High-speed motility originates from cooperatively pushing and pulling flagella bundles in bilophotrichous bacteria.

Bacteria propel and change direction by rotating long, helical filaments, called flagella. The number of flagella, their arrangement on the cell body and their sense of rotation hypothetically determine the locomotion characteristics of a species. The movement of the most rapid microorganisms has in particular remained unexplored because of additional experimental limitations. We show that magnetotactic cocci with two flagella bundles on one pole swim faster than 500 µm·s-1 along a double helical path, making them one of the fastest natural microswimmers. We additionally reveal that the cells reorient in less than 5 ms, an order of magnitude faster than reported so far for any other bacteria. Using hydrodynamic modeling, we demonstrate that a mode where a pushing and a pulling bundle cooperate is the only possibility to enable both helical tracks and fast reorientations. The advantage of sheathed flagella bundles is the high rigidity, making high swimming speeds possible.

Introduction

The understanding of microswimmer motility has implications ranging from the comprehension of phytoplankton migration to the autonomously acting microbots in medical scenarios (Sengupta et al., 2017; Felfoul et al., 2016). The most present microswimmers in our daily lives are bacteria, most of which use flagella for locomotion. Well-studied examples of swimming microorganisms include the peritrichous (several flagella all over the body surface) Escherichia coli with an occasionally distorted hydrodynamic flagella bundling (Turner et al., 2000) and the monotrichous (one polar flagellum) Vibrio alginolyticus, which are pushed or pulled by a flagellum and exploit a mechanical buckling instability to change direction (Xie et al., 2011; Son et al., 2013). The swimming speeds of so far studied cells are in the range of several 10 µm s−1 and their reorientation events occur on the time scale of 50–100 ms (Son et al., 2013; Berg and Brown, 1972).

Magnetococcus marinus (MC-1) is a magnetotactic, spherical bacterium that is capable of swimming extremely fast (Zhang et al., 2014; Fenchel and Thar, 2004; Bazylinski et al., 2013; Garcia-Pichel, 1989). MC-1 as well as the closely related strain MO-1 (Ruan et al., 2012) are equipped with two bundles of flagella on one hemisphere (bilophotrichous cells). The bacterium also features a magnetosome chain, which imparts the cell with a magnetic moment (‘magnetotactic’ cell). They are assumed to swim with the cell body in front of both flagella bundles, which synchronously push the cell forward (Zhang et al., 2014). This assumption leads to helical motion in the presence of a strong magnetic field, which exerts a torque on the cell’s magnetic moment, as seen in hydrodynamic simulations (Yang et al., 2012). In the absence of a magnetic field, this model predicts rather straight trajectories.

Our observations disagree with the above-mentioned model, indicating that an understanding of the physics of their swimming is still missing, even though proof of concept biomedical applications of these bacteria have already emerged (Felfoul et al., 2016). Here, we confirm that MC-1 cells reach speeds of over 500 µm s−1 (Figure 1D) and observe that the cells travel along a double helical path, which has not been reported for bilophotrichous cells so far. In addition, we observed that this rapid movement is complemented by an extremely fast reorientation ability (less than 5 ms). We connect the flagella bundle architecture and the swimming mechanism by hydrodynamic simulations and show that only a striking cooperative movement where one flagella bundle pushes while the other pulls the cell explains these motility characteristics.

Figure 1.

3D tracking in physiochemically controlled conditions.

(A) MC-1 cells were transitioned into agar-free medium, the solution was degassed to remove the oxygen and inserted into a microcapillary. Due to one open and one closed end, an oxygen gradient formed and the cells accumulated near their preferred microoxic conditions (the oxic-anoxic-interface, OAI). (B) The cells were observed near the band using 400 fps phase contrast video microscopy with a spherical aberration, which causes interference patterns around the spherical swimmers. These patterns can be correlated with patterns from silica beads of known height relative to the focal plane. (C) A tracking algorithm enables high-throughput 3D tracking of the microswimmers (Taute et al., 2015). Colors indicate different cells. (D) Individual tracks were analyzed and a clockwise helical travel path with a radius close to the cell diameter was found as well as instantaneous traveling speeds between 100 µm s−1 and 500 µm s−1. Tracks can be interrupted by rapid reorientation events that last only 2.5–5 ms. The helix parameters like pitch and period time do not change before and after an event, but apparently do so in the projected 2D tracks (see projected shadow in D).

Top simulation parameters: Theta (reorientation angle) was 90 °, velocity was 100 µm/s, and runtime was 0.86 s. Bottom parameters: Theta was 170 °, velocity was 14 µm/s and runtime was 0.86 s. The mean time each reorientation lasted was 0.14 s for both cases.

(A) Turning angle θ after a reorientation event, (B) effective traveling speed, (C) runtimes and D) runlengths in-between events.

The radius, pitch and period time of the helix remain constant before and after the event, which is identified as a drop in velocity to 0 and which lasted 3.75 ms ± 1.25 ms (error due to frame rate). While the top view indicates a change in helix parameters, the 3D view reveals no apparent changes.

Results

MC-1 cells were observed in a physicochemically controlled environment that resembled the cell’s natural habitat (Adler, 1973; Figure 1A). The cells were introduced into a flat, rectangular microcapillary tube with a height of 200 µm, where they accumulated in a band near their preferred oxygen conditions. The cells were imaged near this band (Frankel et al., 1997; Fenchel, 1994) at central height to avoid surface interactions. The capillary was placed at the center of three orthogonal Helmholtz coils (Bennet et al., 2014), which were used to cancel the Earth’s magnetic field with a precision of 0.2 µT after the band had formed. Hence, the cells’ motion could be observed in the absence of significant magnetic torques. Tracking was performed in 3D at 400 frames s−1 (fps) (Figure 1B). A high-throughput tracking method was used (Taute et al., 2015) (see Materials and methods) for the reconstruction of the tracks (Figure 1C and D).

Our first set of observations revealed that the cells traveled on helical paths (with a pitch of 5.3 µm ± 1.3 µm, a diameter of 1.7 µm ± 0.2 µm, and a period of 46 ms ± 32 ms, n = 65, errors are standard deviations) in the absence of magnetic torques (parameter extraction was performed with an automated track analysis algorithm, which was validated against simulated swim tracks, see Figure 1—figure supplement 1 and Figure 1—figure supplement 2). In addition, abrupt changes of the direction of the helical axis were observed (e.g. around 90° in Figure 1—figure supplement 3). These directional changes occur within 2.5 ms to 5 ms, at least one order of magnitude faster than any previously analyzed reorientation events (Son et al., 2013; Berg and Brown, 1972). Directional changes did not occur via a continuous modulation of the ratio of radius and pitch (Crenshaw and Edelstein-Keshet, 1993), as it has been observed as a part of the chemotaxis of sperm (Jikeli et al., 2015). Rather, the helix parameters were the same before and after such an event. 3D tracking is essential to obtain this conclusion, as projected 2D tracks exhibit apparent changes in the helix parameters (see projected shadow of the track in Figure 1D and Figure 1—figure supplement 3).

Figure 1—figure supplement 1.

Validation of event identification algorithm for 3D tracking on simulated tracks with known parameters.

Top simulation parameters: Theta (reorientation angle) was 90 °, velocity was 100 µm/s, and runtime was 0.86 s. Bottom parameters: Theta was 170 °, velocity was 14 µm/s and runtime was 0.86 s. The mean time each reorientation lasted was 0.14 s for both cases.

Figure 1—figure supplement 2.

MC-1 swimming statistics.

(A) Turning angle θ after a reorientation event, (B) effective traveling speed, (C) runtimes and D) runlengths in-between events.

Figure 1—figure supplement 3.

Oblique and top view of a typical event in an MC-1 swimming track.

The radius, pitch and period time of the helix remain constant before and after the event, which is identified as a drop in velocity to 0 and which lasted 3.75 ms ± 1.25 ms (error due to frame rate). While the top view indicates a change in helix parameters, the 3D view reveals no apparent changes.

The cells were further examined at 1640 fps in high-intensity dark-field video microscopy to visualize the cell body and the flagella bundle movement in detail (Figure 2, Materials and methods). A representative track of the cell body movement is shown in Figure 2A together with the tracked velocity. At such frame rates, a more complex movement pattern becomes apparent, which was not detectable during the 3D tracking at 400 fps. The cell track can be represented by a superposition of two helices, a small helix on a large helix (Figure 2B), resulting in a position over time . The large helical track featured a pitch p of ~4 µm, a radius r of ~1 µm and a period T of 72 ms. The small helix featured a pitch of ~0.66 µm, a radius r of ~0.125 µm and a period T close to 14.4 ms. The ratio between this specific track’s period times was close to 6, which was later used to choose the flagella bundle’s rigidities and torque in the numerical simulations. The parameters for the small helix have been determined for a single track and do not possess statistical information. However, comparable double helical paths were apparent in all recorded tracks, including the cell track of the flagella imaging attempt (Figure 4, Video 1 and Video 2).

Figure 2.

High-speed dark-field video microscopy at 1640 fps reveals a dual helical travel path of a MC-1 cell during free swimming.

In (A) the tracked path is displayed after smoothing by a 5-point moving average filter and then plotted together with the cell velocity. Green arrows indicate velocity maxima. In (B) it is shown that the projected swimming path and projected velocity can be described by a projection of a large 3D double helix .

Video 1.

Cell observation in a high-intensity dark-field video microscopy experiments at 1424 fps and 60x magnification with cell tracking.

Video 2.

Cell observation in a high-intensity dark-field video microscopy experiments at 1424 fps and 60x magnification with cell tracking and tracking of white spots on the cell surface.

The flagella bundle morphology and movements were imaged in transmission electron microscopy (TEM) and in high-intensity dark-field video microscopy (Figure 3, Video 1 and Video 2). In the video microscopy at 1424 fps, short fibers next to the cell body and bright spots on the cell surface could be observed, which we identified as a part of a flagellum bundle close to the cell surface (Son et al., 2013). Although a state-of-the-art flagella imaging method, a sufficiently high framerate and a photon density close to the cell death limit were chosen, the resulting images reveal only little information about the exact flagella bundle positions and dynamics. This is the result of the combination of the small size of the cells and their extraordinary high swimming speed and possibly their strong flagella bundle movement.

Figure 3.

Flagella bundle imaging.

(A) TEM image of an MC-1 cell stained with uranyl acetate. Two flagella bundles emerge from the MC-1 cell body. Each bundle features seven flagella, which emerge from a cavity on the cell surface. The individual flagella are bundled by a sheath (scale bar is 1 µm). (B-D) Individual frames from the high-speed dark-field video microscopy experiment at different points in time (Video 1, scale bars 1 µm). Bright spots appeared on the cell surface and next to the cell, which were identified as the parts of the flagella bundles that were closest to the cell surface. As the cell swam downwards in the image (blue track), the flagella bundle spots appeared in front of and behind the cell (relative to its overall movement direction). When comparing the times of B and C, the abrupt change in position indicates that another flagella bundle moved into focus of the microscope. When comparing C and D, it becomes apparent that the spots appear at nearly the same position on the cell at nearly the same horizontal position in the helical movement pattern, indicating a periodic flagella movement pattern that causes the helical track.

Despite the difficulties in imaging the flagella bundles in full length, the position of the flagella bundle near the cell surface were tracked together with the cell’s trajectory over 85 ms, which corresponds to 1.6 periods on the large helical trajectory of the cell (Figure 4). The observed movement pattern is more complex than previously assumed (Felfoul et al., 2016; Frankel et al., 1997; Bazylinski et al., 2013; Waisbord et al., 2016). The two flagella bundle positions moved rapidly around the cell. Crucially, one flagella bundle position is often seen in front of the cell (relative to its swimming direction), contrary to the model of two pushing flagella bundles, which remain behind the cell body. Additionally, the bright spots’ movement pattern featured the same periodicity as the large helical swimming track of the cell (also compare Figure 3C and Figure 3D).

Figure 4.

Positions of flagella bundles on the cell over 85 ms at 1424 fps.

(A) The cell swam from top to bottom in the field of view. Moving bright spots were identified as the parts of the flagella bundles that were closes to the cell surface. The cell’s helical traveling path had a period time of 52 ms. (B) The positions were tracked over time and are depicted relative to the center of the cell, represented by different spheres for different time intervals. (C) The schematic of the cell’s large helical track is color-coded according to the different time intervals where the flagella bundle parts were visible to highlight the periodicity of the flagella bundle movement.

We turned to numerical simulations of the cell’s swimming behavior, to develop a deeper understanding of the mechanisms of propulsion and rapid reorientation and to compensate the missing information from flagella bundle imaging. We performed Stokesian dynamics simulations (Adhyapak and Stark, 2015) for a spherical cell body (1.3 µm in diameter) with two discretized helical filaments (4 µm long and 50 nm thick if not stated differently) representing the flagella bundles (see Materials and methods for details). A large helical path of a microswimmer is produced from an off-axis (relative to the swimming direction) torque which continuously changes the direction of the thrust force. In case of a bilophotrichous cell, this requires a significant asymmetry in the propulsion force vectors of the two flagella bundles. Five possible asymmetries in flagella bundle configuration were considered to produce this torque and the resulting swimming paths were compared to experimental data (Figure 5). Three asymmetric configurations were ruled out numerically, as they did not result in a significant match between simulated and experimentally observed helix diameter, helix pitch and speed. These configurations are: A difference in flagella bundle length (Figure 5—figure supplement 1), a difference in motor strength (Table 1) and an asymmetry in the equilibrium angle of the two flagella bundles relative to the cell surface (Shum, 2019) (see Appendix 1). Another scenario would be a periodic, time-dependent movement of the flagella equilibrium angles relative to the cell surface (Nguyen and Graham, 2017). However, constant offsets in these angles already did not produce any significant matches between experiment and simulation. This indicates that this scenario will also fail this test and was not investigated further. Although not fully reaching experimentally observed helix pitches, the asymmetry in sense of motor rotation was the only scenario producing significant matches in helix diameter and cell speeds.

Figure 5.

Simulations of MC-1 swimming dynamics with one pushing and one pulling flagellar bundle.

The principle cell and flagella bundle arrangement is shown in (A) at four distinct time points where the cell body diverges strongly from the helix axis. (B) Shows the projected track of a simulated MC-1 with the flagellar opening angle 60o together with the projected speed. The results are comparable to measured data from Figure 2A. (C) Histogram of turning angles for the reorientation events seen in experiments and in simulations, where reorientation results from periods of synchronous rotation. (D) Validation of the cooperative pushing and pulling model in the presence of a strong magnetic field (3 mT). A third hyper-helix is observed in both experiment and simulation.

No match could be reached for either pitch or diameter in the simulations.

While the pitch is not fully matched for any opening angles between the two flagella, the helix diameter and cell’s speed could be simulated.

The tested flagella rotation patterns are shown together with the resulting cell trajectories for five different conditions: no magnetic field, weak magnetic field (50 μT) for a cell with magnetic moment parallel and perpendicular to the flagellar axes, and strong magnetic field (5 mT), again with magnetic moment parallel and perpendicular to the flagellar axes.. In the absence of a magnetic field, two pushing flagella produce small helices, two pulling flagella cause a strongly distorted movement pattern and only a cooperatively pushing and pulling flagella combination reproduces the experimentally observed double helical motion shown in Figure 2A. A strong magnetic field results in the hyper-helical motion shown in Figure 4D, if the magnetic moment is not parallel to the flagellar axes.

Figure 5—figure supplement 1.

Swimming path parameters from numerical simulations with an asymmetry in flagella lengths compared to experimental values.

No match could be reached for either pitch or diameter in the simulations.

Table 1.

Simulated swim track parameters for an asymmetry in motor torques for MC-1 cells for different opening angles between the two flagella bundles.

The second row gives statistical averages of experimental values for diameter (D), pitch (P), period time (T) and speed (Vt).

Opening angle	Tm2/Tm1	D (μm)	P (μm)	T (ms)	Vt (μm/s)
	Experiment	1.7	5.3	46.0	100.0
60 °	0.1	0.43	0.93	47.4	52.33
	0.5	0.29	0.73	20.2	58.85
	0.9	0.18	1.58	33.4	88.42
80 °	0.1	0.54	1.08	54.4	54.76
	0.5	0.39	1.09	32	60.44
	0.9	0.20	1.61	34.8	84.14

Since TEM images did not allow for a precise determination of the opening angle between the two flagella bundles (with respect to the body center; see, for example, Figure 3A), a wide range of opening angles (30°−120°) was used in the simulations (Figure 5—figure supplement 2). The effect of the flagellar motors was included as a torque at the base of the helices, which rotates the filament, and a counter-torque, which rotates the cell body. We considered the two bundles to rotate independently either counter-clockwise (CCW) or clockwise (CW). The parameters for bending rigidity, torsion rigidity and torque of the flagella bundles were adjusted to reproduce the observed movement characteristics (i.e. the parameters of the small helices and velocities). A motor torque of 12 pN µm, about 3.5 times the motor torque of an E. coli cell, had to be chosen together with high isotropic bending and twisting rigidities of 7 pN μm2, about two times larger than for single flagella (Adhyapak and Stark, 2015). We assume that this increase is attributed to the structure of the flagella bundle, where seven protein motors power seven sheathed flagella cooperatively (Figure 3A). Only this assumption allowed for stable and high swimming velocities in the simulations, indicating that the function of the flagella bundle is to combine high torques with high rigidities.

Figure 5—figure supplement 2.

Comparison of cell path parameters from experiments and the pushper-puller simulation scenario.

While the pitch is not fully matched for any opening angles between the two flagella, the helix diameter and cell’s speed could be simulated.

Different swimming scenarios arising from combinations of CW and CCW rotation of the two bundles were simulated: both flagella bundles pushing the cell (CCW and CCW), both flagella bundles pulling (CW and CW) and one flagellum pushing, one pulling (CCW and CW). The fact that flagella can pull cells is well established (Son et al., 2013; Constantino et al., 2016) and our simulations show that only the CCW and CW model results in the double-helical tracks experimentally observed (Figure 4A and B, Figure 5—figure supplement 2, Video 3 and Video 4). Tracking the simulated cells resulted in time traces of the velocity and the projected position that are strikingly similar to the experimental ones (for a quantitative comparison, see Figure 5 and Figure 2A). We tested the dependence of the trajectories on model parameters, in particular the opening angle and motor torque (Table 2 and Table 3). Double helical trajectories were observed for flagellar opening angles in the whole parameter range, with diameter and speed in agreement with the experiments (but no fully quantitative agreement for the pitch of the large helix).

Video 3.

Simulation of cell swimming with two pushing flagella.

Video 4.

Simulation of cell swimming with one pulling and one pushing flagella.

Table 2.

Swimming features of MC-1 cells using CCW and CW swimming mechanism for simulations with different flagellar opening angles.

The given output parameters are the helix diameter, its pitch, the period time (period), the effective velocity (Vz) and the instantaneous velocity (Vt).

Flagellar opening angle	Diameter (μm)	Pitch (μm)	Period (ms)	Vz (μm/s)	Vt (μm/s)
30°	2.1	3.0	144	21	76
45°	1.7	2.3	88	27	87
60°	1.4	1.7	59	30	96
80°	1.1	1.2	39	32	106
100°	0.9	1.1	34	32	108
120°	0.7	1.1	34	33	106

Table 3.

Swimming features of MC-1 cells for different motor torques.

The 3.5 times increase of motor torque compared to E-coli cells was chosen for the simulations described in the main text due to the best fits of swimming track parameters.

Motor torque (Tm/Tm-Ecoli)	D (μm)	P (μm)	T (ms)	Vt (μm/s)
3.5	1.4	1.7	59	96
3	1.44	1.56	74.4	80.84
2.5	1.43	1.64	84.2	69.41
2	1.49	1.39	101.2	53.27

We also tracked the position of the flagella bundle positions in the simulations (by tracking the first beads of each discretized flagellum). Both flagella bundles rotated with 100 Hz - 150 Hz and one period of the rotation coincides with one period of the large helix in the cell’s trajectory. This agrees with the flagella bundle movement in the experiments (Figure 4).

Our model suggests a pushing flagella bundle together with a pulling flagella bundle rotating in opposite senses. The experimentally observed rapid reorientation events that lasts between 2.5 ms and 5 ms are most reasonably produced by a sharp change in the rotation of at least one flagella bundle. We tested different scenarios (Figure 5—figure supplements 3–5) and found a transiently synchronous rotations of the two flagella bundles to be the most likely process. We tested this by changing the sense of rotation from CCW and CW to CCW and CCW for 4 ms in the simulations and calculating the angle between the trajectory segments before starting and after finishing this transient CCW and CCW step (outtake in Video 5). This procedure indeed resulted in rapid reorientation with a change in direction by 80° ± 8° (Figure 4C), in agreement with the 94° ± 39° change seen in the experiments (errors are standard deviations). The standard deviation of the directional change is small in our simulations, where only the runtime was varied, compared to the experimental value. The mismatch likely arises from biological diversity in flagella lengths and opening angles, but also due to shifts in local physiochemical conditions, which can for example influence the motor torque (Son et al., 2013). A transient buckling deformation was observed in the simulated flagella bundles during an event, which caused the fast reorientations.

Figure 5—figure supplement 3.

Reorientation angle histogram for the following transient flagella bundle configuration: CCW and CW → CCW and 0 → CCW and CW (the rotation of the puller flagella is stopped during reorientation).

Figure 5—figure supplement 5.

Reorientation angle histogram for the following transient flagella bundle configuration: CCW and CW → CCW and CCW → CW and CCW → CCW and CCW → CCW and CW.

Video 5.

Example simulation of reorientation events.

To further validate our simulations, we predicted the cell’s swimming behavior at high magnetic fields using our simulation without changing further parameters. The direction of the cell’s magnetic moment was assumed to be perpendicular to the bisector of the two flagella bundle axes (further scenarios can be found in Figure 5—figure supplement 6). The simulations showed an additional, large hyper-helical movement pattern (with diameter µm and pitch µm). The same pattern could thereafter be found in experimental data with similar parameters ( µm and µm). Typical MC-1 hyper-helical trajectories from simulations and experiments are shown in Figure 5D.

Figure 5—figure supplement 6.

Simulations of different scenarios of MC-1 swimming with varying magnetic fields and magnetic moment orientation.

The tested flagella rotation patterns are shown together with the resulting cell trajectories for five different conditions: no magnetic field, weak magnetic field (50 μT) for a cell with magnetic moment parallel and perpendicular to the flagellar axes, and strong magnetic field (5 mT), again with magnetic moment parallel and perpendicular to the flagellar axes.. In the absence of a magnetic field, two pushing flagella produce small helices, two pulling flagella cause a strongly distorted movement pattern and only a cooperatively pushing and pulling flagella combination reproduces the experimentally observed double helical motion shown in Figure 2A. A strong magnetic field results in the hyper-helical motion shown in Figure 4D, if the magnetic moment is not parallel to the flagellar axes.

Discussion

Magnetococci are exceptional swimmers with respect to both their high speed and their reorientation swiftness. In addition, they are likely to make use of a previously unrecognized pattern of motion of their flagella bundles, with one bundle pushing the cell body and the other pulling it. Key to observing this pattern of motion was the 3D tracking of single cells in the absence of magnetic torques to observe the unexpected two-helix trajectories, together with hydrodynamic simulations. The hierarchically organized flagella bundles provide a high torque and rigidity, necessary to reach record speeds of over 200 body lengths per second, while the unusual type of coordination of the two bundles provides a mechanism for rapid reorientation. Magnetotactic cocci such as MC-1 thrive and generally represent the most abundant MTB in aquatic environments (Lefèvre et al., 2015). Their unusual motility is certainly an adaptation that represents a selective advantage that make them competitive in the highly coveted biotopes that are the oxic-anoxic interfaces (Brune et al., 2000). Indeed, the observed geometry of this resulting swimming pattern is reminiscent of the situation recently reported for magnetospirilla (Murat et al., 2015), despite the difference in body plan. The magnetosprilla have two flagella at opposite cell poles, and a magnetic moment parallel to the flagellar axis, while the flagella of the cocci studied here are attached on one hemisphere of the cell and almost perpendicular to the magnetic moment. Nevertheless, both swim aligned with a magnetic field with one flagellum ahead and one trailing. In summary, record-breaking cells like MC-1 can help to understand physical limits of natural microswimmers and provide design principles for their artificial counterparts.

Materials and methods

Cell medium and culturing

MC-1 was cultured similarly to the procedure reported by Bazylinski et al. (2013). Artificial sea water (ASW) was used as a base medium, containing 20 g NaCl, 6 g MgCl2, 2.4 g Na2SO4, 0.5 g KCl and 1 g CaCl2 per liter H20. To this was added (per liter) the following, in order, prior to autoclaving: 0.05 mL 0.2% (w/w) aqueous resazurin, 5 mL Wolfe’s mineral solution (ATCC, MD-TMS), 0.3 g NH4Cl, 2.4 g HEPES and 1.6 g agar (Kobe I, Carl Roth). The medium was then adjusted to pH 6.3 and autoclaved. After the medium had cooled to about 45°C, the following solutions were added (per liter), in order, from previously sterile-filtered stock solutions: 0.5 ml vitamin solution (ATCC, MD-VS-10mL), 1.8 mL 0.5 M potassium phosphate buffer, pH 7, 3 mL0.01 M FeCl2 and 40% (w/w) Na thiosulfate. Finally, 0.4 g cysteine was added (per liter), which was made fresh and filter-sterilized indirectly into the medium. The medium (12 mL) was dispensed into sterile Hungate tubes after verifying a pH of 7.0. All cultures were incubated at room temperature (~25°C) and, after approximately one week, a microaerobic band of MC-1 formed at the oxic–anoxic interface (pink-colorless interface) of the tubes. The cells were harvested in volumes of 1 mL from that region and magnetically transferred to ASW for experiments. The transfer step was necessary to remove agar for swimming experiments and to minimize background scattering in dark-field microscopy.

Cell morphology analysis

The flagella bundle length was determined with ImageJ from images taken with a Zeiss EM 912 Omega transmission electron microscope using an acceleration voltage of 120 kV. The cells were dried on a carbon film on a regular TEM copper grid and stained with 4% uranyl acetate for 6 min. Due to the staining, the cell walls appeared electron dense and covered the sight on flagella on top or below the cells. Hence, we added the average cell radius to the mean of the flagella length. A mean flagella bundle length of 3.3 µm ± 0.4 µm (n = 27) resulted. The size of the non-dried cells were measured with ImageJ from images taken with a LSM780 (Zeiss; Germany) confocal microscope. The mean size was 1.3 µm ± 0.1 µm (n = 103).

Microcapillary experiments

1 mL of a freshly harvested sample was degassed using nitrogen for 15 min and the sample was introduced into a rectangular micro-capillary (VitroTubes, #3520–050,) by capillary forces. One end of the capillary was sealed with petroleum jelly and the capillary was mounted on a microscope slide that was used to hold the sample on the microscope stage. The oxygen diffusion from the open end caused an oxygen gradient inside the medium, which led together with the oxygen consumption of the cells to the formation of a microaerobic bacteria band. The band formed in the presence of a 50 µT magnetic field towards the sealed end. The tracks were taken after 30 min of microcapillary infiltration at 0 µT.

3d tracking experiments

3D swim tracks were recorded at 400 fps in a microcapillary in the vicinity of the microaerobic band (Nikon, S Plan Fluor ELWD,×40, Ph2, NA 0.6; NA 0.76 condenser lens; Ph2 aperture ring, 635 nm LED illumination). The 3D tracks were reconstructed using the high-throughput phase contrast reference method by Taute et al. (2015). A spherical aberration was introduced using a misalignment of the correction collar of the ×40 phase contrast objective to a cover slip thickness correction of 1.2 mm. The aberration caused inference pattern, which can be correlated with the relative height of the microswimmer. A custom made microscope platform, developed by Bennet et al. (2014), was used, which features three orthogonal Helmholtz coil pairs around the sample position. The setup can generate homogeneous fields at the sample position with arbitrary direction with a precision of 0.2 µT. The Earth’s magnetic field was canceled or an artificial field towards low oxygen conditions was generated during a capillary experiments.

Dark-field microscopy

Flagella bundle positions were visualized at 1424 fps using high-intensity dark-field microscopy (Nikon 60×, 0.5–1.25 NA CFI P-Fluor oil objective at 0.75 NA; 1.2 NA oil condenser; mercury lamp illumination) and an Andor Zyla 5.5 (10-tap) camera (6.5 μm per pixel). The cells were placed inside a 10 µm deep chamber in ASW. A deeper chamber did not allow for successful dark-field imaging of flagella bundles due to an increase in noise from background scattering. The focal plane was adjusted to the center of the chamber, such that interactions between the observed flagella bundles and the chamber surfaces were avoided. Presumably due to the flagella bundle size and the high rotation speed of the cell and of the bundles, a direct observation of the whole flagella bundles was not successful and only a small section of each flagellum could be visualized. A sub-millisecond exposure time set the requirement for high photon intensity at the sample position. The intensity was increased until the cells melted instantly when swimming into focus. A green filter prevented the melting while still facilitating sufficient brightness. Scrupulous cleanliness at all optical interfaces was obligatory. High-speed cell body tracking could be accomplished at the center of a 200 µm deep chamber. Optimized visualization was achieved at 1640 fps using dark-field microscopy without the critical illumination from flagella tracking (Zeiss 60×, 1.0 NA; 1.2 NA oil condenser; halogen lamp illumination). The dark-field setup did not allow for a cancellation of external magnetic fields. Measurements of the magnetic field at the capillary position yielded Bx = −195 µT ± 0.2 µT, By = 60 µT ± 0.2 µT and Bz = 27 µT ± 0.2 µT.

Cell and flagella tracking software

3D track reconstruction was performed in MatLab (The MathWorks) using an adapted version of the code from Taute et al. (2015). For automated analysis of the reorientation angle distribution, the script was successfully tested against simulated data with known reorientation angles (See SI). The helix parameter determination was performed automatically on tracks extending a duration of 0.4 s, with a mean-square-displacement of at least 10 µm2 and a mean opening angle between all consecutive velocity vectors of less than 60 ° to exclude strongly irregular tracks. The same exclusion parameters have been used for the reorientation angle analysis. Simultaneous dark-field cell and flagella tracking was performed using an in-house semi-automatic MatLab program. Tracking of the cell trajectory at 1640 fps without the flagella movement was performed using the TrackMate plugin of ImageJ.

Hydrodynamic simulations

Each flagellum was modeled as a helical filament and a rotary motor. The helical filament was discretized with 20 beads with a discretization distance of 200 nm and bead diameter of 50 nm. Excluded volume interactions between all particles are considered using a truncated Lennard-Jones potential. Hydrodynamic interactions are taking into account using Stokesian dynamics simulation method having the translational anisotropic friction coefficients of  pNs/μm2 and  pNs/μm2 and the rotational friction of  pNs for the flagellum beads and  and for the translational and rotational friction coefficients of the cell body. Irrespective of the high motor torque and isotropic bending and twisting rigidities mentioned in the main text, a stretching rigidity of 1000 pN, comparable to that of single flagella, could be chosen. A Rotne-Prager matrix was used for calculating the cross-mobilities and cross-hydrodynamics (Dhont, 1996). The swimming dynamics of the model cell at low Reynold number was calculated by solving the translational and rotational Stokes equations of motion for the cell body, flagellar beads and the bonds between them. A second-order Runge-Kutta algorithm (Sewell, 1988; Press et al., 1992) and simulation time-steps of 10−7 s were used to solve the equations of motion numerically.

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

This paper reports on the discovery of an unusual and fascinating form of motility in a magnetotactic bacterium. Through a combination of experiment and theory the authors have found that the two flagellar bundles work in opposite ways – one pushing and one pulling the cells through their fluid environment in double-helical paths. The very fast motility and very rapid trajectories changes appear to arise from these features. This work will surely be of interest not only to those interested in the biology of motility, but also physical scientists interested in the fluid dynamics of locomotion.

Decision letter after peer review:

Thank you for sending your article entitled "High-speed motility originates from cooperatively pushing and pulling flagella bundles in bilophotrichous bacteria" for peer review at eLife. Your article is being evaluated by three peer reviewers, and the evaluation is being overseen by a Reviewing Editor and Detlef Weigel as the Senior Editor.

The essence of the criticisms concerns the interpretation of the experimental observations (and the possibility of achieving better imaging of the flagella) and the analysis of the computational results. In both cases further details are needed, and in the former it appears that additional imaging would be in order.

Reviewer #1:

In my opinion the topic of the article is interesting as it identifies a new flagellar arrangement for motility (though similar simultaneous pushing/pulling has been observed for species with bundles at opposite poles rather than the same hemisphere, both in magnetospirilla). However, I am not completely convinced by the authors' evidence. As detailed below, while the story is quite plausible, the experiments seem to not have key evidence which I would have expected to have easily been observed if the scenario were true, and the numerics are not performed (or perhaps just not described) in a way that convincingly rules out other hypotheses.

1) In the dark field imaging, bright spots on the cell body are identified with flagella. This identification is not obvious to me, and it is quite puzzling why the authors do not actually image the flagellar bundles. If the bundles were indeed extended from the body in front and behind the cell, I would expect that they could be visualized which would definitively prove their proposed configuration. In Son, Guasto and Stocker, 2013, sheathed flagella are quite readily imaged. In my experience, there may be difficulty seeing flagella with darkfield at 1640 fps, but they should be readily imaged at <200 fps using their camera. Note that 200 fps is more than fast enough to resolve the larger helical motion with period 72 ms. It is unclear if the authors attempted this. If the authors tried and were not able to see a leading and lagging bundle, I would take that as evidence against their claim.

2) The authors describe the double helical path as "unexplored" in the Introduction, but based on my knowledge, double helical paths are what should be expected from most types of propulsion by flagellar bundles, with the main distinguishing feature being the size of the larger helix. The smaller helix has in the past been attributed to the rotation of the bundle or flagellar helix, see Keller and Rubinow, 1976. The larger helix arises anytime there is non-axisymmetric propulsion, see Hyon, Powers, Stocker, Fu 2012 and has been observed in many species, albeit with varying sizes of the larger helix. As discussed in point 3 below, an important implication of this is that the bar for numerical simulations cannot be simply qualitative as in "straight" vs. "double-helical," but must involve some quantitative analysis of the trajectory pitch and radius.

3) The numerics have not been described in a way that clearly rules out alternative hypotheses.

First, as mentioned above, the quantitative details of the helical trajectory are important, but these are not well-matched. The authors say that their trajectories match the experimental helical diameter and period of the trajectories but not the pitch. The results are strongly dependent on the angle between the bundles, which is used as a fitting parameter. They say that they believe that the pitch could be matched eventually by fitting the opening angle and flagella length but do not actually do it.

Second, changing directions of bundles is also likely able to produce many different types of helical trajectories with varying pitch and radii. These have not been eliminated as possibilities.

Third, the fast reorientation in the simulations is interesting, but again, it is not ruled out whether other configurations could also yield similarly fast reorientations.

Fourth, the speed of the swimming is used as supporting evidence, but this is achieved by increasing the torque on the motor (to a value somewhat higher than some report for E. coli, but not out of the realm of possibility). If one is allowed to adjust the motor torque, then any speed can be reached for any configuration.

Reviewer #2:

The manuscript reports results of a detailed experimental and numerical study of the swimming motion of magnetotactic cocci, bilophotrichous bacteria with flagellar bundles at both poles. The experiments show that bacteria swim along a double helical path, with a very high swim speed and short reorient time compared to other bacteria. Hydrodynamic modelling demonstrates that this is due to a pushing and a pulling bundle. Experimental and numerical results are in good semi-quantitative agreement.

I think this is a very nice investigation, which carefully elucidates the complex swimming motion of a bilophotrichous bacterium. Experiment and hydrodynamic simulation complement each other very well to characterize the geometry of the flagellar organization and swim pattern. Thus, I strongly support publication of this manuscript in eLife.

I have only a few comments, which the authors should consider before publication:

1) In the main text, the authors talk about flagella as well as flagella bundles. I found this pretty confusing, until I saw Figure S1. I recommend clarifying this point early in the main text.

2) The reorientation angle could be discussed in a bit more detail. As far as I understand, as one bundle changes its direction of rotation, both bundles are in pushing (or pulling) mode, which leads to a (roughly) 90° reorientation of the swimming direction. This would explain the particular value of the reorientation angle. This seems to be somewhat similar to the behavior seen in simulations of the early stages of swimming and bundle formation of peritrichous bacteria, when the flagella are initially pointing in arbitrary directions, see, J. Hu et al., Sci. Rep. 5, 9586 (2015). However, the bundle rotation direction has to change back. This would result in another 90° angle, but uncorrelated with the first. Please clarify.

Reviewer #3:

The authors present a very thorough study of the exceptional swimming capacities in terms of speed and possibilities to suddenly change direction, displayed by a magnetotactic bacterium strain (MC-1). They use a holographic method to track in 3D the swimming trajectories and fast camera observations of the rotating body to visualize the positions of the flagellar bundles hooked to the cell. They discover the presence of a double helical path characterizing the swimming kinematics that can be explained by two sets of pushing and pulling flagella bundles working cooperatively and positioned within the same hemisphere of the spherical body. They simulate a hydrodynamic model to corroborate their finding, yielding very reasonable quantitative agreement with the experiments. The model, when set to some limits also helped to discover a new swimming pattern under the application of a magnetic field. Importantly, they also find that to reach such performances, the flagella must assume motive torque and bundle rigidity significantly larger than what is usually obtained for other bacterial strains.

I found the experimental study excellent with important conclusions on an original propulsion model. My opinion is that the paper could deserve publication in eLife. My only concern is sometimes on the pedagogy and clarity of the explanations delivered. I believe this could be significantly be improved. I do not deny that the authors really tried to make an effort to deliver their message, in particular visually. However, it remains that sometimes, it took me several subsequent readings of the same paragraph to really understand what they actually meant. Figure 3 is a particular example of that, and I found in this instance, very hard to follow the reasoning in the caption in association with the visual message. Then I would suggest that the paper could significant gain in pedagogy and impact, if it was critically read by someone who could help to clarify some explanatory sentences.

Reviewer #1:

[…]

1) In the dark field imaging, bright spots on the cell body are identified with flagella. This identification is not obvious to me, and it is quite puzzling why the authors do not actually image the flagellar bundles. If the bundles were indeed extended from the body in front and behind the cell, I would expect that they could be visualized which would definitively prove their proposed configuration. In Son, Guasto and Stocker, 2013 sheathed flagella are quite readily imaged. In my experience, there may be difficulty seeing flagella with darkfield at 1640 fps, but they should be readily imaged at <200 fps using their camera. Note that 200 fps is more than fast enough to resolve the larger helical motion with period 72 ms. It is unclear if the authors attempted this. If the authors tried and were not able to see a leading and lagging bundle, I would take that as evidence against their claim.

We agree on the fact that a clear image of the flagella would be the easiest way to proof our claims. We agree that we should clarify that we were not able to directly observe the full flagella bundles in motion. This, however, has proven to be challenging even for larger cells that additionally move one order of magnitude slower than the here analyzed species (e.g. low signal-to-noise ratio in Son, Guasto and Stocker, 2013).

We wrote:

“The flagella bundle morphology and movements were imaged in transmission electron microscopy (TEM) and in high-intensity dark-field video microscopy (Figure 3, Video 1 and Video 2). […] This is the result of the combination of the small size of the cells and their extraordinary high swimming speed and possibly their strong flagella bundle movement. “(Result section, fourth paragraph)

To explain the flagella images in an understandable way, we introduced a new Figure 3, focusing on convincingly identifying the tracked spots on the cell surface as parts of flagella bundles.

We thereafter wrote:

“Despite of the difficulties in imaging the flagella bundles in full length, the position of the flagella bundle near the cell surface were tracked together with the cell’s trajectory over 85 ms, which corresponds to 1.6 periods on the large helical trajectory of the cell (Figure 4). […] Additionally, the bright spots’ movement pattern featured the same periodicity as the large helical swimming track of the cell (also compare Figure 3C and Figure 3D).” (Result section, fifth paragraph)

And later:

“We turned to numerical simulations of the cell’s swimming behavior, to develop a deeper understanding of the mechanisms of propulsion and rapid reorientation and to compensate the missing information gathered from flagella bundle imaging.” (Result section, sixth paragraph)

2) The authors describe the double helical path as "unexplored" in the Introduction, but based on my knowledge, double helical paths are what should be expected from most types of propulsion by flagellar bundles, with the main distinguishing feature being the size of the larger helix. The smaller helix has in the past been attributed to the rotation of the bundle or flagellar helix, see Keller and Rubinow, 1976. The larger helix arises anytime there is non-axisymmetric propulsion, see Hyon, Powers, Stocker, Fu 2012 and has been observed in many species, albeit with varying sizes of the larger helix. As discussed in point 3 below, an important implication of this is that the bar for numerical simulations cannot be simply qualitative as in "straight" vs. "double-helical," but must involve some quantitative analysis of the trajectory pitch and radius.

Thank you for the clarification. We rewrote the corresponding section (not describing double helical paths as “unexplored”).

We edited the sentence:

“Here, we confirm that MC-1 cells reach speeds of over 500 μm s-1 (Figure 1D) and observe that the cells travel along a double helical path, which has not been reported for bilophotrichous cells so far.” (Introduction, final paragraph)

Towards the quantitative analysis of trajectory and pitch: We have greatly extended the subsection “Discussion of pitch differences between simulation and experimental observation”.

3) The numerics have not been described in a way that clearly rules out alternative hypotheses.

As reviewer 1 correctly pointed out, we should discuss all possible varieties of different asymmetries in flagella bundle configuration in the paper. Five possible configurations are: A difference in flagella length, a difference in motor strength, an asymmetry in the equilibrium angle of the two flagella relative to the cell surface, a time-dependent movement of the flagella equilibrium angles relative to the cell surface and the initially proposed asymmetry in sense of rotation (‘pusher-puller’ hypothesis). Finally, the best match between experiments and simulations has been found for our initially claimed ‘pusher-puller’ hypothesis.

We changed the introduction to the simulation paragraph:

“A large helical path of a microswimmer is produced from an off-axis (relative to the swimming direction) torque which continuously changes the direction of the thrust force. […] Although not fully reaching experimentally observed helix pitches, the asymmetry in sense of motor rotation was the only scenario producing significant matches in helix diameter and cell speeds.” (Result section, sixth paragraph)

Further, we added the mentioned scenarios to the text:

“1-Asymmetry in senses of rotation of motors: This asymmetry is described in the original version of the submitted manuscript. […] So, the results show that asymmetry in flagella direction cannot generate a helix comparable to MC-1’s large helix.” (Subsection “Test of different asymmetries in flagella configurations”)

First, as mentioned above, the quantitative details of the helical trajectory are important, but these are not well-matched. The authors say that their trajectories match the experimental helical diameter and period of the trajectories but not the pitch. The results are strongly dependent on the angle between the bundles, which is used as a fitting parameter. They say that they believe that the pitch could be matched eventually by fitting the opening angle and flagella length but do not actually do it.

We further investigated the parameter space within the pusher-puller scenario and extended the section. More precisely, we varied the opening angle, the flagella length and the motor torque and list the resulting track parameters in the subsection “Discussion of pitch differences between simulation and experimental observation”:

“While the pitch of the simulation result, presented in the main text, is off by a factor of 2, the diameter and period time of helical trajectories are in good agreement with the experiments. […] It can be concluded that the motor torque can be one of the parameters that may help matching between experiment and simulation, although the definitive match was not reached in this study.”

Second, changing directions of bundles is also likely able to produce many different types of helical trajectories with varying pitch and radii. These have not been eliminated as possibilities.

In the previous simulations, we fixed the opening angle for each simulation but the direction of the flagella axes are free to be determined by the dynamics of the system. Therefore, we did not set this parameter manually and the physical model implies the direction of flagella axes based on the physical conditions such as hydrodynamic interaction and other forces in the system. According to the reviewer’s suggestion, we also did additional simulations where the direction of flagella was changed. We observe a very weak second helix in the swimming trajectories with diameter of about an order of magnitude smaller than the experimental values. The weak effect of flagella direction may originate from the flagella’s flexibility. Due to flagella’s flexibility, the two flagella approach each other at their end and the swimming of the bacteria is not very different from the symmetric case. As mentioned above, it is the dynamic of the system that determines the shape and orientation of the flagella. So, the results show that asymmetry in flagella direction cannot generate a helix comparable to the MC-1 large helix.

We described this point in the new version of the manuscript, as described in the new paragraph on different flagella bundle scenarios, as already presented.

Third, the fast reorientation in the simulations is interesting, but again, it is not ruled out whether other configurations could also yield similarly fast reorientations.

We edited the main text and added the section “Transient flagella configurations that produce fast reorientation events”. We found that the originally suggested scenario produces best fits between experiments and simulations.

We wrote in the main text:

“Our model suggests a pushing flagella bundle together with a pulling flagella bundle rotating in opposite senses. […] A transient bending-like deformation was observed in the simulated flagella bundles during an event, which made the fast reorientations possible.” (Results section, ninth paragraph)

We wrote in the text:

“Three possible transient flagella configurations for reorientation events were tested in the simulations (CCW meaning counter-clockwise and CW meaning clockwise).

[…] Based on these results, we concluded that the reorientation angle statics produced by CCW&CW → CCW&CCW → CCW&CW matched the best to the experiment result.” (Subsection “Transient flagella configurations that produce fast reorientation events”)

Fourth, the speed of the swimming is used as supporting evidence, but this is achieved by increasing the torque on the motor (to a value somewhat higher than some report for E coli, but not out of the realm of possibility). If one is allowed to adjust the motor torque, then any speed can be reached for any configuration.

Each MC-1 sheathed bundle of flagella contains several flagella. Since we modeled a sheathed bundle of flagella by a single flagellum, we considered its effective torque to be several times larger than a single motor of E. coli. Unfortunately, there is no reported experimental value for the motor torque of MC-1. Therefore, since we have a bundle of flagella, we tested high torques and tune it in a way that the simulated velocity fit to the experimental values.

We described more clearly in the manuscript the above reasons for why we used higher motor torques for MC-1 than a single motor of E. coli. We wrote:

“The parameters for bending rigidity, torsion rigidity and torque of the flagella bundles were adjusted to reproduce the observed movement characteristics (number of small helices per large helix and velocities). […] Only this assumption allowed for stable and high swimming velocities in the simulations, indicating that the function of the flagella bundle is to combine high torques with high rigidities.” (Results section, seventh paragraph)

Reviewer #2:

[…]

1) In the main text, the authors talk about flagella as well as flagella bundles. I found this pretty confusing, until I saw Figure S1. I recommend clarifying this point early in the main text.

This was indeed confusing. We changed all “flagella” for “flagella bundle” and moved Figure S1 to Figure 3A. Thank you for the hint.

2) The reorientation angle could be discussed in a bit more detail. As far as I understand, as one bundle changes its direction of rotation, both bundles are in pushing (or pulling) mode, which leads to a (roughly) 90° reorientation of the swimming direction. This would explain the particular value of the reorientation angle. This seems to be somewhat similar to the behavior seen in simulations of the early stages of swimming and bundle formation of peritrichous bacteria, when the flagella are initially pointing in arbitrary directions, see, J. Hu et al., Sci. Rep. 5, 9586 (2015). However, the bundle rotation direction has to change back. This would result in another 90° angle, but uncorrelated with the first. Please clarify.

The observations showed reorientation times shorter than 5 ms. In simulation, we showed that one MC-1 reorientation event can be produced by a three-step deformation of flagella (pusher&puller → pusher&pusher → pusher&puller). The reorientation angle is measured by finding the angle between the first and third steps (pusher&puller steps) irrespective of the trajectory of the middle transient pusher&pusher step. The reviewer’s reasoning for having two uncorrelated 90° reorientation is true if we give the middle step enough time such that the flagellum can find a stable configuration but since this middle step is very short, it is not possible to measure a definite angle between the first and second or the second and third steps.

We clarified the definition of reorientation angle and how we calculated it in the manuscript. We edited the following paragraph:

“Our model suggests a pushing flagella bundle together with a pulling flagella bundle rotating in opposite senses. […] A transient bending-like deformation was observed in the simulated flagella bundles during an event, which caused the fast reorientations.” (Results section, ninth paragraph)

Reviewer #3:

[…]

I found the experimental study excellent with important conclusions on an original propulsion model. My opinion is that the paper could deserve publication in eLife. My only concern is sometimes on the pedagogy and clarity of the explanations delivered. I believe this could be significantly be improved. I do not deny that the authors really tried to make an effort to deliver their message, in particular visually. However, it remains that sometimes, it took me several subsequent readings of the same paragraph to really understand what they actually meant. Figure 3 is a particular example of that, and I found in this instance, very hard to follow the reasoning in the caption in association with the visual message. Then I would suggest that the paper could significant gain in pedagogy and impact, if it was critically read by someone who could help to clarify some explanatory sentences.

We corrected the English in many cases and clarified the flagella imaging part. Since Figure 3 was difficult to follow, we added a new figure (now Figure 3), as discussed in the answers to reviewer 1.

We also changed the legend of the former Figure 3 (now Figure 4) to:

“Figure 4. Positions of flagella bundles on the cell over 85 ms at 1424 fps. (A) The cell swam from top to bottom in the field of view. Moving bright spots were identified as the parts of the flagella bundles that were closes to the cell surface. The cell’s helical traveling path had a period time of 52 ms. (B) The positions were tracked over time and are depicted relative to the center of the cell, represented by different spheres for different time intervals. (C) The schematic of the cell’s large helical track is color-coded according to the different time intervals where the flagella bundle parts were visible to highlight the periodicity of the flagella bundle movement.”

Key resources table

Reagent type (species) or resource	Designation	Source or reference	Identifiers	Additional information
Cell line (Magnetococcus marinus)	MC-1	CEA	NCBI:txid156889	Dr. Christopher Lefèvre, CNRS
Software, algorithm	ImageJ	NIH	RRID:SCR_003070
Software, algorithm	MatLab	The MathWorks	RRID:SCR_001622