Type 1 Diabetes

Introduction

Type 1 Diabetes

Type 1 Diabetes (T1D) is a complex, multifactorial, chronic autoimmune disorder in which insulin-producing beta cells in the islets of Langerhans of the pancreas are targeted and destroyed by the body’s own immune system. The disease is characterized by hyperglycaemia and isulinopenia and requires lifelong treatment with insulin administrations for survival.

T1D usually begins in childhood or young adulthood before 20 years of age and is therefore originally termed juvenile-onset diabetes, even though it can occur at any age. According to the American Center for Disease Control, the prevalence of T1D for residents of the United States aged 0-19 years is 1.7/1000. T1D incidence has been globally rising during the past decades by as much as a 5.3% annually in the United States. If present trends continue, doubling of new cases of T1D in European children younger than 5 years is predicted between 2005 and 2020, and prevalence of cases in individuals younger than 15 years will rise by 70%

(Patterson et al, 2009). Epidemiological studies highlight the geographical

differences in the incidence of T1D, with the highest occurring in individuals of

European background, intermediate in Africans and much lower in East Asians

(Diabetes Epidemiology Research International Group, 1988). Within Europe, a

steep gradient exists, and prevalence in northern countries is several-fold higher

than in those on the Mediterranean. Both genetic and environmental differences

are likely to underlie this geographical distribution (Borchers et al, 2010).

M. Grupillo

The etiology and pathogenesis of T1D could be represented by a series of stages beginning with genetic susceptibility, in the absence of autoimmunity, and culminating in the complete loss of insulin due to T-cell-mediated islet destruction (Fig. 1) (Eisenbarth, 1986). It is likely that interplay of inherited and environmental factors is important in defining disease risk.

human and the hypothesis that insulin may be a primary islet autoantigen for the non-obese diabetic (NOD) mouse and human.

Stages of type 1A diabetes

Dr Stuart Soeldner (1) at the Joslin Diabetes Center metabol- ically evaluated over several decades a series of monozygotic twins whose twin mate had type 1A diabetes. The risk of developing diabetes of such a monozygotic twin is approxi- mately 50%, and with the development of the cytoplasmic islet cell autoantibody assay (2), it was possible to correlate the appearance of autoantibodies to the loss of insulin secretion, as twins progressed to overt diabetes (3). These observations led to the hypothesis that type 1A diabetes is a chronic auto- immune disorder, and the disease could be represented by a series of stages beginning with genetic susceptibility in the absence of autoimmunity and culminating in the complete loss of insulin secretion (4) (Fig. 1).

Stage 1: genetic predisposition

It is likely that type 1A diabetes is a result of interactions between genetic susceptibility and environmental influences.

Data from twin studies revealed that identical twins show differences in their concordance rate (3, 5, 6). Of note, the concordance for type 1A diabetes is directly correlated with the age of onset in the twin affected first. Thus, if diabetes occurs in the first monozygotic twin before age 5, the risk of diabetes in the second twin is 50%, whereas if diabetes develops after age 25 in the first twin, the risk to the second twin is only 5% (5). These data suggest that in twins affected

before the age of 5, there is a higher genetic susceptibility and that this susceptibility determines both the probabilities of early onset and higher concordance. Pairs of twins discordant for type 1A diabetes may be concordant for the presence of islet autoimmunity and subclinical impairment of insulin release. To test this hypothesis, we have studied 23 monozy- gotic twins selected because they were discordant for type 1A diabetes when first ascertained (7). In this group, seven have developed diabetes after a period of time ranging between 3 and 31 years from the time of onset in the first twins. Among the still non-diabetic 16 pairs of twins, 12 could be evaluated with intravenous glucose tolerance tests and measurement of autoantibodies to GAD-65, insulin, and IA-2. Among these subjects, two-thirds had autoantibodies and first-phase insulin release lower than the first percentile of normal controls, suggesting a high risk for future progression to diabetes.

In both humans and animal models, the major determinants of type 1A diabetes are genes within the major histocompatibility complex (MHC) [human leukocyte antigen (HLA) in humans]

(8, 9). Type 1B diabetes is defined as severe loss of insulin secretion that is not immune-mediated and does not have a known genetic etiology. While this type has been described as a clinical entity, there is no convincing evidence that it exists.

Most type 1A diabetes patients express HLA-DR3 or DR4 class II alleles, and approximately 30–40% of patients are DR3/DR4 heterozygous and carry the highest risk genotype (10). Sequencing of HLA molecules has shown that the DQ locus has the strongest association with type 1A diabetes (11).

This locus encodes for the HLA-DQ molecule (a heterodimer of two chains termed a and b). This molecule controls immune recognition and antigen presentation to CD4

T cells and thus may affect susceptibility to diabetes by shaping the T-cell receptor (TCR) repertoire, by determining the immune response to peripheral peptides or by both mechan- isms. In the NOD mouse model, the MHC is also the major determinant of diabetes susceptibility. For NOD mice, the absence of I-E expression and specific I-A sequences (I-A

) are associated with susceptibility. As described later in detail, in the mouse, these alleles control the response to the insulin peptide B9-23 as well as development of diabetes.

The insulin gene, the variable nucleotide tandem repeat locus (VNTR), is the next best-characterized diabetes suscepti- bility locus. This locus affects the expression of insulin in the thymus (12, 13), and the available evidence suggests that thymic expression of self-antigens may be crucial for tolerance induction (14). Levels of insulin expression in the thymus may affect negative selection of insulin autoreactive T cells or positive selection of regulatory T cells, thus modulating

Precipitating event

Genetic predisposition

β-cell mass

Overt immunologic abnormalities

Normal insulin release

Glucose normal

C-peptide present

Age (years)

Overt diabetes

No C-peptide Progressive

loss insulin release

Fig. 1. Stages in the development of type 1A diabetes. Modified from (4) Eisen barth GS. Type 1 diabetes mellitus. A chronic autoimmune disease. N Engl J Med 1986;314:1360–1368.

Immunological Reviews 204/2005 233

Figure 1 - Stages in the development of type 1A diabetes. (Eisenbarth, N Engl J Med 1986)

Genetic predisposition

The strongest evidence for a genetic influence comes from comparing

monozygotic and dizygotic twins. All studies show a higher disease concordance

rate in genetically identical monozygotic twins (an average of 50%) (Redondo et

al, 2008) than in dizygotic twins (6%) (Steck et al, 2005), which is similar to what

is found in nontwin siblings. Of note, the concordance for T1D is directly

correlated with the age of onset in the twin affected first. If diabetes occurs in the

first monozygotic twin before age 5, the risk of the second twin developing

diabetes is at least 60%, whereas if diabetes develops after age 25 in the first twin,

the risk to the second twin is only 5% (Redondo et al, 2001). These data suggest

that in twins affected before the age of 5, there is a higher genetic susceptibility

that determines both the probabilities of early onset and higher concordance. For

monozygotic twins discordant for T1D, expression of antiislet autoantibodies

directly correlates with progression to overt diabetes, even though this may not

occur until after decades of follow-up (Redondo et al, 1999; Redondo et al, 2001;

Hyttinen et al, 2003). Indeed, Eisenbarth claims that, with prospective long-term

follow-up, concordance between monozygotic twins increases tremendously, at

least considering as markers of autoimmunity, islet autoantibodies (Redondo et al,

2008).

The search for T1D-associated genes

Over the last three decades, the study of T1D has led the field in the identification

of genes underlying complex multifactorial diseases. Unlike single gene disorders,

which are inherited in distinct predictable Mendelian patterns, in T1D,

identification of the combination of causative genes is challenging. The two

primary approaches used to identify risk loci for T1D have been linkage studies

and association studies. Linkage studies, typically using affected sibling pairs, can

identify regions of the genome that are shared more frequently among affected

relatives. They use markers spanning the genome at a modest density and are the

method of choice when the risk factor have large effect sizes but are relatively

rare. They provide broad information on chromosomal regions that may contribute

to T1D risk. In contrast to linkage studies, association studies can detect alleles

with much more modest effects on risk as long as those alleles are relatively

common. They use specifically selected markers in gene of interest that are

genotyped in case patients and unaffected control individuals (Steck and Rewers,

2011). Recently, genome-wide association (GWA) studies have represented a

paradigm shift in strategies for identifying risk genes for complex (multifactorial)

human diseases, including type 1 diabetes. This research has been made possible

by the developments of high-density single nucleotide polymorphism (SNP)

genotyping arrays, analytical methods that build on the synthesis of population

genetics, statistical genetics and genetic epidemiology, and the use of large

clinically well-characterized case and control populations, as well as family

collections, as that provided by the International Type 1 Diabetics Genetics Consortium (T1DCG) (Pociot et al, 2010).

The recently completed T1DGC GWA study, meta-analysis and replication study included data from >30.000 individuals (Barrett et al, 2009). Although there were

>40 regions in the human genome that provided evidence of association with T1D etiopathogenesis (P<10

) (Figure 2), it is well established that some allele at different loci of the major histocompatibility complex (MHC) in humans, the human leukocyte antigens’ (HLA) complex, show the strongest association by far, with reported odd ratios (ORs) ranging from 0.02 to >11 (Erlich et al, 2008). After HLA, variants in the gene encoding insulin (Ins) confer the highest genetic susceptibility to the disease (Suri and Unanue, 2005; Jones et al, 2006; Barrett et al, 2009; Howson et al, 2009) with OR = 2.38 (Pociot et al, 2010). Polymorphism of other genes is associated with diabetes risk, but only two other loci, the protein tyrosine phospatase, non receptor type 22 (lymphoid) (PTPN22) gene on chromosome 1p13 (Bottini et al, 2004) and the interleukin 2 receptor, alpha (IL2RA) (Qu et al, 2007) on chromosome 10p15.1, have ORs greater than 1.5;

most are in the range of 1.1-1.3 which underscores the importance of the HLA

region compared with other loci.

Figure 2 - Type 1 diabetes–associated loci from GWA studies (Pociot et al, Diabetes 2010).

HLA

An association between a gene of the HLA region and T1D was first reported in 1974 (Nerup et al, 1974) and to date, the MHC remains the genetic region showing the strongest association with disease and is designated IDDM 1 (Davies et al, 1994).

The HLAs are a family of homologous proteins that present antigens to T- cells. They are all encoded within the MHC region, which spans 4 Mb of DNA, on chromosome 6p21 in tight linkage disequilibrium (Mehra and Kaur, 2010) and are arranged in three subregions.

The distal class I region contains genes that encode the alpha peptide chains of the

HLA class I molecules, which dimerize with beta2-microglobulin to form the

classic transplantation antigens HLA-A, B and C. Class I molecules are expressed

on the surface of all nucleated cells and are responsible for antigen presentation to

CD8+ cytotoxic T-lymphocytes.

The centromeric class II region contains A and B genes encoding the alpha and beta chains of the HLA-DR, DQ and DP molecules. Peptide-class II complexes are expressed by specialized antigen-presenting cells (dendritic cells, B-lymphocytes, monocytes, the thymus epithelium) and can activate CD4+ T-helper (Th) cells of Th1 (proinflammatory) or Th2 (regulatory) phenotype. The antigen is bound and presented in a groove whose conformation and amino acid context determines the specificity of each allele for different antigenic epitopes. The groove of class I HLAs accommodates peptides of 8-9 amino acids in length, whereas class II HLAs present longer epitopes of 15 amino acids. A sophisticated proteolityc machinery is responsible for digestion proteins to these sizes (Polychronakos and Li, 2011).

The central class III region encodes molecules with a variety of functions, including complement components (C4A, C4B, factor B and C2), tumor necrosis factor (TNF), heat shock protein Hsp70 and 21-hydroxylase (CYP21).

All HLA proteins are highly polymorphic, as they have dozens or even hundreds

of protein-coding alleles (Hughes, 2008). Sorting out HLA associations is

complicated not only by the extremely large numbers of reported alleles at the

HLA genetic loci but also by differences in allele frequencies and haplotypic

combinations among populations, incomplete penetrance of the HLA susceptibility

loci, and epistatic interactions of HLA with other susceptibility factors (Noble and

Erlich, 2012). This functional sequence diversity is driven by a strong selective

pressure that favours heterozygosity and a wide range of selectivity for antigens to

optimize immune response against a large diversity of current and emerging

pathogens. This recent, environmentally modulated selection is responsible for

broad population differences in HLA alleles, which could partially explain the geographic differences in T1D incidence (Melanitou et al, 2003).

Extensive sequencing of MHC class II alleles in man, the nonobese diabetic (NOD) mouse, and the Bio-breeding rat, as well as the use of NOD mice transgenic for several MHC class II molecules, has revealed a complex interplay between alleles of the two major isotypes of MHC class II molecules, HLA DR and DQ in man (especially the genes encoding their highly polymorphic β-chain genes DRB1 and DQB1) and I-A and I-E in the mouse (Acha-Orbea and McDevitt, 1987; Wicker et al, 1995). Both DR and DQ genes are likely to be important for determining disease risk, but the effect of each allele may be modified by the haplotype on which it is carried. This determines a hierarchy of associations ranging from highly protective to strongly predisposing (Sheehy et al, 1989; Cucca et al, 1993; Erlich et al, 1993; Noble et al, 1996; Bugawan et al, 2002; Lambert et al, 2004; Aly et al, 2006). The DRB1*1501-DQA1*0102- DQB1*0602 haplotype, found in ~20% of the population but only 1% of diabetic patients, confers dominant protection against T1D (Pugliese et al, 1995; Erlich et al, 2008). Conversely, the highest risk is conferred by the DR3/DR4 heterozygous haplotype (DR3 is DRB1*03-DQA1*0501-DQB1*0201 and DR4 is DRB1*04- DQA1*0301-DQB1*0302). Over 90% of young-onset T1D cases carry at least one copy of these haplotypes compared to ~20% of the general European- descendent population (Erlich et al, 2008).

The presence of an amino acid other than Asp at position 57 of the DQB chain has

long been known as a major determinant of T1D pathogenicity. DQB1 alleles that

differed in their T1D association also differed at a codon for Ala, Ser or Val versus Asp. Physical explanation of the unusual importance of this particular single amino acid location for the development of the autoimmune characteristics of T1D came with the elucidation of the crystal structure of the HLA-DQ8 molecule (Lee et al, 2001). Its crystal structure (identical to the homologous I-Ag

molecule present in the NOD strain) indicated that the Asp at position 57, a residue in pocket P9 of the peptide-binding groove, contributes to a salt bridge that is absent in the presence of the other neutral amino acids at this position. A model was proposed in which DQβ position 57 aspartic acid positive alleles mediate resistance to IDDM (Insulin Dependent Diabetes Mellitus), which varies in degree with the sequence of other residues in the DQα and β chains (Todd et al, 1987;

Horn et al, 1988; Corper et al, 2000). In support of this model, studies on

transgenic NOD mice showed that substitution of the Ser57 and His56 residues in

the β chain of the I-A molecule with Asp57 and Pro56 protected the mice from

developing diabetes (Singer et al, 1988). Expression of I-E (β chain position 57

aspartic acid positive) in the NOD mouse, and of DR B1 chains expressing

aspartic acid at position 57, also mediates varying degrees of resistance to type 1

IDDM. The loss of the Asp residue may contribute to diabetic pathology by

affecting the proper lodging and presentation of self-peptides, leading to

ineffective tolerance induction and autoimmunity (Morel et al, 1988; Todd et al,

1988; Trucco, 1992; McDevitt, 2001; Ludvigsson, 2009; Yoshida et al, 2010). As

a result, T-cells with a relatively higher affinity for islet cell autoantigens, such as

insulin, can survive the central negative selection process and populate the

periphery in genetically susceptible individuals. On this basis, Tian and collegues transfected ex vivo the gene encoding a resistant Asp-57

beta chain into the bone marrow cells isolated from the diabetes-prone NOD mouse. The expression of the newly formed diabetes-resistant molecule in the reinfused hematopoietic cells was sufficient to prevent T1D onset in the NOD recipient, even in the presence of the native, diabetogenic, non-Asp-57

molecule (Tian et al, 2004).

Several investigators have reported that variation at DPB1, another HLA class II gene, also contributes to T1D risk, although not as strong as the effect of the established predisposing and protective DR-DQ haplotypes. Results of these studies have shown that susceptibility is increased by DPB1*0202 and DPB1*0301 and reduced by DPB1*0402 (Noble et al, 2000; Cucca et al, 2001;

Valdes et al, 2001; Cruz et al, 2004; Stuchlikova et al, 2006; Baschal et al, 2007;

Varney et al, 2010).

It has long been known that the class II HLA association does not explain all of the linkage to the MHC region; an equally consistent, albeit substantially less prominent, association has been found for class I alleles (Fennessy et al, 1994;

Honeyman et al, 1995; Lie et al, 1999; Nejentsev et al, 2000; Noble et al, 2002;

Valdes et al, 2005; Valdes et al, 2005). A recent study by Nejentsev et al

demonstrates that, after taking into account the dominating influence of class II

genes, most of the residual association in the HLA region can be attributed to

HLA-B and HLA-A genes (Nejentsev et al, 2007). Most notably the presence of

the HLA-B*39 allele was found to be a significant risk factor and is associated

with a lower age at diagnosis of T1D. Additionally, HLA-A*02 increases the risk

in individuals possessing the high-risk class II DR3/DR4 haplotype (Fennessy et

al, 1994; Robles et al, 2002). HLA-A*0201 is one of the most prevalent class I alleles, with a frequency of >60% in T1D patients. There is accumulating evidence for the presence and functionality of HLA-A*02-restricted CD8 T-cells reacting against β-cell antigens such as insulin, glutamate decarboxylase (GAD) and  Islet Amyloid Polypeptide (IAPP) in T1D patients and islet transplanted recipients (Panina-Bordignon et al, 1995; Panagiotopoulos et al, 2003; Parker et al, 2009).

Transgenic NOD mice have been generated expressing human HLA-A*02

molecules (Marron et al, 2002) and their accelerated diabetes onset provides

functional evidence for the involvement of this particular class I allele.

Insulin

The pancreas houses two distinctly different tissues. Its bulk comprises exocrine tissue, which is made up of acinar cells that secrete pancreatic enzymes delivered to the intestine to facilitate the digestion of food. Scattered throughout the exocrine tissue are many thousands of clusters of endocrine cells, the islets of Langerhans, composed by four cell populations, organized in a stereotypical topological order (Trucco, 2005). Within the islet, α cells produce glucagon; β cells, insulin; δ cells, somatostatin; and γ cells, pancreatic polypeptide — all of which are delivered to the blood (Figure 3).

Figure 3 – Cross section of the pancreas (Trucco et al, J. Clin. Invest 2005)

Insulin, a 51 amino acid heterodimer, selectively produced in the β-cells as a pre- pro-hormone, is subsequently processed to proinsulin and finally, after removal of the connecting peptide, to insulin (Figure 4).

Figure 4 - Proinsulin molecule. (A) Tridimensional image. (B) Schematic view. (Luppi et al, Pediatric Diabetes 2011).  

The autoimmune nature of T1D is exquisitely specific for the insulin-producing β- cells, whereas other endocrine cells within the islets and the surrounding exocrine cells are largely spared (Tisch and McDevitt, 1996; Kelly et al, 2001). Indeed, such cellular specificity strongly implicates the fundamental role of β-cell-specific self-antigens in the initiation and progression of the disease.

Although numerous islet autoantigens, such as glutamate decarboxylase 65

(GAD65), insulinoma-associated protein (IA)-2, islet-specific glucose-6-

phosphatase catalytic subunit related protein (IGRP), islet cell autoantigen 69 (ICA69), and Zinc transporter (Znt) 8, have been identified in T1D etiology (Liu and Eisenbarth, 2002; Pietropaolo et al, 2003; Lieberman et al, 2003; Wenzlau et al, 2007; Ludvigsson, 2009; Stadinski et al, 2010), anti-insulin autoimmunity remains as perhaps the most important driving force for the initiation and progression of the disease (Wegmann and Eisenbarth, 2000; Pugliese, 2005;

Bluestone et al, 2010). Autoantibodies specific to insulin can be detected in patients years before the clinical onset of diabetes, and their presence has been used as one of the key biomarkers (together with anti-GAD65 and IA-2 autoantibodies) to predict diabetes onset in genetically high risk individuals (Casu et al, 2005; Pietropaolo et al, 2008; Morran et al, 2010). Similarly, T-cell clones with T-cell receptors (TCRs) specific for insulin have been identified as the predominant population of islet infiltrating T-cells in prediabetic NOD mice.

Further characterization of six independently derived insulin-specific CD4

T-cell-

clones show that they all reacted to a region of the insulin molecule defined by a

synthetic peptide encompassing residues 9-23 of the insulin B chain. Is proven that

all six clones can either accelerate diabetes in young NOD mice or adoptively

transfer the disease to NOD severe combined immunodeficiency (SCID) mice

(Wegmann et al, 1994; Daniel et al, 1995). In addition, multiple other molecules

as well as T-cell clones reacting with unknown molecules are reported to be

autoantigens, e.g. islet glucose-6-phosphatase catalytic subunit-related protein

(Martin et al, 2002; Lieberman et al, 2003). However, insulin may have a primary

function, because the immune response against IGRP could be prevented by

inducing tolerance to (pro)insulin in NOD mice. Insulin-reactive T-cells are still

detected in the absence of IGRP-reactive T-cells, whereas IGRP-reactive T-cells are undetectable when insulin-reactive T-cells are eliminated (Krishnamurthy et al, 2004).

Insulin is the only autoantigen for which transgenic overexpression in antigen presenting cells (APCs) affects the disease progression in NOD mice (Gianani and Eisenbarth, 2005). Moreover, when a metabolically active, mutant form of insulin (which has alanine in place of tyrosine at the 16

amino-acid position of the B chain) is transgenically expressed in insulin knockout NOD mice, insulitis and the development of autoimmune diabetes are totally abolished (Nakayama et al, 1995). The altered sequence presumably changes the antigenicity of the dominant insulin B9-23 peptide. The observation in genetically-manipulated laboratory mice may be relevant to human T1D as B9-23-specific T-cells could be demonstrated in freshly isolated lymphocytes from patients with recent-onset T1D as well as from subjects at high risk for the disease (Alleva et al, 2001). Recently it was shown that strengthening peripheral insulin immune tolerance through immunizing prediabetic NOD mice with a dominant insulin mimetope can effectively prevent diabetes onset and anti-islet autoimmunity progression, further implicating the pivotal roles of anti-insulin autoimmunity in islet destruction (Daniel et al, 2011).

Indeed, the human T1D locus with the second highest effect magnitude (IDDM2),

maps to chromosome 11p15.5 within and just upstream of Ins (the gene that

encodes the pre-proinsulin peptide). At a variable number of tandem repeats

(VNTR) polymorphism located 0.5 kb 5’ to Ins (Lucassen et al, 1993; Bennett et

al, 1995), the short (class I) VNTR alleles (26 to 63 repeats) predispose to T1D

(odds ratio >2, allele frequency 0.7), whereas long (class III) alleles (140 to 210 repeats, frequency 0.3) are codominantly protective (Barratt et al, 2004).

Insulin autoantibodies in newly diagnosed T1D patients were found to be associated with the Ins VNTR polymorphism in some (Graham et al, 2002; Butty et al, 2008) but not all the studies (Perez De Nanclares et al, 2004). One study compared the Ins VNTR polymorphism between Finland and Sweden (Laine et al, 2007). The T1D risk genotypes (Class I/I and I/III) were significantly more common in Finland than in Sweden, both among patients and controls. Class III homozygous genotypes showed varying degrees of protective effect due to polymorphisms within Class III. These observations suggest that heterogeneity between protective Class III lineages could exist. However, it is important to note that the frequency of disease-associated Class I haplotype as such was significantly associated with T1D in Japanese (Awata et al, 2007).

The emerging model is that genetic variants that reduce thymic Ins expression lead to incomplete immune tolerance to insulin. Pancreatic Ins expression is only marginally affected by Ins VNTR alleles (Vafiadis et al, 1996). However, Ins expression in the thymus is more than twofold higher from the protective class III VNTR alleles compared with the class I VNTR alleles (Vafiadis et al, 1997;

Pugliese et al, 1997). Studies of the human thymus is complicated by the relative

inaccessibility of this tissue, but it is interesting to note that in one of the few

studies on the subject thus far the Ins VNTR Class III allele, in a homozygous or

heterozygous state, has been shown to promote regulatory type IL-10-producing

CD4+ T-cell responses (Durinovic-Belló et al, 2005). These findings suggest that

lower levels of insulin in the developing thymus lead to T1D autoimmunity.

Unlike human and other mammals, rodents have two insulin genes, Ins1 and Ins2. The two proteins are very similar in structure; the mRNA varies only by two amino acids in the B chain and several in C peptide and leader sequence of preproinsulin. Although both are expressed in the pancreatic β-cells and are able to sustain glucose homeostasis autonomously, Ins2 is the predominant form of insulin expressed in the thymus (Duvillié et al, 1997; Chentoufi et al, 2002;

Faideau et al, 2006). Using both Ins1 and Ins2 mutant mice, Chentoufi and Polychronakos showed that thymic insulin expression levels are inversely correlated with anti-insulin autoimmunity. In details, they established a very elegant mouse model in which mice lack the expression of either one or two copies of the two insulin genes. While insulin production in pancreatic β-cells is largely unaffected, thymic expression is dependent on the number of gene copies present.

Mice expressing low thymic insulin levels presented spontaneous peripheral

reactivity to insulin and the C-peptide, whereas mice with normal levels showed

no significant response (Chentoufi et al, 2002). Thus, this work provides

functional evidence that thymic insulin levels play a key role in the selection of

insulin specific T-cells and support the concept that variation in Ins levels,

determined by allelic variation at the IDDM2 locus, is the mechanism by which

this locus influences disease risk. Another support to this hypothesis came from

studies that utilized NOD mice overexpressing the Ins2 gene under the MHC class

II promoter (French et al, 1997). The increased levels of Ins2 expression in the

thymus of the transgenic mice compared with non-transgenic NOD mice were

associated with the complete prevention of insulitis and diabetes. Using a

contrasting approach, NOD mice with mutant Ins2 genes (NOD Ins2

) developed

more aggressive forms of autoimmune diabetes, displaying earlier onset (15-20 weeks in NOD Ins2

mice as compared to 25-32 weeks in control females) and a higher penetration (90-100% in both male and females as compared to 30% in males and 80% in females of NOD mice) (Thebault-Baumont et al, 2003).

Furthermore, it was recently shown in a humanized NOD model that Ins2- deficiency augments the cytotoxic CD8

T-cell response to (pro)insulin in the context of human T1D-associated HLA-A*0201 allele, implicating a facilitative role of deficient thymic insulin expression in promoting CD8+ T-cell responses to insulin-secreting beta cells (Jarchum and DiLorenzo, 2010). In contrast, knocking- out the Ins1 gene in NOD mice had only minor effects on thymic insulin expression. Diabetes and insulitis are markedly reduced in Ins1

NOD female mice, with virtually no mice developing insulitis and diabetes. Heterozigous mice also show a decreased incidence of diabetes as well as a delayed appearance of symptoms. However, Ins1

express insulin autoantibodies at level similar to those produced by the wild-type NOD female mice, suggesting that the production of insulin antibodies may not be controlled by the expression of insulin in the thymus. Prevention or delay of disease progression suggested that different disease-promoting epitopes were derived from the proteins products encoded by the Ins1 and Ins2 genes (Moriyama et al, 2003). Altogether, these findings further demonstrate the immunomodulatory roles of thymic insulin expression in T1D progression.

The molecular mechanisms underlying the Ins VNTR haplotype-dependent insulin

expression are still unclear. The predominant hypothesis that these VNTR regulate

the insulin expression levels in the thymus by affecting the autoimmune regulator

(AIRE) binding to its promoter region (Anderson et al, 2002; Pugliese et al, 1997;

Vafiadis et al, 1997), was recently confirmed in vitro (Cai et al, 2011). Direct interactions between AIRE and the VNTR have been reported but need confirmation. The finding that AIRE seems to interact with chromatin characterized by specific modification of histones might suggest another mechanism of interaction with the VNTR (Koh et al, 2008).

PTPN22

A relatively new member of the T1D susceptibility gene set is PTPN22, which

encodes the lymphoid protein tyrosine phosphatase (LYP), a negative regulator of

T-cell kinase signaling that is crucial to the balance between host defense and self-

tolerance. An association of a nonsynonymous SNP in PTPN22 at position 1858

with T1D (Bottini et al, 2004; Zheng and She, 2005; Steck et al, 2009) as well as

many other immune diseases (Criswell et al, 2005) has been reported from many

populations, with an OR of 2-3 for the homozygous TT genotype. The C1858 SNP

results in a missense mutation (R620W) that changes an arginine at position 620 to

a tryptophan and thereby abrogates the ability of the molecule to bind to carboxyterminal SRC kinase (CSK), which inhibits LYP (Begovitch et al, 2004;

Steck et al, 2006). The polymorphism has been associated with a gain-of-function mutation (Vang et al, 2005). As a consequence, upon TCR activation, T-cells that are Arg/Trp heterozygotes show diminished expression of the activation marker CD25 and secrete less interleukin 2 (IL-2) and interleukin 10 (IL-10) compared to Arg/Arg homozygotes (Fiorillo et al, 2010). This apparent downregulation of TCR signalling might interfere with proper tolerance induction in the thymus or the periphery.

IL2RA

Allelic variation in the interleukin (IL)-2 receptor-α gene (IL 2RA) region

accounts for another genetic risk factor implicated in T1D (Vella et al, 2005; Lowe

et al, 2007, Qu et al, 2007). The α chain of the IL-2 receptor complex (IL2Rα,

CD25) is an essential molecule upregulated in effector T-cells upon activation, but

is constitutively highly expressed in forkhead box P3 (FOXP3)

-regulatory CD4

cells (Sakaguchi et al, 2011). Following an initial association report, fine mapping

has pointed to a protective effect (OR = 0.6) of a low-frequency allele (Lowe et al, 2007). A reported second independent SNP effect showed no association in a replication study on >10.000 DNA samples from the T1DGC, and its role remain uncertain. Functionally, the protective allele is associated

with

higher expression of IL2RA mRNA in lymphoblastoid cells and peripheral blood mononuclear cells (PBMCs) (Qu et al, 2009). Examination of subsets of these cells by flow cytometry showed that IL-2Rα expressed from the protective allele was higher in CD4+ memory cells but lower in stimulated CD14+ and CD16+ monocytes (Dendrou et al, 2009), revealing complex, cell-specific regulation. In NOD mice, decreased IL-2 expression from the risk allele results in suppression of regulatory T-cells (Yamanouchi et al, 2007). In multiple sclerosis (MS) (Maier et al, 2009) and other autoimmune conditions (Giordano et al, 1988; Greenberg et al, 1988;

Adachi et al, 1989), increased levels of soluble IL2Rα (sIL2Rα) are found in

circulation. Given the indispensible role of IL-2 in Treg function and the potential

for sIL2Rα to neutralize IL-2, one could argue that IL2RA allelic risk variants

impair Treg functionality by upregulation of sIL2Rα. However, it was recently

found that IL2RA susceptibility genotypes in T1D are associated with lower levels

of sIL2Rα (Lowe et al, 2007; Maier et al, 2009). Furthermore, in vitro stimulated

peripheral blood mononuclear cells (PBMCs) from individuals with T1D make

less sIL2Rα than those from control individuals (Giordano et al 1989). This could

indicate a defect in the cellular subset that is the source of cleaved IL2Rα. An

alternative explanation may be that, even in the presence of normal Treg

frequencies in T1D, IL2RA polymorphisms account for functional defects in the

Treg compartment (Brusko et al, 2005; Qu et al, 2009). In conclusion, it seems

that although genetic variability in the IL2RA gene is associated with several autoimmune diseases including T1D, the mechanisms and extent to which sIL2Rα levels mediate these conditions differs significantly.

Other candidate genes

Besides the well-established non-HLA loci, a number of other associations with T1D have been reported for genes involved in functions that are related to the T - cell mediated adaptive immune response and tolerance mechanisms.

The cytotoxic T-lymphocyte-associated protein 4 (CTLA4) gene encodes a T -cell- specific transmembrane co-receptor. Like LYP, it is also an important negative regulator of T-cell activation, as evidence by the severe lymphoproliferative disorders seen in knock-out mice (Waterhouse et al, 2005). SNPs have been described in the human CTLA4 promoter region and exon 1. The G allele of the first exon (Ala17Thr) has been most consistently associated with T1D (Kavvoura and Ioannidis, 2005) and reduced control of T-cell proliferation (Kouki et al, 2000). This allelic variant affects glycosylation of the mature CTLA4 protein (Anjos et al, 2002), but its role in T1D is unclear, as its genetic effect can be entirely accounted for by more strongly associated SNPs in the 3′ flanking region.

The predominant hypothesis in humans, however, is that these SNPs lower the mRNA levels of the soluble CTLA4 splice variant (Ueda et al, 2003).

Interferon induced with helicase C domain 1 (IFIH1) belongs to a family of

helicases that are capable of eliciting an interferon response upon sensing viral

double-stranded RNA (dsRNA) (Kato et al, 2006), which is consistent with the

possibility that enteroviruses have a triggering role in T1D autoimmunity. A genetic defect in IFIH1 could potentially interfere with proper detection and clearance of viral infections and lead to an abnormal, diabetogenic immune response. A common IFIH1 SNP (Thr946Ala) is associated with T1D (odds ratio

= 0.85) (Zouk et al, 2010). An independent study confirmed the presence of T1D- associated polymorphisms in the IFIH1 gene and showed that gene expression levels in PBMC are higher in individuals with the susceptible genotypes (Liu et al, 2009). A plausible hypothesis is that in these individuals, the IFIH1 pathway, leading to exacerbated antiviral immunity and production of type I interferons, may primarily recognize viral infections. In line with this hypothesis, by resequencing a large number of cases and controls, low-frequency loss-of- function IFIH1 variants were discovered that confer strong protection (odds ratio = 0.5) (Nejentsev et al, 2009). Therefore, it can be assumed that the predisposing allele of the common IFIH1 SNP is associated with gain-of-function. Preliminary in vitro studies did not show an effect of IFIH1 Thr946Ala on a dsRNA-triggered interferon response (Chistiakov et al, 2010), raising the possibility that it is merely a tag for an additional, functional regulatory variant.

Another gene reported to have a functional polymorphism associated with T1D is

Ubiquitin-associated and SH3 domain containing A (UBASH3A; also known as

STS2), specifically expressed in lymphocytes. Like PTPN22, it encodes a tyrosine-

specific phosphatase that is likely to down-regulate the TCR activation response

(Carpino et al, 2004). The UBASH3A allele conferring risk to T1D (OR = 1.15)

(Concannon et al, 2008) is associated with gain-of-function through a higher level

of expression of UBASH3A in lymphoblastoid lines (Dixon et al, 2007).

The effects of these T1D loci on functional aspects of the adaptive immune response are consistent with the emerging model that events triggered by TCR signaling are attenuated in T1D, either through loss-of- function of activating signals or gain-of-function of inhibitory signals, thus compromising self-tolerance.

Environmental triggers

However, epidemiologic evidence of an increase in T1D incidence in many

countries over the past 2-3 decades (Onkam et al, 1999; Patterson et al, 2009)

indicates that environmental factors are also important in modulating the onset of

the disease in genetically susceptible individuals. Among these, infections,

chemicals and components of early childhood diet have been associated with the

onset of T1D. Extensive circumstantial data designate enteroviruses, and more

specifically coxsackieviruses, as prime viral candidates (Luppi et al, 1999; Filippi

et al, 2008). The possibility of a causal relationship has been evaluated in both

human studies (Tuvemo et al, 1989; Frisk et al, 1992; Conrad et al, 1994; Conrad

and Trucco, 1994)) and animal models (Yoon et al, 1986) and it has received new

attention, though there remains considerable controversy with some prospective

epidemiological studies failing to find an association. A study from Frisk showed

that newly diagnosed T1D children have high titers of neutralizing antibodies to a Coxsackie viral strain (CVB-14 VD22921), which has been shown to cause persistent infection in human pancreatic islets (Frisk and Tuvemo, 2004). The islet cell tropism of enteroviruses and its possible association with T1D has been reported in a study (Ylipaasto et al, 2004) that demonstrated enteroviral sequences by in situ hybridization in a subset of pancreas from diabetic patients.

Immunohistochemical detection of enterovirus protein in the pancreatic tissue of diabetic patients, confirmed by sequencing, was shown by Dotta et al (Dotta et al, 2007). Whereas the presence of enterovirus particles in pancreatic islets suggests that T1D is a consequence of selective viral infection of β cells, data favoring alternative mechanisms have been reported. The striking sequence similarities between the 2C protein from coxsackievirus and glutamate decarboxylase, lead to the postulation of viral mimicry in the etiology of T1D (Kaufman et al, 1993).

Subsequent results argued both in favor (Atkinson et al, 1994; Tian et al, 1994) and against such a mechanism (Ritcher et al, 1994; Horwitz et al, 2001; Schloot et al, 2001). The timing of enterovirus infection in relation to T1D onset is a controversial issue. In addition to demonstrable traces of infection in recent-onset individuals (Luppi et al, 2000), enterovirus infections have been also demonstrated in prediabetic autoantibody-positive children (Sadeharju et al, 2001). Taken together these data suggest that viral infection triggers the autoimmune response.

However, data from nonobese diabetic (NOD) mice show that preexisting insulitis

is required for coxsackievirus to induce diabetes (Serreze et al, 2000; Horwitz et

al, 2001; Drescher et al, 2004). Translating this to the human situation, susceptible

individuals may have ongoing subclinical insulitis for years until viral challenge instigates an acceleration of beta-cell destruction and hyperglycemia.

Other viral infections including rubella (Gale, 2008), mumps (Honeyman, 2005;

Goto et al, 2008), rotavirus (Honeyman et al, 2000; Makela et al, 2006), parvovirus (Guberski et al, 1991; Kasuga et al, 1996) and cytomegalovirus (Pak et al, 1988) have been associated with the initiation of T1D in humans (van der Werf et al, 2007) potentially due to molecular mimicry between viral proteins and β cell antigens, cross-presentation of β cell antigens released from damaged β cells or bystander mechanisms during virus-mediated immune-responses (Wen et al, 2005). However they remain to be confirmed in large patient populations. There is evidence that multiple viral infections prevent diabetes in young NOD mice (Oldstone et al, 1990), which supports the notion that microbial infection can protect rather than precipitate autoimmunity. Enhanced incidence of T1D worldwide may be related to cleaner environments reducing the exposure of newborn babies or infants to microbial infection referred to as “hygiene hypothesis” (Kolb et al, 1994; Bach, 2001; Christen et al, 2005). Introduction of general childhood immunizations and the growing prevalence of T1D in developed countries seemed to happen concurrently. However, multiple large studies found no support for any causal relation between vaccination and T1D (Blom et al, 1991;

EURODiab, 2000; DeStefano et al, 2001; Hviid et al, 2004).

Studies are under way to determine whether dietary factors contribute to the

development of T1D. Initial epidemiologic reports that bovine milk, and in

particular its albumin component, contributes to diabetes (Karjalainen et al, 1992)

have not been replicated in studies from Denver (Norris et al, 1996), Germany

(Ziegler et al, 2003) and Australia (Couper et al, 1999). The TRIGR (Trial to Reduce IDDM in the Genetically at Risk) trial will test whether hydrolyzed infant formula compared with cow’s milk-based formula decreases risk of developing T1D in children with genetic susceptibility (TRIGR Study Group, 2011). Recent studies from BABYDIAB study of Germany (Ziegler et al, 2003) and the Diabetes Autoimmunity Study in the Young (DAYSY) study of Denver (Norris et al, 2003) implicate that early ingestion of cereals as a factor increasing development of anti- islet autoantibodies.

Environmental factors therefore appear to have a modulating, and perhaps disease-

triggering role in susceptible individuals, while is well established that a specific

genetic constitution is required for such an event to cause diabetes.

T1D and possible mechanisms of tolerance induction

From immunological studies on human subjects and animal models, it is known that T1D involves cell-mediated adaptive immunity: β-cells are destroyed by infiltrating T lymphocytes whose T-cell receptors (TCRs) recognize β-cell antigens in the context of human leukocyte antigen (HLA). For TCR recognition, antigens must be presented by HLA class I or II molecules. T-cell activation is crucial for the deletion of self-reacting TCRs, and partial loss-of-function in the chain of events that are related to TCR activation may compromise self-tolerance, leading to autoimmunity (Polychronakos and Quan, 2011).

The thymus is the central lymphoid organ responsible for the maturation and differentiation of bone-marrow-derived thymocytes. The random generation of the T-cell repertoire, including autoreactive T-cells, is regulated in the thymus by mechanisms of central self-tolerance. Anatomically, the thymus is divided into subcapsular, cortical and medullary compartments. The stromal cells include a variety of bone-marrow-derived professional antigen-presenting cells (dendritic cells, macrophages and B cells) and endoderm-derived cortical thymic epithelial cells (cTEC) and medullary thymic epithelial cells (mTEC) (Boyd et al, 1993).

Thymic selection of the T-cell repertoire is key in many autoimmune diseases.

Deletion or inactivation of self-reactive T-cell depends on whether developing T-

cells “see” an autoantigen. While it is conceivable that blood-borne antigens can

be captured and engulfed by the bone marrow-derived APCs and be trafficked

back to the thymus, the mechanism of thymic presentation of autoantigens of

tissue-specific nature remained elusive until about a decade ago, when above-the- noise levels of transcripts of tissue-specific-antigens (TSA) were found in the mTEC (Kyewski and Klein, 2006).

The first reports of tissue-specific antigen transcription resulted from the creation of transgenic mice expressing target genes under control of the rat insulin promoter, although these systems were originally designed to target gene expression to the pancreas. Surprisingly, these transgenes and several others with tissue-specific promoters were activated in the thymic stroma, leading to the deletion of antigen-specific T-cells and establishment of tolerance (Hanahan, 1998). This promiscuous gene expression extends the scope of central tolerance to virtually all tissues of the body. Later, the essential role of central tolerance in establishing and maintaining a T-cell repertoire that is tolerant towards peripheral tissue antigens, preventing organ-specific autoimmunity, has been convincingly demonstrated in various experimental models (Klein et al, 2000; Anderson et al, 2002; Chentoufi and Polychronakos, 2002; Miyamoto et al, 2003; Thebault- Baumont et al, 2003; Liston et al, 2004).

One of the master regulators of TSA ectopic expression in the thymus is the Aire

gene, on chromosome 21q22.3. The gene is approximately 13kb in length and the

coding sequence is composed of 14 exons coding for a putative protein of 545

amino acids with a predicted molecular mass of 57.5 kDa. The Aire protein

contains motifs indicative of a transcription regulator including a conserved

nuclear localization signal, two PHD zinc-finger motifs, a SAND domain, four

LXXLL nuclear receptor binding motifs and a proline-rich region (Pitkanen and

Peterson, 2003). Loss-of-function Aire single mutations are responsible for a very

rare autosomal recessive disease named autoimmune polyendocrinopathy, candidiasis and ectodermal dystrophy (APECED) or autoimmune polyendocrine syndrome type 1 (APS-1). The Aire gene is expressed predominantly in the mTEC of the thymus, suggesting that its mechanism of autoimmune prevention could be to facilitate immune tolerance within the thymus, perhaps by maintaining normal thymic architecture or regulating interactions with thymocytes undergoing negative selection (Heino et al, 1999;

). Furthermore, the detection of Aire specifically within nuclear speckles suggested that Aire might function by directly regulating gene expression within the thymus (Heino et al, 1999). Interestingly, mRNA in situ hybridization also showed the presence of Aire transcripts within both the lymph nodes and spleen (Heino et al, 2000), and reverse transcriptase polymerase chain reaction (RT-PCR) of peripheral blood showed that human Aire message was detected in CD14+ monocytes and dendritic cells (Kogawa K et al, 2002).

Because Aire and peripheral antigens were both expressed in mTEC, it was

speculated that Aire might control the transcription of these antigens. Two

independent strains of Aire-deficient mice were generated and both the lines

exhibited multi-organ autoimmunity with lymphoid infiltration of target tissues

and serum autoantibodies (Anderson et al, 2002; Ramsey et al, 2002). It was

found, through bone marrow chimera and thymic transplant experiments, that Aire

deficiency precipitated disease when it was absent from the radioresistant stroma,

but did not substantially affect autoimmune process when lacking from the

hematopoietic compartment (Anderson et al, 2002; Kuroda et al, 2005). Gene

expression profiling of sorted mTECs from the Aire knockout mice and wild-type

littermates revealed that, Aire-deficiency is estimated to result in the downregulation of potentially thousands of genes (Derbinski et al, 2005), 80% of which are restricted in their expression pattern to one or several specific tissues.

Aire-regulated expression of TSAs is postulated to promote tolerance by driving the negative selection of self-reactive thymocytes that recognize Aire-regulated antigens with high affinity. Evidence for this concept stems from experiments using TCR transgenic models in which negative selection of clonal T-cell populations is defective in Aire-deficient mice (Liston et al, 2003; Liston et al, 2004; Anderson et al, 2005; Su et al, 2008). Additional evidence that Aire deficiency results in defective negative selection of self-reactive T-cells comes from studies in which decreased expression of a specific Aire regulated TSA is associated with autoimmunity against the specific tissue expressing the TSA.

Direct links between loss of Aire-regulated TSAs in the thymus and targeting of

these TSA by autoantibody production and cellular infiltration in the associated

tissue have now been demonstrated in a number of organs including the eye

(DeVoss et al, 2006), stomach (Gavanescu et al, 2007), prostate (Hou et al, 2009)

and lung (Shum et al, 2009). Of note, the interphotoreceptor retinoid-binding

protein (IRBP) was expressed within the thymus in an Aire-dependent fashion, and

when the thymus of an Irpb-knockout mouse was transplanted under the kidney

capsule of the thymus atrophic nude mouse, T-cell infiltration was detected in the

eyes after 10 weeks (DeVoss et al, 2006). Thus, the absence of a single self-

antigen in the thymus was sufficient to lead to autoimmune disease, highlighting

the role of this individual thymic tissue-specific antigen in maintaining tolerance.

Although Aire is primarily expressed in the thymus, RNA transcripts encoding Aire have also been detected in the peripheral lymphoid organs, from both mice and humans (Heino et al, 2000; Anderson et al, 2002; Kogawa et al, 2002).

Gardner et al showed that extrathymic Aire-expressing cells (eTACs), resident within the stroma of secondary lymphoid organs, express a diverse array of distinct self-antigens and are capable of interacting with and deleting naïve autoreactive T-cells. In addition an analysis of the role of Aire itself in directing expression of tissue specific antigens in eTACs was conducted by sorting out eTAC-enriched fractions from Aire wild-type and knockout mice and comparing their transcripts by microarray analyses. It was found that in eTACs there were roughly 160 genes differentially regulated by Aire, that these targets were enriched for TSAs, and that the identified targets were distinct from those, which were turned on in the thymus by Aire (Gardner et al, 2008), suggesting that peripheral Aire may complement Aire’s role in the thymus. The ability of eTACs to induce deletional tolerance among cognate T-cells and their expression of putative tissue- specific antigens imply that they may have a physiological role in maintaining peripheral tolerance, although the experiments required to formally make that determination have yet to be conducted.

Purpose of the study

Similar to the other TSAs, insulin expression has been found in the mouse thymus (Jolicoeur et al, 1994) and is regulated by the Aire gene (Faideau et al, 2006;

Chentoufi and Polychronakos, 2002). Mice lacking the Aire protein do not express proinsulin in the thymus, show an increased number of activated memory T-cells and display an autoimmune profile similar to mice subjected to neonatal thymectomy (Anderson et al, 2002). A recent study showed that the risk of autoimmune diabetes in APECED is associated with short alleles of the 5’-insulin VNTR (Paquette et al, 2010), leading to the hypothesis that the Aire protein may function as regulator of the insulin gene through the VNTR region. As above, different VNTR alleles correlate with different thymic insulin expression levels;

the greater the number of VNTRs, the more thymic transcripts, and the less diabetes.

It remains controversial which cells in the thymus are expressing insulin. APCs of bone marrow origin, including thymic dendritic cells and macrophages were reported to express (pro)insulin transcripts (Pugliese et al, 2001). In contrast, studies using purified thymic cell populations showed that (pro)insulin was only expressed in thymic epithelial cells of the endoderm origin (Palumbo et al, 2006).

In an attempt to resolve this controversy, Faideau et al created bone marrow

chimeras in Ins2 knockout mice and demonstrated that tolerance to (pro)insulin-2

was due to radioresistant cells in the thymus, presumably epithelial cells (Faideau

et al, 2006). To address the importance of thymic insulin expression in mediating

immune tolerance towards pancreatic β-cells even in the context of MHC allelic

resistance (Pietropaolo et al, 2002), we first examined the Ins2 ectopic expression in specific thymic cell types and identified endoderm-derived mTECs as the major Ins2-expressing cells in the thymus, consistent with previous publications (Anderson et al, 2002; Palumbo et al, 2006). To elucidate the essential role of this mTEC-insulin expression, we developed an animal model in which the mouse Ins2 gene was specifically deleted in mTEC, whereas its production in the pancreas remained intact. When these mTEC-Ins2-deleted animals also lacked (pro)insulin 1, the other isoform of the mouse insulin gene, spontaneous T1D developed around three weeks after birth.

This first part of the study demonstrated that disruption of thymic expression of a single tissue-specific self-molecule is sufficient to trigger autoimmunity towards the relevant tissue and results in pathologic damage even in the presence of disease-resistant alleles of MHC molecules.

Notably, TSAs expression, including insulin, was also reported in peripheral

lymphoid tissues, such as spleen and lymph nodes (Pugliese et al, 2001; Garcia et

al, 2005; Lee et al, 2007). It would seem reasonable that expression of genes

encoding for self-molecules in peripheral lymphoid organs may contribute to

tolerance. Although little is known about the function of insulin expression in LNs,

recent studies have shown that TSA-expressing stromal cells can effectively

deplete TSA-specific CD8+ T-cells from the peripheral repertoire (Magnusson et

al, 2008; Cohen et al, 2010; Fletcher et al, 2010). As for insulin expression in BM-

derived APCs, discordant results have been reported regarding the specific cell

subsets. Both (pro)insulin transcripts and proteins were detected in human

CD11c+ dendritic cells (DCs) in the thymus and peripheral lymphoid tissues

(Pugliese et al, 2001; Garcia et al, 2005). In contrast, Hansenne et al found neither Ins1 nor Ins2 transcripts in mouse CD11c high DCs, regardless of their maturation status (Hansenne et al, 2006). Nevertheless, transplantation of BM cells harvested from NOD Ins2+/+ mice failed to slow down diabetes progression in NOD Ins2-/- recipients, suggesting that endogenous levels of Ins2 expression in BM-derived cells of NOD mice cannot restore peripheral tolerance to insulin (Martin-Pagola et al, 2009). However, the failure could be attributed to the low levels of Ins2- expression in the transplanted BM cells, as it was shown that the levels of Ins2- expression in spleen and pancreatic lymph nodes (LN) decrease significantly after weaning (3-4 weeks) in NOD mice (Kodama et al, 2008). Thus, the role of insulin expression in secondary lymphoid tissues in regulating peripheral tolerance of β - cells remains elusive.

In the second part of the study, we generated animal models to better investigate the immunomodulatory function of insulin expression in BM-derived cells in establishing and/or maintaining peripheral immune tolerance to pancreatic β -cells.

We identified the insulin-expressing cells of BM origin as a population of

Aire

CD11c

B220

DCs that might play an essential role to restrict activation and

clonal expansion of insulin reactive T-cells in the periphery.