Complement gene variants in relation to autoantibodies to beta cell specific antigens and type 1 diabetes in the TEDDY Study

A total of 15 SNPs within complement genes and present on the ImmunoChip were analyzed in The Environmental Determinants of Diabetes in the Young (TEDDY) study. A total of 5474 subjects were followed from three months of age until islet autoimmunity (IA: n = 413) and the subsequent onset of type 1 diabetes (n = 115) for a median of 73 months (IQR 54–91). Three SNPs within ITGAM were nominally associated (p < 0.05) with IA: rs1143678 [Hazard ratio; HR 0.80; 95% CI 0.66–0.98; p = 0.032], rs1143683 [HR 0.80; 95% CI 0.65–0.98; p = 0.030] and rs4597342 [HR 1.16; 95% CI 1.01–1.32; p = 0.041]. When type 1 diabetes was the outcome, in DR3/4 subjects, there was nominal significance for two SNPs: rs17615 in CD21 [HR 1.52; 95% CI 1.05–2.20; p = 0.025] and rs4844573 in C4BPA [HR 0.63; 95% CI 0.43–0.92; p = 0.017]. Among DR4/4 subjects, rs2230199 in C3 was significantly associated [HR 3.20; 95% CI 1.75–5.85; p = 0.0002, uncorrected] a significance that withstood Bonferroni correction since it was less than 0.000833 (0.05/60) in the HLA-specific analyses. SNPs within the complement genes may contribute to IA, the first step to type 1 diabetes, with at least one SNP in C3 significantly associated with clinically diagnosed type 1 diabetes.

After Bonferroni correction, none of the 15 SNPs remained as significant.
There was no significant interaction between any of the complement SNPs on IA and HLA-risk categories.

Discussion
In the present study we were not able to demonstrate a robust association between IA and SNPs within the complement genes. Nevertheless, we found a significant association with type 1 diabetes time-to-event with one SNP (rs2230199) in the C3 gene, but only among DR4/4 subjects. There was a strong association for C3 rs2230199 among HLA-DR4/4 carriers, the genotype conferring the second highest risk for type 1 diabetes. An infection could precipitate type 1 diabetes in those with genetic predisposition (based upon rs2230199 in C3) since the expression levels of C3 could be different and these subject would not be able to vigorously fight the pathogen. C3 is a key player in the formation of MAC from all three pathways in the activation of complements. An infection within the islets of Langerhans' shortly before the onset of type 1 diabetes could explain the finding of cytotoxic T-cells, which are part of the innate immune response, at the time of diagnosis. The C3 rs2230199 SNP is a functional polymorphism, where the G allele causes an amino acid substitution where arginine is substituted with glycine (R102G) 20 . The glycine variant of C3 moves faster in electrophoresis and has been shown to confer increased adhesion to monocytes 21 . The R102G variant has shown a strong association with age-related macular degeneration (AMD) 20,22,23 . One feature indicating different C3 levels in those with type 1 diabetes is that children with type 1 diabetes have impaired clot lysis due to increased incorporation of C3 into clots thereby prolonging lysis time 24 .
There was a strong correlation between the three SNPs in ITGAM (rs1143678 vs rs1143679, r s = 0.85, p < 0.0001, rs1143678 vs rs1143683, r s = 0.99, p < 0.0001 and rs1143679 vs rs1143683, r s = 0.85, p < 0.0001) consistent with strong linkage disequilibrium in this region. All three SNPs have previously been shown to impair Mac-1-mediated neutrophil phagocytosis and phagocytosis mediated by Fcγ receptors as well as to impair the firm adhesion of neutrophils in healthy subjects 25 . SNP rs17615 in CD21 was associated with an increased hazard ratio using time to type 1 diabetes among HLA-DR3/4 carriers, while those with the rs17615 minor allele have been shown to have a protection from SLE. SNP rs17615 (together with rs1048971 and rs4308977) have a biological function that decrease the splicing of the alternatively spliced exon 11 conferring increased amounts of the short CR2 isoform 26 . ITGAM, also denoted CD11b-integrin is an adhesion molecule expressed on several cells within the innate immune system such as granulocytes, monocytes, macrophages and natural killer cells. Moreover, ITGAM is involved in the complement system due to its capacity to bind inactive complement component 3b (iC3b) 27 . One variant within ITGAM (rs1143679, A-allele) confers an amino acid substitution where arginine is substituted with histidine R77H. This substitution causes a functional dysfunction of ITGAM leading to impaired adhesion and reduced phagocytosis of complement opsonized antigens. Interestingly, the R77H variant of ITGAM is highly associated with systemic lupus erythematosus (SLE) 28 . CD21 is also known as the C3d-receptor or the Epstein-Barr virus receptor 29 . The C3d-receptor is expressed on B-cells and has been shown Chr SNP  to increase the B-cell receptor (BCR) signaling as well as to enhance the processing of antigens 30 . C4BPA has an inhibitory effect within the complement system 31 .

Gene of interest
The major strength of the present study is that we have children followed from birth who, in some cases, developed IA and in some cases progressed to type 1 diabetes, thereby allowing us to study both disease initiation and progression. Additional strengths are that we only have included subjects that have high-risk HLA genotypes, thereby setting an equal baseline for the major genetic susceptibility. The study design with repeated sampling (more often in the toddlers and more dispersed after four years of age) allowed us to use proportional hazard ratio modeling to identify hazard ratios for time to end-points (IA and type 1 diabetes) in this cohort consisting of three European populations and a USA population (three centers combined). This study design increase the power compared with cross-sectional studies of cases and controls. A current limitation on the study, is the relatively low number of participants that have reached the end-points (IA and type 1 diabetes). The Bonferroni correction might be too conservative if some SNPs are in linkage disequilibrium as is most likely the case for the four SNPs within ITGAM, the three SNPs in CD21 the two SNPs in C3 and the two SNPs in C5. Nevertheless, among DR4/4 subjects SNP rs2230199 in C3 achieved Bonferroni-corrected significance. As the predisposition to type 1 diabetes occurred only in those carrying HLA-DR4/4, the proposed apparent contribution is necessary and requires confirmation in functional studies.
As our TEDDY cohort gets older more children will proceed to IA and type 1 diabetes, and we will have more power in our determination of association of complement factor gene variants and hazard ratios using time to IA and type 1 diabetes.
In summary, although complement gene SNPs are not robustly associated with IA or type 1 diabetes, we concluded that variants in the complement genes do have some apparent contribution to IA and type 1 diabetes. However, discrimination of complement pathways effects requires further research.

Subjects and Methods
Newborn children were screened for high-risk HLA genotypes and enrolled into the TEDDY study from 1 September 2004 until 28 February 2010. A total of 424 788 newborns in Finland, Sweden, Germany and the United States (Colorado, Georgia and Washington) were screened as previously described 32,33 . The HLA highrisk genotypes for subjects from the general population (GP) were as follows: DR3/4, DR4/4, DR4/8 and DR3/3. An additional six genotypes were used for first degree relatives (FDRs) to a subject with type 1 diabetes: DR4/4, DR4/1, DR4/13, DR4/4, DR4/9, DR3/9 (Table 5). DR4 subtyping was performed to exclude children from the GP with DRB1* 04:03. DQB1* 03:04 also qualified for inclusion into the TEDDY study. Subtyping was not done to  . Proportional hazard models were adjusted for HLA genotypes, sex and country of residence and population stratification. The first two principal components resulting from a principal components analysis of the US participants were included in the model to adjust for population stratification. First degree relatives were excluded from the analyses. Hazard ratios (HR) and 95% confidence intervals (95% CI) were estimated in this time-to-event analysis. A robust variance estimate was used to account for the dependence within families in the models. For a Chi-square value to remain significant after Bonferroni correction for multiple comparisons of 15 SNPs it must be less than 0.00333. Abbreviations: CD21: C3D-receptor, complement receptor type 2, EBV-receptor, ITGAM: Integrin alpha M = MAC1 or complement receptor 3, C4BPA: C4 binding protein alpha chain.
distinguish BQB1* 02:0X and DQA1* 03:0X subtypes. In the DQB1* 05:01 haplotype, DR10 had to be excluded, only DR1 was eligible. A total of 8667 newborn children were enrolled based on these criteria. The initial blood sample for HLA-screening was obtained either as cord blood in the maternity clinic or as dry blood spot (DBS) on day three to four. If the child carried any of the high-risk genotypes, the family was contacted by a study nurse and invited to participate in the 15 year follow-up study with blood sampling for analysis of autoantibodies (GADA, IA-2A and mIAA) every 3 months between 3 and 48 months of age and every six months thereafter. HLA-genotypes were confirmed in a second blood sample at the 9-month visit. This blood sample was also used for the ImmunoChip SNP-analyses. This study was performed according to the principles expressed in the Declaration of Helsinki. Signed informed consents were obtained for all study participants from their parents or primary caretakers, separately, for genetic screening and participation in the follow-up.  The explicit values of the HR (95% CI) for type 1 diabetes in the HLA stratified analyses can be found in an online Appendix. Panel A HRs and 95% CIs in 65 subjects with type 1 diabetes and 2139 subjects without type 1 diabetes carrying the HLA-DR3/4-genotype (n = 2204). Panel B HRs and 95% CIs in 19 subjects with type 1 diabetes and 1067 subjects without type 1 diabetes carrying the HLA DR4/4-genotype (n = 1086). Panel C HRs and 95% CIs in 16 subjects with type 1 diabetes and 950 subjects without type 1 diabetes carrying the HLA-DR4/8-genotype (n = 966). Panel D HRs and 95% CIs in 15 subjects with type 1 diabetes and 1203 subjects without type 1 diabetes carrying the HLA-DR3/3-genotype (n = 1218).
Scientific RepoRts | 6:27887 | DOI: 10.1038/srep27887 Study outcome -Islet autoimmunity (IA) and type 1 diabetes. The primary outcome was the finding of persistent confirmed IA assessed every 3 months between 3 and 48 months of age and every 6 months thereafter. IA was confirmed if identified in both Reference Laboratories. Persistent autoimmunity was defined by the finding of confirmed IA (any of GADA, IA-2A or mIAA) on two or more consecutive visits. Date of persistent confirmed IA was defined as the draw date of the first of two consecutive samples of at which the child was confirmed as "positive" for a specific autoantibody or any autoantibody. Because children can be born with maternal islet autoantibodies, positive results due to maternal transmission via placenta were excluded when defining the child's IA status. In order to discriminate maternal autoantibodies from IA in the child, the IA status of the mother was assessed when the child was 9 months if IA was detected at 3 or 6 months of age. The child's IA status was defined based on both maternal and child IA. If maternal autoantibodies were identified, the child was not considered persistently IA positive unless the child had a negative sample prior to the first positive sample. All samples with a positive result of IA and 5% of negative samples were re-analyzed for confirmation in both Reference Laboratories. In the U. S., all samples were assayed at the Barbara Davis Center for Childhood Diabetes at the University of Colorado, Denver; in Europe all samples were assayed at the University of Bristol, the U. K. All the autoantibodies were measured using radioimmuno-binding assays [34][35][36][37] . The secondary outcome of the study was the diagnosis of type 1 diabetes as defined by guidelines from the American Diabetes Association 16 .
HLA-typing. HLA-genotype screening was performed using either a DBS punch or a small volume whole blood lysate (WBL) 33,38 . Following PCR amplification of exon 2 of the HLA class II gene (DRB1, DQA1 or DQB1), alleles were identified either by direct sequencing, oligonucleotide probe hybridization, or other genotyping techniques. Typing to certify specific DR-DQ haplotypes were specified for each clinical center. HLA genotypes in eligible subjects were confirmed, using the 9-month sample, by the central HLA Reference Laboratory at Roche Molecular Systems, Oakland, CA.

Single Nucleotide Polymorphisms (SNPs). SNP analysis was performed by the Center for Public Health
Genomics at University of Virginia, using the Illumina ImmunoChip. The ImmunoChip is a custom array for genotyping of SNPs selected from regions of the human genome firmly associated with autoimmune diseases. The final selection of SNPs containing ~186 000 SNPs in 186 regions, for 12 autoimmune diseases was decided by the ImmunoChip Consortium. The 9-month sample was used for SNP genotyping after DNA extraction done by Roche Molecular Systems (Pleasanton, CA). Quality control (QC) steps to assure high quality of the reported SNPs comprised the exclusion of subjects due to low call rate (> 5% SNPs missing) and discordance with reported sex and prior genotyping. Secondly, SNPs were removed from analysis due to low call rate (< 95%), Hardy-Weinberg equilibrium (HWE) p-value < 10 −6 (except for chromosome 6 due to HLA eligibility requirements) as well as being monomorphic or an insertion-deletion.
A total of 17 SNPs residing in loci within genes coding for complement factors were present on the ImmunoChip. Of these 17 SNPs, two did not pass QC (rs1061170 in CFH and rs2274567 in CD35). In the TEDDY study, we have previously performed a confirmation study of 41 SNPs associated with type 1 diabetes and risk of autoantibody positivity 39 .  Statistical analyses. Two outcomes were analyzed: the time to persistent confirmed IA and the time to type 1 diabetes. The time to persistent confirmed IA was defined as the age when the first of confirmed positive samples was taken, and the right censored time when the last negative sample was taken. The time to type 1 diabetes was the child's age at the time of diagnosis of type 1 diabetes. The primary analysis was to investigate the association of the SNPs within the complement genes with the risk of development of persistent confirmed IA/ type 1 diabetes. The minor allele frequency (MAF) was determined in the selected cohort (n = 5474) and used throughout the analyses. This strategy infers that the risk is associated with the minor allele, even though the certain allele that has the lower frequency may vary across populations. Cox proportional hazard model was used for the analyses, adjusting for HLA, sex and country of residence. In addition, the first two principal components (PC) of the ImmunoChip data were used as covariates in the Cox model to adjust for potential population stratification. The axes of the principal components were estimated based on the TEDDY US Caucasian subjects and principle components for all TEDDY Subjects were calculated by projecting their data onto the estimated axes. The differences between countries (especially Finland individuals that typically have different PCs) were adjusted by the country covariate, and other population substructure (such as northern vs southern European of origin) was adjusted by PC1 and PC2. We have re-calculated the data also without correction for the first two PCs and the only marked change was that the p-value for rs4597342 in the IA analysis increased from 0.0406 to 0.0519. Therefore, we present the data with adjustment for the first two PCs throughout this publication. A total of 4850 families participated with only one child, 297 families with 2 children, and 10 families with 3 children. A robust variance estimate was used to account for the dependence within families in the Cox proportional hazard model 40 .
The secondary analysis was to study the tentative interaction between non-HLA SNPs and HLA-genotype and also country-specific association of the SNPs with the time to development of persistent confirmed IA/type 1 diabetes.
The strength and directions of associations were denoted by hazard ratios (HR) with 95% confidence intervals (95% CI). P-values less than 0.05 were considered to indicate nominal statistical significance. Significance after Bonferroni correction was achieved when p-values were less than 0.00333 (0.05/15) in the entire cohort and less than 0.000833 (0.05/60) in the HLA-specific analyses. Statistical analyses were performed using the Statistical Analytical Software (Version 9.3, SAS Institute, Cary, NC). Quality control analyses were performed using PLINK (http://pngu.mgh.harvard.edu/purcell/plink) 41 . The PCA was performed using EIGENSTRAT software 42 .