Conditional Expression Demonstrates the Role of the Homeodomain Transcription Factor Pdx1 in Maintenance and Regeneration of β-Cells in the Adult Pancreas

The basic elements of the Pdx1 tet-off inducible gene repression system are shown schematically in Fig. 1. Pdx1^tTA knock-in and Tg^Pdx1 transgenic mice were generated as previously described (25). To obtain mice in which Pdx1 could be conditionally repressed, the Pdx1^tTA/+ knock-in mice were crossed with Tg^Pdx1 mice and progeny intercrossed to derive Pdx1^tTA/tTA;Tg^Pdx1 mice as well as genotype control mice (see results).

Animals were housed under standard 12-h light/dark conditions and fed and watered ad libitum. Experiments were approved by the Royal Melbourne Hospital Campus Animal Research Ethics Committee. Groups of mice were age and sex matched within each experiment. Doxycycline was administered as a single intraperitoneal dose of 100 mg/kg at time t = 0 and maintained by addition to drinking water at 0.5 mg/ml. BrdU (100 mg/kg) was administered intraperitoneally as a single dose 16 h before the mice were killed.

Mice were fasted overnight and challenged with 2 g/kg i.p. glucose. Glucose levels in retroorbital venous blood samples were measured at 0, 15, 30, 60, 120, and 180 min postchallenge with an Accu-chek Advantage glucometer (Roche Diagnostics, Castle Hill, NSW, Australia). Statistical analyses (Mann-Whitney tests) were performed using Graphpad Prism (GraphPad Software, San Diego, CA).

Adult pancreata were fixed in Histochoice (Sigma, St. Louis, MO) or 4% paraformaldehyde for 3 h before processing, embedding in paraffin, and sectioning (5 μm). The following primary antibodies were used: guinea pig anti-human insulin (Dako, Carpinteria, CA) at 1:1,000, rabbit anti-human amylase (Sigma) at 1:1,000, rabbit anti-human glucagon (Dako) at 1:200, rabbit anti-human somatostatin (Dako) at 1:100, rabbit anti-GLUT2 (a gift from Bernard Thorens, Université de Lausanne, Lausanne, Switzerland) at 1:100, and rabbit anti-human filtrin (Alpha Diagnostic, San Antonio, TX) at 1:20. Rabbit anti-Pdx1 (a gift from Christopher Wright, Vanderbilt University Medical Center, Nashville, TN) was used at 1:500 and rabbit anti-Pdx1 antibody (purified IgG 1:25) was raised against a glutathione S-transferase fusion protein containing amino acids 14–284 of murine Pdx1. Fluorescein isothiocyanate–conjugated anti-guinea pig (ICN Biomedicals, Aurora, OH) or Alexa fluor 568–conjugated goat anti-rabbit (Molecular Probes, Eugene, OR) immunoglobulins were used for immunofluorescence. Nuclei were counterstained with 4′,6′-diamidine-2′-phenylindole dihydrochloride (DAPI). BrdU-labeled cells were detected with a cell proliferation detection kit (Amersham, Buckinhamshire, U.K.) with mouse monoclonal anti-BrdU primary and Alexa fluor 568–conjugated anti-mouse secondary antibodies. Terminal deoxynucleotide transferase-mediated dUTP nick-end labeling was performed with an apoptag fluorescein apoptosis detection kit (Serologicals, Norfolk, GA). Digital images were captured with an Axiocam camera from an Axioplan2 compound microscope (Carl Zeiss, Göttingen, Germany). Dual-color immunofluorescence images were compiled from two separate images of the same section using fluorochrome-specific filters.

BrdU-labeling index was determined by imaging 6–8 fields (200× magnification) of randomly selected pancreas sections from each of three untreated or 21-day doxycycline–treated, or six 14-day doxycycline–treated, mice per group. BrdU⁺ cells were counted with National Institutes of Health image software (National Institutes of Health, Bethesda, MD). Quantitative analysis of glucagon- and insulin-stained pancreata was performed with Axiovision 4.2 image analysis software (Carl Zeiss). Briefly, 4–8 nonoverlapping fields containing islets (coimmunostained with separate flourophores for insulin and glucagon and counterstained with DAPI) were imaged from pancreata of 2–3 animals for each treatment group. Insulin- and glucagon-stained areas were determined by performing identical semiautomated image analysis routines on each of the separate fluorescent image channels. Islet areas were determined by manually delineating the islets, and costained cells were counted manually on the merged immunofluorescent images.

Total RNA was isolated with RNAzol B (Tel-Test, Friendswood, TX) from the pancreata of age- and sex-matched Pdx1^tTA/tTA;Tg^Pdx1 mice either untreated (n = 3) or treated (n = 4) with doxycyline for 14 days. Equal amounts of RNA were pooled to create +dox and −dox groups for comparison on Affymetrix gene chip arrays (Affymetrix, Santa Clara, CA). A total of 30 μg total RNA was used to generate fragmented cRNA as per Affymetrix gene chip protocols. cRNA quality was assessed by hybridization of 5 μg fragmented cRNA to Affymetrix Test 3 arrays before hybridizing 20 μg fragmented cRNA to Affymetrix MGU74Av2 gene chip arrays. Chips were processed and scanned as per protocols in the Affymetrix gene chip expression analysis technical manual. Analysis of the probe data were performed as follows. The perfect match probe intensities were background corrected and normalized across arrays and then summarized for each probe set using robust multichip analysis (26,27). Once expression levels had been calculated, differentially expressed genes were ranked and values expressed as fold change.

Quantitative real-time RT-PCR was performed on the Roche Lightcycler (Roche Diagnostics). For each gene to be assayed, a template was amplified by PCR and subcloned into a pGEM-T vector (Promega, Madison, WI). The templates were used to generate standard curves for quantitative determination of mRNA expression level in the individual RNAs described above. Expression levels were corrected relative to the expression of the of β-actin gene. Sequences of oligonucleotide primers used are available on request.

Islet area was determined for pancreata of mice that were either untreated or treated with doxycycline for 14 days (four mice per group). Three hematoxylin and eosin–stained sections from each pancreas were sampled, each section separated by ∼250–300 μm. Nonoverlapping 40× magnification fields were captured with an Axiocam digital camera (Carl Zeiss), and the total tissue and islet area per field was determined using Axiovision 4.2 image analysis software (Carl Zeiss). All sections were scored blind with respect to treatment. Statistical significance was determined using a Mann-Whitney U test (GraphPad Software).

Two genetic modifications were required to achieve conditional repression of Pdx1 (25) (Fig. 1). First, the endogenous Pdx1 gene was inactivated by replacing its coding region with that of the tetracycline transactivator (tTA off) by homologous recombination. As a result, transcription of the tTA gene is directed by the endogenous Pdx1 transcriptional regulatory sequences. Mice with one inactivated Pdx1 allele develop normally (1–3) and can be maintained in the heterozygous state. Second, mice bearing a tTA-responsive Pdx1 gene were created by introducing a bicistronic transgene containing a Pdx1 mini-gene and an enhanced green fluorescence protein reporter gene under the control of a tTA-responsive promoter. When the two mouse lines were crossed, some of the progeny contained both the knocked-in tTA gene and the tTA-responsive transgene. Mice homozygous for the Pdx-tTA–off knock-in allele are dependent on tTA-mediated transcription of the transgene for expression of Pdx1. In this system, mice with the Pdx1^tTA/tTA;Tg^Pdx1 genotype develop normally because transgenic Pdx1 expression is driven, indirectly, by tTA expressed off the native Pdx1 promoter. Administration of doxycycline prevents the tTA from binding to the tTA-responsive promoter, thereby deactivating transcription of the transgene and depleting Pdx1 (Fig. 1). We previously reported (25) that Pdx1^tTA/tTA;Tg^Pdx1 mice survive to adulthood and maintain normal glucose homeostasis.

Adult mice were treated with doxycycline for 0, 7, 14, or 21 days, and β-cell function was assessed weekly by intraperitoneal glucose tolerance tests (Fig. 2A). Untreated Pdx1^tTA/tTA;Tg^Pdx1 mice had normal fasting blood glucose levels and responded normally to a glucose challenge, recovering to basal levels within 3 h. However, within 7 days of starting doxycycline, the early glucose response of Pdx1^tTA/tTA;Tg^Pdx1 mice was increased and their ability to recover after the glucose challenge was impaired, similar to that of the Pdx1^tTA/+ mice. By 14 days of doxycycline treatment, Pdx1^tTA/tTA;Tg^Pdx1 mice had diabetes (Fig. 2A).

Although pancreas histology of Pdx1^tTA/tTA;Tg^Pdx1 mice treated with doxycycline for 14 days was grossly normal, the islets appeared to be smaller (Fig. 2B). Morphometric analysis confirmed a significant decrease (P < 0.05) in the islet area of treated animals (Fig. 2D). Transferase-mediated dUTP nick-end labeling staining was performed to determine whether impaired β-cell function in doxycycline-treated Pdx1^tTA/tTA;Tg^Pdx1 mice was associated with β-cell apoptosis. Transferase-mediated dUTP nick-end labeling–positive cells were detected at a similar low frequency in both control and doxycycline-treated mice, predominantly in the exocrine pancreas, labeled cells being rarely detected in the islets of either treated or untreated mice (data not shown). Staining for filtrin (28), a member of the nephrin family (29) whose expression is restricted to β-cells (28,30), revealed that β-cells, although lacking insulin (see below), were still present after 14 (Fig. 2C) and 28 days of doxycycline treatment. Quantitative RT-PCR demonstrated a decrease in Pdx1 and insulin transcripts and a corresponding increase in glucagon transcripts in pancreata of doxycycline-treated mice (Fig. 2E). In doxycycline-treated mice, histological analysis of the pancreas revealed total absence of Pdx1 (Fig. 3A, panel f) and a marked reduction of insulin, with only a few cells in each islet staining strongly (Fig. 3A, panel f). Glucagon-positive cells were prominent, as well as cells costaining for glucagon and insulin (Fig. 3A, panel h). Expression of the glucose transporter Glut2 was undetectable in the islets of Pdx1-repressed mice (Fig. 3A, panel i). Amylase was restricted to the exocrine pancreas, although its staining appeared to be reduced (Fig. 3A, panel j), consistent with our previous finding that the level of amylase mRNA is reduced in Pdx1-repressed mice (25). Untreated Pdx1^tTA/tTA;Tg^Pdx1 mice (Fig. 3A, panels a–e) and control Pdx1^+/+;Tg^Pdx1 mice (Fig. 3A, panels k–o) had normal expression of Pdx1, insulin, somatostatin, glucagon, Glut2, and amylase.

Quantitative image analysis of the pancreata of doxycycline-treated mice revealed that the decrease in insulin staining corresponded with an increase in glucagon staining (Fig. 3B) and with an increase in the number of islet cells coexpressing insulin and glucagon (Fig. 3C), confirming the quantitative RT-PCR results. These observations suggest that Pdx1 normally supresses expression of the glucagon gene in β-cells.

A key feature of the tet-off system is the ability to reversibly control expression of the tTA-dependent transgene. Therefore, to demonstrate that derepression of the Pdx1 transgene would restore β-cell function, we withdrew doxycycline after 14 days administration to Pdx1^tTA/tTA;Tg^Pdx1 mice. Blood glucose levels of 8-h fasted mice were then measured every 7 days for 35 days. Blood glucose levels decreased significantly by 14 days and progressively until four of five mice were normoglycemic by 28 days, demonstrating that doxycycline-induced diabetes is reversible (Fig. 4A). Histological analysis revealed Pdx1 protein in some islet cells 7 days after doxycycline withdrawal, although insulin expression remained low (Fig. 4B). However, by 14 days, both Pdx1 and insulin were expressed homogeneously at pretreatment levels throughout the islets (Fig. 4B), and both the glucagon-positive islet area (Fig. 3B) and the number of insulin⁺/glucagon⁺ costained cells per islet (Figs. 3C and 4C) had also decreased.

Foci of duct proliferation comprising small ductules embedded in connective tissue, similar to those observed in other models of pancreas regeneration (10,31), were present in mice treated with doxycycline for 14 days and became more prominent the longer the duration of doxycycline administration (Fig. 5A, upper panel). Staining for proliferating cell nuclear antigen confirmed the presence of proliferating cells within these ductal-like structures (Fig. 5A, lower panel). To quantitate cellular proliferation, Pdx1^tTA/tTA;Tg^Pdx1 mice either untreated or treated with doxycycline for 14 or 21 days were pulsed with BrdU overnight before they were killed. Immunofluorescence analysis revealed increased nuclear labeling in the pancreata of Pdx1-repressed mice at both 14 (P < 0.0005) and 21 (P < 0.0001) days (Fig. 5B). BrdU-positive cells were more frequent within exocrine tissue, although they were also observed proximate to ducts, within foci of duct proliferation and in islets (Fig. 5B).

Following withdrawal of doxycycline, insulin⁺ cells were primarily detected in islets or smaller islet-like cell clusters (Fig. 5C). Pdx1 was observed at high levels in islets and islet-like cell clusters in insulin⁺ cells (Fig. 5C). In addition, insulin⁺ cells expressing Pdx1 were observed embedded in or closely associated with ducts (Fig. 5C). Pdx1⁺/insulin⁻ cells were also observed associated with ducts (Fig. 5C, panel b), but the Pdx1 staining in these cells was generally of lower intensity than in the insulin⁺ cells.

To further characterize the molecular events associated with Pdx1 repression and potentially identify novel Pdx1-regulated genes, pancreatic RNA was isolated from Pdx1^tTA/tTA;Tg^Pdx1 mice that had been treated with doxycycline for 14 days and from untreated mice, and gene expression profiles were compared on Affymetrix MGU74Av2 gene chip arrays. To minimize variation, the gene chip analyses were performed on pooled RNA samples from treated and untreated mice. The experiment was replicated with independent cohorts of mice, and the results obtained were essentially identical. Pdx1 repression was associated with upregulation of 97 transcripts and downregulation of 29 transcripts by at least 1.8-fold (Table 1). Gene ontology associations revealed that >50% of the upregulated genes were involved in metabolism or cell growth and/or maintenance (Fig. 6A). The results were validated by real-time RT-PCR of several genes in individual mouse RNA samples that had been pooled for the gene chip experiments (Figs. 2E and 6B). Despite variation between individual samples, expression levels determined by real-time RT-PCR correlated well with the fold change determined from the gene chip analysis of pooled RNA. Of the downregulated genes, the most notable were the insulin genes ins1 and ins2, validating the results of the gene chip strategy. Upregulated genes notably included several previously identified by differential gene expression screens in models of pancreas regeneration (32,33), in particular the regenerating islet–derived (Reg) family (Reg2, Reg3α, Reg3β, and Reg3γ) that plays a role in the maintenance of progenitor cell populations and endocrine differentiation via the notch signal transduction pathway and transducer of ErbB2.1 (Tob1) involved in the regulation of cell proliferation and transformation. Novel transcripts were also identified in the screen. One of these, 1810015C04Rik (Entrez Gene ID 66270) was confirmed by RT-PCR to be upregulated in all of the Pdx1-repressed mice (Fig. 6B) and found to encode a highly conserved protein of unknown function. It has been previously identified in a number of tissues, including embryonic pancreas (Unigene Mm.25311), suggesting a potential role for this gene in early pancreas development. RT-PCR analysis revealed expression of 1810015C04Rik in two pancreatic β-cell lines, Min1046 and βTC-3, as well as in kidney and liver but not in the α-cell line, αTC-1 (Fig. 6C).

Using genetically modified mice in which Pdx1 transcription can be reversibly repressed, we confirm the critical role of Pdx1 in maintaining β-cell function. Furthermore, we provide evidence that Pdx1 normally suppresses glucagon expression in β-cells and that repression of Pdx1 leads to enhanced proliferation of presumptive β-cell progenitors.

Mutations of Pdx1 leading to impaired β-cell function underlie maturity-onset diabetes of the young type 4 in humans, and several studies (5–7,34,35) have demonstrated a dose-dependent requirement for Pdx1 in the maintenance of normal glucose homeostasis in mice. These studies initially examined pancreatic function in adult mice with one Pdx1 allele inactivated (Pdx1^+/−) and more recently used transgenic approaches to conditionally inactivate (5) or inducibly repress Pdx1 either with antisense Pdx1 (6) or Pdx1-specific ribozyme (7) strategies. However, these models are limited by their dependence on expression of molecules under the control of the insulin promoter, which is downregulated following repression of Pdx1, resulting in relatively mild impairment of glucose tolerance (6,7) or the late onset of diabetes in the case of RIP-Cre-Pdx1 mice (5). In our model, the coding sequence for the tet-off regulatory protein (tTA) is incorporated into the Pdx1 locus so that normal temporal and spatial regulation of the gene is maintained, thereby overcoming a problem inherent in many transgenic systems that utilize partial regulatory sequences out of their genomic context. The genetically modified mice are morphologically and functionally normal, yet doxycycline treatment of adult mice repressed Pdx1 expression and induced marked hyperglycemia associated with reduced expression of insulin and the glucose transporter, Glut2. These effects were reversible upon withdrawal of doxycycline. This is therefore a truly inducible gene knock-out model in which the role of the gene can be studied at any time during development or in mature animals. β-Cells were still present in Pdx1-repressed mice, and some contained residual insulin, which is not surprising as Pdx1 is only one of a number of transcription factors involved in the regulation of insulin transcription (36). Indeed, mice heterozygous for Pdx1 have normal pancreatic insulin content, although their ability to secrete insulin and clear glucose is impaired (5,37). We confirmed that the expression of several genes involved in the maintenance of glucose homeostasis, including insulin (38), Glut2 (39), and glucokinase (40), as is regulated by Pdx1 in the inducible antisense Pdx1 repression model of Lottmann et al. (6). Similarly, we were unable to detect Glut2 in the islets of Pdx1-repressed mice. It is likely that the loss of Glut2 leads to impaired glucose sensing and contributes to the diabetic phenotype of the Pdx1-repressed mice. In addition, we observed a reduction in islet area in Pdx1-repressed mice, indicating loss of the ability to maintain β-cell mass in the absence of Pdx1. It has been recently reported that haploinsufficiency of Pdx1 may be responsible for increased apoptosis of islet cells in Pdx1^+/− mice (41). We did not find evidence for increased β-cell apoptosis in Pdx1-repressed mice. Nevertheless, as apoptosis is a rapid and transient event, it is possible that a population of apoptosis-sensitive β-cells may have been ablated before the 7-day time point at which our first observations were made.

In addition to the progressive loss of insulin expression in β-cells during Pdx1 repression, we observed a concomitant increase in glucagon expression and in the number of cells expressing both insulin and glucagon in islets. These observations are consistent with studies demonstrating that glucagon expression is repressed by Pdx1 in β-cell lines (42,43) and that the α-cell mass is increased in β-cell–specific Pdx1 knock-down mouse models (5,7). Similarly, ectopic expression of Pdx1 in α-cell lines has been shown to repress glucagon gene expression (44), although this probably requires cell line–specific cofactors (45).

We performed microarray analysis to characterize the response to Pdx1 repression. A number of upregulated genes were involved in metabolism, consistent with studies (37,46) in which defects in mitochondrial metabolism have been linked to diminished expression of Pdx1. Several of the most differentially upregulated genes (Reg2, Reg3α, Reg3β, and Reg3γ) were members of the regenerating islet-derived gene family originally identified in models of β-cell regeneration (47–49). In addition, several other upregulated genes, including Numb and Tob, are likely to be involved in islet cell development. Numb is a modulator of the notch signaling pathway involved in the maintenance of embryonic neural progenitor cells (32). It is also expressed in early pancreatic progenitors and mature β-cells and has been implicated in the differentiation of β-cells from duct cells (33). The epidermal growth factor receptor ErbB2 is expressed in the developing pancreas and is upregulated in a model of duct-cell proliferation and islet neogenesis (50). Tob is an ErbB2 interacting protein implicated in the regulation of cell proliferation and differentiation (51). We also identified a novel highly conserved gene, 1810015C04Rik, which is expressed in the developing pancreas and mature β- but not α-cells. Further investigation is required to determine the role of 1810015C04Rik and other novel genes identified in this screen in β-cell viability and regeneration. In addition, it will also be important to determine whether these changes in gene expression are due to Pdx1 repression per se or are secondary to impaired β-cell function and hyperglycemia. Nevertheless, the initiation of a regenerative program by Pdx1 repression is a novel observation and a basis for further understanding β-cell neogenesis in the adult pancreas.

The adult pancreas has the capacity to increase its β-cell mass during times of increased physiological need or in response to injury. However, the nature of putative β-cell progenitors is a subject of ongoing debate. Numerous studies (9–11) support the existence of duct-associated progenitors in models of pancreatic injury, whereas a more recent lineage-tracing study (13) indicates that β-cell progenitors derive from β-cells themselves in islets. These findings are not mutually exclusive and may reflect different means to an end, depending on circumstances. Our model of reversible Pdx1 repression reveals a potential duct-associated progenitor cell population. First, during Pdx1 repression, we observed focal regions of duct proliferation, recognized as a hallmark of pancreas regeneration (10,31). Second, we observed an increase in cell replication, proportional to the duration of Pdx1 repression, primarily in the exocrine pancreas but also associated with duct-like structures as well as in islets. Third, we found that genes known to be associated with β-cell regeneration were upregulated during Pdx1 repression. Finally, after reversal of Pdx1 repression we observed cells that were Pdx1⁺/insulin⁺ or Pdx1⁺(low)/Ins⁻ closely associated with duct epithelium.

In rodent models of pancreas injury (20,23,24), Pdx1 expression has been associated with β-cell neogenesis. The proliferation we observed in response to repression of Pdx1 could be secondary to hyperglycemia rather than the loss of Pdx1 itself, as observed in other models of induced hyperglycemia (31,52). However, our findings show that proliferation of progenitors occurs in the absence of Pdx1 and that Pdx1 is required for their differentiation potentially into β-cells. Gu et al. (53) found that Pdx1 was rarely expressed in the ductal cells of the adult pancreas and were likely to be present only in differentiating cells in the process of delaminating from the ducts. In a separate study (R.J.M., unpublished data), it was shown that Pdx1 is not required for the continued proliferation of duct cells in the developing pancreas but is absolutely required for the differentiation of duct-like cells of the embryonic epithelium into endocrine and exocrine lineages. As suggested by Waguri et al. (20), β-cells could arise from both intraislet and ductal progenitors, depending on the circumstance. During periods of acute hyperglycemia, it may be sufficient for β-cells to enter the cell cycle and thereby increase the β-cell mass. However, following pancreatic injury or chronic hyperglycemia, a more primitive regenerative program may recruit duct-associated β-cell progenitors. Possibly consistent with evidence from lineage tracing (13), we also observed occasional islet cells labeled with BrdU in Pdx1-repressed mice but were unable to determine to what extent these cells contribute to β-cell regeneration. It is worth noting that in Pdx1-repressed mice, filtrin staining revealed the presence of β-cells, but the islet area was decreased, suggesting that Pdx1 may be required for the maintenance of intraislet progenitors. An important caveat, however, is that we have not been able to discriminate between restoration of β-cell function in pre-existing β-cells and β-cell neogenesis.

Combining the conditional Pdx1 expression model with lineage tracing or cell-specific ablation models could further elucidate the origins of β-cell progenitors and contribute to the goal of renewing β-cells for the treatment of type 1 diabetes.

This work was supported by a Juvenile Diabetes Research Foundation (JDRF) Postdoctoral Fellowship (to A.M.H.) and by a partnership Program Grant from the JDRF and the National Health and Medical Research Foundation of Australia (to L.C.H.). R.J.M. was supported by National Institutes of Health Grant DK55266.

The authors thank Ken Simpson for technical assistance with microarray analysis and Fang-Xu Jiang for helpful discussions.