Diabetes Volume 69, July 2020 1573 Motifs of Three HLA-DQ Amino Acid Residues (a44, b57, b135) Capture Full Association With the Risk of Type 1 Diabetes in DQ2 and DQ8 Children Lue Ping Zhao,1 George K. Papadopoulos,2,§ William W. Kwok,3 Antonis K. Moustakas,4 George P. Bondinas,2 Helena Elding Larsson,5 Johnny Ludvigsson,6 Claude Marcus,7 Ulf Samuelsson,6 Ruihan Wang,8 Chul-Woo Pyo,8 Wyatt C. Nelson,8 Daniel E. Geraghty,8 and Åke Lernmark5 Diabetes 2020;69:1573–1587 | https://doi.org/10.2337/db20-0075 GENETICS/GENOMES/PROTEOMICS/METABOLOMICS HLA-DQA1 and -DQB1 are strongly associated with type motif “CAD” captured the risk association of DQ2.5 (OR 1 diabetes (T1D), and DQ8.1 and DQ2.5 are major risk 2.10, P 5 1.96 * 10220). Two risk associations were related haplotypes. Next-generation targeted sequencing of HLA- to GAD65 autoantibody (GADA) and IA-2 autoantibody (IA- DQA1 and -DQB1 in Swedish newly diagnosed 1- to 18 2A) but in opposite directions. CAD was positively asso- year-old patients (n 5 962) and control subjects (n 5 636) ciated with GADA (OR 1.56, P 5 6.35 * 1028) but negatively was used to construct abbreviated DQ haplotypes, con- with IA-2A (OR 0.59, P 5 6.55 * 10211). QAD was negatively vertedintoaminoacid(AA)residues,andassessedfortheir associated with GADA (OR 0.88; P 5 3.70 * 1023)but associations with T1D. A hierarchically organized haplo- positively with IA-2A (OR 1.64; P 5 2.40 * 10214), despite type (HOH) association analysis allowed 45 unique DQ a single difference at a44. The residues are found in and haplotypes to be categorized into seven clusters. The around anchor pockets 1 and 9, as potential T-cell re- DQ8/9 cluster included two DQ8.1 risk and the DQ9 ceptor contacts, in the areas for CD4 binding and putative resistant haplotypes, and the DQ2 cluster included the homodimer formation. The identification of three HLA-DQ DQ2.5 risk and DQ2.2 resistant haplotypes. Within each AAs (a44, b57, b135) conferring T1D risk should sharpen cluster, HOH found residues a44Q (odds ratio [OR] 3.29, 2 2 functional and translational studies. P 5 2.38 * 10 85)andb57A (OR 3.44, P 5 3.80 * 10 84)tobe associated with T1D in the DQ8/9 cluster representing all ten residues (a22, a23, a44, a49, a51, a53, a54, a73, a184, b57) due to complete linkage disequilibrium (LD) of a44 Type 1 diabetes (T1D) is one of the most common chronic with eight such residues. Within the DQ2 cluster and due diseases among children and young adults, and the disease to LD, HOH analysis found a44C and b135D to share the incidence is steadily rising worldwide, posing challenges to risk for T1D (OR 2.10, P 5 1.96 * 10220). The motif “QAD” of affected families and increasing financial stress on health a44, b57, and b135 captured the T1D risk association of care systems (1–4). The lifetime risk is .1% in both North DQ8.1 (OR 3.44, P 5 3.80 * 10284), and the corresponding America and Europe (5,6). Decades of research have shown 1Public Health Sciences Division, Fred Hutchinson Cancer Research Center, Corresponding authors: Lue Ping Zhao, [email protected], George K. Papadopoulos, Seattle, WA [email protected], and Åke Lernmark, [email protected] 2 Laboratory of Biophysics, Biochemistry, Biomaterials and Bioprocessing, Faculty Received 22 January 2020 and accepted 30 March 2020 of Agricultural Technology, Technological Educational Institute of Epirus, Arta, This article contains supplementary material online at https://doi.org/10.2337/ Greece db20-4567/suppl.12053640. 3Benaroya Research Institute at Virginia Mason, Seattle, WA 4Department of Food Science and Technology, Faculty of Environmental Sciences, §G.K.P. has been retired from Technological Educational Institute (TEI) of Epirus, Ionian University, Argostoli, Cephalonia, Greece Arta, Greece since 1 September 2018. The affiliation is given for identification 5Department of Clinical Sciences, Lund University CRC, Skåne University Hospital, purposes only. As of 1 October 2018, the TEI of Epirus has been absorbed by the Malmö, Sweden University of Ioannina. The respective department is now called Department of 6Crown Princess Victoria Children’s Hospital, Region Östergötland, and Division of Agriculture. Pediatrics, Department of Clinical and Experimental Medicine, Linköping Univer- © 2020 by the American Diabetes Association. Readers may use this article as sity, Linköping, Sweden long as the work is properly cited, the use is educational and not for profit, and the 7Division of Pediatrics, Department of Clinical Science, Intervention and Technol- work is not altered. More information is available at https://www.diabetesjournals ogy, Karolinska Institutet, Stockholm, Sweden .org/content/license. 8Clinical Research Division, Fred Hutchinson Cancer Research Center, Seattle, WA 1574 HLA-DQ Amino Acid Residues Modify T1D Risk Diabetes Volume 69, July 2020 that T1D is an autoimmune disease, in which the host that significantly associate with T1D susceptibility by immune system wrongfully attacks pancreatic islet b-cells using sequence-based HLA genotypes. One recent ap- (7). The initial immune cell activation is reflected in the proach is the conditional analysis, i.e., systematically generation of autoantibodies against specific autoantigens searching through all residues for most significantly present in the b-cells. These autoantibody biomarkers associated residues and then conditioning on selected appear first to either insulin (IAA) or GAD65 (GADA), residues and searching for the next significantly associ- representing two different endotypes (8–11). Approxi- ated residues (28). After its application to T1D-associated mately 60% of children with a first autoantibody within residues, the single residue b57 was found to explain 1 year developed a second autoantibody against insulin, .15% of class II association. However, this finding GAD65, IA-2 (IA-2A), or zinc transporter 8 (ZnT8A). Ge- should be interpreted cautiously because the local LD netic studies, from classical HLA studies and genome-wide could condition out “other critical residues” (29). To ease association studies (12–14), have confirmed that the stron- this concern, we recently applied an ROR to investigate gest genetic associations with T1D are with the class II HLA HLA-DRB genes with T1D risk and have been able to find genes on chromosome 6, dwarfing ;50 non-HLA genetic multiple residues and their motifs that capture all known associations (12,15–17). While several recent studies have DR associations (23). Further scrutiny of the ROR ap- attempted to distinguish between two adjacent MHC class II proach, when considering both high LD and high-order loci (HLA-DR and -DQ) (18–21), even extending to the interactions, has suggested that ROR is capable of finding HLA-DP locus (22), we have recently identified 11 residues multiple associated residues in high LD but may miss in HLA-DRB1 and 15 residues in HLA-DRB3, -DRB4, and others in the presence of high-order interactions. Con- -DRB5 that may include critical residues by applying the sequently, residues identified by ROR include critical recursive organizer (ROR) (23). residues but may still leave some critical variants Early T1D investigations showed that HLA-DQ may undiscovered. have greater roles in disease susceptibility than HLA-DR In this study, the aim is to identify potential T1D critical genes (24,25). HLA-DQ includes two genes (DQA1 and residues in HLA-DQA1 and -DQB1 that are mediated DQB1), located, respectively, at (32,637,357–32,643,652) through their interactions with multiple residues corre- and (32,659,467–32,666,684) on human genome hg38 sponding to protein structures. To conquer this aged chal- [UCSC Genome Browser on Human Dec. 2013 (GRCh38/ lenge, we revise the ROR approach to a hybrid approach. hg38) Assembly]. Due to their proximity (only 15,815 Clusters of AA sequences that potentially are indicative of bases apart), they are in high linkage disequilibrium (LD). different protein structures were first identified. This was Functionally, proteins coded by DQA1 and DQB1,respec- followed by searching for specificT1DAAsinahaplotype tively, form a heterodimer, as a typical MHC II molecule, association analysis using their specific AAs. This approach with the DQA1 and DQB1 polypeptides. Their biosynthesis is hereafter referred to as a hierarchically organized haplo- in-vivo occurs in concert with that of the invariant chain type (HOH) association analysis. Briefly, HOH carries out (26), which associates with the MHC II ab heterodimer phylogenic analysis of AA sequences of unique DQA1-DQB1 having the CLIP peptide portion bound into the groove of haplotypes and hierarchically organizes all DQ haplotypes; this heterodimer (27). i.e., the shorter the distance between two DQ haplotypes, Mature DQa1andDQb1 protein molecules consist of, the closer they will be placed (see the flowchart in Supple- respectively, 232 and up to 237 (because of polymorphism mentary Fig. 1 for details). Two clusters of DQ haplotypes in the length of the intracytoplasmic tail) amino acids emerge: one cluster includes two high-risk DQ8.1 haplo- (AAs) (for the intricacies of the HLA-DQA and -DQB AA types, in addition to a resistant DQ9 haplotype, and the numbering system see Research Design and Methods). other cluster includes a high-risk DQ2.5 haplotype. The first Note that AA and residue are used interchangeably here- cluster was referred to thus as DQ8/9 to indicate the after. Within the predominantly Caucasoid population, inclusion of both DQ8 and DQ9 haplotypes, while the many residues are monomorphic, along with some other second cluster was referred to as the DQ2 cluster. Within residues that are polymorphic but not identified with any each cluster, HOH carries out T1D association analysis specific function. The critical residues for T1D risk are with individual residues within their respective cluster, expected to exhibit strong associations with T1D.