An E-cadherin-actin clutch translates the mechanical force of cortical flow for cell-cell contact to inhibit epithelial cell locomotion

Summary

Adherens junctions (AJs) allow cell contact to inhibit epithelial migration yet also permit epithelia to move as coherent sheets. How, then, do cells identify which contacts will inhibit locomotion? Here, we show that in human epithelial cells this arises from the orientation of cortical flows at AJs. When the leader cells from different migrating sheets make head-on contact with one another, they assemble AJs that couple together oppositely directed cortical flows. This applies a tensile signal to the actin-binding domain (ABD) of α-catenin, which provides a clutch to promote lateral adhesion growth and inhibit the lamellipodial activity necessary for migration. In contrast, AJs found between leader cells in the same migrating sheet have cortical flows aligned in the same direction, and no such mechanical inhibition takes place. Therefore, α-catenin mechano-sensitivity in the clutch between E-cadherin and cortical F-actin allows cells to interpret the direction of motion via cortical flows and signal for contact to inhibit locomotion.

Graphical abstract.

Introduction

Epithelial cells possess an intrinsic capacity to move that critically influences their tissue dynamics.¹ Locomoting populations of cells drive morphogenetically dynamic tissues,^2,3 but even apparently quiescent tissues retain a propensity to migrate that is rapidly activated during wound healing and tissue repair.⁴ Clearly then, regulation of locomotility is a central factor in epithelial biology.

Here, a long-standing question is how contact between migrating cells regulates their movement.⁵ Many studies since the pioneering work of Abercrombie and Heaysman^6,7 have documented the phenomenon of contact inhibition of locomotion (CIL).^2,8 This has been implicated in tissue homeostasis and its loss proposed to contribute to disease, such as tumor invasiveness.⁹ In the classical description of CIL, migrating cells collapse their protrusions when they make contact, cease locomotion, then initiate new protrusions at free areas of their margins. This phenotype was first described in mesenchymal cells, such as cultured fibroblasts,¹⁰ and neural crest cells in vivo,¹¹ but it is also seen in epithelia.¹² CIL has been identified in many different tissues, and a variety of intracellular signaling pathways have been implicated in its mechanism.² However, less is known about the proximal steps that allow physical contact between the surfaces of cells to inhibit their locomotility.

In many cases, cell-cell contact is mediated by classical cadherin adhesion receptors. Cadherin adhesion begins with the homophilic trans-ligation of cadherin ectodomains presented on the surfaces of cells. This can then engage a variety of signaling pathways and regulators of the actin cytoskeleton, cellular processes with the potential to influence locomotility. So, it has been tantalizing to consider whether cadherins might also mediate CIL.¹² Indeed, expression of classical cadherins conferred CIL behavior on cadherin-null fibroblasts,¹³ while depletion of cadherins disrupted CIL in epithelia¹⁴ and in the neural crest.¹⁵ These observations establish an important role for cadherins in CIL.

However, cadherin ligation by itself is not sufficient to explain how contact inhibits epithelial migration. Of note, migrating epithelia are linked together by E-cadherin-based adherens junctions (AJs). Epithelial migration is inhibited when the leader cells at the front of migrating sheets encounter one another in a ‘head-on’ fashion (Figure 1A, green head-on contacts). But elsewhere in the monolayer, cells move despite forming cadherin adhesions with one another. This is clearly evident at the migrating front, where leader cells are connected to their neighbors by AJs at their sides (Figure 1A, blue side-side contacts). This implies that epithelial cells have a mechanism to evaluate the orientation of the cells with which they come into contact, discriminating head-on contacts that are perpendicular to the direction of movement from side-side contacts, which are parallel to migration. This could reflect additional surface signals or, more parsimoniously, some other important feature of cadherin adhesions found at head-on but not at side-side contacts. But what this feature might be is unknown.

Figure 1. Nascent adherens junctions downregulate lamellipodia as cell-cell contact inhibits epithelial migration.

(A) Schematic: migrating leader cells are connected to one another at side-side AJs (blue) where retrograde cortical flows are oriented in the same (parallel) direction. Retrograde cortical flows in overlapping lamellipodia (green) of head-on contacts between colliding leader cells have an opposite (antiparallel) orientation. Black solid arrows: migration direction; black dotted arrows: direction of F-actin flows.

(B) Contact inhibition of migration evidenced by a boundary (arrowheads) between cell populations expressing LifeAct-GFP (green) or LifeAct-TagRFPT (magenta). Dotted lines: free sheet edges.

(C) Antiparallel F-actin flow at head-on junctions between cells expressing LifeAct-GFP (green) or LifeAct-TagRFPT (magenta). Rose diagrams: F-actin flow direction by particle image velocimetry (PIV).

(D) Parallel F-actin flow at side-side junctions between cells expressing LifeAct-GFP (green) or LifeAct-TagRFPT (magenta). Representative image and kymograph (at white dotted line) from Video S2A. Wavy dotted lines: cell boundaries; blue lines: overlapping area; arrows and rose diagram: direction of F-actin flows. (E and F) Lamellipodial dynamics after contact between cells expressing LifeAct-GFP (green) or LifeAct-TagRFPT (magenta). (E) Representative image and kymograph (at white dotted line) from Video S1A. Wavy dotted lines: cell boundaries; *first contact. (F) Lamellipodia dynamics; n = 71 contacts from 3 independent experiments. Dots: means; light blue area: SEM.

(G and H) Nascent AJ assembly between cells expressing E-cadherin-GFP^CRISPR (green) and LifeAct-TagRFPT (magenta). (G) Representative image and kymograph (at white dotted line). Wavy dotted line: cell boundary; *first contact. (H) E-cadherin intensity; n = 34 contacts from 4 independent experiments. Dots: means; light blue area: SEM; blue line: sigmoidal fit (R² = 0.61).

(I and J) Lamellipodial dynamics after contact between α-catenin knockout (KO) cells expressing LifeAct-GFP (green) or LifeAct-TagRFPT (magenta). (I) Representative image and kymograph (at white dotted line). Wavy dotted lines: cell boundaries; *first contact. (J) Lamellipodial dynamics; WT (n = 34) or α-cat KO (n = 48) contacts from 3 independent experiments. Dots: means; light color areas: SEM.

(K and L) Contact inhibition of migration measured by cell interpenetration beyond boundaries between populations expressing LifeAct-GFP (green) or LifeAct-TagRFPT (magenta). (K) Representative images from Videos S3A and S3B. White dotted lines: cell boundaries. (L) Cell interpenetration; WT (n = 80) or α-cat KO (n = 80) contacts from 3 independent experiments.

***p < 0.001; Mann-Whitney U test. Data are means ± SD with individual data points indicated. See also Figure S1.

From first principles, we might predict that any mechanism for contact to inhibit migration must act upon the locomotor apparatus of the moving cells. Lamellipodia are major organs of locomotility that are often apparent in migrating epithelial cells, particularly leader cells at the front of migrating sheets.¹⁶ Lamellipodia extend the cell margin forward through branched actin assembly at their leading edges. As well, retrograde flow of cortical actin generates force for migration when engaged with adhesions at the cell substrate.^17,18 Inhibition of actin assembly or retrograde flow compromises cell locomotion, supporting the notion that lamellipodia are important for cell migration. We now report that cell-cell contact inhibits epithelial migration when newly forming cadherin adhesions downregulate the lamellipodia that are necessary for cell locomotion. This involves a tension-sensitive clutch in cadherin-bound α-catenin that is activated when the oppositely directed cortical flows found at head-on contacts become mechanically linked by cadherin adhesion.

Results

We used MCF7 mammary epithelial cells grown to confluence in silicon molds. These migrated toward one another as coherent sheets when the molds were removed, and they formed a single monolayer connected by AJs upon contact with one another (Figures 1B and S1A). However, the line of demarcation between the former populations remained evident for many hours after they had established contact. This was demonstrated by expressing differently tagged LifeAct transgenes in the two populations, which revealed a smooth boundary between the two populations, with minimal interpenetration of cells (Figure 1B). This suggested that epithelial locomotion was inhibited when the migrating populations collided with one another.

Nascent AJs inhibit leader cell lamellipodia

We then focused on the leader cells to analyze how contact inhibited migration. Leader cells within each individual population were connected to one another by AJs at their side-to-side contacts and also extended large dynamic lamellipodia in the direction of movement. These lamellipodia overlapped to form the first cell-cell contacts when the two migrating populations encountered each other head-on (Figures 1A and 1C; Video S1A). Lamellipodia were distinguished by prominent retrograde actin flows, as visualized by particle image velocimetry (PIV) of LifeAct movies (Figure 1C, left). These cortical flows were present both at side-side junctions between leader cells and in the overlapping head-on contacts. However, at overlapping head-on contacts the retrograde cortical flows of the lamellipodia were oriented in opposite directions, away from the line of contact (Figure 1C, right; Video S1A), whereas at side-side contacts the retrograde flows were oriented in the same direction (Figure 1D; Video S2A). For clarity, we describe oppositely directed flows as antiparallel and flows that are oriented in the same direction as parallel (illustrated schematically in Figure 1A).

The overlapping lamellipodia at head-on contacts displayed extension and retraction, becoming progressively less dynamic until they stopped after 30–40 min (Figures 1E, 1F, and S1B; Video S1A). Using MCF7 cells whose endogenous E-cadherin had been tagged with GFP by CRISPR-Cas9 gene editing (E-cadherin-GFP^CRISPR), we found that the downregulation of lamellipodial activity in these overlapping contacts coincided with the appearance of immobile E-cadherin-GFP^CRISPR deposits (Figures 1G and 1H), ceasing shortly after E-cadherin levels had plateaued (Figures 1F and 1H). We refer to these E-cadherin deposits as nascent AJs, as they grew laterally to eventually form the junctional border between the colliding cell populations (Figure S1A). This suggested that the assembly of nascent AJs might contribute to inhibiting lamellipodia.

To test this, we deleted a critical component of the cadherin adhesion complex, by CRISPR-Cas9 gene editing (α-catenin knockout [KO]) (Figures S1C–S1E). Confluent α-catenin KO cells expressed E-cadherin at their cell surface but did not assemble AJs (Figures S1E and S1F). Moreover, in migration assays, contact between α-catenin KO cells failed to inhibit lamellipodia (Figures 1I and 1J; Video S1B) and failed to inhibit locomotion. α-catenin KO populations interpenetrated after contact, forming finger-like projections rather than the smooth borders seen with wild-type (WT) cells (Figures 1K and 1L; Video S3). These findings reinforce the notion that nascent AJs allow contact to inhibit epithelial locomotion. They further suggest that a key might lie in downregulating the lamellipodial activity that is necessary for leader cells to migrate.

Adherent and free surface cadherins couple to cortical flow in migrating leader cells

Retrograde cortical flows can generate shear forces for cell migration.¹⁸ Therefore, to understand how assembly of AJs might inhibit lamellipodia, we analyzed the relationship between E-cadherin and cortical actin flow in the leader cells. As seen elsewhere,¹⁹ E-cadherin adhesions at the side-to-side contacts between the lamellipodia of leader cells displayed retrograde movement, consistent with their being coupled to the flowing cortex (Figure 2A; Video S2B).

Figure 2. Free and adherent E-cadherin is coupled to retrograde flow in migrating leader cells.

(A) E-cadherin-GFP^TG (green) at side-side contacts displays retrograde movement with parallel F-actin flows (LifeAct-TagRFPT, magenta). Representative image and kymograph (at white dotted line) from Video S2B.

(B and C) E-cadherin-GFP^TG (green) in leader cell lamellipodia co-expressing LifeAct-TagRFPT (blue). Surface E-cadherin clusters co-labeled in live cells with extracellular E-cadherin mAbs (magenta) appear white. E-cadherin-GFP^TG puncta (vesicles) that are not surface-labeled appear green. (B) Representative image and kymograph (at straight white dotted line) from Video S4. Wavy dotted line: free cell edge; yellow arrowheads and lines: E-cadherin-containing intracellular vesicles; red arrowheads and lines: E-cadherin surface clusters and tracks. (C) Velocity of intracellular vesicles (n = 94), surface E-cadherin clusters (n = 167), and F-actin flow waves (n = 136) from 3 independent experiments.

(D and E) Effect of blebbistatin, CK666, jasplakinolide (BCJ) treatment on retrograde flow of surface E-cadherin clusters (green) and F-actin (magenta). (D) Representative kymographs (at white dotted line; Figure S2H) and (E) velocity of E-cadherin clusters (n = 554) and F-actin flow waves (n = 410) from 3 independent experiments.

(F and G) α-catenin-F-actin affinity regulates coupling of surface E-cadherin clusters (labeled E-cadherin mAbs, blue) to retrograde cortical flow (LifeAct-TagRFPT, magenta) in leader cell lamellipodia expressing GFP-α-catenin^WT or GFP-α-catenin^A+ (green). (F) Representative images and kymographs (at white dotted lines). (G) Velocity of F-actin flow waves and E-cadherin/α-catenin clusters. F-actin (α-cat^WT) flow waves (n = 117), E-cad/α-cat^WT clusters (n = 49), F-actin (α-cat^A+) flow waves (n = 95), and E-cad/α-cat^A+ clusters (n = 31) from 2 independent experiments.

ns: not significant, ***p < 0.001; Kruskal-Wallis test. Data are means ± SD with individual data points indicated. See also Figures S2 and S3.

Puncta of E-cadherin were also evident on closer inspection of the free lamellipodia, visualized with either E-cadherin-GFP^CRISPR (Figure S2A) or with an exogenously expressed transgene (E-cadherin-GFP^TG) (Figure 2B). The greater signal strength of E-cadherin-GFP^TG revealed that the puncta comprised two subpopulations with distinct patterns of movement (Figures 2B, S2B, and S2C). One population showed rapid (~1.5 μm/min) movements in both anterograde and retrograde directions (Figures 2B and 2C, yellow; Video S4); often appeared to move along microtubules (Figure S2D); and were not accessible to extracellular E-cadherin monoclonal antibodies (mAbs) in live cells (Figure 2B, yellow). We consider these to be E-cadherin-containing intracellular vesicles.

In contrast, ~50% of E-cadherin-GFP^TG puncta formed when diffuse cadherin coalesced at the leading edges of the lamellipodia (Figure S2E). These showed a distinctive, slow processive retrograde movement (~0.2 μm/min; Figures 2B and 2C, red; Video S4) that made them readily distinguishable from the intracellular vesicles. They co-labeled in live cells with extracellular E-cadherin mAbs (Figure 2B, red) but were not due to mAb-induced cross-linking, because their size was unaffected by the mAbs, and this slowly moving subpopulation was also evident without mAb labeling (Figures S2B and S2C). Since they immunostained for β- and α-catenin (Figure S2F), we conclude that they represent cis-clusters of E-cadherin-catenin complexes (CCCs) that form independently of adhesive trans-ligation on the free cell surface, as previously reported.^20,21 (We refer to these as ‘free surface’ or ‘unligated’ cadherin clusters to emphasize that they do not represent trans-ligated complexes.)

Two observations indicated that the retrograde flow of free surface cadherin clusters reflected their association with cortical flow. First, movement of free surface cadherins was inhibited by drugs that reduce cortical flow. In lamellipodia, cortical flow is driven by the combination of Arp2/3-dependent actin assembly at the leading edge and myosin II-based contractility at the bases of lamellae.¹⁷ Consistent with this, we found that cortical flow was reduced by blebbistatin, CK666, and jasplakinolide (BCJ) when used individually (Figure S2G), but especially in combination at doses that did not overtly disrupt cytoskeletal integrity (Figures 2D, 2E, S2G, and S2H). This cocktail also effectively suppressed the retrograde flow of surface cadherin clusters (Figures 2D, 2E, and S2H). Second, unligated cadherin clusters moved more slowly than cortical actin flow, but this was accelerated by expressing an α-catenin^A+ transgene that is mutated to unfold the actin-binding domain (ABD) and increase its affinity for F-actin²² (Figures 2F, 2G, S3A, and S3B). However, the speed of cortical flow was not itself affected (Figures 2F and 2G). By implication, α-catenin coupled free surface cadherins to the flowing cortical cytoskeleton and indeed controlled the efficiency of coupling via its affinity with actin filaments. (We could not test the effect of α-catenin KO as this reduced free cadherin surface clustering [Figures S2I and S2J], making it difficult to measure cadherin movement.) Therefore, in leader cells both free cadherins on lamellipodia and cadherins at side-to-side contacts were coupled to retrograde cortical flow.

Coupling antiparallel cortical flows allows head-on E-cadherin adhesions to inhibit lamellipodia

The distinct movements of surface versus vesicular E-cadherin allowed us to define how nascent AJs assembled. Slow-moving free surface cadherin clusters incorporated into the immobile E-cadherin deposits that represent nascent AJs, leading to lateral extension of the AJs (Figure 3A). But we did not detect incorporation of rapidly moving puncta that would represent intracellular vesicles. This implied that nascent AJs assembled when free E-cadherin clusters that were coupled to cortical flow underwent trans-ligation as lamellipodia made contact with one another. Since E-cadherin in the nascent AJs was immobile, we then asked if contact affected retrograde flow in lamellipodia.

Figure 3. Nascent head-on adherens junctions couple antiparallel cortical flows together.

(A) Nascent head-on AJ assembly between migrating cells expressing LifeAct-TagRFPT (magenta) and E-cadherin-GFP^TG (green). Timelapse (top) and magnified detail (below). E-cadherin-GFP^TG clusters (blue arrowheads) display slow processive movements (red tracks) and fuse with already-static nascent AJs (yellow arrowheads). Black dotted lines: bottom cell edge.

(B) Persistent antiparallel F-actin flows at overlapping head-on contacts between WT cells expressing LifeAct-GFP (green) or LifeAct-TagRFPT (magenta). Representative image and kymograph (at white dotted line). Wavy dotted lines: cell boundaries; blue lines: overlap; arrows: direction of F-actin flows.

(C) Time from first contact to plateau of E-cadherin (n = 34 events from 4 independent experiments) and cessation of antiparallel cortical flows (n = 39 from 3 independent experiments) or of dynamic lamellipodia (n = 58 from 3 independent experiments).

(D) Antiparallel F-actin flows at head-on contacts between α-catenin KO cells expressing LifeAct-GFP (green) or LifeAct-TagRFPT (magenta). Representative image and kymograph (at white dotted line) from Video S1B. Wavy dotted lines: cell boundaries; blue lines: overlap; arrows: direction of F-actin flows.

(E) Schematic: heterologous head-on contacts between migrating WT and α-catenin KO cells. Black arrows: migration direction; colored arrows: F-actin flow directions; black dotted arrow: direction of cadherin-catenin cluster movement.

(F) Heterologous contacts between migrating WT cells (expressing LifeAct-iRFP670, blue) and α-catenin KO cells (expressing LifeAct-TagRFPT, magenta; and E-cadherin-GFP^TG, green) as in (E). Representative image and kymograph (at white dotted line) from Video S5. Note that only E-cadherin in the α-catenin KO cells is visible. Blue dotted line: WT cell edge; white arrowheads: anterograde movement (kymo) of E-cadherin clusters in α-catenin KO cells; red lines: tracks of moving clusters.

(G and H) Dynamic lamellipodia persist at heterologous contacts between WT cells (LifeAct-iRFP670, blue) and α-catenin KO cells (LifeAct-TagRFPT, magenta).

(G) Representative image and kymograph (at white dotted line). Wavy dotted lines: cell boundaries; *first contact. (H) Lamellipodia dynamics; n = 28 contacts from 2 independent experiments. Dots: means; light color area: SEM.

ns: not significant, *p < 0.1; Kruskal-Wallis test. Data are means ± SD with individual data points indicated. See also Figure S2.

To do this, we examined kymographs of LifeAct tagged with different fluorophores in the overlapping lamellipodia that constituted the first head-on contacts. Interestingly, we observed that the antiparallel flows persisted at these overlapping contacts, only ceasing when their lamellipodia became quiescent (Figures 3B and 3C; Video S1A), ~5–10 min after the peak of nascent AJ assembly (Figure 3C). This suggested that AJ assembly might inhibit lamellipodia by downregulating their underlying retrograde flows.

One potential mechanism for downregulation could involve nascent AJs linking the oppositely directed, antiparallel flows together. Because free cadherins were already engaged with the cortex, they would be predicted to physically couple flows together when they underwent adhesive trans-ligation. In support of this idea, we found that cell-cell contact failed to inhibit retrograde flow when E-cadherin was uncoupled from the cortex in α-catenin KO cells (Figure 3D; Video S1B). However, as α-catenin KO also compromises cadherin clustering (Figures S2I and S2J), it remained possible that some other aspect of cadherin signaling inhibited cortical flow.

To pursue this hypothesis, we studied heterologous cell-cell contacts, where E-cadherin was uncoupled from the cortex in only one cell of a contact (Figure 3E). We reasoned that coupling of flows would be disrupted if even one CCC of a trans-ligated pair were unable to bind F-actin. So, we analyzed the overlapping lamellipodia at head-on contacts when separate populations of WT and α-catenin KO cells migrated toward one another. Of note, E-cadherin clustering was restored to the α-catenin KO cells at these heterologous contacts (Figure 3F; Video S5). We interpret this to reflect trans-ligation between cadherin in the KO cells with E-cadherin clusters formed in the WT cells, evidence that cadherins were functional in the α-catenin KO cells. Furthermore, cadherin clusters in the α-catenin KO cells also demonstrated slow processive movement; but this occurred in an anterograde direction, implying that they were being pulled by retrograde flow in lamellipodia of the WT cells (Figure 3F; Video S5). This confirmed that cortical binding of cadherins was compromised in the α-catenin KO cells. Therefore, this setup preserved clusters of trans-ligated cadherins and their connection to cortical flow, but it disrupted the physical engagement of cortical flows across the contacting membranes that is expected when the cadherin adhesion system is intact.

Heterologous contacts failed to downregulate either retrograde cortical flows or the lamellipodial activity that they drive (Figures 3G and 3H). Of note, retrograde flow and lamellipodial activity at these heterologous contacts continued unabated in the WT cells, as well as in the α-catenin KO cells (Figures 3G and 3H). This cell-non-autonomous behavior indicated that trans-ligated cadherin complexes had to be coupled to the cortices on both sides of a nascent AJ to downregulate antiparallel flows. This supported the notion that coupling of cortical flows might allow the assembly of AJs to downregulate lamellipodia at head-on contacts.

Mechanical tension across cadherin adhesions correlates with the topology of cortical flows

But how might mechanical coupling of cortices allow AJs to inhibit retrograde flows at head-on contacts but not at the side-to-side AJs that connected leader cells? An important consideration lies in how the different topologies of flow might affect mechanical tension applied to the cadherins. Cortical flows can exert mechanical force on membrane proteins.^23,24 We confirmed this for trans-ligated cadherins by plating MCF7 cells onto soft PDMS substrata coated with E-cadherin ectodomain ligands (Figures 4A and S4). Here, cellular E-cadherin was immobilized by trans-ligation with the substrate ligand, and retrograde cortical F-actin flow persisted at the cell-substrate interface in the distal, cortical regions of the cells (Figure 4B). Microbeads embedded in the substrata underwent a periodic displacement that coincided exactly with waves of F-actin passing over the beads (Figures 4C and 4D), evidence that force from cortical flow was transmitted across trans-ligated cadherin complexes. Based on this test of principle, we envisaged that antiparallel flows would pull against one another, like a tug-of-war, when they were connected by AJs at head-on contacts. In contrast, parallel flows connected by AJs at side-side contacts would move in the same direction. This predicted that tension across the trans-ligated CCCs would be greater at head-on compared with side-side contacts (Figure 4E).

Figure 4. Cortical flow exerts mechanical tension on trans-ligated E-cadherin complexes.

(A and B) (A) F-actin (phalloidin) organization and (B) F-actin (LifeAct-GFP) dynamics in the cortical region of MCF7 cells spread on E-cadherin ectodomain-coated substrate. White dotted line: site of kymograph.

(C and D) Transmission of F-actin flow forces across trans-ligated cadherin complexes. (C) Cortical zone of a cell expressing LifeAct-GFP (green) spread on an E-cadherin-coated PDMS substrate (2–3 kPa) containing nanobeads (magenta). White dotted lines: site of kymographs. (D) F-actin dynamics and bead displacement in kymograph from (C1).

(E) Schematic: predicted mechanical tension at cadherin-catenin complexes in migration assays. Black solid arrows: migration direction; black dotted arrows: F-actin flow direction; green dots: cadherin-catenin clusters.

(F) Schematic: cadherin-catenin complex under low and high tension, with tension-sensitive α-catenin domains identified by α18 and VD7 Abs.

(G and H) Tension-sensitive conformational changes in α-catenin. (G) Representative side-side (top row) and head-on (bottom row) AJs, co-stained for α18 (M-domain) or VD7 (ABD) (green) and total α-catenin (magenta). Ratiometric images indicate domain-specific conformational changes. White arrowheads: examples of analyzed junctions. (H) Quantification of ratiometric α-catenin M-domain (α18) and ABD (VD7) staining. M-domain parallel (n = 38) and antiparallel (n = 55) flow junctions; ABD parallel (n = 41) and antiparallel (n = 52) flow junctions from 3 independent experiments.

(I–K) E-cadherin-GFP^TG fluorescence recovery after photobleaching in side-side versus head-on junctions. (I) Representative pre- and post-bleaching images, (J) normalized fluorescence intensity, and (K) E-cadherin-GFP^TG immobile fractions after photobleaching. Parallel flow junctions (n = 20), antiparallel flow junctions (n = 9) from 2 independent experiments. Dots: means; black lines: logistic fits (parallel F-actin flow, R² = 0.97; antiparallel F-actin flow, R² = 0.98).

***p < 0.001; Mann-Whitney U test. Data are means ± SD with individual data points indicated in (H) or means + SEM in (K). See also Figure S4.

To test this, we probed for tension-sensitive changes in the conformation of α-catenin. Mechanical tension opens the central M-domain of α-catenin²⁵ and also unfolds the autoinhibited α1 helix in the F-actin-binding site (ABD),²² conformational changes that are detectable with the α18 mAbs²⁵ and VD7 polyclonal Abs,²⁶ respectively (Figure 4F). Consistently, we found that staining for both α18 and VD7, corrected for total α-catenin, was more intense in nascent AJs at head-on contacts compared with side-to-side contacts (Figures 4G and 4H). Therefore, as predicted, tension across the cadherin complex was higher at the head-on contacts where AJs inhibited lamellipodia than at the side-side contacts. This was further supported by differences in E-cadherin stability at these two contacts. After photobleaching, E-cadherin-GFP fluorescence recovered more slowly at head-on contacts than at side-side contacts (Figures 4I–4K). As tension stabilizes E-cadherin,^27,28 this difference in cadherin stability was consistent with greater tension being exerted on cadherin complexes at head-on contacts. Of note, we did not detect cortical myosin II at these nascent AJs (not shown), suggesting that actin flow might be the direct source of mechanical force.

Modeling cadherin-actin interactions as a tension-activated clutch

We then considered how greater tension across the cadherin complexes at head-on contacts could inhibit retrograde flows and lamellipodial activity. One possibility was whether this increased the mechanical coupling of antiparallel flows by nascent AJs. The tension-sensitive conformational changes in α-catenin that we observed at head-on contacts increase its association with the actin cytoskeleton.²⁹ Opening of the ABD increases its affinity for F-actin to mediate a catch bond^22,30; and opening of the M-domain recruits vinculin, an additional F-actin-binding protein.²⁵ We focused on the effect of tension on the ABD, since a role for the M-domain was not evident in screening studies where an α-catenin mutant that cannot recruit vinculin (α-catenin^V—; Figures S3A and S3B) fully restored AJs to α-catenin KO cells (Figure S5A).

We then considered a simple model of mechanical antagonism, where trans-ligated cadherins at head-on contacts would allow antiparallel flows to pull against one another, causing their reciprocal inhibition. Decreased cortical flow would downregulate lamellipodial activity to reduce locomotility, as well as decrease the retrograde flow of cadherin to promote its retention in AJs. Reciprocal inhibition of flows would be further enhanced when tension increased the affinity of the α-catenin ABD for actin filaments, effectively increasing mechanical load in the tug-of-war between oppositely directed cortical flows. This model assumes that regulation of the α-catenin ABD can affect the response of cortical flow to mechanical load. We tested this principle by expressing the α-catenin^A+ mutant to increase affinity for F-actin and by analyzing cortical flow in cells plated on the uniform load of E-cadherin-coated glass. Cortical flow at the adhesive interface was decreased in cells expressing α-catenin^A+ (Figure 5A), confirming that conformational activation of the ABD can influence cortical flow in response to mechanical load.

Figure 5. Mechanical tension activates the F-actin-binding domain of α-catenin to promote cis-clustering of E-cadherin.

(A) Increasing α-catenin-actin affinity reduces cortical flow in response to the mechanical load of immobilized E-cadherin. F-actin flow (LifeAct-TagRFPT, magenta) was measured in the cortical regions of MCF7 cells expressing GFP-α-catenin^WT (n = 120) or GFP-α-catenin^A+ (n = 142) (green) and spread on E-cadherin ectodomain-coated substrates. Representative images and quantification from 3 independent experiments. Dotted white lines: site where flow rates were measured in kymographs.

(B) Clutch model of cell-cell adhesion at overlapping lamellipodia (see text for details). (BI) Pre-coupled cadherin-catenin complexes (CCCs, clutches) bind to antiparallel cortical flows with speed v. (BII) At each time step of the simulation, unbound CCCs can bind stochastically to F-actin on either side, according to binding rate k_on, and they become pulled by the antiparallel actin flows when bound on both sides. This progressively exerts force on CCCs, which also slows the F-actin flow. Force can cause bound CCCs to: (1) unbind from F-actin (with unbinding rate k_off) or (2) unfold α-catenin (with unfolding rate k_unf). (BIII) α-catenin unfolding increases bond stability (by decreasing k_off) and also leads to F-actin bundling. Because actin bundling will increase the likelihood of actin-catenin binding, this is introduced in the model by increasing the density of CCCs (clutches) available to bind. This results in more bound clutches, further building force and slowing actin flows.

(C) Simulations of the clutch model incorporating the effect of an α-catenin-actin catch bond alone on cortical flow. Top and bottom parts of graph show antiparallel velocities in each of the two cells making contact. Inset: the first minute of the simulation, indicating a minor, transient decrease in F-actin flow velocity.

(D and E) The α-catenin ABD promotes cis-clustering of E-cadherin. (D) Representative images of free E-cadherin cis-clusters (blue arrowheads), labeled with extracellular E-cadherin mAbs (magenta), in lamellipodia of leader cells expressing GFP-α-catenin mutants (green). (E) Cluster size measured by α-catenin fluorescence intensity in cells expressing α-catenin^WT (n = 200), α-catenin^A+ (n = 200), α-catenin^ΔβH (n = 180), and α-catenin^A+ΔβH (n = 200). Data from 4 inde-pendent experiments.

(F) Effect of GFP-α-catenin^WT, GFP-α-catenin^ΔβH, or GFP-α-catenin^A+ΔβH (green) expression on coupling of surface E-cadherin clusters (labeled with E-cadherin mAbs, blue) to retrograde cortical flow (LifeAct-TagRFPT, magenta). Representative images and kymographs (at white dotted line). Arrowheads: F-actin flow-driven retrograde movement of cadherin-catenin clusters.

(G) Predicted effect of applying tension to the α-catenin ABD on E-cadherin density in simulations of the enhanced clutch model. α-catenin^A+ represents the tension-stimulated state modeled by increasing the binding rate between cadherin clutches k_on and the amount of cadherin molecules added upon α1 helix unfolding d_add with respect to WT values (Table S1).

(H and I) Nascent AJ assembly in head-on contacts between cells expressing LifeAct-TagRFPT (magenta), E-cadherin-GFP^TG (green), and iRFP670-α-catenin mutants (blue). (H) Representative kymographs. Dotted lines: cell edges; *first contact. (I) E-cadherin fluorescence intensity. Cell-cell contact events: α-cat^WT (n = 16), α-cat^A+ (n = 14), α-cat^ΔβH (n = 18), and α-cat^A+ΔβH (n = 17) from 3 independent experiments. Dots: means; Light color areas: SEM; lines: sigmoidal fits (α-cat^WT, R² = 0.67; α-cat^A+, R² = 0.53).

ns: not significant, ***p < 0.001; Mann-Whitney U test in (A); Kruskal-Wallis test in (E). Data are means ± SD with individual data points indicated. See also Figures S3 and S5.

To quantitatively evaluate this scenario for mechanical antagonism, we then modeled the adherent CCC as a mechanosensitive clutch where tension alters the F-actin-binding properties of the α-catenin ABD (Figures S6A–S6C; Tables S1 and S2; Methods S1). We define a ‘clutch’ as a pre-coupled complex containing E-cadherin and α-catenin, trans-ligated to an equivalent complex on the other cell once the cells make contact. Clutches can bind to the actin cytoskeleton on both sides (Figure 5BI). Upon head-on contact, initial adhesions are thus composed of a given number of clutches. Clutches get submitted to force as they bind to actin and are pulled in opposite directions from both cells due to the antiparallel actin flows. We then incorporated two effects of force exerted on clutches based on change in the ABD.^22,30 First, force can unfold the α1 helix, with an unfolding rate that depends on force as a slip bond. Second, unfolding of the α1 helix increases the affinity of the α-catenin ABD-F-actin bond. Once clutches unbind, they can rebind again according to a given binding rate (Figures 5BI and 5BII).

However, our qualitative predictions were not supported in numerical simulations of the model. Although small decreases in actin flows appeared (Figures 5C and S5B), these occurred over timescales of seconds and did not yield the minutes-scale downregulation of flow or adhesion growth that we observed experimentally. This was because, based on the reported k_off for α-catenin and F-actin,³⁰ even in their high-affinity state clutches were predicted to disengage from F-actin on seconds timescales, so that force would not evolve and increase over minutes timescales (Figure S5B). By tuning model parameters (e.g., by decreasing myosin contractility levels), the model could be adjusted to predict a substantial rather than small reduction in actin flows, but this would always occur in a timescale of seconds and not minutes. Further, the model did not predict the adhesion growth observed in experiments. This implied that enhanced binding to F-actin was not sufficient to explain how the application of tension to α-catenin could downregulate lamellipodia and promote adhesion growth.

Tensile unfolding of the α-catenin ABD promotes the growth of E-cadherin cis-clusters and AJ assembly

We therefore considered other potential mechanisms that could be engaged when mechanical tension is applied to the ABD. In addition to increasing its affinity for F-actin, unfolding of the α1 helix also reveals a β-helical (βH) motif that promotes dimerization of the ABD and potentially F-actin bundling.²² Both these mechanisms could increase the number of CCCs that are incorporated into clusters. Force-induced actin bundling is also observed in the context of cell-matrix adhesion,³¹ and its minutes timescale matches well what we observed for AJs.

To test this possibility, we asked how unfolding of the ABD influenced E-cadherin cis-clustering independent of other signals that are engaged by cadherin trans-ligation. Therefore, we measured how mutations in the α-catenin ABD affected unligated E-cadherin cis-clusters on the free surfaces of cells. First, we analyzed the α-catenin^A+ mutant²²: as this unfolds the ABD, we reasoned that it should mimic the effect of applying tension even without mechanical loading by homophilic trans-ligation. Indeed, expressing α-catenin^A+ in α-catenin KO cells significantly increased the fluorescence intensity of the clusters (Figures 5D and 5E), consistent with enhanced recruitment of CCCs. Then, we further mutated α-catenin^A+ to delete the βH motif (α-catenin^A+ΔβH; Figures S3A and S3B). This abrogated the ability of α-catenin^A+ to increase cluster size (Figures 5D and 5E), implying that the βH motif was necessary for α-catenin to enhance cadherin cis-clustering in response to tension. However, deleting the βH motif alone (α-catenin^ΔβH; Figures S3A and S3B) did not affect the basal levels of free clustering that were seen in cells (Figures 5D and 5E), although clustering was largely abolished by either α-catenin KO or mutations of the actin-binding residues within the ABD (α-catenin^A—; Figures S3A, S3B, and S5C). Mutants lacking the βH motif were capable of binding to the cortex, as they displayed slow retrograde flow at the free surface (Figure 5F). This implied that cadherin-catenin clustering in our experiments reflected the action of two mechanisms: a basal requirement for α-catenin to bind F-actin and an enhancement when tension unfolds the ABD to reveal the βH motif.

We then modified our clutch model so that the application of tension to cadherin clutches induced not only a higher resistance to force of α-catenin-actin bonds but also to the cis-growth of clusters (Figures S6A–S6C; Tables S1 and S2; Methods S1). This was modeled by increasing the number of clutches (i.e., CCCs available to bind actin) upon α-catenin ABD unfolding (Figures 5BI–5BIII). Simulations of this ‘enhanced’ clutch model showed that force now increased (Figure S5D), and cadherin junctions assembled and grew when tension was applied, with a sigmoidal profile and minutes-scale time course that closely resembled what we observed experimentally when AJs first formed (compare Figure 5G with Figure 1H). The assembly of AJs and evolution of force in the model were further accelerated if we increased the rate of clutch addition and the affinity of clutches to actin, thereby simulating the α-catenin^A+ mutant (Figures 5G and S5E). This predicted that tensile activation of α-catenin could support the assembly of nascent AJs, so long as the ABD was induced to promote cis-clustering of E-cadherin.

These predictions of the model were supported experimentally when we measured the kinetics of nascent AJ assembly. First, nascent AJ assembly was substantially compromised when we deleted just the βH motif (α-catenin^ΔβH) (Figures 5H and 5I). This resembled the output of our original model, where AJs failed to assemble if tension did not enhance cis-clustering of E-cadherin. Second, AJ assembly was further increased when we reconstituted α-catenin KO cells with α-catenin^A+, but this enhancement was lost when we deleted the βH motif (α-catenin^A+ΔβH) (Figures 5H and 5I). Therefore, enhanced cadherin cis-clustering allows tensile activation of the α-catenin ABD to mediate nascent AJ assembly.

Enhanced cadherin-catenin cis-clustering allows mechanical antagonism to inhibit cell locomotion

Importantly, simulations also showed that cortical flow became downregulated in a sustained fashion when cis-clustering was incorporated into our enhanced model of reciprocal mechanical inhibition (Figures 6A and 6B). Moreover, this occurred on a timescale consistent with what we observed experimentally. We predicted that this would downregulate lamellipodial activity and allow contact to inhibit epithelial migration.

Figure 6. An extended bifunctional α-catenin-actin clutch allows contact to inhibit epithelial cell migration by downregulating lamellipodia.

(A and B) Predicted effect on cortical flow (A) when tension is applied to the α-catenin ABD in the extended clutch model or (B) when the α-catenin ABD is constitutively unfolded in the α-catenin^A+ mutant. Top and bottom of the graph show antiparallel flow velocities in each of the two cells making contact.

(C and D) Cadherin cis-clustering by the α-catenin ABD is necessary for contact to downregulate lamellipodial activity. (C) Representative images and kymographs (at white dotted lines) in cells expressing LifeAct-GFP (green) or LifeAct-TagRFPT (magenta) from Video S6A. Wavy dotted lines: cell boundaries; *first contact. (D) Lamellipodial dynamics; α-cat^WT (n = 51 events), α-cat^A+ (n = 29), α-cat^ΔβH (n = 40), and α-cat^A+ΔβH (n = 50) from 3 independent experiments. Dots: means; light color areas: SEM.

(E) Cadherin cis-clustering by the α-catenin ABD is necessary for contact to downregulate retrograde cortical flow in cells expressing LifeAct-GFP (green) or LifeAct-TagRFPT (magenta). All cells express α-catenin^WT or α-catenin^ΔβH. Representative images and kymographs (at white dotted lines). Wavy dotted lines: cell boundaries; blue lines: overlapping area; arrows: direction of F-actin flows; *first contact.

(F and G) Contact inhibition of migration measured by cell interpenetration beyond boundaries of cells expressing LifeAct-GFP (green) or LifeAct-TagRFPT (magenta). All cells express either α-catenin^WT, α-catenin^A+, α-catenin^ΔβH, or α-catenin^A+ΔβH. (F) Representative images from Video S6B. White dotted lines: cell boundaries. (G) Cell interpenetration; α-catenin^WT (n = 32), α-catenin^A+ (n = 84), α-catenin^ΔβH (n = 32), and α-catenin^A+ΔβH (n = 60) contacts from 3 independent experiments.

ns: not significant, ***p < 0.001; Kruskal-Wallis test. Data are means ± SD with individual data points indicated. See also Figure S5.

To test this, we examined how mutating the α-catenin ABD influenced the response of lamellipodia to cell-cell contact. First α-catenin KO cells were reconstituted with α-catenin^A+ to mimic the effect of an enhanced mechanical stimulus that unfolds the ABD. This accelerated the downregulation of lamellipodial activity when cells made contact with one another (Figures 6C and 6D; Video S6A). However, it was necessary for the unfolded ABD to be capable of increasing cis-clustering, as the effect was lost if the βH motif was also deleted (α-catenin^A+ΔβH; Figures 6C and 6D; Video S6A). Furthermore, deleting the αH motif alone (α-catenin^ΔβH) abolished the capacity for cell-cell contact to inhibit lamellipodia (Figures 6C and 6D; Video S6A). Consistent with our model of reciprocal inhibition, the ability of contact to inhibit retrograde flow was also disrupted when we deleted the βH motif (Figure 6E). Therefore, enhanced cis-clustering allows tensile activation of the α-catenin ABD to inhibit cortical flow and downregulate lamellipodial activity.

Finally, we found that colliding α-catenin^ΔβH populations failed to inhibit cell migration. α-catenin^ΔβH populations continued to interpenetrate one another upon contact, rather than forming the relatively smoothly defined borders seen in WT populations (Figures 6F and 6G; Video S6B). The α-catenin^A+ mutant did not increase the robust degree of contact inhibition already displayed by the WT cells; however, its ability to mediate contact inhibition was also disrupted by deleting the βH motif (α-catenin^A+-ΔβH) (Figures 6F and 6G; Video S6B). Altogether, these findings indicate that tensile activation of α-catenin ABD allows cell-cell contact to inhibit epithelial migration by downregulating lamellipodia.

Discussion

In this study, we sought to understand how migrating epithelial cells recognize when to stop moving when they come into contact with one another. We now propose the following model (Figure 7). The trigger for CIL occurs when E-cadherin complexes on the free surfaces of migrating cells trans-ligate and mechanically connect together the antiparallel, retrograde cortical flows of colliding lamellipodia. Coupled antiparallel flows apply mechanical tension to the α-catenin ABD, increasing its affinity for actin filaments and triggering the recruitment of further CCCs. Adhesion growth enhances mechanical coupling between antiparallel cortical flows, leading to their reciprocal downregulation. Loss of cortical flow then compromises the lamellipodial activity of leader cells to inhibit epithelial migration. The orientation of flows is a key factor in this model, as antiparallel flows generate greater tension than parallel flows. This antiparallel orientation reflects, in turn, the geometry of head-on collisions.

Figure 7. α-catenin clutch-induced adherens junction assembly inhibits lamellipodial activity and cell migration; schematic overview.

(I) Free surface E-cadherin clusters are coupled to F-actin retrograde flow through β- and α-catenin. (II) E-cadherin trans-ligation allows the antiparallel F-actin flows to apply mechanical load, opening the α-catenin F-actin-binding domain (ABD). (III) This conformational change increases the α-catenin binding affinity for F-actin and causes cis-clustering through the βH-domain, resulting in the bundling of F-actin filaments and growth of the newly formed AJs. (IV) Cis-clustering of cadherin-catenin complexes inhibits cortical F-actin flow by reciprocal mechanical loading to (V) inhibit lamellipodia and antagonize further cell migration.

Several lines of evidence argue that tensile activation of the α-catenin ABD plays a key role in allowing cadherin adhesion to inhibit locomotion. First, mechanical tension across α-catenin was greatest at the head-on contacts that inhibit epithelial migration, compared with the side-side AJs. Second, although several scaffolding domains in α-catenin respond to mechanical tension,^22,25,30,32 mimicking tensile unfolding of the ABD with the α-catenin^A+ mutant was sufficient to enhance AJ assembly and lamellipodial inhibition. Third, CIL required the ability of the tension-activated ABD to promote cis-clustering of cadherin complexes via its βH motif.²² The ability of contact to inhibit migration and lamellipodial activity was abolished when the βH motif of the ABD was deleted in either WT α-catenin or the tension-mimicking α-catenin^A+ mutant. Thus, although increased affinity for actin filaments is the best-understood consequence of applying tension to the α-catenin ABD,^22,30,33 cis-clustering of CCCs is a second effect that is essential for contact to inhibit epithelial locomotion. Our results do not exclude the possibility that other α-catenin partners contribute at later stages of contact inhibition. Nonetheless, they argue strongly that activation of the ABD represents the first, key stage in the process.

Ultimately, cadherin adhesions must inhibit epithelial migration by acting on some critical element in the locomotor apparatus of the cells. Retrograde flow is an important driver of cell locomotility, which generates force for cell movement.^17,18 Therefore, it was noteworthy that retrograde cortical flow became downregulated at head-on AJs during CIL, and this involved tensile activation of the α-catenin ABD and its βH motif. How could activation of the α-catenin ABD allow cadherin adhesion to downregulate cortical flow? One parsimonious explanation is that antiparallel cortical flows reciprocally inhibit one another by mechanical loading. The capacity for enhanced mechanical loading to inhibit flow was demonstrated in principle by finding that expression of α-catenin^A+ downregulated the cortical flows that cells made at adhesions with cadherin-coated substrata. It was reinforced by the observation that flows were not inhibited at heterologous contacts between WT and α-catenin KO cells, a situation where antiparallel flows could not act upon one another because one side of the trans-ligated CCC was unable to bind F-actin. Increased affinity of α-catenin for F-actin would be predicted to mechanically couple antiparallel cortical flows at AJs. However, our computational modeling suggested that this mechanism alone was not enough to inhibit cortical flow on the timescales that we observed experimentally. Instead, sustained inhibition of flow required that the tension-activated ABD also promote adhesion growth, as we confirmed experimentally by deleting the βH motif. This suggests that enhanced cadherin cis-clustering is necessary to initiate CIL, because it increases the mechanical coupling of antiparallel flows for sustained reciprocal inhibition.

Our mechanical model does not exclude contributions from other tension-sensitive signaling events. For example, mature AJs in MCF7 and other epithelial cells are characterized by a contractile cortex.^27,34 Although myosin II was not recruited into nascent AJs until after the initial phase of lamellipodial inhibition, it is possible that the subsequent assembly of a myosin-enriched cortex replaces the tensile forces of retrograde flow to sustain CIL. As well, Arp2/3-mediated actin nucleation drives leading edge protrusion and generates actin for cortical flow in lamellipodia.^17,35,36 Therefore, it is possible that branched actin nucleation is also altered when nascent AJs inhibit lamellipodial activity. Cadherin adhesion has complex effects on Arp2/3, the principal actin nucleator at the leading edges of lamellipodia. VE-cadherin junctions are reported to downregulate Arp2/3-based actin assembly,³⁷ whereas Arp2/3 can also be recruited to E-cadherin to nucleate actin at AJs.³⁸ Interestingly, at the end of our simulations, actin flows stabilized at very low speeds, thus indicating a regime in which lamellipodial dynamics no longer dominate. Whether tensional signals generated by coupling antiparallel flows can also regulate Arp2/3 is an interesting question for future research.

In conclusion, we propose that the α-catenin ABD is a bifunctional clutch that allows the cadherin system to interpret tensional stimuli of cortical flows to inhibit epithelial cell locomotion. There is an attractive simplicity to this mechanical model. It has the capacity to exert rapid effects: since cadherins already interacted with the cortex on the free surfaces of cells, they were poised to couple flows upon contact. Furthermore, this model has the potential for subcellular spatial specificity, as flows will only be inhibited at cadherin adhesions that are exposed to sufficient tension, as occurred at head-on contacts but not at the side-side AJs that hold migrating cells together. This emphasizes the key role that topology of cortical flow plays in our model. We propose that the antiparallel orientation of flow that is found at head-on contacts is responsible for the tensile signal that activates the α-catenin ABD clutch when cadherins first trans-ligate upon cell-cell contact. As cadherin adhesions grow and increase mechanical coupling of the cortices, that tug-of-war will also generate the loading that eventually downregulates flow. Without excluding contributions for other contact-based cell surface signals to CIL,³⁹ this mechanism is ideally placed for cells to make early-immediate responses to contact by coupling the molecular apparatus of adhesion to the cytoskeletal driver of lamellipodia. Since patterns of cortical flow reflect the overall polarity of migrating cells, this mechanosensitive apparatus allows migrating epithelia to infer the orientation of cells with which they collide, discriminating head-on from side-to-side contacts, and alter locomotion accordingly.

Limitations of the study

We used the MCF7 system for our experiments, a well-characterized model of epithelial migration where AJs inhibit locomotility when they assemble at head-on collisions. This system also reproduces many features of locomoting epithelia, including dynamic lamellipodia and a requirement for AJs to also mediate sheet cohesion. It will now be important to test how the model that we have identified may operate in other epithelia, notably in vivo models of sheet migration.

Star⋆Methods

Key Resources Table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Mouse monoclonal anti E-cadherin ectodomain	BD Biosciences	Cat#:563571; RRID: AB_2738283
Mouse monoclonal anti β-catenin	BD Biosciences	Cat#:610154; RRID: AB_397555
Mouse monoclonal anti E-cadherin	Gift from Dr. M. Takeichi, RIKEN Center for Biosystems Dynamics Research	N/A
Rabbit polyclonal anti GAPDH	R&D Systems	Cat#:2275-PC-100; RRID: AB_2107456
Rabbit polyclonal anti α-catenin	Thermo Fisher	Cat#:71-1200; RRID: AB_2533974
Rabbit polyclonal anti α-catenin	Gift from Dr. B. M. Gumbiner, Seattle Children’s Research Institute	N/A
Rabbit polyclonal anti α-catenin (VD7)	Vestweber lab, Max Planck Institute for Molecular Biomedicine	N/A
Rat monoclonal anti E-cadherin	Abcam	Cat#:ab11512; RRID: AB_298118
Rat monoclonal anti E-cadherin	Thermo Fisher	Cat#:13-1900; RRID: AB_2533005
Rat monoclonal anti α-catenin (α18)	Gift from Dr. A. Nagafuchi, Kumamoto University	N/A
Goat anti mouse, rabbit, rat Alexa-Fluor-405	Thermo Fisher	Cat#:A-31553; RRID: AB_221604 Cat#:A-48254; RRID: AB_2890548 Cat#:A-48261; RRID: AB_2890550
Goat anti mouse, rabbit, rat Alexa-Fluor-488	Thermo Fisher	Cat#:A-11001; RRID: AB_2534069 Cat#:A-11008; RRID: AB_143165 Cat#:A-11006; RRID: AB_2534074
Goat anti mouse, rabbit, rat Alexa-Fluor-594	Thermo Fisher	Cat#:A-11032; RRID: AB_2534091 Cat#:A-11037; RRID: AB_2534095 Cat#:A-48264; RRID: AB_2896333
Goat anti mouse, rabbit, rat Alexa-Fluor-647	Thermo Fisher	Cat#:A-21236; RRID: AB_2535805 Cat#:A-21245; RRID: AB_2535813 Cat#:A-48265; RRID: AB_2896334
Goat anti mouse HRP	Bio-Rad	Cat#:1706516; RRID: AB_11125547
Goat anti rabbit HRP	Bio-Rad	Cat#:1706515; RRID: AB_11125142
Bacterial and virus strains
NEB 10-beta Competent E.coli	New England Biolabs	Cat#:C3019H
Stellar Competent E.coli	Clontech	Cat#:636766
Chemicals, peptides, and recombinant proteins
Lipofectamin 3000	Thermo Fisher	Cat#:L3000015
Puromycin	Sigma-Aldrich	Cat#:P8833
G418	Santa Cruz	Cat#:108321-42-2
Lenti-X concentrator	Clontech	Cat#:631232
Alexa Fluor® 488, 594, 647 Phalloidin	Thermo Fisher	Cat#:A12379 Cat#:A12381 Cat#:A22287
SiR-Tubulin	Cytoskeleton	Cat#:CY-SC002
Para-nitroblebbistatin	Optopharma	Cat#:DR-N-111
CK666	Sigma-Aldrich	Cat#:SML0006
Jasplakinolide	Merck	Cat#:420107
Protein A	Thermo Fisher	Cat#:21181
hE-cadherin/Fc	SinoBiological	Cat#:10204
CY-52-276 Part A	Dow Corning Toray	Cat#:CY 52-276 A
CY-52-276 Part B	Dow Corning Toray	Cat#:CY 52-276 A
3-aminopropyl trimethoxysilane	Sigma Aldrich	Cat#:281778
Crystalline Trypsin	Sigma Aldrich	Cat#:T-0303
Chemiluminescent Substrate	Thermo Fisher	Cat#:34579
ProLong Gold with DAPI	Cell Signaling	Cat#:8961
ProLong Gold without DAPI	Cell Signaling	Cat#:9071
Experimental models: Cell lines
MCF7	ATCC	HTB-22
HEK293T	ATCC	CRL-3216
MCF7 LifeAct-TagRFPT	This study	N/A
MCF7 LifeAct-TagRFPT, E-cadherin-GFPCRISPR	This study	N/A
MCF7 LifeAct-TagRFPT, E-cadherin-GFPTG	This study	N/A
MCF7 LifeAct-GFP	This study	N/A
MCF7 α-catenin KO, LifeAct-TagRFPT	This study	N/A
MCF7 α-catenin KO, LifeAct-GFP	This study	N/A
MCF7 α-catenin KO, LifeAct-TagRFPT, E-cadherin-GFPTG	This study	N/A
MCF7 LifeAct-iRFP670	This study	N/A
MCF7 LifeAct-TagRFPT, iRFP670-α-cateninWT	This study	N/A
MCF7 LifeAct-GFP, iRFP670-α-cateninWT	This study	N/A
MCF7 LifeAct-TagRFPT, iRFP670-α-cateninA+	This study	N/A
MCF7 LifeAct-GFP, iRFP670-α-cateninA+	This study	N/A
MCF7 LifeAct-TagRFPT, iRFP670-α-cateninΔβH	This study	N/A
MCF7 LifeAct-GFP, iRFP670-α-cateninΔβH	This study	N/A
MCF7 LifeAct-TagRFPT, iRFP670-α-cateninA+ΔβH	This study	N/A
MCF7 LifeAct-GFP, iRFP670-α-cateninA+ΔβH	This study	N/A
MCF7 LifeAct-TagRFPT, E-cadherin-GFPTG, iRFP670-α-cateninWT	This study	N/A
MCF7 LifeAct-TagRFPT, E-cadherin-GFPTG, iRFP670-α-cateninA+	This study	N/A
MCF7 LifeAct-TagRFPT, E-cadherin-GFPTG, iRFP670-α-cateninΔβH	This study	N/A
MCF7 LifeAct-TagRFPT, E-cadherin-GFPTG, iRFP670-α-cateninA+ΔβH	This study	N/A
Oligonucleotides
CTNNA1 targeting sequence 5’GAAATGACTGCTGTCCATGC‘3	This study	N/A
Recombinant DNA
PX459 CTNNA1 (α-catenin CRISPR KO)	This study	N/A
PLViP-LifeAct-GFP	This study	N/A
PLViP-LifeAct-TagRFPT	This study	N/A
PLViP-LifeAct-iRFP670	Gift from Dr. J van Buul, Sanquin Research and Landsteiner Laboratory	N/A
pLViN-E-cadherin-GFPTG	This study	N/A
GFP-α-cateninWT	This study	N/A
GFP-α-cateninA+	This study	N/A
GFP-α-cateninA-	This study	N/A
GFP-α-cateninV-	This study	N/A
GFP-α-cateninΔβH	This study	N/A
GFP-α-cateninA+ΔβH	This study	N/A
PLViN-iRFP670-α-cateninWT	This study	N/A
PLViN-iRFP670-α-cateninA+	This study	N/A
PLViN-iRFP670-α-cateninΔβH	This study	N/A
PLViN-iRFP670-α-cateninA+ΔβH	This study	N/A
Software and algorithms
ImageJ 1.52	NIH	https://imagej.nih.gov/ij/download.html
Prism 9.0.0	Graphpad	https://www.graphpad.com/scientific-software/prism/
Matlab 9.9	Mathworks	https://www.mathworks.com/products/matlab.html
LASX	Leica	https://www.leica-microsystems.com/products/microscope-software/p/leica-las-x-ls/
Fusion	Andor	https://fusion.help.andor.com/display/fusionum/Home
Zen Blue 3	Zeiss	https://www.zeiss.com/microscopy/int/products/microscope-software/zen.html
Other
4-well silicone inserts	Ibidi	Cat#:F8807
Carboxylated fluorescent beads	Thermo Fisher	Cat#:80469

Resource Availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Alpha Yap: a.yap@uq.edu.au.

Materials availability

All reagents generated in this study will be made available upon request to the lead contact.

Experimental Model And Study Participant Details

Cell culture and transfection

MCF7 cells were obtained from ATCC (HTB-22) and cultured in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% fetal bovine serum (FBS) (GE Healthcare Australia), 100 units/ml penicillin/streptomycin (pen/strep) (Life Technologies Australia) and 1 mM CaCl2. HEK293T cells were cultured in DMEM with L-glutamine and sodium pyruvate (Invitrogen), supplemented with 10% FBS and 100 units/ml Pen/Strep. All cells were cultured at 37°C in 5% CO2. For migration assays, cells were seeded in silicone inserts (Ibidi). Inserts were removed 24 hours after seeding.

Lipofectamin 3000 (Thermo Fisher) was used to transfect MCF7 cells for ectopic expression of transgenes and CRISPR gene-editing constructs. PEI 2500 (BioScientific) was used to transfect HEK293T cells for lentiviral packaging. Both methods were performed according to manufacturer’s protocols.

Stable cell lines

MCF7 E-cadherin-GFP knock-in cells were generated using CRISPR-Cas9 technology and were described before.⁴⁰ MCF7 α-catenin knockout cells were generated using CRISPR-Cas9 technology. Cells were transfected with the pSpCas9(BB)-2A-Puro (PX459) vector bearing the appropriated sequence (5’-GAAATGACTGCTGTCCATGC-’3) to target CTNNA1 (α-catenin). 24 hours after transfection, cells were subjected to 0,8 μg/ml puromycin for 48 hours. After selection, single clones were grown and α-catenin expression was assessed using western blot and IF staining. Genomic sequencing was performed to confirm knockout.

MCF7 cells expressing LifeAct-GFP, LifeAct-TagRFPT, LifeAct-iRFP670, E-cadherin-GFP^TG, iRFP670-α-catenin^WT, iRFP670-α-catenin^A+, iRFP670-α-catenin^ΔβH and iRFP670-α-catenin^A+ΔβH were generated using lentiviral transduction. Lentiviral constructs were packaged into lentivirus in HEK293T cells by means of third generation lentiviral packaging plasmids. Lentivirus containing supernatant was harvested on days 2 and 3 after transfection, concentrated using a Lenti-X concentrator (Clontech #631232) and used to transduce target cells. Lentiviral-transduced MCF7 cells were subjected to either 0,5 μg/ml puromycin or 400 μg/ml G418 for 72 hours. Subsequently, cells were maintained in either 0,25 μg/ml puromycin or 50 μg/ml G418.

Method Details

DNA constructs

pLVX-IRES-Puro (PLViP) and pLVX-IRES-Neo (PLViN) were used for lentiviral packaging. PLViN was constructed by replacing the EF1a promotor for a CMV promotor in pLV-EF1a-IRES-Neo (Addgene #85139). PLViP-LifeAct-GFP and PLViP-LifeAct-TagRFPT were constructed by replacing iRFP670 for GFP- and TagRFPT-coding sequences in PLViP-LifeAct-iRFP670 (Kindly provide by Dr. J. van Buul, Sanquin Research and Landsteiner Laboratory, The Netherlands). pLViN-E-cadherin-GFP^TG was constructed by replacing YFP for a GFP-coding sequence in N1-E-cadherin-YFP. Subsequently, E-cadherin-GFP was cloned into PLViN. GFP-α-catenin^WT was constructed by cloning a GFP-coding sequence into acat 1–906 (Addgene #24194). GFP-α-catenin^A+ (RAIM670-733GSGS), GFP-α-catenin^A- (L785A I792A V796A) and GFP-α-catenin^V- (L344P) were constructed by Gene Universal using GFP-α-catenin^WT. GFP-α-catenin^ΔβH (ΔE795-810S) was constructed using GFP-α-catenin^WT. GFP-α-catenin^A+ΔβH (RAIM670-733GSGS, ΔE795-810S) was constructed using GFP-α-catenin^A+. All α-catenin mutants were properly incorporated into adherens junctions after expression in WT MCF7 cells, indicating β-catenin binding was not affected (Figure S3B). PLViN-iRFP670-α-catenin^WT, PLViN-iRFP670-α-catenin^A+, PLViN-iRFP670-α-catenin^ΔβH and PLViN-iRFP670-α-catenin ^A+ΔβH were constructed by replacing GFP for an iRFP670-coding sequence in the GFP-α-catenin constructs. Subsequently, iRFP670-α-catenin constructs were cloned into PLViN. All constructs were confirmed by sequencing.

Antibodies, drugs and chemicals

Mouse monoclonal antibodies against E-cadherin ectodomain (#563571) (used for IF) and β-catenin (#610154) were purchased from BD Biosciences. Mouse monoclonal antibody against E-cadherin (used for Western) was kindly provided by Dr. M. Takeichi (Riken, Japan). Rabbit polyclonal antibody against GAPDH (#2275-PC-100) was purchased from R&D Systems. Rabbit polyclonal antibody against α-catenin (#71-1200) (used for Western blot) was purchased from Thermo Fisher. Rabbit polyclonal antibody against α-catenin (used for IF) was kindly provided by Dr. B. M. Gumbiner (Seattle Children’s, USA). Rat monoclonal antibodies against E-cadherin were purchased from Abcam (#ab11512) (used for IF) and Thermo Fisher (#13-1900) (used for IF). Rat monoclonal antibody against α-catenin M-domain (α18) was kindly provided by Dr. A. Nagafuchi (Kumamoto University, Japan). Rabbit polyclonal antibody against α-catenin ABD (VD7) has been described before.²⁶

Alexa-Fluor-405- (#A-31553, #A-48254, #A-48261), Alexa Fluor-488- (#A-11001, #A-11008, #A-11006), Alexa-Flour-594- (#A-11032, #A-11037, #A-48264) and Alexa Fluor-647- (#A-21236, #A-21245, #A-48265) conjugated goat antibodies against mouse, rabbit and rat were purchased from Thermo Fisher. For Western blot, goat anti-mouse-HRP conjugate (#1706516) and goat anti-rab-bit-HRP conjugate (#1706515) were purchased from Bio-Rad.

Alexa Fluor® 488- (A12379), 594- (A12381) and 647- (A22287) conjugated phalloidin were purchased from Thermo Fisher. SiR-Tubulin (CY-SC002) was purchased from Cytoskeleton and used at a concentration of 1 μM. Cells were incubated with SiR-Tubulin for 1 hour prior to imaging. Para-nitroblebbistatin (DR-N-111) was purchased from Optopharma and used at a concentration of 10 μM. CK666 (SML0006) was purchased from Sigma-Aldrich and used at a concentration of 50 μM. Jasplakinolide (420107) was purchased from Merck and used at a concentration of 100 nM. In order to stop cortical actin flow, cells were treated with Para-nitroblebbistatin for 1 hour, following by 2-hour treatment with Para-nitroblebbistatin, CK666 and Jasplakinolide (BCJ).

E-cadherin-coated substrate assay

Coverslips were covered with 25 μg/ml Protein A (Thermo Fisher, #21181) and incubated at room temperature (RT) for 3 hours. After this step, the coverslips were kept wet. Coverslips were washed with PBS and 25 μg/ml hE-cadherin/Fc (SinoBiological, #10204) was added, followed by overnight (ON) incubation at 4°C. Following incubation, coverslips were washed with PBS and incubated with 1% BSA for 45 minutes at RT. Finally, the prepared substrates were washed with PBS and movie media (15 mM HEPES, 5 mM CaCl₂, 2 mM L-glutamine and 0,05% FCS) was added.

Soft PDMS substrates with a stiffness of 2-3 kPa were prepared by mixing CyA and CyB (Dow Corning) at a 1:1 ratio. The substrates were cured at 80°C for 2 hours before silanizing with 5% 3-aminopropyl trimethoxysilane (Sigma Aldrich, #281778) in ethanol. 0,05% of 200nm carboxylated fluorescent beads (Thermo Fisher, #F8807) were added to the substrates and the substrates were coated with protein A and hE-cadherin/Fc as described above.

To perform E-cadherin-coated substrate spreading assays, cells were trypsinized with Crystalline Trypsin (Sigma Aldrich, #T-0303) just long enough to see the first cells rounding up (6 minutes for MCF7 cells). Cells were resuspended in movie media and centrifuged for 3 minutes at 0,5 G. Next, supernatant was removed, cells were resuspended and filtered through a 40 μm cell strainer to generate single cells. Lastly, cells were added to the prepared E-cadherin-coated substrates and spreading was monitored for 2 hours before additional treatments were performed.

Trypsin protection assay

Cells were consecutively washed with Hanks’ Balanced Salt Solution (HBSS) and incubation media (WCL, C+: 2 mM CaCl₂ in HBSS; C-: 2 mM EGTA in HBSS) without trypsin. Next, incubation media with trypsin was added to the cells (WCL: 2 mM CaCl₂ in HBSS; C+: 2 mM CaCl₂ + 0,5% Crystalline Trypsin in HBSS; C-: 2 mM EGTA + 0,5% Crystalline Trypsin in HBSS) and cells were incubated for 20 minutes at 37°C. Media and floating cells were transferred to tubes and 1 ml cold (4°C) 2mM CaCl₂ + 1% FCS in HBSS was added to the cells. Cells were scraped and transferred to the same tubes, followed by maximum speed centrifugation for 2 minutes at 4°C. Finally, supernatant was removed and sample buffer + DTT was added to the cells to prepare them for Western blot analysis.

Western blotting

Cell extracts were prepared in sample buffer (62,5 mM Tris-HCl pH6,8; 2,5% SDS; 0,002% Bromophenol Blue; 5% DTT, 8% glycerol), boiled for 5 minutes, and loaded on SDS-PAGE gels. After running the gels, samples were transferred onto 0,22 μm-pore membranes. Next, blots were blocked with 5% BSA, incubated ON with primary antibodies at 4°C and incubated for 1 hour with HRP-conjugated secondary antibodies at RT. After incubation, blots were treated with Supersignal West Pico PLUS Chemiluminescent Substrate (Thermo Fisher, #34579) and imaged on a Bio-Rad ChemiDoc MP.

Immunofluorescence staining

Cells were fixed with either -20°C MeOH for 10 minutes (E-cadherin, β-catenin, α-catenin, α18, VD7) or 4% PFA for 20 minutes at RT (E-cadherin, phalloidin), followed by permeabilization with 0,2% Triton X-100 for 1 minute. Next, samples were blocked with 1% BSA dissolved in PBS supplemented with 0,05% Tween-20 for 45 minutes and sequentially incubated with primary antibodies for 1 hour and fluorescently labelled secondary antibodes and/or phalloidins for 45 minutes. Finally, samples were washed, dried and mounted in ProLong Gold with (Cell Signaling, #8961) or without (Cell Signaling, #9071) DAPI.

Microscopy

Fixed and live samples (FRAP) were imaged on a Leica DMi8 inverted confocal microscope equipped with 405nm, 440nm pulsed, 442nm and 470nm – 670nm white light lasers and HyD detectors. Images were acquired using a 100x (1,40 N.A.) oil HC Plan Apochromat objective.

Fixed and live samples were imaged on an Andor Dragonfly Spinning Disc inverted confocal microscope equipped with 405nm, 445nm, 488nm, 514nm and 637nm lasers and Zyla 4.2 sCMOS detectors. Images were acquired using 40x (0,95 N.A.) air or 60x (1,4 N.A.) oil Plan Apo objectives.

Live samples were imaged on a Zeiss AxioObserver upright widefield microscope equipped with LED illumination and sCMOS detector. Images were acquired using a 20x (0,5 N.A.) air EC Plan-Neofluar objective.

Fixed samples were imaged on a Zeiss AxioImager upright widefield microscope equipped with LED illumination and CCD detector. Images were acquired using a 63x (1,4 N.A.) oil Plan Apochromat objective.

Image analysis

Lamellipodia dynamics was analysed using movies of cells expressing LifeAct-GFP and LifeAct-TagRFPT. Directional changes of the lamellipodium edge were quantified (Figure S1B) and grouped in 10 minute bins. T=0 indicates the first frame in which lamellipodia made contact. Kymographs were generated using the ‘KymoRescliceWide’ ImageJ plugin with a linewidth of 1 μm.

E-cadherin accumulation was analysed using movies of cells expressing LifeAct-TagRFPT and E-cadherin-GFP^CRISPR. E-cadherin-GFP^CRISPR fluorescence intensity was measured in 3,17 μm² circular regions of interest in which the earliest adhesions appeared. T=0 indicates the first frame in which lamellipodia made contact, based on the LifeAct-TagRFPT signal. Background was subtracted and all values were normalized before graphs were plotted. Sigmoidal functions were fitted to data points. Kymographs were generated using the ‘KymoRescliceWide’ ImageJ plugin with a linewidth of 1 μm.

Cell invasion was analysed using movies of cells expressing LifeAct-GFP and LifeAct-TagRFPT. Invasion was measured 24 hours after first contact in 166 μm wide regions of interest.

Relative numbers of intracellular and surface E-cadherin clusters were analysed in images of cells labelled for total and surface E-cadherin using IF. Gaussian blur was applied to images and clusters were detected using automated ‘Moments’ thresholding followed by particle detection using particle analysis without a size cut-off. Intracellular cluster pool was calculated by subtracting the surface cluster pool from the total cluster pool.

Cadherin/catenin cluster velocities were analysed using movies of cells expressing E-cadherin-GFP^CRISPR, E-cadherin-GFP^TG, GFP-α-catenin, iRFP670-α-catenin or by E-cadherin ectodomain-binding antibodies to label E-cadherin surface clusters. Only clusters that could be tracked from start to end within the time of recording were quantified. In the case of a gradual decrease of velocity, average velocity per cluster was defined by drawing a straight line from start to end in the corresponding kymograph. Actin flow velocities were analysed using movies of cells expressing LifeAct-GFP, LifeAct-TagRFPT or LifeAct-iRFP670. Kymographs were generated using the ‘KymoRescliceWide’ ImageJ plugin with a linewidth of 0,7 μm.

α-catenin conformational changes were visualized by α18/ α-catenin IF labelling (to analyse the M-domain), or VD7/ α-catenin IF labelling (to analyse the ABD). Ratiometric images were generated by dividing the α18 or VD7 images by the total α-catenin images. Gaussian blur was applied to images before calculation. For quantification, the ratiometric signal was measured at the junctions of interest. For anti-parallel flow junctions at head-on contacts, cells were fixed between 1 and 2 hours after first contact.

FRAP measurements were performed by bleaching 8-μm diameter circles in the region of interest, followed by 6:40 minutes of imaging with a frame interval of 2 seconds. Logistic functions were fitted to data points.

α-catenin cluster sizes were analysed in images of cells expressing GFP-α-catenin mutants and labelled for surface E-cadherin using IF. GFP-α-catenin fluorescence intensity was measured in 0,82 μm² circular regions of interest in which an E-cadherin surface cluster was visible. Background was subtracted to correct for expression levels.

Directional F-actin flow within lamellipodia was analysed in PIVlab version 2.61, MATLAB. Lamellipodia flow was analysed for 8:20 minutes with a frame interval of 5 seconds. Rose plots of the average flow direction were generated.

Software

All images and movies were analysed, and all movies were labelled using ImageJ 1.52. All plots were generated, and all statistical tests were performed in Prism 9.0.0. All modelling and Particle Image Velocimetry (PIV) analysis was performed using Matlab 9.9 (R2020b). The Leica DMi8 inverted confocal microscope was controlled by LASX software. The Andor Dragonfly Spinning Disc inverted confocal microscope was controlled by Fusion software. The Zeiss AxioImager upright widefield microscope and the Zeiss AxioImager upright widefield microscope were controlled by Zen Blue 3 software.

Quantification And Statistical Analysis

Statistical parameters for individual experiments are listed in the corresponding figure legends. This includes the statistical test performed, sample size (n), number of independent experiments and statistical significance.

Highlights.

• E-cadherin adhesion couples cortical flows when migrating epithelia collide

• Coupled antiparallel flows apply mechanical tension to unfold α-catenin’s C terminus

• α-catenin promotes cadherin adhesion growth to mechanically load the actin flows

• Mechanical coupling downregulates antiparallel actin flows to inhibit migration

In brief.

Noordstra et al. report that α-catenin mechanosensitivity within E-cadherin complexes mediates contact inhibition of epithelial migration. Antiparallel cortical flows found at lamellipodia of colliding cells exert tension on the α-catenin actin-binding domain, causing it to increase adhesion growth, and mechanical loading then downregulates cortical flow, thereby paralyzing lamellipodia and inhibiting locomotion.

Data and code availability

• All microscopy data and original western blot images reported in this manuscript will be shared by the lead contact, Alpha Yap, upon request.

• All original code will be shared by the lead contact, Pere Roca-Cusachs, upon request.

• Any additional information required to reanalyze the data reported in this work paper is available from the lead contact upon request.