The Molecular Circuitry Regulating the Switch between Iron Deficiency and Overload in Mice*

Recent positional cloning of the radiation-induced polycythaemia (Pcm) mutation revealed a 58-bp microdeletion in the promoter region of ferroportin 1 (Fpn1), the sole cellular iron exporter identified to date. Here we report a molecular definition of the regulatory mechanisms governing the dynamic changes in iron balance in Pcm heterozygous mice between 3 and 12 weeks of age. Hepatic and/or duodenal response patterns of iron metabolism genes, such as Trfr, cybrd1, and Slc11a2, explained the transition from early postnatal iron deficiency to iron overload by 12 weeks of age. A significant delay in developmental up-regulation of hepcidin (Hamp), the pivotal hormonal regulator of iron homeostasis, correlated with high levels of Fpn1 expression in hepatic Kupffer cells and duodenal epithelial cells at 7 weeks of age. Conversely, upon up-regulation of Hamp expression at 12 weeks of age, Fpn1 expression decreased, indicative of a Hamp-mediated homeostatic loop. Hamp regulation due to iron did not appear dependent on transcription-level changes of the murine homolog of Hemojuvelin (Rgmc). Aged cohorts of Pcm mice exhibited low levels of Fpn1 expression in the context of an iron-deficient erythropoiesis and profound iron sequestration in reticuloendothelial macrophages, duodenum, and other tissues. Thus, similar to the anemia of chronic disease, these findings demonstrate decreased iron bioavailability due to sustained down-regulation of Fpn1 levels by Hamp. We conclude that regulatory alleles, such as Pcm, with highly dynamic changes in iron balance are ideally suited to interrogate the genetic circuitry regulating iron metabolism.

In mammals, iron bioavailability for erythropoiesis and other vital organismal functions is regulated at three principal sites: placental or duodenal uptake, release from hepatic stores, and recycling of scavenged iron from senescent red blood cells via reticuloendothelial (RES) 3 macrophages (1). Iron recycling from macrophages generates the vast majority of iron for daily consumption, amounting to ϳ22 mg of iron per day under steady state conditions in humans (2). Ferroportin 1 (Fpn1; also known as MTP1, IREG1, SLC11A3, SLC39A1) plays a piv-otal role at all three sites and functions as the sole cellular iron exporter identified to date (3)(4)(5).
Both cellular and systemic hormonal mechanisms impinge upon Fpn1 regulation. Predominant modes of cellular regulation include transcriptional as well as translational control. Several studies document changes in Fpn1 mRNA levels and transcription rates in mice and cultured cells under conditions of iron overload or deficiency (6 -14). Translational regulation involves a highly conserved sequence in the 5Ј-untranslated region of Fpn1 mRNA known as the iron-responsive element (1). Trans-acting iron regulatory proteins bind to the ironresponsive element stem-loop structure, inhibiting Fpn1 mRNA translation under low intracellular iron conditions in tissue culture cells (6,7,15). Hepcidin (Hamp), the principal hormonal regulator of iron homeostasis (1), controls Fpn1 expression levels at the cell surface by a posttranslational mechanism. Upon Hamp binding to cell surface-bound Fpn1, the iron exporter becomes internalized and degraded in lysosomes (16). Importantly, this implicates the Hamp-Fpn1 axis in a homeostatic loop wherein, under conditions of high intracellular iron in hepatic stores, Hamp-mediated degradation of Fpn1 decreases duodenal iron uptake and macrophage iron release, preventing organismal iron overload.
Coding region mutations in human FPN1 cause the autosomal dominant ferroportin disease (also referred to as hemochromatosis, type IV) (17). Furthermore, two mutant mouse models for Fpn1 function were recently described (3,5). Constitutive or conditional deletion of several transmembrane domains confirmed Fpn1 function in maternal-to-fetal iron transport, export from duodenal enterocytes, and iron release from RES macrophages (5). The other mouse model involves a regulatory allele of Fpn1 generated by radiation mutagenesis. Positional cloning identified a 58-bp microdeletion in the Fpn1 promoter in polycythaemia (Pcm) mutant mice (3). Depending on the developmental stage, tissue, and genotype, the microdeletion caused dynamic dysregulation of Fpn1 expression, resulting in both hypo-and hypermorphic phenotypes (3,4). During late gestation, decreased Fpn1 protein expression in placental syncytiotrophoblast cells resulted in a severe neonatal iron deficiency and a hypochromic, microcytic anemia in Pcm homozygotes, providing the first definitive evidence for Fpn1 function in maternal-to-fetal iron transport in mammals (4). Postnatally, aberrant transcription initiation eliminated the iron-responsive element in the 5Ј-untranslated region of Fpn1 mRNA, increasing hepatic and duodenal Fpn1 protein levels (3). The latter governed augmentation of intestinal iron uptake, reversing the perinatal iron deficiency to a tissue iron overload by 12 weeks of age. Strikingly, the majority of Pcm heterozygous animals displayed a transient erythropoietin-dependent polycythaemia with peak hematocrits of up to 80% during young adulthood, eponymous of the mutant strain (3).
Here we report the molecular mechanisms underlying the iron homeostasis defects in Pcm mice. The "IronChip" microarray platform and real-time quantitative RT-PCR accurately defined the hepatic and duodenal gene response patterns that governed the transition from early postnatal iron deficiency to tissue iron overload in young adult Pcm mice. In support of a Hamp-Fpn1 homeostatic loop in vivo, Pcm mice manifested an inverse correlation between Hamp mRNA levels and Fpn1 expression in RES macrophages and duodenal epithelial cells. Aged cohorts of Pcm mutant mice exhibited an iron-deficient erythropoiesis in the context of profound iron sequestration in RES macrophages and other cell types, constituting the end point of Pcm disease pathogenesis.

EXPERIMENTAL PROCEDURES
Mice and Genotyping-Pcm mice on a partially congenic A/J background (N5 and later) were used for analysis and genotyped as described previously (3). Cohorts of Pcm and wild-type animals were aged to 8 -19 months and fed a standard chow. All animal experiments in this study were approved by the Institutional Animal Care and Use Committee of Baylor College of Medicine.
Microarray Analysis-Total RNA was isolated from liver and duodenum samples using TRIzol reagent (Invitrogen) and prepared according to standard methods. Pools of total liver RNA from 6 Pcm heterozygotes and 6 wild-type mice at 3, 7, and 12 weeks of age were analyzed using the Mouse Version 6.0 of the IronChip as described previously (11). Likewise, pools of total duodenal RNA from 6 Pcm heterozygotes and 6 wild-type mice were analyzed at 7 and 12 weeks of age. Expression values were calculated from dye swap experiments (18). Genes were represented by either single or multiple clones on the microarray platform. In the latter case, the average ratios and standard deviations were determined. The entire data set representing the expression values of all genes represented on the IronChip cDNA microarray platform will be submitted to Array express (European Bioinformatics Institute).
Histology and Immunohistochemistry-Liver and duodenal samples from various stages were processed as described previously (3). A polyclonal Fpn1 antibody was generated in rabbit against oligopeptide sequence GPDEKEVTDENQPNTS at the carboxyl terminus of Fpn1 and immunoaffinity purified. Immunohistochemistry for Fpn1 and F4/80 (clone CI:A3-1; Serotec) was performed as described (4). For Trfr antigen retrieval, sections were boiled in 0.01 M citric acid, pH 6.0, and blocking was achieved with horse serum. Primary antibody incubation was performed overnight at 4°C using mouse ␣-TfR at 1:1000 dilution (Invitrogen). Prussian blue staining for iron was accomplished using the Accustain iron staining kit (Sigma) according to the manufacturer's protocol. For comparison of protein expression levels, wild-type and Pcm mutant samples were mounted side by side on the same slide. Images were acquired with a Sony 085 CCD color RGB sensor digital camera mounted on a Zeiss Axioplan 2 microscope.
Iron Determination and Blood Cell Analyses-Non-heme tissue and serum iron were determined as described previously (3). Hematocrit (Hct) measurements, expressed as a percentage, were obtained by standard microcapillary determination. Red cell parameters, including mean corpuscular volume and mean corpuscular hemoglobin concentration, as well as white blood cell (WBC) counts were determined on the ADVIA 120 hematology system (Bayer).
Statistical Analyses-All data are reported as the mean Ϯ S.D. All comparisons were made versus wild-type cohorts and analyzed for significant differences using the Student's unpaired t-test.

RESULTS
To elucidate the consequences of dynamic Fpn1 dysregulation on the genetic circuitry governing iron metabolism, oxidative stress, hypoxia, and inflammation we used a highly sensitive and accurate cDNA microarray platform, the IronChip (11). The analysis focused on gene expression in liver and duodenum, respectively, as sites of iron storage and absorption at 3, 7, and 12 weeks of age. In the context of the striking polycythaemia phenotype by young adulthood, heretofore not described in animal models of dysregulated iron metabolism, molecular analyses were directed toward Pcm heterozygotes. In comparison with previous microarray studies on mouse models of iron balance abnormalities (11,19), Pcm mice presented discrete but highly reproducible differences in hepatic and duodenal gene expression during postnatal development (Fig. 1). To validate the microarray findings based on pooled RNA, the individual samples were subjected to real-time RT-PCR analysis using the most relevant probes. The average changes of differential gene expression along with the respective standard deviations are shown in supplemental Table S1.
The Hamp-Fpn1 Homeostatic Loop-Neither microarray analysis nor quantitative RT-PCR detected statistically significant differences in duodenal and hepatic Fpn1 transcript levels between Pcm heterozygous and wild-type mice at 3, 7, and 12 weeks of age ( Figs. 1 and 2, A and B).  Importantly, hepatic Hamp mRNA levels were significantly reduced in Pcm heterozygotes at 7 weeks of age ( Figs. 1 and 2C). However, this reflected a delay in developmental Hamp up-regulation because at 12 weeks of age Hamp mRNA levels in Pcm heterozygotes were significantly higher (microarrays) or similar (RT-PCR) to wild-type levels ( Figs. 1 and 2D). Thus, delayed developmental up-regulation of Hamp is conducive to persistent elevation in Fpn1 protein expression in polycythaemic 7-week Pcm heterozygotes (3). These results underscore the responsiveness of Hamp regulation to the dynamic changes in hepatic iron levels in the context of the Fpn1 regulatory allele. Therefore, to the best of our knowledge, the Pcm mouse model provides the strongest evidence to date for the Hamp-Fpn1 homeostatic loop in vivo.
Response Patterns of Iron Metabolism Genes-In good agreement with the severe neonatal iron deficiency in Pcm mice (3,4), microarray analysis detected increased hepatic Trfr (transferrin receptor) mRNA expression in 3-and 7-week-old Pcm heterozygotes (Fig. 1). Quantitative RT-PCR showed no statistically significant difference between Pcm heterozygotes and wild-type controls at 7 weeks (Fig. 2E). However, both the IronChip and quantitative RT-PCR demonstrated a strong decrease in Trfr mRNA expression by 12 weeks of age ( Figs. 1 and 2F). This result is consistent with destabilization of Trfr mRNA by a mechanism dependent on iron regulatory proteins (1) in response to hepatic iron accumulation in Pcm liver at 12 weeks of age (3), obviating Trfrdependent cellular iron uptake. Likewise, increased Hmox1 (heme oxygenase 1) mRNA levels in 7-and 12 week-old Pcm heterozygotes can be explained by the increased need for heme catabolism in polycythaemic animals, heralding hepatic iron accumulation (Fig. 1). Consistent with marked systemic iron requirement and low hepatic Hamp mRNA expression, duodenal Trfr, cybrd1 (duodenal cytochrome b), and Slc11a2 (solute carrier 11a2; also known as Nramp2, Dmt1) mRNA levels of Pcm heterozygotes at 7 weeks of age were increased (Figs. 1 and 2G). Increased expression of these iron transporters reverted to wildtype levels at 12 weeks of age ( Figs. 1 and 2H). Therefore, coordinated shifts in transcriptional expression patterns of these genes are sufficient to explain the transitory iron loading in Pcm heterozygotes.
Response Patterns of Selected Iron Metabolism Proteins in Liver and Duodenum-At 3 weeks of age, Trfr was expressed at moderate levels in wild-type liver, whereas Pcm heterozygotes displayed significant upregulation of this protein (Fig. 3A; compare with mRNA results in Fig.  1). Inspection under high magnification suggested significant augmentation of Trfr levels in Kupffer cells and sinusoidal endothelial cells, contrasting with a more subtle increase in expression in hepatocytes (data not shown). At 12 weeks, Pcm heterozygotes demonstrated significantly lower Trfr levels compared with wild-type liver (Fig. 3B), consistent with the mRNA results (Fig. 1). Elevated Trfr protein expression in duodenum from 7-week-old Pcm heterozygotes restored Trfr to levels indistinguishable from wild-type littermates at 12 weeks of age (Fig.  3, C and D; compare with mRNA results in Fig. 1). Thus, immunohistochemistry corroborated the Trfr mRNA response patterns in liver and duodenum. At 7 weeks of age, Pcm heterozygous duodenum exhibited significant elevation of Fpn1 expression compared with wild-type littermates (Fig. 3E). Increased expression was particularly evident at the basolateral membrane of the epithelial cells. Conversely, at 12 weeks of age, Pcm heterozygotes were remarkable for markedly lower Fpn1 expression levels compared with wild-type (Fig. 3F). Because Fpn1 mRNA levels were indistinguishable between Pcm heterozygotes and wild-type at both 7 and 12 weeks of age (Fig. 1), the changes in protein levels strongly supported the notion of Hamp-mediated post-translational regulation of Fpn1 expression.

Down-regulation of Fpn1 Expression and Reticuloendothelial Iron
Overload-We previously demonstrated a significant augmentation of Fpn1 protein expression in Pcm mutant liver during early postnatal development (3). By immunohistochemistry, increased Fpn1 expression localized predominantly to punctate, stellate-shaped cells at 3 weeks of age (Fig. 4A). Based on a similar staining pattern for F4/80 (data not shown), the antibody to which recognizes a macrophage-restricted glycoprotein of the epidermal growth factor 7 family (22), marked upregulation of Fpn1 in Pcm mutants localized to Kupffer cells. Likewise, persistent elevation of Fpn1 protein expression at 7 weeks of age localized predominantly to Kupffer cells in polycythaemic Pcm heterozygotes (data not shown). In contrast, the small subset of Pcm heterozygotes with normal hematocrit as well as Pcm homozygotes displayed wild-type levels of hepatic Fpn1 expression (data not shown). By 12 weeks of age, the Fpn1 expression pattern was indistinguishable between wild-type and Pcm mutant liver (Fig. 4B). This down-regulation of Fpn1 expression correlated with RES iron accumulation in Pcm homozygous liver by 12 weeks of age (Fig. 4C). Furthermore, hepatic  Fpn1 expression remained near base-line levels in aged Pcm homozygotes (Fig. 4D), whereas significant iron accumulation persisted (Fig.  4E). Large accumulations of iron, consistent with hemosiderosis, were observed in cells staining against F4/80 (Fig. 4F). Notably, the overall number of macrophages appeared similar between aged wild-type and mutant livers (Fig. 4F), and collagen staining revealed no evidence of widespread fibrosis (data not shown).
Red cell parameters and white blood cell counts in aged cohorts of Pcm animals are shown in supplemental Table S2. A modest but significant elevation in WBC was observed in both heterozygotes (Pcm/ϩ WBC 1.2 Ϯ 0.42 10 3 cells/l (n ϭ 10) versus ϩ/ϩ WBC 0.87 Ϯ 0.32 10 3 cells/l (n ϭ 15); p ϭ 0.027) and homozygotes (Pcm/Pcm WBC 1.7 Ϯ 0.99 10 3 cells/l (n ϭ 17); p ϭ 0.0044). This was primarily reflective of an increase in absolute lymphocyte count and, to a lesser degree, neutrophil count. No statistically significant differences in monocyte, eosinophil, or basophil counts were observed. In the context of a semidominant defect in spleen development in Pcm animals (4), it is conceivable that the increased presence of circulating white blood cells in aged Pcm mice results from the defects in this significant repository site for immune cells.

DISCUSSION
In recent years, positional cloning of iron disease loci in both humans and mutant mouse models, complemented by transgenic and gene knock-out studies in mice, have identified many genes involved in iron metabolism (1). The challenge now resides in dissecting the function of the encoded proteins and their integration into the regulatory circuits governing cellular and organismal iron homeostasis. These investigations will benefit greatly from the availability of mouse strains, such as Pcm, with regulatory mutations in critical components of the pathways.
The Pcm mouse mutant exhibits the gamut of iron balance disorders, ranging from iron deficiency at birth to tissue iron overload by young adulthood (3,4). The present study defined, in molecular terms, the regulatory interferences underlying the dynamic changes in iron homeostasis in Pcm mice. For example, duodenal and/or hepatic expression of Fpn1 and Trfr protein, as well as Hamp, cybrd1 and Slc11a2 mRNA, explained the transition from early postnatal iron deficiency to iron overload in 12-week old Pcm mice. In comparison with previous studies (10,11), these results demonstrate transcriptional responsiveness of the duodenal iron transport system to organismal and/or cellular iron balance. These changes in gene expression are likely to result directly or indirectly from alterations in Hamp signaling. Recently, it was shown that, under conditions of augmented erythropoiesis and decreased Hamp expression, mRNA and protein levels of Slc11a2 and Cybrd1 increased (24). Furthermore, genetically Hamp-deficient mice demonstrated up-regulation of Slc11a2, Cybrd1, and Fpn1 expression (25). Our results indicate that polycythaemic 7-week-old Pcm heterozygotes demonstrate decreased Hamp expression in the context of significant alterations in mRNA levels of iron-related genes, such as iron transporters. In addition, cellular regulation responsive to Fpn1-mediated elevated iron efflux from intestinal enterocytes leading to increased mRNA expression of iron transporters in Pcm heterozygotes must be considered (26).
Genetic evidence suggests that Rgmc, a second causative gene for juvenile onset hereditary hemochromatosis (type II), functions upstream of Hamp (20). Conceivably, profound down-regulation of Hamp in polycythaemic Pcm heterozygotes at 7 weeks of age constitutes an appropriate in vivo context to discern expression differences in putative upstream regulators, including Rgmc. However, mRNA levels of Rgmc were indistinguishable between wild-type and 7-week old Pcm heterozygotes. Similarly, work by Krijt et al. (21) revealed no differences in Rgmc transcript levels following treatment of mice with iron or erythropoietin, known regulators of Hamp expression. Thus, Hamp regulation in response to iron status is not likely to depend on changes in Rgmc transcription. , duodenum (C), and heart (D). Although no differences in non-heme iron levels were observed in brain (E), Pcm homozygotes showed decreased serum iron levels (F). White columns depict wild-type; gray columns Pcm heterozygous; black columns Pcm homozygotes. *, p Ͻ 0.01; **, p Ͻ 0.001; ***, p Ͻ 0.0001.
Recently, it was shown that Hamp expression is regulated developmentally (21). Compared with fetal development, perinatal Hamp expression decreases by the order of several magnitudes, reaching lowest levels by P8. As postnatal development proceeds, Hamp levels gradually return to high levels by adulthood. Fig. 6 depicts a model wherein delayed up-regulation of Hamp in the context of increased Fpn1 mRNA translation from iron-responsive element-less transcripts (3) forms the basis for elevated Fpn1 protein levels, governing Pcm pathogenesis during early postnatal development. Therein, dysregulated Fpn1-mediated duodenal uptake overcompensates for the perinatal iron deficiency and causes organismal iron overload. Upon up-regulation of Hamp by 12 weeks of age, Fpn1 levels decrease in a distinctly post-transcriptional manner, iron uptake ceases, and iron becomes increasingly sequestered in RES macrophages and duodenal epithelial cells. This heralds the Pcm disease end point at a fairly early stage, because aged cohorts of Pcm mutant mice exhibited an iron-deficient erythropoiesis in the context of marked iron sequestration in RES macrophages, duodenum, and other tissues, as well as reduced serum iron levels. Similar to the anemia of chronic disease caused by other disease mechanisms (27), it remains an enigma why signaling through the iron-deficient erythron is insufficient to reverse Hamp up-regulation and thus ameliorate iron sequestration and anemia in aged Pcm mice. Nonetheless, iron-deficient erythropoiesis marks both the beginning and the end point of the hematopoietic defects in Pcm mice. However, whereas embryonic/perinatal anemia results from primary organismal iron deficiency (4), adult Pcm mice develop anemia because of decreased iron bioavailability despite organismal iron overload. The polycythaemia develops at the transition phase between the two disease states, governed by unimpeded erythropoietin signaling (3).
Recent studies have described several mechanisms by which mutations in human FPN could lead to type IV hereditary hemochromatosis, including unresponsiveness to Hamp-mediated degradation, defects in cell surface localization, and dominant-negative effects on cellular processing of wild-type FPN (28 -31). Phenotypic and molecular analyses of Pcm mice support an additional disease mechanism wherein augmented Fpn1-mediated iron uptake earlier in life is followed by RES iron sequestration and microcytic anemia due to homeostatic Hamp signaling. Interestingly, a loss-of-function mutation in Fpn1 did not result in significant hepatic iron overload in heterozygous mice (5). Therefore, to the best of our knowledge, Pcm represents the first in vivo model system demonstrating disease end points commonly detected in type IV hemochromatosis. Additionally, the present results suggest that, due to the dynamic nature of iron homeostasis, analyses of in vivo models should require consideration of various stages in pre-and postnatal development as the phenotypic richness and regulatory intricacies cannot be appreciated by the description of one or two stages alone.