Direct observation of motor protein stepping in living cells using MINFLUX

Dynamic measurements of molecular machines can provide invaluable insights into their mechanism, but these measurements have been challenging in living cells. Here, we developed live-cell tracking of single fluorophores with nanometer spatial and millisecond temporal resolution in two and three dimensions using the recently introduced super-resolution technique MINFLUX. Using this approach, we resolved the precise stepping motion of the motor protein kinesin-1 as it walked on microtubules in living cells. Nanoscopic tracking of motors walking on the microtubules of fixed cells also enabled us to resolve the architecture of the microtubule cytoskeleton with protofilament resolution. Description Zeroing in on motor proteins The super-resolution microscopy technique MINFLUX enables localization of fluorophores using a minimal number of photons. Two studies now expand on the development and implementation of MINFLUX to track motor protein dynamics in vitro and in cells (see the Perspective by Fei and Zhou). Wolff et al. refined the precision of MINFLUX such that single-fluorophore tracking with nanometer precision was possible with only tens of photons. They tracked the movement of kinesin-1 on microtubules and were able to see individual 4-nanometer substeps and rotation of the protein during stepping in their analysis. Deguchi et al. applied MINFLUX with a labeling and tracking approach called motor-PAINT to monitor stepping of motor proteins on microtubules in living and fixed cells in both two and three dimensions. —MAF Nanoscale conformational dynamics of individual motor proteins are measured in living cells.

M olecular machines drive many processes essential for life. For example, members of the myosin, kinesin, and dynein families are adenosine 5′triphosphate (ATP)-driven processive molecular motors that recognize the intrinsic structural polarity of cytoskeletal filaments to drive directed movement important for cellular processes such as intracellular transport and cell division (1)(2)(3). Over the past decades, the microtubule plus-end directed motor kinesin-1, hereafter called kinesin, has served as a key model both for the understanding of motor dynamics and for the development of improved single-molecule methods such as optical tweezers and single-particle tracking (4)(5)(6)(7). These studies have revealed that kinesin moves in a hand-over-hand manner, in which each step along the microtubule encompasses a 16-nm displacement of the N-terminal motor domain, which leads to an 8-nm displacement of the C-terminal cargo-binding domain (Fig.  1A). Despite the great success of existing singlemolecule techniques in studying purified motors in well-controlled in vitro reconstitution experiments, performing such experiments in the living cell has proven challenging. The large labels required for optical tweezers typically bind multiple motors in undefined ways, and single fluorophores do not provide sufficient spatial and temporal resolution to resolve the fast stepping behavior of motors under phys-iological ATP concentrations. As a workaround, bright nanoparticles taken up through endocytosis into transport vesicles have been used as proxies to study motor dynamics in living cells (8,9). However, in such experiments, the identity and copy numbers of the motors that drive transport are unknown. Therefore, the stepping dynamics of specific motors inside living cells has remained experimentally inaccessible.
MINFLUX, a super-resolution microscopy technology (10), holds great potential to overcome these limitations. By efficiently using the limited photon budget of single fluorophores, MINFLUX enables high spatial resolution for imaging (10)(11)(12) and temporal resolution for fluorophore tracking (13,14). In MINFLUX tracking, a donut-shaped excitation beam is scanned around a single fluorophore (Fig. 1B). From the intensities measured at specific positions, the coordinate of the fluorophore is calculated and the scanning pattern is recentered on this position before the next iteration. Keeping the fluorophore close to the dark center of the beam results in a high localization precision and minimizes photobleaching. In a companion paper in this issue, MINFLUX was used to dissect the conformational stages of kinesin stepping on in vitro polymerized microtubules (15). We now show that this technique can also enable the highresolution tracking of fast molecular motors in living cells.

Optimization of MINFLUX for motor protein tracking
To establish MINFLUX tracking of molecular motors in living cells, we first optimized the workflow on single molecules and fluorescent beads in vitro to maximize precision and speed (fig. S1, A to C) and achieved a localization precision of ≈2 nm with a submillisecond temporal resolution. We then optimized kinesin tracking using motor-PAINT (16). In this method, cells are permeabilized and fixed before fluorescently labeled kinesin motors are added that walk along microtubules toward the plus end ( Fig. 1C and movie S1). Unlike high-resolution in vitro assays that use large beads as labels (5,17), we used small fluorescent tags that reduced the linkage error to ≈3 nm [as predicted by Alphafold2 (18,19)], which was comparable to the system resolution. Motor-PAINT is less challenging than live-cell tracking because it allows us to precisely control the concentration of kinesin motors and, importantly, their speed, by adjusting the ATP concentration. Using MINFLUX motor-PAINT, we were able to reconstruct cellular microtubules with a precision of ≈2 nm (Fig. 1, D and E, and table S1). Additionally, the directionality of kinesin reveals the orientation of the microtubules. Compared with standard motor-PAINT with a wide-field microscope, the use of MINFLUX improved the localization precision 5-fold, the temporal resolution 50-fold, and the number of localizations per track by more than one order of magnitude. In neurons, this allowed us to better resolve individual microtubules inside dendrites compared with our earlier motor-PAINT study (Fig. 1D) (16). In human osteosarcoma (U2OS) cells, we could resolve individual trajectories of the purified motors in the crowded area around the centrosome with near-protofilament resolution (Fig. 1, E and F). Tracks, which were just 12 nm apart, were easily resolvable, and we regularly observed side stepping between different protofilaments ( Fig. 1, F and G, and movie S2). These side steps often occurred after stalling events, suggesting that motors were circumventing obstructions such as microtubuleassociated proteins (MAPs) that became fixed to the microtubules or microtubule defects from the fixation process.
A closer inspection of individual tracks showed clusters of localizations that correspond to the 8-nm steps of the labeled C terminus. Indeed, these steps become obvious when plotting the position of the motor along the microtubule over time (Fig. 1H). This allowed us to quantify the precise step-size and dwell-time distributions under saturating (physiological) ATP concentrations (Fig. 1, I and J). From 956 steps in 49 tracks, we measured a step size of 7.8 ± 2.7 (SD) ± 0.09 (SEM) nm and an average dwell time of 30.8 ms. To investigate the stepping behavior in greater detail, we reduced the ATP concentration to slow down the motors (20), resulting in a similar step size but a reduced rate constant for ATP binding ( fig. S2, A and E, and tables S1 and S2). Under these conditions, we could measure hundreds to thousands of localizations per step (Fig. 1, K and L, and movie S3). Averaging over the coordinates allowed us to calculate the position of each step with subangstrom precision (SEM).

Application of MINFLUX in living cells
We next attempted MINFLUX tracking of kinesin in living cells. To this end, we expressed HaloTag-kinesin (full-length) in U2OS cells (movie S5) and labeled at most a single motor domain per dimer by addition of the dye JF646 at very low concentrations. Individual tracks clearly revealed the 16-nm steps of the motor domains (Fig. 2, A to D, and movie S6). On average, we found a step size of 15.7 ± 3.8 SD ± 0.25 SEM nm and an average dwell time of 46.8 ms (fig. S2, B and F). We also observed tracks with frequent switching between microtubules, back-slipping potentially caused by multiple competing motors (movie S7), and, unlike in motor-PAINT, tracks without clear steps ( fig. S4). The latter potentially stem from kinesins that are attached to dynamic microtubules or cargoes driven by other motors and are thus passively dragged along. However, the number of tracks that we could acquire was low, likely because of kinesins assuming an autoinhibited form with only a low fraction in the processive state (22,23). To increase the throughput, we used the truncated kinesin variant K560, in which cargo binding and autoinhibition are removed, and treated cells with Taxol to increase the number of stabilized microtubules preferred by kinesin (movie S8) (24). With these changes, we could readily observe multiple tracks in a single field of view (Fig. 2, E to H, and movie S9). This allowed us to measure precise in vivo step-size and dwell-time distributions from 2887 steps in 330 tracks (Fig. 2, I and J). These measurements revealed that although the average step size of 16.2 ± 3.8 SD ± 0.07 SEM nm was similar to that of full-length kinesin, the average dwell time of 27.5 ms was much shorter, consistent with the higher speeds that we observed with K560 (table S2). To test whether our approach could be extended to more complex and sensitive cell types, we examined kinesin dynamics in the axons of live primary mouse cortical neurons (Fig. 2, K to M), which critically depend on motor-driven transport. Here, we could clearly quantify the 16-nm stepping dynamics of kinesin without Taxol treatment (step size = 15.7 ± 3.7 SD ± 0.21 SEM nm; average dwell time 29.2 ms; fig. S2, C and G), demonstrating that MINFLUX reveals conformational dynamics of individual motor proteins in complex cellular systems.

Three-dimensional tracking in live cells
Because most biological structures extend into three dimensions, only three-dimensional (3D) tracking can capture the true dynamics and avoid projection artifacts that limit the accuracy in 2D tracking. Unfortunately, standard single-particle tracking provides, at best, poor z resolution (25). MINFLUX has been used to image cellular structures in 3D (11), but for tracking, it has so far remained limited to 2D. We therefore adapted MINFLUX for 3D track-ing by scanning a 3D donut beam in 3D (Fig. 3,  A and B). We achieved a localization precision of 2.5, 3.1, and 3.9 nm in the x, y, and z directions, respectively ( fig. S1D). When used with motor-PAINT, we could resolve many tracks on crossing microtubules (Fig. 3C), including jumps between microtubules (arrow in Fig. 3D). We could also establish 3D tracking in live cells with a similar spatial and temporal resolution (3.9 nm and 3.0 ms, respectively; table S1), allowing us to resolve the 16-nm steps of kinesin in 3D (Fig. 3, E and F; fig. S2, D and H; and movie S10). When we investigated these trajectories in the cross-sectional views, we found dynamic movements along the z axis, including side steps and vertical trajectories. 3D tracking allowed us to extract accurate step sizes from a strongly inclined trajectory (average 15.1 nm), which, when analyzed in 2D, showed a bias toward smaller step sizes (average 9.0 nm) ( fig. S5). Thus, MINFLUX tracking opens the possibility to quantify the precise 3D dynamics of molecular machines in living cells.

Discussion
Here, we established MINFLUX tracking of kinesin with nanometer spatial and submillisecond temporal resolution and demonstrated that we could directly resolve steps of individual motors in live cells. In contrast to recent in vitro MINFLUX measurements (15), we did not observe clear 4-nm substeps of kinesin in motor-PAINT, likely because of insufficient spatial resolution. However, in live-cell experiments with our N-terminally labeled motor, we could occasionally observe 8-nm substeps ( fig. S6). This encouraged us to also image C-terminally labeled kinesin in living cells.
Here, the faster dynamics and higher background fluorescence made the measurements challenging, but we could observe tracks with the expected 8-nm step size ( fig. S7). We found that these C-terminally labeled motors moved with a slightly higher average velocity (table  S2), an effect that could be the result of motor domain labeling, an important consideration for future experiments. Our study paves the way for investigating how the stepping kinetics of motors in cells are modulated by the presence or absence of different MAPs or cargo adaptors (26). Such measurements could help to explain the observed discrepancy of kinesin stepping behaviors such as dwell time, stalling, and side stepping in motor-PAINT (in vitro-like) and live cells (e.g., the presence of regulatory MAPs such as MAP7, higher salt concentrations, etc.) (tables S1 and S2). One parameter that was consistent across studies was the average step size despite the fact that in live cells the motion of the microtubules themselves could contribute to the measured step size. As expected, the average step size was ≈16 nm, both in U2OS cells [which have some microtubule sliding (27)] and in neuronal axons [where microtubules are largely immobile (28)] (table S1). The impact of microtubule movement is thus likely to be minimal at the time scale of kinesin steps.
MINFLUX tracking is not limited to kinesin, but can also be used to study the precise motion of any protein in living cells with high spatiotemporal resolution and minimal perturbance because of its compatibility with single-fluorophore labels (see fig. S8 for tracking of Myosin-V). In the future, developing MINFLUX to simultaneously track two colors will enable monitoring of the relative 3D positions of labeled protein domains with nanometer spatial and submillisecond temporal resolution. Such measurements of conformational changes of molecular machines in their native environment will provide important insights into their function and regulation.