Early in mouse embryo development, the inner cell mass proliferates rapidly and differentiates to generate a population of pluripotent cells known as primitive ectoderm. Shortly after implantation, the innermost cells of the primitive ectoderm cell mass undergo apoptosis and form a cavity through a process known as cavitation (Coucouvanis and Martin 1995). The surviving primitive ectoderm cells that surround the cavity differentiate to form a pseudostratified columnar epithelium, giving rise to a structure known as the epiblast (Fig. 4). Although they retain pluripotentiality, primitive ectoderm/epiblast cells can be distinguished from the inner cell mass on a morphological basis, by the fact that they have upregulated fgf5 expression (Haub and Goldfarb 1991; Hebert et al. 1991) and down-regulated rex1 expression (Rogers et al. 1991), and by the fact that they can no longer contribute to chimera formation following blastocyst injection (Rossant 1977; Beddington 1983). The epiblast responds to extrinsic signals and gives rise to the primary germ layers (Gardner and Rossant 1979) as well as the primordial germ cells (Ginsburg et al. 1990).
The primary germ layers in the embryo are formed during the process of gastrulation, which begins at approximately embryonic day 6.5 (E6.5) in the mouse (Tam and Behringer 1997
). At the onset of gastrulation, the epiblast cells in the region that will define the posterior part of the embryo thicken and form a transient structure known as the primitive streak (Fig. 4
). During gastrulation, epiblast cells traverse the primitive streak and undergo an epithelial to mesenchymal transition, giving rise to mesoderm and definitive endoderm. Fate-mapping studies have demonstrated that the development of subpopulations of mesoderm and endoderm is not random but rather appears to be controlled, both temporally and spatially (Kinder et al. 1999
). For instance, the first mesoderm cells to develop are derived from epiblast cells that move through the most posterior region of the streak. This mesoderm colonizes the developing yolk sac and gives rise to the hematopoietic and vascular cells of the blood islands. Epiblast cells that traverse the streak slightly later and in a more anterior position will form cardiac mesoderm, head mesenchyme, and paraxial mesoderm, whereas epiblast cells that move through the most anterior region of the primitive streak will generate endoderm and axial mesoderm, which gives rise to the notochord (Tam and Behringer 1997
). Cells in the most anterior region of the epiblast that do not move through the primitive streak will form ectoderm. While the mechanisms controlling epiblast migration and mesoderm and endoderm induction are not fully understood in the mouse, expression analysis and gene targeting studies have shown that members of the TGF
family including BMP4 (Hogan 1996
) and nodal (Conlon et al. 1994
; Schier and Shen 2000
) as well as members of the Wnt (Yamaguchi 2001
) and FGF (Yamaguchi and Rossant 1995
; Sun et al. 1999
) families play important roles in these processes.
Modeling embryonic development with ES cells
Cells with characteristics of primitive ectoderm have been identified in ES cell differentiation cultures. Rathjen et al. (1999) discovered that media conditioned by a hepatocellular carcinoma cell line could promote the synchronous differentiation of ES cells to a homogeneous population of cells that grew as a monolayer and expressed fgf5 but not rex1. These cells, named early primitive ectoderm-like (EPL) cells, were unable to contribute to chimerism following injection into blastocysts. When differentiated in culture, EPL cells formed EBs and generated mesoderm and derivative cell populations more rapidly and efficiently than EBs from ES cells (Lake et al. 2000). These findings support the interpretation that EPL cells are similar to the primitive ectoderm of the embryo and, as such, represent the initial stage of differentiation from ES cells. While these studies demonstrate that this crude conditioned medium promotes the development of EPL cells from ES cells, the factors and signaling pathways regulating this differentiation step in culture remain to be determined.
Although it is unlikely that a structure comparable to the primitive streak develops in ES cell cultures, it will be important to mimic the induction events associated with different regions of the streak that lead to the development of specific cell populations. To date, the regulation of early germ layer induction in ES cell cultures remains poorly understood, as most studies have carried out the differentiation in the presence of FCS, which consists of undefined mixtures of growth factors and inhibitors. The limited number of studies that have investigated germ layer induction in ES cell differentiation cultures have provided evidence that the same sets of molecules regulating these decisions in the embryo are also active in vitro.
ES cells will generate neuroectoderm when differentiated as EBs or as a monolayer in the absence of serum and inducing molecules (Wiles and Johansson 1999; Ying et al. 2003b). Using ES cells with the GFP cDNA targeted to the neuroectoderm specific gene, Sox1, Ying et al. (2003b) were able to demonstrate that up to 60% of the cells in monolayer cultures formed neuroectoderm in serum-free cultures. Differentiation to neuroectoderm in these cultures did not appear to represent a "default" pathway but rather was dependent on FGF signaling (Ying et al. 2003b). Addition of BMP4 to serum-free cultures inhibited the development of neuroectoderm and led to the induction of mesoderm (Finley et al. 1999; Wiles and Johansson 1999; Ying et al. 2003b). Addition of serum resulted in a similar inhibition of neuroectoderm development. Further support for a neural inhibitory role for BMP has been provided by studies demonstrating that expression of the BMP antagonist noggin in ES cells promotes neural differentiation (Gratsch and O'Shea 2002). The inhibition of neural differentiation by BMP4 in the ES cell model is consistent with findings in Xenopus that show that noggin and chordin, inhibitors of BMP4, induce neural development (Sasai et al. 1995; Wilson et al. 1997).
Induction with both BMP4 and VEGF in serum-free ES cell differentiation cultures resulted in the development of hematopoietic cells, indicating that the equivalent of posterior streak mesoderm was generated (Nakayama et al. 2000; Park et al. 2004). Further characterization of these populations will be required to determine if other mesoderm types are generated under these conditions. In addition to BMP, Wnt signaling also appears to play a pivotal role in germ layer induction in the ES cell cultures (Aubert et al. 2002). Inhibition of Wnt through the expression of the Wnt inhibitor Sfrp2 led to enhanced neural development in EBs. Conversely, expression of Wnt1 inhibited neural development.
To be able to track mesoderm induction during ES cell differentiation, we targeted the GFP cDNA to the brachyury gene, which is expressed throughout the primitive streak as well as in all nascent mesoderm (Fehling et al. 2003). Using this reporter line, it was possible to demonstrate that batches of serum selected for optimal hematopoietic development also induced significant amounts of mesoderm, as up to 80% of the total EB population expressed GFP by day 4 of differentiation. A typical induction profile over a 3-d differentiation period is shown in Figure 5. With this ES cell line, it is easy to quantify the development of the brachyury-expressing population under different conditions. This model will also enable one to determine if a particular factor is functioning at the level of mesoderm induction (brachyury expression) or its specification to a particular cell type. In serum-free cultures, brachyury was not induced, whereas genes indicative of neuroectoderm development were expressed, an observation consistent with the findings from the studies outlined above (Kubo et al. 2004).
In addition to serum, activin also induced high levels of GFP (Fig. 5). However, the populations that developed following activin induction differed from those following serum induction. Whereas serum efficiently generated mesoderm that gave rise to hematopoietic and vascular cells, activin induced a range of cell types that differed depending on the concentration of factor used (Kubo et al. 2004). Low concentrations of activin gave rise to skeletal muscle indicating induction of paraxial mesoderm, whereas high concentrations led to the development of definitive endoderm and derivative cell types. Increasing concentrations of activin resulted in a down-regulation of neuroectoderm genes, suggesting that, as with BMP4 and Wnt, activin signaling inhibited neural development in these cultures. Although activin did not induce hematopoietic cells directly, when the treated EBs were transferred to serum cultures for several days, this lineage did develop. Cell sorting studies revealed that both the mesoderm and endoderm populations developed from brachyury-positive cells. This is an important observation as it suggests that activin leads to the development of populations equivalent to those that are induced in different regions of the primitive streak (Fig. 6). High concentrations of activin induce the differentiation of cells similar to those generated in the anterior streak in the embryo, whereas lower concentrations induce cell types that are formed in the more posterior regions of the streak.
The demonstration that activin can induce both mesoderm and endoderm in ES cell cultures is consistent with findings in the Xenopus system, where it has been shown to induce these lineages from animal cap cells in culture (Green et al. 1992; Gamer and Wright 1995; Ninomiya et al. 1999). While these studies demonstrate that activin can induce mesoderm and endoderm in model systems, gene-targeting studies in mice indicate that it is not the endogenous factor that regulates these developmental decisions in the early embryo (Vassalli et al. 1994; Matzuk et al. 1995). One interpretation of these findings is that these activities of activin are likely mimicking the function of nodal, as both factors can bind the same receptors and thus initiate the same signaling events (Schier and Shen 2000). Unlike activin, nodal is essential for primitive streak formation as well as early mesoderm and endoderm development in vivo (Schier and Shen 2000; Whitman 2001).
The studies outlined in this section have only begun to probe the mechanisms that regulate the earliest stages of lineage development in ES cell differentiation cultures. The findings from them do, however, highlight the importance of understanding these early events, as the inducing molecules used in the differentiation cultures dramatically influence the cell populations that ultimately develop. These observations also serve a cautionary note that one set of conditions cannot be used to optimally generate progeny of all three germ layers. A model of ES cell differentiation, with the effects of the known inducers, is shown in Figure 6.