Endocytic myosin-1 is a force-insensitive, power-generating motor

Myosins are required for clathrin-mediated endocytosis, but their precise molecular roles in this process are not known. This is, in part, because the biophysical properties of the relevant motors have not been investigated. Myosins have diverse mechanochemical activities, ranging from powerful contractility against mechanical loads to force-sensitive anchoring. To better understand the essential molecular contribution of myosin to endocytosis, we studied the in vitro force-dependent kinetics of the Saccharomyces cerevisiae endocytic type I myosin called Myo5, a motor whose role in clathrin-mediated endocytosis has been meticulously studied in vivo. We report that Myo5 is a low-duty-ratio motor that is activated ~10-fold by phosphorylation, and that its working stroke and actin-detachment kinetics are relatively force-insensitive. Strikingly, the in vitro mechanochemistry of Myo5 is more like that of cardiac myosin than like that of slow anchoring myosin-1s found on endosomal membranes. We therefore propose that Myo5 generates power to augment actin assembly-based forces during endocytosis in cells.

The actin cytoskeleton can produce pushing and pulling force, both of which are 46 required for CME in S. cerevisiae (Sun et al., 2006). When actin filament ends grow against a 47 surface, they push the surface forward Oster, 1996, 2003). During CME, actin 48 filaments, bound by coat proteins, grow against the plasma membrane at the base the CME 49 site, driving invagination (Picco et  Additional power may be provided by myosins, which generate tension on actin 55 filaments. The myosins critical for CME, Myo3 and Myo5 in budding yeast, and Myo1e in 56 vertebrates, are type I myosins (Geli and Riezman, 1996;Cheng et al., 2012;Krendel et al., 57 To distinguish between these possibilities, we measured the force sensitivity of the 80 endocytic myosin Myo5 (not to be confused with the vertebrate type V myosin). Myo5 is 81 insensitive to resistive force compared to related myosins. We therefore propose that Myo5 82 actively powers CME. Because actin and myosin collaborate in a variety of membrane 83 remodeling processes, we expect that these results will be instructive beyond CME. 84 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made where K1' is a rapid equilibrium binding step, k2' is a rate-limiting isomerization to the 132 AM.ATP state, and kdiss is the rapid actin dissociation step. The apparent second order rate 133 constant for ATP binding to phosphorylated actoMyo5 was determined by a linear fit to the 134 data (K1 ' k2 ' = 0.39 ± 0.017 µm -1 s -1 ). The unphosphorylated actoMyo5 data were fit by: 135 2 0 ′ , (Equation 2) 136 and the maximum rate of isomerization (k2' = 290 ± 24 s -1 ) and ATP affinity (K1 ' = 0.006 ± 137 0.0016 µM -1 ) were determined. The apparent second-order rate constant for ATP binding 138 (K1'k2') was determined from a linear fit of the observed rates below 100 µM ATP to be 1.1 139 ± 0.28 µM -1 s -1 (Table 1). 140 The rate constant for ADP dissociation (k+5') was measured by preincubating 141 When myosin active sites are saturated with ADP, the rate of ATP-induced dissociation of 146 actomyosin is limited by ADP's slow dissociation. Light scattering transients were fitted by 147 single exponential functions, yielding rates for ADP release for phosphorylated actoMyo5 148 (k+5' = 74 ± 2.0s -1 ) and for unphosphorylated actoMyo5 (k+5' = 107 ± 5.9 s -1 ) ( Fig. 2F and 149 Table 1). The signal-to-noise ratio of the fast light scattering transients is low, resulting in 150 large uncertainties on these fits. However, these rates are substantially faster than the 151 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted June 30, 2023. ; https://doi.org/10.1101/2023.03.21.533689 doi: bioRxiv preprint steady-state ATPase values, but slower than the maximum rate of ATP-induced actomyosin 152 dissociation. ADP release for actoMyo5 ADP is much faster than ADP release for vertebrate 153 Myo1b and Myo1c (Greenberg et al., 2012;Lewis et al., 2006). It is more similar to the 154 vertebrate endocytic myosin-1, Myo1e (El Mezgueldi et al., 2002). Because ADP release is 155 rate limiting for detachment of Myo5 and Myo1e from actin, fast ADP release by these 156 molecules mean that the unloaded actin-attachment lifetimes for endocytic type I myosins 157 are < 15 ms. This property may make these motors particularly well-suited to function in 158 dynamic actin networks like those at CME sites, where actin filaments elongate and 159 "treadmill" into the cytoplasm (Kaksonen et al., 2003(Kaksonen et al., , 2005. 160 161 Actin gliding is dependent on Myo5 phosphorylation state 162 Our kinetic results suggest that both phosphorylated and unphosphorylated Myo5 have 163 low duty ratios (i.e., the motor spends a small fraction of its ATPase cycle bound to actin). 164 Since ADP release limits the rate of phosphorylated Myo5 detachment from actin at 165 saturating ATP (k+5' = 74 ± 2.0 s -1 ) and since we have measured the overall ATPase rate 166 (Vmax = 3.3 ± 0.15 s -1 ), we can estimate the duty ratio: 167 The calculated duty ratio of phosphorylated Myo5 is 0.045. Unphosphorylated Myo5 has a 169 lower duty ratio (< 0.004). 170 To assess the effect of phosphorylation on Myo5 motility, we performed in vitro 171 motility assays at 1 mM ATP. Motors were attached site-specifically to coverslips coated 172 with anti-His6 antibody. Coverslips were incubated with a range of concentrations of 173 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The kinetics of actin attachment durations and mechanics of single myosin molecules were 199 measured using an optical trapping instrument that can detect sub-nanometer 200 displacements with millisecond temporal resolution (Woody et al., 2018;Snoberger et al., 201 2021). We used the three-bead optical trapping geometry in which a biotinylated actin 202 filament is held between two laser-trapped polystyrene beads coated with neutravidin, 203 creating a bead-actin-bead dumbbell ( (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted June 30, 2023. ; https://doi.org/10.1101/2023.03.21.533689 doi: bioRxiv preprint aligned at their beginnings and forward-averaged in time (Fig 3F-H, left), or aligned at their 219 ends and reverse-averaged in time (Fig 3F-H, right). 220 Ensemble averages of Myo5 interactions showed a two-step working stroke at the 221 three ATP concentrations, but the step-size was most accurately resolved at 10 µM ATP 222 (see Materials and methods). In this condition, an initial substep of 4.8 nm was followed by 223 a second substep of 0.2 nm (Fig. 3G). We determined the lifetimes of the substeps by fitting 224 the ensemble averages with single exponential functions. At 1 µM ATP ( The observed rates at 1 and 10 µM ATP are consistent with the second order rate constant 235 for ATP binding of 0.39 ± 0.017 µM -1 s -1 measured by stopped-flow kinetics (K1'k+2', 236 Table 1). 237 We determined the detachment rates of actomyosin events by plotting the 238 cumulative frequency of individual attachment durations and fitting a single exponential 239 function to the data by maximum likelihood estimation (Fig. 3I). Data from 1 and 10 μM 240 ATP were well fit by single exponentials with rates of 0.88 and 6.87 s -1 , respectively (Fig. 3I, 241 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted June 30, 2023. ; https://doi.org/10.1101/2023.03.21.533689 doi: bioRxiv preprint Fig. 3K). These rates match well with the observed rate of ATP binding (Table 1), as well as 242 the fits for the reverse ensemble averages, indicating that at sub-saturating ATP (1 and 10 243 µM), detachment is limited by ATP binding ( Figure 3J, blue squares & gray diamonds, Fig.  244 3K). Data from 1000 μM ATP were best described as the sum of 2 exponentials, with the 245 major rate of 67.8 s -1 comprising 92.1% of the total, and a minor rate of 11.6 s -1 comprising 246 7.9% of the total (Fig. 3I, Fig. 3K). The major rate is consistent with both the observed ADP 247 release rate and the measured forward ensemble average rates, indicating that at 248 saturating ATP, ADP release limits detachment of actomyosin interactions (

Myo5 is a relatively force-insensitive motor 252
To elucidate the force sensitivity of Myo5, we measured how its actin detachment rate was 253 affected by mechanical force opposing the power stroke using an isometric feedback 254 system that maintained the actin filament near its initial position (Takagi et al., 2006). The 255 initial force applied to Myo5 in this system depends in part on where along the actin 256 filament Myo5 stochastically binds, so this approach allowed measurement of attachment 257 durations at a range of resistive forces (Fig. 4A). Plotting attachment durations as a 258 function of force revealed a general trend of longer attachment durations at higher 259 resisting forces. At each interaction force, attachment durations are exponentially 260 distributed and, as expected based on prior isometric feedback experiments, the data 261 appear noisy when plotted this way (Fig. 4A). Converting these data to detachment rates by 262 binning them by force at every ten points, averaging, and taking the inverse of the 263 attachment duration more clearly reveals the trend (Fig. 4B). 264 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted June 30, 2023. 15 nm), is dramatic. From 0 to 2 pN of resistance, Myo1b attachment lifetimes slow from 283 ~600 ms to ~45 s, resulting in negligible power generation (Fig. 4D). Over the same 284 interval, Myo5 attachment lifetimes slow very modestly from ~15 ms to ~25 ms, allowing 285 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted June 30, 2023. ; https://doi.org/10.1101/2023.03.21.533689 doi: bioRxiv preprint it to generate considerable power (Fig. 4D). Thus, Myo5 is unlikely to act as a force-286 sensitive anchor in cells and is more likely to power movements against a resisting load. 287 288

Proposed function of type I myosin in clathrin-mediated endocytosis 289
Myo5 is one of the best-studied myosin-1 proteins in vivo. Quantitative live cell imaging 290 and electron microscopy have revealed that it is recruited to CME sites simultaneously with 291 initiation of actin assembly, where it concentrates at the base of the site as membrane 292 invagination proceeds (Jonsdottir and Li, 2004;Idrissi et al., 2008). Although it has long 293 been appreciated that the presence (Geli and Riezman, 1996; Goodson et al., 1996) and 294 mechanochemical activity (Sun et al., 2006) of type-1 myosins are required for CME, the 295 mechanistic contribution of motor activity to the dynamic actin network was unknown. 296 When it was discovered that some type I myosins are acutely force sensitive (Laakso et al., 297 2008), it became apparent that these motors could have mechanochemical activities that 298 range from force-dependent anchoring to power generation during CME. However, 299 distinguishing among these possibilities has been difficult in cells. Mutant Myo5 molecules 300 lacking the motor head or bearing mutations intended to lock the ATPase in low and high 301 actin affinity states each block CME, results that do not reveal the mechanochemical role of (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made Here we have shown that Myo5's motor generates power rather than forming force-309 sensitive catch bonds. The overall ATPase rate of Myo5 is slow relative to other power-310 generating myosins, but its power stroke and detachment from actin are fast, and they slow 311 only modestly under load (Fig. 4C). Myo5's relative force insensitivity means it generates 312 steady power against resistance (Fig. 4D). Because Myo3 and Myo5 can each support CME 313 in the absence of the other, we suspect that Myo3 is similarly force-insensitive. at the base of CME sites, and it may move actin filaments at an angle to the membrane it is 329 bound to. Actin subunits "treadmill" towards the cytoplasm in endocytic actin networks at 330 ~50-100 nm/s (Kaksonen et al., 2005(Kaksonen et al., , 2003, so Myo5's motility rate of 700-900 nm/s (Fig.  331 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted June 30, 2023. ; https://doi.org/10.1101/2023.03.21.533689 doi: bioRxiv preprint 2G), which we would expect resistance to slow only modestly, is fast enough to do work on 332 the actin network as it assembles. We therefore expect that the myosins power membrane 333 invagination and relive load to accelerate actin assembly during CME. resolved. Here, we demonstrated that a type I myosin critical for CME, a process well-339 known to be driven by actin assembly, generates power. Implication of endocytic type I 340 myosin as a force-insensitive motor suggests that actin assembly and myosin power 341 generation can be coordinated to do coherent work in membrane remodeling processes. 342 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted June 30, 2023. Optical trapping experiments were performed at room temperature (20 ± 1 °C) 499 using a dual beam 1064 nm trapping laser as described in (Woody et al., 2018(Woody et al., , 2017. A 500 single laser beam was split into 2 beams using polarizing beam splitters and steered into a 501 60x water immersion objective (Nikon). Laser light was projected through an oil 502 immersion condenser and into quadrant photodiodes (JQ-50P, Electro Optical Components, 503 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted June 30, 2023. ; https://doi.org/10.1101/2023.03.21.533689 doi: bioRxiv preprint Inc.), each of which were conjugate to the back focal plane of the objective. Direct force 504 detection from the quadrant photodiodes was achieved using a custom-built high-voltage 505 reverse bias and an amplifier. Data acquisition, beam position control output, and isometric 506 feedback calculations were controlled with custom-built virtual instruments (Labview, 507 Matlab). 508 Individual 0.8 μm diameter neutravidin-coated bead were caught in the two traps 509 and held approximately 5 μm apart. Trap stiffnesses were adjusted to 0.05 -0.1 pN/nm for 510 each trap. A biotinylated actin filament visualized by rhodamine phalloidin was bound to 511 the two trapped beads, creating a bead-actin-bead dumbbell. The dumbbell was 512 pretensioned (3-5 pN) by steering one beam using a piezo controlled mirror conjugate to 513 the back focal plane of the objective, and the surface of pedestal beads were probed for 514 myosins. Putative myosin interactions were detected via drops in variance of the two 515 beads, and the 3-dimensional position of the dumbbell relative to the myosin was refined 516 further by maximizing the rate and size of the observed power stroke deflections. Every 30-517 60 s, the dumbbell was moved axially along the actin filament in ~6 nm steps between 518 trace acquisition to ensure even accessibility of actin-attachment target zones. Stage drift 519 was corrected via a feedback system using a nano-positioning stage and imaging the 520 position of the pedestal bead with nm precision (Woody et al., 2017). In experiments using 521 1 μM ATP, due to the longer actomyosin interactions, stage drift was still observed even 522 with the stage feedback engaged, leading to a presumed underestimation of the 523 displacement size. All data were acquired at a sampling rate of 250 kHz. 524 Isometric optical clamping experiments were performed as described in (Woody et  covariances and a change-point algorithm. Data collected at 1000 μM ATP were analyzed at 537 250 kHz, while data collected at 1 and 10 μM ATP were downsampled to 2 kHz by 538 averaging every 125 points to enhance analysis speed. Events were detected by calculating 539 the covariance of the 2 beads using a smoothing window of 33.3, 15, and 5.25 ms and an 540 averaging window 60, 36, and 12 ms at 1, 10, and 1000 μM ATP, respectively. The 541 instrument deadtime was calculated to be 2 times the covariance averaging window. For 542 each 15 s trace, the detected covariance was plotted and fit to double gaussian 543 distributions, with the smaller mean gaussian corresponding to the actomyosin "bound" 544 portion and the larger mean gaussian corresponding to the "unbound" portion of events. A 545 putative event was defined as an event where the covariance starts above the unbound 546 peak mean, drops below the bound peak mean, and remains below the unbound peak mean 547 for at least the length of the instrument deadtime prior to returning back above unbound 548 peak mean. Event starts and ends were further refined using a changepoint algorithm as 549 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted June 30, 2023. (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made  Each data point represents 3-6 time courses averaged and fit to a single exponential decay 778 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made  Movie S2 is four times faster than the playback rate of Movie S1. 874 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made Anchoring of actin by Myo5

Figure 1
Type I Myosin . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted June 30, 2023. ; https://doi.org/10.1101/2023.03.21.533689 doi: bioRxiv preprint  . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted June 30, 2023. ; https://doi.org/10.1101/2023.03.21.533689 doi: bioRxiv preprint ...

Actin gliding on Myo5 coated coverslips
. CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted June 30, 2023. ; https://doi.org/10.1101/2023.03.21.533689 doi: bioRxiv preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted June 30, 2023. ; https://doi.org/10.1101/2023.03.21.533689 doi: bioRxiv preprint  (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted June 30, 2023. ; https://doi.org/10.1101/2023.03.21.533689 doi: bioRxiv preprint