HNF4alpha and HNF1alpha Dysfunction as a Molecular Rational for Cyclosporine Induced Posttransplantation Diabetes Mellitus

Posttransplantation diabetes mellitus (PTDM) is a frequent complication in immunosuppressive therapy. To better understand the molecular events associated with PTDM we investigated the effect of cyclosporine on expression and activity of hepatic nuclear factor (HNF)1alpha and 4alpha and on genes coding for glucose metabolism in cultures of the rat insulinoma cell line INS-1E, the human epithelial cell line Caco-2 and with Zucker diabetic fatty (ZDF) rats. In the pancreas of untreated but diabetic animals expression of HNF4alpha, insulin1, insulin2 and of phosphoenolpyruvate carboxykinase was significantly repressed. Furthermore, cyclosporine treatment of the insulinoma-1E cell line resulted in remarkable reduction in HNF4alpha protein and INS1 as well as INS2 gene expression, while transcript expression of HNF4alpha, apolipoprotein C2, glycerolkinase, pyruvatekinase and aldolase B was repressed in treated Caco-2 cells. Furthermore, with nuclear extracts of cyclosporine treated cell lines protein expression and DNA binding activity of hepatic nuclear factors was significantly repressed. As cyclosporine inhibits the calcineurin dependent dephosphorylation of nuclear factor of activated T-cells (NFAT) we also searched for binding sites for NFAT in the pancreas specific P2 promoter of HNF4alpha. Notably, we observed repressed NFAT binding to a novel DNA binding site in the P2 promoter of HNF4alpha. Thus, cyclosporine caused inhibition of DNA binding of two important regulators for insulin signaling, i.e. NFAT and HNF4alpha. We further investigated HNF4alpha transcript expression and observed >200-fold differences in abundance in n = 14 patients. Such variability in expression might help to identify individuals at risk for developing PTDM. We propose cyclosporine to repress HNF4alpha gene and protein expression, DNA-binding to targeted promoters and subsequent regulation of genes coding for glucose metabolism and of pancreatic beta-cell function.


Introduction
In organ transplantation there is a need to suppress an immune response against the grafted organ. Immunosuppressive therapies with calcineurin inhibitors result, however, in unwanted secondary effects. This includes risk of infections of all types, lymphomas and other malignancies [1,2]. Posttransplantion diabetes mellitus (PTDM) is a further complication with an incidence of approximately 8-10% for cyclosporine and 16-18% for tacrolimus across renal, liver, heart and lung transplant patients [3,4]. Noteworthy, the DIRECT study reports a 36% incidence of impaired glucose metabolism and a 14% incidence of PTDM with either cyclosporine or tacrolimus [5]. Indeed, cyclosporine caused morphologic and functional alterations of pancreatic beta-cells with subsequent hyperglycemia and hypoinsulinoma in diverse animal studies [6][7][8][9][10][11]. Based on their mode of action cyclosporine and tacrolimus repress interleukin-2, thereby suppressing the early cellular response of T-lymphocytes to an antigenic stimuli. As of today the causes for the diabetogenic potential of calcineurin inhibitors remain uncertain. To better understand the molecular events associated with PTDM we investigated expression and activity of hepatic nuclear factor 1a (HNF1a) and 4a (HNF4a). Notably, dysfunction of these transcription factors have been associated with diabetes mellitus. For instance, the early onset of type II diabetes referred to as MODY (maturity onset diabetes of the young) was mapped to mutations within the HNF1a (MODY3) and HNF4a (MODY1) gene [12]. Moreover, linkage analysis in combination with fine-mapping for susceptibility to multifactorial late-onset type 2 diabetes has identified predisposing variants of HNF4a and HNF1a in a growing number of studies [13][14][15]. The HNF4a-dependent transcription of HNF1a is required for normal b-cell function [16], but there is also a feedback loop of HNF4a and HNF1a to maintain tissue specific metabolic function [16][17][18]. Additionally, in conditional HNF4a knockout mice b-cell function was impaired upon glucose-stimulated insulin secretion [19][20][21] whereas HNF1a knockout mice develop diabetes [22].

Results and Discussion
Initially, we investigated expression of HNF4a in the pancreas of Zucker diabetic fatty (ZDF) rats. This is an established disease model for type 2 diabetes. We observed reduced expression of HNF4a and of genes regulated by this factor in the glucose metabolic pathway, notably phosphoenolpyruvate carboxykinase 1 (PCK1), insulin1 (INS1) and insulin2 (INS2) ( Table 1). Furthermore, HNF4a and HNF1a was significantly reduced in the liver of these animals (Table 1). In the past HNF4a was shown to regulate INS1 [29]. As rodents express two isoforms of insulin (INS1 and INS2) [30] both genes were investigated, but the physiological role of INS2 is not clear as yet [30]. By use of advanced bioinformatics we identified a new HNF4a binding site in the promoter of the INS2 gene at position 2245 to 2232 upstream of the start site of transcription [see Material and Methods for sequence information and electrophoretic mobility shift (EMSA) assay in Fig. 1D]. Loss of HNF4a DNA-binding to targeted promoters resulted in reduced expression of genes coding for glucose transport and metabolism and of insulin secretion from pancreatic ß-cells [28]. Furthermore, in conditional HNF4a knockout mice b-cell function was impaired upon glucose-stimulated insulin secretion [19][20][21]. Conversely, in HNF1a overexpressing beta cell lines increased transcript expression of insulin, glucose transporter 2, L-pyruvate kinase, and aldolase B was observed [26,27] whereas HNF1a knockout mice developed diabetes [22].
To further probe for HNF4a and HNF1a function we cultured the human intestinal cell line Caco-2. This cell line enables mechanistic studies with HNF4a protein expression being comparable to its expression levels in the liver [31]. In cell culture experiments we analyzed the effect of cyclosporine on HNF4a and HNF1a expression and activity. HNF4a gene and protein expression ( Table 2, Fig 1A) as well as HNF1a protein expression ( Fig 1B) was significantly repressed after treatment of Caco-2 cells with 25 mM (30 mg/ml) cyclosporine for 72 h, but HNF1a gene expression remained unchanged (Table 2). For comparison actin western blotting was used as housekeeping protein (Fig. 1C). Additionally, we investigated expression of genes coding for glucose metabolism, i.e. apolipoprotein C2 (ApoC2), aldehyde dehydrogenase 2 (ALDH2), phosphoenolpyruvate carboxykinase 1 (PCK1), glycerol kinase (GK), pyruvate kinase (PKLR) and aldolase B (ALDOB), and found ApoC2, GK, PKLR and ALDOB transcripts to be significantly repressed ( Table 2). We further studied the ability of HNF4a to bind to promoter sequences of HNF1a, ApoC2, GK, PKLR, ALDOB, and INS2 by EMSA supershift assays. As shown in Fig. 1D we observed strong binding of nuclear extracts of untreated cell cultures to all cognate recognition sites. Addition of a specific HNF4a antibody shifted the band, therefore providing clear evidence for the specificity of the assay. Strikingly, cyclosporine reduced binding of HNF4a to all EMSA probes employed to approximately 20% when compared with untreated cell cultures (Fig 1D, 1E). Binding activity of HNF1a to its recognition site in the pancreas specific P2 promoter of HNF4a was reduced as well ( Fig. 2A, 2B), but treatment with equimolar concentrations of the calcineurin inhibitor tacrolimus did not influence HNF4a gene expression (Table 3).
To further confirm cyclosporine mediated dysregulation of HNF4a we analyzed different rat and mouse beta cell lines, i.e. INS-1E, RINm5F and MIN6 cells, for its HNF4a expression. INS-1E cells express HNF4a more abundantly and therefore were used for subsequent experiments ( Table 4). As INS-1E cells are much more sensitive to the cyclosporine induced toxicity effects than Caco-2 cells, cell viability was tested at different cyclosporine concentrations. Treatment of INS-1E cells with 8.3 mM (10 mg/ ml) cyclosporine (one third of the concentration used for Caco-2 cells) resulted in a 55% viability (Fig. 3A). In western blotting experiments actin served as a housekeeping protein, which we found to be constantly expressed (Fig. 3B). HNF4a protein expression of INS-1E cells is much lower than in liver [32]. In nuclear protein extracts HNF4a expression was below the limit of detection but its gene expression was unchanged (Table 5). Nonetheless, HNF4a DNA binding activity could be assayed for in EMSA supershift assays and was significantly reduced to 58% after treatment with 8.3 mM (10 mg/ml) cyclosporine (Fig. 3C, 3D). It is of considerable importance that the gene expression of the HNF4a target genes insulin1 (INS1) and insulin2 (INS2) was significantly repressed (Table 5).
Taken collectively, HNF4a and HNF1a expression and DNAbinding activity was repressed after cyclosporine treatment as was transcription of genes in the glucose and insulin signaling pathways targeted by HNF4a and HNF1a. Our study is the first report to determine a direct connection between cyclosporine treatment and activity of hepatic nuclear factors and our findings provide a molecular rational for PTDM observed in transplant patients. We suggest individual differences in the HNF4a gene and protein expression amongst patients to be of critical importance for the diabetogenic potential of cyclosporine. Indeed, on average 1/10 of cyclosporine treated patients develop PTDM. Consequently, repression of HNF4a by cyclosporine depends on the abundance of HNF4a protein. In Fig. 4 HNF4a gene expression in the liver of 14 patients was plotted; the data are scattered over a wide range. Likely, patients with low HNF4a and HNF1a protein would be at higher risk of developing cyclosporine induced PTDM. Specifically, cyclosporine binds to calcineurin and inhibits Ca 2+dependent serine / threonine phosphatase activity [33]. Normally this phosphatase dephosphorylates nuclear factor of activated Tcells (NFAT), which moves from the cytoplasm to the nucleus to Gene expression was measured by real-time qRT-PCR in 14 weeks and 9 months old ZDF rats and lean controls (n = 10 animals, respectively) and was determined relative to expression of cyclophilin, which served as a housekeeping gene. Gene expression in control rats was set to 100 and values for ZDF rats represent transcript abundance relative to control. Non-parametric Mann-Whitney-U-Test was used to compare ZDF and control groups. Results are considered significant at p,0.05 (gene names and p-values in bold). Gene expression of HNF4a in the liver of this cohort of ZDF rats has been previously reported [43]. doi:10.1371/journal.pone.0004662.t001 associate with other proteins, thereby regulating expression of interleukin-2, granulocyte macrophage colony stimulating factor (GM-CSF), TNFa, IFNc and other interleukins [34,35]. Although inhibition of calcineurin results in immunosuppression, altering activity of NFAT will also impact regulation of INS1 gene transcription. Indeed, this factor is activated by calcineurin in response to increased Ca 2+ -levels [36]. Disruption of the NFAT/ insulin pathway may contribute to the diabetogenic effects of cyclosporine as will be discussed below. Notably, Heit et al [37] reported the b-cell specific deletion of calcineurin to result in agedependent diabetes, while conditional expression of activated NFAT reverted the diabetic phenotype in these mice. Furthermore, expression of genes critical for b-cell endocrine function e.g. HNF4a and HNF1a was increased in mice when NFATc1 was conditionally activated [37]. It is of considerable importance that NFAT cooperates with other transcription factors involved in insulin transcription such as PDX1, NEUROD1 and HNF4a. The evidence for this cooperation stems from chromatin immunoprecipitation assays [37]. The calcineurin/NFAT signaling appears to be essential for the regulation of pancreatic b-cell function; its cooperation with HNF4a could provide a molecular rational for cyclosporine induced PTDM [37]. HNF4a activity differs amongst cell types, in part due to use of alternate promoters. Whilst in hepatocytes the P1 promoter of HNF4a is primarily activated, the P2 promoter is specifically activated in pancreatic b-cells [17,18] Indeed, P2 is exclusively expressed in INS-1E cells, see Table 6. In the study of Heit et al [37] binding of NFAT to the P1 promoter of HNF4a (NM_008261) was observed. The findings of Heit et al [37] are surprising as for normal b-cell function usage of the P2 promoter of HNF4a would have been expected. Notably, we observed NFAT binding at the human P2 promoter of HNF4a at position 2461 to 2450 upstream of the start site of transcription (see Material and Methods for sequence information). Furthermore, binding of NFAT to the HNF4a P2 promoter was reduced in response to cyclosporine treatment (Fig. 5A, 5B), but expression of members of the NFAT gene family (NFATc1, c2, c3, c4) and of calcineurin itself was unchanged after cyclosporine treatment of Caco-2 cells (Table 2). There is clear evidence for a role of NFAT in glucose/insulin homoeostasis [38]. NFAT signaling plays an essential role in the development of diabetes in calcineurin knockout mice [37]. Taken collectively, we report a remarkable repression of HNF4a and HNF1a after cyclosporine treatment and propose cyclosporine to act through a calcineurin/NFAT dependent mechanism on these transcription factors. We further identified a novel NFAT binding site in the human HNF4a P2 promoter and report HNF4a activity and expression of genes of the glucose/insulin signaling pathway to be reduced in the pancreas of ZDF diabetic rats.
In conclusion, cyclosporine repressed HNF4a/HNF1a expression, DNA-binding to targeted promoters and subsequent expression of genes involved in glucose metabolism and pancreatic b-cell function. We propose a molecular mechanism for PTDM based on dysregulation of HNF4a/HNF1a and of NFAT insulin signaling pathway targeted by cyclosporine.

Cell culture and cyclosporine treatment
Caco-2 cells, a human intestinal cell line derived from a colon adeno-carcinoma, were obtained from and cultivated as recom- h] and 32 P labeled oligonucleotides to probe for DNA binding to HNF4a binding-sites within promoters of HNF1a (HNF1a), apolipoprotein C2 (ApoC2), glycerol kinase (GK), pyruvate kinase (PKLR), aldolase B (ALDOB), and insulin2 (INS2). In EMSA supershift assays an antibody directed against HNF4a (+) was added. Shifted (HNF4a) and supershifted bands (HNF4a ss) were marked. (E) Dried EMSA gels were analyzed with a Molecular Imager (BioRad, Muenchen, Germany) using the Quantity One software (BioRad, Muenchen, Germany). HNF4a binding of control extracts to the respective binding sites was set to 100% and inhibition of binding to the respective binding sites after treatment with cyclosporine [

Diabetic disease model
Pancreas (animals aged 9 months) and liver (animals aged 14 weeks) of fa/fa obese Zucker diabetic fatty (ZDF) rats and of +/fa lean nondiabetic control rats were kindly provided by W. Linz and H. Ruetten (Sanofi-Aventis, Frankfurt, Germany) [42]. Pancreatic mRNA degrades quickly, i.e. in less than 1 minute after tissue resection, therefore, pancreas was frozen immediately. All rats were male with mean body weight of 398.8630.2 (obese) and 334.2619.3 (lean) for 14 weeks aged animals and 403.8635.7 (obese) and 463.3630.3 (lean) for 9 months aged animals. Representative phenotype data (e.g. blood glucose, insulin) are provided in Niehof et al [43].

Isolation of nuclear extracts, western blotting analysis and electrophoretic mobility shift assays
Nuclear extracts were isolated by the method of Dignam et al [44] with minor modifications as detailed previously [31]. Details for western blotting analysis and electrophoretic mobility shift assays were given in Niehof and Borlak, 2005 [31]. Antibodies directed against HNF4a (sc-6556), HNF1a (sc-6547), and Actin (sc-1616) were purchased from Santa Cruz Biotechnology (Heidelberg, Germany). Nuclear extracts were prepared mainly in triplicate and used as described in the figure legend. The antigen-antibody complexes were visualized using the enhanced chemiluminescence (ECL) detection system (PerkinElmer Life Sciences, Rodgau-Juegesheim, Germany). Light signal detection was done with the CCD camera Imager system Kodak IS 440 CF (Kodak, Biostep GmbH, Jahnsdorf, Germany) and quantification was performed using the Kodak 1D Image analysis software (version 3.5.). The oligonucleotides were purchased from MWG Biotech (Ebersberg/Muenchen, Germany) with the following sequences: AAG GCT GAA GTC CAA AGT TCA GTC CCT TC (HNF1a, NM_000545), TGT CTA GGC CAA AGT CCT    , NM_001030004] and were 32 Plabeled. Super shift assays were done with HNF4a specific antibody (sc-6556x), HNF1a specific antibody (sc-6547x), and NFAT specific antibody (sc-1149x), all were purchased from Santa Cruz Biotechnology, Heidelberg, Germany and once again details are given in [31].

RT-PCR and real-time semi-quantitative PCR
Total RNA was isolated using the nucleospin RNA Isolation Kit (Macherey-Nagel) according to the manufacturers recommendations. 4 mg total RNA from each sample was used for reverse transcription (Omniscript Reverse Transcriptase, Qiagen, Hilden,    Table 7. doi:10.1371/journal.pone.0004662.g004 Table 6. HNF4a isoform expression in INS-1E cells.

HNF4a isoform Mean6SD
HNF4aP1 0 HNF4aP2 418.186225.99 HNF4a isoform expression was measured by real time qRT-PCR in INS-1E cells after 6 days in culture (n = 3, respectively). Gene expression was determined relative to expression of mitATPase6, which served as a housekeeping gene. Gene expression in rat liver served as positive control for HNF4aP1 expression, gene expression in rat pancreas served as positive control for HNF4aP2 expression. doi:10.1371/journal.pone.0004662.t006 Figure 5. Cyclosporine inhibits NFAT binding to the P2 promoter of HNF4a. (A) Electrophoretic mobility shift assays with 2,5 mg Caco-2 cell nuclear extract [control or cyclosporine treatment, 25 mM (30 mg/ml) for 72 h] and 32 P labeled oligonucleotides to probe for DNA binding to the NFAT binding site within the HNF4a P2 promoter (NFAT-site in HNF4a P2). In EMSA supershift assays an antibody directed against NFAT was added. Control and treated probes were run on same gels. (B) Dried EMSA gels were analyzed with a Molecular Imager (BioRad) using the Quantity One software (BioRad). NFAT binding of control extracts was set to 100% and inhibition of binding after treatment with cyclosporine [

Statistical analysis
All values are expressed as mean6standard deviation. To determine significance between two groups, comparison was made using the non-parametric two-tailed Mann-Whitney-U-Test. Therefore, Statistica software, version 7.1 (StatSoft) was used. The results are considered significant when the p value was less than 0.05.