Load-induced enhancement of Dynein force production by LIS1–NudE in vivo and in vitro

Most sub-cellular cargos are transported along microtubules by kinesin and dynein molecular motors, but how transport is regulated is not well understood. It is unknown whether local control is possible, for example, by changes in specific cargo-associated motor behaviour to react to impediments. Here we discover that microtubule-associated lipid droplets (LDs) in COS1 cells respond to an optical trap with a remarkable enhancement in sustained force production. This effect is observed only for microtubule minus-end-moving LDs. It is specifically blocked by RNAi for the cytoplasmic dynein regulators LIS1 and NudE/L (Nde1/Ndel1), but not for the dynactin p150Glued subunit. It can be completely replicated using cell-free preparations of purified LDs, where duration of LD force production is more than doubled. These results identify a novel, intrinsic, cargo-associated mechanism for dynein-mediated force adaptation, which should markedly improve the ability of motor-driven cargoes to overcome subcellular obstacles.

traces) suggests that typical LD force production in the NudE/LIS1 knock-down backgrounds decreases after attempt 1, and overall appears less robust in these traces, compared to motion in controls (in Fig.1). (j) Typical minus end escape in P150 RNAi cells.   Step detection on single-motor kinesin (top) shows that 8 nm steps are detected when present; such steps are present for plusend moving LDs (second trace), and for minus-end traces (third through fifth). Very little change occurs between steps detected in M1 and M5 in the purified LDs.

Supplementary Note 1: Development of a new system to measure forces
One of the challenges in measuring forces in vivo using optical trap is calibration of the applied force. We employed two different techniques for this purpose and both were found suitable. The first, a method we previously applied, is based on in vitro calibrations using refractive index-matching approach 1 , measuring laser deflection with a QPD. The second is a newer method 2-4 , detecting momentum-changes in the laser beam at back focal plane of the condenser using a PSD (position sensitive detector) to directly measure the beam deflections. It thus measures force directly (Fig. 1b and Fig. 3a & 3b, Supplementary Fig. S2a-2j,   Supplementary Fig. S6d-S6f), and does not need the determination of LD position relative to the trap center, in contrast to the index-matching approach (Supplementary Fig. S6a-S6c). Importantly, because it measures momentum changes of the scattered laser light, Method-2's determination of applied force is insensitive to cargo size and shape, or exact position of the trapped object relative to the trap center. In our measurements both approaches yielded similar results (the average forces measured for LDs are similar, Supplementary Fig.   S6f). Note that in either case, because some of the measurements span over 50 seconds, implementation of an independently running focus lock system in our setup eliminated slow drift in the microscope stage.
We use a Nikon TE200 inverted DIC microscope; on which two optical traps were assembled using single mode diode laser (980 nm, 700 mW from EM4 Inc.) and an 830 nm, 1 Watt, Ti-Sapphire laser (Coherent, Verdi-5). An automated xy-piezo-mirror (Madcity labs) is placed in the back focal plane of the objective to steer the optical trap in the field of view by the application of DC voltage (0-10 V, two independent DAC outputs from 16 bit resolution NI card for X and Y motion, capable of corrections down to 1 nm). The calibration factor for steering the trap using mirror was estimated each time by identifying LD positions inside the cell at minimum and maximum voltages. The setup is designed in such a way that with a mouse click in the vicinity of the moving LD in the DIC video, the droplet is first moved (via motorized stage, Ludl Electronics) close to the trap center, and then the trap is positioned with more accuracy using the real time template matching and XYpiezo-mirror. We restricted the piezo-mirror based steering distance of the trap to within few microns from the center of field of view by using linearly moving XY-stage. The range of trap motion using piezo-mirror was calibrated by trapping a freely diffusing LD in the cell each time the coverslip was changed. Analog voltage was incremented in steps of 2 volts from 0-10 V, using 16 bit D/A outputs of NI card. Signal was simultaneously applied to X and Y channels and LD positions recorded using template matching to obtain the scaling factor. This combination greatly reduced the uncertainty in co-localizing the centers of LD and trap. More importantly, decreased the effective time needed to apply the trapping force after identifying the LD center using real time template matching (Note that we employed LD template matching twice; first, to identify the droplet when it is far away from the trap and second, immediately after the LD is brought to the center of field of view and just before applying the trap). This combination also provides the ability to automatically scan the sample over a wide range and at the same time achieve high accuracy (~10nm) for the trap-droplet relative position. High resolution position information were obtained from either a QPD (Method-1) or a PSD (Method-2, momentum transfer), see Methods. Autofocus lock system was built using an 850nm diode laser (in TIRF mode), a QPD and a piezo z-stage (from PI) on which the sample is mounted.

Supplementary Note 2. Identification of Minus and Plus End Droplets
In principle, there could be local changes to/uncertainty in microtubule orientation. However, we were careful to avoid such complications, because we only analyzed droplets whose motion was approximately perpendicular to the cell periphery, and occurring in locations where the majority of the vesicle traffic was linearly outward from the cell center. Certainly the vast majority of those droplets moving 'out' (and that thus fit this criteria) are going to be moving plus-ends, and the vast majority of those moving 'in' are towards minus-ends. Ours is a statistical argument-and most of time significantly correct with regard to orientation; an occasional incorrectly identified droplet will not drastically alter our conclusions. There is a large body of work amassed over many years that indicates that in non-epithelial cells, the minus-ends are close to the nucleus, and the plus-ends are close to the periphery. This is also supported by our EB1-GFP and MT imaging (see supplementary movies 11 & 12). Further, the differences between our plus-end data and minus-end data unambiguously show that we can appropriately identify each direction of travel. For instance, pooling all the plus-end forces, and all of the minus-end forces, we did a Kolmogorov-Smirnov test, to compare the distributions, and found that the hypothesis that the two distributions are the same can be rejected with a p value smaller than 0.0000001.
Thus, given the extensive work from others in the field establishing that in normal cells such as COS1 the microtubule plus-ends are oriented towards the periphery, combined with the dramatic differences we see in plus-end vs. minus-end motion, there seems little probability that we are in fact unable to differentiate between the two. Critically, our in vivo data is consistent with our in vitro studies along polarity-marked microtubules, where there is absolutely no ambiguity as far as directionality of transport: in vitro as well as in vivo, force adaptation occurs by altering minus-end force persistence.

Supplementary Note 3. Criteria for scoring attempts and Escapes.
A successful escape was easily determined, as the LD mostly walked out of the trap. A failed escape attempt was scored by considering both video tracking and high resolution PSD data (2 kHz). The criteria used for scoring a failed attempt is that LD position must be distinctly away from baseline, and that clear detachments should occur, resulting in the LD quickly falling to within 20% of the maximum displacement from the baseline (Note that in majority of the events the detachments brought the LD to within 10-15% of peak displacement in ~ 0.05 sec, See Fig. 1a & 1b, Supplementary Fig. S2a-2j). We had no difficulty in observing the clean detachments in case of LDs possibly because they are not membrane bound. We hypothesize that this may not be true in membrane bound vesicles as the membrane can in principle act as an elastic tether thus making it difficult to observe clean detachments. In these measurements the clean detachments could be easily identified by the abrupt change or discontinuity in the slope of the LD track both in the PSD and video data.
Errors in escaped fraction (f) were determined as √ for n droplets tested.
Supplementary Note 4: On the apparent increase in escape probability on P2 in the P150 siRNA background. While an increase in plus-end escape probability was observed for the second plus-end escape in the P150 siRNA background (see Fig. 2f), we view this as a statistical fluke. We base this on two observations. First, in additional repeated experiments with P150 RNAi cells, the escape probability for the second plus attempt was not observed to increase, although the minus end escape probabilities went up as found earlier.
(44 LDs tested in the repeat experiment). The results were combined, which brought down the pvalue (to 0.06), but we acknowledge it is still rather large. Second, in addition to the escape probability measurements, we also carried out high-force measurements. In these more quantitative measurements, in the P150 RNAi background, there was no increase in P2 maximum force or duration of force production (see Fig. 3b, 3d), again consistent with the hypothesis that the increased P2 escape seen in Fig. 2f was a statistical fluke. Errors in escaped fractions in each attempt (f) were estimated with √ for n droplets that made the escape attempt.

Supplementary Note 5. Labile Microtubules are unlikely to explain the Adaptation
In the peripheral region where we make measurements we did not detect obvious microtubule motion (see supplementary movie 12). We can eliminate buildup of sustained microtubule deformations as key contributors to adaptation, because of the long (10-second) periods between attempts, where no force is required to keep the droplet positioned in the trap (see e.g. Fig. 1a & 1b, Supplementary Fig. S2a-S2b, S2e-S2j); during such periods any putative microtubule deformations would relax. Further, we also see the time dependent systematic increase in the force persistence of motor complexes with purified LDs in vitro, in the absence of labile microtubules. Thus, adaptation results from changes in the function of the machinery on the cargos themselves, rather than motion of the microtubules. Ultimately, the machinery driving minus-end directed lipiddroplet transport adapts to opposition to motion, and this adaptation requires the combined use of Dynactin, LIS1, and NudE & NudEL. Such a conclusion is consistent with the in vitro experiments which still have force adaptation, though they occur along static taxol-stabilized microtubules stuck to glass coverslips.

Supplementary Note 6. Step sizes of purified LDs under load
Since dynein can take different step sizes 5 , and dynactin can change dynein's step-size distribution 6 , we wondered what happened to step size distributions during adaptation. We first tested our analysis on polystyrene beads driven by single kinesin-1 motors (K-560, Supplementary Fig. S4h, top), confirming that 8 nm steps were detected as expected. Then, we examined stepping behavior of plus-end moving LDs in vitro ( Supplementary Fig. S4h, second). While forward motion was similar, with predominantly 8 nm steps detected, more back-steps were detected. With this calibration done, we examined minus-end motion. On, average step size distributions (Supplementary Fig. S4h) for attempts M1 (n=14), M4 (n=13) and M5 (n=13) were similar, and were centered around 8nm with perhaps a hint of a small peak at 16nm. We found no significant change in step size distributions during the course of minus end adaptation. While the data do not directly support the hypothesis that adaptation involves alteration of dynein step sizes, we should note that the cargo motion here