We performed an elaborate analysis of 102 circulating biomarkers, previously studied in disease conditions such as type 2 diabetes, obesity, and cardiovascular disease (36–39), but hardly in healthy individuals with different lifestyles. Analysis of a selection of these biomarkers across two platforms showed similar results, underpinning their reliability, and indicating the robustness of these platforms. Except for leptin, individual biomarker levels were not significantly different between high and low aerobically fit females. Since leptin levels have been positively correlated to body fat percentage (40,41), the difference in leptin presumably results from a significant difference in body fat percentage between high-fit and low-fit females, further underpinning the validity of our data. Our observation that all other biomarkers were similar between the two groups, while previous studies in high and low aerobically fit individuals found significant differences in e.g., lipid and protein metabolites (42–47), is likely related to our standardized experimental set-up, as compared to other studies. We studied healthy, young-adult females of similar age and body mass index (BMI) in a highly controlled setting, while previous studies were performed with metabolically impaired individuals (38) and individuals with different BMI (42–44,47) or wider age ranges (42,47), in experimental conditions that were less standardized (42–45,47), and especially these factors impact circulating metabolite levels (38,42,43). Given that the levels of the analysed biomarkers were similar among the healthy females in our study, and multiple of these biomarkers showed dysregulation during disease, our findings imply that this biomarker set could be used to monitor progress from a healthy to an unhealthier state and may be use in health improvement interventions.
Studies that focus on recent exercise effects, i.e., effects on the day after exercise completion, are scarce compared to studies on acute or chronic exercise (15,18). Yet, recent exercise is especially relevant for biomarkers, as they can indicate whether physical activity of study subjects should be controlled prior to sampling. Here, we demonstrated that adiponectin, lipid metabolites, and inflammatory markers were most responsive to recent exercise, which is line with data from other studies (48–51). These findings suggest that future biomarker studies should consider standardization of study subjects’ physical activity 24 hours prior to blood sampling, especially when they include hormones, and markers related to lipid metabolism and inflammation.
Multiple separate clusters that were obtained in the heatmap and correlation matrix included biomarkers that corresponded to biomarkers embedded in our predefined, functional biomarker categories. Examples are the BCAAs, fatty acids, ketone bodies, short-chain acylcarnitines, long-chain acylcarnitines, cholesterol metabolites, and lipoproteins, which suggests that the response of biomarkers within these (sub)categories are interdependent. This has two important implications. First, one biomarker within a cluster could be considered as representative of the total cluster (e.g., isoleucine for the BCAAs), which could be of relevance for studies that measure only one or a limited number of biomarkers from one correlated cluster. Second, it provides opportunities for future studies to compute one total, standardized score for all biomarkers within a cluster that are strongly correlated (e.g., a total BCAA score). From a disease risk assessment point of view, such an integrated score will likely have a larger power and stronger predictive value as compared to individual biomarker levels. Previously, Wang et al. have found that BCAA levels could predict type 2 diabetes risk (36). Integrated BCAA analysis is therefore promising as health-status biomarker. Not all biomarkers from functional categories can be integrated because of differences in the individual responses (e.g., peptide hormones, inflammation markers, and short- vs. longer-chain acylcarnitines). Clustering outside the functional category was also observed. The inverse association between CRP and the amino acid glycine has also been demonstrated previously (52,53) and likely results from the inflammation modulating capacity of glycine (54,55). The positive association between N-acetylglycoprotein and lysophosphatidylcholine is also likely mediated via inflammation, since N-acetyglycoproteins plasma levels correlate with lipoprotein-associated phospholipase A2 levels (56), which generates lysophosphatidylcholine to promote inflammation (57,58). Direct positive links between glutamine, tyrosine, C5:0-OH and C5:1 acylcarnitines have not yet been described, but could be mediated by BCAA breakdown (59,60). The positive link between betaine and C18:2 acylcarnitine has not yet been demonstrated in humans, but may be related to fatty acid incorporation, as previously demonstrated in pigs (61). The observed correlations imply some revision of our a-priory functional categorization and, importantly, provide leads for biomarker integration and functional interpretation of changes in biomarker levels.
Next to the functional links between biomarker pairs, the hierarchical clustering models also showed that the degree of clustering for the intraindividual biomarker response i.e., the baseline and post-exercise biomarker values of one subject, was higher than the degree of clustering of the group (high-fit vs. low-fit) and the timepoint (baseline vs. post-exercise) biomarker responses. This finding suggests a considerable level of interindividual variation in our study population, which might also explain our observation that ~ 35% of the biomarkers was significantly impacted by recent exercise, but that clustering did not separate total baseline and post-exercise biomarker profiles. Since Krug et al., also showed that the interindividual variability was increased by using challenge tests (62), one could speculate that the challenged biomarker responses within one individual over time might act as a better predictor of health status, as compared to a singular analysis of the average biomarker levels of a larger group during basal homeostasis.
Our study included some strengths and limitations. One of the strengths is the integrated approach to analyze single as well as joined biomarker behavior in a healthy, homogenous study population at basal as well as challenged conditions, which provides us better insight in the behavior of biomarkers relatively to each other. An understanding of biomarkers in the healthy individuals is a prerequisite for their use in preventive health, for example biomarker guided dietary advise for health improvement. Another strength of our study is the focus on female individuals, since sex can affect metabolic responses (20,42), and females are often underrepresented in biomarker studies (15). One of the limitations of our study is that we could not determine the contribution of intraindividual variation, i.e., the day-to-day variation within an individual, as we sampled only twice in a relatively short time span. Although previous studies have demonstrated that the intraindividual variation for circulating adipokines (63), inflammatory markers (63,64), and metabolites (65,66) is smaller than the interindividual variation, we cannot exclude this source of error in our study. Second, we did not include additional post-exercise sampling timepoints, e.g., immediately post-exercise or a few hours post-exercise. Since the levels of most inflammatory markers, oxidative stress related markers and metabolites change acutely or in the first few hours after exercise, with each marker having its own kinetic profile (15,18) and the fact that biomarker kinetics can also differ between individuals as a result of interindividual variation (62), sampling at multiple timepoints after the exercise bout would have given insight in the exercise-induced biomarker behavior over time. Third, our study focused on a total of 102 biomarkers related to hormone signaling, inflammation and oxidative stress, and metabolism, while fitness level and single exercise stimulation have been associated with alterations in markers that were not included in our study, such as vitamins (44,45), ceramides (38), and individual lysophosphatidylcholines (38,42), which could possibly have provided additional insights in these biomarkers in view of the homogeneity of our study subjects characteristics and high level of study standardization.