The impact of menstrual cycle phase and BMI on mitochondrial dysfunction in PMOS endometrium

Abstract

STUDY QUESTION

Does the endometrium of oligo-ovulatory women with polyendocrine metabolic ovarian syndrome (PMOS) exhibit menstrual cycle phase-specific transcriptomic alterations compared to women without PMOS?

SUMMARY ANSWER

Compared to healthy controls, women with PMOS exhibit endometrial transcriptomic dysregulations which are menstrual cycle phase-specific and, most notably, related to mitochondrial function.

WHAT IS KNOWN ALREADY

Infertility and gestational dysfunction in women with PMOS are increasingly attributed to endometrial abnormalities, yet the underlying mechanisms behind PMOS-related endometrial dysfunction remain unclear. Previous studies have focused on anovulatory phenotypes, resulting in a poor understanding of cycle phase-specific defects in the endometria of women with PMOS.

STUDY DESIGN, SIZE, DURATION

Endometrial biopsies were obtained from oligo-ovulatory women with PMOS and healthy controls at four distinct menstrual cycle phases (proliferative, early-, mid-, and late-secretory phases) and subjected to RNA sequencing. Parallel in vitro experiments were conducted using cultured endometrial stromal cells (ESCs) and epithelial organoids (EEOs) derived from an additional endometrial sample cohort.

PARTICIPANTS/MATERIALS, SETTING, METHODS

RNA sequencing was performed on biopsies from 78 (oligo-)ovulatory women, age- and BMI-matched between PMOS (n = 39) and control (n = 39) groups. Phase-specific transcriptomic profiles and pathway enrichment analyses were performed. For functional assays, endometrial cells were isolated from four PMOS and four control donors per BMI group (lean: < 25 kg/m²; obese: ≥30 kg/m²). Mitochondrial function, including respiration, ATP and reactive oxygen species production, and membrane potential, was assessed in both cell types before and after hormonal exposure.

MAIN RESULTS AND THE ROLE OF CHANCE

Endometrial transcriptomes clustered primarily by cycle phase; however, PMOS-specific dysregulation was observed related to extracellular matrix organization, estrogen receptor (ESR) signaling, ion transport, and mitochondrial respiratory function. Gene trajectory analyses revealed phase-specific alterations in progesterone-, insulin signaling-, and mitochondria-related gene expression during the early-secretory phase, and epithelial function during the mid-secretory phase. Functionally, decidualized ESCs from obese PMOS women displayed reduced mitochondrial respiration, whereas EEOs from lean PMOS patients exhibited increased mitochondrial activity.

LARGE SCALE DATA

Sequencing data are available from the corresponding author upon request.

LIMITATIONS, REASONS FOR CAUTION

Women with PMOS were oligo-ovulatory while controls exhibited regular cycles. The sample size for in vitro studies was modest, warranting validation in larger cohorts. Additionally, the use of bulk RNA sequencing may obscure cell-type-specific transcriptomic signatures, emphasizing the need for future single-cell approaches to dissect cellular heterogeneity in PMOS endometrium.

WIDER IMPLICATIONS OF THE FINDINGS

This study uncovers dynamic, cycle phase-dependent molecular and bioenergetic alterations in the endometrium of oligo-ovulatory women with PMOS. Mitochondrial dysfunction, particularly during the implantation window, emerges as a key pathological feature. The observed BMI-dependent divergence in mitochondrial function suggests that personalized therapeutic strategies may be necessary to improve reproductive outcomes in women with PMOS.

FUNDING

Funded by the Jusélius Foundation, Novo Nordisk Foundation, Research Council of Finland, and the Horizon 2020 Marie-Curie MATER Innovative Training Network (all to T.T.P.), with additional support from the University of Oulu Scholarship Foundation (L.L.), the Estonian Research Council (E.V.M. and I.R.), and the Estonian Research Council grant no. PRG1076, Swedish Research Council grant no. 2024-02530, Novo Nordisk Foundation grant no. NNF24OC0092384 and Horizon Europe grant NESTOR, grant no. 101120075 (A.S.).

DISCLOSURES

The authors declare no competing interests.

Introduction

Polyendocrine metabolic ovarian syndrome (PMOS), formerly known as polycystic ovary syndrome (PCOS), is a complex endocrine disorder affecting up to one in six women of reproductive age. It is diagnosed by the presence of a minimum of two of the following criteria: ovulatory dysfunction, clinical or biochemical signs of hyperandrogenism (HA), and polycystic ovarian morphology (Teede et al., 2023, 2026). Women with PMOS commonly experience reproductive impairments—such as anovulation and impaired embryo implantation—as well as adverse pregnancy outcomes, including miscarriage, preeclampsia, and gestational diabetes, often resulting in smaller family size (Palomba et al., 2015; Persson et al., 2019; Bahri Khomami et al., 2024). Even though the most significant cause of subfertility in women with PMOS is anovulation, the anovulatory phenotype in women with PMOS is variable and often improves with age (Brown et al., 2011; Liang et al., 2011; Piltonen et al., 2026). Endometrial dysfunction, however, represents a significant, but frequently underestimated contributor to impaired reproductive outcomes in women with PMOS (Piltonen, 2016; Palomba et al., 2021; Stener-Victorin et al., 2024). Limited understanding of the causal molecular mechanisms contributing to endometrial dysfunction further complicates fertility treatment strategies.

The uterine endometrium is a hormonally responsive tissue that predominantly comprises endometrial stromal cells (ESCs), epithelial cells (EECs), and immune cells, and undergoes cyclical remodeling during the menstrual cycle. During the proliferative phase, rising estradiol levels stimulate ESC and EEC proliferation and vascular expansion to support endometrial thickening (Pijnenborg et al., 2006; Fitzgerald et al., 2016). Following ovulation, increasing progesterone (P4) levels initiate the secretory phase, promoting glandular and vascular maturation, and increasing the infiltration of immune cells (Jabbour et al., 2006); the uterine natural killer cell population is increased, and the inflammatory M1 macrophage population that predominates the proliferative endometrium shifts to a more anti-inflammatory M2 subtype (Pepe et al., 2018). Progesterone also drives decidualization—the differentiation of fibroblast-like ESCs into secretory decidual cells—and primes EECs for implantation through enhanced secretory activity and expression of adhesion molecules (Wang et al., 2020; Fitzgerald et al., 2023). These coordinated changes establish a receptive endometrial environment for embryo implantation during the window of implantation (WOI), 7–8 days post-ovulation.

PMOS endometrium exhibits inherent abnormalities, as PMOS and non-PMOS samples show distinct clustering in microarray analysis, and the PMOS endometrial transcriptome displays downregulated pathways related to glucose metabolism, cell cycle, extracellular matrix (ECM), and P4 signaling (Kim et al., 2009; Savaris et al., 2011; Piltonen, 2016; Palomba et al., 2021). Fluorescent-activated cell sorting has enabled the study of genetic alterations in different endometrial cell types in the proliferative phase, including mesenchymal stem cells (Piltonen et al., 2013), supported by single-nuclei RNA sequencing analysis of proliferative phase endometrium of women with PMOS, highlighting cell type-specific differences related to integrin signaling (in EECs) and ECM (in EECs and ESCs) (Eriksson et al., 2025). While these findings underscore PMOS-specific endometrial dysfunction, a critical limitation of many previous studies has been a lack of stringent patient and cycle phase matching, potentially masking key cycle-dependent alterations impacting the viability of embryo implantation.

Another exacerbating factor in PMOS-related reproductive complications is the prevalence of systemic metabolic dysfunction, including excess body weight (affecting up to 60% of the PMOS population) and insulin resistance (Stener-Victorin et al., 2024), that diminish endometrial health. Specifically, PMOS endometrium exhibits disrupted glucose metabolism and heightened inflammatory responses (Piltonen, 2016; Khatun et al., 2021; Palomba et al., 2021); the key mechanisms driving these changes, however, remain poorly understood.

Cell metabolism is tightly linked to mitochondrial performance, and mitochondrial dysfunction has been proposed as a contributing factor in PMOS pathophysiology (Zhang et al., 2019), potentially via oxidative stress- and inflammation-related mechanisms (Holmström et al., 2012; Heinonen et al., 2015). While mitochondrial dysfunction and altered mitochondrial DNA (mtDNA) content have been reported in endometriosis and endometrial cancer (Cormio et al., 2012; Wang et al., 2005), studies on mitochondrial function in endometrial cells in PMOS remain scarce. Our previous study uncovered decreased mitochondrial gene expression in PMOS-derived endometrial epithelial organoids (EEOs) (Luyckx et al., 2025b), but despite other data, mainly originating from hormone-treated PMOS-like rodent models, there remains a critical gap in understanding mitochondrial dynamics in the human PMOS endometrium (Zhang et al., 2021).

Despite recent insights into PMOS pathophysiology, molecular characterization of PMOS uterine endometrium remains limited. Here, we applied bulk RNA-sequencing (RNA-seq) to investigate the transcriptomic landscape of the endometrium across different cycle phases and assessed ESC and EEC mitochondrial function using a well-characterized sample set of oligo-ovulatory women with PMOS, which enabled secretory phase analysis alongside well-characterized age- and BMI-matched controls. Our findings provide novel insights into the dynamic molecular alterations in PMOS endometrium across the menstrual cycle, including mitochondrial dysfunction during the critical point of endometrial preparation for embryo implantation. The results establish a foundation for more precise therapeutic strategies improving the fertility and pregnancy outcomes of women with PMOS.

Materials and methods

Ethical approval for the use of human tissues

The Ethical Committee of Northern Ostrobothnia Hospital District approved this study (65/2017). Informed consent forms were obtained from all participants in accordance with the Declaration of Helsinki.

Subjects

Endometrial biopsies from women with a PMOS diagnosis (minimum 2/3 Rotterdam criteria fulfilled, diagnosed by a gynecologist; Teede et al., 2023) who had occasional ovulations (PMOS), and healthy volunteers with regular menstrual cycles, who were age- and BMI-matched to the cases (Ctrl), were collected in Oulu University Hospital. Ovulatory status was determined by LH surge test tracking up to 6 months, where one ovulation during the tracking period was considered occasional. Ovulation was confirmed by ultrasound to detect the corpus luteum and by serum P4 measurement. Exclusion criteria included breastfeeding or pregnancy within 3 months, use of hormonal medication (including metformin and GLP-1 agonists) within 2 months, smoking, and uterine abnormalities (e.g. hyperplasia or polyps, assessed by ultrasound and histology). Participants were aged 21–42 years. Fasting blood samples were collected between menstrual cycle days 2–8. Hirsutism was diagnosed based on the modified Ferriman–Gallway score (FG score) assessed by a clinician (Lizneva et al., 2016). For subgroup analysis, a cut-off of FG score 5 was used to segregate hyperandrogenic (FG ≥ 5) from non-hyperandrogenic (FG < 5) subjects. BMI subgroups were defined as: lean (< 25 kg/m²), overweight (≥ 25 kg/m²), and obese (≥ 30 kg/m²). Fasting glucose, insulin, serum anti-Müllerian hormone (AMH), LH, FSH, P4, testosterone, and sex hormone-binding globulin (SHBG) were measured as described by Lee et al. (2024) and Virtanen et al. (2025). Homeostatic model assessment for insulin resistance (HOMA-IR) was calculated as: fasting glucose (mmol/l) × fasting insulin (mIU/l)/22.5. Free androgen index (FAI) was calculated as: testosterone (nmol/l)/SHBG (nmol/l) × 100. Participant characteristics are presented in Table 1 (for transcriptome analysis), Table 2 (for in vitro assays), Supplementary Tables S1 and S2 (for subgroup analyses for HA and BMI, respectively).

Table 1.

Participant characteristics transcriptome analysis.

	Ctrl (n = 39)	PMOS (n = 39)	P-value
Age	34.0 [29.0; 39.0]	34.0 [30.0; 39.0]	0.674
BMI (kg/m²)	25.4 [22.7; 31.5]	26.1 [22.7; 30.8]	0.479
WHR	0.80 (±0.05)	0.81 (±0.05)	0.230
Glucose (mM)	5.8 [4.8; 5.3]	5.2 [4.9; 5.4]	0.495
Insulin (mIU/l)	7.5 [5.1; 11.1]	7.7 [4.8; 9.7]	0.978
HOMA-IR	1.63 [1.04; 2.76]	1.71 [1.12; 2.32]	>0.999
AMH (ng/ml)	1.74 [1.34; 2.95]	4.93 [2.50; 5.87]	<0.001
LH (IU/l)	7.47 [5.26; 9.20]	9.13 [6.405; 13.07]	0.027
FSH (IU/l)	4.43 [3.05; 6.79]	4.34 [2.8; 6.225]	0.563
LH/FSH	1.50 [1.30; 1.97]	2.03 [1.57; 3.12]	0.006
T (nmol/l)	1.16 (±0.34)	1.51 (±0.52)	<0.001
SHBG (nmol/l)	61.50 [36.20; 85.75]	52.56 [44.27; 70.22]	0.85
FAI	1.81 [1.35; 2.67]	2.88 [2.02; 3.42]	0.002
FG score*	2.5 [1.8; 3.0]	7.5 [4.0; 9.0]	<0.001

Data are presented as mean±SD for normally distributed data or median [25th percentile; 75th percentile] for not normally distributed data. T-test or Mann–Whitney test. P < 0.05 depicted in bold. Ctrl, control; PMOS, polyendocrine metabolic ovarian syndrome; WHR, waist-hip-ratio; HOMA-IR, homeostatic model assessment for insulin resistance; AMH, anti-Müllerian hormone; T, testosterone; SHBG, sex hormone-binding globulin; FAI, free androgen index; FG, modified Ferriman–Gallwey score.

^*n = 36 (Ctrl) or 37 (PMOS) for FG score due to missing values.

Table 2.

Participant characteristics for in vitro studies.

	Lean			Obese			L- vs O-Ctrl	L- vs O-PMOS
	L-Ctrl (n = 8*)	L-PMOS (n = 8*)	P-value	O-Ctrl (n = 6*)	O-PMOS (n = 6*)	P-value	P-value	P-value
Age	37.0 [30.2; 39.0]	38.0 [31.5; 39.2]	0.858	38.0 [30.5; 38.7]	31.5 [29.0; 34.0]	0.787	0.874	0.355
BMI (kg/m²)	22.9 (±2.1)	22.6 (±1.7)	0.772	33.4 (±2.4)	34.3 (±3.8)	0.660	<0.001	<0.001
WHR	0.80 [0.78; 0.82]	0.83 [0.81; 0.84]	0.262	0.81 [0.79; 0.83]	0.87 [0.86; 0.89]	0.058	0.643	0.018
Glucose (mM)	4.8 [4.7; 5.1]	5.1 [5.0; 5.3]	0.454	5.3 [5.2; 5.5]	5.7 [5.4; 6.0]	0.177	0.163	0.050
Insulin (mIU/l)	5.4 [3.7; 6.7]	6.3 [4.6; 7.8]	0.339	10.7 [8.3; 13.3]	14.5 [11.8; 19.3]	0.132	0.007	0.002
HOMA-IR	1.10 [0.78; 1.59]	1.45 [1.08; 1.85]	0.340	2.68 [1.98; 3.07]	3.71 [2.83; 5.17]	0.179	0.006	0.002
AMH (ng/ml)	3.47 (±1.62)	7.45 (±4.85)	0.046	1.45 (±0.90)	5.66 (±2.99)	0.008	0.023	0.450
LH (IU/l)	9.18 [6.18; 9.27]	7.99 [7.40; 13.34]	0.508	6.29 [5.65; 6.67]	8.41 [7.56; 9.42]	0.093	0.558	0.945
FSH (IU/l)	7.28 [6.71; 7.86]	7.03 [5.90; 9.78]	0.915	8.21 [7.69; 9.88]	5.01 [4.92; 6.36]	0.026	0.084	0.180
LH/FSH	1.16 (±1.14)	1.53 (±0.51)	0.152	0.61 (±0.19)	1.51 (±0.37)	<0.001	0.002	0.954
T (nmol/l)	1.09 [0.89; 1.18]	1.35 [0.84; 2.19]	0.508	0.83 [0.19; 0.86]	1.20 [1.04; 1.49]	0.002	0.062	0.653
SHBG (nmol/l)	81.12 (±40.80)	65.30 (±22.27)	0.394	40.03 (±6.89)	37.43 (±13.54)	0.684	0.035	0.022
FAI	1.31 [1.19; 1.40]	2.57 [1.46; 3.46]	0.423	1.94 [1.67; 2.01]	3.23 [2.63; 3.75]	0.015	0.222	0.386
FG score	3 [2.0; 3.5]	6.8 [4.7; 7.2]	0.018	2.5 [2.0; 4.5]	9.5 [7.2; 13.2]	0.047	0.913	0.149

Data are presented as mean±SD for normally distributed data or median [25th percentile; 75th percentile] for not normally distributed data. T-test or Mann–Whitney test. P < 0.05 depicted in bold. Ctrl, control; PMOS, polyendocrine metabolic ovarian syndrome; L, lean; O, obese; WHR, waist-hip-ratio; HOMA-IR, homeostatic model assessment for insulin resistance; AMH, anti-Müllerian hormone; T, testosterone; SHBG, sex hormone-binding globulin; FAI, free androgen index; FG, modified Ferriman–Gallwey score.

^*n = 6–8 depending on availability of serum measurements for AMH, LH, FSH, T, and SHBG.

Endometrial biopsies

Endometrial tissue samples were collected with a plastic suction curette (Pipelle, Cooper Surgical, USA). Secretory phase timing was determined by urine LH surge testing (Clearblue, Switzerland) and dating was confirmed by ultrasonography and histological assessment by an experienced pathologist. Samples were collected from proliferative phase (PE): 6–8 days after the beginning of menstruation, from early-secretory phase (ESE): 2–4 days after LH surge (LH + 2–4), from mid-secretory phase (MSE): LH + 7–8, and from late-secretory phase (LSE): LH + 11–12. Endometrial biopsies for transcriptome analysis were snap-frozen in liquid nitrogen and stored at −80°C until use. Four samples from women with and 4 from women without PMOS during MSE phase were fixed for transmission electron microscopy (TEM) (see below). For transcriptome analysis, 78 endometrial samples were collected across different cycle phases (n = 9–10 per phase per group). Thirteen participants gave two samples from different cycle phases in two separate cycles.

RNA isolation and library preparation for RNA-seq

For RNA-seq, RNA was extracted from endometrial samples by homogenizing frozen tissue with mechanical lysis (bead with lysis buffer (Qiagen, Germany)) and purified with chloroform followed by RNA extraction using the miRNAeasy mini kit with an automated QIAcube workstation, following the manufacturer’s protocol (Qiagen). RNA concentration and purity were measured with a NanoVue Plus (GE Healthcare, IL, USA), RNA integrity values and concentration were measured with Bioanalyzer (Agilent Technologies, CA, USA) and a Qubit fluorometer (Thermo Fisher Scientific, MA, USA), after which the samples were diluted to 50 ng/µl. Samples were stored in −80°C, and library preparation and sequencing were performed at the Core Facility of Genomics, University of Tartu. For this, 10 ng of each sample was used for library preparation using the UPX 3′ Transcriptome kit (Qiagen). Samples were pooled after the addition of sample-specific index and unique molecular identifiers (UMIs) during cDNA synthesis. The average size of the pooled library fragments was measured with a high-sensitivity DNA kit and a 2100 Bioanalyzer (both from Agilent), and the library concentration was determined on a Qubit 4 with the dsDNA HS Assay kit (Thermo Fisher). Sequencing was carried out on a NextSeq 500 instrument (Illumina, CA, USA) using Mid Output Kit v2.5 (Illumina) with paired-end configuration of 100 + 27 bp according to the recommendations of the sequencing library preparation kit provider.

RNA-seq data analysis

Raw sequenced reads were processed using UMI-tools (Smith et al., 2017). Sequenced reads were de-multiplexed by extracting reads containing cell ID barcodes used in the library preparation. One mismatched base in the barcode sequence was allowed, and corresponding reads were retained in the analysis. Next, reads were mapped to the GRCh38.p12 reference genome using STAR aligner version 2.5.2 (Dobin et al., 2013) with default parameters but not allowing for multi-mapping. Reference genome-corresponding GENCODE release version 31 (Frankish et al., 2019) was used to annotate alignments to known gene loci. Read quantification was done by obtaining molecule counts based on UMI barcodes, according to the UMI-tools workflow. Next, the data were imported and pre-processed in R (Foundation for Statistical Computing, Austria). Raw data were filtered to remove genes with missing counts in eight or more samples (∼50% of total) and quantile normalized, then differential expression testing between conditions and subgroups (see above: Subjects) was performed using DESeq2 (Love et al., 2014). Gene lists obtained from differential expression testing were ranked by −log10p × log2FC, where p denotes unadjusted P-values and FC the fold-change. For all heatmap visualizations, genes with P-value <0.05 were plotted using normalized expression values. For unsupervised clustering analysis using Uniform Manifold Approximation and Projection (UMAP), normalized data were projected onto a latent space using 10 principal components. Gene set enrichment analysis (GSEA) was conducted using the ReactomePA package (Yu and He, 2016). Reactome annotations (Jassal et al., 2020) were used as the reference set for GSEA. The list of enriched pathways per cycle phase can be found in Supplementary Data File S1. Cell deconvolution was performed using XCell (Aran et al., 2017). Related P-values can be found in Supplementary Data File S2. Cell trajectories and genes associated with them while different across conditions, were defined with the Slingshot method and subsequent general additive models (GAMs) (Street et al., 2018) (Supplementary Data File S3). For graphically reporting differential mitochondrial gene expression across phases, mean gene expression values per group were plotted against the menstrual cycle phases. The list of targets can be found in Supplementary Data File S4. For subgroup analysis, unadjusted P-values and Benjamini–Hochberg-adjusted P-values (p.adj) were reported in Supplementary Data File S5.

RNA isolation and RT–qPCR

RNA was extracted from endometrial tissue or cells using the RNeasy kit (Qiagen) followed by DNase treatment according to the manufacturer’s instructions. RNA quantity and quality were tested via Dropsense 16 (Trinean, Belgium). cDNA synthesis was performed using the RevertAid first strand DNA synthesis Kit (Thermo Fisher) followed by reverse transcriptase quantitative PCR (RT–qPCR) analysis using a CFX connect Real-Time PCR Detection System (Bio-Rad, CA, USA). Triplicate cDNA samples were used. Forward and reverse primers were designed using PrimerBank or NCBI PrimerBLAST, and sequences are listed in Supplementary Table S3. GAPDH and YWHAZ were used as housekeeping genes. Relative expression levels are shown as 2 − dCt (mean±SD) in which dCt = Cttarget gene−Cthousekeeping gene.

Transmission electron microscopy

Whole endometrial samples were fixed in EM fixative (1% glutaraldehyde, 4% formaldehyde mixture in 0.1 M phosphate buffer, pH 7.4). For ESC fixation, cell medium was removed and cells were incubated in EM fixative for 10 min and gently detached from the cell plate by scratching, followed by additional fixation for 1 h and centrifugation (6000 rpm, 2 min). The cell pellet and endometrial samples were post-fixed in 1% osmium tetroxide, dehydrated in acetone, and embedded in Epon LX 112 (Ladd Research Industries, VT, USA). Thin sections (80 nm) were cut with an EM UC7 ultramicrotome (Leica Microsystems, Germany), stained in uranyl acetate and lead citrate, and examined in a Tecnai Spirit G2 transmission electron microscope (FEI Europe, The Netherlands). Images were captured with a Quemesa camera (Olympus Soft Imaging Solutions GMBH, Germany).

Endometrial cell isolation

For in vitro assays, endometrial cell isolation was performed with PE phase endometrial samples as described in our previous publications (Piltonen et al., 2015; Khatun et al., 2017). Briefly, endometrial samples were digested with collagenase 6.4 mg/ml collagenase type I (Sigma-Aldrich, MO, USA) and 100 U/ml hyaluronidase (Sigma-Aldrich) in Hanks buffered salt solution with Ca²⁺ and Mg²⁺ (HBSS, Gibco, USA) following separation of epithelial glands from stromal cells via passage through a 40 µm cell strainer. The stromal fraction, which passed through the strainer, was incubated for 2 h at +37°C in stromal cell media (SCM, 3:1 Dulbecco’s Modified Eagle’s Medium—high-glucose (DMEM, phenol red-free, Sigma-Aldrich)/MCDB-105, supplemented with 0.676 mM sodium pyruvate (Sigma-Aldrich), 10% charcoal-stripped fetal bovine serum (FBS, BioIVT, New York, USA), 1% penicillin–streptomycin mix (Gibco), 50 µg/ml gentamycin (Gibco), and 5 µg/ml insulin (Sigma-Aldrich)) to separate them from the non-attaching epithelial compartment. The epithelial glands were frozen in liquid nitrogen in 10% dimethyl sulfoxide (Sigma-Aldrich) in keratinocyte serum-free medium (KSFM, 1% FBS, 0.2% defined KSFM supplement (ThermoFisher)) in defined KSFM basal medium (Gibco) for EEO establishment (see below).

ESC culture and hormone treatment

ESCs were cultured in SCM in 10 cm culture dishes at +37°C with 5% CO₂ until confluent and media were changed every 2–3 days. For ESC decidualization, cells were starved for 48 h in SCM containing 2% FBS followed by decidualization for 96 h in 2% SCM media supplied with 10 nM estrogen (E2, Sigma-Aldrich), 1 µM P4 (Sigma), and 0.5 mM 8-bromo cyclic adenosine monophosphate (cAMP, Tocris, UK). Cells from low passage (P1–3) were used.

EEO culture and hormone treatment

EEOs were established as previously detailed in Luyckx et al. (2025b). Briefly, previously isolated and cryo-preserved epithelial glands were thawed and mechanically triturated into small fragments followed by seeding in Matrigel (growth factor reduced, Corning, USA). EEOs were cultured in organoid medium (Luyckx et al., 2025b) at +37°C with 5% CO₂ and split every 7–10 days. Organoids were split and cultured either in no hormones (baseline) for 7–10 days, or hormone treated by culture for 6 days in 10 nM β-estradiol (Sigma-Aldrich) (E2 treatment), or culture for 2 days in 10 nM E2 followed by 4 days in a combination of 10 nM E2, 1 µM P4, 0.25 mM cAMP, and 10 µM XAV-939 (Tocris) (EPCX treatment). Organoids from a low passage (P3–5) were utilized for experiments.

In vitro assays

Seahorse mitochondrial stress assay

XFp (for ESCs) and XF Pro (for EEOs) analyzers (Seahorse Bioscience, MA, USA) were used to assess mitochondrial respiration. About 40 000 or 20 000 ESCs per well were seeded on XFp Cell Culture Microplates (Agilent) for non-decidualized or decidualized conditions, respectively (as experimentally determined), and were incubated for 24 h (+37°C, 5% CO₂) prior to the assay. For EEOs, dissociated EEOs were plated on Seahorse XFe96/XF Pro Cell Culture Microplates (Agilent) in 3 µl drops containing 70% Matrigel/30% DMEM/F12 and cultured for 5–10 days in organoid medium to reach the desired confluency. Assays were performed according to the standard protocol using a Seahorse XF Cell Mito Stress Test kit. Briefly, cells were washed twice with un-buffered cell media (XF DMEM medium pH 7.4, Agilent), supplied with 10 mM glucose, 1 mM pyruvate, and 4 mM glutamine) followed by 1 h incubation in a non-CO₂ incubator. Next, the Mito Stress Test was performed with 1 µM carbonyl cyanide-p-trifluoromethoxyphenylhydrazone (FCCP) and 1.5 µM oligomycin (concentrations experimentally determined). For EEOs, the number of measurement cycles per injection was adjusted as necessary to observe a plateau in oxygen consumption rates (OCRs). For ESCs, the results were normalized to 10 000 cells per well via live cell counting. For EEOs, the results were normalized to EEO area per well, measured from images taken with an Axio Zoom.V16 fluorescence stereo microscope (Zeiss, Germany), using ImageJ Fiji software. For data presentation, some datapoints were merged to provide a consistent view. For data analysis, all datapoints and original scaling were utilized.

CellROX, ATP, and TMRE assays

Upon 90% confluency of ESCs, oxidative stress was measured using the CellROX assay (Invitrogen, CA, USA), ATP production was measured using the ATPlite assay (Perkin Elmer, CT, USA) and mitochondrial membrane potential was measured using a TMRE-Mitochondrial Membrane Potential Assay kit (Abcam, UK), all according to manufacturer instructions.

Statistics

All data are presented as mean±SD or as median and range (25th, 75th percentile). Data analysis was performed using GraphPad Prism 10 (GraphPad Software, MA, USA). A minimum of three different BMI-matched donors was used per group. For in vitro assays and RT–qPCR, the geometric mean of three technical replicates per sample was used. The RT–qPCR data were tested for outliers using the following method: datapoints < Q1 − (1.5 × interquartile range (IQR)) or datapoints > Q3 + (1.5 × IQR) were considered outliers, where the IQR = Q3 − Q1. Normality was assessed using Shapiro–Wilk’s normality test. Differences between two groups were determined by Student’s t-test or the Mann–Whitney test. Differences between more than two groups were determined by one-way ANOVA or the Kruskal–Wallis test. Differences were considered statistically significant at P < 0.05.

Results

Distinct endometrial transcriptome in PMOS is modulated by cycle phase

To study the cycle phase-specific endometrial transcriptome of women with PMOS, endometrial biopsies were collected from age- and BMI-matched women with PMOS and non-PMOS women (Ctrls) across different phases of the menstrual cycle (i.e. PE, ESE, MSE, and LSE) relying on occasional ovulation in PMOS and regular ovulation in controls, followed by RNA-seq analysis (Fig. 1a). Compared to non-PMOS women, women with PMOS had increased serum AMH and LH levels and increased LH/FSH ratios and were more androgenic (increased serum testosterone, FAI, and Ferriman–Gallwey scores) (Table 1). P4 levels did not differ between the groups (data not shown).

Figure 1.

Distinct endometrial transcriptomes in PMOS and across the menstrual cycle. (a) Study setup. Endometrial biopsies were obtained from women with or without polyendocrine metabolic ovarian syndrome (PMOS) across four different menstrual cycle phases and from different BMI groups followed by bulk RNA sequencing or isolation of endometrial stromal cells (ESCs) and epithelial cells (EECs) for culture and in vitro assays to evaluate mitochondrial function. Created in BioRender. Arffman, R. (2026) https://BioRender.com/bt9yb94. (b) Heatmap depicting supervised clustering of normalized gene expression levels of top 10 differentially expressed genes (DEGs) per group and per cycle phase. Red, high expression; blue, low expression. (c) Uniform manifold approximation and projection (UMAP) unsupervised clustering analysis of samples, color-coded by cycle phase (upper panel) or by PMOS status (lower panel). Black line represents the phase transition of PE and ESE to MSE and LSE. Each datapoint represents a biological sample. (d–g) Top 10 gene set enrichment analysis (GSEA) pathways in PMOS versus control samples per cycle phase: PE (d), ESE (e), MSE (f), and LSE (g). N = 8–10 individual donors per phase. Ctrl, control; PE, proliferative phase; ESE/MSE/LSE, early-, mid-, and late-secretory phase.

To provide an unbiased perspective on the RNA-seq data, first, a supervised clustering analysis was performed. The resultant heatmap demonstrated that, as expected, the phase of the menstrual cycle rather than PMOS status was the predominant determinant of the endometrial transcriptomic landscape, with relatively minimal variation observed within samples corresponding to the same phase (Fig. 1b). In line with this, differences between the PMOS and non-PMOS cases also seemed minor in the visualization of sample distribution using UMAP (Fig. 1c). Furthermore, transcriptomic similarity was evident between the PE and ESE phases, as well as between the MSE and LSE phases, indicated by common faint red regions in the heatmaps (Fig. 1b), and this was confirmed by the clustering of PE-ESE and MSE-LSE in the UMAP (Fig. 1c).

Subsequent GSEA, however, revealed distinct cycle phase-specific differences between PMOS and Ctrl endometria (Fig. 1d–g, Supplementary Data File S1). During PE, PMOS endometria showed an upregulation of pathways related to predominantly G2/M cell cycle phase transition as well as pathways related to collagen and ECM degradation (along with upregulation of ADAM9), suggesting increased proliferation in PMOS PE phase endometrium. Downregulation was seen in epigenetic regulatory factor methyl CpG binding protein 2 (MECP2) regulation (Fig. 1d). Interestingly, the largest number of significantly enriched pathways between the study groups was found in the ESE, where several pathways relating to the respiratory electron transport chain were upregulated in PMOS (along with upregulation of NDUFA12 and COX7B), suggesting that mitochondrial function in PMOS endometrium is altered during the differentiation process toward preparation for implantation (Fig. 1e). In the ESE phase, only two pathways were downregulated in PMOS: one related to the small GTPase Rab protein regulation of intracellular membrane trafficking of organelles and vesicles, and the other to neurotransmitter receptors and postsynaptic signal transmission (Fig. 1e). In the MSE phase in PMOS, an upregulation of several epigenetic regulation- (along with MTA2 upregulation) and protein metabolism-related pathways were noted, as well as pathways related to RNA transcription, the latter suggesting sustained proliferation. Additionally, in the MSE GSEA analysis, pathways related to unfolded protein response (UPR) and endoplasmic reticulum (ER) stress, with genes such as SERP1, CALR, and AMFR (Supplementary Data File S1), were upregulated, while lipid metabolism-related pathways were downregulated. Of note, the pathway ‘Metabolism of lipids’—including mitochondrial membrane proteins AGK and AGPAT5—was also downregulated in PMOS (Fig. 1f), possibly linking with dysregulated mitochondrial function. Lastly, PMOS LSE endometrium showed only limited differential pathway enrichment, including hemostasis- and platelet regulation-related pathways (along with increased PROCR expression). On the other hand, pathways related to estrogen-dependent signaling (along with increased CREBBP expression), as well as RNA polymerase function were downregulated compared to Ctrl LSE samples (Fig. 1g). Of note, cellular deconvolution analysis revealed broadly comparable estimated cell-type compositions between PMOS and control endometrium in each phase, with only a limited number of cell populations showing significant differences. For example, smooth muscle cells and several types of immune cells were significantly enriched in PMOS in the ESE (Supplementary Fig. S1a–d and Supplementary Data File S2).

Lastly, RT–qPCR analysis of selected genes (PMEPA1, ESR1, VCAN, and TIMP3) across different menstrual cycle phases confirmed differential expression patterns consistent with those observed in the sequencing data, thereby supporting the validity of the transcriptomic findings (Supplementary Fig. S2a). Collectively, while the cycle phase was the primary determinant of sample clustering, several phase-specific alterations were evident in PMOS endometria compared to controls.

Trajectory mapping uncovers cycle-specific differences in gene expression in PMOS endometria

Given the strong alignment of gene expression with cycle phase, a phase-specific comparison was conducted between control and PMOS groups by incorporating the phase-dependent transcriptional baseline. Genes demonstrating a strong association with the cycle phase and showing group-specific differences were identified using trajectory modeling with gene-specific GAMs to test for divergence between groups.

Overall, the number of differently expressed genes (DEGs) was highest in the PE (72 genes) and ESE (92 genes) phases (Fig. 2a, Supplementary Data File S3). The MSE (53 genes) and LSE (42 genes) phases showed the smallest number of DEGs. The PMOS samples showed clear hierarchical clustering in the PE, ESE, and MSE phases in the heatmap. The LSE phase presented weak hierarchical clustering of Ctrl and PMOS groups, suggesting that differences between Ctrl and PMOS LSE transcriptomes were minor (Fig. 2a).

Figure 2.

Differentially expressed genes (DEGs) are related to cell metabolism and mitochondrial function. (a) Heatmaps depicting normalized gene expression levels of all identified DEGs per phase after trajectory modeling. Each column presents a control (C, blue) or PMOS (P, red) sample clustered hierarchically to highlight expression patterns. Each row corresponds to an individual gene; red, high expression; blue, low expression. Grey scale denotes log2(fold change), white, low fold change; black, high fold change. Genes of interest in bold. (b) Ridgeline plot of DEGs (rug lines depicting individual genes) related to mitochondria across cycle phases. X-axis denotes gene counts; Y-axis depicts cumulated gene expression of all samples. (c) Average expression levels of differentially expressed mitochondria-related genes in endometrial samples from women with and without PMOS across menstrual cycle phases. Blue = Ctrl; Red = PMOS. N = 8–10 individual donors per phase. Ctrl, control; PMOS, polyendocrine metabolic ovarian syndrome; PE, proliferative phase; ESE/MSE/LSE, early-, mid-, and late-secretory phase.

In agreement with GSEA results (see Fig. 1d–g), DEGs in the PE phase of PMOS endometria were predominantly associated with RNA and DNA transcription and chromatin remodeling, including increased expression of the protein phosphatase gene RPAP2 that is involved in small nuclear RNA transcription and the multifunctional non-histone chromosomal protein gene HMGB2, involved in nuclear chromatin remodeling (Fig. 2a).

In the ESE phase, a crucial turning point for the endometrium to transfer from proliferation to secretion, trajectory modeling distinguished several DEGs in the PMOS group, including those implicated in insulin signaling (upregulated IRS1 and downregulated GLIS3) and ECM regulation (increased ITGA1 and MIA3, implicated in collagen secretion (Wilson et al., 2011) and decreased MMP16). Of note was the finding of increased P4 receptor (PGR) expression in PMOS in the ESE phase (Fig. 2a), which is consistent with available literature (Hu et al., 2018).

Given that the embryo makes contact with maternal endometrium during the MSE phase, interestingly, DEGs here were linked to epithelial cell–cell adhesion, with reduced expression of PERP essential for adhesion junction formation and epithelial cell integrity (Ihrie et al., 2005), DDR2 (collagen-activated receptor involved in communication with the ECM; Mohamed et al., 2022) and CDH1, coding for E-cadherin. Notably, several genes associated with cellular and oxidative stress, such as PRKD3 and OXR1, were reduced, indicating an altered oxidative stress response in MSE PMOS endometria (Fig. 2a).

Finally, during the LSE phase, changes in genes related to cellular stress and apoptosis were prominent. These included increased expression of epithelial–mesenchymal transition contributor GGCT (Xu et al., 2020) and angiogenesis regulator BASP1 (Khajavi et al., 2021), and reduced expression of GPX3 (implicated in the defense against oxidative stress and tumor suppression; Falck et al., 2010), SLK, KLF6, and NUPR1, highlighting a disruption in cellular stress response and apoptotic regulation in PMOS endometria (Fig. 2a).

In line with DEGs related to metabolism and oxidative stress, a consistent finding across different phases of the menstrual cycle in PMOS was the differential expression of several genes associated with mitochondrial function. Notably, during the PE phase, mitochondria-related genes such as ACADM and MT-ND6 were upregulated in PMOS endometria, representing nuclear and mitochondrial genomes, respectively (Fig. 2a). Similarly, in the ESE phase, increased expression of genes including LYRM7 (required for mitochondrial complex III assembly; Alfattal et al., 2023) and mitochondrial MT-CO2 (mitochondrial cytochrome C oxidase II) was observed (Fig. 2a). During the MSE phase, the expression of mitochondrial MT-CO2 and nuclear NFU1 and SUCLA2 was significantly decreased in the PMOS group, and in the LSE phase, mitochondrial markers such as SOD2 and TFAM were similarly downregulated, both representing nuclear genome-encoded mitochondrial proteins (Fig. 2a).

These findings were further investigated by an in-depth comparison of mitochondria-related gene expression between PMOS and Ctrl endometria per cycle phase. Mitochondria-related gene expression per cycle phase (list of targets in Supplementary Data File S5) was visualized using a ridgeline plot, which indicated that differences in gene expression were most obvious in the PE and ESE phases (as seen from the divergent shape of the ridgeline plot, Fig. 2b). Similarly, plotting of gene expression of significant mitochondria-related DEGs throughout the cycle phase revealed that most differences between PMOS and Ctrl endometria were found in the ESE phase, including significant upregulation of MT-CO1, MT-CO2, MT-RNR1, COX7B, and UQCRQ in PMOS versus Ctrl samples (Fig. 2c and Supplementary Fig. S2b). In contrast, UQCRB, SOD2, and TFAM were significantly downregulated in PMOS versus control samples in the LSE (Fig. 2c).

Together with GSEA data (Fig. 1e), these findings collectively highlight a phase-specific dysregulation of mitochondrial gene expression, suggesting abnormal mitochondrial function in the pathophysiology of PMOS.

Distinct transcriptomic signatures associated with HA and BMI in PMOS

HA and metabolic dysfunction are key features of PMOS. Therefore, subgroup analyses were performed to dissect effects of androgen excess and high BMI on the PMOS endometrial transcriptome. First, PMOS samples were stratified based on clinical HA (FG cut-off value: 5; N = 5 HA and 5 non-HA samples) to assess the impact of androgen excess (Supplementary Table S1, Fig. 3a, Supplementary Fig. S3). To visualize whether previously identified PMOS-associated transcriptional changes differed by HA status, the genes identified as differentially expressed in PMOS endometria during the ESE phase (Fig. 2a) were plotted in a heatmap. Distinct clustering between controls, non-HA PMOS, and HA PMOS samples was observed (Fig. 3a). In the previously detailed cycle phase analysis, the ESE phase exhibited the highest number of DEGs, including several mitochondria-related DEGs. Here, a differential expression analysis comparing HA and non-HA PMOS samples in the ESE phase (Supplementary Data File S5) identified further downregulation of several mitochondrial genes (NADK2, NDUFV3, DCAKD, MFN1, TOMM22, ALDH6A1, SFXN1) in HA PMOS, raising the possibility that mitochondrial alterations contribute to PMOS pathophysiology in a hyperandrogenic subtype-specific manner (Supplementary Data File S5). Additionally, relating to oxidative stress defense, the catalase antioxidant gene CAT was downregulated in comparison between HA and non-HA PMOS groups.

Figure 3.

Subgroup analysis of endometrial transcriptome. (a) Heatmap depicting normalized expression of previously identified differentially expressed genes (DEGs) during the early-secretory phase (ESE) in control (ctrl, N = 10), hyperandrogenic (HA, N = 5), and non-hyperandrogenic (non-HA, N = 5) PMOS samples. (b) Heatmap depicting normalized expression of previously identified DEGs in lean and overweight Ctrl and PMOS samples during the ESE. Red, high expression; blue, low expression. N = 3 (lean Ctrl), 7 (overweight Ctrl), 3 (lean PMOS), 7 (overweight PMOS) Ctrl, control; PMOS, polyendocrine metabolic ovarian syndrome.

Next, samples were stratified by BMI (lean, BMI < 25 kg/m² (N = 3 Ctrl and 3 PMOS samples); overweight, BMI ≥25 kg/m² (N = 7 Ctrl and 7 PMOS samples); Supplementary Table S2, Fig. 3b, Supplementary Fig. S4) to assess its impact on gene expression (Fig. 3b). As with the HA subgroup analysis, previously identified DEGs from the transcriptomic analysis of PMOS endometria in the ESE phase (Fig. 2a) were visualized in a heatmap for the BMI subgroup (Fig. 3b). Differential clustering between Ctrl and PMOS samples was seen (Fig. 3b). A differential expression analysis for the BMI-defined subgroups showed that transcriptional differences were evident in both the lean and overweight PMOS groups. Here, mitochondria-related DEGs, including COX1 and COX2, were upregulated in lean PMOS compared to lean controls (Supplementary Data File S5). In addition, the lean PMOS group exhibited increased expression of genes associated with cellular stress responses (TXNIP, TXN), as well as PGR. A similar pattern was observed in the overweight subgroup, where mitochondrial-related genes (COX1, COX3, CYTB5A, and NDL4) were differentially expressed between control and PMOS samples. Furthermore, the overweight PMOS group showed upregulation of the estrogen receptor gene ESR1.

Overall, these observations suggest that the mitochondrial-associated transcriptional alterations are an inherent feature of PMOS endometrium, especially to the HA-phenotype.

Mitochondrial dysfunction in PMOS-derived ESCs upon decidualization

Given the aberrant expression of mitochondria-related genes in PMOS endometrium, mitochondrial morphology was analyzed in detail using TEM on MSE phase endometrial biopsies from women with and without PMOS. However, no significant abnormalities in terms of shape, membrane integrity, or cristae structure of the mitochondria were found in in vivo PMOS ESCs (Supplementary Fig. S2c). In addition, analysis of cultured ESCs revealed that, compared to non-decidualized (nDE) ESCs, decidualization led to an apparent reduction in mitochondrial elongation, consistent with previous findings (Ryu et al., 2020). Nevertheless, the morphological decidualization-induced alterations were comparable between control and PMOS ESCs (Supplementary Fig. S2d).

Next, the functional implications of alterations in cellular metabolism- and endometrial function-related gene expression in PMOS were investigated. To this end, ESCs were isolated from endometrial biopsies of age- and BMI-matched women with and without PMOS. Participants were classified as lean (lean control, L-Ctrl; lean PMOS, L-PMOS; BMI < 25 km/m²) or obese (obese control, O-Ctrl; obese PMOS, O-PMOS; BMI ≥30 kg/m²). Lean women with PMOS demonstrated higher AMH levels and were more androgenic compared to L-Ctrls, characterized by significantly elevated FG scores. In obese women with PMOS, serum AMH, testosterone, LH/FSH ratios, and FAI were significantly increased compared to the O-Ctrl group (Table 2). Overall, obese women presented with increased insulin resistance, as defined form increased fasting insulin and HOMA-IR scores, irrespective of PMOS status (Table 2).

ESCs were decidualized in vitro using P4 and cAMP (DE ESCs, see experimental design in Fig. 1a), and no difference in decidualization efficiency was found between PMOS and Ctrl groups, as assessed by the fold change in PRL8A2 and IGFBP-1 gene expression after decidualization (Supplementary Fig. S2e).

To assess whether PMOS affects mitochondrial function, mitochondrial performance of ESCs under conditions of cellular stress and increased energetic demand was assessed in live cells using the Seahorse Mito Stress Test assay. Mitochondrial respiration was unchanged in nDE PMOS ESCs in both lean and obese groups (Fig. 4a and b). In contrast, upon decidualization, differential respiration was observed in both BMI groups. L-PMOS DE ESCs showed significantly reduced proton leak compared to L-Ctrls and a non-significant yet noticeable trend toward lower OCR in all calculated values. In the obese group, PMOS caused a significant decrease in basal and maximal respiration (Fig. 4c and d), again with a slight trend toward lower OCR in most calculated values.

Figure 4.

Dysfunctional mitochondrial respiration in PMOS ESCs. (a, c) Overview of Seahorse Mito Stress Test assay on non-DE (a) or DE (c) ESCs in lean (upper panel) and obese (lower panel) groups. (b, d) Calculation of basal respiration (Basal Resp.), ATP production (ATP prod.), proton leak, maximal respiration (Max Resp.), spare respiratory capacity (Spare Resp. Cap.), and non-mitochondrial respiration (Non-Mito. Resp.) in non-DE (b) or DE (d) ESCs in lean (upper panel) and obese (lower panel) groups. Data normalized to 10 000 cells. N = 3 individual donors per BMI group. One-way ANOVA or non-parametric alternative. *P < 0.05; **P < 0.01. Data presented as mean ± SD. Ctrl, control; PMOS, polyendocrine metabolic ovarian syndrome; OCR, oxygen consumption rate; ESC, endometrial stromal cells; DE, decidualized; FCCP, carbonyl cyanide-4-(trifluoromethoxy)phenylhydrazone.

As a further evaluation of mitochondrial function, ATP content was measured using the ATPlite assay. Decidualization was found to have a major reducing effect on ATP content in both PMOS and Ctrl groups (Fig. 5a). Compared to controls, PMOS nDE ESCs exhibited reduced ATP levels. Stratification by BMI revealed that this reduction was specifically driven by lower ATP content in L-PMOS compared to L-Ctrl (Fig. 5a). Furthermore, both reactive oxygen species (ROS) production and mitochondrial membrane potential were significantly reduced in PMOS ESCs compared to controls in both decidualized and non-decidualized conditions (Fig. 5b and c). As with ATP content, BMI stratification revealed that these reductions were driven by the L-PMOS group in both conditions (Fig. 5b and c).

Figure 5.

Mitochondrial dysfunction in lean-derived PMOS ESCs. (a–c) ATP content (a), CellROX (ROX) fluorescence, an indicator of reactive oxygen species (ROS) production (b), and TMRE fluorescence, an indicator of mitochondrial membrane potential (c), all normalized to total RNA. Data shown with BMI groups pooled (top) and stratified (bottom). N = 3 individual donors per BMI group. One-way ANOVA or non-parametric alternative. *P < 0.05; ***P < 0.001. Data presented as mean ± SD. Ctrl, control; PMOS, polyendocrine metabolic ovarian syndrome; ESC, endometrial stromal cell; TMRE, tetramethylrhodamine ethyl ester; DE, decidualized.

In summary, while mitochondrial morphology was unaffected, PMOS endometrium exhibited dysregulated mitochondrial function regardless of BMI, but the impairment in ATP and ROS production and altered membrane potential seemed to be more profound in the L-PMOS group.

Lean-derived PMOS EECs have altered mitochondrial respiration

Consistent with the analysis of ESCs, we assessed the effects of PMOS and BMI on EEC mitochondrial function by generating EEOs and performing a Seahorse Mito stress test. EEOs were established and maintained following a previously described protocol (Luyckx et al., 2025b).

In the absence of hormonal treatment (i.e. baseline state), L-PMOS EEOs exhibited increased basal respiration and increased proton leak compared to BMI-matched control EEOs, whereas no significant differences were detected in the obese cohort (Fig. 6a and b).

Figure 6.

BMI-specific changes in EEO mitochondrial respiration. (a, c) Overview of Seahorse Mito Stress Test assay on EEOs at baseline (a) or after treatment with estrogen, progesterone, cAMP, and XAV-939 (EPCX) (c) in lean (upper panel) and obese (lower panel) cohorts. (b, d) Calculation of basal respiration (Basal Resp.), ATP production (ATP prod.), proton leak, maximal respiration (Max Resp.), spare respiratory capacity (Spare Resp. Cap.), and non-mitochondrial respiration (Non-Mito. Resp.) in EEOs at baseline (b) or after EPCX treatment (d) in lean (upper panel) and obese (lower panel) cohorts. Data normalized to total EEO area. N = 3 individual donors per BMI group. One-way ANOVA or non-parametric alternative. *P < 0.05; **P < 0.01; ***P < 0.001. Data presented as mean±SD. Ctrl, control; PMOS, polyendocrine metabolic ovarian syndrome; EEO, endometrial epithelial organoid; OCR, oxygen consumption rate; FCCP, carbonyl cyanide-4-(trifluoromethoxy)phenylhydrazone.

Given that EEOs can be pharmacologically driven toward a receptive-like state through exposure to E2, P4, cAMP, and the WNT pathway-inhibitor XAV-939 (EPCX treatment) (Wang et al., 2010), we next assessed mitochondrial function following this EPCX treatment. L-PMOS EEOs again displayed differences in mitochondrial respiration compared to L-Ctrl EEOs, characterized by increased basal respiration, ATP production, and proton leak, while no significant differences between PMOS and Ctrl EEOs were observed in the obese cohort (Fig. 6c and d).

Overall, mitochondrial respiration was altered in L-PMOS EEOs, with the differences becoming more pronounced following hormonal treatment.

Discussion

Previous studies of PMOS endometrium have been limited by suboptimal cohort selection and sample matching, particularly regarding BMI and menstrual cycle phase, factors which can significantly confound interpretation of the effects of PMOS on endometrial dysfunction. Therefore, in this study, we used a rare, well-characterized, BMI- and age-matched endometrial sample cohort from women with PMOS who exhibited occasional ovulations, alongside healthy controls, to investigate how PMOS and the typical characteristics (HA and overweight/obesity) affect the endometrial transcriptome and mitochondrial function in a cycle phase-dependent manner.

Transcriptome analysis revealed that sample clustering occurred primarily based on cycle phase rather than PMOS status. This can be explained by the highly dynamic nature of the endometrium, which undergoes profound transcriptomic and functional changes across the different phases of the menstrual cycle which can obscure the relatively smaller differences between control and PMOS groups.

In women with PMOS during the PE phase, we observed increased RNA transcription-related pathways and DEGs which, considered alongside increased ESR1 expression in ESE, could implicate either prolonged E2-stimulated proliferation or a deficiency to blunt E2-driven proliferation upon transition to the ESE phase (Matsuyama et al., 2024). Additionally, in PE phase, increased enrichment of pathways related to ECM organization was found. Given that ECM remodeling is essential for stromal regeneration after menses, but also during the transition from PE to ESE phase (Gargett et al., 2012; Winkler et al., 2024), these processes might be dysregulated in PMOS endometria. Via trajectory modelling, aimed at accounting for phase shifts or delays in the menstrual cycle, we also identified DEGs related to ECM during the ESE. Supporting this, ECM-related pathways have been found to be enriched in both epithelial and stromal cell populations in the PE phase in PMOS (Eriksson et al., 2025), and abnormal ECM composition with excessive collagen deposition have also been observed in the uteri of PMOS-like prenatally androgenized mice (Luyckx et al., 2025a).

Notably, the ESE phase showed the most DEGs between PMOS and control samples. This highlights how the changes seen in PMOS endometrium occur mostly during this sensitive, highly dynamic phase. Other DEGs in the ESE phase were related to insulin signaling (e.g. IRS1) and again ECM remodeling—both critical for endometrial decidualization and hormone signaling (Neff et al., 2020; Winkler et al., 2024). Of note, cell deconvolution analysis revealed no major differences in estimated cell-type composition between groups, suggesting that the observed transcriptional alterations were unlikely to be driven by differences in cellular composition. Interestingly, pseudo-time trajectory analysis also uncovered increased PGR expression in ESE PMOS endometria, which could support the previously established conclusion of P4 resistance in PMOS (Aghajanova et al., 2010; Savaris et al., 2011; Piltonen et al., 2015). Increased PGR expression could also reflect a temporal shift in the maturation of PMOS endometrium—a finding that may have been previously obscured when not employing pseudo-time trajectory analysis. Future studies that resolve the expression patterns of PGR isoforms across cell types and timepoints are needed to clarify the molecular contributions of P4 resistance to PMOS endometria pathophysiology. With regard to steroid hormones, ESR-related pathways were downregulated in LSE-phase PMOS endometria, again demonstrating altered steroid hormone signaling in PMOS endometrium.

One of the key findings of the transcriptomic analysis was the differential expression of genes related to the respiratory chain and to mitochondria. Specifically, in the ESE phase, PMOS samples showed increased enrichment of pathways related to respiratory electron transport, demonstrated by differential expression of several mitochondria-related genes (e.g. upregulation of MT-CO2, NDUFA12, and COX7B). Mitochondrial function is critical for energy-intensive endometrial processes, including ESC decidualization (Ma et al., 2011; Dalla Torre et al., 2024). Mitochondrial impairment could therefore suggest a cycle-phase-specific vulnerability in PMOS, which could potentially disrupt key preparatory processes required for implantation. In addition, differential expression of oxidative stress genes (OXR1 down in MSE; GGCT up in LSE) may disrupt redox homeostasis, exacerbating mitochondrial dysfunction, potentially amplified by the hyper-estrogenic, ROS-promoting PMOS environment (Chou et al., 2020; Khatun et al., 2021). This environment may also increase the risk of endometrial hyperplasia and cancer noted by several studies (Holm et al., 2012; Johnson et al., 2023) and trigger inflammatory processes, a common feature in PMOS that has been linked to pregnancy complications and prematurity (Palomba et al., 2014; Arffman et al., 2019).

The MSE phase showed upregulation of ER stress- and UPR-related pathways. In PMOS samples, the ER chaperone gene CALR, encoding calreticulin, was upregulated, whereas the key calcium-signaling genes CALM1 and CALM2, encoding calmodulin proteins as well as the sulfate transport protein coding SLC26A2, were downregulated. Calreticulin and calmodulins play central roles in calcium homeostasis and in ER–mitochondrial communication via Ca²⁺ signaling. These transcriptomic alterations may therefore suggest a mechanistic connection between ER stress and mitochondrial stress in PMOS endometrium, noting previous reports of both dysfunctions in granulosa cells of women with PMOS (Zhao et al., 2015; Bouman et al., 2011). Additionally, seen with trajectory modelling, many DEGs in MSE were associated with epithelial cell adhesion (e.g. PERP, CHD1). This finding aligns with (Eriksson et al. (2025) who suggest a central role for integrin-mediated signaling in PMOS endometrial pathogenesis in their transcriptomic study. During the WOI, luminal endometrial epithelium upregulates adhesion molecules to facilitate embryo attachment (Singh and Aplin, 2009; Zhou et al., 2020). Dysregulation of these genes in PMOS may therefore impair endometrial receptivity and diminish the likelihood of successful embryo implantation.

The LSE phase pseudo-time trajectory analysis yielded the fewest DEGs, indicating relatively limited transcriptional divergence between PMOS and control endometrium during this phase. Still, reduced expression of PAEP (Progestagen-Associated Endometrial Protein)—a key regulator of endometrial receptivity and implantation during the WOI—was observed in LSE PMOS samples, which is consistent with its downregulation in PMOS EEOs, further supporting impaired epithelial differentiation and receptivity in the PMOS endometrium (Luyckx et al., 2025b).

Further subgroup analyses of the effects of HA and BMI on PMOS endometrium showed that within the PMOS group, significant differences could be seen between those with clinical HA compared to those without. Interestingly, several mitochondrial DEGs, such as NADK2, a NAD kinase gene, and respiratory complex I subunit coding gene NDUFV3 were identified in the ESE phase. These results suggest that cellular metabolism and mitochondrial function in PMOS may vary according to the degree of HA, although further studies are needed to establish the biological significance of this relationship. In contrast, when analyzing overweight- and lean-BMI PMOS subgroups, several mitochondrial genes were found to be significantly upregulated in both BMI-groups in PMOS compared to the control group, suggesting that altered mitochondrial gene expression is a consistent feature of PMOS across BMI-defined subgroups.

To assess the implications of differential mitochondria- and oxidative stress-related gene expression revealed in the transcriptomic analyses, functional studies using ESCs and EEOs were employed. Mitochondrial respiration was impaired in decidualized PMOS ESCs, regardless of BMI, whereas non-decidualized cells did not exhibit such deficits, suggesting that the energetically demanding decidual state required for embryo implantation represents a particular metabolic vulnerability in PMOS. Further analyses of ATP content, ROS production, and mitochondrial membrane potential, however, demonstrated that mitochondria-related dysfunction extends beyond the decidualized state, as these parameters were also impaired in non-decidualized ESCs, with L-PMOS ESCs exhibiting the most pronounced deficits. Importantly, the differences were not attributable to differences in decidualization capacity between the control and PMOS groups, as assessed with RT–qPCR analysis of IGFBP-1 and PRL8A2. Studies on decidualization report varying effects on endometrial cell respiration (Ryu et al., 2020; Dalla Torre et al., 2024). Noting the heterogeneous nature of PMOS, differences in decidualization capacity of PMOS-derived endometrial cells should consider the ovulatory status of the women, but also note the severity of their phenotypic expression, as was done in the current study. Overall, L-PMOS ESCs appeared to have the poorest performance across all in vitro tests. However, it should be noted that high BMI may in part have a negative obscuring effect on cell metabolism, such that with BMI-matched controls the effects of obesity may conceal the underlying PMOS-specific differences between obese control and PMOS cases, leading to these differences being more apparent in the lean cohort. Still, the findings of reduced mitochondrial respiration and align with data from PMOS-derived induced pluripotent stem cells (iPSCs) (Min et al., 2018; Khatun et al., 2021), supporting mitochondrial dysfunction as an intrinsic feature of PMOS.

When mitochondrial respiration in EEOs was assessed, PMOS-related differences in mitochondrial function were evident in both basal and EPCX-treated conditions, but only in the lean cohort. Interestingly, the increased proton leak, as identified in L-POCS EEOs, has been linked to increased mitochondrial ROS production, since proton leak is hypothesized to act as a regulatory mechanism to balance the mitochondrial transmembrane proton gradient in order to reduce oxidative damage (Divakaruni and Brand, 2011). Another possible contribution to the observed increase in proton leak could be a difference in the function or composition of mitochondrial transport proteins (Jastroch et al., 2010). Again, these results underline the importance of BMI as a determining factor of the endometrial phenotype, which is in line with the observation of reproductive and metabolic subtypes identified by genome-wide association studies (Dapas et al., 2020; Liu et al., 2022). Still, the apparent differences in mitochondrial respiration between lean and obese cohorts should be interpreted cautiously, as direct quantitative comparison was not feasible in the current experimental setup.

Collectively, our transcriptomic analyses revealed cycle phase-specific differences in ECM, steroid hormone regulation, and endometrial adhesion molecule expression (Fig. 7). A key finding arising from the transcriptomic analysis was the differential expression of mitochondrial genes, alongside differential pathway enrichment related to ion transport, indicating dysfunction of the respiratory electron transport chain in PMOS endometrium. Notably, the apparent discrepancy between transcriptional upregulation of mitochondrial pathways and impaired mitochondrial function possibly reflects a compensatory response, whereby defective mitochondrial bioenergetics drive increased expression of mitochondria-related genes. A similar finding has been reported in PMOS-derived iPSCs, supporting this interpretation (Min et al., 2018). Dysregulated mitochondrial gene expression seemed to be more pronounced in the hyperandrogenic PMOS phenotype, but at the endometrial level, it was also shown to be cell type specific, with differences seen between ESCs and EECs, and with BMI-specific defects underscoring the independent role of PMOS in this defect (Fig. 7).

Figure 7.

Graphical summary of study findings showing cycle phase-specific transcriptomic changes in oligo-ovulatory PMOS endometrium and BMI-dependent effects on mitochondrial function and cellular respiration in endometrial stromal cells and endometrial epithelial organoids.

Notable strengths of this study include the stratification by hyperandrogenic and metabolic subtypes, together with precise cycle phase timing, which enabled robust transcriptomic comparisons across the menstrual cycle, and inclusion of both ESC and EEC cell types. In addition, secretory-phase samples were obtained after long-term tracking of occasional ovulations in PMOS, enhancing the translational relevance of findings, also taking into consideration the high number of spontaneous pregnancies in women with PMOS (Persson et al., 2019). Limitations of this study included the use of bulk RNA-seq, which may have obscured cell type-specific transcriptional changes. In this context, immune cell populations, despite constituting a substantial component of the endometrium, were not included in the present analyses. Second, given the exploratory nature of the subgroup transcriptomic analyses, further functional studies are required to determine the biological significance of the observed transcriptomic differences associated with HA. With regard to BMI, the live-cell metabolic analyses performed using Seahorse assays did not permit direct comparisons between lean- and obese-derived samples due to separate assay runs and the lack of precise normalization methods when using 3D models. Finally, the in vitro assays were intended as an initial functional characterization and therefore involved relatively small sample sizes, although significant group differences were nevertheless detected.

Overall, this study represents the first comprehensive assessment of mitochondrial function across different cycle phases in PMOS, incorporating HA and BMI as covariates. This study also identifies BMI-dependent, cell type-specific mitochondrial impairments in PMOS endometrium, providing a basis for future work on the mechanisms of endometrial dysfunction in PMOS.

Supplementary data

Supplementary data are available at Human Reproduction online.

Data availability

Due to General Data Protection Regulation (GDPR) restrictions, sequencing data are available from the corresponding author (T.T.P.) upon request and will be shared following the execution of a signed data transfer agreement. All other data underlying this article are available in the article and in its online supplementary material.

Authors’ roles

TTP, RKA, and JK designed the study. TTP and JK collected the endometrial samples, and RNA-seq was performed by AV-M, IR, and AS. In vitro experiments were performed by LL, MM, US, and RKA, and data were analyzed by VI, LL, MM, and SL. LL and MM wrote the paper under supervision of TTP, RKA, and US. All authors critically reviewed and approved the paper.

Funding

L.L. was supported by a PhD Fellowship from Horizon 2020 Marie-Curie MATER Innovative Training Network (H2020-MSCA-ITN-2018, T.T.P.); University of Oulu Scholarship Foundation Grant, University of Oulu Profi6 Fibrobesity Programme (Research Council of Finland 336449); and Novo Nordisk Foundation (NNF21OC0070372, T.T.P.). L.L. and M.M. were supported by Jusélius Foundation. U.S. was supported by the Novo Nordisk Foundation and Research Council of Finland (321763, T.T.P.). A.V.-M. and I.R. were supported by the Estonian Research Council (PSG608). T.T.P. and R.K.A. were supported by the Research Council of Finland (315921, 321763, T.T.P.; 336449, R.KA.). A.S. was supported by the Estonian Research Council (grant no. PRG1076); Swedish Research Council (grant no. 2024-02530); Novo Nordisk Foundation (grant no. NNF24OC0092384); and Horizon Europe grant NESTOR (grant no. 101120075).

Disclosures

The authors declare no competing interests.