A mesoderm-derived precursor for mesenchymal stem and endothelial cells

Summary

Among the three embryonic germ layers, the mesoderm is a major source of the mesenchymal precursors giving rise to skeletal and connective tissues. These precursors, however, have not been previously identified and characterized. Using human embryonic stem cells directed to mesendodermal differentiation, here we show that mesenchymal stem/stromal cells (MSCs) originate from a population of mesodermal cells identified by expression of the apelin receptor. In semisolid medium, these precursors form FGF2-dependent compact spheroid colonies containing mesenchymal cells with a transcriptional profile representative of mesoderm-derived embryonic mesenchyme. When transferred to the adherent cultures, individual colonies give rise to MSC lines with chondro-, osteo-, and adipogenenic differentiation potentials. Although the MSC lines lacked endothelial potential, endothelial cells could be derived from mesenchymal colonies, suggesting that, similar to hematopoietic cells, MSCs arise from precursors with angiogenic potential. Together, these studies identified a common precursor of mesenchymal and endothelial cells, mesenchymoangioblast, as the source of mesoderm-derived MSCs.

MSCs are fibroblastoid cells capable of long-term in vitro expansion and multilineage differentiation to bone, cartilage, adipose and connective tissues (Bianco et al., 2008; Dennis and Charbord, 2002; Friedenstein et al., 1974; Prockop, 1997). In addition, MSCs give rise to hematopoiesis-supportive stroma and contribute to the formation of the HSC niche (Blazsek et al., 2000; Delorme et al., 2006; Muguruma et al., 2006) and vascular wall (Crisan et al., 2008). Although MSCs are widely used for cellular therapies and tissue engineering, their precursors remain largely unknown. Thus there is a need to define developmentally distinct MSC subsets and the hierarchy of their progenitors to advance our understanding of heterogeneity within MSCs and its implications for the developmental and therapeutic potential of these cells.

Studies in mouse embryo demonstrated the origin of MSCs from neural crest (Morikawa et al., 2009; Trentin et al., 2004). In addition, Sox1⁺ neuroepithelial cells were identified as progenitors that give rise to MSCs through a neural crest intermediate stage (Takashima et al., 2007). Recently, the MSC potential of neural crest stem cells generated from human embryonic stem cells (hESCs) was shown (Lee et al., 2007). The mesoderm is considered to be another and major source of mesenchymal cells giving rise to skeletal and connective tissues (Dennis and Charbord, 2002). An early Flk1⁺ mesodermal precursor, with the potential to differentiate into endothelial cells, blood, muscle and mesenchymal lineage cells (bone and cartilage), was identified in the E9.5 mouse dorsal aorta (Minasi et al., 2002). Evaluation of osteogenic, chondrogenic and adipogenic potential of cells isolated from different anatomical sites in the E11 mouse embryos revealed itraembryonic hematopoietic tissues (aorta-gonad-mesonephros; AGM) as a site of origin for cells with mesenchymal differentiation potential (Mendes et al., 2005). However, immediate mesodermal precursors that give rise to expandable multipotential MSC lines are not identified and characterized.

To identify MSC precursors of mesodermal origin, we employed a human embryonic stem cell (hESC) differentiation system that reproduces many aspects of early embryonic development and provides access to cells otherwise inaccessible in humans that represent the earliest stages of mesodermal commitment (Gadue et al., 2005; Kennedy et al., 2007; Kennedy et al., 1997; Yang et al., 2008; Zambidis et al., 2005). Although multiple studies described successful generation of MSCs from hESCs and iPSCs, including demonstration of fibroblast-colony forming activity (Barberi et al., 2005; Boyd et al., 2009; Karlsson et al., 2009; Lee et al., 2007; Lian et al., 2007; Lian et al., 2010; Olivier et al., 2006; Trivedi and Hematti, 2008), origin of MSCs in these cultures remains obscure. Mesenchymal “precursors” identified in these studies were isolated using typical MSC markers such as CD73 or CD105 and defined as cells with osteo-, chondro-, and adipogenic potential (i.e. representing de facto MSCs rather than cells predetermined to MSC fate or true MSC precursors). Here, using hESCs directed to mesendodermal differentiation through coculture with OP9, we show that mesoderm-derived MSCs arise from a common endothelial and mesenchymal cell precursor, mesenchymoangioblast, a transient population of cells within APLNR⁺ mesodermal subset that can be identified using FGF2-dependent mesenchymal colony-forming cell (MS-CFC) assay in serum-free semisolid suspension culture.

Results

Identification of Mesenchymal Colony-Forming Cells in hESCs Differentiated in Coculture with OP9

To identify human mesenchymal precursors of mesodermal origin we employed hESC coculture with OP9, which is known as an efficient differentiation system for cells of mesodermal lineages including hematopoietic, vascular, and cardiac cells (Nakano et al., 1994; Schroeder et al., 2003; Vodyanik et al., 2005). Molecular profiling of OP9-depleted hESC-derived cells differentiated throughout days 1–7 demonstrated selective commitment toward mesodermal and endodermal lineages with no detectable ectoderm (tropho-, neuro- or surface ectoderm) (Fig. S1). Kinetically, the T, MIXL1, and EOMES transcription factors involved in induction of mesendoderm were upregulated synchronously and peaked on day 2 of differentiation. This was followed by expression of mesoderm- and endoderm-specific genes, and coincided with a maximal cell proliferation on day 3 of differentiation. This stage was also accompanied by sustained expression of SNAI1 and SNAI2 genes involved in epithelial-mesenchymal transition. Among the mesodermal subsets, only genes representing lateral plate/extraembryonic mesoderm (FOXF1, HAND1, GATA2) were upregulated, while no substantial expression of axial (CHRD, SHH), paraxial/myogenic (MEOX1, TCF15, MYOD1, MYF5, PAX3, PAX7), or intermediate (PAX2, PAX8, OSR1) mesoderm genes was noticed. Following 4–5 days of differentiation, specification of endodermal and mesodermal lineages was observed.

Since FGF signaling is involved in specification, migration, and patterning of mesoderm during embryonic development (Ciruna and Rossant, 2001) and MSC expansion (Tsutsumi et al., 2001), and our observation that FGF2 selectively increased the mesenchymal subpopulation of hESC-derived CD34⁺ cells (Fig. S2), we tested if any hESC-derived cells with colony-forming mesenchymal potential could be detected in FGF2-supplemented semisolid media. In accordance with upregulation of genes marking mesodermal commitment on day 2–3 of hESC/OP9 coculture, we observed FGF2-dependent colony-forming activity when hESC-derived cells were plated in serum-free semisolid medium. Day 2 cells generated compact, sharply-circumscribed spheroid colonies formed by tightly packed cells, whereas day 3 cells mostly produced disperse colonies with a morphology similar to the previously described blast (BL) or hemangioblast colonies (Fig. 1A), which are known derivatives of angiohematopoietic mesoderm (Choi et al., 1998; Kennedy et al., 2007; Kennedy et al., 1997). Both colony types were generated at relatively high frequency, comprising up to 2–3% of the total hESC-derived cells (Fig. 1B). Similar to BL colonies (D'Souza et al., 2005), compact colonies were formed through establishment of tightly packed single cell-derived structures (cores) which expand into spheroid colonies (Fig. 1A and Movies 1S–3S). Both compact and BL colony formation was solely dependent on the presence of FGF2, but not other tested factors (VEGF, SCF, IGF1, EGF, and HGF; Fig. 1C and not shown), and required serum-free medium. While previous studies reported the VEGF dependence of BL colonies generated using embryoid body method (Kennedy et al., 2007; Kennedy et al., 1997), we could not detect any colonies when VEGF used alone and its addition to FGF2 had no significant effect on the number of BL colonies from hESCs differentiated on OP9. It is possible that this discrepancy could be attributed to the differences in differentiation systems. For example, a high level of autocrine VEGF production by hESCs differentiated on OP9, or the presence of inhibitors/modulators of VEGF signaling in embryoid body system. The formation of compact and BL colonies was completely abrogated by adding transforming growth factor β1 or activin A (1 ng/ml) to clonogenic cultures (Fig. 1C) or in presence of serum. Although PDGF-BB alone lacked colony-forming activity, its addition to FGF2 significantly increased the frequency and size of compact colonies. In contrast, the addition of VEGF to FGF2 cultures essentially abrogated formation of compact colonies at the core stage (Fig. 1C). It is important to notice that both types of colonies represent a transient population of cells that could not be detected beyond 4 days of differentiation when the first hematopoietic colonies arise (Fig. 1B).

Figure 1. Identification of mesenchymal colony-forming cells.

(A) Development of MS and BL colonies in semisolid culture. Scale bar, 100 μm. (B) Kinetics of MS-, BL- and Hem- (hematopoietic) CFCs in H1 hESC/OP9 coculture. Error bars, standard deviation of 6 independent experiments. (C) The effect of adding cytokines to clonogenic cultures on MS and BL colony formation. ActA is activin A. Bars represent standard deviation of four independent experiments. * p < 0.01. See also Figure S1 and S2.

Phenotypic analysis of fully developed spheroid colonies (day 12 of clonogenic culture) demonstrated that they consist of a uniform population of CD140a⁺CD90⁺CD56⁺CD166⁺CD31⁻CD43⁻CD45⁻ cells, with the majority of cells expressing MSC/perivascular cell marker CD146 (Crisan et al., 2008; Sacchetti et al., 2007), indicating the mesenchymal identity of these colonies (Fig. 2B–2C). Thus we designated these colonies as mesenchymal (MS) colonies and cells forming these colonies as MS-CFCs. It is noteworthy that cells within colonies expressed endomucin (EMCN), endothelial tyrosine kinase (TEK or TIE2), and VEGFR1 (FLT1, Fig 2B–2C) molecules typically associated with endothelial cells. Transcriptional profiling of MS and BL colonies clearly demonstrated that MS colonies were enriched in genes representative of embryonic mesenchyme originating from lateral plate/extraembryonic mesoderm (FOXF1, HAND1, NKX2–5, SNAI2, CDH11, RUNX2, SOX9), while BL colonies were enriched in genes signifying hematopoietic commitment (GATA1, NFE2, TAL1, SPI1, VAV1) (Fig. 2D). RT-PCR analysis of individual MS colonies revealed that essentially all colonies lacked expression of ESC (POU5F1, SOX2), neuroectoderm (SOX2), mesendoderm (T) and endoderm (FOXA2)–specific genes, but showed a high expression of lateral plate/extraembryonic mesoderm-specific marker FOXF1, and genes essential for epithelial-mesenchymal transition (SNAI2) and bone and cartilage development (RUNX2, SOX9). However, expression of genes involved in adipogenesis (PPARG) and myogenesis (MYF5) was not detected (Fig. 2E).

Figure 2. Characterization of mesenchymal colonies and cell lines derived from them.

(A) Derivation of MSC line from individual MS colony. Panels show morphology of MS colony in semisolid media and after placement on fibronectin/collagen-coated dishes, and MCS line after 1^st (p1) and 5^th (p5) passage. Scale bars, 200 μm. (B) Flow cytometric analysis of MS colonies, polyclonal and 24 individual clonal MSC lines; p4 is passage 4, EMCN is endomucin. (C) Summary of phenotype of MS colonies and MSC lines at passage 5 (p5) as detected by flow cytometry. Bars represent standard deviation of 3 independent experiments. (D) Divergent gene expression profile of MS and BL colonies. Selected transcripts defining mesodermal, mesenchymal and hematopoietic lineages are shown. (E) Quantitative RT-PCR analysis of gene expression in individual MS colonies. Each dot on the graphs represents gene expression level in a single colony. Horizontal lines show average expression levels for all tested colonies (n= 30). (F) Expansion potential of polyclonal cell lines (mean doubling time (dT) and standard deviation for 4 lines are shown). (G) Expansion potential of 10 individual clonal cell lines derived from a single MS colony. See also Figure S3, and S6.

Similar to hESCs, MS-CFCs were also generated from induced pluripotent stem cells (iPSCs); although we noted that MS colonies with typical morphology could be obtained only from transgene-free iPSCs. Lentivirally-reprogrammed cells generated colonies with aberrant morphology, indicating that background expression of transgenes could affect the spatial organization of colonies (Fig. S3A).

Single Cell Origin of Mesenchymal Colonies

As observed by time-lapse photography, MS colonies developed from single cells (Movie S1). To confirm single cell origin using alternative assays, we prepared a chimeric hESC (H1) line containing equal proportions of randomly mixed EGFP- and mOrange-marked H1 hESCs and assayed it for MS-CFCs. By prediction, if colonies are derived from single cells, no mixed colonies should be observed. However, if at least 2 or more cells are required, half or more of the colonies should be mixed. When the MS-CFC assay was performed with a differentiated chimeric cell population, almost all of the MS colonies were single-colored, either EGFP- or mOrange-positive. A minority of mixed colonies (<5%) was observed only in cultures with high cell plating density. At the density of 10⁴ cells/ml, all counted colonies (n=189) were of a single color (Fig. 3A), thus demonstrating that MS colonies arise from single cells, but not from cell aggregates. As additional proof of clonality, we performed a single cell deposition assay which revealed that singly-deposited cells generated MS colonies with a frequency similar to bulk cultures (~1%; Fig. 3B).

Figure 3. Demonstration of single cell origin of mesenchymal colonies.

(A) H1 cells expressing either EGFP or mOrange-H2BB fusion protein were used for preparation of chimeric H1 cell line containing equal proportions of randomly mixed EGFP- and mOrange-positive cells. Graph indicates the numbers of single color (red or green) or mixed (black) colonies in semisolid clonogenic culture at different plating densities. Bars show the absolute numbers of single color (red or green) or mixed (black) colonies in semisolid clonogenic culture counted at different plating densities. The numbers above bars indicate percentage of mixed colonies of total colonies counted. (B) A single cell deposition of APLNR⁺ HES2.R26tdRFP cells into methylcellulose medium with feeder culture of colony-forming H1 hESCs (10⁴ cells/ml).

Mesenchymal colonies originate from precursors with endothelial potential

To elucidate the differentiation potential, MS colonies were collected and transferred back onto OP9 for an additional 4 days of culture. As shown in Fig. 4A, the majority of cells differentiated from compact colonies expressed CD146⁺CD31⁻CD43⁻CD45⁻ MSC phenotype (Sacchetti et al., 2007). However, we also noticed a readily identifiable population of CD31⁺/CD144⁺CD43⁻ endothelial cells comprising up to 7% of the total hESC-derived cells. When MS colonies were plated on Matrigel matrix, endothelial cells migrated out of colonies and organized into a network of typical vascular tubes (Fig.4C). The endothelial potential of MS colonies could be significantly enhanced with the addition of bone morphogenic protein 4 to colony-forming medium (BMP4; Fig. 4D). Despite readily detectable endothelial differentiation, CD43⁺ (Vodyanik et al., 2006) or CD45⁺ hematopoietic cells were never detected in these cultures even from colonies generated in presence of BMP4 (Fig. 4A, 4D and Fig. S4A). When individual compact spheroid colonies were transferred onto OP9, about 70% of the colonies generated CD144 (VE-cadherin)⁺ endothelial cells in addition to the mesenchymal cells identified by expression of calponin (Fig. 4B). By contrast, BL colonies generated predominantly CD43⁺ hematopoietic cells with some CD31⁺/CD144⁺CD43⁻ endothelial cells and CD146⁺CD31⁻ mesenchymal cells (Fig. 4A and 4B). Kinetic analysis of the angiogenic potential of MS colonies at different stages of maturation revealed that the formation of mesenchymo-endothelial clusters was the most prominent from colonies collected on day 6 of clonogenic culture. Three-day-old (core stage) and 12-day-old (mature stage) colonies formed predominantly endothelial and mesenchymal clusters, respectively (Fig. S4B and S4C). Based on specific morphology, differentiation potential, and contrasting VEGF response, we concluded that the compact spheroid colonies represent common precursor for mesenchymal and endothelial lineages distinct from the already described BL-CFC (Choi et al., 1998; Kennedy et al., 2007; Kennedy et al., 1997). By analogy with hemangioblast, we designated this novel mesenchymal and endothelial precursor as mesenchymoangioblast.

Figure 4. Demonstration of endothelial potential of mesenchymal colonies.

(A) Differentiation potential of MS and BL colonies collected on day 12 of clonogenic culture after re-plating on OP9. Flow cytometric analysis demonstrates development of CD146⁺CD31⁻ mesenchymal, CD31⁺CD146⁺CD43⁻ endothelial, and CD43⁺ hematopoietic lineages. (B) Immunostaining analysis of cell clusters developed from a single MS (top, scale bar, 100 μm) and BL (bottom, scale bar, 50 μm) colony collected on day 5 of clonogenic culture. Cells are identified as CD144(VE-cadherin)⁺CD43⁻ endothelial, CD43⁺ hematopoietic and calponin⁺ CD144⁻ mesenchymal. Scale bars, 100 μm. (C) Vascular tube formation by MS colonies collected on day 6 of clonogenic culture. Scale bar, 200 μm. (D) Effect of the addition of BMP4 to clonogenic media on endothelial potential of individual colonies on OP9. Left upper quadrant numbers show percentage of CD31⁺ endothelial cells (mean ± SD of 6 independent experiments). (E) Lack of endothelial potential of MSC lines after coculture with OP9. Note MSCs acquire expression of CD34 after coculture with OP9, but CD31⁺ endothelial cells are not observed. See also Figure S4.

Single Mesenchymal Colonies Give Rise to MSC lines

When individual MS colonies were plated on the collagen/fibronectin-coated plastic, immediate attachment and vigorous outgrowth of fibroblast-like cells were observed (Fig. 2A). Following expansion, cells grew intensively during the first 10 passages (doubling time (dT)=18–23h), growth rate was attenuated at 10–15 passages (dT=30–35h), and gradual senescence was observed during 15–25 passages (Fig. 2F). Cultures derived from single MS-CFC accumulated up to 10²² total cells (Fig. 2G). Assuming a single-cell origin of colonies, this number corresponds to the expansion potential of single hESC-derived mesenchymal precursor. Although the cells composing the MS colonies were lacking CD73 and expressed CD105 weakly, colony-derived cell lines upregulated their expression and displayed CD146⁺CD73⁺CD105⁺CD31⁻CD43⁻CD45⁻ phenotype typical of adult bone marrow MSCs and maintained normal karyotype (Fig. 2B–2C, Fig. S3B). Generated MSC lines could be differentiated in the chondro-, osteo- and adipogenic conditions with triple potential detected in the majority of single colony-derived lines (Fig. 5). This finding provides functional confirmation of the MSC nature of the colony-derived cell lines and proves MSC origin from mesenchymoangioblasts. Neither endothelial nor hematopoietic cells could be detected after coculture of MSC lines with OP9 cells (Fig. 4E). The same was true in the feeder-free culture with hematoendothelial growth factors (VEGF, FGF2, SCF, TPO, IL3, IL6; not shown), indicating the restricted differentiation potential of expanded MSCs.

Figure 5. Skeletogenic differentiation potential of MSC lines.

Chondrogenesis is shown in (A), by Alcian blue staining, and in (B), by glycosaminoglycan (GAG) production following differentiation in chondrogenic conditions. Osteogenesis is shown in (C), by Alizarin red staining and (D), by calcium accumulation following differentiation in osteogenic conditions. Adipogensis is shown in (E), by Nile red staining and (F), by fat accumulation following differentiation in adipogenic conditions. In (A), (C), and (E), scale bars are 100 μm. In (B), (D), and (F), bars represent standard deviation of mean obtained at indicated time points (Days 0–36 of differentiation) using polyclonal cell lines (n=3). *p< 0.01 as compared to Day 0. Each dot on the graphs represents values by individual clonal cell line on day 24 of differentiation. Horizontal lines show average values for all tested clonal cell lines (n=24). (G) Kinetic analysis of mesenchymal lineage-specific gene expression in polyclonal MSC lines differentiated during 24 days under chondrogenic, osteogenic, and adipogenic conditions as determined by quantitative RT-PCR. Average expression normalized to RPL13A. Bars represent standard deviation of three independent experiments.

Mesenchymoangioblasts Originate from Population of Mesodermal Cells Identified by Expression of Apelin Receptor

Demonstration of endothelial potential of MS colonies strongly indicates the mesodermal origin of MS-CFCs. However, neural crest origin of these cells could not be completely excluded since the potential of neural crest cells to form corneal endothelium is recognized (Noden, 1978). To more precisely define the source of mesenchymoangioblasts, we decided to identify a cell population enriched in MS-CFCs. Although VEGFR2 (Flk1 or KDR) and PDGFRα in mouse are well-defined markers of mesoderm with the potential to generate hematovascular and mural cells (Orr-Urtreger et al., 1992; Sakurai et al., 2006; Yamaguchi et al., 1993; Yamashita et al., 2000), we noticed that both of them were expressed weakly following 2–3 days of hESC differentiation in OP9 (Fig.6A), complicating isolation of discrete populations by sorting. Moreover, the KDR expression in undifferentiated hESCs limits the utility of this marker for isolation of early-differentiated cell population (Vodyanik et al., 2005; Yang et al., 2008). During analysis of the transcriptional profile of differentiated hESCs in OP9 coculture, we found that expression of apelin receptor (APLNR, also known as angiotensin receptor like-1) was strongly induced and upregulated on days 2–3 of differentiation concurrently with mesodermal commitment and colony-forming activity. Since previous studies showed APLNR expression in lateral plate mesoderm and its precursors in xenopus and zebrafish embryo at gastrula stage (Devic et al., 1996; Zeng et al., 2007) and in primitive streak and adjacent embryonic and extraembryonic mesoderm in mouse embryo (D'Aniello et al., 2009), we studied APLNR expression following hESC differentiation by flow cytometry. Unlike KDR, undifferentiated hESC were homogeneously APLNR⁻. On day 2 of coculture, 15–20% of differentiated hESCs became APLNR⁺. The proportion of positive cells increased to 60–70% on day 3, and gradually decreased in the following days (Fig. 6A). In contrast to the low initial expression of PDGFRα and KDR, the first positive cells expressed APLNR in high density thus allowing for a clean separation of negative and positive populations. Co-staining with anti-T and -FOXA2 antibodies revealed that APLNR⁺ cells were T⁺FOXA2⁻ mesodermal precursors (Fig. 6A). Further proof that APLNR⁺ cells represent mesoderm was obtained by differentiating hESCs in the presence of inhibitors of WNT and TGF-β signaling and mesoderm induction (Gadue et al., 2006). These experiments demonstrated almost an entire abrogation of the APLNR⁺ population in OP9 coculture (Fig. 6B). The molecular profiling analysis shown in Fig. 6C and Fig. 7 demonstrated that transcripts associated with neural crest/neuroectoderm (POU4F1, SOX1, SOX10, SOX3) were not found in APLNR⁺ cells. Genes indicative of primitive streak cells/mesendoderm (MIXL1, T, EOMES) were all expressed but not enriched in APLNR⁺ cells, consistent with their advanced lineage commitment. As expected, APLNR⁺ cells were enriched in TCF21 mesoderm-specific transcripts, whereas transcripts marking pan-endoderm (FOXA2, APOA1), definitive (FOXA1, TMPRSS2), and visceral (TTR, AFP) endoderm were found in APLNR⁻ cells. Interestingly, APLNR⁺ cells expressed the FOXF1, IRX3, BMP4, WNT5A, HAND1, and HAND2 genes representative of lateral plate/extarembryonic mesoderm, but not the markers of paraxial/myogenic (MEOX1, TCF15, PAX3, PAX7) and intermediate (PAX2, PAX8) mesoderm in the embryo. This data indicates that rather than being a total population of cells committed to mesendodermal development, APLNR⁺ cells represent a mesoderm, or likely a subpopulation reminiscent of lateral plate/extraembryonic mesoderm.

Figure 6. Identification of APLNR as a marker of mesoderm generated from hESCs in coculture with OP9.

(A) Characterization of H1 hESC-derived APLNR⁺ cells by flow cytometry. (B) Effect of WNT (DKK-1, dickkopf homolog 1; 150 μg/ml) and TGF-β1 (SB431542; 5 μg/ml) signaling inhibitors on generation of APLNR⁺ cells. Error bars, standard deviation of 4 independent experiments. (C) Enrichment of mesodermal transcripts in APLNR⁺ cells isolated on day 3 of differentiation. (D) Frequency of MS- and BL-CFCs in APLNR⁺ and APLNR⁻ cell fractions isolated at days 2, 2.5 and 3 of hESC/OP9 coculture. Bars represent standard deviation of 3 independent experiments. See also Figure S5.

Figure 7. Gene expression profiling of APLNR⁺ and APLNR⁻ cells, cores, MS and BL colonies and MSC lines (passages p1 and p5) obtained from H1 hESCs differentiated for 2 (D2) or 3 (D3) days in coculture with OP9.

(A) Shown are heat maps for selected sets of genes defining indicated germ layers and their subpopulations/derivatives. Cores were collected on day 3 of clonogenic cultures and fully developed colonies on day 12 of clonogenic cultures. EMT is epithelial-mesenchymal transition. VSMC is vascular smooth muscle cells. (B) Lack of SOX1 neuroepithelium marker expression throughout all stages of differentiation. H1 embryonic bodies differentiated for 14 days were used as positive control. (C) Quantitative RT-PCR analysis of representative transcripts in indicated cell subsets. Bars represent gene expression in pooled samples from 3 experiments normalized to RPL13.

To determine origin of MS- and BL-CFCs, differentiated hESCs were fractionated into APLNR⁺ and APLNR⁻ cells and assayed for FGF2-dependent colony-forming potential. As shown in Fig. 6D, selection of APLNR⁺ cells resulted in almost entire depletion of MS-CFCs and BL-CFCs in APLNR⁻ fraction. However, the enrichment in MS-CFCs in APLNR⁺ fraction was less than expected. To find out whether poor enrichment could be caused by the inhibition of MS-CFCs by antibody binding to APLNR through modulation of apelin signaling, we tested the effect of adding anti-APLNR antibody and APLNR ligand apelin-12 to clonogenic cultures of day 2 unseparated hESCs. As shown in Fig. S5A, apelin-12 inhibited MS colony formation, indicating that signaling through APLNR has a negative effect on MS-CFCs. Because a similar inhibitory effect was observed in cultures containing antibody, we concluded that anti-APLNR monoclonal antibody used in this study possess agonistic properties, which can explain limited MS-CFCs enrichment in APLNR⁺ fraction following magnetic separation. Nevertheless, segregation of MS- and BL-CFCs almost exclusively to APLNR⁺ fraction additionally confirmed mesodermal origin of mesenchymal precursors. Since no expression of neuroepthelial/neural crest markers including SOX1 was detected in APLNR⁺ cells, MS colonies and MSCs (Fig. 7B), neural crest origin of MSCs in our culture system was completely excluded.

Flow cytometric analysis (Fig. S5B) and molecular profiling data (Fig. S1) demonstrated that mesenchymoangioblasts develop within 48 hours of differentiation prior to the expression of CD73 and CD105 MSC markers and upregulation of MSC-related genes, i.e. onset of mesenchymogenesis. Although MS-CFCs arise from a APLNR⁺ mesoderm with silent endothelial genes, their specification toward MSCs in colony-forming cultures proceeds through the core stage at which cells acquire angioblastic gene expression profile (Fig. 7A and 7C) and endothelial differentiation potential, which is mostly detectable during the early stages of colony formation (Fig. S4B). This activation of an endothelial program was evident during the core formation of both MS and BL colonies, although additional divergent activation of mesenchymal versus hematopoietic genes was obvious. Maturation of MS colonies was associated with progressive loss of endothelial and increase of mesenchymal potential, indicating that specification APLNR⁺ cells toward mesenchymal cells in clonogenic cultures proceeds through formation of angiogenic core followed by its progressive transformation into tripotential mesenchymal cells.

Discussion

In the present study we demonstrate for the first time that mesoderm-derived MSCs originate from precursors with angiogenic potential, called mesenchymomagioblasts, which we identified as MS-CFCs with the potential to differentiate into both MSCs and endothelial cells. In embryo, early angiogenic precursors are scattered throughout the mesoderm and some of them possess strong migratory activity (Cleaver and Krieg, 1998; Noden, 1990; Pardanaud et al., 1996). Endothelial precursors originating in paraxial mesoderm vascularize predominantly kidney and body wall, whereas endothelial precursors of lateral plate mesoderm origin vascularize visceral organs and contribute to aortic floor and intra-aortic hematopoietic cell clusters (Pardanaud et al., 1996). Early angiogenic precursors seem to possess a broad differentiation potential including blood, cartilage, bone, smooth, skeletal, and cardiac muscle (Minasi et al., 2002). It is well documented that distinct populations of endothelial cells/progenitors within the floor of the dorsal aorta give rise to hematopoietic stem cells, and that bipotential precursors with endothelial and hematopoietic differentiation potential are identified in ESC cultures (Bertrand et al., 2010; Boisset et al., 2010; Choi et al., 1998; Kissa and Herbomel, 2010; Nishikawa et al., 1998; Taoudi and Medvinsky, 2007; Zovein et al., 2008). However, association between endothelial precursors and MSCs is not well established, although cells with endothelial and mural cell differentiation potential have been demonstrated in Flk1⁺ cells generated from mouse ESCs (Yamashita et al., 2000). In our studies, the identification of a clonal mesodermal precursor mesenchymoangioblast giving rise to endothelial cells and MSCs with robust expansion and trilineage (chondro-, osteo-, and adipogenesis) differentiation potential strongly indicates the branching off of the MSCs from precursors with primary endothelial characteristics. The development of MSCs through endothelial pathway could be a potentially distinctive feature of mesoderm-derived MSCs, since neural crest-derived MSCs initially arise from neuroepithelim (Takashima et al., 2007). Interestingly, neural crest and angioblasts are two major types of cells that migrate extensively during embryonic development (Evans and Noden, 2006). These properties could be important for establishing MSC network within the tissues.

In contrast to the previously described meso-angioblast, a mesodermal progenitor with angiogenic characteristics and broad differentiation potential (Minasi et al., 2002), the mesenchymoangioblast identified in our studies demonstrates a more narrow differentiation capability (MSCs and endothelium). Because mesenchymogenic MS-CFCs and hemogenic BL-CFCs arose in continuity and show more limited differentiation potential, these cells could represent precursors downstream of common mesodermal precursor, defining the stage when cells with multipotent mesodermal potential diversify into angiogenic subsets with more restricted mesenchymal or blood differentiation potential (see graphical abstract available on line).

Flk1 (VEGFR2 or KDR) is a well-established marker of mesoderm and angiogenic precursors in mouse embryo and differentiated ESC cultures (Orr-Urtreger et al., 1992; Sakurai et al., 2006; Yamaguchi et al., 1993; Yamashita et al., 2000). However, KDR is expressed in undifferentiated hESCs (Vodyanik et al., 2005; Yang et al., 2008) and its expression at early stages of differentiation is mostly weak (see Fig. 6A). This limits the utility of KDR for isolation of discretely defined negative and positive populations. Our data indicates that APLNR could be a good alternative to KDR marker for isolation of early mesoderm-committed population from hESCs in OP9 coculture because of its absence in undifferentiated hESCs and much brighter expression following early differentiation. APLNR (also known as angiotensin type I-like receptor AGTRL1 or APJ) is a G-coupled protein receptor activated by its ligand apelin, a peptide originating from preapelin (Habata et al., 1999). Studies in xenopus and zebrafish indicated a critical role APLNR-mediated signaling in normal vascular and cardiac development and migration of myocardial progenitors (Inui et al., 2006; Scott et al., 2007). At gastrula stage in xenopus, expression of apelin receptor (X-msr) was first detected in the prospective lateral plate mesoderm cells and later became restricted to lateral plate mesoderm (Devic et al., 1996). During gastrulation in zebrafish, APLNR is expressed in adaxial, intermediate, and lateral plate mesoderm, including anterior lateral plate mesoderm where cardiac precursors develop (Zeng et al., 2007). In E8 mouse embryos, APLNR is detected in primitive streak, adjacent mesoderm, and extraembryonic mesoderm (D'Aniello et al., 2009). Molecular profiling studies presented here demonstrate that OP9 induces selective commitment of hESCs to mesoderm and endoderm with predominant expansion of lateral plate/extraembryonic mesoderm and their derivates. Selection of APLNR⁺ population enriches cells with a lateral plate/extraembryonic mesoderm gene expression profile. Although the full differentiation potential of APLNR⁺ cells remains to be determined, our studies indicate that these cells can generate at least MSCs, endothelial, and blood cells, thus confirming the identity of APLNR⁺ population as lateral plate/extraembryonic mesoderm. Importantly, APLNR expression was detected before differentiated hESCs acquire expression CD73 and CD105 MSC markers, thus making it possible to separate MSC precursors from already established MSCs.

In conclusion, the identification of mesenchymoangioblast as a precursor for mesenchymal and endothelial cells, and APLNR⁺ mesodermal population enriched in cells with angiohematopoietic and MSC differentiation potential offers a novel opportunity to investigate the cellular and molecular pathways for the development of mesodermal lineage cells. In addition, access to clonal MSC populations with a well-defined origin, differentiation, and robust expansion potential could provide a unique cell source for tissue engineering and MSC-based medical therapies.

METHODS

Maintenance of human ESCs and iPSCs

Human ESC (H1, H13, H9) (Thomson et al., 1998) lines were obtained from WiCell Research Institute (Madison, WI). iPSCs (iPS(IMR)-90-1 and iPS(FSK)-1) were generated by reprogramming fetal and neonatal fibroblasts using lentiviral vectors (Yu et al., 2007). Transgene-free iPS-1(19-9-7T) and iPS-5 (4-3-7T) were produced by episomal vectors (Yu et al., 2009). HES2.R26-RFP cell line (Irion et al., 2007) was provided by Dr. Gordon Keller (Mount Sinai School of Medicine, New York, NY, USA). All hES/iPSC lines were maintained in undifferentiated state on irradiated mouse embryonic fibroblasts as described (Amit et al., 2000; Yu et al., 2007).

hES/iPSCs differentiation in OP9 coculture

Mouse OP9 bone-marrow stromal cell line was provided by Dr. Toru Nakano (Osaka University, Japan). hES/iPSCs were induced to differentiate in coculture with OP9 stromal cells and depleted of OP9 cells using anti-mouse CD29 antibodies (AbD Serotec, Raleigh, NC) as described (Vodyanik et al., 2005; Vodyanik and Slukvin, 2007).

Colony-forming culture for mesodermal precursors

Schematic diagram of the protocol used for generation of MS colonies and MSC lines is depicted in Fig. S6. Single-cell suspension of hES/iPSC-derived cells was prepared at 0.5–2×10⁴ cells/ml in the semisolid colony-forming serum-free medium (CF-SFM) containing 40% ES-Cult™ M3120 methylcellulose (2.5% solution in IMDM; Stem Cell Technologies, Vancouver, BC, Canada), 25% StemSpan™ serum-free expansion medium (SFEM; Stem Cell Technologies), 25% human endothelial serum-free medium (ESFM; Invitrogen), 10% BIT 9500 supplement (Stem Cell Technologies), GlutaMAX™ (1/100 dilution; Invitrogen), Ex-Cyte® supplement (1/1000 dilution; Millipore, Billerica, MA), 100 μM MTG, 50 μg/ml ascorbic acid and 20 ng/ml basic fibroblast growth factor (FGF2). Where indicated, human PDGF-BB (10 ng/ml), VEGF (20 ng/ml), stem cell factor (SCF; 50 ng/ml), epidermal growth factor (EGF; 20 ng/ml), insulin-like growth factor 1 (IGF1; 50 ng/ml), hepatocyte growth factor (HGF; 20 ng/ml), activin A (1 ng/ml), transforming growth factor β1 (TGF-β1; 1 ng/ml), bone-morphogenic protein-4 (BMP-4; 5 ng/ml), apelin-12 (100 ng/ml; Phoenix Pharmaceuticals Inc), or anti-APLNR monoclonal antibody (5 μg/ml; R&D Systems) were added to cultures. All cytokines were purchased from Peprotech (Rocky Hill, NJ). MS and BL colonies were scored on 12^th day of culture. Individual MS colonies were picked from culture under inverted microscope. For bulk collection of MS colonies (>100 μm in diameter), day 12 colony-forming cultures were diluted 1/5 in DMEM/F12 medium and filtered through 100 μm cell strainers (BD Biosciences). Cores from day 3 colony-forming cultures were purified in a similar manner using 30 μm cell strainers.

Colony-derived MSC lines

Fibronectin plus collagen-coated plastic was prepared by incubation of tissue culture grade plastic (BD Bioscience) with human fibronectin (5 μg/ml; Invitrogen) and human collagen I (10 μg/ml; Stem Cell Technologies) solutions in PBS. Individual, or purified by filtration, MS colonies (>100 colonies per culture) were plated on the fibronectin/collagen-coated plastic in mesenchymal serum-free expansion medium (M-SFEM) containing 50% StemLine™ II serum-free HSC expansion medium (HSFEM; Sigma), 50% ESFM, GlutaMAX™ (1/100 dilution), Ex-Cyte® supplement (1/2000 dilution), 100 μM MTG, and 10 ng/ml FGF2. After 3 days, the attached colonies were dissociated by StemPro® Accutase® solution (Invitrogen), and plated on the fibronectin/collagen-coated dishes in M-SFEM medium. Colony-derived MSC lines established either from individual (clonal lines) or multiple colonies (polyclonal lines) were routinely maintained by 3-day subculture on fibronectin/collagen-coated 10-cm dishes using StemPro® Accutase® detachment solution and M-SFEM. The first confluent culture was denoted as passage 1.

Analysis of angiohematopoietic potential of MS and BL colonies and MSC lines

MS and BL colonies or MSC lines were cocultured with OP9 in αMEM supplemented with 10% FBS and cytokines SCF (50 ng/ml), TPO (50 ng/ml), IL-3 (10 ng/ml), and IL-6 (20 ng/ml). After 4 days of culture, cells were harvested and analyzed by flow cytometry or stained in situ with rabbit anti-human CD144 (VE-cadherin; 1 μg/ml; eBioscience, San Diego, CA) in combination with mouse anti-human CD43 (0.5 μg/ml; BD Bioscience), mouse anti-human Calponin (0.5 μg/ml; Thermo Fisher Scientific), anti-human nuclei (Millipore; clone 235-1; 1/1000 dilution) or anti-human OB-cadherin (Invitrogen) primary antibodies, followed by a mix of secondary cross-absorbed donkey anti-mouse IgG-DyLight 594 and donkey anti-rabbit IgG-DyLight-488 (both at 2 μg/ml; Jackson ImmunoResearch Laboratories, Inc., West Grove, PA) antibodies. For vascular tube formation, MS colonies were collected at day 6 of clonogenic culture, transferred onto growth factor-reduced Matrigel (BD Biosciences) and cultured in EGM-2 endothelial cell growth medium (Lonza, Bazel, Switzerland) at 37°C.

Isolation of APLNR⁺ cells

To detect and isolate APLNR⁺ cells, anti-human APLNR (APJ, AGTRL1) mAb (clone 72133, R&D Systems) were conjugated with APC by Lightning-Link-APC kit (Innova Biosciences Ltd., Cambridge, UK). Isolation of APLNR⁺ cells was performed using MACS and anti-APC magnetic beads (Miltenyi Biotec, Bergisch Gladbach, Germany). APLNR⁺ cells were purified using LS+ positive selection columns. Negative cell fraction from LS+ column was additionally passed through high-retention LD depletion column to obtain fraction of APLNR⁻ cells. The purity of APLNR⁺ cells was higher than 95% as verified by flow cytometry. APLNR⁻ cells contained <1% APLNR⁺ cells.

Statistical Tests

The significance of differences between the mean values was determined by paired Student's ttest.