Abstract
Background:
The current research on advanced glycosylation end products (AGEs) and cognitive function is limited.
Objective:
We aimed to investigate the relationship between multiple plasma AGEs and cognitive function and mild cognitive impairment (MCI).
Methods:
Baseline data from The Lifestyle and Healthy Aging of Chinese Square Dancer Study was used in this cross-sectional study. Ultra-high-performance liquid chromatography tandem mass spectrometry was used to determine plasma levels of carboxymethyl lysine (CML), carboxyethyl lysine (CEL), and methyl imidazolinone (MG-H1). Four cognitive tests were used to obtain the four cognitive domain scores and the composite z scores. The Petersen criteria were used to diagnose MCI. The data were analyzed by multivariable linear and logistic regression models.
Results:
This study included 1,018 participants (median age 61.0 years, 87.3% female). After multivariate adjustment, the βs of the highest quartile of CML and CEL compared to the lowest quartile were –0.28 (–0.38, –0.17) and –0.13 (–0.23, –0.03), respectively, for the composite z score. For the four cognitive domains, CML was negatively correlated with memory, attention, and executive function, and CEL was negatively associated with memory and language function. In addition, higher CML was associated with a higher odds of MCI. MG-H1 was not associated with cognitive function.
Conclusions:
High plasma AGE levels were correlated with poorer cognitive function, particularly CML and CEL, higher levels of CML were also associated with higher odds of MCI. To clarify the effects of different AGEs on cognitive function and the underlying mechanisms, further longitudinal and experimental studies are needed.
Keywords
INTRODUCTION
As human life expectancy increases, the prevalence of dementia is rising rapidly, with an estimated 139 million adults worldwide living with dementia by 2050 [1]. Mild cognitive impairment (MCI) is an intermediate stage that lies between the normal aging process and the diagnosis of dementia, which is considered a significant precursor to the onset of dementia [2]. Given the limited effectiveness of available treatments and the huge burden of dementia, efforts to prevent cognitive decline and slow the progression of cognitive impairment could be an important strategy for reducing the incidence of dementia [3, 4].
Advanced glycation end products (AGEs) are a group of irreversible end products formed by exogenous or endogenous non-enzymatic reactions of sugars and macromolecules, including proteins, lipids, and nucleic acids [5]. Carboxymethyl lysine (CML), carboxyethyl lysine (CEL), and methyl imidazolidone (MG-H1) are representative examples of AGEs [6]. Age-related AGE accumulation is a natural part of aging and is also accelerated by hyperglycemia, inflammation, and oxidative stress [5]. In turn, many diseases and pathologies, including cognitive disorders, are strongly associated with the accumulation of AGEs [7]. Epidemiological studies have shown a connection between AGEs and cognition [8, 9]. A prospective study of 920 dementia-free individuals found that those with high urinary pentosidine levels performed worse on cognitive tests over 9 years [8]. Similarly, a cross-sectional study demonstrated a negative association between ELISA-measured serum AGEs levels and Montreal Cognitive Assessment (MoCA) scores in patients with type 2 diabetes mellitus (T2DM) and MCI [9]. Furthermore, there is a negative correlation between dietary AGEs and skin AGEs levels with cognitive performance [10, 11]. However, few studies have simultaneously examined the relationship between different types of AGEs levels and global cognitive function, specific cognitive domains, and MCI. Additionally, the existing studies lack consistency, potentially stemming from restricted sample sizes, variations in population sources, differences in AGEs metrics, and notably variances in testing methods. Compared to other methods, ultra-high-performance liquid chromatography tandem mass spectrometry (UPLC-MS/MS) is a fast and specific method for the detection of plasma-specific AGEs [12, 13]. Therefore, the relationship between different types of plasma AGEs levels measured by UPLC-MS/MS and cognitive function needs to be further studied.
The study was designed to examine the relationship between three plasma AGEs (CML, CEL, MG-H1) and cognitive function and prevalent MCI in the middle-aged and elderly population by UPLC-MS/MS.
METHODS
Study population
This cross-sectional study analyzed baseline data from the Lifestyle and Healthy Aging of Chinese Square Dancer Study (HealthyDance Study). The HealthyDance Study recruited middle-aged and elderly (≥45 years old) participants from China between 2020 and 2021 (ChiCTR2200056477). The Medical Ethics Committee of Medical College, Wuhan University of Science and Technology (ethics number: 201925) provided approval for the study. Each participant gave informed consent after being informed of the content and procedures of the study and then cooperated in completing a series of standardized questionnaires and physical measurements. With informed consent, some participants provided blood samples (n = 1,073). After excluding those with missing covariate information (n = 25) and incomplete cognitive function information (n = 30), a total of 1,018 participants were included in the final analysis (Supplementary Figure 1).
Ascertainment of plasma AGEs
Plasma samples were obtained by centrifuging whole blood samples at 3000 rpm for 5 min at 4°C using a desktop high-speed refrigerated centrifuge (CenLee16 R, Hunan, China). All plasma samples were stored at –80°C until analysis. Standard plasma samples were prepared by adding 25μL of mixed internal standard to 50μL of the plasma sample. The concentrations of three plasma-free AGEs (CML, CEL, MG-H1) were measured using UPLC-MS/MS. After sample preparation, the samples were analyzed using an Xevo TQ-XS (Waters TQD, Milford, USA) in electrospray ionization (ESI) positive multiple reaction monitoring (MRM) mode. CML, CEL, and MG-H1 levels were determined by the peak area ratio of each unlabeled peak area to the peak area of the corresponding internal standard [14].
Ascertainment of outcomes
The participants’ neurocognitive performance in different domains was evaluated using comprehensive neuropsychological tests. Specifically, the memory domain was assessed using the Auditory Verbal Learning Test (AVLT) [15], the language domain was assessed using the Verbal Fluency Test (VFT) [16], the attention domain was assessed using the Digit Symbol Substitution Test (DSST) [17], and the executive function domain was assessed using the Trail-Making Test-B (TMT-B) [18]. To exclude patients with dementia, the Clinical Dementia Rating (CDR) was employed [19]. Additionally, the scores of the neuropsychological tests were converted to z scores separately (higher z scores indicate better cognitive function). The mean of the sum of these z scores was used as the total score of global cognition, which reflected the participants’ global cognitive function. The z scores for cognitive function were calculated as follows: Memory z score = (AVLT-mean AVLT)/SD AVLT Language z score = (VFT -mean VFT)/SD VFT Attention z score = (DSST -mean DSST)/SD DSST Execution z score = (mean TMT - B -TMT-B) /SD TMT - B Composite z score = (memory z score + language z score + attention z score + executive z score)/4.
According to the Petersen Criteria [20], MCI was diagnosed as poor performance in at least one cognitive domain (<1.5 SD below normative means), along with a subjective memory decline complaint, intact daily functioning, and no dementia (CDR≤0.5).
Ascertainment of covariates
Based on the Lancet report on dementia prevention [21], multiple possible risk factors were considered as covariates for adjustment. Information on confounding factors was obtained using a standardized questionnaire. Sociodemographic factors, including age, sex, years of education, and marital status, were assessed. Age was recorded as a continuous variable. Sex was classified as male or female. Years of education were categorized into three groups: 0–6 years, 7–12 years, and > 12 years. Marital status was divided into two groups: married or non-married, which included individuals who were single, divorced, or widowed. Lifestyle factors, such as smoking status, drinking status, total energy intake, sleep quality, and physical activity, were also examined. Smoking status included current smoking and non-smoking. Drinking status included current drinking and non-drinking. Total energy intake was calculated based on information from the food frequency questionnaire. Sleep quality was assessed using the Pittsburgh Sleep Quality Index (PSQI), with a higher score indicating poorer sleep quality [22]. Physical activity was evaluated using the Physical Activity Scale for the Elderly (PASE) [23]. According to the guidelines of the World Health Organization (WHO), individuals were considered to meet the recommended physical activity level if they engaged in at least 150 to 300 min of moderate-intensity aerobic physical activity, at least 75 to 150 min of high-intensity aerobic physical activity, or an equivalent combination throughout the week [24]. Clinically related factors, namely body mass index (BMI), hypertension, and diabetes mellitus, were also considered. BMI was calculated as weight in kilograms divided by height in meters squared (kg/m2). Hypertensive participants were those who reported a diagnosis of hypertension, taking antihypertensive medication, or measurement of high blood pressure (systolic blood pressure≥140 mmHg or diastolic pressure≥90 mmHg) [25]. Similarly, diabetes mellitus was defined as previously diagnosed diabetes, taking hypoglycemic drugs, or hyperglycemia (fasting serum glucose≥7.0 mmol/L, or random serum glucose≥11.1 mmol/L) [26].
Statistical analyses
Baseline characteristics are presented as numbers (percentages) for categorical variables and as medians (interquartile range [IQR]) for continuous variables. The multiple linear regression model was applied to examine the relationship between three plasma AGEs and global cognitive function and four cognitive domains. In order to investigate the relationship between plasma AGEs and MCI, a logistic regression model was applied. In addition, the restricted cubic spline (RCS) model was used to investigate the nonlinearity of the associations.
AGEs levels were divided into quartiles (Q1-Q4) and participants in the lowest group (Q1) were used as the reference group. The results were presented as odds ratio (OR) or β with a 95% confidence interval (95% CI). Our analyses employed three sets of models. We adjusted for age, sex, education, and marital status in model 1. Model 2 was additionally adjusted for smoking status, drinking status, physical activity, total energy intake, and sleep quality. Further adjustments were made in model 3 for BMI, diabetes mellitus, and hypertension. Medians for each AGE category were included as continuous variables to obtain p values for linear trend.
We investigated whether the association of AGEs with overall cognitive function differed in different subgroups. Stratified analyses were performed according to age (<60 and≥60 years), obesity (<24 and≥24 kg/m2), education (≤9 and > 9 years), fasting blood (yes and no), and chronic diseases including hypertension, diabetes, stroke, and coronary heart disease (yes and no). Using SAS version 9.4 (SAS Institute Inc, Cary, NC, USA) software, all analyses were conducted. Statistically significant results were defined as p value less than 0.05 (two-sided).
RESULTS
Baseline characteristics
Participants’ characteristics are presented in Table 1. The study included 1,018 participants with a median age (IQR) of 61.0 (56.0, 66.0) years. Of these subjects, 889 (87.3%) were female, with median CEL, CEL, and MG-H1 levels of 56.1 (39.0, 82.9) nmol/L, 63.9 (50.2, 82.2) nmol/L, and 408.1 (279.4, 593.6) nmol/L, respectively.
Baseline characteristics of participants
Values are numbers (percentages) unless stated otherwise. CML, Carboxymethyl lysine; CEL, Carboxyethyl lysine; MG-H1, Methyl imidazolidone; IQR, interquartile range; PSQI, Pittsburgh Sleep Quality Index; BMI, body mass index (calculated as weight in kilograms divided by height in meters squared).
Association between AGEs and cognitive function according to each cognitive domain
The results of multivariate linear regression analysis for each AGE and composite z score are shown in Table 2. Based on the fully adjusted model, we found that plasma CML levels were negatively correlated with the composite z scores. The βs (95% CIs) for Q3 and Q4 were –0.18 (–0.29, –0.08) and –0.28 (–0.38, –0.17), respectively, compared to the Q1 group. Moreover, we discovered a negative correlation between plasma CEL levels and composite z scores (βQ4 = –0.13; 95% CI: –0.23, –0.03). The median trend tests for CML (p-trend<0.001) and CEL (p-trend = 0.026) are significant. Moreover, the βs for the composite z scores associated with per SD increment in CML and CEL were –0.10 (–0.13, –0.06) and –0.04 (–0.08, –0.002), respectively. Furthermore, we found a significant negative association with the composite z score only in the Q2 group of MG-H1. The results of the subgroup analysis were consistent with the primary outcomes (Supplementary Table 1).
Relationship between the plasma AGEs and composite z scores in multiple linear regression analysis
Data are presented as βs (95% CIs). Model 1: adjusted for age, sex, education, and marital status. Model 2: adjusted for model 1 + smoking status, drinking status, physical activities, Pittsburgh Sleep Quality Index, and total energy intake. Model 3: adjusted for model 2 + body mass index, diabetes, and hypertension. AGEs, advanced glycation end products; CML, Carboxymethyl lysine; CEL, Carboxyethyl lysine; MG-H1, Methyl imidazolidone. *Linear trend test using the median value of each category.
The results of multivariate linear regression analysis for each AGE and cognitive domains are shown in Fig. 1. We found that the group with higher plasma CML levels had worse performance on AVLT, DSST, and TMT-B in the fully adjusted model. The trend test results also confirmed the above associations (all p-trends<0.001). Although no statistically significant association was found in the language domain, per SD increment of CML was significantly negatively associated with the z scores in all four cognitive domains. For CEL, the group with higher plasma CEL levels performed worse on AVLT and VFT. Furthermore, it was only in the Q2 group that MG-H1 was negatively correlated with memory z scores.
Relationship between the plasma AGEs and cognitive domains in multiple linear regression analysis. The βs were adjusted for age, sex, education, marital status, smoking status, drinking status, physical activities, Pittsburgh Sleep Quality Index, total energy intake, body mass index, diabetes, and hypertension. The βs are indicated by the circles and the 95% CIs are reflected by the error bars. AGEs, advanced glycation end products; CML, Carboxymethyl lysine; CEL, Carboxyethyl lysine; MG-H1, Methyl imidazolidone. *Linear trend test using the median value of each category.
Association between AGEs and MCI
Higher plasma CML levels showed a strong positive association with higher odds of MCI after multivariate adjustment (Table 3). The multivariate-adjusted ORs (95% CIs) for the third and fourth quartiles (Q3 and Q4) of CML levels were 2.33 (1.14, 4.78) and 2.57 (1.25, 5.25), respectively, compared to the reference group (Q1). These findings were further supported by the trend test (p-trend = 0.002). In addition, the OR for MCI with per SD increment of CML was 1.49 (1.13, 1.97).
Relationship between the plasma AGEs and MCI in multiple logistic regression analysis
Data are presented as ORs (95% CIs). Model 1: adjusted for age, sex, education, and marital status. Model 2: adjusted for model 1 + smoking status, drinking status, physical activities, Pittsburgh Sleep Quality Index, and total energy intake. Model 3: adjusted for model 2 + body mass index, diabetes, and hypertension. AGEs, advanced glycation end products; MCI, mild cognitive impairment; CML, Carboxymethyl lysine; CEL, Carboxyethyl lysine; MG-H1, Methyl imidazolidone. *Linear trend test using the median value of each category.
Significant non-linear associations were observed between plasma CML and CEL and composite z scores (p for non-linear<0.001 for CML, Supplementary Figure 2A; p for non-linear = 0.010 for CEL, Supplementary Figure 2B) in multivariable-adjusted restricted triple spline analyses. Similarly, there was a significant non-linear association between plasma CML and MCI (p for non-linear = 0.006, Supplementary Figure 3A).
DISCUSSION
This study found that higher plasma CML and CEL levels were associated with poorer performance on global cognitive function and specific cognitive domains. Specifically, CML was significantly negatively associated with memory, attention, and executive function, while CEL was significantly negatively associated with memory and language function. Furthermore, higher plasma CML levels were associated with higher odds of MCI. These associations remained significant after adjusting for sociodemographic, lifestyle, and clinically relevant factors.
Our results are consistent with some previous epidemiological studies [9, 27–34]. In a cross-sectional study of 101 patients with T2DM, individuals with MCI had higher levels of serum total AGEs and a negative correlation with MoCA scores compared with individuals without MCI [9]. Several cross-sectional studies have similarly revealed that individuals with MCI have significantly higher levels of serum total AGEs than those without MCI [33, 34]. Similarly, Zhang et al. discovered a positive correlation between combined exposure to plasma-free AGEs and an elevated risk of cognitive impairment [28]. Given the limited representation of total AGEs at different AGEs levels, the association between single or multiple AGEs and cognitive function has been investigated in several studies. For instance, two observational studies reported a negative correlation between cognitive test scores and plasma CML levels [29, 30]. However, in another cross-sectional study involving 781 individuals, plasma CML and CEL levels were not significantly associated with global cognitive function. Interestingly, CEL was found to have a negative correlation with global cognitive function in 566 participants without T2DM [32]. These discrepancies can be partly attributed to different methods of AGEs measurement and cognitive assessment, as well as different population sources. Combining previous studies, there are significant regional differences in levels of AGEs, with significantly higher levels observed in developed countries, possibly due to high levels of AGEs in Western diets [35–37]. Thus, further studies of the association between plasma AGEs and cognitive function in the middle-aged and elderly population are warranted. We expanded on previous studies by examining the correlation between plasma AGEs and four cognitive domains, including AVLT, a representative episodic memory test [38]. The results showed significant associations between CML, CEL, and MG-H1 and poorer memory performance, indicating that high AGEs accumulation may act as a leading factor in memory impairment, a significant feature of amnestic MCI [4]. It is worth noting that CEL was found to be negatively associated with composite z scores, but not with MCI. This finding is reasonable from the perspective of MCI diagnosis since it is based on clinical assessments that evaluate cognitive function and functional status, rather than just specific test scores. In addition, studies have shown that CML is deposited in astrocytes and neurons of Alzheimer’s disease (AD) patients [39, 40], whereas few studies have been reported on CEL and MG-H1 [41]. Moreover, the strong resistance of glyoxal-derived AGE to AGE-degrading enzymes may also partly explain the association of CML with more cognitive markers [42].
Moreover, the role of AGEs in cognitive function is also reflected in dementia [43]. For instance, a recent large-scale prospective study conducted on 93,830 participants, utilizing the UK Biobank, discovered that individuals in the third tertile group of dietary CML, CEL, and MG-H1 had a 27%, 53%, and 24% higher risk of dementia compared to those in the first tertile group, respectively [37]. Another prospective study including 3,889 participants found a positive association between skin AGEs levels and dementia [11]. Additionally, lower levels of dietary AGEs are significantly associated with a lower incidence of AD [44]. Some studies have also found that the levels of AGEs in serum or cerebrospinal fluid may be a potential biomarker for the early diagnosis of AD [45, 46]. As a precursor to dementia, people with MCI have a 5% to 10% higher risk of developing dementia [47, 48]. This suggests that reducing the occurrence of MCI by identifying risk factors for MCI may be an effective way to prevent dementia. Further studies are necessary to observe the association between AGEs and the onset of dementia in the MCI population and to explore the possibility of AGEs as potential biomarkers for the transition from MCI to AD. In addition, investigation of the cellular and molecular mechanisms of CML and CEL versus other AGEs in AD using vitro or vivo models will provide new ideas for future prevention and treatmentstrategies.
One of the pathogenic mechanisms of AGEs is through cross-linking with proteins thereby leading to alterations in protein structure and function. AGEs have been shown to modify amyloid-β (Aβ) and accelerate soluble Aβ deposition [49]. Vitek et al. found that AGEs in the plaque portion of AD brain were three times higher than in healthy controls [50]. In addition, glyceraldehyde-derived AGEs have been reported to have direct toxic effects on neurons [51]. Another pathogenic mechanism of AGEs is the receptor pathway, which is the main pathway of AGEs action. The receptor of AGEs (RAGE) binds not only AGEs but also Aβ [52]. Firstly, AGEs-RAGE interactions activate the transcription factor NF-κB through multiple pathways (JAK-2-STAT 1, PI3K– AKT, MAPK– ERK, NADPH oxidase– ROS), promoting inflammatory responses and leading to neuronal damage and blood-brain barrier dysfunction (BBB) [53]. Secondly, RAGE binding to circulating Aβ mediates Aβ passage through the BBB into the CNS and deposition in the brain [54].
Identifying and preventing MCI at an early stage holds significant public health implications while exploring the relationship between AGEs and cognition further complements the existing evidence. Our present study found a negative association between plasma AGEs and cognition, indicating a need to improve diet quality and reduce poor lifestyle habits to decrease the intake and production of AGEs. It is worth noting that AGEs levels in the body are influenced not only by lifestyle choices such as sedentary behavior but also by a poor diet such as high in calories and fructose [55]. Epidemiological studies have confirmed that dietary AGEs may play a significant role in cognitive decline [56, 57]. For example, a prospective study of 684 non-demented older adults found that higher baseline dietary AGEs contributed to a faster 3-year overall cognitive decline through episodic memory and perceptual speed [57]. Additionally, Hoscheidt et al. found that cognitively normal participants had improved memory and reduced Aβ levels after a Mediterranean diet, but the effect was reversed after a Western diet with higher levels of dietary AGEs [58]. This is in line with our finding that AGEs levels are negatively correlated with memory. However, they did not find a similar effect in the MCI group, suggesting that dietary improvements may be effective before cognitive impairment. Moreover, our study revealed significant associations between plasma CML and CEL and cognitive function, implying that dietary choices lower in CML and CEL may have greater cognitive benefits. Nevertheless, the mechanisms behind the varying effects of different types of AGEs on cognition need to be clarified in further studies.
The main strength of our study is the detection of plasma levels of multiple AGEs and the assessment of global cognitive function, four cognitive domains, and MCI through a series of neuropsychological tests, providing multiple evidence for the link between multiple AGEs and multi-dimensional cognitive function. Several limitations should be considered in our study. Firstly, the cross-sectional study design prevents us from drawing causal conclusions or determining whether elevated AGEs levels occurred before cognitive decline or MCI. Secondly, the limited number of male participants compromises the representativeness of the population, thus limiting the generalizability of the results to males. To obtain more representative and generalizable findings in the future, it is advised to include a larger male population in the cohort. Thirdly, the cognitively related genes, particularly APOE ɛ4, were not conducted. In addition, although we attempted to account for many potential confounding factors, there may still be residual confounders from unmeasured factors.
Conclusions
In summary, we found that individuals with higher plasma levels of CML and CEL had poorer cognitive performance, and CML levels were positively correlated with the odds of MCI. These findings imply that specific plasma AGEs are related to cognitive function, and the potential mechanisms require confirmation in future studies.
AUTHOR CONTRIBUTIONS
Senli Deng (Formal analysis; Investigation; Writing – original draft; Writing – review & editing); Ruikun He (Formal analysis; Methodology; Writing – review & editing); Zhongbao Yue (Supervision; Validation); Benchao Li (Methodology; Writing – review & editing); Fengping Li (Data curation; Writing – review & editing); Qing Xiao (Software); Xiaoge Wang (Investigation; Visualization); Yuanyuan Li (Data curation; Project administration); Ruilin Chen (Investigation); Shuang Rong (Conceptualization; Funding acquisition; Project administration; Resources; Writing – review & editing).
Footnotes
ACKNOWLEDGMENTS
Thanks to the participants and interviewers in this study for their support and cooperation.
FUNDING
This work was supported by the Square Dance Cohort Fund of CNS Academy of Nutrition and Health (Beijing Zhongyinghui Nutrition and Health Research Institute), National Natural Science Foundation of China (No. 81941016), and the Scientific Research Start-up Fund of Wuhan University. The funder of the study had no role in the study design, data interpretation, writing of the report, or decision of publication.
CONFLICT OF INTEREST
Shuang Rong is an Editorial Board Member of this journal but was not involved in the peer-review process of this article nor had access to any information regarding its peer-review.
All other authors have no conflict of interest to report.
DATA AVAILABILITY
Privacy and ethical restrictions mean that the data is not publicly available, but further information about the datasets is available from the corresponding author upon reasonable request.
