Biophysics of Actin Filament Severing by Cofilin

Abstract

The continuous assembly and disassembly of actin filament networks is vital for cellular processes including division, growth, and motility. Network remodeling is facilitated by cofilins, a family of essential regulatory proteins that fragment actin filaments. Cofilin induces net structural changes in filaments that render them more compliant in bending and twisting. A model in which local stress accumulation at mechanical discontinuities, such as boundaries of bare and cofilin-decorated filament segments, accounts for the cofilin concentration dependence of severing, including maximal activity at sub-stoichiometric binding densities. Real-time imaging of cofilin-mediated filament severing supports the boundary-fracture model. The severing model predicts that fragmentation is promoted by factors modulating filament mechanics (e.g. tethering, cross-linking, or deformation), possibly explaining enhanced in vivo severing activities.

Self-association (i.e. polymerization) of the protein actin into linear double-stranded helical filaments provides forces that drive eukaryotic cell motility (1; 2). Central to developing predictive models of motility is quantitative knowledge of actin behavior and how it is influenced by the vast number of actin regulatory proteins in cells. The majority of actin-based motility models are derived from biochemical and biophysical studies carried out in solution with purified protein components.

Assembled actin filaments elongate (e.g. grow and shrink) only from their ends (3). Sustained motility requires the regeneration of assembled subunits and replenishment of the actin monomer pool. Accordingly, factors that increase the filament end concentration provide a greater number of sites from which subunits can add or dissociate, thereby accelerating the overall subunit flux of filament networks.

Filament subunit interactions are non-covalent, yet filaments in solution assembled from purified actin can reach several microns in length (~365 subunits per micron) because of the stabilizing lateral and longitudinal contacts associated with the double-stranded filament geometry (4; 5). The mechanical properties of filaments are comparable to some familiar commercial plastics, enabling filaments and filament networks to sustain forces associated with cell motility as well as provide cells with mechanical strength and integrity. However, filaments in solution at their assembled microscopic length scales behave as semi-flexible polymers that readily undergo thermally driven bending and twisting shape fluctuations.

The actin regulatory protein cofilin binds and severs actin filaments in vitro and in vivo (6–9). Cofilin is essential (10; 11) and dysfunction is associated with human pathologies (12; 13). The recognition that filaments are mechanically stiff, yet readily fragmented by regulatory proteins, has attracted the utility of physics and engineering formalisms, specifically those developed in polymer mechanics. Severing is driven by cofilin binding interactions and linked equilibria; no external energy sources (e.g. ATP hydrolysis) are required. Elucidation of the cofilin severing mechanism has therefore benefited from interpretive power of physical chemistry, specifically thermodynamics and kinetics.

The contents of this review article focus on the later two aspects - physics and chemistry - of cofilin severing activity. Given space limitations, we focus on recent developments with the understanding that considerable bodies of work preceded and facilitated these advances. We encouragingly direct readers to these earlier works (cited within papers cited in this review) to fully grasp current understanding of the cofilin severing mechanism. We note that an extensive body of literature exists evaluating the interaction of cofilin with actin monomers, but these works are not discussed here.

Cofilin binding thermodynamics and kinetics

Cofilin binds actin stoichiometrically - one cofilin molecule per actin filament subunit (14–17) - and, in some cases, with positive cooperativity. Vertebrate cofilin binds vertebrate actin filaments cooperatively and equilibrium binding isotherms display a characteristic sigmoidal shape, independent of the detection method used to assay binding (e.g. cosedimentation (18; 19), fluorescence (17; 20), or phosphorescence (21)). Some non-vertebrate cofilins bind their native, species compatible actin non-cooperatively (22), while binding of some cofilins (e.g. S. cerevisae) is so strong that it precludes reliable detection of cooperative interactions should they exist (23).

Studies assessing cooperative binding commonly employ the familiar Hill Equation and yield Hill coefficients ranging from 4–40, indicating (relatively) weak positive cooperativity (14–17). This is a surprising behavior because actin-bound cofilins do not directly interact (24). Cooperative interactions must therefore arise from linked changes in (average) filament structure (25–28) and/or dynamics (21; 29; 30; 23).

Despite being an assuring indicator of cooperative interactions, the Hill formalism is limited in its ability to translate observed binding behaviors into meaningful thermodynamic binding parameters, as experimentally measured apparent binding constants represent weighted average composites of individual binding events (31). A one dimensional Ising lattice model with nearest neighbor cooperative interactions (32) has proven to be more informative than the more conventional Hill analysis, despite it not explicitly accounting for actin filament subunit geometry.

A two-state Ising model in which each lattice site (i.e. an individual actin subunit) exists as vacant or occupied by a cofilin molecule (Figure 1) reliably describes both the equilibrium (17) and kinetic (33) binding of vertebrate cofilin and vertebrate actin, and provides intrinsic binding constants (K_int) and cooperativity parameters (ω). According to this model, bound cofilin can exist in one of three distinct binding modes: a) isolated, with neighboring sites vacant and overall binding affinity K_int, b) singly contiguous, with one of the two neighboring sites occupied and overall binding affinity K_intω, and c) doubly contiguous, with both neighboring sites occupied with overall affinity K_intω².

Figure 1.

One-dimensional Ising lattice model with nearest-neighbor cooperative interactions.

The body of equilibrium and kinetic data collected to date suggest that vertebrate cofilin binds with an intrinsic binding affinity of ~10–20 µM and a cooperativity parameter of ~8–20 (20). Cofilin binding is slow when compared to most other actin binding proteins (34) and on the timescale of ligands that bind cryptic filament sites (e.g. phalloidin (35; 36)). Time courses of cooperative association identified a slow isomerization subsequent to cofilin encounter with actin, indicating that bound cofilactin adopts (at least) two distinct conformations (33).

The nearest-neighbor Ising model lacks structural information and considers only nearest neighbor cooperative interactions. Although evidence for non-nearest neighbor cofilactin interactions exists (37; 38), they are not required to account for cooperative equilibrium (17; 8; 31) and kinetic binding data (33). More sensitive assays are needed to determine if non-nearest neighbor effects contribute to cofilin binding cooperativity. We note that although a two-state Ising model accounts for cooperative vertebrate cofilactin interactions, more reliable fits to non-vertebrate cofilin equilibrium binding data are obtained with a four-state conformational ensemble model that considers intrinsic equilibria among two distinct cofilin-occupied conformations as well as two distinct vacant states (31).

When binding is positively cooperative, clusters of bound cofilin will form along filaments. Knowledge of the various cofilin binding mode affinities (defined above) allows for prediction of the average (and distribution)(39) of cofilin cluster sizes at a given concentration (17). The predicted cofilin cluster sizes are small – with an average size of near unity (17)– at cofilin concentrations yielding efficient severing (10; 40), indicating that only one or few cofilin proteins are sufficient to sever filaments.

Cofilin binding is surprisingly weakling dependent on temperature over the 4–37°C range, indicative of a modest interaction enthalpy and a binding equilibrium that is driven primarily by positive entropy changes (20). The entropy changes driving cofilactin interactions must originate from reorganization of solvent, salts and/or protein (cofilin and actin) since these are the only components present in experimental samples. Solvent contributions appear to be small, as assessed from the dependence of binding free energy and cooperativity on inert crowding agents (20), see Ref. (41) for additional discussions.

Cofilin binding is salt-dependent, as one might expect for binding to a charged polymer (42). The intrinsic cofilin binding free energy is affected by salt, but the cooperativity is not (20). Salt-dependent cofilin binding has been interpreted in terms of a competition between cofilin and filament-associated cations (20) that likely bind at specific and discrete sites on filaments (43). Isolated cofilin binding is linked to release of a single Mg²⁺ (or approximately two K⁺) (20). Cooperative interactions have no apparent linkage to ion release, suggesting that actin conformational changes associated with cofilin binding and ion release are not propagated cooperatively beyond sites with bound cofilin.

Actin filament conformational dynamics

Ion release contributes partially to the entropic changes driving cofilin binding. The remaining entropy changes originate from enhanced conformational fluctuations of the cofilactin complex. The molecular origins of these changes manifest as increases in spontaneous, thermally driven actin filament shape fluctuations upon cofilin binding. Such filament conformational dynamics concern both large-scale filament shape changes as well as small amplitude fluctuations of individual subunits, and thus span a broad range of timescales (Figure 2A). Experimental and computational studies indicate that cofilin alter actin conformational dynamics on all of these length and time scales.

Figure 2.

(A) Actin filament conformational fluctuations span broad timescales. Filaments display large-scale tumbling, bending, and twisting motions. Individual actin subunits undergo sub-microsecond timescale dynamics that modulate subunit interfaces. (B) Cofilin binding introduces mechanical discontinuity that promotes filament severing.

Filament bending and twisting deformations occur on the microseconds to seconds time scale, depending on the amplitude of the motion. Single filaments in solution display diffusive tumbling, though these motions are very slow and negligible on the seconds time scale associated with cofilin-mediated severing events. Analysis of actin filament shape fluctuations generally employ conditions that eliminate or reduce tumbling motions.

Conformational dynamics of individual actin subunits span the same breadth of timescales as a typical globular protein, with functional motions ranging from millisecond to picoseconds timescales. Domain rotational motions or local folding/unfolding reactions on millisecond to microsecond timescales modulate subunit stability, the interface between neighboring subunits, and potentially regulate accessibility to binding partners (44–47). Additionally, loop fluctuations and side chain rearrangements on nanosecond or picoseconds timescales may regulate solvent or ion interactions that modulate filament stability (43).

Cofilin modulates actin subunit structure and filament mechanics

Cofilin binding yields net structural changes at both the subunit and filament scale. On the subunit scale, a conformational change occurs in the DNAase I binding loop (48–51) as well as the actin C-terminus (52). Otherwise cryptic sites within subunits become accessible to protease cleavage (44). On the filament scale, a change in twist and subunit tilt is observed (28; 27; 24), which could destabilize both the longitudinal (48) and lateral (53) filament subunit interfaces. These filament structural changes presumably contribute to cooperative binding since no direct contacts exist between bound cofilin molecules (24).

Computational modeling has provided valuable insight concerning the atomic level changes in the actin subunits and filament mechanical properties that result from cofilin binding (30; 54). Coarse-grain and molecular dynamics simulations of actin filaments (30; 54) reveal an outward radial movement of the DNAase I binding loop in actin subdomain 2. This conformational change alters the long-axis subunit interface, potentially destabilizing filament contacts, as inferred from biochemical solution studies (48).

The cofilin-mediated reorganization of actin subdomain 2 presumably narrows the overall filament radial contact distribution between actin subunits (29) yielding cofilactin filaments that bend and twist more readily than their bare counterparts. Direct visualization of filaments undergoing thermally-driven fluctuations in shape allows for measurement of filament flexural rigidity (55–57). The apparent cofilactin filament Young’s modulus is ~20-fold lower (after accounting for the predicted radial mass distribution change) than that of bare filaments (29). As a results, cofilin binding increases the thermally-driven filament bending amplitudes and modes (23)(Figure 2B). Cofilin also makes actin filaments ~18-fold more compliant in twisting (21).

Cofilin promotes severing at mechanical discontinuities

It has been proposed that discontinuities in filament mechanics and dynamics generate local accumulation of stress at boundaries of bare and cofilin-decorated segments (17; 58; 29; 38; 23; 59; 19). Such a model predicts that bare and fully decorated filaments sever less readily than partially decorated, which has been observed (17; 58; 29; 59). Direct visualization of cofilin, actin, and subsequent filament severing events demonstrate that the severing probability is higher at boundaries than within bare or cofilin-decorated segments (38). The model also predicts that cofilins that do not alter filament mechanics will not sever. Consistent with this prediction, vertebrate cofilin minimally alters yeast actin filament flexibility and weakly severs them (23).

Severing at mechanical discontinuities may also contribute to severing of filament bundles and networks. Cofilin severs fascin cross-linked bundles more readily than single filaments, despite reduced access to filament binding sites (60). These results suggest that both the stiffening effect of fascin (61), as well as the non-uniform distribution of cofilin on the bundles due to competitive fascin binding, enhances cofilin severing activity by promoting more pronounced mechanical discontinuity.

In contrast, Arabidopsis thaliana villin 1-crosslinked filament bundles are resistant to ADF/cofilin-mediated depolymerization (62). The factors contributing to the distinct severing of fascin and villin bundles are not understood. Competition and limited site accessibility is likely to influence severing efficiency towards bundles, and may fully account for the inhibiting effect of villin (62). However, the enhanced severing of fascin bundles (60) favors a mechanism that overcomes reduced site accessibility. Differences in crosslinker compliance or effects on filament mechanics could modulate cofilin severing efficiency, as can the crosslinker binding mode (e.g. contacting multiple filament subunits could accumulate stress at the subunit interfaces more effectively than contacts made predominantly with an individual subunit, as in the case of villin (63)).

The critical bending angle (i.e. angle at which bending deformations become irreversible) is smaller for fragmentation at bare and cofilin-decorated segment boundaries compared to bending sites within the segments (23). This behavior suggests that boundaries (either between cofilin-decorated segments and bare or potentially certain actin-binding protein decorated segments) are more susceptible to fracture than non-boundary segments, analogous to the adhesive joint failure of non-protein materials(64) . Such behavior could result from the altered actin subunit interfaces resulting from the filament twist and subunit tilt that occurs upon cofilin binding.

Although the proposed severing mechanism described here applies to vertebrate cofilin severing, not all cofilins may function in the same way. In contrast to vertebrate cofilin, yeast cofilin that is competent to sever yeast actin may act through a different mechanism (23). Other severing proteins such as gelsolin require other cofactors such as calcium and bind to other sites on the actin filament (65; 66).