Structural basis of αE-catenin–F-actin catch bond behavior

  1. Xiao-Ping Xu
  2. Sabine Pokutta
  3. Megan Torres
  4. Mark F Swift
  5. Dorit Hanein  Is a corresponding author
  6. Niels Volkmann  Is a corresponding author
  7. William I Weis  Is a corresponding author
  1. Scintillon Institute, United States
  2. Departments of Structural Biology and Molecular & Cellular Physiology, Stanford University School of Medicine, United States
  3. Department of Structural Biology and Chemistry, Pasteur Institute, France

Abstract

Cell-cell and cell-matrix junctions transmit mechanical forces during tissue morphogenesis and homeostasis. α-Catenin links cell-cell adhesion complexes to the actin cytoskeleton, and mechanical load strengthens its binding to F-actin in a direction-sensitive manner. Specifically, optical trap experiments revealed that force promotes a transition between weak and strong actin-bound states. Here, we describe the cryo-electron microscopy structure of the F-actin-bound αE-catenin actin-binding domain, which in solution forms a five-helix bundle. In the actin-bound structure, the first helix of the bundle dissociates and the remaining four helices and connecting loops rearrange to form the interface with actin. Deletion of the first helix produces strong actin binding in the absence of force, suggesting that the actin-bound structure corresponds to the strong state. Our analysis explains how mechanical force applied to αE-catenin or its homolog vinculin favors the strongly bound state, and the dependence of catch bond strength on the direction of applied force.

Introduction

The development and maintenance of multicellular organisms depends upon specific adhesion between cells. Morphogenetic movements of sheets of cells are driven by changes in the cytoskeleton of individual cells that are linked to adjacent cells by adhesion molecules. Tissue integrity depends upon response of these adhesive structures to external mechanical perturbation (Guillot and Lecuit, 2013; Ladoux and Mège, 2017). Cell-cell junctions, including the adherens junctions (AJ), transmit mechanical forces between cells. In AJs, the extracellular domains of cadherins mediate homophilic cell-cell contact, and their cytoplasmic domains are linked to the actin cytoskeleton by β-catenin and α-catenin (Meng and Takeichi, 2009; Shapiro and Weis, 2009). Specifically, β-catenin binds to the cytoplasmic tails of cadherins and to α-catenin; α-catenin binds to β-catenin and to filamentous (F-)actin (Figure 1A). This architecture enables forces generated by actomyosin constriction to be transmitted to neighboring cells during morphogenesis, and conversely allows the actin cytoskeleton to respond to external loads. Similarly, in cell-extracellular matrix adhesions, the extracellular domains of integrins bind to components of the extracellular matrix. The cytoplasmic protein talin binds to integrins and to vinculin, a homolog of α-catenin.

α-Catenin in adherens junctions.

(A) Schematic of AJ and the role of αE-catenin in the connection to the actin cytoskeleton. The extracellular region of cadherins (green) bind to one another between cells, and their cytoplasmic domains bind to β-catenin (yellow). β-Catenin binds to α-catenin (pink/red), which binds to F-actin (orange) weakly in the absence of force (top panel). Tension (indicated by arrows) favors the strong actin-binding state of α-catenin, and also produces conformational changes in α-catenin that lead to recruitment of vinculin (light blue). While the net direction of the force is likely perpendicular to the junction (black arrows), there will be local force components along the mixed-polarity filaments toward their pointed (-) ends through actomyosin contractility (grey arrows). (B) Primary structure of αE-catenin; binding sites for β-catenin, vinculin and F-actin are indicated. (C) Crystal structure of αE-catenin ABD (Ishiyama et al., 2018) (PDB 6dv1); the five helices H1-H5, the N-terminal capping helix H0, and the C-terminal extension (CTE) are labeled.

α-Catenin appears to be the major sensor of mechanical force in the AJ. Tension on cadherins depends upon actomyosin activity, β-catenin and α-catenin (Borghi et al., 2012). As AJs develop, tension placed on α-catenin promotes conformational changes that enable it to bind to its paralog vinculin (Barrick et al., 2018; Kim et al., 2016; le Duc et al., 2010; Li et al., 2015; Maki et al., 2016; Maki et al., 2018; Terekhova et al., 2019; Yonemura et al., 2010), whose actin-binding activity further strengthens the cytoskeletal linkage (Thomas et al., 2013). In the mature AJ, actin bundles of mixed polarity run parallel to the junction (Hirokawa et al., 1983) and lie in close apposition to the membrane (Buckley et al., 2014). Single molecule force measurements of the cadherin–catenin complex binding to actin that employed an optical trap setup revealed that the complex displays catch bond behavior, wherein the interaction with actin is strengthened under mechanical load (Buckley et al., 2014). This property was observed subsequently in α-catenin and full-length vinculin themselves, as well as the actin-binding domain (ABD) of vinculin, indicating that the homologous ABDs of these proteins confer catch bond behavior (Abore et al., 2020; Huang et al., 2017). In both proteins, the catch bond is asymmetric: force directed toward the (-) end of the actin filament results in a longer lived bond than when force is directed toward the (+) end (Abore et al., 2020; Huang et al., 2017).

α-Catenin has three major domains: an N-terminal β-catenin-binding domain, a middle (M) domain, followed by a flexible linker to the C-terminal ABD (Figure 1B). The three-dimensional structures of the ABD from αE(epithelial)- and αN(neuronal)-catenins have been determined (Ishiyama et al., 2018; Ishiyama et al., 2013), and the ABD has also been visualized in the crystal structure of a nearly full-length αE-catenin (Rangarajan and Izard, 2013). These structures reveal that the ABD comprises a bundle of five helices, preceded by a short N-terminal helix (designated H0) that sits on top of the bundle (Figure 1C). Helices 2, 3, 4 and 5 (H2-5) form an antiparallel four-helix bundle in which hydrophobic residues from each helix contribute to a hydrophobic core. Helix 1 interacts with the side of the four-helix H2-H5 bundle. A long C-terminal extension (CTE), residues 844–906, follows H5. In different structures, the CTE adopts different conformations and is partly disordered. The vinculin ABD likewise adopts a similar five helix bundle architecture, albeit with a shorter H1 and no H0 (Bakolitsa et al., 2004; Bakolitsa et al., 1999; Borgon et al., 2004).

In the optical trap data, the distribution of bound lifetimes of the cadherin/catenin complex or vinculin at any particular force follows a bi-exponential distribution, indicating that there are two distinct actin-bound states, weak and strong (Buckley et al., 2014; Huang et al., 2017). The population of longer lifetimes increases with force, and modeling of these data indicated that the catch bond behavior arises because force enhances interconversion of the weakly- to the strongly bound state (Buckley et al., 2014). These observations explain why binding of the cadherin–catenin complex to actin in solution, that is, under no external load, is weak; force shifts the equilibrium between the weakly and strongly bound states and thereby produces tighter binding (Buckley et al., 2014; Yamada et al., 2005).

While the catch bonding of αE-catenin and vinculin to actin is established, to date there has been no molecular explanation of how force changes the structure of their ABDs to promote strong binding. Recent work in solution demonstrated removal of H0 from the ABD of αE-catenin enhances its affinity for F-actin, suggesting that force-dependent removal of this structural element is important for catch bonding (Ishiyama et al., 2018). However, vinculin lacks H0 yet also displays catch bond behavior (Bakolitsa et al., 2004; Bakolitsa et al., 1999; Borgon et al., 2004; Huang et al., 2017). Here, we present the structure of the αE-catenin ABD lacking H0 bound to F-actin obtained by cryo-electron microscopy (cryo-EM). The structures of the free (Ishiyama et al., 2018; Ishiyama et al., 2013; Rangarajan and Izard, 2013) and the actin-bound forms of the complete αE-catenin ABD, as well as the structures of the vinculin ABD in the absence (Bakolitsa et al., 2004; Bakolitsa et al., 1999; Borgon et al., 2004; Mei et al., 2020) or presence (Mei et al., 2020) of actin provide an explanation for the weak to strong actin-binding transition, and biochemical and mutational data support this model.

Results

Structure of αE-catenin ABD bound to F-actin

To understand the αE-catenin-actin filament interaction in molecular detail, we obtained a three-dimensional cryo-EM reconstruction of ADP-actin filaments bound to a truncated αE-catenin ABD (residues 671–906) at 3.6 Å resolution (Figure 2, Figure 2—figure supplement 1, Table 1), which we reported at lower resolution previously (Hansen et al., 2013). This construct deletes the first half of, and thereby destabilizes, the short H0, and binds 4.5x more strongly than the complete ABD (residues 666–906) (Table 2, Figure 2—figure supplement 2; Ishiyama et al., 2018). Consistent with the previously described cooperative binding by this construct observed in TIRF and cryo-EM (Hansen et al., 2013), we observed either bare actin filaments or stretches of filaments continuously bound by αE-ABD added at 10 μM (Figure 2—figure supplement 1A). At the same concentration, we were unable to observe binding of the complete ABD to actin filaments in the electron microscope (Figure 2—figure supplement 1B). Given the KD values of these two constructs (Table 2), small changes in concentration likely have a large effect on actin decoration. Indeed, a comparable structure using an ABD construct spanning residues 664–906 was produced using 20 μM ABD (Mei et al., 2020).

Figure 2 with 5 supplements see all
Cryo-EM analysis of the αE-catenin ABD–F-actin complex.

(A) Cryo-EM map of the actin-ABD structure. The segmented ABDs are shown in magenta. The (-) end of the filament is shown at the top, and the (+) end at the bottom. (B) Molecular model of a section of an actin filament bound to αE-catenin ABDs, same orientation as (A) and with transparent density map overlaid. Actin protomers are colored according to their long-pitch helix in blue and yellow. The bound ABDs are shown in magenta and pink. (C, D) Closeups of model and cryo-EM map showing residues on H4 (panel C) and the CTE (panel D) that have been studied by site-directed mutagenesis. The ABD is shown in red, and two monomers of actin in different shades of blue. In (D), a neighboring copy of the ABD along the filament is shown in pink. Actin residue labels are italicized.

Table 1
Cryo-EM data collection and model statistics.
Data collection
MicroscopeTitan Krios
Voltage [kV]300
DetectorFalcon II
Magnification75,000
Exposure parameters
 Total dose [e-2]60
 Exposure time [s]1.0
Pixel size [Å]1.035
Defocus range [μm]−0.8 to −2.8
Data processing
Images used4769
Initial segments728,331
Final segments422,822
Helical symmetry
 Rise [Å]27.4
 Twist [°]−166.9
Resolution [Å]3.6
FSC threshold0.143
Sharpening B-factor [Å2]−96
Refinement
Initial models [PDB IDs]6djo, 6dv1
Non-hydrogen atoms25,248
Model resolution [Å]3.5
FSC threshold0.5
RMS deviations
 Bond lengths [Å]0.007
 Bond angles [°]0.92
Rotamer outliers [%]4.4
Mean B-factor [Å2]70.4
Validation
Molprobity score2.08
Clash score6.28
Ramachandran plot
 Favored [%]96.3
 Allowed [%]3.7
 Disallowed [%]0.0
CC (mask)0.86
CC (volume)0.82
EMringer score2.29
Table 2
Affinities of αE-catenin ABD constructs for actin, determined by co-sedimentation.

KD values and standard deviations for αE-catenin 666–906, 671–906, and 692–906 are the average of three replicate measurements. KD values for the other constructs are the average of two measurements. For αE-catenin 671–906 W859A binding did not reach saturation and therefore only a lower limit for the KD is given. N.D., no detectable binding. Representative binding curves and corresponding gels are shown in Figure 2—figure supplements 2 and 4.

αE-catenin ABD variantKD (μM) (SD)
666–906 (full length)8.5 (0.7)
671–9062.0 (0.3)
692–9060.7 (0.3)
696–9060.4 (0.05)
699–9060.5 (0.1)
671–8725.0 (0.1)
671–868N.D.
671–864N.D.
699–8682.8 (0.1)
671–906 W859A>35

The cryo-EM reconstruction allowed us to build an atomic model of the complex (Figure 2), using the structures of bare ADP actin filaments (Chou and Pollard, 2019) and the αE-ABD crystal structure (Ishiyama et al., 2018) as starting points. The EM map was poor in the αE-catenin CTE, and we were able to correct a sequence registration error in our first model based on the coordinates of the 3.2 Å resolution structure reported by Mei et al., 2020 provided by Dr. G Alushin. Only the first three and last residues of actin could not be placed with confidence. For the αE-catenin ABD, there was no detectable density for residues 671–698, or from 872 to 906. In addition, six residues in the loop connecting H4 and H5 could not be modeled. Local resolution analysis shows that the most well-defined region is within the actin filament core and gradually falls off toward larger radii (Figure 2—figure supplement 1C). Importantly, the interface of the ABD and F-actin is well defined in the cryo-EM map, as is the conformation of the helical bundle.

The helical rise (27.4 Å) and twist (−166.9°) of F-actin in the reconstruction are practically identical to those of bare F-actin (Chou and Pollard, 2019; Merino et al., 2018), indicating that the ABD does not induce any major changes into the filament. Consequently, the root-mean square deviation between bare ADP actin filaments and actin with the ABD bound is low (0.76 Å, Cα deviations). The most notable difference is in the conformation of the subdomain two loop (D-loop), a region implicated in changes associated with the ATP hydrolysis cycle (Chou and Pollard, 2019; Merino et al., 2018), stiffness and stability (Kang et al., 2012; Pospich et al., 2017), and in filament disassembly (Grintsevich et al., 2017). In the present structure, this region is in a ‘closed’ conformation similar to that observed in the bare ADP-bound F-actin structures (Chou and Pollard, 2019; Merino et al., 2018). Modeling suggests that the alternative ‘open’ D-loop conformation that occurs in equilibrium with the ‘closed’ conformation in other nucleotide states (Merino et al., 2020) may clash with the bound αE-catenin. Relative to the bare ADP-actin structure, however, D-loop residues 45–50 move significantly, with M47 showing the largest displacement of about 5 Å. This region contacts the αE-catenin ABD. It has been noted that tensile forces imposed on actin by the thin ice needed for cryo-EM imaging may affect the D-loop (Galkin et al., 2012), and we have previously observed similar differences at low resolution upon αE-catenin binding (Hansen et al., 2013), but whether tension has any role in the conformation observed here is unclear given the direct contacts with the ABD. Moreover, the refinement procedure used to generate high-resolution structures from cryo-EM images selects and enforces a single conformation (see Methods), so it is likely that in order to achieve the highest resolution reconstruction possible, other information content including alternative conformations, was lost.

In contrast to the local changes in F-actin, the ABD undergoes large structural rearrangements upon complex formation. Compared to the unbound ABD crystal structure, the N-terminus through the last turn of H1 is disordered (Figure 3A). We note that an F-actin-bound αE-catenin ABD structure has been reported recently for the complete ABD (664-906) (Mei et al., 2020), and the same residues are disordered, demonstrating that the truncation of H0 in our construct has no influence on the actin-bound structure. The remaining four helices rearrange to bind to the filament. A key part of this change involves the long loop that connects H4 and H5, where the first strand of a β-hairpin becomes a two-turn extension of H4 (residues 795–801) that forms contacts with actin (Figures 2C and 3A,B).

Figure 3 with 2 supplements see all
Overall changes in ABD structure upon binding to F-actin.

Comparison of the unbound ABD crystal structure (PDB 6dv1; light orange) with the ABD in the actin-bound state (magenta). (A) Overall comparison; the orientation is rotated approximately 180° from that shown in Figure 1c. (B) Top view of H0 packing interactions lost upon its removal, and rearrangements of helices 2–5. The left panel depicts packing interactions of H0 residues I672, M673 and L676 with H5 residues V809, G811 and A815 (all highlighted in orange), and the right panel overlay shows the resulting changes in H4 and H5.

In addition to the changes in the helical bundle, the remaining turn of H1 moves up from the bottom of the rest of the helical bundle (Figure 3 and Figure 3—figure supplement 1). This moves W705 (H1) away from an aromatic cluster formed with Y837 (H5), and W859 (CTE), and shifts the CTE upward such that M861 of the CTE now packs with Y837 and W859 (Figure 3—figure supplement 1). The CTE past 862, which is disordered in the isolated structure, forms an extended peptide that interacts with actin (Figure 2D). The backbone carbonyl oxygens of αE-catenin K862 and P864 form hydrogen bonds with actin residues R28 and R95 (Figure 2—figure supplement 3A). αE-catenin K866 forms salt bridges with actin D24 and D25, K867 forms a hydrogen bond with the backbone carbonyl of R28 and contacts V30, and L869 packs against actin residues A26, P27, E334 and Y337 (Figure 2—figure supplement 3A). The importance of the change in the CTE and formation of contacts with F-actin is highlighted by the effect of mutating W859 to alanine, which lowers the affinity for actin approximately 10-fold (Table 2, Figure 2—figure supplement 4). Finally, we note that the CTE sits between actin and the next ABD along the long pitch of the filament, and V870 packs against V800 at the top of H4 of the neighboring ABD. This interaction may contribute to the cooperative binding of the ABD to actin (Hansen et al., 2013).

The actin-bound αE-catenin ABD complex structure is consistent with previously reported mutations that weaken its binding to F-actin. Ishiyama et al., 2018 mutated H4 residues L785, I792 and V796 to alanine. L785 had the most severe effect, reducing the KD ~15 x; it interacts at the interface of two longitudinally adjacent actin monomers, including L349 of one monomer and V45 and G46 in the D-loop of the other (Figure 2C). I792A and V796A also had significant effects on affinity; I792 forms a packing interaction with I345 of actin and V796 packs against actin residues I341 and D25 (Figure 2C). Chen et al., 2015 found several point mutants that severely weakened binding, including I792A. K842A eliminates a hydrogen bonds with H87 and Y91 of actin (Figure 2—figure supplement 3B); and K866A eliminates salt bridges with actin D24 and D25 (Figure 2—figure supplement 3A). Pappas and Rimm, 2006 removed sets of positively charged residues and saw only modest effects on binding; the largest effect was the triple mutant K747A/K748A/K797A; K797 packs into the hydrophobic core of the bundle and forms a salt bridge with actin E334, whereas the other two lysine residues point into solvent on the other side of the domain (Figure 2—figure supplement 3C).

Deletion of the C-terminal αE-catenin residues 865–906 compromises actin binding (Pokutta et al., 2002), whereas a construct ending at 883 binds (Chen et al., 2015; Pappas and Rimm, 2006). This observation is consistent with the contacts observed between actin and residues 866–869 (Figure 2D, Figure 2—figure supplement 3A). To more precisely determine which residues of the αE-catenin CTE observed to contact actin are critical for binding, we prepared a series of C-terminal truncations of the αE-catenin 671–906 construct and compared their affinities (Table 2, Figure 2—figure supplement 4). We confirmed that removing residues 865–906 produced no detectable binding. Removal of residues 869–906 eliminated detectable F-actin binding from the variant starting at 671, and when these residues are removed from the higher affinity variant lacking H1 (starting at 699), binding is detectable but 5.6x weaker. These findings confirm the contributions of L869 and V870 to binding (Figure 2—figure supplement 3A). Surprisingly, there is a slight loss of affinity upon removal of residues 873–906, even though we do not observe these residues in the structure. Mei et al., 2020 proposed that these C-terminal residues absent in the structure may mediate a small increase in affinity when actin is placed under tension. However, deleting these residues weakens the affinity (albeit slightly) in a solution assay, which indicates that they have a role independent of tension. It is possible that these are highly dynamic interactions that are not sufficiently stable to be visualized in the cryo-EM structure.

The structural changes and interactions with actin observed here appear to be conserved throughout the α-catenin/vinculin family. Specifically, the residues that mediate the interactions with actin are strongly conserved throughout the α-catenin sequences (Figure 2—figure supplement 5). Moreover, although relatively few of the actin-contacting residues are conserved in vinculin (notably, those in the C-terminal portion of H4), the vinculin ABD undergoes a similar structural transition upon binding to actin (Figure 3—figure supplement 2; Kim et al., 2016; Mei et al., 2020).

H0 and H1 regulate actin affinity

Despite the large changes between the free and F-actin-bound ABD structures, we see no evidence for multiple conformations of the ABD when bound to F-actin, suggesting that the four-helix state is the stably bound one. To assess whether the rearranged state is significantly populated in solution, we compared the proteolytic sensitivity of the H0-deleted ABD used in the EM structure (671-906) in the presence and absence of F-actin, using the protease elastase. We found that in isolation the domain was resistant to digestion, whereas binding to F-actin led to the appearance of smaller, protease-resistant fragments (Figure 4A). N-terminal sequencing of the SDS-PAGE bands corresponding to these fragments revealed cleavage of H1 at residues A689 and S703 (Figure 4B). The cleavage at S703 is consistent with the very weak cryo-EM density observed between residues 699 and 702, which likely indicates that this turn of helix is flexible. Moreover, the cleavage at 689 suggests that H1 becomes unstructured and flexible when dissociated from the H2-5 bundle. The resistance of H1 to protease in the absence of F-actin implies that its association with the H2 and H5 surface is strongly favored in solution. Helix 1 of the vinculin actin-binding domain, which forms a similar four helix bundle when bound to actin (Figure 3—figure supplement 2; Kim et al., 2016; Mei et al., 2020), also becomes proteolytically sensitive upon binding to actin (Bakolitsa et al., 1999). The proteolysis data from both αE-catenin and vinculin suggest that the free energy of binding to actin drives the structural rearrangement of the ABD.

Stability of the H1- H2/H5 interface.

(A) Time course of elastase digestion of αE-catenin 671–906. The two smaller fragments analyzed by N-terminal sequencing are indicated with asterisks. (B) Rainbow diagram of the unbound αE-catenin ABD (PDB 6dv1), colored as in Figure 1c. Residue labels for the H5 (red) helix are shown in white for clarity. The two residues at the elastase cut sites are indicated in gold. Side chains in the H1- H2/H5 interface are shown in stick representation.

If the association of H0 and H1 with the four-helix bundle inhibits the rearrangement of the structure to a stable actin-bound conformation, they should weaken the affinity of the ABD for actin. Therefore, we prepared a series of αE-catenin ABD variants in which H0 and H1 were deleted or truncated. As more N-terminal sequence was deleted, the affinity became stronger, such that deleting H1 through residue 698 results in approximately 18x stronger binding to actin filaments relative to the complete ABD (Table 2, Figure 2—figure supplement 2). Notably, Ishiyama et al. deleted H0 only and saw about a ~ 3 x increase in affinity (Ishiyama et al., 2018), similar to change we observed when H0 is disrupted by deleting residues 666–670 (Table 2, Figure 2—figure supplement 2). This observation is also consistent with the observation that the crystal structure of the αN-catenin ABD lacking H0 still forms the 5-helix assembly observed in the complete ABD (Ishiyama et al., 2018). Overall, the deletion data indicate that removal of H0 and most of H1 produce a strong actin-binding species. Notably, the enhanced binding conferred by deleting H1 can compensate for the loss of residues 869–906 (Table 2; compare 671–868 and 699–868).

Sequence features of the ABD support the idea that the free energy of binding to actin drives dissociation of H0 and H1 and rearrangement of the remaining helices. H0 has three conserved hydrophobic residues (I672, M673 and L676) that pack against three residues at the N-terminal region of H5 (Figure 3B), two of which (V809 and G811) are poorly ordered in the actin-bound structure. H1 binds to the outer face of the H2–H5 four helix bundle, interacting with a surface formed by H2 and H5. Several hydrophobic H1 residues (I684, V688, F691 and L698) are buried in this interface (Figure 4B), which would disfavor dissociation of H1. However, the H1 interaction surface formed by H2 and H5 is not strongly hydrophobic, comprising four methionine residues (M723, M724, M816 and M826), C720, T727, I819, A822 and the aliphatic portion of K823 (Figure 4B), which suggests that there would not be a large destabilization of the four-helix bundle upon removing H1 from this surface. Indeed, the construct starting at 699, which deletes all of the H1 sequence missing in the EM structure, is well behaved in solution (it is monomeric as assayed by size exclusion chromatography-coupled multi-angle light scattering; data not shown), consistent with the idea that exposure of this surface is not energetically disfavored. Notably, the mildly hydrophobic character of this H2/H5 surface is strongly conserved throughout the α-catenin family (Figure 2—figure supplement 5).

Insights into catch bond mechanism

Our structural and biochemical data indicate that removal of H0 and H1 enable the structural transition of the C-terminal half of H4 and movement of the CTE (Figure 3), which result in additional contacts with F-actin and stable binding (Figures 2 and 5). Given that the vinculin ABD lacks H0 but its structure bound to actin shows the same four-helix, rearranged bundle relative to vinculin in solution (Mei et al., 2020), and that the crystal structure of the αN-catenin ABD lacking H0 retains the five-helix bundle architecture of the full ABD (PDB 6duw, 6duy) (Ishiyama et al., 2018), it is clear that removal of H1 is the major determinant in achieving the high-affinity actin-binding state. The binding data in solution, that is, in the absence of applied force, demonstrate that the free energy of binding to actin drives this transition. Optical trap data from both αE-catenin and vinculin indicate that force enhances the stability (specifically, the bound lifetime) of the ABD-actin interaction (Abore et al., 2020; Buckley et al., 2014; Huang et al., 2017). Modeling of the optical trap data indicated that the major effect of force is to prevent the transition from a strongly bound to a weakly bound state (Buckley et al., 2014; Huang et al., 2017). Force also promotes the transition to a strongly bound state, although this is a smaller effect (Buckley et al., 2014; Huang et al., 2017) that is consistent with the binding energy to actin driving the rearrangement to the strong state. Thus, while force is not needed to stably bind actin, it enhances the strength of the ABD-actin interaction by shifting the equilibrium between weakly and strongly bound states toward the strong state.

Figure 5 with 1 supplement see all
Model of the weak and strong actin-binding states of αE-catenin.

(A) Superposition of the isolated αE-catenin ABD on the actin-bound structure reveals no major clashes with F-actin (left panel). When the ABD is bound to F-actin, H0 and H1 dissociate from the H2-5 bundle, which results in the extension and shift of the C-terminal part of H4 as well as ordering and repositioning of the CTE to bind to actin. (B) Schematic diagram of αE-catenin ABD conformational states when unbound, weakly bound, and strongly bound to actin. Cooperative binding of the ABD, as observed in the cryo-EM structure, is illustrated for the strong state. Note that in the strong state, the H0 and H1 regions are drawn as helices when dissociated from the H2-5 bundle, but it is likely that they are unstructured in this case (see text for details).

To gain structural insight into how force affects the weak and strong actin-binding states of the ABD, we superimposed the isolated αE-catenin ABD structure (PDB 6dv1) on the actin-bound version (Figure 5A, Figure 5—figure supplement 1). This revealed clashes of actin residues K328, I330 and P333 (Figure 5—figure supplement 1) with ABD residues D813 and M816 in the first turn of H5, as well as E799 in the H4-H5 connection and K683 of H1. Modeling indicates that a few other minor clashes can be alleviated by changes of side chain rotamers. Given the proteolytic digestion data suggesting that H1 dynamically associates with the H2-H5 region (Figure 4) and that crystal structures of the αE- and αN-catenin ABDs in isolation show variability in the H4-H5 connection (Ishiyama et al., 2018), it seems likely that small changes (on the order of 1 Å) in this region could accommodate the actin surface without the wholesale conformational changes that produce the four-helix conformation. Moreover, as the N-terminal half of H4 and almost all of H5 of the isolated and actin-bound ABD structures superimpose closely (Figure 5A, middle), it is likely that in the five-helix conformation these regions could form interactions similar to those visualized in the EM structure. However, key interactions made by the C-terminal part of H4, including those of I792 and V796 (Figure 2C), would not form. These observations suggest that with small changes of H1 and the first turn of H5, the five-helix conformation could bind F-actin weakly, and we propose that this conformation represents the weak binding state observed in the optical trap (Figure 5B).

Assuming that the four-helix conformation observed in the complex with F-actin represents the strong state, it is likely that force on the ABD prevents H1 and H0 from associating with the rest of the ABD (Figure 6A, Video 1). Tension on the N-terminus of the ABD that is stably bound to F-actin will prevent re-association of H1 and H0, thereby favoring the strongly bound state and enhancing its lifetime. Conversely, if the five-helix conformation binds weakly, force would provide additional energy to drive the dissociation of H0 and H1 from the bundle (Figure 6B, Video 2), thereby facilitating the transition to the strongly bound state. The dissociation of H1 also enables the ordering of the CTE, which contributes to high-affinity binding (Table 2) through its direct interactions with actin and possibly by contributing to cooperative binding through its interaction with the longitudinal ABD neighbor on actin (Figure 2, Figure 2—figure supplement 3A, Figure 5B).

Model of directional catch bonding.

Actin and the αE-catenin ABD in the strong and weak states are illustrated as in Figure 5B. The N-terminus of the ABD is shown tethered to a stationary point, that is, as part of the cadherin/β-catenin/α-catenin complex. The grey arrows indicate the direction of force. (A) Tension applied to the bound strong state prevents re-binding of H0/H1 to the H2-H5 bundle. Force applied in the (-) direction will move the H1 sequence away from the H2-H5 bundle and place this region in an unfavorable orientation for rebinding, whereas force directed in the (+) direction will place the H1 sequence closer to and in a more favorable orientation for rebinding. See Video 1 for an animated version. (B) Tension applied to the bound weak state will remove H0/H1 from the H2-H5 bundle. Force applied in the (-) direction will tend to pull H0/H1 away from the H2-H5 bundle, whereas force in the (+) direction is predicted to have a smaller effect on H0/H1 dissociation. See Video 2 for an animated version.

Video 1
The strong to weak transition.

Animation illustrating changes in the αE-catenin ABD going from the strong to weak actin-bound state. The first half of the movie shows force is applied in the (-) direction, and the second half shows force in the (+) direction, highlighting the difference in distance for reassociation of H0 and H1 in the two directions.

Video 2
The weak to strong transition.

Animation illustrating changes in the αE-catenin ABD in going from the weak to strong actin-bound state. The first half of the movie shows force is applied in the (-) direction, and the second half shows force in the (+) direction, highlighting the difference in distance for reassociation of H0 and H1 in the two directions.

Discussion

Catch bond behavior has been observed in a number of proteins subject to mechanical force in a variety of biological contexts. Examples include the extracellular portions of cell adhesion molecules such as bacterial FimH that attaches to the urinary tract epithelium, selectins that mediate rolling of leukocytes on endothelia, integrins that mediate cell-extracellular matrix adhesion, and classical cadherins (Pruitt et al., 2014). For FimH, selectins and integrins, hinges between domains change position under mechanical load, and these changes can be transmitted to binding domains in a variety of ways to promote a strong ligand-binding state (Le Trong et al., 2010; Pruitt et al., 2014; Rakshit et al., 2012; Springer, 2009; Springer et al., 2008; Xiao et al., 2004). More recent work has revealed catch bonding to F-actin by the intracellular adhesion proteins αE-catenin and vinculin. These proteins also have multiple domains that are likely to change their relative positions upon application of force, as has been demonstrated for αE-catenin (Barrick et al., 2018; Choi et al., 2012; Kim et al., 2016; Li et al., 2015; Terekhova et al., 2019). Attachment of the α-catenin N-terminal domain to β-catenin and the C-terminal ABD to actin (Figure 1B) implies that tension is transmitted through the entire protein.

The structure of the actin-bound αE-catenin ABD presented here provides a molecular-level explanation of the catch bond behavior of αE-catenin as well as the homologous vinculin (Abore et al., 2020; Buckley et al., 2014; Huang et al., 2017). By serving as a bridge between β-catenin and actin, αE-catenin is placed under tension, which likely stretches the linker between the M domain and the ABD and thereby applies tension to the ABD H0 and H1 regions. Likewise, binding of vinculin to talin and actin in focal adhesions will stretch the loop that precedes the ABD. Tension stabilizes the four-helix, strong-binding conformation of these ABDs bound to F-actin by preventing rebinding of H1 (and H0 in the case of αE-catenin) (Figure 6A, Video 1). The stabilization of the αE-catenin CTE in the four-helix conformation is likely an important component of the strongly bound state: it not only forms interactions with actin but also contacts a neighboring ABD on the filament, which may underlie its cooperative binding to F-actin (Buckley et al., 2014; Hansen et al., 2013). Moreover, modeling of the isolated αE-catenin ABD structure on actin shows few clashes and suggests that a five-helix conformation close to that of the isolated structure forms a subset of the interactions observed in the cryo-EM structure, and this likely corresponds to the weakly bound state. If so, then tension applied to the weakly bound state could facilitate dissociation of H0 and H1 and therefore transition to the strongly bound conformation (Figure 6B, Video 2).

It has been noted that tension applied to an extended peptide can remove this element from another part of the same protein or a partner, and that the mechanical properties of the now flexible, disordered polypeptide contribute to catch bond behavior (Guo et al., 2019; Guo et al., 2018; Yuan et al., 2017). While beyond the scope of this work, it is interesting to consider that the CTE has such a role in the catch bonding of the ABD. In this case, its contribution is difficult to clearly define since its ordering in the presence of actin appears to be intimately coupled to the change in H1. Moreover, the proteolysis data (Figure 4A) suggest that H1 becomes disordered in the transition from the strongly to weakly bound state. If H1 (and possibly H0) unfold when detached from the rest of the ABD, it would provide a flexible linker that extends in the direction of force. Future work using molecular dynamics simulations and optical trap studies should allow us to investigate this aspect of the ABD.

Optical trap and biolayer interferometry measurements indicate that in the absence of force, 80–90% of the actin-bound αE-catenin molecules are in the weak state (Buckley et al., 2014; Ishiyama et al., 2018). The presence of a small fraction in the strong state implies that the free energy of binding to actin promotes dissociation of H1 even in the absence of force (Figure 6A). Our deletion data indicate that removal of H1 does not destabilize the rest of the ABD, but the burial of several hydrophobic H1 residues in the H2/H5 interface (Figure 4B) suggests that H1 dissociation would be disfavored. The energy input by mechanical force acting over a certain distance helps to overcome this barrier.

A recent study proposed that removing H0 enables the strongly bound, force-enhanced state of αE-catenin (Ishiyama et al., 2018), although the structure of the actin-bound ABD showing the dissociated H1 and rearranged H2-H5 region was not available to these authors. Their proposal was based in part on the observation that in order to fit biolayer interferometry F-actin binding data for the complete ABD, a model invoking two species with different affinities was needed (Ishiyama et al., 2018). The increased affinity without H0 (Ishiyama et al., 2018; Table 2) shows that the stabilization energy provided by the interaction of H0 with the rest of the molecule contributes to the barrier to the transition to the stably bound conformation in αE-catenin, as H0 likely interferes with the rearrangements of H2–H5 (Figure 3B). However, the fact that the crystal structure of a mutant αN-catenin ABD lacking H0 is not significantly different from the crystal structure of the complete ABD (Ishiyama et al., 2018) indicates that removing only H0 does not produce the rearranged, strong-binding four-helix bundle conformation in solution. Moreover, the vinculin ABD lacks H0, yet also forms catch bonds with actin and forms a four-helix bundle similar to that of αE-catenin when bound to F-actin (Huang et al., 2017; Kim et al., 2016; Mei et al., 2020). Also, both the vinculin ABD and αE-catenin ABD lacking H0 are resistant to protease digestion when not bound to actin (Figure 4; Bakolitsa et al., 1999). Finally, removal of H0 and H1 produces a significantly higher affinity for F-actin than removing only H0 (Table 2). These observations imply that dissociation of H1 is the major determinant of stable binding to actin. The free energy of binding to F-actin drives H1 dissociation, and mechanical force further enhances the dissociated state. Although it is not apparently integral to the catch bond behavior, we speculate that the conserved H0 of α-catenin serves to tune the stability and force response of the ABD; in the disordered state bound to actin, it will provide additional flexible elements that may affect the detailed catch bond behavior.

The catch bonds formed by αE-catenin, vinculin and their ABDs with actin show a strong asymmetry (Abore et al., 2020; Huang et al., 2017) (N. Bax, D. Huang, A. Wang, A. Dunn, and W.I.W., manuscript in preparation). Specifically, force directed toward the (-) end of the filament greatly enhances the strongly bound, long-lived state, whereas force toward the (+) end has a more modest effect on bound lifetimes. (We originally reported [Buckley et al., 2014] that the catch bonding of the cadherin/β-catenin/αE-catenin complex was not asymmetric, but this proved to be due to limitations in the sensitivity of the instrument used in that study; N. Bax, D. Huang, A. Wang, A. Dunn, and W.I.W., manuscript in preparation). The lifetimes of the weak and strong actin-bound states observed in the optical trap were described by a modified Bell model (Bell, 1978Evans, 2001) in which the transition state energy depends on the force acting over the distance from the ground to transition state (Huang et al., 2017). The dependence of bound lifetime on force was the same for force directed toward either the (+) or (-) end of the filament, implying that there is a different distance to the transition state in the two directions (Huang et al., 2017). Although we do not have a detailed molecular model for the transition between weak and strong states, we note that the force vector experienced by α-catenin or vinculin is not necessarily aligned with the actin filament and will depend on the way the molecule is tethered. Thus, with the rest of α-catenin or vinculin bound to a stationary anchor (i.e. cadherin/β-catenin or integrin/talin), force in the (-) direction of the filament would result in pulling the N-terminus of the ABD in the opposite direction, positioning H1 away from the H2-H5 bundle and disfavoring its rebinding (Figure 6A, Video 1). Likewise, force applied to the weakly bound five-helix conformation would result in H1 ‘peeling off’ from the bundle (Figure 6B, Video 2). Force applied in the opposite (+) direction would tend to move the N-terminus of the ABD such that H1 would be more aligned with its orientation found in the five-helix state, giving it a higher probability of rebinding to the H2-H5 bundle (Figure 6A, Video 1), yielding a smaller effect of force in stabilizing the strong state in this direction. While myosin II contractility also applies force in the (-) direction of actin filaments, the actual direction of the applied force vectors for both αE-catenin and vinculin depends on the detailed geometry of the full adhesive complexes and their organization in the cell. Determining these parameters will be needed to fully understand catch bonding by these proteins.

Materials and methods

Key resources table
Reagent type
(species) or
resource
DesignationSource or
reference
IdentifiersAdditional
information
Recombinant DNA reagentpGEX-TEVChoi et al., 2012
https://doi.org/10.1074/jbc.M511338200
Ampicillin resistance; expression in bacterial cultures; pGEX-KG plasmid (ATCC) with a new TEV protease site
Contact Weis lab for distribution
Recombinant DNA reagentpGEX-4T-3GE Healthcare28-9545-52Vector for thrombin-cleavable GST fusion protein expression in bacteria
Strain, strain background (Escherichia coli)BL21 (DE3) Codon-Plus RILAgilent230245Strain for expressing recombinant proteins
Software, algorithmRELION3 3.0.8He and Scheres, 2017
https://doi.org/10.1016/j.jsb.2017.02.003
Scheres, 2012https://doi.org/10.1016/j.jmb.2011.11.010
Zivanov et al., 2018https://doi.org/10.7554/eLife.42166
RRID:SCR_016274
Software, algorithmMotionCor2 1.3.0Zheng et al., 2017
https://doi.org/10.1038/nmeth.4193
RRID:SCR_016499
Software, algorithmGctf 1.06Zhang, 2016
https://doi.org/10.1016/j.jsb.2015.11.003
RRID:SCR_016500
Software, algorithmCTFFIND4 4.1.5Rohou and Grigorieff, 2015https://doi.org/10.1016/j.jsb.2015.08.008RRID:SCR_016732
Software, algorithmpyCoAn 0.3.0Volkmann and Hanein, 1999 https://doi.org/10.1006/jsbi.1998.4074Revision 1419
Software, algorithmPhenix 1.17.1Afonine et al., 2018https://doi.org/10.1107/S2059798318006551RRID:SCR_014224
Software, algorithmCoot 0.8.9Emsley et al., 2010https://doi.org/10.1107/S0907444910007493RRID:SCR_014222
Software, algorithmEMRingerBarad et al., 2015
https://doi.org/10.1038/nmeth.3541
Via Phenix
Software, algorithmMolProbityChen et al., 2010
https://doi.org/10.1107/S0907444909042073
RRID:SCR_014226
Via Phenix
Software, algorithmResMap 1.14Kucukelbir et al., 2014
https://doi.org/10.1038/nmeth.2727
Software, algorithmSBGridMorin et al., 2013https://doi.org/10.7554/eLife.01456RRID:SCR_003511
Software, algorithmUCSF Chimera 1.14Pettersen et al., 2004https://doi.org/10.1002/jcc.20084RRID:SCR_004097
Software, algorithmGraphPad Prism 8.0.2GraphPad Software, IncVersion 263
RRID:SCR_002798

Expression and purification of αE-catenin ABD constructs

Request a detailed protocol

αE-catenin ABD constructs were cloned into a pGEX-4T-3 or pGEX-TEV bacterial expression vector; the latter is a modified pGEX-KG vector with a tobacco etch virus (TEV) protease recognition site inserted after the GST-tag and the thrombin cleavage site. There are four additional amino acids are left at the N-terminus after Thrombin or TEV cleavage (GSPN in case of the pGEX-4T3 vector and GGIL in case of the pGEX-TEV vector). N-terminal GST-fusion proteins were expressed in E. coli BL21cells. Cells were grown at 37°C to an OD600 of 0.8–1.0 and induced overnight at 18°C with 0.5 mM isopropyl-1-thio-β-d-galactopyranoside. Cells were harvested by centrifugation and pellets were resuspended in 20 mM Tris pH 8.0, 200 mM NaCl and 1 mM DTT. Before lysis in an Emusliflex (Avastin), protease inhibitor cocktail (Mixture Set V, Calbiochem) and DNase (Sigma) were added. After centrifugation at 38,500 × g for 30 min, the lysate was incubated with glutathione-agarose beads for 1 hr at 4°C. After washing the beads with PBS containing 1 M NaCl and 1 mM DTT, the beads were equilibrated with either 20 mM Tris pH 8.5, 150 mM NaCl, 1 mM DTT for thrombin cleavage or 20 mM Tris pH 8.0, 150 mM NaCl, 1 mM DTT, 1 mM EDTA, 10% glycerol for TEV cleavage. The protein was cleaved for 2 hr at room temp with thrombin or overnight at 4°C with TEV. Cleaved protein was eluted from the beads and further purified on a cation exchange column (Mono S 10/100, GE Healthcare) in MES pH 6.5, 1 mM DTT buffer with a 0–500 mM NaCl gradient, followed by size exclusion chromatography (Superdex S200, GE Healthcare) in 20 mM HEPES pH 8.0, 150 mM NaCl, 1 mM DTT.

Actin-binding assay

Request a detailed protocol

G-actin prepared from rabbit muscle (Spudich and Watt, 1971) was stored in 40 μM aliquots at −80°C. Frozen aliquots were thawed on ice and centrifuged for 20 min at 140,717 x g in a Beckman TLA 100 rotor. After centrifugation, the concentration was determined by UV absorbance at 290 nm and G-actin was polymerized by addition of 10x F-buffer (100 mM pH 7.5 Tris, 500 mM KCl, 20 mM MgCl2, 10 mM ATP) and incubation for 1 hr at room temperature. Aliquots from the same batch of actin were used for all polymerization assays, and efficient polymerization was confirmed by pelleting at 20 min at 140,717 x g and analysis of the supernatant and pellet by SDS page. F-actin was stored for up to 2 weeks at 4°C. For sedimentation assays, F-actin was diluted to 4 μM with buffer A (20 mM HEPES pH 8.0, 150 mM NaCl, 1 mM DTT, 2 mM MgCl2, 0.5 mM ATP, 1 mM EDTA). A dilution series of purified αE-catenin ABD in 20 mM HEPES pH 8.0, 150 mM NaCl and 1 mM DTT was set up and an equal volume of 4 μM F-actin or buffer A was added. The mixture was incubated for 30 min at room temperature. Samples were centrifuged in a Beckman TLA100 rotor at 140717 x g for 20 min at 4°C. The supernatant was carefully removed and the pellet resuspended in reducing Laemmli buffer. Samples were run on SDS PAGE. Coomassie-stained bands were scanned and quantified on a LI-COR Odyssey scanner (LI-COR Biosciences). To extrapolate concentration from band intensity a dilution series of αE-catenin ABD was run in parallel for each assay and stained and destained under the same conditions as the assay itself. To correct for SDS-PAGE loading errors, for each concentration of αE-catenin, its band intensity was normalized by the ratio of the actin band intensity at that point and the average actin band intensity calculated over all concentration points. The data were analyzed in the program GraphPad Prism (GraphPad Software, La Jolla, CA,) and fitted with a ‘single binding with Hill coefficient’ model, with the exception of the αE-catenin 671–906 W859A mutant. In that case, the curves did not reach saturation, and fitting with a Hill coefficient was not possible, so a 'One site-specific binding’ model was used to obtain KD estimates. In this case the binding is sufficiently weak that we report a lower limit on the KD rather than a specific value (Table 2).

Electron cryo-microscopy sample preparation

Request a detailed protocol

Rabbit skeletal actin was prepared as described (Kang et al., 2012; Spudich and Watt, 1971) and was used within 1 week of preparation. Fresh complete (residues 666–906) or truncated αE-catenin-ABD (residues 671–906) were used within 1–2 days of preparation. Both filamentous actin and the respective αE-catenin ABD construct were diluted into KMEI buffer (10 mM Imidazole pH 7, 50 mM KCI 2 mM MgCI2, 1 mM EGTA, 0.2 mM ATP, 2 mM DTT) at 0.125 mg/ml actin and 0.25 mg/ml ABD, corresponding to an ABD concentration of 10 μM. After 10 min of incubation, 5 or 4 μl from the final 1:2 (wt/wt) mixture was applied to plasma cleaned C-flat copper grids 2/1 or 2/2 (Protochips Inc), respectively. After 1 min of incubation in a humidified chamber, excess liquid was manually blotted, and the samples were plunge-frozen in liquid nitrogen-cooled liquefied ethane using an in-house designed cryo-plunger.

Screening for the best sample mixture ratios and blotting conditions was performed on a Tecnai Spirit T12 electron microscope (ThermoFisher Scientific) equipped with Eagle CMOS imaging device (ThermoFisher Scientific), operated at a voltage of 120 kV and a defocus between −1.5 and −2.5 μm. Micrographs were visually inspected for quality of filaments, filament density, background, and the presence of bound ABD. The choice of samples for data collection was based on evaluation of these parameters. Data sets were acquired on Titan Krios electron microscope (ThermoFisher Scientific) equipped with an XFEG and operated at a voltage of 300 kV. Although the sample preparation protocol was optimized, we had to screen for usable grids and grid squares manually. Images were recorded on a Falcon II direct detection camera (ThermoFisher Scientific) under minimal dose conditions using the automatic data collection software EPU (ThermoFisher Scientific). Within each selected grid hole, two positions were imaged, each with a total exposure of 1 s. A total of 5573 dose-fractionated image stacks with seven frames each were collected with a 1.035 Å pixel size at defoci ranging from −0.8 µm to −2.8 µm in four separate, independent imaging sessions.

Cryo-EM image processing

Request a detailed protocol

Dose weighting and motion correction were applied using MotionCor2 version 1.4.0 (Li et al., 2013) using anisotropic motion correction with 5 × 5 patches. The initial defocus was estimated either using Gctf 1.06 (Zhang, 2016) or CTFFIND4 version 4.1.5 (Rohou and Grigorieff, 2015), depending on the imaging session. 804 images were discarded during real-time screening at data collection time for excessive drift, strong astigmatism, and low visibility of Thon rings. The remainder was processed with the helical reconstruction routines in RELION3 version 3.0.8 (He and Scheres, 2017; Scheres, 2012; Zivanov et al., 2018). Briefly, the helices were divided into overlapping boxes that were essentially treated as individual, independent particles (with modified Bayesian prior accounting for constraints implied by helicity) to allow sorting of the segments into different conformations and selecting the most well-defined of the conformations present in the sample, a prerequisite for reaching high resolution. For the truncated αE-catenin ABD (671-906) bound to rabbit actin, a total of 728,331 filament segments were extracted using a box size of 200 × 200 pixels from 63,480 manually picked filaments. Two-dimensional reference-free classification for the data set was carried out in RELION3 to eliminate bad segments and segments that showed no evidence for bound ABD, reducing the number of segments from 728,331 to 422,822. An in-house rabbit skeletal actin filament reconstruction filtered to 40 Å resolution was used for an initial model. After several rounds of 3D classification and refinement followed by manual removal of bad particles and further enrichment of segments showing clear decoration, the estimated resolution of the reconstruction, using the 0.143 FSC cutoff gold-standard procedure implemented in RELION3, reached 3.6 Å after postprocessing. The helical rise of the reconstruction was 27.4 Å with a helical twist of −166.9°. The postprocessing included RELION3-based CTF refinement, B-factor sharpening, and application of a soft-edged mask generated in RELION3 corrected for helical edge effects using pyCoAn 0.3.0, an extended python version of CoAn (Volkmann and Hanein, 1999). Additional sharpening was applied using pyCoAn. The reconstruction was then symmetrized within pyCoAn using the refined helical parameters. Local resolution estimates were calculated with ResMap 1.1.4 (Kucukelbir et al., 2014) and RELION3.

A molecular model was produced starting from a structure of bare ADP actin filaments (PDB code 6djo) and the αE-ABD crystal structure (PDB code 6dv1), then iteratively adjusted manually with Coot 0.8.9 (Emsley et al., 2010) and subjected to real-space refinement in Phenix 1.17.1 (Afonine et al., 2018). Quality indicators, including MolProbity (Chen et al., 2010) and EmRinger (Barad et al., 2015) scores, were calculated with Phenix. Some of the processing was done in the SBGrid environment (Morin et al., 2013). Figures were generated with UCSF Chimera version 1.14 (Pettersen et al., 2004).

The coordinates and cryo-EM map of the αE-catenin–F-actin complex have been deposited in the Protein Data Bank, identifiers 6WVT and EMD-21925, respectively.

Limited proteolysis

Request a detailed protocol

Limited proteolysis of αE-catenin 671–906 was performed with elastase (Worthington Biochemical). αE-catenin ABD (8 μM) was incubated in the presence or absence of 8 μM F-actin for 1 hr at room temperature in 5 mM Tris pH 8.0, 50 mM potassium chloride, 2 mM magnesium chloride and 0.5 mM DTT. After addition of elastase to a final concentration of 0.009 mg/ml, aliquots were removed after 5’, 15’, 30’, 1 hr, 2 hr, 4 hr and the proteolysis reaction was stopped by addition of SDS sample buffer and boiling. Samples were run on SDS–PAGE and gels were stained with Coomassie Blue. For N-terminal sequencing, bands were transferred on a PVDF membrane. Bands of ~24 and~26 kDa found only in the presence of F-actin were excised and submitted for N-terminal (Edman) protein sequencing.

Data availability

The coordinates and cryo-EM map of the αE-catenin-F-actin complex have been deposited in the Protein Data Bank, identifiers 6WVT and EMD-21925, respectively.

The following data sets were generated

References

    1. Emsley P
    2. Lohkamp B
    3. Scott WG
    4. Cowtan K
    (2010) Features and development of coot
    Acta Crystallographica. Section D, Biological Crystallography 66:486–501.
    https://doi.org/10.1107/S0907444910007493
    1. Spudich JA
    2. Watt S
    (1971)
    The regulation of rabbit skeletal muscle contraction. I. biochemical studies of the interaction of the tropomyosin-troponin complex with actin and the proteolytic fragments of myosin
    The Journal of Biological Chemistry 246:4866–4871.

Decision letter

  1. Christopher P Hill
    Reviewing Editor; University of Utah School of Medicine, United States
  2. Suzanne R Pfeffer
    Senior Editor; Stanford University School of Medicine, United States
  3. Christopher P Hill
    Reviewer; University of Utah School of Medicine, United States
  4. Jie Yan
    Reviewer; National University of Singapore, Singapore

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Thank you for submitting your article "Structural basis of αE-catenin-F-actin catch bond behavior" for consideration by eLife. Your article has been reviewed by three peer reviewers, including Christopher Hill as the Reviewing Editor and Reviewer #1, and the evaluation has been overseen by a Senior Editor. The following individuals involved in review of your submission have agreed to reveal their identity: Jie Yan (Reviewer #2).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

We would like to draw your attention to changes in our revision policy that we have made in response to COVID-19 (https://elifesciences.org/articles/57162). Specifically, we are asking editors to accept without delay manuscripts, like yours, that they judge can stand as eLife papers without additional data, even if they feel that they would make the manuscript stronger. Thus the revisions requested below only address clarity and presentation.

Summary:

A structure is reported for a complex between αE-catenin and F-actin. The catenin construct is a truncated form that promote F-actin binding, apparently by destabilizing an alternative weakly-binding conformation. The actin-bound catenin conformation differs from that of a previously report structure of a longer construct, which presumably displays the autoinhibited weak-binding form. The primary mechanistic insight is a plausible mechanism for the catch-bond behavior of the catenin-actin interaction. The structural-mechanistic model is supported by biochemical data and sequence conservation, and represents an important advance in understanding.

Essential revisions:

Abstract

1) In the Abstract, the authors mentioned "Upon binding to actin, the first helix of the bundle dissociates and the remaining four helices and connecting loops rearrange to form the interface with actin." This conclusion is suggested from the higher proteolytic sensitivity of the H0-deleted ABD in the presence of F-actin, instead of a direct result. However, if the binding to F-actin is enough to dissociate H1, why do the authors propose that there are two populations of the bound ABDs, a weaker bound state where the H1 is not dissociated and a stronger bound where the H1 is displaced, which equilibrium is shifted by the force applied to the ABD? To allow the force play a significant role, it seems to me that in the absence of force, the auto-inhibited weaker species should be the prevalent one, so that the force can modulate the binding by releasing this auto-inhibition. Perhaps it is more reasonable to think that the observed higher proteolytic sensitivity of the H0-deleted ABD in the presence of F-actin indicates that the H1 helix becomes more dynamic, instead of being dissociated, making its cleavage sites dynamically exposed.

General

2) The proposed mechanism is mainly based on that force may stablize the bound ABD in the higher affinity conformation where the H0 and H1 helices are dissociated from H2-H5. Previous study by Huang et al., (2017) revealed that, when the force applied to vinculin ABD is toward the minus end of F-actin, a switch from catch-bond to slip-bond occurred at ~ 8-10 pN. Based on the proposed mechanism in this work, the force needed to stabilize the ABD in the H1 displaced state should be in a similar level. This again suggests that H1 associates with H2-H5 with significant interaction energy when the ABD is bound on F-actin, which requires a few pN forces to dissociate.

3) The study indicated role of the long C-terminal extension (CTE), residues 844-906, which follows H5. According to the structure, the disordered CTE past 860 forms an extended peptide that interacts with F-actin. Previous study by Huang et al., (2017) also revealed that, when the force applied to vinculin ABD is toward the plus end of F-actin, although the mechanical stability of vinculin ABD on F-actin is significantly weaker than that toward the minus end, a switch from catch-bond to slip-bond was still observed at ~ 8-10 pN forces. Apparently, this plus-end catch-to-slip switch behavior cannot be explained by the force-dependent dissociation of the H1 helix from H2-h5. However, I think it might be explained by the presence of the pre-extended CTE peptide on F-actin.

4) Several recent studies (Yuan et al., 2017; Guo et al., 2018; 2020) suggested that if unfolding of a domain or a rupturing of a biomolecular complex follows a transition pathway where a pre-extended peptide is peeled away from the remaining folded core in a shearing force geometry, a catch-to-slip switch behavior can be expected. This phenomenon is related to the highly flexible nature of the peptide, leading to a negative transition distance within certain tension range up to a value of f_s (therefore catch bond) which then switches to positive values (therefore slip bond) when tension exceeds f_s. The more the peptide is pre-extended in the native structure, the larger the switching tension f_s. Based on this physical mechanism, if the pre-extended CTE is the last anchoring site of the ABD stretched toward the plus end of F-actin, a catch-to-slip switch behavior is predicted. If the authors think the plus-end catch-to-slip switch behavior an interesting phenomenon, it will be good to add several sentences on this point in the Discussion section.

5) Citations of published structures are often missing both in the manuscript and figure legends. Both the original publication and PDB accession number should be stated.

Introduction

6) "In the optical trap data" – needs a citation.

7) "Comparison of the crystal structure and the actin bound form of the complete aE-catenin ABD, as well as the structures of the vinculin ABD free or bound to actin (Bakolitsa et al., 2004; Bakolitsa et al., 1999; Borgon et al., 2004; Mei et al., 2020), provides an explanation for the weak to strong actin-binding transition, and biochemical and mutational data support this model." – It is unclear which citation belongs to which structure and what is presented in this paper.

Results section

8) "[…] we obtained a three-dimensional cryo-EM reconstruction of actin filaments bound to a truncated aE-catenin ABD (residues 671-906) […] This construct deletes the first half of, and thereby destabilizes, the short H0, and binds 4.5x more strongly than the complete ABD" – Why was this construct chosen? Was the background too high when using the full ABD and a higher concentration? What is the affinity of the full-length protein? Since this construct results in a sole population of the strong binding state, and thus levers out the catch bond mechanism and possibility to solve the structure of both states (weak and strong binding), this point should be discussed in more detail in the manuscript. Moreover, the authors themselves note that Mei et al. used the full ABD, arising the question why they didn't.

9) "[…] we observed either bare actin filaments or stretches of filaments continuously bound by aE-ABD. Using the same 10 μM concentration, we were unable to observe binding of the complete ABD to actin filaments in the electron microscope." – An additional SI figure showing examples for bare actin and decorated stretches for both cases would be helpful. Also, it's not clear to me, if all this was done using cryo-EM and if decoration was judge by eye on micrograph level or if for example layer lines were used.

10) "Indeed, a comparable structure using an ABD construct spanning residues 664-906 was produced using a higher concentration of ABD (Mei et al., 2020)." – Stating the concentration used by Mei et al., and the corresponding affinity would be helpful.

11) "The visible portion of the aE-catenin ABD starts at residue 699 and ends at 871; 6 residues in the loop connecting H4 and H5, and 8 residues in the connection between H5 and the rest of the C-terminal extension also could not be modeled." – readers without cryo-EM background require an explanation what "visible portion" implies.

12) A short mentioning of the actin isoform and its nucleotide state/ overall buffer condition is necessary to follow the discussion of the D-loop state.

13) " […] modeling suggests that the 'open' D-loop conformation that has been associated with the ATP-bound form of bare actin filaments (Merino et al., 2018) would clash with the bound aE-catenin" – This is interesting. Does aE-catenin prefer a certain nucleotide state of actin?

14) "Moreover, the refinement procedure used to generate high-resolution structures from cryo-EM images selects and enforces a single conformation" – please specify that helical refinement was used and add a short explanation for non-expert readers, how a single conformation is enforced.

15) I am missing a few aspects in the discussion of the binding site/mode: As aE-catenin contacts two actin subunits, does it also stabilize F-actin? Does the binding site overlap with the binding site of other actin binding proteins? Do they colocalize in the cell? Is a full-length crystal structure available? Or data about how the remainder of catenin packs relative to the ABD? Would it introduce clashes/ additional interactions between neighboring molecules? Maybe the authors could add a "sketch" to one of the figures.

16) "[…] N-terminus through the last turn of H1 is disordered (Figure 3A). We note that an F-actin-bound aE-catenin ABD structure has been reported recently for the complete ABD (664-906) (Mei et al., 2020), and the same residues are disordered, demonstrating that the truncation of H0 in our construct has no influence on the actin bound structure." – Can the authord speculate on the function/stabilization of the disordered part in the full-length protein?

17) "This repositions an aromatic cluster formed by conserved residues W705 (H1), Y837 (H5), and W859 (CTE), and shifts the CTE upward (Figure 3C)." Please highlight these residues in Figure 2—figure supplement 2B.

18) "The CTE past 860, which is disordered in the isolated structure, forms an extended peptide that interacts with actin." Interactions with actin need to be specified/ described in more detail. A 2D interaction plot might be helpful.

19) "[...] lowers the affinity for actin approximately 10-fold (Table 1)." Reference to wrong table.

20) "Chen et al., (et al.2015) found several point mutants that severely weakened binding, including I792A. K842A eliminates contacts with actin residues H87 and Y91; and K866A eliminates side chain and main chain contacts with actin residues R28 and V30 (Figure 2D)." Reference of I792 to structure is missing, no figure showing the interactions of K842 and R28.

21. "Surprisingly, removal of residues 869-906 eliminated detectable F-actin binding, even though none of these residues interact with actin in our structure." Can the authors offer a hypothesis for what the function of residues 869-883 could be?

22) "Moreover, although relatively few of the actin-contacting residues are conserved in vinculin (notably, those in the C-terminal portion of H4), the vinculin ABD undergoes a similar structural transition upon binding to actin" a superposition of both structures in a supplementary figure would help the reader to follow this point.

23) "We cannot assess whether the weak state corresponds precisely to the solution conformation, or if it is intermediate between the experimentally determined unbound and actin-bound structures." I agree that it's impossible with the author's construct/ data, but shouldn't it be possible to get both states with a longer construct that still undergoes the conformational change from weak to strong binding?

24) Discussion section

25) What follow-up experiments could help to clarify the mechanism? Structure with longer construct? Interesting target for MD simulations?

Materials and methods section

26) Values in Manuscript/ Methods and Table 1 do not match! Resolution 3.8 vs 3.6 Å, different symmetry values? All values need to be checked and corrected!

27) "The data were analyzed in the program GraphPad Prism (GraphPad Software, La Jolla California USA) and fitted with a 'single binding with Hill coefficient' model. The data were analyzed in the program GraphPad Prism and fitted using the 'specific binding with Hill slope' model, with the exception of the aE-catenin 671-906 W859A mutant." Partially duplicated sentence.

28) "The defocus was estimated either using Gctf (Zhang, 2016) or CTFFIND4 (Rohou and Grigorieff, 2015)" Why were two different programs used to estimate the defocus of a single data set? Which criterion was used to decide which result to use?

29) Is there a reason why particles were not polished and/or CTF refined? Did the authors try 3D classification (focused, possibly in combination with signal subtraction)?

30) " The reconstruction was then symmetrized within pyCoAn using the refined helical parameters. Real space refinement was performed with Phenix (Afonine et al., 2018). Local resolution estimates were calculated with ResMap (Kucukelbir et al., 2014) and RELION3." Remove sentence about real space refinement here (discussed in detail in the next paragraph)

31) Missing version numbers for some software packages including Phenix, CTFFIND, GCTF and others.

32) Cryo-EM resolutions should only be stated with one decimal digit.

Tables

33) Table 1: Please add exposure time, number of non-hydrogen atoms, Rotamer statistics and cross correlation values. A consistent style should be used for numbers i.e. 4,769 mics not 4769. Cryo-EM resolutions should only be stated with one decimal digit.

Figures and Legends

34) Mismatch of capitalization of panel lettering in manuscript and figures i.e. Figure 1a vs 1A.

35. Figure 1: Showing cadherins from different cells in different shades of green would help to understand the figure. Figure legend: "connection to the actin cytoskeleton", "Crystal structure of the aE-catenin ABD" PDB accession numbers should be stated. Highlights of binding sites in B are not explained in the figure legends.

36) Figure 2: Typo in figure caption " aE-catenin ABD-F-catenin complex". Boxes that show regions enlarged in C) and D) should be added in B. Why is G46 shown in ball representation? Why are labels shown in red and not in magenta to match the structure? Sticking to one color code helps the reader. Label for L803 hardly readable, suggest white instead of yellow background for residue labels.

37) Figure 2—figure supplement 1A) missing scale bar, B) FSC curve needs to be shown up to 0.5 (Nyquist) and y axis >1, as not only the value is of importance but also the curve itself especially the oscillation around zero close to Nyquist. C) Color range for local resolution is not meaningful and should be adjusted.

38) Figure 2—figure supplement 2A) Color code not explained properly in legend, important residues i.e. hydrophobic cluster should be highlighted.

39) Figure 3: The color code in the legend does not completely correspond to the figure. Missing verb in "The upper panel packing interactions of H0 residues 672, M673 and L676 with H5 residues V809, G811 and A815, […]". Missing citation and PDB accession number of the unbound state. Side chains are hardly visible in C) View needs to be changed / another one added to emphasize side chains and the movement of the CTE. The order from left to right to bottom is confusing.

40) Figure 4: Enlarge A, Labels in B) not legible. Consider a white outline/background.

41) Figure 5: "This revealed only minor clashes with actin at the N-terminus of H5, and modeling suggests these can be alleviated by changes of side chain rotamers (Figure 5A, left)." Neither the clashes or proposed side chain rotations are apparent. An extra view and especially a movie would help.

42) The manuscript might benefit from the inclusion of movies. For example, to highlight the binding site of aE-catenin and show close-up views of interactions with F-actin and between ABDs. Another movie could show a morph of the unbound and bound structure of aE-catenin, and an animation of the catch bond model.

https://doi.org/10.7554/eLife.60878.sa1

Author response

Essential revisions:

Abstract

1) In the Abstract, the authors mentioned "Upon binding to actin, the first helix of the bundle dissociates and the remaining four helices and connecting loops rearrange to form the interface with actin." This conclusion is suggested from the higher proteolytic sensitivity of the H0-deleted ABD in the presence of F-actin, instead of a direct result. However, if the binding to F-actin is enough to dissociate H1, why do the authors propose that there are two populations of the bound ABDs, a weaker bound state where the H1 is not dissociated and a stronger bound where the H1 is displaced, which equilibrium is shifted by the force applied to the ABD? To allow the force play a significant role, it seems to me that in the absence of force, the auto-inhibited weaker species should be the prevalent one, so that the force can modulate the binding by releasing this auto-inhibition. Perhaps it is more reasonable to think that the observed higher proteolytic sensitivity of the H0-deleted ABD in the presence of F-actin indicates that the H1 helix becomes more dynamic, instead of being dissociated, making its cleavage sites dynamically exposed.

The existence of two bound states comes directly from optical trap data (Buckley et al., 2014, Huang et al., 2017), which showed that for any given force there is a bi-exponential distribution of lifetimes that can be explained only if there are two bound states. The best model for these data was that either state could dissociate from actin (with different kinetics), and that there is an equilibrium between the weak and strong-bound states that is altered by force. This is the underlying premise of the paper and is described in the introduction. As we discuss in the present paper, conformational change in the ABD is coupled to the binding energy to actin, i.e. in the presence of actin the barrier to the conformational change is lower.

We phrased the cited sentence badly – the stable (strong state) form is what is rearranged, but we are proposing that the weak state corresponds to a structure close to that of non-H1 dissociated form. We have modified the abstract to clarify the origin of the two bound states model. We have also further clarified the role of force on the equilibrium in subsection “Insights into catch bond mechanism”.

The proteolysis experiments are performed on a minutes time scale, so although both forms are sampled, clipping the helix will effectively shift the equilibrium, so that over the time of the experiment it is completely lost.

General

2) The proposed mechanism is mainly based on that force may stablize the bound ABD in the higher affinity conformation where the H0 and H1 helices are dissociated from H2-H5. Previous study by Huang et al., (2017) revealed that, when the force applied to vinculin ABD is toward the minus end of F-actin, a switch from catch-bond to slip-bond occurred at ~ 8-10 pN. Based on the proposed mechanism in this work, the force needed to stabilize the ABD in the H1 displaced state should be in a similar level. This again suggests that H1 associates with H2-H5 with significant interaction energy when the ABD is bound on F-actin, which requires a few pN forces to dissociate.

We agree with the reviewer. As noted in the response to point 1, both weak and strong state can bind to actin and force shifts the equilibrium between them, and in the weak state we propose that H1 is associated with H2-H5, albeit with possibly small changes relative to the unbound structure (see text). The slip behavior at 8 pN is due to removing the entire bundle from actin. We note that a-catenin, although homologous, is not vinculin, so there is no a priori reason to believe that the detailed force dependence would be the same. Nonetheless, we did observe a transition to slip at a similar force value for the cadherin/b-catenin/a-catenin complex in Buckley et al., 2014.

3) The study indicated role of the long C-terminal extension (CTE), residues 844-906, which follows H5. According to the structure, the disordered CTE past 860 forms an extended peptide that interacts with F-actin. Previous study by Huang et al., (2017) also revealed that, when the force applied to vinculin ABD is toward the plus end of F-actin, although the mechanical stability of vinculin ABD on F-actin is significantly weaker than that toward the minus end, a switch from catch-bond to slip-bond was still observed at ~ 8-10 pN forces. Apparently, this plus-end catch-to-slip switch behavior cannot be explained by the force-dependent dissociation of the H1 helix from H2-h5. However, I think it might be explained by the presence of the pre-extended CTE peptide on F-actin.

The kinetics observed in the optical trap were described by a modified Bell model in which the transition state energy depends on the force acting over the distance from the ground to transition state (see Buckley et al., 2014 and Huang et al., 2017). As the reviewer notes, the force dependence doesn’t change depending on direction, only the bound lifetime, implying a different distance to the transition state in the two directions. A geometric model can account for force stabilizing the strongly bound state regardless of whether the force is directed toward the plus or minus end of the actin filament (see Huang et al.,). Although at present we do not have a detailed molecular model for the transition between weak and strong states, as we tried to illustrate in Figure 6 and the short animation (Videos 1 and 2), the force vector experienced by a-catenin or vinculin is not necessarily aligned with the actin filament, and will depend in part on the geometry of how the molecule is tethered. In addition, the component of the force needed to remove H1 (and H0 in the case of a-catenin) is unlikely to be precisely coincident with the helix axis of actin. Thus, the projection of force onto the reaction coordinate can have the same sign regardless of whether the actin filament is being pulled on from its (-) or (+) end.

4) Several recent studies (Yuan et al., 2017; Guo et al., 2018; 2020) suggested that if unfolding of a domain or a rupturing of a biomolecular complex follows a transition pathway where a pre-extended peptide is peeled away from the remaining folded core in a shearing force geometry, a catch-to-slip switch behavior can be expected. This phenomenon is related to the highly flexible nature of the peptide, leading to a negative transition distance within certain tension range up to a value of f_s (therefore catch bond) which then switches to positive values (therefore slip bond) when tension exceeds f_s. The more the peptide is pre-extended in the native structure, the larger the switching tension f_s. Based on this physical mechanism, if the pre-extended CTE is the last anchoring site of the ABD stretched toward the plus end of F-actin, a catch-to-slip switch behavior is predicted. If the authors think the plus-end catch-to-slip switch behavior an interesting phenomenon, it will be good to add several sentences on this point in the Discussion section.

Thank you for pointing out these papers and model. The idea that the disorder-order transition of the CTE contributes to the behavior is extremely interesting, and we have added a brief discussion of this idea. Another effect that may contribute to the transition from the strong to weakly bound state, which we did not address in the original manuscript, is that H1 (and H0) may unfold (as suggested by the proteolysis data) when they are detached from the rest of the ABD. If so, they would provide a flexible linker that extends in the direction of force and contributes to the catch bond behavior. More broadly, we plan to do further modeling with MD simulations to assess whether the data can be modeled in a way that provides physical parameters, but that is beyond the scope of this work. Crucially, the situation here is complicated by the fact that the position of the CTE is coupled to the order or disorder of H1. This could mean that the order-to-disorder of the CTE in going from strong to weak state contributes to the behavior, but because this change is concerted with that of H1, its effect can’t be sorted out cleanly. Finally, in our understanding, the specific model presented in the cited papers predict mono-exponential decay at any given force, which is not what was observed for either a-catenin or vinculin.

5) Citations of published structures are often missing both in the manuscript and figure legends. Both the original publication and PDB accession number should be stated.

Citations and PDB accession number were added in the appropriate places.

Introduction

6) "In the optical trap data" needs a citation.

(Buckley et al., 2014; Huang et al., 2017) were both cited in this sentence

7) "Comparison of the crystal structure and the actin bound form of the complete aE-catenin ABD, as well as the structures of the vinculin ABD free or bound to actin (Bakolitsa et al., 2004; Bakolitsa et al., 1999; Borgon et al., 2004; Mei et al., 2020), provides an explanation for the weak to strong actin-binding transition, and biochemical and mutational data support this model." It is unclear which citation belongs to which structure and what is presented in this paper.

We have made these corrections (Introduction).

Results section

8) "[…] we obtained a three-dimensional cryo-EM reconstruction of actin filaments bound to a truncated aE-catenin ABD (residues 671-906) […] This construct deletes the first half of, and thereby destabilizes, the short H0, and binds 4.5x more strongly than the complete ABD" Why was this construct chosen? Was the background too high when using the full ABD and a higher concentration? What is the affinity of the full-length protein? Since this construct results in a sole population of the strong binding state, and thus levers out the catch bond mechanism and possibility to solve the structure of both states (weak and strong binding), this point should be discussed in more detail in the manuscript. Moreover, the authors themselves note that Mei et al. used the full ABD, arising the question why they didn't.

In our hands, the full-length ABD construct produced very sparse binding to actin filaments in the concentration ranges we used. From our earlier studies (Hansen et al., 2013) we already knew that the 671-906 construct results in excellent decoration for helical averaging, making it an obvious choice for improving the reconstruction without increasing the background too much. Mei et al., (2020) used a 2x higher concentration and a somewhat different protocol for preparing the cryo specimens of the full ABD construct, but crucially, they observed the same structure, i.e., starting at 699, with H0 and H1 dissociated and the H2-H5-CTE region rearranged. Thus, both constructs populate the presumed strong state in the reconstructions. It is likely still in an equilibrium but this equilibrium is shifted towards the strong binding state for the 671-906 construct relative to the full-length ABD, reflected in its ~4x higher affinity (Table 2). We cannot rule out that the conditions of the EM specimen preparation may affect the equilibrium and shift it further to the strong state. We have clarified these differences in the text (subsection “Structure of aE-catenin ABD bound to F-actin”).

Regarding the full-length a-catenin, it binds with about the same KD as the ABD alone (but this is a dimer, so we don’t have a direct comparison to monomeric ABD), but unlike the ABD it does not bind with strong cooperativity (Hansen et al., 2013). Consistently, we have only observed sparse decoration with the full-length protein, preventing reconstructions.

9) "[…] we observed either bare actin filaments or stretches of filaments continuously bound by aE-ABD. Using the same 10 μM concentration, we were unable to observe binding of the complete ABD to actin filaments in the electron microscope." An additional SI figure showing examples for bare actin and decorated stretches for both cases would be helpful. Also, it's not clear to me, if all this was done using cryo-EM and if decoration was judge by eye on micrograph level or if for example layer lines were used.

We added representative micrographs for both full-length and truncated ABD data with markings for bare and decorated actin filaments to Figure 2—figure supplement 1 (A and B). All of the analysis was done by cryo-EM, including initial screening. Decoration, which is readily apparent in the micrographs was judged initially by visual inspection using high-defocus data from the 120 keV cryo-TEM microscope while screening conditions. Decoration and the noted cooperativity was confirmed in the data set computationally during data processing. We have now noted these in the Materials and methods section.

10) "Indeed, a comparable structure using an ABD construct spanning residues 664-906 was produced using a higher concentration of ABD (Mei et al., 2020)." Stating the concentration used by Mei et al., and the corresponding affinity would be helpful.

They used 20 mM, i.e. 2x higher. Our measured affinity for our full-length ABD construct (666-906) was 8.5 mM. We have now noted these in the text (subsection “Structure of aE-catenin ABD bound to F-actin”).

11) "The visible portion of the aE-catenin ABD starts at residue 699 and ends at 871; 6 residues in the loop connecting H4 and H5, and 8 residues in the connection between H5 and the rest of the C-terminal extension also could not be modeled." Readers without cryo-EM background require an explanation what "visible portion" implies.

“Visible portion” was perhaps a poor choice of words. We rewrote this section as follows (subsection “Structure of aE-catenin ABD bound to F-actin”):

“For the aE-catenin ABD, there was no detectable density for residues 671-698, or from 872-906. In addition, 6 residues in the loop connecting H4 and H5, and 8 residues in the connection between H5 and the rest of the C-terminal extension also could not be modeled.”

12) A short mentioning of the actin isoform and its nucleotide state/ overall buffer condition is necessary to follow the discussion of the D-loop state.

We now note that ADP actin is used, with its preparation given in the Materials and methods section.

13) " […] modeling suggests that the 'open' D-loop conformation that has been associated with the ATP-bound form of bare actin filaments (Merino et al., 2018) would clash with the bound aE-catenin" This is interesting. Does aE-catenin prefer a certain nucleotide state of actin?

There are no data available on the nucleotide preference. Also, in retrospect we gave a too-definitive statement. In fact, while the ADP state of vertebrate actin appears to be tightly coupled with the ‘closed’ conformation of the D-loop, other nucleotide states, including ATP, clearly show an equilibrium between the ‘closed’ and the alternative ‘open’ D-loop conformation. We replaced the original statement with (this in the text (subsection “Structure of aE-catenin ABD bound to F-actin”)):

“Modeling suggests that the alternative ‘open’ D-loop conformation that occurs in equilibrium with the ‘closed’ conformation in other nucleotide states (Merino et al., 2019), may clash with the bound aE-catenin.”

14) "Moreover, the refinement procedure used to generate high-resolution structures from cryo-EM images selects and enforces a single conformation" please specify that helical refinement was used and add a short explanation for non-expert readers, how a single conformation is enforced.

We have now explained this in detail in the Materials and methods section, and refer to that in the main text.

15) I am missing a few aspects in the discussion of the binding site/mode: As aE-catenin contacts two actin subunits, does it also stabilize F-actin? Does the binding site overlap with the binding site of other actin binding proteins? Do they colocalize in the cell? Is a full-length crystal structure available? Or data about how the remainder of catenin packs relative to the ABD? Would it introduce clashes/ additional interactions between neighboring molecules? Maybe the authors could add a "sketch" to one of the figures.

We have no direct evidence that aE-catenin stabilizes F-actin. More broadly, while interesting, we feel that discussion of overlapping/competing sites and subcellular localization would distract from the focus of this paper. The binding site does overlap other binding sites, in particular that of Arp2/3, and we have published that aE-catenin competes with Arp2/3 for binding (Drees et al., 2005). We are not aware of any other such competition studies. Regarding full-length aE-catenin, small angle x-ray scattering data indicate that the ABD is flexibly linked to the rest of the molecule (Ishiyama et al., 2013; Nicholl et al., 2018; Terekhova et al., 2019), so we cannot provide a sensible model showing the full protein. We previously published that full length aE-catenin by itself does not show strong cooperativity of binding vs. the ABD (Hansen et al., 2013), which could be due to steric interference of neighboring molecules. However, the relevant complex for tension is when aE-catenin is associated with cadherin/b-catenin, and there are no data that speak directly to this point.

16) "[…] N-terminus through the last turn of H1 is disordered (Figure 3A). We note that an F-actin-bound aE-catenin ABD structure has been reported recently for the complete ABD (664-906) (Mei et al., 2020), and the same residues are disordered, demonstrating that the truncation of H0 in our construct has no influence on the actin bound structure." Can the authord speculate on the function/stabilization of the disordered part in the full-length protein?

We have added the following to the Discussion section:

“Although it is not apparently integral to the catch bond behavior, we speculate that the conserved H0 of a-catenin serves to tune the stability and force response of the ABD; in the disordered state bound to actin, it will provide additional flexible elements that may affect the detailed catch bond behavior.”

17) "This repositions an aromatic cluster formed by conserved residues W705 (H1), Y837 (H5), and W859 (CTE), and shifts the CTE upward (Figure 3C)." Please highlight these residues in Figure 2—figure supplement 2B.

Added to the alignment figure as requested (now Figure 2—figure supplement 5).

18) "The CTE past 860, which is disordered in the isolated structure, forms an extended peptide that interacts with actin." Interactions with actin need to be specified/ described in more detail. A 2D interaction plot might be helpful.

Added details to the text (his in the text (subsection “Structure of aE-catenin ABD bound to F-actin”)) and a new figure (Figure 2—figure supplement 3A).

19) "[...] lowers the affinity for actin approximately 10-fold (Table 1)." Reference to wrong table.

Corrected.

20) "Chen et al., (2015) found several point mutants that severely weakened binding, including I792A. K842A eliminates contacts with actin residues H87 and Y91; and K866A eliminates side chain and main chain contacts with actin residues R28 and V30 (Figure 2D)." Reference of I792 to structure is missing, no figure showing the interactions of K842 and R28.

I792 is described in Figure 2C and referenced in the preceding part of the paragraph. Added additional Figure 2—figure supplement 3B and C to show interactions of K842 as well as K797.

21) "Surprisingly, removal of residues 869-906 eliminated detectable F-actin binding, even though none of these residues interact with actin in our structure." Can the authors offer a hypothesis for what the function of residues 869-883 could be?

We do not have a good hypothesis; we think that there may be a dynamic interaction that averages out in the structure. Mei et al. propose that 872-906 mediate additional binding when actin itself is subject to tension. It could be that 869-871 form contacts that we cannot assign due to the limited resolution of our structure. We now discuss this in the text (subsection “Structure of aE-catenin ABD bound to F-actin”):

(Mei et al., 2020) proposed that these C-terminal residues absent in the structure may mediate a small increase in affinity when actin is placed under tension. However, deleting these residues weakens the affinity (albeit slightly) in a solution assay, which indicates that they have a role independent of tension. It is possible that these are highly dynamic interactions that are not sufficiently stable to be visualized in the cryo-EM structure.

22) "Moreover, although relatively few of the actin-contacting residues are conserved in vinculin (notably, those in the C-terminal portion of H4), the vinculin ABD undergoes a similar structural transition upon binding to actin" a superposition of both structures in a supplementary figure would help the reader to follow this point.

We have added comparisons of the free and actin-bound vinculin and aE-catenin structures as Figure 3—figure supplement 2. We used the 8.5 Å resolution vinculin structure from Kim et al., 2016 as the higher resolution structure from Mei et al., (2020) is not available.

23) "We cannot assess whether the weak state corresponds precisely to the solution conformation, or if it is intermediate between the experimentally determined unbound and actin-bound structures." I agree that it's impossible with the author's construct/ data, but shouldn't it be possible to get both states with a longer construct that still undergoes the conformational change from weak to strong binding?

No, because it seems to equilibrate to the state seen here, and the longer construct is even weaker (see above regarding the preparation of samples). We also note that Mei et al., 2020 do not report any heterogeneity in the structures.

24) Discussion section

25) What follow-up experiments could help to clarify the mechanism? Structure with longer construct? Interesting target for MD simulations?

We now note that MD simulations and more optical trap experiments are being done in order to better define the mechanism (Discussion section).

Materials and methods section

26) Values in Manuscript/ Materials and methods section and Table 1 do not match! Resolution 3.8 vs 3.6 Å, different symmetry values? All values need to be checked and corrected!

Values have been checked and corrected in the text.

27) "The data were analyzed in the program GraphPad Prism (GraphPad Software, La Jolla California USA) and fitted with a 'single binding with Hill coefficient' model. The data were analyzed in the program GraphPad Prism and fitted using the 'specific binding with Hill slope' model, with the exception of the aE-catenin 671-906 W859A mutant." Partially duplicated sentence.

Corrected in the text.

28) "The defocus was estimated either using Gctf (Zhang, 2016) or CTFFIND4 (Rohou and Grigorieff, 2015)" – Why were two different programs used to estimate the defocus of a single data set? Which criterion was used to decide which result to use?

The data were collected in several separate imaging sessions. Initially, we used CTFFIND4. We later switched to Gctf, which gives essentially the same results as CTFFIND4 but is substantially faster because it is GPU accelerated. This is now described in the Materials and methods section.

29) Is there a reason why particles were not polished and/or CTF refined? Did the authors try 3D classification (focused, possibly in combination with signal subtraction)?

The CTF was, in fact, refined. Particle polishing was not used because anisotropic motion correction and dose weighting was already applied within motioncor2. We reflected these facts in the updated Materials and methods section.

30) "The reconstruction was then symmetrized within pyCoAn using the refined helical parameters. Real space refinement was performed with Phenix (Afonine et al., 2018). Local resolution estimates were calculated with ResMap (Kucukelbir et al., 2014) and RELION3." Remove sentence about real space refinement here (discussed in detail in the next paragraph).

Removed from the text.

31) Missing version numbers for some software packages including Phenix, CTFFIND, GCTF and others.

Version numbers were added.

32) Cryo-EM resolutions should only be stated with one decimal digit.

Revised in the text.

Tables

33) Table 1: Please add exposure time, number of non-hydrogen atoms, Rotamer statistics and cross correlation values. A consistent style should be used for numbers i.e. 4,769 mics not 4769. Cryo-EM resolutions should only be stated with one decimal digit.

The values were added to the Table.

Figures and Legends

34) Mismatch of capitalization of panel lettering in manuscript and figures i.e. Figure 1A vs 1A.

Capitalization is now consistent with panels.

35. Figure 1: Showing cadherins from different cells in different shades of green would help to understand the figure. Figure legend: "connection to the actin cytoskeleton", "Crystal structure of the aE-catenin ABD" PDB accession numbers should be stated. Highlights of binding sites in B are not explained in the figure legends.

Corrected.

36) Figure 2: Typo in figure caption " aE-catenin ABD-F-catenin complex". Boxes that show regions enlarged in C) and D) should be added in B. Why is G46 shown in ball representation? Why are labels shown in red and not in magenta to match the structure? Sticking to one color code helps the reader. Label for L803 hardly readable, suggest white instead of yellow background for residue labels.

G46 is shown as a ball because showing side chains only on the cartoon representation does not show α-carbons. Labels are color matched to structure; the yellow background makes them more readable but may make the color look slightly different. We prefer to use this scheme.

37) Figure 2—figure supplement 1A) missing scale bar, B) FSC curve needs to be shown up to 0.5 (Nyquist) and y axis >1, as not only the value is of importance but also the curve itself especially the oscillation around zero close to Nyquist. C) Color range for local resolution is not meaningful and should be adjusted.

Scale bars were added to A and B. FSC curve was extended. The color range for local resolution was adjusted to show a narrower resolution range.

38) Figure 2—figure supplement 2A) Color code not explained properly in legend, important residues i.e. hydrophobic cluster should be highlighted.

Corrected in the text.

39) Figure 3: The color code in the legend does not completely correspond to the figure. Missing verb in "The upper panel packing interactions of H0 residues 672, M673 and L676 with H5 residues V809, G811 and A815, […]". Missing citation and PDB accession number of the unbound state. Side chains are hardly visible in C) View needs to be changed / another one added to emphasize side chains and the movement of the CTE. The order from left to right to bottom is confusing.

The changes in the CTE due to repacking of the aromatic cluster are now in a separate figure (Figure 3 – figure supplement 1), where we added an overlay of the free and bound structures to better show the change in the CTE position.

40) Figure 4: Enlarge A, Labels in B) not legible. Consider a white outline/background.

Enlarged and outlined labels in B.

41) Figure 5: "This revealed only minor clashes with actin at the N-terminus of H5, and modeling suggests these can be alleviated by changes of side chain rotamers (Figure 5A, left)." Neither the clashes or proposed side chain rotations are apparent. An extra view and especially a movie would help.

We have added a Figure 5—figure supplement 1 to show the major clash with actin and how small movements of H1 and the start of H5 would relieve them. We also clarified the text (subsection “Insights into catch bond mechanism”) to better explain the model.

42) The manuscript might benefit from the inclusion of movies. For example, to highlight the binding site of aE-catenin and show close-up views of interactions with F-actin and between ABDs. Another movie could show a morph of the unbound and bound structure of aE-catenin, and an animation of the catch bond model.

We have added animated versions of Figure 6 in order to intuitively capture our model (Video 1 and Video 2).

https://doi.org/10.7554/eLife.60878.sa2

Article and author information

Author details

  1. Xiao-Ping Xu

    Scintillon Institute, San Diego, United States
    Contribution
    Formal analysis, Investigation, Methodology
    Contributed equally with
    Sabine Pokutta
    Competing interests
    No competing interests declared
  2. Sabine Pokutta

    Departments of Structural Biology and Molecular & Cellular Physiology, Stanford University School of Medicine, Stanford, United States
    Contribution
    Conceptualization, Formal analysis, Investigation, Writing - review and editing
    Contributed equally with
    Xiao-Ping Xu
    Competing interests
    No competing interests declared
  3. Megan Torres

    Departments of Structural Biology and Molecular & Cellular Physiology, Stanford University School of Medicine, Stanford, United States
    Contribution
    Formal analysis, Investigation
    Competing interests
    No competing interests declared
  4. Mark F Swift

    Scintillon Institute, San Diego, United States
    Contribution
    Investigation, Methodology
    Competing interests
    No competing interests declared
  5. Dorit Hanein

    1. Scintillon Institute, San Diego, United States
    2. Department of Structural Biology and Chemistry, Pasteur Institute, Paris, France
    Contribution
    Funding acquisition, Investigation, Methodology, Project administration, Writing - review and editing
    For correspondence
    dorit.hanein@pasteur.fr
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-6072-4946
  6. Niels Volkmann

    1. Scintillon Institute, San Diego, United States
    2. Department of Structural Biology and Chemistry, Pasteur Institute, Paris, France
    Contribution
    Conceptualization, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Writing - review and editing
    For correspondence
    niels.volkmann@pasteur.fr
    Competing interests
    No competing interests declared
  7. William I Weis

    Departments of Structural Biology and Molecular & Cellular Physiology, Stanford University School of Medicine, Stanford, United States
    Contribution
    Conceptualization, Formal analysis, Funding acquisition, Investigation, Writing - original draft, Project administration, Writing - review and editing
    For correspondence
    weis@stanford.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-5583-6150

Funding

National Institutes of Health (GM118326)

  • Dorit Hanein
  • Niels Volkmann
  • William I Weis

National Institutes of Health (GM131747)

  • William I Weis

National Institutes of Health (S10-OD012372)

  • Dorit Hanein

National Institutes of Health (S10-OD026926)

  • Dorit Hanein

Pew Charitable Trusts (864K625)

  • Dorit Hanein

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We are grateful to Dr. Gregory Alushin for providing coordinates of the αE-catenin ABD bound to actin published in the accompanying paper (Mei et al., 2020), which allowed us to correct an error in the CTE of our model during the proof stage of this manuscript. We thank Max Pokutta for technical assistance. WIW thanks Alex Dunn for discussions. This work was supported by US National Institutes of Health grant GM118326 (DH, NV and WIW) and GM131747 (WIW). NIH grants S10-OD012372 and S10-OD026926, and PEW innovation funds 864K625 to DH funded the purchase of the Titan Krios electron cryo-microscope (ThermoFisher Scientific), Falcon two direct detector (ThermoFisher Scientific), and upgrades for both the T12 and Titan Krios hardware and software including the Eagle CMOS imaging device (ThermoFisher Scientific) on the Tecnai Spirit T12.

Senior Editor

  1. Suzanne R Pfeffer, Stanford University School of Medicine, United States

Reviewing Editor

  1. Christopher P Hill, University of Utah School of Medicine, United States

Reviewers

  1. Christopher P Hill, University of Utah School of Medicine, United States
  2. Jie Yan, National University of Singapore, Singapore

Publication history

  1. Received: July 9, 2020
  2. Accepted: September 9, 2020
  3. Accepted Manuscript published: September 11, 2020 (version 1)
  4. Version of Record published: October 26, 2020 (version 2)

Copyright

© 2020, Xu et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 2,122
    Page views
  • 341
    Downloads
  • 27
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, Scopus, PubMed Central.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Xiao-Ping Xu
  2. Sabine Pokutta
  3. Megan Torres
  4. Mark F Swift
  5. Dorit Hanein
  6. Niels Volkmann
  7. William I Weis
(2020)
Structural basis of αE-catenin–F-actin catch bond behavior
eLife 9:e60878.
https://doi.org/10.7554/eLife.60878

Further reading

    1. Cell Biology
    2. Structural Biology and Molecular Biophysics
    Lin Mei, Santiago Espinosa de los Reyes ... Gregory M Alushin
    Research Article Updated

    The actin cytoskeleton mediates mechanical coupling between cells and their tissue microenvironments. The architecture and composition of actin networks are modulated by force; however, it is unclear how interactions between actin filaments (F-actin) and associated proteins are mechanically regulated. Here we employ both optical trapping and biochemical reconstitution with myosin motor proteins to show single piconewton forces applied solely to F-actin enhance binding by the human version of the essential cell-cell adhesion protein αE-catenin but not its homolog vinculin. Cryo-electron microscopy structures of both proteins bound to F-actin reveal unique rearrangements that facilitate their flexible C-termini refolding to engage distinct interfaces. Truncating α-catenin’s C-terminus eliminates force-activated F-actin binding, and addition of this motif to vinculin confers force-activated binding, demonstrating that α-catenin’s C-terminus is a modular detector of F-actin tension. Our studies establish that piconewton force on F-actin can enhance partner binding, which we propose mechanically regulates cellular adhesion through α-catenin.

    1. Biochemistry and Chemical Biology
    2. Structural Biology and Molecular Biophysics
    Maura Greiser, Mariusz Karbowski ... Liron Boyman
    Research Article

    Mitochondrial ATP production in cardiac ventricular myocytes must be continually adjusted to rapidly replenish the ATP consumed by the working heart. Two systems are known to be critical in this regulation: mitochondrial matrix Ca2+ ([Ca2+]m) and blood flow that is tuned by local ventricular myocyte metabolic signaling. However, these two regulatory systems do not fully account for the physiological range of ATP consumption observed. We report here on the identity, location, and signaling cascade of a third regulatory system -- CO2/bicarbonate. CO2 is generated in the mitochondrial matrix as a metabolic waste product of the oxidation of nutrients that powers ATP production. It is a lipid soluble gas that rapidly permeates the inner mitochondrial membrane (IMM) and produces bicarbonate (HCO3-) in a reaction accelerated by carbonic anhydrase (CA). The bicarbonate level is tracked physiologically by a bicarbonate-activated adenylyl cyclase, soluble adenylyl cyclase (sAC). Using structural Airyscan super-resolution imaging and functional measurements we find that sAC is primarily inside the mitochondria of ventricular myocytes where it generates cAMP when activated by HCO3-. Our data strongly suggest that ATP production in these mitochondria is regulated by this cAMP signaling cascade operating within the inter-membrane space (IMS) by activating local EPAC1 (Exchange Protein directly Activated by cAMP) which turns on Rap1 (Ras-related protein 1). Thus, mitochondrial ATP production is shown to be increased by bicarbonate-triggered sAC signaling through Rap1. Additional evidence is presented indicating that the cAMP signaling itself does not occur directly in the matrix. We also show that this third signaling process involving bicarbonate and sAC activates the cardiac mitochondrial ATP production machinery by working independently of, yet in conjunction with, [Ca2+]m-dependent ATP production to meet the energy needs of cellular activity in both health and disease. We propose that the bicarbonate and calcium signaling arms function in a resonant or complementary manner to match mitochondrial ATP production to the full range of energy consumption in cardiac ventricular myocytes in health and disease.