A Conditional Model Reveals That Induction of Hepatocyte Nuclear Factor-1α in Hnf1α-Null Mutant β-Cells Can Activate Silenced Genes Postnatally, Whereas Overexpression Is Deleterious

A 3.17-kb fragment encoding mouse HNF1α was subcloned into the multiple cloning site of pBI-L (Clontech), downstream of the CMV (cytomegalovirus) minimal promoter and the bacterial Tet operator (TetO) enhancer sequence (Fig. 1A). Vectorless linearized gel-purified DNA was used for pronuclear injection in fertilized oocytes to generate Tg^tet-Hnf1α transgenic mice. Two independent lines were generated, one of which resulted in consistently detectable expression and was chosen for analysis. To generate Tg^RIP-tTA transgenic mice, a 641-bp rat insulin II promoter fragment was inserted in pTet-tTA upstream of a tetracycline-dependent transactivator (tTA) cDNA sequence (17), followed by a SV40 small T antigen intron sequence. Linearized plasmid was microinjected in pronuclei as described (18).

Tg^tet-Hnf1α and Tg^RIP-tTA mice were intercrossed with Hnf1α^+/− mice (6,19) to generate either dizygous Tg^tet-Hnf1α or Tg^RIP-tTA Hnf1α^+/− progeny. These were used to generate Tg^{tet-Hnf1α/RIP-tTA} double-hemizygous transgenics with Hnf1α^+/+, Hnf1α^+/−, or Hnf1α^−/− genotypes. All experiments were carried out in mice with a mixed-strain (CD1/C57BL/6) background, using littermates as controls. Expression of the Tg^tet-Hnf1α transgene was controlled by feeding mice pellets containing tetracycline at 2.3 mg/g (Harlan). Blood glucose and weight did not significantly differ between nontransgenic littermates, hemizygous and dizygous Tg^tet-Hnf1α, Tg^RIP-tTA, and tetracycline-fed Tg^{tet-Hnf1α/RIP-tTA} double transgenics (not shown).

Genotyping of Tg^tet-Hnf1α was carried out with oligonucleotides 5′-CGGGGATCCTCTAGTCAGC-3′ and 5′-CTCTTTGCTCAGGCCAGACT-3′ and, for Tg^RIP-tTA, with 5′-ATTTGAGGGACGCTGTGGGCTCTT-3′ and 5′-ACTTCAATGGCTAAGGCGTC-3′. Hnf1α genotyping was performed as previously described (6). All studies were approved by the institutional animal care and use committee.

Neonatal 7-day-old (P7) and adult mice were fasted 5 h before blood glucose measurements. Then, 6- to 8-week-old mice were injected with glucose (2 mg/g i.p.), and venous blood was obtained from the tail at 0, 15, 60, and 120 min. Blood glucose was measured with an Accutrend sensor (Roche) in at least six mice per age and genotype.

For islet isolation, 6- to 8-week-old Hnf1α^+/+, Hnf1α^−/− Tg^{tet-Hnf1α/RIP-tTA}, and Hnf1α^+/+ Tg^{tet-Hnf1α/RIP-tTA} mice fed tetracycline since conception were used (6). The islets were placed in culture in RPMI (Gibco) supplemented with 10% tetracycline-free FCS (Attendbio), 500 ng/ml tetracycline (Sigma), and penicillin/streptomycin at 1:100 dilution (Gibco) at 37°C, 5% CO₂, for 24 h and then cultured for 60 h in the same medium with or without 500 ng/ml tetracycline.

Total RNA was extracted from adult liver embryonic day (E) 18.5 dissected pancreas or from 50 to 100 cultured islets, using Trizol (Invitrogen), and analyzed with a 2100 Bioanalyzer (Agilent). Total RNA (1 μg for liver and embryonic pancreas, 100 ng for islets) was reverse transcribed as previously described (6). Real-time PCR was carried out with an iCycler (BioRAD), using SybrGreen detection. Relative quantitation was carried out by generating a standard curve by serial dilution of one sample and then extrapolating threshold cycle (Ct) values of all test samples assessed in duplicate. All results were normalized with values obtained with control G9a oligonucleotides. Oligonucleotide sequences were designed to span an intron (supplemental Table, which is detailed in the online appendix [available at http://diabetes.diabetesjournals.org]).

Immunofluorescence in paraffin-embedded tissues was performed as previously described (6,12). Primary antibody dilutions were mouse anti-Ki67 (1:50 dilution; BD Pharmigen), rabbit anti-Ki67 (1:700 dilution; Novacastra), mouse anti-p27 (1:300 dilution; Transduction Laboratories), rabbit anti–activated caspase 3 (1:20 dilution; Abcam), rabbit anti-HNF1α (1:1,000 dilution) (8), mouse anti-HNF1α (1:50 dilution; Transduction Laboratories), rabbit anti-glucagon (1:200 dilution; Dako), goat anti-ghrelin (C-18, 1:3,000 dilution; Santa Cruz Biotechnology), rabbit anti-GLUT2 (1:200 dilution; Bernard Thorens, University of Lausanne), rabbit anti-Idx1 (Pdx1, 1:1,000 dilution; Joel Habener, Massachusetts General Hospital), rabbit anti–Nkx 6.1 (1:1,000 dilution; Ole Madsen, Hagedorn Research Institute), mouse anti-Pax6 (1:20 dilution; Developmental Hybridoma Bank), rabbit anti-Isl1 (1:20 dilution; Developmental Hybridoma Bank), goat anti-Sp1 (1:200 dilution; Santa Cruz Biotechnology), mouse anti–E-cadherin (1:200 dilution; Transduction Laboratories), and guinea pig anti–insulin II (1:2000 dilution; Chris Van Schravendijk, Vrije Universiteit Brussel, Brussels, Belgium). Donkey secondary antibodies conjugated to Cy2, Cy3, and AMCA (7-amino-4-methylcoumarin-3-acetic acid) were obtained from Jackson Laboratory and used at 1:200 dilution.

Three nonserial sections of paraffin-embedded pancreas from three E15.5, E18.5, or P7 untreated Hnf1α^+/+ Tg^{tet-Hnf1α/RIP-tTA} and control mice were stained for insulin and glucagon, and the number of cells per area of pancreatic tissue was calculated. All values correspond to the analysis of 300–1,000 endocrine cells per genotype.

To compare rates of apoptosis, dual stainings for insulin and activated caspase 3 were carried out in two nonserial sections from two E18.5 control and untreated Hnf1α^+/+ Tg^{tet-Hnf1α/RIP-tTA} pancreas. Activated caspase 3–positive cells were counted if they were found to stain for insulin or were contained within an insulin-positive cell cluster.

For β-cell cycle activity, untreated Hnf1α^+/+ Tg^{tet-Hnf1α/RIP-tTA} and control mice were costained for Ki67 or p27 and insulin or HNF1α. For each age and genotype, three nonserial sections of 3–6 specimens were analyzed to calculate the fraction of insulin-positive and HNF1α-overexpressing cells that stained for Ki67 or p27.

Statistical significance was determined by χ² or unpaired Student’s t test.

To achieve conditional expression of HNF1α exclusively in β-cells, we used the tetracycline-dependent expression system based on two transgenes: Tg^RIP-tTA and Tg^tet-Hnf1α (Fig. 1A). In Tg^RIP-tTA, tTA is transcribed specifically in β-cells under the control of the insulin promoter. In Tg^tet-Hnf1α, HNF1α is controlled by the tTA-dependent tetO operator. The two transgenes were crossed with Hnf1α^+/− mice to generate Tg^{tet-Hnf1α/RIP-tTA} mice on Hnf1α^+/+, Hnf1α^+/−, or Hnf1α^−/− backgrounds.

In the presence of tetracycline, there was no expression of HNF1α protein or mRNA in β-cells or liver from Hnf1α^−/− Tg^{tet-Hnf1α/RIP-tTA} by either RT-PCR or immunofluorescence (Figs. 1N and O and 6M). Further supporting that HNF1α was not expressed at significant levels, tetracycline-treated Hnf1α^−/− Tg^{tet-Hnf1α/RIP-tTA} mice had the full-blown Hnf1α^−/− phenotype, including identically low non-Mendelian birth rates, high postnatal mortality, low weight, renal glycosuria, fasting basal hyperglycemia, and fatty liver (19,20) (data not shown). In the absence of tetracycline, transgenic HNF1α in Tg^{tet-Hnf1α/RIP-tTA} mice was highly specific for β-cells, as shown by RT-PCR and immunofluorescence staining of HNF1α in islet, liver, and kidney (Fig. 1B–O).

Within pancreatic islets, the expression of Tg^tet-Hnf1α in induced mice was heterogeneous (Figs. 1D and 6F and supplemental Fig. 1B, which is detailed in the online appendix). Thus, normal HNF1α levels were observed in some β-cells, whereas many others showed mild overexpression or marked overexpression (Figs. 1B and D and 6D and F). Attempts to reduce expression levels in vitro or in vivo by using low levels of tetracycline were unsuccessful because this primarily resulted in changes in the percentage of transgenic HNF1α-expressing cells rather than altering single-cell expression levels (supplemental Fig. 1).

We first studied double transgenics in a wild-type Hnf1α locus background (Hnf1α^+/+ Tg^{tet-Hnf1α/RIP-tTA} mice) that were never exposed to tetracycline and thus had unrestrained expression of the transgene. At birth, mice were of normal weight, yet they strikingly exhibited hyperglycemia (Fig. 2A). At 6–8 weeks, male mice had severe hyperglycemia (Fig. 2A and B), whereas females displayed a more variable phenotype (not shown). Thus, sustained overexpression of HNF1α in β-cells throughout embryonic development resulted in diabetes.

To assess the effects of transgene induction postnatally, Hnf1α^+/+ Tg^{tet-Hnf1α/RIP-tTA} male mice were exposed to tetracycline since conception, then tetracycline was removed at 4 weeks of age to induce the transgene, and mice were analyzed 4 weeks later. This also resulted in a diabetic phenotype, although the results were more variable and less severe compared with mice that expressed the transgene during embryonic development (Fig. 2C). Thus, the consequences of overexpressing HNF1α in β-cells were milder when this was carried out in the adult stage.

To understand the diabetogenic effects of sustained overexpression of HNF1α, a morphologic examination of the pancreas was undertaken at different developmental time points. At E15.5, expression of HNF1α at levels that clearly exceeded those seen in wild-type cells was readily detected in 95% of β-cells of untreated Hnf1α^+/+ Tg^{tet-Hnf1α/RIP-tTA} embryos (Fig. 3J). At this stage the number of β- and α-cells in untreated Hnf1α^+/+ Tg^{tet-Hnf1α/RIP-tTA} embryos were normal (Fig. 3A–C). By E18.5, the islet β-cell population was clearly compromised in Hnf1α^+/+ Tg^{tet-Hnf1α/RIP-tTA} embryos, with a threefold reduction in β-cells, whereas glucagon cell numbers were not significantly different (Fig. 3D–F). Furthermore, β-cells appeared more scattered near ductal structures instead of forming islet aggregates (not shown). By postnatal day 7, there was a fivefold reduction in β-cells in Hnf1α^+/+ Tg^{tet-Hnf1α/RIP-tTA} male mice and a concomitant increase of α-cells (Fig. 3G–I) and ε-cells (not shown). Furthermore, there was a profound alteration of the normal islet architecture, with α-cells invading the central mass of islets (Fig. 3H and supplemental Fig. 2, which is detailed in the online appendix). In adult Hnf1α^+/+ Tg^{tet-Hnf1α/RIP-tTA} males, islets were even more severely compromised, with islets frequently composed mostly of glucagon cells (supplemental Fig. 2). Thus, overexpression of HNF1α in β-cells gradually led to a severe decrease in the number of β-cells and abnormal islet architecture.

Because untreated Hnf1α^+/+ Tg^{tet-Hnf1α/RIP-tTA} embryos had normal β-cell density at E15.5 (Fig. 3A–C), we concluded that the initial genesis of β-cells was unlikely to be affected by overexpression of HNF1α. With the progressive loss of β-cell mass, there was a striking concomitant reduction in the percentage of β-cells overexpressing HNF1α, dropping from 90% of β-cells in E18.5 to 10% in adult mice (Fig. 3K–M). Given that the transgene was mosaic in newly formed β-cells (Fig. 3J), this suggests a strong cell selection against overexpressing β-cells, thus favoring cells that do not express or express only lower levels of Hnf1α.

To understand why overexpression of HNF1α reduced the β-cell mass, proliferation and cell death were assessed. Untreated E18.5 embryos or neonatal mice were studied to ensure that long-standing hyperglycemia did not interfere with the results. At both time points, Hnf1α^+/+ Tg^{tet-Hnf1α/RIP-tTA} mice presented a significant decrease in β-cell proliferation, as measured with Ki67/insulin dual immunofluorescence (four- and twofold reduction, respectively) (Fig. 4A–F). Furthermore, most HNF1α-overexpressing β-cells did not coexpress Ki67 in either postnatal day (P) 0 or P7 Hnf1α^+/+ Tg^{tet-Hnf1α/RIP-tTA} mice (1.2 and 2.7%, respectively) (Fig. 4G and data not shown). This contrasts with a Ki67-labeling index of 16.5% at both ages in control β-cells expressing normal HNF1α (Fig. 4F and data not shown).

Furthermore, Hnf1α^+/+ Tg^{tet-Hnf1α/RIP-tTA} E18.5 mice displayed a nearly threefold increase in the percentage of p27-positive β-cells (Fig. 4H–J). Increased p27 expression was confirmed in HNF1α-overexpressing cells (36% p27-positive HNF1α-overexpressing β-cells vs. 13% in control β-cells) (Fig. 4J–K). Thus, reduced proliferation in HNF1α-overexpressing cells may result in part from increased expression of the cell-cycle inhibitor p27. Overexpression of HNF1α also induced apoptosis. A significant increase in the density of activated caspase 3 cells was observed in untreated Hnf1α^+/+ Tg^{tet-Hnf1α/RIP-tTA} E18.5 compared with nontransgenic littermates (4.41/mm² vs. 0.33/mm² activated caspase 3-positive cells, respectively) (Fig. 4L–O). Similar results were obtained in newborn sections from control and untreated Hnf1α^+/+ Tg^{tet-Hnf1α/RIP-tTA} pancreata (not shown).

In summary, overexpression of HNF1α in β-cells inhibited cell cycle activity and induced cell death. The analysis of preterm mice ensured that this preceded the development of diabetes and thus constitutes a causal factor, rather than a consequence of hyperglycemia.

To further understand the consequences of overexpressing HNF1α in β-cells, we compared the expression of a representative set of key β-cell–specific genes, including Pdx1, Nkx6.1, Pax6, Isl1, and GLUT2, in untreated Hnf1α^+/+ Tg^{tet-Hnf1α/RIP-tTA} and control embryos. Marked downregulation of all studied β-cell–specific genes was observed selectively in HNF1α-overexpressing β-cells, whereas the ubiquitous protein Sp1 and the cell membrane protein E-cadherin remained unaltered (Fig. 5 and data not shown). These inhibitory effects reflected long-term in vivo effects of HNF1α overexpression because it was not observed after transient 60-h ex vivo induction in Hnf1α^+/+ Tg^{tet-Hnf1α/RIP-tTA} isolated islets (not shown).

Despite the deleterious effects of long-term HNF1α overexpression in β-cells, we reasoned that because HNF1α expression in transgenic β-cells was very heterogeneous, some cells should express near-normal HNF1α concentrations. This allowed us to test the cell-autonomous regulatory functions of HNF1α in β-cells and the ability to activate silenced genes postnatally. Hnf1α^−/− Tg^{tet-Hnf1α/RIP-tTA} mice were thus treated with tetracycline since conception, and, at 4 weeks of age, tetracycline was removed to induce transgenic HNF1α for 4 weeks. Hnf1α^−/− and tetracycline-treated Hnf1α^−/− Tg^{tet-Hnf1α/RIP-tTA} mice showed a complete absence of GLUT2 expression in β-cells (Fig. 6B, E, and M). In contrast, Hnf1α^−/− Tg^{tet-Hnf1α/RIP-tTA} mice in which HNF1α had been induced for 4 weeks exhibited GLUT2 expression in a small number of β-cells (Fig. 6C and F). The staining was weaker than in wild-type β-cells, and it most frequently occurred in cells that expressed HNF1α at levels resembling those encountered physiologically (Figs. 6A–F), in keeping with the finding that HNF1α overexpression downregulates GLUT2 expression (Fig. 5I and L). Nonetheless, these mice did not show an improvement of the Hnf1α^−/− diabetic phenotype, presumably because of the deleterious consequences of excessive HNF1α in a large fraction of cells. Partial rescue of GLUT2 expression was also observed in untreated Hnf1α^−/− Tg^{tet-Hnf1α/RIP-tTA} newborn mice (Fig. 6G–L).

Previous studies in isolated mouse hepatic cells from Hnf1α knockout mice showed that HNF1α function could be partially rescued if HNF1α was reexpressed with adenovectors during early embryonic development but not in adult knockout hepatocytes (16). However, it is unclear whether HNF1α has a similar role as an early epigenetic regulator in β-cells. To assess whether HNF1α-dependent gene activity could be rescued postnatally in differentiated Hnf1α^−/− β-cells, we induced Tg^tet-Hnf1α in adult β-cells for short periods, thus circumventing the deleterious effects of long-term overexpression. Whole-animal experiments could not be carried out for this purpose because the in vivo effects of tetracycline persist up to 3 weeks after removal (21 and R.F.L., unpublished observations). We thus isolated and cultured Hnf1α^−/− Tg^{tet-Hnf1α/RIP-tTA} islets from 6-week-old mice in which the transgene had been repressed with tetracycline since conception. Islets were first cultured with tetracycline and then alternatively without tetracycline for 60 h to transiently induce HNF1α expression, or with tetracycline as a control. Importantly, Hnf1α mRNA was completely repressed in treated Hnf1α^−/− Tg^{tet-Hnf1α/RIP-tTA} mice, but it was strongly activated in the absence of tetracycline (Figs. 1O and 6M). Transgene expression was nevertheless heterogenous. Dose- and time-response studies suggested that this was more likely attributable to the variable time required for different cells to respond and the mosaic behavior of transgenes rather than to cell selection (supplemental Fig. 1 and data not shown). Even though only a subset of Tg^{tet-Hnf1α/RIP-tTA} islet cells express HNF1α at levels expected to be transcriptionally productive, HNF1α induction activated target gene expression (Fig. 6M). Thus, although Fxr, Glut2, Dbp, and Hdg2 (HNF1α-dependent gene 2) (J.M. Servitja, unpublished observations) mRNA were undetectable in Hnf1α^−/− islets, and only trace Hnf4α, Hgfa, and Pklr levels were observed, the ex-vivo induction of HNF1α in Hnf1α^−/− Tg^{tet-Hnf1α/RIP-tTA} islets resulted in the activation to 56, 7, 11, 7, 25, 10, and 3% of the expression levels observed in Hnf1α^+/+ islets, respectively (Fig. 6M). Thus, despite gene-specific variability, all tested HNF1α target genes responded to postnatal HNF1α rescue in islets. It is worth noting that Glut2 is β-cell–specific within pancreatic islets (supplemental Fig. 3, which is detailed in the online appendix) (22), and therefore, at least for this gene, the results reflect gene activation in β-cells. These results show that HNF1α has cell-autonomous functions in pancreatic islet cells that can be rescued postnatally.

HNF1α deficiency is the most common form of monogenic diabetes, and it represents a paradigm for a human β-cell transcriptional defect. A crucial question in this field is whether genes that are silenced because of a germ line transcription factor mutation can be reactivated at any point throughout the lifetime of the cell. Alternatively, the transcription factor may need to be present during a restricted developmental time frame, making the silenced target unable to respond at later time points.

Earlier studies using Hnf1α^−/− embryos have shown that HNF1α target gene dependence in the pancreas becomes manifest once the β-cell differentiates during embryonic development (8). However, this observation does not address the developmental time frame during which Hnf1α must exert its effects to activate its targets. In support for a possible early epigenetic role of HNF1α, previous work has shown that transduction of early (E12.5-E13.5) embryonic Hnf1α^−/− hepatocytes with adenovectors expressing HNF1α can restore the expression of the HNF1α target Pah to 30% of normal values (16), whereas no Pah mRNA was detected by RT-PCR after transduction of HNF1α in adult Hnf1α^−/− hepatocytes (16). Partial postnatal induction of Pah mRNA nevertheless occurred in the presence of DNA methylase and histone deacetylase inhibitors, suggesting that early expression of HNF1α in developing hepatocytes may be essential to remove negative regulatory chromatin constraints (16). It is thus remarkable that in our model, postnatal induction of HNF1α in adult Hnf1α^−/− β-cells activates target genes without the need for chromatin-modifying agents. As occurred with Pah induction in embryonic liver, the level of target activation in adult islet cells was incomplete, although this is likely to be markedly underestimated because the concentration of transgenic HNF1α in our system is nonphysiological in a large fraction of β-cells. These findings suggest that early developmental expression of HNF1α may be less critical in β-cells than in the liver. This is consistent with the fact that although HNF1α is present in early hepatoblasts, it is not detectable in putative pancreatic ductal precursors and becomes manifest as β-cells differentiate, whereas only its paralogue HNF1β is present in embryonic duct cells (12). The ability to recapitulate HNF1α function postnatally without the need to induce transcriptional competence during embryonic development is important if HNF1α function is to be replaced in a clinical setting of HNF1α deficiency.

The deleterious effects of overexpressing HNF1α has precluded testing whether β-cell–specific expression of HNF1α can rescue the diabetic phenotype. We believe that the inhibitory effects in Tg^{tet-Hnf1α/RIP-tTA} mice are unlikely to reflect a bona fide physiological gain-of-function of HNF1α in β-cells because HNF1α-deficient mice clearly do not exhibit increased proliferation or decreased apoptosis (3,19 and R.F.L., unpublished observations). Instead, several nonexclusive potential mechanisms can be considered for the inhibitory effects. First, because HNF1α regulates transcription by the recruitment of coactivators (6,23,24), high concentrations of HNF1α could sequester these coactivators, leading to deranged transcription of essential genes. Deleterious effects could also result from the interaction with the dimerization cofactor of HNF1 (DCoH) pterin-4a-carbinolamine dehydratase (PCD), a dual-function protein that acts as a coactivator of HNF1α in the nucleus but regenerates tetrahydrobiopterin in the cytoplasm (25). Immunofluorescence studies showed strong nuclear relocalization of PCD/DCoH in transgenic β-cells (not shown), suggesting that cytoplasmic depletion of PCD/DCoH could underlie the inhibitory effects. Finally, excessive HNF1α protein may trigger endoplasmic reticulum stress and apoptosis, a mechanism known to result in β-cell depletion (26–28). It is nevertheless worth noting that transgenic overexpression per se does not produce this effect in β-cells because many proteins have been overexpressed with an insulin promoter fragment in β-cells without affecting cellular mass, including the single Tg^RIP-tTA transgenic in this study, several Cre recombinase lines, and a diverse set of gene products (29–35). It is possible, however, that transcriptional regulators are particularly prone to give rise to negative consequences when overexpressed. We have identified four examples to support this notion: the transcription factors AP-2, Ets2, mouse Pax6 (small eye), and Pax6 Drosophila homolog Eya (eyes absent), all of which result in cell growth inhibition when overexpressed in other cell types in transgenics, even though their respective knockout phenotypes are not consistent with growth suppressor activity (36–41).

β-Cell depletion was associated with increased islet α-cells. This has been observed in other β-cell–deficient models (42), although the causes are poorly understood. We cannot discard that some β-cells undergo a fate change to α-cells, although insulin has been reported to inhibit α-cell function (43), and it is thus possible that β-cells also repress α-cell growth.

Interestingly, the phenotype of mice overexpressing HNF1α is reminiscent of that seen in transgenic and cell models in which dominant-negative derivatives of HNF1α are overexpressed. Such models exhibit a marked reduction of β-cell proliferation, apoptosis, and an alteration of the islet structure, giving rise to a severe phenotype (44–47) not observed in knockout mice (3,6,19). This indicates that dominant-negative derivatives of HNF1α may have effects in addition to inhibiting HNF1α, some of which may be mediated through mechanisms similar to those elicited by overexpression of full-length HNF1α. This does not negate that dominant-negative models have indeed been shown to suppress HNF1α and revealed bona fide HNF1α-dependent functions in β-cells (7,44,46,47).

The observation that HNF1α in β-cells is vulnerable not solely to deficiency but also to excessive concentrations has several implications. Quite commonly, transcription factor genes are expressed with relatively strong expression systems in both gene therapy settings and in experimental approaches to study gene function. The current study suggests that any attempt to correct HNF1α deficiency in humans needs to ensure that physiological levels of HNF1α are attained. By extension, this knowledge needs to be considered in any attempt to use forced expression of HNF1α, or plausibly of any other transcription factor, to recapitulate β-cell genetic programs artificially in transdifferentiation (48,49) or stem cell differentiation schemes aimed at generating β-cells for replacement therapy (50,51). Importantly, it highlights that the lack of a desired effect resulting from the delivery of a transcription factor gene in any setting cannot be regarded as conclusive unless normal concentrations are attained.

In summary, this study shows that HNF1α has cell-autonomous functions in pancreatic islets that can be elicited postnatally. It also underscores the importance of HNF1α expression levels in β-cells. These findings have potential implications for the treatment of HNF1α-deficient diabetes and for the interpretation of models used to study transcription factors regulating β-cell differentiation and function.

This work was funded by a PhD studentship to R.F.L. from the Generalitat de Catalunya, Juvenile Diabetes Research Foundation Grant 1-2002-21 (to J.F.), Ministerio de Educacion y Ciencia Grant SAF2005-850, and Instituto de Salud Carlos III Grant G03/212.

We thank Joel Habener and Larry Moss for discussions before initiating these studies, Frank Gonzalez (National Cancer Institute) for Hnf1α^−/− mice, Lars Ährlund-Richter (Karolinska Institute) for transgenic microinjection, the University of Barcelona Faculty of Medicine Confocal Microscopy Unit, Gerald Crabtree for Hnf1α cDNA, and Marcelina Parrizas for subcloning Hnf1α cDNA in pBIL.