Liver Specification in the Absence of Cardiac Differentiation Revealed by Differential Sensitivity to Wnt/β Catenin Pathway Activation

Embryonic precursors of liver and heart, whilst not sharing cellular origin, develop in close proximity through a dynamic series of inductive signaling events. During gastrulation anterior endoderm (AE) provides cardiogenic signals that act on adjacent mesoderm, resulting in induction of cardiac precursors. Subsequently cardiogenic mesoderm generates a FGF signal that acts on adjacent AE to induce foregut organ specification. Additional signals such as BMP and Wnt provide further information required for liver specification. Most findings on liver specification were derived from mouse explant studies as well as experiments with Xenopus and zebrafish embryos. To address some of the limitations of these models, here we used two complementary ex vivo models based on Xenopus embryos: pluripotent animal cap explants expressing Gata4 transcription factor and conjugates of gastrula-stage AE with animal caps (AC). We show that in these models liver specification is not sensitive to Wnt signaling manipulation, in contrast to the requirement for Wnt antagonism shown in vivo. FGF pathway is not necessary for Gata4-induced liver specification in animal cap explants but is required for prolonged period in sandwiches of AE and AC. In contrast, BMP signaling is shown to be essential for Gata4-induced liver specification. Our findings may have implications for research on liver differentiation from embryonic stem cells.


INTRODUCTION
During embryonic development liver is induced in foregut endoderm by diverse and dynamic signaling from surrounding mesodermal tissue. Classical embryological experiments in the avian model have identified cardiac mesoderm as a source of an essential signal that specifies liver primordium induction and outgrowth from the adjacent gut tube (Zaret, 2008(Zaret, , 2016Zorn and Wells, 2009). These findings were confirmed in the mouse explant system, which reconstitutes interactions between cardiogenic mesoderm and ventral endoderm (Gualdi et al., 1996). The mouse explant assay has been used to identify FGF signaling as a cardiac mesoderm-derived factor that induces liver-specific gene expression (Jung et al., 1999). Furthermore, BMP derived from the adjacent septum transversum mesenchyme was shown to be required together with FGF for liver specification (Rossi et al., 2001). In addition to FGF and BMP signaling, the Wnt pathway has been implicated in liver specification (McLin et al., 2007;Gordillo et al., 2015;Zaret, 2016). The actions of these signaling pathways in early liver development are highly dynamic and dose-dependent.
The close relationship between embryonic liver and heart likely begins early in development, during gastrulation. Experimental evidence from chick and frog models have suggested that during gastrulation dorso-anterior endoderm, a tissue that will contribute to the liver formation, is required to induce cardiac tissue in adjacent mesoderm (Lough and Sugi, 2000). Later on, after hepatic specification, signals arising from developing liver bud appear to induce the formation of the proepicardium in the mesothelium in later cardiac development (Ishii et al., 2007). Therefore, the fates of developing heart and liver may be tied by several rounds of reciprocal signaling.
Liver-inducing signals regulate the transcriptional program in foregut endoderm via pioneer transcription factors FoxA and Gata4 which have the ability to associate with target genes in compacted heterochromatin (Zaret, 2016). Gata4 and Gata6 zinc-finger transcription factors have conserved roles in liver development in the mouse, zebrafish, and frog (Gordillo et al., 2015). In addition, Gata5 has been shown to regulate liver development in Xenopus (Haworth et al., 2008). The Gata4/5/6 family of transcription factors have well-documented roles in other tissues, notably the heart (Charron and Nemer, 1999).
Of relevance for the current study, Gata4, a hepatic pioneer transcription factor, has cardiogenic activity: gain of function of Gata4 alone, or together with other cardiac factors, can induce cardiogenesis in Xenopus and mouse embryos, respectively (Latinkic et al., 2003;Takeuchi and Bruneau, 2009). In pluripotent animal pole cells from Xenopus blastula embryos Gata4 induces not just cardiac cell fate, but also liver cell fate (Latinkic et al., 2003). This finding provides an experimentally amenable model of co-induction of cardiac and liver fates to study the mechanisms involved.
We have complemented the Gata4-based induction model with another Xenopus model developed for investigating the inductive capacity of anterior endoderm (AE) (Samuel and Latinkic, 2009). In this model, early gastrula anterior endoderm explants are conjugated with pluripotent responding tissue, blastula stage animal caps (AC). AC/AE conjugates were shown to recapitulate cardiogenic signaling between the source, AE, and the responder, AC (Samuel and Latinkic, 2009). Here we show that AC/AE closely mimic cellular and molecular interactions as they occur during liver induction as well. AE explants in isolation retain endodermal characteristics but fail to adopt liver fate, which can be induced if AE is conjugated with AC tissue. An AE-derived signal first induces cardiac precursors in AC, which appear to generate a signal that acts on AE to induce liver cell fate.
Using both the Gata4 and AC/AE models, we show that active Wnt signaling is compatible with hepatic specification despite the well-known inhibitory effect on cardiac differentiation. In addition, we show that Gata4 induces liver cell fate independently of FGF signaling but requires BMP signaling.
Western blotting to detect exogenous (injected) Gata4 protein using HA tag was as described (Gallagher et al., 2012).

Induction of Hepatogenesis in Gata4-Expressing Animal Cap Explants and Animal Cap/Anterior Endoderm Conjugates
Expression of transcription factor Gata4 in animal cap explants from blastula stage embryos has been shown to lead to cardiac differentiation (Latinkic et al., 2003). In this model Gata4 does not act with exclusive cardiac specificity but also induces early endoderm markers as well as a liver marker fabp1 (formerly known as lfabp). High level of liver-specific expression of fabp1 appears in late tadpole stages, beyond the practical limit of culturing AC explants. We have re-examined liver induction in Gata4 mRNA-injected AC explants by using expression of a liverspecific early tadpole marker nr1h5 (formerly known as for1), as well as liver-enriched markers hhex and foxa2. Our analyses demonstrate that Gata4 induces liver cell fate in AC explants, in addition to cardiac cell fate (Figures 1A,B).
Gata4 induces cardiac and liver cell fates in uniformly injected AC explants (Latinkic et al., 2003), suggesting that under conditions of gain of function of Gata4 in pluripotent AC explants, liver cell fate is induced cell-autonomously and that fate acquisition is stochastic. This activity of Gata4 is consistent with its well-known roles in liver specification in vivo (Gordillo et al., 2015;Zaret, 2016). We further explored the question of cell autonomy of liver specification by endoderm specifiers and to that end we used Sox17, which in early vertebrate embryos is an exclusive endoderm determinant. Sox17 induces endoderm in AC explants, both when uniformly expressed and when expressed in half of each explant (Gallagher et al., 2014). Cardiac tissue is induced in hemi-injected explants only (Supplementary Figure S1; Gallagher et al., 2014). Similarly, liver marker nr1h5 expression is induced only in hemi-injected sox17 AC explants, strongly suggesting that liver cell fate is induced non-cell autonomously by a determinant of early endoderm (Supplementary Figure S1).
In addition to Gata4-triggered cardiogenesis in AC explants, we have developed a model that uses endogenous cardiogenic signal produced by the gastrula stage AE to induce cardiac cell fate in juxtaposed (conjugated) AC tissue (Samuel and Latinkic, 2009). AE explants express endodermal markers a2m and sox17 as well as AE marker hhex (Supplementary Figure S2). During culturing period, AE explants retain endodermal characteristics (a2m and sox17) as well as hhex expression. Given that AE express hhex, which at tadpole stages marks both liver and endothelial cells, and that AC/AE conjugates contain endothelial cells (Samuel and Latinkic, 2009), the expression of hhex cannot be used to monitor liver specification in this model.
At tadpole stage (st. 34) AC/AE conjugates showed nr1h5 expression in AE, adjacent to the domain of cardiomyocytes marked by myl7, in a manner resembling the close spatial relationship of the developing heart and liver in the embryo (Figures 1C-F). Upon prolonged culture until st. 43, the conjugates showed evidence of endodermal fate diversification, by expressing liver (fabp1), intestine (fabp2), pancreas (pdia2 and pdx1), and lung/thyroid (nkx2-1) markers (Figure 2). Despite expressing pdia2 and pdx1, AC/AE conjugates did not express insulin or amy2a, suggesting incomplete pancreatic reprogramming. As previously shown (Samuel and Latinkic, 2009), AC/AE expressed cardiac ventricular marker myl3 as well.

Cerberus and hhex Are Required in Anterior Endoderm for Liver and Cardiac Specification
Cerberus (cer1) and hhex have both been shown to be required for normal development of the anterior end of the embryonic axis and for cardiac and AE specification (Brickman et al., 2000;Martinez Barbera et al., 2000;Foley and Mercola, 2005;Foley et al., 2007). We took advantage of the AC/AE model to specifically downregulate hhex or cer1 in AC ( Figure 3A) or AE ( Figure 3B) using previously described MOs against cer1 (Kuroda et al., 2004) and hhex (Smithers and Jones, 2002). We have confirmed effectiveness of cer1MO and hhexMO by showing that they affect heart development (Supplementary Figure S5). Our results demonstrate that both hhex and cer1 are specifically required in AE for cardiac and liver specification, in agreement with previous work (Foley and Mercola, 2005;Foley et al., 2007). Interference with Hhex function in AE by using dominant-negative construct HhexVP2 (Brickman et al., 2000) FIGURE 2 | Endoderm diversification in AC/AE conjugates. Examination of gene expression in st. 43 AC/AE explants by RT-PCR reveals expression of liver (fabp1), intestine (fabp2), pancreas (pdia2 and pdx1, but not ins and amy2a), thyroid/lung (nkx2-1) as well as ventricular cardiomyocyte (myl3) markers. St. 43 E-sibling control embryo at st. 43.
produced the same result as hhexMO ( Figure 3B). Additional experiments have shown that cer1 deficiency in the AE can be rescued by expression of cer1 mRNA in AC, in a strict dosedependent manner ( Figure 3C).

Differential Effect of Wnt/β Catenin Signaling Activation on Cardiac and Liver Specification
Wnt pathway activation has a well-documented attenuating effect on cardiac differentiation in vivo, in embryonic stem (ES) cell differentiation model and in Gata4-expressing AC from Xenopus embryos (Marvin et al., 2001;Schneider and Mercola, 2001;Tzahor and Lassar, 2001;Latinkic et al., 2003;Naito et al., 2006;Liu et al., 2007;Ueno et al., 2007). We have activated Wnt signaling in control animal cap explants or in those expressing Gata4 by zygotic co-expression of Wnt8. Wnt8 had no effect on its own on cardiac or liver markers but it attenuated cardiogenesis induced by Gata4 (Figures 4A,B; Latinkic et al., 2003). At the same time, the expression of the liver marker nr1h5 was unaffected (Figures 4A,B).
We have previously shown that activation of Wnt signaling opposes cardiac differentiation but not specification in AC/AE conjugates (Samuel and Latinkic, 2009). Using the same experimental approach, we have activated Wnt/β-catenin FIGURE 3 | Cerberus and hhex are required in anterior endoderm for liver and cardiac specification. (A) Injection of cer1 or hhex MOs (10 or 20 ng/embryo, respectively) in AC have no effect on heart and liver marker expression in AC/AE conjugates. This is in contrast to (B), where injection of cer1 or hhex MOs in AE leads to downregulation of both cardiac and liver marker gene mRNA levels in conjugates. Injection of mRNA (500 pg/embryo) coding for dominant negative Hhex-VP2 construct had the same effect as hhex MO. (C) Cardiac and liver specification deficiency in AC/cer1 MO-AE conjugates can be rescued by Cerberus. cer1 mRNA was injected in AC at increasing concentrations, from left to right; 100, 200, 500, and 1000 pg per embryo. All samples were cultured until stage control had reached stage 34 where PCR was carried out for indicated markers.
signaling in AC/AE at different time points, either uniformly by LiCl ( Figure 4C) or specifically in AC or AE by activation of an inducible chimaeric construct Lef-β-catenin-GR (LEF-β-GR, Figures 4D,E). Brief LiCl treatment causes a strong but transient activation of Wnt target genes siamois (sia) and nodal3.1, whose expression is undetectable 6 h after treatment, whereas activation of Lef-β-catenin-GR leads to a milder but sustained response (Samuel and Latinkic, 2009). Our results show that treatment with LiCl at an early time point, near the time of cardiac specification (st. 9) has no effect on either cardiac or liver specification, whereas treatment at the late neurula stage (st. 21) abolishes cardiac but not liver marker expression. Similarly, activation of Wnt signaling using Lef-β-catenin-GR in AC but not in AE affects cardiac but not liver markers.
Antagonism of Wnt signaling is required for cardiac differentiation and has been shown to promote specification of AE-derived liver. In Gata4-expressing AC explants, antagonism of Wnt by secreted antagonist Dkk-1 has been shown to enhance cardiogenesis (Latinkic et al., 2003). At the same time, nr1h5 expression is unaffected (Figures 5A,B). In AC/AE explants Dkk-1 does not affect cardio-and hepatogenesis ( Figure 5C). Wnt was suggested to be required for liver bud outgrowth, suggesting that the AC/AE model does not capture those later stages of liver development.

BMP Signaling Inhibition Attenuates Liver Cell Fate Specification
BMP signaling has been shown to be required for liver specification in a wide range of models (Gordillo et al., 2015). In agreement with this we show that cell autonomous inhibition of BMP signaling using truncated BMP receptor (BMPRI) targeted to foregut interferes with liver development in tadpoles (Figures 6A-F). We next wished to test the dependence of liver Frontiers in Physiology | www.frontiersin.org FIGURE 5 | Wnt/β-catenin signaling is not required for liver specification in Gata4-AC and AC/AE. (A) Gata4 and dkk-1 mRNA co-injection in animal pole explants causes a well-described increase in cardiac marker levels but has no effect on nr1h5. (B) qPCR analysis confirms these findings and extends them by showing no significant effect on hhex and foxa2. (C) dkk-1 expression in AC/AE (injected in AC) likewise has no effect on nr1h5. RT-PCR analyses were performed on st. 34 explants and sibling embryo controls. specification on BMP signaling in Gata4-induced hepatogenesis. Cardiogenesis induced by Gata4 in AC explants does not require BMP signaling (Latinkic et al., 2003; Figure 6G), however, inhibition of BMP signaling using truncated BMP receptor or a small molecule inhibitor Dorsomorphin lead to a decrease in expression of nr1h5 at st. 34 and hhex at st. 10 (Figures 6G,H). These results suggest that BMP signaling is required for hepatic, but not cardiac, induction by Gata4 in pluripotent AC explants.

FGF Signaling Is Not Required for Gata4-Mediated Liver Specification
FGF signaling has a well-documented role in liver specification (Zaret, 2016). We have next examined the involvement of FGF signaling in liver specification in Gata4-expressing AC explants. Downregulation of the FGF pathway using dominantnegative FGFR1 (XFD; Figures 7A,B) or small drug SU5402 (Supplementary Figure S3) has no effect on both liver and cardiac specification, suggesting that the FGF pathway is not required in Gata4-induced liver cell fate specification in AC explants. Effectiveness of SU5402 and XFD was shown by their ability to inhibit expression of early mesodermal marker and FGF target tbxt (formerly xbra; Figure 7D) and by induction of characteristic gastrulation defect phenotype (Supplementary Figure S4). In AC/AE explants, the FGF pathway is required for cardiogenesis immediately following formation of conjugates (Samuel and Latinkic, 2009; Figure 7C). Not surprisingly, under these conditions liver specification is also affected. In contrast, inhibition of the FGF pathway from st. 13 until the end of culturing period at st. 34 had no effect on cardiac differentiation but attenuated nr1h5 expression ( Figure 7C). Shorter treatment time windows (st. 16-23, 23-28, 28-34) had no major effect on cardiac and liver cell fate specification (Figure 7C).

DISCUSSION
In this report we have used two experimental models based on Xenopus embryos that permit induction of liver and cardiac fates, to investigate their specific signaling requirements.
The simpler of the models is based on Gata4-mediated induction of liver specification in pluripotent animal cap explants. In this model, when Gata4 mRNA is expressed throughout the explant, both cardiac and liver cell fates are induced, suggesting that fate acquisition is stochastic under conditions when Gata4 is most likely acting cell-autonomously. In comparison, a bona fide endoderm determinant Sox17 induces liver gene expression in animal pole explants only non-cell autonomously. It would be of interest to further explore cell autonomous mode of Gata4 action in liver specification in animal cap explants.
The second model used in the current study is based on heterochronic conjugates of gastrula-stage AE and blastula-stage animal cap explants. In these AC/AE conjugates AE induces cardiac specification in the overlaying animal cap ectoderm (Samuel and Latinkic, 2009), and in this study we have shown that AC/AE conjugates also express a range of endoderm markers. The liver tissue in conjugates is induced in the endoderm, adjacent to the cardiac domain which has been induced in the animal cap explant (Figure 1). This configuration closely resembles the spatial relationship of the heart and the liver in the early embryo and suggests that AC/AE conjugates recapitulate many aspects of cellular and molecular interactions that govern cardiac and liver specification in the embryo. In addition to the liver markers, AC/AE conjugates expressed a marker of lung and thyroid, nkx2-1, as well as subset of pancreatic markers-pdia2 and pdx1, but not ins, suggesting that partial reprogramming toward endocrine pancreas has been achieved by st. 43.
We have used the AC/AE model to examine the roles of hhex and cer1 in AC or AE. As expected (Foley and Mercola, 2005), hhex was found to be required in the AE for cardiac as well as for liver specification (Figure 3). Similarly, cer1 is required in the AE (Figure 3), in agreement with the findings by Foley et al. (2007). The deficiency of cer1 in AE can be rescued by injection of cer1 mRNA in the AC, but only in a narrow concentration range, suggesting that the pathways regulated by the Cerberus protein, BMP, Wnt, and Nodal (Piccolo et al., 1999), operate at a finely tuned level. In future it would be of interest to use AC/AE conjugates to dissect the requirement of each of these pathways, for example by asking whether and when Cerberus function could be replaced by small molecule inhibitors of Wnt, BMP and Nodal pathways, as well as to examine epistatic relationship between hhex and cer1.
Wnt pathway activation interferes with cardiogenesis both in Gata4-expressing animal cap explants and in AC/AE conjugates FIGURE 7 | FGF signaling is not required for liver specification by Gata4 but is essential in AC/AE explants. (A) Gata4 and XFD (dominant-negative FGFR1) mRNA co-expression in animal pole explants has no effect on nr1h5 and myl7 expression. (B) qPCR analysis confirms these findings and extends them by showing no significant effect on hhex and foxa2. (C) Treatment of AC/AE with FGFR inhibitor SU5402 (50 µM) has stage-specific effect. Immediately upon AC/AE conjugate formation, SU5402 inhibits cardiac and liver gene expression. Treatment from st. 13 until the end of incubation at st. 34 has no effect on cardiogenesis but inhibits liver specification. Shorter time windows of treatment are largely without effect. Staging is according to the stage of AE (st. 10 at the time of conjugation). C-Control explants treated with 50 µM DMSO. (D) SU5402 and XFD are effective inhibitors of early mesodermal marker and FGF target gene tbxt. CE-control (untreated) sibling embryos. RT-PCR analyses were performed on st. 34 explants and sibling embryo controls. (Latinkic et al., 2003;Samuel and Latinkic, 2009), but it does not significantly affect liver specification in both models (Figure 4). In Gata4-expressing AC explants liver specification is apparently independent of the presence of differentiated cardiomyocytes. One possibility is that inductive signaling between cardiac mesoderm and endoderm in this model occurs prior to cardiac differentiation and the second one is that liver cell fate is induced cell autonomously by Gata4.
We have previously shown that expression of cardiac precursor markers nkx2-5 and tbx5 is not affected by Wnt activation in AC/AE conjugates, suggesting that cardiac precursors transiently produce a liver-inducing signal. Under the conditions of Wnt pathway activation cardiac precursor are prevented to undergo differentiation into cardiomyocytes, showing that the production of the liver-inducing signal does not require cardiomyocytes and that the liver-inducing signals are likely transiently produced by cardiac progenitors.
Several studies have shown that Wnt antagonism is required for liver specification (Zorn and Wells, 2009;Zaret, 2016). Most relevant for the current study is the work by McLin et al. (2007) who have shown that Wnt antagonizes foregut development in Xenopus embryos. The apparent discrepancy between the two studies is likely due to the differences between the models that were used, in vivo by McLin et al. and explants in the current study. In animal cap explants Gata4 might be acting in parallel to or downstream of hhex in specifying liver, as hhex has been shown to be downstream of Wnt antagonism (Foley and Mercola, 2005).
Our results have suggested that BMP signaling is required for liver specification downstream of Gata4. It will be of interest to examine in more detail how and when BMP antagonism interferes with Gata4-induced hepatogenesis. In addition we have found that BMP signaling is required for liver specification in vivo, as previously reported (Zorn and Wells, 2009;Zaret, 2016).
Unlike BMP inhibition, interference with FGF signaling was found to have no effect on Gata4-driven liver cell fate specification in animal cap explants (Figure 7 and Supplementary Figure S3). In contrast, liver specification in AC/AE explants shows requirement for prolonged FGF signaling (st. 13-34), but shorter treatment windows during neurula and tailbud stages show no effect (Figure 7). This finding suggests that FGF signaling is required for liver specification over a prolonged period rather than within a discrete, well-defined time window. Early FGF inhibition immediately after conjugation of AC/AE explants inhibits cardiogenesis and this likely leads to an inhibition of liver specification as well.
Our results with AC/AE explant conjugates are in good overall agreement with those of Shifley et al. (2012) who reported that prolonged FGF signaling is required for liver specification in Xenopus embryos. Taken together with the results with manipulation of Wnt signaling, our findings suggest that cardiac precursors produce a liver specifying signal, most likely an FGF, which is required over a prolonged period to specify liver fate.
The principles of liver specification and differentiation that were established in various vertebrate embryos have been and are the key to the development and refinement of protocols for liver cell differentiation from pluripotent stem cells. This is an area of intense research due to potential medical applications and our findings may inform future attempts at refinement of differentiation protocols.

AUTHOR CONTRIBUTIONS
KH performed most of the animal cap experiments. LS performed most of the AC/AE conjugate experiments. SB and PK contributed to analysis of several experiments. BL contributed to experimental manipulation of embryos, planned the study, and wrote the manuscript.

FUNDING
This work was supported in part by project grants from Biotechnology and Biological Sciences Research Council grant BB/C517368 and British Heart Foundation (BHF) and Ph.D. studentships from BHF and Medical Research Council.