A Metallomic Approach to Assess the Associations between Plasma Metal Levels with Amnestic Mild Cognitive Impairment and Alzheimer’s Disease: A Cross-sectional and Longitudinal Study

Alzheimer’s disease (AD) involves the abnormal activity of transition metals and metal ion dyshomeostasis. The present study aimed to assess the potential of 36 trace elements in predicting cognitive decline in patients with amnestic mild cognitive impairment (aMCI) or AD. METHODS All participants (controls, aMCI, and AD) underwent baseline cognitive tests, which included the Mini-Mental State Examination (MMSE) and plasma biomarker examinations. We conducted a trend analysis for the cognitive tests and plasma trace elements and examined the correlations between the latter and annual MMSE changes during follow-up.


Background
Alzheimer's disease (AD) is a chronic progressive neurodegenerative disease that is usually mild at rst and gradually worsens over time. It is the most common cause of dementia [1]. AD typically presents with amnestic syndrome, characterized by poor learning ability with memory loss or atypical variants, manifested as an early impairment of language function (aphasic variant), visuospatial function (posterior cortical atrophy), executive function (frontal/behavioral-comportamental variant), and motor function (corticobasal syndrome) [2].

Mild cognitive impairment (MCI) is an intermediate state between the cognitive changes observed in
normal cognition and symptomatic pre-dementia [3]. Patients with MCI demonstrate unexpected cognitive impairments, based on their age and education level. However, in MCI, the disease severity does not meet the criteria for dementia [4]. A meta-analysis of 34 studies reported on a 10-20% prevalence of MCI in adults aged ≥ 65 years [5]. Amnestic MCI (aMCI) is frequently observed as a prodromal stage of AD, with an annual conversion rate of up to 25% [4,6]. The aforementioned statistics highlight the importance of early diagnosis and intervention for patients with aMCI [7].
The National Institute on Aging and Alzheimer's Association (NIA-AA) as well as the International Working Group have focused on a biomarker-based de nition of AD, emphasizing on the importance of β-amyloid (Aβ) deposition, pathological tau, and neurodegeneration in the AD continuum [8,9]. The accepted biomarkers are amyloid positron emission tomography (PET) ligand binding, atrophy on structural magnetic resonance imaging (MRI), hypometabolism in uoro-deoxyglucose (FDG)-PET, and several cerebrospinal uid (CSF) proteins, such as low levels of Aβ42 (or a Aβ 42 / Aβ 40 ratio), elevated total tau (Ttau), and elevated phosphorylated tau (P-tau). Nevertheless, CSF collection is an invasive procedure and should only be performed by a technical physician, which limits access. Similarly, the high cost and laborintensiveness of neuroimaging also limits the wide-spread application in primary care or clinical o cebased settings. For this reason, there is limited evidence for the inclusion of CSF biomarkers or PET markers in MCI and AD diagnosis during routine clinical practice [10]. Therefore, less invasive and economical blood-based biomarker testing, as well as genetic, clinical and demographic information will likely play an important role in population screening in the future.
Abnormal transition metal activity plays a critical role in the pathogenesis of AD [11][12][13]. Dyshomeostasis and the concentration of metal ions in neuro brillary tangles, senile plaques, and the CSF support this concept [14]. Speci cally, dyshomeostasis and the generation of toxic Aβ oligomers are likely responsible for the AD-associated synaptic dysfunction. Therefore, the inhibition or prevention of amyloid plaque aggregation, the pathological hallmark of AD, may be treated through the targeting of metal ions, metal complexes, or metal-protein compounds, such as metal chelators is a potential therapeutic implication. This highlights the signi cant role of metals in the etiology of AD [15]. However, no comprehensive study to date has evaluated the potential of trace metal biomarkers in predicting cognitive decline.
We aimed to assess the potential of 36 trace elements in predicting cognitive decline in patients with aMCI or AD. In addition, we used plasma levels of the trace elements to de ne optimal cut-off values in order to differentiate healthy elderly controls (HCs) from those with aMCI or AD.

Measuring the plasma trace elements
We collected the venous blood in heparinized vacutainer BD tubes (Becton Dickinson Labware, Franklin Lakes, USA) and stored them at − 20°C until the time of analysis. We quanti ed the trace elements using an Agilent 7800 ICP-MS instrument (Agilent Technologies, USA).
Prior to the analysis, 100 µL blood plasma were diluted (1:50 v:v) with a diluent comprising 0.05% Triton X-100 (Sigma-Aldrich, Co., MO, USA) and 1% HNO3 (ULTREX® Ultrapure Reagent, J.T.Baker, Co., Canada) in 18.2 MΩ cm distilled deionized water. We calibrated the system using standard solutions with different concentrations of trace elements prepared from Certipur® Certi ed Reference Material (Merck, Germany). We used metal solutions with a nal concentration of 0.10, 20, 30, 40 and 50µg/L for external calibration of the system. We performed a laboratory quality control via a permanent analysis of the certi ed reference material of blood plasma (ClinChek® Plasma/Whole Blood Control for Trace Elements, RECIPE Chemicals + Instruments GmbH, Germany).

ApoE genotyping
To obtain genetic information from samples collected from Taiwanese patients of Han Chinese ethnicity e ciently, the Taiwan Biobank (TWB) designed the TWB genotype array, based on the Affymetrix Axiom genotyping platform. The TWB genotype array enabled good-quality genotyping. Two single-nucleotide polymorphisms (SNPs, rs429358 and rs7412) de ning Apo E isoforms were genotyped using the TWB array.

Statistical analyses
We compared the demographics and scores of cognitive tests between the binary study groups (i.e., aMCI vs. control; AD vs. control) using independent sample t-test and the Fisher's exact test for the continuous and categorical variables, respectively. The linear trend across the disease groups (from control, aMCI to AD) was assessed using a linear contrast in the general linear model and the Cochran-Armitage trend test for the continuous and categorical variables, respectively. Considering the lack of normality, levels of trace metals between the binary study groups were compared using the non-parametric Mann-Whitney Utest. Likewise, we evaluated the linear trend of trace metals across the disease groups using the Jonckheere-Terpstra test. In addition, we further adjusted the age for evaluating the group difference (pairwise comparison) and the trend analysis in the cognitive tests and trace metals, as the age signi cantly differed between the groups.
We also evaluated the utility of individual trace metals in differentiating between the disease groups (aMCI vs. control; AD vs. control; and AD vs. aMCI) using the area under the receiver operating characteristic (ROC) curve (AUC). The 95% con dence interval of AUC was calculated using the DeLong's method. We used the Youden index to determine the optimal cut-off value.
The trace metals with signi cant differentiation ability in the previous ROC analyses were selected for further analyses. The association between those trace metals and the annual change of the MMSE score ([second assessment-rst assessment]/follow-up year) in either disease group (aMCI or AD) was further assessed using a partial correlation with an adjustment for the age, education level, and body mass index. All tests were two-tailed and a P-value < 0.05 was considered statistically signi cant. Data analyses were conducted using SPSS 25 (IBM SPSS Inc, Chicago, Illinois).

Patient pro les
Of the total 40 subjects, nine were non-disease controls. The remaining were comprised of 23 and eight patients with aMCI and AD, respectively. There was no difference in the sex distribution among the study groups. Patients in the AD group (82.9±8.6 years) were the oldest, followed by those in the aMCI (78.3±7.8 years) and the control group (67.0±6.3 years) (P for trend <0.001). The baseline cognitive tests were the poorest in the AD group, followed by the aMCI and control group (P for trend <0.05 in the baseline MMSE and tCDR tests) ( Table 1).

Trace elements
An increase in the disease severity was associated with lowered boron (B), bismuth (Bi), thorium (Th), and uranium (U) levels (all of adjusted P for trend <0.05) ( Table 2 and Figure 1). The levels of "B", "Bi", "Th", and "U" were signi cantly different between the aMCI and control groups as well as between the AD and control groups. However, the levels of "Mn", "Zr", "Sb", "Ba", and "Pt" were only signi cantly different between the aMCI and control groups. In contrast, the levels of "Cr", "Co", "Ge" and "T" were only signi cantly different between the AD and control groups ( Table 2).
The utility of trace metals to differentiate between the disease groups We evaluated the utility of trace metals in differentiating between the disease groups (aMCI vs. control; AD vs. control; and AD vs. aMCI). The levels of "B", "Hg" and "Th" could differentiate between all disease groups, including that between aMCI and AD. In contrast, the levels of "Ca", "Zr", "W", "Tl", "Bi", and "U" could differentiate both aMCI and AD from the control group. In addition, the levels of "B", "Mn", "Co", "Cu", "Ge", "Se", "Ba", "Pt", "Hg", "Pb", and "Th" could distinguish between the aMCI and AD groups (Supplemental Table 1.). The optimal cut-off value of the selected trace metals with satisfying differentiation ability (AUC>70.0%) and the corresponding sensitivity/speci city are displayed in Supplemental Table 2.

The association between trace metals and an annual change in MMSE scores
We eventually selected the trace metals with signi cant differentiating ability to evaluate their association with the annual change in MMSE scores. Higher levels of "B", "Zr", and "Th" were signi cantly associated with a greater cognitive decline in the aMCI group. In contrast, higher levels of "Ca" were signi cantly associated with a less cognitive decline in the aforementioned group. Higher levels of "Mn" were associated with a greater cognitive decline in the AD group (Table 3).

Discussion
This is the rst clinical study to evaluate the relationship between multiple serum metal levels and cognitive decline in healthy participants as well as those with aMCI or AD. The plasma concentrations of "B", "Bi", "Th", and "U" decreased with an increase in the disease severity. Moreover, "B", "Bi", "Th", and "U" levels were signi cantly different between patients with aMCI as well as AD and the healthy controls. The ROC analyses revealed that the plasma concentrations of "B", "Hg", and "Th" could differentiate between the disease groups (aMCI vs. control; AD vs. control; and AD vs. aMCI). "B" demonstrated high AUCs for aMCI versus the controls (97.6%, cut-off value: ≤73.1 ug/l) and AD versus the controls (100%, cut-off value: ≤47.1 ug/l). "Hg" revealed the highest AUC to differentiate AD from aMCI (79.9%, cut-off value: ≤1.02 ug/l). Following an adjustment for the potential confounding factors in the aMCI group, while higher baseline levels of "Ca" were associated with a smaller cognitive decline, those of "B", "Zr", and "Th" were associated with a rapid cognitive decline. In contrast, higher baseline levels of "Mn" were associated with a rapid cognitive decline in the AD group.
"B" levels were negatively associated with aMCI and AD. "B" is an essential trace element, abundant in fruits, vegetables, walnuts, and pulses. Recent animal and human studies have reported that long-term dietary supplementation with walnuts may reduce the risk or delay the progression of aMCI and AD [19,20]. There is increasing evidence for the bene cial effects of "B" on human health, particularly in promoting hormone and immune response, in ammation, oxidative stress regulation, and central nervous system function [21]. Furthermore, "B" deprivation leads to poor performance in tasks, such as movement speed and exibility, attention, and short-term memory in older adults [22]. In other words, the aforementioned studies highlight an association between "B" levels and cognitive function. In addition, "B" plays an important role in human brain function and cognitive protection.
The plasma "Th" levels were negatively associated with aMCI and AD. Moreover, higher baseline levels of "Th" were associated with a faster cognitive decline. An animal study reported that "Th"-treated mice demonstrated impaired learning and memory performance, similar to our results [23]. Furthermore, it resulted in the activation of acetylcholinesterase in mouse brain [23]. This necessitates further research on humans to reveal the underlying association between "Th" and cognitive function.
The plasma "U" concentrations were negatively associated with aMCI and AD. Daily dietary intake as well as water consumption are the most common ways of ingesting "U". Root crops, such as potatoes and sweet potatoes contribute the highest "U" content in the diet [24]. Moreover, sweet potato anthocyanins can enhance memory and improve cognitive de cits, which in turn may be related to its antioxidant properties [25,26].
Higher baseline levels of "Ca" were associated with less cognitive decline in patients with aMCI. A longitudinal population-based study from Sweden revealed that women receiving "Ca" supplements are at higher risk of developing dementia (odds ratio, 2.10; p = 0.046) [27]. Moreover, according to the unadjusted trend analysis in mixed-gender groups, "Ca" levels were lowest in the healthy control group and highest in the aMCI and AD groups. Recent studies have reported on the association between the disruption of intracellular Ca2 + homeostasis and the neuropathology of AD, memory loss, and cognitive dysfunction [28,29]. Increased intracellular "Ca" in the endoplasmic reticulum (ER) is the possible mechanism by which presenilin mutations disrupt intracellular "Ca" signaling. Furthermore, preclinical studies have revealed that excess ER Ca2 + release through the inositol 1,4,5-trisphosphate receptor or the ryanodine receptor is related to tau and amyloid pathology, and contributes to memory and learning de cits [30,31]. Therefore, calcium dyshomeostasis plays a critical role in the pathogenesis of AD.
We observed an inverse association between "Hg" levels and both aMCI and AD. There are three major groups of "Hg" compounds, namely elemental, inorganic, and organic. "Hg" is converted to methylmercury by bacteria, which enters the food chain and bioaccumulates in predatory sh [32]. Fish consumption is the primary source of methylmercury exposure [32]. Seafood, including shell sh and n sh is the largest contributor to organic "Hg" exposure in the human population. A systemic review mentioned that long-chain omega-3 fatty acids in a high-sh diet can delay cognitive decline in elderly individuals, without dementia [33]. The serum "Hg" levels of our subjects was within the normal range, thus indicating a normal dietary intake (normal value: <20µg/L for women aged ≥ 50 years and men aged > 18 years) [34]. Nonetheless, ICP-MS can only detect total "Hg" and fails to distinguish between the organic and inorganic forms. However, patients with aMCI or AD may likely reduce their seafood intake, thereby reducing organic methylmercury exposure, compared to healthy controls. We did not use a detailed food frequency questionnaire, including the types, frequency, and amount of seafood intake.
This made it di cult to explain the inverse correlation between "Hg" levels and aMCI and AD, thus necessitating further investigation.
Our study established an association between higher baseline levels of "Mn" and rapid annual cognitive decline in patients with AD. "Mn" is an essential metal that maintains the normal functions of the human body. However, increased "Mn" levels in the brain are associated with impaired motor coordination, memory de cits, psychiatric disorders, and Parkinson's disease [35][36][37]. An animal study reported that the overexpression of Aβ in transgenic mice lead to "Mn" accumulation in the brain, thus suggesting the role of Aβ may in "Mn" homeostasis and neurotoxicity [38]. A study conducted in China further mentioned that people with higher plasma "Mn" concentrations were associated with higher plasma Aβ peptides levels, similar to our results [39]. The aforementioned evidences suggested a relationship between "Mn" and AD, and the presence of shared pathophysiological mechanisms.
The strengths of our study include the robust statistical analysis, detailed cognitive examinations, prospective design, and large response rate at follow-up. However, our study had some limitations, which should be considered while interpreting the ndings. First, the NIA-AA research framework de nes AD as a biological construct rather a clinical diagnosis. However, AD diagnosis focuses on the biomarkers of brain Aβ deposition, pathologic tau, and neurodegeneration [8]. Our study population was limited to clinically evaluated patients with positive results for the cognitive tests. Moreover, we did not obtain amyloid and Tau positron emission tomography scans nor any CSF biomarkers. Second, cognitive decline is closely related to neurodegenerative biomarkers, such as hippocampal atrophy [40]. The pathophysiological link between the trace metals and neurodegenerative biomarkers remains to be determined. Third, the participants were recruited in a tertiary medical center. Therefore, our results may not be generalized to other populations, such as elderly people living in the community. However, there was also some strength of the single-center design, which made it possible to systematically and uniformly collect all the data of all participants. In addition, the cognitive assessment conducted by the board-certi ed clinical psychologist, making information bias less likely. Fourth, our sample size was small. Therefore, our ndings need to be veri ed by large-scale studies in future. Fifth, we only measured the metals at a single time point, which may re ect a short period of exposure. Therefore, long-term serial measurements of trace metals may help researchers explore their relationship with cognitive decline.

Conclusions
This was the rst exploratory study to compare the differentiating ability of trace elements biomarkers in patients with aMCI and AD. Several trace elements were signi cantly associated with the abovementioned cognitive tests and annual cognitive changes in the patients. The plasma concentrations of "B", "Hg", and "Th" could satisfactorily detect different stages of cognitive function in healthy controls and patients with aMCI and AD. Moreover, higher baseline levels of "B", "Zr", and "Th" and "Mn" were associated with a rapid cognitive decline in patients with aMCI and AD, respectively. Further large-scale longitudinal studies are required to replicate our preliminary ndings. The protocol was approved by the Institutional Review Board for the Protection of Human Subjects at the Tri-Service General Hospital (TSGHIRB 1-107-05-111). Written informed consent was obtained from all participants.

Consent for publication
All authors have approved of the manuscript and agree with its submission.

Availability of data and materials
The datasets during and/or analyzed during the current study available from the corresponding author upon reasonable request.

Competing interests
All authors declare that they have no competing interests.

Authors' contributions
All authors have contributed substantially to, and are in agreement with the content of, the manuscript. Conception/design, provision of study materials, and the collection and/or assembly of data: YKL and FCY; data analysis and interpretation: all authors; manuscript preparation: all authors; and the nal approval of the manuscript: all authors. The guarantor of the paper takes responsibility for the integrity of the work as a whole, from its inception to publication.