Effects of external strain on the regulation of microtubule sliding induced by outer arm dynein of sea urchin sperm flagella

Axonemal dynein (hereafter called dynein) is a minus-end directed microtubule motor that powers the motility of eukaryotic flagella and cilia. It is established that the cyclical bending of a flagellum is the result of ATP-dependent inter-doublet sliding caused by dyneins (Satir, 1968; Summers and Gibbons, 1971; Shingyoji et al., 1977; Brokaw, 1989). For the flagellar oscillation, the amount and the velocity of sliding induced by dynein must be regulated in a coordinated manner, among the nine doublet microtubules along the flagellum. How are the dynein activities regulated and coordinated in flagella? Studies using intact sperm or demembranated flagellar axonemes have revealed that the sliding activities of dynein can be modulated by external strain or bending imposed to the axoneme when dynein is within the 9+2 structure of a flagellum (Gibbons et al., 1987; Shingyoji et al., 1991; Morita and Shingyoji, 2004; Hayashi and Shingyoji, 2008). This leads to an idea of the ‘feedback regulation’ of dynein in the axoneme, in which the inter-doublet sliding induced by dynein results in the bending of the axoneme, and the bending regulates the dynein activities in turn. The 9+2 structure is thought to be responsible for this regulation of dynein activities to some extent; the axonemal substructures such as the central-pair microtubules (CP) and the radial spokes (RS) are likely to be involved in the regulation of microtubule sliding and wave forms at physiological (1 mmol l⁻¹) concentration of ATP (Hosokawa and Miki-Noumura, 1987; Smith and Sale, 1992; Smith, 2002; Wargo and Smith, 2003; Nakano et al., 2003), and there is a hypothesis that bending of the axoneme may regulate dynein activity through mechanical signal imposed on RS (Porter and Sale, 2000; Oda et al., 2014).

Is the regulation of dynein activity dependent on the 9+2 structure, or are the dynein molecules themselves involved in the strain-dependent cyclical bending? Previous studies on single molecule analyses of dynein isolated from flagellar axonemes have revealed that dynein itself is capable of strain-dependent backward stepping (towards the plus-end of microtubules) (Shingyoji et al., 1998) and strain-dependent backward force generation (Shingyoji et al., 2015). These results suggest that a single dynein molecule can change modes of sliding activities (from the forward mode to the backward mode) in response to external strain. However, when dyneins on a doublet microtubule work in a flagellum, they usually function as a team of hundreds of molecules, and in general, behaviour of the ensemble of motor proteins is not necessarily a simple summation of the single motor proteins, a counterexample of which is the directionality of kinesin 5 motors, Cin8 and Kip1 (Roostalu et al., 2011; Fridman et al., 2013). Thus an in vitro experimental system to investigate the mechanical characteristics of ensemble of dynein molecules is required.

The simplest and most prevalent assay system to study the activity of ensemble of motor proteins in vitro is the so-called ‘in vitro motility assay’, in which motor proteins adsorbed on a glass surface translocate the complementary purified filaments (Vale et al., 1985; Kron and Spudich, 1986; Vale and Toyoshima, 1988). It has been broadly used to study the properties of dynein, such as sliding velocity, duty ratio, modes of interaction with microtubules, and the effects of inhibitors or nucleotides (Vale et al., 1989; Moss et al., 1992; Hamasaki et al., 1995; Sakakibara et al., 1999; Inoue and Shingyoji, 2007; Kotani et al., 2007). Investigating the effect of external strain on dynein in this assay system has been technically difficult, and only shift in sliding velocity has been reported as the effect of small external force applied by fluid flow imposed to a sliding microtubule interacted with Chlamydomonas inner arm dynein species c (Kikushima and Kamiya, 2009). Here, we developed a new assay system with introduction of micromanipulation techniques based on the in vitro motility assay, in which sliding microtubules can be freely bent with a glass microneedle, and the direction of the strain imposed on dynein and microtubules can be estimated from the shapes and motion of the microtubules. We found that dissociation and stoppage of sliding microtubules were induced by external strain. We also found that backward sliding, which has been observed in single molecule analyses (Shingyoji et al., 2015), can also be induced. Modification in sliding velocities was also observed. These results suggest that sliding activities of multiple dynein molecules can be modified and coordinated through external strain in a flexible manner, and that this regulatory mechanism might be involved in the flagellar oscillation.

21S dynein was prepared from sperm flagella of the sea urchins Pseudocentrotus depressus, Hemicentrotus pulcherrimus and Strongylocentrotus nudus, according to procedures described previously (Yoshimura and Shingyoji, 1999; Imai and Shingyoji, 2003; Inoue and Shingyoji, 2007) with some modifications. Sperm suspended in Ca²⁺-free artificial seawater (465 mmol l⁻¹ NaCl, 10 mmol l⁻¹ KCl, 25 mmol l⁻¹ MgSO₄, 25 mmol l⁻¹ MgCl₂ and 2 mmol l⁻¹ Tris-HCl; pH 8.0) was treated for about 1–2 min with demembranating solution [0.1% (w/v) Triton X-100, 150 mmol l⁻¹ potassium acetate, 2 mmol l⁻¹ MgSO₄, 10 mmol l⁻¹ Tris-HCl, 2 mmol l⁻¹ EGTA and 1 mmol l⁻¹ dithiothreitol; pH 8.0] at room temperature. The demembranated sperm were resuspended in 150 mmol l⁻¹ potassium acetate reactivating solution (150 mmol l⁻¹ potassium acetate, 2 mmol l⁻¹ MgSO₄, 10 mmol l⁻¹ Tris-HCl, 2 mmol l⁻¹ EGTA and 1 mmol l⁻¹ dithiothreitol; pH 8.0; without ATP) and were fragmented by passing through a 23-gauge hypodermic needle (Terumo, Tokyo, Japan) with a 1 ml syringe on ice. All the following procedures were performed on ice or at 4°C. After removal of the sperm heads by centrifugation at 2000 g, outer arm dynein was extracted from the axonemes by treatment with 0.6 mol l⁻¹ NaCl (reactivating solution containing 0.6 mol l⁻¹ NaCl instead of potassium acetate). The crude outer arm dynein was centrifuged on a 5–20% sucrose density gradient made in reactivating solution (containing 0.2 mol l⁻¹ NaCl instead of potassium acetate) at 188,000 g for 12 h. The protein concentrations of the fractions were determined by the Bradford method (Bradford, 1976), using bovine serum albumin as a standard. The fractions with the highest protein concentrations were adopted as purified outer arm dynein (21S dynein). 21S dynein in sucrose solution was kept in liquid nitrogen until use.

Tubulin was purified from porcine brains (Castoldi and Popov, 2003) by two cycles of polymerization–depolymerization in the presence of high molarity PIPES buffer (final concentration: 0.5 mol l⁻¹), suspended in MES buffer (80 mmol l⁻¹ MES, 1 mmol l⁻¹ EGTA, 1 mmol l⁻¹ MgSO₄ and 0.5 mmol l⁻¹ dithiothreitol), and stored in liquid nitrogen until use. Purified tubulin was labelled with tetramethylrhodamine (Hyman et al., 1991), using MES as a buffer and one cycle of polymerization–depolymerization instead of two. To obtain seeds, unlabelled tubulin and tetramethylrhodamine-labelled tubulin were mixed at a ratio of 5:1 and incubated at 33°C for 30 min in the presence of 5 mmol l⁻¹ GTP. The seeds were further polymerized by being mixed with unlabelled tubulin and tetramethylrhodamine-labelled tubulin at a ratio of 4:50:5, and incubated at 33°C for 20 min. Polymerized microtubules were stabilized by addition of 0.1 mmol l⁻¹ paclitaxel.

For induction of microtubule sliding, the method of in vitro motility assay of dynein (Paschal et al., 1987; Vale and Toyoshima, 1988) was modified based on the method of the micromanipulation assay described previously (Morita and Shingyoji, 2004). By being diluted with 50 mmol l⁻¹ potassium acetate reactivating solution (50 mmol l⁻¹ potassium acetate, 2 mmol l⁻¹ EGTA, 2 mmol l⁻¹ MgSO₄, 1 mmol l⁻¹ dithiothreitol and 10 mmol l⁻¹ Tris-HCl; pH 8.0), 21S dynein at a concentration of 100–350 µg ml⁻¹ was obtained. Then 21S dynein (10 µl) was introduced into a 1.4 µl-perfusion chamber constructed with two glass coverslips (24 mm×50 mm and 6 mm×18 mm) and 60 µm-thick mending tapes (Nichiban, Tokyo, Japan) used as a spacer. To induce microtubule sliding, microtubules (final concentration, 6 µg ml⁻¹) suspended in assay buffer containing ATP [20 mmol l⁻¹ glucose, 216 µg ml⁻¹ glucose oxidase, 36 µg ml⁻¹ catalase, 1% (v/v) 2-mercaptoethanol, 10 µmol l⁻¹ paclitaxel and 1 mmol l⁻¹ ATP, with or without ATP regeneration system in reactivating solution; 10 µl] were successively introduced into the chamber after the perfusion of 21S dynein. As the ATP regeneration system, 0.5 mg ml⁻¹ creatine kinase and 5 mmol l⁻¹ creatine phosphate were used.

To obtain an open surface over the sliding microtubules, which makes it possible to apply mechanical manipulation with a glass microneedle to the microtubules, 150 µl of assay buffer with 1 mmol l⁻¹ ATP was added on the smaller (upper) coverslip, and the smaller coverslip was carefully slid sideways with tweezers (Fig. 1). Microtubules sliding on the glass surface at the bottom of the 150 µl pool of assay buffer were observed at room temperature (21–28°C) under an inverted fluorescence microscope (IX-70; Olympus, Tokyo, Japan) with a ×100 oil-immersion objective lens (PlanApo; NA=1.4; Olympus), using a mercury arc lamp (USH-102D; Ushio, Tokyo, Japan) for excitation. The glass microneedles were made with a micropipette puller (PP-830; Narishige, Tokyo, Japan) and a glass rod (G-1000; Narishige). They were attached to and controlled by a water hydraulic micromanipulator (MW-3; Narishige), with the angle of approach to the coverslip around 20 deg. A glass microneedle was visualized simultaneously with the microtubules as a phase-contrast image using a halogen lamp for illumination. The movements of microtubules and glass microneedles were recorded with an image-intensified CCD camera (C2400-77; Hamamatsu Photonics, Hamamatsu, Shizuoka, Japan) and a hard disk drive recorder.

Movies were prepared by cropping the recorded data with NIH Image software (National Institutes of Health, Bethesda, MD, USA). Sequential images were prepared by capturing the movies with ImageJ (version 1.49h; National Institutes of Health). For calculation of the sliding velocities of microtubules, the sliding distances of the anterior or posterior end of the microtubules within a certain period of time (a few seconds) were measured either by ImageJ or by tracing the positions on to a transparent film fixed on a display. The time course of the position of microtubule ends were made with MTrackJ Plugin for ImageJ (Meijering et al., 2012). The tracings of the shapes of the microtubules were made using software for automatic tracking of a cilium along its length (Bohboh, BohbohSoft; Shiba et al., 2002).

In the present in vitro motility assay using open surface chambers, microtubule sliding was induced by dynein molecules adsorbed onto a glass surface of a perfusion chamber. In a separate experiment using polarity-marked microtubules (Shingyoji et al., 1998, 2015), we confirmed that the anterior end of a microtubule is always the plus end. In the present study, we tried to manipulate such microtubules moving on dyneins with a microneedle (Fig. 1). Like most of the studies using the in vitro motility assay, we used singlet microtubules polymerized from brain tubulin. There has been a report on using axonemal tubulin (95%)-derived microtubules for the in vitro motility assay of axonemal dynein, but it only shows higher velocity in short (<6 µm) conspecific microtubules than those polymerized from porcine brain tubulin (100%), and does not show any significant difference in longer microtubules (Alper et al., 2013; see the next section for the description of the lengths of the microtubules used in the present study). Micromanipulation was carried out according to our previously developed method (Morita and Shingyoji, 2004; Hayashi and Shingyoji, 2008) with some modifications. In our open-surface chamber, sliding velocity of the microtubules was 4.6±1.1 µm s⁻¹ (mean±s.d. at 27°C; N=15, 150 µg ml⁻¹ dynein from Pseudocentrotus depressus), which is comparable to those shown in the conventional in vitro motility assay (3.5±1.3 µm s⁻¹; Paschal et al., 1987; and around 6 µm s⁻¹; Moss et al., 1992). As reported for the conventional in vitro motility assay (Paschal et al., 1987; Vale and Toyoshima, 1988), microtubules in the present assay system also showed unidirectional sliding towards their longitudinal axis, occasional spontaneous stoppage and spontaneous recovery of sliding; of 367 sliding microtubules observed without manipulation in three separate experiments (by using 150 µg ml⁻¹ 21S dynein obtained from Pseudocentrotus depressus), 79 (22%) microtubules showed spontaneous stoppage (once or more) during crossing the observation field (87×115 µm). The average duration of the stoppage was 5.1 s (N=66) and its peak duration was 0–2 s (Fig. S1). Microtubule sliding was observed for more than 20 min after perfusion with ATP, both in the presence and absence of ATP regeneration system.

Mechanical manipulation was applied to a sliding microtubule with a glass microneedle so as to bend a part of the microtubule in one plane. In other words, microtubules were bent within the plane parallel to the glass surface, thereby interesting responses including stoppage and backward sliding were induced. In a few cases when lower concentration (100 µg ml⁻¹) of dynein was used, micromanipulation provided bending in three-dimensional effects and dissociation of microtubules occurred. Details of the dissociation will be described later (see Results, Three-dimensional effects caused dissociation). Unless otherwise noted, responses induced by imposed bending were not affected by conditions, including the presence or absence of ATP regeneration system, different kinds of sea urchins for dynein and dynein concentrations used for motile assay.

Of 574 trials of planar bending applied to 260 sliding microtubules (length: 25±14 µm, mean±s.d., ranging from 3 to 79 µm, N=574), 434 trials (76%) resulted in continuation of forward sliding, as shown in Fig. 2A. An apparent decrease in sliding velocity was observed in 37 trials of the 434 (see the next section). We found that in 136 (23%) of the 574 trials, microtubules stopped sliding with imposed planar bending and stayed apparently stationary on the glass surface just after the microneedle was removed from the microtubule (Fig. 2B). In the remaining four cases under planar bending, we observed backward sliding (see below). The lengths of the microtubules that showed each of those responses (decrease in velocity, 21±9 µm, N=37; stoppage, 26±14 µm, N=136; backward sliding, 36±9 µm, N=4; means±s.d.) were not significantly different from those (25±14 µm, N=434, mean±s.d.) of the microtubules with continuous forward sliding (Mann–Whitney U-test, P>0.05).

When planar bending was applied to a sliding microtubule with a glass microneedle, in addition to the shape change of the microtubule induced by the movement of the glass microneedle (0–0.8 s in Fig. 2A), changes in mechanical state between the dynein–microtubule interactions on the glass surface would be induced. Here, we consider three categories of strain imposed on dyneins to be important. One is the lateral strain, which shifts a microtubule laterally from the original position (‘a’ in the bottom panel of Fig. 2A). This strain is mainly imposed on a microtubule at the region of contact with the glass microneedle. Secondly, microtubules and dyneins experience pulling strain along the length of a microtubule towards the point of contact with the glass microneedle (‘b’ in the bottom panel of Fig. 2A). This strain results from the tension of the microtubule. Thirdly, elastic strain resulting from bends of a microtubule is imposed on dyneins (‘c’ in the bottom panel of Fig. 2A). This kind of strain works in a direction so as to decrease the curvature of the bend of the microtubule. The first one (lateral strain, a) and the second one (pulling strain, b) are active only during imposed bending of a microtubule with moving the glass microneedle, whereas the third one (elastic strain, c) is always active as long as the bend of the microtubule is present and independent of the glass microneedle and its movement.

Of 434 cases showing continuation of forward sliding by planar bending, 37 brought about apparent decrease in sliding velocity. One example is shown in Fig. 3A. We measured a change of the sliding path distance within 3.2 s of the stable phase before, during and after imposed bending and obtained the sliding velocities. The time-dependent change in the path distance (absolute value) of the anterior end of the same microtubule as shown in Fig. 3A is analysed in Fig. 3B. The slope of this plot, representing the sliding velocity, decreased during bending. In Fig. 3C the average sliding velocities in three states, before bending, during bending and after the bend dissolved, are plotted for nine cases. The average sliding velocities were low during bending while they were similarly high both before and after bending. In the state of ‘after the bend dissolved’ the microtubule bent with a microneedle became nearly straight after being released from imposed bending. The relationship between the percentage of decrease in the average sliding velocities, calculated from the rate before and during bending of each of the nine examples, and the bend angle were examined (Fig. 3D). The average sliding velocities were decreased by 28–93% in response to imposed bending of 44–221 deg. However, there was no statistical significance in fitting the linear regression model to the relationship between the percentage of decrease in the average sliding velocities (a dependent variable) and the bend angles (an independent variable) (F-test for linear regression analysis, P>0.4, N=9). These results suggest that imposed bending is able to decrease the sliding velocity, but the extent of decrease is not solely determined by the bend angles.

Stoppage of sliding microtubules induced by planar bending did not always recover sliding after removal of the microneedle. Of the 136 trials of planar bending that induced stoppage, about half (53%) recovered sliding spontaneously (i.e. without any further mechanical manipulations) (Fig. 4, filled box). In the remaining stoppage of 53 cases (Fig. 4, grey and hatched boxes), when planar bending was further applied, 34 of them recovered sliding (Fig. 4, grey box). Thus 78% (106 out of 136) of the microtubules that showed stoppage in response to the first imposed bending showed recovery of sliding (Fig. 4, filled and grey boxes). Some of the microtubules showing stoppage (19 out of 53 cases) were exposed to irreversible damage of severing a microtubule into two parts by the second manipulation (Fig. 4, hatched box). However, most of the 19 severed microtubules showed active behaviour: in 11 severed microtubules either half of the severed microtubules moved, and in six severed microtubules both parts of the severed microtubules moved.

Fig. 5A shows a typical example of the stoppage induced by imposed bending, accompanied by forward sliding due to spontaneous recovery. The microtubule was bent in the middle region with backward pulling of the anterior region, thereby stoppage occurred only in the anterior region (at 2.7–11.4 s), but after stoppage (for 8.7 s) the anterior region recovered sliding at 11.4 s and the whole length began to move (arrows). The distribution of duration of the stoppage for 34 cases is shown in Fig. 5B. The average duration was 9.6 s. These results indicate that stoppage induced by imposed bending was not the result of irreversible damage to dynein, but reflects a reversible state of stable dynein–microtubule crossbridges.

What kind of mechanical signals cause stoppage? From the observation of the microtubules that showed stoppage, we noticed that the anterior region (or the leading part) of the microtubule was often pulled backwards along with imposed bending, just before the stoppage of sliding. This suggests a possibility that backward-pulling strain caused by the movement of the microneedle (pulling strain, b in the bottom panel of Fig. 2A) may be important in inducing stoppage. To confirm this, we categorized the trials of planar bending according to the regions of bending and the presence/absence of backward displacement of the anterior region, and quantified the relative frequency of stoppage to the number of the trials of bending of each category. The relative frequency of stoppage to the number of trials of bending on the middle region of microtubules was significantly higher than that of bending on the posterior region (P<0.03, chi-squared test; Fig. 6A), and the relative frequency of stoppage to the number of trials of bending on the middle region with a backward displacement of the anterior region was significantly higher than that without one (P=1×10⁻⁵<0.01, chi-squared test; Fig. 6B). This result suggests that backward pulling strain may be important for the induction of stoppage of sliding.

To further test the importance of backward strain on induction of stoppage, short (9±4 µm, mean±s.d.; ranging from 5 to 14 µm, N=10) sliding microtubules were pushed backwards by causing a head-on collision with a glass microneedle without bending. This kind of mechanical manipulation was technically difficult, but we have succeeded in doing so in 10 trials on seven sliding microtubules. While seven trials induced continuation of forward sliding, three trials of backward pushing induced stoppage. One example of those three trials is shown in Fig. 6C. This suggests that the backward strain, not bending of microtubules itself, is important for induction of stoppage.

In 10 trials out of the 136 trials of bending that induced stoppage, only the anterior part of the microtubules became stationary while the posterior part (distal region) continued sliding at least for a period of time. In Fig. 6D imposed bending started at 0 s and the anterior region stopped sliding at 5.7 s (indicated in red in Fig. 6D), while the posterior region continued sliding (indicated in blue in Fig. 6D) between 5.7 and 14.4 s. We focused on the motion of the anterior part of this microtubule just before the stoppage of the anterior region, and found that the anterior part of the microtubule was pulled backwards (5.2–5.7 s) by the pulling strain caused by the glass microneedle (pulling strain, b in the bottom panel of Fig. 2A). Fig. 6E shows the superimposed tracings of the positions of the microtubule before and after backward pulling of the anterior region (at 5.2 and 5.7 s in Fig. 6D). The anterior end of the microtubule, indicated by the arrows, moved 1.0 µm backwards in 0.5 s. In all of the 10 cases where microtubules showed stoppage only in the anterior region (lengths of the microtubules: 42±15 µm, mean±s.d., N=10), the anterior part was pulled backwards by the glass microneedle just before the stoppage of the anterior region. The distances of the backward displacement of the anterior end were 1.5 µm on average, and ranged from 0.3 to 3.3 µm (N=10). Stoppage only in the anterior region of a microtubule was never observed in microtubules that were not mechanically manipulated (N=724 microtubules). This further suggests the importance of backward strain in induction of stoppage of sliding.

Dynein concentration positively affected the frequency of occurrence of stoppage, while the presence of an ATP regeneration system or the species of sea urchins used for preparation of dynein did not. In the experiments using 100 µg ml⁻¹ 21S dynein, microtubules often slid with their one end dissociated and swaying, while in the experiments using more than 150 µg ml⁻¹ 21S dynein almost all the microtubules slid with the whole length within the plane of focus. The results suggest that higher concentrations of perfused dynein, which probably lead to higher densities of dynein molecules on the glass surface, may result in higher affinity between dynein and microtubules, facilitating the conduction of strain from the manipulated microtubules to interacting dynein, causing the stoppage of sliding more frequently induced by mechanical manipulation.

When ATP in the perfusion chamber was almost consumed by ATPase activity of dynein, microtubules stopped sliding and became stationary on the glass surface, as reported previously for the in vitro motility assay without wash of excess dynein (Paschal et al., 1987). When excess amount of 21S dynein was removed with buffer solution before perfusion of ATP and microtubules, microtubules showed diffusive motion and dissociated from dynein after the exhaustion of ATP, which is consistent with the previous report for the conventional in vitro motility assay with wash of excess dynein (Moss et al., 1992). We confirmed that under the condition with wash of excess dynein, imposed bending can also induce stoppage of microtubule sliding. This suggests that stoppage can be induced by backward strain in an independent mechanism from the stationary state observed after ATP exhaustion and that the stoppage induced by mechanical manipulation is the result of reversible changes in modes of motility of active dynein molecules.

Four out of the 574 trials of planar bending brought about a reversal in the sliding direction, which was never observed without mechanical manipulation. Four cases are few, but seem to represent a very important characteristic of dynein.

In three of the cases, backward sliding was induced only in the anterior region of the microtubules by bending the posterior region of the microtubules. One example is shown in Fig. 7A and Movie 1. This microtubule was sliding towards the lower left in the panel in Fig. 7A before bending (from −5.0 to 0 s). However, the anterior region changed the sliding direction towards the upper right when bending was applied in the posterior region. After bending, the glass microneedle was kept in the same position (indicated by an arrowhead, at 0.7–13.7 s in Fig. 7A). Superimposed tracings of the microtubules in Fig. 7A are shown in Fig. 7B. The anterior end of this microtubule moved 5.0 µm towards the upper right, the opposite direction of original sliding, indicating the backward sliding along the anterior region of the microtubule. The position of the glass microneedle for bending was kept almost stable during the backward sliding, resulting in an increase in curvature of the bent region in Fig. 7A and B. The tip of the microneedle shifted a slight distance of about 0.6 µm towards the lower direction between 0.7 and 13.7 s, possibly due to pushing force of the backward sliding in the anterior region of the microtubule. The backward sliding for a few seconds was accompanied by dissociation and displacement of the anterior region of the microtubule at 18.8 s, which then is followed by recovery of the slow forward sliding at 34.5 s. Slow forward sliding recovered the original velocity after sliding microtubule released from bending constriction with the microneedle at 46.1 s. These results show that the backward sliding is not the result of passive movement caused by the movement of the glass microneedle, but is driven by sliding induced by dynein motors.

What kind of signal is important for inducing backward sliding? In all of the three cases where backward sliding was induced only in the anterior region of the microtubules, bend angles were more than 140 deg (1–3 in Table 1), and the glass microneedle was kept stationary at the place of imposed bending, which resulted in keeping the bent region of the microtubules during the backward sliding. In such a condition, elastic strain of the microtubule (c in the bottom panel in Fig. 2A) must be imposed on the anterior region of the microtubule continuously. Elastic strain works in the direction such that it decreases the curvature of the microtubule, so the anterior region of the microtubule and the interacting dynein are estimated to receive elastic strain in an angled backward direction. These suggest that continuous backward strain may be important for inducing backward sliding.

In the remaining case among the four, backward sliding was induced in the whole length of the microtubule (Fig. 7C, Movie 2). The microtubule sliding towards the lower right was bent in the posterior region (at −6.5 to 0 s) and the glass microneedle was kept attached to the microtubule (indicated with arrowheads at 1.9–74.9 s). Both the anterior end and the posterior end of the microtubule moved backward, in the opposite direction of the original sliding (1.9–74.9 s). Superimposed tracings of this microtubule show that the whole length of the microtubule translocated almost along a single track during the backward sliding (Fig. 7D). This microtubule recovered forward sliding after the backward sliding (74.9–87.9 s), and proceeded on the same track as the previous backward sliding (the dotted line in Fig. 7D). This result suggests that backward sliding can be induced as a result of reversible change in sliding direction of a single set of dynein molecules interacting with the microtubule.

In this case, the glass microneedle was slowly shifted towards the bottom of the image for 3.0 µm while the microtubule slid backward for 2.3 µm (1.9–74.9 s in Fig. 7C, number 4 in Table 1). Considering the steric geometry of the microtubule and the glass microneedle, where the tip of the glass microneedle was obliquely attached to the glass surface and the microtubule was located between the microneedle and the glass surface, the posterior region of the microtubule might have been pulled backwards by the motion of the microneedle. Together with the analyses of the three microtubules that showed backward sliding only in the anterior region, this result suggests that continuous backward strain may be important in inducing backward sliding.

Although there was no statistically significant difference (P>0.05, randomized test on matched samples), the velocities of the backward sliding shown during bending tended to be smaller than the forward sliding velocities before bending (Table 1). This does not appear characteristic of backward sliding because the velocities of the forward sliding following the backward sliding (exerted before the bend dissolved) also tended to be slower than that exerted after the bend dissolved. These suggest that imposed bending may induce decrease in sliding velocity in both forward and backward sliding, possibly through a common mechanism such as continuous elastic strain resulting from the bend of the microtubules.

Ten trials of imposed bending with a three-dimensional effect, in which microtubules were bent in a three-dimensional way, not within the plane parallel to the glass surface, induced dissociation of the microtubules. In two of these cases the dissociation occurred only in a short period; the glass microneedle was inserted between a microtubule and the glass surface, thus a part of the microtubule transiently dissociated from dynein on the glass surface, with the rest of the microtubule continuing to slide on the glass surface, then eventually the whole length came back to the plane of focus and recovered forward sliding. In the other eight cases, the whole length of the microtubule dissociated from dynein on the glass surface (complete dissociation). In one of these eight cases, the middle region of a microtubule was pushed down towards the glass surface by the glass microneedle and the microtubule was bent so that the anterior and posterior region dissociated from dynein on the glass surface, followed by dissociation of the whole length. In the other seven cases, the glass microneedle was inserted and moved between a microtubule and the glass surface, causing dissociation of the whole length of the microtubule (Fig. 8). The relative frequency of dissociation to the number of mechanical manipulations was significantly higher when the concentration of perfused dynein was 100 µg ml⁻¹ than when it was 150 or 180–200 µg ml⁻¹, probably because the microneedle was more easily inserted between the microtubule and the glass surface when the dynein density was lower, due to the lower affinity between dyneins and microtubules. These results suggest that adequate strain, which is perpendicular to the glass surface, can induce dissociation of microtubules from dynein.

In the present study, to understand the behaviour of an ensemble of dynein molecules under mechanical signals, the microtubule sliding induced by outer arm dynein was investigated. We found that planar bending can induce three different modes of motility of dynein: continuation of forward sliding, stoppage of sliding and backward sliding. Decrease in sliding velocity was also induced by planar bending. For induction of the stoppage and backward sliding, the importance of backward strain is suggested. Moreover, imposed bending with a three-dimensional effect induced dissociation of the microtubules from dyneins. While this is a novel in vitro experimental system to investigate the mechanical properties of dynein, we should also pay attention to the limitation of the present experimental set-up, in which dyneins interact with singlet microtubules and a glass surface, unlike the intact situation with the doublet microtubules.

Stoppage (stationary mode) induced in the present study was distinct from dissociation. The microtubules that showed stoppage remained within the plane of focus without diffusive motion, implicating the sustained interaction with dynein (Fig. 2B), whereas dissociated microtubules went out of focus and showed diffusive motion (Fig. 8). Based on the cyclical response consisting of bending, pause with a constant curvature and dissociation observed in a pair of doublet microtubules of a frayed flagellar axoneme, dyneins are thought to change their modes from sliding to stationary in response to external strain without detachment from the microtubule at a low concentration (10 µmol l⁻¹) of ATP (Lindemann, 2014; Mukudan et al., 2014). Stoppage of sliding in the present study suggests that this is also the case even at a high, physiological concentration (1 mmol l⁻¹) of ATP. Furthermore, in addition to transverse strain (t-force; Lindemann and Lesich, 2015), the present study indicates that backward, parallel strain may also be important in inducing stoppage of sliding.

It appears that there is a natural fluctuation between the forward sliding mode and the stationary mode of dynein molecules, because 53% (72 out of 136) of the microtubules that showed stoppage in response to mechanical manipulation recovered sliding spontaneously in several seconds (Figs 4 and 5A,B). Microtubules that were not manipulated with the glass microneedle also showed occasional transient stoppage and recovery of sliding (see Results, Microtubule sliding induced by dyneins in the open surface chamber). Transient stoppage and recovery of sliding of microtubules were also reported in the study of conventional in vitro motility assay of dynein (Vale and Toyoshima, 1988). If this occasional stoppage and recovery is the result of fluctuation of modes of motility of individual dynein molecules, the role of external strain may be coordinating the activities of a set of dynein molecules along a microtubule simultaneously to a stationary mode.

Early studies on sliding disintegration of axonemes (Sale and Satir, 1977; Fox and Sale, 1987) or studies on in vitro motility assay (Paschal et al., 1987; Vale and Toyoshima, 1988) reported only unidirectional motility induced by dynein. However, ATP-dependent bidirectional stepping of dynein (Shingyoji et al., 1998) and bidirectional force generation by dynein (Shingyoji et al., 2015) under strain imposed by an optical trap has been reported in single molecule (or two to four molecule) analyses, and bidirectional stepping of a single cytoplasmic dynein molecule under strain imposed by an optical trap has also been reported (Gennerich et al., 2007). Although backward sliding was induced only in rare cases (four cases out of 574 trials of planar bending) in the present study, they suggest that ensemble of dynein molecules can collectively support backward sliding of microtubules in response to external strain, either by strain-dependent backward stepping of dyneins or by backward force generation by dyneins. Backward strain was suggested to be important for induction of backward sliding in the present study, and this direction is consistent with the previous reports on single molecule analyses (Gennerich et al., 2007; Shingyoji et al., 2015). In the present study, dyneins adsorbed onto the glass surface are supposed to be oriented in every direction, and we cannot preclude a possibility that some dyneins are attached to a microtubule in the wrong direction and that such dyneins may be involved in the backward sliding. However, the induction of backward sliding in microtubules interacted with dyneins arrayed on doublet microtubules (Shingyoji et al., 2015; Shingyoji, 2017) indicates that uniformly oriented dyneins are also capable of backward sliding. A possible interpretation of this finding in the context of beating axonemes is discussed below.

Decrease in sliding velocity was induced while the bend was formed, and the velocity recovered after the bend dissolved (Fig. 3). This suggests that an ensemble of outer arm dynein molecules can not only be switched between on/off or stop/go, but can modify collective sliding velocity in a reversible manner in response to a mechanical signal in the presence of a high concentration (1 mmol l⁻¹) of ATP. It is not yet clear whether the curvature of microtubules is important, or if some kind of external strain is important for the induction of velocity change, but considering the previous reports that decrease in sliding velocity of a single axonemal and cytoplasmic dynein molecule can be induced by backward strain imposed by an optical trap (Hirakawa et al., 2000; Rai et al., 2013), backward strain resulting from the elasticity of bent microtubules may be important for induction of decrease in sliding velocity in the present study.

The roles of outer arm dynein and inner arm dynein in the axonemes have been suggested to be somewhat different. While flagella of inner arm-deficient mutants of Chlamydomonas reinhardtii showed a reduction in shear amplitude with only a small reduction in beat frequency, those of outer arm-deficient mutants showed beating with a reduced frequency to about half that of wild-type flagella (Brokaw and Kamiya, 1987). Also, sea urchin sperm flagella that were extracted with demembranating detergent along with 0.5 mol l⁻¹ KCl to remove outer arm dynein could be reactivated with ATP to produce bending waves whose wave forms are not significantly different from normal, but have a reduced beat frequency of about half that of control demembranated sperm without exposition to 0.5 mol l⁻¹ KCl (Gibbons and Gibbons, 1973). These suggest that the primary function of outer arm dynein is increasing propagation velocity (Brokaw, 1994, 1999) by mainly contributing to the metachronal sliding, exerting a maximum shear force and sliding at the bend regions of an axoneme, and decreasing sliding velocity to nearly zero on the straight regions, where the reversal of the inter-doublet sliding direction occurs (Brokaw and Gibbons, 1975; Gibbons, 1981) (see upper panel in Fig. 9A). In the present study, stoppage and backward sliding of purified outer arm dynein on a glass surface were induced by backward strain (Fig. 9B). This might reflect the regulatory mechanism of the sliding activities (such as sliding velocities and directionalities) of outer arm dynein in a beating axoneme, which could be regulated and coordinated by backward strain imposed on dynein at straight regions as elastic resistance (Fig. 9A) (Brokaw, 2001).

This strain-dependent regulatory mechanism alone may not be sufficient to explain the whole picture of flagellar movement. For example, in asymmetrically beating flagella, synchronous sliding occurs throughout the region distal to the proximal principal bend (Goldstein, 1977). The synchronous sliding that accompanies the growth of the proximal principal bend seems to impose shear strain on dyneins in the distal regions, but it does not apparently affect the propagation of the pre-existing bends (Goldstein, 1977). It is likely that the dyneins are capable of integrating various components of information, which include shear strain, as well as transverse force that accompanies the curvature of an axoneme (Lindemann, 1994a,b; Lindemann and Lesich, 2015), and signals related to calcium ion (Brokaw, 1979). The present result suggests that isolated outer arm dyneins have the ability to respond to mechanical signals of strain, which might reflect an important aspect of the communication within a beating axoneme and a basic requirement for the overall flagellar oscillation. We hope that the present experimental system provides a basis for more detailed, quantitative approach for understanding the mechanical properties of dynein in the future.

We thank the staff of the Misaki Marine Biological Station at the University of Tokyo for supplying sea urchins Pseudocentrotus depressus, and the Research Center for Marine Biology at Tohoku University for supplying sea urchins Hemicentrotus pulcherrimus and Strongylocentrotus nudus.