Alzheimer’s disease (AD) is the most prevalent age-related dementia affecting millions of people worldwide. Its main pathological hallmark feature is the formation of insoluble protein deposits of amyloid-β and hyperphosphorylated tau protein into extracellular plaques and intracellular neurofibrillary tangles, respectively. Many of the mechanistic details of this process remain unknown, but a well-established consequence of protein aggregation is synapse dysfunction and neuronal loss in the AD brain. Different pathways including mitochondrial dysfunction, oxidative stress, inflammation, and metal metabolism have been suggested to be implicated in this process. In particular, a body of evidence suggests that neuronal metal ions such as copper, zinc, and iron play important roles in brain function in health and disease states and altered homeostasis and distribution as a common feature across different neurodegenerative diseases and aging. In this focused review, we overview neuronal proteins that are involved in AD and whose metal binding properties may underlie important biochemical and regulatory processes occurring in the brain during the AD pathophysiological process.

1. Alzheimer’s Disease: Hallmark Amyloid Aggregation and Neuronal Dysfunction

Alzheimer’s disease (AD) is a progressive neurodegenerative disorder characterized by cognitive decline. The neuropathology hallmarks are gross atrophy of the cortex and hippocampus, and the accumulation of amyloid-beta (Aβ) into senile plaques and of hyperphosphorylated tau into neurofibrillary tangles. The deposition of Aβ and hyperphosphorylated tau aggregates in the human brain occurs in opposite directions with an orderly neuroanatomical pattern. Amyloid plaques first appear in the neocortex and slowly progress through the striatum, the basal cholinergic nuclei, the brain stem, and finally the cerebellum [1]. The deposition of tangles begins in the brain stem and progresses towards the neocortex [2]. Thus, the common presence of amyloid plaques and tau neurofibrillary tangles in the cortex only happens at late stages of the disease.

AD is heterogeneous and multifactorial with sporadic and familial forms [36]. The large majority of patients have the sporadic form or late onset dementia (later than 65 years). The few remaining patients have the familial form with early onset dementia (around 30 years to 65 years) and may present different symptoms. These patients have mutations in one of three genes encoding proteins essential for Aβ formation: the amyloid precursor protein (APP) and presenilins 1 and 2 (PSEN1/2) [710]. Presenilins are components of catalytic subunit of γ-secretase multicomplex, responsible for the cleavage of APP and formation of Aβ. The origin of the sporadic form is complex involving multiple genetic and environmental risk factors, for example, the presence of apolipoprotein E-ε4 allele, mitochondrial dysfunction, head injury, or a compromised brain blood barrier [3, 11]. Despite the fact that AD is the most common form of dementia of the elderly and affects millions of people worldwide, the exact cause of this disorder is still unknown. The genetic evidence obtained from the rare familial form of AD supports the hypothesis that the accumulation of Aβ plaques is at the origin of the disease. This is the foundation for the amyloid-β cascade hypothesis [12] which has been the central theory in AD research for the last three decades. According to this hypothesis, the deposition of Aβ is the initial event and it is sufficient to trigger the cascade of pathological and clinical changes in AD, which are the formation of senile plaques and neurofibrillary tangles and subsequent neuronal death, vascular damage, and dementia [12]. Although senile plaque deposition is an early event in the disease, as observed in postmortem human brains [1], plaque accumulation in the brain does not correlate with dementia [13] implying that other mechanisms are associated with neurodegeneration. Notably, therapies designed until now that aimed at targeting amyloid plaques and APP proved to be largely unsuccessful. An increasing amount of data challenges the amyloid-β cascade hypothesis.

Therefore, efforts to integrate the other pathogenic features of AD and multiple etiology pathways into a more global model are now needed. During the course of AD, tau is hyperphosphorylated and accumulates in the somatodendritic compartment as paired helical filaments and straight filaments [14]. In neurons, tau is the major microtubule associated protein and stabilizes its structure. Tau interacts with tubulin promoting its assembly into microtubules. The level of phosphorylation regulates the activity of tau and hyperphosphorylation suppresses its microtubule assembly activity. In addition, hyperphosphorylated tau sequesters normal tau and other microtubule associated proteins that further contribute to microtubule disassembly [15]. Therefore, the abnormal phosphorylation of tau results in loss of normal function and gain of toxic function in the AD brain. The formation of neurofibrillary tangles does correlate with cognitive decline and with neuronal and synapse loss [13, 16].

Senile plaques are extracellular deposits composed mainly of amyloid peptides ranging from 39 to 43 amino acids, which are natural metabolites of APP generated by sequential cleavage by β-secretase and γ-secretase l [17]. The APP is a transmembrane protein necessary for neurogenesis, for neurite outgrowth and guidance, and for synapse formation and repair [18]. APP is processed in different ways through different enzymes leading to the formation of amyloidogenic and nonamyloidogenic precursors. The processing of APP results in the formation of soluble α- and β-secreted APP (sAPP) which is cleaved by α- and β-secretase, respectively. As a product in the nonamyloidogenic pathway, sAPPα promotes neuronal survival and neurite outgrowth, among other beneficial neuronal functions. Contrarily, sAPPβ is not involved in the beneficial functions of sAPPα, participating in synapse pruning. Aβ is secreted through sequential APP cleavage by β- and γ-secretases, resulting in peptides that can range from 39 to 43 amino acids. The Aβ peptides are catabolized by multiple amyloid degrading enzymes, for example, neprilysin and insulin-degrading enzyme [19]. It is the imbalance between the production and clearance of Aβ that triggers its deposition as amyloid plaques. However, several studies suggest that Aβ has a physiological role in the synapses and its complete removal induces neuronal cell death [2022]. In addition to the aggregates, Aβ is also present in soluble oligomeric forms in APP-transgenic mice and human diseased brains [20]. Compared to Aβ aggregates, the soluble oligomers are highly neurotoxic [23]. Therefore, it is possible that aggregation of Aβ into plaques is a neuroprotective mechanism that eliminates the toxic oligomeric forms [15].

The normal functions of synapses are impaired during the course of AD. Synapse loss correlates with dementia suggesting that it is important for disease progression and for the degeneration process [24]. Dense plaque deposition causes the surrounding neurites to bend and change trajectory, which can lead to changes in synapse signal transmission. Also, gliosis and oxidative stress are observed in the vicinity of plaques. During normal development of the brain, microglia are involved in synaptic pruning after birth and it is possible that in the diseased AD brain the recruitment of activated microglia around the plaques participates in the synapse loss [24]. In addition to aggregates, the oligomeric forms of Aβ obtained from cultured cells or from human AD brain disturb synapses and lead to cognition impairment in injected mice [2527]. Comparably, evidence also shows that soluble forms of tau are toxic for synapses [28]. The molecular mechanisms that lead to synapse dysfunction and neuronal loss downstream of Aβ and tau are not completely identified but different pathways are implicated such as mitochondrial dysfunction, oxidative stress, inflammation, and dysregulation of metal homeostasis.

2. Metals and Metal Binding Proteins Implicated in AD

Metal ions play essential roles in the brain and there is solid evidence pointing to their homeostatic dysfunction across different neurodegenerative diseases (e.g., [2931]). This includes the first row transition metals, iron, copper, and zinc and also calcium, whose homeostasis is important for neuronal function and during aging [3234]. One major hypothesis for this cross talk, which has been put forward since a number of years and which has been elegantly reviewed in [35], proposes that AD is as much as a metallopathy as a proteinopathy. Indeed, age-related metal ion dysfunction altered levels of neuronal metal ions in AD-affected areas including accumulation in protein deposits, and the interplay between metal ions and AD pathological proteins indicates a close relationship between protein misfolding, aggregation, and metal ion homeostasis. In AD patients, it has been shown that Cu2+, Zn2+, and Fe2+ are found in the core and rims of senile plaques [36, 37] and colocalize with Aβ [38]. This has led to the suggestion that metal ion sequestration into plaques could lead to deficient distribution of these metals in the neighbouring regions [39]. Moreover, it is described that in AD patients Zn2+ is decreased in serum and blood but increased in the cerebrospinal fluid and neocortical tissue [4042]. In addition, Zn2+, Cu2+, and Fe2+ are increased in the neuropil of AD patients [36, 43]. In agreement with a role of metal ions in pathology, molecules designed to chelate Zn2+ and Cu2+ from amyloid-beta aggregates [44, 45] were found to decrease Aβ deposits in mice models due to Aβ solubilisation [45]. Here, as a contribution for a broader molecular and biochemical analysis of protein-metal cross talks in neurodegeneration, we undertake an overview of proteins with metal binding properties which are implicated in AD (Table 1).

2.1. Amyloid-β

Metal ions have been acknowledged as important players of the pathological effects of Aβ aggregation in AD and have been considered as possible modulators of Aβ misfolding and aggregation due to their binding to the Aβ peptide [4649] and to amyloid fibrils [50, 51]. Cu2+, Zn2+, and Fe2+ bind to Aβ influencing its aggregation pathway and are found in and nearby extracellular senile plaques [29, 36]. The binding of metal ions to Aβ invariably results in aggregation which may either be into amyloid fibers or into amorphous aggregates, depending on the metal ion, stoichiometry, and environmental conditions [49]. In spite of contradictory findings, there seems to be a consensus that (a) superstoichiometric levels of Cu2+ and Zn2+ result in insoluble and amorphous aggregates rather than organized fibrils [49, 5255]; (b) equimolar Zn2+ and Cu2+ induce amorphous aggregates, which slowly convert to fibrils [56, 57]; and (c) at subequimolar Cu2+ levels, the kinetics of fibril formation are accelerated [52, 58, 59] (Figure 1). The observation that high levels of Zn2+ and Cu2+ seem to shift aggregation into oligomeric precursors rather than organized fibrils has important consequences in brain function, as these Aβ precursors are now known to be the neurotoxic self-propagating species causing neurodegeneration. Furthermore, Cu2+ and Fe2+ participate in ROS production causing oxidative stress and neuronal damage, thus being one of the causes that potentiate Aβ toxicity [6062]. Indeed, the formation of H2O2 as a product of the interaction between Aβ and Cu2+ can generate hydroxyl radicals, which are related to AD pathology [63]. Superoxide has also been recently shown to be an intermediate of the reaction leading to the production of H2O2 by Cu+-Aβ and O2 [64]. Zinc and copper chelators inhibit Aβ plaque deposition in AD patients [44, 65, 66], further suggesting that amyloid pathology may arise from the dysregulation of these metal ions. Excess of iron increases Aβ production [67] and leads to the formation of annular protofibrils [68] and slows down the formation of ordered cross-β fibrils [69] towards the formation of shorter and less ordered aggregates [53, 69] which are potentially more toxic.

2.2. Tau

Tau is a disordered cytosolic protein involved in microtubule assembly and stability whose aggregation and toxic deposition are triggered by hyperphosphorylation. This results in the formation of intracellular tau paired helical filaments (PHF), which ultimately gather to form the characteristic neurofibrillary tangles (NFT) [70, 71], a process which is modulated by metal ions [30] (Figure 2). Zn2+ binds tau and promotes its hyperphosphorylation [72]; however, low concentrations of zinc induce fibril formation whereas high concentrations induce granular aggregates [73]. Fe3+ also binds to hyperphosphorylated tau and induces its aggregation [74, 75], mostly into PHF [75]; however, reduction to Fe2+ can reverse aggregation of tau [75]. Excess of iron is accumulated in NTF [76, 77] generating oxidative stress due to the Fenton reaction and perpetuating tau hyperphosphorylation [78]. The role of Cu2+ in tauopathies is controversial. Some studies suggest that tau binds Cu2+ [79], inhibiting its aggregation in vitro [80] while promoting tau hyperphosphorylation in hippocampal neurons [81]. Other studies however suggest that addition of copper-bis(thiosemicarbazone) complexes that increase intracellular copper in AD mice brains inhibits tau phosphorylation [82].

2.3. Amyloid-Beta Precursor Protein

Abnormal processing of the amyloid precursor protein leads to neurotoxic Aβ production. The proteolytic processing of APP is influenced by metal ions, by protein ligands, and by the APP oligomerization state. Cu2+ and Zn2+ promote APP expression [8385] and possibly interfere with Aβ metabolism. Cu2+ enhances APP dimerization and increases extracellular release of Aβ [86]; yet, other studies suggest that high copper concentrations modulate APP processing leading to reduced Aβ production [87]. Interestingly, APP contains a copper binding domain and a site that favours Cu+ coordination, which has led to the suggestion that it could act as a neuronal metallotransporter [87]. Recent structural and biochemical studies have uncovered a high-affinity binding site within the E2 domain that binds competitively Cu2+ and Zn2+ at physiological concentrations [88]. Metal binding results in large conformational changes and in different structural states that regulate the function of APP and Aβ metabolism [89]. Indeed, APP can be a mediator of Cu neurotoxicity since it was shown that in primary neuronal cultures APP loaded with Cu2+ induces cell death [90]. This may possibly involve catalytic reduction of Cu2+ to Cu+ leading to an increase in oxidative stress in neurons [91]. The links between APP and metal metabolism are further emphasized by the interaction of APP with ferroportin, to promote iron export and its ferroxidase activity [92, 93]. APP ferroxidase activity is inhibited by Zn2+ binding contributing to Fe2+ accumulation in AD brains [92].

2.4. Presenilin-1

Presenilin-1 (PS-1) is a component of the γ-secretase multicomplex, responsible for the cleavage of APP. Presenilins have an activity as low-conductance passive ER Ca2+ leak channels which is independent of γ-secretase activity [94]. Overexpression of presenilin results in increased Ca2+ release whose levels are restored by γ-secretase inhibitors [95]. Mutations in presenilins as in familial AD forms result in downregulation of Ca2+ channels and Ca2+-dependent mitochondrial transport proteins, strengthening the relationship between Ca2+ homeostasis and presenilin [94, 96, 97]. A recent study based on the effects of metal chelators on γ-secretase suggests that Ca2+ and Mg2+ stabilize γ-secretase and enhance its activity [98].

2.5. Metallothionein 3

Metallothioneins are a family of ubiquitous proteins with metal binding properties and antioxidant activity [99]. Neuronal metallothionein 3 (MT3), which is involved in the transport and homeostasis of Zn2+ and Cu2+, plays an important role in several AD related pathways. MT3 is decreased in AD patients [100] and in Tg2576 mice [101], which can lead to aberrant neuritic sprouting [100]. Additionally, MT3 increases sAPPα (soluble amyloid precursor protein α) levels and reduces Aβ production [102], through an increase in ADAM10 (a disintegrin and metallopeptidase 10). ADAM10 is a protein responsible for the cleavage of APP-derived peptides and activation of the nonamyloidogenic pathway [103]. Mechanistically, it has been reported that the β-domain of MT3 interacts with Aβ, abolishing Cu2+ mediated aggregation [104, 105] and ROS production [104]. It has also been suggested that rapid metal exchange between Zn2+-MT3 and Cu2+-Aβ [106] or Zn2+ release by MT3 [107] promotes structural changes in Aβ aggregates. In agreement with this, in primary neuron cultures, MT3 inhibits the formation of toxic Aβ aggregates alleviating their neurotoxic effects [105, 108]. One possible mechanism for this effect may be related to the observed metal swapping between MT3 and soluble and aggregated Aβ, which abolishes the production of Cu-induced ROS [104, 109].

2.6. Zinc Transporter 3

Zinc transporter 3 (ZnT3) is a synaptic Zn2+ transporter responsible for loading zinc into presynaptic vesicles. This protein is highly expressed in the brains of AD transgenic mice, in which it colocalizes with amyloid plaques [110112], where zinc is also found at high concentrations. Zinc sequestering within amyloid plaques has been suggested to provoke an imbalance in the cellular environment with concurrent effects on overall metal metabolism and protein homeostasis [35]. ZnT3 has been shown to decrease with aging and AD, contributing to the aggravation of zinc-mediated cognitive decline [113]. In the AD Tg2576 transgenic mouse model with a ZnT3 knockout, cerebral Aβ deposition was nearly abolished by the lack of synaptic Zn2+ [58, 59]. ZnT3 and other zinc transporters, such as ZnTs 1, 4, 5, 6, and 7, are also found upregulated in amyloid plaques of human AD brains near Zn2+ enriched terminals [60], revealing a cross talk between zinc induced amyloid plaques and zinc transporters. In ZnT3 knockout mice, the addition of metal chaperones results in restoration of expression of the synaptic proteins PSD-95, AMPAR, and NMDAR2b, due to the restitution of hippocampal zinc content [113].

2.7. ProSAP/Shank Scaffold Proteins

ProSAPs/Shanks are zinc-regulated multidomain proteins that are important scaffolding molecules of the postsynaptic density (PSD), a protein dense structure composed of both membranous and cytoplasmic proteins localized at the postsynaptic plasma membrane of excitatory synapses [114]. Deregulation of ProSAP/Shank has been reported in AD: in patients brains and in transgenic mice models, the accumulation of Aβ oligomers is accompanied by reduction of synaptic scaffold protein levels, such as Shank1 and ProSAP2/Shank3 [115], and disruption of the Homer1b and Shank1 scaffolds [116]. Interestingly, sequestration of Zn2+ by Aβ leads to less mature synapses by decreasing Shank1 protein levels at the postsynaptic density in hippocampal neurons [117]. Future studies will further elucidate the mechanistic cross-links between the presence of Aβ, zinc levels, and the scaffolding PSD proteins in the context of AD [118].

2.8. Ferritin

Ferritin is the major intracellular iron storage protein in the body. It has elevated levels in AD brain tissue [119121] and is found in the vicinity of AD plaques [120], suggesting that ferritin trapped within the plaque inclusions may block the transport of iron between cells. The loss of integrity of hippocampus tissue of AD patients is linked with the increase of ferritin [122] and with a reduction of ferroportin protein levels [123]. Effectively, the impact of iron on AD outcomes is not fully explored but a recent longitudinal study has shown that ferritin is strongly associated with cerebrospinal fluid apolipoprotein E levels; in turn, ferritin is elevated by the Alzheimer’s risk allele, APOE-ε4 [124]. This study speculates that the APOE-ε4 genotype raises the baseline iron load in the AD brain, lowering the threshold for iron-mediated neuronal loss, a hypothesis that remains to be experimentally addressed.

2.9. S100 Proteins

S100 proteins are a family of at least 21 different vertebrate-specific proteins with two Ca2+-binding EF-hand type sites and in some cases additional sites for Zn2+ and Cu2+ [125]. S100 proteins are part of the inflammatory response and a number of these proinflammatory cytokines (S100B, S100A6, S100A7, S100A1, S100A9, and S100A12) have been implicated in neurodegenerative disorders, such as AD.

S100B is a proinflammatory cytokine that triggers glial cell proliferation in a RAGE-dependent manner [141]. RAGE is an immunoglobulin-like cell surface receptor that is upregulated in AD and triggers the expression of proinflammatory cytokines and mediates Aβ transport across the blood-brain barrier [142144]. At high micromolar concentrations, S100B promotes neuroinflammatory processes and neuronal apoptosis [145]. Increased expression of S100B by plaque-associated astrocytes in AD contributes to the appearance of dystrophic neurites overexpressing βAPP in diffuse amyloid deposits [132]. Astrocytic overexpression of S100B is correlated with the degree of neurite pathology in Aβ aggregates and is induced by interleukin-1 (IL-1), which is secreted by activated microglia present in the plaques [146]. TNFα, a cytokine with high levels in AD, decreases S100B expression in astrocytes but increases its extracellular levels which can lead to RAGE activation [147]. Furthermore, studies demonstrated increased susceptibility to neuroinflammation and neuronal dysfunction after infusion of Aβ in transgenic mice overexpressing S100B [148]. Interestingly, S100B interacts with tau in a Zn2+ dependent fashion that could be responsible for neurite outgrowth [133]. Other studies, however, suggest that the S100B:tau interaction is mediated by Ca2+/calmodulin-dependent kinase II and results in the inhibition of tau phosphorylation [134].

S100A6, S100A9, and S100A12 also have consistently high levels in samples of AD patients [135, 149]. In particular, S100A9 is found near neuritic plaques [136, 137] and was found to coaggregate with Aβ in vitro and form toxic aggregates [136, 138]. Knockout of S100A9 in a transgenic mouse resulted in reduced Aβ levels in the brain and the animals presented an improved spatial reference memory [139]. In agreement with these observations, knockdown of S100A9 in the AD Tg2576 mice model reduced Aβ and APP C-terminal levels and decreased BACE activity [137]. Induction of S100A9 levels increased intracellular Ca2+ levels, which in turn upregulated secretion of the inflammatory cytokines IL-1β and TNFα [150]. On the opposite, expression of exogenous S100A7 in primary corticohippocampal neuron cultures derived from Tg2576 transgenic embryos inhibits the generation of Aβ and promotes the activity of α-secretase [140]. Interestingly, S100 proteins have been found to have amyloidogenic properties [151155]. This feature, along with the high abundance of S100 proteins in protein deposits, their metal binding properties, dysregulation of Ca2+ signalling, and the high levels of Cu2+ and Zn2+ in the plaques, will certainly translate into the elucidation of new functions of S100 proteins in AD pathomechanisms.

3. Conclusion

Metal homeostasis and balance depend on a number of biochemical processes and proteins, many of which operate in the neuronal environment and in the extracellular synaptic space or at its interface. The biochemistry of this particular cellular moiety is deeply altered upon aging and under neurodegeneration, with wide changes in protein levels, signalling molecules, and metal ion concentrations. Changes in protein and metal ion homeostasis are hallmark features across amyloid-forming neurodegenerative diseases and as we have here overviewed, a number of proteins implicated in AD are directly regulated by metal-protein interactions; in some cases, metal ions are even directly involved as modulators of aggregation pathways. Uncovering the mechanistic details of this cross talk at the biochemical levels in respect to effects on synaptic protein networks, Aβ metabolism and intra- and extracellular protein aggregation in the context of concurrent affected processes such as oxidative stress and neuroinflammation are thus among the major challenges in modern molecular neurosciences.

Conflict of Interests

The authors declare that there is no conflict of interests regarding the publication of this paper.


This work was supported by the Bial Foundation through grant PT/FB/BL-2014-343 (to CMG), by the Fundação para a Ciência e a Tecnologia through grant PTDC/QUI-BIQ/117789/2010 (to CMG), PhD fellowship SFRH/BD/101171/2014 (to JSC), and grant UID/MULTI/04046/2013 from FCT/MCTES/PIDDAC (to BioISI). CMG is a recipient of a Consolidation Level Investigator FCT, also from the Fundação para a Ciência e a Tecnologia (IF/01046/2014).