Genome-Wide Microarray Expression

Home , CD99L2, Entrez, LANCL1, SUCLG2, ZC3H11A

The Journal of Immunology

Genome-Wide Microarray Expression Analysis of CD4؉ T Cells from Nonobese Diabetic Congenic Mice Identiﬁes Cd55 (Daf1) and Acadl as Candidate Genes for Type 1 Diabetes1

Junichiro Irie,* Brian Reck,† Yuehong Wu,* Linda S. Wicker,‡ Sarah Howlett,‡ Daniel Rainbow,‡ Eleanor Feingold,† and William M. Ridgway2*

NOD.Idd3/5 congenic mice have insulin-dependent diabetes (Idd) regions on chromosomes 1 (Idd5)and3(Idd3) derived from the nondiabetic strains B10 and B6, respectively. NOD.Idd3/5 mice are almost completely protected from type 1 diabetes (T1D) but the genes within Idd3 and Idd5 responsible for the disease-altering phenotype have been only partially characterized. To test the hypothesis that candidate Idd genes can be identiﬁed by differential gene expression between activated CD4؉ T cells from the diabetes-susceptible NOD strain and the diabetes-resistant NOD.Idd3/5 congenic strain, genome-wide microarray expression analysis was performed using an empirical Bayes method. Remarkably, 16 of the 20 most differentially expressed genes were located in the introgressed regions on chromosomes 1 and 3, validating our initial hypothesis. The two genes with the greatest differential RNA expression on chromosome 1 were those encoding decay-accelerating factor (DAF, also known as CD55) and acyl-coenzyme A dehydrogenase, long chain, which are located in the Idd5.4 and Idd5.3 regions, respectively. Neither gene has been implicated previously in the pathogenesis of T1D. In the case of DAF, differential expression of mRNA was extended to the protein level; NOD CD4؉ T cells expressed higher levels of cell surface DAF compared with NOD.Idd3/5 CD4؉ T cells following activation with anti-CD3 and -CD28. DAF up-regulation was IL-4 dependent and blocked under Th1 conditions. These results validate the approach of using congenic mice together with genome-wide analysis of tissue-speciﬁc gene expression to identify novel candidate genes in T1D. The Journal of Immunology, 2008, 180: 1071–1079.

n complex genetic diseases such as type 1 diabetes (T1D),3 cessful in excluding candidate genes and narrowing the list of pos- dozens of allelic variants interact to promote or discourage sible candidate genes (5–14). I disease manifestation (1–3). Similar ﬁndings have been made There is a practical limit, however, to ﬁne-mapping disease in animal models of complex genetic disease such as the NOD genes using congenic strains of mice since recombination in the mouse, which spontaneously develops T1D (4, 5). Positional clon- genome is not random and often the smallest congenic interval that ing to discover genes causing T1D in NOD mice has involved the can be isolated can contain up to 50 genes (4). Expression analysis generation of congenic mice and this approach has been very suc- of genes within the interval can contribute evidence of genetic variation and thereby greatly assist candidate gene prioritization in a manner unbiased by previous biological knowledge of any of the

*Division of Rheumatology and Immunology, School of Medicine, University of genes. The use of genome-wide microarray expression analysis Pittsburgh, Pittsburgh, PA 15261; †Department of Biostatistics, Graduate School of (15, 16) not only allows simultaneous assessment of all of the Public Health, University of Pittsburgh, Pittsburgh, PA 15261; and ‡Juvenile Diabetes genes within the interval (given microarray chips with sufficient Research Foundation/Wellcome Trust Diabetes and Inflammation Laboratory, De- partment of Medical Genetics, Cambridge Institute for Medical Research, University gene substrate to recognize splice variants), but also highlights of Cambridge, Cambridge, United Kingdom any genes controlled in trans by a candidate causal gene. Received for publication September 13, 2006. Accepted for publication October The combination of microarray analysis with congenic strain 31, 2007. fine-mapping has had variable success in genetic mouse models of The costs of publication of this article were defrayed in part by the payment of page autoimmunity. In two different lupus mouse models, microarray charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. analysis identified strong genetic candidates (17, 18). In T1D, 1 W.M.R. was supported by National Institutes of Health (NIH) National Institute of however, an earlier attempt at this analysis was not successful Diabetes and Digestive and Kidney Diseases 60714 and NIH RFA A102-006. L.S.W. when applied to expression in whole, naive spleen (19). The au- is a Juvenile Diabetes Research Foundation (JDRF)/Wellcome Trust Principal Research Fellow and the research in the laboratory of L.S.W. for this study was also thors concluded that analyzing expression in a noninduced whole supported by NIH P01 AI039671. The availability of NOD congenic mice through the organ was not informative, implying that selecting specific im- Taconic Farms Emerging Models Program has been supported by grants from the mune cell subsets may be more productive. Therefore in this re- Merck Genome Research Institute, National Institute of Allergy and Infectious Dis- eases, and the JDRF. port, we have chosen to analyze differential gene expression in ϩ 2 Address correspondence and reprint requests to Dr. William M. Ridgway, Division purified, activated CD4 T cells because a large amount of liter- of Rheumatology and Immunology, School of Medicine, University of Pittsburgh, ature supports a pathogenic role for this cell type in NOD mice S725 Biomedical Science Tower, 200 Lothrop Street, Pittsburgh, PA 15261. E-mail address: [email protected] (20–22), suggesting that some genes causing T1D, known as Idd (for insulin dependent diabetes) genes, are expressed in the CD4ϩ 3 Abbreviations used in this paper: T1D, type 1 diabetes; Idd, insulin-dependent diabetes; SNP, single nucleotide polymorphism; DAF, decay-accelerating factor; Ct, T cell subset. cycle threshold; ACADL, acyl-coenzyme A dehydrogenase, long chain; SAPE, The NOD.Idd3/5 congenic mouse, with the B6-derived Idd3 in- streptavidin-PE. terval on chromosome 3, and the B10-derived Idd5 interval on Copyright © 2008 by The American Association of Immunologists, Inc. 0022-1767/08/$2.00 chromosome 1, is almost completely protected from diabetes www.jimmunol.org 1072 MICROARRAY ANALYSIS OF NOD CONGENIC CD4ϩ T CELLS

(1–2% incidence at 7 mo of age for NOD.Idd3/5 females compared with 80% for NOD females) (23). The genetic basis of T1D protection from diabetes in NOD.Idd3/5 mice has been partially characterized. The Idd3 region is a 650-kb interval containing five known genes (Tenr, Il2, Il21, Centrin4, and Fgf2), two predicted genes of unknown function (KIAA1109 and KIAA1371), and three pseudogenes. The prime candidate gene encodes IL-2 and the causative single nucleotide polymorphisms are hypothesized to subtly alter its mRNA expression levels (24). As detailed in the accompanying article (25), there are four subregions within the FIGURE 1. Dot plot representation of the whole gene chip data set from ϩ larger Idd5 region: Idd5.1 (2.0 Mb), Idd5.2 (1.5 Mb), Idd5.3 (3.5 NOD-, NOD.Idd3/5-, and B6.G7-activated CD4 T cells. Shown is the Mb), and Idd5.4 (78 Mb). The genes accounting for Idd5.1 and distribution of probe sets representing the entire data set from the NOD, NOD.Idd3/5, and B6.G7 gene chips, analyzed by intersecting probe sets Idd5.2 are most likely Ctla4 (6–8) and Nramp1 (4, 8), respec- from different strains represented on the x and y axes (see Results). Each tively. The molecular basis of these two candidate genes has been circle represents ϳ10 probe sets. Probe sets in the top 7% of differential attributed to single nucleotide polymorphisms in the coding re- expression between NOD- and NOD.Idd3/5-activated CD4ϩ T cells (ver- gions which alter splicing in case of Ctla4 and the primary amino tical line) intersected with probe sets in top 7% of differential expression acid sequence for Nramp1. between NOD- and B6.G7-activated CD4ϩ T cells (horizontal line). The To further characterize the genetic basis of T1D resistance in upper right quadrant shows the probe sets that were in the top 7% in both NOD.Idd3/5 mice, we formulated the hypothesis that the geneti- the NOD vs NOD.Idd3/5 and the NOD vs B6.G7 comparisons.; the full list cally controlled resistance to diabetes would be correlated to al- of these genes is found in supplemental table 1. tered gene expression patterns in candidate genes on chromosomes 1 and 3 of activated NOD.Idd3/5 CD4ϩ T cells. Thus, activated CD4ϩ T cells from NOD and NOD.Idd3/5 should show differ- within the B10-derived introgressed region of chromosome 1 in ential expression primarily in the congenic intervals on chro- NOD.Idd3/5 mice. mosomes 1 and 3; conversely, the common NOD genome outside the congenic region on these two chromosomes should be Preparation, purification, and stimulation of splenocytes equivalently expressed. Exceptions to this hypothesis, i.e., dif- ϩ The spleen from each mouse was removed aseptically and minced. After ferential expression of genes from NOD and NOD.Idd3/5 CD4 lysing RBC, the cells were washed three times with PBS. To purify CD4- T cells outside the congenic intervals, could result from down- positive splenocytes, splenocytes were prepared by magnetic separation stream or trans effects of genes in the congenic intervals. As a using a MiniMACS system (Miltenyi Biotec) according to the manufac- ϩ turer’s instructions. Purified CD4-positive splenocytes were suspended in control, we also compared gene expression between CD4 T RPMI 1640 medium (Invitrogen Life Technologies) supplemented with cells purified from NOD and B6.G7 mice, two strains differing 10% heat-inactivated FBS (Invitrogen Life Technologies) and 1 mM L- throughout the genome except that they share the NOD MHC alanyl-glutamine (Invitrogen Life Technologies), 100 U/ml penicillin, 100 region on chromosome 17. ␮g/ml streptomycin (Invitrogen Life Technologies), 1 mM sodium pyru- ␮ Our results demonstrate the potential of this multifaceted ap- vate (Invitrogen Life Technologies), and 50 M 2-ME. The CD4-positive splenocytes (1 ϫ 106) were transferred to each well of a 24-well plate proach to identifying candidate genes in congenic mice. Of over precoated with anti-CD3 Ab (10 ␮g/ml) and anti-CD28 Ab was added (1 22,000 probe sets analyzed on the chip, 16 of the 20 genes most ␮g/ml final concentration). The cells were cultured for the indicated period differentially expressed between both NOD and NOD.Idd3/5 and and harvested. Th2 conditions were: recombinant mouse IL-4 (10 ng/ml) NOD and B6.G7 CD4ϩ T cells were located within the boundaries and/or anti-mouse IFN-␥ Abs (10 ␮g/ml), whereas Th1 conditions were: recombinant mouse IL-12 (5 ng/ml) and/or anti-mouse IL-4 Abs. In some of the NOD.Idd3/5 congenic intervals on chromosomes 1 and 3. studies, anti-mouse IL-4R Abs alone were added (BD Pharmingen). The two most differentially expressed genes in the Idd5 region, Cd55 (formerly Daf1) and Acadl, have not been implicated in T1D Flow cytometry pathogenesis and are novel candidate genes for Idd5.4 and Idd5.3, After culture, cells were incubated with Fc blocker (BD Pharmingen) and respectively. The identification of Cd55 and Acadl as candidate stained with labeled Abs for 20 min at 4°C. Samples were analyzed on a genes illustrates that an unbiased genetic approach to gene iden- FACSCalibur (BD Biosciences). Anti-CD4 and -CD55 Abs were pur- tification using congenic mouse strains, relevant cell populations, chased from BD Pharmingen. and genome-wide microarray analysis can identify unexpected RNA extraction candidate genes and provide valuable insights into the biological processes underlying T1D. Total RNA was extracted from cultured cells using the RNeasy mini kit (Qiagen). The RNA was redissolved in RNase-free water and yield was estimated by spectrophotometry; equal quantities of RNA were used for Materials and Methods analysis. Samples were hybridized to the Affymetrix mouse chips (see Mice below) at the Genomics and Proteomics Core Laboratory at the University of Pittsburgh. NOD.B6 Idd3 B10 Idd5 mice (Ref. 23, hereafter referred to as NOD.Idd3/5 mice), NOD, and B6.H2g7 (hereafter called B6.G7) mice were bred and Real-time RT-PCR analysis of decay-accelerating factor (DAF) housed under specific pathogen-free conditions and all procedures were mRNA expression conducted according to approved protocols of the University of Pittsburgh School of Medicine Animal Care and Use Committee. All mice used were CD4ϩ T cell RNA was reverse-transcribed using an oligo-dT primer and female, aged 8–11 wk. Idd5 congenic strains 974, 1092, 1595, 2574, 1094, reverse transcription system (Promega) according to the manufacturer’s and 2193 were obtained from the Taconic Emerging Models program (Tac- instructions. Real-time PCR was conducted for CD55 and GAPDH (inter- onic Farms). To confirm the purity of the genetic background of the con- nal control) in an ABI Prism 7300 sequence detector (Applied Biosystems). genic strains, DNA from lines 974, 1092, 1595, and 1094 (from which All reactions were performed using TaqMan Universal MasterMix; primer/ 2574 and 2193 were derived) was tested by genotyping usinga5Kmouse probe sets were purchased from Applied Biosystems (CD55 primer/probe single nucleotide polymorphism (SNP) chip, performed by ParAllele Bio- set ϭ Mm00438377_m1; Applied Biosystems). The obtained mRNA level sciences. For all of the strains tested, no non-NOD SNPs were identified was expressed relative to the GAPDH PCR product amplified from the outside of the defined congenic regions. The 5 K mouse SNP chip was also same sample; DAF value ϭ 2ٙ((Ct of GAPDH) Ϫ (Ct of DAF)), where Ct used to determine that the B10 and B6 strains are identical by descent (26) is the cycle threshold. h ora fImmunology of Journal The

Table I. Gene expression ranking by log odds ratio obtained by intersecting the top 7% of probe set expression between NOD and NOD.Idd3/5 with the top 7% of probe set expression between NOD and B6.G7a

NOD vs Idd3/5 NOD vs Idd3/5 NOD vs Idd3/5 NOD vs B6.G7 NOD vs B6.G7 NOD vs B6.G7 Chromosome, Affymetrix Probe Gene Symbol Odds Maximum p Value Odds Maximum p Value Location Idd Region Set ID Locuslink

Suclg2 0.557 IDD 0.00006 0.472 B6.G7 0.00002 c6, 95.4 1427441_a_at 20917 Entrez gene Cd55 0.521 NOD 0.00076 0.45 NOD 0.00014 c1, 132.3 Idd5.4 1418762_at 13136 Entrez gene Acadl 0.518 IDD 0.00018 0.389 B6.G7 0.00145 c1, 66.8 Idd5.3 1448987_at 11363 Entrez gene Lad1 0.514 IDD 0.00106 0.332 B6.G7 0.01168 c1, 137.6 Idd5.4 1418449_at 16763 Entrez gene Tnni1 0.458 IDD 0.00332 0.326 B6.G7 0.00498 c1, 137.6 Idd5.4 1450813_a_at 21952 Entrez gene Exosc9 0.451 IDD 0.00446 0.321 B6.G7 0.00935 c3, 36.7 b 1418462_at 50911 Entrez gene Ptprv 0.451 NOD 0.00257 0.357 NOD 0.00206 c1, 136.9 Idd5.4 1449957_at 13924 Entrez gene Npl 0.428 NOD 0.00565 0.378 NOD 0.00159 c1, 155.2 Idd5.4 1424265_at 74091 Entrez gene Stk25 0.423 IDD 0.01131 0.422 B6.G7 0.00069 c1, 95.4 Idd5.4 1416770_at 59041 Entrez gene D15Wsu75e 0.419 IDD 0.00386 0.391 B6.G7 0.00099 c15, 81.8 1460689_at 28075 Entrez gene Ramp1 0.383 NOD 0.0169 0.28 NOD 0.01706 c1, 93.0 Idd5.4 1417481_at 51801 Entrez gene Gzmd 0.379 IDD 0.02119 0.41 NOD 0.00058 c14, 55.1 . 1420344_x_at 14941 Entrez gene Ugt1a10 0.334 IDD 0.0026 0.205 B6.G7 0.00444 c1, 89.9 Idd5.4 1424783_a_at 394432 Entrez gene Rnpep 0.33 IDD 0.0255 0.265 B6.G7 0.02056 c1, 137.1 Idd5.4 1451243_at 215615 Entrez gene Hspe1 0.319 NOD 0.03005 0.274 NOD 0.0183 c1, 55.0 b 1422579_at 15528 Entrez gene Glrp1 0.302 IDD 0.02314 0.223 B6.G7 0.0178 c1, 90.3 Idd5.4 1421732_at 14659 Entrez gene Zc3h11a 0.302 NOD 0.00988 0.227 NOD 0.02069 c1, 135.4 Idd5.4 1426359_at 70579 Entrez gene Csrp1 0.284 IDD 0.04475 0.211 B6.G7 0.04412 c1, 137.5 Idd5.4 1425811_a_at 13007 Entrez gene Ralb 0.275 IDD 0.02153 0.359 B6.G7 0.00066 c1, 121.2 Idd5.4 1417744_a_at 64143 Entrez gene Nek6 0.263 NOD 0.02092 0.201 B6.G7 0.02148 c2, 384.0 1423596_at 59126 Entrez gene Plekho1 0.243 IDD 0.02635 0.218 B6.G7 0.01025 c3, 95.7 1417128_at 67220 Entrez gene Elac1 0.242 NOD 0.02755 0.161 B6.G7 0.03613 c18, 73.9 1419228_at 114615 Entrez gene Rhov 0.223 NOD 0.03668 0.395 NOD 0.00009 c2, 119.0 1424976_at 228543 Entrez gene Prdm5 0.221 NOD 0.00453 0.284 NOD 0.00076 c6, 65.7 1432057_a_at 70779 Entrez gene Gna14 0.209 IDD 0.02759 0.162 B6.G7 0.03029 c19, 16.5 1449848_at 14675 Entrez gene Exosc1 0.207 IDD 0.02824 0.293 B6.G7 0.01514 c19, 42.0 1432052_at 66583 Entrez gene Cd99l2 0.198 NOD 0.01179 0.18 NOD 0.01321 X, 68.7 1423965_at 171486 Entrez gene Apobec3 0.175 NOD 0.00727 0.483 B6.G7 Ͻ0.00001 c15, 79.7 1417470_at 80287 Entrez gene Akna 0.172 NOD 0.02641 0.248 B6.G7 0.00981 c4, 63.1 1452393_at 100182 Entrez gene Aak1 0.168 NOD 0.01022 0.195 NOD 0.01955 c6, 86.8 1452632_at 269774 Entrez gene Cd160 0.166 IDD 0.01523 0.177 B6.G7 0.00665 c3, 96.6 1420066_s_at 54215 Entrez gene Pdpn 0.161 IDD 0.03106 0.178 B6.G7 0.02627 c4, 142.9 1419309_at 14726 Entrez gene Itpka 0.16 IDD 0.00006 0.159 B6.G7 0.00515 c2, 119.6 1424037_at 228550 Entrez gene 2010305C02Rik 0.16 IDD 0.01727 0.334 NOD 0.00054 c11, 75.3 1432282_a_at 380712 Entrez gene Smad3 0.157 IDD 0.04242 0.168 B6.G7 0.02718 c9, 63.6 1450472_s_at 17127 Entrez gene Il2rb 0.152 NOD 0.02467 0.381 B6.G7 0.00073 c15, 78.3 1448759_at 16185 Entrez gene Pscd3 0.149 NOD 0.04982 0.264 NOD 0.00618 c5, 144.3 1418758_a_at 19159 Entrez gene Ptdss1 0.135 NOD 0.04265 0.165 NOD 0.01559 c13, 67.0 1441866_s_at 19210 Entrez gene Ccl27 0.131 NOD 0.02967 0.151 B6.G7 0.0221 c4, 41.7 1434962_x_at 20301 Entrez gene Map3k3 0.131 NOD 0.04091 0.153 NOD 0.02625 c11, 106.0 1426686_s_at 26406 Entrez gene Pls3 0.101 IDD 0.01235 0.29 B6.G7 0.00032 X, 73.1 1423725_at 102866 Entrez gene Akr1c12 0.093 NOD 0.01869 0.261 B6.G7 0.00042 c13, 42.8 1422000_at 27383 Entrez gene 0610007P08Rik 0.08 IDD 0.048 0.422 B6.G7 0.00026 c13, 63.9 1453985_at 76251 Entrez gene

a Gene expression is ordered by log odds ratio (from empirical Bayes analysis, see Materials and Methods) for the NOD vs NOD. Idd3/5 comparison. The “Maximum” columns indicate which strain showed the higher expression. Chromosomal location, in megabases, was determined by Ensembl (version 46) query. Genes represented by more than one probe set are shown only once; Cd55, D15Wsu75e, and Akna had two probe sets and in each case the top ranked probe set is shown. b The gene is not within the boundaries of Idd3, Idd5.1, Idd5.2, Idd5.3,orIdd5.4 but is within the boundaries of the introgressed genetic interval of the NOD.3/5 congenic strain. 1073 1074 MICROARRAY ANALYSIS OF NOD CONGENIC CD4ϩ T CELLS

Line Number Markers Mb R8 1092 974 1094 1595 2574 2193 50

D1Mit478 52

D1Mit302 60 D1Mit249 Ctla4 62 Idd5.1

D1Mit303 D1Mit5 64

66 Acadl 68 Idd5.3

74 D1Mit181 Nramp1 76 Idd5.2

D1Mit134 80

130 Cd55 (Daf1) Idd5.4

135

140

D1Mit101 145 FIGURE 3. DAF protein and mRNA expression analyses on NOD, NO- D.Idd3/5, and B6.G7 CD4ϩ T cells. CD4ϩ T cells were purified and stim- D1Mit346 150 Idd5.4 ulated with anti-CD3/CD28 for 3 days and were then analyzed by flow D1Mit267 155 cytometry for cell surface expression of the DAF protein (a) and quanti- 160 tative PCR following reverse transcription for mRNA expression, analyzed with the Mann-Whitney U test (b). FIGURE 2. Genetic maps of NOD Idd5 congenic strains used in this study. Vertical arrows indicate the Idd5 subregions as identified in the accompanying article (25). and stain. The GeneChip was then stained for 10 min in streptavidin-PE (SAPE) solution (1ϫ MES stain buffer, 2 mg/ml acetylated BSA, 10 ␮g/ml Real-time RT-PCR analysis of acyl-coenzyme A dehydrogenase, SAPE; 1ϫ MES stain buffer contains 100 mM MES, 1M [Naϩ], 0.05% long chain (ACADL) mRNA expression Tween 20). Nonstringent buffer was used to wash off the first stain solution. Signal amplification was achieved by 10-min incubation with biotinylated RNA was extracted from purified CD4ϩ T cells in TRIzol (Invitrogen Life anti-streptavidin (1ϫ MES stain buffer, 2 mg/ml acetylated BSA, 0.1 Technologies) and 1000 ng of total RNA was used in a cDNA synthesis mg/ml normal goat IgG, 3 ␮g/ml biotinylated anti-streptavidin) followed reaction with Superscript II reverse transcriptase (Invitrogen Life Technol- by a second 10-min incubation with SAPE. The chip was washed and filled ogies). cDNA was used as template in a TaqMan PCR (prepared with with nonstringent wash buffer before being removed from the fluidics sta- TaqMan Universal PCR Master Mix; Applied Biosystems) with the fol- tion and scanned using the GeneArray scanner. lowing primers and probes designed to detect ACADL mRNA: forward GATTTATCAAGGGCCGGAAG, reverse GAAATCGCCAACTCAG Data analysis CAAT and probe FAM-TGTCCGATTGCCAGCTAATGCC-TAMRA. Activated CD4ϩ ␤ -Microglobulin was used to normalize expression levels as described T cell gene expression data was obtained from NOD (four 2 Idd3/5 previously (6). samples), B6.G7 (four samples), and NOD. (three samples) mice. Preprocessing of the data consisted of robust multichip average back- Microarray techniques ground correction, quantile normalization, and robust multichip average expression summarization as described by Irizarry et al. (27). Preprocess- MOE430A Affymetrix high-density oligonucleotide array chips containing ing was implemented using the affy library of the bioconductor package of 506,944 oligonucleotide probes for 22,690 probe sets were used in the R (28). The advantages of this preprocessing procedure over other meth- analysis. Total RNA was converted to double-stranded cDNA according to ods, for instance, the stock Affymetrix MAS5.0, are described in Bolstad standard methods and purified using an Affymetrix cDNA clean-up col- et al. (29). Correlations, box plots, and variability vs mean plots were used umn. An aliquot of the double-stranded cDNA equivalent to 5–7 ␮gof to verify that the preprocessing successfully reduced the variability of the starting RNA was added as template to an in vitro transcription reaction as expression measures between chips. per the ENZO BioArray High-Efficiency RNA Transcript Labeling kit, and After normalization of the microarray data, we created ranked lists of the the resulting biotinylated cRNA was purified using an Affymetrix RNA most differentially expressed genes—one list for each pair of strains. The clean-up column. After elution, the cRNA was quantified by spectropho- ranked lists were created using an empirical Bayes method (30), which is tometry and 20 ␮g of cRNA was incubated at 94°C in fragmentation buffer effectively a t test that is “smoothed” to remove the erratic effects of genes (40 mM Tris-acetate (pH 8.1), 100 mM KOAc, 30 mM MgOAc) for 35 with unusually low-variance estimates. This is very similar to the method min. A 1-␮l aliquot of the sample was run on an Agilent Bioanalyzer to used in the SAM software (31). After application of the empirical Bayes verify that fragmentation resulted in RNA of the desired size distribution. method to remove the low-variance genes from the ranked lists, we as- Fifteen micrograms of the fragmented RNA was added to a final volume sessed the statistical significance of the most differentially expressed genes of 300 ␮l of hybridization mixture, applied to the GeneChip, and incubated using a standard two-sample t test. For statistical analysis of protein ex- overnight at 45°C with rotation. Following hybridization, the sample was pression, the Mann-Whitney and ␹2 tests were performed in GraphPad removed and the GeneChip cassette was filled with nonstringent wash Prism and JMP-IN software. Physical location of genes was established buffer. The chip was loaded onto an Affymetrix Fluidics station for wash using Ensembl version 46. The Journal of Immunology 1075

FIGURE 4. Chromosome one SNP maps generated from the T1D base web site (http://dil.t1dbase.org/page/ PerlegenSNPs) showing (a) Cd55,(b) the Idd5.3 region, and (c) Acadl. SNP data generated for the NOD strain by Perlegen (http://mouse.perlegen.com/ mouse/index.html) was compared with the B6 sequence present on mouse En- sembl (http://www.ensembl.org/Mus_ musculus/index.html). SNPs that are not polymorphic between NOD and B6 are shown as blue vertical lines and polymorphic SNPs between NOD and B6 are shown as red lines. Yellow vertical lines indicate a known SNP that has not been deﬁned in the NOD strain. Regions lacking SNPs often indicate that comparative sequences were not obtained due to an abundance of repetitive sequence in the region or to other technical issues.

Results NOD vs B6.G7, as shown in Table I. Remarkably, 16 of the 20 Empirical Bayes analysis of normalized microarray expression most differentially expressed genes are located in the NOD.Idd3/5 data sets from NOD, NOD.Idd3/5, and B6.G7 purified, congenic intervals on chromosomes 1 and 3 (Table I, column activated CD4ϩ T cells to identify candidate genes in the eight). Fifteen of 43, or 35%, of all the significantly differentially Idd3/5 congenic regions expressed genes were from chromosome 1; in contrast, all the other chromosomes averaged only 1.45 genes per chromosome. To generate candidate genes in the NOD.Idd3/5 congenic inter- This represents a significant enrichment ( p Ͻ 0.0001) of chromo- vals, we purified CD4ϩ T cells from NOD, NOD.Idd3/5, and some 1 genes in the sample, as assessed by ␹2 analysis comparing B6.G7 spleens, activated them with anti-CD3 and CD28, and sub- the number of genes found in the ϳ90 Mb chromosome 1 segment jected the mRNA to genome-wide analysis with Affymetrix mi- (15/90) vs the number (28 genes) found in the ϳ2571 Mb of chro- croarray chips. Using an empirical Bayes approach (see Materials mosome outside the chromosome 1 segment (golden path length, and Methods), an estimated posterior logarithm of odds of differ- based on Ensembl version 46). Using the Idd5 subregion bound- ential expression for each probe set was generated to create ranked aries as defined by the T1D frequencies of Idd5 congenic strains lists of differentially expressed probe sets. The log odds score was (Fig. 2 and Ref. 25), 14 of the 15 differentially expressed genes in determined for each Affymetrix probe set for ranking expression in the NOD.Idd3/5 chromosome 1 congenic interval are still included CD4ϩ T cells between two pairs of strains: NOD vs NOD.Idd3/5, within the boundaries of two of the four known Idd5 subregions: and NOD vs B6.G7. The top 7% most differentially expressed probe sets were then selected from each list (the 7% cutoff being chosen arbitrarily; 7% represents 1588 probe sets). After removing probe sets that were lacking identifiable genes or chromosomal mean: 8.7 mean: 9.1 locations in Ensembl, there were 208 probe sets that were in the median: 5.5 median: 5.3 top 7% in both the NOD vs NOD.Idd3/5 and the NOD vs B6.G7 974 1092 comparisons. We considered that a subset of these 208 probe sets represented potential candidate genes (Fig. 1; the complete list of probe sets is shown in supplemental table I).4 Fig. 1 represents the expression by log odds of all the probe sets on the chip (each circle represents ϳ10 probe sets); the vertical and horizontal lines rep- mean: 54.9 mean: 64.4 resent the cutoffs for our top 7% lists. The upper right quadrant of median: 23.5 median: 21.6 Fig. 1 shows the probe sets resulting after intersecting the top 7% lists from the NOD vs NOD.Idd3/5 and NOD vs B6.G7 compar- 1595 2574 isons. In addition to using the log odds to rank all the probe sets, we assessed the statistical significance of differential expression between two strains at any single probe set by using the usual t statistic to generate p values, and then eliminated probe sets with DAF Expression a p value Ͼ0.05. This resulted in 43 unique genes, all significantly FIGURE 5. Increased DAF expression is observed only on cells having the differentially expressed between both NOD vs NOD.Idd3/5 and ϩ NOD DAF allele. CD4 T cells were purified from NOD.Idd5 congenic strains and stimulated with anti-CD3/CD28. Cell surface analysis of DAF expression was 4 The online version of this article contains supplemental material. performed as in Fig. 3. One representative of three experiments is shown. 1076 MICROARRAY ANALYSIS OF NOD CONGENIC CD4ϩ T CELLS

ϩ Table II. Differential expression of the known genes in the Idd5.3 CD4 T cells compared with similarly activated cells from interval by microarray log odds rankinga NOD.Idd3/5 and B6.G7 mice (Fig. 3b). The differential expression of the DAF protein and its mRNA is consistent with the B6 Location Log Odds 3/5 Affymetrix and NOD Cd55 alleles having different haplotypes that could Gene Symbol (Mb) vs NOD Rank ID alter gene regulation (Fig. 4a). Mtap2 66.12 0.0105 11,506 1421327_at Having confirmed the gene chip results for Cd55, we examined C030018g12Rik 66.40 NA NA NF additional Idd5 congenic strains of mice with progressively BC042720 66.56 NA NA NF smaller B10-derived chromosome 1 regions, some with NOD al- Rpe 66.63 0.0383 4,602 1416706_at 1110028c15Rik 66.65 0.0145 9,840 1436212_at leles at Cd55 and some with B10 alleles, for DAF protein expres- Acadl 66.76 0.5180 3 1448987_at sion (Fig. 2, genetic map, and Fig. 5, expression studies). In each Myl1 66.86 0.0682 2,023 1452651_at case, CD4ϩ T cells from mice with B10 alleles at Cd55 (lines 1092 Lancl1 66.93 0.1278 415 1427012_at and 974) had decreased DAF expression as compared with CD4ϩ Cps1 67.09 0.0004 20,458 1455540_at T cells from mice with the NOD allele (lines 1595 and 2574) (Fig. Erbb4 67.97 0.0180 8,706 1427783_at Zfpn1a2 69.47 0.0098 11,816 1421643_at 5). These results confirm Cd55 as a candidate gene for Idd5.4, although further dissection of this region by making additional a The rank refers to the rank of expression between NOD and NOD.Idd3/5 in the complete microarray database of 22,690 probe sets. NA, Not applicable; NF, not Idd5.4 congenic strains is required to substantiate this candidacy found in the Affymetrix probe sets. because the B10-derived genetic interval comprising Idd5.4 is large, ϳ70 Mb (Fig. 2). Thus, hundreds of genes are located in Idd5.4 including Cd55 and several other differentially expressed Acadl in Idd5.3 and 14 genes, including Cd55,inIdd5.4 (Table I). genes listed in Table I. Four of the top 20 genes, (including Suclg2, which is the most In contrast to the very large Idd5.4 region, Acadl is located in differentially expressed gene between NOD and both NOD.Idd3/5 the relatively small 3.553 Mb Idd5.3 interval (25), which is highly and B6.G7), and all of the remaining 23 genes in Table I, were polymorphic between NOD and B6 for much of the region (Fig. located outside of the congenic intervals, suggesting that these 4b). Notably, of the Idd5.3 genes interrogated for differential ex- expression differences represent downstream effects of differential pression, only Acadl was highly differentially expressed by log gene expression within the introgressed regions (see Discussion). odds ranking (Table II). The observation that Acadl is differentially expressed is again consistent with B6 and NOD having different Confirming differential expression of candidate genes Cd55 and haplotypes throughout much of the Idd5.3 region (Fig. 4b) includ- Acadl ing Acadl (Fig. 4c). The 43 genes shown in Table I represent ϳ0.2% of all probe sets on We confirmed and extended the gene chip results of increased the gene chip. Nonetheless, in experimental terms, investigating this expression of the B10 ACADL allele by assessing ACADL mRNA number of genes represents an enormous effort. We therefore decided expression by quantitative PCR in activated CD4ϩ T cells from to initially focus our investigation on the most differentially expressed congenic mice having smaller B10-derived intervals in the Idd5 candidate genes in the Idd5.4 and Idd5.3 regions, Cd55 and Acadl. region than the NOD.Idd3/5 congenic strain (Figs. 2 and 6). Line Cd55 encodes the protein DAF, also known as CD55. Cd55 is located 1094 CD4ϩ T cells, which have the B10 ACADL allele, had in- in the distal segment of the B10-derived Idd5 region present in the creased expression of ACADL mRNA compared with cells from NOD.Idd3/5 mouse. First, we evaluated DAF protein expression on line 2193 mice having the NOD allele (Fig. 6). These observations purified CD4ϩ T cells (cultured under the same conditions as those highlight Acadl as a major candidate gene for mediating the dis- used for the gene chip analysis) from NOD, NOD.Idd3/5, and B6.G7 ease-causing phenotype localized to the Idd5.3 region (25). mice. As shown in Fig. 3a, DAF was significantly up-regulated on the ϩ cell surface of NOD CD4ϩ T cells compared with NOD.Idd3/5 or DAF up-regulation on CD4 T cells is promoted by IL-4 and B6.G7 CD4ϩ T cells, thereby confirming the gene chip results at the suppressed in Th1 conditions protein level. Reverse transcription of RNA obtained from the CD4ϩ Our discovery of a variation in DAF expression that localizes to a T cells followed by quantitative PCR (TaqMan methodology) analy- chromosome region that includes Cd55 is of interest because the ses also confirmed the results obtained in the microarray experiments; knockout of the DAF/CD55 gene increases T cell activity and sus- there was increased expression of DAF RNA in activated NOD ceptibility to experimental autoimmune encephalomyelitis (see

A B 5.5 5.5

6 6

6.5 6.5

a Ct

a Ct 7 7

delt

delt 7.5 7.5

8 1094 8 1094 2193 2193 8.5 8.5 0 5 10 15 20 0 5 10 15 20 Time (hours) Time (hours) FIGURE 6. Differential expression of Acadl. Expression of Acadl was assessed in RNA isolated from CD4ϩ T cells stimulated in vitro with anti-CD3/ CD28 for 0, 5, and 20 h. Two strains were compared for Acadl expression, line 1094 (f) and line 2193 (F). Line 1094 has B10 alleles at Idd5.3 whereas line 2193 has NOD alleles at Idd5.3. Data from two experiments of six performed are shown (a and b). All six experiments had similar results. Comparisons of the ⌬ Ct values of Acadl mRNA obtained from CD4 T cells from lines 1094 and 2193 demonstrated signiﬁcant differences at the 5 and 20 h time points (p ϭ 0.08, 0.0003, and 0.005 using the Mann-Whitney U test at 0, 5, and 20 h, respectively). The Journal of Immunology 1077

A FIGURE 7. DAF expression on Neutral Th1 NOD CD4ϩ T cells is up-regulated by both Th2 conditions and IL-4, and down-regulated by both Th1 condi- Mean: 28.5 tions and anti-IL-4R Ab. NOD Mean: 6.6 Median: 8.3 Median: 4.4 CD4ϩ T cells were purified and stimulated with anti-CD3/CD28 as in Fig. 3, under neutral, Th1, and Th2 conditions (see Materials and Meth- ods) or with only IL-4 or anti-IL-4R Ab. A, One representative of seven Th2 IL-4 experiments is shown. B, The mean channel fluorescence from all experiments plus SEM is shown for each Mean: 41.5 condition. The complete data set is Mean: 32.0 Median: 12.0 shown in Table III. Neutral condition Median:9.1 DAF expression was significantly different from DAF expression using Th1, Th2, and anti-IL-4R conditions (p ϭ 0.005, 0.048, and 0.01, respec- DAF α-IL4R tively), and Th1 DAF expression was % of Max significantly different from that un- B der Th2 conditions (p ϭ 0.008). Th1 condition DAF expression was not 250 *p = 0.048 *p = 0.02 significantly different from that using Mean: 7.4 Median: 4.6 the anti-IL4R condition nor was Th2 200 condition DAF expression significantly different from that using the 150

IL-4 condition. Cells cultured under *p = 0.005 IL-4 conditions showed signiﬁcantly DAF 100 different expression than Th1 (p ϭ

0.009) and anti-IL-4R conditions Mean channel fluorescence 50 (p ϭ 0.02). All signiﬁcance testing was performed using the Mann- 0 Whitney U test. Neutral Th1 Th2 IL-4 anti-IL4R

Culture conditions

Discussion). We thus hypothesized that the differential expression DAF was significantly different between IL-4 alone vs Th1 con- of DAF by the B10 and NOD alleles would have functional conditions ( p ϭ 0.009) and IL-4 alone vs anti-IL-4R Ab ( p ϭ sequences on the immune response, and initiated studies to inves- 0.02), while expression under Th2 conditions did not differ sta- tigate the regulation of DAF expressed at the cell surface of CD4ϩ tistically from IL-4 alone ( p ϭ 0.66) (Fig. 7, Table III). More- T cells in different cytokine environments (Fig. 7 and Table III). over, anti-IL-4R Abs alone added to culture with anti-CD3/- As noted above, NOD CD4ϩ T cells up-regulate DAF under neu- CD28 stimulation completely prevented DAF up-regulation; the tral conditions. However, as shown in Fig. 7, Th1 conditions were expression levels of DAF with anti-IL-4R Ab were significantly not associated with DAF up-regulation on NOD CD4ϩ T cells different from neutral conditions ( p ϭ 0.01), whereas expres- whereas Th2 conditions strongly enhanced DAF up-regulation. sion under Th1 conditions was not significantly different from Th2 conditions did not, however, increase DAF expression on anti-IL-4R Ab alone. NOD.Idd3/5 CD4ϩ T cells (data not shown). Next, we asked which components of the Th2 culture conditions were sufficient Discussion for up-regulation of DAF. IL-4 alone was sufficient for strong Identification of candidate genes in complex autoimmune diseases, ϩ up-regulation of DAF on NOD CD4 T cells. Up-regulation of even when they are confined to a small, well-defined genetic interval, remains a challenging scientific problem. Hypothesis-driven ap- proaches (e.g., investigating immunological mediators in T1D, or in- Table III. Expression of CD55 on activated NOD CD4ϩ T cellsa vestigating possible autoantigens) will overlook genes in novel and unexpected pathways whereas a “hypothesis-free” whole genome ap- Exp Neutral Th1 Th2 IL-4 Anti-IL-4R proach can miss candidate genes by investigating too many variables 1 111.9 38.3 238 nd nd simultaneously. Indeed, a prior published attempt to discover caus- 2 39.5 18.3 155.8 92.7 48.6 ative genes in NOD congenic mice analyzed expression in whole, 3 28.5 6.6 41.5 32.0 7.4 unstimulated spleen preparations and did not detect compelling can- 4 71.6 22.8 129.6 134.7 22.9 5 76.3 23.6 253.6 116.5 22.4 didate genes in the Idd intervals (19). In the current study, we have 6 118.5 nd nd 296.2 38.1 approached the search for candidate loci by focusing on global ex- 7 89.9 nd nd 196.6 32.3 pression analysis of activated CD4ϩ T cells from both NOD and a nd, Not done. All experiments are after 3 days of culture of NOD CD4ϩ T cells NOD congenic mice to define a subset of differentially expressed under the stated conditions. See Fig. 7 and text for p values. genes located in the introgressed congenic intervals. 1078 MICROARRAY ANALYSIS OF NOD CONGENIC CD4ϩ T CELLS

As summarized in Table I, this approach was highly successful; 16 expression renders T cells more responsive, as suggested by the of the top 20 differentially expressed genes identified were localized knockout studies, the higher CD55 levels in the presence of IL-4 to introgressed non-NOD DNA present in the congenic mice, and 14 could prevent the T cells from responding as efficiently to other IL- of the 16 genes were located within the boundaries of Idd5.3 or Idd5.4 4-mediated differentiation signals, thereby causing the T cells to re- on chromosome 1 and therefore represent candidate genes. Two of the main more Th1-like. We have previously shown that NOD T cells are most differentially expressed genes, Daf1 and Acadl, were confirmed biased to Th1 expression (43, 44), consistent with many other publi- using quantitative PCR and are located within the boundaries of the cations suggesting a Th1 bias in NOD mice (45–50). Idd5.4 and Idd5.3 regions, respectively, thereby establishing them as It is also possible that CD55 expression caused by IL-4 is anti- strong candidate genes. inflammatory and that the genetic program generating Th1 conditions Several of the candidate genes in Table I were not located in the in NOD mice in vivo negates the potentially protective effect of DAF introgressed regions on chromosomes 1 and 3 present in NOD.Idd3/5 by preventing its up-regulation because of a paucity of IL-4. Indeed, mice. Suclg2, for example, emerges as the most differentially many therapeutic interventions associated with the induction of Th2- expressed gene overall, but it is located on chromosome 6. A related phenotypes have prevented T1D (51–53). Our genetic map- reasonable assumption is that the differential expression of Suclg2 ping studies indicate that the NOD allele of Idd5.4 acts as a T1D- is a downstream effect of one or more of the non-NOD-derived resistance allele whereas the B10 allele increases T1D susceptibility genes in the introgressed Idd3/5 intervals. The identification of (25); however, it is likely that the ϳ70 Mb Idd5.4 interval will have Acadl as a candidate gene highlights its potential to mediate such more than one gene affecting T1D, a hypothesis we are currently downstream effects. Acadl encodes an enzyme, acyl-coenzyme A testing. In particular, a NOD congenic strain having a small interval dehydrogenase, which controls the first step in fatty acid ␤-oxida- encompassing the B10 allele of Cd55 will be tested to determine tion (32). Studies from the Thompson laboratory (33, 34) have whether this isolated region contributes resistance or susceptibility highlighted the unique bioenergetic challenges facing T cells upon to T1D. activation: they require enormous energy expenditures, but also face significant anabolic challenges in synthesizing sufficient ma- Acknowledgments terials for cell division, including fatty acids needed for new cell We acknowledge Deborah Hollingshead and the University of Pittsburgh membrane construction. The outcome of these energetic demands Genomics and Proteomics Core Laboratories for their assistance with is that T cell activation directly suppresses fatty acid ␤-oxidation microarrays. via the PI3K/Akt pathway (33, 34). We have shown that the protective B10 allele of Acadl has higher expression compared with Disclosures the NOD allele, presumably leading to increased fatty acid use; The authors have no financial conflict of interest. moreover, Suclg2 mRNA is also higher in Idd3/5 compared with ϩ NOD-activated CD4 T cells. Suclg2 encodes succinyl coenzyme References A ligase, which is a component of the citric acid cycle. Higher 1. The Wellcome Trust Case Control Consortium. 2007. Genome-wide association levels of two enzymes involved in energy production, Acadl and study of 14,000 cases of seven common diseases and 3,000 shared controls. Suclg2 in T cells with a B10-derived protective Idd5.3 region, Nature 447: 661–678. 2. Easton, D. F., K. A. Pooley, A. M. Dunning, P. D. Pharoah, D. Thompson, suggest a change in T cell bioenergetics that could alter the func- D. G. Ballinger, J. P. Struewing, J. Morrison, H. Field, R. Luben, et al. 2007. tion and survival of T cells. The highlighting of such downstream Genome-wide association study identifies novel breast cancer susceptibility loci. expression differences by microarray analysis should provide Nature 447: 1087–1093. 3. Todd, J. A., N. M. Walker, J. D. Cooper, D. J. Smyth, K. Downes, V. Plagnol, significant insights into disease pathogenesis because it is possible R. Bailey, S. Nejentsev, S. F. Field, F. Payne, et al. 2007. Robust associations of that protective alleles at distinct Idd genes would cause partially four new chromosome regions from genome-wide analyses of type 1 diabetes. overlapping downstream patterns. Because the combined activity Nat. Genet. 39: 857–864. 4. Kissler, S., P. Stern, K. Takahashi, K. Hunter, L. B. Peterson, and L. S. Wicker. of protective alleles at Idd3 and Idd5 provide more disease 2006. In vivo RNA interference demonstrates a role for Nramp1 in modifying protection than would be expected if the alleles were acting in a susceptibility to type 1 diabetes. Nat. Genet. 38: 479–483. multiplicative fashion (35), the discovery of gene expression dif- 5. Wicker, L. S., J. A. Todd, and L. B. Peterson. 1995. Genetic control of autoimmune diabetes in the NOD mouse. Annu. Rev. Immunol. 13: 179–200. ferences that occur only when a particular combination of protec- 6. Ueda, H., J. M. Howson, L. Esposito, J. Heward, H. Snook, G. Chamberlain, tive alleles is present would also illuminate key pathways causing D. B. Rainbow, K. M. Hunter, A. N. Smith, G. Di Genova, et al. 2003. Associ- ation of the T-cell regulatory gene CTLA4 with susceptibility to autoimmune autoimmunity. disease. Nature 423: 506–511. We identified Cd55 as one of the most differentially expressed 7. Vijayakrishnan, L., J. M. Slavik, Z. Illes, R. J. Greenwald, D. Rainbow, B. Greve, genes between NOD and NOD.Idd3/5 T cells. Cd55 is well-known L. B. Peterson, D. A. Hafler, G. J. Freeman, A. H. Sharpe, et al. 2004. An autoimmune disease-associated CTLA-4 splice variant lacking the B7 binding as a complement regulatory gene, but recent research has also domain signals negatively in T cells. Immunity 20: 563–575. identified it as a T cell costimulatory molecule that interacts with 8. Wicker, L. S., G. Chamberlain, K. Hunter, D. Rainbow, S. Howlett, P. Tiffen, its ligand, CD97 (36). Splice variants of CD97 produce isoforms J. Clark, A. Gonzalez-Munoz, A. M. Cumiskey, R. L. Rosa, et al. 2004. Fine mapping, gene content, comparative sequencing, and expression analyses support having variable numbers of epidermal growth factor domains that Ctla4 and Nramp1 as candidates for Idd5.1 and Idd5.2 in the nonobese diabetic engage DAF with variable affinity (37). DAF likely binds to the first mouse. J. Immunol. 173: 164–173. CD97 epidermal growth factor domain but also requires domains 2 9. Lyons, P. A., N. Armitage, F. Argentina, P. Denny, N. J. Hill, C. J. Lord, M. B. Wilusz, L. B. Peterson, L. S. Wicker, and J. A. Todd. 2000. Congenic and 5; the alternately spliced variant of CD97 that expresses only mapping of the type 1 diabetes locus, Idd3, to a 780-kb region of mouse domains 1, 2, and 5 binds DAF with the highest affinity (38). chromosome 3: identification of a candidate segment of ancestral DNA by hap- Cd55 knockout mice have been studied in several models of auto- lotype mapping. Genome Res. 10: 446–453. 10. Lyons, P. A., W. W. Hancock, P. Denny, C. J. Lord, N. J. Hill, N. Armitage, immunity and show worsened experimental glomerulonephritis, ex- T. Siegmund, J. A. Todd, M. S. Phillips, J. F. Hess, et al. 2000. The NOD Idd9 perimental myasthenia gravis, and experimental autoimmune enceph- genetic interval influences the pathogenicity of insulitis and contains molecular Ϫ/Ϫ variants of Cd30, Tnfr2, and Cd137. Immunity 13: 107–115. alomyelitis (39–42). Moreover, Cd55 mice demonstrated 11. Olofsson, P., J. Holmberg, J. Tordsson, S. Lu, B. Akerstrom, and R. Holmdahl. significantly enhanced T cell responses with hypersecretion of IFN-␥ 2003. Positional identification of Ncf1 as a gene that regulates arthritis severity in (42). However, it is unclear whether proinflammatory or anti- rats. Nat. Genet. 33: 25–32. 12. Todd, J. A., T. J. Aitman, R. J. Cornall, S. Ghosh, J. R. Hall, C. M. Hearne, inflammatory biological effects will be caused in vivo by high vs low A. M. Knight, J. M. Love, M. A. McAleer, J. B. Prins, et al. 1991. Genetic CD55 expression levels in the presence of IL-4. If the lack of CD55 analysis of autoimmune type 1 diabetes mellitus in mice. Nature 351: 542–547. The Journal of Immunology 1079

13. Wicker, L. S., J. A. Todd, J. B. Prins, P. L. Podolin, R. J. Renjilian, and 34. Deberardinis, R. J., J. J. Lum, and C. B. Thompson. 2006. Phosphatidylinositol L. B.Peterson. 1994. Resistance alleles at two non-major histocompatibility com- 3-kinase-dependent modulation of carnitine palmitoyltransferase 1A expression plex-linked insulin-dependent diabetes loci on chromosome 3, Idd3 and Idd10, regulates lipid metabolism during hematopoietic cell growth. J. Biol. Chem. 281: protect nonobese diabetic mice from diabetes. J. Exp. Med. 180: 1705–1713. 37372–37380. 14. Cervino, A. C., M. Gosink, M. Fallahi, B. Pascal, C. Mader, and 35. Cordell, H. J., J. A. Todd, N. J. Hill, C. J. Lord, P. A. Lyons, L. B. Peterson, N. F. Tsinoremas. 2006. A comprehensive mouse IBD database for the efficient L. S. Wicker, and D. G. Clayton. 2001. Statistical modeling of interlocus inter- localization of quantitative trait loci. Mamm. Genome 17: 565–574. actions in a complex disease: rejection of the multiplicative model of epistasis in 15. Pease, A. C., D. Solas, E. J. Sullivan, M. T. Cronin, C. P. Holmes, and type 1 diabetes. Genetics 158: 357–367. S. P. Fodor. 1994. Light-generated oligonucleotide arrays for rapid DNA se- 36. Hamann, J., B. Vogel, G. M. van Schijndel, and R. A. van Lier. 1996. The seven- quence analysis. Proc. Natl. Acad. Sci. USA 91: 5022–5026. span transmembrane receptor CD97 has a cellular ligand (CD55, DAF). J. Exp. 16. Schena, M., D. Shalon, R. W. Davis, and P. O. Brown. 1995. Quantitative mon- Med. 184: 1185–1189. itoring of gene expression patterns with a complementary DNA microarray. Sci- 37. Gray, J. X., M. Haino, M. J. Roth, J. E. Maguire, P. N. Jensen, A. Yarme, ence 270: 467–470. M. A. Stetler-Stevenson, U. Siebenlist, and K. Kelly. 1996. CD97 is a processed, 17. Liu, K., C. Liang, Z. Liang, K. Tus, and E. K. Wakeland. 2005. Sle1ab mediates seven-transmembrane, heterodimeric receptor associated with inflammation. the aberrant activation of STAT3 and Ras-ERK signaling pathways in B lym- J. Immunol. 157: 5438–5447. phocytes. J. Immunol. 174: 1630–1637. 38. Hamann, J., C. Stortelers, E. Kiss-Toth, B. Vogel, W. Eichler, and R. A. van Lier. 18. Rozzo, S. J., J. D. Allard, D. Choubey, T. J. Vyse, S. Izui, G. Peltz, and 1998. Characterization of the CD55 (DAF)-binding site on the seven-span trans- B. L. Kotzin. 2001. Evidence for an interferon-inducible gene, Ifi202,inthe membrane receptor CD97. Eur. J. Immunol. 28: 1701–1707. susceptibility to systemic lupus. Immunity 15: 435–443. 39. Sogabe, H., M. Nangaku, Y. Ishibashi, T. Wada, T. Fujita, X. Sun, T. Miwa, 19. Eaves, I. A., L. S. Wicker, G. Ghandour, P. A. Lyons, L. B. Peterson, J. A. Todd, M. P. Madaio, and W. C. Song. 2001. Increased susceptibility of decay-acceler- and R. J. Glynne. 2002. Combining mouse congenic strains and microarray gene ating factor deficient mice to anti-glomerular basement membrane glomerulone- expression analyses to study a complex trait: the NOD model of type 1 diabetes. phritis. J. Immunol. 167: 2791–2797. Genome Res. 12: 232–243. 40. Lin, F., S. N. Emancipator, D. J. Salant, and M. E. Medof. 2002. Decay-accel- 20. Wong, F. S., and C. A. Janeway, Jr. 1997. The role of CD4 and CD8 T cells in erating factor confers protection against complement-mediated podocyte injury in type I diabetes in the NOD mouse. Res. Immunol. 148: 327–332. acute nephrotoxic nephritis. Lab. Invest. 82: 563–569. 21. Suri, A., and J. D. Katz. 1999. Dissecting the role of CD4ϩ T cells in autoimmune 41. Lin, F., H. J. Kaminski, B. M. Conti-Fine, W. Wang, C. Richmonds, and diabetes through the use of TCR transgenic mice. Immunol. Rev. 169: 55–65. M. E. Medof. 2002. Markedly enhanced susceptibility to experimental autoimmune myasthenia gravis in the absence of decay-accelerating factor protection. 22. Anderson, M. S., and J. A. Bluestone. 2005. The NOD mouse: a model of im- J. Clin. Invest. 110: 1269–1274. mune dysregulation. Annu. Rev. Immunol. 23: 447–485. 42. Liu, J., T. Miwa, B. Hilliard, Y. Chen, J. D. Lambris, A. D. Wells, and 23. Hill, N. J., P. A. Lyons, N. Armitage, J. A. Todd, L. S. Wicker, and W. C. Song. 2005. The complement inhibitory protein DAF (CD55) suppresses L. B. Peterson. 2000. NOD Idd5 locus controls insulitis and diabetes and overlaps T cell immunity in vivo. J. Exp. Med. 201: 567–577. the orthologous CTLA4/IDDM12 and NRAMP1 loci in humans. Diabetes 49: 43. Koarada, S., Y. Wu, and W. M. Ridgway. 2001. Increased entry into the IFN-␥ 1744–1747. ϩ effector pathway by CD4 T cells selected by I-Ag7 on a nonobese diabetic 24. Yamanouchi, J., D. Rainbow, P. Serra, S. Howlett, K. Hunter, V. E. Garner, versus C57BL/6 genetic background. J. Immunol. 167: 1693–1702. A. Gonzalez-Munoz, J. Clark, R. Veijola, R. Cubbon, et al. 2007. Interleukin-2 44. Koarada, S., Y. Wu, G. Olshansky, and W. M. Ridgway. 2002. Increased nono- gene variation impairs regulatory T cell function and causes autoimmunity. Nat. ϩ bese diabetic Th1: Th2 (IFN-␥: IL-4) ratio is CD4 T cell intrinsic and inde- Genet. 39: 329–337. pendent of APC genetic background. J. Immunol. 169: 6580–6587. 25. Hunter, K., D. Rainbow, V. Plagnol, J. A. Todd, L. B. Peterson, and L. S. Wicker. 45. Katz, J. D., C. Benoist, and D. Mathis. 1995. T helper cell subsets in insulin- 2007. Interactions between Idd5.1/Ctla4 and other type 1 diabetes genes. dependent diabetes. Science 268: 1185–1188. J. Immunol. 179: 8341–8349. 46. Trembleau, S., G. Penna, E. Bosi, A. Mortara, M. K. Gately, and L. Adorini. 26. Ridgway, W. M., B. Healy, L. J. Smink, D. Rainbow, and L. S. Wicker. 2007. 1995. Interleukin 12 administration induces T helper type 1 cells and accelerates New tools for defining the “genetic background” of inbred mouse strains. Nat. autoimmune diabetes in NOD mice. J. Exp. Med. 181: 817–821. Immunol. 8: 669–673. 47. Trembleau, S., G. Penna, S. Gregori, M. K. Gately, and L. Adorini. 1997. De- 27. Irizarry, R. A., B. Hobbs, F. Collin, Y. D. Beazer-Barclay, K. J. Antonellis, viation of pancreas-infiltrating cells to Th2 by interleukin-12 antagonist admin- U. Scherf, and T. P. Speed. 2003. Exploration, normalization, and summaries of istration inhibits autoimmune diabetes. Eur. J. Immunol. 27: 2330–2339. high density oligonucleotide array probe level data. Biostatistics 4: 249–264. 48. Trembleau, S., G. Penna, S. Gregori, H. D. Chapman, D. V. Serreze, J. Magram, 28. Gautier, L., L. Cope, B. M. Bolstad, and R. A. Irizarry. 2004. affy–analysis of and L. Adorini. 1999. Pancreas-infiltrating Th1 cells and diabetes develop in Affymetrix GeneChip data at the probe level. Bioinformatics 20: 307–315. IL-12-deficient nonobese diabetic mice. J. Immunol. 163: 2960–2968. 29. Bolstad, B. M., R. A. Irizarry, M. Astrand, and T. P. Speed. 2003. A comparison 49. Rabinovitch, A. 1994. Immunoregulatory and cytokine imbalances in the patho- of normalization methods for high density oligonucleotide array data based on genesis of IDDM: therapeutic intervention by immunostimulation? Diabetes 43: variance and bias. Bioinformatics 19: 185–193. 613–621. 30. Lin, Y., P. Reynolds, and E. Feingold. 2003. An empirical Bayesian method for 50. Rabinovitch, A., O. Sorensen, W. L. Suarez-Pinzon, R. F. Power, R. V. Rajotte, differential expression studies using one-channel microarray data. In Statistical and R. C. Bleackley. 1994. Analysis of cytokine mRNA expression in syngeneic Applications in Genetics and Molecular Biology, Vol. 2, No. 1, Article 8, pp. islet grafts of NOD mice: interleukin 2 and interferon ␥ mRNA expression cor- 1–18. relate with graft rejection and interleukin 10 with graft survival. Diabetologia 37: 31. Tusher, V. G., R. Tibshirani, and G. Chu. 2001. Significance analysis of microar- 833–837. rays applied to the ionizing radiation response. Proc. Natl. Acad. Sci. USA 98: 51. Rabinovitch, A., W. L. Suarez-Pinzon, O. Sorensen, R. C. Bleackley, 5116–5121. R. F. Power, and R. V. Rajotte. 1995. Combined therapy with interleukin-4 and 32. Matsubara, Y., Y. Indo, E. Naito, H. Ozasa, R. Glassberg, J. Vockley, Y. Ikeda, interleukin-10 inhibits autoimmune diabetes recurrence in syngeneic islet-trans- J. Kraus, and K. Tanaka. 1989. Molecular cloning and nucleotide sequence of planted nonobese diabetic mice: analysis of cytokine mRNA expression in the cDNAs encoding the precursors of rat long chain acyl-coenzyme A, short chain graft. Transplantation 60: 368–374. acyl-coenzyme A, and isovaleryl-coenzyme A dehydrogenases: sequence homol- 52. Fox, C. J., and J. S. Danska. 1997. IL-4 expression at the onset of islet inflam- ogy of four enzymes of the acyl-CoA dehydrogenase family. J. Biol. Chem. 264: mation predicts nondestructive insulitis in nonobese diabetic mice. J. Immunol. 16321–16331. 158: 2414–2424. 33. Fox, C. J., P. S. Hammerman, and C. B. Thompson. 2005. Fuel feeds function: 53. Adorini, L., and S. Trembleau. 1997. Immune deviation towards Th2 inhibits energy metabolism and the T-cell response. Nat. Rev. Immunol. 5: 844–852. Th-1-mediated autoimmune diabetes. Biochem. Soc. Trans. 25: 625–629.