Candidate genes for type 1 diabetes modulate pancreatic islet inflammation and β-cell apoptosis


Correspondence to: Dr Izortze Santin or Dr Decio L. Eizirik, Laboratory of Experimental Medicine, Université Libre de Bruxelles, Route de Lennik, 808, CP618, B-1070 Brussels, Belgium.Email: or


Genome-wide association studies (GWAS) have identified more than 50 loci associated with genetic risk of type 1 diabetes (T1D). Several T1D candidate genes have been suggested or identified within these regions, but the molecular mechanisms by which they contribute to insulitis and β-cell destruction remain to be clarified. More than 60% of the T1D candidate genes are expressed in human pancreatic islets, suggesting that they contribute to T1D by regulating at least in part pathogenic mechanisms at the β-cell level. Recent studies by our group indicate that important genetically regulated pathways in β-cells include innate immunity and antiviral activity, involving RIG-like receptors (particularly MDA5) and regulators of type I IFNs (i.e. PTPN2 and USP18), and genes related to β-cell phenotype and susceptibility to pro-apoptotic stimuli (i.e. GLIS3). These observations reinforce the concept that the early pathogenesis of T1D is characterized by a dialogue between the immune system and pancreatic β-cells. This dialogue is probably influenced by polymorphisms in genes expressed at the β-cell and/or immune system level, leading to inadequate responses to environmental cues such as viral infections. Further studies are needed to clarify how these disease-associated variants affect pancreatic β-cell responses to inflammation and the subsequent triggering of autoimmune responses and progressive β-cell loss.


Recent genome-wide association studies (GWAS) have linked >50 genetic variants to susceptibility to type 1 diabetes (T1D), explaining nearly 80% of heritability [1]. On the basis of these observations some basic concepts have emerged on the inheritance of T1D. First, and except for the major histocompatibility complex (MHC) which has an odds ratio (OR) >6.5, most of the genetic loci identified by GWAS involve relatively frequent alleles each conferring a modest risk (OR < 2) for disease [2]. Second, many of the polymorphisms are shared among different autoimmune diseases, suggesting that clues for disease pathogenesis may reside in the function of these shared genes [3, 4]. Third, most of the identified candidate genes are expressed both in immune competent cells [5, 6] and in pancreatic β-cells [7], suggesting a genetically modulated dialogue between these two key components of T1D [8]. Fourth, these genetic studies indicate potential disease-relevant biological pathways that contribute to the pathogenesis of T1D and might be interesting targets for therapy.

The GWAS approach relies on testing a comprehensive catalogue of common (>5%) genetic variants in affected individuals and controls from a population to detect those variants associated with disease [9]. This is a valuable approach, but it has limitations [10]. Among them, the molecular mechanisms by which putative candidate genes modulate disease risk remain poorly characterized and it is often unclear which of the genes (or non-coding regions) in the identified region that are actually linked to disease. Furthermore, single variants that are closely associated with disease and have a small but relevant role in pathogenesis (for instance, by modulating the speed of progression of multiple autoantibody positive individuals to actual diabetes) may account for only a small part of disease risk and fail to be detected [11]. Thus, to clarify how specific genetic backgrounds contribute to β-cell destruction in T1D it is important to understand the combined effects of associated genes and their products in the context of interacting functional pathways.

Recent reviews have addressed the potential role of candidate genes for T1D at the immune system level [12], and we will focus here on emerging information on genetically regulated pathways in pancreatic β-cells. This field is in its early stages, but recent studies by us [7, 13-15] and others [16] suggest that two genetically-regulated pathways play a role in pancreatic β-cells during the early stages of T1D, namely an excessive innate immune activity involving RIG-like receptors (particularly IFIH1/MDA5 [14]) and regulators of type I interferons (i.e. PTPN2 [15] and USP18 [17]), and polymorphisms in genes related to β-cell phenotype and susceptibility to pro-apoptotic stimuli (i.e. GLIS3) [18, 19].

Type 1 Candidate Genes are Expressed in Pancreatic β-cells and Modulated by Inflammatory Mediators

Many of the susceptibility genes for T1D may act through the modulation of immune-related processes, such as antigen presentation, expansion of self-reactive cells and modification/regulation of the immune function [12]. The human leukocyte antigen (HLA) region in chromosome 6p21 explains more than 40% of the genetic risk of T1D, with the HLA class II alleles (HLA-DQ, -DR and -DP) showing the strongest association with the disease [20]. Genetic variations in HLA-DQ molecules affect antigen recognition by T cells [21]. Another candidate gene, the interleukin-2 receptor (IL-2R) [22], is involved in immune regulation and contributes to the immune tolerance mediated by regulatory T cells (Tregs) [23]. PTPN2, a T1D susceptibility gene with an OR around 1.5 [2], is implicated in IL-2R signalling in CD4+ T cells and an intronic risk allele in the gene correlates with impaired IL-2R signalling and defective Treg activation [6], while the candidate gene ERBB3 is highly expressed in monocytes and dendritic cells (DCs) [5] and modulates the function of antigen presenting cells.

Accumulating evidence indicate that the triggering of insulitis and diabetes depends on a dialogue between the invading immune cells and the target β-cells (reviewed in [8]). In line with this, we have recently shown that more than 60% of T1D candidate genes are expressed in pancreatic human islets; expression of many of these genes is modulated by pro-inflammatory cytokines [7]. This indicates that these candidate genes may also contribute to T1D pathogenesis by acting at the β-cell level. The putative functions of the T1D genes expressed in pancreatic islets are diverse, including genes encoding receptors (e.g. ERBB3 and IFIH1), transcription factors (e.g. STAT4 and GLIS3) and kinases/phosphatases (e.g. SKAP2, PTPN2, PTPN22 and UBASH3A), among others (figure 1). Interestingly, there is a preferential expression of T1D candidate genes in human pancreatic islets when compared to genes that are not associated with the disease (figure 2). Thus, RNA sequencing data from human pancreatic islets revealed that the absolute expression levels of T1D candidate genes in human pancreatic islets is around twofold higher than the expression of all genes detected as expressed, of randomly selected groups of 60 genes, of genes related to height and of genes associated with colitis (figure 2).

Figure 1.

Expression of candidate genes for type 1 diabetes (T1D) in pancreatic human islets. Type 1 candidate genes ranked by odds ratio conferred by the risk allele of the polymorphism putatively associated with the disease [2]. Genes expressed in pancreatic islets are shown in blue (median RPKM > 0.5–1; low expression) or red (median RPKM > 1; medium-high expression) bars, while T1D genes that are not expressed in pancreatic islets are represented by black bars (median RPKM < 0.5). The expression of 60% (RPKM > 1) or 69% (RPKM > 0.5) of T1D candidate genes, depending on the selected cut-off, is detected in human pancreatic islets. Gene expression in human islets was determined by RNA sequencing, and the figure is adapted with permission from [7].

Figure 2.

Preferential expression of type 1 diabetes (T1D) candidate genes in human pancreatic islets. The expression level of T1D genes with RPKM > 1 (n = 24) in pancreatic human islets is around two times higher than the expression of all genes (RPKM > 1) detected by RNA sequencing in human islets (n = 23 308) [7], of a randomly selected group of 60 genes (one group is shown, but this was repeated 10 times with similar results), or of genes associated with colitis (n = 184) [96] or height (n = 88) [97] (all with RPKM > 1). *p < 0.05; T1D genes vs. any other group; Fisher's exact test.

Ongoing studies from our group are starting to unveil the putative contribution of these candidate genes at the β-cell level to the pathogenesis of T1D. PTPN2 is highly expressed in rat and human pancreatic islet cells and plays an important role in the modulation of interferon (IFN) signalling pathways in β-cells [13, 15]. Inhibition of PTPN2 expression in β-cells leads to increased signal transducer and activator of transcription (STAT) activation [13] and to exacerbated IFNα/β/γ-induced apoptosis via activation of the BH3-only protein Bim and consequent induction of the mitochondrial pathway of cell death [15]. These data suggest that polymorphisms leading to a decreased expression of PTPN2 may sensitize β-cells to apoptosis after a triggering event (e.g. type I IFNs produced by β-cells in response to a viral infection). In line with this possibility, analysis of PTPN2 mRNA expression in immune cells (e.g. T cells and transformed B cell lines) indicates that the T1D-associated risk allele reduces mRNA expression of the gene [6].

Other susceptibility genes for T1D may play important roles in antiviral responses at the β-cell level. Thus, IFIH1 (MDA5), a cytoplasmic receptor of viral double stranded (ds)RNA, is expressed in human pancreatic islets and upregulated by enterovirus infection or exposure to polyinosinic : polycytidylic (PIC) acid (a synthetic viral dsRNA) [14, 24]. MDA5 inhibition in pancreatic β-cells decreases PIC-induced chemokine expression and release [14]. This has the potential to decrease immune cells homing to the islets during insulitis, and is in good agreement with observations that polymorphisms leading to decreased MDA5 expression are protective against T1D [25].

In some cases, however, expression of a candidate gene in β-cells is not associated with an obvious function. Thus, SH2B3 (or LNK), which encodes an adaptor protein implicated in growth factor and cytokine signalling, is expressed in β-cells but does not seem to play a role in cytokine-induced β-cell inflammation or death (figure 3). SH2B3 mRNA expression is upregulated in β-cells by the pro-inflammatory cytokines IL-1β + IFNγ (figure 3A), but the inhibition of SH2B3 by a specific small interfering (si)RNA does not influence cytokine-induced chemokine expression (figure 3B, C) or apoptosis (figure 3D). The T1D-associated variant (rs3184504) within SH2B3 is predicted to alter splicing regulation by disrupting an exonic splicing enhancer (ESE) motif for the splicing regulator SRp55 (figure 3E). Whether this polymorphism leads to novel SH2B3 isoforms that affect β-cell inflammation and death in vivo or modulate the immune system remains to be clarified. SH2B3 is genetically associated with other autoimmune diseases (e.g. celiac disease and rheumatoid arthritis) [26], suggesting that its role in T1D pathogenesis may depend on regulation of the immune system. These observations highlight that discovery of the specific contribution of candidate genes for T1D expressed at the β-cell and/or immune system will depend on both in vitro studies (preferentially based on human tissues) and in vivo animal models with specific (and preferentially conditional) inactivation of the candidate genes at either the β-cells or immune cells.

Figure 3.

The type 1 diabetes (T1D) candidate gene SH2B3 is upregulated by cytokines in β-cells but does not play a role in cytokine-induced inflammation or death. INS-1E cells were transfected with an irrelevant small interfering RNA (siRNA) (siCTRL) or with a siRNA targeting SH2B3. After 48 h of recovery cells were left untreated or treated with IL-1β (10U/ml) + IFNγ (100 U/ml) for 24 h as previously described [13]. (A–C) Expression of SH2B3, CXCL9 and CXCL10 were measured by RT-PCR and normalized by the housekeeping gene GAPDH. (D) Cell viability was assessed by HO/PI staining after 24 h of cytokine treatment. Results are means ± s.e.m. of three independent experiments; **p < 0.01 and ***p < 0.001 vs. siCTRL under the same treatment; §p < 0.05 and §§§p < 0.001 vs. untreated and transfected with the same siRNA; ANOVA followed by paired t-test with Bonferroni correction. (E) Single nucleotide polymorphism (SNP) function prediction report of the phenotypic changes by SNP rs3184504 in SH2B3 gene obtained with Fast SNP online free tool ( On the basis of Fast SNP software results, allele C in rs3184504 is predicted to disrupt an exonic splicing enhancer (ESE) motif recognized by the splicing factor SRp55.

Potential Pathogenic T1D Gene Networks in β-cells

Disease-associated pathways are predicted based on gene interactions defined by parameters such as gene ontology terms, protein–protein interactions, possession of common regulatory motifs (e.g. transcription factor binding sites), co-expression data and subcellular co-localization data [27]. The risk to develop T1D most likely depends on the joint effect of susceptibility variants in multiple genes interacting in pathways that alter function or expression of other members of the pathway. This may lead to disease when combined with adequate environmental stimuli: some variants may sensitize the subject to specific environmental factors, e.g. active MDA5 and the interferon regulatory pathways, may render individuals overresponsive to a β-cell viral infection or to ‘danger signals’ provided by dying cells [14, 28]. Pathway analysis of GWAS data indicates that genetic contribution to common autoimmune diseases with a strong inflammatory component, such as T1D, Crohn's disease or rheumatoid arthritis, is determined by multiple gene variants in inflammatory pathways involved in host response to infectious diseases [28]. Thus, pathways related to pattern recognition receptors [e.g. toll-like receptors (TLRs)], signal transduction (e.g. JAK-STAT and MAPK), antigen processing and presentation (e.g. MHC class I and II) are identified as key pathogenic gene networks in these diseases [28]. In line with this, a recent study integrating GWAS and protein–protein interaction data in T1D identified 17 disease-relevant biological gene networks [16]. Three of these networks are significantly enriched in cytokine-regulated genes. Of interest, many of these genes are expressed in human islets and modulated by pro-inflammatory cytokines, suggesting that they contribute for islet inflammation during the early stages of T1D pathogenesis [16].

Pathway analysis of T1D genes expressed in human pancreatic islets [7] reveal that many of them are part of pathways related to antiviral responses and type I IFN signalling (Santin I. and Eizirik DL., unpublished data). Epidemiological and clinical studies in humans support the implication of viral infections, particularly by enteroviruses (e.g. Coxsackie virus), as environmental triggers for the development of T1D [29-33]. Thus, Coxsackie virus B-specific antibodies [29] and enteroviral RNA [30] are more frequently observed in serum samples from T1D patients than in healthy individuals. Moreover, immunohistochemical staining of human pancreatic islets and other techniques revealed that the enteroviral capsid protein VP1 is present at higher frequency in insulin-containing islets from patients with recent-onset T1D as compared to healthy controls [31, 33]. A recent meta-analysis of 33 prevalence studies involving 1931 T1D cases and 2517 controls confirmed a clinically significant association between enteroviral infections and islet autoimmunity and T1D in humans [32].

Type I IFNs are important mediators of antiviral responses. They are produced by many cell types (including pancreatic β-cells) and signal to both immune and non-immune cells the presence of an ongoing viral infection [34]. The type I IFNs IFNα and IFNβ act through the common and ubiquitously expressed type I IFN receptor (IFNAR). The IFNAR triggers signal transduction pathways that establish an antiviral response in target cells by activating IFN-stimulated genes (ISGs) [34], and accumulating evidence points to a pathogenic role of type I IFNs in human T1D. Thus: (i) Type I IFNs and their downstream genes are expressed in pancreatic islets from T1D individuals [35]; (ii) There is a signature of IFN-induced genes in the peripheral blood of affected patients [36]; (iii) An association has been described between risk alleles for T1D and IFN signalling and pathways related to innate immune responses [37]; (iv) In mouse models of T1D exogenous IFNs accelerate disease [38] while blocking IFNs [39] or inactivating its receptors [38] prevents or ameliorates disease.

Of particular relevance, new GWAS-based evidence linked susceptibility to autoimmune diseases (e.g. T1D and inflammatory bowel disease) to regions in the host genome that are involved in host–virus interactions [40]. Furthermore, a combined analysis of gene networks and DNA sequence variation in T1D showed that an interferon regulatory factor 7 (IRF7)-regulated gene network (named IDIN, for ‘IRF7-driven inflammatory network’) is a major contributor to T1D risk [41]. As a whole, these data suggest that specific viral infections (or other ‘danger signals’ that remain to be determined) and the consequent interferon production may function as initial triggers of a severe islet inflammatory response in genetically predisposed individuals, eventually culminating in an autoimmune assault against the β-cells in some individuals.

Human and rodent pancreatic β-cells express several pattern recognition receptors (PRRs) that recognize and respond to viral components, including TLR3, MDA5 (a candidate gene for T1D) and RIG-I [14, 42-45]. Other PPRs are also expressed in human pancreatic islets (reads per kilobase per million mapped reads; RPKM > 1), including NOD1 and the nuclear receptor IFI16 [7]. Some of these receptors may also recognize molecules derived from host tissues, components of cells or induced gene products in specific conditions [46]. For example, cell death releases endogenous DNA/RNA which may be recognized by TLRs, activating innate immune responses and, in some cases, autoreactive T cells. Mammalian DNA and RNA are potent self-ligands for TLR9 and TLR7, respectively, and induce IFNα production by pre-dendritic cells, which may contribute to the pathogenesis of systemic lupus erythematosus [47]. Administration of low doses of CpG (an endogenous ligand of TLR9) accelerates the onset of diabetes in non-obese diabetic (NOD) mice [48], whereas TLR9−/− NOD mice have a significant delay in the onset of diabetes as compared to TLR9+/− or NOD wild-type (WT) littermates [49]. In the mouse model of streptozotocin-induced diabetes, deletion of TLR4 attenuates NF-κB activation, leading to reduced expression of pro-inflammatory cytokines and chemokines [50]. The expression patterns of TLRs and their ligands in T1D individuals support their role in disease pathogenesis. Thus, circulating levels of TLR2 and TLR4 ligands (e.g. endotoxin, Hsp60 and HMGB1) are significantly increased in T1D patients compared to healthy controls [51]. Moreover, expression of TLR2 and TLR4 is significantly upregulated in monocytes from patients with T1D compared to controls; this increase correlates with augmented release of IL-1β and TNFα by monocytes [52].

The effects of PRR activation in β-cells are mainly mediated via three signal transduction cascades, namely the JAK/STAT-, MAPK- and NF-κB-mediated pathways [42, 44, 53]. It is noteworthy that the main signalling pathways perturbed by RNA viruses (the family to which Coxsackie viruses belong) to assure their survival include the JAK-STAT and chemokine signalling pathways [54], in a mirror image of the pathways regulated by T1D candidate genes in pancreatic β-cells. Indeed, several T1D genes interact within the antiviral response and type I IFN signalling pathways in pancreatic β-cells, modulating chemokine production and apoptotic pathways via the JAK-STAT signalling pathway (figure 4). The local release of chemokines will enhance the inflammatory environment in the islets, attracting more infiltrating immune cells which, together with accrued cell death, may amplify the autoimmune attack against β-cells [8].

Figure 4.

Type 1 diabetes (T1D) genes interact in pathogenic pathways linked to antiviral responses in pancreatic β-cells. Upon recognition of dsRNA by the cytoplasmic receptor MDA5 (which is a candidate gene for T1D) several transcription factors, including IRFs and STATs, are activated in β-cells leading to production of pro-inflammatory chemokines and type I IFNs and local inflammation [14]. Both type I and type II IFNs contribute to β-cell destruction in genetically susceptible individuals. PTPN2, another candidate gene for T1D, modulates interferon (IFN)-induced β-cell death by regulating phosphorylation of the pro-apoptotic protein Bim via JNK1 [15]. It also plays an important role as a negative regulator of the signal transducer and activator of transcription (STAT) signalling pathway in pancreatic β-cells [13, 15]. USP18, a member of the T1D-associated gene network IDIN, negatively regulates IFN-induced STAT signalling and mitochondrial apoptotic pathways in β-cells [17].

The viral dsRNA receptor MDA5 initiates virus-induced chemokine production via activation of the transcription factors IRF3, IRF7 and NF-κB [55], whereas genes such as PTPN2 or USP18 (a member of the T1D-associated gene network IDIN [41]) serve as ‘checks and balances’ by negatively regulating the activation of the STAT signalling pathway [13-15, 17] (figure 4). Both PTPN2 and USP18 inhibit IFN-induced β-cell apoptosis via modulation of the mitochondrial pathway of cell death [15, 17]. Rare variants within the sequence of MDA5 predicted to reduce expression or alter function of the receptor in humans are protective against T1D [56]. Partial inhibition of MDA5 expression leads to a decrease in viral dsRNA-induced chemokine production in pancreatic β-cells [14], suggesting that individuals carrying the protective allele in MDA5 will release less pro-inflammatory chemokines in response to a viral infection, leading to a less pronounced inflammation in the islets (figure 4). In line with this, risk alleles in MDA5 are associated with faster progression of T1D development (31% within 5 years) as compared to protective genotypes (11% within 5 years) [57]. Inhibition of PTPN2 in β-cells leads to exacerbated STAT activation and increased PIC- and IFN-induced β-cell death (figure 4) [13-15]. These in vitro data are in agreement with the observations that risk alleles in the PTPN2 gene provoke a decrease in its expression and confer risk for T1D [6]. The observations described above support the hypothesis that key T1D candidate gene networks regulate β-cell mechanisms to detect and counteract viral infections. In line with this, the JAK-STAT signalling pathway was ranked as the pathway most closely associated with T1D by using a sub-pathway-based approach focused on the joint effects of single nucleotide polymorphisms (SNPs) modestly associated with the disease [11].

Apart from the evident role of MDA5 and PTPN2 (discussed above), other candidate T1D genes may also contribute for the regulation of antiviral responses at the β-cell level (figure 5). Tyrosine kinase 2 (TYK2) is a T1D gene encoding a kinase that is bound to the type I IFN receptor (IFNAR) on the cell surface and takes part in the IFN signalling cascade [58]. Various SNPs within and downstream TYK2 sequence have been associated with protection against T1D [59] and other autoimmune diseases (systemic lupus erythematosus, ulcerative colitis or Crohn's disease) [16, 60]. Although the functional effect of these polymorphisms remains to be characterized, a natural loss-of-function mutation in TYK2 in the mouse strain B10.Q/J reduces cellular responsiveness to type I IFNs and other pro-inflammatory cytokines (e.g. IL-12 and IL-23), rendering the mice more susceptibility to infections but markedly resistant to induction of autoimmune arthritis [61]. In agreement with these results, TYK2 knockout mice display reduced type I and II IFN signalling as demonstrated by diminished activation of JAKs and STATs [62]. The protective nature of the T1D-associated polymorphisms in TYK2 suggest that individuals carrying these alleles may have a decreased responsiveness to IFNs in β-cells (and possibly other cell types), leading to a reduced expression of IFN-induced genes (e.g. pro-inflammatory chemokines) and reduced inflammatory environment in the islets.

Figure 5.

Global integration of known T1D genes into pathogenic pathways in pancreatic β-cells. T1D-associated genes and pathways (in red) participate in signalling networks that may contribute to the pathogenesis of T1D. These include type I IFN signalling, PRR signalling, communication with the immune system, apoptosis and metabolism/function. Upon viral infection, recognition of viral double stranded (dsRNA) by the receptor MDA5 leads to secretion of pro-inflammatory cytokines and chemokines by β-cells (e.g. IFNα and IFNβ, CCL2, CXCL10) and infiltrating immune cells (e.g. IFNγ), triggering the activation of several signalling cascades implicated in antiviral responses and inflammation. Viral products, putative endogenous ligands of PRR and pro-inflammatory mediators act as stressors, inducing expression of transcription factors such as BACH2, GLIS3, STAT1 and NF-κB that modulate apoptotic pathways (mainly the mitochondrial pathway of cell death, with a central role for the BH3-only protein Bim). Some of these transcription factors (e.g. NF-κB and STATs) induce expression of HLA class I molecules while others [e.g. GLIS3 and NOVA1 (not shown)] modify AS of some genes. This, together with increased endoplasmic reticulum stress (not shown in the figure) may induce the generation of potential novel autoantigens that further aggravate the autoimmune assault. Some T1D genes, such as GLIS3, also regulate basic processes in β-cells (insulin synthesis and secretion). Abbreviations: AS, alternative splicing; dsRNA, double stranded RNA; IFN, interferon; INS, insulin; PRR, pattern recognition receptor; STAT, signal transducer and activator of transcription; T1D, type 1 diabetes.

Several T1D candidate genes are also involved in PRR signalling (figure 5). Initiation of antiviral responses (e.g. type I IFN production) in pancreatic β-cells through PRRs (such as MDA5) require the activation of STATs and IRFs. These transcription factors regulate potential diabetogenic pathways that include T1D-associated genes (e.g. STAT4 [63]) within the pathway or are part of T1D genetic risk pathways detected by in silico analysis [11, 41]. Another important transcription factor for β-cell PRR signalling, inflammation and apoptosis is NF-κB [8]. NF-κB activity at the β-cell level is potentially regulated by two T1D candidate genes, namely TNFAIP3 (or A20) and PRKCQ [64, 65]. The T1D gene TNFAIP3 encodes a cytoplasmic protein that is induced by NF-κB and inhibits the duration and intensity of NF-κB signalling, providing a negative feedback loop [64]. Polymorphisms within this gene are associated with risk for T1D, rheumatoid arthritis, systemic lupus erythematosus, psoriasis and Crohn's disease [66, 67]. Although the functional impact of T1D-associated polymorphisms in the TNFAIP3 protein remains to be clarified, sequences containing three of the five SNPs associated with rheumatoid arthritis significantly repress TNFAIP3 expression. In vivo, this decreased TNFAIP3 transcription would limit the negative regulation of NF-κB, leading to enhanced pro-inflammatory cytokine and chemokine expression [68]. Considering that NF-κB has also a mostly pro-apoptotic function in pancreatic β-cells [8, 69], alterations in TNFAIP3 expression/function may have an important impact on β-cell survival.

A meta-analysis of GWAS data identified a polymorphism located 79 kb from PRKCQ as significantly associated with T1D genetic risk [65]. This gene encodes a kinase implicated in T cell function, including integration of T cell receptor (TCR) and CD28 signalling and activation of JNK and NF-κB [70]. Several variants of the PRKCQ gene are expressed in pancreatic human islets (our own unpublished data), but its role in NF-κB regulation at the β-cell level remains to be clarified. Interestingly, PRKCQ expression has also been linked to fat-induced insulin resistance in skeletal muscle and liver [71, 72].

Some β-cell-expressed T1D candidate genes have potential roles in the regulation of chemokine production and release by β-cells, regulating the ‘dialogue’ with the immune system cells (figure 5) [8]. Thus, invading immune cells release cytokines such as IL-1β or IFNγ that upon recognition by specific receptors in β-cells activate signalling pathways (e.g. NF-κB and STAT signalling) and increase local expression of pro-inflammatory chemokines. Besides the T1D candidate gene PTPN2 (discussed above), other candidate genes expressed in β-cells, such as SH2B3, may regulate this process by inhibiting JAK2 activation (figure 5). Experiments performed by our group have not confirmed a role of SH2B3 in IL-1β + IFNγ-induced chemokine expression in β-cells (figure 3), but additional studies are required to clarify this issue.

Examination of the candidate genes described above indicate a common theme, namely that polymorphisms leading to a vigorous innate immunity and inflammatory response are usually associated with increased risk for T1D and other autoimmune diseases, while decreased responses in these pathways lowers the risk of developing these diseases. It is conceivable that populations that are today at increased risk for autoimmune diseases evolved under the genetic selective pressure provided by infectious diseases such as plague, tuberculosis and leprosy [73-75]. In this context, autoimmune diseases could be seen as the ‘noise’ of a very active immune system, selected over millennia to provide better protection against life-threatening infectious diseases.

While most of the genes discussed above are related to innate immunity and inflammation, GLIS3 is a candidate gene for both T1D [76] and T2D [77] that has a relevant role for both β-cell function and survival [18, 19]. This transcription factor regulates insulin gene expression both directly and via interaction with Pdx1, MafA and NeuroD1 (figure 5) [78]. Loss-of-function mutations in GLIS3 gene cause a rare syndrome characterized by neonatal diabetes and congenital hypothyroidism [79]. Thus, GLIS3 is an example of genes in which inactivating mutations lead to rare monogenic diseases (e.g. neonatal diabetes) while common variants contribute to the genetic risk for complex diseases (e.g. T1D and T2D). Of special interest is our recent finding that GLIS3 regulates alternative splicing in pancreatic β-cells. Thus, inhibition of GLIS3 expression by specific small interfering RNAs (siRNAs) in rat and human β-cells modifies alternative splicing of the pro-apoptotic protein Bim, inducing formation of the pro-apoptotic variant Bim small (BimS). This increases β-cell apoptosis via the mitochondrial pathway of cell death, which is prevented by parallel knockdown of Bim [19].

Epigenetic Regulation and Alternative Splicing of T1D Candidate Genes and its Implication in Pancreatic β-cell Dysfunction

Environmental factors may alter gene expression via epigenetic modifications of relevant tissues in diabetes [80]. These epigenetic changes can contribute to T1D pathogenesis by inducing persistent activation of ‘pro-inflammatory phenotypes’ in β-cells or immune effector cells. A genome-wide DNA methylation analysis of mononuclear cells from monozygotic twins discordant for T1D revealed the presence of T1D-specific methylation variable positions (T1D–MVPs) [78]. Interestingly, several T1D-MVPs were associated with genes implicated in immune functions or associated with T1D risk, including hypomethylation of HLA-DQB1 which carries the highest predisposing risk allele for T1D (in LD with risk alleles in HLA-DRB1) [78]. Methylation modifications in pancreatic islets from type 2 diabetic individuals have already been described [81], but extension of these findings to T1D is hampered by the scarce availability of human islet material from these patients. The development of nPOD, a network for pancreatic organ donors with diabetes [82], and the progressive improvement of techniques for laser capture of tissues and amplification of mRNA/cDNA ahead of RNA sequencing will hopefully solve this problem.

A ChIP-chip approach to compare genome-wide histone H3K9me2 patterns in peripheral lymphocytes and monocytes from T1D patients and healthy donors showed a significant increase in H3K9me2 in several susceptibility genes for T1D (e.g. CTLA4). These was also altered H3K9me2 in genes grouped in networks related to autoimmunity and inflammation, including NF-κB and JNK signalling [83]. Both NF-κB and JNK have been previously shown to play an important role in cytokine-induced β-cell death [8, 69, 84].

Many T1D genetic associations fall into non-coding regions of the genome [85], suggesting that not only protein-coding genes but also previously considered ‘silent areas’ contribute to disease risk. Technological advances that enable a better understanding of gene regulation by distant regions will hopefully provide information on the role of these non-coding regions. Gene expression can be regulated by non-protein-coding RNAs and in recent years several studies implicated microRNAs (miRNAs) in the pathogenesis of diabetes and its complications [86, 87]. Modified expression of the miRNAs miR-29a/b/c, that regulate expression of genes implicated in β-cell function and survival, has been observed in islet endocrine cells from pre-diabetic NOD mice [88]. Furthermore, lymphocytes from autoantibody positive T1D patients show an increased expression of miR-326 [89]. Rare variants in the 3′-UTRs of the T1D candidate genes CTLA4 and IL-10 may affect T1D pathogenesis by modifying miRNA-binding sites and altering post-transcriptional gene regulation [90]. Independent component and pathway-based analysis of miRNA-regulated gene expression in cytokine-treated INS-1E cells identified miRNAs that directly or indirectly modulate the type I IFN and cytokine-cytokine receptor pathways [91]. Furthermore, expression profiling of miRNAs in mouse and human islets exposed to pro-inflammatory cytokines show modulation of several miRNAs [88].

Another mechanism by which inflammation may cross-talk with candidate genes is via alternative splicing. Recent studies by our group indicate that exposure of rat or human islets to the cytokines IL-1β + IFNγ induce modification in splicing of >500 genes [7, 92]. Interestingly, the ‘alternative splicing signature’ induced by cytokines in human islets [7] seems rather specific, since it is not reproduced when the islets are exposed to a metabolic stress, namely exposure to the free fatty acid palmitate (Cnop M et al., unpublished data). Moreover, the RNAseq data of control and cytokine-treated human islets indicate that several T1D candidate genes undergo alternative splicing ([7]; Santin I. and Eizirik DL., unpublished data). These observations suggest that under inflammatory conditions different transcripts of a given T1D gene may have different functions. As discussed above, recent data from our group [19] show that the candidate gene GLIS3 modulates alternative splicing of the pro-apoptotic protein Bim. The observations described above suggest that pro-inflammatory cytokines may affect the function of T1D genes via modifications in miRNA expression or alternative splicing. Since these candidate genes can in turn modulate inflammation and cytokine production and/or function this points to a complex regulatory system involving candidate genes, triggers of inflammation, the inflammation itself, changes in miRNAs and alternative splicing induced by inflammation, and their impact on candidate genes, β-cell function and survival (figure 6). The outcome of these interactions may be either resolution of inflammation or progression to a full autoimmune assault, progressive β-cell loss and T1D.

Figure 6.

Proposed model for the cross-talk between T1D genes, inflammation, alternative splicing and miRNAs during insulitis. T1D genes interact with environmental triggers to induce islet inflammation. Local inflammation may affect the function of T1D genes via modifications in miRNA expression or alternative splicing. The outcome of this complex regulatory system may be either resolution of inflammation with minor or no β-cell loss, or evolution to autoimmunity, progressive β-cell death and eventual diabetes.


Given the observed heterogeneity among T1D patients regarding both autoimmunity markers [93] and the rate of loss in β-cell function [94], it is conceivable that genetically different individuals exposed to diverse environmental cues will have different pathways for β-cell dysfunction and death and disease initiation/evolution. For instance, in some individuals excessive innate immune responses may prevail, in others the β-cells may be particularly susceptible to immune-mediated dysfunction and apoptosis, while in a third group (probably the most numerous) different combinations of these two broad factors will contribute to disease. This indicates that different therapeutic strategies will be required depending on the genetic background of the affected individuals. As most disease-associated variants have small effects on disease risk, they are presently insufficient to predict disease development or provide immediate diagnostic benefit. Thus, functional analysis of risk variants is needed to determine the precise consequences of the associated variation in disease pathogenesis and to define their translational potential [95].

In the context of T1D, additional studies are required to clarify how risk variants affect pancreatic β-cell responses to environmental challenges and/or inflammatory mediators. As the function of these variants, alone or in interaction with other candidate genes and/or environmental signals, is identified and downstream pathogenic pathways clarified, our understanding of T1D pathogenesis will improve. As is stated in the Dhamapadha (a fundamental Buddhist text) ‘a fool who is aware of his own ignorance is thereby wise.’ Understanding the causes of T1D at its real level of complexity will hopefully open the door for personalized approaches for prevention or therapy.


Work by the Authors discussed in this review was supported by the European Union (Collaborative Projects Naimit and BetaBat in the Framework Program 7), Juvenile Diabetes Research International and Actions de Recherche Concertée de la Communauté Française (ARC) and Fonds National de la Recherche Scientifique (FNRS), Belgium. I. S. was supported by a postdoctoral fellowship from the Education Department of the Basque Country. We thank Dr L. Marroquí, Dr F. Moore and Dr A. Op de Beeck for helpful comments on the manuscript.

Conflict of Interest

No potential conflict of interest declared.