Research ArticleNeuroscience

Thioredoxin Mediates Oxidation-Dependent Phosphorylation of CRMP2 and Growth Cone Collapse

See allHide authors and affiliations

Science Signaling  26 Apr 2011:
Vol. 4, Issue 170, pp. ra26
DOI: 10.1126/scisignal.2001127


Semaphorin3A (Sema3A) is a repulsive guidance molecule for axons, which acts by inducing growth cone collapse through phosphorylation of CRMP2 (collapsin response mediator protein 2). Here, we show a role for CRMP2 oxidation and thioredoxin (TRX) in the regulation of CRMP2 phosphorylation and growth cone collapse. Sema3A stimulation generated hydrogen peroxide (H2O2) through MICAL (molecule interacting with CasL) and oxidized CRMP2, enabling it to form a disulfide-linked homodimer through cysteine-504. Oxidized CRMP2 then formed a transient disulfide-linked complex with TRX, which stimulated CRMP2 phosphorylation by glycogen synthase kinase–3, leading to growth cone collapse. We also reconstituted oxidation-dependent phosphorylation of CRMP2 in vitro, using a limited set of purified proteins. Our results not only clarify the importance of H2O2 and CRMP2 oxidation in Sema3A-induced growth cone collapse but also indicate an unappreciated role for TRX in linking CRMP2 oxidation to phosphorylation.


Proper development of the nervous system depends on various guidance molecules that steer axons by regulating cytoskeletal dynamics in growth cones (14). Semaphorin3A (Sema3A), which induces growth cone collapse to repel axons (24), binds to and activates the receptors neuropilin-1 (NP-1) and plexin-A (PlexA) to regulate the cytoskeleton by means of collapsin response mediator protein 2 (CRMP2) (57). After Sema3A stimulation, cyclin-dependent kinase 5 (CDK5) phosphorylates CRMP2 at Ser522, which acts as a priming site for glycogen synthase kinase–3 (GSK-3)–dependent CRMP2 phosphorylation at Thr509, Thr514, and Ser518 (68). CRMP2 associates with tubulin heterodimers and promotes microtubule polymerization (9); these Sema3A-dependent phosphorylations lead to microtubule disassembly and thereby growth cone collapse. Thus, CRMP2 phosphorylation has been regarded as crucial to Sema3A signaling, but the precise molecular mechanism of how the activities of CDK5 and GSK-3 are regulated remains largely obscure.

Molecule interacting with CasL (MICAL) is another essential mediator of Sema-dependent axon guidance (10, 11). MICAL has a flavoprotein monooxygenase domain that is required for its role in axon steering, suggesting that redox signaling may play a crucial role in Sema3A signaling. Indeed, mammalian MICAL generates H2O2 (11, 12), a reactive oxygen species (ROS), and the flavin monooxygenase inhibitor (−)-epigallocatechin gallate blocks Sema3A-induced axonal repulsion (10, 13); however, there has been no definitive evidence that MICAL produces H2O2 in vivo after Sema3A stimulation. Drosophila MICAL acts through its flavoprotein monooxygenase domain to directly regulate actin reorganization (14), but it is unclear whether this occurs through production of H2O2 or whether there is any functional relationship between MICAL and CRMP2.

Whereas excess ROS cause oxidative stress, ultimately leading to cell death, moderate amounts of ROS mediate physiological phenomena such as cell proliferation and motility (1517). In particular, accumulating evidence indicates that H2O2 functions as a second messenger by oxidizing cysteine residues to form disulfide bonds in various target proteins and thereby regulate their function. Such disulfide bonds are generally reduced to thiols by thioredoxin (TRX), a thiol oxidoreductase conserved in both prokaryotes and eukaryotes (1719). TRX reduces disulfide bonds through thiol-disulfide exchange between the target protein and the CXXC (in which C is cysteine and X is any amino acid) residues at the TRX active site. Oxidized TRX is then reduced and regenerated by thioredoxin reductase (TxR), which is itself reduced through oxidation of NADPH (reduced form of nicotinamide adenine dinucleotide phosphate). Given that H2O2 mediates cell signals through cysteine oxidation, it is likely that proteins involved in ROS signaling are substrates of TRX. Therefore, the identification of TRX substrates should provide important clues to the identities of key unknown players in ROS signaling.

Here, we used a substrate-trapping method to perform an in vivo screen for candidate TRX substrates and identified CRMP2. We showed that Sema3A stimulated the generation of H2O2 by MICAL, leading to CRMP2 oxidation and thereby enabling it to form a disulfide-linked homodimer through Cys504. Unexpectedly, CRMP2 oxidation also promoted a disulfide-linked interaction with TRX, and thereby GSK-3–dependent phosphorylation, resulting in growth cone collapse.


Identification of CRMP2 as a candidate substrate for TRX

Two cysteine residues (Cys32 and Cys35) conserved in TRX are directly involved in its oxidoreductase reactions (1820). The C35S mutant form of TRX, in which Cys35 is substituted with Ser, can form mixed disulfide-linked complexes with TRX substrate proteins (20), which can then be released by reduction with dithiothreitol (DTT), leading to its use in substrate-trap experiments (Fig. 1A). To search for TRX substrate proteins under physiological conditions, we generated NIH 3T3 cell lines stably expressing FLAG–TRX C35S or FLAG–TRX C32/35S, in which both Cys32 and Cys35 are substituted with Ser. The latter mutant does not form such complexes, and thus, it can be used as a negative control. Cell lysates were immunoprecipitated with anti-FLAG antibodies, and candidate proteins were eluted with DTT. Samples of the proteins thus obtained were resolved by SDS–polyacrylamide gel electrophoresis (SDS-PAGE) and visualized by silver staining (Fig. 1B). Mass spectrometry identified an abundant protein in the 65-kD region as CRMP2, which acts as a necessary mediator of Sema3A-induced growth cone collapse (5) and plays a role in establishing neuronal polarity (8, 21). We also identified several well-characterized TRX substrates, such as peroxiredoxin I and II (22), supporting the validity of this experimental approach.

Fig. 1

Identification of CRMP2 as a candidate substrate for TRX. (A) Schematic model of the capture and release of target proteins by TRX C35S. (B) Lysates of NIH 3T3 cells stably expressing the FLAG-TRX mutants were immunoprecipitated with anti-FLAG beads. The beads were treated with DTT, and proteins eluted by DTT (E) or remaining on the beads (R) were subjected to SDS-PAGE and silver staining. (C, E, and G) Lysates from COS-7 cells transfected with the indicated constructs were immunoprecipitated (IP) with anti-FLAG antibodies and analyzed by immunoblotting. WT, wild type. (D) COS-7 cells were transfected with the indicated constructs and treated with 100 μM H2O2 for 5 min. Cell lysates were treated with IAA and immunoblotted under nonreducing or reducing conditions. Asterisk (*) indicates TRX C35S that formed a complex with an unidentified protein. (F) Lysates of N1E-115 neuroblastoma cells treated with 500 μM H2O2 for 5 min were treated with IAA, acetone-precipitated, and dissolved. They were then immunoprecipitated with anti-CRMP2 antibodies and immunoblotted with the indicated antibodies. IgG, immunoglobulin G. (H) GST-CRMP2 pull-down assays of recombinant His-TRX in the presence of 5 mM DTT (reducing) or 100 μM H2O2 (oxidizing). Proteins were subjected to SDS-PAGE and Coomassie Brilliant Blue (CBB) staining.

To confirm the interaction between TRX and CRMP2, we performed coexpression and coimmunoprecipitation analyses. Myc-CRMP2 bound to FLAG–TRX C35S but not to FLAG–TRX WT (wild type) or C32/35S (Fig. 1C), indicating the requirement of Cys32 for CRMP2 binding. Next, we treated the lysates with iodoacetamide (IAA) to block further oxidation of cysteine residues and prevent nonspecific interactions between thiols after cell harvest. Nonreducing SDS-PAGE of the IAA-treated lysates revealed a 75-kD signal, which matches well with the molecular mass of the TRX-CRMP2 disulfide complex (Fig. 1D). When cells were treated with H2O2 to induce formation of disulfide bonds, TRX WT also formed a complex with CRMP2 (Fig. 1D).

Proteins with a TRX-related sequence containing the catalytic cysteine residues constitute a large family (18, 19, 23) that includes nucleoredoxin (NRX) (24), TRX-related protein 32-kD (TRP32) (25), and TRX-related protein 14-kD (TRP14) (26), all of which exist in the cytoplasm. To investigate the specificity of the CRMP2-TRX interaction, we generated mutants corresponding to TRX C35S in NRX, TRP32, and TRP14 and tested their ability to bind to CRMP2. Only FLAG–TRX C35S associated with Myc-CRMP2 (Fig. 1E), indicating that, of the TRX family proteins examined, the interaction is specific to TRX. Next, we used anti-CRMP2 antibodies to perform immunoprecipitation assays on N1E-115 neuroblastoma cell lysates either treated with H2O2 or left untreated to determine whether endogenous TRX interacts with CRMP2. Cell lysates were treated with IAA as above and then precipitated with acetone to denature proteins and dissociate noncovalent bonds (27). TRX coimmunoprecipitated with CRMP2 when cells were treated with H2O2 but not when they were left untreated (Fig. 1F). Because CRMP2 is predicted to interact with TRX Cys32 through a disulfide bond, we tried to identify which cysteine residues of CRMP2 were responsible for the interaction by substituting the cysteine residues conserved between human and mouse with serines (C132S, C133S, C248S, C323S, C334S, C439S, and C504S). Of these mutants, only Myc-CRMP2 C504S failed to associate with FLAG–TRX C35S (Fig. 1G), suggesting that Cys504 is crucial for CRMP2’s interaction with TRX. To further examine the cysteine-dependent interaction, we used recombinant TRX and CRMP2 proteins to conduct glutathione S-transferase (GST) pull-down assays. His–TRX C35S was specifically pulled down by CRMP2-GST, and the interaction was augmented by H2O2 (Fig. 1H). Moreover, His–TRX C35S did not associate with CRMP2-GST C504S, irrespective of its redox state. Collectively, these results support the notion that TRX Cys32 links directly to CRMP2 Cys504 through a disulfide bond. A previous study indicated that S-nitrosylation occurs at CRMP2 Cys504, further supporting the notion that Cys504 is subject to redox reactions (28, 29).

Crucial importance of the TRX-CRMP2 interaction in Sema3A signaling

CRMP2 is necessary for the Sema3A-induced growth cone collapse of dorsal root ganglion (DRG) neurons (57); thus, we next investigated the role of TRX in Sema3A signaling. First, we used immunofluorescence microscopy to confirm the presence of endogenous TRX in growth cones (Fig. 2A). Next, we expressed Myc-CRMP2 C504S, which does not bind TRX, in neurons, as well as a Myc-tagged form of the phosphorylation-inactive CRMP2 T509A/S522A mutant, which acts in a dominant-negative manner (7), which we used as a positive control for the inhibition of Sema3A-induced growth cone collapse. Both Myc–CRMP2 C504S and Myc–CRMP2 T509A/S522A were present in growth cones (fig. S1), and, like Myc–CRMP2 T509A/S522A, Myc–CRMP2 C504S suppressed Sema3A-induced growth cone collapse (Fig. 2B). These results suggest that Cys504, which is crucial for CRMP2 interaction with TRX, is also important for Sema3A signaling. To directly examine the role of TRX in Sema3A signaling, we investigated the effects of overexpressing wild-type or mutant forms of FLAG-TRX on Sema3A-dependent growth cone collapse. The TRX C32/35S mutant, which competes for TxR, inhibits the function of endogenous TRX (30). We found that Sema3A-dependent growth cone collapse was repressed in DRG neurons expressing FLAG–TRX C32/35S, unlike those expressing FLAG–TRX WT (Fig. 2C), suggesting the importance of TRX in mediating Sema3A signaling. In contrast, FLAG–TRX C35S, which, like wild-type TRX binds CRMP2, had no effect. To confirm these findings, we performed knockdown experiments with three different small interfering RNAs (siRNAs) directed against TRX. Western blot analyses indicated that these siRNAs all partially suppressed the expression of TRX (Fig. 2D). The growth cone collapse assays revealed the importance of endogenous TRX in Sema3A signaling. In contrast, there was no effect on serum-induced phosphorylation or dephosphorylation of various proteins (fig. S2), indicating that the effects of TRX knockdown on Sema3A-induced growth cone collapse are specific. We performed rescue experiments to exclude the possibility of off-target effects of the siRNAs, using the TRX Rsc (rescue) mutant, which has three silent mutations in the target sequence and is thus resistant to TRX siRNA. As shown in Fig. 2E, expression of FLAG–TRX Rsc restored the normal collapse response.

Fig. 2

Crucial importance of TRX-CRMP2 interaction in Sema3A signaling. (A) DRG neurons were stained with anti-TRX antibody (green) and phalloidin (red) to visualize cell morphology. (B to E) DRG neurons were transfected with the indicated constructs or siRNAs together with a GFP-expressing plasmid. Sema3A-treated (30 min) cells were fixed and stained with phalloidin. GFP-positive cells were examined. Percent collapsed is given as the mean ± SEM (n = 3 to 4 experiments). *P < 0.05, significant difference from the control. Typical images of growth cones (B) and knockdown efficiencies by TRX siRNAs with quantification (D and E) are also shown. (F) Time-lapse phase-contrast images of chicken DRG growth cones exposed to Sema3A gradients (arrows) in the absence (control) or presence of 10 μM DNCB. Digits represent minutes after the onset of Sema3A application. The graph shows turning angle of growth cones (mean ± SEM, n = 12 to 15 experiments), with positive and negative values indicating attraction and repulsion, respectively. (G) Chicken growth cones transfected with the indicated constructs were exposed to Sema3A gradients. Turning angles of growth cones are shown (mean ± SEM, n = 12 to 15 neurons). (F and G) *P < 0.05, significant difference from control growth cones.

To further investigate the role of the TRX-CRMP2 interaction in Sema3A signaling, we examined the turning behavior of growth cones in response to a Sema3A gradient. Directional application of Sema3A causes repulsive growth cone turning in culture (24, 31). Bath application of the TxR inhibitor 2,4-dinitro-1-chlorobenzene (DNCB), which forces TRX to take an inactive form by inhibiting TRX regeneration by TxR (32), suppressed Sema3A-induced repulsive turning of growth cones in DRG neurons (Fig. 2F). Moreover, expression of Myc–CRMP2 C504S also inhibited growth cone repulsion, whereas Myc–CRMP2 WT had no significant effect (Fig. 2G).

To confirm the in vivo importance of the TRX-CRMP2 interaction, we performed in utero electroporation analyses. Chen et al. reported that Sema3A signaling is important for the radial migration of cortical neurons during development (33); therefore, we examined the radial migration of cortical neurons after transfection of Myc–CRMP2 WT or C504S–expressing plasmids into E14.5 (embryonic day 14.5) mouse brains perturbed the migration of cortical neurons, whereas Myc–CRMP2 WT showed no significant effect (fig. S3, A and B). The migration defects produced by expression of Myc–CRMP2 C504S resemble those seen in NP-1–conditional knockout mice or various Plexin (PlexA2, A4, or D1) knockdown cortical neurons (33). These results support an important role for the TRX-CRMP2 interaction in Sema3A signaling, consistent with the in vitro results obtained from the growth cone collapse assays.

Deoxidization of a disulfide-linked CRMP2 homodimer by TRX

Immunoblot analysis of cell lysates treated with IAA, subjected to nonreducing SDS-PAGE, and probed with anti-FLAG antibody revealed that, in addition to 65-kD signals corresponding to monomeric FLAG-CRMP2, H2O2 treatment led to the appearance of signals at around 135 kD (Fig. 3A). Unlike the 65-kD signals, the latter were not apparent when SDS-PAGE was conducted under conventional reducing conditions. Thus, these 135-kD signals appear to reflect a disulfide-linked oligomeric form of FLAG-CRMP2. Oligomer formation seems to occur inside cells, because IAA treatment of cell lysates blocked in vitro formation of FLAG-CRMP2 oligomer in response to H2O2, excluding the possibility that these oligomers formed after harvesting the cells (fig. S4). H2O2-induced CRMP2 oligomer formation depended on Cys504, because analyses of cells transfected with FLAG–CRMP2 C504S failed to show a 135-kD signal (Fig. 3A). To investigate the involvement of other cysteine residues in CRMP2 oligomer formation, we coexpressed Myc–CRMP2 WT and FLAG–CRMP2 C504S and found that Myc–CRMP2 WT did not form disulfide-linked oligomers with FLAG–CRMP2 C504S (Fig. 3B), indicating that only Cys504 is involved in disulfide bond formation. We found that, like ectopically expressed FLAG-CRMP2, endogenous CRMP2 in DRG neurons also formed a homodimer in response to H2O2 (Fig. 3C).

Fig. 3

TRX chemically reduces disulfide-linked CRMP2 homodimer induced by Sema3A. (A, B, and E) COS-7 cells were transfected with the indicated constructs and then treated with 100 μM H2O2 for 5 min for (A) and (B) or 30 min for (E). Cell lysates were treated with IAA and immunoblotted under nonreducing or reducing conditions. (C) DRG neurons were treated with 100 μM H2O2 for indicated times and analyzed with anti-CRMP2 immunoblotting. (D) Lysates of H2O2-treated COS-7 cells expressing FLAG-CRMP2 were immunoprecipitated with anti-FLAG antibody. The immunoprecipitates were incubated with His-TRX or DTT and then analyzed with anti-FLAG immunoblotting. (F) DRG neurons transfected with TRX siRNA or treated with DNCB (100 μM for 60 min) or H2O2 (100 μM for 5 min) were analyzed with anti-CRMP2 immunoblotting. Graph shows the relative amount of dimer/total CRMP2 (mean ± SEM, n = 3 experiments). *P < 0.01, significant difference. (G) DRG neurons were transfected with GFP-HyPer plasmid and then stimulated with Sema3A or 100 μM H2O2. Pseudo-colored images and calibration bars (arbitrary units) are indicated. Relative intensity in growth cones was analyzed. Data are means ± SEM. For each construct, five to eight neurons were measured. *P < 0.05, significant difference from the control. (H) DRG neurons were stimulated with Sema3A-containing conditioned medium for the indicated times and analyzed with anti-CRMP2 antibody.

Next, we analyzed the effects of TRX on the CRMP2 disulfide bond. We first performed an in vitro assay of CRMP2 reduction. COS-7 cells expressing FLAG-CRMP2 were stimulated with H2O2, and then FLAG-CRMP2 was purified with anti-FLAG beads and incubated with His-TRX. Loss of the dimer revealed that, in the presence of His–TRX WT, FLAG-CRMP2 was reduced, unlike FLAG-CRMP2 mixed with His–TRX C32/35S (Fig. 3D), indicating that TRX can reduce the disulfide bonds in vitro. Moreover, coexpression of Myc–TRX WT or C35S decreased the abundance of the H2O2-induced FLAG-CRMP2 homodimer in cells (Fig. 3E). We also investigated the effects of TRX knockdown on CRMP2 oxidation and found that TRX siRNA augmented CRMP2 oxidation (Fig. 3F), thus confirming that TRX is involved in reduction of CRMP2 in vivo. Treatment with DNCB also resulted in CRMP2 oxidation.

MICAL, which is involved in Sema signaling in both Drosophila and mammalian cells (10, 11), can generate H2O2 (11, 12). Therefore, we used the recently developed H2O2-specific probe, green fluorescent protein (GFP)–HyPer (34), to determine whether Sema3A treatment of DRG neurons leads to production of H2O2. H2O2 treatment of HyPer-transfected DRG neurons resulted in augmentation of fluorescence in neurites (Fig. 3G), and, consistent with localized H2O2 production, Sema3A treatment increased fluorescence at the tips of neurites. Moreover, Sema3A treatment promoted formation of the CRMP2 homodimer (Fig. 3H), indicating that it elicited CRMP2 oxidation.

Stimulation of H2O2 generation through MICAL

We next examined the possibility that MICAL mediates Sema3A-induced H2O2 generation. There are three MICAL isoforms (MICAL1, 2, and 3) in mammals (10, 13). Reverse transcription–polymerase chain reaction (RT-PCR) analyses indicated that MICAL1 and 3 are abundant in DRG neurons but we could detect only slight expression of MICAL2 (Fig. 4A and fig. S5). We transfected DRG neurons with siRNAs directed against MICAL1 and 3, and confirmed the decreased abundance of their cognate mRNAs (Fig. 4A). We found that, unlike control cells, HyPer fluorescence did not increase in response to Sema3A in MICAL knockdown neurons (Fig. 4B), indicating that endogenous MICAL mediates Sema3A-dependent H2O2 generation. To exclude the possibility of off-target effects, we performed rescue experiments using a mouse MICAL1 construct that is resistant to siRNAs directed against rat MICAL1 and 3 (Fig. 4, B and E). DRG neurons transfected with FLAG-mMICAL1 together with siRNAs against MICAL showed normal H2O2 production in response to Sema3A. In contrast, mMICAL1 GW, which is catalytically inactive (10), failed to restore H2O2 production (fig. S6). These results convincingly indicate that Sema3A stimulates H2O2 generation through MICAL. Next, we examined the effect of MICAL knockdown on the oxidation state of CRMP2 and found that, as predicted, CRMP2 in MICAL knockdown neurons did not form the disulfide-linked homodimer in response to Sema3A stimulation (Fig. 4C).

Fig. 4

MICAL regulates Sema3A-induced H2O2 generation and the ensuing CRMP2 oxidation. (A) DRG neurons were transfected with the indicated siRNAs and then subjected to RT-PCR analyses for each MICAL isoform. (B) DRG neurons were transfected with the indicated siRNAs and constructs, together with the GFP-HyPer plasmid, and then stimulated with Sema3A. Representative fluorescent images are pseudo-colored and shown with a calibration bar (left panel; a.u., arbitrary unit). Relative intensity of GFP fluorescence in the growth cones is presented as mean ± SEM (right panel, six neurons for each construct). *P < 0.05, significant difference against control cells. (C) DRG neurons transfected with siRNAs for both MICAL1 and MICAL3 were stimulated with Sema3A for 15 min. Lysates were treated with IAA and analyzed by immunoblotting with anti-CRMP2 antibody under nonreducing and reducing conditions. (D and E) DRG neurons were transfected with the indicated constructs or siRNAs with the GFP-expressing plasmid. After treatment with Sema3A for 30 min, the cells were fixed and stained with phalloidin to visualize growth cones. Data are means ± SEM (n = 3 experiments). For each experiment, more than 50 GFP-positive neurons were measured. *P < 0.05, significant difference against control siRNA–transfected neurons.

Transfection with siRNA directed against either MICAL1 or MICAL3 led to moderate reduction in the fraction of neurons showing growth cone collapse (Fig. 4D). Reduction of growth cone collapse was also observed in neurons transfected with both MICAL1 and MICAL3 siRNAs (Fig. 4D). Again, expression of FLAG-mMICAL1 rescued the defect in collapse response (Fig. 4E). These results implicate H2O2 production by MICAL in mammalian Sema3A signaling.

Oxidation-dependent CRMP2 phosphorylation mediated by TRX

Next, we investigated the role of TRX in CRMP2 phosphorylation as well as the functional implications of CRMP2 phosphorylation. When cells expressing FLAG-CRMP2 were treated with H2O2, a fraction of the FLAG-CRMP2 band shifted upward (Fig. 5A). A similar CRMP2 mobility shift occurs during Sema3A signaling as a result of GSK-3–dependent phosphorylation, which has been implicated in Sema3A-induced growth cone collapse (68). Therefore, we examined whether H2O2-induced CRMP2 phosphorylation occurs at the same site phosphorylated by GSK-3. Western blot analyses with an antibody that specifically recognizes GSK-3–phosphorylated CRMP2 (p-CRMP2) (8) indicated that H2O2-induced CRMP2 phosphorylation occurs at the GSK-3 site (Fig. 5A). Similarly, phosphorylation of endogenous CRMP2 at the GSK-3 site was observed when DRG neurons were treated with H2O2 (Fig. 5B).

Fig. 5

TRX mediates oxidation-dependent CRMP2 phosphorylation. (A) COS-7 cells were transfected with the indicated constructs and then treated with 100 μM H2O2 for 15 min. Cell lysates were subjected to immunoblotting analyses. Graph shows normalized band intensities (p-CRMP2/CRMP2) (mean ± SEM, n = 3 experiments). *P < 0.01, significant difference. (B) DRG neurons were treated with 100 μM H2O2 for indicated times or stimulated with Sema3A, and the cell lysates were subjected to immunoblotting. (C and D) DRG neurons were transfected with siRNAs or constructs and then stimulated with 100 μM H2O2 or Sema3A; the cell lysates were analyzed with the indicated antibodies. (B to D) Quantitative measurements of band intensities (p-CRMP2/CRMP2) with normalization are shown. (E) CRMP2-GST pull-down assays of mouse brain lysates or purified tubulin. Proteins were subjected to immunoblotting and CBB staining. Asterisk (*) indicates the nonspecific signal derived from CRMP2-GST, which reacted with anti-Slp1 antibody.

Next, we assessed the role of TRX in H2O2-induced CRMP2 phosphorylation. Unexpectedly, coexpression of Myc–TRX WT or C35S enhanced CRMP2 phosphorylation, whereas Myc–TRX C32/35S repressed it (Fig. 5A). Moreover, Myc-TRX was less effective at increasing phosphorylation of FLAG–CRMP2 C504S, although some enhancement was observed. These results suggest that TRX somehow promotes CRMP2 phosphorylation. To confirm this possibility, we performed RNA interference knockdown of TRX, which decreased CRMP2 phosphorylation (Fig. 5C). We also found that Myc–CRMP2 C504S expressed in neurons showed less phosphorylation after Sema3 treatment than did Myc–CRMP2 WT (Fig. 5D).

Having confirmed a role for TRX in CRMP2 phosphorylation, we conducted GST pull-down assays to investigate the functional effects of CRMP2 phosphorylation. Mouse brain lysates or purified tubulin proteins were subjected to pull-down assays with CRMP2–GST WT or T514D, which mimics GSK-3β–phosphorylated CRMP2 (8), and the precipitates examined by immunoblotting analyses for known CRMP2-binding proteins, including tubulin, Numb, kinesin light chain 1 (KLC1), actin, dynein heavy chain (DHC), and Synaptotagmin-like protein 1 (Slp1) (9, 3539). We found that the T514D mutation weakened CRMP2 binding to tubulin and Numb, but not actin, as reported previously (8, 40) (Fig. 5E). The mutation also decreased the association of KLC1 and Slp1 with CRMP2-GST, whereas DHC binding was unaffected. CRMP2 regulates not only tubulin assembly by directly interacting with tubulin, but also cargo transport on microtubules through its interactions with Numb, KLC1, and Slp1 (9, 35, 36, 38); thus, these results suggest that CRMP2 phosphorylation may affect microtubule-dependent function by disrupting tubulin assembly and cargo transport.

The molecular mechanisms of TRX-mediated CRMP2 phosphorylation

These results suggest that CRMP2 phosphorylation may be linked to formation of a complex between TRX and CRMP2. Indeed, immunoblotting analyses after nonreducing SDS-PAGE revealed that CRMP2 bound to TRX showed increased phosphorylation compared to the monomer or disulfide-linked homodimer (Fig. 6A). This provides an explanation for the lack of a dominant-negative effect of FLAG–TRX C35S on growth cone collapse (Fig. 2C).

Fig. 6

Molecular mechanisms of TRX-mediated CRMP2 phosphorylation. (A) COS-7 cells transfected with the indicated constructs were treated with 100 μM H2O2 for 15 min and subjected to immunoblotting under nonreducing or reducing conditions. Quantitative measurements of band intensities (p-CRMP2/CRMP2 of the monomer, dimer, complex with TRX, and total) with normalization are also shown. Data are means ± SEM (n = 3 experiments). *P < 0.01. (B and C) In vitro kinase assays were performed with indicated recombinant proteins in the presence of DTT (5 mM) or H2O2 (100 μM). Proteins were analyzed with indicated antibodies. (A to C) Quantitative measurements of band intensities (p-CRMP2/CRMP2) with normalization are shown. (D) Model of TRX-mediated CRMP2 phosphorylation. After Sema3A stimulation, MICAL generates H2O2 at growth cones and oxidizes CRMP2 to form a disulfide-linked homodimer through Cys504. In turn, oxidized CRMP2 forms a disulfide complex with TRX (Capturing by TRX) and becomes preferentially phosphorylated by GSK-3, resulting in growth cone collapse.

Finally, we tried to replicate the TRX-mediated increase in CRMP2 phosphorylation with purified recombinant proteins. CRMP2 was weakly phosphorylated in the presence of CDK5, which acts as a priming kinase for GSK-3–dependent CRMP2 phosphorylation (68), and GSK-3β (Fig. 6B). Addition of His-TRX alone failed to further increase CRMP2 phosphorylation not only in reducing (DTT) conditions but also in oxidizing (H2O2) conditions. We speculated that, because His-TRX should be directly oxidized by H2O2 or rapidly released from CRMP2 through formation of an intramolecular disulfide bond, only a small amount of His-TRX was available to interact with CRMP2. Therefore, we also added TxR and NADPH to regenerate the reduced form of His-TRX, and found that this treatment increased CRMP2 phosphorylation when the reaction was performed in the presence of H2O2 (Fig. 6B). His–TRX C32/35S did not promote CRMP2 phosphorylation, and, regardless of the conditions, wild-type His-TRX failed to stimulate the phosphorylation of CRMP2 C504S (Fig. 6C). Kinase assays with 32P-labeled adenosine 5′-triphosphate (ATP) revealed that phosphorylation of CRMP2 by GSK-3β was augmented by TRX in the presence of the TRX-regeneration system and H2O2, whereas phosphorylation by CDK5 was not (fig. S7). These results indicate that the disulfide-linked interaction between TRX and CRMP2 stimulates CRMP2 phosphorylation by GSK-3β.


Here, we identified a mechanism of H2O2-mediated signal transduction in Sema3A signaling. After Sema3A stimulation, MICAL generates H2O2 at growth cones and oxidizes CRMP2, so that CRMP2 forms a disulfide-linked homodimer through Cys504 (Fig. 6D). Oxidized CRMP2 is then reduced by TRX, and during this catalytic reaction, TRX forms a disulfide-linked intermediate with one molecule of CRMP2. Formation of the complex between CRMP2 and TRX promotes CRMP2 phosphorylation by GSK-3 and thereby growth cone collapse.

Previous studies have elucidated key players in Sema3A-induced CRMP2 phosphorylation. For instance, the tyrosine kinase Fyn associates with PlexA and phosphorylates CDK5 at Tyr15 (41), increasing CDK5’s catalytic activity. However, the specific pathways whereby Sema3A stimulation leads to Fyn activation and CDK5 phosphorylation are unknown. In this context, our study clearly demonstrated H2O2 generation by MICAL. Accumulating evidence indicates that protein tyrosine phosphatases are susceptible to H2O2-induced oxidation, which generally results in inactivation of their phosphatase activity (1517). Therefore, MICAL-generated H2O2 may contribute to sustaining CDK5 activation by inhibiting CDK5 inactivation through Tyr15 dephosphorylation.

An R-Ras–mediated pathway has been reported to participate in CRMP2 phosphorylation by GSK-3 (42, 43). Sema4D-activated PlexB acts as a guanosine triphosphatase–activating protein for R-Ras, thereby inactivating it, resulting in inhibition of Akt. Akt inactivates GSK-3 through phosphorylation of GSK-3 Ser9; thus, Sema4D stimulation ultimately induces GSK-3 activation. Our results (Figs. 5 and 6 and fig. S7) demonstrate the crucial importance of CRMP2 oxidation for its phosphorylation. Therefore, not only kinase activation but also substrate oxidative state is crucial for CRMP2 phosphorylation. In our experimental setting, using Sema3A-stimulated DRG neurons, we observed no Akt inactivation or GSK-3 activation (fig. S8). Thus, CRMP2 phosphorylation appears to be regulated at multiple levels in vivo, allowing precise and complex regulation of CRMP2 in different cell types.

Our study reveals that H2O2 generated by MICAL in response to Sema3A treatment is crucial for CRMP2 oxidation and growth cone collapse. Consistent with these results, mammalian MICAL has recently been reported to play an important role in Sema3A signaling (11). That study showed that CRMP binds to the monooxygenase domain of MICAL (11), suggesting that CRMP2 might be directly oxidized by MICAL. Some flavoprotein monooxygenases catalyze oxidation of thiol compounds, such as cysteamine, cysteine, and the tripeptide glutathione, to form disulfide bonds (44, 45). In addition, structural analysis suggests that stabilization of the monooxygenase domain of MICAL depends on the binding of macromolecules, such as polypeptides, and the authors proposed that proteins acted as these substrates (46). Therefore, formation of the complex between MICAL and CRMP2 may accelerate CRMP2 oxidation.

Cys504 is located in the C-terminal tail region of CRMP2, where many phosphorylation sites also exist. This tail region is predicted to be unfolded (47, 48). Thus, association with TRX at Cys504 would not be expected to prevent GSK-3 from interacting with these phosphorylation sites, because the target Ser residues can rotate freely in the unfolded polypeptide. Alternatively, TRX might recruit GSK-3β to its substrate, CRMP2.

The Cys504 is conserved in all vertebrate CRMP2, but not in Caenorhabditis elegans or Drosophila melanogaster (fig. S9). Moreover, the C-terminal region that includes Cys504 is also rarely conserved in these simpler organisms. This region is predicted to be unfolded (47, 48), and thus, we postulate that other Cys residues in C. elegans or D. melanogaster might compensate the function of vertebrate Cys504. Note that the phosphorylation sites for both GSK-3β and CDK5 (Thr509, Thr514, Ser518, and Ser522 in human CRMP2) are not conserved in C. elegans or D. melanogaster, but the corresponding residues are present in all vertebrate CRMP2s. Therefore, it is conceivable that CRMP2 oxidation and phosphorylation evolved to emerge simultaneously around the time of the advent of vertebrates to connect MICAL-generated H2O2 to CRMP2 phosphorylation.

CRMP proteins are known to form covalent bond–independent oligomers (49). In the protein complex composed of CRMP2 WT and C504S, CRMP2 WT should not be able to form a disulfide bridge with CRMP2 C504S. When CRMP2 C504S is ectopically expressed, the amount of CRMP2 C504S should be much larger than that of endogenous (wild type) CRMP2; thus, most endogenous CRMP2 is presumably trapped in the complex with CRMP2 C504S. Under these conditions, endogenous CRMP2 will not be able to form a disulfide bridge. This may provide an explanation for the ability of ectopically expressed CRMP2 C504S to act as a dominant negative to inhibit Sema3A-induced growth cone collapse.

Materials and Methods


The following commercially available antibodies were used: rabbit anti-TRX (Chemicon and Redox Bio Science); mouse anti-CRMP2 (Immuno-Biological Laboratories); rabbit anti-Akt, rabbit anti–phospho-Akt (Ser473), rabbit anti–phospho–GSK-3α/β (Ser21/9) (Cell Signaling Technology); mouse anti–GSK-3β (BD Transduction Laboratories); mouse anti-FLAG (M2), mouse anti–β-tubulin (Sigma); rabbit anti-Myc (A-14), mouse anti-Myc (9E10), rabbit anti-Slp1, rabbit anti-DHC (Santa Cruz Biotechnology); mouse anti–β-actin (Chemicon); and rabbit anti-Numb (Abcam). The rabbit antibody against CRMP2 was generated by immunizing rabbits with recombinant His-CRMP2 (amino acids 404 to 572) as an antigen and after affinity purification with GST-CRMP2. Specificity of the antibody against CRMP2 and TRX is indicated in fig. S10. Rabbit anti–phospho-CRMP2 (p-CRMP2) antibody that recognizes phosphorylated Thr514 and rabbit anti-KLC1 were as previously described (8, 36).

Plasmid constructs and siRNA

The human GSK-3β and mouse NRX complementary DNAs (cDNAs) were obtained as previously described (50, 51). The human CRMP2 cDNA was obtained from I.M.A.G.E clones (clone ID 6177866; Invitrogen). The cDNAs for mouse TRX, human TRP14, human TRP32, mouse CDK5, and mouse p25 were generated by RT-PCR. The mouse MICAL1 cDNA was provided by Kazusa DNA Research Institute (clone name, msh04044). The GFP-HyPer cDNA was purchased from Evrogen. Site-directed mutagenesis was conducted with the QuikChange Mutagenesis Kit (Stratagene). To construct TRX Rsc, we replaced T at position 73, T at position 76, and C at position 79 in the mouse TRX open reading frame with G, G, and A, respectively. To construct MICAL GW, we replaced three glycine residues at positions 91, 93, and 96 of mouse MICAL1 protein with tryptophan residues. The cDNA fragments were inserted into pEF-BOS, pCAGGS (Clontech Laboratories), pGEX-2T, pGEX-2T modified to generate proteins tagged with GST at the C terminus, pGEX-6p1 (GE Healthcare), pFastBac1 (Invitrogen), and pQE30 (Qiagen) plasmid vectors. Sema3A fused to alkaline phosphatase or control alkaline phosphatase expression plasmid was as previously described (5). Myc-PlexA2 and NP-1 expression plasmids were as previously described (41). The siRNA duplex oligonucleotides against rat TRX (#1, #2, and #3), rat MICAL1, and MCIAL3 were purchased from Invitrogen. The control siRNA had the shuffled sequence of #1 siRNA and was designed as a nonsilencing siRNA that does not correspond to any known mammalian mRNA sequence. The target mRNA sequences are as follows: #1, 5′-CCAATGTGGTGTTCCTTGAAGTAGA-3′; #2, 5′-CCTCTGTGACAAGTATTCCAATGTG-3′; #3, 5′-GAGTTCTCTGGTGCTAACAAGGAAA-3′; control, 5′-GAGTCTCGTGGAATCGAACGTTAAA-3′; MICAL1, 5′-GGCAGAATATGAGTTGGGCATCATA-3′; and MICAL3, 5′-CCCTGTCACTAGGTATCCCAATATT-3′.

Identification of TRX target proteins

For identification of TRX target proteins, lysates of NIH 3T3 cells stably expressing FLAG–TRX C35S or C32/35S were incubated with anti-FLAG M2 agarose beads (Sigma) for 2 hours. The beads were washed five times and then incubated with lysis buffer containing 5 mM DTT. The eluted proteins were separated by SDS-PAGE and stained with the SilverQuest Silver Staining Kit (Invitrogen). The bands of interest were excised from the gel and treated with trypsin. The resulting peptides were purified and analyzed by matrix-assisted laser desorption/ionization–time-of-flight tandem mass spectrometry (MALDI-TOF-MS/MS) (4700 Proteomics Analyzer; Applied Biosystems).

Growth cone collapse assays

DRG neurons were prepared as previously described (5, 7). They were removed from E14 Sprague-Dawley rats. Dissociated neurons were then transfected with various expression plasmids together with GFP-expressing plasmid by means of the Nucleofector device with the rat neuron nucleofector kit and program O-03 (Lonza Group AG) and plated on glass slides coated with poly-d-lysine (Sigma) and laminin (BD Biosciences). DRG neurons were cultured for 16 hours, and then the medium was replaced with a nerve growth factor (NGF)–free medium. Cells were cultured for an additional 3 hours and then stimulated with 0.5 nM Sema3A.

Sema3A proteins were prepared as described previously (7, 41). Briefly, human embryonic kidney (HEK) 293 cells were transiently transfected with Sema3A-expressing plasmid with Lipofectamine 2000 reagent (Invitrogen), and then the culture medium (conditioned medium) was collected after 48 hours to be used for stimulating neurons in culture. The abundance of Sema3A in the conditioned medium was determined by quantitative immunoblotting analyses. After stimulation with Sema3A, neurons were fixed with formaldehyde, and growth cones were visualized by staining with rhodamine-phalloidin. GFP-positive cells were measured for collapse response. For each experiment, more than 50 neurons were measured. Growth cones with fewer than two filopodia were considered collapsed, and the percentage of all growth cones that were collapsed was determined.

Growth cone turning assays

Growth cone turning induced by an extracellular gradient of Sema3A was performed as described previously (31, 52). DRG neurons from E9 chicks were dissociated and plated on a glass dish coated with poly-d-lysine and laminin. After several hours, 10 μM DNCB (Sigma) was bath-applied to culture media at least 60 min before the application of Sema3A gradients. To examine the function of CRMP2 C504S in growth cone turning, we transfected dissociated neurons with either Myc–CRMP2 WT or C504S-expressing plasmids together with GFP-expressing plasmid, using the Nucleofector device, and cultured overnight. The transfected cells were identified by GFP fluorescence. Sema3A gradients were applied with a micropipette containing Sema3A (100 μg/ml in pipette, R&D Systems). Micropipettes were set 100 μm from the growth cone at a 45° angle with respect to the original direction of axon elongation.

In utero electroporation and quantification of fluorescence intensities

Pregnant ICR mice were purchased from SLC Japan. In utero electroporation was performed as previously described (53, 54). Electroporated brains were cut into 16-μm coronal sections with a cryostat (Leica Microsystems). Fluorescence images of frozen sections of enhanced GFP–expressing mouse brains were captured by TCS SL laser scanning confocal microscopy (Leica Microsystems). Fluorescence intensities inside similar-width rectangles in various regions of the cerebral cortex (layers II to IV, V to VI, and IZ and VZ/SVZ) were measured by the TCS SL software as previously described (53). Relative intensities to the total fluorescence were calculated and plotted in the graphs with SEs.

Redox-state analyses

Cell lysates were incubated for 30 min at 37°C in the presence of 15 mM IAA and 1% SDS with frequent mixing and subjected to nonreducing or reducing SDS-PAGE. In the case of DRG neurons, cells were cultured in medium with 15 mM N-acetyl cysteine (NAC, Sigma) for 3 hours and then cultured in NAC-free medium. After 1 hour, cells were treated with H2O2 or stimulated with Sema3A for the indicated time and harvested.

H2O2 imaging analyses

We performed time-lapse imaging analyses with an Olympus IX81 microscope equipped with an Olympus DP30BW camera. GFP-HyPer was excited at 525 nm, and fluorescence from GHP-HyPer was monitored at 480 nm. DRG neurons were cultured in glass dishes (IWAKI) coated with poly-d-lysine and laminin. Images were taken every minute after Sema3A stimulation. We quantified image intensities with DP Controller software (Olympus) and ImageJ software (National Institutes of Health).

Statistical analyses

All statistical analyses were done with the Student’s t test. Error bars in the graphs represent SEM. Turning angle of growth cones, fluorescence intensities of GFP-HyPer, and immunoblots were quantified by measuring scanned photographs in ImageJ software (NIH).

Supplementary Materials

Fig. S1. Immunolocalization of ectopically expressed proteins in growth cones.

Fig. S2. TRX knockdown does not inhibit serum-induced biochemical changes.

Fig. S3. Inhibition of the radial migration of cortical neurons by CRMP2 C504S.

Fig. S4. IAA treatment blocks CRMP2 oligomer formation.

Fig. S5. RT-PCR analyses of MICAL1 to 3.

Fig. S6. Rescue of H2O2 generation by ectopically expressed MICAL.

Fig. S7. TRX promotes GSK-3β–dependent CRMP2 phosphorylation in vitro.

Fig. S8. Sema3A stimulation does not alter the degree of Akt or GSK-3β phosphorylation.

Fig. S9. Alignment of the amino acid sequence of the region of human CRMP2 containing Cys504 with the corresponding sequences.

Fig. S10. Characterization of the rabbit anti-CRMP2 antibody and the rabbit anti-TRX antibody.

References and Notes

  1. Acknowledgments: We thank S. Ohmi and H. Fukuda for technical help with mass spectrometry; J. Takagi, N. Yasui, and T. Tojima for helpful discussion; and T. Kawauchi for technical comments. Funding: This study was supported in part by a Grant-in-Aid for Scientific Research from the Japan Society for the Promotion of Science (H.M.) and the Ministry of Education, Culture, Sports, Science and Technology of Japan (H.M.). Author contributions: A.M. and H.M. are responsible for the overall study design. A.M. performed the experimental work and data analysis. M.Y. and M.H. performed in utero electroporation. Y.F. and Y.Y. trained A.M. in basic experimental procedures. R.I., H.K., F.N., and Y.G. trained A.M. in experiments with DRG neurons. T.Y. and K.K. generated anti–p-CRMP2 antibody. A.M. and H.M. wrote the paper with the help of the other authors. Competing interests: The authors declare that they do not have any competing financial, personal, or professional interest.
View Abstract

Stay Connected to Science Signaling

Navigate This Article