Palmitate Inhibits Insulin Gene Expression by Altering PDX-1 Nuclear Localization and Reducing MafA Expression in Isolated Rat Islets of Langerhans *

Abnormalities in lipid metabolism have been proposed as con-tributing factors to both defective insulin secretion from the pancreatic beta cell and peripheral insulin resistance in type 2 diabetes. Previously, we have shown that prolonged exposure of isolated rat islets of Langerhans to excessive fatty acid levels impairs insulin gene transcription. This study was designed to assess whether palmitate alters the expression and binding activity of the key reg-ulatory factors pancreas-duodenum homeobox-1 (PDX-1), MafA, andBeta2,whichrespectivelybindtotheA3,C1,andE1elementsin the proximal region of the insulin promoter. Nuclear extracts of isolated rat islets cultured with 0.5 m M palmitate exhibited reduced binding activity to the A3 and C1 elements but not the E1 element. Palmitate did not affect the overall expression of PDX-1 but reduced its nuclear localization. In contrast, palmitate blocked the stimulation of MafA mRNA and protein expression by glucose. Combined adenovirus-mediated overexpression of PDX-1 and MafA in islets completely prevented the inhibition of insulin gene expression by palmitate. These results demonstrate that prolonged exposure of islets to palmitate inhibits insulin gene transcription by impairing nuclear localization of PDX-1 and cellular expression of MafA. The prevalence of diabetes mellitus is increasing dramatically in Western countries, in part because of the increase in obesity. Type 2 diabetesmellitus,themostfrequentformofthedisease,ischaracterized by defective insulin secretion from the pancreatic beta cells and peripheral

The prevalence of diabetes mellitus is increasing dramatically in Western countries, in part because of the increase in obesity. Type 2 diabetes mellitus, the most frequent form of the disease, is characterized by defective insulin secretion from the pancreatic beta cells and peripheral insulin resistance. According to the lipotoxicity hypothesis, abnormalities in lipid metabolism contribute to both defects (1) and in particular to the inexorable decline of beta cell function observed during the course of the disease (2). However, the mechanisms of lipotoxicity in the beta cell remain largely unknown.
In vitro, prolonged exposure to excessive concentrations of fatty acids inhibits glucose-stimulated insulin secretion (3)(4)(5)(6)(7) and insulin gene expression (8 -11). Previous studies in our laboratory have shown that deleterious effects of fatty acids appear mediated by distinct mechanisms; whereas inhibition of insulin secretion is observed after culture with palmitate, oleate, and other fatty acids (7), insulin gene expression is only affected by palmitate and is mediated via de novo synthesis of ceramide (11). In isolated rat islets, we have shown that palmitate markedly blunts the activation by glucose of an insulin promoter reporter construct, indicating a transcriptional mode of action (11). However, the mechanisms by which palmitate affects the insulin promoter are unknown.
The aims of this study were 1) to examine whether prolonged exposure of isolated rat islets to palmitate alters the binding activity of MafA, PDX-1, and Beta2 to their cognate sequences on the insulin promoter and, if so, 2) to ascertain whether this effect is because of reduced expression or post-translational modifications of the transcription factors and 3) whether overexpression of the proteins prevents palmitate inhibition of insulin gene expression.
Animals-Normal 225-275-g male Wistar rats were purchased from Charles River Laboratories (Wilmington, MA). Animals were housed on a 12-h light/dark cycle with free access to standard laboratory chow and water. All procedures using animals were approved by the Pacific Northwest Research Institute Animal Care and Use Committee.
Rat Islet Isolation and Culture-Rat islets were isolated by collagenase digestion and dextran density gradient centrifugation as described previously (11) and then allowed to recover overnight in RPMI 1640 supplemented with 10% fetal bovine serum, 100 units/ml penicillin, 100 g/ml streptomycin, and 11.1 mM glucose. Experimental culture conditions for each experiment were as described under "Results." The preparation of culture media containing palmitate was as described previously (11) with a final molar ratio of palmitate/BSA of 5:1. All control conditions contained the same amount of BSA and vehicle (EtOH/H 2 O, 1:1, v/v) as those with palmitate.
Immunoblotting-Proteins (5-30 g) were separated by 10% SDS-PAGE and transferred to nitrocellulose. Membranes were blocked in 5% nonfat milk in TBS-T (10 mM Tris, pH 7.5, 150 mM NaCl, and 0. Immunohistochemistry and Confocal Microscopy-Cultured islets were washed with PBS, fixed with 4% formaldehyde for 30 min on ice, washed twice with PBS, and embedded in 80 l of liquefied HistoGel (Richard-Allan Scientific, Kalamazoo, MI). After hardening, the islet-HistoGel pellet was further fixed in 4% formaldehyde for 3 h on ice. Following standard ethanol dehydration and paraffin embedding, 0.4-m sections were mounted on slides and rehydrated. Antigen retrieval was performed using 10 mM sodium citrate heated to 90°C for 9 min and then allowed to cool to room temperature. The sections were washed with PBS twice, blocked with 5% donkey serum in PBS, 1% BSA for 1 h at room temperature, and then incubated overnight at 4°C with an anti-PDX-1 antibody diluted 1:10,000 in PBS, 1% BSA. An anti-rabbit secondary antibody conjugated to Cy5 (Jackson ImmunoResearch Laboratories) was used at 1:1000 in PBS, 1% BSA for 1 h at room temperature for detection. Nuclei of the sectioned islets were detected by the addition of 1 g/ml propidium iodide just prior to mounting with Aqua Poly/Mount (Polysciences, Warrington, PA). Images were acquired using a Fluoview 500 Olympus confocal microscope.
Electrophoretic Mobility Shift Assay (EMSA)-Double-stranded oligodeoxynucleotide probes against human insulin A3 (5Ј-CCC CTG GTT AAG ACT CTA ATG ACC CG-3Ј), Rat2 C1 (5Ј-AGC TTG GAA ACT GCA GCT TCA GCC CCT CTG-3Ј), and Rat2 E1 (5Ј-TCT GGC CAT CTG CTG ATC CA-3Ј) elements were labeled by end filling with the DNA polymerase large (Klenow) fragment kit (Promega) and [␣-32 P]dCTP (PerkinElmer Life Sciences). In addition, we used an oligodeoxynucleotide probe (Zd: 5Ј-TTT GCT CTC CTG GAG ACA TTT GCC CCC AGC TGT GAG C-3Ј) to the Z mini-enhancer of the human insulin promoter (23). Nuclear extracts (5-10 g) were incubated with 30,000 cpm of labeled probe with or without cold competitors in a final volume adjusted to 30 l with binding buffer (15 mM HEPES, pH 7.5, 60 mM KCl, 5 mM MgCl 2 , 2 mM EDTA, 12% glycerol, 3.3 mM dithiothreitol, and 100 ng of poly(dI-dC)) at room temperature for 30 min. Binding reactions were resolved on 4.5% acrylamide gels run in 0.5% TBE (44.5 mM Tris, 44.5 mM boric acid, 1 mM EDTA) for 2 h at 4°C and visualized by autoradiography. The identity of the protein in the binding complexes was determined by supershift using antibodies against PDX-1 or MafA. The bands were quantified using Optiquant System software (Packard Instrument Co.).
Ribonuclease Protection Assay (RPA)-RPAs were carried out using the direct protect lysate RPA kit (Ambion, Austin, TX) and [␣-32 P]UTPlabeled probes generated from a 360-bp sequence of the rat II preproinsulin gene and a conserved 245-bp sequence of the mouse ␤-actin gene. Protected fragments were resolved on 5% denaturing gels as described previously (11). The probes were determined to be in excess for each experiment. The bands were quantified using the Optiquant System software (Packard Instrument Co.).

Generation of Recombinant Adenoviruses and Islet Transduction-
Recombinant adenoviruses (Ad) encoding PDX-1 and MafA under the cytomegalovirus promoter were generated by the Pacific Northwest Research Institute Adenoviral Core Facility, as described previously (11) using a mouse PDX-1 cDNA provided by M. Montminy (Salk Institute for Biological Studies, La Jolla, CA) and a mouse MafA cDNA provided by R. Stein (Vanderbilt University). A recombinant Ad expressing firefly luciferase under the control of the cytomegalovirus promoter was used as a control. The appropriate titer for each recombinant Ad was assessed in preliminary experiments by infecting 100 islets with increasing viral concentrations (from 10 5 to 10 10 plaque-forming units (pfu)/ islet) overnight. The islets were then washed in PBS prior to initiating experimental conditions.
Statistics-Data are expressed as mean Ϯ S.E. Intergroup comparisons were performed by Student's paired t test or analysis of variance with post-hoc Dunnett t test where appropriate. p Ͻ 0.05 was considered significant.

Palmitate Inhibits MafA and PDX-1 Binding to Their Cognate
Sequences on the Insulin Promoter in Isolated Islets-Isolated rat islets were cultured for 24, 48, or 72 h with 2.8 or 16.7 mM glucose in the absence or presence of 0.5 mM palmitate. Binding of islet nuclear extracts to the A3, C1, E1, and Zd probes was assessed by EMSA. Fig. 1A shows a representative gel after a 72-h culture. Discrete complexes were observed with each probe and disappeared in the presence of excess unlabeled probes. The presence of MafA and PDX-1, respectively, in the complexes formed with the C1 and A3 probes was confirmed by supershifting the bands with specific antibodies.
The presence of 16.7 mM glucose visibly enhanced the intensity of the binding complex to the C1 probe and, to a lesser extent, to the A3 and E1 probes. The addition of 0.5 mM palmitate markedly reduced binding to the C1 and A3 elements in the presence of 16.7 mM glucose (Fig. 1A). Binding to the E1 and Zd sites was not affected by palmitate. The loss of binding to the A3 and C1 elements in the presence of 0.5 mM palmitate at elevated glucose (16.7 mM) was detectable after 24 h of culture and statistically significant after 48 h (p Ͻ 0.05) and 72 h (p Ͻ 0.001) (Fig. 1B). These results demonstrate that palmitate specifically impairs the binding activity of MafA and PDX-1 to their cognate sequences on the insulin promoter in the presence of high glucose levels in beta cells, consistent with our previous observation that palmitate blunts glucose activation of an insulin promoter-reporter (11). Palmitate Alters PDX-1 Nuclear Localization and MafA Gene Expression-To determine whether the loss of binding activity observed in Fig. 1 was because of diminished protein expression or altered localization, expression of PDX-1, MafA, and Beta2 in whole-cell, nuclear, or cytosolic protein extracts of rat islets cultured for 24 h in 2.8 or 16.7 mM glucose in the absence or presence of 0.5 mM palmitate was assessed by immunoblotting ( Fig. 2A).
Expression of PDX-1 in whole-cell lysates was not significantly affected by either glucose or palmitate ( Fig. 2A). At 16.7 mM glucose, PDX-1 was predominantly localized in the nuclear fraction as shown by Western blotting and immunohistochemistry (Fig. 2, A and B). In the presence of palmitate, however, the expression of PDX-1 in the nucleus was reduced (Fig. 2, A-C). These results suggest that palmitate impairs the translocation of PDX-1 from the cytosol to the nucleus that normally occurs upon glucose stimulation (26,27). At 2.8 mM glucose, MafA expression was undetectable in whole-cell lysates by immunoblotting ( Fig. 2A). In contrast, at 16.7 mM glucose MafA protein was readily detected, demonstrating a strong stimulation of MafA protein expression by glucose. In the presence of palmitate, glucose-induced MafA protein expression was reduced, on average, by 44 Ϯ 3% (n ϭ 7; p Ͻ 0.001) in whole-cell lysates (Fig. 2, A and D). MafA protein was undetectable in the cytosolic fraction, and its levels in the nuclear fractions were reduced by 47 Ϯ 7% by palmitate (n ϭ 5; p Ͻ 0.01; Fig. 2, A and D), in parallel with its reduced expression in whole-cell lysates. These results indicate that palmitate inhibits glucose-induced MafA protein expression. Neither Beta2 nor the control protein transcription factor IID was affected by palmitate.
To further investigate the mechanisms by which palmitate affects protein expression, MafA and PDX-1 mRNA levels were measured in islets after a 24-h culture in 2.8 or 16.7 mM glucose with or without 0.5 mM palmitate (Fig. 3A). Glucose significantly increased MafA mRNA expression (n ϭ 7; p Ͻ 0.01). In the presence of palmitate, the stimulatory effect of glucose was no longer significant (Fig. 3A). The time course of this inhibitory effect is shown in Fig. 3B. In contrast, neither glucose nor palmitate had a significant effect on the expression of PDX-1 mRNA (Fig. 3A), consistent with the unaltered total protein expression observed (Fig. 2, A and C). These results indicate that palmitate affects PDX-1 and MafA binding activities by two distinct mechanisms: PDX-1 is altered at the post-translational level in its nuclear localization, whereas palmitate inhibits glucose stimulation of MafA mRNA levels.
Adenoviral Co-expression of PDX-1 and MafA Prevents Palmitate Inhibition of Insulin Gene Expression-To further explore the roles of PDX-1 and MafA in mediating palmitate inhibition of insulin gene expression, we next sought to determine whether overexpression of these transcription factors could restore insulin mRNA expression in the presence of palmitate. First, isolated islets were infected with adenoviruses encoding for PDX-1 or MafA at a titer of 10 5 pfu/islet, and expression of the proteins was examined by immunoblotting after 24, 48, or 72 h. As shown in Fig. 4A, protein expression was sustained for up to 72 h. Next, to determine whether the overexpressed proteins were functional, islets were infected with increasing titers of each virus, and the binding activity of islet nuclear extracts to the A3 and C1 elements was examined by EMSA. As shown in Fig. 4B, binding of overexpressed PDX-1 and MafA to their respective sequences exceeded endogenous levels and increased with increasing viral titers. For subsequent experiments, the lowest titer of virus (10 5 pfu/islet) was used. Islets infected with PDX-1, MafA, or both were then cultured for 72 h in the presence of 16.7 mM glucose with or without 0.5 mM palmitate. Insulin gene expression was measured by RPA (Fig. 4C) and quantified (Fig. 4D). As shown previously (10, 11), palmitate significantly decreased insulin gene expression after 72 h in control, luciferase-infected islets (p Ͻ 0.01). Overexpression of MafA alone increased insulin gene expression relative to that of the luciferase-infected control islets at 16.7 mM glucose. Paradoxically, overexpression of PDX-1 reduced insulin mRNA levels.
In the presence of palmitate, neither MafA nor PDX-1 alone was able to prevent the decrease in insulin gene expression (Fig. 4C). However, combined overexpression of MafA and PDX-1 completely prevented the inhibitory effect of palmitate on insulin gene expression (Fig. 4, C  and D), confirming the key role of these two transcription factors in the mechanisms of lipotoxicity on the insulin gene.

DISCUSSION
The aims of this study were to determine whether prolonged exposure of isolated rat islets to palmitate alters the binding activity of MafA, PDX-1, and Beta2 to the insulin promoter, to ascertain whether this effect was due to reduced expression or post-translational modifications of these transcription factors, and to determine whether overexpression of these proteins could prevent palmitate inhibition of insulin gene expression. Our results show that exposure of rat islets to elevated levels of palmitate inhibits binding of MafA and PDX-1 but not Beta2 to the insulin promoter. Interestingly, the mechanisms of this inhibition differ in that palmitate alters the cellular localization of PDX-1 by preventing its translocation to the nucleus, whereas glucose-stimulated MafA mRNA expression is blunted, resulting in reduced overall MafA protein levels. Importantly, palmitate inhibition of insulin gene expression is prevented by combined overexpression of MafA and PDX-1. We (9 -11) and others (8,28,29) have shown previously that concurrent exposure to chronically elevated levels of glucose and palmitate inhibits insulin gene expression. Importantly, this inhibitory effect of palmitate was observed to occur at the transcriptional level (11). We now demonstrate that the mechanisms of this transcriptional inhibition involve two key transcription factors of the insulin gene, PDX-1 and MafA. Although reduced binding of PDX-1 and MafA to the A3 and C1 elements, respectively, was demonstrated indirectly by EMSA (Fig. 1), the observation that palmitate decreases nuclear localization of PDX-1 and mRNA expression of MafA indicates that the major site of action of palmitate lies upstream of the binding activities of the transcription factors.
Insulin gene expression is highly regulated by glucose, which coordinately enhances both the rate of transcription and mRNA stability (14). Several studies have shown that binding to the A3 and C1 elements is modulated by glucose (reviewed in Ref. 14). A major site of glucose regulation of PDX-1 function is at the post-translational level (14). Although the precise nature of the post-translational modifications is still debated, glucose promotes nuclear translocation of PDX-1 (26, 27, 30 -34) and its binding to the A3 element (27,30). We observed a glucose-induced increase in binding to the A3 element, as reported previously (14,17,27,35) in adult islets, which was abolished in the presence of palmitate (Fig. 1). Also in line with prior studies (26,36,37), we observed no glucose-induced increase in total PDX-1 protein expression (Fig. 2, A and C). Palmitate did not inhibit PDX-1 expression but significantly reduced its nuclear localization (Fig. 2, A-C). Interestingly, at 48 h both Gremlich et al. (28) and Yoshikawa et al. (29) observed that PDX-1 mRNA expression was substantially reduced in the presence of palmitate. In our experiments, we did not observe either a glucose-induced increase or a palmitate-mediated decrease in PDX-1 mRNA expression (Fig. 3A). Rather, the most compelling effect of palmitate after 24 h of exposure was at the level of PDX-1 nuclear localization (Fig.  2, A-C). Because PDX-1 is thought to regulate its own transcription (38,39), we postulate that the early defect seen in the presence of palmitate is a sequestration in the cytosol, which, at later time points, might also result in decreased overall cellular expression of PDX-1, as observed by Gremlich et al. (28) and Yoshikawa et al. (29). Consistent with this hypothesis, Kharroubi et al. (40) failed to observe a decrease in PDX-1 expression in both INS-1 cells and rat islets after 24-h treatment with palmitate.
Binding to the C1 element was markedly increased in response to glucose (Fig. 1). MafA has only recently been identified as the activator of the C1 element (18,20) and is localized exclusively in the nucleus of pancreatic beta cells ( Fig. 2A and Ref. 20). At low glucose concentrations, MafA protein levels are essentially undetectable ( Fig. 2A and Ref. 19). As shown here with rat islets (Figs. 2 and 3) and previously in MIN6 cells (19), glucose significantly enhances mRNA and protein expression of MafA. Importantly, our results demonstrate for the first time that palmitate inhibits glucose stimulation of MafA mRNA (Fig. 3) and, as a consequence, its protein expression (Fig. 2), providing an additional mechanism whereby this fatty acid impairs insulin gene transcription.
As a trans-activator of the insulin gene, MafA alone has a much greater effect than either PDX-1 or Beta2 (19). In this study, adenoviral overexpression of MafA enhanced insulin gene expression, suggesting that MafA might be a factor limiting transcription of the insulin gene in rat islets. This is further supported by the recent observation that MafA binds to and activates elements of the PDX-1 promoter (41). In contrast, overexpression of PDX-1 actually reduced insulin mRNA levels in islets  (Fig. 4C). This phenomenon has been described previously (17,42) and was suggested to result from the titration of co-activators by the overexpressed protein (17). Importantly, combined overexpression of MafA and PDX-1 completely prevented palmitate inhibition of insulin gene expression. This observation identifies MafA and PDX-1 as critical targets of lipotoxicity at the level of the insulin gene.
We previously demonstrated that palmitate inhibition of insulin gene expression requires de novo synthesis of ceramide (11). The results of the present study raise the question as to how ceramide might affect PDX-1 and MafA. Although the signaling mechanisms of PDX-1 regulation by glucose are still debated, involvement of the phosphatidylinositol 3-kinase pathway has been suggested (27,(31)(32)(33)43). Because ceramide has been shown to interfere with this pathway, particularly at the level of protein kinase B in all cell types tested thus far (44), it is tempting to speculate that ceramide might dampen a stimulatory signal from glucose to PDX-1 via the phosphatidylinositol 3-kinase pathway. Alternatively, ceramide may activate c-Jun NH 2 -terminal kinase in beta cells as it does in other cells (45,46). In beta cells, c-Jun NH 2 -terminal kinase activation has been shown to inhibit PDX-1 binding (47) and nuclear localization (48). These hypotheses are currently under investigation in our laboratory. The mode of regulation of MafA by glucose is essentially unknown. Our results suggest that palmitate inhibition of MafA expression is at the transcriptional level. Elucidating the mechanisms of these effects awaits characterization of the promoter/enhancer region of the MafA gene.
In conclusion, this study uniquely demonstrates that the mechanisms by which palmitate impairs insulin gene transcription involve reduced binding of PDX-1 and MafA to the insulin promoter. Interestingly, both factors are affected at different levels by palmitate: PDX-1 is affected post-translationally in its nuclear localization, whereas MafA is affected at the transcriptional level. The importance of these observations for the mechanisms of lipotoxicity is illustrated by the complete prevention of the decrease in insulin gene expression in the presence of palmitate observed after combined overexpression of PDX-1 and MafA. Although the contribution of fatty acid inhibition of insulin gene transcription to beta cell dysfunction in human type 2 diabetes remains to be established, it has been shown in rodents that intact regulation of insulin gene expression by glucose is required to maintain adequate intracellular insulin stores in the face of increased demand (49). Identification of the precise molecular targets of fatty acids should facilitate the design of new therapeutic strategies to prevent the deterioration of beta cell function during the course of type 2 diabetes.