Expressed KIF1A Motors Are Inactive in Mammalian Cells
To study the regulation and motile properties of mammalian KIF1A under native conditions, rat KIF1A (Figure S1
) was tagged with monomeric citrine (mCit), a variant of yellow fluorescent protein (FP), and expressed in COS cells. This approach has been used successfully to study Kinesin-1 motors and avoids potential problems associated with in vitro purification and/or labeling [2
]. In live cells, motor activity can be determined using the nonhydrolyzable ATP analog AMPPNP (5′-adenylyl-beta,gamma-imidodiphosphate) to block the release of active kinesin motors from microtubules [2
]. COS cells expressing mCit-KIF1A were transiently permeabilized with the bacterial toxin streptolysin O (SLO). Upon addition of AMPPNP, mCit-KIF1A did not become trapped on microtubules but remained diffuse and cytosolic (A and B), indicating that KIF1A is not engaged with microtubules when expressed in mammalian cells. Identical results were obtained when the mCit tag was placed at the C-terminus of KIF1A (KIF1A-mCit, Figure S2
). The inability to bind microtubules is inherent to KIF1A, as untagged and Myc-tagged versions of KIF1A also did not become trapped on microtubules (Figure S2
). In contrast, active versions of Kinesin-1 rapidly became locked on microtubules upon addition of AMPPNP ([2
] and Figure S2
). We conclude that mammalian KIF1A is inactive in vivo.
KIF1A Exists as a Dimeric Motor That Is Inactive for Microtubule Binding In Vivo
Expressed KIF1A Is a Dimeric Protein
KIF1A could be inactive due to either of the two proposed models for kinesin motor regulation. To distinguish between these possibilities, we first investigated whether expressed KIF1A motors exist as monomers or dimers using coimmunoprecipitation. COS cell lysates coexpressing mCit- and Myc-tagged KIF1A proteins were immunoprecipitated with control antibodies (immunoglobulin G [IgG]) or antibodies to the Myc tag. Coprecipitation of mCit-KIF1A with Myc-KIF1A was observed only with the Myc antibody (C), suggesting that KIF1A exists in a dimeric state. As an alternative method, we used chemical crosslinking. Myc- or mCit-tagged versions of KIF1A were expressed in COS cells, and lysates were untreated or treated with dimethylpimilimidate (DMP). In the presence of DMP, both Myc-KIF1A and mCit-KIF1A motors migrated at approximately twice the molecular weight (MW) of the corresponding noncrosslinked proteins (D, lanes 1–6). A known dimeric motor, the kinesin heavy chain (KHC) subunit of Kinesin-1, undergoes a similar DMP-induced mobility shift (D, lanes 7 and 8). In contrast, mCit showed no shift in mobility upon DMP treatment (D, lanes 9 and 10). These results indicate that the expressed KIF1A protein exists in a dimeric state.
We next probed the monomer/dimer state of KIF1A using fluorescence resonance energy transfer (FRET) stoichiometry, a method that calculates an average FRET efficiency (EAVE
) and minimizes effects of variable donor and acceptor FP expression [2
]. Control experiments show an EAVE
= 0% for unlinked donor (monomeric cyan FP [mCFP]) and acceptor (mCit) FPs and an EAVE
= 37% for mCFP and mCit linked by 16 amino acids (unpublished data). For donor and acceptor FPs placed on the N-terminus of KIF1A (“motor-to-motor” FRET), a low but measurable FRET efficiency was obtained (EAVE
= 3.0 ± 1.2%, n
= 45 cells, E and F). This value is not compatible with a monomeric state of the motor. Rather, this value indicates that KIF1A motors exist in a dimeric state in live cells. The motor-to-motor FRET efficiency of KIF1A is comparable to that of the Kinesin-1 holoenzyme (EAVE
= 2.1 ± 0.4%, n
= 30 cells, E and F) whose motor domains are “pushed apart” in the inactive state [2
]. By analogy, it is thus possible that the two motor domains of inactive, dimeric KIF1A motors may be pushed apart as part of the regulatory mechanism. FRET stoichiometry was also used to probe the overall conformation of KIF1A. For donor and acceptor FPs localized at the N- and C-termini of a single KIF1A polypeptide (mCit-KIF1A-mCFP), the low but measurable “motor-to-tail” FRET (EAVE
= 4.8 ± 1.1%, n
= 39 cells, Figure S3
) confirms that KIF1A is not a fully extended molecule. We conclude that KIF1A exists in a compact dimeric state but is not active for microtubule binding.
Endogenous KIF1A Motors Exist in a Dimeric State
That KIF1A motors expressed under native conditions exist in a dimeric state was surprising as recombinant and endogenous KIF1A/Unc104 motors have been defined as monomeric kinesins based on hydrodynamic analysis. Thus, we tested whether endogenous KIF1A motors also exist in a dimeric state. We first used crosslinking of cytosolic extracts from rat brain and detergent extracts of murine cortical neurons. In the presence of DMP, the endogenous KIF1A proteins showed a reduced mobility (G), suggestive of a dimeric state. Crosslinking analysis is dependent on the ability of the antibody to recognize the crosslinked species. Unfortunately, the KIF1A antibody was less able to recognize its antigenic sequence after crosslinking than were antibodies to the epitope tags (D). Thus, we sought an alternative method to analyze the monomer/dimer state of endogenous KIF1A motors. Previous reports used hydrodynamic analysis of KIF1A as compared to MW standards [13
]. As the shape of kinesin motors may influence their hydrodynamic properties, we sought to compare the sedimentation of endogenous KIF1A motors in sucrose gradients to motors with known oligomeric states. Cytosolic extracts of rat brain or detergent extracts of COS cells expressing dimeric Myc-KIF1A or mCit-KIF1A motors were separated by sucrose gradient sedimentation. The majority of the expressed Myc-KIF1A was found in fractions 5–7 (H), whereas the majority of the expressed mCit-KIF1A was found in fractions 6 and 7 (H) with the shift likely due to the mCit tag. The endogenous Kinesin-1 protein in rat brain cytosol was also found in fractions 6 and 7 (H). Strikingly, the endogenous KIF1A protein in rat brain cytosol was found in fractions 5–7 (H), a mobility identical to that of the expressed dimeric Myc-KIF1A and mCit-KIF1A proteins. It is interesting to note that Kinesin-1 (~360 kDa) and KIF1A motors (~380–440 kDa) sediment slower than the marker protein catalase (240 kDa) despite their larger size. Thus, sedimentation as compared to MW standards is not a reliable indication of motor mass. Taken together, these results indicate that endogenous KIF1A exists as a dimeric protein.
Two Autoinhibitory Mechanisms Regulate KIF1A
That KIF1A exists in a dimeric and inactive state suggests that dimerization is not sufficient for activation. Thus, we tested an autoinhibitory mechanism for KIF1A regulation as related to the domain structure and dimer state of the protein. In KIF1A, a coiled-coil (CC) segment adjacent to the motor domain, referred to as the neck coil (NC), is predicted by the COILS program only if a 14–amino acid window is used for analysis (A). The NC is followed by a region of strong coiled-coil prediction (CC1), a forkhead-associated (FHA) domain, and coiled-coil segments CC2 and CC3 (A).
Truncation of the FHA+CC2 Region Relieves Autoinhibition of Microtubule Binding but Not Processive Motility
Based on this domain structure, we created truncated versions of KIF1A (B) by placing a mCit tag after CC2 [KIF1A(1–726)], after CC1 [KIF1A(1–491)], or after the NC [KIF1A(1–393)]. We first tested the ability of these constructs to bind to microtubules. KIF1A(1–726)-mCit remained cytosolic and did not localize to microtubules upon addition of AMPPNP (C–E), indicating that this protein retains the autoinhibited state of the full-length (FL) molecule. In contrast, deletion of the FHA and CC2 segments resulted in a motor, KIF1A(1–491)-mCit, that became trapped in a microtubule-bound state upon addition of AMPPNP (C–E). Thus, the FHA and CC2 domains contribute to autoinhibition by blocking productive interactions with microtubules. This is consistent with previous work on truncated KIF1A/Unc104 motors in vitro [26
]. Interestingly, deletion of the CC1 domain resulted in a construct, KIF1A(1–393)-mCit, that accumulated on peripheral microtubules at steady state, suggesting that this may be a processive motor. Addition of AMPPNP resulted in an increase in microtubule localization, most notably in the central regions of the cell (C). These results suggest that CC1 negatively regulates motility of KIF1A.
To directly analyze the processive motility of FL and truncated KIF1A motors, we used two assays. First, motility in vivo was indirectly assessed by the ability of KIF1A motors to accumulate at the tips of neurites in neuron-like CAD cells. Second, processive motility in vitro was directly measured using single-molecule motility assays and a total internal reflection fluorescence (TIRF) microscope. For the latter, KIF1A constructs were tagged with three tandem copies of mCit for improved signal and decreased photobleaching and photoblinking [29
When expressed in differentiated CAD cells, KIF1A-mCit and (1–726)-mCit were diffusely localized in the cell body and did not concentrate at the ends of neuronal processes (F). In addition, very few motility events were observed in vitro for KIF1A-3xmCit or (1–726)-3xmCit motors (G and H, ). These results indicate that truncation of the C-terminal half of KIF1A is not sufficient to relieve autoinhibition of microtubule binding (C–E) or processive motility (F–H). Truncation of the FHA+CC2 region also resulted in little to no processive motility for KIF1A(1–491) in vivo (F) or in vitro (G and H, ). This was surprising as KIF1A(1–491) motors bound to microtubules in vivo (C–E). Thus, the FHA+CC2 region contributes to autoinhibition by blocking microtubule binding, but an additional mechanism(s) controls the ability of the motor to undergo processive motility. This second control mechanism may reside, at least in part, in the CC1 domain as KIF1A(1–393) showed a significantincrease in motility as these motors concentrated at the tips of neurites (F) and displayed a large number of processive motility events in vitro (G and H, ). Analysis of the motile properties of KIF1A(1–393) motors gave an average speed of 1.36 ± 0.04 μm/s and an average run length of 1.24 ± 0.06 μm per event (), comparable to previous studies [13
]. The few motility events observed for the FL, (1–726), and (1–491) motors occurred with decreased velocity and run lengths as compared to (1–393) and were likely due to a dynamic equilibrium between active and inactive states.
Summary of the Motile Properties and Oligomeric State of FL and Truncated KIF1A Motors
Taken together, these results indicate that two regions of KIF1A contribute to autoinhibition. First, the FHA and CC2 domains (amino acids 492–726) prevent microtubule binding (), and second, the CC1 domain (amino acids 394–491) inhibits processive motility ().
Deletion of CC1 Restores the Dimer State and Processive Motility
The CC1 Domain Prevents Processive Motility by Promoting the Monomeric State
To further investigate the relationship between KIF1A domain structure, autoinhibition, and dimerization, we used three assays to test whether truncated motors exist in a dimeric state. We first used chemical crosslinking of KIF1A(1–726)-mCit, KIF1A(1–491)-mCit, and KIF1A(1–393)-mCit motors. In the presence of DMP, most of the (1–726)-mCit and (1–393)-mCit proteins shifted to a higher MW species (A), consistent with a dimeric state. In contrast, very little (1–491)-mCit protein displayed a shift in mobility (A). Rather, in the presence of DMP, (1–491)-mCit motors showed either the same mobility as the uncrosslinked species or a slightly increased mobility (A), perhaps due to an intramolecular crosslink between the CC1 and NC domains [28
We next used coimmunoprecipitation of Myc- and mCit-tagged truncated KIF1A motors. Coexpression of (1–726)-Myc and (1–726)-mCit resulted in precipitation of (1–726)-mCit by the Myc antibody but not a control antibody (B), suggesting that KIF1A(1–726) exists in a dimeric state. Similar results were obtained for the (1–491) and (1–393) truncated motors (B). Thus, (1–726) and (1–393) behaved as dimeric proteins by both chemical crosslinking and coimmunoprecipitation, whereas (1–491) showed a more varied behavior.
We then took the advantage of stepwise photobleaching of FPs to test whether truncated versions of KIF1A exist as dimers. This assay has the advantage of analyzing individual motors rather than ensemble averages. Lysates of COS cells expressing 3xmCit-tagged FL or truncated KIF1A motors were analyzed by TIRF microscopy. For each construct, the fluorescence intensity of 160–200 individual fluorescent spots was recorded over time. The number of bleaching steps was determined for each spot (C) and then plotted in a histogram to show the population distribution (D). In control experiments, the majority of KHC(1–891)-3xmCit motors bleached in four to six steps (D and [29
]), consistent with the presence of six FPs in the dimeric KHC molecule. A similar distribution of bleaching events was obtained for FL, (1–726), and (1–393) motors (D), indicating a dimeric state for these KIF1A constructs. In contrast, the majority of KIF1A(1–491)-3xmCit motors bleached in two to three steps, consistent with a monomeric state (D). Taken together, these data suggest that truncated KIF1A constructs that contain only the NC and CC1 domains exist as monomers that are not capable of processive motility.
The NC Is Required for Dimerization of KIF1A
To directly demonstrate that removal of the CC1 domain restores processive motility as well as the dimeric state, we used two-color TIRF imaging to simultaneously track mCit and mCherry fluorescence in lysates of COS cells coexpressing KIF1A(1–393)-3xmCit and KIF1A(1–393)-3xmCherry. That (1–393) motors move as dimeric molecules is indicated by observations of fluorescent spots labeled with both mCit and mCherry that move together in a linear fashion (representative track, E). We also observed mCit and mCherry fluorescent spots with nonoverlapping motility events, as observed when KIF1A(1–393)-3xmCit and KIF1A (1–393)-3xmCherry motors were expressed separately (unpublished data). We believe that these mCit-only and mCherry-only fluorescent spots are also dimeric motors based on their fluorescence intensity. The average maximum mCit fluorescence of mCit-only spots was 513.4 ± 54.0 arbitrary units (n = 65 spots). This value is significantly greater than the average maximum mCit fluorescence intensity (356.8 ± 51.4 arbitrary units, n = 44 spots) of spots that colabeled and comigrated with mCherry. These results provide the first direct demonstration that KIF1A moves in a directed manner as a dimeric motor.
Dimerization of KIF1A via the NC Domain
To identify the sequences required for dimerization, we generated KIF1A constructs containing various amounts of the NC region (1–369, 1–377, or 1–381, A) or the full NC and several residues of the subsequent hinge region (1–393, A). These constructs were designed to directly correspond to previously studied KIF1A motors (A). The monomer/dimer state of the NC truncations was tested by crosslinking and photobleaching analysis. In the presence of DMP, the majority of (1–393)-mCit shifted to a higher MW species (B). In contrast, (1–381)-mCit, (1–377)-mCit, and (1–369)-mCit showed no mobility change in the presence of DMP (B). In photobleaching experiments, a large proportion of KIF1A(1–393)-3xmCit molecules bleached in four to six steps (C and D), consistent with a dimeric state containing six FPs. However, KIF1A(1–381)-3xmCit and KIF1A(1–369)-3xmCit molecules bleached primarily in two or three steps (C and D), indicating a monomeric state. These results suggest that the entire NC, as well as residues in the subsequent hinge region, are required for dimerization. This is consistent with a study of synthesized NC peptides where several residues beyond G387 were required to prevent dissociation [30
Motile Characteristics of Dimeric KIF1A Motors
We next set out to compare the microtubule-based properties of monomeric and dimeric KIF1A motors. We first confirmed that monomeric motors retain the ability to bind to microtubules. Indeed, upon AMPPNP treatment, monomeric (1–381)-mCit and (1–369)-mCit motors became locked on microtubules, similar to dimeric (1–393)-mCit motors (A and B). We next tested the ability of the NC constructs to undergo processive motility in vivo. The dimeric motor (1–393)-mCit accumulated in neurite tips of differentiated CAD cells, whereas the monomeric motors (1–381)-mCit and (1–369)-mCit remained primarily in the cell bodies (C). These results confirm that the full NC is required for dimerization as well as processive motility in mammalian cells.
Motor Properties of Monomeric and Dimeric KIF1A Motors
Finally, we investigated the motile characteristics of (1–393)-3xmCit, (1–381)-3xmCit, and (1–369)-3xmCit motors by in vitro single-molecule motility assays. Motility events that persisted for at least five frames (500 ms) were analyzed to determine velocities and run lengths. As expected, a large number of motility events were observed for dimeric (1–393) motors (D and E, ). Surprisingly, motility events were also observed for the monomeric (1–381) and (1–369) motors (D and E, ) even though these motors were not processive in vivo (C). However, the motility of the monomeric motors differed from that of the dimeric motor in three ways. First, monomeric motors showed a significant decrease in the number of observed motility events (). Second, qualitative analysis of the velocity and run-length histograms shows that only dimeric motors gave distributions (Gaussian and single exponential, respectively) characteristic of processive motors (D and E). Third, dimeric motors underwent significantly longer run lengths (in some cases >6 μm) as dimeric (1–393) motors averaged 1.24 ± 0.06 μm/run, whereas monomeric (1–380) and (1–369) motors averaged only 0.42 ± 0.02 μm/run and 1.02 + 0.09 μm/run, respectively (E, ). Thus, monomeric motors are significantly impaired in their motile properties.
Dimeric but Not Monomeric KIF1A Motors Display ATP-Driven Processive Motility
We then used two approaches to determine whether the motility of dimeric KIF1A(1–393) motors occurs by one-dimensional (1D) diffusion, as demonstrated for monomeric KIF1A motors, or by ATP-driven processive motility, as demonstrated for other dimeric kinesin motors [19
]. Single-molecule motility assays were analyzed to include fluorescent spots visible and motile for at least three frames (300 ms) to ensure inclusion of diffusive events (representative times series, A). This resulted in an expected decrease in the average run lengths of both dimeric (1–393)-3xmCit and monomeric (1–369)-3xmCit motors (0.90 ± 0.05 μm/run and 0.65 ± 0.04 μm/run, respectively, C) as well as increases in the average velocities (1.88 ± 0.07 μm/s and 2.16 ± 0.10 μm/s, respectively, B).
ATP-Dependent Processive Motility Is a Property of Dimeric but Not Monomeric KIF1A Motors
Our first approach was to compare the biophysical properties of a large number of motility events by plotting the data as a comparison of the velocities or run lengths against the duration of each motility event (time spent in one-directional motion). By this analysis, drastic differences between monomeric and dimeric mechanisms of motility became apparent. The motility events of dimeric (1–393)-3xmCit motors could be segregated into two classes: first, motile events lasting for short periods of time (<1 s) at a wide variety of speeds and distances (D and E, top left panels, red circles) and second, motile events lasting for longer time periods (>1 s) at constant speeds of ~1.2 μm/s (D, top left panel, blue circle) and with run lengths directly dependent on the amount of time spent in motion (E, top left panel, blue circle). We hypothesized that the first class of motility (D and E, red circles) is 1D diffusion, whereas the second class of motility (D and E, blue circles) is processive motility. Indeed, monomeric KIF1A(1–369)-3xmCit motors only moved for short periods of time (<1 s) at a wide variety of speeds and distances (D and E, middle left panels, red circles), indicative of 1D diffusion. In contrast, the processive motility of dimeric KHC(1–891)-3xmCit motors was evident as these motors spent longer periods of time in motion (>1 s) at a constant speed (D, bottom left panel, blue circle) and for distances directly dependent on the time spent in motion (E, bottom left panel, blue circle). Thus, dimeric KIF1A(1–393) motors display motility properties of both 1D diffusion and processive motility.
Our second approach was to distinguish these motility classes by their ATP dependence. Diffusion of monomeric KIF1A motors occurs in the weakly bound or ADP (adenosine diphosphate) state [31
], whereas processive hand-over-hand stepping of kinesin motors requires the energy of ATP hydrolysis. When single-molecule motility assays of monomeric (1–369)-3xmCit motors were carried out in the presence of ADP, no change in motility was observed. Monomeric motors continued to move only for short periods of time (<1 s) at various speeds and run lengths (D and E, middle right panels, red circles), confirming that monomeric KIF1A(1–369)-3xmCit motors moved by 1D diffusion. In contrast, the processive motility of dimeric KHC(1–891)-3xmCit motors was abolished in the presence of ADP (D and E, bottom right panels). The motile properties of dimeric (1–393)-3xmCit motors were also highly dependent on nucleotide. Dimeric (1–393)-3xmCit motors continued to display short motility events (<1 s) of various velocities and run lengths (D and E, top right panels, red circles) in the presence of ADP, similar to the monomeric motors. However, dimeric motors were unable to undergo processive motility in the presence of ADP (D and E, top right panels). This analysis demonstrates that for motility events that last only short time periods, it is not possible to distinguish 1D diffusion from processive, hand-over-hand motor stepping (overlap of red and blue circles). However, for events that last longer than 0.5–1 s, comparisons of velocity and run length to time spent in directional motility enable the separation of motility mechanisms. We conclude that KIF1A motors exist as dimeric molecules that can move by 1D diffusion but show processive motility only in the presence of ATP.