Lysophospholipids and branched chain amino acids are associated with aging: a metabolomics-based study of Chinese adults

A total of 64 inpatients who presented for medical examination at Xuanwu Hospital were enrolled in our study, including 32 older adults aged 60 years or older and 32 sex-matched young adults aged between 36 and 59 years. We excluded individuals with malignant tumors, dysfunction of one or more organs, or who were receiving glucocorticoid or immunomodulatory therapy. We collected data related to general characteristics (age, sex, weight, height, smoking, and drinking), physical capacity (like 4-m gait speed and maximum grip strength), blood pressure, medical history, prescribed drugs, and laboratory test results. The body mass index (BMI) was calculated as the weight in kilograms divided by the square of the height in meters. The 4-m gait speed was calculated by dividing 4 m by the seconds it takes to walk this distance on flat ground without assistance. Grip strength was defined as the maximum strength of both hands while standing upright, with arms hanging naturally down. Creatinine clearance (CCr) was estimated using the Cockcroft–Gault equation.

After at least 8 h of overnight fasting, smoking and alcohol abstinence, blood samples from participants were collected and centrifuged. Serum was drawn and stored in cryogenic tubes at − 80 °C within 2 h.

For the general characteristics and basic clinical data, continuous variables are shown as the mean (standard deviation); categorical variables are shown as frequencies (percentages). Continuous variables were compared using Student’s t test, and categorical variables were compared using the Pearson’s χ² test. Pearson correlation analysis was used to describe correlations between metabolites or physical functions and age. Differences were considered to be statistically significant with two-sided p value < 0.05.

For metabolomics data, metabolites whose content was below the detection limit were imputed with zero. Principal component analysis (PCA) and orthogonal partial least squares discriminant analysis (OPLS-DA) were performed to distinguish old and young groups and to screen potential aging biomarkers (with variable importance in the projection > 1 and p value < 0.05). We performed a random permutation test and computed the area under the receiver operating characteristic (ROC) curve (AUC) on the OPLS-DA model to evaluate it. The peak areas of metabolites were compared using Student’s t test. Metabolites with fold change > 1.5 or < 1/1.5 and p < 0.05, were more meaningful to aging.

Metabolite set enrichment analysis (MSEA) [18] and metabolic pathway analysis (MetPA) [19] based on SMPDB (https://​smpdb.​ca) and KEGG (https://​www.​kegg.​jp/​) were used to identify over-represented metabolite sets and pathways related to aging. Metabolites with two-sided p values < 0.05, as determined by Student’s t test, were included in the pathway analysis, and those without matched HMDB ID were excluded. Metabolite sets with fold enrichment > 2 and raw p value < 0.05, and pathways with raw p value < 0.05, were considered as more significant to aging.

The basic characteristics of the 64 sex-matched participants with serum UPLC–QTOF data are summarized in Table1. Older participants tended to be slower (p = 0.011) and weaker (p < 0.001), had a greater pulse pressure (p = 0.003) and more chronic diseases (p = 0.001), and took more drugs (p < 0.001). Older people showed lower levels of creatinine clearance, albumin, hemoglobin, total cholesterol, and low-density lipoprotein and higher levels of D-dimer. Grip strength (r = − 0.64, p < 0.0001) and gait speed (r = − 0.54, p < 0.0001) were strongly negatively correlated with age.

The QC samples were clustered well in the PCA score plot with good consistency (Additional file 1: Fig. S1). In the paired comparison matrix of 13 QC samples after log10 conversion, the correlation coefficients were all greater than 0.99 (Additional file 1: Fig. S2). After correction by internal standards, the average relative standard deviation value in QCs was 11.0%, indicating high quality of the data in this work.

By comparing with 66 standard compounds (Additional file 1: Table S1), 349 metabolites of 46 classes were identified from serum samples using ultra-performance liquid chromatography (UPLC)–quadrupole time-of-flight (QTOF) mass spectrometry. The class, name, database ID and confidence level [20] of identified metabolites were listed in Additional file 1: Tables S2 and S3. In PCA, the first two principal components (Dimention-1: 17.8%, Dimention-2: 11%) accounted for nearly 30% of the inter-group differences (Additional file 1: Fig. S3Aa–c). The metabolomics data of older adults could be distinguished from those of young individuals in PCA score plot (Fig. 1A). We performed OPLS-DA on the identified metabolites and observed significant separation between the old and young groups (Fig. 1B). The model was proven to work in a random permutation test (Additional file 1: Fig. S3Ba, b), and showed great sensitivity (0.97) and specificity (1.00) in the receiver operating characteristic (ROC) curve (Additional file 1: Fig. S3B, c). There were 76 metabolites with variable importance in the projection (VIP) > 1 and p < 0.05, in the t test, 34 with fold change (FC) > 1.5 or < 1/1.5 and p < 0.05 (Fig. 1C), and 30 with VIP > 1, FC > 1.5 or < 1/1.5, and p < 0.05 (Additional file 1: Table S4), respectively, which were considered as candidate biomarkers of aging with statistical differences (Fig. 1D). Of the 80 differential metabolites we identified, we observed lower levels of lysophosphatidylcholine (LPC) and lysophosphatidylethanolamine (LPE), including LysoPC (18:2, 20:4, 22:4), sn2 LysoPC (18:2, 20:4, 22:4), LysoPE (20:1, 20:4, 20:5, 18:1, 18:2, 22:6), and sn2 LysoPE (18:1, 18:2, 20:4, 22:6), in the serum samples of older adults.

Among the metabolite sets with fold enrichment > 1 (Fig. 2A), valine, leucine and isoleucine degradation (fold enrichment = 2.16) was the only one with raw p < 0.05 in enrichment analysis for aging (with up-regulation of l-valine [Val] and down-regulation of alpha-ketoisovaleric acid, l-isoleucine [Ile], leucine [Leu], and ketoleucine) (Additional file 1: Table S5). The pathways associated with aging were valine, leucine and isoleucine degradation, and valine, leucine and isoleucine biosynthesis (Fig. 2B, Table S6).

In the three BCAAs, l-Leu (r = − 0.32, p-0.010) and l-Ile (r = − 0.30, p = 0.017) were negatively associated with age, while l-Val had no statistical association with age (Fig. 2C).

LPC and LPE are important precursors for the biosynthesis of phospholipids (PLs), including phosphatidylcholine (PC) and phosphatidylethanolamine (PE), which are essential for maintaining the structure and function of biological membranes (including mitochondrial membranes) [21‐23]. We observed that serum levels of six LPC and ten LPE were down-regulated in older adults compared to young. In the Baltimore Longitudinal Study of Aging, lower plasma LPC was correlated with impairment of mitochondrial oxidative capacity after adjusting for age, sex, and height [24]. In terms of physical function, lower LysoPC (18:2) could longitudinally predict a greater decrease in gait speed in older adults [25]. A decline in circulating LPs has been associated with many age-related diseases, such as cardiovascular diseases [26], arterial stiffening [27], Alzheimer's disease [28, 29] and several types of cancer [30]. A similar phenomenon exists in aging animal models; in previous studies, PC and PE levels were reported to decrease with age in Caenorhabditis elegans, rats, and mice [31]. The longitudinal decline of LPC (18:0) and LPE (20:4, 22:6, 36:2, 36:3, 38:4, 38:5, 38:6, 40:6, and 40:7) with age in mouse kidney tissue was also observed by lipidomics analysis [32]. These data suggest that LPs are closely associated with aging. Quantification of circulating LPs could be a promising method to reflect the degree of aging or to influence LP levels through diet or drugs as potential interventions for aging. However, a systematic review showed that the sex differences in LPCs and PLs in aging are notable; indeed, PC, PE, and LPC decreased with age in men, while LPC and PE increased with age in women [33].

BCAAs (Leu, Ile, and Val) are essential amino acids that can only be obtained from the diet, not by endogenous synthesis. We found that valine, leucine, and isoleucine degradation and valine, leucine, and isoleucine biosynthesis were two significant pathways in aging, in which l-Leu and l-Ile were negatively correlated with age. Changes in BCAAs and related pathways in older adults may thus have a two-sided effect on aging.

BCAAs, especially Leu, are powerful stimulants of the mechanistic target of rapamycin (mTOR) pathway [34]. The increase in mTOR activity is considered to be a crucial mechanism of aging [1, 34]. Although accompanied by many negative effects, downregulation of BCAAs can also reduce activation of mTOR signaling and benefit lifespan, which may be an adaptation of organisms to fight aging. On the other hand, a lack of BCAAs may have negative effects on health-span. The decline in relative BCAA content disturbs the amino acid balance in vivo and suppresses appetite, thereby exacerbating age-related malnutrition [35, 36]. Circulating BCAA levels were positively correlated with weight and fat mass in both mice and men [37]. Robust adults tended to consume more BCAAs than frail adults [38]. It is widely believed that BCAAs can stimulate muscle growth as an anabolic signal, and BCAA deficiency may have adverse effects on muscle structure and function in the elderly.

Long-term dietary supplementation of essential amino acids, including BCAAs, may be a powerful method for the prevention and treatment of sarcopenia [39]. However, the effect of pure BCAA supplements in stimulating muscle growth and fighting aging is controversial. Currently, it is widely believed that BCAA supplementation would hardly have benefits to muscle growth, as protein synthesis requires not only signal regulation from BCAAs, but also other signals and abundant amino acid substrates [40]. Similarly, supplementation or restriction of BCAAs hardly has any major effect on aging in the absence of other nutrients [41].

Moreover, we observed up-regulation of l-Val in the old group, without statistical correlation with age, which was different from l-Leu and l-Ile. With similarities in structure and function, the three BCAAs are generally considered to be synchronous. BCAA levels are not only dependent on dietary intake, but are also affected by degradation [41]. Leu is a ketogenic amino acid, Val is a glucogenic amino acid, and Ile is a ketogenic and glucogenic amino acid. The differences in the degradation pathways could explain the unsynchronism of BCAA changes with age to some extent.

This study aimed to screen age-related biomarkers from serum samples using untargeted metabolomics methods. The operation procedures of clinical data and sample collection, physical function measurement, and metabolomics analysis were all standardized, which ensured the credibility of this research. However, chronological age has limitations in assessing aging and physiological function, without considering the influence of genes and the external environment on aging. In addition, the source of the recruited participants was a single-center and the sample size was relatively small. As a result, the extension of our results is limited to the Chinese population from diverse regions, ethnicities, and cultures. There is a gap in the number of participants from different sexes and a lack of comparison of the metabolic profiles between sexes. Besides, there was a lack of adjustment of interference factors, such as diet and environment in the data analysis. Although these limitations restrict the representation of our results, we initially reveal the metabolic features of aging in China, which provide potential targets for the quantification and intervention of aging. Longitudinal research and verification of aging biomarkers and mechanisms based on multiple populations of each sex will be carried out in the future.