|Home | About | Journals | Submit | Contact Us | Français|
Mutations in the tail of the cytoplasmic dynein molecule have been reported to cause neurodegenerative disease in mice. The mutant mouse strain Legs at Odd Angles (Loa) exhibits impaired retrograde axonal transport, but the molecular deficiencies in the mutant dynein molecule, and how they contribute to neurodegeneration, are unknown. To address these questions, we purified wild-type and mutant mouse dynein. Using biochemical, single molecule, and live cell imaging techniques, we find a strong inhibition of motor run-length in vitro and in vivo, and significantly altered motor domain coordination in the mutant dynein. These results suggest a potential role for the dynein tail in motor function, and provide the first direct evidence for a link between single-motor processivity and disease.
Cytoplasmic dynein is a minus-end directed microtubule motor protein responsible for a variety of cellular functions, including retrograde axonal transport1. Two dynein mutant mouse strains, Legs at Odd Angles (Loa) and Cramping1 (Cra1), were identified in a screen for genes involved in late onset motor neuron disease (MND)2. This finding identified a new class of MND genes, which also includes the mouse and human p150Glued gene encoding a subunit of the dynein regulatory complex dynactin3–5, and significantly expanded the limited pool of familial Amyotrophic Lateral Sclerosis (ALS) candidate genes. Of the dynein mutations, Loa has received particular attention6–10. Loa/+ mice were initially reported to exhibit lower motor neuron degeneration, but more recent studies have found severe loss of sensory neurons8,9. Loa mutant mice also exhibited a decreased rate of retrograde axonal transport2,10. The Loa mutation resides in the extreme N-terminal region of the dynein heavy chain (HC) polypeptide (F580Y), within the dynein "tail." This region is responsible for organizing the multiple dynein subunits into a complex and for binding to membranous cargo. Dynein generates force through its two motor domains, each located at the C-terminal end of the HC, 1,500 a.a. (15–20 nm) away from the site of the Loa mutation. Whether and how the Loa mutation affects cytoplasmic dynein function has remained untested.
To approach this problem, we first tested the ability of Loa mutant dynein to remain associated with membranous vesicles isolated by sucrose step gradient flotation from wild-type, Loa/+ and Loa/Loa brains, but found no clear difference (Fig. 1a). The Loa mutation lies in the region of the dynein HC involved in HC-HC dimerization and in HC-intermediate chain (IC) binding (Supplementary Information, Fig. S1). We therefore tested for potential defects in the stability of the mutant complex. Fractionation of whole brain cytosol by sucrose density gradient centrifugation revealed a single major 20S dynein peak for both the wild-type and Loa/+ mutant animals. Loa/Loa dynein, in contrast, showed a small but reproducible decrease in dynein s-value, accompanied by the appearance of a free IC peak (12 ± 2 % of total ICs) at 6S (Supplementary Information, Fig. S2a). To test whether these observations reflect a more general reduction in mutant dynein stability, we exposed brain extracts to potassium iodide (KI), a chaotropic salt to which dynein is particularly sensitive11. In the presence of KI, dynein dissociation increased both as a function of KI concentration and of the proportion of mutant dynein HC (Supplementary Information, Fig. S2b–d). These results indicate that the Loa mutation may impair the interactions between subunits.
To gain further insight into the molecular effects of the Loa mutation, we purified cytoplasmic dynein from wild-type and Loa/+ adult mouse brains1. As we observed in brain cytosol, the purified mutant dynein remained intact by sucrose density gradient centrifugation (data not shown). The purified Loa/+ dynein showed no differences in subunit composition from the wild-type mouse protein, and neither preparation showed detectable levels of the processivity factor dynactin by Coomassie Blue staining or immunoblotting (Fig. 1b; Supplementary Information, Fig. S3a–b). Because the homozygous mutant mice die shortly after birth, the amounts of brain tissue we could obtain were inadequate to purify biochemical amounts of dynein by this procedure. Nonetheless, we were still able to purify small amounts of Loa/Loa dynein for single molecule analysis using a modification of this method (Fig. 2–3, Supplementary Information, Fig. S3c–d).
To test the effects of the Loa mutation on dynein mechanochemical activity, we measured the ATPase activity of wild-type and Loa/+ dynein in the presence and absence of microtubules. The basal ATPase activity was similar for both wild-type and Loa/+ dynein: 51.3 ± 5.3 and 51.7 ± 3.3 nmole Pi/min/mg, respectively (Fig. 1c). However, in the presence of microtubules, the ATPase activities of wild-type and mutant dynein differed markedly. The Michaelis constant for microtubules, KmMT, was 1.5 ± 0.2 µM for the purified wild-type dynein, similar to the value for bovine cytoplasmic dynein12, whereas the KmMT for Loa/+ dynein was considerably higher, 11.8 ± 3.4 µM (P<0.001). In contrast, the maximum rate (Vmax) values were very similar (wild-type = 255 ± 9 nmol/min/mg; Loa/+ = 263 ± 43 nmol/min/mg). The latter results indicate that the maximal ATP turnover rate is unaffected by the Loa mutation. The increased KmMT, however, suggests a pronounced decrease in the effective affinity of the mutant dynein for microtubules during ATP hydrolysis. In support of this possibility, the KmMT for Loa/+ was partially rescued at reduced ionic strength (4.0 ± 0.6 µM vs. 1.2 ± 0.2 µM for wild-type), which increases dynein-microtubule affinity12. Furthermore, microtubule binding by the purified mutant dynein was markedly reduced relative to wild-type dynein in the presence of ATP, though not in its absence (apo state) (Fig. 1d). This surprising result suggests that much of the dynein fraction normally seen to cosediment with microtubules in the presence of ATP is engaged in active, processive movement along the microtubule. These results together suggest an effect of the Loa mutation on the interaction of dynein with microtubules during the ATPase and force-generating portion of the crossbridge cycle, though not in the strong microtubule binding (apo) state.
To gain insight into the underlying molecular defects, we used quantum dot and optical trap assays under single molecule conditions. We attached dynein to quantum dots using antibodies to the tail of the wild-type and mutant molecules and analyzed their velocity and run-length. Movements in the two dynein preparations were ATP-dependent and at a velocity similar to that reported for mouse dynein-dynactin complexes (Fig. 2a; Supplementary Information, Fig. S4a)13. The purified wild-type and mutant dyneins showed predominantly unidirectional movements, but some bidirectional events were also observed, as previously reported (Fig. 2c)13–15. A clear difference, however, was observed in the wild-type vs. mutant dynein run-lengths (Fig. 2b–c). Wild-type dynein exhibited an average run-length of 339 ± 33 nm (Fig. 2c), about half that of mouse dynein-dynactin complexes13, a reflection of the absence of dynactin from our dynein preparations. Loa/+ dynein displayed fewer long runs and a 23 % shorter average run-length (259 ± 9 nm). Loa/Loa dynein was even more impaired, with an average run-length approximately half that of wild-type dynein (175 ± 4 nm) (Fig. 2c). Optical trap experiments using the same dynein preparations adsorbed to polystyrene beads gave comparable results, with no difference in average velocity between Loa/+ and wild-type dynein, but with a similar reduction in Loa/+ single motor processivity (Fig. 3a; Supplementary Information, Fig. S4b–c). Based on the motility data, we calculate a 31–37 % increase in the average off-rate for Loa/+ vs. wild-type dynein at 0.5 – 1 mM ATP, respectively (see Supplementary Information), consistent with the enhanced dissociation of Loa/+ dynein from microtubules in the presence of ATP (Fig. 1). As for bovine cytoplasmic dynein, the stall force for single wild-type mouse dynein molecules was 1.4 ± 0.3 pN, and within experimental error of those measured for Loa/+ (1.7 ± 0.4 pN) and Loa/Loa dynein (1.6 ± 0.4 pN) (Fig. 3b)14,16. We also observed no difference in the step size of the wild-type or mutant dynein motors while they moved along microtubules under load (Fig. 3c; Supplementary Information, Fig. S5a–c).
Physiological cargoes typically use multiple motors. To assess how the observed defects in dynein processivity translate to the in vivo condition, we tested the run-lengths for multiple wild-type and Loa mutant dyneins in vitro and found an increase in run-length for multi-motor conditions (as reported in 17); however, this increase was attenuated for the Loa/+ dynein (Fig. 3d). We also carried out computational modeling of dynein run-lengths at more extensive motor/cargo ratios, and found the mean travel defect to persist to three or more dynein molecules/cargo, with and without assuming potential effects of in vivo factors such as dynactin13,18,19 (Fig. 4b; Supplementary Information, Fig. S6a–b). We find here that an additional consequence of decreased single molecule processivity is a reduction in the number of instantaneously engaged motors per cargo in the multiple motor range, which in turn further limits cargo travel (Fig. S6b). These simulations reveal a previously unappreciated sensitivity of multiple-motor run-lengths to changes in single-motor processivity, especially apparent here where the single-motor processivity is low in comparison to bovine dynein.
The Loa mutation has been linked to defects in retrograde axonal transport2. To compare the effects of the Loa mutation in vivo with our in vitro data, we performed live cell imaging analysis of lysosome/late endosome behavior at high temporal resolution (5 frames/sec; Fig. 4a–c). We then used custom analysis software to precisely identify periods of uninterrupted motion (“runs”, see Supplemental Information). We imaged the far distal region of the axon (>100 um from cell body), where we observed predominantly unidirectional, retrograde runs as previously described20. Retrograde run-lengths were reduced by 53 % and 83 % in Loa/+ and Loa/Loa neurons, respectively (Fig. 4b; Supplementary Information, Fig. S6c). The wild-type in vivo run-length (5.27 ± 0.34 µm) allowed us to use our theoretical model to estimate an average of 7.7 dynein molecules per cargo (Fig. 4b), similar to recent values based on stall force and biochemical isolation21,22. We then predicted the expected mutant run-lengths at the same 7.7 dyneins/cargo, adjusting only the single-motor processivity to reflect the measured in vitro processivity defects. The predicted run-lengths were in excellent agreement with those measured in vivo (2.49 ± 0.13 vs. 2.43 ± 0.21 µm for Loa/+; 0.89 ± 0.05 vs. 0.86 ± 0.07 µm for Loa/Loa, Fig. 4b). As in our in vitro experiments (Fig. 2–3), we see no change in instantaneous lysosome velocity in the Loa mutant neurons (Fig. 4c, left). However, the overall average velocity was reduced by 22 % in Loa/+ and 43 % in Loa/Loa neurons vs. wild-type (Fig. 4c, right), consistent with an increase in run terminations, and comparable to theoretical prediction (11 % decrease in average velocity for Loa/+ and 37 % for Loa/Loa, see Supplemental Information). Together, these results argue strongly that the Loa dynein processivity defect we identify in vitro can account for the observed impairment in retrograde axonal transport.
To gain further insight into the mechanism responsible for altered Loa dynein processivity, we tested for the reported tendency of cytoplasmic dynein to step laterally on the microtubule surface, in contrast to the strict linear travel of kinesin15,23. We confirmed lateral stepping in mouse dynein, and observed a significant increase in its frequency for the Loa/+ mutant protein (Fig. 5a–b). One potential explanation for this result is a disrupted coordination between the two motor domains within the native dynein complex. A gating mechanism between dimeric motor domains is well established for kinesins and myosins, but is not well understood for dynein24–26. Gating contributes to processive motion by ensuring that, as one motor domain advances, the other remains strongly associated with its track in the apo state, thereby preventing premature detachment26. As a test for altered gating in the Loa mutant dynein, we measured its KmATP. Despite the somewhat decreased Vmax for Loa/+ dynein due to its lower KmMT (Fig. 1c), the KmATP was clearly decreased (18.8 ± 4.1 µM vs. 27.0 ± 6.5 µM for wild-type, P<0.05) (Fig. 5c). This result is consistent with a defect in communication between motor domains, allowing premature ATP binding by the apo motor domain, and subsequent release from the microtubule (Fig. 5d), though further kinetic analysis will be needed to confirm this model.
These results have novel implications for intramolecular regulation of cytoplasmic dynein motor activity. Despite the location of the Loa mutation within the dynein tail, we find several lines of evidence for an altered interaction between dynein and microtubules, the first direct indication for communication between the dynein motor and tail domains. The clear defect we observed in mutant dynein processivity raised the possibility of motor domain miscoordination. Supporting this, we identified novel defects in the Loa mutant dynein: both an increase in lateral stepping on the microtubule lattice, and an increased affinity for ATP apparently leading to premature run termination. Consistent with the subtly decreased stability of the Loa dynein complex, we propose that an abnormal linkage within the base of the Loa dynein molecule disrupts coordination between the two motor domains by altering their relative positioning and ability to interact laterally.
Our results also suggest a novel role for processivity defects in disease causation. We argue that neurodegeneration in the Loa/+ mutant mouse is unlikely to result from dynein subunit dissociation, which we do not detect for the Loa/+ complex in cytosolic extracts or following purification. (Only dynein from the Loa/Loa mouse, which dies shortly after birth, shows evidence of dissociation: Supplementary Information, Fig. S2). Our data also argue against a loss in the association of dynein with membranous cargo, which persisted in our biochemical analysis. The most dramatic change we observed to result from the Loa mutation was a decrease in dynein run-length. This effect quantitatively accounted for the observed defect in axonal transport as indicated by a combination of empirical analysis with computational modeling. The neurons that are most likely to be affected by this defect in processivity and impairment in transport would be those with the longest axons, such as motor and sensory neurons. Our study therefore provides the first evidence that altered motor protein processivity can have pronounced consequences for in vivo transport, the impairment of which clearly correlates with neuronal death and disease.
Cytoplasmic dynein was purified from wild-type and Loa/+ adult brains by microtubule-affinity, ATP release, and sucrose density gradient fractionation as previously described1. For purification of dynein from postnatal day zero (P0) brains, cytosol from three P0 brains was subjected to sucrose density gradient fractionation, microtubule-affinity in the presence of GTP (to remove kinesin), then ATP extraction to release dynein (Fig. S3c–d).
Whole brain cytosol from P0 mice was fractionated on a 5–20 % Tris-KCl sucrose gradient (20 mM Tris-HCl, pH 7.6, 50 mM KCl, 5 mM MgSO4, 0.5 mM EDTA), or was incubated with 0.075 M, 0.15 M, or 0.30 M potassium iodide (KI) on ice for 1 hour, then fractionated on a sucrose gradient containing 0.075 M, 0.15 M, or 0.30 M KI. Fractions were analyzed by SDS-PAGE and western blot probing for dynein heavy chain (HC)27 (1:1000) and dynein intermediate chain (IC) (1:3000; clone 74.1, K. Pfister, University of Virginia). A solution of thyroglobulin, ferritin, catalase, lactate dehydrogenase, and albumin was fractionated on a sucrose gradient to determine S-values for the gradient fractions.
ATPase assays were performed using the malachite green method as previously described12 in Tris-KCl buffer containing 10 mM KCl or 50 mM KCl, and incubated at 37°C for 15 minutes in the presence of 1 mM ATP. For ATPase activity as a function of microtubules, taxol-stabilized microtubules (Cytoskeleton, Inc., Denver CO) were added to a solution containing 1 µg of dynein before adding ATP to start the reaction. For ATPase activity as a function of ATP, all reactions were performed in Tris-KCl buffer containing 10 mM KCl and in the presence of 10 µM microtubules. Activities were plotted at the range of microtubule or ATP concentrations. Michaelis-Menten curves were fit to the data to derive Km and Vmax values.
For microtubule binding experiments, 1 µg of dynein was incubated with a 2.5 µM final concentration of taxol stabilized microtubules at 37°C for 15 minutes, then centrifuged. Supernatant and microtubule pellet were either recovered for analysis, or the pellet was resuspended in Tris-KCl buffer containing 10 mM ATP, incubated at 37°C for 15 minutes, and centrifuged. Supernatants and pellets were analyzed by SDS-PAGE and western blot probing for dynein intermediate chain and tubulin (1:5000; Sigma).
Membrane flotation analysis was performed as previously described10. Briefly, two P0 brains from each genotype were homogenized and centrifuged at 30,000 g. The supernatant was centrifuged at 150,000 g. The high-speed pellet was fractionated by flotation through a 2M, 1.5M, and 0.6M sucrose step gradient. Membrane vesicles floated to the 0.6–1.5M interface. Samples were taken from the 0.6M layer, the vesicle (V) interface, and the 1.5M layer for SDS-PAGE and western blot analysis, probing for IC, HC, and synaptotagmin (1:500; StressGen) as a marker for vesicles.
In vitro motility assays were conducted in flow chambers assembled from a glass slide and acid washed cover slip using double-sided adhesive tape (chamber volume ~10 µL). Solutions were incubated in the chamber for 10 minutes. The chamber was incubated with 5 mg/mL Biotin-BSA, washed 2 × 20 µL with blocking buffer (30 mM HEPES pH 7.2, 50 mM KAcetate, 2 mM MgAcetate, 1 mM EGTA, 10 % glycerol, 1 mg/ml BSA, 1 mg/ml casein), incubated with 5 mg/mL streptavidin solution, washed twice, then incubated with rhodamine- and biotin-labeled microtubules and washed twice again with blocking buffer with 5 µM taxol.
Monoclonal anti-74.1 IC antibody (350 nM), polyclonal anti-HC antibody (450 nM), or polyclonal anti-R1B2 antibody (450 nM) was mixed with goat anti-mouse or goat anti-rabbit quantum dots (QD) (350 nM) (Invitrogen) in blocking buffer for 30 minutes on ice. Different antibodies gave similar results. Dynein (10 nM) was added to the antibody-QD mixture and incubated on ice for 30 minutes. Similar results were obtained at 1:50 protein-to-QD ratios. The dynein-QD mixture was diluted 50 times immediately prior to chamber incubation, then washed twice. Blocking buffer containing 500 µM ATP, 5 µM taxol, and an oxygen scavenging system was flowed into the chamber. Differences in run-lengths and velocities between optical trap and QD assays were due to differences in ATP concentrations. Single QD-labeled-dynein molecules were visualized at 25°C on an inverted microscope (DMIRBE, Leica) using a Qdot 525 filter set (Omega Optical, Brattleboro, Vermont). Images were captured on a CCD camera (CoolSnap HQ, Photometrics) at 5 frames/sec. The movement of individual dynein-QDs was analyzed using a custom-tracking program (Gross Lab) to identify the QD position vs. time via 2-d Gaussian fitting of their brightness profile. A processive run is defined here as motion >200 nm that is terminated by detachment from the microtubule. A bidirectional run is a >200 nm run in one direction, then a >200 nm run in the opposite direction. In the case of a bidirectional run, the run-length was measured as the sum of processive run-lengths in both directions.
Optical trap assays, data recording, particle tracking, and stalling force analysis was performed as previously described14,28–30. Dynein was attached directly to polystyrene beads (PolySciences, Inc.); using antibody as a linker impaired the motile behavior of beads. Bead-microtubule binding fraction was used as a read-out for the average number of available motors per bead: 50–100 % binding fractions correspond to the few motor condition, whereas ≤ 30 % binding fractions correspond to the single motor range14,31,32. Dynein-coated beads were positioned above a microtubule for 20–30 seconds using an optical trap. A binding event was scored upon bead binding and processive motion along microtubule. Motility and stalling force measurements were conducted using 500 nm polystyrene beads in assay buffer (10 mM PIPES, 2 mM MgSO4, 1 mM EGTA, 1 mM DTT, 50 mM K-Acetate, 10 % Glycerol, pH 6.9) supplemented with scavenging solution and 1 mM ATP immediately before measurements. An automated program monitored the bead position, then turned off the trap upon detection of processive bead motion. An individual run was defined as the course between a bead binding to, and then detaching from a microtubule. For stall force measurements, a trap stiffness of 2.2 pN/100 nm was used. A stall was scored if the bead proceeded away from trap center and held its plateau position for >300 ms before detachment.
Step size analysis was carried out on continuous segments of force traces where the systematic noise was no greater than ~8 nm, using an objective, model-independent step detection method,33 cross verified using a chi-square reduction method,34 and tested for kinesin stepping in vitro35.
Lateral position experiments were conducted using 200 nm beads23 in assay buffer with scavenging solution and 0.2 mM ATP. We examined dynein’s lateral wandering by quantifying the change in a bead’s lateral position between subsequent frames. Runs longer than 200 nm and 2 sec were analyzed. To limit possible bias toward longer travels, runs for each genotype were grouped into five different travel ranges (increment in 100 nm from 200 to >600 nm). The final distribution of lateral changes describes the mean ± SD from averaging each subgroup.
We adapted a recently developed Monte Carlo model, which closely reproduced force measurements for multiple dyneins in vitro16, constrained by our experimental measurements of single-motor processivity for wild-type, Loa/+, and Loa/Loa dyneins. A “run” is initiated when at least one of the N motors becomes stochastically bound to the microtubule (characterized by on-rate), and terminated when all N motors become detached (characterized by off-rate). For each motor, we used the same dynein stiffness (0.32 pN/nm), viscosity (10−3 N·sec/m2), on-rate estimation (1/sec), step size (8 nm), as well as formulas for stepping direction, velocity, and off-rate under load as described in16. Dynein’s off-rate under no load was the lone adjusted parameter in our simulation, and was assumed to represent the sole differentiating factor between each genotype. To estimate the appropriate off-rate (under no load) for each genotype in vivo, we evaluated the average run-length in vitro when only one motor was available for transport. To account for potential in vivo factors, we carried out simulations assuming a two-fold processivity enhancement of dynactin18,19, and a 0.1 pN constant opposing force to account for viscous drag in cytosol.
Hippocampal neurons were cultured from E19 mouse embryos as previously described and cultured for 4–6 DIV (days in vitro)36. Mice were cared for and treated according to IACUC regulations. For experiments, cultures were incubated in fresh media containing 500 µM lysotracker, washed, then imaged in 2 mL of fresh media (with 10 mM Hepes) at 37°C with 5 % CO2 on an Olympus IX81 inverted microscope. Images were captured at 5 frames/sec for 1.5 minutes using an Hamamatsu ORCA-R2 CCD camera. Experiments were performed at least three times per genotype. For velocity, the motions of single lysosomes were analyzed using kymographs generated with the “Multiple Kymograph” plug-in for ImageJ (EMBL, Heidelberg, Germany). For run-length, individual lysosomes were identified by a 2-d Gaussian fitting of their brightness profile, and custom analysis routines identified periods of un-interrupted motion (runs), where net travel for any three consecutive frames (400 ms) was at least 80 nm in the minus-end direction.
Travel distributions were fit to a single exponential decay to extract the decay constant (and fit uncertainty) as its characteristic run-length. Speed distributions were fit to a Gaussian function to extract the mean peak position (and fit uncertainty) as its average velocity. Stalling force distributions were Gaussian fitted to extract average values for single motor force production and fitted full width at half maximum were used to properly account for systematic noise. Statistical significance was determined using the Student’s t-test.
We thank Richard McKenney for help with the quantum dot assays and useful discussion and Michael Vershinnin for developing the tracking program. This work was supported by NIH grants GM47434 to RBV, RO1GM070676 to SPG and GM008798-09 to KMOM, AHA grant 825278F to JX, and the Columbia University Motor Neuron Center.
Author ContributionsKMOM, JX, SPG, and RBV designed the research. KMOM and JX performed experiments and analyzed data. KMOM, JX, SPG, and RBV wrote the paper.
Competing Interests Statement
The authors declare no competing financial interests.