Hepcidin levels correlate to liver iron content, but not steatohepatitis, in non-alcoholic fatty liver disease

All patients referred to the Unit of Liver Diseases at the Karolinska University Hospital for liver biopsy due to chronic liver disease and/or hemochromatosis, and with an elevated serum ferritin, between January 2008 and April 2013, were asked to participate in the study. Hyperferritinemia was defined as a serum ferritin > 350 μg/L. In addition, patients with chronic liver disease and normal iron parameters undergoing liver biopsy were enrolled for comparison. All patients were over 18 years of age and had given written informed consent. One patient was excluded due to iron deficiency. No patients included had been subject to treatment with iron reduction therapy before entering the study.

A total of 84 patients were enrolled (26 females, 58 males), of which 62 had elevated ferritin levels and 23 a normal serum ferritin concentration. Thirty-eight of the 84 patients had NAFLD, 17 had untreated hereditary hemochromatosis (HH), and 29 had various other causes of chronic liver disease (CLD), such as autoimmune liver disease, alcoholic liver disease, chronic viral hepatitis, alpha-1-antitrypsin deficiency, cryptogenic cirrhosis, porphyria cutanea tarda, methotrexate-induced liver fibrosis, or the hemochromatosis phenotype but without the C282Y or H63D mutations. All these other etiologies (except NAFLD and HFE-associated HH) were thus grouped together as CLD. NAFLD was defined as either Grade 1 or more steatosis in the liver biopsy according to Kleiner et al. [22], or a bright liver with increased echogenicity at abdominal ultrasound investigation. Among the 17 patients with HH, 12 were HFE C282Y homozygotes and five were C282Y/H63D compound heterozygotes. In the group of 29 patients with chronic liver disease, eight had a normal iron content in the liver, and 21 had iron overload, and were classified as “chronic liver disease with iron overload” (CLD-IO). One of these had received oral iron substitution for several years; however, none had been treated with parenteral iron substitution or blood transfusions. Amongst the 21 patients classified as CLD-IO, ten had a clinical phenotype of hemochromatosis (elevated serum ferritin and transferrin saturation, and hepatic iron overload) but without homozygosity for the HFE C282Y mutation or compound heterozygosity for the C282Y and H63D mutations, and without alcohol overconsumption. The other 19 CLD patients had alcohol overconsumption (> 30 g/day) (n = 9), primary biliary cholangitis (n = 2), hepatitis C (n = 1), alpha-1-antitrypsin deficiency (n = 1), porphyria cutanea tarda (n = 1), cryptogenic cirrhosis (n = 2), or methotrexate-treated psoriasis (n = 3). None of the patients with HH, NAFLD or CLD with the clinical phenotype of hemochromatosis had reported a previous or current alcohol consumption exceeding 20 g/day. Two CLD-IO patients (with alpha-1-antitrypsin deficiency and alcohol overconsumption, respectively) were heterozygous for the H63D mutation, and one (with alcohol overconsumption) was heterozygous for the C282Y mutation.

Iron overload was defined as a histologic iron score of ≥1 or an estimated iron content > 40 μmol/g on magnetic resonance imaging (MRI) investigation (see below). The patient groups are displayed in Table 1.

Liver biopsy was performed in 66 out of the 84 patients. MRI was used for iron assessment in 35 cases, and in 14 of these, histology was lacking. In 21 cases, there was both histology and MRI. In four cases (two HH homozygotes, one HH compound heterozygote, and one with CLD and normal ferritin) both liver histology and MRI was lacking.

A total of 40 healthy controls, recruited from hospital staff, participated in the study. None had a history of liver disease. Written consent was given. Of the controls, six individuals were excluded (elevated transaminases in one case, compound heterozygosity of the HFE gene and elevated serum ferritin in one case, iron deficiency with serum ferritin < 15 μg/L in three cases, and elevated ferritin (413 μg/L) in one case). The remaining 34 controls were included in the study (Table 1).

Biochemical data was collected at the time of enrollment in the study. Blood samples were drawn before 10 A.M. in the morning. Subjects were not fasting but had had a light breakfast. Routine blood chemistry analyses as well as HFE mutation analysis were performed on all subjects at the Department of Clinical Chemistry at Karolinska University Hospital. Body mass index was calculated using the formula: weight in kilogram / (height in meters) ².

Freshly drawn samples from the 84 patients and 34 controls were centrifuged and serum was stored at − 70 °C until analysis. Samples were analyzed for hepcidin by a competitive ELISA kit (Bachem, Peninsula Laboratories, LLC, CA, United States) as reported previously [23]. Reference ranges established in 83 normal subjects showed hepcidin levels that ranged 8–76 and 2–50 μg/L for men and women, respectively (2.5–97.5 percentiles). The results were significantly different between genders. As internal controls, pooled sera of 7 and 6 samples representing low (≈0.4 μg/L) and normal (≈3 μg/L) levels respectively were frozen at − 70 °C. Control sera were run in 6 replicates at each assay. The intra-assay variation showed CVs of 18% for low and 13% for normal controls, while inter-assay CVs were 18 and 19%, respectively. The lower limit of detection, calculated as 3 SD above the lowest standard, was 0.05 μg/L and linearity for this kit was determined as between 0,2–5 μg/L (2–50 μg/L for samples diluted 1:10). Samples outside linearity limits were rerun using proper dilution factor, and all samples were run in duplicate.

IL-6 and TNFα were measured using Bio-plex Pro Human Cytokine Group 1 kit (Bio-rad Laboratories, Hercules, CA, USA) according to the manufacturer’s instructions. Briefly, plasma/serum was diluted 1:4 using Bio-plex sample diluents. To obtain the nine point (including blank) standard curve, the kit standard was reconstituted and diluted fourfold. The 10× IL-6 and TNFα coupled beads were diluted in kit assay buffer and added to all standard and sample wells. The plate was incubated on shaker 30 min. After washing IL-6 and TNFα biotinylated detection antibodies were added and the plate was incubated as above. In the final step, PE-conjugated Streptavidin was added and the plate was run on a Magpix instrument (Luminex Corporation, Austin TX, USA) and analyzed with xPonent software (Luminex).

Sixty-six patients underwent liver biopsy, and tissue from 39 of these was collected for hepcidin mRNA analysis. Tissues were immediately immersed in RNAlater and stored at − 70°C until processed. Total RNA was successfully retrieved from 36 of the 39 utilized liver biopsies with a dry weight of 0.3–5.9 mg using the RNAqueous -4PCR kit (Ambion PN AM1914). Recovered quantities of RNA ranged from 13 to 200 ng/μL. The quality and quantity of the extracted RNA was verified with the Bio-Rad Experion 700–7000 electrophoresis system, and only samples with an RQI > 8 were included in the study. cDNA synthesis was carried out with the High Capacity Reverse Transcriptase Kit (Applied Biosystems), using 65–930 ng of total RNA per sample. Determination of specific mRNA levels was performed as described previously [23].

Liver biopsy samples were revalued by an experienced pathologist (O.D.) blinded to clinical data. Samples from NAFLD-patients were evaluated for the degree of steatosis (0–3), lobular inflammation (0–3) and hepatocellular ballooning (0–2) according to Kleiner et al. [22]. The unweighted sum of these three variables were used to calculate the NAFLD activity score (NAS). Patients with NAS ≥5 were diagnosed with NASH.

Siderosis was determined for all patients semi-quantitatively on histopathologic examination of Perls’ stained liver biopsy samples adapted from Deugnier et al. [24] to match available levels of magnification.

An iron score from 0 to 4 for iron in hepatocytes was determined as follows: [0] granules absent or barely discernible at a magnification of 400X; [1] barely discernible granules at a magnification of 200X but easily confirmed at a magnification of 400X; [2] discrete granules at 100X magnification; [3] discrete granules easily confirmed at magnification of 40X, but barely discernible at a magnification of 20X; [4] granules obvious at a magnification of 20X, and barely visible for the naked eye. RES-iron was also determined and scored as [0] none, [1] mild, [2] or more than mild, as described by Nelson et al. [25]. These two scores were transformed into a histologic iron score (HIS) ranging from 0 to 5, comprising the score for iron in hepatocytes (0–4), plus one point for RES iron in those cases where it had been determined as more than mild, or a half point where it has been determined as mild. Iron overload was defined as a histologic iron score of ≥1.

MRI was used for detection and quantification of liver iron overload in 35 patients and correlated to histology in 21 of these (Fig. 1). The liver iron was assessed semi-quantitatively as described by Gandon et al. [26].

In the correlation analyses of serum hepcidin to liver iron content, MRI iron was approximated to histologic liver iron (HIS) score based on the correlation estimated from Fig. 1: < 40 μmol iron/g tissue  = HIS 0; 40–74 μmol/g = HIS 1; 75–129 μmol/g = HIS 2; 130–179 μmol/g = HIS 3; 180–239 μmol/g = HIS 4; ≥240 μmol iron/g tissue = HIS 5.

The relationship between two categorical variables was examined with Chi²-test or Fisher’s exact test (when applicable). Numerical values of laboratory parameters were analyzed using one-way ANOVA and validated for equal variance and normal distribution. Kruskal-Wallis ANOVA was used when the assumptions did not hold. The correlation between two numerical variables was analyzed with simple linear regression validated for linearity, variance between observations and for normal distribution. In the cases where the assumptions did not hold the Spearman’s rank order correlation was used instead. Multiple linear regression was used for variables that were significantly correlated to serum hepcidin in the simple linear regression. A p-value < 0.05 was considered statistically significant.

The distribution of patients, and clinical and laboratory data of patients and controls are demonstrated in Table 1. Controls were significantly younger than patients, and had lower BMI, ALT and serum ferritin levels. BMI was highest in the NAFLD patient group. Transferrin saturation was significantly increased in patients with homozygous HH, and in the 10 CLD-IO patients with a HH phenotype without HFE mutations, compared with the other patient groups and controls. Hepatic iron score did not differ significantly between patients with DIOS and CLD-IO.

The distribution of HFE mutations are shown in Table 2. Among patients with chronic liver disease and iron overload (CLD-IO), four were heterozygous for C282Y, two homozygous and one heterozygous for H63D. The H63D mutation was significantly more frequent in patients with NAFLD as compared with the controls (p < 0.05).

Simple linear regression showed a good correlation between histologic iron score and hepatic iron content determined by MRI, as demonstrated in Fig. 1 (r² = 0.77, p < 0.01).

Serum hepcidin values for the different patient groups and controls are shown in Fig. 2. Serum hepcidin levels were significantly increased in patients with NAFLD with DIOS and in those with chronic liver disease with iron overload (CLD-IO) compared with the other groups. The ratios between serum hepcidin and ferritin are shown in Fig. 3. As expected, this ratio was significantly reduced in homozygous HH compared with the other groups. Among patients with CLD-IO, this ratio was slightly lower in those with alcoholic liver disease (ALD) or hepatitis C (0.049±0.034) as compared with those without alcohol overconsumption (0.058±0.032), or DIOS patients (0.070±0.037), although these differences were not statistically significant.

Figure 4 shows the ratios between serum hepcidin and hepatic iron score, which was similar in patients with CLD-IO and DIOS, and reduced in those with homozygous HH. The hepcidin/iron score ratio was slightly lower in those with ALD or hepatitis C (18.7±8.1) as compared with those without alcohol overconsumption (22.4±10.2), or DIOS (30.8±23.7), however not statistically significant. There was a significant correlation between serum hepcidin levels and hepatic HAMP mRNA (r² = 0.39, p < 0.01).

Serum hepcidin, serum transferrin saturation and hepatic iron score were all significantly higher in NAFLD with DIOS as compared with NAFLD without DIOS (p < 0.05). Serum levels of triglycerides or total cholesterol did not differ significantly between the groups. Levels of TNF-α and IL-6 were highest in NAFLD without DIOS and  elevated serum ferritin (difference not statistically significant). HAMP mRNA in liver tissue correlated to the hepatic iron score (r² = 0.45, p < 0.05) but not to NAFLD activity score (r² = 0.003, p < 0.89). Serum hepcidin correlated significantly to serum ferritin (r² = 0.20, p < 0.01) and serum transferrin saturation (r² = 0.17, p < 0.01) but not to BMI, TNF-α, IL-6, triglycerides or cholesterol. In multiple linear regression analysis only ferritin correlated significantly to serum hepcidin levels when adjusted for other variables. There was no significant difference in stage of fibrosis, grade of steatosis, ballooning, lobular inflammation or NAFLD activity score between the groups (Table 3).