Single Filaments to Reveal the Multiple Flavors of Actin

Abstract

A number of key cell processes rely on specific assemblies of actin filaments, which are all constructed from nearly identical building blocks: the abundant and extremely conserved actin protein. A central question in the field is to understand how different filament networks can coexist and be regulated. Discoveries in science are often related to technical advances. Here, we focus on the ongoing single filament revolution and discuss how these techniques have greatly contributed to our understanding of actin assembly. In particular, we highlight how they have refined our understanding of the many protein-based regulatory mechanisms that modulate actin assembly. It is now becoming apparent that other factors give filaments a specific identity that determines which proteins will bind to them. We argue that single filament techniques will play an essential role in the coming years as we try to understand the many ways actin filaments can take different flavors and unveil how these flavors modulate the action of regulatory proteins. We discuss different factors known to make actin filaments distinguishable by regulatory proteins and speculate on their possible consequences.

Main Text

A striking feature of living cells is the variety of mechanisms involving actin assembly, and the diversity of filament networks being formed. Long-lived structures like the cell cortex, lying flat against the plasma membrane, are maintained and reorganized to allow the formation of specific transient structures such as the cytokinetic ring at the end of cell division. These diverse filament architectures and dynamics are tightly controlled by the cell (see (1) for a recent review). This can only be achieved if actin assembly is precisely regulated in space and time.

At the base of all these filament networks lies a universal assembly mechanism of actin monomers (G-actin, for “globular actin”) into actin filaments (F-actin, for “filamentous actin”) (2). Actin filaments have a double-helical structure with structurally and dynamically distinct ends, the barbed end being more dynamic than the pointed end of the filament (2, 3). G-actin has a strong affinity for ATP and thus filaments mostly assemble from ATP-G-actin. Upon incorporation into a filament, the actin subunit changes conformation, locking its nucleotide in and hydrolyzing it, rapidly becoming ADP-Pi-F-actin and then ADP-F-actin (following the release of inorganic phosphate Pi). The F-actin nucleotide state thus marks the age of the filament and modifies its mechanical and biochemical properties (4).

To form the observed variety of filament networks, these assembly properties are modulated by regulatory proteins. Hundreds of these proteins have been identified, and we will refer to them in general as actin-binding proteins (ABPs). They can bind to filament sides, filament ends, and/or monomers and have a variety of effects (see (2) for a structural review). Typically, nucleators will initiate the assembly of new filaments, whose elongation can be modulated by end-binding proteins that can enhance (e.g., formins, Ena/VASP) or block (e.g., capping proteins) the addition of new monomers, and filaments will eventually be disassembled by side-binding proteins (e.g., actin depolymerizing factor (ADF)/cofilin, Aip1). Filaments can be connected to one another by other side-binding proteins, giving rise to specific filament organizations (e.g., fascin or α-actinin form parallel or antiparallel bundles, respectively). Depending on the regulatory proteins at play, various filament assemblies can be generated, with different geometries, dynamics, and life-spans (1).

Our ability to observe and manipulate single actin filaments in vitro has made tremendous progress, to the point where we now observe individual reactions taking place live (5, 6). In this review, we will begin by summarizing the progress made by single filament techniques and explain how this experimental approach has contributed to our understanding of how ABPs regulate actin assembly.

How different filament networks can emerge, as a result of the action of ABPs, is still not entirely clear. In particular, it seems that many ABPs can tell filaments apart, thus acting on a subset of actin networks in cells (7, 8). This feature questions the uniqueness of actin filaments in cells, suggesting that actin filaments could come in different “flavors.” This will be discussed in the second part of this review, and we will argue that single filament techniques will be essential in bringing answers to this tantalizing question in the near future.

A brief history of single filament revolutions

Being able to manipulate and observe individual actin filaments is of great help in understanding how their assembly is regulated and to what extent molecular details matter. Some of the earliest details on actin filament dynamics were actually obtained by observing single filaments with quantitative electron microscopy (EM) (3, 9), with images obtained by fixing samples at different time points. Following these pioneering works, however, bulk solution measurements based mainly on the increase of signal resulting from the assembly of pyrene-labeled actin monomers into filaments (10) have become a method of choice. They have been essential to improve our understanding of actin assembly dynamics at both ends, of how ATP hydrolysis follows filament assembly, as well as of how regulatory proteins tune actin steady-state properties (see (4) for a review). Static structural details are also required to understand the filament and protein conformational changes that underlie the reactions one can monitor live, and the contributions of crystallography as well as cryo-EM studies (which are undergoing a revolution of their own (11)) are essential (2, 12, 13, 14). Here, our intention is to focus on what the live imaging of single filaments has to offer to the actin field.

Live imaging relies mostly on fluorescent probes to report kinetics events in the sample. The canonical in vitro experimental setup is a chamber of a few tens of microliters with actin filaments anchored or maintained in the vicinity of the coverslip to be imaged by a microscope objective (Fig. 1 A). Since the 1980s, reconstituted actin filaments, from purified actin or cell extracts, can be visualized using fluorescently labeled phalloidin (15). The disadvantage is that it stabilizes filaments and alters the binding of some proteins, and thus does not allow the proper observation of individual reactions live. The first real-time observations of actin dynamics were achieved in the early 2000s (16, 17). This revolution was made possible thanks to the use of total internal reflection fluorescence (TIRF) microscopy that overcomes the limitation of fluorescent background (18) due to monomeric actin present in solution in the micromolar range (Fig. 1 A).

Figure 1.

Some of the main experimental configurations for the live observation and manipulation of single actin filaments and regulatory proteins. (A) In the canonical setup, TIRF illuminates fluorophores in a shallow region above the coverslips, and labeled filaments are anchored to the surface. (B) Microfluidics takes advantage of fluid friction (viscous drag) to align filaments with the flow, close to the coverslip, with only one end anchored to the surface. (C) Tweezer techniques (here, optical tweezers) manipulate filaments by anchoring their ends, or portions near their ends, to microbeads. (D) The background fluorescent signal from solution can be reduced further using convex lens-induced confinement or zero mode waveguides. (E) A low background enables the observation of individual molecules (here, the Arp2/3 complex). (F) High speed AFM can be used to monitor filament dynamics live. (G) Surface patterning techniques have mostly been applied to large populations of actin filaments, but can also be used to study individual filaments. To see this figure in color, go online.

The live observation of dynamic actin filament assembly provided a direct confirmation of the elongation rates determined some 20 years earlier with fixed samples. Compared to bulk solution measurements, the direct observation of individual filaments, as reactions are taking place, gives access to new information. Key results include showing that a formin processively tracks an elongating barbed end (19, 20), showing that Arp2/3-generated filaments branch off existing ones with a well-defined geometry (21) (Fig. 1 E), and seeing that cofilin-induced filament fragmentation takes place near boundaries between bare and cofilin-decorated regions (22). These results, which are essential to our understanding of these molecular mechanisms and their consequences in a cellular context, were made possible by the observation of individual events under the microscope.

Bringing down technical hurdles and limitations

Images are powerful and often give the compelling impression that they faithfully depict reality. Naturally, one should be cautious and a number of controls have to be performed. Two technical limitations, inherent to a single filament experiment, readily come to mind: filaments are stuck to surfaces and labeled with fluorescent molecules. These manipulations could bias results and generate artifacts. Fortunately, there are ways to control and limit these effects as new techniques emerge and combine.

Interactions between filaments and the glass coverslip, either nonspecific or mediated by the anchoring points, may affect protein reactions and filament dynamics. Tethering points can cause uncontrolled mechanical stress, as well as occupy binding sites on the sides or the ends of filaments. Control experiments can be performed by changing surface passivation techniques and anchoring strategies. The density of tethering points can be varied, and brought down to very low values. Surface patterning techniques (Fig. 1 G) allow one to control the distribution of anchoring points, thereby creating specific geometries for filaments (23). Crowding agents such as methylcellulose can be used to maintain filaments close to the coverslip without tethers, but they induce filament bundling (24) and potentially affect reaction rates. Using microfluidics (Fig. 1 B) offers the possibility to maintain filaments close to the surface, while anchoring one end only and without using crowding agents (25).

Fluorescent labeling is a notorious source of artifacts, from classical light-induced filament fragmentation to the recently elucidated light-induced formation of covalent dimers within actin filaments (26). A simple way to evaluate the occurrence of such artifacts is to vary labeling fraction and/or light exposure. Different fluorophores can be used, and their location on the protein of interest can be varied. Microfluidics, which provides time-controlled exposure of filaments to different protein solutions, can be used to generate filaments with unlabeled segments, and to monitor elongation from unlabeled actin monomers (25, 26). High-speed atomic force microscopy (AFM) can also be used to monitor live binding of ABPs on filaments, and does not require the labeling of proteins with fluorophores (27, 28) (Fig. 1 F). In addition, this technique provides useful structural data, such as the local pitch and width of the actin filament (28). A limitation, however, is the requirement to immobilize filaments on the surface (see previous paragraph).

Today, cameras are sensitive enough to detect single molecules and track them over time. Different proteins can be monitored individually, using different labels, and their colocalization and the sequence of their respective arrival on a given site can be determined (5, 29). Such observations are a great asset to decipher molecular mechanisms involving several ABPs.

One important limitation for these observations is background signal, either coming from the surface or from solution (typically, single molecule observations in TIRF require working with <100 nM of fluorescent molecules). Surface background should be minimized by proper passivation. New microscopy techniques are available to reduce background from bulk (30), pushing the enhancement of signal/noise ratio further than TIRF does (Fig. 1 D). Convex lens-induced confinement locally confines the fluorescent molecules in a small volume by lowering the height of the chamber dramatically, down to a few nanometers. Linear zero mode waveguide confines the excitation to a thin layer above the glass coverslip (one order of magnitude shallower than in TIRF) and also confines the sample in a narrow strip, thus reducing the excitation volume. Both techniques enable the use of higher concentrations of labeled proteins, thereby giving access to interactions with weak affinities.

Single filament techniques can also be used to move filaments around and apply forces. Optical and magnetic tweezers (Fig. 1 C) are particular cases of techniques that are very adapted to apply well-controlled forces to single filaments or molecular motors. They have been used successfully to provide a wealth of details on molecular machineries related to actin, myosin motors in particular (31). AFM also allows one to apply forces, but in a range of magnitude and with an orientation that are less adapted to single filament studies. Microfluidics can also be used to apply tension to filaments or motors, taking advantage of the viscous drag of the flowing fluid (Fig. 1 B). The applied force fluctuates more than with tweezer techniques, but it is applied to tens of filaments in parallel and thus generates larger amounts of data (25).

The downside of observing individual events is that a large amount of data needs to be collected to provide statistically significant results. However, single filament techniques allow one to observe several individual filaments simultaneously. Motorized microscope stages further increase the number of observed filaments by collecting data from different fields of view. The ensuing data analysis can be automated, at least partly, by computer software to enhance its efficiency.

Naturally, single filament techniques are not the only tool providing information on actin assembly. Rightfully, an increasing number of studies combine in vitro and in vivo observations to strengthen their conclusions. Moreover, labeling and imaging techniques from both fields are progressively converging, and single filament techniques benefit from developments in live cell imaging. For example, green fluorescent protein fusion proteins are now also used for single molecule experiments in vitro (32) and we expect more superresolution techniques to be used in single filament studies (33). The high spatial resolution reached by these techniques might not suffice to identify protein binding sites and other approaches, such as cryo-EM and crystallography, will remain invaluable to decipher molecular mechanisms.

Changing the way we look at actin filaments

The technical developments of single filament techniques have recently brought a number of striking and unexpected results that could not be inferred from earlier, less accurate methods. Part of these results come from the ability to observe a mechanism involving several regulators, either by tracking single molecules or by using microfluidics to expose filaments sequentially to different proteins. For example, monitoring individual branching reactions with labeled Arp2/3 complexes has revealed that only a small fraction of the Arp2/3 binding events led to the formation of a branch. Moreover, these experiments showed that the departure of the activator of Arp2/3 was required for the branch to form (5, 6). Two different studies monitoring single molecules have shown that Aip1 fragments cofilin-decorated filaments with enhanced efficiency (34, 35). Microfluidics has been used to sequentially expose filaments to formin Fmn2 and Spire, which both compete for filament barbed ends, and show that their direct interaction leads to a ping-pong mechanism where the recruitment of one protein at the barbed end is favored by the presence of the other, which seems to immediately depart, avoiding the formation of a complex at the barbed end (36). In contrast, although Capping Protein (CP) and formins were thought to mutually exclude each other from barbed ends, two very recent studies, using either microfluidics (37) or single-molecule monitoring (29), have revealed that these two proteins could coexist at the barbed end for tens of seconds.

In all the examples we have just cited, quantitative measurements were derived from single-filament observations. It would be a terrible mistake to consider these techniques as mere providers of illustrations for phenomena already quantified through different means, since one would miss out on what they have to offer as powerful biophysical and biochemical tools that can be used to reliably measure rate constants.

Single filament techniques now bring us to a point where we are able to see individual reactions live, and on filaments that we manipulate in a controlled manner (sequential exposure to different proteins, application of mechanical stress). Future improvements will keep on expanding these tools. For example, progress in fluorescent probe chemistry will hopefully provide new ways to monitor conformation changes live under the microscope.

These powerful tools will play a crucial role in addressing the formidable challenge of deciphering the regulation of actin assembly. In the rest of this review, we would like to expand on a key aspect of this challenge: the emergence of the notion that, rather than one generic actin filament, we face a system where filaments can have different flavors, which determine how they are regulated.

The many flavors of actin

As we have mentioned in the beginning, there is more than one type of actin, and we often need to specify which actin we are referring to. Obviously, we distinguish G-actin from F-actin, and the terminal subunit at the barbed end is in an intermediate conformation, as illustrated by its enhanced Pi release rate compared to that of F-actin subunits (38). The nucleotide state of a subunit also matters, and so does the nature of the surrounding divalent cations (Ca²⁺ or Mg²⁺, the latter being considered physiological) because they affect interactions between neighboring subunits in the filament (see (39) for a review).

These different flavors are relevant in cells, where they can be recognized by ABPs and thus modulate their regulatory action. ADF/cofilin for example, will bind preferably to ADP-actin, forming domains on the older portion of the filament while leaving the ADP-Pi-actin region near the barbed end unaffected (22).

In the following paragraphs, we will go through some of the other factors that are emerging as means to distinguish actin filaments, i.e., to give them different flavors, which allow ABPs to tell them apart and thereby enable cells to regulate them in different ways. They are recapitulated in Fig. 2.

Figure 2.

Different flavors of actin filaments, and how they affect the binding of ADF/cofilin. Starting at the top, and going clockwise. ADF/cofilin binds preferably to ADP-actin rather than ADP-Pi-actin, Its binding and fragmentation activity is delayed if the filament is under mechanical tension. Fragmentation by ADF/cofilin is more efficient if filaments are bundled by fascin (though the opposite has been reported on other types of bundles). Specific nucleators and elongators, like formins, generate different filament structures and may affect ADF/cofilin binding as they affect the binding of other ABPs, such as tropomyosins (Tpm3.1 and Tpm1.12 shown here), which in turn regulate ADF/cofilin binding. Posttranslational modifications, especially those targeting ADF/cofilin binding sites on subdomain2 (SD2) are likely to have an impact. The binding of ADF/cofilin is more cooperative on cytoplasmic actin than α-skeletal actin filaments. To see this figure in color, go online.

Differentiation of filaments through the noncovalent action of ABPs

Different actin filament networks are usually associated with a specific nucleator (or elongator), and each network seems to be specifically targeted by some regulatory proteins rather than others. Nucleators are often activated locally, downstream of specific signals. What determines the specificity for other ABPs that subsequently regulate these filaments remains an open question.

A key experiment was done by placing microbeads functionalized by nucleation promoting factor Las17 in yeast extracts, and showing that the resulting filament networks selectively recruited the corresponding ABPs (40). This result shows that nucleators can be enough to give filaments their identity. It also indicates that local gradients (e.g., ATP, cations, and pH) that may exist in cells do not seem to play a role in this particular process. Another key experiment focused on two formins, which elongate distinct filament networks in yeast, each network getting decorated by a specific form (acetylated and unacetylated) of yeast’s unique tropomyosin. The authors of this study forced these two formins to switch locations, and showed that tropomyosins switched locations accordingly, leading to the relocation of myosin motors in the cell (41). Formins can thus dictate which tropomyosin isoform will decorate a particular filament, and this specifies which ABPs can bind to this filament.

Tropomyosins appear to be of particular interest in this filament differentiation scheme. There are many different tropomyosin isoforms in fungi and metazoan cells, with marked differences in their subcellular localization (see (8) for a recent review). Different tropomyosins block or allow the binding of other ABPs, such as ADF/Cofilin (42, 43) (Fig. 2) and myosins (44). Tropomyosins thus appear as a central player in regulating the distribution of ABPs to specific filament assemblies in cells.

How do tropomyosins (or other ABPs) recognize filament populations based on how they were nucleated? An interesting hypothesis comes from the cryo-EM observation that different actin filament conformations are possible (13), suggesting that different nucleators could build filaments with different structures. Reinforcing this idea is the observation that formin mDia1 elongates filaments with a structure that differs from that of freely assembled filaments (45) that would therefore be recognized or even stabilized by tropomyosins (46).

Altogether, these results suggest that different nucleators will generate filaments with specific conformations, which will thus exhibit different affinities for regulatory proteins, which cooperate and compete, with tropomyosin playing a central role. This scheme, described in detail by Michelot and Drubin (7), can be tested further with single filament studies to directly measure the binding rates of proteins on filaments formed with different nucleators or elongators. These measurements could be coupled to cryo-EM observations to connect the different filament structures with specific nucleators and ABPs.

Actin isoforms

For convenience, in vitro studies often use α-skeletal actin (from skeletal muscle) and almost overlook that six actin isoforms exist in mammals, including the two cytoplasmic (or nonmuscle) β- and γ-actin isoforms. This is justified by the fact that actin is extremely conserved, and that its isoforms differ by <7% in sequence, with only four residues differing between β- and γ-actin. This strong similarity makes it both difficult to separate isoforms during purification, and reasonable to assume that relevant conclusions can be drawn by looking at the regulation of α-skeletal actin filaments by cytoplasmic proteins.

In cells, however, the two cytoplasmic isoforms have been reported to locate in specific patterns (see (47) for a brief review), indicating that actin networks preferentially use different isoforms. Their high sequence similarity makes these experiments challenging and selective antibodies need to be designed. A motivation to tackle the challenge of isoform purification for in vitro studies comes from the fact that actin isoforms exhibit different assembly dynamics (48) and that ABPs are able to detect the small differences between isoforms. Cofilin, for example, has a lower affinity for cytoplasmic actin monomers than for α-skeletal actin monomers (49) and binds to cytoplasmic filaments with a greater cooperativity than to α-skeletal actin filaments (50). Similarly, nonsarcomeric myosin motors display marked preferences for cytoplasmic actin, and myosin-7A ATPase activity is enhanced by γ-actin over β-actin (51).

Because yeasts, with only one actin gene, are able to grow differentiated actin networks, it is clear that other factors are essential (7). Nonetheless, actin isoforms may be a contributing factor in other cells. Their ability to copolymerize offers the possibility to form filaments with various isoform compositions (47).

Posttranslational modifications

Covalent alterations of amino acids are potentially an additional cellular mechanism for the spatiotemporal regulation of actin. So far, the effect of posttranslational modifications of actin on the regulation of filament assembly dynamics have been less characterized than noncovalent factors, and this appears to be an expanding research area (see (52) for a recent review).

Some modifications can be the work of pathogens (such as ADP-ribosylation) or be part of a global regulatory process, coupled to other differentiating factors. For instance, N-terminal arginylation of cytoplasmic β-actin regulates cell motility (53) and preventing arginylation results in a number of actin-related defects (52).

Posttranslational modification can change local interactions with ions, which are key in regulating a number of filament properties (39). In particular, a number of these modifications occur on subdomain 2 of actin, whose conformation changes are likely to affect the overall structure of the filament (13). These modifications are likely to change intrinsic filament properties as well as the affinity for ABPs. To our knowledge, such modified filaments have yet to be studied as specific substrates for ABPs, and we expect single filament experiments to contribute to such studies in the coming years.

For example, the redox enzyme Mical specifically oxidizes Met-44 and Met-47 of F-actin, resulting in unstable filaments (54). Recent cryo-EM data, showing the filament-stabilizing role of the actin D-loop, where these amino acids are located, provides a structural interpretation for this result (12). The disassembly dynamics of these oxidized filaments, as well as their modulation by ABPs, could be quantified by single filament experiments.

Mechanical stress and geometrical constraints

In cells, actin networks are exposed to a number of mechanical stresses that appear as another factor that could regulate the conformation and biochemical properties of actin filaments (55). Mechanical data in cells confirm this hypothesis, showing for example that filaments stretched by applying tension with a microcapillary have a higher affinity for myosin II motor domain (56).

Single filament experiments are particularly adapted to address this question; however, this is still an emerging field as these experiments are challenging, and there is only a handful of results so far (see (57) for a recent review). Local filament curvature, randomly applied by sticking filaments to a coverslip, has been shown to favor Arp2/3-mediated branching on the convex side of the filament (58). Tension applied to filaments, using optical and magnetic tweezers, has been shown to delay the binding of cofilin and the subsequent filament severing (59). These latter results still need to be quantitated extensively to test explanations of this mechanosensitivity based on structural changes of filaments (55). Such results are likely to come from single filament experiments in the near future.

Single filament experiments have already provided detailed data in situations where the tension applied to a filament is transmitted to ABPs interacting with this filament, thereby changing their conformations and molecular activity. Single filament microfluidics experiments have shown that applying tension to a filament accelerates its elongation by formin mDia1 (25) and this effect is in quantitative agreement with the formin translocation model derived from structural data (60). Magnetic tweezer experiments have shown that pulling on talin exposed its vinculin binding domains and could lead to the reinforcement of focal adhesions (61). Optical tweezer experiments have shown that the cadherin-catenin complex made more stable bonds with actin filaments that were put under tension, which may explain force transduction in cell-cell adhesions (62).

The application of mechanical stress to filaments in cells is often closely related to the geometry imposed by the organization of actin anchoring points and bundling proteins. The geometrical organization of filaments can also be a factor modulating the action of ABPs. Actin filament bundles, for example, provide a simple geometrical organization that can change how ABPs interact with the filaments. Myosin X recognizes this geometry and walks processively farther on a bundle than on individual filaments (63). The disassembling activity of ADF/cofilin has been reported to be hindered on filaments bundled by methylcellulose (24) but enhanced on fascin-bundled actin filaments (64). These experiments can be difficult to interpret because a number of factors may contribute, including competition between cofilin and fascin, which could also occur on single filaments, as well as specifically geometrical aspects such as filament-filament distance or the rigidity of their bonds.

Because geometrical organization often involves mechanically constraining filaments, the effect of geometry may be hard to separate from that of mechanical stress. Future single filament experiments; for example, applying patterning strategies (23) to smaller filament populations to control geometry, will certainly bring insights into these questions.

Coexistence and evolution of different flavors within one actin filament

The factors giving actin various flavors that we have listed are of a different nature (Fig. 2). One of the challenges for future studies is that these different factors do not act independently. For example, mechanical tension may not have the same effect on an ABP’s affinity for filaments made of different actin isoforms, or following specific posttranslational modifications.

In addition, the flavor of actin subunits in a filament evolves over time. The nucleotide states change irreversibly as F-actin hydrolyses ATP, whereas most of the other factors are reversible, including covalent posttranslational modifications. Mical-induced methionine oxidation, for example, can be reversed by another enzyme, SelR (65).

The structural polymorphism of actin filaments, deduced from cryo-EM observations (13), provides the structural basis to understand these different filament identities. Due to thermal fluctuations, a filament would be able to sample different conformations over time, or breathe—a term introduced some 20 years ago to describe the infrequent availability of some binding sites on actin filaments (66). The different factors giving a filament its flavor would shift the conformational distribution, favoring a specific conformation over the others.

Moreover, these cryo-EM images also report that blocks of different structure coexist within the same filament (13). This is consistent with the idea that a filament could simultaneously contain domains of different flavors. A well-documented example is the existence of ADP-Pi-rich and ADP-rich regions, which can lead to the decoration of the latter by cofilin and change its local filament conformation (22, 67, 68). Actin isoforms readily copolymerize (48, 51), and different filament regions could be made with different isoform contents. One can easily imagine how such domains could exist for other flavor-giving factors; for example, if a posttranslationally modified filament elongates further from unmodified actin monomers, or if mechanical stress is locally applied to a portion of a filament.

Addressing the question of multiple domains with different flavors coexisting within a single actin filament requires the ability to artificially generate hybrid filaments, with a controlled mixture of different flavors. This is precisely what single filament techniques allow one to achieve today. In particular, microfluidics allows one to generate filaments by sequentially using different conditions, and this was initially illustrated by building filaments with an artificial, purely ADP-Pi region (38).

Barbed end region: where does the end begin?

A region of particular interest to regulate actin assembly is the barbed end of the filament. The terminal subunit at the barbed end appears to be in an intermediate state, between that of G- and F-actin, as illustrated by its enhanced Pi-release rate compared to F-actin (38). It is likely that the few neighboring subunits also exhibit some conformational flexibility so that, in addition to the two terminal subunits we refer to as the barbed end, we may have to consider the barbed end region as a domain with specific properties (Fig. 3 A).

Figure 3.

Where does the barbed end begin? (A) Schematic representation of an actin filament, where we speculate that subunits near the barbed end have a greater freedom to change conformations. (B) A formin FH2 dimer tracking an elongating barbed end takes advantage of this conformational freedom to locally impose its preferred filament structure. (C) CP, in contrast, imposes a filament structure that is compatible with the standard F-actin structure. (D) When both CP and a formin coexist at the barbed end (29, 37), they compete through the conflicting conformations they each want to impose to the barbed end region. Eventually, the formin (or CP) will detach from the filament, or the formin will diffuse along the filament (29), away from the barbed end. To see this figure in color, go online.

The crystal structure of formin’s FH2 dimer in interaction with actin indicates that a formin preferably binds actin subunits in an arrangement that differs from the standard F-actin helical conformation, with a short helical twist of 180° instead of 167° (60). It thus seems that formin uses the conformational flexibility of the barbed end region to impose a local conformational change to the actin filament (Fig. 3 B). Consistent with this idea, filaments elongated by formin mDia1 were reported to be more flexible and exhibit a structural rearrangement spanning over tens of subunits (45).

CP, on the other hand, binds the filament barbed end with actin subunits in a conformation compatible with the standard F-actin helical structure (69). Barbed end capping by CP would thus suppress the conformational flexibility of the barbed end region (Fig. 3 C). In the frame of this hypothesis, when CP and a formin coexist at the barbed end, they not only compete for binding sites but also through the local conformation they each try to impose to the barbed end region (Fig. 3 D). Similarly, proteins binding to the filament side have been reported to alter barbed end assembly (70) even when they are bound several tens of subunits away from the barbed end (71).

More experiments along the lines of the single filament studies that have recently reported the existence of the formin-CP-complex (29, 37) will certainly shed more light on the specificities of the barbed end region.

Conclusions

Single filament techniques are a powerful, still expanding field. So far, these new tools have contributed to deciphering a number of mechanisms regulating actin assembly. In many ways, they have changed the way we look at actin filaments, and the amount of quantitative information we can now extract from these observations brings precious insight into molecular mechanisms, simultaneously involving several regulatory proteins.

Although we spontaneously marvel at the cell’s ability to generate and regulate diverse filament networks made from essentially one universal actin monomer, new challenges lay ahead as we now realize that various factors can give different flavors to actin, and modify its assembly properties. Our amazement is now just as strong, but somehow turned around: how can organized structures arise from such a rich ensemble of regulatory factors? How many different types of actin filaments can a cell build?

Fortunately, to tackle this awesome question, we now have at our disposal increasingly sophisticated and efficient single filament tools that allow us to fabricate, manipulate, and observe actin filaments with great accuracy. We believe they will be instrumental in the coming years, to generate quantitative data characterizing the role of actin-binding proteins in different contexts, and to refine our understanding of actin dynamics.

Editor: Brian Salzberg.