ORIGINAL ARTICLE

A novel model for brain iron uptake: introducing the conceptof regulationIan A Simpson1, Padmavathi Ponnuru2, Marianne E Klinger1, Roland L Myers1, Kavi Devraj1, Christopher L Coe3, Gabriele R Lubach3,Anthony Carruthers4 and James R Connor1,2

Neurologic disorders such as Alzheimer’s, Parkinson’s disease, and Restless Legs Syndrome involve a loss of brain iron homeostasis.Moreover, iron deficiency is the most prevalent nutritional concern worldwide with many associated cognitive and neuralramifications. Therefore, understanding the mechanisms by which iron enters the brain and how those processes are regulatedaddresses significant global health issues. The existing paradigm assumes that the endothelial cells (ECs) forming the blood–brainbarrier (BBB) serve as a simple conduit for transport of transferrin-bound iron. This concept is a significant oversimplification, atminimum failing to account for the iron needs of the ECs. Using an in vivo model of brain iron deficiency, the Belgrade rat, we showthe distribution of transferrin receptors in brain microvasculature is altered in luminal, intracellular, and abluminal membranesdependent on brain iron status. We used a cell culture model of the BBB to show the presence of factors that influence iron releasein non-human primate cerebrospinal fluid and conditioned media from astrocytes; specifically apo-transferrin and hepcidin werefound to increase and decrease iron release, respectively. These data have been integrated into an interactive model where BBB ECsare central in the regulation of cerebral iron metabolism.

Journal of Cerebral Blood Flow & Metabolism advance online publication, 15 October 2014; doi:10.1038/jcbfm.2014.168

Keywords: blood–brain barrier; mathematical modeling; neurochemistry; neurovascular unit; neurovascular coupling; receptors

INTRODUCTIONA number of neurologic disorders are associated with either toomuch or too little iron in the brain. Excessive brain ironaccumulation is reported in Parkinson’s and Alzheimer’s diseases,amyotrophic lateral sclerosis, neurodegeneration with brain ironaccumulation, and Huntington’s disease.1 Conversely, in a 2010WHO report, iron deficiency was identified as the most prevalentmicronutrient problem. Iron deficiency is associated with sig-nificant cognitive, performance, and brain structural deficits.2,3

Although brain iron concentrations remain fairly constant, recentstudies by Unger et al4 suggest that maintenance of brain ironhomeostasis is a dynamic process that involves both diurnal andregional regulation. Thus, elucidation of the mechanism(s) bywhich the brain acquires iron and regulates brain iron transport,can provide insights into the adaptive responses involved inmaintaining brain iron homeostasis and those associated withneurobiologic disease.A widely held theory on how the brain acquires iron, the

‘transcytosis mechanism’ (Figure 1, Route 1 (red) proposes thatcirculating transferrin-bound iron (holo-transferrin, Tf) binds totransferrin receptors (TfR) on the luminal membrane of theendothelial cells (ECs) that comprise the blood–brain barrier (BBB).The Tf–TfR complexes undergo endocytosis and traverse the ECs,where they fuse with the abluminal membrane and the Tf isreleased into the brain extracellular space. This perspective, which

views the ECs of the BBB as a simple conduit, has several flaws. Itfails to provide a mechanism for regulating brain iron uptake or toaccount for the substantial iron requirements of the ECs. Thismodel also fails to account for several important aspects ofcerebral iron transport: (1) the observed rate of iron transport issignificantly greater than transferrin import, implying a segrega-tion of iron and Tf within the ECs, (2) there is no mechanism toinduce release of the tightly bound Tf at the abluminal membrane,(3) there is no mechanism to return the receptor to the luminalmembrane,5 (4) there is no mechanism for recharging apo-transferrin after the delivery of iron to neural cells,6 and (5) there isno mechanism for regulating the expression of Tf receptors inresponse to intracellular iron.In the current study, we have considered the mechanism by

which iron is transported across the enterocyte in the gut as apotential model for how iron is transported across the BBB. A keyelement in the regulation of the passage of iron across theenterocytes is the presence of TfRs on the blood-side mem-branes.7 These receptors transfer Tf-bound serum iron into theenterocyte to inform the enterocyte of the body’s iron status andthe release of iron is further regulated by hepcidin that is releasedby the liver when circulating iron levels are high. In the currentstudy, we have investigated whether this logic also applies to thebrain’s ECs. We investigated the distribution of Tf receptors in theEC in an in vivo model with two goals; first to determine whether

1Department of Neural and Behavioral Sciences, Penn State Hershey Medical Center, Hershey, Pennsylvania, USA; 2Department of Neurosurgery, Penn State Hershey MedicalCenter, Hershey, Pennsylvania, USA; 3Harlow Center for Biological Psychology, University of Wisconsin, Madison, Wisconsin, USA and 4Department of Biochemistry and MolecularPharmacology, University of Massachusetts Medical School, Worcester, Massachusetts, USA. Correspondence: Dr JR Connor, Department of Neurosurgery (H110), Penn StateHershey Medical Center, 500 University Dr, Hershey, PA 17033, USA.E-mail: jconnor@psu.eduThe authors are grateful to Xuesheng Zhang for providing the astrocytes cultures used in this study. Kari Duck made valuable contributions to data interpretation and discussion.This work was supported by the following grants from the National Institutes of Health: 1R01NS077678-01 (JC and IS) and P01 HD39386, Brain and Behavior in Early IronDeficiency (Betsy Lozoff, Principal Investigator, JC, CC, and GL co-investigators).Received 13 March 2014; revised 3 September 2014; accepted 4 September 2014

Journal of Cerebral Blood Flow & Metabolism (2014), 1–10© 2014 ISCBFM All rights reserved 0271-678X/14 $32.00

www.jcbfm.com

Tf receptors are present on the abluminal membrane and second,whether the intracellular distribution of Tf receptors changed inEC in response to differing brain iron status. Subsequently, weapplied the in vivo findings to a cell culture model of the BBB toshow the presence of a regulatory system for iron release from theBBB. The in vivo model we have used is the ‘Belgrade rat’ thatexpresses a mutation in the iron transporter, Divalent MetalTransporter 1 (DMT1). This mutation results in loss of the ability totransport iron across the epithelial cells in the gut thus leading toa systemic anemia. The DMT1 is also responsible for the release ofendosomal iron that has been released from Tf in the endosomes.In reticulocytes, this mutation prevents the release of iron from

the endosome and thus 80% of the iron and transferrin recyclesback to the cell surface and accounts for the hypochromia andmicrocytosis associated with the mutation.8,9 Similarly, thismutation results in a systemic cellular iron deficiency, which alsorenders the brains of the Belgrade rats iron deficient.10 Thus, theyprovide an ideal model in which to investigate the intracellulardistribution of Tf receptors in the brain ECs to test the hypothesisthat the receptor distribution is dynamic and responsive to brainiron status. By combining this information with our earlierstudies8,11 and the cell culture data herein, we now propose anew model for regulation of brain iron uptake in which the ECs, far

Figure 1. Model for brain iron uptake and regulation. A schematic representation of potential routes for iron delivery to the brain and sites forregulation. Route 1 (red arrows) describes the widely held ‘transcytosis’mechanism iron, which proposes that circulating transferrin-bound iron(holo-transferrin, Tf) binds to transferrin receptors (TfR) on the luminal membrane of the endothelial cells (ECs) that comprise the blood–brainbarrier (BBB). The Tf/TfR complex undergoes endocytosis and Tf remains bound with iron as the endocytic vesicle traverses the ECs andsubsequently fuses with the abluminal membrane at which point the Tf with iron is released into the brain extracellular space. Routes 2 (greenarrows) and 3 (black arrows) also use endocytosis of the Tf/TfR complex to initiate iron uptake, however, the endocytic vesicles are delivered toa Compartment of Uncoupling of Receptor and Ligand (CURL), which is present in all cells. Here the iron Fe3+ is released from the TfR, reducedto Fe2+, and transported out of the endosome by Divalent Metal Transporter 1 (DMT1) to be available either for incorporation into ironrequiring proteins or for storage with ferritin. The apo-Tf in Route 2/2A remains bound to TfR and is recycled to either the luminal or abluminalmembrane where it is released and the TfR is then available for to repeat the process. Route 3 envisages a separation of Tf and TfR with holo-Tfbeing dispatched to abluminal membrane while TfR is returned to the luminal membrane. In contrast to Routes 1 and 3, where it is proposedthat the iron as Fe3+ is release as holo-Tf, in Route 2 it is envisaged that apoTf is released via TfR on the EC and the Fe2+ is released via thetransporter ferroportin. It is then reoxidized to Fe3+ with either cerruloplasmin or another ferroxidase and subsequently recombined with apoTfoutside the EC. The various neural cells can then take up the reconstituted holo-Tf and any subsequently released apoTf can be recycled.

Brain iron uptakeIA Simpson et al

2

Journal of Cerebral Blood Flow & Metabolism (2014), 1 – 10 © 2014 ISCBFM

from serving as a simple conduit, are the focal point for theregulation of cerebral iron homeostasis.

MATERIALS AND METHODSAll animal protocols described in this study were approved by thePennsylvania State University College of Medicine Institutional Animal Careand Use Committee and are consistent with the 2013 AVMA Guidelines.

Immuno/Electron Microscopy ProtocolMale Belgrade homozygous (B/B) and heterozygous (B/+) rats and threeSprague Dawley 6 to 7 months old were transcardially perfused with 4%paraformaldehyde (Fisher, Thermo Fisher Scientific Inc., Waltham, MA, USA,#T353-500)/0.1% glutaraldehyde (EM Sciences, Hatfield, PA, USA, 111-30-8)and postfixed overnight at 4°C in the same solution. The tissue wascryoprotected by immersion in (10% to 30%) sucrose and 16 μm sectionswere prepared. The sections were blocked in (5%) milk and (5%) goatserum before exposure to the mouse monoclonal antibody against the TfR(Zymed, Life Technologies, Grand Island, NY, USA, #13-6800). Thesecondary antibody was a goat anti-mouse Alexa Fluoronanogold Fabfragment (Nanoprobes, Yaphank, NY, USA, #7202) diluted 1:25. Afterexposure to the secondary antibody, the tissue was fixed with 1%glutaraldehyde in phosphate-buffered saline for 10minutes, beforeexposure to silver enhancer (Nanoprobes, #2012) and then osmicated(Polysciences, Warrington, PA, USA, #223). Finally, the slides were stainedwith 2% uranyl acetate before embedding in resin. Sections were cut at 60to 90 nm for non-blinded ultrastructural analysis.

Isolation of Rat MicrovesselsRat microvessels were isolated from three male homozygous andheterozygous Belgrade rats and three Sprague Dawley rats using themethodologies previously described under non-blinded conditions.11–13

Typically, neuronal and astrocytic contamination was o5% as assessed byglucose transporter distribution.12

In Vitro Experiments on Bovine Retinal Endothelial CellsFor the cell culture model of the BBB, bovine retinal endothelial cells(BRECs) of either sex were grown in MCDB-131 media (Gibco, Rockville, MD,USA) supplemented with 10% FBS (Gemini, West Sacremento, CA, USA),10 ng/mL EGF (Invitrogen, Carlsbad, CA, USA), 0.2mg/mL ENDO GRO (VECTechnologies, Inc., Rensselaer, NY, USA), 0.09mg/mL heparin (FisherScientific, Fair Lawn, NJ, USA) and antibiotic/antimycotic (Gibco). For theexperiments, BRECs were grown to confluence on porous filters (0.4 μmpore size) in transwells (Corning Life Sciences, Wilkes Barre, PA, USA) coatedwith fibronectin (Sigma, St Louis, MO, USA). We added serum-free and EGF-free medium to the BRECs with 138 nm hydrocortisone (Sigma) for 48 hoursbefore the experiments to enhance formation of tight junctions.To analyze the expression of ferroportin, BRECs were seeded in 6-well

plates (1 × 106cells/well) in triplicate and treated with Hepcidin (700 nmol/L), apo-Tf (0.1 mg/mL), and holo-Tf (0.1 mg/mL) in serum-free and EGF-freeMCDB131 medium for 24 hours. Then, the cells were harvested and celllysates were solubilized in RIPA buffer (Sigma) containing a proteaseinhibitor cocktail (Sigma). Total protein was estimated using BCA proteinassay kit (Thermo Scientific, Rockford, IL, USA) and the levels of ferroportindetermined by western blot analysis with a rabbit polyclonal anti-ferroportin antibody MTP 11A and for DMT1 with NRAMP-24A antibodyfrom Alpha diagnostics (San Antonio, TX, USA). Enhanced chemilumines-cence (Perkin-Elmer, Waltham, MA, USA) was performed and the blot wasscanned in the Fuji Film System. Then, the blot was stripped, and probedfor β-actin (rabbit polyclonal, 1 to 5,000 dilution; Sigma).

Iron Release Experiments59Fe-NTA complex was prepared and added to the serum-free medium inthe upper chamber of the transwell and incubated overnight at 37°C. Thecells were rinsed and apo-Tf (0.1 mg/mL) (Sigma), holo-Tf (0.1 mg/mL)(Sigma), hepcidin (700 nmol/L) (Peptide Institute, Inc., Osaka, Japan), apo-Ftand holo-Ft (horse spleen ferritin, Sigma) in serum-free, EGF-free MCDB131medium was placed in the lower chamber. RITC-dextran (Sigma) wasincluded in the media in the upper chamber to monitor the integrity of theintercellular junctions in this model. 59Fe release into the lower chamberwas measured by removing aliquots every hour and counting on a single

channel analyzer (Canberra, Meriden, CT, USA). RITC-dextran from thelower chamber was read at 570 nm in a fluorescence plate reader (SpectraMax Gemini, Molecular Devices, Sunnyvale, CA, USA).

Experiments on Primary Rat Astrocyte Cultures and Collection ofConditioned MediaMixed glial cell cultures were prepared from the cortex of 1- to 2-day-oldSprague-Dawley rat brains of either sex following the method described byMcCarthy and de Vellis.14 Cells were plated at a density of 1 × 106/mL in 6-well plates, 6 wells per experimental condition. The next day, theastrocytes were treated with serum-free medium, serum-free mediumwith deferoxamine (100 μmol/L), or serum-free medium with TMH-Ferrocene (10 μmol/L) for 24 hours. Then, medium was removed and cellswere rinsed and incubated in serum-free medium for 24 hours. Theconditioned media from the three conditions control, iron deficient, andiron loaded were collected and used to study the influence of astrocyteconditioned medium on iron release from BBB and to study the influenceof astrocyte conditioned media on TfR expression in BRECs. Subsequently,the BRECs were harvested and lysed in RIPA buffer. The total protein wasestimated using micro BCA method and immunoassay performed asdescribed previously.1

Statistical AnalysisFor all the data, one-way ANOVA was performed with post-hoc Tukey'sMultiple Comparison Test to compare the means of all the groups using theGraphPad Prism 4 software (GraphPad Software, Inc. San Diego, CA, USA).

Simulation MethodsThe model was developed using Berkeley Madonna X (Mach-O; v 8.3.22,Berkeley Software, Berkeley, CA, USA), which is a general-purpose differe-ntial equation solver. Simulations were run over 10,000 to 200,000simulated minutes using the Rosenbrock (stiff) Integration Method.

Monkey Cerebrospinal Fluid SpecimensCerebrospinal fluid (CSF) was obtained from 12 rhesus monkeys (Macacamulatta) between 6 and 8 months of age. Samples were collected underacute ketamine sedation and postprocedural analgesics administeredfollowing an Institutional animal care and use committee-approvedprotocol. The monkeys were selected on the basis of well-characterizedhematology indices to include six healthy iron-sufficient (two female, fourmale) and six iron-deficient anemic animals (two female, four male).Cerebrospinal fluid specimens (1mL, blood free) were obtained via cervicalpuncture and immediately frozen in an ultracold freezer at − 80°C, beforebeing shipped to the laboratory for analysis of Fe release using the in vitroEC model of the BBB.

RESULTSFigure 2A illustrates western blot analysis for Tf receptors andDMT1 in the ECs derived from Sprague Dawley and heterozygousand homozygous Belgrade rats. Figures 2B and 2C illustrate thesubcellular distribution of the transferrin receptors in correspond-ing ECs in intact brain as detected by EM-immunocytochemistryand reveal the presence of Tf receptors on the abluminalmembrane of the ECs. Moreover, we observed an asymmetrictransferrin receptor distribution between the luminal, the abluminal,and the intracellular membranes that was significantly modulatedin the heterozygous and homozygous Belgrade rats and is quanti-tated in Figures 2D and 2E. If transcytosis was the mechanism foriron transport in the brain (depicted as Route 1 in Figure 1), thenmutations in DMT1 should have had no effect on Tf receptordistribution. However, there was clearly a profound change in Tfreceptor distribution in response to the alterations in DMT1. Theobservation of the presence of Tf receptors on the abluminalsurface led us to a second hypothesis that signaling molecules inthe CSF and extracellular fluid are able to report the brain ironstatus to the ECs of the BBB (Route 2 is green in Figure 1).To test these hypotheses, we used a bichamber cell culture

model of the BBB.15 We obtained CSF from healthy, iron-sufficient,

Brain iron uptakeIA Simpson et al

3

© 2014 ISCBFM Journal of Cerebral Blood Flow & Metabolism (2014), 1 – 10

and anemic iron-deficient infant monkeys. The results from thisexperiment are shown in Figure 3A and clearly show increasediron release from ECs upon exposure to CSF from iron-deficientmonkeys versus CSF from iron sufficient monkeys. It is noteworthythat the extent of iron release correlates positively with thehematological indices (Mean Corpuscular Volume and Hemoglo-bin) across the 12 animals tested (Figures 3B and 3C).We subsequently investigated whether signaling to the BBB

could be directed from astrocytes. As shown in Figure 4A, mediafrom astrocytes exposed to iron chelation are associated with

increased iron release as compared with media from the controlastrocytes. In contrast, the media from iron-loaded astrocytes areassociated with decreased iron release. Moreover, astrocyte-conditioned media also affect TfR regulation. Media from iron-deficient astrocytes increase the expression of EC TfRs whereasmedia from iron-loaded astrocytes decrease the expression of TfRs(Figure 4B). These data not only reveal the capacity of astrocytesto influence iron release, but also show a physiologic responsewithin the EC. These experiments reveal that there is regulation ofTfRs in ECs and the regulation is responsive to the brain’s iron

Brain iron uptakeIA Simpson et al

4

Journal of Cerebral Blood Flow & Metabolism (2014), 1 – 10 © 2014 ISCBFM

status and concur with our recent demonstration of ironregulatory proteins being present in the brain microvasculature.8

The return of the Fe release to baseline levels in both Figures 3and 4 suggests the formation of a steady state between the ECand the basal chamber.We further explored the regulatory mechanisms governing iron

release from ECs using specific candidate proteins found in theCSF: apo- and holo-transferrin, h-ferritin, erythropoietin, andhepcidin (Figure 5). Hepcidin acutely reduces the release of ironFe59, whereas apo-transferrin promotes its release while holo-transferrin, ferritin, and erythropoietin are without detectableeffect. The presence of an iron transporter is clearly required toaccount for these results and ferroportin is the obvious candidategiven that it has already been shown in human and rat brainmicrovasculature.8,16,17 The presence of ferroportin in the BRECswas confirmed (Figure 5) and we further showed that the levelscan be reduced by the presence of hepcidin and holo-transferrin,but not apo-transferrin.The novel findings for iron transport and regulation including

redistribution of TfRs determined from the respective electronmicrographs resulted in a model illustrated in Figure 6, thatsignificantly expands the existing paradigm of receptor mediatedtranscytosis at the BBB. The model was generated using knownvalues for both iron and Tf concentrations and their uptake andrelease kinetics in both control and Belgrade rats. The modelincludes estimates for iron, holo-transferrin and apo-transferrinrelease, holo-transferrin regeneration, and rates of neural celltransferrin/apo-transferrin recycling and CSF clearance. In estab-lishing the model, it became necessary to invoke both an ironstorage component within the EC and a mechanism(s) to curtailiron release to prevent brain iron overload. Moreover, in order foriron to be released into the brain, we were required to invoke thepresence of an iron transporter on the abluminal membraneconsistent with our observation for the presence of ferroportinand a mechanism to regenerate holo-transferrin in the braininterstitium. To integrate the data presented above and toformulate the conceptual model presented in Figure 1, someassumptions are necessary. The first requires that all neural cellsthat acquire iron via Tf will release apo-transferrin, therebyestablishing a steady state between iron released by ECs andneural cell iron turnover. In this context, it is important to notethat an iron excess beyond fully saturated Tf has been consistentlyreported in the CSF under normal conditions.18,19 In our model(Figure 1), Route 1 depicts the transcytosis model of brain ironuptake and Route 2 illustrates the pathway for iron uptake. Theimportant distinctions between the two models are (1) Route 2provides a mechanism to account for the iron requirements ofECs; (2) it provides a mechanism by which the iron released withinECs can become available to the brain and thus accounts for the

asymmetric release of iron and transferrin described pre-viously;18,19 (3) it proposes the presence of an iron reservoirwithin the EC and thus provides a mechanism through which ironregulatory proteins could regulate the expression of iron uptakeand storage proteins in response to iron depletion or excess;8 (4) itprovides an explanation as to why the brains of Belgrade rats areiron deficient despite Fe saturation of the circulating Tf and canalso account for the increase in Tf receptor number andredistribution shown in Figure 2; (5) it provides for the regulationof iron release from the BBB by proteins generated in response tobrain iron status and which act on the abluminal surface of the EC.In our model, we assume the CSF is in dynamic equilibrium withthe extracellular fluid and that molecules in both compartmentswill turn over at a normal rate.18 More findings and detailssupporting this conceptual schema, including simulations of theredistribution of TfRs in the control, heterozygous, and homo-zygous Belgrade rats, in response to modulations in the ability ofDMT1 to release iron from the endosomes of the control andBelgrade rats are simulated in Figure 7 (see also the website foundin Supplementary S1). We altered the extent to which DMT1enables release of iron from the endosome to model irontransport in homo- and heterozygous Belgrade rats. A diagram-matic version of the model in Figure 1 in which the individual rateconstants for each step are identified is illustrated in Figure 6 andsimulated in Figure 7. The corresponding derivations for each rateconstant are described in Supplementary Tables S1 and S2. Wherepossible, we used published concentrations for serum and CSF Tfand iron concentrations observed in both control and B/Brats.18,20–22 We also used binding constants for TfR recycling ratesderived by us,23,24 and others. We predicted concentrations ofholo- and apo-transferrin under the various conditions. Whileinterstitial Fe3+ promotes hepcidin release, which in turn inhibitsferroportin, we simulated this process (Figure 7) in a more simpli-fied manner as hepcidin inhibition of ferroportin via a simpleMichaelis-Menten inhibition in which the Ki for hepcidin inhibitionof ferroportin is 0.1 μmol/L. Apo Tf stimulation of ferroportin ismodeled as a simple Michaelis-Menten stimulation characterizedby a stimulation factor S and a K0.5 for Apo-Tf stimulation of0.3 μmol/L. Supplementary Table S1 illustrates the equations des-cribing the flow of transferrin, iron, and receptors between cellsand subcellular compartments; Supplementary Table S2 providespredictions of iron and transferrin flows across the BBB.

DISCUSSIONOur model for the regulation of iron transport across the BBB(Figure 1) concurs with reports on how iron transport acrossintestinal enterocytes is regulated (for review, see Oates et al7 andMorgan and Oates25). Iron release from enterocytes involves

Figure 2. Distribution of transferrin receptors and Divalent Metal Transporter 1 (DMT1) in brain microvessels cells isolated from SpragueDawley (SD) rats and heterozygous (B/+) and homozygous (B/B) Belgrade rats. (A) A typical western blot of isolated microvessels from therespective rat brains (n= 3), which reflects the total level of Tf receptors and DMT1 in the microvessels from the three different animal models.Note that inactive DMT1 is detected in the Belgrade animals. The presence of transferrin receptors in the microvessels was detected by EM-immunostaining in the luminal, abluminal, and intracellular membranes as indicated by an → in a representative section in (B). Typicalelectron micrographs obtained at the level of the substantia nigra from SD, B/+, and B/B rats, respectively, are shown in (C). The enlarged areareveals the distribution of the Tf receptors among the luminal, intracellular, and abluminal membranes with immunogold labeling. (D) Revealsthe absolute distribution of the receptors among the various membranes. Analysis of number of particles in each membrane was performedon 15 vessels from each of three brains/group. The receptor levels were significantly greater in SD rats than either group of Belgrade rats. Theaverage number of particles per microvessel is SD= 134.6± 8.9, B/+= 77.3± 5.7, B/B= 96.7± 6.9. SD4B/+ (Po0.0001) SD4B/B (P= 0.0006) B/+ versus B/B (NSD P= 0.067). The fractional distribution of total cellular Tf receptors in each membrane in is shown in (E). * denotesintracellular TfR4luminal or abluminal TfR (P≤ 0.0001; two-way ANOVA) comparing expression in all compartments within each animal type, ¥denotes B/B intracellular TfRoS/D intracellular TfR (Po0.005; one-way ANOVA). § denotes abluminal TfR4luminal TfR in the B/B bloodvessels (P= 0.0001; two-way ANOVA). No other abluminal versus luminal differences were statistically significant. ‡ denotes B/B and B/+abluminal TfR4S/D abluminal TfR (Po0.0006; one-way ANOVA). The relative change in the level of Tf receptors in the abluminal membranesof the B/B rats is significantly increased above both the SD and B/+ rats, which is consistent with a redistribution of receptors from theintracellular membranes to the abluminal membranes as proposed in the feedback models illustrated in Figures 1 and 6 and simulated inFigure 7.

Brain iron uptakeIA Simpson et al

5

© 2014 ISCBFM Journal of Cerebral Blood Flow & Metabolism (2014), 1 – 10

extracellular hepcidin and feedback modulated by apo-transferrin/holo-transferrin ratio in the serum that reflects the iron status ofthe body. Herein, we have showed an analogous system in thebrain with regulation of iron release and a feedback mechanismthat is receptive to signals generated by different neural cells inresponse to their respective iron status. The first clue pointing tothe existence of this feedback mechanism was the discovery ofsignificant numbers of TfRs on the abluminal membrane of ECs,which progressively increase in the ECs of the heterozygous and

homozygous Belgrade rats, respectively. These rats have a mutationin the protein DMT1, which is responsible for releasing iron fromthe endosomes that make up the Compartment of Uncoupling ofReceptor and Ligand, in Figure 1. In the absence of functionalDMT1, iron release into the cytoplasm of all the cells of the body,including ECs, is compromised. We previously reported thepresence of DMT1 in ECs of the BBB both in rats26 and humans.8

Although there is not universal agreement on this point,27 therequirement for a mechanism for iron to enter the EC labile iron

Figure 3. (A) Iron release from blood–brain barrier (BBB) culture model after exposure to cerebrospinal fluid (CSF) from normal or iron-deficient (ID) monkeys. 59Fe-NTA was added to the endothelial cell monolayer in the upper chamber and incubated overnight. The mediumfrom upper and lower chambers was removed and fresh media were returned to the upper chamber and artificial CSF or CSF from either ID oriron-sufficient (IS) monkeys was added to the basal chamber. RITC dextran was also added to the upper chamber to monitor the integrity ofthe tight junctions in the bovine retinal endothelial cells (BRECs). Aliquots were taken from the basal chamber at the time periods indicated.Iron release from BRECs increased significantly when CSF from ID monkeys was placed in the lower chamber compared with the iron releasefor control CSF. The increased iron release with IS monkey CSF compared with artificial CSF shows the presence of factors not in artificial CSF.That the iron release returns to baseline over time suggests a saturable signal consistent with our observation of Tf receptors on the abluminalmembrane shown in Figure 1. (B, C) Scatterplot of individual data for CSF used in this figure, showing the consistency of the relationshipbetween iron release from EC and the monkeys’ hematological iron status. CSF was purposefully collected from six iron-deficient ones thatmet standard criteria for anemia (mean corpuscular volumeo60 fL) and compared with specimens from six healthy iron-sufficient ones. Theamount of 59Fe released at the 2-hour peak time point was inversely correlated with the monkeys’ peripheral iron status, suggesting that irontransport proteins in the CSF were upregulated in an attempt to promote iron release from the ECs. Prior studies on other monkeys haveshown that this compensatory response is only partially successful, however, and the brain of an anemic monkey remains iron deficient for atleast 3 to 4 months after its hematology and peripheral iron status have been restored by consuming iron-rich foods or regular ferrous sulfatetreatments. These data indicate that there is communication between brain iron uptake and peripheral iron status.

Brain iron uptakeIA Simpson et al

6

Journal of Cerebral Blood Flow & Metabolism (2014), 1 – 10 © 2014 ISCBFM

pool is acknowledged in the literature28 and required by ourmodel. The studies on Belgrade rats showed that despite highcirculating Tf levels, brain iron uptake was compromised.22 Thisoutcome is clearly contrary to the predictions of the transcytosismodel, but is consistent with the view that the luminal TfR is notrate-limiting for the release of iron into the brain. We used thedata of Knutson et al8 and Farcich and Morgan29 in the current

model, and have indicated that iron release in the EC is reducedby 50% and 100% in B/+ and B/B, respectively, providing acompelling argument that iron is released into the cytoplasm ofthe ECs forming the BBB and can be subsequently released into

Figure 4. Release of iron from the blood–brain barrier (BBB) isregulated by signaling proteins released by astrocytes. Rat primaryastrocytes were grown in standard medium (control) or mediacontaining either TMH-ferrocene to iron load or Desferal, an ironchelator, to iron deplete the cells. After 24 hours, the media werereplaced with fresh medium after rinsing the cells thoroughly. (A)The experimental astrocyte media were allowed to equilibrate for24 hours and then removed and placed in the lower chamber of ECcultures. The EC had been previously loaded with 59Fe. The mediafrom the basal chamber of the EC were sampled for the presence of59Fe over the time course indicated in the graph. Iron release fromthe EC significantly increased with medium from iron-chelatedastrocytes in lower chamber whereas medium from iron-loadedastrocytes in the lower chamber was associated with decreased ironrelease. The RITC dextran levels in the lower chamber weremonitored to determine the integrity of the tight junctions. Thesedata are consistent with our model that there is iron release fromthe EC and that the rate of release can be altered by signalingproteins. These data show that the signaling proteins are releasedby astrocytes. (B) This graph shows that experimental media fromastrocytes altered expression of Tf receptors in EC. Thus, indicatingthat media contained signals that not only affected iron release (asshown in A), but also induced changes in the EC intracellular ironstatus resulting in altered expression of Tf receptors in a predictabledirection (increased when sensing iron deficiency and decreasedwhen iron loaded).

Figure 5. (A) Iron is released from endothelial cells (ECs) and theamount released can be modulated by candidate signaling proteins.To show that the stored iron could be released from the ECs, ECs werepreloaded with 59Fe-NTA for 12 hours. After the 12-hour incubationperiod, apo-Tf (0.1mg/mL), holo-Tf (0.1mg/mL), hepcidin (700 nmol/L), erythropoietin (EPO 4mU/mL), horse spleen apo-Ft (50 nmol/L), orhorse spleen holo-Ft (50 nmol/L) were placed in the lower chamber.These concentrations were chosen as reflective of the levels normallyfound in the cerebrospinal fluid (CSF). After 1 hour, the media fromthe lower chamber were sampled (100 μL) and every subsequenthour for 5 hours. During this 5-hour period, dextran labeled with FITCis present in the upper chamber as a control for integrity of the tightjunctions and the media from the lower chamber are monitored forthe presence of FITC-dextran over the course of the experiment. Theamount of 59Fe (measured as Gamma counts) released into the lowerchamber over 5 -hour culture is plotted against the baseline rate ofrelease over a 5-hour time period. There was no change in dextrancontent over the time period measured (data not shown). (B, C)Ferroportin is expressed in bovine retinal endothelial cells (BRECs)and is responsive to hepcidin and transferrin. Western blot analysiswas performed to determine ferroportin protein expression in BRECsafter 24-hour treatment with hepcidin, apo-Tf, and holo-Tf (B).Ferroportin mRNA expression relative to internal actin control (%)significantly decreased when the cells were treated with hepcidin(Hepc) and holo-Tf compared with the control BRECs (**Po0.01) (C).The expression of ferroportin significantly decreased with holo-Tftreatment compared with apo-Tf-treated BRECs (*Po0.05) and thecontrol BRECs (**Po0.01). The bars represent mean value for eachgroup and the error bars represent the s.e.m. for each group.

Brain iron uptakeIA Simpson et al

7

© 2014 ISCBFM Journal of Cerebral Blood Flow & Metabolism (2014), 1 – 10

the brain. The presence of the membrane iron export proteinferroportin in EC cultures30 (Figure 5) and in vivo 31 indicates atransporter for non-transferrin mediated iron release from the EC.These observations coupled with prior reports on the existence offerritin, the iron storage protein in ECs further strengthen theanalogies between how enterocytes and ECs of the BBB regulateiron release and monitor signaling feedback.Although our BBB model has some similarity to the gut, distinct

differences exist. For example, there is no transferrin-mediated

uptake of iron from the lumen of the gut, but rather the acidicenvironment of the gut promotes iron uptake via DMT1 directly.32

The release of iron from the gut via ferroportin is clearly analogousto the brain, although at least in cultured ECs, there also appearsto be iron export via Tf.33,34 In the current model, we haveaccorded 0%; however, the fraction of the holo-transferrin to apo-transferrin clearly remains to be determined in vivo and amendedwithin the model. The ability of apo-transferrin to induce ironrelease from the ECs has not been shown in the gut models and

Figure 6. The model describes transfer of Fe3+ and transferrin from serum, across the blood–brain barrier (endothelial cell, EC) to theinterstitium and a ‘brain cell’. Key elements include: (1) Fe2+ export from intracellular endosome to cytoplasm (described by k9 and k19),ferroportin-mediated Fe export to interstitium (described by k11 and k20), low affinity/low capacity Fe export from EC to serum (described byJ41) and hepcidin feedback inhibition of EC ferroportin when interstitial Fe2+ rises (described by Ki). This was found to be necessary to preventoveraccumulation of Fe2+. Exported Fe2+ must be reoxidized to Fe3+ by Ceruloplasmin on astrocytes before it can recombine withapotransferrin to regenerate transferrin. Intracellular Fe2+ is sequestered by ferritin (Fr). The contents of the interstitium drain at a defined rate(sink) into the cerebrospinal fluid (CSF) and subsequently out of the brain.

Brain iron uptakeIA Simpson et al

8

Journal of Cerebral Blood Flow & Metabolism (2014), 1 – 10 © 2014 ISCBFM

represents a significant new insight into release of iron from cells.In enterocytes, as Fe2+ iron is exported, it is immediately oxidizedto Fe3+ by hephaestin and then bound to available apo-Tf.35 In thebrain, we proposed that the oxidation of the released Fe2+ viaferroportin is mediated by ceruloplasmin. Ceruloplasmin isreportedly present on the end feet of astrocytes that engulf theECs.36,37 In our experiments in which we used CSF or astrocyteconditioned media, there was a return of Fe release to baselinesuggesting the establishment of a steady state. The return of Fe tothe EC would require conversion of the Fe2+ to Fe3+ and bindingto apoTf. This conversion requires the presence of a solubleferroxidase and ceruloplasmin has recently been proposed as apotential candidate by McCarthy and Kosman.38 Additionalsupport for a role for ceruloplasmin is provided by the observationin a mouse model of aceruloplasminemia, in which ferritinexpression is increased in astrocytes, suggesting that these cellshave accumulated iron. The mechanism by which iron accumu-lates in astrocytes is not known; they do not express TfRs, but havebeen shown to express DMT1.26 The secretome of the astrocytesand how it is affected by their iron status is under investigation.On the basis of our model, if hepcidin is released by astrocytes, asin the liver to signal iron sufficiency, then iron accumulation by

astrocytes would reduce iron release from ECs, which is indeedwhat we observed in Figure 4. It should be noted that theconcentration of apo-transferrin in the basal chamber in ourmodel is entirely comparable to that present in the circulation andthe hepcidin level represents approximately 5 to 7X the circulatingconcentration and 3X the apparent dissociation constant (KD) forhepcidin..39,40 To our knowledge, the concept of ambient ironstatus signaling in the brain has not been investigated. Moreover,the studies using monkey CSF provide striking evidence that thesystemic iron status is related to release of iron into the brain. Thisobservation suggests a heretofore, unappreciated relationshipthat could represent novel insights into communication betweensystemic iron status and brain iron uptake.

DISCLOSURE/CONFLICT OF INTERESTThe authors declare no conflict of interest.

AUTHOR CONTRIBUTIONSDrs Connor, Simpson, Carruthers, Ponnuru, Coe, and Lubach contributed towriting the manuscript. Drs Simpson, Carruthers, Connor, and Devraj developed

Figure 7. (A) Effects of loss of Divalent Metal Transporter 1 (DMT1) on endothelial cell (EC) luminal, abluminal, and intracellular transferrinreceptor content. Total transferrin receptor in each compartment includes unliganded receptor, receptor complexed with transferrin andreceptor complexed with apotransferrin (e.g., in compartment 1, TR1+TRTf1+TRApo1). The simulation model was run using the flows andconstants described in Supplementary Tables S1 and S2. Each run required a simulated time course of 3 × 105 seconds (3.5 days) to achievesteady state. Rate constants (k1 to k7) were carefully adjusted to simulate SD (Sprague Dawley, control) data. The activity of endosomal DMT1(described by rate constant k9) was then adjusted to zero and relative distributions of TR recomputed. Rate constants were then adjusted (seeSupplementary Table S2) until experimental data (homozygous DMT1-null condition, − /− ) were successfully simulated. DMT1 (k9) was thenrestored to full activity, control TR distributions recalculated and rate constants were adjusted again to recapitulate control TR distributions.This process was repeated until SD and homozygous DMT1-null mice data were successfully simulated by a common set of rate constants(Supplementary Table S2). DMT1 activity (k9) was then adjusted to 50% and TR distributions computed for the DMT1 +/− condition. Openbars represent experimental data and closed bars the simulated data—all produced by a single common set of rate constants summarized inSupplementary Table S1. (B) Relative Iron (Fe) and transferrin (Tf ) uptake by the brain (J13 and Tf uptake respectively in Supplementary TableS1) in control (SD), homozygous (− /− ) and heterozygous (+/− ) DMT1 knockouts. After simulating TR distributions in Figure 6A, iron (solidbars) and apo-transferrin (empty bars) uptake was calculated for SD (red), homozygous (green), and heterozygous (blue) DMT1 knockouts. (C)Effect of interstitial [apo-transferrin] on iron uptake by the brain. Ordinate: relative iron uptake; abscissa: interstitial [apo-transferrin] μmol/L.(D) Effect of interstitial [Fe2+] on iron uptake by the brain. Ordinate: relative iron uptake; abscissa: interstitial [Fe2+] μmol/L.

Brain iron uptakeIA Simpson et al

9

© 2014 ISCBFM Journal of Cerebral Blood Flow & Metabolism (2014), 1 – 10

the model and conceived the experiments in this study. Drs Ponnuru andDevraj performed the experiments with EM technical support from Ms Klingerand Mr Myers. Drs Coe and Lubach provided the monkey CSF and data onblood parameters.

REFERENCES1 Zecca L, Youdim MB, Riederer P, Connor JR, Crichton RR. Iron, brain ageing and

neurodegenerative disorders. Nat Rev Neurosci 2004; 5: 863–873.2 Lozoff B, Beard J, Connor J, Barbara F, Georgieff M, Schallert T. Long-lasting neural

and behavioral effects of iron deficiency in infancy. Nutr Rev 2006; 64:S34–S43, discussion S72-91.

3 Beard JL, Connor JR. Iron status and neural functioning. Ann Rev Nutr 2003; 23:41–58.

4 Unger EL, Earley CJ, Beard JL. Diurnal cycle influences peripheral and brain ironlevels in mice. J Appl Physiol (1985) 2009; 106: 187–193.

5 Klausner RD, Van Renswoude J, Ashwell G, Kempf C, Schechter AN, Dean A et al.Receptor-mediated endocytosis of transferrin in K562 cells. J Biol Chem 1983; 258:4715–4724.

6 Dickinson TK, Connor JR. Histological analysis of selected brain regions ofhypotransferrinemic mice. Brain Res 1994; 635: 169–178.

7 Oates PS, Thomas C, Morgan EH. Transferrin receptor activity and localisation inthe rat duodenum. Pflugers Archiv: European journal of physiology 2000; 440:116–124.

8 Knutson M, Menzies S, Connor J, Wessling-Resnick M. Developmental, regional,and cellular expression of SFT/UbcH5A and DMT1 mRNA in brain. J Neurosci Res2004; 76: 633–641.

9 Farcich EA, Morgan EH. Uptake of transferrin-bound and nontransferrin-boundiron by reticulocytes from the Belgrade laboratory rat: comparison with Wistar rattransferrin and reticulocytes. Am J Hematol 1992; 39: 9–14.

10 Burdo JR, Martin J, Menzies SL, Dolan KG, Romano MA, Fletcher RJ et al. Cellulardistribution of iron in the brain of the Belgrade rat. Neuroscience 1999; 93:1189–1196.

11 Burdo JR, Simpson IA, Menzies S, Beard J, Connor JR. Regulation of the profile ofiron-management proteins in brain microvasculature. J Cereb Blood Flow Metab2004; 24: 67–74.

12 Maher F, Vannucci SJ, Simpson IA. Glucose transporter isoforms in brain: absenceof GLUT3 from the blood-brain barrier. J Cereb Blood Flow Metab 1993; 13:342–345.

13 Simpson IA, Appel NM, Hokari M, Oki J, Holman GD, Maher F et al. Blood-brainbarrier glucose transporter: effects of hypo- and hyperglycemia revisited. J Neuro-chem 1999; 72: 238–247.

14 McCarthy KD, de Vellis J. Preparation of separate astroglial and oligodendroglialcell cultures from rat cerebral tissue. J Cell Biol 1980; 85: 890–902.

15 Antonetti DA, Wolpert EB, DeMaio L, Harhaj NS, Scaduto RC, Jr. Hydrocortisonedecreases retinal endothelial cell water and solute flux coincident with increasedcontent and decreased phosphorylation of occludin. J Neurochem 2002; 80:667–677.

16 Yang WM, Jung KJ, Lee MO, Lee YS, Lee YH, Nakagawa S et al. Transientexpression of iron transport proteins in the capillary of the developing rat brain.Cell Mol Neurobiol 2011; 31: 93–99.

17 Wu LJ, Leenders AG, Cooperman S, Meyron-Holtz E, Smith S, Land W et al.Expression of the iron transporter ferroportin in synaptic vesicles and the blood-brain barrier. Brain Res 2004; 1001: 108–117.

18 Bradbury MW. Transport of iron in the blood-brain-cerebrospinal fluid system.J Neurochem 1997; 69: 443–454.

19 Moos T, Morgan EH. Kinetics and distribution of [59Fe-125I]transferrin injectedinto the ventricular system of the rat. Brain Res 1998; 790: 115–128.

20 Ueda F, Raja KB, Simpson RJ, Trowbridge IS, Bradbury MW. Rate of 59Fe uptakeinto brain and cerebrospinal fluid and the influence thereon of antibodies againstthe transferrin receptor. J Neurochem 1993; 60: 106–113.

21 Moos T, Morgan EH. The metabolism of neuronal iron and its pathogenic role inneurological disease: review. Ann NY Acad Sci 2004; 1012: 14–26.

22 Moos T, Morgan EH. The significance of the mutated divalent metal transporter(DMT1) on iron transport into the Belgrade rat brain. J Neurochem 2004; 88:233–245.

23 Davis RJ, Faucher M, Racaniello LK, Carruthers A, Czech MP. Insulin-like growthfactor I and epidermal growth factor regulate the expression of transferrinreceptors at the cell surface by distinct mechanisms. J Biol Chem 1987; 262:13126–13134.

24 Roskams AJ, Connor JR. Aluminum access to the brain: a role for transferrin and itsreceptor. Proc Natl Acad Sci USA 1990; 87: 9024–9027.

25 Morgan EH, Oates PS. Mechanisms and regulation of intestinal iron absorption.Blood Cells Mol Dis 2002; 29: 384–399.

26 Burdo JR, Menzies SL, Simpson IA, Garrick LM, Garrick MD, Dolan KG et al. Dis-tribution of divalent metal transporter 1 and metal transport protein 1 in thenormal and Belgrade rat. J Neurosci Res 2001; 66: 1198–1207.

27 Moos T, Skjoerringe T, Gosk S, Morgan EH. Brain capillary endothelial cells mediateiron transport into the brain by segregating iron from transferrin without theinvolvement of divalent metal transporter 1. J Neurochem 2006; 98: 1946–1958.

28 Moos T, Rosengren Nielsen T, Skjorringe T, Morgan EH. Iron trafficking insidethe brain. J Neurochem 2007; 103: 1730–1740.

29 Farcich EA, Morgan EH. Diminished iron acquisition by cells and tissues ofBelgrade laboratory rats. Am J Physiol 1992; 262: R220–R224.

30 Hadziahmetovic M, Song Y, Ponnuru P, Iacovelli J, Hunter A, Haddad N et al. Age-dependent retinal iron accumulation and degeneration in hepcidinknockout mice. Invest Ophthalmol Vis Sci 2011; 52: 109–118.

31 Connor JR, Ponnuru P, Wang XS, Patton SM, Allen RP, Earley CJ. Profile ofaltered brain iron acquisition in restless legs syndrome. Brain 2011; 134:959–968.

32 Oates PS, Thomas C. Augmented internalisation of ferroportin to late endosomesimpairs iron uptake by enterocyte-like IEC-6 cells. Pflugers Archiv: European journalof physiology 2005; 450: 317–325.

33 Hudson DM, Krisinger MJ, Griffiths TA, Macgillivray RT. Neither human hephaestinnor ceruloplasmin forms a stable complex with transferrin. J Cell Biochem 2008;103: 1849–1855.

34 Burdo JR, Antonetti DA, Wolpert EB, Connor JR. Mechanisms and regulation oftransferrin and iron transport in a model blood-brain barrier system. Neuroscience2003; 121: 883–890.

35 Chen H, Attieh ZK, Dang T, Huang G, van der Hee RM, Vulpe C. Decreasedhephaestin expression and activity leads to decreased iron efflux from differ-entiated Caco2 cells. J Cell Biochem 2009; 107: 803–808.

36 Unger EL, Earley CJ, Thomsen LL, Jones BC, Allen RP. Effects of IV ironisomaltoside-1000 treatment on regional brain iron status in an iron-deficientanimal. Neuroscience 2013; 246: 179–185.

37 Chirasani SR, Markovic DS, Synowitz M, Eichler SA, Wisniewski P, Kaminska B et al.Transferrin-receptor-mediated iron accumulation controls proliferation and glu-tamate release in glioma cells. J Mol Med (Berl) 2009; 87: 153–167.

38 McCarthy RC, Kosman DJ. Ferroportin and exocytoplasmic ferroxidase activity arerequired for brain microvascular endothelial cell iron efflux. J Biol Chem 2013; 288:17932–17940.

39 Zhang DL, Senecal T, Ghosh MC, Ollivierre-Wilson H, Tu T, Rouault TA. Hepcidinregulates ferroportin expression and intracellular iron homeostasis of erythro-blasts. Blood 2011; 118: 2868–2877.

40 De Domenico I, Nemeth E, Nelson JM, Phillips JD, Ajioka RS, Kay MS et al. Thehepcidin-binding site on ferroportin is evolutionarily conserved. Cell Metab 2008;8: 146–156.

Supplementary Information accompanies the paper on the Journal of Cerebral Blood Flow & Metabolism website (http://www.nature.com/jcbfm)

Brain iron uptakeIA Simpson et al

10

Journal of Cerebral Blood Flow & Metabolism (2014), 1 – 10 © 2014 ISCBFM