Article Text

Download PDFPDF

Original article
Association of HFE and TMPRSS6 genetic variants with iron and erythrocyte parameters is only in part dependent on serum hepcidin concentrations
  1. Michela Traglia1,
  2. Domenico Girelli2,
  3. Ginevra Biino3,4,
  4. Natascia Campostrini2,
  5. Michela Corbella2,
  6. Cinzia Sala1,
  7. Corrado Masciullo1,
  8. Fiammetta Viganò1,
  9. Iwan Buetti1,
  10. Giorgio Pistis1,
  11. Massimiliano Cocca1,
  12. Clara Camaschella1,5,
  13. Daniela Toniolo1,3
  1. 1Division of Genetics and Cell Biology, San Raffaele Research Institute, Milano, Italy
  2. 2Department of Medicine, University of Verona, Verona, Italy
  3. 3Institute of Molecular Genetics, CNR, Pavia, Italy
  4. 4Institute of Population Genetics, CNR, Sassari, Italy
  5. 5Vita Salute University, San Raffaele Scientific Institute, Milano, Italy
  1. Correspondence to Daniela Toniolo, Division of Genetics and Cell Biology, San Raffaele Research Institute, Via Olgettina 58, Milano 20132, Italy; daniela.toniolo{at}


Background Hepcidin is the main regulator of iron homeostasis: inappropriate production of hepcidin results in iron overload or iron deficiency and anaemia.

Aims To study variation of serum hepcidin concentration in a normal population.

Results Hepcidin showed age and sex dependent variations that correlated with ferritin but not with serum iron and transferrin saturation. The size of the study population was underpowered to find genome wide significant associations with hepcidin concentrations but it allowed to show that association with serum iron, transferrin saturation and erythrocyte traits of common DNA variants in HFE (rs1800562) and TMPRSS6 (rs855791) genes is not exclusively dependent on hepcidin values. When multiple interactions between environmental factors, the iron parameters and hepcidin were taken into account, the HFE variant, and to lesser extent the TMPRSS6 variant, were associated with ferritin and with hepcidin normalised to ferritin (the hepcidin/ferritin ratio).

Conclusions The results suggest a mutual control of serum hepcidin and ferritin concentrations, a mechanism relevant to the pathophysiology of HFE haemochromatosis, and demonstrate that the HFE rs1800562 C282Y variant exerts a direct pleiotropic effect on the iron parameters, in part independent of hepcidin.

View Full Text

Statistics from


Iron is essential for multiple biological functions in all tissues, but especially for haemoglobin synthesis, as shown by anaemia that results from iron deficiency. Excess iron is toxic, because it favours oxidative stress and cell damage.1 For this reason, the amount of plasma iron is maintained within narrow limits and is tightly regulated by the liver peptide hepcidin according to the body's needs.2 Hepcidin controls the surface expression of the iron exporter ferroportin on enterocytes and iron recycling macrophages.3 Genetic disorders of the hepcidin–ferroportin pathway lead to opposite conditions. Haemochromatosis is caused by mutations in genes which encode upstream hepcidin activating proteins (HFE, TFR2, hemojuvelin) or mutations in hepcidin itself. All these forms of haemochromatosis are characterised by inappropriately high iron absorption, elevated transferrin saturation and serum ferritin, and inappropriately low/undetectable hepcidin expression.4 More rarely mutations affect ferroportin, the downstream target of hepcidin; as a consequence, either iron is not recycled and remains sequestered in macrophages or the mutant is not internalised because is hepcidin resistant.5 Iron refractory, iron deficiency anaemia (IRIDA) is caused by mutations of TMPRSS6, which encodes the liver expressed hepcidin inhibitor serine protease matriptase-2. Mask mice homozygous for a truncated matriptase-2 lacking the serine protease domain,6 Tmprss6 null mice,7 and patients with IRIDA8 do not efficiently absorb oral iron because they are unable to fully suppress hepcidin activation.9 10 They display very low transferrin saturations but moderately decreased serum ferritin because of iron retention in macrophage stores.11

Genetic variants of two of the hepcidin regulatory genes, TMPRSS6 and HFE, affect serum iron concentration12 and transferrin saturation13 14 in normal populations. Furthermore, single nucleotide polymorphisms (SNPs) at TMPRSS6 and HFE loci were found to be associated with quantitative variations of haemoglobin (Hb) concentrations and erythrocyte traits.15–18 However, it remains uncertain whether the association is iron mediated or dependent on a direct effect of the variants on erythropoiesis. In addition, the effect of the ‘iron gene’ variants was ascribed to variations in hepcidin concentrations,15 16 18 but serum hepcidin was not measured.

We report here the analysis of serum hepcidin concentrations, measured by a mass spectrometry based method,19 in 1657 normal individuals from the Val Borbera (VB) genetic isolate in Northern Italy. We explored relationships between hepcidin and a set of anthropometric, haematologic, and iron parameters and tested the association of two common variants rs1800562 and rs855791 in the HFE and TMPRSS6 genes, respectively, with iron, erythrocyte parameters and hepcidin values in 1545 genotyped individuals. We demonstrate a reciprocal control of serum hepcidin and ferritin concentrations that may be relevant to the pathophysiology of HFE haemochromatosis, and demonstrate that the HFE C282Y variant exerts a direct pleiotropic effect on several of the iron parameters, partly independent of hepcidin.


Study subjects

The VB population has been previously described.20 Only individuals aged 18 years or older were eligible to participate in the study. The study was approved by the San Raffaele Hospital and Regione Piemonte ethical committees. The health status of the population was assessed as reported.20

Blood tests

Fasting blood samples were obtained in the early morning. Blood was analysed the same day or aliquoted and stored at −80°C for further analysis. Blood cell counts and erythrocyte indexes were determined with an automated cell counter.21 Other blood tests, serum iron, transferrin, and ferritin were determined by standard methods. Transferrin saturation was calculated dividing serum iron by transferrin (mg) ×1.42.22

Serum hepcidin assay

Serum hepcidin was measured in all samples with a validated mass spectrometry based method, that is surface enhanced laser desorption/ionisation time-of-flight mass spectrometry (SELDI-TOF-MS) using a PCS4000 (Bio-Rad, Hercules, California, USA) mass spectrometer, copper loaded immobilised metal affinity capture ProteinChip arrays (IMAC30-Cu2+), and a synthetic hepcidin analogue (hepcidin-24, Peptides International, Louisville, Kentucky, USA) as an internal standard,19 with recent technical improvements.23 The lower limit of detection was 0.55 nM. The intra- and inter-assay coefficient of variations of this method ranged from 6.1–7.3% and from 5.7–11.7% (mean 7.7%), respectively. In order to produce comparable results and to override the circadian rhythm of hepcidin,24 25 measurements were performed on samples obtained in all cases after an overnight fast.

Statistical analysis

Statistical analyses were performed by using STATA V.9 (StataCorp). Comparisons of all measured parameters in men and women were performed using the t test. Sex specific correlation analysis was used to assess the linear relationship between s-hepcidin and all other parameters. Subsequently, simple and multiple linear regression analyses were employed to find best predictors of serum hepcidin.

Heritability analysis

The heritability analysis was performed using SOLAR (Sequential Oligogenic Linkage Analysis Routines ver. 4.1.2) (, as described.20 As the distributions were not normal, a log10 transformation was performed for hepcidin, hepcidin/ferritin, and ferritin. For all phenotypes, individuals presenting values more than four SDs from the mean were removed.

Genotyping and association analysis

One thousand six hundred and sixty-four DNAs from the VB population were genotyped using the Illumina 370 Quad-CNV array, v3. DNAs with more than 10% missing genotypes, SNPs that failed the Hardy-Weinberg Equilibrium test (p<10−6) and with minor allele frequency <0.01 were removed. 343 866 SNPs, which included common SNPs in HFE, TMPRSS6 and transferrin (TF) genes, were used in the analyses.

Genome-wide association analysis (GWAS) was done on the directly genotyped SNPs using the GenABEL package26 that takes into account the relatedness among the VB population, using genomic kinship. To account for multiple testing, the p value cut-off for GW significance was 1.5E-7. An additive model was used on the standardised residuals of each quantitative trait adjusted for the effects of sex, age, sex*age, squared age and sex*square age. Each trait was checked for normality with non-parametric tests. A log10 transformation was performed for hepcidin, hepcidin/ferritin ratio, and ferritin. All the other traits (serum iron, transferrin, transferrin saturation) did not require normalisation. For all phenotypes, individuals with values more than four SDs were removed.

All the analyses were done on the whole sample and on a selected subset (subset 1) where individuals affected by acquired conditions known to alter iron metabolism and hepcidin concentrations were omitted. This includes subjects with C reactive protein (CRP) >1 mg/dl (as marker of clinical inflammatory conditions) and serum ferritin <30 ng/ml (as marker of iron deficiency).


Serum hepcidin concentrations show age and sex related changes and strongly correlate with serum ferritin

Serum hepcidin was determined in 1657 subjects, 929 females and 728 males, age range 18–98, mean age 55.4±17.8 years (supplemental figure S1). Anthropometric data, red cell parameters, serum iron, transferrin, transferrin saturation, and ferritin concentrations were available for all samples (supplemental table S1).

Hepcidin and most iron parameters showed striking age and sex dependent variations (figure 1 and supplemental table S2), particularly evident for ferritin (figure 1B), hepcidin (figure 1C) and, to a lesser extent, for transferrin saturation (figure 1A). Ferritin concentrations were lower in females aged <50 years and significantly higher in older females, while they remained stable across the different age groups in males (figure 1B). Serum hepcidin concentrations showed variations analogous to those of ferritin in individuals <50 years old, were similar in males and females aged 50–70, whereas among the elderly they were lower in females. The hepcidin/ferritin ratio, used to correct for hepcidin changes according to iron stores,27 28 clearly indicated the large difference between young males and females that sharply decreased with ageing (figure 1D).

Figure 1

Age and sex dependent variations of transferrin saturation (A), serum ferritin (B), serum hepcidin (C), and hepcidin/ferritin ratio (D) in the whole population. Males are indicated by a continuous line, females by a dotted line. Bars indicate SEs.

Serum hepcidin correlation analysis by sex (supplemental table S3) showed that Pearson's correlations were negligible for all measured parameters except for ferritin (r=0.32 and r=0.53 in men and women, respectively) and CRP (r=0.25 in women) and were always greater in females. Hepcidin concentrations are known to be decreased or even suppressed by iron deficiency and increased by iron overload and inflammatory cytokines. To reduce the effect of these confounding environmental variables, 41 males and 296 females with iron deficiency defined by ferritin <30 ng/ml and 75 subjects with CRP values >1 mg/dl (eight with concomitant iron deficiency) or individuals missing the information were excluded from the analyses. We also excluded 50 individuals with undetectable hepcidin who had multiple causes that might account for hepcidin suppression, such as heavy alcohol intake, β-thalassaemia trait, blood donations, and advanced age. The remaining 1203 individuals (642 males and 561 females, mean age 56.86±17) are indicated as subset 1.

To study correlations between hepcidin and iron and red cell parameters, separate linear regression analysis were performed in males and females from subset 1, using log transformed hepcidin and age as covariate (supplemental table S4). Significant variables were tested in multiple regression models (table 1). Age, ferritin, Hb and mean corpuscular haemoglobin concentration (MCHC) were independent predictors of hepcidin concentrations in males, accounting for 13.5% of the total hepcidin variability; in females age, ferritin and total cholesterol accounted for 17.7% of the total phenotypic variation. Only age and ferritin were common to both sexes. Serum iron, transferrin, and transferrin saturation lost their correlations with hepcidin, when adjusted for the other parameters.

Table 1

Multiple regression analysis of serum hepcidin by sex

Consistent with serum ferritin being a predictor of hepcidin concentrations, when we clusterised the subjects according to ferritin values in three classes corresponding to iron deficiency (Ft<30 ng/ml), normal iron balance (30≤Ft≤200 ng/ml in females and 30≤Ft≤300 ng/ml in males) and iron overload (Ft>200 ng/ml in females and >300 ng/ml in males), mean hepcidin concentrations increased progressively and differed significantly among the three groups (p<0.001) (figure 2A). Considering classes of transferrin saturation that define iron deficiency (<16%), normal iron status (16–45%), and iron overload (>45%), no significant differences in hepcidin concentrations were observed (figure2B).

Figure 2

Serum hepcidin in groups of individuals classified according to serum ferritin (A) or transferrin saturation (TfSat) (B) concentrations. Three classes are shown: iron deficiency (ferritin <30 ng/ml and TfSat ≤16%), normal iron status (intermediate values), and iron overload (ferritin >200 ng/ml in females/>300 ng/ml in males and TfSat >45%). Mean values are age and sex adjusted by ANOVA (95% CI).

Hepcidin heritability is low and increases when hepcidin is corrected for ferritin concentrations

The genetic component of the variability (heritability= H2) of all iron parameters is quite relevant and was estimated also for the VB population, thanks to the availability of a complete genealogy.20 The heritability of hepcidin, even calculated in subset 1, devoid of acquired confounding factors, was instead very low (H2=0.098) and non-significant (table 2). In this analysis ferritin was the most significant covariate (p=1.5E-28). Accordingly, if the serum hepcidin concentrations were normalised to ferritin and the hepcidin/ ferritin ratio was considered, the heritability increased (H2=0.219). Age and sex explained about 12% of the variability. None of the other iron parameters contributed to the model.

Table 2

Hepcidin and ferritin heritability

We calculated the heritability of ferritin in the same dataset. Ferritin H2 was significantly higher in subset 1 (H2=0.455) than in the whole population (H2=0.27) (table 2). Consistent with the correlation between hepcidin and ferritin, the heritability of ferritin decreased (H2=0.381) in subset 1 if hepcidin was included as a covariate. In this case the covariates sex and hepcidin explained 29.9% of the variability. None of the other iron parameters was a significant covariate for ferritin heritability.

Association of TMPRSS6 and HFE variants with iron and erythrocyte parameters is replicated in the Val Borbera population

The size of the study population was underpowered to find GW significant associations with hepcidin or hepcidin/ferritin ratio values (supplemental figure S2). However the availability of the hepcidin concentrations allowed us to evaluate better the effect of the common variants in two genes, TMPRSS6 and HFE, involved in monogenic disorders of iron metabolism.1 These variants were previously found associated with iron parameters and red blood cells traits12–14 and their effect was hypothesised to be dependent on hepcidin variation. Most findings of previous GWAS for iron parameters and erythrocyte traits were replicated in the VB population (supplemental table S5). In our series HFE rs1800562, corresponding to the C282Y variant, which at the homozygous state is responsible of hereditary haemochromatosis, was associated with serum iron (p=3.95E-9), transferrin (p=4.95E-11), and transferrin saturation (p=2.64E-15) at GW significance and to lesser extent to mean corpuscular haemoglobin (MCH), mean corpuscular volume (MCV) and MCHC. rs855791, corresponding to the A736V of the serine protease TMPRSS6,13 was associated at GW significance with serum iron (p=9.41E-11) and transferrin saturation (p=3.89E-9). rs3811647 in the TF gene was associated only with transferrin concentrations (p=2.1E-16).14 Association analysis of subset 1 showed an increased genetic effect of HFE and TMPRSS6 SNPs on iron and transferrin saturation and a smaller increase of TMPRSS6 effect on MCV and MCH (supplemental table S5). A large increase of the genetic effect was also found for the Tf SNP on iron and particularly on Tf.

The association of TMPRSS6 rs855791 variant to red cell traits was reported as mostly dependent on the amount of iron available for erythropoiesis.13 By using iron parameters as covariates in the regression analysis for MCV, MCH, and MCHC we found that iron, transferrin saturation and ferritin reduced the effect of both the HFE and TMPRSS6 in subset 1 while transferrin did not. Considering together iron, transferrin saturation and ferritin, HFE association was abolished and that of TMPRSS6 greatly reduced (supplemental table S6).

Association of common TMPRSS6 and HFE variants with iron parameters is not dependent on hepcidin concentrations

Based on the results described above we were able to assess whether the effect of the two common genetic variants in TMPRSS6 and HFE considered was mediated by hepcidin. We used hepcidin as covariate in the association analysis of 1545 genotyped individuals that had serum hepcidin measured. For all iron and red blood cells parameters, the association of the HFE and TMPRSS6 SNPs did not change significantly (supplemental table S7), suggesting that the two variants may exert a direct effect on these parameters.

A novel association of hepcidin/ferritin ratio to TMPRSS6 and HFE variants

The association of HFE rs1800562 variant with ferritin, which was GW borderline significant in the whole cohort (p=3.06E-7), became highly significant (p=7.49E-10) in subset 1, and the significance further increased if hepcidin was used as covariate (p=1.64E-10). TMPRSS6 rs855791 association with ferritin was nominally significant and did not greatly change after adjusting for covariates (table 3).

Table 3

Association of rs1800562 and rs855791 to hepcidin and ferritin

We also tested whether the HFE and TMPRSS6 variants were associated with hepcidin. As expected from the heritability results, hepcidin was not associated unless it was normalised to ferritin (table 3). The hepcidin/ferritin ratio was associated with both variants in subset 1: the HFE rs1800562 variant was significantly associated (p= 6.36E-04) but the effect was smaller than that observed for the other iron parameters. It explained only 1% of the variance compared to around 4% for ferritin and 3% for the other iron parameters (supplemental table S5). The VB cohort does not have enough statistical power to definitively demonstrate association of the TMPRSS6 rs855791 that was only nominally associated (p=1.49E-02).


We report here the first large scale epidemiological and genetic study of serum hepcidin, the main regulator of plasma iron concentration, in the general adult population.

Hepcidin concentrations under normal iron homeostasis showed striking gender differences and variation across ages. While in males concentrations were rather stable, in females more dynamic changes were observed, paralleling the well known age related ferritin changes. Young females had significantly lower concentrations than males. Females aged 50–60 years showed hepcidin concentrations comparable to those of age matched males, but had significantly lower serum ferritin. This underlines that the threshold for hepcidin increase in response to body iron is lower in females. Among the elderly, hepcidin concentrations decreased in both genders paralleling ferritin reduction, even if the hepcidin decrease was more evident in females (figure 1).

Hepcidin regulation was studied in human disorders of iron metabolism and in animal models: it is known to respond rapidly to increased circulating iron loaded transferrin3 29 (measured by transferrin saturation) and tissue iron, whose surrogate index is serum ferritin. In our series serum hepcidin strongly correlated with serum ferritin in both sexes, confirming results previously observed in a small number of individuals,25 30 but did not correlate with serum iron, transferrin, and transferrin saturation. Thus, although an acute increase in transferrin saturation triggers an hepcidin response,27 29 31 in steady conditions hepcidin appears mainly influenced by stored iron. Accordingly, hepcidin concentrations differed in groups of individuals whose iron status was assessed according to ferritin values, but not in groups classified according to transferrin saturation. The strong correlation between hepcidin and ferritin underscores the relevance of normalising the hepcidin concentrations using the hepcidin/ferritin ratio, as proposed in hereditary haemochromatosis.27 28 The correlation between hepcidin and ferritin concentrations was positive, reflecting the response of hepcidin to iron stores concentrations, likely through increased BMP6 and activation of the BMP signalling pathway.32 33 However, the reverse might also be true as hepcidin might modulate the concentration of serum ferritin by degradation of the iron exporter ferroportin,3 thus favouring iron retention in macrophages and an increase in cytosolic and serum ferritin.34

The previously reported associations of two common TMPRSS6 and HFE variants to iron were replicated in the VB cohort where they reached GW significance. We confirmed that association of the TMPRSS6 rs855791variant with all red cell traits was mostly dependent on the amount of iron available for erythropoiesis13 and we showed that HFE C282Y association with all red cell traits was abolished if iron, transferrin saturation and ferritin were considered together as covariates in association analysis.

The availability of serum hepcidin concentrations allowed us to directly test whether hepcidin could be the molecule mediating the association of the two TMPRSS6 and HFE variants. Our results showed that this was not the case. Since including hepcidin as a covariate in association analysis did not change the results, we suggest that the association of TMPRSS6 and HFE variants to iron parameters could result from a direct pleiotropic effect on the iron parameters. Alternatively, as hepcidin concentrations are homeostatically regulated by iron and erythropoietic activity, genetic effects may be masked by the hormone nature of hepcidin, which controls iron by a negative feedback and is strongly influenced by environmental factors. Accordingly, the heritability of hepcidin was negligible. The heritability was higher for the hepcidin/ferritin ratio (H2=0.219), confirming that genetic factors modifying hepcidin values may be masked by the rapid changes of hepcidin concentration in response to increased body iron stores or to other factors. We therefore tested association of the hepcidin/ferritin ratio to the HFE and TMPRSS6 variants considered. This resulted in a significant association of the hepcidin/ferritin ratio with HFE (and borderline with TMPRSS6) (table 3), demonstrating that the HFE C282Y mutation can indeed affect hepcidin, but with a modest effect that could not account for the strong effect of the same mutation on the other iron parameters.

Altogether, the two common variants considered appear to affect iron in part through hepcidin, likely through modulation of the BMP signalling pathway that integrates signals from erythropoiesis and iron stores to activate or repress hepcidin transcription.1 However, it is quite difficult to account for the direct and strong effect of the HFE C282Y mutation on transferrin. We cannot exclude the possibility that the HFE C282Y mutation affects additional and novel pathway(s) (figure 3) able to regulate iron homeostasis in normal situations and cause transferrin downregulation independently from hepcidin.

Figure 3

Model of the effect of HFE C282Y on transferrin saturation and hepcidin (see text for details). The dotted line indicates the hypothetical effect of reduced hepcidin on serum ferritin.

We also studied association of the two variants to ferritin. HFE C282Y was previously reported to be associated with ferritin in a single study.35 In our whole cohort it was borderline significant, but became significant at GW levels (p=7.49E-10) with a strong effect (β=0.556, SE=0.090) in subset 1, strengthening the importance of excluding, in this type of analysis, acquired conditions that modify iron metabolism and particularly hepcidin and ferritin concentrations. In addition, the effect of HFE association on ferritin in subset 1 was increased (p=1.64E-10, β=0.576, SE=0.090) after adjusting for hepcidin concentrations. This finding further confirmed the mutual control between the two variables and suggests that the positive effect of HFE C282Y on total body iron (and thus on ferritin) is in part antagonised by its negative effect on hepcidin. HFE C282Y increases transferrin saturation and cell iron uptake. However, the concomitant hepcidin downregulation favours iron release from macrophages. On one side this translates into the vicious cycle of further enhancing transferrin saturation, but on the other side, if the secreted ferritin is related to macrophage iron content,36 it would reduce serum ferritin (figure 3).

In conclusion, our study shows a complex interplay between hepcidin and ferritin and points to the high transferrin saturation in C282Y HFE haemochromatosis as the major cause of iron loading through increased cell iron uptake, despite increased iron release due to the low hepcidin–ferroportin interaction. The results also show a new association between HFE and TMPRSS6 variants with hepcidin/ferritin ratio, that could represent an index of potential clinical utility to assess adequate hepcidin secretion.


We thank the Val Borbera inhabitants who have made this study possible. We also thank the local administrations and the ASL-22, Novi Ligure (Al) for their continuous support.


View Abstract

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:


  • Funding This work was funded by grants from: Compagnia San Paolo Torino, and Progetto Finalizzato Sanità RF-FSR-2007-647201 to DT, Telethon Foundation Onlus, Rome, Grant GGP080892008, Regione Lombardia Sal-11, ID 17389 and EU Contract LSHM-CT-2006-037296 EURIRON to CC.

  • Competing interests None.

  • Patient consent Obtained.

  • Ethics approval San Raffaele Hospital Ethical Committee.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.