Mechanisms of active wetting and fluidification in epithelial cell collectives

MainAlterations in the mechanical properties of single cells and tissues are increasingly recognized as hallmarks of disease, particularly cancer. Mechanical features such as tension, adhesion, elasticity and viscosity are governed by actomyosin cytoskeletal dynamics and transmitted across tissues through cell–cell and cell–extracellular matrix (ECM) adhesions1,2,3.At the tissue scale, physical phase transitions—most notably, solid-to-fluid-like behaviours and active wetting—are emerging as key regulators of morphogenesis and disease progression4,5,6,7,8,9,10. Epithelial tissues often adopt a solid-like, jammed state at high density due to coordinated adhesion, cortical tension and constrained cell motility10. This jammed state is tumour suppressive, as it maintains tissue cohesion and limits the emergence of invasive clones5,11,12. Conversely, alterations in adhesion, cell shape or contractility can promote tissue fluidization, facilitating transitions from indolent ductal carcinoma in situ (DCIS) to invasive disease5,11,12.Transitions between three-dimensional spheroids and two-dimensional monolayers provide an experimentally tractable framework to study these processes, particularly in the context of active epithelial wetting5,6. Unlike passive wetting, active wetting arises from the balance between cell-generated traction forces and intercellular contractility and is well captured by active polar fluid models9,13. Despite remarkable theoretical progress, the molecular mechanisms controlling these tissue-level phase transitions remain poorly understood.Here we investigate the role of IRSp53, a membrane-associated I-BAR domain protein that links membrane curvature sensing to actin cytoskeletal regulation14,15,16. IRSp53 promotes directional migration by inducing the formation of filopodia and lamellipodia15,17,18, which have also been implicated in shaping cell–cell adhesions in epithelia19,20. Consistent with this, IRSp53 localizes at cell–cell contacts21,22,23, regulates integrin trafficking24, and is required for the polarized architectural organization and morphogenesis of epithelial tissues21. These properties position IRSp53 as a candidate regulator of epithelial mechanics at the interface between membrane remodelling and cytoskeletal force transmission.Using MCF10DCIS.com cells as a model for early breast cancer25, we show that IRSp53 depletion disrupts coordinated collective migration in two-dimensional (2D) monolayers and delays jamming transitions at high cell density. By contrast, the loss of IRSp53 enhances the spreading of three-dimensional (3D) spheroids, indicating a primary effect on epithelial collectives’ viscosity rather than cell–substrate adhesion. Combining live imaging, biophysical measurements and theoretical modelling, we demonstrate that these effects arise from reduced supracellular tension, altered junctional organization and increased fluidity of epithelial collectives. At the molecular level, we identify Afadin (AFD) as a functional interactor of IRSp53. AFD is essential for epithelial development, polarity and cortical tension regulation26,27,28,29,30, and its depletion recapitulates IRSp53-loss phenotypes in wetting assays. Together, our findings identify IRSp53 and AFD as key regulators of epithelial viscoelasticity, linking junctional architecture and single-cell mechanics to tissue-scale phase transitions relevant to tumour progression.Consistent with this framework, the analysis of patient datasets reveals that the reduced expression and aberrant localization of IRSp53 correlate with a poor clinical outcome in breast cancer, underscoring the pathological relevance of IRSp53-regulated tissue mechanics.IRSp53 removal affects DCIS collective motionIRSp53 regulates the interplay between the plasma membrane and the actin cytoskeleton during directional migration and invasion, contributes to cell–cell and cell–ECM adhesions, and is required for the polarized organization of epithelial tissues14,15,16,17,18,31. Perturbation of these processes is associated with invasive breast cancer and the emergence of collective migratory behaviours32.To investigate the role of IRSp53 in this context, we generated inducible shRNA-expressing MCF10DCIS.com cells to downregulate IRSp53 (Fig. 1a). This cell line models early breast cancer progression, as it forms ductal-carcinoma-like structures that can evolve into invasive tumours in vivo25. MCF10DCIS.com cells retain epithelial characteristics, including E-cadherin-based junctions and the ability to form dense, jammed monolayers and exhibiting partial mesenchymal traits12.Fig. 1: IRSp53 removal affects DCIS collective motion.a, Immunoblotting of MCF10DCIS.com SCR, IRSp53_KD and IRSp53_KD rescued with murine IRSp53 (IRSp53_KD + mIRSp53) using the indicated antibodies. b, Scratch-wound migration of SCR, IRSp53_KD and IRSp53_KD + mIRSp53 monolayers (Supplementary Video 1). Left: representative still images at the indicated times. Scale bar, 100 µm. Right: motility quantification. Top: wound coverage rate, measured as % area covered over time (early phase = first 2 h of active migration). Bottom left: velocity correlation length from PIV. Mean ± s.d. (≥15 fields of view; three independent experiments). Bottom right: directionality index from manual tracking of leading-edge cells in ImageJ (example trajectories shown). Mean ± s.d. (n = 32 cells; three independent experiments). c, Left: PIV-derived root mean square velocity (Vr.m.s.) over time for SCR-H2B-GFP and IRSp53_KD-H2B-mCherry MCF10DCIS.com cells seeded at the jamming density and monitored by time-lapse microscopy (Supplementary Video 5). Mean ± s.d. (>20 fields; three independent experiments). The vertical dashed line indicates the time point (t = 50 h) corresponding to the snapshots shown beside. Centre: dashed line marks t = 50 h, corresponding to the velocity-field snapshots shown. Scale bar, 100 µm. Right: Vr.m.s. at t = 50 h. Mean ± s.d. (>20 fields from three independent experiments). d, Left: representative PIV velocity-field snapshots at t = 20 h for empty vector (EV) or RAB5A cells, transfected with control oligos (Luc, EV_Luc, RAB5A_Luc) or human IRSp53 siRNA (EV_IRSp53_RNAi, RAB5A_IRSp53_RNAi), seeded at the jamming density and imaged by time-lapse microscopy (Supplementary Video 6). The colour map reflects the alignment with respect to the mean velocity vCM, quantified by the parameter a(x) = v(x) · vCM/||v(x)||||vCM||. A value a = 1 (a = –1) indicates that the local velocity is parallel (antiparallel) to the mean direction of migration, indicated by the red line in the top-left square of each snapshot. Scale bar, 200 µm. Right: collective motion velocity VCM shown in the log scale. Mean ± s.d. (n = 12 fields; two independent experiments). IRSp53 and RAB5A expressions were verified using quantitative reverse-transcription polymerase chain reaction (Extended Data Fig. 1e). Statistical tests are listed in Supplementary Table 4; the P values are shown in the graphs.Source dataFull size imageWe first examined collective migration during wound healing in confluent monolayers. IRSp53 depletion had minimal effects on the initial rate of wound closure (Fig. 1b, early phase, and Supplementary Video 1). However, it significantly reduced the long-range coordination of cell motion, as evidenced by the decreased correlation length and reduced directionality of single-cell trajectories (Fig. 1b, Extended Data Fig. 1a and Supplementary Videos 1 and 2). These defects resulted in delayed wound closure at later time points. The re-expression of murine IRSp53, resistant to the shRNA, fully rescued coordinated migration and wound closure dynamics (Fig. 1a,b).In particular, IRSp53 loss did not affect the migration of isolated cells in one-dimensional linear motility assays (Extended Data Fig. 1b and Supplementary Video 3), indicating that IRSp53 primarily regulates emergent collective properties rather than intrinsic single-cell motility. Loss of coordination was observed not only at the wound edge but also deep within the monolayer (Supplementary Video 4), suggesting a defect in long-range mechanical or polarity coupling across the cell collective.Dense epithelial monolayers undergo a fluid-to-solid (jamming) transition as proliferation increases crowding and constrains cell motion10. We, therefore, tested whether IRSp53 contributes to this transition. Although control monolayers progressively arrested their motion, IRSp53-depleted monolayers remained fluid-like, displaying sustained velocity fluctuations at late time points. This was quantified by an increased root mean square velocity (Vr.m.s.), measured by both particle image velocimetry (PIV) and single-cell tracking (Fig. 1c, Extended Data Fig. 1c and Supplementary Video 5). Importantly, IRSp53 depletion did not alter the proliferation rates, as nuclear density increased similarly in control and knockdown monolayers (Extended Data Fig. 1d), indicating that delayed jamming was not due to differences in cell growth.Previous work demonstrated that the expression of the endocytic protein RAB5A reawakens motility in jammed epithelial monolayers by inducing long-range coordinated flocking motion7,12,33. We, therefore, asked whether IRSp53 is required for RAB5A-induced collective migration. As expected, RAB5A expression promoted highly coordinated multicellular streams, quantified by an increase in the velocity of the centre of mass (VCM) (Fig. 1d, Extended Data Fig. 1e and Supplementary Video 6). Notably, silencing IRSp53 in RAB5A-expressing monolayers severely impaired the flocking behaviour, resulting in a marked reduction in VCM despite preserved RAB5A expression.Together, these data demonstrate that IRSp53 is dispensable for single-cell migration but is essential for long-range coordination of collective motion, the establishment of a jammed state at high density and the emergence of coherent flocking dynamics in epithelial monolayers.IRSp53 loss decreases epithelial collective viscosityTo identify the physical parameters controlled by IRSp53 during collective dynamics, we investigated the active wetting behaviour of epithelial spheroids, a well-established system to probe active tissue fluidization6,9,13. In this assay, 3D spheroids spread onto adhesive substrates, transitioning into 2D monolayers through an active wetting process.Time-lapse microscopy revealed that IRSp53 depletion markedly accelerated spheroid spreading compared with control spheroids (Fig. 2a, Extended Data Fig. 1f and Supplementary Videos 7 and 8). This effect was reproduced in primary murine mammary epithelial cells lacking IRSp53 as well as in human HaCat keratinocytes with inducible IRSp53 knockdown (Extended Data Fig. 1g–h and Supplementary Videos 9 and 10), indicating a general role for IRSp53 in regulating wetting dynamics.Fig. 2: IRSp53 loss decreases cell spheroid viscosity.a, Representative still images from the time-lapse images of SCR or IRSp53_KD MCF10DCIS.com spheroids seeded on fibronectin-coated six-well plates (Supplementary Video 7). Scale bar, 200 μm. Left graph: spreading area quantified manually (ImageJ). Mean ± s.d. (n = 21 SCR, 20 IRSp53_KD spheroids; three independent experiments). Right graphs: semiautomatic segmentation of area versus time shows faster spreading on IRSp53 knockdown. Mean ± s.d. (n = 7 SCR, 5 IRSp53_KD spheroids). Solid black lines, linear fits with slopes dA/dt = (1.27 ± 0.19) × 104 µm2 h−1 for SCR and dA/dt = (4.77 ± 0.33) × 104 µm2 h−1 for KD. Box chart: dA/dt values of a single spheroid. Mean ± s.d. (n = 16 SCR, 11 IRSp53_KD spheroids; three independent experiments). b, Model framework. Left: monolayer internal stress contributions—active contractile stress (amplitude −ζ, linked to the polarity field) and viscous stress (amplitude η, linked to velocity gradients). Middle (top): spreading geometry defining the initial size R0 (equivalent radius at t = 0), monolayer radius R, thickness h and stiff-core radius R1 (radial velocity expected to vanish). Velocity v and traction T fields are indicated. Middle (bottom): radial velocity and predicted traction profiles. vr intersects the x axis in R1; initial slope \(\approx \frac{{T}_{0}{L}_{{\rm{c}}}}{\eta h}\); slope at \({\rm{r}}=R\approx \frac{\zeta }{\eta }\); velocity peak occurs at ≃Lc from the boundary. Predicted traction profile Tr has a peak amplitude T0 and characteristic length Lc (nematic length). Top right: example at t = 24 h of an IRSp53_KD spheroid—top view with a segmented area (red) and the corresponding PIV velocity field (colour map, µm h−1). Scale bar, 200 µm. Bottom right: representative azimuthally averaged vr profiles over time for SCR and IRSp53_KD. Mean ± s.d. over five consecutive frames. Model fits yield viscosity- and contractility-related parameters. c, Time evolution of core size R1/R0: approximately constant in SCR but rapidly decreases in IRSp53_KD, indicating faster core ‘melting’. Mean ± s.d. (n = 7 SCR, 5 IRSp53_KD). d, Core-melting rate d(R1/R0)/dt from linear fits of single spheroids. Mean ± s.d. (n = 15 SCR, 11 IRSp53_KD; three independent experiments). e,f, Fit parameters A (e) and B (f) from radial velocity profiles evaluated at matched spreading. A, traction to viscosity (T0/η); B, contractility to viscosity (ζ/η). Mean ± s.d. (n = 16 SCR, 11 IRSp53_KD; three independent experiments). g, Traction force microscopy. Left: segmentation and traction map of an IRSp53_KD spheroid (colour map, kPa). Scale bar, 50 µm. Right: representative radial traction profiles over time for SCR and IRSp53_KD. Solid lines, best fits. h, Average evolution of the maximum traction T0 over time for SCR (blue line) and IRSp53_KD (red line). i,j, T0 (i) and Lc (j) evaluated at matched spreading (the front has doubled: R = (2.0 ± 0.2)R0). Mean ± s.d. (n = 29 SCR, 28 IRSp53_KD spheroids; three independent experiments). No significant difference in Lc or T0 is observed. Statistical tests are listed in Supplementary Table 4; the P values are shown in the graphs.Source dataFull size imageIn analogy with liquid droplets, the spreading of cell aggregates reflects a balance between cell–cell cohesion and traction forces exerted on the substrate6,9. Given that IRSp53 depletion impaired coordinated migration in wound-healing assays (Fig. 1b), it was unlikely that enhanced spreading resulted from increased cell–substrate adhesion. We, therefore, hypothesized that IRSp53 primarily regulates collective viscosity and cohesion.Both control and IRSp53-depleted spheroids spread at approximately constant rates of \(\frac{{\rm{d}}{A}}{{\rm{d}}{t}}\); however, the spreading rate was significantly higher on IRSp53 knockdown (Fig. 2a). To gain a mechanistic insight, we quantified the time-resolved velocity fields of spreading spheroids using PIV. Radial velocity profiles displayed a characteristic non-monotonic shape, with a maximum near the spreading front (Fig. 2b), consistent with predictions of active polar fluid theory9.We modelled the spreading monolayer as a two-dimensional active polar fluid characterized by a polarity field p and a velocity field v. Under the adiabatic approximation, the polarity field obeys the steady-state equation$${L}_{{\rm{c}}}^{2}{\nabla }^{2}{p}_{\alpha }={p}_{\alpha },$$ (1) where Lc is the nematic length. In the low-Reynolds-number limit, force balance requires that traction stresses exerted on the substrate balance gradients of the internal stress tensor σs/h:$$\frac{1}{h}{\partial }_{\beta }{\sigma }_{\alpha \beta }^{{\rm{s}}}={T}_{\alpha },$$ (2) where h is the monolayer thickness and σs is symmetric part of the stress tensor. The antisymmetric part, consistent with the adiabatic approximation for the polarity field, is assumed to be negligible. The model is complemented with the following simplified constitutive equations:$${\sigma }_{\alpha \beta }^{{\rm{s}}}=\eta \left({\partial }_{\alpha }{v}_{\beta }+{\partial }_{\beta }{v}_{\alpha }\right)-\zeta {p}_{\alpha }{p}_{\beta }$$ (3) and$${T}_{\alpha }=-{T}_{0}{p}_{\alpha }$$ (4) are the internal stress and the traction stress, respectively. In previous equations, T0 represents the maximum traction stress, which quantifies the maximum force per unit area exerted by cells on the substrate; ζ 20–30 event counts to NaN. For island analysis, event count images were used to create an arbitrary binary mask of cell–cell junctions, to avoid contribution from non-junctional portions of the cells to lifetime calculations. Mean lifetime values for each thresholded image were obtained using Fiji.Blebbing analysisSCR and IRSp53_KD cells were detached by trypsinization and resuspended by gentle pipetting for 5 min. Bleb formation and reabsorption over time was monitored by a DeltaVision microscope with a ×40 oil objective every 10 s for 5 min. The blebbing dynamics was quantified by segmentation of cells with a custom Fiji plug-in63. After choosing the object to measure for each cell, the plug-in applies a WEKA custom model64 to identify the objects and extrapolates the area variation for each frame.Nanomechanical AFM mapping of MCF10DCIS.com monolayerSCR, IRSp53_KD and AFD_KD cells were seeded in confluent conditions on 35-mm glass-bottom dishes (FluoroDish FD35–100, World Precision Instruments), in complete medium with 1 µg ml−1 of doxycycline. The day after, monolayers were rinsed with PBS 1× several times and taken to the NanoWizard 4XP AFM system (Bruker). After calibrating a PFQNM-LC-V2 live-cell probe (Bruker) and letting it equilibrate to 37 °C, the cells were placed on the AFM stage to image. Cells were scanned in the quantitative imaging mode to generate high-spatial-resolution maps to quantify Young’s modulus using a modified Hertz contact model for paraboloidal tips65. Scans were taken at 256 × 256 resolution with a 1-nN force setpoint, 1,500-nm Z range and 100 µm s−1 of ramp speed over an area of 50 × 50 µm2.Laser ablation on junctions and apical cortexSCR and IRSp53_KD cells (stably expressing EGFP–CAAX as membrane reporter) or AFD_KD cells (stably expressing mCherry–CAAX) were seeded in confluent conditions (at the jamming density) onto Nunc Glass Base dish (diameter, 27 mm) in complete medium with 1 µg ml−1 of doxycycline.After 48–72 h, monolayers were subjected to targeted laser ablation. Experiments were performed using a Leica Stellaris DIVE combined with a femtosecond pulsed Spectra Physics Laser capable of priming two-photon excitation. The ablation was performed using a 750-nm wavelength. Imaging was performed with single-photon excitation in a regular confocal fashion. The objective used was an HC PL APO ×63/1,2 W UVIS CS2. For laser ablation on junctions, the distance between vertices (vertex separation) defining the ablated contact was measured as a fraction of time. Distance values were subtracted from the initial contact length. The values were then calculated as a function of time, and the initial recoil values for each contact were obtained by a nonlinear regression of the data to the following equation: f(t) = (initial recoil/k)(1 – e−kt) (refs. 7,66). For laser ablation of the apical cortex, changes in the cellular area (measured by the CAAX signal as a proxy for the cell perimeter) were quantified as a fraction of time using Fiji software29.CLEMElectron microscopy examination and CLEM were performed as previously described21,67,68. In brief, MCF10DCIS.com SCR and IRSp53_KD cells (stably expressing EGFP–E-cadherin as a junctional reporter) were trypsinized to a single-cell suspension at 4 × 104 cells ml−1 in complete medium containing 2% Matrigel (BD Biosciences). Cell–medium–Matrigel suspensions (350 μl) were plated in 35-mm dishes with gridded coverslips (P35G-1.5–14-CGRD, MatTek), precoated with 20 μl of Matrigel (10 mg ml−1). Then, 2% Matrigel complete medium was added to 2 ml of complete medium with 1 µg ml−1 of doxycycline. After 72–96 h, the obtained spheroids were stained with NucBlue Live ReadyProbes and identified on grids by confocal microscopy. The fluorescence images for the CLEM data were acquired on a Leica SP8 DLS confocal microscope, using an HC PL FLUOTAR ×20/0.5 objective.Electron microscopyA brief description of each process is presented below.EmbeddingSpheroids grown on MatTek glass-bottom dishes were fixed with a 2.5% paraformaldehyde and 2.5% glutaraldehyde (EMS) mixture in 0.2 M of sodium cacodylate (pH 7.2) for 2 h at room temperature, followed by three washes in 0.2 M of sodium cacodylate (pH 7.2) at room temperature. Then, the spheroids were incubated in a 1:1 mixture of 2% osmium tetraoxide and 3% potassium ferrocyanide for 1 h at room temperature followed by rinsing for three times in cacodylate buffer. Afterwards, the samples were sequentially treated with 0.3% thiocarbohydrazide in 0.2 M of cacodylate buffer for 15 min and 1% osmium tetraoxide in 0.2 M of cacodylate buffer (pH 6.9) for 30 min. The samples were then rinsed with 0.1 M of sodium cacodylate (pH 6.9) buffer until all traces of the yellow osmium fixative were removed and then it was washed in deionized water, treated with 1% uranyl acetate in water for 1 h and washed in water again. Finally, the samples were subjected to dehydration in ethanol and embedded in epoxy resin at room temperature and polymerized for at least 72 h in a 60 °C oven67.SectioningAs described above, the spheroid of interest was selected during the optical sectioning and Z stacking using confocal microscopy. During Z stacking, the distance between the bottom and surface of the spheroid was estimated. The embedded samples were then sectioned with a diamond knife (DiATOME) using a Leica EM UC7 ultramicrotome. For the trimming and advance to the region of interest, we used the DiATOME diamond trimming blades (trim tool 45 and histo; DiATOME). To avoid mistakes, we took into consideration the possible ‘shrinkage’ of the matrix during its dehydration and embedding into Epon. Thus, when only 3 µm was left before the beginning of the cell surface, we replaced the trimming histo-knife with an Ultra 35 knife (DiATOME), and then cut two 200-nm sections and then a small series of 70-nm sections. Sections were analysed with a Tecnai 20 High Voltage EM (Thermo Fisher Scientific) operating at 200 kV.Immunofluorescence on MCF10DCIS.com monolayersSCR and IRSp53_KD cells were seeded in confluent conditions (at the jamming density) in μ-Slide 8 well ibiTreat chambers (ibidi, 80826) in complete medium with 1 µg ml−1 of doxycycline. After 16 h, monolayers were fixed with 4% paraformaldehyde for 20 min at room temperature. Cells were then permeabilized and blocked with PBS 0.1% Triton X-100 and 0.2% bovine serum albumin for 20 min at room temperature. Incubation with primary antibody solution (1% bovine serum albumin in PBS 1×; Supplementary Table 3) was performed overnight at 4 °C in a wet chamber protected from dark. After three washes of 5 min each with PBS 1×, incubation with secondary antibody solution (PBS 1×; Supplementary Table 3) was performed for 1.5 h at room temperature. After three washes of 5 min each with PBS, nuclei were stained with DAPI (1:1,000 in PBS 1×) and samples were stored in PBS at 4 °C till acquisition.Samples were imaged using a Leica SP8 DLS microscope (lasers used were 405, 488, 561 and 647 nm) with a ×63 (1.4 correction) oil objective. Confocal sections on the z axis were acquired with 0.5-µm step size and resliced on the y axis using ImageJ, to get maximum projections of actin staining (TRITC–phalloidin). A Leica algorithm allowing for computational clearing (Thunder) was also used to image phallodin-positive intercellular spaces in the MCF10DCIS.com monolayer.Clinical cohorts and histopathological evaluationA TMA comprising a retrospective consecutive cohort of 1,755 female patients, from the European Institute of Oncology in Milan53,54, with complete clinicopathological follow-up, was used. All patients provided written informed consent and underwent surgical procedures. Sections obtained from paraffin-embedded tissues were stained with mouse monoclonal anti-IRSp53, as already described21, and analysed with a Scan Scope XT device and the Aperio Digital pathology system software (Aperio, Leica). Patients were stratified according to the immunohistochemical pattern distribution of IRSp53 (apical–membrane distribution versus cytosolic distribution) in the primary tumour. The association between IRSp53 apical–membrane distribution or cytosolic distribution and different clinicopathological parameters was evaluated using Fisher’s exact test.Statistical analysis and data reproducibilityAll quantitative data are presented as scatter plots with bars indicating the mean ± s.d. of independent biological replicates, unless stated otherwise. The number of independent experiments and the number of samples analysed for each experiment are reported in the corresponding figure legends. Exact P values, when statistically relevant, are indicated in the figures and summarized in Supplementary Table 4. No statistical methods were used to predetermine the sample size. No data were excluded from the analyses. Experiments were not randomized, and investigators were not blinded to allocation during experiments or outcome assessment.For all the quantitative analyses, the unit of analysis was defined as the smallest independently measurable biological entity subjected to the experimental intervention. Depending on the assay, this corresponded to individual spheroids, individual cells, individual cell collectives or independent fields of view within confluent monolayers. These units were independently generated and analysed both within and across experiments. As variability between units within a single experiment was comparable with variability observed between independent experiments, data from multiple experiments were pooled for statistical analysis. Pooling was performed only after confirming the reproducibility of the measured distributions across biological replicates, with no detectable differences in effect size or variance between experiments. Accordingly, for each figure, both the number of samples analysed and the number of independent experiments performed are reported.Reporting summaryFurther information on research design is available in the Nature Portfolio Reporting Summary linked to this article.