Address correspondence and reprint requests to Torben Moos, Department of Health Science and Technology, Aalborg University, Fredrik Bajers Vej 3B, 1.216, 9220 Aalborg East, DK-Denmark. E-mail: email@example.com
Iron, an essential element for all cells of the body, including those of the brain, is transported bound to transferrin in the blood and the general extracellular fluid of the body. The demonstration of transferrin receptors on brain capillary endothelial cells (BCECs) more than 20 years ago provided the evidence for the now accepted view that the first step in blood to brain transport of iron is receptor-mediated endocytosis of transferrin. Subsequent steps are less clear. However, recent investigations which form the basis of this review have shed some light on them and also indicate possible fruitful avenues for future research. They provide new evidence on how iron is released from transferrin on the abluminal surface of BCECs, including the role of astrocytes in this process, how iron is transported in brain extracellular fluid, and how iron is taken up by neurons and glial cells. We propose that the divalent metal transporter 1 is not involved in iron transport through the BCECs. Instead, iron is probably released from transferrin on the abluminal surface of these cells by the action of citrate and ATP that are released by astrocytes, which form a very close relationship with BCECs. Complexes of iron with citrate and ATP can then circulate in brain extracellular fluid and may be taken up in these low-molecular weight forms by all types of brain cells or be bound by transferrin and taken up by cells which express transferrin receptors. Some iron most likely also circulates bound to transferrin, as neurons contain both transferrin receptors and divalent metal transporter 1 and can take up transferrin-bound iron. The most likely source for transferrin in the brain interstitium derives from diffusion from the ventricles. Neurons express the iron exporting carrier, ferroportin, which probably allows them to excrete unneeded iron. Astrocytes lack transferrin receptors. Their source of iron is probably that released from transferrin on the abluminal surface of BCECs. They probably to export iron by a mechanism involving a membrane-bound form of the ferroxidase, ceruloplasmin. Oligodendrocytes also lack transferrin receptors. They probably take up non-transferrin bound iron that gets incorporated in newly synthesized transferrin, which may play an important role for intracellular iron transport.
Iron is essential for a plethora of functions in all cells. In the brain these include neurotransmission, myelination and cell division. In the circulation, iron is bound to transferrin with a binding-capacity for iron that only reaches its limit in diseases like hemochromatosis in which non-transferrin bound iron present as a low-molecular weight form can be detected in plasma (Batey et al. 1980; Brissot et al. 1985). The hydrophilic nature of the iron-containing transferrin prevents its passage into the brain, but to circumvent this feature and simultaneously nourish neurons and glial cells, the brain, as the only organ in the body with this capacity, expresses receptors for transferrin (transferrin receptor 1) on the luminal side of its capillaries (Angelova-Gateva 1980; Jefferies et al. 1984; Kawabata et al. 1999). Hence, transferrin receptor-mediated uptake of iron by brain capillary endothelial cells (BCECs) followed by further transport into the brain is the major mechanism by which iron is transported into the brain. Only a small proportion of total transport into the brain occurs via the choroids plexuses (Crowe and Morgan 1992). Even in conditions in which non-transferrin-bound iron is present in the plasma, iron fails to accumulate in the brain. This applies both to a mouse model of hemochromatosis (Moos et al. 2000) and in patients suffering from hemochromatosis (Russo et al. 2004), thus excluding the transport of significant amounts of this form of iron into the brain. Nevertheless, iron is present in approximately normal amounts in the brain of hypotransferrinemic mice, which have extremely low concentrations of circulating mouse transferrin (Beard et al. 2005), which can be because of uptake of non-transferrin-bound iron prior to fully development of the blood-brain barrier (BBB) integrity, transport across the choroid plexus at all ages (Ueda et al. 1993; Takeda et al. 2001), or transport of iron present on the small amount of transferrin naturally present in the plasma or injected in order to allow the animals to survive.
The general consensus on uptake of iron-transferrin by transferrin receptors present on BCECs is not followed by agreement when considering the mechanism of iron release from transferrin and its transport from the interior of BCECs and further into the brain. For a long time researchers have been divided between interpretations claiming that iron would be transported through BCECs either (i) by means of receptor-mediated transcytosis of iron-containing transferrin from the luminal to the abluminal side, or (ii) by receptor-mediated endocytosis at the luminal side followed by detachment of iron from transferrin inside endosomes and subsequent transport into the brain, plus retro-endocytosis of iron-free (apo)-transferrin to the luminal side of BCECs, followed by its release into plasma (cf. Moos and Morgan 2000; Rouault and Cooperman 2006). Emerging data, however, indicate that the transport of iron into the brain is more complicated. In disagreement with the hypothesis on transcytosis, there is no evidence for transport of transferrin through BCECs (Crowe and Morgan 1992; Strahan et al. 1992; Moos et al. 2006). Likewise, the hypothesis predicting that iron is transported out of endosomes suffers from the observation that the molecule responsible for transporting iron from endosomes, divalent metal transporter 1 (DMT1) is absent from BCECs (Moos and Morgan 2004; Moos et al. 2006).
Another topic of considerable interest is the fate of iron transported through BCECs. Recent data point towards an essential role of molecules released from astrocytes for metabolism of iron present in a low-molecular weight form in the brain extracellular space (Moos and Morgan 2004; Jeong and David 2006), which indirectly supports the idea of non-transferrin bound iron being released from BCECs to the brain interior. In this review, we propose hypotheses on (i) how iron is transported through BCECs by means of abluminal detachment from transferrin via active participation of astrocytes, (ii) how iron circulates in the brain extracellular fluid (interstitial fluid and CSF) bound to transferrin and other molecules, (iii) how iron is taken up by neurons and macroglia (astrocytes and oligodendrocytes) and microglia, and (iv) how iron gets transported out of the brain.
Blood to brain transport of iron
Blood to endothelium transport
Several observations indicate that transferrin receptors are essential for iron uptake by the brain. In embryonic life, transferrin receptors are expressed by proliferating neural progenitor cells (Copp et al. 1992) and BCECs from the time they appear (Fig. 1) (Moos et al. 1998). Failure of the transferrin receptor expression in fetal mice results in lethal outcomes during development with severe defects in the CNS most probably because the lack of transferrin receptors prevents iron uptake by dividing brain cells (Levy et al. 1999). The transferrin receptor is continuously expressed by BCECs and neurons in the developing rodent brain (Moos et al. 1998), but the expression pattern is age-dependent with peak expression in BCECs around the second postnatal week. In neurons expression reaches its highest levels from the beginning of the fourth postnatal week, when the BBB integrity is fully developed and the rate of iron transport into the brain is lower than that of the developing brain (Taylor and Morgan 1990).
In conditions with iron deficiency, the transport of iron into the brain is considerably higher than in conditions with normal available levels of iron (Taylor et al. 1991). The synthesis of the transferrin receptor is generally known to be regulated at the mRNA level through five iron responsive elements (IRE) present in the untranslated region of the mRNA (Rouault 2006). If the cellular iron level is low, these IRE’s interact with iron regulatory proteins to protect mRNA from cleavage and degradation and thereby increase protein synthesis (Rouault 2006). Surprisingly iron deficiency is not accompanied by higher expression of transferrin receptor protein in BCECs (Fig. 2) (Moos et al. 1998). A possible explanation for this finding could be that the BCECs do not increase their expression of transferrin receptor molecules but that they instead raise the cycling rate of endosomes containing transferrin receptors. Supporting this notion, targeting the BCEC transferrin receptor by intravenous injection of monoclonal antibodies directed against the receptor in normal and iron deficient rats did not show increased binding in the iron deficient animals (Moos and Morgan 2001), and cultured bovine BCECs were found to contain up to a 80% spare pool of transferrin receptors that could be mobilized in conditions associated with shortage of available iron (Van Gelder et al. 1995; Visser et al. 2004).
The concentration of iron in the brain increases with increasing age and is even relatively higher in brains of subjects with neurodegenerative diseases such as Parkinson’s disease (Dexter et al. 1987; Sofic et al. 1988). Little is known about the expression pattern of transferrin receptors in capillaries of the aging brain. Likewise, the expression of transferrin receptors in brains of neurodegenerative diseases has received little study. Hence, there is currently no evidence on whether the high concentration of iron of aging or neurodegenerative cases can be attributed to increased BCECs transferrin receptor function.
Endothelium to brain transport
Binding of iron-transferrin to the transferrin receptor is followed by docking and formation of an endosome that enters the BCEC. The endosomal pH is slightly acidic, which causes the iron to be liberated from transferrin. As described in many other cell types, the presence of DMT1 allows the transport of divalent cations out of the endosome and into the cytosol while exchanging the cationic load inside the endosome with two protons (Gunshin et al. 1997). Surprisingly DMT1 could not be detected in BCECs in a series of independent investigations using a panel of antibodies raised against different regions of the DMT1 molecule such as the conserved trans-membrane region and the variable C-terminal region with or without an iron responsive element (+IRE vs. −IRE) (Fig. 3) (Moos and Morgan 2004; Moos et al. 2006). In contrast, DMT1 was readily detectable in neurons where it was present in a punctate form in the cytoplasm and choroid plexus epithelial cells. Also, the high requirement for iron in the developing brain was not reflected by expression of DMT1 in BCECs (Moos et al. 2006). Supporting these observations, using a non-radioactive in situ hybridization that allows for high resolution examination of tissue section at high magnification, Gunshin et al. (1997) reported the presence of DMT1 mRNA in neurons and choroids plexus epithelial cells but did not report its presence in BCECs or glial cells. These results conflict with those of Siddapappa et al. (2002, 2003) who reported the presence of DMT1 in BCECs. However they found a uniform distribution of DMT1 throughout the BCEC cytoplasm, which is less likely to reflect its real distribution.
What should not be overlooked is the BCECs own mandatory need for iron. This need is probably particularly high during development of the brain when BCECs are rapidly proliferating (Mato et al. 1989). Given the robust expression of transferrin receptors by BCECs during development, the most likely mechanism of iron acquisition by the BCECs is by DMT1-mediated pumping into the cytosol from iron-containing endosomes. As mentioned above, however, DMT1 was not detected in BCECs in spite of a ready detection in neurons and choroid plexus epithelium. Should DMT1 expression be very weak and too low to be detectable by immunoassays, we propose that this theoretical occurrence of DMT1 in BCECs relates to DMT1 involved with iron-transport into the cytosol to feed the BCECs themselves but does not play a role in iron transport through the BCECs.
Based on these observations we conclude that DMT1 is unlikely to play a role for iron-transport across BCECs, and that iron is transported from the luminal to the abluminal surface of these cells inside vesicles without any step which leads to its release from the endosome into the cytosol and from there into the brain interstitium. Evidence to support this hypothesis would be to detect transferrin receptor-containing vesicles at the abluminal side of BCECs. The transport of transferrin through BCECs has been studied in vivo with techniques such as examination at the ultrastructural level using detection of horseradish peroxidase-conjugated transferrin (Roberts et al. 1993), gold-labeled anti-transferrin receptor monoclonal antibodies (OX26) (Bickel et al. 1994), and co-detection of OX26 with laminin to identify the basal lamina of the BCECs (Gosk et al. 2004). Although these studies did provide some evidence that transferrin receptor-containing vesicles are present at the abluminal side of BCECs there is, to the knowledge of the present authors, still no study that has been able to convincingly address this problem in vivo, although in vitro studies of BCECs grown in polarized conditions do indicate that transferrin-containing vesicles fuse with the abluminal surface (Van Gelder et al. 1997; Burdo et al. 2003). If transferrin receptor-containing vesicles are transported through the BCECs and fuse with the abluminal side, they would offer iron detached from transferrin and ready to be transported further into the brain. The iron atoms would probably be in their ferric form, and the process would not require the function of ferroportin, which is responsible for ferrying iron out of other types of cells (Wessling-Resnick 2006). There is disagreement as to whether ferroportin is present in (Wu et al. 2004) or absent from BCECs (Moos and Rosengren Nielsen 2006).
Astrocytes may influence iron release from BCECs
Astrocytic end-feet form intimate contacts with the abluminal side of BCECs of around 95% of the area denoted by the basal lamina (Fig. 4) (Brightman and Reese 1969; Kacem et al. 1998; Abbott et al. 2006). Although the BBB integrity is determined by BCECs alone, this close relationship is generally believed to denote an important role for astrocytes in maintaining the barrier integrity. The astrocytic end-feet processes could play an important metabolic role for neutralization of solutes transported through the BCECs into the brain to preclude the risks of such solutes impairing the delicate extracellular environment of the brain. In the case of iron, astrocytes could be important regulators of iron import into the brain as proposed in recent publications (Moos and Morgan 2004; Moos et al. 2006) and summarized below.
A hypothetic mechanism of iron release from transferrin on the abluminal surface of BCECs
Iron-containing transferrin is transported to the abluminal surface of BCECs at where it remains bound to its receptor but is exposed to a local microenvironment, which leads to release of the iron (Fig. 5). Several factors which can release iron from transferrin are present in the brain extracellular fluid and may be at relatively high concentrations near the BCEC-astrocyte end-foot junction. These include hydrogen ions, ATP and other nucleotides, and citrate (Morgan 1977, 1979). The rate of release of iron is directly proportional to the hydrogen ion concentration; a reduction in pH of only 0.1 pH unit can cause a marked increase in iron release. The pH of CSF and brain interstitial fluid is lower than that of plasma (Davson and Segal 1969) and in the microenvironment between BCECs and astrocytes may be considerably lower still. ATP and other nucleotides are released from astrocytes and other brain cells (Neary et al. 1996; Guthrie et al. 1999; Montana et al. 2006) and could act as mediators of iron release from transferrin. Also, citrate, another such mediator, is released from astrocytes (Sonnewald et al. 1991) and is present in brain interstitial fluid in relatively high concentrations (Petroff et al. 1986). Hence, the microenvironment at the abluminal surface of BCECs is likely to be conducive to iron release from transferrin. The transferrin would mainly remain bound to its receptor which has a high affinity for apotransferrin at acidic pH (Morgan 1983) and would recycle to the plasma. The iron would pass into the brain interstitium to be bound by citrate, ascorbate (see Bradbury 1997) or the transferrin present in the interstitial fluid.
The areas for future investigations to verify this hypothesis will be to verify the presence of an acidic environment in the brain interstitium, and to prove that citrate in such conditions is capable of binding iron leading to its cellular uptake, although it should not be overlooked that ferric citrate indeed is a reliable cellular donor of non-transferrin bound iron at acidic pH in vitro (Sturrock et al. 1990).
Another clue to the role of astrocytes for transporting iron into the brain comes from studies of the newborn rodent brain in which iron transport is lower than at older developmental ages (Moos et al. 2006). In contrast to older aged rodents, BCECs of the newborn accumulate ferrous and ferric iron, which shows that they take up more iron from the circulation than is transported into the brain (Moos 1995; Cheepsunthorn et al. 1998; Moos et al. 2006). The significance of this iron storage could be to provide a reservoir needed for the robust proliferation of BCECs that takes place during the first postnatal week in the developing rodent brain (Morgan and Moos 2002). The BCECs of the newborn brain also differ from that of older age in that they have not yet formed fully intimate contacts with the end-feet of astrocytes (Xu and Ling 1994). Hence, it is tempting to speculate that astrocytes are needed to invest the abluminal surface of the BCECs to promote release of iron from transferrin through the action of low-molecular weight substances by the mechanisms suggested above, and that this process does not occur until BCECs have reached a state of maturity that allows astrocytes to form these intimate contacts.
Circulation of iron inside the brain
The brain interstitium
Provided molecules released by astrocytes capture iron transported through BCECs, the interstitium will contain non-transferrin bound iron complexed to smaller organic molecules like citrate, ATP and ascorbic acid (c.f. Bradbury 1997). The BCECs are likely to release ferric iron, whereas ferrous iron release is likely to occur by secretion from neurons and oligodendrocytes because of their expression of the iron exporter ferroportin (Burdo et al. 2001; Wu et al. 2004; Moos and Rosengren Nielsen 2006), although the oxidation state of iron transported by ferroportin is not yet firmly established. Supporting the idea of ferrous iron release, McKie et al. (2000) found that extracellular ceruloplasmin, a ferroxidase, enhanced ferroportin-mediated iron release from cells and a study in the hypoxic pig brain using microdialysis provided evidence that ferrous iron indeed is present extracellularly even in the non-hypoxic brain (Savman et al. 2005). The non-transferrin bound iron constituents are likely iron sources for all cells of the brain. Notably, however, ferrous iron carries the potential for causing damage to cell membranes via formation of iron-hydroxyl radicals that are highly toxic (Dunford 1987). In this context, it seems highly significant that the rat brain contains 40–100 times the concentration of the anti-oxidant ascorbic acid as that of plasma (Milby et al. 1982), which may act to prevent the toxicity of ferrous iron (Bradbury 1997). The brain does not form ascorbic acid, but, in some non-primate animals like mice and rats, it is synthesized in the liver and transported into the brain via a high-affinity ascorbic-acid transporter-dependent mechanism in the choroid plexus from where it may diffuse further into the brain and attenuate oxidative stress (Shin et al. 2005). It should be noted, however, that high concentrations of ascorbate may also lead to increased oxidative damage because of redox cycling of iron from the ferric to the ferrous state.
Beside from circulating in the brain interstitium bound to low-molecular weight constituents like ATP and citrate, ferric iron is thought to be mainly bound to transferrin. The most likely source for transferrin in the brain interstitium derives from diffusion from the ventricles (see below), as little transferrin is likely to be released from the abluminal surface of BCECs, as described above, or to be secreted from oligodendrocytes (De Arriba Zerpa et al. 2000; see below). The importance of transferrin for binding ferric iron within the bran interstitium should not be underestimated, as it has equilibrium constants for binding two molecules of ferric iron more than 1010 times higher than that of citrate binding a single molecule of ferric iron (Bradbury 1997). As mentioned above, however, citrate probably plays an important function for binding ferric iron released by BCECs, because the local concentration of citrate is high in the space between astrocytic end-feet and the abluminal side of the BCECs.
Another protein with potential to act as an extracellular iron carrier in the brain is lactoferrin (a.k.a. lactotransferrin). It is structurally closely related to transferrin but has an even higher affinity for ferric iron (Morgan 1981; Brock 1995). It is known to be synthesized in the mammary glands and neutrophil polymorphonuclear leukocytes and to circulate in low concentrations in blood plasma. It is not believed to play a significant role in the plasma transport of iron but, rather, to be involved in inflammatory reactions, at least in part because of anti-bacterial properties. Lactoferrin is present in CSF, in increased concentrations after cerebral bleeding or infarction (Terent et al. 1981). A lactoferrin receptor is present on brain microvessels, and there is evidence from in vitro studies that lactoferrin can traverse BCECs (Fillebeen et al. 1999). Hence, brain lactoferrin may be derived from the blood, although local synthesis probably also occurs (Siebert and Huang 1997). Lactoferrin and its receptor have been detected in neurons and glial cells in many different neurodegenerative diseases (Kawamata et al. 1993; Leveugle et al. 1994; Faucheux et al. 1995). It is likely that, in the brain, lactoferrin acts as a scavenger for iron released from degenerating and damaged cells thereby reducing the potentially toxic effects of such iron. However, it should be noted that there is no evidence that it functions as a quantitatively significant transporter of iron under normal conditions.
The ventricular system
In the ventricles and subarachnoid space, the circulation of the CSF provides a powerful sink that serves to drain the interstitial fluid of the brain into the CSF, which means that solutes present in the CSF are less likely to play a significant role for regulation of the concentration of iron, e.g. by diffusion upstream past the ependymal lining and into the brain interstitial space (cf. Szentistvanyi et al. 1984; Bradbury 1997). However, the conclusions that can be drawn from available data, including those of our own, are somewhat mixed. Radiolabeled iron-transferrin injected into the lateral ventricles, to simulate transferrin derived either from synthesis in choroid plexus epithelial cells (Tsutsumi et al. 1989) or transcytosis from plasma through the choroid plexus (Crowe and Morgan 1992), was detected in peripheral blood within minutes after injection (Moos and Morgan 1998a). In the brain, the radioactive iron mainly distributed to regions situated in the intimate vicinity of the ventricles and subarachnoid space even 24 h after injection (Moos and Morgan 1998a). Essentially the same observation was made when exogenous transferrin or OX26 was injected intraventricularly (Moos 2003). Conversely, endogenous albumin and transferrin could be detected in neurons without projections out of the CNS, hence excluding retrograde axonal transport from the periphery (Moos and Hoyer 1996). Together, these data provide reliable evidence that large molecules like albumin and transferrin are capable of diffusing into the brain interstitium from the ventricles, but that this diffusion process occurs very slowly and for technical reasons is very hard to document.
Iron uptake and export by neuronal cells
Neurons take up iron-transferrin injected into the brain (Moos and Morgan 1998b), and they contain both transferrin receptors and DMT1, which clearly indicates that they can take up iron-transferrin and transport it to endosomes from where iron is pumped into the cytoplasm (Figs. 3 and 6) (Burdo et al. 2001; Moos and Morgan 2004). The neurons potently up-regulate their expression of transferrin receptors in response to iron deficiency, which reflects the functioning of this receptor for the uptake of iron (Fig. 2) (Moos et al. 1998). This is in contrast to the expression of DMT1 that remains at unaltered levels in the brain following iron deficiency (Ke et al. 2005). The neurons probably also take up non-transferrin bound iron present in the brain interstitium, as they take up iron-citrate in culture conditions (Fig. 6) (Rosengren Nielsen and Moos, unpublished observation).
Neurons are thought to regulate their iron levels so that iron not used for metabolic purposes is released from the cells. The observation that neuronal ferroportin is virtually ubiquitously expressed in the brain suggests that iron-export mediated by ferroportin is a permanently active mechanism, which ensures iron-homeostasis inside the neuron. Iron is thought to undergo axonal and dendritic transport, and as ferroportin is found in the somata, axons, and dendrites of neurons, it probably plays an important role to regulate iron levels everywhere in the neuron (Fig. 7) (Moos and Rosengren Nielsen 2006). Neurons of some forebrain nuclei, however, also contain ferritin, showing that neurons are capable of storing iron (Hansen et al. 1999).
The capability to reduce ferric iron and transport it into the cytosol depends on the presence of molecules with ferric reductase activity. Vargas et al. (2003) found that choroid plexus and ependymal cells contain stromal cell-derived receptor 2, a molecule of the so-called b561 family with ferric reductase activity. This is the sole report published so far on this topic, and clearly more research is needed to evaluate whether neurons contain a ferric reductase. It should not be overlooked, however, that the expression of ferroportin may account for the export of residual iron from the neurons in the form of ferrous iron. At the present, there is no evidence of whether the function of the neuronal ferroportin requires the cooperative function of iron oxidases like ceruloplasmin or hephaestin. Ceruloplasmin is generally thought to be specifically expressed by astrocytes (Klomp et al. 1996), but a recent study provided evidence of neuronal pathology in ceruloplasmin null mice (Jeong and David 2006).
Iron uptake and transport in macro- and microglia
In the intact brain, astrocytes are devoid of transferrin receptors, suggesting that they take up iron by a mechanism that does not involve the transferrin receptor (Moos and Morgan 2004). Supporting this notion, astrocytes were shown to take up non-transferrin bound iron in vitro (Schipper et al. 1999). DMT1 probably is absent from astrocytes in vivo (Moos and Morgan 2004), although contrary observations have been made by other investigators (Huang et al. 2004, 2006). The export of iron from astrocytes is thought to involve the copper-containing protein ceruloplasmin that exhibits ferroxidase activity, which is capable of oxidizing ferrous iron to ferric iron and is present on astrocytes in a membrane-bound form (Fig. 6) (Patel and David 1997). Depletion of the ceruloplasmin gene in mice was recently shown to affect astrocytes in that they accumulated iron and showed signs of pathological changes (Jeong and David 2006). These data confirm the current hypothesis that ceruloplasmin is needed for export of iron from astrocytes (Klomp et al. 1996). The mutation of the ceruloplasmin gene leads to the unusual induction of ferritin expression by the affected astrocytes, a phenomenon that is otherwise restricted to neurons, oligodendrocytes, and microglia. Deletion of ceruloplasmin also affected some neurons of the cerebellum, and it was suggested that the failure of astrocytes to oxidize ferrous iron could affect the iron availability for the neurons, causing them to become iron-depleted either by reducing the concentration of iron in the brain extracellular space or by impairment of a direct iron-transfer between astrocytes and neurons (Jeong and David 2006). The basis for the latter hypothesis is that astrocytes form an important part of a so-called neurovascular unit formed between BCECs, astrocytes, and neurons (Fig. 4) (Abbott et al. 2006), and that astrocytes theoretically would take up iron directly from BCECs and direct it to neurons by means of intracellular transport. Increased expression of neuronal transferrin receptors is a hallmark of a state of iron deficiency (Fig. 2) (Moos et al. 1998). As this was not found in ceruloplasmin null mice, it is not justifiable yet to conclude that the neurons suffered from a loss of direct iron-supply from damaged astrocytes as the result of lack of ferroxidase activity, as recently suggested by Jeong and David (2006).
Similar to astrocytes, oligodendrocytes are likely to acquire iron by a mechanism devoid of the involvement of the transferrin receptor (Fig. 8). This could involve uptake of iron-citrate, which subsequently leads to incorporation of iron into transferrin synthesized by the oligodendrocyte (Bloch et al. 1985). Interestingly, evidence of transferrin secretion by oligodendrocytes is doubtful (De Arriba Zerpa et al. 2000). Possibly, transferrin in oligodendrocytes serves as a molecule for intracellular transport of iron along the extreme extensions, which wrap around several axons. Oligodendrocytes increase their content of iron and ferritin with increasing age (Benkovic and Connor 1993). This could reflect a deficiency in their capability to release iron, possibly because they gradually lose their capability to transfer iron back from the peripheral extentions to the soma. In contrast to astrocytes, oligodendrocytes contain ferroportin, indicating that they can export iron as part of their regulation of intracellular iron-homeostasis (Fig. 7) (Wu et al. 2004; Moos and Rosengren Nielsen 2006). It will be of interest to examine the expression pattern of ferroportin in oligodendrocytes of the aging brain to see if they increase their expression to compensate for the increasing level of iron or whether decreased expression may lead to iron accumulation in these cells.
Microglia originates from bone marrow cells of the myelo-monocytic lineage, and during the development of the brain they migrate into the brain from the circulation as monocytes, and then differentiate into quiescence microglia (Milligan et al. 1991). The migrating cells are ferritin-containing because of a high content of iron that is needed for production of free radicals as part of the so-called respiratory burst activity, but during their gradual transformation they lose their content of iron and corresponding expression of ferritin (Moos 1995; Cheepsunthorn et al. 1998). When they reach their resting and fully differentiated state, microglia rarely contain histological detectable levels of iron, transferrin, ferritin or other iron-related proteins such as transferrin receptors or DMT1. Moreover, microglia, in contrast to monocytes and macrophages, do not contain ferroportin (Moos, unpublished observation). Microglia take up non-transferrin bound iron in culture (Takeda et al. 1998).
Iron transport out of the brain
The quantity of iron in the brain increases with age but not as rapidly as would be expected from measurement of the rate of iron uptake from the blood at the different ages (Morgan 1999). Hence some iron must be exported. The major route is almost certainly via the CSF and its reabsorption back into the blood from the subarachnoid space. This was demonstrated in rats by injection of radiolabeled transferrin into the lateral cerebral ventricle (Moos and Morgan 1998a). The transferrin was largely rapidly reabsorbed into the blood stream but a small proportion slowly entered the brain parenchyma and was able to donate iron to the brain cells. Hence, it is likely that transferrin in the brain interstitial fluid could move in the reverse direction carrying with it iron derived from brain cells especially from cells which possess the iron exporter, ferroportin, and export this iron to the ventricles or subarachnoid space.
Within the CSF transferrin is fully saturated with iron (Bradbury 1997; Moos and Morgan 1998b). Nevertheless, the concentration of transferrin in CSF is very low so that the capacity to export iron by this route is limited. The CSF also contains lactoferrin, ferritin and non-protein-bound iron (Terent et al. 1981; Moos and Morgan 1998b), which represent additional means of iron export from the brain. The quantities of iron involved are again very small under normal circumstances but probably can increase considerably under pathological ones. Microglia and other phagocytic cells, which can enter the brain in inflammatory conditions and then exit the brain, are additional important mediators of iron export after cell death and intracerebral hemorrhage that increases the local iron concentration substantially (Hua et al. 2006).
Future research will address whether transferrin receptor-containing endosomes are transported to the abluminal surface of BCECs. Experiments should also be conducted to evaluate whether astrocytes interact with BCECs to promote the release of iron further into the brain either by secretion of iron-binding substances, which diffuse in the brain extracellular space or by intracellular transport from their end-feet via cellular processes. The investigations should also involve research on the mechanisms by which astrocytes respond to increasing concentrations of extracellular iron, and also to see if interactions between astrocytes and neurons are responsible for direct cellular iron-transfer.
Understanding the mechanisms involved in iron homeostasis in the brain as a whole and at the cellular level in neurons and glial cells is of utmost importance, as deficiency or excess of iron is probably intimately involved in the development of many functional and structural neurological abnormalities The mechanisms responsible for iron transport into the brain have been discussed above but those concerned with export from the brain, the other aspect of homeostasis, have so far received little attention and are poorly understood. At the cellular level the problem is no better resolved. How much of the iron taken up by cells is in a non-transferrin-bound form, possibly as iron-citrate or iron-ATP, and how much is transferrin-bound acquired by the function of transferrin receptors? As indicated above, the latter is probably important in neurons but not in glial cells.
Iron efflux from brain cells is also unresolved. The recent studies on the functions of ceruloplasmin implicate this iron oxidase in efflux from astrocytes but exactly how it functions in this process is uncertain. In particular it would be of great interest to see whether the iron export process also requires an iron transporter. In neurons and oligodendrocytes such a transporter, ferroportin, has been identified, but again questions arise as to whether its function requires the cooperative action of an iron oxidase such as ceruloplasmin or hephaestin.
Grazyna Hahn is thanked for excellent photographic assistance. The most recent results by the authors was generated by grant support from the Danish Medical Research Council, The Lundbeck Fund, The Carlsberg Fund, The Danish Parkinson’s Disease Fund, Direktør Jacob Madsens & Hustru Olga Madsen’s Fund, Fonden til Lægevidenskabens Fremme, and Lily Benthine Lunds Fond (TM), and the Medical Research Fund of Western Australia (EHM).