Introduction

The NOD mouse, which spontaneously develops diabetes, is an important model of type 1 diabetes. Since its development, more than 30 years ago, this strain has provided a wealth of information on the development of this complex autoimmune disease [1]. In the prediabetic phase, islets become infiltrated by macrophages and dendritic cells, followed by CD4+ and CD8+ T cells. This process, known as insulitis, starts at about 4 weeks of age, resulting in diabetes onset at 12–14 weeks of age in about 60–80% of female and 10–30% of male NOD mice [2]. The most important type 1 diabetes susceptibility genes are the MHC genes, in particular MHC Class-II [3, 4]. In addition, more than 40 non-MHC loci have been identified as contributors to disease susceptibility [5, 6]. Congenic non-obese-resistant (NOR) mice on the other hand, do not develop diabetes despite sharing 88% of their genome with NOD mice, including the MHC Class-II haplotype H2g7 and other Idd susceptibility genes [7].

The NOD mouse has been used worldwide to investigate the genes, proteins or pathways implicated in type 1 diabetes susceptibility, with a main focus on the role of the immune system [8–12]. However, increasing evidence points towards a role for the beta cells themselves in the initiation of the autoimmune process and attraction/activation of immune cells; however, the exact mechanisms involved remain unclear. Recently, post-translational modifications (PTMs) have been suggested as a mechanism for the generation of auto-antigenic epitopes in type 1 diabetes [13, 14], as also observed in other autoimmune diseases [14–17]. The aim of this study was to investigate the gene and protein landscape of islets from 2–3-week-old NOD mice compared with islets from NOR and C57Bl/6 mice, with special attention on the presence of PTMs [2].

Materials and methods

Animals

C57Bl/6 mice were obtained from Harlan (Horst, the Netherlands). NOR mice were obtained from Jackson Laboratory (Bar Harbor, ME, USA). NOD mice have been inbred in our animal facility under semi-barrier conditions since 1989. One-week-old and 2–3-week-old mice from mothers that were not diabetic during pregnancy or weaning were used. All animal manipulations were in compliance with the principles of laboratory care and were approved by the Institutional Animal Ethics Committee of KU Leuven.

Islets

Islets were isolated as described previously [18]. Briefly, pancreases from ten mice were digested with collagenase and the islets were centrifuged on a dextran gradient and hand-picked to remove exocrine tissue.

RNA isolation and microarray

Total RNA from islets of 1-week-old and 2–3-week-old NOD, NOR and C57Bl/6 mice (150 islets per extraction, n = 4) was extracted using the RNeasy Micro Kit (Qiagen, Hilden, Germany). Starting from 100 ng of islet RNA, sense-strand DNA was generated and gene expression levels were analysed on Affymetrix GeneChip Mouse Gene 1.0 ST Arrays (for full details see ESM Methods)

Quantitative RT-PCR

(qRT-PCR) was performed as described previously [19]. Expression levels of Angptl7, Dpt, Tmem45a, Trnp1, Lbp, Trim12a, Pgap2, Akr1e1, Dio1, Dock10, Vps13d, Lyrm7 and Padi2 were analysed in islets from 1-week-old and 2–3-week-old NOD, NOR and C57Bl/6 mice.

2D-DIGE

Samples of 40 μg protein lysate, obtained from approximately 1000 islets from 2–3-week-old NOD, NOR and C57Bl/6 mice (n = 4), were separated on immobilized pH gradient (IPG) strips in pH range 4–7 (24 cm, GE Healthcare, Machelen, Belgium) (full details are available in ESM Methods).

Protein identification

Spots were picked from preparative gels with 350 μg protein lysate and trypsin digested as described previously [20]. MS analysis was performed by 4800 MALDI-TOF/TOF (Applied Biosystems, Carlsbad, CA, USA) and individual peptides from the MS/MS analysis were manually filtered; those with an individual expected value >0.05 were deleted, as were identifications based on a single peptide. Differentially expressed proteins were linked to Idd loci (see ESM Methods for full details).

Network analysis

Protein–protein interaction (PPI) networks were created for each list of differentially expressed proteins and first-order neighbours using InWeb [21]. The networks were visualised in Cytoscape [22]. Ranking of the differentially expressed mouse proteins based on their assignment to type 1 diabetes relevant protein complexes was performed as described previously [23]. PANTHER was used to classify genes and proteins by biological processes [24]. Gene ontology (GO) enrichment analysis was performed by AmiGO [25] (full details of these analyses are described in ESM Methods). To identify altered pathways, genes/proteins that were differentially expressed or interacted in a network were loaded into the IPA software and database (Ingenuity Systems, www.ingenuity.com, accessed 1 April 2016). The Mouse Gene 1.0 ST Array reference set was used and all tissues and cell lines were included for analysis.

Statistical analysis

Microarray differential expression was calculated using a significance analysis of microarrays (SAM) implemented in the SAMR package. Cut-off values were set on false discovery rate (FDR; q value) 0.01. 2D-DIGE analysis was performed by DeCyder software (version 7.2.1.72) and p ≤ 0.05 was considered significant. The significance of the overlap between differentially expressed genes and proteins, as well as the overlaps between genes and proteins found to be differentially expressed in either NOR or C57Bl/6 islets when compared with NOD islets, was evaluated using a hypergeometric test, considering all protein-coding genes in the mouse genome (GRCm38.p4) as background. See ESM Methods for details. Data are expressed as means ± SEM and were analysed as stated in figure legends.

Results

Gene expression profiling in pancreatic islets of prediabetic 2–3-week-old NOD, NOR and C57Bl/6 mice

To identify early differences in islets of NOD mice, we performed microarray analyses on islets of 2–3-week-old NOD mice and compared the transcriptome profile to islets of age-matched NOR and C57Bl/6 mice. First, we compared the gene expression between female and male NOD islets (ESM Table 1). Since only five genes, all X or Y chromosome linked, were different between both sexes, we decided to focus on islets from female mice only for further investigation. This revealed 213 differentially expressed genes, out of a total of 35,556 genes present on the microarray, between NOD and NOR islets. Of these, 75 had higher and 138 had lower expression levels in NOD islets (q < 0.01). Comparison of NOD with C57Bl/6 islets, revealed a difference in 700 genes (q < 0.01); of which, 212 had higher and 488 had lower expression levels in NOD islets (Table 1). Of those, 53 transcripts were differentially regulated when comparing NOD vs NOR and C57Bl/6 (p < 1 × 10−10) (Fig. 1a, ESM Table 2) No evidence was found for increased expression of IL-1β, IFNγ, TNFα or IL-6 in NOD islets (data not shown), confirming the absence of inflammation in the islets at the time of investigation.

Table 1 Most differentially expressed genes between islets from 2–3-week-old NOD vs NOR and C57Bl/6 mice by microarray
Full size table
Fig. 1
figure 1

Overlap between NOD vs NOR (pink) and NOD vs C57Bl/6 (blue) differentially expressed (a) genes and (c) proteins and respective networks (b) and (d). Overlap, p < 0.05

Full size image

Considering the top five differentially expressed genes in NOD vs NOR islets and NOD vs C57Bl/6 islets, Padi2 was one of the highest ranked, with a 3.40- and 3.09-fold higher expression in NOD islets compared with NOR and C57Bl/6, respectively (Table 1). mRNA expression of Padi2 in age-matched NOD.scid islets revealed a similar expression level to NOD, suggesting that transcription happens in endocrine and not immune cells (Fig. 2). qRT-PCR also confirmed the expression levels of other differentially expressed genes at 2–3 weeks of age (ESM Fig. 1), as well as in 1-week-old mice (ESM Fig. 2).

Fig. 2
figure 2

Padi2 mRNA is highly expressed in 2–3-week-old NOD and NOD.scid islets compared with C57Bl/6 and NOR islets. Statistical analysis was performed by one-way ANOVA. n = 4–10; ***p < 0.001; ****p < 0.0001

Full size image

GO classification of the differentially expressed genes in NOD vs NOR and C57Bl/6 islets (ESM Tables 3, 4) revealed a prominent prevalence of transcripts implicated in biological pathways related to metabolic processes (50.0% and 45.1% of all differentially expressed transcripts compared with NOR or C57Bl/6, respectively), especially primary metabolic processes, and cellular processes (39.7% and 39.3% compared with NOR or C57Bl/6, respectively), with the majority involved in cell communication. Genes that are differentially expressed in NOD islets compared with both NOR and C57Bl/6 represented the same groups (ESM Table 5). Further enrichment analysis of the differential genes in NOD vs C57Bl/6 islets (ESM Table 6) highlighted genes associated with carbohydrate derivative transport (7.79-fold; p = 2.71 × 10−2), response to metal ions (3.38-fold; p = 2.84 × 10−2) and regulation of protein kinase activity (2.49-fold enriched compared with C57Bl/6; p = 4.47 × 10−4) (Fig. 3), with the expression of the majority of genes being lower in NOD islets.

Fig. 3
figure 3

GO classification of differentially expressed genes in NOD vs C57Bl/6 islets showed enrichment of the biological pathways related to (a) carbohydrate derivative transport (GO: 1901264) (7.79-fold enriched; p = 2.71 × 10−2), (b) response to metal ions (GO: 0010038) (3.38-fold enriched; p = 2.84 × 10−2) and (c) regulation of protein kinase activity (GO: 0045859) (2.49-fold enriched; p = 4.47 × 10−4). Genes that are linked to these classes are shown and expression levels in NOD vs C57Bl/6 islets are represented (>1.3-fold higher, green; 1.3–2.5-fold lower, orange; >2.5-fold lower, red). Analysis performed by AmiGO

Full size image

To investigate how the differentially expressed genes in NOD islets connect to each other, PPI network analysis was performed. Among the 213 differentially expressed genes in NOD vs NOR islets, 101 genes clustered together within the PPI network and formed a significant (p = 3.79 × 10−6) subnetwork with inclusion of first-order interaction partners, resulting in a total of 497 genes with 702 interactions. The genes differentially expressed in NOD vs C57Bl/6 islets were significantly connected (p = 2.42 × 10−3), in a subnetwork containing 363 input genes, extended to 1408 genes when including first-order interaction partners and containing 3298 interactions in between. Furthermore, as for the differentially expressed genes, there was a significant overlap of 88 genes when comparing the NOD vs NOR and NOD vs C57Bl/6 networks (p = 1.38 × 10−8) (Fig. 1b).

Ingenuity pathway analysis (IPA) of the identified PPI networks revealed that both in the NOD vs NOR and NOD vs C57Bl/6 network (ESM Tables 7, 8), genes related to endocrine system development were highly represented. In addition, the NOD vs C57Bl/6 network was enriched for genes functioning in carbohydrate metabolism, as well as genes related to cellular movement, cell death and survival. In general, the functional networks identified by IPA were clearly related to the ontological classes that were assigned to the differentially expressed genes by PANTHER and AmiGO, as described above. When evaluating potential upstream regulators by IPA, a significant number of genes were linked to predicted lower expression of Hnf1a in NOD islets compared with both control groups (Fig. 4a,b), which was confirmed by qRT-PCR (Fig. 4c).

Fig. 4
figure 4

Expression of Hnf1a is predicted to be inhibited in (a) NOD vs NOR and (b) NOD vs C57Bl/6 islets by IPA, based on the expression levels of downstream differentially expressed genes. The genes and arrows are coloured according to expression levels, confidence and predicted relationship. (c) Lower mRNA expression in NOD islets was confirmed by qRT-PCR. Statistical analysis was performed by one-way ANOVA. n = 4; *p < 0.05

Full size image

Proteomic profiling in pancreatic islets of prediabetic 2–3-week-old NOD vs NOR and C57Bl/6 mice

In parallel to the microarray analysis, differences in the proteome of NOD vs NOR and C57Bl/6 islets were investigated by 2D-DIGE. Of the 2141 ± 186 spots detected, 124 spots showed a differential expression between at least two groups (n = 4, p < 0.05) (Fig. 5). Of these, 89 unique proteins were identified (identification rate 65%). Similar to the transcriptome analysis, most significant differences were observed between NOD vs C57Bl/6 islets (100 protein spots, 45 proteins identified), while the islet-proteome of the congenic NOR mice only had 39 differential protein spots (19 proteins identified) (Table 2). Eleven proteins were differentially expressed in NOD compared with both C57Bl/6 and NOR islets (p < 1 × 10−10) (Fig. 1c, ESM Table 9)

Fig. 5
figure 5

2D-gel image with indication of differentially expressed protein spots in NOD vs NOR (n = 39) and NOD vs C57Bl/6 islets (n = 100). Analysis performed by Decyder version 7.2.1.72. n = 4; p < 0.05

Full size image
Table 2 Differentially expressed identified proteins between islets from 2–3-week-old NOD vs NOR and NOD vs C57Bl/6 mice by 2D-DIGE and MALDI-TOF/TOF
Full size table

GO classification of the differentially expressed proteins in NOD compared with NOR and C57Bl/6 islets demonstrated a high prevalence of the same biological processes as in the transcriptome analysis, namely metabolic processes (60.9% and 52.6%, respectively) and cellular processes (30.4% and 42.1%, respectively). Furthermore, the majority of proteins that were shared by NOR and C57Bl/6 islets, but different in NOD, were related to metabolic (50%) and cellular processes (62.5%). Although the data set of differentially expressed proteins is relatively small, significant enrichment of proteins implicated in protein folding (p = 3.65 × 10−4), more specifically in positive regulation of protein folding (p = 2.90 × 10−4), was differential in NOD vs C57Bl/6 islets (Fig. 6). In the latter group, the expression of protein disulfide-isomerase A3 (PDIA3), PDIA4 and heat shock cognate 71 kDa protein (HSP7C) was lower in NOD islets.

Fig. 6
figure 6

GO classification of differentially expressed proteins in NOD vs C57Bl/6 islets showed enrichment of the biological pathways related to protein folding (GO: 0006457) (p = 3.65 × 10−4) and more specifically positive regulation of protein folding (GO: 1903334) (p = 2.90 × 10−4). Proteins linked to these classes are shown and expression levels in NOD compared with C57Bl/6 islets are presented (higher expression, green; lower expression, red). The label shape indicates the presence in multiple isoforms (octagon, same regulation; diamond, differential regulation). GO analysis performed by AmiGO

Full size image

PPI networks were generated based on differentially expressed proteins in NOD islets and first-order interaction partners, similarly as for the transcriptome analysis. Out of the 39 differentially expressed proteins between NOD and NOR islets, 13 formed a significant (p = 1.10 × 10−5) network, which was increased to 71 proteins and 80 internal interactions when including first-order interaction partners. Twenty-nine differentially expressed proteins between NOD and C57Bl/6 formed a connected network (p = 4.96 × 10−6), which increased to 236 proteins and 436 internal interactions including the first-order interaction partners. Comparison of the two networks revealed 37 shared proteins (p < 10 × 10−10) (Fig. 1d); of which, most were linked to a subnetwork clustered around PDIA3, ezrin (EZR) and inner membrane protein, mitochondrial (IMMT) (Fig. 7).

Fig. 7
figure 7

Detail of the PPI network of differentially expressed proteins (yellow border) and first-order interaction partners (blue border) in (a) NOD vs NOR and (b) NOD vs C57Bl/6 islets. Proteins that overlap between NOD vs NOR and NOD vs C57Bl/6 network are shown in purple. Proteins with differential mRNA expression are shown in green. The label shape indicates the presence in multiple isoforms (octagon, same regulation; diamond, differential regulation)

Full size image

2D-DIGE analysis revealed that 29% (26/89) of the identified islet proteins were present in multiple isoforms. Seven proteins showed a shift in abundance between two isoforms between NOD and C57Bl/6 mice (Table 3). Among these were two chaperones, the endoplasmic reticulum (ER) chaperone glucose-regulated protein of 78 kDa (GRP78) and the mitochondrial stress protein (GRP75). Three of the PTM proteins were enzymes, namely hydrolase 3′(2′),5′-bisphosphate nucleotidase 1 (BPNT1), mitochondrial phosphoenolpyruvate carboxykinase [GTP] (PCKGM) with a role in gluconeogenesis and arsenite methyltransferase (AS3MT). The latter was present in four different isoforms; of which, expression in NOD was higher in one and lower in three compared with C57Bl/6 islets. Annexin A5 (ANXA5) and UDP-N-acetylhexosamine pyrophosphorylase-like protein 1 (UAP1L) were both present in two isoforms; of which, one was more and one less abundant in NOD compared with C57Bl/6 islets.

Table 3 Differentially regulated PTM proteins between islets from NOD vs C57Bl/6 islets by 2D-DIGE
Full size table

A challenge when analysing the differences between NOD and healthy control islets is to identify the proteins that are relevant to the disease pathogenesis. For this purpose, we performed gene prioritisation by ranking differentially expressed islet proteins according to their potential relevance to type 1 diabetes based on text mining of biomedical records from the OMIM and PubMed databases (Tables 4, 5). This pointed to an important role for PDIAs. PDIA3, which was differentially expressed both in NOD vs NOR and NOD vs C57Bl/6, was the highest ranked protein with regard to type 1 diabetes relevance in both comparisons. In addition, PDIA4, which was only significantly differentially expressed between NOD vs C57Bl/6, was also ranked in the top ten in this comparison. Finally, PDIA6, which had lower expression in NOD compared with both NOR and C57Bl/6 islets, was also retrieved as a relevant candidate protein, ranked in the top ten.

Table 4 Gene prioritisation of differentially expressed proteins between NOD vs NOR islets: ten highest ranked proteins
Full size table
Table 5 Gene prioritisation of differentially expressed proteins between NOD vs C57Bl/6 islets: ten highest ranked proteins
Full size table

Another highly ranked protein involved in protein metabolism was neuroendocrine convertase 2 (NEC2). This endopeptidase, mediating the conversion of proinsulin to insulin in beta cells, had one isoform that had significantly higher expression in NOD vs NOR islets, while four isoforms were higher in NOD compared with C57Bl/6 islets (Table 3).

Several cytoskeletal proteins appeared in the top ten proteins associated with type 1 diabetes (Tables 4, 5). As such, tropomyosin alpha-3 chain (TPM3), important for the stabilisation of actin filaments, had lower expression in NOD islets. In contrast, ezrin (EZR), connecting cytoskeleton structures such as actin and microtubules to the plasma membrane, was more highly expressed in NOD islets.

Discussion

In this study, we aimed to identify early differences in islets of NOD mice compared with congenic NOR and wild-type C57Bl/6 mice by the combination of transcriptional and translational analysis. Although several studies have been performed to characterise diabetes predisposing genes and proteins, most of them have focused on the role of the immune system in this process instead of the islets themselves [8–12, 26] or expression levels of only mRNA [9] or proteins [27] were investigated.

For high-throughput analysis of gene expression levels, microarrays are very appropriate because of their high sensitivity and accuracy. However, mRNA levels do not always correlate with respective protein levels, which are much more relevant to the biological function of cells. For that reason, we combined microarray with proteome analysis by 2D-DIGE. Although this technique also has some constraints, such as limited detection of proteins with low abundance, extreme isoelectric point or high hydrophobicity, an enormous advantage of 2D-DIGE is the possibility to detect the occurrence of PTMs even without knowing the nature of the modification. In addition to the combination of both techniques, this study was completed by performing integrated data analyses, making use of PPI networks, pathway analyses by IPA, PANTHER and AmiGO, and in silico gene prioritisation for type 1 diabetes relevance.

Since diabetes incidence is known to be higher in female NOD mice compared with males, microarray was performed to identify sex-differences that contribute to diabetes predisposition. The expression of only five genes, all X or Y chromosome linked, was different between NOD males and females (ESM Table 1). Therefore, only female mice were used for further investigations. C57Bl/6 mice and NOR mice that display insulitis without the development of diabetes were both used as control strains. Analysis of genes and proteins that are differentially expressed in NOD islets compared with both control strains indicated that there were 53 common genes (ESM Table 3) and 11 common proteins (ESM Table 4). Six of these proteins (TPM3, NEC2, AKP0, EZRI, KAD1, PDIA6) were also highly ranked by gene prioritisation, indicating the importance of these genes in relation to type 1 diabetes.

In line with the reported limited correlation between mRNA and protein levels [28], our results show only a minor overlap between differentially expressed transcripts and proteins. However, when performing more integrated pathway and network analyses on the microarray and proteomics data, the overall groups of biological processes enriched in the differentially expressed genes and proteins in NOD islets were remarkably similar for both comparisons. A major role was premised for proteins related to metabolic and cellular processes. Primary metabolic processes are essential for the normal anabolic and catabolic pathways such as carbohydrate, lipid and protein metabolism. A defect in genes responsible for the supply of carbohydrates, needed for optimal energy production in islets, could be one of the predisposing factors for diabetes development in NOD mice. It has indeed been described that changes in metabolic demands precede type 1 diabetes, both in humans and NOD mice [29]. Concerning the group related to cellular processes, mainly genes/proteins involved in cell communication, genes functioning as cell surface receptors for cytokines and growth factors, protein kinases and proteins involved in the secretion machinery were altered in NOD islets.

Enrichment analysis of the differential genes in NOD vs C57Bl/6 islets revealed dysregulation of genes related to metal ion transport, in line with recent studies showing that pancreatic changes in Zn2+ levels influence the availability and action of insulin [30, 31]. In addition, genes implicated in the regulation of protein kinase activity were affected in NOD islets, which could lead to disturbances in the regulation of several molecular processes. As such, cGMP-specific 3′,5′-cyclic phosphodiesterase (PDE5A), catalysing the hydrolysis of cGMP to 5′-GMP, was found to be lower expressed in NOD islets. Since it was shown that inhibition of this enzyme potentiates beta cell death, a similar effect is expected in NOD islets [32]. Furthermore, lower expression of the Wnt signalling pathway modulator Sfrp5, as observed in NOD islets in the present study, has been described to improve insulin sensitivity but impair beta cell function [33].

Based on the IPA pathway analysis, we retrieved Hnf1a as an upstream regulator of several differentially expressed genes, which was confirmed by qRT-PCR. Mutations in this gene are known to cause MODY and large-scale genetic studies have shown an association of genetic variants with type 2 diabetes [34]. The relationship with these phenotypes implicates an important role for Hnf1a in beta cells. Coherently, experimental studies showed that this transcription factor controls beta cell function and growth by regulating the gene expression of glucose transporter 2, pyruvate kinase, collectrin, hepatocyte growth factor activator and Hnf4a [34].

PDIA3, 4 and 6 are indicated as proteins that play a crucial role in NOD islets, since they were highly ranked by gene prioritisation and were central in PPI networks. The function of these enzymes, which have lower expression in NOD islets compared with both control mice strains, is to rearrange S-S bonds, making them crucial for correct protein folding. An important role for PDIAs in beta cell functioning has already been described, especially for PDIA6. Together with other chaperones such as GRP78 and calreticulin, PDIA6 is responsible for the correct folding of proinsulin, and silencing of PDIA6 in mouse beta cells reduces insulin production [35, 36]. Furthermore, compared with native proinsulin, binding of PDIA6 is ten times higher to misfolded proinsulin containing the Akita mutation, where it plays a key role in targeting this misfolded protein to the ER degradation pathway [35]. Based on this knowledge, it is conceivable that lower expression of PDIA6 in NOD islets leads to higher levels of unfolded proteins, activation of the unfolded protein response and consequently induction of ER stress. It is well known that beta cell death is associated with increased levels of oxidative, ER and mitochondrial stress, and that stress response genes, such as Chop, Jnk, Xbp1s and Puma, are induced in cytokine-treated beta cells, as well as in isolated islets from prediabetic and diabetic NOD mice [37–39]. However, in our study, the expression of these genes was not increased in islets from 2–3-week-old NOD mice. This suggests that intrinsic defects in proper protein folding, associated with the significantly lower expression of PDIAs in NOD islets vs NOR and C57Bl/6, is one of the first triggers for enhanced stress in beta cells, thereby underlying the high susceptibility for beta cell dysfunction and death. These findings are also in line with a report by Yang et al who compared islets from 3-week-old NOD mice with ALR/Lt mice by 2D-gel analysis, demonstrating lower expression of PDIAs and several heat shock proteins [27]. In addition, PDIAs are also involved in regulation of cytoskeleton organisation, especially by modification of beta-actin [40]. As such, lower expression of PDIAs in NOD islets may be central to a significant amount of differential genes/proteins linked to cytoskeleton and cell communication.

One of the cytoskeleton proteins that shows an altered expression in NOD compared with C57Bl/6 and NOR islets is EZR, which is expressed together with other proteins from the ERM scaffolding protein family in beta cells, namely radixin and moesin. These ERM proteins are activated by phosphorylation, induced by glucose and calcium, leading to cell surface translocation, where they participate in traffic and release of insulin granules [41]. However, in contrast to the high expression observed in NOD islets, islets of diabetic ob/ob mice, a model for type 2 diabetes, are characterised by less active ERM [41]. Despite this, perturbations in the cytoskeleton of beta cells can have crucial implications for proper insulin secretion, and disturbed interaction between beta cells and the environment may lead to impaired beta cell functioning in general.

Of interest, a number of the differentially expressed genes and proteins in NOD islets have already been associated with type 1 diabetes since they map to specific Idd loci (Tables 6, 7), such as PDIA3 and EZR described above. Also the majority of the most differential genes in NOD islets (Table 1), confirmed by qRT-PCR, map to these loci (Table 6).

Table 6 Idd loci localisation of differentially expressed genes in NOD vs NOR and NOD vs C57Bl/6 islets
Full size table
Table 7 Idd loci localisation of differentially expressed proteins in NOD vs NOR and NOD vs C57Bl/6 islets
Full size table

Finally, the most remarkable finding in NOD islets at this early preinsulitic age was the high expression of Padi2 mRNA, confirming our earlier findings by qRT-PCR [42]. This suggests that Padi2 is the diabetes susceptibility gene located in Idd25, a region on Chr4 for which the importance in diabetes development was already suggested by the generation of subcongenic strains between NOD and NOR mice. In favour of this hypothesis, no regulation was observed for Ephb2, the gene closely located to Padi2 in the Idd25 locus [43]. As we suspect only very low or no immune cells in the islets at the investigated age, the expression of Padi2 in endocrine cells was supported by similar expression levels in NOD.scid islets, further emphasising the importance of citrullination in NOD mice [39]. Previous results from our group showed that the ER chaperone GRP78 is citrullinated in beta cells upon inflammatory stress and that citrullinated GRP78 is an auto-antigen in NOD mice. Results from the 2D-DIGE analysis in the present study further highlight the presence of several proteins which are PTM modified in NOD islets, a finding that has not been reported by an earlier proteomic study on NOD islets [27]. The presence of different modified isoforms may point to a role for citrullination, not only for GRP78 but also for other islet proteins.

In conclusion, we have shown that islets of preinsulitic NOD mice already have a significantly differential mRNA and protein expression profile compared with control NOR and C57Bl/6 islets. An important role for PDIAs was suggested in NOD islets, where low expression of these chaperones required for S-S bond formation and crucial for correct insulin folding can lead to accumulation of unfolded proteins, activation of the unfolded protein response and generation of ER stress in beta cells. Additionally, high expression of Padi2 mRNA, coding for the enzyme responsible for citrullination, together with the presence of several proteins in multiple isoforms, indicates that PTMs in beta cells are of high importance for type 1 diabetes susceptibility. Modifications can affect protein function but also create neo-epitopes that can be recognised as auto-antigens. In general, the findings from the present study point towards a crucial role of the beta cell itself, independent of differences present at the level of the immune system, in the susceptibility and initiation of type 1 diabetes.