SUMMARY

Graphical Abstract

Open in a separate window

In Brief

INTRODUCTION

Cadherin-mediated adhesion is a defining trait of all metazoans. It is mediated by a transmembrane receptor, a classical cadherin (E-cadherin in many epithelia), which functions in a large multi-protein complex. The unstructured cytoplasmic region of E-cadherin binds directly to β-catenin, which in turn is bound to α-catenin (we denote this as the cadherin–catenin complex, or CCC). The CCC-incorporated α-catenin also interacts with F-actin through its actin-binding domain (Pappas and Rimm, 2006; Harris and Tepass, 2010; Michael and Yap, 2013; Mège and Ishiyama, 2017), providing a bridge between cadherin and the actin cytoskeleton. We recently showed that the extracellular region of punctate adherens junctions (pAJs) consists of dense, paracrystalline nanoclusters formed through cis and trans interactions of cadherin ectodomains, interspersed with less dense cadherin regions. The nanoclusters in this “mosaic model” assemble and disassemble on a time-scale of seconds in a process that depends on the dynamics of junctional actin (Indra et al., 2018). However, the pAJs are linked to F-actin bundles, which are thought to be among the most stable actin structures. How then can the fast dynamics of cadherin clusters and the much slower turnover of actin bundles be reconciled? Further, the pulling forces the bundles apply on the CCC are essential for both the overall stability of AJs and the tension-dependent recruitment of vinculin and other proteins into AJs (Huveneers and de Rooij, 2013; Mège and Ishiyama, 2017). How can these tensile forces be applied between regions with very different dynamical properties? Here we address these questions by studying the spatiotemporal properties of E-cadherin clusters and F-actin bundles in pAJs.

AJs, including pAJs, along with focal adhesions (FAs) and tips of lamellipodia, are sites of rapid F-actin polymerization (Vasioukhin et al., 2000; Zhang et al., 2005; Kovacs et al., 2011). However, the polymerization mechanism and the role and fate of polymerized filaments in AJs are poorly understood. In one model, F-actin at AJs is nucleated by the Arp2/3 complex. Direct binding of Arp2/3, to the cytoplasmic tail of E-cadherin (Kovacs et al., 2002), or indirect binding through cortactin (Han et al., 2014), neogenin (Lee et al., 2016), or α-actinin-4 (Tang and Brieher, 2012) have all been suggested as mechanisms for Arp2/3 recruitment into AJs. The Arp2/3-nucleated filaments have been proposed to grow using Ena/Vasodilator-stimulated phosphoprotein(VASP) proteins (Scott et al., 2006; Leerberg et al., 2014) and to then be reconfigured from a branched to a bundled array by an N-WASP-dependent mechanism (Kovacs et al., 2011; Brieher and Yap, 2013). This model is challenged, however, by observations that mature AJs are depleted of Arp2/3 (Verma et al., 2004; Hansen et al., 2013) and by proteomics analyses that have failed to identify Arp2/3, neogenin, N-WASP, or α-actinin-4 in association with cadherin tails (Van Itallie et al., 2014; Guo et al., 2014; Li et al., 2019).

There are also uncertainties concerning the configuration and dynamics of F-actin in AJs. Platinum replica electron microscopy (EM) identified a pool of branched F-actin at different types of AJs, including at pAJs, of endothelial cells (Efimova and Svitkina, 2018), while other EM studies have suggested that AJs are directly associated with tensile actin bundles (Yonemura et al., 1995; Buckley et al., 2014; Choi et al., 2016; Li et al., 2019). With respect to dynamics, fluorescence recovery after photobleaching (FRAP) and fluorescence loss after photoactivation (FLAP) experiments with Madin Darby Canine Kidney (MDCK) cells showed that AJ-associated F-actin recovers with a half-life of seconds (Yamada et al., 2005), a rate that corresponds to that of the branched F-actin in lamellipodia (Lai et al., 2008; Fritzsche et al., 2013). However, other studies have reported a half-life of ~1 min or even longer (Wu, 2014; Cavey et al., 2008; Kovacs et al., 2011), corresponding to that of bundled rather than branched actin (Hotulainen and Lappalainen, 2006).

Here we show that pAJ-associated bundles are composed of two distinct regions: a stable F-actin bundle, which we refer to as the adherens junction bundle stalk (AJ-BS), and a highly dynamic pAJ-proximal region at the tip of this bundle, the adherens junction bundle tip (AJ-BT). The existence of these two connected but distinct actin bundle regions helps to reconcile previous studies, which have alternatively found both branched and dynamic, and bundled and stable, F-actin in AJs. Since the dynamical properties of the AJ-BTs are similar to those of CCC clusters, the existence of the tips resolves the issue of a mismatch between bundle and CCC dynamics. Finally, bidirectional coupling between CCC and F-actin dynamics coordinates organization of the actin cytoskeleton in adjacent cells. Our work, therefore, provides a dynamical model of pAJs, which explains how they are able to maintain and synchronize their remarkable flexibility under constant actomyosin stretching. In light of their common protein compositions, this model may be applicable to AJs in general.

RESULTS

AJ-linked Actin Bundles Consist of Two Distinct Regions with Differing Compositions of Actin-Binding Proteins

The pAJs are the most prominent AJs of human squamous carcinoma A431 cells (Indra et al., 2013, 2018), and are identified by E-cadherin staining in Figure 1A. Phalloidin staining showed that they are linked to radial actin bundles of diverse length, which approach each pAJ from two adjacent cells. Their pAJ-attached termini exhibit the strongest phalloidin fluorescence (Figure 1A, arrowheads), perhaps due to overlap of the actin bundles from each cell linked at the junction. The extra-junctional end of the bundle appears to integrate into a dense actin cortical network (Figure 1A).

Open in a separate window

Figure 1.

Organization of pAJ-linked Actin Bundles in A431 Cells

(A–D) Projections of all x-y optical slices of A431 cells stained for (A) E-cadherin (Ecad, red) and F-actin (actin, green); (B) Ecad (green), F-actin (red), and myosin IIA (myosin II, blue); (C) Ecad (green), actin, and calponin-3 (Cnn3); and (D) Cnn3 (green) and myosin IIA (red).

Scale bar, 10 μm. The zoomed areas (indicated by dashed boxes) are presented on right panel. Scale bar, 5 μm. Group of pAJs in a left panel in (A) is further magnified in the inset. Arrowheads in (A)–(C) indicate adherens junction bundle tips (AJ-BTs), which are enriched with F-actin and Ecad, but are devoid of myosin IIA and Cnn3. Arrowhead in (D) shows that myosin IIA is not integrated into a Cnn3-positive pAJ-linked bundle.

To better understand the organization of the pAJ-connected bundles, we stained A431 cells for myosin IIA (MyoIIA) and calponin-2 or −3 (Cnn2/3). These proteins have been shown to be abundant in actin bundles in fibroblasts (Tojkander et al., 2012; Ciuba et al., 2018). We found that MyoIIA was typically concentrated several microns away from pAJs (Figure 1B). By contrast, Cnn2/3 were specifically enriched in the pAJ-connected bundles (Figures 1C and S1). However, their pAJ-integrated termini were completely devoid of Cnn2/3 (Figure 1C, arrowhead). Double Cnn3/MyoIIA staining confirmed that calponin-positive pAJ-linked bundles did not incorporate MyoIIA (Figure 1D).

The absence of Cnn2/3 staining in the pAJ-attached termini suggests that these regions are structurally different from the remaining parts of the bundles; however, this observation could also be due to inaccessibility of Cnn2/3 to antibodies at sites of actin-CCC interactions. To distinguish these two possibilities, we expressed recombinant Dendra2-tagged Cnn3 (Cnn3-Dn) in A431 cells. Figures 2A and ​and2B2B show that, consistent with the antibody staining, Cnn3-Dn was efficiently incorporated into the pAJ-linked bundles, but not into their pAJ-integrated termini. Line-scan analysis of fluorescence intensity across the pAJs confirmed that the depressions in Cnn3-Dn incorporation exactly matched the peaks of both F-actin and E-cadherin (Figure 2E).

Open in a separate window

Figure 2.

F-actin in AJ-BT Regions of pAJ-linked Bundles Associates with a Distinct Set of Actin-Binding Proteins

(A–D) Projections of all x-y optical slices of Cnn3-Dn-expressing A431 cells imaged for Dendra2 fluorescence (green, Cnn3-Dn) and stained for (A) Ecad, (B) actin, (C) vinculin (Vcn), and (D) mena. Scale bar, 10 μm. The zoomed areas (indicated by dashed boxes in Cnn3-Dn images) are presented at the bottom. Scale bar, 5 μm. Arrowheads indicate the gaps in Cnn3-Dn distribution along pAJ-linked bundles that are precisely matched by the antibody staining.

(E) Line-scan analysis of the selected pAJs for distribution of Cnn3-Dn (green) and the indicated pAJ-associated proteins revealed by specific antibodies (red).

(F) Structured Illumination Microscopy of Cnn3-Dn-expressing A431 cells double-stained for Dendra2 (green, Cnn3) and Ecad (red). Arrowhead indicates AJ-BT region. Scale bar, 10 μm. The zoomed areas (indicated by dashed boxes in Cnn3-Dn images) are presented at the bottom. Scale bar, 5 μm.

(G) The thickness of the Cnn3-Dn-deficient regions measured for 15 independent pAJs on Ono, 2018; Wu, 2014 SIM and on confocal images (mean +/−SEM). Two-tailed t test.

(H) Schematic representation of pAJ-linked bundles. Each bundle appears to consist of at least two regions. The first region, AJ-BT (AJ-BT), which is Cnn2/3-deficient (CNN2/3-), directly contacts the cadherin-catenin complex (CCC). The second, Cnn2/3-positive (CNN2/3+) region, AJ-BS, is attached to AJ-BT and interconnects the AJs with the actomyosin cytoskeleton.

We next tested whether major actin cytoskeletal proteins associated with pAJs—vinculin, VASP, mena, and FBLIM1—were located at the Cnn3-Dn–deficient regions of the pAJ-linked bundles. Indeed, immunofluorescence microscopy of Cnn3-Dn–expressing cells for these proteins, and line-scan analyses of the resulting images, clearly showed the presence of these proteins in the pAJ-adjacent bundle termini (Figures 2C–2E and S1). Structured-illumination microscopy (SIM) of E-cadherin, which marks the locations of pAJs, further validated localization of E-cadherin within the Cnn3-Dn–deficient regions of the bundles (Figure 2F). The average thickness of this region, assessed using 30 independent pAJ images (obtained by SIM or conventional microscopy), was ~400 nm (Figure 2G).

Overall, these results show that a pAJ-linked actin bundle is structurally heterogeneous and encompasses at least two regions (Figure 2H). The first region overlaps with the CCC at the light-microscopy resolution level. It exhibits highest F-actin deposition, is enriched with VASP, mena, FBLIM1, and vinculin, and is devoid of Cnn2/3. We denote this region as the bundle tip (AJ-BT), which, given the 400-nm total length of the Cnn2/3-deficient area, has a length in each cell on the order of 200 nm. The second region corresponds to the bundle stalk (AJ-BS), which is enriched with Cnn2/3. The distal end of AJ-BSs interacts with the AJ-BT while the opposite end is integrated into the contractile actomyosin network, which ultimately transmits the mechanical forces into this system.

F-actins in AJ-BSs and AJ-BTs Have Different Dynamical Properties

The relatively large size of the tip region allows it to be analyzed using conventional light microscopy. To this end, we produced A431 cells that stably express a photoactivatable form of actin, mCH_PA-actin. Previous studies have shown that actin tagged with fluorescent proteins assembles into filaments and can serve as an excellent probe of F-actin dynamics (Yamada et al., 2005; Tojkander et al., 2015). Indeed, double immunostaining of the transfected cells for mCH_PA-actin and phalloidin showed that the actin bundles of these cells recruited the recombinant actin (Figure 3A). Furthermore, similar to the untagged actin in the control cells, mCH_PA-actin produced bright plaques at sites of E-cadherin-based pAJs (Figure 3B). To specifically photoactivate mCH_PA fluorescence at bundle tips or at bundle stalks, we developed a “pAJ-actin dynamics” assay in which the mCH_PA-actin cells were co-cultured with mGFP-tagged E-cadherin (EcGFP)-expressing A431 counterparts (see Figure 3C). These mCH_PA-actin/EcGFP co-cultures allowed us to study actin dynamics in mixed pAJs, one half of which was highlighted by EcGFP, with the other half expressing the “dark” mCH_PA-actin.

Open in a separate window

Figure 3.

Dynamics of F-actin in AJ-BTs and in AJ-BSs

(A and B) Maximum projection view of A431 cells stably expressing mCH_PA-actin stained for (A) actin using Phalloidin (actin, green) and mCH_PA-actin using anti-mCH antibody (mCH, red) and (B) Ecad (green) and mCH_PA-actin (red). Scale bar (in A and B), 10 μm. The zoomed areas (indicated by dashed boxes in actin and Ecad images) are presented at the bottom. Scale bar, 5 μm.

(C) Time-lapse series of mCH_PA-actin photoactivated in the co-cultures of mCH_PA-actin- and EcGFP-expressing cells. The series were taken in green and red imaging channels: green channel detects EcGFP, while red channel detects only background fluorescence before photoactivation. The full-screen images (left) show groups of cells just before photoactivation. Scale bar, 10 μm. The time-lapse frames of the boxed areas are zoomed at the right panels and the arrowheads show the selected pAJs before photoactivation (−1), immediately after (0) and then at particular following time points (exact time is indicated in s). Red fluorescence of mCH_PA-actin was activated in diffraction-limited spots (~500 nm in diameter) in three different locations: exactly at sites of EcGFP-tagged pAJs (AJ-BT); ~0.5 μk toward the center of the dark cell (AJ-BS), or at the pAJ lateral sides (lateral to AJ). Scale bar, 5 μm.

(D) Kinetics of decay of mCH_PA-actin photoactivated in AJ-BT, AJ-BS, and lateral to pAJ (n = 10, mean +/− SEM).

(E) Parameters for mCH_PA-actin fluorescence decay assessed from curves presented in (D).

(F) Displacement of individual photoactivated spots of actin over 1 min (n = 10). Two-tailed t test.

First, we photoactivated mCH_PA precisely at pAJ sites using the maximum focused laser beam (Figure 3C, AJ-BT; Video S1). Time-lapse imaging showed that at this position, the spot of red mCH_PA-derived fluorescence (its diameter encompassed ~4 to 5 pixels, or 440–550 nm) exhibited a relatively short lifetime, with the half-life of fluorescence decay ~20 s, with low variation between experiments (Figures 3D and ​and3E).3E). The photoactivated spot, usually trackable for up to 1 min, was immobile and tightly associated with the selected pAJ during the entire observation period (Figure 3F). To probe actin dynamics in the AJ-BSs, we targeted the laser beam to sites located ~0.5 μm from the pAJ toward the center of the dark cell (Figure 3C, AJ-BS; Video S2). Remarkably, at these sites the decay of photoactivated spots was significantly slower and furthermore, in most cases the spots showed retrograde movements (Figure 3F). Both of these features, the retrograde movement (median ~400 nm/min) and the rate of fluorescent decay (median t1/2 ~50 s) significantly varied between individual junctions (Figures 3E and ​and3F).3F). To ensure that the F-actin dynamic features were specific for the selected spots along pAJ-linked bundles, we performed control experiments, which targeted sites near to, but outside of, the pAJs. At these locations, where junctional actin is not present (Figures 3C and ​and3D;3D; Video S3), the half-life of the resulting red fluorescence decay (~10 s) was significantly faster than in AJ-BTs.

Taken together, the pAJ-actin dynamics assay confirmed the compound architecture of the pAJ-linked actin bundles. Their AJ-BTs appear to consist of short-lived F-actin whose turnover rate of ~20 s is similar to that of lamellipodium actin (Fritzsche et al., 2013; Lai et al., 2008). By contrast, F-actin in AJ-BSs exhibited characteristics more similar to those in dorsal actin bundles linked to FAs, which display centripetal flow with an average half-life of ~2 min (Hotulainen and Lappalainen, 2006; Tojkander et al., 2015). While we cannot exclude the possibility that a small pool of AJ-BT-linked F-actin eventually moves into the AJ-BS, the bulk of the AJ-BT F-actin was immobile and tightly coupled with the CCC in pAJs.

Cadherin Cluster Stability Regulates F-actin Turnover in AJ-BTs and AJ-BSs

The fast turnover of AJ-BT F-actin suggests that the assembly-disassembly of both F-actin and cadherin clusters is tightly coupled. This idea is supported by previous observations that stabilization of F-actin by Jasplakinolide prolonged the lifetime of cadherin clusters in pAJs (Indra et al., 2018). Does the turnover of F-actin in AJ-BTs depend on the stability of cadherin clusters? To answer this question, we studied how different manipulations known to affect cadherin cluster dynamics influence that of F-actin. The green fluorescence in one half of pAJs in our mCH_PA-actin/EcGFP co-cultures allowed us to use FRAP technique to determine the dynamical properties of cadherin in pAJs. While FRAP significantly underestimates the rate of cadherin cluster assembly and disassembly due to local reincorporation of cadherin into new clusters (Indra et al., 2018), this approach is useful for the general assessment of cadherin cluster stability. First, we expose the co-cultures to hypertonic shock, which we have shown to dramatically reduce cadherin dynamics in pAJs (Troyanovsky et al., 2006). Indeed, after 10 min in hypertonic sucrose media, the recovery of EcGFP fluorescence in pAJs as well as the decay of mCH_PA-actin photoactivated fluorescence in AJ-BTs was nearly completely stalled (Figures 4A and ​and4B).4B). Interestingly, the mCH_PA-actin turnover in AJ-BSs as well as its retrograde flow were also blocked (Figures 4B and ​and4C4C).

Open in a separate window

Figure 4.

Cadherin Regulates F-actin Turnover in pAJ-linked Bundles

(A) FRAP of EcGFP in pAJs after 10 min in hypertonic sucrose media (mean +/− SEM).

(B) Decay of mCH_PA-actin photoactivated in AJ-BT (actin-AJ-BT) or in AJ-BS (actin-AJ-BS) (mean +/− SEM).

(C) Displacements of the photoactivated spot as in Figure 3F.

(D) Maximum projection view of three different co-cultures of A431 cells. One of the sublines stably expresses mCH_PA-actin, which is visualized by anti-mCH staining (mCH, red). These cells form mixed pAJs with cells stably expressing EcGFP (GFP, green), or its mutants, cis mutant EcGFP (cis-GFP) and P5A-EcGFP (P5A-GFP). Scale bar, 10 μm. The zoomed areas (indicated by dashed boxes in green images) are presented on the right panel. Scale bar, 5 μm.

(E) FRAP of pAJs in cells expressing EcGFP or its mutants P5A-EcGFP or cis mutant EcGFP (n = 15). FRAP was performed either in homogeneous cultures of corresponding cells (Homogeneous pAJ) or in co-cultures with “dark” mCH_PA-actin-expressing A431 cells. In the latter case the pAJs located between two types of cells were studies (mean +/− SEM).

(F) Decay of mCH_PA-actin spots photoactivated in pAJs (Actin-AJ-BT) or in bundles (Actin-AJ-BS) in co-cultures with EcGFP or its mutants. Right panel shows parameters of corresponding decays (mean +/− SEM).

(G) Displacements of the photoactivated spots as in (C).

Since hypertonic shock abolished pAJ disassembly through relatively indirect and still unknown mechanisms, we chose to modify cadherin cluster stability directly using specific point mutations that effect cadherin ectodomain trans and cis interactions. Specifically, in the mCH_PA-actin/EcGFP co-culture, we replaced the EcGFP cells with cells expressing mutant EcGFP. The first, involving the P5A mutant, dramatically strengthens the K_D of cadherin trans-bonds, from ~100 μM to 3 μM (Vendome et al., 2011), and thus increases cluster stability. Notably, since P5A forms “extra” hydrogen bonds in trans with distal β-strand residues, the strengthening effect of this mutation should also be observed when binding to wild-type (WT) E-cadherin (Vendome et al., 2011). The second mutant targets the cis-binding interface of E-cadherin and destabilizes pAJs by disrupting the paracrystalline lattice formed by cadherin ectodomains (Harrison et al., 2011; Strale et al., 2015).

We first performed a set of control experiments in order to confirm that in the mixed pAJs in our co-cultures, the P5A-EcGFP mutant would indeed decrease, and the EcGFP cis mutant would increase, the turnover rate of the intact E-cadherin present in the mCH_PA-actin-expressing cells. We showed that both mutants established mixed pAJs with mCH_PA-actin-expressing cells and that these mixed pAJs were linked to the mCH_PA-actin bundles (Figure 4D). Then, using a FRAP assay with the mutant-expressing cells in homogeneous cultures and in co-cultures with mCH_PA-actin-expressing cells, we verified that both mutants changed EcGFP turnover rates as expected. Indeed, fluorescence recovery curves (Figure 4E) showed that the P5A mutation significantly reduced, while the cis mutation, by contrast, increased, the pAJ cadherin turnover rate in both the mixed and homogeneous pAJs. The P5A mutant likely stabilizes mixed pAJs through enhanced trans interactions with WT E-cadherin, whereas the cis mutant destabilizes pAJs likely by disrupting one half of the E-cadherin lattice, thus disrupting the other half.

Finally, we confirmed that turnover of the WT E-cadherin on one half of pAJs matched that of the mutant E-cadherin present on the opposite side of the same pAJ. To show this, we co-cultured the mutant-expressing cells with A431 cells expressing mCH-tagged E-cadherin (EcmCH). Both, GFP and mCH fluorophores in the hetero-chromatic pAJs were simultaneously bleached and recovery of their fluorescence simultaneously monitored by time-lapse microscopy (Figure S2). This two-color FRAP experiment revealed that E-cadherin turnover rates were indeed balanced in both sides of AJs, consistent with previously published data showing that cadherin turnover rates are closely correlated in two halves of the same junction (de Beco et al., 2015).

These observations confirmed that the replacement of EcGFP with P5A-EcGFP or with cis-EcGFP, respectively, either slowed or increased WT E-cadherin turnover in the mCH_PA-actin-expressing cells connected through mixed pAJs in our co-cultures. Next, using our pAJ-actin dynamics assay, we determined actin turnover rates in these junctions. The results clearly showed that AJ-BTs from the pAJs for which the opposite half incorporated the P5A-EcGFP mutant exhibited a significant increase in the lifetime of mCH_PA red fluorescence (t1/2 ~40 s). By contrast, mCH_PA fluorescence in AJ-BTs for which pAJs contacted cis-EcGFP-expressing cells had much shorter lifetimes, with t1/2 of just few seconds (Figure 4F). Surprisingly, both mutations correspondingly changed the F-actin turnover not only in the AJ-BTs, but also in the AJ-BSs (Figure 4F). Finally, the centripetal flow of the red spot photoactivated at the AJ-BSs was nearly completely stalled in the cells connected to P5A-EcGFP but, in contrast, increased in cis mutant EcGFP co-cultures (Figure 4G).

F-actin Dynamics Are Coordinated between Adjacent Cells

Our previously reported data suggest that F-actin turnover regulates the rate of cadherin cluster assembly and disassembly (Indra et al., 2018). Here we have demonstrated that the inverse phenomenon also takes place: cadherin cluster dynamics control the rate of F-actin turnover in pAJ-linked bundles. These observations imply that F-actin polymerization–depolymerization rates in both halves of a pAJ are coordinated. To test this hypothesis we assessed whether, in the mCH_PA-actin/EcGFP co-culture, F-actin stabilization in the EcGFP expressing cells would stabilize F-actin in the adjacent mCH_PA-actin-expressing cells.

To stabilize F-actin in the EcGFP-expressing cells, we knocked out a key actin depolymerizing factor, cofilin 1 (Cfl1), using a CRISPR/Cas9 approach. The lack of Cfl1 has been shown to reduce the F-actin depolymerization rate of the branched actin network of different types of cells, but does not abolish AJs (Lappalainen and Drubin, 1997; Hotulainen et al., 2005; Iwasa and Mullins, 2007; Vitriol et al., 2013; Wang et al., 2016). As has been shown for other cell types, Cfl1 knockout (KO) in the EcGFP-expressing A431 cells resulted in cell shape flattening, which may be caused by unbalanced actin polymerization in cell lamellipodia. Despite these changes, as has been observed previously (Chu et al., 2012; Wang et al., 2016), the Cfl1-KO cells retained their pAJs (Figure 5A) and general actin bundle organization, as revealed by phalloidin and anti-Cnn3 staining (Figure 5B, ​,C).C). As expected from our previous experiments with Jasplakinolide (Indra et al., 2018), the FRAP assay showed a significant stabilization of E-cadherin in AJs upon Cfl1 KO (Figure 5D, homogeneous pAJ).

Open in a separate window

Figure 5.

Knockout of Cfl1 Affects Actin Dynamics in the Neighboring WT Cells

(A–C) Maximum projection view of control A431-EcGFP cells (A431-EcGFP) or after Cfl1-KO (A431-EcGFP-Cfl1-KO). Cells were imaged for GFP (EcGFP) and for (A) Cfl1 (Cfl, red), (B) actin (actin, red), and (C) Cnn-3 (Cnn3, red). Scale bar, 10 μm. The zoomed areas (indicated by dashed boxes in EcGFP images) are presented in the insets (A) or at the bottom (B and C). Scale bar, 5 μm.

(D) FRAP of pAJs in control and in Cfl1-KO A431-EcGFP cells (n = 15). FRAP was performed either in homogeneous cultures of corresponding cells (homogeneous pAJ) or in co-cultures with “dark” mCH_PA-expressing A431 cells. In the latter case the pAJ between dark cells and cells expressing EcGFP were studied. The corresponding FRAP parameters are shown at the bottom (mean +/− SEM).

(E) Decay of mCH_PA-actin spots photoactivated in pAJs (actin-AJ-BT) or in bundle stalks (actin-AJ-BS) in mCH_PA-actin expressing cells co-cultured with A431-EcGFP or A431-EcGFP-Cfl1-KO cells. The corresponding parameters of mCH_PA-actin turnover are shown at the bottom (mean +/− SEM).

EcGFP in the Cfl1-KO cells showed nearly the same turnover rate in the co-culture with mCH_PA-actin-expressing cells as in the homogeneous culture (Figure 5D). As expected from this observation, photoactivation of mCH_PA-actin in such co-cultures showed that the Cfl1-KO EcGFP-expressing cells significantly stabilized actin in both AJ-BTs and AJ-BSs in adjacent cells (Figure 5D). Together, these experiments suggest that the reduction of F-actin dynamics induced by Cfl1-KO in EcGFP-expressing cells was transduced into the neighboring cells through reduction of cadherin turnover in AJs.

pAJs Are Enriched with Factors Driving Dynamics of the Branched Actin Cytoskeleton

To further shed light on the organization of pAJ-associated actin filaments and mechanisms of their fast dynamics, we stained A431-EcGFP cells for proteins known to bind and regulate the dynamics of branched actin filaments. First, we extended our examination of Cfl1. This protein transiently targets and severs actin filaments in collaboration with another protein, Aip1 (Bravo-Cordero et al., 2013; Ono, 2018). While, as observed for other cell types (Abe et al., 1993; Munsie et al., 2012), Cfl1 was mostly cytosolic in A431 cells, many of the well-isolated pAJs exhibited Cfl1-specific puncta (Figure 6A). Line-scan analyses confirmed close spatial association between pAJs and such Cfl1 puncta (Figures 6A and ​and6E).6E). It has been shown that in absence of Aip1, Cfl1 binds to but does not sever filaments, leading to accumulation of Cfl1-bound F-actin (Kato et al., 2008); we exploited this observation to clearly identify the pool of Cfl1-targetted actin filaments in Aip1 knockout A431-EcGFP cells. As shown in Figure 6B, these cells exhibited numerous Cfl1-decorated cortical filaments, which were especially prominent in pAJs (Figure 6B). The strong enrichment of Cfl1 in pAJ-linked filaments, together with the dramatic reduction of actin and E-cadherin turnover in pAJs upon Cfl1 KO, strongly suggests that Cfl1 plays a direct and critical role in F-actin homeostasis in pAJs.

Open in a separate window

Figure 6.

Subcellular Localization of Proteins Involved in Branched Actin Dynamics: Cfl1, Arp2/3, and Cortactin

Projections of all x-y optical slices of A431-EcGFP cells

(A), (C), (D), and A431-EcGFP-Aip1-KO cells (B) imaged for GFP fluorescence (green, EcGFP) and stained for Cfl1 (Cfl), Aip1 (Aip), Arp2/3 (Arp2/3), and cortactin (CTTN). Scale bar, 10 μm. The zoomed areas (indicated by dashed boxes in EcGFP images) are presented at the bottom or in the insets (for anti-Aip staining in A). Arrowheads indicate pAJs, which exhibit clear colocalization with the antibody-stained proteins. Arrows marked pAJs, which colocalization with the stained protein is unclear because of high nearby fluorescence. Scale bar, 5 μm.

(E) Line-scan analysis of the pAJs (selected by arrowheads in A–D) for distribution of EcGFP (green) and the indicated pAJ-associated proteins (red). In the case of Arp2/3 and cortactin, two line-scans are shown—one for “marked by arrowhead” and another for “marked by arrow” as pAJs.

Next we studied localization of Arp2/3, a key protein complex involved in organization and dynamics of the branched actin cytoskeleton. As highlighted in the Introduction, there are strong uncertainties about the mechanisms of Arp2/3 recruitment into AJs and even about its presence in mature AJs (Verma et al., 2004; Hansen et al., 2013). This uncertainty stems in part from the fact that Arp2/3 is abundant in the cell cortex and is especially enriched in cell–cell contact areas. To reduce cytosolic pools of Arp2/3, we fixed cells in cytoskeleton-preserving buffer in the presence of Triton X-100. Figures 6C and ​and6D6D show that Arp2/3 and its cofactor, cortactin, were highly concentrated in the cell–cell contact regions. In contrast to Cfl1, for which peak intensity usually corresponded to that of E-cadherin, Arp2/3 and cortactin were located in relatively large zones encompassing the pAJs (Figure 6D). These results suggest that Arp2/3, in agreement with proteomics data, might not be specifically targeted into mature pAJs. Instead, Arp2/3 might be involved in formation and dynamics of the entire actin cortex of the lateral membranes in which the AJ-BTs are integrated. More work is needed to understand whether Arp2/3 plays a specific role in generation and dynamics of CCC-coupled F-actin.

DISCUSSION

In this work we show that pAJ-linked actin bundles consist of two distinct regions: a bundle stalk, the AJ-BS, which connects the bundle to the actomyosin cytoskeleton; and a bundle tip, the AJ-BT, which serves as a joint between the relatively stable stalk and the highly dynamic adhesive region of pAJs. Our FLAP experiments show that the half-life of F-actin in AJ-BTs is ~20 s as compared to ~1 min in AJ-BSs. Time-lapse recordings of these adjacent F-actin regions suggest that they are generated by distinct F-actin assembly pathways since mCH_PA-actin activated in AJ-BTs does not translocate into the AJ-BS, but remains associated with corresponding pAJs over its entire lifetime. The F-actin in the AJ-BS, by contrast, is not immobile but flows rearward.

These differences in dynamics may reflect the striking structural difference between these two bundle compartments. The AJ-BS apparently consists of parallel actin filaments since it is enriched with the prominent actin bundling proteins Cnn2/3. The AJ-BT lacks these bundling proteins, but instead is enriched with proteins involved in actin dynamics, such as mena and VASP. The F-actin depolymerization factor, Cfl1, which plays a key role in the dynamics of the branched actin networks (Chan et al., 2009; Bravo-Cordero et al., 2013), is also colocalized with AJ-BTs. The role of this factor in AJ-BT dynamics is supported by the dramatic stabilization of these structures upon Cfl1 KO. AJ-BTs are integrated into the cell–cell contact areas enriched with Arp2/3, known to play a role in nucleation of branched actin networks. This dual organization of pAJ-linked actin bundles is consistent with a recent platinum replica EM study of endothelial AJs, which shows that bundles of parallel actin filaments are linked to a VE-cadherin-rich zone through a ~200–300-nm-thick branched actin network (Efimova and Svitkina, 2018).

Our study also reveals that during its entire lifespan F-actin photoactivated in AJ-BTs associates with pAJs. This observation is important since it suggests that assembly-disassembly cycles of F-actin in AJ-BTs are coupled with those of cadherin clusters to which they are linked via α-catenin. This idea is further reinforced by the observed bidirectional relationship between F-actin and cadherin turnover. In all cases when F-actin was stabilized, E-cadherin clusters also showed increased stability. Our study provides two examples of this relationship: KO of the F-actin depolymerization factor Cfl1 and osmotic shock all simultaneously increased the lifetime and immobile fraction of both E-cadherin and F-actin in AJ-BTs. These observations are in line with previous data showing that F-actin stabilization by Jasplakinolide increases stability of E-cadherin clusters in AJs (Indra et al., 2018). Here we have shown that the inverse is also true: increases or decreases in the lifetime of E-cadherin clusters in pAJs induced by mutant E-cadherins correspondingly changed F-actin lifetimes.

The precise spatiotemporal organization of F-actin in AJ-BTs and the mechanisms responsible for the close coupling between actin and cadherin cluster turnover are still unclear and are important avenues for future research. Figure 7 presents a very general schematic model, based on the results published here and previously (Indra et al., 2018; Efimova and Svitkina, 2018), as to how F-actin and cadherin cluster dynamics are coupled and coordinated. The figure illustrates how parallel and stable actin filaments of the AJ-BS interact with apparently branched and dynamic actin filaments in the AJ-BT, which in turn, interact with the CCC. Our results strongly suggest that each cadherin nanocluster associates with its own “tangle” of the branched F-actin network. The top panel depicts two cadherin nanoclusters linked to two such tangles that are separated by a region of lower cadherin and actin density. In the bottom panel, these two tangles and associated CCC clusters depolymerize and a new tangle linked to a new cadherin cluster is formed. The transition between the structures shown in the two panels involves the rearrangement of cadherin clusters that continuously dissociate and reform in pAJs as a result of two coupled processes that are each cooperative: an extracellular process involving cis and trans cadherin interactions (Harrison et al., 2011; Strale et al., 2015) and an intracellular process involving α-catenin binding to F-actin (Hansen et al., 2013; Buckley et al., 2014; Chen et al., 2015). It is possible that Arp2/3 and Cfl1 play key roles in this process. Importantly, new cadherin clusters continuously reutilize E-cadherin of the disassembled clusters, which results in underestimation of cadherin turnover rate by FRAP or FLAP assays (Indra et al., 2018). By contrast, the requirement for ADP–ATP exchange impedes the immediate reuse of the G-actin released from the disassembled F-actin tangles. This difference in assembly mechanisms could be a major reason for significant differences in the half lives of E-cadherin and AJ-BT actin in pAJs (~5 min and ~20 s, respectively).

Open in a separate window

Figure 7.

Schematic Model of Coupling between F-actin and Cadherin in pAJs

A hypothetical cross section of a pAJ is shown. Blue parallel lines represent the actin bundle stalk (AJ-BS) apparently consisting of parallel actin filaments of the same polarity. The stalk is terminated with the bundle tip (AJ-BT) shown by red intersecting lines, which in turn interact with the CCCs (green). Initially (the top image) the bundle tip consists of two cadherin clusters (1 and 2), formed in association with corresponding tangles of branched F-actin. In the bottom panel, these tangles and associated clusters are disassembled and a new CCC cluster (3) and a corresponding new actin tangle are assembled. The new cluster recruits only a few new cadherin molecules (yellow), but the new tangle is built entirely from newly delivered G-actin (magenta). The AJ-BT/AJ-BS interface is positioned to play a key role in dynamics and/or organization of both AJ-BT and AJ-BS.

Close coupling between cadherin clusters and F-actin in AJ-BTs resolves the question highlighted in the Introduction about the apparent mismatch between cadherin and F-actin turnovers in pAJs. Furthermore, it explains how tensile actomyosin-based forces conducted by the AJ-BS can be applied to the “fluid” cadherin-based adhesive interface of pAJs: At each given moment, the interface contains many cadherin-adhesive clusters that are each linked to the tensile AJ-BS through their own tangles of F-actin. A new question now arises, however, as our findings show that there is indeed a mismatch, but it is between AJ-BT and AJ-BS F-actin dynamics. This mismatch suggests that the AJ-BS/AJ-BT interface has novel properties, which are yet to be fully understood. We speculate that the AJ-BS/AJ-BT interface may correspond to a two-sided platform, which continuously produces short-lived tangles of branched F-actin on its AJ-BT side and stable parallel filaments on the opposite, AJ-BS side. Indeed, the centripetal flow we observed for photoactivated actin in AJ-BS suggests active actin polymerization within the bundle, including at the AJ-BS/AJ-BT interface. This polymerization, together with pulling forces generated by MyoIIA, could underlie the rearward treadmilling along bundles we observed for photoactivated AJ-BS actin spots. Interestingly, both the kinetics of the loss of this spot fluorescence and its treadmilling speed depend on E-cadherin stability in pAJs. This dependence suggests that homeostasis of F-actin in the AJ-BS and AJ-BTs as well as that of cadherin clusters is interconnected, despite the fact that, as indicated above, F-actin in AJ-BT and AJ-BS use different actin polymerization pathways. This complex organization of pAJs may also be relevant to other types of AJs. Indeed, branched actin interconnecting CCCs with actin bundles have been detected in linear AJs (Efimova and Svitkina, 2018), and increased rearward treadmilling induced by the cis E-cadherin mutant was noted in actin bundles connected to E-cadherin–coated substrates (Strale et al., 2015).

Our findings reveal remarkable similarities between AJ-BSs and the dorsal actin bundles linked to FAs. Both types of bundles are enriched with Cnn2 and Cnn3, but not with MyoIIA (Tojkander et al., 2012). They both include relatively stable F-actin, which continuously polymerizes at the junction-attached end of the bundle and then flows rearward. In the case of FAs, the most active polymerization occurs at the interface between the bundle and integrin-associated proteins (Hotulainen and Lappalainen, 2006). The opposite ends of these bundles, distal from junctions, are interconnected with the actomyosin contractile system. Analogous to our observations with pAJs, all proteins within FAs turn over rapidly despite the overall morphological stability of the FA (reviewed in Geiger et al., 2009; Case and Waterman, 2015). Moreover, the adhesive regions of FAs consist of small integrin clusters (reviewed in Changede and Sheetz, 2017), while those in pAJs consist of small cadherin clusters (Indra et al., 2018). Nevertheless, specialized connecting structures analogous to AJ-BTs have not been detected for FAs. It should be noted, however, that F-actin directly associated with FAs is enriched with the same proteins as F-actin in AJ-BTs: vinculin, mena, VASP, and FBLIM1 among them.

Both AJs and FAs function as key force transmission complexes of cells. They sense tensile forces exerted on cells and instruct cell behavior in accordance with these forces (reviewed in Geiger et al., 2009; Huveneers and de Rooij, 2013; Mège and Ishiyama, 2017; Yap et al., 2018). However, our results suggest that AJs sense attached cells using a unique mechanism, which allows them to be rigid enough to transmit force while still being fluid enough to facilitate cell migration. Specifically, the half of a pAJ in one cell senses the dynamics of cadherin and F-actin in the half of the pAJ in the adjacent cell, thus allowing for synchronization of their dynamics. Potentially, this sensing, which we term “dynasensing,” may play a critical role in morphogenesis, for example, by accelerating the movement of particular types of cells within a stable epithelial sheet. The mechanism suggests how mobile cells could instruct stationary cells to form flexible AJs, which, rather than impeding cell motion, could facilitate movement. Such sensing and synchronization of junction dynamics could thus facilitate cell intercalating events during normal morphogenesis or the invasion of malignant cells.

In conclusion, we show that pAJ-linked actin bundles consist of a stable stalk and a dynamic tip. In the tip, short-lived F-actin polymerization-depolymerization cycles are tightly coupled with assembly–disassembly cycles of cadherin clusters. The coupling between E-cadherin and actin turnover allows pAJs to function not only as flexible mechanosensing connectors between the actin cytoskeleton of neighboring cells, but also as sensors and synchronizers of their cytoskeletal dynamics.

STARMETHODS

Supplementary Material

ACKNOWLEDGMENTS

Footnotes

REFERENCES