Although the statistical evidence to support the claim that the HFE C282Y polymorphism modifies the association between traffic-related air pollution and cognitive function in our sample of older men is weak, the pattern of response across groups is intriguing and suggests the presence of effect modification by HFE C282Y. Participants possessing a variant HFE C282Y polymorphism appear to be protected from the adverse association between black carbon and cognition, while participants who lacked the C282Y variant exhibit strong adverse associations. We found little support for modification of the association between traffic-related air pollution and cognition by HFE H63D.
This is the first study to investigate effect modification by HFE of the air pollution-cognition relationship. These findings are in line with reported effect modification by HFE of the relationship between exposure to air pollution and cardiovascular outcomes or risk factors in the NAS cohort. Persons lacking HFE variant genotypes exhibit strong adverse associations between exposure to fine particulate matter (PM) and heart rate variability, while there is little evidence for an association among those possessing an HFE variant, and this pattern was driven primarily by the HFE C282Y SNP
. Similarly, higher exposures to black carbon or PM2.5 were associated with higher plasma homocysteine concentrations in participants who lacked the HFE C282Y variant, but no association was observed in participants possessing at least one HFE C282Y variant
. A similar pattern was observed for modification by HFE C282Y of the association between traffic-related air pollution and heart-rate-corrected (QT) interval, an electrocardiographic marker of ventricular repolarization that is associated with arrhythmia and cardiac death
The mechanism by which exposure to traffic-related air pollution may contribute to cognitive impairment is unknown, but likely involves inhalation of particulates. In animal experiments, exposure to fine or ultrafine particulate matter appears to induce inflammation, lipid peroxidation, and neuronal degeneration in the central nervous system
[25–28]. Metals, including iron, are often found adsorbed onto the surface of traffic-related particulates
 and exposure to metals is associated with induction of oxidative stress and inflammation
. Oxidative stress and inflammation may contribute to the development of cognitive impairment directly or through inducement of cardiovascular problems.
HFE modulates intracellular uptake of iron and other metals. Variants of HFE are associated with the disease hemochromatosis, which is characterized by iron overload due to excessive uptake of iron from the gastrointestinal tract
. Studies of HFE modulation of the health effects of air pollution, including the current study, generally find that variants in HFE protect against the adverse impact of air pollution exposure
[19, 20]. These findings may be attributable to HFE modulation of uptake of metals from the lung. Exposure to traffic-related air pollution occurs primarily through inhalation. In order for traffic-related air pollution to adversely impact human health, some component of traffic-related air pollution must enter the body, either through penetration of the respiratory epithelium, ingestion, or through translocation up the olfactory nerve. Variants of HFE C282Y and H63D are associated with increased iron uptake from the gut, and higher body stores of iron may down-regulate overall metal absorption, including absorption from the lung of metals adhered to inhaled air pollution. In animal models, rats fed a high-iron diet or exposed to iron oxide exhibited less absorption of manganese after intratracheal installation compared to those fed a regular diet
[17, 18]. This reduction in pulmonary absorption was paralleled by lower uptake by other organs, including the brain
. Therefore, HFE genotype, associated iron levels, and iron homeostatic mechanisms could influence the toxicity of exposure to traffic-related air pollution through modulation of metal uptake. Lower uptake of adsorbed metals would be expected to reduce the associated systemic and brain-based oxidative stress and inflammation and lessen any adverse effect of air pollution on cognition. This mechanistic hypothesis might explain why we did not see an adverse association between air pollution and cognition in carriers of the HFE C282Y variant.
We also note that the evidence to support effect modification of the association between air pollution and cognition was stronger when we considered the C282Y variant than when we considered the H63D variant, which is consistent with the relative degree of function of these SNPs. The HFE protein produced by the C282Y variant, but not the H63D variant protein, is often degraded before final processing and is not expressed on the cell surface
 and does not associate with transferrin
. As such, the C282Y variant is associated with greater loss of function compared to the H63D variant, resulting in increased iron transfer in the gut. In this case, we may expect lower respiratory intake of adsorbed metals, and therefore lower levels of air pollution-related oxidative stress and inflammation, in those with the C282Y variant than in those with the H63D variant.
We acknowledge that our study has several limitations. Use of BC exposure estimates based on residential address may misclassify long-term exposure. However, given the age and residential stability of our cohort, misclassification of our exposure due to occupation, commuting, or relocation is expected to be minimal and largely unrelated to cognition or HFE status, suggesting bias towards the null. Non-differential misclassification of cognitive status is likely, but is mitigated by the use of all available cognitive data to consider the association between BC and total metrics of cognition. It is possible that residual confounding may bias our results. However, we were able to adjust for many known predictors of cognition, including several measures of socioeconomic status. Furthermore, we expect confounding by other classes of air pollution to be minimal, as the correlation between traffic-related air pollution and regional pollutants is relatively low due to differences in the spatial-temporal distribution of these types of pollutants. Selection bias is a concern, but because both poor cognition and exposure to traffic-related air pollution predict morbidity and mortality
[33–36], we would expect selection bias to mask the expected adverse association or create a false protective association. While this may contribute to the absence of an adverse association among participants with HFE C282Y variants, it would not account for the strong adverse association observed within those who lacked HFE C282Y variants. While we are unable to concretely attribute our findings to a particular aspect of traffic-related exposure, modification of the BC-cognition association by HFE implicates metals adhered to particulate matter as a likely toxic agent. While population substructure may induce spurious associations in gene-environment interaction studies, this is unlikely to account for our findings given our homogeneous study sample of white men. Finally, our sample size and the number of participants with the variant HFE alleles are small, which makes it difficult to detect a subtle effect and increases the possibility that our findings are due to chance. This analysis requires replication in an independent sample.