Global Transcriptional Response to Hfe Deficiency and Dietary Iron Overload in Mouse Liver and Duodenum

Iron is an essential trace element whose absorption is usually tightly regulated in the duodenum. HFE-related hereditary hemochromatosis (HH) is characterized by abnormally low expression of the iron-regulatory hormone, hepcidin, which results in increased iron absorption. The liver is crucial for iron homeostasis as it is the main production site of hepcidin. The aim of this study was to explore and compare the genome-wide transcriptome response to Hfe deficiency and dietary iron overload in murine liver and duodenum. Illumina™ arrays containing over 47,000 probes were used to study global transcriptional changes. Quantitative RT-PCR (Q-RT-PCR) was used to validate the microarray results. In the liver, the expression of 151 genes was altered in Hfe−/− mice while dietary iron overload changed the expression of 218 genes. There were 173 and 108 differentially expressed genes in the duodenum of Hfe−/− mice and mice with dietary iron overload, respectively. There was 93.5% concordance between the results obtained by microarray analysis and Q-RT-PCR. Overexpression of genes for acute phase reactants in the liver and a strong induction of digestive enzyme genes in the duodenum were characteristic of the Hfe-deficient genotype. In contrast, dietary iron overload caused a more pronounced change of gene expression responsive to oxidative stress. In conclusion, Hfe deficiency caused a previously unrecognized increase in gene expression of hepatic acute phase proteins and duodenal digestive enzymes.


Introduction
Iron plays crucial roles in cellular metabolism but, in excess, it can catalyze the formation of free radicals leading to oxidative stress and cell damage [1]. Iron is absorbed in the duodenum, where it crosses the apical and basolateral membranes of absorptive enterocytes to enter the blood stream [2]. There is no regulated mechanism of iron excretion, and thus the absorption of iron must be tightly regulated to maintain iron balance. HFErelated hereditary hemochromatosis (HH, OMIM-235200) is an autosomal recessive disorder in which absorption of iron is inappropriately high [3,4]. HH is characterized by high transferrin saturation and low iron content in macrophages. Iron is deposited primarily in the parenchymal cells of various organs, particularly the liver, but also the pancreas, heart, skin, and testes, resulting in tissue damage and organ failure. Clinical complications in untreated HH patients include hepatic fibrosis, cirrhosis, hepatocellular carcinoma, diabetes, cardiomyopathy, hypogonadism, and arthritis [4].
HH is characterized by inappropriately low expression of the iron-regulatory hormone hepcidin [5]. Hepcidin, a small peptide hormone expressed mainly in the liver, is a central player in the maintenance of iron balance [6]. The only known molecule capable of transporting iron out of cells is ferroportin [7][8][9]. This iron exporter is located in the plasma membrane of enterocytes, reticuloendothelial cells, hepatocytes, and placental cells [7]. Hepcidin binds to ferroportin and induces its internalization and degradation, therefore suppressing the transport of iron into the circulation [10]. The expression of hepcidin is induced by increased iron stores and inflammation, and suppressed by hypoxia and anemia [11,12].
Mice homozygous for a null allele of Hfe (Hfe 2/2 ) provide a genetic animal model of HH [13]. There are several animal models of iron overload based on administration of exogenous iron [14]. According to the route of iron delivery, these can be divided into two main types: enteral (i.e. dietary) and parenteral models. For example, dietary supplementation with carbonyl iron in mice reproduces the HH pattern of hepatic iron loading, with predominantly parenchymal iron deposition [14]. Although both Hfe 2/2 mice and carbonyl iron-fed mice develop iron overload, there are important differences between these two models. Hfe 2/2 mice lack Hfe protein and therefore have decreased expression of hepcidin [15,16], while mice with dietary iron overload express functional Hfe protein and their hepcidin expression is elevated [12].
Current RNA microarray technology allows expression profiling of the whole transcriptome. This methodology has been used to explore the effects of Hfe gene disruption on mRNA expression in the liver and duodenum, two organs with crucial roles in iron metabolism [17]. In the present study, we used this approach to study gene expression in the liver and duodenum of Hfe 2/2 mice and wild-type mice, with or without dietary iron overload. This allowed the identification of genes whose expression is changed during iron overload and those genes whose expression is differentially influenced by lack of Hfe protein.

Results
We used global microarray analysis to study gene expression in the liver and duodenum of Hfe 2/2 mice and carbonyl iron loaded mice, and comparing it with that of wild-type mice fed a standard diet. This approach allowed the identification of genes whose expression is changed during iron overload and those genes whose expression is differentially influenced by lack of Hfe protein. All the mice used were males and all had the same genetic background (C57BL/6).

Hepatic transcriptional response to Hfe deficiency and dietary iron overload
Hepatic RNA from 3 Hfe 2/2 mice and 2 wild-type mice was subjected to microarray analysis. The Pearson correlation coefficient between the knock out mice and between the controls was in both cases 0.989. The results revealed 86 induced genes and 65 repressed genes, using a cutoff value of 61.4-fold (Table 1 and Dataset S1). This cutoff value has been proposed as an adequate compromise above which there is a high correlation between microarray and Q-RT-PCR data, regardless of other factors such as spot intensity and cycle threshold [18]. The fold-changes ranged from 9.83 to 23.47. Functional annotation of the gene lists highlighted the biological processes that may be modified by Hfe deficiency. This analysis revealed enrichment of heat shock proteins and proteins related to inflammatory responses or antigen processing and presentation, among others (Table 2).
Another microarray experiment was performed using hepatic RNA from 3 mice with dietary iron overload and 2 mice fed a standard diet. The similarity between samples from individual mice was measured as the Pearson correlation coefficient, which was 0.989 between iron overloaded mice and 0.991 between control mice. The expression of 123 genes was upregulated and that of 95 genes was downregulated, applying a cutoff value of 61.4-fold (Table 1 and Dataset S2). The fold-changes ranged between 13.58 and 27.46. The list of regulated genes was functionally annotated (Table 3), showing enrichment of cyto-chrome P450 proteins as well as others involved in glutathione metabolism, acute-phase response, organic acid biosynthetic process and cellular iron homeostasis, among others.
There were 11 upregulated and 7 downregulated genes that were affected by both Hfe deficiency and dietary iron overload in similar fashion, while 27 genes were regulated in opposite directions by these two conditions in the liver (Table 4). In some cases, several genes belonging to the same gene family showed divergent regulation (e.g., Saa1, Saa2, Saa3) with upregulation in Hfe 2/2 mice and downregulation by dietary iron overload.
Altered expression of iron-related genes in the liver. The expression of 3 iron-related genes was altered in the liver of Hfe 2/2 mice. The expression of Hamp1 and Tfrc was decreased and that of Lcn2 was induced. We confirmed these results using Q-RT-PCR, and also tested the expression of Hamp2, which was downregulated ( Figure 1). Dietary iron overload changed the expression of 5 ironrelated genes in the liver. The expression of Hamp1, Hamp2, Lcn2 and Cp were upregulated using both microarray analysis and Q-RT-PCR, while Tfrc expression was down-regulated by 1.7-fold ( Figure 2).

Confirmation of hepatic microarray results by Q-RT-
PCR. Microarray analysis for the expression of several genes was confirmed by performing Q-RT-PCR on hepatic samples from 5 Hfe 2/2 mice, 4 wild-type control mice, 5 iron-fed mice and 4 mice fed a standard diet. For this purpose, we selected iron-related genes and others whose expression was substantially altered in the experimental groups. A total of 29 results from the hepatic microarray data, corresponding to 24 different genes, were tested by Q-RT-PCR, and 27 (93.1%) of them showed concordant results by these two methods (Figures 1 and 2). Changes in Foxq1 and Dmt1 expression were false-positives in the microarray analysis for Hfe 2/2 mice and dietary iron overload, respectively. The upregulation of Ltf expression by dietary iron overload observed by microarray analysis could not be confirmed by Q-RT-PCR because the expression levels in samples from all but one of the treated mice and all control mice were below the detection threshold.
Duodenal gene expression response to Hfe deficiency and dietary iron supplementation Microarray analysis of duodenal RNA from 2 Hfe 2/2 mice and 2 wild-type mice revealed that the expression of 143 genes was upregulated and that of 30 genes was downregulated when a cutoff value of 61.4-fold was used (Table 1 and Dataset S3). The foldchanges ranged from 15.67 to 23.14. The Pearson correlation coefficient between knockout mice and between controls was 0.976 and 0.971, respectively. Functional categories overrepresented among the genes regulated by Hfe deficiency included proteins with endopeptidase activity, and others involved in lipid catabolism and antimicrobial activity ( Table 5).  Global transcriptional regulation was also studied in the duodenum of mice fed an iron-supplemented diet, using 3 treated mice and 2 controls. The Pearson correlation coefficient was 0.985 between treated mice and 0.983 between controls. The expression of 49 genes was induced and 59 genes were repressed, applying a cutoff value of 61.4-fold (Table 1 and Dataset S4). The foldchanges ranged between 6.07 and 25.64. Functional annotation of the gene list evidenced enrichment of genes involved in glutathione metabolism, antigen processing and presentation and inflammatory response, among others (Table 6).
We identified genes whose expression was affected by both Hfe deficiency and dietary iron supplementation in the duodenum. There were 4 genes whose expression was induced in both conditions, 3 genes whose expression was decreased, and 4 genes with opposite regulation (Table 7).
Altered expression of iron-related genes in the duodenum. In the duodenum of Hfe 2/2 mice, Hamp2 expression was increased by 2.7-fold using microarray analysis. However, this could not be confirmed by Q-RT-PCR, because Hamp2 mRNA levels in the samples from wild-type mice and in one Hfe 2/2 sample were below the detection threshold. In mice fed the ironsupplemented diet, the duodenal expression of Tfrc was downregulated and that for Hmox1 was upregulated: both of these results were validated by Q-RT-PCR ( Figure 3).

Confirmation of duodenal microarray results by Q-RT-
PCR. Q-RT-PCR analyses were done on duodenal RNA samples from 5 Hfe 2/2 mice, 4 wild-type control mice, 5 ironfed mice and 4 mice fed a standard diet in order to confirm the microarray results. The mRNA expression of a total of 17 different genes was tested and 16 (94.1%) showed concordant results between microarray analysis and Q-RT-PCR (Figures 3 and 4). The sole discrepant result concerned the expression of Ddb1 that was downregulated according to microarray analysis, while Q-RT-PCR revealed a slight induction (1.25-fold) of expression.

Discussion
The goal of this study was to explore and compare the genomewide transcriptome response to Hfe deficiency and dietary iron overload in murine liver and duodenum. This approach allowed the identification of genes whose expression is changed during iron overload and those genes whose expression is differentially influenced by lack of Hfe protein. The global transcriptional response to Hfe deficiency has been explored previously in the liver and duodenum of two mouse strains [17]. However, it is notable that only a few analogous changes in gene expression are seen when comparing our data with those of the previous study, even for mice of the same genetic background. Two other reports have explored expression of selected genes by using dedicated arrays in Hfe 2/2 mice and in mice with secondary iron overload produced by intraperitoneal injection of iron-dextran [19,20]. In one study, duodenum and liver samples were analyzed using an array of ironrelated genes [19]. The results for duodenal gene expression in Hfe 2/2 mice have no concordance with ours. Regulation of hepatic gene expression, on the other hand, is similar for several genes, such as Hamp1, Tfrc and Mt1. The second report focused on gene expression in the duodenum [20], and again, there is little concordance between their observations and ours. The lack of agreement between these studies is probably due to differences in the animal models (parenteral vs. enteral iron loading; mouse strains) and in the microarray methodology.
The hepatic expression of acute phase proteins (APPs) can be induced by inflammatory mediators such as interleukin-6. Interestingly, the liver of Hfe 2/2 mice has upregulated expression of APPs such as serum amyloids, lipocalins and orosomucoids. Notably, the expression of serum amyloid genes (Saa1, Saa2, Saa3) was upregulated in the Hfe 2/2 mice compared to being downregulated in dietary iron overload, suggesting that Hfe deficiency induces this gene expression by an iron-independent mechanism. However, hepatic interleukin-6 mRNA expression was not significantly changed by Hfe deficiency, so the potential involvement of this cytokine in the observed upregulation of APPs remains uncertain.
Lipocalin2 (human Ngal from neutrophil gelatinase-associated lipocalin) is an APP with antimicrobial properties through a mechanism of iron deprivation by siderophore binding [21]. It can donate iron to various types of cells [22,23] and seems to be capable of intracellular iron chelation and iron excretion [24]. Furthermore, a recent study has shown that lipocalin2 is an adipokine with potential importance in insulin resistance associated with obesity [25]. We observed that Lcn2 expression is increased in the liver of both Hfe 2/2 mice and those with dietary iron overload, suggesting that this induction is iron-related.
Dietary iron overload of the liver led to increased expression of both hepcidin genes (Hamp1, Hamp2) as previously reported [26,27], and these results were verified by Q-RT-PCR. In the liver of Hfe 2/2 mice, Hamp1 expression was downregulated as expected [15,16,19]. We also examined the levels of Hamp2 mRNA by Q-RT-PCR and found a -1.92-fold change. The low expression of hepatic Hamp1 in Hfe 2/2 mice is likely responsible for the increased iron absorption and low microphage iron content in these mice [15,16,19].
Inhibitor of DNA-binding/differentiation proteins, also known as Id proteins, comprise a family of proteins that heterodimerize    [28]. Expression of Id1, 2, and 3 is increased during liver disease, with levels that escalate as liver disease progresses from hepatitis to cirrhosis. In hepatocellular carcinoma, high expression is observed in well-differentiated tumors, and it decreases as the tumor cells become undifferentiated [29]. In light of these findings, it has been suggested that Id1, 2, and 3 may play a role in the early stages of hepatocarcinogenesis. Given that, it is notable that we found that the expression of Id1, 2, 3, and 4 was increased in the liver of mice with dietary iron overload, but was unaffected in Hfe 2/2 mice. Increased hepatic expression of Id1 mRNA has previously been reported in mice fed an iron-supplemented diet [30]. The same study showed upregulation of the gene for bone morphogenetic protein 6 (Bmp6) in the same experimental mice. Recent work demonstrates that Bmp6 is a key player in the signalling pathway that controls hepcidin expression [31]. Unexpectedly, upregulation of hepatic Bmp6 mRNA expression by dietary iron overload was not evident in the current study. The gene expression of several heat shock proteins was downregulated in the liver and duodenum by both Hfe deficiency and dietary iron overload, with a considerably greater number of these genes downregulated in the liver of Hfe2/2 mice. Although these genes are induced under certain stress conditions, such as heat shock and ischemia-reperfusion, their expression is decreased by iron overload [19,27,32]. Currently, the physiological implications of this downregulation are unknown.
Our results indicate that disruption of the Hfe gene induces the expression of many genes in the duodenum coding for digestive enzymes, such as elastases, carboxypeptidases, trypsins, chymotrypsins, amylases, and lipases. In contrast, feeding mice with an iron-supplemented diet did not affect the expression of any of these genes. The upregulation of gene expression for digestive enzymes in Hfe 2/2 mice is surprising because overexpression of these enzymes has not been associated with HH.
A common feature of the duodenal response to both Hfe deficiency and dietary iron overload was the transcriptional repression of genes involved in antimicrobial activities, such as cryptdins. In mice fed an iron-supplemented diet, there was also a decrease in mRNA expression for genes involved in antigen processing and presentation, such as some genes of the MHC class II family.
The solute carrier molecules constitute a large family of proteins involved in membrane transport of diverse molecules. The gene expression of many family members was affected by Hfe deficiency or dietary iron overload. In the duodenum, the expression of the sodium-coupled neutral amino acid transporter Slc38a5 was induced in Hfe 2/2 mice and repressed in mice fed an ironsupplemented diet. In the liver, the expression of Slc46a3 was upregulated in Hfe 2/2 mice. This gene belongs to the Slc46 subfamily of heme transporters. It is thus a close relative of Slc46a1 (also known as HCP1), a recently identified, although controversial, heme transporter [33,34]. The iron transporter Dmt1, encoded by Slc11a2, contains an iron-responsive element (IRE) in the 39UTR of its mRNA. This permits the regulation of Dmt1 mRNA levels according to the cellular labile iron pool by mediation of the iron regulatory proteins, IRP1 and IRP2. Under iron-replete conditions, IRP activity is reduced rendering the Dmt1 mRNA vulnerable to degradation. The opposite is true under irondeficient conditions, which is believed to be the situation inside the enterocytes of HH patients and Hfe 2/2 mice [35,36]. Accordingly, in some studies, increased expression of Dmt1 has been observed in the duodenum of HH patients [37] as well as in Hfe 2/2 mice [38]. However, we did not find a significant change in the expression of Dmt1 in the Hfe 2/2 duodenum. This may be explained by the inability of our microarray probes and PCR primers to discriminate between IRE-positive and IRE-negative transcripts.
The post-transcriptional regulation of Tfrc (transferrin receptor 1) by iron is also mediated through the IRE/IRP system [39]. Tfrc is involved in the uptake of transferrin-bound iron by cells. Analogous to our observations, suppression of Tfrc expression in the duodenum [40] and liver [27] of mice fed an ironsupplemented diet, and in the liver of Hfe 2/2 mice [19] has been reported previously. Our microarray analysis indicates that the expression of Tfrc was not significantly changed in the duodenum of Hfe 2/2 mice, a result that agrees with a previous report [19].
Excess free iron increases oxidant production [1]. Subsequently, some antioxidant defense mechanisms are upregulated in order to provide resistance to iron-related toxicity. It is notable from our data that this response is elicited in both liver and duodenum, as seen in the upregulation of glutathione S-transferase genes. Interestingly, dietary iron overload seems to induce a stronger response than Hfe deficiency, especially in the regulation of enzymes involved in glutathione-related detoxification of reactive intermediates.   In conclusion, Hfe deficiency results in increased gene expression of hepatic APPs and duodenal digestive enzymes. In contrast, dietary iron overload causes a more pronounced change of gene expression responsive to oxidative stress.

Ethics Statement
The animal protocols were approved by the Animal Care and Use Committees of Saint Louis University and the University of Oulu (permission No 102/05).

Animal care and animal models
Five male C57BL/6 mice homozygous for a disruption of the Hfe gene and 4 male wild-type control mice were fed a standard rodent diet (250 ppm of iron) and sacrificed at approximately 10 weeks of age. The generation of the Hfe 2/2 mice has been described elsewhere [13]. In addition, 5 male C57BL/6 mice fed an iron-supplemented diet (2% carbonyl iron) and 4 male control mice fed a standard diet (200 ppm of iron) for 6 weeks were used [27]. The mice with dietary iron overload had a hepatic iron concentration that was approximately 2.5 times higher than the Hfe 2/2 mice. The duodenum and liver samples were immediately collected from anesthetized mice and immersed in RNAlater solution (Ambion, Huntingdon, UK).

RNA isolation
Total RNA extraction and quality control have been described previously [27].

Microarray analysis
All microarray data reported in the present article are described in accordance with MIAME guidelines, have been deposited in NCBI's Gene Expression Omnibus public repository [41], and are accessible through GEO Series accession number GSE17969 [42]. Microarray experiments were performed in the Finnish DNA Microarray Centre at Turku Centre for Biotechnology using Illumina's Sentrix Mouse-6 Expression Beadchips. Duodenal and liver RNA samples from 3 Hfe 2/2 mice and 3 mice with dietary iron overload were used. As controls, RNA samples from the duodenum and liver of 4 wild-type mice (2 controls of the Hfe 2/2 mice and 2 controls of the mice with dietary iron overload) were used. All 10 samples were analyzed individually. The amplification of total RNA (300 ng), in vitro transcription, hybridization and scanning have been described before [27].

Data analysis
Array data were normalized with Chipster (v1.1.1) using the quantile normalization method. Quality control of the data included non-metric multidimensional scaling, dendrograms, hierarchical clustering, and 2-way clustering (heat maps). These analyses showed that data from one of the three duodenal samples from Hfe 2/2 mice were highly divergent from the other two. Thus, this sample was excluded from further analyses. The data were then filtered according to the SD of the probes. The percentage of data that did not pass through the filter was adjusted to 99.4%, implicating a SD value of almost 3. At this point, statistical analysis was performed using the empirical Bayes t-test for the comparison of 2 groups. Due to the small number of samples, the statistical results were considered as orientative and thus no filtering was applied to the data according to p-values. The remaining 280 probes were further filtered according to fold-change with 61.4 as cut-off values for up-and down-regulated expression, respectively. The functional annotation tool DAVID (Database for Annotation, Visualization and Integrated Discovery) [43,44] was used to identify enriched biological categories among the regulated genes as compared to all the genes present in Illumina's Sentrix Mouse-6 Expression Beadchip. The annotation groupings analyzed were: Gene Ontology biological process and molecular functions, SwissProt Protein Information Resources keywords, SwissProt comments, Kyoto Encyclopedia of Genes and Genomes and Biocarta pathways. Results were filtered to remove categories with EASE (expression analysis systematic explorer) scores greater than 0.05. Redundant categories with the same gene members were removed to yield a single representative category.  Quantitative Reverse-Transcriptase PCR For this analysis, duodenal and liver RNA samples from 5 mice of each experimental group (Hfe 2/2 and dietary iron overload) and 4 mice from each control group (wild-type and normal diet) were used. Exceptionally, for the analysis of mRNA expression in the duodenum of Hfe 2/2 mice, only 4 samples were used. RNA samples (5 mg) were converted into first strand cDNA with a First Strand cDNA Synthesis kit (Fermentas, Burlington, Canada) using random hexamer primers. The relative expression levels of target genes in the duodenum and liver were assessed by Q-RT-PCR using the LightCycler detection system (Roche, Rotkreuz, Switzerland). The reaction setup, cycling program, standard curve method and primer pairs for Angptl4, Dnajb1 and Tfrc have been described before [27]. Mouse Hamp1 and Hamp2 primers have also been characterized previously [26]. The primer sets for the other target genes (Dataset S5) were designed using Primer3 [45], based on the complete cDNA sequences deposited in GenBank. The specificity of the primers was verified using NCBI Basic Local Alignment and Search Tool (BLAST) [46]. To avoid amplification of contaminating genomic DNA, both primers from each set were specific to different exons, when possible. Each cDNA sample was tested in triplicate. The mean and SD of the 3 crossing point (Cp) values were calculated for each sample and a SD cutoff level of 0.2 was set. Accordingly, when the SD of the triplicates of a sample was greater than 0.2, the most outlying replicate was excluded and the analysis was continued with the two remaining replicates. Using the standard curve method, the Cp values were then transformed by the LightCycler software into copy numbers. The expression value for each sample was the mean of the copy numbers for the sample's replicates. This value was normalized by dividing it by the geometric mean of the 4 internal control genes, an accurate normalization method [47]. The normalization factor was always considered as a value of 100 and the final result was expressed as relative mRNA expression level.

Statistical analyses
We performed statistical analyses of the microarray data using the empirical Bayes t-test for the comparison of 2 groups, and the p-values are shown in supplementary datasets S1-S4. For the Q-RT-PCR results, we used the Mann-Whitney test to evaluate differences in group values for Hfe 2/2 mice vs. wild-type mice and mice with dietary iron overload vs. untreated mice. Due to the small sample sizes, the statistical significance is only considered as orientative. Values are expressed as mean6SD.

Supporting Information
Dataset S1 List of genes differentially expressed in the liver of Hfe knockout mice Found at: doi: 10