Preserving Mafa Expression in Diabetic Islet β-Cells Improves Glycemic Control in Vivo*

Background: MAFA expression is markedly decreased in islet β-cells of type 2 diabetes mellitus. Results: Mis-expression of Mafa in mouse diabetic db/db β-cells ameliorated glucose-stimulated insulin secretion and β-cell mass. Conclusion: Mafa alone is sufficient to improve β-cell function and mass under diabetic conditions. Significance: These results establish how consequential this transcription factor is to islet β-cells under pathological conditions. The murine Mafa transcription factor is a key regulator of postnatal islet β-cell activity, affecting insulin transcription, insulin secretion, and β-cell mass. Human MAFA expression is also markedly decreased in islet β-cells of type 2 diabetes mellitus (T2DM) patients. Moreover, levels are profoundly reduced in db/db islet β-cells, a mouse model of T2DM. To examine the significance of this key islet β-cell-enriched protein to glycemic control under diabetic conditions, we generated transgenic mice that conditionally and specifically produced Mafa in db/db islet β-cells. Sustained expression of Mafa resulted in significantly lower plasma glucose levels, higher plasma insulin, and augmented islet β-cell mass. In addition, there was increased expression of insulin, Slc2a2, and newly identified Mafa-regulated genes involved in reducing β-cell stress, like Gsta1 and Gckr. Importantly, the levels of human GSTA1 were also compromised in T2DM islets. Collectively, these results illustrate how consequential the reduction in Mafa activity is to islet β-cell function under pathophysiological conditions.

Type 2 diabetes mellitus (T2DM) 2 is caused by insufficient insulin production from pancreatic islet ␤-cells in the setting of insulin resistance, with the latter principally reflecting the inability of cells in muscle, liver, and fat to respond adequately to normal insulin levels. Precisely why ␤-cells fail to produce sufficient quantities of insulin under these conditions is unclear. Notably, a subset of ␤-cell-enriched transcription factors essential to ␤-cell development and/or function were recently shown to be inactivated under T2DM stress conditions in rodent models and human islet ␤-cells (1-3), specifically MAFA, PDX1, and NKX6.1. Compelling evidence indicates that reactive oxygen species generated by increased glucose metabolism causes ␤-cell inactivation and even death in T2DM. Significantly, islet ␤-cells have unusually low antioxidant enzyme levels (e.g. glutathione peroxidase-1 and catalase) (4 -6), with antioxidant treatment improving ␤-cell function in human T2DM islets (7)(8)(9) and T2DM animal models (10 -12). For example, transgenic ␤-cell-specific expression of glutathione peroxidase-1 improved Mafa, Nkx6.1, and blood glucose levels in db/db mice, a model of T2DM (2,12).
The change in Mafa was found to occur earlier than Nkx6.1 in mouse db/db ␤-cells, correlating closely with decreased expression of essential regulators involved in cell proliferation, glucose sensing, and insulin secretion (2,13). Reduced levels of such effectors were also found in pancreas-specific Mafa ⌬panc deletion mutant mice (e.g. Insulin, CyclinD2, and Munc18-1 (14)). In addition, Mafa is only produced in embryonic insulin ϩ cells destined to populate the adult, which represents an unusually late and highly specific expression pattern in relationship to other islet-enriched transcription factors (15). Islet ␤-cell dysfunction under T2DM stress conditions likely results from the gradual loss of MAFA followed by either PDX1 or NKX6.1, because Mafa ⌬panc mice are only glucose intolerant (14), whereas islet ␤-cell-specific loss of Pdx1 or Nkx6.1 almost immediately causes overt hyperglycemia (16 -19).
In the present study, we directly evaluated the impact of Mafa insufficiency in T2DM by generating transgenic db/db mice that conditionally expressed this transcription factor in only islet ␤-cells. The Mafa producing db/db mice demonstrated improved glycemic control and ␤-cell function, with restoration coinciding with expression of proteins that reduce oxidative stress. These studies not only provide keen insight into the prominence of Mafa activity in vivo, but also shed light on the significance of developing T2DM therapeutics to ameliorate ␤-cell function by preventing transcription factor inactivation.

EXPERIMENTAL PROCEDURES
Human Pancreas Samples-Pancreatic tissue was obtained from patients at Osaka University Hospital who were undergoing a partial pancreatectomy to remove pancreatic or distal bile duct tumor cells. Glucagon tolerance tests were performed before the surgery. The pancreas sample was fixed in 4% paraformaldehyde and 4-m sections were prepared using routine procedures. The study protocol was approved by the Osaka University Hospital Ethics Committee, and informed consent was obtained from each patient. The clinical donor data are provided in supplemental Table 1.
Generation of Pdx1 PB -CreER TM ;CAG-CAT-Mafa myc ;db/db Mice (termed ␤Mafa myc :db/db)-pCAG-CAT-Mafa myc was constructed from pCAG-CAT-lacZ (20) by replacing the lacZ sequences with a fragment containing mouse Mafa coding sequences linked to a myc tag and the bovine growth hormone polyadenylation signal. A 5.0-kb SalI-SacI CAG-CAT-Mafa myc spanning fragment of this plasmid was purified and microinjected into fertilized eggs of BDF1 mice. A total of 13 lines of CAG-CAT-Mafa myc mice were generated, and the high TMinducible signal to sham-treated Mafa myc expression properties of the b, d, and f lines were selected for further analysis. Pdx1 PB -CreER TM transgenic mice (21), which express TM-activated Cre recombinase under the control of the islet ␤-cell specific Pdx1 Area I/II enhancer, were crossed with the b, d, and f CAG-CAT-Mafa myc lines to generate Pdx1 PB -CreER TM ;CAG-CAT-Mafa myc (i.e. ␤Mafa myc ) mice. These mice were then backcrossed with C57BL/KsJ-db/m (db/m) mice for more than 10 generations to eventually obtain ␤Mafa myc ;db/db mice. Subcutaneous injections of 0.1 mg/1.0 g of BW TM were performed three times within 5 days for the induction of Mafa myc expression. The efficacy of islet ␤-cell expression was determined by anti-myc epitope staining. Because all of three lines demonstrated a similar improvement of plasma glucose levels after crossing with db/db mice, we mainly used the b line of CAG-CAT-Mafa myc . All animal procedures were approved by the Ethics Review Committee for Animal Experimentation of Osaka University Graduate School of Medicine.
Glucose Tolerance Tests-Glucose tolerance tests (0.5 g/kg of BW) were performed 4 and 8 weeks after TM injection on mice fasted overnight. Glucose and insulin levels were measured from tail vein sampled blood with a portable glucose meter and the insulin ELISA Kit (Morinaga Biochemicals, Yokohama, Japan).
Islet Perifusion Analysis-Isolated islets were first cultured overnight in 10% FCS RPMI medium containing 5 mM glucose, and then 20 islets were placed in a chamber and perifused for 1 h with 40 mg/dl of glucose, followed by 30 min with 400 mg/dl of glucose. The effluent was collected every 30 s for 5 min, followed by 1 min for 5 min, and 2 min for 8 min. The sample insulin concentration was normalized to that of the whole cell protein.
Microarray Analysis-The quality of islet RNAs was determined using an Agilent Bioanalyzer (Agilent Technologies, Palo Alto, CA), and samples with RNA integrity number more than 7.0 were used for microarray analysis. Total RNA was amplified with the WT-Ovation TM RNA Amplification system (NuGEN Technologies, San Carlos, CA) and labeled with cyanine 3. Each hybridization contained 1.65 g of fragmented cyanine 3-labeled cDNA, and was hybridized at 65°C for 17 h to the Agilent Mouse GE 4 ϫ 44K v2 microarray (Design ID 026655). Signal intensity was determined with an Agilent DNA microarray scanner. Normalization was performed using Agilent GeneSpring GX version 11.0.2 (per chip, normalization to 75 percentile shift; per gene, normalization to median of all samples). Data filtration was performed, resulting in a total of 30,161 probes as a valid probe set where at least one of the four total samples had a present flag. A 2-fold or greater change in signal intensity was considered a significant difference.
Gene Ontology (GO) Analysis-Gene ontogeny analyses on microarray data were performed with GeneSpring GX software using the annotations provided in the database of the Gene Ontology Consortium. The data were processed with Fisher's exact test and multiple test correction to identify significant over-representation of Gene Ontology annotations belonging to Molecular Functions, and the top 20 of significantly up-regulated (Ն 2.0-fold) GO terms were listed in Table 1.
Statistical Analysis-Data are expressed as mean Ϯ S.D. Statistical analysis was performed using one-way analysis of variance followed by the Scheffe's test. A value of p Ͻ 0.05 was considered to be statistically significant.
␤Mafa myc mice were crossed more than 10 generations into the C57BL/6KsJ db/db background, a T2DM model characterized by insulin resistance and ␤-cell failure due to severe obesity (25). TM injections were performed in 9-week-old Pdx1 PB -CreER TM ;db/db, CAG-CAT-Mafa myc ;db/db, and ␤Mafa myc ; db/db mice, roughly when Mafa levels decline in db/db ␤-cells (1,2). In contrast, Pdx1 expression was unaffected under these circumstances, another key islet-enriched transcription factor sensitive to glucotoxic conditions associated with T2DM (supplemental Fig. S4). Transgenic Mafa myc in ␤Mafa myc ;db/db mice was clearly detectable in insulin ϩ cells within 1 week of TM induction (Fig. 1B), and expression was sustained during the remaining 8 weeks of analysis. In contrast, Mafa myc was not produced in non-diabetic (db/m) and diabetic (Pdx1 PB -CreER TM ;db/db, CAG-CAT-Mafa myc ;db/db) control islets. The total number of Mafa ϩ cells was not only much greater than in diabetic control db/db ␤-cells, but exceeded the number of Mafa myc producing cells (Fig. 1B). These results imply that Mafa myc restored expression of endogenous Mafa.
Blood glucose levels in ␤Mafa myc ;db/db mice were significantly lower than either control diabetic Pdx1 PB -CreER TM ; db/db or CAG-CAT-Mafa myc ;db/db mice ( Fig. 2A). Notably, this improvement was observed after only 1 week of TM treatment, which was associated with Mafa myc production in ␤-cells and the reduction in fed blood glucose levels in ␤Mafa myc ; db/db mice to 327.0 Ϯ 59.4 mg/dl, whereas control diabetic and  (Fig. 2, C  and D), which also manifested reduced HbA1c values (Fig. 2B).
Improved plasma glucose clearance and higher insulin secretion levels were observed in ␤Mafa myc ;db/db mice in intra-peritoneal glucose tolerance tests (Fig. 3, A and B), whereas insulin sensitivity (Fig. 3C) and body weight (supplemental Fig. 5) did  not differ between the transgenic Mafa myc expressing and nonexpressing db/db mice. In addition, both 1st and 2nd phase insulin secretion increased in isolated ␤Mafa myc ;db/db islets (Fig. 3D). Collectively, these results show that Mafa myc synthesis in 9-week-old db/db mice enabled quick recovery of ␤-cell function even in the context of persistent insulin insensitivity and obesity, resulting in better glucose utilization in peripheral tissues.
Islet ␤-Cell Mass Increases in Mafa myc Producing db/db Mice-Insulin resistance and elevated blood glucose levels are first observed at around 4 weeks in db/db mice, whereas hyperglycemia plateaus at around 12 weeks. High-level ␤-cell proliferation occurs at 4 weeks in response (2), with sustained hyperglycemia decreasing islet insulin ϩ cell mass (25). Significantly, ␤-cell mass improved in ␤Mafa myc ;db/db islets relative to diabetic controls (Fig. 4A and supplemental Fig. S6). This resulted from reduced cellular caspase-3-mediated apoptosis, and not a change in proliferation rate of islet ␤-cells (Fig. 4, B and C). These findings illustrated the importance of Mafa in preserving ␤-cell levels in T2DM islets.
Identification of Gene Products Impacted by Maintaining Mafa Expression in ␤Mafa myc ;db/db Islet ␤-Cells-We first investigated how Mafa myc influenced the expression of genes known to be regulated by this factor in ␤Mafa myc ;db/db islets, specifically insulin I, insulin II, Pdx1, and glucokinase (Gck), and Mafa itself (13). As expected from immunostaining (Fig. 1B), total Mafa mRNA was greatly induced over the non-transgenic diabetic controls, essentially now equivalent to euglycemic db/m islets (Fig. 5). Moreover, insulin 1, insulin 2, and Slc2a2 were also increased in Mafa myc expressing db/db samples, with Pdx1 trending toward recovery. In contrast, Gck expression was unaffected by diabetic conditions. These results demonstrated that functionally important Mafa-regulated gene products were limiting in diabetic db/db islet ␤-cells.
To identify other significant targets of Mafa myc control, Agilent Mouse GE 4 ϫ 44K v2 microarray analysis was performed with RNA extracted from 14-week-old ␤Mafa myc ;db/db and diabetic control Pdx1 PB -CreER TM ;db/db islets. Approximately 340 genes had a more than 2-fold level increase in Mafa myc producing islets (supplemental Appendix S1), including, as  expected, Mafa and Slc2a2. Notably, gene ontology analysis indicated that proteins involved in reducing oxidative stress in ␤-cells were up-regulated, like the glucokinase regulatory protein (Gckr) that inactivates Gck by translocating the enzyme into the nucleus (26,27), and glutathione S-transferase ␣1 (Gsta1), which catalyzes the conjugation of glutathione to various molecules for the purpose of detoxification (28,29).
The selective increase in Gckr and Gsta1 expression levels was confirmed upon comparing RNA expression in 14-weekold Mafa myc db/db islets to control db/db and db/m islets (Fig.  5A). This is likely caused by Mafa directly activating Gckr and Gsta1 transcription, because expression of both was significantly reduced upon shRNA-mediated knock-down of Mafa in mouse MIN6 ␤-cells (supplemental Fig. S7). These results imply that factors reducing db/db ␤-cell stress, such as Gckr and Gsta1, were important effectors of ␤Mafa myc ;db/db recovery.
Induction of Gckr and Gsta1 Levels Coincides with Mafa myc Production and Reduced Oxidative Stress in ␤Mafa myc ;db/db ␤-Cells-Further support of a direct role for Mafa myc in recovery of db/db ␤-cells, Gckr, and Gsta1 expression increased within 3 days of TM treatment in ␤Mafa myc ;db/db islets (Fig.  5B). Insulin 1, insulin 2, and Slc2a2 levels were also elevated within this early period, although there was not yet a change in blood glucose levels (supplemental Fig. S8). Because elevated Gckr and Gsta1 could reduce oxidative stress in ␤-cells, for example, by the Gckr protein sequestering Gck in the nucleus and reducing glucose flux (27), production of the oxidative stress marker 4-hydroxy-2-nonenal was compared in nondiabetic, diabetic, and Mafa myc containing db/db islets (supplemental Fig. S9). As predicted from elevated Gckr and Gsta1, 4-hydroxy-2-nonenal staining levels were decreased in Mafa myc islets in comparison to diabetic CAG-CAT-Mafa myc ; db/db and Pdx1 PB -CreER TM islets. However, Mafa myc islet staining was still elevated in relationship to euglycemic db/m islets, implying that restoration of Mafa alone is unable to completely prevent glucotoxicity in db/db ␤-cells. Such was expected from the difference in blood glucose and HbA1c levels between ␤Mafa myc ;db/db and db/m mice (Fig. 2B). Collectively, these results strongly suggest that Mafa myc improves db/db ␤-cell by reducing toxic stress conditions, presumably by elevating expression of factors like Gckr and Gsta1.
There Is a Marked Decrease of GSTA1 in Human T2DM Islets-The presence of MAFA in insulin ϩ cells is markedly decreased in T2DM islets (see supplemental Fig. S1 and Refs. 2 and 24). Immunostaining for GSTA1 was performed in normal and T2DM pancreatic samples to determine whether compromised expression was also associated with the disease state in humans. Gsta1 was clearly detected in normal human ␤-cells, whereas almost undetectable in T2DM islet (Fig. 6). Collectively, our analysis in db/db mice and human T2DM islets illustrates that reduction of MAFA in T2DM not only influences normal mediators of ␤-cell function and mass, but also stress effector levels in islet cells.

DISCUSSION
Transcription factors have been demonstrated to play essential roles in both the embryonic formation and adult function of islet cells in mice. The association of many of these islet-enriched proteins to MODY implies that their function is also conserved within humans (30). Notably, Mafa is distinct among this group of transcription factors in being uniquely produced at the onset of islet ␤-cell formation during development and exclusively within this embryonic and adult islet population. Moreover, gene knock-out studies have demonstrated that Mafa only affects islet ␤-cell maturation, but not euglycemia. Human MAFA has also been found to be markedly decreased in T2DM ␤-cells, with overt cell dysfunction likely caused by decreased NKX6.1 and/or PDX1, whose loss from mouse islets quickly results in hyperglycemia (2,16,19). However, because the expression and/or activity of a variety of factors were affected in islet ␤-cells under these glucotoxic conditions, it is difficult to elucidate what factor(s) is pathophysiological. In this report, we directly examined the significance of Mafa activity in insulin-resistant, diabetic db/db mice by conditionally and specifically producing Mafa myc in islet ␤-cells. Expression levels comparable with endogenous normoglycemic Mafa improved ␤-cell function, ␤-cell mass, and blood glucose levels. These results support MAFA in the molecular underpinnings of islet ␤-cell inactivation during T2DM.
The induction of Mafa myc in ␤Mafa myc ;db/db mice was initiated at 9 weeks of age to maintain production during the period when endogenous stressors would normally cause Mafa inactivation. The expression of many target genes of Mafa increased in Mafa myc ␤-cells relative to diabetic ␤-cells. Improvement of glycemic control by Mafa myc ␤-cells was produced in the context of continued production of obesity-and dyslipidemia-induced stress signals instigated by the leptin receptor mutation in db/db mice. Microarray studies were performed to identify the proteins enhancing ␤-cell activity. As expected, a variety of gene products normally required in glucose sensing and insulin secretion were elevated in Mafa myc ␤-cells, like the Slc2a1 and Slc2a2 (supplemental Fig. S10), glucose transporters essential to rodent and human glucose-stimulated insulin secretion (31). Interestingly, genes encoding proteins protective to the ␤-cell were also found, with Gsta1 and Gckr representative examples (supplemental Appendix S1). Notably, Gsta1 and Gckr expression increased immediately after Mafa myc production, prior to the reduction in blood glucose levels. Moreover, their expression was sustained throughout the time course of experimentation, as predicted of important neutralizing effectors of ␤-cell stress.
Islet ␤-cells have very low hydrogen peroxide scavenging enzyme levels relative to other cell types, like glutathione peroxidase-1 and catalase (4 -6). Significantly, transgenic ␤-cell-specific glutathione peroxidase-1 expression profoundly increased ␤-cell function in db/db mice, coinciding with recovery of nuclear Mafa and Nkx6.1 (2). The insulin secretion defects in human T2DM islets are also improved upon in vitro treatment with reactive oxygen species scavengers (7). Induction of Gsta1 and Gckr by Mafa myc would reduce oxidative stress in ␤-cells, for example, by preventing oxidation of cysteine 277 and 293 in Mafa that cause loss in cis-acting DNA binding activity (2). This could explain why endogenous Mafa levels were also increased in ␤Mafa myc ;db/db islets (Fig. 1B).
Interestingly, the mechanisms of action of these antioxidants are distinct. Thus, the Gsta1 enzyme detoxifies reactive oxygen species by direct conjugation with glutathione, whereas Gckr inhibits glucokinase by binding non-covalently to form an inactive complex (26). As glucokinase is the principal regulator of glucose flux and insulin secretion in ␤-cells (32), Gckr binding would reduce cellular metabolism, and consequentially the generation of active oxygen species. Gckr was identified as a T2DM susceptible gene in genome-wide association studies (33). The increased oxoreductase activity in ␤Mafa myc ;db/db islet ␤-cells would also reduce thioredoxin-interacting protein levels, a cellular redox regulator that both induces cell death and inhibits Mafa mRNA expression (34). GSTA1 levels were decreased in T2DM islets (Fig. 6). Consequently, it will be important to investigate precisely how expression of antioxidant factors like Gckr and Gsta1 influence ␤-cell activity. These findings suggest that screening of pharmacological agents to either enhance antioxoreductase levels is warranted for treatment of pre-and established T2DM patients. For example, could induction of antioxidant factors be protective in islet ␤-cells of obese individuals, as the majority will not become T2DM (35)? Perhaps expression is also limiting within the insulin ϩ cells produced during in vitro differentiation of human embryonic stem cells, which express many transcriptional regulators associated with mature islet cells, yet remain dysfunctional with regards to glucose responsiveness and high insulin production until able to express MAFA (36). Notably, oxidative destruction is still evident in ␤Mafa myc ; db/db islets (supplemental Fig. S9), presumably limiting ␤-cell activity and the complete restoration of blood glucose levels.
Presumably, the failure to produce normal Nkx6.1 levels in ␤Mafa myc ;db/db ␤-cells results in relative inactivity, and prevents the return to euglycemia.
Interestingly, improved blood glucose levels and islet ␤-cell function are observed upon treatment of db/db mice with either phloridzin (37) or insulin (38), which act by effecting glucose utilization (i.e. insulin) or clearance (phloridzin) in peripheral tissues. The consequential reduction in blood glucose levels likely enhances ␤-cell mass and activity by decreasing cellular glucose flux and stress effector levels. The present work strongly suggests that Mafa is a direct mediator of improved ␤-cell activity under these conditions. Thus, sustaining Mafa expression in diabetic ␤Mafa myc ;db/db ␤-cells led to lowered blood glucose levels, increased ␤-cell mass, and improved ␤-cell function. Future efforts will be directed at determining how to prevent MAFA inactivation in T2DM ␤-cells.