IRON METABOLISM

Studies of the Mechanism of Iron Transport Across the Blood-Brain Barrier R. Roberts, MD, PhD,'§ A. Sandra, PhD," G. C . Siek, PhD,? J. J. Lucas, PhDJ and R. E. Fine, PhDt

The mechanism by which iron enters the central nervous system from the blood is not well understood. Iron in blood plasma is totally bound to transferrin (Tf),a major plasma glycoprotein. Tf receptors are present on the blood-brain barrier (BBB) endothelium. It is not known whether iron separates from Tf during its passage across the endothelial cells and then enters the brain by another mechanism, or whether the two proteins enter the brain together. We characterize here the morphological pathway for endocytosis of a monomeric horseradish peroxidase-transferrin conjugate by the rat BBB endothelium. Our results indicate that this conjugate binds to Tf receptors on the luminal BBB, is internalized via clathrin-coated vesicles, enters early or sorting endosomes, and, subsequently, late or recycling endosomes near the Golgi apparatus. No evidence is found for Tf transcytosis. It is likely that iron separates from Tf in early endosomes, which are assumed to be acidic, as they are in other cells, and enters the brain by an as yet undefined pathway. A clonal line of brain capillary endothelial cells that mimics the BBB when grown on permeabilized membranes can transcytose iron provided as Fe55-Tf.This cell line may provide a useful system to determine the pathway that iron uses to enter the brain. We also present evidence that cultured chick embryo forebrain neurons contain a large number of a unique Tf receptor. Roberts R, Sandra A, Siek GC, Lucas JJ, Fine RE. Studies of the mechanism of iron transport across the blood-brain barrier. Ann Neurol 1992;32:S43-S50

The endothelial cells of adult mammal brain form the blood-brain barrier (BBB), which severely restricts passage of water-soluble molecules from blood to brain f l , 2). Because iron is required by all cells to sustain life and in fact some portions of the brain, including the substantia nigra, contain very high iron concentrations, a transport system for iron across the BBB is essential. All plasma iron is normally found complexed to the molecule transferrin (Tf), a major plasma protein C31. The finding that BBB endothelial cells contain a significant concentration of transferrin receptors (TfRs) detected histochemically [4} indicates that the transport of iron into the brain may involve a TfR-mediated pathway. Support for this involvement has come from several biochemical studies C5, 61. We used a horseradish peroxidase-transferrin (HRPTf) conjugate, the data from which indicates for the first time that the Tf endocytic pathway in BBB capillary endothelial cells involves the same compartments previously described in polarized epithelial cells [7, 81. We also present preliminary data regarding a newly described cell culture model of the BBB that provides direct evidence for transcytosis of Fe via a TfR-mediated pathway. Finally, we present data indi-

cating that fetal central nervous system (CNS) neurons, which require Tf and iron for survival 191, possess large numbers of TfRs.

From the "Department of Anatomy, University of Iowa College of Medicine, Iowa City, IA; tGRECC, ENRM VA Hospital, Bedford, MA; and the $Department of Biochemistry and Molecular Biology, SUNY Health Science Center, Syracuse, NY. §Present address: Department of Pathology, Washington University School of Medicine, St Louis, MO.

Address correspondence to Dr Fine, GRECC, Building 18, ENRM VA Hospital, Bedford, MA 01730.

Materials and Methods Preparation of Trans-rrin-Peroxihse Rat Tf (Jackson Labs, West Grove, PA) was iron-loaded and iodinated using the methods described by Yamashiro and colleagues [lo]. Transferrin was conjugated to peroxidase utilizing the heterobifunctional crosslinking agent N-succinimidyl-3-(2-pyridyldithio) propionate [ 11) (SPDP) using the procedure recommended by the manufacturer (Pierce Biochemicals, Rockford, IL). Briefly, 10 mg rat Tf in 1 mL phosphate-buffered saline (PBS) (pH, 7.5) containing a trace amount of IlZ5-Tfwas reacted with 20 pg SPDP at room temperature for 30 minutes. The reaction products were passed over a PD-10 column, and the protein peak containing Tf-PDP was collected. In a separate tube, 20 pg SPDP was added to 5 mg HRP in 1 mL PBS and allowed to react for 30 minutes at room temperature. PDP-HRP was isolated after passage of the products over a PD-10 column. To introduce free sulfhydryl (SH) groups in Tf-PDP, the solution was adjusted to pH 6.0 and dithiotreitol was added to a final coiicentration of 10 mmol/L. After 30 minutes at room temperature the products, including Tf-SH, were passed over a PD-10 column and the protein peak recovered.

s43

The PDP-HRP was then added to the Tf-SH solution and the p H adjusted to 7.5. The solution volume was adjusted to 0.5 mL using a Centricon PMlO concentrator, and the reaction was allowed to continue for 12 hours at room temperature. The reaction products were analyzed by nonreduced SDS-PAGE and autoradiography. Tf-HRP was separated from unreacted PDP-HRP and Tf-SH by passage over a Sephacryl 200 column. The biological activity of Tf-HRP was assessed by competition assays.

Pe&sion Protocol Twenty-one-day-old Sprague-Dawley rats were killed with ether and the vascular system was cleared of blood elements by perfusion through the left ventricle with oxygenated M199 (with 20 mmol/L Hepes) containing 2 mg/mL bovine serum albumin. The animal was then cooled to 4°C by placing it in an ice bath and perfusing with cold M199 for approximately 5 minutes. The ascending aorta was then identified and cannulated, and all subsequent perfusion occurred through this vessel at a flow rate of 5 mL./min. The animal was then perfused with cold buffer containing Tf-HRP (200 Fg/mL) for 15 minutes to allow for receptor-ligand binding. The animal was then placed in a 37°C bath and perfused for various lengths of time with a buffer warmed to 37°C. The animal was then perfused with fixative containing 1% glutaraldehyde and 1% paraformaldehyde in 0.1 mol/L phosphate buffer. Nonspecific (nontransferrin receptor-mediated) endocytosis of Tf-HRP was determined in control experiments in which native rat ferrotransferrin (5 mg/mL) (Sigma Chemical, St Louis, MO) was added to the perfusate to compete for TfR binding.

Tissue Processing for Electron Microscopy After fixation, brains were removed and postfixed in the same fixative for 1 hour. Coronal sections, 50-pm thick, were cut with a vibratome, and included thalamus, hypothalamus, internal capsule, and cerebral cortex. The peroxidase reaction was initiated by incubation of the sections in 50 mmol/L PO4 buffer containing 0.01% H,Oz and 0.05% diaminobenzidine. After 5 to 10 minutes, the sections were washed in PO, buffer and processed for electron microscopy. Brain sections were flat embedded in Epon between opposed sheets of aclar. Polymerized brain sections were remounted on polymerized blocks of Epon with superglue and sectioned for electron microscopy. Ultrathin sections were stained with lead citrate and photographed on a Hitachi H-7000 electron microscope operating at 50 KU.

mounted on small sections of cork with optimum cutting temperature compound and plunged into propane slurry cooled by liquid nitrogen. Ten-micrometer-thick frozen sections were cut, picked up on coverslip, and subjected to indirect immunoperoxidase using MRC OX-26 antitransferrin receptor monoclonal antibody (Serotec, Oxford, UK) {41 and a mouse Vectastain kit (Vector Labs, Burlingame, CA) using procedures recommended by the manufacturers. Control experiments were performed using antiinsulin monoclonal antibody in place of MRC OX-26. After incubation in all reagents and washes, the sections were fixed for 1 hour in 1% glutaraldehyde in 0.1 mol/L PO4 buffer and then washed 3 times in 0.1 mol/L PO4 buffer for 1 hour. The peroxidase reaction was initiated by incubation of the sections in 50 mmol/L PO4 buffer containing 0.01% H,02 and 0.05% diaminobenzidine. After 5 to 10 minutes the sections were washed in PO4 buffer and processed for electron microscopy as described. The coverslips were mounted on Epon-filled beem capsules and polymerized overnight. The coverslips were separated from the tissue sections by immersion in liquid nitrogen. The sections were then cut and photographed as described.

Cell Culture of Bwine Brain Capillary Endotbelial Cells Clonal bovine brain capillary endothelial cells (BCE) were kindly provided by the laboratory of Dr L. L. Rubin. Cells were maintained and plated on tissue culture filter inserts as previously described E 141. Briefly, the cells were maintained on collagen, fibronectin-coated flasks in mirnimal essential medium, 10% plasma-derived horse serum 1: 1 with astrocyte-conditioned medium (prepared from primary cultures of rat embryonic astrocytes cultured in DMEM, 10% fetal calf serum; medium collected every 48 hours), supplemented with 1.25 ng/mL basic fibroblast growth factor. For transcytosis experiments, 250,000 BCE cells were plated on collagen, fibronectin-treated 24-mm Costar Transwell filters (0.4-pm pore size; Costar Corp, Cambridge, MA), which fit into 6-well tissue culture plates. The filters were maintained in astrocyte conditioned medium 1 : 1 with serum-freedefined N2 medium El51 for 15 days until the cells were fully confluent on the filter. The cells were then treated for 4 additional days with 250 pmol/L 8-4-(chlorophenyl-thio)CAMP (CPT-CAMP; Boehringer-Mannheim, Indianapolis, IN) and 17.5 pmol/L RO20-1724, a phosphodiesterase inhibitor (BioMol, Beaverton, OR).

Fe Transcytosis Experiments Ultrastructural lmmunolocalization of Transfewin Receptors Transferrin receptors were localized using the procedure described by Levine and Woods (121. Briefly, rat brains were fixed by vascular perfusion with periodate-lysine-paraformaldehyde (2% paraformaldehyde, 0.075 mol/L lysine, 0.01 mol/L Na104, 0.0375 mol/L NaP04 (pH, 6.2) (131. The brains were removed and fixed for an additional 2 hours in the same fixative. Coronal sections 2 to 3 mm thick were cut, and sections of cerebral cortex and thalamus were cut into 2- to 3-mm squares, which were washed in PBS containing 10% dimethylsulfoxide for 1 hour. The tissue was S44 Annals of Neurology

Fe55(as ferric chloride; Amersham International) was bound to bovine iron-free Tf (Sigma) as previously described { 161. Additional bovine iron-free transferrin was saturated with unlabeled iron (Fe-Tf) using a described method 117). The transcytosis of FeS5was determined by measuring the transport of label across the cell layer and filter in the presence and absence of an excess of unlabeled Fe-Tf. C'*-labeled inulin was included as a measure of BCE barrier integrity. Both the upper and lower chamber of the Transwell filter were washed gently 3 times with 1: 1 DMEM/F12 with 20 mmol/L Hepes to remove any free Tf. Fes5-bound Tf (1.5 pg/mL, approximately 1.9 x mol/L, which is a five-fold

Supplement to Volume 32, 1992

excess of the published K, value for the TfR) was added to the upper chambers in DMEMIF12 1 :1 with either 2.5 mg/ mL purified human serum albumin or 2.5 mg/mL unlabeled Fe-Tf. Aliquots (50 p,L) were taken from the lower chamber beginning 10 minutes after the addition of the label and at 10-minute intervals, up to 60 minutes. The aliquots were added to 6-mL scintillation vials with 5 mL Aquasol and counted for 10 minutes on a Beckman scintillation counter precalibrated for double-label FeS5and C l4 counting.

Characterization of Chick Fetal Neuron TfRs Chicken neuronal cultures were prepared exactly as previously described 191. SDS-polyacrylamide gel electrophoresis (SDS-PAGE) was performed using the method of Laemmli [18]; Western blotting was performed as previously described { 191. Immunoprecipitation was carried out as previously described 1191. Briefly, 4-day cultured neurons in a 100-mm culture dish were washed once in Hank's balanced salts and placed in 10 mL 1: 1 DMEMIFl2 (Sigma) containing onetenth the normal concentration of methionine; 100 uCi S35methionine (ICN Chemicals, Costa Mesa, CA) was added to the medium. After 24 hours, the medium was removed and the cells were scraped from the dish, pelleted at 1,000 g for 5 minutes, resuspended in isolation buffer (0.1% SDS/ 1% Triton X-100 in PBS), and homogenized using a Dounce with a tight pestle. Insoluble material was removed by centrifugation at 20,000 g for 20 minutes. The resulting supernatant was divided into 3 equal aliquots; 3 S anti-chick oviduct TfR was added to one, 3 uL anti-kinesin heavy chain 1191 to another, 3 S preimmuoe rabbit serum to the third. After incubation for 2 hours at 37°C in a shaking water bath, 25 KL protein A-agarose beads ('Sigma) were added, and the incubations continued for an additional 2 hours. The beads were then collected by sedimentation at 15,000 g for 10 seconds, washed 3 times in isolation buffer, and 3 times in isolation buffer minus detergents. The beads were then placed in Laemmli sample buffer, incubated for 10 minutes at 37"C, and the beads removed by sedimentation for 10 seconds at 15,000 g. The resulting supernatants were boiled for 1 minute and subjected to SDS-PAGE. The gels were impregnated with 1 rnollL salicylic acid, dried, and autoradiography was performed.

Photomicrographs of the stained cultures were taken through a Nikon Optiphot microscope at x200 magnification.

Results The properties of the HRP-Tf conjugate will be described elsewhere (manuscript in preparation). It behaved similarly to native transferrin by a number of criteria, as has been previously shown {21). When we employed 100 F g / d HRP-Tf in the rat perfusion model described, we found that after perfusion for 15 minutes at 4"C, the reaction product was entirely localized at the luminal surface of the BBB endothelium in both clathrin-coated and noncoated regions (data not shown). Control perfusions in which excess native rat Fe-Tf was included revealed a great diminution of the labeled luminal membranes (data not shown). After warm-ups of between 2 and 4 minutes, followed by fixation, DAB cytochemistry, and electron microscopy, the reaction product was found primarily in coated pits and vesicles at or near the luminal surface of the BBB (Fig 1). The regular hexagonal clathrin coating can be visualized at high magnification (data not shown). Little or no labeling of noncoated regions of the luminal plasma membrane or of microvilli was seen, indicating that Tf-HRP complexes associated with TfRs congregate in ciathrin-coated pits shortly

immanobistochemistry Nine-day-old chick embryo forebrain neurons were prepared as previously described 191. The neurons were plated on poly-D-lysine-coated, 4-chamber lab-tek culture slides and cultured for 7 days in DMEMlF12, 1: 1. The cells were fixed in 4% paraformaldehyde in PBS with Ca2+for 25 minutes, then briefly washed with cold absolute methanol. The cultures were incubated for 45 minutes in blocking solution (5% Carnation nonfat dry milk in PBS). The cultures were then incubated with a 1:600 dilution of rabbit antichick oviduct transferrin receptor antibody 1201 for 1 hour, followed by 3- to 10-minute washes with PBS. The cultures were then incubated with a 1 :400 dilution of peroxidase-labeled goat antirabbit immunoglobulin G (IgG) with a 1:500 dilution of normal goat serum for 1 hour, followed by 3- to 10-minute washes with PBS. Labeled cells were identified by the diaminobenzidine-peroxidasestaining.

Fig I . Transferrin 2s found in coated vesicles early after warm-up. A 21-day-old rat was killed and perfnsed with a solution containing 100 pglmL horseradish peroxidasetransfewin (HRP-Tf) for IS minutes at 4°C.The perfusate was then continued for an additional 3 minutes at 37"C, followed by fixation, sectioning, peroxidase reaction, and electron microscopy (EM). The arrowhead points to a typical 100-nm coated vesicle labeled with diaminobenzidine near the luminal surface of a blood-brain barrier endothelial ceL. The bar in this and subsequent EMS indicates 100 nm.

Roberts et al: Iron Transport Across the BBB S45

~~

~

~

~~

Fig 3. Transfewin is contained in 50- t o 100-nm vesicles and

tubules surrounding the Golgi region 15 minutes aftevwarmup. Perfusion, warm-up, and subsequent procedures were as described in Figure I . This jigure shows the labeling of many small (50-100 nm) vesicles and tubules at later times following warm-up. Many of these labeled structures surround the Golgi apparatus.

Fig 2. Transfewin is contained in Large (250-500 nm) uncoated vesicles and tubules, 5 to 8 minutes after warm-up. Perfusion, warm-up, and subsequent procedures were as described in Figure 1. (a) Low magnification indicates the labeling of 250 to 500 nrn uncoated structures with the appearance of ear& or sorting endosomes. (b) Higher magnification of a characteristic Lzbeled structure. Arrow points to this structure.

after warm-up and are rapidly internalized in intact coated vesicles. At longer intervals following warm-up (5-8 min), the reaction product was now seen mainly in larger (100-200 nm) noncoated tubular and vesicular structures with the morphology of early and sorting microsomes visualized in other cell types t22, 231 (Fig 2). At warm-up between 8 and 15 minutes, a further change in the DAB-positive structures was noted. The predominantly labeled species was 50- to 100-nm vesicles and tubules in regions near the Golgi apparatus (Fig 3). In other cell types in which these structures have been visualized, they are called recycling endo-

somes, indicating their role in recycling Tf-TfR complexes back to the plasma membrane 1231. Labeled vesicles were never observed near the ablumind plasma membranes or in subendothelial locations, providing no evidence for transcytosis of HRPTf under the experimental conditions we employed. We also localized TfRs in BBB endothelial cells using the MRC OX-26 antirat TfR antibody employed by Jefferies and colleagues 141 in their studies. This antibody recognizes an extracellular domain of the TfR. After indirect immunoperoxidase cytochemistry on sectioned rat brain, we found DAB reaction product solely on the luminal plasma membrane of the BBB endothelium. The most intense label is over clathrincoated pits (Fig 4a). No label is seen over the nonluminal plasma membrane. We also used a second antibody conjugated to 10-nm colloidal gold. This protocol also resulted in visualization of gold particles on the exterior aspect of luminal clathrin-coated pits (Fig 4b). Again, no label was visualized over the nonluminal plasma membrane. The experiments described herein suggest that in BBB endothelium, endocytosed Fe+3 Tf rapidly reaches sorting endosomes, which by analogy with other cells 124, 251 are likely to be acidic. The lowered p H causes iron to separate from Tf. Therefore, iron transcytosis from these structures is likely to involve as yet undiscovered pathways not necessarily involving Tf. We have recently used a newly developed cell culture system that reproduces many properties of the

S46 Annals of Neurology Supplement to Volume 32, 1992

100

0-0 A-A

Fe55-Tf Fe55-Tf

+ cold Fe-Tf

60

0

20

0 0

10

20

30

40

50

60

Time (mins)

Fig 5 . Transcytosis of Fd’ across an in vitro model of the blood-brain bawier using cultured bovine brain capilavy endothelial cells. Procedures are described in the Materials and Methods section. Data presented as cpm of Fd’I50 uL sample from the lower chamber. Fd5 introduced to the upper chamber as Fd’-transfirrin (Fe55-Tf)in the presence and absence of an excess of unlabeled Fe-Tf (cold Fe-Tf).

Fig 4. Transfewin receptors are localized to the luminal plasma membrane of the blood-brain barrier, including coated pits. Brains of killed rats were jxed, and frozen sections were prepared as described in Materials and Methods. Ten-micrometer-thick frozen sections were stained with a monoclonal antibody against the rat transfewin receptor. (a) Results obtained when a horseradish peroxidase-labeled antimouse immunoglobulin G (IgG)second antibody was employed. Arrows point to labeled luminal coated pits. Other areas of the luminal membrane are also labeled. (6) Results obtained when an antimouse IgG second antibody coupled to 10-nm colloidal gold particles is employed. The arrow points to a luminal coated pit labeled by swera1 gold particles.

BBB 1141. Using cloned bovine blood-brain cells cultured on porous filters, which maintain high resistance junctions and greatly retard inulin from reaching the abluminal or brain side of the cells, we added bovine FeSSTfin the presence and absence of a one hundredfold excess of unlabeled Fe-Tf to the luminal medium and measured the rate of passage of FeS5to the abluminal or brain side of the cells. Figure 5 shows the results obtained. We can indeed demonstrate the receptormediated passage of Fe55 from luminal to abluminal side. When we employed I’25-Tf-Feusing the identical protocol we saw no evidence of Tf transcytosis (data not shown). One of us (J.J.L.) recently purified an estrogeninducible integral membrane glycoprotein from chick oviduct which has the ability to bind Tf-Fe+3 1201. Antibodies to this protein do not crossreact with antibodies against the embryonic red blood cell (RBC)TfR 1261. This TfR seems to be localized to tissues which in mammals contain large quantities of TfRs, including the liver and the BBB capillary endothelium 1271. We employed pure cultures of 9-day-old chick embryo forebrain neurons, which can survive, and extended neurites in a chemically defined medium containing only Tf and insulin 191. The Tf requirement can be replaced by the use of iron chelators (data not shown). We performed Western blotting on these neurons using both a polyclonal antibody against the oviduct TfR and a monoclonal antibody against the RBC TfR. Figure 6a demonstrates that the oviduct TfR antibody recognizes a protein with a molecular weight of 90 kd, corresponding to the molecular weight of the oviduct TfR, whereas very little reaction is seen with the RBC TfR antibody. Figure 6b demonstrates that 24-hour Roberts et al: Iron Transport Across the BBB S47

S35-methionine-labeled neuronal cultures synthesize a large quantity of a 90-kd protein, which can be specifically immunoprecipitated with the oviduct TfR antibody. Finally, these cultured neurons were heavily stained with the oviduct TfR antibody (Fig 7). Discussion The morphological data presented herein are consistent with studies in other cell types with respect to the pathway for Fe-Tf uptake by BBB endothelid cells. Fe-Tf is bound specifically to luminal TfRs and is concentrated in clathrin-coated pits, which bud from the plasma membrane as coated vesicles. The vesicles then lose their coats and fuse to form early or sorting endosomes. As the endosomes acidify, iron and the Tf-TfR separate, and the Tf-TfR complexes enter tubular and vesicular vesicles surrounding the Golgi apparatus and are then recycled back to luminal plasma membrane. The fate of iron cannot be determined from these morphological studies but will be considered. No evidence of HRP-Tf-positive vesicles near the nonluminal plasma membrane were visualized and no Tf-R-positive coated pits were seen on the nonluminal plasma membrane. These results therefore contradict earlier results based on biochemical studies, which reported the transcytosis of Tf across the BBB [ S , 6). There are several possible explanations for this contradiction. First, biochemical studies are much more sensitive. For example, in the study by Fishman and associates [ 5 ] , only 1 in 10,000 Tf molecules in the perfusate reached the brain. It is therefore quite likely that the morphological techniques employed herein would not detect this small level of Tf transcytosis. It is also possible that if we were technically able to lengthen the perfusion times, we would have detected transcytosis. In this regard, the kinetics of Tf transcytosis detected biochemically [ 5 ) and of the transcytosis of wheat germ agglutinin-HRP reported by Broadwell and colleagues C28) were only detectable after longer intervals. Whatever the reasons for our failure to detect Tf transcytosis, the critical point is that our morphological results strongly suggest that, as demonstrated in ocher cells, Fe and Tf separate in early endosomes. Therefore, whether a small amount of internalized Tf is transcytosed into the brain is not irrelevant to the fate of the internalized Fe. Fig 6. A transferrin receptor (TfR) isolated from chicken oviduct is also found on cultured fetal chicken forebrain neurons. Nine-day-old chicken embryo forebrain neurons were cultured for four days as described previously {9}. (a) Western blot o f two I 00-pg homogenates of chicken forebrain neurons with a I :I00 dilution of a monoclonal antibody against the chicken embryonic red blood cell TfR (lane 2 ) or with a I :I000 dilution of a rabbit polyclonal anti-oviduct T f R (lane 3). Lanes 1 and 4 contain I0 pg embryonic chick red blood cell membranes. Lane 5

S48 Annals of Neurology

contains 10 pg chick oviduct membrane. Lanes I and 2 are with a {S2i}-antimouse immunoglobulin G (IgG) (Sigma) and lanes 3, 4, and 5 with a antirabbit IgG (Sigma). The bar shows the migration of a 97-kd protein standard. (6) Autoradiogram ofS3’ methionine-labeled chick neuronal cultures immunoprecipitated with: (lane I ) preimmune rabbit serum, (lane 2) anti-chick oviduct TfR, and (kane 3) anti-bovine kinesin, as described in the Materials and Methods section. The bar shows the migration of a 97-kd protein standard.

Supplement to Volume 32, 1992

~~

Fig 7. Immunohistochemisty of chick oviduct transferrin receptor in cultures of forebrain neurons from 9-day-old chick embyos. Cultures were stained using peroxidase-labeling as &scribed in the Materials and Methods section. (a) Control staining with nonimmune rabbit serum ( x 300, before 32% reduction). (b) Neurons stained with a 1.600 dilution of rabbit anti-chick oviduct transferrin receptor ( x 300, before 32% reduction).

There are several possible pathways that Fe could follow after separating from Tf. One possibility is that iron leaves the endosome via an unknown route, albeit a necessary one, probably involving a selective pore or transporter, and enters the cytoplasm. The iron is then transported via a cytoplasmic binding protein to the abluminal membrane, is in some manner released to the extracellular space, and is bound to extracellular Tf to be delivered to various brain cells. Our preliminary results using a cell culture model of the BBB reported herein and those of an in vivo study 161 would be consistent with this model. An analogous transferrin-iron transport system may be the trophoblastic cells of the placenta, which form the blood-fetal barrier. These cells express a large number of TfRs on the blood side of the cells E291. After internalization, the iron is released in acidic endosomes and the apotransferrin (apo-Tf) is released on the

blood side. The BeWo cell, a transformed trophoblastic cell line, has been used to demonstrate that after release from Tf, iron reaches the cytosol, complexes with ferritin, and subsequently is released into the extracellular medium E30l. The rate of iron release is accelerated two-fold by the addition of apo-Tf to the outside medium, and the released iron is in a low molecular weight form capable of binding to apo-Tf. An alternate pathway for iron transcytosis can be postulated as well, based on the well-described mechanism for iron transcytosis across the Sertoli cells, an epithelium which comprises the blood-testes barrier. In this cell type, iron separates from Tf in basolateral endosomes, is transferred to the trans-Golgi network in which it binds to apo-Tf, which is a major secretory product of the Sertoli cell. The Fe-Tf is then secreted into the testicular medium C31, 327. We believe that this is an unlikely mechanism for the BBB endothelial cells because there is no evidence that these cells synthesize Tf. This mechanism may, however, be operable in the epithelial cells of the choroid plexus, which forms the blood-cerebrospinal fluid barrier. These cells do synthesize and secrete Tf and contain blood-facing TfRs E33, 341. Our results also indicate that developing neurons from immature brain require Fe and neurite production for survival and contain a large number of TfRs presumably to internalize and provide iron for enzyme systems that require this metabolism. It is likely that at least one group of Fe-containing enzymes, which will be extremely active during neurite outgrowth, are fatty acid desaturases 135, 361. These enzymes are required in the production of phospholipids containing unsaturated fatty acids needed to produce the huge amount of plasma membrane for neurite outgrowth. We surmise that after neurite extension and synaptogenesis are completed, neurons would require less Fe and contain many fewer TfRs as well. In fact, the TfRs in adult chicken brain are found almost entirely in the BBB endothelium 1271. The recent development of cell culture systems that mimic the properties of the BBB 114, 377 offers the prospect of better understanding the mechanism of iron transport into the brain. The preliminary evidence presented herein that suggests there is Tf-TfRmediated iron transcytosis may allow us to understand better the regulation of iron in the brain in normal and pathological states. This work was supported by grants from the Juvenile Diabetes Foundation and the United States Public Health Service, AM 25295 (A.S.), and by a Merit Review Grant from the VA (R.E.F.). We thank Dr Ann Mason for providing the anti-chick red blood cell antibody and Dr Lee Rubin for the bovine brain endothelial cell line. We also acknowledge the expert editorial assistance of Simona Zacarian.

Roberts et

al:

Iron Transport Across the BBB

S49

References 1. Bradbury MWB. The blood-brain barrier: transport across the cerebral endothetium. Circ Res 1985;57:213-222 2. Broadwell RD. Transcytosis of macromolecules through the blood-brain barrier: a cell biological perspective and critical appraisal. Acta Neuropathol 1989;79: 117-128 3. Aisen P, Listowsky I. Iron transport and storage proteins. Annu Rev Biochem 1980;49:357-393 4. Jefferies WA, Brandon MR, Hunt SV, et al. Transferrin receptor on endothelium of brain capillaries. Nature 1984;312: 162-163 5. Fishman JB, Rubin JB, Handrahan JV, et al. Receptor-mediated transcytosis of transferrin across the blood-brain barrier. J Neurosci Res 1987;18:299-304 6. Banks WA, Kastin AJ, Fasold MB, et al. Studies of the slow bidirectional transport of iron and transferrin across the blood brain barrier. Brain Res Bull 1988;21:881-885 7. Fuller SD, Simons K. Transferrin receptor polarity and recycling accuracy in “tight” and “leaky” strains of Madin-Darby canine kidney cells. J Cell Biol 1986;103:1767-1779 8. Hughson EJ, Hopkins CR. Endocytic pathways in polarized Caco-2 cells: identification of an endosomal compartment accessible from both apical and basolateral surfaces. J Cell Biol 1990;110:337-348 9. Aizenman Y, Weichsel M Jr, de Vellis J. Changes in insulin and transferrin requirements of pure brain neuronal cultures during embryonic development. Proc Natl Acad Sci USA 1986;83: 2263-2266 10. Yamashiro DJ, Tycko B, Fluss SR, et al. Segregation of transferrin to a mildly acidic (pH 6.5) para-GoIg compartment in the recycling pathway. Cell 1984;37:789-800 11. CarlssonJ, Drevin H , Aken R. Protein thiolation and reversible protein-protein conjugation. Biochem J 1978;173:723-737 12. Levine DS, Woods JW. Immunolocalization of transferrin and transferrin receptors in mouse small intestine absorptive cells. J Histochem Cytochem 1990;38:851-858 13. McLean JW, Nakane PK. Periodate-lysine-paraformaldehyde fixative, a new fixative for electron microscopy. J Histochem Cytochem 1974;22:1077 14. Rubin LL, Hall DE, Porter S, et al. A cell culture model of the blood-brain barrier. J Cell Biol 1991;115:1725-1735 15. Bottenstein JE, Sat0 GH. Growth of a rat neuroblastoma cell line in serum-free supplemented medium. Proc Natl Acad Sci USA 1979;76:514-517 16. Bates GW, Schlabach MR. The reaction of ferric salts with transferrin. J Biol Chem 1973;248:3228-3232 17. LarrickJW, Cresswell P. Transferrin receptors on human B and T lymphoblastoid cell lines. Biochim Biophys Acta 1979;583: 483-490 18. Laemmli UK. Cleavage of structural proteins during the assembly of the head of bacteriophage @. Nature 1970;227:680-685 19. Morin PJ, Liu N, Johnson RJ, et al. Isolation and characterization of rapid transport vesicle subtypes from rabbit optic nerve. J Neurochem 1991;56:4 15-42 7 20. Poola I, Lucas JJ. Purification and characterizarion of an estro-

S50

Annals of Neurology

gen-inducible membrane glycoprotein. J Biol Chem 1988;263: 19137-19146 21. Stoorvogel W, Geuze HJ, Griffith JM, et al. The pathways of endocytosed transferrin and secretory protein are connected in the trans-Gobi reticulum. J Cell Biol 1988;106:1821-1829 22. Geuze HJ, SlotJW, Strous GJ, et al. Intracellular receptor sorting during endocytosis: comparative immunoelectron microscopy of multiple receptors in rat liver. Cell 1984;37:195-204 23. Yamashiro DJ, Maxfield FR. Acidification of morphologically distinct endosomes in mutant and wild-type Chinese hamster ovary cells. J Cell Biol 1987;105:2723-2733 24. Dautry-Varsat A, Clechanover A, Lodish HF. p H and the recycling of transferrin during receptor-mediated endocytosis. Proc Natl Acad Sci USA 1983;80:2258-2262 25. Klausner RD, Ashwell G, Van Renswoude J, et al. Binding of apotransferrin to K562 cells: explanation of the transferrin cycle. Proc Natl Acad Sci USA 1983;80:22633-22636 26. Poola I, Mason AB, Lucas JJ. The chicken oviduct and embryonic red blood cell transferrin receptors are distinct molecules. Biochem Biophys Res Comm 1990;171:26-32 27. Fuernkranz HA, SchwabJE, Lucas JJ. Differential tissue localization of oviduct and erythroid transferrin receptors. Proc Natl Acad Sci USA 1991;88:7505-7508 28. Broadwell RD, Batin BJ, Saleman M. Transcytotic pathway for blood borne protein through the blood-brain barrier. Proc Natl Acad Sci USA 1988;85:632-636 29. Parmley RT, Barton JC, Conrad ME. Ultrastructural localization of transferrin, transferrin receptor, and iron binding sites on human placental and duodenal microvilli. Br J Haematol 1985; 6 0 8 1-89 30. Van Der Ende A, du Maine A, Simmons CF, et al. Iron metabolism in Bewo chorion carcinoma cells: transferrin-mediated uptake and release of iron. J Biol Chem 1987;2622:8910-8916 31. Djakiew D, Hadley MA, Byers SW, et al. Transferrin-mediated transcellular transport of 59Feacross confluent epithelial sheets of Sertoli cells grown in bicameral cell culture chambers. J An-. drol 1986;7:355-366 32. Morales C, Clermont Y. Receptor-mediated endocytosis or transferrin by Sertoli cells of the rat. Biol Reprod 1986;35: 393-405 33. Bloch B, Popovici T, Chouham S , et al. Transferrin gene expression in choroid plexus of the adult rat brain. Brain Res Bull 1987;18:573-576 34. Tsutsumi M, Skinner MK, Sanders-Bush E. Transferrin gene expression and synthesis by cultured choroid plexus. J Biol Chem 1989;264:9626-9631 35. Larkin EC, Rao GA. Importance of fetal and neonatal iron: adequacy for normal development of central nervous system. In: Dobbing J, ed. Brain behavior and iron in the infant diet. New York: Springer-Verlag, 1989:43-57 36. Rao GA, Larkin EC. Reduced myelination in rat pups due to parental iron deprivation. Biochem Arch 1989;5:91-99 37. Dehouck MP, Meresse S, Delorme P, et al. An easier reproducible and mass-production method to study the blood-brain barrier in vitro. J Neurochem 1990;54:1798-1801

Supplement to Volume 32, 1992