Bending actin filaments: twists of fate

In many cellular contexts, intracellular actomyosin networks must generate directional forces to carry out cellular tasks such as migration and endocytosis, which play important roles during normal developmental processes. A number of different actin binding proteins have been identified that form linear or branched actin, and that regulate these filaments through activities such as bundling, crosslinking, and depolymerization to create a wide variety of functional actin assemblies. The helical nature of actin filaments allows them to better accommodate tensile stresses by untwisting, as well as to bend to great curvatures without breaking. Interestingly, this latter property, the bending of actin filaments, is emerging as an exciting new feature for determining dynamic actin configurations and functions. Indeed, recent studies using in vitro assays have found that proteins including IQGAP, Cofilin, Septins, Anillin, α-Actinin, Fascin, and Myosins—alone or in combination—can influence the bending or curvature of actin filaments. This bending increases the number and types of dynamic assemblies that can be generated, as well as the spectrum of their functions. Intriguingly, in some cases, actin bending creates directionality within a cell, resulting in a chiral cell shape. This actin-dependent cell chirality is highly conserved in vertebrates and invertebrates and is essential for cell migration and breaking L-R symmetry of tissues/organs. Here, we review how different types of actin binding protein can bend actin filaments, induce curved filament geometries, and how they impact on cellular functions.

Actin is often characterized as being assembled into linear filaments through the action of de novo linear actin nucleation factors such as formins and Spire 22-25 ( Figure 1A). These linear actin filaments can be organized into different assemblies through bundling and/or cross-linking proteins, allowing them to perform functions including filopodia formation. Linear actin filaments also serve as the scaffold for the formation of branched actin through the action of de novo branched actin nucleation factors including Wiskott-Aldrich Syndrome family proteins and the Arp2/3 complex 22,26-33 ( Figure 1B). Branched actin meshes are associated with cell events including lamellipodia formation and cell migration.
Actin filaments themselves are also inherently helical and this helicity has been demonstrated to provide the filament with its mechanical strength and play an important role in facilitating cell chirality and L-R asymmetric morphogenesis 34 ( Figure 1C). Interestingly, this intrinsic twisting is maintained during formin mDia1-dependent actin filament elongation as a result of mDia1 rotating along the filament axis as it actively incorporates actin molecules 35 . The canonical actin filament double helix has a pitch of 72nm and completes 13.9 turns per micrometer. Notably, computational modeling predicts that filaments under 200pN of tension can untwist approximately 2° to elongate the filament 36 . In a separate study, tension on actin filaments was shown to enhance the binding affinity of myosin II motors 37 . Additionally, it was observed that Arp2/3-mediated branching occurred more frequently on the convex side of bent filaments 38 . Multiple studies have now shown  cooperative and preferential binding to different actin binding proteins (ABPs) depending on the physical conformation of the actin filament 34,40-42 . Intriguingly, a recent study showed that the nucleotide state of actin filaments can have a major impact on the rigidity of the filament and influence its twisting and bending 42 . The specific nucleotide state in the presence of bending forces can additionally influence the spatial rearrangements of actin molecules and change the helical twist of actin filaments as predicted by computer simulations 43 . Therefore, there exists a complex feedback loop between the physical stresses applied to an actin filament, its conformation, and the combination of ABPs bound. This complex system of forces and ABPs can dictate the conformation of a given actin filament and/or the organization/geometry of actin filament assemblies to properly sense and react to cellular and environmental events.
In addition to their roles in assembling linear and branched actin filament assemblies, a number of ABPs have recently been shown to alter the properties of actin filaments, triggering them into curved geometries. Here, we review the ability of different ABPs-alone or in conjunction with other ABPs-to alter the conformation of actin filaments and/or the geometry of actin filament assemblies, and their effect on cellular and developmental events.

IQGAP/calponin homology proteins
The IQGAP (IQ motif-containing GTPase-activating proteins) protein family are scaffolding proteins with multifarious interactions that allow them to facilitate actin dynamics at the cell membrane. They have been implicated in dynamic actin processes such as cell migration [44][45][46] . IQGAPs possess recognizable protein motifs, including calponin homology domains (CH), coiled-coil domains (CC), and GTPase R domains (GRD), that contribute to their modulation of actin assemblies 44,45 .
The CC domain has been demonstrated to bind to Ezrin, a cytoskeleton-plasma membrane anchor, and the GRD domain transiently protects the Cdc42 small GTPase in its activated state by reducing GTP hydrolysis. These two domains allow the IQGAP family to localize to the plasma membrane and sustain the interaction between the WASP branched actin nucleator and the Arp2/3 complex through its CH domain, thereby promoting branched actin nucleation 47 . In addition to binding WASP, the CH domain has been shown to bind actin filaments with high affinity and, strikingly, to bend actin filaments to a high curvature 48 . In a recent study, Palani and colleagues creatively used the CH domain of Rng2 (fission yeast IQGAP ortholog) to show that it was capable of bending actin filaments to form ring structures 48 (Figure 2A). In a series of in vitro assays, the authors demonstrated that membrane-tethered His-tagged CH domain not only bent actin filaments, but consistently bent filaments anti-clockwise. This observation suggests a level of inherent chirality in actin filament bending because of bound CH domain-containing proteins. Interestingly, in a separate study, Harris and colleagues showed that the conformation of an actin filament can have a stark effect on the binding kinetics of a tandem CH domain reporter 49 . The authors measured the dwell time of tandem CH domain reporters on actin filaments in the presence of actin stabilizing drugs (phalloidin and Jasplakinolide) and other actin binding proteins (Drebrin, Heavy meromyosin, Cofilin) to show that the conformation of the actin filament can alter how well the reporter bound to actin filaments. As actin filaments are inherently helical, some bound proteins can change the pitch of actin and induce twisting or bending forces to give rise to high-order actin assemblies such as a ring. The bending or twisting of actin filaments can sterically influence the localization of other ABPs or result in preferential binding of specific ABPs (e.g., the Arp2/3 complex) 38 . Thus, there is a striking reciprocity in which the binding of ABPs alters the conformation of actin filaments, while the conformation of actin filaments affects which ABPs can bind.

Cofilin
Cofilin is a member of the ADF/cofilin family of actin binding proteins that are known for severing and/or depolymerizing actin filaments, thereby regulating actin filaments through the generation of dynamic actin disassembly 50-53 . Notably, Cofilin has been shown to sever bent actin filament regions more efficiently 54 . When Cofilin binds to actin filaments, the helical half-pitch of the decorated actin filaments (referred to as "cofilactin" filaments) changes: 37 nm for bare actin filaments and ~27 nm for cofilactin filaments 55-57 . The shorter helical pitch and the transition from bare to cofilactin filaments led to accelerated actin severing. However, saturated cofilactin filaments are no longer severed cofilin-dependently and are more flexible 50,52,58 . While examining actin bundles within filopodia at neuronal growth cones using cryo-electron tomography, Hylton and colleagues recently showed that cofilactin filaments associate with curvature of the actin bundles, thereby regulating filopodia structure and dynamics 59 . Interestingly, this cofilactin filament conformational change results in an exclusion of the Fascin actin-crosslinking protein from cofilactin bundles, leading to less organized actin filament bundling, the bending of filopodial protrusions, and a disruption of filopodial structure and functions ( Figure 2B) 59 . These results led the authors to propose that cofilin-mediated bending of actin filaments in filopodia would provide enhanced filopodia environmental searching/sensing to aid in the efficiency of targeted neural outgrowth 59 .

Septins
Septins are a family of highly-conserved GTP-binding proteins that form hetero-oligomeric complexes that polymerize into filaments and play important functions in many cellular processes, including cytokinesis, cell polarity, epithelial wound healing, and membrane remodeling, and when mis-regulated, are associated with neurodegenerative diseases and cancer 60-67 . These dynamic Septin structures function at the cell cortex where they associate with membrane composed of specific curvatures or lipid compositions, as well as with actin and microtubule cytoskeletal networks composed of specific architectures or properties 60,61,64,65,[68][69][70][71] . Given their various oligomeric assemblies and prime cortical localization, Septin complexes have been implicated in both membrane-and cytoskeleton-associated functions: i.e., aiding in the spatial segregation of cell compartments, crosslinking/bending cytoskeletal filaments, serving as scaffolds, and/or serving as membrane-bound diffusion barriers. Septins are divided into four different classes that assemble into hetero-hexamer or -octamer complexes that form higher-order structures, including rings, gauzes, and collars 64,69,77 . Oligomeric Septin filaments have been demonstrated to cooperate with actin filaments in dynamic structures such as the cytokinetic ring 78 . Septins often colocalize with bundled and curved actin filaments: in vitro assays have shown that both fly and human Septin hexamers can bundle actin filaments and bend them into various curved geometries such as rings and loops with diameters ranging 3-18μm 72 . Intriguingly, both fly and human Septins decorated actin filament bundles and rings without forming full rings themselves in vitro 72 , suggesting that Septins are not acting as a template for actin filaments to form a ring, but are instead bending actin filaments as they bind to the side of the filaments. Septin-mediated actin filament bending results in greater curvatures of the filaments ( Figure 2C). Using the rapid cellularization of the developing Drosophila embryo, Mavrakis and colleagues went on to demonstrate the function of Septin-induced curved actin filaments in vivo. They found that actin filaments formed bar-like structures in septin mutant embryos, rather than the ring-like structures observed surrounding the nucleus in wildtype embryos. Thus, both in vitro and in vivo assays conducted in this study strongly suggested that Septins play a role in bending and circularizing actin filaments by binding to the side of the filaments to facilitate formation of higher-order structures. Intriguingly, using cryo-EM, Mendonça and colleagues recently described a human Septin hetero-hexamer filament that tends to bend at its center where two Septins form a homodimeric interface 79 . This opens the possibility that when this bent Septin hetero-hexamer oligomerizes and binds to actin, it bends actin to form a ring structure.

Anillin
Anillin is a crosslinker of both Septins and actin that contains a pleckstrin homology domain allowing it to anchor to cell membranes 80 . As a prominent Septin binding protein, Anillin has been shown to also play a critical role in cytokinesis, particularly in organizing and anchoring the cytokinetic ring at the plasma membrane 78,81 . Although there is yet to be direct evidence showing how Anillin bends or curves actin filaments, recent studies have demonstrated that Anillin can generate contractile force 73 and that Anillin is required to properly orient Septin filaments at the cytokinetic ring 78 . In a nicely executed series of experiments, Kučera and colleagues showed that Anillin actively moves linear actin filaments and facilitates actin ring constriction by promoting force generation in a myosin-independent manner ( Figure 2D) 73 . Similar to Septins, Anillin formed actin rings in vitro, demonstrating its potential to bend actin filaments. In a second study conducted by Arbizzani and colleagues, the fission yeast Anillin ortholog, Mid2, was shown to participate in aligning Septin filaments parallel to the cytokinetic ring 78 . Using polarized fluorescence microscopy, they found that Septins remained randomly oriented when Anillin was deleted. Given the role of Septins in bending actin filaments (discussed above), Anillin may do more than simply bundling actin: Anillin may act to generate and facilitate organizational changes in Septins and actin to modulate the tunability and optimization achieved by actin assemblies.
α-Actinin and Fascin α-Actinin and Fascin are actin crosslinkers that can influence the mechanical characteristics of actin filaments/bundles and how they respond to different applied stresses or geometrical constraints 74,82,83 . Actin crosslinkers can dynamically regulate the stiffness of actin filament bundles and the stability of actin networks through bundling or bridging actin filaments in many cellular processes, including filopodia formation and focal adhesion assembly 84-86 . While both α-Actinin and Fascin can bundle individual actin filaments to form tightly packed parallel filaments, the structures of these resulting actin filament bundles are different. A combination of TEM and in vitro actin bundling assays showed that actin filaments bundled by α-actinin have ~35 nm space between each filament and form square lattice structures, whereas actin filaments bundled by Fascin have ~8 nm space and form a hexagonal lattice structure 74,83 ( Figure 2E). In vitro actin bundling assays also showed that α-Actinin and Fascin exhibit mutually exclusive localization on actin filaments 74,83 ( Figure 2F). Because these actin crosslinkers generate different spacings among actin filaments and can be present on different regions of these assemblies, actin filaments must bend to accommodate these differences 74,83 . Consistent with in vitro assays, localization of α-Actinin and Fascin is also mutually exclusive within cells. For example, Fascin localizes in the distal portion of filopodia, whereas α-Actinin localizes in the proximal region 87,88 . While the biological function of local actin bending by α-Actinin and Fascin has not been directly addressed, their coordination is thought to be important for cell stiffness 89 .

Myosins: conventional and unconventional
Myosins are a large family of motor proteins that move on actin filaments and are required for many cellular processes including muscle contraction, intracellular transport, endocytotic pathways, and left-right (L-R) asymmetric morphogenesis 90-97 . Among 15 different classes of myosins, Myosin II is known as conventional myosin and has been well-studied with respect to its ability to generate force through the concerted movement of actin filaments. Myosin II is bipolar and generates contractile force by binding to and sliding parallel actin filaments past each other. Surprisingly, recent studies showed that bipolar Myosin II can also bind to a single actin filament 75,98,99 . In this case, one end of the bipolar myosin protein is oriented towards the barbed end of the actin filament, whereas the other end is oriented towards the pointed end. As myosin preferentially moves towards the barbed end of actin filaments, the head oriented towards the barbed end will move faster than the head oriented towards the pointed end. In response to these different velocities either in the same direction or in opposing directions, the region of the actin filament between the myosin filament ends will bend/buckle ( Figure 2G).
Unconventional myosins are actin-based motor proteins that are grouped into distinct classes based on their motor and tail domains, which provide different structural and functional characteristics 94,100-105 . Their conserved N-terminal motor domains bind to actin and their class-specific C-terminal domains regulate the selection of and binding to cargos or other cell components 94,100-105 . Unconventional myosins have been studied with a focus on their function as cargo transporters, molecular anchors (to the actin cytoskeleton or plasma membrane), tension sensors, actin crosslinkers, signal transducers, and mediators of membrane-cytoskeleton interactions needed for a wide range of cellular processes, including endo-/exo-cytosis, organelle and membrane trafficking, cell motility, and L-R asymmetry 94,96,97,[102][103][104]106 . Type I unconventional myosins are tethered to the membrane and regulate cellular processes such as membrane tension and intracellular trafficking in cells 107 . While examining the role of type I unconventional myosins in the generation of L-R asymmetry in different Drosophila tissues, Lebreton and colleagues eloquently showed that two of these, Myo1D and Myo1C, generate de novo directional twisting of cells and organs in opposite directions to establish chirality [108][109][110] . Intriguingly, the motor domain of Myo1D, but not that of Myo1C, produced counterclockwise circular motility of actin filaments in vitro. In contrast to Drosophila Myo1C, mammalian Myo1C does produce a counterclockwise circular motility of actin filaments in vitro ( Figure 2H) 76 . A subsequent study examining L-R asymmetric morphogenesis in Drosophila showed that individual cells in the gut exhibit a directional chiral shape, and that such cell chirality is regulated by MyoID 111 . Intriguingly, this individual cell chirality induces the L-R asymmetric morphogenesis of the Drosophila gut 111 . Overall, directional actin motility regulated by type I Myosin may generate the force to deform the cell into a chiral shape that subsequently determines L-R asymmetry in the tissues.

Conclusions
Actin networks must dynamically change their configurations to generate the forces needed within cells for many of their essential biological processes 1,3-5,7,10-16 . The ability of proteins to bend individual or groups of linear actin filaments is emerging as an important contributor to the dynamic regulation of higher-order actin assemblies required for cellular processes such as the formation of an actin ring, cell migration, and endocytosis 48,72,73,112,113 . Bending actin filaments directly regulates actin network geometries, provides directional actin branching for ABP binding (i.e., Arp2/3 prefers binding to the convex side of actin filaments 38 ), provides regions for more efficient severing 54 , and can be used to generate force by unbending. In addition, a recent study suggests that nucleotidedependent modifications of the actin filament conformational landscape influences the binding specificity of ABPs 42 . Intriguingly, α-Actinin, IQGAP, and unconventional myosins also induce directional actin filament movements such as counterclockwise swirling. This actin-dependent cell chirality is observed in a wide range of animals from Drosophila to mammals 57,111,[114][115][116][117][118][119][120][121] . While much of the evidence for actin bending and its biological consequences to date comes from studies using in vitro systems, recent studies using cells and model organisms revealed that cell chirality has important roles in processes such as cell migration and L-R asymmetric organ development 97,106,122,123 . We still have much to learn about how altering the conformation of actin filaments by bending them into curved geometries affects the functions of actin assemblies, results in chirality, and the means by which it is used to achieve precise developmental outcomes. In addition, it remains unclear how ABP binding induces actin filament bending and directional movement of actin filaments. The combination of cryo-EM, super-resolution microscopy, cutting-edge live-imaging techniques, single-cell RNAseq, and computer simulations can aid in further understanding the mechanistic insights and revealing the mysteries still surrounding the role of actin bending.