Randomised-controlled trials

In the placebo-controlled, randomised trials, there is no clear evidence that supplementation or depletion of Cu or Fe brings beneficial effects on the cognitive performance of patients with AD. All supplementation trials administered Cu as an add-on to normal medication or in combination with other nutrients, which would hamper drawing of any conclusion even if there was a beneficial effect. The depletion studies were small-sampled (n 20–78) and ran over short periods (3–24 months). The trial by Squitti et al. ⁽Reference Squitti, Rossini and Cassetta⁴⁸⁾ was conducted over 24 weeks; this was not sufficient time to detect a cognitive decline in the placebo group. Implicitly, no final conclusion is possible on the clinical benefit of d-penicillamine without a longer study period. This also applies to the PBT-2 trial over 12 weeks in which MMSE and ADAS-cog scores diverged for the placebo and intervention groups, but effects remained insignificant⁽Reference Lannfelt, Blennow and Zetterberg⁴⁶⁾. The early (1993) DFO trial included AD patients diagnosed by ADAS-cog and ran over 2 years. Unfortunately, the dropout rate for performing different cognitive tests (e.g. Wechsler Memory Scale) was too high to analyse anything but the video-recorded behaviour over time, which could be biased by the scoring procedure and behaviour changes induced by the method itself. Furthermore, it is worth noting that DFO is a charged molecule that is rapidly metabolised and does not easily pass the blood–brain barrier, which may limit its applicability; it also is unclear whether the beneficial effects of DFO relied on the chelation of Fe, Al or other metals. In summary, the current evidence arising from randomised controlled trials on the association between Cu or Fe and AD is inconclusive and awaits longer studies with larger samples.

Prospective studies

The paucity of clinical data regarding nutritional trials is countered by a longitudinal study by Morris et al. ⁽Reference Morris, Evans and Tangney³²⁾. The study had a large sample (n 3718), a median follow-up of 5·5 years and deduced cognitive scores from four different tests, but may be limited since the mixed effect models were not adjusted for the apoE genotype that influences both lipid⁽Reference Fotuhi, Mohassel and Yaffe¹⁴⁶⁾ and Cu⁽Reference Zappasodi, Salustri and Babiloni⁷⁷⁾ metabolism in AD. In addition, dietary Cu intake was calculated by multiplying the daily intake of food with the Cu content of foods, which depends on the soil and can vary regionally. However, it is unlikely that this analytical step heavily biased the results, as the findings of the study were strongest for supplementary Cu intake. While no association was found for cognitive decline and Cu and dietary cholesterol⁽Reference Morris, Evans and Tangney³²⁾, a trial with the cholesterol-lowering drug atorvastatin found that AD patients treated with atorvastatin (n 32) over 1 year showed significant improvements in cognitive performance compared with placebo, and that the treatment was also associated with a reduction in circulating Cu levels in blood⁽Reference Sparks, Petanceska and Sabbagh¹⁴⁷⁾. The second prospective study⁽Reference Squitti, Bressi and Pasqualetti⁵⁰⁾ was small-sampled (n 81), with a short follow-up and explored the predictability of cognitive decline in relationship with Cu levels. It provided no information on the size of the effect that simultaneous hyperlipidaemia and high Cu levels have on cognitive impairment or whether apoE influences the association. In summary, the prospective studies suggest that a diet that is rich in both Cu and saturated fats fosters cognitive decline in elderly subjects.

Cross-sectional studies

The findings by Ortega et al. ⁽Reference Ortega, Requejo and Andrés⁵²⁾ are based on comparing groups with MMSE scores ≥ 28 and < 28. The association of Fe intake and MMSE scores was not adjusted for age, which lowered the scores. Also, the conventional cut-off point for dementia is 24⁽Reference Petersen, Smith and Waring¹⁴⁸⁾, so it is difficult to deduce what the findings mean for the association of dietary Fe with AD. In contrast, another study found systemic Fe and Cu levels to be associated with scores of certain cognitive tests but measured neither MMSE scores nor the prevalence of AD⁽Reference Lam, Kritz-Silverstein and Barrett-Connor⁵³⁾. The third cross-sectional study⁽Reference Milward, Bruce and Knuiman⁵¹⁾ did not find an association between dementia and serum ferritin levels. However, serum ferritin may not be a reliable proxy for brain Fe⁽Reference House, St Pierre and Milward¹⁴⁹⁾. In summary, the current evidence of cross-sectional studies between an association of AD and Cu or Fe levels is inconclusive.

Autopsy studies

In general, the heterogeneous outcomes of the autopsy studies might have been influenced by several factors. First, there may be differences in the underlying tissue samples according to non-matched differences in the patients including age, sex and genetic or environmental factors such as exposure to toxic elements. Second, the small sample sizes of the studies (nine to twenty-one AD patients; ten to seventeen controls) come with increased variability. Third, it is possible that fixation of brain tissue samples with formalin⁽Reference Lovell, Robertson and Teesdale^29, Reference Samudralwar, Diprete and Ni^109, Reference Deibel, Ehmann and Markesbery¹⁵⁰⁾ influences the results⁽Reference Schrag, Dickson and Jiffry¹⁵¹⁾. Fourth, the trace elements were measured by different methods such as inductively coupled MS⁽Reference Corrigan, Reynolds and Ward¹¹⁴⁾, instrumental neutron activation analyses⁽Reference Deibel, Ehmann and Markesbery^99, Reference Andrási, Farkas and Scheibler¹¹⁶⁾ or radiochemical neutron activation analysis⁽Reference Tandon, Ni and Ding¹¹⁸⁾. Fifth, there are dissimilarities in the preparation of the post-mortem tissues. For instance, some studies used short-term post-mortem interval tissue specimens and samples from AD cases confirmed by histology⁽Reference Andrási, Farkas and Scheibler^116, Reference Deibel, Ehmann and Markesbery¹⁵⁰⁾, while a few did not⁽Reference Corrigan, Reynolds and Ward¹¹⁴⁾, some measured Cu as μg Cu/g dry brain weight⁽Reference Deibel, Ehmann and Markesbery^99, Reference Plantin, Lying-Tunell and Kristensson¹¹⁷⁾, others referred to wet weight⁽Reference Tandon, Ni and Ding^118, Reference Religa, Strozyk and Cherny¹³⁶⁾. Any conclusion reached should take these methodological heterogeneities into account. It may nonetheless be summarised that the autopsy studies almost consistently revealed that Fe is increased and that homeostatic Fe regulation is disrupted in the brains of patients with AD. The high levels of Cu, Fe and Zn in the amyloid plaques indicate the central roles of the metals in the formation of the central histological features of AD (for details, see Bush⁽Reference Bush¹⁶⁾). The situation with Cu levels in brains of AD patients is less clear. Most studies found no significant differences in Cu concentrations between AD patients and controls. Some of those that suggest otherwise compared AD patients with young controls⁽Reference Loeffler, LeWitt and Juneau¹¹⁹⁾ and used formalin fixation and no short-term post-mortem interval samples (>48 h)⁽Reference Andrási, Farkas and Scheibler¹¹⁶⁾. However, definitive answers on whether Cu is unchanged in the brain during AD remain a matter of scientific debate.

Case–control studies

The presented case–control studies provide four lines of evidence. First, they confirm the findings of autopsy studies that Fe is increased in the brains of AD patients in comparison with controls by applying phase imaging or MRI. Second, seven case–control studies compared Fe, TF and ferritin levels in plasma, serum and CSF of patients with AD with controls. The studies' heterogeneity may arise from differences in the sample collection and preparation, in the methods used for Fe measurement, in the diagnosis of AD, in the characteristics of the study populations and in the limited number of studies. These all hinder drawing general conclusions on systemic Fe levels in AD. Third, genes associated with Fe metabolism (HFE, TF C2) were analysed for an association with AD. For HFE, a meta-analysis negates such an association; for TF C2, the results are inconsistent and necessitate additional scrutiny. Fourth, case–control studies analysed alterations in Cu levels in plasma, serum and CSF of AD patients in comparison with controls. A current meta-analysis reported increased Cu levels in the serum of AD patients and unchanged CSF levels. The apoE ɛ4 genotype was associated with higher Cu levels than in non-carriers and altered brain activities, suggesting that the increased risk of apoE ɛ4 carriers for AD may partly be based on a role in the effects induced by the dyshomeostasis of Cu.

Further studies

Apart from general limitations that are associated with uncontrolled studies and case reports, the interesting study by Pajonk et al. ⁽Reference Pajonk, Kessler and Supprian¹³⁰⁾ may be further limited by its short follow-up of 8 weeks, and the fact that the analysis excluded all patients in the highest tertile of plasma Cu levels. The results of further studies⁽Reference Kessler, Pajonk and Meisser^129, Reference Salustri, Barbati and Ghidoni¹³¹⁾ highlight the current scientific debate on whether it is more appropriate to address total Cu or free Cu levels in the study of mental decline⁽Reference Salustri, Barbati and Ghidoni¹³¹⁾.

Summary

In conclusion, the current trials provide no conclusive evidence that depletion or supplementation of Cu or Fe is beneficial for AD; prospective studies found that a diet concurrently high in Cu and saturated fats may foster cognitive decline in age; Fe has been consistently found at elevated levels in the brains of AD sufferers by both autopsy and case–control studies. The specific outcomes for Cu are more conflicting; while evidence suggests that the systemic Cu level is increased in patients with AD, further research is needed to define the alterations of Cu in the brain during AD. Also, the relevance of certain genes including TF C2 or apoE awaits further investigation.

Molecular basis

Is the prospective studies' finding biologically plausible? The molecular effects of dietary Cu found in cells and animals are complex and understood only in part. They appear to depend on whether only Cu or Cu and further nutritional factors are experimentally altered.

In the case of a single elevation in Cu levels, the amyloidogenic pathway is inhibited⁽Reference Bayer, Schäfer and Simons^152, Reference Borchardt, Camakaris and Cappai¹⁵³⁾. One possible explanation for this is a Cu-induced change in APP processing. APP is an integral membrane protein with two Cu binding domains⁽Reference Multhaup, Schlicksupp and Hesse¹⁵⁴⁾, which is cut by endonucleases⁽Reference Mattson²⁷⁾ into lipid rafts, membrane microdomains enriched in cholesterol and sphingolipids⁽Reference Hung, Robb and Volitakis^155, Reference Simons and Ikonen¹⁵⁶⁾. When Cu binds to APP, a conformational change is induced that is thought to affect the clustering of APP in the cell membrane, which, in turn, alters the rate at which APP is processed⁽Reference Kong, Miles and Crespi¹⁵⁷⁾. In consequence, the concentration of Aβ decreases, which would be in line with the findings of some clinical studies⁽Reference Kessler, Pajonk and Bach^38, Reference Kessler, Pajonk and Meisser^129, Reference Pajonk, Kessler and Supprian¹³⁰⁾, but in contradiction to others⁽Reference Squitti, Lupoi and Pasqualetti^74, Reference Squitti, Barbati and Rossi⁷⁶⁾. Moreover, the processing of the APP depends on the flotillin-2-related endocytosis of APP⁽Reference Ehehalt, Keller and Haass^158, Reference Schneider, Rajendran and Honsho¹⁵⁹⁾, which depends on cholesterol⁽Reference Schneider, Rajendran and Honsho¹⁵⁹⁾ and Cu⁽Reference Hung, Robb and Volitakis¹⁵⁵⁾. Increased Cu levels attenuate Aβ synthesis via the inhibition of APP endocytosis⁽Reference Hung, Robb and Volitakis¹⁵⁵⁾.

On the other hand, Cu overload has been reported to result in the overexpression of APP⁽Reference Armendariz, Gonzalez and Loguinov¹⁶⁰⁾, lipid peroxidation, generation of the reactive aldehyde 4-hydrox-2-nonenal and oxidative stress⁽Reference Hayashi, Shishido and Nakayama¹⁶¹⁾, thus aggravating the vicious circle of AD pathogenesis. A comprehensive understanding of how Cu is involved in AD awaits further research. However, if – in addition to increased Cu intake – further nutritional parameters such as intake of cholesterol or SFA are experimentally altered, the molecular effects appear to be distinct. Hence, the addition of trace amounts (0·12 parts per million) of Cu in distilled water to cholesterol-fed rabbits can induce a reduction in cognitive abilities and increased levels of amyloid plaques⁽Reference Sparks and Schreurs¹⁶²⁾.

In vitro experiments have shown that Aβ can form cation channels in the lipid bilayer membrane⁽Reference Lin, Bhatia and Lal¹⁶³⁾. Enrichment of cholesterol reduces membrane fluidity and results in the destabilisation of the Aβ channels and an exclusion of Aβ in a metal- and pH-dependent manner⁽Reference Curtain, Ali and Smith¹⁶⁴⁾. In the light of the aforementioned pathogenic effects of Cu overload, additional presence of Cu may further foster neurotoxicity via soluble Aβ and lipid oxidation. Furthermore, hypercholesterolaemia is thought to simultaneously heighten the brain levels of Aβ and Fe⁽Reference Ghribi, Golovko and Larsen¹⁶⁵⁾, which exacerbates oxidative stress⁽Reference Zecca, Youdim and Riederer³¹⁾. Hence, the membrane integrity and stability of Aβ in the membrane are delicately balanced, and even small changes in the concentrations of metals, cholesterol or saturated fats may be able to influence the pathogenesis of AD. This conclusion might be limited since brain Fe and Cu homeostases depend on a multitude of molecular players and processes (see for reviews, Hung et al. ⁽Reference Hung, Bush and Cherny³⁰⁾; Zecca et al. ⁽Reference Zecca, Youdim and Riederer³¹⁾) that have not all been taken into account in the present review. With regard to clinical evidence, the cholesterol-promoting effect of a high dietary intake ratio of SFA:PUFA should be kept in mind⁽Reference Keys, Anderson and Grande^166–Reference Mensink and Katan¹⁶⁸⁾. Fig. 1 provides an integration of the effects of Cu and cholesterol into the pathology of AD.

Fig. 1 Effects of cholesterol and Cu on the amyloid β pathology. Considering, in step (1), that cholesterol and Cu levels rise, cholesterol will be integrated into the membrane whose fluidity, in turn, is reduced. The membrane-bound Aβ protein dissociates (2)⁽Reference Curtain, Ali and Smith¹⁶⁴⁾ to form extracellular amyloid plaques at which Fe²⁺ and Cu⁺ generate H₂O₂, which results in lipid peroxidation and the subsequent generation of 4-hydroxynonenal (4HNE), a neurotoxic aldehyde. In the cell, the free Aβ exhibits diverse pathogenic mechanisms (3) including mitochondrial oxidative stress, decreased production of ATP, production of H₂O₂ in the mitochondria, the Fe- and Cu-catalysed generation of the hydroxyl radical, that induces oxidative stress in the endoplasmic reticulum⁽Reference Mattson²⁷⁾. Finally (4), the cholesterol-enriched diet can lead to apoptosis, DNA damage, blood–brain barrier disruption, as well as dysregulation at the level of Fe regulatory proteins⁽Reference Ghribi, Golovko and Larsen¹⁶⁵⁾.

However, we are missing a lot of information that could help us see the complete picture. For instance, the complex homeostases of brain Fe and Cu and the relationship of systemic to brain levels are only understood incompletely. Also, the relevance of apoE in the interplay of Fe, Cu and lipid metabolism during AD is unclear. Supplementary experiments in mural AD models at different life stages could give answers to questions of whether the temporal pattern of dietary combinations are relevant to the development of AD and could lay the ground for additional clinical trials.

Practical implications

In a recent prospective study with elderly people, Gu et al. ⁽Reference Gu, Nieves and Stern¹⁶⁹⁾ identified a dietary pattern that was strongly associated with a lower risk of developing AD. It included higher intakes of salad dressing, nuts, fish, tomatoes, poultry, cruciferous vegetables, fruits, and dark and green leafy vegetables and a lower intake of high-fat dairy products, red meat and butter⁽Reference Gu, Nieves and Stern¹⁶⁹⁾. The identified dietary pattern contains few SFA and is likely to be associated with a low intake of Fe⁽Reference Dwyer, Zacharski and Balestra^170, Reference Buijsse, Feskens and Moschandreas¹⁷¹⁾. The same could be true for a Mediterranean diet (high intake of vegetables, legumes, fruits, fish, nuts and cereals, but a low intake of saturated lipids, meat and poultry, and a moderate intake of ethanol⁽Reference Willett, Sacks and Trichopoulou¹⁷²⁾), which has been reported to be associated with a reduced risk of AD⁽Reference Scarmeas, Stern and Tang²⁴⁾. Against the clear clinical evidence of elevated Fe levels in AD, it is tempting to assume that the benefits of a Mediterranean diet on AD do not exclusively rely on a distinct lipid intake but also on a lower level of Fe intake. In contrast, the foods with the highest contents of Cu (beef liver, oysters, molluscs, certain nuts, almonds and cocoa⁽Reference Lindow, Elvehjem and Peterson^173, Reference Lurie, Holden and Schubert¹⁷⁴⁾) are neither typical nor atypical for a Mediterranean diet. Its benefits for AD⁽Reference Scarmeas, Stern and Tang²⁴⁾ therefore are probably not based on differences in the intake of Cu.

Although circulating Cu relates to the nutritional status of Cu⁽Reference Solomons¹⁷⁵⁾, the origin of the free Cu fraction is still under discussion, potentially relying on inflammation⁽Reference Akiyama, Barger and Barnum¹⁷⁶⁾ or an increased efflux from cortical cells⁽Reference Bush¹⁶⁾. The alteration of systemic and brain Cu level in AD therefore depends on diet to a degree that is not yet known. However, it is startling that the participants on the highest quintile of Cu intake in the Morris Study took Cu in vitamin or mineral supplements⁽Reference Morris, Evans and Tangney³²⁾. This gives significance to the dietary intake of Cu in the elderly regardless of the current status of understanding.

In this light, it is of note that dietary intakes of different metals and their physiological levels in the body are not independent of each other and other factors. Thus, an elevation in Fe levels can be secondary to a high-Cu diet as Cu facilitates Fe intake⁽Reference Dunlap, James and Hume¹⁷⁷⁾, and hypercholesterolaemia might be induced by misbalancing dietary metal intake⁽Reference Squitti, Salustri and Siotto⁹⁷⁾. Moreover, the inverse correlation of the blood ceruloplasmin:TF ratio⁽Reference Squitti, Salustri and Siotto⁹⁷⁾ may also suggest that brain Fe accumulation is related to systemic alterations, which in turn depend on diet.

Therefore, the practical implications could be to avoid Cu-containing supplements and a high intake of SFA or excessive diet of Fe. Fe deficiency in developed countries is most common in certain subgroups of the population including toddlers and women of childbearing age⁽Reference Looker, Dallman and Carroll¹⁷⁸⁾. Data from the nutritional survey 2007–2008, National Health and Nutrition Examination Survey III, showed that the average nutritional intake of Cu is between 1·3 and 1·6 mg/d (1·1–1·3 mg/d) for adult men (women) in different age cohorts⁽¹⁷⁹⁾, whereas the dietary reference intake is 0·9 mg/d of Cu for adults⁽¹⁸⁰⁾. For Fe, the dietary reference intake recommended by the National Academy of Sciences is 8 (15–18) mg/d for adults (women aged 14–50 years)⁽¹⁸⁰⁾, while the actual intake is 15·6–18·1 mg/d for adult men and 12·6–13·2 mg/d for adult women at different ages⁽¹⁷⁹⁾. That means that the intake of Cu and Fe is up to 100 % higher than recommended. All efforts to reduce Fe or Cu levels should target the metals' physiological windows in order to avoid deficiency but achieve levels that are below the threshold levels critical for AD. What these thresholds look like with regard to AD is speculative with respect to the heterogeneity of different case–control studies. For instance, some studies found total Cu serum levels of >13 μg/l for the controls⁽Reference Agarwal, Kushwaha and Tripathi^79, Reference Smorgon, Mari and Atti⁸⁰⁾, while others reported levels of < 10 μg/l both, for controls and AD patients⁽Reference Molina, Jiménez-Jiménez and Aguilar^63, Reference Zappasodi, Salustri and Babiloni⁷⁷⁾. Also, AD is characterised by multiple aetiology and therapeutic ranges of metal reduction may not be generalised without considering further (dietary) parameters.

Despite the eschewal of high dietary intakes of Fe and Cu, various alternative approaches to lowering metal levels in the brain are under investigation. Several molecules acting as chelating agents are currently being developed⁽Reference Lee, Friedman and Angel^181–Reference Mandel, Amit and Bar-Am¹⁸³⁾, while some nutritional constituents have also been examined in relation to their metal-chelating activities.

The green tea polyphenol ( − )-epigallocatechin-3 gallate is being discussed as entailing protective effects on the progression of AD by different mechanisms, including an Fe- chelating function⁽Reference Weinreb, Mandel and Amit¹⁸⁴⁾. Curcumin, a polyphenolic diketone responsible for the yellow colour of turmeric, is also suggested to exert a protective effect against AD by binding Fe²⁺ and Cu²⁺ though the binding affinity for Zn²⁺ is small⁽Reference Jiao, Wilkinson and Christine Pietsch^185, Reference Baum and Ng¹⁸⁶⁾. There are more chelating agents being discussed for the treatment of AD. For instance, ethylenediaminetetraacetic acid has been reported to induce improvement in patients with AD⁽Reference Amit, Avramovich-Tirosh and Youdim¹⁸²⁾. Pharmaceutical compounds that exert their anti-inflammatory effects by interaction with Cu-, Fe- or Zn-dependent proteins have also been found to lower the risk for developing AD⁽Reference McGeer, Schulzer and McGeer¹⁸⁷⁾, and a number of other molecules that act as chelating agents are currently being developed⁽Reference Malecki and Connor¹⁸⁸⁾. An alternative approach of reducing the level of stored Fe in patients with AD has been hypothesised recently with the use of calibrated phlebotomy that could reduce stored Fe without causing anaemia⁽Reference Dwyer, Zacharski and Balestra¹⁸⁹⁾. All practical implications, however, should be taken cautiously in light of the limited clinical evidence and the multiple causality of AD.

Conclusion

In conclusion, the present systematic review suggests that a diet rich in Cu and Fe might aggravate the detrimental effects of a high intake of cholesterol and SFA on the risk of developing AD. The association is biologically plausible. However, since the relationship between dietary metal and fat intake and dementia is clinically not well examined, additional studies are necessary to further address which nutritional patterns best fit to certain risk groups in the population.