Main

Iron deposition unrelated to the HFE mutations of hereditary hemochromatosis is a common finding in cirrhotic livers, occurring in up to a third of explanted livers.1 Hemosiderosis appears to be more common in the setting of advanced liver dysfunction and may be predictive of a poor prognosis. We recently reported that the presence of stainable iron on an index liver biopsy showing cirrhosis was associated with higher baseline MELD score.2 Moreover, even when the more advanced degree of liver dysfunction in these patients was taken into account, they tended to decompensate more rapidly than patients without stainable iron. Ganne-Carrié et al3 have made similar observations in patients with alcoholic cirrhosis. In addition, liver transplant patients whose explant shows heavy iron deposition are reported to be at increased risk of mortality from infection.4 However, despite the prevalence and clinical significance of hemosiderosis, its cause remains unknown.

In the past several years, several important discoveries have provided new insights into the mechanisms governing iron metabolism. These include the identification of the iron transporters divalent metal transporter 1 (DMT1) and ferroportin (Fpn), a novel transferrin receptor 2 (TFR2), hemojuvelin (Hjv), and the iron regulatory peptide hepcidin.5, 6, 7, 8, 9, 10, 11 All of these are expressed in liver, and mutations in the latter four are associated with rare forms of iron overload.10, 12, 13, 14, 15 As the liver is important in the regulation of iron metabolism under normal circumstances, it is reasonable to surmise that hemosiderosis may result from dysregulated expression of one or more of iron-related proteins in cirrhosis.

Studies examining the expression of some of these newly described proteins in patients with chronic liver disease indicate that this may be the case. Stuart et al16 have reported that duodenal expression of DMT1 and Fpn, the iron transporters at the apical and basolateral surfaces of the enterocyte, respectively, are increased in cirrhotic patients. These alterations would be predicted to increase intestinal iron absorption, potentially contributing to hepatic iron deposition. These changes may reflect decreased levels of hepcidin, which downregulates intestinal iron absorption by causing internalization and degradation of Fpn. Consistent with this prediction, Détivaud et al17 found that hepatic hepcidin expression in 16 patients with moderate to severe fibrosis correlated inversely with fibrosis stage. These intriguing findings may provide an explanation for the pathogenesis of hemosiderosis; however, they remain to be confirmed in a larger study. Furthermore, as a comprehensive assessment of the effects of cirrhosis on the expression of genes involved in hepatic iron metabolism is lacking, it is unknown whether there are additional defects in iron regulation that contribute to the development of hemosiderosis. Therefore, the aim of this study was to compare the expression of genes involved in the regulation of iron metabolism in cirrhotic and control human livers.

MATERIALS AND METHODS

Human Livers

Samples from cirrhotic livers removed at liver transplantation were collected at University of Iowa Hospitals and Clinics (‘cirrhotics;’ n=22). Clinical and laboratory data were abstracted from the medical record. The latest laboratory values before transplantation were recorded. Control livers were obtained from patients undergoing resection of liver lesions at the University of Iowa Hospitals and Clinics (n=5). Samples collected from surrounding macroscopically normal-appearing liver were evaluated histologically to confirm the absence of tumor. Complete blood counts and iron studies were obtained on the morning of surgery. Liver samples from both groups were snap-frozen in liquid nitrogen and stored at −80°C. Because of the potential variability of intraorgan iron concentrations and gene expression, iron quantitation and RNA isolation were performed on three separate specimens from each cirrhotic liver. Except where noted, results presented are those obtained from the sample with the median hepatic iron concentration (HIC). Owing to the small size of the control specimens, only one sample was available for each parameter. The study was approved by the Institutional Review Board of University of Iowa.

HIC

Quantitation of HIC was performed as previously described.18 The upper limit of normal using this assay is 1500 μg/g dry wt.

Histochemical Staining for Iron

Perls’ Prussian Blue staining was carried out on formalin-fixed, paraffin-embedded sections of livers using standard techniques. Iron deposition was assessed using the method of Edwards et al.19

Real-Time RT-PCR

Quantitation of hepcidin (HAMP), ferritin (FTH1, ferritin heavy polypeptide 1), transferrin (TF), and TFR2 gene expression were assessed using the TaqMan assay using the primers given in Table 1. PCR reagents for transferrin receptor (TFRC), Fpn (also known as Slc40a1), DMT1 (also known as Slc11a2), and Hjv(also known as hemochromatosis type 2) were purchased from Applied Biosystems (Foster City, CA, USA; TaqMan Assays on Demand, assays Hs99999911_m1, Hs00205888_m1, Hs00167206_m1, and Hs00377108_m1, respectively). Oligonucleotide probes for HAMP, FTH1, TF, and TFR2 were labeled with the fluorescent dye FAM on the 5′ end and with TAMRA or BHQ on the 3′ end as shown in Table 1.

Table 1 Primer and probe oligonucleotide sequences for HAMP, FTH1, TF, and TFR2
Full size table

Real-time RT-PCR was performed as previously described using TaqMan 18S Ribosomal RNA Control Reagents (Applied Biosystems) as an endogenous control.20 Validation experiments showed similar amplification efficiencies for the target gene and the 18S PCRs. The comparative Ct method was used to calculate the amount of target mRNA, normalized to the amount of 18S RNA, and relative to an internal calibrator (control human liver), expressed as 2 -ΔΔ C t . Semiquantitative RT-PCR for DMT1 isoforms was performed using primers described by Hubert and Hentze.21

Western Blots

Portions of frozen liver were homogenized and analyzed by western blot as previously described.20 Primary antibodies include anti-TFRC (13-6800; Zymed Laboratories, San Francisco, CA, USA; dilution 1:500), anti-ferritin (F5012; Sigma, St Louis, MO, USA; dilution 1:1000), and anti-DMT1 (sc-30120; Santa Cruz Biotechnology, Santa Cruz, CA, USA; dilution 1:200). The Fpn antibody has been previously described.22 Peroxidase-conjugated secondary antibodies were obtained from Santa Cruz Biotechnology. Bound antibody was detected using chemiluminescence according to the manufacturer's instructions (ECL; GE Healthcare Bio-Sciences Corp., Piscataway, NJ, USA). The DMT1 and Fpn Westerns were stripped (Antibody Stripping Solution; ADI, San Antonio, TX, USA) and reprobed with an anti-β-actin antibody (A2066; Sigma; dilution 1:10 000). Bands were quantitated by densitometry using ImageJ software (NIH).

Statistical Analysis

Comparisons were carried out by Mann–Whitney U-statistics on medians when data were not normally distributed and unpaired two-tailed t-test when normally distributed. Spearman's rank order correlation test was used to determine nonparametric correlations between parameters and linear regression analysis to evaluate if linear correlation was present. P-values of <0.05 were considered significant.

RESULTS

Clinical Characteristics

The control group was comprised of 1 man and 4 women, with an average age of 63 (range 47–78). Two subjects had pancreatic cancer and liver nodules on imaging studies. They underwent laparotomy to exclude metastatic disease before pancreatectomy. Of the remaining patients, one had metastases from colon cancer, one had a metastasis from renal cancer, and one had a symptomatic liver cyst. Histologic evaluation of control livers revealed completely normal histology in two and mild, nonspecific changes in the remainder (two with mild portal inflammation without fibrosis and one with mild ductular proliferation). The cirrhotic group included 12 men and 10 women. Their average age was 50 years (range 24–67). Chronic hepatitis C was the most common underlying liver disease, followed by alcoholic liver disease and chronic hepatitis B.

Clinical parameters related to iron metabolism of the subjects are shown in Table 2. Transferrin saturations were significantly higher in the cirrhotics than in the controls (59±7 vs 24±4%; P<0.01). This was a consequence of elevated serum iron levels in the cirrhotic group compared to the controls (106±15 vs 61±15; P=0.03), as well as a trend toward decreased total iron binding capacity. On average, serum ferritin levels were higher in the cirrhotics, but this difference was not significant. Hemoglobin and hematocrit levels were similar in the groups.

Table 2 HIC and clinical parameters related to iron metabolism of study subjects
Full size table

Hepatic Iron Concentrations and Iron Staining

HICs tended to be higher in the cirrhotic livers, but this difference was not significant (Table 2). Notably, the range of HICs was much broader in the cirrhotic group (101–3268 μg/g) than among the controls (320–774 μg/g). Consistent with previous reports,23, 24 intraorgan HICs in some cirrhotic livers varied considerably (>5-fold).

There was a strong correlation between HIC and grade of iron staining (rs=0.89; P<0.0001). A total of 13 had ≥1+ stainable iron. Hepatocellular iron deposition predominated in the majority (11 of 13), with all but one also showing sinusoidal lining cell iron.

Hepatic Expression of Genes and Proteins Involved in Iron Metabolism

Levels of HAMP transcripts were similar in the control and cirrhotic groups (P=0.976; Figure 1). In contrast, levels of FPN, DMT1, and ferritin mRNA differed significantly between the groups (Figures 1 and 2). Fpn transcript levels were significantly higher in cirrhotics than in controls (P=0.007). In contrast, the 100 kDa band corresponding to Fpn was generally weaker in the cirrhotic livers than in the controls on western blot (Figure 3). Although there was considerable variability among the cirrhotic livers, Fpn levels did not appear to vary with iron content at the protein level nor did the abundance of its mRNA correlate with HIC (Figure 3; Table 3).

Figure 1
figure 1

Relative mRNA expression of HAMP (P=0.976) and hemojuvelin (P=0.864) did not differ between the groups. DMT1 and FPN mRNA expression were significantly higher in the cirrhotics than in controls, P=0.039 and 0.007, respectively. *P=0.039; **P=0.007.

Full size image
Figure 2
figure 2

Relative mRNA expression of ferritin (FTH1) was significantly higher in the cirrhotics than in controls (P=0.025); transferrin (TF) and transferrin receptor 1 (TFRC) mRNA levels also tended to be higher in the cirrhotics (P=0.055 for both). Transferrin receptor 2 (TFR2) did not differ between the groups (P=0.99). *P=0.025.

Full size image
Figure 3
figure 3

Western blot of ferroportin in control and cirrhotic human livers demonstrates that abundance of ferroportin in the cirrhotic livers is variable but generally lower than in the control livers and has no obvious relationship to hepatic iron concentration. Lysates from cirrhotic human livers are grouped according to iron concentration determined on samples from the same livers. Western blotting was performed as described in the ‘Materials and methods’.

Full size image
Table 3 Summary of alterations in iron regulatory genes and proteins in cirrhotic human livers compared to controls
Full size table

Like Fpn, levels of DMT1 mRNA were significantly higher in cirrhotic livers than in controls (P=0.039), whereas levels of DMT1 protein tended to be diminished in the cirrhotic livers. DMTh HIC or mRNA level (Figure 4; Table 3). Several DMT1 isoforms have been described, including an iron-response element (IRE)-containing form, a non-IRE form, and two additional splice variants.21 Because the primers used in the quantitative PCR are expected to recognize all four isoforms, we performed semiquantitative RT-PCR using isoform-specific primers to determine whether expression of DMT1 isoforms is altered in cirrhotic liver. Both the IRE and non-IRE DMT1 isoforms were present in control and cirrhotic livers, and the IRE-containing transcript remained detectable in cirrhotic livers with elevated iron concentrations (Figure 5). The exon 1B variant was likewise present in all control and cirrhotic livers examined. In contrast, the exon 1A variant was detected in only a few cirrhotic livers; however, no relationship between the presence of this variant and increased hepatic iron was evident when additional livers were examined (data not shown). Thus, apart from the exon 1A variant, the repertoire of DMT1 expression is similar in control and cirrhotic liver, and is not clearly affected by iron content in the latter.

Figure 4
figure 4

Western blot of DMT1 in control and cirrhotic human livers shows variable abundance of DMT1 irrespective of the presence of cirrhosis and of hepatic iron concentration. Western blotting was performed as described in the ‘Materials and methods’ using an antibody that detects all four DMT1 isoforms. The layout of the liver samples the same as in Figure 3.

Full size image
Figure 5
figure 5

RT–PCR showing that the IRE, non-IRE, and exon 1B variants of DMT1 are expressed in control and cirrhotic human livers. A low level of expression of the exon 1A variant is detected only in some cirrhotic livers. Hepatic iron concentrations (HICs) were determined on samples taken from the same livers. ‘Normal’ indicates control human livers; ‘low iron’ and ‘high iron’ indicate HIC <500 μg/g and >1500 μg/g, respectively. RT–PCR was performed as described in the ‘Materials and methods’. The first lane is 100-bp ladder; the lane labeled ‘+’ indicates the positive control (Caco-2 cells), ‘−’ the negative control (no RT).

Full size image

FTH1 mRNA levels were also significantly higher in the cirrhotic livers than in the controls (3.1±0.6 vs 2.3±0.3; P<0.03), whereas TFRC and TF transcripts showed similar trends, which did not achieve statistical significance (Figure 2). Expression of TFRC and ferritin were assessed in cirrhotic livers by western blot (Figure 6). TFRC was readily apparent in livers with HICs <500 μg/g, but its abundance diminished progressively with increasing HIC, becoming undetectable in livers with HICs >1000 μg/g, whereas ferritin showed the inverse pattern. Thus, at the protein level, TFRC and ferritin are expressed in a physiologically appropriate manner relative to iron content in cirrhotic human livers, indicating that dysregulation of these proteins is unlikely to be causally related to hemosiderosis.

Figure 6
figure 6

Western blot of transferrin receptor and ferritin in cirrhotic human livers demonstrates a reciprocal relationship between the abundance of transferrin receptor and ferritin that is physiologically appropriate vis-à-vis iron concentrations. Lysates from cirrhotic human livers are grouped according to hepatic iron concentration (HIC) as determined on samples taken from the same livers.

Full size image

Previous studies indicate that hepatocytes are the predominant source of HJV and TFR2 transcripts in liver.25, 26 The finding that there were no significant differences in the level of HJV and TFR2 transcripts (Figures 1 and 2) among the groups suggests that, despite the variable content of inflammatory and fibrogenic cells in the cirrhotic livers, hepatocytes were the predominant source of RNA in those samples. These data are summarized in Table 3.

Correlations Between Gene Expression, Hepatic Iron Concentrations, and Clinical Parameters

On the basis of analysis of the three separate samples from each liver, a highly significant correlation between HAMP mRNA levels and HIC was observed in the cirrhotic livers (n=66, rs=0.425; P<0.0004). A similar result was obtained when only data from a single sample from each cirrhotic liver was considered (rs=0.508; P=0.016) (Table 4; Figure 7). Although no correlation was observed between HAMP transcript levels and HIC in the control group, this is likely a type II error due to the small number of samples in the control group, as previous work has shown a positive correlation between HIC and HAMP expression in human liver.17

Table 4 Spearman correlations between HIC and transcript levels for genes involved in iron metabolism in cirrhotic livers
Full size table
Figure 7
figure 7

Relationship between HAMP mRNA and hepatic iron concentration in control and cirrhotic human livers. Circles represent cirrhotic livers, and inverted triangles indicate control livers.

Full size image

DMT1 and Fpn transcripts were strongly correlated in cirrhotic livers (rs=0.72; P=0.0001), suggesting that hepatic expression of DMT1 and Fpn may be coordinately regulated. There was no correlation with either of these transcripts and HAMP mRNA or HIC, suggesting that the expression of these iron transporters is not influenced by hepatic iron status, at least in cirrhotic livers (Table 4). Interestingly, a number of highly significant correlations were observed between HJV expression and that of other genes involved in iron metabolism including DMT1 (rs=0.59; P=0.0038), Fpn (rs=0.61; P=0.0027), TFRC (rs=0.76; P<0.0001), and TF (rs=0.48; P=0.02).

Correlations of HIC with clinical parameters of iron status were also assessed (Table 5). Serum iron and transferrin saturation both showed significant correlations with HIC in the cirrhotic livers (rs=0.720; P<0.0002 and rs=0.731; P<0.0002, respectively). A strong correlation between HIC and serum ferritin was observed in the cirrhotic livers (rs=0.781; P<0.0001) with a similar trend seen in the control livers (rs=0.90; P=0.083). These data suggest that serum iron studies provide a relatively good indirect assessment of HIC in cirrhosis, with ferritin being the single best predictor.

Table 5 Spearman's correlations between serum iron parameters and HICs and transcript levels of genes involved in iron metabolism
Full size table

No relationships were identified between HAMP transcripts and serum markers of iron status in the control group. In the cirrhotic group there was a significant negative correlation between HAMP mRNA and total iron binding capacity (rs=−0.500; P=0.021), whereas positive correlations were observed between HAMP mRNA and transferrin saturation (rs=0.428; P=0.023) and HAMP mRNA and serum ferritin (rs=0.704; P=0.0011) (Table 5). The correlation of these parameters with hepcidin expression is consistent with the relationship of hepcidin and iron stores as reflected by the HIC and provides further evidence that hepcidin expression at the mRNA level is responsive to iron status in cirrhosis.

DISCUSSION

This work represents the first comprehensive assessment of iron-related gene expression in human livers. To our knowledge, it is also the only work concerning hepcidin expression in diseased human livers in which multiple samples from individual livers were assessed for both iron content and hepcidin expression, to take into account the variability of hepatic iron deposition within cirrhotic livers. Contrary to our expectations, we found that levels of HAMP mRNA are similar in control and cirrhotic human livers and correlate with HIC in the latter, suggesting that regulation of hepcidin gene expression in response to iron status is preserved in cirrhosis.

In assessing these results, it is important to consider the possibility that iron metabolism was abnormal in our control subjects, most of whom had malignancies. However, anemia of chronic disease or cytokine-driven acute phase physiology associated with malignancy would raise hepcidin expression,27 thereby potentially increasing its levels in this group compared to the cirrhotics, which was not the case. Thus, although this factor seems unlikely to account for the lack of difference in HAMP mRNA levels between the groups, we cannot exclude the possibility that malignancy dampens hepcidin expression. In this context, it is worth pointing out that we originally intended to use donor livers as controls in this study, but abandoned this when our initial data showed a pattern of gene expression suggestive of an acute phase reaction (O Bergmann and K Brown, unpublished results). These observations underscore the challenges of working with scarce human tissues such as livers.

We must qualify our conclusion that hepcidin expression is regulated appropriately in cirrhotic livers by noting that we found no correlation between HAMP mRNA and HIC in control livers in this study, probably because of the small number of samples in this group. Given this limitation, it is possible that cirrhosis alters the relationship between HAMP mRNA and HIC in a quantitative manner, eg, hepcidin expression may be lower at any given HIC in cirrhotic liver than would be the case in normal liver. If so, the correlation between HAMP mRNA and hepatic iron in cirrhotic livers notwithstanding, hepcidin expression in cirrhosis may be deficient in absolute terms and thus contribute to inappropriate iron absorption. This issue may be resolved by examination of a larger number of bona fide normal human livers; however, such an undertaking will be difficult owing to ethical and logistic considerations.

The concept linking downregulated hepcidin expression with hemosiderosis is supported by recent studies showing decreased hepatic hepcidin expression after short-term alcohol administration to mice28 and with increasing age in HCV polyprotein transgenic mice.29 Although these data establish that hepcidin expression can be modulated by alcohol and HCV, two important causes of chronic liver disease in humans, the relevance of these observations to patients with cirrhosis is uncertain. On the one hand, there are no data to indicate that the alcohol-induced reduction in hepcidin expression results in hepatic iron accumulation, which would be unlikely in any case, given the brief (7 days) duration of the experiment.28 Whether longer-term exposure to alcohol would result in persistent downregulation of hepcidin expression and/or dysregulation of iron metabolism is unknown at present, but it is worth noting that previous work in alcohol-treated rodents has failed to demonstrate spontaneous iron overload.30, 31 On the other hand, in contrast to HCV-infected humans, in whom hemosiderosis is uncommon before the development of advanced fibrosis,32 HCV transgenic mice develop mild hepatic iron deposition in the absence of hepatic fibrosis. Thus, whether the causes and consequences of altered hepcidin expression in these animal models mimic those in human chronic liver disease remains to be determined.

In contrast to the lack of an effect of cirrhosis on hepcidin expression, we found that cirrhotic human livers demonstrate increased levels of Fpn and DMT1 mRNA, whereas the levels of the respective proteins tended to be lower in cirrhotic livers than in controls. Neither the mRNA nor the protein levels of the transporters correlated with hepatic iron concentrations, suggesting that the alterations in their expression were a consequence of cirrhosis and not a direct response to the presence of excess iron. The reason for the lack of correspondence between the mRNA and protein levels of the iron transporters is unclear. Little is known about the regulation of these genes in liver; however, the available data suggest that hepcidin and/or iron status has little, if any, influence on levels of Fpn mRNA. Yeh et al have reported that Fpn transcripts are unaltered in the livers of iron-deficient Belgrade rats (in which hepcidin expression is dramatically decreased); these investigators also observed that systemic administration of recombinant hepcidin to rats failed to modify hepatic Fpn mRNA levels.33 We have found that both Fpn and DMT1 mRNA are increased in the livers of C57Bl6 mice fed a high fat diet or chronically administered 20% ethanol in the drinking water. In neither of these models of fatty liver does the expression of the iron transporters correlate with HAMP mRNA or HIC (O Bergmann and K Brown, in preparation). These data strongly suggest that hepatic levels of Fpn and DMT1 mRNA are not directly controlled by either HIC or HAMP expression, but may be influenced by inflammatory or stress-responsive pathways. Although examination of the promoter region of these genes will be helpful to confirm these speculations, it is noteworthy that these findings in murine models of liver injury are consistent with our observations in cirrhotic human livers.

A second question raised by our findings is whether the abundance of Fpn in the liver is regulated directly by hepcidin, as is the case in the intestine.34 Our data showing decreased levels of Fpn in the cirrhotic livers with no increase in HAMP mRNA do not strongly support such a mechanism; however, it is important to note that we did not measure circulating or hepatic levels of hepcidin. Assuming that local hepcidin production is preserved in diseased liver and that it normally mediates Fpn degradation and internalization in hepatocytes and/or Kupffer cells in an autocrine or paracrine manner, an alternative mechanism, such as impaired synthesis or localization of Fpn, may lead to increased turnover of the protein and decreased levels in cirrhosis.

A further obstacle in assessing the potential role of the transporters in the pathogenesis of hemosiderosis is the lack of information concerning their function in the liver. In the intestine, DMT1 and Fpn constitute the ‘mucosal block’ to iron uptake, respectively, regulating the entry and egress of iron from enterocytes. Consistent with this role, the abundance of both proteins increases in the gut in response to iron deprivation and decreases with iron repletion. In contrast, Trinder et al35 have reported that DMT1 behaves in an inverse manner in the liver, with hepatocyte DMT1 immunoreactivity increased in iron overload and decreased in iron deficiency. These investigators have proposed that hepatocyte DMT1 may be involved in the clearance of non-transferrin bound iron (NTBI) from portal blood, preventing the entry of low molecular weight iron into the systemic circulation and diverting it into hepatocytes, which have a robust capacity for iron storage. If so, hepatocyte DMT1 may be particularly important in cirrhosis, as duodenal expression of iron transporters is increased in cirrhotics, thus creating the potential for iron uptake to exceed the binding capacity of transferrin, leading to the presence of NTBI in the portal circulation.16 The decrease in DMT1 protein in the cirrhotic livers appears unlikely to play a primary role in the development of hemosiderosis, but may contribute to the increase in circulating levels of NTBI reported in cirrhotic patients.36

In contrast to DMT1, we speculate that the decrease in Fpn in the cirrhotic livers may be mechanistically linked to the pathogenesis of hemosiderosis. Although Kupffer cells are the major cellular source of Fpn in liver, hepatocytes also express Fpn.6, 37 Decreased Fpn in either or both of these cell types would be expected to result in intracellular iron retention, which is consistent with the variable iron deposition in parenchymal and nonparenchymal cells in siderotic livers.38 However, if decreased Fpn were the sole abnormality, the resulting increase in hepatic iron content would be expected to elicit a physiological response leading to decreased intestinal iron absorption that would prevent progressive iron loading. The fact that this does not occur indicates that there are likely to be additional abnormalities, perhaps involving the sensing of iron content or biosynthesis or secretion of hepcidin, that participate in the pathophysiology of secondary iron overload. Despite the recent advances in this area, much remains to be learned regarding regulation of hepatic iron metabolism.