Dynamic interaction between T cell-mediated ß-cell damage and ß-cell repair in the run up to autoimmune diabetes of the NOD mouse

Sankaranand S. Vukkadapu1, Jenine M. Belli1, Koji Ishii1, Anil G. Jegga2, John J. Hutton2, Bruce J. Aronow2,3 and Jonathan D. Katz1,4

1 Diabetes Research Center, Division of Endocrinology
2 Division of Pediatric Informatics
3 Division of Developmental Biology
4 Divison of Molecular Immunology, Cincinnati Children’s Hospital Research Foundation and College of Medicine, University of Cincinnati, Cincinnati, Ohio


    ABSTRACT
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 GRANTS
 REFERENCES
 
In type 1 diabetes mellitus (T1DM), also known as autoimmune diabetes, the pathogenic destruction of the insulin-producing pancreatic ß-cells is under the control of and influenced by distinct subsets of T lymphocytes. To identify the critical genes expressed by autoimmune T cells, antigen presenting cells, and pancreatic ß-cells during the evolution of T1DM in the nonobese diabetic (NOD) mouse, and the genetically-altered NOD mouse (BDC/N), we used functional genomics. Microarray analysis revealed increased transcripts of genes encoding inflammatory cytokines, particularly interleukin (IL)-17, and islet cell regenerating genes, Reg3{alpha}, Reg3ß, and Reg3{gamma}. Our data indicate that progression to insulitis was connected to marked changes in islet antigen expression, ß-cell differentiation, and T cell activation and signaling, all associated with tumor necrosis factor-{alpha} and IL-6 expression. Overt diabetes saw a clear shift in cytokine, chemokine, and T cell differentiation factor expression, consistent with a focused Th1 response, as well as a significant upregulation in genes associated with cellular adhesion, homing, and apoptosis. Importantly, the temporal pattern of expression of key verified genes suggested that T1DM develops in a relapsing/remitting as opposed to a continuous fashion, with insulitis linked to hypoxia-regulated gene control and diabetes with C/EBP and Nkx2 gene control.

type 1 diabetes; nonobese diabetic mouse; pancreas; microarray; gene expression profile


    INTRODUCTION
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 GRANTS
 REFERENCES
 
TYPE 1 DIABETES MELLITUS (T1DM) is an autoimmune disease resulting in destruction of the insulin-producing pancreatic ß-cells. The nonobese diabetic (NOD) mouse is an excellent model for T1DM (40). Current data from the NOD mouse suggest that T1DM, like other organ-specific autoimmune diseases, results from the interplay of specific predisposing genetic elements with poorly defined environmental risk factors. Like human T1DM patients, NOD mice develop a preclinical condition, termed insulitis (checkpoint 1), during which leukocytes surround and then infiltrate the pancreatic islets of Langerhans. Several studies have suggested that the insulitis undergoes a qualitative change from a relatively benign infiltration to a pernicious insulitis associated with ß-cell apoptosis that leads to hyperglycemia (checkpoint 2) (4, 33, 45). The underlying genetic controls governing each of these processes are complex and only partially understood (54). Selective breeding studies using the NOD mouse have mapped nearly 20 genetic regions and subregions that alter the tempo and severity of insulitis and/or the transition to diabetes, yet most of these crucial genetic elements have not been fully described. Even where gene products are known, for example the unique major histocompatibility complex (MHC) class II gene products, human HLA-DQ2 and -DQ8 and mouse I-Ag7, it is not clear how they act to precipitate T1DM (41, 52).

However, the close association between T1DM and MHC class II has inspired intense investigation into the role that MHC class II-restricted CD4+ T cells play in T1DM. It is now well established that CD4+ T cells play a prominent role, as 1) CD4+ T cells from NOD mice respond naturally to pancreatic ß-cell antigens (53), 2) bulk and cloned CD4+ T cells from diabetic NOD mice can transfer T1DM to young NOD and NOD.scid recipients (21, 31), 3) NOD mice lacking CD4+ T cells fail to develop T1DM (32, 47), and 4) T cell receptor (TCR) transgenic NOD mice harboring CD4+ T cells with islet antigen reactivity develop insulitis and diabetes (30). We and others have used one of these TCR transgenic mice, BDC2.5, to establish the presence of discrete checkpoints along the way to overt diabetes (4, 30).

To understand the complex autoimmune processes affecting the pathogenesis of T1DM, it is absolutely essential to identify, catalog, and investigate the major modulations in gene expression patterns that occur during the course of disease pathogenesis. In light of the rapid advances in genome-based research, it is now possible to undertake a comprehensive molecular study of complex diseases such as T1DM using cDNA microarray analysis. Complementary DNA microarrays provide us with a powerful tool for the simultaneous imaging of the expression of large numbers of genes (36) and are used to delineate the progression of cancer (12), type 2 diabetes (42), asthma (29), and other complex or multiorgan diseases. In the present study, we report the application of cDNA microarray technology to T1DM and the identification of subsets of genes the expression of which changes markedly during the initiation of insulitis, as well as others that are modulated during the progression to ß-cell destruction and overt diabetes in the NOD mouse.


    MATERIALS AND METHODS
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 GRANTS
 REFERENCES
 
Animals.
NOD.scid mice (N.sc), NOD mice (N), BDC2.5 TCR transgenic mice on the NOD genetic background (BDC/N), and BDC2.5 TCR transgenic mice on the NOD.scid genetic background (BDC/N.sc) were used. BDC2.5 mice express the BDC2.5 TCR as described (32). All mice were maintained under specific pathogen-free conditions in our Association for Assessment and Accreditation of Laboratory Animal Care-approved animal facility under Institutional Animal Care and Use Committee approval.

Histology and immunohistochemistry of pancreas.
Pancreata from N.sc, N, BDC/N, and BDC/N.sc were removed and embedded in paraffin, sectioned, and stained with hematoxylin and eosin for the identification of mononuclear cells or stained with guinea pig anti-insulin antibody (Dako) followed by Alexa-568-conjugated goat anti-guinea pig IgG (H+L) antibody (Molecular Probes) as described in Ref. 33. 2-(2-Nitro-1H-imidazol-1-yl)-N-(2,2,3,3,3-pentafluoropropyl)acetamide (EF-5) was detected using ELK3-51, a monoclonal antibody conjugated to Alexa-488 (Dr. Cameron J. Koch, Dept. of Radiation Oncology, Univ. of Pennsylvania, Philadelphia, PA).

Imaging procedure.
Stained sections were imaged with a Zeiss Axioplan 2.0 fluorescence microscope (Zeiss) with a 100-W mercury lamp illumination, and image was analyzed with Axio Vision software (v.4.0, Zeiss).

Interleukin-17 determination.
Interleukin (IL)-17 was quantified, using the IL-17 Quantikine kit (R&D Systems).

EF-5 treatment of mice for the detection of hypoxia.
BDC/N and BDC/N.sc mice were given an intraperitoneal injection of 10 mM EF-5 (obtained from Dr. R. Vishnuvajjala, National Cancer Institute, Bethesda, MD), prepared in 0.9% saline. The mass of solution administered was 1% mouse’s mass; the equivalent whole body concentration was 100 µM.

Diabetes.
Diabetes was assessed by measurement of venous blood as described previously (33).

RNA isolation.
Total RNA was isolated from pancreas (3-wk-old mice, unless otherwise noted) by the guanidinium isothiocyanate method (10). RNA was isolated from 12 individual mice in each group, and 2 RNA pools were made from 6 animals each after verification of RNA quality to minimize expression biases. The integrity of isolated RNA was assessed by formaldehyde-agarose gel electrophoresis and Agilent bioanalyzer chips.

DNA microarrays and analysis of GeneChip data.
Analysis of mRNA expression was performed in duplicate, using the mouse U74Av2 microarray (GeneChip, Affymetrix, Santa Clara, CA). RNA labeling and microarray hybridizations were performed according to the manufacturer’s recommendations by the Cincinnati Children’s Hospital Research Foundation (CCHRF) genomics core. All the experiments were done in duplicate.

Data analysis.
Affymetrix MicroArray Suite version 5.0 was used to scan and quantify the GeneChips, using default scan settings. Intensity data were collected from each chip, scaled to a target intensity of 1,500. For subsequent analysis of gene expression, we used Robust MultiArray (RMA) (23). RMA files were made using RMA Express software (v.0.1, biolstat@stat.berkeley.edu) and analyzed with GeneSpring 6.0 software (Silicon Genetics, Redwood City, CA). All experiments were carried out in duplicate to maximize the statistical significance. Hybridization data were normalized in a two-step process to minimize variation at both chip and gene level. Initially, each chip data point was normalized to the distribution of all genes on the chip to control for variation between samples and then normalized to the specific sample, N.sc. Normalized data were filtered by data file restriction, expression percentage ratio, and statistical significance and clustered (based on expression profile or pattern) using K-means and hierarchical clustering. Subsets of genes such as chemokines, cytokines, cell cycle control, antigen processing, and so forth were made from the filtered gene lists, based on imported pathway schematics designed by our laboratory from published data as well as the simplified gene ontology at CCHRF. Raw and RMA experimental data for N.sc, N, BDC/N, and BDC/N.sc were deposited in the Gene Expression Omnibus database (http://www.ncbi.nlm.nih.gov/geo/) under the accession numbers GSM27453, GSM27454, GSM27456, GSM27457, GSM27446, GSM27447, GSM27451, GSM27452, and GSE1623, respectively.

Verification of gene expression data.
To verify the fidelity of microarray analysis, we selected, for further analysis, representative genes (~20) with a broad range of expression variations from our initial list of genes associated with insulitis and diabetes. Fresh total pancreatic RNA from BDC/N.sc, BDC/N, N, and N.sc was prepared, reverse transcribed, and subjected to semiquantitative PCR using an Eppendorf thermocycler, followed by real-time PCR on a Roche LightCycler with SYBR Green (Roche Diagnostics), according to the manufacturer’s recommendations. Expression of target genes was normalized to ß-actin. In the case of real-time PCR, the relationship between cycle threshold (CT) and gene copy number was determined by a regression equation (log copy no. = –0.283 CT + 11.309) as described in Ref. 13. Dynamic gene expression (kinetics analysis) was performed, as described above, from freshly isolated RNA samples from mice at 1, 2, 3, 4, 6, 8, 12, 16, and 20 wk of age.

Pathway and literature search analysis.
Biological function and associated regulatory pathway search analysis for the selected genes was performed using a mouse U74Av2 annotation database with system identifiers. Gene description, functional categories, and molecular function were identified using databases such as NetAffy (http://www.affymetrix.com) and BLAST (http://www.ncbi.nlm.nih.gov).

GenomeTraFaC analysis.
To identify conserved regulatory elements in mouse and human orthologs, TraFaC analysis was performed as described in Ref. 25, and the results were stored in the GenomeTraFaC database (http://genometrafac.cchmc.org). Briefly, complete genomic and cDNA sequences of mice and humans were downloaded from the UCSC Golden Path (28). With the use of the RepeatMasker program (http://ftp.genome.washington.edu/RM/RepeatMasker.html), repeat elements were masked before computational alignment using Advanced Pipmaker (http://bio.cse.psu.edu). The MatInspector Professional version 4.3 (http://www.genomatix.de) program was used to locate putative transcription factor binding sites in orthologous sequences utilizing the TRANSFAC database (63) (http://www.gene-regulation.com/) to identify matches in DNA sequences. The output consists of a table indicating a list of putative transcription factor binding sites. Exon annotations were based on the mRNAs of the Reference Sequence (RefSeq) database of the National Center for Biotechnology Information (NCBI).

CisMols analysis.
To identify putative consensus cis-acting regulatory sequences in genes that were coexpressed, putative cis-regulatory regions were identified and stored in GenomeTraFaC, using the CisMols server (http://cismols.cchmc.org). Cis-clusters with at least two binding sites and shared across at least five genes in each group were identified.


    RESULTS
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 GRANTS
 REFERENCES
 
The immunopathology of T1DM is both molecularly and cellularly complex. Although modern reductionist approaches have clearly provided new insights into the etiology and pathogenesis of T1DM, especially in the NOD mouse model, the sheer complexity of the disease necessitates, at some level, a more holistic approach to identify critical perturbations in the normal workings of the immune system and the target host tissue to reveal the disease process in its fullest extent. With this in mind, we undertook to apply functional genomics to identify key changes in the NOD mouse during the course of T1DM development. The major objectives of our study were 1) to establish a functional genomics framework to study the pathogenesis of T1DM, 2) to initiate the identification of genes involved in the evolution of T1DM from insulitis to diabetes, 3) to develop methodologies to provide a mechanism-based context for data analysis, and 4) to establish gene expression verification strategies that provide both predictive and mechanistic insights into disease progression.

Establishing a functional genomics approach.
For functional genomics to provide meaningful measurements of stage-specific gene expression, the specific disease stage must be well defined. BDC2.5 TCR transgenic NOD mice all manifest highly reproducible and distinct changes in phenotype during their progression toward T1DM (30, 33). We and others have previously shown that BDC/N and BDC/N.sc develop insulitis from 13 to 16 days of age, yet only the BDC/N.sc mice develop diabetes by 3–4 wk of age (Fig. 1A). BDC/N mice do not become diabetic until 4 mo of age, and then only with an ~40% incidence (Fig. 1A), representing mice with accelerated acquisition of the checkpoint 1 (insulitis) phenotype. The BDC/N.sc mice, on the other hand, represent mice with an accelerated transition through checkpoint 1 to checkpoint 2, overt diabetes (Fig. 1A). Therefore, we examined the gene expression differences in pancreata from four distinct groups of 3-wk-old mice: BDC/N.sc mice with both severe insulitis and recent onset diabetes, BDC/N mice with severe insulitis without diabetes, and NOD and N.sc mice that served as controls for preclinical disease, adaptive immunity, and genetic background. NOD mice in our barrier colony have a very low incidence (~5–7%) of peri-insulitis lesions at 3 wk of age (Fig. 1B). To allow for statistical analysis, two independent pools of RNA (6 mice/pool) were isolated and hybridized to Affymetrix U74Av2 microarrays.



View larger version (19K):
[in this window]
[in a new window]
 
Fig. 1. Defined onset and incidence of insulitis and diabetes in our 4 nonobese diabetic (NOD; N) mouse models. A: accelerated diabetes in BDC/N.sc mice. BDC/N.sc mice are diabetic by 4 wk of age, BDC/N mice develop reduced diabetes after 4 mo of age, NOD mice develop spontaneous diabetes to 97% (females) by 1 yr, and NOD.scid mice do not develop diabetes (not shown). For further explanation of mice, see MATERIALS AND METHODS. B: rapid insulitis in BDC/N and BDC/N.sc by 18–21 days of life. Insulitis is scored as either none, peri-insulitis (open bars), moderate (hatched bars, <50% infiltration), or severe (solid bars, >50% infiltration).

 
We used RMA normalization (23) followed by statistical and expression level analysis to establish a master list of 318 genes (of 12,448), comprising those genes deemed reliably "present" with clear hybridization signals above normalized background controls and with significant replicate expression (by t-test, P ≤ 0.05) by either log or simple normalized ratios corrected for false discovery, using the Benjamini-Hochberg algorithm (7). This master list is available as Supplemental Table S1 (supplemental materials are available at the Physiological Genomics web site)1 . By using whole pancreas analysis, we avoided the potential for artifact and bias due to differences in the enzymatic digestion and isolation of intact and infiltrated islets as disease progressed. Moreover, this technique proved highly reproducible, as can be seen in the comparison of hierarchical clustering of both individually isolated and hybridized samples with their replicate log averages (Supplemental Fig. S1). Mining these data, using K-means and hierarchical clustering with Pearson correlation, showed that our functional genomics approach provided clear sets of genes with tight statistical correlation to specific phenotypes and disease stages (Fig. 2).



View larger version (32K):
[in this window]
[in a new window]
 
Fig. 2. Distinct hierarchical clustering of gene subsets during disease progression. Left: K-means cluster sets. Genes were selected for analysis if their average expression level deviated from that of NOD.scid (N.sc) by at least 1.4-fold in at least 1 sample with P value < 0.05 and corrected for false discovery using Benjamini-Hochberg. Detailed lists are given in Supplemental Table S1. Shown are gene expression profiles of NOD (N), BDC2.5/NOD (BDC/N), and BDC2.5/NOD.scid (BDC/N.sc). Right: gene trees using Pearson correlation of log expression data. A list of the genes is given in Supplemental Table S1. Transcriptionally upregulated genes are indicated in red, and downregulated genes are indicated in blue. Yellow is unchanged. Data normalized to NOD.scid (N.sc).

 
Changes in gene expression during insulitis and diabetes.
Parsing our master list of 318 genes by K-means clustering, we identified genes involved in insulitis, diabetes, and ß-cell death in our four data sets. We observed five major clusters of genes (Fig. 2A), a complete list of which can be found in Supplemental Table S1. Group 1 contained genes with more or less equal expression in NOD, BDC/N, and BDC/N.sc. Genes highly expressed in NOD are in group 2. Group 3 contained genes upregulated in BDC/N coincident with the onset of insulitis but in the absence of diabetes. Group 4 contained genes modulated with late-stage insulitis and diabetes (BDC/N.sc). Lastly, group 5 contained genes with expression variation in both BDC/N and BDC/N.sc and, therefore, represents genes modulated throughout the course of insulitis, regardless of outcome. Given the importance of modulated gene expression during insulitis and diabetes, we chose to focus on the genes expressed in data sets 3, 4, and 5 (shown as gene trees in Fig. 2B). In addition, we analyzed syntactically our data through molecular pathway filters to ascertain which molecular cascades or gene clusters are associated with disease progression, specifically genes involved with islet antigen and antigen processing, T cell activation, differentiation and signaling, cytokine and chemokine production, lymphocyte trafficking, homing and adhesion, inflammation, ß-cell differentiation, cell cycle, and apoptosis. These data are detailed in Supplemental Table S2.

Islet cell autoantigens.
T1DM is an autoimmune disease, and islet cell autoantigens serve as targets for the T cells (52). Our GeneChip data show that most islet antigens, such as ICA-69, peripherin, and insulin, are lost with insulitis. Heat shock proteins (HSP) rose transiently with the onset of insulitis but declined with disease progression (Fig. 3). The 67-kDa isoform of glutamic acid decarboxylase (GAD67) was not detected at all. Expression of the 65-kDa isoform of GAD (GAD65) was only observed in NOD mice, and then only weakly. Islet destruction is an important pathological feature of T1DM, so it was not surprising that islet cell autoantigens were lost at this stage. This is most notable for the two insulin genes, where their levels of expression drop precipitously with the transition to late-stage insulitis and diabetes. In general, these data strongly suggest that known islet antigens are not markedly upregulated with insulitis, and the priming of autoimmunity via these antigens is before significant insulitis. In addition, they suggest that these antigens do not drive the late stages of the disease process.



View larger version (30K):
[in this window]
[in a new window]
 
Fig. 3. Differential expression of genes related to islet antigens, islet differentiation, integrin adhesion, and antigen processing in defined mouse models of type 1 diabetes mellitus (T1DM). Shown are gene trees using Pearson correlation of log expression data. Transcriptionally upregulated genes are indicated in red, and downregulated genes are indicated in blue. Yellow is unchanged. Data normalized to N.sc.

 
Antigen processing genes.
With late-stage insulitis and diabetes in BDC/N.sc mice, there is a marked upregulation of interferon (IFN)-{gamma}-induced antigen processing genes (MHCI + II, Lmp7, ctsS) and downregulation of ctsL, several peptidases, and H2Oa. Although it can be argued that these appreciable rises in IFN-{gamma}-induced genes can be accounted for by a generalized accumulation of host antigen presenting cells (APC) during late-stage insulitis and diabetes, we did not observe a global rise in generalized APC message levels but rather a specific rise in IFN-{gamma}-induced transcripts (Fig. 3). These data are consistent with a prominent role for IFN-{gamma} and activated APC at this critical stage of T1DM pathogenesis.

Leukocyte adhesion molecules.
Infiltration of islets with immune cells is thought to occur in T1DM in response to localized inflammation. As seen in Fig. 3, we observed increased expression of P-selectin ligand in diabetes. It has been reported that Th1 but not Th2 cells express this gene, thereby allowing migration in response to E- and P-selectins (6). We also observed increased expression of transcripts from genes encoding adhesion molecules such as thrombospondin (THBS)1.

Integrins and adhesion molecules.
The integrins and Ig superfamily adhesion molecules are important for stopping leukocyte rolling and mediating leukocyte aggregation and transendothelial migration (65). The migration of T cells during the pancreatic inflammatory response is integrin dependent; however, only vascular cell adhesion molecule (VCAM)1 gene (Vcam1) expression modulated significantly with progression to diabetes (Fig. 3).

Islet cell differentiation genes.
Elevated expression of the regenerating genes Reg3{alpha}, Reg3ß, and Reg3{gamma} was observed during insulitis in both BDC/N and BDC/N.sc (Fig. 3). These genes, first isolated from the rat (51), are IL-6 responsive and showed augmented expression upon inflammation (14). These genes are associated with protection from oxidative damage and may play a role in islet regeneration and islet cell homeostasis (44). Of particular note is a recent study suggesting that Reg3ß may act as a autoantigen in NOD mice and humans (20). Moreover, the prominence of the Reg gene family in the maintenance of islet homeostasis is underscored by the recent gene-targeted disruption of Reg3{gamma}, which exhibited a significant reduction in adult islet cell mass (58).

It is important to note that, with insulitis, Pax1, Pdx1, Neurog2, and other early ß-cell differentiation genes are either unaffected or are downmodulated (Fig. 3), suggesting that the differentiation of new ß-cells from pancreatic stem cells is highly unlikely or at least does not require alterations in these critical differentiation genes.

Cytokines and chemokines and their receptors.
Cytokines alter the local expression of chemokine and adhesion molecules to increase recruitment of specific lymphocyte populations. To our surprise, we could not detect the presence of the Th1-specific marker IFN-{gamma} in our analysis; however, IFN-{gamma} mRNA levels are generally quite low and therefore did not meet our expression cutoffs. Yet there is undeniable evidence that IFN-{gamma} is present, as established by IFN-{gamma}-responsive gene expression (Figs. 3 and 4 and Supplemental Table S1). Elevated IFN-{gamma} mRNA levels are readily observed, however, by real-time PCR (Fig. 5). Moreover, the IFN-{gamma}-driven expression of CXCL9 and CXCL10 are observed during the late stages of insulitis and diabetes seen in BDC/N.sc (Fig. 4A). Increased expression of CCL5 in BDC/N.sc (1.6-fold) is also observed. This is consistent with published data implicating CCL5 as a critical chemokine in T1DM (8).



View larger version (22K):
[in this window]
[in a new window]
 
Fig. 4. Upregulation of CXCL9, CCL5, CCL8, and CCL9 as well as Th1 cytokines during late-stage insulitis and diabetes. Shown are patterns of expression of chemokine receptors and their ligands (A) and cytokines and their receptors (B) in BDC/N.sc mice relative to BDC/N mice.

 


View larger version (31K):
[in this window]
[in a new window]
 
Fig. 5. Relapsing/remitting expression of validated gene set as revealed by quantitative real-time PCR. A kinetic study of a representative set of validated genes is shown from 1 to 20 wk of age. BDC/N mice do not survive past 4 wk. Three independent experiments, in duplicate, were performed from 6 individual mice per group per time point per isolation.

 
With the onset of diabetes, Il17 and Il18, both associated with memory and Th1 T cell responses, are upregulated (Fig. 4B). These cytokines, along with transforming growth factor (TGF)-{alpha} and TGF-ß3 and a number of inflammation-associated cytokine receptors, form the general cytokine mRNA expression pattern seen with late-stage insulitis and diabetes.

The upregulation of the cytokine-related signaling molecule nuclear factor (NF)-IL-6 (Nfil6), also referred to as CCAAT/enhancer binding protein (C/EBP), in BDC/N.sc samples (2.4-fold) suggests that IL-6 plays an important role in T1DM pathogenesis. NF-IL-6 is a DNA binding protein and mediates IL-1-induced IL-6 transcription. Elevated levels of IL-6 have been reported in T1DM and are seen here (Fig. 4B). Downregulation of cytokines such as IL-1ß, IL-2, IL-3, IL-7, IL-11, and IL-13 was observed with the onset of late-stage insulitis (Fig. 4B). Taken as a whole, these data strongly suggest that the initial inflammatory response in T1DM is weakly polarized or represents a mixed Th1/Th2 response, which gradually morphs into a dominant Th1 response with late-stage insulitis.

Th1-specific genes in T1DM.
Where assessed in absolute terms or normalized to TCR levels, the Th1 differentiation genes such as Tbet, GADD45 members, Rac2, and Il12Rb2 are upregulated in BDC/N.sc mice but are largely absent in BDC/N mice, suggesting a molecular switch before onset of diabetes (Fig. 5 and Supplemental Table S1). T-bet is a transcription factor that plays a central role in the development of Th1 T cells (50). T-bet transactivates the IFN-{gamma} gene, induces IFN-{gamma} production when transduced into primary T cells, and redirects polarized Th2 cells into the Th1 pathway (50). GADD45ß is induced by TCR signaling in both naive and effector CD4+ T cells and required for the function and generation of Th1 cells (37). GADD45{gamma} (CR6/OIG37) is induced during T cell activation and is higher in Th1 cells than in Th2 effector cells (38). The small guanosine triphosphatase Rac2 is expressed selectively in mouse Th1 cells (34). The elevated expression of these Th1-specific genes in our mouse model clearly indicates a role for Th1 T cells in the spontaneous development of T1DM in the NOD mouse.

Complement components and inflammation.
Alternative complement pathway components such as C3 and C4 were upregulated (8.5- and 3.2-fold, respectively) during diabetes, in the absence of immunoglobulins, suggesting antibody-independent T cell-mediated inflammatory responses (Supplemental Table S2). Increased expression of Tnfr2 was observed during diabetes in BDC/N.sc. We also observed augmented transcripts for ceruloplasmin and metallothionein genes (2.6- and 2.7-fold, respectively) during diabetes. Ceruloplasmin or multicopper oxidase is an acute-phase responsive enzyme, and its increased expression here likely reflects greater oxidant stress (11).

High expression of Atf4, Atf5, and Hspa5 during insulitis in BDC2.5/N mice suggests ER stress. ER stress is known to occur in several pathophysiological states such as hypoxia. ATF4 is basally repressed in nonstressed cells. Under stress, ATF4 translation is derepressed through the phosphorylation of eIF2{alpha}. The role of ATF5 in islet function or maintenance is not known. ATF5 is expressed in the central nervous system, implicating its role in neuroendocrine as well as pancreatic endocrine development and function (5). The role for HSPA5 in T1DM is likewise unclear, although it is a stress response gene.

Kinetic analysis and gene expression validation via real-time PCR.
Independent verification of our microarray data is vital. Moreover, a dynamic study of these disease-associated genes is more likely to shed new mechanic insights into T1DM than simple static measures of expression. To these ends, we isolated high-quality pancreatic RNA from our mice, quantified mRNA expression using real-time PCR, and standardized expression to ß-actin. As shown in Fig. 5, we followed the expression of Tcrb, VCAM1, casp1, casp12, Gzmb, Il4, Ccl8, Tnfa, and a set of genes specific for Th1 population such as Ifng, Tbet, Cr6/Gadd45g, Il17, and Rac2. We observed a rapid increase in the expression levels for all of the above genes by 3 wk of age in BDC/N.sc mice. As for BDC/N and NOD mice, the expression levels were low at 3 wk compared with BDC/N.sc mice, thereby verifying our previous microarray data, but then underwent significant modulation with time. Although not all genes showed synchronicity in their modulation, there is a generalized trend of periodic expression modulation in these and other genes tested (Fig. 5 and data not shown). These data show that the insulitis in BDC/N and NOD mice is a dynamic waxing and waning process, suggesting that T1DM is a relapsing/remitting autoimmune disease much like multiple sclerosis.

In addition to real-time PCR analysis, we also analyzed the stage-specific expression of Reg genes, Reg3{alpha}, -ß, and -{gamma}, using semi-quantitative PCR, necessitated by the high homology in the Reg gene family that precluded real-time analysis (Fig. 6). The expression of Reg3{alpha}, Reg3ß, and Reg3{gamma} genes is clearly elevated in BDC2.5/N.sc and BDC/N pancreas as early as 1–2 wk of age, at the time coincident with the initiation of insulitis (33). Therefore, these genes appear to serve as excellent markers for insulitis. Expression of Reg3{gamma} above basal levels was not seen in NOD mice until 8 wk (Fig. 6A), at an age when insulitis begins in these animals. It remained at basal levels in N.sc mice throughout the entire assay period. Interestingly, these genes do not wax and wane, suggesting that the protection and repair response may be an ongoing event once infiltration ensues (Fig. 6B).



View larger version (24K):
[in this window]
[in a new window]
 
Fig. 6. Reg3{gamma} gene expression is an excellent marker for insulitis. A: semiquantitative PCR with time in pancreatic RNA samples of BDC/N.sc, BDC/N, and N mice reveals a tight correlation between gene expression and onset of insulitis. Representative experiment of 5 independent experiments. B: relative gene expression for Reg family members by densitometry. BDC/N.sc, solid symbol; BDC/N, shaded symbol; N, open symbol. BDC/N mice do not survive past 4 wk. Densitometry is average ± SD from 3 independent experiments.

 
Because cytokine mRNA levels do not always correlate with cytokine protein production levels, we measured directly plasma-available levels of IL-17 in the of BDC/N and BDC/N.sc mice. We observed high levels of IL-17 as early as day 10 of life in BDC/N.sc mice (Fig. 7), concurrent with the onset of severe insulitis in these animals and preceding diabetes by 7–10 days. In fact, at the time of diabetes onset (day 21), the levels of IL-17 plummeted dramatically (Fig. 7). On the other hand, IL-17 levels tended to vary minimally in BDC/N mice. These data are consistent with the real-time PCR analysis for IL-17 transcripts (Fig. 5).



View larger version (14K):
[in this window]
[in a new window]
 
Fig. 7. Enhanced IL-17 expression during the run-up to diabetes in BDC/N mice. IL-17 levels were quantified in the serum samples of BDC/N and BDC/N.sc mice. Serum samples were collected from 6 individual mice per group per time point over a 16-wk period. BDC/N mice (shaded circle) reveal low-level relapsing modulation of IL-17, whereas BDC/N.sc mice (solid circle) show rapid IL-17 expression in the period directly preceding overt T1DM. BDC/N.sc mice are diabetic after 21 days. BDC/N mice did not develop diabetes during the 16-wk experimental period.

 
Identification of shared regulatory cascades in T1DM using cis-regulatory element cluster analysis.
Identifying the cis-element clusters shared by coordinately regulated gene groupings will provide new mechanistic insight into T1DM pathogenesis by helping to predict salient gene-gene interactions that occur during disease evolution. To this end, we employed GenomeTraFaC and CisMols analyses, developed by the CCHRF Division of Pediatric Informatics (25), to identify cis-regulatory elements in coordinately regulated genes of set 3 and set 4 from Fig. 2. The results of TraFaC analysis consist of 1) a regulogram, a graphical representation of common binding sites (hits) in the context of sequence similarity between two orthologs (Fig. 8A provides an example, using VCAM1), and 2) a TraFaCgram, a graphical representation of shared transcription factor binding sites between the two orthologs (Fig. 8B, again using VCAM1). Forty of the 93 genes from cluster 3 (high expression of genes in BDC/N) and 30 of the 87 genes (high expression in BDC/N.sc) and their human orthologs were examined. With the use of the CisMols server (http:// cismols.cchmc.org), an extension of GenomeTraFaC that identifies compositionally similar cis-regulatory element clusters occurring in groups of coregulated genes, we identified, ranked and statistically compared regulatory clusters from genes in set 3 and set 4 to control genes (20 genes selected at random from the roughly 7,300 genes in the GenomeTraFaC database that were not included in either set 3 or 4 but were expressed in our dataset). The data indicate that coordinately regulated genes in a given cluster or set (in our case, disease stage) share similar transcription factor binding sites or cis-element similarity. As shown in Fig. 8C, our data indicate that, during insulitis (set 3), cis-regulatory modules containing Ets and one or more zinc finger binding factor consensus sites were statistically overrepresented. These cis-regulatory clusters have been reported to drive expression of genes associated with hypoxia (64), including angiogenesis (17) and inflammation, especially tumor necrosis factor (TNF)-{alpha}-induced inflammation (55), as well as expression of superoxide dismutase and nitric oxide synthase genes (57). These findings are consistent with the observed changes in gene expression during insulitis, which showed marked increases in TNF-{alpha}- and IL-1/IL-6-induced inflammatory genes and increased levels of superoxide dismutase and synthase genes and hypoxia- and stress-induced genes. This suggests that the initial phase of insulitis is controlled to a great degree by the localized liberation of TNF-{alpha} and the recruitment of leukocytes via TNF-{alpha} and IL-1/IL-6-induced chemokines.



View larger version (40K):
[in this window]
[in a new window]
 
Fig. 8. GenomeTraFaC and CisMols analyses reveal use of Ets family of transcription factors during insulitis and C/EBP and Nkx2.2 transcription factors during diabetes. A: identification of conserved and shared cis-elements between human and mouse sequences, depicted as a regulogram, for VCAM1. The 2 genomic sequences are represented as horizontal bars with exons in red. The percent identity and abundance of shared transcription factor binding sites (hits) are shown. B: TraFaCgram of the mouse VCAM1 promoter region and its human ortholog. Horizontal bars are genomic sequences, and the nos. indicate nucleotide positions. C: summary table of cis-regulatory elements shared between members of set 3 (n = 40 genes) and set 4 (n = 30 genes) from Fig. 3.; lists were truncated to include only those clusters with 5 or more genes. All bold and colored clusters are statistically significant (P < 0.01, Fischer Exact).

 
Analysis of the cis-regulatory modules from set 4 (late insulitis and diabetes) reveals that the major statistically relevant change in regulatory element usage is the use of C/EBP in concert with the homeodomain transcription factor Nkx2 and Oct1 (Fig. 8C). These clusters are associated with the expression of several effector cytokines and chemokines, including monocyte chemoattractant protein-1, and the expression of the addressin intercellular adhesion molecule-1 (26) as well as critical pancreatic ß-cell functions and glucose-responsiveness (39), including the differentiation of ß-cell tissue (49) and insulin expression (62).

To analyze whether the islet environment is hypoxic, as predicted by GenomeTraFaC and CisMols during insulitis, we conducted experiments using EF-5. EF-5 covalently binds to cells and tissues in a hypoxic state; this binding is easily revealed using an anti EF-5 monoclonal antibody (18). Both BDC/N and BDC/N.sc mice (3 wk old) were given intraperitoneal injections of EF-5 in saline; 12–16 h later, the mice were killed and tissues were sampled and stained for evidence of localized hypoxia. We found clear evidence of a highly hypoxic local environment in and around infiltrated islets in BDC/N mice (Fig. 9). No hypoxia was observed in the liver and spleen tissue sections from BDC/N. Interestingly, pancreatic sections from BDC/N.sc showed little if any regions of localized hypoxia (data not shown); however, at this point in time, BDC/N.sc mice had little intact islet mass and markedly reduced inflammation associated with end-stage disease, consistent with our predicted analysis using CisMols.



View larger version (48K):
[in this window]
[in a new window]
 
Fig. 9. Localized hypoxia during insulitis in BDC/N mice as revealed by EF-5 reactivity. Immunohistochemistry analysis for insulin (middle; AlexaFluor-568) and EF-5 (left; AlexaFluor-488) in the pancreas and liver of BDC/N mice. Hematoxylin and eosin (H/E) parallel sections are at right. Three-week-old BDC/N and BDC/N.sc animals were injected with EF-5 and killed after 12 h, and the tissue was sampled and fixed, followed by staining with anti-EF-5 to reveal areas of localized hypoxia. Pancreatic islets of BDC/N were hypoxic, and no hypoxia was detected in liver. Tissue section of BDC/N.sc animals did not show hypoxia (data not shown).

 

    DISCUSSION
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 GRANTS
 REFERENCES
 
We used cDNA microarray technology to analyze the transcriptome of NOD mice to identify, constrain, and define subsets of genes that modulate with diabetic progression. Our data showed that 1) insulitis is a IL-1/IL-6/TNF-{alpha}-driven process; 2) insulitis caused the upregulation of a family of genes associated with end-stage ß-cell maturation and repair, the Reg gene family; 3) insulitis is a mixed Th1/Th2 infiltration; 4) late-stage insulitis and overt diabetes correlate with a focused Th1 T cell response with unregulated IL-17; 5) IFN-{gamma}-responsive genes play a critical role in diabetes pathogenesis; 6) insulitis correlates with the usage of the Ets family of transcription factors, whereas diabetes correlates with the usage of C/EBP and Nkx2 transcription factors; and 7) islet environment is highly hypoxic during rapid insulitis.

THBS1, a glycoprotein released from platelet granules in response to thrombin stimulation (2), is a transient component of the extracellular matrix known to promote chemotaxis of leukocytes to inflammatory sites. It has been shown that THBS1 provides a costimulatory signal, necessary for the activation of autoreactive T cells (59). The specific expression of THBS1 in NOD T1DM is consistent with the observed upregulation of THBS1 in vessel walls of diabetic Zucker rats and is a direct response of vascular cells to glucose (48). In addition, we observed that high glucose levels upregulate THBS1-dependent TGF-ß activation by altering cGMP-dependent protein kinase activity, itself regulated by a decreased nitric oxide signaling (see Fig. 4B). A role for THBS1 has not been previously reported for NOD diabetes.

Consistent with our observation, the adoptive transfer of diabetes in the NOD mouse can be delayed by treatment with anti-VCAM-1 (24, 56). Therefore, the blockade of VCAM-1 may represent a potential therapeutic treatment for T1DM.

As for chemokines, CCL5 is known to have a number of profound consequences on T cells, including costimulation of cytokine release and T cell proliferation. Similar results were observed by Eaves et al. (15), using congenic strains and microarray gene expression analysis, in NOD mice. CCL5 levels are higher in a wide range of inflammatory disorders and pathologies (19) including pancreatic infiltrates that promote rapid destruction of the insulin-producing ß-cells in the NOD mouse (8). Interestingly, CCL5 maps to the known diabetes control locus, IDD4, which plays a role in modulating the tempo and severity of insulitis in the NOD mouse (61). It has been reported that a pancreatic Th1 cytokine environment greatly accelerates the recruitment of islet-specific CD4+ T cells to the pancreas, especially in the presence of IFN-{gamma}, suggesting that the rapid onset of diabetes in the BDC/N.sc mouse is in part due to enhanced levels of IFN-{gamma}-responsive chemokines (22).

IL-17 is a proinflammatory cytokine, originally cloned from herpes virus (66). It is produced by activated and memory CD4+ T cells, particularly Th1 cells (1). IL-17 has pleiotropic activities including induction of TNF-{alpha}, IL-1, and various other cytokines, chemokines, and adhesion molecules that play important roles in various inflammatory responses (27). In rheumatoid arthritis, IL-17 acts in association with TNF-{alpha} to destroy cartilage; in this respect, IL-17 acts as an effector cytokine (60). Its exact role in T1DM pathogenesis has not previously been explored. Our identification of it here is therefore intriguing. Its presence here is consistent with its action as an effector cytokine, perhaps under the control of IL-12 and IL-23, both of which are present during this period and are known to regulate IL-17 expression (3).

The low level and periodic modulation of Il17 observed in the BDC/N mice is consistent with a relapsing/remitting course of insulitis. Treatment of N.sc recipient mice of activated BDC2.5 T cells with anti-IL-17 antibody did not prevent the onset of diabetes significantly (data not shown), suggesting that activated effector T cells act in an IL-17-dependent manner. However, what is not yet clear is whether IL-17 inhibition can alter the de novo activation of diabetogenic T cells, but this seems unlikely given the phenotype of IL-17-deficient mice, as IL-17 was mainly seen as an effector cytokine (43). Further studies are required to establish the importance of this particular cytokine in T1DM. At present, IL-17 seems to represent an important new marker for progression toward ß-cell destruction and diabetes onset that is reliably detectable in peripheral blood of NOD mice (Fig. 7).

Metallothionein is associated with stress responses; its ectopic overexpression in pancreatic ß-cells protects islets from hypoxia, providing broad resistance to oxidative stress by scavenging most kinds of reactive oxygen species and by reduced nitric oxide-induced ß-cell death (35). The influence of metallothionein upregulation is observed in multiple sclerotic lesions as well (36) . Metallothionein is likewise modulated in experimental autoimmune encephalitis (EAE), where metallothioneine deficiency results in the development of more clinically severe EAE (46).

The recent observation that the Reg3ß serves as an autoantigen (20) suggests that the Reg gene family may drive the observed "antigenic spread" seen in T1DM by providing new antigenic substrates for autoreactive T cell recognition. We propose then that islet infiltration is initiated by Th2 or mixed T cell subsets; this retrograde inflammation induces ß-cell repair/protection response in the absence of additional de novo differentiation of ß-cells from pancreatic stem cells (as evidenced by the lack of expression of early ß-cell differentiation genes and by the lack of proapoptosis gene expression). This in turn produces a state of "reactive homeostasis," where ß-cell damage and repair are in equilibrium. However, once a dominant Th1 response results and the pace of ß-cell death quickens, cellular repair is insufficient to maintain ß-cell mass, and frank diabetes ensues.

To date, studies on differential expression of genes in T1DM using cDNA microarrays have taken advantage of pancreatic ß-cell lines, as they provide an attractive source of starting material for analysis and in vitro manipulation (9, 67). In addition, these cell lines are rather homogeneous, which provides yet another distinct advantage for expression studies. However, these cell lines cannot model the complexity of in vivo interactions among islets, exocrine pancreas, and the host immune response seen in T1DM. Other approaches include the use of NOD mice spleens to define key checkpoints for T1DM (16). Our data, on the other hand, provide a more global survey of gene modulation in vivo over a 20-wk period and indicate that insulitis is not a progressively increasing process but a relapsing and remitting one. Genes examined in our experiments should now afford the immunology community a "first pass" opportunity to link specific phenotypes with known genetic linkages and provide a framework within which to measure how known induced and spontaneous alterations in the NOD perturb specific genetic pathways implicated in the disease process. In conclusion, we have analyzed and verified the differential expression of genes during the evolution of T1DM pathogenesis in the NOD and in genetically altered BDC/N and BDC2.5/N.sc mice using cDNA microarrays. Microarray analysis of large-scale gene expression provides a versatile tool to open new avenues of research and simultaneously provides for the exciting possibilities for the identification of novel drug targets in T1DM that may facilitate the noninvasive staging and treatment of prediabetic individuals.


    GRANTS
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 GRANTS
 REFERENCES
 
The CCHRF Affymetrix Core is supported in part by National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) Grant R24-DK-064403. The Bioinformatics Core is supported in part by grants from the Howard Hughes Medical Institute and the State of Ohio. This work is supported in part by Juvenile Diabetes Research Foundation Grant 1-2002-142 and NIDDK Grant DK-062274 (to J. D. Katz).


    ACKNOWLEDGMENTS
 
We thank Sara Rankin and Shawn Smith, Affymetrix Core, CCHRF, for help with GeneChip experiments and Sarah Williams, Pediatric Informatics, in the preliminary analysis of GeneChip data; Sue Kong and Kristin Stanley for help with RMA analysis; Ashima Gupta for CisMols analysis; and Yuhui Qiu and Laura Spangler Mead for technical help. We wish to thank Dr. Charles Caldwell for the gift of the EF-5 compound and antibody. We thank Dr. Christopher Karp and Dr. Bo Wang for critical review of the manuscript.


    FOOTNOTES
 
Article published online before print. See web site for date of publication (http://physiolgenomics.physiology.org).

Address for reprint requests and other correspondence: J. D. Katz, Cincinnati Children’s Hospital Research Foundation and College of Medicine, Univ. of Cincinnati, 3333 Burnet Ave., Cincinnati, OH 45229-3039 (E-mail: jonathan.katz{at}cchmc.org).

10.1152/physiolgenomics.00173.2004.

1 The Supplemental Material for this article (Supplemental Tables S1 and S2 and Supplemental Fig. S1) is available online at http://physiolgenomics.physiology.org/cgi/content/full/00173.2004/DC1. Back


    REFERENCES
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 GRANTS
 REFERENCES
 

  1. Aarvak T, Chabaud M, Miossec P, and Natvig JB. IL-17 is produced by some proinflammatory Th1/Th0 cells but not by Th2 cells. J Immunol 162: 1246–1251, 1999.[Abstract/Free Full Text]
  2. Adams JC. Thrombospondin-1. Int J Biochem Cell Biol 29: 861–865, 1997.[CrossRef][ISI][Medline]
  3. Aggarwal S, Ghilardi N, Xie MH, de Sauvage FJ, and Gurney AL. Interleukin-23 promotes a distinct CD4 T cell activation state characterized by the production of interleukin-17. J Biol Chem 278: 1910–1914, 2003.[Abstract/Free Full Text]
  4. Andre I, Gonzalez A, Wang B, Katz J, Benoist C, and Mathis D. Checkpoints in the progression of autoimmune disease: lessons from diabetes models. Proc Natl Acad Sci USA 93: 2260–2263, 1996.[Abstract/Free Full Text]
  5. Angelastro JM, Ignatova TN, Kukekov VG, Steindler DA, Stengren GB, Mendelsohn C, and Greene LA. Regulated expression of ATF5 is required for the progression of neural progenitor cells to neurons. J Neurosci 23: 4590–4600, 2003.[Abstract/Free Full Text]
  6. Austrup F, Vestweber D, Borges E, Lohning M, Brauer R, Herz U, Renz H, Hallmann R, Scheffold A, Radbruch A, and Hamann A. P- and E-selectin mediate recruitment of T-helper-1 but not T-helper-2 cells into inflammed tissues. Nature 385: 81–83, 1997.[CrossRef][ISI][Medline]
  7. Benjamini R and Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. JR Stat Soc B57: 289–300, 1995.
  8. Bradley LM, Asensio VC, Schioetz LK, Harbertson J, Krahl T, Patstone G, Woolf N, Campbell IL, and Sarvetnick N. Islet-specific Th1, but not Th2, cells secrete multiple chemokines and promote rapid induction of autoimmune diabetes. J Immunol 162: 2511–2520, 1999.[Abstract/Free Full Text]
  9. Cardozo AK, Heimberg H, Heremans Y, Leeman R, Kutlu B, Kruhoffer M, Orntoft T, and Eizirik DL. A comprehensive analysis of cytokine-induced and nuclear factor-{kappa}B-dependent genes in primary rat pancreatic ß-cells. J Biol Chem 276: 48879–48886, 2001.[Abstract/Free Full Text]
  10. Chomczynski P and Sacchi N. Single-step method of RNA isolation by acid guanidinium thiocyanate-phenol-chloroform extraction. Anal Biochem 162: 156–159, 1987.[CrossRef][ISI][Medline]
  11. Cunningham J, Leffell M, Mearkle P, and Harmatz P. Elevated plasma ceruloplasmin in insulin-dependent diabetes mellitus: evidence for increased oxidative stress as a variable complication. Metabolism 44: 996–999, 1995.[CrossRef][ISI][Medline]
  12. Dhanasekaran SM, Barrette TR, Ghosh D, Shah R, Varambally S, Kurachi K, Pienta KJ, Rubin MA, and Chinnaiyan AM. Delineation of prognostic biomarkers in prostate cancer. Nature 412: 822–826, 2001.[CrossRef][ISI][Medline]
  13. Dolganov GM, Woodruff PG, Novikov AA, Zhang Y, Ferrando RE, Szubin R, and Fahy JV. A novel method of gene transcript profiling in airway biopsy homogenates reveals increased expression of a Na+-K+-Cl cotransporter (NKCC1) in asthmatic subjects. Genome Res 11: 1473–1483, 2001.[Abstract/Free Full Text]
  14. Dusetti NJ, Frigerio JM, Keim V, Dagorn JC, and Iovanna JL. Structural organization of the gene encoding the rat pancreatitis-associated protein. Analysis of its evolutionary history reveals an ancient divergence from the other carbohydrate-recognition domain-containing genes. J Biol Chem 268: 14470–14475, 1993.[Abstract/Free Full Text]
  15. Eaves IA, Wicker LS, Ghandour G, Lyons PA, Peterson LB, Todd JA, and Glynne RJ. Combining mouse congenic strains and microarray gene expression analyses to study a complex trait: the NOD model of type 1 diabetes. Genome Res 12: 232–243, 2002.[Abstract/Free Full Text]
  16. Eckenrode SE, Ruan Q, Yang P, Zheng W, McIndoe RA, and She JX. Gene expression profiles define a key checkpoint for type 1 diabetes in NOD mice. Diabetes 53: 366–375, 2004.[Abstract/Free Full Text]
  17. Elvert G, Kappel A, Heidenreich R, Englmeier U, Lanz S, Acker T, Rauter M, Plate K, Sieweke M, Breier G, and Flamme I. Cooperative interaction of hypoxia-inducible factor-2{alpha} (HIF-2{alpha}) and Ets-1 in the transcriptional activation of vascular endothelial growth factor receptor-2 (Flk-1). J Biol Chem 278: 7520–7530, 2003.[Abstract/Free Full Text]
  18. Evans SM, Joiner B, Jenkins WT, Laughlin KM, Lord EM, and Koch CJ. Identification of hypoxia in cells and tissues of epigastric 9L rat glioma using EF5 [2-(2-nitro-1H-imidazol-1-yl)-N-(2,2,3,3,3-pentafluoropropyl) acetamide]. Br J Cancer 72: 875–882, 1995.[ISI][Medline]
  19. Gerard C and Rollins BJ. Chemokines and disease. Nat Immun 2: 108–115, 2001.[CrossRef][ISI]
  20. Gurr W, Yavari R, Wen L, Shaw M, Mora C, Christa L, and Sherwin RS. A Reg family protein is overexpressed in islets from a patient with new-onset type 1 diabetes and acts as T-cell autoantigen in NOD mice. Diabetes 51: 339–346, 2002.[Abstract/Free Full Text]
  21. Haskins K, Portas M, Bradley B, Wegmann D, and Lafferty K. T-lymphocyte clone specific for pancreatic islet antigen. Diabetes 37: 1444–1448, 1988.[Abstract]
  22. Hill NJ, Van Gunst K, and Sarvetnick N. Th1 and Th2 pancreatic inflammation differentially affects homing of islet-reactive CD4 cells in nonobese diabetic mice. J Immunol 170: 1649–1658, 2003.[Abstract/Free Full Text]
  23. Irizarry RA, Bolstad BM, Collin F, Cope LM, Hobbs B, and Speed TP. Summaries of Affymetrix GeneChip probe level data. Nucleic Acids Res 31: e15, 2003.[Abstract/Free Full Text]
  24. Jakubowski A, Ehrenfels BN, Pepinsky RB, and Burkly LC. Vascular cell adhesion molecule-Ig fusion protein selectively targets activated {alpha}4-integrin receptors in vivo. Inhibition of autoimmune diabetes in an adoptive transfer model in nonobese diabetic mice. J Immunol 155: 938–946, 1995.[Abstract]
  25. Jegga AG, Sherwood SP, Carman JW, Pinski AT, Phillips JL, Pestian JP, and Aronow BJ. Detection and visualization of compositionally similar cis-regulatory element clusters in orthologous and coordinately controlled genes. Genome Res 12: 1408–1417, 2002.[Abstract/Free Full Text]
  26. Ji B, Chen XQ, Misek DE, Kuick R, Hanash S, Ernst S, Najarian R, and Logsdon CD. Pancreatic gene expression during the initiation of acute pancreatitis: identification of EGR-1 as a key regulator. Physiol Genomics 14: 59–72, 2003.[Abstract/Free Full Text]
  27. Jovanovic DV, Di Battista JA, Martel-Pelletier J, Jolicoeur FC, He Y, Zhang M, Mineau F, and Pelletier JP. IL-17 stimulates the production and expression of proinflammatory cytokines, IL-ß and TNF-{alpha}, by human macrophages. J Immunol 160: 3513–3521, 1998.[Abstract/Free Full Text]
  28. Karolchik D, Baertsch R, Diekhans M, Furey TS, Hinrichs A, Lu YT, Roskin KM, Schwartz M, Sugnet CW, Thomas DJ, Weber RJ, Haussler D, and Kent WJ. The UCSC Genome Browser Database. Nucleic Acids Res 31: 51–54, 2003.[Abstract/Free Full Text]
  29. Karp CL, Grupe A, Schadt E, Ewart SL, Keane-Moore M, Cuomo PJ, Kohl J, Wahl L, Kuperman D, Germer S, Aud D, Peltz G, and Wills-Karp M. Identification of complement factor 5 as a susceptibility locus for experimental allergic asthma. Nat Immun 1: 221–226, 2000.[CrossRef][ISI]
  30. Katz J, Benoist C, and Mathis D. Major histocompatibility complex class I molecules are required for the development of insulitis in non-obese diabetic mice. Eur J Immunol 23: 3358–3360, 1993.[ISI][Medline]
  31. Katz JD, Benoist C, and Mathis D. T helper cell subsets in insulin-dependent diabetes. Science 268: 1185–1188, 1995.[ISI][Medline]
  32. Katz JD, Wang B, Haskins K, Benoist C, and Mathis D. Following a diabetogenic T cell from genesis through pathogenesis. Cell 74: 1089–1100, 1993.[CrossRef][ISI][Medline]
  33. Kurrer MO, Pakala SV, Hanson HL, and Katz JD. Beta cell apoptosis in T cell-mediated autoimmune diabetes. Proc Natl Acad Sci USA 94: 213–218, 1997.[Abstract/Free Full Text]
  34. Li B, Yu H, Zheng W, Voll R, Na S, Roberts AW, Williams DA, Davis RJ, Ghosh S, and Flavell RA. Role of the guanosine triphosphatase Rac2 in T helper 1 cell differentiation. Science 288: 2219–2222, 2000.[Abstract/Free Full Text]
  35. Li X, Chen H, and Epstein PN. Metallothionein protects islets from hypoxia and extends islet graft survival by scavenging most kinds of reactive oxygen species. J Biol Chem 279: 765–771, 2004.[Abstract/Free Full Text]
  36. Lock C, Hermans G, Pedotti R, Brendolan A, Schadt E, Garren H, Langer-Gould A, Strober S, Cannella B, Allard J, Klonowski P, Austin A, Lad N, Kaminski N, Galli SJ, Oksenberg JR, Raine CS, Heller R, and Steinman L. Gene-microarray analysis of multiple sclerosis lesions yields new targets validated in autoimmune encephalomyelitis. Nat Med 8: 500–508, 2002.[CrossRef][ISI][Medline]
  37. Lu B, Ferrandino AF, and Flavell RA. Gadd45ß is important for perpetuating cognate and inflammatory signals in T cells. Nat Immun 5: 38–44, 2004.[CrossRef][ISI]
  38. Lu B, Yu H, Chow C, Li B, Zheng W, Davis RJ, and Flavell RA. GADD45{gamma} mediates the activation of the p38 and JNK MAP kinase pathways and cytokine production in effector TH1 cells. Immunity 14: 583–590, 2001.[CrossRef][ISI][Medline]
  39. Lu M, Seufert J, and Habener JF. Pancreatic ß-cell-specific repression of insulin gene transcription by CCAAT/enhancer-binding protein ß. Inhibitory interactions with basic helix-loop-helix transcription factor E47. J Biol Chem 272: 28349–28359, 1997.[Abstract/Free Full Text]
  40. Makino S and Tochino Y. The spontaneously non-obese diabetic mouse. Exp Anim 27: 27–29, 1978.
  41. Mathis D, Vence L, and Benoist C. ß-Cell death during progression to diabetes. Nature 414: 792–798, 2001.[CrossRef][ISI][Medline]
  42. Nadler ST and Attie AD. Please pass the chips: genomic insights into obesity and diabetes. J Nutr 131: 2078–2081, 2001.[Abstract/Free Full Text]
  43. Nakae S, Komiyama Y, Nambu A, Sudo K, Iwase M, Homma I, Sekikawa K, Asano M, and Iwakura Y. Antigen-specific T cell sensitization is impaired in IL-17-deficient mice, causing suppression of allergic cellular and humoral responses. Immunity 17: 375–387, 2002.[CrossRef][ISI][Medline]
  44. Narushima Y, Unno M, Nakagawara K, Mori M, Miyashita H, Suzuki Y, Noguchi N, Takasawa S, Kumagai T, Yonekura H, and Okamoto H. Structure, chromosomal localization and expression of mouse genes encoding type III Reg, RegIII{alpha}, RegIIIß, RegIII{gamma}. Gene 185: 159–168, 1997.[CrossRef][ISI][Medline]
  45. Pakala SV, Chivetta M, Kelly CB, and Katz JD. In autoimmune diabetes the transition from benign to pernicious insulitis requires an islet cell response to tumor necrosis factor alpha. J Exp Med 189: 1053–1062, 1999.[Abstract/Free Full Text]
  46. Penkowa M, Espejo C, Martinez-Caceres EM, Poulsen CB, Montalban X, and Hidalgo J. Altered inflammatory response and increased neurodegeneration in metallothionein I+II deficient mice during experimental autoimmune encephalomyelitis. J Neuroimmunol 119: 248–260, 2001.[CrossRef][ISI][Medline]
  47. Prochazka M, Gaskins HR, Shultz LD, and Leiter EH. The nonobese diabetic scid mouse: model for spontaneous thymomagenesis associated with immunodeficiency. Proc Natl Acad Sci USA 89: 3290–3294, 1992.[Abstract/Free Full Text]
  48. Stenina OI, Krukovets I, Wang K, Zhou Z, Forudi F, Penn MS, Topol EJ, and Plow EF. Increased expression of thrombospondin-1 in vessel wall of diabetic Zucker rat. Circulation 107: 3209–3215, 2003.[Abstract/Free Full Text]
  49. Sussel L, Kalamaras J, Hartigan-O’Connor DJ, Meneses JJ, Pedersen RA, Rubenstein JL, and German MS. Mice lacking the homeodomain transcription factor Nkx2.2 have diabetes due to arrested differentiation of pancreatic beta cells. Development 125: 2213–2221, 1998.[Abstract/Free Full Text]
  50. Szabo SJ, Kim ST, Costa GL, Zhang X, Fathman CG, and Glimcher LH. A novel transcription factor, T-bet, directs Th1 lineage commitment. Cell 100: 655–669, 2000.[CrossRef][ISI][Medline]
  51. Terazono K, Yamamoto H, Takasawa S, Shiga K, Yonemura Y, Tochino Y, and Okamoto H. A novel gene activated in regenerating islets. J Biol Chem 263: 2111–2114, 1988.[Abstract/Free Full Text]
  52. Tisch R and McDevitt H. Insulin-dependent diabetes mellitus. Cell 85: 291–297, 1996.[CrossRef][ISI][Medline]
  53. Tisch R, Yang XD, Singer SM, Liblau RS, Fugger L, and McDevitt HO. Immune response to glutamic acid decarboxylase correlates with insulitis in non-obese diabetic mice. Nature 366: 72–75, 1993.[CrossRef][ISI][Medline]
  54. Todd JA and Wicker LS. Genetic protection from the inflammatory disease type 1 diabetes in humans and animal models. Immunity 15: 387–395, 2001.[CrossRef][ISI][Medline]
  55. Tsai EY, Falvo JV, Tsytsykova AV, Barczak AK, Reimold AM, Glimcher LH, Fenton MJ, Gordon DC, Dunn IF, and Goldfeld AE. A lipopolysaccharide-specific enhancer complex involving Ets, Elk-1, Sp1, and CREB binding protein and p300 is recruited to the tumor necrosis factor alpha promoter in vivo. Mol Cell Biol 20: 6084–6094, 2000.[Abstract/Free Full Text]
  56. Tsukamoto K, Yokono K, Amano K, Nagata M, Yagi N, Tominaga Y, Moriyama H, Miki M, Okamoto N, Yoneda R, Inoue Y, Yagita H, Okumura K, and Kasuga M. Administration of monoclonal antibodies against vascular cell adhesion molecule-1/very late antigen-4 abrogates predisposing autoimmune diabetes in NOD mice. Cell Immunol 165: 193–201, 1995.[CrossRef][ISI][Medline]
  57. Ukkola O, Erkkila PH, Savolainen MJ, and Kesaniemi YA. Lack of association between polymorphisms of catalase, copper-zinc superoxide dismutase (SOD), extracellular SOD and endothelial nitric oxide synthase genes and macroangiopathy in patients with type 2 diabetes mellitus. J Intern Med 249: 451–459, 2001.[CrossRef][ISI][Medline]
  58. Unno M, Nata K, Noguchi N, Narushima Y, Akiyama T, Ikeda T, Nakagawa K, Takasawa S, and Okamoto H. Production and characterization of Reg knockout mice: reduced proliferation of pancreatic ß-cells in Reg knockout mice. Diabetes 51, Suppl 3: S478–S483, 2002.
  59. Vallejo AN, Mugge LO, Klimiuk PA, Weyand CM, and Goronzy JJ. Central role of thrombospondin-1 in the activation and clonal expansion of inflammatory T cells. J Immunol 164: 2947–2954, 2000.[Abstract/Free Full Text]
  60. Van Bezooijen RL, Van Der Wee-Pals L, Papapoulos SE, and Lowik CW. Interleukin 17 synergises with tumour necrosis factor alpha to induce cartilage destruction in vitro. Ann Rheum Dis 61: 870–876, 2002.[Abstract/Free Full Text]
  61. Ward SG and Westwick J. Chemokines: understanding their role in T-lymphocyte biology. Biochem J 333: 457–470, 1998.[ISI][Medline]
  62. Webster NJ, Kong Y, Cameron KE, and Resnik JL. An upstream element from the human insulin receptor gene promoter contains binding sites for C/EBP beta and NF-1. Diabetes 43: 305–312, 1994.[Abstract]
  63. Wingender E, Chen X, Hehl R, Karas H, Liebich I, Matys V, Meinhardt T, Pruss M, Reuter I, and Schacherer F. TRANSFAC: an integrated system for gene expression regulation. Nucleic Acids Res 28: 316–319, 2000.[Abstract/Free Full Text]
  64. Yan SF, Lu J, Zou YS, Soh-Won J, Cohen DM, Buttrick PM, Cooper DR, Steinberg SF, Mackman N, Pinsky DJ, and Stern DM. Hypoxia-associated induction of early growth response-1 gene expression. J Biol Chem 274: 15030–15040, 1999.[Abstract/Free Full Text]
  65. Yang XD, Michie SA, Mebius RE, Tisch R, Weissman I, and McDevitt HO. The role of cell adhesion molecules in the development of IDDM: implications for pathogenesis and therapy. Diabetes 45: 705–710, 1996.[Abstract]
  66. Yao Z, Fanslow WC, Seldin MF, Rousseau AM, Painter SL, Comeau MR, Cohen JI, and Spriggs MK. Herpesvirus Saimiri encodes a new cytokine, IL-17, which binds to a novel cytokine receptor. Immunity 3: 811–821, 1995.[CrossRef][ISI][Medline]
  67. Zimmer Y, Milo-Landesman D, Svetlanov A, and Efrat S. Genes induced by growth arrest in a pancreatic beta cell line: identification by analysis of cDNA arrays. FEBS Lett 457: 65–70, 1999.[CrossRef][ISI][Medline]