Specification of hepatopancreas progenitors in zebrafish by hnf1ba and wnt2bb

The liver and ventral pancreas, and presumably gall bladder and extra hepatopancreatic ducts, are believed to develop from a common multipotent progenitor population (hepatopancreas progenitors) present at early somitogenesis (Deutsch et al., 2001; Chung et al., 2008; Wandzioch and Zaret, 2009). In support of this model, multipotency of cells within this developing organ system appears to be maintained beyond early somitogenesis stages. For example, in ptf1a mutants, presumptive ventral pancreas progenitors will instead adopt intestinal (Kawaguchi et al., 2002), gall bladder, common bile duct (Burlison et al., 2008) or liver fate (Dong et al., 2008) (P.D.S.D., unpublished). Further observations supporting multipotency come from Sox17, Fgf10, and Hes1 loss-of-function studies where cells of the extra hepatopancreas ductal system can adopt pancreas or liver fate (Sumazaki et al., 2004; Fukuda et al., 2006; Dong et al., 2007; Spence et al., 2009). In addition, fgf10 mutants develop ectopic pancreas cells in the liver (Dong et al., 2007), and both fgf10 and sox9 mutants develop liver cells in the pancreas (Dong et al., 2007; Seymour et al., 2012). These examples demonstrating that cells within the developing hepatopancreas organ system can readily switch fate is consistent with the liver, ventral pancreas, gall bladder and associated ducts arising from a common multipotent progenitor pool.

It remains unresolved whether a common genetic mechanism specifies this entire progenitor population. Although significant progress has been made in identifying factors necessary for either pancreas or liver specification, factors uniquely required for induction of the entire ventrally derived hepatopancreas system have yet to be discovered. Several transcription factors including Ptf1a and Pdx1 have been implicated in ventral pancreas specification, but neither factor is required for liver specification (Offield et al., 1996; Kawaguchi et al., 2002). In Hhex or Gata4 mutants, the ventral pancreas is not specified, but early liver markers are expressed, suggesting liver specification has occurred (Bort et al., 2004; Bort et al., 2006; Watt et al., 2007). Although Foxa1 and Foxa2 function redundantly to establish endoderm competence necessary for liver specification (Kaestner, 2005; Lee et al., 2005), whether these factors act broadly to affect foregut endoderm patterning or have a more restricted role in liver and ventral pancreas specification has not been addressed.

Zebrafish and mouse embryos with severe or complete loss of HNF1B (also known as HNF1β, TCF2 and VHNF1) function exhibit profound foregut regionalization defects and subsequently fail to specify the liver and both the dorsal and ventral pancreas (Sun and Hopkins, 2001; Haumaitre et al., 2005; Lokmane et al., 2008). For this reason, it is unresolved whether the liver or pancreas agenesis defects observed in severe hnf1b loss-of-function mutants are secondary to the earlier foregut patterning defects. Loss of wnt2bb in zebrafish can lead to a low penetrant liver specification defect. Although the penetrance of this defect is increased when combined with wnt2 loss of function, swimbladder agenesis is also observed and, importantly, ventral pancreas specification still occurs (Ober et al., 2006; Poulain and Ober, 2011). It is not known whether other Wnt ligands act redundantly with wnt2bb and wnt2 to specify the ventral pancreas. Nevertheless, Wnt signaling has been implicated in pancreas development. Studies in frog showed that Wnt signaling before early somitogenesis plays a negative role in liver and pancreas development by inhibiting foregut identity (McLin et al., 2007). Studies in mice suggest that ectopic Wnt signaling during early pancreas development can antagonize pancreas specification (Heller et al., 2002; Heiser et al., 2006; Murtaugh, 2008). However, between these developmental stages, whether endogenous Wnt signaling regulates ventral pancreas specification has yet to be determined.

HNF1B is the gene implicated in maturity onset diabetes of the young 5 (MODY5), one of ten monogenic, dominantly inherited forms of early-onset diabetes (Horikawa et al., 1997; Borowiec et al., 2009; Nyunt et al., 2009; Wang et al., 2009). Based on genome-wide association (GWA) studies, HNF1B and its closely related paralog HNF1A (MODY3), are also associated with type 2 diabetes (T2D), suggesting broader significance for studying these monogenic diabetes genes (Frayling, 2007; Wang et al., 2009; Voight et al., 2010). Loss of insulin producing pancreatic β-cells is characteristic of nearly all forms of diabetes and is the underlying factor for diabetic insulin dependence. While it has been suggested that HNF1B is important for liver insulin sensitivity, it remains unclear whether its dysfunction in the pancreas, particularly the β-cells, contributes to the etiology of HNF1B-related diabetes (Brackenridge et al., 2006; Kornfeld et al., 2013). Although targeted knockout of Hnf1b from β-cells in mice leads to impaired glucose tolerance, fed or fasted plasma glucose and insulin levels were not affected, suggesting that Hnf1b may not be crucial for maintaining β-cell function (Wang et al., 2004). However, it is not known whether Hnf1b is required for normal β-cell development. In support of a role in β-cell development, lineage-tracing studies in mice reveal that Hnf1b is expressed in pancreas endocrine progenitors that can give rise to β-cells during development (Solar et al., 2009). GWA studies have also implicated the Wnt pathway, including WNT2B, as the highest ranked pathway associated with T2D (Perry et al., 2009). By analyzing the role of wnt2bb and hnf1b in pancreas development, we aim to advance our understanding of β-cell development and diabetes etiology.

The heterozygous nature of the HNF1B mutations in individuals with MODY5 suggests that partial loss of HNF1B function can lead to disease pathologies. However, to date only complete or severe loss-of-function Hnf1b alleles have been examined in mouse and zebrafish (Barbacci et al., 1999; Sun and Hopkins, 2001; Haumaitre et al., 2005; Lokmane et al., 2008). As would be expected, Hnf1b-null mutants exhibit more pleiotropic and severe defects than those observed in individuals with MODY5, limiting their utility as development and disease models. Whether HNF1B plays a distinct role in β-cell development or in pancreas and liver specification has been difficult to study because of the broader endoderm patterning defects associated with its severe loss of function. Our studies provide new insights into HNF1B function by uncoupling these defects. We report here a novel hypomorphic hnf1ba zebrafish mutant that does not exhibit significant foregut endoderm regionalization defects. Our analysis of this mutant leads to several important advances. We find that partial loss of hnf1ba function in zebrafish can cause MODY5-like pancreas defects. We also find that normal levels of hnf1ba function are distinctly required for proper β-cell numbers and for pancreas specification. Furthermore, from a synergistic genetic interaction between hnf1ba and wnt2bb, we uncover a new mechanism for hepatopancreas progenitor specification, revealing that a common genetic program, within a narrow developmental time window, specifies progenitors of the liver, ventral pancreas, gall bladder and associated ducts.

Zebrafish were raised and maintained under standard laboratory conditions. We used the following mutant and transgenic lines: liger^s430 (hnf1ba^s430), hnf1ba^hi2169 (Sun and Hopkins, 2001), wnt2bb^s404 (Ober et al., 2006), Tg(hsp70l:wnt8-GFP)^w34 (Weidinger et al., 2005), Tg(elastase:GFP);Tg(lfabp:dsRed)^gz2 (Wan et al., 2006), Tg(hsp70l:dkk1-GFP)^w32 (Stoick-Cooper et al., 2007) and Tg(ptf1a:eGFP)^jh1 (Godinho et al., 2005).

Genomic DNA was amplified with (F:5′-AGCAGCACAATATCCCTCAGCGCGAGGTCG-3′) and (R:5′-AAAACCGTCACATGCAACAA-3′) primers. Amplicons were then digested overnight with Taq^α1 (NEB). The forward primer introduces a mismatch creating a Taq^α1 restriction site in the hnf1ba^s430 amplicon (Neff et al., 1998). Following digestion, the wild-type allele generates a band of 195 bp and the liger^s430 allele a band of 165 bp.

Genomic DNA was amplified with (F:5′-CGTTCGTATACGCGATCTCC-3′) and (R:5′-TTCCACAGCGGTTGTTATGA-3′) primers followed by digestion with Fok1. The wild-type allele generates a band of 258 bp and the wnt2bb^s404 allele a band of 152 bp. hnf1ba^hi2169 mutants were genotyped as described previously (Sun and Hopkins, 2001).

One-cell stage embryos were injected with varying amounts of a hnf1ba morpholino (0.28-1.7 ng) (5′-GGGAAATGCGGTATTGTGATCTTTC-3′) (Gene Tools).

Embryos were heat-shocked at 21 hpf by adding egg-water pre-warmed to 38°C for Tg(hsp70l:dkk1-GFP)^w32 and 39°C for Tg(hsp70l:wnt8-GFP)^w34 and maintained at 37°C and 38°C, respectively, for 7 hours. Following heat-shock, embryos were sorted, returned to 27.5°C and collected at specified stages.

Whole-mount fluorescent immunohistochemistry was performed as previously described (Dong et al., 2007), using the following antibodies: mouse anti-Islet1/2 [1:10; Developmental Studies Hybridoma Bank (DSHB)], rabbit anti-pan-Cadherin (1:5000; Sigma), guinea pig anti-Insulin (1:500; Biomeda), mouse anti-Synaptic Vesicles (SV2; 1:20; DSHB), guinea pig anti-Pdx1 (1:200), rabbit anti-Prox-1 (1:200; Millipore) and mouse anti-Myosin to label somites following in situ hybridization (1:20; DSHB, F59), fluorescently conjugated Alexa secondary antibodies (Molecular Probes) and DAPI. Control and mutant (or morphant) embryos were stained in the same tube for each experiment. To analyze β-cell numbers, day 3 and 5 hnf1b^s430 and wild-type embryos were processed for confocal imaging using DAPI and antibodies against insulin and synaptic vesicles. Z-stacks (0.8 μm steps) were collected for the entire islet with optical slices captured at a focal depth of 1.6 μm. Samples were imaged using a Zeiss LSM5 Pascal or Zeiss 710 confocal microscope.

Full-length zebrafish wild-type and mutant hnf1ba (V147E) were subcloned into pcDNA3-Myc vector and co-transfected with β28-Luciferase reporter (a kind gift from Dr G. Crabtree, Stanford University, CA, USA) containing three copies of the HNF1-binding site from the fibrinogen promoter and with RSV-β-Gal, into Min6 or HepG2 cells in the absence or presence of both CMV Flag-pCAF or CMV Flag-CBP expression vectors, as indicated. Luciferase was measured using a luminometer, and values normalized for expression of β-Gal as described.

In situ hybridization was performed as described with minor modifications (Yelon et al., 1999; Dong et al., 2007). Following in situ hybridization, immunohistochemistry was performed to fluorescently label somites as previously described (Dong et al., 2007; Huang et al., 2008). Images were captured using a Zeiss 710 confocal microscope with AxioPlan 4.8 software. In situ hybridization and immunofluorescent images were captured for a single sample and merged using Photoshop CS3. Probes for hnf1ba, hhex, prox1, cmyc (myca - Zebrafish Information Network) and wnt2bb were donated and are described elsewhere (Ho et al., 1999; Sun and Hopkins, 2001; Yamaguchi et al., 2005; Ober et al., 2006; Shin et al., 2011). Experimental and control embryos were stained in the same tube and all experiments were carried out at least twice using distinct clutches.

From a zebrafish ENU mutagenesis screen (Ober et al., 2006), we identified liger^s430, a recessive mutant that exhibits a variably smaller pancreas, as marked by Tg(ptf1a:eGFP)^jh1 expression (Godinho et al., 2005; Dong et al., 2008) (Fig. 1A,B, arrows indicate pancreas). ptf1a, which encodes a basic helix-loop-helix transcription factor, is expressed in the pancreatic anlagen and is required for ventral pancreas specification (Kawaguchi et al., 2002; Lin et al., 2004; Zecchin et al., 2004; Dong et al., 2008). The liger^s430 mutation is linked to the simple sequence length polymorphism (SSLP) marker z4396 on chromosome 15, in the same region as hnf1ba (zfin.org).

Complementation analysis with a severe hnf1ba mutant, hnf1ba^hi2169, yielded transheterozygotes that phenocopied the liger^s430 homozygous variable pancreas hypoplasia (with greater severity and penetrance), indicating that the liger^s430 mutation is a hypomorphic allele of hnf1ba (Fig. 1C-F). In addition, liger^s430 homozygous and liger^s430/hnf1ba^hi2169 transheterozygous embryos do not exhibit kidney cysts (not shown) or liver agenesis, as reported in the hnf1ba^hi2169, hnf1ba^hi548, hnf1ba^hi1843, hnf1ba^wi408 and hnf1ba^la550 homozygotes, providing evidence that liger^s430 is hypomorphic to these published alleles (Sun and Hopkins, 2001; Wiellette and Sive, 2003; Song et al., 2007). Morpholino (MO) antisense knock-down of hnf1ba translation in embryos (morphants) phenocopies the range of pancreas hypoplasia and agenesis defects observed in liger^s430 homozygotes and liger^s430/hnf1ba^hi2169 transheterozygotes, depending on the amount injected (Fig. 1G-I; supplementary material Fig. S3), providing further evidence that the liger^s430 mutation is a hypomorphic allele.

Sequence analysis of hnf1ba in liger^s430 mutants revealed a transversion of a thymine to an adenine in the coding region at position 440, resulting in a valine to glutamic acid (V147E) missense mutation (Fig. 1J). This valine is highly conserved among vertebrates and is in the DNA-binding POU-specific domain where MODY5 and MODY3 (Hnf1a) missense mutations are most commonly found (Fig. 1K) (Chi et al., 2002; Gong et al., 2004; Edghill et al., 2006a). Based on crystal structure data of human HNF1B in complex with DNA, V140 in human HNF1B, which is homologous to V147 in zebrafish (Fig. 1K), is predicted to be involved in the formation of the POU-specific domain hydrophobic core but not make direct contact with DNA (supplementary material Fig. S1A) (Lu et al., 2007). Therefore, substitution of a highly charged group (V147E) is predicted to perturb the hydrophobic core and disrupt protein structure and function.

Consistent with the predicted structural model, we find that when compared with wild-type zebrafish Hnf1ba, V147E Hnf1ba fails to efficiently activate transcription of a luciferase reporter driven by a multimerized HNF1-binding element of the fibrinogen gene (supplementary material Fig. S1B) (Chi et al., 2002). In addition, although decreased relative to wild-type Hnf1ba, V147E Hnf1ba transcriptional potential can be enhanced by known co-activators pCAF and CBP (Barbacci et al., 2004), further supporting that V147E Hnf1ba retains some function (supplementary material Fig. S1B). Together with genetic linkage to the hnf1ba locus, phenocopy of liger^s430 homozygous phenotypes by hnf1ba knock-down, failure of complementation by hnf1ba^hi2169 and substitution of a highly conserved residue of Hnf1ba, we conclude that liger^s430 (now designated hnf1ba^s430) is caused by a hypomorphic mutation in hnf1ba. We note that hnf1bb, a paralog of hnf1ba in zebrafish, was not reported to be expressed in the developing foregut endoderm (Choe et al., 2008; Wingert and Davidson, 2011; Naylor et al., 2013). Because loss of hnf1ba in zebrafish and Hnf1b in mouse exhibit similar foregut endoderm phenotypes (Sun and Hopkins, 2001; Haumaitre et al., 2005; Lokmane et al., 2008), we suggest that hnf1ba is the primary functional ortholog of mammalian Hnf1b in foregut development.

Foregut regionalization defects and pancreas agenesis in severe Hnf1b loss-of-function zebrafish and mouse mutants limit their utility as models for Hnf1b function in the developing and adult pancreas (Sun and Hopkins, 2001; Haumaitre et al., 2005). With this newly identified hypomorphic hnf1ba mutant, we have the unique opportunity to examine the consequence of Hnf1ba partial loss of function during pancreatogenesis. We find that the developing hnf1ba^s430 mutant exocrine pancreas, which is thought to arise exclusively from the ventral pancreas in zebrafish (Field et al., 2003), is variably reduced in size based on Tg(ptf1a:eGFP)^jh1 expression (Fig. 1A,B). In fewer than 15% of hnf1ba^s430 homozygotes (n>300), the ventral pancreas can appear either normal in size (<10%) or completely absent (<10%), with variation between clutches. Because we find that MODY5-like pancreas hypoplasia can result from a quantitative loss of Hnf1ba (via translational knock-down), we propose that pancreas hypoplasia in MODY5 is due to decreased levels of HNF1B function. The dorsal pancreas, which comprises primarily endocrine cells in zebrafish (Field et al., 2003; Hesselson et al., 2009), is dysmorphic in hnf1ba^s430 homozygotes. In mutant embryos, endocrine cells, which arise from both the dorsal and ventral pancreas, can fail to form a single islet cluster. Furthermore, the clusters of endocrine cells that do form do not organize into an islet with a central core of β-cells, normally observed in zebrafish and mice. We find a significant and consistent reduction in the number of β-cells in hnf1ba^s430 homozygotes, regardless of exocrine size, although the severity of reduction generally correlates with exocrine size (Fig. 2A). To determine whether the roles of hnf1ba in regulating β-cell number and exocrine size are distinct, we compared β-cell numbers in 80 hpf wild-type embryos with those of 125 hpf mutants with a similar or larger exocrine pancreas. We find that age-matched or older mutant embryos with an exocrine pancreas similar to or larger than that of an 80 hpf wild type have significantly fewer β-cell numbers (n>20) (Fig. 2B-G; supplementary material Fig. S2A-D). This finding indicates that hnf1ba regulation of β-cell number is distinct from its regulation of pancreas size.

Although severe loss of hnf1ba does lead to ventral pancreas agenesis in zebrafish, it also causes loss of adjacent endodermal organs, including the liver, dorsal pancreas and swim bladder, indicating a more profound role for hnf1ba in posterior foregut patterning (Sun and Hopkins, 2001). In Hnf1b-null mice, in which ventral pancreas agenesis occurs, only rudiments of non-differentiating liver and dorsal pancreas buds were observed, also suggesting a broader role for Hnf1b in endoderm development and potentially obscuring a distinct role in pancreas specification (Haumaitre et al., 2005). Consistent with a role for Hnf1b in AP (anterior/posterior) patterning of the foregut endoderm, both zebrafish and mouse Hnf1b-null mutants exhibit a posterior expansion of hnf3b and sonic hedgehog (shh) expression into the presumptive duodenum (posterior foregut), the region from which the ventral and dorsal pancreas and liver develop (Sun and Hopkins, 2001; Haumaitre et al., 2005). Hedgehog signaling in the anterior endoderm is required early to pattern the anterior foregut via repression of more posterior endoderm fate (duodenum/pancreas/liver) (Hebrok, 2003; Wan et al., 2006). Therefore, posterior expansion of shh expression would be expected to inhibit posterior foregut patterning events. Accordingly, zebrafish and mouse hnf1b-null embryos exhibit loss of expression of the posterior foregut endoderm (pancreas/duodenum) marker, pdx1, but not loss of the anterior foregut/endoderm markers hnf3b, nkx2.1 and irx2 (Sun and Hopkins, 2001; Haumaitre et al., 2005; Lokmane et al., 2008). These observations indicate a conserved effect of severe hnf1b loss-of-function mutations on the AP patterning of the vertebrate foregut (Sun and Hopkins, 2001; Haumaitre et al., 2005). In contrast to hnf1ba^hi2169 mutants, AP foregut endoderm patterning appears normal in hypomorphic hnf1ba^s430 embryos with pancreas agenesis. Duodenal Pdx1 expression is not lost in these hnf1ba^s430 mutants, suggesting that the duodenal region (posterior foregut) is properly patterned (Fig. 2J). More important, adjacent endodermal organs (swim bladder, dorsal pancreas and liver) do develop and differentiate in hnf1ba^s430 mutants with ventral pancreas agenesis, further indicating that foregut AP patterning is not significantly impaired (Fig. 2K-M). These results reveal that ventral pancreas development is more sensitive to reduced Hnf1ba function than neighboring tissues, thereby suggesting that hnf1ba is distinctly required for ventral pancreas specification, independent of its role in foregut endoderm regionalization and liver specification. Our in vivo data are consistent with previous molecular studies suggesting that Hnf1b transcriptionally activates Hnf6 to regulate Pdx1 and ventral pancreas specification (Poll et al., 2006).

It is unresolved whether hnf1b also plays a distinct role in liver specification because of the difficulty in uncoupling the liver agenesis phenotype from the foregut regionalization defects exhibited by severe hnf1b mutants (Sun and Hopkins, 2001; Lokmane et al., 2008). Because wnt2bb mutants display rare liver specification (<1%, n=200) and transient hypoplasia defects (Ober et al., 2006) (Fig. 3A,B), and because hnf1ba^s430 mutants exhibit variable but mild hepatic hypoplasia (Fig. 3A,C), we tested for a synergistic genetic interaction between hnf1ba^s430 and wnt2bb^s404 with respect to liver specification. Remarkably, about 80% of the wnt2bb^s404; hnf1ba^s430 double mutants (44/55) completely lack hepatocytes, as assessed by Tg(lfabp:dsRed)^gz2 expression, indicating that hnf1ba is indeed required for liver specification. The penetrance and severity of the liver agenesis and hypoplasia defects in double mutants is significantly greater than the additive penetrance and severity of the single mutants, indicating a synergistic genetic interaction. Surprisingly, pancreatic agenesis/hypoplasia is also more severe and penetrant in double mutants (Fig. 3D), revealing a novel role for wnt2bb signaling in pancreas specification (wnt2bb^s404 mutants do not exhibit obvious pancreas hypoplasia). Strikingly, in 6% of the double mutants (3/48), the entire ventrally derived hepatopancreas system, including the pancreas, gallbladder, extra-hepatopancreas ducts [assessed with 2F11 antibodies (Crosnier et al., 2005; Dong et al., 2007)] and liver, is absent (Fig. 3E,F), suggesting a more profound role for hnf1ba and wnt2bb in specifying these developmentally associated organs (supplementary material Table S1). Islet1 (Isl1) expression in mesenchymal cells that surround the extra-hepatopancreas ducts and adjacent intestine (Dong et al., 2007) appears normal in these double mutants (Fig. 3G,H). Importantly, neighboring endodermal organs, including the swim bladder and dorsal pancreas, do develop in these double mutants, indicating that foregut AP patterning is not significantly affected in the wnt2bb^s404; hnf1ba^s430 double mutants. Other than hepatopancreas organ system defects, wnt2bb^s404; hnf1ba^s430 double homozygous embryos are indistinguishable from their wild-type and single mutant siblings, based on brightfield microscopy (Fig. 3I; Fig. 1A,B). Therefore, agenesis defects observed in the wnt2bb^s404; hnf1ba^s430 double mutants are highly restricted to ventrally derived hepatopancreas organs. Furthermore, unlike wnt2bb^s404 or hnf1ba^s430 single mutants, which can be homozygous viable as adults, double mutants do not survive beyond 9 dpf, indicating a synthetic lethal genetic interaction. To our knowledge, this is the first report of a genetic condition that leads to specific and complete agenesis of the entire ventrally derived hepatopancreas system, demonstrating that the liver and ventral pancreas are specified by a common genetic mechanism.

Current models suggest that genetic events regulating liver and pancreas specification occur during early somitogenesis (Chung et al., 2008; Wandzioch and Zaret, 2009; Shin et al., 2011). However, the liver and ventral pancreas do not bud until after mid-somitogenesis, suggesting that other genetic events may more directly regulate hepatopancreas fate in the foregut endoderm. To assess the temporal requirement of the synergistic genetic interaction between Wnt signaling and Hnf1ba in hepatopancreas specification, we mis-expressed the Wnt/β-catenin antagonist Dickkopf (Dkk1) (Glinka et al., 1998) in a temporally specific manner via heat-shock using the Tg(hsp70l:dkk1-GFP)^w32 line in the hnf1ba^s430 background (Fig. 3J,K). Heat-shock of Tg(hsp70l:dkk1-GFP)^w32 or hnf1ba^s430 embryos for 7 hours beginning at 21 hours post fertilization (hpf, ∼24-somite stage, 3 hours before initial expression of the hepatopancreas progenitor markers hhex and prox1) did not yield hepatopancreas agenesis phenotypes (Fig. 3J; not shown). However under the same conditions, 38% (5/13) of the Tg(hsp70l:dkk1-GFP)^w32 embryos in the hnf1ba^s430 mutant background showed a loss of the entire hepatopancreas system, including the liver, ventral pancreas, and associated ducts (Fig. 3K). These hepatopancreas phenotypes were indistinguishable from those of wnt2bb^s404; hnf1ba^s430 double mutants, though they occurred with greater penetrance. Importantly, we emphasize that all organ agenesis defects observed are permanent and do not recover at later stages, strongly supporting a specification defect rather than a general developmental delay. The complete loss of the entire hepatopancreas organ system in wnt2bb^s404; hnf1ba^s430 double mutants and Tg(hsp70l:dkk1-GFP)^w32; hnf1ba^s430 embryos suggests a possible defect in specification of hepatopancreas progenitors. Consistently, heat-shocked Tg(hsp70l:dkk1-GFP)^w32; hnf1ba^s430 embryos lack early Prox1 and hhex expression in the ventral foregut at 36 hpf (a stage when their expression in both liver and pancreas progenitors are consistently detectable in wild type and single mutants) without detectable reduction of Pdx1 expression in the presumptive duodenum, indicating a specific loss of hepatopancreas progenitors (Fig. 4A-H; n=30). Interestingly, prox1 and hhex, which we demonstrated to be downstream of Hnf1ba and Wnt signaling during hepatopancreas specification, have also been associated with increased diabetes risk by GWA studies (Dupuis et al., 2010). A genetic interaction among these factors in hepatopancreas development may hint at their potential interactions in contributing to diabetes. We note that heat-shock induction at later stages (23 hpf or later) dramatically decreased the penetrance of the hepatopancreas system agenesis phenotype, suggesting that initial Wnt signaling between 21 and 23 hpf is crucial for hepatopancreas specification. And based on when Wnt inhibition was needed to efficiently block hepatopancreas specification, we suggest that between 21 and 28 hpf, just prior to liver budding, Wnt signaling is required for hepatopancreas progenitor specification. Blocking Wnt signaling at later stages (48 hpf) via ectopic dkk1 expression led to a reduction in liver size (Goessling et al., 2008), suggesting a later role for Wnt signaling in liver expansion, distinct from its earlier role in hepatopancreas progenitor specification.

The highly specific loss of the hepatopancreas system in wnt2bb^s404; hnf1ba^s430 double mutants suggests that Hnf1ba and Wnt2bb functions converge at the foregut endoderm to induce hepatopancreas fate. To determine the mechanism of this genetic interaction, we analyzed the expression of hnf1ba, wnt2bb and prox1 at the earliest stage of hepatopancreas development (Sun and Hopkins, 2001; Ober et al., 2006; Huang et al., 2008; Lokmane et al., 2008; Noël et al., 2008). At 24 hpf, hnf1ba expression extends from the presumptive pancreas/liver to the presumptive duodenal regions of the foregut endoderm (from AP region of somite 1-5; Fig. 5A-C). At this stage, wnt2bb shows confined expression (somite 1-2) in the lateral plate mesoderm adjacent to the anterior-most region of the hnf1ba expression domain. Coincidently, it is in this anterior subdomain of hnf1ba expression that initial prox1 expression becomes detectable. Together with our functional data, we hypothesize that hnf1ba expression defines the area of the foregut endoderm that can be induced by Wnt signaling to express Prox1 and adopt hepatopancreas fate (Fig. 5D). According to this model, broad overexpression of Wnt ligands should lead to posterior expansion of Prox1 expression where hnf1ba is expressed. As predicted and consistent with previous studies (Poulain and Ober, 2011; Shin et al., 2011), broad overexpression of Wnt8a via heat-shock of Tg(hsp70l:wnt8a)^w34 embryos (Weidinger et al., 2005) beginning at 21 hpf resulted in a robust posterior expansion of Prox1 expression within the hnf1ba-positive domain of the foregut endoderm (Fig. 5E,F; n=30). Crucially, in hnf1ba^s430/hi2169 transheterozygotes, Prox1 expression is either not induced or is weakly induced in the foregut endoderm following ectopic Wnt8 expression (Fig. 5G; 100%, n=6; and inconsistently induced in hnf1b^s430 homozygotes; not shown). These data demonstrate that hnf1ba function is indeed required for Wnt signaling to induce hepatopancreas fate. A possible mechanism for this genetic interaction is that hnf1ba is necessary for transduction of the Wnt signal in the foregut endoderm. To investigate this, we examined the expression of cmyc (myca), a direct Wnt/β-catenin target (He et al., 1998), which is expressed in hepatopancreas progenitors and whose expression domain can be induced posteriorly by ectopic Wnt8a expression (Fig. 5H) (Yamaguchi et al., 2005; Shin et al., 2011). Intriguingly, this posterior expansion also correlates with the hnf1ba expression domain (compare Fig. 5I with Fig. 5A). To test whether cmyc expression in the foregut is Wnt signaling dependent, we overexpressed the Wnt antagonist Dkk1, starting at 21 hpf and found a severe reduction in cmyc expression (n=25) (Fig. 5J). These data suggest that cmyc expression is a faithful marker of endogenous Wnt signaling activity in the foregut endoderm. In hnf1ba^s430 homozygotes, cmyc foregut expression is markedly reduced, suggesting that hnf1ba function is important for Wnt signaling in the foregut endoderm (Fig. 5K; n>30). Furthermore, in hnf1ba^s430/hi2169 transheterozygotes, cmyc foregut expression cannot be robustly induced by ectopic Wnt8, suggesting that hnf1ba is important for foregut endoderm Wnt signaling activity (Fig. 5L; 100%, n=12; and inconsistently induced in hnf1ba^s430 homozygotes). Because these results demonstrate that hnf1ba is required for the expression of a known transcriptional target of Wnt signal transduction, cmyc, we suggest that hnf1ba functions to generate a permissive domain in the foregut endoderm for Wnt signaling. A role for Hnf1ba in Wnt signaling could explain the severe, yet localized, foregut defect observed when combining the wnt2bb^s404 and hnf1ba^s430 mutations. In wnt2bb^s404; hnf1ba^s430 double homozygous mutants, Wnt2bb loss of function would be compounded by the reduced foregut Wnt signaling that results from Hnf1ba partial loss of function. We suggest that this synergistic reduction of Wnt signaling contributes to the loss of hepatopancreas specification in wnt2bb^s404; hnf1ba^s430 double mutants. It is possible that non-Wnt-related targets of Hnf1ba also contribute to the specification of hepatopancreas progenitors.

Our studies suggest that Hnf1ba and Wnt2bb coordinate the specification of the progenitor population that gives rise to the liver, ventral pancreas, gall bladder and associated ducts, but not neighboring endodermal tissue such as the dorsal pancreas, demonstrating a common genetic mechanism for specification of the entire ventrally derived hepatopancreas system. Although it has been suggested that genetic events regulating liver and pancreas specification occur during early somitogenesis (Chung et al., 2008; Wandzioch and Zaret, 2009), we show the Hnf1ba and Wnt signaling interact later, only several hours prior to liver budding, to induce hepatopancreas fate. It is well established that signals from the mesoderm play a key role in patterning the foregut endoderm during development (Kumar et al., 2003; Ober et al., 2006; Sneddon et al., 2012). But it is less clear how the endoderm regulates the reception of these signals. We provide in vivo evidence that Hnf1ba function is required for the foregut endoderm to respond to Wnt2bb signaling from the mesoderm.

We anticipate that this conceptual advance in understanding how hepatopancreas progenitors are normally specified could contribute to more efficient production of differentiated liver or pancreas cells to potentially treat liver disease and diabetes. HNF1B+ cells are produced in protocols designed to generate pancreatic endocrine cells from embryonic stem cells (Kroon et al., 2008). Based on our discoveries, we suggest that stimulating Wnt signaling in such HNF1B+ endoderm cells would increase the efficiency of current stem cell differentiation protocols by increasing the hepatopancreas progenitor pool. Furthermore, an in vitro study using human embryonic stem cells suggests that a restricted level of Wnt signaling, at a stage when HNF1B expression is robust, can enhance development of pancreas fate by 15-fold (Nostro et al., 2011), supporting a positive role for Wnt signaling in human pancreas development. This apparent functional conservation further validates zebrafish as a reliable vertebrate model for human pancreas development. Notably, our work suggests that there is a narrow developmental window when positive Wnt signaling input is required for specification of both liver and pancreas progenitors. Because Wnt signaling outside this window may have negative effects on pancreas specification (Murtaugh, 2008), precise stimulation of HNF1B+ endoderm cells by Wnt signaling in differentiation protocols may be necessary. Future studies to determine how progenitors specified by Hnf1b and Wnt signaling are subsequently allocated into distinct liver, pancreas, gall bladder, or duct fate will be important to further enhance stem cell differentiation efforts.

Despite the discovery of MODY5 over 14 years ago, the role HNF1B plays in pancreas development and diabetes has remained elusive, hindered by the lack of a suitable animal model system. Homozygous Hnf1b-null mutants exhibit broad defects that preclude pancreas and liver morphogenesis studies. Furthermore, heterozygous null Hnf1b mutants were not reported to exhibit MODY5-like pancreas defects, suggesting that Hnf1b dose regulation and/or requirements are different among vertebrates. Our studies demonstrate the utility of a homozygous hypomorphic hnf1ba mutant model that mimics pancreas pathologies observed in humans with heterozygous HNF1B mutations. Interestingly, variation in the severity of pancreas hypoplasia found in this hnf1ba zebrafish model is also characteristic of MODY5 cases (Bellanné-Chantelot et al., 2004). This variation is even found among patients within the same family (Barbacci et al., 2004; Edghill et al., 2006a). We postulate that, like homozygous hnf1ba^s430 zebrafish mutants, heterozygous HNF1B mutations in individuals with MODY5 result in a reduction of HNF1B function to a crucial threshold that is sensitive to stochastic developmental variation (Gärtner, 1990; Baranzini et al., 2010; Raj et al., 2010), which can be sensitive to variations in genetic background. Using this novel MODY5 model, we reveal that reduced Hnf1ba function can compromise developmental β-cell numbers, and show that this reduction is not secondary to pancreas hypoplasia defects. Furthermore, as we find in hnf1ba^s430 zebrafish mutants, exocrine pancreas hypoplasia and islet disorganization were also observed in a study of two human fetuses harboring heterozygous HNF1B mutations (Haumaitre et al., 2006). These common defects suggest that Hnf1b function in pancreas development is broadly conserved among vertebrates. However, a reduction in the number of β-cells was not reported in that study. We suggest that, similar to the variation in pancreas hypoplasia in both individuals with MODY5 and zebrafish hnf1ba^s430 homozygotes, a decrease in embryonic β-cell numbers in MODY5 is also variable - which may explain the variable onset of MODY5 diabetes. In support of this idea, rare cases of neonatal diabetes have been reported for MODY5, suggesting that β-cell numbers may be developmentally compromised in some people with HNF1B mutations (Edghill et al., 2006b; Beckers et al., 2007). In addition to the previously suggested roles for Hnf1b in β-cell insulin secretion and hepatic insulin signaling (Wang et al., 2004; Brackenridge et al., 2006; Kornfeld et al., 2013), our findings raise the possibility that a reduction in β-cell number may be a significant contributing factor to the etiology of HNF1B diabetes, potentially contributing to insulin dependence in MODY5. Furthermore, using this hypomorphic Hnf1b mutant model, we provided definitive evidence that hnf1ba plays a key role in ventral pancreas specification (Haumaitre et al., 2005), upstream of ptf1a and distinct from its roles in regulating foregut regionalization and liver specification. By revealing specific roles for hnf1ba in β-cell and pancreas development, our findings set the stage for future work to explore the mechanism of these crucial functions.

Our studies revealing an unexpected coordinate role of hnf1ba with wnt2bb in hepatopancreas specification also lays the foundation for future research to explore how Hnf1ba, as a transcription factor, functions to generate a Wnt permissive foregut endoderm and whether this genetic interaction is conserved in other tissues or in disease. Hnf1ba could be functioning in parallel with Wnt signaling by independently regulating Wnt target genes or upstream of Wnt signaling by regulating a component of the Wnt signal transduction pathway, including co-factors. Previous biochemical studies have suggested that Hnf1b can regulate target genes via epigenetic modification (Barbacci et al., 2004; Verdeguer et al., 2010). Future studies that seek to determine the precise molecular nature of this genetic mechanism will need to investigate these avenues of Hnf1b function. An understanding of this interaction may also have implications for their potential interaction in disease, as GWA studies have revealed that both HNF1B and the Wnt effector gene TCF7L2 are strongly associated with T2D (Frayling, 2007). With our finding that Hnf1ba is required for expression of the direct Wnt signaling target cmyc, we suggest that a possible consequence of HNF1B deficiency in diabetes may be the attenuation of Wnt signaling. Consistent with this model, a human hypoplastic MODY5 fetal pancreas was shown to have decreased expression of a Wnt signaling component β-catenin (Haumaitre et al., 2006).

The hypomorphic hnf1ba^s430 allele presented here will be a valuable model for further investigations into the genetic mechanisms of embryonic and post-embryonic pathologies associated with MODY5, including maturity onset diabetes and pancreas, liver and kidney dysfunctions. Because mouse Hnf1b-null heterozygotes were not reported to have MODY5-like phenotypes, we suggest that a mammalian model with a homozygous Hnf1b hypomorphic mutation similar to hnf1ba^s430 may be necessary to phenocopy MODY5 pancreas defects. This approach of using hypomorphic alelles was shown to be necessary to model heterozygous SOX2 mutations in humans with anophthalmia-esophageal-genital (AEG) syndrome (Que et al., 2007). Importantly, an Hnf1b mouse model with a broad and partial reduction in Hnf1b function would be more comparable with the MODY5 genetic condition than would a conditional Hnf1b knockout model. Such a mouse would be valuable to complement our findings on the functions of Hnf1b in vertebrate hepatopancreas progenitor specification, particularly if combined with precise Wnt signaling inhibition.

We thank Jeremiah Gilmore for molecular biology support, Haowen Zhou, David Nguyen, Ana Ayala and Steve Waldron for expert help with the fish, and Pamela Itkin-Ansari, Amnon Schlegel, Elke Ober, Silvia Curado, Silvia Cereghini and Isla Cheung for discussions and/or critical reading of the manuscript. We thank Zhaoxia Sun, Julia Horsfield and Tatjana Piotrowski for reagents.