Department of Pharmacology and the Proteomics Center, Zhongshan School of Medicine, Sun Yat-sen University, Guangzhou, China
Address correspondence and reprint requests to Dr Mingtao Li, Department of Pharmacology, Zhongshan School of Medicine, Sun Yat-sen University, 74 Zhongshan Road II, Guangzhou 510080, China. E-mail: email@example.com
Neuronal activity shapes the dendritic arbour; however, most of the molecular players in this process remain to be identified. We observed that depolarization-induced neuronal activity causes an increase in the phosphorylation of glycogen synthase kinase-3 (GSK-3)α/β on Ser21/9 in cerebellar granule neurons. Using several approaches, including gene knockdown and GSK-3α/βS21A/S21A/S9A/S9A double knockin mice, we demonstrated that both GSK-3β and GSK-3α mediate activity-dependent dendritic growth and that Ser21/9 phosphorylation of GSK-3α/β plays an important role in this process. Collapsin response mediator protein-2 (CRMP-2), which is crucial for axon development, is phosphorylated at Thr514 and inactivated by GSK-3. We found CRMP-2 was located mainly in the dendrites of cerebellar granule neurons, in contrast to the axonal distribution in hippocampal neurons. Over-expression of CRMP-2 promoted and knockdown of CRMP-2 impaired dendritic growth, suggesting that CRMP-2 is necessary and sufficient for activity-dependent dendritic development. Furthermore, silencing CRMP-2 completely blocked the dendritic growth-promoting effects of GSK-3 knockdown, and expression of Thr514 nonphosphorylated form of CRMP-2 counteracted the inhibitory effect of constitutively active GSK-3. This data indicate that CRMP-2 functions downstream of GSK-3. Together, these findings identify a novel GSK-3/CRMP-2 pathway that connects neuronal activity to dendritic growth.
As sites of synaptic integration, dendrites play a critical role in neuronal information processing. The establishment of the dendritic arbour is a highly dynamic process that is regulated by both cell intrinsic programmes and extrinsic influences (Jan and Jan 2003; Emoto 2012). Studies of neuronal morphogenesis have revealed that extracellular cues, such as neuronal activity, control the development of dendrites by regulating intracellular signalling pathways, such as calmodulin-dependent kinases, Rho GTPases and MAPKs (Wong and Ghosh 2002; Konur and Ghosh 2005; Redmond 2008). However, how these pathways operate locally to control cytoskeletal elements within dendrites is not fully understood.
Glycogen synthase kinase-3 (GSK-3) is a multifunctional serine/threonine kinase that is composed of two distinct isoforms, GSK-3α and GSK-3β (Woodgett 1990). Inhibition of GSK-3β promotes dendritic growth in sympathetic, cortical and hippocampal neurons (Naska et al. 2006; Lim and Walikonis 2008). Abnormally increased GSK-3β activity contributes to dendrite degeneration under pathological conditions (Lin et al. 2010). Therefore, strict regulation of GSK-3β is required for normal dendrite development. N-terminal phosphorylation of GSK-3α/β on Ser21/9 (Ser21 in GSK-3α and Ser9 in GSK-3β) is a cardinal feature of GSK-3 inhibition (Frame et al. 2001). We and others have demonstrated that neuronal activity represses GSK-3β activity through CaMKII-mediated phosphorylation at Ser9 in cerebellar granule neurons (CGNs) and through integrin-linked kinase in sympathetic neurons (Delcommenne et al. 1998; Song et al. 2006). However, it is not known if the phosphorylation of Ser9 is required for neuronal activity-dependent dendritic growth. In addition, the role of GSK-3α in dendritic growth has long been neglected, even though GSK-3α and β exhibit similar expression patterns in neurons and share similar substrates.
Collapsin response mediator protein 2 (CRMP-2), also referred to as TOAD-64, Ulip-2 or DRP-2, is a microtubule-binding protein that is necessary for cytoskeletal remodelling (Fukata et al. 2002). CRMP-2 is enriched in the distal parts of axons in developing hippocampal neurons and plays a critical role in axonal outgrowth and neuronal polarity (Inagaki et al. 2001). CRMP-2 is phosphorylated at Ser522 by Cdk5 and subsequently at Ser518, Thr514 and Thr509 by GSK-3β (Cole et al. 2004). Phosphorylation of CRMP-2 by GSK-3β inactivates its activity, leading to inhibition of axonal growth and neuronal polarity (Yoshimura et al. 2005). However, the role of CRMP-2 in dendrites has not been studied. One recent study reported that Ser522-phophorylated CRMP-2 is enriched in dendrites of cortical neurons and is essential for dendritic field organization (Yamashita et al. 2012). As Ser522 is a priming site for GSK-3-mediated phosphorylation of CRMP-2, a functional relationship between GSK-3 and CRMP-2 in regulating dendrite development has yet to be identified.
In this study, we demonstrate that both GSK-3β and GSK-3α regulate dendritic growth in CGNs and that the phosphorylation of GSK-3 at Ser21/9 is critical for activity-dependent dendritic growth. Furthermore, we demonstrate that the GSK-3/CRMP-2 pathway is an important mediator of CGN dendritogenesis, which may have significant implications for neuronal connectivity in the brain.
Material and methods
GSK-3α/βS21A/S21A/S9A/S9A double knockin mice were kindly provided by Dr Dario Alessi of the University of Dundee (McManus et al. 2005). CGNs of Sprague Dawley rats, C57/BL6 or GSK-3α/βS21A/S21A/S9A/S9A double knockin mice were prepared from 7- or 8-day-old pups as described previously (Yuan et al. 2009; Xie et al. 2011). All procedures related to the care of animals were carried out in accordance with the Chinese National Health and Medical Research Council (NHMRC) animal ethics guidelines and approved by the Sun Yat-Sen University Animal Experimentation Ethics Committee. Briefly, neurons were dissociated from freshly dissected cerebella by mechanical disruption in the presence of trypsin and DNase and then seeded at a density of 1.5 × 106 cells/ml in basal modified Eagle's medium containing 10% foetal bovine serum and potassium at a concentration sufficient to induce membrane depolarization (25 mM KCl). For immunofluorescence analysis, CGNs were seeded on coverslips (Thermo Fisher Scientific, Rockford, IL, USA). Hippocampi were dissected from post-natal pups (0- or 1-day, Sprague–Dawley rats), and hippocampal neurons were dissociated with 0.125% trypsin and plated at a density of 1 × 104 cells/cm2 onto coverslips. Cultured hippocampal neurons were maintained in Neurobasal A medium containing 2% B27 and 0.5 mM glutamine supplement. We used poly-d-lysine as cell culture substrate for both CGNs and hippocampal neurons. Human embryonic kidney (HEK) 293 cells were cultured in Dulbecco's modified Eagle's medium with 10% foetal bovine serum as previously described (Ma et al. 2012). All ingredients for cell culture were obtained from Gibco/BRLLife Technologies (Eggenstein, Germany).
Transfection and drug treatments
For CGNs, transfection was performed within 24 h after plating using a previously described calcium phosphate method (Ma et al. 2012). The overall transfection efficiency using the calcium phosphate method in CGN culture was ~1%. One day after transfection, neurons were switched to serum-free medium containing 25 mM (25K) or 5 mM KCl (5K) for 72 h. The neurons were then fixed and subjected to immunocytochemistry. Upon 5K treatment, CGNs undergo apoptotic death; to study neuronal morphology in the absence of apoptotic effects, a Bcl-xl plasmid was co-transfected with a Green Fluorescent Protein (GFP) or β-galactosidase plasmid in all our experiments. As previously reported (Gaudilliere et al. 2004), Bcl-xl completely reversed CGN apoptosis in 5K media but did not affect neuronal morphology, which enabled us to study activity-dependent neuronal development in CGNs. Hippocampal neurons were transfected with the indicated plasmids on days in vitro (DIV) 3, and experiments were performed 3 days after transfection. For the administration of inhibitors, CGNs were incubated in 25K or 5K media in the presence or absence of CT99021 (Tocris Bioscience, Bristol, UK), AR-A014418 (Calbiochem, San Diego, CA, USA) or SB415286 (Sigma-Aldrich, St. Louis, MO, USA). Cells that did not receive inhibitors received dimethylsulphoxide as a control. To avoid toxicity, the final concentration of dimethylsulphoxide was < 0.1%.
Western blotting and antibodies
Western blotting analysis was performed as described previously (Ma et al. 2007; Yuan et al. 2009). Briefly, lysates were separated by sodium dodecyl sulphate–polyacrylamide gel electrophoresis and electrophoretically transferred to a polyvinylidene difluoride membrane. Membranes were blocked in Tris-buffered saline with 5% milk and 0.05% Tween 20 and probed with primary antibodies at 4°C overnight. The following antibodies were used at a dilution of 1 : 1000 unless otherwise indicated: anti-phospho-GSK-3α/β (Ser21/9), anti-CRMP-2 and anti-phospho-CRMP-2 (Thr514) (all from Cell Signalling Technology, Danvers, MA, USA); anti-GSK-3α/β (Upstate, Lake Placid, NY, USA); Tubulin (Sigma-Aldrich); and anti-V5 (Serotec, Kidington, UK). Appropriate horseradish peroxidase-conjugated secondary antibodies (Jackson Immuno-Research, West Grove, PA, USA) were used for detection with enhanced chemiluminescence reagents (GE Healthcare, Chalfont St. Giles, UK).
Constructs and RNA interference
The plasmids pcDNA3.1-V5-GSK-3α wild type (WT), pcDNA3.1-V5-GSK-3α S21A, pcDNA3.1-V5-GSK-3β WT and pcDNA3.1-V5-GSK-3β S9A have been described previously (Song et al. 2010). The pmGFP-C1 vector was a kind gift of Dr Paul De Koninck (Hudmon et al. 2005). Bcl-xl was cloned from rat cDNA into the EcoRI and NotI sites of the pCMV-HA vector. Rat CRMP-2 was digested from the pQE80-CRMP-2 plasmid [described previously (Liu et al. 2009)] and incorporated into the HindIII and XhoI sites of the pcDNA3.1/V5-His A plasmid. The RNAi-resistant CRMP-2 (CRMP-2Res) was generated with a site-directed mutagenesis kit using primer ‘GATCACGGGGTAAATAGTTTCCTAGTGTACATGGCTTTCA' as has been reported (Kawano et al. 2005). The CRMP-2 T514A (Thr514 was replaced by Ala) and T514D (Thr514 was replaced by Asp) mutation were generated by overlap extension PCR. All constructs were confirmed via DNA sequencing.
GSK-3 siRNA fragments were used as previously described (Song et al. 2010). Three CRMP-2 siRNAs (siCRMP-2-a, siCRMP-2-b and siCRMP-2-c) were designed to target rat CRMP-2 mRNA (NM_001105717) sequences: 5′-ACUCCUUCCUCGUGUACA-3′, 5′-GAUGGGUUGAUCAAGCAA-3′ and 5′-ACTCCTTCCTCGTGTACAT-3′ respectively. A non-targeting siRNA was used as a negative control for all siRNA transfection experiments. All siRNAs were synthesized by Shanghai GenePharma (Shanghai, China). The efficacy and specificity of siRNAs were determined as described previously (Song et al. 2010).
Neurons were grown on coverslips and processed according to the immunocytochemistry protocol as described previously (Yuan et al. 2009; Song et al. 2010). CGNs were fixed with freshly prepared 4% paraformaldehyde, followed by permeabilization with 0.1% Triton X-100 in Tris-buffered saline and blocking in 3% donkey serum. The cells on the coverslips were incubated with the following primary antibodies: rabbit anti-GFP (Abcam, Cambridge, UK), mouse anti-β-galactosidase (Cell Signalling Technology); mouse anti-MAP2 (Chemicon, Temecula, CA, USA); mouse anti-Tau-1 (Chemicon); and rabbit CRMP-2 (Cell Signalling Technology). Rabbit primary antibodies were detected with an anti-rabbit secondary antibody conjugated to Alexa Fluor 488 (Molecular Probes, Leiden, the Netherlands). Mouse primary antibodies were detected with an anti-mouse secondary antibody conjugated to Alexa Fluor 555 (Molecular Probes).
Morphometry of CGNs
To analyse the dendritic morphology of transfected CGNs in primary cultures, images of individual neurons were captured randomly and in a blinded manner on an Axio Observer Z1 (Carl Zeiss, Oberkochen, Germany) with a 40× objective and then analysed with Image Pro Plus 6 software (Media Cybernetics). We analysed at least 40 neurons per treatment and repeated three times or more. Axons and dendrites were identified in GFP-transfected neurons based on morphology and selective expression of MAP2 and Tau1 in dendrites and axons respectively (data not shown). Images were acquired in an unbiased manner and were scored blindly without previous knowledge of treatment. For colocalization analysis, neurons were scanned with the 63×oil-immersion objective on a confocal microscope (LSM 710 Meta; Carl Zeiss). Images were recorded with sequential acquisition settings at a resolution of 1024 × 1024 pixels and 12-bit depth and processed with LSM 710 software.
All experiments were repeated at least three times using independent culture preparations and performed blindly. The significance of difference between means was analysed by the anova with Bonferroni/Dunn post hoc tests (for multiple comparisons) and using Student's t-test (for single comparisons). The groups were compared using Student's t-test with control values. Significance is indicated as follows: *p < 0.05; **p < 0.01; ***p < 0.001.
Depolarization is critical for dendritic growth in CGNs
Neuronal activity is crucial for the development of dendrites and the formation of neural circuits (Volkmar and Greenough 1972; Harris 1981; Zhang and Poo 2001). Membrane depolarization acts as an inducer of excitability and various signals in the neurons, which has been used as a model of neuronal activity (Vaillant et al. 2002; Sohya et al. 2007). Chronic depolarization, usually achieved by raising the extracellular concentration of K+ ions (> 20 mM KCl), was applied in various cultured nerve cells, including cerebellar granule neurons (Gallo et al. 1987).
The cultured CGNs faithfully recapitulated the sequential generation of axons and dendrites observed in vivo (Powell et al. 1997) and displayed stereotypical morphology, with a T-shaped axon and multiple dendrites, providing a robust system for studying the development of axons and dendrites (Konishi et al. 2004; Kim et al. 2009; Puram et al. 2011). Membrane-depolarizing concentrations of potassium [KCl]o = 25 mM KCl (25K) mimicking the effect of neuronal activity, is not only required for CGN survival (Gallo et al. 1987; D'Mello et al. 1993) but also critical for dendritic growth (Gaudilliere et al. 2004; Nakanishi and Okazawa 2006).
To investigate activity-dependent dendritic morphology without the interference of apoptosis, CGNs were co-transfected with an anti-apoptotic protein Bcl-xl between all of the groups. The expression of Bcl-xl significantly prevented CGNs from death induced by low potassium (Fig. 1a, left panel), but did not affect dendritic and axonal growth either in complete medium containing serum or in serum-free 25K medium (Fig. 1a, middle and right panels), which is consistent with results from other groups who used CGNs as a model to study dendritic growth (Gaudilliere et al. 2004; Ramos et al. 2007). The anti-apoptotic effect of Bcl-xl in CGNs allowed the assessment of neuronal morphology in serum-free conditions. Under serum-free conditions, which were used in all subsequent experiment, we could clearly investigate underlying mechanisms by which depolarization/neuronal activity regulates dendritic growth.
CGNs differentiated well in serum-free 25K medium, with a T-shaped axon and multiple branching dendrites (Fig. 1b, marked by GFP), suggesting that depolarization is sufficient for CGN development. When neurons were maintained in 5K, it displayed fewer and shorter dendrites, with a total dendritic length of 154.6 ± 21.5 μm in 25K versus 43.6 ± 9.4 μm in 5K and total dendritic tips of 4.6 ± 0.6 in 25K versus 2.3 ± 0.2 in 5K (Fig. 1b and c). Cell body and nuclei were intact, arguing against the interpretation that the dendritic phenotype reflected an apoptotic process. These data together with previous reports (Vaillant et al. 2002; Gaudilliere et al. 2004) support a critical role of neuronal activity in dendritic growth.
Both GSK-3β and GSK-3α inhibit dendritic growth
GSK-3β has been reported to regulate neuronal dendritic growth (Munoz-Montano et al. 1999; Naska et al. 2006; Lim and Walikonis 2008). In the present experiment, we found that depolarizing concentrations of KCl increased the phosphorylation of GSK-3α/β on Ser21/9 in the developing CGNs (Fig. 2a), suggesting that GSK-3α/β activity was inhibited by neuronal activity throughout the CGN development process. We next examined whether GSK-3 activity negatively regulates dendritic growth. Pharmacological inhibition of GSK-3 by two specific and structurally unrelated GSK-3 inhibitors, CT99021 (CT) and AR-A014418 (AR) (Bain et al. 2007) significantly increased the dendritic length and branches of granule neurons in 5K (Fig. 2b and c), supporting that inhibition of GSK-3 promotes dendritic growth. To further distinguish the role of the two isoforms of GSK-3 in dendritic development, we introduced synthetic siRNA fragments specifically targeting GSK-3α or GSK-3β. The specificity and effectiveness of these siRNA fragments was demonstrated previously (Song et al. 2010). Knockdown of either GSK-3α or GSK-3β significantly increased the total dendritic length and tip numbers in 5K (Fig. 2d and e). Taken together, these results indicate that either GSK-3β or GSK-3α is involved in regulating depolarization-dependent dendritic growth.
Phosphorylation of GSK-3α/β (Ser21/9) is important for activity-dependent dendritic growth
Having demonstrated that GSK-3α/β activity inhibits dendritic growth, and that neuronal activity induces an increase in Ser21/9 phosphorylation, we next examined whether the inhibitory phosphorylation of GSK-3α/β at Ser21/9 is critical for activity-dependent dendritic growth. CGNs were transfected with constitutively active GSK-3α S21A or GSK-3β S9A mutants. Both GSK-3α S21A and GSK-3β S9A mutants reduced dendrite branching and tip numbers in 25K (Fig. 3a and b), indicating that the phosphorylation of GSK-3 at Ser21/9 is necessary for activity-dependent dendritic growth.
To further confirm the requirement of Ser21/9 phosphorylation for activity-dependent dendritic growth and exclude any unknown effects of gene over-expression, GSK-3α/βS21A/S21A/S9A/S9A double knockin mice (McManus et al. 2005) were used. CGNs from GSK-3α/βS21A/S21A/S9A/S9A double knockin mice consistently exhibited impaired dendritic growth in 25K medium as compared with wild-type mice (Fig. 3c and d).
Together, these data indicate that GSK-3α/β suppresses dendritic growth and that the inhibitory phosphorylation of GSK-3α/β on Ser21/9 is indispensable for activity-induced dendritic growth.
CRMP-2 is necessary and sufficient for activity-dependent dendritic growth
CRMP-2, a microtubule-binding protein, is responsible for cytoskeletal remodelling and axonal growth (Fukata et al. 2002; Arimura et al. 2004; Kimura et al. 2005; Xiong et al. 2012). Phosphorylation of CRMP-2 by GSK-3β inactivates its activity (Cole et al. 2004), leading to the inhibition of axonal growth and neuronal polarity (Yoshimura et al. 2005). Recently, Yamashita et al. suggested that CRMP-2 also localizes to the dendrites of cortical neurons, regulating dendritic patterning (Bretin et al. 2005; Yamashita et al. 2012).
To clarify the function of CRMP-2 in CGNs, we first observed the CRMP-2 distribution pattern in CGNs by immunostaining. Interestingly, strong immunoreactivity of CRMP-2 was observed in the dendrite shaft (Fig. 4a, line 1 shows the enlarged parts of dendrite, β-gal is in green for marking the neurons, CRMP-2 is in red; dendrite appears yellow in the merged images suggesting higher proportions of CRMP-2 throughout the dendrite), whereas faint staining of CRMP-2 was observed in the middle segments of the axon (Fig. 4a, line 2) as well as the distal region of the axon (Fig. 4a, line 3). This was different from hippocampal neurons in which CRMP-2 largely locates in the axons, especially is enriched in the distal parts of growing axon (Inagaki et al. 2001) and (Fig. 4b). To determine if CRMP-2 regulates activity-dependent dendritic growth in CGNs, we employed three different siRNA fragments to knockdown of endogenous CRMP-2 and observe the effects on CGN dendrites. The knockdown efficiency was first determined in HEK293 cells by western blot (Fig. 4c) and then in CGNs by immunostaining because of the low transfection efficiency in neurons (Fig. 4d and e). Knockdown of CRMP-2 shortened the total dendritic length and decreased dendritic branching (Fig. 4f), with axon morphology rarely affected (Fig. 4g). To exclude the possible off-target effects of CRMP-2 siRNA, we performed siRNA rescue experiment with CRMP-2Res, the siRNA-resistant form of CRMP-2, which has been described in previous report (Kawano et al. 2005). We found that expression of CRMP-2Res completely rescued dendritic development in CRMP-2-siRNA-transfected neurons, whereas transfection with CRMP-2 WT only had partial rescue effect (Fig. 4h), suggesting the CRMP-2 knockdown effect is specific. These results indicate that CRMP-2 is necessary for activity-dependent dendrite development in CGNs.
Over-expression of CRMP-2 in hippocampal neurons has been shown to alter neuronal polarity and lead to the formation of multiple axons. Our results were consistent with these reports (Fig. 5a). However, over-expression of CRMP-2 in CGNs specifically promoted dendritic growth (Fig. 5b and c) instead of affecting neuronal polarity (Fig. 5d) or axonal growth (Fig. 5e). Together, these data indicate a specific role of CRMP-2 in regulating acitivity-dependent dendritic but not axonal growth of CGNs.
CRMP-2 functions downstream of GSK-3 to promote activity-dependent dendritic growth
On the basis of the findings that GSK-3 inhibits and CRMP-2 promotes activity-dependent dendritic growth, we hypothesized that GSK-3 suppresses dendritic growth through phosphorylation and inhibition of CRMP-2 in CGNs.
CRMP-2 has been reported to be primed phosphorylation at Ser522 by Cdk5 and subsequently to be phosphorylated and inhibited by GSK-3β at Ser518, Thr514 and Thr509 (Cole et al. 2004; Yoshimura et al. 2005). In CGNs, the phosphorylation of CRMP-2 at Thr514 was increased in 5K medium compared with 25K medium (Fig. 6a). This increased phosphorylation was abolished by GSK-3 inhibitors (Fig. 6b), consistent with that GSK-3 directly phosphorylates CRMP-2 in CGNs under conditions of potassium deprivation. Furthermore, CGNs double-immunostained with CRMP-2 (green) and GSK-3 (red) showed yellow co-localized staining (Fig. 6c) in the dendrites of CGNs, consistent with that GSK-3 locally regulates CRMP-2 phosphorylation in the dendrites of CGNs.
To investigate their functional relationship in dendrites, CGNs were transfected with GSK-3 siRNAs with or without CRMP-2 siRNAs. Knockdown of GSK-3 enhanced dendritic growth, which was abolished by knockdown of CRMP-2 (Fig. 6d), implying that CRMP-2 functions downstream of GSK-3 to regulate activity-dependent dendritic growth. To further determine whether the regulation of CRMP-2 phosphorylation is the main effector of GSK-3-mediated dendritic growth, we constructed CRMP-2 T514A (Thr514 was replaced by Ala) and T514D (Thr514 was replaced by Asp) mutants which are expected to mimic unphosphorylated and phosphorylated forms of CRMP-2 respectively. We then examined whether the expression of the non-phosphorylated forms of CRMP-2 compensate for the GSK-3β-induced dendrite defect. The expression of CRMP-2 WT or CRMP-2 T514A counteracted the effects of GSK-3β S9A on dendrite outgrowth, with T514A more efficiently, while CRMP-2 T514D had no obvious effects (Fig. 6e). Similar results were also observed using GSK-3α (data not shown). Together, these results indicate that GSK-3 regulates dendritic growth through the phosphorylation of CRMP-2 in CGNs.
This study has identified a critical role of the GSK-3/CRMP-2 signalling pathway in regulating activity-dependent dendritic growth in CGNs. Our findings demonstrate the following: (i) both GSK-3β and GSK-3α inhibit dendritic growth, (ii) the phosphorylation of GSK-3α/β at Ser21/9 is necessary for activity-dependent dendritic growth, (iii) CRMP-2 functions as a downstream effector of GSK-3 to regulate activity-dependent dendritic growth.
CRMP-2 is reported to regulate axonal elongation and neuronal polarity in hippocampal neurons (Inagaki et al. 2001), but its role in dendrites has rarely been studied. Here, we observed that CRMP-2 is largely localized to the dendrites of CGNs, in contrast to the axonal distribution in hippocampal neurons. We and others have observed that, in hippocampal neurons, CRMP-2 over-expression changes the neuronal polarity and induces the formation of multiple axons (Fig. 5a and Inagaki et al. 2001). However, over-expression of CRMP-2 in cultured CGNs extended the dendritic length and increased dendrite branching, leaving axonal morphology unaffected. Thus, it is most likely that the differential distribution of CRMP-2 may account for the cell-type specific functions of CRMP-2. In addition to our findings that show that CRMP-2 is localized in neuronal dendrites, Ser522-phosphorylated CRMP-2 was recently shown to be enriched in the dendrites of cortical neurons and P7 cortical sections. CRMP-2 knockin mutant mice, in which the Ser residue at 522 was replaced with Ala, display increased numbers of dendrites in layer V neurons in vivo but not in cultured cortical neurons (Yamashita et al. 2012). This inconsistency may result from the difference between the in vivo and in vitro extracellular environments. However, in our cultured CGN model, enhanced CRMP-2 activity promoted dendritic growth, which was consistent with the in vivo cortical neuron data. Our results, together with the above reports, suggest that CRMP-2 not only regulates axon development but is also involved in dendritic growth depending on the neuronal type.
However, the mechanisms by which CRMP-2 contributes to dendrite outgrowth remain to be clarified. In cultured hippocampal neurons, CRMP-2 was reported to promote axonal outgrowth through various mechanisms, including (i) binding to tubulin heterodimers to activate microtubule assembly (Fukata et al. 2002); (ii) regulating Numb-mediated endocytosis of adhesion molecules (Nishimura et al. 2003); (iii) acting as a cargo receptor for kinesin-1 to transport its interacting proteins such as Sra-1/WAVE1 (Kawano et al. 2005) and Tubulin heterodimer (Kimura et al. 2005) to axons. GSK-3-mediated phosphorylation of CRMP-2 on T514 lowers the binding activity of CRMP-2 to Tubulin and decreases axonal growth. Although few studies have tested whether dendritic growth depends on tubulin subunit addition to the same extent as axon growth, the over-expression of cypin, a protein that possesses homology to the tubulin binding domain of CRMP-2, has been shown to promote microtubule assembly and thus increase dendrite growth and branching (Akum et al. 2004). Therefore, it is probable that GSK-3/CRMP-2/Tubulin pathway may also play a role in regulating dendritic growth.
GSK-3β has been reported to regulate dendritic growth (Munoz-Montano et al. 1999; Naska et al. 2006; Lim and Walikonis 2008; Shelly et al. 2011). Our results also demonstrated that inhibition of GSK-3 robustly increased dendritic growth in CGNs. Previous studies on the role of GSK-3 in neuronal morphology have predominantly emphasized GSK-3β rather than GSK-3α (Jiang et al. 2005; Yoshimura et al. 2005). We have reported that GSK-3α has equivalent protein abundance and exerts addictive proapoptotic effects with GSK-3β in CGNs (Song et al. 2010). Here, we further demonstrated that, in developing CGNs, GSK-3α Ser21 is phosphorylated in parallel with GSK-3β Ser9 phosphorylation in 25K (Fig. 2a). Moreover, knockdown of either GSK-3α or GSK-3β promoted dendritic growth (Fig. 2d and e). These data support the notion that both isoforms of GSK-3 suppress dendritic development.
What is the upstream regulator of GSK-3 during activity-dependent dendritic growth? Our previous data have demonstrated that CaMKII but not Akt/PKB, p90RSK, or PKA phosphorylates and inhibits GSK-3 during CGN depolarization (Song et al. 2010). However, the role of CaMKII in dendritic growth is complex and cell-type specific. In CGNs, CaMKIIα promotes (Gaudilliere et al. 2004), while CaMKIIβ inhibits (Puram et al. 2011) dendritic growth. In hippocampal neurons, CaMKIIβ controls dendritic morphology and synapse formation, while CaMKIIα regulates synaptic strength (Fink et al. 2003). In sympathetic neurons, CaMKII promotes dendrite formation (Naska et al. 2006), while in cortical (Redmond et al. 2002) and optic tectal neurons of Xenopus (Wu and Cline 1998), CaMKII inhibits dendritic growth. Therefore, whether CaMKII is the upstream kinase of GSK-3 in modulating activity-dependent dendritic development in CGNs requires further investigation.
The elaboration of dendrites is important to the establishment of neuronal circuits, and dendritic abnormalities are the most consistent anatomical correlates of mental retardation (Purpura 1975). Thus, elucidation of the cellular and molecular mechanisms underlying dendritic development is significant. Degeneration of dendritic structure has been observed in many neurological diseases such as Parkinson's disease (Gerfen 2006) and Alzheimer's disease (AD) (Kaufmann and Moser 2000; Knobloch and Mansuy 2008), which are often accompanied by abnormal increased GSK-3 activity (Wang et al. 2007; Hooper et al. 2008). Moreover, hyperphosphorylated CRMP-2, particularly at the GSK-3-sites Thr514 and Ser518, has been observed in both human AD cortex and in animal models of AD (Soutar et al. 2009). In this study, we demonstrated that CRMP-2 functions downstream of GSK-3 to mediate activity-dependent dendritic growth, which may provide new insights for the understanding of brain development and could lead to novel therapeutic targets for these neurodegenerative diseases.
The authors declare no conflicts of interest. We thank Dr Dario Alessi for kindly providing us the GSK-3α/βS21A/S21A/S9A/S9A double knockin mice. We thank Dr Paul De Koninck for providing us the pmGFP-C1 plasmid. This study was supported by the National Basic Research Programme of China 973 Program (2011CB504105), the National Natural Science Foundation of China (81030024, U1201224), the Natural Science Foundation of Guangdong Province (9351008901000003), and the China Postdoctoral Science Foundation funded project (No. 2011M501129).