Collapsin response mediator protein-2 accelerates axon regeneration of nerve-injured motor neurons of rat


Address correspondence and reprint requests to Hiroshi Kiyama, Department of Anatomy & Neurobiology, Osaka City University Graduate School of Medicine, 1-4-3 Asahimachi, Abenoku, Osaka, 545–8585, Japan. E-mail:


The rat collapsin response mediator protein-2 (CRMP-2) is a member of CRMP family (CRMP-1–5). The functional consequence of CRMP-2 during embryonic development, particularly in neurite elongation, is relatively understood; however, the role in nerve regeneration is unclear. Here we examined the role of CRMP-2 during nerve regeneration using rat hypoglossal nerve injury model. Among the members, CRMP-1, CRMP-2, CRMP-5 mRNA expressions increased after nerve injury, whereas CRMP-3 and CRMP-4 mRNA did not show any significant change. In the N1E-115 cells, CRMP-2 has the most potent neurite elongation activity among the CRMP family members. In dorsal root ganglion (DRG) organ culture, CRMP-2 overexpression by adenoviral vector demonstrated substantial neurite elongation. On the other hand, CRMP-2 (ΔC381), which acts as a dominant negative form of CRMP-2, inhibited neurite formation. Collectively, it would be plausible that CRMP-2 has potent nerve regeneration activity after nerve injury. We therefore examined whether CRMP-2 overexpression in the injured hypoglossal motor neurons accelerates nerve regeneration. A retrograde-tracer, Fluoro-Gold (FG), was used to evaluate the number of reprojecting motor neurons after nerve injury. CRMP-2-overexpressing motor neurons demonstrated the accelerated reprojection. The present study suggests that CRMP-2 has potent neurite elongation activity in nerve regeneration in vivo.

Abbreviations used

collapsin response mediator protein


dorsal root ganglion


in situ hybridization


Dulbecco's modified Eagle's medium


fetal bovine serum




neural-restrictive silencer element

Regeneration of injured nerve is a unique feature of the adult peripheral nervous system, as opposed to the central nervous system. The injured neurons are protected from cell death, and subsequently growth cones of proximal axon stumps elongate and contact in appropriate regions in accordance with a variety of extracellular guidance cues (Fawcett and Keynes 1990; Tessier-Lavigne and Goodman 1996; Chisholm and Tessier-Lavigne 1999). Although there are significant differences between neuronal development and regeneration, mechanisms underlying the extracellular neurite guidance cues and growth cone dynamics during development are likely to be implicated in nerve regeneration. In terms of growth cone dynamics during development, Rho, SCG10, and GAP-43 families appear crucial because of their abundance and localization in the growth cone during development. The Rho family plays a crucial role in intracellular actin dynamics (Kaibuchi et al. 1999). The SCG10 family members appear to function in microtubule dynamics. GAP-43 is identified as a growth-associated molecule and enriches the growth cone. Although the precise function of the GAP-43 family is unknown, it seems to be associated with the actin filament in the growth cone. Many molecules in those families were proved to be expressed and have significant functions in nerve regeneration (Strittmatter et al. 1995; Tanabe et al. 2000; Iwata et al. 2002). Recently, implications of the collapsin response mediator protein (CRMP) family during development have been revealed, and several lines of intriguing evidence are accumulating. Initially, it was reported that chick CRMP-62 is required for the collapsin/semaphorin3A signaling process in Xenopus laevis oocytes (Goshima et al. 1995), and there are indications that the CRMP family is involved in plexin-based semaphorin signaling (Liu and Strittmatter 2001). The CRMP family consists of five members, CRMP-1 to CRMP-5, and in the developmental stage is up-regulated in the nervous system (Wang and Strittmatter 1996; Quinn et al. 1999; Fukada et al. 2000). Among these members, also CRMP-2 was identified as a substrate for Rho-associated protein kinase (Rho-kinase), which is a specific effector of Rho (Arimura et al. 2000). In addition, the mutations of unc-33, which is a CRMP-2 homologue of Caenorhabditis elegans, lead to aberrant patterns of axon outgrowth (Hedgecock et al. 1985; Li et al. 1992; Amano et al. 2000). Recently, in the hippocampal neurons of primary culture, CRMP-2 was detected in the distal growing portion of axons. Overexpression of CRMP-2 leads to the formation of additional axons in cultured hippocampal neurons, which suggests CRMP-2 is associated with axon-genesis and neuronal polarity (Inagaki et al. 2001). CRMP-2 is thought to induce axon elongation by promoting microtubule assembly (Fukata et al. 2002). All of these results are intriguing and imply that the CRMP family plays a role in the nerve regeneration process. To understand the implication of CRMP family in nerve regeneration, we examined the functional significance of CRMPs using a rat hypoglossal nerve regeneration model. We revealed that expressions of the CRMP family members are differentially regulated in response to nerve injury and that one member contributes significantly to neurite elongation during nerve regeneration in the rat.

Materials and methods

Operation and in situ hybridization

All the experimental procedures were conducted in accordance with the standard guideline for animal experiments of the Graduate School of Medicine, Osaka City University. Male Wistar rats, weighing approximately 150 g, were anaesthetized with pentobarbital and the right hypoglossal nerve was transected. Following postoperative periods of 1, 3, 5, 7, 14, 21, 28, 35, 42, 49, and 56 days after axotomy, the animals (five rats at each point) were decapitated under anesthesia. Their brains were removed and frozen on powdered dry ice. Sections (18-µm thick) were cut on a cryostat and mounted on 3-aminopropyltriethoxysilane-coated slides. In situ hybridization (ISH) was performed as described previously (Kiryu et al. 1995b). ISH on brain sections was performed using 35S-UTP-labeled complementary RNA probes. For the CRMP family mRNA detection, complementary DNA was obtained from rat brain cDNA library using PCR. Those were complementary to bases 1726–2299 of CRMP-1 (U52102), 1991–2538 of CRMP-2 (Z46882), 1424– 1971 of CRMP-3 (U52103), 816–1349 of CRMP-4 (U52104), and 1010–1448 of CRMP-5 cDNA (NM 023023). The findings were assessed using both film autoradiography and emulsion autoradiography. For the histological control experiment sense probes were used. All sense probes did not show any positive signals.

Cell culture and transfection in the N1E-115 cells

N1E-115 cells were cultured in Dulbecco's modified Eagle's medium (DMEM) with 10% fetal bovine serum (FBS) (Invitrogen, Grand Island, NY, USA) overnight. N1E-115 cells were transiently transfected with pexCA-myc-hCRMP-1, pexCA-myc-rCRMP-2, pexCA-myc-rCRMP-5, and pexCA-LacZ using LipofectAMINE (Invitrogen). Twenty-four hours after transfection, cells were cultured in the 5% serum-containing medium for 24 h. Then they were fixed and stained with anti-c-Myc polyclonal antibody (1 : 500, Santa Cruz Biotechnology, Inc., Santa Cruz, CA, USA) or anti-β-galactosidase polyclonal antibody (1 : 500, Cappel, West Chester, PA, USA). The percentages of cells bearing neurites (length > 20 µm from the cell body) among the transfected cells were measured.

Construction of recombinant adenoviral vector

Recombinant adenoviral vectors were constructed in the following manner. The cDNAs were isolated from whole adult rat brain cDNA library using PCR. Rat CRMP-2 and c-myc tag sequences were fused to the C-terminal end. Then cDNA for the myc-tagged CRMP-2 was inserted into pAxCALNLw, a Cre-loxP system mediated expression cosmid cassette (Sato et al. 1998a, 1998b). It has been reported that CRMP-2ΔC381 functioned as the dominant negative form (Inagaki et al. 2001). Adenoviral vector of myc-CRMP-2ΔC381 was also made by the same methods of the myc-tagged CRMP-2. The modified SCG10 promoter sequence, which has two additional artificially tandem neural-restrictive silencer elements (NRSE) upstream of its sequence (Mori et al. 1990), were inserted into pAxAwNCre (Sato et al. 1998b) for the construction of neuron-specific Cre recombinase expressing adenovirus.

Rat dorsal root ganglion organ culture and adenoviral vector infection

Lumbar DRG was removed aseptically from rats on postnatal day 0 and pooled in ice-cold L15 medium. The DRG were incubated with 100 µL of DMEM with 5% FBS containing pAxCALNLCRMP-2 plus pAxNCre, pAxCALNLCRMP-2ΔC381 plus pAxNCre, or pAxCANLacZ plus pAxNCre at a concentration of 6 × 107 pfu/mL for 2 h in 5% CO2 at 37°C (20 DRGs each). After incubation with adenoviral vector, the adenovirus-containing medium was aspirated, and DRG was buried in 0.4% collagen type 1 gel (Koken, Tokyo, Japan). After the gel had consolidated, DMEM/F12 mixture (1 : 1) with N2 supplement (Invitrogen), and 5 µm BrdU/uridine was added, and the ganglions were cultured in 5% CO2 at 37°C for 3 days.

Immunocytochemical analysis of cultured DRG

After fixation (using 4% paraformaldehyde in PBS), the adenovirus-infected ganglia were incubated with anti-c-Myc polyclonal antibody (1 : 250, Santa Cruz Biotechnology, Inc.) and anti-GAP-43 monoclonal antibody (1 : 250, Sigma, St. Louis, MO, USA). β-Galactosidase was identified using anti-β-galactosidase polyclonal antibody (1 : 500, Cappel). For visualization of the immunocytochemical signals, Alexa 488-labeled anti-rabbit secondary antibodies (1 : 250, Molecular Probes, Eugene, OR, USA) and Alexa 594-labeled anti-mouse secondary antibodies (1 : 250, Molecular Probes) were used. Finally, the length of the longest 20 neurites of each ganglion was measured from the periphery of the ganglia using Zeiss LSM510 software, and the average length of 400 neurites in each group was calculated.

Immunohistochemical analysis of hypoglossal nerve

Under anesthesia, the right hypoglossal nerve was crushed once for 30 s, which yielded a lesion 1 mm in length. Immediately after the operation, the animal's head was positioned in a stereotaxy frame, and using a 33-gauge 10 µL Hamilton syringe either a combination of pAxCALNLCRMP-2 (1.5 µL, 2.0 × 108 pfu/µL) and pAxN2SCGNCre (1.5 µL, 2.0 × 108 pfu/µL) or pAxCALNLNLacZ (1.5 µL, 2.0 × 108 pfu/µL) and pAxN2SCGNCre (1.5 µL, 2.0 × 108 pfu/µL) was injected into the hypoglossal nucleus. Three days after operation, the infected animals were fixed in Zanboni's fixative (2% paraformaldehyde, 0.1% phosphate buffer, and 0.21% picric acid). The brains and right hypoglossal nerves were postfixed in Zanboni's fixative and cryo-protected in 30% sucrose. The brains and nerves were sectioned (14-µm thick). Efficacy of viral infection was assessed by 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside (X-Gal) histochemical analysis (Scholer et al. 1989) and immunostaining with monoclonal anti-c-myc antibody (9E10). Then sections were washed and incubated in FITC-conjugated antibody (Vector Laboratories, Burlingame, CA, USA) for 24 h at 4°C.

Identification of regenerating nerve in the tongue

After anesthesia, the right hypoglossal nerve was crushed, and then these animals were injected with pAxCALNLCRMP-2 (1.5 µL, 2.0 × 108 pfu/µL) and pAxN2SCGNCre (1.5 µL, 2.0 × 108 pfu/µL) at a depth of 2 mm into the hypoglossal nucleus using a 33-gauge 10 µL Hamilton syringe after the head was positioned in a stereotaxy frame. Two weeks after nerve crush, these rats were perfused and their tongues were removed. These tongues were sectioned (20-µm thick), and immunohistochemical analysis was performed to detect CRMP-2 using anti-c-Myc polyclonal antibody (1 : 100, Santa Cruz Biotechnology, Inc.). Alexa 488-labeled anti-rabbit antibodies (1 : 250, Molecular Probes, Inc.) were used as the secondary antibody. Rhodamine-conjugated α-bungarotoxin (1 : 250, Molecular Probe, Inc.) was added to the secondary antibody to detect postsynaptic membrane.

In vivo assay for nerve regeneration

After the right hypoglossal nerve was crushed, a combination of either pAxCALNLCRMP-2 (1.5 µL, 2.0 × 108 pfu/µL) and pAxN2SCGNCre (1.5 µL, 2.0 × 108 pfu/µL) or pAxCALNLNLacZ (1.5 µL, 2.0 × 108 pfu/µL) and pAxN2SCGNCre (1.5 µL, 2.0 × 108 pfu/µL) was injected into the hypoglossal nucleus to allow their infection. Nerve regeneration was assessed, using the retrograde tracer Fluoro-Gold (FG) (Fluorochrom Inc., Englewood, CO, USA), 1, 2, 3, and 4 weeks after nerve crush as described previously (Hirota et al. 1996).


Expression of CRMP family during hypoglossal nerve regeneration

In the normal hypoglossal motor neurons, moderate levels of mRNAs for CRMP-4 and CRMP-5 were observed, and weak hybridization signals for CRMP-1 and CRMP-2 were found. CRMP-3 mRNA expression was not observed in the hypoglossal nucleus under normal conditions. After nerve injury, the mRNA signals for CRMP-1, CRMP-2, and CRMP-5 increased significantly in the injured hypoglossal motor neurons, whereas CRMP-4 mRNA expression level increased slightly, if at all, in the injured hypoglossal nucleus (Fig. 1). CRMP-3 mRNA expression was not found before or after nerve injury. The emulsion autoradiograms counter-stained with thionin clearly demonstrated that the increased hybridization signal (accumulation of silver grains) was localized in injured motor neurons, and not in glial cells (Fig. 1vi–x). A semiquantitative analysis with the film autoradiogram showed that the increased level of mRNAs for CRMP-1, CRMP-2, and CRMP-5 was observed 1–56 days after nerve injury, peaking at around 14 days after axotomy (Fig. 1b). The maximum increases of mRNA level, which mean the maximum ratio of hybridization signal observed between the injured and control sides of hypoglossal nuclei, were 3.5-, 2.5-, and 3.5-fold for CRMP-1, CRMP-2 and CRMP-5, respectively. The expression peaks were relatively later than those of other nerve injury associated molecules, such as growth factor receptors and intracellular signaling molecules.

Figure 1.

(a) Expression of mRNA for CRMP-1 (i,vi), CRMP-2 (ii,vii), CRMP-3 (iii,viii), CRMP-4 (iv,ix), and CRMP-5 (v,x) in the hypoglossal nucleus 14 days after unilateral hypoglossal nerve transection (right side). CRMP-1, CRMP-2, and CRMP-5 mRNAs showed increased expression after axotomy. There was no hybridization signal for CRMP-3 mRNA in both injured (right) and intact (left) hypoglossal nuclei. CRMP-4 mRNA was slightly up-regulated in axotomized nucleus. High-power bright-field photographs counter-stained with thionin demonstrated that localizations of the hybridization signal (accumulation of silver grains; arrows) for CRMP-1 (vi), CRMP-2 (vii), CRMP-4 (ix), and CRMP-5 (x) mRNAs were on injured motor neurons, and not on glial cells. Scale bars = 200 µ m (i–v), and 100 µm (vi–x). C.C, central canal. (b) Semi-quantitative expression profiles of mRNAs for CRMP-1, CRMP-2, CRMP-3, CRMP-4, and CRMP-5 after nerve injury. Each point shows the mean ratio of hybridization signal intensity (axotomized side/control sides). Error bars indicate SEM.

Induction of neurite formation in N1E-115 cells by CRMP-2

As the rat hypoglossal nerve transection study resulted in the increase of CRMP-1, CRMP-2 and CRMP-5 mRNA expression in injured motor neurons, we then examined whether these CRMP family member can attenuate nerve elongation in vitro using N1E-115 cells. Among the three members, CRMP-2 over expression demonstrated substantial neurite elongation activity, whereas CRMP-1 and CRMP-5 did not show such clear neurite elongation (Fig. 2). However it should be mentioned that expressions of CRMP-1 and CRMP-5 sometimes demonstrated a few short microspike-like processes, which were usually not seen (Fig. 2a). Those small processes were apparently distinct from those seen in CRMP-2 expression.

Figure 2.

CRMP-2 among the CRMP family members promotes neurite elongation in the N1E-115 cells. (a) Morphological change of N1E-115 cells after transfection of pexCA-myc-hCRMP-1 (CRMP-1), pexCA-myc-rCRMP-2 (CRMP-2), pexCA-myc-rCRMP-5 (CRMP-5) or pexCA-LacZ (LacZ). Only CRMP-2 promotes neurite elongation. Scale bar = 50 µm. (b) Percentage of N1E-115 cells having neurite whose length is more than 20 µm among the transfected (myc-positive) cells. CRMP-2 transfection significantly induced number of cells, which had longer neurite. Mann–Whitney U-test. p < 0.01.

Promotion of neurite extension of rat DRG by CRMP-2

The above-mentioned experiments strongly suggest the presence of CRPM-2 in neurite elongation of N1E-115 cells, and thus we constructed adenovirus vectors encoding CRMP-2 and its dominant negative form ΔC381. The adenoviral vectors were designed to express CRMP-2 and ΔC381 specifically in neuronal cells by using the Cre-loxP system, and for the expression of Cre recombinase, a neuron-specific promoter designated NCre was used. Using these adenoviral vectors, we next examined the neurite elongation activity in the DRG organ cultures. The DRG, removed from neonatal rats and infected with adenoviruses expressing Cre with CRMP-2 (pAxCALNLCRMP-2), ΔC381 (pAxCALNLCRMP-2ΔC381), or LacZ (pAxCALNLNLacZ), was then embedded into collagen gel and cultured. Three days after infection, neurite lengths were compared among groups. For the detection of exogenous CRMP-2 and ΔC381 expressions, the antibody against c-myc tag was used, and antibody against GAP43 was also used as a neurite marker. In almost all elongated neurites, c-myc and GAP43 immunoreactivities were colocalized. In particular, very intense immunoreactivities for c-myc and GAP43 were found in the distal neurite and growth cone (Fig. 3a). LacZ-expressing DRG demonstrated about 450-µm neurite elongation, whereas CRMP-2 infected DRG showed twofold longer neurites (approximately 800 µm) than LacZ-infected DRG. The neurite length of ΔC381-expressing DRG was significantly shorter than that of LacZ-expressing DRG (Fig. 3b,c).

Figure 3.

CRMP-2 over expression promoted neurite elongation of DRG neurons. (a) High power magnification of neurite tip expressing CRMP-2. In the distal tip of neurites and growth cones, exogenously expressed CRMP-2 is colocalized with a neurite marker GAP-43. Scale bar = 50 µm (i–iii). (b) DRG neurons (postnatal day 0) were infected with pAxCALNLCRMP-2 (CRMP-2), pAxCALNLCRMP-2ΔC381 (ΔC381) or pAxCALNLNLacZ (LacZ), and subsequently buried in collagen gel. After being cultured for 3 days, the elongated neurites were visualized with anti-c-Myc or anti-β-galactosidase antibodies. Apparent neurite elongation was seen in CRMP-2 expression. The neurite length of ΔC381-expressing DRG is shorter than that of LacZ-expressing DRG. Scale bar = 100 µm (iv–ix). (c) The length of the longest 20 neurites of each ganglion was measured from the periphery of the ganglia and the average length of 400 neurites in each group was calculated. Error bars represent SE. Mann–Whitney U-test.

Transportation of CRMP-2 to the tip of regenerating axon and reorganized endplate

We further examined if the newly synthesized CRMP-2 is in fact transported to the regenerating axon and targeted to the axon tip by using c-myc tagged CRMP-2 in vivo. After the hypoglossal nerve crush, the adenovirus (pAxCALNLCRMP-2) was injected into the rat hypoglossal nucleus, and more than half of the hypoglossal motor neurons were infected and expressed exogenous CRMP-2 (Fig. 5a). By this infection, c-myc tagged CRMP-2 could be found only in regenerating axon or endplates. Three days after nerve injury, when most regenerating nerves entered the nerve distal to the injured site, the injured hypoglossal nerve was examined with c-myc antibody. Numerous axons, which were positive for c-myc antibody, were observed in the distal injured nerve (Fig. 4a,b), suggesting newly synthesized CRMP-2 is transported to the distal regenerating axon. Two weeks after nerve crush, when a small number of regenerating axons entered the tongue and reformed endplates, regenerating nerve was found in endplates of the tongue muscle. The postsynaptic part of the endplates was identified by bungarotoxin staining, and regenerating nerve was identified by c-myc immunoreactivity. As shown in Fig. 4(c), newly synthesized CRMP-2 was confirmed to be transported into the nerve terminal and localized in the presynaptic site of the endplates.

Figure 5.

Expression of CRMP-2 accelerates nerve regeneration in adult rat. (i) pAxCALNLCRMP-2 (CRMP-2) and pAxN2SCGNCre (Cre) infected motor neurons in rat hypoglossal nuclei (expression was detected by anti-c-myc antibody 1 week after infection) (i). pAxCALNLNLacZ (LacZ) and pAxN2SCGNCre (Cre) infected motor neurons (expression was detected by X-gal staining 1 week after injection) (ii). Scale bar = 200 µm. (ii) FG-positive neurons after infection of hypoglossal motor neurons with pAxCALNLCRMP-2 plus pAxN2SCGNCre (iii and v) or pAxCALNLNLacZ plus pAxN2SCGNCre (iv and vi). All right sided hypoglossal nuclei are on the injured side, 2 weeks after injury (iii and iv), LacZ expressing motor neurons show no FG-positive neurons (regenerated) on the injured side (iv), whereas CRMP-2 expressing motor neurons show approximately five FG-positive (regenerated) cells per section in injured nucleus (iii). At 4 weeks after nerve injury, FG-positive cells increased in both virus-infected nuclei (v and vi); however, the FG-positive cell number in the right hand nucleus of (v) significantly exceeded that in right hand nucleus of (vi). Scale bar = 200 µm. (iii) Time-course of nerve reprojection rate after nerve crush in rats infected with either CRMP-2 (pAxCALNLCRMP-2 and pAxN2SCGNCre) or LacZ (pAxCALNLNLacZ and pAxN2SCGNCre). The regeneration rate was assessed by calculating the percent ratio of the number of FG-stained cell bodies in the operated side to those in the control side. Five animals were examined at each time point. Asterisks denote statistically significant differences. Mann–Whitney U-test, p < 0.01.

Figure 4.

CRMP-2 is transported to regenerating nerve and its terminal. (a) A low-power view of the right hypoglossal nerve 3 days after crush. The exogenous CRMP-2 was visualized by c-myc tag immunostaining (FITC). (b) A high-power view of the regenerating nerve in (a). Positive CRMP-2 staining was seen in the regenerating nerve distal to the crush site. (c) Exogenously expressed CRMP-2 was also localized at the newly organized neuromuscular junction. Rhodamine-conjugated-α-bungarotoxin (red) was used for visualization of the postsynaptic structure. Alexa 488 fluorescence (green) was used to detect tagged CRMP-2. The tagged CRMP-2 staining in re-projected nerve was observed in the neuromuscular junction. Scale bars = 100 µm (a), 200 µm (b), and 30 µm (c).

Promotion of axon regeneration in rat by CRMP-2

Because neurite elongation was induced in the CRMP-2-expressing DRG neurons and N1E-115 cells, and synthesized CRMP-2 was actively transported to the distal regenerating axons, we expected that the axon elongation activity might be enhanced by CRMP-2 overexpression in regenerating motor neurons as well. To evaluate an enhancement of axon regeneration by CRMP-2, we used the retrograde tracer (FG). When FG is injected into tongue after nerve injury, it is taken from axon terminals, and retrogradely transported into cell bodies, thereby, motor neurons that reproject into the tongue are labeled by FG, but those not yet reprojected into the tongue are not labeled (Hirota et al. 1996; Namikawa et al. 2000). FG was injected into both sides of the tongue 2 days before the rats were killed to allow retrograde transport of the dye into the neuronal cell bodies. Animals were killed at 1, 2, 3, and 4 weeks after nerve crush, and FG-positive cells were counted on both injured and control sides. The axon-regenerating rate was calculated by comparing FG-positive cell numbers in the intact (left) and the injured (right) sides of hypoglossal nuclei. One week after nerve injury, FG-positive neurons were not detected in injured hypoglossal nuclei of both CRMP-2-infected and LacZ-infected rats, although numerous FG-positive cells were found in the intact side of the hypoglossal nuclei. Two weeks after nerve injury, 5–10 FG positive neurons per section were identified in CRMP-2-overexpressing motor neurons of the injured side; whereas almost no FG-positive cells were found in the LacZ-overexpressing injured motor neurons (Fig. 5biii,biv). Four weeks after nerve injury, approximately 80% of injured motor neurons, which were overexpressing CRMP-2, reprojected into the tongue, whereas approximately 50% of LacZ-expressing motor neurons reprojected (Fig. 5bv,bvi). The regeneration ratio in the CRMP-2-expressing neurons exceeded those in the LacZ-overexpressing ones by 20–30% (Fig. 5c). This result indicates that CRMP-2 overexpression has a significant effect on axonal regeneration in vivo.


The present study revealed that CRMP-2 overexpression succeeded in elongating neurites in both N1E-115 cells and organ-cultured DRG and further enhanced regeneration of injured motor neurons.

Previously, using a hypoglossal nerve axotomy model, we revealed the implication of various molecules during the nerve regeneration process (Kiryu et al. 1995a; Namikawa et al. 1998; Nakagomi et al. 1999; Tanabe et al. 1999, 2000). In the present study, a positive involvement of CRMP-2 in nerve regeneration was clarified. Although the function of this family in nerve regeneration has been poorly understood, the family has an association with neurite elongation during neural development and neurite formation of some cell lines (Minturn et al. 1995; Wang and Strittmatter 1996; Byk et al. 1998; Fukada et al. 2000). In addition, Minturn et al. (1995) reported an increase of CRMP-2 mRNA in nerve-injured spinal motor neurons, suggesting an association of CRMP-2 with nerve regeneration. We then attempted to identify gene expression profiles in response to nerve injury among the family members using a hypoglossal nerve regeneration model. This study showed that three members of the CRMP family (CRMP-1, CRMP-2 and CRMP-5) respond to nerve injury and the other two (CRMP-3 and CRMP-4) do not. This may suggest a functional variation in the CRMP family, although precise functional differences among family members have not been elucidated (Ricard et al. 2001). The increase ratio of the mRNAs for CRMP-1 and CRMP-5 after nerve injury appears higher than that of CRMP-2 in Fig. 1(b), however, the ratio demonstrated does not necessarily mean the absolute amount of mRNAs. The biased expression of CRMP-2 mRNA in non-injured motor neurons might be considered. Another intriguing finding is that CRMP-1, CRMP-2 and CRMP-5 mRNA expressions peak around 14 days after nerve injury. Previous studies have shown that molecules associated with cell survival, such as glutamate transporters, growth factor receptors, and intracellular signaling molecules, are expressed with a peak around 3 days or earlier after nerve injury (Kiryu et al. 1995a; Namikawa et al. 2000). As for some transcription factors such as c-Jun (Herdegen et al. 1997), the maximum expression period after nerve injury is much earlier. Therefore, the CRMP family could be categorized into a late gene response family during nerve regeneration. To date, the members of this late gene response family are limited, and typical are those of the SCG10 family (Iwata et al. 2002). SCG10 family members, such as stathmin, Rb3, and SCG10 itself, induce their mRNA peak expression around 10–14 days after hypoglossal nerve injury (Beilharz et al. 1998; Iwata et al. 2002). Some studies demonstrate that the SCG10 family is a microtubule regulation factor (Charbaut et al. 2001), and interestingly the association of CRMP-2 with microtubule dynamics is also suggested (Gu and Ihara 2000). In addition, it has recently been demonstrated that CRMP-2 binds to tubulin heterodimer and promotes microtubule assembly (Fukata et al. 2002). Taken together, the CRMP family and SCG10 family may function at a similar point, such as microtubule dynamics during the nerve regeneration process, and their functions may be opposite in microtubule dynamics and presumably do not function as survival factors.

The present study revealed that CRMP-2 but not CRMP-1 and CRMP-5 promotes substantial neurite elongation in N1E-115 cells. Previously, the neurite-promoting activity of CRMP-2 was demonstrated in PC12 and P19 cells in response to the stimuli of nerve growth factor and retinoic acid, respectively (Minturn et al. 1995). In addition cAMP stimulated NG108 cells express significant amount of CRMP-2 but not CRMP-1 and -5 (our unpublished data). These findings perhaps suggest that whatever the stimulus is, CRMP-2 expression is tightly linked with the phenomenon of neurite formation. This could be the reason why CRMP-2 expression is induced during development, in particular from the late embryonic stage to 1 week after birth (Wang and Strittmatter 1996; Fukada et al. 2000). Although heterophilic oligomerization of CRMPs may be important for neurite formation (Fukada et al. 2000), an increase of CRMP-2 seems pivotal, at least for the neurite formation of N1E-115 cells.

We constructed adenovirus vectors expressing CRMP-2 and a dominant negative form of CRMP-2, and the neurite elongation activity was further examined using organ culture of DRG. In DRG organ culture, CRMP-2 overexpression strikingly promoted neurite extension of DRG neurons, and tended to accumulate in the neurite tips, where marked colocalization of CRMP-2 and GAP-43 was observed. Furthermore, a dominant negative form of CRMP-2, ΔC381-overexpressing DRG was inhibited neurite formation about 100 µm compared with LacZ-overexpressing DRG. This CRMP-2-induced neurite elongation was not apparent in NG108 cells and undifferentiated PC12 cells (our unpublished data), whereas dramatic neurite elongation by CRMP-2 expression is evident in the primary culture of hippocampal cells (Inagaki et al. 2001) and N1E-115 cells (Fukata et al. 2002). This may suggest that CRMP-2 is not the primary factor for neurite formation, and some additional factor, which is already expressed in DRG and N1E-115 cells, may be necessary to elicit neurite elongation activity of CRMP-2.

In this study, we showed that CRMP-2 is accumulated in regenerating nerve and promoted nerve regeneration in vivo. It has been reported that CRMP-2 is associated with microtubule dynamics and may have the stabilizing effect of overexpressed CRMP-2 on bundles of microtubule structures (Gu and Ihara 2000). Furthermore, it has been suggested that CRMP-2 binds to tubulin heterodimer and promotes microtubule assembly in the process of axon outgrowth (Fukata et al. 2002). Microtubule dynamics are crucial for nerve regeneration. When neurites elongate, it is likely that both dynamic assembly and disassembly of microtubules are required near the growth cone. This could be a reason why both the SCG family and CRMP-2, which function as microtubule destabilizing and stabilizing factors, respectively, increase in parallel during nerve regeneration (Iwata et al. 2002).

Finally, we demonstrated the neurite elongation promoting activity of CRMP-2 in hypoglossal nerve regeneration by adenovirus mediated gene transfer. As shown in Fig. 5(a), an efficient preferential expression of CRPM-2 in neuronal cells was achieved by the use of neuron specific SCG10 promoter, which could be further activated after nerve injury because SCG10 expression is also induced by axotomy. Using this efficient gene transfer system, CRMP-2 overexpression accelerated nerve regeneration by approximately 25–30%. This strongly implicates CRMP-2 in nerve regeneration as a neurite elongation regulator, although not as the primary factor. Previously, the highest CRMP-2 expression was observed during neurite elongation, including cell differentiation, neuronal development, and nerve regeneration (Quinn et al. 1999), and our study clearly demonstrated CRMP-2 induced neurite elongation in DRG culture and rat regenerating nerve. Although it is clear that CRMPs represent a major functional link between plexin activation and the resulting growth cone collapse (Liu and Strittmatter 2001), the consequences of CRMP-2 overexpression emphasize the implication of CRMP-2 in nerve regeneration as well. Growth cone collapse might be associated with actin dynamics by CRMP-2, whereas neurite outgrowth might be due to microtubule dynamics by CRMP-2 (Gu and Ihara 2000; Inagaki et al. 2001; Fukata et al. 2002).

In conclusion, CRMP-2 is a crucial neurite elongation factor, not only in cultured cells and development, but also in nerve regeneration, and CRMP-2 overexpression promotes nerve regeneration of rat motor neurons. At present, although the CRMP-2 overexpression is sufficient for promoting axon regeneration in animal, it is uncertain if CRMP-2 is essential for the regenerating axon in animal. The molecular mechanism underlying the nerve regeneration activity of CRMP-2 has to be elucidated further, but it is most likely that microtubule assembly and stabilization by CRMP are involved. The simultaneous regulation of both actin and microtubules dynamics by CRMP may be necessary for proper nerve regeneration.


We thank Dr I. Saito (University of Tokyo) for providing pAxCALNLNLacZ, Dr Yanagi for providing CRMP-5 plasmid. This work was supported in part by Grant-in-Aid MEXT and MHLW, Japan.