bioRxiv preprint doi: https://doi.org/10.1101/277947; this version posted March 7, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.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 available under aCC-BY-NC-ND 4.0 International license. Tuning molecular motor transport through cytoskeletal filament network organization Monika Scholz,1, 2 Kimberly L. Weirich,1, 2 Margaret L. Gardel,1, 2, 3 and Aaron R. Dinner1, 2, 4, ∗ 1James Franck Institute, the University of Chicago, Chicago, IL 60637 2Institute for Biophysical Dynamics, the University of Chicago, Chicago, IL 60637 3Department of Physics, the University of Chicago, Chicago, IL 60637 4Department of Chemistry, the University of Chicago, Chicago, IL 60637∗ The interaction of motor proteins with intracellular filaments is required for transport processes and force generation in cells. Within a cell, crosslinking proteins organize cytoskeletal filaments both temporally and spatially to create dynamic, and structurally diverse networks. The architecture of these networks changes both the mechanics as well as the transport dynamics; however, the effects on transport are less well understood. Here, we compare the transport dynamics of myosin II motor proteins moving on model cytoskeletal networks created by common crosslinking proteins. We observe that motor dynamics change predictably based on the microstructure of the underlying networks and discuss how this can be utilized by cells to achieve specific transport goals. I. INTRODUCTION tures spatially and temporally to control the motor dy- namics. The interior of cells is organized by cytoskeletal struc- Here, we demonstrate experimentally that actin fila- tures. Protein monomers like actin and tubulin organize ment network organization can modulate transport dy- into polymers, resulting in polar filaments that can span namics of myosin II motors. To isolate the effects associ- cellular dimensions. These filaments interact with molec- ated with bundle architecture from those stemming from ular motor proteins to drive intracellular transport, cel- cellular regulation [17], we track myosin II on model net- lular shape change, and cell motility [1]. These different works of actin filaments bundled by biological crosslink- functions rely on different arrangements of the filaments, ing proteins. The dynamics can be interpreted in terms ranging from tight bundles in filopodia to meshes in the of the known features of the crosslinking proteins. cell cortex [2]. Filaments are organized locally by proteins that can bind to multiple actin filaments to crosslink them into a II. TRANSPORT PROPERTIES DEPEND ON network [2]. The type of crosslinking protein determines THE MICROSCOPIC BUNDLE ARCHITECTURE the resulting network structure. Proteins with a small crosslinking distance, such as fimbrin or α-actinin, create We analyze single-particle tracking data from myosin tightly packed bundles with the filaments spaced 8-36 nm II motors moving on a quasi-two-dimensional (quasi- apart [3, 4]. The crosslinking protein fascin creates sim- 2D) actin network (Fig. 1). The actin filaments are ilarly tight bundles, but the actin filaments are polarity- crosslinked with either α-actinin, fimbrin, fascin or fil- sorted as well [5, 6]. In contrast, filamin has a crosslinking amin to create actin bundles with distinct microstruc- distance of 160 nm, resulting in a loosely bundled struc- tures (Fig. 1). The actin networks and skeletal-muscle ture for low concentrations of crosslinking protein [7]. myosin II were prepared as described in [16]. Single- While the consequences of these diverse structures for particle trajectories were obtained using the Python- network mechanics have the focus of several studies [8{ based implementation of the Crocker-Grier algorithm 13], the consequences for motor transport have received Trackpy [18]. The resulting trajectories align well with less attention [14, 15] . the crosslinked actin filaments (Fig. 1). We recently found that the microscopic arrangement For each experimental condition, we characterize the of filaments can have profound effects on the transport motor motion through the time-averaged mean-squared of motor proteins across filament networks [16]. In par- displacement (TA-MSD) as a function of lag time ∆ and ticular, the presence of filament loops can lead to un- measurement time T as previously described [16, 19, 20]: productive cycling of motors. These vortex-like states 1 Z T −∆ give rise to a power-law distribution of dwell times and hR~(T; ∆)2i = [~x(t + ∆) − ~x(t)]2dt; (1) glassy dynamics in intracellular transport. We showed T − ∆ 0 that we could tune the transport properties of the motors by varying properties of the system that control the num- where ~x(t) is the position of a motor at time t. ber of junctions that can support cycling. We thus pre- The exponent α in the scaling relation hR~(T; ∆)2i / dicted that cells could modulate their cytoskeletal struc- ∆α characterizes the motion: α = 1 for simple diffu- sion, and α = 2 for a purely inertial motion; non-integer values are possible as well (e.g., [19, 20] and references therein). We observe that the exponent of the TA-MSD ∗ To whom correspondence may be addressed: din- as a function of lag time changes depending in the un- [email protected] derlying bundle structure (Fig. 2). The polarity-sorted bioRxiv preprint doi: https://doi.org/10.1101/277947; this version posted March 7, 2018. The copyright holder for this 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 available under aCC-BY-NC-ND 4.0 International license. 2 α-actinin Fimbrin Fascin Filamin Actin Myosin II FIG. 2. (A) Mean-squared displacement of myosin II minifil- FIG. 1. Schematic showing the interactions of four cross- aments on actin networks crosslinked with different proteins linking proteins with actin filaments (red arrows). Fluores- as a function of lag time (∆) and (B) as a function of mea- cence microscopy image of the quasi-2D actin networks af- surement time (T ). T = 9 s and ∆ = 0:2 s for fascin (orange) ter crosslinking with either α-actinin (A), fimbrin (B), fascin and filamin (gray). For α-actinin (blue) and fimbrin (red), (C) or filamin (D). The trajectories of myosin II are overlaid T = 137:7 s and ∆ = 3:06 s. The number of trajectories (red). Note that the scale of the images change between pan- is 210, 251, 236 and 256 for α-actinin, fimbrin, fascin and els, due to the large difference in myosin II speeds on different filamin, respectively. The mean trajectory lengths are 167, crosslinked actin networks. 208, 16 and 17 s for α-actinin, fimbrin, fascin and filamin, respectively. bundles created by the crosslinking protein fascin lead to an exponent close to two, indicating strongly directed motion. In contrast, the exponent of the TA-MSD is III. ANGULAR DISTRIBUTION ILLUSTRATES MICROSCOPIC DIRECTIONAL CHANGES closer to one for filamin, α-actinin and fimbrin. We also consider the TA-MSD as a function of mea- surement time, T . This order parameter is sensitive to The TA-MSD cannot reveal directional correlations the ergodicity of the dynamics. If the properties of the within trajectories, but there is evidence of such corre- transport process underlying the observed dynamics are lations in particle-tracking studies of molecular motors unchanged over the course of measurement, the dynamics [19{24]. To quantify such correlations, we use the rela- are ergodic, and the expected exponent is zero. However, tive angle distribution [20]. The relative angle is defined as previously described for a fimbrin-crosslinked network, by the mean-square displacement decreases as the measure- ~v(t; ∆) · ~v(t + ∆; ∆) ment time increases (i.e., the exponent is less than zero) cos θ(t; ∆) = ; (2) [16]. We find similar exponents for both α-actinin and j~v(t; ∆)jj~v(t + ∆; ∆)j fimbrin, which is unexpected considering the relevant where ~v(t; ∆) = ~x(t + ∆) − ~x(t). The normalized his- scales. We previously described a mechanism for these togram (probability density function, PDF) of θ values non-ergodic dynamics that relies on motors interacting for successive vectors within trajectories can be used as with at least two filaments at the same time [16] and get- a directional order parameter. ting effectively trapped. Filaments in fimbrin bundles are The relative angle distribution is flat for simple dif- spaced closer than the binding radius of a myosin motor, fusion because there are no correlations between steps and the myosin minifilament can interact with many fila- of the random walk. A dictionary of relative angle dis- ments at the same time, supporting non-ergodic dynam- tributions for a variety of more complicated transport ics. However, the filament spacing in α-actinin bundles processes can be found in [20]. Consistent with previ- is larger than the myosin binding radius, which would ous observations for molecular motors [20], for the four suggest that the myosin minifilament spans these gaps experimental conditions that we consider here (Fig. 3), via its long axis of 100{500 nm, which is considerably we generally observe peaks at θ = 0 and θ = π, indicat- larger than the bundle spacing. The exponents for filamin ing an apparently inertial motion and frequent reversals, and fascin do not deviate as strongly from zero, indicat- respectively. ing that the dynamics are only weakly non-ergodic. In The relative angle distribution can be calculated at dif- the case of fascin, the unipolar bundles favor directional ferent ∆ to elucidate the timescales contributing to the transport, and do not support trapping, even if the mo- motion. To investigate the effects on the transport of tor protein interacts with multiple filaments at once. In local structure, as opposed to large-scale network topol- the case of filamin the large filament spacing often leads ogy, we chose a small ∆ (∆ = 0:1; 0:1; 1:53; 1:53 s for to the motor protein interacting with a single filament.