Almost 50 years ago, it was realized that aggregates of pluripotent embryocarcinoma cells recapitulated aspects of early embryonic development (Pierce and Dixon, 1959
; Stevens, 1959
). This was quickly recognized as a model system for the understanding of normal embryonic cell differentiation. These aggregates—termed embryoid bodies because of their superficial resemblance to mouse blastocysts—generally show few morphological signs of gastrulation and display unorganized tissue differentiation (Martin et al., 1977
; Wiley et al., 1978
). However, this issue has not been readdressed since the advent of ES cells with their superior ability to generate embryonic tissues (Evans and Kaufman, 1981
; Martin, 1981
). Our data show that embryoid bodies derived from ES cells display a large degree of self-organization: they establish anteroposterior polarity and develop a domain with characteristics of the primitive streak, where cells undergo EMT and form mesendoderm precursors in a process that is dependent on local activation of the Wnt pathway. Embryoid body development therefore resembles normal embryonic development much closer than previously thought, and provides an easily accessible model for the formation of anteroposterior polarity and the establishment of the primitive streak. Moreover, the presence of self-organization and polarity suggests that embryoid bodies can establish morphogen gradients controlling cell differentiation. This would not only explain the wide repertoire of developmental activities present in the embryoid body, but would provide a new tool for investigating the establishment and mode of action of such gradients.
, primitive streak formation and the establishment of anteroposterior polarity depend on interactions between the epiblast and two extraembryonic tissues, the visceral endoderm and the extraembryonic ectoderm (reviewed in (Tam and Loebel, 2007
)). Primitive-streak formation requires specification of the posterior epiblast by Wnt3 (Huelsken et al., 2000
; Liu et al., 1999
). It is thought that expression of Wnt3 is activated by nodal, through BMP4 signaling in the extraembryonic ectoderm (Ben-Haim et al., 2006
; Brennan et al., 2001
). In turn, Wnt3 activates a feedback loop that maintains nodal expression in the epiblast (Ben-Haim et al., 2006
). Since all three factors can start the feedback loop, this explains why they were all able to initiate self-organization in the embryoid body. However, our results indicate that Bmp4 is not required for this process, and that nodal/ActivinA can activate Wnt signaling in absence of a functional Bmp pathway. This conclusion is in agreement with a recent study that found that Bmp4 has a posteriorizing effect on mesoderm, but is not required for generation of mesoderm in embryoid bodies (Nostro et al., 2008
). In vivo
, loss of Bmp4 severely affects gastrulation and establishment of the primitive streak (Winnier et al., 1995
). Combined, these results suggest that Bmp4 is necessary for induction of Wnt3 and nodal signaling in the embryo, but is not longer required once these pathways are active. This also implies that Bmp4, not nodal, is the signal that starts the feedback loop in vivo
During embryogenesis, the visceral endoderm promotes anterior patterning by producing nodal and Wnt antagonists, including Cerl, Lefty1, and Dkk1, in the anterior region of the embryo (Glinka et al., 1998
; Kimura-Yoshida et al., 2005
; Perea-Gomez et al., 2002
; Yamamoto et al., 2004
). Establishment of anteroposterior polarity is therefore the result of a balance between posteriorizing signals and their antagonists. Using Wnt3a and Dkk1 we could manipulate this balance in embryoid bodies to favor either posterior or anterior character. By providing exogenous Wnt proteins we posteriorized the embryoid body and promoted mesendodermal fate. Conversely, Wnt antagonists promoted anterior character and neurectodermal differentiation. The status of the Wnt signaling pathway is therefore a critical parameter in protocols for directed ES cell differentiation.
How is the initial polarity of the embryoid body established? A deterministic mechanism seems unlikely, since the embryoid body consists of cells that cannot be distinguished by virtue of lineage or inductive environment. In a stochastic mechanism, differences in cells arise from developmental noise, which can have multiple sources. These differences are amplified and stabilized, and can lead to asymmetric cell fate decisions in a uniform cellular assembly (Losick and Desplan, 2008
). In the great majority of the embryoid bodies we studied, a single domain of Wnt signaling slowly expanded throughout the embryoid body. This suggests a cell-nonautonomous decision-making system, in which cells nearby the original inducing cell are recruited into the polarizing center, whereas an inhibiting signal, acting over a longer range, prevents formation of multiple polarizing centers. Indeed, in larger embryoid bodies, produced by aggregating more cells, we found more instances of embryoid bodies with two domains of Wnt signaling (not shown and , 3 days vehicle). This suggests that the range of the putative inhibitory signal was insufficient in these larger embryoid bodies. Nodal and the Wnt signal itself are candidates for the recruiting signal, since a single, low intensity pulse of these signals suffices to induce a polarizing center. A better understanding of this phenomenon could have implications for the in vivo
establishment of anteroposterior polarity.
A persistent problem in understanding the regulation of cell differentiation in ES cell cultures is the presence of fetal calf serum, which consists of undefined mixtures of growth factors and inhibitors. We show that an important role of serum in embryoid body differentiation is to provide a Bmp activity that activates Wnt signaling and starts primitive streak formation. For this function, we could replace serum-containing medium with purified growth factors in chemically defined medium. The ability to differentiate ES cells in defined conditions should facilitate the derivation of pure cell populations from ES cells.