Growth Hormone Mediates Its Protective Effect in Hepatic Apoptosis through Hnf6

Background and Aims Growth hormone (GH) not only supports hepatic metabolism but also protects against hepatocyte cell death. Hnf6 (or Oc1) belonging to the Onecut family of hepatocyte transcription factors known to regulate differentiated hepatic function, is a GH-responsive gene. We evaluate if GH mediates Hnf6 activity to attenuate hepatic apoptotic injury. Methods We used an animal model of hepatic apoptosis by bile duct ligation (BDL) with Hnf6 -/- (KO) mice in which hepatic Hnf6 was conditionally inactivated. GH was administered to adult wild type WT and KO mice for the 7 days of BDL to enhance Hnf6 expression. In vitro, primary hepatocytes derived from KO and WT liver were treated with LPS and hepatocyte apoptosis was assessed with and without GH treatment. Results In WT mice, GH treatment enhanced Hnf6 expression during BDL, inhibited Caspase -3, -8 and -9 responses and diminished hepatic apoptotic and fibrotic injury. GH-mediated upregulation of Hnf6 expression and parallel suppression of apoptosis and fibrosis in WT BDL liver were abrogated in KO mice. LPS activated apoptosis and suppressed Hnf6 expression in primary hepatocytes. GH/LPS co-treatment enhanced Hnf6 expression with corresponding attenuation of apoptosis in WT-derived hepatocytes, but not in KO hepatocytes. ChiP-on-ChiP and electromobility shift assays of KO and WT liver nuclear extracts identified Ciap1 (or Birc2) as an Hnf6-bound target gene. Ciap1 expression patterns closely follow Hnf6 expression in the liver and in hepatocytes. Conclusion GH broad protective actions on hepatocytes during liver injury are effected through Hnf6, with Hnf6 transcriptional activation of Ciap1 as an underlying molecular mediator.


Introduction
Growth Hormone (GH) regulates somatic growth, cell regeneration and metabolic activities. GH actions are mediated by Stat5 transcriptional activation of Igf-1 [1] and its downstream mediators, with the liver as a major site of Igf-1 production. GH-Stat5 pathway is also implicated in regulating the promoter function of Hepatocyte nuclear factor Hnf6 (Onecut; Oc-1) [2], a liver-enriched transcription factor critical to normal liver development, differentiation and function. Null mice with global disruption of Hnf6 have aberrant hepatoblast maturation, abnormal bile duct formation and die early of cholestasis [3]. In the differentiated liver, Hnf6 regulates hepatocyte target genes involved in metabolic [4] [5] [6] and regenerative functions [7] [8]. We previously treated mice with recombinant human GH to increase in vivo Hnf6 expression and found that GH treatment improved bile duct ligation (BDL) hepatic injury, enhanced cholesterol clearance and improved hepatocyte-specific cellular proliferation [8]. GH has also previously been shown to attenuate hepatocyte death. In GH receptor Ghr-deficient mice, cholic acid feeding worsens cholestasis and hepatocyte apoptosis [9]. Additionally, compared to mdr2 (multi-drug resistant transporter-2) null mutant mice, cholestasis, hepatic fibrosis and hepatocyte apoptosis were exacerbated in mdr2/Ghr or mdr2/Stat5 double null mice [9] [10] [11]. In these GH-resistant mice, Hnf6 expression was diminished, suggesting that increased susceptibility to hepatic apoptosis in the absence of GH function can be attributed to impaired Hnf6 hepatocyte-specific function, and that Hnf6 biological function is broader than the previously demonstrated Hnf6 regulation of hepatocyte proliferation and metabolic activities. We therefore test the hypothesis that GH apoptosis function during cholestatic liver injury is critically linked to Hnf6 function in hepatocytes. We used mice where Hnf6 was conditionally inactivated in the liver (Hnf6 -/-, referred to as KO mice) to evaluate if Hnf6 deficiency 1) worsens hepatic apoptotic injury; 2) impairs GH pro-survival effects in an in vivo model of hepatocyte apoptosis by BDL cholestatic injury; and 3) suppresses Hnf6 hepatic target genes underlying GH-mediated function.

Animal procedures
The animal study protocol was approved by and conducted in accordance with the Institutional Animal Care and Use Committee (IACUC) at the University of Illinois College of Medicine. Six to eight weeks F6 generation mice received care according to the IACUC guidelines.
Following bile duct ligation (BDL n = 8-10, Sham n = 4), PBS or human recombinant GH was delivered at 5 ug/h by subcutaneous Alzet miniosmotic pumps for 1 week [8]. Because male mice were used in our previous work, as well as in Ghr and Stat5 null models, and because of the higher sensitivity of male mice to GH-responsive hepatic target genes [13,14], male mice were used.

Immunostaining
Paraffin-embedded liver tissues underwent TUNEL staining using TdT-FragEL TM DNA fragmentation kit from Calbiochem (#QIA33), or α-Sma immunostaining [15]. The strength of α-Sma staining was quantified by Image J analysis program. The percentage of TUNEL-positive hepatocytes in 30 random microscopic fields for 1000 hepatocytes/mouse was counted.

Western blot assays
In three independent experiments, liver total or nuclear protein extract immune complexes [15] were detected with horseradish-conjugated secondary antibody (Fisher) followed by chemiluminescence (ECL + plus, Amersham Biosciences, Inc.).

ChiP on ChIP
Hnf6-immunoprecipitated DNA samples (with non-specific IgG as control) from three wild type and three KO liver were hybridized to Affymetrix Mouse Promoter 1.0R array containing over 25,500 mouse promoter regions. Hnf6-bound genes were grouped into functional categories using the biological module-centric algorithm DAVID Bioinformatics Resources 6.7, provided by National Institute of Allergy and Infectious Diseases (NIAID), NIH (https://david. ncifcrf.gov/home.jsp). Genes with Enrichment Scores ! 1.25 (p 0.05) were further investigated by hierarchical cluster analysis using MAT (model-based analysis of tiling-array) scores. The data have been deposited at http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE80498 in MIAME-compliant standard format. (S1 Table and

Electrophoretic mobility shift assay (EMSA)
Crude nuclear extracts from WT and KO liver or in vitro-expressed Hnf6 proteins [16] were prepared with labeled double-stranded oligonucleotides 1, 2 and 3 bearing Hnf6-binding sites ( Table 1). Protein-DNA complexes were separated from unbound DNA probes using 6% native polyacrylamide gels containing 0.5x Tris borate/EDTA in three independent experiments.
Hnf4a, Oc2, Foxa2, C/epb primer sequences have been reported previously [6,15,17]. Relative gene expression was calculated using the mathematical delta delta method by PE Applied Biosystems. Levels were normalized to Cyclophilin and reported relative to group with the lowest expression level.

Primary hepatocyte culture
Primary hepatocytes from 3-5 male KO or WT liver were isolated after collagenase perfusion, purified by Percoll centrifugation, plated at 5.5 x 10 5 cells/well and cultured in DMEM supplemented with 100 nM dexamethasone, 100 nM insulin, 100 U/ml penicillin G, 100 mg/ml streptomycin, and 10% bovine calf serum at 37˚C in 5% CO 2 . After three hours, LPS (1 ug/mL) and/or GH (1 ug/mL) were added for 32 hours before cell harvest.

Statistical Analysis
Data are expressed as mean ± S.D.. Intergroup differences were evaluated by ANOVA for repeated measures. A p value of 0.05 is considered to be significant. All statistical analyses were performed with the software SPSS.

GH diminishes BDL apoptosis in WT but not KO liver
Previously, we reported that GH infusion during BDL increases hepatic Hnf6 and Igf1 expression, with corresponding diminished cholestatic liver injury, enhanced hepatocyte proliferation and improved hepatic metabolic function [8]. Common features in hepatic cholestatic injury are hepatocyte cell death and hepatic fibrosis, with apoptosis driving the fibrogenesis response (19,20). Experimental models of cholestasis in GH-resistant mice provide strong evidence for GH role in protecting against hepatocyte death [9,10,[18][19][20][21]. To evaluate if Hnf6 directly regulates hepatocyte apoptosis and if Hnf6 mediates GH function in apoptosis, BDL CGGAAATTCCCAGCTTCTGG TTTGGTAGTGTTCAGCGAGC Forward primer and reverse primer sequences shown with position upstream of the transcription start site. was performed in Hnf6 -/-(referred to as KO) mice with conditional inactivation of hepatic Hnf6 in hepatocytes and biliary epithelial cells. As previously reported [22], at baseline, KO adult mice have normal liver biochemistries and no intrahepatic abnormality. Except for a 4-fold elevation (p = 0.02) in Hnf1b, a transcription factor regulated by Hnf6 [3], the expression of Hnf6 paralogs Oc-2 and Oc-3 [23,24] and major hepatocyte nuclear factors such as Hnf1a, Hnf4a, Foxa2, and C/ebp are unaffected in KO mice (data not shown). Following 7 days of BDL injury, relative to Sham liver, PBS-treated BDL WT show enhanced hepatocyte apoptosis by TUNEL assay (Fig 1A and 1B), increased expression of Caspase-3, -8 and -9 by real-time PCR (Fig 1C and 1E), and increased cleaved caspase protein expression by western blot ( Fig  1F). With GH treatment, TUNEL hepatocyte apoptosis is reduced (Fig 1A and 1B), as is Caspase-3, -8, -9 gene and protein expression in WT BDL liver (Fig 1C and 1F). This is paralleled by attenuated cholestasis with lower serum alkaline phosphatase [8] (Fig 2A), diminished αsmooth muscle actin (α-Sma) marker for fibrosis (Fig 2B and 2C) and profibrotic genes Ctgf, Tgfb2R (Fig 2D and 2E) in GH-treated BDL WT liver relative to PBS BDL liver. In the absence of GH, KO BDL and WT BDL hepatocytes show similar degree of apoptotic injury by TUNEL assay ( is also similar between WT BDL and KO BDL liver in the absence of GH. In KO mice, previously demonstrated GH-associated reduction of hepatocyte TUNEL labeling, and of hepatic Caspase -3, -8, -9 expression in WT BDL liver is lost (Fig 1C-1E). In fact, we observed increased caspase-3 expression, minimal to no reduction of caspase-8, and -9 proteins ( Fig  1F), undiminished cholestatic injury, no differences in Ctfg, Tgfb2R expression or α-Sma staining relative to PBS-treated KO BDL liver (Fig 2). The data suggest that Hnf6 is essential to GHmediated reduction of BDL hepatocyte apoptosis, hepatic cholestasis and fibrosis.

GH reduction of BDL apoptosis is Hnf6-dependent
Since GH transcriptionally activates Hnf6 and Igf1, we assessed their contribution to GH protective effects. Following GH infusion in WT mice, GH signaling is enhanced as shown by higher hepatic phosphorylated Stat5 protein in GH-Sham relative to PBS-Sham liver; and similarly, in GH-BDL relative to PSB-BDL liver ( Fig 3A). Correspondingly, GH-responsive Igf1 (Fig 3B), and Hnf6 gene ( Fig 3C) and protein expression ( Fig 3D) increase in GH-treated Sham WT relative to PBS-treated WT Sham; and similarly, in GH-BDL relative to PSB-BDL liver. Previous publication indeed showed that GH-induced increase of Hnf6 expression is hepatocyte specific with increased Hnf6 localization to hepatocyte nuclei [8]. While the Stat5 protein and Igf1 response is intact in GH-treated KO Sham and BDL liver, relative to PBStreated Sham and BDL KO liver respectively, as expected, Hnf6 gene and protein (Fig 3C and  3D) expression is absent in KO liver. These data show that despite intact Stat5 pathways and Igf1 expression in KO mice, the loss of Hnf6 impairs GH improvement of BDL apoptosis and fibrosis in KO liver, suggesting that GH-mediated protective response is mediated through Hnf6, independent of Igf1.

Hnf6 binds inhibitor of Apoptosis Protein gene Ciap1 promoter
Genome-wide location analysis of Hnf6-bound genes by ChIP-on-ChIP assays of WT and KO liver (n = 3 each, S1 Table and  induced cell signaling for apoptosis [25] [26] [27]) is markedly reduced in KO liver. To confirm the ChIP-on-ChIP results, we performed real-time PCR of WT and KO liver for apoptosis-related gene profiles. Among members of the Iap families, notably Ciap1, Ciap2, Survivin, and Xiap (Fig 3E), and Bcl2 families (Fig 3F), only Ciap1 and Xiap are significantly downregulated in KO liver. ChIP assays for constitutive in vivo binding of Hnf6 nuclear protein to Ciap1 and Xiap promoters show that Ciap1 promoter sequences are amplified from Hnf6-immunoprecipitated but not mock-treated or irrelevant IgG-treated hepatic chromatin/DNA complex in WT liver (Fig 4A). KO liver have no Ciap1 promoter amplification, showing that baseline Hnf6 binding to Ciap1 promoter in WT liver is lost in KO liver. No difference between IgG-vs Hnf6-bound Xiap complex is found in WT or KO liver. Furthermore, anti-Hnf6 antibody-precipitated Xiap1 promoter amplification is similar to mock-or IgG-treated liver, consistent with the lack of constitutive binding of Hnf6 to Xiap1 promoter. Ciap1 promoter sequence analysis for Hnf6 consensus binding motifs [28] reveals three potential HNF6 binding sites ( Table 2). Electromobility shift assays show that Ciap1 oligonucleotide 1 binding to in vitro synthesized Hnf6 protein is strongest (Fig 4B). These results further suggest that Ciap1 is likely a direct Hnf6-regulated target gene.
Lower case letters represent nucleotides deviating from Hnf6 consensus binding sequence. Nucleotide abbreviations are as follows: D is not C, H is A or T, W is A or G, Y is A or C and N is A, T, C or G.

Ciap1 expression correlates with Hnf6 hepatic expression profile
We next characterized the relationship between in vivo Hnf6 and hepatic expression of Iap during BDL. Previously, GH treatment reduces the extent of BDL-mediated downregulation   Fig 3C and 3D). In the current study, GH treatment in Sham or BDL WT liver is associated with upregulation of hepatic Ciap1 (Fig 5A) relative to PBS Sham or BDL liver respectively, but not of Xiap, Survivin, or Ciap2 (Fig 5C-5E). In the KO liver, the loss of Hnf6 is associated with the loss of induction in the expression of all Iap family of genes during BDL, with or without GH treatment (Fig 5A). Of note, Bcl2 gene expression in both WT and KO liver is similar between PBS-treated BDL and GH-treated BDL samples (Fig 5F). Western blotting shows that paralleling Ciap1 expression pattern in WT liver, Ciap1 protein was higher in GH-treated Sham relative to PBS-Sham liver, and likewise, in GH-BDL liver relative to PBS-BDL liver (Fig 5B). In contrast, Ciap1 expression in KO liver is uniformly lower in Sham, PBS-BDL and GH-BDL relative to WT liver counterparts (Fig 5B). The data showing tight correlations between Hnf6 and Ciap1 patterns of expression are consistent with Ciap1 as an Hnf6-regulated target gene.

Hnf6 binds to Ciap1 promoter in BDL liver
Electromobility shift assay of liver nuclear extracts against Ciap1 promoter oligonucleotide 1 (Fig 4C) shows that in vivo, Hnf6-Ciap1 signal complex is stronger in GH-treated BDL WT liver relative to PBS-BDL WT liver, reflecting enhanced Hnf6 expression and Hnf6 binding to the Ciap1 promoter in GH-treated samples. KO BDL liver shows poor HNF6-Ciap1 signal complex, even with GH treatment. These data are in further support of Hnf6 binding and regulation of the Ciap1 promoter as a target gene during the hepatic adaptive response to injury.

GH enhances Ciap1 and attenuates apoptosis in primary hepatocytes
IAP proteins suppress TNFaαnd TRAIL-induced cell death by inhibiting the activity of initiator and effector caspases [25] [26] [27]. Ciap1 knockdowns with either targeted SiRNA or antagonist to Ciap using endogenous or pharmacologic SMAC (Second Mitochondrial Activator of Caspases) increase stress-mediated apoptosis, while Ciap1 overexpression inhibits cell death in various cell-derived tissues [29] [30] [31] [32]. In primary rat hepatocytes, transfection with adenoviral vector expressing human CIAP1 suppresses LPS/cytokine mediated caspase-3 activation [33]. Conversely, hepatocyte cell lines as well as primary mouse hepatocytes treated either with ShRNA targeting Ciap1 or with SMAC mimetics become sensitized to apoptosis [34] [35]. To further evaluate Hnf6/Ciap1 transcriptional link with hepatocyte-specific apoptosis, primary hepatocytes from KO and WT liver were exposed in vitro to high dose LPS. Western blot (Fig 6)  injury during BDL [15], and in ethanol-or carbon tetrachloride-treated liver (data not shown)). GH treatment however restores Hnf6 expression in LPS hepatocytes, with Ciap1 expression pattern paralleling that of Hnf6 in LPS and GH-LPS samples. In WT hepatocytes, suppression of Hnf6/Ciap1 at high LPS dose is associated with enhanced cleaved caspase-3, while GH-dependent increased Hnf6/Ciap1 expression correlates with suppression of LPSstimulated apoptosis. In spite of Hnf6 loss and parallel reduction in Ciap1 expression, KO hepatocytes respond to LPS with similar ability as in WT hepatocytes to induce caspase-3. In these KO hepatocytes however, GH effect on enhancing Hnf6 and Ciap1 is impaired. This is associated with compromised ability of GH to attenuate LPS-induced apoptosis. In summary, the stress response in KO mice is characterized by the loss of GH-mediated attenuation of apoptotic injury in LPS-treated primary hepatocytes because of suppressed Hnf6-and Ciap1 expression.

Discussion
Hepatocyte apoptosis is a universal finding in all forms of liver injury, including cholestasis. We previously demonstrated that Hnf6 mediates hepatocyte survival functions such as hepatocyte cellular proliferation, hepatic glucose and cholesterol metabolism through Hnf6 transcriptional activation of its target genes [4][5][6][7]36]. We herein evaluated the role of Hnf6 in hepatocyte-specific apoptosis in the previous model of BDL by using mice with conditional inactivation of Hnf6 in the liver. Null mice do not exhibit more severe hepatocyte apoptotic BDL injury, demonstrating that Hnf6 is not essential to protection against hepatic apoptosis. Hepatic Igf1 during BDL injury is unimpaired in KO mice. KO mice thus may have compensated for Hnf6 loss through, as previously described, Igf-1 pro-survival effects on cell proliferation [37,38] and anti-apoptotic activity [39][40][41]. The loss of Hnf6 function in KO mice may be further counterbalanced by preserved expression of Oc2 and Oc3 paralogs, Foxa2, and notably by compensatory increases in Hnf1b. Hnf1b and Foxa2 have been demonstrated to have prosurvival [42] and antiapoptotic [43] function in hepatocytes, respectively. These transcription factors may cooperate with Igf1 to further alleviate KO liver susceptibility to hepatic cell death. Prior cholestatic injury models in Ghr, Ghr/Mdr2 or Stat5/Mdr2 null mice have implicated GH-Stat5-Igf1 pathway in GH-mediated cell growth, survival, metabolism, and particularly protection against hepatocyte apoptosis [9] [10]. GH can also modulate hepatic gene function in an Igf1-independent manner through GH-Stat5 regulation of Hnf6 promoter. We previously demonstrated that GH treatment during BDL in mice attenuates injury-related downregulation of Hnf6 expression. In these animals, enhanced Hnf6 expression is associated with improved hepatocyte proliferation, cholestasis, and cholesterol clearance. This protective response is linked to GH-Stat5 induction of Hnf6 and Hnf6 Cyclin D1 and Cyp7a1 target genes [8]. Our current data show that with GH treatment, broad reduction in hepatic apoptosis (extrinsic and intrinsic pathways) and in cholestasis is observed in WT BDL mice. GH protective function is however abrogated in KO BDL mice, despite intact Stat5 and Igf1 response. Hnf6 is therefore necessary for GH broad protection against apoptosis and cholestasis during BDL injury, independently of Igf1 [8].
Albumin-Cre-mediated liver-specific inactivation of Hnf6 affects both hepatocytes and cholangiocytes in our KO liver. While the contribution of Hnf6 inactivation in cholangiocytes cannot be excluded, the demonstrated GH-Hnf6 link to hepatic apoptosis is likely hepatocytedominant on the basis of hepatocyte-specific TUNEL immunostaining and primary hepatocyte data. Furthermore, hepatocytes constitute 85% of the liver cell mass while cholangiocytes only comprise 3-5% of the epithelial cell population. Interestingly, our previous work demonstrated that while Hnf6 has prosurvival effects on hepatocytes, Hnf6 impairs cholangiocyte proliferation [15]. While we do not observe baseline biliary duct abnormalities in the KO mice, it remains to be determined if the KO liver will respond during BDL injury to GH with an accentuated ductal proliferative reaction over biliary apoptosis.
The liver displays a sexual dimorphic response to GH [44], with Hnf6 participating in this sex-differential network to regulate many metabolic enzymes in a sex-specific fashion, such as CYP2C12, a member of cytochrome P450 family [45].
We have yet to determine if GH-and Hnf6-mediated responses to hepatocyte apoptosis are also sex-specific.
With respect to cholestasis, bile transport gene promoters contain Hnf6 binding sites (data not shown), suggesting that Hnf6-targeted genes may be involved in bile uptake and excretion. Further evaluation of Hnf6 regulation of cholestatic genes is pending. Overall, our data suggest that previously described aggravated hepatic injury in Gh-deficient mice is likely from dual Igf1 and Hnf6 defective functions. The implication is that preservation of both Igf1 and Hnf6 axes is important to hepatic survival responses against apoptosis and cholestasis. Future evaluations of cholestatic injury in double null Igf1/Hnf6 against single Igf1 or Hnf6 null mice will shed further light on their relative biological importance.
Hepatocellular cell death is a driver and sensitizer for liver inflammation and fibrosis [46]. GH attenuation of apoptosis is therefore expected to improve BDL-fibrosis [8]. GH-dependent Igf1 is an important protective mechanism against fibrosis, since hepatic fibrosis is accelerated in Igf-deficient cholestatic mice [10], but attenuated following Igf1 replacement [21]. Despite a normal Igf1 response to GH during BDL, GH does not improve hepatic fibrosis in KO mice. The GH-Hnf6 regulated pathway, independently of Igf1, is therefore an alternative mechanism for GH protective effect in fibrosis. A potential mediator of this pathway is Hnf6-regulated Tgfb2r and Ctgf target genes. We previously demonstrated that in vivo, Hnf6-adenoviral vector suppresses Tgfb2r hepatic expression [8]. With GH treatment, increased Hnf6 and reduced BDL hepatic fibrosis are associated with diminished Tgfb2r gene and protein [8], suggesting that Hnf6 negatively regulates Tgfb2r. Furthermore, Tgfb2r signaling is increased in global Hnf6 null mice embryonic liver [47], while our ChIP-on ChIP data show over-amplified Tgfb2r promoter in KO relative to WT liver (data not shown). These data are consistent with Hnf6 as a transcriptional suppressor of Tgfb2r promoter function. Of note, pancreatic Ctgf expression was upregulated in mice with pancreas-specific Hnf6 inactivation [48], suggesting that HNF6 antifibrotic function may be mediated through its transcriptional inhibition of Tgfb2r and Ctgf. It remains to be seen if GH-dependent alleviation of fibrosis is secondary to suppression of apoptosis by GH/Igf1-, GH/Hnf6-dependent pathways, and/or through GH/ Hnf6-regulated suppression of Hnf6 fibrogenic Tgfb2r or Ctgf target genes.
Among members of the IAP Inhibitor of Apoptosis Protein families modulating TNFinduced apoptosis and survival [25] [26] [27], Ciap1 (Bicr2) promoter expression is markedly reduced in KO liver. Consistent with Ciap1 as an Hnf6-transcriptionally regulated gene, ChiP assays reveal that Hnf6 constitutively binds Ciap1 promoter while baseline hepatic Ciap1 expression is severely diminished in KO mice. Electromobility shift assays additionally show that GH treatment of WT BDL mice increases Hnf6-Ciap1 nuclear complex signal over untreated WT BDL liver, with real time PCR and Hnf6 and Ciap1 protein expression recapitulating the gel shift data. All of these GH-enhancing effects are suppressed in Hnf6 KO mice. Together, the data are in support of Hnf6-regulated Ciap1 transcriptional activation as an underlying mechanism for GH antiapoptotic function during BDL.
During injury, Ciap1 inhibits apoptosis signaling through a Caspase-activation and -recruitment domain, and several Baculo Iap repeat domains to interact with TNF-receptor associated factors TRAF2/3 and caspase-3, -7, -9 [25] [26]. Despite lower baseline Ciap1 expression in KO primary hepatocytes, LPS stimulates caspase expression equally well in both WT and KO hepatocytes, in support of the complexity of LPS-induced apoptosis. These in vitro findings are compatible with our in vivo data showing that BDL KO liver do not exhibit worse apoptotic injury. GH restores Ciap1 and attenuates apoptosis in WT but not KO primary hepatocytes, consistent with Ciap1 as a mediator of GH hepatic antiapoptotic function in a hepatocyte-specific fashion through Hnf6-dependent pathways. These data recapitulate our in vivo whole liver findings that the protective effect of GH on apoptosis is lost in Hnf6-and Ciap1-deficient KO BDL mice. It is worth noting that Ciap is implicated in viral clearance in mouse model of hepatitis B infection [49], as well as in chemoresistance [32,50,51], providing the clinical rationale for IAP antiviral [52] and antitumor therapies [53]. It remains to be determined if Ciap1 link to the protective effect of GH on in vivo hepatocyte apoptosis can be explored as a potential application for liver support therapies.

Conclusion
Hnf6 is not necessary to protection against LPS-apoptotic injury in hepatocytes, or to hepatic protection against apoptosis in BDL cholestasis. Hnf6 is however is critical to GH prosurvival function during LPS-injury in primary hepatocytes, as well as in the cholestastic liver, to attenuate hepatocyte apoptosis independently of GH/Igf1 axis. This is likely mediated through Hnf6 transcriptional target gene Ciap1. The implication is that for patients with hepatic Igf1 dysfunction in clinical settings such as chronic liver diseases [54], therapeutic interventions using Hnf6 and/or Ciap1 gain-of-function approaches may diminish hepatocellular death, improve hepatocyte proliferation and enhance liver function.  Table. ChIP-on-chip Hnf6-bound genes by functional categories. The table depicts Hnf6-bound genes with Enrichment Scores ! 1.25 (corresponding to a 0.05) and Group Enrichment Score ranking by biological significance based on overall EASE (Expression Analysis Systematic Explorer) scores (corresponding to the one-tailed Fisher exact probability of gene over-representation by functional class). Baculoviral IAP repeat-containing 2 (NM_007465, Birc2 or Ciap1) is shown in Group 4, with Enrichment Score of 1.59. (DOC)