Cofilin-induced unidirectional cooperative conformational changes in actin filaments revealed by high-speed atomic force microscopy

1Actin filaments are involved in a variety of important functions in eukaryotic cells, including muscle contraction, amoeboid movement, cytokinesis, intracellular transport, and transcriptional regulation within the nucleus. These diverse functions depend on the interaction between actin and specific actin binding proteins (ABPs), and it is generally assumed that specific biochemical signaling is involved in the spatial and temporal regulation of each actin–ABP interaction. During migration of amoeboid cells, for instance, cofilin plays essential roles in continuous extension of lamellipodia by severing actin filaments to promote filament depolymerization or to initiate polymerization from new barbed ends (reviewed by ). Three independent biochemical mechanisms are known to inhibit cofilin activity; phosphorylation of Ser3, sequestration to phosphatidylinositol 4,5-bisphosphate (PIP2) in plasma membrane, and lower pH. These inhibitory mechanisms are implicated in the regulation of cofilin activity during cell migration, since experimental unleashing of inactive cofilin has been shown to initiate cofilin-dependent processes. Critically speaking, however, these data do demonstrate that active cofilin is required, but there is little experimental evidence that localized activation of cofilin is necessary for proper cofilin functions in cell migration. Indeed, overexpression of constitutively active S3A cofilin did not inhibit motility. Thus, biochemical signaling is not sufficient to explain how cofilin activities are properly regulated spatially and temporally in cells.

2On the other hand, many ABPs alter the atomic structure of actin subunits within filaments, and in certain cases these conformational changes are cooperative. For instance, the pioneering work of Oosawa and his colleagues demonstrated that the increase in fluorescence intensity of a fluorophore on actin saturates when only one molecule of the myosin motor domain is added for each 10 actin subunits within filaments. Similar myosin-induced cooperative conformational changes have been detected in various other assays. In addition, dense binding of cofilin shortens the helical pitch of actin filaments by 25%, and time-resolved phosphorescence anisotropy and differential scanning calorimetry showed that one molecule of bound cofilin changes the structure of ∼100 actin subunits within a filament. Moreover, binding of cofilin to actin filaments is cooperative. This implies that cooperative conformational changes induced by an ABP are propagated to neighboring actin subunits, increasing their affinity for that, or another, ABP. This could provide a novel mechanism by which actin filaments change their function by regulating their affinities for various ABPs.

3However, the currently available information on structural changes to actin filaments is limited to high-resolution static images (electron microscopy), low-resolution dynamic changes (fluorescence microscopy), and bulk biochemical analyses. There is little information available in the molecular mechanism that mediates the propagation of structural changes to neighboring actin subunits. Atomic force microscope (AFM) is unique in that it enables detailed structural analysis of wet protein samples, and recent dramatic improvements in scanning speed now enables real time imaging of conformational changes in protein samples with high-spatial resolution. This high-speed AFM (HS-AFM) has been used to visualize molecular movements such as the stepping motion of myosin V along actin filaments, rotary catalysis of F1 ATPase without a rotor, and light-induced conformational changes in bacteriorhodopsin. Here, we used HS-AFM to directly visualize conformational changes in actin filaments induced by cofilin binding. We found that conformational changes within cofilin clusters unidirectionally propagate to the neighboring bare actin zone in a cooperative manner and that the growth of the cofilin cluster follows this unidirectional cooperative conformational change.

4For AFM, actin filaments must be immobilized on the stage, yet they must have the freedom of movement to bind cofilin and exhibit the resultant changes in the helical twist that accompany cofilin binding. We therefore formed a bilayer of positively charged lipid on the surface of freshly peeled mica fixed to the observation stage. A solution of actin filaments was then placed on the supported lipid bilayer and HS-AFM was performed. Right-handed double helical filaments were clearly visualized (Figure 1), as in earlier reports.

5We approximated the positions of the crossover points, where the two strands of actin filament are aligned vertically, using the position of the highest point (peak) in each half helix identified in the AFM images. Quantitative analysis indicated that the height of those peaks, or the thickness of the filaments, was 8.6 ± 0.8 nm (average ± SD), and the spacing between the peaks, or half helical pitch, was 36.8 ± 4.3 nm (Figure 1D,E). These values, which are consistent with previously reported structural data, will hereafter be referred to as normal height and normal half helical pitch, respectively. The distribution of half helical pitches ranged from ∼26 nm to ∼45 nm and was well fit by a normal distribution with a standard deviation of 4.3 nm. This distribution should reflect natural variation in half helical pitch, as was proposed on the basis of a similar distribution of half helical pitches in electron micrographs of negatively stained specimens and also measurement errors. To assess the overall magnitude of the measurement errors, we prepared paracrystals of actin filaments in the presence of 30 mM MgCl2 (Figure 1C) (Hanson, 1973). The distribution of the half helical pitches of the paracrystals measured under the same conditions was narrower than that of free filaments, with a mean of 36.5 nm and standard deviation of 3.0 nm (Figure 1F). In a hypothetical case where the paracrystals do not undergo spontaneous conformational changes, this distribution is the upper limit of the measurement errors in our system. Furthermore, assuming normal distributions of the true half helical pitches and measurement errors, the true standard deviation of the half helical pitches of control filaments was calculated to be 3.1 nm or larger, by subtracting the variance of the actin paracrystals from that of control filaments. This indicated that the structure of the actin filaments does indeed vary with changes in helical twist.

6When 75 nM cofilin was added to actin filaments immobilized on the lipid surface, cofilin gradually bound to the actin filaments, and sections of filaments bound with cofilin molecules were easily identified, as the bound filaments appeared thicker and the peaks were taller (Figure 1B). Peak heights within long cofilin clusters (longer than eight consecutive half helices) showed a single distribution of 10.6 nm ± 1.0 nm, approximately 2 nm taller than the control filaments (Figure 1D). This is consistent with electron microscopic analysis, which showed that filaments fully decorated with cofilin are ∼2.3-nm thicker than control filaments due to the presence of cofilin molecules (Figure 1—figure supplement 1). Moreover, the height distribution was fit well by a single normal distribution, which suggests that within those long cofilin clusters, both actin strands are homogeneously bound with cofilin molecules.

7Cofilin binding shortened the half helical pitch by 27% to 26.9 nm (Figure 1E), which is again consistent with earlier electron microscopic and AFM analyses. The standard deviation of the distribution of half helical pitches of the fully cofilin-decorated filaments was 3.8 nm, and subtracting the maximum possible measurement error yielded a standard deviation of 2.3 nm or larger. This value is smaller than that of the control filaments (3.1 nm) and suggests that cofilin binding restricts torsional movements of the filaments within cofilin clusters.

8We next analyzed shorter cofilin clusters along otherwise apparently bare filaments (Figure 2) so that both ends of the cluster were visible within the imaged area. Intriguingly, we noticed that the half helical pitch in the apparently bare section on one side of the cofilin cluster was short, while the pitch on the other side of the cluster was nearly normal.

9The observation buffer was then changed to F buffer containing 0.1 mM ATP, 1 mM ADP, hexokinase, glucose, 20 nM subfragment-1 (S1) of skeletal myosin II, and 75–100 nM cofilin, so that the polarity of the actin filaments could be identified from the characteristic tilted binding angle seen when S1 transiently binds the filament. This analysis revealed that the half helical pitch of the bare zone on the pointed-end side of the cofilin cluster was short, while that on the barbed-end side was slightly longer than the normal pitch (Figure 2 and Table 1). Student's t-test indicated that the differences in the mean half helical pitches between the first neighbor bare zone on either side of the cluster and the control filaments are statistically significant (p < 0.00001 and p < 0.001 for pointed-end and barbed-end side, respectively). Pitches of the neighboring half helices second from the cofilin clusters were nearly normal (Table 1).

10The results summarized above demonstrate that cofilin clusters induce distinctively asymmetric conformational changes in bare zones immediately neighboring the clusters, but it is still uncertain whether a stretch of many bound cofilin molecules, as in clusters, is necessary to induce such conformational changes. To answer that question, we needed to visualize individual cofilin molecules bound to actin filaments, and measure the half helical pitch of the filament around those bound molecules. This has not been possible because cofilin molecules are too small to image individually using electron microscopy or AFM. We therefore engineered a fusion protein in which the N-terminal half of the rod domain of α-actinin was attached to the C-terminus of cofilin. In electron micrographs of negatively stained specimens (Figure 3A), as well as in HS-AFM images (Video 1 and Figure 3—figure supplement 4), we were able to see clusters of cofilin-rod with a half helical pitch ∼25% shorter than the bare zones, which is similar to control cofilin molecules bound to actin filaments (Figure 3C).

11Rod-like structures were often observed to stick out from apparently bare sections of the filaments (Figure 3B). These are most likely the rod portions of the cofilin-rod molecules bound sparsely along seemingly bare sections of the filaments, because their length was ∼10 nm, which is expected for half an α-actinin rod, and because similar rod-like structures were rarely observed in control images of cofilin added to actin filaments (Figure 3C). We therefore estimated the helical pitches on both sides of single rod-like structures sticking out from the filaments by measuring the filament length that encompassed three actin subunits along one strand on either side of the bound molecule. In addition, because we did not know the polarity of each filament, we compared the helical pitches on both sides by dividing the longer helical pitch by the shorter one. This yielded a value of 1.02 ± 0.02 (average ± SD, n = 9), which was much smaller than the 1.30 ± 0.18 (n = 4) measured for cofilin clusters in electron micrographs or 1.37 calculated from the AFM data (Figure 1E) and was comparable to the value of randomly selected regions along control actin filaments (1.04 ± 0.05, n = 18). Furthermore, the helical pitches around the apparently singly bound cofilin-rod molecules (36.0 ± 1.0 nm, n = 5) did not significantly differ from the control helical pitch (36.3 ± 1.0 nm, n = 8) measured in electron micrographs.

12In HS-AFM observations, we were able to observe four cases of an apparently single cofilin-rod molecule binding transiently to an actin filament (Figure 3D and Video 2). Consistent with the electron microscopic observation described above, we were unable to detect significant changes in peak heights and half helical pitches, such as those observed around cofilin clusters, around those singly bound cofilin-rod molecules (Figure 3E,F).

13These results suggest that a single cofilin molecule cannot induce conformational changes, either symmetric or asymmetric, that involve detectable changes in helical pitch when it binds to bare sections of the filament.

14Real time AFM imaging taken at 1.5 or 2 frames/s enabled us to follow the growth of individual cofilin clusters along actin filaments. In the case illustrated in Figure 4A and Video 3, observed in F buffer containing 1 mM ADP and 0.1 mM ATP, P4 was already tall when imaging started (∼10.5 nm, average of t = 0–5 s), while P3 had an intermediate height (∼9.5 nm), and P1 and P2 were normal (∼8.5 nm) (Figure 4A, middle). We interpreted this to mean that the initial cofilin cluster extended beyond P4 but ended near P3. P3 gradually became taller and plateaued beyond ∼25 s. Thereafter, P2 started to rise at ∼15 s and plateaued at ∼63 s, whereas P1 started to rise at ∼60 s. The distance between P2 and P3 rapidly shortened to ∼27 nm between 0 and 15 s, before P2 started to rise, while the distance between P1 and P2 shortened between 30 and 40 s, before P1 started to rise (Figure 4A, bottom). This sequence of events implies gradual growth of this cofilin cluster into the neighboring bare zone on the pointed-end side where the helical pitch was shortened.

15Cofilin binds preferentially to actin subunits carrying ADP, compared to those carrying ATP or ADP and Pi. Since actin filaments in F buffer containing ATP should have ATP–actin caps near the barbed ends, and since subunits on the pointed-end side tend to carry ADP only due to ATP hydrolysis and Pi release, the directional growth of cofilin clusters toward the pointed-end might reflect the asymmetric distribution of actin subunits with different nucleotides. To test this possibility, we prepared two different types of actin filaments with homogenous nucleotide states along the lengths. In the first case, filaments polymerized in buffer containing 1 mM ATP were incubated for 30 min in buffer containing 1 mM ADP, which is much longer than the 350 s required to hydrolyze ATP and release the resultant Pi from half of the polymerizing ATP–actin molecules, so that most of the actin subunits should carry ADP only. In the second case, filaments polymerized in buffer containing 1 mM ATP were incubated for 10 min in buffer containing 1 mM ADP and 10 mM Pi. Under this condition, most of the actin subunits should carry ADP and Pi, considering a Kd of 1.5 mM for Pi. HS-AFM observations after the addition of 20 nM S1 and 75 nM cofilin or 150 nM S1 and 900 nM cofilin demonstrated that, in both cases, the growth of cofilin clusters was primarily to the pointed-end direction (Figure 4B and Video 4, and Figure 4C and Video 5). Results of a large number of observations are compiled in Figure 5.

16The rate of formation and growth of cofilin clusters depended on the nucleotide state of actin subunits as well as the concentration of cofilin. In most experiments we used 75 nM cofilin since we were able to observe de novo formation of clusters and tractable growth, after mixing and settling of the system. When 1 µM cofilin was added and mixed, we were unable to find bare zones of actin filaments. When actin filaments were incubated with 1 mM ADP and 10 mM Pi, much higher concentrations of cofilin (e.g., 900 nM) was needed to induce cofilin clusters. The growth rates of cofilin clusters were within tractable range even in the presence of 1 µM cofilin, when 10 mM Pi was present (Figure 4C). These results are consistent with the weaker affinity of cofilin for actin subunits carrying ADP and Pi.

17Cofilin is an actin filament severing protein (reviewed by ), and although our experiments were performed at pH 6.8, which should suppress severing activity, we observed frequent severing of the filaments near or within short cofilin clusters (Figure 6A–C, Videos 6–10 and Figure 6—figure supplements 1 and 2). Because a previous study by found that severing activity is highest at 10 nM of human cofilin and that this activity sharply declines above and below 10 nM, we observed severing events at several different cofilin concentrations, including 10 nM. In the presence of 10 or 20 nM cofilin, severing was very infrequent (Video 11 and Figure 6—figure supplement 3). Severing was often observed between 40 (Video 12 and Figure 6—figure supplement 4) and 200 nM cofilin. In the presence of 650 nM cofilin, many of the filaments were fully decorated with cofilin when observations started, and severing was observed exclusively near the ends of the long clusters (Video 12 and Figure 6—figure supplement 4). When the cofilin concentration was increased to 1 µM, all the filaments were fully decorated with cofilin, and severing was infrequent along those fully decorated filaments (Video 12). Thus, we also observed that cofilin's severing activity was highest in the medium concentration range (i.e., 40–200 nM), but this concentration is at least several-fold higher than that reported by. These two studies used the same pair of proteins (i.e., skeletal muscle actin and human cofilin), and we can only speculate that this discrepancy is due to differences in experimental conditions.

18We next quantitatively analyzed the severing positions of the filaments relative to the short cofilin clusters in the presence of 40–200 nM cofilin (Figure 6D). Approximately 60% of severing events occurred inside the cofilin clusters, mostly within half helices neighboring bare zones. Severing also occurred in half helices in bare zones neighboring a cofilin cluster, which accounted for approximately 40% of severing events. Overall, ∼80% of the severing events occurred within one-half of a helix on either side of the boundary between a bare zone and a cofilin cluster. Severing in bare zones far from cofilin clusters was very rare.

19In the presence of 40 nM cofilin, the lowest concentration we used for quantitative severing assay, we observed a total of 22 cases of severing events in 31 actin filaments. Among those, 18 cases occurred in half helices immediately neighboring the boundary between a bare zone and a cofilin cluster, even though cluster formation was relatively rare at this concentration of cofilin. Three cases occurred in ‘far’ bare zones more than half a helix away from the boundary, and one occurred in an ‘inner’ cofilin cluster more than half a helix away from the boundary (Video 11).

20Based on these observations, we conclude that severing under the present experimental condition preferentially occurs near the boundary between a bare zone and a cofilin cluster, but not necessarily at the boundary. Considering the difference in spatial resolution, this conclusion is also consistent with that made by, who used fluorescence microscopy and concluded that severing occurs at the boundary between cofilin clusters and bare zones.

21Using HS-AFM to visualize wet samples with high-spatial and -temporal resolution, we were able to create an image of conformational changes in actin filaments induced by cofilin binding, including asymmetric changes in the helical twist on either side of a bound cofilin cluster, growth of the clusters toward the pointed-end, and the severing of filaments near the ends of cofilin clusters.

22The short helical pitch in the cofilin clusters was propagated to the immediately neighboring bare zone on the pointed-end side of the cluster. Since our current analysis method only measures heights and distances between peaks, we were unable to examine if the supertwisted structure returns to a normal state abruptly or gradually. Also, we were unable to determine the positions of the ends of cofilin clusters precisely above the spatial resolution of half helices, making it difficult to estimate accurately how far the supertwisting conformational changes propagate into neighboring bare zones. Nonetheless, the fact that the pitches of the neighboring half helices second from the cofilin clusters were nearly normal suggests that the effect does not propagate longer than one-half of a helix.

23The neighboring bare zone on the barbed-end side of the cluster had slightly longer helical pitch than the control. Although a t-test indicated that the untwisting was statistically significant, we are not certain if it was an active process. If interactions between actin filaments and the lipid surface produced resistive force against local rotation of the filament, supertwisting in the cluster would induce passive, compensating untwisting of the nearby helix. If that is the case, the fact that actin filaments are not necessarily free to rotate in vivo suggests similar passive untwisting may occur in vivo as well. In any case, it is noteworthy that cofilin clusters induced distinctly asymmetric cooperative conformational changes in neighboring bare zones. Propagation of the supertwisted conformation from cofilin clusters to neighbor bare zones had been suggested by image analysis of electron micrographs of cofilin–actin complexes.

24Our observation that the cofilin clusters grew in the pointed-end direction supports the conclusion of that cofilin preferentially binds to the supertwisted segments of actin subunits. This tendency of cofilin clusters to grow unidirectionally was unaffected when most of the filament subunits carried ADP or ADP+Pi, demonstrating that the directional growth of cofilin clusters does not depend on a gradient of different nucleotide states on actin subunits within each filament. Instead, it presumably depends on asymmetric cofilin-induced conformational changes in the actin filaments, which is an intrinsic property of the polar structure of the actin filaments.

25The rise of peaks and shortening of half helical pitches that accompanied the growth of cofilin clusters were sometimes rapid, but at other times slow and gradual (Figure 4). If we assume that the rise was due to simple addition of cofilin molecules at the two binding sites closest to the crossover point, and if the cofilin clusters grow in perfect synchrony along two strands of the double helix, then the crossover points should rise abruptly in one step. If, on the other hand, there is a time-lag between cofilin bindings at the two critical binding sites, due to delayed growth along one strand, for example, the rise would occur in two smaller discrete substeps. Indeed, in many cases, the rise of peaks appeared to occur in two discrete substeps (Figure 4), supporting this possibility. In other cases, however, the peaks rose gradually over the course of 40 s (e.g., P2 in Figure 4A), although we have no mechanistic explanation for those events.

26In certain cases the next peak (P1 in the case shown in Figure 4A) only rose when the previous peak (P2) had reached a plateau, whereas in other cases the new peak (P2) started to rise while the previous one (P3) was still rising. This latter case is more consistent with the possibility that cofilin clusters do not necessarily grow in concert along the two strands, and the delay may sometimes exceed one-half of a helix.

27Another uncertainty is whether the new cofilin molecule always binds to the vacant binding site immediately neighboring the cluster. If the unidirectional cluster growth to the pointed-end is driven by the shortened helical pitch on the pointed-end side of the cluster, the new molecule is likely to bind any of the vacant binding sites in the supertwisted, apparently, bare zone on the pointed-end side of the cluster. This view is consistent with the result of recent high-resolution single molecule fluorescence microscopic assays that showed that free cofilin molecules prefer to bind within 65 nm of an already bound cofilin molecule, but not necessarily to the site immediately neighboring the already bound molecule. We thus speculate that there are vacant binding sites behind the advancing front of cofilin clusters, which are eventually filled by other cofilin molecules to form tight clusters, and that the advances are not necessarily in concert between the two strands of the double helix.

28In the presence of cofilin, actin filaments are frequently severed at or near the boundary between cofilin clusters and bare zones, at least when examined at the spatial resolution of fluorescence microscopy. That observation is consistent with the idea that severing preferentially occurs at sites of structural discontinuity, such as the boundary between supertwisted and normal helical pitches. An alternative view is that cofilin binding weakens longitudinal contacts between actin subunits within cofilin clusters and also in neighboring bare zones, but because bound cofilin bridges two actin subunits, strengthening interactions between them, severing occurs in nearby supertwisted bare zones. Roughly 40% of the severing events in our AFM observations occurred at the boundary between a cluster and a bare zone or within the neighboring bare zone, which may be explained by these hypotheses. However, more than half of the severing events occurred within the cofilin clusters, albeit close to the cluster ends. One plausible explanation is that the scattered unoccupied cofilin binding sites near the ends of the cluster, discussed above, may be easy to break due to the lack of bridging cofilin molecules. Alternatively, if growth of cofilin clusters on one of the two filament strands lags the growth on the other, conformational stress may develop between the two strands, leading to breaks in the filament near the ends of cofilin clusters.

29Previous biochemical and simulation studies suggested that one or a few bound cofilin molecules are sufficient to sever actin filaments. However, our AFM observations are more consistent with the view that efficient severing requires a cofilin cluster longer than one-half of a helix, since most of the severing events were observed within or very near cofilin clusters that were recognizable in our AFM images. The need of contiguously bound cofilin molecules for efficient severing is consistent with our earlier mutant analysis. Within filaments in solution, D11Q mutant actin subunits rapidly exchange bound ADP for ATP in solution. Consequently, most subunits within D11Q actin filaments have ATP bound, reducing their affinity for cofilin and protecting them from severing. Interestingly, cofilin was able to bind to copolymers of D11Q and wild-type actins, but severing was inefficient, suggesting that contiguous clusters of cofilin-bound actin subunits are necessary for efficient severing activity.

30A related question is how many bound cofilin molecules are required to induce cooperative supertwisting conformational changes in actin filaments? Answering this question is technically difficult when the imaging power is not sufficient to see individual bound cofilin molecules, but our observations using a cofilin-rod fusion protein showed that one is not enough. Consistent with this view, previous simulation studies suggested that two cofilin molecules bound close to one another along an actin filament serve as a nucleus to initiate cooperative binding of cofilin to form a cluster. Future high-resolution HS-AFM studies using cofilin fused with a rod or some other structural marker to make it visible with AFM will directly test those hypotheses.

31Even if a singly bound cofilin molecule cannot induce supertwisting of the helix or initiate cluster growth, it does not exclude the possibility that single bound cofilin molecules induce subtler, perhaps longer range, cooperative conformational changes, such as those indirectly detected through biophysical measurements. Apparently cofilin is able to induce at least two distinct types of cooperative conformational changes: one that requires cluster formation, involves ∼25% supertwisting of the helix, and propagates over half a helix toward the pointed-end of the filament; and a second type that singly bound cofilin molecules can induce which involves relatively subtle conformational changes, and propagates much longer. Simulation studies by suggest that there are two distinct states of cofilin–actin complexes, which may be correlated with the two types of cooperative conformational changes we propose here.

32Cooperativity in the binding of cofilin to actin filaments could have multiple physiological implications. Generally speaking, cooperativity would amplify small changes in the input (concentration of active cofilin) to a larger difference in output (cluster formation and severing). In addition, the cellular concentration of cofilin is lower than that of polymerized actin, and cooperativity would be a useful means of disrupting selected filaments under those conditions.

33Second, we propose that the propagation of cofilin-induced conformational changes into neighboring cofilin-unbound zones of actin filaments would give cofilin an advantage in competition with other ABPs, once a small cofilin cluster is established as a foothold. For example, cofilin is implicated in severing and depolymerization of aged actin filaments in lamellipodia. However, those actin filaments are often bound with tropomyosin, which inhibits binding of cofilin and protects the filaments from cofilin's severing and depolymerizing activities. If cofilin forms a small cluster at a vacant site on an actin filament that is otherwise decorated and protected by tropomyosin, the cluster would induce, or apply conformational stress to induce, supertwisting on the neighboring tropomyosin-bound segment on the pointed-end side. This would accelerate dissociation of tropomyosin, resulting in faster growth of the cofilin cluster than when cluster growth needed to wait for spontaneous dissociation of neighboring tropomyosin molecules. Furthermore, that cluster growth is directed only to the pointed-end may be beneficial in selective disassembly of aged actin filaments.

34Within cells, formin remains bound to the barbed-end when it catalyzes filament elongation. This causes relative rotation between the formin molecule and the filament. Thus, if the formin molecule and the filament are not free to rotate, the filament will be untwisted, and if cofilin has a lower affinity for untwisted actin filaments, rapidly polymerizing actin filaments in a formin-dependent manner will be protected from severing by cofilin. It should be noted, however, that formin bound to the barbed-end of an actin filament also allosterically changes the structure of the filament, which may interfere with the interactions of the filament with cofilin, independent of mechanically forced untwisting of the filament helix.

35discovered that the untwisted conformation of actin filaments induced by a drebrin N-terminal fragment propagates to neighboring bare zones; however, their findings seem to indicate that this propagation is in both directions from the drebrin clusters, though the authors did not address that point. It thus appears that there are multiple forms of ABP-induced cooperative conformational changes to actin filaments that propagate into bare zones, which implies there are also multiple physiological functions for such cooperative conformational changes to actin filaments.

36cDNA encoding human cofilin 1 was amplified from a human cDNA library using PCR with primers 5′-ggtaccatggcctccggtgt and 5′-tctagacaaaggcttgccctcca. After confirmation of its sequence, the amplified DNA fragment was subcloned into pColdI expression vector (Takara Bio, Otsu, Japan) at the XbaI and KpnI sites. The pColdI vector had been modified to contain a TEV cleavage site between the His tag and the multi-cloning sites, so that the amino acid sequence near the N-terminus was MNHKVHHHHHHIEGRHMENLYFQGTMASGVAVS… (italics indicate the TEV cleavage site).

37The cofilin-rod fusion gene was constructed by cloning the cDNA encoding the first half of the Dictyostelium α-actinin rod downstream of the cofilin gene in pColdITEV. The amino acid sequence at the junction of the two proteins was …SAVISLEGKPLEQTKSDYLKRA…, and the C terminal sequence was …QKIEDSLVSR (italics show extra amino acid residues derived from recognition sites for restriction enzymes). The first half of the Dictyostelium α-actinin rod, corresponding to amino acid residues 265–505 of the parent molecule, forms a monomeric12-nm long rod-like structure and has been used as an artificial lever arm of myosin motors.

38The proteins were expressed in Escherichia coli, purified using Ni-NTA affinity chromatography, and dialyzed against a buffer containing 10 mM HEPES, pH 7.4, 50 mM KCl, 0.1 mM DTT, and 0.01% NaN3 overnight at 4°C. After concentrating with a centrifugal concentrator (Amicon Ultra 4), aliquots were snap-frozen in liquid nitrogen and stored at −80°C. Unless otherwise stated, the experiments used those His-tagged cofilin or cofilin-rod proteins. However, we repeated some key experiments after removing the His-tag by treatments with TEV protease and obtained qualitatively similar results. These experiments included asymmetric conformational changes of actin filaments on either side of a cofilin cluster, unidirectional growth of cofilin clusters in the pointed-end direction, and frequent severing of filaments near the boundary between a bare zone and a cofilin cluster. Cofilin with and without His-tag bound to actin filaments at a 1:1 molar ratio with a similar affinity (Figure 3—figure supplements 1 and 2). Cofilin-rod with and without His-tag also did not sediment on its own, then bound to actin filaments with affinities similar to cofilin without the rod fusion (Figure 3—figure supplement 3).

39Rabbit skeletal muscle actin and chymotryptic subfragment-1, S1, were purified as described previously and stored in liquid nitrogen. Some experiments used G-actin that was further purified by gel filtration column chromatography, yielding identical results.

40Before use, an aliquot of frozen stock was thawed for 1 hr on ice and clarified by ultracentrifugation at 80,000 rpm for 5 min at 5°C. Protein concentration was then measured using an Advanced Protein Assay (Cytoskeleton, Denver, CO), using calibrated skeletal actin as the standard.

41We used a laboratory built high-speed atomic force microscope (HS-AFM) as described previously. HS-AFM imaging was carried out in the tapping mode with small cantilevers (BL-AC10DS-A2, Olympus, Tokyo, Japan) whose spring constant, resonant frequency in water, and quality factor in water were ∼0.10 N/m, ∼400 kHz, and ∼2, respectively. The probe tip was grown on the original tip end of a cantilever through electron beam deposition and was further sharpened using a radio frequency plasma etcher (PE-2000, South Bay Technology, Redondo Beach, CA) under an argon gas atmosphere (typically at 180 mTorr and 15 W for 3 min). During HS-AFM imaging, the free-oscillation peak-to-peak amplitude of the cantilever (A0) was set to ∼2 nm, and the feedback amplitude set point was set at more than 0.9A0. Details of the method for HS-AFM imaging are described elsewhere.

42We prepared small unilamellar vesicles (SUVs) and mica-supported lipid bilayer as described previously. The typical lipid composition was 1,2-dipamitoyl-sn-glycero-3-phosphocholine (DPPC) and 1,2-dipalmitoyl-3-trimethylammonium-propane (DPTAP) at a weight ratio of 9:1. The lipids were purchased from Avanti Polar Lipids (Alabaster, AL). SUVs were dispersed in Milli-Q water at 2 mg/ml and stocked at −20°C. Before use, the SUVs were diluted in 5 mM MgCl2 to 0.5 mg/ml and sonicated with a bath sonicator (AUC-06L, AS ONE, Osaka, Japan) for 1 min. An aliquot of the sonicated SUVs was deposited on the surface of freshly cleaved mica, which had been glued onto a sample stage beforehand, and incubated for more than 3 hr at room temperature (24–26°C) in a humid sealed container to avoid surface drying. Up to 10 sample stages were prepared simultaneously and stored in the sealed container.

43Before deposition of actin filaments, the surface of the sample stage was rinsed with a large amount of Milli-Q water (∼20 µl × five times) to remove excess SUVs and lipid bilayers. Actin filaments were then deposited onto the lipid bilayer using one of the following methods.

44G-actin (5–10 µM) was polymerized in F buffer (40 mM KCl, 20 mM PIPES–KOH, pH 6.8, 1 mM MgCl2, 0.5 mM EGTA, 0.5 mM DTT) containing 1 mM ATP for 30 min on ice. The resultant actin filaments were diluted to 0.5–1.0 μM in F buffer containing 1 mM ATP. Water on the lipid bilayer on a sample stage was replaced with 1–2 µl of F buffer containing 1 mM ATP, to which 2 µl of the diluted actin solution was added. After 5–10 min of incubation at room temperature, unattached actin filaments were removed by exchanging the solution with 50 µl of one of the four different F buffer-based observation buffers each containing (i) 1 mM ATP, (ii) 0.1 mM ATP and 1 mM ADP, (iii) 1 mM ADP, 5 U/ml hexokinase, and 10 mM glucose, and (iv) 1 mM ADP and 10 mM Pi. In the case (iii), incubation was continued for 30 min to ensure that all actin subunits within the filaments carried ADP.

45Alternatively, G-actin (20 µM) was polymerized in F buffer containing 1 mM ATP and 30 mM MgCl2 for 1 hr at room temperature. After introduction of this solution to the sample stage and 10 min of incubation at room temperature, unattached actin filaments were removed by gently exchanging the solution with F buffer containing 1 mM ATP and 30 mM MgCl2.

46Finally, the sample stage was mounted to the z-scanner of a HS-AFM apparatus and immersed in a liquid cell containing the same observation buffer used in the last step, and HS-AFM imaging was performed.

47To follow binding of cofilin to actin filaments, 6 µl of cofilin diluted in the observation buffer was injected into the observation cell during AFM imaging. In some experiments, S1 was added in the observation buffer in the concentration of 20 nM (in cases [i], [ii], and [iii]) or 150 nM (in case [iv]) to identify the polarity of actin filaments.

48HS-AFM images were viewed and analyzed using the laboratory built software, Kodec4.4.7.39. In brief, a low-pass filter to remove spike noise and a flattening filter to make the xy-plane flat were applied to individual images. The position and height of the peak within each half helix were determined semi-automatically using the following steps. First, the most probable highest point near a crossover point was selected manually. Second, the actual highest point was automatically determined by searching a 5 × 5 pixel area (typically 7.5 × 7.5 nm2) around the selected point. Third, the peak position was refined based on a center of mass calculation using information on the heights and positions within the 5 × 5 pixel area around the selected point, after which the refined peak position and height were used to represent the peak of the half helix.

49The Kodec4.4.7.39 for HS-AFM image viewing and analysis software is coded in Visual C# (Visual Studio 2010, Microsoft, USA) and is available as Source code 1. All filters and subroutines for image analysis used in the present study are included in the software. We confirmed the compatibility between the software and computers operated with Windows 7 or 8. Installer of the software, Kodec4_Setup.msi, is available in the subfolder of ‘Kodec 4.4.7.39\Setup\Release’. This software should be cited as: Sakashita M, M Imai, N Kodera, D Maruyama, H Watanabe, Y Moriguchi, and T Ando. 2013. Kodec4.4.7.39.

50Actin filaments were prepared by polymerization of G-actin (20 µM) in F buffer containing 40 mM KCl, 20 mM PIPES, pH 6.8, 1 mM MgCl2, 0.5 mM EGTA, 0.5 mM DTT, and 1 mM ATP for 30 min on ice. Cofilin or cofilin-rod was added to 1 µM actin filaments in 50 µl of F buffer at a molar ratio of 1:2 or 1:1, and the solutions were mixed by gentle pipetting, after which the mixture was incubated for 3–5 min at room temperature. This mixture was then added to F buffer containing sodium phosphate (pH 6.8) so that the concentration of actin was 0.5–1 µM and that of Pi was 5 mM, and a drop of this solution was immediately placed on a copper grid. The samples were fixed and negatively stained using a solution containing 1% uranyl acetate and 20 µg/ml bacitracin. The fixed samples were dried under an incandescent lamp to form films over the holes of the grid. Electron microscopic data were then acquired using a Hitachi H-7650 transmission electron microscope.