Using longitudinal data from the WRAP cohort, we were able to identify several plasma metabolites that were associated with the MIND diet, physical activity, smoking, and caffeine intake. Several of these were associated with AD endophenotypes at a strict Bonferroni-adjusted level of significance.
Among 27 plasma metabolites associated with the MIND diet score, most are derived from specific food groups which compose the MIND diet. Fruit and vegetable consumption in the MIND diet can be directly linked to metabolites related to antioxidants, specifically vitamin A and polyphenols, and gut microbiome by-products from high fiber foods. For example, carotenoids are found in vegetables and fruits that are orange in color and are also present in olive oil[25]. Our results showed that a higher MIND diet score is associated with higher vitamin A in the form of carotene diol(1), carotene diol(2), carotene diol(3), and beta-cryptoxanthin, a vitamin A precursor. S-Methylcysteine sulfoxide is also positively associated with the MIND diet score, and its major food sources are Brassicas (cruciferous vegetables; e.g., Brussels sprout)[26]. Seafood is a common staple in the Mediterranean diet and is designated as a positive component of the MIND diet. Both marine and shellfish are rich sources of a variety of long-chain omega-3 fatty acids[27, 28]. Specific fatty acids that were positively and significantly related to MIND diet scores in our study were docosahexaenoate (DHA; 22:6n3), eicosapentaenoate (EPA; 20:5n3), and stearidonate (stearidonic acid, SDA; 18:4n3), which are all abundant in seafood[27, 28]. Research suggests that DHA protects against AD and other dementias and is also beneficial for human cognitive function[29]. In addition, some vegetables, berries, and red wine also contain polyphenols, another class of antioxidants that includes 3-phenylpropionate (hydrocinnamate)[30]. A major benefit of antioxidants is that they offer protection against free radical damage, which can be beneficial for those at risk for neurodegenerative diseases[31]. Higher consumption of fiber-rich foods, such as whole grains, provides a source of insoluble fiber that is utilized by the gut microbiome to create compounds, including indolepropionate[32]. Overall, these metabolites also provide some biological evidence of how the MIND diet influences the human body through food and potentially protects people from cognitive decline and AD.
Among the metabolites associated with physical activity, glutamate, which refers to the anion of glutamic acid, was previously reported to decrease in plasma with prolonged physical exercise[33, 34], which was consistent with our findings. Alpha-hydroxyisovalerate has been reported to be negatively associated with physical activity energy expenditure in a blood metabolite study based on a Chinese cohort[35]. However, we detected higher alpha-hydroxyisovalerate with higher MET-hours. Future research is needed to clarify this relationship.
As for smoking, we were able to replicate several findings from previous studies for metabolites that were associated with smoking status, specifically o-cresol sulfate, 3-methyl catechol sulfate(1), N-(2-furoyl)glycine, oxalate (ethanedioate), 4-vinylphenol sulphate, piperine, threonate, and indolepropionate [36, 37]. Among them, the o-cresol sulfate and N-(2-furoyl)glycine are marked as respiratory irritants or potential health hazards by the PubChem database[38], which could be related to pesticide residue and chemical preservatives that are present in cigarettes. Other findings from our results such as 4-vinylguaiacol sulfate, 3-hydroxypyridine sulfate are also worth exploring in future studies. For example, according to the PubChem database[38], the 3-hydroxypyridine sulfate also called 3-pyridinol has been detected as a thermal degradation product from the smoke of burning leaves of Salvia divinorum, a Mexican psychoactive plant, and can cause respiratory irritation.
For caffeine intake, six metabolites, paraxanthine, 1-methylxanthine, caffeine, 1-methylurate, 1,7-dimethylurate, and 5-acetylamino-6-formylamino-3-methyluracil (AAMU), belong to the caffeine metabolism pathway. Metabolites associated with caffeine intake in our study, such as AAMU, theophylline, paraxanthine, caffeine, quinate, trigonelline, 1-methylxanthine, and 1-methylurate, replicated previous findings[39, 40]. Methylxanthines are a class of phytochemicals, including caffeine, theophylline, and 1-methylxanthine, that act as vasodilators and can reduce the risk of stroke, preventing further neurological damage[41]. Other findings such as 1,7-dimethylurate, 1,3-dimethylurate, 3-hydroxypyridine sulfate, and hippurate are also of interest. For example, it has been suggested that coffee intake would result in the elevation of the hippurate[42].
We identified four potential mediation effects. The MIND diet was associated with higher levels of beta-cryptoxanthin and a higher PACC score. As mentioned above, food components of the MIND diet, like vegetables and fruits, are rich in beta-cryptoxanthin. Beta-cryptoxanthin is an antioxidant protecting organs and tissues from oxidative damage and is the precursor of vitamin A[43]. Evidence from the literature suggests that vitamin A is crucial for maintaining higher central nervous system function in older people[44], and lower levels of vitamin A have been associated with cognitive decline and higher risk of AD[45–47]. On the contrary, beta-cryptoxanthin was negatively associated with smoking and a significant mediator of the negative association between smoking and PACC. A possible explanation for this mediation effect may be that metabolites or metabolism pathways related to smoking negatively impact levels of beta-cryptoxanthin[48], or that the lower level of beta-cryptoxanthin is due to a relatively unhealthy diet (e.g., having fewer vegetables) among smokers. Another mediation effect we detected was that higher levels of hippurate (hippuric acid) were associated with a higher MIND diet score and higher Immediate Learning score. Hippurate is strongly associated with the consumption of polyphenol-rich foods such as berries and coffee[49, 50]. A recent randomized control study conducted by Rutledge et al.[50] in 2021 showed that blueberry intake was associated with higher levels of blood hippurate and improvements in cognition. Our results also suggested that sustained physical activity may reduce the blood level of glutamate and this was associated with lower CSF NfL. Lower levels of CSF NfL are usually detected in cognitively normal individuals when compared to those with mild cognitive impairment and AD[51]. In the central nervous system, glutamate acts a neurotransmitter, has been linked to NfL[52], and is a key player in excitotoxicity [53], although it is not clear whether changes in circulating glutamate influence levels in the brain.
The WRAP cohort is an ongoing longitudinal cohort and has multiple time points of modifiable risk factor and metabolomics data per participant. The sample size is fairly large for each modifiable risk factor (around 900–1000). Although we did not have access to an independent cohort in which to replicate these findings, our results are based on the most conservative Bonferroni corrected p-values. In numerous instances our data replicate findings from published studies, and our identification of key metabolites associated with the four modifiable behaviors is consistent with what is known from nutritional and biological studies. The adjustment for correlated modifiable risk factors in each model increases confidence that these metabolites are biologically meaningful, and not simply driven by overall health status. However, there are some limitations. When we tested the metabolites with CSF biomarkers, especially in the mediation analysis that also included the modifiable risk factors, the sample size became smaller (around 165, depending on the model), reducing the power to detect significant associations. Future research to replicate our results and test for mediation in a larger independent sample is warranted. Since the relationship between reported dietary intake and the actual absorption and metabolism of various dietary components is complex, a more refined mediation model including multiple mediators or multiple pathways would be a logical next step for future research. Another limitation of this analysis is that the cohort is dominated by non-Hispanic white individuals, with little representation of other racial/ethnic groups. Thus, the findings may not be generalizable to other groups. Further study can focus on identifying or verifying these associations in other under-represented populations[54].
In summary, our study identified several plasma metabolites that are associated with the MIND diet score, physical activity, smoking, and caffeine intake. We suggest that these significant plasma metabolites may be valuable as biomarkers to track the activity of each modifiable risk factor. For example, beta-cryptoxanthin, DHA, and EPA could be used as biomarkers of the MIND diet or other dietary intake. At the same time, these metabolites may also unveil potential biological mechanisms of how modifiable risk factors influence the human body. Testing these hypotheses in intervention studies would be the next important step.