Calcineurin/Nuclear Factor of Activated T Cells and MAPK Signaling Induce TNF-α Gene Expression in Pancreatic Islet Endocrine Cells*

Cytokines contribute to pancreatic islet inflammation, leading to impaired glucose homeostasis and diabetic diseases. A plethora of data shows that proinflammatory cytokines are produced in pancreatic islets by infiltrating mononuclear immune cells. Here, we show that pancreatic islet α cells and β cells express tumor necrosis factor-α (TNF-α) and other cytokines capable of promoting islet inflammation when exposed to interleukin-1β (IL-1β). Cytokine expression by β cells was dependent on calcineurin (CN)/nuclear factor of activated T cells (NFAT) and MAPK signaling. NFAT associated with the TNF-α promoter in response to stimuli and synergistically activated promoter activity with ATF2 and c-Jun. In contrast, the β-cell-specific transcriptional activator MafA could repress NFAT-mediated TNF-α gene expression whenever C/EBP-β was bound to the promoter. NFAT differentially regulated the TNF-α gene depending upon the expression and MAPK-dependent activation of interacting basic leucine zipper partners in β cells. Both p38 and JNK were required for induction of TNF-α mRNA and protein expression. Collectively, the data show that glucose and IL-1β can activate signaling pathways, which control induction and repression of cytokines in pancreatic endocrine cells. Thus, by these mechanisms, pancreatic β cells themselves may contribute to islet inflammation and their own immunological destruction in the pathogenesis of diabetes.

Cytokines are small secreted or membrane-bound signaling proteins classically known for their roles in regulating immunity, inflammation, and hematopoiesis (1). They are central to the growth and differentiation of leukocytes and required for eliciting an immune response. Although they have been detected in numerous cell types, a majority are predominantly expressed in T cells and macrophages. Proinflammatory cytokines contribute to inflammation of the islets, which can result in the targeted destruction of islet ␤ cells in type 1 diabetes or impairment and dysfunction of islet ␤ cells in type 2 diabetes.
The pathogenesis of type 1 diabetes has been associated with an asymptomatic accumulation of inflammatory mono-cytes and lymphocytes around islets, known as peri-insulitis (2)(3)(4)(5)(6). Upon progression to insulitis, macrophages, dendritic cells, T lymphocytes, and B lymphocytes eventually infiltrate the islet and mediate the selective destruction of pancreatic ␤ cells. Several cytokines expressed by mononuclear immune cells have been identified in islets that play a major role in insulitis and destruction of ␤-cells, including IL-1␤, TNF-␣, interferon ␥ (IFN-␥), and interleukin-6 (IL-6) (7)(8)(9)(10)(11). However, the events initiating an inflammatory response and invasion by islet infiltrates have not been clearly defined. Moreover, it is still unknown why ␤ cells are chosen as the primary target of immune destruction, whereas other islet endocrine cells remain intact (12,13).
Recent studies indicate that ␤ cells themselves express cytokines and chemokines that may have local effects on the islet or resident macrophages. Microarray studies show that IL-1␤, IFN-␥, and TNF-␣ can induce proinflammatory cytokine and chemokine mRNA expression in RINm5F and INS-1 ␤-cell lines and purified rat and human ␤ cells (14 -17). In addition, up-regulated IL-1␤ mRNA expression has been observed in purified ␤ cells from diabetic human donors (18). We sought to identify conditions that induce cytokine expression in pancreatic islet endocrine cells, and we now show that glucose and IL-1␤ induce cytokine mRNA and protein in both ␣ and ␤ cells. We further identify signaling requirements and mechanisms of IL-1␤-induced activation of the TNF-␣ gene in pancreatic ␤ cells.
CN/NFAT-mediated transcription also often requires converging MAPK pathways to activate basic leucine zipper * This work was supported, in whole or in part, by National Institutes of (bZIP) proteins, which cooperate with NFAT to regulate promoter activity (26). A classic example of this is the regulation of the IL-2 gene. Expression of IL-2 requires T cell receptor activation of tyrosine kinases Lck and ZAP70 that induces a calcium transient to activate CN/NFAT and stimulates Rasdependent MAPKs. MAPKs in turn activate AP-1 bZIP dimerized proteins c-Fos and c-Jun. AP-1 interacts with NFAT on multiple NFAT-AP1 composite sites within the IL-2 promoter to regulate gene expression. Similarly, NFAT interacts with bZIP Maf proteins to selectively regulate IL-4 expression in T helper 2 cells and insulin gene expression in pancreatic ␤ cells (33)(34)(35)(36).
The TNF-␣ gene is selectively regulated by NFAT and ATF2/Jun bZIP heterodimer in B cells and T cells (37,38). In T cells, ATF2/Jun bind to the cyclic AMP-response element of the TNF-␣ gene promoter and cooperate with NFAT, which binds to multiple NFAT sites, including the adjacent 3 element to regulate transcriptional activity (37). NFAT also binds to an alternative site in B cells and regulates the TNF-␣ promoter with ATF2/Jun in a noncooperative manner (38). C/EBP-␤ and c-Jun have also been identified to induce TNF-␣ expression independent of NFAT in myelomonocytic cells (39).
Here, we show that activation of NFAT and ATF2/c-Jun by CN and p38 JNK, respectively, results in the expression of TNF-␣ in pancreatic ␤ cells. In contrast, activation of the islet ␤-cell-enriched bZIP MafA disrupts NFAT-mediated TNF-␣ gene expression. These studies provide mechanistic insight as to how NFAT differentially regulates expression of genes in normal ␤ cell maintenance (the insulin gene for example), while mediating repression of cytotoxic genes by similar signaling pathways. They also elucidate key signaling components in ␤ cells that induce cytokine expression and potentially contribute to islet dysfunction and diabetes.

Cell Culture and Islet Tissue Isolation
MIN6 and INS-1 cells were cultured in DMEM and RPMI 1640 medium (Invitrogen), respectively, with 10% heat-inactivated fetal bovine serum (FBS), 10 mM HEPES, pH 7.4, 2 mM L-glutamine, 1 mM sodium pyruvate, 50 mM ␤-mercaptoethanol and penicillin (100 units/ml), and streptomycin (100 g/ ml) at 37°C in 95% air, 5% CO 2 ). ␣TC1 cells and RAW264.7 were also maintained in DMEM with the supplements described above. ␣TC1 cells were also supplemented with 1ϫ nonessential amino acids (Sigma). Human islets were obtained from ICR Basic Science Islet Distribution Program and from the Islet Cell Processing Laboratory of the Islet Cell Transplant Program at Baylor University Medical Center, Dallas. Islet preparations were stained with 2 mg/ml diphenylthiocarbazone and islet were isolated to Ͼ95% purity. Pancreatic tissue that did not stain with diphenylthiocarbazone was retained and used for nonendocrine pancreatic tissue controls. Islets were cultured in RPMI 1640 or Krebs-Ringer bicarbonate HEPES buffer (KRBH) media according to experimental design.

Immunoblot Analysis
Cells were harvested in lysis buffer containing 50 mM HEPES, pH 7.5, 150 mM NaCl, 1% Triton X-100, 0.2 mg/ml phenylmethylsulfonyl fluoride (PMSF), 0.1 M NaF, 2 mM Na 3 VO 4 , 10 g/ml aprotinin, 5 g/ml pepstatin A, and 5 g/ml leupeptin. Islets were aspirated through a needle for complete cell lysis. Lysate protein (20 g/lane) was subjected to SDS-PAGE, electrotransferred to nitrocellulose membranes, and blocked with 5% milk in Tris-buffered saline containing 0.1% Tween 20 (TBST), pH 7.4, for 2 h. Primary antibody was then added, and the membrane was incubated for 1 h at 4°C and washed three times with TBST. Membranes were then incubated with secondary antibodies conjugated to DyLight 680 and 800 fluorescent dyes (Cell Signaling) to detect proteins using an ODYSSEY Infrared Imaging System (LI-COR).

Flow Cytometry
␤-Cell Purification-Pancreatic ␤-cells were purified from islets as described previously (41). Isolated human islets were dissociated with 0.25% trypsin, 0.02% EDTA and gentle pipetting for 10 min at 37°C. Dissociated islets were washed with PBS and stained with 25 M Newport Green DCF diacetate for 30 min at 37°C and subjected to fluorescence-activated cell sorting (FACS) on a MoFlo high speed sorter. Newport Green-stained cells (␤-cell fraction) and the remaining cells (non-␤ cell fraction) were collected in KRBH and analyzed for islet hormone and immune cell marker expression.
Purification of MafA shRNA-transfected Cells-INS-1 cells were transfected with shMafA-GFP and trypsinized after 24 h. Fluorescent cells were FACS-purified on a MoFlo high speed sorter.
TNF-␣ Protein Expression-To analyze cells expressing TNF-␣ protein, MIN6 and ␣TC1 were dissociated using Accutase (Innovative Cell Technologies) and fixed with 70% eth-anol. Dissociated cells were stained with anti PE-TNF-␣ antibody and subjected to flow cytometry on a FACScan analyzer.

Enzyme-linked Immunosorbent Assay (ELISA)
TNF-␣ protein was quantified by collecting media from treated cells, concentrating the samples up to 10-fold by Microcon YM-10 centrifugal filter devices (Fisher), and performing an ELISA with a mouse TNF␣ single analyte ELISArray kit (SABiosciences).

Luciferase Assays
Transfected cells were treated and harvested with 1ϫ Reporter Lysis Buffer and assayed with the Dual-Luciferase reporter assay system (Promega) using a TD20/20 bioluminometer (Turner Designs).

cDNA Synthesis
TRI Reagent (Ambion) was used to isolate total RNA from cells. Islets required needle aspiration for complete cell lysis. RNA was reverse-transcribed to cDNA by random primers using the high capacity cDNA reverse transcription kit (Invitrogen).

Real Time Quantitative PCR
DNA recovered from ChIP assays and reverse-transcribed cDNA from treated cells were analyzed by PCR using a Bio-Rad CFX-96 real time PCR system with SYBR Green or Taq

Islet Transplantation in Mice
Nude mice were rendered diabetic using 180 mg/kg streptozotocin. Mice were considered diabetic when fasting blood glucose was greater than 200 mg/dl for 3 or more consecutive days. Diabetic nude mice were then transplanted with 2000 human islet equivalents under the kidney capsule for 3 days or until normoglycemic. Post-transplant recipient mice were then treated with IL-1␤ in the presence or absence of p38 JNK inhibitors for 6 h. After the 6-h treatment period, the transplanted islet cells were recovered from the kidney capsules, and total islet cell RNA was isolated. TNF-␣ and insulin mRNA expression was measured by QPCR using human-specific TaqMan probes.

RESULTS
Pancreatic ␤ Cells Express Cytokines in Response to IL-1␤-IL-1␤ expression has been observed in both type 1 and 2 prediabetic pancreatic islet tissue (42,43). We postulated that IL-1␤ induces expression of cytokines in islet endocrine cells that may contribute to or exacerbate islet inflammation. By real time QPCR, we observed that IL-1␤ (20 ng/ml) acutely (within hours) induced cytokine gene expression in MIN6 ␤-cell and ␣TC1 ␣-cell lines (data not shown). To confirm these results in primary tissue, we purified ␤ cells by FACS from isolated and dissociated human pancreatic islets and analyzed them for expression of cytokines, islet hormones, and immune cell markers as described below.
Primer probes were designed to detect immune cell marker expression by QPCR across human and mouse species to identify the presence of immune cells in islet cell fractions. The probe targets included F4/80 (expressed in macrophages), CD3 (expressed in T cells), and ITGB2 (a subunit of integral cell-surface proteins broadly expressed in immune cells). As expected, F4/80 was expressed over 600-fold in RAW264.7 macrophages compared with Jurkat T cells, ␣TC1, MIN6, and human islets (data not shown). ITGB2 was expressed in both RAW264.7 and Jurkat over 800-and 1800fold higher, respectively, than in ␣TC1, MIN6, and human islets. CD3 was only expressed in Jurkat and not detectable (Ct Ͼ40) in RAW264.7, ␣TC1, MIN6, or human islets. Glucagon and a relatively small quantity of insulin were detected in the non-␤-cell fraction, but only insulin was expressed in the ␤-cell fraction (Fig. 1A). Although ITGB2 and F4/80 were expressed in dissociated, pre-sorted islets and the sorted non-␤-cell fraction, none of the immune cell markers were detectable in the purified ␤ cells (Fig. 1A). The results showed that the probes tested were selective for immune cell markers expressed in their respective cell types and that the ␤-cell fractions were purified to the extent that immune cell markers were no longer detectable by QPCR.
CN/NFAT Signaling Is Involved in TNF-␣ Gene Expression in ␤ Cells-Expression of many cytokines requires CN/NFAT signaling in immune system cells (37, 38, 44 -46). Therefore, we sought to determine whether cytokines are regulated by CN/NFAT signaling in pancreatic ␤ cells. IL-1␤ induced mRNA expression of several cytokines in purified human ␤ cells, including TNF-␣, TNF-␤, IL-1␤, IFN-␥, and MCP-1 (Fig. 1B). In each case, 100 nM FK506 repressed cytokine induction, indicating that CN regulates cytokine gene expression in ␤ cells. These results were also observed in MIN6 and ␣TC1 cells (data not shown). Of the cytokines tested, TNF-␣ induction was most pronounced (over 200-fold) in purified ␤ cells after 4 h of treatment with IL-1␤ (Fig. 1B). TNF-␣ has been clinically implicated in the pathogenesis and progression of both major forms of diabetes (47,48). Moreover, targeted expression of TNF-␣ in ␤ cells in nonobese diabetic transgenic mice results in insulitis and can induce or accelerate diabetes (49 -52). Therefore, we further investigated signaling mechanisms required for the induction of TNF-␣ expression in ␤ cells.
To determine whether IL-1␤ could induce TNF-␣ protein in ␤ cells, we used flow cytometry to detect expression of TNF-␣ in MIN6 (Fig. 1C). Cells treated with IL-1␤ in the presence of TAPI-1 (to prevent TNF-␣ release) showed more than an 8-fold increase in number of ␤ cells expressing TNF-␣. MIN6 cells also secreted TNF-␣ protein into the medium within 4 h of treatment of IL-1␤ (Fig. 1D). In each case, the effect of IL-1␤ was reduced by the CN inhibitor. The requirement of CN for maximal expression of TNF-␣ in ␤ cells suggested that NFAT may be involved. We therefore tested the requirement of NFAT to induce TNF-␣ promoter activity in ␤ cells. Promoter-reporter assays revealed that IL-1␤ induces TNF-␣ promoter activity almost 8-fold in MIN6 and INS-1 cells (Fig. 1E). Exogenous expression of a dominant negative NFAT truncation mutant (dnNFAT) ablated induction of TNF-␣ promoter activity in MIN6 and INS-1 cells by IL-1␤. Moreover, a mutated version of the truncation mutant (dnNFATm), in which the inhibitory PXIXIT motif was changed to AXAXAA, had little or no effect on TNF-␣ promoter activity. Together, these data show that CN/NFAT signaling is required for induction of TNF-␣ gene expression in ␤ cells.
Glucose Sensitizes ␤ Cells to IL-1␤-induced TNF-␣ mRNA Expression-We previously showed that glucose can regulate NFAT-mediated gene expression in pancreatic ␤ cells (36,53,54). Therefore, we tested the effect of glucose on TNF-␣ gene expression in ␤ cells. Human islets cultured in high (16 mM) glucose for 3 days expressed 2.5-fold TNF-␣ mRNA compared with cells cultured in basal (4.5 mM) glucose ( Fig. 2A). TNF-␣ mRNA was detected in islets in basal glucose in the late cycles of QPCR (Ct ϳ37-39), corresponding to near background levels of mRNA detection. Insulin (100 nM) had little effect on TNF-␣ mRNA, suggesting that the modest increase in glucose-induced TNF-␣ mRNA was not attributed to glucose-induced insulin secretion (Fig. 2B). Islets treated with IL-1␤ for 4 h in the presence of high glucose resulted in greater than 150-fold increase in TNF-␣ expression compared with islets cultured in basal glucose (Fig. 2A). The effect of IL-1␤ was almost 40-fold higher than that of high glucose alone, but it produced only approximately half of TNF-␣ mRNA compared with IL-1␤ and high glucose combined. Time course analyses of IL-1␤-induced TNF-␣ expression in MIN6 and ␣TC1 cells revealed that TNF-␣ mRNA was highest (up to 1000-fold) around 2 h and drastically reduced by 4 h (Fig. 2C). High glucose enhanced TNF-␣ expression up to 2000-fold in both MIN6 and INS-1 cells within 1 h of treatment of IL-1␤ (Fig. 2, D and E). The enhanced effect was not observed by 100 nM insulin and was diminished by 4 h. The results show that high glucose enhances acute induction of TNF-␣ mRNA expression in islet endocrine cells and that the response is rapidly reversed within 4 h of IL-1␤ treatment.
NFAT Associates with the TNF-␣ Promoter in Islet Endocrine Cells-NFAT was required for IL-1␤-induced TNF-␣ promoter activity. Therefore, we sought to determine whether NFAT associated with the TNF-␣ promoter in islet endocrine cells in response to glucose and IL-1␤. ChIP assays revealed that treatment with either 16 mM glucose or IL-1␤ resulted in NFAT association with the TNF-␣ promoter within 20 min in islets and MIN6 cells (Fig. 3A). IL-1␤ also stimulated NFAT to associate with the promoter in ␣TC1 and RAW264.7, whereas 16 mM glucose had no effect. Interestingly, lowering glucose to 2 mM was sufficient to induce NFAT-TNF-␣ promoter association in ␣TC1 cells and islets. None of the tested conditions were capable of stimulating the association of NFAT with the TNF-␣ promoter in nonendocrine pancreatic cells. We also used ChIP assays to determine whether NFAT associated with cytokine promoters in primary ␤ cell tissue (Fig. 3B). NFAT associated with the TNF-␣ promoter in purified human ␤ cells in response to IL-1␤ and K ϩ -induced depolarization. The response to depolarization suggested that a calcium transient is sufficient to induce NFAT promoter association in ␤ cells. In each case, NFAT association was sensitive to the CN inhibitor, FK520. Overall, these data suggest that CN/NFAT signaling is required to regulate the TNF-␣ promoter in ␤ cells.
To understand mechanistically how glucose and IL-1␤ induce TNF-␣ gene expression in islet cells, we performed ChIP assays on the TNF-␣ promoter for transcription factors ATF2, c-Jun, and C/EBP-␤, which are known to regulate TNF-␣ gene transcription. Time course analyses showed that glucose alone induced a transient association (between 5 and 15 min) of NFAT with the TNF-␣ promoter in islets (Fig. 3C). NFAT association was sustained for up to 1 h upon addition of IL-1␤. The stabilized NFAT-TNF-␣ promoter complex in response to IL-1␤ correlated with the sustained association of ATF2. c-Jun was also present on the promoter within 1 h. These results suggest that high glucose transiently activates NFAT but is not sufficient to recruit complement transcription factors required for TNF-␣ promoter activation. IL-1␤ stimulates the recruitment of ATF2 and c-Jun to the TNF-␣ promoter and stabilizes the NFAT-TNF-␣ promoter complex.

IL-1␤-induced TNF-␣ Expression in Islet Endocrine Cells
Requires Both p38 and JNK-We previously reported that glucose primarily activates ERK1/2, and IL-1␤ activates all three MAPKs ERK1/2, p38, and JNK in ␤ cells (55,56). We confirmed these results in INS-1 cells (Fig. 4A) and sought to determine whether MAPKs were required for TNF-␣ expression by using inhibitors for ERK1/2 (U0126), p38 (SB203580), and JNK (SP600125). Human islets and INS-1, MIN6, and ␣TC1 cell lines showed only marginal changes in IL-1␤-in-duced TNF-␣ mRNA expression when treated with 10 M U0126, SB203590, or SP600125 singly. In contrast, a combination of p38 and JNK inhibitors potently suppressed TNF-␣ gene expression (Fig. 4B). To determine whether blocking p38 JNK could prevent expression of TNF-␣ in islet endocrine cells in vivo, we transplanted isolated human islets into nude mice and treated them with IL-1␤ (1 g/kg) for 6 h in the absence or presence of 39 and 30 mg/kg of p38 and JNK inhibitor, respectively. IL-1␤ induced TNF-␣ mRNA expression in the transplanted islets, which was largely reduced by combined p38 and JNK inhibitors (Fig. 4C). This contrasted with insulin mRNA, which was relatively unaffected under these conditions. Effects of p38 and JNK inhibitors to suppress TNF-␣ protein expression were also observed in MIN6 and ␣TC1 (Fig. 4, D and E). The results indicate that induction of TNF-␣ by IL-1␤ in ␤ cells is regulated by both p38 and JNK.
MAPKs Differentially Regulate Association of bZIP Transcription Factors with the TNF-␣ Promoter-To identify downstream factors that activate the TNF-␣ gene in response to IL-1␤, we used ChIP assays to analyze association of transcription factors known to bind to and regulate the TNF-␣ promoter. NFAT, c-Jun, and ATF2 were all bound to the TNF-␣ promoter in islets within 1 h of IL-1␤ treatment (Fig.  5A). Both c-Jun and ATF2 were no longer present when C/EBP-␤ associated with the promoter between 2 and 4 h. Islet-enriched factors BETA2 and PDX-1 were not present within the times tested. Interestingly, the ␤ cell-enriched bZIP factor MafA was associated with the TNF-␣ promoter with the same kinetics as C/EBP-␤. To assess requirements of MAPKs for factor association, we did quantitative ChIP analysis of the TNF-␣ promoter in INS-1 cells treated with IL-1␤ for 1 and 4 h in the presence of MAPK inhibitors. Whereas c-Jun, ATF2, and MafA association was relatively resistant to MAPK inhibitors singly, each factor exhibited distinct MAPK requirements for TNF-␣ promoter association (Fig. 5B). Association of c-Jun was sensitive to both U0126/SP600125 and SB203590/SP600125, indicating a requirement for either ERK1/2 and JNK or p38 and JNK. NFAT and ATF2 association also required p38 and JNK. MafA association required ERK1/2 and p38. On the other hand, inhibitor of ERK1/2 alone was sufficient to prevent C/EBP-␤ association. The requirement of multiple inhibitors to prevent induction and transcription factor association indicated that there is overlap in the activation of factors by MAPKs that regulate TNF-␣ expression. These results suggest a complex requirement of MAPK signaling to regulate the activity of factors bound to the TNF-␣ promoter in islet endocrine cells.
MafA Represses TNF-␣ Promoter Activity-To determine whether these transcription factors could regulate the TNF-␣ promoter, we co-transfected them into INS-1 cells with a Ϫ1311-bp TNF-␣ promoter-reporter. NFAT alone activated the promoter almost 15-fold (Fig. 6A). The result was heightened severalfold more when NFAT was co-expressed with ATF2 and c-Jun. Expression of C/EBP-␤ alone or with NFAT had relatively little effect. MafA, however, potently repressed NFAT-mediated TNF-␣ promoter activity. These data sup-port that ATF2/c-Jun enhance NFAT-mediated TNF-␣ promoter activity in ␤ cells, whereas MafA represses it.
To identify the area of the TNF-␣ gene that is regulated by MafA, we analyzed overlapping regions of the 5Ј-flanking TNF-␣ promoter up to Ϫ445 bp relative to the transcriptional start site by quantitative ChIP assays. The assays revealed a relatively high association of Maf at the Ϫ124 to Ϫ55 bp region (or between Ϫ97 and Ϫ70 bp) of the TNF-␣ promoter in response to IL-1␤ (Fig. 6B). This region corresponds to the previously described B-NFAT-C/EBP-␤ DNA-binding sites (Fig. 6B).
To assess if MafA and other transcription factors could regulate this region, we co-expressed them with a reporter containing a three times tandem repeat of the Ϫ95 to Ϫ70 region of the TNF-␣ promoter. Similar to the previous result, NFAT up-regulated this reporter activity almost 15-fold (Fig.  6C). However, no enhancement of NFAT-mediated transcription by ATF2 and c-Jun was observed. In contrast, MafA expression stimulated reporter activity almost 20-fold, and coexpression with NFAT almost doubled the effect. Thus, MafA and NFAT can up-regulate transcriptional activity via this isolated DNA sequence. This was not surprising considering that MafA is generally known as an activator and does not repress transcriptional activity in isolation. 3 To confirm that MafA regulates TNF-␣ gene transcription in ␤ cells, we silenced its expression in INS-1 cells by RNAi and assessed its requirement to repress TNF-␣ gene expression. Transfection of INS-1 cells with MafA shRNA increased IL-1␤-induced TNF-␣ mRNA expression and prevented its repression at 4 h (Fig. 7A). Overall, the results indicate that NFAT and MafA can modulate transcriptional activity from the Ϫ95 to Ϫ70 DNA segment of the TNF-␣ gene and repress its activity in the context of the TNF-␣ promoter in ␤ cells.
C/EBP-␤ Is Required for MafA to Repress TNF-␣ Gene Expression-The kinetics of association of C/EBP-␤ and Maf with the TNF-␣ promoter corresponds to a rapid decline in TNF-␣ mRNA. We previously showed that IL-1␤ induces C/EBP-␤ expression in ␤ cells within 2 h and that C/EBP-␤ can disrupt NFAT-mediated insulin gene transcription (36,56).
To determine whether C/EBP-␤ regulates TNF-␣ expression in ␤ cells, we blocked expression of C/EBP-␤ in INS-1 cells by RNAi and assessed the effect on IL-1␤-induced TNF-␣ mRNA. Preventing expression of C/EBP-␤ in INS-1 cells by adenovirus-mediated C/EBP-␤ shRNA enhanced IL-1␤-induced TNF-␣ mRNA expression at 2 h and attenuated or delayed its rapid reversal observed in controls at 4 h (Fig.  7B). The results suggest that induction of TNF-␣ expression by IL-1␤ in ␤ cells is in part repressed by C/EBP-␤.
MafA is rapidly expressed in ␤-cells in the presence of glucose. However, its association with the TNF-␣ promoter is not observed until C/EBP-␤ is present. Therefore, we sought to determine whether MafA requires C/EBP-␤ to regulate the TNF-␣ promoter. ChIP assays showed that although C/EBP-␤ shRNA prevented the appearance of C/EBP-␤ on the TNF-␣ 3 R. Stein, personal communication. promoter, Maf was still present. Likewise, knocking down MafA by RNAi had no effect on the association of C/EBP-␤ with the TNF-␣ promoter (Fig. 7, C and D). However, reporter assays revealed that knocking down C/EBP-␤ ablated the ability of MafA to repress NFAT-mediated TNF-␣ promoter activity (Fig. 7E). These data indicate that C/EBP-␤ is not required for MafA to associate with the TNF-␣ promoter, but it is required for MafA to repress TNF-␣ gene expression.

DISCUSSION
Proinflammatory cytokines can have detrimental effects on pancreatic islets leading to autoimmune destruction of ␤ cells in type 1 diabetes, ␤-cell dysfunction in type 2 diabetes, and tissue rejection in islet transplantation (57)(58)(59)(60). Exposure of islets to cytokines may arise from multiple sources, including resident or invading immune cells or systemically by nonimmune cells such as adipose tissue (6,(61)(62)(63). It is also known that cytokines are produced directly from islet tissue, and most studies support that these cytokines are produced primarily by infiltrating immune cells (6,42,64,65). However, previous studies have indicated that combinations of cytokines IL-1␤, IFN-␥, and TNF-␣ can induce chemokine and cytokine mRNA in rat and human ␤ cells (14 -16, 66). Additional experiments showed that islets secrete MCP-1, IL-15, IP-10, and MIP-3␣ in the medium in response to cytokines (17,66). However, it is still unclear whether these proteins were produced by the islet endocrine cells or resident immune cells.
In this study, we show that IL-1␤ induces TNF-␣ mRNA and protein expression in islet endocrine cells, and we determine signaling mechanisms by which the TNF-␣ gene is regulated in ␤ cells. We show that MAPKs activate downstream bZIP factors that cooperate with NFAT to induce or repress TNF-␣ gene expression. NFAT associated with the TNF-␣ promoter in islet, MIN6, and ␣TC1, and purified ␤ cells but not in nonendocrine pancreatic cells. This indicates that NFAT regulates TNF-␣ expression in the endocrine cell component of the islets in response to glucose and IL-1␤. ATF2 and c-Jun cooperatively activate the TNF-␣ promoter with NFAT, resulting in an acute accumulation of TNF-␣ mRNA.
The activation required both p38 and JNK and was rapidly reversed by the induction of C/EBP-␤ and activation of MafA, suggesting that ␤ cells have an aggressive counter-regulatory component to limit TNF-␣ production. C/EBP-␤ was required for MafA to repress TNF-␣ expression, and ERK1/2 were required for C/EBP-␤ to associate with the TNF-␣ promoter. Hence, the combination of p38 and JNK MAPKs promotes the induction of TNF-␣, whereas ERK1/2 appears to oppose it. Interestingly, MafA associated with and repressed the Ϫ125 to Ϫ56 region of the TNF-␣ promoter independent of C/EBP-␤ binding. No Maf consensus DNA-binding site was identified, and therefore further analysis will be needed to determine protein interactions and DNA base pair sequences required for MafA to regulate this region.
For example, it has been observed that administration of anti-TNF-␣ antibody can restore glucose homeostasis in both type 1 and 2 diabetes (74 -77). However, exogenous TNF-␣ can accelerate or prevent type 1 diabetes in nonobese diabetic mice, dependent upon the timing and duration of exposure (49,(77)(78)(79). Moreover, ␤-cell-specific expression of TNF-␣ in transgenic mice can induce or prevent type 1 diabetes depending on when TNF-␣ is expressed or repressed (50, 69).  These studies showed that extended localized expression of TNF-␣ in ␤ cells in the early disease stage leads to a breakdown of peripheral T cell tolerance to islets. In contrast, induction of TNF-␣ expression in the late stage protected islets from T cell infiltration. Furthermore, low concentrations of TNF-␣ and IL-1␤ can stimulate insulin secretion in ␤ cells, although higher concentrations and high glucose contribute to TNF-␣ and IL-1␤ cytotoxicity (80,81).
Based on these previous observations and the results of this study, we propose that high glucose conditions induce a low level expression of TNF-␣, which may have beneficial or compensatory effects on ␤ cells. However, exposure of islet endocrine cells to IL-1␤ results in a higher expression of TNF-␣ mediated by CN/NFAT and MAPKs. Extended localized expression of TNF-␣ in ␤ cells results in the recruitment of immune cells and islet inflammation These events contribute to ␤-cell destruction and ultimately lead to diabetes.
CN inhibitors are sufficient to prevent expression of TNF-␣ in islets as well as other cytokines in immune cells that are required for an immune response. However, CN is also re- quired for insulin gene expression and ␤-cell maintenance and survival (36,53,54,82). Not surprisingly, the use of FK506 to suppress the immune system and prevent organ or tissue graft rejection is associated with a high incidence of post-transplant diabetes (83)(84)(85)(86)(87)(88). Therefore, other methods need to be employed to selectively repress cytokine production without affecting islet cell function. We showed that inhibiting both p38 and JNK suppressed TNF-␣ gene expression in islets in vivo. Most importantly, insulin gene expression was unaffected. These results indicate that it is feasible to repress cytokine production in ␤ cells, while sparing ␤-cell function. Further studies will be required to determine whether selectively targeting expression of TNF-␣ in islet endocrine cells can prevent or alleviate immune destruction of ␤ cells and diabetes.