Iron and aluminum homeostasis in neural disorders.

The brain is the most compartmentalized organ. It is also highly aerobic. Because nerve cells grow but do not regenerate, the brain is the organ best suited for the accumulation of metabolic errors colocalized in specific areas of the brain over an extended period. Alzheimer's disease (AD) is primarily a neurological disorder of the elderly. It is suggested that this disorder results from the accumulation of such errors, and that AD onset aluminum and iron contribute to but do not necessarily initiate the onset of the disease. In vitro and in vivo evidence summarized here suggests that this is effected by interfering in the utilization of glucose and glucose-6-phosphate, and sequestration of iron by ferritin. beta-amyloid precusor proteins (beta-APPs) are normal components of the human brain and some other tissues. Proteolysis of these, presumably by serine proteases, generates a 39 to 42 amino acid long peptide, the alpha-amyloid (beta-AP). In AD brains, beta-AP aggregates into plaque, the hallmark of AD brains. Some of the alpha-APPs also contain a 56 amino acid long segment which inhibits serine proteases. We show that in vitro, at pH 6.5, aluminum activates beta-chymotrypsin 2-fold and makes it dramatically resistant to protease inhibitors such as bovine pancreatic trypsin inhibitor (bPTI) or its mimic present in the beta-amyloid precursor proteins (beta-APPs). Iron and oxygen are reported to favor cross-linking of beta-AP in vitro.(ABSTRACT TRUNCATED AT 250 WORDS)


Introduction
Alzheimer's disease (AD) is an age-associated dementia characterized by altered memory, cognition, and behavior. In AD, nerve cells in the specific areas of the brain degenerate. Theories contributing to AD include decreased cholinergic innervation defective protein synthesis, defective protein turnover, and aluminum toxicity (1). None are unequivocally proved or ruled out. Indeed, to date, the observed differences in normal and AD brain are quantitative rather than qualitative, but a search for a single biochemical event initiating AD has been unsuccessful. This is true even of the most current suspect, the amyloid plaques; the aggregates of the 3-amyloid peptides (,B-AP) generated by the proteolysis of the ,B-amyloid precursor pro-An adult brain weighs between 1.2 and 1.5 kg, but utilizes 20% of the oxygen consumed by the body. This highly aerobic tissue uses glucose as a sole or major source of energy. The requirement for glucose is very high-120 g/day, compared to 190 g for the entire body (3). The brain is also a highly compartmentalized organ; the use of glucose is area specific and stimuli dependent. For example, the areas of the brain using glucose in response to visual stimuli are distinct from those using glucose in response to auditory stimuli (4). Furthermore, unlike other tissues, brain cells do not multiply or regenerate. Thus, this vital organ is also well suited for the accumulation of toxins such as aluminum and also of metabolic errors which may result from such accumulation.
Solution chemistry of aluminum is quite complex. Martin has suggested that AI(III) is the biologically relevant species and, based on its affinity constants for citrate and transferrin as well as their circulating concentrations in vivo, these two compounds are primarily responsible for the chelation and transport of aluminum (5). Regardless of the vehicle used for transporting aluminum, it is generally agreed that the concentration of aluminum in AD brains is high and that the distribution of aluminum within the brain is nonuniform (6)(7)(8). Despite highly sensi-tive techniques for quantifying focal concentrations of brain aluminum, some researchers dispute the role of aluminum in AD and some dispute even the elevated levels of aluminum in AD brains (9). Nevertheless, several reports acknowledge that aluminum may play a role in AD (10).
We hypothesized that a critical mass of metabolic errors colocalized in specific areas of the brain is essential to produce a neurological disorder such as AD. Aluminum, a recognized neurotoxin, participates in formulating this critical mass by interfering in the metabolism of glucose, iron and proteolytic processing of P-APPs. Evidence obtained in our laboratory supports this hypothesis. It is summarized below.

Aluminum and Glucose Metabolism
Glucose enters glycolysis as glucose-6phosphate (G6P). This reaction, catalyzed by hexokinase, requires ATP and Mg(II). In the brain, the enzyme is membrane bound and the activity is latent. The activation requires its release from the membrane. This then is the first step in glucose consumption. Aluminum (III) binds to ATP 10 times more tightly than does Mg(II). Thus, in vitro, the concentration of aluminum as low as 160 nM inhibits hex-Environmental Health Perspectives Preincubation (min) Figure 1. Effects of aluminum on human and pig brain G6PD isozymes and human erythrocyte 6-phosphogluconate dehydrogenase. The isozymes were incubated with 5 pM AIC13 in 25 mM HEPES, pH 7.0, at 23°C. At different times, an aliquot was withdrawn and assayed for enzyme activities. Data are percentages of values for the controls. The specific activities (pmole NADPH formed/mg of protein/min) of various preparations were as follows: human G6PD isozyme 1, 210; human G6PD isozyme 11, 52; pig G6PD isozyme 1, 50; and pig G6PD isozyme 11, 201. The specific activity of 6-phosphogluconate dehydrogenase was as follows: human isozyme 11, 21; pig isozyme 11, 51; and human erythrocyte, 48. (A): G6PD activity of human (*) and pig (o) brain isozyme 1. (B G6PD activity of human (.) and pig (o) brain isozyme 11. Open and solid squares indicate 6-phosphogluconate dehydrogenase activity of human and pig brain isozyme 11, respectively. Triangles indicate 6-phosphogluconate dehydrogenase activity from human erythrocytes ( 14).
okinase (11). In view of this, one would anticipate that Al(III) would inhibit all ATP-dependent reactions. However, such is not the case because glucokinase and phosphofructokinase are unaffected by Al(III), whereas pyruvate kinase is activated (12). In the brain, 80 to 85% of the glucose-6-phosphate is used in glycolysis, and the remaining 15 to 20% is used by the hexosemonophosphate shunt which generates T ME (min) Figure 2. Reverse-phase HPLC of human brain ferritin subunits. One mg of brain ferritin in 0.5 ml of water was dissociated by incubation with acetic acid (67%), thioglycolic acid (5%) in a final volume of 0.86 ml on ice for 16 hr followed by dialysis against 0.1% trifluroacetic acid (TFA) and lyopholized, resuspended in 0.1% TFA, and injected onto a Vydac C18 reverse-phase column connected to a Waters dual pump HPLC. The solvent system employed was solvent A-0.1% TFA in water; solvent 8-0.1% TFA in 80% acetonitrite. Flow rate = 1 ml/min. Eluted fractions were monitored at 220 nm. Fractions were pooled, lyopholized and rechromatographed. They ran true (35).
ribose-5-phosphate, CO2, and two molecules of NADPH. Four enzymes that participate in this process are G-6-P dehydrogenase (G6PD), lactonase, 6-Pgluconate dehydrogenase, and phosphopentose isomerase. G6PD from the brain is a tetramer composed of four identical molecular weight (57,000) subunits but two types of amino acid sequences. Therefore, electrophorograms of the homogenates of brain from human or pig show five bands of enzyme activity. Two of these have been purified to near homogeneity from pig and human brain (13). Isozyme I has only G6PD activity, whereas isozyme II also has 6-phosphogluconate dehydrogenase activity. Significantly, the G6PD activity of both isozymes were inhibited by Al(III) (Figure 1). The inactive protein bound a total of four molecules of aluminum, one per subunit. The circular diachroism studies showed that the aluminum-enzyme complex was more randomly coiled than its native counterpart. Bound metal ion was nondialyzable, even in the presence of EDTA. However, sodium fluoride or transferrin removed the bound metal ion and restored full activity (KDAI = 2-3 1M) (14). Thus, unlike hexokinase, aluminum bound to the protein and produced an inactive enzyme.
In the brain, the shunt pathway is associated almost exclusively with myelineated tissue, and the activity varies with the degree of myelineation (15). Because the synthesis of fatty acids requires NADPH, this close association is functionally advantageous.
The rate-limiting step in glycolysis is the modulation of phosphofructokinase and fructose-1,6-bisphosphatase activities by fructose-1,6-bisphosphate, ATP, and fructose-2,6-bisphosphate. However, recent studies have shown that in the brain, the initial modulator is ribose-1,5-bisphosphate (16). Reduced supply of ribose-5phosphate resulting from the aluminummediated reduction in production of NADPH and ribose-5-phosphate would produce a focal deficit of the key component essential for myelin synthesis as well as for regulation of glycolysis.
Experimental metal toxicity can be produced by injecting very high concentrations of the toxicant. While such studies save time, physiologically relevant metal toxicity results from prolonged exposure to a chronic level of the toxicant. We reported that brain homogenates of rats fed 100 pM AlCl 6H20 in the drinking water for 1 year had about a 25% reduced activity of hexokinase and G6PD (17). Examination of the sections of the brains of rats similarly exposed to AlC13.6H20 for two years also showed between 15% and 20% reduction of glucose metabolism in several areas of the brain (18). The results approached statistical significance only for the ventral palladium and temporal cortex. Although more animals were essential to establish a statistically significant difference, the observed reduction is compatible with the reported aluminum-mediated impaired metabolism of glucose by rat brains in vitro (19) and reduced hexokinase in AD brains (20).

Iron and Brain Metabolism
In higher organisms, iron bound to transferrin and stored in ferritin accounts for more than 90% of the total nonheme iron (21). Both proteins also sequester other metal ions, including aluminum (22). Intracellular concentration of iron regulates the synthesis of ferritin and the stability of the transferrin receptors. When the cellular concentration of iron is high, the available ferritin mRNA is utilized more efficiently and at lower concentrations of iron, the transferrin receptor mRNA is more stable; and, therefore, more transferrin receptors are available for iron transport. This regulation is due to a conserved sequence of 28 nucleotides at the 5'untranslated region (5'-UTR) of ferritin mRNA and at the 3'-UTR of the transferrin receptor (23  nucleotides at the 5'-UTR which are absent in the liver message. In the brain, the larger message (1.4 kb) is nine times more abundant than the smaller, liver-type message. In the liver, the smaller message is three times more abundant than the larger message.
Transferrin(s), a family of glycoproteins of mw 80,000 is a single polypeptide chain. It has two metal-binding sites, one each at the N-and C-terminal end. It is the major serum protein that transports iron. It also binds other metal ions including aluminum (24,25). The stability constants for the Nand C-sites are not identical. In vivo, even under extreme iron overload, only 30% of the total iron binding sites are occupied by iron (21). Because a convenient radioactive isotope of aluminum is not available in vitro, radioactive gallium, 67Ga(III), is often used as an aluminum mimic and the data extrapolated to in vivo situations. Farrer et al. showed that transferrin from the sera of AD patients bound less gallium (and, therefore, less aluminum by inference) than the age-matched controls. They suggested that in AD patients transferrin binds less aluminum and, therefore, more serum aluminum is available for transport to the brain (26). This potentially attractive idea is controversial because the same observed reduced gallium binding would result if the presumed iron-free sites were actually occupied by aluminum.
Furthermore, the more recent data suggests that in vivo aluminum and gallium are not transported by the same mechanism (27). In this respect, the data by Connor et al. are noteworthy (28). These workers showed that the access of aluminum to various areas of the brain is probably via the transferrin-mediated receptor system; they suggested that this system may also regulate the transport of other metal ions. They also noted that although transferrin and its receptors are present throughout the brain, their distribution in the central nervous system is not uniform (28). These observations further underscore the possibility of a colocalization of metal ions producing metabolic errors leading to neurological disorders. Transferrin in the serum delivers its metal ions to the brain via a transferrinreceptor mediated system. The olfactory system seems to have its own transferrinlike protein, thus gaining direct access to the brain (DP Perl, personal communication). In the brain, iron and presumably other metal ions are sequestered by ferritin.

Aluminum, Iron, and Ferritin in the Brain
In humans, most of the iron is recycled and very little is excreted. Therefore, the concentration of total iron in the human body increases with age (29). In the male, it increases from about 300 mg at age 20 to 25 years to about 1800 mg at age 80 to 90 years. In the female, the levels remains at about 300 mg until the years of age 50 or until menopause, and increase to about 1200 by the age of 80 to 90 years. Significantly, a fetal brain has very little iron or ferritin which stores it. In the adult brain, a third of the total nonheme iron is stored in ferritin (30). This protein is composed of a total of 24 chains, heavy (H) and light (L), of molecular weights of 21,000 and 19,000, respectively. A fully saturated ferritin can store in its protein shell up to 4500 molecules of iron as Fe(III) hydroxyphosphate. Isoferritins varying in subunit composition exist in different tissues. Ferritin binds several metal ions in vitro and in vivo. Indeed, ferritin aluminum complex can be prepared in vitro, and has been isolated from the brains of two AD patients and from rats chronically exposed for a year to 100 pM AlC13.6H20 in their drinking water (31). Our subsequent studies (unpublished data) showed that ferritin isolated from two AD and one normal brain had between 2 and 4 moles of aluminum bound per mole of protein, and one AD brain ferritin had 12.8 molecules of iron per mole of ferritin. In all AD brains, the concentration of ferritin was consistently higher than in the controls. Cochran and Chawtur (32) also observed similar binding of aluminum to ferritin in vitro. In contrast, Deadman et al. (33) could not observe any difference in the aluminum bound to ferritin in normal and AD brains, but consistent with earlier observations (31) reported elevated levels of ferritin in AD brains and they produced aluminum ferritin complexes in vitro. We observed that in vitro more aluminum bound to holoferritin than to apoferritin and that aluminum reduced the rate of iron uptake by ferritin (34).
To determine whether human brain ferritin is chemically distinct from that found in the liver, we undertook further characterization of human brain ferritin and its subunits. SDS-PAGE of ferritin from normal or AD brain showed two bands, H = 70%, L = 30%. However, isoelectrofocusing on native ferritin showed several bands of isoferritins (31). HPLC chromatography (35) of human brain ferritin from normal or AD tissues yielded a cluster of about five heavy chains and predominantly only one light chain (Figure 2). This appeared to have offered an explanation for several isoforms of brain ferritin resolving after isoelectrofocusing of the native protein. To determine the difference, if any, in the amino acid sequences of H chains, we used a cDNA clone for ferritin heavy chain from the liver and screened the human brain cDNA libraries from 11 week old fetal brain and from adult normal and AD brains. The cDNAs isolated from these sources were sequenced by the "dideoxynucleotide" method. The preliminary results schematically represented in Figure 3  tion. Standard assay conditions (39 were used, except 38 mM PIPES buffer, pH 6.5, was employed, the methanol concentration was reduced to 3%, and calcium chloride was absent from the control assays. Fresh solutions of Sigma enzyme were made daily, calibrated for activity, and kept in dilute HCI on ice to prevent autolysis. One-min preincubations of metal and enzyme were shown to be sufficient for full activation. The synthetic substrate, benzoyl-tyrosine ethyl ester (BTEE), was not precipitated or hydrolyzed by the addition of aluminum trichloride alone. Different orders of addition did not affect activation (39. * none of these cDNAs were full length at the 5' end. * Northern blot analysis of the RNA (poly A+) preparations from human liver, normal adult and AD brain and 11-week-old fetal human brains revealed the presence of two transcripts of 1.4 kb and 1.1 kb for ferritin. * The larger, 1.4 kb, is most abundantly expressed in the brain, while its level of expression in the liver is 10 times lower. * the two transcripts are also expressed differentially in other human tissues like kidney, lungs, pancreas, heart, placenta, and skeletal muscle. * Comparison of the sequence showed that the small transcript (1.1 kb) from the brain was identical to that reported in the liver. *

Aluminum and Plaque Formation
Presence of plaques in an AD brain first discovered by Alzheimer in 1907 (37) has survived the test of time and is now considered a histopathological benchmark of the AD brain. The plaques are aggregates of a 39 to-42-amino-acid-long peptide (J-AP) produced by the proteolysis of a family of a-amyloid precursor proteins (P-APPs).
Serine proteases are suggested to be involved in this processing. Part of the P-A4 segment of )-APP is lodged in the plasma membrane that separates cytoplasm from the extracellular region ( Figure 4). These proteins are also found in other nor- (i) activation by aluminum (i) free radical mediated (Fe2+ +02) cross-linking at pH 6.5 L {ii) damage to the membranes and oxidative (ii) resistance to inhibition modification to produce easily by protease inhibitors proteolysable substrate mal or AD tissues. The 2 f-APPs that contain 750 and 771 amino acids also contain a 56-amino-acid-long segment whose sequence is 60% homologous to the wellknown bovine pancreatic trypsin inhibitor (bPTI). This led to the suggestion that deregulation of the interaction between the serine protease(s) and their substrates (I-APP) leads to P-AP and their subsequent aggregation into plaques. Although any physiological role of 3-APPs or ,B-AP is unknown, they are believed to be involved in the etiology ofAD (2). At physiological pH, calcium (Ca) ions (20 mM) stabilize serine proteases against autolysis. We observed that in vitro at pH 6.5, 200 PM AlCl3.6H20 (total concentration) activated a-chymotrypsin 2-fold ( Figure 5). The activation was immediate and reversible by dilution. One mole of aluminum bound per mole of enzyme (Hill plot) and neither Cl, nor Na,+ nor Fe(III) affected a-chymotrypsin activity (data not shown). Ca(II) also activated the enzyme, yet the KM(app) was 10-fold higher. Metal buffers are routinely used to assure a steady supply of a free metal ion. In the presence of an appropriate concentration of chela-Environment Al (inert) ta tors such as ethylenediaminetetracetic acid or nitrilotriacetate the concentration of A1(III) can be varied between 10 9 to 103M (38). However, when these chelators were employed in the assay system, the activating effect of aluminum was abolished (39). This strongly suggests, although it does not prove, that hydroxylated polynuclear species could be responsible for the activation of the enzyme. Thus, high focal concentrations of probably polynuclear aluminum deposits discovered in the brains of patients with AD may be physiologically significant (39). Figure 6 shows the effect of bPTI or the bPTI-like segment of P-APPs on the activity of a-chymotrypsin and compares it with that of the aluminum-activated enzyme. As seen, aluminum dramatically protected a-chymotrypsin against the inhibition by bPTI or the BX-9 fusion protein (41) which contains the bPTI-like segment of P-APP inserted into P-galactosidase. Aluminum also activated trypsin by 140% with a similar K(app) but produced only a 15-fold decrease in the binding of bPTI (data not shown). In addition to aamyloid peptide(s), AD plaques also contain a-i-antichymotrypsin. Aluminum protected against inhibition by a-iantichymotrypsin, but the change in K, was only about 10-fold (40).
The above results were obtained with synthetic substrates. To verify whether the results hold true for natural substrates, we studied limited proteolysis of transferrin by a-chymotrypsin in the presence of AICI .6H20. Indeed, aluminum produced a 2-?old increase in the rate of proteolysis of transferrin (KMAl(app)-2.1 x 104-M).
Finally, 100,000 g supernatants of brain homogenates at pH 6.5 also showed a 2.8fold increase in the rate of hydrolysis of BTEE in the presence of 5 x 10-4 M AIC13.6H20 (39).

Role of Iron in Plaque Formation
Interaction of metal ions such as Fe(II) with oxygen produce oxygenated free radicals (42). These radicals oxidatively modify proteins and make them more susceptible to proteolysis (43), damage DNA, and peroxidize lipids (42). Aluminum enhances the lipid peroxidation of erythrocyte membranes initiated by iron. The rate of this lipid peroxidation is greater at acidic pH (44). Acidic pH also accelerates the rate of the release of iron from ferritin (21) and favor aggregation of 1-AP (45). More recently Dyrks et al. (46) reported that in vitro, the cross-linking of f-AP, which causes the formation of insoluble plaques, is enhanced by oxygenated free radicals produced in the presence of a metal ion such as iron. Localized acidosis in the brain has been reported during ischemia, and hypoxia, and in AD patients (47)(48)(49). Figure 7 shows how these factors can contribute to plaque formation.
The role of aluminum and iron in deregulating brain metabolism discussed above is summarized in Figure 8. Almost all the changes observed in AD brain are quantitative and not qualitative. Thus, although the presence of increased aluminum in AD brain is generally confirmed, in one recent report its presence in AD (senile) plaques has been attributed to experimental artifact (50). Similarly, the presence of an increased concentration of ferritin in AD brain reported earlier (31,33) and its localization in neuritic plaques of AD patients (51) has been con-Volume 102, Supplement 3, September 1994 211 firmed, but its association with Al was not seen in several AD brains (33). Most reports show that familial cases of AD once considered due to genetic mutations in the AD-gene in chromosome 21 may also arise from mutations in chromosomes 19 and 14 (52). The amyloid plaque unique for AD brains is observed by histopathologic examination of the brains of diseased patients and, therefore, has been considered by many as the end result, rather than the cause, of AD. The P-APPs exist in normal individuals also, but their biological function is unknown. Finally, even in the familial AD, identical afflicted twins are not aflicted by AD simultaneously. Similarly, not every individual exposed to elevated levels of environmental aluminum suffers from AD. It therefore appears that AD probably results from the iron-and aluminum-mediated colocalization of metabolic errors accumulating in specific areas of the brain (Figure 7). Clearly, the reactions shown do not have to occur in the order given. For example, large concentrations of i-AP can be produced with an active protease and sufficiently high concentrations of ,-APPs. Aluminum, iron, oxygen, and an acidic pH can help achieve that level of P-AP faster and at lower concentration of P-APPs by activating the protease, and permitting its function despite the presence of the PPTI-like inhibitor of ,-APPs and crosslinking P-AP. Similarly, the role of aluminum in deregulating various metabolic reactions in Figure 8 (12) as well as numerous others (10) could occur only if the local con-centrations of aluminum and iron are sufficiently high. The same is true for the ill effects of unsequestered iron.
Although the data presented above and summarized in Figures 7 and 8 have been obtained in vitro and need to be verified in vivo, recent reports seem to support the postulate that indeed iron and aluminum contribute to the onset of AD. Accordingly, epidemiologic studies of McLachlan et al. (53) reported partial relief against AD by desferrioxamine. This compound was first used against dialysis dementia caused by aluminum toxicity. It is an effective chelator for aluminum as well as iron. Thus, the observed relief against AD (53) may well be due to the removal of both these neurotoxins.