Fenton chemistry and oxidative stress mediate the toxicity of the β-amyloid peptide in a Drosophila model of Alzheimer’s disease

The mechanism by which aggregates of the β-amyloid peptide (Aβ) mediate their toxicity is uncertain. We show here that the expression of the 42-amino-acid isoform of Aβ (Aβ1–42) changes the expression of genes involved in oxidative stress in a Drosophila model of Alzheimer’s disease. A subsequent genetic screen confirmed the importance of oxidative stress and a molecular dissection of the steps in the cellular metabolism of reactive oxygen species revealed that the iron-binding protein ferritin and the H2O2 scavenger catalase are the most potent suppressors of the toxicity of wild-type and Arctic (E22G) Aβ1–42. Likewise, treatment with the iron-binding compound clioquinol increased the lifespan of flies expressing Arctic Aβ1–42. The effect of iron appears to be mediated by oxidative stress as ferritin heavy chain co-expression reduced carbonyl levels in Aβ1–42 flies by 65% and restored the survival and locomotion function to normal. This was achieved despite the presence of elevated levels of the Aβ1–42. Taken together, our data show that oxidative stress, probably mediated by the hydroxyl radical and generated by the Fenton reaction, is essential for Aβ1–42 toxicity in vivo and provide strong support for Alzheimer’s disease therapies based on metal chelation.


Introduction
There is a growing consensus that smaller, soluble aggregates of the b-amyloid peptide (Ab), rather than mature amyloid plaques, are the pathogenic species in Alzheimer's disease (AD) (Lambert et al., 1998;Walsh et al., 2002). However, the mechanism by which these aggregates mediate their toxicity remains unclear. As Ab is generated in the extracellular space, or more likely within the lumen of endocytic vesicles (Koo & Squazzo, 1994;Refolo et al., 1995), its toxic effects may be mediated by membrane damage or by interactions with membrane-bound proteins. There is evidence that Ab aggregates can degrade the electrical resistance of membranes (Kayed et al., 2004;Demuro et al., 2005), possibly by forming pores (Lashuel et al., 2002), or alternatively they may interact with membrane receptors or even gain access to the cytoplasm. Membranes can also be damaged by the reactive oxygen species that are generated by Ab aggregates in the presence of metals such as copper, zinc or iron (Bush, 2003). Subsequent pathological processes include mitochondrial damage (Abramov et al., 2004), tau phosphorylation with consequent axonal transport dysfunction and the initiation of cell death (Kienlen-Campard et al., 2002;Wei et al., 2002;Jo et al., 2004). However, until recently (Cao et al., 2008) it has been impossible to take a global view to ask which biological processes are essential for the development of the disease and which are downstream consequences of neurotoxicity. Knowing which biological processes are directly involved in initiating AD will allow us to focus on those upstream targets that have the greatest therapeutic potential.
We have developed a model of AD that is based on the expression of the human Ab in fly neurons by coupling it to an N-terminal secretion signal peptide (Crowther et al., 2005). The Ab 1-42 but not the Ab 1-40 control accumulates in the brain and results in decreased lifespan and impaired locomotor performance. These phenotypes are more marked in flies expressing the E22G (Arctic) mutant of the Ab 1-42 , which causes increased aggregation of Ab and is responsible for early onset familial AD (Nilsberth et al., 2001). Here, we use our Drosophila model of AD to identify the pathways and intermediates that are critical for Ab-mediated toxicity in vivo.

DNA constructs
The cDNA for Drosophila ferritin 1 heavy chain (Fer1HC) was isolated from EST clone GH24060 (Berkley Drosophila Genome Project). This cDNA lacked the iron response element resulting in the expression of a constitutively active form of Fer1HC. The cDNA for Drosophila ferritin 2 light chain (Fer2LC) naturally lacks an iron response element and was isolated from the EST clone AT16780 (Berkley Drosophila Genome Project). Both cDNAs were inserted downstream of GAL4 UASs (UAS-Fer1HC and UAS-Fer2LC) in the pUAST plasmid by directional cloning following EcoRI and XhoI digestion.

Affymetrix cDNA microarray
Individual samples were each hybridized to a Drosophila Genome GeneChip Ò Array following standard Affymetrix protocols. Affymetrix Microarray Suite 5 was used to generate signal values and detection calls. Probe level intensity data were adjusted for background, normalized and log transformed using the robust multichip average pre-processing method (Irizarry et al., 2003) using Rosetta Resolver Ò 7.1.
Ratio data for each individual probe were created as follows. Ab 1-40 baseline samples were created for each time-point by pooling all replicates (see Supplementary material, Fig. S1a). The individual replicates for each test condition (Arctic Ab 1-42 and Ab 1-42 ) were then compared with the age-matched control Ab 1-40 pool. All statistical analyses were performed on log 10 ratio data.
Gene expression differences were determined by one-way, errorweighted anova on ratio data by comparing Ab 1-42 and Arctic mutants with age-matched Ab 1-40 controls (factor = genotype, n = 4 per group). Genes were only considered in the analysis of overrepresented biological themes if the differential expression was highly significant (P < 0.01).

Characterization of over-represented biological themes
Gene Ontology (Ashburner et al., 2000) enrichment analysis was performed using the Gene Ontology tree machine (http://bioinfo.vanderbilt.edu/gotm/) (Zhang et al., 2004). In summary, a hypergeometric test was performed to determine whether a particular biological process or molecular function was disproportionately represented in the set of genes that were differentially expressed as compared with all of the genes in the GeneChip Ò array. Where n = number of genes that are differentially expressed between two experimental conditions (e.g. between Arctic Ab 1-42 and Ab 1-40 on day 0), N = total number of genes on the array, K = number of genes on the array that belong to the category of interest and k = number of genes that belong to the category that are differentially expressed, the significance (P) of enrichment for a given category is determined by A ratio of enrichment, R, is calculated as R = k ⁄ k e where k e = (n ⁄ N)K is the expected value for a given category if n = number of genes were a random sample of uniformly selected genes from the reference set of all genes on the chip.

Gene Search element screen
A library of 3000 unique insertions of the Gene Search (GS) element was generated by mobilizing the GS element from the X chromosome (DGRC number 200079) to the autosomes by crossing with flies expressing the Delta2-3 transposase. The first eclosed fly with a stably-jumped GS element from each mobilization cross was used to establish a GS line. Stocks were maintained by monitoring eye colour. The effect of the Gal-4 activated GS element on the longevity of flies expressing the Arctic mutant of Ab 1-42 was determined by crossing male flies with elav c155 -Gal4 (elav-Gal4) on the X chromosome and the Arctic Ab 1-42 transgene on the second chromosome with virgin females with a floating GS element. All of the female offspring had both elav-Gal4 and the Arctic Ab 1-42 transgene but only half had the GS element; the unmodified population (without the GS element) provided an internal control for each longevity assay. The null hypothesis was that the presence of the GS element made no difference to the longevity of the flies expressing the Arctic Ab 1-42 transgene.
Longevity assays for the primary screen were performed at 25°C, blind to the identity of the GS elements. Live flies were counted and the food changed on days 1, 3 and 5 of a 7 day cycle. A mean number of 17 flies was assessed per GS line (total number of flies assessed, 50 320; minimum number of flies assessed per GS line, 10). To detect suppression of the longevity phenotype we determined the time to 75% death, which is the median survival of the flies in which the GS element has prolonged their life. These survival times were normally distributed (n = 2893, mean 21.6 days, SD 3.8 days) and any lifespan that was more than two SDs greater than the mean was defined as significant.
In the secondary screen the chromosomal site of insertion was determined by classical genetics and homozygous GS flies were crossed so that all flies in the longevity assays expressed the activated GS element. This secondary screen used homozygous GS element stocks and assessed more than 30 flies, permitting robust comparison of the survival of GS-modified and unmodified Arctic Ab 1-42 flies. The site of the GS element insertion was identified in those lines with confirmed modifier activity by inverse polymerase chain reaction using the FlyChip facility (http://www.drosdel.org.uk/molecu-lar_methods.php). The molecular function of genes with inserts within the coding sequence was determined from FlyBase. For inserts in noncoding DNA, the molecular function was determined for the genes on either side of the insert. The effect of the activated GS element on the survival of control flies and flies expressing the AlzArc1 transgene was also determined. GS lines in which the modification of the Ab longevity phenotype was not confirmed or in which the GS element had a marked non-specific effect on control flies were discarded.
Longevity assays for the secondary screen and genetic modifier assessment Longevity assays in the secondary screen and for each of the subsequent genetic modifiers followed at least 80 flies per genotype in groups of 10 flies per vial. Live flies were counted and their food changed on days 1, 3 and 5 of a 7 day cycle. For assessing the efficacy of metal chelation, clioquinol (Calbiochem) was dissolved in dimethylsulphoxide and the solution added to fly food to give a final concentration of 0.2% v ⁄ v. Survival curves were plotted using the Kaplan-Meier estimator. The statistical significance was calculated using the log rank test within the spss 11.0 statistical package. The null hypothesis in all of the longevity assays was that the presence of the GS element made no difference to the longevity of the flies expressing the Arctic Ab 1-42 transgene.

Assay of sensorimotor performance
The sensorimotor performance of the flies was determined using a previously described negative geotaxis assay (Rival et al., 2004). Fifteen flies were placed in a sterile plastic column (25 cm tall · 1.5 cm internal diameter) and tapped to the bottom. After 45 s the flies at the top of the column (N top ) and the flies remaining at the bottom (N bot ) were counted. Three trials were performed at 1 min intervals. The performance index was defined as (15 + N top -N bot ) ⁄ 30. Statistical analysis was performed using the two-tailed Student's t-test.

Quantitation of b-amyloid peptides
Drosophila were cultured at 25°C for 5 days after eclosion and then five heads were homogenized in 50 lL of 5 m guanidinium HCl, 5 mm EDTA and 50 mm hepes, pH 7.3. Following centrifugation for 5 min at 12 000 g, 20 lL of the clear supernatant was removed and mixed with 180 lL of 25 mm HEPES, pH 7.3, 1 mm EDTA and 0.1% w ⁄ v bovine serum albumin with protease inhibitors (CompleteÔ, Roche). Triplicate 25 lL aliquots were mixed with an equal volume of phosphate-buffered saline containing 2% w ⁄ v bovine serum albumin, 0.2% v ⁄ v Tween-20 and protease inhibitors (CompleteÔ, Roche) in wells on a MESO microtitre plate (Standard bind, MA6000, no. P11AA-1; Meso Scale Discovery, MD, USA). The reaction was started by adding 25 lL of 4 lg ⁄ mL solutions of biotinylated 6E10 or 4G8 (Signet Laboratories, MA, USA) monoclonal antibodies. After mixing a further 25 lL aliquot of 1 lg ⁄ mL Ruthenium-labelled G2-10 or G2-11 (The Genetics Company, Switzerland), monoclonal antibody solution was added to each well. Following an overnight incubation at 25°C, the plates were washed twice with phosphatebuffered saline, 150 lL of S Read Buffer (R92SC-1, Meso Scale Discovery, MD, USA) was added and the measurement was taken in a Sector PR instrument (Meso Scale Discovery). Statistical analysis was performed using the two-tailed Student's t-test.

Hydrogen peroxide sensitivity assay
Two-day-old flies were cultured in a vial containing only filter paper soaked in an aqueous solution of 2% w ⁄ v sucrose and 10% v ⁄ v H 2 O 2 or a control solution of 2% w ⁄ v sucrose alone. Ten vials of 15 flies were cultured for each genotype at 25°C with the soaked paper being replaced twice per day. Statistical analysis was performed using the two-tailed Student's t-test.

Carbonyl assay
Five adult flies (28 days old, grown at 25°C) were decapitated and the heads were immediately homogenized in water, sonicated and centrifuged at 12 000 g for 10 min and the clear supernatant assayed for protein content using the Bradford method. The protein concentration was adjusted to 5 lg ⁄ mL by the addition of phosphatebuffered saline. Protein carbonyl groups were assayed using an enzyme-linked immunosorbent assay-based protocol described by Alamdari et al. (2005). Protein carbonyl groups were reacted with dinitrophenol hydrazine (Sigma-Aldrich) and the resulting dinitrophenol adducts were detected using an anti-DNP rabbit polyclonal antibody (Sigma-Aldrich). Statistical analysis was performed using the two-tailed Student's t-test.

Iron and zinc determination in fly head extracts by mass spectrometry
Flies were cultured at 29°C and decapitated at 10 days old. Triplicates of 20 fly heads per condition were weighed by difference into a new 15 mL screw-top polypropylene centrifuge tube (part no. 2086-500, Elkay, UK). Nitric acid (500 lL) ('Trace Select', part no. 84385, Fluka, UK) was added and the tubes were sealed and incubated overnight at 65°C such that there was no visible solid matter remaining. The sample was prepared by resuspending the extracts in 5 mL of double-distilled water before analysis using an Elan 6100 DRC ICP ⁄ MS (Perkin-Elmer). The reaction cell was used for all measurements with the following operating conditions: 1.04 L ⁄ min nebulizer flow, 1100 W radio frequency power and a dynamic reaction cell gas flow of 0.7 mL ⁄ min of ammonia. Iron (Mr 55.9349), zinc (Mr 65.9260) and calcium (Mr 43.9555), but not copper (Mr 62.9298), were reliably detected. As calcium is not chelated by clioquinol it was used to control for the quantity of tissue in each sample.

Results
Microarray analysis of gene expression in flies expressing b-amyloid peptide supports the importance of oxidative stress in b-amyloid peptide toxicity Affymetrix microarrays were used to identify gene expression signatures (one-way anova, P < 0.001) for flies expressing wildtype Ab 1-42 at day 0, 3 and 8 of adult life and Arctic Ab 1-42 at day 0 and 3 by comparison to age-matched Ab 1-40 controls. The derived gene expression signatures for flies expressing wild-type and Arctic Ab 1-42 were then used to determine the biological processes, molecular functions and cellular components that were over-represented in each set (in supplementary Fig. S1b). The role of redox genes appeared most apparent in the early stages of Ab 1-42 expression (Fig. 1); this is most clearly seen for flies expressing wild-type Ab 1-42 at day 3 (top part, shaded) where these genes comprised six of the 10 families of enriched genes. Genes that were classified as having 'oxidoreductase activity' were significantly enriched at both day 3 and 8 in flies expressing wild-type Ab 1-42 (P < 0.01). Similarly, carbonyl reductase (NADPH) activity was enriched in the day 3 signature of flies expressing Arctic Ab 1-42 (P < 0.01). Taken together, these observations provide strong support for the role of oxidative stress-related genes in the response of the fly brain to the expression of the Ab. P-element screen for modifiers of b-amyloid peptide toxicity implicate redox regulation as an important target of modifier activity A 3000-line library of unique GS (Toba et al., 1999) element inserts was screened for lines that modified the longevity of Drosophila expressing Ab 1-42 in their central nervous system. The GS elements can either disrupt gene function when they insert within essential coding or non-coding DNA, or else they can bidirectionally upregulate neighbouring genes. In our screen 1.5% of the GS inserts resulted in an increase in median survival that was more than two SDs away from the mean, whereas 0.5% of the inserts significantly reduced survival. A secondary longevity assay was performed to confirm the initial findings but also to determine the effect of the GS elements on the survival of control flies that did not express Ab. The insertions that specifically modified the longevity of Ab-expressing flies, and not controls, were classified into 18 suppressor and three enhancer groups according to the identity of the neighbouring genes. Seven of these 21 classes were adjacent to genes with a predictable role in oxidative stress (Table 1 and Fig. 2). The selected suppressor GS elements increased median survival by 25-96% in flies expressing Ab but had no, or little, effect in control flies (Fig. 2, filled and empty bars, respectively).
Flies expressing b-amyloid peptide 1-42 are more sensitive to oxidative stress and have higher levels of oxidative damage The genetic screen and microarray data implicated oxidative stress as playing a central role in the toxicity of the Ab 1-42 . This was assessed Ab, b-amyloid peptide; GS, Gene Search.

Fig. 2. The percentage increase in median survival in flies expressing Arctic
Ab 1-42 that was attributable to the GS element was calculated (filled bars). Control flies (n = 100) possessing elav-Gal4 ± the GS element but lacking the Arctic Ab 1-42 transgene (n = 100) were also assessed. The percentage increase in median survival in these control flies that was attributable to the GS element was determined (empty bars). The significance of the greater increase in median survival in Arctic Ab 1-42 vs. the effect on control flies was significant (P < 0.001) in all cases. The number of independent GS element inserts that were independently identified in the screen and analysed in this assay is indicated in white on the filled bars. Where three or more GS inserts were available in a particular class the error bars indicate the SD of the estimates. We assumed that the estimates varied normally and the significance was calculated by Student's t-test [ferritin light (L) chain inserts, P < 0.05; ferritin heavy (H) chain inserts, P < 0.001; CG9432, P < 0.001]. DH, dehydrogenase.
Transgenic over-expression of single antioxidative stress genes rescues the b-amyloid peptide-induced longevity phenotype The ability of antioxidative stress genes identified in the screen, and genes from the canonical oxidative stress pathway, to modify the toxicity of Ab was confirmed in flies by specifically over-expressing each transgene in combination with wild-type or Arctic Ab 1-42 . In this way we assessed candidate genes from the GS element screen, i.e. the heavy and light chains of Drosophila ferritin (Fer1HC and Fer2LC) and carbonyl reductase by over-expression of the Drosophila enzyme Sniffer (SNI). We also tested candidate genes including cytoplasmic CuZn-SOD1, the mitochondrial Mn-SOD2 (mitSOD2) and CAT, and GST (Fig. 4). This approach allowed comparison of the efficacy of the upstream enzymes that modulate the generation of free radicals (SOD1, mitSOD2, CAT, Fer1HC and Fer2LC) with the downstream enzymes that repair oxidative damage (GST and carbonyl reductase ⁄ SNI). Over-expression of ferritin heavy chain resulted in a 105% increase in median survival of Arctic Ab 1-42 flies (P < 0.0001), whereas overexpression of ferritin light chain gave a 49% increase in median survival (P < 0.0001) ( Fig. 4a and b). The most powerful canonical oxidative stress-related protein was CAT, which increased the median survival of Arctic Ab 1-42 flies by 56% (P < 0.0001, Fig. 4a and c). MitSOD2 prolonged median survival by 18% (P < 0.0001, Fig. 4a and d). In contrast, and contrary to our expectations, the overexpression of cytoplasmic SOD1 enhanced the toxicity of Arctic Ab 1-42 . Flies expressing mutants of SOD1 that are dominant-negative for activity were then assessed to see if they had a similar modifying activity. Co-expression of both the SOD1 n108 mutant and RNAi for SOD1 resulted in a 43% and 32% increase in median survival, respectively ( Fig. 4a and c), indicating that it is the catalytic activity of SOD1 that potentiates the toxic effect of the Ab. The protection afforded by CAT combined with the toxicity of SOD1 suggests that the uncompensated production of H 2 O 2 is a vital step in the oxidative stress caused by the Ab. An additional stress resulting from increased SOD1 activity may be the co-production of O 2 by the dismutation of superoxide radicals. This oxygen may go on to generate further H 2 O 2 by reacting with iron or copper ions that are complexed with Ab (Huang et al., 1999a,b).
Much smaller protective activities were observed when enzymes, specifically Drosophila carbonyl reductase (SNI) and GST, that are involved in steps further downstream in the pathway were upregulated (Fig. 4a). Control experiments, in which the antioxidative stress genes were over-expressed in flies that were identical except that they lacked the Ab transgene, did not reveal non-specific prolongation of life (Fig. 4d). In flies expressing wild-type Ab 1-42 we were able to confirm the prolongation of lifespan by the heavy and light chains of ferritin and the toxic effect of SOD1 (data not shown).
Transgenic over-expression of single antioxidative stress genes rescues the b-amyloid peptide-induced locomotor phenotype The modifiers were then assessed for their ability to protect against the Ab 1-42 -mediated decline in locomotor function. Flies co-expressing either ferritin subunit with Arctic Ab 1-42 were significantly more mobile than flies expressing Arctic Ab 1-42 alone (Fig. 5a, elav-Gal4  . Flies co-expressing Arctic Ab 1-42 and the heavy chain of ferritin exhibited survival that was similar to that of control flies (b, elav-GAL4). Similarly, co-expression of another gene from the GS screen, carbonyl reductase (a, elav-Gal4 UAS-SNI), yielded a small but significant prolongation of the lifespan of Arctic Ab 1-42 flies. The expression of other canonical antioxidative enzymes also had significant effects on longevity. Both mitSOD2 (a, elav-Gal4 UAS-mitSOD2) and CAT (a and c, elav-Gal4 UAS-CAT ) prolonged the lifespan of the flies. Surprisingly, SOD1 (a and c, elav-Gal4 UAS-SOD1) enhanced the toxicity of Arctic Ab 1-42 . In contrast, the knockdown of endogenous SOD1 protein by UAS-RNAi (a and c, elav-Gal4 UAS-IR.SOD1) protected the fly from Ab toxicity. The enhancer effect of SOD1 appears to be mediated by its catalytic activity because a dominant negative mutant of SOD1 prolonged the lifespan of the flies expressing Arctic Ab (a and c, SOD1 n108 ). In control experiments there was no prolongation of lifespan when SOD1, mitSOD2, CAT, carbonyl reductase, ferritin heavy chain and ferritin light chain were expressed using elav c155 -Gal4 in flies that did not carry the UAS-Arctic Ab 1-42 transgene (d). Kaplan-Meier survival curves were plotted and statistical significance was assessed by the log rank test using the spss 11.0 statistical package. Differences shown were all statistically significant (P < 0.001). In control experiments elav-Gal4 flies had the same lifespan as the background w 1118 flies.
The metal chelator clioquinol reduces b-amyloid peptide-mediated neuronal toxicity and specifically reduces iron levels in the brain That the protection against Ab toxicity afforded by ferritin is mediated in part by the chelation of Fe 2+ and Fe 3+ ions was investigated by treating flies expressing Ab 1-40, wild-type Ab 1-42 and Arctic Ab 1-42 with the metal chelator clioquinol (Kaur et al., 2003) (Fig. 6). Although clioquinol had no effect on the longevity of flies expressing the non-toxic Ab 1-40 (Fig. 6a, circles) there was, in contrast, a clear dose-related increase in longevity for Arctic Ab 1-42 (Fig. 6a, diamonds) flies. Expression of wild-type Ab 1-42 (Fig. 6a, triangles) gave an intermediate response to clioquinol treatment with an optimum dose of approximately of 20 lm. The degree of functional rescue in flies expressing Arctic Ab 1-42 closely reflected the concentration of iron in the extracts of fly heads (Fig. 6b). We found that the presence of Arctic Ab 1-42 (Fig. 6b, triangles) increased the uptake of iron as compared with control flies (Fig. 6b, circles), a striking effect that is only completely reversed by treatment with 200 lm clioquinol. The levels of zinc (1.5-1.7 mm) were the same in Arctic Ab 1-42 -expressing and control flies and remained unchanged following clioquinol treatment. Copper could not be detected in the head extracts.
Oxidative damage in flies expressing b-amyloid peptide is reduced by ferritin and increased by superoxide dismutase 1 The quantity of carbonyl groups was assessed in flies expressing Arctic Ab 1-42 in the presence or absence of the heavy and light chains of ferritin. The expression of Arctic Ab 1-42 significantly increased the concentration of carbonyl groups (Fig. 7a, elav-Gal4 w 1118 vs. elav-Gal4 UAS-Arctic Ab 1-42 , P < 0.001) but this was reduced by 30% following the co-expression of ferritin light chain (Fig. 7a, elav-Gal4 UAS-Arctic Ab 1-42 UAS-FerLC and elav-Gal4 UAS-Arctic Ab 1-42 , P < 0.01). Remarkably, ferritin heavy chain (Fig. 7a, elav-Gal4 UAS-Arctic Ab 1-42 UAS-FerHC) reduced the carbonyl level (P < 0.001) almost to that of control flies (Fig. 7a, control elav-Gal4 UAS-Arctic Ab 1-42 ). These findings are consistent with ferritin exerting its suppression of the Ab phenotype by an antioxidant effect. Conversely, the enhancer effect of SOD1 was accompanied by a significant increase in carbonyl levels in Arctic Ab 1-42 flies (Fig. 7a, elav-Gal4 UAS-Arctic Ab 1-42 vs. elav-Gal4 UAS-Arctic Ab 1-42 UAS-SOD1, P < 0.01). These data suggest that the rapid production of H 2 O 2 by cytoplasmic SOD1 can overwhelm endogenous CAT resulting in the Fe 2+ -mediated generation of the hydroxyl radical (Fenton reaction) that damages proteins as demonstrated by the increased carbonyl load.
Ferritin subunits suppress the toxicity of b-amyloid peptide 1-42 despite increased levels of b-amyloid peptide 1-42 in the brain The effect of over-expressing antioxidant transgenes on the level of Ab 1-42 in the brains of flies was assessed by an enzymelinked immunosorbent assay that measured 5 m guanidinium hydrochloride-soluble Ab 1-42 . Six independent lines of flies expressing Arctic Ab 1-42 and carrying GS elements that cause over-expression of Fig. 6. The iron-chelating compound clioquinol was added to the fly food at final concentrations of 2, 20 and 200 lm and its effects on the Ab-associated longevity phenotype and the accumulation of iron in the fly brain was assessed. The median lifespan of flies expressing Ab 1-40 (a, circles) was not significantly increased by clioquinol at any concentration; however, flies expressing Arctic Ab 1-42 demonstrated a clear concentration-related increase in median survival (a, diamonds). Wildtype Ab 1-42 flies exhibited an intermediate, but significant, response (a, triangles). Kaplan-Meier survival statistics with the log rank test were used to analyse the data (significance of difference from no-clioquinol control, **P < 0.01, ***P < 0.001, unlabelled P > 0.05). Flies expressing Arctic Ab 1-42 (b, triangles) accumulated significantly more iron in their brains than control flies (b, circles) and this was only reversed by treatment with 200 lm clioquinol. Error bars show the SD (n = 3). The significance of the difference between clioquinol-treated and non-treated flies was calculated pairwise using the two-tailed Student's t-test (**P < 0.01).
ferritin subunits were shown to accumulate significantly higher levels of Ab 1-42 in their brains (65 pm in control flies vs. 100 pm for GS elements co-expressing the light chain gene, P < 0.01, and 141 pm over-expressing the heavy chain gene, P < 0.01). This elevation of Ab 1-42 when over-expressing ferritin subunits was confirmed in flies co-expressing Ab 1-42 with UAS-linked transgenes for either the heavy or light chains (Fig. 7b). In contrast, the levels of Ab 1-42 were unaffected by the co-expression of CAT or the over-expression, or the knockdown, of SOD-1. Thus, the rescue mediated by over-expression of oxidative stress genes was not mediated by suppression of the Ab 1-42 .

Discussion
Two genome-wide genetic screens were used in our Drosophila model of AD to identify genes that play an important role in the toxicity of the Ab. The first screen, using Drosophila cDNA microarrays, quantified the changes in gene transcription that occurred in response to the expression of Ab 1-42 as compared with Ab 1-40 . The second complementary screen, using a library of flies with unique GS element insertions, identified genes, which when transcriptionally up-or downregulated modified the lifespan of flies expressing Ab 1-42 . Taken as a whole the GS screen was remarkable for the number of genes adjacent to modifying P-elements that had redox or antioxidant activities. Similarly, the cDNA microarray studies showed that redoxassociated transcripts were the most robustly represented functional group of genes that were differentially regulated by Ab 1-42 expression. This combination of observations is powerful because, whereas the microarray data provide genetic evidence of a response by the brain to oxidative stress, the GS screen, in contrast, points to a direct pathogenic role for oxidative stress in the generation of Ab-related Fig. 7. Oxidative stress mediated by the Arctic Ab 1-42 is reversed by ferritin despite an increase in the levels of Ab 1-42 . A reduction in carbonyl load was apparent with co-expression of the light chain of ferritin (a, UAS-FerLC elav-Gal4 UAS-Arctic Ab 1-42 , P < 0.05); however, the effect was more marked following over-expression of the heavy chain of ferritin (a, UAS-FerHC elav-Gal4 UAS-Arctic Ab 1-42 , P < 0.01). In contrast, co-expression of SOD1 (a, UAS-SOD1 elav-Gal4 UAS-Arctic Ab 1-42 , P < 0.01) significantly increased the oxidative damage as compared with flies expressing Arctic Ab 1-42 alone (a, control elav-Gal4 UAS-Arctic Ab 1-42 ) (**P < 0.01, ***P < 0.001). The concentration of Ab 1-42 in the heads of flies was determined for three independent biological replicates for each line of flies. Co-expression of both the heavy and light chains of ferritin results in significantly increased levels of Ab (b, UAS-FerLC and UAS-FerHC). In contrast, modification of SOD1 or CAT activity (b, UAS-SOD1, UAS-SOD1 RNAi and UAS-CAT) has no effect on Ab levels. Error bars show the SEM.
phenotypes. The role of oxidation in causing disease was further emphasized by the remarkably close correlation between the severity of the phenotypes observed for Ab 1-40 , Ab 1-42 and Arctic Ab 1-42 and the oxidative modification of brain proteins as determined by carbonyl levels. These findings are concordant with the raised levels of oxidized proteins and lipids in post-mortem AD brains where it is known that oxidative damage is present from the earliest clinical stages of the disease (Nunomura et al., 2001;Markesbery et al., 2005). Our work therefore represents an advance on current clinical work and mouse models that have not clarified whether oxidative stress plays a direct role in the pathogenesis of AD or whether it is a consequence of the disease process. Indeed, the view that oxidative stress is a bystander effect is supported by clinical trial data that show, despite early encouraging data from observational studies (Morris et al., 2002a,b;Zandi et al., 2004) and one prospective trial (Sano et al., 1997), that the antioxidant vitamin E is not able to prevent the onset of dementia or prevent progression of established disease (Petersen et al., 2005).
In this work we have undertaken a molecular dissection of the oxidative stress pathway (Fig. 8) in order to gain a more detailed understanding of which oxidative species were of particular importance in Ab-mediated neurodegeneration. This work has emphasized three main sources of oxidative stress: firstly, the generation of hydroxyl radicals via the Fenton reaction (Smith et al., 1997;Dikalov et al., 2004), secondly, mitochondrial superoxide levels and thirdly, reactive aldehyde species produced by lipid oxidation. Although each of these sources contributes significantly to oxidative stress, the conversion of H 2 O 2 to the hydroxyl radical is the most important toxic process. The most immediate support for this comes from the profound protection afforded to flies when CAT is co-expressed with Ab 1-42 because hydroxyl radical generation will predictably be abolished by converting H 2 O 2 to water.
Only co-expression of ferritin subunits offered a more effective rescue of the longevity phenotype than CAT. Ferritin proteins are highly conserved in evolution and have been extensively characterized (Harrison & Arosio, 1996). Functional ferritin complexes may be composed of various proportions of heavy and light subunits. For instance, in the human, ferritin from the brain is predominantly composed of heavy chain, whereas in the liver the ferritin has a high proportion of light chain. In our model, the over-expression of Fer1HC or Fer2LC occurred in the context of endogenous Fer1HC and Fer2LC expression. Thus, in our transgenic animals we should produce ferritin complexes composed of both subunits but with one or the other representing the major component. Remarkably, ferritin heavy chain was able to restore the longevity and locomotor phenotype of flies expressing the highly toxic Arctic Ab 1-42 to that of control flies. That this rescue was mediated by an antioxidant effect is supported by the reductions in carbonyl levels in the brains of flies co-expressing either the light or, again more potently, heavy chain of ferritin. It is likely that the sequestration of iron by both ferritin subunits, and the conversion of Fe 2+ to Fe 3+ by the heavy chain, slows hydroxyl radical production. Our data show for the first time that it is likely to be the removal of Fe 2+ , by the ferroxidase activity of the heavy chain, that is specifically beneficial. Previous in-vitro data have shown that synthetic Ab can directly generate H 2 O 2 in the presence of metal ions (Huang et al., 1999a;Tabner et al., 2005) and in particular iron (Khan et al., 2006). Using molecular oxygen as a substrate, the production of H 2 O 2 by synthetic Ab depends on Fe 2+ ions generated via a redox cycling of iron (Huang et al., 1999a;Khan et al., 2006). Thus, it is possible that ferritin can not only prevent the Fenton reaction but additionally the Fe 2+ -scavenging activity of ferritin heavy chain can also protect neurones against the intrinsic redox properties of Ab (Fig. 8).
The powerful antioxidant properties of ferritin can also rescue the locomotor deficits associated with Ab expression despite an accompanying twofold increase in Ab load. These data make the modulation of brain iron metabolism an attractive therapeutic target not least because the concentration of Fe 2+ in amyloid plaques is 1 mm, almost three times the normal level (Bush, 2003). Furthermore, clioquinol, a metal chelator, reduces plaque deposition in mouse models of AD (Cherny et al., 2001), is safe in clinical trials (Ritchie et al., 2003) and, as we show here, specifically reduces iron levels in the brain and prolongs the survival of flies expressing Arctic Ab 1-42 .
Oxidative stress has been linked with mitochondrial dysfunction in several neurodegenerative disorders including AD and also Parkinson's and Huntington's diseases (Bowling & Beal, 1995;Maier & Chan, 2002). In AD there is evidence that soluble oligomeric aggregates of Ab may damage the insulating properties of the neuronal plasma membrane (Kayed et al., 2004) resulting in calcium influx (Demuro et al., 2005). Mitochondria from patients with AD are impaired in their ability to buffer calcium influxes and also show defects in their respiratory complexes, particularly complex IV, resulting in increased generation of reactive oxygen species (Sheehan et al., 1997;Abramov et al., 2004). Our data are consistent with these hypotheses showing that over-expression of the mitochondrial superoxide scavenging enzyme mitSOD2 offered modest but significant protection against Ab toxicity. Fig. 8. Model of Ab-mediated oxidative stress. CuZn-SOD1 breaks down superoxide free radicals (O 2 )) in the cytoplasm to produce H 2 O 2 and molecular oxygen (O 2 ). In the presence of oxygen, Fe 2+ cycles back to Fe 3+ by the Ab to produce H 2 O 2 . H 2 O 2 is neutralized into water by CAT. The heavy chain of ferritin (Fer1HC) has ferroxidase activity, which catalyses the conversion of Fe 2+ into Fe 3+ ions. Fe 3+ is subsequently stored by the light chain of ferritin (Fer2LC). Thus, ferritin has two effects: it prevents Fe 2+ from interacting with the Ab and producing H 2 O 2 and it prevents Fe 2+ from reacting with H 2 O 2 and producing the free radical hydroxyl (OH . ) via the Fenton reaction. Hydroxyl radicals can also oxidize lipids to generate long-lived reactive aldehydes. Carbonyl reductase (CAR) and GST are downstream antioxidant defences that participate in the detoxification of the reactive aldehyde species.
It was surprising that the co-expression of SOD1 enhanced the toxicity of Ab 1-42 . This result underlines the low toxicity of the superoxide anion and again emphasizes the importance of H 2 O 2 in the oxidative stress pathway. By showing that dominant negative mutants of SOD1 prolong the life of the AD flies, our findings demonstrate that SOD1 catalytic activity mediates its toxicity. Furthermore, RNAi knockdown of SOD1 protein expression did not provide any additional rescue, indicating that our data are not confounded by toxicity caused by SOD1 misfolding and aggregation as is proposed for SOD1-linked familial amyotrophic lateral sclerosis (Lynch et al., 2004). The finding of elevated carbonyl levels in the brains of flies co-expressing SOD1 and Arctic Ab 1-42 suggests that rapid dismutation of superoxide can overwhelm the capacity of endogenous CAT to remove the consequent H 2 O 2 . In addition, the dismutation of superoxide radicals releases molecular oxygen that may become a substrate for the Ab-dependent production of H 2 O 2 (Fig. 8). Nevertheless, the enhancer activity of SOD1 in Drosophila was unexpected because previous studies in a mouse model of AD showed that SOD1 knockdown caused increased activation of apoptotic pathways (Chen et al., 2005). Further work is required to determine whether SOD1 upregulation is toxic in a mammalian model of AD.
Finally, we have evaluated the effect of two enzymes that are involved in the detoxification of reactive aldehydes (GST and carbonyl reductase). Reactive aldehydes, such as 4-oxonon-2-enal, result from hydroxyl radical-mediated lipid oxidation and accumulate in the brains of patients with AD (Lovell et al., 1997;Markesbery & Lovell, 1998). Despite being less reactive than the hydroxyl radical, the greater stability of the aldehydes permits damage to proteins and DNA over longer times and distances within the cell. We have shown that over-expression of GST or carbonyl reductase gave only a modest rescue of Ab toxicity, suggesting that lipid damage is indeed downstream of the most important toxic events.
In summary, we have used genetic screens and a candidate gene approach to dissect the contribution of reactive oxygen species to the toxicity of Ab in our Drosophila model of AD. We have found that the primary oxidative stressors are likely to be H 2 O 2 and the consequent hydroxyl radical. Preventing oxidative stress, specifically by manipulating iron metabolism, provides a powerful strategy for reducing Ab toxicity in AD.

Supporting Information
Additional supporting information may be found in the online version of this article: Fig. S1. (a) Experimental design for the Affymetrix chip analysis of differential gene transcription. Fig. S2. In control experiments, the expression of the key modifiers of A toxicity in the absence of A did not have a beneficial effect on locomotor function at any time point. Please note: Wiley-Blackwell are not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.