To dissect the mechanism of FSM, we directly observed IFT, FMG1-B and gliding motilities using single-molecule imaging techniques. We monitored the movement of individual IFT trains by using total internal reflection fluorescence (TIRF) illumination to image paralyzed-flagella (pf
) mutant cells that had adhered to the glass surface with both flagella. To establish a link between FMG1-B transport and IFT, we simultaneously tracked the movement of FMG1-B antibody-coated fluorescent beads (200 nm diameter, dark red) and IFT27-GFP in the pf18
strain. The beads performed short processive runs with reversals of direction whereas IFT trains moved in a regular and unidirectional manner. Multicolor kymography analysis shows that the beads transiently dissociated from one IFT train, diffused for a period of time and then bound to another IFT train (, Video 1
) similar to the movement of extraflagellar particles observed along the flagellar membrane (Dentler, 2005
). Diffraction-limited images of beads and IFT trains were fitted to a two-dimensional Gaussian to achieve a higher localization precision (Yildiz et al., 2003
). shows the colocalization of a membrane-attached bead with an individual IFT train as it moves unidirectionally for >500 nm. The movements of the bead and the colocalized IFT trains correlate strongly (>0.99) with each other during the processive run (N = 30, see for more examples), excluding the possibility that microspheres coincidentally moved together with IFT trains. We measured the distance between microspheres and the nearest IFT trains moving in the same direction for both the experimental data and randomly generated kymographs. The Kolmogorov-Smirnov test (p=7 × 10−9
, ) demonstrates that IFT trains colocalize with FMG1-B along the flagellar membrane ().
We next investigated whether IFT plays a direct role in gliding motility by monitoring individual IFT trains in IFT20-GFP and IFT27-GFP pf18 cells as they glided on coverslips. A small fraction of IFT trains displayed rare pauses (0.125 s−1 per cell, Ncells = 23, Npauses = 247) as they moved either in the retrograde or anterograde direction. Remarkably, the pausing of IFT trains was required for the initiation of whole-cell movements (, ). We did not observe the start of a gliding event without a paused IFT train. We were able to determine the directionality of 60% of the paused IFT trains, and all of the trains correspondent to the initiation of gliding motility stopped moving during retrograde transport (N = 148). To rule out the possibility that IFT pausing events and initiation of gliding motility are simply coincidental, we quantified the lag time between the pausing of the last retrograde IFT train and the initiation of gliding motility. The average lag time was 0.48 ± 0.05 s. In comparison, we observed individual IFT pausing events (including both anterograde and retrograde pauses) every 8.25 ± 0.96 s per cell (). The Student’s t-test excludes the null hypothesis that the timing of retrograde IFT pausing and gliding motility are independent of each other (p<0.0001). Anterograde trains also displayed rare pauses (~0.02 s−1 per cell, N = 27), but we never observed pausing of an anterograde train before the initiation of gliding motility.
In 24% of all cases (N = 50), a single paused train appeared sufficient to pull the entire cell (, Video 2
). In other cases, multiple trains paused before the start of gliding motility (, Video 2
). Pauses occurred in the leading flagellum, and cells moved in the opposite direction relative to the transport trajectory of the paused trains. In uniflagellate IFT27-GFP pf18
cells, we also always observed pausing of single or multiple IFT trains before the initiation of gliding motility (). Gliding motility either stopped when paused IFT trains resumed movement (), or continued until the cell bodies reached the paused trains (). We never observed the cell body gliding further than the position of the paused IFT train(s).
We found compelling evidence that dynein-1b is the primary motor responsible for gliding motility. First, uniflagellate pf18
cells always glided with the flagellum leading the cell body at 1.49 ± 0.10 µm/s (mean ± SEM) (, Video 3
) (Bloodgood, 2009
), suggesting that gliding motility is driven by pulling forces generated through a minus end directed microtubule motor. In contrast to pf18
cells, uniflagellate dynein-1b ts
) (Engel et al., 2012
) were unable to display robust gliding at restrictive temperatures (, Video 3
). We observed only short-range (<500 nm) gliding-like motion in 54% of these cells, compared to robust unidirectional gliding motility observed over 5 µm in pf18
cells. These short-range motions were bidirectional () and significantly slower (forward: 0.35 ± 0.09 µm/s, backward: 0.47 ± 0.16 µm/s) than the gliding speed of pf18
cells (). Second, inactivation of dynein-1b in biflagellate dhc1b-3ts
cells decreased the fraction of gliding cells from 100% to 52% and reduced the gliding speed from 0.86 ± 0.08 µm/s to 0.18 ± 0.03 µm/s. In contrast, inactivation of kinesin-2 in fla10ts
cells resulted in a twofold increase in gliding speed (), indicating that kinesin plays an inhibitory role during dynein-1b driven gliding motility. These results agree with our observations that the pausing of anterograde trains does not initiate gliding motility.
To further test the role of dynein-1b in gliding motility, we treated the cells with a small molecule inhibitor of dynein, ciliobrevin D (Firestone et al., 2012
). We varied ciliobrevin D concentration between 0–150 µM and monitored IFT in cells adhered both of their flagella to surface 2 min after drug treatment (, , Video 4
). Both anterograde and retrograde IFT train frequencies dropped with increasing concentrations of ciliobrevin D (), and at >100 µM ciliobrevin D we observed accumulation of IFT trains at the flagellar tip (see example kymograph in ). At 150 µM ciliobrevin D, retrograde IFT frequency was reduced by 92% compared to the 70% decrease observed after 6 hr of heat inactivation of dhc1b-3ts
cells (Engel et al., 2012
). The velocities of retrograde and anterograde trains also decreased by 60% and 36%, respectively (). Importantly, inhibition of dynein-1b resulted in a significant reduction in the speed (79%) and frequency (79%) of gliding motility (N = 50, ). These results further support our conclusion that dynein-1b motors produce the force for gliding motility.
Ciliobrevin D inhibits dynein-1b and stops gliding motility.
Based on these results, we propose a model to describe the functions of IFT motors, IFT trains and FMG1-B transport in gliding motility (). Surface adhesion of the FMG1-B cargo through its large extracellular carbohydrate domain (Bloodgood, 2009
) stops the retrograde IFT train. Dynein motors previously engaged in transporting the paused IFT train exert force towards the microtubule minus end, causing the whole cell to move toward the plus-end flagellar tip. Thus, gliding motility works similarly to microtubule gliding assays, in which surface-immobilized dyneins glide microtubules with their plus-end tips in the lead.
To investigate how cells reverse gliding direction, we developed a kymography method for monitoring IFT trains as the cells reorient their flagella during gliding (, Video 5
). During 60% of reversals (N = 40), the cell lifted one of its flagella and the paused IFT trains on the surface-adhered flagellum drove the motility (Bloodgood, 2009
). In other cases, single or multiple paused IFT trains accumulated in one flagellum, and the cell body glided toward this cluster until the paused trains either detached from the glass surface or reached the flagellar base. We also observed cases with paused IFT trains in both flagella where the cell remained immotile, likely due to the balance of forces between bound dynein-1b motors (). There were no indications of coordination of the pausing events between the two flagella. Different modes of reversals in gliding motility may allow cells to search through the environment by a random walk when they adhere both of their flagella, and to move in unidirectional manner by lifting one of the flagella.
Mechanisms of reversal in gliding motility.
Gliding motility in C. reinhardtii
is controlled by a Ca2+
-calmodulin regulated kinase and requires micromolar levels of Ca2+
in the media (Bloodgood and Spano, 2002
). We tested whether the pausing of IFT trains depends on Ca2+
() by analyzing the kymographs of IFT at different Ca2+
concentrations. To quantify pausing in flagella, we used Fourier space direction analysis (FSDA, see ) to separate the traces of paused IFT trains and moving IFT trains into two kymographs. Kymographs of paused IFT trains in show that pausing along the length of the flagellum was significantly reduced in Ca2+
-deprived cells (see additional examples in ). After Ca2+
depletion, IFT trains rarely paused in the middle regions of flagella and accumulated at the base (, middle). plots the total fluorescence intensity of paused IFT trains (Ncells
= 30 for each case) along the lengths of flagella at different Ca2+
levels. The frequency of pausing in Ca2+
-deprived cells was significantly lower than IFT pausing in cells at a normal Ca2+
concentration. In regular TAP media (free [Ca2+
] = 0.34 mM), 95% of all surface-adhered cells displayed gliding motility and individual pausing events were observed every ~8 s per flagellum, on average. In contrast, both the fraction of gliding cells and IFT pausing frequency were reduced by ~10-fold at <1 µM Ca2+
(). Next, we compared the IFT pausing frequencies in gliding and non-gliding cells at different Ca2+
levels. The pausing frequency in non-gliding cells was low and independent of Ca2+
concentration. In contrast, the pausing frequency in gliding cells was high at normal Ca2+
levels and gradually declined to match that of non-gliding cells as the Ca2+
concentration decreased below 1 nM ().
Ca2+ is required for the pausing of IFT trains at flagellar adhesion sites.
-deprived cells, the beads freely diffused on the flagellar membrane but did not move processively by IFT trains (not shown). Thus, we propose that Ca2+
is required for attachment of FMG1-B to IFT trains, not for the activation of a motor protein that provides force for gliding, as previously suggested (Bloodgood, 2009
). As the flagellar membrane is enriched with a PKD2-like Ca2+
channel (Pazour et al., 2005
; Huang et al., 2007
signaling at flagellar adhesion sites may play a role in controlling the directionality and timing of gliding movement.
To measure the forces exerted by motors bound to individual IFT trains, we tracked the movement of FMG1-B-antibody coated bead motility using an optical trap. To rule out the possibility that loads exerted by the trap might disrupt the linkage between IFT and FMG1-B, we performed TIRF imaging of IFT27-GFP and optical trapping of beads simultaneously (). We observed that IFT trains remained colocalized with trapped beads when subjected to external forces (; see for additional examples). The average offset between the positions of the trapped beads and colocalized IFT trains was 280 ± 10 nm (N = 11). This is due to the fact that the beads (920 nm diameter) and IFT trains (200–1000 nm in length) (Pigino et al., 2009
) are large objects relative to the resolution of conventional fluorescence imaging (approximately 200–250 nm) and the forces that stretch the bead-FMG1-B-IFT linkage displace the center of the bead from the IFT train (see the schematics in ). The trap measurements correspond to forces generated by IFT motors and provide direct evidence that IFT transports FMG1-B. Since beads are outside the cell, but are physically linked to the action of IFT trains, the assay combines the advantages of precise in vitro trapping with the ability to manipulate IFT motility under load.
Stall force measurements on single IFT trains.
In a fixed trap assay, ~50% of processive bead movements terminated with a stall before returning to the trap center (, ). We did not observe a significant difference between the force values of stall and release events. This suggests that motors attached to an IFT train may not be able to reach their maximum stall force, defined as the stall force of a single motor multiplied by the number of bound motors (Shubeita et al., 2008
). Histograms of peak forces () show that anterograde and retrograde IFT trains moved against 21.4 ± 0.7 pN and 25.2 ± 1.3 pN (SEM), respectively. Forces exerted on IFT trains well exceed the force-generation capability of a single motor. Previous studies demonstrated that multiple motors produce larger forces, move with higher velocities under load and carry cargos further than a single kinesin or dynein motor (Mallik et al., 2005
; Vershinin et al., 2007
). It remains controversial whether motor stall forces are additive at low motor copy numbers (Vershinin et al., 2007
) or whether multiple motors tend to transport their cargo using only one load-bearing motor at a time (Jamison et al., 2010
). Therefore, we believe that peak forces in our experiment represent a lower boundary for estimates of motor copy number. Assuming that IFT motors produce 6–7 pN forces (Gennerich et al., 2007
; Brunnbauer et al., 2010
), we estimate that at least four motors transport IFT trains at a time, in agreement with previously reported values (Engel et al., 2009b
). It is possible that multiple motor engagements enhance the run length of individual IFT trains, allowing them to traverse the length of a flagellum. In addition, the fast transport of IFT trains in a viscous cellular environment and IFT’s role in FSM may require forces larger than single motor stall forces.
The measured forces in both directions only marginally changed (<20%) under different bead antibody-coating conditions (. This result argues against the possibility that antibody-coating of the beads leads to the crosslinking of more than one IFT trains, which would be expected to multiply the average peak force. 20% of retrograde runs escaped the trap by producing more than 80 pN of force, presumably via clustering of FMG1-B (Bloodgood, 2009
), while <1% escape events were observed in the anterograde direction.
We next investigated why the speed of gliding motility (1.49 ± 0.10 µm/s, SEM) is significantly slower than that of retrograde IFT (2.96 ± 0.14 µm/s), even though both processes are powered by dynein-1b motors actively transporting IFT trains. IFT trains that carry 1-μm beads moved ~30% slower than IFT trains that were not associated with the beads (two-tailed Student’s t-test, p = 2.1 × 10−9 (anterograde) and 1.8 × 10−11 (retrograde), ). We reasoned that the viscous drag of the membrane may slow down IFT trains transporting FMG1-B clusters and led to the observed differences in speed between IFT and FSM. To estimate the drag constant of the bead-IFT complex, we analyzed individual stalling events in the optical trap assay () and measured the recoil time of the bead after a stall (). The average drag constant of the bead-IFT complex was found to be 8.0 ± 2.8 pN s/μm (SEM, N = 30).
Viscous drag of the membrane slows down the motility of IFT trains that carry beads.
We hypothesized that the large drag constants we measured were due to the interaction between the bead and the membrane. Our results rule out the possibility that microtubule motors step backwards during the recoiling of the bead, as the average bead velocity was 20 µm/s (an order of magnitude faster than that of dynein and kinesin) and there were no detectable backward steps. To rule out other possible interactions between IFT trains and axonemes, we oscillated individual beads on a flagellar membrane surface in a square wave pattern and measured the recoiling time when the beads are decoupled from IFT (). The average viscous drag constant was 8.4 ± 4.2 pN.s/µm, which is similar to the drag that beads experience when they interact with IFT trains. Anterograde and retrograde trains moving at 2–3 µm/s would experience 16–24 pN resistive forces, comparable to the total motor force exerted on a single IFT train (). Therefore, IFT is subjected to a high drag force when it carries a large bead along the flagellar surface, which leads to the reduction of transport velocity.
We next investigated how kinesin-2 activity reduces the speed of gliding motility (). This could theoretically be explained by a tug-of-war between active kinesin and dynein motors that are both present on the same IFT train. Alternatively, kinesin-2 and dynein-1b motors may be exclusively active on anterograde and retrograde cargos, respectively, and pausing of anterograde trains could produce forces that oppose the gliding forces of paused retrograde trains. To distinguish between tug-of-war and coordinated transport mechanisms, we examined how inhibition of one class of motors affected the forces exerted on IFT trains traveling in the opposite direction (Laib et al., 2009
). At permissive temperatures, the peak forces of IFT in fla10-1ts
were in close agreement with wild-type (WT) cells. Heat-inactivation of kinesin-2 reduced the frequency of anterograde runs by 66%, but did not alter the peak forces on retrograde runs (Laib et al., 2009
) (). Similarly, heat inactivation of dynein-1b significantly reduced the ratio of retrograde to anterograde transport events (0.15), but had minimal effect on the peak forces of anterograde runs (). These results exclude a tug-of-war mechanism in IFT, which would lead to an increase of force from motors walking in one direction upon inactivation of the motors walking in the opposite direction. Instead, only one type of a motor remains active on IFT trains at a time, which is consistent with the result that retrograde speed of fla10-1ts
does not change after kinesin-2 is inactivated. We propose that kinesin-2 on paused anterograde trains slows down gliding motility by exerting forces in the opposite direction to that of dynein-1b on paused retrograde trains.
Force measurements on temperature-sensitive mutants.