Mice with hepcidin-resistant ferroportin accumulate iron in the retina

Abstract

Because ferroportin (Fpn) is the only known mammalian cellular iron exporter, understanding its localization and regulation within the retina would shed light on the direction of retinal iron flux. The hormone hepcidin may regulate retinal Fpn, as it triggers Fpn degradation in the gut. Immunofluorescence was used to label Fpn in retinas of mice with 4 different genotypes (wild type; Fpn C326S, a hepcidin-resistant Fpn; hepcidin knockout; and ceruloplasmin/hephaestin double knockout). No significant difference in Fpn levels was observed in these retinas. Fpn localized to the abluminal side of the outer plexiform vascular endothelial cells, Müller glia cells, and the basolateral side of the retinal pigment epithelium. Adeno-associated virus (AAV)-hepcidin was injected into the eyes of hepcidin knockout mice, while AAV-lacZ was injected into the contralateral eyes as a control. AAV-hepcidin injected eyes had increased ferritin immunolabeling in retinal vascular endothelial cells. Fpn C326S mice had systemic iron overload compared to wild type and had the fastest retinal iron accumulation of any hereditary model studied to date. The results suggest that physiologic hepcidin levels are insufficient to alter Fpn levels within the retinal pigment epithelium and Müller cells, but may limit iron transport into the retina from vascular endothelial cells.—Theurl, M., Song, D., Clark, E., Sterling, J., Grieco, S., Altamura, S., Galy, B., Hentze, M., Muckenthaler, M. U., Dunaief, J. L. Mice with hepcidin-resistant ferroportin accumulate iron in the retina.

Iron, while essential for living organisms, must be tightly regulated as it can catalyze the formation of reactive oxygen species. Within the retina, labile iron has been shown to cause degeneration (1). Age-related macular degeneration (AMD), the leading cause of irreversible vision loss in individuals over 50 yr old (2), is associated with increased retinal iron (3). We have previously studied mice lacking the 2 ferroxidases ceruloplasmin (Cp) and hephaestin (Heph), which leads to intracellular iron entrapment. These double-knockout (DKO) mice develop retinal iron overload, causing retinal degeneration with some histologic and molecular features of human AMD (4). Retinal degeneration in the ceruloplasmin/hephaestin DKO (Cp/Heph DKO) mice can be ameliorated by systemic iron chelation therapy (5). A role for iron in AMD is further supported by the observation that patients with aceruloplasminemia, an autosomal recessive disease with retinal iron accumulation, can develop early-onset macular degeneration (6, 7).

Ferroportin (Fpn), a transmembrane protein that is the only known cellular iron exporter (8), and hepcidin, an iron regulatory hormone, are 2 key contributors to retinal iron homeostasis. In enterocytes and macrophages, hepcidin binds to Fpn, triggering its internalization and degradation (9). Knockout (KO) of hepcidin causes retinal iron overload and retinal degeneration (10). It is unclear whether the retinal iron overload in hepcidin KO mice results from high serum iron levels, impaired retinal Fpn regulation, or both.

To better understand the role of retinal Fpn regulation in retinal iron homeostasis, we first immunolocalized Fpn, then studied a mouse line with the hepcidin-resistant Fpn (Fpn C326S) mutation (11). This mutation prevents hepcidin from binding to Fpn, rendering Fpn unresponsive to changes in hepcidin levels (12–14). The systemic consequences of the Fpn C326S mutation (high serum and parenchymal iron levels) are comparable to the effects of hepcidin deficiency (11, 13, 15). Fpn is regulated at the transcriptional, posttranscriptional, and posttranslational levels. It is regulated at the transcriptional level by iron via the Nrf/Bach 1 system (16, 17) and posttranscriptionally by an iron responsive element in the 5′-untranslated region of its mRNA. Iron regulatory proteins bind the iron responsive element of the Fpn mRNA, preventing Fpn translation in situations of intracellular iron deficiency (18–20). Fpn is regulated at the posttranslational level by hepcidin.

In this study, we demonstrate that Fpn C326S–mutant mice have the fastest retinal iron accumulation of any hereditary mouse model studied to date. We establish the retinal Fpn localization pattern, which suggests a route of retinal iron flux. We also provide evidence of local hepcidin-Fpn iron regulation in vascular endothelial cells.

MATERIALS AND METHODS

Animals

Hepcidin 1 KO (Hepc1^−/−) mice on a C57BL/6J background were generated as previously described (15). C57BL/6J mice with a targeted mutation in the Cp gene (Cp^−/−) and naturally occurring mutation in the Heph gene (Heph^sla/sla or Heph^sla/Y) were generated as previously published (21) and are referred to herein as Cp/Heph DKO mice. C326S mice on a 129P2/C57BL/6 mixed genetic background (B6N4F1) were generated as described previously (11). Male and female homozygous and heterozygous C326S mice and the respective wild-type (WT) controls were aged until 17 to 23 wk of age. Age- and gender-matched WT littermates were used as controls. WT, Hepc1^−/−, and DKO mice were kept on a 270 ppm iron diet, while the diet of the C326S mice contained 200 ppm iron (LASQCdiet, Rod18-A).

All procedures were in accordance with the European convention for the protection of vertebrate animals used for experimental and other scientific purposes, and were approved by the Institutional Animal Care and Use Committee of the University of Pennsylvania. C326S mice have been processed according to the regulations of the Animal Care and Use Committee of the University of Heidelberg (Project T-17/12).

Immediately after death, eyes were enucleated and samples of spleen, liver, and duodenum were taken. Tissues were fixed for at least 24 h in 4% paraformaldehyde, 0.1 M phosphate buffer, pH 7.4. In an automated tissue processor, samples were dehydrated through graded ethanol, cleared in xylenes, and immersed in paraffin. Paraffin-embedded tissue was cut into 5 µm sections and air dried on Superfrost Plus microscopy slides (Thermo Fisher Scientific, Waltham, MA, USA).

Perls Prussian blue stain

Paraffin was removed from the sections, which were rehydrated through xylenes and descending ethanol and washed in distilled water. Slides were then put into Coplin jars containing 2.5% potassium ferric ferrocyanide and 2.5% hydrochloric acid solution. Closed Coplin jars were heated in a nonshaking 65°C water bath under a fume hood for 45 min. Afterward, slides were washed in distilled water again. Sections were then bleached by incubating them first for 5 min in 0.25% potassium permanganate in 1× PBS, followed by 5 min in 0.1% oxalic acid in 1× PBS, then washed with 1× PBS. A light hematoxylin and eosin counterstain was performed; then sections were dehydrated in ethanol, cleared in xylene, and mounted.

Autofluorescence

Formaldehyde-fixed mouse eyes were dissected and the posterior segment cryoprotected in 30% sucrose solution overnight. Next the eye cups were embedded in Tissue-Tek optimal cutting compound (Sakura Finetek, Torrance, CA, USA) and slowly frozen in 2-methylbutane on dry ice. Sections (10 µm) were cut, air dried at room temperature overnight, and mounted with Vectashield mounting medium with DAPI (Vector Laboratories, Burlingame, CA, USA).

Immunofluorescence

Paraffin-embedded sections were deparaffinized and rehydrated through xylenes and descending ethanol and washed in distilled water. Antigen retrieval on inner organ sections was performed by incubation in 50 mM Tris-HCl (Tris(hydroxymethyl)amino-methane) pH 8.0, 1 mM EDTA, 0.5% Triton X-100, 20 µg/ml proteinase K (Roche Applied Sciences, Indianapolis, IN, USA) for 10 min at room temperature in a humid chamber. Antigen retrieval on retina sections was most effective when performed by incubation in 25 mM Tris-HCl pH 8.5, 1 mM EDTA, 0.05% (w/v) SDS at 97°C for 40 min (22). After antigen retrieval and antibody incubation, sections were washed 3 times with PBS.

For cryosectioning, the mouse eyes were enucleated and the globes were fixed in 4% paraformaldehyde for 10 min. The globes were then rinsed in PBS, and the eye cups were dissected by removing the cornea, iris, and lens. The eye cups were cryoprotected overnight in 30% sucrose and embedded in optimal cutting temperature compound (Tissue-Tek; Sakura Finetek, Torrance, CA, USA). Immunofluorescence was performed on cryosections 10 µm thick, as described previously (23).

Both primary and fluorescently labeled secondary antibodies were applied in PBS with 1% bovine serum albumin at room temperature for 1 h in a humid chamber. Primary antibodies were rabbit anti-Fpn (NBP1-21502; Novus Biologicals, Littleton, CO, USA) as previously used (24) at a 1:400 dilution, rabbit anti-L ferritin (E17) at a 1:200 dilution (the kind gift of Paolo Arosio, University of Brescia), chicken anti-H ferritin (Abcam, Cambridge, MA, USA) at 1:200 dilution, mouse antiglutamine synthetase at a 1:300 dilution (ab64613; Abcam), rat-anti CD31 at 1:100 dilution (ab7388; Abcam), and anti-α smooth muscle actin antibody at 1:200 dilution (ab7817; Abcam). Primary antibody was detected using fluorophore-labeled secondary antibodies (Jackson ImmunoResearch Laboratories, West Grove, PA, USA). Control sections were treated identically, except that primary antibodies were omitted. All sections were mounted with VectaShield mounting medium with DAPI (Vector Laboratories).

Microscopy

Sections were analyzed by light microscopy using a Nikon TE300 microscope equipped with ImagePro Plus version 6.1 (Media Cybernetics, Bethesda, MD, USA) software. Bright field was used for Perls Prussian blue–stained sections; epifluorescence was used for immunofluorescence and autofluorescence. A ×60 oil immersion objective and a ×20 objective were used for the micrographs presented. Sections triple labeled with CD31, ferritin, and α−smooth muscle actin were imaged by a Zeiss LSM-510 Meta confocal microscope (Carl Zeiss GmbH, Jena, Germany).

Gene transfer

Three-week-old hepcidin KO mice were used for virus intravitreal injection. The right eye received intravitreal injection of adeno-associated virus (AAV)-lacZ and served as the control (AAV serotype 8 with the cytomegalovirus promoter). The left eye received AAV-hepcidin (the same AAV serotype 8 containing the entire hepcidin gene coding sequence inserted behind the cytomegalovirus promoter). The viral titer was 10¹³ genomes per milliliter, and 2 μl of virus was injected into each eye. Transduction was confirmed through detection of hepcidin mRNA in the neural retina using quantitative PCR.

RESULTS

Retinal iron overload in C326S mice

Retinal cryosections were immunolabeled for both H and L ferritin as a measure of intracellular iron, as described previously (21). Both H and L ferritin were markedly increased in the retinas and choroids of 23-wk-old C326S mutant mice (Fig. 1C, F) compared to WT (Fig. 1A, D). H ferritin labeled the photoreceptors more prominently than L ferritin, and the change in photoreceptor iron was not as dramatic as in other retinal cell types. The intensity of immunofluorescence in the retinal pigment epithelium (RPE) and choroid of the C326S mutants showed a gene dosage effect, with the strongest ferritin signal in the homozygous C326 mice. The heterozygous C326S mice had weaker ferritin signal throughout the neural retina compared to the homozygous C326 mice, and the WT mice had the lowest ferritin intensity of all (Fig. 1B, E).

Figure 1.

Retinal L and H ferritin in WT and Fpn C326S mutants. Anti–L (A–C) and anti–H ferritin antibodies (D–F) were used for immunofluorescence on paraffin-embedded sections of eyes of 23-wk-old male mice. Genotypes were WT (A, D), C326S heterozygous (B, E), and C326S homozygous (C, F). Samples from at least 3 mice per genotype were stained; 1 representative micrograph each is shown. Exposure times and scaling are equal for all micrographs. There is gene-dosage dependent increase of ferritin throughout neural retina, RPE, and choroid. Increase is more pronounced for l-ferritin than for H ferritin. GCL, ganglion cell layer; IPL, inner plexiform layer; ONL, outer nuclear layer; INL, inner nuclear layer; IS, inner segments; OS, outer segments.

Given the increased ferritin levels in C326S mice, we sought to confirm the presence of iron directly by using Perls Prussian blue stain. Iron was detected in the RPE of 23-wk-old C326S heterozygous (Fig. 2B) and homozygous (Fig. 2C) mice, with an evident gene dosage effect, similar to the gene dosage effect in the ferritin immunofluorescence (Fig. 1). This gene dosage effect was also seen with systemic iron levels where homozygous mice had the highest systemic iron levels, WT mice had the lowest, and heterozygous mice had levels between those of their homozygous and WT counterparts (11). No Perls Prussian blue signal was detected in the WT (Fig. 2A). In homozygous C326S mutant mice, the choroid also showed strong Perls Prussian blue labeling (Fig. 2C), while there was no labeling in the choroid of heterozygous or WT mice (Fig. 2A, B). Because iron levels were high in 23-wk-old retinas, we stained sections of 17-wk-old C326S homozygous mice and found that iron was detectable in the RPE and choroid even at that age (Fig. 2D).

Figure 2.

Perls Prussian blue stain of WT and Fpn C326S mutant retinas. Perls Prussian blue staining was performed on paraffin-embedded sections of eyes of male 23- (A–C) or 17-wk-old (D) mice. Genotypes were WT (A), C326S heterozygous (B), and C326S homozygous (C, D). Samples from at least 3 mice per genotype were stained; 1 representative micrograph each is shown. Scaling for (A–C) is identical, with scale bars shown in (C). There is gene dosage (A–C)-dependent increase of iron deposits in RPE. Only Fpn C326S homozygous mice (C, D) showed pronounced iron overload in choroid.

RPE morphology and autofluorescence

RPE morphology as a function of iron overload was then addressed. The slight iron overload that we found in C326S heterozygous mice did not alter the microscopic appearance of the RPE (Fig. 3B) compared to WT (Fig. 3A). However, in 23-wk-old C326S homozygous mouse eyes, we found areas with abnormal RPE. These RPE cells were hypertrophic with apical nuclei and autofluorescent cytoplasm (Fig. 3C).

Figure 3.

RPE morphology and autofluorescence. Cryosections of eyes of 23-wk-old mice were studied for RPE morphology and autofluorescence. Blue fluorescence is DAPI. Genotypes were WT (A), C326S heterozygous (B), and C326S homozygous (C). Samples from at least 3 mice per genotype were stained; 1 representative micrograph each is shown. RPE cells of C326S homozygous mice were hypertrophic with apical nuclei and autofluorescent cytoplasm (C).

Localization of Fpn in WT mouse retinas

C326S mutants were expected to have more Fpn localized at the cell surface as a result of their resistance to hepcidin. To establish a baseline, Fpn immunofluorescence studies were conducted to determine the localization of Fpn in the WT retina. In WT retinas, we found that Fpn cell surface expression was most prominent on the abluminal side of outer plexiform layer (OPL) vascular endothelial cells (Fig. 4A–C), on Müller cells (Figs. 4A and 5C), and the basolateral side of the RPE (Fig. 4A, F). Colabeling with glutamine synthetase was performed to mark Müller cells and Fpn (Fig. 5A–C). There was colocalization of Fpn and glutamine synthetase throughout the Müller cells except for the innermost portion of the end feet along the internal limiting membrane (Fig. 5C).

Figure 4.

Fpn expression in WT C57BL/6J mouse retina. We studied Fpn cell surface expression in WT retina of C57BL/6J mice. Immunolabeling from 28-wk-old mouse is shown. Rabbit anti-Fpn antibody was used on 5 µm paraffin-embedded sections after heat-induced antigen retrieval and detected with anti-rabbit Cy3 (A). DAPI was used as nuclear counterstain. Green channel shows autofluorescence, especially in red blood cells. Several blood vessels in the OPL are strongly Fpn positive. Magnifications of 2 of these vessels (B, C) show red blood cell surrounded by endothelial cells. This is best visible in (B), where the nucleus is also found in the section. As can be seen at high magnification, the abluminal side of these endothelial cells is strongly positive for Fpn. In contrast, Fpn signal on inner plexiform layer (IPL) vessels is much weaker and nearly invisible at same exposure conditions (D, E). Müller cells have Fpn-positive end feet at the inner and outer limiting membranes. This is shown in detail in Fig. 5. Strong basolateral Fpn signal was consistently observed in RPE (A, F).

Figure 5.

Retinal Müller cell Fpn expression in WT C57BL/6J mouse. Colabeling for Fpn and glutamine synthetase was performed on paraffin-embedded sections of 19-wk-old mice. Fpn is shown in red (A) and glutamine synthetase in green (B). DAPI was used as nuclear counterstain (C, D). Merged fluorescence (C, D) demonstrates by yellowish superposition signal that glutamine synthetase–positive Müller cells are also positive for Fpn, which is most pronounced along external limiting membrane but not along innermost part of internal limiting membrane.

Retinal Fpn levels in Fpn C326S mutant, hepcidin KO, and Cp/Heph DKO mice

The resistance of Fpn to hepcidin caused by the C326S mutation could lead to increased retinal cell surface Fpn, if Fpn is regulated by hepcidin produced within the retina. The same effect should occur in hepcidin KO mice. In Cp/Heph DKO mice, the absence of both hephaestin and ceruloplasmin, as well as the consequent retinal iron overload/hepcidin up-regulation (4, 10), could also affect Fpn levels. We analyzed the effects of these 3 genotypes on retinal Fpn levels by comparing Fpn immunofluorescence in the retinas of these mutant mice to age-matched WT controls processed in parallel with their respective mutant counterparts. Notably, we found no difference in the Fpn immunofluorescence intensity in the Müller and RPE cells in hepcidin KO (Fig. 6A), Fpn C326S mutant (Fig. 6B), and DKO (Fig. 6C) mice compared to their respective WT controls. An increased photoreceptor signal in Fig. 6B compared with Fig. 6A, C was due to autofluorescence, as determined by a no-primary-antibody control (not shown).

Figure 6.

Retinal Fpn immunofluorescence in hepcidin KO, C326S mutant, and Cp/Heph DKO mice. Fpn immunofluorescence was performed on paraffin-embedded sections of hepcidin KO (A), Fpn C326S mutant mice (B), and Cp/Heph DKO (DKO) (C). Left side of each panel shows WT mouse of same sex, age, parallel processing, and fixation, as well as equal exposure conditions. We found same typical pattern of Fpn-positive Müller cells, OPL endothelium, and basolateral RPE described in Figs. 4 and 5. Stronger photoreceptor signal in (B), especially in outer segments, can be ascribed to autofluorescence.

In order to verify the specificity of the Fpn antibody, it was tested in immunofluorescence studies of paraffin-embedded tissues of several organs from hepcidin KO mice. Fluorescence photomicrographs of the liver, spleen, and duodenum of the 3 hepcidin KO (Fig. 7A–C) and 3 WT mice (Fig. 7D–F) were assessed. The resulting labeling reflected the expected expression pattern for both WT and hepcidin KO mouse liver, spleen, and duodenum; WT livers (Fig. 7D–F) showed a typical Kupffer cell pattern, WT spleens showed Fpn-positive macrophages in the red pulp, and WT duodeni displayed a slight basolateral but mostly cytoplasmic staining of enterocytes. Hepcidin KO mice (Fig. 7A–C), in contrast to the WT mice, showed stronger Kupffer cell staining, Fpn-positive hepatocyte membranes, strong Fpn signal throughout the red pulp of the spleen, and a strong basolateral Fpn signal on duodenal enterocytes.

Figure 7.

Fpn immunofluorescence on inner organs of hepcidin KO and WT mice. To test whether immunofluorescence is sensitive enough to demonstrate differences in cell surface Fpn expression, we performed immunofluorescence on liver, spleen, and duodenum of 3 hepcidin KO (A–C) and 3 WT mice (D, E), all 12-wk-old. Images were obtained with ×20 objective (left), and details of same area are shown with ×60 oil immersion (right). Exposure conditions and scaling are equal among all ×20 and among all ×60 micrographs. Immunofluorescence shows expected pattern and pronounced change in hepcidin KO compared to WT mice. Cell surface Fpn signal is stronger in duodenal enterocytes, spleen macrophages, Kupffer cells, and hepatocytes of hepcidin KO mice (A–C) than WT mice (D–F).

AAV-hepcidin causes ferritin accumulation in vascular endothelial cells in hepcidin KO mice

We performed hepcidin gene transfer to determine if hepcidin could have an effect on retinal iron. One month after intravitreal injection, fluorescence photomicrographs of retinas showed that in the presence of AAV-hepcidin there were ferritin-positive blood vessels (Fig. 8C, D, arrowheads) not seen in the controls (Fig. 8A, B). To more precisely localize the ferritin in blood vessels, we then triple-labeled with CD31 (white in Fig. 9), a marker of vascular endothelial cells, ferritin (red in Fig. 9), and anti-α-smooth muscle actin antibody for pericytes (green in Fig. 9). We found colocalization of ferritin with the vascular endothelial cells (yellow in Fig. 9) but not with pericytes, whose label surrounds the endothelial cells.

Figure 8.

Ferritin immunofluorescence in AAV-hepcidin-injected retinas. AAV-hepcidin was injected into left eyes (C, D) and AAV-LacZ into right eyes (A, B). Arrowheads show ferritin-positive blood vessels in eyes that received AAV-hepcidin (C, D) that are not seen in controls (A, B).

Figure 9.

Vascular endothelial cell immunofluorescence in AAV-hepcidin injected retinas. Triple labeling for vascular endothelial cell, pericytes, and ferritin was performed on cryosections of mice that received AAV-hepcidin. CD31 is shown in white, ferritin in red, and pericytes in green. Orange in merged fluorescence photomicrograph demonstrates that CD31-positive vascular endothelial cells are also positive for ferritin. However, ferritin staining was not colocalized with pericytes surrounding endothelial cells.

DISCUSSION

We demonstrated that mice carrying the Fpn C326S mutation develop an early gene dosage-dependent retinal iron overload with Perls Prussian blue–stainable iron in the RPE. This iron overload occurred in 17-wk-old mice, making them the youngest mice to develop RPE iron overload compared to other hereditary murine iron overload models. Importantly, the iron overload was paralleled by hypertrophy and autofluorescence in the RPE, which is also found in AMD eyes (25). We previously described similar RPE hypertrophy and autofluorescence in hepcidin KO mice at 18 mo and Cp/Heph DKO mice (4, 10) at 7 mo. C326S mice will therefore serve as an additional model for oxidative stress–induced early RPE degeneration. Nutritional differences cannot account for the early iron accumulation in C326S retinas, as hepcidin KO and Cp/Heph DKO mice were kept on a 270 ppm iron diet, while the diet of the C326S mice contained only 200 ppm iron. However, the genetic backgrounds of these 3 models were not identical. The genetic backgrounds of hepcidin KO and DKO mice were pure C57BL/6, but the C326S mice had a mixed 129P2/C57BL/6 background (93.75% C57BL/6 congenic). An overview of some of the major phenotypic characteristics of the different gene mutations/KO is outlined in Table 1.

TABLE 1.

Phenotypic characteristics of mutant and KO mice

	Gene
Characteristic	Fpn (Slc40a1)	Hepcidin 1 (Hepc1)	Ceruloplasmin (Cp) and hephaestin (Heph)
Strain	129P2/C57BL6J (11)	C57BL/6J	C57BL/6J
Mutation	C326S point mutation (11)	Targeted disruption (15)	Targeted mutation (Cp) and naturally occurring sla mutation (Heph) (21)
High plasma/serum iron	Demonstrated after 10 wk (11)	Demonstrated after 2 mo (15)	Low serum iron
Inner organ iron status	Parenchymal iron overload (liver, pancreas), iron poor macrophages (spleen red pulp, Kupffer cells after 8 wk) (11)	Parenchymal iron overload (liver after 2 mo, heart, pancreas), iron-poor macrophages (spleen red pulp, Kupffer cells) (2)	Liver and brain iron overload, low serum iron, and anemia (Dunaief lab, unpublished data)
Increased retinal l-ferritin	Demonstrated after 23 wk	Demonstrated after 3 mo (10)	Demonstrated after 7 mo (4)
Perls Prussian blue iron stain in RPE and choroid	Demonstrated after 17 wk	Demonstrated after 18 mo (10)	Demonstrated after 7 mo (4)
RPE autofluorescence	Demonstrated after 23 wk	Demonstrated after 18 mo (10)	Demonstrated after 7 mo (4)

In order to investigate the cell surface Fpn localization in Fpn C326S mutants, we first studied the baseline expression in WT mice and found a strong abluminal expression in the retinal OPL vascular endothelium, in Müller cells, most prominently in their end feet, and on the basolateral side of the RPE. There was no difference in Fpn levels at any retinal location comparing the Fpn C326S mutants to WT.

Therefore, we extended this investigation to hepcidin KO mice and Cp/Heph DKO mice and interestingly did not find a difference in cell surface Fpn compared to the respective controls. It is important to note that the same antibody showed a different, but more expected, Fpn response to hepcidin when used on several inner organs of hepcidin KO versus WT mice, validating the antibody labeling. Therefore, much of the retinal Fpn protein in the Müller cells and RPE seems not to be regulated by hepcidin in WT or Cp/Heph DKO mice (or counteracting mechanisms efficiently neutralize the lack of Fpn regulation by hepcidin in the retina). It was more difficult to assess Fpn levels in vascular endothelial cells, as strong Fpn signal in the OPL tends to obscure the vasculature.

To test whether Fpn in vascular endothelial cells might be regulated by hepcidin, we injected AAV-hepcidin into the vitreous of hepcidin KO mice and found elevated ferritin (indicating elevated iron levels) in these cells. This is consistent with hepcidin-triggered Fpn degradation because cells lacking Fpn, the only known cellular iron exporter, accumulate intracellular iron. The hepcidin levels produced by the virus may have been supraphysiologic, but this experiment suggests that hepcidin, which is normally produced in the retina (26), may be able to inhibit iron influx into the retina through the retinal vascular endothelium.

Our results provide a possible explanation for the tendency of iron to accumulate in the RPE in Fpn C326S and hepcidin KO mice. We found Fpn on the abluminal side of retinal vascular endothelium and the basolateral side of the RPE. This staining pattern suggests that iron is trafficked from retinal capillaries toward the RPE and from the RPE to the choroidal capillaries (Fig. 10). High Müller cell Fpn suggests that Müller cells may participate in intraretinal iron trafficking. The Müller cell Fpn signal was particularly high at the external limiting membrane. Therefore, Müller cells could provide photoreceptors with iron. Photoreceptor function is dependent on sufficient iron supply (27). RPE cells phagocytose photoreceptor outer segments every day and thus accumulate photoreceptor iron. In our proposed retinal iron pathway (Fig. 10), the RPE is the end point of retinal iron trafficking, from which iron is exported back to the circulation. In situations of retinal iron overload, as in Fpn C326S mutants, hepcidin KO, bone morphogenetic protein 6 KO, and DKO (4, 10, 28), or in case of RPE dysfunction, which may occur in AMD (3), the RPE seems to be unable to export excess iron to the circulation and becomes iron overloaded before other parts of the retina are affected.

Figure 10.

Proposed retinal iron transport pathway. Schematic drawing of proposed Fpn-driven iron trafficking through retina based on immunofluorescence and iron staining results. Red dots indicate iron; arrows, Fpn where it shows strongest immunolabeling. Abluminal retinal vascular endothelial cell Fpn may transport iron through blood–retina barrier into retina (A). Müller cells may absorb part of this iron and then release it toward photoreceptors at the external limiting membrane (B). Photoreceptor outer segments are phagocytosed by RPE cells, which can export iron out of retina and toward vessels of choriocapillaris (C).

This typical pattern of RPE-enhanced iron overload is consistent with the Fpn expression data presented herein. RPE cells recycle iron from photoreceptors to the circulation in the same way that Kupffer cells and spleen macrophages recycle iron from red blood cells. However, in hepcidin KO mice, Kupffer cells and spleen macrophages become iron deficient despite high serum iron (15). This correlates well with the high levels of cell surface Fpn on macrophages in the hepcidin KO mice, as we have shown. If the RPE behaved like other phagocytosing cells and increased Fpn expression in the hepcidin KO mice, the RPE may not become iron overloaded. Accordingly, RPE iron overload in hepcidin KO mice and Fpn C326S mice (which also have increased systemic iron levels), and after systemic iron injections (29) could be explained by failure to up-regulate basolateral RPE Fpn.

This hypothesis raises the question of why RPE Fpn is not affected by hepcidin. It has recently been demonstrated that the liver is the primary source of hepcidin (30). Therefore, one answer may be that RPE Fpn may not be accessible to circulating hepcidin of hepatic origin. Moreover, while retinal hepcidin expression has been demonstrated (26), it is unclear whether local retinal hepcidin expression has significant effects on RPE Fpn under physiologic conditions. Access to systemic hepcidin on the basolateral side of the RPE may be possible, as hepcidin is small (∼2.8 kDa) and could penetrate the fenestrae of the choriocapillaris and the Bruch membrane. However, a major fraction of hepcidin might be bound to plasma proteins (31), although this is controversial (32). If hepcidin penetrates the choriocapillary endothelium, it might be taken up by endothelium or could bind to endothelial Fpn. It is possible that the increase in retinal hepcidin expression associated with high levels of intraretinal iron, as seen in eyes injected with holo-transferrin (10), might limit further iron uptake into the retina by triggering degradation of Fpn in the vascular endothelial cells. High, very local concentrations of hepcidin might affect Fpn on specific cell types, such as vascular endothelial cells, without causing any noticeable changes in Fpn levels within the Müller cells.

By immunolocalizing Fpn, the only known cellular iron exporter, we are able to propose a route for retinal iron flux; iron enters the retina through the vascular endothelium, is distributed by Müller cells, and exits from the basal RPE. When the retina senses high iron level, it increases hepcidin production, which triggers Fpn degradation in vascular endothelial cells, limiting retinal iron entry. However, alterative pathways are possible, such as iron flux directly from Müller cells to RPE. Also, neurons may help conduct iron through the retina through their axons. Future studies will need to assess the role of retinal versus systemic hepcidin utilizing retina- or liver-specific conditional KO. Furthermore, the interesting role of the RPE in retinal iron metabolism warrants more detailed studies of RPE iron storage and trafficking.

Glossary

AAV

adeno-associated virus

AMD

age-related macular degeneration

Cp/Heph DKO

ceruloplasmin/hephaestin double knockout

DKO

double knockout

Fpn C326S

hepcidin-resistant ferroportin

Fpn

ferroportin

Hepc1^−/−

hepcidin 1 knockout

KO

knockout

OPL

outer plexiform layer

RPE

retinal pigment epithelium

WT

wild-type