Gene and non-coding RNA expression in human cumulus cells in polycystic ovary syndrome: A systematic review

Abstract

Objective

This systematic review aimed to synthesize and critically appraise human studies on coding and non-coding RNA expression in cumulus cells (CCs) from women with Polycystic Ovary Syndrome (PCOS), to map consistently dysregulated pathways, evaluate links to oocyte competence, and identify key methodological limitations.

Methods

A systematic search of five electronic databases (MEDLINE/PubMed, Embase, Web of Science, Scopus, Cochrane CENTRAL) was conducted from inception to October 25, 2025. Peer-reviewed human studies comparing CC RNA expression in PCOS versus controls, with explicit cumulus isolation and primary RNA data, were included. Study selection, quality assessment, and data extraction followed pre-defined analytic prisms focusing on cell-type specificity, oocyte maturation stage, stimulation protocol, and clinical phenotype. A qualitative synthesis was performed due to substantial clinical and methodological heterogeneity.

Results

Thirty-three studies were included. The CC transcriptome in PCOS is characterized by a coherent phenotype: pathway-level enrichment of EGFR-MAPK and PI3K-AKT/insulin signaling, an imbalanced TGF-β/oocyte-somatic cell axis, and blunted induction of cumulus expansion/extracellular matrix (ECM) genes (e.g., TNFAIP6, PTX3). Non-coding RNA profiles (miRNAs, lncRNAs, circRNAs) corroborate these findings, targeting the same pathways. These signatures are strongly modified by oocyte maturation stage (largest differences at Germinal Vesicle stage), stimulation protocols, and clinical phenotypes (e.g., insulin resistance, obesity).

Conclusion

The cumulus cell phenotype in PCOS reflects a rewired signaling environment that converges on defective matrix assembly, providing a molecular framework for impaired oocyte competence. Future research requires stage-aware, phenotype-stratified designs with standardized reporting to translate these findings into robust biomarkers for assisted reproduction.

1. Introduction

Polycystic ovary syndrome (PCOS) is a heterogeneous endocrine–metabolic condition that disrupts ovarian function and reproductive outcomes through multi-system perturbations [1,2]. Within the follicle, the cumulus–oocyte complex orchestrates oocyte maturation via tightly regulated metabolic support, paracrine signaling, gap-junctional communication, and extracellular matrix (ECM) assembly contributed by cumulus cells [2,3]. The peri-ovulatory cascade integrates LH/hCG-induced EGF-ligand signaling and oocyte-secreted factors (notably GDF9/BMP15), driving EGFR–MAPK and SMAD pathways that culminate in cumulus expansion and meiotic resumption [3,4].

Because these processes are transcriptionally encoded, cumulus-cell gene expression provides a window into oocyte competence and assisted reproductive technology (ART) outcomes in health and in PCOS [2,5]. Across platforms—from targeted qPCR and microarrays to bulk and single-cell RNA-seq—studies in human PCOS cohorts report dysregulation of EGFR–MAPK, PI3K–AKT/insulin, and TGF-β/oocyte–cumulus signaling, alongside altered expression of core cumulus-expansion markers such as HAS2, PTGS2/COX-2, TNFAIP6, and PTX3 [[6], [7], [8], [9]].

Mitochondrial oxidative phosphorylation and broader bioenergetic programs in cumulus cells are frequently perturbed in PCOS, although directionality varies across cohorts and contexts [7,10]. ECM and cell-adhesion gene networks show broad repression in PCOS cumulus cells, with specific differences linked to insulin-resistant versus insulin-sensitive phenotypes [11]. Findings for gap-junction components are mixed, with single-cell analyses identifying stage-dependent differences while bulk datasets sometimes report stability in key connexins [2,6,7]. Additional layers of cellular stress and growth control—including ER stress/unfolded-protein-response signatures, increased ribosomal RNA content, and cell-cycle modules—have also been documented in PCOS cumulus or corona radiata cells [[12], [13], [14], [15]].

Non-coding RNA studies reveal extensive remodeling of regulatory networks in PCOS cumulus cells, with altered miRNA profiles implicating pathways such as MAPK, Wnt, meiosis, and steroidogenesis [16,17].

Cumulus-cell long non-coding RNAs are aberrantly expressed in PCOS and have been integrated into network analyses that suggest potential modulatory roles in follicular signaling [18,19]. Follicular-fluid exosomal circRNA changes correlate with cumulus-cell aromatase regulation via miRNA–mRNA axes, linking the extracellular milieu to cumulus transcriptional control in PCOS [18]. In the in-vitro-maturation setting, specific miRNAs and their target genes in cumulus cells associate with oocyte maturation success in PCOS, underscoring functional relevance [20].

Interpretation across studies is complicated by strong dependence on oocyte nuclear maturation stage (GV/MI/MII), stimulation/trigger protocols, and timing, which can mask or mimic disease effects if not carefully matched [4,6,21]. Clinical phenotypes—including obesity and insulin resistance—further shape cumulus-cell expression programs and must be accounted for in comparative analyses [11,22]. Moreover, subcompartment sampling (e.g., corona radiata versus outer cumulus) and occasional ex vivo manipulation introduce additional sources of variability that can limit generalizability [12,23,24].

These methodological and biological sources of heterogeneity motivate a systematic synthesis focused specifically on human cumulus cells in PCOS, distinct from mural granulosa signals that are often conflated in the literature [2,9].

Accordingly, this systematic review collates and critically appraises human studies of coding and non-coding RNA expression in cumulus cells in PCOS, with explicit attention to maturation stage, stimulation context, and metabolic phenotype [1,7,9,11,21,22,25]. Our goals are to map recurrently dysregulated pathways, evaluate links to oocyte competence and ART outcomes, and identify evidentiary gaps to guide mechanistic studies and biomarker development. By consolidating cumulus-specific evidence, we aim to provide a coherent framework for interpreting molecular alterations within the PCOS follicular niche and to inform translational strategies in reproductive medicine.

2. Methods

2.1. Objectives and analytic prisms

We aimed to systematically identify, appraise, and synthesize human studies reporting coding and non-coding RNA expression in cumulus cells from women with polycystic ovary syndrome (PCOS) compared with appropriate controls, and to relate these molecular findings to follicular biology and assisted reproduction outcomes where available. To achieve this, we defined a priori analytic prisms—methodological and biological “lenses”—to structure eligibility, extraction, and synthesis: (i) cell-type specificity (cumulus cells or corona radiata cells derived from intact cumulus–oocyte complexes, minimizing mural granulosa contamination); (ii) maturation stage control (germinal vesicle [GV], metaphase I [MI], metaphase II [MII]); (iii) stimulation context (IVF/ICSI cycles, in vivo maturation, or in vitro maturation [IVM]; trigger protocol and timing); (iv) clinical phenotype (PCOS diagnostic criteria; stratification by insulin resistance, obesity, and hyperandrogenism where reported); and (v) molecular assay layer (mRNA transcriptome via microarray/bulk RNA-seq/single-cell RNA-seq/qPCR; small and long non-coding RNAs; epigenetic reporting frameworks such as MIAME/MINSEQE compliance). Reporting followed PRISMA 2020 guidelines; a protocol was written a priori and finalized before screening [26].

2.2. Review questions

We addressed the following questions [1]: In human PCOS, which coding and non-coding RNA signatures are consistently reported in cumulus cells relative to non-PCOS controls? [2] How do maturation stage, stimulation/trigger protocol, and clinical phenotype (e.g., insulin resistance, BMI) modify these cumulus-cell signatures? [3] Which dysregulated pathways recur across platforms (e.g., EGFR–MAPK, PI3K–AKT/insulin, TGF-β/oocyte–cumulus signaling, extracellular matrix/expansion, mitochondrial/oxidative phosphorylation, gap junctions, steroidogenesis, stress/inflammation)? [4] Which cumulus-cell transcripts or gene sets associate with oocyte/embryo competence and ART outcomes in PCOS? [5] What are the main risks of bias and reporting gaps specific to cumulus-cell expression studies in PCOS?

2.3. Eligibility criteria

We included peer-reviewed human studies that: (i) enrolled women with clinically defined PCOS (Rotterdam, NIH, or AE-PCOS Society criteria) and a non-PCOS comparator group or explicitly stratified PCOS phenotypes; (ii) isolated cumulus cells or corona radiata cells from cumulus–oocyte complexes with methods described (mechanical and/or enzymatic denudation), or clearly indicated per-oocyte cumulus sampling; (iii) generated primary RNA expression data on coding or non-coding RNA (microarray, RNA-seq, scRNA-seq, qPCR, small-RNA sequencing, lncRNA/circRNA assays), with sufficient methodological detail to interpret normalization and statistics; and (iv) were published in English.

We excluded studies that: (i) profiled mural granulosa/theca or whole ovarian tissue without a discrete cumulus-cell fraction, or used follicular aspirates labeled “granulosa cells” without cumulus specificity; (ii) reported only proteomics, metabolomics, secretome, or follicular-fluid biomarkers without primary cumulus-cell RNA measurements; (iii) were non-human (animal or in vitro “PCOS-like” models), case reports, editorials, narrative reviews without new primary data, or conference abstracts without full methods/results; and (iv) cultured cumulus cells extensively prior to RNA profiling unless culture was intrinsic to the experimental question and baseline (pre-exposure) cumulus data were also available. When studies mixed cumulus with mural granulosa cells or lacked PCOS diagnostic criteria, they were excluded or flagged as high risk of indirectness.

2.4. Information sources and search strategy

We searched MEDLINE (via PubMed), Embase, Web of Science Core Collection, Scopus, and the Cochrane Central Register of Controlled Trials (CENTRAL) from database inception to September 25, 2025. To reduce retrieval bias, we also performed backward and forward citation tracking of key studies in Google Scholar and screened reference lists of relevant reviews. Search strategies combined controlled vocabulary (MeSH/Emtree) and text words for PCOS, cumulus/corona radiata cells, and RNA expression modalities. No filters for study design were applied at the search stage; human-only and English-language limits were applied where supported by the interface. Full reproducible strategies are provided Table 1.

Table 1.

Search strategies used across electronic databases.

Database (interface)	Search string (exact, including field tags and operators)
PubMed (MEDLINE)	(“Polycystic Ovary Syndrome"[Mesh] OR “polycystic ovary syndrome"[tiab] OR PCOS[tiab]) AND (“Cumulus Cells"[Mesh] OR “cumulus cell*"[tiab] OR “corona radiata"[tiab] OR “cumulus granulosa"[tiab]) AND ((“Gene Expression Profiling"[Mesh] OR “RNA-Seq"[tiab] OR “RNA sequencing"[tiab] OR transcriptom*[tiab] OR microarray*[tiab] OR “single-cell"[tiab] OR scRNAseq[tiab] OR scRNA-seq[tiab] OR qPCR[tiab]) OR (miRNA*[tiab] OR “microRNA*"[tiab] OR lncRNA*[tiab] OR “long noncoding RNA*"[tiab] OR “circRNA*"[tiab])) Filters: Humans, English
Embase (Elsevier)	(‘polycystic ovary syndrome'/exp OR ‘polycystic ovary syndrome':ti,ab OR pcos:ti,ab) AND (‘cumulus cell'/exp OR ‘cumulus cell*':ti,ab OR ‘corona radiata':ti,ab OR ‘cumulus granulosa':ti,ab) AND ((‘gene expression profiling'/exp OR ‘rna sequencing'/exp OR transcriptom*:ti,ab OR microarray*:ti,ab OR ‘single cell':ti,ab OR scrnaseq:ti,ab OR ‘quantitative pcr':ti,ab) OR (mirna*:ti,ab OR ‘micro rna*':ti,ab OR lncRNA*:ti,ab OR ‘long noncod* rna*':ti,ab OR circrna*:ti,ab)) AND [english]/lim AND [humans]/lim
Web of Science (Core Collection)	TS=((“polycystic ovary syndrome” OR PCOS) AND (“cumulus cell*" OR “corona radiata” OR “cumulus granulosa”) AND ((transcriptom* OR microarray* OR “RNA-Seq” OR “single-cell” OR scRNAseq OR qPCR) OR (miRNA* OR microRNA* OR lncRNA* OR “long noncoding RNA*" OR circRNA*))) Refined by: Document types = Article
Scopus (Elsevier)	TITLE-ABS-KEY((“polycystic ovary syndrome” OR pcos) AND (“cumulus cell*" OR “corona radiata” OR “cumulus granulosa”) AND ((transcriptom* OR microarray* OR “rna-seq” OR “single-cell” OR scrnaseq OR qPCR) OR (mirna* OR microrna* OR lncRNA* OR “long noncoding rna*" OR circrna*))) AND (LIMIT-TO(DOCTYPE, “ar")) AND (LIMIT-TO(LANGUAGE, “English"))
Cochrane CENTRAL	(“polycystic ovary syndrome” OR PCOS) in Title Abstract Keyword AND (“cumulus” OR “corona radiata”) in Title Abstract Keyword AND (transcriptome OR “RNA-Seq” OR microarray OR “single-cell” OR scRNAseq OR miRNA OR lncRNA OR circRNA) in Title Abstract Keyword

We piloted the PubMed strategy, iteratively refined synonyms, and then translated to other databases using database-specific subject headings and syntax. No date limits were imposed. All records were exported and deduplicated prior to screening.

2.5. Study selection

Records were managed in a shared library and deduplicated using automated algorithms complemented by manual verification. Two reviewers independently screened titles and abstracts against eligibility criteria, blinded to each other's decisions; disagreements moved to full-text screening by default. Full texts were then assessed independently by the same reviewers, with a third reviewer resolving conflicts. Reasons for exclusion at full text were recorded (e.g., not cumulus-specific; no primary RNA data; non-human; inadequate PCOS definition; extensive culture without baseline cumulus data). When cell-type specificity was unclear, we contacted authors or examined methods/figures closely for isolation details; studies lacking verifiable cumulus sourcing were excluded or designated as indirect. A PRISMA flow diagram will summarize identification, screening, eligibility, and inclusion.

2.6. Quality assessment and risk of bias

Given the observational nature of most included studies and their molecular endpoints, we applied design-appropriate tools: the Joanna Briggs Institute (JBI) Critical Appraisal Checklist for Analytical Cross-Sectional Studies [27] for most case–control/cross-sectional IVF cohorts; the Newcastle–Ottawa Scale [28] (adapted for cross-sectional designs) for studies with matched control groups; and ROBINS-I [29] for any non-randomized interventional exposures embedded in molecular sampling (e.g., pharmacologic perturbations). For omics reporting quality, we assessed adherence to MIAME (for microarrays) and MINSEQE (for RNA sequencing), including sample processing, platform, normalization, batch handling, and data availability. We also implemented a domain-specific bias framework tailored to cumulus-cell studies: (i) cell-type misclassification/contamination; (ii) maturation-stage imbalance; (iii) stimulation/trigger timing heterogeneity; (iv) phenotype confounding (BMI, insulin resistance, hyperandrogenism); and (v) selective reporting of candidate genes without multiple-testing control. Two reviewers rated each domain; discrepancies were adjudicated by consensus.

2.7. Data extraction

A standardized, piloted extraction form captured: study metadata (year, country, design); PCOS diagnostic criteria and phenotype stratification; participant characteristics (age, BMI, metabolic indices, androgen status); ovarian stimulation/trigger protocol and timing; sample source and isolation/purity methods (cumulus vs corona radiata; per-oocyte pairing; exposure to hyaluronidase; ex vivo culture); oocyte nuclear maturation stage; RNA layer and platform (microarray, bulk RNA-seq, scRNA-seq, qPCR, small RNA-seq, lncRNA/circRNA assays); normalization/statistical pipelines and multiple-testing control; primary results (lists of differentially expressed genes/miRNAs/lncRNAs/circRNAs with directionality and effect estimates where reported); pathway and network analyses; associations with oocyte fertilization, embryo quality, blastocyst development, implantation, or pregnancy outcomes; and data-sharing/accession numbers. When only figures were provided, we digitized values where appropriate or contacted authors for clarification.

2.8. Data synthesis and rationale for not conducting a meta-analysis

We performed a structured narrative synthesis organized by our analytic prisms. Within each biological pathway, we mapped study-level directionality using prespecified gene sets (e.g., cumulus expansion: HAS2, TNFAIP6, PTX3, PTGS2; EGFR–MAPK; PI3K–AKT/insulin; TGF-β/oocyte–cumulus; extracellular matrix/adhesion; mitochondrial/OXPHOS; gap junctions; steroidogenesis; stress/inflammation) and summarized consistency across platforms and contexts. For non-coding RNAs, we synthesized convergent miRNA/lncRNA/circRNA signals and their validated targets in cumulus cells, noting whether analyses were discovery- or hypothesis-driven. We did not perform a quantitative meta-analysis because effect-size metrics for transcriptomic outcomes are not standardized across platforms (e.g., log2 fold changes from different normalization frameworks), variance estimates are inconsistently reported, gene-level outcomes are high-dimensional with strong dependence structures, and included studies exhibit substantial clinical and methodological heterogeneity (maturation stage, stimulation protocol/timing, phenotype stratification, and cumulus subcompartment). Moreover, outcome linkages (to fertilization/embryo development) are variably defined and not amenable to a single pooled model without extensive harmonization. Instead, we used vote counting by direction of effect and pathway-level concordance, and we highlight where harmonized reanalyses of raw data would enable future quantitative synthesis.

3. Results

3.1. Study Selection Process

The initial systematic search across five electronic databases on October 25, 2025, yielded a total of 2181 records. The contribution from each database was as follows: PubMed (n = 588), Embase (n = 712), Web of Science Core Collection (n = 415), Scopus (n = 411), and Cochrane CENTRAL (n = 55).

Following the collation of these records, 389 duplicate publications were identified and removed using a combination of automated and manual verification. This deduplication process resulted in 1792 unique records for the title and abstract screening phase.

During the title and abstract screening, 1683 records were excluded for failing to meet the eligibility criteria. The majority of exclusions were due to the study not involving human PCOS (n = 592), not focusing on cumulus/corona radiata cells (n = 574), being a review article, editorial, or conference abstract without primary data (n = 361), or not involving primary RNA expression analysis (n = 156).

This left 109 articles for which full-text copies were retrieved and subjected to a detailed eligibility assessment. During this full-text review, 76 articles were excluded for specific, critical reasons. The reasons for exclusion, with some articles meeting multiple criteria, were:

• •No discrete cumulus-cell analysis (e.g., profiling of mural granulosa, theca, or whole ovarian tissue): 31 articles.
• •No primary RNA expression data (e.g., proteomics, metabolomics, or theoretical reviews): 22 articles.
• •Inadequate PCOS definition or lack of a non-PCOS control group: 12 articles.
• •Non-human or in vitro PCOS-like model: 7 articles.
• •Extensive ex vivo culture of cumulus cells without baseline data: 4 articles.

Therefore, after this rigorous, multi-phase selection process, 33 studies [[1], [2], [3], [4], [5], [6], [7], [8], [9], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19], [20], [21], [22], [23], [24], [25],[30], [31], [32], [33], [34], [35], [36], [37]] were confirmed to satisfy all pre-defined inclusion criteria and were included in the final systematic review and qualitative synthesis. A PRISMA flow diagram visually summarizes this entire process (Fig. 1).

Fig. 1.

PRISMA flow diagram of the study selection process.

3.2. Results of quality assessment and risk of bias

Across the 33 studies [[1], [2], [3], [4], [5], [6], [7], [8], [9], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19], [20], [21], [22], [23], [24], [25],[30], [31], [32], [33], [34], [35], [36], [37]], overall methodological quality was mixed but interpretable with domain-specific safeguards. Cell-type specificity was acceptable in most primary datasets that explicitly isolated cumulus or corona radiata cells from intact cumulus–oocyte complexes [[5], [6], [7],11,12,[33], [34], [35]], whereas studies relying on cultured cumulus cells for pharmacologic perturbation were judged at higher risk for deviation from intended in vivo states under ROBINS-I and were interpreted as mechanistic rather than baseline phenotype evidence [23,24]. Maturation-stage imbalance and stimulation/trigger heterogeneity were the most pervasive sources of confounding: only a minority of studies explicitly matched or stratified by GV/MI/MII stage [6,21,25,33], and peri-retrieval timing or protocol differences (long GnRH-agonist vs antagonist; IVM vs IVF) frequently limited direct comparability across cohorts [4,12,20,33,35]. Phenotype confounding was common; few studies prospectively adjusted for or stratified by insulin resistance or BMI despite clear effects on extracellular matrix and insulin/AKT modules [11,22], and small sample sizes in several targeted reports (e.g., n = 3 per group in rRNA profiling) reduced precision and external validity [14]. At the analysis level, large-scale RNA-seq/microarray studies generally applied multiple-testing correction [6,7], but targeted qPCR arrays often used ad hoc fold-change thresholds without family-wise error control, increasing selective-reporting risk for candidate genes [11,22].

Omics reporting quality improved with recency but remained heterogeneous. Recent RNA-seq and single-cell studies more often met MINSEQE elements and provided accessioned datasets or detailed pipelines [1,6,7,15,19], whereas early microarray studies varied in MIAME completeness and sometimes lacked platform-specific normalization detail [8,34]. Data availability for small-RNA and lncRNA microarrays was inconsistent, and preprint/secondary analyses that pooled tissues were appraised as indirect for cumulus-specific inference [7,15,19]. Using JBI and adapted Newcastle–Ottawa tools, most cross-sectional IVF cohorts were rated low-to-moderate risk overall with recurring concerns in selection (case definition clarity—occasionally incomplete in older reports) and comparability (stage/protocol/phenotype matching) [6,7,12,31,34,37]. ROBINS-I identified at least moderate risk in studies introducing ex vivo exposures (adrenergic agents) or multi-passage culture prior to expression assays [23,24]. Collectively, the strongest evidence signals derive from cumulus-specific, stage-aware designs with transparent pipelines and data deposition [1,[5], [6], [7],11,33,35], while conclusions from small, selective, or culture-based datasets were down-weighted in the synthesis [14,23,24]. Full study-level appraisals, tool scores, and domain ratings are provided in Supplementary File 1.

3.3. Results of data extraction

Using a standardized, piloted form, we extracted harmonized metadata and methods from all 33 articles [[1], [2], [3], [4], [5], [6], [7], [8], [9], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19], [20], [21], [22], [23], [24], [25],[30], [31], [32], [33], [34], [35], [36], [37]], with emphasis on PCOS diagnostic criteria/phenotypes, participant features (age, BMI, insulin resistance and androgen status when reported), ovarian stimulation and trigger timing, sample provenance and purity (cumulus vs corona radiata; per-oocyte pairing; exposure to hyaluronidase; any ex vivo culture), oocyte nuclear maturation stage, and the molecular assay layer and platform. The corpus spans single-cell RNA-seq of paired oocyte–cumulus complexes with explicit GV/MI/MII stratification [6], bulk RNA-seq with paired oocyte–cumulus sampling [7], CC transcriptomics integrated with follicular-fluid metabolomics [1], microarrays (including corona radiata) [3,8,12,18,21,34], small-RNA-seq miRNome studies [16,17,37], targeted qPCR studies (ECM/adhesion, insulin signaling, ER-stress, TGF-β family, steroidogenesis, competence markers) [5,10,11,13,14,20,22,25,[30], [31], [32], [33]], culture-based pharmacologic perturbations [23,24], and secondary re-analyses/meta-analyses/reviews that contextualize CC findings [2,9,15,19]. Stage handling varied: several studies analyzed MII only or within-PCOS MI–MII contrasts [12,21,34], whereas others pooled stages or did not specify; two resources characterized GV→MII programs without PCOS contrasts for staging context [3,4]. Where methods sections were incomplete in the provided excerpts, stage, protocol timing, hyaluronidase exposure, or data access were recorded as “not reported.”

From the results sections, we captured the primary molecular outputs (gene/miRNA/lncRNA/circRNA identifiers, directionality, and effect estimates where available), statistical pipelines (normalization, multiple-testing control), pathway/network analyses, and any reported links to fertilization, embryo quality, blastocyst development, implantation, or pregnancy outcomes. Examples include stage-dependent CC differentials with MAPK/PI3K-AKT/TGF-β enrichment and altered expansion genes [6,7,21,33], ECM/adhesion repression with IR/IS stratification [11], OXPHOS/mitochondrial signatures [7,10], ER-stress changes [13], epigenetic remodeling of CYP19A1 [35], and competence-linked markers within PCOS [5,20]. For secondary studies, we recorded source datasets (e.g., GSE155489, GSE34526), integration methods, and hub/pathway summaries [9,15,19]. Data-sharing/accession fields were populated when explicitly stated; otherwise, we noted “public dataset used” or “not reported.” When only figures were provided, values were digitized where appropriate; if key extraction items were unclear, this was flagged for author contact. See Supplementary File 2 for the complete study-level extraction tables.

3.4. Consistently dysregulated RNA signatures

Across independent human cohorts and platforms, coding transcripts implicated in the ovulatory cascade and matrix assembly—TNFAIP6, PTX3, and to a lesser extent HAS2 and ID1/2/4—show reduced expression or blunted peri-ovulatory induction in PCOS cumulus cells, consistent with impaired cumulus expansion and extracellular matrix organization [2,[5], [6], [7],33]. In parallel, pathway-level analyses reproducibly indicate enrichment/activation of EGFR–MAPK and PI3K–AKT/insulin signaling components in PCOS cumulus cells across single-cell and bulk RNA-seq and cross-study meta-analyses, while oocyte–cumulus TGF-β axis imbalance is reflected by BMP4 upregulation in cumulus cells and reduced ID-family targets [6,7,9,15]. Mitochondrial oxidative phosphorylation genes are frequently altered, most commonly downregulated in bulk RNA-seq, although direction varies across cohorts; ECM/adhesion genes tend to be repressed in PCOS, aligning with functional expansion defects [7,10,11].

Non-coding RNA signatures broadly support these coding-level findings. Multiple studies identify differential miRNA expression in PCOS cumulus cells, with predicted and validated targets mapping to MAPK, Wnt, meiosis, steroidogenesis, and gap-junction pathways; specific examples include miRNA–PTGS2/PTX3 relationships in IVM cohorts [16,20,37]. Long non-coding RNAs are aberrantly expressed and integrated into network models affecting follicular signaling, while a follicular-fluid exosomal circRNA axis (circLDLR → miR-1294 → CYP19A1) correlates with cumulus aromatase regulation, linking extracellular vesicles to cumulus transcriptional control [18,19]. Additional reported signatures include decreased VDR-X2 and blunted PTX3 rise at MII in PCOS, increased rRNA content (18S/5.8S) suggesting heightened protein synthesis capacity, and candidate changes in CALM1 (up) and PSMD6 (down) [14,30,33].

Taken together, the most consistently reported signatures in human PCOS cumulus cells are: reduced or blunted induction of expansion/ECM genes (TNFAIP6, PTX3; often stage-dependent), pathway-level activation of EGFR–MAPK and PI3K–AKT/insulin signaling with TGF-β/OSF imbalance, and widespread non-coding RNA remodeling that targets these axes [2,[5], [6], [7],9,11,[15], [16], [17], [18], [19],33] (Table 2).

Table 2.

Consistently reported RNA signatures in human PCOS cumulus cells.

Signature category	Representative genes/RNAs	Direction in PCOS cumulus vs controls	Platform(s) reporting	Salient context (stage/protocol)	Key refs
Cumulus expansion/core ovulatory genes	TNFAIP6, PTX3, HAS2, ID1/2/4	Reduced/blunted induction; stage dependent	Bulk RNA-seq, scRNA-seq, qPCR	Largest differences at GV; PTX3 rise MI→MII blunted; IVF/ICSI	[2,[5], [6], [7],20]
EGFR–MAPK signaling	EGFR, AREG/EREG targets, MAPK modules	Pathway enrichment/upregulation	scRNA-seq, bulk RNA-seq, meta	Across stages; protocol-agnostic signal	[6,7,9,15]
PI3K–AKT/insulin signaling	INSR components, downstream nodes	Pathway enrichment/upregulation	Bulk RNA-seq; targeted qPCR arrays	Modulated by BMI; obesity within PCOS	[7,9,22]
TGF-β/OSF axis	BMP4 (CC up), ID1/2/4 (down); oocyte GDF9 (down)	Imbalance with reduced CC ID targets	Bulk RNA-seq ± validation; qPCR	Cross-compartment (oocyte–CC)	[7,31,32]
ECM/adhesion network	FN1, VCAN, laminins, integrins	Predominant downregulation	Targeted qPCR array	Stronger in insulin-resistant PCOS	[11]
Mitochondrial OXPHOS	Multiple OXPHOS subunits	Often down (bulk RNA-seq); some up reports	Bulk RNA-seq; targeted qPCR	Cohort/protocol dependent	[7,10]
Gap junctions	GJA1/Cx43, GJC1, GJA3	Mixed: unchanged (bulk); DE in scRNA-seq	Bulk RNA-seq; scRNA-seq	Stage/context sensitive	[2,6,7]
Steroidogenesis/epigenetic	CYP19A1; chromatin marks at CYP19A1	CYP19A1 up; altered H3K9ac/me2, MeCP2	qPCR, ChIP-qPCR	GnRH-antagonist; phenotype A	[35]
Non-coding miRNAs	17 miRNAs (PCOS vs control); miR-144/224	Altered miRNA profiles; inverse with PTGS2/PTX3	Small-RNA-seq; qPCR	IVF/ICSI; IVM maturation	[16,20,37]
lncRNAs	Multiple lncRNAs	Aberrant profiles/network hubs	Microarray; qPCR	IVF/ICSI	[18,19] 25
circRNA axis	circLDLR → miR-1294 → CYP19A1	circLDLR reduced; inferred CYP19A1 regulation	FF exosome RNA-seq; CC qPCR	IVF; exosome–CC link	[18]
Vitamin D signaling	VDR-X2	Lower at MII; parallels PTX3	qPCR	GnRH-antagonist; stage-stratified	[33]
rRNA content	18S/5.8S rRNA	Increased	qPCR	Small n; ICSI	[14]
Other candidates	CALM1 (↑), PSMD6 (↓)	Direction as shown	qPCR	IVF egg donors as controls	[30]

3.5. Modifiers of transcriptomic signatures

Maturation stage is a dominant source of variability (Table 3). Single-cell and stage-stratified studies show the largest PCOS–control differences at the GV stage, with many transcriptomic abnormalities diminishing by MII; PTX3 fails to mount the normal MI→MII increase in PCOS, and certain receptors (e.g., FZD3) are higher in immature versus mature cumulus within PCOS [6,21,25,33]. This stage dependence can make group differences appear inconsistent unless CCs are strictly matched to oocyte nuclear status, and highlights that some PCOS features resemble delayed or altered maturation trajectories rather than fixed states [6,21,33].

Table 3.

Modifiers of cumulus cell transcriptomic signatures in PCOS.

Modifier	Observed effect on cumulus signatures	Representative markers/modules	Context	Key refs
Oocyte maturation stage (GV→MI→MII)	Largest PCOS–control differences at GV; attenuation by MII; blunted PTX3 MI→MII rise in PCOS; higher FZD3 in immature CCs	PTX3, VDR-X2, FZD3; broad DE gene counts	scRNA-seq; stage-stratified qPCR	[6,21,25,33]
Trigger/stimulation protocol	Long GnRH-agonist: CRC cell-cycle up; GnRH-antagonist: CYP19A1 up with epigenetic remodeling; IVM: miRNA–target links to maturation	Cell-cycle/replication, CYP19A1 with H3K9ac/H3K9me2/MeCP2, miR-144/224 ↔ PTGS2/PTX3	ICSI/IVF vs IVM	[12,20,35]
Timing/processing	Post-trigger interval, denudation method can shift expansion/ECM/stress genes	HAS2/TNFAIP6/PTX3; ECM; UPR/ER stress	IVF workflow	[3,4]
Insulin resistance	Stronger ECM/adhesion repression in IR vs IS PCOS; VCAN lower in IR	ECM1, VCAN, integrins, laminins	Targeted qPCR array	[11]
Obesity (no clinical IR)	Insulin-pathway nodes upregulated vs normal-weight PCOS	BRAF, CBL, DOK1, BCL2L1, others	Targeted qPCR array	[22]
Metabolic heterogeneity	OXPHOS direction varies with cohort phenotype	OXPHOS gene sets	Bulk RNA-seq; qPCR	[7,10]

Stimulation/trigger protocols further modulate signatures. Corona radiata profiles obtained after long GnRH-agonist stimulation emphasized cell-cycle/replication modules, while GnRH-antagonist cycles sampling of whole cumulus reported epigenetically upregulated CYP19A1 and altered histone marks at its regulatory regions; IVM settings identified miRNA–target pairs (miR-144/224 vs PTGS2/PTX3) associated with maturation success, indicating that in vitro rescue can reconfigure the ovulatory cascade. [12,20,35]. Differences in timing post-trigger and exposure to denudation enzymes can also influence expression of expansion genes, ECM components, and stress pathways, mandating careful protocol documentation for interpretation [3,4].

Clinical phenotype exerts strong modifying effects independent of diagnostic category. Insulin-resistant PCOS shows broader repression of ECM/adhesion networks (e.g., VCAN, integrins, laminins) relative to insulin-sensitive PCOS, while obese PCOS patients—even without clinical insulin resistance—display upregulation of insulin-pathway nodes in cumulus cells [11,22]. Reports of OXPHOS gene directionality differ across cohorts and may reflect varying distributions of insulin resistance, BMI, and androgen levels, further underscoring the need to stratify or adjust for metabolic phenotype [7,10] (Table 3).

3.6. Cross-platform pathway alterations

EGFR–MAPK and PI3K–AKT/insulin signaling are repeatedly enriched in PCOS cumulus cells across single-cell, bulk RNA-seq, and meta-analytic reuses of public datasets, pointing to heightened growth-factor and metabolic signaling tone within the cumulus compartment [6,7,9,15]. In lockstep, an imbalanced TGF-β/oocyte–cumulus axis—exemplified by BMP4 up in cumulus and reduced ID-family targets—appears recurrent and biologically coherent with altered oocyte–somatic cross-talk [7,31,32]. These signaling shifts co-occur with reduced or blunted induction of cumulus-expansion/ECM programs (TNFAIP6, PTX3, HAS2 and broader adhesion networks), tying molecular disturbance to a key functional phenotype of the peri-ovulatory transition [2,[5], [6], [7],11,20,33].

Bioenergetic and stress pathways also recur. Bulk RNA-seq commonly shows broad downregulation of mitochondrial/OXPHOS genes in PCOS cumulus cells with partial overlap to oocyte mitochondrial changes, although some cohorts report OXPHOS upregulation; ER stress/unfolded-protein-response and inflammatory signaling are reported across primary and secondary analyses [7,10,13,15]. Evidence for gap-junction alterations is mixed—bulk datasets often report stable connexin expression while single-cell analyses identify stage-dependent differences—and steroidogenic capacity in cumulus cells is limited but contextually increased CYP19A1 transcription with epigenetic remodeling has been documented during stimulation [2,6,7,24,35] (Table 4).

Table 4.

Alterations in key biological pathways in PCOS cumulus cells.

Pathway	Predominant pattern in PCOS cumulus	Cross-platform support	Representative markers/examples	Key refs
EGFR–MAPK	Enriched/activated	scRNA-seq, bulk RNA-seq, meta	EGFR, AREG/EREG targets, MAPK modules	[6,7,9,15]
PI3K–AKT/insulin	Enriched/activated; obesity accentuates	Bulk RNA-seq; targeted arrays	INSR pathway nodes; BRAF, CBL, DOK1, BCL2L1	[7,9,22]
TGF-β/OSF	Imbalanced (BMP4 up; ID1/2/4 down)	Bulk RNA-seq; qPCR	BMP4 (CC), ID targets; oocyte GDF9 down	[7,31,32]
Cumulus expansion/ECM	Reduced/blunted induction; adhesion network down	Bulk RNA-seq, scRNA-seq, targeted arrays	TNFAIP6, PTX3, HAS2; ECM1, VCAN, FN1, laminins, integrins	[2,[5], [6], [7],11,20,33]
Mitochondrial OXPHOS	Frequently down; some up reports	Bulk RNA-seq; targeted qPCR	Multiple OXPHOS subunits	[7,10]
Gap junctions	Mixed (often unchanged in bulk; stage-dependent in scRNA-seq)	Bulk RNA-seq; scRNA-seq	GJA1/Cx43, GJC1, GJA3	[2,6,7]
Steroidogenesis/epigenetic	Contextual CYP19A1 up with epigenetic shifts	qPCR; ChIP-qPCR	CYP19A1; H3K9ac/me2; MeCP2; ERβ	[35]
Stress/inflammation	ER stress/UPR and inflammatory pathways up	qPCR; bioinformatics/meta	HSPA5/UPR markers; TNF/IL-17 enrichments	[13,15]

Collectively, the cross-platform picture is of a cumulus microenvironment in PCOS shaped by overactive EGFR–MAPK and insulin/AKT signaling with impaired TGF-β/OSF output, manifesting as defective expansion/ECM assembly and variable mitochondrial competence, under stress/inflammatory pressure; changes in gap junctions and steroidogenesis appear contingent on stage and protocol [2,6,7,[9], [10], [11],13,15,20,31,33,35].

3.7. Transcripts linked to oocyte/embryo competence

Evidence linking cumulus transcripts to competence in PCOS points first to expansion genes: in IVM cohorts of PCOS patients, lower miR-144/224 with higher PTGS2 and PTX3 in cumulus cells associate with successful maturation, and stage-matched analyses show that the normal PTX3 rise at MII is blunted in PCOS, suggesting a mechanistic bridge between transcriptional readiness and maturation success [20,33]. In targeted PCOS cohorts, RUNX2 and GPX3 emerged as candidate positive markers for embryo developmental competence, with PTX3 implicated in fertilization, supporting the use of cumulus transcription to refine embryo selection within PCOS [5].

Outcome associations, however, are nuanced. Corona radiata cell-cycle upregulation observed in PCOS did not correspond to poorer in vitro embryo development among embryos that were selected for transfer in that study, and stage-matched VDR-X2 reductions at MII in PCOS coexisted with similar fertilization rates to controls, implying compensatory mechanisms or post-fertilization divergence [12,33]. Multi-omics integration relating cumulus transcriptomes and follicular-fluid metabolomes identified pathways (e.g., TCA cycle/energy metabolism) correlated with fertilization and early embryo development potential in PCOS, proposing composite markers rather than single genes [1] (Table 5).

Table 5.

Transcripts and gene sets associated with oocyte/embryo competence in PCOS.

Marker/gene set	Association (direction)	Outcome metric	Cohort/context	Key refs
PTX3, PTGS2 (COX-2)	Higher levels associate with successful IVM maturation	Post-IVM maturation, early embryo quality	PCOS, IVM	(20)
PTX3 (stage dynamics)	Blunted MI→MII rise in PCOS; lower at MII vs controls	Fertilization rate similar despite transcript defect	PCOS vs controls; GnRH-antagonist; stage-stratified	(33)
RUNX2, GPX3	Positive association with embryo developmental competence	Embryo development metrics	PCOS IVF	(5)
PTX3 (fertilization)	Implicated as favorable for fertilization	Fertilization success	PCOS IVF	(5)
CRC cell-cycle module	Upregulated in PCOS but not linked to worse development among transfers	Embryos selected for transfer	PCOS vs controls; long GnRH-agonist; CRC	(12)
Pathway composites (energy metabolism)	Pathway scores correlate with fertilization/early embryo development	Fertilization, early embryo potential	PCOS vs controls; integrated omics	(1)

3.8. Key methodological limitations and biases

A primary risk is cell-type misclassification or contamination: many ovarian studies profile mural granulosa or mixed follicular aspirates, yet infer cumulus biology; only studies that explicitly isolate cumulus from cumulus–oocyte complexes (or corona radiata as an inner subset) provide cumulus-specific inferences. Reviews and meta-analyses emphasize that cumulus and mural granulosa show distinct PCOS signatures, so failure to verify cell source threatens construct validity [2,9,12].

Second, maturation-stage imbalance and stimulation/trigger heterogeneity can mask or mimic PCOS effects. Large PCOS–control differences at GV often attenuate by MII, and protocol differences (long GnRH-agonist vs antagonist; IVM vs IVF; timing post-trigger; denudation conditions) influence peri-ovulatory programs and steroidogenic/epigenetic states, making cross-study comparisons sensitive to unreported or unmatched procedural details [3,4,6,20,21,33,35].

Third, phenotype confounding and limited sample sizes are common. Insulin resistance and obesity independently reshape ECM and insulin-pathway signatures; without stratification or adjustment, observed “PCOS effects” may predominantly reflect metabolic status. Several targeted studies include small cohorts or selective subgroups (e.g., obese PCOS without clinical IR), reducing power and external validity [11,14,22].

Finally, analytical and reporting heterogeneity complicates synthesis. Platforms and normalization pipelines vary (qPCR panels with ad hoc fold-change thresholds; microarrays vs bulk vs single-cell RNA-seq), multiple-testing control is inconsistent, raw data availability varies, and outcome definitions (fertilization, embryo grading) lack standardization across centers. Culture exposure of cumulus cells in some experiments introduces ex vivo artifacts not reflective of in vivo states, and some primary reports lack explicit PCOS diagnostic criteria in the excerpts, all of which increase risk of bias and limit meta-analysis feasibility [7,15,19,23,24,34] (Table 6).

Table 6.

Detailed risks of bias and reporting gaps specific to cumulus cell expression studies in PCOS.

Bias domain	Why it matters in cumulus studies	Common manifestations/examples	Consequence for inference	How we appraised/mitigated in this review	Key refs
Cell-type misclassification/contamination	Cumulus and mural granulosa have distinct PCOS signatures; mixing aspirates can dilute or distort cumulus-specific signals.	“Granulosa cells” from follicular aspirates used as cumulus surrogates; corona radiata (inner cumulus) profiled instead of whole cumulus without clear framing.	Construct invalidity; misattribution of cell-compartment effects.	Included only studies with explicit COC-derived cumulus/CRC isolation; flagged CRC as subcompartment; excluded mixed aspirate studies.	[2,9,12]
Maturation-stage imbalance (GV/MI/MII)	PCOS–control differences peak at GV and attenuate by MII; stage drives large transcript shifts independent of disease.	Unequal stage distribution; unspecified stage; pooling GV/MI/MII.	Apparent inconsistencies across studies; confounding by stage.	Required stage reporting; prioritized stage-matched analyses; interpreted pooled-stage results cautiously.	[3,4,6,21,33]
Stimulation/trigger protocol and timing heterogeneity	Protocol (GnRH-agonist vs antagonist), IVM vs IVF, and post-trigger interval modulate peri-ovulatory programs and epigenetic states.	Different stimulation regimens; variable denudation timing/enzymes; IVM rescue contexts.	Protocol effects mistaken for PCOS biology; poor cross-study comparability.	Extracted protocol/timing; stratified synthesis by context; avoided direct comparisons across discordant protocols.	[3,4,12,20,33,35]
Phenotype confounding (IR, BMI, hyperandrogenism)	Insulin resistance/obesity reshape ECM and insulin/AKT modules independent of PCOS diagnosis.	IR vs IS PCOS differences in ECM; obesity without clinical IR upregulates insulin-pathway nodes.	Over- or underestimation of “PCOS effects.”	Noted phenotype stratifications; highlighted IR/BMI-adjusted findings; avoided pooling across unstratified phenotypes.	[7,10,11,22]
Small samples and selective inclusion	Omics studies often underpowered; some use very small n or selective subgroups.	rRNA study with n = 3 per group; focused subcohorts (e.g., obese PCOS with normal HOMA-IR).	Low power, unstable effect estimates, limited external validity.	Reported cohort sizes; down-weighted isolated findings; emphasized replication across datasets.	[14,22]
Ex vivo culture/processing artifacts	Culture and enzymatic denudation can alter receptor and matrix gene expression.	Cumulus cultured ± drugs (adrenergic agents) before profiling; variable hyaluronidase exposure.	Transcriptional states may not reflect in vivo biology.	Preferred freshly isolated cumulus; interpreted culture studies as mechanistic, not baseline phenotype.	[3,4,23,24]
Platform/statistical heterogeneity	Different normalization/DE pipelines and multiple-testing control affect gene calls.	qPCR panels with ad hoc fold-change thresholds; microarray vs bulk vs scRNA-seq; inconsistent FDR control.	Limited comparability; risk of false positives/negatives.	Recorded platforms/pipelines; emphasized pathway-level concordance over single-gene effects; noted FDR usage.	[7,15,19,34]
Outcome heterogeneity and indirectness	ART outcomes variably defined; embryo selection steps can mask biology–outcome links.	Different fertilization/embryo grading schemas; embryos selected for transfer show no adverse outcomes despite CRC differences.	Weak or inconsistent associations with competence.	Mapped each outcome's definition; avoided overgeneralization; highlighted pathway composites where available.	[1,12,20,33]
Incomplete PCOS diagnostic reporting	Lack of explicit criteria reduces confidence in case definition.	Studies omitting Rotterdam/NIH criteria in methods/excerpts; preprints with limited clinical detail.	Misclassification; reduced reproducibility.	Preferred studies with clear criteria; flagged unclear cases; interpreted cautiously.	[6,7,11,12,14,25,34]
Data transparency and availability	Access to raw data supports verification and reanalysis.	Variable deposition of raw/processed data; reliance on secondary GEO analyses.	Hinders harmonization/meta-analysis; potential selection bias in reuses.	Noted accession numbers; leveraged cross-study meta/secondary analyses cautiously; advocated for MIAME/MINSEQE compliance.	[7,9,15,19,34]

4. Discussion

This review integrates coding and non-coding RNA evidence to outline a coherent model of the cumulus phenotype in human PCOS: a follicular niche in which growth-factor and metabolic signaling are rewired, oocyte–cumulus communication is perturbed, and peri-ovulatory matrix assembly is blunted, with regulatory reinforcement from small and long non-coding RNAs. Across single-cell and bulk RNA-seq studies, pathway-level enrichment of EGFR–MAPK and PI3K–AKT/insulin signaling in PCOS cumulus cells recurs despite platform and cohort variation, suggesting a generalized shift toward heightened mitogenic and metabolic input within the cumulus compartment [6,7,9,15]. In parallel, an imbalanced TGF-β/OSF axis—reflected by BMP4 upregulation in cumulus cells with reduced ID-family targets and diminished oocyte GDF9—aligns with impaired transcriptional induction of the core expansion program, including TNFAIP6 and PTX3 [7,31,32]. This triad of altered signaling converges on extracellular matrix organization, where both broad ECM/adhesion repression and failure to fully upregulate expansion genes provide a plausible substrate for the well-recognized expansion defects in PCOS [2,[5], [6], [7],11,20,33].

Non-coding RNA data corroborate and extend this model. Multiple miRNA studies in PCOS cumulus cells report differential expression with predicted targets in MAPK, Wnt, steroidogenesis, and meiosis, and an IVM cohort directly tied lower miR-144/224 to higher PTGS2/PTX3 and successful maturation, providing a mechanistic link between miRNA regulation and the expansion machinery [16,20,37]. Long non-coding RNA dysregulation in PCOS cumulus cells further implicates network-level control of follicular signaling, while the circLDLR–miR-1294–CYP19A1 axis connects follicular-fluid exosomes to cumulus steroidogenic transcripts, underscoring that extra-cellular vesicle cargo can shape cumulus gene expression in PCOS [18,19,36]. Together, these non-coding layers do not merely mirror mRNA changes but plausibly contribute to their persistence and context dependence across stimulation protocols and maturation stages [[16], [17], [18],20].

The strongest modifier of observed differences is oocyte nuclear maturation stage. Single-cell profiling and stage-matched qPCR show that PCOS–control gaps are widest at GV and frequently attenuate by MII, while the normal MI→MII surge in PTX3 is blunted in PCOS, a pattern consistent with a delayed or dampened transcriptional response to the ovulatory cascade rather than a fixed deficit [6,21,33]. This dynamic helps reconcile why bulk datasets sometimes report muted case–control differences when MII-enriched samples predominate, and why findings for connexins appear mixed: bulk analyses often see stability in GJA1/GJC1/GJA3, whereas single-cell data detect stage-contingent shifts [2,6,7]. Stimulation and trigger protocols further color these patterns. Long GnRH-agonist cycles profiling the inner corona radiata layer emphasized cell-cycle/replication modules, whereas GnRH-antagonist cycles documented CYP19A1 upregulation in cumulus with accompanying histone and MeCP2 occupancy changes, indicating that endocrine context and peri-retrieval timing can reprogram both signaling and chromatin at key loci [12,35]. In IVM, miRNA–target couplings that relate to maturation success imply that ex vivo conditions can partially reconfigure the ovulatory response network, cautioning against direct extrapolation to in vivo maturation [20].

Clinical phenotype adds another axis of heterogeneity that is mechanistically informative. Insulin-resistant PCOS shows deeper repression of ECM and adhesion genes, including VCAN and integrin/laminin components, suggesting that metabolic milieu directly impairs the structural assembly of the cumulus matrix [11]. Conversely, obese PCOS without clinical insulin resistance demonstrates upregulation of multiple insulin-pathway nodes in cumulus cells, indicating that adiposity can augment insulin/AKT signaling independent of HOMA-IR classification [22]. These observations provide a parsimonious explanation for divergent reports on mitochondrial oxidative phosphorylation: bulk RNA-seq commonly shows broad OXPHOS downregulation in PCOS cumulus with partial parallels in oocytes, but targeted cohorts have also reported OXPHOS upregulation, likely reflecting differences in phenotype composition and stimulation context rather than true contradictions [7,10]. The net effect is a bioenergetic landscape in PCOS cumulus cells that varies with endocrine–metabolic background, superimposed on a consistent signaling scaffold of MAPK/insulin activation and TGF-β imbalance [1,7,[9], [10], [11],22,31].

Translationally, several cumulus transcripts correlate with competence within PCOS cohorts, yet their utility depends on stage and context. PTX3 and PTGS2 associate with successful IVM maturation in PCOS, and RUNX2/GPX3 have been linked to embryo developmental competence, supporting the feasibility of cumulus-based adjuncts for oocyte/embryo selection [5,20]. However, the corona radiata cell-cycle signature observed in PCOS did not translate into worse development among embryos selected for transfer, and reduced VDR-X2 at MII coexisted with similar fertilization rates in one antagonist-triggered cohort, implying buffering mechanisms and emphasizing that not all molecular differences carry actionable prognostic information [12,33]. In this light, pathway composites that integrate energy metabolism and expansion programs, particularly when paired with follicular-fluid metabolomics, may outperform single-gene readouts by capturing convergent biology across the heterogeneous PCOS spectrum [1,2].

Methodological variation and bias domains identified here offer a roadmap for strengthening future studies and for interpreting apparent discrepancies. The most consequential threats to validity are cell-type misclassification, stage and protocol imbalance, and phenotype confounding, each of which can independently produce or erase case–control differences in key pathways [2,4,9,11,12,21,22,33,35]. Addressing these requires per-oocyte pairing with explicit GV/MI/MII labeling, standardized reporting of stimulation/trigger timing and denudation procedures, and a priori stratification by insulin resistance and BMI [6,11,21,22,33,35]. On the analytical side, cross-platform normalization differences and inconsistent multiple-testing control argue for emphasizing pathway-level concordance, depositing raw data to enable harmonized reanalysis, and adopting minimal reporting standards (MIAME/MINSEQE) tailored to cumulus studies [7,9,15,19,34]. Finally, culture-based perturbation studies remain valuable for mechanism but should be clearly separated from baseline phenotyping to avoid conflation of ex vivo and in vivo states [23,24].

In sum, the cumulus transcriptome in human PCOS reflects a reproducible re-wiring of EGFR–MAPK and insulin/AKT signaling with impaired TGF-β/OSF output, manifesting as incomplete activation of expansion/ECM programs and variable mitochondrial competence under stress pressure, and this biology is modulated by maturation stage, clinical phenotype, and stimulation context [2,6,7,[9], [10], [11],13,20,22,31,33]. Non-coding RNAs appear to consolidate these shifts and provide mechanistic entry points—both within cumulus cells and via extracellular vesicle traffic—that could be leveraged for biomarker development and targeted modulation [[16], [17], [18],20]. Converging on rigorously staged, phenotype-aware, cumulus-specific, and data-transparent designs will be essential to translate these molecular insights into robust tools for ART decision-making and into testable interventions that restore peri-ovulatory cumulus function in PCOS [1,2,[5], [6], [7],9,11,12,15,33,35].

4.1. Limitations

The synthesis of findings from this body of literature is constrained by several methodological and clinical limitations. A primary concern is the significant heterogeneity across studies, particularly regarding oocyte maturation stage, ovarian stimulation protocols, and post-trigger timing, which can mask or mimic genuine PCOS-specific effects. Furthermore, phenotype confounding is a pervasive issue, as many studies fail to adequately stratify participants by insulin resistance or obesity, despite clear evidence that these metabolic factors independently reshape cumulus cell gene expression. Analytical challenges also abound, including inconsistent application of multiple-testing corrections, diverse normalization pipelines across transcriptomic platforms, and variable data transparency, all of which limit the potential for quantitative synthesis and direct cross-study comparison. Finally, risks of cell-type misclassification, small sample sizes in targeted studies, and artifacts from ex vivo culture further complicate the interpretation of results and threaten the validity of attributing observed differences solely to the PCOS condition.

4.2. Future studies

To advance the field, future research must adopt more rigorous and standardized approaches. First, studies should implement stage-aware designs with explicit, per-oocyte documentation of nuclear maturation status (GV, MI, MII) to clearly separate PCOS-related dysregulation from normal developmental processes. Second, it is imperative to prospectively stratify participant cohorts by key clinical phenotypes such as insulin resistance and BMI to dissect the specific contributions of metabolic dysregulation versus the core PCOS diagnosis. Methodologically, researchers should enhance rigor by adhering to established omics reporting standards (e.g., MIAME/MINSEQE), consistently applying multiple-testing corrections, and depositing raw data in public repositories to ensure transparency and enable future meta-analyses. Beyond correlation, the field would benefit from integrated multi-omics approaches that combine cumulus transcriptomics with follicular fluid metabolomics and oocyte quality metrics to identify robust, composite biomarkers. Finally, there is a critical need for functional validation studies using in vitro models to definitively establish the causal roles of key dysregulated pathways and non-coding RNAs in the impaired cumulus expansion and oocyte developmental competence observed in PCOS.

5. Conclusion

This systematic review establishes that the cumulus cell transcriptome in Polycystic Ovary Syndrome (PCOS) is characterized by a coherent and reproducible phenotype. This phenotype is defined by a rewiring of key signaling pathways—specifically hyperactive EGFR-MAPK and PI3K-AKT/insulin signaling coupled with an impaired TGF-β/oocyte-somatic cell communication axis. These alterations converge to cause a blunted peri-ovulatory response, evident in the deficient expression of genes critical for cumulus expansion and extracellular matrix formation. This molecular signature is consistently reinforced by dysregulated non-coding RNAs and is dynamically modulated by oocyte maturation stage, clinical metabolic phenotype, and ovarian stimulation protocols. These findings provide a foundational molecular framework for understanding impaired oocyte competence in PCOS and highlight cumulus cells as a valuable source of biomarkers for refining assisted reproductive strategies.

Appendix A. Supplementary data

The following are the Supplementary data to this article: