Serum response factor is essential for endometrial function and prevention of inflammatory fibrosis

To examine the association of SRF expression with endometrial pathology, we performed immunohistochemistry (IHC) on secretory phase endometrial biopsy samples from infertile patients with endometriosis and observed that SRF was diminished in both epithelial and stromal cells compared to controls without endometrial pathology ( Fig. 1 A ). To determine the molecular function of SRF in human endometrial stromal fibroblasts, we performed RNAi-based SRF knockdowns in primary menstrual effluent–derived HESCs (ME-HESCs) and in hTERT-immortalized HESCs (T-HESCs) for independent confirmation and functional validation. After confirming successful knockdown of SRF ( SI Appendix , Fig. S1 A and B ), we performed RNA-seq to assess the impact on gene expression in both ME-HESCs and T-HESCs. The differentially expressed genes (DEGs) were enriched for pathways related to cell cycle and growth response, actin cytoskeleton, and extracellular matrix (ECM) ( Fig. 1 B and SI Appendix , Fig. S1 C and D and Dataset S1 ). Consistent with the transcriptomic data, validation experiments showed that SRF -deficient T-HESCs exhibit a decrease of structured F-Actin filaments and consistently reduced cell viability over time ( Fig. 1 C and SI Appendix , Fig. S1 E ). SRF expression in endometriosis-related infertility and HESC function. ( A ) IHC for SRF protein ( Left ). Bar plot ( Right ) shows H-score of staining intensity in each cell compartment, split by pathology status (control n = 19, infertile with endometriosis n = 16). ** P 0.05, calculated via the Kruskal–Wallis test followed by Dunn’s multiple comparisons test. ( B ) Heatmaps show normalized RNA-seq counts of selected genes differentially expressed in both ME-HESC and T-HESC after SRF knockdown (n = 3/treatment). ( C ) F-Actin staining of T-HESC after SRF knockdown shown in green with DAPI in blue ( Left ; n = 3/treatment)). Bar blot ( Right ) shows mean fluorescence intensity of GFP. ** P < 0.01, calculated via the unpaired t test. (Scale bars, 50 µm.) ME-HESC, menstrual effluent–derived human endometrial stromal cell; T-HESC, hTERT-immortalized human endometrial stromal cell. To assess the consequences of SRF deficiency for decidualization response, we treated SRF -expressing and SRF -deficient T-HESCs with a cocktail of estradiol, medroxyprogesterone acetate, and cyclic AMP (EPC) media or vehicle media for 3 d. While EPC treatment strongly activated IGFBP1 and PRL decidualization marker genes and decreased SRF levels in siNT controls, siSRF-treated cells displayed a blunted response ( SI Appendix , Fig. S1 F ), consistent with previous reports ( 20 , 21 ). These results indicate that though SRF is moderately suppressed during the decidualization process, sufficient predecidual SRF expression is required for HESCs to perform cytoskeletal remodeling and proliferation in the initialization of the decidual response, with relevance for infertility-related endometrial pathophysiology. Given its associations with endometrial pathology and its function in HESCs, we next sought to determine the in vivo role of SRF in uterine function. We profiled SRF protein expression in the mouse uterus throughout early pregnancy and found it was strongest in epithelial and smooth muscle cells from gestation day (GD)0.5-GD3.5 with stromal SRF strengthening to match it by GD4.5. ( SI Appendix , Fig. S2 A ). Because global deletion of Srf causes embryonic lethality, we generated a conditional knockout using the Pgr Cre ( 22 ) to ablate Srf in the Pgr -expressing cells of the uterus ( Pgr Cre/+ Srf flox/flox ; Srf d/d ) and confirmed the knockout efficiency ( Fig. 2 A and SI Appendix , Fig. S2 B ). As expected, SRF was ablated in Srf d/d endometrial epithelial and stromal cells but still expressed in immune and endothelial cells that do not exhibit Pgr Cre activity ( 23 ). In a 6-mo breeding trial, Srf flox/flox (Srf f/f ) controls delivered expected numbers of pups and litters, but Srf d/d females were completely infertile ( Fig. 2 B and SI Appendix , Fig. S2 C ). Since the Pgr Cre is also active in a subset of pituitary, ovary, and oviduct cells, we asked whether Srf d/d mice could produce and transport fertilized embryos to the uterus. Flushing GD3.5 oviducts and uteri demonstrated that young adult Srf d/d females (6 to 8 wk) contained similar numbers of fertilized embryos to controls, but all embryos in Srf d/d females were in the oviducts rather than the uterus ( Fig. 2 C ). IHC identified broad SRF expression in mesenchymal and epithelial cells of the Srf f/f control oviduct, most thoroughly in the isthmus segment ( SI Appendix , Fig. S2 D ). However, in Srf d/d oviducts, SRF was ablated in most cells excepting many epithelial cells of the infundibulum and ampulla where Pgr Cre activity is expected to be weaker ( 24 ). In 6- to 8-wk-old females, ovarian corpora lutea (CL) counts and serum P4 levels were comparable between genotypes ( SI Appendix , Fig. S2 E ). In contrast, older GD3.5 Srf d/d females (9 to 10 wk) showed significantly decreased CLs and serum P4, and by 11 to 14 wk, Srf d/d females did not contain embryos or CLs with serum P4 levels further reduced ( SI Appendix , Fig. S3 A – D ). SRF IHC showed that SRF was expressed widely in Srf f/f ovaries and was not affected by age over the age range studied here ( SI Appendix , Fig. S3 E ). In Srf d/d ovaries, SRF was ablated in luteal cells whenever they formed but comparable to controls in the remainder of the ovary. As a further measure of hypothalamic–pituitary–ovarian axis function, we examined estrous cyclicity of Srf f/f and Srf d/d mice between 6 and 14 wk of age via vaginal cytology ( SI Appendix , Fig. S4 ). At 6 to 8 wk old, cycling was somewhat variable in mice of both genotypes, but Srf f/f and Srf d/d mice both typically reached each stage of the cycle. At 9 to 10 wk old, Srf f/f females cycled very regularly. Srf d/d mice still cycled but in some cases began showing extended estrus or metaestrus phases. Between 11 and 14 wk old, Srf f/f controls still cycled regularly, but Srf d/d females exhibited prolonged estrus and metaestrus and rarely showed diestrus. This disruption in Srf d/d estrous cyclicity with age mirrors the timeline of diminished serum P4 and lack of ovarian CL formation and supports the conclusion that Srf d/d females gradually develop endocrine dysregulation between 9 and 14 wk of age. These findings describe important roles for SRF in Pgr -expressing cells of the female reproductive tract with respect to oviductal embryo transport and maintenance of endocrine function through the normal reproductive lifespan. However, young adult Srf d/d mice (6 to 8 wk) maintained a functional hypothalamic–pituitary–ovarian axis based on serum P4 levels and embryo counts, providing a window to study the impact of Srf deletion on in vivo uterine function with this model. We therefore excluded Srf d/d mice from further analysis once they reached 9 wk of age, except where otherwise specified. Infertility phenotype of female Srf d/d mice. ( A ) IHC for SRF protein in the uterus at GD3.5 (n = 4/genotype). (Scale bars, 100 µm.) ( B ) Bar plot shows litters per female in a 6-mo breeding trial (n = 7/genotype). *** P < 0.001, calculated via the Mann–Whitney test. ( C ) Bar plot ( Left ) displays the number of fertilized embryos (blastocysts and morulae) flushed from the uterus and oviducts on GD3.5 6 to 8 wk of age (Srf f/f n = 18, Srf d/d n = 21). Images ( Right ) show examples of fertilized embryos flushed from the indicated anatomical region. (Scale bars, 100 µm.) **** P < 0.0001; *** P < 0.001, calculated via the Kruskal–Wallis test followed by Dunn’s multiple comparisons test. ( D ) Gross morphology ( Left ) of uteri 3 d after transferring wild-type blastocysts into pseudopregnancy day 2.5 mice of the indicated genotype. Arrowheads added to indicate implantation sites. (Scale bars, 1 cm.) Bar plot ( Right ) displays the rate of successful implantation (Srf f/f n = 10, Srf d/d n = 7). *** P < 0.001, calculated via the Mann–Whitney test. ( E ) Gross morphology ( Left ) of uteri 5 d after intraluminal oil stimulation of one uterine horn of the indicated genotype. Arrowheads added to indicate the stimulated horn. (Scale bars, 1 cm.) Bar plot ( Right ) displays the weight ratio, calculated as the oil-stimulated horn weight divided by the unstimulated horn weight (Srf f/f n = 12, Srf d/d n = 13). *** P < 0.001, calculated via the Mann–Whitney test. ( F ) IHC for COX2 protein in artificial decidualization test uteri with Insets displaying low magnification images (Srf f/f n = 3, Srf d/d n = 4). (Scale bars, 300 µm.) ( G ) Masson’s trichrome staining shows collagen fibers in blue. Young adult samples were collected at GD3.5 between 6 and 8 wk of age (n = 5/genotype). Aged adult samples were collected during random cycling after the 6-mo breeding trial between 32 and 49 wk of age (n = 5/genotype). (Scale bars, 100 µm.) GD, gestation day. To functionally test the Srf -deficient mouse uterus independent of oviductal transport, we performed nonsurgical embryo transfer of wild-type blastocysts into young adult pseudopregnant recipient uteri. Srf f/f controls supported formation of implantation sites with large decidual balls, but Srf d/d uteri displayed no comparable implantation sites ( Fig. 2 D ). In a single Srf d/d case out of seven, two much smaller, reddened swells were noted, which we judged to be resorbed implantation sites. To directly test the ability of Srf d/d uteri to decidualize independent of embryo presence and function, we performed an artificial decidualization test on ovariectomized mice injected with exogenous hormones to mimic the milieu at implantation. Ovariectomies were performed prior to the onset of age-related endocrine dysregulation to prevent unintended preexisting impact on the uterus. While the oil-stimulated decidual horn in control uteri responded with substantial morphological change, increased weight ratio, and stromal cyclooxygenase-2 (COX2) induction, Srf d/d uteri showed no response ( Fig. 2 E and F ). Furthermore, Masson’s trichrome staining revealed remarkable development of fibrosis in young adult (6- to 8-wk-old) Srf d/d uteri ( Fig. 2 G ). The severity of Srf d/d uterine fibrosis increased in older mice (over 8 mo), though there may be a contribution from systemic endocrine dysregulation to the aging phenotype. These data demonstrate that even when controlling for ovarian and oviductal function, Srf d/d uteri are unable to support embryo implantation and exhibit severe stromal fibroblast dysfunction. To ascertain the broad transcriptomic landscape underpinning the dysfunctional Srf d/d uterus at GD3.5, we performed RNA-seq analysis on bulk tissues from four mice of each genotype. The DEGs were headlined by strong activation of proinflammatory genes and pathways across a wide range of immune markers ( SI Appendix , Fig. S5 A – C and Dataset S2 ). We also observed upregulated E2 target genes, mixed dysregulation of growth response, proliferation, epithelial identity, and ECM regulation genes, and downregulated P4 targets and cytoskeletal contractility genes. Despite the partial E2-induced, P4-repressed gene signature in the Srf d/d uterus, ESR1 and PGR protein levels were not notably altered, and as with serum P4 levels, serum E2 levels were statistically comparable between genotypes ( SI Appendix , Fig. S5 D – F ). Together, bulk GD3.5 analyses indicate that the Srf -deficient mouse uterus transcriptome parallels the SRF -deficient HESCs in the broad categories of cytoskeletal, ECM, and growth response dysregulation. However, unique to the in vivo system, Srf deficiency strongly activates a proinflammatory gene signature. To query cell type composition and disaggregate cell type–specific transcriptomic changes in the Srf d/d uterus, we performed single-cell RNA-seq on two pools of two Srf f/f uteri and two pools of two Srf d/d uteri staged at GD3.5. For each sample, we sequenced between 9,986 and 12,579 cells, for a total of 44,040 cells. After filtering out low-quality cells and likely doublets, 33,500 cells were used for clustering analysis. Separation by genotype was evident in several clusters before implementation of sample-based integration ( SI Appendix , Fig. S6 A ). Integrated clustering identified 20 clusters to which broad cell type annotations were assigned based on expression of known marker genes ( Fig. 3 A and B and Dataset S3 ). Substantial differences in cell type composition were evident between genotypes, with the Srf d/d samples containing about 2/3 the fibroblasts, 3/5 the smooth muscle cells, over double the epithelial cells, and over triple the immune cells compared to controls ( Fig. 3 C ). Consistent with the known Pgr Cre activity pattern in the uterus ( 22 , 23 ), Srf levels were diminished in Srf d/d epithelial cells, fibroblasts, perivascular cells, and smooth muscle cells but not in immune or endothelial cells ( SI Appendix , Fig. S6 B ). These findings show that proinflammatory gene expression in the Srf d/d uterus is accompanied by an increase in immune cells that is not due to direct modulation of Srf levels in the immune cells themselves. Cellular composition of Srf f/f and Srf d/d uterus at GD3.5. ( A ) UMAP plot shows 20 cell clusters identified in four integrated scRNA-seq datasets: two Srf f/f and two Srf d/d , each a pool of uterine tissues from two animals. Dashed lines surround clusters of the same broad cell type. ( B ) Dot plot shows top markers used to annotate clusters. ( C ) Fractional bar plot ( Left ) shows the percentage of each broad cell type out of all cells of each genotype. UMAP plot ( Right ) displays the broad cell types color-coded by cell type and split by genotype. ( D ) UMAP plots show reclustered immune cells split by genotype and color-coded by subtype. ( E ) Bar plot shows the percentage of each immune subtype out of all uterine cells of the same genotype. Fill color corresponds to immune subtype, outline color indicates genotype (black = Srf d/d ). ( F ) IHC for murine macrophage marker F4/80. Arrowheads indicate aggregated regions of F4/80+ cells near epithelial layers (n = 6/genotype). ( G ) IHC for murine neutrophil marker Ly6G. Arrowheads indicate aggregated regions of Ly6G+ cells near epithelial layers (n = 6/genotype). (Scale bars, 100 µm.) To characterize immune cell subtypes, we extracted them from the full dataset and reclustered them, annotating subtypes based on known markers ( Fig. 3 D and SI Appendix , Fig. S6 C and Dataset S4 ). All subtypes except proliferating natural killer (NK) cells increased in Srf d/d uteri as a share of total uterine cells relative to controls ( Fig. 3 E ). Most notably, Srf d/d uteri exhibited a twofold increase of macrophages and a 21-fold increase of neutrophils, which were the predominant immune cell subtypes in the Srf d/d uterus. IHC for the murine macrophage marker F4/80 identified regions of highly concentrated F4/80+ cells near epithelial layers in most Srf d/d endometrium samples at GD3.5 but rarely in Srf f/f endometrium (5/6 Srf d/d , 1/6 Srf f/f ) ( Fig. 3 F ). Unlike macrophages, neutrophils are rare in the healthy uterus, except at menstruation in humans or simulated menstruation in mice ( 25 , 26 ). However, substantial Ly6G+ neutrophil infiltration was evident in several Srf d/d cases at GD3.5 (3/6 Srf d/d , 0/6 Srf f/f ) in and around the epithelium ( Fig. 3 G ). The substantial myeloid immune cell infiltration in the Srf d/d uterus is a likely contributor to the fibrotic phenotype ( 27 , 28 ). Because Srf was deleted in epithelial cells, fibroblasts, perivascular cells, and smooth muscle cells of the Srf d/d uterus, we examined these cell types more closely in our scRNA-seq dataset to identify the cell type–specific transcriptomic impacts of Srf ablation. The molecular function of SRF has been thoroughly described in mesenchymal cells with pathophysiological consequences noted for gastrointestinal and vascular smooth muscle ( 13 ). In our scRNA-seq dataset, all Srf f/f smooth muscle cells and most perivascular cells (91%) expressed the smooth muscle marker Acta2 , but only 51% of Srf d/d smooth muscle cells and 31% of Srf d/d perivascular cells expressed it ( SI Appendix , Fig. S7 A ). A similar pattern held true for smooth muscle markers Tagln and Myh11 . However, analysis of markers more specific for the smooth muscle cluster ( Cnn1 and Actg2 ) or the perivascular cluster ( Pdgfrb , Rgs5 , Mcam , and Nes ) showed that smooth muscle-specific markers decreased in Srf d/d cells, where perivascular-specific markers were much more stable. Smooth muscle cell DEGs showed suppression of smooth muscle activity, cell adhesion, and cytoskeleton-related pathways with concomitant activation of proinflammatory and ECM remodeling–related pathways ( SI Appendix , Fig. S7 B and Dataset S5 ). Consistent with downregulation of smooth muscle contractility genes, IHC for contractile smooth muscle protein calponin 1 (CNN1) identified discontinuities in the circular smooth muscle layer of the Srf d/d myometrium ( SI Appendix , Fig. S7 C ). Perivascular cell DEGs exhibited dysregulation of cell cycle progression and cytoskeletal integrity pathways, and IHC confirmed that Srf d/d perivascular cells were negative for smooth muscle actin alpha 2 (ACTA2) ( SI Appendix , Fig. S7 D and E and Dataset S5 ). These results demonstrate that Srf deletion disrupts smooth muscle and perivascular cell integrity and contractility. To assess the impact of Srf deletion on fibroblast subtypes, we extracted the fibroblast clusters from the full dataset, reclustered them, and reannotated them ( Fig. 4 A and SI Appendix , Fig. S8 A and B and Dataset S6 ). Leveraging the cycling mouse uterus mesenchymal scRNA-seq atlas published by Kirkwood et al. ( 29 ) and the preimplantation stage pregnant mouse uterus scRNA-seq atlas published by Yang et al. ( 30 ), we identified subclusters 0 and 1 as stromal (inner) fibroblasts (“F2” in Kirkwood and “S1” in Yang), and subcluster 4 as basal (outer) fibroblasts (“F3” in Kirkwood and “S2” in Yang). Cell cycle analysis distinguished subclusters 2 and 3 as proliferating fibroblasts, with subcluster 2 skewed toward S phase and subcluster 3 skewed toward M phase. Subcluster 5 markers did not include clear known fibroblast subtype identifiers and suggested possible fibroblast/immune cell doublets that escaped filtering. This subcluster was by far the smallest (<1% of fibroblasts) and did not appear to change based on genotype; we therefore did not classify it further. Like the Yang et al. study, we did not identify a distinct subcluster uniquely enriched for “F1” markers from the Kirkwood et al. atlas, described in that study as subepithelial stromal cells. Since our data are from the preimplantation pregnant uterus rather than cycling uterine cells, F1 fibroblasts may lose their distinct identity due to pregnancy-specific transcriptional reprogramming. Fibroblast subtype analysis in scRNA-seq of Srf f/f and Srf d/d uterine cells at GD3.5. ( A ) UMAP plots ( Left ) show cell clustering pattern of fibroblasts extracted from the full dataset and reintegrated, color-coded by subtype and split by genotype. Dot plot ( Right ) displays expression of top marker genes used to annotate each fibroblast subcluster. ( B ) Bar plot ( Left ) displays selected Ingenuity Pathway Analysis Z-scores of Canonical Pathways enriched in Srf d/d stromal fibroblast DEGs. Violin plots ( Right ) show relative RNA expression of key DEGs in stromal fibroblasts. **** P adj < 0.0001, calculated via Seurat’s FindMarkers function, which uses the nonparametric Wilcoxon rank-sum test. DEG, differentially expressed gene; ECM, extracellular matrix. Fibroblast subtype composition was relatively stable between Srf f/f and Srf d/d samples with the notable exception that proliferating fibroblast subpopulations shifted from G2-M phase dominant (64% in Srf f/f uteri) to S-G2 phase dominant (81% in Srf d/d uteri), suggesting cell cycle arrest in a subset of Srf d/d fibroblasts ( SI Appendix , Fig. S8 C ). To examine the fibroblast-specific transcriptomic consequences of Srf deletion in vivo, we performed differential expression analyses between genotypes for the two major nonproliferating populations of fibroblasts. Srf d/d stromal fibroblasts DEGs were enriched for activation of pathways related to cell stress, inflammation, and fibrosis, whereas pathways related to cytoskeletal contractility and growth response were deactivated ( Fig. 4 B and Dataset S7 ). Notably, Type I and VI collagen genes, TGFβ pathway genes, Mmp2 , and Thbs1 were upregulated, consistent with fibrotic ECM overproduction ( 31 , 32 ), and decidualization-critical genes Egr1 , Egfr , Fst , and Zbtb16 ( 33 – 36 ) were suppressed. Furthermore, Srf d/d stromal fibroblasts were Scara5 -low and Dio2 -high, a gene signature associated with aberrant decidual response and recurrent pregnancy loss (RPL) ( 37 ). Srf d/d basal fibroblasts showed similar but more limited changes, with their DEGs centering around inflammation and cytoskeletal contractility genes ( SI Appendix , Fig. S8 D and Dataset S7 ). Overall, the changes seen in Srf d/d fibroblasts demonstrate a breakdown in cytoskeletal regulation, a disruption of decidual growth response, and proinflammatory, profibrotic gene expression reflective of Srf d/d uterine histology. Mesenchymal cell differential expression analyses reflected the smooth muscle disruption and excessive collagen production of the Srf d/d uterus but did not reveal the full picture of the inflammatory phenotype, particularly the wide array of overexpressed cytokines in the bulk RNA-seq data and the impetus for immune cell accumulation at mucosal layers. For that, we had to shift our focus to the epithelial cells. Following the same approach as the fibroblast analysis, we extracted the epithelial clusters from the full dataset, reclustered, and reannotated them ( Fig. 5 A and SI Appendix , Fig. S9 A and B and Dataset S8 ). Based on the atlas of murine epithelial differentiation published by Spencer et al. ( 38 ) and other known markers, we identified subclusters 1 and 4 as LE cells, subclusters 0 and 2 as GE cells, and subcluster 3 as proliferating epithelial cells. Subcluster 5 appeared to be a group of low-quality cells (few counts and features per cell) rather than a true epithelial subtype ( SI Appendix , Fig. S9 C ), so we left it unclassified. Epithelial subtype analysis in scRNA-seq of Srf f/f and Srf d/d uterine cells at GD3.5 with ligand–receptor analysis. ( A ) UMAP plots ( Left ) show cell clustering pattern of epithelial cells extracted from the full dataset, reintegrated, color-coded by subtype, and split by genotype. Dot plot ( Right ) displays expression of top marker genes used to annotate each epithelial subcluster. ( B ) Bar plot ( Left ) displays selected Ingenuity Pathway Analysis Z-scores of Canonical Pathways enriched in Srf d/d LE cell DEGs. Violin plots ( Right ) show relative RNA expression of key DEGs in LE cells. ( C ) Circle plots display CellChat ligand–receptor analysis of selected signaling pathways among selected cell populations, split by genotype. **** P adj < 0.0001, calculated via Seurat’s FindMarkers function, which uses the nonparametric Wilcoxon rank-sum test. DEG, differentially expressed gene; E2, estradiol; ECM, extracellular matrix; GE, glandular epithelial; LE, luminal epithelial; P4, progesterone. The most obvious alteration in epithelial subtype composition in the Srf d/d samples was the proliferating epithelial cluster, which consisted of nearly 100% Srf d/d cells ( SI Appendix , Fig. S9 D ). This cluster broadly expressed gland markers, but a substantial subset expressed luminal markers, leading us to conclude it was a mixture of both GE and LE cells that clustered based on proliferative status. Ki67 IHC confirmed an overall increase in Srf d/d epithelial proliferation ( SI Appendix , Fig. S9 E ). When quantified, the increase in Ki67 positivity was statistically significant in the LE but did not meet the significance threshold in the GE. Another remarkable difference between genotypes in the epithelial clusters was the decreased separation between LE and GE clusters, evident in both the subclustering and broad clustering analyses, which suggests that Srf d/d epithelial cell transcriptomes are altered in such a way as to lessen the distinction between LE and GE. Probing the transcriptomic differences between genotypes specific to the LE cells via differential expression analysis revealed DEGs enriched for activation of neutrophilic, profibrotic, and estrogenic pathways together with deactivation of cytoskeletal pathways in the Srf d/d cells ( Fig. 5 B and Dataset S9 ). A deeper look at the DEGs uncovered overexpression of E2-inducible mucosal innate inflammatory mediators including Lcn2, Ltf , and C3 , as well as known fibrogenic genes Cxcl17 , Mmp7 , and Fbln1 ( 39 – 43 ). We observed significant decreases in key progesterone signaling genes Gata2 , Fkbp5 , and Areg as well as SRF-induced growth response gene Egr1 ( 11 , 15 ). In the GE cells, DEG analysis identified genes that were directionally similar to the LE DEGs at the pathway level and included many of the same upregulated E2 signaling, innate inflammatory, and fibrosis-related genes as well as many overlapping downregulated P4 signaling, cytoskeletal, and growth response genes ( SI Appendix , Fig. S10 and Dataset S9 ). Notably, Hbegf levels were decreased in the Srf d/d GE, with implications for decidualization through stromal Egfr ( 34 , 44 ). In addition, several known markers of endometrial gland identity and function were decreased, including the implantation-critical cytokine Lif , yet expression of classical gland marker and Lif -regulator Foxa2 ( 9 , 45 , 46 ) was slightly elevated. Both Esr1 and Pgr mRNA levels were subtly altered in Srf d/d LE and GE cells ( Esr1 up in LE, down in GE; Pgr up in both), but they did not translate to clear differences in protein levels ( SI Appendix , Fig. S5 D and E ). The LE-specific and GE-specific Srf d/d gene expression profile together with the spatial pattern of macrophage and neutrophil accumulation in the Srf d/d uterus point to the Srf -deficient epithelial cells as the predominant source of cytokines and chemokines that initiate inflammation. To assess the impact of Srf ablation on inferred cell–cell communication among the major endometrial cell types of interest, we applied ligand–receptor analysis to the Srf f/f and Srf d/d scRNA-seq datasets. This analysis identified increased overall signaling strength from LE to macrophages and neutrophils, from GE to neutrophils, and from macrophages and neutrophils to fibroblasts and most other cell types in the Srf d/d dataset ( SI Appendix , Fig. S11 A and Dataset S10 ). Enriched overall signaling information flow among these cell types in the Srf d/d uterus included increased complement, MIF, and TGFβ signaling and decreased EGF signaling. More specifically, complement signaling from Srf d/d LE and GE cells to macrophages and neutrophils was activated, MIF signaling from Srf d/d LE and GE cells to macrophages was increased, TGFβ signaling from Srf d/d macrophages to stromal fibroblasts was increased, and EGF signaling from Srf d/d GE to stromal and basal fibroblasts was decreased ( Fig. 5 C and SI Appendix , Fig. S11 B ). This evidence suggests that estrogenic innate inflammatory signaling from Srf -deficient epithelial cells recruits myeloid immune infiltration, which in turn activates profibrotic signaling to fibroblasts. At the same time, dysregulation of gland to fibroblast interactions important for normal stromal function likely plays a part in the loss of decidual response. Given the associations we identified between SRF expression decrease and human endometrial pathologies, we sought to assess the cell type–specific applicability of our findings from the Srf d/d mouse model to human endometriosis pathophysiology. For this purpose, we took advantage of the Human Endometrial Cell Atlas (HECA), which contains scRNA-seq data from 63 endometrial tissue donors with or without endometriosis ( 8 ). After excluding data derived from women on hormonal therapies, a total of 260,576 cells from 30 donors without endometriosis and 19 endometriosis patients were reintegrated (labeled HECA-NH). We then converted mouse genes to human orthologs in our mouse scRNA-seq dataset and projected the combined Srf f/f and Srf d/d mouse dataset onto the HECA-NH UMAP ( Fig. 6 A ). Cell types with rough equivalence between the mouse and human datasets were defined based on the UMAP projection and the descriptions in the original HECA publication, and appropriate subtypes were pooled to make comparisons between species at broad cell type levels ( SI Appendix , Table S1 ). Comparative analysis of Srf f/f and Srf d/d uterus scRNA-seq with the human endometrial cell atlas (HECA). ( A ) UMAP plot ( Left ) shows cell clustering pattern of color-coded human endometrial cell subtypes from the HECA after reintegration excluding hormone-treated donors. UMAP plot ( Right ) shows cell clustering pattern of color-coded GD3.5 mouse endometrial cell subtypes projected onto the HECA UMAP (beige background). ( B ) Violin plot displays relative SRF RNA expression from selected groups of HECA cell types split by menstrual cycle phase. Crossbar indicates mean expression. ( C ) Violin plot displays relative SRF RNA expression from selected groups of HECA cell types split by pathology status. Crossbar indicates mean expression. ( D ) Proportional Venn diagrams display the number of overlapping and nonoverlapping differentially expressed genes between the human and mouse scRNA-seq datasets from among genes with one-to-one orthologs. The tables show selected genes from among the overlaps that are either upregulated in both endometriosis patient cells and Srf d/d mouse cells compared to controls (red) or downregulated in both (blue). **** P adj 0.05, calculated via Seurat’s FindMarkers function, which uses the nonparametric Wilcoxon rank-sum test. GE, glandular epithelial; HECA-NH, human endometrial cell atlas non-hormone-treated; LE, luminal epithelial. With broadly equivalent cell types defined, we first queried SRF expression in the human data. Split by menstrual cycle phase and broad cell type, we observed that SRF was expressed in all cell types across cycle phases but most strongly, on average, in menstrual phase proliferating fibroblasts, proliferative phase proliferating epithelial cells, and menstrual phase GE cells ( Fig. 6 B ). Next, we performed comparative differential expression analyses to determine which aspects of the endometriosis-specific endometrial cell type transcriptomes (endometriosis vs. no endometriosis) were reflected in the Srf -deficient uterine cells (Srf d/d vs. Srf f/f ). As assessed by adjusted p-value (p adj ), SRF mRNA was significantly changed only in LE and proliferating epithelial cells, where it was decreased in the endometriosis group ( Fig. 6 C ). The largest overall numerical overlap of DEGs between the human and mouse data in the cell types assessed was in the LE cells, where 604 DEGs were shared, including parallel increases in noted E2-induced mucosal innate inflammatory mediators LCN2 and C3 and profibrosis genes MMP7 and FBLN1 ( Fig. 6 D and Dataset S11 ). Parallel decreases were evident in progesterone signaling–related genes GATA2 and FKBP5 and SRF-induced growth response gene EGR1 . The overlap between species for other cell types was comparatively smaller, though the GE cells in both datasets notably underexpressed LIF and HBEGF , paracrine signaling factors important for decidualization, and the nonproliferating fibroblasts in both datasets overexpressed THBS1 , SNAI1 , and SGK1 , suggesting increased TGFβ signaling ( 32 , 47 , 48 ). Together, these data underscore distinct similarities between the Srf d/d and endometriosis patient epithelial cells, especially with respect to their estrogenic, proinflammatory gene signature and their cosuppression of decidualization-critical secretory genes.

Endometriosis affects about 10% of reproductive age women, totaling an estimated 190 million worldwide ( 49 ). Other endometrial pathologies like chronic endometritis and Asherman’s syndrome are less common but still take a devastating toll through painful symptoms and fertility problems. Though these conditions vary widely in clinical presentation, common themes include inflammation, fibrosis, and aberrant steroid hormone signaling ( 4 – 7 ). In this study, we have shown that the transcription factor SRF is essential for endometrial health and function in initiation of pregnancy with relevance to human endometriosis–related infertility. SRF plays a part regulating endometrial stromal fibroblast growth response and cytoskeletal remodeling, but its most prominent role in vivo is to prevent activation of epithelium-linked innate inflammation and development of endometrial fibrosis. SRF’s molecular function has been extensively studied in mesenchymal cells of various nonuterine tissues, especially muscle cells and embryonic fibroblasts. Those studies indicate that it derives much of its transcriptional activity through context-specific cofactor interactions ( 13 ). Binding with ternary complex factors (TCFs) potentiates immediate early growth response gene expression (including the AP-1 and Egr families), whereas interaction with myocardin-related transcription factor (MRTF) family cofactors upregulates myofibroblast genes related to cytoskeletal dynamics and contractility such as the Acta family, Vcl , and Tagln ( 15 ). Our findings in HESCs regarding gene expression profile, F-Actin structure, and cell viability implicate both TCF and MRTF targets as contributors to SRF function in HESC homeostasis and in vitro decidualization response. Several of these genes were also abrogated in the Srf d/d mouse stromal fibroblasts, which fail to respond to decidual stimulus in vivo. Both TCF and MRTF family target genes have been individually linked to endometrial stromal decidualization in separate studies including AP-1 subunits, EGR1 , EGFR , and ACTA2 ( 33 , 34 , 50 – 53 ). MRTF targets are typically associated with cellular contractility, a property of myofibroblasts related to wound healing and tissue remodeling that is defined by expression of cytoskeletal and ECM genes including ACTA2 , ACTG2 , MYL9 , THBS1 , and type I collagens ( 23 , 32 ). This myofibroblast-like contractility is transiently recapitulated by decidual cells but is limited to the early decidual response, deactivating as decidualization progresses, ostensibly to permit successful embryo invasion ( 51 – 53 ). In vivo too, perturbations that cause sustained myofibroblast activation in the mouse uterus impair fertility ( 23 , 54 ). As described in vitro by Afzal et al., SRF cofactor rebalancing from MRTF dominance to TCF dominance accompanies the process of contractile phenotype reversal and is enforced by paracrine HBEGF signaling ( 21 ). Our in vitro HESC decidualization data are consistent with this study; however, we further show that adequate baseline SRF levels before hormone exposure are required to facilitate a complete decidualization response. This finding resonates with published data showing that EGR1 , an SRF-induced gene, is required to transcriptionally prime HESCs for decidualization, even though its expression is subsequently downregulated after hormone treatment ( 33 ). Our in vivo data show that in the absence of SRF, stromal fibroblasts lose Egr1 expression and subsequently fail to decidualize. Therefore, we postulate that to promote healthy decidualization, stromal SRF activity is required to activate TCF-associated growth response but must be appropriately tempered so as not to promote excessive MRTF-associated myofibroblast activation. Past study of SRF function in epithelial cells has been more limited. Several investigations have focused on gastrointestinal cells and tissues, finding that SRF promotes gastric epithelial cell and hepatocyte proliferation and regeneration after injury ( 55 ). In epithelial-origin cancer cells, SRF depletion typically reduces cell proliferation and viability, whereas overexpression tends to enhance invasiveness, motility, and epithelial to mesenchymal transition (EMT) ( 56 – 58 ). Other studies have suggested roles for SRF in apoptotic regulation of mammary epithelial cells, proliferative control of keratinocytes, and promotion of renal fibrogenesis with SRF overexpression ( 13 , 59 , 60 ). Here, we reveal the previously unreported phenomenon of epithelial innate inflammatory factor production and myeloid immune cell recruitment in response to Srf ablation. One of the most striking phenotypes of the Srf d/d mouse was the endometrial fibrosis that began before 8 wk of age and dramatically worsened with further aging. While SRF has been frequently tied to the development of fibrosis of various tissue types, it has been through the paradigm of SRF overexpression causing fibroblast activation and subsequent fibrogenesis ( 61 – 63 ). However, our study is unique in its identification of SRF depletion—not overexpression—as a driver of fibrosis, paradoxically concurrent with downregulation of typical myofibroblast markers like Acta2 , Actg1 , Vcl , and Myl9 in Srf -deficient endometrial fibroblast populations. Our data suggest that paracrine interactions between epithelial, immune, and fibroblast cell populations are required to synergistically prompt excessive collagen deposition in the absence of endometrial Srf . Since SRF is not ablated in the immune cells of the Srf d/d uterus, the inflammatory phenotype is likely due to paracrine signaling from endometrial epithelial cells and fibroblasts where SRF is depleted. Single-cell ligand–receptor analysis identified excessive epithelial innate inflammatory signaling from Srf d/d epithelial cells to macrophages and neutrophils and, in turn, increased TGFβ signaling from macrophages to stromal fibroblasts. Epithelium–macrophage crosstalk in endometriosis patient samples involving C3 and MIF pathways was recently reported based on spatial transcriptomic analysis, and MIF receptor (CD74) overexpression has been independently shown in endometrium from endometriosis patients ( 64 , 65 ). Moreover, complement signaling and increased macrophage numbers are players in both lung and endometrial fibrosis contexts ( 7 , 27 , 32 , 66 , 67 ). In addition to macrophage activity, increased neutrophilic activity also occurs in endometriosis, and therapeutic targeting of neutrophil-attracting cytokines has been successfully used to reduce inflammation and fibrosis in a primate model ( 28 , 68 ). Thus, the endometrial fibrosis we identified in Srf -deficient conditions is presumptively mediated by the hyperinflammatory tissue environment. It remains a matter for future study to determine the direct mechanism by which Srf deletion prompts epithelial overexpression of mucosal inflammatory factors including Lcn2 , Ltf , C3 , Mif , Cxcl17 , and Mmp7 , as we found in Srf d/d LE cells. C3 and complement signaling was one of the major differential interactions in our single-cell ligand–receptor analysis, unique to the Srf d/d LE-macrophage and LE-neutrophil communication patterns. C3 , in addition to Lcn2 and Ltf are known to be E2-inducible in endometrial epithelial cells ( 39 , 40 ). Furthermore, elevated estrogenic action and suppressed P4 activity can drive endometrial inflammation, particularly neutrophil infiltration like we observed in the Srf d/d mouse uterus ( 25 , 69 ). Dysregulation of epithelial E2 and P4 signaling in the absence of SRF is therefore one potential mechanism for the inflammatory conditions we identified in the Srf d/d uterus. In support of this possibility, we observed the overexpression of other estrogen-induced epithelial genes in the Srf d/d LE that closely mapped onto the human endometriosis–effected LE gene signature, including Greb1 , which was recently identified as a driver of E2-driven endometriosis pathophysiology ( 70 ). However, the elevated epithelial inflammation in our Srf d/d epithelial cells cannot be explained exclusively by elevated E2 signaling since innate immune effector MIF is suppressed rather than activated by E2 ( 71 , 72 ). MIF signaling was one of the key interactions identified between Srf d/d epithelium and macrophages in our single-cell ligand–receptor analysis. Furthermore, circulating E2 and P4 levels and endometrial ESR1 and PGR protein levels were not significantly different between Srf f/f mice and Srf d/d mice under 9 wk of age in our study. Therefore, any role played by SRF in modulating steroid hormone signaling is likely through effects on the cell type–specific genomic occupancy of steroid hormone receptors, changes in the stability of larger transcription factor complexes, or alterations in chromatin accessibility, for example. An alternative possibility is that the inflammatory response in Srf d/d uteri is caused by cellular stress from the disruption of actin filament and/or proliferative growth response gene expression outputs, as these genes are known to be direct transcriptional targets of SRF. This type of cell stress can prompt a senescence-associated secretory phenotype, generating an inflammatory response ( 73 ). Though not the focus of this investigation, we found that in vivo Srf deletion in Pgr -positive cells of the mouse disrupted oviductal function and caused premature decline of ovarian function. We showed that SRF was widely expressed in the Srf f/f oviduct but ablated throughout the Srf d/d oviductal mesenchyme and epithelial cells of the isthmus, suggesting a local impact of SRF ablation on embryo transport. PGR signaling performs noteworthy functions in controlling the oviductal inflammatory environment and oviductal smooth muscle contraction ( 19 , 24 ), functions that potentially interact with SRF activity in light of our data. In the ovary, we found that SRF was widely expressed in controls regardless of age but was ablated in Srf d/d CLs whenever they formed. Nevertheless, serum P4 levels and embryo counts remained comparable between 6- and 8-wk-old Srf f/f and Srf d/d mice, only declining in older Srf d/d mice when normal CLs no longer formed and estrous cyclicity was disrupted. This suggests that SRF is dispensable for luteal cell function once CLs form but plays a role in initial luteinization in older mice. We speculate that this broader age-dependent endocrine defect in Srf d/d females may stem from progressive dysfunction of the anterior pituitary, where the Pgr Cre is known to be active ( 22 ). These findings reinforce the need for further study of SRF in hormonally responsive cell types and in reproduction generally. In summary, our data that show macrophage and neutrophil recruitment in the Srf -deficient uterus coinciding with fibrosis development bolster the concept of innate inflammation as a driver of fibrogenesis in the setting of endometrial pathologies. Furthermore, the E2-activated, P4-suppressed epithelial gene expression pattern incited by Srf ablation implicates the involvement of steroid hormone signaling dysregulation in excessive inflammation and ECM deposition. This is a paradigm ripe for further study in the context of infertility-related endometrial pathologies as well as disease processes in other hormonally responsive tissues, and it stresses that efforts to develop antifibrotic therapeutics blocking the activity of SRF-controlled pathways must consider potential for unanticipated adverse effects. SRF is indispensable for female fertility, and insufficient SRF activity incites pathological inflammatory fibrosis of the endometrium.

All use of human donor tissues was performed in accordance with Institutional Review Board (IRB)-approved study protocols. Specifically, formalin-fixed, paraffin-embedded human endometrial tissues used to examine SRF expression via IHC were collected by endometrial biopsy from women with and without endometriosis using approved protocols at Atrium Health Wake Forest Baptist (IRB00067885, IRB00057549, and IRB00080450) and from previous archival samples also collected with IRB approval at Prisma Health, Greenville, SC. Study design and conduct complied with all relevant regulations regarding use of human study participants and was conducted in accordance with the criteria set by the Declaration of Helsinki with written informed consent received from all participants. All samples were taken during secretory phases of the menstrual cycle as assessed by histologic dating according to the criteria of Noyes et al. ( 74 ) and confirmed by histopathological examination by an experienced fertility specialist (B.A.L.). Control endometrial tissues (n = 19) were obtained from normally cycling donors without infertility who were free of hormones for at least 60 d and laparoscopically negative for endometriosis. Endometriosis-related infertility cases (n = 16) were undergoing endometrial sampling prior to surgical removal of endometriosis and were negative for uterine leiomyoma and adenomyosis. All donors were between 18 and 50 y of age. Primary ME-HESCs were obtained from deidentified archives previously banked after noninvasive isolation from menstrual fluids of healthy volunteers with the NIEHS Clinical Research Unit (n = 3) according to IRB-approved, published methods ( 75 ) and analyzed at passage 6. T-HESCs were purchased from American Type Culture Collection (#CRL-4003), and experiments were performed in triplicate at a minimum. All HESCs were cultured in DMEM/F12 (Gibco #11320033) supplemented with 10% fetal bovine serum (FBS, Gibco #10082147) and 1% penicillin–streptomycin (Sigma #P0781) at 37 °C in a 5% CO 2 humidified incubator. All experiments involving animals were approved by the NIEHS Comparative Medicine Branch Animal Care and Use Committee and conducted in accordance with applicable guidelines. Survival surgeries and supportive care were provided by Comparative Medicine Branch Veterinarians. Conditional Srf knockout mice were generated by breeding Pgr Cre/+ mice ( 22 ) (B6.129-Pgr tm2(Cre)Lyd , in-house colony) with Srf flox/flox (Srf f/f ) mice ( 14 ) (B6.129S6-Srf tm1Rmn /J, Jackson Laboratory #006658) to produce Pgr Cre/+ Srf flox/flox (Srf d/d ) conditional knockout and Srf f/f control mice. Mice were housed in the NIEHS vivarium on a 12-h light–dark cycle and fed Lab Diet 5KON/5K20 diet ad libitum. Additional experimental details are described in SI Appendix .

Appendix 01 (PDF) Dataset S01 (XLSX) Dataset S02 (XLSX) Dataset S03 (XLSX) Dataset S04 (XLSX) Dataset S05 (XLSX) Dataset S06 (XLSX) Dataset S07 (XLSX) Dataset S08 (XLSX) Dataset S09 (XLSX) Dataset S10 (XLSX) Dataset S11 (XLSX)