Endometrial cancer (EC) is the sixth most common cancer in women globally, with rising incidence that standard risk factors - obesity, diabetes, prolonged estrogen exposure - cannot fully explain. Researchers have increasingly focused on endocrine-disrupting chemicals (EDCs): synthetic compounds that interfere with the body's hormone system. Phthalates, a family of chemicals used to make plastics flexible, are some of the most widespread EDCs in the modern environment. They are found in food packaging, personal care products, medical devices, and countless plastic consumer goods.
Diethyl phthalate (DEP) is among the most commonly detected phthalates in human urine - its metabolite monoethyl phthalate (MEP) is present in over 95% of individuals tested. Because EC is an estrogen-responsive cancer (estrogen drives tumor growth), DEP's ability to weakly mimic estrogen makes it a plausible contributor to EC development. Epidemiological studies have linked higher urinary DEP levels to more aggressive EC tumors, but these studies only show correlation - the molecular mechanisms by which DEP might actually cause or accelerate EC growth have been unknown.
This study takes a comprehensive approach to answering that question. Using computational biology to generate hypotheses, machine learning to identify key genes, molecular simulations to test physical interactions, and laboratory cell experiments to confirm biological effects, the researchers build a multi-level case for exactly how DEP may contribute to EC at the molecular level.
The study used a systematic approach called network toxicology - combining computational databases and gene expression analysis to map the connections between a chemical and a disease. The first step was identifying EC-associated genes using four public gene expression datasets from the GEO database (GSE17025, GSE36389, GSE63678, GSE115810), which profiled EC tumor tissue versus normal endometrium. Differential expression analysis identified 1,583 genes abnormally expressed in EC. Weighted Gene Co-expression Network Analysis (WGCNA) then grouped these genes into co-expression modules - clusters of genes that tend to rise and fall together, suggesting they participate in common biological processes. Three modules (blue, green, purple) correlated significantly with EC.
The second step was identifying DEP's predicted molecular targets. Three databases were searched: ChEMBL (recorded chemical-target interactions), STITCH (chemical-protein interaction predictions), and SwissTargetPrediction (structure-based target prediction). This yielded 287 unique predicted DEP target genes. The critical analytical step was finding the overlap between EC-associated genes and DEP targets - genes that are both abnormally expressed in EC and predicted to be affected by DEP, suggesting they could be the molecular bridge between DEP exposure and EC development. This overlap contained 19 genes.
Machine learning was applied to prioritize the most important of these 19 genes for EC classification. A total of 113 different algorithm combinations were tested, including Random Forest, Gradient Boosting, XGBoost, LASSO, Ridge regression, Elastic Net, SVM, and others. The best-performing models (RF/GBM-based ensembles) achieved AUCs near or exceeding 0.90. SHAP analysis of the optimal model ranked each gene's contribution to classification, identifying the top five most informative genes for further investigation.
SHAP analysis identified five core genes as the top contributors to EC classification among DEP-associated candidates: FOS (highest contributor), NR4A1, ADRA2C, JUN, and SLC6A2. Each gene achieved individual AUC above 0.79 for EC classification, confirming that their expression alone carries meaningful prognostic information. Together, these genes represent a molecular signature connecting DEP exposure to EC biology.
FOS and JUN are the two subunits of the AP-1 transcription factor complex - a master regulator of cell proliferation, differentiation, and stress responses. AP-1 is activated by MAPK signaling and drives expression of genes like Cyclin D1 that push cells to divide. In endometrial tissue, AP-1 family members have been shown to correlate with cell cycle regulators including Cyclin D1 and CDK4. NR4A1 is an orphan nuclear receptor - a protein that controls gene expression but whose natural activator is unclear - linked to oxidative stress responses and metabolic adaptation in tumor cells. ADRA2C is an alpha-2 adrenergic receptor subtype connected to cAMP/PKA signaling, and SLC6A2 encodes the norepinephrine transporter, linking adrenergic (stress-related) signaling to potential tumor promotion.
Pathway enrichment analysis of all 19 overlapping genes confirmed significant enrichment in the MAPK signaling pathway, cAMP signaling, and cGMP-PKG pathways - all known regulators of cell proliferation, survival, and cell cycle progression. Chemical carcinogenesis pathways and the apoptosis pathway were also enriched, pointing to mechanisms by which DEP could both promote cell division and suppress programmed cell death.
Identifying gene expression associations does not prove that DEP physically interacts with these proteins. To assess whether DEP can actually bind to FOS, NR4A1, ADRA2C, JUN, and SLC6A2, the researchers performed molecular docking simulations - computationally testing whether DEP's three-dimensional structure fits into binding pockets on each protein's surface. The most favorable binding energies were found for SLC6A2 (-6.8 kcal/mol), ADRA2C (-6.0 kcal/mol), and NR4A1 (-5.8 kcal/mol), suggesting these are the most physically compatible interactions. FOS (-4.0 kcal/mol) and JUN (-4.6 kcal/mol) showed weaker predicted binding.
Docking alone shows only a static snapshot. To test whether these complexes remain stable over time, 100-nanosecond molecular dynamics (MD) simulations were performed in GROMACS - essentially simulating the protein-DEP complex swimming in a water environment for a virtual 100 nanoseconds and tracking whether they stay together. SLC6A2, ADRA2C, and NR4A1 complexes showed stable, compact binding: RMSD values (a measure of structural drift) stabilized at 0.25-0.6 nm, and the free energy landscapes showed single deep energy wells indicating stable conformations. FOS and JUN showed greater conformational flexibility and multiple energy states, suggesting less stable binding.
These simulations suggest that DEP is most likely to physically interact with ADRA2C, NR4A1, and SLC6A2 through stable protein-ligand complexes. The interactions with FOS and JUN may be more transient or indirect - potentially mediated through signaling cascades rather than direct binding. The researchers appropriately note that FOS and JUN are transcription factors whose DNA-binding function occurs through a different domain than where DEP docking was modeled, making the FOS/JUN docking results require cautious interpretation.
The computational findings were tested in two human EC cell lines: HEC-1-A and Ishikawa. DEP significantly enhanced cell proliferation at concentrations of 10 nM, 1 micromolar, and 10 micromolar - the biologically active, non-toxic range selected based on dose-response experiments. EdU incorporation assays (which label cells actively copying their DNA) confirmed increased DNA synthesis activity, providing direct evidence that DEP accelerates cell division rather than simply affecting cell death rates.
DEP exposure caused a measurable increase in intracellular reactive oxygen species (ROS) - chemically reactive molecules that damage DNA and proteins and are classic markers of oxidative stress - alongside reduced activity of superoxide dismutase (SOD), the cell's main enzyme for neutralizing ROS. This oxidative stress is important: it can activate stress-responsive signaling pathways including MAPK, linking DEP's oxidative effects to its downstream pro-proliferative signaling.
Western blot protein analysis confirmed the predicted signaling changes: DEP treatment increased phosphorylation (activation) of ERK1/2 (the MAPK pathway output kinase), PI3K and AKT (the parallel PI3K-AKT survival pathway), and GSK3-beta (a target of AKT). The downstream consequence was upregulation of Cyclin D1 and CDK4 - two proteins that physically drive the cell cycle from the G1 (growth) phase into the S (DNA synthesis) phase. Cell cycle analysis by flow cytometry confirmed that DEP-treated cells had a reduced proportion in G1 and increased proportion in S phase, exactly as predicted.
This study proposes a coherent mechanistic model for how DEP may promote endometrial cancer: DEP exposure induces oxidative stress (increased ROS, reduced SOD activity), which activates MAPK/ERK and PI3K/AKT signaling pathways in parallel. These pathways converge on upregulation of Cyclin D1 and CDK4, which drive the cell cycle from G1 into S phase, accelerating cell division. The five core genes (FOS, NR4A1, ADRA2C, JUN, SLC6A2) represent molecular hubs where DEP's effects intersect with EC biology - either as direct binding targets or as transcriptional mediators of the activated signaling pathways.
The study's multi-level framework - from computational screening through machine learning gene prioritization to molecular simulation to cellular validation - represents a methodological template for investigating environmental chemical-cancer links. Most prior studies in this field either establish exposure-disease correlations without mechanism or test one pathway in isolation. The integration here across multiple levels of evidence makes the mechanistic case substantially more convincing, even if it remains preliminary without animal model or clinical validation.
Important limitations include the absence of in vivo (animal) evidence and the reliance on established cancer cell lines rather than primary patient-derived tumor cells. The concentrations of DEP used in cell culture experiments may differ from physiologically relevant tissue exposures in humans. The causal necessity of each signaling pathway - whether ERK or AKT is required for DEP-induced proliferation, or whether both are active independently - was not established with pathway inhibitor experiments. Future studies using MEK/ERK inhibitors, PI3K/AKT inhibitors, and ROS-scavenging agents will be needed to establish which of these pathways is mechanistically required rather than simply co-activated by DEP.