Advertisement
  • Loading metrics

Associations of genetically determined iron status across the phenome: A mendelian randomization study

  • Dipender Gill ,

    Roles Conceptualization, Formal analysis, Investigation, Methodology, Visualization, Writing – original draft, Writing – review & editing

    dipender.gill@imperial.ac.uk

    ‡ These authors contributed equally and are joint first authors.

    Affiliation Department of Epidemiology and Biostatistics, School of Public Health, Imperial College London, London, United Kingdom

  • Beben Benyamin ,

    Roles Conceptualization, Formal analysis, Investigation, Methodology, Writing – review & editing

    ‡ These authors contributed equally and are joint first authors.

    Affiliations Australian Centre for Precision Health, University of South Australia, Adelaide, Australia, Institute for Molecular Bioscience, University of Queensland, Brisbane, Australia, South Australian Health and Medical Research Institute, Adelaide, Australia

  • Luke S. P. Moore,

    Roles Investigation, Writing – review & editing

    Affiliations National Institute for Health Research Health Protection Research Unit in Healthcare Associated Infections and Antimicrobial Resistance, Imperial College London, United Kingdom, Chelsea & Westminster NHS Foundation Trust, London, United Kingdom, Imperial Biomedical Research Centre, Imperial College London and Imperial College NHS Healthcare Trust, London, United Kingdom

  • Grace Monori,

    Roles Investigation, Writing – review & editing

    Affiliation Department of Epidemiology and Biostatistics, School of Public Health, Imperial College London, London, United Kingdom

  • Ang Zhou,

    Roles Data curation, Formal analysis, Investigation, Methodology, Writing – review & editing

    Affiliation Australian Centre for Precision Health, University of South Australia, Adelaide, Australia

  • Fotios Koskeridis,

    Roles Data curation, Investigation, Methodology, Writing – review & editing

    Affiliation Department of Hygiene and Epidemiology, University of Ioannina Medical School, Ioannina, Greece

  • Evangelos Evangelou,

    Roles Methodology, Writing – review & editing

    Affiliations Department of Epidemiology and Biostatistics, School of Public Health, Imperial College London, London, United Kingdom, Department of Hygiene and Epidemiology, University of Ioannina Medical School, Ioannina, Greece

  • Mike Laffan,

    Roles Conceptualization, Writing – review & editing

    Affiliation Centre for Haematology, Imperial College London, United Kingdom

  • Ann P. Walker,

    Roles Conceptualization, Writing – review & editing

    Affiliation Population Science & Experimental Medicine, Institute of Cardiovascular Science, University College London, London, United Kingdom

  • Konstantinos K. Tsilidis,

    Roles Methodology, Writing – review & editing

    Affiliations Department of Epidemiology and Biostatistics, School of Public Health, Imperial College London, London, United Kingdom, Department of Hygiene and Epidemiology, University of Ioannina Medical School, Ioannina, Greece

  • Abbas Dehghan,

    Roles Methodology, Supervision, Writing – review & editing

    Affiliations Department of Epidemiology and Biostatistics, School of Public Health, Imperial College London, London, United Kingdom, Medical Research Council-Public Health England Centre for Environment, School of Public Health, Imperial College London, London, United Kingdom, UK Dementia Research Institute, Imperial College London, London, United Kingdom

  • Paul Elliott,

    Roles Investigation, Methodology, Writing – review & editing

    Affiliations Department of Epidemiology and Biostatistics, School of Public Health, Imperial College London, London, United Kingdom, Imperial Biomedical Research Centre, Imperial College London and Imperial College NHS Healthcare Trust, London, United Kingdom, Medical Research Council-Public Health England Centre for Environment, School of Public Health, Imperial College London, London, United Kingdom, UK Dementia Research Institute, Imperial College London, London, United Kingdom, Health Data Research UK-London, London, United Kingdom

  • Elina Hyppönen,

    Roles Conceptualization, Writing – review & editing

    Affiliations Australian Centre for Precision Health, University of South Australia, Adelaide, Australia, South Australian Health and Medical Research Institute, Adelaide, Australia, Population, Policy and Practice, Great Ormond Street Institute of Child Health, University College London, London, United Kingdom

  • Ioanna Tzoulaki

    Roles Conceptualization, Investigation, Methodology, Supervision, Writing – review & editing

    Affiliations Department of Epidemiology and Biostatistics, School of Public Health, Imperial College London, London, United Kingdom, Department of Hygiene and Epidemiology, University of Ioannina Medical School, Ioannina, Greece, Medical Research Council-Public Health England Centre for Environment, School of Public Health, Imperial College London, London, United Kingdom

Associations of genetically determined iron status across the phenome: A mendelian randomization study

  • Dipender Gill, 
  • Beben Benyamin, 
  • Luke S. P. Moore, 
  • Grace Monori, 
  • Ang Zhou, 
  • Fotios Koskeridis, 
  • Evangelos Evangelou, 
  • Mike Laffan, 
  • Ann P. Walker, 
  • Konstantinos K. Tsilidis
PLOS
x

Abstract

Background

Iron is integral to many physiological processes, and variations in its levels, even within the normal range, can have implications for health. The objective of this study was to explore the broad clinical effects of varying iron status.

Methods and findings

Genome-wide association study (GWAS) summary data obtained from 48,972 European individuals (55% female) across 19 cohorts in the Genetics of Iron Status Consortium were used to identify 3 genetic variants (rs1800562 and rs1799945 in the hemochromatosis gene [HFE] and rs855791 in the transmembrane protease serine 6 gene [TMPRSS6]) that associate with increased serum iron, ferritin, and transferrin saturation and decreased transferrin levels, thus serving as instruments for systemic iron status. Phenome-wide association study (PheWAS) of these instruments was performed on 424,439 European individuals (54% female) in the UK Biobank who were aged 40–69 years when recruited from 2006 to 2010, with their genetic data linked to Hospital Episode Statistics (HES) from April, 1995 to March, 2016. Two-sample summary data mendelian randomization (MR) analysis was performed to investigate the effect of varying iron status on outcomes across the human phenome. MR–PheWAS analysis for the 3 iron status genetic instruments was performed separately and then pooled by meta-analysis. Correction was made for testing of multiple correlated phenotypes using a 5% false discovery rate (FDR) threshold. Heterogeneity between MR estimates for different instruments was used to indicate possible bias due to effects of the genetic variants through pathways unrelated to iron status. There were 904 distinct phenotypes included in the MR–PheWAS analyses. After correcting for multiple testing, the 3 genetic instruments for systemic iron status demonstrated consistent evidence of a causal effect of higher iron status on decreasing risk of traits related to anemia (iron deficiency anemia: odds ratio [OR] scaled to a standard deviation [SD] increase in genetically determined serum iron levels 0.72, 95% confidence interval [CI] 0.64–0.81, P = 4 × 10−8) and hypercholesterolemia (hypercholesterolemia: OR 0.88, 95% CI 0.83–0.93, P = 2 × 10−5) and increasing risk of traits related to infection of the skin and related structures (cellulitis and abscess of the leg: OR 1.25, 95% CI 1.10–1.42, P = 6 × 10−4). The main limitations of this study relate to possible bias from pleiotropic effects of the considered genetic variants and misclassification of diagnoses in the HES data. Furthermore, this work only investigated participants with European ancestry, and the findings may not be applicable to other ethnic groups.

Conclusions

Our findings offer novel, to our knowledge, insight into previously unreported effects of iron status, highlighting a potential protective effect of higher iron status on hypercholesterolemia and a detrimental role on risk of skin and skin structure infections. Given the modifiable and variable nature of iron status, these findings warrant further investigation.

Author summary

Why was this study done?

  • Iron has many vital physiological roles, and variations in its levels can have health implications.
  • In this study, we explored the broad clinical effects of varying iron status.

What did the researchers do and find?

  • We used randomly allocated genetic variants related to iron status to study its effects on a broad range of medical outcomes.
  • Genetic variants related to higher iron status were associated with a lower risk of anemia and hypercholesterolemia and a higher risk of skin and skin structure infections.

What do these findings mean?

  • Given that iron status can be modified, the novel, to our knowledge, associations with hypercholesterolemia and skin and skin structure infections could have clinical implications.
  • Further research is required to confirm these associations and explore how they can be incorporated towards improving clinical practice.

Introduction

Iron is a vital nutrient integral to various physiological processes, including metabolism, erythropoiesis, immune function, and cognitive development [1]. Systemic iron status varies considerably, with serum iron having a coefficient of variation of 30.2% in men (mean 21.2 μmol/L, standard deviation [SD] 6.4 μmol/L) and 36.2% in women (18.5 μmol/L, 6.7 μmol/L) [2]. Given the crucial role of iron, deviations in its levels can have notable health implications [1,2]. At the extremes of iron status are iron deficiency and iron overload. Iron deficiency anemia affects 1.2 billion people worldwide and is responsible for 34.7 million years lived with disability per annum [3]. The etiology of iron deficiency includes inadequate dietary iron intake, impaired absorption, increased losses, and increased requirements such as that due to pregnancy [4]. At the other end of the spectrum, iron overload is most commonly attributed to hemochromatosis and iatrogenic iron overload related to blood transfusions, such as in the treatment of refractory anemia or thalassemia [5]. The modifiable nature and wide variation of iron status, both in the healthy population and among individuals with pathologically low or high levels, make the clinical implications of iron status a research priority.

Observational studies into the effects of iron status can be hindered by confounding from unmeasured and unknown environmental factors and reverse causation bias from outcomes that affect iron status. The use of genetic variants related to systemic iron status to study its effects can overcome these limitations because their random allocation during conception minimizes confounding, and their presence from birth prevents reverse causation [6,7]. By studying the effect on iron status related to randomly allocated alleles, such a mendelian randomization (MR) approach has previously been used in targeted analyses to investigate the effect of iron status on risk of Parkinson’s disease, coronary artery disease, and stroke [810]. MR can also be applied to traits across the human phenome, in an agnostic exploration termed MR–phenome-wide association study (MR–PheWAS) [11]. Such analysis allows for the rapid and efficient investigation of potential health implications attributable to varying an exposure of interest (such as systemic iron status in this case) and provides direction for further targeted study [12].

In this work, we performed an MR–PheWAS of iron status using data from the UK Biobank. As instruments to study the effect of varying systemic iron status, we used genetic variants concordantly related to serum iron, ferritin, transferrin, and transferrin saturation in a pattern consistent with an effect on overall iron status [810,13]. Given the pivotal role of iron across various fundamental physiological processes [1,14] and the opportunity to therapeutically modify systemic levels, the aim of this analysis was to identify a set of health outcomes potentially causally related to iron status. This should guide further clinical research directed towards preventing and treating iron-associated disease.

Methods

This study is reported as per the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline (S1 Checklist). Appropriate patient consent and ethical approval were obtained in the original studies from which data for this work were obtained. Although no formal protocol or prospectively documented analysis plan was used in this study, all the main analyses were decided a priori. At the request of the reviewers, only post hoc sensitivity analyses were performed, as described below.

Genetic instruments for systemic iron status

The exposure phenotype of interest was systemic iron status, which can be measured clinically using the serum iron, ferritin, transferrin, and transferrin saturation biomarkers [15]. We selected genetic instruments for systemic iron status that had relations to these 4 biomarkers in a pattern consistent with an effect on overall iron status, increasing serum iron, ferritin, and transferrin saturation and decreasing transferrin levels [8,10,13,15]. A genome-wide association study (GWAS) performed by the Genetics of Iron Status Consortium on 48,972 European subjects (combined Discovery [N = 23,986] and Replication [N = 24,986] cohorts, 55% female) identified 3 such single-nucleotide polymorphisms (SNPs): rs1800562 and rs1799945 in the hemochromatosis (HFE) gene and rs855791 in the transmembrane protease serine 6 (TMPRSS6) gene [2,8,10]. Both the HFE and TMPRSS6 proteins have established roles in maintaining iron homeostasis (S1 Text), and therefore variants in their respective genes make viable instruments for systemic iron status [16]. The two SNPs in the HFE gene were in low linkage disequilibrium (LD r2 < 0.01) when considering combined European populations with the LDlink resource [2,17]. All three of these SNPs have previously been shown to be strong instruments for MR analysis as measured by F-statistics > 10 [8,18] and collectively explain approximately 3.8% of the variation in serum iron [2,8]. Genetic association estimates for the 3 iron status instrument SNPs with the 4 biomarkers of iron status (serum iron, ferritin, transferrin, and transferrin saturation), respectively, are provided in S1 Table.

PheWAS

The PheWAS was performed in the UK Biobank, a prospective cohort study comprising 503,317 individuals aged 40–69 years recruited between 2006 and 2010 [19]. Participants provided blood samples used for genotyping, and their data were linked to Hospital Episode Statistics (HES) from April, 1995 to March, 2016 [19]. PheWAS analysis was restricted to participants of self-reported European descent in order to maintain consistency with the European population used to obtain instruments for systemic iron status. To avoid bias from related individuals, one participant from each pair of relatives was randomly excluded based on a kinship coefficient of >0.0884. We used the International Classification of Diseases (ICD) versions 9 and 10 to identify cases in the HES data, with both incident and prevalent cases included. Self-reported diagnoses were not considered. Diagnoses were aligned to the phecode grouping system in order to optimize identification of clinically relevant phenotypes [20]. Cases were identified as individuals having at least one documented event and controls as individuals with no record of that outcome or its related phecodes [21]. A series of case-control groups were generated for each phecode, and logistic regression analysis was performed for each instrument SNP separately across all phecodes, adjusting for age, sex, genotyping array, and the first 4 genetic principal components. Analysis was limited to phecodes that had 200 or more cases in order to generate improved statistical power for consequent MR analyses (S1 Text) [22,23].

MR

There is no single biomarker for overall iron status [2], and serum iron levels were used to quantify the genetic associations of the instruments with systemic iron status. PheWAS association estimates for each instrument SNP represent the association of 1 copy of the effect allele with the outcome under consideration. MR estimates for each SNP were calculated as the ratio of this with the corresponding association of the same SNP with serum iron levels (i.e., conventional two-sample ratio method MR) to provide an estimate of the risk of that outcome scaled to a 1 SD increase in serum iron [24]. The SD of serum iron across all individuals included in the Genetics of Iron Status Consortium’s GWAS was 6.1 μmol/L [2]. Standard errors were generated using second-order weights (S1 Text) [24]. Inverse-variance weighted (IVW) meta-analysis of MR estimates for all 3 instrument SNPs was performed to derive the overall MR estimate for the effect of iron status on risk of each considered outcome [18,25]. Statistical significance of MR effect estimates across the considered phenotypes was ascertained using the false discovery rate (FDR) method with a 5% threshold to correct for multiple testing of correlated phenotypes [26].

Sensitivity analyses

Pleiotropy in the context of MR refers to the phenomenon in which genetic instruments affect the outcome of interest through pathways that are at least partly independent of the exposure under consideration and is a source of potential bias [7,27]. Heterogeneity in the MR estimates generated by different instrument SNPs beyond that expected by chance can be used to indicate the presence of such pleiotropy [28], and we assess for this in our MR–PheWAS analysis using the Cochran Q test (interpreting P < 0.05 as evidence of heterogeneity and thus pleiotropy). Only outcomes for which there was no evidence of pleiotropy were taken forward. For such outcomes, sex-stratified IVW MR estimates were also obtained using PheWAS results obtained exclusively from genetically male and female individuals, respectively. Furthermore, MR estimates were also scaled to effects on the ferritin, transferrin, and transferrin saturation biomarkers of iron status.

To further investigate the robustness of the findings to possible pleiotropy, the weighted median MR sensitivity analysis was performed. This orders the MR estimates produced by each instrument SNP by their magnitude weighted for their precision and produces an overall MR estimate based on the median value, with standard error estimated by bootstrapping [29]. It is a robust approach when more than half of the information for the analysis is derived from valid instruments [29].

Statistical analysis was undertaken by DG, BB, GM, and AZ using the software R (version 3.4.2; The R Foundation for Statistical Computing, Vienna, Austria). The TwoSampleMR package was used to facilitate the weighted median MR analysis [30].

Results

Descriptive characteristics of the UK Biobank participants included in PheWAS analyses, along with the number of phenotypes and cases considered in each disease category, are provided in Tables 1 and 2. Results of the PheWAS and MR–PheWAS for each instrument SNP are provided in S2S4 Tables, together with the number of cases and controls available for each outcome. After performing exclusions for related and non-European participants, 424,439 individuals were included in the PheWAS analyses, with genetic association estimates for all 3 instrument SNPs available for 904 distinct phecodes. The IVW meta-analysis pooled MR estimates are given in S5 Table, with results of the Cochran Q test for heterogeneity across the 3 SNPs.

thumbnail
Table 1. Descriptive characteristics of the UK Biobank participants (N = 424,439) included in PheWAS analyses.

https://doi.org/10.1371/journal.pmed.1002833.t001

thumbnail
Table 2. The number of phenotypes and cases considered in each disease category.

https://doi.org/10.1371/journal.pmed.1002833.t002

For the 19 outcomes reaching statistical significance at the 5% FDR threshold (P < 1.1 × 10−3), scatter plots representing the SNP–serum iron and SNP–outcome association estimates are shown in S1S5 Figs. Table 3 details the 14 traits for which there was consistent MR evidence (without suggestion of heterogeneity) across the 3 genetic instruments for a causal effect of higher iron status. Consistent results for these traits were obtained when performing the weighted median MR sensitivity analysis (Table 3) or scaling MR estimates to effects on the different biomarkers of iron status (S6 Table). S7 Table provides the MR odds ratio (OR) per 1 SD increase in genetically determined serum iron level, along with results stratified by sex. Similar estimates were obtained when considering males and females separately, with 95% confidence intervals (CIs) overlapping throughout, although there was some possible suggestion that the association with cellulitis outcomes was stronger for men (S7 Table).

thumbnail
Table 3. IVW MR and weighted median MR estimates for the outcomes reaching 5% FDR significance where there was no evidence of heterogeneity as determined by the Cochran Q test.

https://doi.org/10.1371/journal.pmed.1002833.t003

Higher iron status was most negatively associated with risk of acute posthemorrhagic anemia (OR per 1 SD increase in serum iron 0.35, 95% CI 0.19–0.65, P = 1 × 10−3). In the other direction, higher iron status was most positively associated with glossitis (OR 2.64, 95% CI 1.56–4.46, P = 3 × 10−4), followed by cellulitis and abscess of the leg, arm/hand, and foot or toe, which all produced similar estimates, OR 1.25 (95% CI 1.10–1.42, P = 6 × 10−4).

The identified effects broadly fall into three categories in relation to higher iron status—outcomes related to decreased risk of anemia (Fig 1), decreased risk of hypercholesterolemia (Fig 2), and increased risk of skin and soft tissue infections (Fig 3). Additionally, there was an association of higher genetically determined iron status with increased risk of glossitis and lower risk of poisoning by antibiotics, which were more difficult to categorize (S7 Table).

thumbnail
Fig 1. Forest plot of results for traits related to anemia for which there was evidence across the 3 genetic instruments for a causal effect of higher iron status.

The ORs are reported as MR estimates corresponding to 1 SD increase in serum iron. MR, mendelian randomization; OR, odds ratio; SD, standard deviation.

https://doi.org/10.1371/journal.pmed.1002833.g001

thumbnail
Fig 2. Forest plot of results for traits related to hypercholesterolemia for which there was evidence across the 3 genetic instruments for a causal effect of higher iron status.

The ORs are reported as MR estimates corresponding to 1 SD increase in serum iron. MR, mendelian randomization; OR, odds ratio; SD, standard deviation.

https://doi.org/10.1371/journal.pmed.1002833.g002

thumbnail
Fig 3. Forest plot of results for traits related to skin and skin structure infections for which there was evidence across the 3 genetic instruments for a causal effect of higher iron status.

The ORs are reported as MR estimates corresponding to 1 SD increase in serum iron. MR, mendelian randomization; OR, odds ratio; SD, standard deviation.

https://doi.org/10.1371/journal.pmed.1002833.g003

Discussion

In performing what we believe to be the first MR–PheWAS of systemic iron status, we derived evidence of a protective effect of higher iron levels on risk of traits related to anemia, including aplastic anemia, iron deficiency anemia, anemia from other deficiencies, and acute posthemorrhagic anemia. We additionally found evidence of a protective effect of higher iron levels on risk of hypercholesterolemia and cholesterolosis of the gallbladder. In contrast, our MR–PheWAS provided evidence of a detrimental effect of higher iron status on risk of skin and soft tissue infections, including at the hands, arms, toes, feet, and legs.

Consistent with our findings for infection, iron scavenging systems are common among bacteria that infect the skin, such as Staphylococcus aureus [31] and Streptococcus pyogenes [32]. For these organisms, there is a proposed association between iron metabolism and virulence in human disease [31]. As such, host defense mechanisms act to sequester free iron away from invading pathogens [33]. In the context of abnormalities of iron metabolism such as in hemochromatosis, there is evidence of increased susceptibility to bacterial infections [34], with particular bacterial species showing markedly elevated growth in human serum collected following iron supplementation [35]. Therefore, our finding that higher iron status increases risk of skin and skin structure infections is biologically plausible and consistent with previous evidence. Globally, cellulitis accounted for approximately 598,000 disability adjusted life years across all ages in 2017 [36]. Identifying and modifying potential contributory factors for cellulitis is a global health priority and would potentially serve to decrease antimicrobial use and consequent resistance.

Our findings for higher iron status reducing risk of hypercholesterolemia are also of considerable clinical relevance, with over a third of adult US citizens estimated in 2005–2012 to have cholesterol levels that fall above recommended levels [37], increasing risk of morbidity and mortality related to cardiovascular disease. Similarly, we find evidence that higher iron status decreases risk of gallbladder cholesterolosis, a condition related to the buildup of cholesteryl esters [38]. The HFE rs1800562 variant that we use as an instrument for systemic iron status has previously been associated with low-density lipoprotein cholesterol in GWAS meta-analysis [39]. The mechanism underlying this relationship may include effects related to the HFE protein, other variants in close proximity to rs1800562, or systemic iron status. The consistent evidence we identify across two genetic variants of HFE (rs1800562 and rs1799945) and one variant of TMPRSS6 (rs855791) for a causal effect of higher iron status on lowering risk of both hypercholesterolemia and gallbladder cholesterolosis implicates a mechanism in which high iron status more generally affects cholesterol synthesis. In keeping with this, iron status has previously been suggested to affect lipid metabolism in both rats and humans [40,41]. Further work is required to unravel the mechanistic details of any such effect, and the MR technique may be used to investigate the effect of systemic iron status on different lipid fractions, for example.

Previous work has taken an MR approach to investigate the association of genetic variants related to hereditary hemochromatosis with risk of 11 outcomes that are implicated in iron overload [42]. However, the new, to our knowledge, contribution of our current study is that it performed a hypothesis-free investigation into the causal effects of iron status more widely across the human phenome and thus allowed for the identification of novel, to our knowledge, associations, namely potential effects on risk of cellulitis and hypercholesterolemia. Furthermore, our study pooled MR estimates obtained using all available genetic instruments for iron status, rather than focusing on those related to hereditary hemochromatosis [42], and thus better allowed us to investigate the effects of variation in iron status through any cause. The MR methodology used is less vulnerable to environmental confounding and reverse causation than traditional observational research strategies [6,7]. Using genetic variants that are randomly allocated at conception to instrument the effect of modifying systemic iron status [6], we estimated the cumulative lifetime effects of genetically determined variation across over 900 disease outcomes. Our results for higher iron status protecting against iron deficiency anemia and acute posthemorrhagic anemia support the validity of our methodological approach because the role of iron in generating hemoglobin is well-established and associations of our genetic instruments for systemic iron status with red blood cell traits have previously been described [2,8]. A major challenge to such an MR approach is deciphering effects that are attributable to bias related to pleiotropic variants, which we addressed by considering the heterogeneity in the MR estimates produced by our different instrument SNPs [27,43] and in statistical sensitivity analysis using the weighted median MR approach, which is more robust to the inclusion of pleiotropic variants [29]. Furthermore, by using strong instruments associated with serum iron, ferritin, transferrin, and transferrin saturation in a pattern consistent with their effect on systemic iron status, our analysis aims to reflect the effects of varying iron exposure per se rather than that of some other associated traits [8,10].

Limitations of our approach include the use of HES data, which offered a rich source of clinical outcomes that were linked to the genetic data of UK Biobank participants but possibly also introduced misclassification bias [44]. For example, it is unclear whether the observed protective effect of higher iron status on risk of aplastic anemia is attributable to a misclassification of iron deficiency anemia. Similarly, in a scenario in which iron status may not be a cause of aplastic anemia, it may still contribute to diagnosis by shifting borderline cases beyond the requisite threshold for disease label allocation. Our finding of an increased risk of glossitis with higher iron status contradicts the established protective effect of iron on atrophic glossitis, and it is uncertain whether the diagnosis in this context is related to atrophy [45] or superimposed infection, as would be in keeping with our findings for the effect of iron status on risk of superficial infections. For the observed protective effect of higher iron status on risk of hypercholesterolemia (as defined in HES data), the lack of serum cholesterol measures in UK Biobank meant that it was not possible to quantify the magnitude of effect on actual cholesterol levels nor the particular lipid fractions to which this related. Furthermore, ascertainment bias may be responsible for our finding related to poisoning from antibiotics because individuals with higher iron status appear to be more likely to develop particular types of infection (such as those related to the skin and soft tissue), thus potentially also affecting the spectrum of antibiotics to which they are exposed. Finally, inadequate statistical power may have also resulted in false negative results in our MR–PheWAS. The previously described MR effects of iron status on risk of Parkinson’s disease and coronary artery disease (coronary atherosclerosis) were not statistically significant after correcting for multiple testing in our current analysis, although the directions of effect were consistent with previous work (S5 Table) [8,9]. Similarly, type II error and false negative results may have also arisen because of exclusion of all results that evidenced heterogeneity in the MR estimates from different instrument SNPs when attempting to avoid bias from pleiotropy.

Interventions are available to manipulate systemic iron status. The efficacy of oral iron replacement is limited by low gastrointestinal absorption [46], alongside side effects such as abdominal pain, nausea, and constipation that affect around a third of patients and reduce compliance [46,47]. At a population level, fortification of foods with iron-containing micronutrient powders has been efficacious for treating iron deficiency anemia [48, 49]. Anemia refractory to oral supplementation or severe anemia may be managed with intravenous iron infusion [46]. In contrast, iron overload is primarily treated with venesection in hemochromatosis patients [50], with use of iron chelation to enhance iron excretion an additional option [51]. Clinical trials and guidance on the manipulation of iron status have most often related to treatment of anemia [46], such as in the context of menstruation [52], pregnancy [53], and chronic kidney disease [54]. Weaker and more limited evidence exists on the titration of iron status as a treatment for other clinical outcomes, including acute stroke [55], malaria [56], and restless leg syndrome [57]. However, no trial has so far addressed the effect of manipulating iron status to prevent or treat skin and skin structure infections. Similarly, the possibility of targeting iron status to optimize lipid metabolism has not been explored. Given the findings of our study, further research might focus on the degree to which iron status can be titrated in both the prevention and treatment of disease. However, caution must be taken when extrapolating the findings of such MR analyses, particularly because their estimates relate to small variations in iron status within the normal range rather than at extremes of iron deficiency or overload.

In conclusion, this study used MR to explore the effect of iron status across the human phenome and identified a number of novel, to our knowledge, clinically relevant results. Cellulitis and hypercholesterolemia are widespread and of notable significance. Given that iron status is a modifiable trait, further work is warranted to validate our findings, investigate possible underlying mechanisms, and explore whether directed manipulation of iron levels can be used to optimize health outcomes.

Supporting information

S1 Checklist. STROBE checklist.

STROBE, Strengthening the Reporting of Observational Studies in Epidemiology.

https://doi.org/10.1371/journal.pmed.1002833.s001

(DOCX)

S1 Text. Roles of HFE and TMPRSS6 in maintaining iron status, statistical power calculation, and second-order weight details.

HFE, hemochromatosis gene; TMPRSS6, transmembrane protease serine 6 gene.

https://doi.org/10.1371/journal.pmed.1002833.s002

(DOCX)

S1 Table. Genetic association estimates for the 3 iron status instruments with the 4 biomarkers of systemic iron status [2].

https://doi.org/10.1371/journal.pmed.1002833.s003

(XLSX)

S2 Table. PheWAS results for rs1800562.

All estimates are in log OR units. OR, odds ratio; PheWAS, phenome-wide association study.

https://doi.org/10.1371/journal.pmed.1002833.s004

(XLSX)

S3 Table. PheWAS results for rs1799945.

All estimates are in log OR units. OR, odds ratio; PheWAS, phenome-wide association study.

https://doi.org/10.1371/journal.pmed.1002833.s005

(XLSX)

S4 Table. PheWAS results for rs855791.

All estimates are in log OR units. OR, odds ratio; PheWAS, phenome-wide association study.

https://doi.org/10.1371/journal.pmed.1002833.s006

(XLSX)

S5 Table. IVW MR estimates for all outcomes.

The minimum number of cases and controls for any of the 3 genetics instruments are given. IVW, inverse-variance weighted; MR, mendelian randomization.

https://doi.org/10.1371/journal.pmed.1002833.s007

(XLSX)

S6 Table. Comparison of IVW MR estimates when scaled to 1 SD increase in serum iron, (log10 transformed) ferritin, transferrin saturation, and transferrin.

Higher transferrin levels reflect lower iron status. The minimum number of cases and controls for any of the 3 genetics instruments are given. MR estimates are given in log OR units. IVW, inverse-variance weighted; MR, mendelian randomization; OR, odds ratio; SD, standard deviation.

https://doi.org/10.1371/journal.pmed.1002833.s008

(XLSX)

S7 Table. Combined and sex-stratified MR estimates for the effect of iron status.

MR estimates are given per SD change in serum iron. MR, mendelian randomization; SD, standard deviation.

https://doi.org/10.1371/journal.pmed.1002833.s009

(XLSX)

S1 Fig. Scatter plot of instrument genetic association estimates for traits related to anemia.

https://doi.org/10.1371/journal.pmed.1002833.s010

(TIFF)

S2 Fig. Scatter plot of instrument genetic association estimates for traits related to hypercholesterolemia.

https://doi.org/10.1371/journal.pmed.1002833.s011

(TIFF)

S3 Fig. Scatter plot of instrument genetic association estimates for traits related to skin and skin structure infections.

https://doi.org/10.1371/journal.pmed.1002833.s012

(TIFF)

S4 Fig. Scatter plot of instrument genetic association estimates for the poisoning by antibiotics and glossitis traits.

https://doi.org/10.1371/journal.pmed.1002833.s013

(TIFF)

S5 Fig. Scatter plot of instrument genetic association estimates for the traits in which there was evidence of heterogeneity between instruments.

https://doi.org/10.1371/journal.pmed.1002833.s014

(TIFF)

Acknowledgments

The authors acknowledge the Genetics of Iron Status Consortium and the UK Biobank and their participants for contributing the data used in this work. The authors also acknowledge Joshua Denny and Evropi Theodoratou for the provision of phecode linkage to UK Biobank Hospital Episode Statistics data.

References

  1. 1. Abbaspour N, Hurrell R, Kelishadi R. Review on iron and its importance for human health. J Res Med Sci. 2014;19(2):164–74. pmid:24778671; PubMed Central PMCID: PMCPMC3999603.
  2. 2. Benyamin B, Esko T, Ried JS, Radhakrishnan A, Vermeulen SH, Traglia M, et al. Novel loci affecting iron homeostasis and their effects in individuals at risk for hemochromatosis. Nat Commun. 2014;5:4926. Epub 2014/10/29. pmid:25352340; PubMed Central PMCID: PMCPMC4215164.
  3. 3. GBD 2015 Disease and Injury Incidence and Prevalence Collaborators. Global, regional, and national incidence, prevalence, and years lived with disability for 310 diseases and injuries, 1990–2015: a systematic analysis for the Global Burden of Disease Study 2015. Lancet. 2016;388(10053):1545–602. pmid:27733282; PubMed Central PMCID: PMCPMC5055577.
  4. 4. DeLoughery TG. Iron Deficiency Anemia. Med Clin North Am. 2017;101(2):319–32. Epub 2016/12/08. pmid:28189173.
  5. 5. Sebastiani G, Pantopoulos K. Disorders associated with systemic or local iron overload: from pathophysiology to clinical practice. Metallomics. 2011;3(10):971–86. Epub 2011/09/07. pmid:21901209.
  6. 6. Davey Smith G, Ebrahim S. What can mendelian randomisation tell us about modifiable behavioural and environmental exposures? BMJ. 2005;330(7499):1076–9. pmid:15879400; PubMed Central PMCID: PMCPMC557238.
  7. 7. Davies NM, Holmes MV, Davey Smith G. Reading Mendelian randomisation studies: a guide, glossary, and checklist for clinicians. BMJ. 2018;362:k601. pmid:30002074; PubMed Central PMCID: PMCPMC6041728.
  8. 8. Gill D, Del Greco M F, Walker AP, Srai SKS, Laffan MA, Minelli C. The effect of iron status on risk of coronary artery disease: a mendelian randomization study. Arterioscler Thromb Vasc Biol. 2017;37(9):1788–92. pmid:28684612.
  9. 9. Pichler I, Del Greco MF, Gogele M, Lill CM, Bertram L, Do CB, et al. Serum iron levels and the risk of Parkinson disease: a Mendelian randomization study. PLoS Med. 2013;10(6):e1001462. pmid:23750121; PubMed Central PMCID: PMCPMC3672214.
  10. 10. Gill D, Monori G, Tzoulaki I, Dehghan A. Iron Status and Risk of Stroke. Stroke. 2018;49(12):2815–21. pmid:30571402
  11. 11. Millard LAC, Davies NM, Timpson NJ, Tilling K, Flach PA, Smith GD. MR-PheWAS: hypothesis prioritization among potential causal effects of body mass index on many outcomes, using Mendelian randomization. Sci Rep. 2015;5:16645. Artn 16645 pmid:26568383
  12. 12. Denny JC, Bastarache L, Ritchie MD, Carroll RJ, Zink R, Mosley JD, et al. Systematic comparison of phenome-wide association study of electronic medical record data and genome-wide association study data. Nat Biotechnol. 2013;31(12):1102–10. pmid:24270849; PubMed Central PMCID: PMCPMC3969265.
  13. 13. Taylor AE, Davies NM, Ware JJ, VanderWeele T, Smith GD, Munafo MR. Mendelian randomization in health research: using appropriate genetic variants and avoiding biased estimates. Econ Hum Biol. 2014;13:99–106. Epub 2014/01/07. pmid:24388127; PubMed Central PMCID: PMCPMC3989031.
  14. 14. Puig S, Askeland E, Thiele DJ. Coordinated remodeling of cellular metabolism during iron deficiency through targeted mRNA degradation. Cell. 2005;120(1):99–110. WOS:000226365400014. pmid:15652485
  15. 15. Wish JB. Assessing iron status: beyond serum ferritin and transferrin saturation. Clin J Am Soc Nephrol. 2006;1 Suppl 1:S4–8. pmid:17699374.
  16. 16. Swerdlow DI, Kuchenbaecker KB, Shah S, Sofat R, Holmes MV, White J, et al. Selecting instruments for Mendelian randomization in the wake of genome-wide association studies. Int J Epidemiol. 2016;45(5):1600–16. pmid:27342221; PubMed Central PMCID: PMCPMC5100611.
  17. 17. Machiela MJ, Chanock SJ. LDlink: a web-based application for exploring population-specific haplotype structure and linking correlated alleles of possible functional variants. Bioinformatics. 2015;31(21):3555–7. pmid:26139635; PubMed Central PMCID: PMCPMC4626747.
  18. 18. Palmer TM, Lawlor DA, Harbord RM, Sheehan NA, Tobias JH, Timpson NJ, et al. Using multiple genetic variants as instrumental variables for modifiable risk factors. Stat Methods Med Res. 2012;21(3):223–42. pmid:21216802; PubMed Central PMCID: PMCPMC3917707.
  19. 19. Sudlow C, Gallacher J, Allen N, Beral V, Burton P, Danesh J, et al. UK Biobank: An Open Access Resource for Identifying the Causes of a Wide Range of Complex Diseases of Middle and Old Age. PLoS Med. 2015;12(3):e1001779. ARTN e1001779 WOS:000352096300001. pmid:25826379
  20. 20. Wei WQ, Bastarache LA, Carroll RJ, Marlo JE, Osterman TJ, Gamazon ER, et al. Evaluating phecodes, clinical classification software, and ICD-9-CM codes for phenome-wide association studies in the electronic health record. PLoS ONE. 2017;12(7):e0175508. pmid:28686612; PubMed Central PMCID: PMCPMC5501393.
  21. 21. Li X, Meng XR, Spiliopoulou A, Timofeeva M, Wei WQ, Gifford A, et al. MR-PheWAS: exploring the causal effect of SUA level on multiple disease outcomes by using genetic instruments in UK Biobank. Ann Rheum Dis. 2018;77(7):1039–47. WOS:000438037000026. pmid:29437585
  22. 22. Verma A, Bradford Y, Dudek S, Lucas AM, Verma SS, Pendergrass SA, et al. A simulation study investigating power estimates in phenome-wide association studies. BMC Bioinformatics. 2018;19(1):120. pmid:29618318; PubMed Central PMCID: PMCPMC5885318.
  23. 23. Brion MJ, Shakhbazov K, Visscher PM. Calculating statistical power in Mendelian randomization studies. Int J Epidemiol. 2013;42(5):1497–501. pmid:24159078; PubMed Central PMCID: PMC3807619.
  24. 24. Burgess S, Butterworth A, Thompson SG. Mendelian randomization analysis with multiple genetic variants using summarized data. Genet Epidemiol. 2013;37(7):658–65. pmid:24114802; PubMed Central PMCID: PMC4377079.
  25. 25. Thompson JR, Minelli C, Del Greco M F. Mendelian Randomization using Public Data from Genetic Consortia. Int J Biostat. 2016;12(2):/j/ijb.2016.12.issue-2/ijb-2015-0074/ijb-2015-0074.xml. pmid:27092657.
  26. 26. Benjamini Y, Hochberg Y. Controlling the False Discovery Rate—a Practical and Powerful Approach to Multiple Testing. J Roy Stat Soc B Met. 1995;57(1):289–300. WOS:A1995QE45300017.
  27. 27. Del Greco M F, Minelli C, Sheehan NA, Thompson JR. Detecting pleiotropy in Mendelian randomisation studies with summary data and a continuous outcome. Stat Med. 2015;34(21):2926–40. pmid:25950993.
  28. 28. Burgess S, Bowden J, Fall T, Ingelsson E, Thompson SG. Sensitivity Analyses for Robust Causal Inference from Mendelian Randomization Analyses with Multiple Genetic Variants. Epidemiology. 2017;28(1):30–42. pmid:27749700; PubMed Central PMCID: PMC5133381.
  29. 29. Bowden J, Davey Smith G, Haycock PC, Burgess S. Consistent Estimation in Mendelian Randomization with Some Invalid Instruments Using a Weighted Median Estimator. Genet Epidemiol. 2016;40(4):304–14. pmid:27061298; PubMed Central PMCID: PMC4849733.
  30. 30. Hemani G, Zheng J, Elsworth B, Wade KH, Haberland V, Baird D, et al. The MR-Base platform supports systematic causal inference across the human phenome. eLife. 2018;7:e34408. pmid:29846171; PubMed Central PMCID: PMCPMC5976434.
  31. 31. Dale SE, Doherty-Kirby A, Lajoie G, Heinrichs DE. Role of siderophore biosynthesis in virulence of Staphylococcus aureus: identification and characterization of genes involved in production of a siderophore. Infect Immun. 2004;72(1):29–37. pmid:14688077; PubMed Central PMCID: PMCPMC343950.
  32. 32. Bates CS, Montanez GE, Woods CR, Vincent RM, Eichenbaum Z. Identification and characterization of a Streptococcus pyogenes operon involved in binding of hemoproteins and acquisition of iron. Infect Immun. 2003;71(3):1042–55. pmid:12595414; PubMed Central PMCID: PMCPMC148835.
  33. 33. Parrow NL, Fleming RE, Minnick MF. Sequestration and scavenging of iron in infection. Infect Immun. 2013;81(10):3503–14. pmid:23836822; PubMed Central PMCID: PMCPMC3811770.
  34. 34. Khan FA, Fisher MA, Khakoo RA. Association of hemochromatosis with infectious diseases: expanding spectrum. Int J Infect Dis. 2007;11(6):482–7. pmid:17600748.
  35. 35. Cross JH, Bradbury RS, Fulford AJ, Jallow AT, Wegmuller R, Prentice AM, et al. Oral iron acutely elevates bacterial growth in human serum. Sci Rep. 2015;5:16670. pmid:26593732; PubMed Central PMCID: PMCPMC4655407.
  36. 36. GBD 2017 DALYs and HALE Collaborators. Global, regional, and national disability-adjusted life-years (DALYs) for 359 diseases and injuries and healthy life expectancy (HALE) for 195 countries and territories, 1990–2017: a systematic analysis for the Global Burden of Disease Study 2017. Lancet. 2018;392(10159):1859–922. pmid:30415748.
  37. 37. Mercado C, DeSimone AK, Odom E, Gillespie C, Ayala C, Loustalot F. Prevalence of Cholesterol Treatment Eligibility and Medication Use Among Adults—United States, 2005–2012. MMWR. 2015;64(47):1305–11. WOS:000366102500001. pmid:26633047
  38. 38. Koga A. Fine-Structure of the Human Gallbladder with Cholesterosis with Special Reference to the Mechanism of Lipid-Accumulation. Brit J Exp Pathol. 1985;66(5):605–11. WOS:A1985ATG6700012.
  39. 39. Teslovich TM, Musunuru K, Smith AV, Edmondson AC, Stylianou IM, Koseki M, et al. Biological, clinical and population relevance of 95 loci for blood lipids. Nature. 2010;466(7307):707–13. WOS:000280562500029. pmid:20686565
  40. 40. Stangl GI, Kirchgessner M. Different degrees of moderate iron deficiency modulate lipid metabolism of rats. Lipids. 1998;33(9):889–95. pmid:9778136.
  41. 41. Ahmed U, Latham PS, Oates PS. Interactions between hepatic iron and lipid metabolism with possible relevance to steatohepatitis. World J Gastroenterol. 2012;18(34):4651–8. pmid:23002334; PubMed Central PMCID: PMCPMC3442203.
  42. 42. Pilling LC, Tamosauskaite J, Jones G, Wood AR, Jones L, Kuo CL, et al. Common conditions associated with hereditary haemochromatosis genetic variants: cohort study in UK Biobank. BMJ. 2019;364:k5222. pmid:30651232; PubMed Central PMCID: PMCPMC6334179.
  43. 43. Bowden J, Hemani G, Davey Smith G. Detecting individual and global horizontal pleiotropy in Mendelian randomization: a job for the humble heterogeneity statistic? Am J Epidemiol. 2018;187(12):2681–5. pmid:30188969.
  44. 44. Padmanabhan S, Carty L, Cameron E, Ghosh RE, Williams R, Strongman H. Approach to record linkage of primary care data from Clinical Practice Research Datalink to other health-related patient data: overview and implications. Eur J Epidemiol. 2019;34(1):91–9. Epub 2018 Sep 15. pmid:30219957.
  45. 45. Wu YC, Wang YP, Chang JY, Cheng SJ, Chen HM, Sun A. Oral manifestations and blood profile in patients with iron deficiency anemia. J Formos Med Assoc. 2014;113(2):83–7. pmid:24388269.
  46. 46. Jimenez K, Kulnigg-Dabsch S, Gasche C. Management of Iron Deficiency Anemia. Gastroenterol Hepatol (N Y). 2015;11(4):241–50. pmid:27099596; PubMed Central PMCID: PMCPMC4836595.
  47. 47. Cancelo-Hidalgo MJ, Castelo-Branco C, Palacios S, Haya-Palazuelos J, Ciria-Recasens M, Manasanch J, et al. Tolerability of different oral iron supplements: a systematic review. Curr Med Res Opin. 2013;29(4):291–303. Epub 2013/02/06. pmid:23252877.
  48. 48. De-Regil LM, Jefferds MED, Peña-Rosas JP. Point-of-use fortification of foods with micronutrient powders containing iron in children of preschool and school-age. Cochrane Database Syst Rev. 2017;11:CD009666. Epub 2017/11/23. pmid:29168569.
  49. 49. Arcanjo FPN, da Costa Rocha TC, Arcanjo CPC, Santos PR. Micronutrient Fortification at Child-Care Centers Reduces Anemia in Young Children. J Diet Suppl. 2018:1–10. Epub 2018/06/29. pmid:29958044.
  50. 50. Assi TB, Baz E. Current applications of therapeutic phlebotomy. Blood Transfus. 2014;12 Suppl 1:s75–83. Epub 2013/10/03. pmid:24120605; PubMed Central PMCID: PMCPMC3934278.
  51. 51. Poggiali E, Cassinerio E, Zanaboni L, Cappellini MD. An update on iron chelation therapy. Blood Transfus. 2012;10(4):411–22. Epub 2012/06/27. pmid:22790257; PubMed Central PMCID: PMCPMC3496216.
  52. 52. Low MS, Speedy J, Styles CE, De-Regil LM, Pasricha SR. Daily iron supplementation for improving anaemia, iron status and health in menstruating women. Cochrane Database Syst Rev. 2016;4:CD009747. pmid:27087396.
  53. 53. Pena-Rosas JP, De-Regil LM, Garcia-Casal MN, Dowswell T. Daily oral iron supplementation during pregnancy. Cochrane Database Syst Rev. 2015;(7):CD004736. pmid:26198451.
  54. 54. Padhi S, Glen J, Pordes BA, Thomas ME, Guideline Development G. Management of anaemia in chronic kidney disease: summary of updated NICE guidance. BMJ. 2015;350:h2258. pmid:26044132.
  55. 55. Ma J, You C, Hao L. Iron chelators for acute stroke. Cochrane Database Syst Rev. 2012;(9):CD009280. pmid:22972139.
  56. 56. Neuberger A, Okebe J, Yahav D, Paul M. Oral iron supplements for children in malaria-endemic areas. Cochrane Database Syst Rev. 2016;2:CD006589. pmid:26921618; PubMed Central PMCID: PMCPMC4916933.
  57. 57. Trotti LM, Bhadriraju S, Becker LA. Iron for restless legs syndrome. Cochrane Database Syst Rev. 2012;(5):CD007834. pmid:22592724; PubMed Central PMCID: PMCPMC4070449.