Autoinhibition of kinesin-1 is essential to the dendrite-specific localization of Golgi outposts

Neuronal polarity relies on the axon- or dendrite-specific localization of cargo by molecular motors such as kinesin-1. This study shows how autoinhibition regulates both kinesin-1 activity and localization to keep dendritic Golgi outposts from entering axons.


Introduction
To receive and send signals, neurons rely on the polarized distribution of vesicles, macromolecules, and organelles to dendrites and axons. The majority of transport in neurons is mediated by the microtubule (MT)-based motors kinesin and dynein, which traffic cargo in both compartments. To investigate how motor function is regulated to maintain proper cargo distribution we focused on the MT plus end-directed kinesin-1. Kinesin-driven events must be tightly controlled in the cell body to ensure that both outgoing and incoming cargos reach their proper destinations.
In contrast with mammalian neurons, sorting decisions in the cell body of invertebrate neurons are theoretically simplified as axons contain plus end-distal MTs whereas dendrites contain predominantly minus end-distal MTs (Rolls et al., 2007;Stone et al., 2008). However, for both vertebrate and invertebrate neurons, transport must be highly regulated to prevent kinesin-driven transport in dendrites from continuing unimpeded into axons upon reaching the cell body. Similarly, in the cell body, kinesin-1 bound to dendritic cargo must be inactivated to enable transport out to dendrites by dynein and to prevent axon entry. In both mammalian and invertebrate neurons, MT polarity must be coupled with additional mechanisms that regulate kinesin-1 activity and/or cargo attachment to achieve proper cargo distribution .
Kinesin-1, like other kinesin family members, is regulated by autoinhibition (Verhey and Hammond, 2009). Autoinhibition inactivates kinesin-1 when it is not cargo-bound and may finetune motor activity when kinesin-1 is cargo-bound (Lane and Allan, 1999;Woźniak and Allan, 2006;Verhey and Hammond, 2009;Fu and Holzbaur, 2013;Maday et al., 2014). Although in vitro studies indicate that autoinhibition plays a key role in controlling motor function, the cellular role or roles for autoinhibition of kinesin-1 are still poorly understood. To date, motor autoinhibition has been shown to modulate axonal transport (kinesin-1), axon outgrowth (kinesin-3), and synapse positioning (kinesin-4; Moua et al., 2011;van der Vaart et al., 2013;Cheng et al., 2014;Niwa et al., 2016). It is attractive to consider the possibility that autoinhibition of kinesin-1 would regulate the compartment-specific localization of cargo as well.
We used in vivo structure-function analysis of endogenous kinesin-1 in fruit flies to uncover mechanisms that regulate kinesin-1. We sought to identify mutations that would lead to the axonal mislocalization of dendritic cargo, specifically Golgi outposts, which our data suggest are carried by kinesin-1. We initially tested the idea that kinesin-1 uses its MT-binding domain (MTBD) to "read" positional information encoded on MTs. Studies of truncated kinesin-1 have implicated three different loops in the kinesin-1 MTBD (β5-loop 8, loop 11, and loop 12) in polarized transport; however, there is no consensus on which loops are essential (Nakata and Hirokawa, 2003;Konishi and Setou, 2009;Nakata et al., 2011;Huang and Banker, 2012). Our in vivo studies of endogenous kinesin-1 show that E177 within β5-loop 8 is critical to the dendrite-specific localization of Golgi outposts. This particular glutamate, E177, is part of a short motif previously implicated in kinesin-1's preference for axons in vertebrate neurons (Konishi and Setou, 2009), but it also mediates contact with the C-terminal tail when kinesin-1 is in an autoinhibited conformation (Kaan et al., 2011). Our combined in vivo and in vitro results indicate that mutating E177 results in the axonal mislocalization of Golgi outposts by relieving kinesin-1 autoinhibition. We show that autoinhibition is also critical for the proper localization of kinesin-1 itself. Like constitutively active truncated motor constructs, uninhibited full-length kinesin-1 is enriched in axon terminals and nearly depleted from dendrites, which correlates with the axonal mislocalization of outposts and reduced dendrite growth. Last, our genetic interaction tests suggest that outposts are restricted to dendrites by balancing the activities of kinesin-1 and dynein. Dynein transports outposts into dendrites (Ye et al., 2007;Zheng et al., 2008;Lin et al., 2015), and our data indicate that the selective localization of outposts to dendrites relies on the autoinhibition of axon-targeted kinesin-1.

Results
Survival and neuronal morphogenesis depend on E177 in β5-loop 8 of kinesin-1 We initially set out to test whether kinesin-1 is stopped from carrying dendritic cargo into axons via the recognition of MT-based signals by its MTBD, which is comprised of three loops. To identify the loop or loops essential to transport, we took an in vivo structure-function approach and mutated endogenous kinesin-1 in fruit flies, which express one highly conserved kinesin-1, Kinesin heavy chain (Khc). Most of what is known about kinesin-1-mediated transport is derived from truncated motors. Thus, it is unclear which domain or domains are necessary for the full-length motor to mediate the selective localization of cargo in vivo. We took advantage of a fly strain we previously created that enables the rapid knock-in of Khc alleles via site-directed integration (Fig. 1 A; Winding et al., 2016). As Khc is an essential gene, we used viability to initially identify the loop or loops critical to Khc function.
β5-Loop 8 contains four amino acids (TERF) that were previously implicated in polarized transport (Konishi and Setou, 2009). The glutamate (E177) in this motif stabilizes the motor in an autoinhibited conformation, thus implicating this motif in both motor navigation and autoregulation (Kaan et al., 2011). We substituted the TERF sequence in Khc with the equivalent sequences from unc-104 and Khc-73. Strikingly, swapping Khc TERF with SKLA from unc-104 resulted in lethality, whereas mutating to the Khc-73 SQLA did not (Table 1). Notably, E177K had been previously uncovered in a genetic screen for lethal mutations in Khc (Brendza et al., 1999). Though mutating E177 to alanine (E177A) had no effect, we found that both charge-reversal mutations E177K and E177R caused lethality (Table 1).
We next determined how the Khc E177 mutations affected neuronal morphogenesis. The Drosophila melanogaster class IV dendritic arborization sensory neurons elaborate a highly branched dendritic arbor in the periphery and extend a single unbranched axon into the ventral nerve cord (VNC). Decreasing Khc levels reduced dendrite arborization and resulted in axons with minor ectopic branches (Fig. 1, C-E and I-K), consistent with research showing that Khc is essential for dendrite morphogenesis (Satoh et al., 2008). The viable Khc E177A mutation had no effect on neuron morphogenesis, but Khc E177K phenocopied Khc-RNAi (Table 2 and Fig. 1, F-K). The dendrite arborization phenotype of both the Khc-RNAi and Khc E177K neurons became more pronounced as the animals aged ( Fig. 1, D , E, G, and H). Thus, the lethal E177 charge-reversal mutations disrupt neuron morphogenesis, indicating Khc-mediated transport is disrupted.
The Khc E177K mutation disrupts the dendrite-specific localization of Golgi outposts and human transferrin receptor (hTfR)-positive vesicles The Khc E177K neurons resemble mutants in which dendritic transport is disrupted. Moreover, the ectopic axonal branching is characteristic of mutations that result in the axonal mislocalization of dendrite-specific Golgi outposts (Ye et al., 2007;Satoh et al., 2008;Zheng et al., 2008;Arthur et al., 2015). This led us to test whether Khc transports Golgi outposts and has a role in restricting them to dendrites. Kinesin-1 has been linked to Golgi membranes via its adapter kinesin light chain (Klc), which we found partially colocalizes with a subset of outposts in dendrites (Fig. 2, A-D; Marks et al., 1994;Gyoeva et al., 2000;Allan et al., 2002;Woźniak and Allan, 2006). A third of motile Golgi outposts colocalize with Klc, with most moving in a direction and at a velocity consistent with kinesin-mediated transport (Fig. 2 D). In control and Khc E177A neurons, Golgi outposts localized predominantly to dendrites (Fig. 2 B). In contrast, Golgi outposts mislocalized to axons in Khc E177K , Khc E177R , and Khc-RNAi-expressing neurons (Fig. 2,E and F). The mislocalization of outposts in Khc E177K neurons was rescued by the neuron-specific expression of wild-type Khc, indicating a cell autonomous requirement for Khc in outpost distribution. Mislocalized outposts were visible in both the proximal axon and axon terminals in the VNC (Fig. 2, G and H). These data indicate that Khc transports Golgi outposts and that the E177K and E177R mutations disrupt motor function to cause ectopic outpost localization.
To determine whether additional dendritic cargos are affected, we analyzed GFP-tagged hTfR. hTfR::GFP is a commonly used marker of dendritic vesicles that are transported, at least in part, by kinesin-1 and whose localization in fly neurons is Khc-dependent (Schmidt et al., 2009;Henthorn et al.,  Cartoon of kinesin-1 highlighting β5-loop 8 (green), loop 11 (orange), and loop 12 (blue), which were replaced by equivalent loops from unc-104. β5-loop 8 is highly conserved between flies and humans. TERF sequence is underlined; arrow indicates Khc E177, also highlighted in red. (C-H) Representative images of control (con) and mutant neurons at 72 and 96 h AEL. Bar, 50 µm. (I) Quantification of dendrite length and branch number (mean ± SD) at 72 h AEL of 11 (Khc E177A/− ), 13 (Khc E177K/− ), and 16 (control, Khc-RNAi) neurons in at least three larvae and at 96 h AEL of 10 neurons (control, Khc-RNAi, Khc E177K/− ) in at least three larvae. ****, P < 0.0001 relative to control and evaluated by one-way ANO VA and Tukey post hoc test (top and middle). Sholl analysis (bottom) of dendrite arbors at 72 h AEL (mean ± SEM). Critical radius and maximum branches reported in Table 2 2011). In control neurons, hTfR::GFP was enriched in dendrites with some puncta in axons (Fig. 2,I and J). The Khc E177K mutation and Khc-RNAi resulted in an accumulation of hTfR::GFP in axons and axon terminals (Fig. 2, J-L). Although Golgi outposts and hTfR::GFP mislocalize to axons, we found no aberrant dendritic localization of three different axonal proteins (Fig. S1). Thus, Khc E177 has a critical role in restricting the localization of both Golgi outposts and hTfR::GFP-positive vesicles to dendrites.
Khc E177K does not phenocopy mutations that disrupt kinesin-1 ATPase activity The phenotypic similarity between Khc E177K and Khc-RNAi suggested that the mislocalization of Golgi outposts in the E177K mutant may be a result of decreased motor activity. However, the E177K mutation had no effect on the velocity or MT dwell time of truncated Khc in single-molecule motility assays (Fig. S2). Furthermore, Golgi outpost localization was unaffected by E164K or S246F mutations that alter ATPase activity and strongly affect the Maximum branches 61 ± 6 49 ± 9*** 39 ± 6**** 57 ± 8 (NS) 61 ± 9 (NS) Sholl analysis of dendritic arbors quantifies the number of dendrites that intersect concentric circles centered on the cell body. The radius of the circle with the maximum dendritic intersections is the critical radius and the intersections at this radius are the maximum number of branches. Mean ± SD for both measurements; n (neurons) = 14 control, 16 Khc-RNAi, 10 Khc E177K , 14 Khc S246F , 11 Khc S246F,E177K . *, P = 0.05-0.01; **, P = 0.01-0.001; ****, P < 0.0001 in comparison with control and evaluated by one-way ANO VA with Tukey post hoc test.  Brendza et al., 1999). Similarly, the E177A mutation did not affect outpost localization despite its reported effect on MT-stimulated ATPase activity and velocity (Woehlke et al., 1997). Thus, the E177K mutation is unlikely to disrupt Golgi outpost localization simply by diminishing Khc's motility properties.
Axonal MT polarity is perturbed by Khc-RNAi but not the E177K mutation The normal localization of Golgi outposts in the Khc ATPase mutants raises the question of why outposts mislocalize to axons when Khc is absent. Decreasing kinesin-1 levels or function can disrupt MT organization in axons and dendrites (Yan et al., 2013;  Lee et al., 2017). In Khc-RNAi-expressing neurons, axonal MT orientation was altered ( Fig. 3 C), which may contribute to the ectopic axonal localization of outposts via dynein and/or another kinesin. MT organization in Khc E177K neurons was normal, indicating that a change in MT polarity does not underlie the mislocalization of Golgi outposts (Fig. 3 C). Ectopic Golgi outposts have been speculated to disrupt the polarity of axonal MTs because of their capacity to serve as platforms for MT nucleation (Ori-McKenney et al., 2012;Sanders and Kaverina, 2015). It is possible that the ectopic outposts in the Khc E177K axons may seed minus end-distal MTs, but that such MTs are cleared by mechanisms that normally maintain axonal MT polarity. Regardless, our data indicate that the ectopic outposts in Khc E177K axons are not a secondary consequence of a change in MT polarity.
Localization of truncated constitutively active Khc is not affected by the E177K mutation A motor's ability to navigate in cells is typically assayed using truncated motors that lack cargo-binding and autoregulatory domains. Like the equivalent truncated mammalian kinesin-1 (Nakata and Hirokawa, 2003), truncated fly Khc(1-401)::superfolder GFP (sfGFP) was present at low levels in the proximal axon and dendrites but accumulated in axon terminals ( The E177K mutation disrupts autoinhibition of kinesin-1 When kinesin-1 is in an autoinhibited conformation, E177 mediates an intramolecular salt bridge between the motor and tail (Fig. 4 A; Kaan et al., 2011). We tested whether the E177K mutation might alleviate autoinhibition by disrupting the motor-to-tail interaction. We first turned to an in vitro single-molecule approach. Wild-type or mutant full-length Khc motors tagged with mNeonGreen were added to flow cells containing MTs, and their motility was observed by total internal reflection fluorescence (TIRF) microscopy. As expected, very few motility events were observed for wild-type Khc, which is typically in an autoinhibited state in solution (Fig. 4 B).
When motile, the motor typically moved only a short distance at slow speed before pausing or detaching (Fig. 4, B-D). As a positive control, we deleted the central hinge (Δhinge2) that enables the motor to fold on itself and adopt the autoinhibited state (Friedman and Vale, 1999). The Δhinge2 motors showed smooth movement along the MTs, with few pauses, and with speeds typical of an active kinesin-1 motor (Fig. 4, B-D), consistent with a previous study (Friedman and Vale, 1999). Motors with autoinhibition-disrupting mutations (E177K, R947E, and K944E) displayed motility that was characteristic of active kinesin-1 and was increased compared with the wild-type motor (Fig. 4, B-D; Hackney and Stock, 2000). The single-point mutant motors paused more than the Δhinge2 mutant motor, likely because these motors can still adopt the folded conformation, whereas the Δhinge2 mutant exists in an extended state. If the E177K mutation relieves autoinhibition, then the resultant activation should be ameliorated by decreasing ATPase activity. The double-mutant E177K, S246F motor behaved similarly to the single-mutant S246F and wild-type motors, indicating that reducing motor activity suppresses the effects of the E177K mutation ( Fig. 4, B-D). These findings support the idea that E177 has a role in regulating kinesin-1 autoinhibition to control motor function.
Disrupting autoinhibition perturbs Golgi outpost localization and dendrite morphogenesis Given the single-molecule evidence that E177K disrupts kinesin-1 autoinhibition, we next asked whether Golgi outpost localization would be affected by other mutations that disrupt autoinhibition. Indeed, all the mutations that alleviate kinesin-1 autoinhibition (Δhinge2, R947E, K944E, and P945S) resulted in the ectopic axonal localization of Golgi outposts (Fig. 5, A and B). And, like the E177K mutation, the Δhinge2 and K944E mutations significantly reduced dendrite arborization (Fig. 5, C-G). We also tested whether the S246F mutation would suppress the mislocalization of Golgi outposts caused by E177K. Neurons expressing the double-mutant E177K, S246F Khc did not have any mislocalized Golgi outposts in their axons and had relatively normal dendrite arbors ( Table 2 and Fig. 5, A, B, and E-H). Reducing Khc activity reverses the effects of the E177K mutation, suggesting the E177K phenotypes result from increased motor activity. We further tested this idea by decreasing the levels of the kinesin-1 cofactor ensconsin in the E177K mutant neurons. Ensconsin (also known as MAP7) relieves autoinhibition in Drosophila neurons and embryos to promote kinesin-1 activation (Sung et al., 2008;Barlan et al., 2013). Loss of ensconsin by itself did not disrupt outpost localization, but loss of ensconsin in the Khc E177K neurons partially suppressed the mislocalization of Golgi outposts (Fig. 5, A and B). These data support the idea that a less inhibited, more active kinesin-1 ectopically transports Golgi outposts into axons. We also tested whether Golgi outpost distribution was affected by the mutation of two conserved serines near E177 (S181 and S182) whose phosphorylation has been implicated in regulating kinesin-1 autoinhibition (Morfini et al., 2009). Golgi outpost localization and dendrite arborization were altered by phosphomimetic mutations but not alanine mutations (Fig. S3). This suggests that phosphorylation of these serines may favor the uninhibited active state, whereas dephosphorylation may allow the motor to switch between active and inhibited states.

Loss of autoinhibition enriches Khc in axon terminals and reduces levels in dendrites
Because the E177K mutation enhances kinesin-1 activity, we reasoned the mutant motor might localize primarily to axons like the truncated constitutively active motor. The broad expression of Khc made it difficult to visualize GFP-tagged Khc in individual neurons, so we used split GFP (Cabantous et al., 2005;Kamiyama et al., 2016). We tagged endogenous Khc with sfGFP11 and restricted the expression of sfGFP(1-10) so that sfGFP would be reconstituted specifically in the class IV sensory neurons. Wild-type Khc localized to dendrites, axons, and axon terminals, whereas the E177K mutation resulted in a significant shift in Khc distribution from dendrites to axon terminals (Fig. 6, A, B, and D-F). The double-mutant E177K, S246F Khc localized normally to both dendrites and axons (Fig. 6, C-F). Thus, by lessening autoinhibition, the E177K mutation activates full-length Khc, depleting the motor from dendrites and resulting in its axonal enrichment.
Reduced motility of Golgi outposts and hTfR-positive vesicles in E177K mutant dendrites A significant decrease in dendritic Khc levels is likely to disrupt outpost motility. Consistent with this idea, mutations that alleviate autoinhibition reduced the fraction of motile outposts and increased their pause duration similar to Khc-RNAi (Fig. 7, A-C). The E177K mutation significantly reduced the frequency of retrograde events and the likelihood that outposts moving anterograde switched direction (Fig. 7 C). A significant decrease in anterograde reversals likely underlies the increased anterograde event frequency observed in the Khc E177K neurons (Fig. 7 C). Although pause duration increased, run length was normal in the mutants (Fig. S4 A). Golgi outposts were not the only cargo affected: the E177K mutation and Khc-RNAi also decreased hTfR::GFP flux, most strikingly retrograde flux (Figs. 7 D and S4 B). In contrast, the ATPase mutant S246F had little effect on Golgi outpost or hTfR::GFP motility (Fig. 7, B-D). This suggests that the Khc S246F mutant is sufficiently active to support transport in dendrites, which likely explains why dendrite growth is not affected by this mutation. Combined, our results show that depleting Khc from  Table 2. *, P = 0.05-0.01; ****, P < 0.0001; and n.s. relative to control and evaluated by one-way ANO VA and Tukey post hoc test (G and H).
dendrites correlates with reduced outpost and hTfR::GFPmotility, with pronounced effects on retrograde transport.
Next, we attempted to visualize Golgi outposts being transported by the E177K mutant motor. We found it technically challenging to visualize Khc-E177K::sfGFP in mutant Khc E177K::sfGFP11/− neurons, but we did observe that Khc-E177K::sfGFP colocalized with retrogradely moving outposts in the dendrites of heterozygous Khc E177K::sfGFP11/+ neurons (Fig. 7 E). Because one gene copy of the E177K mutation is not sufficient to drive outposts into axons, we could only observe colocalization in dendrites (only one Khc tail in a motor dimer is needed to inhibit activity). This colocalization is consistent with our model that the E177K mutant motor is able to actively transport Golgi outposts into axons.

Khc interacts with dynein to regulate Golgi outpost localization
Golgi outposts are transported into dendrites by dynein (Ye et al., 2007;Zheng et al., 2008;Arthur et al., 2015), and the dendrite-specific localization of outposts may depend on a regulated balance of dynein and kinesin-1 activities. We used genetic interaction tests to determine whether disrupting the motor balance to favor Khc results in the ectopic axonal localization of outposts.
We used RNAi targeting the dynein light intermediate chain (dlic) subunit (dlic-RNAi) to reduce dynein function. A strong decrease in dynein function (dlic-RNAi and dicer) resulted in the ectopic accumulation of outposts at axon terminals, consistent with the idea that a shift favoring kinesin enables the motor to pull outposts into axons (Fig. 8 A). A weak dynein loss of function (dlic-RNAi alone) did not affect outpost distribution, providing a paradigm to test for genetic interactions (Fig. 8 A). In the weakdynein loss-of-function background, Golgi outpost distribution was not affected by reducing Khc activity by either removing one copy of Khc or decreasing ATPase activity (Fig. 8, B, C, and F). This is consistent with the notion that attenuating the activity of both dynein and kinesin-1 simultaneously maintains a motor balance that restricts Golgi outposts to dendrites.
ectopic Golgi outposts at axon terminals (Fig. 8, E and F). These results support the idea that relieving Khc autoinhibition tips the dynein-kinesin balance toward kinesin-1 and biases transport into axons.
If a balance of dynein and kinesin-1 regulates Golgi outpost distribution, then reducing kinesin-1 levels in dendrites should cause a distal shift of outposts in the arbor. Consistent with this idea, knocking down Khc reduced the number of outposts proximally and increased outposts distally (Fig. 8 G). The E177K mutation, which reduces Khc levels in dendrites, resulted in a similar shift in outpost localization. Although there are more Golgi outposts distally, we did not observe a corresponding shift in dendrite branches such as has been reported for mutations that alter outpost distribution or the dynein-kinesin balance (Ye et al., 2007;Satoh et al., 2008;Zheng et al., 2008;Arthur et al., 2015;Taylor et al., 2015). It is possible that kinesin-1 acts downstream of Golgi outposts to promote dendrite branching, or that loss of Khc disrupts the transport of additional factors needed for branching. Of the Golgi outposts remaining in the proximal dendrites of the E177K mutants, significantly fewer colocalized with RFP::Klc, and overall fewer RFP::Klc puncta were present, paralleling the reduction in Khc levels (Fig. 8 H). Thus, lessening Khc autoinhibition results in Khc, its cofactor Klc, and outposts moving out of dendrites and causes a distal shift of remaining outposts, likely because of an unopposed dynein motor.

Discussion
Neuronal function relies on the accurate delivery of proteins, vesicles, and organelles to axons and dendrites by molecular motors. Our in vivo structure-function analysis of endogenous Khc, combined with in vitro single-molecule experiments, indicate that regulated autoinhibition of kinesin-1 is necessary for the polarized distribution of Golgi outposts and hTfR-positive vesicles to dendrites. Previous studies suggested that autoinhibition may be dispensable for some Khc-mediated activities, whereas the autoinhibitory IAK tail motif is critical to axonal transport (Moua et al., 2011;Williams et al., 2014). Our results indicate that kinesin-1 autoinhibition has a previously unrecognized function in preventing the motor from carrying Golgi outposts into axons. Relieving autoinhibition enhances Khc activity and drives the cargo-bound motor into axons, depleting dendrites of Khc and perturbing dendrite arborization. Disrupting autoinhibition generates phenotypes that resemble kinesin-1 loss of function. However, these effects can be reversed by decreasing ATPase activity, which strongly indicates these phenotypes are a result of a gain in motor activity, not a loss. It is likely that the depletion of the motor from a particular compartment, such as dendrites, contributes to the loss-offunction phenotypes observed when kinesin-1 autoinhibition is disrupted. Thus, tight regulation of kinesin-1 activity is critical to the polarized localization of dendritic cargo, the motor itself, and neuronal morphogenesis.
We had initially sought to determine whether kinesin-1 is stopped from carrying dendritic cargo into axons by reading out MT-based signals such as MT posttranslational modifications, MT-associated proteins, and/or GTP-tubulin subunits (Bentley and Banker, 2016). Mutagenesis of the three loops in the kinesin-1 MTBD revealed that loop 12 and β5-loop 8 are required for normal kinesin-1 function. Swapping loop 12 may decrease animal survival because of the unique K-loop in loop 12 of kinesin-3 increasing the on-rate of the motor onto MTs (Soppina and Verhey, 2014). Swapping β5-loop 8 had a more severe effect on survival. (C) Quantification of outpost motility, pause duration, event frequency, and reversals calculated for 7 (Khc E177K/− ), 9 (control), and 10 (Khc S246F/− ) neurons from at least five larvae (Khc-RNAi and Khc Δhinge2/− neurons had too few events for some calculations). The reversal ratio is the number of events that reversed direction after a pause relative to the number of events that did not; significant differences between the reversal ratios were determined via χ 2 tests (n = number of events after a pause). (D) Quantification of hTfR::GFP flux in dendrites (mean ± SEM) in 9 (Khc-RNAi and Khc S246F/− ), 11 (control), and 12 (Khc E177K/− ) neurons from at least four larvae. *, P = 0.05-0.01; **, P = 0.01-0.001; ****, P < 0.0001; and n.s. relative to control and evaluated by one-way ANO VA and Tukey post hoc test (C and D). (E) Kymograph showing Khc-E177K::sfGFP (green) colocalizes with a Golgi outpost (GalNacT2::RFP, magenta) moving retrograde.
Within β5-loop 8, we found that an essential glutamate (E177) is highly sensitive to the nature of substitution. Charge-removal substitutions (E177A and E177Q) had no effect on survival; however, charge-reversal substitutions (E177K or E177R) were lethal and altered dendritic trafficking. Based on previous studies, we expected that the underlying mechanism would be an alteration of either kinesin-1 responsiveness to MT-based signals such as MT detyrosination (Konishi and Setou, 2009) or autoinhibition (Kaan et al., 2011). We found that the E177K mutation increases motor activity in vitro, causes the full-length motor to localize to axon terminals, and can be combined with dynein loss of function to drive outposts into axons. Additional autoinhibition-disrupting mutations in the Khc tail domain also result in the axonal mislocalization of Golgi outposts. We thus favor the idea that the loss of Khc inhibition is responsible for altered dendritic trafficking in E177K mutants. Although we cannot rule out the possibility that E177 is involved in track selection, structural studies indicate the TERF sequence may not be positioned to directly respond to MT modifications (Gigant et al., 2013;Atherton et al., 2014). Combined, our data suggest that E177 regulates motor activity through autoinhibition, which we propose functions as a brake on the axonal localization of the motor when it is carrying dendritic cargo. It will be interesting to test how the E177K mutation and loss of autoinhibition affect the localization and function of full-length kinesin-1 in mammalian neurons, whose dendritic MTs have a more mixed polarity than fly.
Our study raises the question of how kinesin-1 autoinhibition is spatially regulated. It is likely that autoinhibition of kinesin-1 attached to dendritic cargo would be triggered in the cell body or proximal axon, where dendritic cargo is turned around (Edwards et al., 2013;Rolls and Jegla, 2015;Bentley and Banker, 2016). In the proximal axon of fly sensory neurons, Golgi outposts colocalize with the dynein cofactor NudE, which likely functions with dynein to localize outposts to dendrites (Ye et al., 2007;Zheng et al., 2008;Arthur et al., 2015;Lin et al., 2015). The mammalian NudE orthologue Ndel1 is proposed to "catch" dendritic cargos in the proximal axon and prime them for dynein-mediated transport into dendrites (Kuijpers et al., 2016;Klinman et al., 2017). We found that simultaneously decreasing dynein activity and kinesin-1 autoinhibition resulted in ectopic axonal Golgi outposts. This suggests that the selective transport of Golgi outposts to dendrites relies on coordinating the activation of a dendrite-targeted motor (dynein) and the inhibition of a motor that targets axons (kinesin). A key future goal is to determine how kinesin-1 autoinhibition is spatially regulated to mediate the proper distribution of cargo to axons or dendrites to maintain neuronal polarity.

Molecular cloning Khc knock-in alleles
The Khc knock-in alleles were generated using the fly strain Khc KO-attP , in which Khc is knocked out and replaced with an attP site for rapid site-directed integration of new alleles (Winding et al., 2016). To knock in new Khc alleles, an integration plasmid (pGE-attB-GMR; Huang et al., 2009) containing the new allele was injected into Khc KO-attP embryos and integrated via ϕC31-mediated recombination. The following mutations were introduced by PCR-based mutagenesis using Q5 High-fidelity polymerase (New England Biolabs, Inc.): loop-11 swap, loop-12 swap, SKLA, SQLA, E177A, E177K, E177R, Δhinge2, R947E, K944E, S246F, E177K + S246F, S181D + S182D, and S181A + S182A. To make the β5-loop 8 swap mutant, the unc-104 sequence that encodes β5-loop 8 was ordered as a gBlock (Integrated DNA Technologies) and the final construct was created via Gibson Assembly (New England Biolabs, Inc.). The amino acid sequences that were swapped are as follows: Khc β5-loop 8 (KIR DLL DVS KVN LSV HED KNR VPY VKG ATE RFV SS) was substituted with RVR DLL NPK NKG NLR VRE HPL LGP YVE DLS KLA VTD from unc-104; Khc loop 11 (KVS KTG AEG TVL DEAK) was substituted with RAD STG AKG TRL KEGA from unc-104; and Khc loop 12 (GNKT) was substituted with VAS KKK NTK KAD from unc-104. To make Khc::sfGFP11, the sfGFP11 peptide (multimerized 7×) was added to the C-terminal end of Khc with a GGS GG linker between Khc and sfGFP11 and successive sfGFP11 peptides. All knock-in plasmids were fully sequenced before injection (BestGene Inc.). The oligonucleotides and gene fragments used to make these and all other constructs are listed in Table S1.
Image capture and analysis All neuron imaging was performed on an SP5 confocal microscope (Leica Microsystems). The dorsal class IV dendritic arborization neurons (ddaC) in the abdominal segments of control and mutant larvae were imaged in live animals. Live larvae were mounted on slides in a solution of 50% PBS and 50% glycerol and immobilized by pressing a coverslip mounted on top of two lines of vacuum grease spacers flanking the animal. All images obtained from Drosophila larvae were acquired at 25°C from a hybrid detector (Leica Microsystems) using the Leica Application Suite software.
To quantify the number of ManII-positive Golgi outposts in axons, third-instar larvae expressing ManII::GFP and CD4::Tomato in ddaC neurons were imaged at a resolution of 1,024 × 1,024 pixels using a 40 × 1.3 NA oil-immersion objective. Z stacks 10-20 µm thick (1.5 µm per z step) were maximum projected, and the number of ManII::GFP puncta was quantified using the FIJI particle analysis function (ImageJ; National Institutes of Health). The ManII::GFP signal in black-and-white images was inverted and subject to threshold with a cut-off gray value of 170. ManII::GFP-positive puncta between 0.15-15.00 µm in diameter within 75 µm of the cell body (axon) or a 50-µm radius of the cell body (proximal dendrites) were counted. To calculate distal Golgi outposts per dendrite branch, a box of 125 × 250 µm was drawn around the distal tips closest to the dorsal midline. This area was inverted and subject to threshold with a cut-off gray value of 140. ManII::GFP-positive puncta between 0.10-15.00 µm in diameter were counted. The number of tips in the same region was counted using the CD4::Tomato channel.
Golgi outpost colocalization with Klc was determined by simultaneous imaging of ManII::GFP and RFP::Klc for 2-5 min in ddaC dendrites of third instar larvae. Movies were captured at a rate of 0.667 frames/s and a resolution of 1,024 × 512 pixels using a 40 × 1.3 NA oil-immersion objective. Videos were stabilized using the FIJI Image Stabilizer plugin, and kymographs were generated using ImageJ. Kymographs were used to determine colocalization between Golgi outposts and Klc as well as to calculate the directionality and velocity.
Golgi outpost motility was quantified by imaging ManII::GFP puncta for 4-6 min in one or two ddaC neurons per larva at 72 h after egg laying (AEL). Videos were captured at a rate of 0.895 frames/s and a resolution of 1,024 × 512 pixels using a 40 × 1.3 NA oil-immersion objective. ManII::GFP puncta within 100 µm of the cell body in primary and secondary dendrites were analyzed. Videos were stabilized using the FIJI Image Stabilizer plugin, and kymographs were generated using MetaMorph (Molecular Devices). Position and time data from MetaMorph were exported to Excel to quantify directionality. Puncta that traveled consistently in one direction were classified as anterograde or retrograde, whereas those that switched direction were classified as bidirectional. Puncta that did not move during the video were classified as paused. Kymographs were further in analyzed MetaMorph to quantify the number, frequency, velocity, run length, and pause duration of anterograde and retrograde events. The event frequency was calculated only for videos with motile outposts. Particles that moved, paused for at least three frames (2.62 s), and then resumed movement were used to calculate direction changes and determine the frequency of direction switching.
hTfR::GFP was imaged using similar conditions and protocols as described for Golgi outposts. In brief, hTfR::GFP puncta were imaged for 4-6 min in one or two ddaC neurons per larva at 72 h AEL. Movies were captured at a rate of 0.895 frames/s and a resolution of 1,024 × 512 pixels using a 40 × 1.3 NA oil-immersion objective. GFP puncta within 100 µm of the cell body in primary and secondary dendrites were analyzed. Videos were stabilized using the FIJI Image Stabilizer plugin and kymographs were generated using MetaMorph. Position and time data from MetaMorph were exported to Excel to quantify directionality. Robust hTfR::GFP motility and frequent puncta crossings made single-particle tracking impractical. Flux was calculated based on the number of particles that crossed the midpoint of a 75-µm-long kymograph per minute. Particles were classified as anterograde or retrograde based on their directionality as they crossed the midpoint of the kymograph.
Dendrite length, branch points, and Sholl analysis were quantified by imaging two to four ddaC neurons from segments A3-A6 in larvae at 72 and 96 h AEL at a resolution of 2,048 × 2,048 pixels using a 20 × 0.7 NA oil immersion objective. Z stacks 10-15-µm thick (1 µm per z step) were analyzed using the FilamentTracer module in the Imaris software program (Bitplane) to obtain dendrite length, branch point number, and Sholl crossings. Neurons were analyzed individually by masking signals from neighboring neurons. Data were exported to Excel for statistical analysis. Axon branching was assessed for late third instar larvae by counting branch points in individual axons. Axons were split into three categories: unbranched, fewer than five branches, or five and more branches.
Truncated Khc(1-401)::sfGFP localization was quantified by imaging two to four ddaC neurons from segments A3-A5 in third instar larvae at a resolution of 1,024 × 1,024 pixels using a 40 × 1.3 NA oil immersion objective. Z stacks 10-15-µm thick (1 µm per z step) were analyzed using FIJI software to determine the background-subtracted GFP intensity and Tomato signal in the proximal dendrites and axons 50 µm from the cell body. The ratio of GFP/Tomato was calculated for each neurite analyzed.
MT polarity was analyzed by imaging EB1::GFP puncta within 100 µm of the cell body for 4-6 min in one or two ddaC neurons per larva at 72 h AEL. Videos were captured at a rate of 0.895 frames/s and a resolution of 1,024 × 512 pixels using a 40 × 1.3 NA oil-immersion objective. Videos were stabilized using the FIJI Image Stabilizer plugin, and kymographs were generated using MetaMorph. Position and time data from MetaMorph were exported to Excel to quantify directionality.
The fluorescent signal of reconstituted sfGFP in axon terminals expressing Khc::sfGFP11 and sfGFP(1-10) was quantified by determining the background-subtracted, integrated fluorescence intensity of both Khc::sfGFP and CD4::Tomato within a 20 × 50-µm box drawn around commissures (segments A2-A6). The ratio of GFP/Tomato signal was determined for each commissure analyzed. The Khc::sfGFP/Tomato signal in dendrites was similarly quantified in FIJI by drawing a 50-µm-long line starting from the cell body and tracing the primary dendrite with the brightest GFP signal. The mean intensity over 10-µm segments was statistically analyzed.

Antibody staining
To image bruchpilot, cysteine string protein, and synapsin signal in sensory neurons, third instar larvae were dissected in 1× PBS, fixed in 4% paraformaldehyde for 45 min, washed with 1× PBS, permeabilized in 1× PBS, 0.3% Triton X-100 (PBSTx) for 20 min, washed with PBSTx, quenched in 50 mM NH 4 Cl for 10 min, washed with PBSTx, and incubated in block for at least 1 h. Larval fillets were incubated in primary antibody overnight, washed (3 × 30 min in PBSTx), and then incubated in fluorescent secondary antibodies overnight. Fillets were then washed in PBSTx (3 × 30 min) and mounted in a solution of elvanol. Fixed larvae were imaged using a 40 × 1.3 NA oil-immersion objective. Z stacks 20-30 µm thick (1 µm per z step) were analyzed using FIJI. Mouse antibodies that recognize bruchpilot (nc82), cysteine string protein (DCSP-2 6D6), and synapsin (3C11) were obtained from the Developmental Studies Hybridoma Bank.

S2 cell lysates
Drosophila S2 cells were cultured in Schneider's Drosophila medium (Gibco) supplemented with 10% (vol/vol) FBS (HyClone) at 25°C. Constructs expressing Drosophila Khc tagged with mNeo-nGreen at the C terminus in the pMT vector were transfected into cells using Cellfectin II (Invitrogen), and protein expression was induced by adding 1 mM CuSO 4 to the medium 4-5 h posttransfection. The cells were harvested after 48 h.
Motor-expressing S2 cells were harvested and centrifuged at low speed at 4°C. The cell pellet was washed once with PBS and resuspended in ice-cold lysis buffer (25 mM Hepes/KOH, pH 7.4, 115 mM potassium acetate, 5 mM sodium acetate, 5 mM MgCl 2 , 0.5 mM EGTA, and 1% vol/vol Triton X-100) freshly supplemented with 1 mM ATP, 1 mM PMSF, protease inhibitors cocktail (P8340; Sigma-Aldrich), 10% glycerol, and 1 mM DTT. After clarifying the lysate by centrifugation at 16,000 g at 4°C, aliquots were snap frozen in liquid nitrogen and stored at −80°C until further use. The relative amount of motor across lysates was calculated by a dot blot in which serial dilutions of each S2 lysate were spotted onto a nitrocellulose membrane (GE Healthcare). The membrane was air-dried and immunoblotted with a polyclonal antibody to kinesin-1 generated against the motor domain peptide (KLS GKL YLV DLA GSE KVS KTG AEG; Verhey et al., 1998). The amount of Khc in each spot was quantified (ImageJ), and spots within the linear regimen were used to normalize motor levels for addition to single-molecule motility assays.

Single-molecule motility assays
All single-molecule assays were performed at room temperature in a flow cell (∼10 µl volume) prepared by attaching a clean no. 1.5 coverslip to a glass slide with double-sided tape. HiLyte-647labeled MTs were polymerized from a mixture of HiLy647-labeled and unlabeled tubulins (Cytoskeleton) in BRB80 buffer (80 mM Pipes/KOH, pH 6.8, 1 mM MgCl 2 , and 1 mM EGTA) supplemented with 1 mM GTP at 37°C for 15 min. After an addition of five volumes of prewarmed BRB80 containing 20 µM taxol and an additional 15 min incubation at 37°C, polymerized MTs were stored at room temperature in the dark for further use. Polymerized MTs were diluted in BRB80 Buffer containing 10 µM taxol and then infused into flow-cells and incubated for 5 min at room temperature for nonspecific adsorption to the coverslip. Subsequently, Blocking Buffer (1 mg/ml casein in P12 buffer [12 mM Pipes/KOH, pH 6.8, 2 mM MgCl 2 , and 1 mM EGTA] with 10 µM taxol) was infused and incubated for another 5 min. Finally, kinesin motors in the motility mixture (0.5-1 µl of S2 cell lysate, 2 mM ATP, 3 mg/ml casein, 10 µM taxol, and oxygen scavenging [1 mM DTT, 1 mM MgCl 2 , 10 mM glucose, 0.2 mg/ml glucose oxidase, and 0.08 mg/ml catalase] in P12 buffer) was added to the flow cell. The flow cells were sealed with molten paraffin wax and imaged by TIRF microscopy using an inverted microscope Ti-E/B (Nikon) equipped with the perfect focus system (Nikon), a 100× 1.49 NA oil immersion TIRF objective (Nikon), three 20-mW diode lasers (488 nm, 561 nm, and 640 nm) and an electron-multiplying charge-coupled device detector (iXon X3DU897; Andor Technology). Imaging of full-length motors was performed by continuous imaging at 200 ms per frame for 1 min, and imaging of truncated motors was performed by continuous imaging at 100 ms per frame for 30 s. Image acquisition was controlled by Nikon Elements software.
Analysis of single-molecule motility data Maximum intensity projections were generated from the videos, and kymographs were produced by drawing along motor-decorated tracks (width = 3 pixels) using Nikon Elements software. Run length was calculated by taking the distance moved along the x axis of the kymograph. Dwell time was calculated by taking the time of the event along the y axis of the kymograph. Velocity was defined as the distance on the x axis of the kymograph divided by the time on the y axis of the kymograph. Full-length motors frequently paused during motility events, and thus only the run length and velocity between pauses were analyzed. Motility events that started before image acquisition or finished after image acquisition were included in the analysis, and thus the motility parameters for motors with long dwell times are an underestimate of the true values. Motility data from at least two independent experiments were pooled and analyzed.
Relative landing rates were calculated for truncated motors by normalizing motor protein levels across lysates to ensure equal motor input. The relative landing rate was defined as the number of events per unit MT length per unit time and are reported as number of events (µm/s) ± SEM. Only events that lasted at least five frames (500 ms) were counted. Over 10 different MTs from at least two independent experiments were analyzed for each construct.

Statistical analysis
Data were analyzed using Stata software. The Shapiro-Wilk test was used to determine whether data were normally distributed. For normally distributed data, Student's unpaired t tests were used to compare individual samples, and one-way ANO VA with post hoc Tukey (a = 0.01 was used to determine whether a significant difference existed) and t tests were used to compare multiple samples. Significance levels are represented as follows: *, P = 0.05-0.01; **, P = 0.01-0.001; ***, P = 0.001-0.0001; and ****, P < 0.0001. Fig. S1 shows antibody staining of larval fillets to analyze the distribution of synaptic proteins in control and Khc mutant neurons. Fig. S2 shows single-molecule motility results obtained using truncated wild-type and mutant kinesin-1 motors. Fig. S3 shows neuron morphogenesis and Golgi outpost distribution in Khc phosphorylation mutants. Fig. S4 shows additional Golgi outpost and hTfR motility parameters. Table S1 shows all oligonucleotides used in this study.