Depression is associated with endometrial polyp occurrence and convergent gut microbiota-mitochondrial alterations: a multi-omics case-control study

Endometrial polyps (EPs) affect 7.8%–34.9% of women and cause abnormal uterine bleeding and reproductive failure, yet their pathogenesis is unclear. Depression is over-represented among women with EPs but is usually read as reactive distress; whether that association involves the gut microbiota and mitochondrial biology is unknown. We combined a hospital-based case–control analysis (n = 466), faecal full-length 16S rRNA and blood leukocyte RNA-seq in 30 participants (PD n = 11, PND n = 12, NPND n = 7), and a chronic restraint stress (CRS) mouse model with bacterial supplementation. Depression was significantly associated with higher odds of EP in the full cohort (unadjusted OR 2.28, 95% CI 1.55–3.34; P < 0.001), and this trend remained stable following confounder adjustment, propensity-score matching and multiple sensitivity analyses. Signals were partitioned into EP-specific, depression-specific and dual-associated classes. In the EP-specific class, predicted microbial functions showed less electron-transfer gene content (complexes I–II), while host ETC subunits spanning complexes I, III and IV ( NDUFA1, UQCRH, COX6B1, COX7A2 ) were coordinately downregulated and CCR2, CCR5 and UMPS upregulated, with enrichment for oxidative phosphorylation. Dual-associated signals comprised higher Faecalibacterium prausnitzii and lower Eggerthella lenta , negatively correlated (Kendall τ = −0.37, P = 0.007; PD > PND > NPND), together with downregulation of GDF15, HIC1, TCF7L2 and ICAM1 . In mice, restraint stress was accompanied by stromal proliferation exceeding that in estradiol-treated controls and by reduced uterine TOMM20; oral F. prausnitzii given after stress was associated with higher TOMM20 and lower proliferation, whereas E. lenta was not. These findings place a convergent gut microbiota–mitochondrial axis at the intersection of depression and endometrial proliferation, and nominate candidate taxa and transcripts for hypothesis-driven validation. Limitations: multi-omics profiling included 30 participants (11, 12 and 7) with no depression-only cell; microbial functions were predicted by PICRUSt2 and no metabolite was measured; medication use, diet, physical activity, sleep and hormonal status were incompletely captured; and each bacterial group contained three animals. The axis proposed here is therefore hypothesis-generating and requires longitudinal, metabolomic and tissue-specific validation.