In Vivo Conditional Pax4 Overexpression in Mature Islet β-Cells Prevents Stress-Induced Hyperglycemia in Mice

The pIRES2-DsRedexpress (Clontech) vector was used for the generation of the inducible Pax4 or the mutant variant Pax4R129W cDNA cassette. The final construct contained the tetracycline responsive promoter, the rabbit β-globin intron followed by the Pax4 or Pax4R129W coding sequence. A myc-epitope and polyhistidine tag were added for detection purposes. DsRedexpress was included in the constructs in order to follow Pax4 induction using noninvasive in vivo imaging. Pax transgenic animals were crossed to RIPrtTA mice to generate double transgenic descendants with conditional expression of Pax4 or Pax4R129W specifically in β-cells. Induction of transgene expression was achieved by providing 1 g/L of doxycycline (Sigma-Aldrich) in the drinking water. The GLUT2 knockout mouse has been described elsewhere (15). The Geneva Veterinary Cantonal Office and the CABIMER animal committee approved all experiments.

Animals fasted overnight or for 4 h and were injected intraperitoneally with 2 g glucose/kg body wt or with 1 unit insulin/kg body wt Actrapid (Novo Nordisk Pharma). Blood was collected from tail vein, and glucose levels were measured with a Precision Xceed glucometer. Plasma insulin levels were measured at 0 and 15 min using a mouse insulin ELISA kit (Mercodia, Uppsala, Sweden). Pancreatic insulin content was determined using an insulin enzyme immunoassay kit (SPI-BIO; Bertin Pharma Biotech Division, Brunschwig, Basel, Switzerland).

STZ was prepared in 0.1 mmol/L sodium citrate at pH 4.5 and administered by intraperitoneal injection (200 mg/kg body wt). Blood glucose was determined before STZ injection and then three times per week.

Pancreatic islets were isolated by collagenase (Roche, Switzerland) digestion, handpicked and maintained in 11.1 mmol/L glucose/RPMI-1640 (Invitrogen, Switzerland) supplemented with 10% FCS (Brunschwig), 100 units/mL penicillin, and 100 mg/mL streptomycin (Sigma, St-Gallen, Switzerland).

Total RNA from islets was extracted using the RNeasy Micro Kit (Qiagen), and quantitative real time PCRs (Q-RT-PCRs) were performed as previously described (16). Primer sequences were designed using the Primer Express Software (Applera Europe) and can be obtained upon request.

Transient transfections were performed as previously described (11). The luciferase reporter contructs, pFOX-Luc and pFOX-MafAR3-Luc (region 3 of the MafA promoter specifying β-cell expression), were provided by Dr. R. Stein (Vanderbilt University, Nashville, TN).

Islets were isolated from transgenic animals that had received 1 g/L doxycycline in drinking water for 1 month. Islets were cultured for 48 h in the presence of 1 ug/L doxycycline prior to the addition of 0.25 ng/mL interleukin (IL)-1β, 9.1 ng/mL tumor necrosis factor-α, and 100 ng/mL γ-interferon (R&D Systems). Apoptosis was evaluated 24 h posttreatment using the Cell Death ELISA Plus kit (Roche). Alternatively, cell death was estimated by cytochrome C release. Dispersed islets were concentrated by cytospin onto glass slides, and immunofluorescence was performed using anticytochrome C serum (BD Biosciences) and antiserum for the mitochondrial marker TOM 20 (Santa Cruz). Nitrite production in culture media was determined using the Griess reaction (Sigma).

Mice were injected with 1 mL BrdU solution/100 g body wt (Roche); 24 h later, they were killed. Proliferation was detected by immunohistochemistry on pancreatic sections.

For paraffin sections, pancreata were dissected and fixed in 4% paraformaldehyde. Dehydration, embedding, and sectioning were performed at either the Geneva or CABIMER Histology platform. Sections were rehydrated in ethanol and blocked in PBS containing 1% BSA and 0.1% Tween. The following primary polyclonal antibodies were used: guinea pig anti-insulin, mouse anti-glucagon, rabbit anti-pdx1 (1:200; provided by C. Wright, Vanderbilt University), rabbit anti-GLUT2 (1:200), and rabbit anti-MafA (1:200; Bethyl). A mouse monoclonal BrdU was also used. The following secondary antibodies were added: Alexa Fluor 568 goat anti-guinea pig, Alexa Fluor 488 goat anti-mouse, and Alexa Fluor 488 goat anti-rabbit. For cryostat sections, pancreata were frozen in liquid nitrogen prior to sectioning and mounting. Sections were fixed in 4% paraformaldehyde, and immunodetections were performed with the following primary polyclonal antibodies: rabbit anti-Pax4 (provided by Dr. Sosa-Pineda) and guinea pig anti-insulin. Secondary antibodies used were as follows: Alexa Fluor 568 goat anti-rabbit and Alexa Fluor 488 goat anti–guinea pig.

Islets were harvested in lysis buffer (0.05 mol/L TRIS-HCl, pH 7.5, and 62.5 mmol/L EDTA), cells disrupted by sonication, and proteins resolved by 10% PAGE. Gels were processed for immunoblotting using standard procedures. The following primary polyclonal antibodies were used: rabbit anti-Pax4, goat DsRed (Santa Cruz), and rabbit TFIIE-α (Santa Cruz). A goat anti-rabbit IgG antiserum conjugated to horseradish peroxidase (GE Healthcare) was then added, and immunoreactive products were visualized by chemiluminescence (Pierce, Rockford, IL).

Results are expressed as means ± SEM. Statistical differences were estimated using the unpaired t test, ANOVA with Bonferroni post hoc test, or nonparametric Mann-Whitney test.

To study in vivo the role of Pax4 and the impact of a mutant variant associated with diabetes in β-cell plasticity, we generated inducible double transgenic mouse lines expressing RIP-rtTA along with either Pax4/DsRedexpress or Pax4R129W/DsRedexpress. For simplicity, double transgenic rtTA/Pax4 and rtTA/Pax4R129W animals are denoted as Pax4 and Pax4R129W, respectively, while control animals comprising single transgenic littermates are referred to as ST. Q-RT-PCR revealed that both Pax4 and Pax4R129W transcripts were expressed to similar levels exclusively in islets of doxycycline-treated transgenic Pax4 animals, whereas expression of the transgenes was not detected in either control ST mice or in double transgenics that had not been provided with doxycycline (Fig. 1A). Western blotting confirmed expression of Pax4 and DsRedexpress in islets of the doxycycline-treated animals (Fig. 1B). Immunohistochemistry using either Pax4 or c-myc antisera further substantiated specific expression of Pax4 in islets and revealed colocalization with insulin in most but not all β-cells, suggesting β-cell–restricted expression of the transgene (Fig. 1C and D).

In parallel, induction of DsRedexpress was monitored by noninvasive in vivo imaging in Pax4 transgenic animals. Fluorescence was detected in the abdominal region of anesthetized living animals treated with doxycycline for 48 h (Supplementary Fig. 1A). Imaging of extracted organs confirmed expression of DsRedexpress specifically in the pancreas of treated animals, while other organs exhibited background fluorescence (Supplementary Fig. 1B).

Pax4 transgenic animals that received doxycycline for 1 month exhibited normal glucose tolerance and insulin sensitivity (Fig. 1E and F). Consistent with this finding, glucose-induced insulin secretion was normal in islets isolated from Pax4-expressing mice compared with control ST islets (245 ± 56 vs. 246 ± 55 ng/mL). Similar results were obtained with Pax4R129W animals (data not shown). These results demonstrate that conditional expression of Pax4 can be specifically achieved in islet β-cells and that its forced expression does not alter glucose homeostasis in double transgenic animals.

To assess the protective role of Pax4 in vivo, doxycycline-treated Pax4 or ST animals were challenged with a single high dose of STZ to destroy β-cells. ST animals rapidly developed hyperglycemia while Pax4-overexpressing animals remained normoglycemic throughout the experiment (Fig. 2A). For confirmation that Pax4 was mediating the protection against STZ-induced hyperglycemia, doxycycline was removed on day 27 and animals in group A received, 14 days later, a second STZ injection. Group A animals developed hyperglycemia similar to that observed in ST animals, whereas Pax4 animals that did not receive a second STZ injection (group B) remained normoglycemic (Fig. 2A). In an independent set of experiments, animals continuously exposed to doxycycline and rechallenged with a second high dose of STZ did not develop hyperglycemia, confirming Pax4-dependent protection even after two high doses of STZ (Fig. 2B). Pax4R129W animals challenged with a single dose of STZ developed an intermediate level of hyperglycemia, indicating that the diabetes-associated mutant is less efficient in protecting mice against stress-induced hyperglycemia (Fig. 2C). Immunohistochemistry combined with morphometry analysis revealed a clear reduction in insulin⁺ cells, with a concomitant increase in glucagon⁺ cells in STZ-injected ST animals compared with what was observed in untreated mice (Fig. 2D and Supplementary Fig. 2A and B). In contrast, islet morphology and cell composition in Pax4-overexpressing animals challenged with either one or two doses of STZ were similar to those of islets from control animals (Fig. 2D and Supplementary Fig. 2A and B). Islets from Pax4R129W mice exhibited less insulin staining, suggesting a partial loss of β-cells correlating with intermediate hyperglycemia (Fig. 2D). These results show that overexpression of Pax4 protects β-cells against multiple STZ challenges and prevents development of hyperglycemia, whereas the mutant is less effective.

For further assessment of the protective action of Pax4, islets isolated from either Pax4 or Pax4R129W transgenic animals were treated with cytokines. Pax4 islets were protected against cytokine-induced cell death compared with ST control animals that displayed a threefold increase in cell death in the presence of cytokines (Fig. 3A). Islets isolated from Pax4R129W animals exhibited a similar degree of apoptosis compared with cytokine-treated ST animals (Fig. 3A).

Exogenous IL-1β was shown to induce endogenous levels of IL-1β and to lead to human islet β-cell death (17). Interestingly, IL-1β transcript levels were suppressed by 50% in islets isolated from Pax4-overexpressing animals. In contrast, Pax4R129W mutant mice exhibited a twofold increase in IL-1β transcript levels compared with control untreated ST islets (Fig. 3B). A parallel modulation in transcript levels of the IL-1β antagonist was observed in both transgenic animal models, whereas expression levels of the IL-1β receptor remained unaltered (Fig. 3B).

A hallmark of cytokine-mediated cell death is the activation of the nuclear factor (NF)-κB pathway, which leads to enhanced NOS2 and COX2 expression and the downstream increase in nitric oxide (NO) production, ultimately resulting in cytochrome C release from mitochondria (18). Consistent with this model, islets isolated from Pax4 mice and exposed to cytokines in the absence of doxycycline exhibited a 2.5-fold increase in cytochrome C release. In contrast, cytochrome C release was abrogated in doxycycline-treated Pax4 islets, whereas Pax4R129W-overexpressing islets displayed a threefold increase compared with noncytokine-treated islets (Fig. 4A and B). Interestingly, NOS2 transcript levels were only significantly elevated in Pax4R129W islets, whereas COX2 levels were augmented in both mutant and wild-type Pax4 islets (Fig. 4C). Despite differences in NOS2 expression, both Pax4 and Pax4R129W islets presented a significant increase in cytokine-induced nitrite production, an indicator of NO production (Fig. 4D). However, Pax4 islets displayed a trend toward a small decrease in nitrite production compared with non–doxycycline-treated islets.

To further delineate the mechanism by which Pax4 protects β-cells from either STZ- or cytokine-induced apoptosis, we determined mRNA expression levels of key β-cell markers and those of antiapoptotic genes in islets of Pax4- and Pax4R129W-overexpressing animals (Fig. 5A). Transcript levels of Pdx1 and Pax6 were not altered in Pax4- or Pax4R129W-overexpressing islets, whereas MafA, GLUT2, and insulin levels were decreased in both animal models. Immunohistochemistry confirmed decreased MafA and GLUT2 protein levels in islets of Pax4-expressing mice (Fig. 5B and C). Bcl-2 but not Bcl-xL expression levels were increased threefold specifically in Pax4 islets compared with those in control animals without doxycycline administration (Fig. 5A).

Because insulin and GLUT2 are target genes of MafA (19,20) and because a Pax4 consensus sequence was predicted in region 3 of the MafA promoter (21), we investigated whether Pax4 could directly regulate MafA transcription. Pax4 and Pax4R129W dose dependently inhibited a luciferase reporter construct driven by region 3 of the MafA promoter, reaching 50% inhibition at the highest dose (Fig. 5D). Repression was specific; Pax4 was unable to inhibit the control pFOX-Luc construct. However, Pax4 failed to interact with the putative binding site identified in this MafA region (MafAR3Pax4), whereas it displayed specific binding to the G3 element of the glucagon promoter (data not shown) (22).

STZ is selectively taken up by β-cells through GLUT2 (23). Because GLUT2 levels were decreased in Pax4-overexpressing islets, we determined whether such inhibition would be sufficient to protect β-cells against STZ. To this end, the GLUT2 heterozygous mouse model was used in order to directly measure the impact of the glucose transporter in STZ-mediated cell death. Transcript levels of the transporter were reduced by 60% in heterozygous GLUT2 mice (Fig. 6A). Similarly to Pax4 mice, these animals were refractory to STZ-induced β-cell apoptosis, whereas control wild-type littermates were sensitive to the toxic agent (Fig. 6B). These results suggest that the protective effect of Pax4 against STZ is potentially partially conveyed by reduced GLUT2 expression. However, despite similarly lower levels of GLUT2, Pax4R129W mice are only partially protected against STZ, suggesting that alternative detrimental processes are associated with the mutant phenotype.

Adenoviral-mediated overexpression of Pax4 in cultured rat islets was shown to hamper insulin secretion (11). Although this effect was not observed in Pax4 transgenic animals treated with doxycycline for 1 month, plasma insulin levels in response to glucose were greatly reduced in animals treated for 4 months (Fig. 7A). This was accompanied by glucose intolerance in these animals (Fig. 7B). Furthermore, morphological analysis of pancreas isolated from Pax4-expressing mice revealed altered islet architecture typified by an increased number of enlarged intercellular spaces, reminiscent of capillaries detected in control islets (Supplementary Fig. 3). To determine whether this effect was reversible, we removed doxycycline from the drinking water of 4-month–treated animals to repress Pax4 expression and reevaluated plasma insulin levels 1 month later. Consistent with our premise, the rise in plasma insulin levels in response to glucose was rescued in transgenic animals subsequent to removal of doxycycline (Fig. 7C). Total pancreatic insulin content was significantly increased in Pax4 mice after removal of doxycycline (Fig. 7D).

The increase in insulin content observed after doxycycline removal (Fig. 7D) suggests that long-term expression of Pax4 may induce β-cell replication. Consistent with this premise, islets isolated from Pax4 mice that underwent 4-month but not 1-month doxycycline treatment showed reduced glucose-induced plasma insulin levels (Fig. 7A) and exhibited increased BrdU labeling of β-cells compared with control ST islets (Fig. 8A). This was accompanied by an overall nonsignificant increase in the proportion of β-cells (Fig. 8B). A BrdU⁺/insulin⁻ cell subpopulation was also detected specifically in Pax4-overexpressing islets (Fig. 8A). These cells were also negative for glucagon and somatostatin staining. Given that Pax4 overexpression impedes insulin transcription (Fig. 5A), we reasoned that these cells were likely β-cells with low levels of insulin. Therefore, Pdx1 was used as an alternative β-cell marker. Most BrdU⁺ cells were also Pdx1⁺, and the total percentage of these cells was equivalent to the sum of the insulin⁺ and insulin⁻ cell subpopulations that were BrdU⁺ (compare Fig. 8A and C). Despite a significant increase in overall proliferation (Fig. 8C), we did not detect an increase in the total population of Pdx1⁺ cells (Fig. 8D) or an increase in islet mass due to a low number of proliferating cells. Consistent with increased long-term cell replication, c-myc and Cdk4 but not Cdk2 or -6 were increased in Pax4-overexpressing islets (Fig. 8E).

Our study highlights the physiological role of Pax4 in vivo as a regulator of mature islet β-cell survival and proliferation. Pax4 prevented development of hyperglycemia in animals exposed to environmental insults such as repeated STZ challenges and protected islets against cytokine-mediated apoptosis. The diabetes-associated R129W mutation was less efficient in protecting animals against STZ-induced hyperglycemia. The detrimental effect of the mutation was only revealed in the presence of harmful environmental cues highlighting the interplay between genetic determinants and the environment in the development of hyperglycemia (24). Thus, in response to physiological demand or pathophysiological situations, Pax4 expression may be transiently induced to protect and to compensate β-cell mass, whereas the mutant variant would be unable to achieve this, resulting in the gradual loss of β-cells and ultimately diabetes. The regulation of Pax4 expression by environmental factors provides a likely explanation for the failure of GWAS to identify Pax4 as a diabetic susceptibility gene as well as for discrepancies in different studies to validate the association of Pax4 polymorphisms with type 2 diabetes in various ethnic populations (7).

One mechanism by which Pax4, but not Pax4R129W, protects β-cells against STZ- and cytokine-induced cell death may, in part, be connected to the regulation of selective target genes of the NF-κB signaling cascade (25). Indeed, we found that transcript levels of IL-1β, the predominant cytokine implicated in the activation of the NF-κB pathway, were decreased in Pax4-expressing islets, whereas they were increased in Pax4R129W islets. Consistent with NF-κB activation, transcript levels of its target gene nos2 were selectively higher in mutant animals, whereas levels of the target gene cox2 were increased in both Pax4 and Pax4R129W. NOS2 is involved in NO production and downstream stimulation of apoptosis (25), whereas the COX2-generated metabolite prostaglandine E₂ was found to be antiapoptotic (26). As IL-1β was shown to autostimulate expression levels of endogenous IL-1β (17) and to induce β-cell death in human islets (27), Pax4-mediated inhibition of IL-1β transcription with a concomitant increase in COX2 levels may confer increased resistance to apoptosis. In contrast, elevated basal levels of IL-1β and NOS2 in Pax4R129W-overexpressing mice may sensitize islet cells to NO-induced apoptosis. Furthermore, because NO can also be released from islets exposed to STZ (28), we suggest that Pax4R129W mice are more prone to cell death than GLUT2 heterozygous mice because of an increased predisposition to NO-mediated apoptosis. Interestingly, the levels of the IL-1β antagonist, IL-1Ra, were also increased in mutant islets, suggesting restoration of an IL-1β–to–IL-1Ra ratio that prevents activation of the apoptotic program (29). Consistent with this model, several putative Pax4 binding sites were predicted in both the il-1β and nos2 gene promoters.

Despite differences in Il-1β levels, both Pax4- and Pax4R129W–expressing islets exhibited similar levels of NO production upon exposure to cytokines. This indicates that the NF-κB signaling pathway can still be fully activated upon strong exogenous stimulation and that the protection conveyed by Pax4 is relayed through additional gene product that blocks the NF-κB apoptotic pathway. One such candidate gene could be Bcl-2, which regulates the mitochondrial response to proapoptotic signals, preventing the release of cytochrome C and the subsequent activation of the cell death program (30). Indeed, we found that Pax4- but not Pax4R129W-overexpressing islets exhibited increased expression of Bcl-2 with blunted cytokine-mediated cytochrome C translocation. Corroborating our results, Bcl-2 overexpression was shown to protect mouse and human islets against cytokine-induced apoptosis (31,32). In contrast to our previous study in rat islets overexpressing Pax4 (11), the increased expression of Bcl-xL in mouse islets was not detected, possibly as a result of species differences.

Overexpression of either Pax4 or Pax4R129W markedly perturbed expression of MafA and two of its target genes, glut2 and insulin, which are key markers of mature β-cells. Although both Pax4 proteins suppressed the activity of a luciferase construct harboring the MafA gene promoter region 3, we could not detect Pax4 binding to a putative binding site within this region. Preliminary ChIP assay results revealed no convincing interaction of Pax4 with region 3, suggesting an indirect effect of Pax4 on MafA expression. However, we have previously demonstrated that Pax4 impairs glucagon and insulin gene transcriptions through direct protein interaction with Pax6 (22). Given that Pax6 was recently shown to bind to the mafa gene promoter and to regulate transcription (33), we speculate that Pax4 may be inhibiting MafA expression by a mechanism similar to the one delineated for glucagon and insulin. Inhibition through protein-protein interaction, and not direct DNA binding activity, would explain the inhibitory effect observed with the weaker DNA-interacting mutant variant.

Long-term overexpression of Pax4 resulted in a blunted glucose-induced rise in plasma insulin levels—an effect reversible upon inhibition of Pax4. A recent study also reported impaired glucose-induced insulin secretion in an animal model in which Pax4 was constitutively overexpressed in α-cells. These cells were converted to β-cells that remained functional for up to 4 weeks before an unexpected decline in insulin secretion occurred. Animals ultimately died of hyperglycemia (8). We would therefore like to propose that Pax4-dependent impaired insulin secretion is mediated by the long-term repression of the mature β-cell marker MafA and insulin.

Aberrant expression of Pax genes were shown to convert cells to a less differentiated state and to promote self-renewal and increasing survival (34). Consistent with this premise, a subpopulation of replicating Pdx1⁺/insulin⁻ cells was identified in Pax4-overexpressing islets. These cells were also negative for both glucagon and somatostatin, suggesting that they likely represent insulin-deprived β-cells resulting from repression of MafA and insulin transcription by Pax4. A similar β-cell subpopulation was recently characterized in vitro (35). These cells displayed impaired glucose-induced insulin secretion and were prone to proliferation in response to activin A (36). More importantly, MafA and GLUT2 transcript levels were lower in this subpopulation, whereas Pax4 levels were higher. Taken together, these results suggest that Pax4 under physiological conditions is most likely expressed in a subpopulation of β-like cells prone to proliferation. Substantiating this hypothesis, Cdk4 and c-myc transcript levels were increased in Pax4-overexpressing islets. Adenovirus-mediated Cdk4 expression was shown to increase mouse islet cell proliferation (37), whereas its repression causes insulin deficiency and hyperglycemia (38). Similarly, β-cell–specific overexpression of c-myc caused downregulation of insulin gene expression with a concomitant increase in proliferation and apoptosis (39). That a fraction of islet β-cells responds to proliferation would provide a mechanism by which a pool of mature cells can rapidly undergo regulated replication to compensate for lost cells while quiescent cells retain the ability to regulate blood glucose.

In summary, in vivo overexpression of Pax4 reveals an important role of the transcription factor in islet β-cell proliferation and survival. The detrimental impact of Pax4 mutations is only revealed upon environmental stresses, which reinforces the critical role of gene-environment interaction in the development of diabetes.

This work was funded by the Swiss National Science Foundation (grants 310030-119763 to B.R.G.; 310000-116750/1 to C.B.W. and B.R.G.; 310000-109402, 310000-122423, and CR32I3-129987 to P.M.; and 3100A0103867 NCCR “Frontiers-in-Genetics” to P.L.H.), the Juvenile Diabetes Research Foundation (grants 7-2005-1158 to C.B.W. and B.R.G.; 40-2011-11 to P.M.; and 26-2007-928 and 26-2008-640 to P.L.H.), the EU (FP-7, BETAIMAGE 222980, IMI IMIDIA, and C2008-T7 to P.M. and FP-6, β Cell Therapy Consortium to P.L.H.), the National Institutes of Health/National Institutes of Diabetes and Digestive and Kidney Diseases (the Beta Cell Biology Consortium to P.L.H.), and the Foundation Progreso y Salud, Junta de Andalucia (to B.R.G.). K.H.H.H. was supported by a GeMet/Novo Nordisk fellowship. No other potential conflicts of interest relevant to this article were reported.

K.H.H.H. generated the bulk of the results, conceived and designed the experiments, and wrote the manuscript. P.I.L. contributed to research data and reviewed the manuscript. T.B. provided advice, expertise, mice, or reagents. C.M.J.M., D.A., and J.V.O. contributed to research data. M.C. provided advice, expertise, mice, or reagents. F.T. and A.G. contributed to research data. B.T. provided advice, expertise, mice, or reagents. P.L.H. provided advice, expertise, mice, or reagents; reviewed the manuscript; and contributed to research data. P.M. and C.B.W. provided advice, expertise, mice, or reagents and reviewed the manuscript. B.R.G. conceived and designed the experiments and wrote the manuscript.

The authors acknowledge the technical assistance of D. Brighouse, E. Husi, A. Charollais, J. Cancela, and D. Caille, University of Geneva. The authors also thank Dr. Paloma Dominguez from CABIMER for her expert advice with confocal microscopy.