Previously, we examined 20 non-HLA SNPs for association with islet autoimmunity (IA) and/or progression to type 1 diabetes (T1D). Our objective was to investigate fourteen additional non-HLA T1D candidate SNPs for stage- and age-related heterogeneity in the etiology of T1D. Of 1634 non-Hispanic white DAISY children genotyped, 132 developed IA (positive for GAD, insulin, or IA-2 autoantibodies at two or more consecutive visits); 50 IA positive children progressed to T1D. Cox regression was used to analyze risk of IA and progression to T1D in IA positive children. Restricted cubic splines were used to model SNPs when there was evidence that risk was not constant with age.
Type 1 diabetes (T1D) is a chronic autoimmune disease in which the insulin-producing beta cells of the pancreas are destroyed. There is typically a preclinical phase of circulating autoantibodies, called islet autoimmunity (IA), which precedes the clinical diagnosis of T1D. T1D is widely believed to be caused by an environmental factor on a susceptible genetic background. The major susceptibility locus for T1D maps to the HLA class II genes at chromosome 6p21. These HLA class II alleles account for 30–50% of the familial clustering of T1D [
More than 50 non-HLA T1D susceptibility gene markers have been confirmed. The major non-HLA loci include
Prospective birth cohorts have the unique ability to study two stages in the natural history of T1D: development of IA and progression to T1D in IA positive children. Different exposures have been associated with one or both stages. For instance, DAISY recently identified an association between a gene-gene interaction involving the vitamin D receptor gene (
The Diabetes Autoimmunity Study in the Young (DAISY) is a prospective study composed of two groups of children at increased risk for T1D who were recruited between 1993 and 2004 and are being followed prospectively for the development of IA and T1D. One group is made up of first degree relatives of patients with T1D, identified and recruited between birth and eight years of age, mainly through the Barbara Davis Center for Childhood Diabetes (
Autoantibodies were tested at 9, 15, and 24 months, and annually thereafter, or at their first visit and annually thereafter if the child enrolled after birth. Radioimmunoassays were used to measure serum autoantibodies to insulin, GAD-65, and IA-2 (BDC512), as previously described [
Cases of IA were defined as those children positive for at least one islet autoantibody (IAA, GAD-65, IA-2) on two or more consecutive visits. T1D was diagnosed by a physician and defined as random blood glucose >200 mg/dL and/or HbA1c (A1C) >6.5% with clinical symptoms of T1D.
DAISY children were genotyped for fourteen non-HLA T1D candidate SNPs:
The SNPs were genotyped utilizing the Taqman SNP genotype based OpenArray platform (Applied Biosystems, CA, USA). Custom designed 48-sample arrays and normalized genomic DNA were loaded using the OpenArray AccuFill system and cycling was performed on a GeneAmp 9700 PCR system (Applied Biosystems, CA, USA), all according to manufacturer protocol. Alleles were analyzed using the OpenArray SNP genotyping analysis software v.1.0.3 and Taqman Genotyper Software 2.0 (Applied Biosystems, CA, USA). All fourteen SNPs had a 95% call rate or higher.
Each SNP was tested for consistency with Hardy-Weinberg proportions using a 1-degree of freedom
We obtained genetic data on at least one of the fourteen non-HLA T1D candidate SNPs for 1634 non-Hispanic white children in the DAISY cohort. This included 132 children who developed IA, of whom 50 went on to develop T1D. Fifteen IA cases were positive for autoantibodies on their first clinic visits; these left-censored cases were removed from the development of IA analysis cohort but retained in the progression from IA to T1D cohort. The same 50 IA positive children who went on to develop T1D are the same 50 T1D cases in the development of T1D analyses. However, not all IA positive children went on to develop T1D. All statistical analyses were limited to non-Hispanic whites in the DAISY cohort.
SAS version 9.3 (SAS Institute Inc., Cary, NC, USA) statistical software package was used for all statistical analyses. SNPs were tested for violation of the proportional hazards assumption using a supremum test, with a
We first examined whether non-HLA variants were associated with development of IA. The mean age at first IA positive visit was 6.2 years, and the mean age at last follow-up visit in children who did not develop IA was 9.9 years (Table
Demographic characteristics of DAISY non-Hispanic white population.
Characteristic | Development of islet autoimmunity (IA) ( |
Progression from IA to type 1 diabetes (T1D) ( |
||||||
---|---|---|---|---|---|---|---|---|
Children positive for IA ( |
Children negative for IA ( |
Univariate HR and 95% CI |
|
IA positive children who progressed to T1D ( |
IA positive children who have not progressed to T1D ( |
Univariate HR and 95% CI |
|
|
Mean age (years) |
|
|
N/A | N/A |
|
|
N/A | N/A |
Mean age at first IA positive visit (years) | N/A | N/A | N/A | N/A |
|
|
0.86 (0.77, 0.95) | 0.003 |
HLA-DR3/4, |
43 (36.8%) | 248 (16.5%) | 2.97 (2.01, 4.39) | <0.0001 | 28 (56.0%) | 19 (23.2%) | 2.79 (1.65, 4.72) | 0.0001 |
First degree relative with T1D | 75 (64.1%) | 737 (49.1%) | 1.41 (0.96, 2.07) | 0.08 | 36 (72.0%) | 54 (65.9%) | 1.02 (0.56, 1.85) | 0.96 |
Sex (female) | 62 (53.0%) | 712 (47.4%) | 1.22 (0.85, 1.74) | 0.28 | 24 (48.0%) | 44 (53.7%) | 1.03 (0.59, 1.83) | 0.91 |
CI: confidence interval; DAISY: Diabetes Autoimmunity Study in the Young; HLA: human leukocyte antigen; HR: hazard ratio; IA: islet autoimmunity; T1D: type 1 diabetes.
aAge at first IA positive visit.
bAge at last followup.
cAge at T1D diagnosis.
Unadjusted SNP association analyses are presented in (see Supplemental Table 1 in Supplemetary Material available online at
Association between non-HLA T1D candidate SNPs and development of IA, progression from IA to T1D, and development of T1D adjusted for HLA-DR3/4, DQB1*0302 genotype and first degree relative with T1D.
SNP | Minor allele | Development of IA ( |
Progression from IA to T1D |
Development of T1D |
||||
---|---|---|---|---|---|---|---|---|
MAFa | Adjusted HRb and 95% CI |
|
Adjusted HRc and 95% CI |
|
Adjusted HRb and 95% CI |
|
||
|
A | 0.44 |
|
|
1.13 (0.75, 1.71)d | 0.56 |
|
|
|
G | 0.46 | 1.15 (0.88, 1.51)d | 0.31 | 0.87 (0.61, 1.24)d | 0.44 | 1.03 (0.69, 1.54)d | 0.87 |
|
T | 0.28 | 0.97 (0.73, 1.27)d | 0.80 | 1.09 (0.62, 1.92)e | 0.77 | 0.70 (0.40, 1.23)e | 0.21 |
|
C | 0.29 | 1.08 (0.82, 1.43)d | 0.58 | 1.22 (0.79, 1.87)d | 0.38 | 1.20 (0.79, 1.83)d | 0.40 |
|
G | 0.35 | 1.04 (0.80, 1.34)d | 0.78 | 0.93 (0.67, 1.28)d | 0.65 | 1.18 (0.82, 1.70)d | 0.37 |
|
A | 0.12 | 0.90 (0.58, 1.41)e | 0.65 | 0.52 (0.17, 1.57)e | 0.25 | 0.43 (0.17, 1.08)e | 0.07 |
|
C | 0.27 | 0.85 (0.63, 1.15)d | 0.28 | 1.26 (0.70, 2.27)e | 0.44 | 0.77 (0.43, 1.38)e | 0.38 |
|
T | 0.28 | 0.93 (0.68, 1.28)d | 0.66 | 0.86 (0.49, 1.53)e | 0.61 | 0.88 (0.50, 1.55)e | 0.65 |
|
G | 0.17 | 1.19 (0.80, 1.77)e | 0.40 | 0.76 (0.41, 1.42)e | 0.39 | 0.87 (0.47, 1.64)e | 0.67 |
|
C | 0.25 | 0.90 (0.66, 1.22)d | 0.49 | 1.54 (0.83, 2.85)e | 0.17 | 1.01 (0.57, 1.78)e | 0.99 |
|
T | 0.35 | 0.94 (0.72, 1.22)d | 0.63 | 0.94 (0.57, 1.57)d | 0.82 | 0.73 (0.45, 1.20)d | 0.21 |
|
C | 0.20 | 0.82 (0.64, 1.05)d | 0.11 | 0.84 (0.61, 1.16)d | 0.29 | 0.86 (0.60, 1.24)d | 0.43 |
|
A | 0.44 | 1.19 (0.89, 1.59)d | 0.25 | 1.03 (0.72, 1.48)d | 0.86 |
|
|
rs10517086 | A | 0.30 | * | * | 1.36 (0.77, 2.41)e | 0.30 |
|
|
CI: confidence interval; DAISY: Diabetes Autoimmunity Study in the Young; HLA: human leukocyte antigen; HR: hazard ratio; IA: islet autoimmunity; MAF: minor allele frequency; T1D: type 1 diabetes.
aMinor allele frequency (MAF) calculated for children negative for IA.
bAdjusted for HLA-DR3/4, DQB1*0302 genotype and first degree relative with T1D.
cAdjusted for HLA-DR3/4, DQB1*0302 genotype, first degree relative with T1D, and age at first antibody positive visit.
dSNP analyzed additively with HR representing increase in risk for each additional minor allele.
eSNP analyzed dichotomously with HR representing increase in risk for at least one minor allele.
*SNP rs10517086 did not meet the assumptions of proportional hazards in the development of IA analysis and therefore was modeled using a restricted cubic spline (Figure
Association between SNP rs10517086 and development of IA modeled using a restricted cubic spline. The hazard ratio at different ages is represented by the solid line and the 95% confidence intervals are represented by the dotted lines. Hazard ratios and 95% confidence intervals are shown at 9 months, 2 years, 4 years, and 6 years (denoted with short vertical lines) illustrating the increased risk of developing IA in the younger ages, but not in older ages.
We then examined whether non-HLA variants were associated with progression to T1D in IA positive children. Of the 132 IA positive children in DAISY, 50 developed T1D; the mean age at T1D diagnosis was 8.7 years (Table
In order to evaluate the same outcome as previous GWAS to see if similar associations could be seen using a time-to-event analysis in a prospective birth cohort and to better understand the role these SNPs play in the natural history of T1D, we examined whether these non-HLA variants were associated with T1D in our population of 1619 children. All 50 children who developed T1D had developed IA previously, so the same 50 T1D cases were included in both the progression from IA to T1D (presented in Section
In exploring associations between fourteen previously discovered non-HLA T1D candidate SNPs and the development of IA and progression to T1D in the prospective DAISY cohort, we found that
SNP (rs10517086) exhibits an age-related effect with development of IA, with an increased risk of developing IA before the age of two or in younger ages, and a null effect in older ages. Children carrying a risk allele for SNP (rs10517086) developed IA significantly earlier than children without a risk allele. Based on these epidemiologic analyses, future studies should investigate mechanisms as to how this SNP influences risk of early autoimmunity. SNP (rs10517086), which was first discovered through another GWAS and meta-analysis of T1D, is located within a gene desert on chromosome 4. Loci in or near genes without a known function or in regions not containing annotated genes may indicate involvement of long-range gene expression regulatory elements and/or nonprotein-coding RNA genes [
In combination with those presented in this paper, DAISY has now investigated 34 non-HLA T1D candidate SNPs for association with development of IA, progression from IA to T1D, and/or development of T1D with multiple examples of stage-related heterogeneity, which are presented in Table
Non-HLA T1D candidate SNPs associated with development of IA, progression from IA to T1D, and/or development of T1D in DAISY.
SNP | Development of IA | Progression from IA to T1D | Development of T1D | |||
---|---|---|---|---|---|---|
Adjusted HR and 95% CI |
|
Adjusted HR and 95% CI |
|
Adjusted HR and 95% CI |
|
|
|
|
|
1.13 (0.75, 1.71)abd | 0.56 |
|
|
|
1.12 (0.86, 1.46)aef | 0.42 |
|
|
1.00 (0.70, 1.43)ahc | 1.00 |
|
1.39 (0.99, 1.95)aef | 0.05 | 1.34 (0.72, 2.52)aeg | 0.35 |
|
|
|
|
|
0.65 (0.27, 1.60)ijk | 0.35 | 0.99 (0.60, 1.66)ach | 0.98 |
|
|
|
0.98 (0.50, 1.93)aeg | 0.96 |
|
|
|
|
|
** | ** |
|
|
|
1.19 (0.89, 1.59)abc | 0.25 | 1.03 (0.72, 1.48)abd | 0.86 |
|
|
rs10517086 | * | * | 1.36 (0.77, 2.41)bdj | 0.30 |
|
|
DAISY: Diabetes Autoimmunity Study in the Young; HLA: human leukocyte antigen; HR: hazard ratio; CI: confidence interval; IA: islet autoimmunity; T1D: type 1 diabetes.
aSNP analyzed additively with HR representing increase in risk for each additional minor allele.
bListed in Table
cAdjusted for HLA-DR3/4, DQB1*0302 genotype and first degree relative with T1D.
dAdjusted for HLA-DR3/4, DQB1*0302 genotype, first degree relative with T1D, and age at first antibody positive visit.
eFrom Steck et al. (2009) [
fAdjusted for HLA-DR3/4, DQB1*0302 genotype, ethnicity, sex, and first degree relative with type 1 diabetes.
gAdjusted for HLA-DR3/4, DQB1*0302 genotype, ethnicity, sex, first degree relative with type 1 diabetes, and age at first antibody positive visit.
hFrom Steck et al.(2012) [
iFrom Frederiksen et al. (2013) [
jSNP analyzed dichotomously with HR representing increase in risk for at least one minor allele.
kAdjusted for
**Analysis not conducted.
*SNP rs10517086 did not meet the assumptions of proportional hazards in the development of IA analysis and therefore was modeled using a restricted cubic spline (Figure
GWAS are important for identifying new candidate regions associated with a clinical outcome, such as T1D, in a large number of cases and controls. Prospective birth cohort studies, like DAISY, are then able to take these newly identified candidate regions and look for associations with different stages in the disease process and at different ages. As a prospective birth cohort study following children at increased risk for developing T1D from birth, we are able to capture the preclinical phase of T1D, islet autoimmunity. This allows us to study two separate stages in the natural history of T1D: development of IA and progression to T1D in IA positive children. We are also able to study whether certain exposures are important at one age, but not another.
Due to the cost associated with following a large group of children from birth into adulthood, our sample sizes are much smaller than those obtained for GWAS. Our lack of association for many of these SNPs is not evidence against their association with T1D but may result from limited power, especially in the progression from IA to T1D stage. We also have a very unique population comprised of children with high risk HLA genotypes and it is possible that the effect of these SNPs differs based on one’s HLA risk status. The risk for non-HLA loci appears to be lower in individuals carrying high-risk HLA genotypes, as has been seen with
We believe that taking these GWAS identified candidate regions and studying them in the context of the natural history of T1D are central to better understanding the disease process and where in the disease process genetic loci may be important. This will allow us to create more accurate risk prediction models for both stages in the natural history of the disease, as well as inform the design of targeted interventions to prevent or slow the progression of IA and subsequent development of T1D.
The effect of a SNP may act nonlinearly, with an effect at early ages but not later ages or vice versa. Our results provide evidence that SNP (rs10517086) is acting on early risk of IA, with the age at onset of IA occurring significantly earlier in children with a minor allele compared to children with no minor alleles. By ignoring heterogeneity in the etiology of disease, valuable associations may be missed that could aid in better understanding complex diseases, such as T1D.
Confidence interval
C1q and tumor necrosis factor related protein 6
Chromosome 6 open reading frame 173
Chromosome 14 open reading frame 181
Cytotoxic T-lymphocyte-associated protein 4
Human leukocyte antigen
Hazard ratio
Islet autoimmunity
Interferon induced with helicase C domain 1
Interleukin 2
Interleukin 2 receptor, alpha
Interleukin 7 receptor
Insulin
Protein kinase C, theta
Protein tyrosine phosphatase, nonreceptor type 22
Src kinase associated phosphoprotein
SWI/SNF related, matrix associated, actin dependent regulator of chromatin, subfamily E, member 1
Single nucleotide polymorphism
Type 1 diabetes
Toll-like receptor 8
Ubiquitin associated and SH3 domain containing A.
The authors declare no conflict of interests.
This research was supported by NIH Grants R01-DK49654, R01-DK32493, and R01-DK32083, DERC Molecular Biology Core NIH P30 DK57516, and the NIH/NCRR Colorado CTSI Grant no. UL1 RR025780.