The flatiron mutation in mouse ferroportin acts as a dominant negative to cause ferroportin disease

Hereditary hemochromatosis is a common disorder in humans, characterized by iron overload resulting in tissue injury and ultimately organ failure. Typically, hemochromatosis exhibits an autosomal-recessive pattern of inheritance and is associated with mutations in HFE, hemojuvelin, hepcidin, or transferrin receptor 2.^1,2  Targeted deletion of these genes in the mouse results in hemochromatosis, providing mouse models for most forms of the disease. Hemochromatosis type IV, also referred to as ferroportin (Fpn) disease, results from mutations in the iron transporter ferroportin. Fpn is the only known iron exporter in mammalian cells and is present on the surface of macrophages, intestinal enterocytes, hepatocytes, and placental cells.^3–5  The level of cell surface Fpn is regulated by its interaction with hepcidin, a peptide secreted by the liver in response to iron stores and inflammation. Hepcidin binds to Fpn, inducing its internalization and degradation, thus regulating the export of iron from cells to plasma.⁶ 

Mutations in Fpn lead to iron-overload disease but, in contrast to other forms of hemochromatasis, ferroportin disease exhibits an autosomal-dominant pattern of inheritance.⁷  The disorder has different presentations depending on the Fpn mutation. Mutations leading to Fpn that is not internalized by hepcidin result in iron accumulation in hepatocytes and high transferrin saturation.^8,9  Mutations leading to Fpn that is not appropriately targeted to the cell surface result in iron accumulation in Kupffer cells and low transferrin saturation.^9–11  The mechanism by which the disease mutations exert a dominant effect is unclear. Some groups that study the disease suggest that it results from haploinsufficiency,^10,12  whereas others suggest that the disorder results from a dominant-negative effect of the mutant allele.^9,13  Importantly, all human mutations are missense mutations and mice that are heterozygous for a targeted deletion of Fpn do not show the disease.¹⁴  Treatment for hemochromatosis aims to decrease iron load by repeated phlebotomy and this treatment works well for most patients. Many patients with ferroportin disease, however, become anemic with phlebotomy, highlighting the need for a mouse model to develop better treatments.

We report here on a missense mutation in mouse Fpn that results in a disorder that is identical to classic human ferroportin disease. We show that macrophages isolated from mutant mice have no Fpn on their cell surface and that expression of Fpn constructs containing the missense mutation (H32R) affects the behavior of wild-type Fpn. These results show that Fpn disease is due to a dominant-negative effect of the mutant allele and provide the first mouse model for this disorder.

The ffe mouse line was identified in a screen for recessive ethylnitrosourea (ENU)–induced mutations that cause morphologic abnormalities at embryonic day (E) 12.5.^15–17  The ffe mutation was generated on a C57BL/6J genetic background and backcrossed to C3H/HeJ or 129/SvJ. In a mapping cross of 1078 opportunities for recombination, ffe was mapped between Massachusetts Institute of Technology (mit) simple sequence-length polymorphism (SSLP) markers D1mit213 and D1mit528. For high-resolution mapping, additional polymorphic DNA markers were generated based on nucleotide repeat sequences. D1ski4-L: CCTCTACCACTGCCTATTCTGT; D1ski4-R: ACAGGCTGGACCTGAGCA; D1ski6-L: GTTACCGAGCAACAGCGAAG; D1ski6-R: TGAATGTGCACCTGTCTATGG; D1ski7-L: TTTGGCGATGAATCTTCTGA; D1ski7-R: TGAAAAGATGGCCAATTGCT; D1ski12-L: GGGTTAGAACCAAAAGGGTGA; D1ski12-R: CCAAGAAGCAGGAGTGGGTA; D1ski13-L: GCCTGATGTCTTTGTCATGC; D1ski13-R: ATGGAACTTCAGGGTTGACG; D1ski14-L: TGGACCTGAAAATCATATTATTACA; D1ski14-R: CTGGGTCCCCTCCTTCTTAC. The entire Fpn transcript was sequenced by reverse transcriptase–polymerase chain reaction (RT-PCR; Superscript One-Step RT-PCR; Invitrogen, Carlsbad, CA) using RNA isolated from E10.5 ffe/ffe and C57BL/6 control embryos. Sequencing was confirmed using DNA isolated from 10 additional ffe/ffe mutant embryos.

Livers were fixed overnight in 4% paraformaldehyde, cryopreserved in 30% sucrose, embedded in OCT compound (Tissue-Tek; Electron Microscopy Science, Hatfield, PA) and sectioned at 10 mM. Fixed frozen sections were incubated in hydrochloric acid (10%) and potassium ferrocyanide (5%) as described¹⁸  and counterstained with eosin.

The cloning and expression of mouse Fpn in a cytomegalovirus (CMV)–containing vector (pEGFP-N1 [Clontech, Mountain View, CA] or pCMV-Tag4 [FLAG; Stratagene, La Jolla, CA]) was described previously.⁹  pEGFP-FpnH32R was generated in pFpn-EGFP-N1 by using the QuikChange Site-Directed Mutagenesis Kit (Stratagene), amplified in Escherichia coli and sequence verified. HEK293T cells were maintained in Dulbecco minimal essential media (DMEM) with 10% fetal bovine serum and were transfected with pFpn-GFP and pFpn(H32R)-GFP or pFpn-FLAG using Nucleofector technology (Amaxa, Gaithersburg, MD) according to the manufacturer's directions. Mouse bone marrow macrophages were harvested from femurs and maintained as described previously.¹⁹ 

Erythrophagocytosis was performed as described.³  Macrophages were incubated with IgG-coated red blood cells (RBCs) at a ratio of 20 RBCs/macrophage for 90 minutes. The cultures were then washed free of extracellular RBCs and incubated for 18 hours before measurement of ferritin. Wild-type and ffe/+ macrophages phagocytosed an average of 10 RBCs/macrophage with more than 95% of the cells containing RBCs. There was no difference between wild-type and ffe/+ macrophages in the average number of RBCs/cell or the percent of cells with ingested RBCs.

siRNA oligonucleotide pools, nonspecific and mouse Fpn specific, were obtained from Dharmacon (Lafayette, CO). Mouse bone marrow macrophages were transfected with siRNAs at concentrations up to 100 nM using Oligofectamine reagent (Invitrogen) as described.⁴  At 24 hours after transfection, the cells were trypsinized and plated onto 35-mm plates with or without ferric ammonium citrate (FAC) (10 μM Fe). Cells were grown for 18 hours and then processed for Western blot and ferritin analysis.

Immunofluorescence was performed as described previously.⁹  Fluorescent images were captured on an Olympus BX51 microscope (Olympus, Tokyo, Japan) using an Olympus U-CMAD-Z camera and a 60×/1.30 NA oil-immersion objective lens. Images were acquired using Picture Frame 2.5 software (Olympus America, East Muskogee, OK). Captured images were combined in Adobe Photoshop 6.0 (Adobe Systems, San Jose, CA) to generate multiple-image figures for publication. Fpn was detected using a rabbit anti-Fpn antibody (1:100; a generous gift from Dr David Haile). Transferrin receptor 1 was detected using a mouse anti–human transferrin receptor 1 antibody (CD81 1:100; RDI, Concord, MA) followed by Alexa 594–conjugated goat anti–rabbit IgG (1:750; Molecular Probes, Eugene, OR), Alexa 594–conjugated goat anti–mouse IgG (1:750; Molecular Probes). Mouse anti-FLAG antibody M2 (Sigma, St Louis, MO) was used to detect Fpn-FLAG by immunofluorescence (1:750) followed by Alexa 594–conjugated goat anti–mouse IgG (1:750; Molecular Probes). Cellular protein was extracted with 150 mM NaCl, 10 mM EDTA, 10 mM Tris (pH 7.4), 1% Triton X-100, and a protease inhibitor cocktail (Roche, Indianapolis, IN). Western analysis was performed as described previously⁹  using rabbit anti-Fpn (1:500) and mouse anti–human α-tubulin (1:5000) followed by either peroxidase-conjugated goat anti–rabbit IgG (1:12 500, Jackson ImmunoResearch Labs, West Grove, PA) or peroxidase-conjugated goat anti–mouse IgG (1:12 500, Jackson ImmunoResearch Labs). Mice were bled retro-orbitally and peripheral blood smears were stained with Wright-Giemsa. Ferritin analysis was performed as described.⁶  Transferrin saturation was determined by means of a ferrozine-based iron and total iron-binding capacity assay (Teco Diagnostic, Anaheim, CA). Protein concentration was measured using the BCA assay (Pierce, Rockford, IL). All experiments were repeated a minimum of 3 times.

In a screen for recessive mutations induced by ethylnitrosurea that affect development of the mouse embryo, we identified ffe as a recessive mutation that causes developmental defects on a mixed C3H/HeJ X C57BL/6 background, including neural tube closure defects, microphthalmia, forebrain truncations, generalized edema, severe anemia, and mid-gestation lethality (data not shown). When the ffe mutation was crossed onto the 129/SvJ inbred strain, homozygous mutant embryos showed only severe anemia and mid-gestation lethality (Figure 1A-B). Using a positional cloning strategy, the ffe mutation was mapped to a 2.6-megabase interval on mouse chromosome 1 that contains 16 transcripts, including the Slc40a1 gene encoding the Fpn protein (Figure 1C). Fpn is the only exporter of cellular iron in mammalian cells^5,20,21  and is expressed on placental syncytiotrophoblast cells where it is required for iron transport from the maternal circulation to the fetus.¹⁴  Sequencing of the cDNA encoding Fpn from ffe mutant embryos revealed a c.95A>G nucleotide substitution. The deduced amino acid substitution is H32R in the first putative transmembrane domain (Figure 1D). Most Fpn^null/null embryos die by E7.5¹⁴  because of defects in iron transfer across the visceral endoderm, indicating that Fpn^ffe is a hypomorphic allele of Fpn. It is likely that the more complex phenotype seen in the original mutant was due to the background strain, as the complex phenotype was lost when the mice were bred onto different backgrounds.

The cell surface localization of Fpn is essential for its function,⁶  and the ffe mutation could result in mislocalization of Fpn, thus inhibiting iron transport into the fetal circulation. To determine if the Fpn^ffe mutation results in its mislocalization, this mutation (H32R) was introduced into a plasmid containing a green fluorescent protein (GFP)–tagged Fpn. As shown in Figure 2A, Fpn-GFP expressed in HEK293T cells is localized to the plasma membrane. In contrast, Fpn(H32R)-GFP is predominantly intracellular. To determine if Fpn(H32R)-GFP has iron-exporting activity, cells expressing Fpn-GFP or Fpn(H32R)-GFP were iron loaded by incubation with ferric ammonium citrate (FAC), inducing the expression of the iron storage protein ferritin (Figure 2B). Expression of Fpn-GFP leads to iron export, resulting in decreased ferritin levels,⁶  whereas expression of Fpn(H32R)-GFP results in only a minor decrease in ferritin, indicating that the Fpn^ffe allele has significantly decreased iron-exporting activity.

We have shown that Fpn is a multimer and mutations in Fpn that affect cell surface localization also result in mislocalization of wild-type Fpn.^9,13  To determine if expression of Fpn(H32R)-GFP can change the localization of the wild-type protein, Fpn-FLAG was coexpressed with Fpn-GFP or mutant Fpn(H32R)-GFP and localization of the tagged proteins was determined by immunofluorescence. Coexpression of Fpn-GFP with Fpn-FLAG resulted in plasma membrane localization of both proteins (Figure 2C). In contrast, coexpression of Fpn-GFP(H32R) with Fpn-FLAG resulted in the intracellular localization of GFP and FLAG-tagged proteins, indicating that mutant Fpn can change the localization of wild-type protein. This is specific to Fpn, as expression of Fpn(H32R)-GFP did not affect localization of transferrin receptor 1. To determine if mutant Fpn can affect the iron-exporting activity of wild-type Fpn, Fpn-FLAG was coexpressed with Fpn-GFP or Fpn(H32R)-GFP in iron-loaded HEK293T cells. Coexpression of wild-type proteins resulted in export of cellular iron and decreased ferritin levels (Figure 2D). Cells coexpressing Fpn(H32R)-GFP with Fpn-FLAG had elevated levels of ferritin, indicating that the Fpn^ffe allele (FpnH32R) can act as a dominant negative by changing the localization of wild-type Fpn and its iron-exporting activity.

To determine if endogenous levels of Fpn^ffe also act as a dominant negative in vivo, macrophages were isolated from Fpn^ffe/+ heterozygous mice and localization of endogenous Fpn was examined. Fpn expression in macrophages is induced in response to high levels of cellular iron.³  Little Fpn is expressed in untreated primary cultures of macrophages isolated from wild-type or Fpn^ffe/+ animals (Figure 3A-B). Upon iron loading, high levels of Fpn are detected at the cell surface of wild-type macrophages. Iron loading of macrophages from Fpn^ffe/+ heterozygous mice did not result in expression of Fpn on the cell surface, although equivalent levels of Fpn were detected in wild-type and Fpn^ffe/+ macrophages (Figure 3B). Since Fpn^ffe/+ mice express both a wild-type and mutant allele of Fpn, these results indicate that Fpn^ffe inhibits the cell surface localization of wild-type Fpn.

To determine if macrophages from Fpn^ffe/+ mice are able to export iron, macrophages from wild-type or Fpn^ffe/+ mice were loaded with iron and cellular ferritin levels were measured. Wild-type and Fpn^ffe/+ cells expressed equivalent amounts of ferritin in the absence or presence of FAC. Wild-type macrophages exported cellular iron upon removal of FAC, as indicated by decreased ferritin levels (Figure 3C, black bars); however, macrophages isolated from Fpn^ffe/+ mice were not able to efficiently export iron, and ferritin levels remained high (Figure 3C, gray bars). Similar results were obtained when macrophages were loaded with iron by engulfing RBCs, although the absolute level of ferritin was lower in RBC-fed macrophages (Figure 3C). Together these results indicate that Fpn^ffe acts as a dominant negative in vivo by associating with wild-type Fpn, causing its mislocalization and thus preventing its ability to export iron.

It has been suggested that mutations in Fpn might result in cellular iron overload due to haploinsufficiency^10,12 ; however, mice that are heterozygous for a targeted gene deletion of Fpn do not show evidence of iron-overload disease.¹⁴  To further explore the possibility that haploinsufficiency could explain decreased iron export, we used RNAi to silence mouse Fpn in macrophages. Cells were transfected with nonspecific oligonucleotides or different concentrations of oligonucleotides specific to mouse Fpn. After 24 hours, cells were loaded with iron and levels of Fpn and cellular ferritin were determined 48 hours after transfection. Macrophages showed almost normal levels of iron export even when Fpn protein levels were decreased by 50% (Figure 3D). Reduction in iron-exporting activity (higher levels of ferritin) did not occur until Fpn levels decreased to below 30% of wild-type levels. These results indicate that half normal levels of Fpn, which would be expected in haploinsufficiency, cannot explain the defect in Fpn disease.

We sought to determine if expression of the dominant-negative Fpn^ffe allele in Fpn^ffe/+ heterozygous mice can mimic the human disease. In human patients, Fpn-linked hemochromatosis has a heterogeneous presentation.²²  Some patients exhibit the classic symptoms of hemochromatosis, such as high transferrin saturation and iron accumulation in parenchymal cells. This presentation is associated with mutations in Fpn that affect its response to hepcidin.^8,9  Other patients present with an early rise in ferritin levels, low to normal transferrin saturation, and iron accumulation primarily in Kupffer cells. This presentation is associated with Fpn mutations that affect its plasma membrane localization or iron-exporting activity.^8,9  Serum ferritin levels in Fpn^ffe/+ heterozygous animals were increased 10-fold over 6-month-old control animals and increased with age (Figure 4A). In addition, transferrin saturation in serum from 6-month-old heterozygotes was much lower than that of control mice (Figure 4B). Blood smears from 4-week-old wild-type animals showed normochromic and normocytic RBCs that have a uniform size and shape and no target cells (Figure 4C). Blood smears from 4-week-old Fpn^ffe/+ heterozygous animals revealed some hypochromic red blood cells with variable size and some target cells (Figure 4D). These results show that Fpn^ffe/+ animals have a mild anemia, which is consistent with the extremely low transferrin saturation. To determine if heterozygous animals also present with iron accumulation in Kupffer cells, Prussian Blue staining was used to visualize accumulated ferric iron. Livers from wild-type mice do not accumulate iron (Figure 4E) whereas livers from aged Fpn^ffe/+ animals accumulate high levels of iron in Kupffer cells (Figure 4F). Together, these results indicate that expression of the Fpn^ffe allele affects the ability of wild-type Fpn to export iron by inhibiting the localization and activity of wild-type Fpn. Furthermore, the close similarities between the clinical phenotypes of Fpn^ffe/+ mice and human patients demonstrates the utility of the Fpn^ffe/+ mouse as a model for this disease.

Mutations in Fpn result in an autosomal-dominant disorder with 2 different patient presentations. Mutations that render Fpn insensitive to down-regulation by hepcidin result in iron overload in hepatocytes.^8,11,23  In this form of the disorder there is no regulation of Fpn, leading to high transferrin saturation and iron deposition in hepatocytes. The iron burden of hepatocytes can be reduced by phlebotomy, as liver iron stores are capable of being mobilized through Fpn. A second class of Fpn mutations prevents localization to the cell surface or permits its localization to the cell surface but affects the ability of Fpn to transport iron.^8,11,23  This form of the disorder, referred to as classic Fpn disease, results in low transferrin saturation, high serum ferritin, and excessive iron deposits in Kupffer cells, not hepatocytes. Patients with these types of Fpn mutations develop severe anemia upon repeated phlebotomy and iron stores in liver are not mobilized because mutant Fpn cannot export the excess iron. It has been a source of debate whether haploinsufficiency would explain Fpn disease or whether the disease results from a dominant-negative effect. While studies have supported the view that Fpn is a multimer^9,13,24  and that the mutant allele can affect the behavior of the wild-type allele, other studies have suggested that Fpn is monomeric.^10,12,25  All human Fpn mutations are missense mutations. If haploinsufficiency was the explanation for Fpn disease, then nonsense mutations should also result in the disorder; however, none have been found. Additionally, a targeted gene deletion in the murine Fpn gene has little effect in heterozygous animals.¹⁴  While these data suggest that haploinsufficiency cannot explain the disorder it is formally possible that Fpn disease in mice cannot recapitulate the human disease.

The discovery of the ffe mutation shows that mice can have Fpn disease, as mice heterozygous for the mutation exhibit all of the features of the human disorder: low transferrin saturation, high serum ferritin, and excessive iron in Kupffer cells. Our studies show that macrophages from ffe/+ mice have no detectable Fpn on their cell surface and cannot export iron. Additionally, generation of the H32R mutation in constructs of Fpn-GFP demonstrates that mutant Fpn can exert a dominant-negative effect and prevent the surface localization of wild-type Fpn-FLAG. Finally, using siRNA oligonucleotides to decrease the expression of Fpn by 50%, as would be expected for haploinosufficiency, has no effect on macrophage iron export. Iron export is only compromised when Fpn levels decrease below 30% of wild type. Together, these data show that Fpn disease does not result from haploinsufficiency but rather from a dominant-negative effect of the mutant allele.

The available clinical data,²  supported by the studies reported here, show that Fpn mutations that lead to Kupffer cell iron loading result in low transferrin saturation. The low transferrin saturation may lead to iron-limited anemia during periods of active growth or in the face of phlebotomy. In is unclear, however, whether these Fpn mutations have adverse clinical affects independent of the low transferrin saturation.²⁶  The Fpn^ffe/+ mouse will be a useful system to explore the pathophysiological consequences of Kupffer cell iron loading.

Contribution: I.E.Z. performed the positional cloning, preparation of mouse samples for further analysis, histologic analysis, and wrote the manuscript; I.D.D. generated the Fpn(H32R)-GFP plasmid, performed the immunofluorescence, macrophage culturing, ferritin analysis, and transferrin saturation analysis, and wrote the manuscript; A.P. performed the positional cloning, sequencing of the ffe mutation, and histologic analysis; D.M.W. performed the immunofluorescence and assisted in the writing and preparation of the manuscript; A.J.S. performed the positional cloning; J.F.G. and X.L. analyzed the peripheral blood smears; and L.N. and J.K. guided these studies and assisted in manuscript preparation.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

I.E.Z. and I.D.D. contributed equally to this work.

Correspondence: Jerry Kaplan, Department of Pathology, School of Medicine, University of Utah, Salt Lake City, UT 84132; e-mail: jerry.kaplan@path.utah.edu.

The authors wish to express their appreciation to Dr Richard Ajioka for help in measuring transferrin saturation; Kathryn V. Anderson and members of the Anderson and Niswander laboratories for performing the mutagenesis screen; Lori Bulwith for technical assistance; and Christopher Porter for help with blood smears. The current address for I.E.Z. is Center for Neuroscience Research, Children's Research Institute, Children's National Medical Center, Washington, DC.

This work was supported by National Institutes of Health (NIH) grants DK070 947 (J.K.) and F32-HD08 605 (I.E.Z.). J.F.G. is supported by an NIH training grant (T32CA8608604). L.N. is a Howard Hughes Medical Institute (HHMI) investigator.