Reproducible Association with Type 1 Diabetes in the Extended Class I Region of the Major Histocompatibility Complex

ORIGINAL ARTICLE Reproducible association with type 1 diabetes in the extended class I region of the major histocompatibility complex

MK Viken1,2,10, A Blomhoff3,10, M Olsson4,5, HE Akselsen6, F Pociot7, J Nerup7, I Kockum8, A Cambon-Thomsen9, E Thorsby1,2, DE Undlien3,6 and BA Lie2 1Institute of Immunology, Faculty Division Rikshospitalet, University of Oslo, Oslo, Norway; 2Institute of Immunology, Rikshospitalet University Hospital, Oslo, Norway; 3Institute of Medical Genetics, Faculty Division Ulleva˚l University Hospital, University of Oslo, Oslo, Norway; 4Department of Mathematical Statistics, Chalmers University of Technology, Go¨teborg, Sweden; 5Department of Mathematical Statistics, Go¨teborg University, Go¨teborg, Sweden; 6Department of Medical Genetics, Ulleva˚l University Hospital, Oslo, Norway; 7Steno Diabetes Centre, Gentofte, Denmark; 8Department of Clinical Neurosciences, Neuroimmunology Unit, Karolinska Institutet, Stockholm, Sweden and 9Inserm U 558, University Paul Sabatier, Toulouse, France

The high-risk human leukocyte antigen (HLA)-DRB1, DQA1 and DQB1 alleles cannot explain the entire type 1 diabetes (T1D) association observed within the extended major histocompatibility complex. We have earlier identified an association with D6S2223, located 2.3 Mb telomeric of HLA-A, on the DRB1*03-DQA1*0501-DQB1*0201 haplotype, and this study aimed to fine-map the associated region also on the DRB1*0401-DQA1*03-DQB1*0302 haplotype, characterized by less extensive linkage disequilibrium. To exclude associations secondary to DRB1-DQA1-DQB1 haplotypes, 205 families with at least one parent homozygous for these loci, were genotyped for 137 polymorphisms. We found novel associations on the DRB1*0401- DQA1*03-DQB1*0302 haplotypic background with eight single nucleotide polymorphisms (SNPs) located within or near the PRSS16 gene. In addition, association at the butyrophilin (BTN)-gene cluster, particularly the BTN3A2 gene, was observed by multilocus analyses. We replicated the associations with SNPs in the PRSS16 region and, albeit weaker, to the BTN3A2 region, in an independent material of 725 families obtained from the Type 1 Diabetes Genetics Consortium. It is important to note that these associations were independent of the HLA-DRB1-DQA1-DQB1 genes, as well as of associations observed at HLA-A, -B and -C. Taken together, our results identify PRSS16 and BTN3A2, two genes thought to play important roles in regulating the immune response, as potentially novel susceptibility genes for T1D. Genes and Immunity (2009) 10, 323–333; doi:10.1038/gene.2009.13; published online 19 March 2009

Keywords: association; type 1 diabetes; HLA; xMHC; BTN; PRSS16

Introduction MHC-linked T1D susceptibility genes, as they observed that siblings of DR3/DR4-DQ8 T1D patients, who were Type 1 diabetes (T1D) is a multifactorial disease, caused identical by descent for both extended HLA haplotypes, by an interaction of both genetic and environmental had a considerably higher risk than the DR3/DR4-DQ8 factors. Years of research, as well as a recent large siblings not sharing both of these extended haplotypes. genome-wide association study agree that the major We have earlier performed a microsatellite screen of genetic contribution to T1D resides within the extended the xMHC and found an association with D6S2223 in the major histocompatibility complex (xMHC),1 in which extended class I region, independent of linkage disequi- the human leukocyte antigen (HLA) class II genes, and librium (LD) with the known HLA class II risk alleles.4,6 particularly the DRB1*03-DQA1*0501-DQB1*0201 and The association was observed as a reduced transmis- the DRB1*04-DQA1*03-DQB1*0302 haplotypes confer sion of allele D6S2223*3 from DRB1*03-DQA1*0501- the highest risk.2 However, accumulating evidence DQB1*0201 homozygous parents in a combined T1D shows that these class II genes cannot explain the entire dataset from Norway, Sweden, Denmark, United King- T1D association with the xMHC.3–7 Recently, Aly et al.8 dom and France (11T vs 46NT, P ¼ 0.000004).4,6 Interest- provided further evidence for the existence of additional ingly, others have suggested that if additional T1D risk factors exist on the DRB1*03-DQA1*0501-DQB1*0201 haplotype, these are situated centromeric of HLA-DRB1 Correspondence: Dr BA Lie, Institute of Immunology, Rikshospita- and telomeric of HLA-A.9 let University Hospital, Sognsvannsveien 20, Oslo 0027, Norway. The xMHC is characterized by strong and extended E-mail: [email protected] 10,11 10These authors contributed equally to work. LD, making it difficult to pinpoint the causal 12 Received 23 December 2008; revised and accepted 13 February 2009; variant(s). The LD pattern is also intricate and vary published online 19 March 2009 with regard to the underlying ancestral HLA haplo- Reproducible association with T1D in the xMHC MK Viken et al 324 types.13,14 Thus, attention must be paid to the HLA associated D6S2223 marker and 43 microsatellites span- haplotype distributions in the population being studied. ning the xMHC used in our earlier LD screen.15 A 0 Of relevance for T1D, the LD on the DRB1*03- significant D 40.3 (Pnco0.05) was used as a cut-off DQA1*0501-DQB1*0201 haplotype is generally stretching criterion, as this has been estimated to be the minimal over longer distances than on the DRB1*0401-DQA1*03- usable amount of LD in association studies,22 resulting in DQB1*0302 haplotype, making it harder to capture an a region of B0.8 Mb (between D6S306 and D6S1558). association on the latter haplotype.15 This could explain Single-point associations were obtained by the homo- why the association with D6S2223 was only observed on zygous parent transmission disequilibrium test sepa- the DRB1*03-DQA1*0501-DQB1*0201 haplotype. Further- rately from DRB1*03-DQA1*0501-DQB1*0201 and more, for fine-mapping purposes the DRB1*0401- DRB1*0401-DQA1*03-DQB1*0302 homozygous parents DQA1*03-DQB1*0302 haplotype, with less extensive for the 50 SNPs successfully genotyped in the initial LD, might prove to be especially useful. Finally, these dataset of 205 families (Figure 1). The associated SNPs differential LD patterns make it imperative to map the from this first phase screen were located telomeric of T1D association separately on these HLA haplotypes. D6S2223, except rs9295746. As the associated SNPs on The extended class I region harbors several T1D the DRB1*0401-DQA1*03-DQB1*0302 haplotype were candidate genes involved in immunological processes,16 localized at the very telomeric end of the initially defined including the PRSS16 gene, a cluster of seven butyr- 0.8 Mb region (Figure 1), the SNP screen was extended ophilin (BTN) genes that are members of the immuno- telomerically to cover B1.9 Mb (second phase screen). globulin superfamily,17 and zinc-finger protein genes When analyzing the 134 SNPs from the first and encoding transcriptional regulators involved in cell second phase screen together in the initial dataset, only growth and differentiation.18 The PRSS16 gene is eight showed association (Po0.05) on the DRB1*0401- predominantly expressed in the cortical epithelial cells DQA1*03-DQB1*0302 haplotypic background (Table 1 of the thymus, and has been suggested to play a role in and Figure 1, triangles). Seven of these associated SNPs the positive selection of T cells or in T-cell regulation.19 covered an 85-kb region, in which PRSS16 is the only We have earlier screened this candidate gene for novel known protein-coding gene, and all showed the same polymorphisms,20 but none of the polymorphisms within transmission distortion (13%T (95%CI, 10–35%), P ¼ 0.02) the PRSS16 gene could explain the association observed and were in perfect LD on this haplotypic background. with D6S2223 on the DRB1*03-DQA1*0501-DQB1*0201 The last associated SNP (rs12201890; 0T vs 5NT, haplotypic background.21 P ¼ 0.008) was situated B482 kb telomeric of, and in In the current study, we performed a single nucleotide strong LD with, these seven SNPs (r2 ¼ 0.91). The seven polymorphism (SNP) screen to further limit the asso- SNPs showed a D0 ¼ 0.66 and r2 ¼ 0.11 with D6S2223*3, ciated region defined by the microsatellites analyzed whereas rs12201890 showed D0 ¼ 0.51 and r2 ¼ 0.07. earlier.4,6,15 Initially, we analyzed the transmission of Three additional SNPs were selected as candidate SNPs; SNPs in 205 families with DRB1*03-DQA1*0501- that is, rs9379857 and rs1985732 were selected on the DQB1*0201 and DRB1*0401-DQA1*03-DQB1*0302 homo- basis of reported association with allelic gene expression zygous parents, and paid particular attention to the differences of the BTN3A2 gene,23 and rs996247 (B22 kb association on the DRB1*0401-DQA1*03-DQB1*0302 hap- upstream of PRSS16) was included on the basis of a lotype, as the less extensive LD on this haplotype could reported novel association with T1D.24 None of these guide us closer to the causal variant. Thereafter, candidate SNPs showed association on the DRB1*0401- identified associations were replicated in 725 families DQA1*03-DQB1*0302 haplotype (Figure 1). from the Type 1 Diabetes Genetics Consortium (T1DGC), Twelve polymorphisms from the first and second and were also checked against other associations genotyping phases (11 SNPs and a 15-bp insertion/ observed in the MHC, like the HLA class I loci. deletion) showed association with T1D on the DRB1*03- DQA1*0501-DQB1*0201 haplotype (Table 1), all of which were scattered throughout the region (Figure 1, circles). It is interesting to note that one of the three candidate Results SNPs, rs9379857, presumably influencing the expression Our available dataset has changed over time (mainly of BTN3A2 was associated on the DRB1*03-DQA1*0501- because of the lack of DNA), therefore we reanalyzed DQB1*0201 haplotype (Table 1). Overall, the strongest the earlier reported associated D6S2223*3 allele, and association was observed with rs2393963, showing the transmission distortion from DRB1*03-DQA1*0501- a transmission distortion of 25% (95%CI, 10–40%; DQB1*0201 homozygous parents was 22%, 95% con- P ¼ 0.004), which is comparable with what we observed fidence intervals (CI), 8–36% (7T vs 25NT, P ¼ 0.001) in for D6S2223*3. Variable degrees of LD were seen between the current initial dataset of 205 families from Norway, these thirteen polymorphisms, with r2 ranging from zero Sweden, Denmark and France. We therefore expected the to one (Figure 2a), and seven of these were in strong LD effect size and significance level for the SNP associations with D6S2223*3 (defined as D040.8; Figure 2b). Within to be near this magnitude or stronger. We should be the 85 kb associated region on the DRB1*0401-DQA1*03- able to detect such an association, as the association with DQB1*0302 haplotype, two polymorphisms also showed D6S2223 is relatively pronounced in the initial dataset association on the DRB1*03-DQA1*0501-DQB1*0201 used in the current study. As the association with haplotype, that is, rs3800307 (32%T (95%CI, 17–48%), D6S2223 was observed with an allele being negatively P ¼ 0.04) and rs5030663 (33%T (95%CI, 19–48%), associated, we have focused on the negatively associated P ¼ 0.04). These were the only polymorphisms in perfect SNP alleles and haplotypes throughout the results. LD among those associated on the DRB1*03-DQA1*0501- The genomic region of interest for the fine-mapping DQB1*0201 haplotype. The LD measurements between was defined by pairwise LD between the disease- D6S2223*3 and rs3800307 or rs5030663 (Figure 2), display

Genes and Immunity Reproducible association with T1D in the xMHC MK Viken et al 325 The extended MHC

Classical MHC

Extended Class I Class I Class III Class II

HFE PRSS16 HLA-G HLA-C TNF HLA-DR HLA-DP BTN-genes D6S2223 HLA-A HLA-B HLA-DQ

25600000 27600000 29600000 31600000 33600000

The 1.9 Mb region screened in the first and second phase

The 0.8 Mb region screened in the first phase

3.000 P-values (DRB1*03-DQA1*0501-DQB1*0201) P-values (DRB1*0401-DQA1*03-DQB1*0302) P-value=0.05 2.500 P-values for candidate SNPs (DRB1*03-DQA1*0501-DQB1*0201) P-values for candidate SNPs (DRB1*0401-DQA1*03-DQB1*0302)

rs12201890 2.000 rs9379857

rs5030663 rs9295746 1.500 rs3800307 -log (P)

1.000

0.500

0.000 Position 26.1 26.6 27.1 27.6 28.1 (Mb)

BTN gene cluster PRSS16 D6S2223

Figure 1 A schematic overview of the extended major histocompatibility complex (xMHC), including some key genes (upper panel), and the association results from homozygous parent transmission disequilibrium test for all 137 polymorphisms investigated in our study (lower panel). The 134 polymorphisms tested in both the first and second phase single nucleotide polymorphism (SNP) screen were analyzed separately on the DRB1*03-DQA1*0501-DQB1*0201 (circles) and DRB1*0401-DQA1*03-DQB1*0302 (triangles) haplotypes. The same analyses were performed for the three candidate SNPs, denoted by squares (DRB1*03-DQA1*0501-DQB1*0201) and diamonds (DRB1*0401-DQA1*03- DQB1*0302). Polymorphisms with o4 informative transmissions were excluded (that is, nine SNPs on the DRB1*0401-DQA1*03-DQB1*0302 haplotype). Points above the dashed line have P-valueso0.05. Some key genetic loci in the region are denoted, and the seven DRB1*0401- DQA1*03-DQB1*0302 associated SNPs (rs6931711, rs858988, rs9368492, rs9393796, rs9348756, rs9368493 and rs6938397), in perfect linkage disequilibrium, are encircled in the plot. high D0 (0.91 and 0.90, respectively), but low r2 (0.15 and DRB1*03-DQA1*0501-DQB1*0201 (159T vs 126NT) and 0.14, respectively), indicating the existence of reduced DRB1*0401-DQA1*03-DQB1*0302 (117T vs 75NT) haplo- haplotype diversity between the microsatellite and these types available in the families (Table 1). Eight of the SNPs on the DRB1*03-DQA1*0501-DQB1*0201 haploty- thirteen SNPs found associated on the DRB1*03- pic background. DQA1*0501-DQB1*0201 haplotype still showed signifi- As the number of parents homozygous for DRB1*03- cant association. On the DRB1*0401-DQA1*03- DQA1*0501-DQB1*0201 (N ¼ 111) and DRB1*0401- DQB1*0302 haplotype, none of the SNPs were signifi- DQA1*03-DQB1*0302 (N ¼ 62) was restricted, and the cant, but a trend was observed for rs12201890 (Odds number of informative transmissions was relatively ratios (OR) ¼ 0.41 (95%CI, 0.16–1.03), P ¼ 0.06) and small, polymorphisms showing association by homo- rs9368492 (OR ¼ 0.43 (95%CI, 0.17–1.05), P ¼ 0.06). zygous parent transmission disequilibrium test were also Nevertheless, the size of the transmission distortions tested by investigating the distribution of the associated and the ORs were in the same range as for the DRB1*03- SNPs on all the transmitted (T) and non-transmitted (NT) DQA1*0501-DQB1*0201 haplotype, which may indicate

Genes and Immunity Reproducible association with T1D in the xMHC MK Viken et al 326 Table 1 Associated SNP on the DRB1*03-DQA1*0501-DQB1*0201 haplotype and the DRB1*0401-DQA1*03-DQB1*0302 haplotype from both the homozygous parent TDT (HPTDT) and the haplotype method

SNP Chr pos HPTDT Haplotype method

T vs NT %T (95%CI) P-value % on T %onNT OR (95%CI) P-value haplotypes haplotypes

From DRB1*03-DQA1*0501-DQB1*0201 Distribution on T (N ¼ 159) and NT (N ¼ 126) homozygous parents (N ¼ 109) DRB1*03-DQA1*0501-DQB1*0201 haplotypes

rs7749823 26266058 8 vs 19 30 (12–47) 0.03 23.5 34.9 0.57 (0.33–0.99) 0.05 rs9379857 26475883 10 vs 25 29 (14–44) 0.01 24.2 39.5 0.49 (0.29–0.83) 0.008 rs1624064 26486660 8 vs 18 31 (13–49) 0.05 62.1 73.7 0.59 (0.34–1.01) 0.05 rs3757148 26489185 5 vs 16 24 (6–42) 0.01 80.8 92.4 0.36 (0.16–0.78) 0.01 rs9393738 26728637 9 vs 21 30 (14–46) 0.03 33.1 52.1 0.46 (0.27–0.78) 0.004 rs3800307 27293771 11 vs 23 32 (17–48) 0.04 33.2 49.7 0.51 (0.30–0.85) 0.01 rs5030663 27331044 13 vs 26 33 (19–48) 0.04 33.8 47.7 0.56 (0.34–0.92) 0.02 rs7754411 27467563 4 vs 12 25 (4–46) 0.04 81.2 90.8 0.45 (0.22–0.94) 0.03 rs13207689 27477683 12 vs 24 33 (18–49) 0.04 26.8 38.1 0.60 (0.36–1.00) 0.05 rs2393963 27583079 8 vs 24 25 (10–40) 0.004 56.1 74.3 0.45 (0.26–0.76) 0.003 rs7775817 27656071 8 vs 18 31 (13–49) 0.05 77.2 86.9 0.52 (0.27–0.99) 0.05 rs6917366 27742619 6 vs 18 25 (8–42) 0.01 73.1 89.0 0.34 (0.18–0.67) 0.002 D6S2223*3 27765822 7 vs 25 22 (8–36) 0.001 71.8 87.3 0.38 (0.20–0.72) 0.003 rs9295746 27838043 7 vs 17 29 (11–47) 0.04 34.4 47.2 0.59 (0.34–1.02) NS (0.06)

From DRB1*0401-DQA1*03-DQB1*0302 Distribution on T (N ¼ 117) and NT (N ¼ 75) homozygous parents (N ¼ 61) DRB1*0401-DQA1*03-DQB1*0302 haplotypes

rs12201890 26770605 0 vs 5 0.008 7.7 17.3 0.41 (0.16–1.03) NS (0.06) rs6931711 27253565 1 vs 7 13 (10–35) 0.02 9.4 16.8 0.52 (0.22–1.21) NS (0.1) rs858988 27287174 1 vs 7 13 (10–35) 0.02 9.5 16.8 0.52 (0.22–1.21) NS (0.1) rs9368492 27329497 1 vs 7 13 (10–35) 0.02 7.7 16.6 0.43 (0.17–1.05) NS (0.06) rs9393796 27334944 1 vs 7 13 (10–35) 0.02 9.5 16.5 0.53 (0.23–1.24) NS (0.1) rs9348756 27336238 1 vs 7 13 (10–35) 0.02 9.3 16.6 0.52 (0.22–1.21) NS (0.1) rs9368493 27336279 1 vs 7 13 (10–35) 0.02 9.4 16.5 0.53 (0.23–1.23) NS (0.1) rs6938397 27339203 1 vs 7 13 (10–35) 0.02 8.6 16.5 0.48 (0.20–1.13) NS (0.09)

Abbreviations: Chr pos, chromosome position; CI, confidence intervals; HPTDT, homozygous parent transmission disequilibrium test; NT, non-transmitted; NS, non-significant; OR, odds ratios; SNP, single nucleotide polymorphism; T, transmitted; TDT, transmission disequilibrium test.

that lack of power could be the reason for inability to rs9393795-rs9368492) showed a frequency of 6.6% on reach statistical significance. Finally, the allelic SNP the transmitted and 18.6% on the non-transmitted distributions on the DRB1*03-DQA1*0501-DQB1*0201 DRB1*0401-DQA1*03-DQB1*0302 haplotypes (OR ¼ 0.3 vs DRB1*0401-DQA1*03-DQB1*0302 haplotypes high- (95%CI, 0.13–0.81), P ¼ 0.01). The haplotype associations lighted the differences in the constitution of these two in this region coincided with the single SNP associations haplotypes (data not shown) and underlined the across the PRSS16 gene on the DRB1*0401-DQA1*03- advantage of initially mapping associations separately DQB1*0302 haplotypic background. The telomeric peak, on these haplotypes. covering parts of the BTN-gene cluster, particularly the In an effort to clarify the association picture, BTN3A2 gene, was stretching over an area of B106 kb we constructed five-SNP haplotypes separately on (between SNP numbers 18 and 25; rs9393691 and the DRB1*03-DQA1*0501-DQB1*0201 and DRB1*0401- rs1624064). The most significantly associated five-SNP DQA1*03-DQB1*0302 haplotypic background for the haplotype (SNP numbers 21–25; rs9358929-rs11758089- 134 first and second phase SNPs (Figure 3). All rs2073526-rs13218591-rs1624064) showed a frequency of significantly associated five-SNP haplotypes were 5.3% on the transmitted and 16.1% on the non-trans- located telomeric of D6S2223 on both haplotypic back- mitted DRB1*0401-DQA1*03-DQB1*0302 haplotypes grounds, but again, more confined on the DRB1*0401- (OR ¼ 0.3 (95%CI, 0.10–0.89), P ¼ 0.03). Notably, none of DQA1*03-DQB1*0302 than on the DRB1*03-DQA1*0501- these SNPs showed any association in the single-point DQB1*0201 haplotypes. SNP analyses on the DRB1*0401-DQA1*03-DQB1*0302 The associations on the DRB1*0401-DQA1*03- haplotypic background. DQB1*0302 haplotype peaked in two regions In contrast, the associated five-SNP haplotypes were (Figure 3a). The centromeric peak covered an area of scattered on the DRB1*03-DQA1*0501-DQB1*0201 back- B48 kb across the PRSS16 gene (between SNP numbers ground (Figure 3b), with several less defined peaks of 50 and 63; rs9393794 and rs9379968). The most variable sizes across almost the entire region. Associated significantly associated five-SNP haplotype fragment SNP haplotypes were also present in the two regions (SNP numbers 50–54; rs9393794-rs9468015-rs10946904- found on the DRB1*0401-DQA1*03-DQB1*0302 haplo-

Genes and Immunity Reproducible association with T1D in the xMHC MK Viken et al 327

Figure 2 Linkage disequilibrium plots of the 13 associated polymorphisms on the DRB1*03-DQA1*0501-DQB1*0201 haplotypic background given as either (a) r2 values or (b) D0-values. The multiallelic marker D6S2223 has been converted into a biallelic marker with the negatively type 1 diabetes (T1D)-associated allele 3 as one allele, and all others as the second allele. type, but covering larger areas. Exploration of the LD analyses than for those restricted to the DRB1*03- between the 134 first and second phase SNPs revealed DQA1*0501-DQB1*0201 haplotype in the initial dataset. the presence of longer range LD, especially when It is interesting to note that in the replication study of the investigating pairwise D0-values, indicating the presence T1DGC families, logistic regression analysis conditional of more extensive haplotype structures on the DRB1*03- on DRB1-DQA1-DQB1 (Table 2) revealed significant DQA1*0501-DQB1*0201 compared with the DRB1*0401- association with the two SNPs, rs9368492 (P ¼ 0.003) DQA1*03-DQB1*0302 haplotypic background (data not and rs9393796 (P ¼ 0.001), being in perfect LD in this shown). dataset and located in the region close to, or containing, Nine SNPs showing association on the DRB1*0401- the PRSS16 gene. Three other SNPs showed marginally DQA1*03-DQB1*0302 haplotype, either in the single SNP significant associations, namely rs12201890 B55 kb or in haplotype analyses (or both), were selected for telomeric of PRSS16 and close to a zinc-finger gene replication. This strategy was chosen as the DRB1*0401- (ZNF322A, P ¼ 0.03), as well as rs11758089 and DQA1*03-DQB1*0302 haplotype showed more confined rs13218591 in the BTN-gene cluster (in the BTN3A2 gene, LD patterns, even though we had less power for these P ¼ 0.05).

Genes and Immunity Reproducible association with T1D in the xMHC MK Viken et al 328 BTN gene cluster PRSS16 D6S2223

2 p-value p=0.05 1.8

1.6

1.4

1.2

1 -log (P) 0.8

0.6

0.4

0.2

0 381318 23 28 33 38 43 48 53 58 63 68 73 78 83 88 93 98 103 108 113 118 123 128 SNP number

3.5 p-value p=0.05 3

2.5

-log (P) 1.5

0.5

0 3 8 13 18 23 28 33 38 43 48 53 58 63 68 73 78 83 88 93 98 103 108 113 118 123 128 SNP number Figure 3 The most negatively transmitted five-single nucleotide polymorphism (SNP) haplotypes on the (a) DRB1*0401-DQA1*03- DQB1*0302 haplotype and (b) DRB1*03-DQA1*0501-DQB1*0201 haplotype from the first and second phase. One diamond represents a five-SNP haplotype and is placed above the middle SNP number. Diamonds above the dashed line have a P-value o0.05.

Earlier analyses by our research group of the T1DGC reported association with SNPs near HLA-G, that is, families for associations within the classical MHC using rs4122198, rs1619379 and rs2394186,25 and others have HLA data and SNPs from an extensive screen (not reported association with SNPs in the UBD/MAS1L extending into the region covered by the current study) region (particularly with rs1233478).24 When condition- have shown associations independent of DRB1-DQA1- ing on each of these SNPs (together with DRB1-DQA1- DQB1 in three regions.25 Therefore, we also added loci DQB1 and HLA-B), the association with the SNPs in the from these regions in the conditional logistic regression extended class I region remained significant and vise analyses (Table 2). The associations with the two SNPs versa, except for the PRSS16 SNP rs9393796 when (rs9368492 and rs9393796) in the PRSS16 region were still adjusting for rs2394186 (data not shown). Both these significant after conditioning for DRB1-DQA1-DQB1 SNPs seemed to reduce each other’s effect, thereby together with either HLA-A (Po0.01), HLA-B (Po0.01) indicating some sort of dependency between these or HLA-C (Po0.003). One of the BTN3A2 SNPs, associations. Overall, the LD between the PRSS16 SNPs rs13218591, lost its borderline association when either and the HLA-G or UBD SNPs in the entire T1DGC the HLA-A or -B locus was included in the main effects dataset, and also between rs2394186 and rs9393796 test, whereas the other, rs11758089, remained significant (D0 ¼ 0.15; r2 ¼ 0.01), is low (D0o0.4; r2o0.01). Further- after inclusion of either of the HLA class I loci. more, when examining pairwise haplotypes of SNPs in Furthermore, in the earlier T1DGC screen we have also the extended class I region with SNPs in the classical

Genes and Immunity Reproducible association with T1D in the xMHC MK Viken et al 329 Table 2 Association analyses in the T1DGC families by the main rs5030663) associated on the DRB1*03-DQA1*0501- effects test adjusting for different HLA loci DQB1*0201 haplotype in this region, were also in perfect LD, and showed similar transmission distortions, which Test SNP Conditional loci; DRB1-DQA1-DQB1 together with were comparable, and not superior, to the association seen with D6S2223*3 from DRB1*03-DQA1*0501- — HLA-A HLA-B HLA-C DQB1*0201 homozygous parents in the current initial dataset. Taken together, these observations could indicate rs11754384 NS NS NS NS that we have not genotyped the causal variant, and are rs9358929 NS NS NS NS still relying on LD to detect an association. rs11758089 0.05 0.004 0.03 0.05 rs2073526 NS NS NS NS The ORs for the associations, observed by the rs13218591 0.05 0.11 0.14 0.05 haplotype method, were between 0.51 and 0.56 on rs1624064 NS NS NS NS the DRB1*03-DQA1*0501-DQB1*0201 haplotype and rs12201890 0.03 0.15 0.01 0.03 between 0.43 and 0.53 on the DRB1*0401-DQA1*03- rs9368492 0.003 0.01 0.01 0.003 DQB1*0302 haplotype (albeit not reaching statistical rs9393796 0.001 0.007 0.002 0.001 significance), which are in line with other non-HLA- DRB1-DQA1-DQB1 risk loci. Besides the HLA class II DR Abbreviations: HLA, human leukocyte antigen; NS, non-significant; and DQ genes, other T1D-associated loci demonstrate SNP, single nucleotide polymorphism; T1DGC, Type 1 Diabetes medium to low disease risk, with most loci having ORs Genetics Consortium. o2.26 Hence, the susceptibility locus (or loci) we are Non-significant P-values are given for SNPs, which were significant searching for is likely to display an effect size typical for in the first test when conditioning only on DRB1-DQA1-DQB1. non-HLA-DRB1-DQA1-DQB1 susceptibility loci. It is interesting to note that the associations were restricted to the telomeric side of D6S2223 (except for class I region on various DRB1-DQA1-DQB1 back- rs9295746 on the DRB1*03-DQA1*0501-DQB1*0201 hap- grounds, the disease associations conferred by the SNP lotype). This is in accordance with the mapping efforts alleles generally map to different DRB1-DQA1-DQB1 by others on the DRB1*03-DQA1*0501-DQB1*0201 hap- haplotypes (data not shown). lotype, which have concluded that if additional T1D risk factors are present on this haplotype they are likely to be situated telomeric of HLA-A.9 Furthermore, it indicates that our susceptibility locus maps away from the Discussion classical MHC, although additional risk loci are also In this study, we fine-mapped a region in the extended believed to exist there, in particular for HLA-B.25,27–29 It is MHC class I region, which has earlier been shown to interesting to note that conditional logistic regression be negatively associated with allele D6S2223*3 only on analyses in the T1DGC dataset showed that the associa- the DRB1*03-DQA1*0501-DQB1*0201 haplotypic back- tions with the PRSS16 SNPs and one of the BTN SNPs ground.6 Here, we report associations also on the were independent of HLA-B, and also of the other class I DRB1*0401-DQA1*03-DQB1*0302 haplotypic back- loci. The associations with the PRSS16 SNPs also seemed ground with novel associations in two independent independent of the earlier reported UBD and HLA-G datasets for SNPs both in the PRSS16 gene region and associations, though some dependency was seen in the close to the BTN3A2 gene. regression analyses between one of the HLA-G SNPs and The observed SNP association patterns were more one of the PRSS16 SNPs, even though there was hardly confined on the DRB1*0401-DQA1*03-DQB1*0302 hap- any LD between these SNPs when calculated on the basis lotypic background compared with the more disorderly of the entire T1DGC dataset. Nevertheless, the LD associations on the DRB1*03-DQA1*0501-DQB1*0201 pattern could be different on specific DRB1-DQA1- background. This most likely reflects the underlying DQB1 haplotypes, but the haplotype analyses showed LD patterns on these two haplotypes, and we have that the SNP associations mainly mapped to different earlier shown more extensive LD on the DRB1*03- DRB1-DQA1-DQB1 haplotypes. Ultimately, whether DQA1*0501-DQB1*0201 compared with the DRB1*0401- these SNPs to some extent pick up the same association DQA1*03-DQB1*0302 haplotypic background in the signal will have to be readdressed when the causal initial dataset.15 In line with this, the DRB1*03- variant(s) have been identified. DQA1*0501-DQB1*0201 haplotype with its scattered In this study, we show the presence of additional risk association signals is probably difficult to use for fine- factors in the extended class I region, not only on the mapping purposes even though the D6S2223 association DRB1*03-DQA1*0501-DQB1*0201 haplotype, but also was found on this haplotype. In contrast, the DRB1*0401- likely to be present on the DRB1*0401-DQA1*03- DQA1*03-DQB1*0302 haplotype should provide a better DQB1*0302 haplotype. The associated regions are partly chance of getting closer to the causal variant(s), as the LD overlapping with those identified on the DRB1*03- stretches over shorter distances, which could also explain DQA1*0501-DQB1*0201 haplotype, indicating that these the lack of an association with D6S2223 on this novel risk factors are not restricted to a specific DRB1- haplotype. DQA1-DQB1 haplotype. Nevertheless, the variable LD The associations observed in the PRSS16 gene region patterns and allelic distributions on the two haplotypes on both haplotypic backgrounds were not with the same supported our choice to map these associations sepa- SNPs. The seven SNPs associated on the DRB1*0401- rately until a possible common denominator has been DQA1*03-DQB1*0302 haplotype were in perfect LD, and detected. all of them showed the same transmission distortion The associations on both the DRB1*03-DQA1*0501- (Table 1). The two polymorphisms (rs3800307 and DQB1*0201 and DRB1*0401-DQA1*03-DQB1*0302

Genes and Immunity Reproducible association with T1D in the xMHC MK Viken et al 330 haplotypes in the PRSS16 region may point towards a study. This segmental duplication has, to our knowledge, common risk variant for these two T1D high-risk not been studied regarding human diseases. On the basis haplotypes. Both single SNPs and five-SNP haplotypes of publicly available information from the Copy Number showed association in this region, and we also replicated Variation Project (http://www.sanger.ac.uk/humgen/ the association observed here. Recently Aly et al.24 cnv/) the segmental duplication was only reported reported an association with T1D upstream of PRSS16, within one of the CEPH families (HapMap), and hence but this SNP (rs996247) showed no association in our seem to be uncommon. Preliminary efforts by using a dataset. However, overall there is a strong evidence for relative quantification method to assess the presence of an association around the PRSS16 gene. The causal the segmental duplication have been unsuccessful (data variant is unlikely to be situated within the exons, not shown). Structural variations and copy number introns or promoter region, as our earlier investigation of alterations have recently been found to influence the the gene included a comprehensive screen for poly- risk of complex diseases, for example, a copy number morphisms.20 Instead, it is more likely that the risk variation in the FCGR3B gene was recently found to be variant(s) may be a regulatory variant influencing the associated with systemic lupus erythematosus.39,40 expression of PRSS16 or/and some other gene(s). Hence, the possible involvement of the segmental Analyses of SNP haplotypes on the DRB1*0401- duplication in our associated region should be further DQA1*03-DQB1*0302 background revealed, in addition investigated. to associations around the PRSS16 gene, an association In this study, we have for the first time found a covering part of the BTN-gene cluster, and SNPs in the reproducible association on the DRB1*0401-DQA1*03- region containing the BTN3A2 gene showed the most DQB1*0302 haplotype in the extended class I region pronounced association. It is interesting to note that the telomeric of D6S2223. Hence, this study supports the BTN3A2 gene has shown allelic imbalance in gene existence of additional T1D predisposing factors situated expression depending on the haplotype.23 However, in the extended MHC class I region, independent of the none of the two SNPs shown to be correlated with already known HLA DRB1-DQA1-DQB1 risk alleles. Our high- and low-expression transcripts were associated results suggest that the candidate regions for the on the DRB1*0401-DQA1*03-DQB1*0302 haplotype, but novel associations identified, possibly resides within rs9379857 showed association on the DRB1*03-DQA1* the BTN-gene cluster, but most appreciably around the 0501-DQB1*0201 haplotype. PRSS16 gene. Both PRSS16 and the BTN3A2 are good candidate genes for autoimmune diseases like T1D. PRSS16 encodes a thymus-specific serine protease that is highly Materials and methods expressed in cortical thymic epithelial cells, and believed to be involved in the regulation and/or positive selection Datasets of T cells.19 The functions of the proteins encoded by the The initial dataset was selected from 41000 T1D BTN cluster are not yet fully understood, but they are families, resulting in 205 families with at least one parent expressed on the cell surface in several immunological homozygous for the DRB1-DQA1-DQB1 haplotypes (66 tissues like the spleen, lymph nodes and thymus and Norwegian, 50 Danish, 69 Swedish and 20 French). Seven also in the pancreas,30,31 and an immunological role has families were removed after genotyping because of been implicated. The BTN genes have a strong domain incorrect inheritance patterns, bad DNA quality or poor similarity with other genes in the MHC, and with the genotyping success, resulting in 198 families for statis- B7.1 (CD80) and the B7.2 (CD86) costimulatory receptors tical analyses. A total of 81% of the families had at least on antigen-presenting cells.32 one parent homozygous for the DRB1*03-DQA1*0501- The history of genetic association studies is full of false DQB1*0201 (111 parents) or the DRB1*0401-DQA1*03- positive reports, as many fail to be reproduced. Over- DQB1*0302 (62 parents) haplotypes. All individuals were estimation of the effect in the initial study and/or lack of already genotyped for HLA-DRB1, -DQA1 and -DQB1 21 power in the replication studies may be at fault.33 The alleles and numerous microsatellites, and have earlier 4,6,15 fact that the association we identified with the PRSS16 been used in microsatellite screens of the xMHC. gene region was replicated in an independent family A replication dataset of 2746 samples (725 families; 258 material consisting of several different nationalities, families only having one parent) was obtained from the support that this finding is a true positive result. Still, T1DGC (http://www.t1dgc.org). For association ana- ideally the initial replication should have been per- lyses only one affected child per family was used (that is, formed by the same analyses used in the initial dataset.34 the proband when DNA was available or a randomly This was not possible, as we had no independent HLA- selected affected child (N ¼ 54)). As the underlying LD DRB1-DQA1-DQB1 homozygous T1D family material pattern is of great importance in this study, only available. In contrast, the association over the BTN-gene individuals of Caucasian origin were used. Genotypes cluster was not convincingly reproduced, and the for the HLA loci were available through the T1DGC, and association was only identified with SNP haplotypes in DRB1-DQA1-DQB1 haplotypes were manually recoded the initial dataset. into one locus on the basis of the strong LD and Furthermore, genetic association studies can be com- confirmed by the transmission patterns in the families 25 plicated by genetic structural variants, which are using the procedure described in Eike et al. increasingly being studied.35–43 Within our associated region, a segmental duplication is reported,44 which Selection of polymorphisms is situated between the two associated regions on In the first phase screen, 46 SNPs were selected on the the DRB1*0401-DQA1*03-DQB1*0302 haplotypic back- basis of the availability of SNPs in the Celera database in ground and caused a 223-kb gap not genotyped in our 2004 (http://www.celeradiscoverysystem.com, notably

Genes and Immunity Reproducible association with T1D in the xMHC MK Viken et al 331 Applied Biosystems closed this site for customers in Statistical methods July 2005). Particular interest in the PRSS16 gene led to Association analyses were performed using the Un- 17 SNPs and one insertion/deletion (rs5030663) being phased software package v2.403.47 To exclude confound- selected, within or in the proximity of PRSS16, on the ing associations because of LD with the DRB1-DQA1- basis of allele frequencies and LD patterns from a pilot DQB1 risk variants in the initial dataset, and to seek study (unpublished data). The second phase screen associations separately on the two high-risk and most extended the area of interest in the telomeric direction abundant DRB1-DQA1-DQB1 haplotypes, we employed and closed gaps in the first phase of the SNP screen. At the homozygous parent transmission disequilibrium this time more information was available in the test,6 that is, only transmissions from parents hetero- databases, and 97 SNPs were selected on the basis of zygous for the test marker, but homozygous for either allele frequencies, localization obtained from dbSNP DRB1*03-DQA1*0501-DQB1*0201 or DRB1*0401-DQA1*03- build 125 (http://www.ncbi.nlm.nih.gov/projects/SNP/) DQB1*0302, were included. or tagging ability based on the HapMap project data For LD analyses and Hardy–Weinberg equilibrium release #19 (CEPH, Utah residents with ancestry from tests, the Haploview software v3.2 and v3.32 were northern and western Europe)45 using pairwise tagging used (http://www.broad.mit.edu/mpg/haploview/).48 with r2X0.8 for SNPs with minor allele frequencies Two-loci haplotypes, comprising DRB1-DQA1-DQB1 (MAF)X0.05. Not all SNPs selected by pairwise tagging combined with one SNP, was estimated using the were genotyped owing to different selection strategies, the expectation-maximization algorithm in Unphased nature of the surrounding sequences or lack of multi- v2.403. Longer haplotypes were constructed using the plexing abilities in the genotyping system. Three candidate FAMHAP program with a one-directional progressive SNPs, that is, rs9379857 and rs1985732 were also included. extension technique and estimations with respect to Altogether, 164 SNPs were genotyped, covering an area of affection status.49 Individuals/families with missing B1.9 Mb (between D6S306 and D6S2233). genotypes were included when at least one genotype was available. Haplotype fragments of the 134 poly- Genotyping morphisms from the first and second phase, consisting of The majority of the 164 SNPs were genotyped using the five SNPs in a sliding window (131 sliding windows), SNPlex technology on an ABI3730 DNA sequencer showing a likelihood weight equal to one were included (Applied Biosystems, Foster City, CA, USA). Seventeen in the analysis. The distribution of SNPs on the SNPs were genotyped using the MassARRAY technology transmitted and non-transmitted haplotypes being (Sequenom, San Diego, CA, USA), four by allelic dis- either DRB1*03-DQA1*0501-DQB1*0201 or DRB1*0401- crimination (TaqMan) on an ABI7900 PRISM (Applied DQA1*03-DQB1*0302 was analyzed separately using Biosystems) and one insertion/deletion by fluorescent- the haplotype method.50 OR with corresponding 95%CI based fragment-length analysis on an ABI377 DNA were calculated using Woolf’s formula with Haldane’s analyzer (Applied Biosystems). Genotyping data for correction. A comparison between the association on the rs10946904, rs9393795, rs9368492 and rs5030663 from an DRB1*03-DQA1*0501-DQB1*0201 and the DRB1*0401- earlier study of the PRSS16 gene,21 performed only on the DQA1*03-DQB1*0302 haplotype was performed to DRB1*03-DQA1*0501-DQB1*0201 haplotypic background, search for overlapping associated regions. were included in the current study. Whole genome P-values o0.05 identified in the initial dataset were amplification of genomic DNA was performed before not corrected for multiple testing, but tried replicated in genotyping of the initial dataset using the GenomiPhi the independent T1DGC family material, not selected to Amplification Kit (GE Healthcare, Little Chalfont, UK).46 have DRB1-DQA1-DQB1 homozygous parents. Only 18 Single nucleotide polymorphisms with a genotype parents were homozygous for the DRB1*0401-DQA1*03- success rate o80% in the initial data set, MAFo0.05 or DQB1*0302 haplotype, hence the homozygous parent with unreliable genotype calling, were omitted, leaving transmission disequilibrium test could not be used 137 SNPs, in Hardy–Weinberg equilibrium, for statistical owing to the lack of power. Owing to the larger analyses. The average distance between the SNPs was population heterogeneity in the replication dataset 14 kb on the basis of the UCSC Genome Browser, May (individuals from North America, Australia and all over 2004 assembly (http://genome.ucsc.edu/). The mini- Europe were included), we were reluctant to embark on mum distance was 41 bp and the maximum distance was the haplotype method. Hence, we opted for a conditional a B223-kb gap (between rs9393768 and rs12201890), logistic regression analysis, thus employing the main caused by the difficulties in finding SNPs validated by effects test in Unphased v2.40347 with DRB1-DQA1- frequency in the database (dbSNP) and not aligning to DQB1 as the conditional locus. The procedure for the more than one position on chromosome 6, because of a grouping of DRB1-DQA1-DQB1 haplotypes is described segmental duplication in this region.35,44 in Eike et al.25 Nine SNPs (rs11754384, rs9358929, rs11758089, rs2073526, rs13218591, rs1624064, rs12201890, rs9368492 and rs9393796) were selected for replication in the T1DGC family dataset, and genotyped on genomic Acknowledgements DNA using either Taqman or SNPlex technology (Applied Biosystems). The genotyping success rate was This study was supported by the Juvenile Diabetes 495%. rs9368492 and rs9393796 were in perfect LD Research Foundation International (1-2004-793, 1-2000- (D0 ¼ 1 and r2 ¼ 1) on the DRB1*0401-DQA1*03- 514, 1-2000-515), the University of Oslo, the Functional DQB1*0302 haplotype in the initial dataset, but both Genomics program (FUGE), the Norwegian Research were replicated to ensure that information was not lost Council, the Norwegian Diabetes Association, the Novo- owing to differences in LD in the replication dataset. Nordisk Foundation and the Eastern Norway Regional

Genes and Immunity Reproducible association with T1D in the xMHC MK Viken et al 332 Health Authority. We thank the Norwegian Study Group 8 Aly TA, Ide A, Jahromi MM, Barker JM, Fernando MS, Babu for Childhood Diabetes, the Danish Study Group of SR et al. Extreme genetic risk for type 1A diabetes. Proc Natl Diabetes in Childhood, the Danish IDDM Epidemiology Acad Sci USA 2006; 103: 14074–14079. and Genetics Group, the Swedish Childhood Diabetes 9 Ide A, Babu SR, Robles DT, Wang T, Erlich HA, Bugawan TL Study, the Diabetes Incidence Study in Sweden, the et al. ‘Extended’ A1, B8, DR3 haplotype shows remarkable French Network Inserm for IDDM and MS in Atlantic linkage disequilibrium but is similar to nonextended haplo- Pyrenees (Dr A Cambon-Thomsen) and the Regional types in terms of diabetes risk. Diabetes 2005; 54: 1879–1883. Observatory of health in Aquitaine region (Dr J Doutreix) 10 de Bakker PI, McVean G, Sabeti PC, Miretti MM, Green T, Marchini J et al. A high-resolution HLA and SNP haplotype for the collection of samples used in this study and map for disease association studies in the extended human the bilateral French Norwegian collaborative program MHC. Nat Genet 2006; 38: 1166–1172. AURORA for allowing exchanges and the discussion of 11 Miretti MM, Walsh EC, Ke X, Delgado M, Griffiths M, Hunt S analyses. The MassARRAY genotyping service was et al. A high-resolution linkage-disequilibrium map of the provided by the national technology platform CIGENE human major histocompatibility complex and first generation (Centre For Integrative Genetics) supported by the of tag single-nucleotide polymorphisms. Am J Hum Genet 2005; Functional Genomics Program (FUGE) in the Research 76: 634–646. Council of Norway. This research uses resources pro- 12 Undlien DE, Lie BA, Thorsby E. HLA complex genes in type 1 vided by the Type 1 Diabetes Genetics Consortium, a diabetes and other autoimmune diseases. Which genes are collaborative clinical study sponsored by the National involved? Trends Genet 2001; 17: 93–100. 13 Ahmad T, Neville M, Marshall SE, Armuzzi A, Mulcahy- Institute of Diabetes and Digestive and Kidney Diseases Hawes K, Crawshaw J et al. Haplotype-specific linkage (NIDDK), the National Institute of Allergy and Infectious disequilibrium patterns define the genetic topography of the Diseases (NIAID), the National Human Genome Re- human MHC. Hum Mol Genet 2003; 12: 647–656. search Institute (NHGRI), the National Institute of Child 14 Yunis EJ, Larsen CE, Fernandez-Vina M, Awdeh ZL, Romero Health and Human Development (NICHD), and the T, Hansen JA et al. Inheritable variable sizes of DNA stretches Juvenile Diabetes Research Foundation International in the human MHC: conserved extended haplotypes and their (JDRF) and supported by U01 DK062418. Resources fragments or blocks. Tissue Antigens 2003; 62: 1–20. from the freely available Bioportal (http://www.bioportal. 15 Blomhoff A, Olsson M, Johansson S, Akselsen HE, Pociot F, uio.no) were used for some of the statistical analyses. We Nerup J et al. Linkage disequilibrium and haplotype blocks in are grateful to Kristina Gervin and Siri T Fla˚mfortheir the MHC vary in an HLA haplotype specific manner assessed excellent technical assistance, especially with the SNPlex mainly by DRB1*03 and DRB1*04 haplotypes. Genes Immun 2006; 7: 130–140. genotyping, Hege Dahlen Sollid for help with the 16 Horton R, Wilming L, Rand V, Lovering RC, Bruford EA, haplotype analyses and Morten C Eike for assistance with Khodiyar VK et al. Gene map of the extended human MHC. the compilation of the HLA data for the T1DGC families. Nat Rev Genet 2004; 5: 889–899. 17 Williams AF, Barclay AN. The immunoglobulin superfamily— domains for cell surface recognition. Annu Rev Immunol 1988; 6: 381–405. 18 Partridge J, Wallace DF, Robertson A, Fox MF, Simons JP, References Dooley JS et al. Cloning and molecular characterization of a cross-homologous zinc finger locus ZNF204. Genomics 1998; 1 The Wellcome Trust Case Control Consortium. Genome-wide 50: 116–118. association study of 14 000 cases of seven common 19 Bowlus CL, Ahn J, Chu T, Gruen JR. Cloning of a novel MHC- diseases and 3000 shared controls. Nature 2007; 447: encoded serine peptidase highly expressed by cortical 661–678. epithelial cells of the thymus. Cell Immunol 1999; 196: 80–86. 2 Sheehy MJ, Scharf SJ, Rowe JR, Neme de Gimenez MH, Meske 20 Lie BA, Akselsen HE, Bowlus CL, Gruen JR, Thorsby E, LM, Erlich HA et al. A diabetes-susceptible HLA haplotype is Undlien DE. Polymorphisms in the gene encoding thymus- best defined by a combination of HLA-DR and -DQ alleles. specific serine protease in the extended HLA complex: a J Clin Invest 1989; 83: 830–835. potential candidate gene for autoimmune and HLA-associated 3 Hanifi Moghaddam P, de Knijf P, Roep BO, Van der Auwera B, diseases. Genes Immun 2002; 3: 306–312. Naipal A, Gorus F et al. Genetic structure of IDDM1: two 21 Lie BA, Viken MK, Akselsen HE, Fla˚m S, Pociot F, Nerup J separate regions in the major histocompatibility complex et al. Association analysis in type 1 diabetes of the PRSS16 contribute to susceptibility or protection. Belgian Diabetes gene encoding a thymus-specific serine protease. Hum Registry. Diabetes 1998; 47: 263–269. Immunol 2007; 68: 592–598. 4 Johansson S, Lie BA, Todd JA, Pociot F, Nerup J, 22 Abecasis GR, Noguchi E, Heinzmann A, Traherne JA, Cambon-Thomsen A et al. Evidence of at least two type 1 Bhattacharyya S, Leaves NI et al. Extent and distribution of diabetes susceptibility genes in the HLA complex distinct linkage disequilibrium in three genomic regions. Am J Hum from HLA-DQB1, -DQA1 and -DRB1. Genes Immun 2003; 4: Genet 2001; 68: 191–197. 46–53. 23 Pastinen T, Sladek R, Gurd S, Sammak A, Ge B, Lepage P et al. 5 Koeleman BV, De Groot KN, Van Der Slik AR, Roep BO, A survey of genetic and epigenetic variation affecting human Giphart MJ. Association between D6S2223 and type I diabetes gene expression. Physiol Genomics 2004; 16: 184–193. independent of HLA class II in Dutch families. Diabetologia 24 Aly TA, Baschal EE, Jahromi MM, Fernando MS, Babu SR, 2002; 45: 598–599. Fingerlin TE et al. Analysis of single nucleotide polymorph- 6 Lie BA, Todd JA, Pociot F, Nerup J, Akselsen HE, Joner G et al. isms identifies major type 1A diabetes locus telomeric of the The predisposition to type 1 diabetes linked to the human major histocompatibility complex. Diabetes 2008; 57: 770–776. leukocyte antigen complex includes at least one non-class II 25 Eike MC, Becker T, Humphreys K, Olsson M, Lie BA. gene. Am J Hum Genet 1999; 64: 793–800. Conditional analyses on the T1DGC MHC dataset: novel 7 Robinson WP, Barbosa J, Rich SS, Thomson G. Homozygous associations with type 1 diabetes around HLA-G and parent affected sib pair method for detecting disease predis- confirmation of HLA-B. Genes and Immunity 2009; 10: 56–67. posing variants: application to insulin dependent diabetes 26 Todd JA, Walker NM, Cooper JD, Smyth DJ, Downes K, mellitus. Genet Epidemiol 1993; 10: 273–288. Plagnol V et al. Robust associations of four new chromosome

Genes and Immunity Reproducible association with T1D in the xMHC MK Viken et al 333 regions from genome-wide analyses of type 1 diabetes. Nat high copy number is a protective factor against SLE suscept- Genet 2007; 39: 857–864. ibility in European Americans. Am J Hum Genet 2007; 80: 27 Nejentsev S, Gombos Z, Laine AP, Veijola R, Knip M, Simell O 1037–1054. et al. Non-class II HLA gene associated with type 1 diabetes 39 Fanciulli M, Norsworthy PJ, Petretto E, Dong R, Harper L, maps to the 240-kb region near HLA-B. Diabetes 2000; 49: Kamesh L et al. FCGR3B copy number variation is associated 2217–2221. with susceptibility to systemic, but not organ-specific, auto- 28 Nejentsev S, Howson JM, Walker NM, Szeszko J, Field SF, immunity. Nat Genet 2007; 39: 721–723. Stevens HE et al. Localization of type 1 diabetes susceptibility 40 Aitman TJ, Dong R, Vyse TJ, Norsworthy PJ, Johnson MD, to the MHC class I genes HLA-B and HLA-A. Nature 2007; Smith J et al. Copy number polymorphism in Fcgr3 predis- 450: 887–892. poses to glomerulonephritis in rats and humans. Nature 2006; 29 Valdes AM, Erlich HA, Noble JA. Human leukocyte antigen 439: 851–855. class I B and C loci contribute to type 1 diabetes (T1D) suscepti 41 van Ommen G-JB. Frequency of new copy number variation in bility and age at T1D onset. Hum Immunol 2005; 66: 301–313. humans. Nat Genet 2005; 37: 333–334. 30 Rhodes DA, Stammers M, Malcherek G, Beck S, Trowsdale J. 42 Sebat J, Lakshmi B, Troge J, Alexander J, Young J, Lundin P The cluster of BTN genes in the extended major histocompat- et al. Large-scale copy number polymorphism in the human ibility complex. Genomics 2001; 71: 351–362. genome. Science 2004; 305: 525–528. 31 Ruddy DA, Kronmal GS, Lee VK, Mintier GA, Quintana L, 43 Goidts V, Cooper D, Armengol L, Schempp W, Conroy J, Domingo Jr R et al. A 1.1-Mb transcript map of the hereditary Estivill X et al. Complex patterns of copy number variation at hemochromatosis locus. Genome Res 1997; 7: 441–456. sites of segmental duplications: an important category of 32 Linsley PS, Peach R, Gladstone P, Bajorath J. Extending the B7 structural variation in the human genome. Hum Genet 2006; (CD80) gene family. Protein Sci 1994; 3: 1341–1343. 120: 270–284. 33 Colhoun HM, McKeigue PM, Smith GD. Problems of report- 44 Mungall AJ, Palmer SA, Sims SK, Edwards CA, Ashurst JL, ing genetic associations with complex outcomes. Lancet 2003; Wilming L et al. The DNA sequence and analysis of human 361: 865–872. chromosome 6. Nature 2003; 425: 805–811. 34 Clarke GM, Carter KW, Palmer LJ, Morris AP, Cardon LR. 45 The International HapMap Consortium. The International Fine-mapping vs replication in whole genome association HapMap Project. Nature 2003; 426: 789–796. studies. Am J Hum Genet 2007; 81: 995–1005. 46 Barker DL, Hansen MST, Faruqi AF, Giannola D, Irsula OR, 35 Redon R, Ishikawa S, Fitch KR, Feuk L, Perry GH, Andrews Lasken RS et al. Two methods of whole-genome amplification TD et al. Global variation in copy number in the human enable accurate genotyping across a 2320-SNP linkage panel. genome. Nature 2006; 444: 444–454. Genome Res 2004; 14: 901–907. 36 Stranger BE, Forrest MS, Dunning M, Ingle CE, Beazley C, 47 Dudbridge F. Pedigree disequilibrium tests for multilocus Thorne N et al. Relative impact of nucleotide and copy haplotypes. Genet Epidemiol 2003; 25: 115–121. number variation on gene expression phenotypes. Science 48 Barrett JC, Fry B, Maller J, Daly MJ. Haploview: analysis and 2007; 315: 848–853. visualization of LD and haplotype maps. Bioinformatics 2005; 37 Sebat J, Lakshmi B, Malhotra D, Troge J, Lese-Martin C, Walsh 21: 263–265. T et al. Strong association of de novo copy number mutations 49 Becker T, Knapp M. Maximum-likelihood estimation of with autism. Science 2007; 316: 445–449. haplotype frequencies in nuclear families. Genet Epidemiol 38 Yang Y, Chung EK, Wu YL, Savelli SL, Nagaraja HN, Zhou B 2004; 27: 21–32. et al. Gene copy-number variation and associated polymorph- 50 Valdes AM, Thomson G. Detecting disease-predisposing isms of complement component C4 in human systemic lupus variants: the haplotype method. Am J Hum Genet 1997; 60: erythematosus (SLE): low copy number is a risk factor for and 703–716.

Genes and Immunity