Single Molecule Imaging Reveals Differences in Microtubule Track Selection Between Kinesin Motors

1Understanding how cells generate intracellular structures and overall morphologies is one of the major goals of cell biology. For the cytoskeleton, strikingly different structures can be assembled from a set of highly conserved building blocks. For example, all microtubules are generated by the polymerization of a common α/β-tubulin subunit yet diverse microtubule populations can be generated (e.g., axonemes, spindles, and radial arrays) that carry out distinct functions.

2One way microtubule diversity can be characterized is based on dynamic properties. Some microtubules are dynamic and turn over rapidly by alternating between periods of microtubule growth (polymerization) and shrinkage (depolymerization). Other microtubules are stable (low turnover) and are defined by their resistance to drugs that result in depolymerization of microtubules, such as nocodazole. In vivo, microtubules frequently pause, undergoing neither polymerization nor depolymerization. Microtubule diversity can also be characterized by structural differences, for example alterations in protofilament number, as well as by chemical differences between tubulin subunits due to differences in the expression of tubulin genes (isotypes) or the presence of post-translational modifications (PTMs),–.

3What are the biological functions of microtubules diversity? Dynamic instability allows microtubules to explore three-dimensional space for rapid remodeling of the cytoskeleton during processes such as spindle assembly and cell migration –. Stable microtubules likely play important roles in cellular morphogenesis but how and why remain unclear –. The chemically diverse PTMs that mark stable microtubules may affect morphogenesis by stabilizing microtubules and/or by influencing distinct intracellular transport events –. The expression of tubulin isotypes can influence polymerization dynamics and plays a role in the formation of specific microtubule assemblies such as the flagellar axoneme,.

4The challenge is to explain how the diversity of microtubule structures is translated into specific cellular functions. This likely requires the functions of a large number of microtubule associated proteins (MAPs). Of special interest are motor proteins of the kinesin and dynein families that use ATP hydrolysis to move cellular cargoes along microtubule tracks,. The molecular and mechanistic properties of motor proteins have typically been studied in vitro using homogeneous microtubule assemblies. Thus, understanding how motor proteins could read microtubule diversity has been difficult to answer. We recently developed techniques for tracking kinesin motors at the single molecule level in the cytoplasm of live cells. Here, we extend these techniques to two-color tracking and evaluate the motility of kinesin motors along heterogeneous populations of microtubules tracks in COS cells. We show that Kinesin-1 motors move preferentially along stable microtubules marked by PTMs whereas Kinesin-2 (KIF17) and Kinesin-3 (KIF1A) motors can utilize dynamic microtubule tracks. These results indicate that kinesin motors have evolved to recognize specific microtubule subpopulations and, thus, segregate membrane trafficking events within cells.

5Constitutively active kinesin motors can be generated by truncations that remove autoinhibitory and cargo-binding regions of the polypeptide. For this work, we generated KHC(1-560) (Figure 1A), a dimeric motor that has been well characterized in vitro and in vivo,,. KHC(1-560) motors were tagged with three tandem copies of monomeric Citrine (mCit), a variant of enhanced yellow fluorescent protein (FP) (Figure 1A), and expressed in COS cells (Figure 1B). Single Kinesin-1 motors were tracked in live cells using a modified TIRF microscope (Figure 1C) in which the angle of illumination was varied to enable deeper imaging as described. KHC(1-560)-3xmCit motors were observed to undergo both free diffusion and linear movement (Video S1). Linear motility occurred with an average speed of 0.83±0.08 µm/sec and average run length of 0.91±0.23 µm in live cells (Table 1, n = 372 events), consistent with previous work,.

6To gain an understanding of the ensemble characteristics of Kinesin-1 motility events, we developed methods to sum all motility events of a time series into one image called the standard deviation map (SD Map). We compute this map by determining the standard deviation (SD) of the fluorescence intensity change for each pixel over the entire time series (Figure S1). Thus, pixels with little to no fluorescence variation over the time series have low SDs whereas pixels with large fluctuations in intensity (e.g., where motility events occurred) are highlighted in the SD Map. As seen in the SD Map of a single molecule TIRF movie of KHC(1-560)-3xmCit in live COS cells (Figure 1D), multiple individual Kinesin-1 motors moved repeatedly on linear tracks. Only a few Kinesin-1 tracks were identified in each image series suggesting that the SD Map reveals “hot spots” of Kinesin-1 motility. That these hot spots are not due to TIRF imaging of microtubules at the bottom of the cell is indicated by the observation of many microtubules in a comparable area of COS cytoplasm (Figure 1E). We hypothesized that Kinesin-1 motors move preferentially on only a subset of microtubule tracks.

7To directly test the possibility that Kinesin-1 motors distinguish microtubule populations in cells, we used two approaches. We first performed two-color TIRF imaging (Figure 1F) of live cells expressing FP-tagged Kinesin-1 motors and microtubules. This approach allows us to simultaneously visualize motors and their tracks and can account for shifts in the position of individual microtubules during live cell imaging (see and Figure S2). COS cells were first transfected with plasmids encoding mCherry-tubulin and 24 h later with plasmids encoding KHC(1-560)-3xmCit. After an additional 5 h of expression, the cells were imaged by TIRF microscopy. KHC(1-560)-3xmCit motility events were observed along mCherry-tubulin microtubules (Video S2). Merged images of the SD Map of KHC(1-560)-3xmCit motility events and the mCherry-tubulin fluorescence (Figure 1F, representative of n = 4 cells in two experiments) indicate that mCherry-tubulin microtubules overlapped with 96.6%±5.2% of the Kinesin-1 tracks. In contrast, Kinesin-1 motility events were observed on only 9.1%±8.1% of the mCherry-tubulin microtubules (Table 1). These data indicate that Kinesin-1 motors utilize only a subset of the available microtubule tracks.

8We then performed retrospective immunofluorescence (Figure 1G) after single molecule TIRF imaging. This approach avoids difficulties with double transfection and the possibility that mCherry-tubulin does not incorporate into all microtubules yet is hindered by the fact that cells shrink during fixation, often resulting in a shift in position of the entire microtubule population. For retrospective imaging, COS cells expressing KHC(1-560)-3xmCit motors were imaged by TIRF microscopy, fixed, and then stained with an antibody to total tubulin (Figure 1G). The previously imaged cell was again observed by TIRF microscopy. A comparison of the SD Map of Kinesin-1 motility events and the total tubulin staining confirmed that the motility events occurred on only a subset of microtubules present in the imaging field (Figure 1G, representative of n = 16 cells in six experiments). Taken together, these results confirm that Kinesin-1 motors preferentially utilize only a subset of the microtubules present in COS cells.

9We first tested whether Kinesin-1 motors move preferentially on dynamic microtubules. This population can be observed in live cells expressing FP-tagged plus-end tracking proteins (+TIPs). Two-color TIRF microscopy was used to analyze COS cells coexpressing KHC(1-560)-3xmCit with a mCherry-labeled version of the +TIP protein end binding (EB)3. Very few Kinesin-1 motility events could be observed on microtubules extending back from the EB3-mCherry-labeled plus ends (Video S3). A comparison of the SD Map of KHC(1-560)-3xmCit motility events with the average EB3-mCherry fluorescence (Figure 2A, representative of n = 12 cells in four experiments) demonstrates that the microtubule tracks utilized by KHC(1-560)-3xmCit motors are distinct from the dynamic microtubules marked by EB3-mCherry. In some cases, multiple Kinesin-1 motility events occurred on a microtubule track that appeared to lie directly adjacent to an EB3-marked dynamic microtubule (boxed region in Figure 2A, kymographs in Figure 2B, 2C). Kinesin-1 motility events overlapped with only 2.5%±1.1% of the EB3-mCherry-marked microtubules (Table 1). These results indicate that the preferential motility of Kinesin-1 motors does not occur on dynamic microtubules.

10An avoidance of dynamic microtubules is likely to be advantageous to the motor since the speed of Kinesin-1 is greater than that of microtubule polymerization. In the representative cell shown in Figure 2, KHC(1-560)-3xmCit motors moved at an average speed of 0.73±0.16 µm/sec (Figure 2D, red traces) whereas EB3-labeled microtubules grew at an average rate of 0.08±0.03 µm/sec (Figure 2D, black traces). Thus, Kinesin-1 motors that move along dynamic microtubules could rapidly run off the end of the track.

11We then tested whether Kinesin-1 motors move preferentially on stable microtubule tracks. To do this, we performed retrospective immunofluorescence staining using antibodies that recognize the PTMs that mark stable microtubules. Cells expressing KHC(1-560)-3xmCit were imaged in the TIRF microscope, fixed, stained with antibodies to acetylated α-tubulin and total tubulin, and the previously imaged cells were again viewed on the TIRF microscope. The pattern of KHC(1-560)-3xmCit motility events in the resulting SD Map was similar to the pattern of acetylated microtubules (Figure 3A, representative of 11 cells in six experiments). Kinesin-1 motility events colocalized with 90.3%±5.5% of microtubules marked by acetylated tubulin (Table 1). This suggests that Kinesin-1 moves preferentially along stable microtubules marked by acetylation of α-tubulin.

12We then used retrospective immunofluorescence to test whether Kinesin-1 motility events correlate with the presence of other PTMs that mark stable microtubules. The pattern of KHC(1-560)-3xmCit motility events in the SD Map was similar to that of the microtubule tracks marked by detyrosination (Figure 3B, representative of six cells in five experiments), a modification that appears to mark the same microtubule tracks as acetylation (see and Figure S3). Kinesin-1 motility events did not colocalize with microtubules marked by polyglutamylation (Figure 3C, representative of eight cells in three experiments), most likely due to the low levels of glutamylation on cytoplasmic microtubules in these (Figure S4) and other non-neuronal cells. We conclude that Kinesin-1 motors move preferentially along microtubules marked by acetylation and detyrosination.

13Is preferential motility on stable microtubules a general feature of kinesin motors that drive vesicular transport events? To test this, we performed single molecule imaging of 3xmCit-tagged KIF17, a homodimeric member of the Kinesin-2 subfamily. KIF17 has been implicated in the transport of cargoes in dendrites of neuronal cells and in cilia of invertebrates and vertebrates –. Single molecule TIRF imaging of a constitutively active version of KIF17 [KIF17(1-490)-3xmCit, Figure 4A] showed that these motors moved with an average speed of 1.31±0.05 µm/sec and average run length of 0.56±0.22 µm in live COS cells (Table 1, n = 233 events), consistent with the motile properties of the C. elegans homologue OSM-3,.

14To test whether KIF17 motors move preferentially on a subset of microtubules, we performed two-color TIRF imaging of COS cells co-expressing mCherry-tubulin and KIF17(1-490)-3xmCit (Video S4). A comparison of the SD Map of KIF17(1-490)-3xmCit motility events with the average mCherry-tubulin fluorescence (Figure 4B) demonstrates that KIF17 motility events occurred on nearly all available microtubule tracks (Table 1). In addition, retrospective immunofluorescence imaging indicated that KIF17 motility occurred on both acetylated and non-acetylated microtubules (Table 1 and Figure S5). Together these results indicate that KIF17 is a non-selective motor as it does not show preferential motility when presented with a heterogeneous population of microtubules in COS cells.

15Since motility on dynamic microtubules would appear to be disadvantageous to plus end-directed motors, we performed two-color TIRF imaging of COS cells co-expressing KIF17(1-490)-3xmCit and EB3-mCherry to directly test whether KIF17 motors move on dynamic microtubules. A comparison of the SD Map of KIF17 motility events to the average EB3-mCherry fluorescence demonstrates that KIF17 moved on dynamic microtubules (Figure 4C and Table 1). Multiple individual KIF17(1-490)-3xmCit motors could be observed moving on the same EB3-marked microtubule (Figure 4D, 4E). Surprisingly, KIF17 motors that reached the plus end of the microtubule did not dissociate immediately, but rather lingered at the growing plus end (Figure 4E). Whether this ability of KIF17 to track the plus ends of growing microtubules is associated with its cellular functions is presently unknown.

16We then tested whether a member of the Kinesin-3 family, KIF1A, moves preferentially on a subset of microtubules. KIF1A has been implicated in the axonal transport of synaptic vesicle precursors. Constitutively active dimeric KIF1A(1-393)-3xmCit motors (Figure 5A and ) imaged by single molecule TIRF microscopy moved with an average speed of 1.82±0.04 µm/sec and average run length of 0.55±0.19 µm in live COS cells (Table 1, n = 305 events), consistent with the measured properties of this motor in vitro and its cargoes in vivo,.

17To analyze KIF1A motility on microtubules in vivo, two-color TIRF imaging of COS cells co-expressing mCherry-tubulin and KIF1A(1-393)-3xmCit was performed (Video S5). A comparison of the SD Map of KIF1A(1-393)-3xmCit motility to the average mCherry-tubulin fluorescence indicates that KIF1A motility events were observed on all of the available microtubules (Figure 5B and Table 1) indicating that KIF1A is also a non-selective motor. Consistent with this, KIF1A motility events were observed on both acetylated and non-acetylated microtubules (Table 1 and Figure S5). That KIF1A motors utilize dynamic microtubules for motility was demonstrated by two-color TIRF imaging of COS cells co-expressing KIF1A(1-393)-3xmCit and EB3-mCherry (Figure 5C). Multiple KIF1A motility events were observed to occur on individual dynamic microtubules labeled at their plus ends with EB3-mCherry (Figure 5 D, 5E and Table 1). KIF1A(1-393)-3xmCit motors that overtook the plus end of the microtubules did not fall off but, surprisingly, remained at the plus ends of growing microtubules for significant periods of time (Figure 5E). Thus, the Kinesin-3 motor KIF1A does not select specific microtubule tracks for motility in COS cells and, when using dynamic microtubules, remains localized to the growing plus ends.

18How does the preferential motility of single Kinesin-1 motors on stable microtubules relate to Kinesin-1-driven transport events inside cells? To test whether stable microtubules provide preferred tracks for motility of vesicular cargoes transported by Kinesin-1, we tracked the Golgi-to-plasma membrane transport of a variant of the vesicular stomatitis virus G protein that can be restricted to Golgi-derived vesicles using a temperature shift protocol (VSVG-GFP,,). After imaging (Video S6), we performed retrospective immunofluorescence with antibodies to total and acetylated tubulin. A comparison of the SD Map of the motility events to the acetylated microtubules in the same cell (Figure 6A, representative of 12 cells in four experiments) demonstrates that 68.0%±15.2% of VSVG-GFP-positive vesicles moved along microtubules marked by acetylated α-tubulin. Multiple VSVG-GFP-marked vesicles were observed to move independently along the same acetylated microtubule (Figure 6B, 6C). These results indicate that Golgi-derived vesicles moved by Kinesin-1 are transported preferentially along microtubules marked by acetylation.

19We then tested whether cargoes transported by “non-selective” motors, such as KIF1A or KIF17, reflect the properties of these motors. KIF1A transports presynaptic vesicles in neuronal cells, but a cargo for this motor has not been described in fibroblasts. For KIF17, we found that the steady-state cell surface levels of the voltage-gated potassium (Kv) channel Kv1.5 in HL-1 atrial myocytes was decreased by expression of a dominant negative (DN) version of KIF17 but not Kinesin-1 (Figure S6). This was surprising as KIF17 has so far only been described as a dendritic or ciliary motor,. Western blot analysis shows that the KIF17 protein is expressed in mouse brain and heart tissues as well as HL-1 myocytes (Figure S7).

20KIF17 participation in the motility of Kv1.5-GFP labeled vesicles was examined in live cells by coexpressing Kv1.5-GFP with DN versions of KIF17 or Kinesin-1 (KHC) in HL-1 cells (Figure 7A–7C). To synchronize the Kv1.5-GFP vesicle population spatially and temporally, we used a temperature shift protocol to first restrict Kv1.5-GFP to the trans-Golgi by a 19°C incubation and then initiate post-Golgi transport by incubation at 37°C. Golgi-derived Kv1.5-labeled vesicles were observed to move in a linear fashion interspersed with pauses. Expression of DN KIF17, but not DN Kinesin-1, resulted in a significant decrease in Kv1.5 vesicle motility, both in the average distance traveled and average net velocity of the vesicles (Figure 7D, 7E). These results indicate that KIF17 contributes to the microtubule-based transport of Golgi-derived Kv1.5 vesicles in HL-1 myocytes.

21To examine whether KIF17-driven motility of Kv1.5 channels occurs along a subset of microtubules, retrospective immunofluorescence imaging was applied to HL-1 myocytes expressing Kv1.5-GFP. After live cell imaging (Video S7), the cells were fixed and stained with antibodies for total or acetylated tubulin. A comparison of the SD Map of Kv1.5-GFP vesicle motility to the image of the total microtubule population demonstrates that Kv1.5 vesicles move along microtubule tracks (Figure 7F, representative of 20 cells in six experiments) but only 13.9%±12.7% of the motility events occurred on microtubules marked by acetylated α-tubulin (Figure 7G, n = 3 cells in three experiments). Thus, the motility of Kv1.5 vesicles, like that of the KIF17 motor, does not occur preferentially on stable microtubules marked by acetylation.

22That eukaryotic cells contain heterogeneous microtubule populations is widely appreciated. Yet how microtubule diversity is translated into specific microtubule functions is not clear. We show that visualization of single molecules under native physiological conditions reveals new information about how cytoskeletal components interact with each other. Specifically, we show that kinesin motors can translate microtubule diversity into a functional segregation of secretory cargoes.

23Our results demonstrate a new property that distinguishes kinesin families—the ability to respond to microtubule heterogeneity in cells. We show that individual Kinesin-1 motors undergo preferential motility along stable microtubules marked by PTMs whereas individual Kinesin-2 (KIF17) and Kinesin-3 (KIF1A) motors are not selective as they undergo motility on both dynamic and stable microtubules.

24Transport along stable microtubules would prevent the undesirable situation where a dynamic microtubule track “disappears” under Kinesin-1 and its associated cargoes. Why then do Kinesin-2 and Kinesin-3 motors not avoid dynamic microtubules? One possibility suggested by our live cell imaging of single KIF1A and KIF17 motors (Figures 4 and 5) is that, upon overtaking the plus end of the microtubule, these motors do not dissociate but rather remain at the plus ends of growing microtubules. This may be an important requirement for motors whose cargoes function at the interface of microtubule plus ends and the cell cortex. Interestingly, computer simulations and cell staining have suggested that Kinesin-6 motors may remain attached rather than fall off upon reaching the tips of dynamic microtubules,. A second possibility is that these motors and/or their cargoes can prevent depolymerization of the microtubule track. For example, in yeast, transport of +TIP proteins can prevent microtubules from depolymerizing under minus end-directed kinesin motors –.

25Differences have been reported between kinesin motors in their transport direction in neuronal cells. Kinesin-1 motors accumulate at the tips of axons whereas Kinesin-2 and Kinesin-3 motors accumulate in both axonal and dendritic compartments,. Recent work suggests that track selectivity is likely related to the ability to undergo polarized transport. In this work, substitution of residues in the Kinesin-1 motor domain with the corresponding residues of the KIF1A motor domain resulted in a motor that could no longer distinguish between tyrosinated and detyrosinated microtubules in vitro and could not undergo polarized transport to axons in vivo. Thus, a selective motor was converted into a non-selective motor by mutation of the microtubule-binding surface. Together with our work on imaging single motors, these data support the hypothesis that track selection is an inherent property of the motor-microtubule interaction.

26What biochemical cues enable Kinesin-1 motors to distinguish microtubule populations? Our results show that Kinesin-1 selects stable microtubules marked by detyrosination and acetylation for preferential motility. One possibility is that it is the PTMs themselves that influence Kinesin-1. This possibility has gained support from recent work in vitro and in vivo –. However, the recognition of PTMs by the Kinesin-1 motor is likely to be complex as mutation of motor surface abolished the ability of Kinesin-1 to recognize detyrosinated but not acetylated microtubules. While glutamylation could play an important role in guiding motor-based transport in epithelial and neuronal cells,–, it is not likely to play a critical role in fibroblasts. Recent work has indicated that the situation may be different for fungal motors as Kinesin-3 motors, but not Kinesin-1 motors, show track selectivity. Thus, how different PTMs create a tubulin code that can guide motor protein transport events is an important area for future studies.

27The correlation of Kinesin-1 motility with specific PTMs does not rule out the possibility that other microtubule-based mechanisms influence this or other motors. Structural changes that occur in the microtubule lattice after polymerization and/or stabilization may influence kinesin motors. Also, MAPs that stabilize microtubules have been shown to negatively influence kinesin-based transport events in cells. Thus, it may be that the modifications that occur along stable microtubules serve to decrease binding of MAPs to microtubules and thus clear the way for motor-based transport.

28Our work provides the first demonstration that transport events can be segregated between stable and dynamic microtubules via kinesin motors that select these subpopulations of microtubule tracks. Stable microtubules are critical for morphogenesis during diverse biological events such as cytokinesis, cell motility, and neuronal polarity,,,. The preferential motility of Kinesin-1 along stable microtubules may serve to direct Kinesin-1 transport during morphogenesis and maintenance of polarity in neuronal and epithelial cells,,,,,. A transport module comprised of cargo/Kinesin-1/microtubule subsets is likely involved in other polarized trafficking events such as the delivery of mRNA complexes to the vegetal pole in Xenopus oocytes.

29Dynamic microtubules are important for microtubule search and capture events in mitotic cells as well as in interphase cells during cell polarity and motility,,,. Our results imply that dynamic microtubules can serve as tracks for kinesin-based transport of cargoes that likely function at the microtubule-cortex interface. Indeed, a transport module that comprises cargo/Kinesin-2,3/dynamic microtubule components may be important in trafficking of connexins, cadherins, +TIPs, and channels to cell-cell junctions –. This transport module may also influence retrograde trafficking events such as cytoplasmic dynein-driven movement of endoplasmic reticulum (ER)-derived vesicles to the central Golgi complex.

30This work is the first to analyze the segregation of kinesin motors and their cargoes to distinct microtubules populations and subcellular destinations. Recent work using TIRF microscopy of detergent-extracted cells has indicated that unconventional myosins also differ in their ability to select actin filament tracks. Thus, these types of experiments provide a starting point for exploring the ability of motor proteins to respond to structural and/or biochemical changes in cytoskeletal filaments inside cells as well as the relationship between track selection and cellular function.

31COS and HL-1 cells were cultured and transfected as described,. The following antibodies were purchased: total β-tubulin (E7, Developmental Studies Hybridoma Bank, Univ. Iowa), acetylated α-tubulin (Sigma B-11-61), detyrosinated α-tubulin (Chemicon), GFP (Invitrogen), and fluorescently marked secondary antibodies (Jackson ImmunoResearch). A monoclonal antibody to polyglutamylated tubulin (GT335) was a gift from C Janke (CNRS, France). A polyclonal antibody that recognizes acetylated α-tubulin was generated against amino acids 29–52.

32Constitutively active versions of kinesin motors were generated by PCR cloning of the relevant sequences (aa 1-490 of human KIF17, aa 1-393 of rat KIF1A, and aa 1-560 of rat KIF5C) into the 3xmCit-N1 vector. DN versions of KHC (aa 566-955) and KIF17 (aa 488-846) were cloned into mCherry-C1. All plasmids were verified by DNA sequencing. Plasmids encoding VSVG-GFP and EB3-mCherry were gifts from A Akhmanova and N Galjart (U Rotterdam). Kv1.5-GFP has been described.

33Objective-based TIRF microscopy was carried out as described. Briefly, transfected COS cells on a glass-bottomed 35 mm dish (MatTek) were carefully rinsed with Ringers buffer (10 mM HEPES/KOH, 155 mM NaCl, 5 mM KCl, 2 mM CaCl2, 1 mM MgCl2, 2 mM NaH2PO4, 10 mM glucose, pH 7.2) and imaged on a custom-modified Zeiss Axiovert 135TV microscope equipped with a 1.45 NA a-Plan Fluor objective, 2.5× optovar, 505DCXR dichroic and HQ510LP emission filter (Chroma Technology), 488 nm line of a tunable single-mode fiber-coupled Argon Ion Laser (Melles Griot), and a back-illuminated EMCCD camera (Cascade 512B, Roper Scientific). The angle of illumination was adjusted for maximum penetration of the evanescent field into the cell (near-TIRF), enabling an imaging depth of ∼500 nm, which is sufficient to image nearly all the microtubules in the periphery of flat COS cells. For two-color TIRF, a yellow diode pumped solid-state laser (593 nm, CrystaLaser) was combined with the 488 nm laser using a dichroic mirror (Z488RDC). Fluorescence emissions were first passed though a FF495/605 dual-band dichroic mirror (Semrock) and then projected onto separate halves of the CCD camera by a Dualview beam-splitter (Optic Insights) equipped with a T585LP dichroic beam splitter and ET525/50M and HQ610LP emission filters (all Chroma Technology). In general, cells were imaged ∼6 h post-transfection to maintain low expression levels representative of endogenous protein behavior and optimal for single molecule TIRF imaging. Images were captured every 50–100 ms for 30–35 s. All experiments were carried out at room temperature (18–21°C). For measurement of EB3-mCherry microtubules, the lower average microtubule growth rate compared to those reported in other studies (e.g. and references therein) is likely due to this lower temperature.

34Immediately after imaging, cells were fixed (3 min) in 3.7% paraformaldehyde (Ted Pella), quenched (10 min) with 50 mM NH4Cl, and permeabilized (3 min) with 0.2% Triton X-100. An equal volume of 1.0% glutaraldehyde was then carefully added for an additional 7 min of fixation. After quenching with freshly prepared 1.5 mg/ml NaBH4, primary antibodies in 0.2% fish skin gelatin were incubated for 1–2 h at room temperature or overnight at 4°C.

35COS or HL-1 cells in glass-bottom dishes (MatTek) were imaged live on a Nikon TE2000 microscope with a Plan-APO 100×/NA 1.4 objective and Photometrics CS ES2 camera. VSVG(tsO45)-GFP-expressing cells were incubated overnight at 39°C to accumulate protein in the ER. Two h after shifting cells to 33°C for synchronous protein transport through the Golgi complex and to the plasma membrane, images were obtained every 1 s. For Kv1.5-GFP, cells were incubated at 19°C for 3 h to accumulate secretory proteins in the trans-Golgi and then shifted to 37°C and imaged every 5 s. Statistical analysis was done using one-way ANOVA analysis.

36Videos and images were prepared with ImageJ (NIH) and Photoshop and Illustrator (Adobe). Generation of the SD Maps is described in Figure S1 and. Home-made plug-ins for ImageJ were used for measuring the speed and run length of motors and vesicles. For motors, only diffraction-limited fluorescence spots (5×5 pixels) were selected for analysis that were clearly separated from the neighboring fluorescence and moved in a linear fashion on microtubules tracks identified in the SD Maps. Motile events that did not appear in the SD Map were usually short and/or blurry events that could not be separated from diffusion. For colocalization of motor and microtubule tracks (Table 1), the SD Map of the motility events was overlaid with a static image of the mCherry-tubulin fluorescence. The relative overlap (expressed in %) was calculated as the ratio of microtubules with motility events to total number of microtubules. For colocalization of VSVG-GFP or Kv1.5-GFP vesicles with acetylated tubulin, the SD Map of the vesicle motility was overlaid with the fixed acetylated tubulin image. The overlap (expressed in %) was calculated as the (number of vesicles on microtubules)/(number moving vesicles).

37Supplemental methods are described in Text S1.