Metabolomic profiling of exosomes reveals age-related changes in ovarian follicular fluid

All of the females underwent in vitro fertilization (IVF) or intracytoplasmic sperm injection (ICSI) at Shanghai First Maternity and Infant Hospital from February 2022 to December 2022. Women with medical disorders (such as polycystic ovary syndrome [PCOS], endometriosis, cancer, or a history of ovarian surgery), that could affect follicular development were excluded from the study. Participants were divided into two groups: a young control group (age < 35, n = 30) and an advanced age group (age ≥ 40, n = 30).

This study was approved by the Scientific and Ethical Committee of the Shanghai First Maternity and Infant Hospital affiliated with Tongji University. All of the participants gave their written informed consent and the collection of samples was approved.

All of the patients received controlled ovarian hyperstimulation by using a combination of gonadotropin-releasing hormone (GnRH) agonist or antagonist and recombinant follicle-stimulating hormone (FSH). FF was collected by using transvaginal ultrasound-guided aspiration from dominant follicles (> 18 mm), (yielding 2–3 ml) at 34 to 36 h after the administration of 10,000 IU human chorionic gonadotropin (hCG). The supernatants were stored at − 80 °C after centrifugation at 3000 ×g for 15 min to remove cell debris and other particles. The FF exosomes were collected with an ExoQuick-TC Exosome Precipitation Solution Kit (System Biosciences, Palo Alto, CA), which is widely used to obtain exosomes from different body fluids. However, the obtained exosome sample was not homogenous. In brief, 250 µl of ExoQuick Solution was added to 1 ml of FF in a 1:4 ratio. The mixture was thoroughly mixed and left to incubated overnight at a temperature of 4 °C. After incubation, the mixture was centrifuged at 1500 ×g for 30 min at 4 °C and the supernatant was carefully removed. The resulting exosome pellet was then resuspended in 500 µl of PBS and passed through a 0.22 µm filter to obtain the desired solution.

Exosomes were analysed via transmission electron microscopy (TEM) as previously described [33, 34]. Twenty microlitres of exosome suspension (5 µg/µl) was fixed on a continuous grid, after which it was negatively stained with 2% uranyl acetate solution for 1 min and air-dried. The samples were then observed by using an FEI Tecnai G2 spirit transmission electron microscope (FEITM, Hillsboro, OR) at an acceleration voltage of 120 kV [35].

Nanoparticle tracking analysis (NTA) measurements were performed by using a NanoSight NS300 instrument (Malvern Panalytical, Malvern, UK) with a 488-nm laser and sCMOS camera module (Malvern Panalytical). Flow measurements were performed at a flow rate of 50. Three 60-s measurements were taken and the captured data were analysed by using NTA 3.2 software [35].

The concentration of proteins was quantified by using the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific, Waltham, MA) following the manufacturer’s instructions. Proteins were separated via gel electrophoresis on 10% SDS-PAGE gels and then transferred to PVDF membranes via electroblotting [17]. After blocking with 5% BSA, the blots were probed with exosome-specific antibodies (CD9, CD63, CD81, and Hsp70; System Biosciences) at 4 °C overnight. After washing the membranes, they were incubated with secondary antibodies, and proteins were visualized by using enhanced chemiluminescence reagents (Thermo Fisher Scientific) [17].

Two internal standards were added to exosomes from 1 ml of FF in each serum sample. Gas chromatography–time of flight mass spectrometry (GC–TOFMS) was used to analyse the samples in the order of "control and aging" by using the electron ionization mode (utilized instrument used: Pegasus HT, Leco Corp., St. Joseph, MI). One quality control (QC) sample and one blank vial were run after each 10 exosome samples. The sample was injected using a splitless mode with a volume of 1 μl. A DB-5 ms capillary column, which was 30 m long and 250 μm in diameter with a 0.25-μm film thickness consisting of 5% diphenyl cross-linked with 95% dimethylpolysiloxane, was used to separate the metabolites. Helium (99.9996%) was used with a constant flow rate of 1 ml/min. The GC oven temperature was initially set to 80 °C and held for 2 min. The temperature was gradually increased from the initial temperature to 180 °C at a rate of 10 °C per minute. It was then further increased to 230 °C at a rate of 6 °C per minute and finally to 295 °C at a rate of 40 °C per minute. The temperature was maintained at a constant level at 295 °C for 8 min. The temperatures for the injection, transfer interface, and ion source were set to 270 °C, 260 °C, and 220 °C, respectively. The mass range was set from 30 to 600 with electron impact ionization of 70 eV. The acquisition rate was 20 spectra per second. The obtained files from the GC–TOFMS analysis were exported in NetCDF format by using ChromaTOF software. (v4.44, Leco Co., Los Angeles, CA) [36]. The CDF files underwent several preprocessing steps, including baseline correction, denoising, smoothing, alignment, time-window splitting, and multivariate curve resolution. These steps were performed by using a toolkit developed by MATLAB 7.0 (The MathWorks Inc. Natick, MA), R 10.2 (Lucent Technologies), and JavaSE 1.6 (Sun Microsystems). Data quality control was ensured by using the internal standard and QC, whereas data normalization was performed by using QC. The ion peaks that were generated by the internal standard were removed to avoid any interference in the final results.

By using the GC–TOFMS analysis and by comparing mass fragments and recent times (RTs) with the National Institute of Standards and Technology (NIST) 05 and our laboratory libraries, we annotated 127 metabolites. Our libraries encompassed over 800 metabolites and are continuously expanding. A principal component analysis (PCA) score plot based on the samples from the aging group and the normal control group did not exhibit distinct clusters, thus indicating that they were not separated from each other. Partial least-squares-discrimination analysis (PLS-DA) and orthogonal partial least-squares-discrimination analysis (OPLS-DA) models were used to visualize the metabolic differences among the aging and normal control groups. The R2X, R2Y, and Q2Y values of the OPLS model were nearly 1.0 for the aging and normal control groups, thus indicating a good ability to explain and predict variations in the X and Y matrices. The resultant P values for all of the metabolites were subsequently adjusted to account for multiple tests by using a false discovery rate (FDR) method. Metabolites with both multivariate and univariate statistical significance (variable importance in projection, VIP > 1 and P < 0.05) were considered to be potential markers capable of differentiating PC from the controls. Additionally, the KEGG database was utilized to identify the affected metabolic pathways.

Statistical analysis for acquiring data was performed by using SPSS (SPSS, Chicago, IL) and SIMCA-P 12.0.1 + (Umetrics, Umea, Sweden). For the normally distributed continuous variables, Student’s t test was used for comparisons between the different groups. Otherwise, the Mann − Whitney U test was used. The OPLS models were validated by using a permutation test. Qualitative data were compared by using either the Chi-square test or Fisher's exact test. Spearman’s rank correlation analysis was performed to estimate the correlation between the detected metabolites and the clinical characteristics of all of the subjects, we performed. A statistically significant result was defined as P < 0.05.

The mean basal serum FSH (mIU/ml) was lower in the young age group than in the advanced age group. The number of retrieved oocytes was significantly higher in the young age group than in the advanced age group. There was no significant difference between the two groups in basal serum luteinizing hormone (LH, mIU/ml) or basal serum estradiol (pg/ml) (Table 1).

Exosomes derived from FF were characterized via Western blot, TEM, and NTA. The NTA and TEM findings were in conjunction with the previously documented traits of exosomes (Fig. 1A and Fig. 1C). The NTA and TEM results were consistent with previously reported characteristics of exosomes [17, 35, 37, 38]. Western blot results showed that three exosomal markers (CD63, TSG101, and HSP70) were detected in exosomes derived from FF (Fig. 1B). In addition, we found that exosome concentration (number of particles per ml FF) and exosomal protein concentration (µg exosomal protein/µl FF) from FF in the advanced age group were slightly lower than those from the young age group; however, the difference was not significant (Fig. 1D, E, all P > 0.05).

Multivariate analysis via principal component analysis (PCA) did not demonstrate a perfect separation between the analysed data groups for each group (PC1 variance: 26.83%; PC2 variance: 12.61%) (Fig. 2A). To further distinguish the differences in metabolic profiles between the two groups, we applied an advanced supervisory discriminant model known as OPLS-DA (Fig. 2B). A clear discrimination between the NC-Exo and A-Exo groups was apparent in the scatter plot of the OPLS-DA model (Fig. 2B). Of all of the samples, identified metabolites accounted for 68%, unknown metabolites accounted for 16%, and other components accounted for 16% (Fig. 2C). The identified metabolites mainly included amino acids (33%), carbohydrates (18%), organic acids (20%), fatty acids (13%), lipids (5%), and nucleotide (3%) (Fig. 2C). To further understand the alterations in metabolites in the two groups, we used a Z score heatmap to provide an overview of the metabolic profile in all of the samples, which showed the relative change in each metabolite across all of the samples (Fig. 2D).

We detected 157 metabolites. Of these, 127 metabolites were identified by using JiaLib™ (one of the largest metabolomics libraries in the world) and 30 metabolites were not identified. Table 2 shows the differentially abundant metabolites that were obtained based on unidimensional statistical analysis in FF exosomes between patients in the young age group and advanced age group, including alcohols, alkylamines, amino acids, carbohydrates, fatty acids, indoles, nucleotides, and organic acids. In particular, we found that the levels of the 13 identified metabolites (one alcohol [2-hydroxypyridine], four amino acids [i.e. ratio of l-glutamic acid/l-glutamine, l-threonine, beta-alanine, and glutathione], two carbohydrates [alpha-lactose, D-maltose], one fatty acid [oleic acid], two nucleotides [inosine and cytidine] and three organic acids [the ratio of homogentisic acid/4-hydroxyphenylpyruvic acid, taurine and fumaric acid]) were significantly higher in the advanced age group than in the young age group, with fold changes ranging from 1.075 to 2.723 (P ≤ 0.05). Moreover, four identified metabolites (three amino acids [the ratio of 4-hydroxyphenylpyruvic acid/l-tyrosine, the ratio of l-glutamine/l-glutamic acid and the ratio of ketoleucine/l-leucine], and one organic acid [malonic acid]) were lower in the advanced age group than in the young age group, with fold changes ranging from 0.403 to 0.694 (P ≤ 0.05).

The enhanced volcano diagram shown in Fig. 3A shows the differentially abundant metabolites that were screened based on the one-dimensional criteria (P value and fold change value). In this figure, P values set thresholds of 0.05 and 0.1, and log_1.2 FC (FC is a fold change, multiple of intergroup change) set thresholds of ± 1. As shown in the volcano map based on multidimensional statistics, the highlighted metabolites in the upper right corner were significantly increased in the sample of the advanced age group versus the young age group, whereas the highlighted metabolites in the upper left corner were significantly decreased in the sample of the advanced age group versus the young age group (Fig. 3A). Figure 3B shows a Z score heatmap of significantly differentially abundant metabolites between the two groups. The results were consistent with the differentially abundant metabolite table (Table 2). The pathway analysis that was conducted on differentially abundant metabolites demonstrated significant KEGG pathway enrichment. Differentially expressed metabolites in the pathways were mainly related to pyrimidine metabolism, taurine and hypotaurine metabolism, and valine, leucine, and isoleucine biosynthesis (Fig. 3C).

Given the robust associations of some exosomal metabolites with aging, we, therefore, explored whether exosomal metabolites could serve as biomarkers to differentiate between patients in the advanced age group and the young age group. A total of 29 metabolites were subjected to a random forest classifier for potential metabolite biomarker analyses, and a set of 15 metabolites (the ratio of l-glutamine/l-glutamic acid, the ratio of l-glutamic acid/ l-glutamine, inosine, the ratio of 4-hydroxyphenylpyruvic acid/ l-tyrosine, glutathione, 2-hydroxypyridine, taurine, beta-alanine, the ratio of ketoleucine/ l-leucine, the ratio of homogentisic acid/4-hydroxyphenylpyruvic acid, alpha-lactose, d-maltose, l-threonine, cytidine, and oleic acid) were identified as the optimal set of metabolites to discriminate between the advanced age group and young age group (Fig. 4A). In the variable importance in projection (VIP) variable importance analysis, the differentially abundant metabolite importance between the young age group and the old group was projected (Fig. 4B). We used the 15 metabolites to draw a receiver operating characteristic (ROC) curve, and the area under the curve (AUC) was used as an estimate of the predictive accuracy of the panel of biomarkers. By using the top four important metabolites as indicated by the VIP model (Fig. 4B), an AUC higher than 0.70 was obtained (Fig. 4C). This favourable AUC value demonstrated that these metabolites have a high predictive ability to differentiate patients in the advanced age group from the those in young age group. This result corroborates with the previous models that were built (OPLS-DA), which all performed well in differentiating between the two groups with low error rates. These results suggested that the 15 FF exosomal metabolites may be biomarkers to differentiate between patients in the advanced age group and the young age group.

To analyse the correlation between the detected metabolites and the clinical characteristics of all of the subjects, we performed Spearman’s rank correlation analysis. The overall analysis displayed a clear separation between some clinical characteristics and the concentrations of the metabolites. Among the metabolites that are differentially expressed in FF, some are related to oocyte count and hormone levels (Fig. 5). Several metabolites (i.e. taurine, lactose, glutathione, maltose, oleic acid, melatonin, oxoglutaric acid, 2-hydroxypyridine, citrulline, arabinose, l-arabitol, and pyroglutamic acid) increased with age, and other metabolites (i.e. amino adipic acid, alpha-ketoisovaleric acid, uric acid, hydroxyphenyl pyruvic acid, keto leucine, malonic acid, and glutamine) decreased with age. These results demonstrated an age-related trend in metabolites.