GATA4 and GATA6 regulate pancreatic endoderm identity through inhibition of hedgehog signaling

The pancreas is a vital organ that functions to regulate digestive processes and glucose homeostasis (Hegyi and Petersen, 2013; Mastracci and Sussel, 2012). In adults, the three major diseases associated with the pancreas are pancreatitis, pancreatic cancer and diabetes. In addition, congenital genetic defects contribute to defective organ development, such as annular pancreas, and pathological conditions, including pancreatic agenesis and neonatal diabetes (Etienne et al., 2012; Rubio-Cabezas and Ellard, 2013).

Pancreas development is initiated in two distinct regions of the foregut endoderm in response to signals derived from adjacent tissues (Chen et al., 2004; Kim and MacDonald, 2002; Kumar et al., 2003; Zaret and Grompe, 2008; Zaret et al., 2008). In particular, several studies have shown that a major function of these signals is to repress hedgehog signaling in the presumptive pancreatic endoderm. In the prospective dorsal pancreatic endoderm, notochord-derived activin and fibroblast growth factor (FGF) signals are necessary for the inhibition of sonic hedgehog (Shh) to allow the induction of bud morphogenesis and the pancreatic transcriptional program (Apelqvist et al., 1997; Hebrok et al., 1998). In the prospective ventral pancreatic endoderm, signals secreted by the lateral plate mesoderm, cardiac mesoderm and septum transversum play important roles in inducing pancreatic versus liver fates; the inhibition of WNT and bone morphogenetic protein (BMP) signaling is necessary for ventral pancreas induction, and FGF induces the local expression of Shh to inhibit pancreatic fates in favor of liver lineages (Deutsch et al., 2001; Zaret and Grompe, 2008). As development proceeds, the dorsal and ventral pancreatic buds merge, pancreatic cell specification is initiated and the diverse pancreatic cell types differentiate and proliferate to form the mature functional organ (reviewed by Jorgensen et al., 2007; Pan and Wright, 2011; Pictet and Rutter, 1972).

The GATA regulatory proteins belong to a highly conserved six-member family of zinc finger transcription factors that play essential distinct and overlapping roles during embryonic development, including germ layer specification, organ formation and cell lineage determination. GATA1, GATA2 and GATA3 are important for hematopoiesis, whereas GATA4, GATA5 and GATA6 are important for the development of mesoderm- and endoderm-derived organs, including heart, liver and pancreas (Zhou et al., 2012). Gata4 null mice die at around embryonic day (E) 9.5 owing to defects in heart morphogenesis (Kuo et al., 1997; Molkentin et al., 1997; Narita et al., 1997) and Gata6 null mice die before E7.5 due primarily to defects in extra-embryonic endoderm (Koutsourakis et al., 1999; Morrisey et al., 1998). In the pancreas, Gata4 and Gata6 have overlapping expression in the pancreatic endoderm, but gradually become expressed in separate domains: Gata4 becomes restricted to the exocrine compartment and Gata6 is predominantly expressed in the endocrine compartment (Decker et al., 2006; Ketola et al., 2004). Using a Cre-lox approach, we previously demonstrated that simultaneous deletion of Gata4 and Gata6 from pancreatic progenitor cells leads to pancreatic agenesis in newborn mice (Carrasco et al., 2012; Xuan et al., 2012). The importance of GATA4 and GATA6 in human pancreas development has also been highlighted in reports of genetic cases of human pancreatic agenesis: GATA6 haploinsufficiency contributes to the majority of pancreatic agenesis cases (Lango Allen et al., 2012), and GATA4 haploinsufficiency has been documented in a small number of patients with pancreatic agenesis (Shaw-Smith et al., 2014).

To characterize the molecular changes underlying GATA4/6-mediated pancreas development, we performed global transcriptome analysis on pancreatic buds isolated from E12.5 embryos from control and Gata4^fl/fl; Gata6^fl/fl; Pdx1:Cre (pDKO); R26R:Tomato mice. Consistent with impaired pancreas development, the majority of downregulated genes were pancreatic progenitor markers. Surprisingly, however, many of the upregulated genes included a large number of intestinal and gastric transcription factors. Furthermore, genetic lineage tracing of cells expressing the pancreatic progenitor marker Pdx1 indicated that the pancreatic lineages were converted to stomach and intestinal cell fates. In addition, there was a notable upregulation of many components of the hedgehog pathway, suggesting that the major mechanism of GATA4/6-regulated pancreas development is through suppression of the hedgehog pathway. Previous studies have demonstrated that GATA factors can negatively regulate Shh expression in the stomach, limb bud, and somites (Daoud et al., 2014; Jacobsen et al., 2005; Kozhemyakina et al., 2014). Our studies suggest that GATA4 and GATA6 pattern the foregut endoderm through the repression of hedgehog signaling.

Simultaneous deletion of Gata4 and Gata6 in the pancreatic progenitor domain resulted in severely aplastic pancreatic buds (Carrasco et al., 2012; Xuan et al., 2012). To characterize the molecular pathways that function downstream of GATA4 and GATA6 to mediate the regulation of pancreatic differentiation and morphogenesis, we performed genome-wide transcriptome analysis of the arrested pancreatic rudiments from pDKO embryos compared with littermate control embryos. A R26R;Tomato fluorescent reporter was introduced into the strain to facilitate purification of the PDX1-derived lineages from pooled pancreata of each genotype (Fig. 1A). Quantitative real-time PCR (qPCR) was used to confirm deletion of Gata4 and Gata6, and validate the reduction in pancreatic progenitor markers Pdx1 and Ptf1a (Fig. S1). RNA-Seq analysis of pooled wild-type versus pDKO Tomato+ pancreata identified ∼251 genes that were significantly differentially expressed in the mutant embryos: 118 genes were downregulated and 133 genes were upregulated (Fig. 1B,C). Strikingly, hedgehog signaling was revealed to be the most affected pathway [using ingenuity pathway analysis (IPA) software], with many of the transcriptionally regulated components of the hedgehog pathway being dramatically upregulated (Table 1). qPCR confirmed a 30- to 50-fold upregulation of the two major hedgehog ligands Shh and Indian hedgehog (Ihh), and a 7- to 10-fold upregulation of the receptor patched homolog 1 (Ptch1) and the downstream transcriptional activator Gli1 (Fig. 1D). In situ hybridization for Shh on E10.5 embryos revealed the erroneous expression of Shh expression throughout the pDKO pancreatic domain (Fig. 1E). As inhibition of Shh expression by notochord-derived activin and FGF or cardiac mesoderm-derived FGF has been postulated to be an early event in pancreas induction (Deutsch et al., 2001; Hebrok et al., 1998), our findings suggest that repression of hedgehog signaling downstream of these mesodermal signals is mediated by GATA4 and GATA6 activity.

Previous studies have demonstrated that the absence of Shh expression in the early pancreatic domain is required for normal pancreas development (Hebrok et al., 2000). Furthermore, ectopic upregulation of Shh (Apelqvist et al., 1997; Haumaitre et al., 2005) is associated with diminished pancreas formation and expanded stomach or gut regionalization. Consistent with the elevation of hedgehog signaling in the pDKO pancreatic endoderm domain, there was a large reduction of pancreatic progenitor markers in the pDKO, whereas there was a notable increase in genes encoding stomach progenitor markers, such as Sox2 and Nkx6-3 (5.8- and 19.7-fold, respectively), and intestinal progenitor marker genes, such as Cdx2 and Isx (20-fold) (Table 2).

To determine whether the reciprocal changes in marker expression resulted from re-specification of the pancreatic endoderm in the pDKO embryos, we assessed pancreatic, stomach and intestinal lineage markers in the pDKO; R26R:Tomato mice using the lineage label to track the potentially re-specified pancreatic lineages. As expected, in control E10.5 pancreatic endoderm, Tomato-labeled pancreatic lineage cells in the dorsal pancreatic anlage express the pancreatic lineage marker Pdx1, but do not express the stomach marker gene Sox2 (Fig. 2A-D′). However, in the pDKO embryos, Tomato-labeled pancreatic lineage cells begin to express Sox2 in the place of Pdx1 (Fig. 2E-H′), indicating that pancreatic lineage cells in the dorsal pancreas have switched to a stomach identity. Similarly, in the ventral pancreatic endoderm, the pancreatic lineages appear to adopt intestinal cell fates (Fig. 2I-Y). In pDKO E9.5 embryos, a small number of pancreatic lineage cells begin to co-express Pdx1 and Cdx2 or express Cdx2 in place of Pdx1 (Fig. 2M-P″). By E10.5, there is an increasing number of Pdx1 lineage-labeled cells that express Cdx2 (Fig. 2V-Y′). At E18.5, conversion of the Pdx1 lineage to mature stomach fates is widespread and many of the Tomato lineage-labeled cells have become incorporated into the stomach epithelium as previously demonstrated (Xuan, et al., 2012) where they co-express mature stomach markers, including MUC5AC and ATP4B (Fig. S2). These results suggest that the erroneous upregulation of the hedgehog pathway in pDKO embryos results in re-specification of the dorsal and ventral pancreatic lineages to the adjacent stomach and intestinal fates, respectively.

GATA regulation of Shh expression has been reported in several organ systems. During gastric development, Gata4 and Shh expression is mutually exclusive, suggesting that GATA4 might inhibit the expression of Shh (Jacobsen et al., 2005). In the limb bud, GATA6 was shown to inhibit Shh gene expression through a limb bud-specific enhancer of the Shh gene (Kozhemyakina et al., 2014), and in the somitic system, GATA4/5/6 were found to inhibit Shh signaling by inhibiting Gli1 expression, although direct inhibition of Shh was not required (Daoud et al., 2014). The upregulation of hedgehog pathway expression in the pDKO mice suggested that Shh might be repressed by GATA4 and/or GATA6. An 806-bp MACS1 enhancer located ∼740 kb upstream from the Shh transcriptional start site has been shown to be sufficient for driving Shh expression specifically in the foregut endoderm (Anderson et al., 2014; Sagai et al., 2009). Using position weight matrix analysis we identified four putative GATA consensus elements within the MACS1 enhancer (Fig. 3A, red text). Chromatin immunoprecipitation (ChIP) analysis of these sites revealed strong binding of GATA4 and GATA6 to the two most upstream GATA consensus sites (Gata site 1 and Gata site 2) and minimal, if any, binding to the 3′ sites (Gata site 3 and Gata site 4) (Fig. 3B,C).

To determine whether GATA4 and GATA6 could inhibit Shh expression through the MACS1 enhancer element, we performed luciferase experiments in a Shh-expressing pancreatic αTC6 cell line that lacks endogenous Gata4 and Gata6 (Fig. S3A). Transfection of pGL4.27-MACS1 led to an approximately sixfold increase of luciferase activity compared with pGL4.27 vector alone, suggesting that MACS1 has high activity in these cells (Fig. 3C), which is consistent with previous in vivo expression analysis (Kawahira et al., 2005). Furthermore, co-transfection with either Gata4 or Gata6 reduced MACS1 activity by >50%, with GATA6 having the largest effect (∼75%). Furthermore, simultaneous transfection of half the amount each of Gata4 and Gata6, such that the total amount of GATA protein remained constant (Fig. S3B,C) demonstrated that GATA4 and GATA6 could function together to repress MACS1 activity (Fig. 3D). Surprisingly, however, overexpression of Gata4 and/or Gata6 was still able to partially repress a MACS1 fragment that contained mutations in the four putative Gata consensus sites (Fig. 3A, blue text; Fig. 3D, ‘MACS1mut’). This might indicate that overexpression of GATA factors can bind and activate through cryptic Gata sites (Fig. 3A, green text) present within the MACS1 enhancer or that GATA factors mediate MACS1 repression through an indirect mechanism, as previously described in the heart (Rivera-Feliciano et al., 2006).

Repression of the hedgehog pathway in the presumptive ventral and dorsal pancreatic endoderm in response to signals from adjacent tissues has been well documented (Apelqvist et al., 1997; Chung and Stainier, 2008; diIorio et al., 2007; Haumaitre et al., 2005; Hebrok et al., 1998; Roy et al., 2001). However, the transcriptional regulators that mediate these mesodermal signals upstream of the hedgehog pathway have not yet been identified. In this study, we demonstrate the importance of GATA4 and GATA6 in repressing hedgehog pathway components within the pancreatic anlage. Furthermore, our in vitro data suggest GATA-mediated repression of Shh occurs partially through the endoderm-specific MACS1 enhancer. Consistent with a crucial role for GATA4 and GATA6 in repressing hedgehog signaling in the pancreatic endoderm, there is a conversion of pancreatic lineages to stomach or intestinal cell fates when GATA function is absent. The timing of Shh activation relative to the observed fate changes supports a model in which GATA-mediated repression of hedgehog signaling is necessary to delineate and maintain foregut endoderm fates (Fig. 3E). These findings also reinforce the proposed role for GATA4 and GATA6 as pioneer factors that are positioned at the top of the gene regulatory cascade that patterns tissue-specific gene expression pathways (Zaret et al., 2008).

Recently, it has been reported that haploinsufficiency of GATA4 or GATA6 genes account for more than half of all human pancreatic agenesis cases (Lango Allen et al., 2012), suggesting their important role in human pancreas development. Our discovery that GATA4 and GATA6 are essential for maintaining repression of hedgehog signaling provides important implications for the pathways regulated by the GATA factors during human pancreas development and could lead to novel strategies to detect and/or prevent pancreatic agenesis before birth.

Fluorescently labeled cells from E12.5 control and pDKO pancreata were manually dissected using a Leica MZ16F fluorescence dissecting microscope. Six controls and 12 pDKO pancreata were pooled for RNA-Seq (n=1 pools per genotype). RNA was generated (RNeasy Micro kit; Qiagen) and tested for quality (RIN values >8; Agilent Bioanalyzer 2100). RNA-Seq was performed on the Illumina HISEQ 2000V3 Instrument (Columbia Genome Center) at a depth of 25-30 million 100-bp single-end reads. FPKM values were used to measure RNA expression level and 23,700 genes were compared between control and mutant samples. DEseq analysis (DESeq2, R software package) was performed to identify differentially expressed genes (P<0.0008). RNA-Seq data have been deposited in Gene Expression Omnibus under accession number GSE77083.

RNA from E12.5 pancreata was isolated (RNeasy Micro, Qiagen) to generate cDNA (Invitrogen). qRT-PCR was performed using Taqman AOD probe sets or SYBR green Premix (Bio-Rad). Taqman AODs: Pdx1, Mm00435565-ml; Ptf1a, Mm04203788_gl; Gata4, Mm00484689_ml; Gata6, Mm00802636_ml; Cdx2, Mm01212280_m1; Sox2, Mm03053820_s1 (Open Biosystems-Thermo Scientific). Specific primers: Shh ex1 F, 5′-GGAGCAGACCGGCTGATGAC-3′; Shh ex2 R, 5′-TCGGTCACTCGCAGCTTCAC-3′; Ihh ex1 F, 5′-TCTTCAAGGACGAGGAGAACACG 3′; Ihh ex2 R, 5′-CACCCGCAGTTTCACACCAG-3′; Gli1 F, 5′TGGTACCATGAGCCCTTCTT 3′; Gli1 R, 5′-GTGGTACACAGGGCTGGACT-3′; Ptch1 F, 5′ATCTCGAGACCAACGTGGAG 3′; Ptch1 R, 5′-GCCTCTTCTCCTATCTTCTGACG-3′; Gata4 F, 5′-TAGTCTGGCAGTTGGCACAG-3′; Gata4 R, 5′-ACGGGACACTACCTGTGCAA-3′; Gata6 F, 5′-AGTTTTCCGGCAGAGCAGTA-3′; Gata6 R, 5′-AGTCAAGGCATCCACTGTC-3′.

RNA in situ hybridization was performed as previously described (Prado et al., 2004). Shh antisense riboprobe was prepared from a pBSK-Shh plasmid containing a full-length Shh cDNA. T3 RNA polymerase was used to transcribe a HindIII-linearized plasmid. Brightfield images were acquired using a Leica DM5500 microscope.

All immunofluorescence analysis was performed on frozen sections as previously described (Xuan et al., 2012). Primary antibodies were: rabbit anti-Pdx1 (1:1000; 07-696, Millipore), mouse anti-Sox2 (1:250; MAB4343, Santa Cruz), mouse anti-Cdx2 (1:80; CDX2-88, BioGenex), mouse anti-Muc5AC (1:500; ab3649, Abcam) and rabbit anti-ATP4b (1:1000; MA3-923, Thermo Fisher Scientific). Secondary antibodies were Alexa Fluor 488-, 594- or 697- conjugates (1:500; Jackson ImmunoResearch). Fluorescence and confocal images were acquired using a Zeiss LSM 710 confocal microscope. Images and z-stack images were analyzed using ZEN software (Zeiss).

Cell lysates from a-TC cells were analyzed using mouse anti-Gata4 (1:1000; SC-25310, Santa Cruz), rabbit anti-Gata6 (1:200; SC-9055, Santa Cruz ) and rabbit anti-Gapdh (1:1000; ab9845, Abcam). Quantification of the relative intensity was performed using ImageJ software.

E14.5 pancreata were manually dissected. Chromatin was prepared from these tissues as reported previously (Xuan et al., 2012). The following primers were used for PCR analysis: MACS1 5′ primers: forward 5′-GTGTACAGAAAGCCTAGTGTGTC-3′, reverse 5′-GCAGCAATAAAAGACTATGCCTC-3′; MACS1 3′ primers: forward 5′-ACATTCTGTTGCTAACTTGAAGTG-3′, reverse 5′-AAGCCTGGAATTTATAGCATCTCA-3′; Gata site 2 primers: forward 5′-GAGGCATAGTCTTTTATTGCTGC-3′, reverse 5′-GCATCTGACGGATTGTTAGCCT-3′; Gata site 3 primers: forward 5′-ATGGTCCTGAAGTTTGTTCATCC-3′, reverse 5′-CACTTCAAGTTAGCAACAGAATGT-3′; CPA1exF, CGGAGCTAGTAGCAACCCCT; CPA1exR, CAGGAGCTGGTTCTGATGTG.

The 806-bp MACS1 genomic fragment was PCR amplified (5′ GGTACCTTGTACTGGTGAGTGT, 3′ AGATCTATAACCAGGAGTCCAGG) and cloned into pGL4.27 luciferase reporter plasmid. A DNA fragment containing the 806-bp MACS1mut with point mutations in the four putative Gata sites was synthesized by Integrated DNA Technologies. Full-length cDNAs of mouse Gata4 and Gata6 were cloned into the pcDNA3 vector. These plasmids were transfected along with Renilla luciferase construct into α-TC6 cells using X-treme (Roche) transfection reagent according to the manufacturer's instructions. Forty-eight hours after transfection, cell lysates were collected and luciferase samples were prepared using the Dual Luciferase Reporter Assay System (Promega). Luciferase activities were measured using an Orion II Luminometer and luciferase activity was normalized to Renilla activity.

ImageJ software was used to quantify the band intensities in the images for PCR gel and western blotting.

All values are expressed as mean±s.e.m. Statistical analysis was performed using a two-tailed Student's t-test. Results were considered significant at P<0.05.

We would like to thank Ms Irene Yu for assistance with the model illustration and Ms Jayne Martin (Leibel lab, Columbia University) for providing hedgehog real-time PCR primers. We also thank Matthew Borok and the other members of the Sussel lab for helpful discussions and critical reading of the manuscript. We thank Ms Katie Sinagoga and Dr James Wells for their assistance with whole-mount immunofluorescent staining and 3D imaging of the gut tube, which was not included in the final manuscript.