Stratifications of metabolite concentrations by age and sex. We examined the two significant factors affecting metabolite concentration: the number of days from blood collection to processing to storage (Same Day: DateDiff0 or Next Day: DateDiff1) and sex (Male or female), as in the flow shown in Fig. 1a. If significant differences were found between these, we would divide the groups by each factor for the analysis. Our test results are summarized in Supplementary Table S1 and described in detail below.
First, we checked for any systematic differences between DateDiff0 and DateDiff1. Since the results of the K-S test showed that metabolites lacked normality, a Wilcoxon rank sum test was performed to check for differences in metabolite concentrations between DateDiff0 and DateDiff1. As shown in Supplementary Table S1, all metabolite concentrations significantly differ between DateDiff0 and DateDiff1 samples (p< 0.0001); therefore, separate statistics are used in the following analysis.
Next, we checked the difference in the coefficient of variation, and there were 17 types where DateDiff0 < DateDiff1, two types where DateDiff0 = DateDiff1, and 23 types where DateDiff0 > DateDiff1, indicating that there were many metabolites where the coefficient of variation of DateDiff0 was larger than that of DateDiff1. Specifically, among the metabolites with a difference of 0.01 or more, the metabolites with a larger coefficient of variation in DateDiff0 than in DateDiff1 were Acetone, Succinate, Threonine, Alanine, Leucine, 3-Hydroxybutyrate, Lactate, Glycerol, and Isoleucine. There could be several reasons for these differences. First, when the number of participants is larger, the likelihood of outliers increases, which could have been a factor in increasing the variability of the data. The presence of different subgroups can also increase variability. Sex differences (male-female differences) are one example of this, and in this study, significant differences were found in all metabolites in both DateDiff0 and DateDiff1 because of statistical testing. In Rist et al. (2017) 25, BCAA (branched-chain amino acids) and their metabolites have been shown to have higher concentrations in males than females, contributing to sex differences. In this study, BCAA (Valine, Leucine, Isoleucine) and Glutamine, a metabolite of Isoleucine, showed higher concentrations in all age groups in males than in females in DateDiff0. This suggests that the variability within DateDiff0 may be influenced by the presence of different subgroups due to sex differences. Similar results regarding sex differences have been reported in many previous studies 11, 21-24, and it is widely known that there are differences between males and females in metabolite concentrations.
From these results, the analysis was conducted separately for DateDiff0 and DateDiff1 data and divided by sex. To confirm the differences due to sex, we again used the Wilcoxon rank-sum test to the separated date set. As expected, significant differences were found in all metabolites (p< .0001).
Change of the concentration of each metabolite against age. For the divided data sets, we observed trend changes for each age group, and a metabolite index was calculated as the mean and the standard deviation of metabolite concentration (Supplementary Table S2). From the viewpoint of the diversity and reliability of the concentrations, the standard deviation for each metabolite is focused, and the trend by age is considered. The smaller the standard deviation, the higher the reliability of the data in the age group, and it is deemed to have stability in interpretation and prediction.
First, the metabolome indexes were compared in the male age groups from 20 to over 80. As a result, the metabolites Acetate, Glucose, Glutamine, Tyrosine, and Lysine showed the most minor standard deviations (SD) in the 20-24 age group for both DateDiff0 and DateDiff1. Notably, Glucose (Fig. 2a) was found to have the lowest mean value across all age groups. This suggests a consistent trend of lower values for these metabolites in the early 20s. Also, the age group with the most significant SD for Glucose was 50-54 for DateDiff0 and over 80 for DateDiff1. On the other hand, the highest mean value was seen in the 75-79 age group for both DateDiff0 and DateDiff1. Similarly, Glycerol shows the most minor standard deviation and mean values in the 25-29 age group for both DateDiff0 and DateDiff1.
In the typical middle age of humans, from 40 to 60 years old, very few metabolites showed the most minor standard deviation in DateDiff0. Creatine showed the slightest standard deviation for ages 45-49 in both DateDiff0 and DateDiff1. For ages 55-59, 3-Hydroxyisobutyrate only showed the slightest deviation in DateDiff0, while many metabolites exhibited the highest standard deviation. For example, in the 50-54 age group, Acetate, Glucose, 2-hydroxybutyrate, Cysteine, Alanine, Methionine, Glutamate, Leucine, and Phenylalanine showed the highest standard deviation. This age group tended to have the highest standard deviation in many metabolites, second only to the group over 80.
In the age group 75-79 for males, the metabolites with the most minor standard deviation in both DateDiff0 and DateDiff1 were Lactate, 2-Oxoisocaproate, and Proline. Among these, the average values of 2-Oxoisocaproate and Proline in DateDiff0 were the smallest among the age groups, while Lactate had the smallest average value in DateDiff1. Although the number of males in the age group over 80 years old was small (63 for DateDiff0, 105 for DateDiff1), Glutamate, Threonine, N, N-Dimethylglycine, and Pyruvate showed the most minor standard deviation in both DateDiff0 and DateDiff1. Glutamate had the smallest average value in both DateDiff0 and DateDiff1.
The study by Ko et al. (2006) 40 investigated the relationship between age and blood glucose levels. The results indicated that blood glucose levels gradually increase with age, as age increases. Generally, younger people tend to have higher metabolic activity, with more active synthesis and breakdown of substances. As a result, metabolites such as Glucose may be consumed more quickly, leading to lower concentrations. On the one hand, it has been reported that glucose tolerance is reduced and glucose metabolism is decreased in the elderly 13, 41, 42. According to Basu et al.(2003) 14, the mechanism of age-related impaired glucose tolerance is that older people generally tend to have reduced muscle mass and increased body fat mass, and these body composition. These changes in body composition tend to improve insulin resistance and decrease insulin secretion with aging. This age-related decline in glucose tolerance is also a concern for the increased incidence of metabolic diseases such as diabetes 42, 43. For these reasons, indicators of metabolites associated with the glycolytic system, such as glucose and pyruvate, are essential in old age (Fig. 2a-c).
Next, the trends of metabolites according to age in females were evaluated. The SD of Proline and Glycine was the smallest in the 20-24 age group in both DateDiff0 and DateDiff1. Furthermore, regarding average values, Proline had the most significant value in the age group, while Glycine had the smallest. For Glycine, the age group with the highest standard deviation and average values was 55-59 in both DateDiff0 and DateDiff1. For Glucose, which showed the most minor standard deviation in males in the 20-24 age group, females showed the most minor standard deviation and average values in DateDiff1 only (in DateDiff0, the most minor standard deviation was in the 25-29 age group). For Acetate, the most minor standard deviation and average value were shown in DateDiff0 for females.
In females aged 35-39, 2-aminobutyrate, Isoleucine, Phenylalanine, and Valine showed the most minor standard deviation in both DateDiff0 and DateDiff1. Among these, Phenylalanine and Valine also had the smallest average values in both DateDiff0 and DateDiff1 (Fig. 2d-e). The reasons for the decrease in the metabolite concentrations of phenylalanine and valine in females during this age period could be changes in nutrient intake. Phenylalanine and Valine are essential amino acids in foods such as soy, dairy products, fish, meat, and eggs. Changes in dietary habits, appetite, nutrient intake, and meal content could decrease the intake of these amino acids. Moreover, Phenylalanine and Valine are components of proteins and are involved in protein synthesis and energy production in the body. Therefore, decreasing metabolite concentrations may suggest protein and energy metabolism changes (Supplementary Fig. S1-1, S1-2).
Also, in the age group 70-74, 3-Hydroxybutyrate showed the smallest standard deviation in both DateDiff0 and DateDiff1. In the age group 75-79, the standard deviation of 2-Oxoisocaproate was the smallest. In the group over 80 years old, the standard deviation of Pyruvate and Carnitine was smallest in both DateDiff0 and DateDiff1. Compared to males, the same results were seen in females, with the smallest standard deviation of 2-Oxoisocaproate in the 75-79 age group and Pyruvate in the group over 80.
In this way, trends by age became clear for each metabolite. The trend of small standard deviations in metabolite concentration represents the degree of individual differences or outliers among the participants in that age range, and generally, the smaller the standard deviation, the less variation in metabolite concentrations, indicating a more stable state. This suggests metabolite concentrations fluctuate within a specific range, showing relatively consistent values. When comparing metabolite concentrations by age group, the age groups with smaller standard deviations showed common trends in both DateDiff0 and DateDiff1. In this way, similar results and trends were obtained in the datasets of DateDiff0 and DateDiff1, which have separate distributions, which can be said to support the validity of the data. This is because consistent results were obtained regarding the values of metabolite concentrations despite the different datasets. This suggests the reliability and consistency of the data, demonstrating the robustness of our research results. This has clarified the similarities and expected trends in metabolite concentrations between the datasets of DateDiff0 and DateDiff1. This information will enhance the reproducibility and generalizability of our research results. From these results, it was shown that the datasets used in our research have high validity. This indicates that our research results are reliable and can be expected to provide helpful information for other research and clinical applications.
The transition of concentrations among age. In our analysis of age-related alterations in metabolite concentration, we segmented our study population into specific age groups. We conducted a comparative analysis using the Wilcoxon Rank Sum Test (with Bonferroni-adjusted p-value < .05).
To underscore the significance of these age-driven metabolic shifts, we limited our analysis to comparisons between consecutive age groups (e.g., between the [20–24 years group and the 25–29 years group]). Instances of significant differences are interpreted as drastic transitions in metabolite concentration attributable to aging. This approach allows us to pinpoint precise age intervals where substantial metabolic alterations occur, thereby spotlighting the pervasive impact of aging on metabolite levels. Since the results of DateDiff0 and DateDiff1 are quite small, we will focus on the description for DateDiff0 hereafter.
As a result of the Wilcoxon rank sum test, there were 22 metabolites where significant differences were seen in any age group comparisons for males in DateDiff0. Supplementary Table S3-1 summarizes the results of the age group comparison in DateDiff0 for males. Wilcoxon test p-values for age groups where significant differences were seen are shown in red, and effect sizes are displayed in green. Similarly, Supplementary Table S3-2 summarizes the results for females in DateDiff0. In DateDiff0 for females, there were 37 metabolites where significant differences were seen in any age group comparisons.
Fig. 3a(left side) and 3b(left side) detail the significant metabolite differences across age groups. For males, the most drastic metabolic changes were observed between ages 55-59 and 60-64 with 11 important metabolites, followed by ages 65-69 and 70-74 with eight, and 70-74 and 75-79 with seven. Meanwhile, females showed the most substantial changes between ages 45-49 and 50-54, with 17 significant metabolites, and between 50-54 and 55-59, with 13. Notably, the male-dominant age range of 55-59 to 60-64 showed five significant metabolites (Glycerol, Cysteine, 3-Hydroxyisobutyrate, Formate, and Phenylalanine) in females, four of which excluding 3-Hydroxyisobutyrate, were also effective in males. Older female age groups also exhibited many significant changes, with 10, 12, and 13 important metabolites in the age range 60-64 to 65-69, 65-69 to 70-74, and 70-74 to 75-79, respectively. Interestingly, while 20 metabolites showed no age group differences in males, only five metabolites (3-Hydroxybutyrate, Proline, Glycine, Alanine, N.N-Dimethylglycine) in females displayed similar traits. Four metabolites other than 3-Hydroxybutyrate were common to both sexes, indicating they might be less influenced by aging. For proline and glycine, the MGWAS analysis of Koshiba et al. (2016) 9 showed significant associations between genetic factors and these metabolites. Therefore, the effects of this genetic regulation may have a moderating effect on age-related changes. Fig. 3a(right side) and Fig. 3b(right side) show the frequency of significant differences for each metabolite at the top and between age groups where significant differences were found at the bottom. Males had ten and females nine metabolites, with significant differences in only one age group. Glutamate, Glycerol, Citrate, Cysteine, Pyruvate, and Succinate in males, and Glutamate, Glycerol, Carnitine, Glucose, Glutamine, Ornithine, Pyruvate, and Cysteine in females demonstrated significant differences in four or more times, suggesting these metabolites are likely age sensitive.
Focusing on the male age range [55-59 years: 60-64 years], 22 metabolites showed significant differences in some age groups, and half showed differences in this age group. Five metabolites (Glucose, Creatinine, Phenylalanine, 3-hydroxybutyrate, and Serine) showed no significant changes after this range. Six metabolites (Threonine, Glycerol, Citrate, Formate, Cysteine, and Acetate) showed significant differences in some older age groups. The remaining 11 metabolites demonstrated differences in older age ranges after this group, except 3-methyl-2-oxovalerate, 2-oxoisocaproate, and 3-methyl-2-oxobutyricacid.
In females, the age range [45-49 years: 50-54 years] was notable for metabolic shifts. Of the 37 metabolites with differences across age groups, 17 showed significant differences in this range. Five metabolites (Creatine, Methionine, Serine, Uridine, and 3-methyl-2-oxobutyricacid) showed changes unique to this range. Other metabolites began to change significantly after this range or showed long-term changes before and after. A few metabolites (Asparagine, 2-hydroxybutyrate, 3-hydroxyisobutyrate, and 2-aminobutyrate) showed significant differences only in specific ranges. In contrast, others (Phenylalanine, Glycerol, Formate, Valine, Isoleucine, Tryptophan, Leucine, and 3-methyl-2-oxovalerate) exhibited unique trends in specific age ranges. The period from the late 40s to early 50s, where the most significant differences in metabolite concentrations were seen, typically corresponds to the perimenopause period in females. Many preceding studies have established the influence of hormonal changes on metabolites during this period 18, 20, 36. In Watanabe et al. (2022) 18, a study was conducted comparing plasma metabolite concentrations between premenopausal and postmenopausal females aged 40-60. The trends observed in this age group (45-49 and 50-54 years) in our study were found to be like those in Watanabe et al. (2022) 18. The metabolites were Ornithine, Glutamine, Lysine, Carnitine, Betaine, and Arginine. All these metabolites showed higher average values in the postmenopausal group in Watanabe et al. (2022) 18; in our study, the age group of 50-54 showed significantly higher average values than the age group of 45-49. Furthermore, while Ornithine, Glutamine, and Lysine showed significant differences between premenopausal and MT (transition period) groups, no significant difference was observed with Carnitine, Betaine, and Arginine in the MT group, which lies between premenopause and postmenopause. In our study, Ornithine, Glutamine, and Lysine showed significant differences when comparing the age groups 40-44 and 45-49, and Betaine and Carnitine showed significant differences when comparing the age groups 50-54 and 55-59. These results suggest that our study has detailed the influence of age-related and longitudinal changes on these metabolites. From these results, similar trends were confirmed with the study by Tsuruoka (Watanabe et al. (2022) 18), which supports the objectivity and reliability of our study and demonstrates that our study results can capture age-related changes.
Impact of BMI. We analyzed how age and BMI affect metabolite levels using multiple regression. Age was independent, and BMI was an adjustment variable. Where interaction effects between age and BMI were found, they were interpreted as interacting with each other, and a Simple Slope Analysis was performed.
Supplementary Table S4-1 shows that in males, 17 metabolites were affected by age and BMI interactions. Among them, 11 metabolites, namely 3-Hydroxyisobutyrate, Cysteine, Creatinine, Tyrosine, Glycerol, 3-Hydroxybutyrate, Serine, Betaine, 2-Oxoisocaproate, Tryptophan, and 3-Methyl-2-Oxobutyricacid, showed significant differences across all BMI ranges. Three metabolites, including 2-oxoisocaproate, Tryptophan, and 3-methyl-2-oxobutyricacid, decreased with age. The four metabolites 2-Hydroxybutyrate, Threonine, Proline, and Glycine differed only in the low BMI range, while Threonine, Proline, and Glycine decreased with age. Finally, Asparagine and Lactate only differed in the higher BMI range, with Lactate decreasing with age.
As in Supplementary Table S4-2 in females, nine metabolites showed significant effects and interactions between age and BMI. Seven metabolites (2-aminobutyrate, Cysteine, Glutamate, Leucine, Serine, Valine, and Citrate) differed across all BMI ranges and increased with age. Asparagine, only differing in the lower BMI range, and Lactate, only differing in the higher BMI range, decreased with age (Supplementary Fig. S2-1, S2-2).
In terms of the effect of BMI, we evaluated the trend of the variable under conditions where the interaction effect with age was confirmed by performing a simple slope analysis (Supplementary Fig. S2-1, S2-2). In Moore et al. (2014) 35, a linear regression analysis was conducted after adjusting for age, sex, and smoking status, and 37 types of metabolites were reported to be related to BMI. Of the metabolites that were related to BMI in that study, those common to our study were eight types in males (3-hydroxyisobutyrate, 3-methyl-2-oxobutyrate, Glycerol, Tyrosine, Asparagine, lactate, 2-hydroxybutyrate, Glycine) and three types in females (Glutamate, Leucine, Valine). Our study also compared the partial regression coefficients (effect sizes) of BMI in the high BMI group for these metabolites in Supplementary Tables S4-1 and S4-2 (only metabolites with significant main effects and interactions are listed). The contributions of these metabolites to the model (signs of β) all matched. Also, in Ottosson et al. (2018) 37, a linear regression analysis was performed after adjusting for age and sex, and 19 types of metabolites were reported to have a significant association with BMI. Of these, the metabolites common to our study were four types in males (Tyrosine, Serine, Proline, Asparagine) and six types in females (Glutamate, Leucine, Serine, Valine, Citrate, Asparagine). All these metabolites matched the signs of the partial regression coefficients in the high BMI group in our study. From these, it is suggested that BMI consistently influences the fluctuations of these metabolites, indicating that our study has obtained highly reliable results. Our study clarified how the interaction of BMI and age regulates the concentrations of metabolites. For example, Tyrosine (in males) and Glutamate (in females), which showed a relationship with BMI in all the studies compared, and Lactate, which showed similar results in males and females in our study, are further explained.
In males, the Tyrosine concentration was higher in the high BMI group than in the low BMI group (mean +1SD, β = 0.176, p < 0.001) and showed an increasing trend with age. Similarly, in females, Glutamate concentration was higher in the high BMI group compared to the low BMI group (mean +1SD, β = 0.124, p < 0.001) and increased with age. As for Lactate, significant differences were observed between the age groups of 35 to 39 and 40 to 44 in both males and females, with an increase observed in males and a decrease in females. Subsequently, females reached a concentration peak in the age range of 35 to 39, while males had the highest peak between the ages of 50 and 54. Significant differences were observed between the age groups of 70 to 74 and 75 to 79 in both males and females, where Lactate concentration significantly decreased. Moreover, in both males and females, higher BMI values combined with increasing age showed a tendency for Lactate concentration to decrease (Males: mean +1SD, β = -4.253, p < 0.001; Females: mean +1SD, β = -3.629, p < 0.001).
By elucidating the detailed impact of BMI and its interaction with age on metabolites, we can gain insights into the biological processes and disease mechanisms associated with these metabolites. For instance, the variations in Tyrosine and Glutamate concentrations influenced by the interaction of BMI and age suggest their potential relevance to weight management and age-related diseases. Furthermore, when metabolite changes manifest more prominently under specific combinations of BMI and age, it indicates a higher likelihood of specific health risks for individuals. Such information can contribute to individual risk prediction and aid in developing tailored treatment and management strategies, fostering the advancement of personalized medicine. The metabolic concentration indices we have established can accommodate various conditions and variables in research and can be generalized to other studies and population groups, ensuring flexibility and generality.
Impact of Meals. The influence of age and meals on metabolite concentrations was also investigated. A multiple regression analysis was performed with each metabolite concentration as the dependent variable. The independent variables are age and two groups based on the 'Empty stomach group' postprandial time (Empty stomach and any time), incorporating interaction terms for the two factors of age and postprandial time.
As a result, there were seven types of metabolites in males where significant main effects and interactions were observed between age groups and post-mealtime (Supplementary Table S5-1); Formate, Pyruvate, 2-oxoisocaproate, 3-methyl-2-oxovalerate, 3-methyl-2-oxobutyricacid, Citrate, and Methionine. In females (Supplementary Table S5-2), 15 types of metabolites showed significant main effects and interactions between age groups and post-mealtime; Acetate, Betaine, Formate, Glutamine, Leucine, Ornithine, Phenylalanine, Succinate, Tyrosine, Valine, 2-Aminobutyrate, 3-hydroxybutyrate, 3-methyl-2-oxovalerate, 3-methyl-2-oxobutyricacid, and Creatine.
Among these are formate, 3-methyl-2-oxovalerate, and 3-methyl-2-oxobutyric.Acid were the three metabolites that showed significant differences in both males and females.