Microtubule-associated septin complexes modulate kinesin and dynein motility with differential specificities

Long-range membrane traffic is guided by microtubule-associated proteins and posttranslational modifications, which collectively comprise a traffic code. The regulatory principles of this code and how it orchestrates the motility of kinesin and dynein motors are largely unknown. Septins are a large family of GTP-binding proteins, which assemble into complexes that associate with microtubules. Using single-molecule in vitro motility assays, we tested how the microtubule-associated SEPT2/6/7, SEPT2/6/7/9, and SEPT5/7/11 complexes affect the motilities of the constitutively active kinesins KIF5C and KIF1A and the dynein-dynactin-bicaudal D (DDB) motor complex. We found that microtubule-associated SEPT2/6/7 is a potent inhibitor of DDB and KIF5C, preventing mainly their association with microtubules. SEPT2/6/7 also inhibits KIF1A by obstructing stepping along microtubules. On SEPT2/6/7/9-coated microtubules, KIF1A inhibition is dampened by SEPT9, which alone enhances KIF1A, showing that individual septin subunits determine the regulatory properties of septin complexes. Strikingly, SEPT5/7/11 differs from SEPT2/6/7, in permitting the motility of KIF1A and immobilizing DDB to the microtubule lattice. In hippocampal neurons, filamentous SEPT5 colocalizes with somatodendritic microtubules that underlie Golgi membranes and lack SEPT6. Depletion of SEPT5 disrupts Golgi morphology and polarization of Golgi ribbons into the shaft of somato-proximal dendrites, which is consistent with the tethering of DDB to microtubules by SEPT5/7/11. Collectively, these results suggest that microtubule-associated complexes have differential specificities in the regulation of the motility and positioning of microtubule motors. We posit that septins are an integral part of the microtubule-based code that spatially controls membrane traffic.

Long-range membrane traffic is guided by microtubuleassociated proteins and posttranslational modifications, which collectively comprise a traffic code.The regulatory principles of this code and how it orchestrates the motility of kinesin and dynein motors are largely unknown.Septins are a large family of GTP-binding proteins, which assemble into complexes that associate with microtubules.Using singlemolecule in vitro motility assays, we tested how the microtubule-associated SEPT2/6/7, SEPT2/6/7/9, and SEPT5/ 7/11 complexes affect the motilities of the constitutively active kinesins KIF5C and KIF1A and the dynein-dynactin-bicaudal D (DDB) motor complex.We found that microtubule-associated SEPT2/6/7 is a potent inhibitor of DDB and KIF5C, preventing mainly their association with microtubules.SEPT2/6/7 also inhibits KIF1A by obstructing stepping along microtubules.On SEPT2/6/7/9-coated microtubules, KIF1A inhibition is dampened by SEPT9, which alone enhances KIF1A, showing that individual septin subunits determine the regulatory properties of septin complexes.Strikingly, SEPT5/7/11 differs from SEPT2/6/7, in permitting the motility of KIF1A and immobilizing DDB to the microtubule lattice.In hippocampal neurons, filamentous SEPT5 colocalizes with somatodendritic microtubules that underlie Golgi membranes and lack SEPT6.Depletion of SEPT5 disrupts Golgi morphology and polarization of Golgi ribbons into the shaft of somato-proximal dendrites, which is consistent with the tethering of DDB to microtubules by SEPT5/7/11.Collectively, these results suggest that microtubule-associated complexes have differential specificities in the regulation of the motility and positioning of microtubule motors.We posit that septins are an integral part of the microtubule-based code that spatially controls membrane traffic.
Long-range transport of membrane vesicles and organelles takes place on the microtubule cytoskeleton, a meshwork of α-/ β-tubulin polymers, and is mediated by the kinesin and dynein motors (1, 2).Spatial navigation of the microtubule network is critical for the accurate and timely delivery of cargo, but how motor movement is spatially controlled on microtubules is poorly understood.Growing evidence indicates that microtubules are a heterogeneous network, a mosaic of polymers with varying compositions of tubulin isotypes, posttranslational modifications, and microtubule-associated proteins (MAPs) (3)(4)(5).Collectively, this biochemical diversity provides a spatial code, which directs membrane traffic by determining the microtubule tracks and motile properties of motors and their cargo (6)(7)(8)(9).
In vitro reconstitution of kinesin and dynein motility has begun to provide key insights into the mechanisms by which the binding and movement of motors is regulated on microtubules.Recent work has revealed that MAPs selectively and differentially regulate the motility of kinesin motors (5,7,10,11).MAP7/ensconsin promotes the recruitment and motility of kinesin-1, which is inhibited by most other MAPs (12)(13)(14)(15)(16)(17)(18).Similarly, the motility of the kinesin-3 motor KIF1A is restricted or permitted by different MAPs (13,(19)(20)(21).Dynein-dynactin motor complexes are also differentially affected by MAPs (12)(13)(14)(22)(23)(24)(25).These findings have raised the importance of a comprehensive understanding of the selective regulation of motors by MAPs, which guides the movement and positioning of membrane cargo in the microtubule network.
Septins are GTP-binding proteins, which assemble into multimeric nonpolar oligomers and filaments that associate with microtubules and actin filaments (26)(27)(28).Mammalian septins comprise a large family of 13 paralogous genes, which are categorized into four distinct groups based on sequence similarity (29).Septin complexes consist of subunits from each of the four groups, which can assemble minimally into heterohexamers or hetero-octamers, head-to-head dimers of heterotrimers or hetero-tetramers (30)(31)(32)(33).Subunits of the same septin group can replace one another within their corresponding positions, generating a diversity of complexes (34,35).Alternative complexes with multiple subunits from the same group have been reported and may arise from disproportionate expression of different septins in certain cell types (36).The evolutionarily expansion of septin genes and isoforms suggests that septin subunits bestow distinct properties and functions upon their respective complexes (37)(38)(39)(40).However, the functional specificity of distinct septin subunits and complexes is little understood and explored in mammalian cells.
Mammalian septins associate with subsets of microtubules regulating their spatial organization and dynamics (41).In hippocampal neurons, septin 9 (SEPT9) localizes to dendritic microtubules and reinforces the polarity of membrane traffic by impeding the transport of axonal cargo of kinesin-1 (KIF5C) and promoting the movement of dendritic cargo of kinesin-3 (KIF1A) (19,42).In vitro motility assays revealed that microtubuleassociated SEPT9 differentially regulates kinesin motility, inhibiting KIF5C, enhancing KIF1A, and having no impact on kinesin-2 KIF17 (19).Additionally, SEPT9 reduced the velocity and run length of the activated dynein-dynactin-bicaudal D (DDB) motor complex (19).Unexpectedly, these findings showed that a microtubule-associated septin can differentially modulate the motility of microtubule motors, raising the questions of whether SEPT9 functions similarly in heteromeric septin complexes and whether different septin complexes have distinct regulatory properties.Investigating these questions may not only shed light on the functional specificity of septin complexes and subunits but also reveal a septin-based code for the spatial control of membrane traffic (43).
Here, we sought to examine how three microtubuleassociated septin complexes with distinct subunit compositions affect the motility of kinesin and dynein motors.Using in vitro reconstitution assays of motor motility, we compared the microtubule-associated SEPT2/6/7 and SEPT5/7/11 with one another, and we also asked how septin subunits with opposing effects on KIF1A motility function when they are part of the same complex.We found that SEPT5/7/11 differs from SEPT2/6/7 in being largely permissive to KIF1A motility and promoting microtubule tethering of dynein and kinesin motors.We also report that in SEPT2/6/7/9 complexes, the SEPT9 and SEPT2/6/7 subunits counter each other in enhancing and inhibiting KIF1A motility, which results in a blended effect.In agreement with the distinct properties of SEPT5/7/11 complexes, SEPT5 localizes to a subset of neuronal microtubules that lack SEPT6 and promotes Golgi ribbon morphology and localization in the somatodendritic compartment of hippocampal neurons.Collectively, our results reveal that microtubule-associated septins have differential specificities in modulating kinesin-and dynein-driven motility, which are determined by the combinatorial identity of their subunits.
Sept5/7/11 permits KIF1A motility while inhibiting KIF5C and DDB Given that the motility of KIF1A is differentially impacted by SEPT9 and SEPT2/6/7, we reasoned that septin complexes may have distinct regulatory properties based on the composition and identity of their subunits.To test this hypothesis, we generated a recombinant mCherry-tagged septin complex consisting of septins 5, 7, and 11 (Fig. S1C), which has been identified as a unique complex in neurons (52,53).Incubation of taxol-stabilized microtubules with 50 nM mCherry-SEPT5/ 7/11 resulted in uniform and saturable coating of microtubules, which contained SEPT7 and SEPT11 along with mCherry-SEPT5 (Fig. S1, J-L).
Neuronal Golgi ribbons align with SEPT5-coated microtubules and depend on SEPT5 for their somatodendritic localization The differential effects of SEPT2/6/7 and SEPT5/7/11 on kinesin and DDB motility suggest that these septin complexes may have distinct physiological functions.We explored this possibility by probing for the localization of SEPT2/6/7 and SEPT5/7/11 in the microtubule network of hippocampal neurons, where SEPT5/7/11 was originally identified as a biochemically distinct complex (52).Primary embryonic rat hippocampal neurons, which had developed axonal and dendritic processes after 14 days in culture (DIV14), were stained with antibodies against SEPT5 and SEPT6 as proxy subunits for SEPT5/7/11 and SEPT2/6/7, respectively.To better resolve microtubules and septins in the somatodendritic compartment, we extracted neurons prior to fixation and costained for the MAP2 which decorates the microtubule bundles of dendrites.Using shallow angle TIRF microscopy, we observed SEPT5 colocalizing with MAP2-labeled microtubules in the somato-proximal segments of dendritic shafts (Figs. 5, A and B and S4A).SEPT6, however, was largely absent from the long MAP2-labeled tracks and consisted of short filaments and puncta (Figs. 5, C and D and S4B).Thus, SEPT5-containing complexes are enriched on dendritic microtubule bundles, which lack SEPT6.
We next probed for the localization of the Golgi complex with respect to the dendritic SEPT5-coated microtubules.We focused on the neuronal Golgi, because following axon initiation, it polarizes toward the base of the presumptive apical dendrite and Golgi ribbons align along the microtubules of the somato-proximal segment of dendrites-a phenomenon known as Golgi deployment that precedes the formation and dispatch of Golgi outposts (54)(55)(56)(57).How Golgi ribbons assume their polarized localization is poorly understood.We reasoned that SEPT5/7/11 complexes might play a role because of the enrichment of SEPT5 on somato-proximal microtubules and the in vitro tethering of DDB motors to SEPT5/7/11-coated microtubules.Of note, DDB associates with Golgi membranes and is critical for Golgi morphology and localization (58)(59)(60).
Using wide-field microscopy with advanced optical clearing and super-resolution confocal microscopy, we imaged the localization of the Golgi marker protein GM130 with respect to SEPT5 in rat hippocampal neurons.We observed Golgi stacks anastomosing with the somato-proximal ends of SEPT5 tracks, which extend into the dendritic shaft (Fig. 5, E and F).
In confocal 3D image stacks, 66 ± 3% (n = 12 neurons) of Golgi volume overlapped with SEPT5, and the Manders overlap coefficient was 0.55 ± 0.11.In dendrites with Golgi ribbons, which are deployed well into the shaft, there was a tight coalignment of GM130-labeled tubules with SEPT5enriched tracks, which appeared to be zippered together (Fig. 5G).Golgi ribbons made no contacts with SEPT6 filaments in dendritic shafts, but occasional end-to-end and orthogonal contacts were observed in the cell body (Fig. S4, B  and C).
containing complexes have a distinct function in neuronal Golgi localization and organization, which is in agreement with the in vitro immobilization of dynein and kinesin on SEPT5/7/11-coated microtubules.

Discussion
In the microtubule network, spatial control of membrane traffic and organelle positioning is largely mediated by MAPs and the post-translational modifications (PTMs) of microtubules (4-7).Subsets of microtubules have unique combinations of MAPs and PTMs, which determine the type of motors that bind a microtubule and their motile behaviors (3,8,10,65).Recent work has begun to shed light on the selective regulation of kinesin and dynein motility by distinct MAPs, some of which can function as either suppressors or enhancers of motor motility (12)(13)(14)(15)(19)(20)(21).In combination with microtubule PTMs, these MAPs determine the intracellular routes and positions of various membrane cargos.
We previously identified SEPT9 as a MAP that differentially regulates motility of kinesins KIF5C, KIF1A, and KIF17 (19).This finding raised the possibility that other members of the septin family selectively regulate kinesin and dynein motility (42).Here, we sought to explore this hypothesis by testing how two different septins complexes (SEPT2/6/7, SEPT5/7/11) impact the kinesins KIF5C and KIF1A and the dynein motor complex DDB.Moreover, we tested whether septin subunits influence the collective property of their respective complexes by examining how the SEPT2/6/7/9 complex impacts KIF1A motility, which is enhanced by SEPT9 but inhibited by SEPT2/6/ 7 (19).This is an important question of broader significance for septin biology as it remains poorly understood how the function of septin complexes is determined by their individual subunits (66).In rat hippocampal neurons, the existence of a SEPT5/7/11 complex has been independently confirmed (52,53,63), but it is unclear whether it has distinct or overlapping functions with SEPT2/6/7.SEPT2 and SEPT6 can replace SEPT5 and SEPT11, respectively, in the SEPT5/7/11 complex, though SEPT2 expression is reportedly lower than the other septin subunits Figure 6.Golgi localization and morphology are disrupted by SEPT5 depletion.A, images show Golgi (GM130; magenta) localization with respect to MAP2 (green)-labeled dendrites in primary embryonic rat hippocampal neurons (DIV14) after a 4-day transfection with plasmids that express GFP (blue) and shRNAs.Arrowheads point to Golgi tubules which are deployed into dendrites.Scale bars represent 10 μm.B-D, bar graphs show quantification of Golgi polarization (B), which was scored as Golgi localization at the base of a single dendrite, and Golgi deployment into a dendrite (C; Golgi tubules stretching along the beginning of the dendritic shaft).Golgi morphology (D) was categorized into clustered/condensed, fragmented, or tubulated with ribbon-like appearance.Quantifications were performed in rat hippocampal neurons (DIV14) after 3 days of transfection with control shRNA (n = 75) and shRNAs against SEPT5 (n = 36) and SEPT6 (n = 99).Statistical analysis for pairwise comparisons was done with the chi square test (n.s., not significant; *p < 0.05; **p < 0.01; ****p < 0.0001).The chi square group test result for Golgi polarity was X 2 (2, N = 30) = 13.67,p = 0.001076.The chi square group test result for Golgi deployment was X 2 (2, N = 30) = 11.94,p = 0.00251 and for Golgi morphology, X 2 (4, N = 45) = 69.08,p = 0.00001.Scale bars represent 10 μm.MAP, microtubule-associated protein.(63,67,68).In the neuronal microtubule network, the localization and functions of these septin complexes have not been explored.Because microtubule lattices are coated with combinations of MAPs, heteromeric septin complexes can serve as a model for how combinations of MAPs regulate kinesin and dynein motility along the microtubule lattice.
Our results reveal that the microtubule-associated SEPT5/ 7/11 complex has two fundamental differences from SEPT2/6/ 7. In contrast to SEPT2/6/7, which strongly inhibits the microtubule binding and motility of KIF5C, KIF1A, and DDB, SEPT5/7/11 is first permissive to the microtubule binding and motility of KIF1A and second, enhances the fraction of microtubule-tethered motors that remain immotile without processive movement.The latter anchoring-like effect was most pronounced for DDB, which exhibited no motility while bound on microtubules coated with a SEPT5/7/11 density that allowed kinesin movement, albeit at lower levels.In contrast to SEPT5/7/11, microtubule-associated SEPT2/6/7 and SEPT2/ 6/7/9 inhibited microtubule-DDB binding and thus, there was no DDB dwelling on microtubules.
Tethering of motors to the microtubule ends and lattice is critical for cellular structures such as motile cilia and flagella and the positioning of the spindle during mitosis (69)(70)(71)(72).DDB motility is inhibited on microtubules with detyrosinated α-tubulin, and dynein accumulates at microtubule minus ends (73,74).Previous studies suggested that MAPs such as MAP2, MAP4, tau, and MAP7 inhibit dynein motility (13,14,(21)(22)(23)(24)(25), but some of the evidence is controversial and it is unclear whether these MAPs can anchor cytoplasmic dynein to the microtubule lattice.Mechanistically, SEPT5/7/11 may tether DDB in a similar manner to how MAP7 keeps KIF5 tethered to the microtubule lattice while occluding the motor domain of KIF5 from stepping due to mutually exclusive microtubule-binding sites (14).In support of this possibility, we found a SEPT5/7/11 concentration-dependent increase in DDB immobilization, which occurs concomitantly with a concentration-dependent decrease in processive movement.
Dynein immobilization and inhibition of its motility on microtubules is hard to achieve owing to the ability of dynein to maneuver around MAPs by side-stepping and/or moving backwards (22,51,75).Septins appear to be potent inhibitors of DDB, but they differ mechanistically with SEPT2/6/7 and SEPT2/6/7/9 blocking microtubule binding and SEPT5/7/11 immobilizing dynein upon microtubule attachment.Independently of the mechanism, this inhibition is stronger for DDB than KIF5C and KIF1A, which are motile on microtubules coated with septin concentrations that do not permit DDB motility.On intracellular microtubules, we posit that these septin complexes may bias membrane traffic toward microtubule plus ends, being more restrictive to dynein-than kinesin-driven motility.Moreover, the motility of KIF1Abound cargo might be selectively favored on microtubules with SEPT5/7/11 and SEPT2/6/7/9, which are more permissive to KIF1A than DDB and KIF5C.Thus, septin complexes may impact bidirectional transport, resolving the tug-of-war between groups of cargo-bound motors with opposing directionalities.
The SEPT2/6/7/9 complex is less potent than SEPT2/6/7 in reducing the velocity of KIF1A and enhancing its pausing, which is due to the presence of SEPT9, a KIF1A enhancer.Addition of SEPT9 to SEPT2/6/7 dampens the inhibitory effects of SEPT2/6/7 on motile KIF1A motors, but it does not appear to reduce the inhibition of KIF1A landing on microtubules by SEPT2/6/7.At the microtubule-bound density of SEPT9, which previously enhanced KIF1A motility, SEPT2/6/ 7/9 caused a modest decrease in the landing rates and run lengths of KIF1A and slightly enhanced KIF1A velocity.Thus, SEPT2/6/7 counteracts SEPT9 as an enhancer of KIF1A motility.We posit that this combinatorial blending of inhibitory and enhancing effects may underlie the modulation of kinesin and dynein motility on microtubule lattices, which are coated by combinations of different MAPs.
The distinct properties of SEPT5/7/11 and in particular the microtubule tethering of DDB led us to examine whether SEPT5/7/11 has a septin-specific role in the positioning of the Golgi complex in neurons.We focused on the Golgi complex because its morphology and distribution are largely dependent on dynein (76).Additionally, septins have been previously implicated in Golgi organization and post Golgi membrane traffic in non-neuronal cells (77,78).We found Golgi stacks and ribbons localizing along SEPT5-coated microtubules at the somato-proximal segments of neuronal dendrites.SEPT6 was absent from these microtubule tracks, which indicated that SEPT5/7/11 may have a septin-specific function in Golgi association with the microtubules of proximal dendrites.Indeed, SEPT5 depletion resulted in the loss of Golgi polarization and Golgi ribbon localization along the basal segments of dendrites, while SEPT6 depletion did not have an effect.These data indicate that SEPT5/7/11 complexes promote tethering of Golgi membranes on a subset of microtubules that span the somato-proximal segment of dendrites.
In vitro tethering of DDB to SEPT5/7/11-coated microtubules suggests that SEPT5/7/11 immobilizes Golgi-associated dynein motors on somatodendritic microtubules.Golgi membranes associate with dynein via BICD adapters, and thus neuronal Golgi localization may involve regulation of DDB motility by SEPT5/7/11 (58)(59)(60).In parallel to tethering of Golgi-bound DDB to microtubules, SEPT5/7/11 may prevent association of Golgi-bound kinesin-1 motors with microtubules.Inhibition of kinesin-1 is critical for the dendritic localization of Golgi outposts (79), and inhibition of the motor domain of KIF1C prevents Golgi fragmentation (80).We posit that microtubule-associated SEPT5/7/11 tethers and/or constrains the motility of Golgi stacks and ribbons, which is conducive to the homotypic fusions that maintain the Golgi ribbon morphology (81,82).In the absence of a restricted motility, Golgi membranes would haphazardly scatter throughout the soma, leading to a dispersed morphology of Golgi in SEPT5-depleted cells.
In sum, our results have revealed that microtubuleassociated complexes can differentially and selectively modulate the motility of kinesin and dynein-dynactin motors.Importantly, this differential regulation stems from the properties of the individual septin subunits, which function combinatorially in modulating the microtubule attachment, dwelling stepping of microtubule motors.The modularity of septin assembly and the existence of multiple septin subunits with various isoforms point to a septin code, which functions in synergy with the tubulin and MAP codes for the spatial control of membrane traffic and organelle positioning.

Kinesin and DDB motors
COS-7 cells were transfected with truncated constitutively active motor domain constructs KIF1A(1-393)-GCN4-3xmCit or KIF5C(1-560)-mCit using Lipofectamine 2000 (Thermo Fisher Scientific).After 24 to 48 h, media was removed, cells were gently washed with ice cold PBS or buffer A (30 mM Hepes pH 7.4, 50 mM K-acetate pH 7.4, 2 mM Mg-acetate, 1 mM EGTA pH 7.4, 10% glycerol), and incubated for 20 min in lysis buffer (buffer A, 1 mM PMSF, 0.5% Triton X-100, EDTA-free protease inhibitor III (Calbiochem),1 mM ATP).Cell lysis was followed by centrifugation of extracts at 16,000g for 10 min.Additional ATP was added to extracts at a final concentration of 1 mM, and small aliquots were flash frozen for long term storage at −80 C. DDB purification and motility assays were performed as previously described with the exception that endogenous dynein-dynactin adapter complexes were isolated from HEK293T cells rather than RPE-1 (84).In brief, bacteria were transformed with the plasmid pET28a-Strep-sfGFP-BICD2N, and the purified sfGFP-BICD2N was mixed with HEK293 lysates and Strep-Tactin beads overnight at 4 C. Beads were washed and eluted with 3 mM desthiobiotin and 0.5 mM ATP.
Microtubule coating with recombinant septin complexes and immunolabeling of microtubule-bound septins Acid-washed glass coverslips (0.15 mm thick) were mounted on a glass slide using double sided tape to create 8 to 10 μl motility chambers.Taxol-stabilized microtubules were prepared by incubating unlabeled (80%), HiLyte647 (10%), and biotin conjugated (10%) porcine brain tubulin (Cytoskeleton Inc) in BRB80 (80 mM Pipes pH 6.9, 1 mM EGTA pH 6.9, 2 mM MgCl2, 10% glycerol) supplemented with 1 mM GTP at 37 C for 35 min.The concentrated microtubule mix was incubated at 37 C for 20 min following the addition of 10 μM taxol and then kept at room temperature while protected from light.To adhere biotinylated microtubules to the glass, each chamber was incubated with 5 mg/ml biotinylated bovine serum albumin (BSA) (Sigma) followed by 0.5 mg/ml neutravidin (Thermo Fisher Scientific) which was diluted in cold PBS.Chambers were incubated with taxol-stabilized microtubules diluted in BRB80 containing 1 mM GTP for 15 min, followed by blocking buffer (BRB80, 1 mg/ml BSA, 1% w/v Pluronic F-127 (Sigma), 10 μM taxol) for an additional 5 min.Chambers were then washed with 80 to 100 μl Hepes buffer (30 mM Hepes-KOH pH 7.4, 50 mM of KOAc pH 7.4, 2 mM of MgOAc, 1 mM of EGTA-KOH pH 7.4, and 10% glycerol).Septin complexes were diluted in the Hepes buffer supplemented with 0.1% w/v Pluronic F-127, 0.1 mg/ml BSA, and 10 μM Taxol (Hepes dilution buffer) before applying to the chambers and incubated for 15 min at room temperature.A final wash with 80 to 100 μl Hepes dilution buffer was done before imaging with TIRF microscopy using the DeltaVision OMX V4 imaging platform (GE Healthcare) with 60X/1.49NA objective (Olympus), sCMSO pco.edge cameras.
After microtubule coating with septin complexes, microtubule-bound septins were immunolabeled by first blocking for 10 min at room temperature with Hepes buffer supplemented with 1% w/v Pluronic F-127, 2% BSA, and 10 μM taxol.Prior to labeling, rabbit anti-SEPT6 (S6CU; a gift from Dr Makoto Kinoshita, Nagoya University), rabbit anti-SEPT7 (IBL America 18991), rabbit anti-SEPT9 (ProteinTech 10769-I-AP), or rabbit anti-SEPT11 (MilliporeSigma ABN1342) antibodies were diluted in Hepes buffer and conjugated with 5 μl of the Zenon Alexa Fluor 488 rabbit IgG labeling reagent for 5 min at room temperature.The Zenon blocking reagent (rabbit IgG; 5 μl) was then added to the mix and incubated for 5 min at room temperature before further dilution in 30 μl of Hepes buffer.The diluted mix was applied to the chamber and incubated for 15 min at room temperature before a final wash with 80 to 100 μl of Hepes buffer.

Single molecule motility assays
Stabilized microtubules were made by incubating unlabeled (80%), HiLyte647 (10%), and biotin-conjugated (10%) porcine brain tubulin (Cytoskeleton Inc) in BRB80 (80 mM Pipes pH 6.9, 1 mM EGTA pH 6.9, 2 mM MgCl2, 10% glycerol) supplemented with 1 mM GTP at 37 C for 0.5 to 1 h.The concentrated microtubule mix was incubated at 37 C for 15 to 30 min following the addition of Taxol (10 μM) and then kept at room temperature while protected from light.Acid-washed glass coverslips (0.15 mm thick) were mounted on a glass slide using double sided tape in order to create 8 to 10 μl motility chambers.To adhere biotinylated microtubules to the glass, each chamber was incubated with 5 mg/ml biotinylated BSA (Sigma) followed by 0.5 mg/ml Neutravidin (Thermo Fisher Scientific) which was diluted in cold PBS.Chambers were incubated with taxol-stabilized microtubules diluted in BRB80 containing 1 mM GTP for 10 min, followed by blocking buffer (BRB80, 1 mg/ml BSA, 1% w/v Pluronic F-127 (Sigma), 10 μM taxol) for an additional 10 min.
Extracts containing kinesin motors were diluted in buffer A supplemented with (0.1 mg/ml BSA, 0.1% Pluronic F-127, 2 mM ATP, 10 μM taxol) and a glucose-based oxygen scavenging system (0.035 mg/ml catalase, 4.5 mg/ml D-glucose, 0.2 mg/ml glucose oxidase, 30 mM b-mercaptoethanol in buffer A) in the final motility mix.For motility assays performed in the presence of septin complexes, blocked microtubules were incubated with mCherry-SEPT2/6/7, mCherry-SEPT5/11/7, or mCherry-SEPT2/6/7/9 diluted in buffer A supplemented with 0.1 mg/ml BSA, 0.1% Pluronic F-127, and 10 μM taxol for 10 min.Chambers were washed with buffer A prior to the addition of the final motility mix containing kinesin motors.The motility of microtubule motors was recorded using time-lapse TIRF microscopy and was performed at room temperature (kinesins acquired at five frames/s for 2 min, DDB acquired at one frame/s for 2 min).

Immunofluorescence
Cultured DIV10-DIV14 neurons were fixed with warm fixation buffer containing PBS supplemented with 4% sucrose and 3% paraformaldehyde for 10 min, permeabilized with 0.1% Triton X-100 in GDB (0.2% gelatin, 450 mM NaCl, 30 mM sodium phosphate pH 7.4) for 10 min, and blocked with GDB for an additional 30 min (Fig. 6).Primary and secondary antibodies were diluted in GDB and centrifuged at 100,000g at 4 C for 10 min before adding onto cells.Following a blocking step, primary antibodies were incubated overnight at 4 C.The following day, cells were washed with GDB and secondary antibodies diluted in GDB were incubated at room temperature for 1 to 1.5 h.Neurons transfected with pSUPER plasmids were also incubated with 1 mg/ml DAPI (Sigma) for 10 min prior to mounting.FluorSave mounting medium (EMD Millipore) was used to preserve fixed samples, which were allowed to dry for 1 to 2 h before imaging or storage at 4 C overnight.

Microscopy
All TIRF microscopy (in vitro motility assays, Fig. 5, A-D) was performed with the TIRF module of a Deltavision OMX V4 inverted microscope using a 60X/1.49NA oil immersion objective and cCMOS pco.edge cameras (PCO) and the soft-WoRx software.In vitro motility assays were performed at room temperature.In Figure 5, E and F, images DIV10 neurons were acquired on a Leica THUNDER imager with an inverted DMI8 microscope stand, HC PL APO 63x/1.40 oil objective, and a Leica-K8-A21M726013 camera.Images were processed with the Large Volume Computational Clearing module of the Leica LAS X software (https://www.leica-microsystems.com/products/microscope-software/p/leica-lasx-ls/).In Figure 5G, images were acquired with the Leica Stellaris 5 confocal using a 63X/1.4NA oil immersion objective, zoom of 1X-3X, and 0.2 μm step size for z-stack collection.Images were processed with LIGHTNING convolution module for enhanced lateral resolution.In Figures 6 and 3D, images of DIV14 neurons were acquired on a Zeiss AxioObserver Z1 inverted microscope equipped with a 63x/1.4NA oil immersion objective and a Hamamatsu Orca-R2 CCD camera.

Quantification of microtubule-bound septin complexes
Quantification of the mCherry fluorescence of recombinant septin complexes was performed in Fiji/ImageJ using the plot profile function.A straight line with thickness of 1 was drawn along individual microtubules with >10 μm length.For fluorescence background subtraction, a 10 μm-long straight was drawn at a distance of at least 2 μm away from any microtubule.The mean fluorescence intensity of each microtubule was plotted after subtracting the mean fluorescence intensity of the background.All statistical analyses were performed in GraphPad Prism software 9 (https://www.graphpad.com/)and Microsoft Excel to calculate mean values, S.E.M., and SD, and statistical significance (p values) was derived using a Welch's one way ANOVA test with post hoc Dunnett's T-3 test for multiple pairwise comparisons.

Analysis of in vitro motility
Individual motors were tracked manually using frame-byframe analysis in ImageJ/Fiji.The movement path of each motor (start and end frame) was traced using the segmented line tool, saved as a region of interest, and used to determine motor run lengths.Only kinesin motors that moved processively for at least 0.6 s (3 frames) and DDB motors that were processive for at least 3 s were tracked.For both kinesin and DDB, only microtubules of 10 μm length or longer were considered for analysis and velocities were determined by dividing the run length by total event duration.Landing rates, pausing, and immotile events were determined by generating kymographs created using a region of interest along the length of each microtubule.Landing events were defined as events of motor landing onto the microtubule lattice followed by sustained unidirectional movement that lasted for more than three consecutive frames without reversal of direction or diffusive movement and resolved by a diagonal line in kymographs of movement.The number of landing events were scored for individual microtubules and divided by the length of the microtubule and the duration of the video to extract the landing rates.Pause events were defined as processive events which contained at least a single pause for three frames (600 ms) or longer.The percentage of pausing events was calculated by dividing the total number of pausing events by the total number of their respective processive events per microtubule and multiplying it by 100.Immotile events were defined as particles that bound to the microtubule for at least three frames (600 ms) or more and did not move processively at all.The frequency of immotile events was calculated by dividing the number of immotile events by the length of the microtubule and the duration of the time-lapse capture.

Golgi polarity and morphology analysis
DIV14 neurons transfected with pSUPER plasmids (GFP) were stained with antibodies to GM130, which labeled the Golgi complex, MAP2 to mark neuronal dendrites, and DAPI.Cells with condensed DAPI-labeled chromosomes and other brightly labeled nuclear abnormalities, which are indicative of apoptosis, were excluded from analysis.Golgi complexes (GM130) that were positioned at the base of a single MAP2labeled dendrite were scored as polarized, and Golgi complexes which span the entry points of multiple dendrites were scored as nonpolarized.Golgi complexes that were either scattered throughout the cell body or condensed near the nucleus were also scored as nonpolarized.Morphology of the Golgi complex was classified as tubulated/ribbon if it consisted of elongated ribbons, clustered/condensed if Golgi ribbons or stacks were tightly coalesced into a kidney-shaped bolus, and fragmented if consisted of numerous smaller stacks that were disconnected and dispersed throughout the cell body and dendrites.The Golgi was categorized as deployed if tubular ribbons were extended at least five microns into the shaft of one or more dendrites.Within each classification (polarity, morphology, deployment), the percentage of cells was determined by dividing the n in each category by the total number of cells scored and multiplying by 100.

Quantification of Golgi (GM130) colocalization with SEPT5
Three-dimensional z-series stacks of confocal microscopy images of DIV10 hippocampal neurons were imported into the IMARIS (9.9.1) analysis software (https://imaris.oxinst.com/).Background subtraction was performed on both the GM130 and SEPT5 channels with a filter width of 1 μm.A 3D mask was generated for each channel by thresholding fluorescence signals to a range that best fit the raw signal without including background fluorescence.Prior to fluorescence segmentation, the approximate center z-plane was determined visually by selecting the plane where the majority of the GM130 signal was visible.The colocalization tool was then used to generate a channel representing the 3D overlap between the GM130 and SEPT5 channels.Using this channel, the IMARIS software automatically generated the percentage volume overlap and Mander's coefficients between the GM130 and SEPT5 channels.

Quantification of dynein levels on Golgi (GM130) stacks
Quantifications were performed on the midplane optical sections of three-dimensional z-series stacks, which were collected with the Zeiss AxioObserver microscope and Slidebook 6.0 software (https://www.intelligent-imaging.com/ slidebook).The center of the z-plane was determined visually by selecting the plane where the majority of the GM130 signal was visible.Fluorescence thresholding was used to create a mask around the GM130 and dynein signals after applying no-neighbors deconvolution to improve signal-to-noise ratio.Due to background staining of the nucleus with the rabbit GM130 antibody, the nuclear area was excluded from quantification by generating a second mask that overlapped with the nuclear area.Using the mask operations tool, the nuclear mask was subtracted from the Golgi mask to generate the Golgi area where dynein fluorescence signal was quantified.The total surface area of the masked Golgi (GM130) ribbons and stacks and the sum fluorescence intensity of dynein were derived with the mask statistics tool of the Slidebook 6.0 software.

Statistical analysis
The GraphPad Prism 8 or Microsoft Excel software was used to calculate mean values, SEM, SD, and to perform statistical tests for deriving p values.Data were first analyzed with the D'Agostino and Kolmogorov-Smirnov tests to test for normal distribution.For pairwise comparisons of data with normal distributions, a student's t test was used if SDs were equal or a Welch's t test if SDs were unequal.The Mann-Whitney test was used for non-normally distributed data.For statistical analysis of multiple groups with normally distributed data, a one-way ANOVA was performed with a post hoc Dunnett test for multiple pairwise comparisons.For data that were not normally distributed, a nonparametric Kruskal-Wallis ANOVA test was performed with a post hoc Dunn's test for pairwise comparisons.In Figure 6, categorical data were statistically analyzed with the chi square test.Statistical significance (p values) of pairwise comparisons was derived in GraphPad Prism, while statistical significance for the entire group cohort was derived using an automated online chi square calculator (https://www.socscistatistics.com/tests/chisquare2/default2.aspx) and results were reported in the format: X 2 (degrees of freedom, N = sample size) = chi square statistic value, p = p value.Results with p values of < 0.05 were considered to be statistically significant.Data with p values < 0.05 were determined as statistically significant.

Figure 5 .
Figure 5. Golgi membranes align along SEPT5-coated microtubules in neuronal dendrites.A and B, representative TIRF microscopy images of a primary rat hippocampal neuron (DIV14), which was stained with antibodies to SEPT5 (A; inverted forest green), SEPT6 (B; inverted forest green), and MAP2 (inverted magenta).The same cell is stained and shown in panels A and B, and the MAP2 image is reused.Areas in dashed rectangles (1, 2) are shown in higher magnification.Yellow lines 2 and 3 designate the linescan regions in C and D, respectively.Scale bars represent 5 μm and 1 μm (magnified regions).C and D, plot profiles of the fluorescence intensities of SEPT5, SEPT6, and MAP2 across the lines 2 (C) and 3 (D) shown in panels A and B, respectively.Dashed vertical lines mark peaks of fluorescence that correspond to microtubule bundles.E and F, deconvolution wide-field microscopy images of embryonic rat hippocampal neurons (DIV10) stained for GM130 (magenta) and SEPT5 (green).Arrows point to GM130-labeled Golgi membranes, which are docked on SEPT5 filaments at the base of MAP2-positive (not shown) dendrites.Scale bars represent 5 μm and 2 μm (magnified regions).G, super-resolution confocal microscopy of a primary embryonic rat hippocampal neuron (DIV10), which was stained for SEPT5 (green), MAP2 (cyan), and GM130 (magenta) shows tubular Golgi membranes deployed into the principal dendrite.Regions outlined with dashed lines (I, II) are shown in higher magnification.Scale bars represent 5 μm and 1 μm (magnified regions).MAP, microtubule-associated protein; TIRF, total internal reflection fluorescence.