The maintenance of intracellular processes, like organelle transport and cell division, depend on bidirectional movement along microtubules. These processes typically require kinesin and dynein motor proteins, which move with opposite directionality. Because both types of motors are often simultaneously bound to the cargo, regulatory mechanisms are required to ensure controlled directional transport. Recently, it has been shown that parameters like mechanical motor activation, ATP concentration and roadblocks on the microtubule surface differentially influence the activity of kinesin and dynein motors in distinct manners. However, how these parameters affect bidirectional transport systems has not been studied. Here, we investigate the regulatory influence of these three parameters using in vitro gliding motility assays and stochastic simulations. We find that the number of active kinesin and dynein motors determines the transport direction and velocity, but that variations in ATP concentration and roadblock density have no significant effect. Thus, factors influencing the force balance between opposite motors appear to be important, whereas the detailed stepping kinetics and bypassing capabilities of the motors only have a small effect.
Intracellular transport is essential for cell division or organelle transport (Lodish et al., 2000; Verhey and Hammond, 2009; Soppina et al., 2009), and dysfunction leads to neurodegenerative diseases like Alzheimer's disease or amyotrophic lateral sclerosis (ALS) (De Vos et al., 2008; Goldstein, 2001; Hurd and Saxton, 1996; Chen et al., 2014; Karki and Holzbaur, 1999). In particular, microtubule-based transport is carried out by teams of the opposite-directed motor proteins kinesin and dynein. By actively moving cargo back and forth along microtubules, kinesin and dynein deliver cargo to where it is needed. Mitochondria, for instance, are transported to locations of low ATP concentration (Morris and Hollenbeck, 1993) and chromosomes are perfectly aligned on spindle microtubules during cell division (She and Yang, 2017; Goshima and Vale, 2003). Importantly, however, teams of kinesin and dynein motors are known to be often simultaneously bound to the cargo (Gennerich and Schild, 2006; Welte, 2004; Soppina et al., 2009; Hendricks et al., 2010). Without any regulatory mechanism the cargo might be transported in the kinesin or dynein direction, might randomly switch direction or might get stuck at a random position. However, regulatory mechanisms ensuring targeted transport remain poorly understood.
In the past, different regulatory mechanisms have been proposed. One mechanism suggests coordinating the motor activity (Gross, 2004). In this model, motors are assumed to be a priori in a passive state. By activating one motor team, targeted cargo transport occurs in the direction of the active team (Gross, 2004). One such activation mechanism involves adaptor proteins (McKenney et al., 2014; Schroeder and Vale, 2016; Blasius et al., 2007; Elshenawy et al., 2019). Another hypothesizes a mutual mechanical activation to trigger cargo transport (Monzon et al., 2019; Ally et al., 2009; De Rossi et al., 2017). Mechanical dynein activation has been shown to determine the velocity in unidirectional dynein-driven transport (Monzon et al., 2019). In that study, mechanical dynein activation was shown to strongly depend on the number of involved dynein motors (Monzon et al., 2019) suggesting that in bidirectional transport mechanical activation might also be linked to the number of motors. The influence of varying the number of motors has been studied previously. Rezaul et al. (2016), for instance, reversed a dynein-driven membrane organelle in vivo by adding a large number of kinesin motors. Moreover, Vale et al. (1992) showed that the transport direction in bidirectional gliding assays depends on the number of kinesin motors. However, for a complete understanding of the role of mechanical dynein activation in bidirectional transport, a systematic analysis is needed.
Another model proposes modifying the motor properties as a regulatory mechanism. Müller et al. (2008) showed that modified motor properties lead to different motility states. The motor velocity, for instance, is known to be modified by ATP concentration (Schnitzer et al., 2000; Ross et al., 2006; Torisawa et al., 2014; Nicholas et al., 2015a). If the ATP concentration asymmetrically modified the kinesin and dynein velocities, a bidirectionally moved cargo would likely change its net direction. Another motor property influenced by ATP concentration is the motor stall force (Mallik et al., 2004; Visscher et al., 1999). While the dynein stall force is known to increase linearly with ATP concentration (Mallik et al., 2004), the kinesin stall force is slightly reduced for very low ATP concentrations but invariant otherwise (Visscher et al., 1999). Thus, an increased ATP concentration might strengthen the dynein team and might reverse a cargo mainly transported in the kinesin direction. Consistent with this, a directional change as a function of ATP concentration is predicted by the theoretical work of Klein et al. (2014). However, whether the transport direction can be changed with a change in ATP concentration has not been tested experimentally.
Yet another regulatory mechanism is using hindering roadblocks to control bidirectional transport. Single motor proteins have been observed to have different reactions when encountering a roadblock (Telley et al., 2009; Dixit et al., 2008; Schneider et al., 2015). While a single kinesin detaches, for instance, when encountering the microtubule-associated protein tau, dynein continues stepping (Vershinin et al., 2007; Siahaan et al., 2019; Tan et al., 2019). Moreover, different mechanisms to bypass roadblocks have been observed in single-molecule experiments (Ferro et al., 2019). While kinesin has to detach and reattach behind the roadblock, dynein uses its ability to take sidesteps (Ferro et al., 2019). However, how a cargo that is bidirectionally transported by many motors reacts when encountering a roadblock, and whether roadblocks change the transport direction of such cargo remains unclear.
Moreover, parameters such as second messengers, viscosity, inhibitors or even size changes of the cargo could have an effect on bidirectional transport (Gennerich and Schild, 2006; Soppina et al., 2009; Hendricks et al., 2012; Coy et al., 1999; Firestone et al., 2012). However, here we focus on the relative contribution of number of motors, ATP concentration and roadblocks to regulate bidirectional transport.
To fully understand the relative contribution of different parameters, the parameters need to be varied systematically, and the bidirectional transport needs to be measured without affecting the activity of the motors themselves. The use of microtubule gliding assays is one way to achieve this (see Fig. 1A for an illustration). To measure the collective behavior of motors without affecting the motors themselves, microtubules are labeled and tracked. The number of motors involved in the transport can be systematically changed by varying the motor density on the surface. Moreover, the ATP concentration in the surrounding medium can be systematically changed, and microtubules can be coated with roadblocks at different concentrations. Consequently, microtubule gliding assays are highly suitable for a systematic analysis of the effects of motor number, ATP concentration and hindering roadblocks on bidirectional transport. Additionally, to get a detailed picture of the role of factors such as mechanical activation, we performed simulations of mathematical kinesin and dynein models, which are based on all known single-molecule parameters.
Regulating transport direction by varying the motor number
Previous studies have shown that changing the number of kinesin motors involved in bidirectional transport alters the transport direction (Rezaul et al., 2016; Vale et al., 1992). To test whether our set-up (Fig. 1A) shows a similar behavior, we varied the surface density (0 to 100 μm−2) of Drosophila kinesin heavy chain (hereafter referred to as kinesin, see Materials and Methods for details) at a constant high density (64 μm−2) of recombinant human cytoplasmic dynein (referred to as dynein, see Materials and Methods for details). Because we only take into account microtubules with lengths in a certain range (see figure legends for values), the motor density directly translates into motor number. The particular dynein density was chosen because we previously observed saturated velocities values in unidirectional dynein gliding assays under similar conditions (Monzon et al., 2019). To determine the transport direction, microtubules were polarity-marked and gliding trajectories were measured. Example trajectories (Fig. 1B) show dynein-driven transport for the ‘only dynein’ case and very low kinesin densities of 0.1 μm−2. For higher kinesin densities of 1 μm−2 almost no net movement was observed. In this case, forces between the kinesin and dynein team are balanced and the transport is halted. For higher kinesin densities of 20 μm−2 and the ‘only kinesin’ case, kinesin-driven transport was observed. In summary, three motility states could be distinguished depending on the kinesin density: (1) the dynein-driven state, (2) the balanced state (almost no net movement) and (3) the kinesin-driven state. Thus, by varying the kinesin density the transport direction can be changed.
When seeking to understand the regulation by the kinesin density and the role of mechanical dynein activation, an insight into the state of the motors at the molecular level is required. Experimentally, it is difficult to determine the number of attached kinesin and dynein motors, and it is not possible to know whether individual motors are active or passive. We therefore used the mathematical models of single kinesin and dynein molecules developed by Monzon et al. (2019) and Klein et al. (2014). The models are based on all known kinesin and dynein single-molecule properties (see Table S1). In Monzon et al. (2019), the models are used for unidirectional kinesin and dynein gliding assay simulations. Here, we adjusted the models to the bidirectional set-up (see Fig. 2A for an illustration, and the Materials and Methods for detailed model descriptions).
To be able to rely on the information provided by the bidirectional gliding assay simulation, it has to be shown that the simulation reproduces the experimental observations. Therefore, velocity histograms from simulation and experiment were compared (see Fig. 3). Here again, the kinesin densities were varied at a constant high dynein density. For both simulation and experiment, kinesin-driven transport was observed for high and intermediate kinesin densities, characterized by a narrow peak at high positive velocities. At low kinesin density, a velocity distribution around zero is observed for simulation and experiment indicating the balanced state. For the ‘only dynein’ case, simulation and experiment showed a broad distribution of negative velocities (see Fig. 3). Thus, the simulation reproduces the experiment, and we consider our model a reliable tool to obtain a deeper insight into bidirectional transport at the molecular level.
We observed that kinesin and dynein teams balanced each other at high dynein density (64 μm−2) but low kinesin density (1 μm−2). This raises the question of why more dynein than kinesin motors are needed to balance each other. To understand the mutual interplay between kinesin and dynein motors, and especially the influence of the number of dynein motors, we next considered different dynein densities. Different dynein densities have been shown to strongly influence the transport velocity of unidirectional dynein transport (Monzon et al., 2019). However, the influence of the number of dynein motors on bidirectional transport has not been studied. Hence, we measured the median microtubule gliding velocities for different constant dynein densities as a function of the kinesin density in experiment (Fig. 4A) and simulation (Fig. 4B). For intermediate (13 μm−2) and high (64 μm−2 and 128 μm−2 for simulation only) dynein densities, a transition from the dynein-driven (negative velocities) to the kinesin-driven state (positive velocities) was observed with increasing kinesin density. Comparing the kinesin densities of the balanced states, we find, the lower the dynein density the more the balance was shifted to lower kinesin densities. Thus, the dynein density also regulates the directionality of bidirectional transport.
When varying the kinesin density at the lowest dynein density (3 μm−2), no dynein-driven state was observed but rather the balanced state extended to very low kinesin densities (see Fig. 4B). For the non-existence of a dynein-driven state, two explanations are possible: (1) either no dynein is attached to the microtubules or (2) attached dynein is passive and therefore not transporting the microtubule directionally. To understand which scenario is at play, the median number of attached dynein motors was determined as a function of the kinesin density using the simulation (Fig. 4D, upper panel). We see that one dynein motor is attached for the lowest dynein density and kinesin densities smaller than 0.5 μm−2. This rules out the first possibility, namely, that no dynein motor is attached at all. To test whether the second scenario holds, we need to know the internal state of the attached dynein. Unlike kinesin, which is always active, dynein is known to have an active and a passive state (Zhang et al., 2017; Torisawa et al., 2014; Monzon et al., 2019). In our previously published model (Monzon et al., 2019), dynein is mechanically activated, in agreement with previous studies (Zhang et al., 2017; Torisawa et al., 2014; Belyy et al., 2016). Mechanically activated dynein (from now on called active dynein) performs directional motion on the microtubule, and passive dynein diffuses in the harmonic potential of the motors modeled as linear springs (see also Materials and Methods). In our model, a passive dynein mechanically activates when it is stretched and a mechanically activated dynein deactivates when it is unstretched again. In the gliding assay set-up, motors are tightly bound to the surface. When attaching to the microtubule, they are mechanically coupled via the microtubule. Owing to this coupling, a passive dynein motor is stretched and activated when the microtubule is transported by other attached motors. Consequently, when only one (a priori passive) dynein is attached, like in the ‘only dynein’ case at the lowest dynein density, it will never activate or walk directionally towards the microtubule minus-end. Thus, no net movement occurs. This is in agreement with previous studies stating that single mammalian dynein does not perform processive and directional motion towards the microtubule minus-end (Nicholas et al., 2015a; Brenner et al., 2020; McKenney et al., 2014). However, as soon as a kinesin motor is attached, the kinesin motor transports the microtubule and thereby stretches and activates the passive attached dynein motor. When determining the median number of active attached dynein (Fig. 4D, lower panel) and kinesin motors (Fig. 4C) using the simulation, it turns out that, for the lowest dynein density, the median number of active dynein motors is always zero. Thus, the activated state has a very short life time before being pulled off by kinesin. This means one dynein cannot resist one kinesin and, therefore, no balanced state exists.
To understand how many dyneins are needed to resist against kinesin, we looked at a slightly higher dynein density. At the intermediate dynein density (13 μm−2), we observed a clearly separated balanced state (Fig. 4A,B) indicating that dynein is able to resist kinesin. Looking again at the number and internal states of the attached motors (Fig. 4C,D), we find two attached dynein motors in the dynein-driven state. Although the median number of active attached dynein motors (Fig. 4D, lower panel) is zero, the median velocity in the dynein-driven state is greater than zero. Thus, temporarily, a passive attached dynein motor activated (the mean number of active attached dynein is 0.43) and drove the microtubule. In the balanced state, where additionally one kinesin is attached, one of the two dynein motors is active. This means that the kinesin activated one dynein motor, which then holds against the kinesin. To understand how one active dynein can hold against one kinesin, we need to have a detailed look at the motor forces. Previously, a motor was defined as ‘strong’ when having a large ‘stall force to detachment force ratio’ and as ‘weak’ when this ratio is small (Müller et al., 2008). For kinesin, the force ratio is one because the stall force is similar to the detachment force (see Fig. 2D,E; Table S1). For dynein, we have to distinguish between forward and backward loads because dynein detaches significantly faster under forward loads (smaller detachment force) than under backward loads (higher detachment force) (Fig. 2D,E; Table S1) (Gennerich et al., 2006; Cleary et al., 2014; Nicholas et al., 2015b; Rao et al., 2019). When resisting against kinesin (backward load), the force ratio is 0.3125. Therefore, dynein is the weaker motor in the competition with kinesin. Consequently, an active dynein is likely to be pulled off by a kinesin, coinciding with what we have seen for the lowest dynein density where one dynein competed against one kinesin. However, for the studied intermediate dynein density, there is still the passive attached dynein, when the active dynein is pulled off. The passive dynein is then activated. A balanced state can then be reached when in the meantime, a new passive dynein attaches to the microtubule, ‘helping out’ when the active dynein is again pulled off by kinesin. We conclude the role of the passive dynein is helping out when active dynein is pulled off by kinesin. Consequently, an active dynein having a passive dynein as a substitute is able to temporarily hold against one kinesin.
Our findings show that two dynein can temporarily hold against one kinesin. However, this balance is not stable, because the active dynein is continuously pulled off. Once a new passive dynein does not attach fast enough to help out, kinesin will take over. Obviously, a higher number of dynein motors are expected to stabilize the balanced state. Indeed, at high dynein densities of 64 μm−2 (same for 128 μm−2 in simulation) a stable balanced state with only small fluctuations was observed (Fig. 4A,B). Here, two kinesin motors competed against six active dynein motors in the balanced state (Fig. 4C,D). Looking at the number of active dynein motors, we find it reaches its maximum in the balanced state. This means kinesin activated more dynein and thereby stabilized the balanced state. Moreover, the total number of attached dynein (active and passive) also reached its maximum in the balanced state. This means that having kinesin pulling back dynein, reduces the overall dynein detachment. This could be interpreted as an effect originating from a dynein catch-bond. However, rather than a catch-bond, a slowly increasing detachment rate (slip-ideal bonding) has been observed previously (Cleary et al., 2014; Rao et al., 2019) and is implemented in our model for single dynein. Therefore, the reason for the reduced overall detachment has to be a collective effect. Imagine having kinesin transporting the microtubule, then, the attached dynein motors will be aligned under backward load. Under backward load, the dynein detachment increases more slowly with the force than under forward load (see Fig. 2E, lower panel) (Nicholas et al., 2015b; Cleary et al., 2014; Rao et al., 2019). Consequently, being aligned under backward load, fewer dynein motors detach compared to the number of dynein motors that detach in the ‘only dynein’ case, where the deflections of attached dynein motors are in random directions (forward and backward), as the individual motors step stochastically with different velocities (see Fig. 2B,C for an illustration). In conclusion, kinesin stabilizes the balanced state first by (1) activating passive dynein and (2) reducing the dynein detachment when aligning dynein motors under backward load.
In summary, the directionality of bidirectional transport can be regulated by varying the number of kinesin or the number of dynein motors.
Stable balanced state upon different ATP concentrations
Single-molecule velocities of dynein and kinesin strongly depend on ATP concentration (Schnitzer et al., 2000; Ross et al., 2006; Torisawa et al., 2014). To see whether the ATP concentration could have an effect on bidirectional transport performed by many motors, we first studied the influence of ATP on individual teams of kinesin and dynein motors (intermediate motor densities of 18 μm−2) in unidirectional gliding assays (Fig. 5A,B). We observed that the median velocities of kinesin- and dynein-driven transport increased with increasing ATP concentration for simulation and experiment. While the median velocities of the kinesin assay could be fitted by a Michaelis–Menten equation, velocities of the dynein assay did not show such a behavior. In the dynein assay the velocity increased more in a linear manner and did not saturate at the highest applied ATP concentrations. This means, kinesin and dynein, indeed, react differently to ATP concentration, and changing the ATP concentration might regulate the directionality of bidirectional transport.
To investigate the potential regulation of bidirectional transport by ATP concentration, we varied the kinesin density at a constant dynein density of 18 μm−2 (the same as for the unidirectional assay) for ATP concentrations of 1000 μM and 5000 μM in experiment (Fig. 5C) and simulation (Fig. 5D). When increasing the ATP concentration from 1000 μM ATP to 5000 μM, we found that in the unidirectional kinesin assay (Fig. 5A), the gliding velocity remained almost constant, while in the unidirectional dynein assay (Fig. 5B), the velocity nearly doubled. We reasoned that this behavior may give rise to a shift of the balanced state. For both ATP concentrations, we observed the three formerly described motility states in the bidirectional assay. However, the balanced state was invariant to ATP concentration, indicating that the ATP concentration does not change the transport direction.
To understand why the balanced state does not shift with ATP concentration, we again used the simulation for an insight into the molecular level. We found that, in the balanced state, mainly diffusive stepping of passive dynein occurs (Fig. S1). Although diffusive dynein stepping also depends on ATP concentration, passive dynein does not exert directed forces to the microtubule. Therefore, changing the diffusive stepping of passive dynein does not shift the force balance. Consequently, the balanced state is stable upon changing ATP concentration as long as the applied forces do not depend on the ATP concentration (see Fig. S2 for more ATP concentrations). In conclusion, ATP concentration cannot regulate the directionality of bidirectional transport.
Stable balanced state in the presence of roadblocks
Previous work has suggested that roadblocks might regulate bidirectional transport by asymmetrically inhibiting the motor activity (Siahaan et al., 2019; Henrichs et al., 2020; Tan et al., 2019; Monroy et al., 2018). Single kinesin and dynein motors indeed show different behavior upon encountering a roadblock (Telley et al., 2009; Dixit et al., 2008; Schneider et al., 2015). A single kinesin is known to either pause when encountering a roadblock or completely detach from the microtubule (Dixit et al., 2008; Schneider et al., 2015). A single dynein frequently side steps (Wang et al., 1995) and circumvents roadblocks without detaching. Consequently, single dynein motors overcome roadblocks more successfully than single kinesin motors (Ferro et al., 2019). It is therefore of interest, to study whether roadblocks alter the transport behavior in bidirectional gliding assays.
To know whether roadblocks could have an effect on bidirectional transport by multiple motors, we first studied how multiple dynein and kinesin motors react to roadblocks using unidirectional gliding assays. We coated microtubules with rKin430-T39N, a rigor mutant of rat kinesin-1 (hereafter referred to as roadblocks) (Schneider et al., 2015) at different concentrations and measured the median gliding velocities at constant motor densities of 50 μm−2 (see Fig. 6A for experiment and Fig. 6B for simulation). In both kinesin and dynein assays, the median velocity decreased with increasing roadblock concentration for experiment and simulation. Comparing kinesin and dynein gliding assays, we find dyneins are more affected than kinesins. This is in contrast to previous studies stating that, during cargo transport, multiple kinesins are as affected as multiple dyneins (Ferro et al., 2019). Since dynein likely uses side steps to circumvent roadblocks (Reck-Peterson et al., 2006), we hypothesize that dynein might not be able to take side steps in the gliding assay set-up and is therefore more affected. To test this hypothesis, we used the simulation to implement multiple protofilaments and allowed dynein to take side steps. Using this simulation, we observed that dynein was as affected by roadblocks as kinesin (Fig. S3a) in agreement with findings of Ferro et al. (2019). However, when not allowing dynein to take side steps in the multiple protofilament simulation, dynein is again more affected than kinesin. Consequently, if dynein does not take side-steps, both motors have to detach and reattach to overcome roadblocks. Because the kinesin attachment rate is much higher than the dynein attachment rate (see Table S1), multiple dyneins are affected more strongly by roadblocks in unidirectional gliding assays than multiple kinesins.
We observed that multiple dynein and kinesin motors react differently to roadblocks in unidirectional gliding assays. This raises the question of whether the different reaction to roadblocks influences bidirectional transport in gliding assays. Therefore, we applied the same constant dynein density as for the unidirectional assay with roadblocks and varied the kinesin densities at different roadblock concentrations. In the simulation, for all roadblock densities, clear motility states were observable (Fig. 6D). In the experiment, the dynein-driven state ‘merged’ with the balanced state at very high roadblock concentrations (Fig. 6C). However, for simulation and experiment, the balanced state stayed at the same kinesin density for all roadblock concentrations. Looking at the simulation results of the bidirectional gliding assays with multiple protofilaments, we see that even though dynein takes side steps, the balanced state stayed at the same kinesin density for all roadblock concentrations (Fig. S3b). As a consequence, independently of the side stepping ability of dynein, the balanced state remains at the same kinesin density for all roadblock concentrations.
To obtain an explanation for the balanced state being stable in the presence of roadblocks, we used again the simulation for insights at the molecular level. As stated before, kinesin and active dynein almost did not move at all in the balanced state. In detail, the median moved distance is 65 nm for kinesin and 16 nm for dynein (see also Fig. S4), which is small compared to the mean distance between roadblocks (∼166.67 nm at a density of 6 μm−1). This strong motor localization reduces their interactions with the roadblocks and leads to a stable balanced state in the presence of roadblocks. In conclusion, roadblocks do not regulate the directionality of bidirectional transport.
In this work, we showed that the direction of bidirectional transport is robust (i.e. is not substantially altered) against variation of ATP and roadblock concentration in microtubule gliding assays. We have seen that while ATP and roadblock concentrations influence the dynein- and kinesin-driven states, the balanced state remains stable. In the balanced state, active dynein and kinesin almost do not step at all. This suggests that parameters influencing the stepping of the motors cannot be used to change the balanced state and therefore cannot regulate the directionality of bidirectional transport. Consequently, the inability of dynein to take side steps in our bidirectional gliding assays does likely not change our result. However, when the number of involved motors changes, the balanced state can be shifted towards the kinesin- or dynein-driven state. Unlike ATP and roadblock concentration, the number of motors influences the force balance between kinesin and dynein motors. Having more kinesin motors means the kinesin team is stronger and vice versa. Therefore, factors that influence the force balance between the kinesin and dynein team would be expected to play a key role in the regulation of bidirectional transport.
Our results show that the kinesin activity influences the force balance. Using a mathematical model, we find that the kinesin activity strengthens dynein. The kinesin activity, on one hand, activates passive attached dynein motors and on the other hand increases the number of total attached dynein motors. As a first consequence, the number of attached dynein motors in the balanced state is higher than the number of kinesin motors. This is in agreement with previous studies reporting that more dynein motors than kinesin motors are involved in the tug-of-war (Soppina et al., 2009; Hendricks et al., 2010). A second consequence is that kinesin stabilizes the balanced state by strengthening the weaker and partly passive dynein motors. This explains why we find a stable force balance between kinesin and dynein. Thus, in addition to antagonistic effects between kinesin and dynein teams, there are also cooperative effects. The cooperative effect of kinesin stabilizing the balanced state might be involved in holding a cargo at a specific location inside the cell.
We find that from the three tested regulation factors, number of motors, ATP concentration and roadblocks, that only the number of motors can regulate bidirectional transport. However, the latter might not be the only way that the stable force balance could be regulated. Adding adaptor proteins that activate passive attached dynein could likely also change the force balance. Belyy et al. (2016) showed that a single dynein with adaptor proteins (a dynein–dynactin–BicD2 complex) can hold against a single kinesin, while a single dynein without adaptor proteins cannot. The latter is in agreement with our work showing that a single dynein without adaptor proteins cannot hold against a single kinesin. Moreover, measuring the dynein stall force, Belyy et al. (2016) found that a dynein plus adaptor proteins has a stall force of 4.4 pN, which has to be compared to the stall force of a single dynein without adaptor proteins (1−2 pN) (Mallik et al., 2004; Kunwar et al., 2011; McKenney et al., 2010; Ori-McKenney et al., 2010; Nicholas et al., 2015a; Brenner et al., 2020; Belyy et al., 2016), which we assume for our mechanically activated dynein. Thus, dynein activated by adaptor proteins is stronger than our mechanically activated dynein motors. Based on our work showing that factors influencing the force balance are key regulators of bidirectional transport, we would predict that adaptor proteins would be able to regulate bidirectional transport by increasing the dynein stall force. This suggests that even transport in the kinesin direction could be reversed by adding adaptor proteins.
Beside adaptor proteins changing the dynein stall force and therefore affecting the force balance, it has been shown that ATP concentration can also change the dynein stall force. Mallik et al. (2004) showed that for ATP concentrations lower than 1000 μM, the dynein stall force increases linearly with ATP concentration and remained constant above 1000 μM. In our work, we showed that for ATP concentrations equal to or greater than 1000 μM, the ATP concentration does not influence the force balance. However, the force balance could be influenced by lower ATP concentrations. When simulating the bidirectional gliding assay at lower ATP concentration, we did in fact see a tendency of the balanced state to be moved towards lower kinesin densities (Fig. S2b). But the shift was as small, as shifts arose from fluctuations in the number of motors involved in the transport. However, in Klein et al. (2014), the transport direction could be changed upon changes in ATP concentration in cargo transport simulations. Unlike in microtubule gliding assays, in cargo transport simulations, the number of involved kinesin and dynein motors is constant and dynein is always active. That is why fluctuations are smaller, and the effect of the ATP concentration altering the force balance is visible. Thus, we predict that, in microtubule gliding assays, shifts of the force balance stemming from altered ATP concentrations are small compared to shifts stemming from fluctuating numbers of motors.
Our study shows that the presence of roadblocks hindering both motors does not change the force balance. However, having roadblocks that act exclusively just on one kind of motor could have an effect. Tau islands, for instance, have an asymmetric effect on motor proteins. While kinesin detaches when encountering a tau island, dynein continues walking (Siahaan et al., 2019; Henrichs et al., 2020; Tan et al., 2019). This means tau islands change the motor number and therefore the force balance. Having a kinesin-driven transport that encounters a tau island, for instance, would cause a shift to a dynein-driven state because kinesin motors would detach and dynein could take over. Therefore, roadblocks that asymmetrically detach one kind of motor can be understood to alter the number of attached motors and therefore shift the force balance. As a consequence, such roadblocks could be used to regulate bidirectional transport.
Another mechanism that could alter the force balance by varying the number of motors, are temporarily changing binding affinities of the motors. Kinesin is known to alter the microtubule structure in a way that following kinesin motors show a higher binding affinity (Peet et al., 2018; Shima et al., 2018). Using a floor field model, Jose and Santen (2020) show that this temporarily changed binding affinity leads to the formation of kinesin and dynein lanes during bidirectional axonal transport. This suggests that altered binding affinities of the motors shift the force balance and regulate bidirectional transport.
Our finding that dynein gliding assays do not show a Michaelis–Menten-like dependence on ATP concentration is in contrast to the results from previous studies of Torisawa et al. (2014) and Nicholas et al. (2015a), which both found such a dependence. To understand why we do not see such a dependence, we simulated the unidirectional gliding assay at a higher dynein density and various ATP concentrations. At a higher dynein density (Fig. S2a), the simulation indeed shows a Michaelis–Menten-like dependence on ATP concentration. We know from our previous work (Monzon et al., 2019) that, at low dynein densities, passive dynein motors slow down microtubule gliding. However, having a lower ATP concentration, the activity of passive motors is lower. This means, the lower the ATP concentration, the more the passive motors slow down the microtubule gliding. This explains why the increase in gliding velocity with the ATP concentration at lower dynein density is slower than predicted by a Michaelis–Menten equation. Consequently, we do not see a Michaelis–Menten-like dependence on ATP concentration at low and intermediate dynein densities. At high dynein density, the slowing down by passive dynein motors is negligible and we see a Michaelis–Menten dependence on ATP concentration. This implies the ATP dependence of dynein-driven transport depends on the number of dynein motors.
In our study, we have provided key insights into the factors influencing bidirectional transport in the absence of adaptor proteins. Here, the factors that influence the force balance seem to be most important, whereas other factors, such as stepping or bypassing mechanisms, have little effect. Future work should now focus on the influence of adaptor proteins in light of this or roadblocks that differentially influence the number of attached motors.
MATERIALS AND METHODS
Materials and reagents
All reagents unless otherwise stated were purchased from Sigma. cDNA encoding full-length D. melanogaster Kinesin Heavy Chain (KHC) was sub-cloned into an in-house-generated insect cell expression vector in frame with a C-terminal 6×His tag. Codon optimized gene sequence of R. norvegicus Kif5c (GenArt, gene synthesis, Invitrogen) truncated to the first 430 amino acids with a T39N rigor mutation was cloned upstream of a 6×His affinity tag (with and without eGFP). Plasmids encoding H. sapiens dynein subunits (DYNC1H1, DYNC1I2, DYNC1LI2, DYNLT1, DYNLL1 and DYNLRB1) were a kind gift from Max Schlager and Andrew Carter (MRC Laboratory of Molecular Biology, UK).
Tubulin was purified from pig brains obtained from a local slaughter house by two cycles of polymerization and depolymerization in high concentration PIPES buffer (Castoldi and Popov, 2003), labeled with Alexa Fluor 488 dye, diluted to 4 mg/ml, aliquoted and stored at −80°C.
Polarity-marked microtubules were prepared by preferentially adding a short brightly labeled seed to plus-ends of GMPCPP and taxol-polymerized microtubules (referred to as ‘double-stabilized’ microtubules) in the presence of N-ethylmaleimide (NEM)-labeled tubulin. This was carried out as follows. First, 4.8 μM tubulin was freshly labeled with 1 mM NEM in the presence of 0.5 mM GTP for 20 min on ice. Then, excess NEM reagent was quenched with 8 mMβ-mercaptoethanol. An elongation mix, consisting of 1 mM GTP, 4 mM MgCl2, 1 mM Alexa-Fluor-488-labeled tubulin and 0.64 μM NEM-tubulin in BRB80 buffer was incubated on ice for 5 min, shifted to 37°C for 40 s followed by the addition of 250 μM double stabilized microtubules and incubated at 37°C for 1 h. Resultant microtubules were stabilized with 20 μM taxol. Polarity-marked microtubules were always prepared fresh every day at the beginning of an experiment.
Expression and purification of D. melanogaster KHC
KHC was purified from SF9 cells as described in Korten et al. (2016). Briefly frozen cell pellets infected with FlexiBAC baculoviral particles for 72 h were lysed via ultra-centrifugation at 4°C. Lysates were pre-clarified over a cation exchange column and the resultant elute was passed over an His-tagged IMAC column. Column-bound KHC dimers were eluted with 300 mM imidazole, desalted and snap frozen.
Expression and purification of H. sapiens dynein
Recombinant cytoplasmic dynein expression and purification was performed based on a published protocol (Schlager et al., 2014). The MultiBac plasmid was a generous gift by Max Schlager and Andrew Carter.
The MultiBac plasmid was integrated into the baculoviral genome of DH10EMBacY cells by using Tn7 transposition. 50 ng of MultiBac plasmid was incubated with 100 μl of chemical competent DH10EMBacY cells for 45 min on ice. The mix was heat shocked for 45 s at 42°C and incubated with 600 μl LB medium for 6 h at 37°C. 150 μl cell suspension was streaked out on agar plates containing 50 μg ml−1 Kanamycin, 10 μg ml−1 Gentamycin, 10 μg ml−1 Tetracycline, 20 μg ml−1 X-gal and 1 μg ml−1 IPTG and incubated at 37°C for 2–3 days. Plasmids from white colonies were isolated and checked for presence of all subunits via PCR. SF9 insect cells were infected with bacmids at a 1:1000 virus-to-cell ratio and grown for 4 days. Cells were harvested at 300 g for 15 min at 4°C and pellets were snap frozen in liquid nitrogen at stored at −80°C.
For purification of recombinant dynein, a frozen pellet corresponding to 250 ml insect cell culture was thawed on ice and resuspended in lysis buffer [50 mM HEPES pH 7.4, 100 mM NaCl, 1 mM DTT, 0.1 mM ATP, 10% (v/v) glycerol] supplemented with 1× protease inhibitor cocktail (complete EDTA free, Roche) to a final volume of 25 ml. Cells were lysed in a Dounce homogenizer with 20 strokes. The lysate was clarified at 504,000 g for 45 min at 4°C and added to 3 ml equilibrated (in lysis buffer) IgG–Sepharose beads in a column and incubated on a rotary mixer for 6 h. Protein-bound beads were washed with 50 ml lysis buffer and 50 ml TEV buffer [50 mM Tris-HCl pH 7.4, 148 mM potassium acetate, 2 mM magnesium acetate, 1 mM EGTA, 10% (v/v) glycerol, 0.1 mM ATP, 1 mM DTT and 0.1% Tween 20]. The ZZ affinity tag was cleaved off with 25 μg ml−1 TEV protease in TEV buffer at 4°C on a rotatory mixer overnight. After TEV cleavage, beads were removed and recombinant dynein was concentrated in a 100 kDa cut-off filter (Amicon Ultracel, Merck-Millipore) to a final volume of 500 μl. The TEV protease was removed by size-exclusion chromatography using a TSKgel G4000SWXL column equilibrated in GF150 buffer (25 mM HEPES pH 7.4, 150 mM KCl, 1 mM MgCl2, 5 mM DTT, 0.1 mM ATP and 0.1% Tween 20). Peak fractions were collected, pooled and concentrated to a maximum concentration of 1 mg/ml with a 100 kDa cut-off filter. All purification steps were performed at 4°C. Recombinant dynein was frozen in liquid nitrogen in the presence of ∼10% (v/v) glycerol. Protein concentration was determined with Bradford reagent.
Expression and purification of the rigor mutant rKin430-T39N
A preculture of rKin430-T39N-transformed E. coli BL21 pRARE cells was grown overnight in LB medium at 37°C. 750 ml fresh LB medium, supplemented with 50 μg/ml Kanamycin, was inoculated with 5 ml preculture and incubated in a shaker at 180 rpm, 37°C till optical density reached 0.6. Protein expression was induced with 0.5 mM IPTG and incubated at 18°C overnight. Cells were harvested at 7500 g for 10 min at 4°C. Cell pellet was resuspended in PBS with 10% glycerol (volume equivalent to weight of cell pellet in grams) and either snap frozen in liquid nitrogen or used immediately.
Cells were resuspended in lysis buffer (50 mM sodium phosphate buffer, pH 7.4, 300 mM KCl, 5% glycerol, 1 mM MgCl2, 10 mM β-mercaptoethanol and 0.1 mM ATP) supplemented with 30 mM Imidazole and protease inhibitor cocktail and lysed by 4–5 passages in an Emusiflux French press. Lysate was clarified at 186,000 g for 1 h at 4°C and passed through a 0.45 μm membrane filter to further get rid of particulate matter. A His-trap column was equilibrated with 10 column volumes (CV) of lysis buffer, followed by lysate application at a flow rate of 1 ml min−1. The column was washed with 10 CV of lysis buffer supplemented with 60 mM imidazole and protein eluted in lysis buffer supplemented with 300 mM imidazole. Protein was desalted, aliquoted, snap frozen and stored at − 80°C.
Glass coverslips (22×22 mm and 18×18 mm) were cleaned as follows: sonicated in 1:20 Mucasol for 15 min, rinsed in distilled water for 2 min, sonicated in 100% ethanol for 10 min, rinsed in distilled water for 2 min, rinsed in double distilled water for 2 min and blow-dried with nitrogen. Flow channels were prepared by placing 1.5 mm parafilm strips on a cleaned 22×22 mm coverslip (∼3 mm apart; four strips to prepare three channels) and covered with a 18×18 mm coverslip. This flow cell was placed on a heat block maintained at 55°C to melt the parafilm thereby making channels water tight.
Aliquots of dynein complexes were subjected to microtubule sedimentation to get rid of dead/rigor-binding motors before being used in gliding assays. Briefly, dynein complexes were incubated with unlabeled microtubules in dilution buffer (10 mM PIPES, pH 7.0, 50 mM potassium acetate, 4 mM MgSO4, 1 mM EGTA, 0.1% Tween20, 10 μM taxol, 2 mM Mg-ATP and 10 mM dithiothreitol) for 10 min followed by centrifugation at 120,000 g for 10 min at room temperature. Active dynein complexes retained in the supernatant were kept on ice until further use.
Flow channels were perfused with 2.5 mg ml−1 Protein A in double distilled water and incubated for 5 min followed by washing with dilution buffer. Solutions with various concentrations (prepared in dilution buffer) of kinesin and dynein were incubated in these channels for 5 min and excess motors were washed out with dilution buffer. To block the rest of the surface, motor-coated channels were incubated with 500 μg/ml casein in dilution buffer for 5 min followed by washing with dilution buffer. Double stabilized microtubules prepared in motility buffer (dilution buffer supplemented with 40 mM glucose, 110 μg ml−1 glucose oxidase and 10 μg ml−1 catalase) were incubated for 1 min and excess unbound microtubules were washed out with motility buffer.
Gliding microtubules were imaged on an inverted fluorescence microscope (Axiovert 200M, Zeiss) with a 40×, NA 1.3 oil immersion objective maintained at 27°C. Samples were illuminated with a metal arc lamp (Lumen 200, Prior Scientific) with an excitation 534/30 and emission 653/40 filter set in the optical path. A total of 50–200 frames were acquired with an iXon Ultra EMCCD (Andor) camera with a 100 ms exposure time at a frame rate of 1 Hz. Images were acquired with MetaMorph (Universal Imaging). A minimum of three independent sets of gliding assays (per experimental condition) were performed to obtain statistically significant results.
Microtubules were tracked with Fluorescence Image Evaluation Software for Tracking and Analysis (FIESTA; Ruhnow et al., 2011) which automates Gaussian fitting of fluorescence signals to extract position coordinates. All tracks were manually curated to filter out tracks of microtubules with significant changes in length between successive frames as observed when two microtubules cross paths during gliding. Data analysis and visualization were carried out in MATLAB2014b (MathWorks, Natick, MA). Instantaneous velocity was defined as the ratio of distance traversed by a microtubule and difference in time between consecutive frames. Velocities were smoothed with a rolling frame average over a window of five frames. Mean, median and quantiles of velocity distributions were calculated with in-built MATLAB functions.
Microtubule filament length was an important parameter that influenced gliding velocities (especially with dynein). Unless otherwise indicated, only 10−15 μm long microtubules (filtered from length values acquired from FIESTA) were used for analysis to obtain a more homogenous velocity distribution.
Surface density estimation
The density of kinesin and dynein motors was estimated by taking into a consideration the protein concentration of motors perfused (7 μl) into a 2×18×3 mm2 flowcell; the highest dynein concentration of 55 μg ml−1 yielded a density of . Motility measurements performed on axonemal dynein-coated surfaces have shown that only 10% of the total motor population contributes to active motion (Kotani et al., 2007). Applying dynein concentrations of 55, 28, 11, 6, 3 and 1.5 μg ml−1 were thus estimated to yield surface densities of 128, 64, 26, 13, 6 and 3 μm−2, respectively (Monzon et al., 2019). Similar assumptions were also made for the estimated surface densities of kinesin.
The linear relationship between applied motor concentrations and surface density was confirmed by observing landing events of microtubules on motor-coated surfaces (Katira et al., 2007; Agarwal et al., 2012). The number of microtubules landing on the surface were counted every 10 s for 100 frames. The plot of microtubule number versus time was fit to the curve, where N(t) is number of landed microtubules, Ninit is number of microtubules non-specifically adsorbed to the surface, Nmax is maximum number of microtubules in the field of view, R is the landing rate and t is time elapsed. Landing rate R derived from fit, microtubule area A (=Length of microtubule × 25 nm) and maximal diffusion-limited landing rate Z (assumed to be equal to landing rate of microtubules on very high densly coated motor surfaces) were plugged into the equation σ=−[ln(1-R=/Z)]/A to yield surface density σ.
The surface density of kinesin measured by the landing rate method was in agreement with the density estimated by the first method (7 μl of 6 μg ml−1 of kinesin solution yielded a density estimate of 102 μm−2 versus 100 μm−2 by the landing rate method).
Simulation of kinesin and dynein models
Here, we first describe details of the kinesin and dynein model and then present a simulation run. The dynein model was previously published in Monzon et al. (2019) and the kinesin model in Monzon et al. (2019) and Klein et al. (2014). Here, some parameter values were slightly changed and the models were slightly expanded. All expansions compared to the models published in Monzon et al. (2019) are marked in the text. All parameter values are listed in Table S1 including references to the literature whenever possible.
The gliding assay is modeled as a one-dimensional system. This means only one protofilament of the microtubule is taken into account. The microtubule itself is described as a rigid one-dimensional object, which is situated at constant height above the one-dimensional surface (glass coverslip). Attached motors move the microtubule back and forth.
If detached kinesin and dynein motors are under the microtubule (within the attachment area), they bind to the microtubule with the constant rates ka,kin and ka,dyn, respectively.
Mechanical dynein activation
Gliding assay with seven protofilaments
To compare our results to cargo transport by multiple motors on multiple protofilaments (Ferro et al., 2019), we implemented a gliding assay with seven protofilaments in the presence of roadblocks. In this modification, not only one protofilament is considered, like above, but seven protofilaments. The roadblocks and motors are randomly distributed on the seven protofilaments. If a motor encounters a roadblock or another motor on the same protofilament, it cannot continue walking on this protofilament. However, when a motor encounters a roadblock or another motor protein that is on another protofilament, it is not influenced by this obstacle and continues walking normally. From previous single-motor experiments, it is known that dynein frequently takes side steps (Reck-Peterson et al., 2006), while kinesin stays on the same protofilament (Ray et al., 1993). Here, we implemented a side-stepping rate sside for dynein to change the protofilament. However, kinesin stays on the same protofilament as long as the kinesin motor is attached. Thus, to circumvent a roadblock, dynein can use its side-stepping ability, whereas kinesin has to detach and reattach after the roadblock. The results of the gliding assay with seven protofilaments is shown in Fig. S3.
In the following, one run of the simulation is described.
As a first step the one-dimensional surface is coated with motors. At the beginning the microtubule is situated at position XMT(t=0)=0 and no motor is attached to the microtubule.
Using Gillespie's first reaction sampling (Gillespie, 1977), the next event [attachment, detachment, stepping or (de)activation] is chosen. Then the chosen motor event is performed, meaning the chosen motor is updated. After the motor event, the microtubule is moved to its nearest equilibrium position using a bisection search algorithm.
Measurement and output data
After the relaxation time trelax, the microtubule position and instantaneous velocity are measured. Each second (like in the experiment) the time, the velocity and the position of the microtubule are measured and output as well as the number of attached motors (distinguishing between kinesin, active and passive dynein motors). To mimic the measurement uncertainty of the experiment, a white noise with standard derivation σpos is added to the actual simulation position.
The complete simulation is terminated after Nsamples program runs or if Nmes measurements were performed. A single program run is terminated either after a specific amount of measurements nmes or after the simulation time Tend, or when no motor is attached to the microtubule at a time point greater than zero. The reason for the last termination condition is that in the experiment a microtubule trajectory is also terminated once no motor is attached anymore. If no motor is attached anymore in the experiment, the microtubule diffuses vertically away from the surface and cannot be measured.
We thank Andrew Carter and Max Schlager for kindly providing the plasmids encoding genes for H. sapiens dynein subunits and the multiBac plasmid.
Conceptualization: G.A.M., L. Scharrel, L. Santen, S.D.; Formal analysis: G.A.M., L. Scharrel, L. Santen, S.D.; Investigation: G.A.M., L. Scharrel, A.D., V.H., L. Santen, S.D.; Resources: L. Scharrel, S.D.; Data curation: G.A.M., L. Scharrel, L. Santen, S.D.; Writing - original draft: G.A.M.; Writing - review & editing: A.D., L. Santen, S.D.; Supervision: L. Santen, S.D.; Project administration: L. Santen, S.D.; Funding acquisition: L. Santen, S.D.
This work was supported by Deutsche Forschungsgemeinschaft (SFB1027) and Technische Universität Dresden. We also acknowledge funding from Boehringer Ingelheim Fonds to A.D. (PhD stipend).
Peer review history
The peer review history is available online at https://jcs.biologists.org/lookup/doi/10.1242/jcs.249938.reviewer-comments.pdf
The authors declare no competing or financial interests.