A Fibronectin (FN)-Silk 3D Cell Culture Model as a Screening Tool for Repurposed Antifibrotic Drug Candidates for Endometriosis

This study advances sustainable pharmaceutical research for endometriosis by developing in vitro 3D cell culture models of endometriotic pathophysiology that allow antifibrotic drug candidates to be tested. Fibrosis is a key aspect of endometriosis, yet current cell models to study it remain limited. This work aims to bridge the translational gap between in vitro fibrosis research and preclinical testing of non-hormonal drug candidates. When grown in a 3D matrix of sustainably produced silk protein functionalized with a fibronectin-derived cell adhesion motif (FN-silk), endometrial stromal and epithelial cells respond to transforming growth factor beta-1 (TGF-β1) in a physiological manner as probed at the messenger RNA (mRNA) level. For stromal cells, this response to TGF-β1 is not observed in spheroids, while epithelial cell spheroids behave similarly to epithelial cell FN-silk networks. Pirfenidone, an antifibrotic drug approved for the treatment of idiopathic pulmonary fibrosis, reverses TGF-β1-induced upregulation of mRNA transcripts involved in fibroblast-to-myofibroblast transdifferentiation of endometrial stromal cells in FN-silk networks, supporting pirfenidone's potential as a repurposed non-hormonal endometriosis therapy. Overall, endometrial stromal cells cultured in FN-silk networks—which are composed of a sustainably produced, fully defined FN-silk protein—recapitulate fibrotic cellular behavior with high fidelity and enable antifibrotic drug testing. 1 Introduction Gender equality is a fundamental right,[1] and true empowerment of women can only be achieved when advancements in women's health are prioritized, adequately funded, and made visible.[2-6] Despite affecting 10% of women and people assigned female at birth, treatments for endometriosis remain underfunded and under-researched.[4, 7, 8] This work addresses the urgent need for pharmacological therapies for endometriosis, aligning with Sustainable Development Goal (SDG) #3 (Good Health and Well-Being) and SDG #5 (Gender Equality),[9] illustrating how these SDGs can be integrated into preclinical research.[2] We emphasize the importance of sustainable research practices, including the use of non-animal-derived materials, and hereby present novel three-dimensional (3D) in vitro models of endometriotic stroma and epithelium that model fibrosis using a recombinant silk protein functionalized with the arginine-glycine-aspartate (RGD) cell adhesion motif from fibronectin (FN-silk). By adopting this sustainable material, we embrace our responsibility as scientists to address SDG #13 (Responsible Consumption and Production),[9] particularly regarding the materials used in our laboratories. FN-silk networks enable in vitro modeling of fibrosis in endometriosis and testing of an antifibrotic repurposed drug candidate for endometriosis, aiming to reduce reliance on animal experiments and improve the clinical translatability of our research. There is ample histopathological evidence for dense, fibrous tissue in and around endometriotic lesions,[8, 10-17] yet the role of fibrosis in the pathophysiology of endometriosis is only just beginning to be emphasized by the research community,[10-14] with Viganò et al. even suggesting the definition of endometriosis be changed to “a fibrotic condition in which endometrial stroma and epithelium can be identified.”[11] During repeated injury caused by the shedding of ectopic endometrium in response to a withdrawal of progesterone – a bleeding internal wound[18] – the body's healing response becomes pathological, causing extracellular matrix (ECM) proteins to be deposited recurrently in an attempt to heal tissue damaged by the decidualized endometriotic lesion.[13, 19-22] This process of repeated tissue injury and repair leads to scarring, fibrosis, and, in severe cases, adhesions.[11-13, 23] Encouragingly, recent preclinical studies in non-human primate and murine models have used the presence and extent of fibrotic tissue as an indicator of disease progression against which to evaluate the efficacy of drug candidates.[24, 25] When developing anti-fibrotic pharmacological treatments for endometriosis, the biggest challenges are how to reverse or prevent fibrosis, and how to reach fibrotic endometriotic lesions in the pelvis and on the peritoneum.[11, 26] Cellular mechanisms involved in the fibrogenesis of endometriosis are fibroblast-to-myofibroblast transdifferentiation and epithelial-to-mesenchymal transition[14, 18, 19, 27-31] (Figure 1A). These cellular mechanisms are common to fibrotic pathologies in multiple organs. However, in endometriosis, these processes are intricately linked to repeated tissue injury and repair cycles in response to cyclical changes in steroid hormone levels.[13] On the one hand, fibrosis in endometriosis is driven by stromal fibroblasts, which transdifferentiate into myofibroblasts upon exposure to cytokines such as transforming growth factor beta 1 (TGF-β1),[12, 16, 32] neuropeptides,[33-35] or mechanical cues.[36-39] In fibrosis, secreted ECM proteins accumulate over time, forming fibrotic lesions, which disrupt the tissue architecture and are at least in part responsible for typical symptoms of endometriosis such as infertility, chronic pelvic pain, and dyspareunia.[8, 11, 13, 15] On the other hand, epithelial-to-mesenchymal transition enhances cell migration and adhesion to the peritoneum, thereby increasing the invasive capacity of endometriotic epithelial cells.[30, 40-43] Unlike other fibrotic diseases, fibrosis in endometriosis correlates with higher pain scores.[10, 33, 44] More recently, 3D models in which cells grow in a matrix have raised interest to evaluate fibrotic mechanisms, as cells are able to interact with the matrix and other cells from all sides, deposit ECM proteins, and create their own biomechanical microenvironment, mimicking endometriotic foci found in humans more closely.[10, 27, 38, 42, 45, 51-56] To model endometriosis and associated fibrosis, we used a recombinantly produced spider silk protein derived from the four poly-Ala/Gly-rich repeats and the non-repetitive C-terminal domain of Major Ampullate Spidroin 1, 4RepCT, genetically functionalized to include the arginine-glycine-aspartate (RGD)β cell adhesion motif from fibronectin (FN), known as FN-silk (Figure 1C).[46-49, 57-60] FN-silk possesses the unique capability to self-assemble into biodegradable and biocompatible microfibers in aqueous, physiological-like buffers at room temperature.[46, 50] FN-silk has the additional advantage of having a fully defined composition and being produced using biotechnological methods. Compared to animal-origin materials, such as collagens, FN-silk has a reduced environmental impact and, due to recombinant production, low batch-to-batch variability. The addition of a cell adhesion motif from ECM proteins facilitates cellular attachment to and proliferation on the FN-silk protein compared to non-functionalized silk proteins.[48] Harnessing the ability of the FN-silk protein to self-assemble into fibers of β-sheet structures formed at the air–liquid interface,[50] Johansson, Widhe et al.[46] established a method to produce networks from FN-silk with integrated cells (Figure 1B). Here, we explore 3D cell culture to model fibrotic pathophysiology in endometriotic stroma and epithelium, comparing FN-silk network and spheroid formats. The immortalized human endometrial stromal cell (T-HESC) and epithelial (12Z) cell lines were chosen because they are non-cancerous, stem from patients with leiomyoma and endometriosis, respectively, are readily available, and have been used previously, allowing us to compare results to previously published work.[29, 52, 53, 56] 3D cell culture in a matrix may have advantages in modelling the fibrotic condition of endometriosis, as cells are exposed to biomechanical cues arising from cell–cell and cell–matrix contacts mimicking the ECM. We hypothesized that FN-silk networks could provide this ECM-mimicking material by giving cells fibers to adhere to, remodel, and migrate on, as well as to establish focal adhesion points.[46] To test this hypothesis, we included matrix-free spheroids, which have been employed previously as in vitro models of endometriosis,[29, 52, 53, 56] to be able to compare scaffolded (FN-silk network) and scaffold-free (spheroid) 3D formats. We demonstrate the use of the in vitro models developed herein to test pirfenidone, an antifibrotic drug indicated for pulmonary fibrosis that has the potential to be repurposed for endometriosis.[61, 62] 2 Results and Discussion 2.1 Formation and Characterization of 3D Systems: FN-Silk Networks and Spheroids FN-silk networks with integrated cells were produced using a method adapted from Collodet et al.[47] as illustrated in Figure 2A, where the FN-silk protein assembled into β-sheets during foaming. The bursting of the air bubbles within the foam in the first two days after production caused the formation of a continuous network of β-sheets of FN silk and the network's immersion in the culture medium. As described previously,[46, 47] cells were integrated into the FN-silk prior to formation to allow a high level of integration and a homogeneous distribution of cells throughout the FN-silk network. Spheroids were formed by spontaneous aggregation of cells in suspension in an ultralow attachment plate (Figure 2B). Both the T-HESC (endometrial stromal) and the 12Z (endometriotic epithelial) cell lines were characterized in terms of their metabolic activity and morphology when grown in monolayers, to inform us about design parameters and the context in which to evaluate results from the 3D systems (Figures S1 and S2, Supporting Information). Stromal T-HESCs were seeded at a density of 10 000 cells per unit and epithelial 12Z cells were seeded at a density of 20 000 cells per unit (where one unit was an FN-silk network or a spheroid). We observed that 96 ± 7% of stromal T-HESCs and 96 ± 4% of epithelial 12Z cells seeded were integrated into the FN-silk network (mean ± SD) (Figure 2C), as measured by counting the number of cells left in the supernatant after transferring the FN-silk network to a well with fresh medium (Figure 2A). To evaluate cell metabolic activity in FN-silk networks and spheroids, cells from three different passage numbers were seeded in each 3D format and the metabolic activity was measured using a CellTiter Glo 3D Viability Assay at 1, 4, and 7 days post-seeding (day 0 was the day on which systems were seeded). On day 1 post-seeding, both spheroids and FN-silk networks showed similar metabolic activity, confirming equal cell seeding (Figure 2D,E). By day 4, stromal T-HESC metabolic activity increased slightly or remained constant in FN-silk networks but dropped in spheroids (Figure 2D). By day 7, metabolic activity in FN-silk networks decreased slightly for stromal T-HESCs, while spheroids showed very low activity, suggesting cell death or dormancy (Figure 2D). For epithelial 12Z cells, we observed increased metabolic activity in FN-silk networks by day 7, with variability based on passage numbers, while spheroids showed a decline (Figure 2E). For both cell types, FN-silk networks maintained cellular metabolic activity and supported 3D proliferation, whereas spheroids exhibited a decline in metabolic activity. The results presented here are in agreement with previously published work by Stejskalová et al., who found that at day 4, endometrial stromal cell spheroids (St-T1b cell line) had lower metabolic activity than endometriotic epithelial cell spheroids (12Z cell line) produced using the hanging drop method, with both cell types seeded at 20 000 cells per spheroid.[56] These spheroids showed sprouting and migratory behavior, indicating that at least the cells at the outer shell of the spheroid remained alive and responsive to molecular cues.[56] Wendel et al.[63] produced 12Z spheroids seeded at 18 000 cells per spheroid using the same method described here (Figure 2B). At day 5 of Wendel et al.’s epithelial 12Z spheroid culture, a section cut through the center of the spheroid showed cellular expression of Antigen Kiel 67 (Ki67), a cell division marker, and no expression of cleaved Caspase 3 (cCAS3), an apoptotic marker, even at the spheroid core,[63] suggesting that 12Z cells proliferated in spheroids.[63] This contradicts the observations described here (Figure 2E), which may be due to differences in passage number. The CellTiter Glo 3D Viability Assay was compared to a manual count of cells, and the ATP levels measured were in good agreement with the number of cells counted (Figure S4, Supporting Information). To assess the distribution of cells throughout the FN-silk network and morphological changes during the culture period, FN-silk networks with green fluorescent silk were prepared. On days 1, 4, and 7, networks were fixed and the actin filaments of the cell cytoskeleton were stained with phalloidin 594, and nuclei with 4′,6-diamidino-2-phenylindole (DAPI). Figure 3A shows that stromal T-HESC FN-silk networks contracted between culture days 1 and 4, and subsequently maintained their architecture over a 7-day culture period. Cells were uniformly distributed throughout the FN-silk matrix, with slightly more cells at the outer edges of the network (Figure 3A). The projected area of stromal T-HESC FN-silk networks was calculated from bright field images and was observed to decrease from 5.9 ± 1.2 × 106 µm2 at day 1 to 2.3 ± 0.3 × 106 µm2 at day 4 and 1.5 ± 0.3 × 106 µm2 at day 7 (mean ± SD from n = 6 FN-silk networks, bright field images not shown). This reduction in projected area is attributable to the bursting of bubbles during the formation of the FN-silk network at culture days 1 and 2, after which all bubbles had burst. Subsequently, stromal T-HESCs remodeled the FN-silk network, as cells were oriented along the FN-silk fibers and contracted the FN-silk matrix. The thickness of the networks was in the range of 100−150 µm, estimated from z-stack images taken using confocal microscopy, which was thin enough to allow sufficient oxygenation of cells throughout the network.[64] Stromal T-HESCs tended to stretch along the silk fibers, similar to results reported previously on the SK-BR-3 and MDA-MB-231 breast cancer cell lines.[47] The establishment of strong cell–matrix contacts by stromal T-HESCs was further evidenced by experiments in which cells were detached from FN-silk networks (Figure S3, Supporting Information). It took relatively harsh conditions of 0.5% Trypsin-EDTA incubation for 30 min to detach cells that had been cultured in FN-silk networks for 7 days; shorter incubation periods, lower Trypsin concentrations, and more gentle cell dissociation reagents were unable to detach cells from the FN-silk matrix. This indicates that cells established strong contacts with the FN-silk, mimicking interactions with the ECM. Similar to stromal T-HESCs, epithelial 12Z FN-silk networks maintained their architecture over a 7-day culture period (Figure 3B). On day 1, epithelial 12Z cells were uniformly distributed on the fibers of the FN-silk network. During proliferation in the FN-silk network, epithelial 12Z cells appear to have grown at the outer edges of the silk networks, after filling the center. To cross-check results from the metabolic activity assay, which works by ATP measurement, we evaluated cell viability in 3D culture formats visually using a live/dead staining kit, shown in Figure 3C–F. A homogeneous distribution of live cells was observed throughout the FN-silk network, with few dead cells visible, and some empty regions corresponding to the locations of the bubbles produced during the manufacturing procedure (Figure 3C,E). Spheroids of stromal T-HESCs (Figure 3D) and epithelial 12Z cells (Figure 3F) showed mostly live and some dead cells, while spheroids showed more dead cells than FN-silk networks with each respective cell type. While qualitative, these images support the fact that cells in FN-silk networks had higher levels of metabolic activity, while spheroids had lower levels of metabolic activity, and suggests that this may be in part attributable to more dead cells in the spheroids. The FN-silk networks enabled higher levels of metabolic activity and lower levels of cell death compared to culture in a spheroid format. Deng et al. compared spheroids and silk-scaffolded 3D models of primary human urinary stem cells and observed that spheroids had higher levels of oxidative stress and more cell senescence, consistent with lower metabolic activity of spheroids.[65] To characterize the architecture of the cells in FN-silk networks more closely, samples were fixed on day 4 followed by immunofluorescent staining. Figure 4A shows the morphology of stromal cells in FN-silk networks, stained for Vimentin, a mesenchymal cell marker. Figure 4B shows 12Z cells stained for epithelial cell adhesion molecule (EpCAM) in an FN-silk network. Stromal and epithelial cells were uniformly distributed throughout the FN-silk matrix. Close-up images (Figure S6, Supporting Information) of the FN-silk networks show that cells were closely localized to the matrix. As described by Widhe et al., the interaction between FN-silk fibers and cells is facilitated by integrins, which bind to the RGD cell adhesion motif on the FN-silk.[48] 2.2 Biomechanical Characterization of FN-Silk Networks The biomechanical properties of stromal T-HESC FN-silk networks were characterized using bioindentation, a technique that allowed us to probe the mechanical properties at the whole-network scale (in the millimeter range, Figure 5A). Figure 5B shows a representative force-depth plot of an FN-silk network at culture day 6. The force-depth curve of the FN-silk network showed a linear increase in the applied compressive force during the loading phase, with a penetration depth of 46 µm at the maximum compressive load (Figure 5B), corresponding to ≈30–50% of the height of the FN-silk network. The unloading segment shows the FN-silk network's response to the release of the compressive force, providing insights into its elastic recovery behavior (Figure 5B). The FN-silk network showed a relatively low gradient of the unloading curve (Figure 5B), indicating a very elastic, low-stiffness material. The control curve (Figure 5B) shows a lower penetration depth of 5.5 µm and a steeper unloading curve gradient, indicating a higher stiffness material, in this case, glass. The compressive elastic modulus E of stromal T-HESC FN-silk networks was in the range of 0.41–0.57 kPa at culture day 3 and 0.40–0.52 kPa at culture day 6 (Figure 5C), indicating that no measurable differences in the compressive elastic modulus were detected at different culture periods. This suggests that the primary elastic properties stem from the FN-silk matrix itself, rather than any action of cells in the network. Figure 5D shows the estimated elastic modulus EIT of stromal T-HESC FN-silk networks. We estimated the elastic modulus of stromal T-HESC FN-silk networks to lie in the range of 1.8–4.6 kPa (Figure 5D). This elastic modulus is similar to that reported for endometrial tissue and materials used to model healthy/low stiffness conditions of the endometrium and endometriosis in vitro,[16, 38, 42, 52, 66-69] underscoring the ability of FN-silk to provide a physiologically relevant matrix in terms of the biomechanical environment that cells reside in. 2.3 Effect of Culture Formats on the Expression of Fibrotic Marker Genes Next, we explored how the culture format affected messenger RNA (mRNA) transcript levels of genes relevant to pathophysiological processes in endometriotic stroma and epithelium. When studying fibrosis, the biomechanical microenvironment is a highly relevant disease characteristic, as cellular fibrosis is in part regulated by biomechanical cues arising from the heterogeneous fiber network of the ECM.[37, 38, 53, 71] Modelling this 3D biomechanical environment is therefore desirable when studying fibrosis associated with endometriosis and testing drug candidates. For T-HESCs, we chose to assess the mRNA levels of the following genes: i) COL1A1, which codes for Collagen I α1, an extracellular matrix protein, ii) ACTA2, which codes for α-Smooth muscle actin (α-SMA), a marker of fibroblast-to-myofibroblast transdifferentiation, iii) FN1, which codes for Fibronectin-1, a protein involved in cell adhesion to the ECM which was shown to be elevated in patients with endometriosis compared to healthy patients,[72-74] and iv) SMAD3, which codes for the Smad3 protein, a signal transduction molecule in the canonical TGFβ signaling pathway.[31, 32, 40, 75-77] While a transcript-level upregulation of COL1A1, ACTA2, and FN1 indicated a pro-fibrotic state, SMAD3 was shown to be downregulated at the mRNA level upon exposure to TGF-β1 in T-HESCs.[78] For 12Z cells, we chose to study the following genes: i) SNAI1, which codes for the Snail transcription factor, ii) SNAI2, which codes for the Slug transcription factor, both of which repress E-cadherin to regulate epithelial-to-mesenchymal transition, and were identified as being the transcription factors most strongly affected by TGF-β1-induced epithelial-to-mesenchymal transition in 12Z cells.[29] Further, we studied the genes iii) CDH1, which codes for E-cadherin, a marker of epithelial cells involved in the formation of adherens junctions that characterize intact epithelium, and iv) CDH2, which codes for N-cadherin, a marker of mesenchymal cells that functions as a promoter of cellular motility. A pro-fibrotic state in epithelial 12Z cells was characterized by an upregulation of SNAI1, SNAI2, and CDH2 and a downregulation of CDH1 at the mRNA level. Cells of each type, stromal T-HESC, and epithelial 12Z were harvested from monolayers and brought into suspension. This suspension of “mother cells” was used to seed FN-silk networks, spheroids, and monolayers, separately for each cell type (Figure 6A). Cells in FN-silk networks, spheroids, and monolayers were cultured for 3 days in a complete growth medium and subsequently for 1 day in a serum-free medium. The mRNA levels of cells from FN-silk networks and spheroids were measured by reverse transcription quantitative polymerase chain reaction (RT-qPCR) and normalized to mRNA levels of cells cultured in monolayers (Figure 6A). For stromal T-HESCs, culture in FN-silk network and spheroid formats had similar effects, with mRNA levels differing by magnitudes of factor 2–3 compared to monolayers (Figure 6B). COL1A1 mRNA levels were upregulated by culture in 3D compared to monolayers, especially in spheroids, which may be attributable to more cell–cell interactions and a physiological stiffness, which may also be responsible for the downregulation of ACTA2 in 3D[38, 45] (Figure 6B). FN-1 mRNA, on the other hand, was upregulated in 3D culture compared to monolayers, as was SMAD3 mRNA (Figure 6B). While SMAD3 mRNA upregulation is consistent with anti-fibrotic effects in T-HESCs,[78] FN-1 mRNA upregulation is attributable to the formation of cell–cell contacts, as FN-1 mRNA was upregulated to similar magnitudes for cells in FN-silk networks and spheroids. Collagen I α1 and α-SMA proteins were expressed ubiquitously by cells in FN-silk networks (Figure 6C,D, respectively). The immunostaining provides spatial context, demonstrating the distribution and localization of these key proteins within the FN-silk network. Collagen I α1 showed higher expression on the outer edges of the FN-silk network (Figure 6C), which may be due to a higher number of cells present on the outside as evidenced by the vimentin stain (Figure 4A). α-SMA, on the other hand, was expressed at similar levels throughout the network (Figure 6D), which indicates that the biomechanical microenvironment provided by the FN-silk was homogeneous throughout the network. For epithelial 12Z cells, mRNA levels were affected similarly in both 3D culture formats relative to monolayers (Figure 6E). However, for these cells, culture in 3D upregulated mRNA levels of transcripts involved in fibrosis, namely, SNAI1, SNAI2, and CDH2 mRNAs, while CDH1 mRNA was downregulated (Figure 6E). This suggests that culture in 3D appears to favor a mesenchymal-type character of epithelial 12Z cells at the mRNA level. This observation contrasts with results from Stejskalová et al., who observed an upregulation of CDH1 mRNA and a downregulation of CDH2 mRNA in 12Z spheroids relative to monolayers, albeit using β-actin as a reference gene.[56] Epithelial-to-mesenchymal transition can be caused by hypoxia, although this is unlikely in FN-silk networks, as they are thin enough (100–150 µm) to allow sufficient oxygenation of cells in the network. It may be that different mechanisms are causing epithelial-to-mesenchymal transition in each 3D system, such as hypoxia in spheroids and increased cell proliferation in the FN-silk networks. 2.4 Induction of Fibrosis by TGF-β1 Even though stromal T-HESCs and epithelial 12Z cells were originally isolated from patients with leiomyoma and endometriosis, respectively, and therefore already show a degree of fibrosis, fibrosis can be induced further by treating cells with TGF-β1.[17, 29, 41, 78-81] First, we sought to characterize the effect of TGF-β1 treatment on cell metabolic activity to identify a suitable concentration to treat cells with. Figure 7A,B show the effect of TGF-β1 concentration on the viability of stromal T-HESC and epithelial 12Z cell monolayers, respectively. For T-HESCs, only 20 ng mL−1 TGF-β1, the highest concentration tested, reduced the viability compared to vehicle-treated cells. Other types of myofibroblasts have been observed to have higher proliferation upon exposure to TGF-β1;[82] however, this does not appear to be the case for stromal T-HESCs. For epithelial 12Z cells, TGF-β1 concentrations at 5 ng mL−1 and higher caused an increase in viability compared to vehicle-treated cells. This is consistent with an increase in proliferative/migratory behavior of 12Z cells observed during epithelial-to-mesenchymal transition induced by TGF-β1.[31] Figure 7C,D shows that TGF-β1 treatment at 10 ng mL−1 for 24 h had no effect on the metabolic activity of stromal T-HESC and epithelial 12Z cells cultured in FN-silk network and spheroid formats. Thus, we proceeded with treating cells with 10 ng mL−1 TGF-β1. This concentration of TGF-β1 was not only chosen based on our observations on metabolic activity but also because a concentration of 10 ng mL−1 TGF-β1 lies in the pathophysiological range of peritoneal fluid in endometriosis patients.[32, 79, 83-87] Next, we verified whether TGF-β1 treatment induced a response at the mRNA level indicative of the fibrotic behavior of cells. TGF-β1 upregulated COL1A1 and ACTA2 mRNA levels for cells in monolayers and FN-silk networks, but not in spheroids (Figure 7E). The behavior of stromal T-HESCs in monolayers agrees with results described previously for primary human endometrial stromal cells for COL1A1, ACTA2, FN-1, and SMAD3 mRNA level fold changes.[78] This indicates that stromal T-HESCs are a suitable cell line to model fibrotic pathophysiology. Even though stromal T-HESCs were inherently in a myofibroblast-like state, as they were isolated from a donor with leiomyoma,[88] TGF-β1 further augmented the myofibroblast-like character of these cells in terms of COL1A1 and ACTA2 mRNA levels. Culturing stromal T-HESCs in an FN-silk network format made them less responsive to TGF-β1 than monolayers (Figure 7E), which may be because the FN-silk matrix has a lower, more tissue-relevant stiffness, cells experience more cell–cell contacts, and/or integrin-mediated signaling. Studying fibrosis in monolayers may cause us to overestimate effect magnitudes compared to conditions mimicking the in vivo environment more closely, although the trends are similar. Figure 7E also shows that FN-1 mRNA was upregulated to a greater magnitude for cells in monolayers (factor 160–170%) than in FN-silk networks (factor 30–60%), although this difference was not large enough to be statistically significant. This may be because the mRNA levels of FN-1 were already upregulated by culture in an FN-silk network format relative to a monolayer (Figure 6B), as more cell–cell, and cell–ECM contacts are formed. SMAD3 mRNA levels, on the other hand, were downregulated by TGF-β1 treatment in all three formats, consistent with results from monolayers of primary human endometrial stromal cells, where TGF-β1 treatment induced downregulation of SMAD3 mRNA levels by ≈60%.[78] Stromal T-HESC spheroids showed little response to TGF-β1 at the mRNA level, showing no upregulation in COL1A1, ACTA2, and FN-1 mRNA levels compared to vehicle-treated spheroids (Figure 7E). This suggests that either stromal T-HESCs in spheroids were in a state where they did not respond to cytokines (either dead, or in a dormant state), or that the spheroid architecture was so dense that TGF-β1 was unable to penetrate to cells inside the spheroid and only acted on cells in the outer layer of the spheroid. Either way, our results suggest that spheroids are unsuitable to model TGF-β1 induced fibrosis in endometrial stroma in 3D. For epithelial 12Z cells, TGF-β1 upregulated SNAI1 mRNA levels in all three formats (Figure 7F), and to the greatest extent in FN-silk networks (upregulation by 80–100%) (Figure 7F). Epithelial 12Z cells cultured in all three formats showed a slight upregulation of SNAI2 mRNA levels, however, the magnitude was small, ranging from 10–50%, suggesting that TGF-β1 acted primarily via SNAI1 and to a lesser extent via SNAI2 at the mRNA level in epithelial 12Z cells (Figure 7F), consistent with observations from Ma et al.[29] Epithelial 12Z cells in spheroids responded to TGF-β1 exposure at the mRNA level (Figure 7F), hinting that 12Z cells in spheroids were more responsive than T-HESC spheroids, and that epithelial 12Z spheroid architecture allowed TGF-β1 to penetrate to the center of the spheroid. Our observations suggest that for epithelial 12Z cells, a loss of epithelial character was not coupled with a gain in mesenchymal character. Despite epithelial 12Z monolayers showing the smallest upregulation in SNAI1 mRNA, CDH1 mRNA was downregulated by the largest magnitude for cells in monolayers (downregulation by 50–70%) compared to FN-silk networks (downregulation by 10–23%) and spheroids (no change) (Figure 7F). This shows that culture in the monolayer caused TGF-β1 to have a greater effect on loss of epithelial character than culture in the FN-silk networks. Unexpectedly, we observed only minimal upregulation in CDH2 mRNA levels upon TGF-β1 treatment in all of the three culture formats (Figure 7F). For a complete epithelial-to-mesenchymal transition, we would expect CDH2 mRNA levels to be upregulated, as CDH2 codes for N-cadherin, a mesenchymal marker. Taken together, these results suggest that TGF-β1 treatment caused a loss in epithelial character of the 12Z cells, to the greatest extent in monolayers, but little gain of mesenchymal character. In contrast, Ma et al. observed an upregulation of CDH2 in 12Z cell monolayers by a factor of ≈50% upon TGF-β1 treatment.[29] Our observations are consistent with the partial epithelial-to-mesenchymal transition process observed in primary human endometriotic epithelial cells[42] and suggested by Konrad et al.[30] The effect of TGF-β1 observed in monolayer cultures was reproduced in FN-silk networks for stromal T-HESCs, where it induced fibroblast-to-myofibroblast transdifferentiation, but not in spheroids. In epithelial 12Z cells, TGF-β1 affected epithelial-to-mesenchymal transition across all formats, indicated by a loss of epithelial character with minimal mesenchymal gain. The distinct behavior of stromal cells in different culture formats indicates their sensitivity to the biomechanical microenvironment, with FN-silk networks offering a more physiologically relevant environment than stiff substrates or spheroids, allowing the fibrotic behavior of endometrial stromal cells to be modeled in 3D. Epithelial cells, however, show less dependence on the 3D culture format, with both FN-silk networks and spheroids responding similarly to TGF-β1 in terms of epithelial-to-mesenchymal transition. 2.5 Testing Pirfenidone, an Antifibrotic Drug Candidate with the Potential to be Repurposed for Endometriosis Treatment Exploring the similarities between endometriosis and other fibrotic diseases, such as idiopathic pulmonary fibrosis, could provide valuable insights for the development of pharmacological treatments for endometriosis.[10] Pirfenidone is a small-molecule drug indicated for idiopathic pulmonary fibrosis, a disease that shares common mechanisms with endometriosis. Pirfenidone acts by regulating the Wnt/β-catenin and TGF-β1/Smad2/3 signaling pathways.[89, 90] In a randomized, double-blind, prospective clinical trial with 210 patients undergoing laparoscopic surgery for endometriosis, it was found that 1800 mg of oral pirfenidone administered daily for six months post-surgery led to a reduction in fibrotic adhesions compared to the placebo group.[62] We chose to test the antifibrotic activity of pirfenidone on in vitro models of fibrosis in endometriosis developed here to evaluate their suitability as screening tools for antifibrotic drug candidates. Part of pirfenidone's mechanism of action is that it inhibits the proliferation of myofibroblasts. However, we were interested in pirfenidone's effect on the expression of genes involved in fibrotic pathophysiology, rather than cell proliferation. As such, we sought to identify a pirfenidone concentration that was high enough to be likely to affect mRNA levels, but low enough not to inhibit cell proliferation. To this end, we assessed the effect of the drug concentration on the viability of stromal T-HESCs and epithelial 12Z cells in monolayers. For both cell types, viability remained close to untreated cells at pirfenidone concentrations of up to 1 mg mL−1, while concentrations above 1 mg mL−1 reduced viability (Figure S7, Supporting Information). Therefore, we chose to treat cells with pirfenidone at a concentration of 0.8 mg mL−1 for drug testing experiments. We investigated the effect of pirfenidone on mRNAs involved in fibrosis in stromal T-HESCs and epithelial 12Z cells cultured in monolayers, FN-silk networks, and spheroids (Figure 8A). Each cell type/format combination was pre-treated with TGF-β1 or vehicle solution for 24 h and subsequently treated with either pirfenidone in serum-free medium or serum-free medium only for 24 h (Figure 8B). Stromal T-HESCs were treated with 0.82 ± 0.03 mg mL−1 pirfenidone and epithelial 12Z cells with 0.85 ± 0.04 mg mL−1 pirfenidone (mean ± SD from N = 3 independent experiments). We chose to pre-treat cells with TGF-β1 to assess whether pirfenidone can reverse fibrosis caused by the pre-treatment, making the model more relevant to in vivo conditions in which fibrosis is already present in endometriotic tissue before a drug is given. Figure 8C shows the effect of pirfenidone on COL1A1, ACTA2, and FN-1 mRNA levels for stromal T-HESCs pre-treated with TGF-β1. Pirfenidone downregulated COL1A1 mRNA levels in stromal T-HESCs in all three formats. Stromal T-HESC spheroids did appear to respond to pirfenidone treatment in terms of COL1A1, suggesting that the small molecule was able to diffuse into the core of the spheroid. Pirfenidone downregulated ACTA2 mRNA levels to similar magnitudes for stromal cells in monolayers (reduction by 50–70%) and FN-silk networks (reduction by 29–70%) (Figure 8C). For vehicle pre-treated cells, pirfenidone hardly downregulated ACTA2 mRNA in monolayers (reduction by 25–38%), and not at all in FN-silk networks, indicating that pirfenidone reversed upregulation of fibrosis-related mRNA transcripts induced specifically by TGF-β1 (Figure 8D). T-HESC spheroids, however, did not react to pirfenidone treatment in terms of ACTA2 mRNA levels (Figure 8C,D), suggesting that stromal T-HESC spheroids may be less well suited for testing antifibrotic drugs. Further, Figure 8C,D shows that pirfenidone had no effect on FN-1 mRNA levels of stromal T-HESCs neither in monolayers nor in FN-silk networks (Figure 8C,D). Surprisingly, T-HESC spheroids treated with pirfenidone showed a small downregulation of FN-1 mRNA levels in both TGF-β1 and vehicle pre-treatments (reduction by 29–50%). Together with our observations that stromal T-HESC spheroids were unresponsive to TGF-β1 at the FN-1 mRNA level (Figure 7E), this suggests that likely the culture in serum-free medium caused in reduction in FN-1 mRNA levels in stromal T-HESC spheroids.

shown in Figure 8C,D validated our screening platform: cellular fibrosis was reversed at the mRNA level when stromal cells were exposed to a known antifibrotic compound, pirfenidone. Pirfenidone downregulated mRNA levels of COL1A1 and ACTA2, hinting that this drug may have the potential to prevent ECM protein deposition and possibly reverse TGF-β1-induced fibroblast-to-myofibroblast transdifferentiation in endometriotic stroma. Pirfenidone's anti-fibrotic activity at the mRNA level was indicated similarly for stromal T-HESCs in monolayers and FN-silk networks, and to a lesser extent in spheroids. Although pirfenidone's efficacy has been mainly attributed to its action on fibroblasts,[91] we were interested in studying the effect of pirfenidone on epithelial cells, which form a substantial fraction of cells in the endometrium and endometriotic lesions.[92] Figure 8E,F show that 12Z cells generally behaved similarly in all three culture formats in terms of SNAI1, SNAI2, CDH1, and CDH2 mRNA levels, relative to the no-pirfenidone treated epithelial 12Z cells in each respective culture format. In TGF-β1 pre-treated 12Z cells, pirfenidone treatment upregulated SNAI1 mRNA levels by a factor of 40–160% and did not affect SNAI2 mRNA levels (Figure 8E). The trend of SNAI1 mRNA upregulation by pirfenidone was similar although slightly smaller in vehicle pre-treated 12Z cells across all culture formats (Figure 8F). CDH1 mRNA levels were unchanged in TGF-β1 pre-treated 12Z cells across all culture formats (Figure 8E) and downregulated by a factor of 20–71% in control pre-treatments (Figure 8F). However, CDH2 transcription remained largely unaffected by pirfenidone treatment. Taken together, these mRNA-level results from 12Z cells across all culture formats provide tentative evidence that pirfenidone facilitated epithelial-to-mesenchymal transition in 12Z cells independent of pre-treatment, albeit to a small magnitude. Together with evidence of pirfenidone's effect on other types of epithelial cells,[93-97] our results suggest that this drug's primary anti-fibrotic mechanism of action stems from its effect on fibroblasts by reversing fibroblast-to-myofibroblast transdifferentiation, rather than its effect on epithelial cells. 3 Conclusion and Outlook In this study, we have developed 3D culture models of fibrosis in endometriosis and used these models to evaluate a candidate antifibrotic drug with the potential to be repurposed for endometriosis. We demonstrated that FN-silk networks, developed by Johansson, Widhe et al.[46] and Collodet et al.[47] and adapted slightly here, were a suitable matrix to culture endometrial stromal cells (T-HESCs) in 3D, while spheroids were less suitable in terms of their ability to mimic the fibrotic character of stromal cells found in endometriosis. In addition to being fully defined, non-animal origin, and readily available, FN-silk as a scaffold material was shown to be advantageous due to its tissue-mimicking stiffness. We showed that stromal cells cultured in FN-silk behaved physiologically in terms of cell growth and mRNA-level response to the profibrotic cytokine TGF-β1, while the culture of stromal T-HESCs in a spheroid format did not enable this physiological behavior to be modeled as comprehensively. For epithelial cells, FN-silk and spheroid formats partially modeled fibrotic processes in 3D culture, as probed at the mRNA level, indicating that both formats were suitable to model fibrosis in epithelial 12Z cells cultured in 3D. Testing pirfenidone on these in vitro models provided evidence for the drug's anti-fibrotic effects on endometriotic stroma by downregulating mRNA levels of transcripts implicated in fibroblast-to-myofibroblast transdifferentiation. This demonstrated that the stromal cell FN-silk network 3D culture format is an effective drug screening tool for evaluating antifibrotic candidates. However, pirfenidone upregulated epithelial-to-mesenchymal transition-promoting mRNA levels in epithelial 12Z cells in 3D, albeit to a small magnitude, highlighting the drug's distinct effects on stromal and epithelial components of endometriotic lesions. These observations underscore the utility of in vitro systems to closely characterize the behavior of different cell types in response to drug treatment. In this study, we used qPCR to reliably and quantitatively compare culture formats in terms of changes to their gene expression patterns in response to TGF-β1 and pirfenidone treatment. For future applications of these models in drug testing, additional protein-level methods should be considered to provide a more comprehensive understanding of the candidate drug's effects. This would allow novel endometriosis therapies[103, 104] to be evaluated more comprehensively in vitro. In terms of developing the model further, future work should focus on bringing stromal and epithelial cells into co-culture on an FN-silk network, as this co-culture model may allow the precise orchestration of TGF-β signaling between stromal and epithelial cells[92] to be interrogated in vitro. Additionally, there is an opportunity for the endometriosis research community to set benchmarks against which in vitro models of endometriosis should be compared, to allow model developers to validate their in vitro systems and to enable comparison between different models. We suggest the fibrotic character of endometriosis be included as a criterion of physiological relevance. By continuing to develop and refine these sustainable in vitro models, we can enhance pharmaceutical research on endometriosis while aligning with the UN SDGs, ultimately contributing to both innovative treatments for women's health and the promotion of environmentally responsible research practices. Our work demonstrates that an SDG-guided framework can be effectively applied at various stages of pharmaceutical research and development, from project selection to methodology. By focusing on women's health and minimizing animal experiments, we ensure that our research contributes to sustainable practices that are both ethically sound and scientifically rigorous. This work paves the way for more efficient, responsible, and potentially meaningful approaches to tackling complex diseases like endometriosis, underscoring the importance of integrating sustainable development goals into the heart of pharmaceutical research. 4 Experimental Section Cell culture Immortalized human endometrial stromal cells (T-HESC cell line,[88] Cellosaurus accession no. CVCLC464, female sex) were purchased from Applied Biological Materials, Canada (Ref T0533). T-HESCs were cultured at 37 °C in a humidified atmosphere containing CO2 (5%). The complete growth medium composition was DMEM/F12 without phenol red (Gibco Ref 21041025, Paisley, UK), charcoal-stripped fetal bovine serum (10 v/v%, Gibco Ref 12676029, Grand Island, NY, USA), ITS+ Premix Universal Culture Supplement (1 v/v%, Corning Ref 354352, Bedford, MA, USA [final concentrations in medium: human recombinant insulin (5 µg mL−1); transferrin (5 µg mL−1); selenous acid (5 ng mL−1)]) and penicillin–streptomycin (1v/v%, Gibco Ref 15140122, Grand Island, NY, USA). Subcultivation of monolayers was performed at 70–80% confluence by detachment using TrypLE (Gibco Ref 12604013, Paisley, UK). For experiments, cells from passage numbers 5–18 were used, where passage number 0 was defined as the delivered vial. Genomic profiling was done at passage 8 using highly polymorphic short tandem repeat loci (STRs). STR loci were amplified using the PowerPlex 16 HS System (Promega), fragment analysis was done on an ABI3730xl (Life Technologies), and the resulting data were analyzed using the GeneMarker HID software (Softgenetics). A 100% match to the DNA profile of the T-HESC line was found. Full STR profile results are shown in Table S1 and Figure S8 (Supporting Information). Immortalized human epithelial endometriotic cells (12Z cell line,[98] Cellosaurus accession no. CVCL0Q73, female sex) were purchased from Applied Biological Materials, Canada (Ref T0764). 12Z cells were cultured at 37 °C in a humidified atmosphere containing CO2 (5%). The complete growth medium composition was Dulbecco's Modified Eagle Medium (DMEM) with glucose (4.5 g L−1), without sodium pyruvate (Gibco Ref 11960044, Paisley, UK), fetal bovine serum (10 v/v%, Merck EMD Millipore, Ref ES-009-B, Burlington, MA, USA), l-glutamine (200 mm, Gibco 25030024, Paisley, UK) and penicillin–streptomycin (1 v/v%, Gibco Ref 15140122, Grand Island, NY, USA). Subcultivation of monolayers was performed at 70–90% confluence by detachment using TrypLE (Gibco Ref 12604013, Paisley, UK). For experiments, cells from passage numbers 9–22 were used, where passage number 0 was defined as the delivered vial. Genomic profiling was done at passage 14 and a 96% match to the STR profile of the 12Z cell line was found. Differences were found in the D8S1179 and FGA loci (Table S2, Figure S9, Supporting Information). Cultures were routinely tested for mycoplasma (MycoStrip Mycoplasma Detection Kit, InvivoGen, San Diego, CA, USA) and no mycoplasma contamination was detected. 3D Culture—FN-Silk Networks FN-silk networks were prepared as described previously[46, 47] with adaptations, as shown in Figure 2A. Cells were harvested and brought into suspension in a complete growth medium. FN-silk in PBS (Spiber Technologies AB, Stockholm, Sweden) was freshly thawed and mixed with the cell suspension at a 2:1 volumetric ratio to make a master mix containing liquid FN-silk protein and suspended cells. The final concentration of FN-silk in the master mix was 2.2 mg mL−1 and the cell concentration in the master mix was 1 × 106 cells mL−1 for T-HESC and 2 × 106 cells mL−1 for 12Z cells, such that the number of cells per FN-silk network was 10000 cells for T-HESC and 20000 cells for 12Z. A droplet of master mix (10 µL) was placed at the bottom of each well in a U-shaped-bottom ultralow attachment plate (Nunclon Sphera-Treated U-Bottom Microplate, Thermo Fisher Scientific Ref 174929, Japan). A multichannel pipette set to 28 µL was used to rapidly pipette air into the mastermix droplet 30 times, to allow the silk protein to assemble into β-sheets at the air-liquid interface and thereby form a foam. The foam was incubated in a humidified incubator at 37 °C for 12 min. Each foam was suspended in a complete growth medium (180 µL) and, using a μ-spoon, transferred to a new well containing fresh complete growth medium (180 µL). The FN-silk foam transformed into a network as the sheets around the bubbles burst. Cells in FN-silk networks were cultured for up to 7 days and the medium was changed every 2−3 days. At day 2, all air bubbles had disappeared and FN-silk networks were submerged below the surface of the medium in the center of the well. Integration of cells in FN-silk networks was measured by counting the cells left in the medium after transferring the FN-silk network to a well containing fresh medium. Cells were stained with a Trypan blue solution (0.4%, Merck Ref T8154, Switzerland) and counted using a Neubauer plate. 3D Culture—Spheroids Spheroids were prepared by seeding a cell suspension into a U-shaped-bottom ultralow attachment plate (Nunclon Sphera-Treated U-Bottom Microplate, Thermo Fisher Scientific Ref 174929, Japan) and allowing cells to spontaneously aggregate into spheroids, as described previously[63] and illustrated in Figure 2B. The number of cells seeded per well was equivalent to the FN-silk networks. Spheroids were cultured in 100 µL medium per well for up to 7 days and the medium was changed every 2−3 days. Metabolic Activity—Monolayers Metabolic Activity—3D systems The metabolic activity of cells cultured in 3D formats was measured using the CellTiter Glo Cell 3D Viability Assay (Promega Ref G9681, Madison, WI, USA). The CellTiter Glo 3D Viability Assay was performed according to the manufacturer's instructions with minor modifications. All materials were equilibrated to 24 °C and all steps performed in low-light conditions. Each 3D cell unit was transferred to a well of a white plate (pureGrade S polystyrene flat-bottom 96 well plates, Brand Ref 781665, Wertheim, Germany) containing complete growth medium (80 µL) using a μ-spoon for FN-silk networks and wide-orifice pipette tips (Finntip Wide Orifice Pipette Tips, Thermo Fisher Scientific Ref 9405163) for spheroids. Cell systems were equilibrated to 24 °C for 25 min in the dark. CellTiter Glo 3D Viability reagent (80 µL) was added to each well and the contents were mixed vigorously for 5 min to induce cell lysis. The plate was incubated for 25 min in the dark and luminescence was measured using a plate reader (Tecan Infinite 200 PRO F Plex/M Nano+, Switzerland). The CellTiter Glo Cell 3D Viability Assay was validated by comparison to a manual count of cells (Figure S4, Supporting Information). In addition to metabolic assays, the viability of cells in 3D systems was assessed visually using a live/dead staining kit (LIVE/DEAD Viability/Cytotoxicity Assay Kit, Invitrogen Thermo Fisher Scientific Ref L32250, Eugene, OR, USA). The live/dead assay was performed according to the manufacturer's instructions. Briefly, a working solution with a composition SYTOX Deep Red Nucleic Acid Stain (0.25 µM) and calcein AM (2 µM) in a complete growth medium was prepared. Cells were incubated in the working solution at 37 °C, 5% CO2 for 30 min in the dark. FN-silk networks and spheroids were visualized directly in the well using a wide-field microscope (Kinetix Ti2, Nikon, Japan), with the GFP filter set for calcein-stained cells and the deep red/Cy5 filter set for SYTOX Deep Red-stained cells. Fibrosis Induction A profibrotic state was induced in both T-HESC and 12Z cells by treatment with human recombinant TGF-β1 (Stemcell Ref 78067, USA). TGF-β1 was reconstituted in hydrochloric acid (HCl, 10 mm) and diluted in a vehicle solution of bovine serum albumin (BSA, final concentration 0.1 w/v%) according to the manufacturer's instructions. Unless stated otherwise, cells were treated with TGF-β1 at a concentration of 10 ng mL−1 in a serum-free medium for 24 h, as described by.[29, 31, 32, 79, 83] The corresponding composition of the vehicle, which was used as a control, was 0.0001 w/v% BSA, 1 µM HCl (final concentration in serum-free medium). Serum-free medium was chosen to allow the medium composition to be fully defined during TGF-β1 treatment. Pirfenidone Treatment Pirfenidone (5-methyl-1-phenylpyridin-2(1H)-one) was purchased from BLD Pharm, Germany. Pirfenidone was dissolved in a serum-free medium and the concentration was measured by absorbance at λ = 311 nm in a black/clear bottom quartz microplate using a plate reader (Tecan Infinite 200 PRO F Plex/M Nano+, Switzerland). Unless otherwise stated, T-HESCs were treated with 0.82 ± 0.03 mg mL−1 and 12Z cells with 0.85 ± 0.04 mg mL−1 pirfenidone (mean ±SD, N = 3 independent experiments) in serum-free medium for 24 h. First, cells were treated with TGF-β1 for 24 h to induce a fibrotic state, followed by treatment with pirfenidone for 24 h. Control pre-treatments were vehicle in serum-free medium and complete growth medium. Immunofluorescent Staining, Wide Field, and Confocal Microscopy Whole FN-silk networks were stained for EpCAM, Vimentin, Collagen I α1 (COL I α1), and α-Smooth muscle actin (α-SMA) using immunofluorescent staining. FN-silk networks were fixed in paraformaldehyde (4%, PFA, Thermo Fisher Scientific Ref J61899-AK) in PBS for 20 min at 24 °C and stored in PBS at 4 °C until staining. Networks were incubated in Sudan Black (SB, Thermo Fisher Scientific Ref 190160250, 0.3 w/v% in 70 v/v% ethanol, filtered through a 0.2 µm polyethersulfone membrane) for 15 min to minimize autofluorescence of the FN-silk protein. After washing with PBS three times, cells were permeabilized with Triton-X100 (0.2% in PBS, Thermo Fisher Scientific Ref A16046.AP) for 10 min. FN-silk networks were washed with Tween 20 (0.1% in PBS, Sigma–Aldrich Ref 822184) for 10 min while agitating. Cells were blocked by incubating in goat serum (1 v/v% in PBS/Tween 20 0.1%, Capricorn Scientific Ref GOA-1A) for 1 h. The primary antibody (Table 1) was diluted in goat serum (1 v/v% in PBS/Tween 20 0.1%) and FN-silk networks incubated with the primary antibody overnight at 4 °C. FN-silk networks were washed with Tween 20 (0.1% in PBS) for 10 min while agitating. The secondary antibody (Table 1) was diluted in 1 v/v% goat serum (1 v/v% in PBS/Tween 20 0.1%) and FN-silk networks incubated with the secondary antibody for 2 h at 24 °C in the dark. FN-silk networks were incubated with SB for 10 min. SB was rinsed off the networks by washing with Tween 20 (0.1% in PBS) three times. Nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI, Carl Roth, Karlsruhe, Germany, Ref 6335.1) and F-actin was stained with Alexa Fluor 594 Phalloidin (Invitrogen Ref A12381, Eugene, OR, USA). FN-silk networks were washed with Tween 20 (0.1% in PBS) for 10 min while agitating. As a control, cells were stained with secondary antibody only (no primary antibody) to check for any unspecific binding (Figure S11, Supporting Information). | Name | Class, isotype | Dilution | Ref no. | Supplier | |---|---|---|---|---| | Primary antibodies | |||| | Vimentin Rabbit anti-human | Monoclonal IgG | 1:500 | ab92547 | Abcam | | EpCAM Rabbit anti-human | Monoclonal IgG | 1:500 | ab223582 | Abcam | | COL I α1 Rabbit anti-human | Polyclonal IgG | 1:500 | ab34710 | Abcam | | α-SMA Rabbit anti-human | Polyclonal IgG | 1:500 | PA5-19465 | Thermo Fisher Scientific | | Secondary antibody | |||| | AlexaFluor 488 Goat anti-rabbit | Polyclonal IgG | 1:1000 | A32731 | Thermo Fisher Scientific | For wide-field microscopy, a drop of DAKO fluorescence mounting medium (Agilent Ref S302380-2, Santa Clara, CA, USA) was placed on a glass slide, the network placed into the mounting medium using a µ-spatula, covered with a glass cover slip and sealed around the edges using nitrocellulose (nail varnish). Networks were visualized using a Ti2 Kinetix wide-field microscope with a 10X/0.45 air objective (Nikon, Tokyo, Japan). For confocal microscopy, DAKO mounting medium (40 µL) was placed in each well of a glass-bottom µ-well slide (15 well 3D glass bottom µ-slide, ibidi Ref 81507, Gräfelfing, Germany). Each FN-silk network was transferred to a well using a µ-spatula and visualized using a TCS SP8 DLS Light Sheet microscope in confocal mode with a 10X air objective (Leica Microsystems, Wetzlar, Germany). 4.1 RNA Extraction, cDNA Synthesis, and qPCR—RNA Extraction Total RNA was extracted using the ReliaPrep Cell and Tissue Miniprep System (Promega Ref Z6211, Madison, WI, USA) according to Section 5 of the manufacturer's protocol (RNA Isolation and Purification from Cell Samples), with minor adaptations. Cells were harvested from monolayers using TrypLE. 3D systems were pooled by placing 14−16 units (FN-silk networks or spheroids) in a microcentrifuge tube with PBS. Cells were centrifuged, the supernatant removed, and the pellet frozen at −80 °C. Cells in a frozen pellet were lyzed and 3D systems were broken up by pipetting up and down 100 times. The nucleic acid/cell debris mix was diluted and centrifuged at 12 000 × g for 5 min to pellet FN-silk and cell debris. Taking the supernatant, nucleic acids were bound to a Reliaprep MiniColumn, washed, eluted, and treated with DNase. RNA was bound to a second Reliaprep MiniColumn. After washing, RNA was eluted and quantified using spectrophotometry at λ = 260 nm (NanoDrop 2000, Thermo Fisher Scientific, USA). RNA Extraction, cDNA Synthesis, and qPCR—cDNA Synthesis RNA was reverse transcribed to complementary DNA (cDNA) using the GoScript Reverse Transcriptase Random Primers kit (Promega Ref A2801, Madison, WI, USA) according to the manufacturer's instructions. RNA Extraction, cDNA Synthesis, and qPCR—qPCR cDNA levels were measured by quantitative PCR using the GoTaq qPCR kit (Promega Ref A6002, Madison, WI, USA) according to the manufacturer's instructions under standard cycling conditions. Thermal cycling and fluorescence acquisition were performed in a Rotor-Gene Q 2Plex System (Qiagen, Germany). RT minus reactions were run in every experiment to control for DNA contamination. Relative quantification was performed using the ∆∆Ct method[100] using GAPDH as a reference gene. Primers were synthesized by Microsynth (Balgach, Switzerland) and had amplification efficiencies of 80–98%, similar to GAPDH, over a 100-fold concentration range, allowing the ∆∆Ct method to be used.[100] Primer sequences and validations are provided in Table S3, (Supporting Information). GAPDH was found to be a suitable reference gene, as its mRNA levels were unaffected by TGF-β1 treatment (Figure S12, Supporting Information). Mechanical Characterization Morphological Analysis The size of cells in monolayers, of FN-silk networks and spheroids was measured by an automated method using CellPose, a generalist algorithm for cellular segmentation.[102] Code was written and executed in Python 3.10.13 in a Conda 23.11.0 environment and is available on GitHub at https://github.com/Luciani-Group. Immunofluorescent staining images were visualized using Imaris 10.1.0 (Oxford Instruments, Abingdon, UK). Confocal microscopy images are displayed as maximum-intensity projections of the z-stack. Statistical Analysis All results are reported as mean ± standard deviation (SD) of three independent replicates unless stated otherwise. Statistical analysis was performed in GraphPad Prism 10.3.0 for Windows (GraphPad Software, San Diego, USA). Pairs were compared using the Mann–Whitney test. For experiments with three or more groups, the Kruskal–Wallis test with Dunn's multiple comparisons was used. To check the significance of differences between different formats, each group (monolayers, FN-silk networks, or spheroids) was compared to every other group. To check the significance of differences between the treatment groups and control, each treatment group was compared to the control. Significance values were chosen as *p < 0.1; **p < 0.05.

The authors wish to thank Spiber Technologies AB for generously providing the FN-silk protein. S.T. wishes to thank Dr. Simone Aleandri for helpful discussions on statistical analysis and Astrid Källén, Savvina Gkouma, Dr. Kelly Blust, Dr. Inês Marquez, Dr. Yury Belyaev, Dr. Cristina Zivko, Dr. Thomas Tapmeier, Dr. Assad Riaz, Dr. Anna Stejskalová, and Dr. Shannon Hawkins for sharing their expertise on cell culture methods. S.T. wishes to thank Dr. Jiri Nohava from Anton Paar TriTec SA for the Bioindenter measurements. S.T., M.C.B., and P.L. thank Prof. Dr. Oliver Mühlemann and Nicole Kleinschmidt for providing resources for RT-qPCR analysis. S.T. acknowledges a Short Travel Grant for (Post) Docs from the University of Bern and the Microscopy Imaging Center at the University of Bern. P.L. acknowledges the Swiss National Science Foundation, grant number 215227, for funding part of this project. M.H. acknowledges the Swedish Research Council, grant number 2022-04209, for funding part of this project. Some figures were created with BioRender.com. Conflict of Interest No private study sponsors had any involvement in the study design, data collection, or interpretation of data presented in this manuscript. P.L. declares the following competing interests: she has consulted and received research grants on unrelated projects from Lipoid, Sanofi-Aventis Deutschland and DSM Nutritional Products Ltd. M.H. has shares in Spiber Technologies AB, a company that aims to commercialize recombinant silk. S.T., M.C.B., L.A.G. and M.W. declare no competing interests. Author Contributions S.T. performed conceptualization (equal), data curation (equal), formal analysis (equal), funding acquisition (supporting), investigation (equal), methodology (lead), validation (supporting), visualization (lead), wrote the original draft (lead), and wrote, reviewed, and edited the final manuscript (equal). M.C.B. performed data curation (equal), formal analysis (equal), investigation (equal), methodology (supporting), and validation (lead), and wrote, reviewed, and edited (equal) the final draft. M.W. developed the methodology (supporting), performed supervision (supporting), and wrote, reviewed, and edited (equal) the final draft. L.A.G. developed the methodology (supporting), performed supervision (supporting), and wrote, reviewed, and edited the final draft (supporting). M.H. acquired funds (supporting), performed project administration (supporting), acquired resources (supporting), performed supervision (supporting), and wrote, reviewed, and edited the final draft (equal). P.L. performed conceptualization (equal), funding acquisition (lead), project administration (lead), supervision (lead), acquired resources (lead), and wrote, reviewed, and edited the final draft (equal). Data Availability Statement The data that support the findings of this study are available from the corresponding author upon reasonable request.