Sanford Children's Health Research Center, Sanford-Burnham Medical Research Institute, La Jolla, California, USA
Correspondence: Fred Levine, M.D., Ph.D., Sanford Children's Health Research Center, Sanford-Burnham Medical Research Institute, La Jolla, California 92037, USA. Telephone: 858-795-5179; Fax: (858) 795-5387; e-mail: email@example.com
Author contributions: S. L. and R.P.: conception and design, collection and/or assembly of data, data analysis and interpretation, manuscript writing, and final approval of manuscript; E.K.: financial support; A.P.: provision of study material BI6015; F.L.: conception and design, data analysis and interpretation, manuscript writing, and final approval of manuscript. S.-H.L. and R.P. contributed equally to this article.
Increasing the number of β cells is critical to a definitive therapy for diabetes. Previously, we discovered potent synthetic small molecule antagonists of the nuclear receptor transcription factor HNF4α. The natural ligands of HNF4α are thought to be fatty acids. Because obesity, in which there are high circulating levels of free fatty acids, is one of the few conditions leading to β-cell hyperplasia, we tested the hypothesis that a potent HNF4α antagonist might stimulate β-cell replication. A bioavailable HNF4α antagonist was injected into normal mice and rabbits and β-cell ablated mice and the effect on β-cell replication was measured. In normal mice and rabbits, the compound induced β-cell replication and repressed the expression of multiple cyclin-dependent kinase inhibitors, including p16 that plays a critical role in suppressing β-cell replication. Interestingly, in β-cell ablated mice, the compound induced α- and δ-cell, in addition to β-cell replication, and β-cell number was substantially increased. Overall, the data presented here are consistent with a model in which the well-known effects of obesity and high fat diet on β-cell replication occur by inhibition of HNF4α. The availability of a potent synthetic HNF4α antagonist raises the possibility that this effect might be a viable route to promote significant increases in β-cell replication in diseases with reduced β-cell mass, including type I and type II diabetes. Stem Cells2013;31:2396–2407
The control of β-cell replication is stringent. Once the β-cell mass is established by a process that involves β-cell neogenesis from precursors in embryonic ducts followed by a wave of β-cell replication during prenatal and early postnatal development, there are few physiological settings in which it is stimulated, including obesity, pregnancy, and hyperglycemia [1, 2]. In addition to the physiologically relevant stimulators, β-cell replication also occurs in models of pancreatic damage, most notably partial pancreatectomy [3, 4], but also pancreatic duct ligation . Interestingly, in mice and humans [6-10], the potential of β cells to replicate appears to decline rapidly with age.
Because β-cell mass is decreased but not absent in both type I and type II diabetes [11, 12], there has been great interest in determining the factors that stimulate β-cell replication, with the goal of developing a pharmacologic modulator of that process. In particular, a large variety of growth factors have been tested [13-20]. While this has been successful in some cases (reviewed in ), the induction of β-cell replication has been limited primarily to very young animals due to a poorly understood but profound age-related decrease in β-cell replicative capacity [8-10].
Recently, there has been increasing interest in small molecule inducers of β-cell replication. Recent studies found that glucokinase activators, BACE2 inhibitors, and adenosine modulators induced β-cell replication [17-19, 21]. Glucokinase activators presumably act on β-cell replication by mimicking the effect of hyperglycemia, a known stimulator of β-cell proliferation, but the pathway by which adenosine metabolism acts on β-cell replication appears to be distinct and remains unclear.
In the course of high-throughput screening (HTS) for small molecule insulin promoter modulators, we discovered a potent inhibitor of HNF4α activity . Our initial HTS hit, 1-(2′-chloro-5′-nitrobenzenesulfonyl)−2-methylbenzimidazole (BIM5078), was a dual HNF4α antagonist and PPARγ agonist . This compound was converted by structure-activity studies into a more metabolically stable and HNF4α-specific analog, the benzimidazole phenylsulfonamide BI6015, that lacked activity on PPARγ and that had appropriate pharmacokinetic properties for in vivo administration .
As a member of the nuclear receptor family of transcription factors, HNF4α is a desirable pharmacologic target, as a higher percentage of nuclear receptors have been developed into targets for approved drugs than any other category, including G-protein coupled receptors (GPCRs) . In the β cell, HNF4α plays an important role in glucose-responsive insulin secretion, and is mutated in a dominant, monogenic form of diabetes, MODY1 . Genetic deletion of HNF4α at the time of insulin promoter activation during development resulted in mild β-cell dysfunction but did not affect the β-cell mass, implying that HNF4α does not affect the wave of β-cell replication that occurs during development and in early postnatal life . However, the normal stimulation of β-cell replication that occurs during pregnancy was absent in mice lacking HNF4α in their β cells, which was interpreted as demonstrating a role for HNF4α in β-cell replication . Overexpression of HNF4α in β cells led to a DNA damage response rather than replication, leading to confusion about the role of HNF4α in β-cell replication . In the liver, HNF4α deletion during development led to profound hepatocyte dysfunction , but conditional deletion during adulthood led to hepatocyte replication, demonstrating that HNF4α may play different roles at different times .
The endogenous ligands of HNF4α are thought to be fatty acids [29, 30]. High fat diet produces a biphasic effect on β-cell mass, with an early compensatory increase followed by β-cell apoptosis . This suggests that fatty acids might play a role in the control β-cell replication, but the mechanism is unclear. Based on our previous finding that fatty acids act as HNF4α inhibitors , we hypothesized that our synthetic HNF4α antagonists might promote β-cell replication. BI6015 strongly stimulated β-cell replication in two species, mice and rabbits. In normal mice, only β cells within the islet replicated, while in mice that had undergone chemically induced β-cell ablation, α- and δ-cells replicated as well. There was marked specificity for islet cells and β-cell number was substantially increased, demonstrating that the compound induced true replication.
Materials and Methods
Four-month-old male imprinting control region (ICR) mice (Harlan Sprague Dawley, Inc. Hayward, CA (www.harlan.com)). were injected once per day subcutaneously with dimethyl sulfoxide (DMSO) (n = 8) or BI6015 (10 mg/kg, n = 8, or 30 mg/kg per day, n = 8) for 2 weeks, followed by pancreas harvest. To study the effect of aging, 1-, 4-, and 11-month-old mice were treated with DMSO (n = 3 in each age group) or BI6015 (30 mg/kg per day, n = 3 in each age group) for 5 days, followed by pancreas harvest. For the diabetes model, 4-month-old mice were injected intraperitoneally (IP) with streptozotocin (STZ, 150 mg/kg, Sigma-Aldrich, St. Louis, MO (www.sigmaaldrich.com)) in sodium citrate buffer (pH 4). Blood glucose level was monitored from the tail vein with a glucometer (OneTouch Ultra, Lifescan, Milpitas, CA (www.lifescan.com)) and mice with a blood glucose [mt]400 mg/dl for more than 2 consecutive days were randomized to control or BI6015 (30 mg/kg per day) for 10 days, followed by sacrifice and pancreas harvest 12–24 days after STZ injection (n = 6–8 in each group). BrdU (1 mg/ml, Sigma-Aldrich) was administered continuously in the drinking water and was replaced every fourth day. All procedures were performed in accordance with Sanford-Burnham Institute IACUC regulations.
Six-month-old Dutch-Belted rabbits (Covance, San Diego, CA (www.covance.com)) were injected subcutaneously with DMSO (n = 3) or BI6015 (30 mg/kg per day, n = 3). BrdU was administrated continuously in the drinking water for 10 days, followed by pancreas harvest. All procedures were performed in accordance with TSRI Institute IACUC regulations.
Immunocytochemistry and Image Quantification
Samples were harvested from BI6015 and DMSO-treated mice and rabbits, fixed in 4% paraformaldehyde (USB; OH), and embedded in OCT freezing media (Sakura Finetek; Torrance, CA (www.sakura-america.com)). Slides of 5 µm thickness were washed four times with phosphate buffered saline (PBS) and treated with 0.3% Triton in PBS for 10 minutes. Antigen retrieval was done using with CitriSolv (Thermo Fisher Scientific, PA) for 7 or 10 minutes in sub-boiling temperature. After washing with PBS for 10 minutes, slides were incubated in blocking solution with 5% normal donkey serum (Jackson Immuno Research, West Grove, PA (www.jacksonimmuno.com) PA) for 60 minutes at room temperature. Antibodies are listed in Supporting Information Table S1.
For fluorescent imaging, samples were incubated with the indicated primary antibody (Supporting Information Table S1) and Alexa 488 (Invitrogen; CA La Jolla, CA (www.invitrogen.com)), Rhodamine or Dylight (Jackson ImmunoResearch Laboratories, Inc.) fluor-labeled anti-mouse, rabbit, guinea pig, rat, or goat antibodies and nuclei were counterstained with 4',6-diamidino-2-phenylindole (DAPI) (Invitrogen). Controls using secondary antibodies alone were done to ensure specificity of immunostaining. Fluorescently labeled sections were analyzed with a conventional inverted microscope (Olympus, PlanFl 40x/0.60; PA www.olympusamerica.com).
All the image analysis was done with Image J (NIH). For the normal mice, more than 10 islets with at least 50 endocrine cells per islet were analyzed for each mouse. For the diabetic mice, total endocrine cell number was determined by counting the particular cell type in similar size pancreatic sections, with three randomly selected sections analyzed from each mouse. For rabbits, 100 islets per rabbit were analyzed. P16 density was measured in DAPI-positive nuclei of each islet cell type and normalized to the DAPI intensity.
RNA was purified using the RNeasy Kit (Qiagen, Valencia, CA (www.qiagen.com)). Two micrograms of RNA was used to synthesize cDNA using the qScript cDNA SuperMix (Quanta BioSciences; Gaithersburg, MD). Q-PCR was conducted on cDNA corresponding to 100 ng of RNA using the Opticon Real-Time System (MJ Research, BioRad, Hercules, CA; MA) and Q-PCR SuperMix (BioPioneer, San Diego, CA (www.biopioneerinc.com)). All mRNA values were normalized to 18S rRNA or GAPDH values and are expressed as fold change over vehicle-treated control.
HNF4a Binding Assay
HepG2 cells were treated with fatty acid-free bovine serum albumin (BSA) (Research Organics) or 500 µM fatty acids (C8, C10, palmitate, oleate), DMSO (Sigma-Aldrich, St Louis, MO), or 20 µM BI6015 for 18 hours, with additional compound added 1 hour before harvest. Nuclear protein was extracted using the Nuclear Extract Kit (Active Motif). The protein concentration of nuclear extraction was determined using a BCA protein assay kit (Thermo Fisher Scientific, Waltham, MA (www.thermofisher.com)). HNF4α DNA binding was assessed using the HNF TransAM ELISA assay (Active Motif, #46296) with an anti-HNF4α-specific antibody supplied with the kit. For the negative controls, competitor and mutated consensus oligonucleotides of HNF binding sites were used to verify HNF4α specificity in assay. Three micrograms of nuclear extract was added into wells with immobilized oligonucleotides containing HNF4α (5′-TGGACTTAG-3′) consensus-binding sites. HNF4 α specific (1:1,000) and anti-rabbit horseradish peroxidase-conjugated antibodies (1:1,000) were used to detect HNF4α binding, detected by absorbance on a spectrophotometer within 5 minutes at 450 nm with a reference of 655 nm (OptiMAX, Molecular Devices, Sunnyvale, CA (www.moleculardevices.com)).
Cell Culture and Chemical Treatment
T6PNE cells were maintained in Roswell Park Memorial Institute medium (RPMI) (5.5 mM glucose, Hyclone) supplemented with 10% fetal bovine serum (FBS, Hyclone) and 1% penicillin-streptomycin (pen-strep, Gibco) and grown in 5% CO2, 37°C. To induce E47 activity, 0.5 µM tamoxifen (Sigma-Aldrich) was added to culture media. MIN6 cells were maintained in Dulbecco's modified Eagle's medium (DMEM) (high glucose, Hyclone, Thermo Scientific) supplemented with 15% FBS (Hyclone), 1% pen-strep, and β-mercaptoethanol (Sigma-Aldrich) and grown in 10% CO2, 37°C. HepG2 cells were cultured in DMEM (high glucose, Hyclone) supplemented with 10% FBS and 1% pen-strep and grown at 5% CO2, 37°C on collagen plate (BD Bio Coat). Fatty acid-free BSA (Research Organics, Sigma Aldrich) or 200 µM fatty acids (C10, C16) and DMSO or 0.5 µM BI6015 were added to T6PNE cells for 2 days and to MIN6 cells for 5 hours.
Data are presented as mean ± SEM of three or more independent cultures and animals. Statistical significance was assessed using two-tailed unpaired Student's t test.
Fatty Acids Inhibited HNF4α Activity
While fatty acids are found in the HNF4α ligand binding pocket, there has been little consensus on the extent to which HNF4α activity is regulated by fatty acids. Previously we found that fatty acids inhibited insulin promoter activity. The effect of fatty acids and the structure-activity relationship was consistent with their being weak HNF4α antagonists (Supporting Information Fig. S1) . Here, we extended those findings by examining the effect of fatty acids on HNF4α DNA binding and expression. Both oleic and palmitic acids inhibited HNF4α DNA binding, albeit not as well as the much more potent synthetic HNF4α antagonist BI6015. C8 and C10 fatty acids exhibited lesser degrees of inhibition, consistent with our previous data on the structure-activity relationship of fatty acids on insulin promoter activity (Fig. 1A) . Oleate and palmitate repressed HNF4α gene expression , which is regulated by HNF4α binding through a positive feedback loop .
An HNF4α Antagonist Induced β-Cell Proliferation
To test the hypothesis that antagonizing HNF4α activity would induce β-cell replication, we administered BI6015 to mice subcutaneously for 2 weeks. At that point, we harvested the pancreases and analyzed them for β-cell replication. We found a dramatic increase in β cells positive for Ki67 (Fig. 1B–1E) and BrdU (Fig. 1F–1H, 1O). The increase in KI67 expression in addition to BrdU uptake is important, as BrdU uptake in β cells can result from a DNA damage response rather than true replication [26, 34]. To further rule out DNA damage as opposed to true cell replication as an explanation for the BrdU uptake and Ki67 expression, we assayed for the presence of the DNA damage marker γH2AX, which we previously showed to be present in β cells undergoing DNA damage rather than true β-cell replication . BI6015 did not induce an increase in γH2AX (Supporting Information Fig. S2).
Given the increase in β-cell replication, it was of interest to examine the other cells in the islet, as well as in the surrounding exocrine pancreas. Interestingly, proliferation of α-cells (Fig. 1I–1K, 1P) and δ-cells (Fig. 1L–1N, 1Q) was not increased by BI6015. Similarly, acinar cell replication was not induced by BI6015 in 4-month-old mice (Fig. 1R–1T, Supporting Information Fig. S3). The blood glucose concentration did not differ between DMSO- and BI6015-treated mice (70–130 mg/dl in both groups).
The Effect of BI6015 on β-Cell Replication Was Age Dependent
The mice used for the studies described above were 4 months old, well past the age at which β-cell expansion due to replication occurs in the course of maturation of the animals, and at a point where the control level of β-cell replication has fallen to an extremely low level [6, 7, 35]. Thus, the effect of BI6015 is not merely to increase already ongoing replication, but rather to stimulate replication de novo. However, the effect of BI6015 on β-cell replication was highly age-dependent, with increased replication in 1-month-old (Fig. 2A–2E) and 4-month-old (Fig. 2F–2J) mice, but not in 11-month-old (Fig. 2K–2O) or 20-month-old mice (Supporting Information Fig. S4).
BI6015 Inhibited the Expression of Multiple Cyclin-Dependent Kinase Inhibitors In Vitro and In Vivo
Having found that an HNF4α antagonist induced β-cell replication, it was of interest to determine the mechanism by which that occurred. We began by examining possible effects of BI6015 on the expression of cell cycle regulatory genes. Since the control of β-cell replication is stringent, it is not surprising that many cell cycle regulators are expressed in the β cell [1, 36]. Because HNF4α is primarily a transcriptional activator, we focused our analysis on cell cycle genes that repress cell cycle entry and that might be downregulated by BI6015-induced HNF4α inhibition, leading to cell cycle entry. Prominent among the negative regulators of cell cycle entry are cyclin-dependent kinase inhibitors (CDKIs), which act by inhibiting cyclin-dependent kinases .
In the β cell, a number of CDKIs play important roles, including p57 that causes focal β-cell hyperplasia in infants , and p16 that plays an essential role in the age-related decrease of β-cell replication in humans and mice [39, 40]. BI6015 had effects on CDKIs in both human and murine cells, although—not surprisingly—there were some interspecies differences. In the human islet cell line T6PNE , C10 and C16 fatty acids and BI6015 potently repressed the expression of p57, p27, p21, and p16 (Fig. 3). In the highly differentiated murine insulinoma cell line MIN6 , BI6015 inhibited p21 and p16 expression, but had no effect on the p27 mRNA level (Supporting Information Fig. S5).
The effect of BI6015 on positive regulators of the cell cycle was complex. It had no effect on cyclinD2, CDK4, or CDK6 in either MIN6 or T6PNE cells. The effect of BI6015 on cyclinD1 mRNA was discordant, with an increase in MIN6 but a decrease in T6PNE (Supporting Information Fig. S5). CyclinD1 protein and mRNA levels are under complex control throughout the cell cycle, with the mRNA being regulated primarily by changes in stability rather than transcription , so the effect of BI6015 on cyclinD1 mRNA may reflect the cell cycle status of the two cell lines rather than direct effects of BI6015.
Because of the central role played by p16 in controlling β-cell replication, and particularly the age-dependent decline in β-cell replication [39, 40], we focused on the effect of BI6015 on p16 expression in vivo. BI6015 had a dramatic effect on p16 expression in β cells, causing a profound decrease (Fig. 4). Consistent with the age-dependent effect of BI6015 on β-cell replication, the effect of BI6015 on p16 expression was similarly age-dependent, with the greatest effect occurring in 1-month-old animals (Fig. 4A–4C, a lesser effect at 4 months (Fig. 4D–4F), and no decrease in p16 in 11-month-old mice (Fig. 4G–4I). Interestingly, while α- and δ-cells expressed p16 at a level similar to β cells, BI6015 did not cause a statistically significant decrease in p16 expression in those cells (Supporting Information Fig. S6), consistent with the lack of effect on replication.
The Effect of BI6015 on Replication Was Restricted to Specific Cell Types
In 4-month-old mice, the effect of BI6015 on cell replication was limited to β cells within the pancreas. No effect on the replication of other islet cells or exocrine cells was detected (Fig. 1). However, an increase in acinar cell replication was seen in very young (1-month old) mice (compare insulin negative area in Fig. 2A, 2B). Thus, we examined the effect of BI6015 on other tissues, focusing on those in which HNF4α is expressed at high levels. No increase in replication was detected in the liver, kidney, or intestine of 4-month-old mice, demonstrating a high degree of specificity for the effect of BI6015 and BI6015 had no effect on p16 expression in the liver, kidney, or intestine (Supporting Information Fig. S7). Interestingly, the level of p16 protein was substantially lower or undetectable in the exocrine pancreas, where BI6015 did not have an effect on replication, than in the islet cells, where it did (Supporting Information Fig. S8).
HNF4α Antagonism Increased β-Cell Number and Induced Proliferation of α- and δ-Cells Following β-Cell Ablation
Having shown that an HNF4α antagonist could induce β-cell replication in normal mice, it became of interest to extend that result to a setting in which most β cells had been eliminated. Thus, we examined the effect of BI6015 following STZ-induced β-cell ablation (Figure 5). Even with high-dose STZ, we noticed that a small but significant number of β cells survived, providing a substrate upon which BI6015 could act. β cells that survived STZ were distinct phenotypically from susceptible β cells that were killed by STZ. MafA was present in β cells from control animals untreated with BI6015 (Fig. 6A) and was unaffected by BI6015 (Fig. 6B). However, it was absent from β cells following treatment with STZ (Fig. 6C, 6D). In addition to MafA, GLUT2 and Nkx2.2, important markers of functional β cells, [43, 44] were absent from β cells that survived STZ (compare Fig. 6G, 6K and 6E, 6I). BI6015 inhibited MafA and NKX2.2 gene expression in the T6PNE cell line (Fig. 6Q, 6R), but it did not have an effect on MafA, GLUT2, or NKX2.2 protein expression in vivo (compare Fig. 6B, 6F, 6J and 6A, 6E, 6I). Both MafA and NKX2.2 remained absent from the insulin-positive cells following STZ plus BI6015 treatment in the short time of this experiment (compare Fig. 6D, 6L and 6A, 6I). Similarly, the GLUT2 glucose transporter, which is required for glucose-responsive insulin secretion in murine β cells, remained low in treated mice (compare Fig. 6H and 6E). However, the expression of the transcription factor PDX-1 remained unchanged after STZ and/or BI6015 treatment (Fig. 6M–6P). Thus, it appears that STZ, which requires GLUT2 to enter cells, selected for a population of pre-existing β cells that lacked expression of proteins critical for β-cell function.
Treatment with BI6015 of mice that had undergone STZ-induced β-cell ablation caused a dramatic increase in replication of the small number of remaining β cells (compare Fig. 5A, 5D, 5G with 5B, 5E, 5H, quantified in 5C). Within the 12–24-day period of the experiment, the number of insulin-positive cells increased from 6% of normal to 22% of normal.
In contrast to the results in normal mice, where only β cells within the islet exhibited increased replication (Fig. 1), under conditions of β-cell ablation we observed replication of α- (compare Fig. 5D and 5E, quantified in 5F) and δ- (compare Fig. 5G and 5H, quantified in 5I) cells. α- and δ-cell expansion in response to BI6015 in β-cell ablated mice led to supraphysiological numbers of cells, with α-cells expanding to 244% of control and δ-cells expanding to a similar degree—204% of control (Fig. 5J–5L). Since we previously found evidence for α- to β-cell transdifferentiation in mice that had undergone pancreatic duct ligation plus β-cell ablation , we examined the mice for evidence that BI6015 might play a role in endocrine cell transdifferentiation, but no islet cells expressing multiple hormones were found in STZ-treated BI6015 mice.
BI6015 Induced the Expression of MafB, a Marker of Replicating β Cells
MafB is a marker of replicating β cells, being induced in a number of settings in which β cells are replicating, including pregnancy, obesity, high glucose, and ablation of Men1 expression . Thus, we examined its expression in β cells exposed to BI6015. MafB was absent as expected  from β cells of STZ-treated mice not injected with BI6015 (Supporting Information Fig. S9A–S9C). However, BI6015 treatment induced MafB expression in a subset of β cells (Supporting Information Fig. S9D–S9I).
BI6015 Induced β-Cell Replication in Rabbits
To extend our findings in mice to a larger animal, we chose the rabbit. There is increasing interest in using rabbits as a model of human diseases, including diabetes , as they have proven to mimic aspects of human physiology better than mice in a number of cases .
The basal replication level in the 6-month-old rabbit pancreas was lower than in mice—more similar to humans. Importantly, BI6015 induced β-cell replication in the rabbit pancreas, with a greater than twofold increase in the number of β cells incorporating BrdU (compare Fig. 7D–7F with 7A–7C, quantified in G). Similar to what we found in the exocrine pancreas of 4-month-old mice, BI6015 did not induce an increase in replication in the rabbit exocrine pancreas (quantified in Fig. 7H).
The central finding reported here is that a potent HNF4α antagonist selectively induced the replication of pancreatic β cells. The identification of fatty acids as the natural ligands of HNF4α [29, 30] and the fact that one of the few physiological stimuli for β-cell replication is high fat diet formed the basis for the hypothesis that a potent and selective HNF4α antagonist  might promote β-cell replication. Having pharmacological modulators of HNF4α activity allowed us to inhibit HNF4α in a temporally controlled manner in adult animals. Beginning with a specific small molecule HNF4α antagonist, we were able to test the hypothesis that inhibition of HNF4α in adult animals would result in β-cell replication, finding robust and reproducible β-cell proliferation induced by the compound.
Of critical importance in evaluating interventions to increase β-cell replication is to demonstrate that there is an increase in the number of β cells following treatment. Recent studies from our laboratory  and others [26, 48], in which entry of β cells into the cell cycle was stimulated by overexpression of a transcription factor or cell cycle regulatory gene, led to a DNA damage response rather than true replication. Consistent with true β-cell proliferation, a substantial increase in the number of β cells was found following BI6015 administration in STZ-treated mice. Because of the large number of pre-existing β cells in normal mice and the short duration of BI6015 administration, we were unable to measure an increase in the number of β cells in that setting, but no evidence of DNA damage of apoptosis was evident.
We believe that the effect of BI6015 is due to a direct effect on HNF4α rather than to an off-target effect, that is always a concern with small molecules. BI6015 was not discovered through an unbiased screen for inducers of β-cell replication. Rather, it was discovered in a screen for insulin promoter modulators  and based on its identification as an HNF4α antagonist, we hypothesized that it might induce β-cell replication. Thus, it is not plausible to postulate that its effect on β-cell replication is due to an off-target effect. A previous study where HNF4α was conditionally deleted in β cells using an insulin promoter-cre driver found a defect in the ability of β cells to expand during pregnancy [25, 49]. However, since the β-cell mass at baseline was not affected by conditional HNF4α deletion, this must not have inhibited β-cell replication during development and in the early neonatal period. Rather, it is likely that absence of HNF4α during development led to a defect in a specific pathway leading to β-cell replication during pregnancy. Two recent reports that HNF4α deletion in the adult liver led to hepatocyte replication [28, 50] support our contention that BI6015 is acting to promote β-cell replication through a direct effect on HNF4α.
While many growth factors and peptides have been shown to play a role in β-cell replication, inducing β-cell replication with a small molecule is uncommon [18, 19], and the ability to do so at an age where the background level of β-cell replication has fallen essentially to zero is even more so. We believe that the ability of our compound to induce β-cell replication at a relatively late age is due to its novel ability to inhibit the expression of multiple CDKIs, particularly p16, which plays a critical role in the age-dependent decline of β-cell replication . However, it should also be noted that not all β cells replicated in response to BI6015. This may reflect β-cell heterogeneity or a requirement for more than one factor to act in concert to induce β-cell replication.
The effect of HNF4α antagonists on multiple CDKIs was surprising initially, as the CDKI promoters do not contain HNF4α binding sites, and HNF4α has not been found to bind to CDKI promoters in ChIP-Chip assays [51-53]. However, we found previously that many genes affected by HNF4α antagonism lack HNF4α binding sites in their promoters . Statistical analysis of the pattern of gene expression induced by our compounds revealed highly significant effects on genes containing the core CANNTG E-box sequence to which bHLH transcription factors bind . The promoters of all the CDKIs regulated by BI6015 are dependent for activity on E-boxes in their promoters [32, 54-57]. DNA binding independent activity of HNF4α on the p21 promoter has also been found to occur through an SP1 site .
The effect of HNF4α antagonism on p16 in 1 and 4, but not 11- or 20-month-old mice, suggests that the control of p16 expression occurs on least two levels; one that is HNF4α responsive and one that is unresponsive. This is consistent with the complex control of the p16 gene , and the likelihood that the control of β-cell replication is multifactorial. Multiple lines of evidence point to a loss of the ability of human β cells to replicate at approximately 30 years of age [8-10] and p16 is known to be expressed in adult but not fetal human β cells . While BI6015 inhibited CDKI expression in a human islet cell line, the extent to which primary human β cells can be induced to replicate remains to be determined. Unfortunately, attempts to induce replication in human islets in vitro were confounded by effects of HNF4α antagonism on cell adhesion , leading to islet cell dissociation. However, the induction of β-cell replication by BI6015 in two species in vivo, rabbit and mouse, is promising.
The differential response of α- and δ-cells to BI6015 in the presence and absence of β cells illustrates the importance of signaling between islet cells and is reminiscent of our previous finding that α-cells proliferated and transdifferentiated into β cells only when adjacent β cells were ablated. The nature of the signal that is sensed by the non-β-islet cells is unknown. However, the fact that part of the response of those cells to β-cell ablation, that is, proliferation, is caused by a specific inhibitor of HNF4α raises the possibility that modulation of HNF4α activity in response to inflammation  may be occurring in the pancreatic duct ligation plus alloxan model that we studied previously . However, we did not detect any evidence of transdifferentiation, that is, multihormonal islet cells, following treatment with BI6015, suggesting that the signals promoting β-cell neogenesis by transdifferentiation are distinct from those stimulating proliferation.
It was striking that β cells surviving high-dose STZ treatment prior to the initiation of BI6015 treatment lacked MafA, Nkx2.2, and GLUT2 expression, all of which are important for β-cell function. β cells with undetectable levels of MafA are increased in human type II diabetes , but whether those cells arise as a result of a diabetogenic environment, or whether they are pre-existing and are better able to survive that environment has not been clear. Our data suggest that β cells lacking those proteins exist in the normal mouse pancreas.
We did not observe an effect of BI6015 on cell replication in other organs, including kidney, liver, and intestine. Even pancreatic acinar cells did not exhibit an increase in replication in 4-month-old mice, although they did respond to BI6015 to a small extent in very young mice. The reason for the differential response to BI6015 remains to be determined. Regardless, the highly selective nature of the induction of replication by BI6015 addresses a concern with any potential therapy that stimulates cells to replicate, that is, promotion of tumorigenesis.
In the liver, conditional genetic ablation of HNF4α in adult hepatocytes has been reported to induce replication , so one might have expected BI6015 to induce hepatocyte replication. However, we reported previously that BI6015 is highly metabolized in the liver , and that may have limited the extent to which cells in the liver were exposed to the compound. In fact, administration of BI6015 by the IP route, where most compound is taken up by the portal circulation, did not lead to an increase in β-cell replication (not shown), suggesting that hepatic metabolism may limit systemic exposure. The extensive hepatic metabolism and limited effect of BI6015 in other organs may have been advantageous in limiting undesired effects, particularly in the liver. It is important to note that induction of β-cell replication would involve transient rather than chronic administration of BI6015, further limiting potential undesirable effects of inhibiting that important transcription factor.
Overall, the data presented here are consistent with a model in which the well-known effects of obesity and high fat diet on β-cell replication occur by inhibition of HNF4α. The availability of a potent synthetic HNF4α antagonist raises the possibility that this effect might be a viable route to promote significant increases in β-cell replication in diseases with reduced β-cell mass, including type I  and type II  diabetes.
The principal conclusion of this study is that a synthetic small molecule HNF4α antagonist efficiently induced b-cell replication. Combined with the data presented here and previously by us that fatty acids, the natural ligands for HNF4α, function as weak antagonists, our data suggest that downregulation of HNF4a may be the mechanism by which b-cells undergo hyperplasia in response to obesity.
This work was supported by the Sanford Children's Health Research Center, BetaBat (in the Framework Program seven of the European Community), and CIRM Grant TG2-01162. This article does not involve the products or services of a commercial interest.
Disclosure of Potential Conflicts of Interest
The authors indicate no potential conflicts of interest.