|Home | About | Journals | Submit | Contact Us | Français|
The lissencephaly protein Lis1 has been reported to regulate the mechanical behavior of cytoplasmic dynein, the primary minus-end-directed microtubule motor. However, the regulatory mechanism remains poorly understood. Here, we address this issue using purified proteins from Saccharomyces cerevisiae and a combination of techniques, including single-molecule imaging and single-particle electron microscopy. We show that rather than binding to the main ATPase site within dynein's AAA+ ring or its microtubule-binding stalk directly, Lis1 engages the interface between these elements. Lis1 causes individual dynein motors to remain attached to microtubules for extended periods, even during cycles of ATP hydrolysis that would canonically induce detachment. Thus, Lis1 operates like a “clutch” that prevents dynein's ATPase domain from transmitting a detachment signal to its track-binding domain. We discuss how these findings provide a conserved mechanism for dynein functions in living cells that require prolonged microtubule attachments.
► Single-molecule and structural data reveal how Lis1 regulates dynein motility ► Lis1 binds between dynein's ATPase and microtubule-binding domains ► Lis1 alters communication between these elements, prolonging microtubule attachment ► Mutating an arginine finger motif in dynein mimics aspects of Lis1's effect
To help create the complex internal organization within eukaryotic cells, the myosin and kinesin motor protein families have undergone widespread gene duplication and evolution, presumably so they can fulfill particular functional niches (e.g., transporting cargos, bearing tension, and sliding filaments; Vale, 2003). Accordingly, their members show a spectrum of properties, with different catalytic rates and structural adaptations. A striking departure from this pattern is one of the least understood cytoskeletal motors, cytoplasmic dynein (Figure 1A). This one motor powers nearly all movement toward the microtubule minus end in most eukaryotic cells. Its many functions include transporting and positioning diverse cargos (e.g., mRNAs, proteins, and organelles) during interphase, exerting tension between the microtubule network and cell cortex during cell migration, and helping to construct the spindle during mitosis and meiosis (Karki and Holzbaur, 1999; Vale, 2003). Regulation of dynein is therefore critical, and human diseases arise from its dysfunction (Gerdes and Katsanis, 2005). In addition to task-specific regulators, cytoplasmic dynein has three ubiquitous cofactors that are required for most, if not all, of its functions across eukaryotes (Kardon and Vale, 2009; Vallee et al., 2012): the dynactin complex and two proteins analyzed in this study, Lis1 and Nudel.
Dynein's ATP-hydrolyzing and track-binding elements are key potential targets for regulation and differ in three significant ways from those of myosin and kinesin. First, rather than a compact fold, dynein's ATPase domain is ring shaped, containing six covalently linked AAA+ modules (Neuwald et al., 1999; Roberts et al., 2009). Second, rather than a single site of ATP binding/hydrolysis, four of dynein's AAA+ modules (AAA1–AAA4) can bind nucleotide, with AAA1 being the main site of ATP hydrolysis (Cho et al., 2008; Gibbons et al., 1991; Kon et al., 2004, 2012; Schmidt et al., 2012). Finally, instead of close integration of the track-binding and hydrolysis sites, these domains in dynein are spatially separated by a ~10 nm intramolecular coiled-coil stalk, which protrudes from AAA4 (Cho and Vale, 2012).
Together, dynein's architectural features make interdomain communication a central part of its mechanism and an additional potential target for regulation. For example, structural changes within dynein's AAA+ ring, driven by ATP binding and hydrolysis at AAA1, produce cyclic changes in the affinity of the microtubule-binding domain at the stalk's tip (Imamula et al., 2007; Kon et al., 2004, 2009). In some distantly related AAA+ ATPases, pulses of structural changes are transmitted around the ATPase ring with the assistance of arginine finger motifs that reach from one subunit into the active site of the neighboring subunit (Erzberger and Berger, 2006). However, the extent to which dynein uses similar mechanisms is unknown.
A strong candidate for altering the structural changes within dynein is the evolutionarily conserved protein Lis1. Mutations in the human LIS1 gene cause classical lissencephaly, a severe brain development disorder (Reiner et al., 1993). In brain slices and cultured cells, perturbing Lis1 levels causes defects in dynein-mediated processes such as nuclear migration, mitosis, and cargo transport (Ding et al., 2009; Faulkner et al., 2000; Pandey and Smith, 2011; Smith et al., 2000; Tai et al., 2002; Tanaka et al., 2004; Tsai et al., 2007; Yi et al., 2011). Like cytoplasmic dynein's force-generating subunit, Lis1 is dimeric. Each Lis1 protomer contains an N-terminal dimerization domain and a C-terminal β-propeller domain that binds to the dynein motor domain (Figure 1A; Sasaki et al., 2000; Tai et al., 2002; Tarricone et al., 2004). A coiled-coil protein, Nudel (or its paralogue, NudE), is thought to further tether Lis1 to dynein (McKenney et al., 2010), allowing Lis1 to function at lower concentrations in vivo (Efimov, 2003; Li et al., 2005; Wang and Zheng, 2011; Żyłkiewicz et al., 2011). In Saccharomyces cerevisiae, the Lis1 and Nudel orthologs (Pac1 and Ndl1, respectively) serve to concentrate dynein at the plus ends of microtubules (Lee et al., 2003; Li et al., 2005; Sheeman et al., 2003). This is a critical step in targeting dynein to the cell membrane, where it positions the nucleus during cell division.
A range of properties has been ascribed to Lis1, including slowing dynein-driven microtubule sliding in multiple motor assays, prolonging dynein stall events under load, and enhancing the microtubule affinity of dynein when ADP and Vi (a phosphate analog) are bound (McKenney et al., 2010; Torisawa et al., 2011; Yamada et al., 2008). However, Lis1's mechanism of action remains largely mysterious, in part because of the lack of a recombinant system in which both Lis1 and dynein can be manipulated and studied at the single-molecule level. Although reports have suggested that Lis1 binds the AAA1 module of dynein (Sasaki et al., 2000; Tai et al., 2002), this has not been tested with functional proteins, and structural information on the dynein/Lis1 complex is not yet available. Moreover, reconciling the reported in vitro effects of Lis1 and Nudel with specific dynein activities in vivo has been controversial (Allan, 2011).
Here, using functional, recombinant proteins from S. cerevisiae, we show that Lis1 engages dynein at AAA3/AAA4. From this site, Lis1 acts like a “clutch” to regulate communication between dynein's catalytic ring and microtubule-binding stalk, promoting a microtubule-bound state. Supporting these results, we identify mutations at the AAA3/4 junction that drastically impair Lis1 binding and motility regulation in vitro and dynein function in vivo. In addition, we identify an arginine finger motif within AAA4 and find that its mutation mimics aspects of Lis1's effect in vitro. Previous genetic studies in an evolutionarily distant filamentous fungus showed that the same mutation can partially rescue Lis1 loss in vivo (Zhuang et al., 2007). These results allow us to propose how Lis1 biases dynein to a microtubule attached-state and assists in a variety of cellular functions across eukaryotes.
To dissect how Lis1 and Nudel regulate dynein's motility, we wanted a system in which all three components could be manipulated and studied at the single-molecule level (Figure 1A). We previously developed methods to purify native cytoplasmic dynein complexes from S. cerevisiae, fluorescently label them with tetramethylrhodamine (TMR; via a C-terminal HaloTag), and visualize their movement along immobilized microtubules (Reck-Peterson et al., 2006). Here, we purified Lis1 and Nudel (Figure 1B), which were expressed from their genomic loci in S. cerevisiae. We inserted cleavable tags for affinity purification and a C-terminal SNAP-tag (giving the option of covalent labeling). The 20-kDa SNAP-tag did not interfere with the function of either protein, as revealed in nuclear segregation assays (Figure 1C). We eliminated the risk of endogenous regulators (Lis1, Nudel, and dynactin) from copurifying with dynein via genomic deletions (Table S1 available online).
We began by observing dynein motility at the single-molecule level over a range of Lis1 concentrations. In the absence of Lis1, dynein bound to and traveled along microtubules with a mean velocity of ~100 nm/s (Figure 1D), similar to previous reports (Reck-Peterson et al., 2006). As the Lis1 concentration increased, two trends were apparent. First, dynein velocity became progressively slower, reaching ~7 nm/s at the highest Lis1 concentration tested (1,200 nM). This can be appreciated in the kymographs in Figure 1D, in which steeper lines indicate slower movement. Second, Lis1 increased the length of time that dynein remained attached to the microtubule. For example, at 1,200 nM Lis1, many dynein molecules were attached for the entire duration of a 10 min movie (seen in Figure 1D as lines that span the kymograph's y axis; see Figures S1A–S1D for quantification). Control assays with kinesin showed that the effect of Lis1 is specific to dynein (Figures S1E and S1F). Thus we conclude that Lis1 can convert dynein to a mechanochemical state in which its velocity is slowed and its microtubule attachments are prolonged.
Notably, at a given Lis1 concentration, the population of dynein molecules responded similarly: velocity histograms remained unimodal as the mean decreased (Figure 1F). We plotted dynein's velocity as a function of Lis1 concentration. The plot is well fitted by a hyperbolic curve with a maximal velocity reduction of 95.7 ± 3.1% (Figure 1G; Table S2A). These data show that saturating concentrations of Lis1 can almost entirely arrest dynein on the microtubule. The Lis1 concentration giving half-maximal velocity reduction (K1/2) in this assay is 60.1 ± 10.0 nM. Hereafter, we refer to dynein in the presence of near-saturating Lis1 concentrations (>5-fold higher than the K1/2) as dynein/Lis1.
We next tested whether Nudel, Lis1 and dynein's binding partner, affects the behavior of single dynein molecules in vitro. At 37.5 nM, Nudel alone had little effect on dynein motility (Figures S1G and S1H). By contrast, a mixture of 37.5 nM Nudel and 37.5 nM Lis1 elicited a change in dynein behavior akin to that induced by a ~5-fold higher Lis1 concentration. The mean velocity was slowed to ~29 nm/s, and microtubule encounters were markedly prolonged (Figures 1E and andS1H).S1H). These results show interesting similarities and differences compared with earlier work. In previous in vitro experiments, mammalian Nudel was found to dissociate dynein from microtubules and suppress Lis1's effect on the velocity of unloaded dynein (McKenney et al., 2010; Yamada et al., 2008). In contrast, we find that Nudel from S. cerevisiae allows Lis1 to exert its effect on dynein at lower concentrations even in unloaded conditions. This matches observations from previous experiments in living fungi and frog egg extracts, in which increasing Lis1's concentration rescued Nudel deletion or depletion, respectively (Efimov, 2003; Li et al., 2005; Wang and Zheng, 2011; Żyłkiewicz et al., 2011).
To test which structural domains are required for Lis1 to regulate dynein motility, we analyzed the motility of a well-characterized dimeric motor domain construct (GST-dynein331 kDa; Reck-Peterson et al., 2006). GST-dynein331 kDa, which lacks the N-terminal tail, endogenous dimerization motif, and binding sites for associated dynein subunits, contains two motor domains dimerized by GST (Figure 2A) and displays processive motility similar to that of native dynein. In the absence of Lis1, GST-dynein331 kDa traveled along microtubules with a mean velocity of ~116 nm/s (Figure 2B). Strikingly, Lis1 also reduced the velocity and prolonged microtubule attachments of GST-dynein331 kDa in a dose-dependent manner (Figures 2B and andS2A–S2D;S2A–S2D; Table S2B). The maximal velocity reduction was similar to that observed in native dynein (Figure 2C), but the K1/2 of Lis1 is lower (~15 nM). This difference is consistent with the notion that the affinity of Lis1 for dynein increases when dynein's tail domain is removed (Markus et al., 2009; Reck-Peterson et al., 2006). These results indicate that Lis1 regulates dynein motility by acting on elements within the motor domain.
We next tested whether dimerization of either Lis1 or dynein is an essential part of the regulatory mechanism, using purified monomeric dynein331 kDa, whose motor activity can be analyzed using a microtubule gliding assay (Figure 2D). To produce monomeric Lis1 (Lis1ΔN), we removed Lis1's N-terminal dimerization domain. The monomeric state of Lis1ΔN was verified by size-exclusion chromatography (Figure 2E) and crosslinking (Figure S2E). Monomeric dynein331 kDa drove microtubule gliding at a velocity of 55.5 nm/s (Figure 2F). Significantly, in the presence of 800 nM Lis1ΔN, gliding velocity slowed by ~60% (to 21.5 nm/s) and decreased by ~70% (to 16.2 nm/s) when the Lis1ΔN concentration was increased to 1,600 nM. These data indicate that a single β-propeller domain of Lis1 can regulate the activity of a single dynein motor domain, provided it is supplied at an elevated concentration (perhaps explaining why similar effects were not observed in a previous study; Torisawa et al., 2011).
We next tested these findings in living cells. Yeast expressing monomeric Lis1ΔN as the sole source of Lis1 exhibited a strong nuclear segregation defect (~22%; Figure 2G), consistent with malfunction of the dynein pathway. However, when the cellular concentration of Lis1ΔN was increased >10-fold with the use of a strong, galactose-inducible promoter (Figure S2G), the percentage of cells with a nuclear segregation defect was rescued to ~10% (Figure 2G). This suggests that a monomeric Lis1 construct comprising the β-propeller domain can regulate dynein both in vitro and in vivo, provided its concentration is sufficiently high. Thus, dimerization serves to enhance the apparent affinity between dynein and Lis1.
In principle, dynein could be converted to a microtubule-attached state by arresting its ATPase cycle at a stage corresponding to strong microtubule affinity (e.g., with ADP or no nucleotide bound). Previous ATPase results obtained with brain-purified dynein in the presence of Lis1 have varied (McKenney et al., 2010; Mesngon et al., 2006; Yamada et al., 2008), perhaps due to heterogeneity between and within dynein preparations. To test for Lis1-induced catalytic arrest in our system, we measured the ATPase activity of dynein with different concentrations of Lis1 and microtubules. The basal ATPase rate of GST-dynein331 kDa alone was activated by microtubules to a maximal rate of 10.8 motor-domain−1 s−1 (Figure 3A). Addition of Lis1 also activated the basal ATPase rate (Figure 3B), indicating that Lis1 binding alters turnover in one or more of dynein's catalytic AAA+ modules. Interestingly, in the presence of microtubules, the maximal ATPase rate of GST-dynein331 kDa/Lis1 retained ~90% of the value for GST-dynein331 kDa alone (Figure 3B). This was the case even in extremely high Lis1 concentrations (2 μM; data not shown). Thus, in the presence of microtubules, dynein/Lis1 can continue to hydrolyze ATP, indicating that Lis1-mediated motility regulation is more complex than a simple arrest of the ATPase cycle.
In a canonical mechanochemical cycle, dynein's motor domain dissociates from the microtubule after ATP binding. In a dynein dimer, this allows one motor to detach and undergo a forward excursion along the microtubule while the partnering motor remains bound. To test whether Lis1 alters this mechanochemical coupling, we developed an assay to visualize the ATP-sensitive interactions of individual motor domains with microtubules (Figure 3C). Monomeric, fluorescently labeled dynein331 kDa molecules were bound to surface-immobilized microtubules in the absence of nucleotide, inducing strong “rigor” binding. The molecules appeared as stationary, diffraction-limited spots by total internal reflection fluorescence (TIRF) microscopy and as vertical lines in kymographs (Figure 3D). Addition of ATP caused monomeric dynein331 kDa molecules to rapidly dissociate from the microtubule (Figure 3E, arrow). Dissociation was not due to laminar flow or dilution, as the molecules remained attached after perfusion with buffer (Figure 3D, arrow). After ATP-induced dissociation, monomeric dynein331 kDa was free to diffuse away from the microtubule. Subsequent rebinding events were short and probably correspond to single turnovers of ATP (Figure 3E). The presence of Lis1 caused two main differences in these behaviors. First, instead of dissociating rapidly after ATP addition, dynein331 kDa remained bound to the microtubule for extended periods (average duration of ~6 s) before detaching (Figures 3F and andS3A).S3A). Second, the duration of rebinding events was similarly prolonged (Figure S3B). Thus, Lis1 slows the microtubule off-rate of dynein in ATP conditions. Multiple ATP hydrolysis cycles are likely to occur during these extended periods of microtubule attachment (see Figure 3B and data below). These results suggest a model in which Lis1 alters communication between dynein's catalytic AAA+ modules and microtubule-binding stalk, allowing the motor domain to remain tightly bound to the microtubule even in high ATP concentrations.
A prediction of this model is that Lis1 will cause the motor domains in a dynein dimer to remain attached to the microtubule for longer periods between steps, giving rise to slower motion. If this prediction is correct, then it might be possible to observe individual steps even at a physiological ATP concentration (~1 mM) without hindering load, conditions under which dynein's steps have previously been too rapid to detect. To test this, we attached a bright, photostable quantum dot (Qdot) to the C terminus of one of dynein's motor domains (Figure 4A; Reck-Peterson et al., 2006). Dynein's movement along microtubules could then be tracked with nanometer precision and a temporal resolution of 100 ms.
As expected, free GST-dynein331 kDa moved rapidly in 1 mM ATP, and high-precision traces showed smooth rather than stepwise motion, consistent with a stepping rate that matches or exceeds the temporal resolution (Figure 4B, green trace). In contrast, movement in the presence of Lis1 was slow and exhibited distinct steps, with pauses (“dwells”) intervening between abrupt forward and occasional backward displacements (Figure 4B, red traces). This Lis1-induced stepping behavior is reminiscent of free dynein in 1 μM ATP, in which the stepping rate is limited by the rate of ATP binding (Figure 4B, blue trace). The dynein step size distribution in the presence of 1 mM ATP and Lis1 (Figure S3C) was similar to that previously reported for dynein alone at low ATP concentration (Reck-Peterson et al., 2006). Occasional larger displacements in the Lis1 traces may correspond to “bursts” of multiple rapid steps at the rate observed for free dynein in 1 mM ATP (Figure 4B, green trace) and could reflect Lis1 dissociation events. In summary, we conclude that the slow movement of dynein/Lis1 is not due to smaller forward steps or more frequent back steps (Figure S3C); rather, it is accounted for by longer dwells between steps.
Most frequently, dynein/Lis1 paused for ~1 s between steps (Figure 4C), and longer dwells lasting up to 15 s can be seen in the traces (Figure 4B, asterisk). For comparison, the average duration of an ATP turnover in steady-state assays was ~0.1 s (Figure 3B). The dwells we observe between individual steps in the presence of Lis1 may be related to the pauses previously seen between longer movements in mammalian dynein (see Figure S3 of McKenney et al., 2010). During Lis1-induced pauses, the displacement fluctuations of the Qdot were equivalent to those seen for dynein alone while tightly bound to the microtubule waiting for ATP to bind (in each case, SD = ±4 nm). Together, these results show that Lis1 causes dynein to remain tightly attached to the microtubule for typically more than a second in high concentrations of ATP. This is a pronounced change from dynein's canonical stepping pattern.
We next explored the structural basis for the striking effects of Lis1 on dynein's mechanochemistry. Using size-exclusion chromatography, we were able to purify the complex of GST-dynein331 kDa and Lis1 in buffer supplemented with ATP + Vi, ATP, or no added nucleotide (Figures 5A and 5B). Complex formation was evident from the coelution of dynein and Lis1 (Figures 5A, blue arrow, and 5B) and depletion of free Lis1 (Figure 5A, gray arrow). This differs from a previous report in which interactions were detected in the presence of ATP + Vi (thought to trap the prepowerstroke state of dynein) but not in nucleotide-free conditions (McKenney et al., 2010). It may be that Lis1 affinity in the no-nucleotide state is stronger with dimeric, yeast dynein compared with the monomeric, mammalian dynein construct used in earlier work. We used electron microscopy (EM) to investigate structural differences between the free dynein and dynein/Lis1 fractions in the absence of nucleotide, allowing direct comparison with the yeast motor domain crystal structure (Schmidt et al., 2012) obtained in the same state.
Figure 5C shows negative-stain EM images of GST-dynein331 kDa both alone and bound to Lis1. Molecules from the two samples have the same general form, with the paired motor domains adopting a range of relative orientations about the GST moiety. However, the two motor domains are slightly closer and less variably spaced in the presence of Lis1 (Figure 5D), suggesting that Lis1 might bind between them (see Figure S4A).
We next mapped the Lis1 binding site on dynein using single-particle image processing. Images of individual motor domains were extracted from the dimer micrographs, aligned, classified into similar groups, and averaged, revealing a range of molecular views in the data sets. The two main views of the motor domain are close to those previously seen in other dyneins (Figures 5E and 5F; Burgess et al., 2003; Roberts et al., 2009). Detail was sufficient to match these views unambiguously to orientations of the yeast dynein crystal structure (Figures S4B and S4C). Both the coiled-coil stalk and the microtubule-binding domain at its tip are also resolved in class averages (Figures 5E and 5F, insets), further aiding interpretation of the images.
In the presence of Lis1, there is a pronounced extra density on the margin of dynein's motor domain, near the emergence of the stalk (Figure 5E, arrowhead). This observation is quantified in a difference map, where the additional density gives rise to a strong difference peak (Figure 5E, right panel). Analysis of the second main view (Figure 5F) yields a difference peak in the same position. This extra Lis1 density in both views overlaps with AAA3/4 (Figures 5E and 5F, lower; Figures S4B–S4E). We detect no significant density changes near AAA1 in the presence of Lis1, contrary to earlier reports that Lis1 binds AAA1 (Sasaki et al., 2000; Tai et al., 2002).
To determine whether Lis1 binding at AAA3/4 is responsible for regulating dynein motility, we mutated a series of highly conserved, charged amino acids in dynein at this putative interface (Figures 6A and andS5A).S5A). Whereas these mutations had little or no effect on dynein's intrinsic motility, Lis1-induced velocity reduction was impaired by a single point mutation (K2721E) and nearly abolished by a quadruple substitution (K2721A, D2725G, E2726S, and E2727G) near the AAA3/4 junction (Figure 6B). These four mutations also virtually abolished Lis1 binding in a gel filtration experiment (Figures S5B and S5C). Furthermore, in living yeast cells, the quadruple substitution produced a nuclear segregation phenotype indistinguishable from that caused by Lis1 deletion (Figure 6C). Thus, we conclude that Lis1 engages dynein at AAA3/4, near the interface between its catalytic ring and microtubule-binding stalk.
A pre-existing clue to the involvement of AAA3/4 in Lis1's mechanism comes from genetic studies in Aspergillus nidulans. In this filamentous fungus, a single Arg→Cys mutation within AAA4 can partially suppress the phenotype associated with deletion or mutation of the Lis1 gene (nudF) (Zhuang et al., 2007). On close inspection, this residue is a strong candidate to form the “arginine finger” motif of AAA4. By analogy to related ring-shaped AAA+ ATPases, this motif is expected to reach from AAA4 into the adjacent nucleotide-binding pocket of AAA3 and help transmit structural changes around the ring (Figure 6D; Erzberger and Berger, 2006; Kon et al., 2012; Schmidt et al., 2012). To investigate the role of AAA4's arginine finger and the consequences of the suppressor mutation, we engineered the equivalent mutation (R2911C) in S. cerevisiae dynein. In single-molecule assays, GST-dyneinR2911C retained the ability to bind to and move along microtubules. However, its mean velocity was slowed to ~5 nm/s, and its microtubule encounters were extended in time compared with the parental construct with an intact AAA4 arginine finger (Figure 6E). Strikingly, these properties of the mutant closely resemble dynein behavior in saturating Lis1 concentrations (Figure 6E). As expected, the Lis1Δ phenotype in S. cerevisiae is not rescued by the arginine finger mutation (Figure S5D), because here Lis1 is also required for proper dynein localization (Lee et al., 2003; Markus et al., 2009; Sheeman et al., 2003). By contrast, in A. nidulans, this mutation partially bypasses the need for Lis1, likely because dynein localization is less dependent on Lis1 in these cells (Zhang et al., 2002). In summary, mutating the arginine finger motif of AAA4 confers motile properties similar to those produced by Lis1; moreover, the same mutation can in part circumvent the requirement for Lis1 in A. nidulans (Zhuang et al., 2007).
Our single-molecule and ensemble data illuminate the mechanism by which Lis1 alters cytoplasmic dynein's mechanical behavior and assists in its wealth of cellular functions. Based on these findings, models for Lis1's mechanism must take the following observations into account. First, Lis1 converts single dynein molecules to a microtubule-attached state in a dose-dependent fashion, bringing dynein to a virtual standstill at saturating concentrations. Second, a monomeric Lis1 β-propeller domain, acting on a single dynein motor domain, can elicit these effects. Third, Lis1 alters dynein's mechanochemical coupling: with Lis1 bound, dynein's motor domain can remain attached to the microtubule in physiological ATP concentrations, even while undergoing multiple rounds of hydrolysis. This indicates that Lis1 can alter communication between dynein's ATP hydrolyzing ring and its microtubule-binding stalk. Fourth, Lis1 engages dynein close to the interface between these elements, at AAA3/4.
We bring these observations and previous results together into a model for Lis1's mechanism of action (Figure 7). We first consider a canonical (Lis1-free) dynein stepping cycle that begins with the binding of ATP at dynein's main hydrolytic site (AAA1; Figure 7A, blue). This initiates motions of AAA+ subdomains within the ring, which are translated into sliding between the α helices of the coiled-coil stalk. In turn, the affinity of the microtubule-binding domain at the stalk's tip is weakened; it rapidly detaches from the microtubule and then undergoes a diffusive search for a new binding site.
The main concept of our model is that when Lis1 binds via its β-propeller domain to dynein's ring at AAA3/4, the flow of structural changes from the ring to the stalk is severed (Figure 7B). This function of Lis1 is reminiscent of a “clutch” that alters transmission between dynein's AAA+ “engine” and its track-binding region. Thus, while cycles of ATP binding and hydrolysis can persist within the ring, the microtubule-binding domain at the tip of the stalk retains a strong-binding conformation. As a consequence, the “duty ratio” of the motor (i.e., the fraction of the catalytic cycle spent tightly attached to the track) approaches unity. From a kinetic standpoint, the microtubule off-rate of the motor domain following ATP binding is dramatically slowed, consistent with enhanced microtubule affinity in the ADP-Vi state (McKenney et al., 2010). Although the results here show that the dynein-Lis1 complex can be purified in a variety of nucleotide states, our data do not exclude nucleotide-dependent changes in dynein-Lis1 affinity (McKenney et al., 2010); indeed, we think changes in affinity are likely as dynein's ring structure flexes during the ATPase cycle. Interestingly, aspects of Lis1's effects (e.g., constitutively strong microtubule binding and activated ATPase) have been achieved by locking the helices of the stalk's coiled coil in a defined alignment by crosslinking (Kon et al., 2009). Collectively, these results suggest that Lis1 may affect the mobility of the helices in dynein's coiled-coil stalk using a unique allosteric mechanism involving binding near the stalk's base.
Our structural data, showing that Lis1 binds dynein at AAA3/4 near the interface between its ATP-hydrolyzing ring and microtubule-binding stalk, provide an explanation for how communication between these elements can be regulated. This is in contrast to an earlier report based on interactions between overexpressed fragments of each protein that suggested that Lis1 binds AAA1 (Sasaki et al., 2000; Tai et al., 2002). Our functional studies also support a role for AAA3/4 being the site of Lis1 interaction; mutation of four charged amino acids in AAA4's N-terminal helix severely impairs Lis1 binding, renders dynein nearly completely insensitive to Lis1 in motility assays, and phenocopies a Lis1 null allele in living cells. Furthermore, these charged amino acids are highly conserved in available cytoplasmic dynein sequences (Figure S5A), suggesting that binding at AAA3/4 is likely to be a common feature of Lis1-dynein interactions. A notable exception to this pattern of conservation is found in Schizosaccharomyces pombe, which lacks an obvious Lis1 ortholog. The AAA3/4 binding site also provides a simple explanation for why the Lis1-dynein affinity is sensitive to the “neck” region of dynein's tail (Markus and Lee, 2011; Markus et al., 2009), as these domains are closely apposed (Burgess et al., 2003; Roberts et al., 2012).
Although the ATP-driven domain motions within dynein's ring have not yet been delineated, our data and recent structural studies (Kon et al., 2012; Schmidt et al., 2012) suggest that an “arginine finger” motif, reaching from AAA4 into the catalytic pocket of AAA3, contributes to the normal function of cytoplasmic dynein's ring. Our functional studies also show that mutation of this AAA4 arginine finger mimics aspects of Lis1's impact on dynein motility. Based on comparisons with related ring-shaped AAA+ machines, this motif is expected to be important for the hydrolysis of ATP at AAA3 and/or the transmission of structural changes from AAA3 to AAA4 (Erzberger and Berger, 2006). Consistent with this, the motile properties that we observe in the AAA4 arginine finger mutant closely resemble those of mutants with disrupted ATP binding/hydrolysis motifs in AAA3 (Cho et al., 2008; Kon et al., 2004). An important mechanistic distinction between the AAA4 arginine finger mutant and Lis1 is that the mutation appears to substantially reduce ATPase turnover (Zhuang et al., 2007), whereas Lis1 does not. A possible role of ATP hydrolysis in the dynein-Lis1 complex is to keep the system dynamic and therefore adaptable (see below). Because AAA1 is the principal ATPase site in dynein's ring, AAA3 and AAA4 have previously been referred to as “regulatory” AAA+ modules. Here we show that not only do AAA3 and AAA4 regulate the motions within dynein's ring, but they also form the binding site for Lis1, a ubiquitous cytoplasmic dynein regulator.
We found that Nudel allowed Lis1 to act on dynein at lower concentrations in our single-molecule experiments. This supports a tethering role for Nudel, in which it attaches to dynein's tail (via the intermediate chain of dynein) with one end of its coiled coil and binds to Lis1 with the other (Feng et al., 2000; McKenney et al., 2010; Niethammer et al., 2000; Sasaki et al., 2000; Tarricone et al., 2004; Wang and Zheng, 2011; Żylkiewicz et al., 2011), thus increasing Lis1's effective concentration. However, evidence suggests that within Nudel's tethering role exist additional regulatory layers. For example, an intriguing property of purified mammalian NudE/Nudel proteins is to suppress Lis1's effects on unloaded dynein (Torisawa et al., 2011; Yamada et al., 2008) while still enabling Lis1 to prolong dynein's microtubule attachments under force (McKenney et al., 2010). One possibility here is that Nudel positions Lis1 such that it engages AAA3/4 when dynein is strained. This would allow dynein/Lis1/Nudel to move rapidly until a high load is encountered, at which point Lis1 would engage and resist microtubule detachment. This differs from our results from S. cerevisiae, in which Nudel enhanced Lis1's effects on dynein even in unloaded conditions. Therefore, the load-dependent behavior of dynein, Lis1, and Nudel may vary among species or as a function of posttranslational modification.
Lis1 has been proposed to facilitate two general steps in the cytoplasmic dynein pathway: preparing dynein for transport and allowing dynein to move large, high-load cargos. Our model (Figure 7) provides a plausible mechanism for both types of Lis1 function and also identifies additional possible roles for dynein/Lis1 in living cells. A common theme in these functions is that by acting as a molecular clutch, Lis1 prolongs dynein's attachments to microtubules.
An initial step in many dynein-mediated transport events is the targeting of dynein to the plus ends of microtubules, which grow and shrink near the cell periphery (Kardon and Vale, 2009). Here, dynein is thought to be loaded with cargo before transporting these cargo molecules toward the microtubule minus end. In cells, Lis1 is also concentrated at microtubule plus ends and can be targeted there either in complex with dynein (as in S. cerevisiae and possibly mammalian cells) or separately (as in A. nidulans and Ustilago maydis; Lee et al., 2003; Lenz et al., 2006; Sheeman et al., 2003; Yamada et al., 2008; Zhang et al., 2002). In either scenario, our data indicate that binding of Lis1 to dynein at the microtubule plus end will bias dynein toward a microtubule-attached state. This could increase dynein's residence time at the microtubule plus end, assist in the kinetics of cargo loading, and/or form the starting configuration for motility toward the minus end. By binding near the neck region of dynein's tail, Lis1 may also be involved in “unmasking” the tail to promote interactions with cargo (Egan et al., 2012; Lenz et al., 2006; Markus and Lee, 2011; Markus et al., 2009).
In S. cerevisiae, the cue for transferring dynein onto its cargo protein could also involve the controlled assembly of dynein with the dynactin complex (Markus and Lee, 2011; Markus et al., 2009; Woodruff et al., 2009). Formation of the dynein/dynactin complex may help displace Lis1 and Nudel (McKenney et al., 2011) and convert dynein from a strongly microtubule-attached state (with Lis1 bound) to a rapidly moving processive state (with dynactin bound).
A second, more recently proposed role of Lis1 is to adapt dynein for tasks involving high loads, such as transporting and exerting tension on large organelles and cellular structures (McKenney et al., 2010). By binding near the base of dynein's microtubule-binding stalk and causing it to resist detachment, Lis1 appears ideally suited to facilitate dynein's tension-bearing roles. It remains to be seen whether the ATP-driven movements of dynein's mechanical element (the “linker”) persist with Lis1 bound (Burgess et al., 2003; Kon et al., 2005; Roberts et al., 2009, 2012). If so, this might allow dynein/Lis1 to perform repeated “tugs” on cargo while attached to the microtubule, thereby maintaining tension. Alternatively, Lis1 might regulate the linker domain's motions, as the distal end of the linker lies in close proximity to the Lis1 binding site at AAA3/4 (Figure S4E). The impressive duration of dynein's microtubule attachments in the presence of Lis1 also provides clues to additional possible functions in vivo. A subset of macromolecules and organelles, including mRNAs and the Golgi apparatus, appears to require dynein for retention at specific sites in the cytoplasm (Delanoue and Davis, 2005; Ding et al., 2009; Lam et al., 2010). An idea compatible with our data is that tension-bearing dynein/Lis1 complexes are responsible for anchoring these cargos in place.
An important question is how Lis1 might assist in the rapid, dynein-driven transport of large organelles, as was recently reported in neuronal cells (Pandey and Smith, 2011; Yi et al., 2011). Given that the velocity of these movements (~300 nm/s) exceeds those expected from dynein in saturating Lis1 concentrations (Yamada et al., 2008), one possibility is that Lis1 is substoichiometrically associated with dynein on such cargos. Alternatives include Lis1 allowing the summation of multiple individual dynein forces (McKenney et al., 2010), and Nudel modulating the dynein-Lis1 interaction in a load-sensitive manner. Quantifying the endogenous copy number of dynein, Lis1, and Nudel on these cargos by live-cell imaging would help distinguish among these possibilities.
In the other classes of cytoskeletal motors, different motile properties have been achieved by gene duplication and divergent evolution. In the case of cytoplasmic dynein, our results reveal that Lis1 can generate mechanical diversity by regulating the structural changes that propagate through dynein's large motor domain.
The S. cerevisiae strains used in this study are listed in Table S1. Dynein constructs were purified and labeled essentially as described previously (Reck-Peterson et al., 2006). For modifications and details of the Lis1 and Nudel purifications, see Extended Experimental Procedures. Protein concentrations were determined by comparisons with standards using Bradford protein assays or SDS-PAGE with SYPRO Red (Invitrogen) staining. All protein concentrations (dynein, Lis1, Nudel, and α/β-tubulin) are expressed for the dimer, with the exception of Lis1ΔN, for which the monomer concentration is given.
The S. cerevisiae strains used in this study are listed in Table S1. The endogenous genomic copies of DYN1, PAC1, NDL1, and NIP100 (encoding the dynein heavy chain, Lis1, Nudel, and dynactin subunit p150Glued orthologs, respectively) were deleted or modified using PCR-based methods as previously described (Longtine et al., 1998). Transformations were performed using the standard lithium acetate method (Guthrie and Fink, 1991) and screened by colony PCR. Point mutants were generated using either the PCR stitching method or QuikChange site-directed mutagenesis (Stratagene) and verified by DNA sequencing.
Log-phase cells growing at 30°C were transferred to 14°C for 16 hr. Cells were fixed with 75% ethanol and visualized using a Nikon 80i upright microscope equipped with a Hamamatsu C8484-03 monochrome camera and controlled by Metamorph software. The percentage of aberrant binucleate cells was calculated as the number of binucleate cells divided by the sum of dividing wild-type and binucleate cells.
Cultures of S. cerevisiae for protein purification were grown, harvested, and frozen as previously described (Reck-Peterson et al., 2006). Dynein constructs were purified via their ZZ tag, labeled with HaloTag-TMR or HaloTag-PEG-biotin ligands (Promega), and eluted into a modified TEV buffer (50 mM Tris-HCl [pH 8.0], 150 mM potassium acetate, 2 mM magnesium acetate, 1 mM EGTA, 10% glycerol, 1 mM DTT, 1 mM PMSF, and 0.1 mM Mg-ATP). Nudel was similarly purified via its ZZ tag, with the exception that the lysis buffer additionally contained a phosphatase inhibitor cocktail (Roche), and all buffers lacked Mg-ATP.
Lis1 constructs were purified via His8 and ZZ tags. Lysed cells were resuspended in Buffer A (final concentrations: 50 mM potassium phosphate [pH 8.0], 150 mM potassium acetate, 150 mM NaCl, 2 mM magnesium acetate, 5 mM β-mercaptoethanol, 10% glycerol, 0.2% Triton X-100, 0.5 mM Pefabloc, and 1 mM PMSF) supplemented with 10 mM imidazole (pH 7.5). Subsequent steps were at 4°C unless indicated. The lysate was clarified by centrifugation at 264,900 g for 1 hr. The supernatant was incubated with Ni-NTA agarose (QIAGEN) for 1 hr, transferred into a column, washed three times with Buffer A + 20 mM imidazole, and eluted with Buffer A + 250 mM imidazole. Eluted protein was then incubated with IgG sepharose beads (Amersham Pharmacia) for 1 hr, transferred into a column, and washed twice with Buffer A + 20 mM imidazole and once with TEV buffer (10 mM Tris-HCl [pH 8.0], 150 mM KCl, 10% glycerol, 0.2% Triton X-100, 1 mM PMSF, and 1 mM DTT). Lis1 was released from beads via incubation with TEV protease for 1 hr at 16°C, resulting in cleavage from the His8-ZZ tag.
Fluorescently labeled molecules were visualized using an Olympus IX-81 TIRF microscope with a 100X 1.45 N.A. oil immersion TIRF objective (Olympus) and four CW diode-pumped solid-state lasers (405 nm and 640 nm cubic lasers [Coherent], and 491 nm and 561 nm lasers [Cobolt]). Laser power at the objective was 1.5–4 mW. Images were recorded with a 100 ms exposure on a back-thinned electron multiplier CCD camera (Hamamatsu) controlled by Metamorph software.
Single-molecule motility assays were performed using flow chambers as previously described (Case et al., 1997). For standard single-molecule dynein motility assays, dynein was labeled with TMR, and microtubules contained ~10% biotin-tubulin for surface attachment and ~10% Atto647-, HyLite488-, or fluorescein-tubulin (Cytoskeleton) for visualization. The imaging buffer consisted of 30 mM HEPES (pH 7.2), 50 mM potassium acetate, 2 mM magnesium acetate, 1 mM EGTA, 10% glycerol, 1 mM DTT, 20 μM taxol, 1.25 mg/ml casein, 1 mM Mg-ATP, and an oxygen scavenger system. Images were recorded every 2 s for 10 min, and velocities, durations of microtubule association, and run lengths were calculated from kymographs generated in ImageJ (National Institutes of Health). Runs > 3 s or > 0.5 μm were measured in duration or run length analysis, respectively. All curve fitting was performed in Prism5 (GraphPad). For microtubule gliding assays, GFP-tagged dynein constructs were attached to the coverglass using a polyclonal anti-GFP antibody, and microtubules contained 10% TMR-tubulin. For high-precision stepping experiments, dynein-PEG-biotin molecules were labeled with 655 Qdot streptavidin (Invitrogen) under conditions that yield monovalent Qdot attachment (Reck-Peterson et al., 2006). Images were recorded every 100 ms, and fluorescent spots were fitted with a 2D Gaussian function to precisely localize their position. Steps were detected from the displacement records using a custom MATLAB (The MathWorks) program (Qiu et al., 2012).
For single-molecule microtubule release assays, we used modified flow chambers (Tanner et al., 2009) with inlet and outlet tubes such that buffer conditions could be changed during imaging via a syringe. Monomeric, TMR-labeled dynein331 kDa molecules were bound to immobilized microtubules or axonemal microtubules in the absence of Mg-ATP for 2 min, and then washed with low-salt imaging buffer (lacking potassium acetate) supplemented with 800 nM Lis1 where indicated. After 10 s of imaging (frame rate of 100 ms), the flow chamber was perfused with buffer supplemented with 5 mM ATP and 800 nM Lis1 as indicated.
Glutaraldehyde crosslinking was used to probe the dimerization status of intact Lis1 and Lis1ΔN. Proteins (0.5 μM) in crosslinking buffer (30 mM HEPES [pH 7.2], 50 mM potassium acetate) were incubated ± 0.05% glutaraldehyde for 5 min at ambient temperature. Reactions were quenched by the addition of 25 mM Tris-HCl (pH 8.0). As a negative control, proteins were pretreated with 4% LDS to denature them.
ATPase assays were performed using an EnzChek phosphatase kit (Molecular Probes) as previously described (Cho et al., 2008; Reck-Peterson et al., 2006). The final reaction consisted of 20 nM GST-dynein331 kDa, 0 or 140 nM Lis1, 0-10 μM taxol-stabilized microtubules, 2 mM Mg-ATP, 200 μM 2-amino-6-mercapto-7-methyl-purine riboside, 1 U/ml purine nucleoside phosphorylase, and assay buffer (30 mM HEPES [pH 7.2], 50 mM potassium acetate, 2 mM magnesium acetate, 1 mM EGTA, 1 mM DTT, and 10 μM taxol). A SpectraMax384 plate reader (Molecular Devices) was used to monitor the coupled reaction at OD360 every 5 s for 5 min. Data were fit according to Nishiura et al. (2004).
For size-exclusion chromatography, GST-dynein331 kDa (350–400 nM), Lis1 (700–800 nM), or a mixture of both proteins was preincubated for 10 min at 4°C under the indicated nucleotide conditions (200 μM Mg-ATP + 200 μM sodium vanadate, 200 μM Mg-ATP alone, or no added nucleotide). Samples were fractionated on a Superose 6 PC 3.2/30 column using an ÄKTAmicro system (GE Healthcare) that had been equilibrated with gel filtration buffer (50 mM Tris-HCl [pH 8.0], 150 mM potassium acetate, 2 mM magnesium acetate, 1 mM EGTA, 5% glycerol, and 1 mM DTT) with nucleotide as required. Fractions (50 μl) were analyzed by SDS-PAGE on 4%–12% Tris-Bis gels (Invitrogen) with SYPRO Red staining (Invitrogen), and imaged using an ImageQuant 300 gel imaging system (BioRad). Peak fractions were pooled and flash frozen in liquid nitrogen. Stokes radii were determined by comparisons with protein standards (thyroglobulin, apoferritin, β-amylase, alcohol dehydrogenase, albumin, and carbonic anhydrase [BioRad]).
Peak chromatography fractions containing the GST-dynein331 kDa/Lis1 complex or GST-dynein331 kDa alone were diluted to 30 μg/ml in gel filtration buffer lacking DTT and glycerol. Specimen (5 μl) was applied to freshly glow-discharged, carbon-coated EM grids and stained with 1% uranyl formate solution. Micrographs were acquired using a Tecnai G2 Spirit microscope operating at 120 kV with a LaB6 electron source and a 4k × 4k CCD camera (Gatan US4000). The nominal magnification was 49,000X, giving a sampling of 2.2 Å/pixel at the object level. Single-particle images were selected from the micrographs using BOXER (Ludtke et al., 1999) and processed using SPIDER (Frank et al., 1996) as described previously (Roberts et al., 2009). For measurement of the motor-motor separation, images of dynein dimers were aligned by reference-free methods and classified into groups with an average of five images per class. The centroid positions of the two motor domains in each class average were recorded, and the separation was calculated using arithmetic commands in SPIDER. For mapping of the Lis1 binding site, individual motor domains were extracted from the dimer images using a soft-edged circular mask. The two data sets (±Lis1) were coaligned based on features within the motor domain and separated by classification into characteristic views. Difference maps were calculated between normalized averages from each data set (mean pixel intensity = 0, SD = 1; equal number of particles in each average). Crystal structures were displayed with the use of PyMOL (v18.104.22.168; Schrödinger) and Chimera (Pettersen et al., 2004) software.
Assays are described in detail in the Extended Experimental Procedures. Briefly, motility assays were assembled using flow chambers, and fluorescently labeled dynein molecules and microtubules were visualized by TIRF microscopy (Qiu et al., 2012; Reck-Peterson et al., 2006). Microtubule-activated ATPase assays were performed using the EnzChek phosphatase kit (Molecular Probes; Cho et al., 2008; Reck-Peterson et al., 2006).
GST-dynein331 kDa (350–400 nM), Lis1 (700–800 nM), or a mixture of both proteins was preincubated for 10 min at 4°C under the indicated nucleotide conditions (200 μM Mg-ATP + 200 μM sodium vanadate, 200 μM Mg-ATP alone, or no added nucleotide). Samples were fractionated on a Superose 6 PC 3.2/30 column equilibrated with buffer and nucleotide as required. For negative-stain EM, peak fractions were stained with 1% uranyl formate and imaged using a Tecnai G2 Spirit microscope, as detailed in the Extended Experimental Procedures.
We thank Andrew P. Carter, Peter J. Knight, Tom A. Rapoport, James R. Sellers, Katerina Toropova, and Xin Xiang for critical comments on the manuscript; the Nikon Imaging Center at Harvard Medical School for technical assistance; Priscilla Yang's lab for sharing equipment; Stephen Hinshaw for initial experiments and suggesting the clutch analogy; Weihong Qiu for initial experiments; Brian Goodman for kinesin; and the Reck-Peterson lab for support. A.J.R. is funded by a Sir Henry Wellcome Postdoctoral Fellowship (092436/Z/10/Z) sponsored by Peter J. Knight and Stan A. Burgess (University of Leeds). A.E.L. is funded in part by a research fellowship from the Alfred P. Sloan Foundation. S.R.-P. is funded by the Rita Allen Foundation, the Giovanni Armenise-Harvard Foundation, and a National Institutes of Health New Innovator award (1 DP2 OD004268-01).