Overexpression of Alzheimer's disease amyloid-beta opposes the age-dependent elevations of brain copper and iron.

Increased brain metal levels have been associated with normal aging and a variety of diseases, including Alzheimer's disease (AD). Copper and iron levels both show marked increases with age and may adversely interact with the amyloid-beta (Abeta) peptide causing its aggregation and the production of neurotoxic hydrogen peroxide (H(2)O(2)), contributing to the pathogenesis of AD. Amyloid precursor protein (APP) possesses copper/zinc binding sites in its amino-terminal domain and in the Abeta domain. Here we demonstrate that overexpression of the carboxyl-terminal fragment of APP, containing Abeta, results in significantly reduced copper and iron levels in transgenic mouse brain, while overexpression of the APP in Tg2576 transgenic mice results in significantly reduced copper, but not iron, levels prior to the appearance of amyloid neuropathology and throughout the lifespan of the mouse. Concomitant increases in brain manganese levels were observed with both transgenic strains. These findings, complemented by our previous findings of elevated copper levels in APP knock-out mice, support roles for APP and Abeta in physiological metal regulation.

Metals have been postulated to play a role in the pathogenesis of AD 1 (1,2). Copper, zinc, and iron are concentrated in and around amyloid plaques in AD brain (3), and high levels of zinc (4) and iron (5) have been reported in the amyloid plaques of the Tg2576 mouse model for AD. Numerous reports have also demonstrated transition metal imbalances in AD brain, such as decreased copper, and increased iron, zinc, and manganese (6 -11).
Studies in mice and humans show that iron and copper levels increase with normal aging in several tissues, including brain (12)(13)(14)(15)(16)(17), while zinc levels either remain unchanged or show a slight decrease (14,16,18,19). Therefore, a breakdown of metal regulation could be an inevitable consequence of aging.
Copper and iron are redox active metals that play important catalytic roles in many enzymes. Their levels must be strictly regulated to prevent aberrant reactive oxygen species production resulting in cellular toxicity. AD brain exhibits marked oxidative damage of proteins, lipids, and nucleic acids (20,21).
APP possesses a copper binding site in its NH 2 -terminal cysteine-rich domain, which reduces Cu 2ϩ to Cu 1ϩ (22). APP also has a zinc binding site, which is believed to have a structural role (23). APP and amyloid precursor-like protein 2 (APLP2) knock-out mice show specific elevations in brain and liver copper levels (24), which suggests that APP has a role in copper homeostasis.
A␤, a product of APP proteolytic processing, accumulates in the neocortex in AD. This peptide also possesses selective high and low affinity Cu 2ϩ and Zn 2ϩ binding sites. A␤ reduces Cu 2ϩ to Cu 1ϩ and Fe 3ϩ to Fe 2ϩ , catalyzing the O 2 -dependent production of H 2 O 2 (25). This interaction of A␤ with copper mediates toxicity (26), while zinc inhibits A␤-mediated H 2 O 2 production and toxicity (27). Interaction with copper, zinc, or iron mediates the aggregation of A␤ (28 -30). Chelation of metal ions reverses the aggregation of synthetic A␤ peptide and dissolves amyloid in post-mortem human brain specimens (29,31,32). Treatment of the Tg2576 transgenic mouse model for AD with clioquinol, an orally bioavailable metal chelator induced a marked inhibition of cortical amyloid accumulation (33).
To investigate the effects of aging and APP and/or A␤ overexpression on metal levels, we measured copper, zinc, iron, copper, and manganese levels in the brains of normal and transgenic (Tg) mice across the majority of their lifespan (2.8 -18 months (mo)). We utilized transgenic mice overexpressing human APP695.K670N-M671L (Swedish mutation) (Tg2576) as well as two lines of mice expressing the carboxylterminal 100 residues of APP (C100), with and without the familial AD mutation V717F (TgC100.V717F and TgC100.wt, respectively). TgC100 mice express human A␤ at lower levels than the Tg2576 line, and display no A␤ accumulation nor amyloid plaques up to 20 months (34). These mice provide a model to study the effect of increased human A␤ production without holoAPP overexpression. The TgC100.V717F line produces relatively more A␤x-42, which is of interest, since A␤1-42 binds copper with much greater affinity than A␤1-40 (29,30), and is more readily precipitated (29), more redox active, and more toxic (26) than A␤1-40 when bound to copper.
Here we show age-related increases in copper, iron, and cobalt levels in the brains of all mouse lines studied. These increases may contribute to the age-dependent formation of amyloid and oxidative damage in Tg2576 mice, and possibly also in AD brain. We also show that APP and A␤ expression modulates metal levels, particularly copper, in transgenic mouse brain. These data suggest that the corrupted metabolism of A␤ in AD may cause severe perturbances of essential metal homeostasis. These imbalances may contribute to the neurodegenerative phenotype.

EXPERIMENTAL PROCEDURES
Mice-All mice were housed according to standard animal care protocols and fed standard laboratory chow and tap water ad libitum. We maintained the Tg2576 colony (35) by crossing Tg2576 males with C57BL6/SJL F1 females and determined transgene status by PCR of tail DNA, using primers as described previously (36). NTg littermates were used as controls (referred to as BL6/SJL). We analyzed mice at 2.8 months (2.81 Ϯ 0.002 months), 3.6 months (3.59 Ϯ 0.002 months), 11 months (11.14 Ϯ 0.01 months), and 18 months (18.21 Ϯ 0.03 months). Transgenic mouse lines TgC100.V717F and TgC100.wt were generated as described previously (34) and bred to homozygosity in a C57BL6/ DBA background. NTg controls for the C100 lines were of the same genetic background (C57BL6/DBA F2) (referred to as BL6/DBA). We studied these mice at 2.8 months (2.76 Ϯ 0.02 months), 8 months (7.8 Ϯ 0.1 months), and 18 months (17.6 Ϯ 0.1 months). All groups contained similar numbers of males and females.
Preparation of Brain Tissue-To minimize metal contamination of samples, we presoaked all tubes and equipment in 1% nitric acid and rinsed them in distilled water prior to use. We sacrificed mice by anesthetization with halothane, followed by transcardial perfusion with phosphate-buffered saline at 100 -120 mm Hg until the perfusate ran clear. We then dissected brains to remove olfactory bulb, cerebellum, and brain stem; weighed the remaining wet brain tissue; and snap-froze samples and stored them at Ϫ80°C until use.
Metal Analysis-We dissolved the freeze-dried brains overnight in 0.6 ml of concentrated HNO 3 (Aristar, BDH), followed by heating to 80°C for 20 min, and allowed them to cool to room temperature. To dissolve lipid components, we added 0.6 ml of H 2 O 2 , and once effervescing had ceased (approximately 30 min) we heated samples to 70°C for 15 min and allowed them to cool. We diluted samples in triplicate by 1/51 in 1% HNO 3 (60-l sample plus 3 ml of 1% HNO 3 ) and measured metal levels by inductively coupled plasma mass spectrometry (ICP-MS) with an Ultramass 700 (Varian, Victoria, Australia) in peak-hopping mode with spacing at 0.100 atomic mass unit, 1 point per peak, 50 scans per replicate, and 3 non-consecutive replicates per sample. Plasma flow was 15 liter/min with an auxiliary flow 1.5 liter/min. R F power was 1.2 kilowatts. Each sample was introduced using a glass nebulizer at a flow of 0.88 liter/min. We calibrated the instrument using a 1% HNO 3 mixed calibration standard (Merck Pty. Ltd.) containing 10, 50, and 100 ppb of all metals measured in 1% HNO 3 .
Statistical Analysis-We diluted and measured each brain sample in triplicate and used the average value for analysis. The metal values presented are the mean ng or g/g wet weight of the original prefrozen brain sample. Error bars represent S.E. of the mean of each group of mouse brains analyzed. We performed statistical analyses using Statistica TM for the Macintosh (StatSoft TM ). To determine whether metal levels changed significantly with age within each mouse line, and whether this differed for males and females, we performed two-way analysis of variance (ANOVA) on data from all age groups, with age group and sex as independent variables. If a significant age-sex interaction was found, males and females were re-analyzed separately. To determine whether metal levels in each age group were significantly different from the youngest age group (2.8 months), we performed post-hoc Scheffe tests. To determine whether brain metal levels in each line of transgenic mouse differed from their corresponding NTg controls, and whether this differed between males and females, we performed two-way ANOVA on metal level data from the two mouse lines in all age groups, including sex and age group as independent variables. If a significant mouse type-sex interaction was found, males and females were analyzed separately. Significant differences between transgenic and NTg mice in each age group individually were determined by applying simple contrasts to the data. Average percentage changes in metal levels due to transgene expression represent the average of the percentage differences in each age group Ϯ S.E., hence giving equal weighting to data in each age group.

RESULTS
Normal Age-related Changes in Brain Metal Levels-We first examined how brain metal levels are affected by age in both strains of non-transgenic mouse. Two-way ANOVA of metal levels in both BL6/SJL and BL6/DBA mouse lines, with age group and sex as independent variables, revealed marked and significant age dependent increases (p Ͻ 0.001) in copper, iron, and cobalt levels ( Fig. 1). No significant age-sex interaction was found for any of these metal level increases, indicating that the effects of age on brain copper, iron, and cobalt levels is similar for both males and females. In comparison with the youngest (2.8 mo) group, copper, iron, and cobalt levels in the BL6/SJL line had not changed significantly by 3.4 months, but by 11 and 18 months, significant increases were observed in each metal (p Ͻ 0.001) (Fig. 1A). Copper levels increased by 46% from 5.1 Ϯ 0.1 g/g (wet weight) at 2.8 months to 7.5 Ϯ 0.2 g/g at 18 months; iron levels increased 51% from 17.3 Ϯ 0.3 g/g at 2.8 months to 26.1 Ϯ 0.7 g/g at 18 months; and copper levels increased 66% from 11.6 Ϯ 0.3 ng/g at 2.8 months to 19.3 Ϯ 0.7 ng/g at 18 months. Zinc levels showed no significant overall change with age, but a significant age sex interaction was found (p Ͻ 0.01). When male and female zinc levels were analyzed separately, we found that males displayed a relatively small increase (10%, p Ͻ 0.01) in brain zinc levels across the age groups from 20.5 g/g at 2.8 months to 22.5 g/g at 18 months. Females zinc levels, however, remained between 20.1 and 21.1 g/g in all age groups measured. Manganese levels showed no significant overall change with age, and no significant age-sex interaction, with levels remaining between 0.44 and 0.46 g/g in all age groups measured.
In the BL6/DBA line (Fig. 1B), copper levels increased by 34% from 4.9 Ϯ 0.1 g/g at 2.8 months to 6.5 Ϯ 0.1 g/g at 18 months; iron levels increased 34% from 16.5 Ϯ 0.2 g/g at 2.8 months to 21.9 Ϯ 0.6 g/g at 18 months; and copper levels increased 41% from 12.1 Ϯ 0.2 ng/g at 2.8 months to 17.6 Ϯ 0.7 ng/g at 18 months. Manganese levels, in contrast to the BL6/ SJL strain, showed a significant overall decrease with age (p Ͻ 0.01), dropping by 14% from 0.51 Ϯ 0.02 g/g at 2.8 months to 0.44 Ϯ 0.01 g/g at 18 months (p Ͻ 0.001). No significant age-sex interaction was found for any of these metal level changes. Zinc levels showed no significant overall change with age, but a significant age-sex interaction was found (p Ͻ 0.05). Separate analysis of male and female zinc levels revealed no more than a 5% fluctuation in zinc levels across the three age FIG. 1. Age-related changes in brain metal levels. Metal levels were measured in BL6/SJL and BL6/DBA mouse lines at several ages from 2.8 to 18 months. Percentage change with respect to 2.8-month age groups is presented for each metal. Significance of the effect of age on metal levels was determined by two-way ANOVA across all age groups. ***, p Ͻ 0.001; **, p Ͻ 0.01; *, p Ͻ 0.05. Error bars represent S.E. of the mean. A significant age group-sex interaction was found in zinc levels. Therefore zinc data are presented separately for males and females. groups in both males and females. These changes reached significance only in females (p Ͻ 0.05), where a 5% drop in zinc levels was observed from 2.8 to 8 months (p Ͻ 0.05), whereas 18-month zinc levels were unchanged from 2.8-month levels. Zinc levels remained between 19.0 Ϯ 0.2 and 20.1 Ϯ 0.6 g/g in males and females in all age groups analyzed. The three Tg mouse lines (Tg2576, TgC100.wt, and TgC100.V717F) displayed the same direction of age-related changes in copper, iron, cobalt, and manganese levels as their corresponding NTg background controls, but exhibited transgene-related adjustments in metal levels.
The Effect of APP Overexpression on Metal Levels in Tg2576 Mouse Brain-Tg2576 mice express relatively stable human APP695.swe levels across the lifespan, but human A␤ levels in their brains increase exponentially with age, resulting in amyloid plaques by around 10 -12 months (35,37). To determine the effect of APP695.swe overexpression and A␤ accumulation on brain metal homeostasis, we compared brain metal levels between Tg2576 and BL6/SJL NTg littermates at 2.8, 3.6, 11, and 18 months. Two-way ANOVA of metal levels in Tg2576 compared with BL6/SJL brain, with age group and sex as independent variables, revealed a significant overall reduction in copper levels by 14 Ϯ 1% in Tg2576 compared with BL6/SJL brain (p Ͻ 0.001) (Fig. 2). This reduction showed no dependence on sex, but a significant age group effect was observed (p Ͻ 0.05). Significantly reduced (p Ͻ 0.001) copper levels were observed in each age group, and this effect increased with increasing age (11% at 2.8 months, 14% at 3.6 months, 14% at 11 months, and 16% at 18 months). We also found a small but significant reduction in zinc levels (4 Ϯ 1%, p Ͻ 0.001) in Tg2576 compared with BL6/SJL brains. While the percentage reduction in zinc levels was relatively small compared with the percentage reduction in copper levels, the average absolute decreases in zinc and copper levels were identical (both 13 nmol/g). In contrast, manganese levels were significantly increased in Tg2576 compared with BL6/SJL brain, by 5 Ϯ 3% (p Ͻ 0.001). Neither of these metal level changes showed any significant dependence on age or sex. Iron levels showed no significant overall difference between Tg2576 and BL6/SJL groups when males and females were analyzed together; however, significant mouse type-sex (p Ͻ 0.01) and mouse type-age (p Ͻ 0.05) interactions were found. Separate analysis of males and females revealed a significant change in iron levels in males that was dependent on age (p Ͻ 0.001), a significant reduction (13%, p Ͻ 0.001) was observed in the 11-month age group, but no other age groups. Females, in contrast, showed no significant overall change in iron levels due to APP695.swe expression; however, a significant effect of age group was observed (p Ͻ 0.05), whereby iron levels were significantly increased (14%, p Ͻ 0.01) only in the oldest (18 month) age group compared with age-matched BL6/SJL controls. Cobalt levels showed no significant overall difference between the Tg2576 and BL6/SJL groups and no significant effect of sex or age.
The Effect of A␤ Overexpression on Metal Levels in TgC100 Mouse Brain-The alterations to metal levels in the Tg2576 brain could be attributed either to the metal binding sites in the APP NH 2 -terminal domain or to the metal binding sites on A␤. To test whether A␤ expression alone could modify metal levels, we also measured metal levels in the brains of the TgC100.V717F (Fig. 3) and TgC100.wt (Fig. 4) mouse lines at 2.8, 8, and 18 months and compared them with BL6/DBA NTg controls.
We found significant overall reductions in copper and iron levels in the TgC100.V717F line compared with BL6/DBA controls (p Ͻ 0.001) that showed no dependence on sex. Significantly reduced copper and iron levels were found in each age group (p Ͻ 0.001), and only in the case of iron was a significant effect of age observed (p Ͻ 0.05), whereby the decrease was FIG. 2. Effect of APP overexpression on brain metal levels. Metal levels were measured in Tg2576 and NTg brain at 2.8 months (n ϭ 22 and 25 respectively), 3.6 months (n ϭ 21 and 24, respectively), 11 months (n ϭ 12 and 19, respectively), and 18 months (n ϭ 12 and 12, respectively) by ICP-MS. Each group contained similar numbers of males and females. Metal levels are given as g/g or ng/g wet brain weight. ANOVA results represent significant differences between Tg2576 and BL6/SJL by two-way ANOVA of data in all groups, including age group and sex as independent variables. Where significant differences were found a contrast analysis was performed on each age group. Asterisks represent significant differences between Tg2576 and BL6/SJL within each age group. a represents a significant sex-mouse type interaction. In this case, male and female data were reanalyzed separately. greatest in the older age groups. A small but significant (3.5 Ϯ 0.3%, p Ͻ 0.05) overall increase in zinc levels that bore no effect of sex or age was found in the TgC100.V717F line compared with BL6/DBA controls. Cobalt levels, however, showed a significant (p Ͻ 0.01) increase that was dependent on age. TgC100.V717F mice at 18 months had 19% higher cobalt levels than age-matched BL6/DBA controls (p Ͻ 0.001), while in the younger age groups, there were no significant differences. Manganese levels were significantly increased in TgC100.V717F brain compared with BL6/DBA controls (11 Ϯ 3%, p Ͻ 0.001). This increase showed no dependence on sex or age, and significant increases were observed in each age group (p Ͻ 0.01).
The TgC100.wt line, similarly to the TgC100.V717F line, showed significantly reduced copper and iron levels (p Ͻ 0.001) in conjunction with significantly increased manganese levels (p Ͻ 0.001) when compared with the BL6/DBA controls. However, a significant effect of sex was observed for each of these metals. On separate analysis of males and females, we found that although copper levels were significantly reduced in both sexes, the decrease was greater in females (13 Ϯ 3%, p Ͻ 0.001) than males (7 Ϯ 4%, p Ͻ 0.01). In addition, iron levels were significantly decreased in female TgC100.wt mice (9 Ϯ 1%, p Ͻ 0.001) but were unchanged in males. Conversely, manganese levels were significantly increased in males (20 Ϯ 1%, p Ͻ 0.001) but unchanged in females. No significant effect of age was found on any of these metal level changes. We found no significant difference in zinc levels in the TgC100.wt compared with the BL6/DBA line and no effect of sex. Cobalt levels showed the same alterations as in the TgC100.V717F line, with a significant increase (p Ͻ 0.01) that was dependent on age. The oldest (18 months) age group displayed 19% higher cobalt levels than BL6/DBA controls (p Ͻ 0.001), while levels in the younger age groups were unchanged. DISCUSSION This study demonstrates age-dependent increases in copper, iron, and cobalt levels in bulk brain tissue from two normal mouse strains and three strains of APP-or A␤-overexpressing Tg mice. We hypothesize that the marked elevations in copper and iron as a product of age could explain the age-dependent onset of amyloid neuropathology in the Tg2576 model (35,37). Studies in humans suggest that the aging human brain follows a similar pattern of age-related changes, at least for iron (13,15,17,38,39) and cobalt (40). In the following model, if the changes in metals we observed in mice are also reflected in the aging human brain, then a senescent rise in brain copper and iron could be the neurochemical basis for age being the major risk factor for AD neuropathology (41).
Synaptic zinc released by the glutamatergic synapses (42) is critical for A␤ plaque formation (43), although we find that zinc concentrations averaged through the whole brain remain relatively constant with age ( Fig. 1). However, A␤ plaque indeed contains a mixture of supraphysiological concentrations of copper (Ϸ0.4 mM), iron (Ϸ1 mM), and zinc (Ϸ1 mM) (3). We hypothesize that excess binding of copper and iron to A␤ could alter the metabolism of A␤ leading to its precipitation by the constitutively high ambient zinc concentrations in the synaptic (and corticovascular) milieu. Copper and iron binding to A␤ engenders H 2 O 2 production by A␤ (25), which may inhibit LRPmediated clearance mechanisms (44), leading to A␤ accumulation. An alternative possibility is that A␤ is oxidatively modified by reaction with excess copper or iron (2) and that these modified forms of A␤ are more vulnerable to zinc-(or other metal) induced precipitation. Such oxidative modification inhibits catabolic degradation of polypeptides (45,46), which may also contribute to plaque accumulation. Of the biometals that have been observed to precipitate A␤ in vitro (28,29,47), zinc, copper, and iron are the only ones with sufficient abundance and availability in the neocortex to affect A␤ aggregation and plaque formation in vivo. Cobalt, which is also elevated with age in the mice, has an effect on A␤ precipitation similar to that of zinc (29), but the concentration of cobalt in the brain is 1000-fold lower than copper, iron, and zinc and is mostly found in a non-ionic form within vitamin B 12 (cyanocobalamin). The increasing levels of cobalt with age are therefore unlikely to interact significantly with A␤.
AD is more prevalent in women, and A␤ neuropathology is more prevalent and abundant in female Tg2576 mice (48). This has recently been linked to an age-dependent relative increase in hippocampal synaptic zinc concentrations in female Tg2576 mice (43). In contrast, our current findings demonstrate no significant change in zinc in bulk brain tissue in female BL6/ SJL mice, while males show a small relative increase. This   FIG. 3. Effect of C100.V717F overexpression on brain metal levels. Metal levels were measured in TgC100.V717F and BL6/DBA brain at 2.8 months (n ϭ 16 and 17, respectively), 8 months (n ϭ 15 and 23, respectively), and 18 months (n ϭ 16 and 14, respectively) by ICP-MS. Each group contained similar numbers of males and females. Metal levels are given as g/g or ng/g wet brain weight. ANOVA results represent significant differences between Tg-C100.V717F and BL6/DBA brain metal levels by two-way ANOVA of data in all age groups, including age group and sex as independent variables. Significant differences in metal levels at each age group are indicated by asterisks. These represent contrast analyses of TgC100.V717F compared with BL6/DBA brain within each age group.
suggests an age-dependent change in the distribution of zinc might occur in females, with an enrichment in neocortical synaptic zinc levels, which facilitate amyloid formation, accompanied by a relative depletion of non-cortical zinc levels. Further studies of dissected subregions of the brains of these mouse strains are needed to confirm possible region-specific alterations in metal levels.
Overexpression of APP and C100 resulted in altered metal homeostasis in transgenic mouse brain (Table I). The TgC100 mouse models utilize the ubiquitous ␤-actin promotor (34), while the Tg2576 model uses a relatively brain-specific prion protein promotor (36,49). Moreover, the two models have different genetic backgrounds and differences in transgene expression levels. Despite these differences, a consistent finding in all three transgenic mouse lines was a decrease in copper and an increase in manganese levels across their lifespan.
Brain copper levels in TgC100 and Tg2576 mice were lowered to similar degrees (Table I); and since A␤ is the only known copper binding domain in the C100 construct, our findings are compatible with an independent role for unprecipitated A␤ in lowering brain copper levels. The dose-dependent influence of the expression of APP or its derivatives in lowering brain copper levels is also evidenced by an increase in cortical copper levels in mice that have had genetic ablation of APP (APP Ϫ/Ϫ ) or APLP2 (APLP2 Ϫ/Ϫ ) (24). However, indirect interactions of APP/C100 with other proteins that affect metal homeostasis, such as the neuronal adaptor protein X11␣ (50), is an alternative explanation for our findings that cannot yet be excluded. Our current data are limited to analysis of the bulk effects of the transgene upon metal levels using post-mortem tissue. These data are useful in illuminating the most conspicuous effects of APP expression upon metal homeostasis, but FIG. 4. Effect of C100.wt overexpression on brain metal levels. Metal levels were measured in TgC100.wt and BL6/DBA brain at 2.8 months (n ϭ 20 and 17, respectively), 8 months (n ϭ 28 and 23, respectively), and 18 months (n ϭ 20 and 14, respectively) by ICP-MS. Each group contained similar numbers of males and females. Metal levels are given as g/g or ng/g wet brain weight. ANOVA results represent significant differences between TgC100.wt and BL6/DBA brain metal levels by two-way ANOVA of data in all age groups, including age group and sex as independent variables. Significant differences in metal levels at each age group are indicated by asterisks. These represent contrast analysis of TgC100.wt compared with BL6/DBA brain within each age group. a represents a significant sex-mouse type interaction. In these cases, male and female data were re-analyzed separately.
dynamic studies that could dissect associations between the concentration of APP derivatives and their effects on metal transport would help elucidate the mechanism of these changes. Such studies of the transport of metal ions in cell cultures transfected with appropriate constructs are currently being pursued.
The decrease in copper and increase in manganese in the brains of APP and C100 Tg mice mirrors changes in the AD brain, which has also been reported to have decreased levels of copper relative to age-matched controls (7,8,10,11), and increased levels of manganese (11). Brain coppper concentration is related to plasma copper concentration (51), and both plasma (52,53) and CSF (54) copper levels are significantly elevated in AD. Taken together, these findings imply that there is a pooling of extracellular copper, and a deficiency of intracellular copper, in the AD brain. Due to the catalytic nature of reactive oxygen species generation by redox active metals such as copper and iron, small changes in the levels or distribution of these metals could cause severe oxidative stress. Therefore, elaboration of the compartments of metal ions altered in AD or Tg mouse brain, in contrast to the compartments of metal ions (particularly copper and iron) that are increased due to the aging process, warrants further investigation.
In vitro studies have demonstrated no significant interaction between A␤ or APP and manganese (23,28,29,47). Therefore the increased manganese levels we observed in transgenic mice may be a result of secondary effects of altered metal homeostasis or an up-regulation of manganese-binding proteins such as mitochondrial manganese-superoxide dismutase, in response to increased intracellular oxidative stress. Alternatively, in the brain microenvironment, A␤ may interact with manganese in a manner not yet observed by in vitro studies.
The consistent decrease in iron levels in both TgC100.wt and TgC100.V717F lines suggests a role of A␤ in iron homeostasis. This may be occurring via direct interaction with A␤, but may alternatively reflect a homeostatic adjustment to the reduction in copper levels. The Tg2576 line, in contrast, did not show this decrease. Human AD brain exhibits substantially increased iron levels (10,11,55). Although a portion of this increase may be due to the iron content of accumulated A␤, a more generalized increase in brain tissue may be a consequence of other AD-associated pathogenic changes affecting iron homeostasis such as elevated ferritin levels (6). The exaggerated retention of iron in A␤ deposits or in increased ferritin deposits in the Tg2576 mouse brain may oppose the tendency of human A␤ expression to decrease brain iron levels in the C100 mouse models, explaining why there is no net decrease in brain iron in Tg2576 mice.
Copper and zinc level decreases induced by transgene expression showed little enhancement with age in the Tg2576 brain, despite the fact that A␤ levels accumulate several hundredfold from 2.8 to 18 months, with plaque formation becoming conspicuous from 10 months in these mice (37). The decrease in copper and zinc levels that we observed (Fig. 2) are therefore not a consequence of insoluble A␤ aggregates or plaque pathology. This decrease must either be due to secreted APP and/or A␤ promoting the efflux of the metal ions or APP/A␤ preventing their uptake. Supporting this latter possibility is evidence that A␤ scavenges extracellular Cu 2ϩ , possibly to prevent oxidation (56). Furthermore, treatment of 21month-old Tg2576 mice with clioquinol, an antibiotic with copper/zinc chelation properties, both inhibited plaque formation and paradoxically elevated soluble brain copper and zinc levels. Iron, cobalt, and manganese levels were unaltered (33). In NTg mice, clioquinol treatment decreases copper, iron, and cobalt levels (57). In light of our current findings, this paradoxical increase in copper and zinc in clioquinol-treated Tg2576 mice may be explained by clioquinol preventing Cu 2ϩ and Zn 2ϩ from complexing with extracellular A␤, so securing metal for uptake into metal-deficient brain tissue instead of being sequestered into amyloid. The consequent lowering of extracellular metal concentrations inhibited the formation, or possibly facilitated the dissolution, of amyloid deposits.
Taken together, our findings demonstrate that overexpression of human A␤ in TgC100 mice replicates the lowering of copper and raising of manganese levels that is observed due to APP overexpression in Tg2576 brain (Table I). Different effects on iron, zinc, and cobalt in these Tg mice suggest discrete and perhaps opposing roles for A␤ and APP ectodomain metal binding sites. Assuming the effects of C100 expression on metal levels are due to A␤, and that the APP expression involves the combined actions of both A␤ and APP ectodomain metal binding sites, we can estimate the effect of the APP ectodomain on brain metal metabolism (Table II). Our study reveals that A␤ is involved in the reduction of copper levels. A compounding role of both A␤ and the APP ectodomain in reducing copper levels is  We predict the contribution of the APP ectodomain to brain metal levels by comparing the effect of A␤ overexpression in the TgC100 lines against the effect of APP.swe overexpression in the Tg2576 line. 2 represents decrease, 1 represents increase, and n.c. represents no significant change. The parentheses represent variable effects relating to age and sex. APP  supported by the observation that APLP2 Ϫ/Ϫ mice, like APP Ϫ/Ϫ mice, have increased brain copper levels (24). APLP2 does not produce A␤, so its influence on copper levels is probably mediated by its ectodomain metal binding sequence, which is homologous with that of APP (23,58). The role of APP ectodomain in regulation of brain manganese is unknown, but A␤ is able to increase manganese levels. The A␤ and the APP ectodomain appear to have opposing effects with respect to iron, zinc, and cobalt levels. This differential activity could be due to amyloid deposition either sequestering zinc and cobalt or indirectly altering metal homeostasis in the Tg2576 mice. Alternatively, it could be due to the export of zinc and cobalt by the overexpressed APP ectodomain. Correlation of these findings to human aging and AD is limited by differences between humans and mice in their metal regulatory machinery. In addition, the Tg2576 model does not possess the full spectrum of AD pathology, in particular neurofibrillary tangles. With these caveats, our findings suggest that amyloid pathology in AD may represent the corruption of a compensatory system for preventing the entry of excess copper, which rises as a consequence of aging, into brain tissue. Treatment with certain metal chelators, such as clioquinol, which has been recently shown benefit to AD patients in a phase two clinical trial (59), may exert their therapeutic effects not just by clearing amyloid deposition, but also by restoring brain metal homeostasis.