A programmable DNA origami nanospring that reveals force-induced adjacent binding of myosin VI heads

Mechanosensitive biological nanomachines such as motor proteins and ion channels regulate diverse cellular behaviour. Combined optical trapping with single-molecule fluorescence imaging provides a powerful methodology to clearly characterize the mechanoresponse, structural dynamics and stability of such nanomachines. However, this system requires complicated experimental geometry, preparation and optics, and is limited by low data-acquisition efficiency. Here we develop a programmable DNA origami nanospring that overcomes these issues. We apply our nanospring to human myosin VI, a mechanosensory motor protein, and demonstrate nanometre-precision single-molecule fluorescence imaging of the individual motor domains (heads) under force. We observe force-induced transitions of myosin VI heads from non-adjacent to adjacent binding, which correspond to adapted roles for low-load and high-load transport, respectively. Our technique extends single-molecule studies under force and clarifies the effect of force on biological processes.

M echanical forces regulate the function of numerous mechanosensory protein molecules including motor proteins 1 , ion channels 2 , cytoskeletal filaments 3 and cadherins 4 . These molecules are involved in essential cellular behaviours such as cell growth, differentiation, shape formation and cell death 5 . Optical trapping (OT) provides a powerful methodology to explore the molecular mechanics and mechanoresponse of these biomolecules at the single-molecule level 1,3,4,[6][7][8] . OT functions by applying force to nano-sized molecules via optically trapped particles (generally, 0.2B1 mm in diameter). Although displacements with nanometre accuracy can be acquired using OT, it is difficult to determine intramolecular or subdomain motion within the molecule. Instead, single-molecule fluorescence (SMF) imaging provides a powerful method to monitor structural dynamics within biomolecules by attaching fluorophores to the subdomains [9][10][11] . Therefore, simultaneous OT and SMF has been developed [12][13][14][15] , but it is technically difficult to apply due to a number of factors caused by large-scale devices that use a high-power infrared beam to trap microspheres including low throughput (that is, only one biomolecule can be analysed at a time) and complicated experimental geometry, preparation and optics that minimize dye photobleaching induced by the infrared trapping beam. Furthermore, the use of fluorescent quantum dots (QDs) or gold nanoparticles (GNPs) to circumvent dye bleaching issues is problematic due to their attraction to the trapping beam's focus point.
To overcome these issues, we developed a nanoscale forceapplication device using three-dimensional DNA origami 16,17 . Our device is a programmable DNA nanostructure with a coil shape (nanospring) and we applied it to study the mechanoresponse of the actin-based processive motor protein, myosin VI. Myosin VI is a ubiquitously expressed mechanosensitive dimeric motor protein, which is thought to act as an anchor to maintain the structural integrity of many cellular structures such as stereocilia 18 and the Golgi apparatus 19 . In addition, myosin VI is a vesicle transporter that moves cargo along cytoskeletal actin filaments coupled to the hydrolysis of ATP 20,21 . It has been proposed that mechanical forces cause myosin VI to switch between these roles of anchor and transporter by slowing its ATP hydrolysis cycle at the strongly bound state with actin 20,22 , as well as causing a transition from hand-over-hand to inchworm-like stepping 23 . In inchworm-like stepping, step sizes of the motor domains (or heads) are 44 nm, suggesting the rear head becomes the lead head and forms adjacent binding state (that is, adjacent motor domains close together), which could better resist detachment due to an even distribution of load between the two heads, and therefore could be seen as an adaptation for anchor function. Therefore, Nishikawa et al. 23 predicted that high load should lead to a higher fraction of adjacent binding states observed. To test this hypothesis, SMF imaging of myosin heads in the presence of load is required; however, combined OT and SMF imaging was difficult to implement for the study of molecular machines; therefore, no direct test has yet been reported.
An ideal nanospring for myosin VI would exhibit a relatively small spring constant at the stall force of the motor under investigation, which in the case of myosin VI, is reported to be B2 pN 20,22 ; otherwise, only very few steps could be tracked close to the stall force. In addition, the nanospring should not be overly long (that is, 4 41 mm) to avoid the requirement of extended motor run lengths that stretch the nanospring; this is problematic, as the average run length before spontaneous detachment of myosin VI is B700 nm (ref. 24). Our programmable nanospring can fully satisfy the desired characteristics unlike natural elastic components such as double-stranded DNA (dsDNA). In addition, SMF imaging of the nanospringbiomolecule complex is superior to simultaneous OT and SMF imaging, in that we can visualize over ten complexes at a time using conventional total internal reflection fluorescence microscopy (TIRFM), and QDs and GNPs are also available. Furthermore, a nanospring can control the number and species of biomolecules in the complex and is compatible with other useful methods such as high-speed atomic force microscopy (AFM) 25 , dark-field microscopy and electron microscopy without complicated optics or geometries.
Here we demonstrate SMF imaging of single-and doublecolour QDs attached to the heads of myosin VI dimers tagged with our nanospring and also the tug-of-war between myosin VI and myosin V (opposite directional motors). We find force and ADP synergistically trigger transitions from hand-over-hand to inchworm-like movement of myosin VI and stalling in the adjacent binding state, thus providing insights into the mechanism of the physiological dual function of myosin VI as a vesicle transporter and anchor.

Results
Design and mechanical properties of nanospring. Our nanospring consists of a two-helix bundle programmed with negative superhelical strain to form a coil structure ( Fig. 1a and Supplementary Figs 1 and 2). To construct the nanospring, 7,308-nucleotide single-stranded DNA (ssDNA; scaffold) and 4150 species of short (o50 nucleotide) ssDNA (staples; Supplementary Data) were self-assembled by hybridization under rapid heating followed by gradual cooling. To attach specific protein molecules to the ends of the nanospring, extra ssDNA handles were projected out from each end. The length or number of bundles and turns of the nanospring is programmable ( Supplementary Fig. 2b), resulting in control of the spring constant.
To estimate the force-extension curve, we applied force to the nanospring using OT ( Fig. 1b and Supplementary Fig. 3). We attached a fluorescent polystyrene bead at the end of the nanospring and optically trapped it. The other end was fixed to a glass slide and moved back and forth ( Fig. 1c and Supplementary Movie 1) to obtain a force-extension curve (Fig. 1d). The curve was reversible between stretching and relaxing phases ( Supplementary Fig. 3f), and the change in the slope (spring constant) was more gradual than that of dsDNA over a range of several pN. We also confirmed the force-extension curve was not affected by the rate of stretching ( Supplementary  Fig. 3g). Our two-helix bundle nanospring exhibited a spring constant of 0.012 ± 0.002 pN nm À 1 (mean ± s.d., N ¼ 19) and a length of 841 nm at 2 pN of force (contour length of 1.1 mm), which were estimated from the force-extension fit curve (Fig. 1d). For myosin VI, which takes a 36 nm step 20 , these properties increases the force by 0.43 pN, which is a moderate change in force and useful for facilitating better resolution of the forcevelocity relation near the stall force. In contrast, a dsDNA spring of the same length would be far less sensitive. At 2 pN of force, a 1.1 mm contour length of dsDNA would exhibit a spring constant of B0.09 pN nm À 1 , or roughly an order of magnitude larger 26 . To achieve a comparably compliant spring would require a dsDNA of length B5 mm (ref. 26), which would then impose a limitation of analysis to run lengths of that magnitude, which are very rare (B0.08%) 24 .
Nanospring dynamics driven by a single myosin VI. Myosin VI is composed of an amino-terminal motor domain (head) and tail domain. The tail consists of four subdomains: the proximal, medial and distal tails and the carboxy-terminal cargo-binding ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13715 domain 27 . The proximal and medial tails can potentially act as a lever arm that contributes to force generation and the cargo-binding domain regulates dimerization 24,27 . Myosin VI is thought to function as a dimer in cells 24 and adaptor proteins at the cargo-binding domain are necessary for regulating the equilibrium between the monomeric and dimeric forms 28 . In the present study, to make a stable myosin VI dimer, we engineered a myosin VI monomer fusion protein in which the cargo-binding domain was replaced with SNAP-tag, whereas preserving all other myosin components (Fig. 2a). SNAP-tag can be efficiently conjugated to an oligonucleotide (oligo) that serves as an artificial dimerization domain 29 . The myosin VI monomer was also fused to HaloTag at the N terminus of the head to enable fluorophore labelling. We covalently labelled 21-mer or 42-mer oligos to our myosin VI monomers and confirmed 490% labelling efficiency by a gel-shift assay (Fig. 2b). As the 42-mer oligo contains a complementary sequence to the 21-mer oligo, the mixture showed a dimer fraction due to hybridization. The remaining sequence in the 42-mer oligo retains a domain that captures the ssDNA handle in the nanospring (capture domain).
To observe the stepping dynamics of myosin VI under tension, we designed a myosin VI-nanospring complex (myosinVI-NS; Fig. 2c), where one end of the nanospring was attached to a DNA-coupled myosin VI dimer via the nanospring capture domain on the myosin VI 42-mer oligo, whereas the other end of the nanospring was attached to an immobilized myosin II dimer, which was chemically inactivated by N-ethylmaleimide and labelled with ssDNA. Immobile myosin II strongly binds to actin filaments in the presence of ATP and plays a role in fixing myosinVI-NS on the actin filament.
We used TIRFM to observe the dynamics of a fluorescently labelled nanospring driven by a single myosin VI (Fig. 2d). Thirteen unlabelled staples were replaced with fluorescently labelled staples (5 0 -end was labelled with TAMRA; Supplementary Data) that were homogeneously positioned in the nanospring. As the equilibrium length of the nanospring without tension is B300 nm (that is, close to half the wavelength of the fluorescence emission), the fluorescence image is seen as a diffraction-limited circular spot (Fig. 2d, compressed). When myosin VI stretches the nanospring against tension, the shape becomes elliptical (Fig. 2d stretched and Supplementary Movie 2). By fitting the images to a two-dimensional (2D) Gaussian distribution 30 , we were able to determine the centre position and length of the nanospring to an accuracy of 14±2 nm (mean ± s.d., N ¼ 11) when extended longer than 500 nm total spring length ( Fig. 2e and Supplementary Fig. 4). Typically, myosin VI stretched the nanospring B800 nm and stalled. In this state, it eventually detached from the actin and the nanospring returned to the compressed state after the release of tension. This sequence of events was repeatedly observed. We quantified ARTICLE the tension using a force-extension curve (Fig. 1d). From this curve, we estimated the stall force to be 2.1±0.6 pN (mean±s.d., N ¼ 126; Fig. 2f,g), which is in agreement with previous force measurements using conventional optical tweezers 20,22 .
Tug-of-war between myosin V and myosin VI via nanospring. Myosin VI has been proposed to function as an anchor and a transporter, whereas myosin V has been proposed to function exclusively as a transporter 31 . Our DNA nanospring afforded us the opportunity for direct comparison of their motile properties by a tug-of-war between single myosin V and VI ( Fig. 3a and Supplementary Movie 3). As expected, myosin V and VI (indicated by ATTO647N and Cy3 fluorescence, respectively) moved processively along the same actin filament but in opposite directions 32,33 . The nanospring was stretched further and further until one myosin or the other detached from the actin filament. We observed myosin VI to detach less than half as frequently as myosin V (Fig. 3b), consistent with the dual role proposed for myosin VI.
Detection of head motion in myosin VI dimer under load. Next we observed the stepping dynamics of QD-labelled myosin VI in the absence and presence of the nanospring using TIRFM at B2 nm and 30 Hz spatio-temporal resolution 23,30 . In the absence of the nanospring and at saturating ATP concentration (2 mM), our DNA-coupled myosin VI dimer moved processively along actin filaments (Supplementary Fig. 5 and Supplementary Movie 4). The step-size distribution for the myosin head ( Supplementary  Fig. 5b) was fit to a three Gaussian function with peaks at 75 nm (large forward steps), 43 nm (short forward steps) and À 40 nm (backward steps), consistent with previous reports 23, 34 . The distribution of dwell times just before the forward steps ( Supplementary Fig. 5c) were well fit to the formula tk 2 exp( À kt) if assuming two rate-limiting steps having the same rate constant, k. We obtained k ¼ 4.8 s À 1 , which is also consistent with previous reports 20,22,23 . In conclusion, we confirmed the fusion of SNAP-tag and DNA to induce dimerization did not alter the stepping dynamics or coordination between heads. Further, we also examined the stepping dynamics of myosinVI-NS ( Supplementary Fig. 6a  and confirmed the labelling of nanospring also did not alter the motion. When a myosinVI-NS was linked with an actin filament via immobile myosin II, myosin VI exhibited processive walking, stalling and detachment coupled with the extension and retraction of the nanospring (Fig. 3c and Supplementary Movie 5). The QD-myosin VI trajectory (red circles in Fig. 3c) was in agreement with the nanospring displacement (grey line in Fig. 3c) and stall forces calculated from QD displacements ( Supplementary Fig. 6c) were consistent with previous reports 20, 22 and those estimated from nanospring displacements (Fig. 2g). Further, we could clearly measure the step sizes of individual myosin VI heads at 30 Hz recording rate (Fig. 3d-f). In the absence of ADP at 1-2 pN load (Fig. 3g), the stepping behaviour was similar to that in the absence of the nanospring (that is, 0 pN load) and large forward steps (75 nm) were the most frequent step type. However, when we added 100 mM ADP (physiological nucleotide concentration), the most frequent step type changed to short forward steps (43 nm) and the frequency of backward steps ( À 40 nm) slightly increased (Fig. 3h). Step size (nm) Step size (nm) Large forward steps were dominant both with and without 100 mM ADP at no load (Fig. 3d), which is in agreement with a previous report 23 . We found a similar distribution at 1-2 pN load in the absence of ADP. However, the distribution shifted to short forward steps at 1-2 pN load and 100 mM ADP (Fig. 3e), and at stall force (B2 pN) only short forward and backward steps were observed (Fig. 3f). Thus, our results show a transition in the myosin VI step size distribution in response to ADP and load.
Dynamics of both heads of a myosin VI dimer under load. To fully clarify the stepping behaviour, we monitored both heads of a myosin VI dimer by attaching different colour QDs to them. Simultaneous nanometric observation 23,34,35 of the heads allowed us to determine the coordination and distance between them. We confirmed that the large forward steps we observed were caused by a hand-over-hand movement (Fig. 4a, dashed-line box) and the short forward steps were caused by inchworm-like movement 23,34 (Fig. 4a, dashed-line box); therefore, our results show a transition in myosin VI stepping behaviour in response to ADP and load. Further, the results indicated two binding states: the distant binding state, where the two heads span the actin helical pitch (36 nm) when bound to the actin filament (typically observed in hand-over-hand steps), and the adjacent binding state, where the two head are close (o10 nm apart) when bound to the filament (typically observed in inchworm-like steps). These two states are consistent with previous reports 23, 34 . In the stalled state, we observed that myosin VI repeatedly transitioned between the distant binding and adjacent binding states (Fig. 4b, dashed-line box). The probabilities of short forward and backward steps were almost equal (Fig. 3i), similar to the stalled state of the Brownian ratchet, showing myosin VI is not completely locked onto the actin filament at the stalled state. This dynamic binding and release of the head cannot be observed in conventional force measurement systems such as optical tweezers, as these assays monitor the tail position 20,22 . The adjacent binding state contributed more to the stalled state than did the distant binding state (Fig. 4c). This is in contrast to the bound state in the absence of load, where the distant binding state is dominant (Fig. 4d). The adjacent binding state may evenly distribute the external load to both heads, which could reduce the risk of diffusing away from the actin. Supporting this notion, we found only 26% of detachments from actin occurred in the adjacent binding state (12 observations/46 detachment events). Consequently, efficient anchoring of myosin VI can be promoted by both the load dependence of the ATP hydrolysis cycle 20,22 and by the load dependence of the bias towards the adjacent binding state.

Discussion
The combination of our DNA nanospring technology with SMF imaging reveals that higher forces induce a transition from hand-over-hand to inchworm-like motion in myosin VI. Furthermore, the adjacent binding state was observed to be predominant at stall force, which is consistent with myosin VI anchoring in the cell. These two stepping patterns (or bound states) and the transition between them can only be distinguished by independent tracking of the two head domains under load. The results demonstrate that our DNA nanospring system can reveal the force-dependent functional changes of molecular machines.
We summarize the stepping mechanism of myosin VI under varying loads in Fig. 5. At low load, the hand-over-hand mechanism, where non-adjacent binding states are occupied, is used for the efficient transport of cargos in the cell, because it converts the chemical energy of one ATP into a large forward step (B72 nm head trajectory or 20-40 nm tail trajectory/ATP molecule hydrolysed 32 ) (Fig. 5b). In contrast, an inchworm-like mechanism, where adjacent binding states are occupied half the time, would be inefficient for transport at low load, because it consumes two ATP molecules to produce two short forward steps (43 nm head trajectory each step) to obtain a 20-40 nm tail trajectory 23,34 (Fig. 5c). At higher loads, the stepping mechanism transitions to inchworm-like steps, which may be more efficient for transport, because Brownian ratchets can oppose larger forces when steps are smaller. At the stall force, the myosin heads continue to bind and release, visiting both adjacent and distant binding states (Fig. 4a,b). This is unexpected in terms of the physiological role as an anchor, because robust binding to actin is ideal for anchoring and, in fact, some invertebrate smooth muscles robustly catch actin over long periods with little energy expenditure 36 . We speculate the dynamic binding and release observed in myosin VI at stall force is useful as an adaptation to changes in membrane tension. This feature is similar to skeletal muscle, where dynamic binding and release is thought to allow for quick rearrangement of the binding sites in response to a sudden change in length, reducing or recovering tension 37 .
How does the observed transition to inchworm-like stepping occur at physiological ADP and loading condition? We know that ADP affinity of the heads increases with load 20 , and that when in the ADP bound state, the front head prefers the non-tiltedforward lever arm state 25 (state ii in Fig. 5c and reported by Menetrey et al. 38 ). Thus, under load, the front head is more likely to be in the ADP bound state, which is less likely to tilt forward, and it has been suggested that this leads to more frequent inchworm-like steps 23 . This mechanism is consistent with our observation of a transition to inchworm-like stepping under load with physiological ADP.
The nanospring is a bottom-up force-application nanoscale device, which does not require large-scale equipment, unlike conventional force-application biophysical tools, such as OT, AFM and fluidic chambers. Because of this, our system provides a simple and flexible approach to study single-molecule events under variable force. Our assay system can be extended to other actin-based and microtubule-based motor proteins (myosin II, kinesin and dynein and so on) so long as the ssDNA is labelled to the target proteins. For microtubule-based motors, mutant kinesin 39 may be used as the anchor for the microtubule. Furthermore, our technique is compatible with other single molecule observation systems (for example, high-speed AFM, dark-field microscopy and electron microscopy), which significantly expands the range of potential biophysical experiments. For example, high-speed AFM observation 25 of macromolecules tethered with a nanospring could directly elucidate the force dependence of their structural dynamics and stability. Similarly, dark-field imaging of GNPs tagged with a macromolecule 23,40 and the nanospring would enable us to observe the subdomain motion of proteins perturbed by force with high spatio-temporal resolution. In addition, our nanospring is highly compatible with other DNA origami nanostructures, such as synthetic cargo (or 'chassis') 41 , which can be used to control motor protein assembly, and could be used to investigate the load-dependent mechanical properties of such assemblies. More generally, our DNA nanospring has a tunable spring constant and can be attached to various types of molecules, allowing for the control of species, number, spacing and orientation during force-dependent measurements (for example, Supplementary Fig. 7). The programmability of nanospringmotor protein complexes could also be the basis for nanoactuators to study other mechanosensory proteins or be used as components for synthetic nanoscale devices that require dynamic buildup of mechanical potential energy.
To fold the spring, 10 nM of scaffold was mixed with 100 nM of core staples (Supplementary Data). Oligos were obtained from Hokkaido System Science or IDT. The folding reaction was carried out in Folding buffer (5 mM Tris pH 8.0, 1 mM EDTA and 2-30 mM MgCl 2 ) with rapid heating to 80°C and cooling in single degree increments to 60°C over 2 h followed by additional cooling in single degree increments to 25°C over another 2 h.
The folded nanospring was purified by agarose gel electrophoresis. Structures were loaded into 2% agarose gels and run at 50 V for 3 h in TBE buffer (45 mM Tris, 45 mM boric acid and 1 mM EDTA) supplemented with 11 mM MgCl 2 . The highest migration bands (see Supplementary Fig. 2) were excised, crushed and spun through a Freeze and Squeeze column (Bio-Rad) for 10 min at 20,000 g and 4°C. The nanospring was stored at 4°C.
The purified nanospring was adsorbed for 5 min onto glow discharged formvar-and carbon-coated copper grids, stained for 1 min with 2% uranyl formate and 25 mM NaOH. Images were acquired at 80 kV under low-dose conditions in a Tecnai T12 equipped with a LaB 6 filament and a 4k Â 4k CCD (charge-coupled device) camera. The nominal magnification was Â 50,000 for a pixel size of 3.66 Å at the sample level.
Bead-nanospring conjugation. To estimate the force-extension curve of the nanospring, biotin-modified staples and digoxigenin (DIG)-modified staples Carboxylate-modified polystyrene beads (0.2 mm in diameter, Invitrogen) were crosslinked to anti-DIG polyclonal antibody (Roche) and BSA using the Polylink Protein Coupling Kit (Polysciences, Inc.). One microlitre of anti-DIG antibodycoated beads (B3 nM) and 10 ml of nanospring (B1 nM) were mixed and incubated for 1 h on ice before use in an optical tweezers assay.
A single flow chamber was made using double-sided transparent tape (Scotch) and coverslips (Matsunami). Five microlitres of neutravidin (5 mg ml À 1 , Invitrogen) was flowed into a chamber and incubated for 3 min. Unbound neutravidin was washed out by Assay buffer (AB; 20 mM HEPES-KOH pH 7.8, 25 mM KCl, 5 mM MgCl 2 and 1 mM EGTA) and 10 ml of biotin-coated bead (B3 pM and 0.2 mm in diameter, Invitrogen) was flowed into the cell, which was then incubated for 5 min. Twenty microlitres of bead-nanospring (bead-NS) was diluted ten times in AB plus an oxygen scavenger system 44 and flowed into the cell, which was then sealed with nail polish. The chamber was observed by our microscope after 5 min.
Optical tweezers assay. Fluorescence images of bead-NS were detected by an electron modifying CCD (EMCCD) camera (DL658M-NIT or DU860D-CS0, Andor) at a 30 Hz or 300 Hz recording rate. Bead-NS underwent tethered Brownian motion and an isotropically diffusing bead-NS was chosen (see Supplementary  Fig. 3c) for analysis, to avoid the effects of nonspecific interactions with the glass surface. The bead-NS motion was recorded for 500 frames by the EMCCD camera and the fluorescence images were fitted by a 2D Gaussian distribution to precisely determine the centre position of the fluorescent bead with nanometre accuracy by FIONA 30 and a laboratory-written Labview programme (National Instruments). We estimated the fix point of bead-NS on the glass surface as the mean value of the probability density distribution of the bead position.
When the OT centre was positioned in the vicinity of the fix point, the trapping laser (J15-BL-RW, Spectra-Physics Lasers, Inc.) was switched on to capture the bead (Supplementary Fig. 3d, left). A piezo-driven stage (p-733.3DD, Physik Instrumente) was moved by a triangle wave (frequency, 0.2 Hz or 10 Hz; amplitude, 2.5 mm) in one axis and the trapped bead image was tracked by FIONA. We obtained the force-extension curve according to the procedure described in Supplementary Fig. 3.
Design of myosin constructs. Human myosin Va complementary DNA (Kazusa Product ID ORK07567) was truncated at Gly924. This fragment included the motor domain, lever arm domain (IQ1-6) and a small part of the coiled-coil domain (11 a.a., SVERYKKLHIG). HaloTag (Promega) was attached at the N-terminal via a linker (15 a.a., LRRRPTRPAMDPPSK). For oligo labelling and protein purification, SNAP-tag (NEB) and 6 Â His-tag were attached at the C-terminal via linkers (2 a.a., RA). This Halo-hMyosinVa-SNAP-His fragment and the human calmodulin gene (Met1-Lys149, Kazusa Product ID ORK01403) were introduced downstream of the PH and p10 promoters of the Baculovirus expression vector pFastBac Dual, respectively.
Human myosin VI cDNA (Kazusa Product ID ORK01080) was truncated at Ala1021. This fragment included the motor domain, calmodulin-binding domain, and the proximal, medial and most of the distal tail domain (41 a.a., IQAEVEAQLARQKEEESQQQAVLEQER RDRELALR IAQSEA). For QD labelling, HaloTag was attached at the N-terminal via a linker (15 a.a., LRRRPTRPAMDPPSK). For oligo labelling and protein purification, SNAP-tag (NEB) and 6 Â His-tag were attached at the C-terminal via linkers (2 a.a., RA). This Halo-hMyosinVI-SNAP-His fragment and the human calmodulin gene (Met1-Lys149, Kazusa Product ID ORK01403) were introduced downstream of the PH and p10 promoters of the Baculovirus expression vector pFastBac Dual, respectively.
Protein expression and purification with oligonucleotide labeling. Monomeric forms of myosin V and VI were expressed and purified as follows. Recombinant viruses for myosin V or VI heavy chain and calmodulin were produced by homologous recombination using the Bac-to-Bac Baculovirus Expression System (Life Technologies). Sf9 insect cells were maintained in a monolayer culture in Oligo labelling reactions were performed just after His-tag affinity purification. Amine-modified DNA oligos (Hokkaido System Science) (Oligo A*, B*, C* and D* in Supplementary Data) were linked to the SNAP substrate, benzylguanine (BG; NEB), and 25 mM of BG-oligonuculeotides was labelled with B1 mM myosin V or VI containing an C-terminal SNAP f tag (NEB) in His tag affinity Elution buffer for 30 min at room temperature. Oligo-labelled myosin V and VI were purified by actin filament affinity, aliquoted and stored at À 80°C until use. Just before the experiment, B1 mM oligo-labelled monomeric myosin V or VI was mixed and incubated for 30 min at 4°C, to form DNA-hybridized myosin dimers. The efficiencies of labelling the oligos to and the dimerization of myosin VI were estimated by a gel-shift assay (4-15% gradient gel, Biorad) and determined as 490 and 430%, respectively.
Immobilization of myosin II. Myosin II was extracted and purified from rabbit skeletal muscle 44 . Heavy meromyosin (HMM) was obtained by enzymatically digesting myosin with a-chymotrypsin 45 . Amine-modified DNA oligo (100 mM; Hokkaido System Science; Oligo D* in Supplementary Data) was mixed with 50 mM sulfo-SMCC (Thermo Scientific) and incubated for 1 h at room temperature. Excess SMCC was removed three times by gel filtration (Micro BIO-SPIN P-6, Biorad). HMM was mixed with SMCC-oligo at the stoichiometry 1:10, incubated overnight at 4°C and then mixed with N-ethylmaleimide at 1:2 for 30 min at 4°C. Excess SMCC-oligo was removed twice by ultrafiltration (Amicon Ultra 50 K, Merck Millipore) and Oligo D*-labelled HMM was aliquoted and stored at À 80°C until use. Labelling efficiency was estimated to be B50% by a gel-shift assay. Immobilization was confirmed by the in vitro motility assay 46 .
Myosin VI dimer (2 ml; B500 nM) was mixed with 1 ml HaloTag ligand-QD conjugation (B2 mM) and incubated for 3 h on ice before use for single-molecule imaging experiments. In the case of dual colour imaging, myosin VI monomer was used when mixing with QD. Then, just before the experiment, B500 nM oligolabelled monomeric myosin VI with Qdot 565 or 655 was mixed together and incubated for 30 min at 4°C to form DNA-hybridized myosin VI dimers labelled with different colour QDs.
Myosin-NS conjugation. For the myosin-NS conjugation, handle staples (Supplementary Data) were added to the 100 nM core staples, which were then folded to attach optional 21 bp ssDNA to both ends of the nanospring. One microlitre of DNA-hybridized myosin dimer or immobile myosin II (0.5B1 mM) with anti-handle oligos was mixed with 10 ml of nanospring (B10 nM condensed by ultrafiltration (Amicon Ultra 100 K, Merck Millipore)) and incubated for 30 min on ice.
Single-molecule imaging. To avoid nonspecific interactions of myosin-NS with the glass surface, a glass coverslip was coated with functional PEG according to Schroeder et al. 47 with some modifications. Briefly, coverslips were cleaned with a plasma cleaner for 10 min (FEMTO, Diener electronic). The coverslips were then soaked in freshly prepared 2% (3-Aminopropyl) trimethoxysilane (KBM-603, Shin-etsu Chemical) in acetone for 45 min with gentle shaking at room temperature and rinsed with ddH 2 O. Coverslips were dried and 30 ml of PEG solution (100 mg ml À 1 (PEG 5000 (ME-0500HS, NOF Corp.):PEG 3400 (DE-034HS, NOF Corp.):biotin-PEG (BI-050TS, NOF Corp.) ¼ 100:10:1) in freshly prepared 0.1 M bicarbonate buffer pH 8.3) was put between a pair of coverslips and incubated for 3 h. After rinsing with ddH 2 O, 20 ml of sulfodisuccinimidylartrate (Solteck Ventures) solution (30 mg ml À 1 in freshly prepared 1 M bicarbonate buffer pH 8.3) was put between a pair of coverslips and incubated for 45 min. The coverslips were rinsed, dried and stored at À 80°C in a vacuum desiccator with desiccant.
A single flow chamber was made using double-sided transparent tape (Scotch) and PEG-coated coverslips. Five microlitres of neutravidin (Invitrogen, 0.25 mg ml À 1 ) was flowed into a chamber and incubated for 3 min. Unbound neutravidin was washed out by AB. Ten microlitres of biotinylated actin filament solution (0.01 mg ml À 1 , 10% biotinylation) was flowed into the cell and incubated for 5 min. Unbound actin was washed out by AB and 10 ml of Blocking buffer (10 mg ml À 1 BSA, 0.1% Tween 20 in 5 mM EDTA, 10% glycerol, 0.3 M NaCl and 0.1 M phosphate buffer pH 7.5) was flowed into the chamber, which was then incubated for 2 min. The chamber was then washed with 30 ml of Motility buffer (AB plus 2 mM ATP (Oriental Yeast); 0.436% ADP contaminants B8 mM in 2 mM ATP) and an oxygen scavenger system 44 . For 0 mM ADP experiments, stocked ATP was purified by liquid chromatography 22 . An ATP regeneration system 48 was added in the 0 mM ADP experiments and 100 mM ADP was added in the 100 mM ADP experiments. Ten microlitres of myosin-NS or myosin-QDs-NS diluted 10 or 100 times in Motility buffer was flowed into the cell. The chamber was sealed with nail polish and observed immediately.
Accurate and precise co-localization of different coloured fluorescent probes was achieved in accordance with SHREC 35 . Briefly, multicolour fluorescent beads (Ultra Rainbow Fluorescent particles, 0.2 mm, Spherotech) stuck on a coverslip were used as fiducial markers for each imaging channel. An alignment grid by the fluorescent beads was created with a piezo stage (Physik Instrumente, P-517. 3CD) and the fiducial registration error for each marker was calculated by MATLAB software and estimated to be o7 nm over 12 h, which ensured sufficient stability during the single-molecule experiments.
Photobleaching assay. To quantify the number of myosin VI attached to a nanospring, 10 mM HaloTag TMR ligand (Promega) was labelled with B1 mM myosin VI during the labelling process of 21-mer BG-oligo. Labelling efficiency was spectrophotometrically estimated to be 495%. The TMR and 21-mer oligo-labelled myosin VI was mixed with 42-mer oligo-labelled myosin VI and incubated for 30 min at 4°C to form myosin dimers. Myosin-NS was prepared as described in Myosin-nanospring (myosin-NS) conjugation section and purified with Sephacryl S-500 HR resin (GE Healthcare) twice. The resulting myosin-NS was diluted in Motility buffer, flowed into the cell and observed by TIRFM.
Analysis. The force-velocity relation and the precision of the nanospring extension were determined using Origin 8.5 (Origin Lab.). For the force-velocity relation, individual displacements obtained from tracking a fluorescently labelled nanospirng were averaged to reduce Brownian noise. Line fits were made to successive 1 s time segments and the velocity for each segment was determined as well as the mean force during the segment. Data points were clustered at fixed intervals of force and the average velocities were plotted against average force. The s.e. of the force includes the error from the force-extension curve (average standard deviation is 0.19 pN at 0-2 pN range). The force-velocity curve was constructed from 38 processive motions.
To estimate the precision of the nanospring extension, traces obtained from a single 2D Gaussian fit of TAMRA nanospring images at low ATP concentration (20 mM) were analysed. The precision was defined as the s.d. of the centre positions in a 1 s window. Traces at low load regimes (o500 nm extension) were excluded from the analysis, because the myosin VI stepping could have affected the precision (the stepping rate was 1.1 s À 1 at 20 mM ATP and without load 23 ). We plotted the precision versus the nanospring extension ( Supplementary Fig. 4).
Data availability. The authors declare that the data supporting the findings of this study are available within the article and its Supplementary Information files, or are available from the authors upon request.