This systematic review was conducted in strict adherence to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (8). The study protocol was prospectively registered with the International Prospective Register of Systematic Reviews (PROSPERO) to ensure transparency and reproducibility.

A comprehensive and systematic search of the literature was executed across three major electronic databases: PubMed/MEDLINE, Embase, and Web of Science. The search window encompassed all records published from inception up to September 2025, with no restrictions placed on the language or publication year. The search strategy employed a combination of Medical Subject Headings (MeSH) and free-text keywords joined by Boolean operators (AND/OR) to maximize sensitivity. Key terms included variations of disease nomenclature (“Endometriosis”, “Endometrioma”, “Adenomyosis”), genetic parameters (“Gene Expression”, “Transcriptome”, “RNA-seq”, “Susceptibility Gene”, “GWAS”), and specific phenotypic classifications (“Deep Infiltrating”, “Peritoneal”, “Ovarian”, “Rectovaginal”). To ensure exhaustiveness, the reference lists of all eligible articles and relevant review papers were manually scrutinized to identify any additional studies that might have been missed by the electronic search.

Eligibility was determined based on pre-defined inclusion and exclusion criteria focusing on the molecular differentiation of endometriosis subtypes. We included primary observational studies—encompassing case-control, cross-sectional, and cohort designs—that analyzed human tissue samples obtained from women undergoing surgical management for histologically confirmed endometriosis. The core inclusion requirement was the availability of quantitative gene expression data (mRNA or protein levels) for known susceptibility candidates, specifically stratified by disease phenotype (Superficial Peritoneal Endometriosis [SUP], Ovarian Endometrioma [OMA], or Deep Infiltrating Endometriosis [DIE]). Studies comparing these ectopic tissues against eutopic endometrium (from women with or without the disease) or performing direct head-to-head comparisons between phenotypes were selected.

Conversely, we excluded studies restricted to animal models or in-vitro cell cultures lacking validation in human tissue, to prioritize clinical translational relevance. Studies that aggregated all lesions under a generic “endometriosis” label without phenotypic stratification were excluded, as they would not contribute to the specific objectives of this review. Editorials, conference abstracts, case reports, and narrative reviews were also omitted from the final analysis.

Studies involving mixed phenotypes (e.g., combined OMA and DIE groups) were excluded unless data were reported separately for each subtype to ensure phenotype-specific analysis.

Data extraction was performed independently by two investigators using a standardized data extraction form to ensure consistency. Extracted variables included bibliometric details (author, year, country), study characteristics (sample size per phenotype, menstrual cycle phase), and methodological specifics (gene expression platforms such as RT-qPCR, Microarray, RNA-Seq, or IHC). Quantitative outcomes, specifically fold-change values or relative expression levels, were recorded for synthesis. Any discrepancies arising during the extraction process were resolved through consensus or consultation with a third senior reviewer.

The methodological rigor of the included observational studies was appraised using the Newcastle-Ottawa Scale (NOS) (9). This tool evaluates studies across three domains: the selection of study groups, the comparability of the groups, and the ascertainment of the exposure or outcome. Studies were scored on a star system, with those achieving a score of 7 or higher deemed to be of high quality. This quality assessment was critical for interpreting the strength of evidence regarding gene dysregulation across phenotypes.

The initial search yielded 412 citations. After removing duplicates and screening titles/abstracts, 58 full-text articles were assessed for eligibility. Finally, 15 observational studies met the strict inclusion criteria of stratifying gene expression by phenotype (Figure 1).

The included studies analyzed a total of 1,240 tissue samples. The majority of studies focused on OMA (n = 11) and DIE (n = 8), with studies isolating SUP lesions (n = 9). The analytical methods primarily included RT-qPCR (n = 10) and immunohistochemistry (n = 8), with one recent singlecell RNA sequencing study (11). The mean NOS score was 7.4, indicating moderate-to-high methodological quality, though common biases included lack of menstrual cycle matching. The characteristics data were provided in Table 1 and the analysis data were provided in Table 2.

The transcriptomic landscape of Ovarian Endometrioma (OMA) is profoundly shaped by its unique anatomical containment. Unlike superficial lesions that shed into the open peritoneal cavity, OMA lesions are encapsulated within the ovarian cortex, resulting in the sequestration of menstrual debris. This creates a distinct microenvironment characterized by chronic iron overload and accumulated hemoglobin breakdown products. Consequently, the genetic profile of OMA is consistently dominated by pathways regulating iron homeostasis and oxidative stress response. Multiple independent studies included in this review demonstrated a robust upregulation of HMOX1 (Heme Oxygenase-1) and FTH1 (Ferritin Heavy Chain) within the cyst walls relative to both eutopic endometrium and peritoneal lesions (10). This specific gene activation suggests that the ectopic tissue undergoes a critical metabolic adaptation to survive the cytotoxic effects of free heme, which would otherwise induce cell death via the Fenton reaction and reactive oxygen species (ROS) generation.

Functionally, this overexpression of antioxidant genes acts as a cytoprotective mechanism that may facilitate the persistence of the lesion. The high expression of HMOX1 enables the degradation of toxic heme into biliverdin, carbon monoxide, and iron, while the concomitant upregulation of FTH1 facilitates the safe sequestration of this labile iron. This coordinated genetic response indicates that OMA cells acquire a “ferroptosis-resistant” phenotype, allowing them to thrive in a pro-oxidant environment that would typically trigger apoptosis in normal endometrial cells. This adaptive signature is a defining molecular feature of the ovarian phenotype, distinguishing it biologically from the fibrotic nature of deep infiltrating nodules.

Beyond metabolic adaptation, OMA tissues also exhibit a specific dysregulation of developmental susceptibility loci identified by GWAS. Dysregulation of genes located at the 7p15.2 locus was frequently observed in ovarian samples. Furthermore, WNT4 expression was found to be significantly elevated in OMA tissues compared to controls across four distinct studies (6). As WNT4 is critical for the embryological development of the female reproductive tract and ovarian follicles, its aberrant activation in adult endometriomas suggests a potential reversion to a more primitive or stem-like state. This elevation may also reflect a dysregulated interaction between the ectopic endometrial stroma and the ovarian surface epithelium, pointing to a phenotype-specific pathway driven by both genetic susceptibility and the distinct organ-specific host environment.

The genetic signature of DIE was characterized by invasive and fibrotic pathways, distinct from the OMA profile. In stark contrast to the cystic and hemorrhagic nature of ovarian lesions, the genetic architecture of Deep Infiltrating Endometriosis (DIE) is defined by aggressive invasiveness and profound fibrotic remodeling, often resembling the molecular profile of adenomyosis rather than superficial implants. The most prominent feature of the DIE transcriptome is the hyperactivation of genes governing extracellular matrix (ECM) deposition. Specifically, FN1 (Fibronectin 1) exhibits its highest expression levels in DIE nodules compared to all other phenotypes, particularly in rectovaginal lesions (11). Recent single-cell RNA sequencing data has elucidated the cellular mechanics behind this, revealing that FN1 overexpression is driven by a pathological crosstalk between ectopic stromal cells and the host mesothelial cells. This interaction triggers a process of Epithelial-Mesenchymal Transition (EMT), leading to the formation of the dense, nodular fibrosis that is clinically characteristic of deep lesions and responsible for anatomical distortion.

A critical clinical feature of DIE is its association with severe, debilitating pain, and this is mirrored at the transcriptomic level by a distinct neuroangiogenic signature. DIE lesions exhibit a significant upregulation of NGF (Nerve Growth Factor) and VEGF (Vascular Endothelial Growth Factor) compared to ovarian or superficial subtypes (p < 0.05) (12). Unlike OMA, where gene expression is focused on survival against oxidative stress, the gene program in DIE appears directed toward the active recruitment of sensory nerve fibers and new blood vessels. This phenomenon of neurogenic inflammation provides a molecular explanation for the specific symptomatology of DIE, such as deep dyspareunia and dyschezia, suggesting that the lesion actively modifies the local neural network to perpetuate pain signals. Significant upregulation (>2-fold change) was consistently observed for FN1 in DIE lesions.

The invasive capacity of DIE nodules—often penetrating the muscularis layer of the bowel or bladder—is further supported by dysregulation in structural integrity genes, particularly VEZT (Vezatin). VEZT encodes a transmembrane protein essential for adherens junctions. Its aberrant expression in DIE samples suggests a compromise in cell-cell adhesion, which facilitates collective cell migration and infiltration into surrounding organs (13). When combined with the observed upregulation of Matrix Metalloproteinases (e.g., MMP-2), this genetic profile confirms that DIE possesses a “semi-malignant” invasive phenotype. These findings strongly support the hypothesis that DIE is a biologically distinct entity, potentially originating from a different progenitor cell population or undergoing a divergent evolutionary path compared to non-invasive peritoneal lesions.

Fibrosis and Adhesion: FN1 (Fibronectin 1), a key susceptibility gene involved in cell adhesion, showed the highest levels of activation in DIE nodules, specifically in rectovaginal lesions (11). Hou et al. (2024) confirmed via single-cell analysis that FN1 expression is driven by the interaction between ectopic stromal cells and mesothelial cells, a mechanism unique to invasive nodules.

SUP lesions exhibited a more variable and inflammatory profile, often resembling eutopic endometrium more closely than DIE or OMA.

Inflammation: High expression of IL-6, COX-2, and TNF-alpha was observed in active red lesions. Unlike DIE, SUP lesions showed significantly lower expression of fibrotic markers like ACTA2 (Smooth Muscle Actin) (14).

The central finding of this systematic review is that the traditional classification of endometriosis phenotypes—Superficial (SUP), Ovarian (OMA), and Deep Infiltrating (DIE)—corresponds to distinct molecular pathologies rather than merely anatomical variations. Our synthesis challenges the historical notion of endometriosis as a single, progressive disease continuum where superficial lesions inevitably evolve into deep nodules. Instead, the transcriptomic divergence observed suggests that these phenotypes may represent separate biological entities, likely originating from different progenitor cells or diverging early in their pathogenesis due to specific microenvironmental interactions (16, 17).

The extensive upregulation of FN1 and VEZT dysregulation in DIE provides molecular support for the hypothesis that DIE shares a closer pathogenesis with adenomyosis (“adenomyosis externa”) than with superficial peritoneal implants. The fibrotic drive and smooth muscle metaplasia characteristic of DIE appear to be intrinsic genetic properties of the lesion, facilitated by Epithelial-Mesenchymal Transition (EMT) pathways that are largely absent in SUP. This implies that DIE lesions possess a semi-malignant, invasive phenotype that requires aggressive surgical resection, as hormonal suppression alone may not reverse the established fibrotic architecture (18).

Variability in analytical methods contributes to heterogeneity. IHC provides spatial localization of proteins (critical for verifying stromal vs. epithelial expression), whereas RT-qPCR and RNA-seq offer precise quantification but lack spatial context. Future studies should combine spatial transcriptomics to bridge this gap.

Conversely, the distinct gene profile of OMA underscores the critical role of the “soil” (the ovarian microenvironment) in shaping the “seed” (the endometrial tissue). The profound upregulation of HMOX1 and iron-transport genes indicates that the primary driver of OMA pathology is oxidative stress resulting from cyclic hemorrhage within a closed cavity. This suggests that OMA is not inherently invasive like DIE, but rather metabolic and inflammatory (19). The frequent dysregulation of WNT4 in ovarian lesions further points to a potential embryological origin or a specific susceptibility of the ovarian coelomic epithelium to metaplasia, supporting the genetic basis of the coelomic metaplasia theory for ovarian subtypes (20).

The identification of phenotype-specific gene signatures necessitates a paradigm shift from a “one-size-fits-all” management strategy to a precision medicine approach. Current medical management, primarily relying on nonspecific hormonal suppression (e.g., oral contraceptives or progestins), fails to account for the molecular heterogeneity of the disease (21).

A major strength of this review is the strict eligibility criterion that required stratification by phenotype. Most prior reviews have aggregated molecular data under a generic “endometriosis” label, diluting phenotype-specific signals. By isolating these subtypes, we have unmasked trends, such as the specific association of FN1 with DIE—that were previously obscured in bulk tissue analysis.

However, several limitations must be acknowledged. First, the transcriptomic data analyzed is derived from bulk tissue samples, which include a mixture of endometrial stroma, epithelium, immune cells, and fibrotic tissue. This heterogeneity can introduce bias; for example, the high immune signal in OMA may reflect macrophage infiltration rather than intrinsic endometrial cell changes. Future research must prioritize single-cell RNA sequencing to deconstruct these signals. Second, the impact of the menstrual cycle phase remains a confounding variable. Gene expression is highly hormone-sensitive, and not all included studies matched controls for the cycle phase (proliferative vs. secretory). Finally, the comparison groups varied across studies, with some using autologous eutopic endometrium and others using disease-free controls, which complicates the interpretation of baseline expression levels, the significant heterogeneity in gene measurement techniques and study designs precluded a formal meta-analysis. And there is a potential for publication bias, as studies with non-significant findings may be underreported.