Introduction

Alzheimer’s disease (AD) is a global public health threat in the context of population aging [1]. Identifying risk or protective factors associated with AD is highly critical. The deposition of neurotoxic β-amyloid (Aβ) in the brain is one of the key neuropathologic feature of AD development [2]. Aβ42 and Aβ40 are two predominant isoforms of Aβ [3], which have been examined in an increasing number of research studies for AD pathogenesis and prognosis. Abnormal Aβ status in cerebrospinal fluid (CSF) analysis has been incorporated in the diagnostic criteria for AD [4]. However, CSF acquisition may not be feasible in large observational studies of older adults. Previous studies have demonstrated that peripheral blood Aβ levels were positively correlated with CSF Aβ levels and could be regarded as predictive biomarkers for brain amyloidosis [5]. Moreover, higher Aβ42 and Aβ40 levels in peripheral blood were associated with higher risk of AD [6,7,8], and higher Aβ42/Aβ40 ratio was typically associated with lower AD risk [9].

Selenium (Se) is an essential trace element for cerebral functions and the brain is the last Se-deficient organ in Se deficiency [10]. Changes in Se concentration and selenoprotein activity in blood and brain have been reported in AD and other brain diseases [11, 12]. Se had been found to inhibit the activity of γ-secretase, attenuate the alterations in APP expression and Aβ production, and have a protective effect on the neurotoxicity of Aβ [13, 14]. Selenoproteins are the major forms for Se to perform numerous biological functions [15]. Glutathione peroxidase (GPX) plays a unique role in protecting cells from free radical-induced oxidative damage [16], and GPX gene knockout mice has been reported to exhibit higher sensitivity to Aβ-induced oxidative stress [17]. Selenoprotein P (SELENOP) is an important Se transporter [18], which was found to be co-localized with Aβ plaques in the postmortem tissue from individuals with hallmark AD lesions [19]. Reduction of SELENOP gene expression rendered the N2A cell more sensitive to the toxicity of Aβ [20]. However, the function and mechanism of Se in AD development remain elusive, and there is a paucity of research in the association between Se and Aβ levels in bio-samples from the elderly population.

Considering the difference in Se nutrition status due to geologically uneven soil Se distribution [21] and the increasing concern on Se and AD [12], this study is focused on exploring the association between Se and Aβ levels in older adults with an extensive range of Se exposure.

Methods

Study Design and Participants

A cross-sectional study of 469 elderly people aged 60 and older from four rural counties in Enshi prefecture with diverse soil Se levels was conducted. Participants were enrolled between 2016 and 2017; biomarker measurements were conducted between 2018 and 2019. Data from 433 participants with serum Se status and Aβ measures was used for statistically analysis in this study. The four sites were from extremely high Se area (> 3.00 mg/kg), high Se area (0.40–3.00 mg/kg), Se-sufficient area (0.18–0.40 mg/kg), and Se-deficient area (< 0.18 mg/kg), respectively [21]. Residents were asked to enroll in the study if they met the following eligibility criteria: (a) had lived in the area for at least 30 years; (b) consumed local food and with no dietary supplement; (c) had no language communication problem; (d) agreed to complete a face-to-face interview and provide blood samples. All participants signed written informed consents before each interview. The study was approved by the Institute for Environmental Health and Related Safety, Chinese Center for Disease Control and Prevention.

Blood Se Measures

Fasting blood samples were collected in the morning using 5-mL purple top (EDTA) Vacutainer® tubes. All samples were stored in polypropylene tubes at – 80 °C before laboratory analysis. Blood samples were digested with concentrated nitric acid for 1 h and then placed in a boiling water bath for 4 h until the samples were clear.

Se concentrations in whole blood were measured by inductively coupled plasma mass spectrometry (ICP-MS) (Agilent 7700x, USA). The instrument parameters were as follows: RF power, 1550 W; collision mode, helium flow 4 mL/min; carrier gas flow, 1.0 L/min; plasma gas flow, 15 L/min; and auxiliary gas flow, 0.15 L/min. External calibration curves were applied for quantification. And, two quality control samples (Seronorm™ Trace Elements Whole Blood L-1 and L-2, Norway) were used to ensure the accuracy of the Se measurement. The detection limit for Se was 0.54 μg/L. The relative standard deviation (RSD) of replicate analysis of samples was less than 10%. In order to monitor the detection procedure, blank samples were analyzed for every 20 samples in all batches; 10% duplicated samples were also included; the samples were reanalyzed if the two results differed by > 10%.

Serum Selenoproteins Measures

Selenoproteins were measured by the enzyme-linked immunoassay (ELISA) method. Commercially available kits were selected according to pretest results. Serum GPX kits (Nanjing Jiancheng Bioengineering Institute, Nanjing, China) and SELENOP kits (Shanghai Enzyme-linked Biotechnology Co., Ltd., Shanghai, China) were used to determine the concentration according to the manufacturer’s instructions. The absorbance of each well was read on ELISA reader using 412 nm as primary wavelength for serum GPX and 450 nm as primary wavelength for serum SELENOP. As per manufacturer, for GPX, the intra- and inter-assay CV were 3.56% and 6.80%. For SELENOP, the reported intra- and inter-assay CVs were all less than 10%. In addition to quality control samples, 10% parallel samples were measured in each batch for quality assurance.

Serum Aβ Peptides Measures

Serum Aβ levels were determined using ELISA kits (Wako Pure Chemical Industries, Ltd., Osaka, Japan) according to manufacturer’s instructions. The absorbance was measured at a wavelength of 450 nm with a microplate reader (Bio-Tek, USA). In addition to quality control samples, 10% parallel samples were measured in each batch for quality assurance. Sensitivities were 0.12 pmol/L (assay range 1.0–100 pmol/L) in the Aβ40 assay and 0.08 pmol/L (assay range 1.0–100 pmol/L) in the Aβ42 assay. Intra- and inter-assay coefficients of variation were less than 10% in both assay systems.

Covariates

Information on socio-demographic characteristics (age, gender, and education) and lifestyle factors (alcohol consumption and smoking) were collected by standardized questionnaire from face-to-face interviews. Height and weight were also measured during the interview. The height-measuring instrument and electronic weighing scale were calibrated before measurement. The precision was 0.1 cm for height and 0.1 kg for weight. Body mass index (BMI, defined as body weight in kilograms divided by height in meters squared) was derived from height and weight measurements. The above variables will be used as covariates in the statistical analysis according to previous study [22].

Statistical Analysis

The normality of the data was assessed by the Shapiro–Wilk test. Except for age and BMI, the distribution of other parameters was skewed in our study. The continuous variables of the normal distribution were presented as mean ± standard deviation (SD), and continuous variables of skewed distribution were presented as median (interquartile range, IQR). Categorical variables were expressed as proportion (percentage). Comparisons of demographic characteristics among Se quintiles were performed by analysis of variance or Kruskal–Wallis rank-sum test for continuous variables and chi-square tests for categorical variables. Spearman rank correlation analysis was used to estimate associations among Se, SELENOP, and GPX.

Quantile regression models were used to examine the association between Se and Aβ levels. Quantile regression is a distribution-free method overcoming the limitations of traditional linear regression when the outcome variables fail to satisfy the normal distribution assumption [23, 24]. Conditional serum Aβ percentiles as functions of Se levels were estimated at the 5th, 25th, 50th, 75th, and 95th percentiles. The models were adjusted for age, gender, education, smoking, alcohol consumption, and BMI. Quantile regression coefficient graphs were used to present the changing trend of the association between Se levels and serum Aβ levels.

Statistical analyses were performed using STATA 16.0. All tests were two-tailed, and a P-value of less than 0.05 was considered statistically significant.

Results

Description of Participants

For the 433 participants included in this analysis, mean age was 72.38 ± 5.25 years, and 50.23% were females. Median and interquartile range (IQR) of blood Se, serum SELENOP, and GPX concentrations were 0.11 (0.08, 0.18) mg/L, 4.59 (3.62, 6.75) μg/mL, and 0.20 (0.16, 0.25) mmol/L, respectively. Median and IQR serum of Aβ42 and Aβ40 concentrations were 2.48 (1.18, 4.08) pmol/L and 25.15 (8.56, 49.30) pmol/L, respectively.

Demographic characteristics of study participants by blood Se quintiles were presented in Table 1. There were significant differences in age, SELENOP, GPX, Aβ42, Aβ40, and Aβ42/Aβ40 ratio, while no significant difference was observed in BMI, gender, education, smoking, and alcohol consumption among five Se quintile groups. The blood Se level was negatively correlated with age (rs =  − 0.098, P = 0.044). Non-smokers had higher SELENOP and GPX levels than that in smokers (P < 0.001). Participants who do not consume alcohol had higher serum SELENOP and GPX levels and lower Se levels than drinkers (P < 0.05). Blood Se was positively correlated with serum SELENOP (rs = 0.26, P < 0.001) and GPX (rs = 0.11, P = 0.03).

Table 1 Demography characteristics of study participants by blood selenium quintile groups
Full size table

Blood Se and Serum Aβ

Regression coefficients using quantile regression were presented for the 5th, 25th, 50th, 75th, and 95th percentiles of Aβ distribution (Table 2). Significant negative associations were observed between blood Se and serum Aβ42 and Aβ40 levels at all percentiles (P < 0.05). There was an overall increasing trend of the absolute value of β-coefficients between blood Se and Aβ42 and Aβ40 percentiles, indicating that these associations were generally stronger at the higher Aβ42 and Aβ40 percentiles than lower Aβ42 and Aβ40 percentiles (Fig. 1A and B). For example, for every 1 mg/L increase in blood Se level, Aβ42 decreased by 1.58 pg/mL at the lowest Aβ42 percentile, and Aβ42 decreased by 3.71 pg/mL at the highest Aβ42 percentile. Importantly, Se was positively correlated with Aβ42/Aβ40 ratio at the 25th (β: 0.03, 95%CI: 0.01, 0.05), 50th (β: 0.06, 95%CI: 0.02, 0.10), and 75th (β: 0.07, 95%CI: 0.01, 0.14) percentiles. Interestingly, only at the highest and lowest percentiles for Aβ42/Aβ40 ratio, the association between blood Se and Aβ42/Aβ40 ratio was not significant.

Table 2 Results of quantile regression models of the association between the blood selenium (mg/L) and serum β-amyloid (pmol/L) levels
Full size table
Fig. 1
figure 1

a–i Graphics of quantile regression coefficient between serum Aβ and blood Se, serum GPX and SELENOP levels. The green solid line represented the quantile regression estimated coefficients and the gray area represented 95%CI of the corresponding parameters. The black dashed line represented the ordinary least squares (OLS) estimate regression of the corresponding explanatory variable, and the area between the two black dashed lines represented the 95%CI of the OLS regression value. All estimations were adjusted for age, gender, education, BMI, tobacco smoking, and alcohol consumption.

Full size image

Serum GPX and Aβ

Significant positive associations were observed between serum GPX, serum Aβ42, and Aβ40 levels at all percentiles (Table 3). There was an overall increasing trend of the β-coefficients between serum GPX and Aβ42 and Aβ40 percentiles, indicating that the positive associations were generally stronger at the higher Aβ42 and Aβ40 percentiles than lower percentiles (Fig. 1D and E). For example, for every 1 mmol/L increase in serum GPX level, Aβ42 increased by 11.21 pg/mL in the lowest Aβ42 percentile, and Aβ42 increased by 23.51 pg/mL in the highest Aβ42 percentile. In addition, serum GPX was negatively associated with Aβ42/Aβ40 ratio at 25th (β: − 0.11, 95%CI: − 0.17, − 0.06), 50th (β: − 0.31, 95%CI: − 0.31, − 0.15), 75th (β: − 0.38, 95%CI: − 0.50,0.26), and 95th (β: − 0.88, 95%CI: − 1.75, − 0.02) percentiles, but there was no association at 5th percentile (Table 3). The negative associations were generally stronger at the higher Aβ42/Aβ40 percentiles than lower percentiles (Fig. 1F).

Table 3 Results of quantile regression models of the association between serum glutathione peroxidase (mmol/L) and serum β-amyloid (pmol/L) levels
Full size table

Serum SELENOP and Aβ

No statistically significant associations were observed between serum SELENOP, Aβ42, Aβ40, and Aβ42/Aβ40 ratio (Table 4 and Fig. 1).

Table 4 Results of quantile regression models of the association between serum selenoprotein P (mg/mL) and serum β-amyloid (pmol/L) levels
Full size table

Discussion

In this cross-sectional study, we observed that blood Se was positively associated with serum SELENOP and GPX levels. Higher Se levels were significantly associated with lower serum Aβ42 and Aβ40 levels and associated with higher Aβ42/Aβ40 ratio. Higher serum GPX levels were significantly associated with higher serum Aβ42 and Aβ40 levels and associated with lower Aβ42/Aβ40 ratio. No association with serum SELENOP and Aβ levels was observed.

Total Se and Aβ

Our results were consistent with a cohort study in Italy involving 71 subjects, showing that higher inorganic Se concentrations in baseline CSF samples were inversely associated with lower CSF Aβ concentration in older adults [25]. However, a case–control study involving 30 AD patients, 35 vascular dementia patients, and 40 healthy controls in Indian reported that there was no association between plasma Se and Aβ42 [26]. Nevertheless, the results of some in vitro or in vivo studies could also support our results. A cellular experiment showed that selenite could reduce Aβ40 generation in APP695 stably transfected Chinese hamster ovary cells [27]. An animal study reported that Se combined with fish oil could inhibit the activity of serum β- and γ-secretase in a mouse model of aging [28]. Another in vitro study also observed that selenite could reduce Aβ40, Aβ42, and sAPP beta production by reducing Aβ producing β-secretase and γ-secretase activities [29]. In addition, Se could attenuate lead exposure-induced changes in APP expression and Aβ production in Wistar rat hippocampus [30]. These findings suggest that blood Se may affect the APP processing by regulating the activity of β- and γ-secretase and further influence the serum Aβ levels.

Metabolic disturbances of Aβ homeostasis are the initial culprit in the pathogenesis of AD [31]. Recent reports suggested that high-accuracy assays of plasma Aβ42/Aβ40 ratio strongly predicted brain amyloidosis, and higher plasma Aβ42/Aβ40 ratios were correlated with lower Aβ deposition in brain tissue [5, 32]. However, no observational studies had focused on the association between Se and Aβ42/Aβ40 ratio before. As far as we know, our study is the first report showing that higher blood Se levels were associated with higher Aβ42/Aβ40 ratio, and we could further infer that higher blood Se levels could be associated with higher cognitive performance. Actually, this inference has been confirmed by our previous observational study in China [33] and another cross-sectional study in the USA [34]. Therefore, we speculate that Se could affect cognitive function by affecting Aβ metabolism; higher Se could reduce the generation of Aβ and increase Aβ42/Aβ40 ratio, which may delay the onset of cognitive impairment.

GPX and Aβ

Higher serum GPX was associated with higher serum Aβ and lower Aβ42/Aβ40 ratio in our study, which was opposite to the findings from total blood Se. This difference could be partly explained by the feedback effect of oxidative stress. As a group of antioxidant enzymes, the main biological role of GPX is to protect the organism from oxidative damage, for example, GPX4 could protect cortical neurons from oxidative injury and amyloid toxicity [35], and neurons of GPX1 gene knockout mice exhibited higher sensitivity to Aβ-induced oxidative stress [17]. Some studies suggested that the increase in antioxidant enzymes may represent a compensatory upregulation in response to increased oxidative stress [36, 37]. Interestingly, several studies have confirmed that oxidative stress could increase the expression of APP and the intracellular and secreted Aβ levels [38], and the increase of Aβ levels can aggravate oxidative stress in turn, forming a vicious cycle [39, 40]. This feedback effect could be found in other population-based studies; higher plasma GPX activity was observed in patients with AD and in those with vascular dementia compared with healthy controls [41, 42].

In addition, other selenoproteins may be responsible for the opposite finding in our study. Except for GPX and SELENOP, there are several more selenoproteins which could be associated with AD. In vitro studies showed that selenoprotein S could prevent Aβ production by degrading C99 which could be cleaved to Aβ [43], and overexpression of selenoprotein M could inhibit the α/γ-secretase activity related to the protection of Aβ42 production [44]. The thioredoxin reductase (TrxR) protein family may also have a protective roll in AD [10]. Therefore, we hypothesized that other selenoproteins or Se compounds might have covered up the inverse association of GPX compared with total Se.

SELENOP and Aβ

In the current study, we did not find significant association between serum SELENOP and Aβ levels. However, as the main Se transport protein, SELENOP has been supposed to be responsible for maintaining Se homeostasis and be vital for proper brain development and function [11, 45]. A postmortem study did observe an association that SELENOP was co-localized with Aβ plaques after examining postmortem brain tissue from individuals with the hallmark lesions of AD and individuals without these lesions [19]. In the triple-transgenic AD (3 × Tg-AD) model of mice, overexpression of SELENOP can inhibit the aggregation of Aβ in the brain of mice [46]. Meanwhile, some studies found consistent results as ours. An animal experiment found no effect on the number of amyloid plaque nor intracellular Aβ-containing neurons [47]. And, a cross-sectional study reported that CSF SELENOP levels were not associated with CSF Aβ levels [25]. The conflicting findings suggested that SELENOP may not directly affect the formation and deposition of Aβ; the function of SELENOP in brain health needs to be verified in future studies.

Strengths and Limitations

Our study has several strengths. Firstly, this is the first cross-sectional study with a relatively large sample to explore the relationship between peripheral blood Se levels and Aβ levels in the general population. Compared with CSF sample collection in other studies, blood sample collection in our study is less invasive. Therefore, blood Aβ measures are more suitable for epidemiological studies on general population. Secondly, our study design ensures an extensive range of Se exposure as subjects came from areas with different soil Se levels. Finally, the rural elderly Chinese participants were mostly lifelong residents consuming local food and without Se supplementation; hence, the ascertained Se levels could be inferred to be lifelong Se exposure without the influence of supplements.

There are also some limitations. On the one hand, we did not determine oxidative stress biomarkers to explore the relationship between Se and Aβ. On the other hand, with the cross-sectional study design, we cannot make causal inferences on the association between Se and Aβ levels. Therefore, more longitudinal studies are necessary to confirm our results.

Conclusion

Our results suggest that higher Se levels are associated with lower serum Aβ42 and Aβ40 levels and with higher serum Aβ42/Aβ40 ratio, and the results are specific for different selenoproteins. More longitudinal evidence is needed to confirm our findings, and further studies are needed to elucidate the underline mechanisms.