A. I. Bush, The Mental Health Research Institute of Victoria, 155 Oak Street, Parkville, Victoria 3052, Australia Fax: +61 39387 5061 Tel: +61 39389 2962 E-mail: email@example.com
The postmortem Alzheimer's disease brain is characterized histochemically by the presence of extracellular amyloid plaques and neurofibrillary tangles. Also consistent with the disease is evidence for chronic oxidative damage within the brain. Considerable research data indicates that these three critical aspects of Alzheimer's disease are interdependent, raising the possibility that they share some commonality with respect to the ever elusive initial factor(s) that triggers the development of Alzheimer's disease. Here, we discuss reports that show a loss of metal homeostasis is also an important event in Alzheimer's disease, and we identify how metal dyshomeostasis may contribute to development of the amyloid-β, tau and oxidative stress biology of Alzheimer's disease. We propose that therapeutic agents designed to modulate metal bio-availability have the potential to ameliorate several of the dysfunctional events characteristic of Alzheimer's disease. Metal-based therapeutics have already provided promising results for the treatment of Alzheimer's disease, and new generations of pharmaceuticals are being developed. In this review, we focus on copper dyshomeostasis in Alzheimer's disease, but we also discuss zinc and iron.
Alzheimer's disease (AD) destroys the mental health of millions of people worldwide. Sufferers lose their independence and require dedicated on-going care, usually from members of their own family and at enormous economic and social cost. There is no cure for AD, and the development of effective therapeutic strategies is hampered by a paucity of information on the biological mechanisms underlying the disease. The severity of AD relates to a multitude of age-related cellular processes culminating in neuronal and synaptic dysfunction. Successful therapeutic strategies that target the causative neuropathological events will have an enormous impact on treatment of AD patients.
Four consistent features characterize the AD brain: (a) the presence of extracellular amyloid plaques comprised mainly of aggregated, insoluble amyloid-β (Aβ) peptide; (b) the presence of intracellular neurofibrillary tangles (NFTs) containing hyperphosphorylated tau; (c) increased oxidative damage to lipids, proteins and nucleic acids; and (d) a loss of biometal homeostasis. The deposition of aggregated Aβ and the hyperphosporylation of tau have been shown to cause neuronal damage and contribute substantially to the pathology of AD . Therefore, both have received considerable research attention as potential therapeutic targets. Plaques and NFTs, however, cannot be regarded as ‘up-stream’ causative factors in the development of AD. While aggregated, insoluble Aβ found within plaques can cause neurotoxicity, soluble intermediate Aβ oligomers are substantially more toxic . Similarly, NFTs can also contribute to neurodegeneration, but their development is the result of an aberrant shift in activity of tau kinases and phosphatases . Thus, although targeting plaques and NFTs may ameliorate some of the consequences of AD and no doubt lessen the burden of the disease, the biological mechanisms that caused them to develop will remain unchecked. The fundamental research aim in AD research is to identify the up-stream biological mechanisms that trigger the development of AD. Identifying these mechanisms will substantially facilitate the development of more effective therapeutic strategies.
Metal dyshomeostasis and AD
Transition metals such as Cu and Fe are essential for normal cell functionality. Due to their capacity to move between transition states, they are most abundant within redox enzymes such as Cu/Zn-superoxide dismutase (Cu/Zn-SOD), tyrosinase, and cytochrome c oxidase (COX). In these enzymes, the redox potential of the metals is harnessed to provide the enzymes with their electron transfer capabilities. Paradoxically, it is the same redox potential of the metals that makes them potentially toxic to the cell. Under conditions where they are allowed to accumulate freely, redox active metals contribute directly to cellular oxidative damage by generating the highly reactive and toxic OH• via Fenton and Haber–Weiss reactions . Cells have therefore developed sophisticated regulation and transfer systems to ensure tight control of metals within the cell [5,6]. The toxic effects of Fe and Cu and their role in numerous neurodegenerative and age-related diseases have been reviewed recently .
Equally detrimental to the cell is metal deficiency. Deficient metal bio-availability causes decreased activity of critical enzymes because activity of the enzymes is dependent on optimal metal loading. Cu deficiency is central to the recessive Menkes and Wilson diseases, and has been implicated in AD. Just as excess Cu can contribute to oxidative damage by catalysing the production of OH•, so too can deficient Cu by preventing normal activity of cuproenzymes important in maintaining cellular oxidative homeostasis.
The cortical glutamatergic synapse, where amyloid pathology first commences in AD, contains exceptionally high concentrations of Zn and Cu, which are released during neurotransmission. Zn2+ is released together with glutamate from presynaptic terminals to achieve concentrations in the order of 300 µm. The Zn2+ is concentrated into glutamate vesicles by Zn transporter-3 (ZnT3), which is only expressed in glutamatergic neurons . Ionic Cu is released into the cleft following postsynaptic stimulation of the N-methyl-d-aspartate (NMDA) receptor [10,11] and is concentrated into postsynaptic vesicles by the Menkes Cu7aATPase .
Cu dyshomeostasis is evident within the brain during normal aging [12–15], but is substantially more pronounced within the aged AD-affected brain [16,17]. Lovell et al.  demonstrated that Cu levels in the unaffected (i.e. plaque-free) neuropil of the AD brain are approximately 400% higher than in the neuropil of the healthy brain, and that within the AD brain itself Cu levels are approximately 30% higher within the amyloid plaques compared to plaque-free regions. These data indicate that an accumulation of Cu within the brain, be it a cause or consequence, is consistent with the development of AD. Despite these gross increases in cerebral extracellular Cu, intracellular Cu appears to be deficient in the AD brain . The cuproenzymes COX and peptidylglycine α-amidating monooxygenase have significantly decreased activities in AD brain and cerebrospinal fluid (CSF), respectively [19,20], and deficiencies in COX activity may be responsible for the deficit in energy metabolism characteristic of AD brain . There are conflicting reports in the literature about levels of the cuproprotein ceruloplasmin (Cp) in AD brain [22,23], but Cp activity has been reported to be decreased in plasma , despite being elevated in CSF  and in plasma . Furthermore, the antioxidant Cu/Zn-SOD shows decreased activity in both AD brain and transgenic (Tg) animal models of AD despite increased protein expression . The loss of Cu/Zn-SOD activity likely results from a deficiency in active site Cu because activity can be restored by dietary Cu supplementation .
Homeostasis of the transition metal Fe is also altered in the AD brain. It accumulates within extracellular amyloid plaques [17,28] and localizes within NFTs . Like Cu, Fe is essential within the cell because of its redox potential, and Fe dyshomeostasis can similarly contribute to cellular dysfunction when in excess (by catalysing OH• production) and when deficient (decreased enzyme activity). Of particular relevance to Fe dyshomeostasis in AD is Cp. As described above, Cp is a cuproprotein, but its functional role within the cell is to detoxify and remove excess Fe. Decreased Cp levels in the brain , possibly due to decreased Cu homeostasis, may contribute to Fe accumulation in AD  because Jeong and David  have recently shown that Cp deficiency leads to increased Fe levels in the CNS.
By contrast to Cu and Fe, Zn is redox-silent, and therefore does not contribute directly to redox reactions. Its role within the brain, however, is essential nonetheless. It is required for the activity of enzymes such as Cu/Zn-SOD  and matrix metalloproteinases  where it is required in a structural role rather than a redox-active role. Perhaps the most critical role for Zn within the brain is in neurotransmission across the glutamatergic synapse . Within the synaptic cleft, Zn concentrations can reach approximately 300 µm where it is believed to function as a counter ion for the high concentrations of glutamate present and quenches the response of the NMDA receptor . Like Cu and Fe, considerable data indicates a loss of Zn homeostasis in AD. Abnormally high concentrations of Zn are associated with amyloid plaques in AD brain [17,37,38] and AD Tg mice . In a pertinent study performed in vivo, Lee et al.  crossed AD Tg mice (Tg2576) with mice deficient in ZnT3, the protein responsible for loading Zn into synaptic vesicles for release into the synaptic cleft. These mice exhibited a 50% decrease in amyloid plaque burden compared to Tg2576 littermates, indicating that the pool of Zn essential for glutamatergic neurotransmission may contribute to plaque formation. The synaptic Zn has recently been demonstrated to be in communication with the plasma, and contributes to amyloid congophilic angiopathy, which is abolished in Tg2576 mice where the gene for ZnT3 is ablated .
After the 4.5 kDa Aβ peptide was identified as a major component of the amyloid plaques in AD brain [42,43], global AD research focused on this peptide as a causative agent in the disease. The 39–43 amino acid cleavage product of the Aβ A4 precursor protein (APP) is initially present as a soluble, unaggregated species, and it is only through the processes of oligomerization, aggregation and fibrilization that Aβ forms amyloid plaques. As amyloid plaques are prominent in the postmortem AD brain, early research theories placed the accumulation of extracellular, insoluble forms of Aβ as central to the disease process . However, as studies emerged reporting that soluble, intermediate Aβ oligomers were more toxic than fibrillar Aβ[2,45–47], it became evident that, although amyloid plaques no doubt contribute to neuronal dysfunction, the occurrence of plaques may be several steps downstream from the more critical causes of AD.
Aβ readily binds Cu and Zn via its three N-terminal histidine residues [48–51], and several compelling studies have shown that the interaction between Aβ and these metals can promote the formation of Aβ oligomers, aggregates and fibrils [48,52–55]. Metal mediated oligomerization of Aβ may therefore contribute to the potent inhibition of synaptic transmission mediated by Aβ. Several studies have now shown that Aβ mediated inhibition of synaptic transmission is dependent on the presence of Aβ oligomers, and that Aβ monomers are relatively nontoxic in these assays [56,57]. As described above, synaptically released Zn is required for the formation of amyloid plaques in Tg mice . Although amyloid plaques contain predominantly higher order Aβ aggregates, an initial formation of toxic Aβ oligomers within the synaptic cleft may be determined by Zn released from the presynaptic terminus. Similarly, Cu released into the synaptic cleft following activation of postsynaptic NMDA receptors [10,11] may also facilitate extracellular Aβ oligomerization. Shankar et al.  have reported that the loss of hippocampal synapses in rat organotypic slices is mediated by Aβ oligomers and is dependent on the activity of NMDA-type glutamate receptors.
The concentrations of metal required to induce Aβ oligomerization and aggregation are relatively low, and well within the physiological ranges that could be expected within the brain. The capacity for metals to facilitate this process may therefore be a critical factor in the Aβ mediated pathology of the AD brain. Subsequent to an early report demonstrating that the Cu- and Zn-induced aggregation of Aβ could be prevented by EDTA , disrupting Aβ–metal interactions has been an attractive therapeutic target. In this regard, a salient study demonstrated that the amyloid plaque burden in brains decreased by 49% when Tg2576 mice were treated orally with the 8-hydroxyquinoline derivative clioquinol (CQ) . CQ is a moderate metal chelator capable of crossing the blood–brain barrier, and it was believed that CQ solubilized the Aβ plaques by stripping them of their metal content. This supported the in vitro work previously reported , and was consistent with the notion that excess extracellular metals contribute to the amyloid pathology of AD.
Relative to the soluble Aβ burden of the AD brain, a recent study described a mechanism by which decreased intracellular metal bioavailability may contribute to the accumulation of soluble Aβ outside the cell. White et al.  treated Chinese Hamster Ovary cells over-expressing human APP (CHO-APP) with CQ complexed to Cu or Zn, and found that the metal-CQ treatment substantially decreased the levels of soluble Aβ present in the cell culture medium. Metal-CQ treatment decreased the levels of extracellular Aβ not by preventing an Aβ–metal interaction outside the cell, but by facilitating the delivery of metals into the cell. White et al.  demonstrated that CQ facilitated the delivery of Cu and Zn across the plasma membrane, as determined by inductively coupled plasma mass spectrometry. Once inside the cell, Cu and Zn, but not Fe, activated phosphoinositol 3-kinase mediated protein kinase pathways, which ultimately led to an increase in the secretion of matrix metalloproteinases (MMPs). The capacity for MMPs to degrade Aβ has been reported by several groups [61–64]. Treatment with CQ may therefore have a two-fold effect on Aβ; by binding extracellular metals, it prevents metal mediated Aβ aggregation and toxicity and, by then delivering the bound metals into the cell, it activates specific protein kinases that induce an increase in the production of Aβ-degrading MMPs.
Neurotoxicity generated by the interaction between Aβ and metals may be more complex than the catalysis of Aβ aggregation. Numerous studies have now shown that several potential mechanisms of neurotoxicity for soluble Aβ are exacerbated by, if not dependent on, the presence of metals. This indicates that Aβ–metal interactions, possibly occurring within the cell, may induce mechanisms of neurotoxicity that involve soluble Aβ oligomers, and that the mechanisms of toxicity precede Aβ aggregation and accumulation. Curtain et al. [49,65] demonstrated that the capacity for Aβ to bind Zn and Cu determined its ability to penetrate and disrupt membranes; Crouch et al. demonstrated that Aβ-mediated inhibition of cytochrome c oxidase requires the presence of at least equimolar concentrations of Cu , and that the inhibition was not supported by Zn or Fe ; and Huang et al.  demonstrated that the potential for Aβ to generate neurotoxic H2O2 is dependent on the presence of Cu. These metal-mediated toxic effects of Aβ were abolished by preventing the Aβ–metal interaction with chelators such as EDTA, raising the possibility that therapeutics designed to disrupt Aβ–metal interactions may prevent more than Aβ aggregation and plaque formation.
The native function of the microtubule-associated protein tau is to maintain integrity of the cytoskeleton by promoting assembly and stability of microtubules. Tau isolated from a healthy brain is partially phosphorylated [69,70], indicating that the normal function of tau requires some phosphorylation. However, in AD, tau is hyperphosphorylated, and hyperphosphorylated tau is the form that aggregates in NFTs . The loss of functional tau from the microtubule network can be compensated for by the other microtubule-associated proteins, MAP1A/MAP1B and MAP2. It is the toxic gain of function exhibited by hyperphosphorylated tau that renders it most harmful towards the cell. Hyperphosphorylated tau is capable of sequestering normal tau as well as MAP1A/MAP1B and MAP2 [72,73], and this compounding loss of essential proteins destabilizes the microtubule network, contributing to neurofibrillary degeneration.
Tau hyperphosphorylation occurs because of an imbalance in the activity of tau kinases and phosphatases . One particular tau kinase pertinent to metal dyshomeostasis in AD is glycogen synthase kinase-3 (GSK3). GSK3 has recently been implicated as a critical kinase involved in the hyperphosphorylation of tau . Only active (nonphosphorylated) GSK3 contributes to tau hyperphosphorylation, and Plattner et al.  demonstrated that the negative regulation of GSK3 (i.e. its phosphorylation) is lost in aged, but not young, Tg p25 mice. This loss of regulation resulted in an increase in GSK3 activity and tau hyperphosphorylation. Whether the change in GSK3 regulation in these mice occurred in response to an age-related decline in intracellular metals was not examined, but the study by White et al. , described above, provided evidence for a possible connection. When the bio-availability of intracellular Cu and Zn was increased in CHO-APP cells by treating with CuCQ or ZnCQ complexes, a downstream target of the activated protein kinase pathways was GSK3. By contributing to an increase in GSK3 phosphorylation, this metal mediated effect therefore decreased potential phosphorylation of tau by GSK3. The study of White et al.  did not present data on tau phosphorylation, but the possibility that metal mediated modulation of GSK3 represents a strong candidate therapeutic target for preventing tau hyperhosphorylation has been strengthened by a recent study. Malm et al.  treated AD Tg mice with the Cu ligand pyrrolidine dithiocarbamate (PDTC) and reported that the treatment increased brain Cu levels and activated the same protein kinases previously reported by White et al. . Treatment with PDTC led to an increase in GSK3 phosphorylation and a substantial decrease in tau phosphorylation . Therapeutic modulation of metal bio-availability, such as that described by White et al.  and Malm et al. , may therefore represent a potential therapeutic strategy for preventing the tau hyperphosphorylation and NFT formation characteristic of AD.
Oxidative stress in AD
A significant, consistent feature of AD is that the affected brain is under chronic oxidative stress. An early report in 1986  described an increase in activity of enzymes from the hexose monophosphate pathways in postmortem AD brain samples compared to age-matched controls, and proposed that this reflected increased oxidative stress in the AD brain. Numerous reports have since provided direct data to show extensive oxidative damage in the AD brain . Oxidatively damaged lipids, proteins and nucleic acids have all been reported [78–80].
A critical factor in oxidative stress within the AD brain is intracellular Cu. Insufficient intracellular Cu can contribute to an increase in oxidative stress (described above), and several lines of evidence indicate that intracellular Cu deficiency in AD may involve Aβ and APP. Aβ and its precursor APP both bind Cu, and over-expression of a C-terminal fragment of APP or full length APP, both containing the Aβ domain, results in an overall decrease in Cu within the brain of Tg mice . Conversely, APP knockout mice show a 40% increase in Cu levels within the cerebral cortex . Furthermore, APP gene expression is down-regulated by decreased availability of intracellular Cu  and up-regulated by increased availability of Cu . Collectively, these data present a strong case for the native role of APP/Aβ in regulating intracellular Cu: Cu alters APP gene expression [83,84], and the APP/Aβ produced binds then transports Cu out of the cell. However, once intracellular Cu levels become too low, possibly because of the aberrant increase in Aβ production consistent with AD, the antioxidant capacity of the cell may be compromised, leading to an increase in oxidative stress. The study by Busciglio et al.  in this regard is of particular relevance. These authors demonstrated that an increase in oxidative stress alters APP processing and generates an increase in Aβ production. If an oxidative stress-induced increase in Aβ production promotes excess Cu transport out of the cell, further oxidative stress due to deficient cellular Cu may be created, therefore creating a vicious cycle. Support for this possibility was presented in a recent review .
In addition to promoting oligomerization and aggregation, interactions between Cu and Aβ result in free radical generation in vitro. Synthetic Aβ reduces Cu(II) to Cu(I) with subsequent reduction of O2 giving rise to H2O2. H2O2 is itself toxic and can diffuse through the cell membrane to oxidize lipids and intracellular proteins. However, greater oxidative damage is induced when H2O2 interacts with Aβ-bound Cu(I) resulting in OH•. OH• reacts with lipids, proteins and nucleic acids, resulting in extensive modifications that are often irreversible and impede normal cellular turnover of these components. Furthermore, OH• interaction with Aβ itself can increase Aβ aggregation through the di-tyrosine-mediated cross-linking of Aβ peptides [16–18]. This is consistent with the high di-tyrosine content observed in AD brain tissue .
Oxidative stress within the AD brain is also closely related to tau hyperphosphorylation. For example, Gomez-Ramos et al.  demonstrated that the presence of acrolein, a peroxidation product from arachidonic acid, induces considerable tau phosphorylation and, in a subsequent review article, this group proposed that tau hyperphosphorylation and the formation of NFTs may even represent a normal, protective cellular response to increased oxidative stress . Furthermore, protein kinase signalling pathways sensitive to oxidative stress and known to be altered in AD have been implicated in the phosphorylation of tau . Such data indicate that an increase in cellular oxidative stress, be it through the generation of products of oxidative damage or the activation of specific cell signalling pathways, leads to tau hyperphosphorylation and NFT formation.
Oxidative stress, tau hyperphosphorylation and the Aβ biology of AD are all intricately linked, and considerable research data now exist to indicate that they interact in a series of dysfunctional mechanisms that can ultimately lead to cognitive decline. The early event(s) that initiates this neurodegenerative cycle has not been established, but the role for metal dyshomeostasis in all aspects is clear. In the AD affected brain, metal dyshomeostasis is evident in the form of a substantial increase in the levels of extracellular metals and a decrease in the levels of intracellular metals. Here, we have presented evidence to show that decreased metal bio-availability within the cell is consistent with increased oxidative stress, a loss of regulation of Aβ production, and an increase in tau hyperphosphorylation. Furthermore, an increase in extracellular metals can catalyse Aβ oligomerization and aggregation, and the amyloid plaques that subsequently form may then exacerbate intracellular metal deficiency by sequestering metals outside the cell. Figure 1 summarizes these interdependent dysfunctional events. With the loss of biometal homeostasis placed central to all of these AD-related neurodegenerative mechanisms, the modulation of metal bio-availability has strong potential in the therapeutic treatment of AD.