OBJECTIVE— The nonclassical major histocompatibility complex (MHC) class I chain-related molecules (MICs), encoded within the MHC, function in immunity. The transmembrane polymorphism in MICA (MICA-STR) has been reported to be associated with type 1 diabetes. In this study, we directly sequenced both of the highly polymorphic MIC genes (MICA and MICB) in order to establish whether they are associated with type 1 diabetes independently of the known type 1 diabetes MHC class II genes HLA-DRB1 and HLA-DQB1.

RESEARCH DESIGN AND METHODS— We developed a sequencing-based typing method and genotyped MICA and MICB in 818 families (2,944 individuals) with type 1 diabetes from the U.K. and U.S. (constructing the genotype from single nucleotide polymorphisms in exons 2–4 of MICA and 2–5 of MICB) and additionally genotyped the MICA-STR in 2,023 type 1 diabetic case subjects and 1,748 control subjects from the U.K. We analyzed the association of the MICA and MICB alleles and genotypes with type 1 diabetes using regression methods.

RESULTS— We identified known MICA and MICB alleles and discovered four new MICB alleles. Based on this large-scale and detailed genotype data, we found no evidence for association of MICA and MICB with type 1 diabetes independently of the MHC class II genes (MICA P = 0.08, MICA-STR P = 0.76, MICB P = 0.03, after conditioning on HLA-DRB1 and HLA-DQB1).

CONCLUSIONS— Common MICA and MICB genetic variations including the MICA-STR are not associated, in a primary way, with susceptibility to type 1 diabetes.

Approximately 50% of the familial clustering of type 1 diabetes is attributable to the major histocompatibility complex (MHC) region on human chromosome 6p21, and the MHC class II genes HLA-DRB1 and HLA-DQB1 account for a large proportion of this clustering (1). However, association studies in the MHC region indicate that other genes have effects in type 1 diabetes independently of the class II genes (2–4). Identification of these genes is complicated because genes in the MHC have multiple alleles and exhibit extensive linkage disequilibrium (LD), and the class II genes HLA-DQB1 and HLA-DRB1 show complicated dominance and epistatic effects. Importantly, any putative new effect must be distinguished from the effect of these class II genes (2,3).

The nonclassical MHC class I chain-related molecule (MIC) genes MICA and MICB are located 46.4 and 141.2 kb, respectively, centromeric of HLA-B and ∼2 Mb telomeric of HLA-DRB1 (5) and have been associated both with immunity and with autoimmune diseases (5). In common with most MHC HLA genes, alleles of MICA and MICB are named based on haplotypes of nonsynonymous single nucleotide polymorphisms (nsSNPs). However, the most studied polymorphism in either of these genes is a short tandem repeat (STR) in the transmembrane domain of MICA (MICA-STR). Whether these genes are associated with type 1 diabetes is uncertain, with multiple conflicting reports in the literature (3,6–14). Most of these studies failed to demonstrate convincingly that the association is independent of the MHC class II genes, and all were restricted to analyses of the MICA-STR. The MICA-STR is a GCT repeat microsatellite in MICA exon 5. (Alleles of the MICA-STR are named based on the number of repeat units.) In European populations, the most common allele of the MICA-STR is MICA-STR*A5.1 (frequency 0.50), which contains five GCT repeats and a G insertion that causes a frame shift, leading to a premature stop codon and the truncation of the cytoplasmic tail. (MICA-STR*A5 contains five GCT repeats but not the G insertion and is in frame.) Here, we genotyped the MICA-STR in a large case/ control collection. However, different alleles of the MICA-STR associate with multiple alleles of other polymorphisms in MICA (5). Therefore, to capture the allelic variation of MICA more comprehensively, we designed a sequencing-based typing (SBT) method that yields the genotype information for both MICA and MICB for exons 2–5 and used it to genotype two collections of type 1 diabetic families. Various approaches have been described to obtain a complete genotype of the MICA and MICB genes (15,16). Our method is similar to other SBT methods but incorporates a novel scoring method that increases throughput.

MICA and MICB share 83% DNA sequence homology. Each gene consists of six exons: exon 1, which codes for a signal peptide, exons 2–4 for three extracellular domains (α1, α2, and α3), exon 5 for a transmembrane domain, and exon 6 for a carboxy-terminal cytoplasmic tail (5).

Genotyping of MICA-STR.

Genotyping of MICA-STR was done using PCR (Note 1 of the online appendix [available at http://dx.doi.org/10.2337/db07-1402]) followed by capillary electrophoresis of the amplified product (17). We used primers CCTTTTTTTCAGGGAAAGTGC (FAM labeled) and CCTTACCATCTCCAGAAACTG.

SBT of MICA and MICB.

Our aim was to maximize throughput in order to enable analysis of large sample sets; we therefore used a system that gave the most genotype information while minimizing the amount of sequencing and the cost. For this reason, we limited ourselves to the information that could be captured with a single pair of external primers. In both genes, exon 1 is ∼8.5 kb from exon 2, exons 2–5 are in close proximity to each other in an approximately 2-kb segment, and exon 6 is ∼2.25 kb centromeric of exon 5. Our MICA primers amplified a 1.9-kb fragment containing exons 2, 3, and 4; the MICA-STR is at the beginning of the exon 5 of MICA, and therefore it was not possible to sequence through exon 5. Our MICB primers amplified a 2.1-kb fragment containing exons 2, 3, 4, and 5. PCR (Note 1 of the online appendix) was done using MICA primers CCCCCTTCTTCTGTTCATCA (forward) and TGACTCTGAAGCACCAGCAC (reverse) and MICB primers GGACAGCAGACCTGTGTGTTA (forward) and AAAGGAGCTTTCCCATCTCC (reverse); MICA sequencing primers: TCCTGCCCAGGAAGGTT and CCTGCTGAGTTCCACTGAC for exon 2, AGGAATGGGGGTCAGTGGAA and GAGGGTTTCCCTGGACACAT for exon 3, and CTGTTCCTCTCCCCTCCTTA and CCATCCCTGCTGTCCCTAC for exon 4; MICB sequencing primers: GGACAGCAGACCTGTGTGTTA and GCCTCCCTGACCCTATTCC for exon 2, GAGTAATGGGAGGCCTTCT and TGCATCCATAGCACAGGG for exon 3, and CAGGAGTCCACCCTTGACAT, CGTTGACTCTGAAGCACCAG, and AAAGGAGCTTTCCCATCTCC for exons 4 and 5. For sequencing conditions, see online appendix Note 1.

We aligned and scored sequence reads using the Gap4 program from the Staden package (http://www.mrc-lmb.cam.ac.uk/pubseq/). To increase throughput we included a consensus trace in the alignment. Sequence data are not phased, and therefore a method was needed to determine which pair of alleles was present. While there are many nsSNPs in the MIC genes, they give rise to relatively few alleles per gene. The common alleles are listed on the IMGT/HLA Sequence Database (http://www.anthonynolan.org.uk/HIG/data.html) (18). Since the possible genotypes were limited and known, we used in-house software similar to the Helmberg Score software (19). Positions that varied were extracted from the static alignments of the IMGT/HLA Sequence Database. The alleles were reduced to just those SNPs that are present in the regions we sequenced. A file was then constructed of all possible pairs of MICA or MICB alleles. Our sequence data were extracted in the same format, and the two files were compared. Ambiguity created by missing SNP values was clarified by examining familial relationship and the MICA-STR data.

We validated this method in a reference panel of 40 cell lines received from the International Histocompatibility Working Group (IHWG) International Cell and Gene Bank (http://www.ihwg.org/cellbank/dna/refpan.html) and eight artificial heterozygotes made by pooling DNA from different homozygous cell lines. We obtained 100% concordance with the published MICA genotypes.

Sample populations.

Using our SBT method, we genotyped both MICA and MICB in 818 type 1 diabetes–affected families. These families were from two collections: the Diabetes UK Type 1 Diabetes Warren collection (478 four member families comprised of an affected sib-pair and both parents) (20) and the USA Human Biological Data Interchange (HBDI) collection (340 four- and five-member families including more than one affected sibling and both parents) (21). In addition, we genotyped the MICA-STR in 2,023 U.K. type 1 diabetic case subjects and 1,748 geographically matched control subjects (22). All collections comprised subjects of white European origin. HLA-DRB1 and HLA-DQB1 had been genotyped previously in these subjects using Dynal RELI SSO assays (Invitrogen, Paisley, U.K.) (2).

Statistical analysis.

Statistical analyses were carried out in the statistical package STATA v9 (http://www.stata.com), and the recursive partitioning (rpart) library in R was used for the recursive partitioning (23,24).

Each MIC locus was first tested for association with type 1 diabetes without conditioning on the MHC class II loci. Sets of case and matched pseudo-control subjects were generated from the affected sib-pair families and analyzed using conditional logistic regression (25) with Huber-White sandwich estimators to correct for the nonindependence of siblings. In the case-control collection, case and control subjects were matched to 12 broad geographical regions across the U.K. (Northern, East and West Ridings, North Midland, Eastern, Southeastern, Southern, Southwestern, Wales, Midlands, Scotland, London, and Northwestern) (22) and analyzed using logistic regression (2). The alleles of the MIC loci were modeled assuming a multiplicative mode of inheritance.

To establish whether any observed associations of MICA and MICB were independent of, and not due to LD with, the class II genes HLA-DRB1 and HLA-DQB1, the effects of these highly associated loci must be taken into account in the analysis of MICA and MICB. The overall number of class II alleles, which includes many rare alleles, and the well-established nonmultiplicative effects of these loci complicate their modeling. We found that rpart was an effective method to model the genotypes of HLA-DRB1 and HLA-DQB1 (2). This is a risk-based grouping method that categorizes individuals as affected or unaffected based on their MHC class II genotypes (2). The resulting groups of MHC class II genotypes define strata within which additional, non–class II associations can be tested, using either conditional logistic regression (families) or logistic regression (case/control) (2). The rpart method was performed both in the case/controls and in the families (cases/pseudo-controls). The additional MHC class II independent effect of each MIC locus was tested by adding the alleles of the test locus to the logistic model and testing their independent effects by a Wald test (families) or a likelihood ratio test (case/control). Note that all the alleles of MICA or of MICB were included in the logistic model, which allows testing of the conditional association of the locus and calculation of the conditional relative risks (RRs)/odds ratios (ORs) of the individual alleles simultaneously, relative to a common reference allele. For the most common allele of the MICA-STR (MICA-STR*A5.1), which has a minor allele frequency of 0.5, our power to detect an effect size of OR 2 was >98% at α = 10−5.

Alleles and haplotypes.

In MICA, 40 SNPs plus the MICA-STR yielded 55 distinct alleles in IMGT/HLA release 2.8 (Jan 2005) (18). We limited ourselves to SNPs in exons 2–4 so that MICA*009 and MICA*049 were indistinguishable, as were MICA*027 and MICA*048. Genotypes of the three loci were in Hardy-Weinberg equilibrium in unaffected parents and control subjects (P > 0.05). Frequencies for the MICA and MICB alleles were consistent in the U.K. and U.S. families and similar to those reported in other European populations (9,10,15,16). In online appendix Table 1, we list the combinations of MICA gene alleles and MICA-STR alleles observed in this dataset.

Novel alleles.

In MICB we found four novel alleles that were due to nsSNPs in common alleles: at position 800 in MICB*004, 813 in MICB*008, 508 in MICB*010, and 641 in MICB*014 (Table 1). These SNPs were, however, extremely rare (each was found in only one family). We also identified a synonymous SNP (in 64 individuals) in MICB (G > T at position 762), which is not present on the IMGT/HLA alignment.

Association with type 1 diabetes.

We analyzed the MICA and MICB alleles for association with type 1 diabetes in the family collections and the MICA-STR in the family and case control collections. For each locus we calculated two statistics: the P value of association for each locus overall and the RR/OR for each individual allele. When the MIC loci are analyzed without conditioning on HLA-DRB1 and HLA-DQB1, several alleles of both genes showed a strong association with type 1 diabetes (MICB P = 4.21 × 10−19, MICA P = 3.74 × 10−15, MICA-STR P = 5.44 × 10−9). However, if the effect of these genes is conditioned on HLA-DRB1 and HLA-DQB1, the association with type 1 diabetes disappears (MICB P = 0.03, MICA P = 0.08, MICA-STR P = 0.76). Note that P = 0.03 cannot be considered as evidence of association because it is produced as a result of using a Wald test that is artificially biased toward the alternative hypothesis with rare data (rare alleles, below 0.01, were grouped together, the resulting group had frequency 0.014). Use of a likelihood ratio test without robust variance estimates produces a less significant P value (P = 0.1).

In Table 2 we present the RR for the alleles of MICA and MICB in the U.K. and U.S. families using the most common alleles (MICA*008 and MICB*005) as reference, as they give the tightest 95% CIs. Since none of the other alleles were significantly different to the reference, the reference alleles themselves are also not associated. MICA-STR was also not associated with type 1 diabetes after conditioning on HLA-DRB1 and HLA-DQB1 in the case/control collection (P = 0.65 for the individual alleles and P = 0.91 for the genotypes) or in the families (P = 0.76) (Table 3). We also tested the association of MICA-STR*A5.1 genotypes against all the other genotypes grouped together. They were not associated with type 1 diabetes risk after conditioning (Table 4). We found that all MICA and MICB alleles were in strong LD (D′ >0.9) with one or more HLA-DRB1 or HLA-DQB1 alleles (online appendix Tables 2 and 3).

Our SBT method allows for large-scale and detailed genotyping of the full MICA and MICB alleles. SBT also allows discovery of novel alleles and, indeed, we found four new rare variants for MICB. These results illustrate the high level of allelic variability found in the MICA and MICB genes, implying that this variability has yet to be fully characterized.

Previous and ongoing studies of type 1 diabetes associations, including the Type 1 Diabetes Genetics Consortium (www.t1dgc.org) and the IHWG (http://www.ncbi.nlm.nih.gov/projects/mhc/ihwg.cgi?cmd=DS&ID=11), have either omitted MICA and MICB or focused on the MICA-STR because of the costs and complications associated with obtaining the full genotypes of MICA and MICB. Many published reports have identified one or more of the alleles or genotypes of the MICA-STR as being associated with type 1 diabetes. However, some of these studies did not condition adequately on the MHC class II genes (12–14), while others used subgroup analysis to analyze the MICA-STR in samples chosen to have particular class II genotypes (7–9,11). This subgrouping causes a multiple testing problem and reduces the sample set and hence the power to detect MHC class II–independent effects. We conclude that, in the European populations analyzed here, the common MICA and MICB alleles are unlikely to affect type 1 diabetes susceptibility, nor are they markers for the strong independent associations of the common alleles of HLA-B (HLA-B*39) and HLA-A (2).

Published ahead of print at http://diabetes.diabetesjournals.org on 10 March 2008. DOI: 10.2337/db07-1402.

Additional information for this article can be found in an online appendix at http://dx.doi.org/10.2337/db07-1402.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

This work was funded by the Wellcome Trust (WT), the Juvenile Diabetes Research Foundation (JDRF), the Medical Research Council, and the National Institute for Health Research Cambridge Biomedical Resource Center.

We acknowledge use of DNA from the British 1958 Birth Cohort collection (D. Strachan, P. Barton, S. Ring, W. McArdle, and M. Pembrey), funded by Medical Research Council Grant G0000934 and Wellcome Trust Grant 068545/Z/02, and DNA from the Human Biological Data Interchange and Diabetes UK for the U.S. and U.K. multiplex families, respectively.

We thank B. Widmer, H. Stevens, and the JDRF/WT Diabetes and Inflammation Laboratory DNA team for collecting and preparing the type 1 diabetes case samples and D. Smyth and R. Bailey for help with genotyping.

S.N. is a Diabetes Research and Wellness Foundation Non-Clinical Fellow.

1.
Risch N: Assessing the role of HLA-linked and unlinked determinants of disease.
Am J Hum Genet
 
40
:
1
–14,
1987
2.
Nejentsev S, Howson JM, Walker NM, Szeszko J, Field SF, Stevens HE, Reynolds P, Hardy M, King E, Masters J, Hulme J, Maier LM, Smyth D, Bailey R, Cooper JD, Ribas G, Campbell RD, The Wellcome Trust Case Control Consortium, Clayton DG, Todd JA: Localization of type 1 diabetes susceptibility to the MHC class I genes HLA-B and HLA-A.
Nature
 
450
:
887
–892,
2007
3.
Nejentsev S, Gombos Z, Laine AP, Veijola R, Knip M, Simell O, Vaarala O, Akerblom HK, Ilonen J: Non-class II HLA gene associated with type 1 diabetes maps to the 240-kb region near HLA-B.
Diabetes
 
49
:
2217
–2221,
2000
4.
Valdes AM, Wapelhorst B, Concannon P, Erlich HA, Thomson G, Noble JA: Extended DR3–D6S273-HLA-B haplotypes are associated with increased susceptibility to type 1 diabetes in US Caucasians.
Tissue Antigens
 
65
:
115
–119,
2005
5.
Collins RW: Human MHC class I chain related (MIC) genes: their biological function and relevance to disease and transplantation.
Eur J Immunogenet
 
31
:
105
–114,
2004
6.
Alizadeh BZ, Eerligh P, van der Slik AR, Shastry A, Zhernakova A, Valdigem G, Bruining JG, Sanjeevi CB, Wijmenga C, Roep BO, Koeleman BP: MICA marks additional risk factors for type 1 diabetes on extended HLA haplotypes: an association and meta-analysis.
Mol Immunol
 
44
:
2806
–2812,
2007
7.
Ide A, Babu SR, Robles DT, Wang T, Erlich HA, Bugawan TL, Rewers M, Fain PR, Eisenbarth GS: Homozygosity for premature stop codon of the MHC class I chain-related gene A (MIC-A) is associated with early activation of islet autoimmunity of DR3/4-DQ2/8 high risk DAISY relatives.
J Clin Immunol
 
25
:
303
–308,
2005
8.
Van Autreve JE, Koeleman BP, Quartier E, Aminkeng F, Weets I, Gorus FK, Van der Auwera BJ: MICA is associated with type 1 diabetes in the Belgian population, independent of HLA-DQ.
Hum Immunol
 
67
:
94
–101,
2006
9.
Nikitina-Zake L, Ghaderi M, Park Y, Babu S, Eisenbarth G, Sanjeevi CB: MICA gene polymorphism in HBDI multiplex families.
Ann N Y Acad Sci
 
1037
:
150
–156,
2004
10.
Gupta M, Ludvigsson J, Sanjeevi CB: Frequency of MICA in all babies in southeast Sweden (ABIS) positive for high-risk HLA-DQ associated with type 1 diabetes.
Ann N Y Acad Sci
 
1037
:
138
–144,
2004
11.
Allcock R, Cheong K, Christiansen F, Witt C: Comment to: Gambelunghe G, Ghaderi,: Cosentino A et al. (2000) association of MHC class I chain-related A (MIC-A) gene polymorphism with type I diabetes: Diabetologia 
43
:
507
–514.
Diabetologia
 
44
:
514
–516,
2001
12.
Lee YJ, Huang FY, Wang CH, Lo FS, Tsan KW, Hsu CH, Huang CY, Chang SC, Chang JG: Polymorphism in the transmembrane region of the MICA gene and type 1 diabetes.
J Pediatr Endocrinol Metab
 
13
:
489
–496,
2000
13.
Shtauvere-Brameus A, Ghaderi M, Rumba I, Sanjeevi CB: Microsatellite allele 5 of MHC class I chain-related gene a increases the risk for insulin-dependent diabetes mellitus in latvians.
Ann N Y Acad Sci
 
958
:
349
–352,
2002
14.
Bilbao JR, Martin-Pagola A, Calvo B, Perez de Nanclares G, Gepv N, Castano L: Contribution of MIC-A polymorphism to type 1 diabetes mellitus in Basques.
Ann N Y Acad Sci
 
958
:
321
–324,
2002
15.
Ahmad T, Marshall SE, Mulcahy-Hawes K, Orchard T, Crawshaw J, Armuzzi A, Neville M, van Heel D, Barnardo M, Welsh KI, Jewell DP, Bunce M: High resolution MIC genotyping: design and application to the investigation of inflammatory bowel disease susceptibility.
Tissue Antigens
 
60
:
164
–179,
2002
16.
Gao X, Single RM, Karacki P, Marti D, O'Brien SJ, Carrington M: Diversity of MICA and linkage disequilibrium with HLA-B in two North American populations.
Hum Immunol
 
67
:
152
–158,
2006
17.
Mizuki N, Ota M, Kimura M, Ohno S, Ando H, Katsuyama Y, Yamazaki M, Watanabe K, Goto K, Nakamura S, Bahram S, Inoko H: Triplet repeat polymorphism in the transmembrane region of the MICA gene: a strong association of six GCT repetitions with Behcet disease.
Proc Natl Acad Sci U S A
 
94
:
1298
–1303,
1997
18.
Robinson J, Waller MJ, Parham P, Bodmer JG, Marsh SG: IMGT/HLA Database–a sequence database for the human major histocompatibility complex.
Nucleic Acid Res
 
29
:
210
–213,
2001
19.
Helmberg W, Zahn R, Keller E, Weinmair B, Lanzer G, Albert E: Virtual DNA analysis as a platform for interlaboratory data exchange of HLA DNA typing results.
Tissue Antigens
 
54
:
379
–385,
1999
20.
Bain SC, Todd JA, Barnett AH: The British Diabetic Association–Warren Repository.
Autoimmunity
 
7
:
83
–85,
1990
21.
Lernmark A, Ducat L, Eisenbarth G, Ott J, Permutt MA, Rubenstein P, Spielman R: Family cell lines available for research.
Am J Hum Genet
 
47
:
1028
–1030,
1990
22.
Clayton DG, Walker NM, Smyth DJ, Pask R, Cooper JD, Maier LM, Smink LJ, Lam AC, Ovington NR, Stevens HE, Nutland S, Howson JM, Faham M, Moorhead M, Jones HB, Falkowski M, Hardenbol P, Willis TD, Todd JA: Population structure, differential bias and genomic control in a large-scale, case-control association study.
Nat Genet
 
37
:
1243
–1246,
2005
23.
R Development Core Team:
A Language and Envoronment for Statistical Computing.
R Foundation for Statistical Computing, Vienna,
2006
24.
Therneau TM, Atkinson EJ:
An Introduction to Recursive Partitioning Using the rpart Routine.
Minnesota, Mayo Clinic, Section of Statistics,
1997
25.
Cordell HJ, Clayton DG: A unified stepwise regression procedure for evaluating the relative effects of polymorphisms within a gene using case/control or family data: application to HLA in type 1 diabetes.
Am J Hum Genet
 
70
:
124
–141,
2002
DIABETES
2008

Supplementary data

Supplementary Tables 1-3- doc file