The PubChem database is a publicly accessible repository that offers an extensive collection of chemical information at no cost, thereby providing a comprehensive resource for drug discovery endeavors. (Kim, 2016). Hence, the PubChem database was employed to extract pertinent details regarding AC, aiding in the discovery of potential targets. SwissTargetPrediction, a web-based tool that utilizes both 2D and 3D similarity metrics of known ligands, effectively forecasts target molecules with biological activity (Gfeller et al., 2014). PharmMapper proves to be a valuable tool for the discovery of promising drug targets. It is an online tool that utilizes reverse pharmacophore matching and allows for compound querying against an internal database containing a model of pharmacophores (Wang et al., 2017). Potential targets were predicted independently with these two web servers: SwissTargetPrediction (2-D/3-D ligand similarity, probability ≥0.7) and PharmMapper (reverse pharmacophore matching, Z-score ≥0.9). Only proteins identified by both tools were retained, and their gene symbols were standardised with UniProt (UniProt, 2023).

The Genecards database functions as an extensive and reliable collection of annotated data related to genes. (Safran et al., 2010). It excels in its capacity to comprehensively calculate the relevance score of diseases and genes, enabling it to identify gene-associated diseases with precision. Meanwhile, OMIM (Online Mendelian Inheritance in Man) functions as a highly dependable and current research tool, meticulously elucidating the intricate relationships between human genes and phenotypes (Amberger et al., 2015). Therefore, DOR-associated genes were extracted from OMIM (curated entries) and GeneCards (relevance score ≥10) by searching the term “diminished ovarian reserve”. The final AC-versus-DOR target list was generated as the intersection of the two drug-target and two disease-target sets with Venny 2.1, ensuring that only well-supported candidates were analysed further.

To explore the association between overlapping genes, we utilized the STRING database and specified “Homo sapiens” as the species to import the predicted therapeutic targets (Szklarczyk et al., 2021). A confidence score threshold of ≥0.7 was applied to obtain a Protein-Protein Interaction (PPI) network diagram, which was then exported in TSV format and subsequently refined using Cytoscape 3.10.1 for further analysis (Liang et al., 2024). The Cytoscape software, an application for visualizing and analyzing network biology (Huang, 2023), was utilized to calculate node parameters and visualize molecular connections in the network diagram. To isolate significant targets prior hub gene identification, calculate the significant values by using the cytoNCA plugin to assess degree centrality, between centrality and closeness centrality. To investigate the associations among target genes, we utilized the cytoHubba plugin to assess node degree centrality. Afterwards, we chose the top 20 nodes and arranged them in the inner circle for visualization purposes. Lastly, we employed the MCC algorithm to identify and screen the top 20 genes (Chin et al., 2014).

The therapeutic targets identified for treatment of DOR in AC were examined using Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis and Gene Ontology (GO) enrichment analysis. Biological processes (BP), cellular components (CC), and molecular functions (MF) terms were retrieved with the Metascape database (adjusted P < 0.01) (Zhou et al., 2019). To ensure reliability, each target list was simultaneously uploaded to DAVID 6.8 and STRING (Szklarczyk et al., 2021; Huang da et al., 2009), only terms recovered by at least two independent platforms and exhibiting identical leading genes were retained. Enrichment results were visualised with bubble plots generated through the bioinformatics web server (https://www.bioinformatics.com.cn) (Liang et al., 2024; Gene Ontology, 2015).

The STRING database was utilized to import all targets that were enriched in the three pathways related to ‘production of female gametes, response to hormones, and sex differentiation’. Subsequently, these targets were imported into Cytoscape 3.10.1 software for network analysis using the MCC algorithm provided by the cytoHubba module. The top 20 key genes were selected again. After merging and eliminating duplicate entries from the two sets of 20 primary objectives acquired, a final collection of 22 crucial targets was obtained for subsequent examination.

The 3-dimensional configuration of the AC molecule was acquired from the PubChem database (Kim et al., 2016) and transformed into Mol2 format utilizing Open Babel program (O'Boyle et al., 2011). The template structures of the target proteins were acquired from the UniProt database (UniProt, 2023) and RCSB Protein Data Bank (Chen et al., 2023; Berman et al., 2000). To prepare the target proteins for further analysis, water molecules and original ligands were removed using Pymol before importing them into AutoDock Tools 1.5.7. Subsequently, hydrogenation, charge calculation, and non-polar hydrogen combination were performed. After determining the Grid Box size and genetic algorithm parameters, molecular docking was performed using AutoDock Vina through CMD command characters. Finally, Pymol was utilized to visualize the results (Huang, 2023).

The KGN cells used in this study were obtained from Procell Life Science Technology Co., Ltd. for in vitro research purposes. The KGN cells were cultured in a culture medium composed of DMEM/F12, which was supplemented with 10% FBS (fetal bovine serum). The culture conditions were maintained at a temperature of 37 °C in an environment containing 5% CO₂, and the culture medium was renewed every 2 days. Upon reaching approximately 80% confluency and displaying favorable condition, the cells were divided into three groups (n = 6 wells per group, three independent passages on different days): The normal control (NC group): complete medium only; CTX model (CTX group): 10 μg/mL cyclophosphamide (Sigma, C3250000) dissolved in serum-free DMEM; α-Cyperone (AC group): pre-treated with varying concentrations α-Cyperone (MedChemExpress, HY-N0710) for 6 h, followed by 10 μg/mL CTX without wash-out. Cyclophosphamide (CTX; C3250000) was purchased from Sigma (United States) and dissolved in DMEM as per manufacturer’s instructions. α-Cyperone was purchased from MedChemExpress (HY-N0710, Monmouth Junction, NJ, United States) and dissolved in Methyl sulfoxide (DMSO; Cat# D8371, Solarbio Life Sciences, Beijing, China), and in the final cultivation system, the content of DMSO should not exceed 0.01%.

The KGN cells were cultured in a 96-well plate with a cell concentration of 1 × 10⁴ cells/mL. Each well was supplemented with 100 µL of cell suspension. A minimum of three replicate wells were included in each experimental group. The plate was then placed in an incubator set at 37 °C and enriched with 5% CO₂ until the cells reached approximately 80% confluence. Subsequently, the culture medium of the AC groups was aspirated and gradient concentrations of AC were introduced into the wells, followed by an incubation period of 6 hours. CTX (300 μM) was subsequently added to each well in both the CTX and AC groups. The cells were incubated for an extra 24 h in the culture medium. Next, 10 µL of CCK-8 enhancement solution (Meilunbio, China) was added to each well and allowed to incubate for 1 hour before measuring optical density (OD) at 450 nm using a microplate reader according to standard protocols for colorimetric assays.

Using the AMH assay kit (Nanjing BYabscience technology Co.,Ltd., BY-EH111084), the optical density (OD) at 450 nm was read and the concentration of AMH in the supernatant of KGN cells was determined by comparing with the standard curve.

Intracellular reactive oxygen species (ROS) were quantified using the 2,7-dichlorodihydrofluorescein diacetate (DCFDA) fluorescent probe (Beyotime Biotech, Nanjing, China). A cell suspension containing 2 × 10⁵ cells was exposed to a 50 mM DCFDA solution in pre-warmed PBS for 30 min at 37 °C, following which the fluorescent signal was captured under excitation at 488 nm and emission at 525 nm.

The JC-1 probe was employed to detect mitochondrial depolarization of KGN cells, following the manufacturer’s protocol. Briefly, after the indicated treatments, the cells cultured in 6-well plates were incubated with 5 μg/mL JC-1 for 20 min at 37 °C. The staining solution was removed and cells were washed with JC-1 staining buffer twice. The distribution and fluorescence intensity of JC-1 monomers and aggregates were visualized by fluorescence microscopy, and the images were assessed using the ImageJ software. Mitochondrial depolarization was evaluated by the decrease in the red/green fluorescence intensity ratio.

The KGN cells in the NC, CTX, and AC groups were subjected to total RNA extraction using an RNA extraction kit (#G3640-50T, Servicebio). The spectrophotometric analysis was conducted to evaluate the concentration and purity of the extracted RNA, measuring absorbance at 260 nm and 280 nm. In order to ensure optimal quality, RNA samples with a D(λ)260/D(λ)280 ratio falling within the range of 1.8–2.0 were stored at −80 °C for future utilization. Subsequently, cDNA synthesis was performed using FastKing gDNA Dispelling RT SuperMix (Tiangen Biotech; Beijing; China), followed by PCR amplification on a lightcycler480II Real-time PCR instrument with SYBR Green PCR master mix (#G3326-05, Servicebio). Meanwhile, the β-actin gene was concurrently assessed as an internal reference in conjunction with the samples. The 2^−ΔΔCT method was employed to determine the relative levels of gene expression. All experiments were performed in triplicate parallel assays. The primers utilized in this investigation were custom-designed and synthesized by TsingKe (Wuhan, China), and their specific sequences can be found in Table 1.

This part of the experiment and data analysis were performed by APT Biotechnology (Shanghai, China).

Data are expressed as mean ± SEM of n = 18 replicates per condition (six wells per group in each of three independent experiments). Normality was confirmed for every data set by Shapiro–Wilk test and homoscedasticity by Brown–Forsythe (≥3 groups) or F-test (2 groups); all P > 0.05. Consequently, group means were compared with two-tailed unpaired t-test (two groups) or one-way ANOVA followed by Tukey’s HSD (≥3 groups) in GraphPad Prism nine and R 4.2.0. If either assumption was violated (P ≤ 0.05), Welch’s correction or Kruskal–Wallis test was substituted. Statistical significance was accepted at α = 0.05 and denoted *p < 0.05, **p < 0.01, ***p < 0.001.

The workflow for screening and identifying key targets for AC treatment of DOR is shown in Figure 1. The 2D and 3D configurations, along with associated details of AC, as depicted in Figure 2A, were acquired from the PubChem database (PubChem CID:89528182). Pymol was utilized to generate visual representations of the 3D structure. In order to investigate the potential mechanisms underlying the improvement of DOR by AC, we conducted a search for relevant targets of both AC and DOR, followed by standardizing the obtained gene names using information from the UniProt database. In SwissTargetPrediction and Pharmmapper databases, a total of 198 and 267 AC drug targets were identified respectively. After eliminating duplicates, we selected 466 unique drug targets for further analysis. In Genecards and OMIM databases, there were 1,356 and 202 targets associated with DOR respectively. Following the removal of redundant targets, we obtained a total of 1529 DOR-related targets. Given the scarcity of databases dedicated to identifying targets for both AC and DOR, we constructed separate interaction networks linking AC- and DOR-related targets and assigned distinct color codes to the nodes (targets) retrieved from specific websites (Supplementary Figure S2).

We analyzed the drug targets for AC and compared them with DOR targets to identify potential therapeutic targets for treating DOR with AC. Our analysis resulted in the identification of a total of 257 potential targets (Figure 2B). The protein interaction network of the remaining genes was established by excluding unrelated genes from the STRING database and enhanced using Cytoscape software. The left side of Figure 2C illustrates the construction of the “AC target-DOR target” network. Subsequently, we applied the MCC algorithm to identify the top 20 genes, as depicted on the right side of Figure 2C, which included AKT1, ESR1, CTNNB1, IGF1R, ERBB2, HSP90AA1, among others. The detailed data regarding the top 20 genes can be found in Table 2.

The Metascape database was utilized to conduct a comprehensive analysis on the targets, revealing significant enrichment in GO terms and KEGG pathways (Figures 3A,B). The findings indicate that AC is essential in multiple biological functions such as the production of female reproductive cells, hormonal responses, sexual differentiation, regulation of hormone levels, and development of glands. Molecular functions include nuclear receptor activity, ATP-dependent activity, acting on DNA, transcription factor binding and monocarboxylic acid binding.

Since DOR patients usually have abnormal sex hormone levels, and the results of GO enrichment suggest that the targets of AC action on DOR are closely related to hormone regulation, we focused on the targets enriched in the biological pathway of “female gamete production, responses to hormone, sex differentiation”. We uploaded all targets enriched in the aforementioned three pathways to the STRING database, followed by their importation into Cytoscape 3.10.1 software, and performed a network analysis using the MCC algorithm of the cytoHubba module, again selecting the top 20 key genes (Figure 3C). After consolidating and eliminating duplicate entries from the initial pool of 20 crucial targets identified in two separate ways, a refined set of 22 key targets was derived for subsequent investigation. Table 2 contains comprehensive data on the 22 primary objectives mentioned.

The molecular docking analysis was conducted to investigate the interactions between AC and a set of 22 specific genes. Employing AutoDock software, all the docking results were generated and displayed low binding energies (Delta G), indicating strong affinities between the compound and the targets (Figure 4). The molecular docking results of AC with target proteins are shown in Table 2. Notably, all target proteins exhibited robust binding to AC with respective binding energy all <0 kcal/mol, which indicated that AC and each target can bind spontaneously, signifying their significant roles in the molecular mechanism of AC treatment on DOR. The visualizations of the lowest binding energies between each target and AC were created using Pymol. Furthermore, we also employed visualizations to depict the surface configurations of the target proteins in the molecular docking outcomes. Notably, it is evident that all AC molecules are situated within the active pockets of their respective proteins (Supplementary Figure S1).

The KGN cells were exposed to different levels of CTX or AC for 24 h, after which the cell viability was evaluated using the CCK-8 assay (Figures 5A–B). The findings indicated that the IC50 of CTX on KGN was approximately 500μM, and AC concentrations below 80 μM did not significantly affect KGN cell viability, suggesting that AC at concentrations below 80 μM is non-cytotoxic to KGN cells, and 300μM CTX was selected for subsequent experiments. Following this, KGN cells were treated with gradient concentrations of AC for 6 h and then induced with CTX (300 μM) for 24 h to establish DOR models (Figure 5C). The cell viability was assessed using the CCK-8 assay, which indicated a significant reduction in cell viability observed in the CTX group compared to the NC group, while the viability of KGN cells significantly improved upon treatment with different concentrations of AC. Notably, even at the lowest effective concentration of AC, 40μM, the AC group showed a significant recovery effect similar to that of the NC group. Therefore, a concentration of 40 μM AC was opted for further investigation. We examined the amount of AMH in the supernatants of the three groups of cells, NC, CTX, and AC, and the results showed that the secretion of AMH was significantly reduced in the CTX group compared with the NC group, while it was significantly increased in the 40 μM AC treatment group (Figure 5D). In addition, we examined the effects of AC on the ROS level and mitochondrial membrane potential (MMP) in CTX-treated KGN cells (Figures 5E–H). By DCFH staining, ROS levels were found to be raised in CTX-treated KGN cells. Treatment with AC for 24 h significantly reduced ROS level. Decreased mitochondrial membrane potential, which can be reflected by the decreased fluorescence intensity ratio of JC-1 monomers to aggregates, is an early indicator of apoptosis (Smiley et al., 1991). In normal KGN cells, the JC-1 probe mainly aggregated in the mitochondrial matrix in the form of aggregates and emitted red fluorescence with little green fluorescence. By contrast, treatment with CTX for 24 h significantly reduced red fluorescence and increased green fluorescence, suggesting the inhibited polymerization of JC-1 and mitochondrial membrane potential. Whereas, AC treatment moderated the mitochondrial depolarization induced by CTX. These proves the successful establishment of DOR cell model and the recovery of DOR granule cell function by AC.

We have identified the top 12 genes with the strongest binding affinity based on our molecular docking findings. To assess the mRNA statement levels of the target proteins, we utilized RT-qPCR analysis in both the NC and CTX groups. The outcomes obtained from our experiments (Figure 6) demonstrated a significant increase in mRNA expression levels of MAPK14, ERBB2, CDH1, and ESR1 in the CTX group compared to the NC group. On the contrary, a significant reduction in the expression levels of MAP2K1, SRC, AKT1, ESR2, CYP19A1 and MAPK8 was observed in the CTX group compared to the NC group. AC treatment at 40 μM completely reversed these abreeant expression patterns for MAP2K1, AKT1, ESR2, ERBB2, CDH1, CYP19A1, ESR1, and MAPK8, restoring transcript levels close to—or even beyond—those of healthy controls. These directional changes imply that AC re-balances oestrogen synthesis and survival signalling within granulosa cells, providing a plausible transcriptional basis for its observed improvement in ovarian function.

In vitro, limited proteolysis–mass spectrometry (LiP–MS) has provided a way to observe changes in both protein abundances and structures on a proteome-wide scale. We extracted whole-cell protein samples from CTX-induced KGN cells, divided the samples into two aliquots, and one sample was incubated with AC for 24 h. Two samples were treated with proteinase K (PK), and standard sample preparation was performed on two duplicate samples. All peptides were subjected to liquid chromatography-tandem MS (LC-MS/MS) in data independent acquisition (DIA) mode for DIA analysis (Figure 7A, Created with BioGDP.com (Jiang et al., 2025)).

In order to show the significant difference of peptides between the two groups, the volcano plot of peptides in the two groups was plotted according to two factors: Fold change and P value (T-test). The significantly downregulated peptides were marked in blue (FC < 0.6 7 and p < 0.05). Significantly upregulated peptides are marked in red (FC > 1.5 and p < 0.05) (Figure 7B). The asymmetric cut-offs were chosen to account for the higher technical variability observed in low-abundance peptides. Hierarchical Cluster was used to group and classify the differentially expressed peptides of the two groups, and the Heatmap was displayed. Based on the similarity basis, in the clustering grouping results, the data pattern similarity within the group is generally high, while the data pattern similarity between the groups is low, so the groups can be effectively distinguished. According to the screening criteria of fold change >1.5 times and P value < 0.05 (T-test or other), the significantly differentially expressed peptides could effectively separate the comparison groups (Figure 7C). In order to reveal the overall metabolic pathway enrichment characteristics of proteins corresponding to all differentially abundant peptides, proteins corresponding to all differentially abundant peptides were compared with all proteins of reference species (or all proteins identified experimentally) by KEGG annotation results (Figure 7D). The results showed that, the proteins corresponding to the differentially abundant peptides between the two groups were significantly enriched in Regulation of actin cytoskeleton, Focal adhesion, Proteoglycans in cancer, and Cytoskeleton in muscle cells and other pathways.

Based on the principle of limited enzyme digestion method to find small molecule binding targets, if the abundance of the long peptide of a sequence can be detected at the same time, and the abundance of the corresponding short peptide can be detected, or the short peptide on the corresponding sequence cannot be detected, it means that the sequence may be closed due to the binding of drugs, and the sequence is relatively likely to be the target sequence. By corresponding all detected peptides to the complete sequence of the protein, the situation of peptides with differential abundance on the protein and their position on the sequence can be visualized. We found that among the top three proteins with the highest binding energies in the molecular docking experiment (MAP2K1, MAPK14 and GSK3B, with PG. Cscore (50, 47 and 25, respectively) far above the background threshold (5), indicating the strongest intracellular affinity), significantly upregulated peptides could be detected, while no corresponding significantly downregulated peptides could be detected (Figures 7E–G). It can be understood that in the granule cells of the DOR group, these three proteins were digested into extremely short peptides so that they could not be detected, while in the cells of the AC-treated group, these three proteins could not be digested after binding to AC, so that the significantly upregulated long peptides could be detected. In other words, MAP2K1, MAPK14 and GSK3B may be the key targets of AC binding.