Transcription factor GATA6: a novel marker and putative inducer of ductal metaplasia in biliary atresia

Abstract

Biliary atresia (BA), a neonatal liver disease, is characterized by obstruction of extrahepatic bile ducts with subsequent cholestasis, inflammation, and progressive liver fibrosis. To gain insights into the pathophysiology of BA, we focused attention on GATA6, a transcription factor implicated in biliary development. Early in fetal development GATA6 expression is evident in cholangiocytes and hepatocytes, but by late gestation it is extinguished in hepatocytes. Utilizing a unique set of BA liver samples collected before and after successful portoenterostomy (PE), we found that GATA6 expression is markedly upregulated in hepatocytes of patients with BA compared with healthy and cholestatic disease controls. This upregulation is recapitulated in two murine models simulating bile duct obstruction and intrahepatic bile ductule expansion. GATA6 expression in BA livers correlates with two established negative prognostic indicators (age at PE, degree of intrahepatic bile ductule expansion) and decreases after normalization of serum bilirubin by PE. GATA6 expression in BA livers correlates with expression of known regulators of cholangiocyte differentiation (JAGGED1, HNF1β, and HNF6). These same genes are upregulated after enforced expression of GATA6 in human hepatocyte cell models. In conclusion, GATA6 is a novel marker and a putative driver of hepatocyte-cholangiocyte metaplasia in BA, and its expression in hepatocytes is downregulated after successful PE.

NEW & NOTEWORTHY A pathological hallmark in the liver of patients with biliary atresia is ductular reaction, an expansion of new bile ductules that are thought to arise from conversion of mature hepatocytes. Here, we show that transcription factor GATA6 is a marker and potential driver of hepatocyte ductal metaplasia in biliary atresia. Hepatocyte GATA6 expression is elevated in biliary atresia, correlates with bile duct expansion, and decreases after successful portoenterostomy.

INTRODUCTION

Biliary atresia (BA) is a rare neonatal condition in which the extrahepatic bile ducts are obstructed by a sclerosing, inflammatory process (13). Typical histological findings include cholestasis, inflammation, and fibrosis, rapidly leading to cirrhosis if untreated (13). The characteristic pathological process underlying these changes is expansion of intrahepatic bile ductules, also called the ductular reaction (DR), different forms of which are seen in a wide range of hepatobiliary diseases (11). The cellular origin of the newly formed bile ductules in BA livers is not completely understood (11); some of the neoductules arise from proliferating cholangiocytes, but hepatocytes also participate in this process via ductal metaplasia (19, 22, 38).

A mainstay of therapy for BA is portoenterostomy (PE). This procedure can reverse cholestasis and reduce the DR in patients with BA. Patients who undergo PE at a younger age have improved odds for normalization of hyperbilirubinemia and a better prognosis (35, 36). Unfortunately, despite clearance of jaundice, the majority of patients who undergo PE eventually require liver transplantation because of ongoing liver injury (20). Extensive liver fibrosis and biliary proliferation in the native liver at the time of PE portend a poor outcome (30, 34).

The etiology of BA remains largely unknown, although infectious agents, immunological factors, and toxins have been implicated in its pathogenesis (13). There is no strong evidence for familial inheritance in BA, but genetic modifiers are thought to influence the onset of the disease (13). Irrespective of its precise cause, BA is often accompanied by alterations in the molecular pathways that regulate hepatobiliary development (25).

GATA transcription factors comprise a six-member family of highly conserved zinc-finger DNA-binding proteins that control cell proliferation, differentiation, and gene expression in many organs, including the liver and extrahepatic biliary system (18, 40). Dysregulated GATA factor expression has been linked to developmental disorders and oncogenesis in several tissues (18, 32, 40, 43). In the murine liver, Gata4 and Gata6 are abundantly expressed during liver bud formation (E8–E14) and are necessary for hepatic cell fate commitment (26). After the early fetal period, expression of Gata6 diminishes in liver and is mostly restricted to nonhepatocyte cell population although small amounts of Gata6 RNA can be detected in isolated murine hepatocytes (26, 44). There are no detailed reports about GATA6 expression in postnatal human liver although GATA6 heterozygous mutations are associated with hepatobiliary malformations (1, 8). Altered expression of GATA6 in other tissues has been reported in a variety of human disease states, including gastrointestinal and pancreatobiliary metaplasias (14, 16).

Here, we characterize the hepatic expression of GATA6 during normal human liver development, in patients with BA and in two mouse models of DR induced by extrahepatic obstruction or intrahepatic bile duct insufficiency. Our analysis includes paired liver specimens obtained at the time of PE and during follow-up. We also assess the in vitro effects of enforced expression of GATA6 in human primary hepatocytes and a hepatocellular carcinoma cell line. Our findings suggest a role for GATA6 in ductal metaplasia of hepatocytes in BA.

MATERIALS AND METHODS

Ethical statement.

This study was approved by the Ethical Committee of Helsinki University Hospital and by the Finnish National Authority of Medicolegal Affairs and Health.

Patient samples.

BA patient samples obtained at the time of PE (BA samples) and follow-up biopsies (BA-post-PE samples) were obtained from Helsinki University Hospital. Patients’ ages at the time of PE ranged from 7 to 141 days (median, 43 days). Patients of BA-post-PE samples underwent a successful PE with clearance of jaundice, i.e., their serum bilirubin was normalized (<20 µmol/l) after PE. Ultrasound-guided core-needle follow-up biopsies were taken 302–6,887 days after PE (median, 1,596 days). The subtype of BA in the samples was as follows: type 3 = 88%, type 2 = 10%, type 1 = 2%. The BA and BA-post-PE samples with laboratory values and other clinical information are listed in Data Supplement 1A (supplemental material for this article is available online at the American Journal of Physiology-Gastrointestinal and Liver Physiology website).

Healthy liver transplantation donor samples were used as normal adult liver (AL) controls (aged, 1–62 yr). Normal fetal liver (FL) samples were parsed into two groups: early FL [aged gestational weeks (GW) 12–13] and late FL (aged, GW 32–40). Disease control (DC) samples were derived from pediatric patients with cholestatic conditions other than BA (aged, 32–2,498 days, median 59 days; Data Supplement 1B). The number of samples analyzed was as follows: paraffin-embedded tissue: BA n = 31, BA-post-PE n = 36, AL n = 30, FL n = 4, DC n = 24; frozen fresh tissue: BA n = 8, BA-post-PE n = 13, AL n = 3, and DC n = 6. For the qRT-PCR analysis, we also used one commercially available human total AL whole tissue lysate (Novus Biologicals, Littleton, CO), two human total AL RNA (aged, 21 and 60 yr), and three FL RNA (pooled from multiple donors aged GW 18–40) specimens (Agilent Technologies, Santa Clara, CA; Ambion, Foster City, CA; Clontech, Mountain View, CA; Stratagene, La Jolla, CA) as controls.

RNA in situ hybridization.

RNA in situ hybridization was performed on 5-μm formalin-fixed paraffin embedded (FFPE) tissue sections using RNAscope 2.5 HD detection kit-RED (no. 322350; ACDBio, Milan, Italy) for chromogenic target mRNA detection and RNAscope Multiplex Fluorescent Reagent Kit Version 2 (no. 323100, ACDBio) for fluorescent target detection. Sample pretreatment and probe hybridization conditions were the same for both detection methods. First, tissue sections were baked for 1 h at 60°C and then deparaffinized and treated with hydrogen peroxide for 10 min at room temperature. Target retrieval was performed for 15 min at 100°C, followed by protease plus treatment for 15 min at 40°C. The probes Hs-JAG1 (no. 546181), Hs-HNF6/ONECUT1 (no. 490081), Hs-HNF1b (no. 490071), Hs-PPIB (no. 313901, positive control), and negative control probe DapB (no. 310043) were hybridized for 2 h at 40°C followed by signal amplification. For chromogenic detection, the samples were incubated for 45 min with AMP 5-RED reagent. The samples were then treated with fast red for 10 min at room temperature followed by counterstaining with 50% hematoxylin. The sections were dipped in ammonium water and dehydrated for 15 min at 60°C before mounting. For fluorescent detection, Hs-GATA6-C1 (no. 603131) probe was cohybridized with either Hs-KRT7-C3 (for CK7, no. 550151-C3) or Hs-CFTR-C2 (no. 603291-C2) for 2 h at 40°C. RNAscope 3-plex Positive Control Probe (no. 320861) and RNAscope 3-plex Negative Control Probe (no. 320871) were used as controls. Signal amplification and development for horseradish peroxidase channels were performed according to the manual. TSA Plus Cyanine 3 fluorophore (NEL744001KT; Perkin Elmer, Waltham, MA) was used for Hs-GATA6 probe at 1:1,500 dilution, and TSA Plus Cyanine 5 fluorophore (NEL745001KT) was used for Hs-CFTR and Hs-KRT7 probes at 1:1,500 dilution. The sections were counterstained with DAPI.

Immunohistochemistry.

FFPE liver sections were deparaffinized, hydrated, and treated with citrate buffer (Target retrieval solution; Dako, Glostrup, Denmark) in 97°C for 30 min. Endogenous peroxidase activity was blocked with 3% hydrogen peroxide, and nonspecific binding was prevented by using 1.5% normal serum. Immunoperoxidase staining was performed using a polymerized reporter enzyme staining system (ImmPRESS reagent kit; Vector, Burlingame, CA or Novolink Polymer Detection System; Leica Biosystems, Newcastle, UK) to visualize the bound antibody. Primary antibodies were incubated overnight at 4°C. In control experiments, nonimmune serum replaced the primary antibody. Bile duct epithelial cell nuclei served as an internal positive control for GATA6 in all liver samples. Primary antibodies used were as follows: rabbit antihuman GATA6 IgG (sc-9055; Santa Cruz Biotechnologies, Santa Cruz, CA) at dilution 1:1,200 for human samples, goat anti-human GATA6 IgG (AF1700; R&D Systems, Minneapolis, MN) at dilution 1:1,200 for mouse samples. Immunostaining for cytokeratin 7 (CK7) was performed using a rabbit SP52 anti-human monoclonal antibody and ultraView Universal DAB Detection Kit (Ventana Medical Systems, Tucson, AZ). Images were collected using 3DHISTECH Panoramic 250 FLASH II digital slide scanner at Genome Biology Unit (Research Programs Unit, Faculty of Medicine, University of Helsinki Biocenter, Helsinki, Finland).

RNA extraction and quantitative real-time PCR.

Total RNA was extracted from fresh tissue samples using the RNeasy Mini Kit (Qiagen, Hilden, Germany) or from cell culture lysates using Nucleospin RNA/Protein Kit (Macherey-Nagel, Düren, Germany). Reverse-transcription reaction was performed by Reverse Transcriptase Core Kit and qRT-PCR by MESA GREEN qPCR MasterMix Plus for SYBR Assay (both from Eurogentec, Seraing, Belgium) according to manufacturer’s instructions. PPIG and GAPDH served as reference genes. Primer pairs used are listed in Table 1.

Table 1.

Quantitative real-time PCR primer sequences

Gene	Reference	Oligonucleotide Sequence 5′->3′
AFP	NM_001134.2	F: GAGGGAGCGGCTGACATTAT
		R: CATGGCCTCCTGTTGGCATA
ALB	NM_000477.6	F: GCTGTCATCTCTTGTGGGCTGT
		R: AAACTCATGGGAGCTGCTGGTT
BMP4	NM_001202.4	F: CTGCGGGACTTCGAGGCGACACTTCT
		R: TCTTCCTCCTCCTCCTCCCCAGACTG
CK19	NM_002276.4	F: CTGCGGGACAAGATTCTTGGT
		R: CCAGACGGGCATTGTCGAT
CTNNB1	NM_001098209.1	F: ATTGAAGCTGAGGGAGCCAC
		R: TGCATATGTCGCCACACCTT
PPIG	NM_004792.2	F: CAATGGCCAACAGAGGGAAG
		R: CCAAAAACAACATGATGCCCA
DKK1	NM_012242.2	F: TCACACCAAAGGACAAGAAGG
		R: ATCTTGGACCAGAAGTGTCTAGC
GAPDH	NM_001256799.2	F: GGTCATCCATGACAACTTTGG
		R: CCATCCACAGTCTTCTGGGT
GATA6	NM_005257.5	F: GTGCCCAGACCACTTGCTAT
		R: TGGAATTATTGCTATTACCAGAGC
Gata6 (mouse)	NM_010258.3	F: ATTCACCAGCAGCGACTACG
		R: TTGATTCCTCGAGCGATGTG
HNF1β	NM_000458.3	F: GGCCTACGACCGGCAAAAGA
		R: GGGAGACCCCTCGTTGCAAA
HNF4α	NM_000457.4	F: TGTCCCGACAGATCACCTC
		R: CACTCAACGAGAACCAGCAG
HNF6	NM_004498.2	F: TGTGGAAGTGGCTGCAGGA
		R: TGTGAAGACAACCTGGGCT
JAG1	NM_000214.2	F: TGCCCTCCAGGACATAGTGG
		R: ACTCTCCCCATGGTGATGCA
NOTCH2	NM_001200001.1	F: GGCATTAATCGCTACAGTTGTGTC
		R: GGAGGCACACTCATCAATGTC
SMAD5	NM_001001419.2	F: CTATGTTGGAGAGGTGTATG
		R: CAGCACGTGGTGGGATGAAA
SOX9	NM_000346.3	F: GTACCCGCACTTGCACAAC
		R: TCTCGCTCTCGTTCAGAAGTC
TAT	NM_000353.2	F: GGGGACCCTACTGTGTTTGG
		R: ACTGGATAGGAAGCCGATGG

Protein extraction and Western blotting.

Total protein was extracted from fresh tissue samples or from cell culture lysates using Nucleospin RNA/Protein Kit (Macherey-Nagel, Düren, Germany). Protein (15 µg) was separated by electrophoresis using Mini-Protean TGX Stain-Free gels (Bio-Rad, Hercules, CA) and transferred onto a polyvinylidene fluoride membrane. Nonspecific binding was blocked with 5% nonfat milk in 0.1% Tris-buffered Tween-saline buffer for 1 h. Primary antibody against GATA6 (rabbit anti-human GATA6 IgG, AF1700; R&D Systems) was incubated for 1 h at 37°C. After incubation with secondary antibody for 1 h at room temperature, protein bands were visualized by the Enhanced Chemiluminescence detection kit (Amersham ECL reagent; GE Healthcare, Barrington, IL). Normalization and quantification of the protein band intensity were carried out using Image Laboratory 6.0 software. The intensity of the GATA6 band was normalized to total protein amount of each lane.

Experimental mice.

All mice were housed under barrier conditions. All procedures were approved by the Institutional Animal Care and Use Committee at Vanderbilt University Medical Center or Cincinnati Children’s Hospital Medical Center. Adult 4- to 6-wk-old female mice on a mixed genetic background (C57Bl/6-6-SJL-Swiss Black) were subjected to bile duct ligation (BDL) and allowed to recover for 14 days. Control mice for BDL mice were littermates subjected to a sham surgery. Surgeries were performed as previously described (3). Alb-Cre;Rbpj^flox/flox;Hnf6^flox/flox mice on a mixed genetic background (C57BI/6 and CD1-ICR) were previously reported (39, 42). Control mice were Alb-Cre negative Rbpj^flox/flox;Hnf6^flox/flox mice.

Primary hepatocytes and HepG2 cell line.

Commercially available cryopreserved human primary hepatocytes (Lonza, Basel, Switzerland) were derived from a 62-yr-old Caucasian female with a nonhepatic cause of death. Primary hepatocytes were cultured in hepatocyte-specific plating and maintenance media (MP250 and MM250, Lonza). An established human hepatocellular carcinoma cell line, HepG2, was cultured at 37°C in Dulbecco's Modified Eagle's Medium (DMEM) (Lonza) supplemented with 10% fetal calf serum, 2 mM l-glutamine, 100 U/ml penicillin, and 100 µg/ml streptomycin sulfate.

Plasmid-mediated transfections.

Human primary hepatocytes and HepG2 cells were transfected with a pcDNA3-GATA6 expression plasmid (23) or corresponding control plasmid. jetPEIHepatocyte DNA transfection reagent (Polyplus Transfection, Illkirch, France) was used for human primary hepatocytes and Lipofectamine LTX with Plus Reagent (Thermo Fischer Scientific, Waltham, MA) for HepG2 cells. Cells were incubated 48 h after transfection and subsequently analyzed for gene expression.

Statistical analysis of the data.

The differences in mRNA expression levels of different sample groups were statistically analyzed with nonparametric Wilcoxon test for each pair of sample groups. Student’s t-test was used to analyze the expression alterations in matched pairs of samples. The clinical parameters were compared with the immunohistochemistry data using nonparametric Wilcoxon test. Linear regression analysis was used for correlation of mRNA expression of different genes in the same samples. The statistical analyses were carried out with JMP software (JMP, Cary, NC). All in vitro experiments were performed in triplicate and statistically analyzed using one-way ANOVA, followed with comparison to control using Student’s t-test. P < 0.05 was considered significant in all experiments.

RESULTS

GATA6 expression during normal liver development in humans.

Immunohistochemistry demonstrated prominent GATA6 staining in all cell types of human FL at GW 13 (Fig. 1A). GATA6 immunoreactivity decreased at later stages of liver development; by GW 37, GATA6 protein was limited to bile duct epithelial cells, inflammatory cells, arterial smooth muscle cells, and venous endothelial cells (Fig. 1B). This restricted cellular distribution persisted into adulthood (Fig. 1C). The expression pattern of GATA6 in fetal (GW 37) (Fig. 1, D and E) and adult (Fig. 1, F and G) liver samples was verified with RNA in situ hybridization, which demonstrated strong signal in biliary epithelial cells, cholangiocytes, and low or absent signal in normal hepatocytes.

Fig. 1.

GATA6 is expressed in human hepatocytes during early gestation but not in normal perinatal or adult hepatocytes. A–C: immunohistochemical staining of GATA6 protein in normal human liver. Brown indicates positive staining in nuclei. In fetal liver (FL) at gestational week (GW) 13 (FL GW13), GATA6 protein is expressed in all hepatocytes (arrowhead) and other cell types of the liver (A). At GW37, GATA6 protein expression is diminished in hepatocytes (arrowhead) and restricted to cholangiocytes lining the bile ducts (arrow) (B). The expression pattern of GATA6 protein in adult liver (AL) is similar to that of late FL with positive cholangiocytes (arrow) and negative hepatocytes (arrowhead) (C). D–G: in situ hybridization of GATA6 mRNA in FL GW37 (D and E) and AL (F and G). Green indicates positive signal in cytoplasm. Cholangiocytes show positive signal (arrow); hepatocytes show weak or negative signal (arrowhead). Scale bar = 200 µm (A–C) and 20 µm (D–G).

GATA6 expression is elevated in BA.

Quantitative RT-PCR analysis revealed increased GATA6 mRNA in BA samples taken at the time of PE compared with normal FL (GW 37) or AL (Fig. 2A). Specifically, there was a 13.2-fold higher median expression level of GATA6 mRNA in BA than in AL samples (P = 0.009). GATA6 expression was also significantly higher in BA (median, 2.6-fold; P = 0.024) than in cholestatic DC samples. The statistical comparison of GATA6 mRNA levels between different patient groups is shown in Table 2. In BA samples, GATA6 mRNA expression localized to both hepatocytes and bile duct epithelium, as documented by in situ hybridization (Fig. 2, B and C). Immunohistochemistry confirmed abundant GATA6 protein expression in hepatocytes of BA liver samples (Fig. 2G, Table 3); GATA6 expression was especially high in the periportal areas where DR was present. GATA6 protein was detected in hepatocytes of some DC livers, albeit at a lower frequency and intensity than in BA samples (Table 3).

Fig. 2.

Hepatocyte expression of GATA6 is elevated in biliary atresia (BA) and decreased after portoenterostomy (PE). A: boxplot showing the relative GATA6 mRNA expression, as determined by qPCR, in different liver sample groups. Dots represent individual samples, the box represents the interquartile range, and the whiskers represent the 1st and 4th quartile. The line inside the box is the median, and the dashed line represents the mean. *P < 0.05, **P < 0.01 compared with BA group. The P values of other paired comparisons are indicated in Table 2. B–E: in situ hybridization of GATA6 in BA samples demonstrates strongly positive hepatocytes (arrowhead) (B and C), whereas in situ hybridization in BA-post-PE samples shows weak or negligible signal in hepatocytes (arrowhead) (D and E). F–H: GATA6 immunohistochemistry from normal adult liver (AL) with negative hepatocytes and positive cholangiocytes (F), BA with strong immunoreactivity in hepatocyte nuclei (G), and BA-post-PE with less immunoreactivity in hepatocytes compared with BA (H). I: Western blotting of GATA6 protein in AL, BA, BA-post-PE, and disease control (DC) samples. NBI, normalized band intensity. J: paired sample analysis of GATA6 protein expression from patients before BA and after PE (BA-post-PE). Green (in situ hybridization) and brown (immunohistochemistry) indicate positive staining. Scale bars = 150 µm (B and C), 20 µm (C and E), and 200 µm (F and H).

Table 2.

P values from comparisons of mRNA levels of different liver sample types (in Figs. 2A and 7A, C, and E) using Wilcoxon method

	FL	AL	BA	BA-post-PE
GATA6
AL	0.377	—	—	—
BA	0.019	0.009	—	—
BA-post-PE	0.180	0.068	0.001	—
DC	0.028	0.014	0.024	0.100
HNF1β
AL	0.216	—	—	—
BA	0.032	0.009	—	—
BA-post-PE	0.861	0.148	0.008	—
DC	0.156	0.025	0.561	0.085
HNF6
AL	0.377	—	—	—
BA	0.019	0.009	—	—
BA-post-PE	0.924	0.114	0.0002	—
DC	0.093	0.043	0.561	0.020
JAG1
AL	0.052	—	—	—
BA	0.019	0.009	—	—
BA-post-PE	0.097	0.007	0.043	—
DC	0.093	0.014	0.061	1.000

FL, fetal liver; AL,  adult liver; BA, biliary atresia; BA-post-PE, BA postportoenterostomy; DC, disease control.

Table 3.

GATA6 immunoreactivity in liver sample groups

GATA6 expression	AL, % (n)	BA, % (n)	BA-post-PE, % (n)	DC, % (n)
Low	84 (25)	0 (0)	0 (0)	13 (3)
Low intermediate	13 (4)	7 (2)	24 (7)	25 (6)
High intermediate	3 (1)	32 (10)	52 (15)	17 (4)
High	0 (0)	61 (19)	24 (7)	45 (11)
Total	100 (30)	100 (31)	100 (29)	100 (24)

Percentage of GATA6-positive nuclei in immunohistochemistry. Low <30%, low intermediate 30–50%, high intermediate 50–70%, high >70%. AL, adult liver; BA, biliary atresia; BA-post-PE, BA postportoenterostomy; DC, disease control.

GATA6 expression decreases in hepatocytes after successful PE.

Following PE, there was a significant decrease of liver GATA6 expression both at the mRNA (Fig. 2, A–E) and protein levels (Fig. 2, G–I, and Table 3). In the BA-post-PE samples, GATA6 mRNA and protein expression levels decreased specifically in the hepatocyte cell population, as shown by RNA in situ hybridization and immunohistochemistry, respectively (Fig. 2, B–E and 2, G and H). Western blotting demonstrated abundant GATA6 protein expression in BA livers but low or negligible levels in the BA-post-PE, AL, and DC samples (Fig. 2I). In a blinded immunohistochemical analysis of 20 patient sample pairs before (BA) and after (BA-post-PE) PE, GATA6 expression (scored as described in Table 3; 0 = low, 1 = low intermediate, 2 = high intermediate, 3 = high) decreased in 10 patients, remained unaltered in 9 patients, and increased in 1 patient (Fig. 2J). Statistical analysis of the paired samples confirmed a significant decrease in GATA6 immunoreactivity (P < 0.005) following PE. It is unclear why GATA6 immunoreactivity was unaltered in some of the paired specimens; age at PE, length of follow-up time, or native liver survival did not correlate to the change in GATA6 level.

GATA6 is expressed in BA hepatocytes lacking the cholangiocyte markers CFTR and CK7.

To confirm that the elevation in GATA6 mRNA levels in BA samples is not solely due to intrahepatic bile ductule expansion, we performed a double fluorescence in situ hybridization for GATA6 and two markers of differentiated cholangiocytes [cystic fibrosis transmembrane conductance regulator (CFTR) and cytokeratin-7 (CK7)] (Fig. 3, A–J). Both GATA6 and CFTR were detected in differentiated bile ducts (Fig. 3, F and H), but GATA6 was expressed in hepatocytes negative for CFTR (Fig. 3, F and J). CK7 was found in some periportal hepatocytes undergoing ductal metaplasia. GATA6 was highly expressed in these CK7-positive cells but was also found in adjacent hepatocytes negative for CK7 (Fig. 3, E and J). Altogether, these experiments demonstrated the presence of a GATA6-positive but CK7/CFTR-negative hepatocyte population adjacent to DR.

Fig. 3.

Double in situ hybridization in biliary atresia (BA) liver. A and B: double in situ hybridization was performed on 2 different samples of BA liver. GATA6 expression is high in both ductular reaction (DR) area and liver parenchyma (hepatocytes); these 2 histological compartments are denoted in the figures by dashed lines. C and D: CK7 (C) is strongly expressed in DR area, whereas its expression is weaker in hepatocytes. CFTR (D) expression is limited to DR area. E and F: merged images of DAPI, GATA6, and CK7 (E) or DAPI, GATA6, and CFTR (F). Bile duct epithelium (arrow) is positive for both GATA6 and CK7/CFTR, but hepatocytes (arrowhead) express only GATA6. G and I: higher-magnification images from E. H and J: higher-magnification images from F. Green indicates positive signal for GATA6, and red indicates positive signal for CK7 or CFTR. Blue indicates DAPI staining in the nuclei of all cell types. Scale bars = 50 µm (A–F) and 10 µm (G–J).

Increased GATA6 expression in murine hepatocytes following BDL.

We next assessed GATA6 expression in murine livers as a consequence of BDL. Consistent with published reports (10), BDL murine livers showed histologically visible DR. Using immunohistochemistry, we found GATA6 protein expression in DR cells and periportal hepatocytes (Fig. 4B). In sham-operated mice, the GATA6 immunoreactivity pattern was similar to that of normal human liver, localizing to bile duct epithelium and other nonhepatocyte cell types (Fig. 4A).

Fig. 4.

GATA6 is overexpressed in hepatocytes in 2 mouse models of biliary obstruction. A: GATA6 protein is expressed in bile duct epithelium (arrow) but not in hepatocytes (arrowhead) of normal murine liver (4 wk after sham surgery). B: in mice subjected to bile duct ligation (BDL), hepatocytes (arrowhead) in the periportal area strongly express GATA6. C and D: at postnatal day 15 (P15), before the onset of ductular reaction (DR), there remains weak expression of GATA6 in normal hepatocytes (arrowhead), and there is no difference in GATA6 expression between control (Ctrl) and Alb-Cre;Rbpj^flox/flox;Hnf6^flox/flox double knockout (DKO) mice. E and F: at P30 Ctrl hepatocytes are GATA6 negative, whereas the Alb-Cre;Rbpj^flox/flox;Hnf6^flox/flox liver with severe DR shows strong GATA6 immunoreactivity in hepatocytes. G and H: at P120 in Alb-Cre;Rbpj^flox/flox;Hnf6^flox/flox liver, the DR has diminished, and liver histology, as well as GATA6 expression, is similar to Ctrl liver with immunoreactivity only in bile duct epithelial cells (arrow). n = 3 in each group. Scale bars = 50 µm.

Gata6 expression in Alb-Cre;Hnf6^flox/floxRbpj^flox/flox double knockout mice.

As another animal model for DR, we used Alb-Cre;Hnf6^flox/floxRbpj^flox/flox mice that fail to develop peripheral bile ducts during gestation and have an abnormal intrahepatic biliary tree at postnatal day 15 (P15) (39, 42). By P30, these mice develop a severe DR as the bile ductules regenerate. At P120, bile flow is established, as newly formed intrahepatic ductules communicate with peripheral biliary tree branches. Immunohistochemical staining visualized small amounts of GATA6 protein in P15 Alb-Cre;Hnf6^flox/floxRbpj^flox/flox and control murine hepatocytes (Fig. 4, C and D). At P30, the DR areas and hepatocytes outside these areas were strongly positive for GATA6 (Fig. 4F), whereas, in P30 control livers, the hepatocyte nuclei were negative and GATA6 was expressed only in biliary epithelium (Fig. 4E). By P120, when the bile flow had normalized, GATA6 expression in hepatocytes decreased, and the expression pattern in Alb-cre;Hnf6^flox/floxRbpj^flox/flox mice (Fig. 4H) was similar to that of control mice (Fig. 4G). The expression of Gata6 mRNA in these samples paralleled the observed protein expression (data not shown).

Correlation of GATA6 expression to histopathology in BA samples.

The protein expression score of GATA6 in hepatocytes based on immunohistochemical analysis was divided into the following two groups: high  ≥70% of hepatocyte nuclei GATA6 positive; intermediate/low <70% of hepatocyte nuclei GATA6 positive. We compared the GATA6 protein expression to bile duct expansion (BDE) (high vs. low), fibrosis stage, and portal inflammation stage in BA and BA-post-PE specimens. These histological variables were evaluated by two experienced pathologists blinded to clinical data. BDE scores of 0–1 were considered low stage, and a score of 2 was high stage. In the BA group, there was a statistically significant correlation between hepatocyte GATA6 protein expression and BDE (Fig. 5A); in the BA-post-PE group, no such correlation was seen (data not shown). Fibrosis stage and portal inflammation stage did not correlate to GATA6 protein expression in either of the groups (data not shown).

Fig. 5.

In patients with biliary atresia (BA), GATA6 protein expression in hepatocytes correlates to bile duct expansion (BDE), age at portoenterostomy (PE), and the liver injury marker alanine aminotransferase. Hepatocyte GATA6 protein expression was divided into 2 groups (<70% positive nuclei = low/intermediate vs. >70% positive nuclei = high) and correlated to BDE rate (P = 0.0094) (A), age at PE (B), and plasma (P)-alanine aminotransferase (C). A: contingency tabling coupled with χ² test was employed to test the statistical significance. B and C: dots represent individual samples, the box represents the interquartile range, and the whiskers represent the 1st and 4th quartile. The line inside the box is the median, and the dashed line represents the mean. *P < 0.05.

GATA6 expression correlates to patient age at PE.

High expression of GATA6 protein in patient tissue samples correlated significantly with increased age at the time of PE (Fig. 5B). The median age of patients with low GATA6 expression was 40.5 days, whereas it was 71.5 days with high GATA6 expression. Furthermore, all patients who underwent PE at age of 70 days or older, had a high hepatocyte GATA6 expression (Fig. 5B).

Hepatocyte GATA6 expression correlates significantly with a hepatocyte injury marker in patients with BA.

GATA6 protein expression level (low/intermediate vs. high) was compared with different diagnostic peripheral blood laboratory values including bilirubin, conjugated bilirubin, alanine aminotransferase (ALT), pre-albumin, γ-glutamyltransferase, and fasting total bile acids in BA and BA-post-PE sample groups separately (laboratory values from patient samples are shown in Data Supplement 1A). In the BA group, the GATA6 expression level showed significant correlation to the hepatocyte injury marker plasma alanine aminotransferase (P-ALT) (Fig. 5C). The median P-ALT in patients with low GATA6 expression was 42 U/l, whereas it was 112 U/l in those with high GATA6 expression. All patients who had P-ALT of >200 U/l, had a high hepatocyte GATA6 expression (Fig. 5C). No significant correlations between GATA6 expression level and the other aforementioned laboratory markers were found (data not shown).

Enforced expression of GATA6 leads to increased expression of cholangiocyte lineage markers in HepG2 cells and human primary hepatocytes.

To explore whether GATA6 has a functional role in the BA hepatocyte ductular metaplasia, plasmid transfection was used to overexpress GATA6 in the human hepatocellular cell line HepG2 (Fig. 6A) and human primary hepatocytes (Fig. 6B). Two days posttransfection, we measured the expression of GATA6, genes known to promote cholangiocyte differentiation [cytokeratin 19 (CK19), SRY-box 9 (SOX9), hepatocyte nuclear factor-1β (HNF1β), HNF6, NOTCH2, JAGGED1 (JAG1), Dickkopf 1 (DKK1), β-catenin (CTNNB1), bone morphogenic protein 4 (BMP4), and SMAD family member 5 (SMAD5)], and hepatocyte lineage markers [HNF4α, albumin (ALB), α-fetoprotein (AFP), and tyrosine aminotransferase (TAT)]. HNF1β, HNF6, and JAG1 were significantly upregulated in both cell types after GATA6 overexpression (Fig. 6, A and B). Additionally, DKK1 and HNF4α were significantly upregulated in primary hepatocytes (Fig. 6B).

Fig. 6.

Enforced expression of GATA6 causes alterations in expression of genes regulating cholangiocyte and hepatocyte differentiation. The human hepatocellular cell line HepG2 and primary human hepatocytes were transiently transfected with pCDNA3-GATA6 or pCDNA3 plasmid alone. After 48 h, RNA was harvested and subjected to qRT-PCR analysis. A: in HepG2 cells, 3 genes related to cholangiocyte differentiation (HNF1β, HNF6, and JAG1) were significantly upregulated in cells overexpressing GATA6 compared with control (Ctrl) cells. B: in primary human hepatocytes, the relative expression of 4 genes related to cholangiocyte differentiation (HNF1β, HNF6, JAG1, and DKK1) and 1 gene related to hepatocyte differentiation (HNF4α) were upregulated in cells with GATA6 overexpression compared with control cells. For both cell types, 5 independent experiments were performed in triplicate. Each bar depicts the logarithm of the ratio of mRNA expression GATA6 overexpression vs. control cells. *P < 0.05, **P < 0.01.

Alterations in the expression of cholangiocyte lineage markers in BA.

We next examined the expression of genes induced by GATA6 in both cell transfection models (HNF1β, HNF6, and JAG1) in samples from patients with BA and in control liver samples (Fig. 7, A, C, and E). Each of these genes was more highly expressed in BA samples than in normal FL or AL. Moreover, the expression of each was significantly lower in BA-post-PE samples than in BA samples. Expression of each of these genes positively correlated with that of GATA6 (Fig. 7, B, D, and F). In normal liver (FL GW37), HNF1β, HNF6, and JAG1 mRNA expression was localized in bile duct epithelium (Fig. 8, A, D, and G). HNF6 was also expressed in hepatocytes of normal liver (Fig. 8D). The expression of these three genes was increased in BA samples and localized in the GATA6-positive hepatocytes in addition to bile duct epithelium (Fig. 8, B, E, and H). Like GATA6, hepatocyte expression of HNF1β, HNF6, and JAG1 was diminished after successful PE (Fig. 8, C, F, and I).

Fig. 7.

Expression of HNF1β, HNF6, and JAG1 is elevated in biliary atresia (BA) liver, decreases after successful portoenterostomy (PE), and correlates to GATA6. A, C, and E: boxplots showing relative mRNA expression of HNF1β (A), HNF6 (C), and JAG1 (E), as measured by qPCR, in BA vs. other liver specimens. Dots represent individual samples, the box represents the interquartile range, and the whiskers represent the 1st and 4th quartile. The line inside the box is the median, and the dashed line represents the mean. *P < 0.05, **P < 0.01. The P values of other paired comparisons are indicated in Table 2. B, D, and F: linear regression analyses of HNF1β (B), HNF6 (D), and JAG1 (F) mRNA expression vs. GATA6 mRNA expression in BA-post-PE samples. FL, fetal liver; AL, adult liver; DC, disease control.

Fig. 8.

Localization of HNF1β, HNF6, and JAG1 in normal fetal liver (FL), biliary atresia (BA) and postportoenterostomy (post-PE) liver samples. In situ hybridization demonstrates expression of HNF1β (A–C), HNF6 (D–F), and JAG1 (G–I) in human FL at gestational week 37 (FL GW37), BA, and BA-post-PE. An arrow demonstrates a positive cholangiocyte, and an arrowhead shows a positive hepatocyte. Purple color indicates positive staining.

DISCUSSION

A histological hallmark in the liver of patients with BA is DR (33), an excessive bile ductule expansion with inflammatory components, in which intermediate hepatobiliary cells and primitive bile ductules are formed as a consequence of bile accumulation in the liver (11). The origin of these newly formed ductular cells remains controversial, but studies suggest that they can arise from ductal metaplasia of mature hepatocytes (6, 19, 22). This type of DR is reversible when the causative trigger (e.g., extrahepatic bile duct obstruction) is eliminated.

Our findings suggest that GATA6 is a key player in the DR that accompanies BA. We found that GATA6 is expressed in all FL cell types, but later in gestation GATA6 expression is restricted to nonhepatocyte cell compartment, with strongest expression in bile duct epithelium. By analyzing a unique set of paired human tissue samples, we showed that hepatic expression of GATA6 is upregulated in patients with BA and that this expression decreases after surgical reestablishment of bile flow, coinciding with the dissolution of the DR in the operated patients. Although GATA6 is strongly expressed in bile duct epithelium, its expression pattern in BA livers differs from that of prototypical cholangiocyte markers (CFTR and CK7). In addition to cholangiocytes, GATA6 is expressed in hepatocytes undergoing ductal metaplasia. This association between the DR and Gata6 expression was recapitulated in two mouse models, surgical bile duct obstruction and the Alb-Cre;Hnf6^flox/floxRbpj^flox/flox mouse. Other mouse models of BA, such as rhesus rotavirus A-induced extrahepatic biliary obstruction, are available and could be used to confirm our findings (31).

Supporting the premise that GATA6 drives the ductal metaplasia of hepatocytes, GATA6 expression in hepatocytes correlated with histologically visible BDE stage in samples from patients with BA. Histologically limited BDE is associated with a more favorable prognosis (34). It is also known that the patients with BA who are diagnosed early and undergo PE at a young age survive longer without liver transplantation (35). In this study, patient age at the time of PE also correlated with GATA6 expression level, suggesting that GATA6 increases in hepatocytes, as biliary obstruction prolongs with advancing age. In keeping with these findings, there was also a correlation between GATA6 expression and P-ALT concentration, a marker of hepatocyte injury.

Using cell transfection models, we demonstrated that enforced expression of GATA6 in HepG2 cells or primary human hepatocytes induces a switch in gene expression pattern toward cholangiocyte phenotype, suggesting a functional role for GATA6 in the DR and the pathogenesis of BA. GATA6 overexpression induced several genes important for early bile duct development including HNF1β, HNF6, JAG1, and DKK1. The changes in gene expression following GATA6 overexpression may well occur gradually, and more cholangiocyte lineage markers could be upregulated after longer culture period; because of transient effect of the plasmid transfection and rapid dedifferentiation of cultured hepatocytes, this issue could not, however, be addressed. Upregulation of HNF-4α, a known hepatocyte marker, was also observed in hepatocytes following GATA6 overexpression. This is not surprising given that GATA6 has been shown to directly regulate HNF4α in visceral endoderm of embryoid bodies (24).

DKK1, HNF family members, and Notch pathway components, including JAG1, participate in the differentiation of hepatoblasts to cholangiocytes in FL, and the Notch pathway and HNF-1β also have been implicated in the pathogenesis of BA (4, 5, 7, 25, 37). In vitro studies have shown that Notch signaling induces biliary lineage marker expression in hepatic cells in a dose-dependent manner (15, 28, 46). JAG1 ligand interacts with NOTCH2 receptor to initiate a cascade of proteolytic cleavages, activating the transcription of Notch pathway target genes [reviewed previously (12)]. Alterations in Notch signaling component expression and function are evident in several biliary diseases (9, 21, 27, 29). Previous studies have linked GATA6 to the above-mentioned pathways in the gastrointestinal tract. Accordingly, GATA6 regulates DKK1 and cooperates with members of the HNF family, including HNF6 in the liver (6a, 24, 45). Furthermore, GATA6 is known to regulate Notch signaling pathway components in the murine intestinal epithelium (2, 41). In this study, we found that the expression profiles of Notch ligand JAG1, as well as those of hepatic nuclear factors HNF1β and HNF6, were similar and positively correlated to that of GATA6 in BA livers. The present findings on the expression of HNF1β, HNF6, and JAG1 in the patient samples thus further support their putative functional connection to GATA6.

All in all, the present study suggests a role for GATA6 in BA pathophysiology. Our findings, using human BA samples, mouse models for cholestasis, and in vitro cell models, implicate GATA6 as a novel marker and a potential inducer of ductal metaplasia of hepatocytes in BA. The present results suggest that factors other than cholestasis alone regulate GATA6 because its expression does not correlate with clinical laboratory measures of cholestasis. Furthermore, the expression of GATA6 in hepatocytes seems to be somewhat specific to BA compared with samples obtained from other patients with cholestasis. Given the obvious involvement of transcription factor GATA6 in the molecular pathology of BA, one of the future challenges is to identify factors regulating GATA6 expression in this disease. Cholangiocyte-hepatocyte crosstalk may play a role in the induction of GATA6 into the hepatocytes in conjunction with DR. Ultimately, combining a hepatocyte-specific Gata6 knockout mouse with a mouse model for BA could shed further light on the pathophysiological role of GATA6 in this disease.

GRANTS

This work was supported by the Academy of Finland; Emil Aaltonen Foundation, Finland; Finnish Pediatric Research Foundation, Finland; Helsinki University Central Hospital Research Grants, Finland; Sigrid Jusélius Foundation, Finland; and Pediatric Cancer Foundation Väre, Finland. S. Huppert was supported by funding from NIH R01 DK078640, NIH R01 DK107553, and the Integrative Morphology Core P30 DK078392 (CCHMC Digestive Health Center).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

T.S., M.P., N.A., and K.A.H. performed experiments; T.S., M.P., and N.A. analyzed data; T.S., M.P., N.A., J.L., D.A.R., S.S.H., D.B.W., M.P.P., and M.H. interpreted results of experiments; T.S. and M.P. prepared figures; T.S., M.P., and M.H. drafted manuscript; T.S., M.P., N.A., D.A.R., S.S.H., D.B.W., M.P.P., and M.H. edited and revised manuscript; T.S., M.P., N.A., J.L., D.A.R0., S.S.H., D.B.W., M.P.P., and M.H. approved final version of manuscript; D.B.W., M.P.P., and M.H. conceived and designed research.