The ability of the kinesin-2 heterodimer KIF3AC to navigate microtubule networks is provided by the KIF3A motor domain

Heterodimeric kinesin family member KIF3AC is a mammalian kinesin-2 that is highly expressed in the central nervous system and has been implicated in intracellular transport. KIF3AC is unusual in that the motility characteristics of KIF3C when expressed as a homodimer are exceeding slow, whereas homodimeric KIF3AA, as well as KIF3AC, have much faster ATPase kinetics and single molecule velocities. Heterodimeric KIF3AC and homodimeric KIF3AA and KIF3CC are processive, although the run length of KIF3AC exceeds that of KIF3AA and KIF3CC. KIF3C is of particular interest because it exhibits a signature 25-residue insert of glycine and serine residues in loop L11 of its motor domain, and this insert is not present in any other kinesin, suggesting that it confers specific properties to mammalian heterodimeric KIF3AC. To gain a better understanding of the mechanochemical potential of KIF3AC, we pursued a single molecule study to characterize the navigation ability of KIF3AC, KIF3AA, and KIF3CC when encountering microtubule intersections. The results show that all three motors exhibited a preference to remain on the same microtubule when approaching an intersection from the top microtubule, and the majority of track switches occurred from the bottom microtubule onto the top microtubule. Heterodimeric KIF3AC and homodimeric KIF3AA displayed a similar likelihood of switching tracks (36.1 and 32.3%, respectively). In contrast, KIF3CC detached at intersections (67.7%) rather than switch tracks. These results indicate that it is the properties of KIF3A that contribute largely to the ability of KIF3AC to switch microtubule tracks to navigate intersections.

Our single-molecule quantum dot (Qdot) 2 motility assays revealed that in the absence of load, KIF3AC is highly processive (run length ϭ 1.23 Ϯ 0.09 m) with its velocity at 169 Ϯ 5.6 nm/s (33). In contrast, the engineered homodimers were quite different from each other. Although the velocity of KIF3AA was quite fast at 239 Ϯ 4.2 nm/s with a run length of 0.98 Ϯ 0.09 m, the velocity of KIF3CC was exceedingly slow at 8.2 Ϯ 0.3 nm/s, and the run length decreased to 0.57 Ϯ 0.03 m. Similarly, the ATPase kinetics of KIF3AC and KIF3AA were significantly faster than the ATPase kinetics of KIF3CC (33)(34)(35).
We pursued single molecule experiments to quantify the behavior of KIF3AC at microtubule intersections compared with the performance of homodimeric KIF3AA and KIF3CC. Although this microtubule intersection geometry is quite simple, it asks whether the behavior of KIF3AC is driven by KIF3A or KIF3C or whether it is an emergent mechanochemical property of KIF3AC not predicted by the motile properties of either KIF3A or KIF3C. Moreover, the ability to navigate microtubule intersections and roadblocks on microtubules is a critical requirement for intracellular transport and has been tested for other processive kinesins (36 -52).
The results presented reveal that KIF3AC and KIF3AA were both capable of switching from one microtubule to another at a microtubule intersection. In contrast, KIF3CC detached at microtubule intersections rather than switch microtubules. The results support the interpretation that it is the mechanochemical properties of KIF3A that contribute largely to the ability of KIF3AC to switch microtubule tracks to navigate intersections and that the behavior of KIF3CC at microtubule intersections is more consistent with a role in microtubule plus end targeting and microtubule dynamics than a role as a cargo transporter.

Kinesin motors exhibit three possible behaviors when encountering microtubule intersections
We pursued single-molecule Qdot experiments using total internal reflection fluorescence (TIRF) microscopy to quantify the behavior of KIF3AC, KIF3AA, and KIF3CC at microtubule intersections composed of two microtubules (Fig. 1, A and B). The design of the truncated kinesins is shown in Fig. S1, and these motor constructs have been characterized previously (31, 33-35, 53, 54). Fig. 1B illustrates the three potential fates of the motor upon encountering a microtubule intersection. A motor can continue on the same microtubule track through the intersection (Fig. 1B, path 1), detach from the microtubule at the intersection (Fig. 1B, path 2), or turn onto the second intersecting microtubule (Fig. 1B, path 3). Pausing at the intersection point was also observed and scored. As shown in Fig. 1C, dualcolor composite kymographs provided a visual representation of motor runs on each microtubule as separate colors. Individual kymographs for each constituent microtubule were constructed: for one microtubule, motor runs were represented in red on the kymograph, and for the second microtubule of the intersection motor runs were represented in green. The red and green kymographs for the constituent microtubules were overlaid at the points along their x axes corresponding to the point at which the two microtubules intersected. This process generated a composite kymograph in which a red run color represents Qdot motion along one of the microtubules, whereas a green run color represents Qdot motion on the second microtubule. These composite kymographs were constructed for each intersection, as described under "Experimental procedures", to confirm Qdot behaviors observed in the experimental videos (Movies S1-S5). Qdot behavior at a given intersection was observed in the video and compared with the corresponding composite kymograph to confirm which of the three possible outcomes occurred in the motor's run. Fig. 1C provides examples of the three potential fates of a motor at the intersection. Qdots that continued on the same microtubule through the intersection showed the same color run on the kymograph before and after the intersection (Fig. 1C, panel 1). Qdots that detached at the intersection ended their runs at the intersection point on the kymograph x axis (Fig. 1C, panel 2). Qdots that switched microtubule tracks exhibited a change in run color after encountering the intersection (Fig. 1C, panel 3). The change in run color from red before the intersection point to green thereafter represents a track switch or a turn onto the intersecting microtubule. For all three outcomes, a motor that pauses at the microtubule intersection exhibits a yellow vertical line at the point of intersection along the x axis of the composite kymograph (Fig. 1C, panel 1). Because the Qdot is paused at the point of microtubule overlap, its run appears in both the red and green kymographs. Therefore, the overlaid, dual color kymograph shows the yellow pause. Qdots were also observed to be paused on microtubules but not at the intersection. These appear as either a red or green vertical line on the kymographs.

Differential fluorescence labeling allows a distinction between two intersecting microtubules based on their relative position
Two populations of fluorescent X-rhodamine-labeled microtubules were incorporated into the perfusion chamber for imaging. One microtubule population was diluted with unlabeled tubulin, as described under "Experimental procedures," yielding a microtubule population with a relatively low fluorescence intensity ("dim" microtubules). Upon TIRF imaging, these microtubules were distinct from the second population of microtubules, which was polymerized from undiluted X-rhodamine tubulin and appeared brighter upon imaging ("bright" microtubules). In these experiments, the dim microtubules were always introduced into the perfusion chamber first, followed by the addition of the bright microtubule population. This procedure yielded microtubule intersections with a dim microtubule on the bottom and a bright microtubule on top. The TIRF images shown in Figs. 2-4 and the corresponding videos (Movies S1-S4) demonstrate the contrasting fluorescence intensities of the bottom microtubule versus the top microtubule that form an intersection. This distinction allowed an analysis of whether the relative position of the microtubule from which a motor approaches an intersection has any impact on the outcome executed by the motor at the intersection.

KIF3AC navigation of microtubule intersections
green was assigned to the kymograph representing the bright (top) microtubule. The composite kymograph for these intersections therefore represented Qdot runs along the bottom microtubule in red and Qdot runs along the top microtubule in green.  Fig. 2D. The overall KIF3AC population (n ϭ 310) exhibited a 36.1% frequency of switching microtubule tracks at an intersection. This proportion was statistically significantly different (p Ͻ 0.001) from the 52.6% of the population, which continued straight at the intersection. As shown in pie chart 1 of Fig. 2 (C and D), the 11.3% probability of detaching at the inter- A motor that continues on the same microtubule has the same color run before and after the intersection (1). A motor that detaches will end its run at the intersection (2). A motor that switches microtubule tracks will exhibit a change in run color after encountering the intersection (3). A motor that pauses at the intersection under any of these circumstances exhibits a yellow vertical line at the intersection as illustrated in example 1 of C.

KIF3AC navigation of microtubule intersections
section was significantly different from the population, which switched tracks, as well as the population, which continued straight (p Ͻ Ͻ 0.001).
However, when the results were subdivided into KIF3AC motors that began on the top microtubule versus the bottom microtubule, distinct behavioral differences were observed. To determine the behavior of KIF3AC at intersections, microtubules of lower fluorescence were added first to the coverslip followed by perfusion of microtubules of higher fluorescence. Therefore, the behavior of KIF3AC at intersections could be assessed based on whether the run started on the top microtubule or the bottom microtubule track. For the dual color composite kymographs (left panels), the top microtubule was assigned the color green, and the bottom microtubule was assigned the color red, with the yellow vertical dotted line indicating the point of intersection of the two microtubules. A, a KIF3AC run initiated from the top microtubule. The left panel shows an example in which KIF3AC began its run on the top microtubule and at the intersection, and it remained on the top microtubule. The right panel shows sequential frames captured through TIRF microscopy with the illustrations below each of the frames. T corresponds to the top microtubule, and B corresponds to the bottom microtubule. The dual color kymograph (left panel) shows a lack of color change before and after the intersection, indicating that the motor remained on the same microtubule track. B, a KIF3AC run initiated from the bottom microtubule. The left panel shows the dual color kymograph, which indicates that this KIF3AC run began on the bottom microtubule (red) and at the intersection switched to the top microtubule (green). The right panel presents the TIRF microscopy frames with the corresponding illustrations below. These motor runs can be viewed in the corresponding video (Movie S1). C and D, individual KIF3AC motors were scored based on the three behaviors at microtubule intersections, and the results are shown as either a pie chart or table. Pause durations at intersections are reported as means Ϯ S.E.

KIF3AC navigation of microtubule intersections
KIF3AC motors that started on the top microtubule continued straight through the intersection in 71.8% of observations, compared with 34.8% of motors that continued straight when starting on the bottom microtubule (p Ͻ Ͻ 0.001; Fig. 2, C, pie charts 2 and 3, and D). Motors that started on the bottom microtubule demonstrated a 57.1% tendency to switch tracks, as opposed to 13.4% of motors that started on the top microtubule (p Ͻ Ͻ 0.001; Fig. 2, C, pie charts 2 and 3, and D).
KIF3AC exhibits a statistically significant likelihood to continue straight through the intersection when approaching an intersection from the top microtubule ( Fig. 2A), as opposed to "stepping down" from the top microtubule to switch onto the bottom microtubule track. KIF3AC also exhibits a significant tendency to turn from the bottom microtubule onto the top microtubule (Fig. 2B), as opposed to negotiating such an intersection by maneuvering around the top microtubule to remain on its incident bottom microtubule. KIF3AC behavior at intersections is therefore influenced by the position of the microtubule upon which the motor initiates its run to approach the intersection. However, the likelihood of detaching at the intersection point appeared to be independent of the microtubule position, remaining relatively consistent whether the Qdot run began on the top (14.8%) or bottom (8.1%) microtubule (p ϭ 0.0929; Fig. 2, C, pie charts 2 and 3, and D).
Furthermore, KIF3AC exhibited an overall average pause duration of 7.8 Ϯ 0.7 s at microtubule intersections (n ϭ 100; Fig. 2D). A significant difference was observed (p ϭ 0.00436) between the mean pause durations for runs beginning on the top (9.6 Ϯ 1.3 s) versus those initiated on the bottom microtubule (5.7 Ϯ 0.5 s; Fig. 2D). Mean pause durations prior to detaching at the intersection were longer than the overall average, and this trend was observed regardless of whether the run initiated on the top or bottom microtubule (Fig. 2D). However, the longer mean pause durations and standard errors reported may be due to the extremely low number of observations (n ϭ 12) for pause durations prior to detaching. Furthermore, KIF3AA demonstrated no statistically significant difference in behavior from KIF3AC when divided into motor populations that initiated runs on the top microtubule versus those that began runs on the bottom microtubule. Similar to KIF3AC, KIF3AA exhibited a 76.7% probability of continuing straight through the intersection when starting on the top microtubule (p ϭ 0.408) and a 42.5% probability of continuing straight when starting on the bottom microtubule (p ϭ 0.192; Fig. 3C, pie charts 2 and 3). Additionally, like KIF3AC, KIF3AA demonstrated a 49.4% tendency to switch tracks when starting on the bottom microtubule (p ϭ 0.199) and a 14.0% likelihood of turning when starting on the top microtubule (p ϭ 1.0; Fig. 3, C, pie charts 2 and 3, and D). Furthermore, KIF3AA demonstrated no statistically significant difference (p ϭ 0.861) in the probability of detaching at the intersection, regardless of whether the run began on the top or bottom microtubule, and the motor detached with a likelihood of 9.3 and 8.1%, respectively ( Fig. 3, C, pie charts 2 and 3, and D). This consistent frequency of detachment despite the incident microtubule position is an additional common feature observed in both KIF3AC and KIF3AA (see Fig. 6 also). Fig. 3 (A and B) shows dual color kymographs and corresponding TIRF images that illustrate KIF3AA behavior (see also Movie S2). KIF3AA exhibits a preference for continuing straight when approaching an intersection from the top microtubule (Fig. 3A), as well as a tendency to turn from the bottom microtubule onto the top microtubule (Fig. 3B). The behavior of KIF3AA at intersections is consistent with the observed behavior of KIF3AC, with both motors impacted by the orientation position of the microtubule from which the motor approaches the intersection.

KIF3AA exhibits a similar behavioral pattern as KIF3AC at microtubule intersections
Pause durations for KIF3AA, with an overall average of 2.5 Ϯ 0.1 s (n ϭ 54), were significantly shorter than those exhibited by KIF3AC at 7.8 Ϯ 0.1 s (p Ͻ Ͻ 0.001; Fig. 3D). There was no significant difference between the overall mean pause duration and the mean pause duration prior to detaching, nor was there a significant difference between mean pause durations at intersections when starting on the top versus the bottom microtubule (Fig. 3D). However, the extremely low numbers of observed pauses prior to detaching (n ϭ 3) and when starting on the top microtubule (n ϭ 14) render statistical tests between these subpopulations unreliable. Rather, the KIF3CC population exhibited a 67.7% likelihood of detaching from its associated microtubule at the intersection point, which was significantly greater than the likelihood of KIF3AC and KIF3AA detaching at intersections (p Ͻ Ͻ 0.001, for both comparisons).

KIF3CC exhibits unique tendency to detach at microtubule intersections
KIF3CC exhibited a high probability of detachments at the intersection point, and this was observed regardless of whether the run was initiated on the top or on the bottom microtubule. Fig. 4 (C, pie charts 2 and 3, and D) shows this trend, with KIF3CC exhibiting a 62.5% probability of detaching when starting on the top microtubule and a similar 73.3% probability of detaching when starting on the bottom microtubule (p ϭ 0.0552). Furthermore, the likelihood of KIF3CC to detach is relatively consistent regardless of whether the run was initiated on the top or bottom microtubule.

KIF3AC navigation of microtubule intersections
Additionally, there was a statistically significant difference between the proportion of the KIF3CC population (33.8%), which continued straight when starting on the top microtubule, and the proportion (16.0%), which continued straight when starting on the bottom microtubule (p Ͻ 0.001; Fig. 4, C, pie charts 2 and 3, and D). KIF3CC, therefore, displayed the same tendency exhibited by KIF3AC and KIF3AA to continue straight when approaching an intersection on the top microtubule. A, the right panel shows sequential frames captured through TIRF microscopy with a motor initiating a run (arrowhead) on the top (bright) microtubule with illustrations shown beneath. The corresponding dual color kymograph (left) shows a lack of color change before and after the intersection, indicating that the motor did not switch to the intersecting microtubule. A second motor on the top microtubule, indicated by an asterisk, leaves the field of view of the TIRF frames and therefore was not scored. This motor corresponds to the green run on the kymograph marked with an asterisk. A third motor, marked with an asterisk and not scored, runs along the bottom microtubule, and the corresponding red run in the kymograph is marked with an asterisk. These motor runs can be viewed in the corresponding video (Movie S2). B, the dual color kymograph (left panel) indicates that this motor began on the bottom microtubule (red) and switched to the top microtubule at the intersection (green). TIRF microscopy frames and representative illustrations show a motor initiating a run on the bottom microtubule (right panel; see Movie S2 also. C and D, the behavior of the KIF3AA motors at microtubule intersections were scored and presented as pie charts or the table. Pause durations at microtubule intersections are reported as means Ϯ S.E.

KIF3AC navigation of microtubule intersections
KIF3CC also exhibited the longest pause durations of all three kinesin-2 motors tested, with an overall average pause duration of 126.4 Ϯ 11.3 s (n ϭ 67; Fig. 4D). Unlike the outcomes observed in KIF3AC, statistical tests for KIF3CC showed no significant difference between the mean pause duration for runs initiated on the top microtubule and the mean pause duration for KIF3CC runs initiated on the bottom microtubule (p ϭ 0.264; Fig. 4D). For KIF3CC, the likelihood of pausing appeared to be highest prior to detaching at the intersection (Fig. 4D). However, the small number of observed pauses before switch- A, on the right panel, sequential frames captured through TIRF microscopy show a KIF3CC motor (arrowhead) initiating a run on the top (bright) microtubule with illustrations shown beneath. On the left, the corresponding dual color kymograph shows that the run (green) ended at the intersection point. The short red run on the kymograph, marked with an asterisk, indicates a second motor moving along the bottom microtubule. This run is not seen in the TIRF frames on the right but can be observed in the corresponding video (Movie S3). B, in the right panel, TIRF microscopy frames and representative illustrations show a KIF3CC motor (arrowhead) initiating a run on the bottom microtubule. On the left, the corresponding dual color kymograph shows the run (red) ends at the intersection point. A second motor on the top microtubule is indicated with an asterisk, and the corresponding green run on the kymograph is marked with an asterisk. The red run in the kymograph marked with an asterisk indicates an additional motor running along the bottom microtubule. This run is not seen in the TIRF frames on the right but can be observed in the corresponding video (Movie S4). C and D, the KIF3CC behaviors at microtubule intersections were scored and shown in pie charts and the table. Pause durations at intersections are reported as means Ϯ S.E.

KIF3AC navigation of microtubule intersections
ing tracks or continuing straight through an intersection on the same microtubule renders statistical comparisons between pausing subpopulations unreliable.

Motor behaviors are independent of the angle of intersection
Intersections comprised of microtubules at an array of acute and obtuse angles were selected for analysis. The angle formed by the intersecting microtubules was tabulated to examine the impact of the angle of incidence upon motor behavior at the intersection. "Acute" intersections were defined as those in which the intersecting microtubules formed an angle less than 90°. "Obtuse" intersections were defined as those in which the intersecting microtubules formed an angle greater than 90°. For these experiments, we included a similarly designed kinesin-1, KIF5B K439 as a control (54).
For each of the four motors, the subpopulations that executed a microtubule track switch were further divided based on the angle created by the intersecting microtubules (Fig. 5,  A-D). For each motor population, the number of total events at acute and obtuse intersections were tabulated, as well as the number of switching events at each intersection type. These data were used to calculate the percentage of track switching events at acute intersections versus at obtuse intersections, which enabled direct comparison of KIF3AC, KIF3AA, and KIF3CC with kinesin-1 KIF5B (Fig. 5E; see also Fig. S2 and Movie S5).
Each of the four motor populations failed to exhibit a statistically significant difference in the probability of switching tracks, whether the intersecting microtubules lay at acute or obtuse angles. KIF3AC demonstrated a 38.9% likelihood of switching at acute intersections and a 34.2% likelihood of switching at obtuse intersections (p ϭ 0.473; Fig. 5B). KIF3AA demonstrated a 29.8% likelihood of switching at acute intersections and a 35.2% likelihood of switching at obtuse intersections (p ϭ 0.368; Fig. 5C). KIF3CC demonstrated a 7.0% likelihood of switching at acute intersections and a 7.2% likelihood of switching at obtuse intersections (p ϭ 1.0; Fig. 5D). Similar to the kinesin-2 motors, kinesin-1 KIF5B failed to demonstrate a statistically significant difference in the probability of track switching, with a 14.5% likelihood of switching at acute intersections and a 22.5% likelihood of switching at obtuse intersections (p ϭ 0.141; Fig. 5A).
This consistency across all four motor types indicates that the probability of track switching at intersections is independent of the angle formed by the two intersecting microtubules with a nearly equal likelihood of turning at acute and obtuse intersections. Therefore, the angle of incidence appears to be a characteristic of the intersection yet not a significant influence on the motor's behavior at the intersection. Note, however, that these results are in contrast to those reported in other studies in which multiple motors or different motors were associated with the cargo or the motors were performing under load (42,45,48).

Navigation abilities appear to be influenced by the specific kinesin motor domain
The results presented show that although KIF3AC and KIF5B are both processive kinesins, they differ in their ability to navigate microtubule intersections, even though both transport cargoes in neurons. KIF5B tends to do so by continuing on the same microtubule track (Fig. S2), whereas KIF3AC has a greater capacity to switch tracks to navigate obstacles posed by intersecting microtubules (Fig. 2). These distinct behaviors represent two mechanisms to successfully transport cargo to their cellular destination.
Among the kinesin-2 motors examined, KIF3AC and KIF3AA exhibited the greatest tendency to execute a track switch at a microtubule intersection and the lowest likelihood of detaching at the intersection point. The strong similarities between the single molecule results for KIF3AC and KIF3AA and not KIF3CC suggests that the track-switching capability is encoded by the motor domain and that it is the KIF3A motor domain that enables KIF3AC to switch microtubules at intersections (Figs. 2-4 and 6). The results for KIF5B also reinforce the hypothesis that it is the specific motor domain that encodes the track switching behavior with KIF5B exhibiting a very low percentage of track switch events at intersections (Fig. 5 and Fig. S2) In contrast, dimerization of the KIF3C polypeptide results in a homodimer that lacks the same ability as KIF3AC and KIF3AA to switch microtubule tracks at intersections. KIF3CC predominantly detached at microtubule intersections and exhibited substantially longer mean pause durations at the intersection point than either KIF3AA or KIF3AC. These single-molecule results indicate a tendency of KIF3CC to stall its motion and end its run at the more restrictive geometry of a microtubule intersection. These motility properties would render KIF3CC an inefficient motor for long-range cargo transport in a cellular environment, yet these properties are consistent with a role in promoting microtubule dynamics at the

KIF3AC navigation of microtubule intersections
microtubule plus end as indicated by previous studies (13,30,31). Guzik-Lendrum et al. (31) reported that KIF3CC, but not KIF3AC, was targeted to the microtubule plus ends, acted as a potent catastrophe factor, and modulated the tapered structure typically seen in growing dynamic microtubules. Moreover, this activity required ATP turnover, and KIF3CC tended to pause at the microtubule tip prior to catastrophe. In these microtubule dynamics assays, KIF3CC maintained the ability to collide with the microtubule and step processively, but the velocity was extremely slow and the microtubule plus end events resulted from KIF3CC association at the microtubule plus end rather than KIF3CC stepping to the microtubule plus end. These combined results support the interpretation that the mechanochemical capability of KIF3C is very different from that of KIF3A.

Navigation abilities, across all motors, appear to be similarly influenced by the structure of a microtubule intersection
The single molecule results also indicate that kinesin-2 KIF3AC, KIFAA, and KIFCC share common behaviors at intersections (Fig. 6), suggesting that the geometry or structure of the intersection plays a role in the navigation ability of the motor. In each case, if the approach to the microtubule intersection is from the top microtubule, all three kinesins tend to continue through the intersection on the top microtubule. However, if the motor is approaching the intersection from the bottom microtubule, it is less likely to continue on the bottom microtubule through the intersection. Rather, each motor that approached the microtubule intersection from the bottom microtubule exhibited a significantly higher probability to switch or turn onto the upper microtubule. These behaviors were also consistent for KIF3CC, although it predominantly detached at microtubule intersections.
The influence of the microtubule intersection on a motor's behavior can be explained at the level of microtubule protofilaments. The majority of protofilaments of the top microtubule do not contact the bottom microtubule. Therefore, it is much more likely that a motor will continue stepping on the top microtubule because of the low probability of contacting the bottom microtubule for a track change at the microtubule intersection. Thus, a motor's approach to the intersection has a large influence on how the motor will navigate the intersection.
This study has also revealed a profound difference in the behavior of KIF3AC and KIF3AA compared with homodimeric KIF3CC at microtubule intersections. A recent report by Bensel et al. (55) provides insight. The authors showed that in the absence of load, KIF3CC homodimers moved much slower than KIF3AA and KIF3AC, and both KIF3CC and KIF3AC backstep as single back steps more frequently than observed for KIF3AA. Moreover, both KIF3A and an assisting load accelerate the stepping kinetics of KIF3C in a strain-dependent manner. The results from Bensel et al. suggest that it is the decreased interhead tension within KIF3CC that decreases its run length and its ability to switch tracks at microtubule intersections. Rather, KIF3CC detaches at microtubule intersections, which is consistent with its role as a catastrophe factor that is targeted to the microtubule plus end to promote microtubule dynamics after injury. Moreover, the results of Bensel et al. support the argument presented here that the ability to navigate microtubule intersections is encoded in the catalytic motor domain of KIF3A rather than KIF3C. In conclusion, the results presented emphasize unique adaptations that have occurred to optimize kinesin motors for specific cellular functions.

Kinesin-1 K439 plasmid design and expression
The expression construct for human kinesin-1 K439 was a custom gene synthesis from GenScript and inserted into the pET24d expression vector between NcoI and BamHI restriction sites as reported previously (54). This construct encodes the first 439 residues of Homo sapiens KIF5B, followed by the C-terminally fused end binding protein 1 (EB1) dimerization motif (bold type) whose sequence matched the coiled-coil registry of the helix to generate stable dimers. A TEV protease site (italicized type) followed with linker residues (plain font) and His 8 tag (underlined type): Hs KIF5B K439 (Met 1 -Ile 439 ), DFYFGKLRNIELICQENEGENDPVLQRIVDILYATDETT-SENLYFQGASHHHHHHHH (predicted MW ϭ 56,538).
Escherichia coli BL21-CodonPlus (DE3)-RIL cells (Stratagene) were transformed with the K439 plasmid followed by selection on lysogeny broth (LB) plates containing 50 g/ml kanamycin (selection for K439 plasmid) plus 10 g/ml chloramphenicol. Colonies were selected and grown in LB liquid culture containing the same selective antibiotics at 37°C until they reached A 600 of 0.4 -0.5. The cultures were chilled in an ice bath to lower the temperature to 16°C (ϳ30 min). Subse-

KIF3AC navigation of microtubule intersections
quently, 0.1 mM isopropyl ␤-D-1-thiogalactopyranoside was added dropwise to induce protein expression with continued shaking at 16°C at 185 rpm for ϳ15 h. The cells were collected by centrifugation and resuspended in lysis buffer at 10 ml/g of cells. Lysis buffer contained 10 mM sodium phosphate buffer, pH 7.2, 300 mM NaCl, 2 mM MgCl 2 , 0.1 mM EGTA, 0.02 mM ATP, 1 mM DTT, 10 mM phenylmethylsulfonyl fluoride, and 30 mM imidazole.

Kinesin-2 KIF3 plasmid design and expression
The Mus musculus KIF3A and KIF3C plasmids for KIF3AC, KIF3AA, and KIF3CC protein expression were described in detail previously (33). Briefly, each expression plasmid encoded the native N-terminal motor domain, neck linker, helix ␣7 to initiate coiled-coil formation, followed by the EB1 dimerization motif (bold type) that matched the coiled-coil registry of the native ␣7 helix, followed by the TEV protease site (italicized type) with linker residues (plain font) and affinity purification tags (underlined type). For KIF3AC, the use of the StrepII tag on KIF3A combined with a His 8 tag on KIF3C allowed for sequential purification steps to generate pure, stable heterodimeric KIF3AC: KIF3A (Met 1 -Leu 374 ), DFYFGKLRNIELICQENEG-ENDPVLQRIVDILYATDETTSENLYFQGASNWSHPQFEK (predicted MW ϭ 48,559). KIF3C when expressed for heterodimeric KIF3AC or homodimeric KIF3CC includes the native N-terminal motor domain, neck linker, and helix ␣7, followed by the EB1 dimerization motif, and the C-terminal TEV protease cleavable His 8 tag: KIF3C (Met 1 -Leu 396 ), DFYFGKLRNI-ELICQENEGENDPVLQRIVDILYATDETTSENLYFQGAS-HHHHHHHH (predicted MW ϭ 49,759). Expression of homodimeric KIF3AA is similarly designed with the KIF3A motor domain: (Met 1 -Leu 374 ), DFYFGKLRNIELICQENEG-ENDPVLQRIVDILYATDETTSENLYFQGASHHHHHHHH (predicted MW ϭ 48,502) KIF3AC, KIF3AA, and KIF3CC were expressed in E. coli BL21-CodonPlus (DE3)-RIL cells (33)(34)(35)53). KIF3AC heterodimers resulted from cotransformation of the KIF3A and KIF3C plasmids each with different antibiotic resistance (KIF3A, ampicillin; and KIF3C, kanamycin). Transformed cells were selected on LB plates containing 100 g/ml ampicillin, 50 g/ml kanamycin, and 10 g/ml chloramphenicol. For homodimeric KIF3AA and KIF3CC, each plasmid was selected by 50 g/ml kanamycin and 10 g/ml chloramphenicol. The colonies were selected and grown in LB liquid culture in the presence of the antibiotics at 37°C until they reached A 600 of 0.4 -0.5. Cultures were chilled in an ice bath to 16°C, and then expression was induced by dropwise addition of 0.1 mM isopropyl ␤-D-1-thiogalactopyranoside. Incubation continued at 16°C with shaking at 185 rpm for ϳ15 h. The cells were harvested by centrifugation and resuspended in lysis buffer at 10 ml/g cells. The homodimers of KIF3AA and KIF3CC were expressed following the same protocol with the appropriate antibiotic selection. The construct design for each of the expressed motor proteins is illustrated in Fig. S1.

Purification of kinesin dimers
The purification strategy for these motors has been described previously (33, 34, 53, 54) with modifications included here. An aliquot of the cell suspension (ϳ10 g/100 ml lysis buffer plus 0.1 mg/ml lysozyme) was incubated with gentle stirring in an ice bath for 45 min. Cell lysis was achieved by three replicates of freezing in liquid N 2 and thawing in a 37°C water bath. Subsequently, the lysate was clarified through ultracentrifugation and applied to a HisTrap TM FF Ni 2ϩ -NTA column (GE Healthcare) that was pre-equilibrated with Ni 2ϩ -NTA binding buffer (20 mM sodium phosphate buffer, pH 7.2, 300 mM NaCl, 2 mM MgCl 2 , 0.1 mM EGTA, 1 mM DTT, 0.02 mM ATP, and 30 mM imidazole). After loading, the column was washed with Ni 2ϩ -NTA-binding buffer until the absorbance returned to baseline. The His 8 -tagged kinesins were eluted with a linear gradient (30 -300 mM imidazole in Ni 2ϩ -NTA binding buffer, pH 7.2. The fractions enriched in the specific His 8tagged kinesin were identified through SDS-PAGE. For KIF5B, KIF3AA, and KIF3CC, these fractions were pooled, concentrated, and dialyzed in 20 mM HEPES, pH 7.2, with KOH, 0.1 mM EDTA, 0.1 mM EGTA, 5 mM magnesium acetate, 50 mM potassium acetate, 1 mM DTT plus NaCl in sequential buffers containing 300, 200, and 100 mM NaCl. The final dialysis buffer included 100 mM NaCl plus 5% sucrose (w/v).
For KIF3AC, the fractions from the HisTrap TM FF Ni 2ϩ -NTA column that were enriched in both KIF3A with the StrepII tag and KIF3C with the His 8 tag were identified through SDS-PAGE, pooled, and transferred directly to the StrepTrap TM HP column (GE Healthcare), which was pre-equilibrated with Stre-pII column buffer: 20 mM sodium phosphate buffer, pH 7.2, 300 mM NaCl, 2 mM MgCl 2 , 0.1 mM EGTA, 1 mM DTT, and 0.02 mM ATP. The loaded StrepTrap TM HP column was washed with excess buffer to return to baseline absorbance, followed by elution with StrepII column buffer plus 2.5 mM desthiobiotin. Fractions were analyzed by SDS-PAGE to identify fractions that contained a 1:1 ratio of KIF3A and KIF3C polypeptides. These fractions were then pooled, concentrated, and dialyzed at 4°C in 20 mM sodium phosphate buffer, pH 7.2, 2 mM MgCl 2 , 0.1 mM EGTA, and 1 mM DTT plus NaCl in sequential dialysis steps: 200 mM NaCl, 100 mM NaCl, and 100 mM NaCl plus 5% sucrose (w/v).
The proteins were concentrated, clarified by ultracentrifugation and aliquoted prior to freezing in liquid N 2 and storage at Ϫ80°C. The purity of all preparations was confirmed by both analytical gel filtration on an HPLC gel filtration column (Superose TM 10/300; GE Healthcare Life Sciences) and SDS-PAGE. Note that the purification tags were not cleaved for the experiments reported herein. Typical yields for these purified motors were 1.7-3 mg/g of E. coli. Predicted molecular weights of dimeric motors based on amino acid residue sequence are as follows: KIF5B at 113,076; KIF3AC at 98,318; KIF3AA at 97,004; and KIF3CC at 99,518.
Before each experiment, the purified motors were clarified through high-speed centrifugation for 5 min at 4°C (Beckman Coulter Optima TLX Ultracentrifuge, TLA-100 rotor, 313,000 ϫ g). Motor concentration was determined using the Bio-Rad protein assay with IgG as the protein standard.

KIF3AC navigation of microtubule intersections
nylated-Penta-His antibody (Qiagen) for 60 min in PME80 buffer (80 mM PIPES, pH 6.9, with KOH, 5 mM MgCl 2 , and 1 mM EGTA) at room temperature (33,53). Subsequently, His 8tagged kinesin motors were added to the Qdot-antibody complex to a final concentration of 20 nM kinesin dimer. This mixture results in a 1:10 ratio of kinesin dimer to Qdot complex (20 nM kinesin dimer, 200 nM Qdot-antibody complex). Based on a Poisson distribution, this procedure yields a working stock in which 9% of the Qdots were estimated to have one motor bound and 0.5% of the Qdots with at least two motors bound. Additional single molecule experiments at 1:20 ratio of kinesin dimer to Qdot complex showed that there was no difference in the run length or velocity at 1:10 or at a 1:20 ratio of kinesin dimer to Qdots, thereby providing additional evidence that the experiments at a 1:10 ratio of kinesin dimer to Qdots were at single molecule conditions (33).

Microtubule preparation
Microtubules of different fluorescence intensities were prepared using X-rhodamine tubulin (Cytoskeleton, Inc.). The microtubules of higher fluorescence intensity were prepared with X-rhodamine tubulin and polymerized to yield microtubule tracks of ϳ13-30 m for the experiments. The X-rhodamine tubulin was resuspended in PME80 buffer containing 10% glycerol, incubated on ice with 1 mM MgGTP for 5 min, and then centrifuged at 16,000 ϫ g for 10 min at 4°C (Galaxy 16D Micro 1816 centrifuge, VWR). The supernatant containing the soluble tubulin was transferred to a new tube, and microtubule polymerization was initiated at 37°C for 25 min. The microtubules were stabilized with paclitaxel (33 M final concentration) and were incubated for an additional 10 min at 37°C, yielding a final stock of highly fluorescent microtubules at 30 M tubulin polymer.
To generate microtubules of lower fluorescence intensity, both X-rhodamine tubulin and bovine brain tubulin were combined to yield a 1:10 ratio of X-rhodamine tubulin to unlabeled tubulin. The bovine brain tubulin was mixed with PME80 buffer containing 10% glycerol and incubated on ice with 1 mM MgGTP for 5 min, followed by centrifugation at 16,000 ϫ g for 10 min at 4°C. The supernatant containing soluble unlabeled tubulin was transferred to a second tube. In a separate tube, 9 l of unlabeled bovine brain tubulin was mixed with 1 l of X-rhodamine tubulin, and polymerization was initiated at 37°C for 25 min. These microtubules were stabilized with paclitaxel (33 M final concentration) and were incubated for an additional 10 min, resulting in a final microtubule stock of low-fluorescence intensity microtubules at 30 M tubulin polymer. Note that control experiments showed that the relative velocity and run length using either population of fluorescent microtubules were not statistically significantly different.

Microtubule intersection assay
Perfusion chambers were formed by mounting a silanized coverslip on a glass slide separated by strips of double-sided tape to generate 10 l of flow cells (33,53). The chamber was incubated with 0.4% rat anti-␣-tubulin antibody (ABD Serotec) for 5 min, followed by a 5-min incubation with 5% Pluronic F-127 (Sigma-Aldrich) to block the surface. A dilution of the lower-fluorescence intensity microtubule stock (1:200 in PME80 supplemented with paclitaxel to 22 M) was introduced into the chamber and incubated for 6 min. PME80 with 10 mM DTT and 20 M paclitaxel was then flowed into the chamber to remove unbound microtubules. A dilution of the brighter-fluorescence intensity microtubule stock (1:800 in PME80 supplemented with paclitaxel to 22 M) was introduced into the chamber and incubated for 6 min, followed by perfusion of PME80 with 10 mM DTT and 20 M paclitaxel into the chamber to remove unbound microtubules. The working stock of Qdotmotor complexes (complex of 20 nM kinesin dimer, 200 nM Qdot, and antibody) was diluted in activity buffer (PME80, 0.5% Pluronic F-127, 30 M paclitaxel, 125 g/ml BSA, 50 M DTT, 25 mM glucose, 0.2 mg/ml glucose oxidase, 175 g/ml catalase, 0.3 mg/ml creatine phosphokinase, 2 mM phosphocreatine, 1 mM ATP) to achieve the final concentration in the perfusion chamber of 2 nM kinesin-Qdot complex. Note that only the high-fluorescence intensity microtubules were used for the kinesin-1 KIF5B control experiments (Fig. S2).

Total internal reflection fluorescence microscopy and image acquisition
Immediately upon addition of the kinesin-Qdot complexes, the chambers were imaged by TIRF microscopy using a Zeiss Inverted Axio Observer Z1 MOT fluorescence microscope with the 100ϫ oil 1.46 N.A. Plan-Apochromat objective (Carl Zeiss Microscopy) at 25°C. A Hamamatsu electron multiplier EM-CCD digital camera was used to collect digital images using the ZEN 2.3 software package. This setup yielded 512 ϫ 512pixel images with 0.16 m/pixel in both x and y planes. The kinesin-Qdot complexes were tracked by imaging at 488 nm (5% laser power) every ϳ0.6 s for 5 min using 150 ms of exposure. Video frames were collected for kinesin-2 KIF3AC and KIF3AA, as well as kinesin-1 KIF5B, at an acquisition rate of 1.7 Hz. Because of the extremely slow velocity of KIF3CC (33), videos were acquired over 30 min with 488-nm exposure (150 ms) at 5-s intervals, at an acquisition rate of 0.2 Hz.
Reference images of the X-rhodamine microtubule tracks were acquired at 561 nm (2% laser power) with a 180-ms exposure. These microtubule images were overlaid with the corresponding kinesin-Qdot videos using National Institutes of Health ImageJ software.

Data analysis
Analysis of the kinesin-Qdot motility was conducted using the MultipleKymograph plugin for ImageJ (J. Rietdorf and A. Seitz, European Molecular Biology Laboratory, Heidelberg, Germany). For a given intersection of two microtubules, a dual color composite kymograph was generated, enabling the visualization and tabulation of all kinesin-Qdots that encountered the intersection of interest.
For each of the two distinct microtubules that formed an intersection, a unique kymograph was generated. To create these individual kymographs, microtubule polarity was determined through observation of the kinesin-Qdot motility in the microscopy videos. Using the reference images of the microtubule tracks, the full length from the minus end to the plus end of a single microtubule was measured. The length from the minus

KIF3AC navigation of microtubule intersections
end to the intersection of interest was also measured. This provided a measurement of the distance along the kymograph x axis at which the microtubule intersected with the second microtubule. This process generated a kymograph depicting all kinesin-Qdot motion along a single microtubule, with the point of microtubule intersection marked on the x axis. This process was repeated for each of the two intersecting microtubules, resulting in two distinct kymographs.
The two single kymographs were differentiated by assigning the color red to one kymograph, and the color green to the other. In these experiments, the kymograph corresponding to the dim microtubule was always assigned the color red, whereas the kymograph corresponding to the bright microtubule was always assigned the color green. The distances along the x axis from the microtubule minus end to the point of intersection were compared among the two kymographs. The kymograph x axis lengths were then adjusted by adding the appropriate number of pixels to the ends of the shorter kymograph. This rendered the two kymographs equal in overall length and allowed them to be overlaid by aligning the kymographs at the x axis coordinates corresponding to the point of intersection. This process of merging the red and green kymographs generated a single dual color composite kymograph.
This composite kymograph provided a visual representation of all kinesin-Qdot runs that occurred at a particular microtubule intersection. The point of intersection between two microtubules could be clearly represented as a single coordinate along the x axis of the composite kymograph. Furthermore, on such a composite kymograph, kinesin-Qdot runs on the dim microtubule were represented in red and were distinguishable from runs on the bright microtubule, represented in green. This allowed a distinction between kinesin-Qdot runs initiated on the dim microtubule (the bottom microtubule in the intersection) versus runs initiated on the bright microtubule (the microtubule lying on top).
For the kinesin-1 KIF5B experiments, the same process was used to generate a dual color composite kymograph to represent a microtubule intersection. However, a single population of microtubules with the same fluorescence intensity was used, rendering the orientation of the intersecting microtubules unknown. Therefore, the single kymographs were assigned the colors red and green arbitrarily without regard to microtubule orientation.
For each intersection examined, a dual color kymograph was constructed and was used to score the behavior of all kinesin-Qdots as they encountered the intersection. The N value for each motor includes only kinesin-Qdot runs that were visible within the time of imaging and reached the point of intersection. Intersections of two distinct microtubules oriented at a variety of angles were examined. However, intersections with microtubule bundles were excluded, as well as intersections with both microtubules of either high or low fluorescence intensity for the kinesin-2 experiments. Moreover, the minimum pause duration scored for KIF3AC, KIF3AA, and KIF5B was defined as a vertical on the kymograph for at least two frames or ϳ1.1 s. An event at the intersection was scored as a detachment provided the frame was not significantly different from the background for at least one frame or ϳ0.6 s. Because of the very slow velocity of KIF3CC, the shortest pause scored was three frames or 15 s, and a detachment was defined as not being significantly different from background for one frame or greater, which corresponds to Ն5 s.
For each of the three kinesin-2 motors, the overall N value of 310 was composed of 150 -160 kinesin-Qdots, which approached the intersection from the dim (bottom) microtubule, and 150 -160 Qdots, which approached the intersection from the bright (top) microtubule. For KIF5B, the overall N value of 251 was not further subdivided based on the orientation of the incident microtubule. These data resulted from 4 -6 days of independent experiments for KIF3AC, KIF3AA, and KIF5B and 9 days of independent experiments for KIF3CC.

Statistical analysis
The R language and environment for statistical computing was used (R Core Team, R Foundation for Statistical Computing, Vienna, Austria) to test for the statistical significance of differences in the behaviors of the KIF3AC, KIF3AA, and KIF3CC populations. The R software was used to compare the kinesin-2 populations by conducting a two-sample test for equality of proportions with a continuity correction. This test was also applied for each motor population to compare the proportions of track switching at acute versus at obtuse angles of intersecting microtubules. The IBM SPSS Statistics for Windows software (version 25.0) was used to remove outliers from the sets of pause duration data for each motor population, compute mean pause durations and standard errors, and conduct independent sample t tests to compare and determine statistically significant differences in mean pause durations. For each comparison, a two-tailed p value was computed and evaluated at a significance level of ␣ ϭ 0.01