Molecular Basis for Specific Regulation of Neuronal Kinesin-3 Motors by Doublecortin Family Proteins

1Microtubule (MT)-based transport uses molecular motors to carry cargos over long cellular distances within neurons. The large number of MT motors, especially kinesins (encoded by 45 genes in humans) with diverse cargo specificities, provides a potential means of fine regulation of trafficking, but it is not fully understood how MT-based transport systems achieve specificity with regard to cargo load and targeted transport to specific domains within the neuron. Interaction with MT-associated proteins (MAPs) has been proposed as one means to target transport through complex neuronal structures, as some MAPs show spatially restricted localization in either dendrites or axons. The molecular basis, potential regulatory impact, and degree of specificity of such MAP-motor crosstalk at the MT surface, however, are unknown.

2Mutations in a gene encoding an unusual MAP, doublecortin (Dcx), cause a neuronal migration disorder leading to intellectual disability and epilepsy. Dcx and related doublecortin domain protein genes, including doublecortin-like kinase 1 (Dclk1), encode proteins with tandem MT binding domains, referred to as N-DC (or R1) and C-DC (or R2), but the exact role of each of these dual domains is unknown. In contrast to other MAPs that bind directly on the surface of the MT protofilament, existing evidence demonstrates Dcx binding in the recess between protofilaments.

3Although the migratory disruption caused by mutations in Dcx has widely been regarded as a defect in cytoskeletal regulation, Dcx/Dclk1-deficient neurons also show defects in the transport of presynaptic vesicles in the absence of comparable defects in MT organization. The transport deficiency suggests an attractive alternative hypothesis, that Dcx/Dclk1 may regulate transport of membrane and cellular components, perhaps through kinesin motor proteins, and that specific trafficking of membrane constituents to various neuronal domains may in turn regulate cell shape, as well as the presentation of guidance molecules.

4Here we show that Dcx/Dclk1-deficient neurons have unexpectedly specific defects in Kif1a-mediated transport of presynaptic vesicles, and that RNAi knockdown of Kif1a in neurons mimics several effects of Dcx/Dclk1 deficiency. We demonstrate specific increases of Kif1a MT binding and run length mediated by Dcx, and our subnanometer structural analysis of the Dcx:MT:kinesin complex suggests a model for how Dcx and Dclk1 facilitate Kif1a-MT association to regulate MT-based transport of cellular components. Our findings thus suggest a mechanism in which local control of Dcx-MT binding might in turn regulate kinesin-based transport of cellular components in developing and adult neurons.

5Although overexpression of Dcx and Dclk1 induces MT polymerization and sometimes bundling, we found that absence of Dcx and Dclk1 does not impact MT organization (data not shown) but instead resulted in defective Vamp2 localization in neurons. Pursuing a previous observation that the presynaptic vesicle protein Vamp2 failed to localize normally at 7 days in vitro (DIV) in Dcx/Dclk1-deficient axons, we tested whether Vamp2 was mislocalized in dendrites as well. Indeed, Vamp2 was retained in the cell body with defects in axonal and dendritic transport (Figure 1A), which are rescued through expression of an shRNAi-resistant, HA-tagged human Dcx construct (Figure 1B).

6Since Vamp2 localization requires trafficking by molecular motors, we examined the distribution of candidate kinesin motors for Vamp2 transport in Dcx/Dclk1-deficient neurons, including conventional kinesin and Kif1a, a kinesin-3 family motor that transports presynaptic vesicles. We found that in Dcx/Dclk1-deficient neurons, Kif1a, but not conventional kinesin, is strikingly mislocalized. In wild-type (WT) neurons, Kif1a is present in the cell body and throughout the neurites, whereas Dcx/Dclk1-deficient neurons have less Kif1a in neurites, while the cell body is brightly immunoreactive (Figure 1C). Quantitative immunofluorescence confirms loss of Kif1a staining from Dcx/Dclk1-deficient neurites >4 μm from the cell body, compared to control cells, and this loss can be rescued through expression of an shRNAi-resistant, HA-tagged human Dcx (Figures 1C and D, see Figure S1A available online). In contrast to Kif1a, immunostaining for conventional kinesin reveals no difference between WT and neurons deficient for Dcx/Dclk1 (Figure 1E, Figure S1B), suggesting that Dcx/Dclk1 specifically regulates Kif1a localization.

7The mislocalization of Kif1a seen in Dcx/Dclk1-deficient neurons may account for defective Vamp2 localization, since knockdown of Kif1a itself causes very similar defects in Vamp2 localization. Using shRNAi sequences targeting Kif1a (Figure 2, Figure S2), we found that in most Kif1a knockdown neurons, Vamp2 expression was confined to the cell body (Figure 2B, in contrast to normal neurons shown in Figure 2A). A subset of Kif1a knockdown neurons demonstrated abnormally large accumulations of Vamp2 vesicles in neurites (Figure 2C), a defect also classically observed in transport failure, thus strongly suggesting defective transport of Vamp2 in the absence of Kif1a. Live-cell imaging in Kif1a knockdown cells showed near-total loss of observable mobility for Vamp2-GFP (Figures 2D–2G, Movie S1) in both anterograde and retrograde directions, with the number of mobile vesicles being <5% for the knockdown compared with control (Figure 2D), which is consistent with previous reports that Kif1a knockdown decreases bidirectional cargo transport in neurons. This phenotype could be rescued through expression of an shRNAi-resistant, Myc-tagged human Kif1a (Figure 2D), strongly suggesting that Kif1a is involved in the transport of Vamp2. We confirmed the close physical relationship between Kif1a and Vamp2 through coexpression of Kif1a-mCitrine with Vamp2-RFP, showing that the Kif1a motor colocalizes extensively with Vamp2 (Figures 2H–2J). While all of the vesicular Vamp2-RFP appears to colocalize with Kif1a, a fraction of Kif1a-mCitrine is associated with non-Vamp2-positive vesicular structures, suggesting that the motor is also associated with other cargos. In contrast, coexpression of Kif1a-mCitrine with the mitochondrial marker Mito-RFP shows little colocalization of the motor with mitochondria (Figures 2K–2M), which in neurons are transported by Kif1b and conventional kinesin. Our results therefore suggest that, at the developmental stages examined, Kif1a is the major motor for Vamp2 trafficking and that other motors, such as conventional kinesin, play only a minor role.

8Since our findings that Dcx/Dclk1 deficiency disrupts Kif1a localization imply that Kif1a may mediate Dcx function, we investigated whether neurons deficient for Kif1a show defects in migration and morphology similar to those seen in Dcx/Dclk1-deficient neurons. shRNAi constructs targeting Kif1a (Figure S2A) or Dcx were electroporated into the cortex of E15.5 WT and Dclk1−/− embryos, respectively, after microinjection into the lateral ventricles. While control slice cultures maintained for 4 DIV showed GFP-labeled neurons demonstrating significant migration into the cortical plate (CP), we found that shRNAi constructs disrupting Kif1a protein expression (Figures S2A and S2B) phenocopy Dcx/Dclk1 deficiency in their effects on neuronal migration (Figure S2C), consistent with previous results. At the cellular level, Dcx/Dclk1 deficiency and Kif1a knockdown have similar effects on neuronal process length and polarity, with an overall decrease in neurite length when measuring all processes (Figure S2G), a reduction in length in the three longest neurites (Figure S2H), and finally an increase in the number of primary neurites directly arising from the soma (Figure S2I). These changes likely reflect an early disruption of polarization of the neuroblast and are potentially causative for the migration defects observed, and thus Kif1a deficiency appears to closely phenocopy Dcx/Dclk1 deficiency in early stages of cortical development.

9Since Kif1a and Vamp2 colocalize extensively, we used time-lapse imaging of Vamp2-GFP as a proxy to characterize Kif1a-mediated vesicle transport in WT and Dcx/Dclk1 single and double deficient neurons, as well as following Dcx rescue (Figure 3, Movie S2). We found significantly fewer Vamp2-GFP vesicles exiting the cell body toward the neurites in both Dcx- and Dcx/Dclk1-deficient neurons (Figures 3A–3D); however, this effect can be rescued by overexpression of RNAi-resistant human Dcx (Figures 3D and 3E). Dcx/Dclk1-deficient neurons also show fewer Vamp2-GFP vesicles in neurites (Figure 3F). Extensive control experiments indicate that the transport defects caused by loss of Dcx/Dclk1 do not reflect changes in actin structure—including the “actin filter”—MT structure; common posttranslational modifications of MTs such as glutamylation, (de)tyrosination, and acetylation;, or alterations in other MAPs, such as MAP2 or tau-1 (data not shown).

10Since endogenous Dcx is primarily localized to MTs in distal regions of the neurites, we sought to understand how Kif1a function changes distally in Dcx/Dclk1-deficient neurons by imaging and tracking Vamp2-GFP anterograde transport in distal neurites (Figures 4A–4E, Movie S3). In Dcx-deficient neurons, vesicle tracings were shorter than in neurons treated with a scrambled control shRNA. Quantification of vesicle behavior in both the single and double deficient Dcx/Dclk1 condition demonstrated statistically significant decreases in the number of mobile vesicles in the distal neurite, which can be rescued by expression of shRNAi-resistant human Dcx (Figure 4C). In addition, despite the paucity of mobile vesicles in Dcx-deficient neurons, we were able to determine anterograde run lengths for Vamp2 transport in a sufficient number of vesicles in these neurons to observe a significant decrease compared to WT (Figure 4D). On the other hand, the anterograde velocity of WT, Dcx RNAi, rescue, and Dcx overexpression does not change (Figure 4E). The highly abnormal Kif1a-mediated transport is all the more striking when compared to conventional kinesin function, which is unaffected in Dcx/Dclk1-deficient neurons (Figure 1, Figure S1). Cargo that is transported by conventional kinesin rather than Kif1a, such as mitochondria, was unaffected by Dcx deficiency, in terms of percentage of mobile mitochondria and run length between WT control and Dcx-deficient neurons (Figures 4F–4J, Movie S4). Consequently, we conclude that MTs in Dcx/Dclk1-deficient neurons are unimpaired in their ability to support motor-mediated transport related to conventional kinesin, but severely deficient for Kif1a, suggesting a kinesin subtype-specific role for the Dcx domain proteins Dcx/Dclk1 in the regulation of MT-based transport.

11Since point mutations in Dcx are known to cause defects in neuronal migration, we tested whether some mutant Dcx proteins could potentially disrupt specific Kif1a transport functions. Dcx is known to decorate MTs in a gradient, with highest levels bound in a polarized manner in the cell body and in distal neurites (Figure 5A; ). We hypothesized that Dcx association with MTs may serve to prevent the Kif1a:cargo complex from dissociating from the MT, thus permitting longer run lengths along the MTs and resulting in efflux of the Kif1a:cargo complex from the soma into the neurites. We investigated the Dcx S47R mutation that causes retention of Dcx in the soma to determine its effect on Kif1a transport from the cell body into the neurite. Dcx−/y neurons were transfected with plasmids expressing HA-tagged WT or mutant Dcx constructs, then permeabilized in MT-stabilizing buffer to determine where the exogenously expressed Dcx was bound in these neurons. Dcx−/y neurons rescued with the WT HA-Dcx construct showed the greatest amount of Dcx in the distal regions of the neurites, as well as polarized binding in the soma (Figures 5A and 5B). In contrast, rescue with Dcx S47R results in retention of the mutant Dcx in the soma of the neuron without polarized Dcx binding in the cell body (Figure 5C). Live imaging of Vamp2-GFP to characterize Kif1a function in the same neurons rescued with HA-tagged Dcx S47R constructs demonstrates Dcx retention in the cell body and a significant reduction in the number of Vamp2-GFP vesicles exiting into neurites (Figures 5F and 5G). We conclude that Kif1a-dependent Vamp2 transport from the cell body into neurites is dependent on proper distribution of Dcx on MTs, i.e., polarized Dcx MT binding in the cell body and gradient in the neurites, suggesting that the critical interaction between Dcx and Kif1a is likely to occur at the MT surface.

12We also examined human mutations in the sequence linking the N-DC and C-DC MT binding domains of Dcx. As the DC domains bind to MT recesses, we predict this “linker” sequence to be potentially exposed on the MT surface and interact with Kif1a. While MT binding of the Dcx W146C mutant appears to be intact in neurons, rescue of Dcx shRNAi-treated neurons with the Dcx W146C and K174E linker mutants did not rescue the dendritic polarity defects in neuronal morphology (Figures S3A–S3E). Live-cell imaging of Vamp2-GFP in Dcx shRNAi-treated neurons rescued with Dcx W146C shows a defect in vesicle transport (Figures 5H–5J, Movie S5). We tracked Vamp2-GFP vesicular transport in these neurites and found a decrease in the number of mobile vesicles compared with WT Dcx rescue neurons. Analysis of the mobile vesicles showed that run lengths were decreased in the Dcx W146C mutant rescue compared to WT, but that velocity was not significantly altered, remarkably similar to the effects of Dcx deficiency, suggesting that the interactions between Kif1a and Dcx may require specific amino acid residues in this linker segment.

13The changes in run length caused by alterations in Dcx levels in our mutational analysis suggested that Dcx might regulate Kif1a interactions with MTs. Coimmunoprecipitation (coIP) using a primary antiserum to the C terminus of Dcx on protein lysates of human fetal cortex (23 weeks of gestation) enriches for a complex that includes Kif1a, providing evidence of Dcx association in vivo (Figure 6A). This interaction was confirmed by the direct pull-down of Dcx using an N-terminal HaloTag human KIF1A (amino acids 1–361) fusion protein in the absence of MTs or tubulin using purified protein components (Figure 6B). Indeed, crosslinking experiments using bis(sulfosuccinimidyl)suberate (BS3) to reconstitute this complex using purified protein components in vitro followed by mass spectrometry analysis identified a range of molecular complexes containing both Dcx and Kif1a in the presence of MTs (Figure 6C). Interestingly, formation of the Dcx:Kif1a complex was also observed in the absence of MTs and nucleotide at concentrations >15 μM (data not shown). Therefore, our data suggest that Kif1a and Dcx can interact directly and independently of MTs, though when MTs are present that interaction likely occurs at the MT surface.

14While we show that Dcx, Kif1a, and tubulin form a ternary complex both in vivo and in vitro, we asked whether Dcx has any effect on the direct interaction of Kif1a with MTs. Using a traditional MT pull-down assay and the nonhydrolyzable nucleotide AMP-PNP, which promotes high-affinity motor-MT interactions, lysates from Dcx/Dclk1-deficient mouse brain (Dcx−/y;Dclk1−/−) show significantly less Kif1a bound to MTs compared to WT (Figure S4A). Similarly, lysates from HEK cells expressing the motor domain of Kif1 (amino acids 1–365) and only the MT binding domain of Dcx (amino acids 1–270) show a modest, but significant, increase in binding of Kif1a to MTs in presence of excess MTs by 15%–20% over control (Figures S4B–S4D).

15Because the effect of binding with AMP-PNP was relatively small in contrast to the run length effect observed in vivo, we assessed the effect of Dcx on the Kif1a-MT interaction in the presence of other nucleotides, i.e., ADP and ATP, using purified protein components. We bound purified N-terminally Halo-tagged, truncated human Kif1a (amino acids 1–361) to magnetic HaloLink beads under saturated conditions and coincubated it in the presence or absence of Dcx-decorated MTs and either ATP, ADP, or AMP-PNP (Figures 6D and 6E). Strikingly, while Dcx does not appear to enhance motor binding to MTs in the ATP and AMP-PNP binding state, a significant increase is observed in the ADP binding state when Dcx is present. The Dcx-mediated increase in the presence of ADP is approximately 2-fold compared to a Dcx-negative control and compared to ATP or AMP-PNP in the presence of Dcx. Interestingly, since the experiment is performed under saturating conditions, excess Kif1a protein is pulled down in the presence of Dcx and ADP by a factor of 2-fold over ATP and AMP-PNP and/or lack of Dcx (Figure 6E, right panel). This suggests the possible existence of two binding sites on Dcx, one that is nucleotide independent and one that is specific for ADP-bound Kif1a. Similar results are observed when performing the reverse experiment, where human Dcx is bound to the magnetic beads first, followed by coincubation with Kif1a (C351) in the presence of all other components (data not shown), thus confirming our results shown in Figure 6D.

16To investigate the molecular basis of our cellular and biophysical observations, we first examined whether the conformation of MT-bound Dcx is influenced by the absence or presence of kinesin. In a subnanometer-resolution cryo-EM reconstruction of the binary Dcx:MT complex, we clearly observed a DC core at the Dcx binding site (Figure 7A, top; Figure S5A). This was also previously observed in a reconstruction of a ternary Dcx:MT:kinesin complex (Figure 7A, bottom; ), but due to technical limitations in our reconstruction method, this could have corresponded to N-DC, C-DC, or a mixture of both. Strikingly, in our new structure we found that linker regions on either side of the well-defined DC core adopt a significantly different conformation in the absence compared to the presence of bound kinesin, providing important insight into the Dcx-MT interaction (Figure 7A).

17In the Dcx:MT binary complex (Figure 7A, top), the pre-DC linker region shows only diffuse density, demonstrating that this region is flexible when bound to MTs, as it is in solution (; Figure S5B). Crucially, in the absence of kinesin, there is clearly post-DC linker density docking along the DC core (Figure 7A, top). A distinctive feature of N-DC is the presence of W146 in its post-N-DC linker, which has been shown to dynamically dock against the N-DC core; point mutations at this residue cause lissencephaly and defects in intracellular transport (Figure 5). An equivalent hydrophobic residue is not present in the post-C-DC linker, nor is the linker seen docked against the C-DC in its solution structure (PDB ID code 2DNF; Figure S5B). Flexible docking of available N/C-DC structures—including the post-DC linker—into our cryo-EM reconstruction confirmed that W146 apparently contributes to docking of the extra density against the DC core when bound to MTs. This observation strongly supports the idea that the Dcx density observed in our reconstructions corresponds to N-DC.

18Comparison of the Dcx:MT and DCX:MT:kinesin complexes also provided significant insight into the nature of the contacts between Dcx and kinesin on the MT surface. In the Dcx:MT:kinesin ternary complex (Figure 7B; ), although Dcx binds at the corner of four tubulin dimers, and therefore four kinesin motor domains (MDs I–IV), the Dcx density is more closely associated with kinesin motors along one of the protofilaments (Figure 7B, MDs II and III). In this reconstruction, the MD is in a high-affinity nucleotide-free state and enabled docking of a Kif1a MD crystal structure (Figure 7B, Figure S5C). Residues in kinesinII (loop 2) and kinesinIII (loop 8)—which may be important for axonal specificity of some kinesins—but not kinesinI or kinesinIV, are closer than 5Å to the Dcx density (Figures 7A and B). Intriguingly, the N-terminal linker region outside the proposed N-DC core interacts with the MT wall and lies close to kinesinII loop 2 (Figure 7A), while the C-terminal linker is completely displaced from N-DC due to the presence of the bound motor (kinesinIII). Thus, because their conformation is significantly different in the absence and presence of motor protein, it is likely that residues in the linkers adjacent to the N-DC domain interact with kinesin at the MT surface (Figure 7B, table) and dynamically respond to the presence or absence of bound motor (Figure 7C). By its nature of loose attachment, the motor's low-affinity ADP-bound state is hard to access by subnanometer-resolution cryo-EM structure determination. However, our structural analysis suggests specific residues on Dcx and Kif1a that could also act selectively to enhance the binding of the low-affinity ADP-bound motor for MTs (Figure S5D). Although not visible in our reconstruction due to its flexibility, C-terminal portions of Dcx could also be involved in this interaction.

19Here, we show that the Dcx domain proteins Dcx and Dclk1 regulate the function of the neuronal kinesin-3 Kif1a. Dcx- and/or Dclk1-deficient neurons show impaired Kif1a-mediated transport of Vamp2. Lack of Dcx and/or Dclk1 decreases run length of the motor protein and its associated cargo. We show that these changes in motor behavior are correlated with enhanced binding of the ADP-bound Kif1a motor domain to MTs in the presence of Dcx. In addition, we show that mutations in the linker region of Dcx impair Kif1a motility. Finally, using cryo-EM and subnanometer-resolution reconstruction, we visualize the kinesin-dependent conformational variability of the pre- and post-N-DC linker region of MT-bound Dcx, which likely contributes to regulation of motor function.

20Dcx/Dclk1 regulation at the MT surface represents a mechanism of Kif1a regulation—distinct from cargo binding and release of autoinhibition, dimerization, or interactions with polyglutamylated tubulins. Instead, our data suggest that a MAP, in this case Dcx, can enhance motor function by increasing run length. This increase in run length correlates with an approximately 2-fold increase in affinity of the ADP-bound Kif1a motor domain to MTs in the presence of Dcx, suggesting that fine regulation of the weak affinity state of Kif1a can titrate motor activity locally in critical regions of the neuron where Dcx is enriched on MTs. Our findings suggest that Dcx may regulate run length of kinesin-3 motors through specific reduction of the “off-rate” kinetics of the motor protein, thus reducing the likelihood that the motor domain detaches from the MT after completion of its ATPase cycle that drives processive movement along the protofilament. Differential binding of Dcx and/or Dclk1 to the MT, due to local concentration and/or posttranslational modification of Dcx/Dclk1, then has the potential to provide local control of Kif1a-mediated vesicle trafficking.

21The comparative analysis of the Dcx:MT complex versus the Dcx:MT:kinesin complex allows us to identify residues that might be involved in the specific regulation of the Dcx-Kif1a interaction (Figure 7B, table). Our data show that the region of Dcx bound to the MT corresponds to the N-DC (R1) domain. Interestingly, in the absence of motor binding to the MT, the post-N-DC linker region of Dcx is clearly docked on the MT through W146, which is consistent with previously reported results. Addition of a kinesin MD, however, results in the undocking of the post-N-DC linker and, in addition, orientation of the pre-N-DC density in close proximity between two kinesin motor domains (Figure 7A, bottom; and Figure 7B). Taken together with our live-cell imaging data reported for the human patient mutations W146C and S47R in Dcx (Figure 5), which show both a reduced overall mobility of presynaptic vesicles and a decrease in run length, our structural analysis suggests a crucial role of the N-DC domain of Dcx for regulating access of the Kif1a motor domain to MTs. It is likely that residues adjacent to the N-DC domain of Dcx interact with loop 2 and loop 8 of Kif1a at the MT surface (Figure 7B, table), thus allowing a dynamic response of Dcx to the presence or absence of the motor protein (Figure 7C, Movie S6). Although ultimately ternary reconstructions with Dcx and Kif1a will be necessary to reveal the structural and functional details of the MAP-motor interactions in the presence of different nucleotides, our current analysis yields a first interpretation of how Dcx could act selectively to enhance the binding of the low-affinity ADP-bound motor for MTs. In addition, the docking and undocking of the post-N-DC linker may be a mechanism for maintaining specificity of Kif1a binding to certain MTs that are decorated with Dcx, and thus perhaps excluding trafficking of cargo carried by other molecular motors not regulated by Dcx.

22Dcx MT binding can be mediated by dephosphorylation of Dcx through a specific interaction with spinophilin and protein phosphatase 1 in distal regions of developing neurites. The binding of Dcx on MTs in these regions may therefore dynamically regulate vesicle trafficking in these regions, potentially allowing local mechanisms at the growth cone to regulate trafficking into the growth cone. Spinophilin itself is an actin-binding protein with a role in spine formation. Thus, positive signals directing Dcx MT binding may be downstream of actin effects in spines or dendrites and may stimulate Kif1a transport potentially in a dynamic and localized fashion in both dendrites as well as axons. In contrast, release of Dcx from MTs is mediated by kinases that include many with important roles in neuronal development and function, such as CDK5, MARK, and JunK. Therefore, spatial and temporal regulation of Dcx phosphorylation and dephosphorylation presents a potential means of regulating Kif1a transport in a local fashion.

23The interaction between Kif1a and Dcx/Dclk1 for locally regulating transport suggests mechanisms for the role of Dcx/Dclk1 in neuronal migration and axonal outgrowth. Neurons undergo changes in shape during migration which require dynamic regulation of trafficking and membrane remodeling in response to extracellular signals. Newly born neuroblasts are initially bipolar but become multipolar in the subventricular zone before becoming bipolar migrating neurons that undergo saltatory movements of the leading process and the nucleus as they traverse the intermediate zone and CP. Our data (Figure S3 and data not shown), as well as previously published studies, show that Kif1a and Dcx/Dclk1 are clearly essential for remodeling the multipolar shape into a bipolar shape, which may reflect roles in dynamic regulation of membrane remodeling through changes in vesicle trafficking, though Kif1a-based motor function may also have other roles in migratory mechanisms.

24Dcx and Dclk1 are part of an enlarging family of Dcx domain-containing MAPs that all appear to localize to MTs. The widespread expression of Dcx domain proteins suggests a broader role of this unique family of MAPs in neuronal function. While Dcx is a developmentally expressed protein, Dclk1 is present in both immature and adult neurons. Though the Kif1a-binding linker sequence is highly conserved between Dcx and Dclk1, and appears to regulate Kif1a but not conventional kinesin, this linker sequence is highly divergent in sequence in other Dcx domain-containing proteins such as RP1 and DCDC2 (Figure S3A). This lack of conservation in a potential kinesin-binding interface suggests a model in which DC domain proteins interact with kinesins or other motors in combinatorial fashion, potentially even forming a broader “MAP-kinesin code,” in which MAPs and kinesins show diverse yet specific interactions. Other MAPs bind MTs at various sites and with structurally diverse MT-binding domains, and it is an open question whether non-Dcx domain-containing MAPs may regulate more structurally diverse kinesin family members. Although we have by no means explored the full diversity or specificity of interactions between MAPs and kinesins, our data provide an initial step that may ultimately enlighten the great diversity of these large protein classes.

25Detailed methods can be found in the Supplemental Information.

26Animal work was conducted in compliance with protocols approved by the Children's National Medical Center Institutional Animal Care and Use Committee. We bred Dcx+/−;Dclk1+/− females with Dclk1−/− males to generate Dcx−/y;Dclk1−/− mice as described previously.

27E15 cortices or E17.5 hippocampi were dissected and dissociated using the Worthington Papain dissociation system. Plasmids were transfected by using the Amaxa system of electroporation at the time of dissociation according to instructions from the mouse neuronal transfection protocol or with Lipofectamine 2000 overnight prior to imaging. Details of DNA constructs are in the Supplemental Experimental Procedures.

28Images were acquired using a Zeiss LSM510 confocal microscope. The measurement of pixel intensity in the neurite as a function of distance from the cell body was conducted using MetaMorph. For the tracing and measurement of neurites using the ImageJ plug-in NeuronJ, 20 randomly selected neurons were measured in each category from more than five representative experiments.

29E17.5 neurons were transfected with shRNAi constructs by the standard Amaxa transfection protocol after dissociation and plating. They were again transfected with either Lipofectamine 2000 with VAMP-2 or Mito-DsRed 18 hr prior to imaging, which commenced 96 hr after initial transfection under controlled atmospheric conditions at 37°C. Images were acquired on an inverted epifluorescence microscope (IX-81, Olympus America Inc., Melville, NY) equipped with high numerical aperture lenses (Apo 60× NA 1.49, Olympus) and a stage top incubator (Tokaihit, Japan) maintained at 37°C at a rate of one capture per second for the VAMP2 and one capture every 2 s for the mitoDsRed. Fluorescence excitation was carried out using solid-state lasers (Melles Griot, Carlsbad, CA) emitting at 488 nm (for green) and 561 nm (for red) fluorophores. Emission was collected through appropriate emission band-pass filters obtained from Chroma Technologies Corp. (Brattleboro, VT). Images were acquired with a 12-bit cooled CCD ORCA-ER (Hamamatsu Photonics) with a resolution of 1280 × 1024 pixels (pixel size = 6.45 μm2). The camera, lasers, and shutters were all controlled using Slidebook 5 (Intelligent Imaging Innovations, Denver, CO). Analysis of time-lapse imaging was performed with MetaMorph for tracking and the ImageJ Manual Tracking plugin as described (http://rsbweb.nih.gov/ij/plugins/track/track.html).

30Crosslinking experiments were performed using a 1:1 mixture of BS3-d0 and BS3-d4 (Pierce). For each reaction, 10 μM of each purified protein component (Dcx, Kif1a, MTs) and 4 mM ATP were incubated in a 25 μl reaction in BRB80 buffer (80 mM PIPES, 1 mM EGTA, 1 mM MgCl2 [pH 6.9]) for 30–45 min at room temperature (rocking) to allow ternary complex formation. After this initial incubation period, the crosslinking reagent BS3 was added in 100-fold excess (1 mM), and the reaction was incubated at room temperature for another 60 min (rocking). Reactions in which MTs and/or BS3 were omitted were carried out exactly as described above, but lacking the indicated components. The results were assessed by SDS-PAGE and silver staining, followed by protein identification by mass spectrometry.

31Nucleotide-dependent coimmunoprecipitation of Dcx and Kif1a was performed using purified Halo-tagged human Dcx and Kif1a protein (as described above) and HaloLink Magnetic Beads (Promega). MTs and unpolymerized porcine tubulin were purchased from Cytoskeleton. For each reaction, 4–10 μM of each purified protein component (Dcx, Kif1a, MTs, or tubulin) and 4 mM nucleotide (ATP, ADP, AMP-PNP) were incubated in a 50 μl reaction in HaloLink binding/washing buffer (100 mM Tris [pH 7.4], 150 mM NaCl, 4 mM MgCl2, 1 mM EGTA, 1 mM DTT, 0.5% Triton X-100) for 15–45 min at room temperature (rocking) to allow ternary complex formation. After this initial incubation period, 25 μl of BSA-blocked magnetic beads were added to each reaction, followed by further incubation at room temperature for another 15–45 min (rocking). Reactions in which MTs, tubulin, and/or Dcx or Kif1a were omitted were carried out exactly as described above, but lacking the indicated components. Each reaction was washed three times with HaloLink binding/washing buffer, and after addition of SDS-PAGE sample buffer, proteins were eluted off the beads by boiling the samples at 95°C for 3–5 min. The results were assessed by SDS-PAGE and silver staining, followed by densitometry analysis and quantification. Results shown in Figure 6I were quantified by normalizing the bound amount of Kif1a or tubulin as p = P/(S+P) to the ATP condition in absence of Dcx.

32Recombinant full-length human Dcx (amino acids 1–366) was expressed and purified from Sf9 insect cells, and Dcx-stabilized MT cryo-EM samples were prepared as described previously, except that no kinesin motor domain was added. Low-dose images were collected on a FEG electron microscope (Tecnai F20) operating at 200 kV, 50,000× magnification and with a defocus of 0.9–3.6 μm. Micrographs were recorded on films (Kodak SO-163) and binned to final sampling of 2.8 Å/pixel. MTs (222) were selected and processed following a single particle procedure described elsewhere. A subselection of 146,000 tubulin dimers (75% of total data set) went into the final reconstruction. The resolution of 8.3 Å was assessed using the Fourier shell correlation (FSC) 0.5 criterion, calculated between two reconstructions from independent half data sets (see Figure S7B). EM maps are deposited in the EMDB: Dcx-MT EMDB ID 2095 and Dcx-MT-kinesin EMDB ID 2098.

33UCSF Chimera was used for visualization of 3D models, atomic model building, and rigid-body fitting of atomic models in the cryo-EM density maps. The N-DC domain atomic model (1MJD.pdb, model 11, amino acids 38–150) was extended at its C terminus (amino acids 151–156). The resolution of our reconstruction is not sufficiently high to unambiguously determine the conformation of the extended C-terminal linker. Nevertheless, we modeled an extended polypeptide chain into the Dcx cryo-EM density to evaluate its approximate position on the MT surface and with respect to the kinesin binding site. Briefly, the Dcx cryo-EM density was isolated, the extended N-DC domain docked into it, and the C-terminal polypeptide chain was moved into the EM density using flexible fitting with Flex-EM (the core DC domain, amino acids 53–129; and the docked W146 region, amino acids 145–146, were treated as a single rigid body; ). Finally multiple subunit fitting (extended N-DC domain with four surrounding tubulin monomers) was performed using Flex-EM. The kinesin neck-linker (amino acids 324–329) was modeled into the Dcx:MT:kinesin cryo-EM map in a similar way. Because of apparent local displacements between 1BG2.PDB and the kinesin density seen in our reconstruction, we refined the position of loop 2, loop 9, and helix α6 using Flex-EM, but the larger conformational changes of loop 11 could not be modeled simply; also seen by.