Conditional deletion of ferritin H in mice induces loss of iron storage and liver damage†
Potential conflict of interest: Nothing to report.
Abstract
Ferritin plays a central role in iron metabolism by acting both as iron storage and a detoxifying protein. We generated a ferritin H allele with loxP sites and studied the conditional ferritin H deletion in adult mice. Ten days after Mx-Cre induced deletion, ferritin H messenger RNA (mRNA) was below 5% in the liver, spleen, and bone marrow of deleted mice compared to control littermates. Mice lost their cellular iron stores indicating the requirement of ferritin H in iron deposition. Serum iron and transferrin saturation were slightly increased and correlated with a two-fold increased liver hepcidin 1 mRNA and a reduced duodenal DcytB mRNA level. Under a normal iron regimen, deleted mice survived for 2 years without visible disadvantage. Mice fed on a high iron diet prior to ferritin H deletion suffered from severe liver damage. Similarly, ferritin H deleted mouse embryonic fibroblasts showed rapid cell death after exposure to iron salt in the medium. This was reversed by wild-type ferritin H but not by a ferritin H mutant lacking ferroxidase activity. Cell death was preceded by an increase in cytoplasmic free iron, reactive oxygen species, and mitochondrial depolarization. Conclusion: Our results provide evidence that the iron storage function of ferritin plays a major role in preventing iron-mediated cell and tissue damage. (HEPATOLOGY 2009.)
Protein shells of 24 ferritin H and L subunits can accumulate up to 4500 iron atoms as Fe3+ in all tissues, but most prominently in hepatocytes and reticuloendothelial cells of liver and spleen.1 The ferritin H subunit has a ferroxidase activity needed for iron deposition.2, 3 Translation of both subunits is regulated by iron regulatory proteins 1 and 2 (IRP1 and IRP2) in response to iron availability.4 Thus, the amount of ferritin adapts to body iron levels. Iron stored in ferritin can be mobilized prior to ferritin degradation.5 Ferritin's iron scavenging function is also thought to prevent iron-mediated catalysis of reactive oxygen species (ROS), which provoke tissue damage6 and may cause cancer7 and neurodegeneration.8
Previous studies in mice found that a straight knockout of the ferritin H (Fth) gene provokes embryonic lethality.9 Until now, this has hampered studying potential other functions of ferritin in vivo. Ferritin might contribute iron for cell proliferation of lymphoid cells10 and complement extracellular iron uptake from transferrin by transferrin receptor.11 Ferritin was also postulated to scavenge excess iron from nutrition in duodenal cells, known as the “mucosal block” hypothesis.12, 13 Under this hypothesis, when body iron stores are high and ferritin synthesis in crypt cells increased, mature villus cells might retain iron and prevent its basolateral release to serum. This view is not formally excluded, but challenged by the discovery of the peptide hormone hepcidin 1.14, 15 The relative importance of ferritin versus hepcidin 1 remains therefore to be investigated. Hepcidin 1 expression is regulated by body iron levels and essential for the control of iron absorption.14-16 Hepcidin 1 may either repress messenger RNA (mRNA) of iron transporters through signaling17, 18 and/or interfere directly with basolateral iron export by binding to ferroportin.19 Ferritin also retains iron in reticuloendothelial cells during anemia of acute or chronic inflammation, but it remains unknown whether ferritin synthesis is the cause or result of iron retention. Again, hepcidin 1, induced by inflammatory cytokines, might favor iron retention by controlling iron efflux by way of ferroportin.19, 20 Finally, it was debated whether ferritin is important in providing iron for heme biosynthesis.21, 22
Here, we generated mice carrying loxP sites 5′ of the Fth gene promoter and 3′ of exon 1 in order to delete ferritin H expression in adult mice by Cre recombinase. We have conditionally deleted the Fth gene in adult male mice using Cre under the control of the poly-IC inducible Mx gene promoter.23 This deletes ferritin H almost completely in liver, bone marrow, spleen, and thymus, but less prominently in other tissues. We report here the effects on iron storage, iron toxicity, hematological parameters, and cell viability. We have explored effects of iron overload before or after ferritin H deletion and found rapid cell death in the liver and in mouse embryonic fibroblasts derived from our mice. In cultured cells, the cause of death could be attributed to ROS, mitochondrial depolarization, and permeability transition.
Abbreviations
ALT, alanine aminotransferase; AST, aspartate aminotransferase; Fth gene, ferritin H gene; IRP, iron regulatory protein; ROS, reactive oxygen species; TUNEL, transferase dUTP nick end labeling.
Materials and Methods
All experimental procedures are available as Supporting Materials and Methods.
Results
Conditional Ferritin H Deletion Is Lethal in Embryos.
To study the function of ferritin H in adult mice we generated, by embryonic stem cell recombination, mice with a modified ferritin H allele referred to as FthloxNeo. In this allele the ferritin H promoter and exon 1 are bordered by loxP sites that also flank a neomycin selection cassette between FRT sites (Supporting Fig. 1). To verify our ferritin H targeting strategy, we crossed FthloxNeo mice with hemizygous nestin-Cre1 mice.24 The nestin promoter shows a complex expression pattern with activity in the brain as well as in the germline.24 Therefore, complete deletion of ferritin H was expected to occur in FthloxNeo/+;Nes-Cre1 animals. Of 38 F2 mice, 17 (44.7%) were Fth+/+ and 21 (55.3%) were Fth+/−, confirming nestin-Cre induced deletion. No Fth−/− mice were born, indicating an essential function of ferritin H in embryogenesis. These results are best explained by a germline deletion of ferritin H in the FthloxNeo/+;Nes-Cre1 F1 generation24 with embryonic lethality when the Fth gene is absent from germ cells.9 This is in agreement with previously published studies on embryonic lethality of ferritin deletion in mice.
Ferritin H Deletion in Adult Animals Induces Loss of Iron Storage.
For all subsequent experiments we used Fthlox/lox mice derived by breeding FthloxNeo/loxNeo mice with mice transgenic for Flp recombinase. Fthlox/lox mice were crossed with transgenic Mx-Cre mice to obtain experimental Fthlox/lox;Mx-Cre and control Fthlox/lox littermates. Initially, mice were of mixed genetic background (C57 BL/6 × 129/Sv), but with advancement of the analysis, mice were 10 times backcrossed with C57BL/6J mice. Upon induction by poly I/C, the ferritin H gene deletion was analyzed in adult male mice (Fig. 1A). Liver, spleen, and bone marrow showed an almost complete loss in the ferritin H gene and mRNA. In parallel, ferritin H protein was rapidly lost (Fig. 1B,C). However, staining for ferritin L was enhanced (Fig. 1B), possibly due to increased ferritin L translation after IRP1 and IRP2 inactivation by iron liberated upon ferritin H degradation.4 Concomitant with the loss of ferritin H mRNA, liver and spleen showed a marked decrease in tissue iron (Fig. 1A). In duodenum and heart, ferritin H mRNA was significantly reduced, but not enough to affect iron storage (Fig. 1A), whereas in kidney and lung, ferritin H was poorly deleted, as expected.23 Histochemical analysis showed also a complete loss of iron deposits in spleen reticuloendothelial cells of FthΔ/Δ mice (Fig. 1D). This was not due to the loss of reticuloendothelial cells (Fig. 1E). Liver iron deposits were too low to be detected by this method, or by 3,3′diaminobenzidine (DAB)-enhanced iron staining.
Conditional ferritin H deletion by Mx-Cre reduces iron storage in liver and spleen. Ten-week-old Fthlox/lox;Mx-Cre and control Fthlox/lox mice were injected with poly-IC to activate Cre expression. (A) At day 10 the ferritin H deletion was assessed in different tissues by real-time PCR on genomic DNA (green) and ferritin H complementary DNA (cDNA) (blue). Tissue iron content (red) was measured by the bathophenanthroline method. Values in FthΔ/Δ mice (n = 4) are expressed in percent of values in Fthlox/lox mice (n = 4) set as 100% ± standard deviation (SD). (B) Immunofluorescence staining of ferritin H and L in frozen liver sections (day 10). Scale bar = 200 μm. (C) Immunoblot analysis of liver ferritin H in one Fthlox/lox mouse and two FthΔ/Δ mice at days 3 and 10. (D) Frozen spleen sections stained by Perl's Prussian blue at day 30. Scale bar = 200 μm. (E) Frozen spleen sections stained for macrophages with anti-CD11b and anti-F4/80 antibodies, as well as 4′,6-diamidino-2-phenylindole (DAPI). Scale bar = 100 μm.
Ferritin Is Not Required for Hemoglobin Synthesis.
Most iron is recycled from hemoglobin of senescent erythrocytes, by the reticuloendothelial cells, and subsequently reincorporated into fresh hemoglobin in erythroid precursor cells. Therefore, it was of interest to observe whether Mx-Cre-induced ferritin H deletion (Fig. 1A) would modify hemoglobin and hematocrit levels. FthΔ/Δ mice showed no significant difference compared to Fthlox/lox mice (Supporting Table 1), and they survived for 2 years without noticeable disadvantage, whereas the extent of deleted cells remained unchanged. This suggests that ferritin H is not essential for iron recycling by macrophages and erythroid heme-biosynthesis. We further tested whether mice without ferritin H in liver and spleen showed changes in serum iron levels and transferrin saturation at 3, 10, and 30 days after ferritin H deletion (Fig. 2B). A mild but significant increase was observed for both parameters in experimental versus control mice.
Ferritin H deletion causes an increase of serum iron, transferrin saturation and liver hepcidin 1 mRNA, and repression of duodenal Dcytb mRNA. (A) mRNA expression was tested by real-time PCR at days 3, 10, and 30 after ferritin H deletion using primers indicated in Supporting Table 2. Average expression (arbitrary units) for Fthlox/lox mice (n = 6;3;3) and FthΔ/Δ mice (n = 6;3;3) was normalized to the geometric average of two control genes (glyceraldehyde 3-phosphate dehydrogenase [GAPDH] and hypoxanthine-guanine phosphoribosyltransferase [HPRT] in liver; GAPDH and β-actin in intestine) ± SD. ***P < 0.0005; *P < 0.05. (B) Serum iron levels and transferrin saturation. *P < 0.05. (C) Transferrin saturation correlated with hepcidin 1 mRNA expression in Fthlox/lox (○) and FthΔ/Δ (●) mice (R = 0.701 and P < 0.0002).
Effects of Mx-Cre–Mediated Ferritin H Deletion on Gene Expression.
Real-time polymerase chain reaction (PCR) in FthΔ/Δ versus Fthlox/lox mice at various timepoints revealed a significant increase of liver hepcidin 1 mRNA (Fig. 2A). This increase correlated with increased serum iron (Fig. 2B) and transferrin saturation (Fig. 2C). Among mRNAs coding for proteins of duodenal iron transport, only DcytB mRNA was significantly repressed, whereas mRNA of DMT1, ferroportin, and hephaestin showed no significant change (Fig. 2A and data not shown). Liver mRNAs of ferritin L, transferrin, and Hfe were unchanged. Transferrin receptor 1 mRNA was slightly reduced in most tissues, compatible with mRNA destabilization due to IRP inactivation.25 The mRNAs for transferrin receptor 2, superoxide dismutase 2, and glutathione peroxidase 1 remained unchanged in liver and intestine.
Severe Liver Damage by Ferritin H Deletion in Animals with High Iron Diet.
As FthΔ/Δ mice on C57BL/6 background showed only mild changes, we investigated how well they would endure increased iron levels after feeding with an iron-rich diet. As expected, they accumulated 8 times more iron in the liver and showed a 3.2-fold increase in hepcidin 1 mRNA (Fig. 3B,C). Concomitantly, ferritin H was strongly induced (Fig. 3A, lane 2). Upon Fth gene deletion, ferritin H disappeared rapidly (Fig. 3A, lane 3) and tissue iron levels dropped to 39% within 5 days (Fig. 3B). Perl's staining of nondeleted liver sections showed diffuse iron staining in hepatocytes near blood vessels. Five days after deletion the staining was weaker, with a more granular appearance but no change in distribution (Fig. 3G). Iron release from iron-loaded tissue did not sustain hepcidin 1 expression (Fig. 3C). Most prominently, these mice showed severe signs of illness and had to be sacrificed 5 days after poly-IC injection. Their livers were swollen, with a 42% increased wet weight. Liver tissue sections showed signs of irreversible cell damage with enlarged nuclei, macrosteatosis, hemorrhage, and infiltration by polymorphonuclear cells (Fig. 3E). Liver damage was confirmed by the terminal transferase dUTP nick end labeling (TUNEL) assay, which showed significantly more DNA strand breaks in iron-loaded FthΔ/Δ mice than iron-loaded Fthlox/lox mice or mice on a normal iron diet (Fig. 3F). Quantification of TUNEL staining in seven fields of four independent animals (≅1,500 cells) showed 9.4 ± 5.3% stained cells for iron-loaded FthΔ/Δ mice versus 1.8 ± 0.4% for iron-loaded Fthlox/lox mice (P < 0.025). For normally fed FthΔ/Δ and Fthlox/lox mice the values were 1.0 ± 0.7% and 0.9 ± 0.3%, respectively. Alanine aminotransferase (ALT) and aspartate aminotransferase (AST) activity in serum was also strongly increased (Fig. 3D). In addition, mRNA of heme oxygenase 1, a gene responding to oxidant signals, was induced 13-fold in iron-loaded FthΔ/Δ versus Fthlox/lox mice. When nonloaded FthΔ/Δ mice were injected with iron-dextran 2 months after ferritin H deletion, they showed the same acute liver damage as those preloaded with iron (not shown).
Severe physiological consequences of ferritin H deletion in iron-loaded mice. Mice were fed 14 days with 2.5% carbonyl iron diet prior to Mx-Cre induction and analyzed at day 5. (A) Immunoblot of liver proteins probed with anti-ferritin H and control anti γ-tubulin antibodies. (B) Liver iron in Fthlox/lox mice without (−Fe) (n = 4) or with (+Fe) high iron diet (n = 4) was compared to iron-loaded FthΔ/Δ mice (n = 4). (C) Hepcidin 1 mRNA (arbitrary units) of the same mice as in (B) was normalized to TATA box binding protein mRNA. (D) Cytoplasmic liver enzyme activity of ALT and AST in serum. ***P < 0.0005; **P < 0.005; *P < 0.05. (E) Morphological changes in liver tissue sections stained with hematoxylin-eosin of iron-loaded Fthlox/lox (a) and FthΔ/Δ mice (b-d) analyzed at day 5. Scale bar = 100 μm. The sections of FthΔ/Δ mice show extensive macrosteatosis (b), hemorrhage (c), and polymorphonuclear cell infiltration (d) with neutrophils (insert scale bar = 5 μm). Enlarged version in Supporting Fig. 3. (F) TUNEL assay on frozen liver sections from normal and iron-loaded Fthlox/lox and FthΔ/Δ mice. Iron-loaded FthΔ/Δ mice (d) show significantly more apoptotic or necrotic nuclei with DNA strand breaks. Scale bar = 100 μm. (G) Perl's Prussian blue staining of liver in iron-loaded Fthlox/lox and FthΔ/Δ mice at day 5 after Mx-Cre activation by poly-IC. Scale bar = 200 μm. (H) Perl's Prussian blue staining of liver in iron-loaded Fthlox/lox and FthΔ/Δ mice at day 10 after SA-Cre-ERT2 activation with tamoxifen. Scale bar = 200 μm. Enlarged version of panels (G) and (H) in Supporting Fig. 4.
We questioned whether poly-IC was responsible for liver damage, possibly by triggering an inflammatory response. Therefore, we crossed Fthlox/lox mice with SA-Cre-ERT2 mice expressing tamoxifen-inducible Cre under control of the hepatocyte-specific serum albumin promoter, which leaves the ferritin H gene intact in liver macrophages.26 Tamoxifen activation of Cre-ERT2 in 10-week-old iron-loaded mice reduced liver ferritin H mRNA to 6%–15% at day 10, reflecting the hepatocyte-specific deletion. Perl's staining prior to the deletion was similar to the one in Fthlox/lox;Mx-Cre mice (Fig. 3H). At day 10 after deletion, however, the staining in FthΔ/Δ;SA-Cre-ERT2 mice looked very different from the one in FthΔ/Δ;Mx-Cre mice. Noticeably, hepatocytes did not show iron staining, but some iron remained highly concentrated in a few cells, which we identified as macrophages and in rare instances as hepatocytes. Importantly, unlike FthΔ/Δ;Mx-Cre mice, the FthΔ/Δ;SA-Cre-ERT2 mice showed no signs of liver damage (Supporting Fig. 2). In order to create conditions similar to the Cre activation in Mx-Cre mice, poly-IC was injected together with tamoxifen. These mice showed again no signs of liver damage (not shown). We conclude that liver damage in FthΔ/Δ;Mx-Cre mice is not due to poly-IC, but rather iron toxicity.
Iron-Induced Cell Death of FthΔ/Δ Mouse Embryonic Fibroblasts.
To study the mechanism of cellular iron toxicity, we derived FthΔ/Δ and Fth+/Δ mouse embryonic fibroblasts, which showed identical cell proliferation rates and viability in the absence of iron salts. However, upon exposure to ferric ammonium citrate, FthΔ/Δ cells were 30-fold more sensitive to iron (median lethal dose [LD50] = 1.3 μg/mL) than FthloxNeo/loxNeo cells (LD50 = 36 μg/mL) and 120-fold more than Fth+/Δ cells (LD50 = 151 μg/mL) (Fig. 4A). FthΔ/Δ cell death could be rescued by wild-type ferritin H using an inducible vector (Fig. 4B). No rescue was observed upon expression of mutant ferritin H lacking ferroxidase activity (Fig. 4B). In all cell populations tested LD50 values correlated well with ferritin mRNA levels (Fig. 4C). Together, these data indicate an essential role of ferritin in the protection against iron toxicity.
Survival of mouse embryonic fibroblasts in iron-rich medium depends on the presence of functional ferritin H. Mouse embryonic fibroblasts were derived from FthloxNeo/loxNeo and Fth+/loxNeo mice, transfected with CMV-Cre-ERT and exposed to tamoxifen to derive FthΔ/Δ and Fth+/Δ cells. (A) FthΔ/Δ (○), FthloxNeo/loxNeo (●) and Fth+/Δ (⋄) cells were incubated with increasing concentrations of ferric ammonium citrate and cell viability measured after 4 days by the MTS assay. (B) FthΔ/Δ cells were transfected with cDNA of wild-type ferritin H (▪) or a ferritin H without ferroxidase activity due to 62E→K and 65H→G mutations (□). Iron toxicity was assessed as in (A). (C) Correlation between ferritin H mRNA expression and LD50 dose of ferric ammonium citrate.
To analyze directly whether increased cytoplasmic iron correlates with cell death, we measured IRP activity by an mRNA translation assay with an unstable d2EGFP (d2 enhanced green fluorescent protein) reporter linked to the 5′-iron responsive element of mouse ferritin L. In this assay, low cytoplasmic iron increases IRP activity, which represses d2EGFP translation. When free iron increases, IRP is inactivated and translation repression relieved. Thus, measuring d2EGFP steady-state levels by fluorescence-activated cell sorting (FACS) (Supporting Fig. 5) reflects cytoplasmic free iron levels. d2EGFP expression was increased 80% or more with iron concentrations at the LD50 or above. To reach an increased free iron pool, FthΔ/Δ cells needed much lower extracellular iron concentrations than FthloxNeo/loxNeo cells (Fig. 5A). The increase in free iron correlated with an enhanced accumulation of ROS as measured by dichlorofluorescein (DCF) already 3 hours after the addition of ferric ammonium citrate to cells (Fig. 5B). After 16 hours cells showed signs of mitochondrial depolarization and membrane permeability at or above the LD50 iron level (Fig. 5C and Supporting Fig. 6). Several inhibitors of oxidant activity were able to prevent mitochondrial depolarization and permeability transition (Fig. 5D). Notably, mitoquinone diminished iron toxicity strongly. Cells also underwent mitochondrial permeability transition, which was clearly visible in FthloxNeo/loxNeo cells exposed to high iron concentrations needed for mitochondrial depolarization (Supporting Fig. 7). This effect, however, was not visible in FthΔ/Δ at lower iron concentrations, which nonetheless induced depolarization. This indicates that mitochondrial permeability transition was not a prerequisite for mitochondrial depolarization.
Iron toxicity in mouse embryonic fibroblasts involves increased cytoplasmic iron levels, ROS formation, and mitochondrial depolarization. (A) FthloxNeo/loxNeo (●) and FthΔ/Δ (○) cells were infected with pSH-FTL-d2EGFP and unstable d2EGFP expression analyzed by the FACS at different iron concentrations (Supporting Fig. 5). Values report the fluorescence increase in d2EGFP-positive cells compared to conditions without iron addition (n = 5). (B) ROS formation was assessed by DCF at 3, 6, and 16 hours with different iron concentrations (n = 3). (C) Cells were stained at 6 or 16 hours after ferric ammonium citrate addition with tetramethyl rhodamine methyl ester (TMRM) for mitochondrial depolarization (black bars) and 7-amino-actinomycin D (7aaD) for loss of membrane integrity (white bars) and analyzed by FACS (Supporting Fig. 6). Values reflect the percent of cells that shifted from normal to depolarized or to membrane permeable after subtracting background from untreated cells (n = 5). (D) Same assay as in (C) in the presence of inhibitors mitoquinone (MitoQ), Ebselen (Ebs), and trifluoroperazine (TFP) for 16 hours (n = 5). Inhibition was significant (P < 0.05 by paired t test) with all three inhibitors in FthΔ/Δ cells both for depolarization and cell death, whereas in FthloxNeo/loxNeo cells only depolarization was significantly inhibited by Ebselen and trifluoroperazine.
Discussion
Our data provide clear evidence for the requirement of ferritin H in iron storage and detoxification. We conclude that iron storage in vivo cannot be maintained by ferritin L alone. The importance of ferritin H and its ferroxidase activity for ferritin assembly and iron deposition was previously described in vitro and for cell lines with diminished or increased ferritin H levels.27, 28 Our results extend these studies to tissues under physiological conditions. As iron stores and ferritin H are rapidly reduced in deleted tissues, we conclude that ferritin is continuously degraded. This agrees with reports of ferritin half-lives of 3 days in rat liver29 and <24 hours in tissue culture cells.30 Thus, iron stores are dynamic, making the metal continuously available for the biosynthesis of iron-containing proteins. Upon Fth gene knockout, iron cannot be redeposited into ferritin and leaves the tissue as observed for spleen and liver. This explains the increased serum iron and transferrin saturation. However, about 20% to 40% of the nonheme liver iron remained with a granular appearance, and further work is required to resolve the identity of these nondegradable iron deposits. Attempts to trace the redistribution of iron showed no significant increase in other tissues (Fig. 1A), possibly because C57BL/6 mice accumulate modest hepatic iron stores.31 Studies using 59Fe should detail this further.
Hepcidin 1 is central to the regulation of intestinal iron absorption and iron recycling from reticulocytes.14-16, 19, 20 Our results showing a correlation between hepcidin 1 mRNA induction and the increase in serum iron and transferrin saturation seem to reflect an iron-dependent or holo-transferrin-dependent signaling to the hepcidin 1 gene.14, 32 More surprising is the diminished hepcidin 1 mRNA expression in animals fed with high iron and then deleted of ferritin H (Fig. 3C). However, proper cell responsiveness and signaling to the hepcidin 1 gene might be altered by severe liver damage. We and others have observed that isolated primary hepatocytes in cell culture show diminished hepcidin 1 mRNA expression and are unresponsive to ferric iron in the medium (L. Vanoaica, unpubl. obs.)14, 33 unless treated within a few hours with iron-loaded holo-transferrin.32
The acute liver damage observed within just a few days of the ferritin H deletion in iron-loaded Mx-Cre mice can neither be attributed to iron loading alone, nor to activation of the interferon-α pathway by poly-IC, as control mice showed no signs of damage. Our results in the hepatocyte-specific SA-Cre-ERT2-mediated ferritin H deletion with tamoxifen, or with tamoxifen plus poly-IC, also exclude that poly-IC alone provokes liver damage. Cell death appears, therefore, directly related to iron liberated upon ferritin H deletion and degradation. Whether liberated iron is sufficient to provoke liver damage or involves signals due to ferritin deletion in other tissues remains to be investigated. The fact that ferritin H-deleted mice fed a normal diet for 2 months and then injected with iron dextran show acute liver failure argues in favor of a direct effect of iron toxicity on liver cells. Interestingly, the hepatocyte-specific SA-Cre-ERT2-mediated ferritin H deletion does not provoke damage (Supporting Fig. 2). This suggests that hepatocytes are protected because adjacent macrophages absorb large amounts of liberated iron (Fig. 3), which is not the case in Mx-Cre mice.
We have gained mechanistic insight into the cause of iron toxicity in embryonic fibroblasts and show again that ferritin H protects against iron toxicity (Fig. 4). The protection depends directly on the capacity of ferritin to store iron, because the ferritin H deletion can be rescued by wild-type ferritin H, but not a mutant that lost its ferroxidase activity and iron storage capacity.3, 27 As a first consequence of the ferritin H deletion, the intracellular free iron pool is increased as assessed by the relief of translation regulation of an iron responsive element containing GFP reporter construct (Fig. 5). This completes previous conclusions drawn in K562 cells where ferritin H overexpression reduced the free iron pool.34 Probably as a direct consequence, ROS increased rapidly within 3 hours in FthΔ/Δ cells, even at low extracellular iron concentrations. Although the causative connection between free iron and ROS is well documented, we confirm that ferritin H expression is absolutely essential to prevent ROS formation. Others have shown that small interfering RNA (siRNA)-mediated knockdown of ferritin H in K562 cells increased free iron and ROS production,35 and rendered cells more sensitive to H2O2.28 Moreover, ferritin H induction by nuclear factor kappa B (NF-κB) can counteract tumor necrosis factor alpha (TNFα)-induced apoptosis by scavenging iron and preventing ROS formation.6 We extend these previous observations to the death mechanism with evidence that mitochondrial depolarization and permeability transition precede cell death. Mitochondrial depolarization appears to be a direct consequence of iron-mediated ROS formation as they show similar iron dose-response curves with a marked difference between FthΔ/Δ and FthloxNeo/loxNeo cells. The causative relationship is further underlined by the time-delay between ROS formation and mitochondrial depolarization, as well as the fact that inhibitors with antioxidant functions inhibit mitochondrial depolarization significantly. Iron was previously implicated in cold-induced apoptosis and shown to trigger mitochondrial permeability transition in hepatocytes.36 ROS and mitochondrial permeability transition also plays a role in ischemia-reperfusion injury and iron poisoning.37 Whether the liver damage observed in our Mx-Cre-deleted mice is a direct consequence of mitochondrial depolarization remains to be studied. Our mice should become particularly useful to evaluate the role of iron-catalyzed ROS in specific tissues, and to establish the function of ferritin not only in acute iron toxicity, but also in long-term protection against DNA mutagenesis and aging.
Acknowledgements
We thank Olav Zilian, Fabienne Seiler, and Michel Aguet for advice and the Mx-Cre mouse, Andreas Trumpp for the Nes-Cre1 mouse and pKI-Cre ERT, and Daniel Metzger and Pierre Chambon for the SA-Cre-ERT2 mouse, Michael Murphy for mitoquinone, and Michael Reth for pAN-MerCreMer. We thank Sanjiv Luther for the macrophage staining in spleen and the MIM facility of ISREC for histology.
