Molecular viewing of actin polymerizing actions and beyond: Combination analysis of single-molecule speckle microscopy with modeling, FRAP and s-FDAP (sequential fluorescence decay after photoactivation)


Author to whom all correspondence should be addressed



Live-cell single-molecule imaging is a powerful tool to elucidate the in vivo biochemistry of cytoskeletal proteins. However, it is often somewhat difficult to interpret how a bulk population of the observed molecule might behave as a whole. We review our recent studies in which the combination of image analysis with modeling and bulk kinetics measurements such as FRAP (fluorescence recovery after photobleaching) clarified basic problems in the regulation of actin remodeling pathways.


Development of both vertebrates and invertebrates is achieved by continuous protrusive and contractile movement of cells where the actin cytoskeleton plays essential roles. The actin cytoskeleton is the major component beneath the cell cortex. Actin is highly conserved and one of the most abundant proteins in eukaryotic cells (Dominguez & Holmes 2011). It exists in two states; monomeric G-actin (globular actin) and F-actin (filamentous actin) which forms a polymer arranged in a long-pitch double helix (Holmes et al. 1990). F-actin provides cells with mechanical support, contractile force with myosin motor proteins and myosin-independent force by actin polymerization. Unlike the static picture implied by the name “skeleton”, the actin cytoskeleton dynamically moves and exchanges its components often on the order of a second (Watanabe 2010). There are numerous actin regulatory proteins which bind either G-actin or F-actin to rebuild various actin-based structures on demand. Actin forms unique structures in differentiated cells, such as myofibrils of striated muscle cells, dendritic spines of synapses and actin bundles in stereocilia of auditory cells.

To elucidate the regulation of cell morphogenesis, real-time monitoring of molecular behavior in the actin cytoskeleton offers tremendous potential. Under the conventional 2D culture environment, fluorescence single-molecule speckle (SiMS) microscopy provides a powerful means for high-resolution analysis of assembly/disassembly and movement of the actin network (Watanabe & Mitchison 2002; Watanabe 2012). Fluorescent speckle microscopy (FSM) was first reported in 1998, as a method to visualize movement of cytoskeletal polymers especially along the axis, which is enabled by low-density fluorochrome labeling (Waterman-Storer & Salmon 1998). Our subsequent work has further shown that at the ultralow label density, individual cytoskeleton-associated fluorescent probes can be visualized in living cells (Watanabe & Mitchison 2002).

Curiously enough, while SiMS microscopy directly visualizes assembly and disassembly of actin and its associated proteins, it has been argued that the SiMS method does not yield consistent results with other conventional imaging techniques such as FRAP (fluorescence recovery after photobleaching), especially in measuring actin filament turnover near the cell leading edge (Lai et al. 2008). This argument prompted us to revisit several unsolved issues in the previous SiMS analyses.

In this review, we review our recent studies which have elucidated cellular actin turnover mechanisms through the comparison of the SiMS method with other techniques that measure “bulk” kinetics such as dual-color FRAP (an analogous method to FLAP, fluorescence localization after photobleaching; Zicha et al. 2003) and s-FDAP (sequential fluorescence decay after photoactivation) (Kiuchi et al. 2011b). Combination of SiMS analysis with other approaches has elucidated actin turnover mechanisms more vividly than SiMS data alone. Currently, SiMS microscopy is not recommended for most developmental biologists who need to analyze dynamics in thick three dimensional specimens. Nonetheless, our review illustrates several potential problems that could give rise to apparently contradictory interpretations between direct molecular behavior viewing and bulk structural remodeling observation, which one should bear in mind when interpreting data obtained by bulk kinetics observation of the molecule of interest.

The principle of SiMS microscopy

SiMS microscopy was first introduced in 2002 (Watanabe and Mitchison). In this method, individual enhanced green fluorescent protein (EGFP)-actin molecules in the F-actin state are visualized in cells expressing a very low level of EGFP-actin under the control of the defective cytomegalovirus (CMV) promoter. While signals from EGFP-actin assembled with F-actin appear as a small spot, signals from freely diffusing probes are blurred on the image with approximately 2 s exposure time. Since this method was developed from FSM, we call these spots “speckles”. With this simple trick, cytoskeletal polymers and their associating proteins can be visualized in a state specific manner.

There are two practical approaches to measure the probe's lifetime (duration time) for association with cell structures (Watanabe & Mitchison 2002; Watanabe 2012). One is the Lifetime distribution analysis, in which duration time of each molecule from its appearance to disappearance is measured. The Lifetime method provides precise information on the turnover kinetics of the probe, although the number of observations needs to be 500 or more for accurate interpretation of underlying mechanisms. This method is suited to the analysis of the molecules with relatively rapid turnover kinetics because a large number of time-lapse series are required to cover the duration from appearance to disappearance of long-lived probes (typically twice the longest lifetime). The other method is called the Regression analysis. It follows the surviving fraction of speckles, which pre-exist in a certain frame of time-lapse images. Although the Regression method does not readily reveal the multiple dissociation kinetics of the molecule, it is suited for monitoring the effects of a bioactive compound because this method requires only several consecutive images to derive the overall dissociation rate of the molecule, thus enabling repeated measurements before and after the treatment.

To aid analysis of SiMS data, we have recently developed semiautomatic tracking software, Speckle TrackerJ (Smith et al. 2011) and JFilament (Smith et al. 2010). These are freely-available open-source ImageJ plug-ins. They are designed to be flexible, allowing manual operation of a number of error-correction commands in combination with the computer-based particle tracking and the cell contour detection. These features greatly help the collection of reliable results from images of mixed qualities. Basic protocols for SiMS microscopy (Watanabe 2012) and data analyses using these software packages (Smith et al. 2011; Ryan et al. 2013) were introduced in recent publications.

In the following sections, we introduce our recent studies aimed at unifying the SiMS analysis with other approaches to solve the complex actin turnover mechanisms in cells.

Quantitative modeling of multiple SiMS data on actin and actin-associated proteins to investigate actin treadmilling in lamellipodia

There has been a long-standing debate as to whether “treadmilling” governs actin filament turnover in lamellipodia. Animal cells crawl by extending lamellipodia, thin sheet-like structures that protrude from the cell body. In cultured cells, F-actin flows toward the cell center. This centripetal flow was recognized in 1960s (Ingram 1969; Abercrombie et al. 1970) and its identity was revealed as the retrograde actin flow by the actin FRAP experiment in 1980s (Wang 1985). Electron microscopy unambiguously demonstrates the polarity of F-actin whose fast growing barbed end faces the cell edge (Small et al. 1978; Narita et al. 2012). From these observations, the “treadmilling model” in which actin assembles at the tip of lamellipodia, flows inward and finally depolymerizes at the base of lamellipodia has been supported by many researchers. On the other hand, photoactivation of caged fluorescent actin (Theriot & Mitchison 1991) and our SiMS analysis (Watanabe & Mitchison 2002) demonstrated that F-actin disassembles prematurely before reaching the lamellipodium base. Other studies both in vivo and in vitro have implicated either filament treadmilling (Small et al. 1995; Carlier et al. 1997; Borisy & Svitkina 2000) or frequent filament severing and disruption (Ghosh et al. 2004; Andrianantoandro & Pollard 2006; Brieher et al. 2006; Kueh et al. 2008) as the major actin remodeling mechanism.

We recently examined whether ‘treadmilling’ is the sole actin filament turnover mechanism by unifying SiMS kinetics data of actin and its major regulators into a mathematical model (Miyoshi & Watanabe 2013). In this study, we restricted the condition only to the filament disassembly without disruption in the middle of the filament. In this simple scenario, individual lifetime of actin subunits in the filament can be calculated based on the queuing theory. We were then able to test the validity of the treadmilling model without predicting the kinetic parameters based on in vitro biochemical data. Our results suggest that exclusive filament treadmilling of actin is not likely to operate in lamellipodia. Although the questions concerning the frequency of filament severing and its location in individual filaments still remain unsolved, this example highlights the potential of the data derived from SiMS analysis to quantitatively validate the hypotheses, which are otherwise treated phenomenologically in most biological sciences.

Comparison between SiMS and FRAP data implies recycling with actin oligomers

As mentioned above, Rottner and his colleagues pointed out the apparent discrepancy between their actin FRAP data (Lai et al. 2008) and our SiMS kinetics data (Watanabe & Mitchison 2002) on actin filament turnover in lamellipodia. Certainly, their FRAP data appear to show nearly exclusive actin incorporation at the extreme leading edge. In marked contrast, our SiMS data show approximately one third of the newly-polymerized F-actin disassembling within 10 s, pointing to the frequent re-polymerization of actin throughout lamellipodia.

The discrepancy prompted us to carry out direct comparison between SiMS and FRAP in the same XTC cells (Smith et al. 2013). During the course of this study, we learned an important lesson about the difficulty in actin FRAP experiments. We realized that actin FRAP analysis yields wide variations in the recovery rates and levels of photobleached actin labels between the measurements. Normally in FRAP experiments, if the local amount of the molecule of interest is steady, the recovery curve of FRAP labels should return to nearly 100% of the original signal. In lamellipodia, however, F-actin exhibits a graded decrease toward the lamellipodium base. Therefore the recovery curves climb only halfway when the FRAP labels migrate along the retrograde flow and reach the lamellipodium base (Fig. 1a). This makes it difficult to exclude photobleaching causing damage to the local actin network and preventing its recovery. Moreover, we noticed significant variations in the actin FRAP recovery rates between the measurements. Presumably because of the rapid local actin turnover, actin assembly could be either up- or downregulated temporally, which may cause a large fluctuation in the F-actin density at the photobleached spot fast enough to affect the apparent actin FRAP recovery.

Figure 1.

Dual-color FRAP (fluorescence recovery after photobleaching) (an analogous method to fluorescence localization after photobleaching [FLAP]) of actin reports more accurate F-actin turnover rates than single-color actin FRAP. Fluorescence recovery of enhanced green fluorescent protein (EGFP)-actin after photobleaching with a 435 nm laser pulse generated by Micropoint (Photonic Instruments) was measured in XTC cells simultaneously expressing mCherry-actin. The recovery of EGFP-actin fluorescence was monitored in the photobleached area moving along the retrograde actin flow (a). mCherry-actin fluorescence was used to monitor the changes in the local actin content in the same area (b). The ratio of GFP-actin to mCherry-actin fluorescence in the photobleached area is shown (c). Each colored line indicates the data from the same cell (n = 7 cells). The photobleaching was carried out at time = 0. Cell images in GFP channel (top panels), mCherry channel (middle panels) and the ratio images of GFP-actin to mCherry actin (bottom panels) are shown (d). The measured areas are indicated by dotted lines. Note that without normalization by mCherry-actin fluorescence (a), the recovery rate and the final intensity of photobleached GFP-actin fluorescence vary because of the variations and the changes in the local actin content as revealed by mCherry-actin fluorescence (b). After normalization, the final ratios become relatively constant between the measurements (c). Scale bar, 5 μm.

To avoid such complications, we turned to dual-color FRAP, an analogous approach to the FLAP (fluorescence localization after photobleaching) (Zicha et al. 2003). This method uses counter staining of the network by RFP-actin (mCherry-actin) as a reference for local F-actin amounts in addition to photobleached GFP-actin. The ratio between GFP- and mCherry-actin in the photobleached spot revealed more constant FRAP recovery kinetics than GFP-actin alone (Fig. 1). It is important to note that without the data on mCherry-actin gradients and changes (Fig. 1b), it is difficult to correctly determine how much of the FRAP recovery (Fig. 1a) reflects the recovery due to filament turnover or due to the local fluctuation in the F-actin density.

In this paper (Smith et al. 2013), dual-color FRAP helped us to select and analyze cells with minimal fluctuations during fluorescence recovery. We then simulated the dual-color FRAP experiments with the molecular kinetics data obtained by SiMS analysis, and found that the measured dual-color FRAP kinetics was marginally slower than the simulated FRAP kinetics with the SiMS data. As a possible mechanism to reconcile the remaining difference, we propose that actin might recycle as slowly diffusing O-actin (oligomeric-actin) released from the filament network. This hypothesis originates from our finding of actin turnover dependent fast capping protein dissociation from the lamellipodial actin network (Miyoshi et al. 2006). Capping protein (CP) which binds the barbed end tightly in vitro dissociates from actin fast at 0.57 s−1 in vivo, in a manner sensitive to several actin stabilizing treatments. Another actin capper AIP1 also decreases its actin dissociation rate upon treatment with an actin stabilizing drug jasplakinolide (Tsuji et al. 2009). We postulated that those actin end interacting molecules in part dissociate from the actin network in association with O-actin (Miyoshi et al. 2006; Watanabe 2010). The comparison between actin FRAP and SiMS kinetics has added supportive evidence to this O-actin hypothesis (Smith et al. 2013). Without the data obtained from dual-color FRAP experiments, it might have been difficult to draw a conclusion because the FRAP kinetics before normalization with mCherry-actin were quite diverged. Dual-color FRAP (or FLAP) should be considered as a routine investigation for in vivo dynamics measurements.

s-FDAPplus elucidates the mechanosensitive G-actin increase that leads to actin nucleation burst of formins

We finally introduce a study where bulk measurement of the soluble fraction of the molecule complemented our SiMS data. We recently reported using SiMS microscopy that formin homology proteins (formins) including mDia1 rapidly show an increased actin nucleation frequency in response to the mechanical perturbation of the cell cortex (Higashida et al. 2013). Formins are the major actin nucleators and processive actin polymerizers required for cytokinesis and polarity formation of eukaryotes including yeast and plants (Goode & Eck 2007; Blanchoin & Staiger 2008). Because formins processively polymerize actin over long distances (Higashida et al. 2004), SiMS microscopy is an ideal tool to monitor the actions of formins in cells. We found that multiple formins initiate processive movement shortly after the cell cortex is deformed by microneedles. This activation occurs within 10 s and lasts for less than 100 s. Mechanosensitive actin nucleation by mDia1 requires neither Ca2+ nor kinase signaling. Because multiple formins which share the conserved structures only in formin homology 1 and 2 domains are activated similarly by the needle manipulation, we predicted a certain global mechanism responsible for this mechanosensitive actin nucleation.

We then tested a hypothesis that the concentration of G-actin, which is the substrate of formin-catalyzed actin filament nucleation, might surge to increase the rate of actin nucleation by formins in mechanically perturbed cells. We previously proposed a similar mechanism for the low-dose latrunculin-induced activation of multiple formins. In this case, low-dose latrunculin was found by pharmacokinetic simulation to induce a rapid, several-fold increase in free G-actin, which we call “latrunculin paradox”; a rare situation in which an inhibitor increases its target in a drug-free state (Higashida et al. 2008).

Several experiments provided evidence for G-actin-regulation of formin activation. First, the extent of dissolution of actin stress fibers showed significant correlation with the increase in the frequency of mDia1-mediated actin nucleation during microneedle manipulation. Second, actin associated AIP1 monitored by SiMS microscopy increased by approximately 20% after needle manipulation. AIP1 interacts with the barbed end of the cofilin-severed actin filament (Okada et al. 2002; Tsuji et al. 2009). We used AIP1 as a marker to monitor the activity of the major actin depolymerizer, cofilin. Third, inactivation of cofilin by overexpression of LIM-kinase (Yang et al. 1998) abolished actin nucleation by mDia1. However, it remained uncertain from these observations whether G-actin indeed increases by mechanical perturbation of the cell cortex.

We therefore directly tested whether G-actin increases after needle manipulation by developing an improved s-FDAP method called s-FDAPplus. s-FDAP has been developed for measuring the sequential change in the mobile fraction of actin (Kiuchi et al. 2007, 2011a,b). s-FDAP uses a photoactivatable green fluorescent protein, Dronpa (Ando et al. 2004) tagged with actin (Dp-actin). By photoactivating and erasing Dronpa fluorescence, s-FDAP enables to measure the fast decay of local actin labels, which is used for calculating the fraction of G-actin repeatedly in a single cell.

The results provided evidence that G-actin increases immediately upon needle manipulation to a degree enough for promoting the actin nucleation frequency of mDia1 and other formins. s-FDAPplus analysis revealed that G-actin increased by more than 20% in three cells and by 10–20% in four cells out of 12 manipulated cells (Higashida et al. 2013). Although these increases in G-actin are not large, calculation of the equilibrium between G-actin and its sequestering proteins including profilin, thymosin β4 and others indicates that the approximately 20% increase in total G-actin leads to a surge of free G-actin from 0.5 μmol/L to 1.0–1.5 μmol/L. Since the actin nucleation frequency of the mDia1 FH2 domain depends on the cube of the free G-actin concentration (Li & Higgs 2003), this modest increase of free G-actin may give rise to the marked increase in actin nucleation by formins in mechanically stimulated cells.

SiMS that represents the kinetics of cell structure-bound molecules and s-FDAPplus that reports the soluble fraction of actin provided compelling evidence of G-actin regulation of mechanosensitive actin nucleation of formins. In practice, we had to carefully improve the sensitivity of the s-FDAP method (Fig. 2) as the estimated increase in the total G-actin was as small as 5–10%. We improved s-FDAP in several points. First, s-FDAPplus implements 11 frame serial image acquisition at 100 ms intervals, instead of two images in original s-FDAP (Kiuchi et al. 2011a), for the decay analysis of photoactivated Dp-actin. The mobile fraction is calculated by curve fitting with a model (Kiuchi et al. 2011a). Second, s-FDAPplus estimates the total actin amount in the measured area from Dp-actin fluorescence acquired during the photobleaching step. Third, red fluorescent proteins are used to monitor cell thickness change. These improvements reduced the error to within a few percent range, which was essential for testing our hypothesis of G-actin regulation of formin-catalyzed actin nucleation. The combination of SiMS and s-FDAPplus analyses provided compelling evidence of G-actin regulation of formin-catalyzed actin nucleation by illuminating the two sides of a coin in the actin turnover cycle (Higashida et al. 2013).

Figure 2.

The improved-FDAP (sequential fluorescence decay after photoactivation)plus analysis to measure the cytoplasmic G-actin concentration before and after micromanipulation. (a) Image acquisition of mPlum, photobleaching of Dp-actin, photoactivation of 6.8 μm diameter region (white circle) by 390 nm laser and fast time-lapse imaging of Dp-actin were repeated 15 times at 4.3 s intervals. Cells were physically stimulated by microneedle manipulation for a few seconds before the 6th set. Green lines show schematic representation of photoactivation and decay of Dp-actin fluorescence. (b) In each set, mobile fraction was calculated by fitting the Dp-actin fluorescence decay (red dots) to the model curves (blue lines). Local actin content and cell volume were estimated by Dp-actin fluorescence during photobleaching and mPlum fluorescence, respectively. These normalizations were crucial for the accurate measurement. The G-actin concentration was calculated by (mobile fraction) × (local actin content)/(cell volume). Scale bar, 10 μm. Modified from Higashida et al. 2013.


In this article, we have introduced our three recent studies in which SiMS analysis of actin and its regulators were combined with other techniques. Although SiMS microscopy is able to directly visualize assembly and disassembly of the proteins in living cells, comparison of the data with other approaches has given us deeper insights than SiMS data alone. Importantly, in all cases, quantitative analyses contributed crucial information to the solution of the problems. Even in cases where direct visualization at the molecular level is difficult, we suggest that careful comparison of bulk observation results such as FRAP kinetics data with quantitative modeling of molecular behavior might offer precise understanding of the molecular and structural remodeling pathways in complex biological systems.


This work was supported by Human Frontier Science Program (RGP0061⁄2009-C), the Cabinet Office, Government of Japan, through the Funding Program for Next Generation World-Leading Researchers (LS013) and a grant from the Takeda Science Foundation.