that runs parallel to the MT axis about the length of a tubulin dimer (Fig. 1, A to C, and fig. S1B). Unlike MAP2, MAP4, and tau, which bind along the outer ridges of the protofilaments (12)(13)(14), MAP7 runs halfway between the outer ridge and the site of lateral contact. The costructure of MAP7's MT binding domain (MTBD; residues 60 to 170) and FL tau on the MT (fig. S2) illustrated their distinct MT footprints and confirmed that the helical segment corresponds to the MTBD of MAP7. We identified a single MAP7 sequence register that corresponds to a well-conserved segment of the MTBD (residues 87 to 139; fig. S3,A to C) through Rosetta modeling (15) and validated this registry by determining the structure of MTs decorated with a shorter MAP7 construct (residues 83 to 134; fig. S3D). Because MAP7 MTBD could potentially form a helix longer than the length of a tubulin heterodimer (fig. S2A), we cannot exclude the possibility of a larger footprint of MAP7 on the MT (13) (fig. S4; see materials and methods).

The a-helical density for MAP7 is not uniform (Fig. 1B). Segment I (residues 113 to 139)the best-resolved region (Fig. 1B)-interacts extensively with tubulin (Fig. 1D); Q113 and E117 of MAP7 are within hydrogen-bonding distance of N197 and S155 of b-tubulin, respectively; R114 and K127 of MAP7 engage in electrostatic interactions with E159 and D414 of b-tubulin, respectively. Y108 of b-tubulin inserts into a hydrophobic pocket formed by R120, R121, and V124 of MAP7. We also identified potential hydrogen bonds between R128, R131, and K136 of MAP7 and the mainchain oxygens of E411 and G410 of b-tubulin and V159 of a-tubulin, respectively. Segment III (residues 87 to 99) interacts with a-tubulin. Segment II (residues 100 to 113) faces a cavity at the intratubulin dimer and has the weakest density in our map (Fig. 1B). All-atom molecular dynamics (MD) simulations verified these pairwise interactions, and identified additional potential interactions between MAP7 and tubulin (fig. S5 and movie S2). Segment II fluctuated more than other segments in simulations because it only makes transient contacts with tubulin (Fig. 1, E and F, and fig. S5), explaining why this segment has weak density in our cryo-EM map.

We next determined how MAP7 affects the motility of kinesin and the dynein-dynactin-BicDR1 complex [(DDR), hereafter dynein] (16) (Fig. 2A). FL MAP7 uniformly decorated MTs with a dissociation constant (K D ) of 111 ± 12 nM (±SE) under physiological salt (Fig. 2, B andC, and fig. S6). As previously reported (9,11), the addition of 50-nM MAP7 rescued FL kinesin from autoinhibition (17) and substantially increased its run frequency and length (fig. S7). MAP7 also enhanced the motility of constitutively active kinesin (K560, residues 1 to 560) (Fig. 2, D and E and movie S3) (9,11). Unlike run length and frequency, kinesin velocity decreases even at low MAP7 concentrations (Fig. 2E), which could be a result of pausing at MTBD obstacles or binding to the projection domain. Previously, dynein has been reported not to be inhibited by 5-nM MAP7 (10); however, we found that it was inhibited by MAP7 with a half-maximal inhibition constant (IC 50 ) of 10 ± 3 nM (Fig. 2, D andE). MAP7 decoration of MTs also switched the direction of an assembly that links FL kinesin to dynein (Fig. 2F) (18). 80% of kinesin-dynein assemblies were minus-end-directed on undecorated MTs, whereas 93% moved toward the plus end upon addition of 10 nM MAP7 (Fig. 2, G and H, fig. S6C, and movie S4).

Unexpectedly, kinesin run frequency started to decrease when MAP7 concentration was increased further (100 to 1000 nM; Fig. 2, D andE, and movie S3). We reasoned that the nonlinear relationship between MAP7 decoration and kinesin motility could arise when the motor is subjected to simultaneous activation and inhibition that dominate at different concentrations (Fig. 2E) (19). To determine the source of these opposing inputs, we disrupted the kinesin-MAP7 interaction by truncating either the MAP7-binding domain of kinesin (K490, residues 1 to 490) or the kinesin-binding domain of MAP7 (MAP7-N and MAP7-MTBD; Fig. 3A). In all cases, MAP7 inhibited kinesin (Fig. 3B, fig. S8, and movies S5 and S6) as strongly as it inhibited dynein (Fig. 2E). Thus, MTBD inhibits whereas the projection domain activates kinesin (10).

MTBD may inhibit kinesin by competing for the same tubulin binding site, because superimposing the MT-bound structure of kinesin (20) onto our model reveals an apparent clash between kinesin and segment II of MAP7 (fig. S9A). Alternatively, the flexible segment II may accommodate kinesin binding by shifting away from the intradimer interface, as proposed for DCX and MAP4 (14,21).

To distinguish between these possibilities, we determined the structure of MTs incubated with rigor kinesin [K350, residues 1 to 350, E236A mutation (K350 E236A )] and MAP7 (Fig. 3C). Because distinct binding sites of MAP7 and kinesin can be artifactually averaged during reconstruction (fig. S9B), we performed focused 3D classification around the putative shared binding site of MAP7 and kinesin (fig. S10; see methods). The classification resulted in two distinct maps of the binding site, one occupied by MAP7 only, and the other occupied by kinesin only (Fig. 3D). Thus, kinesin and MAP7 cannot simultaneously bind to the same tubulin dimer.

Consistent with cryo-EM, MTBD was unable to bind MTs predecorated with K350 E236A (Fig. 3E). However, FL MAP7 or a construct lacking the MTBD (DMTBD; fig. S11,A to C) still bound to K350 E236A -decorated MTs, presumably through the P123 domain (residues 175 to 316) (8,9). MT binding of DMTBD was nearly abolished upon cleaving the flexible tails of tubulin (Fig. 3E and fig. S11D), explaining why P123 was invisible in our structure. To verify that kinesin is inhibited at high levels of MAP7 decoration as a result of binding site overlap, we also replaced the MTbinding regions of MAP7 with that of tau, which also overlaps with the tubulin binding site of kinesin and inhibits kinesin motility (fig. S12) (13,22). Similar to FL MAP7, binding of this chimeric MAP to the MT resulted in biphasic regulation of kinesin motility (Fig. 3, F andG).

To reveal how kinesin "walks" along MAP7decorated MTs despite competing for the same binding site, we fluorescently labeled the motor domain of K560 and tracked kinesin stepping with nanometer precision under limiting adenosine triphosphate (ATP) conditions (Fig. 4A). On undecorated MTs, kinesin took 16-nm steps in the forward direction (23), whereas steps in the sideways and backward directions were rare (9 and 3%, respectively; Fig. 4, A andB, and fig. S13A) (24). At 36 nM MAP7, we observed a modest increase in the probability of sideways and backward stepping (Fig. 4, A andB). At nearly saturating MAP7 concentrations, we observed 16-to 64-nm displacements in our trajectories, as well as increased stepping in sideways (35%) and backward (26%) directions (Fig. 4, A andB). Because kinesin cannot take such large steps on its own (Fig. 4, A andB), these large displacements likely represent transient detachment and reattachment of the motor to the MT.

In the absence of the projection domain, kinesin was stuck on MTs decorated with 36-nM MTBD and dissociated from MTs at increased ATP concentrations (fig. S13, B andC). Furthermore, the addition of the C-terminal half of MAP7 (MAP7-C) was unable to stimulate kinesin motility on MTs decorated with the N-terminal half of MAP7 (MAP7-N, fig. S14). Thus, the projection domain is needed to be tethered to the MTBD to enable to bypass MTBD obstacles.

On the basis of our results and previous reports (9,11), we propose a model for kinesin stepping on MAP7-decorated MTs. MAP7 recruits kinesin-1 to the MT and activates subsequent motility (9,11). MT binding of MAP7 also obstructs kinesin stepping along the protofilament, which results in kinesin dissociation from the MT. However, the MAP7 projection domain tethers kinesin to the MT surface, allowing it to rebind the MT at nearby sites not blocked by its MTBD. This "tethered diffusion" of kinesin appears as large forward, sideways, or backward displacements in our trajectories. When the MT surface is nearly saturated with MAP7, the frequency and length of kinesin runs are reduced because of the scarcity of empty tubulin sites to which the motor can rebind after it dissociates (Fig. 4C). Unlike kinesin-1, MAP7 inhibits dynein motility because its MTBD overlaps with the dynein binding site (fig. S15) but its projection domain does not tether dynein to MTs. This biphasic regulation mechanism may enable precise control of kinesin-1-driven transport by varying MAP7 density on cellular MTs. Although MAP7 is required for kinesin-1driven processes in many cell types (10,25,26), dense MAP7 localization was shown to slow and pause organelle transport at brunch junc-tions in rat neurons (8). Increased accumulation of MAP7 may reroute cargos to their destinations by facilitating the detachment of kinesin from its MT track and rebinding to neighboring MTs at these junctions (8).