The role of metallobiology and amyloid-β peptides in Alzheimer’s disease


  • Blaine R. Roberts,

    1. The Mental Health Research Institute, The University of Melbourne, Parkville, VIC, Australia
    Search for more papers by this author
  • Timothy M. Ryan,

    1. The Mental Health Research Institute, The University of Melbourne, Parkville, VIC, Australia
    Search for more papers by this author
  • Ashley I. Bush,

    1. The Mental Health Research Institute, The University of Melbourne, Parkville, VIC, Australia
    2. Department of Pathology, The University of Melbourne, Parkville, VIC, Australia
    Search for more papers by this author
  • Colin L. Masters,

    1. The Mental Health Research Institute, The University of Melbourne, Parkville, VIC, Australia
    2. Centre for Neuroscience, The University of Melbourne, Parkville, VIC, Australia
    Search for more papers by this author
  • James A. Duce

    1. The Mental Health Research Institute, The University of Melbourne, Parkville, VIC, Australia
    2. Centre for Neuroscience, The University of Melbourne, Parkville, VIC, Australia
    Search for more papers by this author

Address correspondence and reprint requests to Dr James Duce, Oxidation Biology Laboratory, Mental Health Research Institute, 155 Oak Street, Parkville, Victoria 3052, Australia. E-mail:


J. Neurochem. (2012) 120 (Suppl. 1), 149–166.


The biggest risk factor for Alzheimer’s disease is the process of ageing, but the mechanisms that lead to the manifestation of the disease remain to be elucidated. Why age triggers the disease is unclear but an emerging theme is the inability for a cell to efficiently maintain many key processes such as energy production, repair, and regenerative mechanisms. Metal ions are essential to the metabolic function of every cell. This review will explore the role and reported changes in metal ions in Alzheimer disease, particularly the brain, blood and cerebral spinal fluid, emphasizing how iron, copper and zinc may be involved through the interactions with amyloid precursor protein, the proteolytically cleaved peptide amyloid-beta (Aβ), and other related metalloproteins. Finally, we explore the monomeric makeup of possible Aβ dimers, what a dimeric Aβ species from Alzheimer’s disease brain tissue is likely to be composed of, and discuss how metals may influence Aβ production and toxicity via a copper catalyzed dityrosine cross-link.

Abbreviations used

Alzheimer’s disease


disintegrin and metalloprotease



β-APP-site cleaving enzyme




sodium dodecyl sulfate


superoxide dismutase


zinc transporter

Introduction: metals in biology

In eukaryotes, the most abundant biochemically functional metals are iron (Fe) zinc (Zn), and copper (Cu). These elements are often referred to as “trace metals”; however, this description invokes a sense of insignificance. Although they exist in a small proportion compared to carbon, hydrogen, nitrogen, oxygen and phosphorous they are just as essential.

Metals are critical for enzymatic function, playing important roles in catalysis, structural stability, transport of oxygen and cellular signaling. Therefore, the value of metal ions in biology cannot be overstated and maintaining a healthy balance of essential elements like Fe, Cu and Zn is required for a functional, disease free system. While metal ions play an important role in biology, knowledge of disruption to metal homeostasis is limited to disorders such as Wilson’s disease, anemia and hemochromatosis. The continued development of specialized tools and a growing interest of the role metal ions in normal biological function and disease will provide much-needed insight into potentially a wide range of disorders. This review will focus on aspects of the role of metals in Alzheimer’s disease (AD), which has developed a large body of research regarding metal dyshomeostasis, prompted by a known interaction of metals with the neurotoxic peptide amyloid-beta (Aβ).

Altered biologically functional metal levels in the brain

Copper levels

Although the presence of Cu within the brain may be regionally different, both animal and human studies of healthy brain tissue have reported a rise in total Cu levels from youth to adulthood (Maynard et al. 2002), followed by a decrease through middle age and old age. A normal, aged neuropil contains ∼79 μM Cu, but in AD-affected cortical tissue the level is decreased (Deibel et al. 1996) and this is expected to be due to Cu’s known association with senile plaques (Miller et al. 2006). Thus, it is hypothesized that in AD Cu is abnormally redistributed to the plaques, leaving the tissue and cells deficient in Cu.

At present, the synaptic cleft is the only known microenvironment within the brain where free Cu is known to be present. At clefts containing post-synaptic NMDA receptors, Cu reaches local concentrations of 15 μM. NMDA plays an important role in achieving these concentrations through the ability to stimulate the release of Cu upon activation (Schlief et al. 2005). In these same neurons, the release of Cu has been shown to protect against excitotoxic death (Schlief et al. 2006).

Zinc levels

Of all tissues, the brain has one of the highest levels of Zn, typically ∼150 μM in grey matter (Frederickson 1989; Frederickson et al. 2005). Zn is substantially enriched in many of the glutamatergic nerve terminals, where it can reach millimolar concentrations before being released upon neuronal activation (Frederickson 1989). Once released into the synaptic cleft, where Zn can again achieve millimolar concentrations, it interacts with neuronal receptors such as NMDA (Smart et al. 2004) as well as various neuronal ion channels and transporters to regulate neuronal transmission (Weiss et al. 1993; Inoue et al. 2010). Although there is little to no change in global Zn content within the ageing brain (Maynard et al. 2002), specific areas known to be high in Zn (e.g. the highly glutamatergic innervated hippocampus) do show age related decreases in the zinc transporter (ZnT3) transporter that concentrates Zn into the glutamatergic vesicles (Adlard et al. 2010). In AD, it is well established that Zn is highly enriched within AD plaques compared with normal age-matched neuropil (Lovell et al. 1998; Miller et al. 2006; Leskovjan et al. 2011) and thus the entrapment of Zn within the plaque would potentially lead to a mislocalization of functional Zn from the rest of the brain.

Iron levels

An average adult brain contains approximately 60 mg of non-heme bound Fe, with the substantia nigra, globus palladus, caudate nucleus and putamen retaining well above the average levels (Bartzokis et al. 1997; Martin et al. 1998). In addition to being important for several biological functions (e.g. oxygen transport and electron transfer), Fe is specifically required within the brain for neurotransmitter synthesis and myelin production (Takeda 2004). Recent studies have demonstrated that Fe is essential for ryanodine receptor-mediated calcium release after NMDA receptor stimulation, which in turn promotes ERK1/2 activation, an essential step of sustained hippocampal long-term potentiation (Munoz et al. 2011). Age-related increases in brain Fe has been found in all species examined, including humans (Bartzokis et al. 1994; Roskams and Connor 1994; Maynard et al. 2002; Hardy et al. 2005), and electron paramagnetic imaging has shown that these increases are regionally clustered (Wender et al. 1992).

In AD, Fe associated with senile plaques increases to nearly three times that of normal neuropil levels (940 μM compared with 340 μM) (Goodman 1953; Connor et al. 1992b; Lovell et al. 1998; Bishop et al. 2002; Collingwood and Dobson 2006). However, outside of the plaque, there has been contention within the literature about the level of Fe in AD brain (Schrag et al. 2011). While doubts in measurement accuracy of total Fe levels within the brain continue, it is important to note that the absolute changes in the total level of Fe may not be as important as the balance between ferric and ferrous iron species. Although unligated Fe is the main component of the labile iron pool that drives translation of several mRNA species, it can also catalyze dangerous reactive oxygen species. Indeed, a mere excess in the amount of Fe may not be toxic to the cell if the correct mechanisms for managing the Fe are fully functioning.

Changes in biologically functional metals in CSF, serum and plasma

Copper levels

Levels of plasma and CSF Cu display an initial sharp rise immediately postnatal, and then proceed to increase steadily with age (McMaster et al. 1992; Milne and Johnson 1993; Madaric et al. 1994; Ekmekcioglu 2001). According to a recent meta-analysis, this increase is exacerbated in AD, and is attributed to increased non-ceruloplasmin Cu (Squitti et al. 2002; Bucossi et al. 2011). These results are consistent with an age-related systematic loss in Cu regulation (Rossi et al. 2007).

Zinc levels

In contrast to Cu, Zn levels in plasma are reported to steadily decrease with age from birth (Bunker et al. 1987; McMaster et al. 1992; Martinez Lista et al. 1993; Madaric et al. 1994) and the presence of Zn ions in serum and blood is further decreased in AD compared with age-matched control patients (Basun et al. 1991; Baum et al. 2010; Brewer et al. 2010). Conversely, the level of Zn in CSF has been reported to be elevated, reflecting the rise observed within the brain (Deibel et al. 1996; Cornett et al. 1998; Religa et al. 2006).

Iron levels

In plasma and serum, Fe is mostly bound to hemoglobulin and serum transferrin. Accordingly serum transferrin levels are used clinically to assess Fe status. Plasma/serum levels of Fe decrease with age, as do iron-containing proteins such as hemoglobin (Ahluwalia et al. 2000). Reports of Fe levels in plasma, serum and CSF have yet to show a consistent change in AD (Smorgon et al. 2004; Strozyk et al. 2009).

Additional metals of interest

Cobalt is an important component for the cofactor cobalamin (vitamin B12) and is required for a normal functioning nervous system, DNA synthesis and energy metabolism. Deficiency in cobalamin can result in neurological symptoms (Weir and Scott 1999) and accelerate deposition of Aβ in transgenic mouse models of AD (Zhuo and Pratico 2010). In human serum, the level of cobalt has been shown to positively correlate with mini-mental state exam (Smorgon et al. 2004), supporting a role for cobalt in cognition. Furthermore, cobalamin deficiency impairs homocysteine catabolism and induces folate deficiency. Elevated levels of homocysteine can double the risk of developing AD (Seshadri et al. 2002) whereas folate deficiency may contribute to the progression of AD (Prodan et al. 2009; Siuda et al. 2009).

A biological role for chromium in mammals was first discovered in rats (Schwarz and Mertz 1959) and later in humans (Anderson 1995) where a deficiency led to impaired efficiency to remove glucose from the blood stream and the development of diabetic symptoms. In the CSF of AD patients, chromium levels are inversely correlated with Aβ1-42 concentration similar to what is observed with Cu, Fe, Zn and manganese (Strozyk et al. 2009), whereas in AD serum chromium, cobalt, selenium and Fe were positively correlated with cognition (Smorgon et al. 2004). Thus, depressed levels of chromium or Fe in either CSF or serum appear to be indicators of a disease phenotype.

Arsenic in humans is typically regarded as toxic and the two main sources for humans are drinking water and food (Tao and Bolger 1999; Smith et al. 2002). With respect to AD, it is hypothesized that environmental exposure to arsenic could play a causal role in the development of AD (Gharibzadeh and Hoseini 2008). Although AD and control patients display no significant difference in the serum level of arsenic, arsenic levels are inversely correlated with mini-mental state exam scores (Baum et al. 2010) and high levels can induce hyperphosphorylation of tau (Giasson et al. 2002) and APP transcription (Dewji et al. 1995). Further study is required to understand the relationship between arsenic and cognition.

Lead and mercury are the two most common metals associated with acute heavy metal toxicity and acute exposure to either metal results in well-defined neurological symptoms (Clarkson et al. 2003; Sanders et al. 2009) However, lower level exposure of mercury and lead are implicated in the etiology of AD (for review see Mutter et al. 2010). In addition, there is evidence that early exposure to lead can contribute to the elevation of DNA methylation, Aβ production and APP expression in old age (Basha et al. 2005; Wu et al. 2008). Inorganic mercury has a very high affinity for selenoproteins (Kd = 10–45) (Ganther 1980) as well as thio-groups (Carvalho et al. 2008; de Freitas et al. 2009), both heavily involved in maintaining the redox balance of a cell, and implicates mercury as an environmental insult that may contribute to the oxidative stress observed in AD. Mercury’s effect in AD is exacerbated by the brain being a major target organ for this metal and, once in the brain, it has a half-life estimated to be as long as 20 years. There are conflicting reports about the elevation of lead and mercury in blood, CSF, urine and other biological tissues in AD and future studies on both mercury and lead are required to determine how these two metals may contribute to the incidence of AD.

Changes in level of metalloproteins

Perturbation of the levels of total metal and a redistribution of metals to plaques that leave the cell deficient, lead to changes of metalloprotein stability and a deficiency in metalloenzymes (De Deyn et al. 1998; Omar et al. 1999; Maurer et al. 2000; Cottrell et al. 2001).

Copper levels

Ceruloplasmin, albumin, and transcuprein are the major copper-binding proteins that transport Cu in plasma. Although increases in Cu in AD have been suggested to be independent of protein levels (Squitti et al. 2005), expression of the predominant Cu-binding protein ceruloplasmin correlates well with the changes in plasma Cu during ageing (Milne and Johnson 1993).

At the cell surface, transport across cellular membranes is predominantly through the copper importer CTR1 and the exporters ATP7a (Menkes) or ATP7b (Wilson). ATP7a is required to facilitate Cu transport across the blood–brain barrier via endothelial cells and its expression peaks during synaptogenesis (Qian et al. 1998; El Meskini et al. 2007). ATP7a concentrates Cu into post-synaptic vesicles in glutamatergic synapses, which release Cu into the synaptic cleft after NMDA receptor activation (Schlief et al. 2005). The specific mechanism involved in the reuptake of Cu released into the synapses is still unknown but is likely to be energy-dependent (Giese et al. 2005), and potentially mediated by presenilin and CTR1 (Greenough et al. 2011).

In addition, there are several intracellular cytosolic Cu chaperones that deliver Cu to specific molecules for normal cellular function (Tapiero et al. 2003) and if Cu is incorrectly delivered then significant consequences occur that may lead to cell death.

Zinc levels

A multitude of ZnTs, zinc-importing proteins and buffering proteins such as the metallothioneins bind cytosolic Zn to prevent excessive levels of free Zn from becoming toxic (for review see Sensi et al. 2009). In addition, diverse classes of proteins also require Zn for normal cellular function, including Zn metalloenzymes, transcription factors and signaling proteins (Vallee et al. 1991; Szallasi et al. 1996). The zinc transporter ZnT3 is present in glutamatergic synapses of hippocampal and neocortical neurons (Palmiter et al. 1996) and is essential for loading Zn into synaptic vesicles (Linkous et al. 2008). ZnT3 levels decrease with age (Adlard et al. 2010), and, as ZnT3 is the only confirmed synaptic vesicle Zn transporter, its impact on cognitive function (Adlard et al. 2010) is likely to be mediated through the downstream effects of synaptic Zn on signaling pathways. In AD brain tissues, ZnT3 levels are decreased (Adlard et al. 2010), and the levels of the other Zn transporter proteins ZnT1, ZnT4 and ZnT6 are altered (Lyubartseva et al. 2009). A study by Lee et al. (2002) examined the influence of ablating synaptic vesicular Zn on Aβ burden by crossing an AD model (transgenic mice carrying the Swedish mutation in APP, Tg2576) with ZnT3 KO animals. The resulting double-mutant mice demonstrated deficits in synaptic Zn, and a marked reduction in Aβ plaques and precipitated Aβ in the cortex and hippocampus. The plaques that were present in the double-mutant mice were smaller than the standard Tg2576 mice, and soluble Aβ levels elevated, supporting a role for synaptic Zn in the precipitation of Aβ. A subsequent study of ZnT3 ablation in this model showed that Zn egress from the synapse also contributes to congophilic angiopathy (Friedlich et al. 2004).

Iron levels

The regulation of Fe by iron-associated proteins is of equal or greater importance to total levels of Fe in the brain. Dysregulation of brain Fe metabolism is multifactorial and can comprise of non-genetic and genetic factors. It may also occur at multiple levels, including Fe uptake and release, storage, intracellular metabolism and regulation (Ke and Qian 2007). Fe transport into tissue, including across the blood brain barrier, requires the Fe carrier, transferrin. Circulating Fe, once oxidized to the safer ferric state through a ferroxidase (e.g. ceruloplasmin or APP), is complexed with transferrin, which in turn binds to the transferrin receptor on the extracellular surface of the membrane (or the lumen side of endothelial cells in the cardiovascular system; Morris et al. 1992). The transferrin complex is then endocytosed and either utilized, compartmentalized safely within ferritin, or immediately effluxed from the cell via ferroportin, which also requires a ferroxidase to facilitate the release of Fe into the extracellular space. Fe homeostasis is highly regulated and the major proteins of importance in the Fe pathway tend to be controlled translationally. Cellular translation is responsive to the cytoplasmic free Fe levels (the Labile Iron Pool), which governs the binding of iron regulatory proteins to iron responsive protein mRNA in a canonical cis-trans iron regulatory system (Klausner et al. 1993). When cellular Fe levels are high, translation of APP as well as ferritin and ferroportin is increased to ensure the safe storage and efflux of Fe, while RNA for the Fe importer transferrin receptor is degraded to stop Fe import (Kikinis et al. 1995; Rogers et al. 2002).

Thus, it is not surprising that there is an age-related increase in ferritin expression which correlates with the increase in Fe, presumably to store the surplus intracellular Fe (Connor et al. 1990, 1992a; Zecca et al. 2001). Accordingly, ferritin levels are further increased in AD, whereas transferrin is decreased (Connor et al. 1992a). The increased Fe in AD is found primarily complexed with ferritin in the plaque-associated neuritic processes (Grundke-Iqbal et al. 1990) and within neurons with neurofibrillar tangles (Morris et al. 1994; LeVine 1997) suggesting that there is association between AD pathological markers and Fe accumulation.

The translational control of APP by Fe, and the ability of Fe to affect APP processing and formation of Aβ oligomers (discussed further below), implies it may be involved in Fe homeostasis and, while there are presently no known high affinity Fe binding sites in APP, this may increase Aβ toxicity in experimental models (Liu et al. 2011) and we recently observed that APP contains a REXXE ferroxidase consensus motif also found within another ferroxidase, heavy-chain ferritin (Duce et al. 2010). Like most ferroxidases, APP’s ability to oxidize Fe2+ to Fe3+ facilitates the efflux of Fe out of the cell, and suppression of APP in neurons (and other cell-types) induces marked Fe retention. Interestingly, APP-specific ferroxidase activity is inhibited by Zn, an event of significance within AD tissue (Duce et al. 2010). Also related to Fe, APP is able to bind heme oxygenase independent of the ferroxidase motif, inhibiting the antioxidant and neuroprotective functions of this protein. Familial AD-related mutations within the APP sequence enhance this inhibitory role and may contribute to neuronal death in AD (Takahashi et al. 2000). APP’s role in Fe homeostasis may also be a clue as to why recent studies have shown that there are genetic risk factors for AD that are also critical to Fe metabolism. These include polymorphisms of the hereditary hemochromatosis gene, which is responsible for excess Fe in hemochromatosis, and the transferrin C2 allele (Namekata et al. 1997; Moalem et al. 2000). Individuals with compound C2 transferrin and hemochromatosis polymorphisms have a 500% increased risk of developing AD (Robson et al. 2004; Bertram et al. 2007).

Disruption in brain Fe homeostasis through alterations of Fe regulatory proteins can increase the vulnerability of cells to oxidative stress (Connor et al. 1992b; Pinero et al. 2000), which the pioneering research of the late Mark Smith and colleagues, proposed to be involved in the onset, progression and pathogenesis of AD (Castellani et al. 2007; Zhu et al. 2007). Notably, over-expression of heavy chain ferritin affords a complete rescue of the locomotor deficits that result from mutant APP over-expression in AD-related mouse models (Rival et al. 2009) and divalent metal transporter 1 may be involved in this AD-related altered Fe homeostasis through its ability to alter APP processing and Aβ generation (Zheng et al. 2009). Changes in superoxide levels caused by alteration of superoxide dismutase (SOD1) activity may also affect Fe metabolism in glial and neuronal cells (Culotta et al. 2006; Danzeisen et al. 2006). Lactoferrin, a Fe storage protein, which is up-regulated in AD (Wang et al. 2010), also exerts an anti-inflammatory function via its inhibitory effect on hydroxyl radical formation and, through its anti-oxidative properties, prevents DNA damage (Sacharczuk et al. 2005).

APP processing and metals

APP is predominantly processed in healthy brain through a route known as the ‘non-amyloidogenic pathway’ whereby APP is cleaved by the α-secretase within the Aβ region producing a secreted ectodomain fragment (sAPPα) and a membrane bound C-terminal fragment, which is subsequently cleaved by the γ-secretase complex to produce smaller fragments that can be safely degraded. However, APP is also processed via an ‘amyloidogenic pathway’. In this pathway APP is sequentially cleaved by the β-secretase, instead of α-secretase, and then γ-secretase. β-secretase cleavage results in the formation of secreted APP beta (sAPPβ) and a membrane bound C-terminal fragment that with further processing by the γ-secretase complex leads to Aβ peptides of varying lengths.

APP processing through either the non-amyloidogenic or amyloidogenic pathways may be affected by metals (Cater et al. 2008). α-Secretase activity is attributed to members of the cell surface metalloproteinase family of disintegrin and metalloprotease (ADAM) proteins. ADAMs require Zn to function and are known to have a long zinc-binding consensus sequence containing several Zn ligands (Edwards et al. 2008). ADAM 9, 10, 17 (also known as tumor necrosis factor-alpha converting enzyme) and 19 all have α-secretase properties (Allinson et al. 2003; Asai et al. 2003). ADAM 10 is currently the strongest candidate as it is actively present in primary neurons (Kuhn et al. 2010) and over-expression in an AD transgenic mouse model increases sAPPα production while reducing Aβ and cognitive deficits (Postina et al. 2004). However, other members of the ADAM family may have a compensatory effect, as there was no evidence of change in sAPPα production in fibroblasts deficient in ADAM10 (Hartmann et al. 2002). As well as α-secretases dependence on Zn, Fe can modulate activity through furin, a proconvertase involved in the regulation of α-secretase-dependent processing (Silvestri and Camaschella 2008). Currently, Cu does not seem to have a direct relationship with α-secretases; however, it has been shown to affect β-secretases. β-APP-site cleaving enzyme (BACE) is a membrane anchored aspartyl protease (Sinha et al. 1999; Vassar et al. 1999; Yan et al. 1999; Cai et al. 2001; Luo et al. 2001) that is able to bind Cu through C-terminal cysteines. BACE1 also interacts with domain I of the copper chaperone for superoxide dismutase via Cu and this interaction may directly compete with copper chaperone for superoxide dismutase’s ability to incorporate Cu into SOD1 as over-expression of BACE1 reduces SOD1 activity (Angeletti et al. 2005). Depending on the cell type, Cu deficiency has been shown to alter the processing of APP to favor the amylodogenic pathway or reduce the clearance of Aβ leading to an elevation in secreted Aβ (Cater et al. 2008). The third secretase involved in both APP processing pathways is the high molecular weight γ-secretase complex comprising of at least four components, presenilin 1 or 2, anterior pharynx defective-1 (Aph-1), presenilin enhancer 2 and nicastrin (Nct) (De Strooper 2003). As with α-secretases, Zn may again be important, as it has been shown to enhance the synthesis of the presenilin subunit (Park et al. 2001) as well as facilitate the oligomerization and inhibit the cleavage of a γ-secretase substrate (Hoke et al. 2005). Recently, presenilins have been shown to play a major role in the turnover of cellular Cu and Zn, influencing the activity of copper- and zinc-dependent proteins such as SOD1 (Greenough et al. 2011).

Furthermore, dimerization of APP has been shown to affect cleavage by α-, β- and γ-secretases whereby increasing homodimerization of APP leads to reduced Aβ production (Scheuermann et al. 2001). To date at least four domains have been reported to promote APP dimerization, the heparin-binding motif in the E2 domain (Gralle et al. 2006; Lee et al. 2010), the APP juxtamembrane region (Shaked et al. 2006), the GXXXG motif near the luminal face of the transmembrane region (Kienlen-Campard et al. 2008) and the Cu binding motif within the amino-terminal E1 domain (Kong et al. 2008) that consists of four ligands (His-147, His-151, Tyr-168 and Met-170) (Barnham et al. 2003). In addition, dimerization may also depend on Zn binding through a conserved region of amino acids between position 170 and 188 (Bush et al. 1993; Scheuermann et al. 2001). This Zn-binding domain consists of two key cysteine ligands at position 186 and 187, as well as other potential ligands (e.g. Cys-174, Met-170, Asp-177 and Glu-184) (Ciuculescu et al. 2005).

Aβ binding to metal

With the discovery that Cu, Zn and Fe ions cause Aβ to precipitate (Bush et al. 1994a; b; Atwood et al. 1998, 2000) it has become clear that Aβ is a metalloprotein (Opazo et al. 2002; Dong et al. 2003; Bolognin et al. 2011). Aβ has both a low and high affinity-binding site for either Cu or Zn (Bush et al. 1994a; b; Atwood et al. 1998, 2000) but the Kd of Aβ affinity for these metals has been an issue of debate. Originally, the high affinity binding was reported to be ≈100 nM, whereas ≈5 μM was shown for the low affinity binding site (Bush et al. 1994a; b). However, it is now understood that the buffer conditions (e.g. the presence of NaCl; Huang et al. 1997), the aggregation state of the peptide (Bush et al. 1994a; Garai et al. 2006) and how the bound and free metal ions are assayed (Tougu et al. 2008) are critical for the values observed. Interestingly, the apparent high affinity Cu binding constant for Aβ is different between Aβ1-42 (7.0 × 10−18 M) and Aβ1-40 (5.0 × 10−11 M) but may reflect a perturbed equilibrium brought about by the precipitation of the peptide (Atwood et al. 2000). At pH 7.4 Aβ binds equimolar amounts of Cu and Zn, however, under slightly more acidic conditions (pH 6.6) Cu completely displaces the Zn from Aβ (Atwood et al. 2000). In similar acidic conditions, Fe is able to bind Aβ and this may be an important factor for Aβ toxicity, as lysosomes (pH 5.5) are key for Fe entry into neurons, and have been implicated as a major factor in the toxicity of Aβ (Ditaranto et al. 2001; Liu et al. 2010). The recently observed in vivo phosphorylation of Aβ has strong potential implications for the binding of Fe to Aβ, as phosphate groups dramatically enhance a peptides ability to bind Fe (Kumar et al. 2011), and Fe is suggested to promote toxicity within experimental models of AD through its ability to delay the ordered Aβ aggregates found in plaques (Liu et al. 2011). Aβ’s affinity for metal is compatible with the concentrations of free ionic release by synaptic neurotransmission (Lee et al. 2002; Schlief et al. 2005; Munoz et al. 2011). This could account for the synaptic localization of Aβ precipitation (Deshpande et al. 2009) and the effect of transition metals in synaptic damage induced by Aβ (Uranga et al. 2010).

The stoichiometry of metal binding to Aβ is 2.5 equivalents, the fractional binding of suggests that Aβ oligomers mediate an additional metal binding site (Atwood et al. 2000). Binding of the oxidized forms of Cu, Fe and Zn (Bush et al. 1994b; Atwood et al. 2000; Syme et al. 2004; Danielsson et al. 2007; Chen et al. 2011) as well as the reduced form of Cu (Shearer and Szalai 2008) to Aβ is mediated by nitrogen ligands from the three histidine residues at positions 6, 13 and 14, along with an oxygen ligand (Curtain et al. 2001, 2003). The candidate oxygen ligand for metal coordination is most likely from aspartate at amino acid 1 (Dorlet et al. 2009; Drew et al. 2009b) and alanine 2 (Dorlet et al. 2009; Drew et al. 2009a). While other oxygen ligands have been proposed, such as aspartate 5 (Zirah et al. 2006), serine 8 (Danielsson et al. 2007), tyrosine 10 (Curtain et al. 2001) and glutamate 11 (Zirah et al. 2006) these have been contended (Syme et al. 2004; Karr and Szalai 2007; Faller and Hureau 2009), especially as peptides lacking the first one to three amino acids do not bind Cu in the same fashion as the native peptide (Karr et al. 2005). More detailed reviews of the metal binding chemistry of Aβ have been recently published (Faller and Hureau 2009; Rozga and Bal 2009).

Metal and Aβ-mediated production of reactive oxygen species

When the oxidized forms of Fe or Cu bind Aβ they become reduced and can produce hydrogen peroxide through subsequent reduction of molecular oxygen (Huang et al. 1999; Opazo et al. 2002; Tabner et al. 2002; Nelson and Alkon 2005). The redox chemistry that is facilitated by the binding of Fe or Cu to Aβ is critical to the oxidative stress induced toxicity that is observed in cell culture and a large body of evidence shows that oxidative injury, mediated by hydrogen peroxide, has a significant role in AD. The generation of the oxidative markers observed in AD can also stem from the reaction of hydrogen peroxide and reduced Fe or Cu to generate hydroxyl radicals by Fenton chemistry. Hydroxyl radicals are highly chemically reactive and in turn generate lipid peroxidation products, protein carbonyl modifications, and nucleic acid adducts such as 8-hydroxy guanosine, all of which are characteristic of AD neuropathology (Smith et al. 1996, 1997).

Aβ’s redox cycling produces hydrogen peroxide in the presence of biological reducing agents. Cholesterol and long-chain fatty acids are the most likely reductants (Opazo et al. 2002; Barnham et al. 2004; Haeffner et al. 2005; Murray et al. 2005; Nelson and Alkon 2005; Puglielli et al. 2005; Smith et al. 2006) as the toxicity of Aβ is associated with the membrane (Ciccotosto et al. 2004) and lipid oxidation products, including oxysterols and 4-hydroxynonenal (HNE), are elevated in AD brain tissue and mouse models of the disease (Opazo et al. 2002; Haeffner et al. 2005; Nelson and Alkon 2005; Puglielli et al. 2005; Smith et al. 2006). Cellular Cu deficiency causes Aβ to localize to cholesterol rich lipid rafts (Hung et al. 2009). In addition, Cu deficiency increases the level of Cu in the lipid rafts and thereby provide a favorable environment for the formation of Aβ:Cu complexes (Hung et al. 2009). Interestingly, HNE can covalently modify the histidines of Aβ resulting in a HNE-Aβ species having a further affinity for lipid membranes (Murray et al. 2005). Conversely, isoforms of apolipoprotein E can prevent the metal mediated toxicity of Aβ in a manner that is inversely proportional to their disease risk (e.g. ApoE4 prevents the least amount of Aβ aggregation) (Moir et al. 1999) suggesting a relationship between Aβ, lipids and the binding of Cu or Fe (for a further review on this see Adlard and Bush 2011).

Copper catalyzes the oxidation of a number of Aβ side chains, promoting the addition of oxygen atoms to form methionine sulfoxide and methionine sulfone at position 35 (Ciccotosto et al. 2004; Ali et al. 2005) as well as modifying lysines at 16 and 28 (Chen et al. 2007). In the presence of Cu and H2O2, tyrosine at position 10 is an additional target for metal-mediated redox chemistry (Barnham et al. 2004; Haeffner et al. 2005) and multiple tyrosine modifications including dityrosine and nitrotyrosine have been observed (Hensley et al. 1998; Reynolds et al. 2005). Oxidized products of histidine and N3-pyroglutamate have also been isolated from AD plaques and interestingly the positron emission tomography imaging ligand Pittsburgh compound B (PIB) has a particularly high affinity for the pyroglutamate modified form of Aβ (Maeda et al. 2007).

Under ideal conditions, synthetic Aβ1-42 peptide treated with excess amounts of Cu and ascorbate yields less than ∼10% Aβ dimeric material (Atwood et al. 2004) and, in general, enzymatically catalyzed dityrosine formation with peroxidases only achieves a ∼30% yield (Malencik et al. 1996). Despite the formation of dityrosine being relatively low yielding in vitro, there is a 5- to 8-fold increase in dityrosine observed in AD brain which suggest that conditions are altered to favor oxidative modifications such as dityrosine formation and nitration (Smith et al. 1997; Hensley et al. 1998). In AD brain, where both Aβ and Cu are predominantly localized in amyloid plaques, there is evidence that the fibrillar structure of Aβ is favorable for dityrosine formation (Yoburn et al. 2003), suggesting that fibrils may still play a role in the mechanism of small oligomer formation. However, it is still unknown if the increased dityrosine involves Aβ or other proteins, and future studies will be required to determine if dityrosine cross-linking between Aβ peptides actually occurs within the brain.

The length of the Aβ species has also been regarded as an important factor in AD pathogenesis, largely due to its enrichment in amyloid deposits (Masters et al. 1985a; b; Kang et al. 1987) despite having relatively low abundance in biological fluids (Vigo-Pelfrey et al. 1993). Interestingly, the redox activity of Aβ is greatest for Aβ42human > Aβ40human >> Aβ40mouse ≈ 0 (Huang et al. 1999) and this relationship is also relevant because Aβ1-42 is overproduced in some familial forms of AD. Redox activity and the length of Aβ also correlate with the neurotoxicity observed in neuronal cultures, which is largely mediated by the Cu : Aβ interaction (Huang et al. 1999; Opazo et al. 2002). Chelators, like TETA and clioquinol prevent the redox activity of Cu or Fe and thus block the toxicity of Aβ in cell culture (Puglielli et al. 2005).

Metals involvement in aggregated Aβ neurotoxicity

Plaques are the defining pathological hallmark of AD and were initially thought to be central to the neurodegenerative process that occurs in AD. Many different species of Aβ from dimer to fibril have been subsequently proposed to be the toxic form through mechanisms that vary from receptor-mediated disruption via NMDA (De Felice et al. 2007; Venkitaramani et al. 2007) and gamma-aminobutyric acid (GABA) (Paula-Lima et al. 2003; Louzada et al. 2004) through to were Aβ causes membrane leakage (Demuro et al. 2005; Chang et al. 2011). As there is little correlation between plaque content and cognitive impairment within the brains of human and animal models, there has been increasing interest in soluble oligomers, which appear to be especially toxic (Lesne et al. 2006; Shankar et al. 2008). There are a wide range of oligomeric preparations described in the literature including amyloid derived, diffusible ligands, globulomers, 56*, and amylospheroids to name just a few. While in all cases it is uncertain how these oligomers form in vivo, there is increasing evidence that metals can play a pivotal role.

The aggregation propensity of Aβ is dramatically increased by Zn, Cu or Fe and analyses of plaques and congophilic angiopathy from both AD and related mouse models have demonstrated that these metals are highly enriched in these structures (Lovell et al. 1998; Dong et al. 2003; Friedlich et al. 2004; Miller et al. 2006; Stoltenberg et al. 2007). Upon binding Aβ, Zn rapidly precipitates the peptide over a broad pH range (Bush et al. 1994b; Atwood et al. 1998; Cherny et al. 1999), while Aβ aggregation induced by either Cu or Fe occurs only under mildly acidic conditions (e.g. pH 6.8–7.0) similar to that found in endosomes (Atwood et al. 1998, 2000; Ha et al. 2007).

At levels detected in the synapse, Zn is able to facilitate aggregation by binding to Aβ hisitidines at position 13 and 14 for spontaneous coordination of both intra- and inter-molecular bridging between peptides (Dong et al. 2006; Deshpande et al. 2009; Miller et al. 2010). Cu and Fe under mild acidic conditions also intramolecularly bind these histidines (Syme et al. 2004; Minicozzi et al. 2008). Interestingly, the rat and mouse Aβ sequence have structural changes at histidine 13, arginine 5 and tyrosine 10 that mitigate metal ion coordination (Gaggelli et al. 2008) and lower the aggregation propensity (Bush et al. 1994b). This may explain why these animals are exceptional among mammals for not forming cerebral Aβ deposits with age (Vaughan and Peters 1981).

Since the discovery of Aβ as a primary component of plaques found in AD and downs syndrome patients (Glenner and Wong 1984; Masters et al. 1985b) the Aβ1-40/Aβ1-42 amino acid residue peptides have been regarded as the primary cause of the disease. The Aβ peptide from the human brain is hard to define, as there are over 40 Aβ peptides that have been described from in vivo extraction (Sergeant et al. 2003; Van Vickle et al. 2008; Portelius et al. 2010). Despite this the overwhelming majority of research has focused on Aβ1-40 and Aβ1-42 with the latter being regarded as the most neurotoxic of the pair. A greater affinity of Aβ1-42 for Cu over smaller Aβ peptide lengths lacking the C-terminal also corresponds with enhanced precipitation (Atwood et al. 1998, 2004), increased formation of sodium dodecyl sulfate (SDS)-resistant dimerization (Atwood et al. 1998) and increased redox activity and toxicity of the bound Cu (Huang et al. 1999). However, recent analysis of the Cu-binding domain in the 3–42 truncated Aβ (Aβ3-42) sequence has also demonstrated that it self-aggregates and nucleates more rapidly than Aβ1-42, and these properties are exaggerated by Cu (McColl et al. 2009).

Determining the most physiological dimeric species

An 8 kDa Aβ species from AD brain extracts is observed to be associated with toxicity (Walsh and Selkoe 2007; Shankar et al. 2008). This band is believed to be an SDS-resistant dimer of Aβ; however, what the monomeric components of this dimer are, and what stabilizes this species against denaturation, are yet to be identified. There are at least two possible mechanisms that may stabilize this species, either the formation of a strong ionic complex or the formation of a covalent cross-link (e.g. dityrosine, see above). Soluble species of Aβ from brain are typically defined as the material remaining in solution after a 100 000 g centrifugation and range in molecular weight from large oligomeric species (> 100 kDa) to a portion of dimer and monomeric material. In the presence of denaturants such as 5 M guanidine or 7 M urea with 2 M thiourea (data not shown), the large molecular weight species of soluble Aβ are reduced to monomers (∼4–5 kDa) and the 8 kDa band commonly described as a dimer (Fig. 1). As this 8 kDa dimeric band is resistant to a variety of denaturants such as hexafluoroisopropanol, urea, guanidine, SDS and formic acid (Podlisny et al. 1995; Walsh et al. 2002; Lesne et al. 2006) the most likely explanation is that this species is covalently cross-linked.

Figure 1.

 Size exclusion chromatography of TBS soluble Aβ species. AD brain was homogenized in Tris-buffered saline (TBS) (pH 7.4) (1 : 4 w/v) and centrifuged at 100 000 g to obtain soluble Aβ. The TBS soluble material was separated based on molecular size under native (a) or denaturing (b) conditions using a superdex 75 GL size exclusion column (10 × 300 mm, GE) equilibrated either in TBS (a) or 6 M guanidine, 50 mM Tris (pH 8.0) (b). The inclusion of denaturing chaotropic salts such as guanidine or urea (data not shown) result in the breakdown of Aβ with apparent molecular weight > 100 kDa into two species of ∼8 and ∼4 kDa, referred to as dimeric and monomeric Aβ, respectively.

As explained earlier, Cu and Fe bound to Aβ can catalyze the oxidative modification of Aβ to form covalently cross-linked dimers. In the presence of cellular reductants, these metals catalyze the reduction of oxygen to H2O2 (Huang et al. 1999; Opazo et al. 2002) and, in the case of Cu, produce a dityrosine cross-linked Aβ species (Atwood et al. 2004; Barnham et al. 2004). Transgenic mice expressing the Swedish mutation of APP do not form apparent dimers until 6–8 months of age (Kawarabayashi et al. 2001), which may reflect the elevation of brain Cu that becomes prominent in the brains of mice at this age (Maynard et al. 2002). In humans, an Aβ species with a mass consistent of an oxidized dimer of Aβ1-42 has been observed in plasma to correlate well with disease progression (Villemagne et al. 2010) and the AD brain contains the hallmarks for high level oxidative and nitrative stress (Smith et al. 1997). The elevated levels of reactive species like hypochlorous acid (HOCl), peroxynitrite (ONOOH), hydrogen peroxide and hydroxyl radicals all have the potential to cause the formation of oxidized and dityrosine cross-linked Aβ species. Furthermore, peroxidases can catalyze the dityrosine cross-link between Aβ (Galeazzi et al. 1999; Nagano et al. 2004) in a similar way that peroxidases catalyze dityrosine formation for other proteins such as calmodulin (Malencik and Anderson 1994).

Characterizing the chemical composition of the 8 kDa dimeric band of Aβ in AD brains may be critical to the future development of therapies and diagnostic tools. This is particularly true as studies have shown that small oligomers have the ability to bind to membranes that assemble to form larger oligomers (Nag et al. 2010) thus dimers may represent the smallest oligomeric unit of Aβ (Streltsov et al. 2011). Although the formation of dimeric species of Aβ is currently an area of intense research, there are a few peculiar properties that may give a clue to its form. For example, the average molecular weight of Aβ1-42 is 4514.1 Da thus a dityrosine cross-linked dimer of Aβ1-42 would have a mass of 9026 Da and as a branched peptide it would be reasonable to expect that the dityrosine cross-linking of this peptide may retard migration on a SDS–polyacrylamide gel electrophoresis gel. However, the calculated molecular weight from SDS–polyacrylamide gel electrophoresis and western blot techniques demonstrate that the molecular weight of the potential dimer band is ∼8 kDa (Fig. 1) and not the predicted 9 kDa. In the literature, this is well established and some publications now even use two arrows to distinguish between AD brain derived dimer and the synthetic Aβ standards (Barry et al. 2011). As we know that in the brain there are over 40 different APP fragments that are classified as ‘Aβ’ (mostly by antibody reactivity) and that Aβ1-42 is not the only peptide present, it is plausible that this species is not a homodimer. With at least 44 monomeric species of Aβ characterized from AD and transgenic (Tg2576) mouse brains (Sergeant et al. 2003; Portelius et al. 2010) there are 1035 unique dimers that could potentially exist (Fig. 2). By calculating the mass of all of the possible dimers a distribution of molecular weights range from 6.4 up to 11.5 kDa; however, the vast majority of dimers are contained between 7.3 and 8.8 kDa. As the molecular weight of the dimeric species observed in AD brain is consistent with the range of dimers that are possible, it would suggest that if the 8 kDa band observed in AD brain is a dimer, then it would likely be a heterodimer of Aβ species.

Figure 2.

 Molecular weight distribution of possible dimers. There are at least 45 monomeric forms of ‘Aβ’ that have been identified (Masters et al. 1985b; Sergeant et al. 2003; Van Vickle et al. 2008; Portelius et al. 2010). The calculated mass of each of the 45 peptides was used to determine all possible combinations of dimers. A histogram (bin width of 50 Da) of the calculated molecular weights of the 1035 potential dimers that are possible from these peptides indicates that ∼80% of to potential dimers have a molecular weight between 7.5 and 8.9 kDa (the histogram was generated using KaleidaGraph Synergy Software).

However, the discrepancy with the model described is that it assumes that each possible dimer combination is equally possible and due to the low abundance of many of the monomeric peptides within the AD brain, the probability of most of these dimers is unlikely. Thus to derive a potentially more accurate distribution of masses we must take into account the relative proportion of each Aβ peptide previously reported (Masters et al. 1985b; Sergeant et al. 2003; Portelius et al. 2010) and assume that all Aβ peptides have equal affinity and opportunity to associate. From these calculations, four peptides compose ∼70% of the total monomeric Aβ species (Aβ1-42, Aβ4-42, pGluAβ3-42, and Aβ1-40) and as such there are 10 unique dimer combinations that are most likely to occur (Fig. 3). The mass of these dimers ranges from 8396 to 9026 Da and all have fairly equal probability of formation under the given assumptions. Thus, if a dimeric species of Aβ exists in AD brain it is likely composed of Aβx-42 peptides (Aβ1-42:Aβ4-42 as this has a calculated weight of 8.7 kDa), which is consistent with previous reports of amino-truncated containing dimers from AD brain (Sergeant et al. 2003). It is our belief that the presence of amino truncated containing dimers could potentially explain why the dimeric material in AD brain has a molecular weight slightly less than the expected or observed Aβ1-42 dimeric species, especially considering that while the Aβ1-42 dimer has a molecular weight of 9 kDa, the remaining nine combinations have a mass closer to the molecular weights observed in AD brain (∼8.6 kDa; Fig. 3). However, resolution of the identity of the 8 kDa species is still to be achieved.

Figure 3.

 Distribution of dimers based on the abundance of monomeric Aβ. The relative abundance of Aβ monomeric species was taken from (Portelius et al. 2010) and used to determine the probability of each monomer associating to form a dimer. This predicts that there are 10 dimers (inset) that are likely to reach a significant proportion of total dimeric species. For example, based on Aβ1-42 being the most abundant monomer the most likely dimer to form is an Aβ1-42/Aβ1-42 homodimer with an average molecular weight of 9026 Da (4514 + 4514 − 2 for dityrosine formation).

Despite the potential for Cu catalyzed formation of cross-linked Aβ there could be a simple alternative explanation for the production of the 8 kDa Aβ immunoreactive band found in AD brain tissue. Another plausible explanation for the appearance of SDS resistant 8 kDa band is not the formation of a cross-linked dimer but altered proteolytic processing of APP. It is well established that the proteases that cleave APP and their proteolytic cut sites within APP are not exact, e.g. the variable cleavage points of γ-secretase (Zhao et al. 2007). A large abundance of fragments are easily detectable due to a high copy number (e.g. C99, C83 AICD, Aβ1-40, Aβ1-42 and p3) while others are less well known due to their lower expression, difficulty in detection by conventional methods and masking by other fragments or oligomeric species of Aβ. There are 27 monomeric peptides (Table 1) that have a molecular weight between 8 and 9 kDa upon N-terminal and C-terminal cleavage by known proteases that contain the WO2 antibody epitope commonly use to detect Aβ (for review of cleavage sites see Naylor et al. 2008). At present, it is extremely difficult to decipher these larger monomeric APP fragments from Aβ covalently cross-linked dimers using conventional techniques and more recent techniques in resolving Aβ species, such as surface enhanced laser desorption ionization mass spectrometry (SELDI), still lack the capability to resolve the mass of the potential dimers from some of the APP cleavage fragments.

Table 1.   Linear APP peptides with MW between 8–9 kDa with defined cleavage sites
N-terminal cleavageAmino acid sequenceAverage MWC-terminal cleavage

Future perspective for metals

Although agnosticism continues for the significance of metallobiology in neurobiology, substantial evidence is emerging to suggest that Fe, Cu and Zn may be instrumental in modulating the toxicity of Aβ at physiological concentrations, particularly through its increased redox capabilities. Even so, there is still a great deal that remains to be determined. A better understanding of metal ions in neurobiology and neurodegenerative diseases such as AD, will come from the continued development of new tools such as fluorescent indicators (Xu et al. 2010; Chan et al. 2011), and techniques that directly monitor metal distribution such as X-ray absorption fine structure (Liu and Bai 2010) and various inductively coupled mass spectrometry techniques (Hagege et al. 2004; Bettmer et al. 2009) to directly measure metal–protein complexes. The implementation and development of these techniques will help to answer questions about how metal ions shift in a diseased brain and how metal ions such as Fe or Cu contribute to the oxidative stress observed in AD.

Presently, the majority of research into drug development continues to target the removal of Aβ from the brain and is based around the conventional basis of the amyloid cascade hypothesis, which has had limited success. The evidence covered by this review suggests that abnormal interactions with biological metals are upstream in the AD pathophysiology, and represent an attractive novel drug target. Changing the metabolism of metal ion homeostasis appears to significantly alter the normal course of APP processing and Aβ generation, deposition and degradation. Pre-clinical and clinical studies in AD using metal chaperones have been impressively successful (Cherny et al. 2001; Adlard et al. 2008; Lannfelt et al. 2008; Faux et al. 2010), supporting the wealth of in vitro data characterizing the interactions between metal ions and Aβ (for a detailed review see Duce and Bush 2010). Normalizing metal ion homeostasis and disrupting resistant dimeric Aβ are logical targets for intervening in the progression of AD and should continue to be developed as a therapeutic approach to the treatment of AD. However, specificity and selectivity are important factors that must be considered when testing new drug candidates. With further understanding of metals relationship with AD pathogenesis, using newer and more accurate techniques, Aβ:metal-based drug design remains a promising treatment for AD in the future.


The authors are supported by the National Health and Medical Research Council of Australia, the Australian Research Council, Operational Infrastructure Support (OIS) from the Victorian State Government and the Alzheimer’s Association.

Conflict of interest

AIB and CLM are both shareholders and consultants to Prana Biotechnology. AIB is a shareholder of Cogstate Ltd and Eucalyptus Biosciences Pty Ltd. BRR, TMR, and JAD have no competing interests to declare.