|Home | About | Journals | Submit | Contact Us | Français|
The endoderm is a multipotent progenitor cell population in the embryo that gives rise to the liver, pancreas, and other cell types and provides paradigms for understanding cell type specification. Studies of isolated embryo tissue cells and genetic approaches in vivo have defined FGF/MAPK and BMP signaling pathways that induce liver and pancreatic fates in the endoderm. In undifferentiated endoderm cells, the FoxA and GATA transcription factors are among the first to engage silent genes, helping to endow competence for cell type specification. FoxA proteins can bind their target sites in highly compacted chromatin and open up the local region for other factors to bind; hence they have been termed "pioneer factors". We recently found that FoxA proteins remain bound to chromatin in mitosis, as an epigenetic mark. In embryonic stem cells, which lack FoxA, FoxA target sites can be occupied by FoxD3, which in turn helps maintain a local demethylation of chromatin. By these means, a cascade of Fox factors helps endow progenitor cells with the competence to activate genes in response to tissue-inductive signals. Understanding such epigenetic mechanisms for transcriptional competence coupled with knowledge of the relevant signals for cell type specification should greatly facilitate efforts to predictably differentiate stem cells to liver and pancreatic fates.
The activation of a particular cell type program within multipotent progenitor and stem cells is perhaps the most dramatic of gene regulatory events: it enables all subsequent gene regulatory events specific to a lineage while generally excluding all other cell type programs available to the progenitor cell. While cells within a blastula or embryonic stem cells are pluripotent and thus have all embryological fates available to them, after gastrulation, cells of the ectoderm, endoderm, and mesoderm lineages are more restricted in their potential fates; and derivatives of each of these germ layers have successively fewer fates choices available. Nonetheless, any cell with an alternate fate choice has at least two parameters governing the cell type decision: signals that provide a "go" to make or allow a decision, and the intrinsic competence of the genome, in terms of its chromatin state, to respond to the signal. Our laboratory investigates both of these areas for the initiation of the liver and pancreatic programs from the endoderm. Understanding the basis for cell type specification will provide insight into normal development, homeostatic self-renewal within the adult tissues, regeneration upon tissue damage, and the prospective programming of stem cells and other progenitor cells to these biomedically relevant tissue types.
Liver and pancreas cells are derived from the foregut endoderm. Our fate mapping studies demonstrated that the liver bud is derived from paired lateral domains of foregut endoderm as well as a physically separated domain of ventral-medial endoderm (Tremblay and Zaret 2005). Although both the lateral and ventral-medial domains give rise to liver bud cells that express early liver genes, including Alb1, afp, and Hnf4, it remains to be determined if descendants from the different progenitor domains have different functions or regenerative capabilities in adult tissues. The earliest cells to express liver genes are called hepatoblasts; later they differentiate into hepatocytes and cholangiocytes (bile duct cells) (Shiojiri 1981; Zaret 2008). Similarly, the pancreas is derived from two domains of endoderm. In this case, the caudal portion of the paired lateral, prospective liver domains of endoderm also give rise to the ventral pancreatic bud (Tremblay and Zaret 2005) and a separate domain of dorsal endoderm, positioned near the notochord, gives rise to the dorsal pancreatic bud (Slack 1995). Later in development, both pancreatic buds merge to create the gland and descendants of both embryonic origins give rise to exocrine and endocrine cell types. Endocrine cells differentiate into five different cell types that are each specialized to express a single hormone. Beta cells are the most abundant and important pancreatic endocrine cell type; they secrete insulin into the bloodstream in response to high blood glucose concentrations, causing body tissues to store glucose after a meal. By contrast, each hepatocyte in the liver has many functions, including the secretion of hormones, serum proteins, and bile salts, the metabolism of nutrients and toxicants, and the storage of glucose.
Tissue explant studies in mouse and chick demonstrated that the foregut endoderm cells are a multipotent population, in that the cells can be induced to initiate gene expression programs for the liver or pancreas fate in response to FGF, BMP, and shh signaling from different overlying types of mesodermal cells (Kim et al. 1997; Hebrok et al. 1998; Jung et al. 1999; Deutsch et al. 2001; Rossi et al. 2001). Specifically, cardiogenic mesoderm cells secrete FGFs as a hepatogenic signal (Jung et al. 1999), which activates liver genes in the adjacent ventral foregut endoderm via MAPK pathway signaling (Calmont et al. 2006); such signaling suppresses the ventral pancreatic program (Deutsch et al. 2001) (Figure 1). These findings provide a molecular explanation for the original discovery with chick embryos that cardiogenic mesoderm induces hepatogenesis in the endoderm (Le Douarin 1975).
We found that septum transversum mesenchyme cells in the vicinity of the cardiogenic mesoderm cells secrete BMPs, which also promote hepatic gene induction in the endoderm (Rossi et al. 2001) (Figure 1). Such BMP signaling also promotes ventral pancreas induction (E. Wandzioch and K. Zaret, in preparation). Lateral plate mesoderm cells also promote ventral pancreatic induction (Kumar et al. 2003). Recent data from Xenopus indicates that Wnt signaling must be suppressed in the foregut endoderm to allow hepatic specification (McLin et al. 2007). Later, Wnt signaling promotes the outgrowth of the liver bud (Ober et al. 2006). The dorsal endoderm expresses shh, which in turn is inhibitory to the dorsal pancreatic program; signals from the notochord suppress endodermal shh expression and allow dorsal pancreas specification (Kim et al. 1997; Hebrok et al. 1998). Liver and pancreas both have endocrine function, and their development is tightly coordinated with that of the vasculature; more specifically, after the initial specification of the liver and pancreas progenitors, their morphogenetic bud development is dependent upon signals from adjacent endothelial cells (Lammert et al. 2001; Matsumoto et al. 2001; Yoshitomi and Zaret 2004). In summary, diverse signals and cell interactions induce the initial genetic programs for liver and pancreas. Indeed, this information has been used to prospectively differentiate embryonic stem cells to liver and pancreas fates (Fair et al. 2003; Teratani et al. 2005; D'Amour et al. 2006; Gouon-Evans et al. 2006).
Evidence that the foregut endoderm cells are truly multipotent, capable of initiating either liver or pancreas fates, comes from tissue explant studies as well as genetic studies of the Hex mutation in mice. Isolated foregut endoderm, along with associated septum transversum mesenchyme cells, readily induce early pancreatic genes in culture (Deutsch et al. 2001). However, inclusion of cardiac mesoderm in the endoderm explants, or treatment of the explants with low concentrations of FGF-2, induces liver genes in the explants and suppresses pancreatic gene induction. Changes in proliferation or cell death are not observed. Thus, the default program for foregut endoderm explants is to initiate the pancreatic program, and cardiac-FGF signals seem to divert the cells to a hepatic fate.
In a different line of research, homozygous null Hex mutants exhibit a defect in liver development after the initiation of the hepatic program and formation of the liver bud (Bort et al. 2004). Interestingly, the liver bud cells fail to continue their differentiation and revert to a gut-like fate (Bort et al. 2006). However, in the Hex null embryos, ventral pancreas genes exhibit a complete failure to be activated (Bort et al. 2004). Further studies showed that the Hex mutation causes cell morphogenetic and movement defects, so that the prospective ventral pancreatic endoderm domain fails to move beyond the cargiogenic domain, which, in turn, normally induces the liver (see above). We found that isolation of the foregut endoderm from Hex mutant embryos and culturing it in vitro, in the absence of cardiogenic mesoderm, allowed the normal induction of early pancreatic genes in the mutant endoderm (Bort et al. 2006). Differences in growth or cell apoptosis were not observed. It thus appears that in Hex null embryos, the ventral pancreatic fate is suppressed in the endoderm by cardiac, hepatogenic signaling, but the endoderm cells retain the competence to initiate the pancreatic program. Thus, foregut endoderm cells are bipotential with regard to liver and pancreas fates; and in the Hex mutant embryos, the nascent liver cells later revert to a gut fate, indicating further multipotency. These findings raise the question of how the cells gain the potential to activate the different cell fates.
Upon discovering that the Alb1 locus in mouse embryos is activated in the endoderm by the earliest hepatogenic signals (Gualdi et al. 1996; Jung et al. 1999), we have used regulatory sequences of Alb1 as sentinels of transcription factor occupancy during hepatic cell type specification. The Alb1 gene contains an upstream enhancer sequence that binds numerous liver-enriched transcription factors and confers liver-specific transcription upon linked genes in transgenic mice and transfected cells (Pinkert et al. 1987; DiPersio et al. 1991; Liu et al. 1991; Jackson and Benyajati 1993). Our laboratory performed in vivo footprinting studies on the enhancer in adult liver cells, embryonic liver buds, and undifferentiated endoderm, as well as in control tissues in which Alb1 is not expressed. We found that most of the key transcription factor binding sites in adult hepatocytes are occupied all the way back, temporally, to the embryonic liver bud stage (McPherson et al. 1993;Gualdi et al. 1996). Extending further to the endoderm, where the Alb1 gene is silent, we discovered only two of the binding sites occupied; for FoxA (formerly HNF3) and GATA factors (Bossard and Zaret 1998) (Figure 2).
Mammals have three unlinked FoxA genes (FoxA1, FoxA2, and FoxA3), of which FoxA2 is expressed in and necessary for endoderm development in apparently all metazoans (Zaret 1999; Davidson and Erwin 2006). Indeed, ectopic FoxA2 expression promotes endoderm development in ES cells (Ishizaka et al. 2002) and endogenous FoxA2 expression is used to efficiently monitor for endoderm formation from ES cells (Kubo et al. 2004). These results underscore the importance of developmental gene regulation studies for being able to predictably manipulate the fates of stem cells. Furthermore, after endoderm formation, FoxA1 and FoxA2 are redundantly necessary in the endoderm for liver induction (Lee et al. 2005). FoxA factors represent a subclass of the Fox transcription factor family; together, Fox factors function in diverse developmental and signaling contexts (Katoh 2004). GATA-4 and GATA-6 factors are expressed in the endoderm as well as other developing tissues and are redundantly necessary for early liver development (Holtzinger and Evans 2005; Zhao et al. 2005; Watt et al. 2007). Notably, throughout the developmental period in which the Alb1 enhancer sites for FoxA and GATA are occupied in the endoderm, the endoderm remains competent to activate the Alb1 gene (Bossard and Zaret 2000). The chromatin occupancy studies, together with the genetics, indicate that FoxA and GATA either mark or help maintain the competence of the endoderm to activate liver genes in response to inductive signals. The early engagement of these factors in a progenitor cell chromatin, prior to target gene activation, along with the factors' ability to help open chromatin structure (see below), has led them to be termed "competence" or "pioneer" factors (Zaret 1999). Furthermore, the mere engagement of target genes by a subset of enhancer binding factors may facilitate a more rapid and homogeneous activation of a genetic program in a field of progenitor cells in development, in response to inductive signals.
The crystal structure of the FoxA DNA binding domain immediately suggested a role for FoxA factors in chromatin, in that the fold of the protein was found to highly resemble that for linker histone (Clark et al. 1993; Ramakrishnan et al. 1993). Further studies showed that purified FoxA1 protein binds its target sites on reconstituted nucleosome core particles in vitro (Cirillo et al. 1998) and MNase mapping and sequential chromatin immunoprecipitation studies of liver chromatin showed that FoxA1 occupies its Alb1 enhancer target sequence on a nucleosome in vivo (McPherson et al. 1993; Chaya et al. 2001). Purified FoxA1 binding to compacted polynucleosome array templates in vitro creates a nuclease hypersensitive site where FoxA1 binds the DNA (Cirillo et al. 2002), showing that the protein opens up the local chromatin (Figure 3). Binding of FoxA1 protein to nucleosomal templates in vitro enables binding by GATA-4 and NF1 at adjacent sites (Cirillo and Zaret 1999; Cirillo et al. 2002). FoxA1 binding to chromatin in vivo also creates hypersensitive sites and enables estrogen receptor binding (Carroll et al. 2005). Nucleosome binding by FoxA1 and FoxA2 in vitro occurs equally well on substrates made with mouse histones or recombinant histones expressed in E. coli (Sekiya and Zaret 2007); thus FoxA1 does not require a prior histone modification for its intrinsic chromatin binding capacity. Yet in vivo, FoxA1 does not bind most of its consensus sites in chromatin and certain histone modifications may exclude or enhance FoxA1 binding to potential targets (Lupien et al. 2008). The ability of FoxA proteins to engage target sites in nucleosomal DNA and enable other factors to bind is consistent with its early, pioneer function in development.
Various Fox factors have now been found to possess chromatin opening activity and enable gene expression. Pha4, a FoxA homolog, enables the activation of the pharyngeal program in C. elegans embryos, with the strongest Pha4 binding genes turning on earliest in pharyngeal development, and genes that bind Pha4 weaker turning on later in pharyngeal development, as Pha4 levels continue to rise (Gaudet and Mango 2002). These findings would suggest that simple protein concentration in the nucleus is a major determinant of Pha4 binding. FoxI1 creates hypersensitive sites at its genetic targets in vivo (Yan et al. 2006), FoxE1 binds compacted chromatin in vitro and in vivo and generates local hypersensitivity (Cuesta et al. 2007), and FoxO1 binds and opens chromatin in a phosphorylation-sensitive fashion (Hatta and Cirillo 2007). Together, these findings indicate that Fox factors generally appear high in the hierarchy of the trascriptional regulatory mechanism, exposing target DNA sequences in chromatin for gene activity.
When a cell divides, in mitosis or meiosis, the extreme state of chromatin compaction excludes most regulatory proteins and leads to gene silencing (Gottesfeld and Forbes 1997). After mitosis, regulatory proteins must once again engage their genomic target sites and reestablish appropriate gene expression states. The means by which mitotic and meiotic chromatin are marked so that genes can be appropriately re-activated is a central aspect of epigenetic regulation. There has been intense focus on DNA methylation and certain histone modifications as being retained through mitosis and hence serving as epigenetic marks (Jenuwein and Allis 2001; Turner 2002; Bird and Macleod 2004; Goll and Bestor 2005; Kouzarides 2007). Yet, cytology, fluorescence recovery after photobleaching, and in vivo footprinting studies indicate that a small subset of transcription factors are also retained on mitotic chromatin (Martinez-Balbas et al. 1995; Michelotti et al. 1997; Chen et al. 2002; Christova and Oelgeschlager 2002). Such factors would be predicted to be able to bind highly compacted chromatin in other contexts, exactly has been found for Fox factors (Figure 3). Indeed, FoxI1 has been directly visualized on metaphase chromosomes, suggesting that the factor could serve as an epigenetic mark (Yan et al. 2006).
To investigate whether FoxA factors could be maintained in mitotic chromatin, we created GFP fusion proteins with FoxA1 and with C/EBPα and the nuclear localization signal (NLS) of SV40 large T antigen, as controls. Plasmids encoding the constructs were transiently transfected into HUH7 hepatoma cells and the cells were treated with aphidicolin (1 ug/ml) for 14 hr, then treated with nocodozole (0.2 ug/ml) for 16 hr, to enrich for cells in mitosis. The resultant cells were treated with DAPI and viewed under fluorescence optics to visualize the nuclear DNA state and location of GFP fusion proteins.
As seen in Figure 4A, B, cells which expressed the GFP-NLS and GFP-C/EBPα control proteins contained GFP fluorescence in nuclei that were decondensed and hence in metaphase. Cells that contained highly condensed chromatin, indicative of mitosis, had GFP-NLS and GFP-C/EBPα excluded from the bulk DNA (Figure 4A, B). By contrast, the GFP-FoxA1 protein was readily visualized in nuclei that were decondensed as well as in cells containing highly compacted mitotic chromatin (Figure 4C). As a secondary assessment of the mitotic state of the FoxA1-GFP transfected cells, we treated cultures with an antibody to histone H3 phosphorylated on serine 10, a mitotic marker. Cells containing highly compacted chromatin with bound FoxA-GFP also stained for the H3ser10 marker, demonstrating FoxA1 occupancy in mitotic chromatin (Figure 4C). Taken together, these studies demonstrate that FoxA1 can bind to mitotic chromatin, and hence its binding to DNA could serve as an epigenetic mark. We speculate that FoxA binding to mitotic chromatin could enable target genes to be more rapidly or synchronously activated as they progress toward interphase.
DNA methylation constitutes a major epigenetic mark that is typically associated with inactive genes (Bird and Macleod 2004). Like many genes, Alb1 is heavily methylated during early development, where the gene is silent, but becomes less methylated as the gene becomes activated, during hepatocyte differentiation (Kunnath and Locker 1983). Despite otherwise heavy methylation in early development, a single unmethylated CpG was discovered at the Alb1 enhancer in various embryonic stem (ES) cell lines, underlying one of the FoxA binding sites (Xu et al. 2007). Given that ES cell lines do not express FoxA proteins (data not shown), we investigated the potential role of FoxD3, which is expressed in ES cells (Sutton et al. 1996). FoxD3 is necessary to maintain ES cell pluripotency (Hanna et al. 2002; Tompers et al. 2005; Pan et al. 2006) and it functions in various other progenitor cell contexts in the embryo (Guo et al. 2002; Perera et al. 2006). Interestingly, chromatin immunoprecipitation studies revealed that FoxD3 engages the Alb1 enhancer in ES cells, where Alb1 is silent (Xu et al. 2007). This resembles the aforementioned occupancy by FoxA in endoderm cells, where Alb1 is also silent, but competent to be activated. Recent studies indicate that FoxD3 is necessary to maintain demethylation of the Alb1 enhancer site in ES cells, and that ectopic FoxD3 expression in fibroblasts induces Alb1 enhancer site demethylation (Xu et al., in preparation). In summary, it appears that a cascade of Fox factors engage the Alb1 gene, starting with FoxD3 prior to gastrulation and continuing with FoxA in the endoderm and its liver descendants (Figure 5). Coordinate with this cascade is the striking and selective demethylation at the Alb1 enhancer Fox binding site.
The combined chromatin mark consisting of a local CpG demethylation amidst highly methylated chromatin and its attendant bound transcription factor extends beyond FoxD3 and its targets. Studies of the Ptcra and Il12 genes, which are normally expressed in the thymocyte and macrophage lineages, show that their respective enhancers are also demethylated in ES cells, while other regions of these genes are methylated (Xu et al. 2007). Introduction of bacterial artificial chromosomes harboring the Il12 gene into ES cells, where CpG sequences within the BAC were methylated prior to transduction, showed that the enhancer sequences became demethylated selectively (Xu et al. 2007). In vivo footprinting reveals diverse factors bound to the demethylated enhancers in ES cells (Xu et al. 2007). Although the trans-acting factors that elicit demethylation are not yet known, mutagenesis of clusters of transcription factor binding sites causes the failure of transfected, methylated Ptcra and Il12 BACs to be demethylated at the enhancers (Xu et al., in preparation). The binding sites that were mutated do not appear to be Fox targets. These findings suggest that distinct trans-acting factors can elicit the local demethylation mark in embryonic stem cells.
Nature has evolved detailed cellular mechanisms by which multipotent cells come to possess particular developmental competencies as well as by which inductive signals and response networks elicit cell type specification. Further elucidation of the marks of developmental competence and their potential mechanisms of action will ultimately allow the prediction of differentiation capacity of a given progenitor or stem cell population. Thus, we wish to better understand chromatin states in embryonic endoderm cells, their progenitors, and in different stages of their descendants to the liver and pancreatic fates. To this end, we are currently employing fluorescence activated cell sorting of different endodermal and embryonic liver and pancreas cell populations to isolate progenitors, for detailed chromatin analysis. Given that a major difficulty with current stem cell differentiation protocols is to develop cells that fully express hepatocyte and beta cell phenotypes, we anticipate that the information we obtain can be used to assess whether cells at early stages in the differentiation protocol have been properly programmed; e.g. by possession of the appropriate chromatin competence marks. This could provide a novel dimension to prospectively programming stem cells to desired fates.
We are also taking a different perspective on how the engagement of pioneer transcription factors at silent genes may mark or endow the potential for gene activity. That is, we have performed a sub-genome wide location analysis of FoxA2 bound sites in the adult mouse liver, and focused on the approximately one-third of FoxA2 targets that occur at silent genes (J. Watts and KZ, in preparation). Typically, these silent FoxA2 bound genes are active in other endoderm-derived cell types. Notably, at least one of the targets is a regulatory gene whose activation can cause cell type conversion, or metaplasia, among gut tissues. Further analysis has revealed a network of repressive transcription factors that help keep the FoxA2 target gene silent in liver cells and that may be disrupted during gut pathologies associated with metaplasia. These studies reveal how pioneer factors enable dysregulated cell differentiation events that may underlie disease.
With regard to our fate mapping of the foregut endoderm, revealing different endoderm domains that together give rise to the embryonic liver bud: we have performed laser-capture microdissection of different endodermal domains, followed by RNA isolation, amplification, and microarray analysis. With the resulting RNA expression profiles, we subtracted the expression profiles of adjacent mesodermal tissue, yolk sac, and the total embryo, providing us with lists of genes whose expression is highly enriched in different domains of the foregut endoderm. We are now using bacterial artificial chromosomes harboring these genes to drive the expression of CRE recombinase in different endodermal domains of transgenic mice. This will allow us to perform genetic lineage marking analysis and determine whether descendants of different endodermal progenitors have different growth, regenerative, and stem cell capacities in the adult liver and pancreas.
A major application of developmental biology studies is to use the tissue-inductive signals that were identified from studies of embryos to prospectively program stem cells. Furthermore, understanding the signal transduction pathways that mediate tissue induction events, and how the pathways interact to form a network, can reveal agonist, antagonist, and other small molecule targets to promote efficient stem cell differentiation. To this end, we are investigating the signal transduction pathways and interactions within endoderm cells during the period preceding and culminating in liver and pancreas induction, as well as within cells at the subsequent steps of differentiation. Understanding how such pathways converge on pioneer factors at target genes and other chromatin parameters and induce new regulatory events leading to cell type specification will provide a cohesive view of how to control cell fates at will.
Research in the Zaret laboratory described in this review has been supported by the National Institutes of Health (R01-GM36477; R01-GM74903; U01-DK072503; CA-06927), the Searle Scholars Program, the Human Frontiers Science Program, and the Mathers Charitable Foundation.