Impact of iron homeostasis biomarkers in non-communicable diseases using Mendelian randomization

Abstract

Iron, an essential mineral for biological functioning contained in many foods and dietary supplements, is associated with non-communicable diseases (NCDs) of global health importance, such as type 2 diabetes (T2D), liver steatosis, and chronic kidney disease (CKD). However, evidence is mainly from conventional observational studies, which are vulnerable to residual confounding and selection bias. Findings from Mendelian randomization (MR) studies, a design less prone to confounding, are also unclear given previous MR studies used few, potentially pleiotropic, genetic instruments. To assess the interrelations between iron homeostasis biomarkers and these NCDs and their biomarkers, as well as to ascertain whether previous findings were biased, I conducted a series of MR studies.

Two-sample MR approach with genetic determinants of iron homeostasis biomarkers (ferritin, serum iron, total iron-binding capacity (TIBC), transferrin saturation) and other exposures (e.g., glycaemic traits, liability to T2D, fibroblast growth factor 23 (FGF23)) using relevant summary statistics obtained from genome-wide association studies (GWAS) (n≤933,970) was used in this thesis. The main analysis was inverse-variance weighting, with sensitivity analyses (e.g., MR-Egger, weighted median) to assess the robustness of the findings, with two-step MR used for mediation analysis and use of cis-variants and colocalization for drug target analysis.

In a bidirectional two-sample MR study assessing the relation between iron homeostasis biomarkers and T2D and glycaemic traits, it was shown fasting glucose, fasting insulin (FI) and genetic liability to T2D may have a positive impact on TIBC, serum iron, and ferritin, respectively, albeit with no mediation by hepcidin, the primary iron homeostasis regulator. Using two-step MR to assess the role of iron homeostasis biomarkers in the associations of liability to T2D and glycaemic traits with liver steatosis and related outcomes, it was shown ferritin mediated the positive association of FI with liver steatosis. Using standard two-sample MR, it was shown FGF23, a downstream effect of iron, protected against coronary artery disease (CAD) and T2D. In further assessment using specifically cis-variants as instruments, FGF23 inhibition likely improved estimated glomerular filtration rate based on creatinine (eGFRcrea). FGF23 inhibition also mediated the positive association of ferritin with eGFRcrea, but was not associated with CAD or T2D. However, the evidence from MR was not substantiated by colocalization.

These findings provide genetic evidence that amongst four iron homeostasis biomarkers, ferritin is likely most relevant to NCDs, although it has differential effects on different NCDs. These studies suggest FGF23 inhibition, albeit lacking evidence from colocalization, and iron chelation could be novel interventions to improve kidney function and prevent liver steatosis, respectively. These studies also show that, whilst MR design is a cost-effective way to assess causal relations, careful design is necessary to avoid generating misleading evidence, which is detrimental to public health. Future studies with different designs, described in this thesis, will be helpful in elucidating the role of iron homeostasis biomarkers in NCDs by triangulating the evidence.