Identification of Non-HLA Genes Associated with Celiac Disease and Country-Specific Differences in a Large, International Pediatric Cohort

Abstract

Objectives: There are significant geographical differences in the prevalence and incidence of celiac disease that cannot be explained by HLA alone. More than 40 loci outside of the HLA region have been associated with celiac disease. We investigated the roles of these non-HLA genes in the development of tissue transglutaminase autoantibodies (tTGA) and celiac disease in a large international prospective cohort study.

Methods: A total of 424,788 newborns from the US and European general populations and first-degree relatives with type 1 diabetes were screened for specific HLA genotypes. Of these, 21,589 carried 1 of the 9 HLA genotypes associated with increased risk for type 1 diabetes and celiac disease; we followed 8676 of the children in a 15 y prospective follow-up study. Genotype analyses were performed on 6010 children using the Illumina ImmunoChip. Levels of tTGA were measured in serum samples using radio-ligand binding assays; diagnoses of celiac disease were made based on persistent detection of tTGA and biopsy analysis. Data were analyzed using Cox proportional hazards analyses.

Results: We found 54 single-nucleotide polymorphisms (SNPs) in 5 genes associated with celiac disease (TAGAP, IL18R1, RGS21, PLEK, and CCR9) in time to celiac disease analyses (10-4>P>5.8x10-6). The hazard ratios (HR) for the SNPs with the smallest P values in each region were 1.59, 1.45, 2.23, 2.64, and 1.40, respectively. Outside of regions previously associated with celiac disease, we identified 10 SNPs in 8 regions that could also be associated with the disease (P<10-4). A SNP near PKIA (rs117128341, P = 6.5x10-8, HR = 2.8) and a SNP near PFKFB3 (rs117139146, P<2.8x10-7, HR = 4.9) reached the genome-wide association threshold in subjects from Sweden. Analyses of time to detection of tTGA identified 29 SNPs in 2 regions previously associated with celiac disease (CTLA4, P = 1.3x10-6, HR = 0.76 and LPP, P = 2.8x10-5, HR = .80) and 6 SNPs in 5 regions not previously associated with celiac disease (P<10-4); non-HLA genes are therefore involved in development of tTGA.

Conclusions: In conclusion, using a genetic analysis of a large international cohort of children, we associated celiac disease development with 5 non-HLA regions previously associated with the disease and 8 regions not previously associated with celiac disease. We identified 5 regions associated with development of tTGA. Two loci associated with celiac disease progression reached a genome-wide association threshold in subjects from Sweden.

Fig 1. SNPs in the previously reported celiac disease associated regions.

Manhattan plot of P-values on the −log₁₀ scale for SNPs (±400kb) previously associated with celiac disease (A) and persistent tissue transglutaminase autoantibody (tTGA) positivity (B). HRs and p-values are calculated using three possible genotypes and adjusted for family history of celiac disease, HLA-DR-DQ genotype, gender, HLA-DPB1, population stratification (ancestral heterogeneity) and country of residence (as strata). The red dashed line represents p = 1x10⁻⁴. Kaplan-Meier plots of the three most significant SNPs associated with celiac disease (C) and tTGA (D) are plotted by dividing the subjects in two groups: (i) Major homozygous (black curves) and (ii) Heterozygous combined with minor homozygous (red curves).

Fig 2. Associations with risk of celiac disease and risk of persistent tissue transglutaminase autoantibody (tTGA) positivity.

Manhattan plot of 133,620 SNPs with MAF>0.01, displaying the P-values on the −log₁₀ scale for SNP associations with celiac disease (A) and persistent tTGA positivity (B). HRs and p-values are calculated using three possible genotypes and adjusted for family history of celiac disease, HLA-DR-DQ genotype, gender, HLA-DPB1, population stratification (ancestral heterogeneity) and country of residence (as strata). The red dashed line represents p = 1x10⁻⁴, the red solid line represents Bonferroni correction threshold. Kaplan-Meier plots of selected SNPs associated with celiac disease (C) and persistent tTGA (D) are plotted by dividing the subjects in two groups: (i) Major homozygous (black curves) and (ii) Heterozygous combined with minor homozygous (red curves).

Fig 3. Associations with risk of celiac disease in the Swedish population.

A: Manhattan plot of 133620 SNPs with MAF>0.01, displaying the P-values on the −log₁₀ scale for the SNPs associated with celiac disease in the Swedish TEDDY population. B: Regional association plots at the PKIA locus generated by LocusZoom, showing the significance of association and the recombination rate. Colors represent HapMap CEU linkage disequilibrium r² values with the most significantly associated SNP (rs117128341; shown in purple). C: Pairwise LD plot for five SNPs in the region of PKIA. The five most significant SNPs from this region are in high LD with each other.

Fig 4. Country-specific associations with risk of celiac disease.

Kaplan-Meier plots of five SNPs mapped to the PKIA region and one SNP mapped to the PFKFB3 region, in the Swedish TEDDY population (A) and in the other TEDDY countries (B). Kaplan-Meier plots clearly indicate country-specific differences. HRs and p-values are calculated using three possible genotypes and adjusted for family history of celiac disease, HLA-DR-DQ genotype, gender, HLA-DPB1 and population stratification (ancestral heterogeneity).

Fig 5. Flow chart of study participants.

The Environmental Determinants of Diabetes in the Young (TEDDY) is an international multicenter study that screened over 420,000 newborns from the general population in four different countries. The present study genotyped 195,806 SNPs on ImmunoChip in 6,010 TEDDY children to identify potential genetic factors responsible for the development of CD and country-specific differences in genetic predisposition. As shown in flow chart, a total of 6,010 subjects were included in the analysis of time-to-CD, and 5379 subjects were included in the analysis of time-to-tTGA.