INTRODUCTION

Iron plays an important role as a cofactor for various enzymes, such as the cytochromes of the electron transport chain and ribonucleotide reductase. On the other hand, free iron can generate highly toxic free radicals because it is a redox-active transition metal (1). A number of enzymes, binding proteins, and transporters have been identified that are involved in mobilizing, transporting, and sequestering iron (1-3). Recent studies on the yeast Saccharomyces cerevisiae have resulted in the identification of several proteins, such as Fet3 and Ftr1, which directly participate in iron transport in this organism (4-6). The mammalian homologues of many of these proteins have yet to be identified. Ceruloplasmin, the major ferroxidase of plasma (300-450 µg/ml), is required for iron transport by transferrin. The oxidation of ferrous iron (Fe(II)) to ferric iron (Fe(III)) mediated by ceruloplasmin is necessary for iron incorporation into transferrin, since transferrin only binds the ferric form of iron. As a ferroxidase, ceruloplasmin might also play a role in a transferrin-independent iron uptake system, such as the one identified by Kaplan and colleagues (7), which requires reduction of iron at the cell surface (reviewed in Ref. 1).

Direct evidence for the role of ceruloplasmin in iron metabolism comes from studies of individuals with aceruloplasminemia, a hereditary deficiency of ceruloplasmin (8-15). These individuals have very little or undetectable levels of ceruloplasmin and severe intracellular iron accumulation in a number of organs, including the brain, particularly in the deep extrapyramidal motor nuclei, where it is associated with neurodegeneration. The neurodegeneration is likely to be a consequence of oxidative stress induced by the oxidation of ferrous iron by agents such as hydrogen peroxide (1). In support of this, Miyajima et al. (16) reported a dramatic increase in the levels of lipid peroxidation in the plasma of individuals with aceruloplasminemia. A number of other neurodegenerative diseases, including Parkinson's disease, Alzheimer's disease, amyotrophic lateral sclerosis, multiple sclerosis, and Hallervorden-Spatz disease are also associated with altered brain iron metabolism and free radical injury (17). It is therefore possible that ceruloplasmin might contribute to the pathology seen in these neurodegenerative diseases as well.

Although generally considered a soluble plasma protein of hepatic origin, we now provide evidence using a monoclonal antibody, mAb¹ 1A1, of a novel GPI-anchored form of ceruloplasmin that is localized to the surface of astrocytes in the central nervous system. The cell surface localization of ceruloplasmin is not seen on hepatocytes and cells of the choroid plexus, both of which are known to secrete ceruloplasmin. Since iron deposition occurs in the brain in aceruloplasminemia and because the level of the secreted form of ceruloplasmin in the cerebrospinal fluid is very low, this novel membrane-associated form of ceruloplasmin is likely to play an important role in iron metabolism in the central nervous system.

EXPERIMENTAL PROCEDURES

Astrocyte cultures were prepared from neonatal rat cerebral cortex as described previously (18). Astrocytes and C6 glioma cells were cultured in Dulbecco's modified Eagle's medium (DMEM) containing 10% fetal bovine serum, vitamins, and penicillin and streptomycin (Life Technologies, Inc.) in tissue culture flasks (Nunc). For certain immunocytochemistry experiments and the ferroxidase assay, cells were replated in serum-free Neurobasal Medium with G-5 supplement (Life Technologies, Inc.).

Membrane protein extracts were prepared from C6 glioma cell tumors raised in nude rats. Tumors were generated by injecting approximately 5 × 10⁵ cultured C6 glioma cells subcutaneously into Nu/Nu nude rats (Charles River). After 1-2 weeks, tumors were harvested and homogenized using a motor-driven Dounce homogenizer in ice-cold hypotonic buffer (10 mM HEPES, 1.5 mM MgCl₂, 5 mM KCl, pH 7.4) containing phenylmethylsulfonyl fluoride and soybean trypsin inhibitor (both at 50 µg/ml), leupeptin and aprotinin (both at 2 µg/ml), N-ethylmaleimide (1 mM), and NaN₃ (0.02%). The homogenate was first centrifuged at 1000 × g for 10 min, and the resulting supernatant centrifuged at 100,000 × g for 60 min. The pellet was then solubilized in ice-cold 1% Nonidet P-40 solubilization buffer (150 mM NaCl, 25 mM Tris, 0.2 mM MgCl₂, 0.2 mM CaCl₂, pH 7.4 (TBSS), 1% Nonidet P-40, 0.02% NaN₃) containing protease inhibitors (as above), and the supernatant clarified by centrifugation at 100,000 × g for 30 min. Detergent extracts of purified membranes of cultured C6 glioma cells were also made using a similar protocol.

An immunoaffinity column consisting of mAb 1A1 conjugated to cross-linked agarose beads (Affi-Gel-10; Bio-Rad) was made according to the manufacturer's instructions. Briefly, 2 ml of Affi-Gel-10 beads was added to 5 ml of purified mAb 1A1 in phosphate-buffered saline (3 mg/ml) and incubated for 4 h at 4 °C. After washing with phosphate-buffered saline, the beads were blocked with 0.5 M ethanolamine (pH 7.4) for 30 min. A 5-ml column of Affi-Gel-10 beads conjugated to ovalbumin (10 mg/ml beads) was also made using the same protocol.

The membrane protein extracts prepared from the C6 glioma tumors or the cultured C6 glioma cells were passed through the ovalbumin column. Eight ml of the flow-through from the ovalbumin column was added to the mAb 1A1 immunoaffinity column and incubated for 18 h at 4 °C. The beads were then washed with 20 bed volumes of TBSS containing 1% Nonidet P-40. Bound protein was eluted with a low pH buffer (200 mM glycine, pH 2.5, 0.1% Nonidet P-40), collected in 1-ml fractions, and immediately neutralized with 90 µl of 1 M Tris (pH 11). The fractions were subsequently dialyzed, concentrated, and analyzed by SDS-PAGE (19) and silver staining (20).

Dialyzed and concentrated fractions of the immunoaffinity-purified 1A1 antigen were pooled and subjected to SDS-PAGE using previously described protocols for protein sequencing (21). Following electrophoresis, the protein was electroblotted onto a polyvinylidene difluoride (PVDF) membrane (21). The protein band was visualized by staining with Coomassie Brilliant Blue (R-250; Bio-Rad), and the 135-kDa band was excised for sequencing. N-terminal amino acid sequencing by automated Edman degradation was performed at the Sheldon Biotechnology Center (McGill University) using a Porton Instruments PI 2090E Integrated Microsequencing System consisting of a gas phase sequencer on-line with a high pressure liquid chromatograph. A protein data base homology search was performed using the BLAST program (22).

Proteins transferred onto PVDF membranes according to published protocols (23) were blocked with 3% ovalbumin and then incubated with goat anti-human ceruloplasmin (1:200 to 1:400; Sigma) for 2 h at room temperature or overnight at 4 °C. Membranes were washed with TBSS containing 0.05% Tween 20, incubated with biotinylated rabbit anti-goat IgG (1:400; Vector) for 90 min at room temperature, washed, and incubated with alkaline phosphatase conjugated to avidin (1:400; Pierce) for 90 min at room temperature. Membranes were washed as before and developed using nitro blue tetrazolium and 5-bromo-4-chloro-3-indolyl phosphate (Promega) in alkaline phosphatase buffer. Control blots were incubated with normal goat serum (1:200 to 1:400) in place of primary antibody.

Ninety 75-cm² flasks of C6 glioma cells were harvested with Hanks' balanced salt solution (Life Technologies, Inc.) containing 5 mM EDTA. Following centrifugation, the cells were resuspended in DMEM containing 600 milliunits/ml PI-PLC (Boehringer Mannheim) in a volume of 4 ml and incubated for 60 min at 37 °C to release GPI-anchored proteins. The samples were then centrifuged and 1A1/ceruloplasmin purified from the supernatant as described above. The purified protein was Western blotted with an affinity-purified antibody that recognizes the inositol 1,2-cyclic phosphate moiety, which is uniquely found only in PI-PLC-cleaved GPI-anchored proteins (24) (anti-CRD antibody, 5 µg/ml; Oxford Glycosystems). The blots were then incubated with the secondary antibody (1/400 donkey anti-rabbit conjugated to biotin; Pierce) for 60 min, incubated with avidin-alkaline phosphatase for another 60 min, and reacted with nitro blue tetrazolium and 5-bromo-4-chloro-3-indolyl phosphate in alkaline phosphatase buffer with appropriate washing between incubations. Control blots were treated identically, except for the omission of the primary (anti-CRD) antibody. For immunofluorescence labeling, astrocytes or C6 glioma cells were treated with 600 milliunits/ml or 5 units/ml PI-PLC as described above and then stained with the mAb 1A1 and Nuclear Yellow.

Immunocytochemistry was performed on cultures of astrocytes, C6 glioma cells, choroid plexus, and liver grown on poly-L-lysine-coated glass coverslips. All incubations were carried out for 30 min at room temperature. Cell surface labeling was performed by incubating live cultures with mAb 1A1 ascites (1:200) or goat anti-human ceruloplasmin (1:500; Sigma), followed by a rhodamine-conjugated goat anti-mouse IgG (1:250; Cappel) or a fluorescein isothiocyanate-conjugated donkey anti-goat IgG (1:250; Cappel), respectively. Cultures were then fixed with acetic acid:ethanol (1:9) for 20 min at 4 °C. Astrocyte and C6 glioma cultures were also double-labeled for glial fibrillary acidic protein (GFAP) using either a rabbit anti-human GFAP and fluorescein isothiocyanate-conjugated goat anti-rabbit IgG (Cappel) or monoclonal anti-GFAP (Sigma) and rhodamine-conjugated goat anti-mouse IgG (Cappel). Nuclear Yellow (0.001%; Hoechst) was included with the secondary antibody to visualize nuclei.

Peptide mapping was performed as described by Cleveland et al. (25). Briefly, equal amounts of ceruloplasmin, immunoaffinity-purified from either serum or cultured C6 glioma cells, were subjected to SDS-PAGE on a 10% acrylamide gel. The gel was stained with Coomassie Brilliant Blue, and the 135-kDa protein bands were excised and equilibrated in buffer containing 125 mM Tris and 0.1% SDS. The gel slices were loaded into the wells of a 12% gel, overlaid with the same buffer containing 20% glycerol, and overlaid with this buffer containing 10% glycerol and 0.5 µg of Staphylococcus aureus V8 protease (Boehringer Mannheim). Following electrophoresis, the peptide fragments were visualized by silver staining.

The oxidase activity of GPI-anchored ceruloplasmin was assayed by a modification of the method described by Osaki et al. (26) for plasma ceruloplasmin. C6 glioma cells cultured in serum-free medium were harvested by trituration in Hanks' balanced salt solution, centrifuged, resuspended in DMEM, and incubated with or without PI-PLC as described above. The PI-PLC-treated supernatant was divided into two equal samples, and one of these was incubated with mAb 1A1-conjugated beads to immunodeplete GPI-anchored ceruloplasmin. Supernatants were then concentrated using Centriprep and Centricon devices (Amicon) and subjected to SDS-PAGE without the addition of -mercaptoethanol to the sample buffer or boiling. The gels were then incubated in 0.1 M sodium acetate, pH 5.7, containing 1 mg/ml p-phenylenediamine (Sigma), which upon oxidation yields a purple precipitate.

RESULTS

Immunoaffinity chromatography was used to purify the 1A1 antigen in sufficient quantity for N-terminal amino acid microsequencing. We have previously shown that only astrocytes express this antigen in the central nervous system (18). Because of the difficulty in using astrocyte cultures for large scale purification, we sought an astrocytic cell line that expressed this molecule. C6 glioma cells were found to express the 1A1 antigen, both by immunocytochemistry and by immunoprecipitation after [³⁵S]methionine labeling (18). These cells were therefore used for the purification of the molecule, since they can be grown easily in culture or be injected subcutaneously into nude rats to generate tumors that yield large amounts of material. Nonidet P-40 extracts of a membrane preparation from either C6 glioma tumors or cultured C6 glioma cells were used to purify the 1A1 antigen. Fractions eluted from the mAb 1A1 immunoaffinity column, and analyzed by SDS-PAGE and silver staining, showed a single band, of approximately 135 kDa (Fig. 1A). This molecular mass is similar to that reported previously for the 1A1 antigen immunoprecipitated from rat cortical astrocytes (18). Two-dimensional gel electrophoresis of the immunoaffinity-purified protein revealed the presence of only one polypeptide species (data not shown).

Partial N-terminal amino acid sequence data of the 1A1 molecule immunoaffinity-purified from C6 glioma tumors and electroblotted onto a PVDF membrane yielded a 13-amino acid sequence. Two of the positions gave weak signals (italics). A search of protein data bases showed that this sequence (REKHYYIGITEAV) was identical to that of rat ceruloplasmin, a copper-binding protein of the same molecular weight as the 1A1 molecule (27, 28).

To further confirm that the immunoaffinity-purified protein was ceruloplasmin, the ability of a polyclonal anti-ceruloplasmin antibody to recognize this protein was determined. To this end, the immunoaffinity-purified 1A1 protein was Western blotted with a goat anti-human ceruloplasmin antibody. These Western blots demonstrated that the protein, purified from either C6 glioma tumors or cultured C6 glioma cells (Fig. 1B), is recognized by the anti-ceruloplasmin antibody. These results therefore provide additional evidence that the immunoaffinity-purified 1A1 protein is likely to be identical or highly similar to ceruloplasmin.

Since ceruloplasmin is one of the major protein components in serum (300-450 µg/ml), immunoprecipitations from serum with the mAb 1A1 were done to establish whether this antibody could recognize serum ceruloplasmin. The polyclonal anti-ceruloplasmin antibody could immunoprecipitate a 135-kDa molecule from rat and bovine serum. However, the mAb 1A1 precipitated ceruloplasmin from rat serum but not from bovine serum (data not shown).

Immunofluorescence labeling with the mAb 1A1 showed labeling of ceruloplasmin on the surface of cultured neonatal rat astrocytes (Fig. 2, E and F), but not on cultured rat choroidal cells (Fig. 2, G and H) or hepatocytes (data not shown). To further confirm the cell surface ceruloplasmin labeling, astrocytes and C6 glioma cells were labeled with a polyclonal goat anti-ceruloplasmin antiserum. These experiments also showed that ceruloplasmin was localized to the surface of these cells (Fig. 2, A-D). Since this polyclonal anti-ceruloplasmin antibody could recognize ceruloplasmin present in fetal bovine serum, the astrocytes and C6 glioma cells were grown in serum-free defined medium. Our earlier work has shown that the mAb 1A1 does not label other cells, i.e. neurons, oligodendrocytes, or endothelial cells, in the central nervous system (18).

The immunofluorescence staining indicates that ceruloplasmin is present on the surface of astrocytes and C6 glioma cells. The cell surface localization of ceruloplasmin on astrocytes and C6 glioma cells is not likely to be the result of ceruloplasmin spanning the cell membrane, as it does not have a sufficiently long hydrophobic amino acid stretch that could serve as a membrane-spanning domain. It is therefore likely that the surface localization of ceruloplasmin is due to (i) association of ceruloplasmin with the extracellular matrix, (ii) association with a cell surface receptor, or (iii) a covalent attachment to the cell membrane. Because a sequence near the C-terminal region of ceruloplasmin meets the minimal requirement for a GPI anchor addition (29-31), we examined whether the cell surface localization of ceruloplasmin could be removed by treatment with PI-PLC, which specifically cleaves GPI anchors from GPI-anchored proteins. Astrocytes purified from the neonatal rat cerebral cortex and C6 glioma cells were treated with PI-PLC and stained with mAb 1A1. Incubation with this enzyme eliminated cell surface labeling by mAb 1A1 (Fig. 3), indicating that the localization of ceruloplasmin on the surface of astrocytes and C6 glioma cells is via a GPI-anchor. There are two possibilities regarding the association of ceruloplasmin to the surface of C6 glioma cells by a GPI anchor. (i) Ceruloplasmin might be associated with the cell surface via a receptor which is GPI-anchored to the membrane, or (ii) ceruloplasmin could be directly GPI-anchored to the cell surface. To resolve this issue, PI-PLC-released ceruloplasmin from C6 glioma cells was purified by mAb 1A1 immunoaffinity chromatography and then Western blotted with an antibody that specifically recognizes GPI anchors. This antibody strongly reacted with the ceruloplasmin purified from C6 glioma cells, but did not react with ceruloplasmin purified from rat serum (Fig. 4). Two-dimensional gel electrophoresis of the protein used for Western blotting revealed the presence of only one protein at 135 kDa, indicating that the anti-GPI antibody is not reacting to a protein copurifying with ceruloplasmin (data not shown). These results, therefore, suggest that ceruloplasmin on the surface of astrocytes and C6 glioma cells is directly GPI-anchored to the membrane.

We carried out peptide (Cleveland) mapping to determine if major differences in amino acid sequence exists between the GPI-anchored form and the secreted form of ceruloplasmin. Ceruloplasmin, immunoaffinity-purified (using the mAb 1A1) from either C6 glioma cell membranes or rat serum, was subjected to Staphylococcus V8 protease digestion. This treatment produced similar peptide maps of the GPI-anchored form and the soluble form of the molecule (Fig. 5), indicating that these two forms of ceruloplasmin are highly homologous.

To assess whether the GPI-anchored form of ceruloplasmin has functional oxidase activity like the secreted form, the following experiment was carried out with C6 glioma cells. These cells were used since large numbers of cells were required (i.e. 90 confluent 75-cm² flasks). The cells were harvested and treated with PI-PLC to remove GPI-anchored molecules from the cell surface. The supernatant containing the latter was divided into two. One of these samples was immunodepleted of GPI-anchored ceruloplasmin with the mAb 1A1. The concentrated supernatants were separated by SDS-PAGE under nondenaturing conditions and the gels stained with p-phenylenediamine, which when oxidized produces a purple precipitate. A single 135-kDa band was observed in the PI-PLC-treated sample (Fig. 6). This band was lost in samples that were immunodepleted with the mAb 1A1 (Fig. 6). Nor was this band seen in control samples not treated with PI-PLC. These experiments, therefore, demonstrate that the GPI-anchored form of ceruloplasmin has oxidase activity. Since other bands were not visible, these studies also suggest that ceruloplasmin may be the major GPI-anchored oxidase on these cells.

DISCUSSION

We provide evidence that the 135-kDa cell surface molecule recognized by the mAb 1A1, which is expressed exclusively by astrocytes in the rat central nervous system (18, 32), is a novel GPI-anchored form of ceruloplasmin. The molecule recognized by the mAb 1A1 was purified by immunoaffinity chromatography using detergent-solubilized membrane extracts of C6 glioma tumors or cultured C6 glioma cells that also express this molecule. N-terminal microsequence analysis of the immunoaffinity-purified 135-kDa band indicated that the molecule recognized by the mAb 1A1 is identical (or homologous) to ceruloplasmin, which is classically considered a plasma protein of hepatic origin. Two-dimensional gel electrophoresis revealed the presence of only one 135-kDa polypeptide in the immunoaffinity-purified material, suggesting that mAb 1A1 recognizes only one molecule, namely ceruloplasmin. Additional evidence that the molecule recognized by this monoclonal antibody is ceruloplasmin (or homologous to it) was provided by the following experiments. (i) Western blot analysis demonstrated that the mAb 1A1-purified protein is recognized by a polyclonal anti-ceruloplasmin antibody, (ii) the monoclonal antibody 1A1 immunoprecipitates a 135-kDa protein from rat serum, and (iii) no differences were observed in the peptide fragments generated by Cleveland mapping of ceruloplasmin from serum and that from membrane preparations of C6 glioma cells.

We have also shown previously by metabolic labeling with [³⁵S]methionine that this 135-kDa molecule is synthesized by astrocytes (18). Immunofluorescence labeling of cells in vitro and iodination of cell surface proteins followed by immunoprecipitation showed that this molecule is associated with the plasma membrane (18). In addition, as shown for ceruloplasmin synthesized by liver cells in vitro (28), we have shown that there is only a small reduction in the molecular weight of this molecule when astrocyte cultures are treated with tunicamycin (18), suggesting that it is poorly glycosylated. We have also reported previously that the 1A1 antigen, which we have shown here to be ceruloplasmin, increases in the cerebellum with postnatal development (32). Several earlier studies have reported the presence of ceruloplasmin mRNA in the brain (33-36). More recently, Klomp et al. (37) have reported ceruloplasmin gene expression by astrocytes. We now provide evidence of a novel GPI-anchored form of ceruloplasmin expressed on the surface of astrocytes in the mammalian central nervous system and that it has oxidase activity.

The cell surface localization of ceruloplasmin is unique to astrocytes, since cells of the choroid plexus and hepatocytes, both of which secrete ceruloplasmin, do not show surface labeling with mAb 1A1. Fibroblasts that form the fibroblastic capsule of various organs are the only other cell type to express this molecule on the cell surface (18). This cell surface localization of ceruloplasmin cannot be the result of ceruloplasmin spanning the cell membrane, since it does not have a sufficiently long hydrophobic amino acid stretch that could serve as a membrane-spanning domain (27, 28). Our demonstration that the surface labeling with the monoclonal antibody can be removed by PI-PLC treatment shows that ceruloplasmin is anchored to the cell surface by a GPI anchor. Experiments in which immunoaffinity-purified ceruloplasmin, from PI-PLC cleaved material obtained from C6 glioma cells, was labeled on Western blots with an antibody that specifically recognizes the GPI anchor provide direct evidence that it is itself GPI-anchored to the cell surface. These results therefore provide the first evidence of a GPI-anchored form of ceruloplasmin.

Interestingly, near the C-terminal end, plasma ceruloplasmin contains a potential site for GPI anchor attachment, consisting of small amino acids followed by a short hydrophobic sequence, that appears to satisfy the minimal requirement for a GPI anchor addition signal (29-31). GPI anchor addition would result in a protein with a molecular weight similar to the secreted form since the cleaved C-terminal sequence would be replaced with a GPI-anchor precursor of similar molecular weight (29-31). PIG-A, a gene that encodes a protein required to initiate GPI anchor assembly, is expressed at a much higher level in the brain than in other tissues (38). Thus, astrocytes that are found in the brain may be able to initiate GPI anchor addition more readily than other cell types, such as hepatocytes. Alternatively, the GPI-anchored and secreted forms of ceruloplasmin might be generated through differential splicing, perhaps in a manner similar to the different isoforms of decay-accelerating factor (39). Interestingly, Mollgard et al. (40) have reported that ceruloplasmin expressed by Xenopus oocytes following injection of mRNA from fetal human liver was secreted by the oocytes, whereas injection of mRNA from fetal human brain led to the expression of ceruloplasmin that was retained within the cells. Although these investigators did not localize the ceruloplasmin expressed from brain mRNA to a specific region of the cell, our studies suggest that it is likely to have been the membrane-bound GPI-anchored form of ceruloplasmin. Whether alternative splicing might be responsible for the different isoforms of ceruloplasmin cannot be determined based on the genomic sequence of ceruloplasmin, since it has not yet been fully characterized, and at present, consists only of the exons and intron/exon boundaries of the human liver cDNA (41).

An important finding is the evidence that the GPI-anchored form of ceruloplasmin has oxidase activity. This finding suggests that this unique form of ceruloplasmin on the surface of astrocytes in the central nervous system is likely to play a role in iron metabolism (1, 3, 26) and antioxidant defense (42, 43). Convincing evidence for a role for this form of ceruloplasmin in iron metabolism comes from studies of individuals with hereditary ceruloplasmin deficiency in which there is marked deposition of iron in various organs including the brain (8-15). Since there is no evidence that ceruloplasmin from plasma crosses the blood-brain barrier, and the levels of the secreted form of ceruloplasmin in the cerebrospinal fluid is normally very low (1 µg/ml; Ref. 44), the GPI-anchored form of ceruloplasmin on astrocytes, which comprise 25% of the total volume of the brain (45), is likely to be the major form of ceruloplasmin in the central nervous system. This GPI-anchored form of ceruloplasmin is similar to a recently identified ferroxidase in yeast called Fet3 (4, 5). Fet3, like ceruloplasmin, is a copper-binding ferroxidase. In addition, Fet3 is also attached to the cell surface, but unlike the GPI-anchored form of ceruloplasmin, it has a single transmembrane domain and has a molecular mass of 72 kDa (4, 5). Interestingly, recent studies show that Fet3 ferroxidase activity is required for iron transport into the cell via a newly described iron transporter called Ftr1 (6). Whether the GPI-anchored form of ceruloplasmin in the mammalian central nervous system also functions along with a transporter similar to Ftr1 to transport ferric (Fe(III)) iron from the extracellular compartment into astrocytes is not known at present. Such a mechanism may contribute, along with other non-transferrin uptake systems (1, 7), to the influx of iron into astrocytes, which lack both the transferrin receptor (46, 47) and melanotransferrin (48). It is also possible that the egress of ferrous (Fe(II)) iron from neurons, oligodendrocytes, and astrocytes may occur via some as yet unidentified transporter, such as that proposed for other mammalian cells (3). The Fe(II) iron exiting these cells could become oxidized by the GPI-anchored form of ceruloplasmin on the surface of astrocytes. Since astrocyte processes are distributed throughout the central nervous system, ceruloplasmin located on astrocytes is ideally positioned to effectively oxidize the highly toxic ferrous iron to the ferric form. The latter may then be available for reutilization or cleared from the central nervous system by binding to transferrin or melanotransferrin.

Besides aceruloplasminemia, iron deposition has been observed in the substantia nigra in Parkinson's disease (49, 50), in the cortex and amyloid plaques in Alzheimer's disease (51, 52), in amyotrophic lateral sclerosis (53, 54), and in Hallervorden-Spatz disease (55). Evidence that iron may contribute to the neurodegeneration in these diseases is provided by the 6-hydroxydopamine-lesioned rat model of Parkinson's disease, in which the iron chelator desferroxamine protects neurons from injury (56). Increased levels of free radicals, which may underlie the neurodegeneration, have been reported in the brains of Alzheimer's and Parkinson's patients (57-60). Reduction in ceruloplasmin has been reported in the cortex of patients with Alzheimer's disease (61). It is possible that damage to astrocytes, resulting in reduced levels of the GPI-anchored form of ceruloplasmin in the affected gray matter regions, might lead to the iron deposition and free radical generation that causes the neuronal degeneration seen in these diseases. This is further supported by the finding that the ferroxidase activity of ceruloplasmin has been shown to inhibit ferrous iron-catalyzed phospholipid peroxidation in vitro (42, 43, 62, 63). The GPI-anchored form of ceruloplasmin on astrocytes which has this important ferroxidase activity may regulate iron transport in and out of neurons and glia in the central nervous system, and may help limit lipid peroxidation in a tissue that is highly susceptible to oxidative injury.