Prenatal iron exposure and childhood type 1 diabetes

Abstract

Iron overload due to environmental or genetic causes have been associated diabetes. We hypothesized that prenatal iron exposure is associated with higher risk of childhood type 1 diabetes. In the Norwegian Mother and Child cohort study (n = 94,209 pregnancies, n = 373 developed type 1 diabetes) the incidence of type 1 diabetes was higher in children exposed to maternal iron supplementation than unexposed (36.8/100,000/year compared to 28.6/100,000/year, adjusted hazard ratio 1.33, 95%CI: 1.06-1.67). Cord plasma biomarkers of high iron status were non-significantly associated with higher risk of type 1 diabetes (ferritin OR = 1.05 [95%CI: 0.99-1.13] per 50 mg/L increase; soluble transferrin receptor: OR = 0.91 [95%CI: 0.81-1.01] per 0.5 mg/L increase). Maternal but not fetal HFE genotypes causing high/intermediate iron stores were associated with offspring diabetes (odds ratio: 1.45, 95%CI: 1.04, 2.02). Maternal anaemia or non-iron dietary supplements did not significantly predict type 1 diabetes. Perinatal iron exposures were not associated with cord blood DNA genome-wide methylation, but fetal HFE genotype was associated with differential fetal methylation near HFE. Maternal cytokines in mid-pregnancy of the pro-inflammatory M1 pathway differed by maternal iron supplements and HFE genotype. Our results suggest that exposure to iron during pregnancy may be a risk factor for type 1 diabetes in the offspring.

Figure 1

Research questions and study design. (a) Illustration of exposures and outcomes (sTfR: soluble transferrin receptor) This figure is not covered by the CC BY licence. [Picture: ^© Can Stock Photo/Eraxion, ^© Can Stock Photo/adekvat and from http://www.ebi.ac.uk/.]. All rights reserved, used with permission. (b) Flow chart illustrating the formation of the study sample in the MoBa cohort, including sub-study of biomarkers. By 1^st of January 2013, cases (n = 276) and random controls (n = 1010) were identified for the study of biomarkers, out of whom 177 and 508 respectively had cord blood samples available. The whole cohort was followed up until 1^st of May 2017, and 373 developed type 1 diabetes.

Figure 2

Perinatal iron exposures and risk of type 1 diabetes. (a) Risk of type 1 diabetes in the offspring by use of iron-containing supplements during pregnancy. Picture: ^©Can Stock Photo/Eraxion, ^©Can Stock Photo/adekvat and from http://www.ebi.ac.uk/. (b) Maternal iron supplement use, type of supplement used, duration of iron supplements and risk of type 1 diabetes. (c) Cord blood ferritin and soluble transferrin receptor (TfR) among children who later developed type 1 diabetes (cases) and randomly selected controls from the same cohort. (d) Maternal and fetal HFE genotype among children who later developed type 1 diabetes (cases) and randomly selected controls from the same cohort. Panel b and c Adjusted for maternal age and education, smoking, parity, pre-pregnancy BMI, mode of delivery, birth weight and prematurity, maternal type 1 diabetes, maternal celiac disease and diagnosed maternal anaemia (<17 weeks). Panel c additionally adjusted for year of birth. Ferritin quartiles: 1^st quartile < 96 mg/L, 2^nd quartile 96–142 mg/L, 3^rd quartile 142.1–207 mg/L, 4^th quartile >207 mg/L. sTFR quartiles: 1^st quartile <2.15 mg/L, 2^nd quartile 2.15–2.63 mg/L, 3^rd quartile 2.631–3.31 mg/L, 4^th quartile >3.31 mg/L. Panel d Reciprocally adjustment for maternal/fetal genotype Standard genotype: Wild type HFE allele at both loci (C282 and H63). Intermediate genotype: p.63D/63D homozygotes and 63D or 282Y heterozygotes High genotype: p.282Y homozygous or p.282Y/63D compound heterozygotes.

Figure 3

Perinatal iron exposure and genome-wide cord blood DNA methylation. (a) No genome-wide significant association between maternal iron supplement use in pregnancy and genome-wide DNA methylation assayed by Illumina 450 K chip, in n = 1,062 mother/child-pairs in the MoBa cohort. (b) Methylation in cord blood samples from the MoBa cohort by the coding SNP rs1800562 (p.C282Y) in the HFE gene at chromosome 6. (c) Methylation in cord blood samples from the MoBa cohort by the coding SNP rs1799945 (p.H63D) in the HFE gene at chromosome 6. (d) Fetal HFE rs1800562 genotype and genome-wide cord blood DNA methylation in n = 1,062 children in the MoBa cohort showed multiple differentially methylated sites near the HFE gene with genome-wide significance. rs1800562 is the hemochromatosis associated non-synonymous p.C282Y SNP. The red marks represent associations for the six methylation sites (CpGs) that were replicated in the ALSPAC cohort in the UK, for the same SNP, reported in their metQTL database. The Manhattan plot is split according to direction of association, with positive associations (hypermethylation) associated with the p.C282Y variant) at the top and negative associations (hypomethylation) at the bottom. The mid panel indicates the position of the genes in this region of chromosome six. Numbers on the x-axis indicate genome position.