Abstract
The differentiation of hepatic and pancreatic progenitor cells during embryogenesis is determined by inductive factors secreted by neighboring cells. These factors stimulate and repress the expression of key regulatory genes in progenitor cells, thereby establishing unique genetic programs that determine cell fate. The signaling network that controls liver and pancreas development is highly dynamic with respect to both concentration and timing of exposure to several key inductive factors. Not only do large changes occur within short time frames, multiple signaling pathways also converge on the same target genes. Given the intense effort under way to generate certain differentiated cell types from both embryonic and induced pluripotent stem cells, greater understanding of how different inductive signals interact with each other may be essential for the eventual success of such efforts.
During embryogenesis, both the liver and the ventral pancreas arise by the budding of diverticula from the ventral foregut endoderm that contain multipotent progenitor cells. These progenitor cells undergo rapid changes in their gene expression profiles as they proliferate, invade the surrounding mesenchyme, become progressively fate restricted, and ultimately form complex, three-dimensional architectures that characterize the adult liver and pancreas. In mice, the expression of genes that identify either liver- or pancreas-specific progenitor cells begins around embryonic day 8 to 8.5, a time that coincides with the presence of seven to eight somites (1, 2). However, beginning as early as day 7.5 or 8 or at the one-to-three (1-3) somite stage, spatially separated regions of definitive endoderm (3) are already being programmed by a panoply of signals originating from surrounding cells of mesodermal origin. These signals stimulate gene expression changes that restrict the ability of foregut endodermal cells to contribute to the hepatic or pancreatic buds. A study provides insight into the complexity and dynamic nature of this signaling network (4) and has substantial implications for mimicking these and other developmental processes in the culture dish.
The signaling network essential for programming the liver and the pancreas consists of several canonical signaling pathways, and probably other factors that remain to be identified. These signals may have different effects depending on the endodermal domain. For instance, studies in embryonic explants have shown that bone morphogenetic proteins (BMPs) from the septum transversum mesenchyme (5) and fibroblast growth factors (FGFs) from the cardiac mesoderm (6) stimulate formation of the liver while suppressing pancreas development. Moreover, both the concentration threshold and the isoform specificity of BMP and FGF signals are critical for their effects (5, 6). Indeed, different FGFs contribute to patterning of the ventral foregut into lung, liver, and pancreas (7). During early developmental stages, when the ventral foregut endoderm is in close proximity to cardiac mesoderm, the pancreas program is suppressed because of the induction of Sonic hedgehog (Shh) by FGF (8). However, the suppressive effect of FGF is relieved as prospective pancreatic progenitor cells migrate to reside at the midline, away from the cardiac mesoderm, thereby allowing expression of pancreas-specific genes. Although the combination of different signaling gradients has provided a model for explaining how the ventral foregut is partitioned into liver and pancreas, this view now appears to be overly simplistic.
Experiments reported by Wandzioch and Zaret (4) indicate that there is much greater complexity and temporal precision to the signaling events required for hepatic and pancreatic organogenesis. By using a half-embryo culture system, the authors made use of pathway-specific agonists and antagonists to investigate the necessity, sufficiency, and interdependence of the signaling networks that determine specification of hepatic and pancreatic progenitor cells. The half-embryo culture system enabled precise monitoring of foregut endoderm morphogenesis ex vivo, including counting of the number of somites. Because somites expand in number every 2 hours in the mouse, they provide a highly accurate index of developmental stage (9). This strategy enabled va.rious genetic and pharmacological perturbations to be performed while preserving the necessary three-dimensional cellular relationships that are vitally important in the developing embryo.
By using phospho-specific antibodies to known targets of FGFs [extracellular signal-regulated kinase (ERK) 1 and 2], BMPs (Smad1, 5, and 8), and transforming growth factor–β (TGF-β) (Smad2), Wandzioch and Zaret (4) demonstrated a highly dynamic interplay between spatially separated FGF and BMP activation domains (Fig. 1). At the 3-4 somite stage of development, FGF target genes are activated in lateral endoderm, whereas BMP targets are activated in medial endoderm. These differences indicate that, at the 3-4 somite stage, pancreatic progenitors initially reside in a region of low BMP activity and eventually merge to form a ventral pancreatic bud at the midline in a region of high BMP target gene expression. This correlates with inhibition of BMP signaling by the antagonist Noggin at the 3-4 somite stage, resulting in increased expression of pancreas-specific genes. However, administering Noggin only a few hours later, at the 5-6 somite stage, results in decreased pancreas-specific gene expression. These results highlight the crucial importance of timing for inducing the pancreatic genetic program.
A complex signaling network governs specification of hepatic and pancreatic progenitor cells. (A) At the 3-4 somite stage, FGF and BMP target genes are expressed in separate domains of the ventral foregut endoderm, reflecting the presence of a signaling gradient. Initially, pancreatic progenitors reside in a region of high FGF and low BMP activation. However, at the 5-6 somite stage, as these progenitors migrate to the midline to form the ventral pancreatic bud, they come to reside in a region of high BMP activation. In contrast, hepatic progenitors initially reside within spatially separate domains (medial and lateral endoderm) and are influenced by different signaling networks. Together, these early signals are crucial for specifying liver and pancreatic cells. (B) Liver gene expression is decreased by inhibition of BMP and increased by stimulation of BMP signaling. Because the amplitude of liver markers is increased if BMP stimulation occurs later, BMP signaling is prohepatic, reinforcing the liver program. Modulating BMP at distinct time points over a span of a few hours induces differential expression of pancreas-specific genes. Although BMP inhibition at the 3-4 somite stage increases pancreatic markers, BMP inhibition at the 5-6 somite stage decreases pancreatic markers. This difference suggests that presumptive pancreatic progenitors require BMP signaling as they move to the ventral midline to contribute to the pancreatic bud. Similarly, TGF-β signaling may also influence the formation of pancreas progenitors at the 5-6 somite stage: Pancreatic markers are increased by inhibition of ALK (activin receptor-like kinase) 4, 5, and 7 and decreased by TGF-β2 treatment.
In marked contrast, spatially distinct hepatic domains initially reside in a region of high BMP activity at the ventral midline and high FGF activity at the lateral endoderm. Although one might expect that modulating BMP would differentially affect these hepatic domains, exogenous induction of BMP signaling at both the 3-4 and 5-6 somite stages results in a pro-hepatic signal, as evidenced by increased liver-specific gene expression. Thus, the timing of BMP signals for generating hepatic progenitor cells may be less critical than it is for the pancreatic program.
Additionally, Wandzioch and Zaret (4) reveal a requirement for TGF-β signaling in confining the number of pancreatic progenitors that are specified from the ventral foregut endoderm. Strikingly, the addition of TGF-β2 to the half-embryo cultures at both 3-4 and 5-6 somite stages had a stronger inhibitory effect on pancreatic versus hepatic early genetic programs. Thus, TGF-β signaling may play a role in either the specification or expansion of a specific pancreatic progenitor cell state. In any case, these experiments clearly point to greater complexity in how different BMP, FGF, and TGF-β signals induce the ventral foregut endoderm to form liver and pancreas, thereby raising many new questions.
Clarification of the dynamic nature of the signaling process that specifies liver and pancreas is of great interest, given the well-established and documented need to develop sources of hepatocytes and pancreatic β cells for therapeutic use (10–12). Although rapid and considerable progress is being reported in developing methods to differentiate embryonic stem (ES) and induced pluripotent stem cells toward hepatic and pancreatic cell fates, the efficient generation of liver and pancreatic lineages may require precisely mimicking the in vivo signaling network. Current protocols for making pancreatic β cells from ES cells use 1- to 3-day exposure times to factors at fixed concentrations. However, the resulting cells are immature and do not secrete insulin in response to changes in the concentration of glucose unless they are matured in vivo for an extensive period of time (13, 14). As suggested by Wandzioch and Zaret (4), a possible explanation may be our inability to apply key developing factors in a manner that fully mimics how hepatocytes and β cells are generated in the embryo. Indeed, the accurate programming of stem cells toward hepatic and pancreatic fates may not only require the activation of particular pathways in a temporally precise manner, but also the simultaneous inhibition of other signaling pathways.
The results obtained by Wandzioch and Zaret (4) clearly indicate that much remains to be learned about the programming and expansion of liver and pancreatic progenitor cells. For instance, current protocols for making pancreatic β cells do not result in nearly enough differentiated cells for clinical use. Thus, we need more knowledge about the expandability of progenitor cells at different times, and we need to extend the timing of the protocols in ways that will allow a greater mass of differentiated cells to be made. Similarly, we need to determine whether other factors are involved and how they contribute to both the specification and differentiation of hepatic and pancreatic progenitor cells. Also, what is the role for Wnt and Notch, both of which contribute to both hepatic and pancreatic organogenesis (15, 16)? Lastly, there is a need for new methods that provide real-time readouts of multiple signaling pathways. Such tools might open an even wider window into the myriad of signaling events that may ultimately be involved in the specification and maturation of endodermal progenitors.
