Kinesin-1 autoinhibition tunes cargo transport by motor ensembles

Abstract

Intracellular vesicular transport by kinesin-1 motors through numerous three-dimensional (3D) microtubule (MT) intersections must be regulated to support proper vesicle delivery. Knowing kinesin-1 can be regulated via autoinhibition, does kinesin-1 exhibit autoinhibition on cargo, and could this regulate vesicular transport through 3D MT intersections in vitro? To answer these questions, we compared liposome transport by ~10 nearly full-length kinesin-1 motors with kinesin light chains bound (KinΔC) versus a constitutively active control (K543). In 3D MT intersections, KinΔC-liposomes terminate (48%) or go straight (43%) but rarely turn (9%), starkly contrasting K543-liposomes, which go straight (57%) or turn (31%) but rarely terminate (12%). On single MTs, KinΔC-liposomes have reduced run lengths and detachment forces versus K543-liposomes, suggesting autoinhibition reduces MT engagement, as supported by threefold lower KinΔC MT landing rates versus K543 and mechanistic in silico modeling. Furthermore, Kinesore, a small molecule that overcomes kinesin-1 autoinhibition, restores KinΔC’s MT engagement. Thus, we propose that partial kinesin-1 autoinhibition while cargo bound may fine-tune cargo delivery to support physiological demands.

INTRODUCTION

Secreted and transmembrane proteins are sorted within the Golgi network into vesicles that are transported by kinesin-1 (KIF5) motors along microtubules (MTs) to their destination on the plasma membrane (1–7). To meet the cell’s physiological demands, vesicular transport and delivery must be regulated spatially and temporally (3,8). Although multiple mechanisms exist to regulate how kinesin-1 motor teams maneuver vesicular cargo through the complex three-dimensional (3D) MT network (9–11), the most basic mode of regulation is by activating or inhibiting individual motors (11,12). Heterotetrameric kinesin-1, consisting of two heavy chains (KHCs) and two kinesin light chains (KLCs) (13,14), adopts a “folded” autoinhibited state with numerous intramolecular interactions along the length of the KHC (1,15–18). The KLCs further stabilize the inhibited state by forming a six-helix bundle with the KHC near the hinge 2 region (4,15–17). Cargo binding to KHC’s C-terminal domains or KLC is thought to disrupt these autoinhibitory interactions, allowing the motor to become “extended” and active (i.e., able to engage the MT) (16). However, once bound to cargo, is kinesin-1 stably active, or is it in an equilibrium between the autoinhibited “folded” and active “extended” states, which could be tuned to optimize cargo transport and delivery by kinesin-1 motor teams (8,19)?

Here, we reconstitute vesicular cargo transport in vitro to determine if kinesin-1 autoinhibition affects multimotor transport of physiologically relevant, 350-nm diameter, fluid-like liposomes through suspended 3D MT intersections (20), much like those encountered in cells (21–24). We took advantage of our previously characterized, nearly full-length (1–888 amino acids) kinesin-1 construct (KinΔC) with bound KLC that lacks the so-called IAK region near the C-terminus that is necessary for full inhibition of activity, but retains the hinge region in the coiled coil that allows the molecule to bend and engage in a head-tail interaction (25). Directional outcomes in 3D MT intersections for liposomes transported by teams of ~10 KinΔC motors were in stark contrast to those of liposomes transported by teams of a constitutively active truncated kinesin-1 lacking KLC (K543) (20). Specifically, KinΔC-liposomes either went straight through or terminated in intersections, but they rarely turned onto the intersecting MT. K543-liposomes preferred to go straight rather than turn, but they rarely terminated in the intersection (20). These differences in directional outcomes may be due to the autoinhibition of the cargo-bound KinΔC motors effectively reducing the number of motors available to engage the MT (16,17,25). This hypothesis is supported by the observation that cargos transported by KinΔC ensembles along single MTs have shorter run lengths and generate less force against an optical trap compared with K543 ensembles. Treating KinΔC with Kinesore, a small molecule that binds KLC and activates kinesin-1 (26), causes the motor to adopt a more “extended” conformation, making KinΔC MT engagement and transport identical to constitutively active K543 motors. To explain why KinΔC autoinhibition modulates liposome transport in 3D MT intersections, we modified our previously published in silico mechanistic model of kinesin-1 transport (20) to implicitly include autoinhibition. The model reproduces our observations of liposome transport through 3D MT intersections by KinΔC when modified either by reducing the number of active motors in the ensemble or reducing the maximum rate of MT association, supporting our hypothesis. We propose that kinesin-1 autoinhibition provides the cell with a mechanism to expand and fine-tune its dynamic range of potential cargo transport and delivery properties.

MATERIALS AND METHODS

Kinesin motor constructs and purification

Two kinesin motor constructs were used in this study, KinΔC and K543. KinΔC contains the N-terminal 888 amino acids of the mouse KIF5B kinesin heavy chain (NCBI accession number: Q61768) followed by a biotin ligase recognition sequence and FLAG tag used for purification, coexpressed with full-length mouse kinesin light chain KLC2 (NCBI accession number: BC014845). K543 contains the N-terminal 543 amino acids of the mouse KIF5B kinesin heavy chain (NCBI accession number: Q61768) followed by a biotin ligase recognition sequence and FLAG tag used for purification. The biotin ligase recognition sequence used in both constructs is an 88-amino-acid sequence from the E. coli biotin carboxyl carrier protein and is singly biotinylated during expression (27).

Both kinesin constructs were expressed in the baculovirus/Sf9 cell system and purified by affinity chromatography as described previously (20). To express KinΔC, Sf9 cells were infected with two recombinant baculoviruses, one encoding the sequence of the KinΔC heavy chain and the other encoding full-length KLC2. Each baculovirus was used at a ratio of five viral particles per cell. Importantly, prior work demonstrated that this strategy yields a 1:1 stoichiometry of KLC to KinΔC heavy chain in the final sample (25,28). Additional kinesin constructs with an encoded YFP were expressed and purified as described previously for motor counting experiments (20,29). For motor counting KinΔC, a C-terminal YFP was added to the KLC2. Kinesin aliquots were in 10 mM imidazole (pH 7.4) with 200 mM NaCl, 55% glycerol, 1 mM DTT, 1 μg/ml leupeptin, and 50 μM MgATP. Aliquots were flash frozen before storage at −80° C.

Microtubule preparation

MTs were prepared using a mixture of unlabeled porcine brain tubulin (Cytoskeleton) with rhodamine-labeled tubulin (Cytoskeleton) or Alexa-647 labeled tubulin (PurSolutions). For rhodamine-labeled microtubules, 20% of the tubulin was labeled, whereas Alexa-647 microtubules contained 33% labeled tubulin. MTs were assembled and stabilized in vitro as described previously (20). The tube containing the assembled MTs was wrapped in foil for storage in a dark drawer, thus minimizing potential photodamage. Polymerized MTs were stable at room temperature for up to 1 week.

Single-molecule quantum dot motility assays

Single-molecule motility assays were performed as described previously (20). Briefly, a kinesin aliquot was rapidly thawed and brought to a final volume of 50 μL with ice cold Buffer 1 (25 mM imidazole (pH 7.4), 300 mM KCl, 4 mM EGTA, and 4 mM MgCl₂). The protein was then clarified by ultracentrifugation at 392,000 × g for 10 min at 4°C, transferred to a fresh tube on ice, and then diluted to a final concentration of 100 nM in Buffer 2 (25 mM imidazole (pH 7.4), 25 mM KCl, 4 mM EGTA, and 4 mM MgCl₂). 2 μL of 100 nM kinesin was mixed with 2 μL of 1 μM Qdot 655 Streptavidin Conjugate (Thermo Fisher) and 6 μL Buffer 2 and incubated on ice for 1 h. The 10:1 ratio of Qdots to kinesin was selected to ensure that experiments are truly in single-molecule conditions. These assays were performed identically for K543 and KinΔC (20).

Single-molecule experiments were performed on silanized cover glass, which was prepared as described previously (20). Motility chambers were prepared by attaching a 22 × 22 mm No. 1 cover glass to a silanized 24 × 60 mm silanized cover glass, separated by 125-μm Mylar shims, using UV-curable adhesive (Norland Optical Adhesive 68). The motility surface was prepared by first coating with 0.8% antitubulin antibody (Bio-Rad YL1/2) in BRB-80 and incubating for 5 min. The chamber was then washed with Wash Buffer (BRB-80 plus 20 μM paclitaxel) before blocking with 5% w/v Pluronic F-127 in BRB-80 for 5 min and again washed. Rhodamine-labeled MTs diluted 1:400 in BRB-80 with 20 μM paclitaxel were flowed into the motility chamber and incubated for 10 min before a final wash with Wash Buffer. Kinesin-Qdot complex was diluted 1:20 in Motility Buffer (Buffer 1 with 2 mM MgATP, 20 μM paclitaxel, 0.5 mg/mL Casein, 0.5% w/v Pluronic F-127, 5 mM creatine phosphate, 0.4 mg/mL creatine phosphokinase, 10 mM DTT, 3.5 mg/mL glucose, 40 μg/mL glucose oxidase, and 27 μg/mL catalase) and flowed into the motility chamber. For experiments utilizing the kinesin activator Kinesore (Tocris Bioscience Cat. No. 6664), a Kinesore stock of 50 mM was prepared in DMSO and kept at −20°C. On the day of the experiment, Kinesore was thawed, diluted 1:10 in DMSO, and further diluted 1:100 in Motility Buffer alongside a control buffer containing an equivalent volume of DMSO. Motility videos and still images of MTs were acquired using a custom-built dual-camera TIRF microscope (29,30) with 532-nm excitation to image MTs and 639-nm excitation to image Qdots. Videos of Qdot motility were acquired at 10 frames per second.

Data analysis

Qdot motility was analyzed using the Multi Kymograph plugin for ImageJ (31), as published previously (20). Run length and velocity data are plotted as dot plots with the median and upper and lower quartiles overlaid on the plot. To determine MT landing rates, the number of events seen on a single kymograph was divided by the product of the length of the MT in micrometers and the duration of the movie in minutes. For each condition of interest, multiple kymographs were analyzed from multiple independent videos, and a mean landing rate and standard deviation were determined based on the landing rate observed in each kymograph in each condition.

Liposome and kinesin-liposome complex preparation

Liposomes used in this study were prepared identically to those described previously (20,29,30,32). The liposomes are stored at room temperature in a dark drawer for up to 1 week of experiments. This liposome preparation yields 400 μL of liposomes at a final concentration of 10 nM. For fluorescence-based motor counting assays (see below), liposomes were prepared with the fluorescent dye omitted to ensure that no background fluorescence interfered with motor counting.

To prepare kinesin-liposome complexes, either KinΔC or K543 aliquots were thawed and clarified as described above and diluted in Buffer 1 to the desired concentration (250 nM–1 μM) depending on the kinesin-liposome ratio required for the experiment. 4 μL of diluted kinesin is mixed with 16 μL of liposomes (10 nM) and 20 μL of Buffer 2, and the mixture is incubated on ice for at least 1 hour. Motor counting experiments were performed as described previously (20,29,30). Briefly, KinΔC with YFP-KLCs or a K543 with an N-terminal YFP was clarified, diluted in Buffer 1 as described above (300 nM–2.5 μM), and incubated with nonfluorescent liposomes. Flow chambers were prepared using plasma-cleaned cover glass. MTs, prepared as described previously (20) with the single change of omitting fluorescently labeled tubulin, were affixed to the flow cell surface via an antitubulin antibody as described above, to promote landing of the kinesin-liposome complex onto the slide surface. YFP-KinΔC-liposome complexes were diluted 1:20 into Motility Buffer without MgATP, perfused into the chamber, and immediately imaged by TIRF microscopy with 488-nm excitation on a Nikon N-STORM microscope at ~60 frames per second.

Monitoring the integrated YFP intensity over time for a spot corresponding to a liposome yields the photobleaching transient for that liposome (Fig. S1 A). Because the number of fluorescent molecules on each liposome is too large to allow for clear observation of individual photobleaching steps (29), we used a statistical photobleaching technique developed by Nayak and Ruttenberg (33), which we have used and described the analysis routine for in detail previously (20,29,30,32). By performing this analysis across a range of kinesin to liposome ratios, we generated a standard curve of the number of bound kinesins versus the kinesin to liposome ratio in the incubation tube (Fig. S1 B and C).

Multimotor liposome motility assays and data analysis

Liposome motility assays were performed using DiI-labeled liposomes as described previously (20). Briefly, on the day of the experiment, an aliquot of kinesin was thawed, clarified, and diluted as described above. 4 μL of diluted kinesin (concentration range 0.4–1.6 μM) was mixed with 16 μL of liposomes (10 nM) and 20 μL of Buffer 2 and set to incubate on ice for at least 1 h. Motility chambers were prepared as described above for single-molecule motility assays. Immediately before imaging, liposomes were diluted 1:20 into Motility Buffer and perfused into the motility chamber before imaging by TIRF microscopy on the same custom microscope described above with 532-nm excitation to image MTs and 639-nm excitation to image the DiI liposomes, and motility was recorded at a framerate of 10 frames per second.

Data analysis

As in the single-molecule motility experiments, liposome motility data were analyzed via kymography. Multimotor runs were much more likely to reach the end of the MT (34–36), so end events were clearly marked in the analysis spreadsheet. Run length and velocity data are shown as dot plots with overlaid median and upper and lower quartiles. End event frequencies are shown as bar graphs.

Optical trapping assays

We use 500-nm lipid-coated silica microspheres as model cargo for both single- and multimotor trapping experiments, prepared as described previously (20,30), based on published protocols (37,38). Briefly, DiI liposomes are prepared but not extruded through the 200-nm-pore-size extrusion filter (T&T Scientific). Instead, liposomes are washed three times by ultracentrifugation (392,000 × g for 10 min) to remove excess SH-NaV and resuspended in Buffer 3 (10 mM HEPES (pH 7.2) and 150 mM NaCl). The liposomes are sonicated in an ice bath at low power with 0.5-s pulses for a total sonication time of 10 min, centrifuged for 10 min at 5000 × g, and transferred to a fresh tube. 50 μL of 500-nm-diameter silica microspheres (Duke Standard) are washed with methanol, dried in a speedvac (Rotovap;Eppendorf), and resuspended in 200 μL of Buffer 3. The sonicated liposomes are incubated at 60° C for 2 min, mixed with the silica microspheres, and immediately vortexed and briefly sonicated. The mixture is then shaken at room temperature for 1 h, washed three times by low-speed centrifugation, and resuspended in PBS (pH 7.4) before storage at 4° C with agitation for up to 3 days.

Optical trapping assays were performed as published previously (20). Briefly, flow chambers were prepared using silanized cover glass. On an experiment day, an aliquot of kinesin was mixed with lipid-coated silica microspheres in 20-fold molar excess, whereas single-molecule force assays were performed with limiting kinesin, as confirmed by screening microspheres to find an incubation ratio where ~1 of every 10 beads generated force. Flow cells were coated with 0.8% w/v antitubulin antibody (Bio-Rad YL1/2) and were blocked with 1 mg/mL Casein (Sigma Aldrich Cat. No. C0406) in Buffer 2, and rhodamine-labeled MTs (1:400 in BRB-80 with 20 μM paclitaxel) were introduced in two sequential flows to align the MTs along a single axis. The motility chamber was then perfused with Motility Buffer before adding a small volume of kinesin-bead complex to one end of the chamber. Doing this creates a “bead front” and makes it easier to trap a single lipid-coated silica microsphere at a time. For experiments performed with Kinesore, a Kinesore stock of 50 mM was prepared in DMSO and kept at −20° C. On the day of the experiment, Kinesore was thawed and diluted 1:10 in DMSO, and the resultant Kinesore dilution was further diluted 1:100 in Motility Buffer alongside a control buffer containing an equivalent volume of DMSO. Optical trapping assays were performed on a commercially available optical trap (Lumicks C-Trap System), which is controlled using the supplied software, BlueLake (Lumicks), as published previously (20). The trap stiffness used throughout this study ranged from ~0.04 to 0.06 pN/nm.

Data analysis

Optical trapping data were analyzed using a custom-written R script as published previously (20). Detachment force distributions were fit using the fitdistr function of the MASS packing in R (39), using a single Gaussian for single-motor experiments and a Gaussian mixture model for multimotor experiments. Log-likelihood ratio testing was performed to determine if the inclusion of a third Gaussian was statistically justifiable (39).

3D MT intersection assay

3D MT intersection assays were performed as described previously (20). Briefly, crossflow motility chambers with two perpendicular flow channels were assembled as published previously (20,30,32). Poly-L-lysine-coated pedestal beads were prepared as described previously (20,30,32) and introduced to the flow cell to suspend MTs above the glass surface. After this, to block the surface in both flow channels, 1 mg/mL BSA in Buffer 2 was infused and incubated for 5 min and then washed with Wash Buffer. MTs labeled with Alexa-647 (PurSolutions) were prepared as described previously (20) and diluted 1:100 in BRB-80 buffer (pH 7.2) with 20 μM paclitaxel. Diluted MTs were perfused through one channel of the flow cell, incubated for 2 min, and then washed with Wash Buffer. This process was then repeated in the same channel, before being repeated twice more in the perpendicular channel. Immediately before imaging, both channels were perfused with STORM buffer (Buffer 2 supplemented with 20 μM paclitaxel, 50 mM beta-mercaptoethanol, and 20 mM Cysteamine). After the stochastic optical reconstruction microscopy (STORM) image was acquired, both flow channels were washed with Motility Buffer. DiO-labeled fluorescent liposomes decorated with an average of 10 kinesins each were then diluted 1:100 in Motility Buffer and perfused into the flow cell and imaged.

Imaging was performed on a Nikon N-STORM microscope equipped with a precise piezoelectric stage and a cylindrical lens as published previously (20,32). Z-calibration was performed as described previously (40,41). STORM images of the suspended Alexa-647 MTs were collected by exciting with 405- and 647-nm light and collecting a stack of 20,000 images at a high (~60 Hz) framerate. Liposomes were imaged using 561-nm excitation at a framerate of 10 Hz. Because the pedestal beads were visible in both the 561-nm and 647-nm channels, they are used as fiducial markers to align the two channels in the X-Y plane, whereas the built-in Nikon Perfect Focus system ensured that both image stacks were taken at identical Z heights.

Data analysis

Data analysis was performed as described previously (20). Briefly, super-resolution particle localizations for the STORM reconstruction of the MTs were performed using the DoM (Detection of Molecules) plugin for ImageJ (42). High-resolution tracking of liposomes was also performed using DoM. Liposome localizations were linked into tracks using the “Link Particles to Tracks” function built into DoM. To analyze intersection outcomes and the spatial relationship of each liposome to the intersecting MTs, custom-written R scripts were used as described previously (20). Only intersections where the angle between the MTs was between 60° and 120° were included in the analysis. Liposome pausing in 3D intersections was measured by kymography. For a pause to count, motion had to appear stalled in the kymograph for at least 3 frames. Outcomes were determined by comparing the last tracked position of the liposome to the lines defining the intersecting MTs and the point of the intersection itself. If the tracked liposome position remained within one liposome diameter (350 nm) of the intersection, the outcome was scored as a termination. Otherwise, the liposome was assigned a straight or turn outcome if the liposome trajectory ended associated with the starting MT or switched to the intersecting MT, respectively.

Size exclusion chromatography

200 μg of 1.6 mg/mL mouse kinesin-ΔC with bound KLC was applied to a 10 × 300 mm Superose 6 (Amersham Biosciences) using an AKTA FPLC (Amersham Pharmacia Biotech). The column was equilibrated with BRB80 (80 mM Pipes (pH 7.2), 1 mM EGTA, 1 mM MgCl₂, and 1 mM dithiothreitol) and the protein eluted at a speed of 0.4 mL/min. Each 0.5-mL fraction was concentrated, and the total amount in each fraction was applied to a Nupage 4%–12% Bis-Tris SDS Gel (Invitrogen). The same experiment was repeated in the presence of 50 μM Kinesore (50 mM stock in DMSO) (Tochris Bioscience) in the column equilibration and elution buffers.

Mechanistic mathematical modeling

The mathematical model and simulation used in this study to predict single and multimotor transport along single MTs and in 3D MT intersections are identical to that which we previously developed to describe similar transport properties of the shorter, constitutively active kinesin-1 construct, K543 (20). Briefly, the liposome and MTs are modeled as rigid bodies, with each kinesin a 25-nm spring-like rod that freely diffuses on the liposome surface and can rotate about its attachment point to the liposome. Kinesin MT attachment occurs at a maximum rate (k_a,0) when the distance between the liposome and the MT equals the length of the kinesin and more slowly at shorter and longer distances (20), which differs from other models that allow maximal MT binding at any distance shorter than the tether length (43,44) and is why the k_a,0 we use is higher than in other models. Diffusion of the kinesins on the liposome surface and positional and rotational fluctuations of the liposome are modeled explicitly, based on the liposome and kinesin physical properties using an Euler-Maruyama scheme (43–45). Additionally, the force experienced by each kinesin is calculated in each timestep of the simulation, Δt = 5 × 10⁻⁷s. Each kinesin obeys a three-state mechanochemical kinetic scheme (46), which includes the force dependence of each state transition rate, with chemical reactions simulated at a timestep Δt = 1 × 10⁻⁵ s.

To simulate liposome transport by KinΔC, all parameters in the model were unchanged (20), save either the number of kinesins (N) or the attachment rate (k_a,0) as described below to reflect the threefold reduction in the attachment rate we observed in the landing assay (see Results and Discussion). Using this model, we simulate unloaded liposome transport along a single MT, transport along a single MT hindered by the load from an optical trap, and liposome transport through a 3D MT-MT intersection. Each simulation begins with a single kinesin bound to the MT and ends when all kinesins detach from the MT, or after 15 s has elapsed. We performed two sets of simulations. In the first set, we performed the simulations with N = 3 kinesin motors and then again with N = 4 kinesin motors (single MT run length and velocity and MT-MT intersection outcomes) and with N = 7 kinesin motors (optical trap detachment force assay). This approach allowed us to investigate whether our experimental observations with KinΔC could be explained by a threefold reduction in motor number with respect to our observations with K543, since we can explain these latter observations with the model having N = 10 kinesin motors in the single MT and MT intersection measurements and N = 20 kinesin motors in the laser trap (20). In the second set of simulations, we performed simulations with a threefold reduction in attachment rate, k_a,0 = 50 s⁻¹, compared with the model that successfully described our observations with K543 that used k_a,0 = 150 s⁻¹.

RESULTS AND DISCUSSION

KinΔC-liposomes go straight or terminate but rarely turn in 3D MT intersections

Kinesin-1 motors must navigate their vesicular cargo through multiple 3D MT intersections en route to the cargo’s destination (21–24). If kinesin-1 motors exist in an equilibrium between the autoinhibited and active states (Fig. 1 A) while attached to a vesicle, does this equilibrium impact how the vesicle is maneuvered through a MT intersection? To address this question in vitro, we suspended MTs between poly-L-lysine-coated silica beads on a coverslip surface, creating orthogonal 3D MT intersections (Fig. 1 B and C). We then decorated 350-nm fluid-like (DOPC) liposomes (Fig. 1 B) with one of two kinesin-1 constructs that could freely diffuse on the liposome surface, which were present at ~10 motors per liposome, based on a photobleaching-based counting assay (Fig. S1). The ubiquitously expressed murine kinesin-1 heavy chain, KIF5B, was the basis for both constructs (Fig. 1 A). The KinΔC construct (1–888 amino acids) with full-length KLC2 bound retains the capacity to be partially autoinhibited, based on a threefold lower landing rate compared with the constitutively active truncated construct K543 (Fig. 3 C). Both constructs are C-terminally biotinylated to provide an attachment handle to the liposome (20,25).

FIGURE 1. KinΔC terminates frequently in 3D MT intersections.

(A) Schematic of constructs of interest. KinΔC (left) contains aa1–888 of murine KIF5B and copurifies with full-length KLC2. Critically, KinΔC is able to adopt the folded conformation associated with autoinhibition. K543 contains aa1–543 of murine KIF5B and serves as a constitutively active control. Both constructs are C-terminally biotinylated (black circles). (B) Schematic cartoon of 3D MT intersection assay. Liposome (gold) is transported by KinΔC (purple) along MTs (blue), whereas KinΔC motors can diffuse on the liposome surface. The liposome is moving out of the page and approaches the horizontal MT in a 3D intersection. The geometry of the interaction between the liposome and the intersection is described by two parameters, d and α, illustrated here, where d is the vertical gap between the intersecting MTs, and α is the angle of approach of the liposome coming in to the intersection, where an α of 0° indicates the liposome is pointed up toward the crossing MT, and an α of 180° indicates the liposome is pointed down and away from the crossing MT. (C) 3D STORM reconstruction of suspended MT intersections. Z-position of MTs is shown via the color bar on the right. The white circles show the approximate outlines of the pedestal beads from which the MTs are suspended. Scale bar = 3.0 μm. (D) Example of a liposome (yellow) passing straight through a 3D intersection. Color bar as in (C). Gap = 34 nm. Scale bar = 500 nm. (E) Example of a liposome (yellow) turning in a 3D intersection. Color bar as in (C). Gap = 174 nm. Scale bar = 500 nm. (F) Bar graph of 3D intersection outcomes for liposomes with 10 KinΔC (left) compared with 10 K543 control (right). 10 K543 control data are replotted from Bensel et al., 2024. For KinΔC data, N = 57 intersection outcomes from three independently performed intersection assays. Error bars represent 95% confidence intervals determined by bootstrapping with 2000 sampling repetitions. (G) Cumulative distribution plot of the pause lifetimes observed for liposomes with 10 KinΔC (cyan) and 10 K543 (magenta) in 3D intersections. Events with no discernible pause are represented as a pause lifetime of 0 s. K543 lifetime data are replotted from Bensel et al., 2024. Pauses are shown as mean ± SEM. The pause lifetime distributions were compared using a Kolmogorov-Smirnov test, which gives a p-value of 0.124.

FIGURE 3. KinΔC and K543 are indistinguishable once engaged with the MT.

(A, upper) Schematic of landing rate assay as performed with KinΔC. KinΔC is bound to Qdots (red) and flowed into a microscopy chamber containing MTs (blue). (A, lower) Example kymograph of single-molecule Qdot assay with KinΔC. Runs are boxed in yellow. Distance along MT is represented by the vertical axis (scale bar = 4 μm), whereas time is represented as the horizontal axis (scale bar = 4 s). A moving particle appears as a slanted line, whereas a stuck particle appears as a hotizontal line. (B, upper) Schematic of landing rate assay as performed with K543. K543 is bound to Qdots (red) and flowed into a microscopy chamber containing MTs (blue). (B, lower) Example kymograph of single-molecule Qdot assay with K543. Runs are boxed in yellow. Axes and scale bars are as in (A). (C) Dot plot with overlaid boxplot for landing rate determined for KinΔC (left) and K543 (right). Each dot represents the landing rate determined for a single MT in a single video with the equation: Landing Rate = N_runs/[MT length (μm) × video length (min)]. Statistical analysis was performed using Kruskal-Wallis with Dunn’s post hoc test using Benjamin-Yuketieli adjustment, p =0.01. N =12 MTs (KinΔC) or seven MTs (K543) from three independent videos each. (D) Schematic representation of single-molecule optical trapping assay as performed. A lipid-coated silica bead (gray) with a bound KinΔC (purple) is brought close to a surface-bound MT (blue) by the laser trap (red). (D, inset) Sample force ramp collected from a single KinΔC molecule in the optical trapping assay. Median-filtered trace (black) is overlaid on top of raw data (gray). (E) Histogram with overlayed fit of detachment forces measured for single KinΔC molecules. KinΔC has a fitted mean detachment force of 6.3 ± 1.7 pN. N_events = 109 from four independent experiments. (F) Histogram with overlayed fit of detachment forces measured for single K543 molecules. K543 has a fitted mean detachment force of 5.7 ± 1.1 pN. K543 run detachment force data are replotted from Bensel et al., 2024.

Using 3D STORM, the liposome’s spatial relationship to the MT intersection was defined, that is, the liposome approach angle (α) and vertical gap (d) between the MTs (Fig. 1 B and C) (20,30). This ensured that we only considered directional outcomes (i.e., whether the liposome went straight (Fig. 1 D, Video S1), turned (Fig. 1 E, Video S2), or terminated its run) that resulted from motors on the liposome surface being able to reach the intersecting MT. For each event that met our criteria, we recorded if the liposome paused, how long it paused, and the directional outcome.

At 3D MT intersections, KinΔC-liposomes were nearly as likely to terminate (48%) as to go straight (43%) but rarely turned onto the crossing MT (9%) (Fig. 1 F). These directional outcomes were distinctly different than we previously reported for K543-liposomes, which preferred to go straight (57%) or turned occasionally onto the crossing MT (31%) but rarely terminated (12%) (Fig. 1 F) (20). Could pausing in the intersection offer insight to these constructs’ different directional outcomes? KinΔC-liposomes were less likely to pause (48% paused) compared with K543-liposomes (70% paused) (20), but there was no statistically significant difference between the pause lifetimes (Fig. 1 G). We observed slightly longer pause durations with both KinΔC-liposomes and K543-liposomes compared with a recent study that used full-length KIF5A without KLC affixed to silica microspheres, which could be explained either by the different kinesin construct used or by the different cargo (45). Although 43% of the pauses recorded for KinΔC-liposomes and 45% of K543-liposome pauses correspond to turn outcomes, only 26% of KinΔC-liposome pauses correspond to straight outcomes, with 30% corresponding to terminate outcomes versus 52% and 3% of K543-liposome pauses corresponding to turn and terminate outcomes, respectively (Fig. S2). Our previous in silico modeling of K543 transport in 3D MT intersections suggested that pausing originates from motors on the liposome surface binding simultaneously to both MTs and engaging in a tug-of-war, which must be resolved before a directional outcome is determined (20,23,45). The observation that KinΔC-liposomes rarely turn suggests that the motors’ autoinhibition reduces the number of available KinΔC motors that can engage the crossing MT, resulting in less frequent tug-of-wars, pausing, and turning events. Furthermore, the higher proportion of pauses leading to termination outcomes for KinΔC-liposomes suggests that the lack of available KinΔC motors due to autoinhibition reduces the ability of KinΔC ensembles to resolve a tug-of-war (Fig. S2).

Another factor that may be important in regulating cargo outcomes at 3D MT intersections is the role of forces perpendicular to the MT surface. In brief, recent experimental (47) and theoretical (48) studies have generated evidence suggesting that kinesin’s detachment rate from the MT is primarily determined by the component of the force applied to the kinesin that is perpendicular to the MT surface rather than the component that is along the length of the MT. During a tug-of-war, a kinesin with a longer tail could have a smaller component of its load perpendicular to the MT surface and therefore might detach from the MT more slowly, which could also be important for explaining why KinΔC teams generate different outcomes from K543 teams in 3D MT intersections. However, the exact parameter describing the vertical load sensitivity of kinesin has not yet been determined, so it remains challenging to explore this possibility further.

KinΔC ensembles have reduced MT engagement

For kinesin-1 transported cargos, the run length, or characteristic travel distance, scales with the number of motors that are MT engaged (49–52), whereas velocity slows due to mechanical coupling creating internal loads between the motors (20,53,54). Therefore, we measured KinΔC-liposome run lengths and velocities along single MTs adhered to a glass slide for varying size of KinΔC ensembles (~5, ~10, or ~20 motors, Fig. 2 A) compared with a single KinΔC motor transporting a quantum dot (Qdot) cargo. As expected, KinΔC ensembles universally had longer run lengths (1.99–2.98 μm) and slower velocities (637–702 nm/s) than single KinΔC motors (1.19 μm and 890 nm/s) (Fig. 2 B and C). However, compared with K543-liposomes transported by similar ensemble sizes, KinΔC-liposomes had shorter median run lengths (Fig. 2 B; Table S1), although velocities were somewhat similar (Fig. 2 C; Table S1). In fact, these run length differences may be underestimated, as run lengths can be limited by the MT length itself (34–36). Knowing that MTs were prepared identically with similar length distributions (Fig. S2), we compared the frequency at which liposomes reached the MT end (Fig. 2 D). For all motor ensemble sizes, K543-liposomes were ~46% more likely than KinΔC-liposomes to reach the MT end (Fig. 2 D), suggesting that the K543-liposomes run lengths are more highly underestimated than those of the KinΔC-liposomes. Therefore, the reduction in KinΔC-liposome run lengths compared with K543-liposomes is even greater. These data support the hypothesis that fewer KinΔC motors from the ensemble engage the MT compared with the constitutively active K543 control.

FIGURE 2. KinΔC ensembles favor fewer engaged kinesins than K543 ensembles.

(A) Schematic representation of multimotor motility assay with fluid-like liposome as cargo. Liposome (gold) has 5–20 kinesins (purple) bound to its surface. The fluorescence of the liposome is tracked by TIRF microscopy as it is moved along MTs (blue) by the bound kinesins. (B) Dot plots with overlaid violin plots representing multimotor run lengths measured for KinΔC (cyan) and K543 (magenta) with up to 20 bound kinesins. Black overlaid boxplot shows median and upper and lower quartiles per condition. K543 control data are replotted from Bensel et al., 2024. p-values are as follows: N.S., p > 0.05; *p < 0.05; **p < 0.01. Run length data are compared using a Kruskal-Wallis with Dunn’s post hoc test using the Benjamin-Yuketieli adjustment. See Table S1 for exact p-values. For KinΔC data, N_events ranges from 137 to 480 from at least three independent experiments. (C) Dot plot with overlaid violin plot representing multimotor velocities measured for KinΔC (cyan) and K543 (magenta). Black overlaid boxplot shows median and upper and lower quartiles per condition. K543 control data are replotted from Bensel et al., 2024. p-values are as follows: N.S., p > 0.05; *p < 0.05; **p < 0.01. Velocity data are compared using a Kruskal-Wallis with Dunn’s post hoc test using the Benjamin-Yuketieli adjustment. See Table S1 for exact p-values. N values are the same as in (B). (D) Bar graph depicting the percentage of runs for each condition (single kinesin Qdot, and 5, 10, and 20 kinesins per liposome), which reach the end of the MT before detaching. K543 data are replotted from Bensel et al., 2024. End event frequencies are compared using a two-proportions Z-test. p-values are as follows: N.S., p > 0.05; *p < 0.05; **p < 0.01. See Table S1 for exact p-values. N values are the same as in (B) and (C). (E) Schematic representation of multimotor optical trapping assay with lipid-coated bead as cargo. Kinesins (purple) are bound to lipid-coated beads (gray) in a 20 to 1 excess. The laser trap (red) is used to position the bead close to an MT (blue), and force ramps are recorded. (F) Sample force ramp collected for a lipid-coated bead coated with 20 KinΔC motors. Raw data are shown in gray with median-filtered data overlaid in black. (G) Detachment force histogram with overlaid double-Gaussian fit collected for KinΔC experimental results (cyan). Fit parameters ± fitting errors are shown in inset. A₂ is defined as 1 − A₁ and is not an independently fit parameter with error determined by error propagation. N_events = 169 from five independent experiments. (H) Detachment force histogram with overlaid triple Gaussian fit collected for K543 (magenta). K543 data and fit are replotted from Bensel et al., 2024. Fit parameters ± fitting errors are shown in inset. A₃ is defined as 1 − (A₁ + A₂) and is not an independently fit parameter with error determined by error propagation.

To effectively count how many KinΔC motors in the ensemble are MT engaged, we measured the cargo’s detachment force in an optical trapping assay, as this force should scale with the engaged motor number (50,51,55) (Fig. 2 E). Because the surface area of the 500-nm lipid-coated beads used for optical trapping is twice that of the 350-nm liposomes used for conventional motility assays, we incubated 500-nm lipid-coated beads with 20-fold excess KinΔC to match the number of motors per unit surface area of a 350-nm liposome with 10 bound motors and then measured the detachment force as the KinΔC motors step against the hindering load of the trap (Fig. 2 E and F). The KinΔC-bead detachment forces were distributed as the sum of two Gaussians centered at 6.5 and 12.6 pN, with 80% of the events in the first peak (Fig. 2 G), whereas K543-bead detachment forces were described by a triple Gaussian with peaks at 5.7, 11.4, and 16.4 pN (Fig. 2 H). The initial detachment force peaks for the KinΔC and K543 constructs were equivalent to that measured for a single motor (KinΔC: 6.4 ± 1.9 pN; K543: 5.7 ± 1.1 pN) (Fig. 3 E and F) and similar to that in the literature (56–58). Therefore, often only one but at most two KinΔC motors are MT engaged compared with K543 ensembles where up to three motors can be MT engaged. These data once again suggest that autoinhibition effectively reduces the number of active motors.

KinΔC has a reduced MT landing rate but once MT engaged is indistinguishable from K543

If kinesin-1 can transition between autoinhibited and active states, does this dynamic equilibrium affect how often the motor engages a MT and its motile properties once engaged with the MT? To quantify MT engagement, we measured the landing rate (number of processive runs per unit time per unit MT length) of single Qdot-labeled KinΔC (1 nM) relative to K543 (1 nM) in saturating MgATP (Fig. 3 A and C). KinΔC initiated processive runs at a rate (0.08 μm⁻¹ min⁻¹) about threefold less than K543 (0.23 μm⁻¹ min⁻¹) (Fig. 3 B and C, Video S3), suggesting that kinesin-1 autoinhibition reduces MT engagement. Once engaged, the mechanochemical properties of single KinΔC motors were not impacted by autoinhibition, with median run lengths (RLs), mean velocities (Vs), and detachment forces (Fs) (RL = 1.19 μm; V = 890 nm/s; F = 6.3 ± 1.7 pN) (Figs. 2 B, C and 3 E) that were indistinguishable from K543 (RL = 1.19 μm; V = 868 nm/s; F = 5.7 ± 1.1 pN) (Figs. 2 B, C and 3 F). Once KinΔC engages with the MT, the motor’s processivity and force generation are governed by the mechanochemistry of the fully active motor, suggesting that transitions back into the autoinhibited state must occur only in solution, after MT detachment. If the reduced rate of MT association reflects a dynamic equilibrium between the autoinhibited and active states, these data do not inform how rapidly the two states exchange (see mechanistic mathematical modeling section below). Nonetheless, the reduced landing rate of KinΔC suggests that autoinhibition can effectively reduce the number of MT-engaged motors in an ensemble and thus the ensemble’s cargo transport capacity as described above.

Kinesore treatment activates an autoinhibited KinΔC motor

If autoinhibitory intramolecular interactions are responsible for KinΔC’s reduced MT engagement, disrupting these interactions should activate the motor. To test this idea, we used Kinesore, a small molecule that interacts with the KLC to mimic cargo binding and disrupt inhibitory intramolecular interactions to activate kinesin-1 (26,59). In the presence of 50 μM Kinesore, KinΔC’s landing rate increased significantly to 0.26 μm⁻¹ min⁻¹, equal to that of K543 (0.22 μm⁻¹ min⁻¹) (Fig. 4 A and B; Videos S3 and S4), which was unaffected by Kinesore, as this construct lacks KLC. By restoring the KinΔC MT landing rate, would Kinesore treatment of a KinΔC ensemble increase its force-bearing capacity (i.e., increased number of MT-engaged motors)? This was in fact the case as the distribution of detachment forces measured for KinΔC-beads in the optical trap with 50 μM Kinesore present was now best fit by the sum of three rather than two Gaussians, centered at 7.2 pN, 13.7 pN, and 20.0 pN (Fig. 4 C and D), similar to that for K543-beads (Fig. 2 H). Therefore, the number of MT-engaged KinΔC motors increased from at most two in the absence of Kinesore to up to three engaged motors with Kinesore present.

FIGURE 4. Kinesore restores the MT engagement activity of KinΔC to that of K543.

(A) Representative landing rate assay kymographs for KinΔC in the presence (left) and K543 in the absence (right) of 50 μM Kinesore. Distance is shown on the horizontal axis (scale bar = 4 μm), whereas time is shown on the vertical axis (scale bar = 4 s). (B) Dot plot of landing rates measured in the absence (−) and presence (+) of 50 μM Kinesore for KinΔC (cyan, turquoise) and K543 (magenta, burgundy). Each dot represents the landing rate determined from a single MT in a single video. Statistical analysis of landing rate data was performed using Kruskal-Wallis with Dunn’s post hoc test using the Benjamin-Yuketieli adjustment. p-values are as follows; N.S., p > 0.05; **p < 0.01. KinΔC +Kinesore, N_events =11 MTs from three independent videos; K543 + Kinesore, N_events =10 MTs from three independent videos. (C) Sample force ramp recorded from a lipid-coated bead being moved by 20 KinΔC motors in the presence of Kinesore. Raw data are shown in gray, and median-filtered data are overlaid in black. (D) Histogram with overlaid triple-Gaussian fit for detachment forces measured KinΔC in the presence of Kinesore (turquoise) with the double-Gaussian fit to detachment forces measured for KinΔC in the absence of Kinesore (cyan) shown as reference. Fit parameters ± fitting errors are shown in inset. A₃ is defined as 1 − (A₁ + A₂) and is not an independently fit parameter with error determined by error propagation. N_events = 418 from four independent experiments. (E) SDS-PAGE of size exclusion chromatography elution fractions for KinΔC in the absence (top) or presence (bottom) of 50 μM Kinesore. Elution fractions are shown increasing from left to right, with the band corresponding to the KinΔC heavy chain (KHC ΔC) shown. Given the same molecular weight, a molecule will elute sooner the more extended it is.

Implicit in the activation of KinΔC by Kinesore is that disrupting the autoinhibitory intramolecular interactions should result in a more extended molecule (15,17,26). We utilized size exclusion chromatography to assess the relative shape of KinΔC molecules in the presence or absence of 50 μM Kinesore. A molecule with an extended structure elutes sooner than a molecule of the same molecular mass with a folded structure (Fig. 4 E). Elution fractions from the size exclusion chromatography column equilibrated and eluted in the absence (Fig. 4 E, top) versus presence of 50 μM Kinesore (Fig. 4 E, bottom) show that KinΔC elutes earlier in the presence of Kinesore. These data suggest that Kinesore binding to KLC results in a more extended conformation, consistent with an active KinΔC motor.

KinΔC autoinhibition: Mechanistic model and impact on liposome transport

The transport data for single KinΔC motors and ensembles together suggest that KinΔC can adopt the folded and extended states even when attached to cargo. To understand how KinΔC autoinhibition impacts ensemble transport, we took advantage of our previously published mechanistic in silico model that predicted single-motor and ensemble K543 transport (Fig. 5 A) on single MTs and in 3D MT intersections (see materials and methods for details of simulations) (20). Crucially, the model is predictive because it is constrained only by a three-state model of kinesin’s mechanochemistry (46) and the physical properties of the liposomes and MTs. Adopting this model to predict KinΔC transport was simplified knowing that once engaged with the MT, the mechanochemical properties of KinΔC are identical to K543. Therefore, the in silico model was modified to include KinΔC autoinhibition as a dynamic equilibrium between an inhibited and an active state where the exchange rates between states are either slow on the timescale of the simulation such that KinΔC exists predominantly in the inhibited state (i.e., effectively reducing the number of active motors in the ensemble) (Fig. 5 E) or the exchange rates are much faster than the timescale of the simulation such that autoinhibition can be modeled simply as a reduced apparent attachment rate (Fig. 5 F). In support of these two concepts, others in the field have proposed both motor number (60) and reattachment kinetics (52,61) as important determinants of ensemble transport properties.

FIGURE 5. Mechanistic modeling of kinesin-1 autoinhibition predicts experimental differences observed between KinΔC and K543.

(A) Schematic representation of our mechanistic model of kinesin-1 transport as published in Bensel et al., 2024. All motors in the model are active and bind MTs with a maximum rate k_a = 150 s⁻¹. We refer to this as the “fully active” model. (B) Dot plot with overlaid boxplot and violin plot of KinΔC (left) and modeled (right) single-molecule run length. Model N_events = 46. (C) Dot plot with overlaid boxplot and violin plot of KinΔC (left) and modeled (right) single-molecule velocity. Model N_events = 46. (D) Histogram of modeled detachment force distribution with overlaid single-Gaussian fit (black, dashed). Single-Gaussian fit to KinΔC detachment force experimental data is shown for reference (black, solid). Model N_events = 99. (E) Schematic representation of our modified mechanistic model of kinesin-1 that emulates an autoinhibited motor where exchange between the active and inhibited states occurs slowly. We call this model “Slow k₊/k₋,” and it is implemented by reducing the number of motors in the simulation by a factor of 3 (i.e., N/3 motors). All other model parameters remain the same. (F) Schematic representation of our modified mechanistic model of kinesin-1 that is modified to capture an autoinhibited motor that exchanges rapidly between the active and inhibited states. We call this model “Fast k₊/k₋,” and it is implemented by reducing maximum rate of MT association by a factor of 3 to k_a = 50 s⁻¹, but no other parameters of the model are changed. (G) Dot plot with overlaid violin plot and boxplot of run lengths measured for experimental (Expt) and modeled liposomes with ~10 bound motors. KinΔC experimental results are shown for reference (cyan) next to the model of Slow k₊/k₋ (purple) (N_events = 93) and Fast k₊/k₋ (dark blue) (N_events = 50). K543 experimental results (magenta) and fully active model results (orange) are replotted from Bensel et al., 2024 and are shown for reference. Statistical analysis of run length data was performed using Kruskal-Wallis with Dunn’s post hoc test using the Benjamin-Yuketieli adjustment. p-values are as follows; N.S., p > 0.05; *p < 0.05. (H) Dot plot with overlaid violin plot and boxplot of velocities measured for experimental (Expt) and modeled liposomes with ~10 bound motors. KinΔC experimental results are shown for reference (cyan) next to the model of Slow k₊/k₋ (purple) and Fast k₊/k₋ (dark blue). K543 experimental results (magenta) and fully active model results (orange) are replotted from Bensel et al., 2024 and are shown for reference. Statistical analysis of run length data was performed using Kruskal-Wallis with Dunn’s post hoc test using the Benjamin-Yuketieli adjustment. p-values are as follows; N.S., p > 0.05. (I) Histogram of experimentally measured detachment forces for ensembles of 20 KinΔC with overlaid model prediction fits. The model of Slow k₊/k₋ is shown in purple (N_events = 200), Fast k₊/k₋ is shown in dark blue (N_events = 200), and the fully active model is shown in orange. Fully active model fit is replotted from Bensel et al., 2024. (J) Bar graph of 3D intersection outcomes for experiments and simulations. Slow k₊/k₋ N_events = 285; fast k₊/k₋ N_events = 260. Error bars represent 95% confidence intervals determined by bootstrapping with 2000 sampling repetitions. K543 experimental results and fully active model results are replotted from Bensel et al., 2024.

For single KinΔC motors, the model predicts run lengths, velocity, and detachment forces (RL = 1.66 μm; V = 787 nm/s; F = 6.5 pN) as observed experimentally and as reported previously for K543 (20) (Fig. 5 B–D). To model autoinhibition and its effect on KinΔC MT engagement, we assumed that the threefold lower KinΔC MT landing rate compared with K543 (Fig. 3 C) was due to autoinhibition, as supported by the observation that Kinesore restores the landing rate of KinΔC to that of K543 (Fig. 4 B). Therefore, we first ran liposome transport simulations, as described above, with autoinhibition reducing the number of active motors in the ensemble by a factor of 3 to model a slow exchange rate between the active and inhibited states (Fig. 5 E). Since liposomes were decorated with a 10-motor ensemble, we simulated liposome transport by either three or four motors and combined the results for each transport condition to best reflect the desired 3.33 motors (i.e., 10 motors divided by a factor of 3). Alternatively, we set the maximum MT attachment rate (k_a,0 = 50 s⁻¹) for all motors in the ensemble to one-third of the rate that predicted the K543 observations (k_a,0 = 150 s⁻¹), reflecting a threefold reduction in the apparent attachment rate and modeling rapid exchange between the active and inhibited states (Fig. 5 F).

Simulations for reduced motor number and MT attachment rate reproduced the experimental trends in ensemble RLs and velocities for KinΔC versus K543 (Fig. 5 G and H). Interestingly, the reduced motor number simulations were consistent with the experimentally observed detachment force distributions showing most often one or at most two MT-engaged KinΔC motors (Fig. 5 I). Simulations with a reduced MT attachment rate, although broadly similar to the reduced motor number case, did have a minor population of three simultaneously engaged motors, though less frequently than in the active model (Fig. 5 I). For KinΔC-liposomes in 3D MT intersections, both the reduced motor number and MT attachment rate simulations qualitatively reproduced the predominant directional outcomes of going straight through or terminating in the intersection that were dramatically different than the K543-liposome directional outcomes and those predicted by a model of fully active motors (Fig. 5 J) (20). Interestingly, the 95% confidence intervals of the straight and terminate outcomes for both models overlap with those corresponding to the experimental data set for KinΔC-liposomes (Fig. 5 J), suggesting that the differences between the models, and between the model and the experiment, are small. Given that the model is constrained simply by the biophysical properties of our system and we did not optimize parameters by fitting them to the data, qualitative agreement between the model prediction and experimental observation with only a single parameter changed strongly supports the proposed mechanism.

To understand why reduced motor number or MT association rate affects directional outcomes in 3D MT intersections, we must rely on our in silico model to provide a mechanistic basis for how each directional outcome occurs. In brief, as a liposome enters the intersection, motors on the liposome surface can bind simultaneously to both MTs, and thus the liposome may pause if the motors engage in a tug-of-war. In a tug-of-war, motors stochastically attach and detach from both MTs with the tug-of-war resolved when motors on one of the intersecting MTs all detach, allowing the winning team to pull the liposome straight (Video S5) or turn (Video S7). Generally, if no tug-of-war occurs, the liposome proceeds straight (Video S6). Finally, if all MT-engaged motors detach, transport is terminated (Video S8). If we reduce motor-MT engagement by reducing either the number of available motors or the MT attachment rate, a tug-of-war is less likely to occur, favoring straight outcomes. Furthermore, since both experimental and simulated data suggest that KinΔC-liposome transport is driven most often by a single engaged motor, when the liposome collides with the crossing MT the motor may detach, terminating the run (Video S9). Thus, our model provides a mechanistic basis for kinesin autoinhibition impacting directional outcomes at 3D MT intersections. Interestingly, modeling autoinhibition as an equilibrium between a folded and extended state where the exchange rates between the states are either slow or rapid (see above) captures much of the KinΔC transport behavior (Fig. 5), suggesting that this mode of motor regulation is not sensitive to the exchange rates. The observation that intersection outcomes are better predicted by the model with slow exchange rates suggests that the folded, autoinhibited state resulting from multiple intramolecular interactions is most likely quite stable, as reported in the literature (15–18).

CONCLUSION

Here we provide evidence that kinesin-1 autoinhibition has profound effects on liposome transport in vitro whether it be on a single MT or when challenged by a 3D MT intersection. The partial autoinhibition exhibited by the KinΔC construct is only one example of what is likely a range of motor activation that is possible in cells and depends on kinesin-1 isoform (16), whether or not KLCs are bound (16,17), the cargo adapter that kinesin is bound to (16,19), as well as MAPs such as MAP7 that enhance engagement of kinesin with the MT (16,62,63). Although this study focuses on kinesin-1, the mechanochemical differences in kinesin subfamilies may provide the cell additional capacity to tune cargo transport (61,64,65). Given the cell’s 3D MT network with numerous intersections, optimizing the directional outcome at each intersection, which is sensitive to the effective number of active motors, provides the cell a mechanism for ensuring vesicular cargo delivery to the proper destination.

Supplementary Material

Supporting material can be found online at https://doi.org/10.1016/j.bpj.2025.08.032.

SIGNIFICANCE.

Kinesin-1 vesicular transport through the cell’s complex microtubule (MT) network must be spatially and temporally regulated to meet the cell’s physiological demands. With kinesin-1 being autoinhibited, does this form of regulation impact the directional outcomes of vesicles encountering three-dimensional (3D) MT-MT intersections as they do in cells? To address this in vitro, liposomes transported by kinesin-1 motors (team of ~10), which exhibit autoinhibition, are challenged with 3D MT intersections and, interestingly, preferentially terminate or go straight and rarely turn, whereas liposomes transported by constitutively active motors preferentially go straight or turn but rarely terminate. Therefore, autoinhibition reduces the number of MT-engaged motors within the team and provides the cell a means of fine-tuning cargo transport to its destination.

DATA AND CODE AVAILABILITY

Due to the size of the super-resolution imaging datasets, data are stored on University of Vermont servers. The data are available from the corresponding author upon reasonable request. Modeling code and analysis scripts are available from the corresponding author upon request.