Actin bends over backward for directional branching

The actin cytoskeleton is the major force-generating machinery in the cell that produces pushing, pulling, and resistance forces. To perform these functions, actin filaments, with the help of many accessory proteins, form architecturally distinct structures designed for specific purposes. Thus, pushing forces are frequently generated by branched networks that assemble in the vicinity of a load and exert force using energy of actin polymerization (1, 2). Although the current level of molecular and biophysical understanding of this process is exemplary (3, 4), a key remaining question is how to maintain the directionality of the constantly branching network and prevent it from expanding into unwanted cell areas. The report by Risca et al. in PNAS (5) links the directionality of actin branching to the load-imposed curvature of actin filaments. This connection supports a direct mechanosensing role of actin filaments and explains the tunneling of actin polymerization toward the load.

The molecular machinery responsible for the assembly of branched actin networks consists of a handful of proteins that were sufficient to reconstitute motility in vitro from purified components (6). A key component of the machinery is the Arp2/3 complex, a heteroheptameric protein that nucleates a new “daughter” filament as a branch on the side of a preexisting “mother” filament at a defined angle of 70°, a process called “dendritic nucleation.” After a period of elongation, growth of branches is terminated by capping proteins that bind to the growing “barbed” ends of actin filaments, and new filaments are nucleated by the Arp2/3 complex to maintain force generation (2).

The dendritic nucleation machinery has many advantages for pushing force generation, explaining its broad repertoire of cellular functions that includes protrusion of lamellipodia in migrating cells, rocketing motility of membrane organelles and intracellular pathogens, formation of cell–cell junctions and synapses, and biogenesis of various intracellular organelles (1). Indeed, anchorage to mother filaments allows daughter filaments to do useful work immediately after birth instead of slipping backward in the absence of traction. Moreover, these nascent branches produce force while they are still relatively short and stiff, whereas longer and less productive filaments are capped. An angled orientation of pushing actin filaments relative to the load is also useful, as it makes it easier to insert an actin monomer between the growing barbed end and the load (7). A major problem with dendritic networks, however, is that the direction of filament elongation, and thus force application, changes by 70° with each nucleation event. If the direction of new branching events is not properly controlled, the network may reverse the direction of pushing in just a few rounds of branching or expand into an unpolarized 3D sphere (Fig. 1A, Left). However, neither of these possibilities seems to occur in biological settings, where most actin filaments remain oriented with their growing barbed ends roughly toward the load (8, 9).

Fig. 1.

Actin bending under load promotes branching in a forward direction. (A) Predicted architecture of branched actin networks depending on a mode of regulation. Combination of NPF targeting, barbed end capping, and curvature-dependent branching predicts optimal directionality of dendritic networks. (B) Actin filament bending creates binding sites for the Arp2/3 complex on the convex surface. (Left) Resistance of the load causes the actin filament to bend, thus creating a binding site (actin subunits outlined in red) for the Arp2/3 complex on the convex side. (Right) Activated Arp2/3 complex binds to the convex side of the mother filaments and nucleates a new branch toward the load.

Several nonexclusive ideas have been developed over time to explain directionality of branching. One idea is based on the subcellular localization of nucleation-promoting factors (NPFs) that activate the intrinsically inactive Arp2/3 complex. Targeting of NPFs to a load (the plasma membrane or a propelled organelle) would favor nucleation on the nearby mother filaments because of greater availability of the active Arp2/3 complex (10, 11). However, even when nucleation occurs on a mother filament close to the load, the nascent branch may grow away from the load, where it would find less resistance. This concern is addressed by another idea that suggests that barbed ends facing away from the load are more likely to be capped (10), because the barbed ends abutting the load may be protected from capping either directly due to physical obstruction by the load or indirectly through load-associated anticapping proteins (12). This mechanism does not prevent unwanted polymerization in a wrong direction, but helps to quickly terminate it (Fig. 1A, Center).

Here, Risca et al. present a unique model of how actin filament branching may be biased to the forward direction (5). The model is built on their observation that the Arp2/3 complex preferentially forms branches from a convex side of a curved mother filament, whereas the concave side is relatively inferior. To obtain these data, Risca et al. “pin down” preformed actin filaments to the glass surface through intermittent biotin–streptavidin linkages. This procedure creates randomly curved actin filaments, which serve as mother filaments for subsequent nucleation that is induced by addition of the activated Arp2/3 complex and actin monomers. By labeling mother and daughter filaments with red and green fluorophores, respectively, the authors can observe the resulting branched actin filaments by fluorescence microscopy.

How does nucleation from a convex side of a mother filament orient branches forward? If we consider the typical geometry of branched networks, where the majority of filaments is oriented at ±35° relative to the overall direction of the pushing force (10, 13), it is easy to see that the resisting force from the load should bend pushing actin filaments in such a way that their convex surfaces face the load. When the Arp2/3 complex preferentially binds these convex surfaces, it forms new branches toward the load and against the resisting force, exactly as required for the best system performance (Fig. 1A, Right). Although the magnitude of the experimentally observed preference for convex mother filaments is quite small, Monte Carlo simulations performed by Risca et al. demonstrate that even such a small bias in nucleation is sufficient to account for a significant shift in the bulk actin polymerization toward the load in just a few rounds of branching.

What is the biochemical basis of the Arp2/3 complex preference for the convex side of the mother filament? In fact, findings of Risca et al. fit perfectly well to the available structural and biochemical data concerning Arp2/3 complex interaction with the mother filament. Although the crystal structure of the Arp2/3 complex within the branch is not available, fitting the crystal structure of the inactive Arp2/3 complex (14) into electron tomography images of branch junctions shows that to accommodate the Arp2/3 complex, three subunits on one side of the mother filament need to partially break their longitudinal interactions and move slightly away from each other with the formation of two gaps in the filament lattice, thus exposing the binding surfaces for the Arp2/3 complex (15). Spontaneous separation of actin subunits, especially at two neighboring positions, is unlikely in an unstressed actin filament, explaining a very slow rate of Arp2/3 complex binding to actin filaments in biochemical assays (16). However, filament bending is expected to

Risca et al. present a unique model of how actin filament branching may be biased to the forward direction.

increase the probability of gap formation on the convex side of the filament, thus creating binding sites for the Arp2/3 complex (Fig. 1B).

The beauty of the curvature-dependent control of branching is that it does not involve a complicated signaling cascade, but provides what appears as the most direct and elegant way to orient force toward the load. Indeed, the key regulator of this system is the actin filament itself. Accumulating evidence suggests that actin filaments exist in multiple structural states and subunits within a filament have a certain degree of conformational freedom (17). This flexibility allows actin filaments to change conformation in response to force. These subtly different, but functionally distinct, structural states can preferentially bind particular actin-binding proteins, while disfavoring others. The study by Risca et al. demonstrates this concept for the Arp2/3 complex, but other examples have also emerged recently. Thus, the actin-severing protein cofilin preferentially binds to and severs relaxed actin filaments, whereas actin filaments under tension are relatively resistant to cofilin (18). Conversely, the actin-dependent motor myosin II preferentially interacts with stretched actin filaments (19). Together, these studies point to an emerging “actin-centric” concept of cellular mechanosensing, according to which the conformational state of an actin filament serves as a readout of the mechanical environment and is detected by actin-binding proteins, which then execute proper actions.