Collagen I triggers directional migration, invasion and matrix remodeling of stroma cells in a 3D spheroid model of endometriosis

Endometriosis is a painful gynecological condition characterized by ectopic growth of endometrial cells. Little is known about its pathogenesis, which is partially due to a lack of suitable experimental models. Here, we use endometrial stromal (St-T1b), primary endometriotic stromal, epithelial endometriotic (12Z) and co-culture (1:1 St-T1b:12Z) spheroids to mimic the architecture of endometrium, and either collagen I or Matrigel to model ectopic locations. Stromal spheroids, but not single cells, assumed coordinated directional migration followed by matrix remodeling of collagen I on day 5 or 7, resembling ectopic lesions. While generally a higher area fold increase of spheroids occurred on collagen I compared to Matrigel, directional migration was not observed in co-culture or in 12Z cells. The fold increase in area on collagen I was significantly reduced by MMP inhibition in stromal but not 12Z cells. Inhibiting ROCK signalling responsible for actomyosin contraction increased the fold increase of area and metabolic activity compared to untreated controls on Matrigel. The number of protrusions emanating from 12Z spheroids on Matrigel was decreased by microRNA miR-200b and increased by miR-145. This study demonstrates that spheroid assay is a promising pre-clinical tool that can be used to evaluate small molecule drugs and microRNA-based therapeutics for endometriosis.

capture the heterogeneity of endometrial cells found in lesions, the cells we employed in this study were an immortalized eutopic stromal cell line St-T1b 28 , primary endometriotic stromal cells (ESCs) and the ectopic light red peritoneal lesion derived epithelial 12Z cell line that was previously shown to be invasive in a Matrigel invasion assay 29 .
First, we validated that the cells retained their stromal and epithelial morphology in culture. Figure 2A shows that while the St-T1b and ESCs cells have an elongated, fibroblast-like stromal morphology, 12Z cells have a mostly polygonal shape and grow in clusters. Furthermore, on tissue culture (TC) plastic, the stromal cells exhibit more defined actin fibers compared to the 12Z cells. Quantitative analysis (Fig. 2B) confirmed that 12Z cells are significantly smaller (p < 0.0001) than St-T1b and ESCs, where the average area for St-T1b, 12Z and ESCs cells on TC plastic were 2086 ± 904.1 µm 2 (n = 29), 787.7 ± 380.9 µm 2 (n = 32) and 1989 ± 889.5 µm 2 (n = 30).
Recent studies suggested that spheroid culture offers several advantages over 2D culture and confirmed that 12Z cells 26 and endometriotic stromal cells 27 can assemble into spheroids using the U-bottom 96 well plates 27 . However, it has not been investigated whether also the hanging drop method can be used to fabricate endometrial spheroids and whether there are any differences between spheroids fabricated from epithelial and stromal endometrial cells alone or their co-culture. We, therefore, evaluated the hanging-drop method, each drop containing 20,000 of stromal or epithelial cells or their co-culture in 20 µL of standard media and selected day 4 as the harvesting day.
Bright-field images (Fig. 2C) show that all the studied cell types self-organized into spheroids. Interestingly, the morphology of the spheroids varied across cell types. St-T1b and ESCs cells assembled into compact, roundspheroids, while the 12Z spheroids were larger and sometimes exhibited slightly branching morphology. We also generated co-culture spheroids from the epithelial 12Z and stromal St-T1b cell lines combined at 1:1 ratio (Fig. 2D). Cell Tracker staining and confocal imaging suggest the two cell populations were homogeneously distributed throughout the spheroid on day 4. Interestingly, while the size of individual 12Z cells in 2D is significantly smaller compared to the ESCs and St-T1b cells, 12Z spheroids were significantly larger compared to St-T1b and ESCs (n = 14, p < 0.0001 and p < 0.001) (Fig. 2E). To exclude that this is due to a cell-counting error, the spheroid size was measured on spheroids prepared three independent times. The co-culture spheroids were also significantly larger compared to St-T1b spheroids (n = 11) and had a higher metabolic activity that is indicative of higher cell count and proliferation over the spheroid formation period (Fig. 2F,G).
Next, we evaluated whether the condensation into spheroids induces changes in gene expression. We analysed a subset of genes related to ectopic tissue invasion. Gene expression analysis revealed that while organisation into www.nature.com/scientificreports/ spheroids alters the gene expression of several markers in St-T1b cells, none of these markers was significantly altered in 12Z spheroids compared to monoculture across three independent preparations (Fig. 2H,I).  www.nature.com/scientificreports/ First, we examined the expression of Ras-related C3 botulinum toxin substrate 1(RAC1/Rac1), a small signalling G protein that directs actin-driven cellular protrusion, microtubule prolongation and the formation of lamellipodia 30 both in single cells and at the leading edge during collective migration 31 . The expression of RAC1 was significantly downregulated in 3D compared to 2D St-T1b (p < 0.01, n = 3) (Fig. 2H).
As the epithelial to mesenchymal transition (EMT) and mesenchymal to epithelial transition (MET) have been implicated in the progression of the disease, we further investigated the expression of mesenchymal markers vimentin (VIM) and cadherin-2 (CDH2) and the epithelial marker cadherin-1 (CDH1) (Fig. 2H). Vimentin expression remained unchanged in both cell lines (p > 0.05, n = 3). The expression of CDH2, a cadherin known to promote invasion in many cell types 32 , was downregulated in St-T1b spheroids (p < 0.01, n = 3) while the expression of CDH1 was upregulated in St-T1b spheroids (< 0.05 = n = 3) compared to the 2D control.
Matrigel and collagen I trigger distinct phenotypes in single cells and spheroids where stromal condensates create defects on collagen I. Having confirmed that endometrial stromal and epithelial endometriotic cell line as well as their co-culture were able to form spheroids, we evaluated their invasive behaviour on two different ECM-derived hydrogels: Matrigel and collagen I using confocal imaging.
Single cells of all studied cells on Matrigel formed cellular aggregates by day 3 (Fig. 3A). While these aggregates remained mostly rounded in St-T1b and ESCs groups, the 12Z cell line aggregates consistently developed multiple multicellular protrusions across several preparations. Cells seeded on collagen I were invading collagen I as single cells (Fig. 3A).
We next evaluated the spheroid behaviour on Matrigel and collagen I. On the basement membrane (BM) mimic Matrigel, the stromal St-T1b spheroids remained rounded with ESCs exhibiting few protrusions and only the 12Z spheroids consistently developed multiple multicellular protrusions across several preparations. Confocal imaging on day 7 (Fig. 3B) revealed that the 12Z protrusive edges consisted of tightly packed cells (DNA in blue) with scant cytoplasm (actin staining in red).
The response of all studied cell types to collagen I as spheroids was markedly different compared to single cells (Fig. 3B). St-T1b and ESC spheroids on collagen I developed into invasive lesion-like structures (Fig. 3B). More specifically, the St-T1b and ESC spheroids gradually invaded collagen I, leaving behind a circularly remodeled matrix with a ring of tightly adhering cells at its margins (Fig. 3B,C). These rings appeared to stabilize the defect and to limit further random cellular spreading outside of the defect in many but not all spheroids. www.nature.com/scientificreports/ Interestingly, no matrix defect or directional spreading was detected in co-culture St-T1b:12Z spheroids on collagen I (Fig. 3D). Co-culture spheroids on Matrigel developed protrusive edges similar to the 12Z-only spheroids.

Directional invasion followed by the formation of a circular defect occurs in St-T1b and ESCs spheroids but not in St-T1b
:12Z co-culture. Next, we quantified the invasive and migratory patterns on Matrigel and collagen I using bright-field imaging and a parameter that we termed 'Fold change in the area' that we defined as the overall projected area, including matrix defects on the day of interest divided by the area on the day 0 or 1 without any sprouts (Fig. 4A). All analysis was done manually in FIJI using the freehand selection tool. While manual measurement has its limitation, especially when the ' Area' increases and its margins become irregular, no significant difference in measured areas was observed between different assessors (Fig. 4B).
Our data show that the 'Fold change in area' is significantly higher on collagen I compared to Matrigel across all studied cell types by day 5 (Fig. 4C,D). Confocal imaging combined with brightfield microscopy suggested the stromal spheroids invade (Fig. 3B,C) and migrate on the collagen I matrix directionally (Fig. 4C). To quantify this, we used the parameter 'Directionality' that is calculated as the ratio of the distance of the centre of the spheroid core from the centre of the overall migrated area b to the semi-major axis of the overall migrated area a (Fig. 4E). The normalized directionality increased for St-T1b but not for 12Z or co-culture spheroids with time on collagen I, especially between days 3 and 5 (Fig. 4F). The directional invasion was typically followed by matrix remodeling resulting in a circular defect at the area with the densest stromal cell population (Fig. 4G). In our system (3 mg/ mL, 40 µL/well) this typically occurred around day 5 or 7 with 84.6% and 53.3% of St-T1b and ESCs, respectively, having a defect on day 7 (n = 13-15 per time point) (Fig. 4G). The defects formed both on 1 mg/mL and 3 mg/ mL collagen I hydrogels, suggesting this behavior occurs across a range of collagen I concentrations (Fig. 4H).
Spheroid 3D culture as an effective tool to screen small molecule drug and microRNA-based therapeutics. We then evaluated the potential of the here presented endometrial spheroid in vitro assay to screen the potential therapeutic effect of mechanoregulatory small molecules and micro RNAs (Supplementary Table ST1).

The broad-spectrum MMP inhibitor NNGH limits the invasive behaviour of stromal spheroids on collagen I.
Previous studies implicated that MMP signalling plays a role in the formation of early endometriotic lesions 15 . Our study shows that the broad-spectrum MMP inhibitor 15 µM N-isobutyl-N-(4-methoxyphenylsulfonyl) glycyl hydroxamic acid (NNGH) significantly reduced 'Fold change in the area' on collagen I from 10.4 fold to 2.3 fold (n = 6-9) and 9.2 fold to 3.3 fold (n = 6-9) in St-T1b and ESCs, respectively, but did not significantly affect the 'Fold change in the area' in 12Z cells (n = 6-9) (Fig. 5A, Supplementary Figure S1). Furthermore, it can be seen from Fig. 5B, that while NNGH treatment prevents the formation of the circular defect on collagen I even after 7 days in culture, the migration of St-T1b and ESCs is not completely eliminated. The effect of NNGH inhibitor on the St-T1b:12Z co-culture was less pronounced and neither the control nor NNGH group formed matrix defects by day 5 (Fig. 5C).

ROCK inhibition significantly enhances spreading and invasion of all studied endometrial cell types on Matrigel.
The ROCK inhibitor Y27632 significantly (p < 0.01) increased the 'Fold change in the area' of all studied cell types on Matrigel compared to DMSO (Fig. 5D). The area occupied by St-T1b, 12Z and ESCs was 17.3, 6.6 and 22.3 fold larger compared to day 0 ( Fig. 5E). Y27632 also affected the numbers of metabolically active cells, which were significantly higher compared to controls for St-T1b and 12Z cells on day 5 on Matrigel. Moreover, Y27632 affected spheroid morphology (Fig. 5F). Y27632 on Collagen I resulted in a disaggregation of the spheroid core in St-T1b and ESCs as shown in the Supplementary Figure S2. Treatment with Y27632, in contrast to the MMP inhibitor NNGH, did not prevent Collagen I matrix remodeling in ESCs (Fig. 5G) suggesting the directional remodeling is rather due to proteolytic action than acto-myosin contraction.

The spheroid model reveals context-dependent roles of the mechanoregulatory microRNAs miR-200b and miR-145 on the invasive behaviour of the endometriotic epithelial cell line 12Z on Matrigel.
We next investigated whether our in vitro model can be used as a tool to screen the functional effect of various microRNAs on endometrial phenotype. In particular, we selected two microRNAs, miR-200b 33 and miR-145 34 , that have been previously shown to be dysregulated in endometriosis 35 and to modulate the invasion and migration of 12Z cells in 2D and Transwell assays. miR-200b acts as a transcriptional repressor of ZEB1/2 and thus downregulates EMT transition 36 . The miR-145 is upregulated in endometrial lesions and has been described to modulate cytoskeletal dynamics in several cell types, including endometrial, and has many validated targets, including beta and gamma actin, cofilin, fascin, myosin light chain 9 and Rho kinase Rock1 34,37 . The transfection was performed in monolayer culture before the fabrication of spheroids and the effect of micro-RNAs on spheroid spreading was assessed after 3 days on Matrigel to minimize the effect of miR dilution and degradation 38 (Fig. 6A). It can be seen from Fig. 6B that microRNA transfection did not significantly alter the ability of cells to form spheroids and the area of individual spheroids was not significantly different (p > 0.05) across the treatment groups nor was the proliferation (Fig. 6C). We observed spheroid fragmentation of miR-200b transfected cells on Collagen I which resulted in a discontinuous nature of the projected area the size of which could not be reliably quantified (Fig. 6D). MiR-145 significantly reduced the spheroid area compared to scr. miR controls on day 3 on collagen I (Supplementary Figure S3). On Matrigel, the microRNAs, affected sprouting characteristics behaviour of 12Z cells as seen in the bright-field images in Fig. 6E  www.nature.com/scientificreports/ increased the number of sprouts per spheroid to ~ 34 (Fig. 6F,G) and increased the overall sprouting area from 56.12 × 10 3 ± 21.87 × 10 3 µm 2 per scrambled control miR spheroid to 130.86 × 10 3 ± 43.47 × 10 3 µm 2 per miR-145 treated spheroids (p < 0.0001) (Fig. 6F,H). In line with previous findings on EMT-marker analysis in 2D-cultured 12Z cells 31,33 , qPCR analysis of miR-200b-treated 12Z spheroids indicated strong upregulation of CDH1 expression levels, however, the data were not significant due to high variability, since only minute amounts of RNA could be isolated from the spheroids (Supplementary Figure S4).

Discussion
Endometriosis is a complex multifactorial disease 1 . The overall goal of this study was, therefore, to develop a modular 3D in vitro model that makes it possible to study the interplay of different factors that have been proposed to contribute to the pathogenesis of endometriosis and screen potential therapeutics in vitro. www.nature.com/scientificreports/ First, we demonstrate that the hanging drop method makes it possible to generate endometrial spheroids of reproducible size and thus provides a good alternative to the low-adhesion plate method 26,39 . Our data show that the spheroid size is consistently cell-type specific, with stromal cells generating smaller spheroids than the epithelial 12Z cells or their co-culture. This is likely due to proliferation of 12Z cell in spheroids as suggested by the cell proliferation assay on spheroids on day 4. qPCR analysis revealed that the spheroid culture affects gene expression. The stromal St-T1b had enhanced expression of the MMP2, MMP14 compared to 2D culture. RAC1, on the other hand, was downregulated in St-T1b spheroids. Spheroids in which Rac1 production was either inhibited or the gene was constitutively expressed had suppressed or enhanced migration in 3D matrices, respectively 40 . We speculate that it is possible that RAC1 is temporarily downregulated in stromal cells cultured as a suspension spheroid culture. While basal CDH1 expression was very low with a Ct value of 27, as expected for a mesenchymal cell line, we observed a significantly increased expression in 3D culture. We could previously (n = 9-10 independent wells across two independent preparations, ANOVA, Tukey's multiple comparisons). (H) The overall area occupied by sprouts was significantly larger and smaller when treated with miR-145 and miR-200b, respectively, compared to scr.miR after 3 days on Matrigel (n = 8-10 independent wells across two different preparations, ANOVA, Tukey's multiple comparisons, two independent experiments), *p < 0.05; **p < 0.01; ***p < 0.001 and n.s. p > 0.05; data expressed as mean ± s.d.  26 , this difference could be due to different spheroid size and culture time. Endometriosis is marked by the growth of endometrium at ectopic locations 1 . We, therefore, investigated how epithelial 12Z, ESCs and St-T1b and co-culture spheroids; and single cells interact with two ectopic ECM mimics Matrigel resembling the basement membrane and collagen I mimicking the exposed stroma. The 'fold increase in the area' of the spheroids was markedly higher on collagen I than on Matrigel on day 5. Similarly, single cells seeded on top of these hydrogels preferentially invaded collagen I hydrogels. Our results are in agreement with previous studies conducted on cancer cells suggesting that collagen I alone can increase the invasive cellular phenotype and show that this effect is significant across cell types [42][43][44] . These data also tie well with the previously reported clinical observations that tissue scarring either due to surgery or persistent microtrauma could contribute to the pathogenesis of endometriosis [7][8][9] .

Scientific Reports
While the 12Z cell line was created from lesion-derived cells 29 based on their ability to penetrate through Matrigel coated invasion chambers, the 12Z single cells in our study only assembled into cellular aggregates with processes and 12Z spheroids developed invasive edges. Previously, Pollock and colleagues also observed only low levels of basal invasion in 12Z cells on Matrigel hydrogels 45 . We speculate that the limited invasive capacity of 12Z cell observed in this study could be due to the chemotactic gradient that is a key part of the invasion chamber setup. Our group has indeed previously demonstrated that 12Z are invasive under a fetal calf serum gradient 33 .
Unexpectedly, there was a marked difference between the behaviour of stromal single-cell suspension and spheroids on collagen I. The St-T1b and ESC spheroids but not single cells consistently migrated on, invaded and remodelled collagen I in a directional manner leaving behind a circular defect in the material encircled by the cells that visually resembled peritoneal endometriotic lesions. Given that this was the case for both the St-T1b cell line derived from healthy cells and ectopic ESCs suggests such invasive behavior might be an inherent property of stromal endometrial menstrual condensates and could be critical not only for the pathophysiology of endometriosis but also for normal regeneration of endometrium.
Directional migration followed by matrix remodeling was not observed in the 12Z-spheroid or the 12Z: St-T1b co-culture groups, suggesting the stromal-epithelial interactions modulate stromal invasiveness. While we did not investigate the MMP levels of the co-culture spheroids, previous research determined that the co-culture between endometrial Ishikawa epithelial and telomerase-immortalized stromal cells reduces the MMP2 levels in stromal cells both in the absence of hormonal stimulation and in the presence of 10 nM estradiol concentration 46 .
In this paper, we further demonstrate that the endometrial spheroid-ECM platform can be used for drug screening of small molecule drugs and micro-RNAs (Fig. 7). We show that the collagen I circular defect caused by stromal cells arises due to matrix degradation via MMPs rather than due to cellular contraction. Both eutopic and ectopic stromal cells had significantly upregulated MMP expression and the MMP inhibitor, NNGH, significantly reduced the size of in vitro stromal lesions on collagen I. These results are in good agreement with Nap and colleagues that demonstrated that inhibiting MMP activity prevents the development of endometriotic lesions in a model combining chicken chorioallantoic membrane model and biopsies of menstrual stage endometrium obtained from healthy donors 15 . Our results refine this model and show that while, in agreement with the previous studies 47 , the MMP inhibitor significantly slows down the invasion of spreading of stromal cells on collagen it has little effect on the collective migration of 12Z cells. Another signaling molecule we targeted is ROCK, which is a key regulator of the cytoskeleton 30,48 . On Collagen I, ROCK inhibitor Y27632 treatment led to a rapid loss of the spheroid core structure compared to controls and Y27632 did not prevent Collagen I remodeling by ESCs suggesting the matrix remodeling is not primarily driven by matrix contraction but rather by MMP proteolytic action. Y27632 further significantly increased the 'fold change in area' and cell numbers in vitro on Matrigel in all studied cell types. Similar increase in cellular spreading following the treatment with ROCK inhibitors have been described in microvascular endothelial cells 49 , retinal pigment epithelial cells 50 and osteoblastic cells 51 . It www.nature.com/scientificreports/ needs to be noted that Y27632 has a complex effect on phenotype 52,53 . For example, prior studies demonstrated that Y27632 reduces endometriosis associated fibrosis in vitro 54 . Another promising class of therapeutics targets microRNA signaling 35,55 . Given that a typical micro-RNA has tens of targets, sequencing studies need to be accompanied by reliable functional assays to be biologically meaningful 56 . In this study, we demonstrate that the spheroid assay can be used to reproducibly evaluate the effect of individual microRNAs on the complex, multicellular spreading of endometriosis-mimicking constructs over several days. We show that miR-200b treatment of 12Z cells resulted in a reduction of sprout formation, which may be indicative of a less invasive phenotype. Our previous 2D data suggest that miR-200b may have reverted the 12Z phenotype to an epithelial-state 33 , however, the paucity of RNA in the spheroids did not allow us to unequivocally confirm this hypothesis, as we saw only a non-significant increase in expression of the epithelial marker E-cadherin (Supplementary Figure S4). Our spheroid model further revealed that while miR-145 reduces the migrated area on Collagen I compared to controls (Supplementary Figure S3), results which are in agreement with previous in vitro 2D assays 34 , the microRNA miR-145 up-regulated in ectopic lesions in vivo increases 12Z sprouting on Matrigel in vitro. These findings were unexpected and investigating this into more detail is beyond the main focus of this study. Nevertheless, there is an increasing appreciation that cells adopt a host of invasive and migratory strategies that are highly context-dependent and enabled by distinct signaling pathways. Liu and colleagues observed that miR-145 upregulation enhances angiogenesis, including the sprouting from aortic rings and linked this to the suppression of tropomodulin 3 (TMOD3) 57 while we observed that miR-145 inhibits proliferation and migration in breast cancer and endometriotic cells using the Transwell migration and scratch assays 32,58 . Therefore, miR-145 might influence cellular invasive behaviour not only in cell-type but also invasive/migratory-mode manner and ECM-substrate-dependent manner. While the majority of oncological studies on miR-145 function suggest that it reduces invasive growth by targeting a variety of mRNAs, two studies in trophoblast cells have described invasion-promoting functions of miR-145, which were attributed to a targeting of mucin 1 (MUC1) and leukemia inhibitory factor receptor (LIFR), respectively 59,60 . We can only speculate that the 3D spheroid culture compared to 2D culture of 12Z cells may have altered the expression patterns of miR-145 target mRNAs in a way that alters the response to this epigenetic regulator. For example, miR-145 may target new mRNAs that are not expressed in the 2D setting (or vice versa), resulting in a different net response. Overall, we demonstrate that the spheroid assay can be used as an additional assay to screen for both small molecule and RNA-based therapeutics.
A major limitation of our study is that it relies on cell lines that have been transformed and represent only a limited subset of disease phenotypes and a more extensive primary cell pool will be required to confirm and fully elucidate the here reported findings. We also did not investigate the influence of decidualization. Notably, our study did not incorporate primary endometrial epithelial cells with purely epithelial characteristics. Additionally, the wider implementation of this assay for the study of endometriosis will rely on future advances in the molecular characterization of spheroids and high-throughput image analysis. Furthermore, automated image analysis would significantly increase the throughput of this assay. In recent years, the quality of image processing algorithms has approached that of trained humans while significantly decreasing the time needed to evaluate individual samples 61 . It needs to be noted that for such algorithms either large training datasets or pre-defined criteria are needed. Given the wide array of spheroid phenotypical responses, we have only started to identify such criteria.
Overall, our screening platform provides evidence that the physiological condensation of endometrial stromal cells into spheroids might play an important role in the development of a subset of endometriotic lesions. As such a directional invasive phenotype in vitro is unlikely to arise by chance, endometrial stromal condensation might also have currently unknown but likely important biological role in the cyclical regeneration of normal endometrium. At the same time, our results show that the epithelial lesion-derived 12Z spheroids also rapidly migrate on collagen I and stromal-epithelial interactions modulate the invasiveness of stromal cells. Previous studies indeed revealed significant heterogeneity and variability among different endometriosis subtypes with several sub-types staining predominantly for stromal markers 62 .
In conclusion, this study documents that endometrial stromal cell line St-T1b and primary endometriotic stromal cells engage in directional migration with significant collagen I remodeling when cultured in spheroid culture and that this behaviour is inhibited by the broad-spectrum MMP inhibitor NNGH. We anticipate that this assay will be used to gain further insights into invasive processes involved in endometriosis and for the screening of both small molecule and RNA-based drug candidates and their off-target effects.

Methods
Cell culture. The 12Z ectopic epithelial cell line 17,29 was maintained in DMEM media (Sigma-Aldrich, cat. No. D0819, Deisenhofen, Germany,) supplemented with 10% FBS (Biochrom GmbH, cat. no. S0615, Berlin, Germany) and 1% Pen/Strep (Sigma-Aldrich, cat. No. P4333). The St-T1b cell line 28 and primary ectopic lesionderived stromal cells (ESCs) were maintained in 70% DMEM/18% MCDB 105 media (Sigma-Aldrich, cat. No. 117-500) supplemented with 10% FBS, 1% Pen/Strep, 1% Glutamine and 5 µg/mL insulin (Sigma-Aldrich, cat. No. 10516). Cells were routinely split twice a week. ESCs were prepared from ectopic lesions and characterized as previously described 63 . Primary endometriotic stromal cells were prepared from a biopsy of a woman with endometriosis who underwent surgical treatment at the Department of Gynecology and Obstetrics of Münster University Hospital in 2013, and stored as aliquoted stocks in liquid nitrogen, which were freshly thawed and passaged in routine culture two times prior to usage in the experiments described. The modified American Society for Reproductive Medicine classification was used to assess endometriosis 64 . For all ESC experiments, stroma cells derived from a lesion located at the pelvic wall (rASRM score II) of a 19-year-old patient were employed. The study was carried out following the Declaration of Helsinki and approved by the local ethics commission Image analysis. All images were analysed in FIJI 66 . Confocal images are depicted as maximal intensity projections. The spheroid area was measured manually by tracing the spheroids using the freehand tool and measure function on Bright-field images of spheroids on Petri Dishes, glass slides or in a 96-well plate. Fold increase in area was calculated as the spheroid area on a given day divided by spheroid size on day 0 or day 1. If on day 1 any protrusions were present and the spheroid was used as a reference size for the given experiment, the protrusions on day 1 were excluded from the analysis to better reflect the size of the original spheroid core. The parameter directionality was calculated as the ratio between the distance in pixels between the centre of the overall migrated area and the centre of the spheroid, divided by the semi-major axis of the overall migrated area of the spheroid (Fig. 4E). The number of sprouts per image was counted manually and the sprouting area was calculated as the total area occupied by an expanding spheroid with sprouts minus the area occupied by the spheroid without any protrusions (Fig. 6F).  Tables 2 and 3.

RNA extraction and
Statistical analysis. Data were analysed using GraphPad Prism8 (GraphPad Software, San Diego, USA).
Normal distribution was tested using the Shapiro-Wilk test. A two-tailed unpaired Student's t tests were used to analyse statistical significance between two conditions in an experiment. For experiments with three or more comparisons, an ordinary one-way ANOVA with a Tukey's multiple comparisons test was used. For data that were not normally distributed, the Kruskal-Wallis test followed by Dunn's multiple comparisons test was used. A two-way repeated-measures (RM) ANOVA with Šidák's multiple comparisons test was used to evaluate the effect of Matrigel and collagen I on spheroid size over time. Significance values were chosen as *p < 0.05; **p < 0.01; ***p < 0.001, ****p < 0.0001. Error bars represent the mean ± s.d or mean + s.d. All figure panels were assembled in Inkscape 0.92.  ACTB  TCA AGA TCA TTG CTC CTC CTGAG  ACA TCT GCT GGA AGG TGG ACA   RAC1  CGC CTC CTG TAG TCG CTT TG  CAC GCT GTA TTC TCG CCA GTG   MMP14  CCA TTG GGC ATC CAG AAG AGAGC GGA TAC CCA ATG CCC ATT GGCCA   MMP2  GCC GTG TTT GCC ATC TGT TT  CTG CAG GGA GCA GAG ATT CG   VIM  TCA GCA TCA CGA TGA CCT TGAA  CTG CAG AAA GGC ACT TGA AAGC   CDH2  TTC TGA CAA CAG CTT TGC CTCTG  TTT ATT CAG AAC GCT GGG GTCA   CDH1 CAA AGC CCA GAA TCC CCA AG CAC ACC TGG AAT TGG GCA AA