Global profiling of double stranded RNA- and IFN-γ-induced genes in rat pancreatic beta cells
- First Online:
- Cite this article as:
- Rasschaert, J., Liu, D., Kutlu, B. et al. Diabetologia (2003) 46: 1641. doi:10.1007/s00125-003-1245-y
- 326 Views
Viral infections and local production of IFN-γ might contribute to beta-cell dysfunction/death in Type 1 Diabetes. Double stranded RNA (dsRNA) accumulates in the cytosol of viral-infected cells, and exposure of purified rat beta cells to dsRNA (tested in the form of polyinosinic-polycytidylic acid, PIC) in combination with IFN-γ results in beta-cell dysfunction and apoptosis. To elucidate the molecular mechanisms involved in PIC + IFN-γ-effects, we determined the global profile of genes modified by these agents in primary rat beta cells.
FACS-purified rat beta cells were cultured for 6 or 24 h in control condition or with IFN-γ, PIC or a combination of both agents. The gene expression profile was analysed in duplicate by high-density oligonucleotide arrays representing 5000 full-length genes and 3000 EST’s. Changes of greater than or equal to 2.5-fold were considered as relevant.
Following a 6- or 24-h treatment with IFN-γ, PIC or IFN-γ and PIC, we observed changes in the expression of 51 to 189 genes. IFN-γ modified the expression of MHC-related genes, and also of genes involved in beta-cell metabolism, protein processing, cytokines and signal transduction. PIC affected preferentially the expression of genes related to cell adhesion, cytokines and dsRNA signal transduction, transcription factors and MHC. PIC and/or IFN-γ up-regulated the expression of several chemokines and cytokines that could contribute to mononuclear cell homing and activation during viral infection, while IFN-γ induced a positive feedback on its own signal transduction. PIC + IFN-γ inhibited insulin and GLUT-2 expression without modifying pdx-1 mRNA expression.
This study provides the first comprehensive characterization of the molecular responses of primary beta cells to dsRNA + IFN-γ, two agents that are probably present in the beta cell milieu during the course of virally-induced insulitis and Type 1 Diabetes. Based on these findings, we propose an integrated model for the molecular mechanisms involved in dsRNA + IFN-γ induced beta-cell dysfunction and death.
KeywordsDouble stranded RNAmicroarray analysisapoptosispancreatic beta cellsinterferon-γnitric oxideNF-κBdiabetes mellitus
RNA-specific adenosine deaminase
bcl-2 associated x protein
- BB rats
diabetes-resistant BioBreeding rats
bcl-2 homology domain-3
BH3 interacting domain death agonist
double stranded RNA
eukaryotic initiation factor 2α
expressed sequence tags
growth arrest and DNA damage
glyceraldehyde 3-phosphate dehydrogenase
gastric inhibitory peptide
3-hydroxy 3-methylglutaryl coenzyme A
heat shock protein
interleukin converting enzyme/caspase 1
inducible nitric oxide synthase
interferon regulatory factor
Janus tyrosine kinase 1 and 2
mitogen activated protein
macrophage chemoattractant protein-1
major histocompatibility complex
macrophage inflammatory protein
manganese superoxide dismutase
- NOD mouse
non obese diabetic mouse
pancreatic duodenal homeobox factor-1
dsRNA dependent protein kinase
regulated upon activation, normal T-cell expressed, and presumably secreted
sarco(endo)plasmic reticulum Ca+2 ATPase type 2
signal transducers and activators of transcription
type 1 diabetes mellitus
Type 1 diabetes mellitus is an auto-immune disease associated with a progressive and selective destruction of the insulin-producing pancreatic beta cells [1, 2]. Although a genetic susceptibility seems to be a prerequisite for the development of Type 1 Diabetes [3, 4], it is now clear that environmental factors such as viral infections are also important aetiological determinants [2, 5].
There is extensive epidemiological evidence for the involvement of viral infections in the pathogenesis of Type 1 Diabetes . Up to now, 13 different viruses, most of them belonging to the enterovirus family, have been found to be associated with the onset of Type 1 Diabetes in humans and in various animal models . Different mechanisms have been proposed for the role of viruses in the pathogenesis of Type 1 Diabetes. These include: (i) infection and rapid destruction of beta cells ; (ii) triggering of local inflammation, leading to destruction of the beta cells through the production of NO, cytokines and other immune mediators in a mechanism referred to as “innocent bystander killing” [9, 10]; (iii) molecular mimicry, based on a partial sequence homology between a protein of the infected cells (i.e. GADD65) and a viral antigen, leading to autoimmune destruction of the beta cell ; (iv) viral infection, coupled with one or more of the factors described above, acting in conjunction to induce beta cell death . For example, mouse infection with a high titre of the D variant of the encephalomyocarditis (EMC-D) virus leads to beta-cell destruction and diabetes mainly as a result of viral replication within beta cells, while mouse infection with a low titre of EMC-D virus leads to diabetes as a chronic process, caused by the destruction of beta cells by soluble mediators such as IL-1β, TNF-α/β and NO produced by macrophages or the beta cells themselves .
The molecular mechanisms involved in beta cells damage by viruses, alone or in combination with soluble mediators, remain to be elucidated. During viral infection, accumulation of the viral replicative intermediate double stranded RNA (dsRNA) in the cytosol of the infected cell stimulates antiviral activities, such as dsRNA-dependent protein kinase (PKR) activation, type I interferons and NO production and a general inhibition of protein translation [13, 14]. These antiviral responses can be mimicked by treatment of cells with the synthetic dsRNA polyinosinic-polycytidylic acid (PIC) [15, 16]. PIC triggers the development of hyperglycaemia in diabetes-resistant BioBreeding (BB) rats and accelerates the development of the disease in diabetes-prone BB rats [17, 18]. In vitro, PIC inhibits glucose-stimulated insulin biosynthesis in mouse islets  and when used in combination with IFN-γ affects rat islet cell function and viability by a mechanism involving, at least in part, increased inducible nitric oxide synthase (iNOS) expression and NO production [20, 21].
We have previously shown that exposure of fluorescence-activated cell sorting-(FACS)-purified rat beta cells to PIC alone does not induce cell death. However, when these cells are exposed to PIC + IL-1β or PIC + IFN-γ they die mostly by apoptosis. The mechanisms of death induction are either NO-dependent in the case of PIC + IL-1β, or NO-independent, in the case of PIC + IFN-γ . Moreover, PIC regulates the expression of several genes that could participate in the induction of islet inflammation and beta-cell death, such as Fas, iNOS, IL-15 and diverse chemokines [21, 22]. The complete range of genes induced by PIC and cytokines in pancreatic beta cells remains, however, to be clarified.
Evaluation of complex patterns of gene expression is now feasible by the use of microarray analysis . We have previously used high-density oligonucleotide arrays to analyse FACS-purified rat beta cells exposed for 6 or 24 h to IL-1β + IFN-γ [24, 25]. Based on the findings obtained, we proposed that beta-cell fate following cytokines exposure, namely death by apoptosis or survival with or without complete functional recovery, depends on the intricate pattern of dozens of genes up- or down-regulated in parallel and/or sequentially. Moreover, these and subsequent data on INS-1 cells  allowed us to start an annotated “Beta Cell Gene Bank”, including information on nearly 3000 genes expressed in beta cells and in INS-1 cells, of which 700 are modified by cytokines. Against this background, we carried out a microarray analysis of FACS-purified rat pancreatic beta cells exposed for 6 or 24 h to PIC, IFN-γ or a combination of both agents. The data obtained provide the first broad picture on how a primary beta cell responds to dsRNA and the inflammatory cytokine IFN-γ, an experimental condition which bears similarity to the in vivo situation during the course of a viral infection. The results obtained allowed us to propose a comprehensive hypothesis for the mechanisms involved in dsRNA and IFN-γ induced beta-cell dysfunction and death.
Materials and methods
Islet cell isolation and culture
Pancreatic islets were isolated from 10–12-week-old male Wistar rats by collagenase digestion, and subsequently dissociated into single cells in a calcium-free medium containing dispase (0.5 mg/ml). Single beta-cells were purified by autofluorescence-activated cell sorting (FACS) . These preparations contain around 90–95% viable beta cells [data not shown, 27]. The purified beta cells were cultured in HAM’s F-10 medium (Invitrogen, Paisley, Scotland) supplemented with 10 mmol/l glucose . For determination of viability, FACS-purified single beta cells (104 cells per well) were cultured for 6 days in Falcon 96-well microtitre plates (Becton Dickinson, New Jersey, N.J., USA) pre-coated with poly-L-lysine and containing 200 µl of medium. Culture medium was changed every 3 days and fresh IFN-γ or PIC was added. For RNA extraction for microarray or RT-PCR analysis, single beta cells were re-aggregated for 3 h in a rotatory shaking incubator , cultured for 14–16 h in suspension, and then exposed for 6 or 24 h to IFN-γ (1000 U/ml; 10 U/ng; Invitrogen) or PIC (100 µg/ml; Sigma Chemical, St Louis, Mo., USA). In some experiments, recombinant human IL-1β (50 U/ml, 38 U/ng, a kind gift of Dr. C.W. Reynolds from the National Cancer Institute, Bethesda, Md., USA) was also utilized as a positive control. The concentrations of cytokines and PIC, and the time points for array analysis, were selected based on our previous studies in beta cells [21, 22, 24, 25], and aimed to analyse beta cells at time points which precede non-specific changes in mRNA expression induced by early apoptosis. Culture media were collected after 24 h for nitrite determination (nitrite is a stable product of NO oxidation), which was done spectrophotometrically at 546 nm wavelength after coloured reaction with the Griess reagent .
Assessment of beta cell protein synthesis and viability
Total protein biosynthesis was determined at 10 mmol/l glucose using L-[4,5-3H] leucine incorporation and trichloroacetic precipitation . The experiments were carried out in duplicate, using 6×104 re-aggregated beta cells per condition. The cells were exposed for 1, 5 and 23 h to IFN-γ + PIC (same concentrations as described above) or left untreated (control) before determination of protein biosynthesis for 2 h in the absence (control) or presence of IFN-γ + PIC. These time points were selected to cover an early time point (3 h), and then two time points placed 1 h after the time points used for microarray analysis (6 and 24 h), the rationale being that protein synthesis usually lags behind mRNA expression. As a positive control for protein biosynthesis inhibition, some control cells were exposed to cycloheximide (10 µmol/l) during the final 2 h of the incubation period.
The percentage of viable, apoptotic and necrotic beta cells was determined after 6 days exposure to IFN-γ and/or PIC [21, 22]. For this purpose, beta cells were incubated for 15 min with propidium iodide (PI, 10 µg/ml) and Hoechst (HO) 342 (10 µg/ml) . This fluorescence assay for single beta cells is quantitative and has been validated by systematic comparisons with electron microscopy observations [32, 33]. The method has been successfully used to evaluate apoptosis/necrosis in rat [21, 22, 32], mouse [34, 35] and human  beta cells.
For microarray analysis, total RNA was isolated from beta cells (at least 10 µg/sample) and used to prepare biotinylated cRNA. The labelled cRNAs were hybridized in duplicate to the rat U34-A oligonucleotide array (Affymetrix, Santa Clara, Calif., USA) [24, 25]. Due to difficulties in obtaining a sufficient number of primary beta cells in a single isolation, and to decrease putative biases caused by biological variation, the cells were pooled from four independent experiments, using in each experiment 3.5–3.7×105 cells/group. We have shown before that microarray analysis, carried out in duplicate on pooled beta cell samples, provides a reliable estimation of massive changes in mRNA expression, as confirmed by a greater than 90% confirmation by RT-PCR of genes observed as modified in the array [24, 25, 37, present data]. Analysis of differential expression was carried out using the GeneChip Suite software, version 4.0.1 (Affymetrix, Santa Clara, Calif., USA). Arrays were normalized by global scaling, with the arrays scaled to an average intensity of 150. Genes were considered as modified by PIC and/or IFN-γ in case they fulfilled the following criteria [24, 25]: (i) the mRNA was present in either control or cells exposed to PIC and/or IFN-γ in both duplicates; (ii) the mean average fold change (experimental group vs control) was greater than or equal to 2.5; (iii) the fold change in each individual duplicate was greater than or equal to 2.0. We have used our own curated “Beta Cell Gene Bank” to assign the filtered genes into their respective functional clusters. The expressed sequence tags (ESTs) that had homology to a known sequence were annotated using the Resourcerer 6.0 database . Non-identified ESTs are not shown here.
RT-PCR and real time RT-PCR
RT-PCR was done using poly(A)+ RNA as described . The number of cycles was selected to allow linear amplification of the cDNA under study. The primer sequences used for amplification of rat cDNAs for GAPDH, GADD153 , pdx-1, GLUT-2 and c-Myc  and for sarco(endo)plasmic reticulum Ca2+ ATPase type 2 (also called SERCA-2)  were as described in the indicated references. For the other genes studied, the primer sequences and their respective PCR fragment lenghts were as follows: dsRNA-dependent protein kinase (PKR), forward (5′-AATCACGCCAACATTGTTCA-3′) and reverse (5′-CACCGGGTCTTGTATCGACT-3′) (107 bp); RNA-specific adenosine deaminase (ADAR), forward (5′-AAGAAACAGGGCAAGCAAGA-3′) and reverse (5′-TGTTGGTCAGAGCGTTGAAG-3′) (244 bp); CEBP/β, forward (5′-CAAGCTGAGCGACGAGTACA-3′) and reverse (5′-CAGCTGCTCCACCTTCTTCT-3′) (147 bp) and Bip/GRP78, forward (5′-CTCAAAGAGCGCATTGACAC-3′) and reverse (5′-GCCACTTGGGCTATAGCATT-3′) (446 bp). The identity of the PCR fragments of each gene was confirmed by size after electrophoretic migration on ethidium bromide-stained agarose gels photographed under UV-transillumination using Kodak Digital Science DC120 camera (Kodak, Rochester, N.Y., USA). The data are shown as a representative figure for four similar experiments. Expression of the “housekeeping” gene GAPDH is not affected by exposure to cytokines in both whole islets and FACS-purified beta cells [21, 22, 40].
Real-time RT-PCR was done as described  using a Lightcycler instrument in a 20 µl reaction containing 3 mmol/l MgCl2, 0.5 µmol/l forward and reverse primers, 2 µl FastStart SYBR Green mix (Roche), and 2 µl template cDNA. The primer sequences and their respective PCR fragment lenghts were as follows: GADD153, forward (5′-CCAGCAGAGGTCACAAGCA-3′) and reverse (5′-CGCACTGACCACTCTGTTTC-3′) (126 bp); SERCA-2, forward (5′-TTGTGGCCCGAAACTACCT-3′) and reverse (5′-TTCATAATGAGCAGCACAAAGGG-3′) (121 bp). The method used for quantification is the standard curve approach [42, 43]. To obtain the standard curve, the primer sequences and their respective PCR fragment lenghts were as follows: GADD153, forward (5′-GTCTCTGCCTTTCGCCTTTG-3′) and reverse (5′-CTACCCTCAGTCCCCTCCTC-3′) (605 bp); SERCA-2, forward (5′-TCTAGTCACCATAGAGATGTG-3′) and reverse (5′-TACTGACTGAGGTAGCAGGA-3′) (912 bp).
A plasmid construct containing the human pdx-1 gene promoter linked to a luciferase reporter gene was kindly provided by Dr D. Melloul, Department of Endocrinology, Hadassah University Hospital, Jerusalem, Israel . Transfected beta cells were exposed to cytokines and/or PIC for 24 h (same concentrations as above). Luciferase activity was assayed with the Dual-Luciferase Reporter Assay System (Promega) as previously described [21, 41]. Test values were corrected for the luciferase activity value of the internal control plasmid, pRL-CMV.
Results of microarray analysis are shown as means of two similar determinations. The results for other experiments are presented as means±SEM of at least three independent experiments. Statistical differences between the groups were determined by paired Student’s t-test or ANOVA, as indicated. A p value of less than 0.05 was considered statistically significant.
Viability and nitrite production of beta cells exposed to PIC and/or IFN-γ
To identify by microarray analysis early and late IFN-γ and/or PIC induced or decreased genes in pancreatic beta cells, the cells were exposed for 6 or 24 h to the following conditions: control condition; IFN-γ (1000 U/ml); PIC (100 µg/ml); or the combination of both agents. When the beta cells were treated for 6 h with PIC and/or IFN-γ, there was no detectable nitrite production. However, after a 24-h treatment the combination of PIC and IFN-γ significantly increased nitrite production as compared to control, non-treated cells. Thus, the values for PIC + IFN-γ were 5.6±1.1 pmol nitrite·10-3cells·h (means ± SEM; n=6; p<0.05 vs control) while control values were 0.7±0.4 pmol nitrite·10–3 cells·h (mean ± SEM; n=6). When tested alone, neither PIC nor IFN-γ affected nitrite production (data not shown). These data on viability and nitrite production are in good agreement with our previous observations [21, 22] and confirm that both IFN-γ and PIC were biologically active.
Identification of IFN-γ and/or PIC -modified genes in beta cells by microarray analysis
Cells from four separate experiments conducted as described above were pooled for RNA extraction, and the resulting biotinylated cRNAs were hybridized in duplicate to the Affymetrix rat U34-A oligonucleotide array containing about 8000 probes (77% known genes and 23% ESTs). Approximately 3759 (3300 to 4217) genes or ESTs were scored as present in each of the six conditions, in fair agreement with our previous observations [24, 25].
Total number of up- and down-regulated genes in rat pancreatic beta cells after 6 or 24 h
IFN-γ + PIC
Percentages of differentially expressed genes following a 6 or 24 h exposure to IFN-γ and/or PIC
IFN-γ + PIC
IFN-γ + PIC
Protein synthesis, modification and secretion
Ionic channels, ions transporters and related proteins
Hormones, growth factors and related genes
Cytokines, chemokines and related receptors
Cytokine processing and signal transduction
MHC and related genes
Cell adhesion and cytoskeleton
Transcription factors and related
RNA synthesis and splicing factors
Total number of genes (absolute values)
List of selected genes modified after a 6- or 24-hour exposure to PIC and/or IFN-γ
Creatine kinase—ubiquitous *
Glucose transporter type 1 (GLUT 1)
Glucose transporter type 2 (GLUT 2) *
Lactate dehydrogenase A
Phosphofructokinase c (PFK-c)
1.2 Arginine metabolism and NO formation
Cationic amino acid transporter-1
1.3 Aminoacids (other than arginine)
Glutamic acid decarboxylase 65 (GAD 65)
L-amino acid decarboxylase
3-hydroxy-3 methylglutaryl coenzyme A (HMGcoA) reductase
Long-chain acyl-CoA synthetase
Lysosomal acid lipase
Mitochondrial cytochrome P450 (P450C27)
Stearyl-CoA desaturase 2
Trycarboxilate carrier (mitochondrial)
1.5 ATP production and processing
Cytochrome b5 reductase (NADH)*
NADH dehydrogenase (ubiquinone) 1α subcomplex 4
Cytochrome P450 monooxygenase
2.0 Protein synthesis, modification, and secretion
Elongation factor 1 α
Mannosidase α type II
SNAP-25 interacting protein hrs-2
Synaptosomal associated protein (SNAP-25a)
3.0 Ionic channels, ions transporters and related proteins
Calcium-activated potassium channel rSK3 (SK)
Multidrug resistance protein (ID:1)
Voltage dependent anion channel (RVDAC 1)*
4.0 Hormones, growth factors and related genes
Cholecystokinin (CCK) precursor
Endothelial growth factor form 3
Gastric inhibitory peptide receptor
Growth hormone receptor (short isoform)
Growth hormone receptor
Mitochondrial vitamin D(3) 25-hydroxylase
Prohormone convertase 2
Somatostatin receptor type 2
Somatostatin receptor type 3
Thyrotropin releasing hormone*
Vascular endothelial growth factor (VEGF)*
Vascular endothelial growth factor B
5.0 Cytokines, chemokines and related receptors
Interleukin 15 *
Macrophage chemoattractant protein-1 (MCP-1) *
Macrophage inhibitory cytokine-1 (MIC-1)
6.0 Cytokine and dsRNA processing and signal transduction
3CH134/CL100 protein tyrosin phosphatase
ER transmembrane protein
Guanylate nucleotide binding protein 2
INF-γ induced GTPase
Janus protein tyrosine kinase 2 (JAK-2)*
MAP-kinase phosphatase (cpg21)
Pim-3 serine threonine kinase
Protein kinase C delta subspecies
Protein tyrosine phosphatase
Protein tyrosine phosphatase TD14 (PTP-TD14)
7.0 MHC and related genes
Mature MHC class Ib α chain
MHC-I non-RT1 α chain
MHC-Ib RT1.S3 *
MHC class II A-β RT1.B-b-β
MHC-II-assoc. invariant chain γ *
Proteasome subunit RC1
Proteasome subunit RING 12 *
Proteasome activator rPA28-b *
8.0 Cell adhesion, cytoskeleton and related genes
Antigen CD-47 *
36Kd β-galactoside binding lectin *
2,3-cyclic nucleotide 3-phosphodiesterase (CNPII)
Intercellular adhesion molecule-1 (ICAM-1)
Neurofilament protein middle (NF-M)
9.0 Transcription factors and related genes
CREM transcriptional repressor*
CREMδ C-G gene
Heat shock transcription factor 1
High mobility group protein I
Hypoxia-inducible factor 1 (HIF1)
Interferon regulatory factor-1 (IRF-1) *
Interferon regulatory factor-7 (IRF-7)*
Leucine zipper protein
Leucine-rich acidic nuclear protein
Small nuclear RING finger protein 4
Transcription factor UBF-2
Transcription factor USF-1
Zinc finger protein 9
10.0 RNA synthesis and splicing factors
Survival motor neuron*
11.0 Cell cycle
GADD34 (mouse MyD116; rat PEG-3) *
Gas-6 growth arrest specific
Hsp 70 gene 1/2 *
Hsp 70 gene 3
O-6-methylguanine-DNA methyltransferase (MGMT)
Heat shock protein 70-1
13.0 Apoptosis and ER stress response and related genes
Calbindin d28 K
GADD 153 (Growth arrest DNA damage 153)
150 kDa oxygen regulated protein (ORP150)
(pBUS30) with repetitive elements
Sec 61 homolog
TRAP-complex gamma subunit
14.0 Anti-viral response
Double-stranded RNA-dependent protein kinase (PKR)
Double-stranded RNA-specific adenosine deaminase (ADAR)
2’,5’-oligoadenylate synthetase (OAS)
P glycoprotein 1
Serine protease inhibitor 15 (spi15)
Among the metabolism-related genes, a decrease in the expression of mRNAs encoding for genes involved in glucose metabolism, such as the beta-cell specific glucose transporter GLUT-2 and lactate dehydrogenase (isoforms A and B) was observed (Table 3; item 1.1). Interestingly, expression of glucokinase was not affected by any of the tested conditions. While expression of glycolytic enzymes was down-regulated, several genes involved or related to lipid metabolism, such as ATP-citrate lyase, 3-hydroxy-3 methylglutaryl-coenzyme A (HMGCoA) reductase, LDL receptor, mitochondrial cytochrome P450 and stearyl-CoA desaturase were up-regulated following exposure for 6 or 24 h to both IFN-γ and PIC (Table 3, item 1.4). Of note, exposure to IFN-γ and PIC led to modifications in key genes related to arginine metabolism and nitric oxide formation (Table 3, item 1.2). The modified genes are arginase (decreased), argininosuccinate synthetase, iNOS and the cationic amino acid transporter-1 (all increased) and correlate well with the observed increased production of nitrite after 24-h treatment by both agents (see above).
After 24-h treatment with IFN-γ and PIC, there was down-regulation of mRNAs encoding mitochondrial subunits of respiratory chain genes involved in ATP production, such as NADH dehydrogenase and cytochrome b5 reductase, coupled with an up-regulation of uridine kinase (Table 3, item 1.5). IFN-γ + PIC also down-regulated expression of mRNAs encoding receptors for the incretins cholecystokinin-A and gastric inhibitory peptide and of growth hormone receptor (Table 3, item 4.0). This, associated with the observed decrease in insulin and pro-hormone convertase (PC) 1 mRNA levels, and of genes involved in ATP production, could contribute for the decreased insulin secretion observed after treatment of pancreatic islets or beta cells with IFN-γ + PIC [20, 45].
Several chemokines, cytokines, as well as genes involved in signal transduction, were modified by IFN-γ or PIC, alone and specially in combination (Table 3, items 5.0 and 6.0). When tested alone, IFN-γ affected expression of the chemokines macrophage inhibitory cytokine-1 (MIC-1) and interferon inducible protein (IP)-10 and of the cytokine IL-15 (Table 3, item 5.0). IL-15, IP-10, MIP-3α, macrophage chemoattractant protein (MCP)-1 and fractalkine mRNAs were induced by PIC and/or the combination of IFN-γ + PIC. Induction of RANTES and TNF-β mRNAs were also observed in pancreatic beta cells exposed to PIC and/or PIC + IFN-γ. It has been previously shown that RANTES and IP-10 are secondarily up-regulated by autocrine production of IFN-β in RAW 264.7 cells, a murine macrophage cell line . In the present microarray analysis however, neither PIC nor IFN-γ, nor PIC and IFN-γ together, modified the expression of IFN-α, IFN-β and IL-1β, at least at our selected time points (data not shown).
Interestingly, the concomitant presence of PIC and IFN-γ led to up-regulation of Notch1 receptor and one of his ligands, Delta-1. Presenilin-2, an enzyme required for intramembraneous proteolysis of Notch was also up-regulated (Table 3; item 6.0). PIC and IFN-γ, however, did not modify expression of Jagged, another Notch ligand (data not shown). There was also up-regulation of JAK-2, Pim-3 serine threonine kinase and CL100 protein tyrosine phosphatase which, associated with a decrease in MAP-kinase phosphatase (cpg21) by exposure to PIC alone or in combination with IFN-γ (Table 3; item 6.0), could affect beta-cell signal transduction by acting on key phosphorylation/dephosphorylation steps.
Genes related to antigen presentation were induced to a major extent after both 6 and 24 h (Table 3, item 7.0). Increased expression of MHC-Ib RT1, MHC-II and proteasome subunit RC1 and RING12 were observed with both IFN-γ and PIC alone, but the two agents often showed additive effects. MTP-1 and MTP-2, proteins involved in the “machinery” for MHC class I presentation, were also up-regulated by both agents.
IFN-γ and/or PIC led to up- and down-regulation of numerous transcription factors and associated proteins (Table 3, item 9.0). Of note, among the 41 transcription factors modified, 63% of them were affected only in the presence of both agents, the sole condition leading to beta cell apoptosis. There was up-regulation of c-jun, c-Myc, C/EBPβ, NF-κB, Lim-1 and STAT-1. Up-regulation of these transcription factors was also observed after beta cell exposure to IL-1β and IFN-γ, another treatment leading to apoptosis. The microarray results also showed up-regulation of other transcription factors not described before in beta cells exposed to cytokines, including heat shock transcription factor 1, hypoxia-inducible factor 1, STAT-3, Max and USF-1 (Table 3; item 9.0). Nuclear Ring Finger protein 4 (RNF4) and SWI/SNF, mRNAs encoding for multiprotein complexes remodelling chromatin, which are required for positive and negative control of various cellular pathways, were also up-regulated following beta-cell exposure to PIC and IFN-γ (Table 3; item 9.0).
IFN-γ alone induced an early (6 h) decrease in two mRNAs potentially involved in beta-cell defence/repair, namely glutathione reductase and hsp70 (Table 3; item 12.0), and of two endoplasmic reticulum (ER) chaperones that could contribute to defence against ER stress, i.e. calnexin and calreticulin (Table 3; item 13.0). In contrast, PIC, especially in combination with IFN-γ, up-regulated several defence/repair genes after both 6 and 24 h. Among them, IFN-γ + PIC modified expression of MnSOD, haeme oxygenase, hsp 70, PEG-3/MyD116 (the rat and mouse homologs of human GADD34) and the DNA repair enzyme O-6-methylguanine-DNA methyltransferase (Table 3; item 12.0).
Among genes related to apoptosis, we observed induction of several putative pro-apoptotic genes by PIC alone or combined to IFN-γ. These included Bax, caspase 2, the cyclin-dependent kinase inhibitor p21/WAF1 and GADD153/CHOP, an ER stress-response transcription factor (Table 3, items 11 and 13.0). Of interest, Bip/GRP78 and bcl-2 expression were unaffected by IFN-γ and/or PIC, at least at our selected time points (see below).
Treatment of beta cells for 6 or 24 h with PIC alone induced expression of genes promoting resistance to viral infection, such as dsRNA-activated protein kinase (PKR, also named eIF-2α kinase), 2′,5′-oligoadenylate synthetase (OAS) and RNA-specific adenosine deaminase (ADAR) (Table 3; item 14.0); addition of IFN-γ neither modified the pattern of expression nor the magnitude of this response. Moreover, IFN-γ alone did not induce any of these genes.
Confirmation by RT-PCR and real-time RT-PCR of genes identified as modified by IFN-γ and/or PIC
The present study is the first attempt to comprehensively define the repertoire of dsRNA- and IFN-γ-induced genes in primary pancreatic beta cells. Using microarray analysis, 3759 genes were detected as expressed in beta cells. Of these, 348 (nearly 10% of the total) were found as changed by dsRNA and IFN-γ, alone or in combination, after 6- or 24-h exposure. The large number of genes modified by dsRNA and IFN-γ, encoding for proteins involved in a broad range of beta-cell functions, emphasises the complex nature of beta-cell responses to two putative mediators of early insulitis. From the genes detected as changed, 38% were either induced by IFN-γ or by dsRNA, while 58% showed an additive or a potentiating effect by the combination of both agents; only 4% of the modified genes were similarly increased/decreased by dsRNA and IFN-γ, without an additive effect by both agents. This, and the fact that dsRNA mostly increases mRNA expression at 6 h (94% of the genes defined as increased), while IFN-γ has mainly an inhibitory effect at this time point (72% of the genes defined as decreased), suggests that dsRNA and IFN-γ signal via different and complementary pathways. Comparison of the present data with our previous microarray analysis of beta cells exposed to IL-1β, alone or in combination with IFN-γ [24, 25], indicates a nearly 50% difference in the pattern of gene expression, suggesting that dsRNA has also important points of differences in signalling as compared to IL-1β. Of note, and in agreement with our previous observations , neither dsRNA nor IFN-γ, nor a combination of both agents, changed beta-cell expression of IL-1β or IL-1α in the microarray analysis. As previously shown [21, 22], and confirmed in the present study, beta cells exposed for 3–6 days to dsRNA + IFN-γ, but not to either agent alone, undergo cell death by apoptosis. These beta cells also present initial adaptive responses that are part of the early host reaction to a viral infection and contribute to amplify immune recognition and immune response against the infective agent . How can we integrate these functional responses with the complex pattern of dsRNA and/or IFN-γ-induced gene expression observed in our microarray analysis? A general model, based on the present data, is proposed in Fig. 5 and discussed below. Due to space limitations, not all genes shown in Fig. 5 and Table 3 are discussed here; additional information on these genes is presented in our previous publications dealing with microarray analysis of pancreatic beta cells [24, 25, 26, 37].
Activation of PKR and of the transcription factor NF-κB are important mediators of dsRNA signal transduction in other cell types [13, 49, 50], and these pathways have been shown to transduce at least part of the effects of dsRNA, including apoptosis, in beta cells [21, 22, 45, 50]. We observed both an induction of the NF-κB precursor p105 and of several well defined beta-cell NF-κB-dependent genes, such as iNOS , MCP-1  and MnSOD ; there was also a nearly 10-fold increase in PKR mRNA expression. Another gene induced to a major extent (more than 50-fold) by dsRNA, in an effect potentiated by IFN-γ, is the transcription factor interferon regulatory factor (IRF)-7. IRF-7 is induced by dsRNA and/or interferons (mostly IFN-α and IFN-β) in other cell types, an effect mediated by the transcription factors NF-κB and STAT-1 [53, 54]. dsRNA increased STAT-1 expression by more than 50-fold, an effect potentiated by IFN-γ at 6 h (present data). IRF-7 and IRF-3 (not found as modified in the array; data not shown) play critical roles in the innate response to a viral infection  and IRF-7 also contributes to IFN-γ-mediated apoptosis . Among the downstream genes regulated by IRF-7, in cooperation with NF-κB, is the CC chemokine RANTES . We observed that PIC increased RANTES expression by more than 100-fold. The agreement between the effects of dsRNA on the expression of genes upstream (STAT-1) and downstream (RANTES) of IRF-7 suggests that this transcription factor participates in the signal transduction of dsRNA in beta cells. Another pathway of dsRNA signalling is via activation of the toll-like receptor 3 (TLR3) , but it is unknown whether this receptor is expressed and functional in beta cells.
While the observations described above suggest a positive feedback on IFN-γ signal transduction, there seems to be a negative feedback operating for dsRNA signalling. Activation of the transcription factor NF-κB has a pro-apoptotic role in beta cells exposed either to IL-1β + IFN-γ [58, 59] or to dsRNA + IFN-γ . MnSOD and IκBα are two genes that might participate in beta-cell defence against apoptosis by decreasing NF-κB activation [60, 61], and both mRNAs were induced by dsRNA. MnSOD is a NF-κB dependent  mitochondrial antioxidant enzyme, and overexpression of MnSOD protects beta cells against immune-mediated damage . NF-κB also regulates IκBα, and increased IκBα concentration both prevents NF-κB migration to the nucleus and removes NF-κB already present in the nucleus .
As discussed above, another effect of dsRNA is to up-regulate PKR . Once activated, PKR phosphorylates, among other substrates, the small subunit of the eukaryotic initiation factor 2α (eIF2α), reducing translation initiation and severely decreasing total cellular protein synthesis. This effect hampers viral replication but, if prolonged, could trigger apoptosis . We observed, however, that beta cells exposed for different time points to dsRNA and IFN-γ have only a 30–40% inhibition of total protein biosynthesis. These apparently divergent findings can be explained by the observation that GADD34 (human homolog of mouse Myd116 or rat progression elevated gene-3), a stress-inducible phosphatase, dephosphorylates eIF2α and induces partial translational recovery after 2–8 h of cellular stress . We presently observed that GADD34 mRNA is up-regulated by dsRNA + IFN-γ, suggesting that this phosphatase provides an additional negative feedback on dsRNA effects.
Despite the up-regulation of IκBα, MnSOD, GADD34 and other putative defence/repair genes (Fig. 5), prolonged exposure of beta cells to dsRNA and IFN-γ eventually culminates in apoptosis. This could be at least in part due to inhibition of several important “defence/repair” genes, and induction of genes that directly contribute to beta-cell death (Table 3; Fig. 5). Decreased expression of “defence” genes  seems to be mostly an early effect of IFN-γ. Two of these genes, calnexin and calreticulin, are chaperones located in the endoplasmic reticulum (ER) , and their early inhibition by IFN-γ could render the beta cells more susceptible to ER stress induced by the subsequent (after 6 h) production of nitric oxide [24, 25, 67, 68]. GADD153, a transcription factor involved in the execution of ER-mediated apoptosis , is up-regulated by dsRNA + IFN-γ at 24 h. IL-1β + IFN-γ also induce GADD153 up-regulation in beta cells [24, 25], an effect mediated by NF-κB activation and consequent increase in iNOS expression and nitric oxide formation . Of note, blocking iNOS activity prevents IL-1β + IFN-γ-induced GADD 153 expression [25, 26], but does not prevent apoptosis in human or rodent beta cells [34, 36] or in insulin-producing cells . Similarly, the use of iNOS blockers decreased dsRNA + IFN-γ-induced beta-cell necrosis, but not apoptosis [21, 22]. Thus, it seems that cytokines or dsRNA + IFN-γ lead to ER stress in beta cells via NO production, but ER stress is not the main mechanism leading to beta cell apoptosis.
Probes for several members of the bcl-2 family of pro- and anti-apoptotic genes , and for different caspases , were present in the array, including bcl-2, bcl-xL, Bak, Bax, Bid, Bad, caspase 1 (ICE), caspase 2, caspase 6 and caspase 7. Of these, only Bax and caspase 2 were found to be modified, both maximally induced by dsRNA and IFN-γ at 24 h. The pro-apoptotic Bax homodimerizes through its BH3 domain, and forms heterodimers with bcl-2 and other proteins. An increased ratio between Bax and bcl-2 contributes to the mitochondrial release of cytochrome c, and other pro-apoptotic proteins, triggering the “execution” phase of apoptosis . dsRNA and IFN-γ increase Bax expression without modifying bcl-2, which could tilt the balance in favour of cell death. Up-regulation of c-Myc was observed under the same experimental conditions and at the same time point as Bax, and it is conceivable that, as described for other cell types , the pathways mediated by both proteins synergize to induce cell death. Caspase 2 is an upstream caspase, contributing to apoptosis by activating executioner caspases, such as caspases 3 and 7 . Caspase-2 activity is also required for translocation of Bax to the mitochondria and the consequent release of cytochrome c . Finally, dsRNA-induced Fas expression might render the beta cells more susceptible to death induced by FasL-expressing mononuclear cells .
Beta cells exposed to dsRNA and IFN-γ have a functional inhibition that precedes cell death [20, 21, 22]. This might be due to excessive production of NO and consequent impairment in glucose oxidation, but other potentially contributory elements were observed in our microarray analysis. Thus, there was a decrease in the glucose transporter GLUT-2, of both insulin and PC-1, an enzyme involved in the conversion of proinsulin to insulin, and in the receptors for the incretins GIP and CCK. We have previously observed that IL-1β + IFN-γ decrease the expression of several genes related to differentiated beta-cell functions and preservation of beta-cell mass [24, 25]; and inhibition of these genes was associated to a 50% decrease in the expression of pdx-1. PDX-1 has a crucial role in maintaining the differentiated phenotype of beta cells [24, 25, 74]. In contrast, dsRNA + IFN-γ decrease insulin mRNA expression and release, without affecting neither pdx-1 mRNA expression nor activity of the pdx-1 promoter (present data). If pdx-1 is not involved in this process of loss of beta-cell differentiated functions, which other genes could participate in it? c-Myc was detected as up-regulated in the array, and increased expression of this oncogene suppresses insulin gene transcription by inhibiting NeuroD/BETA2 . Another intriguing finding was the up-regulation of Notch1, delta-1 (both induced by dsRNA + IFN-γ at 6 h), and of presenilin-2 (induced by dsRNA and IFN-γ at 24 h). Delta-1 is a ligand of the Notch receptor, while presenilins-1 and 2 are enzymes responsible for the intramembraneous proteolysis and activation of Notch [74, 76, 77, 78]. Differentiation is inhibited in endocrine precursor cells expressing activated Notch receptors, whereas the signalling cells (expressing delta-1) are free to differentiate into endocrine cells . It is thus conceivable that dsRNA and IFN-γ-induced re-expression of genes of the Notch signalling pathway contributes to both the loss of the differentiated beta-cell phenotype, and, together with the observed inhibition of GH receptor, prevents in a paracrine fashion the growth/differentiation of newly generated beta cells.
The first line of cellular defence against a viral infection is provided by the local innate immunity, which is followed by the adaptive immune response. Both processes are, at least to some extent, integrated [79, 80]. We have detected induction of several genes that might impair intracellular viral proliferation, including PKR, MX3, double-stranded RNA-specific adenosine deaminase (ADAR) and 2′,5′-oligoadenylate synthetase (OAS). These mRNAs were up-regulated by dsRNA at both 6 and 24 h, with little or no additive effect by IFN-γ. MXs are large GTPases which interfere with viral replication and spread , while ADAR hyperedits dsRNAs by converting adenosine to inosine, targeting the dsRNAs for cleavage and removal from the cytosol . OAS activates RNAseL, which both decreases total protein synthesis and accelerates the degradation of RNA, affecting viral replication but also contributing to cell dysfunction and eventually apoptosis . To mount the “second line” of defence, namely the adaptive immune response, the infected cells must present the viral antigens bound to HLA molecules and attract immune competent cells to the site of infection. As can be seen from Fig. 5, beta cells exposed to dsRNA and/or IFN-γ express several genes related to antigen processing and presentation in the context of MHC class I molecules, and also up-regulate several chemokines, adhesion molecules and cytokines that contribute to homing and activation of macrophages, dendritic cells and T-cells. We and others have described before, by differential display and microarray analysis, induction of several of these chemokines, cytokines and MHC-related molecules by IL-1β and IFN-γ [24, 25, 84], and also confirmed their expression at the mRNA and protein level in both rodent and human islets, and in islets isolated from pre-diabetic NOD mice [68, 84, 85]. Moreover, by use of RT-PCR and “gene-by-gene’ analysis we showed that beta cells exposed to dsRNA and IFN-γ express IP-10, MCP-1, MIP-3α and fractalkine , findings confirmed in the present microarray analysis. One chemokine, however, was not described before in beta cells, namely RANTES. The C-C chemokine RANTES (CCL5) attracts monocytes, activated T-cells and immature dendritic cells during inflammation and immune responses, suggesting a role for RANTES in virus-related diseases . Of special interest, RANTES expression in microglia correlates with the initial symptoms of experimental autoimmune encephalomyelitis [87, 88] and the chemokine, together with IP-10, MCP-3 and MCP-5, contributes to the distinct Th1 islet inflammatory infiltrate leading to beta-cell destruction in NOD.scid mice infused with islet-specific TCR transgenic CD4 cells .
Most information available on the broad molecular effects of dsRNA have been obtained in tumoural cell lines , and little is known about the effects of dsRNA or actual viral infection on gene regulation by non- or poorly-dividing cells, such as beta cells. It is conceivable that some of the mechanisms that allow one cell type to eradicate a viral infection in a non-cytopathic and cytokine-dependent way might cause death in another cell type . This possibility is of special relevance for Type 1 diabetes, where both viruses and their product, dsRNA, together with locally produced cytokines, such as IFN-γ and IL-1β, probably play an important role in the initiation and progression of insulitis. Why and how the cellular attempts to eradicate/neutralise the invading virus go wrong in some individuals, giving rise to progressive inflammation and beta-cell death, remain to be determined. Our study, by showing large scale evaluation of mRNAs modified by dsRNA and IFN-γ in beta cells, provides valuable information to answer this question, and allowed us to propose a comprehensive hypothesis for the molecular regulation of the different cellular responses involved (Fig. 5). The present “data driven” hypothesis needs now to be tested by both targeted “hypothesis driven” experiments and by new microarray and proteomic analysis of rodent and human islet cells exposed either to dsRNA, in the presence of blockers of key signalling pathways, or infected with viruses with a putative pathogenic role in human Type 1 diabetes.
This work was supported by grants from the Juvenile Diabetes Foundation International (JDRF) and the Fonds National de la Recherche Scientifique (FNRS), Belgium. We thank the personnel from the Laboratory of Experimental Medicine, ULB, MA. Neef, J. Schoonheydt, M. Urbain and G. Vandenbroeck for technical assistance and C. Demesmaeker for secretarial help. We thank also R. Leeman and the personnel involved in beta-cell purification at the Diabetes Research Center, Vrije Universiteit Brussel, for help in the initial part of the study. This work has been conducted in collaboration with and supported by the JDRF Center for Prevention of Beta-cell Destruction in Europe under grant number 4-2002-457.
Table S1 Complete list of genes modified after a 6 or 24 hours exposure to IFN-gamma and/or PIC