The sequence of the ~450 residues immediately following the catalytic head and the known lever arm of M6 can be divided into four regions (Fig. 1 and Supplementary Fig. 1 online). The characteristic heptad repeat pattern of a coiled coil with hydrophobic residues dominating the a and d positions is identifiable only in the first ~70 residues, which we call the proximal tail (PT, Pro835–Glu907; numbering is based on the human sequence). A transition to the second domain, also ~70 residues, occurs somewhere between Glu907 and Gln915. After Gln915, an extended run of alternating charge appears with only a few hydrophobic residues, the medial tail (MT, Glu908–Arg980) (Fig. 1). This alternating-charge pattern abruptly ends at residue Arg980, and the C-terminal tail sequence of ~50 residues just before the cargo binding domain has no obvious primary-sequence motifs; we call this region the distal tail (DT, Ile981–Arg1035). We refer to the fourth and final region of ~250 residues as the cargo binding domain (CBD, Arg1036–Lys1285).

The PT has been proposed to be a random coil acting as the flexible element in the M6 tail that spans the remaining ~24-nm gap20. To explore this model, we created three M6 mutants with altered PT domains (Supplementary Fig. 2a online). The three constructs changed the sequence to varying degrees ranging from point mutations that removed nine core hydrophobic residues, proposed to be a and d positions on the basis of coiled-coil prediction algorithms26 (altered A&D), to randomization of up to 58 residues (850–907 random and 857–907 random). These changes were not expected to impair function if the proximal tail is a random coil, and hence these mutants would be expected to show the same stepping behavior of the control M6 dimer1; however, they would be expected to disrupt any organized tertiary structure. By labeling the motors with fluorophores, the motion of single molecules along actin filaments on the surface can be observed in a total internal reflection fluorescence (TIRF) microscope, allowing the rate of motion to be determined. Each of the three mutant constructs showed velocities in TIRF motility assays at 80 μM ATP that were only ~50% of that of the control M6 dimer (Supplementary Fig. 2b,c). Because velocities can be altered by changing either step size or the rate of stepping, detailed optical trap assays were undertaken for one of the mutants. In this case, two beads are attached to an actin filament, one at each end, and held in two separate optical traps. The filament is then positioned above a single motor on the surface, and individual interactions between the motor and the filament are observed. Processive motors produce a characteristic staircase pattern where the displacement between plateaus measures the step size and the duration of the dwell measures the rate of stepping. These assays showed that the altered velocities result from a shortened step size and not altered stepping rates (Supplementary Fig. 2d,e). Therefore, the model of the PT as a random coil is unlikely.

CD spectroscopy was used to determine what secondary-structure elements were present in the PT. CD revealed that the PT has a classic α-helical spectrum with minima at 222 nm and 208 nm and a maximum at 194 nm (Fig. 2a). We estimated that ~80% of the residues in the PT are in an α-helical conformation. The strong minimum at 222 nm was measured as a function of temperature to track the stability of the PT. It unfolds in a cooperative manner characteristic of compact proteins, where tertiary contacts stabilize the secondary structure (Fig. 2a). These results indicate that the PT is a compact domain with a defined structure. This conclusion agrees with the hydrodynamic radius determination from gel filtration, which is that predicted for a spherical compact protein27 (Table 1). An estimate of the PT structure (residues 835–915, in the human sequence) from the Rosetta 2.2.0 structure prediction algorithm28 consistently yielded (7 out of 10 runs) a three-helix bundle with an end-to-end distance of ~3 nm, which is in good agreement with all experimental data and is considered to be a good estimate of the PT structure (Supplementary Fig. 3a,b online).

Two previous models have been put forward for the MTof M6. First, it has been suggested that at high concentrations M6 dimerizes via coiled-coil formation25. Second, M6 contains a charge-repeat pattern homologous to a domain in myosin X that has been shown to be a single α-helix in solution21, suggesting that M6 may take the same conformation. To test these models, the MT and DT domains were expressed together in Escherichia coli and purified to homogeneity. A clear difference between the two models is that the coiled coil would be dimeric and the single α-helix would be monomeric. Multiple-angle light scattering (MALS) is capable of determining molecular weights in solution by measuring scattering as a function of angle while simultaneously measuring the protein concentration. Using MALS, we found that the MT-DT is monomeric, consistent with the single α-helix hypothesis. This was corroborated by dynamic light scattering, which measures light scattering as a function of time to determine the hydrodynamic radii (R_h). The measured R_h of 3.6 nm is consistent with an extended structure rather than a globular one (1.8 nm predicted27; Table 1). Furthermore, the CD spectrum and thermal melt curves of the MT-DT are those expected of a single α-helix, with a strongly α-helical spectrum and a broad thermal unfolding transition spanning ~70 °C, as shown for charged synthetic peptides29 (Fig. 2b). To directly assess the global structure of the domains, we recorded small-angle X-ray scattering (SAXS) profiles and used ab initio reconstructions to develop low-resolution structural models of the MT-DT construct. The reconstruction for MT-DT (Fig. 3a), as well as that for the PT-DT construct (Fig. 3b), shows a long, bent, slender cylinder, similar in length to that expected with the 70-residue MT configured as a single α-helix.

Although the MT-DT is monomeric under these conditions, it is possible that it was not assayed at a high enough concentration to dimerize. To increase the effective concentration, a Gly-Gly-Cys sequence was placed at the C terminus of the natively cysteine-free MT-DT construct, so that two copies could be held in close proximity by a specific disulfide bridge. This mechanism, which has been used in the study of the GCN4 coiled coil30, does not inhibit coiled-coil formation and results in an apparent concentration of at least 4 mM31. The CD spectrum and melt profile of this oxidized dimer were essentially identical to that of the MT-DT monomer (Fig. 2b), showing that even under these conditions the MT-DT is unable to form a dimeric coiled coil.

The above results are consistent with the MT providing the reach needed for the 36-nm step of M6. If the MT-DT domain is the long element that allows the two heads of M6 to separate during processive stepping, removing that reach will alter that processive stepping. A GCN4 sequence was placed just beyond the PT-MT boundary at residue 920 (Fig. 1, MT locked). This locks the two copies of M6 together in a coiled coil at this position and eliminates any contribution of the MT to processive stepping. The impact on motion was tested in gliding filament assays, where fluorescently labeled actin filaments glide across a motor-covered surface. This geometry is useful as it can assay both processive and nonprocessive motors, whereas the TIRF motility assay can assay only processive motors. Control M6 dimer at 2 mM ATP moved at 110 ± 21 nm s⁻¹, and MT locked moved at 75 ± 26 nm s⁻¹, significantly different results as assessed by a t-test (n = 20 per group, P-value > 0.001). The 75 nm s⁻¹ velocity is that expected for a stroke size of ~20 nm, which derives from the ~180° swing of the lever arm extended by the globular PT.

Gliding filament assays are capable of testing more than just the velocity of a motor. Other parameters, such as the rate at which filaments land on the surface and the fraction of filaments that move greater than their length, can also be counted. The theoretical framework of the landing and continuous movement assays allows for quantitative predictions of the landing rate and the fraction of filaments that will move greater than their length as a function of motor density and the number of motors it takes to attach a filament to the surface32. We created models in which either one motor (processive movement) or two motors (nonprocessive movement) were needed for filament attachment. Data from the control M6 dimer were more consistent with the processive model, whereas the MT locked construct was more consistent with the nonprocessive model in both assays (Fig. 4a,b). Optical trap assays also indicate that a single motor of MT locked is incapable of producing the characteristic staircases of a processive motor. Instead, single bind-and-release events, characteristic of nonprocessive motors, were seen with a mean displacement of 23 ± 26 nm (Fig. 4c) and kinetics similar to the control M6 dimer with a mean dwell of 0.9 ± 0.02 s. From these data, we conclude that the MT is the primary element in M6 that allows the heads to sample a large space and processively step along the 36-nm pseudorepeat of the actin filament.

Neither the PT nor the MT dimerize at ~1 μM (Table 1). To further explore M6 dimerization, we examined the role of the CBD. Indeed, MALS analysis showed that a construct extending from the beginning of the PT to the end of the CBD is dimeric at concentrations in the micromolar range, which indicates reasonable affinity (Table 1). Because the PT and the MT-DT do not self associate at these concentrations, we conclude that the CBD is needed for this dimerization. However, under the same conditions, the full-length M6 is monomeric (Table 1), in agreement with previous work22. From these results, we conclude that the CBD dimerization must be inhibited by the M6 head. This is corroborated by the compactness of the SAXS envelope of full-length M6 (Fig. 5). If the model for the PT-MT (Fig. 3b) is fused to the structure of the M6 head and a globular CBD, the resultant full-length M6 monomer cannot fit into the experimental SAXS envelope without a high degree of bending (Fig. 5). This implies that the full-length M6 monomer is in a folded-back state, with an interaction occurring between the CBD and the M6 head, and that this interaction inhibits dimerization. It is likely that the binding of cargo to the CBD helps regulate the equilibrium between this folded-back monomeric state and the dimeric form of M6, especially given the orientation of the CBD when bound to the membrane23 and the observation that M6 dimerizes when binding to membrane24.

Further details for all of the methods are included in the Supplementary Methods online. The temperature was 22 °C, unless otherwise noted. The appropriate sequences from the human M6 cDNA were cloned into a modified pET28a vector (EMD Chemicals). The modified vector contained a hexahistidine (His₆) tag, maltose binding protein (MBP) and a tobacco etch virus (TEV) protease cleavage site 5′ to the cloning site. Proteins were expressed in E. coli Rosetta (DE3) cells (EMD Chemicals) at room temperature, with all subsequent purification steps being conducted at 4 °C. Cells were lysed by sonication. Lysates were clarified by centrifugation at 100,000g for 30 min and bound to nickel–nitrilotriacetic acid (Ni-NTA) resin (Qiagen). The eluted proteins were dialyzed in the presence of TEV protease (1:100 by weight) overnight to cleave the His₆ and MBP portions from the M6 domains. M6 domains were further purified by running the cleavage reaction over a Ni-NTA column to remove His₆-containing fragments, running a MonoQ column (GE Healthcare) and running a Superdex 200 column (GE Healthcare) in either CD buffer (10 mM phosphate, pH 7.4, 25 mM NaCl) or scattering buffer (10 mM phosphate, pH 7.4, 150 mM NaCl).

The porcine control M6 dimer construct described previously40 was modified to create the mutants, and full-length M6 was made using the human M6 cDNA. M6 altered A&D, 850–907 random, and 857–907 random sequences were synthesized by DNA 2.0 (Menlo Park). The altered protein sequences are described in Supplementary Figure 2. M6 MT locked was made with a GCN4 sequence inserted in place of residues 919–950 of the control dimer. All sequences were placed under the control of the polyhedron promoter of the pFastBac Dual vector (Invitrogen), with sea urchin calmodulin (P05934) under the control of the p10 promoter. Recombinant baculoviruses were created as per the Invitrogen protocol. Proteins were purified as described previously40. Briefly, Sf-9 cells were infected with recombinant virus at 28 °C, with all subsequent purification steps being conducted at 4 °C. Cell lysates were clarified by centrifugation for 1 h at 200,000g, and proteins were bound to Flag M2–affinity gel (Sigma) and eluted with Flag peptide. For scattering analysis, full-length M6 was gel filtered using a Superdex 200 column in scattering buffer. For all constructs, absorbance at 280 nm in 6 M guanidinium chloride was used to determine the protein concentration.

CD spectra were acquired using an Aviv 62DS instrument (Aviv Biomedical) with a 1-mm path-length cell in CD buffer. Spectra were taken at 10 °C, with data collected every 1 nm with a 20-s averaging time; results are the average of 3 repeat scans. Concentrations ranged from 1–17 μM. Melt data was collected every 1 °C with a 30-s averaging time and a 2-min equilibration. In all cases, the reverse melt showed at least 90% reversibility. The percent helical content was determined as described previously41.

In solution, molecular weights were determined using a size-exclusion chromatography system coupled to a MALS detector. Protein concentrations were determined with an Optilab rEX refractive index detector, and scattering was detected with a Dawn 18 angle MALS light-scattering instrument (Wyatt Technology Corporation).

Protein samples at ~10 μM in CD buffer were mixed in a 2:1 ratio with sinapinic acid and spotted onto a stainless steel MALDI plate. Analysis was conducted on a Voyager-DE RP (Applied Biosystems) instrument in positive linear mode.

Dynamic light scattering measurements were made using a DynaPro instrument (Protein Solutions) running Dynamics version 6 software. Samples at 1–10 mg ml⁻¹ in scattering buffer were assayed at 25 °C with an acquisition interval of 10 s. Samples were spun at ~15,000g for 10 min immediately before analysis. Results were derived using a regularization fit and with PBS buffer settings.

Proteins were concentrated to 10–300 μM and loaded onto a 25-ml Superdex 200 equilibrated in scattering buffer at 4 °C. The elution volume was determined from the average of at least three runs and converted to R_h.

SAXS measurements were carried out at the XOR/BESSRC undulator beamline 12-ID of the Advanced Photon Source, using a sample-detector distance of 2 m and CCD detector readout (MAR USA). The data were collected using a custom-made sample cell42 at an X-ray energy of 12 keV. Details of the beamline are as described previously42-44.

Samples were centrifuged at 10,000g for 10 min before measurement, and three 0.5-s exposures were obtained. Data were image corrected, normalized by incident flux and circularly averaged. The three profiles for each condition were averaged to improve signal quality. Buffer profiles were collected using identical procedures and subtracted for background correction, and the data showed no signs of radiation damage, based on comparison of consecutive scattering profiles from the same sample (data not shown).

Scattering intensities as a function of the momentum transfer were obtained at different protein concentrations. The SAXS profiles for the MT-DT, PT-DT and full-length M6 constructs are superimposable after scaling by forward-scattering intensity, suggesting that there are no detectable aggregation or interparticle interference effects (Supplementary Fig. 4 online). Radii of gyration were determined from Guinier analysis of the small momentum transfer scattering data45 (Supplementary Fig. 5 online). The radii of gyration obtained from Guinier analysis agree within experimental errors with the values from the real-space distribution function P(r) computed using the regularized transform method implemented in the program GNOM46.

The programs DAM-MIN47 and GASBOR48 were used to construct three-dimensional bead models that fit the scattering data (Supplementary Fig. 6 online). Both programs use a simulated annealing procedure and a compactness criterion. We performed ten independent DAMMIN and GASBOR runs for each scattering profile, using default parameters, the ‘slow’ mode for DAMMIN, no symmetry assumptions (P1 symmetry) and the full recorded scattering profiles. The models resulting from independent runs were superimposed and compared using the program SUPCOMB49 based on the normalized spatial discrepancy (NSD) criterion. Models with NSD values <1 are considered similar. For all data presented in the main text, the ten independent repeat runs yielded models with pairwise NSD values <1, indicating that the algorithms converged reproducibly to similar structures. The ten independent structures for each scattering profile were subsequently averaged, and ‘filtered’ consensus models were computed using the program DAMAVER with default settings50. Consensus models constructed with DAMMIN and GASBOR gave similar results. For visualization, the reconstructed bead models were converted to electron-density maps with the program Situs51.

Assays were conducted as described previously32. Briefly, motors were attached with monoclonal anti-GFP antibody (Chemicon MAB3580) to nitrocellulose-coated cover slips. Tetramethyl rhodamine iso-thiocyanate (TRITC) phalloidin-labeled actin was observed using a total internal reflection microscope52. Movies were scored for the number of filaments landing and moving more than 0.5 μm, the fraction of filaments running greater than their length and the velocity with which filaments moved.

Assays were conducted as described previously52. Briefly, actin biotinylated at 1 in approximately every 25 monomers was attached to a coverslip using steptavidin and nonspecifically absorbed biotinylated BSA. Motors were labeled at a 1:1 ratio with anti-GFP antibody conjugated with multiple Cy3 fluorophores to enhance signal were assayed at 80 μM ATP. Movies were collected on the microscope described in ref. 52 and analyzed using imageJ (National Institutes of Health).

Assays were conducted as described previously3,53 with a few exceptions, notably that trapping was done without feedback. Briefly, motors were attached with monoclonal anti-GFP antibody (Chemicon, MAB3580) to nitrocellulose-coated cover slips. Actin dumbbells were formed between two trapping beads and pulled taught. Positional data was collected at 10 kHz and trap stiffness ranged from 0.006–0.012 pN nm⁻¹. Binding events were determined by eye using a drop in the positional variance of the beads and the bead-to-bead . Transitions between processive steps were also scored by eye.