Glycosylphosphatidylinositol-anchored Ceruloplasmin Is Required for Iron Efflux from Cells in the Central Nervous System*

Ceruloplasmin (Cp) is a ferroxidase that converts highly toxic ferrous iron to its non-toxic ferric form. A glycosylphosphatidylinositol (GPI)-anchored form of this enzyme is expressed by astrocytes in the mammalian central nervous system, whereas the secreted form is expressed by the liver and found in serum. Lack of this enzyme results in iron accumulation in the brain and neurodegeneration. Herein, we show using astrocytes purified from the central nervous system of Cp-null mice that GPI-Cp is essential for iron efflux and not involved in regulating iron influx. We also show that GPI-Cp colocalizes on the astrocyte cell surface with the divalent metal transporter IREG1 and is physically associated with IREG1. In addition, IREG1 alone is unable to efflux iron from astrocytes in the absence of GPI-Cp or secreted Cp. We also provide evidence that the divalent metal influx transporter DMT1 is expressed by astrocytes and is likely to mediate iron influx into these glial cells. The coordinated actions of GPI-Cp and IREG1 may be required for iron efflux from neural cells, and disruption of this balance could lead to iron accumulation in the central nervous system and neurodegeneration. Mechanisms maintain iron homeostasis at the cellular level are crucial for the viability of cells. Excess or inappropri-ately shielded cellular iron can lead to cell death. The effects of this toxicity are especially noticeable in the brain, spinal cord, modified Eagle’s medium (DMEM; Invitrogen) twice and incubated in DMEM containing vitamins and penicillin/streptomycin for 1 h at 37 °C to remove any transferrin-bound iron. Influx Study— After 1 h, DMEM was replaced with Sato’s modified chemically defined serum-free medium (18) without transferrin. Radio-*

Mechanisms to maintain iron homeostasis at the cellular level are crucial for the viability of cells. Excess or inappropriately shielded cellular iron can lead to cell death. The effects of this toxicity are especially noticeable in the brain, spinal cord, and other parts of the central nervous system (CNS), 1 because the mature CNS lacks regenerative capabilities. Although iron is essential for a variety of biological functions, such as oxygen transport, mitochondrial respiration, and DNA synthesis, it can generate highly toxic free radicals because it is a transition metal.
In its divalent state (Fe 2ϩ ), iron is highly toxic when it reacts with hydrogen peroxide and molecular oxygen to produce free radicals. Free radical formation can promote lipid peroxidation, DNA strand breaks, degradation of biomolecules, and eventually cause cell death (1). Therefore, organisms have developed mechanisms to prevent increase of the iron-pool while maintaining sufficient levels for metabolic use. However, these ho-meostatic mechanisms can get misregulated and cause iron deficiency or iron overload. The safe conversion of Fe 2ϩ to Fe 3ϩ is catalyzed predominantly by a copper-binding glycoprotein, ceruloplasmin [Cp, EC 1.16.3.1 (2)]. Humans with mutations of the ceruloplasmin gene (aceruloplasminemia) show iron accumulation in various organs including the liver and brain, which is noticeable by the age of 45-55 years (3,4). Ceruloplasmin null mutant mice also show accumulation of iron in the liver (5,6) and CNS (6). Moreover, increased levels of iron and lipid peroxidation have been observed in the cerebrospinal fluid (CSF) of these patients (7). The accumulation of iron in the CNS correlates with neurodegeneration in humans and mice (3,6). We have shown previously that the rat brain expresses mainly the GPI-anchored form of ceruloplasmin, which is predominantly expressed by astrocytes (8). Human GPI-Cp was also cloned recently (9). DMT1 (also Nramp2/DCT1/SLC11A2) is a divalent metal transporter found in duodenal enterocytes (10). Mutations in DMT1 seen in mk mice with microcytic anemia and the Belgrade (b) rat cause defects in iron transport from the lumen of the gut into enterocytes and from plasma transferrin into erythroid precursors (11,12). DMT1 transports non-heme, ferrous iron (Fe 2ϩ ) instead of ferric (Fe 3ϩ ) iron, which is bioavailable. Therefore, the existence of a membrane reductase to convert ferric iron to ferrous iron was proposed. A mammalian ferric reductase, duodenal cytochrome b (Dcytb/Cybrd1) was cloned recently from mouse duodenum (13). In the gut, Dcytb is mainly localized in the brush border membrane and proposed to function as a ferric reductase. An iron exporter called IREG1 (also MTP1/ferroportin1/SLC11A3) is found on the basolateral side of mouse duodenal enterocytes and several other cell types (14 -16). Although much is known about iron absorption and efflux in enterocytes in the gut, the mechanisms controlling iron homeostasis in neural cells in the CNS is still not well understood.
Here, we report on experiments done to assess how GPI-Cp, which is expressed predominantly by astrocytes in the CNS, regulates iron levels in astrocytes and to determine its interactions with two divalent metal ion transporters, DMT1 and IREG1.

EXPERIMENTAL PROCEDURES
Iron Influx/Efflux Studies-Astrocytes were purified from the brains of neonatal wild-type (Cp ϩ/ϩ ) and Cp-null mice (Cp Ϫ/Ϫ ) obtained from littermates and cultured as described previously (17). The Cp-null mice were generated in this laboratory as reported previously (6). Cultured astrocytes were plated on poly-L-lysine (Sigma)-coated 24-well plates 2 days before the experiment at a density of 3 ϫ 10 5 cells/well. Cells were washed with serum-free Dulbecco's modified Eagle's medium (DMEM; Invitrogen) twice and incubated in DMEM containing vitamins and penicillin/streptomycin for 1 h at 37°C to remove any transferrin-bound iron.
Influx Study-After 1 h, DMEM was replaced with Sato's modified chemically defined serum-free medium (18) without transferrin. Radio-labeled iron ( 59 FeCl 3 , 1 Ci/well; PerkinElmer Life Science) mixed with non-labeled FeCl 3 (total of 40 M) and L-ascorbate (Sigma) was added to keep iron in its ferrous form (molar ratio of FeCl 3 to L-ascorbate was 1:44). After different culture periods, cells were treated with PRONASE protease (Calbiochem) for 1 h at 4°C to remove membrane-bound iron and then lysed in 1 N NaOH. Additional experiments with EDTA (500 M) and PRONASE treatment gave results similar to those of PRONASE alone (data not shown). The amount of radioactivity bound nonspecifically to the cells was estimated by adding the media containing radiolabeled iron to culture wells and removing it within 1-2 min, washing the wells, and measuring with a ␥-counter (Amersham Biosciences). This radioactivity level, which was found to be extremely low, was considered background value and was subtracted from all values at each data point. This value is considered zero at the 0-h time point. Cells were cultured in 5% CO 2 at 37°C in the media containing radiolabeled iron until the desired time points (i.e. 12, 24, and 48 h). Sister cultures were treated in the same manner without radioactive iron, and viable cell numbers were estimated by trypan blue exclusion. The amount of radioactivity that was taken up by the cells was converted into picomoles of iron using a standard graph and normalized to value per 10 6 cells. The standard graph was plotted using counts per minute versus serial dilution of 1 l of 59 FeCl 3 .
Efflux Study-Cultured astrocytes in 24-well plates were washed and incubated in serum-free medium for 1 h, as was done for the influx study. Cells were then loaded with medium containing radiolabeled iron for 24 h (same condition as above). After 24 h, cells were washed twice with DMEM, and serum-free Sato's chemically defined medium without transferrin was added to the cultures. At each time point (0, 12, 24, and 48 h) cells were detached, pelleted, and lysed in 1 N NaOH. In addition, a 200-l aliquot of culture medium from each time point was collected to measure the amount of iron released into the medium. Radioactivity in both cell pellet and culture medium was measured.
All radioactivity measurements at each time point were done in quadruplicate and repeated in three separate experiments. Results are shown as mean Ϯ S.E. Two-sample Student's t test was used to determine statistical significance.
Immunocytochemistry-Purified neonatal rat astrocytes were plated on PLL-coated round glass coverslips and cultured in DMEM containing 10% fetal bovine serum. Cells were washed in Hank's balanced salt solution and stained with anti-ceruloplasmin monoclonal antibody (1A1; 1:100) for 30 min at room temperature. Cells were washed, and rhodamine-conjugated goat anti-mouse IgG was added for 30 min. Cells were fixed in acetic acid/ethanol (5:95 (v/v)) at Ϫ20°C for 20 min. Permeabilized cells were stained with either anti-IREG1 or anti-DMT1 polyclonal rabbit antibodies (1:200; Alpha Diagnostics) for 30 min at room temperature and visualized by fluorescein-conjugated goat antirabbit IgG (all secondary antibodies from Jackson Immunoresearch). Confocal microscopy was used to generate 0.8-m sectioned images of cells.
Coimmunoprecipitation Assay-Purified rat astrocytes were soni-cated and centrifuged at 500 ϫ g for 10 min to pellet the nuclei. Supernatant was centrifuged at 137,000 ϫ g for 1 h at 4°C. The pellet was resuspended in extraction buffer (10 mM Tris/HCl, pH. 7.4, 150 mM NaCl, 5 mM EDTA, 1.25% Triton X-100, and protease inhibitors). Protein concentration in the extracts was measured (DC Protein Assay kit; Bio-Rad). Ceruloplasmin was immunoprecipitated using 200 g of the mouse monoclonal anti-ceruloplasmin antibody (1A1) conjugated to an immunoprecipitation column (Pierce Chemical), and 300 g of the protein extract was incubated overnight. After washing steps, the bound protein was eluted according to the manufacturer's protocol. Immunoprecipitated samples were denatured and separated on 7.5% SDS-PAGE gel and Western blotted to detect Cp, DMT1, and IREG1. A column without conjugated primary antibody was used to obtain the negative control.

RESULTS
How does GPI-Cp regulate iron levels in neural cells so that in its absence, iron accumulates in cells in the CNS, as seen in aceruloplasminemia and Cp Ϫ/Ϫ mice? GPI-Cp could regulate normal iron levels in the CNS by either limiting how much iron enters the cell or facilitating iron efflux. To distinguish between these two possibilities, we assessed iron influx and efflux with 59 Fe in vitro using astrocytes purified from the Cp Ϫ/Ϫ and Cp ϩ/ϩ mice. Astrocytes were used because our earlier work indicated that GPI-Cp is expressed mainly in astrocytes in the rat CNS (20). For the influx studies, astrocytes were cultured in medium containing 59 Fe/ascorbate in Sato's modified serumfree medium. At various time intervals ranging from 12 to 48 h, the cells were harvested and the amount of radiolabeled iron within the cells was estimated. These studies showed that the amount of iron influx is similar in Cp Ϫ/Ϫ and Cp ϩ/ϩ mice (Fig.  1). The rate of influx is 0.1 pmol/10 6 cells/h, which is ϳ2% of the influx reported for macrophages via the transferrin receptor (21). For efflux studies, astrocyte cultures were first loaded with 59 Fe in medium for 24 h. The radiolabeled medium was then removed, cultures were washed, and fresh Sato's serumfree, transferrin-free medium was added. The supernatant and cells were sampled at varying intervals from 12 to 48 h. In cultures of wild-type astrocytes, almost 70% of the iron within the cell was effluxed within 48 h ( Fig. 2A). In contrast, iron efflux was severely impaired in astrocyte cultures from Cp Ϫ/Ϫ mice. Less than 5% of the iron was effluxed in the 48-hour period (Fig. 2B). To assess whether the secreted form of ceruloplasmin can compensate for the lack of GPI-Cp, cultures of astrocytes from Cp Ϫ/Ϫ mice were first loaded with 59 Fe for 24 h, as noted above. After removing the radiolabeled medium, fresh serum-free medium containing soluble ceruloplasmin was added to the cultures. Ceruloplasmin at 1 g/ml (i.e. the concentration found in CSF) increased iron efflux minimally by 9% FIG. 1. Role of ceruloplasmin in iron influx in astrocytes. Nontransferrin-mediated uptake of 59 Fe into astrocytes at 12, 24, and 48 h after addition to cultures of serum-free medium containing radiolabeled iron. Iron uptake is similar in cultures from Cp ϩ/ϩ (f) and Cp Ϫ/Ϫ (Ⅺ) mice. The results are shown as means Ϯ S.E. (Fig. 3), whereas ceruloplasmin at 300 g/ml (i.e. the concentration in serum) enhanced efflux to 50% by 48 h (Fig. 3). Increasing the ceruloplasmin concentration to 400 g/ml increased efflux further to 65% (data not shown). These results indicate that the low amount of ceruloplasmin in CSF is insufficient to mediate iron efflux, and concentrations of soluble ceruloplasmin equal to or higher than serum levels are need to efflux iron to the levels achieved by GPI-Cp.
Because ceruloplasmin is an enzyme, it must regulate iron efflux via iron transporters. We first carried out experiments to assess whether the iron efflux transporter IREG1 and the iron influx transporter DMT1 are expressed by astrocytes. RT-PCR and Western blotting analysis showed that DMT1 and IREG1 mRNA and protein are expressed by astrocytes (Fig. 4, A, B, and D). RT-PCR analysis also indicates that the ferrireductase Dyctb is expressed by these cells, which could make soluble ferrous iron available for influx via DMT1 over prolonged periods in vitro (13) (Fig. 4C). Confocal microscopy of doubleimmunofluorescence labeling of cultured astrocytes showed that IREG1, but not DMT1, is colocalized with GPI-Cp on the surface of astrocytes (Fig. 5). These data suggest that there might be a specific interaction of GPI-Cp with the iron exporter IREG1. To confirm these immunocytochemical results and to assess whether there is a direct physical interaction between these two molecules, GPI-Cp was immunoprecipitated with the monoclonal antibody against Cp (1A1) and Western blotted to detect DMT1 and IREG1. These results showed that IREG1 but not DMT1 is coimmunoprecipitated with GPI-Cp, indicating a direct interaction of this GPI-anchored ferroxidase with IREG1 ( Fig 6A).
Given the evidence of interaction between GPI-Cp and IREG1, it is possible that the lack of iron efflux in cultures of astrocytes from Cp Ϫ/Ϫ mice could be caused by a down-regulation of the expression of IREG1 in Cp Ϫ/Ϫ mice. To determine whether this is the case, total protein from the brains of Cp Ϫ/Ϫ and Cp ϩ/ϩ mice were Western blotted to detect IREG1. The level of expression of IREG1 protein is similar in Cp Ϫ/Ϫ and Cp ϩ/ϩ mice (Fig. 6B), indicating that this transporter requires ceruloplasmin to efflux iron from astrocytes. DISCUSSION Although there has been an increase in our understanding of iron transport and regulation of iron levels in hematopoietic tissue, macrophages, and enterocytes in the gut, there is little direct evidence for the molecular mechanisms underlying iron transport across neural cell membranes in the CNS (reviewed in Ref. 22). Once iron gets past the endothelial cells in the CNS, its uptake into neural cells could occur via transferrin and/or non-transferrin-mediated mechanisms. Transferrin receptors are expressed by oligodendrocytes and neurons in the CNS, whereas their expression in astrocytes has been more difficult to detect, although some recent studies suggest its presence in astrocytes (23). However, transferrin levels in the CSF, which reflects the amount available to CNS tissue, is extremely low (about 1% compared with serum; 0.1-0.28 mol/L in humans (24)), suggesting that under normal conditions, transferrinmediated uptake may not be significant, particularly in astrocytes. This is further supported by the findings in hypotransferrinemic mice, which show a normal distribution of iron in the brain (25,26). These data suggest that non-transferrinmediated mechanisms are likely to be involved in iron influx into cells in the brain. We show here that astrocytes can indeed take up iron though non-transferrin-mediated mechanisms. We also show that astrocytes express both the divalent metal transporter DMT1, which has specificity for divalent metals, including ferrous iron, and the ferric reductase Dcytb, indicating a role for these molecules in mediating iron uptake in the CNS. Our data show that the rate of iron uptake into astrocytes is ϳ2% that of macrophages (21) and ϳ3.5% that of hepatocytes (27). This slow rate of uptake into astrocytes might explain the slow accumulation of iron and late onset of neurological symptoms in aceruloplasminemia.
Iron deficiency states do not seem primarily to affect the brain, suggesting that it is capable of possibly reusing iron or mobilizing it from other sources to maintain normal physiological functions. In addition, high levels of serum iron, as occurs in hemochromatosis, does not lead to iron accumulation in the CNS (reviewed in Ref. 28) possibly indicating a unique homeostatic mechanism at either the level of the endothelial cells in the CNS or in cells that surround CNS capillaries, namely astrocytes. In contrast, iron accumulates in the brain in aceruloplasminemia in humans and mice (3-6), in which serum iron FIG. 2. Role of ceruloplasmin in iron efflux from astrocytes. A, iron efflux from astrocytes from Cp ϩ/ϩ mice. Cells were loaded with 59 Fe for 24 h, washed, and cultured in non-radioactive, serum-free, transferrin-free medium. The amount of radiolabeled iron was measured in the cell pellet and culture medium at 0, 12, 24, and 48 h. About 70% of the radiolabeled iron in the cells is effluxed by 48 h (q). A corresponding increase in radiolabeled iron is detected in the culture medium (E), indicating iron efflux. Results are shown as means Ϯ S.E. B, iron efflux from astrocytes from Cp Ϫ/Ϫ mice. Cells were loaded with 59 Fe and cultured as in A. The percentage of radiolabeled iron in the cells remains unchanged through out the 48-h period (q). Radiolabeled iron is also not detected in the culture medium (E) during this period, indicating that iron does not efflux from astrocyte from Cp Ϫ/Ϫ mice. Efflux at each time point was compared for statistical significance with the corresponding cell pellet or supernatant values from Cp ϩ/ϩ mice shown in A using a two-sample Student's t test. *, p Ͻ 0.05; **, p Ͻ 0.008.

FIG. 3. Effect of soluble Cp on iron efflux from astrocytes from
Cp ؊/؊ mice. To assess whether secreted ceruloplasmin can compensate for GPI-Cp to efflux iron, astrocytes from Cp Ϫ/Ϫ mice were loaded with iron, as in Fig. 2, and cultured in serum-free, transferrin-free medium, to which was added serum ceruloplasmin at a concentration of 1 g/ml (equivalent to CSF levels) and 300 g/ml (equivalent to serum levels). Control Cp Ϫ/Ϫ cells (q) were cultured in the same medium without Cp. Very minimal iron efflux occurred with 1 g/ml of ceruloplasmin (OE), whereas a 50% efflux was seen with 300 g/ml of ceruloplasmin (). Results represent means Ϯ S.E. *, p Ͻ 0.02. and serum transferrin saturation levels are very low (5,6). We show here that a lack of ceruloplasmin expression by astrocytes leads to disruption of iron efflux. The requirement of GPI-Cp for iron efflux from astrocytes indicates that oxidation of ferrous iron transported across the cell membrane, probably via the transmembrane transporter IREG1, is an essential step. Recently, soluble ceruloplasmin was shown to be essential for iron transport across the oocyte membranes when IREG1 was expressed in these cells in vitro (16). Earlier studies on the yeast showed that a transmembrane multicopper oxidase Fet3 with homology to ceruloplasmin interacts with an iron permease, Ftr1p, to transport iron across the cell membrane (29,30). We now show that IREG1 is expressed by astrocytes from the brain and is physically associated with GPI-Cp. Furthermore, we also provide evidence that the expression of IREG1 alone is insufficient to allow iron efflux in the absence of GPI-Cp. BecauseIREG1 transports ferrous iron that is highly toxic, ceruloplasmin as the major ferroxidase in the CNS plays a crucial role in detoxifying it to the ferric state. The inability of IREG1 to efflux ferrous iron in the absence of ceruloplasmin may therefore serve as a protective mechanism to prevent efflux of toxic ferrous iron, leading to the rapid generation of free radicals. The slow accumulation of iron intracellularly may eventually surpass the ability of the intracellular sequestering capacity of the cell and lead to cell damage and death.
Because soluble ceruloplasmin in the CSF, which is produced by the choroid plexus, is extremely low (1 g/ml compared with 300 g/ml in the serum), it is likely to contribute minimally to the ferroxidase activity in the CNS. Astrocytes are known to be the only cell type in the CNS to express ceruloplasmin (20,31), which is of the GPI-anchored form (8,20). The severe accumulation of iron in the brain in cases of aceruloplasminemia indicates that ceruloplasmin expressed on the surface of astrocytes plays an important role in the maintenance of normal iron levels in the CNS and its mobilization out of the CNS. Furthermore, GPI-Cp expressed by astrocytes also seems to be capable of effluxing iron from neurons, because iron accumulates in neurons in aceruloplasminemia. How GPI-Cp regulates neuronal iron levels is not yet known but may involve the transfer of GPI-Cp from the astrocyte to the neuronal cell membrane, because GPI-anchored proteins can transfer from one cell to another by cell-to-cell contact (32). This might explain why Cp is found in both astrocytes and neurons in the CNS (33), whereas Cp mRNA is found only in the astrocytes (31,33). One advantage of the GPI-anchored form or ceruloplasmin in the brain is that it reduces the need to have high levels of ceruloplasmin in the CSF. Our results indicate that at least 300 to 400 g/ml of ceruloplasmin would be needed in the CSF to efflux iron from astrocytes. This contrasts sharply with the total protein concentration of 350 g/ml in human CSF (34). Some of the other features that GPI anchors confer to proteins may also aid in the physiological function of ceruloplasmin; e.g. GPI-anchored proteins, which are located in lipid-rich microdomains, have a much greater degree of lateral mobility (35) and may also help serve as an apical targeting signal (36). The latter feature may lead to the targeting of GPI-Cp in vivo to the astrocytic end-feet that surround capillaries in the CNS and thus position the iron efflux mechanism near blood vessels via which iron can be mobilized out of the CNS.
This work provides the basis for understanding why iron accumulates in the CNS in aceruloplasminemia and may have