Pancreatic and duodenal homeobox gene 1 (Pdx1) down-regulates hepatic transcription factor 1 alpha (HNF1α) expression during reprogramming of human hepatic cells into insulin-producing cells

Abstract

Ectopic expression of Pdx1 triggers rapid hepatocyte dedifferentiation by down-regulating liver-enriched transcription factors and liver-specific functional genes such as hepatic nuclear factor-1α (HNF1α), albumin, and AAT. However, the links between Pdx1 over-expression and hepatic gene down-regulation are incompletely understood. HNF1α and HNF4α are important transcription factors that establish and maintain the hepatocyte phenotype. The human HNF4α gene contains two promoters (P1 and P2) that drive expression of P1-(HNF4α 1-6) or P2-(HNF4α 7-9)-derived isoforms, which are used in different tissues and at different times during development. We hypothesized that the relative expression of HNF1α and HNF4α following ectopic Pdx1 expression may promote hepatic cell dedifferentiation and transdifferentiation toward pancreatic beta-cells. We produced lentiviruses expressing Pdx1, Pdx1-VP16, and Ngn3, along with dual-color reporter genes to indicate hepatic and pancreatic beta-cell phenotype changes. Using these PTF alone or in combinations, we demonstrated that Pdx1 not only activates specific beta-cell genes but down-regulates HNF1α. Pdx1-mediated reduction of HNF1α is accompanied by altered expression of its major activator, HNF4α isoforms, down-regulating hepatic genes ALB and AAT. Pdx1 up-regulates HNF4α via the P2 promoter. These P2-driven isoforms compete with P1-driven isoforms to suppress target gene transcription. In Huh7 cells, the AF-1 activation domain is more important for transactivation, whereas in INS1 cells, the F inhibitory domain is more important. The loss and gain of functional activity strongly suggests that Pdx1 plays a central role in reprogramming hepatocytes into beta-cells by suppressing the hepatic phenotype.

Introduction

One approach for safe long-term control of blood glucose in patients with type 1 diabetes (T1D) is to establish an effective source of patient-specific autologous insulin-producing cells (IPCs) for transplantation. However, differentiated or mature beta cells cannot be efficiently expanded in vitro [1]. One of the most successful approaches for long-term treatment is islet cell transplant therapy, but its application is severely limited by the lack of donor tissue and the requirement of toxic immunosuppressant [2-7]. Cellular reprogramming has the potential to avoid these limitations by converting and expanding patients’ own tissues in vitro into the needed functional tissues [2-7]. Reprogramming studies for treatment of diabetes have mainly focused on using the liver [8-19] as a tissue source due to its high level of regenerative capacity [20] and plasticity [8,10,16], and common developmental kinship with the pancreas [21]. The liver and pancreas share a strikingly similar gene expression profile including expression of many specific transcription factors [22,23] and both tissues are responsive to changes of blood glucose [24,25]. Most studies by forced expression of key pancreatic transcription factors (PTFs), either alone or in combination, delivered into hepatic cells by a variety of means in addition to external factors such as glucose and nicotinamide have demonstrated PTF-directed hepatic progenitors to differentiate into IPCs [14,16,17].

The wide variety of differentiated tissues that arise in a developing organism are due to selective expression and/or suppression of certain sets of transcription factors that regulate the downstream gene expression in a given cell type [26-28]. Much focus has been placed on activation of the PTF to obtain beta cell phenotype that is necessary to generate functional IPCs. However, how the PTFs-mediated inactivation of the host hepatic gene program has not been well studied and is arguably of equal importance. The liver plays a major role in metabolism, glycogen storage, plasma protein synthesis, hormone production, and detoxification. Ectopic expression of Pdx1 triggers rapid hepatocyte dedifferentiation by down-regulating several liver-enriched transcription factors and liver-specific functional genes such as hepatic nuclear factor-1α (HNF1α), albumin (ALB), alcohol dehydrogenase-1β (ADH1β), glucose 6-phosphatase (G6PC), glutamate synthase (GLUL), α-1-antitrypsin (AAT), and hexokinase 2 (Hxk 2) at the mRNA and protein levels [29,30]. However, the molecular links between Pdx1 over-expression and down-regulation hepatic genes are not entirely understood.

HNF1α and HNF4α are important liver-enriched transcription factors with an important role in establishing and maintaining the hepatocyte phenotype [22,23,31]. HNF1α and HNF4α regulate the expression of hundreds of downstream target genes in both hepatocytes and beta cells [22] and the expression of specific isoforms, regulated by alternative splicing, differs significantly between hepatocytes and beta cells [22,23,32]. The human HNF4α gene contains two promoters (P1 and P2) that drive the expression of P1-derived isoforms (HNF4α 1-6) or P2-derived isoforms (HNF4α 7-9) by alternative splicing and usage of different promoters [33], which are used in different tissues and at different times during development. HNF4α P2-isoforms are exclusively detected in adult pancreatic islets and positively regulated by Pdx1, HNF1β and HNF1α [32-34]. In contrast, P1-derived isoforms are most abundant in adult hepatic tissues with relatively low levels of P2 isoforms [32,34,35]. Due to the tissue-specific expression of HNF4α in hepatocytes, HNF1α is regulated mainly by HNF4α P1 isoforms; in contrast it is regulated by P2 isoforms in beta cells. Previous studies indicate that P2 isoforms transactivate HNF1α more weakly than P1 isoforms [36,37]. Differences in the relative expression of the HNF1α and HNF4α isoforms following ectopic expression of Pdx1 may be responsible for hepatic cell dedifferentiation and transdifferentiation toward pancreatic beta cells.

To successfully create lasting beta cell surrogate applicable to the treatment of patients with T1D, the key events involved in the beta cell reprogramming process, especially the under-studied parallel process-how to suppress the host cell phenotype need to be understood. In this study, we specifically focused on studying the role of Pdx1 over-expression in human hepatocellular carcinoma cells (Huh7) in the activation of endocrine pancreatic genes and, at same time, the effects on shutting down the hepatic genes during hepatic cell reprogramming into IPCs. This work sheds insight into the mechanism of reprogramming from hepatic cells into pancreatic endocrine IPCs. Understanding the molecular and cellular events in cell type conversions may help to explain the mechanisms underlying tissue regeneration and plasticity.

Methods

Cell lines and cell culture

3T3 mouse fibroblast cells, 293 human embryonic kidney cells (ATCC cat#CRL-32610), Huh7 (JHSRRB, cat#JCRB0404, Japan), and HepG2 (Sigma, cat#85011430) human hepatocellular carcinoma cell lines were cultured in DMEM supplemented with 10% FBS, and 1% Penicillin/Streptomycin in a 37°C incubator with 100% humidity and 5% CO₂. Huh7 is a well differentiated hepatocyte-derived cellular carcinoma cell line and is commonly used studying liver cell biology, cancer and its potential therapies. In addition, Huh7 cells are easily transduced and exhibit some characteristics of hepatic stem cell properties. Therefore, here we chose Huh7 cells as our model cell line for this study. Rat insulinoma (INS1) cells were cultured in RMPI 1640 supplemented with 200 mM L-glutamine, 100 mM sodium pyruvate, 2-mercaptoethanol, and 1% Penicillin/Streptomycin in the same incubator.

Construction of plasmids

The human Pdx1 expression plasmid (pCMV6-XL5) was purchased from Origene. The mouse Pdx1 expression plasmid was constructed by insertion of mouse Pdx1 cDNA into the BamHI/XbaI site of the pCDNA3 vector (Invitrogen). The truncated mouse Pdx1 expression vectors were constructed by introduction of a stop codon at the amino acid positions of 121, 161, and 201, using the standard site-directed mutagenesis method. The HNF4α2 expression plasmid (pCMV-Sport6) was purchased from Open Biosystems. Exon 1A was removed from this plasmid and it was replaced with exon 1D in order to construct the HNF4α8 expression plasmid. The HNF4α3 and HNF4α9 expression plasmids were constructed by modifying the HNF4α2 and HNF4α8 plasmids respectively by introducing a stop codon at the 3’ end of exon 8 (note: The HNF4α3 and HNF4α9 expression plasmids code for proteins lacking the 41 amino acids from the extended exon 8). The pRL-TK expression vector was purchased from Promega. The mouse HNF1α luciferase reporter was constructed as previously described) [38]. The rat insulin I promoter (RIP) luciferase reporter was constructed by removing the TK promoter from the pRL-TK plasmid and cloning the RIP promoter into this site using the BglII/HindIII restriction sites. The Pax4, Ngn3, NeuroD1, and Nkx2.2 luciferase reporters (pFOXluc) were generous gifts from Dr. Michael German, the University of California, San Francisco, CA.

Lentiviral vector preparation, titration, and transduction

The insulin I promoter driving the green fluorescence protein (RIP-GFP) reporter was constructed as previously described [17]. The lentiviral vector (LV) containing the mouse Pdx1-VP16 fusion gene was constructed as previously described [17,39]. The LV containing rhGFP, and mouse Pdx1, were constructed as previously described [14]. The LV containing mouse Ngn3 was constructed by inserting the cDNA of mouse Ngn3 into the pTYF vector cassette under the control of the elongation factor-1 alpha (eEF-1α) promoter. The LV containing dual promotors of AFP-GFP/RIP-DsRed were constructed with similar approaches [17]. LVs were generated and titrated as previously described [40-42]. For lentiviral transduction, Huh7 cells were transduced with different virus as indicated at a multiplicity of infection (MOI) of 10 in the presence of 10 μg/ml polybrene for 24 hours. The transduction efficiency was monitored by transducing Huh7 cells with LV encoding rhGFP. The insulin promoter activity in Huh7 cells was detected by observing RIP-GFP expression under fluorescence microscopy.

Western blotting and immunocytochemistry (ICC)

Equal amount of cell lysates was loaded on 10% SDS-PAGE and the proteins were transferred to nitrocellulose membrane. Western blotting was performed using primary antibodies to either PV or Pdx1 as previously described [14]. For ICC, cultured cells were released by trypsin digestion and 10,000 cells/cytospin slides were prepared by centrifugation. Cells were stained with rabbit polyclonal antibodies to Ngn3 at a dilution of 1:1000 for 2 hours at room temperature (generous gift from Dr. Michael German) and followed by goat-anti-rabbit secondary antibody linked with HRP and visualized with DAB.

RT-PCR and real time RT-PCR analysis

Total RNA was extracted from Huh7 cells transduced with different combinations of PTFs or from human islets (generous gift from Dr. Xiaoping Deng at The University of Pennsylvania) using Trizol according to the manufacturer’s protocol. Gene expression was detected by RT-PCR. The forward and reverse PCR primers were designed (IDT Technologies) to be located in different exons. Amplification was performed for 35 cycles at 94°C for 30 s, 56°C for 30 s, 72°C for 30 s, and 72°C for 7 min. The PCR products were separated in 2% agarose gels by electrophoresis in TAE buffer. Digital images were captured and analyzed with a Quantity One Imager (BioRad). All of the PCR products were confirmed by Big-Dye DNA sequence analysis in an ABI 377 sequencer (Global Medical Instrumentation, Inc.) following the manufacturer’s protocol. Real-time RT-PCR was performed on selected samples, collected as described above, using SYBR Green PCR Master Mix (Applied Biosystems) according to the manufacturer’s protocol. Primers were designed to detect HNF4α P1 products (exon 1A) or HNF4α P2 products (exon 1D) as well as the C-terminal modifications by detecting exon 9, exon 9+, or exon 8+. Primer sequences are available upon request.

Transfections and luciferase assay

Cells were cultured as previously indicated in 12-well plates and transfected with 0.1-1.0 μg/well DNA (as indicated) using Lipofectamine 2000 Reagent (Invitrogen) according to the manufacturer’s protocol. 0.02 μg/well TK-Luc plasmid was used as a transfection control in all experiments. Cell lysates were harvested and measured 24 hours post transfection using the Dual Luciferase Reporter Kit (Promega) according to manufacturer’s protocol except that only 50 ul of each substrate reagent was used (as optimized by our lab) to read samples using a Lumat LB 9507 Luminometer (Berthold Technologies). All luciferase assays were done in triplicate.

Statistical analyses

Statistical analyses were performed with a 2-sample Student’s test assuming equal variances. Statistical significance was determined by requiring a P value of less than 0.05 for the data to be considered statistically significant.

Results

Pdx1-VP16 and Ngn3 together strongly induced insulin promoter activity

We first assessed if our LVs designed to express PTFs were functional. Transgene expressions at gene and protein levels were confirmed by RT-PCR, Western blotting, and ICC (Figure 1A-C). In order to online monitor the activation of the insulin gene during the reprogramming of hepatic cells towards IPCs, a LV-RIP-GFP was used as indicator. The LV transduction efficiency of Huh7 cells was determined by LV-mediated expressing hGFP driven by the EF1α promoter, a constitutive promoter without cell specificity. We observed more than 99% of transduced Huh7 cell expressing GFP at day 2 (data not shown). RIP-GFP was transduced into Huh7 cells, along with Pdx1, Pdx1-VP16, Ngn3, or with combinations of Pdx1/Ngn3 or Pdx1-VP16/Ngn3 as indicated. At 96 hours of transduction, expression of GFP was examined (Figure 1D). Green fluorescence was observed in less than 2% of Huh7 cells transduced with RIP-GFP only; indicating that insulin was not expressed in the Huh7 cell line. Green fluorescence was observed in rare Huh7 cells transduced with single gene Pdx1, Pdx1-VP16, or Ngn3. Therefore, Pdx1, Pdx1-VP16, or Ngn3 alone was not sufficient to activate the insulin gene at day 4 post-transduction. Green fluorescence was observed in nearly 50% of Huh7 cells transduced with Pdx1/Ngn3 and in more than 90% of Huh7 cells transduced with Pdx1-VP16/Ngn3. This result is consistent with our previous studies that the insulin promoter can be more effectively activated by Pdx1-VP16, a super active form of Pdx1 [14,17].

Figure 1.

Confirmation of LV-PTF gene expression by RT-PCR (A), Western blotting (B), and immunocytochemistry (C) and activation of rat insulin I gene in Huh7 cells (D&E). (A) Total RNA was collected from Huh7 cells following transduction with LV expression vectors for Pdx1, Pdx1-VP16, Ngn3, or combinations as indicated. Indicated gene expression was measured by RT-PCR. (B) Huh7 cells were transduced with expression vectors for Pdx1, Pdx1-VP16, Ngn3, or combinations as indicated. Following transduction, Huh7 cells were scraped off and lysed in RIPA buffer. Equal amounts of cell lysate were separated by SDS-PAGE and immunoblotted with polyclonal antibodies against Pdx1 or VP16 as indicated. INS-1 cells were used as a control for Pdx1. (C) Huh7 cells were transduced with expression vectors for Pdx1, Pdx1-VP16, Ngn3, or combinations as indicated. Following transduction, Huh7 cells on cytospin slides were fixed and stained with Anti-Ngn3 antibody by ICC. (D) Huh7 cells were transduced with LV-RIP-GFP reporter alone or with Pdx1, Pdx1-VP16, Ngn3, Pdx1/Ngn3, or Pdx1-VP16/Ngn3 expression LVs for 4 days. Expression of RIP-GFP was observed and photographed at the day 4 or (E) at different days (2, 3 and 4) following Huh7 cells transduced with the combination either LV-Pdx1/Ngn3 or LV-Pdx1-VP16/Ngn3 under fluorescence microscopy.

To further compare the efficiency of Pdx1 and Pdx1-VP16-mediated activation of insulin gene, GFP expression was observed at 48, 72, and 96 hours after Huh7 cells were transduced with Pdx1/Ngn3 or Pdx1-VP16/Ngn3 (Figure 1E). The results showed that Pdx1-VP16 was much more effective than Pdx1 in activating RIP-GFP. Green fluorescence was observed in 50% of Huh7 cells transduced with Pdx1-VP16/Ngn3 at day 2 compared with only 2% of cells treated with Pdx1/Ngn3.

Examination of the Huh-7 cell transdifferentiation progress by using a dual-color reporter system AFP-GFP/RIP-dsRed

Cell transdifferentiation means that the original cell genes have to be gradually shut-down along with the new-type of cell marker genes turned-on. To observe the progress of transdifferentiation from liver cell to IPC, we constructed plasmid of a dual-color reporter system by introducing the GFP reporter under control of α-fetal protein promoter and dsRed under control of rat insulin-1 gene promoter (AFP-GFP/RIP-dsRed) (Figure 2A) and produced its lentivirus that was used for simultaneously observing the liver-to-IPC transdifferentiation process. Using this dual-color reporter, liver cells would exhibit only green fluorescence, whereas IPCs would exhibit red fluorescence, and the intermediate phase cells would exhibit both red and green (yellow) fluorescence. When Huh-7 cells were transduced with LV-Pdx1-VP16 and LV-Ngn3, along with the dual-color reporter, at day 2 the cells only displayed green fluorescence. At day 4 to 10, approximately half of transduced cells (green cells) produced red fluorescent protein and the other cells remained green color (Figure 2B), indicating that the activation of the insulin gene was not equivalent to a complete cell transdifferentiation. After 20 days, the cell morphology started to change into smaller size. In the transduced cells, 45-50% of cells remained as liver cells (green cells), 25-30% of the cells displayed yellow cells, and 20-25% of the cells displayed only red cells. Therefore, this dual-color reporter system allowed us to observe a dynamic progression of transdifferentiation of hepatic cells into pancreatic IPCs, as it occurred as early as day 4, and became more dominant in one or the other phenotype after 20 days post transgene expression, with an estimated transdifferentiation efficiency of 20-25%. However, we cannot exclude that the low efficiency was caused by unequal transduction of the three lentiviruses. Compared to Huh-7 cells, the ability of transdifferentiation of HepG2 cells was much lower (Figure S1), likely due to low efficiency of cell transduction by lentivirus.

Figure 2.

Identification of insulin producing cells by using a dual-color reporter system. A. Schematic picture of the dual-color reporter system that contains a human alpha fetal protein promoter (AFP) driving expression GFP and a rat insulin promoter (RIP) driving expression of dsRed (AFP-GFP/RIP-dsRed). B. Time-dependent activation of AFP-GFP/RIP-dsRed reporter genes following Huh7 cells transduced by lentiviruses encoding Pdx1, Ngn3 and the dual-color reporter. Cells of red fluorescence only as indicated by white arrows at 20 days are fully transdifferentiated insulin-producing cells.

Pdx1 and Ngn3 induced gene expression related to endocrine pancreas

Given the limitations imposed by the chromatin structure on endogenous genes, next we questioned whether the activation of the RIP-GFP reporter might be applicable to the endogenous human insulin gene. As shown in Figure 1E, although LV transduction with Pdx1-VP16/Ngn3 led to activation of the ectopic promoter RIP-GFP reporter as early as 24 hours, the activation of the endogenous human insulin gene could only be detected at 96 hours by RT-PCR (Figure 3). Notably, the level of activation of the human insulin gene in the cells still remained low, implying that limitations may be due to the chromatin structure of the endogenous gene or the different regulation requirements of the human and rat insulin genes.

Figure 3.

RT-PCR analysis of PTFs and liver specific gene expression induced by various LV-PTF treatments. Huh7 cells were treated with LV-PTFs as indicated for 96h. The total RNA was isolated and reverse transcribed into cDNA. Equal amounts of each cDNA sample were used for RT-PCR analysis with gene specific primers.

To investigate the gene expression profile of Huh7 cells with different treatment after 96 hours, cellular RNA samples were collected and RT-PCR analysis was performed (Figure 3). Several endocrine pancreas functional genes were screened and it was found that endogenous expression of h-Pdx1 and Nkx6.1 could be activated by all treatments. Ngn3, Pdx1/Ngn3, and Pdx1-VP16/Ngn3 activated Pax4, NeuroD1, and Isl1, which are involved in endocrine pancreas development, consistent with previous findings that these genes are direct targets of Ngn3 [43]. Pdx1-VP16 alone or combined with Ngn3 can induce expression of somatostatin (SS). Pancreatic type glucokinase (GK) was only expressed in cells transduced with Pdx1/Ngn3 or Pdx1-VP16/Ngn3. No glucagon or pancreatic polypeptide was detected in any sample. Huh7 cells did have basal expression of Pax6 and GLUT-2, and continued expressing hepatic marker genes such as HNF1β, HNF4α, HGFR, AFP and OCT4 following all treatments. Expression of AFP and OCT4 genes are consistent with Huh7 cells possessing some properties of hepatic stem cells.

Quantitative analysis of PTF promoter activity and the relationship between upstream and downstream PTFs induced by Pdx1 or Pdx1-VP16

Using the dual-luciferase assay we quantitatively analyzed the promoter activity of PTFs in the reprogrammed Huh7 cells. Huh7 cells were transduced with LV carrying Pdx1, Pdx1-VP16, Ngn3, Pdx1/Ngn3, or Pdx1-VP16/Ngn3 for 72 hours and then separately transfected with PTF-luciferase reporter genes (Pax4, Ngn3, NeuroD1, RIP, or Nkx2.2). The relative luciferase activities are shown in Figure 4. Ngn3 alone can directly activate Pax4, NeuroD1, INS1, and Nkx2.2 promoters. However, Ngn3 alone failed to activate its own promoter, which seems inconsistent with the observation that Ngn3 autoactivates its own expression via cooperation with Foxa2 in vitro [44]. Ngn3 combined with Pdx1 or Pdx1-VP16 did not further increase Ngn3 activation, indicating that both Pdx1 and Pdx1-VP16 have no synergistic effect on the activation of Pax4, NeuroD1, and Nkx2.2 genes. Pdx1 can activate Ngn3 promoter, consistent with Pdx1 as upstream factor controlling Ngn3 expression. Previous studies have shown that a modified form of Pdx1 carrying the VP16 transcriptional activation domain from the herpes simplex virus more efficiently induces insulin gene expression in the human HepG2 cell line and the rat WB cell line [14,17], but whether this attribute is a general characteristic of other PTF genes or limited only to the insulin gene was not explored. Our results showed that insulin promoter activity was about 3-fold higher in cells transduced with LV-Pdx1-VP16/Ngn3 than in those treated with LV-Pdx1/Ngn3. However, the Pdx1 modified with VP16 showed lower activation of Pax4, Ngn3, and Nkx2.2 promoters when compared to Pdx1 alone. Therefore, the addition of VP16 restricts Pdx1 transactivation in most cases, indicating the mechanism of activation of insulin by Pdx1 different from activation of the Pax4, Ngn3, and Nkx2.2 promoters.

Figure 4.

Luciferase reporter gene analysis during reprogramming. Relative activity of promoter genes was measured for Pax4, Ngn3, NeuroD1, RIP, or Nkx2.2 luciferase reporters following Huh7 cell transduction with LV-Pdx1, Pdx1-VP16, Ngn3, Pdx1/Ngn3, or Pdx1-VP16/Ngn3 for 72 h. Huh7 cells transduced with LV-rhGFP were used as a control.

Effect of Pdx1 on HNF1α promoter activity in hepatic cells

Previous studies have shown that several hepatic genes are down-regulated following over-expression of Pdx1 in hepatic cells [29,30]. In the current study, we have shown a down-regulation of endogenous HNF1α and its target genes of ALB and AAT in Huh7 cells and HepG2 cells following treatment with Pdx1 or Pdx1-VP16 by RT-PCR (Figure 3 and S2) and Western blot analysis (Figure 5A). This prompted us to evaluate the effect of Pdx1 on expression of HNF1α in hepatic cells. Huh7 cells were co-transfected with the HNF1α -luciferase reporter and various concentrations of human Pdx1 or mouse Pdx1 expression plasmid as indicated for 24 hours. Transfection of both Pdx1 plasmids decreased HNF1α promoter activity in a concentration dependent manner (Figure 5B). With the 0.8 ug Pdx1 expression plasmid, approximately 60% of the activity of the HNF1α promoter was lost. The specificity of inhibition of HNF1α promoter activity by Pdx1 was supported by the fact that C-terminal truncated Pdx1 constructs did not show any inhibitory activity (Figure 5C).

Figure 5.

Down-regulation of HNF1α, ALB, and AAT in Huh7 cells over-expressing Pdx1. A. Huh7 cells were transduced with or without LV-Pdx1 or LV-rhGFP for 48 h. Equal amounts of Huh7 cell lysate were separated by SDS-PAGE and immunoblotted with polyclonal antibody against Pdx1, HNF1α, ALB, AAT, or β-actin. B. Huh7 cells were transfected with 1.0 µg of HNF1α-luciferase reporter plasmid and 0.2-0.8 µg Pdx1 expression plasmid as indicated, with pcDNA3 used as a DNA quantity control. C. Huh7 cells were transfected with 1.0 µg of HNF1α-luciferase reporter plasmid and 0.8 µg of each truncated Pdx1 (120, 160, or 200 amino acids) or full length (283 amino acids), along with pcDNA3 as a DNA quantity control. All experiments were done in triplicate for three times.

Pdx1 increases HNF4α P2 transcripts that compete with P1 transcripts

The major regulator of HNF1α is HNF4α [45]. HNF4α is a complex gene regulated by two distinct promoters (P1 and P2) and alternative splicing [33,46-48] that shows tissue specific expression [32,34,35]. In the adult human liver, HNF4α isoforms are primarily expressed from the P1 promoter whereas, in the adult human pancreas and islets, P2 driven isoforms are predominant [32,34,35]. In addition, Pdx1 is a known regulator of HNF4α P2 driven isoforms and has a binding site in the P2 proximal promoter [33]. Thus, real time RT-PCR was performed in Huh7 cells and HepG2 cells following treatment with LV-Pdx1-VP16 to examine the effects on the HNF4α promoter and isoform expression. As expected, a pronounced increase in HNF4α P2 transcripts was observed compared to P1 transcripts (Figures 6A and S3). Using primers to identify specific HNF4α P2 driven isoforms, an increase in HNF4α7 and HNF4α8, and a decrease in HNF4α9 were observed (Figure 6A).

Figure 6.

Pdx1 increases expression of HNF4α P2 transcripts that compete with P1 transcripts. A. Total RNA was collected from Huh7 cells following transduction with LV-Pdx1-VP16 or control (LV-GFP). Indicated gene expression was measured by real time RT-PCR. Primers were designed to detect general products of HNF4α from the P1 promoter (Exon 1A) or the P2 promoter (Exon 1D) as well as the three HNF4α P2 transcripts (7, 8, & 9). B. Relative activity was measured for the mouse HNF1α-luciferase reporter (1 µg/µl/well) following transfection of various HNF4α isoforms (1 µg/µl/well) in Huh7, INS1, and 293 cells. Activity was normalized in each cell type to the pcDNA3 control. C. Relative activity was measured for the mouse HNF1α-luciferase reporter following transduction of HNF4α2 and Pdx1 alone, and in combination. D. Relative activity was measured for the mouse HNF1α-luciferase reporter following transfection of HNF4α2 and HNF4α8 alone, and in combination. + indicates 1 µg/µl/well and the concentration gradient shows three concentrations (0.1 µg/µl, 0.5 µg/µl, and 1.0 µg/µl)/well.

HNF4α P2 driven isoforms were previously reported to be weaker transactivators of their target genes [36,37]. We therefore investigated the transactivation potential of various HNF4α isoforms from both the P1 and P2 promoters in Huh7, INS1, and 293 cells using the HNF1α-luciferase reporter (Figure 6B). In all of these cell lines it is clear that HNF4α P1 driven isoforms are stronger activators than their corresponding (same c-terminal domain) P2 driven isoforms. However, HNF4α mediated regulation of HNF1α is different between Huh7 and INS1 cells. In Huh7 cells, both HNF4α P1 isoforms function as stronger activators whereas in INS1 cells, the strongest activators are the isoforms containing the same truncated C-terminal end (HNF4α3 and HNF4α9).

To further investigate the mechanism by which Pdx1 influences down-regulation of HNF1α, we set up competition assays in 3T3 cells because they do not express any transcription factors related to our system (data not shown). It was demonstrated that Pdx1 has no direct effect on the HNF1α-luciferase reporter (Figure 6C). We then activated this reporter using the liver specific HNF4α isoform HNF4α2 and demonstrated that Pdx1 can then suppress this activation (Figure 6C). Since Pdx1 does not directly interact with the HNF1α promoter, it was hypothesized that it may function indirectly by altering the regulation of its major activator HNF4α. Since Pdx1 can increase expression of HNF4α P2 driven isoforms, we set up a competition assay in 3T3 cells to examine the effect of expressing multiple HNF4α isoforms (Figure 6D). In this case, the HNF1α-luciferase reporter was activated using the liver specific HNF4α isoform HNF4α2 and then the effect of competition with the corresponding beta cell specific isoform HNF4α8 was examined (both have the same C-terminal modification). As expected and similar to the effect of Pdx1, HNF4α8 can competed with HNF4α2 and significantly reduced the activity of the HNF1α-luciferase reporter. With a 1:1 ratio of HNF4α2/HNF4α8, the activity of the HNF1α-luciferase reporter is almost the same as the HNF4α8 base level. Therefore, it appears that HNF4α P2 driven isoforms are able to compete with P1 driven isoforms to reduce the activation of target genes.

Discussion

In the present study, we established an effective model by LV expression of Pdx1, Pdx1-VP16, Ngn3, and combinations of Pdx1/Ngn3 or Pdx1-VP16/Ngn3 in Huh7 cells in which to study the process of Pdx1-mediated reprogramming of hepatic cells into IPCs. We have demonstrated that co-expression of Ngn3 with Pdx1 is important for the activation of several endocrine pancreas genes such as Pax4, NeuroD, Isl-1, and Nkx2.2. We also examined the role of a Pdx1 fusion protein (Pdx1-VP16) that strongly activates insulin gene expression and increases the efficiency of hepatic to pancreatic reprogramming [8,14,17]. In the present study we further investigated the effects of Pdx1-VP16 on other pancreatic endocrine genes, in addition to the insulin gene, in order to determine if this fusion gene has the same synergistic potential on the transactivation of other PTF genes. It was shown that while Pdx1-VP16 synergistically increased insulin gene expression as previously demonstrated, it failed to have the same effect on several other key genes that are essential to the pancreatic endocrine phenotype. The results show that the Pdx1-VP16 fusion gene only increases transactivation of the insulin gene and suggests that the insulin gene may use a different mechanism of transcription than other pancreatic endocrine genes.

The liver is largely composed of hepatocytes, which occupy 78% of parenchymal liver volume [49,50]. Hepatocytes carry out the primary functions of the liver such as metabolism, detoxification, and protein synthesis of several essential compounds including serum albumin, fibrinogen, and transferrin. It has been suggested that the dominant mechanism for controlling expression of hepatocyte specific genes is at the transcriptional level [51], however, the molecular mechanism by which Pdx1 regulates expression of these hepatic genes is not well established. Using our reprogramming model, it was demonstrated that over-expression of Pdx1 can down-regulate expression of HNF1α in hepatic cells. Because HNF1α is very important in the regulation of a wide array of hepatic genes [22], it may play a fundamental role in the process of dedifferentiation of the hepatic cell phenotype during the reprogramming into IPCs. The RT-PCR and Western blotting data showed that down-regulation of endogenous HNF1α resulted in decreased expression of downstream target genes ALB and AAT during reprogramming. Our findings are consistent with a previous study [29], where adenovirus-mediated expression of Pdx1 led to down-regulation of several mature hepatocyte-specific genes including ALB, ADH1B, G6PC, GLUL, and AAT. Several key hepatic genes are direct downstream transcriptional targets of HNF1α including ALB, AAT, AFP, α- and β-fibrinogen, transthyretin, and pyruvate kinase [52-58]. Therefore, Pdx1 induced down-regulation of HNF1α expression can affect these downstream target genes and may be important in the process of dedifferentiation of hepatic cells by down-regulating expression of an array of genes that determine the hepatocyte phenotype.

Our results also suggest that the mechanism for down-regulation of HNF1α is by affecting expression of its major activator HNF4α. (Note: See The Sladek Lab webpage for detailed information about HNF4α structure and function at http://www.sladeklab.ucr.edu/HNF4.shtml). It was demonstrated that specific HNF4α isoforms function differently in liver than in beta cells (Figure 6A, 6B). In Huh7 cells, the activity of specific HNF4α isoforms appears to rely on the N-terminal domain that results from using different promoters (P1 vs. P2). HNF4α isoforms that are expressed using the P1 promoter contain exon 1A and are the strongest activators in Huh7 cells. The C-terminal domains that resulted from alternative splicing did not significantly alter the function of HNF4α in the context of our experiment in Huh7 cells. In INS1 cells the C-terminal modifications appear to play a major role in the function of HNF4α (Figure 6B). P1 driven isoforms are still stronger activators in INS1 cells when compared to their corresponding P2 driven isoforms (with similar C-terminal ends) but having the truncated C-terminal end that stops in exon 8 increases the activity independent of the first exon. The isoform driven by the P2 promoter that terminates in exon 8 (HNF4α9) is a stronger activator than the P1 driven isoform that include exons 9 and 10 (HNF4α2). This suggests that the C-terminal modifications of HNF4α due to alternative exon splicing may have a beta cell-specific function.

According to the real time RT-PCR data (Figure 6A), HNF4α isoforms that are stronger transcriptional activators containing exon 1A (AF-1 activation domain) or terminating in exon 8 (HNF4α2, HNF4α3, or HNF4α9) are suppressed by treatment with Pdx1 while HNF4α isoforms that are weaker transcriptional activators that contain exon 1D and exons 9 and 10 (HNF4α8). Exons 9 and 10 comprise the inhibitory F domain. It appears that HNF4α dependent regulation of HNF1α in hepatocytes requires stronger activators (P1 driven) than in beta cells which predominantly rely on P2 driven isoforms. This may be due to cell specific mechanisms for regulating this gene. For example, the transcription factor Nkx6.1, which is expressed in beta cells but not in hepatocytes [59], has been shown to regulate HNF1α by binding to its distal promoter [38]. We also produced evidence (Figure 6D) that promoter specific isoforms can compete with each other for activation of their target genes. It is possible that induced expression of HNF4α from the P2 promoter may suppress P1 expression during the reprogramming process of hepatocytes to IPCs, as the opposite has been shown during rodent liver development [60]. Briefly, ectopically expressed Pdx1 protein in hepatocytes activates HNF4α P2 isoforms by binding directly to the P2 promoter resulting in increased expression of P2 isoforms that are weak activators of HNF1α, resulting in suppression of HNF1α activity via a competitive binding with HNF4α P1 isoforms. Ectopic Pdx1 expression may also induce P2 CpG demethylation and increase P1 DNA methylation resulting in increased P2 transcripts in hepatocytes. By studying HNF gene regulation in future studies, we also may gain insight into the epigenetic mechanisms of Pdx1-mediated hepatic phenotype dedifferentiation.

Taken together, our findings indicate that transcription factors that function in multiply cell types must be regulated at specific levels appropriate for each cell type. These new findings may prove useful in the field of cell reprogramming and provide insight for the mechanism of reprogramming hepatic cells into IPCs by Pdx1.

Disclosure of conflict of interest

None.