Dynamic Signaling Network for the Specification of Embryonic Pancreas and Liver Progenitors

See allHide authors and affiliations

Science  26 Jun 2009:
Vol. 324, Issue 5935, pp. 1707-1710
DOI: 10.1126/science.1174497

Integrating Organ Induction

During animal development, multiple signaling pathways specify the induction of organ progenitors such as those in the pancreas and liver. Wandzioch and Zaret (p. 1707) investigated how three signaling pathways converge on the earliest genes specifying these organs. Within hours, multiple changes were observed in the inductive network with different signals operating in parallel. The findings may help to explain the incomplete programming seen in various stem cell differentiation protocols.


Studies of the formation of pancreas and liver progenitors have focused on individual inductive signals and cellular responses. Here, we investigated how bone morphogenetic protein, transforming growth factor–β (TGFβ), and fibroblast growth factor signaling pathways converge on the earliest genes that elicit pancreas and liver induction in mouse embryos. The inductive network was found to be dynamic; it changed within hours. Different signals functioned in parallel to induce different early genes, and two permutations of signals induced liver progenitor domains, which revealed flexibility in cell programming. Also, the specification of pancreas and liver progenitors was restricted by the TGFβ pathway. These findings may enhance progenitor cell specification from stem cells for biomedical purposes and can help explain incomplete programming in stem cell differentiation protocols.

Although stem cells are defined by their ability to self-renew and to differentiate down specific cell lineages, there are limitations in the capability of stem cells to be programmed in vitro. Cells differentiated in culture often lack the full spectrum of properties normally exhibited by their in vivo counterparts (1, 2). One reason for the limitation could be a lack of understanding of how inductive signals interrelate to properly program, for example, pancreas and liver progenitors (13).

In the mouse embryo, progenitors to the pancreatic beta cell and hepatocyte lineages arise from neighboring domains of ventral foregut endoderm. Pdx1 (Ipf1) and alb1 (Alb) are activated at the seventh and eighth somite pair (7-8S) stage in nascent pancreas and hepatocyte progenitors, respectively, in mouse development (47). At the same stage and spanning both the pancreatic (Pdx1+) and hepatic (alb1+) domains, the regulatory genes Hnf1b (Tcf2), Hnf6 (Onecut1), and Prox1 are induced; each is required for the subsequent development of both the liver and pancreatic buds (812). We focused on the signaling network that induces all five of these earliest genes specific to the pancreatic and hepatic lineages.

Embryonic tissue explant studies have established that the different progenitor cell domains are induced by signals emitted from overlying mesoderm cells (6, 1316). In mouse and zebrafish embryos, redundant fibroblast growth factors (FGFs) and bone morphogenetic proteins (BMPs) secreted from the cardiac mesoderm and septum transversum mesenchyme, respectively, induce alb1 in the ventral endoderm and suppress Pdx1 (5, 1721). How these signals coordinately feed into the induction of Hnf1b, Hnf6, and Prox1 is not known, nor is their relationship to TGFβ signaling, which is also present in the foregut (22).

Establishing a foregut endoderm fate map (23) is crucial for monitoring endoderm signaling events before cell type specification. Paired lateral domains of endoderm cells move ventrally at the anterior intestinal portal, helping to close off the developing foregut and contribute to the liver bud (Fig. 1, A and B). The liver bud is also constituted from a distinct population of medial endoderm cells (Fig. 1, A and B); the prospective lateral and medial hepatic domains fuse near the heart at the 7-9S stage (Fig. 1C) (23). Ventral pancreatic progenitors reside in the caudal-most domain of lateral liver progenitors; they migrate toward the midline during gut closure and reside caudal to the liver and heart domains at 7-8S (23). To examine BMP, TGFβ, and FGF signaling within ventral endoderm before pancreatic and liver specification, we stained embryos at different stages (24) with antibodies specific to the phosphorylated pathway intermediates SMAD1,5,8 (for the BMP pathway), Smad2 (for the TGFβ pathway), and extracellular signal–regulated kinases ERK1 and 2 [for the mitogen-activated protein kinase (MAPK) pathway, downstream of FGF (17)]. The staining patterns were blocked by pathway inhibitors to ensure specificity of the antibodies (fig. S1).

Fig. 1

Dynamic signal transduction molecule activation in the foregut endoderm revealed by whole-mount immunofluorescence (IF). (A to C) Fate map of foregut endoderm at different stages (23). (D to O) Images with the foregut facing the viewer. Arrows indicate regions with positive signal; brackets indicate regions lacking signal. m.e., medial endoderm; l.e., lateral endoderm; d.e., dorsal endoderm; hf, head fold; ht, heart; fg, foregut; liv., liver; v.p., ventral pancreas.

Whole-mount immunofluorescence of 3-4S embryos revealed that the prospective lateral hepatic progenitors contained activated FGF-MAPK signaling (Fig. 1D), as reported previously (17). The ventral-medial endoderm, also containing prospective hepatic progenitors, was negative for phosphorylated ERK (pERK) at 3-4S (Fig. 1D). Activated BMP signaling exhibited the opposite pattern: pSMAD1,5,8 was evident in ventral-medial endoderm at 3-4S and not in the ventral-lateral endoderm (Fig. 1G). Thus, the prospective ventral-medial hepatic progenitors have activated BMP signaling (as evidenced by SMAD1,5,8) at the 3-4S stage, but these progenitors do not yet have an active FGF-MAPK pathway, whereas prospective ventral-lateral hepatic progenitors have activated FGF-MAPK at the 3-4S stage, but no active BMP signaling.

A few hours later in development (5-6S), activated MAPK signaling tracked with the prospective lateral hepatic progenitors as they moved ventral-medially (Fig. 1E). During this period, Bmp4 and Bmp2 were expressed in septum transversum mesenchyme across the ventral-medial region (20, 25) (fig. S2, A and B), and pSMAD1,5,8 began to spread laterally (Fig. 1H). At the time of activation of the liver program (7-9S), the two lateral hepatic progenitor domains had migrated ventrally, joining the midline progenitors, and thus the entire liver domain became positive for pERK and pSMAD1,5,8 (Fig. 1, F and I).

At 3-6S stages, the prospective ventral pancreas progenitors lacked activated pSMAD1,5,8 (Fig. 1, G and H). However, by the 7-8S stages, the lateral cells moved ventrally and were in the domain of pSMAD1,5,8 (Fig. 1I). In summary, the ventral pancreatic progenitors are initially negative for BMP signaling and, several hours later, become positive.

BMP signaling via pSMAD1,5,8 utilizes SMAD4, which is a partner for TGFβ-activin signaling via phosphorylated SMAD2,3. Because TGFβ2 is expressed in the foregut and its receptors are expressed in the endoderm (22, 26), we investigated pSMAD2. Notably, during all stages, we detected pSMAD2 throughout the ventral foregut endoderm (Fig. 1, J to L). Similarly, SMAD4 was expressed throughout the ventral foregut endoderm, as well as in the cardiac mesoderm (Fig. 1, M to O). In the dorsal endoderm, we detected SMAD4 and pSMAD2, but not pERK or pSMAD1,5,8 (Fig. 1, J to L, N, and O). Several hours later (11S), ventral SMAD4 expression was confined to the ventral-medial endodermal domain, extending anteriorly to the pharyngeal region of the foregut (Fig. 2A).

Fig. 2

Genetic and pharmacologic perturbation of signaling pathways. (A and B) Whole-mount IF of designated genotypes. Red arrow indicates presence of SMAD4 in ventral foregut endoderm; dashed arrow indicates lack of SMAD4. DAPI, 4′,6′-diamidino-2-phenylindole. (C to E) Images of 5S half embryo before (C) or after (D and E) culture. (F to J) qRT-PCR on endoderm from half embryos ± inhibitors or from FoxA3cre;SMAD4CA/Δ (Smad4 null) embryos at 11S compared with that from 11S wild-type embryos. Each sample (n = 2 to 8) was analyzed in triplicate; *P ≤ 0.05 determined by t test; **P ≤ 0.005; “nd,” not determined. Error bars indicate SD calculated with the comparative Ct (ΔCt) method.

To investigate the role of each pathway, we used genetic and pharmacologic perturbants. We used a FoxA3CRE transgene, which becomes functional in the ventral foregut endoderm at the 5-6S stage (17, 27) and a Smad4 conditional allele (Smad4CA) (28) to impair SMAD1,5,8 (BMP) and SMAD2,3 (TGFβ) signaling. As anticipated, ventral foregut expression of SMAD4 was completely ablated in FoxA3CRE;Smad4CA/Smad4Δ embryos, although persisted in the pharyngeal region (Fig. 2, A and B). We also developed a new half-embryo culture system that allows foregut development (24). Briefly, yolk sacs and tissues caudal to the first somite were removed from 2-6S embryos, and the anterior half embryos were cultured 24 hours while floating. These half embryos reproducibly underwent gut tube closure, then induction of early liver and pancreatic markers, and formed a large, beating heart (Fig. 2, C to E). We processed the ventral gut domain for mRNA analysis by quantitative reverse transcription polymerase chain reaction (qRT-PCR). Gene expression in 24-hour cultures was normalized to that of the gene for glyceraldehyde-3-phosphate dehydrogenase, Gapdh, and compared with that in ventral foregut tissues isolated from intact embryos at 5S, 9S, and 13S stages, with the 5S stage as a reference. Foxa2 expression, a marker of foregut endoderm (27), was consistent throughout the embryonic stages and in 24-hour half-embryo cultures, whereas Pdx1, Hnf1b, Hnf6, Prox1, and alb1 were activated in half embryos to levels observed in 9-13S native embryos (fig. S3). Thus, the half-embryo explants recapitulate the induction of the earliest known genes for the pancreatic and hepatic programs.

Treating half embryos isolated at the 3-4S stage with the BMP inhibitor noggin suppressed induction of the hepatocyte early gene alb1, whereas exogenous BMP4 treatment slightly enhanced alb1 induction (Fig. 2F). With half embryos isolated at the 5-6S stage, stronger effects were observed (Fig. 2F). Similarly, noggin inhibited Prox1 induction, whereas BMP4 enhanced it (Fig. 2G). Consistent with the latter results, the embryos in which Smad4 had been deleted in the endoderm at 5-6S exhibited a complete loss of alb1 induction and impaired Prox1 induction (Fig. 2, F and G), although some Afp-positive cells were eventually produced (see Fig. 3F below). Thus, BMP signaling initially is pro-hepatic with this experimental system, as seen in our previous studies and elsewhere (1, 20, 21).

Fig. 3

Defective ventral pancreatic development in embryos deficient for Smad4 in the endoderm. (A to F) Transverse IF sections for Pdx1 or AFP (red), and DAPI. (G and H) Whole-mount IF for Pdx1 with DAPI. d.p., dorsal pancreas; lb, liver bud.

Treating half embryos isolated at 3-4S with noggin induced the pancreatic beta cell early gene, Pdx1, and treatment with BMP4 inhibited it (Fig. 2J), which together imply a reciprocal effect compared with liver induction, as we saw previously (18, 20). Similarly, at the 3-4S stage, Hnf6 was induced by noggin and inhibited by BMP (Fig. 2H).

In contrast, half-embryo cultures established several hours later, at the 5-6S stage, consistently exhibited the opposite phenotype. Noggin suppressed Pdx1 and Hnf6 expression, whereas exogenous BMP4 enhanced it (Fig. 2, H and J). These results were confirmed by genetics, as deletion of Smad4 at 5-6S wiped out Pdx1 induction and impaired Hnf6 induction (Fig. 2, H and J), which underscored the rapidity with which programming signals naturally change in their inductive effects. The latter findings are consistent with studies on endoderm explants from later-stage chick embryos (29).

The gene expression phenotypes in the FoxA3CRE;Smad4CA/Smad4Δ embryos were accompanied by profound organogenic defects. Immunofluorescence staining revealed virtually no Pdx1-positive cells at embryonic day 9.5 (E9.5) in the ventral gut, nor was there morphological evidence of a ventral pancreatic bud (Fig. 3, C and D). By E10.5, a few ventral Pdx1-positive cells were evident (Fig. 3, G and H). Liver development was also impaired (Fig. 3, E and F). However, there was only a weak effect of Smad4 deletion on dorsal pancreatic development (Fig. 3, A, B, G, and H), consistent with our observation that pSMAD1,5,8 was not observed in the dorsal endoderm (Fig. 1, G to I). Another study used Pdx1CRE, which is activated after 7-8S, to delete a conditional Smad4 allele, and no phenotype was seen during dorsal or ventral pancreatic development (30). Taken together, it appears that BMP-SMAD4 signaling is crucial solely for a several-hour period of ventral pancreatic development, from the 5-6S stage to the 7-8S stage, when Pdx1 becomes activated.

To assess the role of activated pSMAD2,3 (TGFβ signaling) throughout the foregut endoderm (Fig. 1, J to L), we treated half embryos at 3-4S and 5-6S stages with SB431542, an inhibitor of the activin receptor–like kinases (ALK) 4,5,7, and also treated half embryos with exogenous TGFβ2. TGFβ signaling was strongly inhibitory to pancreatic Pdx1 and Hnf1b and modestly inhibitory to the hepatic alb1 and Hnf6 at 5-6S (Fig. 2, F to J). Because Hnf1b was strongly induced in the Smad4 null embryos and was not affected by the other signaling pathways, it appears that Hnf1b suppression could be a primary effect of TGFβ signaling.

The extensive inhibitory effect of TGFβ signaling suggested that it could be restraining cell type specification in the foregut. To test this, we marked nascent pancreatic progenitor cells by pulsing pregnant females from a Pdx1cre-ER × Rosa26R cross (31) with tamoxifen at E7.5, isolating Pdx1cre-ER;Rosa26R embryos at E8.25 and setting up 5-6S half-embryo cultures with or without the ALK4,5,7 inhibitor or exogenous TGFβ2. The ALK4,5,7 inhibitor greatly increased the number of Pdx1CRE-positive cells along the gut endoderm, and excess TGFβ2 virtually prevented their appearance (Fig. 4, A, C, and E). To determine whether the ALK4,5,7 inhibitor markedly changed the proliferative rate of the Pdx1CRE-positive cells, we sectioned the embryo halves sagittally and stained for Ki-67. Extensive Ki-67–positive cells were seen within the gut endoderm, but in both the treated and untreated tissues, Pdx1CRE;Rosa26R–positive cells exhibited little or no staining (Fig. 4, B, D, and F). Furthermore, the ALK4,5,7 inhibitor did not induce apoptosis (fig. S4). Taken together, the data indicate that continuous TGFβ signaling within the foregut endoderm restrains the number of ventral pancreas cells that are specified.

Fig. 4

TGFβ signaling restrains the specification of ventral pancreatic progenitors. (A, C, and E) Whole-mount X-gal staining (blue) for pancreatic progenitors. (B, D, and F). Sagittal sections, counterstained for Ki-67 (brown). (G) Dynamic network inducing the earliest genes for liver and pancreas specification. Arrows, positive signaling; blunt-ended arrows, negative. The thicker lines for BMP and thinner lines for TGFβ at 5-6S indicate BMP as the dominant Smad4-dependent pathway for alb1, Hnf6, and Pdx1.

During foregut closure, the movement of undifferentiated ventral endoderm cells ventrally and caudally, away from pancreas-inhibitory FGF signaling from the cardiac mesoderm, is necessary for pancreatic specification (32). We speculate that TGFβ signaling restrains lateral endoderm specification until the cells move sufficiently far from the heart and into a BMP signaling domain, with the latter then becoming the dominant SMAD4-dependent pathway leading to pancreatic induction.

Our findings underscore the importance of a detailed understanding of the signaling events for cell type specification. In particular, it would be crucial to modulate the timing of BMP and TGFβ signaling to optimize for pancreatic progenitors from stem cells. Our work also suggests possible flexibility in hepatic programming, in that the prospective lateral hepatic progenitors receive FGF signaling before that for BMP, whereas the prospective medial hepatic progenitors receive BMP signaling before that for FGF. Regardless of this marked difference, descendants of both progenitor cell types activate Afp and Hnf4 in the nascent liver bud (23).

Finally, we find that the signals that induce the earliest hepatic and pancreatic genes operate in parallel, independently tracking to different gene targets that together are crucial for cell specification (Fig. 4G). The parallel model helps explain incomplete cell programming in stem cell studies, in that, if a key signal is not provided or not provided at the appropriate time, various markers of correct differentiation could still be induced, but not all of them. A careful investigation of the inductive networks at each step of hepatocyte or beta cell differentiation should allow prospective cell derivation that is useful for research and therapies.

Supporting Online Material

Materials and Methods

Figs. S1 to S4

Table S1


  • * Present address: Department of Cell and Developmental Biology, University of Pennsylvania, 421 Curie Boulevard, Philadelphia, PA 19104, USA.

References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. We thank E. Robertson for SmadCA mice; K. Kaestner for FoxA3cre mice; A. Hines for help with breeding; the Fox Chase Cancer Center (FCCC) Laboratory Animal Facility; F. Roegiers, E. Robertson, H. Fan, R. Lake, and D. Freedman-Cass for comments on the manuscript; and E. Pytko for help in its preparation. E.W. was a Postdoctoral Fellow of the Swedish Research Council. The work was funded by NIH grants R37GM36477 and U01DK072503 to K.S.Z. K.S.Z. is on a Scientific Advisory Board for Johnson & Johnson in the area of embryonic stem cells.

Stay Connected to Science

Navigate This Article