Address correspondence and reprint requests to Department of Cell Pharmacology, Graduate School of Medicine, Nagoya University, 65 Tsurumai, Showa-ku, Nagoya, Aichi, 466–8550, Japan. E-mail: email@example.com
The transport of tubulin and microtubules in a growing axon is essential for axonal growth and maintenance. However, the molecular mechanism underlying the linkage of tubulin and microtubules to motor proteins is not yet clear. Collapsin response mediator protein-2 (CRMP-2) is enriched at the distal part of growing axons in primary hippocampal neurons and plays a critical role in axon differentiation through its interaction with tubulin dimer and Numb. In this study, we show that CRMP-2 regulates tubulin transport by linking tubulin and Kinesin-1. The C-terminal region of CRMP-2 directly binds to the tetratricopeptide repeat domain of kinesin light chain 1 (KLC1). Soluble tubulin binds to the middle of CRMP-2 and forms a trimeric complex with CRMP-2/KLC1. Furthermore, the movement of GFP–tubulin in the photobleached area is weakened by knockdown of KLCs or CRMP-2. These results indicate that the CRMP-2/Kinesin-1 complex regulates soluble tubulin transport to the distal part of the growing axon.
sodium dodecyl sulfate polyacrylamide gel electrophoresis
small interfering RNA
The mechanism underlying tubulin/microtubule transport in axons has been a controversial issue for a few decades (Hirokawa 1997; Hollenbeck and Bamburg 1999; Galbraith and Gallant 2000; Shah and Cleveland 2002; Baas and Buster 2004). Based on studies of metabolically labeled tubulin, it appears that tubulin is synthesized within the cell body and transported in the form of microtubules (Lasek and Hoffman 1976). However, there is still debate about which form of tubulin – microtubules or soluble tubulin – is transported down the long axon. Although the rapid axonal transport system has been thoroughly examined by reviewing the large type of kinesin with a number of cargo molecules that move rapidly and form a vesicular conformation (Hirokawa et al. 1998), the mechanism of cytoskeletal component transport has not been well studied. There are few data regarding the molecular mechanism. In particular, the nature of the motors for tubulin transport has long been a point of speculation (Galbraith and Gallant 2000; Baas and Buster 2004).
We previously found that collapsin response mediator protein-2 (CRMP-2), also known as TOAD-64, Ulip2, and DRP-2, is crucial in axon induction in hippocampal neurons (Inagaki et al. 2001; Fukata et al. 2002; Nishimura et al. 2003). CRMP-2 is one of at least five isoforms (CRMP-1–4 and CRAM), and its expression is up-regulated during neuronal development (Goshima et al. 1995; Minturn et al. 1995; Gaetano et al. 1997; Byk et al. 1998; Inatome et al. 2000). Overexpression of CRMP-2 induces multiple axons and elongation of a primary axon, whereas the dominant negative form of CRMP-2 inhibits the formation of axons (Inagaki et al. 2001). Moreover, CRMP-2 interacts with tubulin and Numb (Fukata et al. 2002; Nishimura et al. 2003). Through its interaction with tubulin dimers, CRMP-2 promotes microtubule assembly (Fukata et al. 2002). Numb interacts with α-adaptin (a subunit of AP2 adaptor complex) and is involved in clathrin-dependent endocytosis (Santolini et al. 2000; Berdnik et al. 2002). CRMP-2 is involved in the polarized Numb-mediated endocytosis of neuronal adhesion molecule L1 (Nishimura et al. 2003). Because CRMP-2 is enriched at the distal part of the growing axon of stage 3 cultured hippocampal neurons (Inagaki et al. 2001), we speculated that the local concentration of CRMP-2, which forms a complex with tubulin heterodimer, determines the fate of axons and dendrites in stage 3 neurons. However, it is unclear how the CRMP-2/tubulin heterodimer is transported in the distal portion of axons during neuronal polarization.
Axonal proteins are thought to be transported by microtubule-dependent motor proteins, such as kinesins. Kinesins are a family of motor proteins that use the energy of adenosine triphosphate (ATP) hydrolysis to move cargo along microtubules (Kamal and Goldstein 2002). The kinesin superfamily consists of 27 kinesin families (Miki et al. 2001). Among them, the most is known about Kinesin-1 (conventional kinesin, Kinesin-I, KIF5/KLC), Kinesin-2 (Kinesin-II, KIF3), Kinesin-4 (KIF4), and Kinesin-13 (KIF2) in nerve tissue. Kinesin-1 is a tetramer of two kinesin heavy chains (KHCs, KIF5s) and two kinesin light chains (KLCs) that are associated in a 1 : 1 stoichiometric ratio (Brady 1985; Vale et al. 1985; Johnson et al. 1990). Kinesin-1 transports several cargo proteins, such as synapsin-1 and growth-associated protein-43 (GAP-43; Ferreira et al. 1992). Previous reports have shown that KLCs are required for either cargo binding (Brady and Pfister 1991; Stenoien and Brady 1997; Khodjakov et al. 1998) or the negative regulation of KHCs (Verhey et al. 1998). The KLC C-terminal domain that consists of six imperfect repeats of a 34-amino acid tetratricopeptide repeat (TPR) module (Gindhart and Goldstein 1996) could be part of a protein interaction interface with a target molecule, such as amyloid precursor protein (APP) and c-jun NH2-terminal kinase (JNK)-interacting protein (JIS/SYD), on vesicular or organellar cargoes (Ferreira et al. 1993; Lamb et al. 1995; Verhey et al. 1998; Blatch and Lassle 1999).
In the current study, we identify Kinesin-1 as a novel CRMP-2 interacting molecule and show that CRMP-2 regulates tubulin transport by linking tubulin and Kinesin-1.
Materials and methods
Materials and chemicals
cDNA encoding human CRMP-2 was generated as described previously (Arimura et al. 2000). cDNAs encoding mouse KLC1 and KIF5A were kindly provided by Dr L. S. Goldstein (Howard Hughes Medical Institute, San Diego, CA, USA). pCAGGS vector was kindly provided by M. Nakafuku (Tokyo University, Tokyo, Japan). pEGFP–N3 vector was purchased from Clontech Laboratories, Inc. (Palo Alto, CA, USA). pQE9 vector was purchased from Qiagen (Hilden, Germany). The following antibodies were used: anti-CRMP-2 monoclonal antibody (C4G), kindly provided by Dr Y. Ihara (Tokyo University, Tokyo, Japan); anti-CRMP-2 polyclonal antibody raised against glutathione-S-transferase (GST)-CRMP-2 (Inagaki et al. 2001); antiα-tubulin monoclonal antibody (DM1A) and antitubulin polyclonal antibody (T3526; Sigma, St Louis, MO, USA); Tau-1 antibody (Boehringer Mannheim, Mannheim, Germany); anti-RhoGDI monoclonal antibody (Transduction Laboratories, Lexington, KY, USA); anti-KHC (KIF5) monoclonal antibody (H2), anti-KLC monoclonal antibody (L1), and anti-KLC monoclonal antibody (L2; Chemicon International, Inc., Temecula, CA, USA); anti-GAP-43 monoclonal antibody (91E12; Roche, Mannheim, Germany); anti-His monoclonal antibody (Qiagen). Anti-KLC polyclonal antibody was raised against GST-KLC1. Because this antigen contains the conserved sequence of KLCs, this polyclonal antibody recognizes two bands with apparent molecular weights of 72 000 (KLC2) and 66 000 (KLC1), respectively (Stenoien and Brady 1997).
The cDNAs encoding human CRMP-2 ΔC440, ΔC381, ΔC275, ΔN348, ΔN416, and ΔN440 were amplified by polymerase chain reaction (PCR). The amplified cDNAs were subcloned into pEGFP–N3 (Clontech) to fuse GFP with the C-terminus of CRMP-2–deleted fragments. CRMP-2 mutants tagged with GST at the C-terminus were generated by subcloning into modified pGEX-4T-1 (Amersham Pharmacia Biotech, Buckinghamshire, UK) or pQE9 (Qiagen). KLC1-heptad repeat (HPT) and TPR were amplified by PCR using the mouse KLC1 cDNA described above as a template. These fragments were subcloned into pGEX-4T-1 (Amersham Pharmacia Biotech) or pQE9 (Qiagen). A fragment lacking the motor domain of KIF5A, KIF5A DN (402–1028 amino acids), was amplified by PCR. The amplified cDNAs were subcloned into pCAGGS-Myc.
GST fusion or His-tagged proteins were purified according to the manufacturer's protocol. GST-fused proteins were then dialyzed against TED buffer [20 mm Tris, 1 mm ethylene diaminetetraacetic acid (EDTA), and 1 mm dithiothreitol (DTT)], and His-fused proteins were dialyzed against 20 mm Tris. Tubulin was prepared from porcine brain by three cycles of polymerization and depolymerization before undergoing diethylaminoethyl-Sepharose column chromatography that used FPLC (Amersham Pharmacia Biotech; Fukata et al. 2002). Native porcine CRMP-2 was purified from porcine brain extracts using a method described previously (Arimura et al. 2000).
Immunoprecipitation was performed as described previously (Fukata et al. 2002). In brief, post-natal day 7 rat brain was extracted by the addition of lysis buffer [20 mm Tris, 150 mm NaCl, 1 mm EDTA, 1 µm phenylmethylsulfonyl fluoride (PMSF), 10 µg/mL leupeptin, 10 µg/mL aprotinin, and 1% NP-40] and then clarified by centrifugation at 100 000 g for 20 min at 4°C. The soluble supernatants were incubated with rabbit IgG, anti-CRMP-2 polyclonal antibody, mouse IgG, anti-KLC antibody (L2), or anti-KHC antibody (H2) for 6 h at 4°C. The immunocomplexes were then precipitated with protein A Sepharose 4B (Amersham Pharmacia Biotech). The resulting eluates, which contained several candidate bands from the immunoprecipitation of CRMP-2, were divided by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS–PAGE), then transferred to a polyvinylidene fluoride membrane. The bands were analyzed by mass spectrometry, as described previously (Fukata et al. 2002). The immunocomplexes were subjected to immunoblotting with anti-KHC antibody (H2), anti-KLC antibody (L1, for Fig. 1a; L2, for Fig. 1b), anti-CRMP-2 polyclonal antibody, antitubulin antibody (DM1A, for Fig. 1a; T3526, for Fig. 1b), and anti-RhoGDI antibody.
Brains from E18 rats were homogenized in 20 mm HEPES, 100 mm K-aspartate, 40 mm KCl, 5 mm ethyleneglycol-bis (B-aminoethyl ether)-N,N′-tetraacetic acid (EGTA), 5 mm MgCl2, 2 mm Mg-ATP, 1 mm DTT, 10 µg/mL leupeptin, 10 µm PMSF, 1 µg/mL pepstatin, and 2 µg/mL aprotinin (pH 7.2). Then, the homogenate was centrifuged at 3000 g for 10 min and divided into supernatant (S1) and precipitant (P1). The S1 fraction was centrifuged at 10 000 g for 20 min (S2 and P2), and the S2 fraction was centrifuged at 100 000 g for 60 min (S3 and P3). The P2 fraction from E18 brain is thought to be a premature synaptosomal fraction because synapses are a characteristic of the mature brain (Saito et al. 1992; Arai and Cohen 1994). Each fraction was subjected to immunoblotting with anti-CRMP-2 (R2), antitubulin (DM1A), anti-KHC (H2), anti-KLC polyclonal, anti-GAP-43 (91E12), and anti-RhoGDI antibodies. Furthermore, the S3 fraction was subjected to sucrose density gradient centrifugation in 0–2 m sucrose gradients. In brief, the S3 fraction was layered on top of the sucrose gradients, and the tubes were spun at 200 000 g for 11 h at 4°C. Fourteen 0.2-mL fractions were collected and then analyzed by SDS–PAGE as described above.
In vitro binding assay
GST, GST-KLC1-full, GST-KLC1-HPT, and GST-KLC1-TPR were separately immobilized onto glutathione-Sepharose 4B (Amersham Pharmacia Biotech). The immobilized beads were incubated with His-CRMP-2 (5 µm) in 20 mm Tris buffer that contained 0.2 mg/mL bovine serum albumin (BSA) for 1 h at 4°C. The beads were then washed six times with 20 mm Tris buffer, and the washed beads were suspended with SDS–PAGE sampling buffer. The bound proteins were subjected to immunoblot analysis with anti-CRMP-2 antibody (R2). The in vitro binding assays using His-KLC1 (1 µm) and GST-CRMP-2 mutant immobilized beads were done the same way. The in vitro binding assay with tubulin (3 µm) and GST-KLC1 was done in the presence or absence of 3 µm His-CRMP-2.
Cell culture, transfection, and time-lapse observation
Dissociated hippocampal neurons prepared from E18 rats by use of papain were suspended in culture medium [neurobasal medium, B-27 supplement (Invitrogen, Carlsbad, CA, USA), and 1 mm glutamine] and plated on polylysine- and laminin-coated 13-mm coverslips at low density (1 × 105 cells per 13 mm diameter). For GFP–protein expression, day in vitro 2 (DIV2) neurons were transfected with plasmids by a calcium phosphate method described previously (Arimura et al. 2000). For the examination of CRMP-2 localization in neurons expressing the dominant negative KIF5A (KIF5A DN), neurons were transfected with pCAGGS-Myc-KIF5A DN, which lacks the motor domain of KIF5A, by a calcium phosphate method before plating. For time-lapse observation, neurons were transfected with GFP–tubulin on DIV3. At 6–10 h after transfection, the area 50 µm distant from the cell body was photobleached in 30 µm square with 75 laser irradiations. Images were acquired using a 63 × oil objective. Images of neurons were collected every 2 s for 2 min. All time-lapse imaging was performed on a confocal laser microscopy system and software (LSM 510; Carl Zeiss, Oberkochen, Germany) built around an Axiovert 100M (Carl Zeiss).
Immunofluorescence study and microscopic observation
Hippocampal neurons were fixed with 3.7% formaldehyde in phosphate-buffered saline (PBS) for 30 min at 37°C, washed with PBS, and permeabilized with PBS containing 0.05% Triton X-100 for 5 min at 4°C. After permeabilization, the cells were blocked in PBS that contained 0.1% BSA and then were doubly stained with anti-CRMP-2 monoclonal antibody and anti-KLC polyclonal antibody, followed by Cy2- or Cy3-conjugated second antibody. For GFP protein expression, 24 h after transfection, neurons extracted as described were fixed and double-stained with each indicated antibody overnight at 4°C. After washing, the samples were incubated with the appropriate Cy3- or Cy2-conjugated second antibody. Neurons were observed using a confocal laser microscopy system (LSM 510) built around an Axiovert 100M. Images illustrating antibody labeling and distribution of the marker protein were acquired using a 40 × 0.75 NA dry objective or a 63 × oil objective. For the examination of the transport of GFP–CRMP-2 mutants, neurons were incubated with CellTrackerTM Orange CMTMR [5-(and-6)-((4-chloromethhyl)benzoyl)amino) tetramethylrhodamine] mixed isomers (Molecular Probes, Eugene, OR, USA) as a cytosolic marker for 30 min at 37°C just prior to fixation. The fluorescence intensity captured with the LSM510 was quantified by ImagePro analysis (Media Cybernetic, Silver Spring, MD, USA) and Excel (Microsoft Corporation, Redmond, WA, USA) as described previously (Nishimura et al. 2004).
siRNA preparation and transfection
A 21-oligonucleotide siRNA duplex was designed as recommended elsewhere (Elbashir et al. 2001) and was synthesized by Japan Bio Service (Saitama, Japan) to target the rat CRMP-2 sequence 5′-ACTCCTTCCTCGTGTACAT-3′, the rat KLC1 sequence 5′-ATACGACGACGACATCTCT-3′, and the rat KLC2 sequence 5′-TCTGGTGATCCAGTATGCT-3′. A scramble sequence, 5′-CAGTCGCGTTTGCGACTGG-3′, was used as a negative control. The transfection of siRNA in primary culture before plating was carried out with calcium phosphate transfection methods described previously (Arimura et al. 2000).
Measurement of axonal length
The measurement of axonal length was done as described previously (Nishimura et al. 2003). In brief, DIV3 neurons were fixed, and the transfected neurons were visualized by immunostaining with anti-KLC polyclonal and Tau-1 antibodies. The images of immunostained neurons were captured with an LSM 510 microscope, the axons were traced on a video screen, and their length was analyzed using the LSM 510 software. At least 40 cells transfected with siRNA were examined in each experiment. The data from at least three independent experiments were subsequently averaged, and the SD was calculated.
Kinesin-1 is a novel interacting protein of CRMP-2
We first searched for the CRMP-2-interacting proteins that determine the specific localization of CRMP-2 in neurons by means of a co-immunoprecipitation assay with developing rat brain extracts (P6 and P7). A protein with a relative molecular mass (Mr) of 120 000 was co-immunoprecipitated with anti-CRMP-2 antibody. We identified the 120 kDa protein as KIF5c by mass spectrometry (data not shown). KIF5c is a KHC of Kinesin-1 (Kamal and Goldstein 2002). Kinesin-1 is a tetramer of two KHCs (KIF5) and two KLCs, which are associated in a 1 : 1 stoichiometric ratio (Brady 1985; Vale et al. 1985). Consistent with this, the results of immunoblot analysis revealed that KLC1 was co-immunoprecipitated with anti-CRMP-2 antibody as well as with KHC (Fig. 1a). Under these conditions, tubulin was co-immunoprecipitated by anti-CRMP-2 antibody. CRMP-2 was reciprocally co-immunoprecipitated with anti-KHC and anti-KLC antibody (Fig. 1b). One of the cytosolic proteins, Rho-GDI, was used as a negative control. Rho-GDI was not co-immunoprecipitated with anti-CRMP-2, anti-KHC, or anti-KLC antibody (Figs 1a and b).
To further examine the in vivo interaction of CRMP-2 with Kinesin-1, subcellular fractionation of brain lysate of 18-day-old-rat embryos (E18) was carried out as described previously (Arai and Cohen 1994). In brief, fractions corresponding to unlysated cells and a nuclear pellet (P1), a crude premature synaptosomal and mitochondrial pellet (P2), and a crude microsomal pellet (P3; see Materials and methods), along with each corresponding supernatant (S1, S2, and S3), were prepared and subjected to immunoblot analysis (Fig. 1c; Arai and Cohen 1994). The majority of CRMP-2 was seen in the cytosol fraction (S1, S2, and S3) and some was found in the P2 fraction (Fig. 1c). KHC and KLC1 were seen in the S3 cytosolic fraction, suggesting that some proportion of these proteins are soluble, as has been reported previously (Rahman et al. 1998). Tubulin was seen in all fractions, probably because the fragmented microtubules were fractionated in the pellet. Most GAP-43 was found in the P2 fraction, and a small amount was seen in the crude microsomal fraction (P3; Rahman et al. 1998).
Using this cytosolic S3 fraction, we performed sucrose density gradient centrifugation. The results of immunoblot analysis revealed that the majority of CRMP-2, KHC, and KLC1 existed in the same fractions, fractions 4 (11.3S) and fraction 5 (6.6S; Fig. 1d). Tubulin existed from fractions 3–8, with the largest amount occurring around fraction 4.3S; this S-value is consistent with that of purified tubulin dimers fractionated in sucrose density gradient (Fukata et al. 2002). Tubulin also existed in the fractions containing CRMP-2 and Kinesin-1 (Fig. 1d). These results suggest that CRMP-2 and Kinesin-1 form a complex in the cytosol fraction and that tubulin is part of this complex.
CRMP-2 has been reported to play an important role in axonogenesis and to accumulate at the distal portion of the growing axon in hippocampal neurons. Therefore, we examined the distribution of KLCs and CRMP-2 in hippocampal neurons by indirect immunofluorescence analysis. Hippocampal neurons at stage 3 showed a typical morphology; neurons had a single long axon and several immature short neurites, as has been reported elsewhere (Craig and Banker 1994). KLCs were diffusely distributed in cell bodies and immature neurites and was accumulated at the distal part of axons, as was CRMP-2 in stage 3 hippocampal neurons (Fig. 2). It has been previously reported that the immunoreactivity of KHC tends to be enriched in the growing tip of only one of the processes in developmental cortical neurons (Morfini et al. 2001), suggesting that KLCs localized in axon can at least form a functional complex with KHC and that this Kinesin-1 complex can efficiently transport the cargos into axon. CRMP-2 has a diffuse distribution throughout the distal part of axonal shafts and is localized at the growth cone (Inagaki et al. 2001). In addition, some punctate CRMP-2 immunoreactivities are colocalized with KLC-containing dot-like structures in axonal shafts. Judging from subcellular fractionation assay, CRMP-2 is mainly cytosolic protein. Some is in the vesicular structure (presumably in synaptosomal-like vesicles; Fig. 1c), and the localization of both forms overlaps with that of KLCs.
Direct binding between CRMP-2 and KLC1
To examine the direct interaction of CRMP-2 with Kinesin-1, an in vitro binding assay was performed using purified recombinant proteins. Previous reports showed that KLCs are required for binding some cargo (Kamal and Goldstein 2002). Thus, we examined the direct binding of CRMP-2 and KLC1, a neuronally enriched KLC isoform. Purified His-CRMP-2 was incubated with GST-KLC1-full (amino acids 1–542, full length), GST-KLC1-HPT (1–167, which includes the heptad repeats domain), GST-KLC1-TPR (168–542, which includes the six-tetratricopeptide repeat domain), or GST immobilized beads. CRMP-2 bound to the full-length KLC1 and the C-terminal TPR domain but not to the N-terminal HPT domain (Fig. 3a). Although the reason that CRMP-2 preferentially associates with the TPR domain in comparison with full-length KLC1 is unclear, it is possible that CRMP-2 is more accessible only to the TPR domain (Fig. 3a).
Next, we determined the binding site of CRMP-2 to KLC1 by in vitro binding assay. Purified His-KLC1 was incubated with various deletion mutants of CRMP-2. KLC1 bound to the full-length CRMP-2 (amino acids 1–572), CRMP-2 N-terminal-deleted mutants ΔN348 (348–572), ΔN416 (416–572), and ΔN440 (440–572), but not to C-terminal-deleted mutants ΔC275 (1–275), ΔC381 (1–381), and ΔC440 (1–440; Fig. 3b). The binding constant between CRMP-2 and KLC1 was determined using the same assay described in Fig. 3(a). The binding of His-CRMP-2 to GST-KLC1-full was dose-dependent and saturable (Fig. 3c). Scatchard analysis revealed a single class of affinity binding sites with a Kd of ∼43 ± 5 nm.
To determine which region is essential for CRMP-2 transport into axons, we examined the localization of green fluorescent protein (GFP)-tagged deleted constructs of CRMP-2 in stage 3-transfected hippocampal neurons. Twenty-four hours after the transfection of GFP–tagged deleted constructs, the distribution of GFP–fused proteins at the distal part of axons was examined in hippocampal neurons. The control GFP–GST protein was localized throughout the axons, and some signal was observed at the tips of the axons (Fig. 4a). In contrast, neurons that expressed GFP–wild-type CRMP-2 showed a clear gradation of fluorescence at the distal part of the axon, and the strongest signals were found at the tips of axons (Fig. 4a). The C-terminal-truncated construct (ΔC381) was not accumulated at the tips of axons. The level of ΔC381 accumulation was clearly lower than that of controls, which suggests that ΔC381 was strongly held within the cell bodies (Figs 4a and b). The C-terminal construct ΔN440 was transported to the distal part of axons (Figs 4a and b). Taken together, these results suggest that the KLC1-binding region of CRMP-2 (amino acids 440–572) is responsible for CRMP-2 accumulation at the distal part of axons.
Transport of CRMP-2 in hippocampal neurons at early stage
We performed RNA interference (RNAi) using siRNA of KLCs to examine whether the transport of CRMP-2 into axons was mediated by Kinesin-1. The results of an immunoblot analysis of cell extracts showed the decrease in KLC1 and KLC2 72 h after transfection (Fig. 5a). Scramble is control siRNA designed to have GC content comparable to that of functional siRNA, but it lacks homology with known gene targets. The viability of KLC siRNA-treated neurons was not different from the scramble siRNA-treated neurons, and the amounts of KHC, CRMP-2, and actin protein were not changed (Fig. 5a). The results of immunofluorescence analyses revealed that the immunoreactivity of KLCs was reduced by KLC-siRNA in almost all neurons and that the immunoreactivity of KLCs was markedly decreased compared to control culture in about 10% of the neurons (Fig. 5b). In these neurons, CRMP-2 was not accumulated at the distal part of longest process and was mainly present within the cell body (Fig. 5b). KLC knockdown neurons maintained the longest process, but the length of such process was decreased compared with that of control neurons (179 ± 33 µm vs. 338 ± 26 µm, respectively; Fig. 5c). The same result was observed in neurons expressing the dominant negative KIF5A (KIF5A DN), which lacks the motor domain of KHCs (KIF5A; Fig. 6). In the neuron expressing KIF5A DN, CRMP-2 was not accumulated at the distal part of longest process and was mainly present within the cell body (Fig. 6a). The length of process in KIF5A DN-expressing neurons was decreased compared with that of control neurons (155 ± 23 µm vs. 281 ± 41 µm, respectively; Fig. 6b). The longest process of KLC knockdown or dominant negative KIF5A-expressing neurons were thinner and tapered toward the distal part of process, and interestingly these processes showed reduced axonal marker Tau-1 staining at the distal part of the longest process (Figs 5b and 6a). These results are consistent with previously reported data that the knockdown of CRMP-2 causes a decrease in the longest process length and affects neuronal polarization (Nishimura et al. 2003). Taken together, these results strongly suggest that CRMP-2 is transported via Kinesin-1 into axons, and Kinesin-1 is important for axonogenesis and neuronal polarity by means of the transport of certain proteins, including CRMP-2.
We previously reported that CRMP-2 binds to a tubulin heterodimer (Fukata et al. 2002). The region of CRMP-2 responsible for the association with tubulin is the 323–381 amino acid region (Fukata et al. 2002), which is different from the KLC1 binding region (440–572). Thus, we examined whether KLC1, CRMP-2, and tubulin could form a trimeric complex in vitro, using each purified protein. In the absence of CRMP-2, tubulin did not bind to KLC1, whereas in the presence of CRMP-2, tubulin bound to the TPR domain of KLC1 (Fig. 7a). These results indicate that tubulin binds to KLC1 in a CRMP-2–dependent manner. This result raises the possibility that the CRMP-2/tubulin complex might be transported by Kinesin-1.
We then examined whether tubulin transport was affected by the function of KLCs or CRMP-2 using GFP–tubulin. Hippocampal neurons were transfected with siRNA of KLCs or CRMP-2 conjugated with Cy3 (Cy3-siRNA). We confirmed that KLCs or CRMP-2 was markedly depleted in siRNA of KLC- or CRMP-2-transfected cultures by immunoblot analysis (Fig. 7b). Under these conditions, some neurons showed the fluorescence of Cy3 inside the cell body (Fig. 7c), suggesting that such neurons were transfected with siRNA. In the neurons containing Cy3-siRNA in the cell body, we performed photobleaching of GFP–tubulin followed by time-lapse analysis. GFP–tubulin was transfected in DIV3 neurons, and photobleaching was performed 6–10 h after transfection. During this period, GFP–tubulin was in a soluble form and was extracted by 0.5% Triton X-100 in the presence of a low concentration of Taxol (Fukata et al. 2002). The control neurons transfected with scramble siRNA showed a steady recovery of fluorescence in the photobleached area within 2 min (Figs 7d and e). The transport velocity of GFP–tubulin was about 0.1–1.0 µm/s and was almost consistent with fast axonal transport (Brady et al. 1985). In contrast, a delay in recovery was observed in CRMP-2 or KLC knockdown neurons (Figs 7d and e). This inhibition of recovery in KLC- or CRMP-2 siRNA–treated neurons was not complete but significant, suggesting that CRMP-2 and KLCs are among the regulators of GFP–tubulin transport. Under our conditions, GFP–tubulin appeared to be moved in both anterograde and retrograde directions, and both directional movements were prevented by the depletion of KLCs or CRMP-2 (see Fig S1). When we treated the neurons with nocodazole and placed them on ice for 5 min, the movement of GFP–tubulin was prevented (data not shown), and neither placing on ice for 5 min nor nocodazole treatment alone prevented tubulin movement, suggesting that GFP–tubulin moves on the rail of microtubules and that the contribution of diffusion is negligible (Galbraith et al. 1999). Taken together, these results suggest that CRMP-2 as well as KLCs are required for GFP–tubulin transport into axon. The roles of CRMP-2 and KLCs in retrograde flow of GFP–tubulin remain to be clarified (see Discussion). Note that the majority of bidirectional movement of GFP–tubulin is not likely to be the result of the assembly of microtubules (Galbraith et al. 1999). As described above, nocodazole treatment alone did not prevent the movement of GFP–tubulin, suggesting that the assemble of microtubules does not contribute to the recovery of GFP–tubulin in photobleached area (Galbraith et al. 1999).
The function of CRMP-2 at the distal portion of axons
Neuronal polarization appears to start the bulk elongation of a minor process as an axon. It is thought that there are prominent rearrangements of cytoskeletons and membrane traffic at this stage. We have previously reported that CRMP-2 participates in axon formation (Inagaki et al. 2001). CRMP-2 forms a trimeric complex with a tubulin heterodimer. The results of in vitro experiments have provided evidence that CRMP-2 promotes microtubule assembly. The overexpression of CRMP-2 induces axon elongation, whereas a CRMP-2 mutant lacking microtubule-assembly activity has no ability to promote axon elongation (Fukata et al. 2002). These results indicate that CRMP-2 enhances axon formation through its microtubule assembly. CRMP-2 associates with Numb and thereby regulates Numb-mediated L1 endocytosis (Nishimura et al. 2003). Both microtubule assembly and Numb-mediated L1 endocytosis occur mainly at the growth cones, where CRMP-2 accumulates. Judging from the result that the knockdown of KLCs delocalizes CRMP-2 in an axon and reduces the axon length, it is likely that the transport of CRMP-2 to the distal part of axons is essential for axon elongation.
The function of Kinesin-1 and neuronal polarity
In the current study, we show that the knockdown of KLCs delocalizes CRMP-2, and inhibits axon elongation. This result is consistent with the results of studies that demonstrated the inhibition of Kinesin-1 by an antisense oligonucleotide treatment of KHCs in cultured hippocampal neurons (Ferreira et al. 1992). Treatment of the neurons with the antisense oligonucleotide affected both the distribution of synapsin-I and GAP-43 and decreased the length of neurites (Ferreira et al. 1992). On the other hand, a KLC1 knockout mouse did not show any defects in neurite formation and axon guidance (Rahman et al. 1999). We speculate that this discrepancy between our result and that in the KLC1 knockout mouse is caused by the redundant effects of KLC2, whose expression was decreased at a certain level in our study. This is the first time that the function of KLC1 and KLC2 in developmental neurons has been revealed clearly.
The knockdown of KLCs inhibited axon elongation and diminished Tau-1 staining, which is a marker of axons. The axonal shaft became thin and tapered toward the distal part of axons in KLC-knocked down neurons. As reported previously, the knockdown of CRMP-2 inhibited axon elongation (Nishimura et al. 2003). In CRMP-2-knocked down neurons, the Tau-1 immunoreactivity was also decreased (data not shown). Tau-1 antibodies recognize certain dephosphorylated forms of Tau, and Tau-1 immunoreactivity at the growing axon is thought to reflect the state of polarization in early stage neurons (Mandell and Banker 1996). Thus, the reduction in Tau-1 immunoreactivity suggests that neuronal polarization is not completely established in KLCs- or CRMP-2-knocked down neurons. Taken together, KLC1-mediated CRMP-2 transport to the distal part of axons appears to play an essential role in axon elongation and neuronal polarization.
The transport of tubulin through CRMP-2/Kinesin-1
Under the conditions in the current study, GFP–tubulin appeared to move in both anterograde and retrograde directions, and both directional movements were prevented by the depletion of KLCs or CRMP-2. We think that the retrograde movement of GFP–tubulin is dependent upon cytoplasmic dynein. There are several types of evidence that the two opposite-directed motors are interdependent for each movement (Brady et al. 1990; Stenoien and Brady 1997; Waterman-Storer et al. 1997; Martin et al. 1999). Recently, the direct interaction between cytoplasmic dynein intermediate chain and KLCs was reported (Ligon et al. 2004). Bidirectional movements of GFP–tubulin raise the possibility that cytoplasmic dynein interacts with the Kinesin-1/CRMP-2/tubulin complex. The prevention of bidirectional movements by siKLCs or siCRMP-2 may reflect the lack of linking of tubulin from these motor protein complexes. The role of the retrograde flow of the CRMP-2/tubulin complex remains to be clarified.
Recently, it has been reported that cytoskeletal movements in the slow transport system are, in fact, rapid, infrequent, and highly asynchronous (Wang and Brown 2002), implying that the slow transport system shares the motor protein with the fast transport system. Some reports have implied the association of kinesin and cytoskeletal component transport (Terada et al. 2000; Xia et al. 2003). Terada et al. (2000) reported that directional movement of tubulin was prevented by functional blocking of motor protein using the antikinesin antibodies in squid giant axon. We found that tubulin bound to KLC1 through CRMP-2 in vitro (Fig. 7a). Sucrose density gradient analysis indicated that a certain population of tubulin forms a complex with CRMP-2/Kinesin-1 (Fig. 1d). Judging from these results, it appears that CRMP-2 is a cargo receptor of tubulin dimers/oligomers and Kinesin-1. Thus, CRMP-2 may serve as an escort protein that carries a tubulin dimer/oligomer on Kinesin-1 along microtubules and enhances microtubule polymerization by co-polymerization (Fukata et al. 2002).
We thank Y. Gu, Y. Ihara, and L. S. Goldstein for kind gifts of materials. We also thank M. Amano, M. Yoshizaki, and T. Nishimura (Nagoya University) for helpful discussion and the preparation some materials and T. Ishii for secretarial and technical assistance. This work was supported by Grants-in-Aid for Scientific Research from the Ministry of Education, Science, Sports and Culture of Japan and by grants from the Novartis Foundation for the Promotion of Science, the Organization for Pharmaceutical Safety and Research (OPSR/Kiko), Research Fellowships of the Japan Society for the Promotion of Science for Young Scientists, Special Coordination Funds for Promoting Science and Technology (SCFPST), the Uehara Memorial Foundation, and the Mitsubishi Foundation.
Figure S1. Time-lapse observation using GFP–tubulin. GFP–tubulin movement in hippocampal neurons transfected with scramble (a), CRMP-2 (b) or KLCs (c) siRNA conjugated with Cy3 was observed. Transfection of siRNA was performed before plating. At DIV3, neurons were transfected with GFP–tubulin. At 6–10 h after transfection, neurons transfected with both GFP–tubulin and Cy3-siRNA were selected, and the area 50 µm distance from cell body was photobleached in a 30-µm2 with 75 laser irradiations in axon. Images of neurons were collected every 2 s for 2 min. Red colour shows the distribution of Cy3-siRNA.