Identification of a Serum-Induced Transcriptional Signature Associated With Type 1 Diabetes in the BioBreeding Rat

Brown Norway (BN) and BB rats were propagated as described (5,6). Before euthanization, animals were fasted for 16 h, and those possessing blood glucose levels >130 mg/dl were disqualified from studies of pre-diabetes. Rats were anesthetized under isofluorane, and blood was collected by heart puncture. Serum was separated by centrifugation then stored at −80°C until use.

Human recombinant IL-1Ra (hIL-1Ra) (350 μg/kg/day) or vehicle (saline) was delivered to DRlyp/lyp rats by intraperitoneal injection. Dosage was based upon previous hIL-1Ra administration to rodents (15,16). Treatment was initiated by day 30 and continued either through day 40 or through diabetes onset. Diabetes onset was defined as the first of two consecutive blood glucose measurements >250 mg/dl. Survival was analyzed with the Kaplan-Meier method. All procedures were approved by the Medical College of Wisconsin Institutional Animal Care and Use Committee.

Fresh PBMCs of healthy BN rats (∼180-day-old males, to avoid variation introduced by estrous or pubertal status) were isolated by density gradient centrifugation. As described (13), transcription was induced by culturing PBMCs for 6 h at 37°C in 5% CO₂ with 20% autologous BN (self-baseline control), allogeneic BN (healthy unrelated control), DRlyp/lyp, or DR+/+ serum. Indicated cultures were supplemented with 1 ng/ml rat IL-1β or 1 μg/ml rat IL-1Ra. RNA was extracted using TRIzol reagent, amplified/labeled using the express kit (Affymetrix, Santa Clara, CA), and hybridized to the Affymetrix Rat Genome 230 2.0 Array. RNA from each culture was independently analyzed. Image data were quantified with Affymetrix Expression Console Software and normalized with Robust Multichip Analysis (www.bioconductor.org/) to determine signal log ratios. ANOVA was conducted and false discovery rates (FDRs) were determined using Partek Genomics Suite version 6.2. To capture the most reliable data, limit the length of gene lists, and facilitate focused pathway analyses, differentially expressed probe sets were defined as those possessing an FDR <10% and |log₂ ratio| >0.5 between the compared groups. Ontological analysis was performed with the Database for Annotation, Visualization, and Integrated Discovery (DAVID) (17). Hierarchical clustering was conducted with Genesis (18). Orthologous rat and human probe sets were mapped using the Affymetrix support document (HG-U133_Plus_2.na29.orthologue.csv.zip). All raw data files have been deposited in The National Center for Biotechnology Information Gene Expression Omnibus (accession no. GSE19537).

Sera of day 60 BB rats and day 180 BN rats were assayed with the BeadLyte cytokine assay kit (Millipore, Billerica, MA), per the manufacturer's protocol, and a Bio-Plex Luminex 100 XYP instrument. Concentrations were calculated with the Bio-Plex Manager 4.1 software and a five-parameter curve-fitting algorithm applied for standard curve calculations. IL-33 levels in sera and IL-1β/-α levels in conditioned culture medium were measured with Quantikine enzyme-linked immunosorbent assays (ELISAs) (R&D Systems, Minneapolis, MN), and gram-negative bacterial endotoxins (lipopolysaccarides [LPS]) were measured by the Limulus amoebocyte lysate assay (Associates of Cape Cod, Falmouth, MA).

hIL-1Ra (17 kDa) was electrophoresed through polyacrylamide gels (10–20% Tris-HCl) and transferred to polyvinylidene difluoride membranes. Membranes were blocked for 1 h with 5% BSA and 0.1% Tween-20 at room temperature then probed with a 1:2,000 dilution of sera from hIL-1Ra or saline-treated DRlyp/lyp rats overnight at 4^ο C. Blots were washed three times with 1 × TBS with 0.1% Tween 20 then incubated with a 1:2,000 dilution of anti-rat IgG, horseradish peroxidase–linked antibody in blocking solution for 1 h at room temperature. After three washes, the Enhanced Chemiluminescence (ECL) Plus System (Amersham, Piscataway, NJ) was used for visualization.

To determine whether inflammatory mediators related to diabetogenesis could be detected through their ability to induce transcription, PBMCs of healthy, day 180 BN rats were cultured with day 60 DRlyp/lyp (representing the period immediately before onset), day 60 DR+/+, allogeneic BN, or autologous BN serum. To ensure induced transcription was not related to elevated blood glucose levels, only sera of normoglycemic, preonset rats were used. In preliminary studies, we observed that day 60 DRlyp/lyp serum, like human recent-onset sera, induced many genes regulated by IL-1. Therefore, PBMCs were also cultured with day 60 DRlyp/lyp serum supplemented with IL-1Ra as well as autologous BN serum supplemented with IL-1β, to respectively block or induce IL-1–mediated gene expression. Culture of BN PBMCs with DRlyp/lyp or DR+/+ sera regulated, within the threshold values, 1,979 and 2,904 probe sets, respectively, relative to culture with allogeneic BN sera.

As reflected in the Venn diagram illustrated in Fig. 1,A, culture of BN PBMCs with either DRlyp/lyp or DR+/+ sera regulated a total of 3,971 probe sets (supplement A Fig. 1,A, available at http://diabetes.diabetesjournals.org/cgi/content/full/db10-0372/DC1). Consistent with type 1 diabetes susceptibility in both strains, the signatures share a significantly nonrandom (P < 10⁻⁶, χ² test), commonly regulated intersection of 912 probe sets that represents 46.1 and 31.4% of the probe sets induced by either DRlyp/lyp or DR+/+ serum, respectively. Hierarchical clustering (Fig. 1,B) shows the relatedness of the five experimental conditions for this subset. These 912 shared probe sets were annotated with DAVID, which identified significantly regulated gene ontology biological processes related to immunological activation, antigen presentation, and intracellular signaling, in particular through the nuclear factor (NF) κB pathway (Table 1 and supplement B).

NFκB is a transcriptional regulator of innate and adaptive immunity (19) that is activated through events such as binding of IL-1 to IL-1R1 or LPS to toll-like receptor (TLR) 4. These events converge at a set of I-κB kinases that phosphorylate the inhibitory I-κB proteins (NFKBIA and NFKBIB) leading to their degradation, allowing release and nuclear translocation of cytoplasmic NFκB, where it facilitates target gene transcription. Both DRlyp/lyp and DR+/+ sera induced Nfkb2, which encodes the p100 precursor that is cleaved to produce the NFκB p52 protein; the receptor-interacting serine-threonine kinase 2 (Ripk2), a potent activator of NFκB; as well as tumor necrosis factor (TNF) receptor superfamily member 25 (Tnfrsf25), which stimulates NFκB activity (20). Sera of either BB rat induced transcription of genes related to intracellular signaling, including signal transducer and activator of transcription 2 (Stat2) and mitogen-activated protein kinase activated protein kinase 5 (Mapkapk5), which is activated by MAPKs in response to exposure to proinflammatory cytokines. DRlyp/lyp and DR+/+ sera also induced transcription of genes encoding chemokines (Ccl2, Cxcl16) and receptors including Tlr4, Tlr7, Ccr1, and Cd14, which are important in TLR4 signaling. Other genes related to immune function included IL-1R–associated kinase 3 (Irak3) and intercellular adhesion molecule 1 (Icam1). Both DRlyp/lyp and DR+/+ sera induced the transcription factor Kruppel-like factor 2 (Klf2), an inflammatory inhibitor that functions through the binding of coactivators necessary for optimal NFκB activity (21), and B-cell CLL/lymphoma three (BclIII), an IL-10–inducible gene that impairs binding of NFκB complexes to DNA (19,22). Importantly, all 912 probe sets of the intersection exhibited directionally concordant inductions by either DRlyp/lyp or DR+/+ sera relative to allogeneic BN sera. Furthermore, when directly comparing expression levels between the DRlyp/lyp and DR+/+, only two probes sets (2 of 912, 0.2%) were differentially expressed (Klf4 and Map4k2, supplement A). Overall, analysis of the intersection revealed that culturing BN PBMCs with either DR+/+ or DRlyp/lyp sera, compared with allogeneic BN sera, induced transcription consistent with immunological activation (Fig. 1 C).

To reveal processes associated with progression to type 1 diabetes, the DRlyp/lyp and DR+/+ serum signatures were independently analyzed. The mean expression from each of the five experimental conditions was subjected to hierarchical clustering using the 1,979 or 2,904 probe sets respectively regulated by DRlyp/lyp sera or DR+/+ sera relative to the BN allogeneic induction (Figs. 2,A and C). Ontological analysis of the DRlyp/lyp signature identified biological processes related to proinflammatory processes possessing higher numbers of associated genes with greater significance compared with the analyses of the DRlyp/lyp:DR+/+ intersection or the DR+/+ signature (Table 1 and supplement B). Specifically, DRlyp/lyp sera regulated biological processes related to antigen presentation, inflammation, and leukocyte migration. Annotated genes related to these functions included induction of Il1b, Tlr2, Il18, Icam2, proteasome subunit β type 1, (Psmb1), Stat1, and numerous MHC class II genes. Among the 1,979 probe sets regulated by DRlyp/lyp sera were central components of the NFκB cascade, including Nfkbia and Nfkbib, as well as positive regulators of NFκB activity, including casein kinase 2 β subunit (Csnk2b), a potent NFκB activator that directly phosphorylates NFKBIA (23), baculoviral IAP repeat-containing 2 (Birc2) (20), and Cd40 (Fig. 3 A and supplement A).

When examining the 1,067 probe sets uniquely regulated by DRlyp/lyp sera (i.e., those excluding the intersection) (Fig. 2,B), a role for IL-1 in induction of the signature became evident, as blocking IL-1R1 by adding IL-1Ra to cultures modulated the signature, yielding one more similar to that induced by DR+/+ sera. Specifically, after adding IL-1Ra to the culture, expression of 83.3% (889 of 1,067) of the DRlyp/lyp-specific probe sets no longer met the threshold values (P < 10⁻⁶, χ² test). Among 912 genes of the intersection, only 40% (374) were influenced by the addition of IL-1Ra (P < 10⁻⁶, χ² test). When examining 1,992 probe sets uniquely regulated by DR+/+ sera (Fig. 2,D), addition of IL-1Ra to the DRlyp/lyp culture resulted in induction of a signature more similar to that of the DR+/+ culture in terms of fold of change. However, the overall expression levels were less influenced, in that among 1,992 genes regulated in the DR+/+ culture, only 26.2% (521 of 1,992) of the probe sets in the DRlyp/lyp + IL-1Ra culture were regulated to meet the threshold values. These statistical findings are reflected in Figs. 2,B and D and show gene expression arising through IL-1R1 signaling accounts for a large part of the difference between the DRlyp/lyp and DR+/+ signatures. Furthermore, addition of IL-1Ra to DRlyp/lyp cultures reduced gene expression annotated as proinflammatory (Fig. 2,B) and enhanced gene expression annotated as being regulatory (Fig. 2,D). Addition of IL-1β to PBMCs cultured with autologous BN sera induced a modest signature, more similar to that induced by that of BB rat sera than cultures possessing BN allogeneic sera (Fig. 2), regulating 209 of 3,971 probe sets regulated by DR+/+ or DRlyp/lyp serum relative to BN allogeneic serum.

Overall, the ontological analysis of 2,904 probe sets regulated by DR+/+ sera revealed the greatest number of highly significant biological processes, and these annotations were primarily related to regulation of gene expression (Table 1 and supplement B). Numerous genes related to negative regulation of inflammation and NFκB signaling were regulated by DR+/+ sera, including Nfkbie, an inhibitor of NFκB (24); the transcriptional repressors zinc finger E-box binding homeobox 1 (Zeb1); zinc finger and BTB domain containing 7a (Zbtb7a); interferon regulatory factor 2 (Irf2); ETS domain-containing protein Elk-3; the inhibitor of activated STAT 2 (Pias2); suppressor of cytokine signaling-5 (Socs5); the cell cycle inhibitor; myeloid/lymphoid or mixed-lineage leukemia 5 (Mll5); and others (Fig. 3 B).

While the sera of either BB substrain induced gene expression consistent with immune activation, DRlyp/lyp sera induced an inflammatory signature consistent with the presence of IL-1, whereas that of the DR+/+ sera can be characterized as largely immunoregulatory. Among the genes regulated by DRlyp/lyp and DR+/+ sera were 87 orthologues previously identified among 587 genes regulated by human type 1 diabetes sera (Figs. 1 and 3 and supplement A) (13).

In an effort to account for the induced signatures, cytokine levels were measured by ELISA in the DRlyp/lyp, DR+/+, and BN sera used in the expression studies (Table 2). Measurable differences in IL-1α or IL-1β levels between DRlyp/lyp, DR+/+, or BN rat sera were not detected, raising the question whether the induced transcription measured after 6 h of culture was a primary effect of IL-1 on the PBMCs or a secondary effect of IL-1 produced by the PBMCs in response to other dilute mediators in DRlyp/lyp sera. Conditioned medium of cultures supplemented with the various sera were assayed for IL-1α and IL-1β at 0, 1, 3, 6, 12, and 24 h. While IL-1α was not detected at all, significant detectable levels of IL-1β were measured in only DRlyp/lyp cultures at 24 h but not at or prior to 6 h (Table 3). This supports the hypothesis that the 6 h transcriptional response is primary and is due in part to IL-1 itself, as lower IL-1β protein was not significantly detected in the DR+/+ and DRlyp/lyp + IL-1Ra cultures. This parallels the significant detection of Il1b transcript only in cultures supplemented with DRlyp/lyp sera. Significant differences in serum IL-10 were not observed; however, consistent with the transcriptional signatures, levels in DR+/+ and DRlyp/lyp serum were greater than that measured in the BN rat. IL-13 was the only significantly upregulated cytokine detected in serum of the day 60 DRlyp/lyp rat.

Recent studies have identified increased levels of LPS in human type 1 diabetic patients (25). Given that TLR4 and IL-1R1 signaling both are mediated through the adaptor protein MyD88 to activate NFκB, we investigated LPS serum levels. On average, significantly higher LPS levels were detected in the DRlyp/lyp and DR+/+ rat serum compared with that of BN rats (Table 2), consistent with the significant induction of Il1b transcript in cultures possessing with DRlyp/lyp sera even when supplemented with IL-1Ra (supplement A).

Given the parallels between the signatures induced by human type 1 diabetes and DRlyp/lyp sera, in terms of overrepresentation of transcripts regulated by IL-1 and modulation of the DRlyp/lyp signature upon addition of IL-1Ra to cultures, we tested the effectiveness of IL-1Ra in delaying type 1 diabetes onset in BB rats. The treatment of BB rats with rat IL-1Ra is cost prohibitive, while hIL-1Ra shares >70% amino acid homology with rat IL-1Ra and is bioactive in rat (15,26). Thus, we reasoned that if IL-1–mediated inflammation was relevant to diabetogenesis in DRlyp/lyp rats, treatment with hIL-1Ra should delay onset for approximately the amount of time required to mount a robust antibody response (∼10–14 days). Treatment was initiated by day 30 and continued through type 1 diabetes onset. Treated rats survived 71 ± 11 days, whereas controls survived 61 ± 6 days (P = 0.007) (Fig. 4,A). Significant differences in weight or growth rate were not observed between treated and control rats (P > 0.05). As expected, the animals possessed antibody titer to hIL-1Ra (Fig. 4 B).

To determine whether hIL-1Ra treatment of DRlyp/lyp rats would result in modulation of the serum signature, additional rats were treated with hIL-1Ra or saline for 10 days, beginning at day 30. Serum was collected at day 40 and assayed to determine 1) whether an inflammatory signature was present at this time prior to insulitis (27), 2) whether day 40 and day 60 signatures were similar, and 3) whether a day 40 signature was modulated in hIL-1Ra–treated rats. Furthermore, after only 10 days of treatment, the effect of hIL-1Ra treatment would be less likely masked by the high concentrations of neutralizing antibodies observed at onset (Fig. 4 B, lower panels).

Regulated probe sets between cultures possessing serum of saline- or hIL-1Ra–treated DRlyp/lyp rats compared with allogeneic BN serum were identified and subjected to hierarchical clustering (Fig. 4,C and D). In the PBS-treated DRlyp/lyp versus BN allogeneic comparison, 1,526 probe sets met these criteria, and among these, a significant percentage (638 of 1,526, 41.8%; P < 10⁻⁶ χ² test) were also regulated by day 60 DRlyp/lyp sera. Ontological analysis again identified biological processes related to antigen presentation, inflammation, leukocyte migration, and activation of the NFκB cascade (Table 1 and supplement 2). Selected genes related to these pathways are shown (Fig. 4,E, supplement A). Compared with BN sera, sera of day 40 DRlyp/lyp rats treated with hIL-1Ra regulated only five probe sets. The two probe sets of the intersection exhibited directionally concordant inductions by either saline-treated DRlyp/lyp or IL-1Ra–treated DRlyp/lyp sera relative to allogeneic BN sera. As reflected by the fold changes illustrated in Fig. 4 D, treatment reduced induction of the proinflammatory signature observed in PBS-treated rats. Overall, these results show that like human type 1 diabetes (13), a serum-induced signature with identity to that observed at onset is detected prior to onset in the DRlyp/lyp rat. Moreover, this signature is modulated by treating rats with hIL-1Ra, suggesting that this approach may prove useful in monitoring the effect of therapeutic interventions in human type 1 diabetes.

Previously, we defined a transcriptional signature induced by sera of human type 1 diabetic patients (13). Here, parallel studies were conducted that defined unique signatures for diabetes-inducible DR+/+ and spontaneously diabetic DRlyp/lyp rats. Like human type 1 diabetes, the signature associated with disease progression in DRlyp/lyp rats includes many genes regulated by IL-1 and differential regulation of NFκB signaling is a key feature distinguishing the signatures induced by sera of the DRlyp/lyp and DR+/+ substrains.

Identity is observed between signatures induced by day 60 serum of DRlyp/lyp and DR+/+ rats, in particular the 912 probe sets of the DRlyp/lyp:DR+/+ intersection. Identity is also evident in 1,992 probe sets regulated to threshold levels by DR+/+ serum. These probe sets, annotated as immunoregulatory, are regulated by DRlyp/lyp serum to a lesser degree that does not meet threshold values. This suggests the presence of endogenous, albeit insufficient, immunoregulatory activity in DRlyp/lyp rats and may explain why relatively few (10⁶) adoptively transferred DR+/+ T_REG cells prevent type 1 diabetes in DRlyp/lyp rats (8). Conversely, from the perspective of the probe sets regulated by DRlyp/lyp sera, the DRlyp/lyp and DR+/+ signatures are distinct, with the DRlyp/lyp signature possessing an IL-1–driven component that is downregulated by adding IL-1Ra to the culture.

Among human type 1 diabetic patients, a significant percentage will develop autoimmune thyroid disease (28). Likewise, lymphopenic BB rats develop autoimmune thyroiditis. Since the method measures the sum of the inflammatory factors present in serum, we cannot exclude that thyroid autoimmunity has not contributed to the signature defined for DRlyp/lyp rats. In terms of other models, we have examined serum signatures associated with type 1 diabetes in LEW.1WR1 rats (29) and find the inflammatory states distinct but sharing partial identity (supplement C) (30).

Despite the challenges of directly measuring peripheral cytokine levels in human type 1 diabetes, studies have established that a complex cytokine milieu exists. This includes elevated IL-1 family members in patient cohorts before and after disease onset (31,,,–35). Here, ELISA analysis was unable to detect differences in IL-1β/-α levels between DRlyp/lyp, DR+/+ or BN rat sera. This may be related to assay sensitivity, the limited number of rats analyzed, or presence of soluble IL-1 receptors, which impairs detection of free bioactive IL-1 and highlights the need for new, more sensitive biomarkers. Given that the amount of IL-1β (1 ng/ml) spiked into the autologous BN cultures exceeded the sensitivity of the multiplex cytokine analysis for IL-1α (>27 pg/ml) and IL-1β (>27 pg/ml), the signatures induced by DRlyp/lyp and DR+/+ sera (in particular genes of the intersection) are not likely an effect of IL-1 concentrations exclusively but involve the contribution of other inflammatory mediators. Binding of IL-1 and LPS to their respective receptors initiate transcriptional programs similar to those observed. We found LPS levels in BB rats nearly twice that of BN rats, consistent with reported intestinal hyperpermeability in BB rats (36), which may lead to the translocation of bacteria and/or endotoxin and a heightened systemic inflammatory state. IL-13 was elevated in DRlyp/lyp compared with DR+/+ and BN serum. This is consistent with elevated IgE levels and eosinophilia prior to onset in DRlyp/lyp rats (37), as IL-13 induces immunoglobulin isotype switching to IgE in B-cells and regulates eosinophilic inflammation.

In vitro, human and rodent pancreatic β-cells are highly susceptible to the actions of IL-1 (38), and IL-1Ra can protect β-cells from the downstream consequences of IL-1 exposure (39). While IL-1 transcript and protein are detected within immune infiltrated pancreatic islets of BB rats and NOD mice (40,–42), in vivo modulation of either ligand or receptor in these models has yielded mixed results. Administration of high doses of IL-1 to diabetes-prone BB rats induced higher blood glucose concentrations before and at type 1 diabetes onset and accelerated onset (43). In contrast, administration of IL-1 to the NOD mouse delayed onset and reduced incidence (44,45). IL-1R deficiency in the NOD mouse slowed progression to type 1 diabetes but did not prevent it (46), and treatment of NOD mice with IL-1Ra inhibited recurrence of type 1 diabetes after islet transplantation (47). IL-1 has been shown to exacerbate autoimmunity by promoting expansion of effector T-cells and attenuating T_REG cell function (48), raising the possibility of IL-1–mediated impaired regulatory function of T_REG cells in DRlyp/lyp relative to DR+/+ rats. Treatment of DRlyp/lyp rats with hIL-1Ra delayed onset for the time required to mount a neutralizing humoral immune response. These results are consistent with a previous study in which hIL-1Ra delayed BB rat type 1 diabetes onset but did not affect growth or modify lymphocyte counts (16). As reported here, anti–hIL-1Ra antibodies developed, to which short-term protection (2–3 weeks) was attributed. Importantly, with delayed onset we observe modulation of the serum-induced transcriptional profile in hIL-1Ra–treated rats.

These findings support that this bioassay may not only have utility as an early mechanistically informative inflammatory biomarker of type 1 diabetes but may prove useful in monitoring changes in inflammatory state during therapeutic interventions, including those targeting IL-1 (49).

This work was supported by National Institute of Allergy and Infectious Diseases (NIAID) Grants R01AI078713 (to M.J.H.) and P01AI42380 (to Å.L.); National Institute of Diabetes and Digestive and Kidney Diseases Grant R01DK080100 (to X.W.); Juvenile Diabetes Research Foundation International Grant 1-2008-1026 (to M.J.H); Advancing a Healthier Wisconsin Initiative Grant 5520065 (to M.J.H.); and The Children's Hospital of Wisconsin Foundation.

No potential conflicts of interest relevant to this article were reported.

M.K., R.G., J.B., S.P., and S.K. conducted all laboratory studies. S.J. and X.W. conducted all bioinformatic and statistical analyses. Comparative studies of serum signatures between BB and LEW.1WR1 rats were done in collaboration with J.P.M. M.K., S.P., S.K., X.W., and Å.L. reviewed/edited the manuscript. M.J.H designed/directed the study and wrote/reviewed/edited the manuscript.

The authors are grateful to Sanja Glisic, PhD, and Parthav Jailwala, MS, at The Medical College of Wisconsin for their critical review of this manuscript. We thank Shouguo Gao, PhD, at the University of Alabama at Birmingham for helpful discussions on statistical issues. Multiplex cytokine assays were conducted at the Baylor Institute for Immunology Research NIAID Cooperative Centers Luminex Core Facility under the direction of John Connolly, PhD.