IV. IN VITRO DIFFERENTIATION POTENTIAL OF EMBRYONIC STEM CELLS
During mouse embryogenesis, the primitive ectoderm of the epiblast forms three primary germ layers: the ectoderm, the mesoderm, and the definitive endoderm. These germ layers interact to form all tissues and organs of the developing embryo. The complex interactions that control the transition of ectoderm to visceral and parietal endoderm in the postimplantation embryo, followed by the formation of mesoderm at the gastrulation stage (days 3 to 7 post coitum), are only beginning to be defined. The in vitro differentiation potential of mES cells has facilitated the examination of these processes.
Differentiation is induced by culturing ES cells as aggregates (EBs; Fig. 6) in the absence of the self-renewal signals provided by feeder layers or LIF, either in hanging drops (40, 394, 395, 398), in liquid "mass culture" (98), or in methylcellulose (390). Moreover, coculture with stromal cell line activity (i.e., of PA6 cells, Ref. 186), and recently, even adherent monolayer cultures in the absence of LIF (411) have been used to differentiate mES cells in vitro. Scaleable production of ES-derived cells can furthermore be achieved through the use of stirred suspension bioreactors with encapsulation techniques (92).
Once differentiation has begun, cells representing primary germ layers spontaneously develop in vitro. Initially, an outer layer of endoderm-like cells forms within the EB, followed over a period of a few days by the development of an ectodermal "rim" and subsequent specification of mesodermal cells. Transfer of these EBs to tissue culture plates allows continued differentiation into a variety of specialized cell types including cardiac, smooth, and skeletal muscle as well as hematopoietic, pancreatic, hepatic, lipid, cartilage, or neuronal and glial cells (see Table 3 and Fig. 6). The temporal expression of tissue-specific genes and proteins in ES-derived cells during in vitro differentiation indicates that early processes of in vivo development into ectoderm, mesoderm, and endoderm lineages are recapitulated in vitro (204, review in Ref. 306).
Both the pattern and the efficiency of differentiation are affected by parameters like ES cell density, media components (high glucose concentration, i.e., at least 4.5 g glucose/l is required) and amino acids, growth factors and extracellular matrix (ECM) proteins, pH and osmolarity, and the quality of the fetal calf serum (FCS). Because the differentiation efficiency depends on the presence of FCS, and even the "batch" of serum used, many efforts have been taken to avoid these uncertainties: dextran-coated charcoal (DCC)-treated FCS [to remove ECM and growth factor activity from FCS (397)], chemically defined medium (176, 279), and recently by substitution of FCS with BSA fraction V (411). Furthermore, different ES cell lines display unique developmental properties in vitro (see Ref. 395).
Another model to study early events of differentiation are "early primitive ectoderm-like" (EPL) cells derived from mES cells by adherent culture in medium conditioned by human hepatocellularcarcinoma HepG2 cells (MEDII-CM) (288, 289). EPL cells exhibit many properties consistent with embryonic primitive ectoderm, but are distinct from ICM and ES cells (compare Tables 1 and 2 with Fig. 1 of Ref. 302). The cells do not participate in embryogenesis following blastocyst injection. But, EPL cells allow modeling of early differentiation events without genetic modification. The aggregation of EPL cells into EBs results in a loss of visceral endoderm and neuroectoderm differentiation, whereas late primitive ectodermal, parietal endodermal, and mesodermal cells develop (302). This pattern suggests that the EPL-EB differentiation model may be suitable for studying mesoderm development in vitro and that failure to appropriately form visceral endoderm in EPL-derived EBs is responsible for the lack of ectoderm lineage formation. The defect in ectoderm differentiation, however, can be achieved by culture of EPL-EBs in the presence of MEDII-CM, which results in the formation of neuroectoderm (primitive ectoderm, neural plate, and neural tube) and an almost complete inhibition of endodermal and mesodermal differentiation (287) (see also sect. IVA).
hES cells differentiate when removed mechanically ("cut and paste") or by enzymatic dissociation from feeder layers and cultured as aggregates in suspension. Cystic EBs formed under these conditions are heterogeneous and express markers of various cell types, including those of neuronal, cardiac, and pancreatic lineages (168, 293, 323; Table 4). However, none of the factors known to influence mES cell differentiation directs hES cells exclusively into a single cell type (323). For instance, activin-A and transforming growth factor (TGF)- were found to induce mainly mesoderm; retinoic acid (RA), epidermal growth factor (EGF), BMP-4, and basic fibroblast growth factor (bFGF) elicited both ectodermal and mesodermal differentiation; whereas nerve growth factor (NGF) and hepatocyte growth factor (HGF) promoted differentiation of hES cells into all three primary germ layers. Interestingly, BMP-4 induced hES cells to develop into extraembryonic, trophoblast-like cells (403), a property clearly different from mES cells.
In section IV, A–D, we describe principal pathways and properties of differentiating mouse and human ES cells into derivatives of the three primary somatic and germ cell lineages. For methodical details such as differentiation protocols and differentiation factors, we refer the reader to the recent publications (79, 369, 382).
A. Ectodermal Differentiation
Among the various lineages produced by the embryonic ectoderm during normal mouse development, the neuroectodermal lineage gives rise to the peripheral and central nervous systems (review in Ref. 212) and to the epithelial lineage, which is committed to becoming epidermal tissue (review in Ref. 130). Vascular smooth muscles are also partially derived from embryonic ectoderm.
Epithelial cell differentiation from ES cells can be identified by the presence of cytokeratin intermediate filaments and keratinocyte-specific involucrin (20, 367). After in vitro differentiation of mES cells, enrichment of keratinocytes and seeding onto various ECM proteins in the presence of BMP-4 and/or ascorbate promotes formation of an epidermal equivalent, which is composed of stratified epithelium (when cultured at the air-liquid interface on a collagen-coated acellular substratum). The resulting tissue displays morphological patterns similar to normal embryonic skin. The cells express late differentiation markers of epidermis and markers of fibroblasts, consistent with those found in native skin. The data suggest that ES cells have the capacity to reconstitute in vitro fully differentiated skin (86).
Of specific importance with regard to cell therapies of neurodegenerative disorders are neuronal and glial cells. The differentiation of mES cells into neuronal cells was published independently by three groups in 1995 (22, 122, 350). The spontaneous differentiation of ES cells into neuronal cells was rather limited (see Ref. 350) but has improved significantly by a number of strategies, involving the use of RA (review in Ref. 306), lineage selection (210, 411), and stromal cell-derived inducing activity (for review, see Refs. 141, 186). Whereas high concentrations of RA originally promoted efficient neuronal differentiation, characterized by the expression of tissue-specific genes, proteins, ion channels, and receptors in a developmentally controlled manner (122, 350), the survival and development of neurons derived in response to RA is limited. Furthermore, the teratogenicity of RA (see Ref. 306) makes it unsuitable for therapeutic applications. For these reasons, alternative protocols, involving multiple steps of differentiation and selection of neural progenitor cells, have been established. Following EB formation, serum is withdrawn to inhibit mesodermal differentiation. The proliferation of neural precursor cells is then induced by the addition of bFGF and EGF. Thereafter, neuronal cell differentiation can be supported by the addition of neuronal differentiation factors (22, 253) and maintained in vitro by neurotrophic differentiation (206) and survival-promoting factors. These include the glial cell line-derived neurotrophic factor (GDNF), neurturin (NT), TGF-3, and IL-1 (311). Significant improvements in the generation and in vitro survival of dopaminergic neurons have been achieved using these factors. Neurons can also be generated from mES cells by RA treatment combined with the genetic selection of lineage-restricted precursors (see Ref. 210), by using EPL-derived EBs in the presence of MEDII-CM (287), or by the cocultivation of ES cells with PA6 stromal cells in serum-free medium (186). In the latter case, the stromal cells produce an inducing activity, which efficiently activates neuronal differentiation, including dopaminergic cells.
Gene expression and electrophysiological studies of cell derivatives indicate the presence of all three major cell types of the brain: neurons [dopaminergic, GABAergic, serotonergic, glutamatergic and cholinergic neurons (22, 116, 122, 186, 206, 311)], astrocytes, and oligodendrocytes (8, 366; see Table 3). An elegant genetic approach to identify and validate ES cell neural regulatory genes was recently described (14). In these experiments, the earliest known specific marker of mouse neuroectoderm (early neural plate and neural tube), Sox1, was targeted with a construct containing GFP. In Sox1-GFP positive ES cell progeny, fluorescence was observed only in early neural precursors. This strategy provided a robust quantitative assay for early steps in neural differentiation. By then using an episomal expression system (see sect. VD) for uniform expression of candidate cDNAs in RA-induced ES cell derivatives, the authors identified one gene, sfrp2, that could strongly stimulate the production of neural progenitors. SFRP2 is an extracellular antagonist of the Wnt family of signaling proteins. Transfection of ES cells with Sfrp2 resulted in enhanced neural differentiation in response to RA (and neural differentiation was obtained even in the absence of RA). Overexpression of Wnt-1 in ES cells inhibited neural differentiation, thus confirming a role of Wnt signaling in ES-derived neuronal differentiation (90, see also Ref. 307). Recently, the authors went on to show that for efficient differentiation into the neural lineage, neither multicellular aggregation nor coculture is necessary. In these experiments, targeted Sox1-GFP ES cells cultured in adherent monolayers, following an efficient neural differentiation regime (N2/B27 medium) and sorting by FACS, differentiated into a highly enriched Sox1-GFP fraction of neural progenitor cells. These selected cells were further differentiated into specific neuronal, glial, and oligodendrocytic cell types (15).
The ability of human ES cells to generate derivatives of the neural epithelium was demonstrated soon after their isolation (362); however, the selective derivation of a given neuron subtype (e.g., dopamine neuron fate) had, until recently, been unsuccessful. Neural progenitor cells derived from hES cells (292) may be specifically enriched (69) and directed to differentiate into mature neurons (e.g., dopaminergic, GABAergic, serotoninergic), astrocytes, and oligodendrocytes (69; see Table 4). Growth factors, mitogens (such as RA, NGF, bFGF, and EGF) (322), ECM proteins (Matrigel, laminin; Ref. 401), and stromal cell lines (MS5, S2) as well as Wnt1-expressing stromal cells (MS5-Wnt1; Ref. 266) all serve as potent enhancers of neuronal differentiation from hES cells. Coculture of hES cells on MS5 stroma and exposure to differentiation factors, such as FGF8, SHH, and BDNF, leads to efficient differentiation of neuroepithelial structures termed "neural rosettes." Replating of these rosettes followed by terminal differentiation produces midbrain dopaminergic neurons that express the neuronal transcription factors Pax2, Pax5, and engrailed-1; release dopamine; and show characteristic properties of dopaminergic neurons by electrophysiological and electron microscopical methods. High-yield dopaminergic neuron derivation was confirmed for both human and monkey ES cell lines (266). The availability of unlimited numbers of midbrain dopaminergic neurons represents a first step towards exploring the potential of hES cells in animal models of Parkinson's disease.
B. Mesodermal Differentiation
Mesoderm is the germ layer that develops into muscle, bone, cartilage, blood, and connective tissue. Blood and endothelial cells are among the first differentiated mesodermal cell types to form in the developing vertebrate embryo at around day E6.5, leading to the formation of yolk sac, an extraembryonic membrane composed of adjacent mesodermal and primitive (visceral) endodermal cell layers, which give rise to blood and endothelial cells (review in Ref. 26). Hematopoietic cells and blood vessels are believed to arise from a common progenitor cell, the "hemangioblast." As with ectodermal lineages, cultured ES cells have been successfully used to recapitulate these mesodermal developmental processes in vitro. Differentiation of ES cells in complex cystic EBs permits the generation of blood islands containing erythrocytes and macrophages (98), whereas differentiation in semisolid medium is efficient for the formation of neutrophils, mast cells, macrophages, and erythroid lineages (390). Application of FCS and cytokines such as IL-3, IL-1, and granulocyte-macrophage colony stimulating factor (GM-CSF) to ES cells generates early hematopoietic precursor cells expressing both, embryonic z globin (H1) and adult major globin RNAs. Use of OP9 cells, which secrete an inducing activity, also leads to the development of all hematopoietic cell types of the erythroid, myeloid, and lymphoid lineages (244) and of natural killer (Nk) cells (review in Ref. 159). Experiments to identify potential inducers of the hematopoietic lineage furthermore indicate that Wnt3 is an important signaling molecule that plays a significant role to enhance hematopoietic commitment during in vitro differentiation of ES cells (199).
The use of endothelial cell restricted promoters illustrates how in vitro analyses of EBs can be used to define complex mesodermal-derived cells. Quinn et al. (281) used the flt-1 promoter to regulate EGFP in PECAM-1 positive ES-derived endothelial cells. The expression of this transgene, at least temporally, coincided with the expression of endogenous flt-1. Further analyses of EGFP expression relative to Sca-1 positive cells suggested that the flt-1 promoter is active in ES-derived endothelial cells, but that it is downregulated during hemangioblast differentiation to the hematopoietic lineage (281). Similarly, Marchetti et al. (220) employed the vascular endothelium-specific promoter tie-1 to drive both EGFP and pacR expression to isolate endothelial cells from genetically modified ES cells. Puromycin (pacR)-resistant cells were positive for the endothelial cell surface markers, but release from puromycin selection resulted in the appearance of -smooth muscle actin positive cells, showing that endothelial cells or their progenitors could also differentiate towards smooth muscle. Finally, the expression of vascular endothelial growth factor receptor 2 (VEGF-2, known in the mouse as fetal liver kinase 1, Flk1) in early mesodermal progenitor cells also enabled the isolation of a Flk1+ cell population that includes endothelial and hematopoietic precursors (127, 249).
A similar strategy was used to study the specification of ES cells into the "hemangioblast." ES cell lines were created that express GFP targeted to the mesodermal gene brachyury (114), a transiently expressed mesoderm-specific transcription factor (176). Analysis of brachyury-GFP targeted cells permitted discrimination between mesoderm and neuroectoderm progenitors. Coexpression analysis of GFP with FLK1, furthermore, revealed three distinct mesodermal cell populations: premesoderm (GFP-/Flk1-), prehemangioblast mesoderm (GFP+/Flk1-), and the "hemangioblast" (GFP+/Flk1+) population, the precursor cells of primitive and definitive hematopoiesis and endothelium (114).
The cellular phenotypes of ES-derived hematopoietic cells have been characterized by specific gene expression patterns and by cell surface antigens (380, 390); however, the most important definition for these cells is functional. ES cell derivatives must demonstrate long-term multilineage hematopoietic repopulating properties to be considered true hematopoietic stem cells. Early reports suggested that the repopulating ability of ES-derived hematopoietic progenitors may be restricted to the lymphoid system (236), but subsequent studies showed a long-term multilineage hematopoietic repopulating potential of ES-derived cells (160, 259).
Another mesodermal cell type that has been extensively analyzed is ES cell-derived cardiomyocytes. These cells readily differentiate from aggregates composed of initially 400–800 starting cells that form in the presence of high FCS (20%) and display properties similar to those observed in cardiomyocytes in vivo or in primary cultures. These cells 1) express cardiac gene products in a developmentally controlled manner (40, 113, 230), 2) show characteristic sarcomeric structures (146, 228), and 3) demonstrate contractility triggered by cardiac-specific ion currents and the expression of membrane-bound ion channels (40, 154, 216–218, 394). The cardiomyocytes develop spontaneously (review in Ref. 43; see Ref. 395) or could be induced by differentiation factors including dimethyl sulfoxide (DMSO) and RA (394) and small molecules, such as Dynorphin B (374) and cardiogenol derivatives (399).
Electrophysiological analyses indicate that early differentiated cardiomyocytes are typical of primary myocardium (216), which subsequently differentiate to atrial-, ventricle-, Purkinje-, and pacemaker-like cardiomyocytes (review in Ref. 154). Importantly, patch-clamp and Ca2+ imaging techniques have permitted a thorough temporal-dependent analysis of electrical activity and the dynamics of ion channel expression and signaling cascades during cardiomyogenesis (1, 167, 174, 227). Microelectrode arrays (MEA) have furthermore been employed to temporally analyze excitation generation within ES-derived cardiac clusters. When EBs are plated onto MEAs, the electrical signals of the field potentials can be recorded over a period of several days from a multitude of electrodes beneath the spontaneously contracting cardiac clusters (24).
Cardiomyocytes differentiated from hES cells show similar properties to those derived from mES cells. Cardiac clusters have been identified on the basis of spontaneous contractions. The cell clusters are composed initially of small mononuclear cells with round or rod-shaped morphology that progress to form highly organized sarcomeric structures at later stages. The cardiac-specific gene expression pattern, electrophysiological properties, and chronotropic responses to adrenergic and muscarinic agonists are also typical of cardiomyocytes (188, 239, 240, 402). Cardiomyocytes differentiated from mouse and human ES cells show similar responses to -adrenergic and muscarinic modulation (290). The differentiation protocols with hES cells, however, yield an insufficient quantity of cardiac cells for experimental analyses. In this context, the recent discovery of cardiac-inducing signals from the endoderm (239) represents a step forward to the generation of cardiomyocytes from hES cells in vitro. The authors cocultured nonbeating EBs of hES cells on a monolayer of END-2 cells, an endodermal derivative generated from P19 embryonic carcinoma cells (241). This procedure resulted in the development of functional cardiomyocytes from hES cells. The continued identification of the molecular nature of the endoderm-derived factors and the application of efficient lineage selection strategies are requirements for the derivation of cardiac tissue from hES cells.
mES cells efficiently differentiate into several other mesodermal cell types, including mesenchymal cell-derived adipogenic (93), chondrogenic (194), osteoblast (61), and myogenic (309) cells (see Table 3). In all cases, the derivation of these cell types was induced by specific differentiation factors. Although all the protocols differ, they involve the successive treatment with specific growth and matrix factors, followed by a coordinated pattern of successive steps of differentiation. A sophisticated spinner culture system has also been established to generate vascular endothelial cells useful as a murine in vitro model for blood vessel development (381). Differentiation induction of mES cells by RA and dibutyryl cAMP resulted in the development of functional vascular smooth muscle cells typical of cells found in large arteries (99). These data show that complex vascular structures, as part of the cardiovascular system, originating in vivo from both mesoderm and neural-crest lineages, can be generated from ES cells in vitro.
C. Endodermal Differentiation
Pancreas and liver cells are derivatives of the definitive endoderm. During embryogenesis, the pancreas develops from dorsal and ventral regions of the foregut, whereas the liver originates from the foregut adjacent to the ventral pancreas compartment. Pancreatic and hepatic cells are of special therapeutic interest for the treatment of hepatic failure (147) and diabetes mellitus (337), and both pancreatic endocrine and hepatic cells develop in vitro from ES cells.
ES-derived hepatic cells show hepatic-restricted transcripts and proteins (149, 177) and can successfully integrate and function in a host liver following transplantation (78, 80, 404, 405). These data suggest that mES cells differentiate into all three lineages of the liver (hepatocytes as well as bile duct epithelial and oval cells). Differentiation strategies have begun to identify specific progenitor cells in the ES cell progeny, which may be of further use to isolate hepatic precursor cells of the liver (181, 182).
Hepatocyte-like endodermal markers were also detected in hES cell derivatives (285, 323). The successful differentiation and isolation of hepatic-like cells from hES cells has been demonstrated by using hES cells stably transfected with the reporter gene EGFP fused to an albumin promoter (203).
The generation of ES-derived insulin-producing pancreatic endocrine cells may be critical to the treatment of diabetes. The first successful induction of pancreatic differentiation from ES cells was obtained by stable transfection with a vector containing a neomycin-resistance gene under the control of the insulin promoter. This enabled lineage selection and maturation of insulin-expressing cells which, upon transplantation, resulted in the normalization of glycemia in streptozotocin-induced diabetic mice (338). In contrast, the spontaneous differentiation of mES cells in vitro generated only a small fraction of cells (0.1%) with characteristics of insulin-producing -like cells (329). This percentage has been increased by the selection of nestin-positive progenitor cells, the products of which showed regulated insulin release in vitro. The insulin-positive clusters, however, failed to normalize high blood glucose levels in transplantation experiments (213). Indeed, subsequent analyses revealed that these insulin-positive cells may be partially committed to a neural fate (330) and are characterized by small, condensed nuclei and are apoptotic. Rather than producing insulin themselves, most of the cells took up this hormone from the culture medium (283).
By modifying the differentiation protocols and using genetically modified mES cells, two groups successfully generated insulin-producing cells (38, 207). Blyszczuk et al. (33) showed that constitutive expression of the pancreatic developmental control gene Pax4 and histotypic differentiation were essential for the formation of insulin-expressing cells, which were found to contain secretory granules typical of both embryonal and adult -cells. Importantly, these cells coexpressed C-peptide and normalized blood glucose levels after transplantation into diabetic mice (37, 38). Similarly, lineage selection using mES cells transfected with a plasmid containing the Nkx6.1 promoter upstream of a neomycin-resistance gene could be used to generate insulin-producing cells that normalized glycemia after transplantation into diabetic animals (207).
Also, the treatment of mES cells with a phosphoinositide 3-kinase (PI 3-K) inhibitor during terminal stages of differentiation generated ES cell progeny expressing various -cell-specific markers. Following engraftment into diabetic mice, these cells also improved the glycemic status and enhanced animal survival (162).
Initial experiments with hES cells indicate that in vitro differentiation generates 1% insulin-secreting cells that show at least some characteristics of pancreatic endocrine cells (13). Treatment of hES cells with NGF results in upregulation of the Pdx-1 gene, the product of which controls insulin transcription and regulates insulin release (323). A modification of the differentiation protocol (see Refs. 213, 283) allowed the generation of insulin-producing clusters from hES cells (324), but further improvements are necessary for generating functional isletlike cells.
D. Germ Cell Differentiation
Only recently has the use of a suitable reporter system allowed the visualization of germ cell formation in vitro. Hübner et al. (164) used regulatory elements (CR2 and CR3) within the germ-line specific (gc) Oct4 gene to visualize initial steps of germ cell formation. To restrict expression of an Oct-3/4-based reporter to germ cells, a genomic gcOct-3/4-GFP construct was introduced into ES cells and cultured at high density. Colonies of variable size formed after 12 days, and GFP-positive cells that expressed Vasa (a marker of postmigratory germ cells) formed small aggregates in the supernatant. The isolation and further culture of these aggregates resulted in well-organized structures, morphologically similar to early ovarian follicles. The formation of these oocyte-like structures was paralleled by estradiol synthesis providing evidence for functional activity of somatic (granulosa) cells in the cultures. A detailed analysis of the oocyte-like cells showed that they were fragile and expressed the zona pellucida proteins (ZP) 2 and 3. The loss of ZP1 expression may account for the fragile zona of in vitro-derived oocyte-like cells, because loss of ZP1 has been shown to perturb folliculogenesis. Continued cultivation of oocyte-like cells, until day 43, revealed structures similar to mouse preimplantation embryos. It is likely that these blastocyst-like structures represent parthenotes, as suggested by the similarity of the follicle outgrowths (164).
Two other reports describe the formation of male germ cells that have the capacity to participate in spermatogenesis in vivo after engraftment (365) and to fertilize oocytes (136). In the latter study, EBs supported the maturation of primordial germ cells into haploid male gametes, which when injected into oocytes restored the somatic diploid chromosome complement and developed into blastocysts. EG cells show erasure of the methylation markers (imprints) of igf2r and H19 genes, a property characteristic of the germ line. Because these data would essentially close the developmental circle that connects ES cells with the germ line, it may be necessary to redefine the ES cell potential in vitro (totipotency versus pluripotency). Moreover, this property of ES cells provides an accessible in vitro model system for studies of germ-line epigenetic modifications and mammalian gametogenesis and should reveal whether the in vitro generated oocytes may be used as starting material to reprogram somatic cell nuclei. If similar processes can be induced in hES cells, this would open a new perspective to the generation of therapeutically relevant tissues by the "therapeutic cloning" approach (see sect. IXB). Indeed, hES cells are able to spontaneously develop into cells representative of meiotic and postmeiotic germ cell development. The in vitro differentiation of hES cell lines as EBs resulted in the formation of VASA-positive cells and the upregulation of transcripts of the meiotic markers SCP1 and SCP3 (synaptonemal complex protein) and the postmeiotic markers GDF9 (growth and differentiation factor) and TEKT1 (tektin). In contrast to mES cells, in vitro differentiated hES cells express both the male and female genetic programs regardless of whether they were karyotypically XX or XY: both GDF9 (oocyte-specific) and TEKT1 (spermatid-specific) expression was detected with differentiation of hES cells (82).