Journal of 64 (2015) 101e112

Contents lists available at ScienceDirect

Journal of Autoimmunity

journal homepage: www.elsevier.com/locate/jautimm

Review article Immunogenetics of : A comprehensive review

* Janelle A. Noble

Children's Hospital Oakland Research Institute, Oakland, CA 94609, USA article info abstract

Article history: Type 1 diabetes (T1D) results from the autoimmune destruction of insulin-producing beta cells in the Received 24 July 2015 pancreas. Prevention of T1D will require the ability to detect and modulate the autoimmune process Accepted 29 July 2015 before the clinical onset of disease. Genetic screening is a logical first step in identification of future Available online 10 August 2015 patients to test prevention strategies. Susceptibility to T1D includes a strong genetic component, with the strongest risk attributable to genes that encode the classical Human Leukocyte (HLA). Other Keywords: genetic loci, both immune and non-immune genes, contribute to T1D risk; however, the results of de- HLA cades of small and large genetic linkage and association studies show clearly that the HLA genes confer KIR Polymorphism the most disease risk and protection and can be used as part of a prediction strategy for T1D. Current Disease association predictive genetic models, based on HLA and other susceptibility loci, are effective in identifying the Type 1 diabetes highest-risk individuals in populations of European descent. These models generally include screening Prevention for the HLA haplotypes “DR3” and “DR4.” However, among racial and ethnic groups reduces the predictive value of current models that are based on low resolution HLA genotyping. Not all DR3 and DR4 haplotypes are high T1D risk; some versions, rare in Europeans but high frequency in other populations, are even T1D protective. More information is needed to create predictive models for non- European populations. Comparative studies among different populations are needed to complete the knowledge base for the of T1D risk to enable the eventual development of screening and intervention strategies applicable to all individuals, tailored to their individual genetic background. This review summarizes the current understanding of the genetic basis of T1D susceptibility, focusing on genes of the , with particular emphasis on the HLA genes. © 2015 Elsevier Ltd. All rights reserved.

Contents

1. Introduction ...... 102 2. HLA and ...... 102 2.1. HLA structure ...... 102 2.2. HLAgenes ...... 102 2.3. HLA polymorphism and nomenclature ...... 103 2.4. HLA genotyping ...... 104 3. HLA association with T1D ...... 105 3.1. DRB1-DQA1-DQB1 haplotype association ...... 105 3.2. DP and T1D risk ...... 106 3.3. HLA class I association with T1D risk ...... 106 3.4. HLA variation among populations ...... 107 4. Association of other immune genes with T1D ...... 108 4.1. Genes in the HLA region ...... 108 4.2. Genes outside the HLA region ...... 108 4.3. KIR association with T1D ...... 108 5. Genetic prediction of T1D ...... 109

* Children's Hospital Oakland Research Institute, 5700 Martin Luther King Jr. Way, Oakland, CA 94702, USA. E-mail address: [email protected]. http://dx.doi.org/10.1016/j.jaut.2015.07.014 0896-8411/© 2015 Elsevier Ltd. All rights reserved. 102 J.A. Noble / Journal of Autoimmunity 64 (2015) 101e112

6. Conclusions ...... 109 Acknowledgments ...... 109 References ...... 109

1. Introduction repertoire of peptides that can bind to a given HLA. The TCR rec- ognizes the combination of HLA and peptide, creating the tri- Susceptibility to type 1 diabetes, in which the immune system molecular complex that initiates the immune response (Fig. 2). destroys insulin-producing beta cells, has a strong genetic compo- nent, as evidenced by strong disease concordance in monozygotic 2.2. HLA genes twin pairs [1]. Four decades of study have shown that immune genes, specifically, those that encode classical Human Leukocyte The genes that encode HLA molecules are located on Antigens (HLA); confer the strongest genetic risk for disease. This review describes the complex genetics of the HLA region of the human (Section 2) and summarizes the current state of knowledge of HLA-associated T1D susceptibility, including differ- ences among populations (Section 3). Associations of other immu- nologically relevant genes are described, and recent studies investigating genes encoding Killer-cell Immunoglobulin-like Re- ceptors (KIR) on natural killer (NK) cells are presented in the context of T1D susceptibility (Section 4). Finally, the utility of immunoge- netics in the prediction of T1D risk is addressed (Section 5).

2. HLA and autoimmune disease

The association of Human Leukocyte Antigens (HLA) with type 1 diabetes was first reported in the 1970s [2e4]. Studies of HLA as- sociation with disease were prompted by the striking observation of the association of HLA-B*27 (called HL-Aw27 at the time) with ankylosing spondylitis [5]. At about the same time, HLA association was observed for several other autoimmune diseases, including (RA), celiac disease (CD), systemic lupus ery- thematosus (SLE), and (MS) [6e9], underscoring the importance of HLA in immune regulation. More than forty years Fig. 1. Schematic representation of HLA class I and class II molecules. The two chains of and thousands of genetic association studies later, genes in the HLA the HLA class II heterodimer are shown, both attached to the cell membrane, with a a b b region of the human genome remain the primary genetic risk de- immunoglobulin-like domains 1, 2, 1, and 2 labeled, and a peptide depicted in the binding groove. The single-chain HLA class I molecule is shown in a complex with b2- terminants for T1D, accounting for approximately half of the ge- microglobulin, with a peptide depicted in the binding groove. netic susceptibility to disease, estimated from studies of affected sibling pairs [10,11]. The genetics of HLA are complex and can be difficult to interpret. Understanding of HLA association with T1D requires a basic understanding of the of HLA. Details about the polymorphisms, structure, organization, and nomenclature of the classical HLA genes, as well as descriptions of available HLA genotyping methods, are presented in the following sections (2.1 through 2.4).

2.1. HLA structure

Classical HLA molecules are cell-surface proteins that bind and present peptide antigens for recognition by the T cell receptor (TCR). HLA is categorized into two classes, with three types of an- tigens in each class. Class I molecules are called A, B, and C and consist of a polypeptide chain that forms a heterodimer with the relatively invariant b-2 microglobulin. Class II molecules are called DR, DQ, and DP and consist of a heterodimer created from two polypeptides (a and b). Despite these differences, the overall structure of HLA class I and class II molecules is quite similar and is illustrated schematically in Fig. 1. The structure of HLA-A2 was solved in 1987, illustrating that the portion of the protein distal to the cell surface folds into a b-pleated sheet on which two a-helices Fig. 2. Schematic representation of the tri-molecular complex. The HLA and peptide create a groove into which peptides bind [12]. The shape of the complex on the presenting cell (APC) is shown binding to the T cell receptor to peptide binding groove and charges within it determine the form the tri-molecular complex. J.A. Noble / Journal of Autoimmunity 64 (2015) 101e112 103 6p21. The organization of the exons and introns is Table 1 similar among the HLA-encoding genes, with exon 1 containing 50 Distribution of additional coding DRB loci on according to DRB1 serogroup. untranslated sequence and encoding the signal peptide, and exons 2 and 3 (and 4 for class I genes) each encoding an immunoglobulin- DRB1 group Second DRB locus like domain that contributes to the extracellular portion of the 1,8,10 None molecule (Fig. 3). The peptide binding groove is formed by the 2 (15 or 16)a DRB5 domains encoded by exons 2 and 3 for class I (e.g., HLA-A) and by 3,11,12,13,14 DRB3 exon 2 from each of the two genes encoding a class II molecule (e.g., 4,7,9 DRB4 HLA-DQA1 and -DQB1); thus, DNA-based HLA genotyping systems a DR2 is split into two serologic groups, DR15 and DR16. Haplotypes are “ ” minimally include those exons. Each class I HLA (A, B, and C) is commonly referred to as DR2; however, individual allele names have the format DRB1*15:xx or DRB1*16:xx. encoded by a single gene (HLA-A, -B, and -C, respectively). For class II antigens, the heterodimer is formed from the products of two genes, e.g., HLA-DQA1 encodes the a chain and HLA-DQB1 encodes Table 2 the b chain of the DQ antigen, HLA-DPA1 encodes the a chain and Extreme polymorphism of the classical HLA loci. b HLA-DPB1 encodes the chain of the DP antigen, and HLA-DRA1 Locus Number of allelesa encodes the a chain and HLA-DRB1 encodes the b chain of the DR A 3107 antigen. Additional loci encoding DRb chains are found on some B 3887 chromosomes. Some additional DRB genes are psuedogenes, but C 2623 HLA-DRB3, -DRB4, and -DRB5 all produce a functional DRb chain DRA 7 that can dimerize with the product of the DRA1 gene. HLA-DRB3, DRB1 1726 -DRB4, and -DRB5 genes segregate as a single locus, i.e., a maximum DRB3 59 DRB4 15 of one of them is found on any given chromosome. HLA-DRB1 is DRB5 21 found on all copies of chromosome 6, but the other three functional DQA1 54 genes are only present on some chromosomes, dependent on the DQB1 780 DRB1 gene on the chromosome. Table 1 shows which of the three DPA1 39 DPB1 520 additional DRB coding genes is found on the same chromosome with each DRB1 gene group. a Numbers are taken from the IMGT/HLA database release 3.20.0, April, 2015.

2.3. HLA polymorphism and nomenclature peptide-binding groove. The basis for the polymorphism in these exons is thought to result from selection. The rapid increase in the HLA genes are the most polymorphic known in the human number of recognized HLA alleles has led to an evolution in the genome. While most genes are monomorphic or have a few variant nomenclature used to describe them. The current nomenclature sequences, classical HLA genes can have thousands (Table 2) [13,14]. was adopted in 2010 and is depicted in Fig. 4. Numeric fields As the table shows, most of the polymorphism for class II is found in separated by colons describe four levels of resolution. Previous the -B, rather than the -A gene. This is particularly true for DR- versions of HLA nomenclature did not employ colon delimiters, encoding genes, where the DRA1 gene is essentially invariant, and which were added because the number of variants in some allele the DRB1 gene is highly polymorphic. In cases where a second DRB groups exceeded 99. In general, the first field describes a serolog- coding gene is present, its pairing with the DRA1 gene represents a ically defined allele group, and the second field identifies a unique mechanism to increase the diversity of the DR antigens on the cell protein sequence encoded by the allele within that group. Fields 3 surface. For DQ, the genes encoding both polypeptides are poly- and 4 describe silent polymorphisms and non-exonic poly- morphic, increasing the number of potential cell-surface DQ mol- morphisms, respectively. The extremely large number of alleles ecules to four via combinatorial diversity (see Section 3.1). For DP, leads to the inability to make allele-level genotype calls with all but nearly all of the diversity is encoded in the DPB1 gene. The DPA1 the highest-resolution genotyping technology (discussed in Section gene has limited polymorphism; in general, only small number of 2.4, below), which can lead to difficulty in data interpretation DPA1 alleles is observed in a given population. Most of the poly- among studies performed with different genotyping methods. morphic sites in HLA genes are found in the exons encoding the Fig. 5 illustrates that a given allele can have multiple designations,

Fig. 3. Organization of the HLA-encoding genes. The structure of the single chain of HLA class I genes is shown, along with the class II A and B genes, encoding a and b chains of the HLA molecule, respectively. *for class I, the HLA-B gene lacks exon 8, and for class II, the DPB1 gene lacks exon 6. 104 J.A. Noble / Journal of Autoimmunity 64 (2015) 101e112

blanks. Adoption of DNA-based genotyping technology in the 1980s allowed better ability to identify and resolve the individual HLA alleles present in a subject and began an explosion in the number of known alleles that is still going on today. Early DNA technology used restriction enzymes and Southern blotting and was known as Restriction Fragment Length Polymorphism (RFLP) [15]. The advent of the Polymerase Chain Reaction allowed the development of PCR- based genotyping methods, including sequence-specific oligonu- cleotide probe hybridization (SSOP) [16], sequence-specific priming (SSP) [17], and sequence-based typing (SBT) [18]. Currently, next-

Fig. 4. HLA nomenclature. The figure illustrates the designation of an allele with the generation sequencing (NGS)-based methods are available for current standard nomenclature. HLA genotyping, using either exons only or larger gene segments, up to and including whole HLA genes [19,20]. Each of these ad- vances in technology has increased the level of resolution of the resulting genotype data. Fig. 6 depicts the relative levels of reso- lution of various HLA genotyping methods. The method selected for a given study depends on the intended use of the data. For example, full sequencing of HLA-B to support a diagnosis of ankylosing spondylitis would not be cost-effective when a simple clinical test for HLA-B*27 could be used [21]. On the other hand, the severe drug hypersensitivity to the nucleoside reverse transcriptase inhibitor abacavir exhibited by HIV patients carrying HLA-B*57:01 is not seen in patients carrying HLA-B*57:03, underscoring the need for at least two-field resolution of HLA genotyping for treatment decisions for Fig. 5. Alternate designations for HLA alleles. Five different designations are shown HIV positive patients [22]. that might be reported for each of two HLA-A alleles. In this case, only the highest level For a disease association study, the decision of which of the HLA genotyping can distinguish the two alleles from one another. many available HLA genotyping methodologies to use greatly af- fects the results. Low resolution genotyping results in a small depending on the level of resolution of the genotyping method, and number of categories with large numbers of data points in each that certain allele pairs, such as HLA-A:01:01:01:01 and HLA- category, leading to good statistical power, with the caveat that A:01:01:01:02N, can only be distinguished by the highest level of each category may contain alleles with different effects on disease genotyping resolution, including sequencing of intronic and un- susceptibility, possibly masking true effects of individual alleles translated sequence. Fortunately, for most disease association within the group. Conversely, allele-level resolution will allow the studies, the relevant information for function resides in the first effects of each individual allele, haplotype, or genotype to be two fields (usually four-digits), which include information on the examined; however, the number of data points for each category, amino acid sequence of the encoded protein. In most cases, silent especially when analyzing genotype associations, may be so low polymorphisms and polymorphisms in intronic and untranslated that statistical power will be insufficient to reveal disease associ- regions do not affect function; thus, two-field resolution is suffi- ation effects. Particularly for studies involving large numbers of cient. In the example given, however, the designation “N” in the samples, cost is a major factor in determining the genotyping allele name HLA-A:01:01:01:02N designates that the allele is not method. In general, higher resolution means higher cost. A well- expressed (null); thus, HLA-A:01:01:01:01 and HLA-A:01:01:01:02N designed study will incorporate a method that provides adequate are expected to be functionally different. resolution to address the scientific hypotheses being tested while remaining within the resources available for the study. For these 2.4. HLA genotyping reasons, HLA genotype resolution varies widely among published studies, and the interpretation of and comparisons among studies The technology for HLA typing is evolving at a rapid rate. His- must take this into account. tocompatibility testing began in the 1960s and was performed by In recent years, methods have been developed to impute ge- cell-based methods using serum from multi-parous women, notype data, i.e., assigning a genotype call to a given polymorphic particularly those with pregnancies from more than one father, position without direct experimental testing using data from which contained to one or more paternal HLA molecules nearby single polymorphisms (SNPs) and knowledge of that were present in the fetus. Extensive standardization of re- haplotype patterns within a given population. This is a highly cost- agents and protocols among HLA typing laboratories was required, effective method of data generation, since it does not depend on and typing assays resulted in very low resolution with frequent expensive laboratory genotyping procedures. Imputation can be

Fig. 6. Resolution of HLA genotyping methods. HLA genotyping methods are listed in order of increasing resolution. NGS ¼ Next generation sequencing. J.A. Noble / Journal of Autoimmunity 64 (2015) 101e112 105 effective in cases where a limited number of known haplotypes are controls, from populations around the globe. All samples were well defined in a homogeneous population. Both complexity of the genotyped for classical HLA loci, including HLA-DRB1, -DQA1, -DQB1, number of haplotypes and admixture or substructure of pop- -DPA1, -DPB1, -A, -B, and -C using PCR-SSOP methods [32] to provide ulations can greatly decrease the efficacy of these methods. Many HLA context for the discovery of other T1D associated genes. Much groups have tried to impute HLA alleles from SNPs in the HLA re- of the association data described below (Sections 3.1e3.3) come gion, with varying levels of success [23e28]. However, all groups from the T1DGC. realize the need for large datasets of classically determined HLA sequences from multiple populations to inform imputation strate- gies. Even in a Finnish population, for which HLA alleles are very 3.1. DRB1-DQA1-DQB1 haplotype association well characterized, imputation methods varied among loci, with the worst performance observed for DRB1 and HLA-B genes [23]. The highest HLA-associated T1D risk is attributable to class II Although calling HLA genotypes from existing genome-wide as- DR- and DQ-encoding loci. Genes in the HLA region exhibit sociation study (GWAS) data is attractive, one must remember that extensive linkage disequilibrium (LD), which is the tendency for these tag SNPs are simply surrogates for given HLA alleles. Very few specific alleles at two loci to be found together more often than SNPs on GWA chips lie within the HLA alleles themselves because of expected, given the distance between them and expected recom- the tremendous amount of variation within the genes. As with low- bination frequencies. This is particularly true for the genes encod- resolution HLA genotyping, most imputed HLA calls are consistent ing DR and DQ, such that, for a given population, an individual allele with a group of alleles (an ambiguity string), rather than a single is usually found in only one or a few haplotype combinations. In allele. Resolution for HLA genotyping by imputation may be as low many cases, the presence of one allele allows assumption of other or lower than that for serologic methods. In addition, although HLA alleles on the haplotype. For example, DRB1*03:01 is found coupled alleles and haplotypes are well established for populations of Eu- to DQA1*05:01 and DQB1*02:01 almost exclusively, to create the ropean descent, information for most other populations is far more haplotype DRB1*03:01-DQA1*05:01-DQB1*02:01, commonly called sparse, and, in most cases, not adequate to inform imputation al- “DR3” and known to be a risk haplotype for T1D. DRB1*04:01 is gorithms properly. found with DQA1*03:01 but can have either DQB1*03:01 or Current efforts in the HLA community are directed toward DQB1*03:02 included in the haplotype. Both are referred to as standardization of data reporting in HLA studies to allow compar- “DR4” haplotypes; however, the T1D risk for the two haplotypes is isons to be made among studies [29]. In the future, the time may very different, as shown in Table 3. This might lead to the come when becomes standard for every assumption that the DQ-encoding genes are the main risk factor for individual, negating the need for specific genotyping of any locus, T1D, and that the DRB1 locus is less, or not at all, important. including those encoding HLA. Currently, however, an under- However, comparison of risk for risk for the same DRB1*04:01- standing of HLA genotyping methods and nomenclature is essential DQA1*03:01-DQB1*03:02 haplotype to that for DRB1*04:03- for correct interpretation of disease association studies. DQA1*03:01-DQB1*03:02 haplotype could lead to the conclusion that the DRB1 locus is the primary contributor to T1D susceptibility, and the DQB1 locus matters less (Table 3). In fact, the total T1D risk 3. HLA association with T1D is most likely due to a combination of DR and DQ molecules [10,33,34]. Thus, assessing a single locus for T1D susceptibility can The earliest reported HLA association for T1D was with HL-A be misleading; analysis of DRB1-DQA1-DQB1 haplotypes is far more antigen “specificity W15” [2], which corresponds to what we now informative and desirable. Like most studies of T1D genetics, the call HLA-DRB1*04:01, a known risk allele for T1D. Subsequently, T1DGC was primarily comprised of samples from individuals of Nerup et al. demonstrated association with W15 and what was then European descent. Examples of the most predisposing and most called “HL-A8” [3], which corresponds to what is now called HLA- protective DRB1-DQA1-DQB1 haplotypes are given in Table 4 with B*08:01 and is found on the common, conserved “A1-B8-DR3” predisposing Odds Ratios shown in red font, and protective shown haplotype that is positively associated with T1D [54]. Family-based in green. studies confirmed these associations [4]. By 1980, the very high risk T1D risk depends not only on the haplotypic context but also on of having a heterozygous genotype with a haplotype including the genotypic context of the alleles being tested. One example of DRB1*03 on one chromosome and DRB1*04 on the other this is the extreme risk of the “DR3/DR4” heterozygous genotype (commonly referred to as DR3/4), was reported [30]. Thousands of (where the DR4 haplotype does not include either the protective reports have been published on the association of HLA with T1D. DRB1*04:03 or DQB1*03:01 allele), which is higher than the com- One of the largest T1D genetics studies was the Type 1 Diabetes bined risk for the individual DR3 and DR4 haplotypes. In meta- Genetics Consortium (T1DGC), an international effort to establish analysis, this genotype was reported to have an Odds Ratio (OR) resources to discover all of the genes responsible for susceptibility of 16.6 [35]. One hypothesis for the increased risk is the putative to T1D [31]. The T1DGC, conceived in 2000 and started in 2001, presence of DQ heterodimers encoded in trans, in addition to the collected over 16,000 subjects, including families, cases, and DQ molecules encoded in cis, on the cell surface. Each chromosome

Table 3 HLA T1D risk can be attributable to either DRB1 or DQB1.

DRB1 DQA1 DQB1 Odds Ratio Effect

04:01 03:01 03:01 0.35 protective

04:01 03:01 03:02 8.39 susceptible

04:03 03:01 03:02 0.27 protective 106 J.A. Noble / Journal of Autoimmunity 64 (2015) 101e112

Table 4 Selected susceptible and protective DRB1-DQA1-DQB1 haplotypes from the Type 1 Diabetes Genetics Consortium data [34].

haplotype DRB1 DQA1 DQB1 OR p value

DR3 03:01 05:01 02:01 3.64 2 x10-22

DR4 04:05 03:01 03:02 11.37 4 x10-5

DR4 04:01 03:01 03:02 8.39 6 x10-36

DR4 04:02 03:01 03:02 3.63 3 x10-4

DR2 15:01 01:02 06:02 0.03 2 x 10-29

DR6 14:01 01:01 05:03 0.02 1 x 10-6

DR7 07:01 02:01 03:03 0.02 3 x 10-4

DR7 07:01 02:01 02:01 0.32 2 x 10-9

DR4 04:03 03:01 03:02 0.27 0.017

at 5% [37]. In most studies of DP association with T1D, only the DPB1 locus is genotyped. The nomenclature for DPB1 alleles differs from that of other classical loci in that few serologic reagents were developed for DP typing prior to the availability of DNA-based genotyping. Thus, while alleles for other loci are clustered by serologic reactivity into groups designated in the first field of nomenclature, DPB1 alleles were numbered in order of their dis- covery, regardless of reactivity. With the exceptions of DPB1*02:02 and DPB1*04:02, allele names were designated by the next available number in the first field, followed by “01 ” in the second field (e.g., Fig. 7. Trans-encoded DQ heterodimers of the DR3/4 genotype. The DRB1-DQA1-DQB1 … … haplotypes in the common, high T1D risk DR3/4 genotype, found in Europeans, are *06:01, *07:01, *08:01, ). The discovery of more than 99 DPB1 shown. The trans-encoded DQA1*05:01-DQB1*03:01 haplotype is shown in the blue alleles was one of the factors leading to the adoption of the colon ellipse, and the trans-encoded DQA1*03:01-DQB1*02:01 haplotype is shown in the red delimiters in the latest nomenclature convention (Fig. 4). ellipse. (For interpretation of the references to color in this figure legend, the reader is Like DR and DQ, DP is an HLA class II molecule that can bind referred to the web version of this article.) exogenous peptides and present them to CD4þ T cells. DPB1 alleles have been reported to be associated with several diseases, including pauciarticular juvenile rheumatoid arthritis and chronic encodes an a polypeptide chain from the DQA1 gene and a b beryllium disease [38,39]. For T1D, analyses of disease association polypeptide chain from the DQB1 gene. The a and b chains encoded for DPB1, or for other loci in the region, must account for LD with on the same chromosome can form heterodimers, and, in most the DR- and DQ-encoding genes on the chromosome [40,41]. Many cases, the a and b chains encoded on different chromosomes can studies report DPB1 association with T1D. Taken together, the data form heterodimers as well, thus allowing the possible expression of from these studies show consistently that DPB1*04:02 is protective four different DQ molecules on the surface of the cell. Fig. 7 illus- for T1D, while DPB1*03:01 and DPB1*02:02 increase T1D risk trates the DQA1 and DQB1 pairs found in trans on a typical T1D- [37,42e47]. predisposing DR3/DR4 genotype. The combination of DQA1*05:01 and DQB1*03:02 has not been observed encoded in cis but has been hypothesized to confer very high T1D risk [34]. The other trans- 3.3. HLA class I association with T1D risk encoded combination, DQA1:03:01 and DQB1*02:01, is encoded in cis on some haplotypes, particularly in African Americans, and The autoimmunity that leads to T1D results in the immune appears to increase risk for T1D on all haplotypes on which it is destruction of the insulin-producing beta cells in the pancreas, a found [36]. Of note, DQB1*02:01 can only be distinguished from process that is thought to occur by cytotoxic (CD8þ) T cell killing. In DQB1*02:02 by a polymorphism in exon 3 which does not addition to helping shape the T cell repertoire, class I HLA molecules contribute to the peptide binding groove. This ambiguity is some- present antigens to CD8þ T cells to initiate cytotoxic T cell killing; times referred to as “02:01g” with “g” representing “group.” thus, specific combinations of class I HLA and peptide are likely to DQB1*02:01 and DQB1*02:02 are presumed to have identical im- influence beta cell destruction. In support of this notion, HLA-A*24 mune function. correlates with low residual beta cell function in T1D patients [48]. Although HLA class II genes, specifically, those encoding DR and DQ, 3.2. DP and T1D risk are the most strongly associated with T1D risk, well-powered studies, including the T1DGC, have demonstrated class I associa- The DP molecule is encoded by the DPA1 and DPB1 genes. The tion with T1D, even after accounting for LD with the DR and DQ- DPA1 gene is far less polymorphic than most other HLA genes. In encoding genes [49e52]. HLA-B*39:06, with an odds ratio (OR) of fact, in the T1DGC data, only three alleles accounted for 98% of 10.31 after accounting for LD, appears to have the strongest pre- nearly 15,000 genotypes (30,000 allele calls), with DPA1*01:03 disposing effect on T1D risk, while HLA-B*57:01 is strongly signifi- present at 76%, DPA1*02:01 present at 17%, and DPA1*02:02 present cantly T1D protective (OR ¼ 0.19, p ¼ 4 10 11) [51]. J.A. Noble / Journal of Autoimmunity 64 (2015) 101e112 107

B*39:06 association has been observed in multiple studies, Table 6 with one study suggesting that it can improve T1D risk prediction, Possible DR3/4 genotype in a mixed-race individual with two significantly protective haplotypes. particularly in patients who carry DRB1*04:04-DQA1*03:01- DQB1*03:02, a moderately predisposing DR4 haplotype [53].The Haplotype DRB1 DQA1 DQB1 class I allele HLA-A*01:01 provides an illustrative example of the DR3 (African) 03:02 04:01 04:02 requirement for accounting for LD in T1D association analyses for DR4 (Asian) 04:03 03:01 03:02 genes in the HLA region. HLA-A*01:01 is part of the conserved, extended HLA haplotype known as “A1-B8-DR3,” which is asso- ciated with T1D risk [54].Thus,A*01:01 appeared predisposing for disease. However, when expected allele frequencies were individuals of European descent (~0.6%) [67]. The protective effect adjusted based on the strong LD of this allele with the DR3 of DRB1*04:03 on T1D is an excellent example of why any T1D haplotype, A*01:01 was actually significantly protective for T1D genetic screening program should resolve genotypes to at least two fi [50]. In addition to effects on T1D risk, HLA class I alleles have elds. < been shown to be associated with age of onset for T1D African Americans are a recently-admixed population ( 20 [52,55e59]. generations) that has been informative in the study of HLA- associated T1D risk [36]. Haplotypes categorized as DR3 and DR7 illustrate why the study of this population is particularly informa- 3.4. HLA variation among populations tive. DR7, a designation for haplotypes including the allele DRB1*07:01, is commonly seen as DRB1*07:01-DQA1*02:01- Among world populations, prevalence (proportion of popula- tion affected) of T1D is thought to be highest in people of European DQB1*02:02 in Europeans and is known to be protective for T1D [34]. However, a closely-related version of DR7, DRB1*07:01- descent at approximately 1 in 300. Not surprisingly, to date, the vast majority of T1D genetic studies have been performed on sub- DQA1*03:01-DQB1*02:02, found in African populations, is not pro- tective for T1D but is, in fact, predisposing for T1D [36,68,69]. jects of European descent. However, T1D is a worldwide problem. fi Incidence (cases per year) estimates for T1D in children under 15 Similarly, an African-Speci c DR3, DRB1*03:02-DQA1*04:01- DQB1*04:02, has the opposite T1D susceptibility effect (protec- years old vary from as high as 64 per 100,000 per year in Finland [60] to as low as 0.1 or less per 100,000 children per year in parts of tion) to that of the common, predisposing DR3 haplotype, DRB1*03:01-DQA1*05:01-DQB1*02:01, seen in most other pop- China, Venezuela [61], and Papua New Guinea [62]. Although more than 13,000 HLA alleles have been reported ulations [36,70]. The opposite effects of European and African DR3 [13,14], more than 90% of these alleles are rare, having been and DR7 haplotypes are illustrated in Table 5. observed only once or twice, with the remaining alleles designated Genetic screening for T1D risk has been a goal of T1D re- common and well-documented (CWD) [63]. Some common alleles searchers for decades. Many large studies (e.g., DAISY [71], TEDDY appear in nearly every population studied, e.g., DRB1*03:01; how- [72], TrialNet [73], TRIGR [74], and others) use a limited genetic ever, others are population specific. The set of alleles present in any screen to select subjects for follow up autoantibody testing and given population differs from, but usually overlaps with, the set intervention studies. World populations are becoming increasingly found in other populations. Even for alleles common to multiple admixed, especially in the United States. The genetic heritage of an populations, allele frequencies and haplotype combinations vary individual is not necessarily apparent from phenotype, nor can it be among populations. determined, in some cases, from interview. The potential danger of The study of HLA association with T1D in varied populations can a limited, low-resolution genetic screen is illustrated by Table 6,in reveal effects that are not apparent from the study of a single which a possible DR3/4 genotype is shown for a hypothetical in- population. To understand the effect of an allele or haplotype on dividual of mixed Asian and African ancestry. If this individual were T1D, the study population must include a high enough frequency of genotyped with a limited, low-resolution genetic screening test for fi the allele or haplotype to reveal the susceptibility effect. For DR, (s)he would be classi ed as very high diabetes risk, based on fi example, unlike most DRB1*04 alleles, the allele DRB1*04:03 is the nding of a DR3/4 genotype. However, both of the haplotypes actually protective for T1D (Tables 3 and 4) [64], but that effect was that comprise this particular DR3/4 genotype are protective for fi difficult to detect in studies of European populations until the large T1D. Genotyping of both DR- and DQ-encoding genes at two- eld resolution would clearly show that this individual is highly un- T1DGC data set was published [34,65,66]. DRB1*04:03 is seen at far higher frequency in individuals of Asian descent (~3.5%) than for likely to get T1D.

Table 5 Comparison of T1D susceptibility effects for DR3 and DR7 halotypes in two populations.

A. DR3 Population DRB1 DQA1 DQB1 OR*

European 03:01 05:01 02:01 4.64

African American 03:02 04:01 04:02 0.16

B. DR7 European 07:01 02:01 02:01 0.34

African American 07:01 03:01 02:01 3.96

*red font denotes T1D susceptible effect, and green font denotes T1D protective effect. 108 J.A. Noble / Journal of Autoimmunity 64 (2015) 101e112

The DR3/4 heterozygous genotype only accounts for up to likely attributable to LD effects of TNFA promoter alleles with DR3 approximately 40% of T1D patients; therefore, even for populations haplotype [81,83]. MHC class I chain-related gene A (MIC-A), a of European descent, simply searching for high-risk DR3/4 in- stress-induced antigen in gut epithelium, has also been studied for dividuals will exclude the remaining 60þ% of future patients that T1D association. Several reports implicate MIC-A5 as a T1D sus- do not have the highest-risk genotype. Understanding the risk ceptibility allele and MIC-A5.1 as a T1D protective allele [84e87]. conferred by individual alleles in haplotypic and genotypic context Complement C4 genes (C4A and C4B) exhibit copy number varia- will allow risk assessment for all individuals, regardless of racial or tion that is associated with autoimmunity [88]. Both the C4A null ethnic background. allele and the C4B “short” allele have reported association with T1D The need for the study of non-European populations to fully [89e93]. understand HLA susceptibility to T1D has been noted for more than 20 years [75]. The number of published studies on non-European 4.2. Genes outside the HLA region populations and T1D is increasing; however, the need for study of underserved and understudied populations remains a critical Cytotoxic T-lymphocyte-associated protein 4 (CTLA4) (gene component of a full understanding of HLA-associated T1D risk. located at 2q33) is a down-regulator of the immune response. T1D association for CTLA4 in multiple ethnic groups was noted nearly 4. Association of other immune genes with T1D two decades ago [94]. A recent meta-analysis of 58 studies confirmed CTLA4 association with T1D but noted that the associa- More than 40 genetic loci have been implicated in T1D risk tions vary among ethnic populations [95]. The immune regulator [76,77]. Many of these genes are relevant to immune function, and PTPN22 (gene located at 1p13) was shown to be associated with they lie both within and outside the HLA region. The associations of T1D in 2004 [96,97]. This association was replicated in many T1D these genes are of far lower magnitude than those of HLA genes; studies as well as in studies of other autoimmune diseases [98]. however, their contribution to T1D autoimmunity is reproducible. Most of these genes are associated with other autoimmune disor- 4.3. KIR association with T1D ders as well. HLA context must be considered in association ana- lyses for these genes, especially for those located in the HLA region, An emerging literature is focused on finding association of where LD with T1D associated DR-DQ haplotypes can be mistaken Killer-cell Immunoglobulin-like receptor (KIR) genes with T1D. for true disease association. KIRs are a family of cell-surface receptors that are found on natural killer (NK) cells and regulate their function [99e101]. HLA class I 4.1. Genes in the HLA region molecules are ligands for KIR, and HLA class I is associated with T1D; thus, KIR genes might be expected to exhibit T1D association. The arrangement of HLA class I and class II genes on chromo- Most KIR T1D association studies to date are small case-control some 6 is represented on Fig. 8. Class I genes are telomeric of class II studies that focus on individual KIR genes and test only presence genes, and between them lies a region commonly referred to as or absence of KIR loci, and results are largely inconsistent class III, although it contains no genes encoding classical HLA. The [102e112]. A large, recent study reported a method for imputation class III region is very gene dense and encodes several genes of of KIR copy number for one KIR locus (3DL1/3DS1) using Genome- immunological relevance including TNFA, MIC-A, and genes Wide Association Study (GWAS) SNP data but saw no significant encoding complement proteins. T1D associations for these genes type 1 diabetes association for that locus [113]. have been reported; however their location in the HLA region The genes that encode KIR are highly polymorphic [13,114].In precludes simple association analysis. As for HLA-DPB1 or class I contrast to HLA, where diversity is almost entirely due to extensive genes, LD with DR- and DQ-encoding genes must be accounted for allelic polymorphism in the genes, much of the known KIR genetic when analyzing data. Polymorphisms in the promoter of the TNFA diversity comes from variations in gene content, i.e., the presence or gene at positions 238 (rs361525) and 308 (rs1800629) have absence of a gene, in addition to allelic polymorphism within each been heavily studied for T1D association with conflicting results gene [115]. The KIR region lies on chromosome 19q13 in the [78e83]. Most observed genetic association for TNFA with T1D is Lymphocyte Receptor Complex (LRC) and is comprised of 14 genes;

Fig. 8. Schematic representation of the HLA region on chromosome 6p21. Selected loci in the HLA region are represented. Classical HLA loci are shown in dark blue. Loci with reported T1D associations are shown in tan. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) J.A. Noble / Journal of Autoimmunity 64 (2015) 101e112 109

Fig. 9. Schematic representation of KIR A and B haplotypes. Eight centromeric and six telomeric KIR genes are represented. General structures for A and B haplotypes are shown. reported numbers of KIR loci vary due to uncertainty about whether and convenience of procuring DNA samples and the need for only a some pairs of named loci are separate genes or multiple alleles of a single test. HLA genes are the obvious place to start for T1D. single gene. With the exception of 4 “framework” genes, other KIR T1D prediction is becoming feasible for high-risk Europeans, genes may or may not be present on a chromosome. Haplotypes of with reports demonstrating good clinical predictive value for the KIR genes fall into two categories: A, containing mostly inhibitory highest-risk HLA alleles, in combination with selected SNPs genes; and B, containing one or more activating KIR genes, sche- [119e121]. In fact, only 4 SNPs, 2 tagging HLA (DR3 and DR4), and matically represented in Fig. 9. KIR haplotypes include centromeric one each in the insulin and CTLA4 genes, can give a clinically rele- and telomeric gene groups, exhibiting strong linkage disequilib- vant predictive value (receiver operating characteristic AUC ¼ 0.72) rium within, but not between, groups [116]. KIR gene content varies to predict the highest risk T1D patients for potential intervention among populations [117]. The extent of KIR alleleic diversity is just [119]. A model incorporating up to 40 SNPs was able to achieve beginning to be understood [13]. even greater predictive value (AUC ¼ 0.87) [120]. However, the Not all KIR ligands have been elucidated; however, HLA class I highest-risk HLA genotype in Europeans, HLA-DR3/4, only accounts molecules are among known ligands for KIR [100]. Based on the for a maximum of about 40% of patients. Predicting risk from other identity of the amino acid residue at position 80, HLA-C alleles can HLA alleles and haplotypes is essential to create predictive models be categorized into 2 groups, C1 (80N) and C2 (80K). Similarly, HLA- that include all. We cannot predict genetic T1D risk for non- B alleles can be classified as Bw4 (80I or 80T) or Bw6 (80N). Effects European populations without performing HLA disease associa- of individual KIR loci on disease are, of course, not only dependent tion studies in those populations, again emphasizing the value of on whether or not that particular KIR locus is present in an indi- genetic studies in non-European populations. vidual but also on whether or not an allele encoding the HLA ligand for that KIR is present. For example, KIR2DL2 binds HLA-C alleles in 6. Conclusions the C1 group. Some studies have reported increased T1D risk for KIR2DL2 [103,105,108]. However, in individuals lacking a C1 allele T1D is a devastating disease that, in the best case, with state of (i.e., both alleles have HLA-C alleles in the C2 group rather than the the art care, is a chronic condition requiring intense management C1 group), any risk from KIR2DL2 is necessarily absent. The need to and, in the worst case, with limited resources, leads to early death. stratify based on disease, KIR, and HLA genes creates a requirement Despite decades of study and much recent progress, an effective for large population sizes to address the association of KIR with intervention or cure remains elusive. Modulation of the immune T1D. As KIR genotyping methodology evolves, more and larger response is necessary to slow or stop beta cell destruction and cohorts will be genotyped, eventually allowing sufficient sample represents a key component of any prevention or intervention size for robust T1D association analysis. strategy. A thorough understanding of the immunogenetics of T1D will allow the informed selection of subjects, at early stages of 5. Genetic prediction of T1D disease with most or all beta cell function preserved, for enrollment in clinical trials. Thorough immunogenetic characterization will The overarching goal for the study of type 1 diabetes is to find a also enhance the ability to predict response to a given therapeutic, cure, or better yet, to prevent the disease altogether. However, thereby allowing selection of appropriate interventions for at-risk prevention requires prediction. Effective identification and use of individuals of any racial or ethnic background. Immunogenetic disease-specific prevention or intervention strategies before overt characterization represents a key factor on the path to T1D pre- disease presentation depends on the ability to identify future pa- vention for all. tients, unless the intervention is benign and inexpensive enough for use in the entire population. For T1D, autoimmunity starts Acknowledgments months to years before overt disease is observed; identification of future patients is not straightforward. To develop therapies for T1D The author thanks Brenda Stallworth for assistance with figures prevention, demonstration of a significant effect on T1D develop- and tables and Julie Lane for critical reading of the manuscript. ment in a well-powered study period is necessary to show efficacy of the novel therapeutic being tested [118]. Given the relatively low References incidence of T1D, trials aimed at preventing disease require 1) screening a very large population to find at-risk individuals and 2) [1] M.J. Redondo, J. Jeffrey, P.R. Fain, G.S. Eisenbarth, T. Orban, Concordance for having a sufficient number of “progressors” in the control arm of islet autoimmunity among monozygotic twins, N. Engl. J. Med. 359 (2008) 2849e2850. the trial who will get disease in a reasonable time frame to enable [2] D.P. Singal, M.A. Blajchman, Histocompatibility (HL-A) antigens, lymphocy- sufficient power to show the benefit of a tested intervention. To totoxic antibodies and tissue antibodies in patients with diabetes mellitus, reduce the cost, scope, and length of such clinical studies, opti- Diabetes 22 (1973) 429e432. [3] J. Nerup, P. Platz, O.O. Andersen, M. Christy, J. Lyngsoe, J.E. Poulsen, et al., HL- mized initial screening is essential. Genetic testing represents a A antigens and diabetes mellitus, Lancet 2 (1974) 864e866. cost-effective choice for an initial population screen due to the ease [4] A.G. Cudworth, J.C. Woodrow, Letter: HL-A antigens and diabetes mellitus, 110 J.A. Noble / Journal of Autoimmunity 64 (2015) 101e112

Lancet 2 (1974) 1153. Contribution of DRB1*04 variants to predisposition to or protection from [5] L. Schlosstein, P.I. Terasaki, R. Bluestone, C.M. Pearson, High association of an insulin dependent diabetes mellitus is independent of dq, J. Autoimmun. 9 HL-A antigen, W27, with ankylosing spondylitis, N. Engl. J. Med. 288 (1973) (1996) 411e414. 704e706. [34] H. Erlich, A.M. Valdes, J. Noble, J.A. Carlson, M. Varney, P. Concannon, et al., [6] H. Isomaki, M. Nissila, K. Koota, J. Martio, A. Tiilikainen, Letter: HL-A 27 and HLA DR-DQ haplotypes and genotypes and type 1 diabetes risk: analysis of rheumatoid arthritis, Lancet 2 (1974) 1212e1213. the type 1 diabetes genetics consortium families, Diabetes 57 (2008) [7] L. Degos, J. Dausset, Letter: histocompatibility determinants in multiple 1084e1092. sclerosis, Lancet 1 (1974) 307e308. [35] J.A. Noble, A.M. Valdes, Genetics of the HLA region in the prediction of type 1 [8] N. Suciu-Foca, J. Buda, P. Almojera, K. Reemtsma, Letter: HL-A antigens and diabetes, Curr. Diabetes Rep. 11 (2011) 533e542. M.I.C. responsiveness in systemic lupus erythematosus, Lancet 2 (1974) [36] J.A. Noble, J. Johnson, J.A. Lane, A.M. Valdes, HLA class II genotyping of African 726e727. American type 1 diabetic patients reveals associations unique to African [9] D.J. Mulder, Letter: HL-A antigens and coeliac disease, Lancet 2 (1974) 727. haplotypes, Diabetes 62 (2013) 3292e3299. [10] J.A. Noble, A.M. Valdes, M. Cook, W. Klitz, G. Thomson, H.A. Erlich, The role of [37] M.D. Varney, A.M. Valdes, J.A. Carlson, J.A. Noble, B.D. Tait, P. Bonella, et al., HLA class II genes in insulin-dependent diabetes mellitus: molecular analysis HLA DPA1, DPB1 alleles and haplotypes contribute to the risk associated with of 180 Caucasian, multiplex families, Am. J. Hum. Genet. 59 (1996) type 1 diabetes: analysis of the type 1 diabetes genetics consortium families, 1134e1148. Diabetes 59 (2010) 2055e2062. [11] N. Risch, Assessing the role of HLA-linked and unlinked determinants of [38] A.B. Begovich, T.L. Bugawan, B.S. Nepom, W. Klitz, G.T. Nepom, H.A. Erlich, disease, Am. J. Hum. Genet. 40 (1987) 1e14. A specific HLA-DP beta allele is associated with pauciarticular juvenile [12] P.J. Bjorkman, M.A. Saper, B. Samraoui, W.S. Bennett, J.L. Strominger, rheumatoid arthritis but not adult rheumatoid arthritis, Proc. Natl. Acad. Sci. D.C. Wiley, Structure of the human class I histocompatibility antigen, HLA- U. S. A. 86 (1989) 9489e9493. A2, Nature 329 (1987) 506e512. [39] K.D. Rosenman, M. Rossman, V. Hertzberg, M.J. Reilly, C. Rice, E. Kanterakis, [13] F.F. Gonzalez-Galarza, L.Y. Takeshita, E.J. Santos, F. Kempson, M.H. Maia, et al., HLA class II DPB1 and DRB1 polymorphisms associated with genetic A.L. da Silva, et al., Allele frequency net 2015 update: new features for HLA susceptibility to beryllium toxicity, Occup. Environ. Med. 68 (2011) epitopes, KIR and disease and HLA adverse drug reaction associations, 487e493. Nucleic Acids Res. 43 (2015) D784eD788. [40] G. Thomson, HLA DR antigens and susceptibility to insulin-dependent dia- [14] J. Robinson, J.A. Halliwell, J.D. Hayhurst, P. Flicek, P. Parham, S.G. Marsh, The betes mellitus, Am. J. Hum. Genet. 36 (1984) 1309e1317. IPD and IMGT/HLA database: allele variant databases, Nucleic Acids Res. 43 [41] M.C. Eike, M. Olsson, D.E. Undlien, K. Dahl-Jorgensen, G. Joner, (2015) D423eD431. K.S. Ronningen, et al., Genetic variants of the HLA-A, HLA-B and AIF1 loci [15] F.C. Grumet, L. Fish, J. Moossazadeh, D. Ness, B.W. Duceman, An HLA-B locus show independent associations with type 1 diabetes in Norwegian families, probe clarifies endonuclease polymorphism of major histocompatibility Genes Immun. 10 (2009) 141e150. complex class I genes, Mol. Biol. Med. 1 (1983) 501e509. [42] E.E. Baschal, T.A. Aly, S.R. Babu, M.S. Fernando, L. Yu, D. Miao, et al., HLA- [16] R.K. Saiki, T.L. Bugawan, G.T. Horn, K.B. Mullis, H.A. Erlich, Analysis of DPB1*0402 protects against type 1A diabetes autoimmunity in the highest enzymatically amplified beta-globin and HLA-DQ alpha DNA with allele- risk DR3-DQB1*0201/DR4-DQB1*0302 DAISY population, Diabetes 56 (2007) specific oligonucleotide probes, Nature 324 (1986) 163e166. 2405e2409. [17] O. Olerup, H. Zetterquist, HLA-DR typing by PCR amplification with [43] T.D. Cruz, A.M. Valdes, A. Santiago, T. Frazer de Llado, L.J. Raffel, A. Zeidler, et sequence-specific primers (PCR-SSP) in 2 hours: an alternative to serological al., DPB1 alleles are associated with type 1 diabetes susceptibility in multiple DR typing in clinical practice including donor-recipient matching in cadav- ethnic groups, Diabetes 53 (2004) 2158e2163. eric transplantation, Tissue Antigens 39 (1992) 225e235. [44] F. Cucca, F. Dudbridge, M. Loddo, A.P. Mulargia, R. Lampis, E. Angius, et al., [18] P. Santamaria, M.T. Boyce-Jacino, A.L. Lindstrom, J.J. Barbosa, A.J. Faras, The HLA-DPB1eassociated component of the IDDM1 and its relationship to S.S. Rich, HLA class II “ typing”: direct sequencing of DRB, DQB, and DQA the major loci HLA-DQB1, -DQA1, and -DRB1, Diabetes 50 (2001) genes, Hum. Immunol. 33 (1992) 69e81. 1200e1205. [19] G. Bentley, R. Higuchi, B. Hoglund, D. Goodridge, D. Sayer, E.A. Trachtenberg, [45] B. Freeman, J. Powell, D. Ball, L. Hill, I. Craig, R. Plomin, DNA by mail: an et al., High-resolution, high-throughput HLA genotyping by next-generation inexpensive and noninvasive method for collecting DNA samples from sequencing, Tissue Antigens 74 (2009) 393e403. widely dispersed populations, Behav. Genet. 27 (1997) 251e257. [20] C. Lind, D. Ferriola, K. Mackiewicz, S. Heron, M. Rogers, L. Slavich, et al., Next- [46] M. Stuchlikova, D. Kantarova, D. Michalkova, L. Barak, M. Buc, Association of generation sequencing: the solution for high-resolution, unambiguous hu- HLA-DPB1 alleles with type I diabetes mellitus in Slovak population, Bratisl. man leukocyte antigen typing, Hum. Immunol. 71 (2010) 1033e1042. Lek. Listy 107 (2006) 73e75. [21] N.J. Sheehan, The ramifications of HLA-B27, J. R. Soc. Med. 97 (2004) 10e14. [47] A.M. Valdes, J.A. Noble, E. Genin, F. Clerget-Darpoux, H.A. Erlich, G. Thomson, [22] C. Darke, S.A. Corbin, External quality assessment of patient HLA-B*57:01 Modeling of HLA class II susceptibility to type I diabetes reveals an effect testing prior to abacavir prescription, Int. J. Immunogenet. 41 (2014) associated with DPB1, Genet. Epidemiol. 21 (2001) 212e223. 277e280. [48] K. Nakanishi, T. Kobayashi, T. Murase, T. Nakatsuji, H. Inoko, K. Tsuji, et al., [23] E. Vlachopoulou, E. Lahtela, A. Wennerstrom, A.S. Havulinna, P. Salo, Association of HLA-A24 with complete beta-cell destruction in IDDM, Dia- M. Perola, et al., Evaluation of HLA-DRB1 imputation using a Finnish dataset, betes 42 (1993) 1086e1093. Tissue Antigens 83 (2014) 350e355. [49] S. Nejentsev, J.M. Howson, N.M. Walker, J. Szeszko, S.F. Field, H.E. Stevens, et [24] A.R. Hsieh, S.W. Chang, P.L. Chen, C.C. Chu, C.L. Hsiao, W.S. Yang, et al., al., Localization of type 1 diabetes susceptibility to the MHC class I genes Predicting HLA genotypes using unphased and flanking single-nucleotide HLA-B and HLA-A, Nature 450 (2007) 887e892. polymorphisms in Han Chinese population, BMC 15 (2014) 81. [50] J.A. Noble, A.M. Valdes, T.L. Bugawan, R.J. Apple, G. Thomson, H.A. Erlich, The [25] X. Jia, B. Han, S. Onengut-Gumuscu, W.M. Chen, P.J. Concannon, S.S. Rich, et HLA class I A locus affects susceptibility to type 1 diabetes, Hum. Immunol. al., Imputing amino acid polymorphisms in human leukocyte antigens, PLoS 63 (2002) 657e664. One 8 (2013) e64683. [51] J.A. Noble, A.M. Valdes, M.D. Varney, J.A. Carlson, P. Moonsamy, A.L. Fear, et [26] A.M. Levin, I. Adrianto, I. Datta, M.C. Iannuzzi, S. Trudeau, P. McKeigue, et al., al., HLA class I and genetic susceptibility to type 1 diabetes: results from the Performance of HLA allele prediction methods in African Americans for class Type 1 Diabetes Genetics Consortium, Diabetes 59 (2010) 2972e2979. II genes HLA-DRB1, -DQB1, and -DPB1, BMC Genet. 15 (2014) 72. [52] A.M. Valdes, H.A. Erlich, J.A. Noble, Human leukocyte antigen class I B and C [27] A. Madbouly, L. Gragert, J. Freeman, N. Leahy, P.A. Gourraud, J.A. Hollenbach, loci contribute to Type 1 Diabetes (T1D) susceptibility and age at T1D onset, et al., Validation of statistical imputation of allele-level multilocus phased Hum. Immunol. 66 (2005) 301e313. genotypes from ambiguous HLA assignments, Tissue Antigens 84 (2014) [53] M.L. Mikk, M. Kiviniemi, A.P. Laine, T. Harkonen, R. Veijola, O. Simell, et al., 285e292. The HLA-B*39 allele increases type 1 diabetes risk conferred by HLA- [28] S.S. Khor, W. Yang, M. Kawashima, S. Kamitsuji, X. Zheng, N. Nishida, DRB1*04:04-DQB1*03:02 and HLA-DRB1*08-DQB1*04 class II haplotypes, H. Sawai, H. Toyoda, T. Miyagawa, M. Honda, N. Kamatani, K. Tokunaga, Hum. Immunol. 75 (2014) 65e70. High-accuracy imputation for HLA class I and II genes based on high- [54] T.A. Aly, E. Eller, A. Ide, K. Gowan, S.R. Babu, H.A. Erlich, et al., Multi-SNP resolution SNP data of population-specific references, Pharmacogenomics J. analysis of MHC region: remarkable conservation of HLA-A1-B8-DR3 (2015 Feb 24), http://dx.doi.org/10.1038/tpj.2015.4 (Epub ahead of print). haplotype, Diabetes 55 (2006) 1265e1269. [29] R.P. Milius, S.J. Mack, J.A. Hollenbach, J. Pollack, M.L. Heuer, L. Gragert, et al., [55] J.M. Howson, J.D. Cooper, D.J. Smyth, N.M. Walker, H. Stevens, J.X. She, et al., Genotype list string: a grammar for describing HLA and KIR genotyping re- Evidence of gene-gene interaction and age-at-diagnosis effects in type 1 sults in a text string, Tissue Antigens 82 (2013) 106e112. diabetes, Diabetes 61 (2012) 3012e3017. [30] J. Barbosa, M.M. Chern, V.E. Anderson, H. Noreen, S. Johnson, N. Reinsmoen, [56] A. Okruszko, B. Szepietowska, N. Wawrusiewicz-Kurylonek, M. Gorska, et al., Linkage analysis between the major histocompatibility system and A. Kretowski, M. Szelachowska, HLA-DR, HLA-DQB1 and PTPN22 gene insulin-dependent diabetes in families with patients in two consecutive polymorphism: association with age at onset for autoimmune diabetes, Arch. generations, J. Clin. Investig. 65 (1980) 592e601. Med. Sci. e AMS 8 (2012) 874e878. [31] S.S. Rich, P. Concannon, H. Erlich, C. Julier, G. Morahan, J. Nerup, et al., The [57] C.L. Roark, K.M. Anderson, L.J. Simon, R.P. Schuyler, M.T. Aubrey, B.M. Freed, Type 1 Diabetes Genetics Consortium, Ann. N. Y. Acad. Sci. 1079 (2006) 1e8. Multiple HLA epitopes contribute to type 1 diabetes susceptibility, Diabetes [32] J.C. Mychaleckyj, J.A. Noble, P.V. Moonsamy, J.A. Carlson, M.D. Varney, J. Post, 63 (2014) 323e331. et al., HLA genotyping in the international Type 1 Diabetes Genetics Con- [58] A.M. Valdes, H.A. Erlich, J. Carlson, M. Varney, P.V. Moonsamy, J.A. Noble, Use sortium, Clin. Trials (Lond., Engl.) 7 (2010) S75eS87. of class I and class II HLA loci for predicting age at onset of type 1 diabetes in [33] E. Harfouch-Hammoud, J. Timsit, C. Boitard, J.F. Bach, S. Caillat-Zucman, multiple populations, Diabetologia 55 (2012) 2394e2401. J.A. Noble / Journal of Autoimmunity 64 (2015) 101e112 111

[59] A.M. Valdes, G. Thomson, H.A. Erlich, J.A. Noble, Association between type 1 polymorphism with Type I diabetes, Diabetologia 43 (2000) 507e514. diabetes age of onset and HLA among sibling pairs, Diabetes 48 (1999) [86] M. Gupta, L. Nikitina-Zake, M. Zarghami, M. Landin-Olsson, I. Kockum, 1658e1661. A. Lernmark, et al., Association between the transmembrane region poly- [60] V. Harjutsalo, L. Sjoberg, J. Tuomilehto, Time trends in the incidence of type 1 morphism of MHC class I chain related gene-A and type 1 diabetes mellitus diabetes in Finnish children: a cohort study, Lancet 371 (2008) 1777e1782. in Sweden, Hum. Immunol. 64 (2003) 553e561. [61] Incidence and trends of childhood Type 1 diabetes worldwide 1990e1999, [87] N. Kumar, G. Sharma, G. Kaur, N. Tandon, S. Bhatnagar, N. Mehra, Major Diabet. Med. 23 (2006) 857e866. histocompatibility complex class I chain related gene-A microsatellite [62] G.D. Ogle, J. Lesley, P. Sine, P. McMaster, Type 1 diabetes mellitus in children polymorphism shows secondary association with type 1 diabetes and celiac in Papua New Guinea, P N G Med. J. 44 (2001) 96e100. disease in North Indians, Tissue Antigens 80 (2012) 356e362. [63] S.J. Mack, P. Cano, J.A. Hollenbach, J. He, C.K. Hurley, D. Middleton, et al., [88] Y. Yang, E.K. Chung, Y.L. Wu, S.L. Savelli, H.N. Nagaraja, B. Zhou, et al., Gene Common and well-documented HLA alleles: 2012 update to the CWD cata- copy-number variation and associated polymorphisms of complement logue, Tissue Antigens 81 (2013) 194e203. component C4 in human systemic lupus erythematosus (SLE): low copy [64] X.M. Zhang, H.Y. Wang, Y.Y. Luo, L.N. Ji, HLA-DQ, DR allele polymorphism of number is a risk factor for and high copy number is a protective factor type 1 diabetes in the Chinese population: a meta-analysis, Chin. Med. J. 122 against SLE susceptibility in European Americans, Am. J. Hum. Genet. 80 (2009) 980e986. (2007) 1037e1054. [65] D.E. Undlien, T. Friede, H.G. Rammensee, G. Joner, K. Dahl-Jorgensen, [89] N.J. Caplen, A. Patel, A. Millward, R.D. Campbell, S. Ratanachaiyavong, O. Sovik, et al., HLA-encoded genetic predisposition in IDDM: DR4 subtypes F.S. Wong, et al., Complement C4 and heat shock protein 70 (HSP70) geno- may be associated with different degrees of protection, Diabetes 46 (1997) types and type I diabetes mellitus, Immunogenetics 32 (1990) 427e430. 143e149. [90] A. Marcelli-Barge, J.C. Poirier, R. Chantome, I. Deschamps, J. Hors, [66] H. Abid Kamoun, S. Hmida, H. Kaabi, A. Abid, H. Slimane Houissa, M. Maamar, J. Colombani, Marked shortage of C4B DNA polymorphism among insulin- et al., HLA polymorphism in type 1 diabetes Tunisians, Ann. Genet. 45 (2002) dependent diabetic patients, Res. Immunol. 141 (1990) 117e128. 45e50. [91] S. Rich, G. O'Neill, A.P. Dalmasso, C. Nerl, J. Barbosa, Complement and HLA. [67] L. Gragert, A. Madbouly, J. Freeman, M. Maiers, Six-locus high resolution HLA Further definition of high-risk haplotypes in insulin-dependent diabetes, haplotype frequencies derived from mixed-resolution DNA typing for the Diabetes 34 (1985) 504e509. entire US donor registry, Hum. Immunol. 74 (2013) 1313e1320. [92] F. Jenhani, R. Bardi, Y. Gorgi, K. Ayed, M. Jeddi, C4 polymorphism in multiplex [68] C.H. Mijovic, A.H. Barnett, J.A. Todd, Genetics of diabetes. Trans-racial gene families with insulin dependent diabetes in the Tunisian population: stan- mapping studies, Baillieres Clin. Endocrinol. Metab. 5 (1991) 321e340. dard C4 typing methods and RFLP analysis, J. Autoimmun. 5 (1992) 149e160. [69] J.A. Noble, J. Johnson, J.A. Lane, A.M. Valdes, Race-specific type 1 diabetes risk [93] K. Lhotta, M. Auinger, F. Kronenberg, K. Irsigler, P. Konig, Polymorphism of of HLA-DR7 haplotypes, Tissue Antigens 78 (2011) 348e351. complement C4 and susceptibility to IDDM and microvascular complications, [70] J.M. Howson, M.S. Roy, L. Zeitels, H. Stevens, J.A. Todd, HLA class II gene Diabetes Care 19 (1996) 53e55. associations in African American Type 1 diabetes reveal a protective HLA- [94] M.P. Marron, L.J. Raffel, H.J. Garchon, C.O. Jacob, M. Serrano-Rios, DRB1*03 haplotype, Diabet. Med. 30 (6) (2013) 710e716. M.T. Martinez Larrad, et al., Insulin-dependent diabetes mellitus (IDDM) is [71] M. Rewers, T.L. Bugawan, J.M. Norris, A. Blair, B. Beaty, M. Hoffman, et al., associated with CTLA4 polymorphisms in multiple ethnic groups, Hum. Mol. Newborn screening for HLA markers associated with IDDM: diabetes auto- Genet. 6 (1997) 1275e1282. study in the young (DAISY), Diabetologia 39 (1996) 807e812. [95] J. Wang, L. Liu, J. Ma, F. Sun, Z. Zhao, M. Gu, Common variants on cytotoxic T [72] W.A. Hagopian, H. Erlich, A. Lernmark, M. Rewers, A.G. Ziegler, O. Simell, et lymphocyte antigen-4 polymorphisms contributes to type 1 diabetes sus- al., The Environmental Determinants of Diabetes in the Young (TEDDY): ceptibility: evidence based on 58 studies, PLoS One 9 (2014) e85982. genetic criteria and international diabetes risk screening of 421 000 infants, [96] N. Bottini, L. Musumeci, A. Alonso, S. Rahmouni, K. Nika, M. Rostamkhani, et Pediatr. Diabetes 12 (2011) 733e743. al., A functional variant of lymphoid tyrosine phosphatase is associated with [73] J.L. Mahon, J.M. Sosenko, L. Rafkin-Mervis, H. Krause-Steinrauf, J.M. Lachin, type I diabetes, Nat. Genet. 36 (2004) 337e338. C. Thompson, et al., The TrialNet Natural History Study of the Development [97] N. Bottini, E.J. Peterson, Tyrosine phosphatase PTPN22: multifunctional of Type 1 Diabetes: objectives, design, and initial results, Pediatr. Diabetes 10 regulator of immune signaling, development, and disease, Annu. Rev. (2009) 97e104. Immunol. 32 (2014) 83e119. [74] T.S. Group, H.K. Akerblom, J. Krischer, S.M. Virtanen, C. Berseth, D. Becker, et [98] A.B. Begovich, V.E. Carlton, L.A. Honigberg, S.J. Schrodi, A.P. Chokkalingam, al., The Trial to Reduce IDDM in the Genetically at Risk (TRIGR) study: H.C. Alexander, et al., A Missense Single-Nucleotide Polymorphism in a Gene recruitment, intervention and follow-up, Diabetologia 54 (2011) 627e633. Encoding a Protein Tyrosine Phosphatase (PTPN22) Is Associated with [75] P.L. Balducci-Silano, Z. Layrisse, E. Dominguez, R. Amaro, P. Gunczler, Rheumatoid Arthritis, Am. J. Hum. Genet. 75 (2004) 330e337. R. Lanes, et al., HLA-DQA1 and DQB1 allele and genotype contribution to [99] L.L. Lanier, B. Corliss, J.H. Phillips, Arousal and inhibition of human NK cells, IDDM susceptibility in an ethnically mixed population, Eur. J. Immunogenet. Immunol. Rev. 155 (1997) 145e154. 21 (1994) 405e414. [100] K.S. Campbell, A.K. Purdy, Structure/function of human killer cell [76] J.C. Barrett, D.G. Clayton, P. Concannon, B. Akolkar, J.D. Cooper, H.A. Erlich, et immunoglobulin-like receptors: lessons from polymorphisms, evolution, al., Genome-wide association study and meta-analysis find that over 40 loci crystal structures and , 132 (2011) 315e325. affect risk of type 1 diabetes, Nat. Genet. 41 (2009) 703e707. [101] B. Rolstad, The early days of NK cells: an example of how a phenomenon led [77] J.A. Todd, N.M. Walker, J.D. Cooper, D.J. Smyth, K. Downes, V. Plagnol, et al., to detection of a novel immune receptor system e lessons from a rat model, Robust associations of four new chromosome regions from genome-wide Front. Immunol. 5 (2014) 283. analyses of type 1 diabetes, Nat. Genet. 39 (2007) 857e864. [102] M. Jobim, P. Chagastelles, P.H. Salim, P. Portela, T.J. Wilson, A.G. Curti, et al., [78] L. Bouqbis, O. Akhayat, H.J. Garchon, F. Calafell, H. Izaabel, TNFA-TNFB hap- Association of killer cell immunoglobulin-like receptors and human leuko- lotypes modify susceptibility to type I diabetes mellitus independently of cyte antigen-C genotypes in South Brazilian with type 1 diabetes, Hum. HLA class II in a Moroccan population, Tissue Antigens 61 (2003) 72e79. Immunol. 71 (2010) 799e803. [79] J. Ilonen, H. Merivuori, H. Reijonen, M. Knip, H.K. Akerblom, F. Pociot, et al., [103] K.L. Mehers, A.E. Long, A.R. van der Slik, R.J. Aitken, V. Nathwani, F.S. Wong, Tumour necrosis factor-beta gene RFLP alleles in Finnish IDDM haplotypes. et al., An increased frequency of NK cell receptor and HLA-C group 1 com- The Childhood Diabetes in Finland (DiMe) Study Group, Scand. J. Immunol. binations in early-onset type 1 diabetes, Diabetologia 54 (2011) 3062e3070. 36 (1992) 779e783. [104] S. Mogami, G. Hasegawa, I. Nakayama, M. Asano, H. Hosoda, M. Kadono, et [80] M. Nishimura, H. Obayashi, I. Mizuta, H. Hara, T. Adachi, M. Ohta, et al., TNF, al., Killer cell immunoglobulin-like receptor genotypes in Japanese patients TNF receptor type 1, and allograft inflammatory factor-1 gene poly- with type 1 diabetes, Tissue Antigens 70 (2007) 506e510. morphisms in Japanese patients with type 1 diabetes, Hum. Immunol. 64 [105] E. Ramos-Lopez, F. Scholten, F. Aminkeng, C. Wild, H. Kalhes, C. Seidl, et al., (2003) 302e309. Association of KIR2DL2 polymorphism rs2756923 with type 1 diabetes and [81] J.A. Noble, A.M. Valdes, J.A. Lane, A.E. Green, H.A. Erlich, Linkage disequilib- preliminary evidence for lack of inhibition through HLA-C1 ligand binding, rium with predisposing DR3 haplotypes accounts for apparent effects of Tissue Antigens 73 (2009) 599e603. tumor necrosis factor and lymphotoxin-alpha polymorphisms on type 1 [106] S. Sanjeevi, C. Sun, A. Kanungo, C.B. Sanjeevi, Killer immunoglobulin receptor diabetes susceptibility, Hum. Immunol. 67 (2006) 999e1004. genes and their HLA-C ligand are associated with Type 1 diabetes in an [82] C. Perez, F.E. Gonzalez, V. Pavez, A.V. Araya, A. Aguirre, A. Cruzat, et al., The Eastern Indian population, Diabet. Med. (2015 May 29), http://dx.doi.org/ -308 polymorphism in the promoter region of the tumor necrosis factor- 10.1111/dme.12815 (Epub ahead of print). alpha (TNF-alpha) gene and ex vivo lipopolysaccharide-induced TNF-alpha [107] I. Santin, G.P. de Nanclares, B. Calvo, A. Gaafar, L. Castano, J.R. Bilbao, Killer expression in patients with aggressive periodontitis and/or type 1 diabetes cell immunoglobulin-like receptor (KIR) genes in the Basque population: mellitus, Eur. Cytokine Netw. 15 (2004) 364e370. association study of KIR gene contents with type 1 diabetes mellitus, Hum. [83] T.A. Patente, M.B. Monteiro, S.M. Vieira, M.E. Rossi da Silva, M. Nery, Immunol. 67 (2006) 118e124. M. Queiroz, et al., Linkage disequilibrium with HLA-DRB1-DQB1 haplotypes [108] A. Shastry, S.K. Sedimbi, R. Rajalingam, L. Nikitina-Zake, I. Rumba, H. Wigzell, explains the association of TNF-308G>A variant with type 1 diabetes in a et al., Combination of KIR 2DL2 and HLA-C1 (Asn 80) confers susceptibility to Brazilian cohort, Gene 568 (2015) 50e54. type 1 diabetes in Latvians, Int. J. Immunogenet. 35 (2008) 439e446. [84] J.R. Bilbao, A. Martin-Pagola, B. Calvo, G. Perez de Nanclares, N. Gepv, [109] A. Shastry, S.K. Sedimbi, R. Rajalingam, I. Rumba, A. Kanungo, C.B. Sanjeevi, L. Castano, Contribution of MIC-A polymorphism to type 1 diabetes mellitus Different KIRs confer susceptibility and protection to adults with latent in Basques, Ann. N. Y. Acad. Sci. 958 (2002) 321e324. autoimmune diabetes in Latvian and Asian Indian populations, Ann. N. Y. [85] G. Gambelunghe, M. Ghaderi, A. Cosentino, A. Falorni, P. Brunetti, Acad. Sci. 1150 (2008) 133e138. C.B. Sanjeevi, Association of MHC Class I chain-related A (MIC-A) gene [110] A.R. van der Slik, B.Z. Alizadeh, B.P. Koeleman, B.O. Roep, M.J. Giphart, 112 J.A. Noble / Journal of Autoimmunity 64 (2015) 101e112

Modelling KIR-HLA genotype disparities in type 1 diabetes, Tissue Antigens haplotypes of the human KIR genes encoding NK cell receptors, Genome Res. 69 (Suppl. 1) (2007) 101e105. 22 (2012) 1845e1854. [111] A.R. van der Slik, B.P. Koeleman, W. Verduijn, G.J. Bruining, B.O. Roep, [117] J.A. Hollenbach, I. Nocedal, M.B. Ladner, R.M. Single, E.A. Trachtenberg, Killer M.J. Giphart, KIR in type 1 diabetes: disparate distribution of activating and cell immunoglobulin-like receptor (KIR) gene-content variation in the inhibitory natural killer cell receptors in patients versus HLA-matched HGDP-CEPH populations, immunogenetics 64 (10) (2012) 719e737, http:// control subjects, Diabetes 52 (2003) 2639e2642. dx.doi.org/10.1007/s00251-012-0629-x (Epub 2012 Jul 1). [112] D. Zhi, C. Sun, S.K. Sedimbi, F. Luo, S. Shen, C.B. Sanjeevi, Killer cell [118] H.R. Thomas, S.E. Gitelman, Altering the course of type 1 diabetes: an update immunoglobulin-like receptor along with HLA-C ligand genes are associated on prevention and new-onset clinical trials, Pediatr. Diabetes 14 (2013) with type 1 diabetes in Chinese Han population, Diabetes/Metab. Res. Rev. 311e321. 27 (2011) 872e877. [119] A.M. Valdes, M.D. Varney, H.A. Erlich, J.A. Noble, Receiver operating char- [113] N. Pontikos, D.J. Smyth, H. Schuilenburg, J.M. Howson, N.M. Walker, acteristic analysis of HLA, CTLA4, and insulin genotypes for type 1 diabetes, O.S. Burren, et al., A hybrid qPCR/SNP array approach allows cost efficient Diabetes Care 36 (2013) 2504e2507. assessment of KIR gene copy numbers in large samples, BMC Genomics 15 [120] C. Winkler, J. Krumsiek, F. Buettner, C. Angermuller, E.Z. Giannopoulou, (2014) 274. F.J. Theis, A.G. Ziegler, E. Bonifacio, Feature ranking of type 1 diabetes sus- [114] D. Middleton, F. Gonzelez, The extensive polymorphism of KIR genes, ceptibility genes improves prediction of type 1 diabetes, Diabetologia 57 (12) Immunology 129 (2010) 8e19. (2014) 2521e2529. [115] P. Parham, MHC class I molecules and KIRs in human history, health and [121] C. Winkler, J. Krumsiek, J. Lempainen, P. Achenbach, H. Grallert, survival, Nat. Rev. Immunol. 5 (2005) 201e214. E. Giannopoulou, et al., A strategy for combining minor genetic susceptibility [116] W. Jiang, C. Johnson, J. Jayaraman, N. Simecek, J. Noble, M.F. Moffatt, et al., genes to improve prediction of disease in type 1 diabetes, Genes Immun. 13 Copy number variation leads to considerable diversity for B but not A (2012) 549e555.