Alzheimer’s disease (AD) is the primary cause of dementia in the elderly and currently affects more than five million Americans (Alzheimer’s Association, 2013). The present research focuses on how the biometals, such as zinc (Zn), copper (Cu), and iron (Fe), interact with amyloid beta (Aβ)—the key constituent of the plaques that are characteristic of the disease, which mediates plaque formation and consequent oxidative damage.

Zn, Cu, and Fe are found in high concentrations in and around amyloid plaques (Lovell et al., 1998; Maynard et al., 2005), consistent with the notion that Aβ is a metalloprotein that possesses binding sites for both Zn and Cu (Bush et al., 1993; Hesse et al., 1994). Zn has been shown to be particularly effective in promoting Aβ aggregation (Bush et al., 1993), and the sequestration of Zn by Aβ leads to an intracellular deficiency of this metal (Grabrucker et al., 2011). Cu, on the other hand, causes a shift in the processing of the amyloid precursor protein (APP) towards the non-amyloidogenic pathway (Borchard et al., 1999) but once Aβ is present, it binds both Cu and Fe. Therefore, Cu and Fe binding is problematic as these metals are redox-active and Aβ reduces both metals, resulting in the production of hydrogen peroxide (H₂O₂; Huang et al., 1999). H₂O₂ further reacts with the reduced metal to produce hydroxyl radicals that can cause oxidative damage (Bush, 2003). Both aging and plaque-related inflammation may promote a more acidic brain environment, which is conducive to Fe and/or Cu binding (Atwood et al., 1998).

Evidence suggests that Zn is crucial for Aβ deposition, and the presence of Zn in vitro prevents the proteolytic degradation of Aβ_1–42 by matrix metalloprotease 2 (Crouch et al., 2011). Zn also inhibits the activity of alpha-secretase and promotes gamma- and beta-secretase activity (Capasso et al., 2005). The zinc transporter, ZnT3, is responsible for loading zinc into the synaptic vesicles (Palmiter et al., 1996; Cole et al., 1999; Linkous et al., 2008). Eliminating synaptic Zn by genetic ablation of zinc transporter 3 (ZnT3) in an AD mouse model resulted in a 50% reduction in amyloid (Lee et al., 2002). ZnT3 levels, however, have been shown to decrease as a function of age and AD (Adlard et al., 2010). Interestingly, Bjorklund et al. (2012) found that in brains exhibiting AD neuropathology, ZnT3 levels were lowest in those patients who received a diagnosis of AD, while those who did not receive a diagnosis had ZnT3 levels similar to those of healthy controls.

Given the complexity of biometals in AD, the behavioral outcome due to changes in metal levels is difficult to predict, but this has been studied in various mouse models. Long-term dietary Zn enhancement in transgenic (Tg) mouse models of AD (TgCRND8, Tg2576, and APP/presenilin 1 (PS1) mouse models) has been shown to result in cognitive deficits (Linkous et al., 2009; Railey et al., 2011), as well as an increase in both brain Zn levels and the number and size of Zn-positive plaques in the cortex and hippocampus (Wang et al., 2010). Other findings have found that a severe Zn deficiency (<10 ppm Zn) enlarged amyloid plaques (Stoltenberg et al., 2007); thus either an overabundance or deficiency in Zn can potentiate amyloid deposition.

The observations derived from these early-onset mouse models may be of diminished utility for the majority of AD cases that are classified as late-onset (manifest after age 65 years) (Harvey et al., 2003). The emphasis on early-onset genetics dominating AD research is due to the strong relationship between those genes and development of the disease. Genetic linkages to the late-onset form of AD are less specific, but the ε4 allele of the apolipoprotein E (ApoE) gene has been identified as a major risk factor (Saunders et al., 1993; Blacker et al., 1998; Huang and Mucke, 2012). The ApoE gene has three alleles (ε2, ε3, and ε4) resulting in three isoforms of the protein (Zannis et al., 1981), with ApoE ε4 overrepresented in patients with late-onset AD (Weisgraber and Mahley, 1996). While ApoE ε4 is considered a susceptibility factor it does not guarantee the development of disease (Weisgraber and Mahley, 1996). The exact mode of action of ApoE ε4 in AD remains elusive. One proposed mechanism is the interaction between ApoE ε4 and Aβ, which may increase plaques and/or impede Aβ clearance (Mahley et al., 2006). Amongst those with AD, ApoE ε4 carriers typically have a larger number of plaques than ApoE ε3 carriers (Schmechel et al., 1993). Similarly, ApoE ε4 mice that overexpress human APP have an elevation in Aβ plaque deposition, reduced numbers of presynaptic terminals, and an impairment in learning and memory tasks as compared to their ε3 counterparts (Huang, 2006).

The ApoE alleles differ at two residues: ε2 (Cys¹¹², Cys¹⁵⁸), ε3 (Cys¹¹², Arg¹⁵⁸), and ε4 (Arg¹¹², Arg¹⁵⁸) (Rebeck et al., 2002). Cysteine binds to Zn and Cu with high affinity and the Zn binding site on APP is located in a cysteine-rich area (Bush and Tanzi, 2002). The increased number of cysteine residues in ε2 and ε3 correlates with their ability to bind zinc in the synaptic cleft and may contribute to the protective effect of these alleles by preventing zinc binding to Aβ (Lee et al., 2010). ApoE ε4 is also implicated in the formation of neurofibrillary tangles (NFTs). It may increase the phosphorylation of the tau protein (Harris et al., 2004) and Zn may be a key modulator of this process (Craddock et al., 2012). Through their relatively high-binding affinity for zinc, Aβ plaques will cause a reduction in intracellular levels of zinc that will in turn lead to a destabilization of microtubules and subsequent liberation of tau (Craddock et al., 2012). These findings suggest that Zn and ApoE ε4 may act together to affect multiple processes in AD through the promotion of NFT formation and increased Aβ deposition, which may be due to increased Zn binding in the presence of ApoE ε4.

In order to examine the effect of excess zinc in the more prevalent form of AD we have developed a late-onset model, examining two types of Tg mice with the (1) humanized ApoE ε4 and mutated hAPP cross (CRND8/E4); or (2) those only with mutated hAPP (CRND8), and investigated the effect of Zn supplementation on spatial memory in these mice. Protein levels of soluble and insoluble Aβ, and ZnT3 were measured to elucidate molecular mechanisms that may be associated with excess dietary zinc in AD.

These data show that the performance of the mice modeling late-onset AD depended significantly upon the presence or absence of Zn in the drinking water. Overall, CRND8/E4 mice did not perform worse than the CRND8 mice, as had initially been predicted. However, CRND8/E4 mice raised on Zn water made the most errors in the BM of all six groups, and also had significantly increased errors compared to those raised on lab water. On the primary latency measure, the CRND8 mice performed significantly worse than the WT mice. The CRND8/E4 group’s performance was intermediate between the other groups, and did not differ significantly from either. In contrast to the effect on primary errors, there was no significant effect of Zn water on primary latency. The observation that primary errors, and not primary latencies, detected a Zn effect in the late-onset mouse model is consistent with O’Leary and Brown (2013) who have suggested that errors are a more sensitive measure of memory than latency in the Barnes maze.

The effect on errors may be due to enhanced Zn binding to Aβ when in the presence of ApoE ε4. ApoE ε4 has the weakest affinity for binding Zn, due to the amino acid substitution of arginine which is not a ligand for Zn (Lee et al., 2010). In contrast, the ApoE ε2 and ApoE ε3 isoforms with cysteine residues strongly bind Zn, and may prevent it from binding with Aβ, which could in turn prevent oligomerization and deposition (Lee et al., 2010). While there are numerous in vitro studies investigating the association between Zn and ApoE ε4, this is a novel behavioral study examining the relationship between the two in a mouse model of late-onset AD.

Zn water did not result in significant latency deficits in the CRND8 and WT groups. For the WT group, Zn significantly reduced the percent time spent in the target quadrant for the 24-h probe trial; thus the WT mice on Zn water were impaired on the most difficult memory task. Neither the CRND8/E4 or CRND8 groups performed above chance on either water condition (Figure 4); thus, additional impairments caused by Zn were difficult to assess. Given the difficulty of the 24-h probe task for the Tg mice, running probe tests as the last run on alternate days, as is done in the Morris water maze (MWM), could be a useful strategy.

We have previously demonstrated significant latency impairments in the MWM in both Zn water rats and Tg (CRND8 and Tg2576) mice (Linkous et al., 2009; Railey et al., 2010, 2011). The MWM is a more stressful test than the BM (Hölscher, 1999) and this may account for the different effects of zinc on latency seen in this experiment. Overall, data from our lab indicate that Zn water has adverse behavioral effects on WT and CRND8 mice but that additional mechanisms are at work for CRND8/E4 mice. Although Zn supplementation exacerbated behavioral deficits in the Tg2576 and CRND8 mice, this led to a paradoxical decrease in amyloid deposits in the Tg2576 mice (Linkous et al., 2009), which was also observed by Harris et al. (2014); however they found no significant behavioral differences.

The deficits in spatial memory that were observed in the Barnes maze, in connection with dietary Zn supplementation, may be attributable to the effect of Zn on other metals in the brain, specifically Cu. Excess dietary Zn prevents adequate absorption of Cu through the intestinal wall (Milne et al., 2001). Therefore, Zn supplementation may potentiate a Cu deficiency that is inherent in Tg mice overexpressing mutant hAPP (Phinney et al., 2003). Brewer (2014), however, found a non-significant delay in cognitive impairment in AD patients who received Zn supplements, and suggests that AD is the result of a Zn deficiency. He also noted that these subjects had relatively low Zn serum levels and suggested that this would lead to an overload of copper. Brewer (2014) considers zinc deficiency a risk factor for AD, due to its interaction with copper (Brewer, 2014).

In contrast, Klevay (2008) has suggested that AD is due to a Cu deficiency instead of Cu toxicity and spatial memory deficits in Zn-treated Tg mice can be remediated by the concurrent administration of small amounts of Cu (Railey et al., 2011). APP-overexpressing mice have lower levels of Cu in the brain (Bayer et al., 2003) which would support increasing dietary Cu to treat AD. This is consistent with another study where APP23 mice were given Cu supplementation which resulted in a reduction of Aβ and an increase in superoxide dismutase 1 (SOD1) activity (Bayer et al., 2003). These conflicting findings from different labs suggest that metal dyshomeostasis in AD can be due to either a deficiency or overabundance of metal ions, and that alterations to the levels of one metal can impact levels of another. The initial metal status of the patient should be an important factor in considering any form of dietary metal-based therapy.

In contrast to our observations, Zn supplementation in the 3xTg mouse actually improved performance on behavioral tasks, and decreased both Aβ and tau pathology (Corona et al., 2010), while decreased dietary Zn elevated plaque volume in APP/PS1 mice (Stoltenberg et al., 2007). Carrying mutations in PS1, or in the case of the 3xTg, a mutation in tau, can alter the ability of these proteins to interact with metals. The divergent results may stem from differences in the mouse model used, chemical composition of zinc solutions, and age of onset and duration of Zn supplementation. Our mice were given a standard rodent chow which contains adequate levels of Zn, in addition to the Zn supplementation, which would be in contrast to the Zn deficiency that is reported in the elderly population. While Zn therapy may be beneficial for those with a Zn deficiency, excess Zn, and Zn in the presence of ApoE ε4, may actually be harmful.

In a mouse model carrying the three different human ApoE alleles in the 5xTg mouse, the ApoE transgenes delayed pathology typically observed in the 5xTg strain (Youmans et al., 2012). Although having an expanded window of time to study the progression of AD pathology could be useful, using either the 3xTg or 5xTg strains may be contraindicated for studying the effects of Zn or other metals in late-onset AD. These multiple transgene models have mutations in PS1, which affects cellular turnover of Zn and Cu, since PS1 has Zn binding sites, which will affect SOD1 activity (Sensi et al., 2008; Greenough et al., 2011; Southon et al., 2013). Therefore, studies looking at the relationship between metals and pathobiology of late-onset AD should avoid incorporating mutated PS1 transgenes. Also, enzymes that modulate tau are Zn-dependent.

The current data did not support the predicted outcome of more behavioral impairment in CRND8/E4 mice as compared to CRND8 mice. On most measures the CRND8/E4 mice performed slightly, but not significantly, better than CRND8 mice. This is consistent with other recent behavioral data in younger Tg mice modeling late-onset AD (Moreau et al., 2013). Similar observations of intermediacy were reported in circadian rhythm and nest building behavior, as well as in markers for inflammatory cytokines (Graybeal et al., 2014). Recent studies have shown that any isoform of human ApoE delays AD pathology in early-onset Tg mice (Youmans et al., 2012; Tai et al., 2013).

One possible explanation for the greater impairment associated with the ApoE ε4 allele in old age is due to its interaction with metals and their dysregulation with increased age. Zn and Cu have an interdependent relationship and they exist in a delicate homeostasis whereby a shift towards either excess or deficiency may influence Aβ neurotoxicity. Lee et al. (2012) have shown that Zn accumulates around dystrophic neurons in aged Tg2576 mice and suggest that this excess Zn “crowded around amyloid plaques” and could directly cause the death of neurons. Such an effect would likely be greater in the case of ApoE ε4 which does not bind Zn and Cu as efficiently as the other forms of ApoE.

Another interesting finding was that ZnT3, which is the transporter responsible for loading zinc into synaptic vesicles, was negatively correlated with soluble Aβ levels, which is the more toxic form of Aβ. Although a general decline in levels of ZnT3 has been associated with cognitive decline (Bjorklund et al., 2012), it has also been demonstrated that the regions of brain that express ZnT3 may change so that the overall level of expression remains unaltered, but the localization of expression may vary and be centered more around areas of pathology in the Tg CNS (Lee et al., 2012). This may explain why no overall differences for ZnT3 expression were detected. It is notable that within the Zn water group we observed a correlation between insoluble Aβ and primary errors, as this may be due to the ability of Zn to aggregate Aβ. The relationship between amyloid load and learning and memory remains controversial; for example, Gruart et al. (2008) found no relationship between hippocampal-dependent learning and amyloid load, suggesting that other mechanisms may be involved.

ApoE ablation reduces the amount of ZnT3 and results in significantly lower levels of synaptic Zn compared to WTs (Lee et al., 2010). Zn is required for both presynaptic and postsynaptic mossy fiber long-term potentiation, and dysregulation of synaptic Zn alters the excitatory input of the mossy fiber-CA3 synapse in the hippocampus (Pan et al., 2011). Furthermore, ApoE-ε4 targeted replacement mice have altered levels of glutaminase and vesicular glutamate transporter 1 (Dumanis et al., 2013). ApoE-ε4 is correlated with decreased dendritic spines (Dumanis et al., 2009), and treatment of neurons with ApoE-ε4 recombinant protein reduces expression of glutamate receptors (Chen et al., 2010). Future research should examine the effects of Aβ, Zn supplementation, and ApoE status on glutamatergic pathways.

Since this study demonstrates that CRND8/E4 mice are sensitive to dietary Zn-enhancement, this suggests that therapies targeting metals should consider ApoE allele status in their analyses. Currently, the only FDA-approved drugs for treating AD work by alleviating symptoms instead of targeting the underlying etiology of the disease. A newer class of drugs currently under evaluation, the 8-hydroxyquinolines (8-OHQ), are selective Zn/Cu chaperones (Greenough et al., 2013). Two drugs in this class, Clioquinol (CQ) and PBT2 have a higher affinity for binding Zn and Cu when compared to Aβ. Instead of chelating these metals from the system, they are able to return metals back inside the cell, thereby restoring the bioavailability of metal ions necessary for cellular processes (Adlard and Bush, 2012). This restoration of intracellular metal concentrations has been shown to increase synaptic proteins, reduce Aβ accumulation, and improve cognitive functioning in Tg mice (Adlard et al., 2008, 2011; Crouch et al., 2011).

This experiment showed that Zn water caused significant impairments in a mouse model of late-onset AD, containing the ApoE ε4 allele; consequently, studies on the role of metals in AD need to take into account the ApoE status of either the human subjects or the mouse models. Furthermore, due to the interaction between different metals such as Cu and Zn, care needs to be taken when conducting both in vitro and in vivo studies, where the outcomes and interpretations of studies may be confounded by this inter-relationship. Future studies examining the effects of Zn and/or Cu deficiencies and excesses in the different human ApoE alleles of a late-onset model of AD is warranted.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.