Dynamic forces shape the survival fate of eliminated cells

Abstract

Tissues eliminate unfit, unwanted, or unnecessary cells through cell extrusion, and this can lead to the elimination of both apoptotic and live cells. However, the mechanical signatures that influence the fate of extruding cells remain unknown. Here we show that modified force-transmission across adherens junctions inhibits apoptotic cell eliminations. By combining cell experiments with varying levels of E-cadherin junctions and three-dimensional modelling of cell monolayers, we find that these changes not only affect the fate of the extruded cells but also shift extrusion from the apical to the basal side, leading to cell invasion into soft collagen gels. We generalize our findings using breast cancer patient-derived xenografts and cysts cultured in matrigel. Our results link intercellular force transmission regulated by cell-cell communication to cell extrusion mechanisms, with potential implications during morphogenesis and invasion of cancer cells.

Introduction

Cell extrusion is a checkpoint mechanism by which unwanted or dead cells are eliminated from a monolayer through cooperation of neighboring cells either mediated by actomyosin contractile rings or lamellipodial protrusions^1–5. Elimination of cells is critical for regulating cell numbers, sculpting tissues, or eliminating damaged cells throughout developmental programs^6–9. During cell extrusion, a cell within an epithelial monolayer is eliminated by its neighbours through loss of apico-basal polarization into the basal or luminal side, respectively^10,11. Importantly, cells can be extruded alive or as dead cells^6,12,13. The extrusion of live or dead cells has been reported during apoptosis¹⁴, epithelial-mesenchymal transition (EMT)¹⁵, or cancer cell invasion¹⁶. This makes cell extrusion a key regulator of homeostatic pressure, i.e. pressure at which cell extrusion compensates cell division¹⁷, morphogenesis⁹ and tumor progression⁷. A large variety of biochemical and biophysical cues including crowding, apoptotic stimuli or mechanical forces can thus alter the fate of these extruding cells^{6,12,13,18,19}. Furthermore, the extrusion of live cells is often associated with a switch from apical to basal direction, leading to tumour cell escape and extra-cellular matrix invasion^10,11. Investigating how the fate of extruding cells is orchestrated offers key clues for revealing physiological and pathological processes that remain poorly understood.

Several studies have established that cell extrusion is linked to the remodeling of neighboring cells through compensatory proliferation^12,20–22, generation of mechanical forces^8,9,23,24, emergence of topological defects¹⁸ as well as the activation of signaling pathways^25,26. This communication between an extruding cell and its neighbors crucially depends on cell-cell junctions^5,27. E-cadherin junctions that regulate cell-cell interactions have been identified as platforms of mechanosensing capable of transmitting forces to neighboring cells^28–31. The biochemical and mechanical regulations of adherens junctions through E-cadherin have been implicated during both homeostatic and oncogenic extrusions^5,16,27,32 and tumour metastasis³³. Thus, we hypothesize that intercellular forces mediated by adherens junctions could play a role in determining the fate of extruding cells as well as their mode of extrusion, basal versus apical. In this work, we employed a comprehensive approach combining in vitro cell culture models, patient-derived xenografts and agent-based models to unveil the role of E-cadherin-based junctions in the establishment of extrusion programs. We find that the absence of E-cadherin in epithelial cells and their increased cell contractility inhibits apoptotic cell extrusion while promoting live cell invasion into the ECM when cultured on thick collagen gels or in 3D gels. We use a 3D phase-field model³⁴ to identify the mechanical pathway that promotes various extrusion mechanisms depending on intercellular forces. In silico and experimental results reveal that cells expressing E-cadherin are subjected to higher and more persistent compressive stresses from their neighbors than cells without E-cadherin, culminating in the extrusion of live cells in monolayers without E-cadherin. We further find that live cell extrusions are associated with non-apoptotic blebbing³⁵ further confirmed by gene expression characterization and protein analysis. In contrast, apoptotic cells upon caspase activation display apoptotic blebbing that is followed by cell fragmentation. Our work demonstrates that different modes of cell extrusion processes are attributed to alterations in the generation, exertion, and transmission of mechanical forces within the tissue leading to genetic and protein level changes.

Results

Live versus apoptotic extrusion driven by weakened cell-cell adhesions

We first used MDCK cells grown on fibronectin coated plastic surfaces as a model system and investigated the impact of E-cadherin based adhesions during cell extrusion. Upon E-cadherin loss of function (hearafter referred to as E-cad KO) we observed no significant difference in apico-basal polarity establishment through immunostaining of ZO1 and podocalyxin (Extended Fig. 1a, b) and TEER measurements (Extended Fig. 1c), the temporal evolution of monolayer density (Fig. 1a) nor the number of extrusions (Fig. 1b), within the first 24 hours. We thus focused on the first 24 hours period to exclude any effects from the differences in density or rate of extrusion. Here, extrusion refers to detachment of a single cell from the monolayer. Inhibition of caspase only partially inhibited extrusion events (Extended Fig. 2a) suggesting that some extrusions might be non-apoptotic. Surprisingly, we found a significant increase in the fraction of live cell extrusions from E-cad KO monolayers (Fig. 1c, d, e) through three different assays (see Methods for details): anti-caspase immunostaining of fixed monolayers (Fig. 1c), (ii) flow cytometry of extruded cells (Fig. 1d, Extended Fig. 2b) and (iii) live imaging using annexin V as a reporter dye (Fig. 1e). In addition, plating the extruded cells led to colony growth of extruded cells from E-cad KO monolayers (Extended Fig. 2c) while cells extruded from WT monolayers did not grow upon replating. Live imaging using annexin V as an apoptosis readout confirmed a significant increase in the fraction of live extrusions within E-cad KO monolayers that constituted more than 90% of all extrusion events at early time points (Fig. 1e, Extended Fig. 2d, e). Here live extrusion is defined as an event that does not express annexin V within 60 mins after extrusion from the monolayer. These results were then verified by live imaging with caspase dye (Extended Fig. 2f) and an endogenously tagged caspase reporter (Extended Fig. 2g). Moreover rescue experiments, where E-cadherin was re-introduced, showed live extrusion profiles similar to WT monolayers (Fig. 1e, Extended Fig. 2d,e). This confirms that the outlined effects are indeed due to E-cadherin loss of function (Fig. 1e, Extended Fig. 2d,e). Altogether, our data reveals that E-cadherin loss of function affects the fate of extruding cells and increases the proportion of live cell extrusions.

Figure 1. E-cadherin loss of function leads to an increase in live cell extrusion within epithelial monolayers:

a) Bar plot showing the time evolution of mean density (number of cells/mm²) over an area of 0.5mm² over the analyzed time window of 0 and 24 hours for both WT (blue) and E-cad KO (red) monolayers. Error bars represent standard error mean (SEM). Here * represents p<0.05, ** p<0.01 and p>0.05 represents n.s. n=21 for E-cad KO and n=22 for WT using unpaired t-test from 2 independent experiments. b) Bar plots showing the number of extrusions averaged over 0-24 and 24-48 hours for both WT (blue) and E-cad KO (red) monolayers. Error bars represent standard error mean (SEM). Here * represents p<0.05 and no significance (n.s.) (p>0.05), n= 5 for E-cad KO and n=4 for WT from an unpaired t-test from 2 independent experiments. c) Maximum intensity projections of immunostained MDCK WT (top) and MDCK E-cad KO (bottom) monolayers showing apoptotic (caspase positive- magenta) and live (caspase negative- magenta) cell extrusions stained for actin (green), and nuclei (yellow). Scale: 50μm. d) Average percentage of live cell (propidium iodide negative) and dead (propidium iodide positive) extrusions obtained through cytometry based sorting from 9 different samples from 2 independent experiments after 24 hours of plating (100% confluency). *** indicates p < 0.001 from an unpaired t-test. Error bars represent SEM. e) Fraction of live (annexin negative) and apoptotic (annexin positive) extrusions obtained from MDCK WT, MDCK E-cad KO, MDCK E-cad KO+ E-cad GFP averaged over the first 10 hours and time period 20-30 hours. Data presented is averaged over 36 (WT), 62 (E-cad KO) and 55 (E-cad KO+ E-cad GFP) events from 2 independent experiments. p values were obtained from Kruskal Wallis test where, **** p<0.0001 and p<0.05 is n.s (not significant and p > 0.05). Error bars here represent standard deviation (S.D).

Mechanical forces determine the fate of extruding cells

Previous work has shown that physical mechanisms based on crowding effects^6,12, mechanical instability^11,18,24,34, microtubule disassembly³⁶ and topological rearrangements¹⁸ can lead to cell extrusion. Thus, we hypothesized that E-cadherin loss³⁷ could impact the fate of extruding cells through changes in mechanical stresses. To test this hypothesis, we first focused on mechanical stress patterns around extrusion sites using a recently developed 3D phase-field model of the cell monolayer³⁴, in which cell-cell and cell-substrate adhesion strengths are considered explicitly and tuned independently (see Methods for model details and parametrization) (Extended Fig. 2h). Cell extrusion, in this model, is an emergent behavior from modeling the full 3D dynamics of cell shape changes in a monolayer that affects the in-plane and out-of-plane forces acting on a cell. As such extrusion events occur without any explicit threshold, or external artificial means of favoring extrusion. Once the out-of-plane forces acting on a cell overpower the forces keeping it in the monolayer and on the substrate, a cell extrusion occurs. The main differences in mechanical imprints of WT relative to E-cad KO cells are a concomitant decrease in cell-cell adhesion, increase in cell matrix adhesion as previously shown³⁷, reduced cell velocity (Extended Fig. 3a), increase in cell traction forces (Extended Fig. 3b), no change in average isotropic stress (Extended Fig. 3c) and an increase in cell stiffness (Extended Fig. 3d). Therefore, using the model, we compared the evolution of mechanical stress around extrusion sites for two different scenarios: a (i) model wild-type monolayer (mWT) with strong cell-cell contacts, weak cell-substrate interaction forces and low stiffness (Video S1) relative to a (ii) model E-cad KO monolayer (mE-cad KO) with weaker cell-cell contacts, stronger cell-substrate interaction forces and higher stiffness (Video S2).

In silico modeling revealed a more pronounced compressive stress buildup over time near extruding mWT cells compared to mE-cad KO cells in a region of about two cell sizes (Fig. 2a). We thus computed the ensemble average of time-averaged isotropic stress fields around extrusions in our model (Fig. 2b). Interestingly, the corresponding probability density function is non-symmetric (Fig. 2c) and together with the cumulative distribution function (Fig. 2d) indicate a pronounced and persistent compressive stress build-up near extruding mWT cells, given the distribution tails and the time-averaged nature of these local fields. At the same time, the probability density corresponding to the averaged stress fields around extruding mE-cad KO cells is almost symmetric with respect to its peak, with a slight shift towards compressive (negative) stresses. Comparing the shape and the range of distributions indicate relatively high local field fluctuations near extruding mWT cells (Fig. 2c-2d and Extended Fig. 3e). Since E-cad KO cells have a higher stiffness relative to their WT counterparts, we hypothesized that the large and persistent compressive stresses in mWT cells (Fig. 2c-2d) could be due to the longer shape relaxation times (t_shape) for more compliant cells given by, t_shape ~ γ⁻¹ where γ corresponds to cell stiffness in the model (see Methods for details). Indeed, the measured shape relaxation times are longer for WT cells (Extended Fig. 3f) and on a similar order of magnitude as predicted by the linear stability analysis of the shape relaxation from the model (see Methods). Thus, longer shape relaxation time combined with high fluctuating stresses can lead to localization of compressive stresses as seen for extruding mWT cells whereas stiffer mE-cad KO cells can relax into their preferred shapes on a shorter time scale.

Figure 2. Local field statistics unravel the mechanical determinants for the fate of an extruding cell:

a) Time evolution prior to an extrusion event and hence the negative sign of the spatially-averaged isotropic stress, in a square domain of two cell sizes, based on 20 simulations representing 22 mE-cad KO (pink) and 21 mWT (blue) extrusion events. Also, included are randomly chosen non extrusion events (n=100, grey) for the same temporal window. Error bars represent standard error. b) Time-averaged over the temporal window in (a) and ensemble-averaged (over n events) isotropic stress fields representing n=21 extrusion events for mWT (left) and n=22 extrusion events for mE-cad KO (right) obtained from 20 simulations normalized by the maximum value of compression in mE-cad KO cells. c) Probability density function corresponding to the averaged isotropic stress maps for (b) simulations (mWT (solid blue), mE-cad KO (solid pink)) and (f) experiments (WT − apoptotic (dotted blue), E-cad KO − live (dotted pink)). d) Cumulative distribution functions corresponding to probability density in (c) clearly showing the range of the data and the persistence of compressive stresses in mWT and WT − apoptotic extruding cells. e) Time evolution same as (a) of the spatially averaged isotropic stress in a square domain with dimensions of 50μm × 50μm (2-3 cell sizes) representing n = 144 E-cad KO − live (pink) and n = 48 WT − apoptotic (blue) extrusion events based on 3 independent experiments. Also, included are random non extrusion sites (n = 96, grey) for comparison. Error bars represent standard error. The second axes show the actual experimental time and stress while the first axes (left,bottom) show the normalized time and stress. f) Time-averaged over the temporal window in (e) and ensemble-averaged (over n events) isotropic stress fields corresponding to n=37 for WT − apoptotic (left) and n=109 E-cad KO − live (right) based on 3 independent experiments and normalized by the maximum compression value in E-cad KO − live cells. The temporal window for the analyses of simulations and experiments are similar and correspond to 70 mins prior to cell detachment. The mapping of the stresses in simulations and the normalization of experimental data is based on a maximum stress of −250 Pa.μm, i.e. compressive. The mappings between simulation units and physical ones are further detailed in the Methods.

To test these predictions, we turned to experimentally measured stress fields in WT and E-cad KO monolayers. We measured monolayer stress³⁸ and tracked how local stresses evolved in a small region of 50 μm (approximately spanning two-three cell sizes) around the cell extrusion site. The comparison of spatially averaged stress patterns around dead and live cell extrusions for both WT and E-cad KO cells revealed distinct mechanical signatures. As predicted by the model, the isotropic stress evolution with time prior to extrusion showed a three-fold increase of compression in WT apoptotic extrusions compared to E-cad KO live extrusions in the time interval 25 mins before cell extrusion (Fig. 2e) where time zero marks the time point at which the cell detached from the monolayer (Extended Fig. 3g,h). This peak in stress precedes caspase activation by 20-40 mins in WT apoptotic extrusions (Extended Fig. 3i). A similar trend of high isotropic stress was observed for E-cad KO apoptotic extrusion in comparison to WT live extrusions (Extended Fig. 3j-l). However, it does not exclude a potential contribution from caspase at low undetectable levels³⁹. Furthermore, the ensemble average of the time-averaged spatial isotropic stress fields around extrusion events revealed the dominance of compressive stresses around apoptotic extrusions in comparison to live extrusion events (Fig. 2f). Remarkably, both the probability density function (Fig. 2c) and the cumulative distribution function (Fig. 2d) associated with the local averaged experimental fields (Fig. 2f) were consistent with modeling results. Altogether, our experimental and computational observations highlight the roles of cell stiffness and stress fluctuations in localizing compressive stresses around extruding cells and consequently forging cell fate.

Cytoskeletal regulation during apoptotic and live extrusion

Next, we explored how these mechanical patterns could be linked to different signatures at sub-cellular levels within extruding cells. Since actin cytoskeleton has been shown to be remodeled during cell extrusion^1,2,40, we thus analyzed actin dynamics during both apoptotic and live cell extrusions. During apoptotic extrusions in both WT (Fig. 3a, Extended Fig. 4a, Video S3, S4) and E-cad KO (Extended Fig. 4b, Video S5, S6) monolayers, extruding cells undergo a reduction in basal area and aspect ratio prior to caspase activation, and we observed actin enrichment and breakage of the actin structures (Fig. 3a, Extended Fig. 4a) which led to apoptotic blebbing (highlighted with arrows) and eventually cell fragmentation (Fig. 3a’) as previously described^35,41,42. For live extrusions, we observed non-apoptotic blebbing of the extruding cell with actin enrichment, prior to cell rounding up and elimination from the cell monolayer for both cell types (Fig. 3b, b’, Extended Fig. 4c, d, Video S7-10) as evident from the change in basal area and aspect ratio prior to cell extrusion. In addition, we observed a higher percentage and number of live extrusions linked to non-apoptotic blebbing (Fig. 3c, Extended Fig. 4e). In agreement with previous measurements³⁷ we observed an increase in basal myosin contractility upon E-cadherin loss of function, which was validated through staining (Extended Fig. 5a) and western blot (Extended Fig. 5b, c). This increase in contractility could contribute to non apoptotic bleb formation as reported previously⁴³. Taken together, our results suggest that apoptotic cells subjected to high compressive stresses activate caspase, leading to cytoskeleton rupture and membrane fragmentation while live extrusion favoured in E-cad KO monolayers, occurs under lower compressive stresses. Currently, we cannot dismiss the potential of inherent resistance to anoikis in E-cad KO cells, which will need more exploration. However, the quantitative examination of the relationship between caspase activation and stress patterns (see Extended Fig. 3i) suggests a probable correlation, with high compressive stresses potentially occurring before caspase activation in apoptotic extruded cells.

Figure 3. Actin dynamics regulate apoptotic vs live extrusion:

a) Live imaging of LifeAct GFP (black) MDCK WT monolayers along with caspase dye (magenta) over time where, time shown is prior to cell extrusion along with plots of basal area change, basal aspect ratio and caspase expression for this apoptotic extrusion event. Orange lines on the plot indicate corresponding time points from the live imaging shown above. a’) is a zoomed in version of the extruding cell marked with arrows highlighting apoptotic blebbing. b) Live imaging of LifeAct GFP (black) MDCK E-cad KO monolayers along with annexin dye (magenta) over time where, time shown is prior to cell extrusion along with plots of basal area change, basal aspect ratio and annexin expression for this live extrusion event. Orange lines on the plot indicate corresponding time points from the live imaging shown above. b’) is a zoomed in version of the extruding cell, where the arrows show non-apoptotic blebbing. Scale: 10μm. c) Averaged percentage of live or apoptotic cell extrusions that display non-apoptotic blebbing, apoptotic blebbing and neither condition. Error bars represent SEM. Averaged over n=6 (live) and n =7 (apoptotic) different experiments. *** p < 0.001, ** p < 0.01, * p < 0.05 and n.s indicates non-significant (p > 0.05) using t-test.

Beyond mechanics: upregulation of pro survival factors

To further investigate the gene regulatory pathways associated with live cell extrusion, we conducted a comparative analysis of global-level genes and proteins regulations involved in cell survival, cell death, and apoptosis in MDCK WT and E-cad KO monolayers (Methods). At both the transcriptional (Fig. 4a) and translational (Extended Fig. 5d) levels, we observed a reduction in cell-cell adhesion gene expression in E-cad KO cells, validating our approach. We observed an upregulation of anti-apoptotic factors, as there was a significant decrease in the expression of negatively regulating genes at the transcriptional level in E-cad KO monolayers (Fig. 4b) and translational level (Fig. 4c, Extended Fig. 5e). Some of the upregulated genes such as SRC⁴⁴, PTGIS⁴⁵, CD74⁴⁶, SNCG⁴⁷ are also known to suppress apoptosis. Additionally, we observed a substantial regulation of actomyosin cytoskeletal proteins at the protein level, with a notable upregulation of MYH10 (gene name for myosin IIB) (Extended Fig. 5f) and a phosphorylation of ML12B (protein name for MRLC) (Extended Fig. 5g) in E-cad KO cells validating our earlier observation of increased contractility. This analysis provides further molecular support for the observed increase in live cell extrusion as well as for the heightened contractility (Extended Fig. 5a, b, c) of E-cad KO in comparison to WT cells, warranting further in-depth exploration in future studies.

Figure 4. Cell death gene expression favors live cell extrusion upon E-cadherin loss of function:

a) Plot of -log (adjusted pvalue) and normalized enrichment score (NES) of cell-cell adhesion gene expression obtained from bulk sequencing. NES > 1 indicates higher expression in E-cad KO and NES < 1 indicates lower expression in E-cad KO monolayers in comparison to WT monolayers. Green dots represents positive regulation, red dots negative regulation and grey dots bifunctional role of cell-cell adhesions. b) Plot of -log(adjusted pvalue) and normalized enrichment score (NES) of genes regulating cell death obtained from bulk sequencing. Green dots represent positive regulation, red dots negative regulation and grey dots bifunctional role of cell death genes. NES >1 indicates higher expression in E-cad KO and NES < 1 indicates lower expression in E-cad KO monolayers in comparison to WT monolayers. c) Plot of log (fold change (FC)) and -log (p value) of cell death regulatory protein levels obtained from proteomics. FC is the ratio of E-cad KO / WT values, meaning FC > 1 indicates higher expression in E-cad KO and FC < 1 indicates lower expression in E-cad KO in comparison to WT monolayers. Green represents positive regulation, red negative regulation and grey bifunctional role of cell death regulatory proteins. d) Plot of log (fold change) and -log (p value) of cell matrix interaction proteins obtained from proteomics. FC is the ratio of E-cad KO / WT values, meaning FC > 1 indicates higher expression in E-cad KO and FC < 1 indicates lower expression in E-cad KO in comparison to WT monolayers.

Adherens junctions regulate apical versus basal mode of cell extrusion

We then questioned if the ability of cells to be extruded live or dead could have an impact on their mode of extrusion, basal versus apical. While cells are eliminated apically into the lumen during apoptotic homeostatic extrusion⁴⁸, oncogenic cell extrusion occurs via basal elimination^16,49,50. Weakening of E-cadherin based junctions is often associated with epithelial mesenchymal transition (EMT)⁵¹, and cancer invasion⁵². Importantly, at both transcriptional (Extended Fig. 5g) and translational (Fig. 4d) levels, we observed significant changes in cell matrix adhesion, with an upregulation of ITGAV (integrin), COL5A2 (collagen) and SRC (Src kinase) in E-cad KO monolayers which could favor cell invasion. To test this in cell culture, we plated MDCK monolayers on thick 2D type I collagen gels allowing basal extrusion as stiff substrates such as glass do not allow basal extrusion. Staining for focal adhesions, showed an increase in the length of focal adhesions of E-cad KO monolayers when grown on these gels (Extended Fig. 6a). In addition, we noticed extensive remodelling of collagen by E-cad KO monolayers (Extended Fig. 6a). Remarkably, both apical and basal extrusions into collagen were observed in E-cad KO monolayers, while all extrusions in WT monolayers remained apical (Fig. 5a, Extended Fig. 6b-d). Importantly, basal extrusions in E-cad KO monolayers displayed both apoptotic (annexin positive) and live (annexin negative) phenotypes (Extended Fig. 6b), the latter being potentially involved in cell invasion program. This observation was further confirmed by our in silico approach where the direction of cell extrusion switched from apical to basal for a weakened cell-cell adhesion, stronger cell-substrate adhesion and enhanced contractility as observed in E-cad KO cells (Fig. 5b), capturing the competing mechanical forces that determine the mode of extrusion.

Figure 5. E-cadherin loss of function promotes basal cell extrusion reminiscent of tumor cell extravasation:

a) Immunostaining of MDCK WT (left) and MDCK E-cad KO (right) monolayers grown on 2D type I collagen gels, actin (magenta), collagen (yellow) and nuclei (cyan). Arrows indicate extruding cells and bottom images show the orthogonal view. Scale: 10μm. b) Snapshots (top) and phase diagram (bottom) showing the relation between activity (contractile (ζ_Q < 0), intercellular activity), and cell-cell adhesion, and the direction of cell extrusions obtained from simulations demonstrating basal (down, magenta) and apical (up, blue) extrusions. Indeed, there is a finite activity threshold for basal extrusion from the model. Too small a contractility does not deform the cells. The coordinate $\left({\tilde{\zeta }}_Q,{\omega }_c/{\omega }_w\right)=(-5,1.2)$ does not go up nor down. Empty spot indicates the absence of any extrusion. c) Triple negative breast cancer xenografts grown on collagen gels showing the maximum projection of few basal planes (left), apical planes (right) and the orthogonal view (bottom) immunostained for collagen (yellow), actin (magenta) and nuclei (cyan). Scale: 20μm. d) Immunostaining of MDCK WT (top) and MDCK E-cad KO (bottom) cysts grown in Matrigel for 10 days, immunostained for actin (magenta), myosin IIA (white), nuclei (blue) and caspase (green) and quantification of cysts with just apoptotic or the presence of live extrusions from n = 40 (WT) and n = 48 (E-cad KO) cysts from 5 independent experiments. Scale: 20μm. White arrows indicate extruded cell. e) Distribution of extrusion events based on apoptotic or live cells for WT (Blue) and E-cad KO (red). n =40 (WT) and n = 48 (E-cad KO) cysts from 5 independent experiments. f) Distribution of extrusion events based on apical or non-apical extrusions for MDCK WT (blue) and E-cad KO (red). Scale bar: 20m. g) Immunostaining of MDCK WT (top) and MDCK E-cad KO (bottom) cysts grown in Matrigel for 21 days, stained for actin (magenta), nuclei (cyan) and caspase (yellow). Scale: 10μm.

Previous studies have shown that E-cadherin levels promote single or collective cell invasion representing the early stages of the spread of cancer^51–56. E-cadherin expression is known to be heterogeneous between different types of breast cancer cells (ductal (E-cad+) and intra-lobular (E-cad-) carcinoma)⁵⁷ and is observed in very aggresive breast metaplastic cancers⁵⁸. This cancer is characterized by the histological presence of at least two cellular types, typically epithelial (E-cad+) and mesenchymal (E-cad-/vimentin+) cells. To further probe our hypothesis of E-cadherin loss of function driven basal extrusion, we analyzed tumoroids isolated from triple negative metaplastic Patient Derived Xenografts (PDX) grown on 2D type I collagen gels⁵⁹. Over time, we observed basal extrusion of E-cad-/-cells with intact nuclei indicating live extrusions (Fig. 5c) further demonstrating the generality of our findings from epithelial cell lines to human derived cancer cells. Our observation of basal extrusion of E-cad KO cells and E-cad -/- breast cancer cells aligns with earlier findings associated with cancer cell invasion^53,54, which precedes the dissemination of cancer cells. Nevertheless, cancer progression is a highly intricate, pleiotropic, and multi-faceted process, and there are instances where E-cadherin may be necessary for cancer cell survival and metastasis⁵², while E-cadherin loss has been associated with invasion^53–56.

We finally asked whether our key findings of enhanced live (instead of apoptotic) extrusions and promotion of basal (instead of apical) extrusions upon E-cadherin loss of function were limited to flat epithelial monolayers. To this end, we extended our experiments to three-dimensional cysts by growing both WT and E-cad KO cells in 3D matrigel, for 10 days and 21 days. After 10 days, we observed E-cad KO cysts were heterogeneous in shape presenting both apical and basal extrusions (Fig. 5d, Extended Fig. 6e). Among the basally extruded cells, we observed both caspase-positive and -negative cells (Fig. 5d-f, Extended Fig. 6c). In contrast, in WT MDCK cysts, we only found apical and apoptotic extrusions (Fig. 5d-f). After 21 days MDCK E-cad KO cysts showed emergence of clumps of cells accumulating on the basal side, which was never observed in MDCK WT cysts (Fig. 5g). Altogether, our experimental and in silico observations show that low cell-substrate adhesion and high cell-cell adhesion (WT) favour apical extrusions while high cell-substrate adhesion, low cell-cell adhesion and high contractility (E-cad KO) favours basal extrusions.

Discussion

Understanding how the fate of extruding cells, dead or alive, is regulated in an epithelial tissue is crucial for understanding a range of diverse fundamental biological processes including tissue growth, morphogenesis and tumor progression. Our work reveals that the weakening of cell-cell adhesion mediated by the loss of E-cadherin is a key determinant shaping not only the fate of extruding cells but also the direction of extrusion, apical versus basal. These differences between WT and E-cad KO monolayers are driven by intercellular force transmission. Intercellular force transmission modulates the mechanical stress patterns in two critical ways: (i) stress field fluctuations and (ii) shape relaxation times which change as a function of cell-cell adhesion and cell stiffness. Remarkably, our 3D modeling, purely based on physical interactions with no biochemical input, predicts distinct patterns of stress localization around extrusion sites, that are corroborated in our experiments. High compressive stress in extruding WT cells, a consequence of large stress fluctuations and long shape relaxation times, correlates with caspase activation, cytoskeleton fracture and apoptotic cell fragmentation in most cases. On the other hand, an increase in contractility leads to membrane blebbing that may confer anoikis resistance and thus, live cell extrusion as summarized in Fig. 6. This observed blebbing in our experiments is in agreement with a recent study linking bleb formation in single cells to the promotion of resistance to cell death (anoikis) program⁶⁰, and was indicative in our gene expression analysis of cell survival pathways and cell death markers that were differentially regulated in E-cad KO cells. However, a detailed characterization to establish a causal link between blebbing and cell survival in these cell lines future research. We cannot rule out the possibility that differences in mechanical patterns could be linked to a potential higher resistance of E-cad KO cells to apoptosis. While our study has mainly focused on the correlation between physical forces and the fate of extruding cells, exploring the combined roles of both physical and non-physical triggers in cell extrusion mechanisms is an exciting area of study for future works. The analysis of gene expression also led to the discovery of upregulated genes associated with cell-matrix adhesion, demonstrating an intriguing balance between the strength of cell-cell and cell-matrix adhesions that initiates basal extrusion. Along this line, our computational approach (Fig. 5b) showed that model E-cad KO cells with enhanced contractility, decreased cell-cell and increased cell-substrate adhesion strengths can lead to basal cell extrusions. Our experimental observations confirmed the switch from an apical extrusion mode in WT cells to apical and basal extrusions in E-cad KO cells on thick collagen gels in 2D and in 3D matrices, basal extrusions being often associated with live cells in both epithelial cell lines and patient-derived xenografts.

Figure 6.

Schematic summarizing the modes of apoptotic (top) and live (bottom) extrusion within a monolayer. Apoptotic and non-apoptotic blebbing has been highlighted with arrows and a red cell indicates caspase activation.

In summary, in this work, we propose a mechanism to explain the determination of the fate of extruding cells based on the transmission of intercellular forces and cellular stiffness. Cellular extrusion mechanisms can be seen as the emergence of 3D structures from cellular monolayers and, as such, are reminiscent of morphogenetic processes¹⁵. The mechanism provided by our study, based on the alteration of stress patterns around the extruded cell, could therefore represent a widely applicable principle not only for understanding tissue homeostasis, but also for the transition of tissues from a 2D to a 3D shape. Ultimately, our discovery that stress fluctuations can trigger different modes of cell extrusion could also provide insight into epithelial multilayering in skin tissues²³, cell competition mechanism⁶¹ and tumour initiation¹⁰. Overall, we anticipate that this work will have important implications in our understanding of the regulation of tissue homeostasis but also of cell shape changes during morphogenesis^62,63 through, for instance, EMT mechanisms where E-cadherin is differently regulated as well as cancer cell invasion^64–66.

Extended Data

Extended Figure 1.

Immunostaining of WT (top) and E-cad KO (bottom) monolayers for a) gp135 (grey) (podocalyxin) an apical marker and nuclei (cyan) showing the orthogonal view. b) ZO1 a tight junction marker (magenta) and nuclei (cyan) along with their orthogonal view, showing that both WT (top) and E-cad KO (bottom) monolayers are polarized at the stages analyzed in this manuscript. Scale: 10μm. c) TEER values from 2 different samples grown on transwell filters 24 and 48 hours after confluency representative of the densities presented in this work for both WT (blue) and E-cad KO (red) monolayers.

Extended Figure 2.

a) Number of extrusions per mm² per hr for both MDCK WT (blue) and MDCK E-cad KO (pink) monolayers that are treated with caspase inhibitor Z-VAD FMK and the control untreated samples (n=10, WT control, n=9 wt caspase inhibition, n=7 E-cad KO control, n=10 E-cad KO caspase inhibition) from 2 independent experiments. b) Average of absolute number of live extrusions obtained from cytometry where the error bar represents the SEM from 9 samples from 2 independent experiments. p value = 0.038 obtained from Kolmogorov-Smirnov test. c) Examples of extruded E-cad KO cells from confluent monolayers that were replated and grew. Scale: 100μm. d) Snapshots obtained from a time lapse image of MDCK WT (top), E-cad KO (middle) and E-cad KO rescue (bottom) showing cell extrusion that are annexin V labelled (apoptotic, left) and annexin V negative (live, right). Arrows indicate the extruded cell. Scale: 50μm. e) Fraction of live extrusions over time obtained from live imaging of MDCK WT (n=11, green), E-cad KO (n=13, pink) and E-cad KO with E-cadherin rescue (n=12, blue). f) Snapshots of MDCK WT and MDCK E-cad KO monolayers, stained with caspase dye (right) along with the phase contrast image (left) showing the site of extrusion. Arrows indicate the extruded cell. Caspase positive extrusions are labelled as +ve and caspase negative extrusions are labelled as -ve. Scale: 50μm. g) Live imaging of MDCK WT (top) and MDCK E-cad KO (bottom) monolayers that are endogenously tagged with a caspase reports (VC3Ai/ RC3Ai) and co-stained with caspase dye. These monolayers were stained for nuclei with Hoescht during live imaging. Scale: 20μm. h) An example cross-section from a simulation. The arrows show, schematically, how cell-cell and cell-substrate adhesions are explicitly accounted for in our modeling approach and mathematically outlined in the Methods.

Extended Figure 3.

a) Scatter plot of velocity magnitude averaged over a 10 hours time period for both WT (blue) and E-cad KO monolayers (red). Error bars represent standard deviation (SD) for velocity where n =26 for WT and n= 23 for E-cad KO from 2 independent experiments. b) Box plot traction force magnitudes averaged over the 1st hour of imaging and the 15th hour of imaging for both WT (blue) and E-cad KO (red) monolayers. n= 30 for WT and n= 16 for E-cad KO from 2 independent experiments for traction force. c) Box plots of averaged monolayer isotropic stress averaged over the 1st hour of imaging and 15th hour of imaging for both WT (blue) and E-cad KO (red) monolayers. For stress, n =46 for WT and n= 29 for E-cad KO from 2 independent experiments. Here no significance (n.s.) refers to p > 0.05 and **** represents p < 0.0001 from unpaired t-test for velocity and Kolomogrov Smirnov test (stress and force). d) Stiffness of cells within a monolayer obtained via AFM from 4 different samples for each condition. Error bars represent standard deviation (SD). e) The susceptibility of the local isotropic stress fields near extruding cells at and prior to the extrusions showing higher field fluctuations for mWT (in-silico) and WT − apoptotic (experiments) cases relative to mE-cad KO (in-silico) and E-cad KO − live (experiments). The values are normalized by the maximum for mWT and WT − apoptotic cases. The temporal window for the analyses of simulations and experiments are similar (Methods). Furthermore, simulation time and computed stress are non-dimensionalized (Methods). f) Relaxation time of confluent WT and E-cad KO cells obtained from n=9 (WT and E-cad KO) samples. * represents p < 0.05. Error bars represent standard deviation. g, h) Representative example of the evolution of spatially averaged isotropic stress over a region of 50x50μm (pink) and basal area (blue) over time where time zero is point at which basal area becomes zero for g) apoptotic extrusion (WT) and h) live extrusion in an E-cad KO monolayer. i) Histogram showing the time difference between increase in stress and caspase activation during WT apoptotic extrusions from n= 24 events from 2 independent experiments. j) Time evolution of averaged isotropic stress around a square region of 50x50μm around the extruding cell prior to extrusion averaged over 26 (E-cad KO apoptotic) and 96 (WT live) extrusions from 3 independent experiments. Error bars represent standard error. k, l) Representative example of the evolution of spatially averaged isotropic stress over a region of 50x50μm (pink) and basal area (blue) over time where time zero is point at which basal area becomes zero for k) apoptotic extrusion (E-cad KO) and l) live extrusion in a WT monolayer.

Extended Figure 4.

a) Confocal images in the midplane from a time lapsed sequence of the apoptotic cell extrusion labeled with actin-GFP (grey), and Caspase-3 (red). At t = 0 the apoptotic cell is eliminated from the monolayer. The emergence of cavities inside the cytosolic F-actin is observed in this example. Scale: 10μm. b) Time lapse example of apoptotic extrusion of lifeact GFP (grey) labeled MDCK E-cad KO monolayers at the basal (top), mid-lateral (middle)and apical (bottom) planes obtained through spinning disk confocal imaging, stained with caspase (magenta) dye, where time t=0 indicates the time point at which basal area is zero. Side view showing extrusion across time. Time is in minutes and indicated with ‘ symbol. Quantifications show the evolution of basal area, aspect ratio against caspase intensity over time where time zero is time at which basal area is zero and time prior defined as -t mins. Scale: 10μm. c) Time lapse example of live extrusion of lifeact GFP (grey) labeled MDCK E-cad KO monolayers at the basal (top), mid-lateral (middle)and apical (bottom) planes obtained through spinning disk confocal imaging, stained with caspase (magenta) dye, where time t=0 indicates the time point at which basal area is zero. Side view showing extrusion across time. Time is in minutes and indicated with ‘ symbol. Quantifications show the evolution of basal area, aspect ratio against caspase intensity over time where time zero is time at which basal area is zero and time prior defined as -t mins. d) Time lapse example of live extrusion of lifeact GFP (grey) labeled MDCK WT monolayers at the basal (top), mid-lateral (middle)and apical (bottom) planes obtained through spinning disk confocal imaging, stained with caspase (magenta) dye, where time t=0 indicates the time point at which basal area is zero. Side view showing extrusion across time. Time is in minutes and indicated with ‘ symbol. Quantifications show the evolution of basal area, aspect ratio against caspase intensity over time where time zero is time at which basal area is zero and time prior defined as -t mins. Scale: 10μm. e) Total number of live or apoptotic cell extrusions that display non-apoptotic blebbing, apoptotic blebbing and neither condition. Error bars represent SEM. Averaged over n=6 (live) and n =7 (apoptotic) different experiments. *** p < 0.001, ** p < 0.01, * p < 0.05 and n.s indicates non-significant (p > 0.05) using t-test.

Extended Figure 5. Cytoskeletal changes trigger live cell extrusion upon E-cadherin loss of function:

a) Immunostaining of ppMLC (T18/S19) (magenta), actin (yellow) and nuclei (cyan) of WT (top) and E-cad KO (bottom) monolayers highlighting both apical and basolateral regions. Scale: 10μm. b) Western blot of WT and E-cad KO monolayers for ppMLC and MLC total and the b’) fold change of phosphorylation quantified as the ratio of ppMLC/ MLC total obtained from 4 different samples. c) Plot of log (fold change) and -log (p value) of cell adhesion proteins obtained from proteomics. Fold change (FC) is the ratio of E-cad KO / WT values, meaning FC > indicates higher expression in E-cad KO and FC < 1 indicates lower expression in E-cad KO in comparison to WT monolayers. d) Plot showing the proteins identified in the GO term (cell death regulators) where the FC represents the ratio of protein levels in E-cad KO vs WT and a positive value in red indicates upregulation in E-cad KO and a negative value in blue indicates downregulation in E-cad KO. e)Plot of log (fold change) and -log (p value) of cytoskeletal proteins obtained from proteomics. Fold change (FC) is the ratio of E-cad KO / WT values, meaning FC > indicates higher expression in E-cad KO and FC < 1 indicates lower expression in E-cad KO in comparison to WT monolayers. Circled gene is MYH10 a known contractility regulator. f) Plot of log (fold change) and -log (p value) of the proteins obtained from proteomics that were phosphorylated. Fold change (FC) is the ratio of E-cad KO / WT values, meaning FC > indicates higher expression in E-cad KO and FC < 1 indicates lower expression in E-cad KO in comparison to WT monolayers. Circled gene is ML12B a known contractility regulator. g) Plot of -log(p value) and normalized enrichment score (NES) of cell-substrate adhesion gene expression obtained from bulk sequencing. NES >1 indicates higher expression in E-cad KO and NES < 1 indicates lower expression in E-cad KO monolayers in comparison to WT monolayers. Green represents positive regulation, red negative regulation and grey bifunctional role of cell death regulation.

Extended Figure 6.

a) Immunostaining of WT (top) and E-cad KO (bottom) monolayers stained for nuclei (cyan), collagen I (yellow), and paxillin (magenta). Scale: 10μm. Plots on the right show the area and length of focal adhesion quantified from 3 different samples. b) Snapshot from a movie of lifeact GFP (grey) tagged MDCK E-cad KO monolayers stained with Hoescht (nuclei, cyan), annexin (magenta) grown on 2D collagen type I gels (yellow). Position i) and ii) are XZ views through the position i) and ii) labelled on the image. Scale: 10μm. c) Immunostaining of MDCK E-cad KO monolayers grown on collagen gels depicting the apical (top) and basal (bottom) planes stained for actin (magenta), collagen (yellow), nuclei (blue). d) Distribution of the number of apical or basal extrusions from 3 individual samples. Scale bar: 10μm. e) Immunostaining of MDCK WT (top) and MDCK E-cad KO (bottom) cysts grown in Matrigel for 10 days, stained for actin (magenta), myosin IIA (white), nuclei (blue) and caspase (green). Arrows indicate extruded cells. Scale: 20μm.

Extended Figure 7.

The field statistics for the computational model: For simulations and experiments, the susceptiblity values are normalized by the maximums for WT - apoptotic and mWT, respectively. The sensitivity of the computational model and its dimensionless parameters on isotropic stress fields, characterized by mean and susceptibility, for (a) adhesion, (b) elasticity and (c) compressiblity. The values are normalized by the bench case, here defined as the lowest for each studied parameter. Time evolution for the global field statistics, characterized by its mean, variance, skewness and kurtosis, to quantify the how elastic contrast between mWT and mE-cad KO cells affects field statistics, (d) with no contrast and (e) with an elastic contrast. Each curve corresponds to a unique realization for mE-cad KO (solid, red) and mWT (dotted, blue).

Data availability

Sequencing dataset (GSE234079) is publicly available. Proteomics dataset is also publicly available via PRIDE (PXD043091 and 10.6019/PXD043091).

Code availability

The code is available on github: https://github.com/siavashmonfared/celadro_three_dimensional.