Drosophila class-I myosins that can impact left-right asymmetry have distinct ATPase kinetics

Myosin-1D (myo1D) is important for Drosophila left-right asymmetry, and its effects are modulated by myosin-1C (myo1C). De novo expression of these myosins in nonchiral Drosophila tissues promotes cell and tissue chirality, with handedness depending on the paralog expressed. Remarkably, the identity of the motor domain determines the direction of organ chirality, rather than the regulatory or tail domains. Myo1D, but not myo1C, propels actin filaments in leftward circles in in vitro experiments, but it is not known if this property contributes to establishing cell and organ chirality. To further explore if there are differences in the mechanochemistry of these motors, we determined the ATPase mechanisms of myo1C and myo1D. We found that myo1D has a 12.5-fold higher actin-activated steady-state ATPase rate, and transient kinetic experiments revealed myo1D has an 8-fold higher MgADP release rate compared to myo1C. Actin-activated phosphate release is rate limiting for myo1C, whereas MgADP release is the rate-limiting step for myo1D. Notably, both myosins have among the tightest MgADP affinities measured for any myosin. Consistent with ATPase kinetics, myo1D propels actin filaments at higher speeds compared to myo1C in in vitro gliding assays. Finally, we tested the ability of both paralogs to transport 50 nm unilamellar vesicles along immobilized actin filaments and found robust transport by myo1D and actin binding but no transport by myo1C. Our findings support a model where myo1C is a slow transporter with long-lived actin attachments, whereas myo1D has kinetic properties associated with a transport motor.

Myosin-1D (myo1D) is important for Drosophila left-right asymmetry, and its effects are modulated by myosin-1C (myo1C). De novo expression of these myosins in nonchiral Drosophila tissues promotes cell and tissue chirality, with handedness depending on the paralog expressed. Remarkably, the identity of the motor domain determines the direction of organ chirality, rather than the regulatory or tail domains. Myo1D, but not myo1C, propels actin filaments in leftward circles in in vitro experiments, but it is not known if this property contributes to establishing cell and organ chirality. To further explore if there are differences in the mechanochemistry of these motors, we determined the ATPase mechanisms of myo1C and myo1D. We found that myo1D has a 12.5-fold higher actin-activated steady-state ATPase rate, and transient kinetic experiments revealed myo1D has an 8-fold higher MgADP release rate compared to myo1C. Actin-activated phosphate release is rate limiting for myo1C, whereas MgADP release is the rate-limiting step for myo1D. Notably, both myosins have among the tightest MgADP affinities measured for any myosin. Consistent with ATPase kinetics, myo1D propels actin filaments at higher speeds compared to myo1C in in vitro gliding assays. Finally, we tested the ability of both paralogs to transport 50 nm unilamellar vesicles along immobilized actin filaments and found robust transport by myo1D and actin binding but no transport by myo1C. Our findings support a model where myo1C is a slow transporter with long-lived actin attachments, whereas myo1D has kinetic properties associated with a transport motor.
Myosin-Is are single-headed cytoskeletal motors that interact with actin filaments to carry out cellular functions related to membrane trafficking, dynamics, and organization (1). Recent studies have shown that two Drosophila myosin-I paralogs, myosin-1C (myo1C; also known as myo61F) and myosin-1D (myo1D; also known as myo31DF), are expressed in tissues that undergo left-right (L/R) asymmetry during development. L/R asymmetry is the process in early embryonic development that breaks the normal symmetry in the bilateral embryo and in flies occurs when organs such as the hindgut turn to become chiral (2)(3)(4)(5). Knockdown of myo1D results in situs inversus in chiral cells and tissues, and myo1C appears to negatively regulate myo1D activity when overexpressed (6,7). Interestingly, overexpression of either paralog in nonchiral tissues results in cell and tissue chirality, with the handedness dependent on the paralog. Thus, myosin-Is are sufficient to induce cell, tissue, and organismal chirality in Drosophila (6,8).
Directionality of induced L/R asymmetry in Drosophila tissues is determined by the identity of the myosin-I motor domain (6). Expression of protein constructs that contain the myo1D motor domain in the nonchiral epidermis or trachea results in dextral rotation of the tissue, and expression of the myo1C motor domain results in sinistral rotation (6). Interestingly, purified, recombinant myo1D turns actin filaments in a counterclockwise direction in in vitro actin gliding assays, whereas myo1C-powered actin filaments do not have a directional preference. While chiral turning of actin filaments by myosin may play a role in development of L/R asymmetry in cells and tissues, the lack of chirality from myo1C implies that it is not sufficient on its own, and so differences in the kinetics of the paralogs may also be important for differences in chiral development (9,10).
In this study, we used transient and steady-state biochemical techniques to determine the rate constants that define the ATPase pathways of myo1C and myo1D (Fig. 1). Our goal was to test if there are differences in the ATPase pathways that could lead to functional specializations of the paralogs. We found the steady-state ATPase activities differ >10-fold. Importantly, we found that both motors have duty ratios >0.4 with the attachment lifetimes in force-bearing states differing by $9-fold. Even though both paralogs have high duty ratios, only myo1D could support transport of 50 nm vesicles in vitro. We propose that these kinetic differences could be pivotal for their function and dynamics in L/R asymmetry during Drosophila development.

Protein purifications of full-length myosin-Is and light-chain binding properties
Recombinant full-length myo1C and myo1D constructs containing C-terminal Avi and FLAG tags were expressed using baculovirus in Sf9 insect cells and copurified with calmodulin (CaM) as described (6) (Fig. 2). Sequence analysis of myo1C and myo1D suggest the presence of three and two light-chain-binding IQ motifs, respectively (5). We determined the stoichiometry of the light chain to motor bound to each paralog by pulling down biotinylated myosin-Is with streptavidin beads and assessing CaM concentration by SDS-PAGE. We found an average of 2.8 ± 0.3 and 1.7 ± 0.2 light chains bind to myo1C and myo1D, respectively (Fig. S1), consistent with the number of available light chain binding sites for both paralogs. Non-CaM light chains were not found to copurify with either myosin.

Myosin-Is steady-state ATPase activity
The steady-state ATPase activities of myo1C and myo1D were determined using the NADH-coupled enzyme reaction in the presence of 0 to 120 μM phalloidin-stabilized F-actin. Actin substantially activates the ATPases activities of both paralogs (Fig. 3). The actin concentration dependence of the ATPase rates (k obs ) was fitted by, where the actin concentration at half-maximum of the ATPase rate (K ATPase ) rate is 48 ± 5 μM for myo1C and 26 ± 4 μM for myo1D, and the maximum ATPase rate (V max ) is 0.44 ± 0.02 s −1 for myo1C and 5.1 ± 0.3 s −1 for myo1D (Table 1). Minimizing contaminating MgADP from protein preparations was crucial, as MgADP inhibited the ATPase rates due to its very tight actomyosin affinity (see below and Experimental procedures; (11)).

Actin-activated phosphate release
Phosphate binding protein (P i BP) fluorescently labeled with 7-diethylamino-3-((((2-Maleimidyl)ethyl)amino)carbonyl) coumarin was used to measure the actin-activated phosphate (P i ) release rate (k +4 0 ; Fig. 1) in sequential-mix, stopped-flow  . The rate of P i release reports the transition of myosin from weak actin-binding to strong actin-binding states. Myo1C preincubated with apyrase-VII, to remove contaminating MgADP, was rapidly mixed with 60 μM MgATP and aged for 5 s to allow for MgATP binding and hydrolysis. The time course of phosphate release in the absence of actin followed a single exponential function with a rate of 0.03 ± 0.02 s −1 . The observed rate increased hyperbolically with increasing actin, which was also pretreated with apyrase-VII (Fig. 6A). No lag was apparent in the transients, and a faster phase was not observed within the first second after mixing. The mechanism of the reaction was modeled as a two-step binding reaction as in Figure 7.
The maximum rate of P i release obtained from fitting the observed rate versus actin by: is k +4 0 = 0.28 ± 0.03 s −1 for myo1C (Fig. 6C). The affinity of myo1C.ADP.P i for actin filaments is K 9 0 = 27 ± 7 μM. Contaminating phosphate and competition for phosphate binding from the "phosphate mop" in the buffer during the very slow transient resulted in variability in the k obs. Nevertheless, we find the actin-dependent rate of P i release rates to be similar to the actin-dependence of the steadystate ATPase activity, pointing to k +4 0 as the rate-limiting step (Fig. 6C, green points).
Apyrase-treated myo1D was mixed with 60 μM ATP and aged for 5 s to allow MgATP hydrolysis. A range of 0 to 60 μM actin was then rapidly mixed with that MgATP-bound myosin. Phosphate release in the absence of actin had a rate of 0.02 ± 0.02 s −1 , and k obs increased hyperbolically with actin ( Fig. 6B). The observed rapid phase of phosphate release was followed by a slow phase that is the result of multiple ATPase turnovers. The rates of the fast phase in the presence of actin were substantially faster than the myo1C transients. The rates of the fast phase were fit by Equation 3. The maximum rate of P i release for myo1D is 96-fold faster than myo1C (k +4 0 = 27 ± 4 s −1 ), and the affinity of myo1D.ADP.P i for actin filaments is K 9 0 = 12 ± 5 μM (Fig. 6D). The maximum rate is more than 5-fold faster than V max , suggesting that the rate-limiting-step of myo1D occurs after P i release.

MgADP release
The rate of MgADP release (k +5 0 ) was determined by MgATP-induced dissociation of actomyo1C.ADP and acto-myo1D.ADP from pyrene-actin as shown in Figure 8.
When myosin active sites are saturated with MgADP, the rate of MgATP-induced dissociation of actomyosin is limited by the slow dissociation of MgADP (11). MgADP (2.5 μM; see below) was preincubated with 0.5 μM pyrene-actomyosin-I and mixed with 1 mM MgATP. Fluorescence transients were best fit by single exponential functions. The rate of MgADP release (k +5 0 = 1.0 ± 0.1 s −1 ) from actomyo1C is $2-fold faster than V max , and rate of MgADP release from actomyo1D (k +5 0 = 8.7 ± 1.9 s −1 ) is similar to V max , suggesting that this step is ratelimiting for myo1D (Fig. 9, A and B). Table 1 Rate and equilibrium constants for key steps of Drosophila myo1C and myo1D ATPase cycle at 20º C a

Rate and equilibrium constants
Myo1C Myo1D   In the presence of subsaturating MgADP concentrations, MgATP-induced pyrene-actomyosin fluorescence transients were best modeled as the sum of two exponential components: a fast component resulting from MgATP binding to nucleotide-free actomyosin-I and a slow component that is the result MgADP dissociation (k +5 0 ) (Fig. 9C). The affinity of MgADP for pyrene actomyosin (K 5 0 ) was determined by measuring the fraction of the amplitude corresponding to the slow component (A slow ) as a function of MgADP concentration (Fig. 9D). The high affinity of MgADP for myosin required fitting by the quadratic equation, where y is the fraction of myosin-I bound to MgADP (12). These myosin-Is have nanomolar affinity for MgADP (K 5 0 ): 89 nM for myo1C and 44 nM for myo1D, which are among the tightest MgADP affinities observed for characterized myosins (13)(14)(15)(16).

Duty ratio
The duty ratio is defined as the fraction of the ATPase cycle in which myosin dwells in actin-attached, force bearing states. Our kinetic characterization points to MgADP release as the step that limits exit from the strongly bound states; thus, we can estimate the duty ratio as: The calculated myo1C (0.44) and myo1D (0.59) duty ratios at saturating actin concentrations are substantially higher than other characterized myosin-I paralogs (1,10,14,15,17).

Actin gliding assays
To assess motor activity, we carried out actin gliding assays by attaching a range of biotinylated myosin-I concentrations (50 nM -200 nM) to neutravidin adsorbed to a glass coverslip ( Fig. 10A; see Experimental procedures). Myo1C propelled actin filaments with an average velocity of 33 ± 6 nm/s, while myo1D propelled filaments with an average velocity of 130 ± 3 nm/s, consistent with kinetics showing myo1D is a faster motor than myo1C. Increasing the motor concentration in the motility chamber resulted in decreases in the average velocities of myo1C (21 ± 2) and myo1D (94 ± 8 nm/s; Fig. 10B). The increase of motor concentration likely results from multiple motors attaching to a single filament at higher densities, inhibiting actin gliding. Both myosin-Is slow with increasing motor density (Fig. S2), suggesting mechanical load slows gliding; however, further mechanochemical studies need to be performed to assess force sensitivity (18)(19)(20)(21)(22)(23). The observed gliding velocities in this experimental setting are faster compared to velocities on fluid lipid bilayer substrates (6), and circular actin gliding promoted by myo1D is attenuated due to rigid attachment to nitrocellulose, as previously described (23).
Kinetically, we observe that both myosin-Is have nanomolar affinity for MgADP ( Fig. 9; Table 1), so we explored the effect of MgADP on actin gliding speeds. We observed a sharp MgADP concentration-dependent decrease in gliding velocity in the presence of 2 mM MgATP (Fig. 10C). The data were fitted by the following equation to determine the MgADP inhibition constant (K i , (24)): where v 0 is the gliding velocity in the absence of MgADP and was fixed at 36 nm/s for myo1C 110 nm/s for myo1D. Fits yielded K i values of 913 ± 90 μM and 222 ± 88 μM for myo1C and myo1D, respectively (Table 1), which indicates that MgADP effectively competes with 2 mM MgATP for binding.

Drosophila myo1D supports small unilamellar vesicles trafficking
Our kinetic characterization indicates that myo1C and myo1D have high duty ratios, which may suggest an ability to transport cargo at low densities. Thus, we tested the ability of myo1C and myo1D to transport small unilamellar vesicles (SUVs) created by extrusion through 50 nm filters composed of 4% PtdIn (4,5)P 2 , 96% 18:1 (Δ9-Cis) PC, and 0.04% lissamine rhodamine-labeled PE along immobilized actin filaments ( Fig. 10D; see Experimental procedures). Mixing 10 nM or 50 nM myo1D with 5 μM SUVs resulted in processive vesicle motility, with many vesicles traversing the entire length of actin filaments and switching tracks (Movie S1). Automated fluorescent particle tracking revealed experiments at both concentrations had similar average speeds and run lengths of the vesicles at (53 ± 32) nm/s and (1.8 ± 1.3) μm (Fig. 10E). Addition of 10 nM myo1C showed no vesicle motility, although occasional actin binding was observed (Movie S1) and increasing motor concentration to 50 nM increased actin binding but did not support motility. These data suggest that myo1D is a bona fide high duty ratio motor that can support vesicle trafficking even at low motor concertation, while myo1C can support limited vesicle binding to actin but not promote motility under the same concentrations.

Myo1C and myo1D have substantially different biochemical kinetic properties
In this study, we define key kinetic rate constants and establish important biochemical differences between two myosin-I paralogs that have opposing roles in cell and tissue chirality in Drosophila. Myo1C and myo1D share the same ATPase pathway found for other myosins (9, 10, 12, 25), but the kinetic rate constants that define their pathways result in the two motors having distinct motile properties (Table 1). Kinetic differences include: (1) >10-fold faster V max for myo1D, (2) phosphate release as the rate-limiting step from actomyo1C and MgADP release from actomyo1D, (3) $9-fold faster rate of MgADP release from actomyo1D, and (4) a greater duty ratio for myo1D.
The different duty ratios and attachment lifetimes of the two myosins likely reflect enzymatic adaptations that evolved for specific biological functions. Thus, a notable finding is that although myo1D has a higher duty ratio than myo1C, the actin-attachment lifetime of myo1C ($1 s) is predicted to be $9-fold longer than myo1D ($110 ms). Additionally, calculated actin-attachment rates of the M.ADP.P i states (k +4 0 /K 9 0 ; Fig. 6) are substantially different for myo1C (0.010 μM −1 s −1 ) and myo1D (2.3 μM −1 s −1 ). These parameters suggest that myo1D may be suited for processes that include rapid actin attachment and detachment kinetics or vesicle transport, whereas myo1C may be suited for sustained tension maintenance. This prediction agrees with our finding that myo1D can processively transport 50 nm vesicles along single actin filaments, but myo1C cannot under experimental conditions (Movie S1). Additionally, myo1D motors will detach from actin filaments much more rapidly, possibly relieving tension in the actin cytoskeleton (see below).
Myo1D moves actin filaments in gliding assays at a greater rate than myo1C but not to the extent predicted by the $9fold difference in the rate of MgADP release. This discrepancy in rates could be explained by differences in lever arm compliance. It is also possible that the speed of myo1C is limited by slow transition rate of the motor to the strong binding states (26). Further mechanochemical studies such as optical trapping experiments and structural studies such as cryo-EM are necessary to address this question.

Relating myosin-I biochemistry to cellular function
During development in Drosophila, class-I myosins are transiently expressed in tissues that undergo L/R asymmetry, and knocking out myo1D results in situs inversus (2,4,6). The mechanisms by which myosin-I paralogs participate in cell and tissue chirality have not been determined. However, there are two important facets to the myosin-dependent chirality mechanisms. First, the presence of either myo1C or myo1D is  ATPase kinetics of class-I Drosophila myosins sufficient to induce chirality in nonchiral tissues (6). Second, the properties of the motor domain determine the chiral direction, so there must be an intrinsic property of the motors that lead to chirality and the differences in turning direction. Although we showed previously that the motors have differences in the ability to power asymmetric gliding of actin filaments in vitro when bound to fluid bilayers (6), it is not known if this turning activity is important for cell chirality. In this study, we revealed that there are substantial differences in the ATPase kinetics of the two motors (Table 1). We propose that these differences in mechanochemistry are important for their differing abilities to promote L/R asymmetry.
Cell biological and biochemical experiments in Drosophila suggest a direct link between myosin-I and cell-cell adhesion proteins, DE-cadherin and β-catenin (7). In this role, myosin-Is may work as a tether in conjunction with other actin cytoskeleton components to induce chirality. Notably, the formin DAAM has been shown to be essential for L/R asymmetry development in Drosophila, where it interacts with both myo1C and myo1D (27). Thus, the interaction between myosin-I and the DAAM-dependent F-actin network appears to create the chiral cytoskeleton.
It has been proposed that vertebrate formin proteins induce cell chirality through their spinning at the actin filament end as a result of their processive elongation activity (28,29). Torque generated by this spinning has been modeled to be transmitted to the cellular actin network to induce chirality via the crosslinking protein, alpha-actinin 1. Strikingly, varying the expression levels of alpha-actinin 1 changes the handedness of cell chirality (28,30), which is likely the result of altered attachment lifetimes between the cytoskeletal linkages. These altered kinetics are proposed to result in changes in cellular twisting direction (30). Given the >8-fold differences in the actin-attachment lifetimes of myo1C and myo1D, as determined by MgADP release rates (k +5 0 ), we propose that myosin-I may be playing a similar role in communicating the chiral Transients are composed of a fast phase corresponding to MgATP-induced actomyosin-I dissociation, followed by a slow phase corresponding to MgADP release. D, MgADP affinity (K 5 ') of myo1C (red) and myo1D (blue) dependent on nucleotide concentration. The observed rate constants were obtained at each nucleotide concentration by fitting the stopped-flow data to a square fit equation. The solid lines represent the best fits by Equation 4 with K 5 ' = 89 ± 2 nM for myo1C and 44 ± 2 nM for myo1D (11). The error bars correspond to the standard deviation of three independent experiments using three different protein preparations per paralog. myo1C, myosin-1C; myo1D, myosin-1D. activity of DAAM to the Drosophila cytoskeleton by crosslinking actin to adhesion proteins and/or the cell membrane.
Alternatively, myosin-I may have a role in cadherin transport and membrane recycling that affects generation of L/R asymmetry. In Drosophila, knockdown of myo1D expression leads to a decrease in Rab-11 endosomes and less E-cadherin in the cell-cell contacts in the hindgut (31). Given our finding that myo1D is capable of powering vesicle transport in vitro, it is possible that in conjunction with formin activity, changes in myo1D-dependent trafficking could transport and concentrate E-cadherin at the adhesion site which may affect chirality. Interestingly, although not a planar cell polarity mechanism, the vesicle transport activity of myo1D is believed to be important for the development of nodal flow structures and development of L/R asymmetry in zebrafish (32,33). Further cell-and organism-based experiments designed to alter the myosin kinetics should help distinguish between these possibilities.
Myosin-I constructs were biotinylated at the C-termini Avi tag sequence using BirA biotin-protein ligase (Avidity). Supernatants collected post-FLAG columns were spinconcentrated using Amicon Ultra Centrifugal Filters (100 kDa cut-off) for 10 min until the protein volume was decreased 3-fold. BirA enzyme (10 μl of 3 mg/ml) and Biomix B buffer were added to final concentrations of 10 mM ATP, 10 mM Mg 2 + acetate, 50 μM d-biotin (Avidity). The mixture was incubated at 25 C for 30 min, followed by final purification on a Mono Q column. Myosin-I purity was confirmed through SDS-PAGE gel, and the protein was dialyzed against KM100 + 50% glycerol buffer as described above. Biotinylation was confirmed through chemiluminescence (SuperSignal West Pico PLUS Chemiluminescent Substrate; Thermo #34580) by HRP-streptavidin (Thermo, #21130).
Rabbit skeletal muscle actin was purified as previously described (34) and used without further modification for steady-state ATPase and phosphate release experiments. Pyrene-labeled F-actin was prepared as described (35,36). Both unlabeled and pyrene-labeled actin filaments were dialyzed against KMg25 buffer (10 mM Mops pH 7.0, 25 mM KCl, 1 mM EGTA, 1 mM MgCl 2 , and 1 mM DTT) to minimize free nucleotide. To further decrease the nucleotide contamination, both unlabeled and labeled actin were treated with saturating phalloidin concentrations (Cayman Chemicals, #18039) to stabilize filaments, and the stabilized filaments were sedimented at 146,944g for 45 min at 4 C. Pellets containing the filaments were resuspended in KMg25 and treated with Apyrase-VII (Sigma Aldrich, A6535) and incubated for 30 min at 25 C to remove free nucleotides. The F-actin was then centrifuged as above, and the pellet was resuspended in KMg25 and dialyzed against the same buffer overnight. Dialyzed F-actin was collected and stored on ice until used for experiments. Nucleotide concentrations for stopped-flow experiments were determined spectrophotometrically (ε 259 = 15,400 M −1 cm −1 ).

Stopped-flow kinetics
A stopped-flow apparatus (SX20 Stopped Flow Spectrometer) was used to acquire all transients. The dead time of the instrument is <3 ms with a 120-μl sample volume. Fluorescence excitation was provided by a 100-W Hg lamp. For steady-state ATPase activity, NADH absorbance was monitored at 340 nm. Pyrene-actin for MgATP-induced actomyosin-I dissociation and MgADP release was excited at 365 nm, and the fluorescence emission peak was detected using a 405 nm long-pass filter. For actin-activated phosphate release, fluorescently labeled mutant phosphate binding protein (MDCC-labeled PiBiP; (7-diethylamino-3-((((2-maleimidyl) ethyl)amino)carbonyl) coumarin)-labeled phosphate binding protein) was excited at 430 nm, and fluorescence was detected with a 440 nm long-pass filter (12,37). Data were acquired and analyzed using Pro Data-SX software. Stopped-flow data were fitted to exponentials functions by a nonlinear least-squares curve fitting. All the reagent concentrations reported are postmixing. For Pi release experiments, 375 nM-3 μM myo1C and 375 to 500 nM of myo1D were used.

Actin gliding assays
In vitro actin gliding assays were conducted in KMg25 buffer (also known as Motility Buffer 1X). Purified G-actin (34) was polymerized using the same buffer and stabilized with Rhodamine Phalloidin (Invitrogen, R415). Recombinant chicken CaM was expressed and purified as previously described (38).
The open ends of the motility chamber were sealed using vacuum grease to prevent drying while acquiring data. Fluorescent actin filaments were visualized using a Leica DMIRB microscope with a 100-fold magnification Leica oil-immersive objective of numerical aperture 1.4. The leading edge of actin filaments were tracked using the Manual Tracking plugin from ImageJ (39), and the average speeds were determined via displacement over time using Microsoft Excel.
An immobilized actin network in a flow chamber to assay SUV motility was created by flowing 0.2 mg/ml neutravidin, followed by two washes of 2 mg/ml BSA and followed by 100 nM F-actin stabilized with 90:10 Alexa-488:Biotin Phalloidin (Alexa-488 Phalloidin; ThermoFisher Scientific A12379 and Biotin Phalloidin; Thermo Fischer Scientific B7474) into a 0.5% nitrocellulose-coated coverslip.
Specified concentrations of myosin-Is were preincubated with 5 μM SUVs to ensure binding via myosin-I tail domain with PI(4,5)P 2 (6,23,40). The mixture was added to a solution containing final concentrations of 5 μM CaM, 2 mM MgATP, 20 mg/ml glucose oxidase, 4 mg/ml catalase, 5 mg/ml glucose, and BSA to avoid nonspecific binding in KMg25. The final solution added to chambers containing the immobilized actin network, and the chamber was sealed with vacuum grease. Image stacks were acquired by fluorescence microscopy at a rate of 1 frame/second for 5 min. Image states were processed using the Cega filtering program (41) and the Trackmate plugin from ImageJ (39).

Data availability
Data to be shared upon request to E. Michael Ostap (ostap@ pennmedicine.upenn.edu).
Supporting information-This article contains supporting information (5).