A comprehensive analysis of cytokine-induced and nuclear factor-kappa B-dependent genes in primary rat pancreatic beta-cells.

Type 1 diabetes mellitus results from an autoimmune destruction of pancreatic beta-cells. Cytokines, such as interleukin-1 beta and interferon-gamma, are putative mediators of immune-induced beta-cell death and, under in vitro conditions, cause beta-cell apoptosis. We have recently shown that interleukin-1 beta + interferon-gamma modifies the expression of >200 genes in beta-cells. Several of these genes are putative targets for the transcription factor nuclear factor-kappa B (NF-kappa B), and in subsequent experiments we showed that NF-kappa B activation is mostly pro-apoptotic in beta-cells. To identify cytokine-induced and NF-kappa B-regulated genes in primary rat beta-cells, we presently combined two experimental approaches: 1) blocking of NF-kappa B activation in cytokine-exposed beta-cells by a recombinant adenovirus (AdI kappa B((SA)2)) containing an inhibitor of NF-kappa B alpha (I kappa Bac) super-repressor (S32A/S36A) and 2) study of gene expression by microarray analysis. We identified 66 cytokine-modified and NF-kappa B-regulated genes in beta-cells. Cytokine-induced NF-kappa B activation decreased Pdx-1 and increased c-Myc expression. This, together with NF-kappa B-dependent inhibition of Glut-2, pro-hormone convertase-1, and Isl-1 expression, probably contributes to the loss of differentiated beta-cell functions. NF-kappa B also regulates several genes encoding for chemokines and cytokines in beta-cells. The present data suggest that NF-kappa B is a key "switch regulator" of transcription factors and gene networks controlling cytokine-induced beta-cell dysfunction and death.

Type 1 diabetes mellitus results from a progressive autoimmune destruction of the insulin producing pancreatic ␤-cells (1). Pro-inflammatory cytokines, such as IL-1␤, 1 IFN-␥, and TNF-␣, are released in the islets by activated macrophages and T-cells in the early stages of the disease (insulitis) and probably contribute for ␤-cell death (2)(3)(4). Blocking cytokine signaling with the use of cytokine antagonists, soluble receptors, or neutralizing antibodies prevents type 1 diabetes mellitus in BB rats and NOD mice (4). Under in vitro conditions, IL-1␤ induces functional impairment in mouse and rat ␤-cells (5) and, in combination with IFN-␥ and/or TNF-␣, leads to human, mouse, and rat ␤-cell death, mostly by apoptosis (4,5). This form of cell death seems to be the main mode of ␤-cell destruction in type 1 diabetes mellitus, as shown by studies in several animal models of immune-mediated diabetes (4).
␤-Cell apoptosis requires prolonged exposure to cytokines (6 -9 days) and is preceded by complex modifications in gene expression (5). To identify these genes, we have recently utilized microarray analysis of fluorescence-activated cell sortingpurified rat ␤-cells exposed for 6 or 24 h to IL-1␤ or IL-1␤ ϩ IFN-␥ (6). ␤-Cell exposure to cytokines leads to induction or down-regulation of ϳ200 genes and ESTs (6). There was a decrease in the expression of several genes related to differentiated ␤-cell functions, probably caused by down-regulation of genes encoding the transcription factors Pdx-1 and Isl-1, whereas cytokines up-regulated stress response genes. Interestingly, cytokines induced the expression of mRNAs for several chemokines, cytokines, and adhesion molecules. These peptides have the potential to increase mononuclear cell homing and activity during insulitis.
Several of the modified genes found by microarray analysis or candidate-gene approach are putative targets for the transcription factor NF-B (5,6). For three of these genes, namely iNOS (7), MnSOD (8), and Fas (9), the role of NF-B has been confirmed by functional studies with promoter-luciferase constructs and gel shift analysis. NF-B is a widely used transcription factor that plays a pivotal role in many cellular responses to environmental changes. NF-B is formed by homodimers or heterodimers of members of the Rel/NF-B family of proteins (10). They can bind to a set of related DNA target sites (B sites) and directly regulate gene expression. In nonstimulated cells, NF-B is associated with the inhibitory protein IB␣ and remains sequestered in the cytosol (10). NF-B-regulated genes have been shown to inhibit the apoptotic program in diverse cell types. In a few cell types, however, the transcription factor also acts as apoptosis inducer (11). Note that most studies suggesting an anti-apoptotic role for NF-B were performed in tumoral cell lines (11), and little information is available on the effects of NF-B activation on primary, nondividing cells.
There is no detectable NF-B activity in resting ␤-cells, but, upon exposure to IL-1␤, NF-B is activated and translocates to the nucleus (12). Inhibition of cytokine-induced NF-B activation by either a recombinant adenovirus (AdIB (SA)2 ), containing a nondegradable mutant form of IB␣ (S32A, S36A), or by stable transfection with a dominant negative inhibitor of NF-B, prevents cytokine-induced cell death in respectively human islet cells (13) and purified rat ␤-cells (14), and in mouse insulin-producing MIN6 cells (15). In addition, intravenous administration of a NF-B "decoy" inhibits alloxan-induced ␤-cell death and diabetes mellitus in mice (16). Thus, it seems that NF-B activation has mostly pro-apoptotic effects in pancreatic ␤-cells. However, the nature of the cytokine-induced and NF-B-regulated genes in pancreatic ␤-cells remains to be identified. To address this issue, we presently combined two experimental approaches, namely the blocking of NF-B activation by AdIB (SA)2 in cytokine-exposed ␤-cells and a comprehensive study of gene expression by microarray analysis. We identified 66 cytokine-induced genes, which are apparently regulated by NF-B. This collection of genes is an exciting resource for understanding the function of NF-B in primary ␤-cells and provides new insights into the signal transduction of cytokineinduced ␤-cell dysfunction and death.

EXPERIMENTAL PROCEDURES
Islet Cell Isolation and Culture-Pancreatic islets were isolated from 10-week-old male Wistar rats by collagenase digestion, and islet ␤-cells were purified by autofluorescence-activated cell sorting (17) (FACStar, Becton-Dickinson, Sunnyvale, CA). ␤-Cell aggregates were cultured in suspension at 37°C in Ham's F-10 medium (Invitrogen, Paisley, Scotland), as described previously (18). For the microarray analysis, purified rat ␤-cells were precultured in Ham's medium for 16 h and then infected either with a control virus (AdLuc) or with an IB␣ superrepressor virus (AdIB (SA)2 ) (described below), or, alternatively, left uninfected. 24 h after infection, the ␤-cells were exposed to IL-1␤ ϩ IFN-␥ for 24 h. IL-1␤ (tested at 50 units/ml; 38 units/ng) was a kind gift from Dr. C. W. Reinolds (NCI, National Institutes of Health, Bethesda, MD), and IFN-␥ (tested at 1000 units/ml; 10 units/ng) was purchased from Invitrogen. The choice of cytokine concentration and time of exposure is based on our previous data (6). For the confirmation experiments with RT-PCR (see below), one additional control virus, AdGFP, was added. To evaluate the role of NO for some of the cytokine-induced and NF-B-dependent genes, the iNOS inhibitor N G -methyl-L-arginine (L-MA, 1.0 mM; Sigma, Bornem, Belgium) was used together with IL-1␤ ϩ IFN-␥. We have shown previously that this concentration of L-MA prevents cytokine-induced NO formation by ␤-cells (19). Culture media from the cells used for the microarray analysis and iNOS blocker experiments were collected after 24 h of IL-1␤ ϩ IFN-␥ exposure for nitrite determination (nitrite is a stable product of NO oxidation), as described previously (20).
Transfection with Recombinant Adenoviruses-The recombinant replicative-deficient adenovirus containing a mutated nondegradable IB␣, with serines 32 and 36 mutated to alanines (AdIB (SA)2 ), and the control virus containing the luciferase gene (AdLuc) were prepared as described previously (21), and were a kind gift from Dr. C. Jobin (University of California, San Diego, CA). The additional control virus containing the green fluorescent protein (AdGFP) was prepared as described (22). ␤-Cells were infected for 2 h at 37°C, at m.o.i. 7.5. We have shown previously that AdIB (SA)2 at this m.o.i. protects ␤-cells against IL-1␤/IFN-␥-induced NO production and apoptosis, but does not affect insulin content or glucose-induced insulin release (14).
Microarray Analysis-For microarray analysis the cells were harvested, total RNA isolated, and biotinylated cRNA samples prepared and hybridized in duplicates to the U34-A oligonucleotide array (Affymetrix, Santa Clara, CA) as described previously (6). Because of difficulties in obtaining a sufficient number of rat ␤-cells in a single occasion, and to decrease eventual biases because of biological variation, the cells were pooled from four separate experiments, using in each experiment 3.5 ϫ 10 5 cells/group. We have observed previously that GeneChip analysis, performed in duplicate on pooled ␤-cell samples, provides a reliable estimation of massive changes in mRNA expression (6).
Analysis of differential expression was performed by the software GeneChip Suite (version 4.0.1). Normalization was performed by global scaling, with the arrays scaled to an average intensity of 150. Duplicate hybridizations, using separate sets of chips, were performed for all conditions. A gene was considered to have modified expression if it averaged Ն2.5-fold change in the duplicate arrays. Cytokine-modified genes were considered as NF-B-dependent when expression was induced or inhibited (Ն2.5-fold) in noninfected and AdLuc-infected ␤-cells, but was prevented by AdIB (SA)2 . Genes whose induction or decrease by cytokines was diminished by at least 50% following AdIB (SA)2 treatment were also considered NF-B-dependent. Genes that had similar patterns of expression in the three different comparisons were considered as NF-B independent. The selected genes were allocated to different functional clusters (Tables I and II) based on the putative biological function of the encoded protein following a previously described classification (6).
mRNA Isolation and RT-PCR-RT-PCR, using specific primers, was performed to confirm some of the changes in mRNA expression observed in the microarray analysis. The selection of RT-PCR, instead of Northern blot analysis, was motivated by the limited availability of primary ␤-cells. mRNA isolation and RT-PCR were performed as described previously (23,24). The number of cycles was selected to allow linear amplification of the cDNAs under study. For semiquantitative PCR, the GAPDH housekeeping gene was used as control. We have shown previously (25) and confirmed in the current experiments that IL-1␤ and IFN-␥ do not affect GAPDH mRNA expression in insulinproducing cells. The primer sequences and their respective PCR fragment lengths were as follows: Glut-2, forward (5Ј-GGTGTGATCAATG-CACCTC-3Ј) and reverse (5Ј-GTATCTGGGGCTTTCTGGAC-3Ј) (646 bp); Pdx-1, forward (5Ј-GGTGCCAGAGTTCAGTGCTA-3Ј) and reverse (5Ј-TTATTCTCCTCCGGTTCTGC-3Ј) (369 bp); c-Myc, forward (5Ј-AA-GGCAGCTCTGGAGTGAGA-3Ј) and reverse (5Ј-TTCTCTTCCTCGTC-GCAGAT-3Ј) (410 bp); calbindin, forward (5Ј-ATGCCAGCAACTGAAG-TCCT-3Ј) and reverse (5Ј-CCGACAAGGCCATTATGTTC-3Ј) (449 bp); cholecystokinin-A receptor, forward (5Ј-TCATGACTCCGTACCCCATT-3Ј) and reverse (5Ј-ATCAGGTTGGCAGCTGAACT-3Ј) (406 bp). The primers for GAPDH, iNOS, Isl-1, IP-10 (rat mob-1), Gadd153/CHOP, and IL-15 were as described in Ref. 6; MCP-1 as described in Ref. 23, Ca 2ϩ ATPase type 2 (also called SERCA2) as described in Ref. 26, and MnSOD as described in Ref. 8. The identity of the PCR fragments of each gene were confirmed by size and DNA sequencing (data not shown). The ethidium bromide-stained agarose gels were photographed under UV transillumination using a Kodak Digital Science DC 120 camera (Eastman Kodak Co.). The data are presented as a representative figure for three to four similar experiments. Abundance of the GAPDH products were assessed by Biomax one-dimensional image analysis software (Kodak) and expressed in pixel intensities (optical density).
Statistical Analysis-Data are presented as means Ϯ S.E., and comparisons between groups were performed by analysis of variance followed by paired t test with the Bonferroni correction.

Identification of Cytokine-induced and NF-B-dependent Genes in Rat Pancreatic ␤-Cells by Microarray Analysis-For
the microarray experiments, noninfected ␤-cells, ␤-cells infected with AdIB (SA)2 , and ␤-cells infected with AdLuc (control virus) were either exposed for 24 h to the combination of the cytokines IL-1␤ ϩ IFN-␥ or left untreated. The selected time of cytokine exposure (24 h) is based on our previous data, showing that 24-h exposure to IL-1␤ ϩ IFN-␥ induces major changes in ␤-cell gene expression, but precedes a significant decrease in ␤-cell viability (6). As described previously (14), AdIB (SA)2 , but not AdLuc, prevented IL-1␤/IFN-␥-induced NO production (data not shown). Cells from four separate experiments were pooled for total RNA extraction, and the resulting biotinylated cRNAs were hybridized in duplicate to the Affymetrix rat U-34A oligonucleotide array containing ϳ8,000 probes (77% known genes and 23% ESTs). As observed previously (6), ϳ3,000 genes and ESTs were scored as present in each of the six conditions (2853-3626 genes). The following comparisons were analyzed: 1) noninfected cells ϩ cytokines versus noninfected cells, 2) AdIB (SA)2 ϩ cytokines versus AdIB (SA)2 , and 3) AdLuc ϩ cytokines versus AdLuc. Similar to our previous findings (6), 24-h exposure to IL-1␤ ϩ IFN-␥ modified expression of 224 genes and ESTs in noninfected ␤-cells and 211 genes and TABLE I Cytokine-induced and NF-B-dependent genes in rat pancreatic ␤-cells Cytokine-induced differences in gene expression were considered as present when the mean -fold change was Ն2.5. Genes were considered as cytokine-modified and NF-B-dependent when their expression was modified in both noninfected ␤-cells and ␤-cells infected with AdLuc (control virus), but this cytokine-induced change was either not present or prevented by at least 50% in ␤-cells infected with AdIB (SA)2 . The genes are ordered according to the -fold variation in gene expression observed in noninfected cells. Asterisk (*) indicates that similar results were obtained by more than one group of probes. Increased, Ϫ, decreased compared with respective controls (i.e. ␤-cells noninfected, infected with AdIB (SA)2 , or infected with AdLuc and not exposed to cytokines). Data are the mean -fold variation of two hybridizations for the gene with the indicated access number.  A stringent criterion was used to accept genes as NF-B-dependent. Thus, we considered as cytokine-induced and NF-Bdependent those genes that were modified Ն2.5-fold by IL-1␤ ϩ IFN-␥ in non-virus-infected ␤-cells and in the AdLuc (both considered as controls), but had this modification prevented by AdIB (SA)2 . In cases where the induction/inhibition was lowered by at least 50% by AdIB (SA)2 , the genes were also considered as NF-B-dependent. The NF-B-dependent genes were classified according to the putative function of their encoded proteins and are listed in Table I. Note that, in the present series of experiments, ESTs, including those with significant homology to known genes (as evaluated by BLAST analysis) are included in Tables I and II. ESTs were not included in our previous microarray analysis of cytokine-induced genes (6).
It is noteworthy that several cytokine-induced modifications in the expression of transcription factors were NF-B-depend-   (Table I). Thus, the induction of both C/EBPs ␤ and ␦, Gadd153/CHOP, and the D-binding protein by cytokines was prevented by AdIB (SA)2 . NF-B blockage by AdIB (SA)2 inhibited the induction of three "defense/repair" genes in ␤-cells, namely MnSOD and the heat shock proteins 27 and 70. On the other hand, AdIB (SA)2 prevented cytokine-induced decrease of the Gas-6 growth arrest-specific gene (Table I).
A list of cytokine-induced and NF-B independent genes of special interest is shown in Table II. As suggested for other tissues (31,32), most of the genes listed in Table II are probably induced by IFN-␥, including STAT-1, IRF-1, IRF-7, several MHC-related genes, and proteasome subunits.

Confirmation by RT-PCR of Genes Identified as Cytokineinduced and NF-B-dependent-Eight of the genes detected by microarray analysis as cytokine-induced and NF-B-dependent
were selected for confirmation by RT-PCR. We also evaluated the role of NF-B for the expression of the transcription Pdx-1. Pdx-1 was not present in the microarray, but it is an essential regulator of pancreatic endocrine cell development and adult islet ␤-cell function (33).
RT-PCR analysis confirmed the microarray results for all but one (calbindin) selected gene (Fig. 1). Thus, AdIB (SA)2 blocked cytokine-induced Gadd153/CHOP, IP-10, c-Myc, MnSOD, and iNOS mRNA expression, whereas it prevented the inhibitory effect of IL-1␤ ϩ IFN-␥ on Glut-2 and Isl-1. In the case of calbindin, the cytokine-induced down-regulation observed by microarray analysis in noninfected cells was not reproduced by RT-PCR (data not shown). IL-1␤ ϩ IFN-␥ down-regulated Pdx-1 expression in noninfected ␤-cells or in ␤-cells infected with the control viruses. Expression of the IB␣ (SA)2 superrepressor prevented this decrease, suggesting that Pdx-1 inhibition depends on NF-B activation (Fig. 1).

Identification of Nitric Oxide-regulated Genes among the Cytokine-induced and NF-B-dependent Genes-IL-1␤
in combination with IFN-␥ induces iNOS expression and the synthesis of the radical NO in ␤-cells. The production of NO depends on the transcription and translation of the iNOS gene and is already detected 6 h after exposure to the cytokines (12). NO has been shown previously to modify gene and protein expression in rat ␤-cells (34 -37). Blocking NF-B expression by AdIB (SA)2 prevents IL-1␤/IFN-␥-induced NO production (Ref. 14 and present data). Thus, modifications in some of the genes listed in Table I could be secondary to inhibition of NO synthesis. Indeed, cytokine induced changes in two of the genes identified in the present array as NF-B-dependent, namely PLD-1 and Hsp 70 (Table I) has been previously shown to depend on NO formation (36,38). To further investigate this issue, rat ␤-cells were exposed for 24 h to the following conditions:  II Cytokine-induced and NF-B-independent genes in ␤-cells Cytokine-induced differences in gene expression were considered as present following the same criteria outlined in Table I. Genes were considered as cytokine-modified and NF-B-independent when their expression was similar in noninfected ␤-cells and ␤-cells infected with AdLuc (control virus) or AdIB (SA)2 . The genes are ordered according to the -fold variation in gene expression observed in noninfected cells. *, Similar results obtained by more than one group of probes. Increased, Ϫ, decreased compared with respective controls (i.e. ␤-cells noninfected, infected with AdIB (SA)2 , or infected with AdLuc and not exposed to cytokines). Data are the mean -fold variation of two hybridizations for the gene with the indicated access number. 10 3 cells ϫ 24 h, p Ͻ 0.01 versus IL-1␤ ϩ IFN-␥).
The cytokine-induced decrease in CCK-A receptor, SERCA2, Glut-2, Pdx-1, and Isl-1 mRNAs and increase in Gadd153/ CHOP mRNA was totally or partially prevented by addition of L-MA (Fig. 2). This suggests that NF-B blocking prevents modifications in the expression of these genes in ␤-cells via inhibition of NO production. On the other hand, the induction of IP-10, IL-15, MCP-1, c-Myc, MnSOD, and iNOS mRNAs by cytokines was not hampered by iNOS inhibition, suggesting a direct effect of NF-B on these genes.
Microarray comparisons between control (i.e. cells not exposed to cytokines) noninfected ␤-cells and ␤-cells infected with AdLuc or AdIB (SA)2 did not show consistent changes in gene expression. Moreover, infection of ␤-cells with AdLuc or AdIB (SA)2 in the absence of cytokines did not modify expression of Pdx-1, Glut-2, Isl-1, and GAPDH mRNAs ( Fig. 1; present data), and did not affect insulin content or glucose-induced insulin release (14). This suggests that the adenoviral vectors used in the present study have minor effects per se on ␤-cells, and that the effects induced by AdIB (SA)2 are mostly related to NF-B blocking.
Similar to our previous microarray analysis (6), we observed that IL-1␤ ϩ IFN-␥ modifies the expression of ϳ200 genes in fluorescence-activated cell sorting-purified rat ␤-cells. We confirmed by RT-PCR the microarray results of seven of eight cytokine-induced dependent genes. Taking these results together with previous observations by our group (6), we have now confirmed by RT-PCR 25 of 27 genes detected in microarray analysis as modified by IL-1␤ ϩ IFN-␥ (Ͼ90% confirmation rate). This suggests that our approach to microarray analysis, i.e. pooling four to six independent experiments and performing the chip analysis in duplicate, provides a reliable method for comprehensive analysis of gene expression in primary ␤-cells.
To identify NF-B-dependent genes, parallel comparisons between the microarray results for cytokine-induced genes in noninfected cells, cells infected with AdLuc and AdIB (SA)2 were performed. Using this approach, 66 genes that were both IL-1␤/IFN-␥-induced and NF-B-dependent were identified. The criteria used to accept genes as NF-B-dependent were rather stringent (see "Experimental Procedures"), and we probably underestimated the real number of cytokine-induced genes that are, at least in part, dependent on NF-B activation. Among the identified NF-B-dependent genes, iNOS and Mn-SOD have already been demonstrated as NF-B-regulated in pancreatic ␤-cells (7,9). Eleven additional genes were identified previously as NF-B targets in other tissues (see "Results"), but most of the 53 remaining genes have not been previously described as NF-B-dependent.
IL-1␤ induces a rapid (30-min) NF-B translocation to the nucleus in rat ␤-cells (9,12), which is independent of IFN-␥ (39). Because the main goal of the present study was to identify NF-B-dependent gene patterns with a potential role in cytokine-induced ␤-cell dysfunction and death, we selected a relatively long time exposure to IL-1␤ ϩ IFN-␥, i.e. 24 h. At this time point, a fraction of ␤-cells is already committed to undergo apoptosis, but there is not yet a detectable increase in the number of dead cells (6). The selected time point does not allow discrimination between early "primary" effects of NF-B and late "secondary" effects of the transcription factor, mediated via induction of other genes and proteins. For instance, AdIB (SA)2 prevention of some of the cytokine-induced genes is probably secondary to inhibition of iNOS expression and NO production. Thus, AdIB (SA)2 decreases iNOS expression (Ref. 14; present data) and the effects of NF-B blocking in some target genes were reproduced by an iNOS inhibitor (Fig. 2). Some of the deleterious effects of cytokines in rat ␤-cells are attributed to NO production, including inhibition of insulin secretion, mitochondrial dysfunction, and DNA damage (34), but cytokineinduced apoptosis in human, rat, and mouse ␤-cells is mostly NO-independent (38,40,41).
As described previously (6), cytokines induced c-Myc mRNA expression in ␤-cells. This effect was prevented by AdIB (SA)2 but not by an iNOS blocker, suggesting a direct effect of NF-B on c-Myc expression. Increased c-Myc mRNA content was also observed in ␤-cells exposed in vivo and in vitro to supraphysiological glucose levels (47), and adenovirus-mediated c-Myc overexpression suppresses both insulin gene transcription and glucose-stimulated insulin secretion (48). Thus, it is conceivable that increased expression of the oncogene c-Myc participates in both high glucose and cytokine-induced ␤-cell dysfunction. Besides contributing to ␤-cell "de-differentiation," transgenic c-Myc expression in ␤-cells is associated with parallel cell growth and apoptosis, eventually culminating in islet involution and diabetes mellitus (49). The presently observed cytokine induction of ornithine decarboxylase (previously confirmed at the protein level; Ref. 50), also shown to be NF-Bdependent, is probably secondary to c-Myc up-regulation, because ornithine decarboxylase is a direct transcriptional target of c-Myc in ␤Ϫcells and other cell types (47). Besides c-Myc, Pdx-1, and Isl-1, cytokine-induced NF-B activation also modulates the expression of the transcription factors C/EBP␤, C/EBP␦, and Gadd153/CHOP. C/EBP factors cooperate with NF-B for cytokine-induced Fas expression in pancreatic ␤-cells (9), and this may be another mechanism by which cytokines, via NF-B activation, contribute to ␤-cell apoptosis in early type 1 diabetes mellitus.
Cytokine-induced decrease in the expression of mRNA for SERCA2, a gene that is responsible for Ca 2ϩ transfer into the endoplasmic reticulum (ER), is also NO-mediated and blocked by AdIB (SA)2 (present data). Depletion of ER Ca 2ϩ by thapsigargin or NO triggers the ER stress pathway in MIN6 cells, leading to Gadd153/CHOP expression and apoptosis (51,52). We presently observed cytokine-induced Gadd153/CHOP expression in primary ␤-cells, a phenomenon secondary to NF-B activation and NO production (6, present data). Of interest, TRAF-2, an adaptor protein with a crucial role for activation of caspase-12 during ER stress-induced apoptosis (53), is also induced by cytokines in ␤-cells via an NF-B-dependent pathway (present data).
AdIB (SA)2 reduced expression of the cell adhesion molecule ICAM-1, the chemoattractant proteins MCP-1 and IP-10 and the cytokine IL-15. These genes are known NF-B targets (28), and their induction was not hampered by an iNOS blocker (present data). The expression of MCP-1 (54), IP-10, and IL-15 have already been confirmed in human pancreatic islets, both at the mRNA and protein levels (data ot shown). We have suggested previously that cytokine-induced chemokine expression by ␤-cells contribute to the activation and recruitment of inflammatory cells to the area of insulitis (6,54). Because this process is NF-B-dependent (present data), it will be of interest to evaluate whether in vivo NF-B blocking decreases mononuclear cell infiltration in early insulitis or following islet allografting into diabetes-prone NOD mice.
Several relevant cytokine-induced genes are NF-B independent (Table II). Most of these genes are probably regulated by the IFN-␥ signaling pathway, which is independent of NF-B activation (31). This group includes the transcription factors STAT-1, IRF-1, and IRF-7. In line with these findings, IFN-␥ alone induces STAT-1 expression and phosphorylation in insulin-producing cells (55), and increases IRF-1 expression in human and rat islets (39). IRF-1 seems to be of minor importance for cytokine-induced ␤-cell death in vitro and in vivo (56,57). On the other hand, STAT-1 activation leads to antiproliferative and pro-apoptotic events in some cell types (58), and its induction has been associated with the deleterious effects of IFN-␥ in insulin-producing cells (55,59). Considering that both IL-1␤ and IFN-␥ are required for triggering apoptosis, it will be of interest to identify the nature of the STAT-1induced genes in ␤-cells. These genes, together with the presently described NF-B-dependent genes, may play a key role in the ␤-cell decision whether or not to undergo apoptosis. Fig. 3 provides an overview of the main findings of the present microarray analysis, and the putative consequences of the observed modifications in gene expression. We suggest that NF-B functions as a "master switch," controlling distinct networks of transcription factors and effector genes that are important for maintaining the ␤-cell differentiated state, cytosolic and ER calcium homeostasis, apoptosis, and attraction and activation of immune cells.
There has been increasing evidence that ␤-cell dysfunction in type 2 diabetes mellitus is caused, at least in part, by defective expression of key transcription factors (60). As suggested by the present findings, it is conceivable that the process of ␤-cell death in type 1 diabetes mellitus is also a "transcription factor malaise."