Alzheimer's disease (AD) is the most common form of dementia in older adults, and the number of affected individuals is set to escalate in the coming decades. Currently, there is no cure for AD, and available pharmaceutical therapies are focused on relieving the severity of associated symptoms. Extracellular plaques primarily comprising beta-amyloid (Aβ) deposits are one of the major hallmark pathologies of AD. The deposition of Aβ in the brain has been demonstrated to occur many years before the onset of clinical symptoms and contributes to neuronal death and loss of cognitive abilities (Villemagne et al., 2013). Thus, this research is focused on understanding and identifying interventions that can slow amyloid deposition and improve cognition in cognitively normal (CN) older adults who are currently at-risk for AD (based on high brain amyloid deposition) and in individuals with an AD diagnosis.

High levels of physical activity (PA) have been reported as one lifestyle factor that may protect against the development of AD pathology and, in addition, can slow brain atrophy and the associated progressive cognitive decline (Schultz et al., 2015; Delli Pizzi et al., 2020; Dougherty et al., 2021). Substantial evidence exists from animal studies to support the role of exercise in reducing brain Aβ levels, likely through multiple mechanisms. These mechanisms affect the Aβ production, by shifting the processing of amyloid precursor protein (APP) toward the non-amyloidogenic pathway via increases in a disintegrin and metalloproteinase (ADAM10) and reductions in beta-site APP cleaving enzyme (BACE). In addition, exercise in animal models has also demonstrated increased Aβ catabolism by upregulating Aβ-degrading enzymes such as insulin-degrading enzyme and neprilysin, among others (Moore et al., 2016; Koo et al., 2017; Khodadadi et al., 2018; Brown et al., 2019; Zhang X. et al., 2019; Zhang X. L. et al., 2019). Building on this animal work, several reports from human studies have linked higher levels of PA with lower brain amyloid deposition, measured under positron emission tomography (PET) using amyloid-binding ligands (Liang et al., 2010; Brown et al., 2012, 2013; Okonkwo et al., 2014; Rabin et al., 2019). As expected, CSF-related biomarkers were also affected by PA in cognitively healthy adults (Law et al., 2018).

Based on the current literature, there appears to be individual variability in the relationship between PA and AD-related pathologies, with carriage of the apolipoprotein E (APOE) ε4 alleles, the greatest known genetic risk factor for sporadic (i.e., non-familial) AD. PA has been associated with lower brain amyloid (Head et al., 2012; Brown et al., 2013) and Tau burdens (Brown et al., 2018) and preserved cognitive functions (Jensen et al., 2019) to a greater extent in APOE ε4 carriers, compared with non-carriers in both healthy controls and patients with AD. However, studies that have reported higher PA levels to be associated with reduced dementia risk years later and have provided inconsistent results regarding whether ε4 carriers or non-carriers receive the greatest benefit (Podewils et al., 2005; Rovio et al., 2005).

In addition to amyloid burden measured via neuroimaging, PA is associated with AD-related blood biomarkers such as plasma Aβ measured by enzyme-linked immunosorbent assays (ELISA) (Baker et al., 2010; Brown et al., 2013; Stillman et al., 2017). More recently, a high-performance immunoprecipitation-mass spectrometry (IP-MS) assay quantifying plasma Aβ peptides was validated in two independent cohorts (Nakamura et al., 2018). This assay was able to differentiate between individuals with high brain amyloid deposition from those with low brain amyloid deposition. The high specificity and sensitivity are key features of this assay and may highlight differences in Aβ levels that could go undetected when using less sensitive assays. As a consequence, this more sensitive Aβ assessment could be used to more specifically evaluate the efficacy of medical/physical therapies with regard to blood-based biomarkers. As different techniques of Aβ measurement may yield different observations, the goal of this study was to assess the relationship between brain Aβ deposition, AD-related blood biomarkers (assessed with a high-performance mass spectrometry assay), and habitual PA levels. We also wanted to evaluate whether the association between PA and AD-related biomarkers was more marked in APOE ε4 allele carriers.

Demographic details of the study participants are summarized in Table 1. No significant differences in age, sex, brain amyloid deposition, APOE ε4 carriage, and tracer used were found between participants with LMPA and HPA levels.

We evaluated the differences in plasma AD-biomarker levels (i.e., Aβ₁₋₄₂, Aβ₁₋₄₀, and APP_669−711) between the PA groups in all participants and after stratifying by APOE ε4 carriage status or brain amyloid status (Table 2). Lower plasma Aβ₁₋₄₂ levels were observed in the HPA group compared with those in the LMPA group, in all participants (p = 0.017 and p = 0.024, unadjusted and adjusted, respectively). After stratifying by APOE ε4 status or brain amyloid status, this difference was evident only in non-ε4 carriers (p = 0.003 and p = 0.004, unadjusted and adjusted, respectively) and in the Aβ- group (p = 0.014 and p = 0.012, unadjusted and adjusted, respectively). A total of 60 individuals (out of 81) who are brain Aβ- were also APOE non-ε4 carriers (out of a total of 91). This may explain why brain Aβ- and APOE non-ε4 carrier groups appear to have similar results.

Similarly, lower plasma Aβ₁₋₄₀ levels were observed in the HPA group compared with the LMPA group, in all participants (p = 0.043, unadjusted and a trend toward statistical significance upon adjustment, p = 0.057). Further, after stratifying by APOE ε4 status or brain amyloid status, this difference was evident only in the non-ε4 carriers (p = 0.020 and p = 0.019, unadjusted and adjusted, respectively) and in the Aβ- group (p = 0.043 and p = 0.031, unadjusted and adjusted, respectively) (Table 2). Plasma APP_669−711 was observed to be lower in the HPA group compared with that in the LMPA group, in all participants (p = 0.006 and p = 0.007, unadjusted and adjusted, respectively). After stratifying by APOE ε4 status or brain amyloid status, this difference was significant only in the non-ε4 carriers (p < 0.001 for both unadjusted and adjusted) and in the Aβ- group (p = 0.012 for both unadjusted and adjusted) (Table 2).

We have also assessed the effect of PA with regard to brain amyloidosis, and we did not find any significant effect of PA before any stratification. The same results were observed after stratification for APOE ε4 status or brain Aβ status (Supplementary Table 1).

In Table 3, the ratios of plasma APP_669−711/Aβ₁₋₄₂ and Aβ₁₋₄₀/Aβ₁₋₄₂ were also assessed with regard to PA and APOE ε4 status or brain Aβ status, and no differences between the PA groups were observed in the whole cohort, nor after stratifying by APOE ε4 status or brain Aβ status. We also assessed the composite z score for plasma ratios of APP_669−711/Aβ₁₋₄₂ and Aβ₁₋₄₀/Aβ₁₋₄₂, and we did not observe any significant differences related to different intensities of PA. Stratifying by APOE ε4 status or brain amyloid status, like in our previous analyses, did not result in significant composite z-score differences in any of the subgroups (Table 3).

However, levels of plasma Aβ₁₋₄₂ were significantly lower (p = 0.013 and p = 0.031, unadjusted and adjusted, respectively), and the ratios APP_669−711/Aβ₁₋₄₂ and Aβ₁₋₄₀/Aβ₁₋₄₂ were significantly higher (APP_669−711/Aβ₁₋₄₂: p < 0.001 for both unadjusted and adjusted; Aβ₁₋₄₀/Aβ₁₋₄₂: p = 0.001 and p < 0.001, unadjusted and adjusted, respectively) in APOE ε4 carriers compared with the ε4 non-carriers (Table 4). Similarly, the levels of Aβ₁₋₄₂ were significantly lower (p = 0.001 for both unadjusted and adjusted), and the ratios APP_669−711/Aβ₁₋₄₂ and Aβ₁₋₄₀/Aβ₁₋₄₂ were significantly higher (APP_669−711/Aβ₁₋₄₂: p < 0.001 for both unadjusted and adjusted; Aβ₁₋₄₀/Aβ₁₋₄₂: p < 0.001 for both unadjusted and adjusted) in the Aβ+ group compared with the Aβ- group (Table 4). More detailed analysis indicated that in most cases, the differences in Aβ₁₋₄₂ levels and APP_669−711/Aβ₁₋₄₂ and Aβ₁₋₄₀/Aβ₁₋₄₂ ratios are affected by APOE genotype and are irrespective of the intensity of the PA (Supplementary Table 2).

The linear regression analysis indicated that the levels of plasma Aβ₁₋₄₂ were significantly and negatively associated with brain amyloid deposition in the whole cohort and both PA groups (p < 0.001, p = 0.010 and p < 0.001 for the whole cohort, LMPA and HPA, respectively) (Table 5). These results were confirmed in partial correlation analyses upon correction for age, sex, and APOE ε4 status, (p = 0.001, p = 0.027, and p = 0.001 for the whole cohort, LMPA, and HPA, respectively) (Table 5). Conversely, no statistically significant associations were observed between brain amyloid and Aβ₁₋₄₀ or APP_669−711, regardless of whether the analysis was performed with or without adjustment for age, sex, and APOE ε4 status (Table 5).

In this report, we assessed the association between PA and blood biomarkers (plasma Aβ₁₋₄₀, Aβ₁₋₄₂, APP_669−711 levels) in CN older adults. Additionally, we examined how the carriage of the APOE ε4 allele (i.e., an important risk factor for sporadic AD) (Corder et al., 1993) affects the relationship between PA and AD biomarkers. We observed that (a) individuals reporting higher levels of PA had lower plasma AD biomarkers in APOE ε4 noncarriers and brain Aβ deposition groups and (b) plasma biomarker ratios are associated with APOE ε4 carrier status and with brain Aβ deposition status.

To date, the linkage between PA, risk of dementia, and APOE ε4 status are not clear. PA has shown to be inversely associated with brain amyloid deposition and to a greater extent in APOE ε4 carriers (Head et al., 2012; Brown et al., 2013); however, contradictory results have been reported when examining dementia risk as an outcome measure (Podewils et al., 2005; Rovio et al., 2005). Our results indicate that high PA has a trend-level association with lower brain amyloid levels only in individuals with high brain amyloidosis. Additionally, high PA is associated, albeit nonsignificantly, with lower SUVR in noncarriers of the APOE ε4 allele, while it was not a factor in APOE ε4 carriers. These data are in partial contradiction with previously published findings, also using AIBL study data (Brown et al., 2013), in which the effect of PA on brain amyloid deposition was restricted to APOE ε4 carriers. Such discrepancies may come from the fact that these analyses were performed on different numbers of healthy controls, which may affect the final results. The main reason for using a different cohort was our interest in evaluating the effect of PA on AD biomarkers when these were assessed using a more specific assay. Such biomarkers were assessed using a high-performance mass spectrometry analysis (Nakamura et al., 2018), which has the advantage of having a higher specificity and sensitivity compared with commercial ELISAs, which have been widely used for the analysis of Aβ₁₋₄₀ and Aβ₁₋₄₂. Utilizing this more sensitive and more specific assay for the measurement of Aβ₁₋₄₂, Aβ₁₋₄₀, and APP_669−711, Nakamura et al. were able to identify individuals (i.e., CN, MCI, and AD) with aberrant brain amyloidosis. However, this study could only be performed using healthy controls from the AIBL cohort who had AD biomarkers assessed by this new technique, and while our results may suggest a novel approach, one of the limitations of this study is the size of the cohort used.

The analysis of our biomarkers indicated that higher PA was significantly associated with lower levels of Aβ₁₋₄₀, Aβ₁₋₄₂, and APP_669−711 in CN ε4 non-carriers, while no relationship was observed in those carrying at least one APOE ε4 allele. It must be noted, however, that while Aβ₁₋₄₂ levels were significantly higher in ε4 carriers, APOE ε4 status had no effect on Aβ₁₋₄₀ and APP_669−711 levels. These data are in accordance with other reports that show associations of higher PA with lower levels of plasma Aβ in humans and mouse models of AD (Baker et al., 2010; Stillman et al., 2017; Khodadadi et al., 2018), although our original report did not observe a similar association between plasma Aβ and PA (Brown et al., 2013). Our original report, however, indicated an effect of HPA on the Aβ₁₋₄₂/Aβ₁₋₄₀ ratio in APOE ε4 noncarriers (Brown et al., 2013). While the size of the cohort may still be a limiting factor, the more sensitive Aβ assays may also play a role in justifying these discrepancies. Increased specificity and sensitivity may detect forms of Aβ that could go undetected in regular ELISA kits, and this could lead to assessing Aβ levels more accurately. This, in turn, would allow for more reliable analyses involving Aβ levels that could lead to a more appropriate assessment of biomarkers, as seen with the findings of Nakamura et al., where plasma Aβ and a related fragment reflected brain amyloidosis with high accuracy.

Additionally, an inverse correlation between brain amyloid deposition and Aβ plasma levels observed in AD participants within the AIBL cohort suggested that patients with AD with higher brain amyloid deposition had lower plasma Aβ₁₋₄₂ levels (Lui et al., 2010). These results indicate that in AD, Aβ₁₋₄₂ is sequestered in the brain in the form of amyloid plaques resulting in lower plasma Aβ₁₋₄₂ levels. As data from this study indicated that increased PA was associated with lower levels of brain amyloid and plasma levels of Aβ₁₋₄₀, Aβ₁₋₄₂, and APP_669−711, we then evaluated if such inverse correlation was retained. As shown, we have reported that there is a significant inverse correlation between brain amyloid and plasma Aβ₁₋₄₂, which was not affected by the level PA. One possible mechanism could be that the main effect of PA was associated with lower brain amyloid and plasma Aβ₁₋₄₀, Aβ₁₋₄₂, and APP_669−711 levels (in non-ε4 carriers only) and is more likely a consequence of a shift toward a non-amyloidogenic pathway, while it has no effect on the transport of Aβ through the blood-brain barrier into the circulation.

The ratio Aβ₁₋₄₀/Aβ₁₋₄₂ (or its inverse) and the ratio APP_669−711/Aβ₁₋₄₂ have also been indicated in previous reports as specific predictors of brain amyloidosis (Ovod et al., 2017; Nakamura et al., 2018; Chatterjee et al., 2019; Schindler et al., 2019) or AD (Lui et al., 2010). Our study, as all others, reported that these plasma biomarker ratios can significantly differentiate CN healthy controls with high brain amyloidosis vs. CN healthy controls with low brain amyloidosis; however, PA does not significantly alter these ratios. Similar results were obtained when stratified for APOE ε4 carrier status, where these ratios were significantly different in APOE ε4 carriers vs. noncarriers (due to the underlying APOE ε4 effect on plasma Aβ₁₋₄₂), but these results were not affected by PA. A schematic representation of how PA can affect plaque formation and Aβ₁₋₄₂ transport across blood-brain barrier (BBB) is illustrated in Figure 1.

To summarize, our data indicated that (a) PA was associated with lower levels of AD-related plasma biomarkers in healthy control APOE ε4 noncarriers and Aβ-individuals, (b) plasma levels of Aβ₁₋₄₂, but not Aβ₁₋₄₀ or APP_669−711, inversely correlated with brain amyloidosis, and (c) PA was associated with lower brain amyloidosis in healthy controls at risk of AD, although the analysis approaches statistical significance. Although we have indicated that our study has some limitations, we have reported that PA influences AD biomarker levels, likely affecting the process underlying the amyloidogenic pathway. Further studies are, therefore, necessary to confirm the validity of our findings in a larger cohort and to determine the involvement of different levels of PA with regards to plasma AD biomarkers.

The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author/s.

The studies involving human participants were reviewed and approved by The Australian Imaging, Biomarkers and Lifestyle (AIBL) study of aging was approved by the Human Research Ethics Committees of St. Vincent's Health, Hollywood Private Hospital, Austin Health and Edith Cowan University (Australia). The patients/participants provided their written informed consent to participate in this study.

SR-S, VD, NK, and BB carried out experiments and collected data. SP, PC, and BB analyzed data and wrote the manuscript. MT, EH, SG, KT, PB, AN, SR-S, VV, CR, DA, BF, IM, HS, CM, and RM provided scientific input. RM supervised the project. All authors contributed to the article and approved the submitted version.

Funding support from Australian Alzheimer's Research Foundation (AARF), Alzheimer's Australia (AA), the Science and Industry Endowment Fund, CSIRO, Brightfocus, USA and the WA Department of Health, as well as industry sources. We acknowledge the financial support of the CRC for Mental Health and the Cooperative Research Centre (CRC) program is an Australian Government Initiative.

NK was employed by Shimadzu Corporation. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.