Diabetes Page 2 of 55

1 A genome-wide association study confirming a strong effect of HLA and 2 identifying variants in CSAD/lnc-ITGB7-1 on chromosome 12q13.13 associated 3 with susceptibility to fulminant type 1 diabetes 4 5 Running title: GWAS of fulminant type 1 diabetes 6 7 Yumiko Kawabata1,*, Nao Nishida2,3,*, Takuya Awata4,†, Eiji Kawasaki5,†, 8 Akihisa Imagawa6, Akira Shimada7, Haruhiko Osawa8, Shoichiro Tanaka9, 9 Kazuma Takahashi10, Masao Nagata11, Hisafumi Yasuda12, Yasuko Uchigata13, 10 Hiroshi Kajio14, Hideichi Makino15, Kazuki Yasuda16,†, Tetsuro Kobayashi17,‡, 11 Toshiaki Hanafusa18, ‡, Katsushi Tokunaga3,†,§, and Hiroshi Ikegami1,†,§ 12 13 1 Department of Endocrinology, Metabolism and Diabetes, Kindai University Faculty of Medicine, Osaka, Japan 14 2 Research Center for Hepatitis and Immunology, National Center for Global Health and Medicine, Chiba, Japan 15 3 Department of Human Genetics, Graduate School of Medicine, The University of Tokyo, Tokyo, Japan 16 4 Department of Diabetes, Metabolism and Endocrinology, International University of Health and Welfare Hospital, Tochigi, Japan 17 5 Diabetes Center, Shin-Koga Hospital, Fukuoka, Japan 18 6 Department of Internal Medicine (I), Osaka Medical College, Osaka, Japan 19 7 Department of Endocrinology and Diabetes, Saitama Medical University, Saitama, Japan 20 8 Department of Laboratory Medicine, Ehime University School of Medicine, Ehime, Japan 21 9 Ai Home Clinic Toshima, Tokyo, Japan 22 10 Faculty of Nursing and Graduate School Nursing, Iwate Prefectural University, Iwate, Japan 23 11 Department of Internal Medicine, Takasago Municipal Hospital, Hyogo, Japan 24 12 Division of Health Sciences, Department of Public Health, Kobe University Graduate School of Health Sciences, Hyogo, Japan 25 13 Diabetes Center, Tokyo Women's Medical University School of Medicine, Tokyo, Japan 26 14 Division of Diabetes, Metabolism and Endocrinology, Department of Internal Medicine, National Center for Global Health and 27 Medicine Hospital, Tokyo, Japan 28 15 Diabetes Center, Shiraishi Hospital, Ehime, Japan 29 16 Department of Metabolic Disorder, Diabetes Research Center, Research Institute, National Center for Global Health and 30 Medicine, Tokyo, Japan 31 17 Okinaka Memorial Institute for Medical Research, Tokyo, Japan 32 18 Sakai City Medical Center, Osaka, Japan 33 34 * Y.K. and N.N. contributed equally to this work. 35 † Core member for genetic analysis in the committee on type 1 diabetes, Japan Diabetes Society. 36 ‡ Cochair of the committee on type 1 diabetes, Japan Diabetes Society. 37 38 § Corresponding authors: 39 Hiroshi Ikegami, Email: [email protected]; TEL: +81-72-366-0221 40 Department of Endocrinology, Metabolism and Diabetes, Kindai University Faculty of Medicine, Osaka 589-8511, Japan 41 Katsushi Tokunaga, Email: [email protected]; TEL: +81-3-5841-3303 42 Department of Human Genetics, Graduate School of Medicine, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-033, 43 Japan 44 Word counts: 4083, Number of tables: 3, Number of figures: 2

1

Diabetes Publish Ahead of Print, published online December 14, 2018 Page 3 of 55 Diabetes

1 Abstract 2 3 The first genome-wide association study of fulminant type 1 diabetes mellitus was 4 performed in Japanese individuals. As previously reported using a candidate gene 5 approach, a strong association was observed with multiple SNPs in the HLA region, and 6 the strongest association was observed with rs9268853 in the class II DR region (P=1.56 7 x 10-23, odds ratio [OR] 3.18). In addition, rs11170445 in CSAD/lnc-ITGB7-1 on 8 chromosome 12q13.13 showed an association at a genome-wide significance level 9 (P=7.58 x 10-9, OR 1.96). Fine mapping of the region revealed that rs3782151 in 10 CSAD/lnc-ITGB7-1 showed the lowest P value (P=4.60 x 10-9, OR 1.97 [95% CI 1.57- 11 2.48]). The risk allele of rs3782151 is a cis-expression quantitative trait locus (eQTL) for 12 ITGB7 that significantly increases the expression of this gene. CSAD/lnc-ITGB7-1 was 13 found to be strongly associated with susceptibility to fulminant, but not classical, 14 autoimmune type 1 diabetes, implicating this locus in the distinct phenotype of fulminant 15 type 1 diabetes. 16

2 Diabetes Page 4 of 55

1 Type 1 diabetes is caused by the destruction of the insulin-producing -cells of the

2 pancreas in genetically susceptible individuals. Etiologically, type 1 diabetes consists of

3 two subtypes: autoimmune (type 1A) and idiopathic (type 1B) (1, 2). In contrast to the

4 extensive studies on the genetics, pathogenesis, prevention, and treatment of autoimmune

5 type 1 diabetes, studies on idiopathic type 1 diabetes are very limited due to the

6 heterogeneous and ambiguous nature of this subtype. Among idiopathic type 1 diabetes

7 subtypes, fulminant type 1 diabetes is an established entity with well-characterized

8 clinical phenotypes (3-5).

9 Fulminant type 1 diabetes is clinically distinct from autoimmune type 1 diabetes;

10 onset is remarkably abrupt, as reflected by near-normal glycated hemoglobin (HbA1c)

11 levels despite very high blood glucose levels, which results in the complete destruction

12 of -cells within a few days. Diabetes-related autoantibodies are essentially negative in

13 patients with fulminant type 1 diabetes (3). In addition to β-cells, α-cell areas are also

14 decreased (6, 7), and mononuclear cell infiltration is observed in the exocrine and

15 endocrine pancreas in recent-onset patients with fulminant type 1 diabetes (5, 7). These

16 observations suggest that the whole pancreas is involved in fulminant type 1 diabetes,

17 which is distinct from the selective destruction of β-cells in autoimmune type 1 diabetes.

18 The genetic basis of fulminant type 1 diabetes is also distinct from that of classical

19 autoimmune type 1 diabetes. This distinction is evident from the marked difference in

20 incidences among different populations. The frequencies of type 1 diabetes in Japan and

21 most East Asian countries are very low; the frequency is typically less than 1/10 of that

22 in white populations of European descent (8). In contrast, most fulminant type 1 diabetes

23 cases are from East Asian countries (4, 5), and only a limited number of cases were

24 reported in white European populations (9). However, increased attention has recently

3 Page 5 of 55 Diabetes

1 been focused on this disease because of the high frequency of fulminant type 1 diabetes

2 in subjects undergoing cancer immunotherapy with immune-checkpoint inhibitors, such

3 as anti-PD-1 and anti-PD-L1 antibodies, in both European and Asian populations (10,

4 11).

5 An accelerated immune reaction triggered by viral infection in genetically

6 susceptible individuals has been proposed to cause the rapid and massive destruction of

7 pancreatic islets in patients with fulminant type 1 diabetes (5, 12), but the etiology of the

8 disease remains largely unknown. The identification of susceptibility genes for fulminant

9 type 1 diabetes is therefore important to clarify the pathogenesis and molecular

10 mechanisms of the disease and to establish effective prediction, prevention, and

11 intervention methods. Information on the molecular mechanisms of fulminant type 1

12 diabetes will also provide novel insights into the molecular mechanisms of type 1 diabetes

13 in general, including type 1 diabetes associated with cancer immunotherapy. However,

14 with the exception of HLA (13-15), the genetic susceptibility to fulminant type 1 diabetes

15 is largely unknown. To identify susceptibility genes for fulminant type 1 diabetes, we

16 performed a genome-wide association study (GWAS) in the Japanese population.

17

18 RESEARCH DESIGN AND METHODS

19

20 Study participants

21 Unrelated Japanese patients with fulminant type 1 diabetes (n=257) were recruited

22 through a nation-wide effort orchestrated by a committee of the Japan Diabetes Society.

23 Fulminant type 1 diabetes was diagnosed by experts in diabetes according to the criteria

24 of the Japan Diabetes Society (16) and confirmed by a review committee on fulminant

4 Diabetes Page 6 of 55

1 type 1 diabetes, which is part of the committee on type 1 diabetes of the Japan Diabetes

2 Society.

3 The diagnostic criteria for fulminant type 1 diabetes were as follows: 1) occurrence of

4 diabetic ketosis or ketoacidosis soon (within approximately 7 days) after the onset of

5 hyperglycemic symptoms, 2) a plasma glucose level ≥16.0 mmol/L (≥288 mg/dL) and a

6 glycated hemoglobin level <8.7% at the first visit, and 3) urinary C-peptide excretion <10

7 μg/day or a fasting serum C-peptide level <0.3 ng/mL (<0.10 nmol/L) and <0.5 ng/mL

8 (<0.17 nmol/L) after intravenous glucagon (or after a meal) load at onset. A diagnosis of

9 fulminant type 1 diabetes mellitus was confirmed if all three criteria were present (16).

10 The mean (± SD) levels of these variables in the present study were as follows: plasma

11 glucose, 46.3 ± 21.6 mmol/L (833.2 ± 389.8 mg/dL); HbA1c, 6.57 ± 0.72% (48.3 ± 7.9

12 mmol/mol); urinary C-peptide excretion, 3.3 ± 2.4 μg/day; and fasting C-peptide, 0.031 ±

13 0.021 nmol/L (0.093 ± 0.062 ng/mL). The control subjects for the GWAS were 419

14 healthy Japanese volunteers used in a previous GWAS (17, 18).

15 Patients with classical autoimmune type 1 diabetes (n=410) were also recruited

16 through the committee of the Japan Diabetes Society (19). The characteristics of the study

17 participants are summarized in Supplementary Table 1.

18 This study was approved by the ethics committees of the Japan Diabetes Society and

19 each institute that participated in this project. Informed consent was obtained from all the

20 participants.

21

22 Genotyping and data cleaning

23 Genotyping for 600,307 SNPs was performed with 257 genomic DNA samples extracted

24 from fulminant type 1 diabetes patients using the AxiomTM Genome-Wide ASI 1 Array

5 Page 7 of 55 Diabetes

1 (Affymetrix, Santa Clara, CA, USA). The genotype calls for these 600K SNPs obtained

2 with genotype data from 257 fulminant type 1 diabetes patients and 419 healthy

3 volunteers were determined using Genotyping Console software (version 4.2). The

4 GWAS genotype data from the healthy volunteers were previously acquired using the

5 AxiomTM Genome-Wide ASI 1 Array and are commonly used for various studies as

6 general population data. All the samples had an overall call rate of more than 97%, with

7 an average overall call rate of 99.42% (min, 97.86; max, 99.82%), and passed a

8 heterozygosity check. No related individuals (PI ≥0.1) were identified by identity-by-

9 descent testing. A principal component analysis was carried out to check the genetic

10 background of the 257 fulminant type 1 diabetes samples, 419 healthy Japanese

11 volunteers and samples from the International HapMap Project (43 Japanese in Tokyo

12 (JPT), 40 Han Chinese in Beijing (CHB), 91 Yoruba in Ibadan (YRI) and 91 Utah

13 Residents (CEPH) with Northern and Western European ancestry (CEU) samples)

14 (Supplementary Fig. 1). Data cleaning was performed for SNP quality control according

15 to the following criteria: SNP call rate of 95% or higher in both the cases and controls,

16 minor allele frequency (MAF) of 5% or higher in both the cases and controls, and no

17 extreme departure from the Hardy-Weinberg equilibrium P-value ≥0.001 in the controls

18 (Supplementary Table 2). All cluster plots for SNPs with P < 0.0001 based on a Chi-

19 square test of the allele frequency model were checked by visual inspection, and SNPs

20 with ambiguous genotype calls were excluded. Of the SNPs on autosomal chromosomes,

21 426,851 SNPs finally passed the quality control filters and were used for the association

22 analysis. A quantile-quantile plot of the distribution of test statistics for the comparison

23 of genotype frequencies in fulminant type 1 diabetes cases and healthy controls showed

24 that the inflation factor lambda was 1.061 for all the tested SNPs, including those in the

6 Diabetes Page 8 of 55

1 HLA region (from HLA-F to KIFC1; chr. 6: 29,645,000-33,365,000; 3.72 Mb [hg19]),

2 and was 1.046 when SNPs in the HLA region were excluded (Supplementary Fig. 2).

3 Because early inflated test statistics were obtained after exclusion of the SNPs in the HLA

4 region, we performed a logistic regression test using the top two components as covariates

5 as well as an association test correcting for stratification using the genomic control

6 approach (20).

7

8 Fine mapping

9 We selected 13 additional SNPs in the region of the top-hit SNP on chromosome

10 12q13.13 based on the linkage disequilibrium (LD) and the MAF (>5%, except for

11 rs3817537 in ITGB7, for which the MAF was 2.93%). The tagging SNPs were selected

12 based on genotype data from the International HapMap Project (21) and the NBDC

13 Human Database (URL: https://humandbs.biosciencedbc.jp/en/). These SNPs and the 3

14 SNPs, rs11170445, rs4606556 and rs2272299, with the lowest P values from the GWAS

15 were genotyped by real-time PCR analysis with TaqMan probes (Applied Biosystems,

16 Tokyo, Japan). The haplotypes were estimated using the PHASE program version 2.1

17 (22, 23).

18

19 Genotype imputation

20 The imputation was conducted with the IMPUTE2 (v2.3.2) program using the haplotype

21 reference panel from Phase I of the 1000 Genomes Project without singleton sites (24).

22 To perform genome-wide imputation while avoiding the restrictions of the program, 10-

23 Mbp intervals were set along the chromosomes, and imputation was repeated for all

24 intervals.

7 Page 9 of 55 Diabetes

1

2 Stratification by HLA

3 The association of the top-hit SNP rs3782151 in CSAD with fulminant type 1 diabetes

4 was evaluated relative to HLA with the top-hit SNP, rs9268853, in the HLA-DR region

5 in all subjects and with HLA haplotypes in a subset of subjects whose HLA genotypes

6 were available (419 control subjects and 216 patients with fulminant type 1 diabetes).

7 HLA-DRB1 and DQB1 were genotyped by PCR using sequence-specific primers and

8 PCR sequence-specific oligonucleotide methods as previously reported (13, 15).

9

10 Re-sequencing of the candidate gene

11 The CSAD region was resequenced in 32 participants who were homozygous for the risk

12 allele at rs3782151. A 23-kb region of CSAD (Chr.12: 53,551,446-53,574,693) was

13 sequenced with the Ion AmpliSeqTM technology. The target region was enriched with the

14 AmpliSeqTM custom DNA panel, the Ion AmpliSeqTM Library Kit 2.0, and the Ion

15 XpressTM Barcode Adapters Kits (TaKaRa Bio, Japan). All coding exons and exon-intron

16 junctions and 70% of the untranslated regions (UTRs) and introns were successfully

17 amplified and sequenced with the Ion ProtonTM System, resulting in sequence results of

18 17.16-kb region (Supplementary Fig. 3, green). Due to technical difficulty, sequence

19 results in some of the introns and UTRs were not obtained (Supplementary Fig. 3, red).

20 The alignment and identification of variants were performed using Torrent Suite software

21 with NCBI build 37 (GRCh37/hg19) as the reference human genome. The annotation of

22 nucleotide variants was performed with Ion ReporterTM Software.

23 The His288Arg variant of CSAD was genotyped using real-time PCR analysis with

24 TaqMan probes (Applied Biosystems, Tokyo, Japan) in 257 patients with fulminant type

8 Diabetes Page 10 of 55

1 1 diabetes, 410 patients with autoimmune type 1 diabetes, and 357 control subjects (13).

2

3 Statistical analysis

4 Allele data were analyzed in 2 x 2 contingency tables using the Chi-square test. The LD

5 and haplotype analyses were performed using Haploview 4.2 software (25). Here, each

6 haplotype was assumed to be one of the alleles at a biallelic locus, and the other

7 haplotypes were assumed to be the other allele. For example, the haplotype

8 ACACTGAAGAGTC and the other haplotypes were designated “A allele” and “B allele”,

9 respectively. The meta-analysis was performed using the Mantel-Haenszel method

10 (fixed-effects models). The P values for the heterogeneity among the panels joined in the

11 Mantel-Haenszel tests were all >0.05.

12

13 Results

14 A highly significant association was observed for multiple SNPs in the HLA region on

15 chromosome 6, and the strongest association was found for rs9268853 in the HLA class

16 II DR region (P=1.56 x 10-23, odds ratio [OR] 3.18, 95% confidence interval [CI] 2.53-

17 4.01) (Fig. 1, Supplementary Fig. 4). In addition, a total of 11 SNPs outside the HLA

18 region showed some evidence of association (P < 1.0 x 10-5) (Supplementary Table 3). In

19 particular, rs11170445 on chromosome 12q13.13 showed an association with genome-

20 wide significance (P=7.58 x 10-9, OR 1.96, [95% CI: 1.56-2.46]), and this evidence

21 provides the first indication of a region outside the HLA region that exhibits a genome-

22 wide significant association with fulminant type 1 diabetes. Two SNPs (rs4606556 and

23 rs2272299) located in the vicinity of rs11170445 also showed low P values

24 (Supplementary Table 3). To remove a potential effect of population stratification, we

9 Page 11 of 55 Diabetes

1 performed a logistic regression test using the top two components as covariates and an

2 association test correcting for stratification using the genomic control approach. Both

3 analyses showed a significant association for rs11170445 with P=6.39 x 10-8 and P=7.58

4 x 10-9, respectively.

5 Fine mapping of the region surrounding rs11170445 identified multiple SNPs with low

6 P values, and the strongest association was observed for rs3782151 (P=4.60 x 10-9, OR

7 1.97 [95% CI 1.57-2.48]) (Table 1, Fig. 2, Supplementary Fig. 5). Association tests using

8 imputed genotypes with data from the 1000 Genomes Project revealed similar results

9 (Table 1, Supplementary Table 4). An LD analysis identified a 65-kb LD block containing

10 three protein-coding genes (CSAD, ZNF740 and ITGB7) (Fig. 2C, D, Supplementary Fig.

11 6). The top-hit SNP in the GWAS, rs11170445, and that identified by fine mapping,

12 rs3782151, are in complete LD in healthy individuals and strong LD (r2=0.98) in

13 fulminant type 1 diabetes patients. When conditional analysis by controlling for

14 rs3782151 in CSAD was carried out, neither rs2272299 in ZNF740 nor rs4606556 in

15 ITGB7 was associated with fulminant type 1 diabetes (Supplementary Table 5). A

16 haplotype association test using all 13 SNPs in LD block 1 showed that all the risk

17 haplotypes associated with fulminant type 1 diabetes contained the minor A allele of

18 rs3782151. Furthermore, no haplotype showed a lower P value than the top-hit SNP,

19 rs3782151 (Supplementary Table 6), which suggests that the minor A allele of rs3782151

20 in CSAD is primarily associated with susceptibility to fulminant type 1 diabetes.

21 Because HLA confers strong susceptibility to fulminant type 1 diabetes, we examined

22 the interaction of SNPs in CSAD with HLA. A regression analysis using a top-hit SNP in

23 the HLA region, rs9268853, as a covariate showed that rs11170445, a top-hit SNP outside

24 the HLA region identified in the GWAS, exhibited a strong association with fulminant

10 Diabetes Page 12 of 55

1 type 1 diabetes (P=6.23 x 10-7) (Supplementary Fig. 7). No significant interaction was

2 observed for rs3782151, a top-hit SNP in the CSAD identified in the fine mapping, with

3 rs9268853 in the HLA region by the heterogeneity test (P=0.332).

4 In addition to the above-mentioned comparison with the top-hit SNP in the HLA

5 region, the interactions of rs372151 with HLA haplotypes and genotypes were also

6 assessed. The Asian-specific DR4 (DRB1*04:05-DQB1*04:01) and DR9 (DRB1*09:01-

7 DQB1*03:03) haplotypes were significantly associated with fulminant type 1 diabetes

8 (Supplementary Table 7), as previously reported (13-15). No heterogeneity in the strength

9 of the association was identified depending on the presence or absence of susceptible

10 HLA haplotypes (Supplementary Table 7). A top-hit SNP in the HLA region, rs9268853,

11 was in linkage disequilibrium with DR4 and DR9 haplotypes (Supplementary Table 7C).

12 The differences in the HLA region between patients with fulminant type 1 diabetes who

13 were or not pregnant have been previously reported (26). To minimize the heterogeneity

14 in fulminant type 1 diabetes, however, only nine pregnant patients with fulminant type 1

15 diabetes were included in the present study and thus, a subgroup analysis was not

16 performed.

17 We also evaluated the association of rs3782151 with classical autoimmune type 1

18 diabetes. The association of rs3782151 with autoimmune type 1 diabetes was weak (OR

19 1.31, P=0.011) compared with its very strong association with fulminant type 1 diabetes

20 (OR 1.97, P=4.60 x 10-9) (Table 2). The frequency of the minor allele at rs3782151 was

21 significantly higher in fulminant type 1 diabetes compared with autoimmune type 1

22 diabetes (0.451 vs. 0.352, P=3.13 x 10-4) (Table 2), suggesting that the association of

23 rs3782151 in CSAD with type 1 diabetes is unique to the fulminant subtype.

24 In addition to rs3782151 in CSAD, we investigated the association of SNPs in ITGB7

11 Page 13 of 55 Diabetes

1 with type 1 diabetes because rs11170466 in ITGB7, which is located 33 kb distal to

2 rs3782151, has been reported to be associated with autoimmune type 1 diabetes in

3 populations of European descent (27). Two SNPs in ITGB7 (rs2272299 in the present

4 study and rs11170466 in reference 27) are in the same LD block and were in complete

5 linkage disequilibrium in the 419 control samples included in the present study. The

6 association of rs3782151 in CSAD (P=4.60 x 10-9) with fulminant type 1 diabetes was

7 much stronger than the association of rs2272299 in ITGB7 (Table 2). The frequencies of

8 the minor allele at rs2272299 in ITGB7 were not significantly different between fulminant

9 and autoimmune type 1 diabetes (P=0.342), which is in clear contrast to the above-

10 described findings obtained for the risk allele of rs3782151 in CSAD in fulminant type 1

11 diabetes (Table 2).

12 To investigate the contribution of the CSAD-ITGB7 region to type 1 diabetes in

13 different ethnic groups, we genotyped rs3782151 in CSAD and rs2272299 in ITGB7 in

14 populations of European descent. Due to the near absence of fulminant type 1 diabetes in

15 European populations (4, 5), only autoimmune type 1 diabetes was studied. Autoimmune

16 type 1 diabetes was associated with rs2272299 in ITGB7, but not with rs3782151 in CSAD,

17 in a large scale study of the European population in Type 1 diabetes Genetics Consortium

18 (Supplementary Table 8). This tendency is similar to that found in the Japanese

19 population, in which ITGB7 rs2272299 (OR 1.63, P=0.0001) showed a stronger

20 association than CSAD rs3782151 (OR 1.31, P=0.011). A meta-analysis of the two

21 populations showed an association between rs2272299 in ITGB7 (summary OR 1.21

22 [1.10-1.34], 9.68 x 10-5), but not rs3782151 in CSAD (summary OR 1.05 [1.00-1.12], NS),

23 and autoimmune type 1 diabetes.

24 An analysis of the CSAD-ITGB7 haplotypes indicated that the A-A, but not the A-G,

12 Diabetes Page 14 of 55

1 haplotype was associated with autoimmune type 1 diabetes (Supplementary Tables 9 and

2 10A), indicating a primary association for ITGB7 with autoimmune type 1 diabetes. In

3 contrast, a positive association was found for the A-A and A-G haplotypes with fulminant

4 type 1 diabetes, but a negative association was obtained for the C-G haplotype, indicating

5 a primary association for CSAD with fulminant type 1 diabetes (Supplementary Table 9).

6 This finding suggests the existence of two distinct loci for type 1 diabetes in the CSAD-

7 ITGB7 region—one in CSAD for the fulminant subtype and the second in ITGB7 for the

8 autoimmune subtype. A haplotype analysis in subjects of European descent indicated that

9 the A-A haplotype was significantly associated with susceptibility to autoimmune type 1

10 diabetes, similarly to the findings obtained for the Japanese population (Supplementary

11 Tables 9 and 10A). The A-G haplotype, however, was associated with protection against

12 autoimmune type 1 diabetes in the European population, in contrast to its neutral effect

13 on autoimmune type 1 diabetes and its association with susceptibility to fulminant type 1

14 diabetes in the Japanese population. The frequencies of the various haplotypes were

15 markedly different between the two populations, with a much lower frequency of the A-

16 A haplotype and a higher frequency of the A-G haplotype in the population of European

17 descent compared with those in the Japanese population (Supplementary Table 10B).

18 To clarify the contribution of CSAD to disease susceptibility, we sequenced the CSAD

19 region in 32 individuals with fulminant type 1 diabetes who were homozygous for the

20 risk allele at rs3782151 and identified 31 single nucleotide variants (SNVs), including

21 one non-synonymous (Chr12:53161148 (hg19), His288Arg) variant in the coding region

22 (Supplementary Table 11). To study the contribution of the Chr12:53161148 (hg19)

23 nucleotide change to susceptibility to fulminant type 1 diabetes, a total of 1,024 subjects

24 were genotyped. The allele frequency of this variant was 0.2% in fulminant type 1

13 Page 15 of 55 Diabetes

1 diabetes [1/514 chromosomes], 0% in autoimmune type 1 diabetes [0/820 chromosomes],

2 and 0% in the control subjects [0/714 chromosomes].

3 The CSAD region encodes not only CSAD but also a long non-coding RNA (lncRNA)

4 termed RP11-1136G11.7-001 (also known as lnc-ITGB7-1:1) (28). The top-hit SNP

5 rs3782151 is located within RP11-1136G11.7 (lnc-ITGB7-1) (Supplementary Fig. 8). To

6 clarify the contribution of the CSAD/lnc-ITGB7-1 region to the expression of nearby

7 genes, we searched a database of cis-expression quantitative trait loci (cis-eQTLs). The

8 top-hit SNP rs3782151 has been reported to be a cis-eQTL of ITGB7 in populations of

9 European descent (OR 1.97, P=4.60 x 10-9) (Table 3) (29). In addition, several SNPs

10 flanking rs3782151 in the CSAD/lnc-ITGB7-1 region have been reported to be cis-eQTLs

11 of ITGB7 in the Japanese population (Table 3, Supplementary Fig. 8) (30, 31).

12

13 Discussion

14 The first genome-wide association study of fulminant type 1 diabetes mellitus was

15 performed in Japanese individuals. In addition to HLA, which was previously identified

16 by a candidate gene approach, variants in CSAD/lnc-ITGB7-1 on chromosome 12q13.13

17 were associated with fulminant type 1 diabetes at a genome-wide significance level

18 (Table 1, Fig. 2).

19 CSAD encodes cysteine sulfinic acid decarboxylase, which is a key enzyme in taurine

20 synthesis. Taurine has been reported to exert anti-inflammatory and cytoprotective effects

21 by attenuating apoptosis and stimulating antioxidant activity (32-36). The contribution of

22 taurine to the protection of pancreatic islets from destruction has been reported in both

23 type 1 diabetes and streptozotocin-induced apoptosis (37, 38), suggesting that CSAD

24 variants might contribute to fulminant type 1 diabetes by impairing the protection of

14 Diabetes Page 16 of 55

1 pancreatic islets. Since the top-hit SNP rs3782151 is located in an intronic region of

2 CSAD (Fig. 2, Supplementary Fig. 8), we sequenced the CSAD/lnc-ITGB7-1 region and

3 identified 31 single nucleotide variants, including one non-synonymous

4 (Chr12:53161148 (hg19), His288Arg) variant in the coding region (Supplementary Table

5 11). This variant was found in 0.2% of patients with fulminant type 1 diabetes but was

6 not observed in autoimmune type 1 diabetes or control subjects. In addition, the variant

7 is not present in the Genome Aggregation Database (gnomAd) derived from 123,136

8 exome sequences and 15,496 whole-genome sequences from unrelated individuals

9 sequenced as part of various disease-specific and population genetic studies

10 (http://gnomad.broadinstitute.org/) (39) and we detected only one heterozygote among

11 the 3,408 Japanese individuals in the Tohoku Medical Megabank, which is the largest

12 genome database based on whole-genome sequences of the Japanese general population

13 (https://ijgvd.megabank.tohoku.ac.jp/) (40). These findings indicate that variants in the

14 protein-coding region of CSAD are unlikely to be a common cause of fulminant type 1

15 diabetes.

16 lncRNAs are generally involved in the regulation of gene expression in many

17 biological systems, including the immune system (41, 42), and the contribution of

18 lncRNAs to inflammatory and immune-related diseases has also been reported (41-43).

19 The top-hit SNP rs3782151 identified in this study is located within lnc-ITGB7-1

20 (Supplementary Fig. 8) and is reportedly a cis-eQTL of ITGB7 obtained from peripheral

21 blood in populations of European descent (Table 3) (29). In addition, recent studies of

22 peripheral blood samples from the Japanese population found that several SNPs flanking

23 rs3782151 in the CSAD/lnc-ITGB7-1 region are cis-eQTLs of ITGB7 (Table 3,

24 Supplementary Fig. 8) (30, 31). In contrast, the effect of these SNPs on the expression of

15 Page 17 of 55 Diabetes

1 CSAD is minimal (Table 3). ITGB7 encodes integrin  subunit 7 (ITGB7), which is

2 expressed in leukocytes and forms heterodimers with 4 or E chains. ITGB7 is involved

3 in the migration, entry, and adhesion of lymphocytes in inflamed organs, including the

4 pancreas (44-48). The expression of MAdCAM-1, which is a ligand for the ITGB7

5 4 heterodimer, has been reported to be upregulated in the inflamed pancreas (47, 48).

6 Disease-associated minor alleles increase ITGB7 expression (30, 31), suggesting that the

7 CSAD/lnc-ITGB7-1 region contributes to fulminant type 1 diabetes through an increase

8 in ITGB7 expression and the acceleration of tissue destruction. Antibodies against ITGB7

9 have been explored for the treatment of inflammatory diseases (49, 50). Further studies

10 are necessary to clarify underlying mechanisms of the contribution of ITGB7 expression

11 to the development of fulminant type 1 diabetes.

12 The present study suggested the existence of two distinct loci for type 1 diabetes in the

13 CSAD-ITGB7 region: one in CSAD/lnc-ITGB7-1 for the fulminant subtype and the second

14 in ITGB7 for the autoimmune subtype (Table 2, Supplementary Tables 8-10). It is

15 currently unknown why CSAD/lnc-ITGB7-1 and ITGB7, both of which affect the

16 expression of the same gene, are associated with distinct subtypes of type 1 diabetes.

17 Among the possible explanations are differences in their expression levels, tissue

18 distribution and/or interactions with other susceptibility genes. Although coding variants

19 in CSAD are unlikely to be a common cause of fulminant type 1 diabetes, the possibility

20 that the responsible variant is located within an extended LD region of the CSAD gene

21 cannot be excluded. Further studies are necessary to clarify these hypotheses.

22 In addition to CSAD/lnc-ITGB7-1 on chromosome 12q13.13, several SNPs throughout

23 the genome have also been suggested to be associated with fulminant type 1 diabetes

24 (Supplementary Table 3). With the exception of ITGB7 on chromosome 12q13.13, as

16 Diabetes Page 18 of 55

1 described above, none of the regions that showed some evidence of association with

2 fulminant type 1 diabetes overlapped with previously reported susceptibility loci for type

3 1 diabetes (ImmunoBase: http://www.t1dbase.org/disease/T1D/) (Supplementary Table

4 3B). Further studies are needed to clarify the contribution of these SNPs to fulminant as

5 well as classical type 1 diabetes.

6 In conclusion, we conducted the first genome-wide association study of fulminant type

7 1 diabetes patients and identified CSAD/lnc-ITGB7-1 on chromosome 12q13.13 as the

8 first non-HLA susceptibility locus for fulminant type 1 diabetes. The present study also

9 suggested the possibility that two distinct loci for type 1 diabetes exist in the CSAD-

10 ITGB7 region on chromosome 12q13.13: one in CSAD/lnc-ITGB7-1 for the fulminant

11 subtype, and the second in ITGB7 for the autoimmune subtype. Elucidating the genetic

12 landscape of fulminant type 1 diabetes will provide novel insights into the molecular

13 mechanisms of not only fulminant type 1 diabetes but also type 1 diabetes in general,

14 including type 1 diabetes associated with immune-checkpoint therapy.

15

16 ACKNOWLEDGMENTS

17 We dedicate this article to Dr. Taro Maruyama, who was actively involved in this

18 project but unfortunately passed away before its completion. We thank all of the

19 participants in the project; the Japan Diabetes Society for supporting the committee on

20 type 1 diabetes; Koichiro Higasa and Fumihiko Matsuda at the Center for Genetic

21 Medicine, Kyoto University Graduate School of Medicine, for providing detailed eQTL

22 data for the Japanese population; Shinsuke Noso for the support and helpful discussions

23 provided throughout the project; and S. Hayase and M. Shiota for their technical

24 assistance.

17 Page 19 of 55 Diabetes

1

2 Funding: This study was supported by a grant from the National Institute of Biomedical

3 Innovation; a grant from the Leading Project of Ministry of Education, Culture, Sports,

4 Science and Technology, Japan; a grant from the Japan Society for the Promotion of

5 Science (JSPS); Grants-in-Aid for Scientific Research (KAKENNHI); grants from the

6 Japan Science and Technology Agency (JST); a Grant for Research on Intractable Disease

7 from the Ministry of Health, Labor and Welfare of Japan; and grants from the National

8 Center for Global Health and Medicine of Japan.

9

10 Conflict of interest: The authors declare no conflict of interest.

11

12 Author contributions: Y.K., N.N., T.A., E.K., K.Y, K.T., and H.I. conducted the data

13 analyses. N.N. and K.T. conducted the genotyping and data quality control. Y.K., T.A.,

14 E.K., A.I., A.S., H.O., S.T., K.T., M.N., H.Y., Y.U., H.K., H.M., T.K., and T.H. collected

15 the samples and discussed the results. H.I drafted the manuscript. T.H. and T.K. managed

16 and organized the consortium on the committee on type 1 diabetes of the Japan Diabetes

17 Society. H.I. is the guarantor of this work and, as such, had full access to all the data in

18 the study and takes responsibility for the integrity of the data and the accuracy of the data

19 analysis.

20 21

18 Diabetes Page 20 of 55

1 References 2 3 1. American Diabetes Association. Diagnosis and classification of diabetes mellitus. 4 Diabetes Care 2014; 37:S81-S89 5 2. Alberti KG, Zimmet PZ. Definition, diagnosis and classification of diabetes 6 mellitus and its complications. Part 1: diagnosis and classification of diabetes 7 mellitus provisional report of a WHO consultation. Diabet Med 1998;15:539–553 8 3. Imagawa A, Hanafusa T, Miyagawa J, Matsuzawa, Y. A novel subtype of type 1 9 diabetes mellitus characterized by a rapid onset and an absence of diabetes-related 10 antibodies. N Engl J Med 2000;342:301-307 11 4. Hanafusa T, Imagawa A. Fulminant type 1 diabetes: a novel clinical entity 12 requiring special attention by all medical practioners. Nat Clin Pract Endocrinol 13 Metab 2007;3:36-45 14 5. Imagawa A, Hanafusa T. Fulminant type 1 diabetes – an important subtype in East 15 Asia. Diabetes Metab Res Rev 2011;27:959-964 16 6. Sayama K, Imagawa A, Okita K, et al. Pancreatic beta and alpha cells are both 17 decreased in patients with fulminant type 1 diabetes: a morphometrical assessment. 18 Diabetologia 2005;48:1560-1564 19 7. Shibasaki S, Imagawa A, Tauriainen S, et al. Expression of toll-like receptors in the 20 pancreas of recent-onset fulminant type 1 diabetes. Endocr J 2010;57:211-219 21 8. Karvonen M, Viik-Kajander M, Moltchanova E, Libman I, LaPorte R, Tuomilehto 22 J. Incidence of childhood type 1 diabetes worldwide. Diabetes Care 2000;23:1516- 23 1526 24 9. Moreau C, Drui D, Arnault-Ouary G, Charbonnel B, Chaillous L, Cariou B. 25 Fulminant type 1 diabetes in Caucasians: a report of three cases. Diabetes Metab 26 2008;34:529-532 27 10. Gaudy C, Clévy C, Monestier S, et al. Anti-PD1 pembrolizumab can induce 28 exceptional fulminant type 1 diabetes. Diabetes Care 2015;38:e182-183 29 11. Ikegami H, Kawabata Y, Noso S. Immune checkpoint therapy and type 1 diabetes. 30 Diabetol Int 2016;7:221-227 31 12. Tanaka S, Nishida Y, Aida K, et al. Enterovirus infection, CXC chemokine ligand 32 10 (CXCL10), and CXCR3 circuit: a mechanism of accelerated beta-cell failure in 33 fulminant type 1 diabetes. Diabetes 2009;58:2285-2291. 34 13. Kawabata Y, Ikegami H, Awata T, et al. Committee on Type 1 Diabetes, Japan 35 Diabetes Society. Differential association of HLA with three subtypes of type 1

19 Page 21 of 55 Diabetes

1 diabetes: fulminant, slowly progressive acute-onset. Diabetologia 2009;52:2513- 2 2521 3 14. Imagawa A, Hanafusa T, Uchigata Y, et al. Different contribution of class II HLA 4 in fulminant and typical acute-onset type 1 diabetes mellitus. Diabetologia 5 2005;48:294-300 6 15. Tsutsumi C, Imagawa A, Ikegami H, et al. Class II genotype in fulminant type 1 7 diabetes: a nationwide survey with reference to glutamic acid decarboxylase 8 antibodies. J Diabetes Investig 2012;3:62-69 9 16. Imagawa A, Hanafusa T, Awata T, et al. Report on the committee of the Japan 10 Diabetes Society on the Research of fulminant and acute-onset type 1 diabetes 11 mellitus: new diagnostic criteria of fulminant type 1 diabetes mellitus (2012). J 12 Diabetes Investig 2012;3:536-539 13 17. Nishida N, Koike A, Tajima A, et al. Evaluating the performance of Affymetrix 14 SNP Array 6.0 platform with 400 Japanese individuals. BMC Genomics 2008;9:431 15 18. Ueta M, Sawai H, Sotozono C, et al. IKZF1, a new susceptibility gene for cold 16 medicine–related Stevens-Johnson syndrome/toxic epidermal necrolysis with 17 severe mucosal involvement. J Allergy Clin Immunol 2015;135:1538-1545 18 19. Kawasaki E, Maruyama T, Imagawa A, et al. Diagnostic criteria for acute-onset 19 type 1 diabetes mellitus (2012): Report of the Committee of Japan Diabetes Society 20 on the Research of Fulminant and Acute-onset Type 1 Diabetes Mellitus. J 21 Diabetes Investig 2014;5: 115–118 22 20. Devlin and Roeder. Genomic control for association studies. Biometrics 23 1999;55:997-1004 24 21. International HapMap 3 Consortium. Integrating common and rare genetic variation 25 in diverse human populations. Nature 2010;467:52-8 26 22. Stephens M, Scheet P. Accounting for decay of linkage disequilibrium in haplotype 27 inference and missing-data imputation. Am J Hum Genet 2005;76:449-462 28 23. Stephens M, Smith NJ, Donnelly P. A new statistical method for haplotype 29 reconstruction from population data. Am J Hum Genet 2001;68:978-989 30 24. Howie BN, Donnelly P, Marchini J. A flexible and accurate genotype imputation 31 method for the next generation of genome-wide association studies. PLoS Genetics 32 2009;5: e1000529 33 25. Barrett JC, Fry B, Maller J, Daly MJ. Haploview: analysis and visualization of LD 34 and haplotype maps. Bioinformatic. 2005;21:263-265 35 26. Shimizu I, Makino H, Imagawa A, et al. Clinical characteristics of fulminant type 1 36 diabetes associated with pregnancy. J Clin Endocrinol Metab 2006;91:471-476

20 Diabetes Page 22 of 55

1 27. Evangelou M, Smyth DJ, Fortune MD, et al. A method for gene-based pathway 2 analysis using genomewide association study summary statistics reveals nine new 3 type 1 diabetes associations. Genetic Epidemiol 2014;38:661–670 4 28. Volders PJ, Verheggen K, Menschaert G, et al. An update on LNCipedia: a 5 database for annotated human lncRNA sequences. Nucleic Acids Res. 6 2015;43:D174-180 7 29. Westra H-J, Peters MJ, Esko T, et al. Systematic identification of trans eQTLs as 8 putative drivers of known disease associations. Nat Genet 2013;45:1238-1243 9 30. Narahara M, Higasa K, Nakamura S, et al. Large-scale East Asian eQTL mapping 10 reveals novel candidate genes for LD mapping and the genomic landscape of 11 transcriptional effects of sequence variants. PLoS One 2014;9:e100924 12 31. Higasa K, Miyake N, Yoshimura J, et al. Human genetic variation database, a 13 reference database of genetic variations in the Japanese population. J Hum Genet 14 2016;61:547-553 15 32. Sirdah MM. Protective and therapeutic effectiveness of taurine in diabetes mellitus: 16 A rationale for antioxidant supplementation. Diab Metab Syndr Clin Res Rev 17 2015;9:55–64 18 33. Lambert IH, Kristensen DW, Holm JB, Mortensen OH. Physiological role of 19 taurine - from organism to organelle. Acta Physiologica 2014;213:191-212 20 34. Huxtable RJ. Physiological actions of taurine. Physiol Rev 1992;72:101-163 21 35. Marcinkiewicz J, Kontny E. Taurine and inflammatory disease. Amino Acids 22 2014;46:7-20 23 36. Kim C, Cha Y-N. Taurine chloramine produced from taurine underinflammation 24 provides anti-inflammatory and cytoprotective effects. Amino Acids 2014;46:89- 25 100 26 37. Arany E, Strutt B, Romanus P, Remacle C, Reusens B, Hill DJ. Taurine supplement 27 in early life altered islet morphology, decreased insulitis and delayed the onset of 28 diabetes in non-obese diabetic mice. Diabetologia 2004;47:1831-1837 29 38. Lin S , Yang J, Wu G, et al. Inhibitory effects of taurine on STZ-induced apoptosis 30 of pancreatic islet cells. Adv Exp Med Biol 2013;775:287-297 31 39. Lek M, Karczewski K, Minikel EV, et al. Analysis of protein-coding genetic 32 variation in 60706 humans. Nature 2016;536:285-291 33 40. Nagasaki M, Yasuda J, Katsuoka F, et al. Rare variant discovery by deep whole- 34 genome sequencing of 1,070 Japanese individuals. Nat Commun 2015;6:8018 35 41. Atianand MK, Caffrey DR, Fitzgerald KA. Immunobiology of long noncoding 36 RNAs. Annu Rev Immunol 2017;35:177-198

21 Page 23 of 55 Diabetes

1 42. Chen YG, Satpathy AT, Chang HY. Gene regulation in the immune system by long 2 noncoding RNAs. Nat Immunol 2017;18:962-972 3 43. Zemmour D, Pratama A, Loughhead SM, Mathis D, Benoist C. Flicr, a long 4 nonocoding RNA, modulates Foxp3 expression and autoimmunity. Proc Natl Acad 5 Sci USA 2017;114:E3472-E3480 6 44. Berlin C, Berg EL, Briskin MJ, et al. 4 integrin mediates lymphocyte binding to 7 the mucosal vascular addressin MAdCAM-1. Cell 1993;74:185-195 8 45. Streeter PR, Berg EL, Rouse BT, et al. A tissue-specific endothelial cell molecule 9 involved in lymphocyte homing. Nature 1988;331:41-46 10 46. Hanninen A, Taylor C, Streeter PR, et al. Vascular addressins are induced on islet 11 vessels during insulitis in nonobese diabetic mice and are involved in lymphoid cell 12 binding to islet endothelium. J Clin Invest 1993;92:2509-2515 13 47. Faveeuw C, Gagnerault MC, Lepault F, et al. Expression of homing and adhesion 14 molecules in infiltrated islets of Langerhans and salivary glands of nonobese 15 diabetic mice. J Immunol 1994;152:5969-5978 16 48. Yang XD, Sytwu HK, McDevitt HO, Michie SA. Involvement of beta 7 integrin 17 and mucosal addressin cell adhesion molecule-1 (MAdCAM-1) in the development 18 of diabetes in obese diabetic mice. Diabetes 1997;46:1542-1547 19 49. Ley K, Rivera-Nieves J, Sandborn WJ, Shattil S. Integrin-based therapeutics: 20 biological basis, clinical use and new drugs. Nat Rev Drug Discov 2016;15:173-183 21 50. Neutath MF. Current and emerging therapeutic targets for IBD. Nat Rev 22 Gastroenterol Hepatol 2017;14:269-278 23 24

22 Diabetes Page 24 of 55

Table 1. Association of SNPs on chr12q13.13 with fulminant type 1 diabetes.

Risk/ * § Gene(s) or RAF RAF Imputation Chr. Position SNP reference OR (95%CI) † P nearby gene(s) allele Case Control P OR 12q13.13 53129030 rs7965802 G/A SOAT2 0.471 0.35 1.65 (1.32-2.07) 1.02 x 10-5 - - 12q13.13 53151908 rs11170445 T/C CSAD 0.447 0.294 1.94 (1.55-2.44) 1.06 x 10-8 7.58 x 10-9 1.96 12q13.13 53158691 rs12161793 G/A CSAD 0.272 0.163 1.93 (1.47-2.52) 1.22 x 10-6 1.00 x 10-6 1.94 12q13.13 53158877 rs3782151 A/C CSAD 0.451 0.294 1.97 (1.57-2.48) 4.60 x 10-9 8.82 x 10-9 1.95 12q13.13 53159448 rs2272305 C/T CSAD 0.273 0.162 1.94 (1.49-2.54) 9.28 x 10-7 1.00 x 10-6 1.94 12q13.13 53160499 rs2272306 A/G CSAD 0.175 0.129 1.43 (1.06-1.95) 0.020 0.039 1.38 12q13.13 53180119 rs2293429 C/A CSAD/ZNF740 0.273 0.165 1.90 (1.45-2.48) 2.11 x 10-6 9.84 x 10-7 1.93 12q13.13 53181775 rs3814777 G/A CSAD/ZNF740 0.276 0.164 1.94 (1.49-2.53) 8.24 x 10-7 9.84 x 10-7 1.93 12q13.13 53187018 rs4606556 T/G ZNF740 0.275 0.163 1.94 (1.48-2.53) 9.95 x 10-7 - - 12q13.13 53188711 rs2272298 G/A ZNF740 0.06 0.056 1.08 (0.68-1.72) 0.754 0.337 0.71 12q13.13 53193038 rs2272299 A/G ITGB7 0.245 0.149 1.85 (1.40-2.44) 1.11 x 10-5 - - 12q13.13 53193684 rs2272300 G/T ITGB7 0.245 0.15 1.85 (1.40-2.44) 1.22 x 10-5 9.62 x 10-6 1.86 12q13.13 53194568 rs2272301 G/C ITGB7 0.18 0.134 1.42 (1.05-1.92) 0.022 0.016 1.47 12q13.13 53200147 rs3825084 C/A ITGB7 0.202 0.162 1.31 (0.99-1.74) 0.061 0.078 1.31 12q13.13 53200476 rs3741435 T/C ITGB7 0.246 0.16 1.71 (1.31-2.25) 9.76 x 10-5 1.99 x 10-5 1.82 12q13.13 53209954 rs1554753 G/A RARG 0.453 0.389 1.30 (1.04-1.63) 0.020 0.220 1.29 * RAF, risk allele frequency; † OR, odds ratio for risk allele; CI, confidence interval. The SNPs with genome-wide significant associations are shown in bold. § Genotype imputation using the 1000 Genomes Project data was performed using IMPUTE2 software with the default parameters.

23 Page 25 of 55 Diabetes

Table 2. Association of SNPs in the CSAD and ITGB7 regions with different type 1 diabetes subtypes (fulminant [F] and autoimmune [A]).

Control (C) Type 1 diabetes P OR (95%CI) *

Fulminant (F) Autoimmune (A) F vs. C A vs. C F vs. A F vs. C A vs. C F vs. A

SNP (gene) n frequency n frequency n frequency

rs3782151 genotype CC 210 0.502 73 0.284 174 0.424

(CSAD) CA 170 0.407 136 0.529 183 0.458

AA 38 0.091 48 0.187 53 0.129 2.72 x 10-8 0.044 8.59 x 10-4

2n frequency 2n frequency 2n frequency

allele C 590 0.706 282 0.549 531 0.648

A 246 0.294 232 0.451 289 0.352 4.60 x 10-9 0.011 3.13x 10-4 1.97 (1.57-2.48) 1.31 (1.06-1.60) 1.51 (1.21-1.89)

n frequency n frequency n frequency

rs2272299 genotype GG 312 0.745 145 0.569 248 0.606

(ITGB7) GA 89 0.212 95 0.373 140 0.342

AA 18 0.043 15 0.059 21 0.051 1.04 x 10-5 8.33 x 10-5 0.623

2n frequency 2n frequency 2n frequency

allele G 713 0.851 385 0.755 636 0.778

A 125 0.149 125 0.245 182 0.222 1.11 x 10-5 1.23 x 10-4 0.342 1.85 (1.40-2.44) 1.63 (1.27-2.10) 1.13 (0.87-1.47)

* OR, odds ratio for risk allele; CI, confidence interval.

24 Diabetes Page 26 of 55

Table 3. SNPs in the CSAD region on chr12q13.13 as cis-eQTLs of ITGB7 and CSAD.

Risk/reference Gene or Position* SNP Association with the disease P for eQTLs P for eQTLs P for eQTLs allele nearby gene OR (95%CI) † P ITGB7‡ ITGB7§ CSAD§ 53129030 rs7965802 G/A SOAT2 1.65 (1.32-2.07) 1.02 x 10-5 3.96 x 10-15 - - 53151908 rs11170445 T/C CSAD 1.94 (1.55-2.44) 1.06 x 10-8 5.05 x 10-10 1.48 x 10-5 0.021 53158691 rs12161793 G/A CSAD 1.93 (1.47-2.52) 7.30 x 10-6 4.00 x 10-93 1.89 x 10-10 0.016 53158877 rs3782151 A/C CSAD 1.97 (1.60-2.48) 4.60 x 10-9 5.28 x 10-10 - - 53159448 rs2272305 C/T CSAD 1.94 (1.49-2.54) 9.28 x 10-7 - - - 53180119 rs2293429 C/A CSAD/ZNF740 1.90 (1.45-2.48) 2.11 x 10-6 2.43 x 10-93 2.03 x 10-10 0.016 53181775 rs3814777 G/A CSAD/ZNF740 1.94 (1.49-2.53) 8.24 x 10-7 2.11 x 10-95 2.03 x 10-10 0.016 53187018 rs4606556 T/G ZNF740 1.94 (1.48-2.53) 9.95 x 10-7 3.74 x 10-99 2.03 x 10-10 0.016 53193038 rs2272299 A/G ITGB7 1.85 (1.40-2.44) 1.11 x 10-5 - 1.46 x 10-10 0.054 53193684 rs2272300 G/T ITGB7 1.85 (1.40-2.44) 1.22 x 10-5 1.06 x 10-114 1.46 x 10-10 0.054

* Genome Reference Consortium Human Build 38 patch release 7 (GRCh38.p7), annotation release 108. † OR, odds ratio for risk allele; CI, confidence interval. ‡ Data for populations of European descent obtained from the Blood eQTL browser (http://www.genenetwork.nl/bloodeqtlbrowser/). § Data for the Japanese population obtained from the Human Genetic Variation Database (http://www.hgvd.genome.med.kyoto-u.ac.jp/); the data included in the database were initially obtained from peripheral blood cells. A dash indicates that no data were available for these SNPs.

25 Page 27 of 55 Diabetes

Figure Legends

Figure 1. Manhattan plot presenting the P values across the genome.

The -log10 P values from 426,851 SNPs in 257 fulminant type 1 diabetes cases and 419 controls plotted according to their physical positions on successive chromosomes. An SNP associated with fulminant type 1 diabetes with genome-wide significance outside the HLA is highlighted.

Figure 2. Linkage disequilibrium structure for the risk locus for fulminant type 1 diabetes in an approximately 80-kb genetic region on chromosome 12q13.13.

The top panel (A) shows the –log10 (P values) for 16 tagged SNPs compared between 257 fulminant type 1 diabetes patients and 419 healthy individuals. The second panel (B) shows the genomic location of the RefSeq genes and their intron and exon structures (NCBI [National Center for Biotechnology Information]). The third panel shows a Haploview plot of the linkage disequilibrium (LD) between markers measured based on r2 (C) and D’ (D) in the 419 healthy individuals and LD blocks, which were defined using Gabriel’s algorithm. Two SNPs associated with fulminant type 1 diabetes with genome- wide significance are highlighted.

26 Figure1 - log10 (P value) 10 15 20 0 5

1

2

3

4

5

6

7

Diabetes 8

9

10 11 12 rs11170445 13 14 15 16

17 18 19 20 21 22 Page 28of55 PageFigure 29 of 55 2 Diabetes

A 10 9 rs3782151 rs11170445 8

7 6 5 4

- log10 (P value) (P log10 - 3 2 1 0 53130K 53140K 53150K 53160K 53170K 53180K 53190K 53200K 53210K B

C

r2

D

D’ Diabetes Page 30 of 55

Supplementary Table 1. Characteristics of the study participants.

Cases Controls fulminant autoimmune GWAS stage Candidate gene N 257 409 419 357 Sex male/female 154/103 168/241 350/69 223/134 Disease onset (years) mean 43.5 20.5 range 1-92 4-34 Page 31 of 55 Diabetes

Supplementary Table 2. SNP filtering for statistical analysis

QC parameter Threshold numbers of SNPs excluded (Autosomal chromosomes) SNP call rate < 0.95 9157 Minor allele frequency (MAF) < 0.05 143057 Hardy-Weinberg equilibrium P value in controls < 0.001 2822 Poor clustering 218 Diabetes Page 32 of 55

Supplementary Table 3. A. SNPs in non-HLA regions showing a suggestive association with fulminant type 1 diabetes in the genome-wide association study (P ≤ 1.0 x 10-4). SNPs with P ≤ 1.0 x 10-5are shown in bold.

Risk/ ‡ Gene or 95%CI Chr Cytoband Position* SNP Reference P OR† nearby gene allele lower upper 1 p36.32 4820420 rs10915616 C/T NPHP4 1.36 x 10-5 2.39 1.60 3.58 1 q42.13 226915712 rs12041000 G/A ADCK3 5.76 x 10-5 1.58 1.26 1.98 2 p25.3 3881302 rs7591948 A/C DCDC2C 4.08 x 10-5 2.41 1.57 3.72 2 p25.3 3906092 rs4850005 A/G LINC01304 2.42 x 10-5 2.49 1.61 3.85 2 p22.1 40979035 rs6544376 T/C LOC388942 9.38 x 10-5 1.62 1.27 2.06 2 p16.2 53653800 rs6727876 T/G ASB3 2.42 x 10-5 1.61 1.29 2.01 2 p16.2 53699166 rs13414216 T/C LOC100302652 7.69 x 10-5 1.57 1.25 1.96 2 p16.2 53791787 rs2542588 C/T LOC100302652 9.33 x 10-5 1.56 1.25 1.94 2 p16.1 55166432 rs10172326 A/G C2orf63 5.87 x 10-5 1.58 1.26 1.98 2 q11.1 95368615 rs2320432 G/A KCNIP3 1.62 x 10-5 1.78 1.37 2.31 2 q34 212714053 rs13393014 G/A IKZF2 3.80 x 10-5 1.61 1.28 2.02 2 q37.3 238930543 rs11680343 G/A HDAC4 3.96 x 10-5 1.89 1.39 2.58 3 p26.1 5093886 rs3864055 T/C ARL8B 5.93 x 10-6 1.79 1.39 2.31 3 p14.3 55497931 rs1160047 A/G ERC2 3.66 x 10-5 1.73 1.33 2.24 3 q21.3 127431182 rs4637240 T/C TPRA1 4.34 x 10-5 1.88 1.38 2.55 3 q21.3 127441926 rs3948104 G/C TPRA1 1.89 x 10-5 1.93 1.42 2.63 3 q21.3 127474584 rs2594220 T/G TPRA1 8.16 x 10-5 1.83 1.35 2.47 4 p15.2 27495853 rs905279 T/C MIR4275 1.44 x 10-5 1.64 1.31 2.05 4 q31.22 147527718 rs4639051 A/G EDNRA 4.37 x 10-5 2.44 1.57 3.80 4 q34.3 181408638 rs6857563 G/T MGC45800 2.16 x 10-5 1.62 1.30 2.02 4 q35.1 184755529 rs12644905 C/T ACSL1 3.60 x 10-6 1.91 1.45 2.52 5 p14.3 21839979 rs16888321 C/T CDH12 5.48 x 10-5 2.15 1.47 3.15 5 p14.3 21863845 rs8180523 G/A CDH12 4.34 x 10-5 2.19 1.49 3.21 5 q11.2 52890164 rs1904163 T/C ITGA1 7.51 x 10-5 1.61 1.27 2.04 5 q11.2 57044912 rs2063260 C/T LOC100130001 4.04 x 10-5 2.08 1.46 2.97 6 p22.3 17120009 rs12203596 C/A FLJ23152 5.90 x 10-6 2.84 1.78 4.53 6 p22.3 24114366 rs13215261 A/G NRSN1 3.45 x 10-5 2.15 1.49 3.11 6 p22.2 26177054 rs12664415 T/C HIST1H2BE 1.96 x 10-5 2.70 1.69 4.33 6 p22.1 29644108 rs3131854 C/T MOG 7.34 x 10-5 1.77 1.33 2.36 6 p21.31 33764485 rs6904991 G/T LEMD2 4.71 x 10-5 1.72 1.32 2.23 6 q16.3 103607414 rs9404439 G/A HACE1 1.49 x 10-5 1.81 1.38 2.38 6 q21 105256293 rs1189924 T/G PREP 6.45 x 10-5 1.58 1.26 1.97 6 q22.1 117303435 rs210966 G/C ROS1 1.97 x 10-5 1.75 1.35 2.28 6 q26 163922743 rs9356171 A/G C6orf118 9.19 x 10-6 1.90 1.43 2.52 7 p21.2 13877714 rs17167612 C/T ETV1 6.50 x 10-5 2.34 1.53 3.59 7 q22.1 101472937 rs1008064 C/T EMID2 7.45 x 10-5 1.56 1.25 1.95 7 q31.1 109189772 rs10953597 G/A EIF3IP1 6.19 x 10-5 1.71 1.31 2.23 Page 33 of 55 Diabetes

7 q36.2 154439470 rs6965651 G/T DPP6 3.51 x 10-5 1.60 1.28 2.01 8 p23.2 3477514 rs13250793 A/C CSMD1 5.43 x 10-5 1.69 1.31 2.18 8 p23.2 3477767 rs13248265 G/T CSMD1 6.46 x 10-5 1.68 1.30 2.16 8 p23.2 4627847 rs17070883 C/A CSMD1 5.05 x 10-5 1.88 1.38 2.56 8 q11.21 51042700 rs16915742 A/G PXDNL 3.66 x 10-5 1.73 1.33 2.24 8 q11.23 53517084 rs13279206 G/A ATP6V1H 4.30 x 10-5 1.63 1.29 2.06 8 q11.23 53523733 rs13270667 G/A ATP6V1H 6.18 x 10-5 1.62 1.28 2.04 8 q22.2 98143304 rs16896685 A/G POP1 5.29 x 10-5 1.71 1.32 2.22 8 q22.2 98150406 rs7000738 C/T POP1 3.86 x 10-5 1.70 1.32 2.18 8 q24.3 140036525 rs7011932 T/C TRAPPC9 9.33 x 10-5 1.56 1.25 1.94 9 p24.1 7762216 rs6477230 G/A C9orf123 3.60 x 10-5 2.52 1.61 3.97 9 p21.3 23166744 rs7871386 C/T ELAVL2 4.60 x 10-6 2.54 1.69 3.83 10 p14 8596272 rs4747806 C/T SFTA1P 5.54 x 10-5 2.35 1.53 3.59 10 q24.1 96325581 rs3789940 C/T DNTT 5.83 x 10-5 1.59 1.27 1.99 11 p15.1 20627519 rs10833369 G/T SLC6A5 3.54 x 10-5 1.65 1.30 2.10 11 p14.1 27778324 rs5790666 -/G KIF18A 8.07 x 10-7 2.32 1.65 3.26 11 p12 41237630 rs729210 T/C LOC399881 3.23 x 10-5 1.62 1.29 2.04 11 p12 41267500 rs4445616 G/T LOC399881 4.39 x 10-5 1.65 1.29 2.09 11 p12 41267818 rs10837603 A/G LOC399881 9.74 x 10-5 1.56 1.25 1.95 12 p13.1 14189684 rs10845905 A/G ATF7IP 1.99 x 10-5 1.63 1.30 2.05 12 q13.13 53151908 rs11170445 T/C CSAD 7.58 x 10-9 1.96 1.56 2.46 12 q13.13 53187018 rs4606556 T/G ZNF740 9.84 x 10-7 1.93 1.48 2.53 12 q13.13 53193038 rs2272299 A/G ITGB7 9.62 x 10-6 1.86 1.41 2.45 12 q15 70516944 rs12582352 T/C PTPRB 3.74 x 10-6 2.58 1.71 3.90 12 q15 70543663 rs919594 C/T PTPRB 1.46 x 10-5 2.42 1.61 3.64 12 q22 93616192 rs11107130 T/C CRADD 2.12 x 10-5 1.66 1.31 2.10 12 q22 93622966 rs7300268 A/C CRADD 4.50 x 10-5 1.62 1.28 2.04 12 q22 93626791 rs7968701 T/C CRADD 3.33 x 10-5 1.61 1.28 2.02 16 p12.1 26144398 rs11866894 G/A C16orf82 9.14 x 10-5 1.72 1.31 2.27 16 p12.1 26147341 rs12598369 A/G C16orf82 6.16 x 10-5 1.75 1.33 2.30 16 q12.2 55697756 rs36010 A/C SLC6A2 5.50 x 10-5 1.67 1.30 2.15 20 q13.31 57723635 rs6025742 T/C C20orf85 9.03 x 10-5 1.56 1.25 1.94 22 q11.21 20814301 rs178050 A/G PI4KA 7.37 x 10-6 1.66 1.33 2.07 22 q11.21 20899436 rs178085 A/G CRKL 1.92 x 10-5 1.62 1.30 2.03

* Nucleotide positions based on GRCh38.p7. † OR, odds ratio; ‡ CI, confidence interval. Diabetes Page 34 of 55

B. SNPs in non-HLA region showing a suggestive association (P ≤ 1.0 x 10-5) with fulminant type 1 diabetes in the genome-wide association study relative to the location of known susceptibility loci for autoimmune type 1 diabetes.

Gene or Other OR† Type 1 Chr Cytoband Position SNP nearby P autoimmune (95%CI‡) diabetes gene disease 113288123- rs6679677 PHTF1 ATD CRO JIA 1 1p13.2 T1D 114009223 rs2476601 PTPN22 RA SLE AA VIT 200599616- 1 1q32.1 rs6691977 T1D 200896640 206709013- rs3024493 CRO SLE UC 1 1q32.1 IL10 T1D 206867593 rs3024505 IBD 110225473- 2 2q13 rs4849135 T1D 111045908 rs1990760 162104363- PSO SLE UC 2 2q24.2 rs35667974 IFIH1 T1D 162504293 IBD VIT rs2111485 203749263- rs3087243 2 2q33.2 CTLA4 T1D ATD CEL RA 203951852 rs11571316 24795892- 2 2p23.3 rs478222 EFR3B T1D 25264789 99926678- 2 2q11.2 rs9653442 AFF3 T1D RA 100432653 1.79 3 3p26.1 5093886 rs3864055 ARL8B 5.93 x 10-6 FT1D (1.39-2.31) 46061249- 3 3p21.31 rs113010081 CCR5 T1D CEL UC 46593299 26026175- 4 4p15.2 rs10517086 T1D 26130530 rs17388568 IL21 122017479- rs4505848 4 4q27 ADAD1 T1D CEL CRO UC 122644147 rs75793288 IL2 rs6827756 165571673- 4 4q32.3 rs2611215 T1D 165682849 1.91 4 4q35.1 184755529 rs12644905 ACSL1 3.60 x 10-6 FT1D (1.45-2.52) 35798580- T1D 5 5p13.2 rs11954020 IL7R 36036080 T1DJ* 2.84 6 6p22.3 17120009 rs12203596 FLJ23152 5.90 x 10-6 FT1D (1.78-4.53) rs11755527 90097116- 6 6q15 rs597325 BACH2 T1D ATD MS RA 90320436 rs72928038 rs9375435 126129075- 6 6q22.32 rs9388489 CENPW T1D 127098552 rs1538171 137568592- 6 6q23.3 rs6920220 TNFAIP3 T1D RA SLE UC IBD 137806067 158897742- 6 6q25.3 rs1738074 TAGAP T1D CEL MS 159119181 1.90 6 6q26 163922743 rs9356171 C6orf118 9.19 x 10-6 FT1D (1.43-2.52) 169879562- 6 6q27 rs924043 T1D 170130554 26618343- 7 7p15.2 rs7804356 T1D 27162670 50327041- 7 7p12.2 rs62447205 IKZF1 T1D 50624014 Page 35 of 55 Diabetes

50833203- 7 7p12.1 rs4948088 COBL T1D 51066332 rs7020673 4238265- 9 9p24.2 rs6476839 GLIS3 T1D 4315442 rs10758593 2.54 9 9p21.3 23166744 rs7871386 ELAVL2 4.60 x 10-6 FT1D (1.69-3.83) rs2104286 rs61839660 5988280- rs7090530 RBM17 T1D 10 10p15.1 MS RA 6146375 rs10795791 IL2RA T1DJ* rs12251307 rs41295121 6388017- 10 10p15.1 rs11258747 PRKCQ T1D 6503142 32983821- 10 10p11.22 rs722988 NRP1 T1D 33222119 88245291- rs12416116 10 10q23.31 RNLS T1D 88511262 rs10509540 2026410- 11 11p15.5 rs7928968 INS T1D 2051274 rs72853903 2092701- T1D 11 11p15.5 rs7111341 INS 2260001 T1DJ rs689 2.32 11 11p14.1 27778324 rs5790666 KIF18A 8.07 x 10-7 FT1D (1.65-3.26) 64076473- 11 11q13.1 rs694739 BAD T1D CRO MS AA 64456055 9366576- rs4763879 12 12p13.31 CD69 T1D 9820167 rs10492166 1.96 12 12q13.13 53151908 rs11170445 CSAD 7.58 x 10-9 FT1D (1.56-2.46) CSAD/ 1.93 12 12q13.13 53187018 rs4606556 9.84 x 10-7 FT1D ZNF740 (1.48-2.53) 1.86 12 12q13.13 53193038 rs2272299 ITGB7 9.62 x 10-6 FT1D (1.41-2.45) 53058065- 12 12q13.13 rs11170466 ITGB7 T1D 53209207 rs11171739 rs705704 IKZF4 55957562- T1D 12 12q13.2 rs2292239 DGKA AA 56404651 T1DJ rs11171710 ERBB3 rs705705 2.58 12 12q15 70516944 rs12582352 PTPRB 3.74 x 10-6 FT1D (1.71-3.90) rs3184504 CEL CRO JIA 111278572- NAA25 12 12q24.13 rs653178 T1D PBC RA AA 112592683 SH2B3 rs17696736 PSC VIT 99240634- 13 13q32.3 rs9585056 GPR183 T1D 99534324 68071518- 14 14q24.1 rs911263 T1D PBC 68485527 68696738- 14 14q24.1 rs1465788 T1D 68851345 97895009- rs4900384 14 14q32.2 T1D 98138364 rs1456988 100817324- rs941576 14 14q32.2 DLK1 T1D 100862402 rs56994090 38522176- rs12908309 15 15q14 RASGRP1 T1D CRO 38701912 rs72727394 rs3825932 78709357- 15 15q25.1 rs12148472 CTSH T1D CEL 78968794 rs34593439 Diabetes Page 36 of 55

rs12927355 10923201- DEXI T1D 16 16p13.13 rs193778 MS PBC 11372654 CLEC16A T1DJ* rs12708716 rs4788084 28283985- 16 16p11.2 rs9924471 IL27 T1D AS CRO IBD 29014657 rs151234 75182342- rs8056814 16 16q23.1 T1D 75487132 rs7202877 39226421- rs2290400 ORMDL3 17 17q12 T1D CRO UC IBD 40084508 rs12453507 GSDMB 40563547- 17 17q21.2 rs7221109 T1D UC 40722575 45903164- 17 17q21.31 rs1052553 T1D 46788073 12738414- rs2542151 CEL CRO UC 18 18p11.21 PTPN2 T1D 12925254 rs1893217 IBD 69831437- rs1615504 18 18q22.2 CD226 T1D MS 69902924 rs763361 CRO JIA MS 10280033- rs12720356 19 19p13.2 TYK2 T1D PBC PSO RA 10517872 rs34536443 IBD 46648121- rs402072 19 19q13.32 T1D 46811943 rs425105 48592641- rs516246 20 19q13.33 FUT2 T1D CRO IBD 48763133 rs602662 1508984- rs6043409 20 20p13 T1D 1741341 rs2281808 42389067- rs11203202 21 21q22.3 UBASH3A T1D RA VIT 42458550 rs11203203 1.66 22 22q11.21 20814301 rs178050 PI4KA 7.37 x 10-6 FT1D (1.33-2.07) 29670355- rs5753037 22 22q12.2 T1D 30273198 rs4820830 37171803- C1QTNF6 22 22q12.3 rs229533 T1D 37262764 RAC2 154425779- X Xq28 rs2664170 T1D 155057576

The data for fulminant type 1 diabetes (FT1D) obtained in the present study are shown in bold. The data for autoimmune type 1 diabetes (T1D) were obtained from ImmunoBase: http://www.t1dbase.org/disease/T1D/ The data for autoimmune type 1 diabetes in Japanese (T1DJ) individuals were obtained from the following reference: Diabetes Metab Res Rev 2011;27:844-848. The SNPs associated with T1DJ are underlined. * The associated SNPs in T1DJ differe from those in ImmunoBase: IL7R, rs6897932; IL2RA, rs706778; and CLEC16A, rs2903692. AA: alopecia areata, ATD: autoimmune thyroid disease, CEL: celiac disease, CRO: Crohn’s disease, IBD: inflammatory bowel disease, JIA: juvenile idiopathic arthritis, MS: multiple sclerosis, PBC: primary biliary cirrhosis, PSO: psoriasis, RA: rheumatoid arthritis, SLE: systemic lupus erythematosus, UC: ulcerative colitis, VIT: vitiligo † OR, odds ratio; ‡ CI, confidence interval. Page 37 of 55 Diabetes

Supplementary Table 4. Seven variants identified by genotype imputation that exhibit a stronger association than rs3782151

Gene(s) or nearby Reference Alternative SNP Chr. Position OR P gene(s) allele allele rs7304044 12q13.13 53142385 CSAD G A 1.975 7.09E-09 rs11170439 12q13.13 53147705 CSAD C T 1.952 8.69E-09 rs140878838 12q13.13 53147884:53147885 CSAD C CAT 1.954 7.99E-09 rs7305472 12q13.13 53153307 CSAD C T 1.949 8.82E-09 rs11170448 12q13.13 53154203 CSAD A G 1.949 8.82E-09 rs3782151 12q13.13 53158877 CSAD C A 1.949 8.82E-09 rs10876414 12q13.13 53165245 CSAD A G 1.957 8.12E-09 rs1465056 12q13.13 53171673 CSAD T C 1.974 5.52E-09

Genotype imputation using the 1000 Genomes Project data was performed using IMPUTE2 software with the default parameters. rs3782151, which was genotyped in the present study, is shown in bold. Diabetes Page 38 of 55

Supplementary Table 5. Conditional analysis by controlling for top hit SNPs, rs11170445 and rs3782151.

A. Conditioned by rs3782151

Predictor FvR Model Odds Ratio OR Lower OR Upper P-value* rs2272299 0.605 1.10 0.77 1.56 rs4606556 0.302 1.21 0.84 1.73

B. Conditioned by rs11170445

Predictor FvR Model Odds Ratio OR Lower OR Upper P-value* rs2272299 0.568 1.11 0.78 1.57 rs4606556 0.277 1.22 0.85 1.74

*Numeric regression analysis was carried out to obtain a full-vs-reduced-model (FvR Model) p-value using rs3782151 and rs11170445 as covariates. Page 39 of 55 Diabetes

Supplementary Table 6. Haplotype analysis of 13 SNPs in LD block 1 in the CSAD-ITGB7 region.

Haplotype Fulminant type 1 diabetes (2n=514) Control (2n=838) OR (95%CI) * P* SNP #1-#13 2n frequency 2n frequency ACACTGAAGAGTC 261 0.508 533 0.636 0.59 (0.47-0.74) 3.32 x 10-6 GTGACGCGTAAGC 117 0.228 116 0.138 1.83 (1.38-2.44) 2.49 x 10-5 GTAATAAAGAGTG 58 0.113 61 0.073 1.62 (1.11-2.36) 0.0116 GTAATAAAGGGTG 31 0.060 46 0.055 1.11 (0.69-1.77) 0.6764 GCACTGAAGAGTC 18 0.035 56 0.067 0.51 (0.29-0.87) 0.0126 GTGACGCGTAGTC 15 0.029 12 0.014 2.07 (0.96-4.46) 0.0578 ATGACGCGTAAGC 7 0.014 8 0.010 1.43 (0.52-3.97) 0.4877 others 7 0.014 6 0.010

Haplotypes of SNP #1-13 in LD block shown in Figure 2: #1, rs7965802; #2, rs11170445; #3, rs12161793; #4, rs3782151; #5, rs2272305; #6, rs2272306; #7, rs2293429; #8, rs3814777; #9, rs4606556; #10, rs2272298; #11, rs2272299; #12, rs2272300; and #13, rs2272301. SNPs showing a genome-wide significant association with fulminant type 1 diabetes and their corresponding disease-associated allele are shown in bold. The haplotypes were estimated using the PHASE program (version 2.1). * Chi-squared test in two-by-two cross tables was used to examine the associations between haplotypes and fulminant type 1 diabetes. In this analysis, each haplotype was assumed to be one of the alleles at a biallelic locus, and the other haplotypes were assumed to be the other allele. For example, the first haplotype (ACACTGAAGAGTC) and the other haplotypes were designated “A allele” and “B allele”, respectively. Diabetes Page 40 of 55

Supplementary Table 7. Effect of the HLA DR-DQ haplotypes on the association of rs3782151 in CSAD with fulminant type 1 diabetes. A. Susceptible HLA haplotypes for fulminant type 1 diabetes. The top and middle panels show the association of HLA with the disease, and the bottom panel shows the association of the minor A allele of rs372151 with the disease relative to the HLA genotype.

Control Fulminant type 1 diabetes OR* [95%CI†] P 2n=838 frequency 2n=432 frequency DR-DQ haplotype DRB1*04:05-DQB1*04:01 121 0.144 142 0.329 2.90 [2.20-3.83] 2.82 x 10-14 DRB1*09:01-DQB1*03:03 124 0.148 116 0.269 2.11 [1.59-2.81] 2.01 x 10-7

n=419 frequency n=216 frequency Haplotype combination S/ S‡ 40 0.095 78 0.361 5.36 [3.49-8.22] 8.58 x 10-16 S/ N§ 165 0.394 102 0.472 1.38 [0.99-1.92] 0.058 N/ N 214 0.511 36 0.167 0.19 [0.13-0.29] 8.66 x 10-17

2n frequency 2n frequency A allele of rs372151 Total 246/836 0.294 193/432 0.447 1.94 [1.52-2.46] 6.32 x 10-8 Subjects with S/S 22/80 0.275 70/156 0.449 2.15 [1.20-3.85] 0.010 Subjects with S/N 101/328 0.308 94/204 0.461 1.92 [1.34-2.76] 0.000374 Subjects with N/N 123/428 0.287 29/72 0.403 1.67 [1.00-2.80] 0.049 Heterogeneity 0.916#

* OR, odds ratio; † CI, confidence interval. ‡ S: susceptible haplotypes, DRB1*04:05-DQB1*04:01 or DRB1*09:01-DQB1*03:03 § N: non-susceptible haplotypes, haplotypes other than DRB1*04:05-DQB1*04:01 and DRB1*09:01-DQB1*03:03 # The P value for heterogeneity assessed by using the random-effect (DerSimonian and Laird) method. Page 41 of 55 Diabetes

B. HLA-DR4 and DR9 haplotypes. The upper panel shows the association of HLA genotypes with the disease, and the lower panel shows the association of the minor T allele of rs372151 with the disease relative to HLA genotypes.

Control Fulminant type 1 diabetes OR* [95%CI†] P n=419 frequency n=216 frequency Haplotype combination DR4/DR4 11 0.026 31 0.144 6.22 [3.06-12.63] 1.78 x 10-8 DR9/DR9 12 0.029 21 0.097 3.65 [1.76-7.58] 2.25 x 10-4 DR4/DR9 17 0.041 26 0.120 3.24 [1.71-6.11] 1.50 x 10-4 DR4/others 82 0.196 54 0.250 1.37 [0.93-2.03] NS DR9/others 83 0.198 48 0.222 1.16 [0.77-1.73] NS Others/others 214 0.511 36 0.167 0.19 [0.13-0.29] 8.66 x 10-17

n frequency n frequency A allele of rs372151 Total 246/836 0.294 191/432 0.442 1.90 [1.49-2.42] 1.51 x 10-7 Subjects with DR4/DR4 5/22 0.227 25/62 0.403 2.30 [0.75-7.03] NS Subjects with DR9/DR9 7/24 0.292 20/42 0.476 2.21 [0.76-6.43] NS Subjects with DR4/DR9 10/34 0.294 25/52 0.481 2.22 [0.89-5.56] NS Subjects with DR4/others 46/164 0.280 51/108 0.472 2.30 [1.38-3.82] 0.001 Subjects with DR9/others 55/164 0.335 41/96 0.427 1.48 [0.88-2.48] NS Subjects with others 123/428 0.287 29/72 0.403 1.67 [1.00-2.80] 0.049 Heterogeneity 0.969#

* OR, odds ratio; † CI, confidence interval. DR4: DRB1*04:05-DQB1*04:01, DR9: DRB1*09:01-DQB1*03:03, others: haplotypes other than DR4 and DR9 # The P value for heterogeneity assessed using the random-effect (DerSimonian and Laird) method. Diabetes Page 42 of 55

C. Relationship between a top-hit SNP in the HLA region, rs9268853, and disease-associated HLA haplotypes.

HLA-DR4 haplotype (DRB1*04:05-DQB1*04:01) Control Fulminant type 1 diabetes rs9268853 rs9268853 with C allele without C allele P with C allele without C allele P with DR4 haplotype 146 72 2.12 x 10-22 179 43 5.60 x 10-9 without DR4 haplotype 180 436 114 96

HLA-DR9 haplotype (DRB1*09:01-DQB1*03:03) Control Fulminant type 1 diabetes rs9268853 rs9268853 with C allele without C allele P with C allele without C allele P with DR9 haplotype 154 70 5.85 x 10-26 148 42 7.66 x 10-5 without DR9 haplotype 172 438 145 97

Data are number of subjects with or without C allele of rs9268853 relative to the presence or absence of disease-associated HLA haplotypes, DR4 (DRB1*04:05-DQB1*04:01) and DR9 (DRB1*09:01-DQB1*03:03). Page 43 of 55 Diabetes

Supplementary Table 8. Association of CSAD (rs3782151) and ITGB7 (rs2272299) with classical autoimmune type 1 diabetes in European and Japanese populations.

Population CSAD (rs3782151) ITGB7 (rs2272299) Control Type 1 OR [95%CI] * Control Type 1 OR [95%CI]

autoimmune fulminant autoimmune fulminant autoimmune fulminant autoimmune fulminant European§ 2n=11,826 2n=17,656 2n=11,826 2n=17,656 C 0.80 0.794 G 0.951 0.944 A 0.20 0.206 1.04 [0.98-1.10] A 0.049 0.056 1.15 [1.04-1.28] (P=0.365) (P=0.0063) Japanese 2n =836 2n=820 2n=514 2n=838 2n=818 2n=510 C 0.706 0.648 0.549 G 0.851 0.778 0.755 A 0.294 0.352 0.451 1.31 [1.06-1.60] 1.97 [1.57-2.48] A 0.149 0.222 0.245 1.63 [1.27-2.10] 1.85 [1.40-2.44] (P=0.011) (P=4.60 x 10-9) (P=0.0001) (P=1.11 x 10-5)

Total † A 1.05 [1.00-1.12]‡ A 1.21 [1.10-1.34]‡ P=0.060 P=9.68 x 10-5

The data show the allele frequencies, and data for fulminant type 1 diabetes in the Japanese population are shown for comparison purposes. * OR: odds ratio; CI, confidence interval. † Meta-analysis by using the Mantel-Haenszel method. ‡ Summary odds ratio [95% confidence interval], the P value for heterogeneity was >0.05 for both rs3782151 in CSAD and rs2272299 in ITGB7. §The data were obtained from the Type 1 Diabetes Genetics Consortium (https://repository.niddk.nih.gov/studies/t1dgc/). Diabetes Page 44 of 55

Supplementary Table 9. Frequencies of CSAD-ITGB7 (rs3782151-rs2272299) haplotypes in Japanese participants.

Haplotype Control Fulminant Autoimmune OR P (C) (F) (A) [95%CI] * 2n=836 2n=510 2n=818 F vs. C A vs. C F vs. A F vs. C A vs. C F vs. A A-A 125 125 176 1.85 1.56 1.18 1.22 x 10-5 5.42 x 10-4 NS (0.150) (0.245) (0.215) [1.40-2.44] [1.21-2.01] [0.91-1.54] A-G 121 105 113 1.53 0.95 1.62 0.004 NS 0.001 (0.145) (0.206) (0.138) [1.15-2.04] [0.72-1.25] [1.21-2.17] C-G 590 280 523 0.51 0.74 0.69 5.38 x 10-9 0.004 0.001 (0.706) (0.549) (0.639) [0.40-0.64] [0.60-0.91] [0.55-0.86] C-A 0 0 6 (0) (0) (0.007)

The haplotypes were estimated using the PHASE program (version 2.1). The number of subjects is shown, and the frequency is indicated in parenthesis. * OR, odds ratio; CI, confidence interval. Page 45 of 55 Diabetes

Supplementary Table 10A. Frequencies of CSAD-ITGB7 (rs3782151-rs2272299) haplotypes in the European population.

Haplotype Control Autoimmune type 1 diabetes OR P 2n=450 2n=458 [95%CI] * A-A 11 34 3.20 5.47x 10-4 (0.024) (0.074) [1.60-6.40] A-G 113 86 0.69 0.021 (0.251) (0.188) [0.50-0.95] C-G 326 338 1.07 NS (0.724) (0.738) [0.80-1.44] C-A 0 0 (0) (0.000)

The haplotypes were estimated using the PHASE program (version 2.1). The number of subjects is shown, and the frequency is indicated in parenthesis. * OR, odds ratio; CI, confidence interval. Due to the near-absence of fulminant type 1 diabetes in European populations, only data for autoimmune type 1 diabetes are shown. Samples were from Noso S et al. J Genet Syndr Gene Ther 2013;4:204. Diabetes Page 46 of 55

Supplementary Table 10B. Comparison of CSAD-ITGB7 (rs3782151-rs2272299) haplotype frequencies between Japanese and European populations.

Haplotype Controls OR P Autoimmune type 1 diabetes OR P Japanese European [95%CI] * Japanese European [95%CI] * 2n=836 2n=450 2n=818 2n=458 A-A 125 11 7.02 3.49 x 10-12 176 34 3.42 7.41 x 10-11 (0.150) (0.024) [3.75-13.14] (0.215) (0.074) [2.32-5.03] A-G 121 113 0.50 2.41 x 10-6 113 86 0.69 0.019 (0.145) (0.251) [0.38-0.67] (0.138) (0.188) [0.51-0.94] C-G 590 326 0.94 0.48 523 338 0.63 3.09 x 10-4 (0.706) (0.724) [0.71-1.18] (0.639) (0.738) [0.49-0.81] C-A 0 0 6 0 (0) (0) (0.007) (0)

The haplotypes were estimated using the PHASE program (version 2.1). The number of subjects is shown, and the frequency is indicated in parenthesis. * OR, odds ratio; CI, confidence interval. Due to the near-absence of fulminant type 1 diabetes in European populations, only data for autoimmune type 1 diabetes are shown. Page 47 of 55 Diabetes

Supplementary Table 11. Variants identified by re-sequencing of the CSAD region in 32 individuals with fulminant type 1 diabetes who were homozygous for the disease-associated allele of the top- hit SNP, rs3782151.

variant Gene or Amino Acid MAF MAF MAF update Chr. position* Type Location Ref. Variant dbSNP§ P for disease¶ number nearby gene Change Japanese† European‡ Japanese ǁ 001 chr12 53158691 CSAD SNV intronic A G 0.2163 0.0556 rs12161793 0.197 7.30 x 10-6 002 chr12 53158877 CSAD SNV intronic G T 0.3846 0.2525 rs3782151 0.3407 4.60 x 10-9 003 chr12 53159448 CSAD INDEL intronic T - novel 004 chr12 53159448 CSAD SNV intronic T C 0.2163 0.0556 rs2272305 9.28 x 10-7 005 chr12 53160499 CSAD SNV intronic G A 0.1683 0.1869 rs2272306 0.1341 0.02 006 chr12 53161148 CSAD SNV exonic His288Arg T C 0.0001 007 chr12 53161674 CSAD SNV intronic A T 0.2163 0.0556 rs11170451 008 chr12 53164591 CSAD SNV intronic G A 0.0052 009 chr12 53164687 CSAD SNV intronic C T 0.0001 010 chr12 53165187 CSAD SNV intronic T A novel 011 chr12 53165191 CSAD SNV intronic A T novel 012 chr12 53165195 CSAD SNV intronic G A novel 013 chr12 53165245 CSAD SNV intronic A G 0.3846 0.2525 rs10876414 014 chr12 53166711 CSAD SNV intronic G A 0.0048 0.0000 rs11170452 015 chr12 53169893 CSAD SNV intronic G A novel 016 chr12 53169913 CSAD INDEL intronic - C novel 017 chr12 53170081 CSAD SNV exonic synonymous G A 0.02115 0.0505 rs11170453 0.1988 018 chr12 53170150 CSAD SNV intronic A G 0.0096 0.0000 rs146549996 0.0034 019 chr12 53171195 CSAD INDEL intronic - G novel 020 chr12 53171673 CSAD SNV intronic T C 0.3846 0.2525 rs1465056 0.3404 021 chr12 53171898 CSAD SNV exonic synonymous G A 0.0001 022 chr12 53174848 CSAD SNV intronic A C 0.0001 023 chr12 53175280 CSAD SNV intronic G A 0.0192 0.0000 rs75249111 0.0326 024 chr12 53176875 CSAD SNV intronic G C 0.2163 0.0556 rs11170454 Diabetes Page 48 of 55

025 chr12 53177190 CSAD SNV intronic T C 0.0001 026 chr12 53177469 CSAD SNV intronic A T novel 027 chr12 53178076 CSAD SNV intronic A C 0.2163 0.0556 rs12580404 0.1997 028 chr12 53178346 CSAD SNV intronic T G 0.2163 0.0556 rs10876417 029 chr12 53180119 CSAD,ZNF740 SNV intronic, upstream A C 0.2163 0.0556 rs2293429 0.2023 2.11 x 10-6 030 chr12 53180260 CSAD,ZNF740 SNV intronic, upstream G C 0.2115 0.0556 rs2293430 031 chr12 53180261 CSAD,ZNF740 SNV intronic, upstream A C 0.2115 0.0556 rs2293431 032 chr12 53180484 CSAD,ZNF740 SNV intronic, upstream C T novel 033 chr12 53180582 CSAD,ZNF740 SNV intronic, upstream C T 0.2163 0.0556 rs34628172 0.2014 034 chr12 53180706 CSAD,ZNF740 SNV intronic, upstream G A 0.1635 0.1818 rs3108403 0.1341 035 chr12 53180708 CSAD,ZNF740 SNV intronic, upstream C A 0.0385 0.0000 rs138922341 0.0533 036 chr12 53180823 CSAD,ZNF740 SNV utr_5, utr_5 G C 0.0385 0.0000 rs187048132 0.0537 037 chr12 53180847 CSAD,ZNF740 SNV utr_5, intronic A C 0.2163 0.0657 rs59218377 0.1991

Accession number: JGAS00000000144, the DDBJ Japanese Genotype-phenotype Archive for genetic and phenotypic human data. Chr., Chromosome, Ref., reference; SNV, single nucleotide variant; INDEL, insertion/deletion; utr, untranslated region. * Genome Reference Consortium Human Build 38 patch release 7 (GRCh38.p7), annotation release 108. † Frequency in Japanese in the 1000 Genomes Project browser (n=208, JTK: Japanese in Tokyo, Japan). https://www.ncbi.nlm.nih.gov/variation/tools/1000genomes/ ‡ Frequency in Europeans in the 1000 Genomes Project browser (n=198, CEU: Utah residents with Northern and Western European ancestry from the CEPH collection). § The SNPs used for fine mapping are shown in bold and italics. rs10876414 and rs1465056 were in complete linkage disequilibrium with the top-hit SNP, rs378215; these SNPs are shown for reference (underlined). Because the samples used for re-sequencing were selected from patients homozygous for the disease-associated allele at rs378215, no variant was identified for these SNPs during the re-sequencing. ǁ Frequency in Japanese in the database from the Tohoku Medical Megabank Organization (TMMO, n=3554). https://ijgvd.megabank.tohoku.ac.jp/ The term “novel” indicates that variants were identified by re-sequencing in this study. ¶ P values for association with fulminant type 1 diabetes obtained in this study Page 49 of 55 Diabetes Supplementary Figure 1. Principal component analysis of 676 samples used in the GWAS together with HapMap samples (43 JPT, 40 CHB, 91 YRI, and 91 CEU samples)

A B

CEU CHB YRI JPT Studied CASE CONTROL

(A) all groups, (B) focused on Japanese samples in the present study. CEU: Utah Residents (CEPH) with Northern and Western European ancestry; CHB: Han Chinese in Beijing, China; YRI: Yoruba in Ibadan, Nigeria; JPT: Japanese in Tokyo, Japan Diabetes Page 50 of 55 Supplementary Figure 2. Quantile-quantile plot for test statistics (Cochran-Armitage trend test) for 425,767 SNPs that passed quality control. The quantile-quantile plot after removal of the loci in the HLA region is shown in the inset.

8 inflation factor 6 λ = 1.046 20 4

2

15 -log10 (observed P-value) 0 0 1 2 3 -log10 (expected P-value)

10

inflation factor λ = 1.061 5 -log10 (observed P-value)

0

0 1 2 3 4

-log10 (expected P-value) Page 51 of 55 Diabetes

Supplementary Figure 3. Exon-intron structure of the CSAD gene and re-sequenced region.

Submitted region to design amplicon sequencing (23 kb) Sequenced region (17.16 kb) Uncovered region by amplicon sequencing Diabetes Page 52 of 55 Supplementary Figure 4 Linkage disequilibrium structure for the risk HLA-DR locus for fulminant type 1 diabetes in about 64 kb genetic region around GWAS top hit SNP rs9268853. A

B

r2

The top panel (A) shows the –log10 (P values) for 177 SNPs compared between 257 fulminant type 1 diabetes patients C and 419 healthy individuals.

The second and third panels show a Haploview plot of linkage disequilibrium (LD) between markers measured by r 2 D’ (B) and D’(C) in the 676 samples including 257 fulminant type 1 diabetes and 419 healthy individuals. The top hit SNP rs9268853 associated with fulminant type 1 diabetes with genome-wide significance is highlighted in red in the top panel Page 53 of 55 Diabetes Supplementary Figure 5 Locus zoom plot of 16 tagged SNPs in about 90 kb genetic region on chromosome 12q13.13

53.52Mb Chromosome 12 53.61Mb

Locus zoom plot* compared between 257 fulminant type 1 diabetes patients and 419 healthy individuals, using 1000 Genomes LD Asian population (Nov 2014) and the hg19 genome build as references. An top-hit SNP rs3782151 is highlighted. Colors of each circle represent r2 values. Blue line shows recombination rate. * Pruim RJ, Welch RP, Sanna S, et al. Bioinformatics 2010;26:2336-2337 Diabetes Page 54 of 55 Supplementary Figure 6 Linkage disequilibrium structure for the risk locus for fulminant type 1 diabetes in about 80 kb genetic region on chromosome 12q13.13 in samples with fulminant type 1 diabetes. A

r2

B

D’

A Haploview plot of linkage disequilibrium (LD) between markers measured by r2 (A) and D’(B) in the 257 samples with fulminant type 1 diabetes Page 55of

- log10 (P value) 10 12 14 0 2 4 6 8 1 diabetes with genome-wide significanceoutsideofHLAinFig.1is highlighted. the HLAascovariate.A top-hit SNP, rs11170445, associatedwithfulminanttype1 Genome-wide Manhattan plotapplyingaregressionanalysiswithtophitSNPrs9268853 in Supplementary Figure7. 01 12 11 10 9 8 7 6 5 4 3 2

Diabetes rs11170445 13 14 5181615 17 19 20 122 21 Supplementary Figure 8. Diabetes Page 56 of 55 Location of disease-associated SNPs in the CSAD/lnc-ITGB7-1 region and their effects on ITGB7 expression.

A 10 rs3782151 rs11170445 9

8

7

6

5

4 - log10 (P value)

3

2

1

0 53130K 53140K 53150K 53160K 53170K 53180K 53190K 53200K 53210K The top panel is the same as described in

B Figure 3 and shows the -log10 (P values) from 16 tagged SNPs from the CSAD-ITGB7 region. Genes (Long-non-coding RNA) The second panel is the genomic locations of the genes in this region. The top hit SNP rs3782151 is located in CSAD and in lnc- ITGB7-1, which encodes a long non-coding- Genes (protein coding genes) RNA (RP11-1136G11.7) also known as lnc- ITGB7-1:1.

The third panel shows the cis-eQTL of ITGB7 in the CSAD/lnc-ITGB7-1 region (Human C cis-eQTL (genotype) Genetic Variation Database [HGVD] Browser [http://www.hgvd.genome.med.kyoto- Probe: ITGB7 u.ac.jp/cgi-bin/searchBydbsnp.cgi]). SNPs in the CSAD/lnc-ITGB7-1 region are cis-eQTLs of ITGB7. eQTLs with a significant effect on cis-eQTL (exome) ITGB7 expression are connected with the Probe: ITGB7 ITGB7 probe (green box) using lines.