Cancer-Associated Fibroblasts Actively Compress Cancer Cells (2023) - PDF

Article https://doi.org/10.1038/s41467-023-42382-4 Cancer-associated fibroblasts actively compress cancer cells and modulate...

Article https://doi.org/10.1038/s41467-023-42382-4 Cancer-associated fibroblasts actively compress cancer cells and modulate mechanotransduction Received: 20 April 2022 Jorge Barbazan 1,2,9, Carlos Pérez-González1,9, Manuel Gómez-González 3 , Mathieu Dedenon4, Sophie Richon1, Ernest Latorre3, Marco Serra4, Accepted: 9 October 2023 Pascale Mariani5, Stéphanie Descroix 4, Pierre Sens 4, Xavier Trepat 3,6,7,8 & Danijela Matic Vignjevic 1 Check for updates 1234567890():,; 1234567890():,; During tumor progression, cancer-associated fibroblasts (CAFs) accumulate in tumors and produce an excessive extracellular matrix (ECM), forming a cap- sule that enwraps cancer cells. This capsule acts as a barrier that restricts tumor growth leading to the buildup of intratumoral pressure. Combining genetic and physical manipulations in vivo with microfabrication and force measurements in vitro, we found that the CAFs capsule is not a passive barrier but instead actively compresses cancer cells using actomyosin contractility. Abrogation of CAFs contractility in vivo leads to the dissipation of compressive forces and impairment of capsule formation. By mapping CAF force patterns in 3D, we show that compression is a CAF-intrinsic property independent of cancer cell growth. Supracellular coordination of CAFs is achieved through fibronectin cables that serve as scaffolds allowing force transmission. Cancer cells mechanosense CAF compression, resulting in an altered localization of the transcriptional regulator YAP and a decrease in proliferation. Our study unveils that the contractile capsule actively compresses cancer cells, mod- ulates their mechanical signaling, and reorganizes tumor morphology. Cancer progression is the result of complex interactions between One of the most abundant cell types in the stroma are cancer- cancer cells and their microenvironment1. Cancer cells continuously associated fibroblasts (CAFs)5. CAFs play multiple roles in cancer sense signals from the surroundings that could be either biochemical, progression, promoting cancer cell survival and proliferation and such as soluble molecules or membrane receptors on stromal cells, or modulating cancer invasion and immune response5,6. CAFs are highly physical, including stiffness, the microarchitecture of the surrounding contractile cells that serve as the main producers of the ECM7,8. extracellular matrix (ECM), fluid pressure and solid stress2. Solid stress Together with the ECM, they form a capsule around the tumor9,10. is a consequence of the proliferation of cancer cells against a viscoe- However, whether this capsule is just a barrier that passively opposes lastic boundary, the stroma, which resists tumor growth and prevents tumor growth or has an active role in the generation of tumor stresses its expansion, resulting in the buildup of internal pressure3,4. remains unknown. In this work we investigated the role of CAFs’ 1 Institut Curie, PSL Research University, CNRS UMR 144, F-75005 Paris, France. 2Translational Medical Oncology Group (ONCOMET), Health Research Institute of Santiago de Compostela (IDIS), University Hospital of Santiago de Compostela (SERGAS), 15706 Santiago de Compostela, Spain. 3Institute for Bioengi- neering of Catalonia (IBEC), The Barcelona Institute for Science and Technology (BIST), 08028 Barcelona, Spain. 4Institut Curie, PSL Research University, CNRS UMR 168, F-75005 Paris, France. 5Institut Curie, Department of surgical oncology, Curie Institute, F-75005 Paris, France. 6Facutltat de Medicina, Universitat de Barcelona, 08036 Barcelona, Spain. 7Institució Catalana de Recerca i Estudis Avançats (ICREA), Barcelona, Spain. 8Centro de Investigación Biomédica en Red en Bioingeniería, Biomateriales y Nanomedicina (CIBER-BBN), 08028 Barcelona, Spain. 9These authors contributed equally: Jorge Barbazan, Carlos Pérez-González. e-mail: [email protected]; [email protected] Nature Communications | (2023)14:6966 1 Article https://doi.org/10.1038/s41467-023-42382-4 a nGFP pMLC b GFP SHG GFP SHG SHG Infiltrated areas Control Stroma-free areas 150 Relative tumor area (%) 100 50 αSMA:MyosinIIA-/- 0 -/- l ro IIA nt sin Co yo M A: M αS c PAA gel PDMS stencil d t=0 h t=17,5 h t=36 h Cancer cells CAFs Traction (Pa) cancer cell clusters CAFs Fs lls CA e f er ce Integrated radial traction force (μN) Radial traction (T) (Pa) Canc CAFs ring area (μm2) Time (h) Distance (μm) Time (h) g h i Traction peak magnitude (Pa) pMLC F-actin DNA pMLC Pre-Bleb 6h +Bleb Normalized CAFs ring area Traction (Pa) F-Actin Time (h) Fig. 1 | CAFs contractility is required for capsule formation and tumor com- co-culture, evolving over time. Black line-the contour of the cancer cell cluster Scale partmentalization. a Representative (from n = 8 positions from N = 2 mice) bar, 100 µm. e Representative kymograph of circumferentially averaged radial Phospho-Myosin Light chain (pMLC) staining (magenta) in a tumor tissue section tractions as a function of the distance to center of the cancer cell cluster (outwards from N/p53/mTmG mice. Tumor cells express a nuclear GFP (nGFP). Scale bar pointing tractions are positive, and inwards pointing tractions negative). Black line- 100 µm. b Tumors generated after subcutaneous engraftment of N/p53 organoids the boundary of the cancer cell cluster. f Time evolution of the integrated radial into control and mice containing myosin IIA-knockout CAFs (quantified in right traction force near the boundary of the cancer cell cluster in E (orange) and CAFs plot). Middle panel: Cancer cells (nuclear GFP), Myosin IIA KO CAFs (membrane- ring area (purple). g Representative (from N = 2 independent experiments) image GFP, green). Collagen I, SHG (magenta). Scale bar: 300 µm. Left panel, magnified of CAFs-cancer cells co-culture stained for pMLC (green), F-actin (magenta) and boxed regions. Scale bar, 100 µm. Right panel, stroma-free areas based on SHG DNA (blue). Insets, magnified boxed region. Scale bar, 100 µm. h Traction maps signal. Right bar chart: quantification of stroma-free (purple) or infiltrated (orange) (quantified in panel i) overlaid on a DIC image of cancer cell-CAFs after ≈40 h co- tumor areas. Data, mean ± SD, N = 2 mice/group. c Top panel, in vitro co-culture culture (Pre-Bleb) and 6 h after addition of blebbistatin (6 h + Bleb). Solid red line, protocol and right panel: representative image (of N = 3 independent experiments) the contour of CAFs ring; dashed line, CAFs ring contour before blebbistatin. Scale of a cluster of primary cancer cells (green) surrounded by CAFs isolated from PDX bar, 100 µm. i Average radial traction peak magnitude near the cluster cell (magenta). Scale bar, 100 µm. Lower panel, 3D rendering of co-cultures after 48 h. boundary (orange) and CAFs ring area (purple) normalized to the initial ring size, as Lower panel, orthogonal Z section. Scale bar: 100 µm. d Representative (of N = 3 a function of time. Dashed black line, blebbistatin addition. Data mean ± SD, n = 30 independent experiments) traction maps overlaid on DIC image of cancer cell-CAFs clusters, N = 3 independent experiments. Nature Communications | (2023)14:6966 2 Article https://doi.org/10.1038/s41467-023-42382-4 contractility in capsule formation and how CAFs mechanically interact To understand the mechanics of cancer cell-CAF interaction, we with cancer cells. quantified tissue forces using traction force microscopy (TFM) (Fig. 1d, Supplementary Fig. 6, Supplementary Video 5). Forces fluctuated Results and discussion across the tissue but systematically accumulated at the boundary To study the organization of CAFs in tumors, we used a transgenic between the CAFs and the cancer cell cluster. Given the radial sym- mouse model that spontaneously develops tumors in the intestinal metry of the system, we decomposed tractions into radial (Tr) and epithelium due to the expression of an activated Notch1 (NICD) and tangential (Tt) components, respective to the cluster boundary. We the deletion of the tumor suppressor p5310 (hereafter referred to as circumferentially averaged these tractions for each timepoint and N/p53/mTmG) (Supplementary Fig. 1a). Cancer cells were visualized plotted them as a function of the distance to the cluster center to build through the expression of a nucleus- and membrane-targeted GFP, kymographs (Fig. 1e, Supplementary Fig. 6b). We found that average while all other cells expressed a membrane-targeted tdTomato. The tangential forces, which likely correspond to migration of CAFs around stroma, composed of CAFs and extracellular matrix (ECM), formed a cancer cell clusters, remained low and constant over time (Supple- thick capsule enveloping the tumor. Interestingly, the stroma mentary Fig. 6b, c). In contrast, radial tractions were maximal at the penetrated the tumor, compartmentalizing it into smaller clusters boundary between cancer cells and CAFs (Fig. 1d, e, Supplementary (Supplementary Fig. 1b; Supplementary Video 1). Each cluster was Fig. 6c). Initially, radial tractions were positive (pointing away from enwrapped with aligned CAFs forming intratumoral capsules. This cancer cells), probably due to CAFs crawling towards the cluster was remarkably similar to the typical organization of human color- (Fig. 1d–f). Later, when CAFs aligned parallel to the cluster boundary, ectal cancers9. Given that intratumoral capsules were rich in phos- radial tractions became negative (pointing towards cancer cells). phorylated, thus active, myosin II (Fig. 1a), we wondered whether These negative radial forces progressively increased as CAFs formed CAFs contractility could play a role in tumor compartmentalization. the supracellular ring, slid on top of the cancer cells and induced To test this, we transplanted N/p53 tumor organoids into budding, until a point where these tractions stabilize to a plateau value transgenic mice in which we conditionally ablated myosin IIA in (Fig. 1f). Notably, the same traction pattern was observed on softer αSMA-expressing cells (αSMA-CreERT2; R26mT/mG; myosin IIAfl/fl). In substrates (2 kPa) (Supplementary Fig. 7a). this mouse model, cancer cells expressed nuclear GFP. All stromal To better understand how CAFs induce the remodeling of cancer cells, including CAFs with wild-type myosin IIA, expressed a cell clusters, we developed a physical model that relates the closure membrane-targeted tdTomato. In αSMA-expressing cells, tdTomato dynamics of the CAF ring and the traction forces to the transmission of switched to GFP concomitantly with the induction of myosin IIA stress between CAFs and cancer cells (Supplementary Theoretical knockout (Supplementary Fig. 2). In control mice, as in sponta- Note 1). In the model, CAF closure is assumed to be driven by a purse- neously developed tumors, CAFs generated intratumoral capsules string mechanism11, which corresponds to a line tension acting at the that compartmentalized and confined cancer cells (Fig. 1b). These inner edge of the CAF monolayer. The increase of inward-pointing capsules were largely absent from mice in which CAFs lacked myosin traction forces as the ring closes reflects the increase of tension at the IIA and, instead, cancer cells and CAFs appeared mixed. To quantify CAF inner edge, which is, to a first approximation, proportional to the this, we segmented large tumor areas using the collagen signal ring curvature for a constant line tension (Supplementary Theoretical visualized by second harmonic generation (SHG) imaging as a proxy Note 1). The observation of inward-pointing traction forces under- for tumor-stroma boundary. We found that control tumors con- neath the cancer cells (Fig. 1e) suggests that the CAFs generate shear tained larger stroma-free areas than tumors containing myosin IIA stress on the cancer cell, modeled as viscous friction between the two depleted CAFs (Fig. 1b). These data show that CAFs’ contractility is layers. We also observed the long-term persistence of traction forces required for the formation of intratumoral capsules, and thus seg- after the ring had stalled (Fig. 1f), which prompted us to adopt an regation of cancer cells into compartments. elastic description of the CAFs and cancer cell cluster and an elastic To study the mechanism by which CAFs confine cancer cells, we interaction with the substrate. This yields a spatial localization of the followed a bottom-up approach and developed an in vitro co-culture traction forces outside the cancer cell cluster, as seen in the experi- system. First, we isolated cancer cells and CAFs from patient-derived ments (λs~50 μm in Supplementary Theoretical Note 1, Fig. 1b). xenografts (PDX) of colorectal cancer (Supplementary Fig. 3). As in the The shear stress induced by the CAF motion on top of the cancer tumors, most cultured CAFs expressed αSMA (Supplementary Fig. 3), cell cluster drives an inward flow of cancer cells and cluster deforma- together with other markers characteristic of myCAFs and iCAFs sub- tion, eventually leading to cancer cell multilayering and bud formation populations (Supplementary Fig. 4). To mimic tumor organization, we as the CAF ring closes. Bud formation can be seen as a yield-stress fabricated flat 11 kPa polyacrylamide gels coated with collagen I and we problem by considering the cluster of cancer cells as an elastoplastic generated circular clusters (150 µm radius) of cancer cells surrounded material where multilayering occurs beyond a critical cluster com- by CAFs (Fig. 1c). CAFs aligned to each other and parallel to the cancer pression. This model yields a phase diagram for bud formation, which cell boundary, encapsulating cancer cells and restraining their predicts the formation of stable buds - as observed in our experiments spreading. Surprisingly, after 8 h, CAFs assembled into a multicellular - for a limited range of ring line tension (Supplementary Theoretical ring that slid on top of the cancer cells. As the ring advanced, it Note 1, Fig. 1f, g). The model shows that the appearance of the bud can deformed clusters, ultimately inducing the multilayering of cancer be directly inferred from the integrated traction force, which plateaus cells and forming a three-dimensional bud (Fig. 1c, Supplementary prior to ring stabilization (Fig. 1f), indicating that a fraction of the ring Video 2). At this point, the ring ceased to advance, and the bud tension is resisted by bud compression (Supplementary Theoretical remained stable. CAFs also encapsulated and induced budding of N/53 Note 1, Fig. 1d). tumor organoids (Supplementary Fig. 5a–c, Supplementary Video 3). Similar to in vivo intratumoral capsules (Fig. 1a), in vitro CAF rings Notably, the same sequence of events was observed ex vivo, where were enriched in phosphorylated myosin II (Fig. 1g). This suggests that CAFs and cancer cells spontaneously exited from PDX tumor frag- ring constriction is driven by actomyosin contractility. Indeed, myosin ments (Supplementary Fig. 5d, Supplementary Video 4). This shows IIA knockout CAFs exerted significantly lower traction forces than that CAFs can form capsule-like structures to confine cancer cells even control CAFs (Supplementary Fig. 8a–c) and failed to compress and in reductionist 2D in vitro environments. The fact that this capsule deform cancer cell clusters even if forced to encircle them due to the deforms and reshapes tumor cell clusters suggests that CAFs are not experimental setup geometry (Supplementary Fig. 8d, e). Besides a just a passive barrier against cell spreading but rather an active entity decrease in contractility, myosin IIA knockout CAFs proliferate less that exerts forces on cancer cells. than control CAFs (Supplementary Fig. 8f), probably due to the role of Nature Communications | (2023)14:6966 3 Article https://doi.org/10.1038/s41467-023-42382-4 Fig. 2 | CAFs compress cancer cells using actomyosin contractility. a Images histograms of displacement angle probability relative to the cut 40 s after ablation. before (left) and after laser ablation (right) in fresh tumor slices (quantified in b). n = 74 ablations from N = 7 mice. e Montage showing images before (left) and after Cancer cells (green), stroma (CAFs) (magenta). Dashed cyan line - contour of cancer laser ablation (right) in in vitro co-cultures. Cancer cells (green), CAFs (magenta). cells. Solid cyan lines delineate ROIs. Laser cut region represented in white. White Dashed cyan line - contour of cancer cells. Solid cyan lines delineate ROIs. Laser cut vectors, tissue displacement 40 s after ablation. Scale bar, 100 µm. Scale vector, region (white box). White vectors, tissue displacement 40 s after ablation. Yellow 1 µm. b Displacement magnitude in ROIs as a function of time after ablation (left). vectors, average cumulative displacement for each ROI. Scale bar, 100 µm. Scale Vertical dashed line (t = 0) indicates the ablation time. Polar histogram of dis- vector, 5 µm. f Displacement magnitude of cancer cells and CAFs as a function of placement angle probability relative to the cut (right) 40 s after ablation. n = 31 time (left) after ablation. Vertical dashed line (t = 0) indicates the ablation time. ablations from N = 7 mice. c, Montage showing images before (left) and after laser Polar histograms of displacement angle probability relative to the cut, for CAFs and ablation (right) in tumor slices (quantified in panel d). Cancer cells (green), stroma cancer cells (right) 40 s after ablation. n = 13 ablations from N = 2 independent (CAFs) (magenta). Dashed cyan line - contour of cancer cells. Solid cyan lines experiments g Displacement magnitude of cancer cells and CAFs as a function of delineate ROIs. Laser cut region, white. White vectors, tissue displacement maps time after ablation, in control tumor slices, slices treated with blebbistatin or tumor 40 s after ablation. Yellow vectors, average displacement for each ROI (for visua- slices contacting Myosin IIA-KO CAFs (left). Polar histograms of displacement angle lization purposes, yellow vectors are not scaled). Scale bar, 100 µm. Scale vector, probability relative to the cut 40 s after ablation. n = 74 ablations from N = 7 mice 1 µm. d Displacement magnitude of cancer cells and CAFs as a function of time after (control), n = 79 ablations from N = 6 mice (Myosin IIA KO), and n = 58 ablations ablation (left). Vertical dashed line (t = 0) indicates the ablation time. Polar from N = 5 mice (Blebbistatin). Data represented as mean ± SEM for (b, d, f, g). Myosin IIA in cytokinesis12. Lower CAFs’ proliferation may decrease the induced a fast relaxation of CAF rings, almost complete disappearance compression of cancer cells due to a decreased mitotic pressure of traction forces, and the flattening of the bud (Fig. 1h, i, Supple- independently of contractility. To detangle the contribution of CAFs mentary Fig. 6d, e, Supplementary Video 6). The relaxation of com- contractility and mitotic pressure to compression, we inhibited con- pression was much faster than any expected effect of the drugs on cell tractility using either blebbistatin to inhibit myosin II (Fig. 1h) or proliferation, thus we conclude that CAF contractility is the main driver Y27632 to inhibit ROCK (Supplementary Fig. 6d). Both treatments of compressive forces. Nature Communications | (2023)14:6966 4 Article https://doi.org/10.1038/s41467-023-42382-4 a b c Beads/F-Actin/DNA 2 DNA 1 1 2 ctin/ s/F-A 1 Be a d d F-Actin/pMLC/DNA F-Actin e CAFs/trypsin f 60 CAFs 30 Pillar height (μm) 0 0 60 120 180 240 300 360 60 400Pa pMLC 30 0 0 0 120 180 240 300 360 Angle (degrees) -1000 -500 0 500 1000 Normal traction (Pa) g h Normal traction 800 Normal traction (Pa) 600 400 200 0 0 0.43 0.95 1.27 1.70 2.13 2.56 2.98 3.41 -200 Traction Disp. magnitude (μm) 600Pa eb eb Bl Bl e- Pre-Bleb Bleb Pr Fig. 3 | Compression is an intrinsic property of CAFs. a PAA pillars containing 50 µm. f Representative unwrapped pillar (from n = 9 pillars, N = 2 independent fluorescent beads (gray). Top, 3D view. Bottom, lateral view. Scale bar, 100 µm. b 3D experiments) showing CAFs occupancy (top, magenta) and 3D traction force maps rendering of CAFs stained for F-actin (phalloidin, magenta) and DNA (DAPI, cyan) (bottom). Color map represents normal tractions (compression is defined as surrounding a pillar (fluorescent beads, green). Scale bar, 100 µm. N = 2 indepen- negative, pulling is defined as positive); yellow vectors represent tangential trac- dent experiments. c Top, x-z pillar cross-section (fluorescent beads, green) sur- tions. g Representative 3D mapping of deformations (orange) and traction forces rounded by CAFs (F-actin, magenta; DNA, cyan). Dashed line represents two (black arrows) exerted by CAFs on a pillar, from n = 9 pillars, N = 2 independent selected x-y planes shown at the bottom (1 and 2). Scale bars, 100 µm. N = 2 inde- experiments). Top (left) and side (right) views. Deformations are magnified 5 times pendent experiments. d Maximum intensity projection of CAFs surrounding a pillar for visualization purposes. h CAFs normal tractions averaged across pillar height, and stained for F-actin (phalloidin, green), active myosin (pMLC, magenta) and DNA on a representative pillar before (magenta) and after (blue) blebbistatin treatment. (DAPI, blue). Inset, magnified boxed region showing staining for pMLC and F-actin. Scale vector, 600 Pa. Right dot plot, quantification of mean normal tractions for Scale bar, 100 µm. e Cross-section of a representative (from n = 9 pillars, N = 2 each pillar before and after blebbistatin. n = 9 pillars, from N = 2 independent independent experiments) pillar at its base, in the presence of CAFs (magenta) and experiments. after removal of CAFs (trypsin, cyan). Inset, magnified boxed region. Scale bar, We also tested whether normal fibroblasts could compress cancer areas nearby the ablated region, whereas almost no displacement was cells. Interestingly, while NIH/3T3 were not compressive and exerted observed in a distant control region (Fig. 2b). These data show that tractions pointing away from the cancer cell cluster, intestinal fibro- cancer cells are compressed in tumors. To address the tensional state blasts compressed cancer cells even more than CAFs (Supplementary of the CAFs, we also performed cuts perpendicular to the tumor Fig. 7a). Thus, compression is not a CAF-specific ability, suggesting that boundary (Fig. 2c; Supplementary Video 7). Again, cancer cells dis- intestinal fibroblasts (many of those are αSMA+ and referred as placed towards the cut, confirming that they were under compression, myofibroblasts13) may also compress the surrounding epithelial cells in whereas CAFs recoiled away from the cut, indicating that they were the healthy gut, as observed in other tissues during development14. under tension (Fig. 2d). Similar results were obtained using in vitro co- Indeed, both CAFs and intestinal fibroblasts exerted high tractions at cultures of cancer cells and CAFs (Fig. 2e, f; Supplementary Video 7). the single cell level, compared to NIH/3T3 (Supplementary Fig. 7b). Abrogation of contractility using blebbistatin or knocking out CAFs Overall, our in vitro force measurements and theoretical predictions myosin IIA in tumor tissue slices decreased CAF recoil and direction- suggest that CAFs intratumoral capsules not only passively confine ality upon ablation. The immediate effect of blebbistatin cannot be cancer cells but also actively compress them. explained by changes in cell proliferation, ruling out mitotic pressure To analyze the mechanical interactions between CAFs and cancer as the main driver of cancer cell compression in vivo (Fig. 2g; Sup- cells in tumors we performed two-photon laser ablations on ex vivo plementary Video 8). Furthermore, tumors containing myosin cultured thick tumor slices. First, we disconnected cancer cells from IIA-knockout CAFs exhibited a small displacement of cancer cells that the surrounding stroma by performing ablations parallel to the edge of was not directed toward the cut (Fig. 2g, Supplementary Video 8). As in cancer cells (Fig. 2a, Supplementary Video 7). Upon ablation, we this model cancer cells have WT levels of myosin IIA, this shows that observed a rapid displacement of cancer cells towards the cut in tumor cancer cell compression is a direct consequence of CAFs contractility. Nature Communications | (2023)14:6966 5 Article https://doi.org/10.1038/s41467-023-42382-4 Fig. 4 | Fibronectin scaffolds allow CAFs supracellular coordination. a CAFs vectors, 300 Pa. Right dot plot, quantification of mean normal tractions per pillar. normal tractions averaged across pillar height, on a representative pillar for control Data represented as mean ± SD. n = 16 pillars, N = 4 (control), n = 14 pillars, N = 4 (magenta) and N-cadherin depleted (cyan) CAFs. Scale vectors, 300 Pa. Right dot (SiRNA#1 Fn1), and n = 9 pillars, N = 2 (SiRNA#2 Fn1). ***p < 0.001, *p < 0.05, plot, quantification of mean normal traction per pillar. Data represented as Kruskal–Wallis ANOVA test, Dunns multiple comparisons test. d Representative (of mean ± SD. n = 9 pillars, N = 2 (control), n = 15 pillars, N = 3 (SiRNA#1 Cdh2), and data plotted in d) traction maps overlaid on a DIC image of cancer cell and CAFs n = 8 pillars, N = 2 (SiRNA#2 Cdh2). Non-significant, Kruskal-Wallis ANOVA test, control (top) or fibronectin depleted (bottom), evolving over time. Black solid line Dunns multiple comparisons test. b Top left panel, representative image of cancer represents the contour of the CAFs ring. Scale bar, 100 µm. e Upper panel, radial cells (cell tracker, green) and CAFs in vitro co-cultures (from n = 5 images, N = 2 traction peak magnitude at the boundary of the cluster for control (magenta) and independent experiments), stained for fibronectin (magenta) and DNA (DAPI, fibronectin depleted (brown) CAFs, as a function of time. Lower panel, CAFs ring blue). Right top panel, inset. Lower left panel, representative image (from n = 18 area normalized to the initial cluster size, as a function of time. Data represented as images, N = 3 mice) of cancer cells (membrane GFP, green) and CAFs in tumors from mean ± SEM. n = 12 (control) and n = 16 (Fibronectin knockdown) clusters, from N/p53/mTmG mice, stained for fibronectin (magenta) and DNA (DAPI, blue). Scale N = 3 independent experiments. ***p < 0.001, **p < 0.01, Mann Whitney two-tailed bars, 50 µm. c CAFs normal tractions averaged across pillar height, on a repre- test for time point t = 36. sentative pillar for control (magenta) and fibronectin depleted (brown) CAFs. Scale We then asked whether the generation of compressive forces is a particle image velocimetry and computed CAFs exerted forces using CAF intrinsic property or induced by cancer cells. To test this, we traction force microscopy15. This provided 3D traction force fields that microfabricated 50 µm radius soft polyacrylamide pillars (11kPa) to we decomposed into a normal component (Tn, perpendicular to the mimic the presence of cancer cells (Fig. 3a, Supplementary Fig. 9a, pillar surface) and a tangential component (Tt, parallel to the pillar Supplementary Video 9). CAFs aligned parallel to the pillar border, surface). Normal forces exhibited a larger magnitude than tangential assembled an actomyosin ring (Fig. 3b–d; Supplementary Video 9) and forces (~350 Pa vs ~200 Pa on average) and were mostly compressive deformed the pillar (Fig. 3e). We measured pillar deformation using 3D (negative) (Fig. 3f, g, Supplementary Fig. 9b, Supplementary Video 10). Nature Communications | (2023)14:6966 6 Article https://doi.org/10.1038/s41467-023-42382-4 These forces decreased dramatically after inhibition of CAF actomyo- cancer cells was even more cytoplasmic. Lack of CAFs compression, sin contractility using blebbistatin or Y27632 (Fig. 3h, Supplementary either by inhibition of contractility (blebbistatin and myosin IIA KO) Fig. 9c). Altogether, these data show that CAFs can even compress (Fig. 5a, b) or supracellular coordination (Fibronectin KD) (Fig. 5c), inert materials and thus, compressive forces emerge from CAFs triggered the nuclear translocation of YAP to levels comparable to intrinsic contractility, independently of the presence of cancer cells. confinement. This shows that contractility-driven compression pro- To effectively compress and deform cancer cells, supracellular motes YAP nuclear exit to levels that cannot be achieved by passive CAF rings should be sufficiently stable to maintain integrity over time. confinement or low substrate stiffness. What mediates connections between CAFs in the ring? Mesenchymal YAP nuclear exclusion may be driven by a specific subpopulation cells, such as CAFs, generally lack stable cell-cell junctions. Instead, of CAFs that is lost upon perturbations in contractility or fibronectin they are interconnected through N-cadherin zipper-like adhesions that expression. To exclude this possibility, we performed RNA sequencing allow cell-cell contacts while permitting fast neighbor exchange16. To of Myosin IIA KO and Fibronectin KD CAFs and found that the test if N-cadherin mediates connections between CAFs in rings, we expression of CAF subtype markers was largely unaffected (Supple- depleted N-cadherin using siRNAs (Supplementary Fig. 10a). Surpris- mentary Fig. 12). ingly, we found that N-cadherin-depleted CAFs were still able to To further understand the mechanism triggering YAP localization, compress pillars (Fig. 4a) and exerted the same levels of forces as we quantified cell density and nuclear area across all conditions. We control cells (Supplementary Fig. 10b). used the CAF radial traction peak (in vitro) and the CAF displacement CAFs produce excessive amounts of ECM proteins, especially upon ablation (in vivo) as readouts of compression. Both in vitro and fibronectin9,17–19. Fibroblasts can use fibronectin to connect to each in vivo, compression correlated with an increase in cancer cell density other using specialized cell-cell contacts, named stitch adhesions20. (Fig. 5d, e, Supplementary Fig. 13b), a decrease in cancer cell nuclear This led us to hypothesize an alternative mechanism for CAF supra- area (Fig. 5d, e) and the nuclear exclusion of YAP (Fig. 5d, e). Impor- cellular coordination indirectly via ECM. We observed that CAFs tantly, CAF density did not correlate with YAP localization, again ruling deposited abundant amounts of fibronectin both in vitro and in vivo out CAFs mitotic pressure as the main source of compression (Sup- (Fig. 4b). Consistent with the fact that CAFs contractility is required for plementary Fig. 13a). Overall, these data suggest that CAF compression fibronectin fibrillogenesis17, myosin IIA-knockout CAFs produced triggers cancer cell packing, decreasing nuclear tension and inducing lower amounts of fibronectin (Supplementary Fig. 10c, d). The fibro- YAP nuclear export. This hypothesis aligns with previous findings nectin network was isotropic and disorganized in the bulk of CAFs but indicating that pulling forces exerted on the nucleus can open nuclear became aligned at the boundary with cancer cells forming supracel- pores and allow YAP to enter the nucleus22,25,26. lular cables that spanned several CAFs (Fig. 4b). Similar reorganization We next investigated the functional implications of CAFs com- of the fibronectin network was previously seen in fibroblasts growing pressing cancer cells. We observed that the net number of cancer cells in macroscopically engineered clefts21. This led us to hypothesize that increased at a much slower rate in compressed cancer cell clusters fibronectin could act as a mechanical scaffold for CAF supracellular compared to freely growing ones (Fig. 6a). In turn, cancer cell density coordination and force generation during compression. Indeed, the increased upon compression, while it decreased during cancer cell- depletion of fibronectin in CAFs significantly reduced their ability to free growth (Fig. 6a). These findings suggest CAF compression affects compress pillars (Fig. 4c). Similarly, in the absence of fibronectin, CAFs cancer cell growth, either by reducing their proliferation or by could not stabilize multicellular rings, which led to a reduction in the increasing apoptosis. In vitro, cancer cell proliferation decreased upon radial traction peak when co-cultured with cancer cells (Fig. 4d, e, CAFs compression and recovered in the presence of non-contractile Supplementary Fig. 10f, Supplementary Video 11). Importantly, this myosinIIA-knockout CAFs based on Edu pulses and Ki67 and Phospho- loss of tractions is not due to impaired force generation because histone H3 (PHH3) staining. In contrast, confinement only decreased control and fibronectin-depleted CAFs exerted similar tractions when the number of mitotic cells (PHH3) (Fig. 6b, Supplementary Data cultured as single cells (Supplementary Fig. 10e). Furthermore, the Figs. 14a, 15a). Importantly, conditioned media either from WT or proliferation of fibronectin-depleted CAFs was unaffected (Supple- myosin IIA-knockout CAFs did not affect cancer cell growth (Supple- mentary Fig. 10g). This shows that fibronectin is dispensable for force mentary Fig. 15b), excluding indirect effects on cancer cell prolifera- generation but required for intercellular force transmission. A direct tion through CAFs nutrient consumption27. In contrast, cancer cell consequence of this mechanical hindrance was the inability of CAFs to apoptosis was not affected by compression (Fig. 6d). We thus conclude compress cancer cells (Fig. 4d, e). Altogether, these data show that that CAFs compression restricts cancer cell proliferation, probably fibronectin cables support force transmission between CAFs, allowing through YAP mechanosensing28, although other mechanisms may also the mechanical coordination required to compress cancer cells. be involved. In fact, cancer cell proliferation and apoptosis were not To address if cancer cells sense and functionally respond to CAF affected in tumors containing myosin IIA-knockout CAFs, despite compression, we analyzed the cellular localization of a well- changes in YAP localization (Fig. 6c, e, Supplementary Figs. 14b, 16). established mechanosensor, the transcriptional co-activator YAP, The discrepancy between in vitro and in vivo results could be due to which shuttles in and out of the nucleus depending on the mechan- additional factors regulating cancer cell proliferation in tumors that ical stress to which cells are subjected22,23. First, we characterized the are absent in our simplified in vitro model, such as cancer cell access to mechanosensing ability of the cancer cells by seeding them on soft nutrients or stroma-secreted factors. (0.2 kPa) and stiff (11 kPa) substrates. As observed in other cell Overall, we found that CAFs assemble intratumoral capsules sta- types24, YAP nuclear to cytoplasmic ratio was increased on stiff sub- bilized by fibronectin scaffolds, allowing coordinated supracellular strates, and this effect was abolished upon inhibition of contractility contraction. These contractile capsules actively compress cancer cells, using blebbistatin (Supplementary Fig. 11). To measure YAP locali- triggering mechanical signaling and inducing reorganization of the zation at the collective level in cancer cell clusters, we quantified the tumor architecture. Thus, in contrast to the generally accepted con- 3D correlation between DAPI and YAP signals (Fig. 5a). A positive cept, the tumor capsule is not just a passive barrier that prevents correlation indicates that YAP is preferentially in the nucleus, tumor expansion, and our data rather points towards capsules as whereas a negative correlation reflects YAP cytoplasmic localization. active determinants of tumor solid stress. Although these results need We found that, in the absence of CAFs, YAP was mostly nuclear in to be validated on more complex in vitro 3D environments, our find- cancer cells growing without any constraint (Fig. 5a). Spatial con- ings could have diverse effects on tumor progression29. It could slow finement of cancer cells on micropatterns lead to an increase in down tumor growth but, at the same time, promote cancer cell cytoplasmic YAP. Upon compression by CAFs, YAP localization in stemness30, drug-resistance31, collapse of blood vessels4 and metastatic Nature Communications | (2023)14:6966 7 Article https://doi.org/10.1038/s41467-023-42382-4 traits by increasing DNA damage32, as well as tumor budding, a poor Methods prognosis factor for colorectal cancer patients33–36. If those processes Ethical statement occur simultaneously or at different stages of tumor progression Animal care and use for this study were performed in accordance with remains to be understood, and new mouse models coupled with the recommendations of the European Community (2010/63/UE) for advanced real-time in vivo imaging techniques would help to unam- the care and use of laboratory animals. Experimental procedures were biguously ascertain the effects of CAF compression on tumor specifically approved by the ethics committee of the Institut Curie progression. CEEA-IC #118 (Authorization NUMBER −25603-2020053122444776- Nature Communications | (2023)14:6966 8 Article https://doi.org/10.1038/s41467-023-42382-4 Fig. 5 | CAFs compression of cancer cells triggers YAP nuclear exit. a Patterned nuclear correlation in cancer cells in vivo. Data represented as mean ± SD, each dot cancer cells (Cell tracker, Green) cultured on a collagen-coated polyacrylamide gel is the average of 3–6 image quantifications per tumor of n = 7 control and n = 6 growing freely, under confinement (by a PDMS stencil), surrounded by control or myosin IIA KO mice, from N = 2 independent experiments. **p < 0.01, Mann Whitney Myosin IIA KO CAFs (F-Actin, phalloidin, magenta), or treated with blebbistatin two-tailed test. c Cancer cells (Cell tracker, Green) surrounded by CAFs treated with (quantified in right dot plot). Cells are stained for YAP. Right upper panel: YAP SiRNA Control and SiRNA fibronectin. Scale bar, 100 µm. Right dot plot: mean YAP nuclear correlation ranging from 0.5 (red, high correlation) to −0.5 (blue, anti- nuclear correlation in cancer cells. Data represented as mean ± SD, n = 12 (SiRNA correlation). Lower panel: schematic representation of how correlations are per- Control), n = 13 (SiRNA Fibronectin) clusters per condition, from N = 3 independent formed. White line outlines the boundary of the cancer cell cluster. Scale bar, experiments. ***p < 0.001, Mann Whitney two-tailed test. d Left graph: Average 100 µm. Right dot plot: mean YAP nuclear correlation of cancer cells in vitro. Data radial traction peaks plotted against YAP nuclear correlation for all conditions. represented as mean ± SD. Free growth (n = 20 clusters, N = 5), confinement (n = 19 Right graph: Cancer cell nuclear area plotted against YAP nuclear correlation for all clusters, N = 5), CAFs control (n = 42 clusters, N = 10), CAFs αSMA:Myosin IIA−/− conditions. Correlations were tested by linear regression (p < 0.05). Data repre- (n = 17 clusters, N = 4), Blebbistatin (n = 12 clusters, N = 3). **p < 0.01, ***p < 0.001, sented as mean ± SD. n and N values correspond to the ones in (a, c). e Left graph: Kruskal–Wallis ANOVA test, Dunns multiple comparisons test. b Tumors from CAF normal displacement upon ablation (respect to the ablation) plotted against control or αSMA:Myosin IIA KO mice showing cancer cells (nuclear-GFP, green), YAP nuclear correlation in control and αSMA:MyosinIIA−/− tumors. Right graph: CAFs expressing myosin IIA (membrane-tdTomato, magenta) and myosin IIA KO Cancer cell nuclear area plotted against YAP nuclear correlation in control and CAFs (membrane-GFP, green), stained for YAP (yellow) and DNA (DAPI, blue). αSMA:MyosinIIA−/− tumors. Data represented as mean ± SD. n and N values corre- Bottom panel, YAP nuclear correlation. Scale bar, 100 µm. Right dot plot: mean YAP spond to the ones in (b). Fig. 6 | CAFs compression restrains cancer cell growth. a Evolution of cancer cell αSMA:MyosinIIA-/- tumors. Data are presented as mean ± SD, n = 6 (Control), 4 clusters that freely grow (left) or grow compressed by CAFs (right). Cell nuclei are (αSMA:MyosinIIA−/−), from N = 1 experiment. Not significative, Mann Whitney two- labeled with Hoestch to allow live imaging. White line outlines the boundary of the tailed test. d Immunostaining of apoptotic cells (labeled by cleaved-caspase 3) in cancer cell cluster. Graphs: Quantification of the evolution of cancer cell number cancer cell clusters that are freely growing, confined or surrounded by control or (left) and density (right). n = 6 clusters per condition from N = 2 independent MyosinIIA KO CAFs. Graph: quantification of CC3+ area normalized to the total experiments. b Immunostaining of cells in M-phase (labeled by Phospho-histone number of cancer cells per cluster. Data are presented as mean ± SD, n = 9 (free H3) in cancer cell clusters that are freely growing, confined or surrounded by growth), 6 (confined), 11 (CAFs control), and 8 (CAFs Myosin IIA KO), from N = 2 control or MyosinIIA KO CAFs. Graph: quantification of the percentage of PHH3+ independent experiments. e Quantification CC3+ area in control and cells in each condition. Data are presented as mean ± SD, n = 18 (free growth), 15 αSMA:MyosinIIA−/− tumors. Data are presented as mean ± SD, n = 5 (Control), 4 (confined), 17 (CAFs control), and 17 (CAFs Myosin IIA KO), from N = 3 independent (αSMA:MyosinIIA−/−), from N = 1 experiment. Representative examples of immu- experiments. *p < 0.05, **p < 0.01, Kruskal–Wallis ANOVA test, Dunns multiple nostainings quantified in c and e can be found in Supplementary Data Fig. 15. Scale comparisons test. c Quantification of percentage of PHH3+ cells in control and bars = 100 µm. Nature Communications | (2023)14:6966 9 Article https://doi.org/10.1038/s41467-023-42382-4 given by National Authority) in compliance with the international described above. Prior to injection in mice, tumoroids from 12 24-well guidelines. Matrigel drops were harvested and mechanically disaggregated using a pipette tip, centrifuged and resuspended in 100 µl of 1:1 Matrigel/cul- Mouse models ture medium. This solution was injected into the interscapular fat pad All mice were kept under Specific Pathogen Free (SPF) conditions for of bone marrow transplanted mice. For this, under Isoflurane gas breeding. Double fluorescent pVillin-CreERT2, LSL-NICD-nGFP; mT/mG; anesthesia, a small skin incision was performed at the level of the p53fl/fl mice were generated as previously described10. Four-weeks old interscapular region, the fat pad was exteriorized and tumoroids were male and female mice were injected intraperitoneally with tamoxifen injected directly into it using a 25 G needle. Mice were then sutured (50 mg/kg) for 5 consecutive days for induction of Cre recombinase using wound clips (7.5*1.5 mm, Phymep), and placed into standard activity. Approximately 8 months after tamoxifen injection, mice housing conditions during the time of tumor development. 3 weeks spontaneously develop invasive intestinal carcinomas37. All cells from post engraftment mice were injected daily with tamoxifen (50 mg/kg) these mice express a membrane-targeted tdTomato, while tumor cells for two consecutive days, to induce Cre-mediated gene recombina- express a nuclear-targeted GFP. tion. Tamoxifen was re-administered either every week or every other Double fluorescent SMA-CreERT2; R26mT/mG; myosin IIAfl/fl were week throughout the entire duration of the experiment to ensure generated by crossing mice containing the MyoIIA heavy-chain (Myh9) efficient knockout of possibly newly generated CAFs. For tumor floxed38 with mice expressing Cre recombinase under the control of compartmentalization analyses we used mice injected every week with the SMA promoter39, and with Rosa26-mTmG40 mice. Rosa26-mTmG tamoxifen. For laser ablation experiments, all mice were analyzed, as mice express a membrane-targeted tandem dimer Tomato (mT) prior ablations were performed specifically in areas with high content of to Cre-mediated excision and a membrane-targeted GFP (mG) fol- knockout CAFs (mGFP+) To account for possible secondary effect of lowing excision. Membrane targeting was achieved using the MARCKS tamoxifen in tumor development, all mice (Cre+/− and Cre−/− controls) membrane tag. were injected with the same doses of tamoxifen. 9 weeks after tumoroid injection mice were sacrificed, and tumors were excised and Generation and culture of tumor organoids prepared for downstream analysis. Tumors from pVillin-CreERT2, LSL-NICD-nGFP; mT/mG; p53fl/fl male and female mice were excised and dissociated using a scalpel in medium Immunostaining of tumor tissue sections containing DMEM-F12 (ThermoFisher Scientific) supplemented with Tissue was fixed in 4% paraformaldehyde (Electron Microscopy Sci- 2,5% (v/v) GlutaMAX (Gibco), 2% (v/v) Antibiotic-Antimycotic (Gibco) ences)/PBS (v/v) for 20 min at RT, and then dehydrated first in 15% and 300 units/ml of Collagenase III (StemCell). Tissue pieces were sucrose (Sigma-Aldrich)/PBS (w/v) solution for 1 h and then in 30% incubated for 2 h at 37 °C under agitation (180 rpm), and then filtered sucrose/PBS (w/v) solution for 2 h at RT. After, tissue was embedded first through 100 µm and 40 µm filters. Dissociated cells were cen- with OCT compound (Sakura) in plastic gaskets (Euromedex), frozen trifuged and resuspended in a 100 µl drop containing a mix of 50% at −20 °C and cut on the cryostat using SuperFrost Plus™ Adhesion Matrigel (Corning)− 50% (v/v) tumoroid medium (DMEM-F12, supple- slides (VWR, Menzel Gläser). Tissue sections were then permeabilized mented with 2,5% (v/v) GlutaMAX, 2% (v/v) antibiotic-antimycotic, with 0.2% Triton x100 (Sigma-Aldrich)/PBS (v/v) for 1 h at RT, blocked 100 ng/mL Noggin (Peprotech), 50 ng/mL EGF (Peprotech), 10 ng/mL with 3% BSA (w/v) (IgG-Free, Protease-Free, Jackson Immuno Research) (Peprotech), 1% (v/v) B27 supplement (ThermoFisher Scientific) and 1% in 0.05% Triton x100/PBS (v/v) solution for 1 h at RT and stained with (v/v) N-2 supplement (ThermoFisher Scientific)). Matrigel drops were primary antibodies overnight in humidified chambers at RT. Sections allowed to polymerize for 40 min at 37 °C and 5% CO2 and were then were then washed 3 times with 0.05% Tx100/PBS (v/v) solution for 1 h, covered with 2 mL of tumoroid medium. Once formed, tumoroids incubated with secondary antibodies, DAPI and phalloidin, depending were split once a week. on the staining, for 4 h at RT, washed 3 times in 0.05% Tx100/PBS (v/v) solution for 1 h and mounted using AquaPolyMount (Polysciences). Tumor establishment in αSMA MyoIIA KO mice Antibodies references and dilutions are listed in Table 1. Bone marrow transplantation experiments were performed in order to render the immune system of αSMA MyoIIA KO mice compatible with Whole-mount staining of tumor tissue tumor growth. Fixed tissue was sliced to 250 µm thick slices in a vibratome (Leica VT1200S) as described before10,41. Staining was performed as Irradiation. 6–8 weeks old recipient male and female mice (SMA- described above with minor modifications: permeabilization was CreERT2; R26mT/mG; myosin IIAfl/fl) were placed in an acrylic container in done with 1% Triton X-100/PBS (v/v) for 1 h at RT (500 µl per tube), all continuous airflow between two opposite X-ray sources (CIXD Dual antibody solutions and washing steps were performed using 0.2% Irradiator, Xstrahl). Cre−/− mice were used as controls. Mice were Triton X-100/PBS (v/v), under mild shaking conditions (1 mL per exposed to a Fractioned Total Body Irradiation (FTBI), at a rate of tube) and concentration of antibodies was increased (see Table 1, in 1,18 Gy/min for a total dose of 10 Gy, fractionated in two 5 Gy doses 150 µl per tube). Incubations with antibodies were done without with a 4 h interval. agitation. Bone marrow transplantation. Donor mice (non-induced pVillin- Confocal imaging CreERT2, LSL-NICD-nGFP; mT/mG; p53fl/fl, 4-10 months old, male and Stained cryosections and whole-mount tissue, as well as stained in vitro female) were sacrificed and bone marrow cells (BMCs) were collected cell cultures were imaged using an inverted confocal microscope from the tibias, femurs and humerus, and resuspended in 300 µl of PBS (Zeiss LSM880NLO) using laser lines 405, 488, 561 and 633 nm and the containing 2% (v/v) FBS. Immediately after irradiation, sex-matched following objectives: 25×/0.8 OIL, W, Gly LD LCI PL APO (UV) VIS-IR, recipient mice were injected retro-orbitally with 100 µl of the BMCs 40×/1.30 OIL DICII PL APO (UV) VIS-IR and 63×/1.4 OIL DICII PL APO. suspension, containing approximately 2 × 106 BMCs, using a 27-gauge Images were processed using ImageJ. needle. Mice were left for 8 weeks to allow efficient grafting of BMCs Second harmonic imaging of non-stained whole-mount tissue and reconstitution of the immune system. sections was performed in an inverted AOBS two-photon laser- scanning confocal microscope (Leica SP8), coupled with a femtose- Tumor establishment and MyosinIIA KO induction. Tumoroids from cond laser (Chameleon Vision II, Coherent Inc.) using a 40×/1.10 HC pVillin-CreERT2, LSL-NICD-nGFP; mT/mG; p53fl/fl mice were cultured as PL APO CS2 water immersion objective. The excitation wavelength Nature Communications | (2023)14:6966 10  Article Table 1 | Antibodies used in the study Antigen Antibody/Chemical Dilution Reference Vendor FACS Cryosections Whole-mount Cultured cells Western blot Fibronectin Anti-Fibronectin rabbit polyclonal antibody – 1–100 1–100 1–200 1–10,000 F3648 Sigma Aldrich N-Cadherin Anti-N-Cadherin mouse monoclonal antibody – – – – 1–500 33-3900 Thermo Fisher Scientific MyosinIIA Anti-MyosinIIA rabbit polyclonal antibody – – 1–100 1–100 1–100 909801 Biolegend GAPDH Anti-GAPDH rabbit polyclonal antibody – – – – 1–5000 G9545 Sigma Aldrich αSMA Anti-αSMA mouse monoclonal antibody 1–200 – – 1–100 – A2547 Sigma Aldrich pMLC Anti-pMLC rabbit polyclonal antibody – – 1–100 1–200 – 3674 Cell Signaling Nature Communications | (2023)14:6966 YAP Anti-YAP rabbit monoclonal antibody – 1–200 - 1–200 – 14074 Cell Signaling CC3 Anti- Cleaved Caspase 3 rabbit monoclonal antibody – - 1–100 1–200 – 9661 Cell Signaling PHH3 Anti-Phospho Histone H3 rabbit monoclonal antibody – 1– 200 – 1–200 – H0412 Sigma Aldrich Ki67 Anti-Ki67 rabbit monoclonal antibody – 1–200 – 1–400 – 9129S Cell Signaling EpCAM PE/Cy7 anti-human CD326, mouse IgG2b, κ 1–10 – – – – 324221 Biolegend EpCAM APC anti-mouse CD326, mouse, rat IgG1 1–10 – – – – 130-102-969 Miltenyi Biotec CD45 Brilliant-Violet 421 anti-human CD45, mouse IgG1, κ 1–20 – – – – 304032 Biolegend CD45 Brilliant-Violet 421 anti-mouse CD45, rat IgG2b, κ 1–20 – – – – 103133 Biolegend CD31 FITC anti-human CD31, mouse IgG1, κ 1–20 – – – – 303104 Biolegend CD31 FITC anti-human CD31, rat IgG2a, κ 1–20 – – – – 102405 Biolegend Vimentin APC anti-human Vimentin, recombinant IgG1 1–250 – – – – 130-106-370 Miltenyi Biotec Isotype controls FITC mouse IgG1 1–10 – – – – 130-092-213 Miltenyi Biotec Brilliant-Violet 421 rat IgG2b, κ 1–20 – – – – 400639 Biolegend FITC rat IgG2a, κ 1–20 – – – – 400505 Biolegend Anti-mouse Alexa Fluor 488 Goat anti-mouse IgG Alexa Fluor 488 polyclonal antibody – 1–200 1–200 1–400 – A11029 Thermo Fisher Scientific Anti-mouse Alexa Fluor 546 Goat anti-mouse IgG Alexa Fluor 546 polyclonal antibody – 1–200 1–200 1–400 – A11030 Thermo Fisher Scientific Anti-Rabbit Alexa Fluor 488 Goat anti-Rabbit IgG Alexa Fluor 488 polyclonal antibody – 1–200 1–200 1–400 – A32731 Thermo Fisher Scientific Anti-Rabbit Alexa Fluor 568 Goat anti-Rabbit IgG Alexa Fluor 568 polyclonal antibody – 1–200 1–200 1–400 – A11011 Thermo Fisher Scientific Anti-rabbit HRP Goat anti-Rabbit polyclonal antibody 32260 Thermo Fisher Scientific Against Fibronectin – – – – 1–10,000 Against MyosinIIA – – – – 1–5000 Against GAPDH – – – – 1– 2500 Anti-mouse HRP Goat anti-Mouse polyclonal antibody 32230 Thermo Fisher Scientific Against N-Cadherin – – – – 1–500 F-actin Phalloidin-Rhodamine – 1–200 1–100 1–200 – R415 Thermo Fisher Scientific Phalloidin-Alexa Fluor 488 – 1–200 1–100 1–200 – A12379 Thermo Fisher Scientific Phalloidin-Alexa Fluor 633 – – – 1–50 A22284 Thermo Fisher Scientific DNA DAPI (4′,6-Diamidino-2-Phenylindole, Dihydrochloride) – 1–400 1–400 1–400 – D1306 Thermo Fisher Scientific https://doi.org/10.1038/s41467-023-42382-4 11 Article https://doi.org/10.1038/s41467-023-42382-4 was set at 947 nm and signals were acquired using 3 non-descanned since CAFs/stroma express mTomato. For myosin-IIA KO tumors HyD detectors: 525/40 nm (for GFP), 585/40 nm (for tdTomato) and where the stroma has a heterogeneous labeling, threshold-based 1,−1 were considered as differentially expressed. Volcano stamp in an inverted phase contrast microscope. and bubble plots were built using SRplot. Gene Ontology enrichment Once coated with fibronectin, SU8 masters were used to micro- analyses were performed using GeneCodis446. fabricate PAA pillars. For this, 33 mm bottom-glass dishes (World Precision Instruments) were treated as before (Silane + Glutar- Data availability aldehyde). PAA gels of 11 kPa (Young modulus) were produced using a All data are available upon request. RNAseq datasets can be accessed solution of 7.5% (v/v) acrylamide, 0.1% (v/v) bis-acrylamide, 0.5% (w/v) through NCBI GEO, accession number: GSE242846. Source data are ammonium persulphate, 0.05% (w/v) tertamethylethylenediamine, provided with this paper. 0.6% (w/v) Acrylic acic N-hydroxysuccinimide ester (stock 10 mg/mL in DMSO) and 2% (v/v) of 200-nm-diameter red fluorescent carboxylate- References modified beads, in PBS. A 25 µl drop of this solution was added to the 1. Joyce, J. A. & Pollard, J. W. Microenvironmental regulation of treated glass and the SU8 master was placed on. PAA gels were allowed metastasis. Nat. Rev. Cancer 9, 239–252 (2009). to polymerize for 1 h at RT. After, 3 ml of PBS were added and incu- 2. Nia, H. T., Munn, L. L. & Jain, R. K. Physical traits of cancer. Science bated for 30 min to facilitate lifting-off of the SU8 master, which was 370. https://doi.org/10.1126/science.aaz0868 (2020). achieved with the help of a scalpel. Gels were then washed once with 3. Nia, H. T. et al. Solid stress and elastic energy as measures of PBS and incubated with a solution of 100 µg/mL of fibronectin for 1 h at tumour mechanopathology. Nat. Biomed. Eng. 1. https://doi.org/10. RT to improve cell attachment. Gels were then washed twice with PBS 1038/s41551-016-0004 (2016). and used immediately or stored at 4 °C for a maximum of 24 h. 4. Stylianopoulos, T. et al. Causes, consequences, and remedies for growth-induced solid stress in murine and human tumors. Proc. Pillar compression assay Natl Acad. Sci. USA 109, 15101–15108 (2012). To assess CAF pillar compression, a 100 µl drop containing ~200.000 5. Sahai, E. et al. A framework for advancing our understanding of pre-stained CAFs (CellTracker Green, ThermoFisher Scientific) was cancer-associated fibroblasts. Nat. Rev. Cancer 20, 174–186 (2020). added on top of each pillar-containing PAA gel and incubated for 30- 6. Kalluri, R. The biology and function of fibroblasts in cancer. Nat. 60 min at 37 °C, 5% CO2 to allow attachment of cells. Non-attached cells Rev. Cancer 16, 582–598 (2016). Nature Communications | (2023)14:6966 15 Article https://doi.org/10.1038/s41467-023-42382-4 7. Attieh, Y. & Vignjevic, D. M. The hallmarks of CAFs in cancer inva- 31. Rizzuti, I. F. et al. Mechanical control of cell proliferation increases sion. Eur. J. Cell Biol. 95, 493–502 (2016). resistance to chemotherapeutic agents. Phys. Rev. Lett. 125, 8. Barbazan, J. & Matic Vignjevic, D. Cancer associated fibroblasts: is 128103 (2020). the force the path to the dark side? Curr. Opin. Cell Biol. 56, 32. Nader, G. P. F. et al. Compromised nuclear envelope integrity drives 71–79 (2019). TREX1-dependent DNA damage and tumor cell invasion. Cell 184, 9. Glentis, A. et al. Cancer-associated fibroblasts induce 5230–5246 e5222 (2021). metalloprotease-independent cancer cell invasion of the basement 33. De Smedt, L., Palmans, S. & Sagaert, X. Tumour budding in color- membrane. Nat. Commun. 8, 13 (2017). ectal cancer: what do we know and what can we do? Virchows Arch. 10. Staneva, R. et al. Cancer cells in the tumor core exhibit spatially 468, 397–408 (2016). coordinated migration patterns. J. Cell Sci. 132. https://doi.org/10. 34. Ohike, N. et al. Tumor budding as a strong prognostic indicator in 1242/jcs.220277 (2019). invasive ampullary adenocarcinomas. Am. J. Surg. Pathol. 34, 11. Vedula, S. R. et al. Mechanics of epithelial closure over non- 1417–1424 (2010). adherent environments. Nat. Commun. 6, 6111 (2015). 35. Prall, F. Tumour budding in colorectal carcinoma. Histopathology 12. Andrade, V. & Echard, A. Mechanics and regulation of cytokinetic 50, 151–162 (2007). abscission. Front. Cell Dev. Biol. 10, 1046617 (2022). 36. van Wyk, H. C. et al. The relationship between tumour budding, the 13. Choo, J., Glisovic, N. & Matic Vignjevic, D. Gut homeostasis at a tumour microenvironment and survival in patients with primary glance. J. Cell Sci. 135. https://doi.org/10.1242/jcs.260248 (2022). operable colorectal cancer. Br. J. Cancer 115, 156–163 (2016). 14. Goodwin, K. et al. Smooth muscle differentiation shapes domain 37. Chanrion, M. et al. Concomitant Notch activation and p53 deletion branches during mouse lung development. Development 146. trigger epithelial-to-mesenchymal transition and metastasis in https://doi.org/10.1242/dev.181172 (2019). mouse gut. Nat. Commun. 5. https://doi.org/10.1038/ 15. Latorre, E. et al. Active superelasticity in three-dimensional epithelia ncomms6005 (2014). of controlled shape. Nature 563, 203–208 (2018). 38. Jacobelli, J. et al. Confinement-optimized three-dimensional T cell 16. Shih, W. & Yamada, S. N-cadherin as a key regulator of collective amoeboid motility is modulated via myosin IIA-regulated adhe- cell migration in a 3D environment. Cell Adh. Migr. 6, 513–517 (2012). sions. Nat. Immunol. 11, 953–961 (2010). 17. Attieh, Y. et al. Cancer-associated fibroblasts lead tumor invasion 39. Wendling, O., Bornert, J. M., Chambon, P. & Metzger, D. Efficient through integrin-beta 3-dependent fibronectin assembly. J. Cell temporally-controlled targeted mutagenesis in smooth muscle Biol. 216, 3509–3520 (2017). cells of the adult mouse. Genesis 47, 14–18 (2009). 18. Erdogan, B. et al. Cancer-associated fibroblasts promote directional 40. Muzumdar, M. D., Tasic, B., Miyamichi, K., Li, L. & Luo, L. A global cancer cell migration by aligning fibronectin. J. Cell Biol. 216, double-fluorescent Cre reporter mouse. Genesis 45, 3799–3816 (2017). 593–605 (2007). 19. Gopal, S. et al. Fibronectin-guided migration of carcinoma collec- 41. Staneva, R., Barbazan, J., Simon, A., Vignjevic, D. M. & Krndija, D. tives. Nat. Commun. 8, 14105 (2017). Cell migration in tissues: explant culture and live imaging. Methods 20. Pankov, R., Momchilova, A., Stefanova, N. & Yamada, K. M. Char- Mol. Biol. 1749, 163–173 (2018). acterization of stitch adhesions: Fibronectin-containing cell-cell 42. Calvo, F., Hooper, S. & Sahai, E. Isolation and immortalization of contacts formed by fibroblasts. Exp. Cell Res. 384, 111616 (2019). fibroblasts from different tumoral stages. Bio Protoc. 4, 21. Kollmannsberger, P., Bidan, C. M., Dunlop, J. W. C., Fratzl, P. & e1097 (2014). Vogel, V. Tensile forces drive a reversible fibroblast-to- 43. Zhao, J. J. et al. Human mammary epithelial cell transformation myofibroblast transition during tissue growth in engineered clefts. through the activation of phosphatidylinositol 3-kinase. Cancer Cell Sci. Adv. 4, eaao4881 (2018). 3, 483–495 (2003). 22. Elosegui-Artola, A. et al. Force triggers YAP nuclear entry by reg- 44. Perez-Gonzalez, C. et al. Active wetting of epithelial tissues. Nat. ulating transport across nuclear pores. Cell 171, 1397–1410 Phys 15, 79–88 (2019). e1314 (2017). 45. Brugues, A. et al. Forces driving epithelial wound healing. Nat. Phys 23. Panciera, T., Azzolin, L., Cordenonsi, M. & Piccolo, S. Mechan- 10, 683–690 (2014). obiology of YAP and TAZ in physiology and disease. Nat. Rev. Mol. 46. Garcia-Moreno, A. et al. Functional enrichment analysis of reg- Cell Biol. 18, 758–770 (2017). ulatory elements. Biomedicines 10. https://doi.org/10.3390/ 24. Dupont, S. et al. Role of YAP/TAZ in mechanotransduction. Nature biomedicines10030590 (2022). 474, 179–183 (2011). 25. Andreu, I. et al. Mechanical force application to the nucleus reg- Acknowledgements ulates nucleocytoplasmic transport. Nat. Cell Biol. 24, We thank all members of the DMV lab and G. Montagnac for helpful 896–905 (2022). discussions and Raimon Sunyer for discussions on the microfabrication 26. Dowbaj, A. M. et al. An optogenetic method for interrogating YAP1 of PAA pillars. We thank Jooske Monster for kindly sharing the Python and TAZ nuclear-cytoplasmic shuttling. J. Cell Sci. 134. https://doi. script to segment nuclei using CellPose. We acknowledge V. Fraisier and org/10.1242/jcs.253484 (2021). O. Renaud from the Cell and Tissue Imaging facility (PICT-IBiSA) and 27. Park, J. S. et al. Mechanical regulation of glycolysis via cytoskeleton Animal Facility, Institut Curie. This work was funded by the European architecture. Nature 578, 621–626 (2020). Union’s Horizon 2020 research and innovation program: European 28. Zanconato, F., Cordenonsi, M. & Piccolo, S. YAP and TAZ: a signal- Research Council (ERC) under the grant agreement CoG 772487 ling hub of the tumour microenvironment. Nat. Rev. Cancer 19, (D.M.V.), AdvG 883739 (X.T.) and the Marie Skłodowska-Curie grant 454–464 (2019). agreement No 797621 (M.G.G.) and 659776 (J.B.), the Spanish Ministry 29. Fernandez-Sanchez, M. E. et al. Mechanical induction of the for Science and Innovation, Juan de la Cierva IJC2018-036875-I (J.B.), the tumorigenic beta-catenin pathway by tumour growth pressure. Inserm ITMO Cancer N°20CR110-00 grant (D.M.V., P.S.), the Fondation Nature 523, 92–95 (2015). ARC pour la Recherche sur le Cancer (C.P.G.), the GEFLUC, Les entre- 30. Cheung, P. et al. Regenerative reprogramming of the intestinal prises contre le cancer (C.P.G.), the Agence Nationale de la Recherche stem cell state via hippo signaling suppresses metastatic colorectal ANR-10-IDEX-0001-02 PSL, ANR-10-EQPX-34, ANR-10-LABX-31 (S.D., cancer. Cell Stem Cell 27, 590–604 e599 (2020). M.S.), ANR-11-LABX-0038 (C.P.G., J.B., D.M.V.), the Spanish Ministry for Nature Communications | (2023)14:6966 16 Article https://doi.org/10.1038/s41467-023-42382-4 Science, Innovation and Universities MICCINN/FEDER, PGC2018- Peer review information Nature Communications thanks the anon- 099645-B-I00 (X.P.), the Generalitat de Catalunya, Agaur, SGR-2017- ymous reviewer(s) for their contribution to the peer review of this work. 01602 (X.P.), the CERCA Programme (X.P.), the Obra Social “La Caixa”, ID 100010434 and LCF/PR/HR20/52400004 (X.T.), the Fundació la Marató Reprints and permissions information is available at de TV3, project 201903-30-31-32 (X.T.) and the Severo Ochoa Award of http://www.nature.com/reprints Excellence from the MINECO (X.T., IBEC). Publisher’s note Springer Nature remains neutral with regard to jur- isdictional claims in published maps and institutional affiliations. Author contributions Conceptualization (J.B., C.P.G., D.M.V.), Methodology (J.B., C.P.G., Open Access This article is licensed under a Creative Commons M.G.G., S.R., E.L., M.S., S.D., P.M.), Data analysis (J.B., C.P.G., M.G.G., E.L., Attribution 4.0 International License, which permits use, sharing, M.D., P.S.), Funding acquisition (X.T., D.M.V.), Writing – original draft (J.B., adaptation, distribution and reproduction in any medium or format, as C.P.G., X.T., D.M.V.), Writing – review & editing: all authors. long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if Competing interests chang