Exploring biomarkers of mitochondrial dynamics in polycystic ovary syndrome and the effect of mesenchymal stem cell transplantation on ovarian mitochondria

Abstract

Background

Polycystic ovary syndrome (PCOS) is a prevalent gynecological disorder commonly associated with endocrine, reproductive, and metabolic dysfunctions. Mitochondrial dynamics (MD) is essential for cellular homeostasis and is implicated in various metabolic disorders. However, its role in PCOS pathogenesis remains underexplored. Mesenchymal stem cells (MSCs) have shown significant promise in ovarian function restoration.

Methods

Eight differentially expressed mitochondrial dynamics-related genes (DE-MDRGs) were identified through analysis of PCOS transcriptomic data from public databases. Biomarkers for PCOS were validated based on expression levels. Gene set enrichment analysis (GSEA) was conducted to investigate the functional role of these biomarkers. The expression of these biomarkers was examined across distinct cell clusters in both PCOS and control groups. A PCOS rat model was created, MSCs were transplanted, and the relative expression of the biomarkers was confirmed via RT-qPCR.

Results

DDHD2 and MRAS were identified as potential biomarkers, both of which are co-enriched in immune response-related pathways. Granulosa cells (GCs) were categorized into four distinct clusters, with clusters GC2 and GC3 showing higher expression levels in PCOS, while GC1 was more prevalent in the controls. MRAS expression in GC1 was notably lower in PCOS compared to controls. RT-qPCR analysis further revealed that DDHD2 was down-regulated, while MRAS was up-regulated in PCOS. MSC transplantation ameliorated these imbalances.

Conclusion

DDHD2 and MRAS are promising novel biomarkers for PCOS, and MSCs appear to protect mitochondrial kinetic homeostasis, offering new insights and potential therapeutic avenues for clinical intervention in PCOS.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13048-026-01974-6.

Background

Polycystic ovary syndrome (PCOS) is the most prevalent endocrine disorder, affecting approximately 3%-15% of women of reproductive age [1]. The condition is characterized by hallmark symptoms, including pronounced hyperandrogenism, menstrual irregularities, and polycystic ovarian morphology, often accompanied by insulin resistance and disturbances in glucose and lipid metabolism. Despite extensive research, the underlying pathogenesis of PCOS remains unclear [2]. Recent studies point to a multifactorial etiology, involving genetic predisposition, epigenetic modifications, environmental influences, oxidative stress, chronic low-grade inflammation, mitochondrial dysfunction, and metabolic disturbances, all of which may contribute to impaired ovarian function [3]. Treatment strategies remain largely symptomatic due to a limited understanding of these mechanisms. Thus, identifying novel biomarkers is critical for improving the accuracy and timeliness of PCOS diagnosis and treatment.

Mitochondria, as the core organelles of cellular energy metabolism [4], and their function and dynamics balance are critical for the maintenance of cellular homeostasis, and the associated abnormalities have been linked to a variety of disease pathogenesis [5]. In PCOS, granulosa cells often show abnormally high numbers of mitochondria, decreased function, and disturbed dynamics [6]. It was shown that Drp1-mediated excessive mitochondrial division promotes granulosa cell apoptosis [7], SYVN1 inhibits this process by promoting Drp1 degradation [8]. Furthermore, the mitochondrial fusion gene Mitoguardin2 is involved in the regulation of steroidogenesis in human ovarian granulosa cells [9]. In contrast, Mdivi-1 promotes steroidogenesis in granulosa cells by inhibiting mitochondrial division [10]. In terms of protective mechanisms, melatonin has been found to attenuate apoptosis in bovine ovarian granulosa cells by promoting mitochondrial autophagy [11]. These findings suggest that targeting the regulation of mitochondrial dynamics may provide potential targets for PCOS therapy. However, systematic studies of mitochondrial dynamics-related genes in PCOS are still relatively lacking, and their potential as disease biomarkers and therapeutic targets has not been fully explored.

Contemporary therapeutic strategies for patients with PCOS encompass a variety of interventions, including dietary changes, physical activity, and medications such as metformin, which are aimed at alleviating symptoms of hyperandrogenism and metabolic dysfunction. A promising addition to the therapeutic landscape is cellular therapy, which has emerged as a novel treatment approach for PCOS [12]. Mesenchymal stem cells (MSCs) are considered an ideal candidate for cell therapy due to their ability to modulate inflammation, reduce apoptosis, enhance antioxidant activity, and regulate immune responses. Recent studies suggest that MSC transplantation may offer a potential therapeutic avenue by restoring mitochondrial dynamics and biogenesis in ovarian cells [13]. Menstrual blood-derived stem cells and exosome therapy can activate pathways related to mitochondrial biogenesis and antioxidative defense [14]. Additionally, extracellular vesicles derived from umbilical cord MSCs (UCCC-MSCs) inhibit the mitosis of ovarian GCs, improving PCOS symptoms [15]. These findings highlight the therapeutic promise of MSC-based strategies in managing PCOS.

In this study, biomarkers related to mitochondrial dynamics in PCOS were identified through bioinformatics analysis. A PCOS animal model was established using female SD rats, and the effect of UC-MSC transplantation on mitochondrial dynamics was explored. Using single-cell transcriptome data, GCs were classified into subtypes, and biomarker expression was analyzed at the single-cell level. The results provide a theoretical foundation for the diagnosis and treatment of PCOS.

Methods

Data acquisition

In this study, PCOS-related datasets, including two RNA sequencing (RNA-seq) datasets and one single-cell RNA-seq (scRNA-seq) dataset, were downloaded from the Gene Expression Omnibus (GEO, https://www.ncbi.nlm.nih.gov/geo/) database. The training dataset, GSE34526 (platform: GPL570), included GCs from 7 patients with PCOS and 3 normal controls. The validation dataset, GSE168404 (platform: GPL16791), comprised GCs from 5 patients with PCOS and 5 normal controls. Additionally, the GSE240688 (platform: GPL24676) dataset was a scRNA-seq dataset containing GCs from 3 patients with PCOS and 3 normal controls. Mitochondrial dynamics, which include fission, fusion, and mitophagy, were central to this analysis [16]. A total of 15 mitochondrial fission-related genes and 9 mitochondrial fusion-related genes were obtained from the MitoCarta 3.0 database (http://www.broadinstitute.org/mitocarta) [17]. From the Molecular Signatures Database (MSigDB, https://www.gsea-msigdb.org/gsea/msigdb), 91 mitochondrial fission-related genes were retrieved, including those from the GOBP_REGULATION_OF_MITOCHONDRIAL_FISSION, GOBP_POSITIVE_REGULATION_OF_MITOCHONDRIAL_FISSION, and GOBP_MITOCHONDRIAL_FISSION categories. Mitochondrial mitophagy-related genes (MMRGs) were sourced from the Reactome (https://reactome.org/) and Kyoto Encyclopedia of Genes and Genomes (KEGG, https://www.kegg.jp/) databases. In total, 28 MMRGs were obtained from the Reactome database [18], including R-HSA-5,205,647, R-HSA-5,205,685, and R-HSA-8,934,903 pathways, while 72 MMRGs were identified from the hsa04137_Mitophagy pathways in KEGG. After removing duplicates, a total of 125 MDRGs were collected (Additional file S1).

Identification of differentially expressed genes (DEGs) and function enrichment analysis

DEGs between GCs of patients with PCOS and normal GCs were identified. In GSE34526, DEGs were filtered using the limma (v. 3.48.3) package (|log₂FC| ≥ 0.5, p-value < 0.05) [19]. A volcano plot of DEGs was generated using the ggplot2 (v. 3.3.5) package [20], and a heatmap of DEGs was created with the ComplexHeatmap (v. 2.14.0) package [21]. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were performed to explore the function of DEGs (adj. p-value < 0.05) using the clusterProfiler (v. 4.0.2) package [22]. Multiple comparisons were corrected using the Benjamini-Hochberg method.

Expression levels of differentially expressed MDRGs (DE-MDRGs)

To identify DEGs associated with mitochondrial dynamics, the intersection of DEGs and MDRGs was calculated, resulting in DE-MDRGs. The expression levels of DE-MDRGs were compared between the PCOS and normal controls in both GSE34526 and GSE168404 datasets using the Wilcoxon test (p-value < 0.05). DE-MDRGs with significant expression differences and consistent trends across both datasets were selected as biomarkers for PCOS. Finally, based on the GSE168404 dataset and GSE168404 data, receiver operating characteristic (ROC) curve analysis was employed to evaluate the diagnostic value of candidate biomarkers for PCOS.

Chromosome localization and subcellular localization

The chromosomal distribution of biomarkers was analyzed using the RCircos (v. 1.2.0) package [23]. Fasta files of biomarkers were downloaded from the National Center for Biotechnology Information (NCBI) gene database (https://www.ncbi.nlm.nih.gov/gene/?term=). These fasta files were then uploaded to the mRNALocater database (http://bio-bigdata.cn/mRNALocater) [24] for prediction of the subcellular localizations of the biomarkers.

Gene set enrichment analysis (GSEA) and gene-gene interaction (GGI) network

The background gene set (c5.go.bp.v2023.1.Hs.symbols.gmt) was downloaded from the MSigDB database. Spearman’s correlation coefficients were then calculated between the biomarkers and all other genes, with genes ranked according to the coefficient values. GSEA was performed using the clusterProfiler (v. 4.0.2) package (adj. p-value < 0.05) to explore the functional role of biomarkers. Multiple comparisons were corrected using the Benjamini-Hochberg method. Additionally, the GeneMANIA (http://genemania.org/) database was used to predict genes interacting with the biomarkers, and a GGI network was constructed based on these interactions.

Molecular regulatory network

To investigate the molecular mechanisms upstream of the biomarkers, biomarker-related microRNAs (miRNAs) and transcription factors (TFs) were predicted, resulting in the creation of a TF-mRNA-miRNA network. First, miRNAs were predicted from the miRDB (https://mirdb.org/) and Targetscan (https://www.targetscan.org/) databases. The hub miRNAs were identified by overlapping predictions from both databases (predicted.cutoff = 500,000). The top 20 TFs were retrieved from the ChEA3 (https://amp.pharm.mssm.edu/ChEA3) database. A TF-mRNA-miRNA network was then constructed, with biomarkers serving as a bridge linking the three components into a cohesive network.

Processing of scRNA-seq data

The GSE240688 dataset was processed using the Seurat (v. 4.1.0) package [25]. Initially, low-quality cells and genes were filtered out (min.cells = 3, min.features = 200). High-quality cells and genes were further selected using stricter criteria (200 < nFeature_RNA < 6000, 500 < nCount_RNA < 20,000, percent_mt ≤ 20%). After normalizing the data, the top 2000 highly variable genes were selected using the ‘vst’ function. Principal Component Analysis (PCA) was performed to identify meaningful principal components (PCs), using the ‘JackStrawPlot’ function (resolution = 0.4). The ‘FindNeighbors’ and ‘FindClusters’ functions were applied to categorize the high-quality cells into distinct clusters based on Uniform Manifold Approximation and Projection (UMAP) clustering. Marker genes of GCs, such as IGFBP7, STAR, and SERPINE2 [26], were used for cluster annotation. The top 5 DEGs, ranked by |log2FC|, were selected as marker genes for each cluster, using the ‘FindAllMarkers’ function. To explore the specific functions of distinct cell clusters in GCs, functional enrichment analysis was conducted using the ReactomeGSA (v. 1.12.0) package [27]. Finally, biomarker expression was analyzed between PCOS and normal controls in various cell clusters.

Cell-cell communication analysis and pseudotime analysis

In the GSE240688 dataset, the fraction and number of distinct cell clusters were quantified in GCs. The interactions between cell clusters were explored using the CellChat (v. 1.6.1) package [28]. Additionally, pseudotime analysis of GCs was performed with the Monocle (v. 2.22.0) package [29].

Animals and models

The sample consisted of 15 healthy, three-week-old female Sprague-Dawley (SD) rats (weights ranging from 46 to 70 g) obtained from Spearfish (Beijing) Biotechnology Co., Ltd at three weeks of age. The experimental protocol was approved by the Experimental Animal Ethics Committee of Tianjin Medical University Chu Hsien-I Memorial Hospital (220629002) and strictly adhered to ARRIVE guidelines. The SD rats were housed in the Animal Center under a 12-hour light/12-hour dark artificial lighting cycle, with a temperature-controlled environment set at 23 ± 2 °C and a relative humidity of 55 ± 5%. The rats had ad libitum access to a standard rodent diet and water. After a one-week acclimatization period, the rats were randomly assigned to two groups: the control group (n = 5), which received subcutaneous injections of 0.2 milliliters of sesame oil, and the model group (n = 10), which was induced with 6 milligrams of DHEA per 100 g of body weight, dissolved in sesame oil and administered via subcutaneous injection once daily for 23 consecutive days. The model group was provided with a high-sugar, high-fat diet, while the control group received a standard diet. After the induction and confirmation of the PCOS model by vaginal smear assessment, 8 rats with successful modeling were randomly divided into two groups: the PCOS group (n = 4) and the PCOS + hUC-MSCs group (n = 4). The rats in the PCOS + hUC-MSCs group were injected with 2 × 10⁶ hUC-MSCs (human UC-MSCs) by tail vein injection on Days 1 and 7. The hUC-MSCs were provided by LONGTAIYINXIN Biotechnology Co., Ltd. All rats were anesthetized with pentobarbital (35 mg/kg) via intraperitoneal injection after two weeks, followed by euthanasia by cervical dislocation on the same day. Serum samples and ovaries were collected for further analysis.

Determination of the estrous cycle

To assess the four phases of the estrous cycle, vaginal smears were microscopically examined and stained with Wright-Giemsa stain during two key periods: the final ten days of the DHEA induction phase and the intervention period [30].

RNA extractions and RT-qPCR

Total RNA from frozen tissue samples of the PCOS (n = 4), PCOS + hUC-MSCs (n = 4) and normal group (n = 5) was isolated using TRIzol Reagent (Ambion, Shanghai, China) according to the manufacturer’s protocol. Total RNA was then reverse-transcribed into cDNA using the SweScript First Strand cDNA synthesis kit (Servicebio, Wuhan, China). RT-qPCR was performed using the Universal Blue SYBR Green qPCR Master Mix (Servicebio, Wuhan, China) and the Applied Biosystems CFX96TM Real-Time PCR Detection System (BIO-RAD, USA). Primer sequences are provided in Additional file S2. Expression levels of biomarkers were normalized to GAPDH as an endogenous control. Data were analyzed using the 2^−ΔΔCT method [31], and each sample was measured in triplicate.

Statistical analysis

All statistical analyses were performed using R (v. 4.1.0) software. A p-value of < 0.05 was considered statistically significant (two-tailed). Spearman’s correlation (|cor| > 0.3, p < 0.05) was used for correlation analysis. Data visualization and analysis were carried out using GraphPad Prism 9 software. *p < 0.05; **p < 0.01.

Results

A total of 8 DE-MDRGs were identified in GSE34526

A total of 2231 DEGs were identified in GSE34526 between PCOS GCs and normal controls, with 1,542 upregulated and 689 downregulated genes (Fig. 1a, b). Functional enrichment analysis revealed significant associations with 1017 GO terms and 70 KEGG pathways (adj. p-value < 0.05) (Additional file S3, S4). GO enrichment analysis indicated that the DEGs were linked to terms such as ‘secretory granule membrane’, ‘tertiary granule’, and ‘myeloid leukocyte activation’ (Fig. 1c). KEGG enrichment analysis suggested involvement in pathways like ‘phagosome’, ‘osteoclast differentiation’, and ‘tuberculosis’ (Fig. 1d). Finally, eight DE-MDRGs were obtained by intersecting the 2231 DEGs with the 125 MDRGs (Fig. 1e).

Fig. 1.

Screening of mitochondrial dynamics related genes in PCOS and normal groups a. The Volcano map of differentially expressed genes (|log2FC| ≥ 0.5 and p-value<0.05, limma). b. The heat map distribution of differentially expressed genes c. GO enrichment analysis of differentially expressed genes (adj. p-value<0.05). d. KEGG enrichment analysis of differentially expressed genes (adj. p-value<0.05). e. Identification of mitochondrial dynamics related differentially expressed genes

DDHD2 and MRAS were determined as the biomarkers for PCOS

The eight DE-MDRGs identified were DDHD2, FUNDC1, LRRK2, MRAS, RAB24, RAB7B, UCP2, and ULK1. Significant differences in the expression levels of DDHD2 and MRAS were observed between the PCOS and control groups in both the GSE34526 and GSE168404 datasets (p-value < 0.05) (Fig. 2a, b). DDHD2 expression was significantly lower in PCOS compared to normal controls in both datasets (p-value < 0.05), whereas MRAS expression was significantly higher in PCOS (p-value < 0.05). Thus, DDHD2 and MRAS were selected as biomarkers. Moreover, both DDHD2 and MRAS demonstrated area under the curve (AUC) values exceeding 0.8 across both datasets, indicating their superior diagnostic efficacy (Additional figure S1). DDHD2 is located on chromosome 8, and MRAS on chromosome 3 (Fig. 2c). Subcellular localization analysis revealed that DDHD2 is predominantly localized in the nucleus, while MRAS is primarily found in the cytoplasm (Fig. 2d). Notably, both DDHD2 and MRAS were co-enriched in pathways related to immune response regulation, such as the ‘immune response regulatory signaling pathway’ and the ‘signaling pathway for immune response regulation of cell surface receptors’ (Fig. 2e, f). A GGI network was constructed based on the GeneMANIA database, including interactions such as DDHD2-RAPGEF5, DDHD2-AFDN, and MRAS-SHOC2 (Fig. 2g). Additionally, 15 hub miRNAs and the top 20 TFs were obtained from online databases to generate a TF-mRNA-miRNA network, including AR-MRAS-has-miR-1271-5p, LHX8-DDHD2-has-miR-495-3p, and MTF2-MRAS-has-miR-1297 (Fig. 2h).

Fig. 2.

DDHD2 and MRAS were determined as the biomarkers for PCOS a. b. The expression levels of 8 DE-MDRGs in the PCOS and normal groups (Wilcoxon test, p-value<0.05, n = 7 PCOS vs. 3 controls for GSE34526; n = 5 per group for GSE168404). c. The chromosomal distribution of DDHD2 and MRAS d. The subcellular localization of DDHD2 and MRAS e. The gene set enrichment analysis of DDHD2 (adj. p-value<0.05). f. The gene set enrichment analysis of MRAS (adj. p-value<0.05). g. GeneMANIA analyzes : The network diagram of genes related to DDHD2 and MRAS and their interactions. h. The transcription factors and miRNA of DDHD2 and MRAS

The granulosa cells were divided into 4 distinct cell clusters

In GSE240688, a total of 27,593 cells and 29,166 genes were obtained. After quality control, 20,132 high-quality cells and 29,166 genes remained for further analysis (Fig. 3a). UMAP clustering was performed using the top 2000 highly variable genes and the top 30 PCs (Fig. 3b, c, d). The high-quality cells were categorized into four distinct clusters (Fig. 3e). Annotation results confirmed that the marker genes of GCs were highly expressed across all four clusters (Fig. 3f), validating that the samples in GSE240688 were indeed GCs, which consisted of four distinct cell clusters. Marker genes specific to each cluster were highly expressed in their corresponding cell populations (Fig. 3g). Notably, a higher proportion of GC2 (cluster 2) was detected in PCOS, while more GC1 (cluster 1) cells were present in normal controls. Additionally, GC3 was more abundant in PCOS than in normal controls (Fig. 3h, i). Functional enrichment analysis revealed that these cell clusters were primarily involved in alanine metabolism. GC3 was notably enriched in pathways such as ‘sterols are 12-hydroxylated by CYP8B1’, ‘Activation of Na-permeable kainate receptors’, and ‘NEIL3-mediated resolution of interstrand cross-links (ICLs)’, compared to the other three clusters, while the ‘Cox reactions’ pathway showed lower enrichment in GC3 (Fig. 3j).

Fig. 3.

The granulosa cells were divided into 4 distinct cell clusters a. The first three images show the genes express in granulosa cells of PCOS and normal groups before quality control. The last three images show the genes express in granulosa cells of PCOS and normal groups after quality control (200<nFeature_RNA<6000, 500<n Count_RNA<20,000, percent_mt≤20%). b. Screening for highly variable genes c. The distribution of granulosa cells in the principal component analysis d. The principal component inflection points (PCIP) chart of granulosa cells e. The granulosa cells were divided into distinct cell clusters based on uniform manifold approximation and projection clustering f. The cell clusters were annotated into distinct cell subgroups based on the marker genes g. The expression levels of marker genes in different cell clusters h. The proportional and numerical distribution of granulosa cells subtypes and other different cell types in PCOS and normal groups (n = 3). i. The granulosa cells were divided into distinct cell clusters based on UMAP clustering in PCOS and normal groups (n = 3). j. The biological functions in which subclusters of granulosa cells are primarily involved

The expression level of MRAS was much lower in GC1 of PCOS

To investigate cell crosstalk among the four cell clusters, cell-cell communication analysis was performed. Interaction effects were observed among the clusters, with GC2 in PCOS exhibiting stronger interactions with other cells than in normal controls (Fig. 4a). Pseudotime analysis revealed that GCs were divided into seven distinct states. The trajectory of GC1 and GC2 began in state 1, and at the end of differentiation, the GCs differentiated into two separate states: State 6, associated with GC3, and State 7, associated with GC1 (Fig. 4b, c, d). Interestingly, State 6 with GC3 was more prevalent in PCOS, while State 7 with GC1 was more abundant in normal controls (Fig. 4e). Based on these observations, it is hypothesized that GCs differentiate from GC1 to GC3, potentially contributing to the increased risk of PCOS. The expression levels of biomarkers are shown in Fig. 4f. Consistent with the GSE34526 and GSE168404 datasets, DDHD2 was highly expressed in normal controls (p-value < 0.05), while MRAS was highly expressed in PCOS (p-value < 0.05) (Fig. 4g). The expression levels of biomarkers in distinct cell clusters were compared between the PCOS and normal controls. No significant difference in DDHD2 expression was observed between the groups (Fig. 4h). In GC1, MRAS expression was significantly lower in PCOS than in normal controls (p-value < 0.0001), whereas in GC4, MRAS expression was significantly higher in PCOS compared to normal controls (p-value < 0.05) (Fig. 4i).

Fig. 4.

The expression level of MRAS was much lower in GC1 of PCOS a. The cell-cell communication analysis of the 4 cell clusters in PCOS and normal groups b-e. The pseudotime analysis of the granulosa cells in PCOS and normal groups f. The expression levels of DDHD2 and MRAS in granulosa cells g. Differential expression of DDHD2 and MRAS in granulosa cells in PCOS and normal groups (Wilcoxon test, p < 0.05). h. Differential expression of DDHD2 in subclusters of granulosa cells in PCOS and normal groups (Wilcoxon test, p < 0.05). i. Differential expression of MRAS in subclusters of granulosa cells in PCOS and normal groups (Wilcoxon test, p < 0.05)

Effect of MSCs transplantation on the expression of DDHD2 and MRAS in PCOS

As shown in Fig. 5a, b, c, d, a regular estrous cycle is a key indicator of normal ovarian function. Elevated androgen levels have been identified as a potential cause of PCOS, disrupting the natural rhythm of the estrous cycle. The confirmation of the PCOS model was validated by observing at least two consecutive cycles that deviated from the typical estrous cycle pattern. Disruption and stagnation of the estrous cycle were observed in the PCOS group compared to the control group (Fig. 5e).

Fig. 5.

Effect of MSCs transplantation on the expression of DDHD2 and MRAS in PCOS (normal: model: MSCs transplantation = 5:4:4) a-d. The estrous cycle of rats: Diestrus, Proestrus, Estrous, Metestrus (Wright-Giemsa stain) e. The estrous cycle of rats in PCOS and normal groupsImages 1-5 are the normal group, images 6-13 are PCOS group. In the image, 1=diestrus, 2=metestrus, 3=estrous, 4=proestrus f.g. The expression of DDHD2 and MRAS in PCOS, NC(normal group) and Y (MSCs transplantation group), Data were analyzed by one-way ANOVA and presented as mean ± SD (*p < 0.05, **p < 0.01)

As shown in Fig. 5f, g, DDHD2 expression was significantly downregulated in PCOS compared to normal controls, and MRAS expression was upregulated in PCOS, consistent with the bioinformatics findings. MSC transplantation was able to upregulate DDHD2 and downregulate MRAS expression in PCOS. Additionally, MSC transplantation modulated the expression of biomarkers related to mitochondrial dynamics genes.

Discussion

Mitochondrial dysfunction at the cellular level can significantly disrupt systemic metabolic balance. Recent studies increasingly suggest that disorders in mitochondrial genes and function may contribute to the pathogenesis of PCOS [32]. This dysfunction is believed to underlie several characteristics of PCOS, including androgen excess, insulin resistance, obesity, abnormal follicular development, and inflammation [33]. Jiang et al. observed mitochondrial membrane swelling and rupture in rat models of PCOS [34], while Sreerangaraja et al. reported decreased mitochondrial count and mass, as well as increased fragmentation and restricted cell expansion in human GC samples from patients with PCOS [35]. PCOS has also been shown to reduce oxidative capacity and energy production in GCs [36–38]. These findings suggest a potential role for mitochondria in the development of PCOS. However, the current understanding of this relationship remains inadequate, highlighting the need for further research. Identifying mitochondrial-related biomarkers is essential for improving the diagnosis, prognosis, and understanding of the pathophysiology of PCOS. In this study, two biomarkers (DDHD2 and MRAS) associated with mitochondrial genes were identified, and their expression at the single-cell level was examined.

DDHD2 is a mammalian intracellular phospholipase A1 (iPLA1γ) containing a DDHD domain. Also known as KIAA0725p, DDHD2 exhibits phospholipase and lipase activities and plays an essential role in lipid homeostasis in the central nervous system [39]. It is involved in phospholipid degradation [40], vesicle trafficking [41], and mitochondrial function [42], in addition to its triacylglycerol (TAG) hydrolase activity. Junjie Chu’s study highlighted the role of circular RNA circRUNX1 in promoting papillary thyroid cancer progression and metastasis by sponging microRNA-296-3p and regulating DDHD2 expression [43]. Previous research has identified DDHD2 as a genetic factor interacting with epigenetics, a process linked to treatment responses in patients with schizophrenia [44]. DDHD2, a key regulator of the cardiolipin remodeling process, is essential for the maintenance of mitochondrial function [42]. The cardiolipin remodeling process has been shown to be decisive in safeguarding the structural stability and catalytic activity of the mitochondrial respiratory chain complex. It has been shown that disturbances in cardiolipin metabolism are associated with imbalances in mitochondrial division and fusion processes [45, 46]. Deficiency of DDHD2 has been shown to be associated with excessive generation of reactive oxygen species in mitochondria and subsequent apoptosis [47]. Knockdown of DDHD2 in neurons also impairs mitochondrial respiratory chain function and hinders ATP synthesis [48]. These findings reveal a protective role for DDHD2 in mitochondrial function. In granulosa cells with PCOS, downregulation of DDHD2 expression may ultimately lead to granulosa cell dysfunction and follicular developmental arrest by disrupting cardiolipin homeostasis, which in turn affects mitochondrial membrane stability, promotes excessive division, or inhibits normal fusion.

MRAS, a member of the RAS subfamily, shares regulatory and effector interactions with classical oncogenic RAS proteins [49]. In this study, we found that MRAS expression was significantly upregulated in PCOS granulosa cells. MRAS has been demonstrated to be a direct target of YAP [50] and that the Hippo-YAP pathway is involved in the regulation of granulosa cell function in the ovarian environment [51], we hypothesized that the initial up-regulation of MRAS in PCOS granulosa cells may possibly be related to altered activity of the Hippo-YAP signaling axis. Combined with the fact that YAP itself has a role in maintaining the stability of the mitochondrial network and mitigating mitochondrial dysfunction and oxidative damage [52, 53], this up-regulation of MRAS may initially be a compensatory response initiated by the cell to cope with abnormal mitochondrial functions. However, in the pathological setting of PCOS, the persistent abnormally high expression of MRAS may be beyond normal physiological regulation and may instead disrupt the kinetic balance of mitochondrial fission and fusion, thereby exacerbating metabolic stress and dysfunction in granule cells. Further single-cell analysis revealed that MRAS showed specific expression in subpopulations of granulosa cells with different functional states. MRAS was significantly under-expressed in the GC1 subpopulation with high expression of the anti-oxidative stress-related gene MT3 [54]; in contrast, it was significantly up-regulated in the GC4 subpopulation with high expression of the metabolic stress marker GDF15 with the pro-apoptotic factor BBC3 [55–57]. We hypothesized that abnormal MRAS upregulation focused on GC4 in the stress state may be involved in abnormal follicular development in PCOS by amplifying mitochondrial dysfunction through synergistic stress and apoptosis pathways. However, it should be noted that whether the differences in MRAS expression among single-cell subpopulations reflect real biological heterogeneity or are affected by single-cell technical variation still needs to be confirmed by cross-platform experiments. In addition, the specific molecular mechanism by which MRAS regulates mitochondrial dynamics in PCOS granulosa cells requires further experimental validation and functional exploration.

In this study, bioinformatics approaches were utilized to identify two biomarkers associated with mitochondrial dynamics, derived from the GEO database. The analysis emphasized the key roles of GC1 and GC3 cell clusters in GC differentiation. The results suggest that differentiation from GC1 to GC3 may elevate the risk of PCOS. Our investigation in PCOS model rats further demonstrates that alterations in the expression of DDHD2 and MRAS are strongly linked to the development of PCOS. Specifically, differential mRNA expression of MRAS and DDHD2 was observed in the ovaries of PCOS rats compared to controls, providing additional evidence of impaired mitochondrial dynamics in PCOS.

MSCs exhibit unique biological functions, including antioxidant, anti-inflammatory, anti-apoptotic, and immunomodulatory properties, making them optimal candidates for cell therapy. Consequently, MSC-based therapies have gained increasing attention for treating various female reproductive disorders [13]. MSCs have shown promise in tissue repair and ovarian function enhancement [58], and their potential in PCOS therapy is being explored. Research has demonstrated that MSCs improve mitochondrial dynamics in PCOS by promoting mitochondrial biogenesis and reducing oxidative stress. Specifically, studies indicate that mitochondrial biogenesis decreases and oxidative stress increases in PCOS GCs, accompanied by reduced estrogen production. Administration of MSCs has been shown to alleviate these abnormalities significantly [14]. Our qPCR data showed that MSCs transplantation affected the expression of genes related to mitochondrial dynamics. This finding echoes the findings of Abdi et al. in a PCOS model, whose study showed that MSCs can be involved in regulating mitochondrial dynamics, promoting mitochondrial biogenesis, and ameliorating the redox state and inflammatory response through the PI3K-AKT pathway [13]. However, the effects of MSCs on mitochondrial function in ovarian tissues were not assessed in this study, and future studies are needed to further validate whether MSCs can directly regulate mitochondrial dynamics at the structural and functional levels by directly detecting mitochondrial morphology, fission/fusion markers, autophagy, and related functional parameters.

This study identified DDHD2 and MRAS as biomarkers associated with mitochondrial dynamics, a finding that was validated through bioinformatics analysis and confirmed in PCOS rat models. Moreover, MSCs transplantation regulates the expression of genes related to mitochondrial dynamics in PCOS. However, there are some limitations of this study. First, the limited sample size may lead to insufficient statistical efficacy, affecting the comprehensiveness and reliability of the results, which needs to be further validated in a larger cohort. Second, the population data showed that MRAS was up-regulated in PCOS as a whole, whereas the single-cell level showed heterogeneity with down-regulation of GC1 and up-regulation of GC4. Whether the difference originated from the true subpopulation-specific regulation or from the variation of the single-cell technology itself still needs to be confirmed by cross-platform experiments. In addition, animal experiments were performed using a rat model, which differs species-wise from humans in DDHD2/MRAS homology and granulosa cell mitochondrial phenotype. Finally, the regulation of mitochondrial morphology, fission/fusion, autophagic processes, and functional parameters (e.g., ATP, ROS, membrane potential) by MRAS has not been directly confirmed by functional experiments in this study. To this end, we plan to simultaneously validate the regulatory function of MRAS and the therapeutic effects of MSCs in cellular and animal models using a combination of gene editing and stem cell transplantation technologies in subsequent studies; and integrate high-resolution live cell imaging, real-time functional assays and multi-omics analyses to analyze its regulatory network at the dynamic system level. Ultimately, the translational potential of targeting this pathway as a new strategy for PCOS treatment will be evaluated in combination with the validation of expanded clinical samples.

Conclusion

In conclusion, DDHD2 and MRAS could serve as valuable biomarkers for PCOS, aiding in diagnosis and treatment. In the future, regulatory strategies targeting DDHD2/MRAS and related pathways may be further explored to correct mitochondrial dysfunction and provide a new direction for precision treatment of PCOS. In addition, MSCs therapy shows potential in regulating mitochondrial dynamic genes, suggesting that the combination of mitochondria-targeted intervention and stem cell therapy may be an effective way to improve the reproductive and metabolic prognosis of PCOS patients.

Abbreviations

Abbreviations

Full name

PCOS

polycystic ovary syndrome

MD

mitochondrial dynamics

MSCs

mesenchymal stem cells

UC-MSCs

umbilical cord mesenchymal stem cell

DEGs

differentially expressed genes

MDRGs

mitochondrial dynamics related genes

GSEA

gene set enrichment analysis

GEO

Gene Expression Omnibus

MSigDB

Molecular Signatures Database

MMRGs

mitochondrial mitophagy-related genes

KEGG

Kyoto Encyclopedia of Genes and Genomes

GO

Gene Ontology

DE-MDRGs

the intersection of differentially expressed genes and mitochondrial dynamics related genes

GGI

gene-gene interaction

TFs

transcription factors

PCA

principal component analysis

PCs

principal components

UMAP

uniform manifold approximation and projection

SD

Sprague-Dawley

GC

granulosa cell

Authors’ contributions

ZH designed the research. NT and OC performed most of the experiments. DC, WZ, WT, YW, YC, JF assisted with experiments and data analysis. NT and OC drafted the manuscript. ZH edited and revised the manuscript. All authors read and approved the final manuscript.

Funding

This work was supported by Tianjin Natural Science Foundation (23JCYBJC00720), National Natural Science Foundation of China (82171629), Tianjin Key Medical Discipline Construction Project (TJYXZDXK-3-007B).

Data availability

The datasets analyzed during the current study are available in the Gene Expression Omnibus (GEO) database with the following accession numbers: GSE34526 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE34526), GSE168404 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE168404), and GSE240688 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE240688).

Declarations

Ethics approval and consent to participate

The experimental protocol was approved by the Experimental Animal Ethics Committee of Tianjin Medical University Chu Hsien-I Memorial Hospital (file number 220629002) and strictly followed ARRIVE guidelines.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.