Synaptic Synthesis, Dephosphorylation, and Degradation

Background: Although associated with intellectual disability, CDKL5 functions and regulation are poorly understood. Results: Upon neuronal activation, CDKL5 is dynamically regulated through local synthesis, dephosphorylation, and degradation. Conclusion: Neuronal activity and maturation control CDKL5 phosphorylation state and expression. Significance: This is the first report showing the activity-dependent modulation of CDKL5 in the neuronal periphery, further linking it to synapse development and plasticity. Mutations in the X-linked CDKL5 (cyclin-dependent kinase-like 5) gene have been associated with several forms of neurodevelopmental disorders, including atypical Rett syndrome, autism spectrum disorders, and early infantile epileptic encephalopathy. Accordingly, loss of CDKL5 in mice results in autistic-like features and impaired neuronal communication. Although the biological functions of CDKL5 remain largely unknown, recent pieces of evidence suggest that CDKL5 is involved in neuronal plasticity. Herein, we show that, at all stages of development, neuronal depolarization induces a rapid increase in CDKL5 levels, mostly mediated by extrasomatic synthesis. In young neurons, this induction is prolonged, whereas in more mature neurons, NMDA receptor stimulation induces a protein phosphatase 1-dependent dephosphorylation of CDKL5 that is mandatory for its proteasome-dependent degradation. As a corollary, neuronal activity leads to a prolonged induction of CDKL5 levels in immature neurons but to a short lasting increase of the kinase in mature neurons. Recent results demonstrate that many genes associated with autism spectrum disorders are crucial components of the activity-dependent signaling networks regulating the composition, shape, and strength of the synapse. Thus, we speculate that CDKL5 deficiency disrupts activity-dependent signaling and the consequent synapse development, maturation, and refinement.

athies characterized by the onset of intractable epilepsy within the first weeks of life, severe developmental delay, hypotonia, and some Rett syndrome-like features (1). CDKL5 encodes a serine/threonine kinase that is characterized by an N-terminal catalytic domain and a long C-terminal tail regulating the catalytic activity, subcellular localization, and stability of the protein (2,3). In human and mouse, the CDKL5/Cdkl5 mRNA and protein are widely expressed with a well recognized enrichment in brain (4). In mouse brains, CDKL5 is weakly expressed during embryogenesis and gets markedly up-regulated during postnatal development (3,4); furthermore, Cdkl5 transcription is regulated by various stimuli, depending on the specific brain district. Indeed, cocaine treatment of rats significantly reduces Cdkl5 mRNA levels in the striatum but not in the frontal cortex (5). At the cellular level, CDKL5 is easily detectable in all neuronal compartments, including the nucleus, cytoplasm, and postsynaptic fraction (6). Loss of function studies have demonstrated that CDKL5 is required for neurite outgrowth, dendritic spine development, and excitatory synapse stability (4,6,7). The first Cdkl5 knock-out mouse model is characterized by motor, social, and anxiety deficits similar to those observed in other autism and Rett syndrome mouse models; Cdkl5-null mice display also impaired learning and memory and altered event-related potentials (8,9). Deficits were also found in neural circuit communication and in multiple signaling pathways, including that of AKT-mammalian target of rapamycin (mTOR), already linked to MeCP2 and Rett syndrome (10).
Neuronal activity is known to influence dendritic outgrowth, synaptic maturation, and plasticity. In particular, membrane depolarization and the subsequent calcium influx trigger a variety of cellular changes, including post-translational modification of synaptic molecules, the induction of localized protein synthesis within dendrites and gene transcription in the nucleus. Altogether, these events regulate synaptic function and the response of neuronal circuits to experience (11). These neuronal activity-dependent mechanisms are involved in learn-ing and memory and adaptive behavioral responses; accordingly, autism spectrum disorders have often been associated with crucial components of the activity-dependent signaling networks regulating synaptic maturation and function (12). Considering that CDKL5 (i) is associated with autism spectrum disorders; (ii) is required for learning and memory and neuronal maturation; (iii) gets transiently phosphorylated upon BDNF stimulation (4); and (iv) gets excluded from the nucleus upon extrasynaptic NMDA receptors (NMDAR) activation (13), we hypothesized that it might be involved in activity-dependent signaling networks. We therefore investigated how neuronal activity regulates CDKL5, at both transcriptional and posttranscriptional levels.
Herein, we show that neuronal depolarization induces CDKL5 within 5 min through NMDARs and mainly relying on local protein synthesis. The increase of CDKL5 is transient and is rapidly followed in mature neurons by its proteasomal degradation. We suggest that this response depends on changes in the phosphorylation state of CDKL5 and requires the activation of NMDARs and protein phosphatase 1 (PP1). Indeed, only in neurons beyond a critical developmental stage, NMDAR activation induces the dephosphorylation of CDKL5, prompting its degradation. Eventually, Cdkl5 transcription is activated, although with slower kinetics, resembling those of immediate early genes. We speculate that the kinase might be a crucial component of activity-dependent signaling pathways and, therefore, might affect synapse development and plasticity.

EXPERIMENTAL PROCEDURES
Mice-Mice (CD1 genetic background) were housed and treated according to the regulations on mouse welfare and ethics and with the approval of the institutional animal care and use committee of the University of Insubria.
Primary Neuronal Cultures and Cell Lines-Primary cortical and hippocampal cultures were prepared from brains of CD1 mouse embryos at 17 days, considering the day of the vaginal plug as embryonic day 0, as described previously (3), and plated on poly-L-lysine-coated dishes at different densities (hippocampal neurons, 16,000 cells/cm 2 ; cortical neurons, 26,000 cells/cm 2 ). After 4 days in vitro (DIV4), cytosine-1-␤-D-arabi-nofuranoside (Sigma-Aldrich) was added at the final concentration of 2 M to prevent astroglial proliferation. The murine neuroblastoma cells, Neuro 2a (N2a), were grown in MEM supplemented with 10% fetal bovine serum, penicillin (100 IU/ml), and streptomycin (100 g/ml) at 37°C in a humidified 5% CO 2 atmosphere.
Electrophysiology-Whole cell voltage or current clamp recordings were obtained from cultured hippocampal neurons at DIV7-DIV19 using a Multiclamp 200B amplifier. Current/ voltage traces were filtered at 2 KHz and acquired at 10 KHz with a Digidata 1440A interface. During recordings, cells were bathed in standard bicarbonate-buffered external saline (125 mM NaCl, 2.5 mM KCl, 1.25 mM NaH 2 PO 4 , 26 mM NaHCO 3 , 2 mM CaCl 2 , 1 mM MgCl 2 , 10 mM glucose, pH 7.4, after saturation with 5% CO 2 , 95% O 2 ) at room temperature. For high K ϩ stimulation, a modified external saline was obtained from bicarbonate-buffered external saline by decreasing NaCl to 78.5 mM and increasing KCl to 50 mM. All solutions were bath-applied at 2 ml/min. Patch pipettes (3)(4)(5)  Cultured hippocampal neurons were treated with KCl for 5 min or 3 h or, as controls for cell death, with 0.5 mM Cd 2ϩ or 200 M H 2 O 2 for 3 h and washed with ice-cold KRH solution. Subsequently, neurons were incubated for 15 min with 2.5 g/ml propidium iodide in KRH on ice, washed twice, and fixed with 4% paraformaldehyde. Fixed neurons were immunostained with anti-synapsin 1 and counterstained with DAPI. Images were taken with an IN Cell Analyzer 1000 at the ALEMBIC facility (San Raffaele Scientific Institute, Milan, Italy), and the percentage of propidium iodide positive neurons with respect to the total number of neurons (synapsin 1 positive) was calculated. For each time point, an average of 150 cells were analyzed.
Apoptosis Assays-TUNEL stain was used to analyze the presence of apoptosis in cortical slices (100 m) perfused with 30 mM KCl (1 h) as described above or as a control with 0.5 mM Cd 2ϩ (3 h). The slices were fixed with 4% paraformaldehyde as above, and the NeuroTACS TM II in situ apoptosis detection kit (Trevigen) was used following the manufacturer's instructions. Images were taken with an Axio ImagerA2 equipped with an AxioCam MRm camera (Zeiss), keeping a fixed exposure time.
Somatodendritic Fractionations-High density cortical neurons (100,000/cm 2 ) at DIV16 were treated for 5 min with 50 mM KCl (or as control with 50 mM NaCl) at 37°C, rinsed with Hanks' balanced salt solution, and treated for 2 min with papain (0.1 mg/ml; Sigma-Aldrich) and DNase (50 g/ml; Sigma-Aldrich) in Hanks' balanced salt solution at 37°C. In some experiments, MG132 (50 M) was included to avoid protein degradation. The cultures were then rinsed three times with ice-cold PBS and covered with 1.5 ml/dish (60 mm) of an ice-cold buffer (10 mM Tris-HCl, pH 7.4, 150 mM NaCl, 0.5 mM EGTA, protein inhibitor cocktail, 1 mM NaF, 1 mM NaVO 4 , 1 mM PMSF, 0.5 mM DTT) containing 0.35 M sucrose, transposed to a new tube, and then mixed 1:1 with a similar solution containing 0.8 M sucrose. The final solution was then passed through a plastic wide tip to break down the neuronal network. The nuclearsomatic pellet was isolated after centrifugation at 1500 ϫ g for 10 min at 4°C, whereas the supernatant was further centrifuged at 28,000 ϫ g for 10 min, producing the dendritic pellet enriched in pre-and postsynaptic structures. At the end the pellets were resuspended in lysis buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, Nonidet P-40 1%, 1 mM EDTA, 0.5 mM DTT, protein inhibitor cocktail, 1 mM PMSF) and analyzed by WB.
Synaptoneurosome Preparation-Synaptoneurosomes were prepared mainly as described in De Rubeis et al. CaCl 2 ), whereas the synaptoneurosomes were recovered in the interphase of the 1.25 M/1 M sucrose fractions and pelleted by centrifugation at 15,000 ϫ g for 10 min at 4°C and finally resupended in 1 ml of resuspension buffer. The synaptoneurosomes were equilibrated by incubation at 37°C for 30 min before stimulation with 50 mM KCl for 5 min at 37°C; an isomolar exposition to NaCl was used as control. In the end, the synaptoneurosomes and the somatic/nuclear fraction were analyzed by Western blotting.
Immunofluorescence-Hippocampal neurons were processed for immunofluorescence as previously described (3). Images were taken with a Nikon Eclipse Ni microscope equipped with a Nikon Digital Sight DS-2MBWc camera. For the quantification of CDKL5 induction in dendrites, images were taken with a Leica TCS SP2 confocal microscope (ALEMBIC, San Raffaele Scientific Institute). Image analysis was performed with ImageJ software: the MAP2 signal was used to select the dendritic profile excluding the soma, and the corresponding CDKL5 fluorescence intensity was quantified.
Western Blot Analysis-Approximately 10 g of protein extracts were heated (at 70°C to preserve phosphorylation, otherwise at 95°C) and separated by 8% SDS-PAGE (37.5:1 acrylamide:bis-acrylamide) using Tuj1 as internal standard. To enhance the separation of phosphorylated and nonphosphorylated CDKL5 isoforms, we exploited 7% SDS-PAGE with 77:1 acrylamide:bis-acrylamide. Filters were developed using Super-Signal West Pico Chemiluminescent Substrate (Pierce) or ECL Prime Western blotting detection reagent (GE Healthcare) and quantified with a CCD camera (Syngene, G-Box Chemi XRQ) where each sample was measured relative to the background or by scanning densitometry using the QuantityOne software package (Bio-Rad). To evaluate the phosphorylation status, a ratio between the normalized phospho-protein and the normalized total isoforms of the same protein was calculated.
RNA Isolation and Quantitative RT-PCR Analysis-Total RNA was isolated using the miRNeasy Kit (Qiagen) according to the manufacturer's instructions. Contaminating DNA was removed with RQ1 RNase-free DNase (Promega), and the obtained RNA was quantified with a Biotech Ultrospec 2000 UV-visible spectrophotometer, and its quality was assessed through agarose gel electrophoresis. cDNA was synthetized from 200 ng of RNA using the SuperScript II reverse transcriptase kit (Invitrogen) as indicated by the manufacturers. Quantitative real time PCR was performed with a Chromo 4 TM detector real time PCR machine with GoTaq quantitative PCR Master Mix (Promega). The quantitative RT-PCR was performed in 25 l using 10 ng of cDNA and the following cycling parameters: 15 s at 95°C, 30 s at 60°C, and 30 s at 72°C. Each independent sample was analyzed in triplicate from RNA collected from at least three independent neuronal cultures. All quantitative RT-PCR values were normalized to GAPDH. A melting curve was automatically generated for each sample and confirmed that a single amplicon was generated in each reaction.
Statistical Analysis-All values are expressed as the average of at least three different experiments Ϯ S.E. The significance of results was evaluated by Student's t test, one-way analysis of variance-Dunnett, or two-way analysis of variance test (Bonferroni post hoc analysis), and statistical significance was established as p Ͻ 0.05.

RESULTS
Neuronal Activation Triggers a Rapid, Protein Synthesis-dependent Increase in CDKL5 Levels-As described in the introduction, CDKL5 might play a pivotal role in converting the effects of transient neuronal stimuli into changes in the molecular pathways that are important for the brain to process information. To analyze whether CDKL5 expression might be affected by neuronal membrane depolarization and in analogy with several studies based on long and short KCl treatments as a paradigm for neuronal activation (16,17), we treated primary hippocampal neurons, prepared from embryonic day 17 mouse embryos and cultured for in vitro for 7 days, with KCl (50 mM) for 5 min. Neuronal activation was confirmed by WB, using phospho-specific antibodies against CaMKII␣ (T286), CREB (S133), and ERK1/2 (Thr-202 and Tyr-204), respectively, with respect to total protein levels ( Fig. 1A). Neuronal viability was assessed by current clamp or voltage clamp whole cell recordings in which the recovery of spiking activity during KCl-induced depolarization block was verified (Fig. 2, A-C) together with the presence of spontaneous postsynaptic potentials or currents with frequency similar to the control (Fig. 2D). As further controls, we used propidium iodide staining to verify that short or prolonged treatments with KCl are not toxic to neuronal cultures at different maturation stages (Fig. 2E).
The response of CDKL5 to neuronal depolarization was analyzed by using a specific antibody (6), and as shown in Fig. 1 (B and C), CDKL5 levels significantly increased after 5 min of treatment; as expected, no increment could be detected in the protein levels of ERK1/2 and c-Fos ( Fig. 1C and data not shown). To confirm the KCl-dependent induction in an ex vivo model, cortical slices from young adult mice (P30 -P35) were stimulated with KCl (30 mM) for 5 min. As shown in Fig. 1 (D and E), WB analyses confirmed that CDKL5 levels are promptly increased in response to membrane depolarization also in a complex neuronal network. As before, we controlled the toxicity of short and long term KCl treatments in brain slices of different developmental ages by analyzing cell mortality through propidium iodide staining and DNA fragmentation through TUNEL assays (Fig. 2, F and G, respectively).
We investigated whether the observed fast increase in CDKL5 protein levels was regulated at the transcriptional or translational level. Pretreatment of hippocampal neurons with the protein synthesis inhibitor cycloheximide (40 M, 30 min) completely prevented the KCl-induced increase in CDKL5 levels (Fig. 1F). Complementary, real time PCR was employed to investigate whether membrane depolarization induced Cdkl5 transcription and/or the induction of well known CREB-responsive immediate early genes, such as Akt1, Egr1, Glur2, and c-Fos (Ref. 18 and Fig. 1G). A modest but significant increase in Cdkl5 transcription was observed in neurons treated with KCl; this induction was comparable with that observed for the other Biphasic Modulation of CDKL5 by Neuronal Activity FEBRUARY 13, 2015 • VOLUME 290 • NUMBER 7 early responsive genes. When transcription was blocked through actinomycin D treatment (Fig. 1H, upper panel), the KCl-mediated induction of CDKL5 was nonetheless comparable with that of the control (Fig. 1H, lower panels, compare  lanes b and c with lane a), demonstrating that the activity-dependent CDKL5 increase relies on translation of pre-existing mRNA.
Synaptic Stimulation Induces the Localized Translation of CDKL5 mRNA-The subcellular distribution of CDKL5 in the nuclear and cytoplasmic compartments is modulated by specific stimuli (13); in the dendrites, the protein colocalizes in the postsynaptic compartment with PSD-95 (6, 7). Thus, we employed a biochemical approach to investigate whether the KCl-dependent induction of CDKL5 occurs in the soma or whether, in analogy with other factors involved in synaptic activity, it preferentially occurs in the dendrites (Fig. 3, A-C). Treated and untreated high density neuronal cultures (DIV16; see "Experimental Procedures" and Ref. 19) were fractionated, and two protein extracts were obtained enriched in cell bodies or dendrites, respectively. The successful fractionation was verified by WB (Fig. 3A), which showed the presence of histone H3 or PSD-95 mainly in the somatic or dendritic fractions, respectively. We found that CDKL5 protein levels increase preferentially in the dendritic fraction upon KCl stimulation, whereas its levels remain rather constant in the soma (Fig. 3, B and C). We hypothesized that Cdkl5 mRNA might get translated in the  neuronal periphery, and we therefore analyzed synaptoneurosomes (see "Experimental Procedures"), which are fractions enriched in isolated functional pre-and postsynaptic terminals, from adult mouse cortices (P30 -P35). The enrichment of PSD-95 and synaptophysin (SYP1) in the synaptic extract verified the fractionation, whereas a significant contamination with cell bodies was excluded staining for histone H3 (Fig. 3D). By RT-PCR, we revealed the presence of Cdkl5 mRNA in the synaptoneurosomal fraction together with Bdnf transcripts, used as positive control; on the contrary, and as expected, we were unable to reveal the presence of Mecp2 mRNA in this fraction (Fig. 3E). After 5 min of stimulation of synaptoneurosomes with KCl, CDKL5 levels were subtly but significantly increased together with the activation of ERKs, demonstrated by the increased ratio of phosphorylated to total ERKs (Fig. 3, F and  G). Eventually, the dendritic induction of CDKL5 upon depolarization was confirmed also in DIV7 neurons by confocal microscopy using MAP2 as dendritic marker (Fig. 3H).

Membrane Depolarization Induces Dephosphorylation of CDKL5 and Its Proteasome-mediated Degradation-Having
demonstrated the depolarization-dependent stimulation of CDKL5 synthesis, we studied the time course of the induction and of its maintenance, treating DIV7 neurons with KCl (50 mM) for 0, 5, 20, 40, 60, and 180 min. CDKL5 levels (Fig. 4A) rapidly increased but subsequently declined, reaching basal levels within 20 min after stimulation and completely disappearing after 3 h of treatment. This result was confirmed in a complex native neuronal network by comparing CDKL5 levels in young adult (P30 -35) cortical slices perfused for 1 h with KCl with those in unstimulated slices (Fig. 4, B and C). To verify whether a sustained stimulation is required for the down-regulation of CDKL5, we treated DIV7 hippocampal neurons for 5 min with KCl and returned them to basal conditions for 3 or 24 h before WB analysis. As shown in Fig. 4D, a short stimulation is sufficient to down-regulate CDKL5. However, protein expression returned to base line levels after 24 h recovery in resting conditions, thus demonstrating that CDKL5 levels are dynamically tuned in response to neuronal activity.
Because we have previously demonstrated that CDKL5 levels can be regulated through proteasomal degradation (13,20), we tested whether this mechanism is also involved in the observed down-regulation. DIV7 neurons were incubated with the proteasomal inhibitor MG132 (50 M) prior to KCl induction. Indeed, the presence of MG132 contained CDKL5 reduction after sustained KCl treatment (Fig. 4, E and F). As already proposed (13), in the presence of MG132, CDKL5 levels rise in unstimulated neurons as well, therefore confirming that CDKL5 turnover is constitutively regulated by proteasomal degradation.
While we were performing these experiments, we noticed that, depending on the electrophoretic conditions, CDKL5 shows an increased mobility upon 20 min KCl induction. Because the migration of some proteins is affected by their phosphorylation status and protein phosphorylation has frequently been associated with neuronal activation and synaptic plasticity, we investigated whether CDKL5 phosphorylation is modulated upon membrane depolarization. As shown in Fig.  5A, after 20 min of KCl stimulation and under appropriate electrophoretic conditions (7% SDS-PAGE with 77:1 acrylamide: bis-acrylamide), CDKL5 migrates mainly as a single band with a higher mobility than the prominent band from untreated neurons. We incubated unstimulated or stimulated neuronal extracts with -phosphatase and confirmed that the mobility of the dephosphorylated kinase is identical to that caused by prolonged KCl stimulation, therefore confirming that the acquired higher mobility is caused by CDKL5 dephosphorylation (Fig.  5B).
We proceeded analyzing whether a brief stimulation was sufficient to induce the appearance of the faster isoform of CDKL5. Neurons were thus depolarized by KCl for 5 min and allowed to recover for another 15 min in resting medium. The brief stimulation was sufficient to induce CDKL5 dephosphorylation that gets more pronounced upon longer incubation in the resting medium (Fig. 5C). To reproduce this observation in a more complex neuronal network, we analyzed the electrophoretic mobility of CDKL5 from depolarized cortical slices (Fig. 5D). In immature cortices (P6.5), the mobility of CDKL5 is not affected even after a prolonged depolarization (1 h); on the contrary, at P11.5 the same stimulus induces the appearance of small amounts of a faster migrating isoform of CDKL5. At P21 and P30, CDKL5 dephosphorylation is visible after only a 5-min pulse of KCl, therefore suggesting that this modification correlates with neuronal maturation. Accordingly, in undifferentiated neuronal cells, such as mouse Neuro 2a, the mobility of CDKL5 is affected by phosphorylation (Fig. 5E), but prolonged KCl treatment induces neither its dephosphorylation nor its degradation (Fig. 5, F and G). Altogether, these data indicate that only in mature neurons, membrane depolarization is associated with a rapid and transient increase in CDKL5 levels followed by an event of dephosphorylation; further, they suggest that in addition to being rapidly synthesized, CDKL5 gets quickly degraded after neuronal activation.
NMDARs Mediate CDKL5 Induction-As schematized in Fig. 6A, neuronal activation can induce local protein synthesis mainly through NMDARs, AMPARs, or voltage-dependent calcium channels; thus, we proceeded investigating the upstream signals that regulate the activation of CDKL5 protein synthesis upon membrane depolarization. To block KCl-mediated calcium influx, we pretreated hippocampal cultures with EGTA and demonstrated that calcium is required to induce CDKL5 (Fig. 6B). This increase was prevented by blocking NMDA or AMPA receptors with the specific antagonists AP5 or CNQX, respectively. It is widely accepted that local protein synthesis requires the activation of ERKs and PKA. Consistently, the KCl-dependent increase in CDKL5 levels was prevented by selectively inhibiting ERKs through preadministration of U0126. Accordingly, we detected an increase in CDKL5 levels, comparable with that of KCl-stimulated neurons, after  Fig. 6C). In accordance with the dependence of the ERKs for the activity-dependent induction of CDKL5, the kinase was significantly induced after 5 min of treatment with BDNF that leads to ERK activation through the TrkB receptors (Fig. 6D).
Altogether these experiments suggest that a robust Ca 2ϩ influx through the NMDA/AMPA receptors induces the observed synaptic synthesis of CDKL5. It is interesting to observe that, in immature neurons (DIV3; see Fig. 8B), the KClinduced expression of CDKL5 is calcium-dependent (data not shown) and maintained in the presence of AP5 and CNQX, therefore suggesting that it does not rely on NMDARs (Fig. 6E). Thus, in young neurons, CDKL5 induction appears mediated by voltage-dependent calcium channels (21); to the best of our knowledge, this is the first time that such a switch in the involved intracellular pathways gets proposed.
NMDARs Trigger a PP1-dependent CDKL5 Dephosphorylation Coupled to Its Subsequent Degradation-We then investigated the signals that mediate CDKL5 dephosphorylation in hippocampal neurons. Dephosphorylation was inhibited by preventing extracellular calcium entry, via EGTA, or blocking NMDARs, using AP5; on the contrary, pretreatment of the cultured neurons with CNQX and U0126 did not affect CDKL5 dephosphorylation, indicating that MAP kinases and AMPARs are not involved in the observed phosphate removal (Fig. 7A). By incubating hippocampal neurons with NMDA for 20 min, we further demonstrated that Ca 2ϩ influx through NMDARs triggers CDKL5 dephosphorylation (Fig. 7B). On the contrary, other stimuli such as forskolin or BDNF, which can rapidly induce the expression of CDKL5 (Fig. 6, C and D), are not able to promote its dephosphorylation (Fig. 7C).
Neuronal activation and synaptic plasticity events are mediated by affecting the functions and interactions of a multitude of proteins, mainly through changes in their phosphorylation status; thus, phosphatases, in addition to kinases, play a pivotal role for synaptic plasticity. The majority of these phosphorylation events occur on serine and threonine residues, and in general, the protein phosphatases PP1, PP2A, and calcineurin (PP2B) are considered the most important dephosphorylating enzymes (21). To exclude the involvement of tyrosine dephosphorylation in the KCl-dependent change of CDKL5 electrophoretic mobility, we inhibited tyrosine phosphatases by pretreating primary hippocampal neurons with Na 3 VO 4 . As clearly shown in Fig. 7D, even if tyrosine phosphatases were efficiently inhibited (see "Experimental Procedures"), no inhibition of KCl-dependent CDKL5 dephosphorylation was observed. Hippocampal neurons were then treated with the potent calcineurin inhibitor deltamethrin; as shown in Fig. 7E, the drug does not impede CDKL5 dephosphorylation although the efficient inhibition was verified (see "Experimental Procedures"). On the contrary, the faster migrating, dephosphorylated band is not present when 100 nM calyculin A, which inhibits PP1 and PP2A, is administered to neurons, therefore implying the involvement of either or both enzymes (Fig. 7E). The presence of 20 nM okadaic acid (OA), which selectively inhibits PP2A and not PP1 (21), does not impede the KCl-dependent CDKL5 dephosphorylation (Fig. 7F), suggesting the critical involvement of PP1. Accordingly, 1 M OA, which inhibits both PP1 and PP2A, impedes the KCl-dependent CDKL5 dephosphorylation, confirming the pivotal involvement of PP1.
Because the prolonged activation of Neuro 2a cells with KCl or of neurons with BDNF induces neither the dephosphory- lation nor the degradation of CDKL5 (Figs. 5, F and G, and 7C), we suggest the existence of a neuronal-specific molecular link between the activity-dependent dephosphorylation and the degradation of CDKL5 after membrane depolarization. Consequently, we addressed whether dephosphorylation is required for CDKL5 demolition. To this purpose, a prolonged depolarization was applied to DIV7 hippocampal neurons pretreated with MG132, OA, or a combination of both inhibitors, and CDKL5 levels were estimated by Western blotting (Fig. 7G). As before, the proteasome inhibition blocked the degradation induced by sustained depolarization (compare column d with column b); of relevance, under these conditions the faster isoform of CDKL5 gets accumulated (Fig. 7H). As shown in column f with respect to column b, PP1/2A inhibition partially blocks the KCl-induced degradation; in the presence of both inhibitors and KCl (column h), CDKL5 levels are rescued without any additional effect (CDKL5 levels in columns h and d are not statistically different), therefore suggesting that the protea- some and the phosphatase PP1 work in the same pathway. Eventually, by testing CDKL5 levels in depolarized neurons pretreated with deltamethrin or different concentrations of OA (Fig. 7I), we demonstrated that only by blocking PP1, the enzyme that causes the appearance of the faster migrating CDKL5 isoform, the kinase degradation was impeded (compare lanes c and b). Thus, we suggest that CDKL5 dephosphorylation by the phosphatase PP1 is mandatory for its subsequent demolition.
Neuronal Maturation Refines the Mechanisms Controlling CDKL5 Levels-As already mentioned, CDKL5 levels increase in response to neuronal maturation (3,4); further, the protein activity and its subcellular localization are modulated by neuronal development (13). We thus proceeded investigating whether the observed modulation of CDKL5 levels depends on neuronal maturation (from DIV3 to DIV21; Fig. 8). To this purpose, aware that the maturation state of cultured neurons depends on the growth conditions, we characterized our hippocampal neuronal cultures by costaining for the pre-and post-synaptic markers synapsin 1 and GluR2, respectively (Fig. 8A). As expected, with both markers DIV3 neurons do not exhibit dendritic punctae, therefore indicating their immaturity. DIV7 neurons are more progressed in the maturation process but are still lacking a clear punctated postsynaptic pattern. From DIV14, the formation of functional synapses between neurons is proved by the discrete GluR2-positive dendritic punctae juxtaposed to synapsin 1-positive presynaptic dots. As shown in Fig. 8 (B and C), the extent of CDKL5 induction appears to inversely correlate with neuronal maturation, whereas CDKL5 demolition gets progressively faster. Indeed, in young neurons (DIV3), 5 min of KCl treatment is sufficient to significantly induce CDKL5 expression; however, the degradation is retarded upon prolonged KCl treatment and is not accompanied by the appearance of the faster migrating isoform. On the contrary, a very brief depolarization of DIV21 (inset in Fig. 8C) permits visualization of the induction of CDKL5 that at 5 min is already masked by the very rapid degradation. Altogether, these data reinforce the observation that CDKL5 dephosphorylation  FEBRUARY 13, 2015 • VOLUME 290 • NUMBER 7 JOURNAL OF BIOLOGICAL CHEMISTRY 4523 ensues at a certain maturation stage and suggest that a novel mechanism of control is imposed on CDKL5 upon neuronal maturation, permitting the kinase levels to be more rapidly and finely tuned. Accordingly, as shown in Fig. 8D, in immature neurons, CDKL5 levels are only slightly affected by the proteasome system. Interestingly, the differential regulation of CDKL5 at different developmental stages was recapitulated in a more complex ex vivo system: a 1-h KCl stimulation is required in newborn mice to induce CDKL5 expression, whereas at further developmental stages, CDKL5 promptly responds to brief KCl stimulation and undergoes degradation if the stimulus is prolonged, with a kinetic resembling that observed in cultured neurons (Fig. 8E).

DISCUSSION
In this study, we reveal that neuronal activation and the consequent calcium influx up-regulates CDKL5 levels within 5 min of stimulation. Although Cdkl5 transcription is induced by membrane depolarization, localized activation of protein synthesis appears as the main mechanism involved in the observed up-regulation. In fact, the kinetics of induction of CDKL5 protein levels is not compatible with its dependence on prior transcription. Moreover, the response occurs predominantly in the dendritic fraction and can be obtained by depolarizing purified synaptoneurosomes. In immature neurons, this induction does not depend on NMDARs, whereas in more mature neurons AMPARs and NMDARs mediate the KCl-dependent increase of CDKL5. In addition, the kinetics of the decline of this activation changes dramatically with neuronal maturation. In fact, in DIV3 neurons CDKL5 levels remain above the basal ones for 1 h after KCl stimulation, whereas at further developmental stages the kinase returns to basal or even lower levels within few minutes (10 min at DIV14; 20 -30 min at DIV7). This suggests that CDKL5 levels need to be tightly regulated and that both gain and loss of functions might be detrimental for proper brain development. Accordingly, duplications in CDKL5 have very recently been identified in patients with neurologic deficits (22). The developmental regulation of the response of CDKL5 to neuronal activation is further supported by the fact that the KCl-mediated depolarization induces the dephosphorylation of CDKL5 only beyond a certain developmental stage. CDKL5 dephosphorylation depends mostly on PP1 activity, occurs in soma and dendrites (data not shown), and is mandatory for its proteasome-dependent degradation. A similar regulation of phosphorylation, occurring only in mature neurons and depending on effectors downstream of NMDARs, has already been described for CREB, a transcription factor that is widely implicated in synaptic plasticity and memory formation (21). Dephosphorylation and subsequent proteasomal degradation is also commonly used for a prompt regulation of protein functions: among others, the levels of the fragile X syndrome-associated protein FMRP are controlled at the synapses through a dephosphorylation-dependent proteasome-mediated degradation (23).
Recently, Ebert and Greenberg (12) have postulated that networks controlling synapse development and functions are of high relevance for the pathogenesis of autism spectrum disorders. We suggest that CDKL5 might belong to these neuronal activity-dependent signaling pathways whose full characterization should help in understanding the molecular basis of autism spectrum disorders and in developing future therapies.
Molecular biology supports this hypothesis. Indeed, a link between CDKL5 and spine development and plasticity has been established demonstrating that the kinase (i) affects neuronal morphogenesis through cytoskeleton rearrangements (4); (ii) is localized at excitatory synapses where it contributes to correct dendritic spine structure and synapse activity (6, 7); (iii) binds the scaffolding protein PSD-95 (7); and (iv) binds amphiphysin 1, a protein involved in synaptic vesicle endocytosis (24).
Neuronal activity triggers changes in the synapse composition, shape, and strength. This can be achieved by inducing the post-translational modification of synaptic molecules, by the localized synthesis of dendritic proteins, and by altering nuclear gene expression. Locally activated CDKL5 might phosphorylate diverse targets therefore affecting spine maturation and plasticity. For instance, upon neuronal activity, AMPA receptors modify their conductivity through a specific CaMKII-dependent event of phosphorylation (25). Interestingly, CaMKII, the most studied kinase involved in the regulation of long-term potentiation (26), is locally synthetized in the postsynaptic compartment after NMDAR induction (27).
Further, distally transported mRNAs are bound by a number of translational inhibitors that impose a repression that has to be removed locally to permit protein synthesis (28). Few kinases have been identified as being locally up-regulated; their enzymatic activity can affect receptor cycling and activation, translation initiation, or elongation or can be relevant to remove RNA binding repressors or to induce splicing. Thus, in the future, it will be highly relevant to reveal the proteome and the phospho-proteome of activated spines devoid of CDKL5. Of relevance, the vast majority of missense pathogenic mutations localize in the catalytic domain of CDKL5, therefore confirming the relevance of its kinase activity for proper brain function (1).
Synaptic stimulation also transmits signals to the nucleus affecting specific transcription programs. So far, few pieces of evidence suggest that CDKL5 has a role in regulating gene FIGURE 8. The response of CDKL5 to depolarization depends on neuronal maturation. A, immunofluorescence of hippocampal neurons at the indicated DIV with antibodies against the pre-and postsynaptic markers, synapsin 1 (red) or GluR2 (green), respectively. Scale bar, 100 m. B, WB and graphs showing CDKL5 levels in hippocampal neurons at DIV3 (left panels) and DIV21 (right panels) after treatment with KCl for the indicated time points. Tuj1 was used as loading control. Asterisk in the WB indicates an unspecific band. The statistical significance (t test) was tested by comparing CDKL5 levels at each time point with basal levels (gray bars). C, graph comparing the changes in CDKL5 levels in neurons at DIV3, 7, 14, and 21 treated with KCl as indicated. The expression of CDKL5 at DIV3, 14, and 21 was compared with the expression at DIV7 by using a two-way analysis of variance test (Bonferroni post-tests; n Ն 3); no statistical significance could be determined at any time point between DIV7 and DIV14. The WB in the small inset shows the change in CDKL5 levels in DIV21 neurons treated with KCl for the indicated time points. D, CDKL5 levels in DIV3 hippocampal neurons treated for 3 h with KCl, for 6 h with MG132, or with both. The graph shows the fold change in CDKL5 levels in treated neurons. n ϭ 3, means Ϯ S.E. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001. E, WB and graphs showing CDKL5 levels in cortical slices at the indicated ages perfused with KCl for 5 min or 1 h. (n Ն 3). expression such as its capability to interact with the transcriptional repressor MeCP2 (29), the DNA methyl transferase DNMT1 (30), and with the nuclear speckles involved in RNA splicing (31). Considering the soma, we have demonstrated that neuronal activity first induces Cdkl5 transcription with kinetics similar to those of immediate early genes and then causes CDKL5 dephosphorylation and degradation. These results reinforce the already hypothesized involvement of CDKL5 in influencing gene expression programs indicating the importance of defining the transcriptional signature of activated neurons devoid of CDKL5. Therefore, in the future we will investigate which pathways affect the observed regulation of CDKL5 and its possible involvement in synaptic plasticity, particularly in long-term potentiation and long-term depression.