Collapsin response mediating protein-2 (CRMP2) has been identified as an intracellular protein mediating Semaphorin3A (Sema3A), a repulsive guidance molecule. In this study, we demonstrate that cyclin-dependent kinase 5 (Cdk5) and glycogen synthase kinase 3β (GSK3β) plays a critical role in Sema3A signalling. In In vitro kinase assay, Cdk5 phosphorylated CRMP2 at Ser522, while GSK3β did not induce any phosphorylation of CRMP2. Phosphorylation by GSK3β was exclusively observed in Cdk5-phosphorylated CRMP2, but barely in CRMP2T509A. These results indicate that Cdk5 primarily phosphorylates CRMP2 at Ser522 and GSK3β secondarily phosphorylates at Thr509. The dual-phosphorylated CRMP2, but not non-phosphorylated or single-phosphorylated CRMP2, is recognized with the antibody 3F4, which is highly reactive with the neurofibrillary tangles of Alzheimer's disease. 3F4 recognized the CRMP2 in the wild-type but not cdk5−/− mouse embryonic brain lysates. The phosphorylation of CRMP2 at Ser522 caused reduction of its affinity to tubulin. In dorsal root ganglion neurones, Sema3A stimulation enhanced the levels of the phosphorylated form of CRMP2 detected by 3F4. Over-expression of CRMP2 mutant substituting either Ser522 or Thr509 to Ala attenuates Sema3A-induced growth cone collapse response. These results suggest that the sequential phosphorylation of CRMP is an important process of Sema3A signalling and the same mechanism may have some relevance to the pathological aggregation of the microtubule-associated proteins.
Neurones form precise patterns of connections that emerge through the interaction between the growth cone and extracellular signals in the developing nervous system. Neuronal extension from somas and navigating extensive length to their crucial targets are controlled by four types of molecular guidance cues, which can be either short-range or long-range, along the pathway (Tessier-Lavigne & Goodman 1996). As in other cell types, morphological changes and motility of neuronal growth cones are closely related to reorganization of actin, tubulin and other cytoskeletal proteins, and interplay between actin and microtubule cytoskeletons has been shown to be critical for growth cone navigation. Most of the research has focused on the actin, because it underlines growth cone motility, the extension and retraction of filopodia and lamellipodia, and these structures are the first to encounter guidance cues during growth cone advance. More recently, accumulating evidence suggests that microtubule cytoskeleton also plays a pivotal role in growth cone path finding and is also regulated by guidance cues operating through intracellular signalling pathways (Mack et al. 2000; Buck & Zheng 2002; Gordon-Weeks 2004; Zhou & Cohan 2004). However, the whole picture of the mechanisms involved in transducing various kinds of external signals to these cytoskeletons for neuronal guidance remains obscure.
The semaphorins constitute a major family of axon guidance cues in central as well as peripheral nervous system (Kolodkin & Ginty 1997; Raper 2000). Semaphorin3A (Sema3A) repulses axons through the co-receptor protein neuropilin-1 (Kolodkin & Ginty 1997) and plexin-As. Neuropilins and plexin-As are ligand binding and signal-transducing subunits of class 3 semaphorin receptor complexes, respectively. Plexin-A includes a cytoplasmic segment that transduces the Sema3A signal into cytoskeletal reorganization. The chicken collapsin (formerly called semaphorin) response mediator protein (CRMP)-62 molecule (also known as CRMP2) was originally identified as a signalling molecule of Sema3A. Antibodies to a specific region of CRMPs block Sema3A-induced growth cone collapse (Goshima et al. 1995). CRMP2, which has also been independently identified, is one of at least five isoforms (CRMP1–5). The interactions among CRMP isomers favor heterophilic oligomerization over homophilic oligomerization (Wang & Strittmatter 1997). CRMP2 binds tubulin heterodimers with higher affinity than it binds microtubules and promotes microtubule assembly (Fukata et al. 2002). This raises the possibility that CRMPs act as downstream components of semaphorin-PlexAs signal transduction pathway through regulating cytoskeletal reorganization.
CRMPs are extensively phosphorylated during neuronal development, and the phosphorylation states of CRMPs are altered in response to NGF and lysophosphatidic acid (Byk et al. 1996; Yoshida et al. 1998; Arimura et al. 2000). CRMP2 was also identified as an antigen for 3F4 monoclonal antibody in the brain of patients affected with Alzheimer's disease (AD) (Yoshida et al. 1998; Gu et al. 2000). 3F4 was obtained from a number of monoclonal antibodies raised against crude paired helical filaments (PHF). 3F4 recognizes CRMP2 phosphorylated at Thr509, Ser518, and Ser522 (Gu et al. 2000). The mean level of 3F4 antigen in AD brains is significantly higher than that in control brains (Yoshida et al. 1998). 3F4 intensely reacts with neurofibrillary tangles (NFT) and occasionally with senile plaque neurites. Immunoelectron microscopy showed that 3F4 decorates twisted PHF-like and straight tubules in perikaryal tangles. This suggests that hyperphosphorylated CRMP2 is a building block of PHF, and raises the possibility that hyperphosphorylated CRMP2 as well as tau is involved in the development of neurofibrillary pathology (Lee et al. 2001). The biological and/or pathophysiological significance of phosphorylation of CRMP2, however, has not been established.
To elucidate cytoskeletal reorganization mediated with external guidance signals, we previously examined growth cone collapse induced by Sema3A and kinase inhibitors and mutant mice. We found that sequential activation of Fyn and Cdk5 is involved in Sema3A responses (Sasaki et al. 2002). Cdk5 is one of proline-directed Ser/Thr kinases which phosphorylate serine or threonine followed by a proline residue, and plays critical roles for brain development by regulating the migration of neurones as well as axon guidance. This led us to investigate whether CRMP2 is a substrate of Cdk5, and whether phosphorylation of CRMP2 is involved in responses to Sema3A. Further, in order to identify in vivo substrates of Cdk5, we compared the phospho-proteins in cdk5−/– brain homogenates with those in the wild-type. We here identify CRMP2 as a substrate of Cdk5. The Cdk5-primed CRMP2 phosphorylation by GSK3β is a critical step in mediating Sema3A-signalling and in generating a specific antigenicity of AD brains. Introduction of non-phosphorylated mutants or small-interfering (si) RNA of CRMP2 into neurones suppresses the Sema3A-induced response. This finding demonstrates the role of CRMP2 in mediating external signals of the axon guidance cues, and has implication for neurofibrillary tangles in AD.
CRMPs are substrates of Cdk5
To examine whether CRMP was a substrate for Cdk5, we performed in vitro kinase assay using purified CRMP2, Cdk5, and p25 or p35, an activator of Cdk5, with [γ-32P]ATP. CRMP2 was phosphorylated in the presence of both Cdk5 and p25 (Fig. 1A). Cdk5/p35 enhanced phosphorylation of CRMP2 in HEK293T cells metabolically labelled with [32P]orthophosphate (Fig. 1B). Phosphorylation of CRMP1, 4 and 5 was also enhanced by Cdk5/p35 in HEK293T cells, a result consistent with the fact that CRMP1, 4 and 5 have the consensus sequence of phosphorylation by Cdk5 near 522 (Fig. 2A). In the present study, we focused on phosphorylation of CRMP2, because CRMP2 has been implicated in the Sema3A-signalling (Goshima et al. 1995), and was also identified as an in vivo substrate for Cdk5 (see below).
Cdk5 is one of proline-directed Ser/Thr kinases, which phosphorylate serine or threonine, followed by a proline residue. We searched for in vivo Cdk5 substrates using cdk5-deficient mice and a monoclonal antibody, which binds to phospho-threonine followed by proline (pThr-Pro antibody). We applied Immunoblot analysis with this pThr-Pro antibody to the comparison of phospho-proteins in the brain lysates from cdk5+/+ and cdk5−/– embryos. We found that three bands were in low intensity or missing in cdk5−/– brain lysate. These bands corresponded to approximately 65 kDa, 70 kDa and 120 kDa in SDS-PAGE electrophoresis (Fig. 1C). 2-D gel electrophoresis followed by Western blot with pThr-Pro antibody showed that five spots were in very low intensity or absent in cdk5−/– brain lysate (Fig. 1D). Spot 1 and 2 corresponded to 65 kDa, spot 3 and 4 to 70 kDa, and spot 5 to 120 kDa bands. MALDI-TOF mass spectrometry (MS) analysis and database search indicated that spot 3 fitted to CRMP2.
We tested the possibility that 70 kDa band was derived from CRMP2. We performed immunoblot analysis of 2-D gel using pThr-Pro antibody and C4G, which is phosphorylation-independent anti-CRMP2 monoclonal antibody. CRMP2 blot made multiple signals in 2-D gel electrophoresis, and two of them matched spot 3 and 4 (Fig. 1E). Phosphorylation of CRMP2 was previously reported in AD patient brains using monoclonal antibody 3F4. 3F4 recognizes three sites of phosphorylation of CRMP2 at Thr509, Ser518 and Ser522 (Gu et al. 2000). Immunoblot analysis with this monoclonal antibody confirmed that spots 3 and 4 completely matched those detected by 3F4 and C4G antibodies (Fig. 1E). Furthermore, immunoblotting with these antibodies revealed that CRMP2 was not phosphorylated at 3F4 recognition site in cdk5–/– embryonic brain (Fig. 1F). These results indicate that CRMP2 was phosphorylated by Cdk5 in the embryonic mouse brain, and the phosphorylation site(s) was (were) Thr509 and/or Ser518 and/or Ser522.
Cdk5 phosphorylates CRMP2 at Ser522 in vitro and in vivo
The consensus sequence of phosphorylation by Cdk5 has been reported to be amino acid sequence (S/T)PX(K/H/R), where S or T is the phosphorylatable serine or threonine (Songyang et al. 1996). Amino acids 522–525 (SPAK) of CRMP2 match with the consensus sequence (Fig. 2A). To test whether Ser522 was the phosphorylation site of CRMP2 by Cdk5, we produced a non-phosphorylated CRMP2 mutant, CRMP2S522A, in which Ser522 was replaced by Ala. In in vitro kinase assay, Cdk5 phosphorylated wild-type CRMP2 (CRMP2wt), but not CRMP2S522A (Fig. 2B). The result was also confirmed in HEK293T cells. Myc-tagged CRMP2wt or CRMP2S522A was expressed alone or with Cdk5/p35 in HEK293T cells, and then labelled with [32P]orthophosphate. Co-expression of Myc-tagged CRMP2wt and Cdk5/p35 increased phosphorylation of CRMP2wt as compared with CRMP2wt alone, while no labelling was observed when CRMP2S522 A and Cdk5/p35 were co-expressed (Fig. 2C). These results indicate that Ser522 was the major site of CRMP2 phosphorylation by Cdk5 in vitro.
Next, we prepared a rabbit polyclonal antibody that specifically recognizes CRMP2 phosphorylated at Ser522. We used a phosphopeptide corresponding to amino acids 516–528 of CRMP2 in which Ser522 was phosphorylated as an antigen. The specificity of this antibody was examined by immunoblot analysis of CRMP family expressed in HEK293T cells. The immunoreactive band intensities of CRMP1 and CRMP2 increased under condition of co-expression of Cdk5/p25, compared to expression of either CRMP1 or CRMP2 alone. This antibody recognized neither CRMP3-5 (Fig. 2D) nor CRMP2S522A (data not shown) even in the presence of Cdk5/p25. These results indicate that this antibody specifically recognized CRMP1 and CRMP2 phosphorylated at Ser522 in vitro. We thus named this antibody anti-pS522-CRMP1/2.
Immunoblot analysis using anti-pS522-CRMP1/2 antibody revealed that the bands were in low intensity or missing in lysate of embryonic brains and neuronal culture from cdk5–/– when compared to cdk5+/+ mice. This fact indicates that CRMP1 and/or CRMP2 were phosphorylated at Ser522 by Cdk5 in vivo (Fig. 2E).
Cdk5 primed GSK3β phosphorylation of CRMP2
The threonine phosphorylation was not seen with brain lysate from cdk5−/– mice (Fig. 1C,D). This implies that CRMP2 was also phosphorylated at the threonine residue(s) in vivo. In addition, no band shift was observed with CRMP2S522A co-expressed with Cdk5/p35 (Fig. 2C). We thus speculated that phosphorylation of Ser522 was required for phosphorylation of other sites, and that other Ser/Thr kinase(s) might be involved in phosphorylation of CRMP2 at threonine residue. CRMP2 has been recently shown to be associated with PHF, microtubules, and bind to tubulin dimers (Yoshida et al. 1998; Gu & Ihara 2000; Fukata et al. 2002), being reminiscent of tau protein. Tau is also known for its Cdk5-primed phosphorylation by a Ser/Thr kinase GSK3β (Arioka et al. 1995; Lee et al. 2001).
When CRMP2 was co-expressed with GSK3β, CRMP2 was phosphorylated and intensity of upper band increased compared with CRMP2 co-expressed with Cdk5/p35 in HEK293T cells (Fig. 3A). When co-expressed with GSK3β and Cdk5/p35, CRMP2 was more phosphorylated and the upper band intensity relatively increased, thereby indicating phosphorylation at additional sites. No phosphorylation and band shift were observed in the co-expression of CRMP2S522A with Cdk5/p35, GSK3β or both. This result suggests that GSK3β as well as Cdk5 phosphorylated CRMP2 at Ser522, or Cdk5 phosphorylation of Ser522 was essential for the additional phosphorylation of CRMP2 by GSK3β. To test this idea, we performed in vitro kinase assay using purified protein without endogenous kinase. GSK3β alone did not phosphorylate CRMP2 (Fig. 3B). We examined whether the phosphorylation by Cdk5 enhances additional phosphorylation of CRMP2 by GSK3β. For this purpose, we prepared GST-CRMP2c (aa486 to carboxyl terminus), and then phosphorylated this GST-fusion protein by immobilized Cdk5/p35 bound to anti-p35 antibody and protein-A beads, with non-labelled ATP in vitro. After removing the immobilized Cdk5/p35, we mixed the phosphorylated GST-CRMP2c and GSK3β with [γ-32P]ATP for kinase reaction (Fig. 3C,a). Incorporation of [32P]phosphate was exclusively observed in Cdk5-phosphorylated GST-CRMP2cwt, but not GST-CRMP2cS522A, confirming the essential role of Cdk5 phosphorylation for the further phosphorylation of CRMP2 by GSK3β (Fig. 3C,b).
3F4 recognized the CRMP2 in the wild-type mouse embryonic brains but not cdk5–/– mouse brain (Fig. 1F). We therefore deduced that the Thr509 and/or Ser518 were putative GSK3β phosphorylation site(s). Immunoblot analysis using pThr-Pro antibody suggested that CRMP2 was phosphorylated at threonine residue in wild-type mouse brain (Fig. 1C,D). We thus produced GST-CRMP2cT509A and performed kinase assay to evaluate Cdk5-primed GSK3β phosphorylation. Incorporation of [32P]phosphate into the CRMP2cT509A was markedly lower when compared to CRMP2cwt. This result indicates that Thr509 was the phosphorylation site of GSK3β in this region of CRMP2 (Fig. 3C,b). The residual level of phosphorylation may be attributed to another consensus site for GSK3β, Ser518 and Thr514. We next examined whether CRMP2 acquired antigenicity for 3F4 through phosphorylation by Cdk5 and GSK3β. Myc-tagged CRMP2wt and the mutant proteins were expressed in HEK293T cells, and immunoprecipitated with anti-Myc antibody. CRMP2wt and the mutants thus prepared were incubated with both Cdk5 and GSK3β. In immunoblot analysis, 3F4 recognized CRMP2wt more than CRMP2T509A, and did not recognize CRMP2S522A or CRMP2T509/S522A (Fig. 3D). To confirm this result, we performed immunoblot analysis using 3F4 in HEK293T cells expressing Myc-tagged CRMP2, Cdk5/p35 and/or GSK3β. 3F4 recognized CRMP2 expressed with Cdk5/p35 and GSK3β, but only faintly with Cdk5/p35 alone (Fig. 3E). When Cdk5/p35 and GSK3β were co-expressed, 3F4 barely recognized CRMP2T509A, and never CRMP2S522A, suggesting that this antibody mainly recognized CRMP2 phosphorylation at Thr509 and Ser522. These results show that Cdk5-primed GSK3β phosphorylation occurred in vitro and in vivo, giving CRMP2 the antigenicity for 3F4.
Co-expression of CRMP2 with GSK3β alone caused CRMP2 phosphorylation, the level of which was comparable to that with Cdk5 and GSK3β (Fig. 3E). This was probably attributed to intrinsic Cdc2 kinase activity in HEK293T cells. Indeed, our in vitro study indicated that Cdc2 kinase could phosphorylate CRMP2 at Ser522 (data not shown).
The phosphorylation of CRMP2 disrupts association of CRMP2 with tubulin
We examined whether Cdk5-primed phosphorylation of CRMP2 regulated cytoskeletal reorganization for semaphorin-induced responses. It is well known that microtubule associated protein tau, when phosphorylated by Cdk5 and GSK3β, decreases its affinity to microtubules and reduces their stability (Jameson et al. 1980; Lindwall & Cole 1984). Recently, CRMP2 was shown to be associated with microtubules, and bound to tubulin heterodimers (Gu & Ihara 2000; Fukata et al. 2002). We therefore examined whether the interaction between CRMP2 and tubulin was altered by CRMP2 phosphorylation by Cdk5 and GSK3β. In HEK293T cells expressing CRMP2 alone, CRMP2 was co-immunoprecipitated with endogenous tubulin. This immunoprecipitation with anti-α tubulin antibody became barely detectable when CRMP2 was co-expressed with Cdk5/p35 or Cdk5/p35 plus GSK3β. When CRMP2S522A was expressed with Cdk5/p35 and GSK3β, the association of the mutant CRMP2 with tubulin was not altered when compared to CRMP2S522A alone (Fig. 4A). This indicates that Cdk5 and GSK3β phosphorylated CRMP2, and this lowered the affinity of CRMP2 to tubulin in vitro.
We performed co-immunoprecipitation experiment using brain lysates from cdk5+/+ and cdk5–/– embryos. The lysates were immunoprecipitated with anti-α tubulin antibody and subjected to immunoblotting with anti-CRMP2 antibody (Fig. 4B,a). The association of CRMP2 with tubulin was enhanced by 2-fold in cdk5–/– embryonic brain, when compared to wild-type brain (Fig. 4B,b), thereby indicating that CRMP2 phosphorylated by Cdk5 reduced its affinity for tubulin in the embryonic mouse brain. To determine the role of GSK3β, we estimated the affinity of CRMP2 to tubulin in cell lysate of cortical neurone primary culture with or without LiCl, a GSK3β inhibitor. LiCl induced band shifting of CRMP2 to the lower molecular weight, suggesting that LiCl inhibited phosphorylation of CRMP2 (Fig. 4C, a). The treatment of LiCl increased the amount of CRMP2 co-immunoprecipitated with tubulin when compared to the untreated control (Fig. 4C,b). These results support the idea that sequential phosphorylation of CRMP2 by Cdk5 and GSK3β is necessary to negatively regulate the association of CRMP2 and tubulin.
Sema3A causes dual phosphorylation of CRMP2, and introduction of non-phosphorylated mutants or knockdown of CRMP2 suppresses Sema3A-induced growth cone collapse
To determine the physiological significance of phosphorylation of CRMP2, we first examined the phosphorylation levels of CRMP after Sema3A stimulation. In immunoblot analysis using anti-pS522-CRMP1/2 antibody, Sema3A (1 nm) increased phosphorylation of endogenous CRMP1 and/or CRMP2 at Ser522 within 10 min, and the increased phosphorylation persisted for up to 30 min. Olomoucine, a cyclin-dependent kinase inhibitor, inhibited Sema3A-induced phosphorylation of CRMP1 and/or CRMP2 (Fig. 5A,a). These results indicated that Cdk5 was involved in phosphorylation of CRMP1 and/or CRMP2 at Ser522 after Sema3A stimulation. Importantly, Sema3A stimulation enhanced the levels of the phosphorylated form of CRMP2, which was detected by 3F4 (Fig. 5A,b). To localize CRMP1 and CRMP2 in DRG neurones, we performed immunocytochemical examination using specific antibodies raised against CRMP1 and CRMP2, respectively (Ricard et al. 2001). CRMP1 and CRMP2 were found to be differentially expressed in the growth cones. CRMP1 was mainly localized in proximal shafts, while CRMP2 was in the central domain of growth cones and some filopodia as well as axon shafts (Fig. 5B,a). We then explored the immunofluorescence intensity with anti-pS522-CRMP1/2 antibody and anti-CRMPs antibodies in the growth cones after Sema3A stimulation. The immunofluorescence with anti-pS522-CRMP1/2 antibody in the growth cones may reflect with the levels of phosphorylated CRMP2, because the relative intensity of immunofluorescence signal at the growth cones detected with anti-CRMP2 antibody predominated over that with anti-CRMP1 antibody (Fig. 5B,a). The levels of phosphorylated CRMP2 at Ser522 was increased by Sema3A stimulation (Fig. 5B,b and B,c). Interestingly, the immunofluorescence levels with anti-CRMP2 antibody, which recognizes both non-phosphorylated and phosphorylated form of CRMP2, was slightly but significantly enhanced by Sema3A (Fig. 5B,b), suggesting that CRMP2 may be recruited to the growth cone area in response to Sema3A. Further, the ratio of the mean value of immunofluorescence intensity of the phosphorylated CRMP2 to that of total CRMP2 was enhanced by 1.5–2-fold at the growth cones 1 and 10 min after Sema3A treatment (Fig. 5B,c). The immunofluorescence with anti-pS522-CRMP1/2 antibody at the shafts also enhanced in response to Sema3A stimulation (see Fig. 5B,a, data not shown). We confirmed that olomoucine suppressed Sema3A-induced increase in the fluorescence levels at the growth cones (Fig. 5B,b). These morphological findings are consistent with those of immunoblot analysis.
To examine the roles of CRMP2 phosphorylation by Cdk5 and GSK3β in growth cone morphology, we expressed the CRMP2wt, unphosphorylated mutants T509A, S522A, or T509A/S522A in DRG neurones using herpes simplex virus (HSV) gene transfer. Introduction of CRMP2T509A, CRMP2S522A and CRMP2T509A/S522A into DRG neurones significantly suppressed Sema3A-induced growth cone collapse, while that of CRMP2wt did not show any effect (Fig. 6A,a,b). To further determine whether both Cdk5 and GSK3β were the same pathway through CRMP2 in Sema3A signalling, we examined the effect of olomoucine (10 µm) and a GSK3β inhibitor TDZD-8 (125 nm) on Sema3A-induced growth cone collapse. Either olomoucine or TDZD-8 alone suppressed Sema3A-induced growth cone collapse, but combined application of olomoucine and TDZD-8 showed no additive effect on Sema3A-induced growth cone collapse (Fig. 6A,c). These results together suggest that the dual phosphorylation of CRMP2 by Cdk5 and GSK3β is involved in Sema3A-induced growth cone collapse.
To determine whether both CRMP1 and CRMP2 are involved in Sema3A-induced collapse response, we employed siRNA method. Specific suppression of CRMP1 or CRMP2 expression by corresponding siRNAs was confirmed by immunocytochemistry (data not shown). Knockdown of CRMP1 or CRMP2 inhibited Sema3A-induced growth cone collapse, indicating that both CRMP1 and CRMP2 were required for mediating the response to Sema3A (Fig. 6B).
CRMP2 is an in vivo substrate for Cdk5 and acts as an intracellular mediator of axon guidance molecules
Using cdk5-deficient mice, we discovered CRMP2 as an in vivo substrate of Cdk5 among CRMP family proteins. Our study demonstrates ordered phosphorylation process of CRMP2 by Cdk5 and GSK3β, with the former being prerequisite for the latter phosphorylation, and that this is involved in mediating the response to Sema3A. The phosphorylation gave CRMP2 the AD-related antigenicity recognized by 3F4 monoclonal antibody (Yoshida et al. 1998). Introduction of non-phosphorylated forms of CRMP2 suppressed Sema3A-induced response. These findings together suggest that the phosphorylation of CRMP2 play a critical role in mediating Sema3A-signalling.
Lysophosphatidic acid, but not Sema3A, induces phosphorylation of CRMP2 at Thr555, which is a single Rho/ROCK kinase phosphorylation site (Arimura et al. 2000). Our study demonstrates that Sema3A promotes phosphorylation of CRMP2 at Ser522, which is a Cdk5 phosphorylation site. We reported that both Rho/ROCK kinase and Cdk5 mediate ephrin-A5-induced signalling in retinal ganglion cells (Cheng et al. 2003). Recently, ephrin-A5 was found to induce phosphorylation of CRMP2 at Thr555 (Arimura et al., unpublished observation). Thus, CRMPs are phosphorylated through Rho/ROCK, Cdk5 and/or GSK3β at multiple sites, and probably with different efficacies, depending on the receptor–ligand interactions involved. Although the physiological significance of multiple phosphorylation of CRMPs by these kinases is unknown, in any case, to respond to axon guidance signals, neurones may require intracellular mechanisms to differentiate these signals. Our finding provides evidence that CRMP is a convergence point of different signalling pathways for axon guidance molecules.
Phosphorylation of CRMP2 regulates its affinity to tubulin
Actin cytoskeleton and microtubules both play critical roles for growth cone steering. It was demonstrated that Sema3A regulates actin dynamics via phosphorylation of cofilin by LIMK (Aizawa et al. 2001). Recent experiments provide evidence that signalling pathway can directly regulate microtubule dynamics (Gundersen & Cook 1999; Zhou & Cohan 2004). In fact, local and selective modification of dynamic microtubules can initiate and instruct directional growth cone steering (Mack et al. 2000; Buck & Zheng 2002; Gordon-Weeks 2004). CRMP2 binds to tubulin heterodimers and promotes microtubule assembly (Gu & Ihara 2000; Fukata et al. 2002). Sema3A facilitates Cdk5 activity, and activated Cdk5 phosphorylates tau, a microtubule associated protein (Sasaki et al. 2002). Our study indicates that phosphorylation of CRMP2 by Cdk5 and GSK3β reduced the association of CRMP2 with tubulin. This effect was not observed with CRMP2S522A, thereby indicating that CRMP2 at Ser522 was a critical phosphorylation site for regulation of its association with tubulin. We also found the reduced association of CRMP2S522A mutant with tubulin, when compared to the wild-type CRMP. This may suggest that mechanisms other than phosphorylation may influence the interaction of CRMP2 with tubulin. In cdk5–/– mouse brain, however, phosphorylation of CRMP2 was not detected with antibodies, pS522-CRMP1/2 and 3F4 and association of CRMP2 with tubulin was enhanced in this mutant. This argues for the idea that phosphorylation of CRMP2 at Ser522 mainly regulate association of CRMP2 with tubulin. Phosphorylation of CRMP2 by Rho/ROCK kinase also reduces association of CRMP2 with tubulin dimers (Arimura et al., unpublished observation). Together, these findings suggest that Sema3A signalling induces microtubule reorganization through the activities of tau and CRMP2, which regulate dynamics of microtubule and tubulin dimers, respectively. The phosphorylation of CRMP2 by Cdk5, GSK3β, and Rho/ROCK kinase may play a role in coordinating cytoskeleton activities in response to multiple axon guidance cues.
Phosphorylation of CRMP2 is involved in Sema3A-signalling
Previous independent reports using kinase inhibitors support the idea that Cdk5 and GSK3β are playing roles in mediating Sema3A signalling (Sasaki et al. 2002; Eickholt et al. 2002). In cdk5-deficient DRG, Sema3A-induced growth cone response is markedly attenuated (Sasaki et al. 2002). Sema3A activates GSK3β at the leading edge of neuronal growth cones, and different GSK3β antagonists can inhibit the growth cone collapse response induced by Sema3A (Eickholt et al. 2002). Our present study provides evidence that Cdk5-primed CRMP2 phosphorylation by GSK3β is involved in Sema3A-signalling. Our idea is consistent with the detection of phosphorylated CRMP2 in immunoblot analysis with anti-pThr-Pro antibody, and with 3F4 antibody in wild-type, but not in cdk5-deficient brain lysates (Fig. 1). Upon phosphorylation by Cdk5, a conformational change probably occurs in CRMP2, leading to CRMP2 phosphorylation by GSK3β. This may well explain the previous finding that activation of GSK3β alone cannot induce growth cone collapse (Eickholt et al. 2002). Generating knock-in mice of non-phosphorylated form of CRMPs will be necessary to elucidate the in vivo significance of CRMPs phosphorylation.
We found that CRMP1 was also a substrate for Cdk5 in vitro. Sema3A stimulation enhanced the level of immunofluorescence at the growth cone detected with anti-pS522-CRMP1/2 antibody, which recognizes both CRMP1 and CRMP2 phosphorylated at Ser522. In fact, introduction of a non-phosphorylated mutant of CRMP1 into DRG also suppressed Sema3A-induced growth cone collapse (unpublished observation). This suggests that phosphorylation of CRMP1 by Cdk5 is involved in Sema3A-induced response as well. Yeast two-hybrid and in vitro binding analysis suggest that the interactions among CRMP isoforms favor heterophilic oligomerization over homophilic oligomerization (Wang & Strittmatter 1997). CRMP1 and CRMP2 prefer hetero-oligomerization (CRMP1-CRMP2) to homo-oligomerization (CRMP1-CRMP1 or CRMP2-CRMP2). Therefore, CRMP1 and CRMP2 could function in concert with each other for physiological responses. This idea is consistent with our present observation that the Sema3A-induced response was attenuated by siRNAi knockdown of either CRMP1 or CRMP2 (Fig. 6B). Possible involvement of CRMP4 and CRMP5, other Cdk5 substrates, in Sema3A-signalling remains unclear at present.
Recently, two related papers have been published independently of our paper. Brown et al. (2004) showed that Cdk5 phosphorylates CRMP2 at Ser522, and non-phosphorylated mutant of CRMP2 suppresses Sema3A-induced growth cone collapse. Cole et al. (2004) showed that GSK3β phosphorylates CRMP2 at serine and threonine residues found as a hyperphosphorylated epitope first identified in plaques isolated from Alzheimers brain. All of these observations are consistent with our present study.
CRMP2 phosphorylation and its implication for AD
NFTs are a common feature of many neurodegenerative diseases, including AD. In prefrontal cortex of AD brain, increase in Cdk5 activity has been observed (Lee et al. 1999). In the mouse model, aberrant Cdk5 activation triggers pathological events such as accumulation of hyperphosphorylated tau, leading to neurodegeneration and NFTs (Cruz et al. 2003). Both tau and CRMP2 are dually phosphorylated by Cdk5 and GSK3β. The Cdk5-primed phosphorylation by GSK3β is also observed with tau (Arioka et al. 1995; Lee et al. 2001; Lucas et al. 2001; Noble et al. 2003). Tau and CRMP2 are associated with microtubules and promote their assembly (Weingarten et al. 1975; Fukata et al. 2002). The site-specific phosphorylation of tau and CRMP2 regulates their ability to bind and stabilize microtubules. Increased phosphorylation of tau and CRMP2 tends to negatively regulate their interactions with tubulin (Jameson et al. 1980; Lindwall & Cole 1984). High levels of phosphorylation of tau (Lee et al. 2001) and CRMP2 (Yoshida et al. 1998) are found in the NFTs. Sema3A promotes activation of Cdk5 in growth cone, followed by phosphorylation of tau (Sasaki et al. 2002) and CRMP2 (Fig. 5A).
Very early pathological alterations in the superficial layers of the entorhinal cortex have been documented in AD and other forms of dementia (Braak & Braak 1992). Interestingly, Sema3A mRNA expression is most prominent in stellate cells in layer preα of the entorhinal cortex in human CNS (Giger et al. 1998). The Sema3A-expressing stellate cells in layer preα appear to be particularly susceptible to degeneration in AD (Braak & Braak 1992). Our study demonstrates that Sema3A stimulation enhanced the levels of the phosphorylated form of CRMP2, which can be recognized by 3F4 (Yoshida et al. 1998; Gu et al. 2000). Sema3A also could induce neuronal cell death, and a specific anti-Sema3A antibody rescue retinal ganglion cells from cell death following optic nerve axotomy (Shirvan et al. 2002). In this context, it is worth noting a recent observation that accumulation of an internalized form of Sema3A is associated with degeneration of neurones in vulnerable fields of the hippocampus during AD (Good et al. 2004). A multiprotein complex containing phosphorylated MAP1B, CRMP2, PlexsA1 and A2 from the hippocampus of patients with AD has also been isolated and characterized. Although it is therefore tempting to speculate that semaphorin-related signalling may be involved in the process of the tangle formation in AD, a causal relationship remains to be determined.
In conclusion, Cdk5-primed GSK3β phosphorylation of CRMP mediates Sema3A signalling. Further work will be required to elucidate the in vivo significance of the phosphorylation of CRMP.
Chemicals and antibodies
Olomoucine and LiCl were purchased from Sigma, and TDZD-8 (4-Benzyl-2-methyl-1,2,4-thiadiazolidine-3,5-dione) was from Calbiochem. Other antibodies also used: anti-CRMP-62 (Goshima et al. 1995); Myc-tagged proteins, 9E10 (Sigma); V5-tagged proteins, anti-V5 (invitrogen); α-tubulin, DM1A (Sigma) anti-pThr-Pro antibody (Cell signal), Alexa 488 and 594-labelled goat anti-mouse and anti-rabbit antibodies were from Molecular Probes, Inc. An antiserum against the phosphorylated form of CRMP2 was raised by injection of asynthetic phosphopeptide (ASSAKTpSPAKQQAC: amino acids 516–528 plus Cys for conjugation) into rabbits. Polyclonal antibody was raised against IVAPPGGRANITSLG (amino acids 557–572) for CRMP2 (Ricard et al. 2001).
Mouse Cdk5 and His-tagged p35 cDNA were provided by Dr A. B. Kulkarni (National Institute of Dental and Craniofacial Research). Rat CRMP1, 2, 3 and mouse CRMP4 cDNA were kindly provided by Dr S. M. Strittmatter (Yale University). Rat CRMP5 was cloned using reverse transcript-polymerase chain reaction (RT-PCR). GSK3β cDNA was purchased from Open Biosystems clone.
cdk5 Mutant mice
All procedures were conducted in accordance with NIH guidelines concerning the Care and Use of Laboratory Animals and with the approval of the Animal Care Committee of the Yokohama City University. All animals were handled in accordance with institutional guidelines and housed in a pathogen-free environment on a 12 : 12 h light:dark cycle. cdk5 Mutant mice were generated as described and maintained in C57BL/6 J background as described (Ohshima et al. 1996).
Immunoblot analysis for mouse brain lysate
Brain samples and cultured cells were homogenated in RIPA buffer (50 mm Tris-HCl buffer, [pH 8.0], Nonidet P-40, 0.5% sodium deoxycholate, 0.1% sodium dodecyl sulphate (SDS), 10 µg/mL leupeptin, 10 µg/mL aprotinin, 1 mm phenylmethylsulphonyl fluoride) with phosphatase inhibitors (phosphatase inhibitor cocktail from Sigma and 5 µm okadaic acid). Immunoblot analysis was performed as described (Sasaki et al. 2002).
For 2-D electrophoresis, mouse embryonic brain at E16.5 from cdk5+/+ and cdk5–/– mice were homogenized in 2.7 m Urea, 0.5% Triton X-100, 0.6% Dithiothreitol (DTT), 0.5% IPG buffer (Ampholine pH 4–7, Amersham Bioscience, Piscataway, NJ, USA) with phosphatase inhibitors (50 mm NaF, 0.2 mm Na3VO4 and phosphatase inhibitor cocktail II from Calbiochem). Homogenates were subjected to 2-D Clean-up Kit (Amersham), after protein concentration was determined with DC Protein Assay Kit (Bio-Rad Laboratories, Hercules, CA, USA). First dimensional isoelectric focusing (IEF) was carried out on a non-linear 13-cm immobilized pH gradient (IPG) strips (pH 4–7) for 16 h using an IPGphor unit (Amersham). Each strip was rehydrated for 16 h with sample lysate (1 mg) in a final volume of 400 µL of IEF solution containing 10 µL bromophenol blue (0.25% w/v), 8 m Urea, 0.5% Triton X-100, 0.6% DTT, 0.5% IPG buffer (Ampholine [pH 4–7], Amersham). IEF was then carried out. Strips were subjected to a two-step equilibration (50 mm Tris [pH 6.8]; 2% SDS; 30% glycerol) in 0.5% DTT and 4.5% iodoacetamide (Sigma) buffers before proceeding to SDS-PAGE. Ten percent SDS-PAGE was used for the second dimension.
In-gel digestion for MALDI-TOF MS and mass analysis
After 2-D electrophoresis was terminated, gels were stained by SYPRO-Ruby (Molecular Probe). Spots corresponding to the immunoblot signals of the pThr-Pro antibody were excised from a gel in which brain lysate of cdk5+/+ embryos was applied. The gel pieces were washed with water for 10 min at 37 °C. Gels were then destained using 50 mm ammonium bicarbonate solution in 50% acetonitrile at 37 °C for 10 min, dehydrated with 50 mL of acetonitrile for 10 min at 37 °C and dried completely in a Speedvac evaporator. In-gel trypsin digestion was performed as described (Shevchenko et al. 1996). Tryptic peptides were loaded on to a Matrix Assisted Laser Desorption/Ionization (MALDI)-mass spectrometry (MS) (BIFLEX IIIU, Bruker Daltonics). Peptide mass calculation was done with BIOWORKS software and used for searching mouse proteins in the NCBInr database (National Center for Biotechnology Information, Bethesda, MD, USA) by Mascot search software (Matrix Science).
In vitro kinase assay for Cdk5
GST-fusion proteins (Cdk5, p25, CRMPwt, and mutants) were expressed in E. coli BL21 strain. Purified Cdk5 and p25 were pre-incubated together for 2 h at 20 °C. In some cases, purified GSK3β (Upstate) was used with or without Cdk5/p25. 5 µg of CRMP2wt or CRMP2S522A and preincubated Cdk5/p25 were mixed. Kinase reaction was initiated by addition of equal to volume of 2 × reaction buffer (100 mm HEPES [pH 7.2], 20 mm MgCl2, 2 mm DTT, 20 µm[γ-32P]ATP (2 µCi)). After incubation for 1 h at 30 °C, the proteins were resolved by SDS-PAGE and autoradiographed. In the experiment of Cdk5-primed phosphorylation by GSK3β, Cdk5 and p35 were expressed in HEK 293T cells and immunoprecipitated with anti-p35 antibody. Purified GST-CRMP2c and immobilized Cdk5/p35 bound to anti-p35 antibody and protein-A beads were mixed with non-labelled ATP (2 mm) for the first kinase assay. After first assay, the samples were spun down to remove Cdk5/p35. For the second kinase assay, the supernatants were mixed with purified GSK3β (upstate) and [γ-32P]ATP (10 µm), and were incubated for 30 min at 30°C.
Metabolic labelling of CRMPs in HEK293T cells with [32P]phosphate
HEK293T cells were seeded at 1.5 × 106 cells per 6-well plate. Two days after, the cells were transfected with expression vector using Fugene6 transfection reagent (Roche). After 20 h, the transfected cells were serum-starved for 4 h. For last 2 h, the cells were labelled with 1.5MBq/mL [32P]orthophosphoric acid. After labelling, the cells were lysed in IP-buffer (20 mm Tris-HCl [pH 7.4], 150 mm NaCl, 1%NP-40, 1 mm EDTA, 50 mm NaF, 1 mm Na3VO4, 50 µmρ-amidinophenylmethanesulphonyl fluoride [ρ-APMSF], 0.1 units/mL aprotinin, 50 µm leupeptin). The lysates were incubated with anti-Myc antibody for 2 h, followed by further incubation with protein G beads. After washing three times with IP-buffer, the samples were analyzed by immunoblot and autoradiography.
Cell culture and immunoprecipitation
HEK293T cells were seeded at 3 × 105 cells per 6 cm dish. The next day, the cells were transfected with 1 µg of expression vectors for indicated receptors and intracellular molecules. After 1–2 days of incubation, cells were lysed in NP-40 buffer (20 mm Tris-HCl [pH 7.4], 150 mm NaCl, 1% NP-40, 1 mm EDTA, 50 mm NaF, 20 mm sodium pyrophosphate, 1 mm Na3VO4, 50 µmρ-amidinophenylmethanesulphonyl fluoride [ρ-APMSF], 0.1 unit/mL aprotinin, and 50 µm leupeptin). Cortical neuronal culture was performed as described (Sasaki et al. 2002). Cells from E15.5 C57BL/6 mouse embryos 3DIV were lysed in NP-40 buffer for immunoblot analysis.
Chick E7 DRG explants were exposed to 1 nm Sema3A for the indicated time, and then fixed by adding equal volume of 8% (v/v) paraformaldehyde in phosphate-buffered saline (PBS) for 1 h. After permeablization with 0.3% (v/v) Triton X-100 in PBS for 2 min, they were blocked with 1% (w/v) bovine serum albumin and 0.1% Triton X-100 in PBS for 1 h. Explants were then incubated with anti-CRMP-1 (1 : 100), anti-CRMP2 (1 : 100) or anti-pS522-CRMP1/2 (1 : 500) antibody in the above blocking solution for 12 h at 4 °C, followed by incubation with Alexa 594-labelled goat anti-rabbit antibody (1/1000) for 1 h. Finally, they were analyzed using OLYMPUS IX71 microscopy.
Recombinant HSV preparations, infection and growth cone collapse assay
Recombinant HSV preparations and infections of chick E7 DRG explants were performed as previously described (Sasaki et al. 2002). In brief, recombinant viruses possessing wild-type and mutant crmp2 genes were added to explants at a concentration of about 106 PFU/mL during 20 h before the collapse assay. The percentage of axons expressing Myc-tagged CRMP2 ranged 60–80% in infected cultures. Growth cone collapse assays using chick DRGs were performed with purified recombinant chick Sema3A (collapsin-His6) as previously described (Goshima et al. 1995).
siRNA preparation and transfection
A 21-ologonuculeotide siRNA duplex was synthesized by Japan Bio service to target rat CRMP1 sequence 5′-GCAGCAGACACCAAAUCCUTT-3′ and control sequence 5′-GUGGGAGCGCGUGAUGAACTT-3′. For rat CRMP2 target sequence 5′-GGGUAAACUCCUUCCUCGUGUACAU, we used stealth RNA which is chemically modified siRNA by invitrogen. A scramble sequence 5′-GGGAACUCCUUCCUCGUGUAUACAU was used as a negative control. Transfection of siRNA in primary culture of rat embryonic DRG (E14) was performed with Lipofectamin 2000 (invitrogen) and CombiMag (OZ BIOSCIENS), according to the manufacturer's instructions. Four h after plating, dissociated neurones were co-transfected with siRNA and a plasmid encoding the enhanced green fluorescent protein (EGFP). After overnight culture, growth cone collapse assay was performed in CRMPs knockdown neurones expressing co-transfected EGFP.
The authors thank Dr A.B. Kulkarni for cdk5–/– mice. The authors also thank RRC stuff in RIKEN BSI for MS analysis and the laboratory members for technical assistance. We also thank Y. Sugiyama in YCU for preparing Sema3A and cell culture, and the other laboratory members for instructive inputs. Finally, the authors thank Dr S.M. Strittmatter (Yale University), Dr K. Kaibuchi (Nagoya University), Dr S. Ohno (Yokohama City University) and Dr T. Hirata (National Institute of Genetics) for useful discussion. This work was supported by Grants-in-aid from the Ministry of Education, Culture, Sport, Science and Technology, Japan (to T.O. and Y.S.), CREST (Core Research for Evolutional Science and Technology) of JST (Japan Science and Technology Corporation) (to Y.G. and K.T.), Uehara Memorial Foundation (to Y.G.), Yokohama Medical Foundation (to Y.U. and Y.G.) and The Yokohama City University Center of Excellence Program of the Ministry of Education, Culture, Sports, Science and Technology of Japan.