TCF21 promotes epithelial-to-mesenchymal transition and cytoskeleton reorganization in uterine development and endometriosis

Abstract

Endometriosis is a common gynecological disease associated with pelvic pain and infertility. Despite several existing theories, the etiology and molecular mechanisms of endometriosis remain to be investigated. Here we report that transcription factor 21 (TCF21) regulates uterine development and endometriosis pathogenesis by promoting epithelial-to-mesenchymal transition (EMT) and cytoskeleton reorganization. Uterine-specific knockout of Tcf21 in mice promotes EMT of the endometrium and dysplasia of the uterus. Accordingly, patients with endometriosis exhibit high TCF21 expression and an expanded population of stromal cells in both eutopic and ectopic endometria. Integrative epigenomic and transcriptomic analyses in patient-derived ectopic stromal cells reveal that TCF21 transcriptionally activated a cohort of genes, including LIMK2, which is critically involved in cytoskeleton organization. Indeed, TCF21-activated LIMK2-cofilin signaling in stromal cells is associated with actin-cytoskeleton reorganization and increased cell invasion and adhesion. In a surgically constructed mouse model, depletion of Tcf21 in eutopic stromal cells alleviates endometriotic lesions, whereas treatment of mice with AAV-Pgr-Tcf21 leads to increased endometriosis incidence, which could be mitigated by administering the LIM kinase inhibitor LIMKi 3. These observations uncover the importance of the TCF21-LIMK2-cofilin axis in uterine development and endometriosis, supporting the pursuit of TCF21-LIMK2-cofilin targeting in the diagnosis and therapeutics of endometriosis.

Endometriosis causes pelvic pain and infertility, but how uterine development and disease are linked is unclear. Here, the authors show that TCF21 shapes uterine development and drives endometriosis by promoting epithelial–mesenchymal transition and reorganization of the cell skeleton.

Introduction

Endometriosis is a chronic gynecological disorder characterized by the ectopic presence of endometrial-like tissue outside the uterus^1,2. Approximately 5–10% of women of reproductive age globally suffer from endometriosis, with 30–50% of the patients experiencing life-impacting pain during periods, chronic pelvic pain, and infertility³. Several theories exist to explain the cause of endometriosis, including retrograde menstruation, abnormal stem cell differentiation, systemic inflammation, and metabolic dysregulation^1,3. However, the molecular mechanism of endometriosis remains elusive, raising significant challenges for the accurate diagnosis and developing targeted therapy of the disease. Interestingly, while endometriosis is a benign disease, the feature of ectopic endometrial tissues in remote organs is reminiscent of the metastatic behavior of cancer, in which epithelial-to-mesenchymal transition (EMT) and mesenchymal-to-epithelial transition (MET) play critical roles in cancer cell invasion and metastatic outgrowth in the distant organs, respectively⁴. A recent single-cell sequencing analysis revealed that the cell type composition in the eutopic endometrium of endometriosis patients differs significantly from the endometrium of healthy women, with much of the epithelial layers replaced by the stroma⁵. The cell type composition changes more prominently in ectopic lesions, with 70% of stromal cells and only 2% of epithelial cells in the ectopic endometrium.⁵ The predominance of stromal-like cells in ectopic lesions suggests that the occurrence and progression of endometriosis could be related to EMT, which leads to a loss of epithelial identity and acquisition of mesenchymal traits in the endometrium.

While retrograde menstruation of endometrial tissue is routinely observed in almost all ovulatory women, only 10% of these women develop endometriosis⁶, suggesting that the ability of endometrial tissue to implant outside the uterus is crucial for the occurrence of endometriosis. Active endometrial fragments carried by retrograde menstruation must adhere to and invade susceptible remote tissues, followed by local proliferation, angiogenesis, the establishment of an inflammatory milieu, and ultimately the formation of ectopic lesions⁷. The initial step of adhesion and invasion is primarily mediated by actin cytoskeleton reorganization, which provides the primary force for cell migration and focal adhesion with the extracellular matrix (ECM)⁸. Among factors mediating cytoskeleton reorganization, LIM domain kinases (LIMK1 and LIMK2) are serine/threonine and tyrosine dual-specificity kinases primarily responsible for phosphorylating and inactivating actin-depolymerizing factors/cofilin family proteins, thereby altering the ratio of filamentous (F) actin to globular (G) actin⁹. Although recent genome-wide association studies (GWAS) have linked endometriosis to dysregulated cytoskeleton organization¹⁰, the underlying molecular mechanism remains unknown.

As a member of the basic helix-loop-helix transcription factor family, TCF21 plays a key role in regulating differentiation during the development of mesoderm-derived organs, such as the gonads, the glomerulus, and the epicardium^11–13. Consistently, GWAS analysis conducted with multiple ethnic groups has linked the TCF21 mutation to increased risks of coronary artery disease, and TCF21 is required for the phenotypic modulation of smooth muscle cells in atherosclerotic tissues^14–16. Tcf21 also marks visceral adipose mesenchymal progenitors and promotes the transformation of these cells into a fibrotic or quiescent state¹⁷. The uterus originates from the mesoderm-derived müllerian ducts. However, whether TCF21 regulates uterine development is unknown¹⁸. Previously, we and others have found that TCF21 expression is upregulated in ectopic endometrial tissues and stromal cells derived from ectopic endometrium in endometriosis patients^19,20. However, the underlying mechanism of TCF21 in endometriosis is still elusive.

Here, we report a role for TCF21 in both uterine development and the promotion of endometriosis. Uterine-specific knockout of Tcf21 in mice led to explicit uterine dysplasia and EMT of the endometrium. Through a genome-wide screen of TCF21 transcriptional targets and subsequent molecular and functional validation in vitro and in vivo, we further demonstrated that TCF21-activated LIMK2-cofilin signaling in stromal cells is critical in mediating the occurrence of endometriosis by regulating cytoskeleton reorganization.

Results

TCF21 is essential for normal uterine development

Previously, we found that increased SUMOylation of TCF21 stabilized the protein in endometriotic stromal cells²¹, whereas FOXM1 and microRNA-92a-3p suppressed the expression of TCF21^22,23. Highly expressed TCF21 could promote endometriotic lesion growth by forming a heterodimer with USF2 to activate the transcription of SF-1 and ERβ¹⁹. However, these findings were primarily derived from in vitro systems or heterologous transplantation models, leaving the in vivo significance of TCF21 in endometrial development unresolved. In addition, the genomic targets, thus, the primary molecular pathways mediated by TCF21 in endometriosis, have not been systematically investigated. To address these issues, we first generated a uterine-specific Tcf21 knockout mouse model to examine whether Tcf21 plays a role in uterine development. This was established by crossing Tcf21^loxP/loxP mice with progesterone receptor Cre knock-in (Pgr^Cre/+) mice^24,25, with both strains on a C57BL/6 genetic background (Fig. 1a). As it was shown that Pgr^cre/+ edits alleles in multiple female reproductive organs including the pituitary gland, ovary, oviduct, and uterus²⁵, we examined the efficiency of Tcf21 knockout using Western blotting and immunohistochemical (IHC) staining to analyze the above tissues in 6-week-old adult mice. The results confirmed that Tcf21 was effectively depleted in Tcf21^loxP/loxPPgr^cre/+ (Tcf21^d/d) group compared to the control Tcf21^loxP/loxPPgr^+/+ (Tcf21^f/f) group (Fig. 1b, c). While Tcf21 depletion was detected in the ovary and oviduct (pituitary gland exhibited very low endogenous Tcf21 level) in Tcf21^d/d mice, the gross morphology and histology of these tissues appeared normal (Fig. 1d). We further evaluated the function of the hypothalamic-pituitary-ovarian (HPO) axis of Tcf21^f/f and Tcf21^d/d female adult mice by examining the serum levels of follicle-stimulating hormone (FSH), anti-mullerian hormone (AMH), estradiol (E₂), and progesterone (P₄). The results showed no significant differences in the levels of these hormones between the two groups (Fig. 1e), indicating intact pituitary function in the Tcf21^d/d mice. Additionally, Tcf21^d/d mice exhibited normal ovarian follicles compared to the Tcf21^f/f mice (Fig. 1f), as shown by serial sectioning of the ovaries and counting of the various stages of follicles from the two groups. We also examined ovulation numbers and fertilization rates of these mice. The results showed that the number of ovulated oocytes was slightly increased in Tcf21^d/d mice, while the fertilization rate was comparable between Tcf21^f/f and Tcf21^d/d mice (Fig. 1g). These findings indicate that despite partial depletion of Tcf21 in the ovary and oviduct in Tcf21^d/d mice, the basic functions of progesterone production, ovulation, and fertilization are largely preserved.

Fig. 1. Deficiency of Tcf21 in the uterus leads to reduced endometrial thickness and decreased stromal cell proportion.

a Schematic of conditional Tcf21 deletion using the Pgr-Cre transgenic system. Conditional knockout mice: Tcf21^loxP/loxPPgr^cre/+, Tcf21^d/d. Control mice: Tcf21^loxP/loxPPgr^+/+, Tcf21^f/f. b Verification of Tcf21 knockout efficiency by Western blotting in the indicated tissues from 6-week-old Tcf21^f/f and Tcf21^d/d mice, respectively. Expression levels were quantified and normalized to Gapdh (right panel). c Representative images of IHC staining for Tcf21 in the indicated tissues from mice in B. Scale bars, 20 μm. d Representative images of the H&E staining of indicated organs from mice in (b). Scale bars: 100 μm. e Circulating level of FSH, AMH, P₄, or E₂ in Tcf21^f/f and Tcf21^d/d mice. f Measurement of the follicle number in Tcf21^f/f and Tcf21^d/d mice. g Quantification of ovulated oocytes and 2-cell embryo formation rate per mouse in Tcf21^f/f and Tcf21^d/d mice following in vitro fertilization. h Representative images of the uterus from 6-week-old Tcf21^f/f and Tcf21^d/d mice, and quantification of uterine weight normalized to body weight. i Uterine histology of Tcf21^f/f and Tcf21^d/d mice. Scale bars: 100 μm. Statistical analysis of endometrial thickness and myometrium thickness. j Representative Ki-67 IHC in endometrium and quantification of Ki-67-positive cells. k TUNEL staining and quantification of apoptotic cells (green) in the endometrium. Scale bar: 100 μm. l Multiplex IF staining of cytokeratin (CK, red), smooth muscle actin (SMA, pink), and vimentin (green) in the uterus from Tcf21^f/f and Tcf21^d/d mice. Nuclei were stained with DAPI. Scale bar: 50 μm. Quantitative analysis of the IF intensity of CK, vimentin, and the epithelial/stromal ratio is shown on the right. m Western blot analysis of Tcf21, E-cadherin, and vimentin in Tcf21^f/f and Tcf21^d/d uteri. n Representative images of IHC staining and quantification in Tcf21^f/f and Tcf21^d/d uteri. Scale bars: 50 μm. Unless otherwise indicated, n = 6 mice per genotype (g = 5; 1, m = 3). Two-sided Student’s t tests were used for statistical analysis, and exact P values are shown in the panels.

Interestingly, the relative uterine weight (the ratio of uterine weight to body weight) of the 6-week-old Tcf21^d/d mice was significantly lower compared to that of match-aged Tcf21^f/f mice (Fig. 1h). Histological examination further revealed that the uteri of Tcf21^d/d mice exhibited significantly reduced thickness of the endometrium. In contrast, the thickness of the myometrium was roughly the same as that of the control group (Fig. 1i). To clarify the mechanism by which Tcf21 affects endometrium thickness, we first examined the proliferation state of the endometrial cells by immunostaining of Ki67, and the results revealed no difference between the Tcf21^f/f and Tcf21^d/d uterus (Fig. 1j). TUNEL analysis also revealed no significant difference in the apoptotic rate between Tcf21^f/f and Tcf21^d/d uteri (Fig. 1k). We next performed multiplex immunofluorescence (IF) staining to further examine the cell type composition of the uteri of these mice. Cytokeratin, vimentin, and α-smooth muscle actin (α-SMA) were used as molecular markers to label endometrial epithelial cells, stromal cells, and smooth muscle cells, respectively, and cell nuclei were labeled with DAPI (Fig.1l, left panel). The intensity of immunofluorescence staining of cytokeratin and vimentin was quantified to represent the expression levels of these markers, and the epithelial to stromal cell ratio was calculated based on the number of cytokeratin⁺/DAPI⁺ cells and vimentin⁺/DAPI⁺ cells. The results revealed an overtly decreased IF intensity of vimentin, concurrent with increased IF intensity of cytokeratin and epithelial/stromal cell ratio, suggesting that the significantly reduced endometrium thickness in Tcf21^d/d mice was primarily caused by impeded stromal formation (Fig. 1l, right panel). Consistently, Western blot analysis of the adult Tcf21^d/d uteri revealed significantly decreased expression of vimentin and increased expression of E-cadherin, representing mesenchymal phenotypes and epithelial phenotypes, respectively (Fig. 1m). We also performed IHC staining to examine the spatial distribution of these molecular markers in the uteri of the two groups. The results confirmed that vimentin is predominantly localized in endometrial stromal cells and is markedly reduced in Tcf21^d/d mice. In contrast, E-cadherin expression is mainly confined to epithelial cells and is elevated in Tcf21^d/d uteri (Fig. 1n). These findings suggest that a compromised EMT process may contribute to the reduced endometrial stromal cells in the Tcf21^d/d mouse uterus.

We further performed transcriptomic profiling to investigate the potential mechanisms underlying the impaired endometrial stromal development in Tcf21-deficient uteri. RNA-seq analysis was performed on uteri collected from 6-week-old Tcf21^f/f and Tcf21^d/d mice. We identified a total of 1795 differentially expressed genes (DEGs) upon Tcf21 deletion, including 407 upregulated and 1388 downregulated genes in Tcf21^d/d mice compared to controls (p < 0.05, |log₂foldchange| ≥ 0.8). Among classic EMT regulators, Snai2 (also known as Slug) was the only one found to be significantly decreased in Tcf21^d/d uteri, whereas other factors, such as Snail, Twist, or Zeb1, did not show significant changes (Fig. S1a). The results were validated using both RT-qPCR and Western blot analyses (Fig. S1b, c). Therefore, the defect of endometrial stromal development observed in Tcf21-deficient uteri could be mediated by suppressed transcription of Snai2.

Notably, the neonatal mouse uterus only comprises simple luminal epithelium and undifferentiated mesenchyme. Glandular epithelium buds emerge by postnatal day 6 (PND6), and full endometrial glands appear by PND12 and extend into the stroma. Meanwhile, the outer layer of the myometrium matures into organized smooth muscle bundles. The basic uterine histoarchitectural configuration in the adult mouse is established by PND15-20²⁶. To further clarify the role of Tcf21 in regulating uterine development, we next examined Tcf21 expression at different stages using both RT-qPCR and Western blotting. At both mRNA and protein levels, the expression of Tcf21 was detected from PND6, peaked at PND15, and gradually decreased from PND21 to PND42 (Fig. 2a, b). Consistently, IHC staining revealed that Tcf21 is expressed in the epithelium, stroma, and myometrium of the uterus, and the expression peaks at PND15 and gradually decreased after the uterine development is complete (Fig. 2c). In line with this, histological analysis showed that Tcf21^d/d mice exhibited a significant reduction in endometrium thickness from PND12 to PND42 compared to controls (Fig. 2d). We further performed multiplex IF staining, in which cells of epithelial and stromal traits were labeled with cytokeratin and vimentin, respectively, and cell nuclei were labeled with DAPI. The epithelial-to-stromal cell ratio was calculated based on the number of cytokeratin⁺/DAPI⁺ cells and vimentin⁺/DAPI⁺ cells. During key stages of postnatal uterine development, a significant reduction in vimentin signal intensity was observed in the uteri of Tcf21^d/d mice, concomitant with a marked increase in the epithelial-to-stromal cell ratio, suggesting that loss of Tcf21 predominantly impairs stromal cell formation during uterine development. Note that the expression of cytokeratin in the uteri was largely unchanged since PND12 in both Tcf21^f/f mice and Tcf21^d/d mice, implicating that epithelial cell development is in earlier stage and not significantly affected by Tcf21 (Fig. 2e). In addition, measurement of the relative uterine weight (uterine weight/body weight) across developmental stages showed that Tcf21^d/d mice exhibited a significant reduction in uterine weight starting at PND15, which persisted into adulthood (Fig. 2f).

Fig. 2. TCF21 is essential for normal uterine development and female fertility.

a RT-qPCR analysis of Tcf21 mRNA in the uterus of wild-type C57BL/6 mice at the indicated PNDs (n = 5 for each group). b Western blot analysis of Tcf21 protein in the uterus of wild-type C57BL/6 mouse at the indicated PNDs (n = 3 for each group). c Representative images of IHC staining and H-score in wild-type mouse uteri at indicated PNDs (PND6 and PND21, n = 5 mice per group; all other time points, n = 6 mice per group). Scale bars: 50 μm. d Uterine morphology and quantification of endometrial thickness at indicated time points from Tcf21^f/f and Tcf21^d/d mice. Tissues were stained using H&E (PND42, n = 6 mice per group; all other time points, n = 3 mice per group). Scale bars: 100 μm. e Multiplex IF staining of cytokeratin (CK, red), smooth muscle actin (SMA, pink), and vimentin (green) in Tcf21^f/f and Tcf21^d/d uteri at indicated PNDs. Scale bar: 50 μm. Quantitative analysis of the IF intensity of CK, vimentin, and the epithelial/stromal ratio is shown on the bottom (n = 3 for each genotype). f Relative uterine weight (uterine weight/body weight) measured at PND12, PND15, PND21, and PND42 in Tcf21^f/f and Tcf21^d/d mice (n = 3 per genotype). Data are represented as mean ± SEM, and exact P values are shown in the panels. Data were analyzed by one-way ANOVA (a, c) or two-tailed Student’s t test (d–f), respectively.

We further examined whether the defect of uterine anatomical structure would affect the fertility of Tcf21^d/d mice. For this purpose, 6-week-old Tcf21^f/f or Tcf21^d/d female mice (n = 6, each group) were bred to 12-week-old wild-type male mice. During the 6-month breeding trial, the Tcf21^f/f group produced 3.833 ± 1.041 litters per female mouse with an average of 5.913 ± 2.372 pups per litter. In contrast, the Tcf21^d/d group produced only 0.833 ± 0.289 litters per female mouse with an average of 3.200 ± 1.643 pups per litter, indicating significantly compromised fertility of Tcf21^d/d mice (Table 1). Together, these results indicated that Tcf21 plays a crucial role in the morphological and functional development of the uterus.

Table 1.

Fertility analysis of Tcf21 conditional knockout mice

Genotype	Number of litters	Number of pups	Average of pups/litter	Average number of litters/mouse
Tcf21f/f (n = 6)	23	136	5.913 ± 2.372	3.833 ± 1.041
Tcf21d/d (n = 6)	5	16	3.200 ± 1.643	0.833 ± 0.289

Genome-wide identification of TCF21 transcriptional targets associated with endometriosis

It has been recognized that EMT, a phylogenetically conserved mechanism that endows cells with plasticity during embryonic development, may be involved in endometriosis. On the other hand, MET drives the secretory transformation of the endometrium in preparation for embryonic implantation and appears to be impaired in endometriosis^27,28. As we have demonstrated that Tcf21 promotes EMT during uterine development, we hypothesize that dysregulation of TCF21 could contribute to the progression from abnormal uterine development to the occurrence of endometriosis. As mentioned earlier, recent single-cell sequencing revealed a gradual decrease in the epithelial/stromal cell ratio in normal endometrium, eutopic endometrium, and ectopic endometrium of endometriosis patients⁵. To confirm this finding, we collected 6 samples of endometrium from healthy women (controls) and 6 paired eutopic and ectopic endometrium from patients with endometriosis, all of which were obtained during the proliferative phase of the menstrual cycle. Multiplex IF staining was performed using cytokeratin as a marker for epithelial cells, vimentin for stromal cells, and cell nuclei were labeled with DAPI. Quantification of the IF staining intensity revealed that vimentin expression was significantly increased while cytokeratin was decreased in ectopic lesions compared to the control or eutopic endometrium, leading to a gradual increase in the proportion of stromal cells from the normal endometrium, eutopic endometrium, and ectopic endometrium, supporting the presence of EMT-like changes during endometriosis (Fig. 3a). In association with the increased proportion of stromal cells, IHC staining showed that the expression of TCF21 gradually increased from normal endometrium, eutopic endometrium, to ectopic endometrium (Fig. 3b), a result consistent with the observation by another group²⁰, supporting that TCF21 could contribute to the difference of cell type composition and is involved in regulating EMT. We next cross-analyzed multiple transcriptome data of paired eutopic and ectopic endometrium samples (GSE105764, GSE179640, and GSE213216), along with previous results from our group²⁹. The results revealed a consistent and significant increase in TCF21 expression in ectopic lesions compared to the eutopic endometrium (Fig. 3c). Consistently, analysis of single-cell sequencing results (GSE179640, n = 2; GSE213216, n = 1) conducted with paired eutopic endometrium (EM) and ovarian endometrioma (OMA) revealed pronounced expansion of mesenchymal cell cluster alongside marked diminishment of the epithelial cell cluster in OMA compared to EM (Fig. 3d, left). Further examination of TCF21 expression in these populations revealed an overt increase of TCF21 levels in the mesenchymal cell population, particularly in ectopic lesions (Fig. 3d, middle and right). We further performed multiplex IF staining using ectopic endometrium samples to confirm the above analysis. Cytokeratin or vimentin was used to label cells from epithelial or mesenchymal origin, respectively, and cell nuclei were labeled with DAPI. The results confirmed that TCF21 is mainly localized to the nuclei of vimentin-positive stromal cells in the ectopic lesions (Fig. 3e).

Fig. 3. Genome-wide identification of TCF21 transcriptional targets associated with endometriosis.

a A multiplex IF staining of cytokeratin (red) and vimentin (green) in normal endometrium, eutopic endometrium (EM) and ectopic endometrium (proliferative phase). Nuclei, DAPI (blue). Scale bar: 200 μm. Quantitative analysis of average fluorescence intensity of cytokeratin and vimentin, and the epithelial/stromal ratio is shown (right, n = 6 per group). b Representative images of IHC and H-score of TCF21 in a (n = 10 per group). Scale bars: 50 μm. c Heatmap of differentially expressed genes in paired EM and ovarian endometriomas (OMA) from GSE105764, GSE179640, and GSE213216 (q < 0.05). d t-SNE visualization of cell clusters from paired EM and OMA (n = 3) with heatmap showing TCF21 expression. The left panel illustrates annotated cell clusters identified across all samples, color-coded by cell type. The middle and right panels display TCF21 expression levels, indicated by color intensity (purple to yellow gradient; yellow indicates higher expression). e Multiplex IF for TCF21 (green), cytokeratin (yellow) and vimentin (red) in OMA; representative images from three independent experiments with similar results. Scale bar, 5 μm. f Genomic distribution of TCF21 CUT&Tag peaks in endometriotic stromal cells (ESCs). g Venn diagram of TCF21 target genes by cross-analysis of CUT&Tag and RNA-seq conducted in ESCs. h KEGG pathway analysis of TCF21 target genes obtained from G, pathway significance values are presented as −log₁₀(P value). i RT-qPCR of indicated genes in paired EM and OMA (n = 7 pairs). j RT-qPCR of indicated genes in paired eutopic stromal cells (EMs) and ESCs (n = 5 pairs). k qChIP validation of representative TCF21 target genes in ESCs using the indicated antibodies (n = 3 patient-derived primary cells). l RT-qPCR analysis of ESCs with or without overexpression of TCF21 (n = 5 patient-derived primary cells). m RT-qPCR analysis of ESCs with or without depletion of TCF21 (n = 5 patient-derived primary cells). Data were analyzed by one-way ANOVA (a, b) or two-tailed Student’s t test (i–m), respectively. Data are represented as mean ± SEM, and exact P values are shown in the panels.

To further understand the molecular mechanism by which elevated TCF21 expression drives endometriosis, we next performed cleavage under targets and tagmentation (CUT&Tag) experiments with antibodies against TCF21 in ectopic stromal cells. Bound DNA fragments were amplified using non-biased conditions, labeled, and then sequenced via HiSeq 2500. Using MACS version 2 with a corrected p < 0.05, we identified 15,574 TCF21-specific peaks across the genomic landscape. The largest group (50%) of the total peaks was located at gene promoters, 29% of the peaks were located in gene introns, and 18% of the peaks were in the intergenic region (Fig. 3f). We then cross-analyze the above data with the RNA-sequencing results (p < 0.05, |log₂foldchange| ≥ 0.26) from paired eutopic and ectopic endometrial samples we generated previously²⁹. Of the identified 1268 potential TCF21 target genes, 549 genes were up-regulated, and 719 genes were down-regulated in ectopic lesions (Fig. 3g). Subsequent KEGG analysis revealed that the upregulated genes were primarily enriched in pathways related to the regulation of the actin cytoskeleton and focal adhesion. In contrast, the downregulated genes were enriched in pathways associated with neurological diseases, including amyotrophic lateral sclerosis, Alzheimer’s disease, and Huntington’s disease. Since we previously showed that TCF21 primarily regulates transcriptional activation during endometriosis¹⁹, we focused on the most significantly upregulated pathways—the regulation of the cytoskeleton and focal adhesion—for further mechanistic studies (Fig. 3h).

To validate the above sequencing results, we first collected 7 paired eutopic and ectopic endometria from patients with endometriosis. As expected, RT-qPCR analysis revealed that the expression of representative genes in the top classified pathways, including LIMK2, ITGA8, APC, Col6A2, and PTK2, was significantly upregulated in ectopic endometrium compared to the matched eutopic endometrium (Fig. 3i). We further isolated stromal cells from the above paired samples and performed RT-qPCR assays. The results confirmed that these genes were significantly upregulated in ectopic stromal cells compared to the matched eutopic stromal cells (Fig. 3j). We next performed quantitative chromatin immunoprecipitation (qChIP) analysis in endometriotic stromal cells using specific antibodies against TCF21. The results confirmed the enrichment of TCF21 at its target promoters. The promoter of the GAPDH gene was used as a negative control (Fig. 3k). In addition, RT-qPCR analysis revealed that the overexpression of TCF21 in endometriotic stromal cells led to increased expression of its target genes, whereas the knockdown of TCF21 decreased their expression (Fig. 3l, m), supporting the notion that TCF21 transcriptionally activates these genes in endometriotic stromal cells.

TCF21 regulates cytoskeleton reorganization and focal adhesion via the LIMK2-cofilin signaling

Among TCF21 target genes, LIMK2 is a serine/threonine protein kinase critically involved in the regulation of cytoskeletal organization. To further demonstrate that TCF21 directly regulates LIMK2 transcription, we performed luciferase reporter assays using a 425 bp fragment of the LIMK2 promoter containing the TCF21 binding motif, as well as a mutant construct in which the E-box motif was fully mutated. The results showed that TCF21 enhanced the transcription of the wild-type LIMK2 promoter but not the mutant one, supporting that TCF21 directly activates LIMK2 transcription by binding to its promoter (Fig. 4a). It is known that LIMK2 can phosphorylate cofilin at Ser3 thereby disable the function of cofilin to depolymerize actin, leading to rearranged ratio of filamentous (F) actin to globular (G) actin favoring to cell invasion³⁰. To examine the impact of TCF21 on the LIMK2-cofilin signaling pathway, we first performed Western blotting experiments by depleting or overexpressing TCF21 in endometriotic stromal cells. The results showed that knockdown of TCF21 led to decreased levels of LIMK2 and phosphorylated cofilin (p-cofilin). Conversely, overexpression of TCF21 led to increased LIMK2 expression and elevated p-cofilin (Fig. 4b). Overexpression of LIMK2 in TCF21-depleted stromal cells rescued the reduction of p-cofilin, whereas knockdown of LIMK2 or treating with LIMKi3, a synthetic antagonist of LIMKs³¹, in TCF21-overexpressing stromal cells abolished the elevation of p-cofilin (Fig. 4c, d). Notably, while LIMK1 shares high structural homology with LIMK2 and also phosphorylates cofilin³², LIMK1 was not identified as a TCF21-regulated gene in the RNA-seq data, and no enrichment of TCF21 was found in the ±3 kb region surrounding the LIMK1 gene in the CUT&Tag results. We further conducted RT-PCR and Western blotting experiments to examine whether LIMK1 is subjected to regulation by TCF21. The results showed that the expression of LIMK1 exhibited no significant difference in ectopic versus eutopic endometria (Fig. S2a–c), and the level of LIMK1 was unaffected upon overexpression or knockdown of TCF21 in cultured endometriotic stromal cells (Fig. S2d, e). Therefore, TCF21 activates LIMK2 transcription but not LIMK1 in our system. Together, these results demonstrated that TCF21 activated LIMK2-cofilin signaling by transcriptional regulation of LIMK2 in stromal cells from endometriotic lesions.

Fig. 4. TCF21 induces cytoskeleton rearrangement and focal adhesion formation via LIMK2-cofilin signaling.

a Luciferase reporter assays showing that TCF21 significantly activates the wild-type LIMK2 promoter but not the mutant promoter lacking the TCF21 binding motif identified by Cut&Tag (n = 3 patient-derived primary cell cultures). b Total cellular proteins were prepared from endometriotic stromal cells overexpressing or depleted of TCF21, and Western blotting was performed to examine LIMK2, cofilin, and p-cofilin (n = 3). c Western blotting analysis of endometriotic stromal cells with indicated treatment for the measurement of the cofilin and p-cofilin (n = 3). d Endometriotic stromal cells were transfected with TCF21 overexpression plasmids for 24 h and treated with 2.5 µM LIMKi3 for another 48 h. The expressions of cofilin and p-cofilin were tested by Western blotting (n = 3). e Western blotting analysis and quantification of the F-actin/G-actin ratio in endometriotic stromal cells transfected with indicated siRNAs or plasmids (n = 3 patient-derived primary cell cultures). f Endometriotic stromal cells were transfected with siTCF21 or TCF21 overexpression plasmid for the morphological examination of F-actin (red) and focal adhesion (green dot) by confocal microscopy. The mean fluorescence intensity of F-actin, focal adhesion numbers, and area were quantified. Thirty cells were analyzed per sample. Scale bars: 50 μm. The mean fluorescence intensity of F-actin, focal adhesion numbers, and area were quantified. Thirty cells were analyzed per sample. Two-sided Student’s t tests were used for statistical analysis, and exact P values are shown in the panels.

We then examined the effect of TCF21 on the conversion between G-actin and F-actin. Ultracentrifugation assays were performed with endometriotic stromal cell lysates to separate F-actin and G-actin. Depletion of TCF21 led to a significantly decreased ratio of F-actin to G-actin, whereas overexpression of TCF21 increased this ratio (Fig. 4e), indicating that TCF21 regulates the balance between actin polymerization and depolymerization. We next performed IF staining with antibodies against actin and vinculin, a marker for focal adhesion. The results showed that knockdown of TCF21 in ectopic stromal cells caused a decrease in F-actin intensity and total focal adhesion numbers and areas, whereas overexpression of TCF21 led to an increase in these indicators (Fig. 4f), supporting the notion that TCF21 regulates cytoskeleton and focal adhesion dynamics in endometrial stromal cells.

Actin cytoskeleton rearrangement and focal adhesion are crucial for cell migration and invasion³³. Fibroblasts utilize focal adhesions for anchoring to the ECM, thereby facilitating the alignment of stress fibers for migration and invasion^8,34. We next investigated whether TCF21-regulated cytoskeleton reorganization can lead to increased cell invasion and adhesion. To test this, we first titrated the effect of jasplakinolide (Jasp), a cyclo-depsipeptide commonly used to promote actin filament polymerization and stabilization³⁵. Western blotting analysis revealed that treatment with 500 nM Jasp induced marked F-actin aggregation in endometriotic stromal cells (Fig. 5a). We next investigated whether TCF21-regulated cytoskeleton reorganization can lead to increased cell invasion and adhesion. Knockdown of TCF21 in endometriotic stromal cells resulted in a significant decrease in invasion and adhesion to Collagen I, as assessed by transwell assays and adhesion assays, respectively. Treating cells with Jasp could abolish the above effects (Fig. 5b, c). In addition, overexpression of LIMK2 in TCF21-depleted cells rescued the decreased invasion and adhesion (Fig. 5d, e). Conversely, knockdown of LIMK2 or treatment of LIMKi3, a potent LIM kinase inhibitor for both LIMK1 and LIMK2, in TCF21-overexpressed stromal cells abolished their boosted ability of adhesion and invasion (Fig. 5f, g). Importantly, cell viability assessed by the CCK-8 assay demonstrated that treatment with 500 nM Jasp or 2.5 μM LIMKi3, which are the concentrations applied in the aforementioned experiments, had no significant impact on endometriotic stromal cell viability (Fig. 5h). Together, these results supported that TCF21-induced cell invasion and adhesion are mediated by upregulation of LIMK2.

Fig. 5. TCF21 promotes endometriotic stromal cell adhesion and invasion via LIMK2-mediated cytoskeletal rearrangement.

a Representative Western blot of F- and G-actin in endometriotic stromal cells treated with F-actin-stabilizing drug jasplakinolide (Jasp) at different concentrations and quantification of the F-actin/G-actin ratio (n = 3 patient-derived primary cells). b Endometriotic stromal cells were transfected with siTCF21 for 24 h and treated with 500 nM Jasp for another 24 h. Transwell experiments were performed to examine the cell invasion ability (n = 3 patient-derived primary cells). Scale bar: 500 μm. c Endometriotic stromal cells were transfected with siTCF21 for 24 h and then treated with 500 nM Jasp for another 24 h, trypsinized and plated on plates coated with fibronectin (FN) and collagen I for 2 h, and adhered cells were examined (n = 3 patient-derived primary cells). Scale bar: 500 μm. d, e Endometriotic stromal cells transfected with TCF21 expression plasmids and siLIMK2 were tested for cell invasion and cell adhesion (n = 3 patient-derived primary cells). Scale bar: 500 μm. f Transwell experiments were performed with endometriotic stromal cells transfected with TCF21 overexpression plasmids for 24 h and treated with 2.5 µM LIMKi3 or siLIMK2 for another 48 h (n = 3 patient-derived primary cells). Scale bar: 500 μm. g Cell adhesion assays of endometriotic stromal cells transfected with TCF21 overexpression plasmids for 24 h and treated with 2.5 µM LIMKi3 or siLIMK2 for another 48 h (n = 3 patient-derived primary cells). h The ESCs were treated with DMSO, LIMKi3 (2.5 μM), or Jasp (500 nM) for 24 h. The survival rates were normalized to the control group values (DMSO) (n = 4 patient-derived primary cells). Scale bar: 500  μm. Data are represented as mean ± SEM, and exact P values are shown in the panels. Data were analyzed by one-way ANOVA.

TCF21-activated LIMK2-cofilin signaling induces endometriosis in vivo

To further investigate the role of TCF21-activated LIMK2-cofilin signaling in vivo, we employed a surgically induced endometriosis model in wild-type C57BL/6 mice as previously described³⁶. Using an autologous transplantation approach, we grafted endometrial fragments onto the mouse abdominal wall with sufficient vascularization (Fig. 6a). To verify the successful establishment of the endometriosis model, we collected eutopic endometrium and ectopic lesions from post-surgery animals for analysis. Histological examination by H&E staining confirmed that the ectopic lesions were composed of both epithelial and stromal cell components. IHC staining of paired eutopic and ectopic endometria revealed significantly increased expression of COX-2 and ERβ (Fig. 6b), consistent with the previously reported upregulation of these two molecules in human endometriotic tissues^36,37, and supports the fidelity of our murine model in recapitulating essential features of human endometriosis. RT-qPCR and Western blotting analysis revealed that Tcf21 was upregulated in ectopic lesions compared to the eutopic uterus in these mice (Fig. 6c, d), an expression pattern similar to that observed in endometriosis patients. We then constructed the above endometriosis model using the Tcf21^d/d and Tcf21^f/f mice. Four weeks after the auto-transplantation, ectopic lesions were successfully developed from endometrial tissue fragments in both Tcf21^d/d and Tcf21^f/f mice (Fig. 6e, left panels). A significant decline in lesion frequency along with a reduction in lesion volume and weight was observed in Tcf21^d/d mice compared to the Tcf21^f/f counterparts (Fig. 6e, right panels), supporting that Tcf21 promotes the development of endometriosis at ectopic sites. Notably, Western blotting and IHC staining showed that the levels of Limk2 and p-Cof were significantly lower in ectopic lesions of Tcf21^d/d mice compared to those of Tcf21^f/f mice, whereas the level of total cofilin was unchanged, supporting that Tcf21-regulated Limk2-cofilin signaling is functional in the mouse endometrium (Fig. 6f, g).

Fig. 6. LIMK2-cofilin signaling mediates TCF21-induced endometriosis in vivo.

a A schematic drawing of experimentally induced endometriosis. b H&E staining and immunohistochemical analysis of the eutopic endometrium and ectopic lesions collected from C57BL/6N mice with surgically induced endometriosis (n = 3). Representative histological sections show epithelial and stromal compartments, along with immunostaining for COX-2 and ERβ. Scale bars: 50 μm. c RT-qPCR analysis of the eutopic endometrium and ectopic lesions of C57BL/6N mice with endometriosis (n = 5). d Western blotting analysis of the eutopic endometrium and ectopic lesions of endometriosis mice (n = 3). e The number of established endometriotic lesions, volume, weight, and representative images of ectopic lesions in TCF21^f/f mice and TCF21^d/d after 4 weeks of induced endometriosis (n = 15 for each genotype). f The protein levels of Limk2, cofilin, and p-cofilin the ectopic lesions of Tcf21^d/d and Tcf21^f/f mice with endometriosis (n = 3 for each genotype). g IHC analyses of Limk2, cofilin, and p-Cof in the ectopic lesions of Tcf21^d/d and Tcf21^f/f mice with endometriosis (n = 3 for each genotype). Micrographs show representative histological sections of endometriotic lesions. Scale bars: 50 μm. h Immunostaining analysis of Tcf21 in different tissues (n = 3 for each genotype). Scale bar: 20 μm. i Validation of uterine-specific targeting by Western blot analysis (n = 3 for each genotype). j Schematic drawing of experimentally induced endometriosis treated with LIMKi3. k Ectopic lesions of the indicated groups. l The number of established endometriotic lesion, volume and weight were assessed in the indicated groups (n = 6). m Western blotting of Limk2, cofilin, and p-Cof in the ectopic lesions of indicated groups (n = 3). Data were analyzed by one-way ANOVA (l) or two-tailed Student’s t test (c, e), respectively. Data are represented as mean ± SEM, and exact P values are shown in the panels.

To further examine whether LIMK2 mediates the effect of TCF21 on ectopic lesion formation, we next performed rescue experiments. For this purpose, we first constructed a uterine-specific Tcf21 expression system, in which the open reading frame of Tcf21 was cloned under the control of the Pgr promoter in an adeno-associated virus 9 (AAV9) vector. C57BL/6N mice were injected with AAV-Pgr-Tcf21 or control AAV-Pgr-vector in the tail vein to generate systemic viral distribution. After 2 weeks, the sampled mice were sacrificed, and IF staining was performed with representative organs or tissues. The results confirmed strong Tcf21 expression in the uterus, with very weak staining in the kidney or the liver (Fig. 6h). Western blotting also revealed that AAV-Pgr-Tcf21 injection resulted in increased protein levels of Tcf21 in the mouse uterus relative to the control AAV-Pgr-vector group, whereas Tcf21 levels were not affected in the liver, spleen, kidney, or heart (Fig. 6i), supporting that AAV-Pgr-Tcf21 injection successfully induced Tcf21 overexpression in the uterus.

We then examined whether Tcf21-induced endometriosis was dependent on Limk2-cofilin in vivo. For this purpose, we performed ovariectomy on 6-week C57BL/6 mice (n = 24) to eliminate the influence of endogenous hormone fluctuations between groups. On day 2 post-surgery, experimental mice were randomly divided into two groups. Each mouse received a tail vein injection of 2 × 10¹¹ vg/kg of either the AAV-Pgr-Tcf21 or AAV-Pgr-GFP, respectively. Mice were then administered an intramuscular injection of 0.1 mg/kg estrogen every 2 days to synchronize the endometrium. Fourteen days after the virus injection, we performed auto-transplantation surgery in these mice to induce endometriosis. These mice were further divided into four groups with matched body weight (n = 6 for each group): AAV-Pgr-GFP + PBS, AAV-Pgr-GFP + LIMKi3, AAV-Pgr-Tcf21 + PBS, and AAV-Pgr-Tcf21 + LIMKi3. Each group received daily intraperitoneal injections of either LIMKi3 (1 mg/kg/d) or an equivalent volume of PBS. To maintain lesion growth, estrogen was administered intramuscularly every 2 days. At the end point, ectopic lesions were evaluated on day 35 (Fig. 6j). Importantly, overexpression of Tcf21 led to significantly increased growth of endometriotic-like lesions, an effect successfully abolished by administration of LIMKi3 (Fig. 6k, l). It should be noted that LIMKi3 treatment alone led to a reduced lesion growth in the AAV-Pgr-GFP group compared to the control AAV-Pgr-GFP group treated with PBS, indicating that LIMK2 signaling contributes to lesion development under basal conditions, under which Tcf21 is also expressed (Fig. 6k, l). Meanwhile, endometriotic tissues from the AAV-Pgr-Tcf21 groups showed increased Limk2 and p-Cof levels compared with the AAV-Pgr-vector groups, and LIMKi3 treatment abolished the Tcf21-induced phosphorylation of cofilin (Fig. 6m). Together, these results indicate that Tcf21 induces endometriosis in vivo by activating the LIMK2-cofilin signaling pathway, and that LIMKi3 can inhibit the lesion growth in endometrial mouse models with or without overexpression of Tcf21.

Clinicopathological relevance of TCF21-LIMK2 axis in endometriosis

To further extend the role of the TCF21-LIMK2 axis in endometriosis to a clinicopathologically relevant context, we collected 18 paired samples of eutopic and ectopic endometrial tissues. RT-qPCR analysis revealed that both TCF21 and LIMK2 were upregulated in ectopic lesions (Fig. 7a), and a significantly positive correlation was found between the expression of TCF21 and LIMK2 (Fig. 7a). Similarly, Western blotting showed that the protein levels of TCF21 and LIMK2 were higher in ectopic lesions (Fig. 7b), and the level of TCF21 and LIMK2 protein were positively correlated (Fig. 7b). In addition, IHC staining revealed elevated protein levels of TCF21 and LIMK2 in stromal cells of ectopic lesions compared to those in matched eutopic endometrial tissues, and the positive correlation between TCF21 and LIMK2 expression was also confirmed (Fig. 7c). Together, the above evidence supported the involvement of TCF21-LIMK2 signaling in the development and progression of endometriosis (Fig. 7d).

Fig. 7. Clinicopathological relevance of TCF21 and LIMK2 in endometriosis.

a RT-qPCR analysis of paired samples of eutopic endometrial tissues and ectopic endometrial tissues for the expression of TCF21 and LIMK2 (left). The relative level of LIMK2 in both eutopic and ectopic endometrium was plotted against that of TCF21 (right). P = 0.014 by Spearman correlation (n = 18). b Western blot analysis of paired samples for the expression of TCF21 and LIMK2 (left). The relative protein level of LIMK2 in both eutopic and ectopic endometrium was plotted against that of TCF21 (right). P = 0.0008 by Spearman correlation (n = 18). c IHC analysis of paired endometriosis samples for the expression of TCF21 and LIMK2. Representative images are shown (Scale bar: 50 μm). The positively stained nuclei were analyzed, and the mean staining intensity was scored by Image software. Box plots on the right show TCF21 and LIMK2 staining scores in EM and OMA (n = 5 paired patient samples). In the box plots, the center line denotes the median, box bounds indicate the 25th and 75th percentiles, and whiskers indicate the minimum and maximum values. Statistical differences between EM and OMA were analysed using two-tailed Student’s t test, and the correlation between TCF21 and LIMK2 expression in eutopic and ectopic tissues was assessed by Spearman correlation. Exact P values are shown in the panels. d The proposed model of the TCF21 mediates EMT in uterine development and promotes endometriosis by regulating cytoskeleton reorganization. The schematic diagram in Fig. 7d was created using FigDraw (www.figdraw.com) under a publication license (authorization no. UYYPSb11bd).

Discussion

Whether endometriosis is intrinsically linked to aberrant uterine development has long been debated. Here, we present the first comprehensive study demonstrating that TCF21, a transcription factor of the bHLH family, is essential for normal uterine development and a driving force behind endometriosis. The compromised EMT in the uterus of Tcf21^d/d mice is reminiscent of an earlier report showing that Tcf21 deficiency in the heart, another mesoderm-derived organ, impedes EMT required for cardiac fibroblast development³⁸. Compromised EMT and reduced endometrial thickness in the uterus of the Tcf21^d/d mice could contribute to their reduced fertility. It was shown that EMT of epithelial cells can facilitate the adhesion between endometrium and the embryo by altering cell polarity, while stromal cells can secrete multiple ligands to mediate embryo adhesion to the luminal epithelium³⁹. Stromal cells also engage in cell communication with immune cells, suppressing the cytotoxic effects of T cells and NK cells to create a favorable uterine environment for embryo implantation⁴⁰. Endometrial stromal cells can further differentiate into decidual cells, which form the decidua to provide a favorable environment for the embryo to support pregnancy⁴¹. A clear positive correlation between endometrial thickness and successful pregnancy outcomes has been reported^42,43, and single-cell sequencing has identified stromal cells as the most abundant cell group in the endometrium⁴⁴, consistent with our observation that Tcf21^d/d mice exhibit a thinner endometrium with reduced endometrial stromal cells. Therefore, we propose that aberrant uterine development, infertility, and increased incidence of endometriosis could be intrinsically linked by dysregulated TCF21 expression, at least for some patients. While we demonstrated that the suppressed transcription of Snai2 may mediate the defect in endometrial stromal development observed in Tcf21-deficient uteri, further mechanistic studies should be conducted to clarify how TCF21 regulates Snai2 expression and EMT during uterine development. Whether it is through the direct activation of the transcription of Snai2? Are there other downstream factors of TCF21 in mediating this process? What are the essential co-regulators of TCF21? What is the critical developmental window of TCF21’s function in promoting EMT? Answers to these questions will shed new light on understanding the complex regulatory machinery of uterine development. The dynamic expression pattern of TCF21 in the postnatal uterus, characterized by a peak at PND15 and subsequent decline, is discordant with the progressive increase in endometrial thickness between PND6 and PND42 in the Tcf21^f/f mice. Notably, the observed peak of Tcf21 expression at PND15 coincides with the critical window of endometrial gland formation and stromal organization, since the histoarchitecture of the uterus is formed mainly by PND15, and resembles that of the adult by PND21^45,46. The subsequent decline in Tcf21 levels in the mature uterus is consistent with our hypothesis that the primary function of Tcf21 is to regulate stromal cell formation in uterine development. In addition, postnatal uterine morphogenesis can be regulated by multiple factors, including secreted signaling molecules, transcription factors, epigenetic regulators, ECM remodeling enzymes, and growth factors, which function coordinately in a temporally and spatially specific manner^47–51. Therefore, the declined expression of TCF21 after PND15 could be compensated by other mechanisms for a sustained endometrial expansion.

The increased percentage of stromal cells in the endometrium could be the pathological foundation of the occurrence and progression of endometriosis. Using biophysical measurements, an earlier study showed that eutopic stromal cells exhibited lower mechanical stiffness and higher elasticity/deformability, which could contribute to their migratory and invasive nature compared to counterparts from disease-free women⁵². Consistently, our genome-wide CUT&Tag and RNA-seq analysis showed that TCF21-regulated cytoskeleton reorganization and focal adhesion are essential in promoting endometriosis. Specifically, we provided multiple evidence showing that LIMK2, a serine-threonine kinase crucial for cytoskeleton dynamics, is the major downstream target gene of TCF21 in endometriotic stromal cells. Application of LIMKi3, a small molecular inhibitor of LIMK, to the endometriosis mouse model can mitigate ectopic stromal cell invasion and reduce lesion size. For potential future clinical applications, while TCF21 is a multifunctional transcription factor, it might be “undruggable”; targeting LIMK-cofilin signaling by LIMKi3, or a more specific LIMK2 inhibitor, could be a promising therapeutic strategy towards endometriosis. LIMKi3 has been shown to be effective in multiple preclinical disease models, including bladder outlet obstruction⁵³ and breast cancer xenografts⁵⁴. However, further evaluation of pharmacological safety, tissue-specificity, and potential off-target actions is still needed before clinical application of LIMKi3 or other LIMK2 inhibitors to treat patients with endometriosis. Additionally, epidemiological studies have reported increased risks of hypertension, hypercholesterolemia, and coronary heart disease in women with endometriosis^55–57. It would be interesting to further investigate whether the mechanism we discovered here for TCF21-promoted endometriosis also plays a role in cardiovascular comorbidities^{14,16,58–62}.

Methods

Ethics statement

The research presented herein secured the approval of the Ethics Committee of Peking University First Hospital, documented under approval number 2022[510-001]. All methodologies involving human subjects were executed in strict alignment with the ethical guidelines issued by both institutional and national oversight bodies and conformed to the Declaration of Helsinki of 1964 and its subsequent modifications or comparable ethical benchmarks. Written informed consent was obtained from all subjects involved in the study. Mouse management and experimental protocols were conducted in strict adherence to the institutional guidelines of the Animal Care and Use Committee (ACUC) and were approved by the ACUC of Peking University First Hospital (J201889).

Generation of uterine-conditional knockout mice

Uterine-specific knockout mice were generated by breeding transgenic mice expressing Cre recombinase driven by the Pgr promoter (Jackson Laboratory) onto the C57BL/6N background. A Tcf21^flox allele was constructed by placing two LoxP sites flanking the first exon of the Tcf21 gene. Homozygous Tcf21^flox mice were generated by BIOCYTOGEN (Beijing, China). Pgr-Cre mice were bred with homozygous Tcf21^flox mice, and offspring expressing Pgr-Cre and heterozygous for Tcf21^flox (Tcf21^flox/+, Pgr^Cre/+) were then crossed to obtain males expressing Pgr-Cre and homozygous for Tcf21^flox (Tcf21^flox/flox, Pgr^Cre/+). These males were mated with Tcf21^flox/flox, Pgr^+/+ females. Tcf21 gene was deleted in Pgr-expressing tissues of the obtained female offspring with Tcf21^flox/flox Pgr^Cre/+ genotype (Tcf21^d/d). Female littermates with Tcf21^flox/flox Pgr^+/+ geneotype (Tcf21^f/f) served as experimental controls. All mice were reared in standard conditions with controlled temperature (21–25 °C), humidity (40–70%), free access to water and food, and the housing facility was illuminated between 7:00 am and 7:00 pm. Animal studies were approved by the First Hospital of Peking University Animal Care Committee.

Total RNA extraction and RT-qPCR

Total RNA was isolated using the TRIzol reagent (Invitrogen) according to the manufacturer’s instructions. cDNAs were synthesized from 2 μg of total RNA using the ABI high-capacity cDNA archive kit (Applied Biosystems), and qPCR analysis was conducted with the ABI 7500 sequence detection system with the ABI power SYBR green PCR master mix (Applied Biosystems). The expression of GAPDH was used as the internal control. Assays were carried out with cells isolated from at least three distinct individuals and replicated three times. Primer sequences are provided in Supplementary Table S1.

Western blotting

Proteins were isolated from tissues or cells using RIPA buffer (KeyGen Biotech) supplemented with protease inhibitor cocktail (Amresco) and phosphatase inhibitor (KeyGen Biotech). The protein concentration was determined using the BCA protein assay kit (KeyGen Biotech). Equal amounts of protein samples were separated on 12% SDS-PAGE and transferred to nitrocellulose membranes (Millipore) by a trans-blot apparatus (Bio-Rad). The membranes were blocked with 5% bovine serum albumin for 1 h at room temperature (RT), followed by overnight incubation with primary antibodies at 4 °C, and incubated with appropriate secondary antibodies at 1:5000 for 1 h at 37 °C. Protein bands were visualized using an enhanced chemiluminescence (ECL) solution (Syngene). Quantification of band intensity was conducted with ImageJ software. All antibodies used in the Western blotting are listed in Supplementary Table S2.

Immunohistochemistry (IHC) staining

Tissues were fixed overnight in 4% paraformaldehyde (PFA) and thoroughly washed in 70% ethanol before they were processed, embedded in paraffin, and sectioned. Sections were cut at 4 μm and mounted on silane-coated slides, deparaffinized, and rehydrated with the graded alcohol. Sections were preincubated with 10% normal goat serum and then incubated with primary antibodies at 4 °C overnight. On the next day, sections were washed in PBS and incubated with biotinylated secondary antibody for 1 h at RT. Immunoreactivity was detected using the DAB Substrate kit (ESGB-BIO) and analyzed by ImageJ. All antibodies used are listed in Supplementary Table S2.

Mouse fertility test

For fertility examination, adult female Tcf21^f/f or Tcf21^d/d mice were continuously mated with wild-type male C57BL/6N mice from the age of 12 weeks (n = 6 per genotype) for 6 months. The number of pups and litters were recorded. Vaginal plug formation was checked the following morning to confirm successful mating (designated as gestation day 0.5). Pregnancy was monitored by body weight gain, and litter size was recorded at the time of delivery. To control for potential variability in male fertility, the same fertile male was used to sequentially mate with multiple females in a rotation schedule. All males were confirmed to be fertile before the experiment.

ELISA

Blood collected from experimental mice was incubated for 2 h at RT before being centrifuged for 20 min at 2000 × g. The supernatant was harvested. The levels of AMH, E₂, and FSH were quantified with the ELISA kit (Cusabio) following the instructions from the manufacturer and analyzed using a microplate reader (Bio-Rad).

Histological analysis

Mouse tissue samples were stained with a standard H&E staining after fixation in 4% neutral buffered formalin, paraffin embedding, and sectioning at 4 µm. The sections were processed, stained, and analyzed with ImageJ for endometrial images and light microscopy photographs. The mean thickness of the tissue was calculated from the area and perimeter ratio.

Ovaries were collected and fixed in 4% PFA for 24 h. Then, they were embedded in paraffin and sliced into 4 μm thick pieces for H&E staining. Follicles were counted in every fifth serial section from ten mice in each group. The mean number per section was calculated. The sections were viewed, and follicles were counted at different stages using a Nikon digital microscope (Nikon). Follicles were classified as follows: type 1 (primordial) follicles consisted of one layer of flattened granulosa cells surrounding the oocyte; type 2 (primary) follicle consisted of one to two complete layers of cuboidal granulosa cells; type 3 (secondary) follicles consisted of an oocyte surrounded by more than one layer of cuboidal granulosa cells with no visible antrum; and type 4 (antral) follicles consisted of an oocyte surrounded by multiple layers of cuboidal granulosa cells and containing one or more antral spaces, possibly with cumulus oophores and a thecal layer.

Ovulation assay

To assess ovulation, adult female mice (Tcf21^f/f and Tcf21^d/d) were superovulated by intraperitoneal injection of 5 IU of pregnant mare serum gonadotropin (PMSG), followed 48 h later by 5 IU of human chorionic gonadotropin (hCG). Mice were sacrificed 14–16 h after hCG injection, and the oviducts were excised. Cumulus-oocyte complexes (COCs) were released from the ampullae into M2 medium, and the total number of ovulated oocytes per mouse was counted under a stereomicroscope.

Fertilization rate assay

Epididymal spermatozoa were retrieved from the cauda epididymis of 12–15-week-old C57BL/6N male mice and incubated in human tubal fluid (HTF) medium supplemented with 10% serum protein substitute (SPS) for 90 min at 37 °C under 5% CO₂ and 95% humidity to allow capacitation. Ovulated oocytes with intact cumulus masses were collected from the oviducts and maintained in HTF medium. Capacitated sperm were then added to the cumulus-oocyte complex (COC) droplets at a final concentration of 1–2 × 10⁶/ml and co-cultured at 37 °C in 5% CO₂ and 95% humidity. After 24–48 h, oocytes were collected and examined for fertilization by the presence of two pronuclei (2PN). The fertilization rate was calculated as the percentage of oocytes with 2PN formation.

Multiplex immunofluorescence (IF) staining

Paraffin-embedded sections of mouse uteri from different developmental stages were processed according to the procedures in the multi-IF kit (Panovue). The sections were immunostained for cytokeratin (CK) to identify epithelial cells, vimentin for stromal cells, and α-smooth muscle actin (α-SMA) for smooth muscle cells. Initially, the sections were dewaxed and rehydrated, followed by high-pressure antigen retrieval. After a blocking step, sections were incubated overnight with the primary antibody at 4 °C, exposed to the secondary antibody at RT for 30 min, and incubated with a fluorescence signal amplification solution for 15 min to enhance signal visibility. Sections were applied with a background noise reduction step and further incubated with either another primary antibody or DAPI (Molecular Probes) for nuclear staining. Cell quantification was performed using Aipathwell® (Servicebio), an AI-based digital pathology software. The software automatically identified tissue regions, segmented nuclei based on DAPI, and recognized epithelial and stromal cells based on CK and vimentin staining, respectively. Consistent fluorescence thresholding was applied across samples. The number of CK⁺/DAPI⁺ and vimentin⁺/DAPI⁺ cells was calculated per section, and the epithelial-to-stromal cell ratio was determined accordingly. Three discontinuous cross-sections of each mouse or patient were counted. For signal intensity analysis, the mean fluorescence intensity of CK and vimentin was quantified using ImageJ (NIH) across experimental groups.

TUNEL staining

To examine apoptosis within uterine tissues, deparaffinized tissue sections were treated with 10 μg/mL proteinase K dissolved in 10 mM Tris HCl for permeabilization. Sections were stained using the TUNEL method outlined by the manufacturer (Proteintech). For nuclear visualization, the sections were further stained with DAPI. Observations and image acquisitions were conducted with an epifluorescence microscope (Axio Imager 2, Carl Zeiss) equipped with the ZEN imaging software.

Analysis of single-cell RNA-seq data

Raw single-cell RNA-seq data were obtained from GSE179640 and GSE213216. The cell ranger (V2.1.1) was used to compare the single-cell expression matrix of each sample to the human reference genome hg38. Data were analyzed using R software (Seurat package [V4.0.6]) for quality control and downstream analysis. Each cell cluster was visualized using uniform manifold approximation and projection. Seurat find all markers was employed to identify gene markers in different cell clusters and differentially expressed genes in the process of screening marker genes.

Clinical samples

This study enrolled 18 patients diagnosed with endometriosis and 10 disease-free women aged between 20 and 35. These women had regular menstrual cycles and had not received any hormonal treatments in the 3 months before their surgical interventions. Normal endometrium was obtained from disease-free women laparoscopically. Self-matched pairs of ectopic and eutopic endometrial tissues were collected during laparoscopic removal of ovarian endometriomas and hysterectomy procedures. The cyst wall from cases of ovarian endometriomas, excluding adjacent ovarian tissue, was isolated, and the inner lining was utilized for further analysis. All endometrial tissues were collected during the proliferative phase of the menstrual cycle, as confirmed by histological criteria and the patient menstrual history. Written informed consent was obtained from all participants, adhering to the ethical standards by the Institutional Review Board of Peking University First Hospital. Detailed clinical information is summarized in Supplementary Table S3.

Primary cell culture

Human eutopic endometrial stromal cells and endometriotic stromal cells were extracted using a modified method described previously⁶³. Surgically obtained tissues were cleansed in PBS and finely chopped into 1 mm³ fragments, followed by digestion with collagenase (1 mg/mL) (Sigma-Aldrich) and DNase (0.04 mg/mL) (Sigma-Aldrich) for 1 h at 37 °C. The dissociated tissues were passed through nylon meshes of 70-μm and 20-μm to eliminate epithelial cells. After centrifugation, the cells were suspended in DMEM/F12 (1:1) (HyClone) supplemented with 10% FBS (Gibco), 100 U/mL penicillin (Lonza), 100 U/mL streptomycin (Lonza), and 250 ng/mL amphotericin B (Lonza), and cultured at 37 °C with humidified 5% CO₂. All experimental procedures were performed within three passages.

Transient cell transfection

On the day before transfection, endometriotic stromal cells were seeded in 6-well plates. Transfection was performed with 70% cell confluency, with the culture medium replaced with DMEM/F12 only. Transfection reagent mixture containing Opti-MEM, Lipofectamine RNAiMAX (Invitrogen), and TCF21 siRNA, LIMK2 siRNA, or scrambled control siRNA (GenePharma) was added dropwise to each well. After 6 h, the culture medium was replaced with normal culture medium, and the cells were collected after culturing for an additional 48 h. The siRNA sequences were as follows: TCF21 sense: ATGATCAGCTGGGCCAAGAA; TCF21 antisense: CCACATCAGCCCCATCATTAA; LIMK2 sense: CGGACAAACGGCTCACTCT; LIMK2 antisense: GGACCCGCATGAATCGACTAT. For overexpression of TCF21 and LIMK2, pCMV-TCF21 or pENTER-LIMK2 (WZ Biosciences) was transfected into endometriotic stromal cells using Lipofectamine 3000 (Invitrogen) according to the manufacturer’s protocol. Treated cells were harvested for RNA or protein isolation after 48 h of incubation.

CUT&Tag

CUT&Tag library was prepared by NovoNGS® CUT&Tag high-sensitivity kit (Novoprotein). In brief, approximately 1 × 10⁵ endometriotic stromal cells were isolated and incubated with NovoNGS® ConA Beads. Subsequently, antibodies and ChiTag transposons were sequentially attached to the chromatin. Following this step, DNA fragments were obtained, amplified, and purified. Genomic fragments with adapters at both ends were enriched by PCR and sequenced by Novogene Bioinformatics Technology on the HiSeq 2500 platform (Illumina). Trimmed reads were aligned to hg38 using Bowtie2, and peaks were called by MACS2. Peak distribution at genomic regions of interest was depicted using the Integrative Genomics Viewer (IGV). KEGG analysis was conducted based on the Database for Annotation, Visualization and Integrated Discovery (DAVID, https://david.ncifcrf.gov/).

Chromatin immunoprecipitation (ChIP)

The ChIP assay was conducted as previously described¹⁹. Endometriotic stromal cells were cultured in 15-cm dishes to 90% confluency, and ChIP was performed using a ChIP assay kit (Pierce Chromatin Prep Module; Thermo Scientific) according to the manufacturer’s instructions. Briefly, cells were cross-linked by 1% formaldehyde and collected in PBS containing 1% protease inhibitors. Cross-linked cells were then lysed and enzymatically digested with micrococcal nuclease to shear the genomic DNA. Five microliters (5%) of the digested chromatin was used as “input”, and the remaining was immunoprecipitated with anti-TCF21 antibody (1:50; Santa Cruz Biotechnology; sc-377225X) at 4 °C overnight with rotation, followed by IP elution and DNA recovery. IgG was used as a negative control. The purified DNA was analyzed by qPCR. The sequences of primer pairs are provided in Supplementary Table S4.

Luciferase reporter assay

A 425 bp genomic fragment corresponding to the human LIMK2 promoter region (chr22:31,248,084–31,248,509, hg38) was PCR-amplified and cloned into the upstream region of the firefly luciferase gene in the pGL3-Basic vector (Promega, USA). This region contains the TCF21 CUT&Tag binding peak, and includes a canonical E-box sequence (5′-CANNTG-3′, −117 to −112), the established DNA motif recognized by bHLH transcription factors, including TCF21. To functionally assess this motif, a mutant construct was generated by substituting all nucleotides within the E-box motif with unrelated bases via overlapping PCR, thereby abolishing potential TCF21 binding. All plasmids were verified by Sanger sequencing. Ishikawa cells were plated in 24-well plates and co-transfected with 200 ng of either wild-type or mutant LIMK2 promoter-reporter plasmids along with 20 ng of pRL-TK Renilla luciferase plasmid (Promega), using Lipofectamine 3000 (Invitrogen) according to the manufacturer’s instructions. Cells were also co-transfected with either a TCF21 expression vector or an empty vector as a control. After 48 h, cells were lysed, and luciferase activity was measured using the Dual-Luciferase Reporter Assay system (Promega). Firefly luciferase values were normalized to Renilla activity in each well. All transfections were performed in biological triplicate, and data are expressed as mean ± standard deviation (SD).

Immunofluorescence staining

Paired eutopic endometrial stromal cells and endometriotic stromal cells were fixed with 4% PFA for 15 min, permeabilized using 0.5% Triton X-100 in PBS for another 15 min, and then blocked with 2% BSA for 30 min. Subsequently, slides were treated with anti-vinculin (1:500; Abcam) followed by an Alexa Fluor 488-conjugated secondary antibody (Abcam). F-actin was visualized with rhodamine phalloidin (Cytoskeleton), and cell nuclei were stained with DAPI. Imaging was carried out with a Leica TCS SP8 confocal laser scanning microscope. The F-actin quantification and number of focal adhesions were performed by ImageJ. Thirty cells were evaluated per experimental group across three independent experiments.

F-actin/G-actin ratio measurement

The F-actin/G-actin ratio of endometriotic stromal cells was analyzed with the G-actin/F-actin in vivo assay kit (cytoskeleton). Briefly, cells were lysed in LAS2 buffer (pH 6.9). Unbroken cells were pelleted by a short centrifugation, and supernatants were pipetted into clear tubes and centrifuged at 100,000 × g, by which the F-actin (filamentous actin) was efficiently partitioned into the pellet, whereas the G-actin (globular actin) was retained in the supernatant. The actin content at both fractions was quantified through immunoblotting with an anti-actin antibody.

Treatment with LIMKi or jasplakinolide

After starvation overnight, endometriotic stromal cells were incubated in serum-free DMEM/F-12 medium containing different concentrations of the LIMKi3 (2.5, 5, 10, and 20 μM, R&D) or DMSO (Sigma) for 24 h. Cell lysates were collected, and the expression of the interested genes was examined by Western blotting. The minimum effective concentration (2.5 μM) was used for transfected cells in Fig. 4. Jasplakinolide (Jasp) was purchased from Millipore and diluted in DMSO. Similar experiments were performed with endometriotic stromal cells to determine the minimum effective drug concentration as 500 nM, which was applied to cells in Fig. 5. Cells were transfected with siRNA, treated with 500 nM Jasp after 24 h, and incubated for another 24 h before downstream assays.

Cell viability assay

Cell viability was analysed by the CCK-8 assay (Beyotime, Shanghai, China) according to the manufacturer’s protocol. The cells were cultured at a density of 0.5 × 10⁴ cells/well in 100 μl of medium in 96-well microplates (Corning, USA) and allowed to adhere overnight. The following day, cells were treated with DMSO (vehicle control), 2.5 μM LIMKi3, or 500 nM jasplakinolide (Jasp) for 24 h. After treatment, 10 μl of CCK-8 (Dojindo, Japan) reagent was added to each well, and the cells were then cultured for 2 h. All the experiments were performed in triplicate. The absorbance was analysed at 450 nm using a microplate reader (Bio-Rad, Hercules, CA, USA), and wells without cells were used as blanks. The proliferation of cells was assessed by measuring the absorbance.

Matrigel invasion assay

In vitro invasion assays were conducted using transwell chambers (24-well, 8-μm pore size, 6.5-mm diameter, Corning) coated with Matrigel at a 1:8 dilution. 2 × 10⁵ endometriotic stromal cells transfected with siRNAs or plasmids were seeded onto the upper chamber in serum-free medium. Six hundred microliters of DMEM/F12 supplemented with 20% FBS was placed in the lower chamber. Cells were incubated at 37 °C for 48 h before the upper surface of the filter, along with any remaining Matrigel, was cleansed using a cotton swab. Subsequently, the inserts were subjected to 4% PFA fixation for 30 min at RT and stained with 0.1% crystal violet solution. Cells migrated to the lower surface were then observed and captured using an Olympus DP71 microscope, with five fields randomly chosen for quantification in each experimental setup. Assays were independently replicated three times using cells cultured from three distinct individuals.

Cell adhesion assay

96-well flat-bottomed plates were coated with 5 μg/cm² collagen I at 37 °C for 2 h and blocked with 100 μl of 2% BSA. Endometriotic cells transfected with siRNA or plasmid were harvested by EDTA treatment, collected by centrifugation, and suspended in serum-free DMEM/F12. Next, 100 μl of solution containing 10,000 cells/well was plated in triplicate and incubated at 37 °C for 2 h. Nonadherent cells were removed by washing with PBS, fixed in 4% PFA for 20 min at RT, and washed with PBS twice more. The cells were stained with 0.1% crystal violet solution for 20 min at RT, then rinsed with PBS three more times. Images were obtained using an inverted phase-contrast microscope and a photomechanical system (Nikon). To quantify adhesive cells, six randomly selected fields were photographed, and the mean number of cells was recorded. Each experiment was performed in triplicate wells and repeated three times.

Induction of endometriosis

Autologous lesion transplantation for endometriosis construction was conducted using C57BL/6N mice according to a procedure previously described³⁶. Mice aged at 6–8 weeks and weighing at least 20 g were selected. To synchronize the endometrium, estrogen was administered intramuscularly at a dose of 0.05 mg/kg/day for 3 days before transplantation. On the day of transplantation, mice were anesthetized with tribromoethanol via intraperitoneal injection. A unilateral uterine horn was isolated, the muscular layer was removed, and the endometrial tissue was placed in DMEM/F12 medium supplemented with 100 U/ml penicillin and 100 μg/ml streptomycin. Subsequently, the endometrial tissue was surgically attached to a well-vascularized area of the abdominal wall, and the abdomen was sutured closed. Postoperatively, estrogen was administered intramuscularly every 2 days at a dose of 0.1 mg/kg/day to support lesion growth, culminating in the establishment of an endometriosis mouse model 1 month later. The size of the transplanted lesions remained consistent across all groups. The mice were euthanized at 28 days following inoculation. The volume of the implanted tissue was determined as volume = 0.5 × length × width². The implanted endometrial lesions were then collected to extract RNA and protein for subsequent quantification.

AAV9 production and delivery

The Pgr-promoter was constructed by excising the CMV bGlobin sequence from the GV551 vector with XbaI/XhoI restriction enzymes. The control adeno-associated virus (AAV) element was conducted as Pgr-promoter-EGFP-MCS-hGH polyA. The sequence Tcf21 was inserted downstream of EGFP with XhoI/XhoI to construct Pgr-promoter-EGFP-Tcf21-hGH polyA (GeneChem). AAV (2 × 10¹¹ vg/mouse) was delivered via tail vein injection into C57BL/6 mice (female, aged 6 weeks), which were used 2 weeks later to establish the endometriosis model. Subsequently, the mice were divided into 4 groups according to intergroup weight matching principles: AAV-Pgr-EGFP + PBS, AAV-Pgr-EGFP + LIMKi3, AAV-Pgr-Tcf21 + PBS, and AAV-Pgr-Tcf21 + LIMKi3. Each group received intraperitoneal injections of either 1 mg/kg/day LIMKi3 or an equivalent volume of PBS. After 3 weeks, the animals were euthanized for tissue collection.

Statistical analysis

Results from triplicate biological experiments are depicted with error bars representing the mean ± standard error of the mean (SEM), except specifically indicated otherwise. Each experimental condition was replicated a minimum of three times to ensure reliability. Group comparisons were conducted using a two-tailed Student’s t test for two groups, and analysis involving more than two groups was performed with one-way ANOVA using SPSS v13.0. Statistical significance was determined as P < 0.05.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

Q.X, J.L, and J.W.Z. conceived and designed the project. J.W.Z. performed most of the experiments. P.L.W., D.Y.Z. contributed to patient enrollment and information collection. Y.L.M., H.W.F., Z.J.P., L.Y.Q., M.Y.G., R.H.L., C.Z., N.Wu and C.Y.Z. assisted experimental conduction and data interpretation. C.P., Y.F.Z. performed laparoscopic surgery in patients with endometriosis. J.W.Z., J.L., and Q.X. wrote the manuscript. Y.F.S. reviewed and edited the manuscript.

Peer review

Peer review information

Nature Communications thanks Nikhil Kumar and Jae-Wook Jeong for their contribution to the peer review of this work. A peer review file is available.

Data availability

All data associated with this study are present in the paper. CUT&Tag data for TCF21 and RNA-seq data for paired eutopic and ectopic endometrium have been deposited in the GEO with the accession numbers GSE282592 and GSE282532, respectively. RNA-seq data from the uteri of 6-week-old Tcf21^f/f and Tcf21^d/d mice have been deposited in the GEO with the accession numbers GSE304889. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains Supplementary material available at 10.1038/s41467-026-69551-5.