Abstract Background Ovarian aging is the main cause of reduced reproductive life span, yet its metabolic profiles remain poorly understood. This study aimed to reveal the metabolic homogeneity and heterogeneity between physiological and pathological ovarian aging. Methods Seventy serum samples from physiological ovarian aging participants, pathological ovarian aging participants (including diminished ovarian reserve (DOR), subclinical premature ovarian insufficiency (scPOI) and premature ovarian insufficiency (POI)), as well as healthy participants were collected and analyzed by untargeted metabolomics. Results Five homogeneous differential metabolites (neopterin, menaquinone, sphingomyelin (SM) (d14:1/24:2), SM (d14:0/21:1) and SM (d17:0/25:1)) were found in both physiological and pathological ovarian aging. While five distinct metabolites, including phosphoglyceride (PC) (17:0/18:2), PC (18:2e/17:2), SM (d22:1/14:1), SM (d14:1/20:1) and 4-hydroxyretinoic acid were specific to pathological ovarian aging. Functional annotation of differential metabolites suggested that folate biosynthesis, ubiquinone and other terpenoid-quinone biosynthesis pathways, were mainly involved in the ovarian aging process. Meanwhile, dopaminergic synapses pathway was strongly associated with scPOI, vitamin digestion and absorption and retinol metabolism were associated with POI. Furthermore, testosterone sulfate, SM (d14:0/28:1), PC (18:0e/4:0) and 4-hydroxyretinoic acid, were identified as potential biomarkers for diagnosing physiological ovarian aging, DOR, scPOI, and POI, respectively. Additionally, SM (d14:1/24:2) strongly correlated with both physiological and pathological ovarian aging. 4-hydroxyretinoic acid was strongly correlated with pathological ovarian aging. Conclusions Metabolic homogeneity of physiological and pathological ovarian aging was related to disorders of lipid, folate, ubiquinone metabolism, while metabolic heterogeneity between them was related to disorders of lipid, vitamin and retinol metabolism. Clinical trial number Not applicable. Supplementary Information The online version contains supplementary material available at 10.1186/s13048-025-01625-2. Keywords: Untargeted metabolomics, Serum metabolic biomarkers, Lipids, Physiological ovarian aging, Pathological ovarian aging Introduction Reproductive decline is the earliest onset symptom of aging in females [[40]1, [41]2]. The ovaries, which produce oocytes and steroid hormones, are essential reproductive organs for female reproductive development, fertility maintenance, and endocrine homeostasis [[42]3, [43]4]. Ovarian aging leads to infertility and menopause due to reduced oocyte quality and quantity, and decreased ovarian function. Physiological ovarian aging gradually occurs with advancing development with age, with 35 years being commonly regarded as the onset of this process [[44]5, [45]6]. In contrast, pathological ovarian aging can arise from various causes, including hereditary, immune, medical, endocrine, inflammatory, lifestyle, and socio-psychological factors [[46]7–[47]9]. Pathological ovarian aging is a progressive condition that is subdivided into several stages: diminished ovarian reserve (DOR), poor ovarian response (POR), subclinical premature ovarian insufficiency (scPOI) and premature ovarian insufficiency (POI). The incidence of pathological ovarian aging is rising and it often occurs earlier in life. Among infertile women, 29.3 – 30% are diagnosed with DOR [[48]10, [49]11], with up to one -third of these patients also experienced POR [[50]12]. Notably, DOR can progress to POI within months to several years. The prevalence of POI is 1.8 – 3.7%, and 0.4% of cases occurring in women younger than 35 years old [[51]13, [52]14]. Although physiological and pathological ovarian aging share similar clinical symptoms, including hormonal changes and eventual ovarian failure, the extent to which these two types of ovarian aging are similar, as well as the underlying mechanisms driving their differences and progression, remain largely unknown. A better understanding of these similarities and differences could enhance early detection of ovarian aging, even in its early or subclinical stages, and provide critical insights for developing effective disease prevention and treatment strategies. Metabolomics, a promising approach in systems biology, enables the comprehensive characterization of circulating metabolites. As metabolic changes can reflect biological physiology and pathology, metabolomics hold great potential for discovering clinical biomarkers, monitoring disease progression and evaluating therapeutic outcomes [[53]15, [54]16]. Women with ovarian aging are prone to metabolic disorders, particularly those involving lipid, glucose, and bone metabolism [[55]17]. Previous studies on physiological ovarian aging have demonstrated changes in lipid, amino acid, and indoleacetic acid metabolism in perimenopausal women, highlighting potential cardiovascular and metabolic risks [[56]18, [57]19]. In the context of pathological ovarian aging, metabolomic studies have uncovered specific alterations. Liang et al. [[58]20] linked oxylipin metabolites and arachidonic acid (AA) metabolic pathway with oocyte development. Oxylipins could be produced by the autooxidation of AA, and AA is essential for oocyte maturation, development and fertility [[59]21]. Moreover, a recent plasma metabolomics study identified lipids such as arachidonoyl amide, 18-hydroxyeicosatetraenoic acid (HETE), and phosphoglyceride (PG) (16:0/18:1) as biomarkers with potential for diagnosing POI [[60]22]. Another study found that POR women with decreased AMH levels showed downregulated serum prostaglandin H2, cortexolone, and tetracosanoic acid [[61]23]. Despite these findings, the distinction between physiological and pathological ovarian aging remains unclear. While metabolomics provides dynamic insights into both physiological and pathological changes, few studies have examined the progression of ovarian aging over time. A long-term study [[62]24] revealed that women experiencing rapid declines in anti-Müllerian hormone (AMH) levels exhibited higher serum concentrations of metabolites such as phosphate, N-acetyl-D-glucosamine, branched chained amino acids (BCAAs), proline, urea, and pyroglutamic acid. In this study, we employed untargeted metabolomics to investigate differential serum metabolites associated with both physiological and pathological ovarian aging. We further explored homogeneity and heterogeneity in metabolic characteristics between the two forms of aging. Our findings contribute to a deeper understanding of the mechanisms underlying ovarian aging and provide valuable insights for clinicians to improve the diagnosis and prevention of this condition. Materials and methods Participants and sample collection Serum samples were collected from 70 participants at the First Affiliated Hospital, Guangzhou University of Chinese Medicine (Guangzhou, China), between March 2022 and October 2022. The participants were grouped as follows: 13 women with physiological ovarian aging (old), 12 with DOR, 14 with scPOI, 19 with POI, and 12 healthy women (control, Ctrl). This study was approved by the Ethics Committee of First Affiliated Hospital of Guangzhou University of Chinese Medicine (No. JY2022-002), and is in accordance with the Declaration of Helsinki. Informed consent was obtained from all the participants prior to sample collection. The inclusion criterion for the old group was (1) age ≥ 45 years; (2) without the history of diminished ovarian function. DOR, scPOI, and POI are different stages of pathological ovarian aging. Considering the complex diagnostic criteria of DOR and the overlap of DOR and scPOI [[63]25, [64]26], inclusion criteria of pathological ovarian aging were set as below: (1) age 20–40 years old; (2) AMH < 1.1 ng/mL and basic follicle-stimulating hormone (bFSH) < 15 mIU/mL (DOR), 15 ≤ bFSH < 25 mIU/mL (scPOI), bFSH ≥ 25 mIU/mL (POI). Participants who met the following criteria were excluded from the study: (1) diagnosis of other endocrine diseases, such as polycystic ovarian syndrome, hyperprolactinemia, or thyroid dysfunction; (2) diagnosis of ovarian endometriosis; (3) hormone or ovarian stimulation treatment in the past 3 months; and (4) history of oophorectomy. The participants’ characteristics are shown in Table [65]1. Serum was collected on the 2nd or 3rd day of menstruation. In cases of amenorrhea or menopause, serum was collected randomly. Then serum samples were stored at -80 ℃. Table 1. Characteristics of participants Item ctrl (n = 12) old (n = 13) DOR (n = 12) scPOI (n = 14) POI (n = 19) Age (years) 30.17 ± 5.65 51.85 ± 3.39^*** 33.42 ± 3.15 32.21 ± 6.78 32.89 ± 7.45 BMI (kg/m^2) 21.63 ± 0.86 22.70 ± 1.71 21.17 ± 1.18 22.13 ± 0.96 21.29 ± 1.24 Age at menarche (years) 12.92 ± 0.67 13 ± 0.58 13.08 ± 0.67 13.07 ± 0.73 12.95 ± 0.71 Gravidity 1.5 (0, 2.75) 3 (2, 4)^* 2 (0.25, 3) 2 (0.75, 2.25) 2 (0, 3) Parity 1 (0, 1) 2 (1, 2.5)^** 1 (0.25, 2) 1 (0, 2) 0 (0, 2) pregnancy loss 0 (0, 0) 0 (0, 1) 0 (0, 0.75) 0 (0, 0) 0 (0, 0) Diabetes (n(%)) 0 (0%) 0 (0%) 0 (0%) 0 (0%) 0 (0%) Cardiovascular disease (n(%)) 0 (0%) 0 (0%) 0 (0%) 0 (0%) 0 (0%) Hormones bFSH (mIU/mL) 5.26 ± 0.98 54.72 ± 26.00^*** 10.05 ± 2.75^*** 19.4 ± 2.78^*** 67.27 ± 28.7^*** bLH (mIU/mL) 4.92 ± 1.18 38.65 ± 18.09^*** 5.66 ± 2.36^* 8.26 ± 3.46 34.76 ± 20.58^*** bFSH/bLH 1.13 ± 0.36 1.49 ± 0.53 2.06 ± 1.01^* 2.84 ± 1.51^*** 2.17 ± 0.74^*** bE[2] (pmol/L) 185.67 ± 53.69 74.97 ± 53.59^** 188.66 ± 162.76 165.36 ± 82.27 74.73 ± 40.76^*** bT (nmol/L) 0.74 ± 0.39 0.37 ± 0.23 0.63 ± 0.37 0.35 ± 0.27^* 0.48 ± 0.29 AMH (ng/mL) 3.90 ± 1.88 0.01 ± 0.01^*** 0.43 ± 0.38^*** 0.33 ± 0.51^*** 0.03 ± 0.04^*** Lifestyles Smoking (n(%)) 1 (8.33%) 2 (15.38%) 1 (8.33%) 0 1 (5.26%) Coffee (n(%)) 1 (8.33%) 1 (7.69%) 1 (8.33%) 0 0 Tea (n(%)) 0 1 (7.69%) 0 0 1 (5.26%) [66]Open in a new tab Comparing to the Ctrl group, ^*P < 0.05, ^**P < 0.01, ^***P < 0.001. Ctrl: control; BMI: body mass index; bFSH: basic follicle-stimulating hormone; bLH: basic luteinizing hormone; bE[2]: basic estradiol; bT: basic testosterone; AMH: anti-Müllerian hormone Pre-treatment of serum samples Serum samples were thawed at 4 ℃. Then, 100 µL of serum was resuspended with 400 µL of 80% methanol using well vortex. After sonicating for 5 min in an ice-water bath, the serum samples were centrifuged at 15,000 ×g, 4℃ for 20 min. The supernatant was diluted with water to a final concentration containing 53% methanol. After centrifuging at 15,000 ×g, 4 ℃ for 20 min, 140 µL of the supernatant was collected for further analysis. Additionally, 2 µL of each sample was mixed for quality control (QC). Additionally, same reagent but without serum was defined as blank sample and tested. Ultra-high performance liquid chromatography-tandem mass spectrometry (UPLC-MS/MS) analysis UHPLC-MS/MS analysis was performed using a Vanquish UPLC system (Thermo Fisher Scientific) coupled with an Orbitrap Q ExactiveTM HF mass spectrometer (Thermo Fisher Scientific). Chromatographic separation was performed on a ThermoFisher Hypersil Gold C18 column (100 × 2.1 mm, 1.9 μm) at 40 ℃. Precisely 0.1% (v/v) formic acid (A) and methanol (B) were used as mobile phases for positive-ion-mode, and 5 mM ammonium acetate (pH = 9.0, A) and methanol (B) were used as mobile phases for negative-ion mode. The flow rate was 0.2 mL/min. After equilibration, 5 µL of each sample was injected, and the flowing gradient elution was as below: 2% B from 0 to 1.5 min, 2–85% B from 1.5 to 3 min, 85–100% B from 3 to 10 min, 100–2% B from 10 to 10.1 min, 2% B from 10.1 to 11 min, and 2% B from 11 to 12 min. Mass spectrometry analysis was performed in both positive- and negative-ion modes. The source ionization parameters were set as below: spray voltage = 3500 V; sheath gas flow rate = 35 psi; aux gas flow rate = 10 mL/min; capillary temperature = 320 ℃; aux gas heater temperature = 350 ℃. Data were acquired over the m/z range of 100–1500. To removing the background noise, the blank sample was injected before injecting serum samples. To confirm the stability and repeatability of the system, samples were injected randomly, and QC samples were repeatedly injected after injecting eight samples. Data collection and processing Raw data were processed using Compound Discoverer (Version 3.1, Thermo Fisher Scientific) to perform peak alignment, peak picking, metabolite identification, and quantification. Parameters for peak alignment and picking were set as below: retention time tolerance ± 0.2 min; actual mass tolerance ± 5 ppm; signal intensity tolerance ± 30%; signal to noise ratio ≥ 3 (when signal to noise ratio ≤ 3, peaks would be defined as background). The molecular weight was determined based on m/z, the molecular formula was predicted based on ppm and additive ions, then automatically matched with mzClould ([67]https://www.mzcloud.org/), mzVault (Version 2.3, local database) and Masslist (local database) for MS1 analysis [[68]25, [69]26]. Further, MS2 analysis were made by automatically matching the molecular ions, fragment ions and collision energy in mzClould and mzVault databases. Then identified metabolites were annotated by The Kyoto Encyclopedia of Genes and Genomes (KEGG, [70]https://www.genome.jp/kegg/pathway.html), The Human Metabolome Database (HMDB, [71]https://hmdb.ca/metabolites) and LIPID MAPS ([72]https://www.lipidmaps.org/). Compound classes were obtained from these databases. Normalized quantification results were obtained according to the QC-based method: raw metabolite quantification of each sample / (sum of raw metabolite quantification/sum of metabolite quantification of QC sample). Statistical analysis Data were statistically analyzed using SPSS software (version 26.0). Continuous variables with normal distributions were described as mean ± standard deviation (SD) and analyzed using one-way analysis of variance (ANOVA). For parametric ANOVA, Dunnett was used in multiple comparison test; For variables without normal distribution or multiple comparison test with unequal variances, Kruskal-Wallis H test was used. Spearman’s rank correlation test was performed for the correlation analysis. P < 0.05 was considered statistically significant. MetaX software ([73]http://metax.genomics.cn) was used to perform multivariate analysis, including principal component analysis (PCA) and orthogonal projections to latent structure discriminant analysis (OPLS-DA). Variable importance in the projection (VIP) of OPLS-DA was determined. Between groups, metabolites with variable importance in projection (VIP) > 1, fold change (FC) > 1.5 or < 0.667, and P < 0.05 were recognized as differential metabolites. Moreover, the validation of the OPLS-DA model was assessed by 7-fold cross-validation (the model was considered validated especially when R^2Y were close to 1) and a permutation test (the model was considered validated when R^2 > Q^2 and Q^2 < 0). The correlation between differential metabolites was calculated using the R language (Version 4.0). KEGG databases was used for metabolic pathway enrichment analysis. P < 0.05 was considered statistically significant. Results Clinical characteristics of participants Significant clinical characteristics, including age, body mass index (BMI), menstruation / pregnancy / metabolic disease histories, hormone levels and lifestyles, were collected and shown in Table [74]1. The differences between groups were mainly focus on the hormones. Compared to the control group, bFSH levels in the old, DOR, scPOI, and POI groups were significantly increased to various degrees (P < 0.001). Accordingly, the AMH levels in these groups significantly decreased to various degrees (P < 0.001). The basic luteinizing hormone (bLH) levels significantly increased in the old, DOR, and POI groups. However, a bFSH/bLH ratio > 2 was only observed in the DOR (P < 0.05), scPOI (P < 0.001), and POI groups (P < 0.001), which may represent a new difference between physiological and pathological ovarian aging. Additionally, decreased baseline estradiol (bE[2]) levels were observed in the old (P < 0.01) and POI groups (P < 0.001), and scPOI group showed a significant decrease in baseline testosterone (bT) levels (P < 0.05). In summary, these serum hormone levels indicate diminished ovarian function during physiological and pathological ovarian aging. Multivariate analysis of metabolites After relative standard deviation denoising, 448 and 325 metabolites were identified in the positive-ion and negative-ion modes, respectively. The unsupervised PCA of all groups showed a similar distribution of variations, especially in the scPOI and POI groups (Fig. [75]1B; Supplementary Fig. [76]1A). Unlike the other ovarian aging groups, variations in the DOR group were more similar to those in the control group. Further PCA of each ovarian aging group and the control group was performed, which showed distinct separation tendencies (Fig. [77]1C, F, I, L; Supplementary Fig. [78]1B, E, H, K). Fig. 1. [79]Fig. 1 [80]Open in a new tab Principal component analysis (PCA) score plots, orthogonal projections to latent structure discriminant analysis (OPLS-DA) score plots and corresponding validation plots in ESI^+ mode. (A) The workflow of the study. PCA score plots of all groups (B), control (n = 12) and old (n = 13) groups (C), control and DOR (n = 12) groups (F), control and scPOI (n = 14) groups (I), control and POI (n = 19) groups (L). OPLS-DA score plots of control and old groups (D), control and DOR groups (G), control and scPOI groups (J), control and POI groups (M). Permutation test of OPLS-DA model in control and old groups (E), control and DOR groups (H), control and scPOI groups (K), control and POI groups (N). Ctrl: control. DOR: diminished ovarian reserve. scPOI: subclinical premature ovarian insufficiency. POI: premature ovarian insufficiency Supervised OPLS-DA was applied to maximize group separation and identify discriminating metabolites, which showed marked improvements in group separation (Fig. [81]1D, G, J, M; Supplementary Fig. [82]1C, F, I, L; Supplementary Fig. [83]2B, E, H, K, N, Q; Supplementary Fig. [84]3B, E, H, K, N, Q). 7-fold cross-validation of OPLS-DA models showed R^2Y (0.926–0.986) and Q^2Y (0.38–0.826). Two hundred random permutation tests showed that R^2 > Q^2 and Q^2 intercepting the Y axis at -0.5~-1.0 (Fig. [85]1E, H, K, N; Supplementary Fig. [86]1D, G, J, M; Supplementary Fig. [87]2C, F, I, L, O, R; Supplementary Fig. [88]3C, F, I, L, O, R). These data indicated stable and not overfitting OPLS-DA models. The OPLS-DA results demonstrated significant metabolic differences between the participants with ovarian aging and healthy participants, the participants with physiological ovarian aging and pathological ovarian aging, and the participants with different pathological ovarian aging. Identification of significant differential metabolites in physiological and pathological ovarian aging Given the success of the OPLS-DA model in classifying the ovarian aging groups and the control group, significant differential metabolites from positive-ion and negative-ion modes were integrated and are shown in volcano plots. Compared to the control group, metabolites marked with bright red color were significantly upregulated in the ovarian aging groups, and metabolites marked with bright blue color were significantly downregulated in the ovarian aging groups. Overall, 20, 26, 18 and 21 differential metabolites were identified in the old vs. control group (Fig. [89]2A), DOR vs. control group (Fig. [90]2B), scPOI vs. control group (Fig. [91]2C), and POI vs. control group (Fig. [92]2D), respectively. Fig. 2. [93]Fig. 2 [94]Open in a new tab The identification of differential metabolites. Volcano plot shows the differential metabolites between old and control groups (A), DOR and control groups (B), scPOI and control groups (C), POI and control groups (D). Comparing to the control group, the up-regulated and down-regulated metabolites were marked in red and blue, respectively. (E) The heatmap analyzes hierarchical clustering of the top 10 differential metabolites in all groups The top10 differential metabolites (|log2FC| value-based) were yellow-marked and showed in Supplementary Table [95]1. We then clustered these significant differential metabolites into groups (Fig. [96]2E), menaquinone, neopterin and SM (d22:1/14:1) were clustered. Interestingly, these metabolites were significantly downregulated both in physiological and pathological ovarian aging groups. Alpha-benzylsuccinic acid and SM (d27:0/14:1) were clustered and significantly upregulated in the old group. In the pathological ovarian aging groups, SM (d17:0/25:1), 9-[(2-hydroxyethoxy)methyl]-1,9-dihydro-6 H-purin-6-one and 15-deoxy-Δ12,14-prostaglandin A1 were clustered and significantly upregulated in the DOR group, while taurocholic acid and SM (d14:0/20:2) were clustered and significantly downregulated. Tetranor-12(R)-HETE, (+/-)8(9)-DIHET and 3-methoxytyramine were clustered and significantly upregulated in the scPOI group but significantly downregulated in the POI group. Homogeneity and heterogeneity between physiological and pathological ovarian aging To further explore the homogeneity of physiological and pathological ovarian aging, significant differential metabolites identified in the old vs. control group, DOR vs. control group, scPOI vs. control group, and POI vs. control group were compared using a venn diagram (Fig. [97]3A). Five metabolites have been recognized to be common in physiological and pathological ovarian aging. Among these metabolites, neopterin (Fig. [98]3B, P < 0.05), menaquinone (Fig. [99]3C, P < 0.01), SM (d14:1/24:2) (Fig. [100]3E, P < 0.01) and SM (d14:0/21:1) (Fig. [101]3F, P < 0.05) were significantly downregulated both in the physiological and pathological ovarian aging groups. SM (d17:0/25:1) (Fig. [102]3D, P < 0.01) was significantly upregulated both in the physiological and pathological ovarian aging groups. Fig. 3. [103]Fig. 3 [104]Open in a new tab The identification of common aging metabolites between physiological and pathological ovarian aging, physiological ovarian aging-specific metabolites. (A) Venn plot shows the differential metabolites in all groups. And there are five common aging metabolites between physiological and pathological ovarian aging (B-F). (G-L) The physiological ovarian aging-specific metabolites. ^*P < 0.05, ^**P < 0.01, ^***P < 0.001 Among these common metabolites, only downregulated extent of SM (d14:0/21:1) both in physiological and pathological ovarian aging showed no difference. Directly comparing with the old group, neopterin (P < 0.05) and menaquinone (P < 0.01) in the POI group showed a larger extent of downregulation, while SM (d17:0/25:1) in the DOR group showed a higher extent of upregulation (P < 0.05). Additionally, downregulation of SM (d14:1/24:2) in all pathological ovarian aging groups was less than that in the physiological ovarian aging group (P < 0.05). In the pathological ovarian aging groups, upregulated extent of SM (d17:0/25:1) in the DOR group was higher than those in the scPOI (P < 0.01) and POI groups (P < 0.05). Differential metabolities specifically changed in the physiological ovarian aging group were further studied (Supplementary Table [105]2). Testosterone sulfate (Fig. [106]3G, P < 0.001), PC (16:0/19:2) (Fig. [107]3J, P < 0.01) and PC (18:2/19:2) (Fig. [108]3K, P < 0.05) were significantly downregulated. While alpha-benzylsuccinic acid (Fig. [109]3H, P < 0.05), C-8 ceramide-1-phosphate (Fig. [110]3I, P < 0.05) and PC (16:2e/20:0) (Fig. [111]3L, P < 0.05) were significantly upregulated. Among these metabolites, testosterone sulfate level in the old group was also lower that those in the DOR (P < 0.001) and POI group (P < 0.01). PC (16:0/19:2) in the old group was downregulated when comparing to the control group, but the level was still higher when comparing to the scPOI group (P < 0.05). In those three physiological ovarian aging specifically upregulated metabolites, alpha-benzylsuccinic acid level in the old group was also higher than those in the DOR group (P < 0.05) and the POI group (P < 0.01), C-8 ceramide-1-phosphate level in the old group was also higher than those in the DOR group (P < 0.001) and the scPOI group (P < 0.01), PC (16:2e/20:0) level in the old group was also higher than those in the scPOI group (P < 0.01) and the POI group (P < 0.001). The correlation between differential metabolites reflects synergistic or repulsive effects on functional regulation. Pearson correlation coefficient analysis was performed in the physiological ovarian aging differential metabolites (Supplementary Fig. [112]4A). Results showed that menaquinone, one of the common differential metabolites, was positively correlated with neopterin (P < 0.001). Testosterone sulfate, one of the physiological ovarian aging specific differential metabolites, was positively correlated with SM (d15:1/20:0) (P < 0.001), but negatively correlated with PC (16:2e/20:0) (P < 0.05). Homogeneity and heterogeneity between different stages of pathological ovarian aging Based on the venn diagram analysis (Figs. [113]3A and [114]4A), we further compared the homogeneity and heterogeneity between three pathological ovarian aging groups. The results showed that the PC (17:0/18:2) (Fig. [115]4B, P < 0.05), PC (18:2e/17:2) (Fig. [116]4C, P < 0.05) and SM (d22:1/14:1) (Fig. [117]4D, P < 0.05) were significantly downregulated both in the DOR and scPOI groups. And downregulated extents of these three metabolites both in the DOR group and scPOI group showed no difference. When directly comparing to the old group, PC (17:0/18:2) level in the POI group (P < 0.01), PC (18:2e/17:2) level and SM (d22:1/14:1) level in the DOR group were significantly lower (P < 0.05). Fig. 4. [118]Fig. 4 [119]Open in a new tab The identification of common aging metabolites in pathological ovarian aging. (A) Venn plot shows the common metabolites in three pathological ovarian aging groups. (B-F) There are five common aging metabolites in pathological ovarian aging. (G-I) The examples of DOR, scPOI or POI-specific aging metabolites. ^*P < 0.05, ^**P < 0.01, ^***P < 0.001 SM (d14:1/20:1) (Fig. [120]4E, P < 0.05) and 4-hydroxyretinoic acid (Fig. [121]4F, P < 0.05) were significantly downregulated both in the DOR and POI groups, and 4-hydroxyretinoic acid level in the POI group showed a higher downregulation extent (P < 0.001). Although downregulation extents of SM (d14:1/20:1) showed no difference between the DOR and POI groups, SM (d14:1/20:1) levels in these two groups were significantly lower than that in the scPOI group (P < 0.01). As for the pathological ovarian aging specific metabolites, 9-[(2-hydroxyethoxy)methyl]-1,9-dihydro-6 H-purin-6-one was specifically upregulated in the DOR group (P < 0.05), 3-(3-methoxyphenyl)propionic acid was specifically downregulated in the scPOI group (P < 0.05), and vitamin A was specifically downregulated in the POI group (P < 0.05). When comparing to the old group, 3-(3-methoxyphenyl)propionic acid levels in pathological ovarian aging groups were significantly lower (P < 0.01). Pearson correlation coefficient analysis (Supplementary Fig. [122]4B) of ten common differential metabolites of pathological ovarian aging showed that menaquinone, one of the common differential metabolites, was positively correlated with 4-hydroxyretinoic acid (P < 0.05). SM (d22:1/14:1), one of the common differential metabolites of the DOR and scPOI groups, was positively correlated with menaquinone (P < 0.001) but negatively correlated with SM (d17:0/25:1) (P < 0.01). Metabolic pathways associated with physiological and pathological ovarian aging KEGG pathway analysis was performed to investigate enriched metabolic pathways associated with physiological and pathological ovarian aging. Significant differential metabolites in both physiological and pathological ovarian aging were mainly enriched in folate biosynthesis, ubiquinone and other terpenoid-quinone biosynthesis (Fig. [123]5A-D. P < 0.05). Other metabolic pathways were also enriched in pathological ovarian aging groups. Significant differential metabolites in the scPOI group were enriched in dopaminergic synapses pathway(P < 0.05). Significant differential metabolites in the POI group were also mainly enriched in vitamin digestion and absorption and retinol metabolism (P < 0.05). Fig. 5. [124]Fig. 5 [125]Open in a new tab Metabolitic pathways in physiological and pathological ovarian aging. Bubble plots show the KEGG pathways enriched with differential metabolites between the old and control groups (A), the DOR and control groups (B), the scPOI and control groups (C), the POI and control groups (D) Predictive value of typical metabolites on physiological and pathological ovarian aging ROC curve analysis was used to assess the predictive value of differential metabolites for ovarian aging. In physiological ovarian aging, testosterone sulfate showed high predictive value with an area under the curve (AUC) of 0.929 (95% CI: 0.82–1, Youden’s index: 0.84) (Fig. [126]6A), indicating that decreased serum testosterone sulfate levels could predict physiological ovarian aging. In pathological ovarian aging, increased levels of SM (d14:0/28:1), PC (18:0e/4:0) and 4-hydroxyretinoic acid showed high sensitivity and specificity for DOR, scPOI, and POI (Fig. [127]6B-D), respectively. The AUC of these markers were 0.847 (95% CI: 0.563-1, Youden’s index: 0.763), 0.923 (95% CI: 0.81-1, Youden’s index: 0.762), and 0.921 (95% CI: 0.822-1, Youden’s index: 0.789), respectively. These new markers showed nearly equal predictive values compared to accepted markers bFSH and AMH (AUC = 1). Fig. 6. [128]Fig. 6 [129]Open in a new tab ROC analysis of metabolites in physiological and pathological ovarian aging. ROC curves show the clinical significance of testosterone sulfate in the old group (A), SM (d14:0/28:1) in the DOR group (B), PC (18:0e/4:0) in the scPOI group (C) and 4-hydroxyretinoic acid in the POI group (D) The predictive value of five common metabolites in pathological ovarian aging were also analyzed. Result showed that PC (17:0/18:2), SM (d22:1/14:1) and 4-hydroxyretinoic acid showed highest predictive values for POI with an AUC of 0.851, 0.921 and 0.921 (Supplementary Fig. [130]5A, C, E). PC (18:2e/17:2) and SM (d14:1/20:0) showed highest predictive values for DOR with an AUC of 0.799 and 0.743 (Supplementary Fig. [131]5B, D). Additionally, combining these five metabolites also showed high predictive values for pathological ovarian aging with the AUC>0.900 (Supplementary Fig. [132]5F). Correlation analysis of common differential metabolites and significant clinical parameters To further investigate the credibility of significant differential metabolites in predicting ovarian aging, Spearman’s correlation analyses were performed between five common metabolites of physiological and pathological ovarian aging, five common metabolites of pathological ovarian aging, and diagnostic standards (age, bFSH, and AMH, Table [133]2). SM (d14:1/24:2) (P < 0.01), 4-hydroxyretinoic acid ((P < 0.01) and SM (d14:0/21:1) (P < 0.05) negatively correlated with age. Table 2. Correlation analysis between the concentration of metabolites with age, basic FSH (bFSH) and AMH level Metabolites Age bFSH AMH Neopterin 0.1 -0.30^* 0.22 Menaquinone 0.05 -0.30^* 0.28^* SM (d17:0/25:1) 0.15 0.21 -0.18 SM (d14:1/24:2) -0.34^** -0.29^* 0.35^** SM (d14:0/21:1) -0.24^* -0.21 0.26^* SM (d22:1/14:1) -0.23 -0.09 0.13 PC (18:2e/16:0) 0.04 0.11 -0.21 PC (17:0/18:2) 0.02 -0.18 0.12 4-Hydroxyretinoic Acid -0.33^** -0.46^*** 0.58^*** SM (d14:1/20:1) -0.20 -0.30^* 0.23 [134]Open in a new tab ^*P < 0.05, ^**P < 0.01, ^***P < 0.001. - were found to be negatively correlated Neopterin (P < 0.05), menaquinone (P < 0.05), SM (d14:1/24:2) (P < 0.05), 4-Hydroxyretinoic Acid (P < 0.001) and SM (d14:1/20:1) (P < 0.05) were negatively correlated with bFSH. Menaquinone (P < 0.05), SM (d14:1/24:2) (P < 0.01), SM (d14:0/21:1) (P < 0.05) and 4-hydroxyretinoic acid (P < 0.001) were positively correlated with AMH. Combining the correlation results with diagnostic standards, SM (d14:1/24:2) showed strong correlations with both physiological and pathological ovarian aging. Additionally, 4-hydroxyretinoic acid levels strongly correlated with pathological ovarian aging. Discussion Decreased fertility rates, various complications, and poor quality of life related to ovarian aging are critical issues. Discovering reliable biomarkers to assess ovarian aging and developing safe drugs to prevent and cure ovarian aging have become significant challenges for promoting healthy aging [[135]27]. In this study, we investigated the metabolic characteristics of both physiological and pathological ovarian aging and found five common differential metabolites of ovarian aging and five specific differential metabolites of pathological ovarian aging. Metabolic pathways, such as folate biosynthesis, ubiquinone and other terpenoid-quinone biosynthesis pathways, are mainly involved both in physiological and pathological ovarian aging. While dopaminergic synapses, vitamin digestion and absorption and retinol metabolism, were specifically associated with pathological ovarian aging. We also found that several metabolites, including testosterone sulfate, SM (d14:0/28:1), PC (18:0e/4:0) and 4-hydroxyretinoic acid showed a close correlation with ovarian aging and could be new biomarkers in diagnosing of physiological ovarian aging, DOR, scPOI and POI, respectively. The homogeneity between physiological and pathological ovarian aging could reflects the essence of ovarian aging. In this study, the majority of common differential metabolites were SM and PC. They are not only fundamental structural elements of cell membranes but also key molecules in regulating energy metabolism and signal transduction. In the ovarian aging groups, SM (d17:0/25:1) was significantly increased, whereas SM (d14:1/24:2) and SM (d14:0/21:1) were significantly decreased. In particular, SM (d14:1/24:2) was significantly and negatively correlated with ovarian aging. The sphingosine base is a common structure in sphingolipids. Sphingosine with a longer chain appears to be positively correlated with ovarian aging. Longer chains would cause lower water solubility and decreased cell membrane mobility [[136]28, [137]29]. In contrast, sphingolipids, including SM, accumulate linearly with age, and although these accumulations are necessary for early development, they may promote age-related diseases by increasing reactive oxygen species or inhibiting Bcl-2 [[138]30, [139]31]. Another lipid, menaquinone, was significantly decreased in the ovarian aging group. Menaquinones (also known as vitamin K2), enriched in ubiquinone and other terpenoid-quinone biosynthesis pathway, are a family of redox-active small molecules that are crucial for energy production in bacteria and the post-translational γ-carboxylation of some proteins [[140]32]. Menaquinone deficiency not only impairs blood coagulation but also leads to age-related diseases such as Alzheimer’s disease and osteoporosis. Supplementation with menaquinone can reverse structural and cognitive deterioration by regulating the nucleotide-binding oligomerization domain-like receptor protein 3 (NLRP3)/caspase-1/Nrf-2 signaling pathway [[141]33], promote bone mineralization, and indirectly increase bone strength [[142]34]. However, the effect of menaquinone on ovarian aging remains unclear. In addition to lipids, lower neopterin levels were observed in the ovarian aging groups. Neopterin, enriched in folate biosynthesis pathway, is a metabolite of guanosine-5’-triphosphate (GTP) and is recognized as a marker of macrophage activation. Contrary to our study, previous studies [[143]35, [144]36] showed that neopterin level was positively associated with age. During aging, the number of cells, including macrophages, decreases. However, age-related chronic inflammation occurs as immunity decreases. A previous study [[145]37] found that old mice showed an increased frequency of C-X-C Motif Chemokine Receptor 3 (CX3CR1) expressing macrophage and a reduction in Ly6C^+ macrophages, which suggested that macrophages would adopt a more immunosuppressive phenotype. Additionally, the homogeneity of physiological and pathological ovarian aging is reflected in folate biosynthesis. Neopterin is the end product of pterin metabolism, and folate is required to form the starting molecule for pterin metabolism. Folate showed a strong positive association with neopterin concentration in older people with depression [[146]38]. Folate is essential for oocyte development, as it is taken up by reduced folate carrier (RFC1) and folate receptor 1 (FOLR1), both of which are highly expressed in cumulus-oocyte complexes (COC) and oocytes. Folate deficiency can impair the activation of the TGFβ1/Smad signaling pathway, which in turn disrupts autophagy and reduces ROS scavenging ability, contributing to ovarian dysfunction [[147]39–[148]40]. Additionally, folate acts as an important methyl donor; folate deficiency can disrupt DNA methylation patterns, leading to gene expression alterations, base substitutions, DNA breaks, and gene deletions within follicles [[149]41]. Maintaining high levels of DNA methylation is essential for promoting oocyte development [[150]42]. In summary, these common differential metabolites suggest that the homogeneity of physiological and pathological ovarian aging may related to oxidative stress, immune disorders, and inflammation. The onset of ovarian aging is insidious, and the diagnostic criteria in the early stages are not sufficiently specific. Therefore, it is difficult to diagnose it physiologically or pathologically in the early stages. Testosterone sulfate, one of the specific differential metabolites of physiological ovarian aging, showed high diagnostic efficacy and accuracy, with an AUC of 0.929 and a Youden’s index of 0.84. Testosterone sulfate is the main metabolite of testosterone, and decreased testosterone levels are associated with poor ovarian function [[151]43]. Additionally, testosterone sulfate was negatively correlated with menaquinone and neopterin levels. Currently, there are no specific biomarkers for diagnosing physiological ovarian aging, apart from age itself. Our findings suggest that testosterone sulfate could serve as a potential diagnostic marker for early-stage physiological ovarian aging, offering clinicians a more precise tool for early detection and intervention. The DOR, scPOI, and POI represent different stages of pathological ovarian aging. DOR and scPOI overlap in diagnostic criteria. Although the DOR patients selected in this study showed different bFSH levels than scPOI patients, these two groups shared three common differential lipid metabolites. All showed significantly decreased PC (18:2e/17:2), PC (17:0/18:2) and SM (d22:1/14:1) levels. PC, classified as a glycerophospholipid, is another lipid found in common differential metabolites. Previous studies [[152]44, [153]45] showed that PC could promote inflammatory cytokine expression and polarizes macrophage activation toward the M1 phenotype. In recent years, more and more studies [[154]46, [155]47] have shown that chronic inflammatory aging is closely related to decreased ovarian function. The DOR and scPOI groups showed significant differences compared to the POI group. SM (14:1/20:1) and 4-hydroxyretinoic acid levels significantly decreased in both the DOR and POI groups. Additionally, 4-hydroxyretinoic acid demonstrated high diagnostic efficacy in diagnosing POI, with an AUC of 0.921 and a Youden’s index of 0.789. These values exceed those reported for previously studied plasma markers, which typically show AUCs ranging from 0.804 to 0.901 [[156]22]. This highlights the superior diagnostic potential of 4-hydroxyretinoic acid in distinguishing POI from other forms of ovarian aging, further enhancing the accuracy of clinical assessments. 4-Hydroxyretinoic acid is an NADPH-dependent hydroxylated metabolite of retinoic acid. Retinoic acid, a derivative of vitamin A, plays an important role in germ cell development. Retinoic acid is critical for initiating meiosis in germ cells of fetal ovaries [[157]48]. It could also promote folliculogenesis directly or through regulating gonadotropin-releasing hormone (GnRH) production and release [[158]49, [159]50]. Additionally, it can regulate ovarian steroidogenesis by influencing the expression or activity of steroidogenic enzymes, such as steroidogenic acute regulatory protein (STAR), cytochrome P450 family 17 subfamily A member 1 (CYP17A1), and cytochrome P450 family 11 subfamily A member 1 (CYP11A1) [[160]51]. These common differential metabolites represented the homogeneity of pathological ovarian aging. Differences were also observed between pathological ovarian aging groups. 9-[(2-hydroxyethoxy)methyl]-1,9-dihydro-6 H-purin-6-one and taurocholic acid were representative metabolites in the DOR group. Taurocholic acid is classified to conjugated bile acid and could effectively relieve aging. Cassandra et al. found that taurocholic acid could protect against both degenerative and neovascular age-related macular degeneration through inhibiting vascular endothelial growth factor (VEGF)-induced choroidal endothelial cell migration and tube formation [[161]52]. Tauroursodeoxycholic acid, derivative metabolites of taurocholic acid, could relieve ovarian aging through rebuilding the damaged endoplasmic reticulum in ovarian surface epithelium [[162]53]. Upregulated cholesteryl sulfate and downregulated choline were representative metabolites in the scPOI group. Cholesteryl sulfate is one of the most important known sterol sulfates in human plasma. It serves as not only a part of many biological membrances but also a substrate for the synthesis of sulfonated adrenal steroids such as pregnenolone. In vitro study showed that cholesteryl sulfate could modulate brain energy and its neuroprotective effects may be related to activating AKT signaling [[163]54]. Another study found that cholesteryl sulfate levels in women’s hair would decrease with age [[164]55], which is different from the pathological ovarian aging. Excessive levels can also inhibit ovarian steroidogenesis by regulating the expression of mitochondrial P450 side chain cleavage (P450scc) and STAR [[165]56]. Additionally, a targeted metabolomics study showed that polyunsaturated choline plasmalogens in the follicle fliuid were significantly downregulated in women with DOR [[166]57]. Vitamin A was specifically downregulated in the POI group. Vitamin A, also named retinol, is the precursor of retinoic acid. Retinol metabolism has received considerable attention in the context of ovarian aging. In addition to promoting the meiosis of germ cells, folliculogenesis, and steroidogenesis, retinoic acid can also stimulate the differentiation of oogonial stem cells and oogenesis [[167]58]. The importance of retinol metabolism in egg regeneration inspired us to investigate treatments for ovarian aging. Additionally, the metabolic pathways involved in vitamin digestion and absorption and retinol metabolism were significant. It has been reported that vitamins C and E have anti-oxidative stress effects and can relieve mitochondrial function, which could restore the follicle reserve in ovarian aging mice [[168]59, [169]60]. Conclusion This study identified that both physiological and pathological ovarian aging share metabolic similarities involving disruptions in lipid, folate, and ubiquinone pathways. However, distinct differences were observed, with physiological and pathological aging displaying unique alterations in lipid, vitamin, and retinol metabolism. These findings not only contribute to better diagnostic capabilities but also open avenues for therapeutic advancements, as early identification of ovarian aging can inform personalized treatment strategies aimed at preserving ovarian function. Limitation Although untargeted metabolomics enables the detection of a broader spectrum of metabolites compared to targeted metabolomics, its accuracy may be compromised due to the inherent complexity of the technique. Serum markers, while useful for systemic metabolomic profiling, may not fully capture ovarian-specific metabolic changes, as they can be influenced by broader systemic factors. Follicular fluid, in contrast, offers a potentially more accurate reflection of ovarian metabolism. However, obtaining follicular fluid in the early stages of ovarian aging presents significant challenges, making serum analysis a more practical alternative for this study. Additionally, the relatively small sample size of this study may limit the statistical power and generalizability of the findings. Given the progressive and potentially systemic nature of ovarian aging, larger, long-term studies are necessary to enhance the robustness of diagnostic capabilities and to better understand the broader impacts and complications associated with ovarian aging. Electronic supplementary material Below is the link to the electronic supplementary material. [170]Supplementary Material 1^ (1.8MB, jpg) [171]Supplementary Material 2^ (2.2MB, jpg) [172]Supplementary Material 3^ (2.2MB, jpg) [173]Supplementary Material 4^ (1MB, jpg) [174]Supplementary Material 5^ (1.2MB, jpg) [175]Supplementary Material 6^ (13.7KB, xlsx) [176]Supplementary Material 7^ (11.2KB, xlsx) Acknowledgements