MAFA controls genes implicated in insulin biosynthesis and secretion
- First Online:
- Cite this article as:
- Wang, H., Brun, T., Kataoka, K. et al. Diabetologia (2007) 50: 348. doi:10.1007/s00125-006-0490-2
Effects of the transcription factor v-maf musculoaponeurotic fibrosarcoma oncogene homologue A (MAFA) on the regulation of beta cell gene expression and function were investigated.
Materials and methods
INS-1 stable cell lines permitting inducible up- or downregulation of this transcription factor were established.
MAFA overproduction enhanced and its dominant-negative mutant (DN-MAFA) diminished binding of the factor to the insulin promoter, correlating with insulin mRNA levels and cellular protein content. Glucose-stimulated insulin secretion was facilitated by MAFA and blunted by DN-MAFA. This is partly due to alterations in glucokinase production, the glucose sensor of beta cells. In addition, the expression of important beta cell genes, e.g. those encoding solute carrier family 2 (facilitated glucose transporter), member 2 (formerly known as GLUT2), pancreatic and duodenal homeobox factor 1 (PDX1), NK6 transcription factor-related, locus 1 (NKX6-1), glucagon-like peptide 1 receptor (GLP1R), prohormone convertase 1/3 (PCSK1) and pyruvate carboxylase (PC), was regulated positively by MAFA and negatively by DN-MAFA.
The data suggest that MAFA is not only a key activator of insulin transcription, but also a master regulator of genes implicated in maintaining beta cell function, in particular metabolism–secretion coupling, proinsulin processing and GLP1R signalling. Our in vitro study provides molecular targets that explain the phenotype of recently reported Mafa-null mice. We also demonstrate that MAFA is produced specifically in beta cells of human islets. Glucose influenced DNA-binding activity of MAFA in rat islets in a bell-shaped manner. MAFA thus qualifies as a master regulator of beta-cell-specific gene expression and function.
KeywordsBeta cellsHumanIslet insulin secretionMAFAMetabolism
dominant-negative mutant of MAFA
electrophoretic mobility-shift assay
glucagon-like peptide 1 receptor
3-hydroxyacyl-CoA dehydrogenase type II
hepatocyte nuclear factor 4 α
ISL1 transcription factor LIM/homeodomain (islet-1)
v-maf musculoponeurotic fibrosarcoma oncogene homologue A
neurogenic differentiation factor 1
NK6 transcription factor-related, locus 1
paired-box protein 4
prohormone convertase 1/3
pancreatic and duodenal homeobox factor 1
solute carrier family 2 (facilitated glucose transporter), member 2
transcription factor 1, hepatic
transcription factor A, mitochondrial
uncoupling protein 2 (mitochondrial, proton carrier)
Type 1 diabetes is characterised by absolute insulin deficiency caused by autoimmune destruction of the insulin-producing pancreatic beta cells. Patients with this disease require daily injections of insulin or islet transplantation. Because islet donors are scarce, the generation of insulin-producing cells from human progenitor/stem cells has become a key challenge in the treatment of type 1 diabetes. The reduction in insulin secretory capacity and beta cell mass are the precipitating factors for development of type 2 diabetes in addition to insulin resistance in the liver, muscle and adipose tissues. Therefore, the identification and characterisation of key transcription factors regulating beta cell gene expression and insulin secretion are essential not only for treatment of type 1 diabetes but also for understanding the pathogenesis of beta cell dysfunction in type 2 diabetes.
The beta cell specificity and glucose responsiveness of insulin expression are conferred by three conserved enhancer elements in the promoter region of the insulin gene. These are E1, A3, and RIPE3b/C1, which, respectively, bind three transcription factors: neurogenic differentiation factor 1 (NEUROD1), pancreatic and duodenal homeobox factor 1 (also known as insulin promoter factor 1, homeodomain transcription factor [IPF1]) and v-maf musculoaponeurotic fibrosarcoma oncogene homologue A (MAFA) [1–5]. NEUROD1 is produced in all pancreatic endocrine cell types, whereas PDX1 and MAFA display beta-cell-restricted production [1–5]. Mouse models with targeted gene disruption have revealed the essential roles of Neurod1 and Pdx1 in pancreatic development and beta cell differentiation. In addition, mutations in human NEUROD1 and PDX1 are associated, respectively, with MODY, subtypes 4 and 6. Moreover, adenovirus vector-mediated gene therapy using either Neurod1 in combination with betacellulin  or Pdx1 alone  has been shown to induce insulin production in mouse liver and cure streptozotocin-induced diabetes. Similar gene therapy was recently shown to yield encouraging results, taking advantage of the synergistic actions of Mafa, Pdx1 and Neurod1 . Furthermore, beta cell replacement therapy in similar animal models has been achieved using in vitro-differentiated insulin-producing cells, which are generated through expression of Pdx1 in human fetal liver progenitor cells .
Large MAF proteins represent a subgroup of the MAF family of basic-leucine zipper transcription factors and include MAFA, c-MAF, MAFB and neural retina-specific leucine zipper protein (NRL) [4, 5]. MAFA is a key regulator of insulin gene transcription, whereas c-MAF and MAFB are produced specifically in adult alpha cells and control the glucagon promoter activity [4, 10]. Recent studies, however, suggest that MAFB is not only important for islet alpha cell function, but may also be involved in regulating genes required for both alpha and beta cell differentiation during development [11–13]. It is noteworthy that reduced MAFA function has been implicated in the development of beta cell dysfunction [14–19]. Hyperglycaemia suppresses Mafa expression in vivo, which is prevented by constitutive overexpression of forkhead box O1 (Foxo1) in beta cells . In addition, glucotoxicity and lipotoxicity inhibit insulin expression through regulation of the DNA-binding activity and/or levels of MAFA and PDX1 . Moreover, reduced expression of Mafa has been reported in beta cells deficient in both insulin and IGF-1 receptors and is associated with reduced beta cell mass and diabetes .
The role of MAFA in regulation of pancreatic beta cell gene expression and insulin secretion has not been fully characterised. Using inducible INS-1 beta cell models, we have previously defined the function of PDX1 in the regulation of target gene expression and insulin secretion [20, 21], consistent with the results obtained in mice with beta-cell-specific deletion of Pdx1 [22, 23]. Similar approaches were employed in the present study to establish the regulatory role of MAFA in beta cell function.
Materials and methods
Establishment of inducible INS-1 stable cell lines
The first-step stable and INS-1-derived INS-rbeta (also referred to as r9) cell line, which carries the reverse tetracycline/doxycycline-dependent transactivator  has been described previously [20, 25]. The plasmids used in the secondary stable transfection were constructed by subcloning the cDNA encoding either wild-type MAFA  or its dominant-negative mutant (DN-MAFA)  into the expression vector PUHD10-3  (a generous gift from H. Bujard, University of Heidelberg, Germany). Both MAFA and DN-MAFA were epitope-tagged with haemagglutinin. The procedures for stable transfection, clone selection, and screening have been described previously .
The standard rat insulinoma INS-1-derived stable cell lines were cultured in RPMI 1640 containing 11.2 mmol/l glucose , unless otherwise indicated.
Immunofluorescence in INS-1 cells
Cells grown on polyornithine-treated glass coverslips were treated for 24 h with or without 500 ng/ml doxycycline. Cells were then fixed and incubated with anti-haemagglutinin tag (a kind gift from MBL/LabForce, Nunningen, Switzerland) (1:100 dilution), followed by the second antibody labelling.
Nuclear protein extraction and electrophoretic mobility-shift assay
Nuclear extracts from INS-1 cells were prepared as previously described . Rat islets were isolated by collagenase digestion and their nuclear proteins were extracted as reported . The double-stranded oligonucleotide corresponding to the RIPE3b/C1 element of the rat insulin II promoter, 5′-TGGAAACTGCAGCTTCAGCCCCTCTG-3′, was used as a probe . Electrophoretic mobility-shift assay (EMSA) procedures were performed using the same conditions for probe labelling and binding reactions as previously described . Anti-MAFA antibody (Bethyl, Montgomery, TX, USA) was used for supershift experiments.
For western blotting, cells were cultured with 0, 75, 150, and 500 ng/ml doxycycline for 24 h. Nuclear extracts and total cellular proteins were fractionated by 11% SDS-PAGE. The dilution for MAFA antibody was 1:4,000. Immunoblotting procedures were performed as described previously  using enhanced chemiluminescence (Pierce, Rockford, IL, USA) for detection.
Measurements of insulin secretion and cellular insulin content
Insulin secretion in INS-1-derived cells was measured in 24-well plates over a period of 30 min, in KRB-HEPES buffer (KRBH; mmol/l: 140 NaCl, 3.6 KCl, 0.5 NaH2PO4, 0.5 MgSO4, 1.5 CaCl2, 2 NaHCO3, 10 HEPES, 0.1% BSA) containing the indicated concentrations of glucose. Cellular insulin content was determined after extraction with acid ethanol following the procedures of Wang et al. . Insulin was detected by radioimmunoassay using rat insulin as a standard .
Total RNA isolation and northern blotting
Total RNA was extracted and blotted onto nylon membranes as described previously . The membrane was prehybridised and then hybridised to 32P-labelled random primer cDNA probes according to Wang and Iynedijian . To ensure equal RNA loading and even transfer, all membranes were stripped and re-hybridised with a house-keeping gene probe (cyclophilin). cDNA fragments used as probes for Mafa, glucokinase (Gck), solute carrier family 2 (facilitated glucose transporter), member 2 (Slc2a2, formerly known as Glut2), insulin, transcription factor 1, hepatic (Tcf1, formerly known as hepatocyte nuclear factor 1 α [Hnf1α]), hepatocyte nuclear factor 4 α (Hnf4a), NeuroD1 and Pdx1 mRNA detection were digested from the corresponding plasmids. cDNA probes for rat glucagon-like peptide 1 receptor (Glp1r), NK6 transcription factor-related, locus 1 (Nkx6-1), glucagon, prohormone convertase 1/3 (Pcsk1), pyruvate carboxylase, uncoupling protein 2 (mitochondrial, proton carrier) (Ucp2), transcription factor A, mitochondrial (Tfam), short-chain 3-hydroxyacyl-CoA dehydrogenase type II (Hadh2), acyl-CoA oxidase and cyclophilin were prepared by RT-PCR and confirmed by sequencing.
Human islet isolation and culture
Freshly isolated human islets from three different donors, who had provided written informed consent, were obtained from D. Bosco (Cell Isolation and Transplantation Center, Department of Surgery, Geneva, Switzerland) and maintained in CMRL-1066 (Invitrogen) at 5.6 mmol/l glucose supplemented with 10% FCS, 100 Units/ml penicillin, 100 μg/ml streptomycin and 100 μg/ml gentamycin (Sigma) for 2–4 days before experiments. The present investigation was approved by the institutional ethics committee. Handpicked islets were then exposed to 5.6 or 11 mmol/l glucose with or without the GLP1R agonist exendin-4 (10 nmol/l) for 24 h.
After stimulation, human islets were partially dispersed by trypsin treatment and cells were fixed and incubated with primary rabbit antibodies for MAFA, ISL1 transcription factor, LIM/homeodomain, (islet-1) (ISL1) and PDX1 (dilutions 1:250) and mouse anti-insulin (1:1,000, Sigma), followed by labelling with secondary antibodies and staining with 10 μg/ml 4′,6-diamidino-2-phenylindole. Images were acquired with a Zeiss Axiocam Imaging System (Bioimaging Core Facility, Medical Faculty, Geneva University, Geneva, Switzerland).
Results are expressed as mean±SEM and statistical analyses were performed by unpaired Student’s t test.
Establishment of stable INS-1 cell lines permitting inducible expression of Mafa and DN-Mafa
We established more than ten clones positively expressing Mafa and DN-Mafa, respectively, using parental INS-rbeta cells . The clones, designated as Mafa*22 and DN-Mafa*39, were selected for the present study because they displayed robust levels of Mafa or DN-Mafa mRNAs after doxycycline induction and showed undetectable background expression under non-induced conditions.
Characterisation of INS-1 clones producing MAFA and DN-MAFA
EMSA with the RIPE3b/C1 element of the insulin promoter demonstrated that doxycycline-induced MAFA production increased its DNA-binding activity dose-dependently (Fig. 2b). The intensity of retarded MAFA binding complexes, performed with nuclear extracts from cells cultured for 24 h with 75, 150, and 500 ng/ml doxycycline, was estimated to be fivefold, 15- and 15-fold, respectively, compared with that of endogenous MAFA (Fig. 2b). DN-MAFA lacks the N-terminal transactivation domain, but contains intact DNA-binding activity . It therefore exerts its dominant-negative function through competing with endogenous MAFA for cognate DNA binding. Indeed, induction of DN-MAFA almost completely abolished MAFA DNA-binding activity (Fig. 2b). Again, no leakage binding activity of DN-MAFA was observed under non-induced conditions. The specificity of retarded binding complexes of MAFA and DN-MAFA were confirmed by the supershift experiments with MAFA antibody.
MAFA controls the expression of beta-cell-specific genes
Accordingly, induction of DN-MAFA led to a marked reduction in the transcript levels of insulin, Glp1r, Pdx1, Nkx6-1 and Pcsk1, whereas the expression of Tcf1 and Hnf4a was unaltered (Fig. 3c). However, Neurod1 expression was suppressed by DN-MAFA (Fig. 3c).
We and others have previously shown that PDX1 and NKX6-1 suppress glucagon expression, causing a shift from an alpha to a beta phenotype [20, 31]. We therefore examined whether MAFA has a similar function. Unexpectedly, induction of MAFA did not inhibit the expression of glucagon, as demonstrated by northern blotting after prolonged film exposure (Electronic supplementary material [ESM] Fig. 1). More intriguingly, induction of DN-MAFA diminished levels of glucagon mRNA (ESM Fig. 1), an effect opposite to that observed in response to suppression of PDX1 or NKX6-1 activity in INS-1 cells. Therefore, the effects of MAFA cannot be simply explained as functioning through PDX1 or NKX6-1.
MAFA regulates the expression of genes implicated in glucose metabolism
Although Gck is expressed in both islet alpha and beta cells, it has been shown to confer beta cell glucose-sensing [25, 40]. The mRNA levels of Gck were elevated by MAFA and reduced by DN-MAFA (Fig. 4). Such effects seem to be unique to MAFA, since Gck expression remained constant in our INS-1 cells permitting either induction or dominant-negative suppression of other pancreatic transcription factors, such as PDX1 , TCF1 , and HNF4α , as well as NEUROD1 or PAX4 (unpublished data, H. Wang and C. M. Wollheim).
The expression of other genes encoding proteins implicated in metabolism–secretion coupling, such as UCP2, TFAM, HADH2, and acyl-CoA oxidase [41–43], was not regulated by manipulation of MAFA function in INS-1 cells (Fig. 4). UCP2 and TFAM modulate, respectively, the membrane potential and biogenesis of mitochondria, whereas HADH2 and acyl-CoA oxidase are involved in fatty acid metabolism.
MAFA regulates glucose-stimulated insulin secretion
Using two INS-1 stable cell lines permitting inducible overexpression or dominant-negative suppression of MAFA, we assessed the effects of MAFA on insulin secretion. Overproduction of MAFA caused a left shift of insulin secretion in response to glucose, whereas induction of DN-MAFA blunted nutrient-stimulated insulin release. This was not due to alterations in insulin content, since the data were expressed as percentage of cellular insulin. Recently, Zhang and colleagues  reported that mice deficient in MAFA developed diabetes due to impaired insulin secretion. However, the authors did not define the molecular targets responsible for the phenotype. We have defined the target genes of MAFA through quantitative gene expression profiling of our novel INS-1 cell lines under induced and non-induced conditions. The results demonstrate that MAFA is not only a key activator of insulin transcription, but also a master regulator of many beta cell genes implicated in maintaining beta cell phenotype, metabolism–secretion coupling, proinsulin processing and GLP1R signalling.
Overexpression of Mafa alone is sufficient to raise endogenous insulin mRNA levels. This is unique to MAFA, because induction of PDX1  or NEUROD1, either separately or combined in the similar INS-1 cell system (unpublished data, H. Wang and C. B. Wollheim), did not reproduce the effects of MAFA. Similarly, other key beta cell transcription factors such as TCF1 and HNF4α, as well as PAX4, do not show such potency in our systematic series of INS-1 cell models [26, 36]. Transduction of MAFA has been shown to increase endogenous insulin transcript expression in rat islets . We therefore propose that MAFA could be a potent and target transcription factor employed for beta cell generation either through in vivo gene therapy or in vitro differentiation of stem/progenitor cells.
Accordingly, dominant-negative suppression of MAFA function decreased insulin production as revealed by mRNA and protein contents. We have also identified the beta-cell-enriched PCSK1and GLP1R as two novel targets of MAFA. DN-MAFA drastically suppressed the expression of Pcsk1 and Glp1r, whereas induction of MAFA enhanced their transcript levels. PCSK1 is required for insulin maturation , whereas GLP1R is necessary for GLP1-mediated beta cell growth and insulin secretion . A similar reduction of PCSK1 and GLP1R levels has also been observed in our INS-1 cells expressing DN-Pdx1, which causes defective proinsulin processing and GLP1R signalling . MAFA has been previously reported to bind the Pdx1 promoter . Indeed, we found that PDX1 is robustly up- or downregulated following induction and suppression of MAFA function. Decreased expression of Pdx1 has also been demonstrated in MAFA-deficient islets . We therefore postulate that MAFA regulates expression of Pcsk1 and Glp1r, at least in part by acting through PDX1.
In addition, we have also identified two other common target genes of MAFA and PDX1: Slc2a2 and Nkx6-1. Like PDX1, MAFA also controls the expression of these two beta-cell-restricted genes. Decreased levels of SLC2A2 have also been reported in the islets of Mafa-null mice . However, this does not explain the impaired insulin secretion . As the capacity of this hexose transporter is much greater than the activity of GCK and other glycolytic enzymes, this decrease would not affect glucose metabolism . To delineate the molecular targets responsible for the regulatory effects of MAFA on insulin secretion, we evaluated the expression patterns of genes involved in metabolism–secretion coupling with or without induction of MAFA or DN-MAFA. The beta-cell-enriched PC has been identified as another novel target of MAFA in the present study. PC plays an essential role in metabolism–secretion coupling in both native beta cells and insulinoma cells [38, 39]. Approximately 50% of glucose-derived pyruvate enters mitochondrial metabolism via carboxylation catalysed by PC, while the remainder is converted into acetyl CoA by the pyruvate dehydrogenase complex . This provides anaplerotic input into the tricarboxylic acid cycle . INS-1 cells, which express low levels of monocarboxylate transporter that are nonetheless sufficient to take up pyruvate, provide a suitable model to assess the impact of decreased PC production on insulin secretion. Induction of DN-MAFA not only impaired glucose-stimulated but also pyruvate-induced insulin release. As pyruvate is metabolised exclusively by the mitochondria, the downregulation of PC certainly contributes, together with GCK, to the insulin secretory defect. On the other hand, the MAFA-evoked overproduction of PC may partially contribute to the left shift of insulin secretion in response to glucose. However, the beta cell glucose-sensor, GCK, should be the major player, since a similar left shift of insulin secretory response has been documented in INS-1 cells overproducing GCK . At variance with Mafa-deficient mice, we found that mRNA levels of Gck were regulated positively by MAFA and negatively by DN-MAFA. This is also unique to MAFA because Gck expression remains constant in our INS-1 stable cell lines allowing induction or suppression of other pancreatic transcription factors, including PDX1 , TCF1 , HNF4α , NEUROD1 or PAX4 (unpublished data, H. Wang and C. B. Wollheim). The reason for the apparent difference in GCK regulation in Mafa-mutant mice and INS-1 cells is not clear. Moreover, K+-induced insulin secretion was inhibited in islets from Mafa−/− mice but not affected in INS-1 cells with suppressed MAFA function. This could be due to effects on L-type voltage-gated Ca2+channels in native beta cells. INS-1 cells express a broader spectrum of such Ca2+channels than native beta cells , which may explain the preserved insulin secretion following membrane depolarisation.
Another intriguing result with MAFA is that it does not promote the alpha to beta phenotype shift observed with PDX1 and NKX6-1 [20, 31], although our data suggest that MAFA regulates their production. We have previously shown that PDX1 suppresses glucagon production and regulates the alpha/beta phenotype shift in INS-1 cells . A recent elegant study in INS-1-derived cells suggests that PDX1 acts through NKX6-1 . Therefore, the effects of MAFA on beta-cell-specific gene expression cannot be explained simply by an action on PDX1. Another argument supporting this notion is that overexpression of Pdx1 does not potentiate glucose-stimulated insulin secretion . Furthermore, in agreement with findings in Mafa-null mice , induction of DN-MAFA diminished the expression of Neurod1 in INS-1 cells, although MAFA induction did not affect its transcript levels.
We also demonstrated that MAFA is specifically expressed in the beta cells of human islets and that its subcellular distribution is not clearly regulated by physiological concentrations of glucose. However, the DNA-binding activity of MAFA is indeed regulated in a bell-shape manner that is dependent on glucose concentration. We therefore clarified the apparent controversies in the literature regarding the glucose responsiveness of MAFA DNA binding [5, 16, 17, 46]. To summarise data from the present study and previous reports, we suggest that glucose induces MAFA DNA-binding activity at physiological concentrations, but suppresses the binding at both sub- and supra-physiological levels.
In conclusion, we employed an inducible INS-1 cell system, which allows direct assessment of MAFA target genes. We have demonstrated that MAFA controls many beta-cell-specific genes, including insulin, Slc2a2, Pdx1, Nkx6-1, Glp1r, Pcsk1 and Pc. PC in combination with GCK may represent the molecular targets responsible for the regulatory effects of MAFA on metabolism–secretion coupling. Furthermore, MAFA regulates expression of genes essential for insulin transcription and processing, as well as for GLP1R signalling. Therefore, we propose that MAFA is a candidate gene for diabetes and a target gene for restoring beta cell function in type 2 diabetes. In addition, it is also a potential target for treatment of type 1 diabetes through MAFA-mediated in vitro beta cell generation from progenitor/stem cells or in vivo gene therapy.
We are grateful to D. Cornut-Harry, D. Nappey, and N. Aebischer for expert technical assistance. We are indebted to H. Bujard (for the PUHD 10-3 vector), and N. Quintrell (for the pTKhygro plasmid). This work was supported by the Swiss National Science Foundation (grant no. 32-66907.01 to C. B. Wollheim), by NIH/NIDDK RO1 DK06127 grant (to A. J. Sharma), by the Olga Mayenfisch Foundation (funding to C. B. Wollheim), and by the Roche Research Foundation, Basel. We are grateful to D. Bosco for providing human islets. This study was supported by the European Union (Integrated Project EuroDia.
LSHM-CT-2006-518153 in the Framework Programme 6 [FP6] of the European Community).
Duality of interest
The authors have declared that they have no conflicts of interest.