The net charge of the K-loop regulates KIF1A superprocessivity by enhancing microtubule affinity in the one-head-bound state

KIF1A is an essential neuronal transport motor protein in the kinesin-3 family, known for its superprocessive motility. We determined that superprocessivity of KIF1A dimers originates from a unique structural domain, the lysine rich insertion in loop-12 termed the ‘K-Loop’, which enhances electrostatic interactions between the motor and the microtubule. In 80 mM PIPES buffer, replacing the native loop-12 of KIF1A with that of kinesin-1, resulted in a 6-fold decrease in run length, and adding additional positive charge to loop-12 enhanced the run length. Interestingly, swapping the KIF1A loop-12 into kinesin-1 did not enhance its run length, consistent with the two motor families using different mechanochemical tuning to achieve persistent transport. To investigate the mechanism by which the KIF1A K-loop enhances processivity, we used microtubule pelleting and single-molecule dwell times assays in ATP and ADP. First, the microtubule affinity was similar in ATP and in ADP, consistent with the motor spending the majority of its cycle in a weakly-bound state. Second, the microtubule affinity and single-molecule dwell time in ADP were 6-fold lower in the loop-swap mutant compared to wild type. Thus, the positive charge in loop-12 of KIF1A enhances the run length by stabilizing the motor binding in its vulnerable one-head-bound state. Finally, through a series of mutants with varying positive charge in the K-loop, we found that the KIF1A processivity is linearly dependent on the charge of loop-12.

coil domain  was inserted just upstream of Ala345 at the start of the neck-coil of 159 Kin1 to generate K11Anc (Fig. 1A). 44 In single-molecule assays, the Kin1 run length was 0.8 ± 0.02 160 µm/s (Fig. 1E), which is consistent with previous work and is four-fold shorter than KIF1A (Fig.  161   1C). 17,35,36 K11Anc had a ~2-fold shorter run length than Kin1 (0.4 ± 0.002 µm; Fig. 1D) and also 162 had a slightly slower velocity (0.5 ± 0.2 µm/s versus 0.7 ± 0.2 µm/s; Fig. 1E). It is unlikely that 163 the shorter run length of K11Anc results from differences in positive charge in the neck-coil domain, 164 because both the kinesin-1 and KIF1A neck-coil domains are negatively charged (-3 for Kin1 and 165 -2 for KIF1A; Fig. S2). Instead, the shorter run length of K11Anc is consistent with the KIF1A neck-166 coil dimerizing only weakly and acting as an extended neck linker in kinesin-1, loosening the 167 connection between the two motor domains and reducing its performance. 17,40-42 In light of this, it 168 is surprising that replacing the KIF1A neck-coil with the more stable kinesin-1 neck-coil does not 169 enhance the KIF1A run length. However, the native neck linker domain of KIF1A is longer than 170 that of kinesin-1, 43 and so one possibility is that stabilizing the KIF1A neck region by adding the 171 kinesin-1 coiled-coil is not sufficient to establish a tight connection between the two heads. It 172 follows that the superprocessivity of KIF1A does not result from tight mechanical connection 173 between the heads to achieve coordinated stepping, but rather from other aspects of the motor's 174 chemomechanics. 175 176

Influence of ionic strength on KIF1A motility 177
To better understand the role of electrostatic interactions in KIF1A motility and to reconcile diverse 178 studies across the literature, we investigated the effect of buffer ionic strength on KIF1A motility. 179 We chose two buffers commonly used in the literature: BRB80, which contains 80 mM PIPES and 180 has a 173 mM ionic strength, and BRB12, which contains 12 mM PIPES and has a 26 mM ionic 181 strength (both buffers also include 1 mM MgCl2, 1 mM EGTA, pH 6.9). Although much of the 182 published work on KIF1A was performed in BRB12, 19,27,31,32 the ionic strength of BRB80 (173 183 mM) is closer to the ~200 mM ionic strength estimated in cells. 45 The low ionic strength of BRB12 7 is expected to maximize electrostatic interactions between motors and microtubules; for instance, 185 the Debye length in BRB80 is 0.7 nm in BRB80 and 2 nm in BRB12. 46 186 187 Consistent with enhanced electrostatic interactions, we found that 1A393 had a nearly 5-fold longer 188 run length in BRB12 compared to BRB80 (14.3 ± 0.4 µm versus 3.0 ± 0.02; Fig. 1C). Additionally, 189 we observed increased pausing behavior at low ionic strength, as reported previously. 27 Thus, we 190 quantified the segmented velocity (Fig. 1D) and found that the velocity decreased from 1.5 µm/s 191 in BRB80 to 1.1 µm/s in BRB12. The longer run length and slower velocity correspond to a 192 substantially higher affinity of KIF1A for microtubules in BRB12 buffer. To compare the 193 microtubule affinity of KIF1A in its weak-binding ADP state in these different buffers, we 194 measured the microtubule dwell time in 2 mM ADP and found that the dwell time in BRB12 was 195 10-fold longer than in BRB80 (Fig. 1F). Thus, the enhanced run length of KIF1A  we compared the effects of the loop swap on the microtubule on-rate of the two motors. Using 259 stopped-flow, the bimolecular on-rate (k !" #$ ) was measured by incubating the motors with 260 fluorescently labeled mant-ADP in BRB80 and flushing the solution against varying 261 concentrations of taxol-stabilized microtubules in 2 mM ADP. Upon mixing, the motors bind to 262 the microtubule and release the mant-ADP, which results in a fall in fluorescence. Because at low 263 microtubule concentrations mant-ADP release is rate limited by microtubule binding, a linear fit 264 of the rates to the microtubule concentration yields the bimolecular on-rate. 15,42,47,48 Notably, 265 swapping the K-loop out of KIF1A had little effect on the bimolecular on-rate of KIF1A,266 contrasting with its effect on the run length (Fig. 2D). However, swapping the KIF1A loop-12 into 267 kinesin-1 caused a 5-fold increase in the on-rate (Fig. 2G). To investigate the mechanism by which the K-Loop enhances KIF1A processivity, we quantified 272 key rate constants in the KIF1A chemomechanical cycle that determine motor run length. Previous 273 work established that kinesin processivity is determined by a kinetic race as the motor takes a 274 forward step, as follows. 17,18 Following ATP hydrolysis, kinesin is in a vulnerable one-head-bound 275 state that can resolve either by the tethered head completing a forward step by binding to the next 276 tubulin binding site and transitioning to a tight binding state, or by the motor dissociating from the 277 microtubule and terminating the run (Fig. 3A). 15,17 Thus, in principle the K-loop could enhance 278 KIF1A processivity by some combination of increasing the on-rate that the tethered head binds to 279 its next binding site (kon TH ) and decreasing the rate that the bound head detaches from the 280 microtubule (kdetach). hydrolysis cycle in a weakly-bound ADP-like state, in agreement with previous work. 15 To 296 calculate the detachment rate from the weak-binding state, we used the KD in ADP together with 297 the k !" #$ results from Fig. 2D, which were 10.6 ± 0.5 µM -1 s -1 for 1A393 and 9.1 ± 2.5 µM -1 s -1 for 298 1A-K1L12. From these values, we calculated kdetach in ADP to be 18 ± 11 s -1 for 1A393 and 66 ± 299 50 s -1 for 1A-K1L12, matching the 4-fold difference in the run lengths in Fig. 2B. 300

301
To more directly test whether the K-loop slows dissociation of KIF1A from the microtubule in the 302 weakly-bound post-hydrolysis state, we used single-molecule TIRF microscopy to measure the 303 dwell time of landing events in 2 mM ADP. For 1A393, the dwell time distribution was well fit by 304 a single-exponential with a time constant of 1.8 ± 0.08 s (Fig. 1F), corresponding to an off-rate of 305 0.56 ± 0.02 s -1 . When we repeated the experiment for 1A-K1L12, the kymographs showed a 306 population of very transient events along with a population of longer duration events (Fig. S3). 307 The dwell time distribution was well fit to a double exponential with a fast population of 0.03 ± 308 0.01 s (koff = 33 s -1 ) that constituted ~40% of the events, and a slow population of 0.3 ± 0.03 s (koff 309 = 3.3 s -1 ) that constituted the other ~60 % of the events (Fig. 3C). Relative to the 1A393 dwell time 310 of 1.8 s, these 1A-K1L12 dwell times correspond to a 60-and 6-fold faster off-rate when the K-311 loop of KIF1A is swapped out. Thus, both the microtubule pelleting assay and the single-molecule ± 30, and 116 ± 13 s -1 , respectively; KM were 119 ± 20, 96 ± 53, and 64 ± 28 µM, respectively (fit 337 ± 95% confidence interval). Because hydrolysis and mant-ADP are thought to be fast, 15 kmax serves as a proxy for the tethered 351 head binding rate. The kmax for our 1A393 construct (154 ± 30 s -1 ; fit ± 95% confidence interval) 352 was similar to 1A368 (172 ± 10 s -1 ), indicating that the neck-coil does not alter the tethered head 353 on-rate. For 1A-K1L12, kmax decreased slightly to 116 ± 13 s -1 , indicating that, in addition to 354 slowing the microtubule off-rate, the K-loop may enhance KIF1A superprocessivity by increasing 355 the tethered head on-rate, k !" '( . However, a caveat of this conclusion is that because the tethered-356 head attachment is thought to be the rate limiting step of the KIF1A chemomechanical cycle, 15 a 357 decrease in this rate should also decrease the overall motor velocity. Instead, the velocity for 1A-358 K1L12 was 20% faster than the 1A393 (Fig. 2C). Additionally, the calculated k !" '( for 1A-K1L12 359 of 116 ± 13 s -1 is slower than the overall stepping rate of 212 s -1 (calculated as 1.7 µm/s velocity the observed run length in ATP and the measured detachment rate in ADP, as follows. From the 365 kinetic race shown in Fig. 3A, 17 the probability of detaching during each cycle is: 366 The number of steps a motor takes before dissociating can be estimated as the inverse of the 367 detachment probability: 368 Thus, the tethered head attachment rate can be estimated by multiplying the measured detachment 369 rate in ADP by the measured run length in ATP: 370 Table 1 shows the off-rates in ADP calculated from measured dwell times in Fig. 3 C and D, along 371 with the run lengths calculated in number of steps from Fig. 2 B and E, and the resulting calculated 372 tethered head attachment rate. The first result is that the calculated tethered head on-rates are 373 roughly three-fold faster for KIF1A than kinesin-1, consistent with the faster stepping rate of 374 KIF1A. The key result from this analysis is that for KIF1A, swapping out the K-loop has no effect 375 on the calculated tethered-head on-rate. In summary, the K-loop contributes to the 376 superprocessivity of KIF1A by slowing the off-rate of the motor from the vulnerable 1HB state, 377 and the K-loop does not modulate the processivity of KIF1A by enhancing the tethered head 378 attachment rate. 379 380 381

KIF1A run length scales with charge of the K-loop 388
In principle, swapping in the K-loop of kinesin-1 could be reducing the KIF1A run length either 389 solely due to differences in positive charge, or through some combination of charge and the length 390 of the loop. To test whether the run length data can be accounted for exclusively based on the 391 charge of loop-12, we designed three additional loop-12 mutant constructs having the same length 392 as wild type but having varying net charge in the loop-12 domain. First, we increased the total 393 charge of loop-12 in our 1A393 construct by replacing three of the native uncharged residues with 394 lysines, resulting in a net charge of +7 in loop-12; we refer to this construct as SuperK (Fig. 4A). 395 Using single-molecule motility assays in BRB80 at 2 mM ATP, we found that the run length of 396 SuperK was ~2-fold longer and the segmented velocity was slightly slower compared to the control 397 Table 2). We also used stopped-flow to measure the bimolecular on-rate, k !" #$ , 398 and found that the SuperK on-rate was 1.5-fold faster than 1A393, but the values were within fit 399 error of one another (Fig. 4D). 400 401 Next, to test whether reducing the net charge of loop-12 reduces the run length, we replaced a 402 portion of the of the lysine residues in loop-12 with glutamines. Glutamine was chosen because it 403 is uncharged in our buffer (pH 6.9) and the side chain is of similar size to lysine, minimizing 404 potential steric effects. We substituted four or five lysines in 1A393 by glutamines, creating 4Q and in the kinesin-1 K-loop (K1L12; +1 net charge) led to short but detectable processivity, reducing 408 the net charge of the K-loop further led to undetectable processivity in BRB80 buffer. 409 410 Because lowering the ionic strength enhanced the run length of other KIF1A constructs, we tested 411 the processivity of these K-loop charge mutants in BRB12 buffer and found that they had 412 measurable run lengths under these conditions (see representative kymographs in Fig. S4). This 413 result confirms that the lack of events in BRB80 was not due to protein misfolding or other off-414 target effects. In BRB12, the run lengths of 4Q and 5Q scaled with the net charge of loop-12, as 415 follows (Fig. 4E). Wild-type 1A393 (+4 charge) had a run length 14 ± 0.4 µm; K1L12 (+1 charge) 416 had a run length of 6.7 ± 0.1 µm; 4Q (neutral) had a run length of 2.6 ± 0.1 µm; and 5Q (-1 net 417 charge) had a run length of 1.2 ± 0.1 µm) (see also Fig. 6 and Table 3). The velocities in BRB12 418 also differed, but not in a charge-dependent manner ( Fig. 4F; Table 3). Interestingly the 419 bimolecular on-rates of the three mutants in BRB12 buffer were similar to one another and they 420 were all ~5-fold slower than 1A393 (Fig. 4G). Thus, the key feature of KIF1A loop-12 that enhances 421 the motor's processivity is the positive charge rather than the longer length of the loop relative to  14 We 448 investigated, three pathogenic human mutations: R307P, R307Q, and P305L (Fig. 5A). The four 449 patients harboring the R307Q mutation had moderate to severe KAND with hypertonia/spasticity, 450 and the pair of twins that harbored the R307P mutation displayed brain atrophy and seizures. 13,14 451 In C. elegans, R307Q was able to partially rescue a null mutant; 51 however, to our knowledge the 452 single-molecule properties of R307Q and R307P have not been evaluated. The four patients 453 harboring the P305L mutation ranged from mild to severe KAND; all showed 454 hypertonia/spasticity, but only two of four showed brain atrophy and one of the four exhibited 455 seizures. In contrast to R307Q, P305L was unable to rescue a C. elegans null mutant; however, 456 P305L was shown to be motile in single-molecule assays albeit with impaired performance. 52 Thus 457 the P307 mutants, which decrease positive charge near the K-loop, present more severe clinical 458 phenotypes, but the P307Q can partially rescue worms. In contrast, P305L, which has been 459 proposed to alter the conformation of a helix adjacent to the K-loop, has a less severe clinical 460 phenotype and the motor retains some motility, but is unable to rescue mutant worms. 52 461 462 We first examined R307P, R307Q, and P305L mutants using single-molecule tracking in BRB80 463 and 2 mM ATP, and observed no motility for any of the three disease mutants (Fig. 5B). When the 464 ionic strength was lowered using BRB12 buffer, R307P and R307Q showed a higher frequency of 465 landing events and longer duration of diffusive events, but still no persistent directional movement run length and ~2-fold slower velocity than the 1A393 control in BRB12 (Fig. 5C-D). A previous 468 study found that that the P305L mutation strongly reduced the microtubule landing rate, and 469 suggested that the mutation alters the interaction of the K-Loop with the C-terminal tail of 470 tubulin. 52 To directly measure the microtubule on-rate of P305L, we used the stopped flow k !" #$ 471 assay in BRB12 and found that the P305L mutant had a ~3-fold lower on-rate than 1A393 (Fig. 5E). 472 Interestingly, the P305L on-rate of 9.3 ± 3 µM -1 s -1 (Fig. 5E) was faster than either the loop swap 473 mutant 1A-K1L12 or the two glutamine mutants in BRB12 ( Fig. 4G; Table 3). This ~3-fold 474 decrease in the on-rate was in the same direction, but it was much smaller than the ~35-fold on-rates of 1A393 (blue squares) and P305L (green squares) in BRB12. Linear fits give k !" #$ (fit ± 485 95% confidence interval). Points are mean ± SEM.  The positively charged loop-12 in kinesin-3 motors, known as the K-Loop, has been a topic of 494 interest in the field for many years, but there has yet to be a consensus on the role of the K-loop in 495 the mechanism of KIF1A superprocessivity. In this work, we have established a comprehensive 496 understanding of the mechanism of the K-Loop in the KIF1A chemomechanical cycle. We 497 conclude that the unique superprocessivity of KIF1A dimers originates from the charge-dependent 498 interaction of loop-12 with the microtubule, resulting in a reduction in off-rate from the post-499 hydrolysis one-head-bound (1HB) state (Fig. 6). This stabilization of the weak-binding state allows 500 time for the tethered head to bind to the next binding site and complete the step, and thus 501 maximizes the number of steps the motor takes before terminating a processive run. 502 regarding the role of the K-loop in KIF1A superprocessivity. Early work made the striking finding 505 that KIF1A monomer constructs move processively through a combination of intermittent forward 506 steps and 1D diffusion along the microtubule. 25,53 The microtubule affinity of these monomers was 507 shown to depend on the negatively charged C-terminal tail of tubulin, scale with the amount of 508 positive charge in loop-12, and be enhanced in low ionic strength buffers. Subsequently, it was 509 shown that full-length KIF1A is dimeric and the motility of dimer constructs was fast, 510 superprocessive, and lacked the diffusional behavior of the monomers. 30,54 However, using a 511 KIF1A dimer truncated after the neck-coil (1-393 aa), it was found that replacing the KIF1A loop-512 12 with that of kinesin-1 did not diminish the run length, 31 a result that was puzzling in light of the 513 monomer results. Subsequent work showed that, unlike kinesin-1, the neck-coil domain of KIF1A 514 is insufficient to stably dimerize the motor, and stabilized dimers could be created by adding a 515 leucine zipper sequence downstream of the neck-coil. 19,27,30,32 Upon closer inspection, the apparent 516 lack of influence of loop-12 on KIF1A processivity can be explained by the fact that the construct 517 used in that work (KIF1A(1-393)) lacked a distal coiled-coil region that is needed to stabilize the 518 dimer. 31 In that work, KIF1A(1-393) had a run length of 2.6 µm, the loop-swap mutant had a run 519 length of 3.6 µm, and the stably dimerized KIF1A(1-393)-LZ had a run length of 9.8 µm. 31,32 Thus, 520 the apparent lack of influence of the K-loop in the truncated dimer lacking a stabilizing LZ domain 521 can be explained by run lengths being terminated by the motor reverting to monomers and 522 dissociating from the microtubule rather than termination of a processive run by normal 523 dissociation of the dimer. As a competing process, this premature termination would mask any 524 change in run length due to the loop swap. 525 526 By measuring key transitions in the kinetic race that defines processivity, we find that the principal 527 role of the K-loop is to stabilize the weak-binding, post-hydrolysis state of KIF1A. The finding 528 from the pelleting assay that the KD for microtubules is similar in ADP and ATP for both wild-529 type and Loop-12 swapped KIF1A emphasizes the importance of the weak binding state for KIF1A 530 processivity. Their similar residence times are consistent with KIF1A spending the majority of its 531 cycle in a vulnerable one-head-bound ADP state, a key finding from a previous study. 15 The 532 parallel reduction of the microtubule affinity in ADP and ATP upon swapping out the K-loop 533 clearly demonstrates how decreasing the microtubule affinity in the ADP state dictates a reduction 534 in run length. Because the other component of the kinetic race, tethered head attachment, also 535 involves microtubule binding and hence might be expected to be modulated by charge in the K-536 loop, it was notable that swapping out the K-loop did not alter the tethered head on-rate. Tethered 537 head binding involves both a mechanical component of the tethered head stretching to the next 538 binding site, as well as a biochemical component of ADP release to achieve tight binding, and so 539 the simplest interpretation of this result is that the tethered head binding rate is not substantially 540 mediated by electrostatic interactions between the motor domain and the microtubule. previous figures (fit ± 95% CI). Linear fits to the points in BRB12 and BRB80 give slopes of 2.7 552 ± 0.9 and 0.9 ± 0.6, respectively (fit ± 95% CI). B, Diagram depicting how increasing positive 553 charge in Loop-12 leads to a slower off-rate in ADP that results in longer run lengths. 554 555 By systematically replacing lysines in KIF1A loop-12 with glutamines, we find that the run length 556 of KIF1A scales linearly with the net charge of loop-12 in both BRB80 and BRB12 (Fig. 6). Three 557 published studies using KIF1A or the C. elegans ortholog Unc104 in BRB12 buffer found similarly 558 that reducing the charge of the K-loop by either swapping with the kinesin-1 sequence or replacing 559 lysines with alanines decreased the run length. 27,30,31 Arpag et. al. found that swapping the KIF1A 560 in run length, and Lessard found a 2.6-fold reduction in run length when the net charge of the K-562 loop was reduced from +4 to +1 by replacing three lysines in the K-loop with alanines. 27 These 563 values are similar to the 2.1-fold reduction in run length we measured between 1A393 (net charge 564 +4) and 1AK1L12 (net charge +1) (Fig. 6A). Using the C. elegans KIF1A ortholog Unc104, 565 Tomishige found that swapping the K-loop (net charge +4) with the human kinesin-1 sequence 566 (net charge 0) resulted in a 5.4-fold reduction in run length, which matches our 5.5-fold shorter 567 run length for K1-4Q (net charge 0). The similarity across these studies (shown graphically in Fig.  568 S5) reinforces our finding that KIF1A processivity scales linearly with the charge of the K-loop, 569 and also argues that this reduction does not depend on the specific sequences, but solely due to the 570 electrostatic charge. This linear relationship and x-intercept around a -1 charge also helps to 571 explain why a construct in which all six lysines in the K-loop were replaced with alanines, resulting 572 in a net charge of -2, was not measurably processive. 27,31 Finally, the shallower slope and more 573 positive x-intercept in BRB80 in Fig. 6A is consistent with charge shielding at higher ionic strength 574 reducing the impact of electrostatic interactions on the run length. 575 576 We also found that in near-physiological ionic strength buffer, swapping the KIF1A K-loop into 577 kinesin-1 did not confer superprocessivity on this motor (Fig. 2). This result suggests that the 578 KIF1A chemomechanical cycle is tuned such that it relies on the K-loop to achieve 579 superprocessivity, whereas the kinesin-1 chemomechanical cycle is tuned to rely on different 580 mechanisms to achieve processivity. One potential explanation is that because kinesin-1 spends a 581 much smaller fraction of its ATP hydrolysis cycle in a one-head bound vulnerable state than 582 KIF1A,15,36 altering the microtubule affinity of this state has a negligible effect on the run length. 583 However, the microtubule detachment rate of kinesin-1 in the weak-binding ADP state was not 584 altered by swapping in the KIF1A K-loop (Fig. 3D), arguing against this mechanism and 585 suggesting instead that the positive charge in the K-loop cannot stabilize this vulnerable state in 586 kinesin-1 the way it can in KIF1A. One possibility is that during transient episodes when the motor 587 domain is tethered to the microtubule solely by its K-loop, KIF1A rebinds rapidly through its 588 canonical microtubule binding site to maintain association, whereas kinesin-1 rebinds more slowly 589 and instead dissociates from this tethered state despite the added electrostatic interactions. This 590 difference in weak-binding characteristics suggests that binding of kinesin-1 to the microtubule in 591 the ADP state is dominated by a different region of the microtubule binding domain, such as Loop-592 8 or Loop-11/a-4. 55 593 594 There are a number of documented KAND mutations in the well-conserved 'PYRD' sequence at 595 the carboxyl end of loop-12, 14 but how these mutations alter the chemomechanical cycle or the 596 interaction of KIF1A with the microtubule is not clear. Despite a recent report that an R307Q 597 mutant partially rescues vesicle transport in a null-mutant worm, 51 we found that R307Q and 598 R307Q were incapable of productive movement in either BRB80 or BRB12 (Fig. 5). Published 599 molecular dynamics simulations found that R307 in KIF1A (and the equivalent R278 in kinesin-600 1) interact electrostatically with residues in the tubulin core, and thus likely contribute to the 601 strength of microtubule binding. 55,56 The calculated binding free energies between R307 and 602 residues in tubulin were similar between the strong-binding ATP and Apo states and the weak-603 binding ADP state in that work, 55 suggesting that R307 is not involved in nucleotide-dependent 604 changes in microtubule binding affinity that occur through the KIF1A mechanochemical cycle. 605 However, the lack of motility of the R307 mutants suggests this residue plays a key role in 606 mechanochemical coupling in the motor domain and/or the strong binding interaction needed for 607 stepping. On the other hand, the diffusive binding of both R307P and R307Q suggest that these 608 mutations do not prevent the K-loop from interacting with the C-terminal tail of tubulin. A nearby 609 mutation, P305L, was previously proposed to alter the interaction of the K-loop with the 610 microtubule based on a ~35-fold reduction in the single-molecule landing rate. 52 That work was 611 carried out in a ~160 mM ionic strength HEPES buffer at pH 7.4, though a different study using 612 ~80 mM ionic strength HEPES buffer at pH 7.2 found only a two-fold decrease in the landing rate. 613 changes with alterations in motility, as well as how changes in the motile properties translate to 623 defects in axonal transport. 624

625
The present study highlights different chemomechanical tuning strategies that kinesin-1 and 626 kinesin-3 employ to carry out their intracellular transport functions. KIF1A is notable in being fast 627 and superprocessive, and it does this by spending most of its hydrolysis cycle in a vulnerable one-628 head-bound state that is stabilized by electrostatic interactions between the K-loop and the 629 microtubule. Importantly, this strategy causes the motor to detach readily under load, which is 630 seemingly not an advantageous property for a transport motor. 19-22 However, KIF1A binds to the 631 microtubule from solution at a much faster rate than kinesin-1, which may partly compensate for 632 this detachment. 15 In contrast, kinesin-1 is able to walk processively against substantial loads, 20,22 633 but it walks more slowly, it binds to the microtubule out of solution more slowly, and in the absence 634 of load has a considerably shorter run length. 15,31 As the mechanistic features of these motors 635 become clarified, the next step is to understand how these functional properties scale up to 636 multimotor cargo transport on diverse microtubules and bidirectional tug-of-war transport with 637 dynein. Similarly, to understand how specific mutations lead to KAND disease states, it will be 638 necessary to define the effect of other mutations on KIF1A chemomechanics and to extrapolate 639 how those changes affect neuronal function. 640 used as the 'wild type' throughout this study, included the KIF1A motor head, neck linker, and 645 neck-coil domains (residues 1-393) followed by the D. melanogaster kinesin-1 neck-coil and 646 coiled-coil 1 domains (residues 345 to 560), and C-terminal GFP tag for single-molecule assays, 647 or D. melanogaster kinesin-1 neck-coil (residues 345 to 406) for biochemical assays. The 1A368 648 construct consisted of the KIF1A motor head and the neck linker domains (residues 1-368) 649 followed by the D. melanogaster kinesin-1 coil domains for stable dimerization (same residues 650 and tags as used above for 1A393). The 1A-LZ construct was composed of R. norvegicus KIF1A 651 residues 1-393 followed by a leucine zipper domain (see Fig. S2 for sequence) for stable 652 dimerization and a C-terminal GFP tag. 32 Subsequent constructs used throughout this study 653 introduced mutations (as described in the Results) using the 1A393 construct as the template. All 654 proteins contained a C-terminal His6 tag. Plasmids were designed in SnapGene and mutants were 655  for each study. Line represents fit to the relative run length versus charge for the present study (see Figure 5 for plot of non-normalized data). All experiments were performed in 12 mM PIPES buffer using constitutively active KIF1A dimers stabilized by an added coiled-coil domain.
(Bottom) Actual and normalized run lengths from the four studies.