial potential of isolated cells was evaluated before and after secondary coculture with OP9 using CFC assay, endothelial culture, and flow cytometry (see Supplemental Information).
The significance of differences between the mean values was determined by paired Student’s t test.
We thank Dr. Toru Nakano for providing OP9 cells, Dr. Dietmar Vestweber for providing endomucin antibodies, Su-Chun Zhang for providing H9-EGFP cells, and Gene Uenishi for editorial assistance. This work was supported by funds from the National Institute of Health (R01 HL081962, U01HL099773, P01 GM081629, and P51 RR000167), Charlotte Geyer Foundation and Lupus Foundation of America. K.S. is supported by funds from the Department of Pharmacology, Faculty of Science, Mahidol University, Bangkok, Thailand. J.A.T. owns stock, serves on the Board of Directors, and serves as Chief Scientific Officer of Cellular Dynamics International. J.A.T. also serves as Scientific Director of the WiCell Research Institute. I.S. owns stock and is scientific founder of Cellular Dynamics International. I.S. is also scientific founder of Cynata. M.V. is scientific founder of Cynata.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Supplemental information The supplemental information includes extended experimental procedures, seven supplemental figures, three supplemental tables, supplemental references and one movie.