Extracellular matrix composition and stiffness regulates progesterone responsiveness and decidualization

D.K. carried out the experiments and analyzed data; L.H. assisted with and guided the collection of de-identified patient samples, D.A. supervised the work. D.K., L.H., and D.A. wrote the manuscript with contributions from all authors.

To determine whether the anatomic location of the lesion alters progesterone responses in endometriosis, we combined and analyzed two publicly available single-cell RNA sequencing datasets ( GSE213216 , 30 GSE179640 31 ). To ensure independent biological replicates and prevent pseudoreplication (i.e., artificial inflation of sample size), only one lesion per patient was included. This design is meant to avoid a hierarchical data structure in which multiple samples from the same individual could bias location-based comparisons. Non-endometriosis endometrium samples were used as controls from a separate dataset. A full list of samples included in the analysis is included in Table S1 . Given that endometriosis lesions are characterized by enrichment of the stromal compartment, 31 we focused our analysis on this cell population. Uniform manifold approximation projection (UMAP) of the whole dataset shows a transcriptionally diverse stromal population ( Figure 1 A), defined by increased expression of common fibroblast markers and a lack of PTPRC , EPCAM , and PECAM1 ( Figure 1 B). Performing UMAP dimensionality reduction on stromal cells alone showed a clear separation between control and endometriotic phenotypes ( Figure 1 C). The two distinct control clusters originate from endometrium and healthy ovaries of patients without endometriosis. These clusters are clearly separated, showing distinct stromal phenotypes between healthy tissues ( Figure 1 D). Meanwhile, endometriotic stromal cells from different locations display greater transcriptional overlap. Figure 1 Public scRNA-seq analysis shows stiff substrates are defined by lowered decidual and laminin-binding responses (A) UMAP of whole lesions ( n = 11), endometrium of subjects without endometriosis ( n = 3), and healthy ovary ( n = 1). (B) Identification of major cell types by expression of canonical markers. (C) UMAP of just the stromal compartment, colored by disease state and (D) location. (E) Diagram showing range of stiffnesses of various locations, of which endometriotic lesions are commonly found. Young’s moduli of 5-26 kPa are considered soft, 26 , 32 , 33 30 kPa middle, 34 and >100 kPa stiff. 35 , 36 (F) Differences in enrichment scores of PR-B gene sets and PR-A gene sets. Individual points represent patients, plotted atop group median and violin plots. Kruskal-Wallis test with Benjamini correction, ∗ = p < 0.001 and ∗∗ = p < 0.0001. (G) Heatmap of genes promoted by PR-A and (H) PR-B isoform binding to transcription factor FOXO1, associated with the inhibition and promotion of decidualization, respectively. Values based on differential expression compared to endometrium for each location. Hierarchical clustering used for locations and genes. Enrichment scores based on relative expression of genes downstream of integrin recognition of (I) laminin and (J) fibronectin. OE and ovary consistently show the lowest scores across both ligands. Laminin scores are highest within the healthy endometrium, followed by all intermediate locations, then the stiffest locations. In contrast, fibronectin scores are highest in healthy endometrium, followed by USL, then intermediate stiffness locations. ACDS, anterior cul-de-sac; OE, ovarian endometriosis; PCDS, posterior cul-de-sac; PW, pelvic wall; USL, uterosacral ligament. For (I and J), violin plots represent individual cells with Kruskal-Wallis test with Bonferroni correction, ∗ = p < 0.001 and ∗∗ = p < 0.0001. Public scRNA-seq analysis shows stiff substrates are defined by lowered decidual and laminin-binding responses (A) UMAP of whole lesions ( n = 11), endometrium of subjects without endometriosis ( n = 3), and healthy ovary ( n = 1). (B) Identification of major cell types by expression of canonical markers. (C) UMAP of just the stromal compartment, colored by disease state and (D) location. (E) Diagram showing range of stiffnesses of various locations, of which endometriotic lesions are commonly found. Young’s moduli of 5-26 kPa are considered soft, 26 , 32 , 33 30 kPa middle, 34 and >100 kPa stiff. 35 , 36 (F) Differences in enrichment scores of PR-B gene sets and PR-A gene sets. Individual points represent patients, plotted atop group median and violin plots. Kruskal-Wallis test with Benjamini correction, ∗ = p < 0.001 and ∗∗ = p < 0.0001. (G) Heatmap of genes promoted by PR-A and (H) PR-B isoform binding to transcription factor FOXO1, associated with the inhibition and promotion of decidualization, respectively. Values based on differential expression compared to endometrium for each location. Hierarchical clustering used for locations and genes. Enrichment scores based on relative expression of genes downstream of integrin recognition of (I) laminin and (J) fibronectin. OE and ovary consistently show the lowest scores across both ligands. Laminin scores are highest within the healthy endometrium, followed by all intermediate locations, then the stiffest locations. In contrast, fibronectin scores are highest in healthy endometrium, followed by USL, then intermediate stiffness locations. ACDS, anterior cul-de-sac; OE, ovarian endometriosis; PCDS, posterior cul-de-sac; PW, pelvic wall; USL, uterosacral ligament. For (I and J), violin plots represent individual cells with Kruskal-Wallis test with Bonferroni correction, ∗ = p < 0.001 and ∗∗ = p < 0.0001. Initially, we hypothesized that differences in stromal phenotypes are caused by shared mechanical characteristics that differ from healthy endometrium. Reported stiffness values for pelvic tissues vary depending on the measurement technique, and thus absolute values are not directly comparable to each other. However, studies consistently demonstrate large differences between tissues that often serve as the locations for endometriosis lesions. For example, shear wave elastography (SWE) measurements of pelvic sidewalls and cul-de-sac regions have reported similar stiffness values of 30 kPa, 34 while ovarian tissues have been reported to be much stiffer ( Figure 1 E). Although the central follicles of healthy ovaries are known to be soft (3 kPa), 35 ovarian endometriosis (OE) forms first upon the surface of the ovary, which has a Young’s modulus of 100 kPa. 35 Although there are no published SWE measurements of uterosacral ligaments (USLs), uniaxial tensile testing has reported Young’s moduli exceeding 6 MPa. 36 These values are notably stiffer compared to the endometrium, whose SWE stiffness ranges from 12 to 26 kPa depending on cycle stage. 32 , 37 , 38 Based on these consistent order-of-magnitude differences, we hypothesized that location stiffness predicts stromal phenotypes, with stiffer locations encouraging progesterone resistance among fibroblasts compared to healthy and softer endometrium. Given that a key difference between lesions and endometrium is hormonal receptivity, we investigated indicators of decidualization across lesion locations. Decidualization is the progesterone-dependent differentiation specific to endometrial fibroblasts and is dynamically regulated by two PR isoforms: PR-A, which slows, and PR-B, which enhances decidual differentiation 18 mediated by the transcription factor FOXO1. Balanced expression of these isoforms is essential for appropriate progesterone responsiveness, and disruptions in this balance have been previously implicated in endometriosis. To determine lesion-specific patterns of PR-FOXO1 signaling, we curated gene sets downstream of PR-A—FOXO1 and PR-B—FOXO1 interactions ( Table S2 ). We then calculated single-cell enrichment scores for each gene set using UCell, 39 which ranks all genes by expression within a cell, then calculates a score reflecting the activity of the genes listed in the given gene set. For each cell, we then computed the PRB-PRA difference score by subtracting the PRA enrichment score from the PRB enrichment score. Thus, a more positive score indicates PR-B-dominant signaling, while a negative score indicates PR-A-dominant signaling. The resulting PRB–PRA difference score was significantly lower in only the stiffest ectopic sites ( Figure 1 F), suggesting a stiffness-dependent shift toward PR-A-dominant signaling. To investigate this further, we performed differential expression analysis of decidualization-inhibiting ( Figure 1 G) or decidualization-promoting ( Figure 1 H) programs between endometrium and each ectopic location. Most ectopic locations showed a mixed response of PR-A targets relative to healthy endometrium, while OE and USL locations showed a moderate increase in the inhibitory program ( Figure 1 G). The upregulation of IGFBP1 and PRL indicates that some progesterone-responsive genes remain functional. However, in OE, the loss of WNT4 -related programs alongside WISP1 and CRISPLD1 , integral for ECM adhesion and remodeling, hints at failure to initiate essential downstream decidualization programs. Meanwhile, USL lesions show a loss of WNT4 and PRL , alongside upregulation of WISP1 and CRISPLD1 . This signature shows enhanced ECM remodeling indicative of fibroblast activation rather than decidualization. Lesions upon the cul-de-sacs show inverted transcriptional signatures compared to those on stiff matrices, suggesting a matrix-remodeling, non-secretory phenotype. Thus, while all locations show some progesterone responsiveness, the overall transcriptional profile supports a shift away from decidualization and toward ECM remodeling phenotypes. When examining PR-B targets that correspond to decidualization, all locations showed unique but mixed transcriptional programs. OE fibroblasts show lower PDLIM1 , FOXO1 , and CDKN2B , suggesting impaired commitment to differentiation. The increase in IGFBP1 , CDKN1B , and BIRC3 may be explained by alternative activation of survival- and stress-induced programs. It also shows increases in transcription factors critical for promoting migration and ECM remodeling, including ETS1 and SNAI1 . Together, this supports dampened decidual programs in favor of sustained proliferation and tissue remodeling. In contrast, PW shows comparable levels of FOXO1 and IGFBP1 to the endometrium, and increased levels of transcription factors to support differentiation: CEBPA , PAX6 , HNF4A . Increases in CDKN2B along with downregulated TEAD4 support exit from the cell cycle, also indicative of differentiation. Together, this program shows dampened proliferation and decidualized states. ACDS again shows the inverse fingerprint of OE, instead showing dramatically lower CDKN2B , IGFBP1 , CDKN1B , BIRC3, and unchanged levels of PDLIM1 , FOXO1, and GZMA . Overall, this shows that FOXO1 is present but shows no ability to promote progesterone-dependent cell-cycle arrest and survival programs that normally accompany effective decidualization. Finally, USL fibroblasts show enhanced levels of FOXO1 , IGFBP1, and BIRC3 alongside PDLIM1 , CDNK2B , CDKN1B , and GZMA reduction. Enhanced FOXO1 and IGFBP1 show early differentiation but again a lack of commitment to cell-cycle exit. Together, these data indicate that while lesions maintain some degree of hormonal responsiveness, the level of response is dependent on location and conflicting survival cues. These findings mirror the clinical reality in which progestin efficacy varies widely across patients, 40 suggesting that microenvironment transcriptional programs may determine decidualization capacity. Because impaired decidualization was seen on intermediate as well as stiff substrates, we next examined whether variations in ECM composition might also influence progesterone responses. To test this, we calculated enrichment scores for integrin-dependent gene programs responsive to fibronectin, laminin, and collagen ( Table S3 ). Across all ECM components, endometrium showed the highest response scores ( Figures 1 I and 1J; Figure S1 H). Collagen responses only showed decreases within ACDS, OE, USL, and healthy ovary sites ( Figure S1 H), whereas laminin and fibronectin showed alterations across all ectopic sites. The stiffer ectopic sites (i.e., OE and USL) displayed lower laminin response scores, followed by middle stiffness sites ( Figure 1 I), while USL had the highest fibronectin score across all ectopic sites ( Figure 1 J). These findings suggest that certain ECM ligands are more varied depending on location, and that these variations, rather than stiffness alone, may reinforce ectopic stromal phenotypes. To directly investigate stiffness and ECM effects on progesterone responses, we use a physiologically relevant, polyacrylamide substrate with a Young’s modulus of 4 kPa for an in vitro monolayer decidualization assay. This value was chosen to be representative of the stiffest values found upon the endometrium through direct nanoindentation, 38 which occurs during the luteal phase, when progesterone initially peaks. 37 Successful initiation of the decidualization pathway was confirmed on plastic by increased expression of FOXO1, PR, and pAkt after 96 h of estradiol, cAMP, and progesterone (ECP) treatment ( Figure S2 A). However, 96 h timepoints on 4 kPa gels lost significant differences ( Figure S2 B) and instead showed an increase in protein and secretory markers at 72 h timepoints ( Figures S2 C and S2D), which is the timepoint used for further experiments. We confirmed that gels do not sequester treatment components from human uterine fibroblasts (hUFs). In doing so, we are confident that any observed phenotype differences are a result of biological mechanisms rather than differences in hydrogel interactions with media ( Figure S2 E). To clarify the effect of fibronectin on stiffer substrates, we examined whether stiffness alone is sufficient to inhibit early progesterone responses in vitro. On soft fibronectin-coated hydrogels, canonical decidual markers FOXO1, total PR, WNT4, and PRL showed slight but significant increases when exposed to decidualization cocktail treatment ( Figure 2 ). FOXO1, total PR, and WNT4 showed median fold change (FC) increases of 1.5, 2.5, and 1.2, respectively, while PRL levels averaged to about 150 pg/mL. While significant, these increases are relatively low and indicate early entry rather than completion of decidualization. 41 , 42 Interestingly, despite total PR increase, relative proportions of PRB and PRA did not show any significant changes in response to treatment nor stiffness. Cells also exhibited increased activation as indicated by pAkt, early decidual programs via PLZF transcription factor, and increased presence of PR co-regulator SRC2 with FCs of 1.5, 1.5, and 3.5, respectively. Figure 2 Cells on stiff fibronectin substrates lose early progesterone receptivity hUFs cultured on fibronectin-coated matrices with Young’s moduli of 4 kPa (black) or 100 kPa (gold) were exposed to estradiol (E) with or without decidualization cocktail (ECP) for 72 h. Canonical markers of decidualization (FOXO1, PR, pAKT, WNT4, PRL) and progesterone receptor co-modulators (PLZF, SRC2) were measured via flow cytometry and reported as fold change (FC) relative to E-treated hUFs on 4 kPa collagen-coated gels (not shown). Data are presented as median with 95% confidence interval analyzed using the aligned Rank Transform ANOVA with Tukey post-hoc. ∗ = p < 0.05, ∗∗ = p < 0.01, ∗∗∗ = p < 0.001, and ∗∗∗∗ = p < 0.0001. Cells on stiff fibronectin substrates lose early progesterone receptivity hUFs cultured on fibronectin-coated matrices with Young’s moduli of 4 kPa (black) or 100 kPa (gold) were exposed to estradiol (E) with or without decidualization cocktail (ECP) for 72 h. Canonical markers of decidualization (FOXO1, PR, pAKT, WNT4, PRL) and progesterone receptor co-modulators (PLZF, SRC2) were measured via flow cytometry and reported as fold change (FC) relative to E-treated hUFs on 4 kPa collagen-coated gels (not shown). Data are presented as median with 95% confidence interval analyzed using the aligned Rank Transform ANOVA with Tukey post-hoc. ∗ = p < 0.05, ∗∗ = p < 0.01, ∗∗∗ = p < 0.001, and ∗∗∗∗ = p < 0.0001. In contrast, stiff hydrogels show present but dampened PR engagement and PRL secretion (1.5 FC and 50 pg/mL, respectively) and a loss of FOXO1 and WNT4 responses, indicating a dampened entrance into the decidualization pathway. This suggests that while initial progesterone signaling and kinase activation still occur, stiff conditions dampen responses to decidualization cues that promote FOXO1 but not PLZF expression. This unresponsiveness is not due to differences in total PRs nor their isoforms, as baseline marker levels are comparable across stiffness groups. Rather, stiffness interferes with coreceptor expression, as indicated by significantly lower SRC2 levels before and after treatment on stiff compared to soft hydrogels (0.75 and 1.0 FC relative to cells on soft fibronectin, respectively). Together, these findings match patterns observed in stiff lesion sites from our single-cell analysis, in which fibroblasts isolated from stiff lesion sites only exhibited partial decidual activation. However, stiffness alone did not fully explain the impaired decidualization observed in lesions, suggesting that ECM composition may further alter these effects. We next compared the effects of decidualization cocktail treatment between collagen and laminin ( Figure 3 A) and between collagen and fibronectin ( Figure 3 B) upon soft hydrogels. Across all ECM ligands, we see the induction of PRL secretion but not WNT4 expression, showing entry but not the stabilization of decidualization. However, we also see ligand-dependent modulation of decidual signaling. Figure 3 ECM cues tune progesterone responses on soft hydrogels hUFs cultured on 4 kPa hydrogels coated with either collagen, laminin (A), or fibronectin (B) after exposure to E or ECP for 72 h. Flow cytometry data are reported as FC relative to E-treated hUFs on 4 kPa collagen-coated gels. ELISA data are reported as raw prolactin (PRL) concentration in pg/mL. E = Estradiol, ECP = Estradiol, cAMP, and progesterone. Data are presented as median with 95% confidence interval analyzed using the aligned Rank Transform ANOVA with Tukey post-hoc. ∗ = p < 0.05, ∗∗ = p < 0.01,∗∗∗ = p < 0.001, and ∗∗∗∗ = p < 0.0001. ECM cues tune progesterone responses on soft hydrogels hUFs cultured on 4 kPa hydrogels coated with either collagen, laminin (A), or fibronectin (B) after exposure to E or ECP for 72 h. Flow cytometry data are reported as FC relative to E-treated hUFs on 4 kPa collagen-coated gels. ELISA data are reported as raw prolactin (PRL) concentration in pg/mL. E = Estradiol, ECP = Estradiol, cAMP, and progesterone. Data are presented as median with 95% confidence interval analyzed using the aligned Rank Transform ANOVA with Tukey post-hoc. ∗ = p < 0.05, ∗∗ = p < 0.01,∗∗∗ = p < 0.001, and ∗∗∗∗ = p < 0.0001. Soft laminin-coated hydrogels encouraged cell activation and priming even in the absence of progesterone, showed increased activation and stability, but had lower secretory responses compared to collagen. hUFs cultured on laminin exhibited significantly 2 FC higher levels of total PR, SRC2, and pAkt in response to E2 alone compared to collagen, indicating a heightened basal activation state. Upon treatment with the decidualization cocktail, both ECM ligands showed expected increases in FOXO1, pAkt, PLZF, and PRL markers. Of these markers, both FOXO1 and pAkt show heightened levels after treatment on laminin compared to collagen, indicating increased engagement in early differentiation signals by increasing from 2 to 5 FC and 2 to 12 FC relative to collagen. Conversely, PRL levels after treatment are significantly lower on laminin at just 45 pg/mL, indicating a decreased late decidual program. While laminin failed to induce increases of PR and SRC2 after treatment, this may be due to higher basal levels limiting induction of further receptors rather than indicating decreased PR engagement with transcription factors. This idea is supported by the increase in both FOXO1 and PLZF transcription factors, which require PR for induction. Laminin also shows trends of increased PRA isoforms compared to collagen. While not significant, the median 2.5FC may be enough to slow initial responses compared to collagen, which would explain the lower PRL induction. Additionally, while WNT4 expression in cells upon laminin did not increase upon treatment, WNT4 did show 1.3 FC increase compared to collagen, perhaps indicating the initiation of decidual stabilization. Together, these results suggest that while both ligands allow for decidualization entry, laminin may slow downstream secretory signaling in favor of upstream and stabilization signaling. In contrast, soft fibronectin-coated hydrogels dampened early FOXO1 but not PR induction relative to collagen. Under E2 treatment, fibronectin reduced pAkt expression at FC of 0.5, indicating lower basal activation. Following decidualization cocktail treatment, cells on fibronectin-coated hydrogels failed to significantly upregulate FOXO1, despite increases in PR (1.5–2.7FC), PLZF (1–1.3FC), SRC2 (1 to 3FC), pAkt (1 to 2FC) and PRL (120 pg/mL). These findings suggest that fibronectin impairs the ability of hUFs to induce FOXO1 expression through both FOXO1-PR and cAMP-CREB transcription pathways. Both collagen and fibronectin showed similar levels of PR isoforms, so the lack of FOXO1 induction may not be due to imbalanced isoform ratios. This is further supported by the increase in PLZF, which would be promoted by PRB isoforms. PRL would then still be able to increase through the cAMP-CREB pathway. 43 Although FOXO1 and PR levels were not significantly lower than the collagen ECP group, overall pAkt remained reduced, reflecting diminished activation. Overall, these results again highlight that ECM identity influences progesterone responses under soft mechanical cues, with laminin encouraging early programs dependent on FOXO1, and fibronectin bypassing FOXO1 in favor of cAMP-dependent pathways. Given that stiffness constrains progesterone responsiveness, we next asked whether these ECM-dependent patterns are maintained or lost under stiff conditions. Stiff conditions greatly altered overall progesterone responsiveness, though ligand identity continues to influence signaling patterns. On stiff collagen, progesterone treatment shows increased upstream markers such as pAkt (15FC), PLZF (1.5FC), PR (1.3FC), and SRC2 (1.2FC), but decreased FOXO1 (0.8FC) and slight changes in downstream markers PRL (25 pg/mL), and WNT4 (1.2FC) ( Figure 4 A). Overall, this shows a critically low induction of decidual programs. Similarly, hUFs on stiff laminin hydrogels showed increased pAkt (1.8FC), PLZF (1.5FC), PR (1.6FC), and SRC2 (2FC) in response to progesterone, but neither WNT4, PRL, nor FOXO1 expression changed significantly. Notably, laminin significantly reduced WNT4 and pAkt compared to collagen, overall indicating an activation of upstream but not downstream signals. These patterns show that stiffness overrides ligand-specific cues in FOXO1 induction and basal PR levels and dampens downstream secretory signals. Figure 4 Stiff substrates inhibit decidualization responses on all ligands, though each ligand shows different dynamics hUFs cultured on 100 kPa hydrogels coated with either collagen vs laminin (A) or collagen vs fibronectin (B) after exposure to E or ECP for 72 h. Flow cytometry data are reported as FC relative to E-treated hUFs on 100 kPa collagen-coated gels. E = Estradiol, ECP = Estradiol, cAMP, and progesterone. Data are presented as median with 95% confidence interval, analyzed using the aligned Rank Transform ANOVA with Tukey post-hoc. ∗ = p < 0.05, ∗∗ = p < 0.01, ∗∗∗ = p < 0.001, and ∗∗∗∗ = p < 0.0001. Stiff substrates inhibit decidualization responses on all ligands, though each ligand shows different dynamics hUFs cultured on 100 kPa hydrogels coated with either collagen vs laminin (A) or collagen vs fibronectin (B) after exposure to E or ECP for 72 h. Flow cytometry data are reported as FC relative to E-treated hUFs on 100 kPa collagen-coated gels. E = Estradiol, ECP = Estradiol, cAMP, and progesterone. Data are presented as median with 95% confidence interval, analyzed using the aligned Rank Transform ANOVA with Tukey post-hoc. ∗ = p < 0.05, ∗∗ = p < 0.01, ∗∗∗ = p < 0.001, and ∗∗∗∗ = p < 0.0001. We next asked whether collagen and fibronectin differ in the molecular machinery underlying their inability to support decidualization on stiff hydrogels ( Figure 4 B). In contrast to Figure 3 , no significant differences were detected between cells cultured on stiff collagen and fibronectin hydrogels. This significantly implicates mechanotransduction pathways in their ability to override ligand cues. Both stiff matrices again show slight induction of decidual responses, with increases in PR (1.5FC), SRC2 (2.5 and 1.5FC), and PRL (25 pg/mL and 48 pg/mL), but not commitment through substantial WNT4 induction. Thus, stiffness suppresses or at least slows decidualization, independent of ECM identity. To assess whether our findings are relevant to gynecological pathologies, we analyzed ECM ligand enrichment in patient samples of endometriosis compared across regions of differing stiffness. Within images, CD10 + areas were used to define lesion areas from surrounding tissue, as done previously ( Figures 5 A and 5B). Masks encompassing CD10 regions ( Figure S4 ) were then used to quantify collagen, laminin, and fibronectin. Interestingly, areas enriched with CD10 in some lesions showed significantly altered ECM composition, as shown in the USL showed in Figure 5 B. Proportions of ECM ligands show no significant differences across stiffness groups ( Figure 5 C). However, total intensity of ECM ligands normalized to the area of the CD10 + region consistently shows enrichment of fibronectin in stiff compared to soft lesion sites ( Figure 5 D). Figure 5 Lesions are enriched in fibronectin regardless of location, but show lowest decidualization markers on stiff substrates Immunofluorescent (IF) analysis of deidentified human endometriotic tissues, n = 3 for each stiffness group. Scale bars, 100 μm. Stiffness groups are defined as follows: Stiff- USL, Mid- ACDS, PW, retrocervical, Soft-sigmoidal colon, perimetrium, ovarian fossa. (A) Representative images of uterine fibroblast marker CD10, collagen 1, and laminin or (B) fibronectin. (C) Proportions of ECM ligands within CD10 + masks. Signals not overlapping CD10 markers are not included in quantitative analysis. (D) Total intensity of ECM ligands over the area of the CD10 + region. Focus areas are defined by separate images of the same biopsy, but of different CD10 + locations. Number of focus areas vary per biopsy, ranging from 1 to 4. Data are presented as mean +/− SEM, and C through D analysis used one-way ANOVA and Tukey’s post-hoc tests. (E) Representative images of uterine fibroblast markers CD10, FOXO1, and PR. Signals not overlapping CD10 not included in quantitative analysis. (F) Mean fluorescent intensity (MFI) per cell of FOXO1 and PR. Violin plots of individual cell values, points average of image. Data are presented as median with 95% confidence interval and the statistics run on images use Kruskal-Wallis test and Dunn’s post-hoc. ∗ = p < 0.05, ∗∗ = p < 0.01, and ∗∗∗ = p < 0.001. Scale bars for images in A, B, and E are 100 microns. Lesions are enriched in fibronectin regardless of location, but show lowest decidualization markers on stiff substrates Immunofluorescent (IF) analysis of deidentified human endometriotic tissues, n = 3 for each stiffness group. Scale bars, 100 μm. Stiffness groups are defined as follows: Stiff- USL, Mid- ACDS, PW, retrocervical, Soft-sigmoidal colon, perimetrium, ovarian fossa. (A) Representative images of uterine fibroblast marker CD10, collagen 1, and laminin or (B) fibronectin. (C) Proportions of ECM ligands within CD10 + masks. Signals not overlapping CD10 markers are not included in quantitative analysis. (D) Total intensity of ECM ligands over the area of the CD10 + region. Focus areas are defined by separate images of the same biopsy, but of different CD10 + locations. Number of focus areas vary per biopsy, ranging from 1 to 4. Data are presented as mean +/− SEM, and C through D analysis used one-way ANOVA and Tukey’s post-hoc tests. (E) Representative images of uterine fibroblast markers CD10, FOXO1, and PR. Signals not overlapping CD10 not included in quantitative analysis. (F) Mean fluorescent intensity (MFI) per cell of FOXO1 and PR. Violin plots of individual cell values, points average of image. Data are presented as median with 95% confidence interval and the statistics run on images use Kruskal-Wallis test and Dunn’s post-hoc. ∗ = p < 0.05, ∗∗ = p < 0.01, and ∗∗∗ = p < 0.001. Scale bars for images in A, B, and E are 100 microns. We next assessed markers of decidualization within the same lesions. Lesions located in stiffer tissue environments exhibited significantly lower FOXO1 levels and a trend toward reduced PR expression. These findings mirror stiffness-dependent repression of early decidualization observed in vitro ( Figures 2 , 3 , and 4 ), as well as the lowered PR responses of OE and USL within public datasets ( Figure 1 ). Together, this work suggests that both matrix stiffening and fibronectin accumulation jointly reinforce progesterone resistance in endometriosis.

Further information and requests for reagents may be directed to and will be fulfilled by the lead contact Daniel Abebayehu ( [email protected] ). All reagents and materials will be made available upon request. The single-cell transcriptomic datasets used in this study are publicly available in the NCBI GEO database (accession nos. GSE179640 and GSE213216 ). All code used in the analysis of this paper have been deposited to and are available at Mendeley (Mendeley data: https://doi.org/10.17632/4nrrgjc932.1 ).

Our work shows that ECM composition directly influences progesterone responses in healthy hUFs ( Figures 3 and 4 ). We show that hUF progesterone responses can be tuned by ECM ligand alone, with laminin slowing secretory responses but showing enhanced stabilization through FOXO1 and WNT4, and fibronectin showing quicker but less stable secretory differentiation. Combined with previous work corroborating our findings that fibronectin is enriched 9 , 29 while FOXO1 and PR are diminished in endometriosis lesions on stiffer substrates ( Figure 5 ), our data suggest that the stiffer lesion microenvironment is sufficient to promote progesterone resistance. How this resistance develops may be directly tied to ECM ligands’ roles in wound healing and tissue development. Fibronectin is often the first matrix protein deposited in response to injury, producing a scaffold for further matrix deposition and cell invasion. Because of this, fibroblast engagement of fibronectin often promotes cell proliferation, 10 further fibronectin deposition, 44 and myofibroblast activation independent of stiffness. 23 Similarly, laminin serves as a basal ECM protein, promoting adhesion, invasion, 45 , 46 , 47 , 48 , 49 and survival. 46 , 47 , 48 , 49 However, laminin is also well established in lineage determination. 50 , 51 , 52 Specifically, laminin is known to activate notch signaling, 52 which is necessary for decidualization. 53 This suggests that while fibronectin may be able to initiate enhanced initial secretory responses, it may not be able to stabilize these responses over proliferative signals, whereas laminin primes fibroblasts for progesterone responses and stability by the induction of notch signaling. We also show that stiffer endometriosis lesion sites show increased fibronectin and the lowest decidualization markers ( Figures 1 F, 3 , 4 , and 5 F). These patterns in progesterone resistance may be further explained by the role of stiffness in the menstrual cycle and wound healing. In healthy endometrium, stiffness is cyclical to promote proliferation or differentiation depending on cycle phase. 26 , 27 In response to endometrial injury, uterine fibroblasts will exit this hormone-driven cycle to adopt wound healing phenotypes to restore uterine repair. Healthy uterine fibroblasts will then reengage with the hormone-driven cycle to reestablish uterine homeostasis. 28 , 34 Fibroblasts within endometrial disorders are unable to reenter this continuum, especially within lesions. Our consistent findings of dampened progesterone signaling on stiffer substrates demonstrate that reentry into progesterone-responsive programs is dependent on location stiffness. This is consistent with previous work that demonstrates how the phenotypes of different fibroblast populations are driven by biochemical and biomechanical ECM cues. 54 , 55 Across our data, we show varying levels of progesterone responses with partial activation signatures depending on location, stiffness, and ligand. Public datasets show increased stress responses and proliferation ( Figures 1 G and 1H) signatures consistent with an inability to engage downstream FOXO1 signaling. Decidualization experiments show similar incapability to induce FOXO1 upregulation between fibronectin hydrogels and stiff collagen and stiff laminin hydrogels. These findings are further validated through imaging analysis, in which we find that lesions from stiff locations show significantly lower FOXO1 than PR. These patterns demonstrate that deficient engagement of FOXO1, rather than PR, may be responsible for progesterone resistance in endometriosis. Additionally, the stiffness-driven effects may also occur within the fibrotic microenvironment characteristic of endometriotic lesions. Clinical studies using SWE report that lesions frequently reach stiffness values ranging from 40 to 295 kPa, 28 all significantly different from surrounding healthy areas. This elevated stiffness is consistent with enrichment of fibrotic signaling pathways, including increased expression of TGFβ1, ECM deposition, 56 , 57 and crosslinking enzymes, 58 which develop as a result of unresolved inflammation and heightened estrogen. 59 , 60 Our data suggest that this mechanical stiffening may ultimately override ECM ligand-specific differentiation cues. Although laminin engagement appears to promote the stabilization of progesterone responses through FOXO1 and WNT4 signaling, these programs are suppressed on stiff substrates across multiple ligand conditions. So, even if lesion locations are soft, if the lesion develops a stiffer matrix over time, the stiffness will act as a dominant regulator that dampens progesterone signaling programs even in the presence of differentiation-promoting ligands. A potential explanation for stiffness-driven patterns may lie in alternative survival mechanisms for stiff environments. Specifically, pAkt was consistently responsive to decidualization cocktails ( Figure 4 ) and is typically activated to prevent apoptosis in stiff environments. 61 While Akt phosphorylation occurs during different modes of cell activation, it also plays a critical role in regulating differentiation. 61 Previous work has shown that Akt directly inactivates FOXO1, 62 , 63 and that active FOXO1 is necessary for decidualization-promoting PR-B programs. 18 This stiffness-dependent mechanism shows a preference for pAkt survival programs, which allow PR but not FOXO1 engagement. Many gynecological pathologies show development of progesterone resistance within stromal compartments, but previously the link between microenvironment composition, mechanics, and progesterone responses has not been fully defined. Specifically, within endometriosis, prior work has established that inflammatory cytokines and oxidative stress contribute to both fibrosis and impaired decidualization in endometriotic fibroblasts, 59 , 60 , 64 but these studies do not address the role of tissue stiffness and composition in regulating hormonal responses. Inflammation-driven fibrosis is known to promote ECM deposition and mechanotransduction pathways, 57 , 65 yet the effect of this altered ECM on progesterone responses has not been identified. Our study identifies that both ECM composition and substrate stiffness directly alter progesterone responses by inhibiting FOXO1 engagement, and that stiffness overrides ECM signals. Ultimately, this work suggests that the biochemical and biomechanical features of the lesion microenvironment significantly impact progesterone responses. Due to the variability in patient responses to hormonal therapies, we propose that lesion location and ECM composition may be used as an indicator for hormonal therapy success after surgery. Further, investigations into nonhormonal treatments of endometriosis may be better served by looking for pathways unique to lesions in stiff locations. In doing so, we may better identify patient-specific approaches to disease management. Our integration of public datasets with in vitro assays and validation among a patient cohort shows consistent trends of stiffness-dependent inhibition of decidualization. However, further work is needed to confirm isoform-level roles in these findings. Our computational work relies on inference of PR-isoform activity rather than direct analysis. For this reason, we cannot definitively conclude that stiffness drives engagement of PR-A over PR-B isoforms, but rather that stiffness specifically inhibits FOXO1 progesterone responses. Additionally, our patient analysis relies on deidentified samples that only consider location as a variable, without considering age, phase of the menstrual cycle, or previous hormonal treatments. Future work that considers these other variables may help identify other relationships that contribute to progesterone resistance and regulate decidualization. In summary, our investigations demonstrate that both anatomical location and microenvironment ECM mechanics shape progesterone responses in endometriosis. scRNA-seq analysis across 5 lesion sites revealed distinct stromal transcriptional programs, with stiffer sites (OE and USL) exhibiting the lowest PRB-PRA difference scores (−0.4 to −0.6) and marked reductions in FOXO1 expression (median FC 0.7–0.8) relative to endometrium, indicating a shift toward PR-A-dominant signaling. Cul-de-sac lesions displayed mixed PRB-PRA programs, while pelvic wall lesions maintained partial FOXO1 activation (1.1 FC) but showed elevated proliferation and ECM remodeling signatures. In vitro , soft laminin hydrogels supported FOXO1 and WNT4 induction (up to 5 FC) and early PRL secretion (∼45 pg/mL), whereas fibronectin promoted PRL (∼150 pg/mL) via cAMP-dependent pathways despite limited FOXO1 and WNT4 engagement. In contrast, stiff hydrogels suppressed FOXO1 and WNT4 responses across all ECM ligands, limiting PRL secretion (25–50 pg/mL) even when upstream markers pAkt, PLZF, PR, and SRC2 were induced (1.5–15 FC). Patient lesion analysis corroborated these trends: fibronectin-enriched, stiff lesions consistently showed lowest FOXO1 and PRL levels, suggesting that mechanical stiffening overrides local ECM differentiation cues. Collectively, these results reveal that lesion stiffness is a dominant regulator of progesterone responses in healthy fibroblasts, impeding FOXO1-dependent decidualization programs while allowing partial PR- and survival-mediated signaling, providing a framework for the observed variability in progestin efficacy in patients.

The endometrium (i.e., the lining of the uterus) is a unique tissue that undergoes complete and cyclical regeneration with no fibrosis in normal physiological settings. 1 Endometrial disorders display a dysfunctional regenerative program within the endometrium and can promote ectopic endometrial tissue growth. Among endometrial disorders (e.g., endometriosis, adenomyosis, endometrial cancer, PCOS, and endometrial hyperplasia), there are two key common features: chronic inflammation and progesterone resistance. 2 The incidence rate of endometrial disorders among reproductive-aged women is on the rise, 3 though the true prevalence is difficult to determine due to vague symptoms, lack of awareness, and difficulty in diagnosis using current diagnostic strategies (i.e., physical examination, laboratory evaluation, and imaging). 4 Consequently, costly and invasive exploratory laparoscopic surgery remains the gold standard for definitive diagnosis, leading to an average seven-year delay in diagnosis. 5 Patients experience substantial morbidity, including reduced quality of life, increased healthcare utilization, and chronic pain. 4 In particular, endometriosis is an endometrial disorder defined by ectopic growth of endometrial tissue lesions outside of the uterus which are characterized by excessive stroma and chronic inflammation. Additionally, these lesions display noteworthy heterogeneity in location, composition, and hormone responsiveness 6 that contributes to inconsistent therapeutic outcomes and underscores the need for more targeted treatment strategies. Current therapies rely on non-specific hormone therapies, sometimes combined with surgical excision. Combined oral contraceptives (e.g., estrogen and synthetic progestins) are routinely prescribed to suppress ovulation and reduce estrogen-driven inflammation, yet a third of patients exhibit poor or no response to hormone therapy due to progesterone resistance. 7 Ectopic lesions frequently display out-of-phase hormonal activity 7 , 8 and reduced progesterone receptor (PR) expression 8 , 9 compared to patient-matched eutopic endometrium. This progesterone resistance, as well as dysfunctional PR activity, is not unique to endometriosis, but is often found in several gynecological pathologies including adenomyosis, uterine fibroids, and polycystic ovarian syndrome. 2 However, the mechanisms underlying these distinct lesion characteristics remain poorly understood. Across the gynecological pathologies listed above, diseased tissue is frequently found within a fibrotic extracellular matrix (ECM), characterized by excessive collagen deposition and increased stiffness, 7 , 10 which has the potential to influence the plasticity of fibroblast responses to hormones. In healthy endometrium, fibroblasts drive cyclic tissue remodeling that is regulated by hormone signaling. In response to estrogen, uterine fibroblasts are proliferative and motile with myofibroblast-like properties, while progesterone promotes their differentiation into a decidualized, secretory phenotype that supports epithelial growth and angiogenesis. 11 , 12 , 13 This process, termed decidualization, depends on progesterone and cyclic-AMP to counteract estrogen signaling. Adequate progesterone responses require PRs to form complexes with co-activators, SRC2 or SRC1, 14 , 15 before binding to transcription factors to initiate differentiation. The transcription factor PLZF is essential for early decidual programs, acting as a direct PR target to induce chromatin remodeling and cell cycle exit to prime for the transcription of downstream decidual markers. 16 As a result, PRs bound to other transcription factors, such as FOXO1, better target and increase expression of downstream decidualization markers including FOXO1, 13 , 17 , 18 PR, 17 WNT4, 14 and secreted prolactin (PRL). 14 Cyclic-AMP is critical to promoting this process, specifically through the PKA/CREB pathway, which can increase both FOXO1 expression and PRL secretion. 16 , 19 Traditional endpoint markers of decidualization are PRL, which marks the convergence into a secretory phenotype, and WNT4, which serves as a stability signal to maintain decidual phenotypes. 14 Meanwhile, the phosphorylation of Akt (pAkt) denotes early cellular activation and survival. 18 The inability of fibroblasts in endometrial disorders to decidualize is therefore closely linked to a loss of hormonal plasticity, potentially reinforced by the fibrotic microenvironment. It is well established that ECM stiffness 20 , 21 , 22 , 23 , 24 , 25 and ECM composition 22 can profoundly influence fibroblast phenotype and responsiveness to external cues. In the healthy uterus, matrix stiffness fluctuates dynamically—temporarily increasing after menstruation or injury to promote repair before softening to restore homeostasis. 26 , 27 In contrast, pathological fibroblasts reside within a persistently stiff 28 and compositionally altered 29 ECM. Previous work has shown that when endometriotic fibroblasts are removed from their native environment and cultured on softer substrates, they lose their inflammatory and contractile phenotype, 22 indicating that the lesion microenvironment is a key driver of pathogenic fibroblast behavior in endometrial disorders. However, it remains unclear whether lesion stiffness and ECM cues alone can inhibit decidualization in otherwise healthy uterine fibroblasts. While both progesterone resistance and fibrotic microenvironments are hallmark features of gynecological pathologies such as endometriosis, the link between these processes remains unclear. We hypothesized that ECM stiffness and composition regulate PR signaling and FOXO1 activation in uterine fibroblasts, thereby controlling their ability to decidualize. To test this, we used physiologically relevant 2D hydrogel cell culture platforms that mimic either the stiffness of the soft, healthy endometrium or the stiffer fibrotic locations, examining how mechanical and biochemical ECM cues influence early stages of decidualization in healthy fibroblasts. In doing so, we determined that ECM ligands directly modulate decidualization pathways, with fibronectin dampening and laminin enhancing progesterone responses on soft matrices. Stiffer matrices override these ECM cues, dampening decidualization across all ligands. Further, we establish that these findings mirror activity in endometriosis as a validation in endometrial disorders, showing that lesions across all locations display greater fibronectin expression and that stiffest sites show lowest decidual markers. Altogether, these data demonstrate a role for lesion microenvironment in the mechanism of inhibited decidualization and progesterone responses in gynecological pathologies.

The authors declare no competing interests.

REAGENT or RESOURCE SOURCE IDENTIFIER Antibodies Alexa Fluor 647 Conjugated anti-FOXO1 BioLegend Cat#947209; Clone: W20064D; RRID: AB_2927910 Alexa Fluor 488 Conjugated anti-PR A/B Cell Signaling Technology Cat#35591S; Clone D8Q2J Biotinylated anti-pAkt (Ser 473) Cell Signaling Technology Cat#4060T; Clone D9E Alexa Fluor 594 Conjugated Streptavidin Invitrogen Cat#S1127 Live/Dead Near-Infrared Fixable Viability dye Biolegend Cat# L34993 Rabbit anti-NCOR2 ThermoFisher Cat# PA5-141074; RRID: AB_2932526 Mouse anti-SRC1 ThermoFisher Cat# MA1-840; RRID: AB_2150919 Rat anti-WNT4 R&D Systems Cat# MAB4751; Clone 55025 Alexa Fluor 594 Conjugated anti-Rabbit IgG ThermoFisher Cat# A32740; RRID: AB_2762824 Alexa Fluor 488 Conjugated anti-Mouse IgG ThermoFisher Cat# A-11001; RRID: AB_2534069 Alexa Fluor 647 Conjugated anti-Rat IgG ThermoFisher Cat# A-21247; RRID: AB_141778 Rabbit anti-SRC2 Cell Signaling Technology Cat#96687S; Clone D2X4M Mouse anti- PRB Abcam Cat# Ab2765; Clone alpha PR6 Alexa Fluor 647 Conjugated anti-Mouse IgG ThermoFisher Cat# A-21235; RRID: AB_2535804 PE-Cy7 Conjugated anti-PLZF ThermoFisher Cat#25-9322-82; Clone 9E12 Donkey Serum Sigma Aldrich Cat#D9663 Tween 20 Sigma Aldrich Cat#P1379 Goat anti-CD10 ThermoFisher Cat# PA5-47075; RRID: AB_2607390 Alexa Fluor 488 Conjugated Donkey anti-Goat IgG ThermoScientific Cat# A32814 ; RRID: AB_2762838 Triton X- Sigma Aldrich Cat#X100 Goat Serum Vector Laboratories Cat#S-1000-20 Rabbit anti-COL1A1 Cell Signaling Technology Cat#81375SF; RRID: AB_3083034 Mouse anti-Laminin Sigma Aldrich Cat#L8271 Mouse anti-Fibronectin Sigma Aldrich Cat#F7387 Mouse anti-PR A/B Life Sciences Cat#C133381 Rabbit anti-FOXO1 Cell Signaling Technology Cat#2880S Alexa Fluor 594 Conjugated anti-Mouse IgG ThermoFisher Cat# A-11005; RRID: AB_2534073 Alexa Fluor 647 Conjugated anti-Rabbit IgG ThermoFisher Cat# A-21244; RRID: AB_2535812 DAPI BioLegend Cat#422801 Biological samples Control Endometrium scRNAseq data Tan et al. 31 GSE179640 Endometriosis scRNAseq data Fonseca et al. 30 GSE213216 Human Endometriotic Tissue Department of Obstetrics and Gynecology, University of Virginia IRB Exemption #598 Chemicals, peptides, and recombinant proteins 17beta-Estradiol (E2) Sigma Cat#3301-1 gm 8-Bromoadenosine 3’:5′-cyclic monophosphate Sigma-Aldrich Cat#B7880 Progesterone Sigma Cat#5341-25 gm Methanol Sigma Cat#34885 Collagen I Cell Guidance Systems Cat#M16S Recombinant Human Fibronectin, CF Free R&D Systems Cat#4305-FNB-200 CellAdhere Laminin-521 Stemcell Technologies Cat#77003 Critical commercial assays Human Prolactin ELISA kit Invitrogen Cat#EHIAPRL FIX & PERM Cell Permeabilization Kit ThermoFisher Cat#GAS004 Deposited data Raw and analyzed data This Paper, Mendeley Data https://doi.org/10.17632/4nrrgjc932.1 Code for scRNAseq processing and scoring This Paper, Mendeley Data https://doi.org/10.17632/4nrrgjc932.1 Human reference genome NCBI build 37, GRCh37 Genome Reference Consortium http://www.ncbi.nlm.nih.gov/projects/genome/assembly/grc/human/ Raw fluorescent images of human lesions This Paper, Mendeley Data https://doi.org/10.17632/4nrrgjc932.1 Experimental models: Cell lines Human Uterine Fibroblasts ATCC Cat#PCS-460-010 Fibroblast Basal Medium ATCC Cat#PCS-201-030 Fibroblast Growth Kit- Low Serum Lifeline Cell Technologies Cat#LS-1038 4 kPa Polyacrylamide Gel Cell Culture Plates Matrigen Cat#SW-24-EC-4 100 kPa Polyacrylamide Gel Cell Culture Plates Matrigen Cat#SW-24-EC-100 Software and algorithms Seurat v5 Hao et al. 29 – Harmony Korsunsky et al. 34 – Harmonizome Diamant et al. 35 – UCell Andreatta and Carmona 36 – OMIQ Dotmatics www.omig.ai CellProfiler v4.2.6 BMC Bioinformatics 38 – Prism v10 GraphPad, San Diego, CA – Other 3-Laser Cytek Northern Lights Aurora Spectral Flow Cytometer UVA Flow Cytometry Core https://cytekbio.com//pages/northern-lights Leica Mica Microscope Leica – Tecan Spark Plate Reader Tecan https://lifesciences.tecan.com/multimode-plate-reader Human uterine fibroblasts (hUFs) were purchased from ATCC (PCS-460-010). Cells were cultured in fibroblast basal medium (ATCC, PCS-201-030) supplemented with fibroblast growth kit-low serum (Lifeline Cell Tech, LS-1038). Cells were incubated at 37°C in humidified atmosphere of 5% CO2 for at least three passages prior to experimentation. All cells used in experiments are below passage ten. The hUF cell line was not authenticated but tested negative for mycoplasma contamination. Efficacy of treatments were first validated in hUFs cultured on tissue culture treated plastic ( Figure S2 A). hUFs were seeded and allowed to adhere for 1 h prior to introduction of either phospho-buffered saline (PBS) (Alkali Scientific, PBD500-CS), 10 μM 17beta-Estradiol (E2) (Sigma, 3301-1 gm) with or without 0.5 mM 8-Bromoadenosine 3’:5′-cyclic monophosphate (cAMP; Sigma-Aldrich, B7880) and 1 μM progesterone (P4) (Sigma 5341-25 gm). Medias were changed every 48 h for a total of 96 h, as described previously. 91 Four samples of each group were lifted for flow cytometry using TypLE Express Enzyme (ThermoFisher) and subsequently stained with the Live/Dead Near-Infrared Fixable Viability dye (Biolegend) then fixed and permeabilized using the FIX & PERM Cell Permeabilization Kit (ThermoFisher). Cells were then stained for one of three panels and analyzed via flow cytometry. For panel 1, ice-cold methanol was added to cells after fixation and left to incubate at 4°C for 10 min before spinning down and incubating in permeabilization buffer with biotinylated anti-phospho-Akt (Ser 473;Rockland, 200-306-268) for 20 min at room temperature. hUFs were then incubated in permeabilization buffer with Alexa Fluor 594 conjugated streptavidin (Invitrogen, s1127), Alexa Fluor 647 conjugated anti-FOXO1 (BioLegend 947206), and Alexa Fluor 488 conjugated anti-Progesterone Receptor (Cell Signaling Technology 35591S) for another 20 min. For panels 2 and 3, cells were incubated with primary antibody solutions diluted in permeabilization buffer, followed by a second incubation step with secondary antibodies. Panel 2 primary antibodies included anti-NCOR2 (ThermoFisher PA5-141074), anti-SRC1 (ThermoFisher MA1-840), and anti-WNT4 (R&D Systems MAB4751), and secondary antibodies were Alexa Fluor 594 anti-rabbit IgG (ThermoFisher A32740), Alexa Fluor 488 anti-mouse IgG (ThermoFisher A-11001), and Alexa Fluor 647 anti-rat IgG (ThermoFisher A-21247). Panel 3 primary antibodies included anti-SRC2 (Cell Signaling Technology 96687S) and anti-alpha PR6 (Abcam ab2765), and secondary antibodies were Alexa Fluor 594 conjugated anti-rabbit IgG and Alexa Fluor 647 conjugated anti-mouse IgG (ThermoFisher A-21235). Panel 3 also had an additional conjugated antibody incubation step with Alexa Fluor 488 conjugated anti-Progesterone Receptor and PE-Cy7 conjugated anti-PLZF (ThermoFisher 25-9322-82). Detailed breakdowns of all panels are listed in Table S6 . Cells were then analyzed on a 3-laser Cytek Aurora Spectral Flow Cytometer through the UVA Flow Cytometry Core Facility and analysis done using the OMIQ software. Progesterone receptor isoform A signal is reported as cells positive for total progesterone receptor and negative for progesterone receptor isoform B. Timepoint experiments were conducted on polyacrylamide gels (Matrigen, SW-24-EC-4) prepared with a 1-h incubation of 10 μg/mL collagen I (Cell Guidance Systems, M16S). hUFs were seeded as described above, then exposed to treatment for 24, 48, 72, and 96 h. All treatment groups were lifted and fixed at the same time, then prepared for flow as described above for panel 1. Medias were preserved and stored at −20°C for later ELISA analysis. For sequestration experiments, 4 kPa and 100 kPa polyacrylamide gels (Matrigen, SW-24-EC-4 or SW-24-EC-100) were coated with collagen, then exposed to PBS or E2, cAMP, and P4 (ECP) -supplemented medias for 24 h. Medias were then collected and used to treat hUFs seeded on tissue culture plastic for 72 h prior to flow. For stiffness and ECM analyses, 4 kPa and 100 kPa gels were prepared prior to cell seeding through a 1-h incubation with 10 μg/mL ECM ligand-either collagen I, fibronectin (R&D Systems, 4305-FNB-200), or laminin-521 (Stemcell Technologies, 77003). hUFs were seeded and allowed to adhere for 2 h prior to introduction of treatment. hUFs were then treated with E2 with or without cAMP and P4 as described above. Medias were changed every 48 h for a total of 72 h. Endpoint medias were then saved and stored at −20°C for later ELISA analysis. Triplicate samples of each group were used for each experiment and normalized against the average of the E2 group. Each experiment was repeated a total of 3 times. Prolactin supernatant concentration was determined using Invitrogen Human Prolactin ELISA kit (Invitrogen EHIAPRL) according to developer instructions. For timepoint and one round of stiffness and ECM experiments, medias were thawed within two weeks of storage and divvied into three technical replicates. Standard curves were developed using serial dilutions ranging from 500 to 7.8 pg/mL and interpolated using a sigmoidal 4 parameter model. All samples were collected according to IRB-exemption #598, and thus no identifiable data on race, sex, age were collected. Lesions were placed immediately in ice-cold DMEM with 10% FBS and 1% P/S after excision via robot-assisted laparoscopy and labeled with the anatomic location of the lesion. Within 8 h of collection, samples were transferred to 15% sucrose (ThermoFisher, 036508-A1) diluted in PBS for overnight incubation, followed by a 24-h 30% sucrose incubation. Samples were then fixed for 24 h within of 4% PFA prior to flash freezing in optimal cutting temperature medium and subsequent storage at −80C. All samples were cryosectioned within three months of collection. Public scRNA-seq datasets ( Table S1 ) were downloaded and processed using the Seurat v5 92 package. Each file was annotated with location metadata as provided by source material. 30 , 31 Filtered feature barcode matrices were imported and filtered to remove cells with less than 200 genes and greater than 20% mitochondrial reads, resulting in a counts matrix comprising of 69,309 cells. Counts were normalized with NormalizeData using log-normalization, scale-factor of 10,000. Highly variable features were identified with FindVariableFeatures using the vst method for the top 20,000 hits. Data was scaled using SCTransform with vst2 settings before performing principal component analysis with RunPCA. Batch effects and datasets were then integrated using the Harmony package 93 with default parameters. The first 15 principal components were used for FindNeighbors and FindClusters with a resolution of 3, followed by RunUMAP, resulting in a total of 68 clusters. Major cell types were determined using markers shown in Figure 1 using the default settings of the DotPlot function. Stromal cells were isolated and subclustered using the same settings as the global UMAP, resulting in 15 clusters for downstream analysis. FOXO1 downstream gene sets were downloaded as JSON files from Harmonizome 94 and parsed in R using the jsonlite package. For each ectopic location, we computed differential expression against endometrium using Seurat’s FindMarkers with Wilcoxon rank-sum and default parameters. The average log2 fold-change (avg_log2FC) values were scaled across locations prior to plotting. Heatmaps were generated using pheatmap with Euclidean distance determinants for hierarchical clustering. Overall heatmaps, and heatmaps filtered for specific gene sets ( Tables S2 and S4 ) are shown in supplementary data. For single-cell signature scoring we computed module scores from Table S3 . We used UCell 39 for rank-based, robust scoring across gene sets. Three panels were used across serial sections of lesions from n = 9 patients, 3 for each stiffness group. For all panels, 10 μm sections were blocked with 2% Donkey Serum (Sigma Aldrich, D9663) followed by overnight incubation of CD10 primary antibody (Thermo, PA5-47075). After washes with 0.1% Tween (Sigma Aldrich, P1379), samples were incubated with 1:1000 dilution of donkey anti-goat Alexa Fluor 488 (Thermo Scientific, A32814 ) for 1 h at room temperature. After washes, samples were then permeabilized with 0.1% Triton X-(Sigma Aldrich, X100) for 10 min and a second blocking step with 2% Goat serum (Vector Laboratories, S-1000-20). The ECM panels were then incubated with 1:300 Rabit anti-COL1A1 (Cell Signaling Technology, 81375SF) and either 1:50 Mouse anti-Laminin (Sigma Aldrich, L8271) or 1:100 Mouse anti-Fibronectin (Sigma Aldrich, F7387) primary cocktail overnight at 4C. The decidual marker panel was incubated with 1:50 Mouse anti-PR A/B (LS-C133381) and 1:50 Rabbit anti-FOXO1 (Cell Signaling Technology, 2880S) in the same conditions. All samples were then washed and incubated with 1:1000 dilution of Goat anti-Mouse Alexa Fluor 594 (Thermo Scientific, A-11006) and Goat anti-Rabbit Alexa Fluor 647 (Thermo Scientific, A-21235) for 1 h at room temperature. Samples were then incubated with 500 ng/mL DAPI (BioLegend, 422801) for 15 min, then imaged with Leica Mica microscope within 2 h of staining. All quantification was performed using CellProfiler v4.2.6. 95 Raw single channel images underwent illumination correction prior to object identification. CD10 channels were used to identify areas of interest using Global Otsu Thresholding, then confirmed manually. The MaskImage module was then used to quantify intensity of markers within CD10 positive areas using MeasureImageIntensity and MeasureObjectIntensity. Untransformed FOXO1 and PR values are used. For ECM panels, MeasureImageAreaOccupied was used to quantify area of CD10 masks. CalculateMath was then used to divide total marker intensity by CD10 area. In excel, these values were converted from MFI/px 2 to MFI/mm 2 using microscope-derived calibration of 100 μm per 175 pixels. ECM proportions were calculated from total intensity per mm 2 for each focus area. Collagen values from matched focus areas between ECM panels A and B were used as a quality control metric to confirm similar expression levels. If collagen values from A and B differed by greater than >3 MFI/mm 2 , focus areas were excluded from analysis. Final focus areas used within analysis are listed in Table S5 . Statistics were performed using GraphPad Prism 10 (GraphPad, San Diego, CA). For all data, outliers were identified and removed using ROUT test with a conservative Q set to 0.2%. Normality of results was determined using Shapiro-Wilk tests, with p -values <0.05 resulting in nonparametric testing. All single-cell data were nonparametric, so all enrichment scores were analyzed with Kruskal-Wallis tests with Dunn’s post-hoc, with significance defined by p -values less than 0.001. For treatment and timepoint validation analysis, one-way ANOVA with Tukey post-hoc tests. For sequestration, stiffness, and ECM analysis, two-way ANOVA with Sidak correction for multiple comparisons was used and plotted as mean and standard error if data passed normality tests. For nonparametric data values are plotted as median with 95% confidence interval, and aligned rank transform ANOVA with Tukey post hoc was performed. Percent positive values across repeat experiments are normalized according to E2 groups and plotted together, with means and standard error or medians with 95% confidence intervals for parametric and nonparametric data, respectively. For WNT4 and PLZF, MFI rather than percent positive values are reported due to constitutive initial expression of both markers ( Figures S2 F–S2H). ELISA absorbance values were interpolated using GraphPad Prism 10, then plotted as absolute prolactin concentration values in pg/mL. Image analysis used either one-way ANOVAs with Tukey post-hoc tests or Kruskal-Wallis with Dunn’s post-hoc tests.

This work is supported by the Wallace H. Coulter Center for Translational Research at the University of Virginia and the University of Virginia School of Engineering . The flow cytometry data for this manuscript were generated in the University of Virginia Flow Cytometry Core Facility (RRID: SCR_017829 ) and were partially supported by the NCI grant no. ( P30-CA044579 ). The authors acknowledge the University of Virginia Flow Cytometry Core for their assistance in analysis and in panel design.