and Immunity (2007) 8, 232–238 & 2007 Nature Publishing Group All rights reserved 1466-4879/07 $30.00 www.nature.com/gene

ORIGINAL ARTICLE Transcriptional profiling of type 1 diabetes genes on 21 in a rat beta-cell line and human pancreatic islets

R Bergholdt1, AE Karlsen1,2, PH Hagedorn3, M Aalund4, JH Nielsen5, M Kruhøffer6, T Ørntoft6, H Wang7, CB Wollheim7, J Nerup1,2 and F Pociot1,2 1Steno Diabetes Center, Gentofte, Denmark; 2Institute for Clinical Science, University of Lund, Lund, Sweden; 3Biosystems Department, Risø National Laboratory, Technical University of Denmark, Roskilde, Denmark; 4Neurotech A/S, Copenhagen, Denmark; 5Department of Medical Biochemistry and Genetics, University of Copenhagen, Copenhagen, Denmark; 6Molecular Diagnostic Laboratory, Aarhus University Hospital, Skejby, Aarhus, Denmark and 7Department of Cell Physiology and Metabolism, University Medical Center, Geneva, Switzerland

We recently finemapped a type 1 diabetes (T1D)-linked region on , indicating that one or more T1D-linked genes exist in this region with 33 annotated genes. In the current study, we have taken a novel approach using transcriptional profiling in predicting and prioritizing the most likely candidate genes influencing beta-cell function in this region. Two array- based approaches were used, a rat insulinoma cell line (INS-1ab) overexpressing pancreatic duodenum homeobox 1 (pdx-1) and treated with interleukin 1b (IL-1b) as well as human pancreatic islets stimulated with a mixture of cytokines. Several candidate genes with likely functional significance in T1D were identified. Genes showing differential expression in the two approaches were highly similar, supporting the role of these specific products in cytokine-induced beta-cell damage. These were genes involved in cytokine signaling, oxidative phosphorylation, defense responses and . The analyses, furthermore, revealed several factor binding sites shared by the differentially expressed genes and by genes demonstrating highly similar expression profiles with these genes. Comparable findings in the rat beta-cell line and human islets support the validity of the methods used and support this as a valuable approach for gene mapping and identification of genes with potential functional significance in T1D, within a region of linkage. Genes and Immunity (2007) 8, 232–238. doi:10.1038/sj.gene.6364379; published online 1 March 2007

Keywords: transcriptional profiling; type 1 diabetes; human islets; ; chromosome 21; candidate gene

Introduction genes for further analysis. In the present study, we have used transcriptional profiling in selecting the most likely Identification of susceptibility genes in complex genetic candidate genes based on functional significance. T1D diseases is a major challenge. In the process of identify- (MIM 222100) is a complex genetic disease of unknown ing disease-associated genes, classically two strategies etiology. In a previous T1D genome scan,1 a region on have been used, the functional candidate gene approach chromosome 21 showed some evidence of linkage to and the positional approach. The functional candidate T1D, especially in the Danish population (LOD 2.33, gene approach takes advantage of existing knowledge P ¼ 0.009). We have recently finemapped this region, also about the disease and genes are selected based on this in the Danish population, to a 6.3 Mb interval with a peak knowledge. This approach has been extensively used, NPL score of 3.61 (P ¼ 0.0002).2 Thirty-three known although, with varying success in type 1 diabetes (T1D). genes were identified in this region (www.ncbi.nlm.nih. With the positional approach, linkage analysis is used to gov), and all known single-nucleotide polymorphisms determine location of disease-linked regions. The (SNPs) in coding regions of these genes (dbSNP, http:// gene(s), thus identified, are subsequently further exam- www.ncbi.nlm.nih.gov/SNP/index.html, build 116) ined and their functional significance is evaluated. were examined. Globally, none of the SNPs showed Linkage peaks typically cover megabase (Mb)-wide association to T1D.2 However, subsets of data (e.g. regions, but workload, data management and costs for families responsible for the linkage signal) demonstrated finemapping can be significantly reduced by prioritizing T1D association of four genes: TTC3, OLIG2, KCNE1 and CBR1. From a statistical point, all 33 genes, positioned underneath the linkage peak, are equally good candidate Correspondence: Professor F Pociot, Steno Diabetes Center, Niels genes. Consequently, they should all be extensively Steensensvej 2, DK-2820 Gentofte, Denmark. E-mail: [email protected] studied. Expression levels of the genes, may, however, Received 13 November 2006; revised 21 January 2007; accepted 22 act as intermediate between genomic DNA January 2007; published online 1 March 2007 sequence variations and more complex disease pheno- Transcriptional profiling of T1D genes on Chr. 21 R Bergholdt et al 233 types;3 and variation in gene expression has been shown critical region 1, -1) and CRYZL1 in large part as an outcome of polymorphisms in DNA (quinine -like 1) genes were downregu- sequence.3,4 We have, therefore, taken a novel approach lated when comparing pdx-1-overexpressing cells stimu- for gene mapping, which uses transcriptional profiling as lated with IL-1b vs unstimulated pdx-1-overexpressing a method to select candidate genes within a region, that cells (Figure 1 and Table 1). In contrast, the ATP5O (ATP- is, prioritizing the genes of interest for further investiga- synthase O subunit) and IFNGR2 (interferon-gamma tion. In the present study, we have used model systems receptor beta chain precursor) genes were upregulated in primarily addressing beta-cell function and response. the presence of IL-1b (Figure 1 and Table 1). Additionally, Gene expression was evaluated by means of two array- the IFNAR1 (interferon-alpha/beta receptor alpha- based approaches. The first approach was based on a rat chain precursor) (P ¼ 0.01), DSCR5 (phosphatidylinositol insulinoma cell line (INS-1ab) overexpressing pancreatic N-acetylglucosaminyltransferase subunit P, PIGP) duodenum homeobox 1 (pdx-1).5,6 Beta-cell maturation is (P ¼ 0.03), TIAM1 (T-lymphoma invasion and metasta- dependent on sequential activation of different transcrip- sis-inducing protein 1) (P ¼ 0.01) and OLIG1 (oligoden- tion factors, including pdx-1,5 and is accompanied by drocyte transcription-factor 1) (P ¼ 0.001) genes were all increased sensitivity to the toxic effects of interleukin 1b upregulated by pdx-1 overexpression at 2 h and/or 24 h, (IL-1b).7 Cytokines have been shown to play a key role in as compared with the control situation without pdx-1 the destruction of beta cells,8 and we have demonstrated overexpression, independently of IL-1b stimulation. The an increased sensitivity to IL-1b in INS-1ab overexpres- CHAF1B (chromatin assembly factor 1 subunit-B) gene sing pdx-1.6 Extensive microarray analysis was carried was downregulated (P ¼ 0.003) by pdx-1 overexpression, out with these cells treated with IL-1b for different Table 1. periods of time. In the second approach, human Second, human pancreatic islet preparations from pancreatic islets from eight donors were examined in eight donors were examined in a separate array-based another array-based evaluation of gene expression of the evaluation of gene expression of all 33 positional human 33 positional human candidate genes with and without candidate genes with and without stimulation with a stimulation with a mixture of cytokines, mimicking the mixture of cytokines. Several of the chromosome 21 presumed pathogenesis of T1D in humans. candidate genes demonstrated expression levels that were downregulated in human islets following cytokine exposure, that is, ITSN1 (intersectin-1), TTC3 (tetratrico- peptide repeat protein-3), CHAF1B, HUNK (hormonally Results upregulated neu tumor-associated kinase), OLIG1 and Gene expression was evaluated by means of two CBR1 (carbonyl reductase 1) (Figure 2 and Table 1). Of different array-based approaches focusing on beta-cell these, HUNK and ITSN1 did not have a homologous function. Expression was evaluated in a rat beta-cell line gene in the rat and TTC3 was not on the rat Affymetrix overexpressing pdx-1, exposed to IL-1 for 2 and 24 h. Of chip. Genes demonstrated to be differentially expressed the 33 human genes, 27 have homologues in the rat. in both approaches were CHAF1B, OLIG1 and CBR1/ Twenty of these were represented by one or more probe CBR3. sets on the Affymetrix rat chip used, and were subjected Common expression patterns of genes may reflect to analysis in this rat beta-cell system. Of these, the CBR3 shared regulatory mechanisms. Promoter regions of the (carbonyl reductase 3), DSCR1 (Down’s syndrome selected genes were, therefore, compared for common

Figure 1 Expression levels in rat genes (corresponding to human chromosome 21 genes), which are differentially expressed after IL-1b stimulation. Expression levels are expressed relative to unstimulated condition, and are provided for 2 and 24 h of IL-1b stimulation. *Comparisons with P-values o0.05. The CBR3 (P ¼ 0.04), DSCR1 (P ¼ 0.001) and CRYZL1 (P ¼ 0.03) genes were downregulated after IL-1b stimulation, whereas the IFNGR2 (P ¼ 1.4 Â 10À7) and ATP5O (P ¼ 0.006) genes were upregulated in the presence of IL-1b. Findings observed in this model system of course need confirmation in additional functional studies using other systems.

Genes and Immunity Transcriptional profiling of T1D genes on Chr. 21 R Bergholdt et al 234 Table 1 Results in human pancreatic islets and rat insulinoma cells

Human isletsCorresponding Represented on Rat insulinoma INS-1ab pdx-1 system rat gene Affymetrix Human gene Expression change in rat chip Genes significantly Genes significantly human islets by cytokine regulated by IL-1 regulated by pdx-1 stimulation stimulation of pdx-1 cells stimulation

CLDN8 —+ + — — TIAM1 —+ + — Yes SOD1 —+ + — — OLIG2 —+ + — — IFNAR1 —+ + — Yes IFNGR2 — + + Upregulated — KCNE1 —+ + — — KCNE2 —+ + — — ATP50 — + + Upregulated — DSCR1 — + + Downregulated — SON —+ + — — SYNJ1 —+ + — — RUNX1 —+ + — — OLIG1 Downregulated + + — Yes CBR1 Downregulated + + — — CBR3 — + + Downregulated — CHAF1B Downregulated + + — Yes CLDN14 —+ + — — DSCR5 —+ + — Yes CRYZL1 — + + Downregulated — CLDN17 —+— GART —+— CLIC6 —+— SIM2 —+— HLCS —+— DSCR6 —+— TTC3 Downregulated + — HUNK Downregulated None — TCP10L — None — IL10RB — None — IFNAR2 — None — MRPS6 — None — ITSN1 Downregulated None —

Thirty-three human chromosome 21 genes are listed, with indication of whether a corresponding rat gene exists, and whether represented by a probe set on the Affymetrix rat chip used. Genes showing differential expression in either system (human islets or rat insulinoma cells) are indicated. Genes represented on the Affymetrix rat chip is listed in physical map order.

transcription factor binding sites. We initially compared pdx1-system (Table 1), were used as input for genes having a human chromosome 21 annotation searches in a protein interaction database (potential (Ensembl), which demonstrated a differential level of interactions of proteins, www.bmm.icnet.uk/~pip).9 For expression in either the pdx-1 system or the human each of the interacting proteins, the corresponding rat islets. Several transcription factor binding sites shared by gene, if present, was identified on the chip. Genes, whose several of the genes were identified (Supplementary corresponding protein could be predicted to interact Table 1), but no statistically significant overrepresenta- with the input chromosome 21 gene and, which in tion of any transcription factor binding site in these addition demonstrated expression profiles similar to the promoters was detected, although data clearly suggested original gene (correlation coefficient40.7) were selected shared regulatory mechanisms. Additionally, analyses (Table 2). Furthermore, genes, which could not be shown were carried out separately in the group of genes to interact at the protein level with the input gene, but downregulated by IL-1b/cytokine mix, the group upre- which showed a highly similar expression profile gulated by IL-1b, as well as the group upregulated by (correlation coefficient40.95) to one or more of the 10 pdx-1 overexpression. In none of the groups, we were input genes were identified as well (Supplementary able to demonstrate significantly overrepresented tran- Table 2). In total, 68 additional rat genes (with LocusLink scription factor binding sites. IDs) were hereby identified. Each of these groups of For the pdx-1-overexpressing INS-1ab cell system, the genes was tested for overrepresented expression levels of more than 30 000 rat genes were (GO) terms. One term, biological process: ‘GO:0006952 measured on the rat chip. We used several approaches to Defense response’, was significantly overrepresented extract additional information from this system. The among the group of genes highly similar in expression corresponding proteins, for the 10 differentially ex- to IFNGR2 (Supplementary Table 2) (corrected P- pressed chromosome 21 homologous genes in the valueo0.032). Specifically, four genes were annotated

Genes and Immunity Transcriptional profiling of T1D genes on Chr. 21 R Bergholdt et al 235

Figure 2 Expression levels of differentially expressed genes in human pancreatic islets. Grey bars represent unstimulated conditions (expression level defined as 1). Black bars represent cytokine-stimulated conditions related to expression levels in unstimulated condition of the same islet preparation (numbered 1–8 on the x-axis). The y-axis shows fold up- or downregulation. Significant upregulation after cytokine exposure, as expected, for several control genes, for example, NOS2A (Figure 2), IL6 and CASP1 (data not shown) was demonstrated. The following chromosome 21 candidate gene transcripts demonstrated expression levels that were all downregulated in human islets following cytokine exposure: ITSN1 (P ¼ 0.003), TTC3 (P ¼ 0.02), CHAF1B (P ¼ 0.02), HUNK (P ¼ 0.02), OLIG1 (P ¼ 0.05) and CBR1 (P ¼ 0.22). Of these, HUNK and ITSN1 did not have a homologous gene in the rat and TTC3 are not found on the rat Affymetrix chip.

with this GO term: IL15, FAS, ICAM1 as well as SOD2 molecular basis of complex traits, like T1D. As our (rat gene Sodm), which have all previously been recent fine-mapping data strongly suggest the existence associated with T1D.10–16 FAS, ICAM1 and SOD2 shared of T1D-linked gene(s) in a region on human chromosome transcription factor binding sites in their promoter 21, further characterization of the candidate genes herein regions for NFKAPPAB65 and NFKAPPAB_01, which is important. Transcriptional profiling is a novel and are also found in other genes in this group of genes with innovative approach, which has proven its feasibility. By similar expression, that is, IFNGR2 and MAPK6. Other selecting genes that change expression levels after genes in the group of genes with expression patterns cytokine exposure in human islets and in a rat beta-cell highly similar to IFNGR2, have common binding sites for line, we believe it is possible to select genes with a likely NFkbib. functional significance in T1D pathogenesis. Functional human candidate genes are genes that exhibit changes in expression level during cytokine exposure, modeling the Discussion pathogenetic process in the target organ of T1D, that is, up- or down-regulation, as well as changes from not Identification of susceptibility genes in complex genetic being expressed to being expressed after cytokine diseases is a major challenge. High heritability of exposure. variation in gene expression has been demonstrated3 In the current study, we have used models, which and suggests that identification of genetic determinants preferentially focus on beta-cell function and response. of gene expression may provide insights into the Despite the fact that T1D is an immune-mediated disease

Genes and Immunity Transcriptional profiling of T1D genes on Chr. 21 R Bergholdt et al 236 Table 2 Interacting proteins with similar expression profile studies,2 substantiating the relevance of this gene in T1D. CBR1/CBR3 are two very similar neighboring Ensembl ID Distance Gene name link genes, believed to be able to substitute each other in at least some biological processes. We have previously, CBR3 in proteome analyses of IL-1b-exposed islets from ENSRNOG00000001701 0.000 Cbr3 304078 diabetes-prone rats, demonstrated Cbr-1 protein to be ENSRNOG00000001134 0.073 RFC5 304528 b 18 — 0.086 CHEK2 114212 downregulated fivefold by IL-1 exposure. The CBR1, ENSRNOG00000008755 0.132 ACOX1 50681 CBR3 and ATP5O genes are furthermore involved in ENSRNOG00000008841 0.196 ORC1L 313479 oxidative phosphorylation with potential implication — 0.239 COPB2 60384 for , believed to be of importance in ENSRNOG00000005762 0.246 RAB22A 366265 beta-cell destruction. Of additional interest are the ENSRNOG00000012710 0.285 PHGDH 361094 IFNGR2 and IFNAR1 genes involved in interferon CHAF1B signaling and map kinase signaling cascades. Cytokine ENSRNOG00000001692 0.000 Chaf1b 288242 and cytokine receptor genes are obvious T1D functional ENSRNOG00000029061 0.064 HIRIP3 361650 candidate genes because of the presumed key role of ENSRNOG00000001245 0.147 PCBP3 294336 cytokines in the pathogenesis of the disease. A main ENSRNOG00000019857 0.166 GNG7 58979 finding in the rat beta-cell line was the IFNGR2 gene, ENSRNOG00000010386 0.220 H2AFX which was significantly upregulated by IL-1b. The — 0.237 SMARCC1 301020 ENSRNOG00000009799 0.254 NP_955420.1 300668 IFNGR2 gene was, however, not differentially expressed in the human pancreatic islets, which might represent TIAM1 differences in signaling between human and rat.19 ENSRNOG00000021569 0.000 Tiam1 304109 Furthermore, interferon-g (IFN-g) was used for stimula- ENSRNOG00000024667 0.227 ZCCHC7 298086 tion of the human islets as opposed to the rat beta-cell ENSRNOG00000008921 0.295 Dlc2 140734 line studies, where only IL-1b was used. The two Down’s IFNGR2 syndrome critical region proteins (DSCR1 and DSCR5), ENSRNOG00000002032 0.000 Ifngr2 360697 differentially expressed in the IL-1b stimulated pdx-1 ENSRNOG00000032948 0.096 TYK2 cells, may also be of potential interest. We have recently ENSRNOG00000015547 0.254 JAK2 24514 demonstrated the prevalence of Down’s syndrome in T1D to be four times increased, supporting genes on ATP5O chromosome 21 to be responsible for this higher ENSRNOG00000001991 0.000 Atp5o 192241 20 ENSRNOG00000019223 0.250 Atp5c1 116550 prevalence. Such genes might also be important in T1D in general. The analyses revealed several transcription factor Interacting proteins with similar expression profile (correlation binding sites shared by several of the promoters of the coefficient40.7) are shown. The Pearson correlation coefficient, differentially expressed genes; none, however, demon- which captures similarity in shape, was calculated (based on all time points and treatments) between each of the input genes and all strated statistically significant overrepresentation. This genes whose protein interacts with the corresponding protein for suggests common regulation and functional interaction that input gene. The threshold of 40.7 was chosen after visual of the chromosome 21 region with regions on other comparison of expression profiles. ‘Distance’ refers to 1-correlation (e.g. via transcription factors), rather than coefficient. among the chromosome 21 genes themselves. In conclusion, we demonstrated several candidate genes with a likely functional significance in T1D. Several of the genes, identified by these approaches, and the few validated T1D genes to date are immune were genes involved in T1D relevant pathways as signal system genes, we believe the immune system destroys transduction, oxidative phosphorylation, defense re- beta cells because of features in beta cells themselves, sponses and apoptosis. Data prove the feasibility and for example, the way they react to the immunological usefulness of using expression profiling in relevant challenge.17 To explore this further, we therefore focused tissues for gene mapping, and in selecting functional on beta-cell genes and used the cytokine stimulation candidate genes presumed to underlie complex genetic model, mimicking what is happening in the target tissue diseases. in T1D pathogenesis. We can, of course, not rule out that important disease genes are expressed only in immuno- logical tissue. Although it is not always straightforward to extra- Materials and methods polate from one species to the other, high concordance was observed between expression changes in the rat Pdx-1-overexpressing INS-1 cells beta-cell line and human pancreatic islets. Genes Cell culture: The tetracycline-inducible, stably transfected demonstrated to be differentially expressed in both INS-1ab cell line has been described previously.5 Cells approaches were CHAF1B, OLIG1 and CBR1/CBR3. were cultured in RPMI 1640 with or without 500 ng/ml Additionally, TTC3 was downregulated in human doxycycline. After 24 h, IL-1b (40 ng/ml)6 was added, pancreatic islets and demonstrated T1D association in and cells were harvested at 2, 4, 6, 12 and 24 h, our genetic studies.2 The molecular function of the respectively. Insulin and nitric oxide measurements were protein encoded by the TTC3 gene is, however, largely carried out on media to confirm an effect of the IL-1b unknown. The CBR1 gene was also associated to T1D in exposure.6 RNA and array processing were carried out as families demonstrating linkage in our previous genetic described in.5

Genes and Immunity Transcriptional profiling of T1D genes on Chr. 21 R Bergholdt et al 237 Gene expression in INS-1 cells Searching for common transcription factor binding sites Affymetrix RG 230 (Affymetrix Inc., Santa Clara, CA, Searches for transcription factor binding sites were USA) 2.0 oligonucleotide arrays were scanned (Affy- carried out using an algorithm based on a frequency metrix GeneChip Scanner) and probe intensities were residue model as implemented in Toucan version 2.3.2 21 normalized and log2-transformed. Significant differen- (www.homes.esat.kuleuven.be/~saerts/software/toucan. tial expression was assessed by using multivariate php)26 with links to relevant databases, containing analysis considering all arrays simultaneously to judge human and rat background models and frequency files, statistical significance.22 Briefly, for each gene, empirical respectively. Testing for statistically overrepresented Bayes-moderated F-statistics were used to test for transcription factor binding sites were carried out using differences between samples. Genes were considered to a binomial test (Po0.05 and significance value 40), also be significantly differentially regulated, if P-values implemented in the Toucan software. adjusted for multiple testing using the false discovery rate (FDR)23 were o0.05. Microarray studies do test many hypotheses; however, use of other methods than conservative mass significance-controlling methods are Acknowledgements generally accepted.24,23 A gene-specific FDR is inter- preted as the Bayesian probability that a gene, declared We thank Susanne Munch and Bodil Bosmann Jørgensen to be differentially expressed, is a false positive.24,23 The for excellent technical expertise. Support from the Pearson correlation coefficient (which captures similarity Foundation of 17-12-1981, the Sehested-Hansen Founda- in shape), based on all time points and treatments, was tion, the Danish Diabetes Association, the AP Moller used to evaluate similarity in expression profiles among Foundation for the Advancement of Medical Science, the genes. Overrepresented GO terms (www.geneontology. European Foundation for the Study of Diabetes (Type 1 org) among groups of genes were detected by calculating Diabetes Research Grant), the Juvenile Diabetes Research P-values using the hypergeometric distribution and Foundation (RFA DK-99-002) as well as the Danish adjusting for multiple testing, using the Bonferroni Medical Research Council is greatly acknowledged. correction. When overrepresented GO terms were found in this manner, the corresponding biological processes were assumed to be of special importance among the group of genes tested. References 1 Nerup J, Pociot F, Holm P, Julier C, Kochum I, Senee V et al. A Human pancreatic islets genomewide scan for Type 1-diabetes susceptibility in Scandinavian families: identification of new loci with evidence Human pancreatic islets were obtained as samples of interactions. Am J Hum Genet 2001; 69: 1301–1313. from a multicenter European Union-supported program 2 Bergholdt R, Nerup J, Pociot F. Fine mapping of a region on on beta-cell transplantation in diabetes directed by chromosome 21q21.11–q22.3 showing linkage to type 1 Professor D Pipeleers. The program has been approved diabetes. J Med Genet 2005; 42: 17–25. by central and local ethical committees. Eight human 3 Cheung V, Conlin L, Weber T, Arcaro M, Jen K, Morley M et al. pancreatic islet preparations were used, five from male Natural variation in human gene expression assessed in donors and three from female donors, aged 8–57 years. lymphoblastoid cells. Nat Genet 2003; 33: 422–425. The islets were stimulated with a mixture of cytokines, 4 Morley M, Molony C, Weber T, Devlin J, Ewens K, Spielman R consisting of TNF-a (5000 U/ml), IFN-g (750 U/ml) and et al. Genetic analysis of genome-wide variation in human b gene expression. Nature 2004; 430: 743–747. IL-1 (75 U/ml). cDNA was prepared from total RNA 5 Wang H, Maechler P, Ritz-Laser B, Hagenfeldt K, Ishihara H, by oligo-dT-primed reverse transcription, as described Philippe J et al. Pdx1 level defines pancreatic gene expression by the manufacturer (TaqMan RT reagents, Applied pattern and cell lineage differentiation. J Biol Chem 2001; 276: Biosystems, Foster City, CA, USA). 25279–25286. 6 Nielsen K, Kruhoffer M, Orntoft T, Sparre T, Wang H, Wollheim C et al. Gene expression profiles during Gene expression in human islets maturation and after IL-1 beta exposure reveal important roles Gene expression quantification in eight human islet of Pdx-1 and Nkx6.1 for IL-1 beta sensitivity. Diabetologia 2004; preparations was evaluated by Low Density Array Cards 47: 2185–2199. (Applied Biosystems), containing assays for all 33 genes 7 Nielsen K, Karlsen AE, Deckert M, Madsen OD, Serup P, Mandrup-Poulsen T et al. Beta-cell maturation leads to in vitro as well as several control genes. These are custom arrays, sensitivity to cytotoxins. Diabetes 1999; 48: 2324–2332. very sensitive (less than 1% divergence among dupli- 8 Mandrup-Poulsen T, Helqvist S, Molvig J, Wogensen L, Nerup cates) and based on real-time PCR and TaqMan (7900HT, J. Cytokines as immune effector molecules in autoimmune Applied Biosystems) technology. Low Density Array endocrine diseases with special reference to insulin-depen- Cards were processed according to instructions by the dent diabetes mellitus. Autoimmunity 1989; 4: 191–218. manufacturer. Data were analyzed using the SDS 2.1 9 Jonsson PF, Cavanna T, Zicha D, Bates PA. Cluster analysis of software (Applied Biosystems) and evaluated by the networks generated through homology: automatic identifica- 2ÀDDCt method,25 normalizing all gene expressions to the tion of important protein communities involved in cancer expression of the housekeeping gene GAPDH, and metastasis. BMC Bioinformatics 2006; 7:2. 10 McInnes IB, Gracie JA. Interleukin-15: a new cytokine target subsequently comparing each islet preparation for each for the treatment of inflammatory diseases. Curr Opin gene in cytokine-stimulated condition vs unstimulated Pharmacol 2004; 4: 392–397. condition. Expression levels were compared using a 11 Cardozo AK, Proost P, Gysemans C, Chen MC, Mathieu C, paired t-test, P-values of o0.05 were considered statis- Eizirik DL. IL-1beta and IFN-gamma induce the expression of tically significant for differential expression. diverse chemokines and IL-15 in human and rat pancreatic

Genes and Immunity Transcriptional profiling of T1D genes on Chr. 21 R Bergholdt et al 238 islet cells, and in islets from pre-diabetic NOD mice. 19 Eizirik DL, Pavlovic D. Is there a role for nitric oxide in Diabetologia 2003; 46: 255–266. beta-cell dysfunction and damage in IDDM? Diabetes Metab 12 Savinov AY, Tcherepanov A, Green EA, Flavell RA, Rev 1997; 13: 293–307. Chervongsky AV. Contribution of fas to diabetes develop- 20 Bergholdt R, Eising S, Nerup J, Pociot F. Increased prevalence ment. Proc Natl Acad Sci USA 2003; 100: 628–632. of Down’s syndrome in individuals with type 1 diabetes in 13 Vence L, Benoist C, Mathis D. Fas deficiency prevents type 1 Denmark: a nationwide population-based study. Diabetologia diabetes by inducing hyporesponsiveness in islet beta-cell- 2006; 49: 1179–1182. reactive T-cells. Diabetes 2004; 53: 2797–2803. 21 Wu Z, Irizarry R, Gentleman R, Murillo F, Spencer F. A model- 14 Hohmeier HE, Thigpen A, Tran VV, Davis R, Newgard CB. based background adjustment for oligonucleotide expression Stable expression of manganese superoxide dismutase arrays. Technical report, John Hopkins University, Department of (MnSOD) in insulinoma cells prevents IL-1beta-induced Biostatistics Working Papers. Baltimore, MD, USA, 2003. cytotoxicity and reduces nitric oxide production. J Clin Invest 22 Smyth GK. Linear models and empirical bayes methods for 1998; 101: 1811–1820. assessing differential expression in microarray experiments. 15 Martin S, van den Engel NK, Vinke A, Heidenthal E, Schulte Stat Appl Genet Mol Biol 2004; 3: Article 3. http://www. B, Kolb H. Dominant role of intercellular adhesion molecule-1 bepress.com/sagmb/vol3/iss1/art3. in the pathogenesis of autoimmune diabetes in non-obese 23 Benjamini Y, Hochberg Y. Controlling the false discovery rate – diabetic mice. J Autoimmun 2001; 17: 109–117. a practical and powerful approach to multiple testing. J Roy 16 Camacho SA, Heath WR, Carbone FR, Sarvetnick N, LeBon A, Sta B 1995; 57: 289–300. Karlsson L et al. A key role for ICAM-1 in generating effector 24 Allison DB, Cui X, Page GP, Sabripour M. Microarray data cells mediating inflammatory responses. Nat Immunol 2001; 2: analysis: from disarray to consolidation and consensus. Nat 523–529. Rev Genet 2006; 7: 55–65. 17 Nerup J, Mandrup-Poulsen T, Helqvist S, Andersen HU, 25 Livak K, Schmittgen T. Analysis of relative gene expression Pociot F, Reimers JI et al. On the pathogenesis of IDDM. data using real-time quantitative PCR and the 2(-Delta Delta Diabetologia 1994; 37: S82–S89. C(T)) method. Methods 2001; 25: 402–408. 18 Sparre T, Christensen UB, Larsen PM, Fey SJ, Wrzesinski K, 26 Aerts S, Van-Loo P, Thijs G, Mayer H, de-Martin R, Moreau Y Roepstorff P et al. IL-1 beta induced protein changes in et al. TOUCAN 2: the all-inclusive open source workbench for diabetes prone BB rat islets of Langerhans identified by regulatory sequence analysis. Nucleic Acids Res 2005; 33: proteome analysis. Diabetologia 2002; 45: 1550–1561. W393–W396.

Supplementary Information accompanies the paper on Genes and Immunity website (http://www.nature.com/gene)

Genes and Immunity