Engineering myosins for long-range transport on actin filaments.

Cytoskeletal motors act as cargo transporters in cells and may be harnessed for directed transport applications in molecular detection and diagnostic devices. High processivity, the ability to take many steps along a track before dissociating, is often a desirable characteristic because it allows nanoscale motors to transport cargoes over distances on the scale of micrometres, in vivo and in vitro. Natural processive myosins are dimeric and use internal tension to coordinate the detachment cycles of the two heads. Here, we show that processivity can be enhanced in engineered myosins using two non-natural strategies designed to optimize the effectiveness of random, uncoordinated stepping: (1) the formation of three-headed and four-headed myosins and (2) the introduction of flexible elements between heads. We quantify improvements using systematic single-molecule characterization of a panel of engineered motors. To test the modularity of our approach, we design a controllably bidirectional myosin that is robustly processive in both forward and backward directions, and also produce the fastest processive cytoskeletal motor measured so far, reaching a speed of 10 µm s(-1).

Previous investigations have shown that processive myosin motion can be tolerant to dramatic changes in lever arm structure and mechanics[9,10]. These studies challenged the idea that processivity depends critically on head-head coordination mediated by internal strain: while natural dimeric motors such as myosin V and myosin VI make use of strain-dependent gating of nucleotide exchange to prevent unproductive detachment of the front head[6-8], this mechanism may yield only modest gains in processivity relative to uncoordinated stepping. Processivity without coordination has also been observed and discussed in recent studies of cytoplasmic dynein[11,12]. Processive motion of a dimeric motor can be explained by the statistics of independently cycling monomers, dependent on kinetic properties of the catalytic head: specifically, the motor variant must have a high duty ratio, defined as the fraction of the mechanochemical cycle spent bound to the filament[3].

Models of uncoordinated stepping suggest that trimeric or tetrameric myosins — which have not been observed in nature — could be dramatically more processive than dimers. Potential gains are illustrated by a toy kinetic model that has been useful for conceptual discussions of processivity[7,10], in which the mechanochemical pathway of a myosin head is approximated by a two-state cycle governed by an attachment rate kon and a detachment rate koff. With three or four heads and a high duty ratio r=kon/(kon+koff), the termination of a run by simultaneous detachment of all heads becomes an unlikely event (Supplementary Modeling). This simple picture ignores potential detrimental effects of additional heads, which could include disruption of natural strain-mediated nucleotide gating, unintentional introduction of unproductive gating, or interference with actin rebinding due to mechanical or steric interactions between heads or due to site occlusion.

To test whether increasing the number of heads can improve the processivity of engineered motors, we designed a series of constructs based on myosin VI. Myosin VI has a high duty ratio[5] of 0.85–0.95[10,13] and is processive in dimeric constructs[5,10], including chimeric dimers with artificial lever arms composed of 2 spectrin repeats[9,10]. We created M6DI8162R~DIM, M6DI8162R~TRI, and M6DI8162R~TET by fusing the chimeric myosin VI motor to engineered coiled coil variants[14] GCN4-pIL, GCN4-pII, and GCN4-pLI (DIM, TRI, and TET) to generate dimeric, trimeric, and tetrameric motors, respectively (Fig. 1) (Supplementary Figure 5). We used single-fluorophore tracking (Supplementary Movies 1–2) to quantify the processivities of the engineered constructs. For all ATP concentrations investigated, motor trajectories (Fig. 1a) show run lengths that increase steeply with increasing oligomerization state (Fig. 1b). M6DI8162R~TRI and M6DI8162R~TET suffer only minor reductions in velocity relative to their dimeric counterpart (Fig. 1c).

Figure 1

Multimerization effects on engineered myosin VI processivity and stepping behaviour. (a) Single molecule fluorescence trajectories for M6DI8162R dimers (red), trimers (green), and tetramers (blue) recorded at 2mM ATPtogether with corresponding cartoons and block diagrams, showing differences due to oligomerization state. (b) Mean run lengths as a function of [ATP]. M6DI8162R~DIM is fit to a 3-state model10, yielding krebind=290 ± 30 s−1, kADPoff=5.8 ± 0.1 s−1, kATPon=0.015 ± 0.001 s−1μM−1, and the defect parameter d=63 ± 6 steps. (c) Motor velocities. M6DI8162R~DIM is fit to the 3-state model, and M6DI8162R~TRI and M6DI8162R~TET are fit to Michaelis Menten kinetics. Trimer: KM=740 ± 40 μM and Vmax=120 ± 2 nm/s; tetramer: KM=630 ± 50 μM and Vmax=99 ± 3 nm s−1. Data are displayed as mean ± s.e.m. (d-e) Gold nanoparticle tracking motor stepping traces. (d) Raw trajectories (colored traces, collected at 500 Hz) are shown along with fits (black) generated by an automatic stepfinding procedure (see Methods). (e) Step size distributions. Histograms of displacements generated by stepfinding (N > 1000 steps for each construct) are shown together with fits (black) to sums of multiple Gaussian distributions. Peak locations for dimer: −12 ± 0.1 nm and 20 ± 0.3 nm; trimer: −9 ± 0.1 nm, 9 ± 0.1 nm, and 21 ± 0.3 nm; tetramer: −10 ± 0.1 nm, 10 ± 0.1 nm, and 23 ± 0.3 nm. Mean displacements per step for dimer: 17.5 ± 0.4 nm; trimer: 14.4 ± 0.4 nm; tetramer: 13.2 ± 0.5 nm.

To further characterize the behaviour of the M6DI8162R constructs, we analysed motor trajectories at high resolution using evanescent darkfield gold nanoparticle tracking[15]. Individual steps were observable in M6DI8162R~DIM, TRI, and TET trajectories (Fig. 1d). Step size distributions depend on the oligomerization state (Fig. 1e); trimers and tetramers show increased frequencies of backsteps and short steps relative to dimers. Overall, trimers and tetramers have slightly shorter mean displacements per step, but many more steps per diffusive encounter.

Shortening the lever arms of our dimers, trimers, and tetramers results in substantially reduced run lengths. M6DI8161R~DIM, M6DI8161R~TRI, and M6DI8161R~TET (Fig. 2a) were generated by truncating the lever arms of the M6DI8162R constructs to only one spectrin repeat (1R). Only M6DI8161R~TET yields measurably processive runs, and these runs are observable only at low [ATP] (Fig. 2a, b). Engineered myosins with shortened lever arms have displayed low processivities in a number of previous studies[16,17], including our work on controllable bidirectional myosins[18]. Possible causes include inhibition of rebinding due to increased strain in 2-head-bound states[16,17,19], aberrant gating due to high and/or off-axis strain, and inaccessibility of preferred off-axis sites due to surface-bound filaments.

Figure 2

Effects of increased flexibility on myosin VI constructs with short lever arms. Single molecule fluorescence trajectories for dimers (red), trimers (green), and tetramers (blue) were recorded at 2mM ATP or (*) 50 μM, showing differences due to addition of a slack element (~1R~) or insertion of a flexible hinge (~).

To address possible causes of reduced processivity, we created modified M6DI8161R~ constructs in which we added a “slack” element consisting of a spectrin repeat flanked with flexible (Gly-Ser-Gly)4 hinges on both sides(Fig. 2). Without extending the rigid lever arm, this slack element is expected to provide increased conformational freedom, reduce internal strain, and provide easier access to potential binding sites. The additional compliance should abrogate strain-mediated gating, which could reduce processivity if gating is beneficial but may increase processivity if unproductive gating is present. The addition of slack results in dramatic run length increases relative to the original M6DI8161R constructs (Fig. 2, Table 1). M6DI8161R~1R~TET exhibits a mean run length of =6.87 ± 0.80 μm (mean ± s.e.m.) at 2 mM ATP. In comparison, M6DI8161R~TET is not processive at 2 mM ATP, and has a mean run length of only =0.24 ± 0.02 μm at 50 μM ATP.

Table 1

Effects of increased flexibility on myosin VI constructs with short lever arms.

Construct	Run Length, <L> (μm)
M6DI8161R~DIM	not measurable (*)
M6DI8161R~1R~DIM	0.64 ± 0.05 (*)
M6DI8161R~TRI	not measurable
M6DI8161R~1R~TRI	2.20 ± 0.14
M6DI8161R~TET	not measurable
M6DI8161R~1R~TET	6.87 ± 0.80
M6DI8161R~TRI	not measurable
M6DI816~1R~TRI	0.44 ± 0.05
M6DI8161R~TET	0.24 ± 0.02 (*)
M6DI816~1R~TET	7.66 ± 0.35 (*)

Mean run lengths are compared between 1R and 1R~1R (top) and between 1R and ~1R (bottom) for single molecule trajectories (Fig. 2) acquired at 2mM or (*) 50 μM ATP. Data are shown as mean ± s.e.m.

We created additional constructs to verify that the beneficial effect of additional compliance does not depend on having more total spectrin repeats in the motor. M6DI816~1R~TRI and M6DI816~1R~TET (Fig. 2, Table 1) differ from M6DI8161R~TRI and M6DI8161R~TET only via the insertion of a (Gly-Ser-Gly)4 hinge (~) between the motor domain and lever arm. Both constructs have much longer mean run lengths than their M6DI8161R counterparts (Fig. 2, Table 1) (Supplementary Movie 3).

We tested the generalizability of our multimerization and slack strategies by constructing engineered motors based on myosin XI, a class with properties that differ significantly from myosin VI. Myosins in this fast andplus-end directed class[20,21] are found in plants and algae and are responsible for transport of golgi stacks, peroxisomes, and mitochondria[22], and for cytoplasmic streaming[23]. Processive motion of wild-type myosin XI dimers purified from Nicotiana tabacum has been recorded[21] at 7 μm/s, the fastest previously observed processive movement of a cytoskeletal motor. No processive behaviour has been reported for other native myosin XI complexes (such as the well-studied Chara corallina motor, which is reported to have a low duty ratio of ~0.34–0.49[20]) or for any recombinantly produced myosin XI protein.

We created engineered myosin XI motors that lack any rigidly attached lever arms, fusing only our slack element ~1R~ and engineered coiled coil domains (DIM, TRI, TET) to the catalytic domain (CD) of myosin XI from Nicotiana tabacum (NM11) or Chara corallina (CM11). We measured processive motion for NM11CD738~1R~DIM, TRI, and TET, and CM11CD746~1R~DIM and TET (Fig. 3a). As seen with our myosin VI constructs, processivity increased with oligomerization state (Fig. 3b). These motors move substantially faster than the myosin VI multimers (Fig. 3c) (Supplementary Movie 4).

Figure 3

Characterization of engineered myosin XI motors. (a) Cartoons and block diagrams for constructs using the Nicotiana tabacum (light brown) and Chara corallina myosin XI (dark brown) catalytic domains. (b) Run lengths and (c) velocities are shown (mean ± s.e.m.) as a function of [ATP].

Unexpectedly, a comparison of the closely related CM11CD746~1R~ and NM11CD738~1R~ constructs also illuminates specialized properties of the Chara motor. At most ATP concentrations, CM11CD746~1R~ constructs are both faster and more processive than their NM11CD738~1R~ counterparts. When comparing myosin classes, there is often a trade-off between processivity and speed[24], which may be understood on the basis of simple kinetic competition (Supplementary Modeling): a short-lived actin-bound state favours high velocities, but (because detachment of the bound head competes with attachment of the free head) disfavours processivity. The simultaneous advantage in speed and processivity of Chara M11 over Nicotiana M11 may be related to the unusual charge distribution in Chara myosin actin binding loops[25], which has previously been proposed to enable high speed without sacrificing actin affinity.

Having developed a general approach for engineering processivity, we applied it to the design of two processive motors with two different properties desirable for potential applications: (i) controllable bidirectionality and (ii) high speed. For both design targets, we chose to make tetramers with slack elements to achieve maximum processivity while counteracting potential detrimental effects from short lever arm lengths.

We previously engineered a chemically controlled bidirectional myosin that switches direction in response to calcium concentration[18]. MCaR-2IQ-DIM exhibits processive motion in both (+) and (−) directions. However, this construct is not measurably processive at saturating ATP, and shows only moderate run lengths even at 50 μM ATP:=550 nm for (−)-end directed motion and =170 nm for (+)-end directed motion. Here, we created MCaR-2IQ~1R~TET (Fig. 4a) by fusing MCaR-2IQ to a flexible (Gly-Ser-Gly)4 hinge, a spectrin repeat (1R), an additional hinge, and the tetrameric GCN4-pLI coiled coil (TET). MCaR-2IQ~1R~TET is highly processive and has mean run lengths at 2mM ATP that exceed those of MCaR-2IQ-DIM at 50 μM ATP (Fig. 4b). High processivity facilitates new experiments in which individual motors are dynamically switchable via in situ buffer exchange (Fig. 4c) (Supplementary Movies 5A–C).

Figure 4

Characterization of design targets for combining processivity with other desirable characteristics. (a) MCaR-2IQ~1R~TET schematic. (b) MCaR-2IQ~1R~TET is processive at 2mM ATP in both low (EGTA) and high (pCa 4) calcium conditions. Randomly selected motor trajectories illustrate predominantly minus-end directed movement in EGTA (99.4% minus (dark blue), 0.6% plus, N=165) and predominantly plus-end directed movement in pCa 4 (83.3% plus (light blue), 2.1% mixed (gray), and 14.6% minus (dark blue), N=96). Minus-end directed run lengths in EGTA and plus-end directed run lengths in pCa 4 are displayed as mean ± s.e.m. Filament orientations were determined using M6DI8162R~DIM as a minus-end directed control18. Average velocities were =8 ± 1 nm/s for minus-end directed runs in EGTA and =1 ± 0.1 nm/s for plus-end directed runs in pCa4 (mean ± s.e.m.). (c) Dynamic switching of individual motors was accomplished using in situ buffer exchange. Duration of buffer exchange is approximate. (d) CM11CD7462R~1R~TET schematic. (e) CM11CD7462R~1R~TET is processive over a range of [ATP], processive velocities up to =10.0 ± 0.3 μm s−1. Velocity and run length data are shown as mean ± s.e.m.

Chara corallina myosin XI is the fastest known cytoskeletal motor[26]. Our results (Fig. 3) showed that Chara myosin can be incorporated into processive tetramers, although no processive behaviour has previously been reported. We reasoned that if we could increase the speed of our processive Chara constructs by adding a rigidly attached lever arm, we might achieve velocities exceeding any known processive cytoskeletal motors. Building on previous work in which the lever arm of the Chara myosin was successfully replaced with two spectrin repeats (2R)[27], we generated a tetrameric construct containing a 2R lever arm and a ~1R~ slack element. CM11CD7462R~1R~TET (Fig. 4d) remains measurably processive at velocities up to =10 μm/s (Fig. 4e) (Supplementary Movie 6). To our knowledge, this is the fastest processive speed measured for a cytoskeletal motor.

Previous work on multimotor cargo transport has established that dimeric processive motors have longer run lengths when assembled into teams of two or more dimers[28-32]. Using bead assemblies[29,33] or DNA scaffolds[28,30,31], this effect has been observed for kinesin-1[28,29,31,32], myosin V[30], and dynein[28,33]. Here, we have shown that engineered monomers can be assembled into processive complexes; fine-tuned coordination and native dimer architectures are disposable[9,10], and effective processivity enhancement can be achieved by forming multiheaded constructs and introducing flexible spacers designed to increase diffusive freedom while abrogating internal tension. Quantitative improvements in processivity, while smaller than seen for a toy model that ignores interference effects and other details (Supplementary Figure 3), are larger than has generally been reported forcomparable multimotor assemblies: for both M6DI8162R and CM11CD746 constructs, the tetramer is ~10X more processive than the dimer, which may be compared with a ~2X improvement in processivity for coupling two myosin V dimers together using a DNA scaffold[30]. Our measurements complement the recent discovery of processive motion in small defined assemblies of kinesin-14 dimers, which are non-processive in isolation[32].

This work provides a general approach for increasing the processivities of engineered myosins in synthetic biology or nanotechnology applications. The modularity of the approach allows high processivity to be combined with complex engineered properties such as directionality control, or with higher velocities than are normally associated with processive motors. Engineered complexes are simply and efficiently assembled (Supplementary Figure 5) upon expression of a single polypeptide chain, and should be deployable in vivo for modifying the parameters of cargo transport processes[34]. Engineered motors may also be used for in vitro imitations of these cellular systems, which achieve effective long-range transport through a combination of processive stepping, high filament densities (to favour rebinding after dissociation), and high motor velocities[35]. In diagnostic device designs that use molecular motors to overcome diffusion limits for sensitive detection[36] or perform molecular sorting tasks[37], long-range transport has often been implemented using filament-based cargo carriers[2]. New processive motors will enhance the capabilities of alternative designs using motor-based cargo carriers[38], allowing for further miniaturization and molecular customization and control.

METHODS

Proteins

DNA constructs were assembled from modules encoding porcine myosin VI (residues 1–816), Chara corallina myosin XI (residues 1–746), Nicotiana tabacum myosin XI (residues 1–738), Dictyostelium α-actinin (residues 266–388 for 1R and residues 266–502 for 2R), and engineered GCN4 leucine zipper variants pIL, pII, and pLI[14] (Supplementary Figure 4). All constructs contained a C-terminal flexible linker, HaloTag protein sequence (Promega), and flag tag (DYKDDDDK) and were cloned into pBiEx-1 (Novagen). Proteins were expressed via direct transfection of SF9 cells and purified and labeled with TMR HaloTag ligand (Promega) or PEG-biotin HaloTag ligand (Promega) as previously described[9,10].

Single-molecule fluorescence tracking assays

Assays were carried out in Assay Buffer (AB) containing 25 mM imidazole, 25 mM KCl, 1 mM EGTA, and MgCl2. The concentration of MgCl2 added varied depending on final assay ATP concentration to give 2mM free Mg2+, according to MAXCHELATOR (http://maxchelator.stanford.edu). Calmodulin (4μM) was added when assaying myosin VI constructs. Blocking buffer consisted of AB with 1 mg/mL BSA (Sigma). Imaging buffer consisted of AB with the desired ATP concentration, 1 mM phosphocreatine, 100 μg/mL creatine phosphokinase, 0.4% glucose, 0.2mg/mL glucose oxidase, 36 μg/mL catalase, and 1.8mM trolox. MCaR-2IQ~1R~TET constructs were assayed in buffers with 2 μM calmodulin and either pCa 4 or EGTA as described[18], with MgCl2 added to yield 2 mM free Mg2+upon addition of 2 mM ATP in imaging buffer. Assays were performed essentially as described[10], except that plasma-cleaned coverslips were used without any nitrocellulose coating. Briefly, N-ethyl maleimide inactivated full-length skeletal muscle myosin (a gift from R. Cooke) was used to immobilize actin filaments stabilized with excess Alexa 633 phalloidin (Invitrogen). Nonspecific binding was minimized by incubation with blocking buffer. Motors were added in imaging buffer. Videomicroscopy data were collected on an electron multiplying charge coupled device camera (Andor) under objective-side total internal reflection fluorescence excitation using a 532 nm diode-pumped solid-state laser (Coherent) through a 1.49 NA 100x objective (Nikon). Actin filaments were imaged prior to recording assay videos with a separate emission filter set. Motor trajectories were tracked using custom software as described[10]. To avoid artefacts due to undersampling of short runs[10], an analysis cut-off was used: only trajectories lasting at least three frames and traveling more thantwo pixels (216 nm) were scored for velocity, run length, and directionality. For the high-velocity motorCM11CD7462R~1R~TET assayed at 100 μM and 500 μM ATP, the cut-off was increased to three pixels (324 nm) (Supplementary Table 2). Mean run lengths were corrected for runs that terminated at the end of a filament using the Kaplan-Meier survivor function, and the reported errors of the mean (s.e.m.) are the error estimates from this function (Supplementary Figure 6). Constructs with mean run lengths greater than two s.e.m. above one pixel (108 nm) were deemed “measurably processive”. “Mixed” runs for MCaR-2IQ~1R~TET were defined as runs that showed >216 nm movement in one direction and subsequently >216 nm in the other direction. Velocity error estimates were calculated as described[10], except that s.e.m. weighting was not used and the number of sets used in bootstrapping was equal to the number of runs scored. Fit error estimates for model fit parameters were calculated as described[10]. A summary of run length and velocity data can be found in Supplementary Table 1.

Gold nanoparticle tracking assays

NeutrAvidin-coated 50 nm gold nanospheres (Nanopartz) were washed into binding buffer (500 mM NaCl, 40 mM Tris pH 7.5, 0.2% (v/v) tween-20 (Sigma), and 1 mg/mL BSA) by centrifugation and resuspension. PEG-biotin labeled proteins were incubated with washed beads for 15 min at 25°C before dilution into imaging buffer. Assays were carried out as for single-molecule fluorescence tracking assays above with the addition of 0.2% (v/v) tween-20 in the blocking buffer and 1 mg/mL BSA and 0.2% tween-20 (v/v) in the imaging buffer, which contained 50 μM ATP. Gold particles were imaged using objective-side evanescent darkfield microscopy. Laser illumination was provided by a 532 nm diode-pumped solid-state laser (Spectra-Physics) through a 1.49 NA 60x objective (Nikon) and totally internally reflected at the sample interface; additional details of the microscope design will be presented elsewhere (Lebel et al., submitted). Scattered light was imaged on a high-speed CMOS camera (Mikrotron Ecosens CL) at 500 Hz. Analysis details are provided in Supplementary Methods.