Different components and mechanisms of the immune system determine autoimmunity (). T cells reactive with self-antigens are believed to be eliminated by apoptosis in the thymus in a process referred to as central tolerance. In the thymus, the transcription factor AIRE (autoimmune regulator) is expressed in the medulla controlling transcription of a wide selection of organ-specific genes (including, for instance, preproinsulin) that generate proteins that are usually only expressed in peripheral tissues, creating an “immunological self-shadow” in the thymus (Anderson et al. 2002
). This should, theoretically at least, be sufficient to enable exclusion of autoreactive T lymphocytes. Yet, many T cells with autoreactive potential evidently slip through the thymus unharmed. It is not completely clear which factors contribute to the incomplete clearance of self-reactive T cells in the thymus, but this may relate to low binding affinity of self-peptide epitopes to human leukocyte antigen (HLA), low avidity of the T-cell receptor recognizing self-epitope in the context of HLA, and differences in posttranscriptional and posttranslational expression regulation in peripheral tissue versus thymus, as discussed below.
Figure 1. Representation of the role of β-cell autoantigens in the development of type 1 diabetes. Thymic events that predispose to impaired tolerance to β-cell proteins include low-level expression or presentation, crypticity, and low affinity. (more ...)
A backup mechanism controlling self-reactivity in the periphery is in place, referred to as peripheral tolerance, that involves regulation of autoreactivity by naturally occurring, nonspecific, and autoantigen-specific regulatory T cells (Treg). Because tissue cells only express major histocompatibility complex (MHC) class I molecules, direct recognition of tissue cells by the immune system occurs through presentation of self-epitopes on MHC class I molecules to CD8 T cells, but following removal and processing of tissue cell debris, autoantigenic fragments can also be expressed and presented by macrophages and dendritic cells to both CD4 and CD8 T cells in the pancreas-draining lymph nodes (through a process referred to as cross-presentation in the case of CD8 T cells). Consequently, both CD4 and CD8 T cells may contribute to both autoreactivity and the control thereof.
It is unclear which factors are involved in loss of tolerance to β-cell antigens in type 1 diabetes, but it has become evident that the β cell is actively involved in its own demise, rather than behaving as an innocent victim of immune aggression (Eizirik et al. 2009
). Viral infection or endoplasmic reticulum (ER) stress may lead to abnormal immunogenicity of β cells, leading to “danger” signals requiring activation of the immune system. Interestingly, one of the first abnormalities in pancreatic islets that seems to precede infiltration of leukocytes (insulitis) is hyperexpression of HLA class I and expression of cytokines of the innate immune system, such as interferon-α (possibly pointing to viral infection) and the chemokine CXCL10, attracting immune cells expressing its receptor CXCR3 into the islet microenvironment (Stewart et al. 1993
; Willcox et al. 2009
; Roep et al. 2010
). Presentation of islet antigen to the immune system may be followed by activation of additional T cells with different islet antigen specificities (determinant or epitope spreading) (Tisch et al. 1993
). It does not yet appear the case that there is a single primary islet antigen driving the autoimmune process in type 1 diabetes, nor is it known which islet antigens lead to pathogenic autoreactivity or bystander islet autoimmunity (Achenbach et al. 2005
; Steck et al. 2011
In terms of repertoire of (candidate) autoantigens, it is still a mystery why loss of tolerance to certain proteins expressed in islets as well as other tissues leads to tissue-specific pathology and disease. Whereas type 1 diabetes shows comorbidity with other inflammatory conditions, such as thyroiditis, celiac disease, Sjögren’s syndrome, and Addison’s disease, the vast majority of patients show islet destruction only. Yet, all known islet autoantigens are expressed elsewhere: preproinsulin (PPI) and IGRP are also expressed in the thymus; GAD65 and IA-2 are neuroenzymes expressed variably in the CNS, neurons, testis, and ovary; chromogranin A is expressed in neurons and non-β endocrine cells, and heat shock protein 60 is expressed in all cells carrying mitochondria. Perhaps even more of an enigma, murine β cells do not express GAD65, yet, immunotherapy with this protein prevents autoimmune diabetes (Karlsen et al. 1992
; Kaufman et al. 1993
). Whatever the mechanism, tolerance is broken and there follows activation and recruitment of cohorts of T lymphocytes recognizing islet autoantigens, which have an intricate involvement in the disease process.
Targets of CD4 T Cells
The autoreactive CD4 T cell is likely at the heart of this disease, as orchestrator of the immune attack on the β cell. It seems most probable and plausible that the breaking of CD4 T-cell tolerance to β-cell autoantigens is the key step on the pathway to disease. Several strategies have been applied to the investigation of CD4 T cells in type 1 diabetes, notably candidate autoantigen-driven cloning of T cells from the peripheral blood, as well as a variety of approaches to measure autoantigen-directed T-cell reactivity in blood samples immediately ex vivo (Atkinson et al. 1992
; Durinovic-Bello et al. 1996
; Hawkes et al. 2000
; Schloot et al. 2007
). To date, these approaches have offered important but limited data on antigens and epitopes. Cloning strategies have enabled numerous incisive mechanistic studies to be performed, but the reliance of this approach on the proliferative capacity of cells may lead to an inherent bias in the information it provides regarding cell type, epitopes, and HLA restriction elements. The opportunity to identify potential early or subdominant epitopes is also limited by this approach. Measurement of reactivity in blood samples ex vivo has been performed using whole antigen preparations, although these have well-documented nonspecific effects, or using panels of overlapping peptides (Roep et al. 1991
; Brooks-Worrell et al. 1996
; Schloot et al. 1997
; Dang et al. 2011
). The latter approach is successful when test and control populations are well matched, but it provides limited information as to HLA restriction elements, requires excessive volumes of blood for optimal performance, largely restricting its use to adult patients, and has the potential to mask or miss subtle issues of peptide length and register, according to the design of the overlapping peptide set. It is also inherently biased in its outcome by the nature of the method (e.g., proliferation) used to measure responses.
For these reasons, we contend that understanding the role of CD4 T cells in type 1 diabetes and translating this into tissue-specific immunotherapy requires a systematic dissection of the autoimmune T-cell response, addressing the nature of the epitopes targeted, their restriction elements, and the functional and phenotypic characteristics of responder T cells, as well as their frequency and temporal dynamics (Peakman et al. 1999
; Arif et al. 2004
). One of the problems in systematizing these studies is the interdependence of several of the parameters involved; for example, studying the frequency of autoreactive T cells may require knowledge about epitopes, HLA restriction, and functional responder type. As a consequence, although the CD4 T cell has been the object of numerous studies in type 1 diabetes, the overall advances in understanding have been modest.
Targets of CD8 T Cells
The epitopes identified to date for type 1 diabetes-specific human autoreactive CD8 T cells are primarily derived from β-cell antigens. Among the most important appear to be those from pre- (pro-) insulin (Pinkse et al. 2005
; Toma et al. 2005
; Skowera et al. 2008
). Previously, we showed that the presence of CD8 T cells reactive to the naturally processed insulin peptide B10–18
presented by HLA-A2 correlates with islet cell destruction, because increases in numbers of circulating CD8 T cells against insulin B10–18
are increased in type 1 diabetic islet allograft recipients that subsequently lose graft function and insulin independency (Pinkse et al. 2005
). Recently, the signal peptide of preproinsulin (the 10-mer peptide ALWGPDPAAA; PPI15–24
) was uncovered as being a naturally produced and presented HLA-A2 epitope (Skowera et al. 2008
). Cytotoxic CD8 T lymphocytes (CTL) against this peptide were cloned and killed β cells in vitro in a glucose concentration-dependent fashion, linking β-cell immunogenicity with its functional activity. This finding underscores the relevance of CD8 T-cell islet autoreactivity in the pathogenesis of T1D, whereas it also indicates that β cells are actively involved in their own demise. Interestingly, this PPI epitope provided high sensitivity and specificity in terms of frequent and preferential recognition by T cells from type 1 diabetes patients versus healthy subjects. The peptide appears to be cotranslated with HLA, and as a consequence this peptide-HLA complex showed the property of glucose regulation. Future studies should address whether this model is generally true for other epitopes derived from preproinsulin or other reported β-cell-derived antigens (e.g., GAD65, IA-2, IGRP, and preproIAPP) (Ouyang et al. 2006
; Velthuis et al. 2010
). A further intriguing aspect to the PPI studies is that an 8-mer peptide that coterminates with PPI15–24
) was also eluted from HLA-A2 and induced CD8 T-cell interferon-γ responses when added to patients’ peripheral blood mononuclear cells in vitro; and yet, this peptide has no measurable binding affinity for HLA-A2 in vitro, and as a result cannot be loaded into HLA-A2 multimers to investigate the autoreactive CD8 T-cell repertoire (unpublished observations). This illustrates the limitations to our understanding of the immunobiology of an autoimmune disease such as type 1 diabetes, when our approach and technologies are wholly shaped by those used to investigate infectious diseases, in which antigen-receptor interactions are typically high affinity/avidity.
Ideally, monitoring for the presence of CD8+
T cells reactive to all the above-mentioned epitopes simultaneously would be desired, posing considerable constraints on blood volumes accessible for monitoring of islet autoimmunity with conventional immune assays (Velthuis et al. 2010
). Initial approaches to monitoring of CD8 T cells reactive to β-cell-derived antigens required staining of a large number of, usually fresh, cells with HLA tetramers loaded with a single peptide, or in vitro culture for functional immune assays (proliferation, cytokine production [ELISPOT]) (Roep et al. 1999
; Peakman et al. 2001
; Schloot et al. 2003
). Furthermore, such detection of islet autoreactive T cells is hampered by their low precursor frequencies in the circulation, the broad possible range of candidate epitopes, the putative existence of regulatory T cells, and finally, as discussed above, the probable low binding affinity of peptide epitopes for HLA and low T-cell receptor (TCR) affinity for peptide-HLA (Naik et al. 2004
; Ouyang et al. 2006
; Velthuis et al. 2010
; Unger et al. 2011
). To cope with these limitations, new approaches have been developed for simultaneous detection of multiple islet cell-specific CD8+
T-cell responses, using multidimensional HLA class I multimers. Using this approach the sensitivity and specificity of previously reported HLA-A2 restricted epitopes were determined in recent onset type 1 diabetes patients, their siblings, control blood donors, and islet cell transplant recipients (Velthuis et al. 2010
). Frequencies of CD8 T cells reactive with insulin B10–18,
, and IA-2797–805
were found to be significantly increased in the circulation of recent-onset patients, whereas healthy subjects rarely showed autoreactive CD8+
T-cell frequencies >1/10,000 CD8 T cells in peripheral blood. Nonetheless, direct ex vivo responses as detected by HLA multimers often contrast with reports using functional assays, such as ELISPOT. This may partly reflect differences in activation status, as well as features influenced by in vitro phenomena during culture.
Functional studies on the engagement of T cells with these epitopes are less extensive (Ouyang et al. 2006
; Unger et al. 2011
). The relevance of insulin B10–18
in type 1 diabetes is underlined by observations that peripheral blood mononuclear cell cultures (PBMC) from a type 1 diabetic patient produced IFN-γ in response to this peptide, that expression of insulin B10–18
renders target cells sensitive to killing by CTL lines, and most importantly, that the presence of insulin B10–18
T cells correlates with destruction of β cells (Pinkse et al. 2005
). Even though it cannot be excluded that some of the T-cell reactivity to this particular epitope results from priming by exogenous therapeutic insulin containing this motif, the epitope is naturally processed and presented by β cells, rendering these as targets for autoimmune mediated cytolysis (Pinkse et al. 2005
It remains possible that not all of the β-cell specificities that are detected as being recognizable by CD8 T cells represent robust targets of autoimmunity. We speculate that some specificities could be redundant (i.e., the CD8 T cells exist), and may even be expanded or primed by antigen-presenting cells cross-resenting β-cell proteins; but that the β cell itself does not naturally process and present the cognate ligand. We would therefore advocate that as far as is possible, epitopes are verified by T-cell cloning and assays that show natural processing by β cells, the most sensitive of which is probably in vitro cytotoxicity using human β cells as targets.
Islet Epitopes and Thymic Selection
The thymus plays an important role in deletion of potentially autoreactive T cells. However, the presence of islet-specific T cells in the circulation of type 1 diabetes patients and healthy subjects indicates that the thymus is not capable of efficiently deleting all
autoreactive T cells (Mathis and Benoist 2009
). In addition to TCR affinity, a further influence on autoreactive CD8 T-cell escape from negative deletion is the strength of peptide binding to HLA, which influences occupancy and thus the strength of signal 1 provision. Based on experiences from pathogens, alloantigens, and tumor antigens, peptides binding with strong affinity to HLA are believed to be main epitopes for T cells. Yet, this rule of thumb most probably cannot apply to epitopes of autoreactive T cells, because peptides binding with high affinity would drive HLA occupancy in the thymus and hence autoreactive T-cell deletion (Ouyang et al. 2006
; Unger et al. 2011
). Peptides with relatively weak affinity for HLA may be more relevant in the periphery during autoimmune disease precisely because relevant T-cell specificities are allowed to escape. In the periphery, the density of expression of the index autoantigen may be much greater than in the thymus, enabling tolerance to be broken. Indeed, several recent reports point to unexpectedly low binding affinities of islet epitopes to HLA class I and II molecules (Ouyang et al. 2006
; Unger et al. 2011
). In the pancreas, where expression of antigens and epitopes is extended or amplified by the metabolic activity of the β cell, peptides with weak affinity for HLA may become highly represented, leading to recognition by T cells that escaped thymic deletion.
Posttranslational and Posttranscriptional Modifications
Potentially, there are further mechanisms through which islet-specific T cells may escape negative deletion. One important example of emerging interest is that epitopes may never be presented in the thymus, simply because they are generated as neoantigens in the periphery as a result of posttranslational modifications (Diez et al. 2001
; Mannering et al. 2005
; Dogra et al. 2006
). From our understanding of the pathogenesis of other autoimmune and inflammatory diseases it has become clear that proteins can be altered from their germline encoding sequence (which is thymically expressed) by a variety of cell biological modifications, and as a result give rise to new antigens against which tolerance does not exist. In rheumatoid arthritis, citrullination renders self-proteins subject to recognition by autoantibodies (van Venrooij and Pruijn 2000
). In celiac disease, enzymatic deamidation of glutamine into glutamate by tissue transglutaminase increases the immunogenicity of gluten peptides (Arentz-Hansen et al. 2000
). It is conceivable that similar or distinct modifications occur in the case of islet proteins creating immunogenic epitopes. The most notable proof of principle in the context of type 1 diabetes was provided by Mannering and colleagues (Mannering et al. 2005
). T cells recognizing the proximal A chain of human insulin were isolated from pancreatic lymph nodes of subjects with type 1 diabetes. T-cell recognition was dependent on the formation of a vicinal disulfide bond between adjacent cysteine residues at A6 and A7, which did not alter binding of the peptide to HLA-DR4, implying a novel posttranslational modification that is required for T-cell recognition of the insulin A chain.
Messenger RNA splice variation may also give rise to differential protein expression, distinguishing central (thymic) and peripheral (islet) proteomes. This has indeed been shown for the islet autoantigens IA-2 and IGRP (Diez et al. 2001
; Dogra et al. 2006
). Their expression in the pancreas is subject to posttranscriptional modifications, in which exon splicing, frame shifting, and truncations lead to changes in the original amino acid sequence. As a result, expression of neopeptide sequences in the pancreas, but not in the thymus, may generate epitopes for autoreactive CD8 T cells. Lack of full expression of the islet protein IGRP in the thymus may impair deletion of IGRP-specific CD8 T cells that are typically detectable in the peripheral blood of patients with type 1 diabetes (Velthuis et al. 2010
Polymorphisms in Islet Autoantigens
It is also plausible that polymorphisms in islet autoantigens may contribute to immunogenicity and autoimmunity, but any relationship with disease is currently still poorly understood. The most documented polymorphism relates to insulin itself. Insulin gene transcription levels in the human thymus correlate with allelic variation at the promoter region (INS VNTR-IDDM2
) that acts as a genetic susceptibility locus for T1D (Pugliese et al. 1997
; Chentoufi and Polychronakos 2002
). VNTR alleles correlate with differential INS
messenger RNA (mRNA) expression in the thymus where in contrast to the pancreas, protective class III variable number of tandem repeats (VNTRs) are associated with higher steady-state levels of INS
mRNA expression. This may explain a protective effect of class III VNTRs: higher levels of class III VNTR-associated INS
mRNA in the thymus leads to elevated levels of preproinsulin protein, thus enhancing central immune tolerance to preproinsulin.
With regard to polymorphisms in islet autoantigens themselves, a common nonsynonymous single-nucleotide polymorphism (SNP) in the SLC30A8
gene encoding ZnT8 influences ZnT8 autoantibody specificity in T1D (Wenzlau et al. 2007
). This SNP affects amino acid 325 and lies within the region of highest ZnT8 autoantibody binding. Patients with type 1 diabetes react primarily with the ZnT8 variant that they carry. Any relevance of polymorphisms of islet autoantigens with regard to immunogenicity and diabetogenicity is unresolved, but it is conceivable that variants not expressed by islet allograft recipients could act as minor histocompatibility (allo-)antigens, increasing the potential to immunize the recipient against β-cell autoantigens.