Initiation of Mammalian Liver Development from Endoderm by Fibroblast Growth Factors

See allHide authors and affiliations

Science  18 Jun 1999:
Vol. 284, Issue 5422, pp. 1998-2003
DOI: 10.1126/science.284.5422.1998


The signaling molecules that elicit embryonic induction of the liver from the mammalian gut endoderm or induction of other gut-derived organs are unknown. Close proximity of cardiac mesoderm, which expresses fibroblast growth factors (FGFs) 1, 2, and 8, causes the foregut endoderm to develop into the liver. Treatment of isolated foregut endoderm from mouse embryos with FGF1 or FGF2, but not FGF8, was sufficient to replace cardiac mesoderm as an inducer of the liver gene expression program, the latter being the first step of hepatogenesis. The hepatogenic response was restricted to endoderm tissue, which selectively coexpresses FGF receptors 1 and 4. Further studies with FGFs and their specific inhibitors showed that FGF8 contributes to the morphogenetic outgrowth of the hepatic endoderm. Thus, different FGF signals appear to initiate distinct phases of liver development during mammalian organogenesis.

Identifying the molecular signals that initiate organogenesis from the gut is important for understanding the fundamental mechanisms of developmental regulation, hereditary digestive disorders, and tissue regeneration. Different segments of the mammalian gut endoderm give rise to the liver, lung, pancreas, thyroid, and gastrointestinal tract. Typically, a portion of the endoderm will begin to express genes specific to one of these tissues, and then the newly specified cells will proliferate out of the endoderm layer to form a tissue bud, initiating morphogenesis (1, 2). InDrosophila, the initial specification of tissues within the gut endoderm is caused by signals from overlying mesoderm (3). In the chick, transplanted foregut endoderm will develop into liver only if it is with its adjacent cardiac mesoderm (4); mesoderm from other areas of the embryo is not hepatogenic (4, 5). Gene inactivation studies in mice have identified many signaling molecules and transcription factors that are required for development of gut-derived organs, but the factors are critical either for the formation of the endoderm itself (6) or for tissue development after the morphogenetic bud stage (7), leaving open the question of how the tissue types are initially specified.

We have modified the tissue transplantation approach to identify signals that control the initial specification of the liver. The first known evidence of hepatic differentiation of the ventral foregut endoderm is activation of the liver-specific serum albumin gene and enhanced expression of α-fetoprotein (AFP) mRNA at the seven- to eight-somite stage in the mouse (2). Transcription of these genes in embryo tissues is detectable by reverse-transcriptase polymerase chain reaction (RT-PCR) analysis of RNA (2); specific mRNA accumulation at this stage has not been detectable by in situ hybridization (2, 8). Later in development, the newly specified hepatic cells exhibit morphogenetic outgrowth to form the liver bud (2, 8). When ventral foregut endoderm is isolated from mouse embryos at ∼8.25 days of gestation (two- to six-somite stages; Fig. 1A) and cultured in microwells for 2 days, expression of albumin mRNA is activated and the amount of AFP mRNA is increased only if the endoderm is in contact with cardiac mesoderm (2). Cardiac mesoderm in the same microwell as the endodermal cells, but separated from the latter, fails to induce liver gene expression, which indicates that either cell contact or one or more locally secreted factors is critical (2). Hepatic induction in this in vitro assay system is similar to that in chicken endoderm transplantation studies (4, 5). In both chicken and mouse embryos, endogenous hepatic induction by the cardiac mesoderm takes place at about the seven-somite stage (2, 4).

Figure 1

Expression patterns of FGF signaling molecules in five- to eight-somite mouse embryos. (A) Side view of mouse embryo at the seven- to eight-somite stage. Squared region indicates sections viewed in (D) to (G); dashed line indicates orthogonal section in (H) to (J). (B and C) FGF2 in situ hybridization (19); front view of embryos with optic lobes at top. MRNA-positive cardiac region is denoted by red arrow. Note the lack of staining in the endodermal lip beneath the cardiac region (orange arrow). (D to J) Antibody staining of seven- to eight-somite embryo sections. Red arrows denote FGF-positive segments of the cardiac region. Orange arrows denote ventral foregut endoderm. Nonspecific, dark staining was occasionally observed in the foregut pocket (D and E). Normal mouse IgG was used as a control in (G); normal rabbit IgG was used in (J). (H) Localized FGF8 staining at the boundary between the cardiac mesoderm and the ventral foregut endoderm from an embryo sectioned orthogonal to the arrows in (F). (K to M) Antibody staining of cocultures of ventral foregut endoderm and cardiac mesoderm. Dashed circles represent areas of beating cardiac tissue. FGFR-1 was present throughout the culture, whereas FGFR-4 was present outside the cardiac areas and was especially prominent at the leading edge of the explants.

We investigated the role of fibroblast growth factors (FGFs) as potential hepatogenic signals because chicken embryos express FGF1 and FGF2 in the myocardium at the 7- to 13-somite stages (9) and mouse embryos express FGF8 mRNA in the cardiac mesoderm from the presomitic to the 7-somite stage (10). Also, FGF receptor 4 (FGFR-4) mRNA is expressed exclusively in the endoderm at 8.5 days of gestation in mice (11) and FGFR-1 is in both the cardiac and endoderm regions (12). FGFs are secreted locally into the extracellular matrix (13) and can promote differentiation of the embryonic mesoderm (14) and the anterior pituitary (15) as well as induce morphogenesis of limbs, lungs, and teeth (16). Inactivation of the FGFR-1 and FGF8 genes results in early embryonic lethality before hepatogenesis (17), whereas FGFR-4 and FGF2 gene inactivation results in minimal embryonic phenotypes, which suggests redundancy (18). We therefore sought to precisely define the expression patterns of FGF signaling components at the time of hepatogenesis in the mouse embryo.

FGF2 mRNA was expressed strongly in the cardiac region at the seven- to eight-somite stage (Fig. 1C, red arrow) but not in the endoderm below the cardiac region (Fig. 1C, orange arrow) (19). FGF2 mRNA was barely detectable at the five-somite stage (Fig. 1B), which shows that it is induced at the time of hepatogenesis. To address whether FGFs themselves are present in the cardiac mesoderm, we used antibodies to stain mouse embryo sections at the seven- to eight-somite stage (20). FGF2 was expressed throughout the cardiac mesoderm (Fig. 1D, red arrows) adjacent to the ventral foregut endoderm (Fig. 1D, orange arrow; Fig. 1G, control). Although Crossley and Martin (10) and we (21) found that cardiac mesoderm expression of FGF8 mRNA declines by the seven-somite stage, we found that FGF8 antigen persists in the cardiac mesoderm near the endoderm (Fig. 1F, red arrow). Interestingly, there was very localized and intense FGF8 staining at the boundary between the cardiac mesoderm and the endoderm (Fig. 1H, red and orange arrows, respectively). FGF1 staining of the cardiac mesoderm was just beginning to appear in seven- to eight-somite embryos (Fig. 1E) and was undetectable at earlier stages (21). FGFR-1 was present in the foregut endoderm as well as in the cardiac mesoderm (Fig. 1I, red and orange arrows; Fig. 1J, control). We also found that FGFR-1 and FGFR-4 are expressed abundantly in the endodermal/epithelial portions of cocultures of ventral foregut endoderm and cardiac mesoderm (Fig. 1, K and L; Fig. 1M, control). We conclude that multiple FGF signaling components are expressed in the relevant tissues precisely at the time of hepatogenesis.

To test whether FGFs alone are sufficient to induce the first step of hepatogenesis, we isolated ventral foregut endoderm from two- to six-somite stage mouse embryos (22) and treated the cells with or without FGFs in the presence of a heparan sulfate carrier (23). Tissue explants were cultured in microwells for 2 days, and then RNA was isolated from individual explants and analyzed for liver gene expression by RT-PCR (24). Of 39 control endoderm explants cultured alone, without FGFs, including 19 tissues grown in the presence of the heparan sulfate carrier, none expressed serum albumin mRNA (Fig. 2A, lane 1; Table 1). Only 1 of 12 tissues was AFP-positive (Table 1). In contrast, of a total of 12 ventral foregut endoderm tissues grown in the presence of FGF-1 (50 or 500 ng ml−1), 11 expressed albumin mRNA and of the 8 tested all expressed AFP mRNA (Fig. 2A, lanes 5 to 8; Fig. 2B, lanes 2 to 5; Table 1). Of nine ventral endoderm fragments cultured in the presence of FGF1 at 5 ng ml−1, five exhibited albumin mRNA expression (Fig. 2A, lanes 2 to 4; Table 1). This FGF1 concentration is known to be close to the threshold for cell responses (25). Although FGF2 at 50 ng ml−1 was inefficient (Table 1), FGF2 at 5 ng ml−1 efficiently induced albumin (Fig. 2D, lanes 2 to 6; Table 1). Such sharp dosage thresholds with FGF2 have been well documented (26). FGF8b was inefficient in albumin induction at either 5 or 50 ng ml−1 (Table 1). As a third marker for hepatogenesis, we analyzed the expression of transthyretin (TTR) mRNA, which is expressed in both the yolk sac and early liver (27). TTR mRNA was detectable by RT-PCR throughout the endoderm during the two- to eight-somite stages (21) and its expression was extinguished when the endoderm was cultured alone (Fig. 2C, lane 1). However, treating the endoderm with FGF1 maintained or enhanced TTR expression (Fig. 2C, lanes 2 to 4; Table 1), as did culturing the endoderm in contact with cardiac mesoderm (Fig. 2C, lane 5). The amounts of FGF-induced liver gene expression were comparable to or greater than that observed when cardiac mesoderm was in contact with the ventral endoderm, relative to the β-actin mRNA internal control.

Figure 2

Induction of liver gene expression in ventral endoderm cultures treated with FGFs. (A to D) RT-PCR analysis of RNA from ventral endoderm tissues cultured for 2 days either with heparan sulfate (50 ng ml−1) and the designated concentrations of FGFs or with cardiac mesoderm (cardiac mes.). Actin and liver-specific mRNA for albumin (A and D), AFP (B), or TTR (C) were assayed simultaneously. (E) Failure of FGF1 to induce albumin expression in embryonic neural tube, midsections, and head tissues.

Table 1

Induction of hepatic mRNAs in embryo tissue cultures. Cumulative number of embryo tissue fragments in our laboratory that expressed (+) or did not express (−) mRNAs for the designated liver-specific genes after 48 hours in culture, as assayed by RT-PCR, is indicated. ND, not done.

View this table:

The hepatogenic effect of FGF1 and FGF2 was specific to the endoderm. Neither factor induced hepatic gene expression in cultured embryo neural tubes, midsections, and head cells (Fig. 2E, lanes 1 to 6; Table 1), all of which originate from the ectoderm, the mesoderm, or both. We conclude that multiple FGFs are expressed in the cardiac mesoderm and are individually sufficient to initiate hepatic gene expression from the endoderm.

To determine whether FGFs are necessary for hepatic specification by the cardiac mesoderm, we sought to perturb FGF signaling by using molecular hybrids between FGF receptors and immunoglobulin G (FR-IgG) (Fig. 3A). These FR-IgG constructs contain the extracellular, ligand binding domain of either FGFR-1 or FGFR-4 fused to the hinge and the dimerization domain of IgG heavy chain, resulting in soluble receptors that sequester FGFs (28). FR1-IgG and FR4-IgG molecules were secreted from COS-7 cells, purified from the culture medium (Fig. 3B, lanes 3 and 4), and tested for biological activity (29) (Fig. 3C). As would be predicted from the known FGF binding capacity of the intact receptors (30), FR1-IgG inhibited the ability of FGF2 to stimulate protein tyrosine kinase activity in NIH 3T3 cells, but FR4-IgG did not (Fig. 3C, lanes 3 to 5). In contrast, FGF8b was efficiently inhibited by FR4-IgG and, to a lesser degree, by FR1-IgG (Fig. 3C, lanes 6 to 8). Thus, the inhibitors exhibit different FGF substrate preferences, as expected from studies with intact FGF receptors (30).

Figure 3

Soluble dominant-negative FGF receptor fusion proteins can inhibit albumin gene induction in foregut cocultures. (A) FR-IgG fusion protein strategy. Loops denote immunoglobulin-like domains; thick bars denote tyrosine kinase domain; zigzag denotes hinge domain; and V and C denote variable and constant domains of IgG heavy chain. (B) SDS-PAGE with silver stain of FR1-IgG and FR4-IgG purified from COS-7 cell supernatants compared with IgG. Mock denotes material from mock-transfected cells. (C) Immunoblot analysis of NIH 3T3 cell lysate after cell incubation for 5 hours with FGFs and FR-IgGs as shown. Lane 2, untreated cell lysate showed basal level of protein phosphorylation. Lanes 3 and 6, treatment with FGF2 or FGF8b caused a shift in the phosphoprotein profile. (D toF) Summary of RT-PCR analysis of albumin gene expression, relative to actin, in cocultures of ventral endoderm and cardiac mesoderm in the presence of FR-IgG at either 500 or 200 ng ml−1 with heparan sulfate at 100 ng ml−1. Similar results were obtained from either condition; results were pooled. For a control, we used the same amount of human IgG or plain buffer.

Ventral foregut endoderm was dissected from two- to four-somite mouse embryos along with its associated cardiac mesoderm and was cultured in the presence of FR1-IgG or FR4-IgG (500 ng ml−1) alone or in the presence of either protein at 200 ng ml−1 with heparan sulfate at 100 ng ml−1. Similar results were obtained under both conditions. As a control, we used either human IgG or buffer containing no fusion proteins. After 2 days in culture, we isolated RNA from individual explants and categorized the amounts of albumin mRNA, relative to actin mRNA as barely detectable (low), comparable to the actin concentration (medium), or greater than the actin concentration (high). In 15 of 20 samples treated with FR1-IgG, albumin concentrations were markedly lower than the control group (Fig. 3, D and E). FR4-IgG was also inhibitory but perhaps slightly less effective (Fig. 3F). The high efficiency of FR1-IgG is consistent with its ability to efficiently inhibit FGF2, which was able to induce liver gene expression on its own. We conclude that endogenous FGF signaling is critical for hepatic induction.

In principle, FGF1 or FGF2 could cause outgrowth of a subpopulation of ventral endoderm cells that were already committed to a hepatic fate and expressing liver-specific genes. Although the high sensitivity of the RT-PCR assay made this unlikely, cardiac mesoderm cultured with foregut endoderm does cause vigorous outgrowth of the endoderm (Fig. 4, A and B), which implicates the presence of morphogenetic signals produced by the cardiac mesoderm (7). However, neither FGF1, FGF2, nor FGF8b had a discernible effect on endodermal cell outgrowth (Fig. 4, C to H) (31), even though FGF1 and FGF2 potently induced hepatic gene expression (Fig. 2).

Figure 4

FGF is necessary for morphological outgrowth of newly specified hepatic endoderm. Micrographs of tissue explants at the designated times after cultures were established. Dashed circles indicate beating cardiac cells. Usually beating areas were not seen at 5 hours. (A andB) Extensive endodermal outgrowth in cocultures of cardiac mesoderm and foregut endoderm. (C to H) Ventral foregut endoderm alone shows little outgrowth regardless of the presence of FGFs. (I to N) Cocultures of ventral foregut endoderm and cardiac mesoderm incubated with the designated components. Original magnification for (A to F, I to N), ×40, and for (G and H) ×100.

Interestingly, in 10 of 14 cultures, FR4-IgG strongly inhibited the morphogenetic response of the endoderm to the cardiac mesoderm (Fig. 4, K and L), whereas FR1-IgG did so in only 3 of 14 samples (Fig. 4, I and J). To test the hypothesis that FR4-IgG was antagonizing an FGF8 isoform, a preferred substrate, we cultured ventral endoderm and cardiac mesoderm fragments in the presence of FR4-IgG and exogenous FGF8b. Remarkably, in five of seven cultures, FGF8b restored the outgrowth response of the foregut endoderm (Fig. 4, M and N). Because FGF8b alone was insufficient to elicit morphogenesis, we conclude that FGF8b, or an FGF with a similar receptor specificity, permits the endoderm to respond to another signal that promotes outgrowth of the newly specified hepatic cells. Preliminary experiments with foregut endoderm alone cultured in FGF2 and FGF8b (n = 3) showed that the other signal apparently is not FGF2 (21).

We have distinguished two signaling phases in the initiation of liver development from the endoderm. First, FGF signaling from the cardiac mesoderm induces the initial step of hepatogenesis; that is, it initiates the hepatic gene expression program in the ventral foregut endoderm. Based on the cardiac expression of FGF1, FGF2, and FGF8 at the time of hepatic specification and the independent activity of FGF1 and FGF2, hepatic induction appears to result from redundant FGF signaling. Second, early expression of FGF8b or a related molecule appears to potentiate the morphogenetic activity of the nascent hepatic cells. That is, FGF8 works in conjunction with a signal that has not been identified to stimulate cell outgrowth. The two-phase model may apply to other examples of mesodermal patterning of the endoderm during organogenesis, in which initial changes in gene expression are tightly coordinated with, but separable from, the morphogenesis of tissue buds (1, 2).

  • * To whom correspondence should be addressed: E-mail: zaret{at}


Stay Connected to Science

Navigate This Article