Activity-dependent synaptic plasticity represents a cellular mechanism suitable for lasting adaptation of network activities during memory formation (Bliss and Collingridge,1993; Frey and Frey,2008; Malenka and Nicoll,1999; Matthies et al.,1990). The molecular and structural changes that cause these long-lasting functional changes in synaptic efficacy require remodeling of synaptic protein ensembles by regulated phosphorylation (Malinow et al.,1989; Roberson et al.,1999), protein translocation (Krucker et al.,2000), and turnover of proteins (Behnisch et al.,2004; Deadwyler et al.,1987; Frey et al.,1988, 1996; Mochida et al.,2001; Stanton and Sarvey,1984). Regulated turnover of proteins is accomplished by de novo protein synthesis (Bramham and Wells,2007; Hernandez and Abel,2008) as well as via targeted degradation through the ubiquitin-proteasome system (UPS) (Hegde,2004; Yi and Ehlers,2007). In fact, inhibition of either protein ubiquitination (Foley et al.,2000), proteasome activity (Lopez-Salon et al.,2001) or ubiquitin ligase (Jiang et al.,1998) leads to deficits in hippocampal LTP and behavioral performance. Moreover, we previously demonstrated that UPS-dependent protein degradation is a prerequisite for induction and input specificity of hippocampal LTP in the CA1 area (Cai et al.,2010; Karpova et al.,2006).
Postsynaptic proteins that are targeted by UPS include several prominent scaffolding molecules such as PSD-95, GKAP/SAPAP, ProSAP/Shank and AKAP (Ehlers,2003). Little is known, however, about the precise sequence of proteasome-dependent degradation events relative to the onset of synaptic stimulation and subsequent reorganization of the synapse. In several cases, activity-dependent protein degradation has emerged as a second order event related to the structural remodeling of synapses. For instance, degradation of SPAR—a key regulator of dendritic spine morphology that acts at the interface between the PSD and the actin-based cytoskeleton (Pak et al.,2001)—results from its phosphorylation by Polo-like kinase 2 (Plk2; also known as serum-induced kinase, SNK; Pak and Sheng,2003). Expression of the kinase largely depends on the level of synaptic activity (Kauselmann et al.,1999), implying that Plk2-induced degradation of SPAR by UPS represents a relatively late event.
The role of SPAR in activity-dependent synaptic plasticity is of great interest. SPAR appears to be selective for spiny neurons and is present in about 65% of the mature mushroom-shaped spines, inducing an increase in complexity of spine shape when overexpressed (Pak et al.,2001). In hippocampal neurons, SPAR is localized in dendritic spines and its over expression causes enlargement of spines, many of which adopt an irregular appearance associated with putative multiple synapses. Dominant negative SPAR constructs cause narrowing and elongation of spines. In addition, it was reported that the EphA4 receptor regulates neuronal morphology through SPAR-mediated inactivation of Rap GTPases (Richter et al.,2007) implicating Rap signaling in the regulation of postsynaptic structures.
Previous studies in primary cell culture systems, suggested that SPAR degradation induced by Plk2 contributes to homeostatic regulation of synaptic transmission. Homeostatic plasticity refers to changes in synaptic function and structure that are induced by prolonged increases or decreases of synaptic activity (Jakawich et al.,2010; Perez-Otano and Ehlers,2005) and which normalize the overall synaptic drive of neurons and thereby stabilizes network dynamics (Seeburg and Sheng,2008). Mechanisms associated with LTP are, in contrast, considered to occur at a much faster time scale.
It remains to be established whether activity-dependent SPAR degradation takes place during synaptic input-specific induction of LTP in CA1 neurons of acute hippocampal slices. Here, we employed a viral gene delivery system to express eGFP-tagged SPAR proteins in CA1 neurons. We combined electrophysiological recordings and confocal microscopy to monitor eGFP-SPAR fluorescence in apical dendrites of CA1 neurons while applying a LTP inducing stimulation paradigm. We found that tetanization induced a reduction of eGFP-SPAR fluorescence that was diminished by proteasome-, cyclin-dependent kinase 5 (CDK5)- and protein synthesis inhibitors. We conclude that tetanization induces CDK5-dependent phosphorylation of SPAR that, in turn, caused SPAR degradation by UPS. These results link LTP induction to mechanisms of activity-dependent structural remodeling of synapses.
Preparation of Semliki forest particles carrying an eGFP-SPAR fusion gene
A fragment encoding enhanced green fluorescence protein (eGFP) fused to spine-associated Rap GTPase activating protein (SPAR) was excised from a pCDNA3.1-based construct (kind gift by Dr. Christina Spilker, Leibniz Institute for Neurobiology, Magdeburg, Germany) through digestion with NheI and SmaI. The fragment was blunted by a fill-in reaction using Klenow polymerase and inserted into the SmaI restriction site of the pSFV1 expression vector. The recombinant pSFV-eGFP-SPAR plasmid, a pSFV-eGFP control plasmid and the pSFV-Helper2 plasmid (which encodes SFV structural proteins) were linearized and reverse-transcribed to yield RNAs for the generation of Semliki Forest particles as described by Ehrengruber et al. (1999). In brief, RNA derived from pSFV-eGFP-SPAR or pSFV-eGFP were co-transfected with RNA derived from pSFV-Helper2 into Human Embryonic Kidney 293 (HEK-293T) cells with Lipofectamine 2000 (Invitrogen). After 48 h, the culture medium containing budded particles was harvested, passed through 0.22 μm filters and stored at −80°C. Thirty minutes before conducting intrahippocampal injection, particles were activated by alpha-chymotrypsin (Sigma, US; CAS No. 9004-07-3; 500 mg/l) for 30 min at room temperature and the reaction inactivated by aprotinin (Sigma, US; CAS No 9087-70-1; 250 mg/l).
Animal care and procedures were approved and conducted under established standards of the Institutes of Brain Science and State Key Laboratory of Medical Neurobiology of Fudan University, Shanghai, China.
Intrahippocampal virus injections
Wistar rats (2–3 weeks old) (Experimental Animal Center, Shanghai Medical School, Fudan University, People's Republic of China) were anesthetized with chloral hydrate (450 mg/kg, Sigma, MO) by intraperitoneal (i.p.) injection and then secured in a stereotaxic frame (Stoelting, Wood Dale, IL). Glass pipettes (Harvard Apparatus, Holliston, MA) with a tip length of about 10 mm and a tip diameter of about 30 μm were filled with activated virus particle containing solution and positioned in the hippocampus through a 0.5 mm diameter hole in the skull (vertical coordinates V: −4.0 mm). The injection coordinates for anterior–posterior (AP) and lateral (L) positions were −5.4 mm and ±5.2 mm, respectively. Pipettes were connected over a paraffin oil-filled plastic tube with a Hamilton syringe (CR-700-50, Hamilton Co., Höchst, Germany). A total of 1.5 μl of the particle containing solution was injected in three portions at 5 min intervals. The pipette was redrawn by 0.2 mm after each injection. The injection pipette remained at the final position for additional 10 min. This procedure was effective to transduce a high number of cells expressing eGFP or eGFP-SPAR (Fig. 1A).
Hippocampal slices preparation
Acute hippocampal slices were prepared as described previously (Cai et al.,2010; Leutgeb et al.,2003, 2005). Briefly, 20–24 h after intra-hippocampal injection, rats were anesthetized using ether and the isolated brains were immersed in carbogenated ice-cold Gey's solution (composition in mM: 130 NaCl, 4.9 KCl, 1.5 CaCl2·2H2O, 0.3 MgSO4·7H2O, 11 MgCl2·6H2O, 0.23 KH2PO4, 0.8 Na2HPO4, 5 glucose, 25 HEPES, 22 NaHCO3, pH 7.32). Transverse hippocampal slices (300 μm) were cut (Vibratome 3000, St. Louis, MO) and immediately transferred to an interface type slice chamber to recover for at least 1 h at 32°C in carbogenated ACSF (containing in mM: 110 NaCl, 5 KCl, 2.5 CaCl2·2H2O, 1.5 MgSO4·7H2O, 1.24 KH2PO4, 10 glucose, 27.4 NaHCO3, pH 7.3).
Hippocampal slices were transferred to a submerged slice chamber and perfused constantly with carbogenated ACSF at 32°C. Field excitatory postsynaptic potentials (fEPSPs) were evoked by stimulation of Schaffer-collateral fibers with biphasic rectangular current pulses (200 μs/polarity) in a range of 15–30 μA through ACSF filled glass pipettes with a resistance of 0.5 MΩ. The tip of the stimulation electrodes was positioned in close proximity to clearly visible fluorescent dendrites (Fig. 2A1) (Jager et al.,2002). The fESPSs were recorded using ACSF filled pipettes and amplified by an AxonClamp 200B amplifier (Molecular Devices, Shanghai, China). The strength of synaptic transmission was estimated by calculating the initial slope of fEPSPs. Stimulation strength was adjusted to 40% of the maximum fEPSP-slope values. Responses to test stimuli were measured every minute or 5 min throughout the experiment. The averaged fEPSP-slope values were depicted in diagrams as Mean ± SEM. The recorded field potentials were digitized at a sample frequency of 10 kHz by Digidata 1400plus AD/DA converter (Molecular Devices, Shanghai, China).
Late-LTP was induced by a tetanization consisting of three 1-s trains at 100 Hz (Cai et al.,2010) every 6 min (Fig. 2A2).
Image acquisition and analysis of time-lapse recordings
The time course of fluorescence changes in GFP-expressing CA1 neurons was acquired by scanning of several optical planes using a confocal microscope (Olympus, Fluoview 1000, Shanghai Branch, China) mounted on an upright fixed stage microscope every 3 min. The distance between the first and the last plane was between 30 and 90 μm. Fluorescence intensity values (averages over regions of interest, ROIs, as indicated) were calculated for each time point according to the following mode of image processing: (i) averaging of the planes that encompass dendrites, (ii) crop of dendrite, (iii) alignment of averaged planes across time points, (iv) measuring of fluorescence intensities of ROIs located on dendrites (including their spines) as well as background. ROIs were placed around clearly visible and in focus dendritic segments (Fig. 2D). Dendrites were chosen that are clearly visible and localized at the same horizontal and vertical positions as the stimulation electrode (Fig. 2A1) (Jager et al., 1998,2002; Karpova et al.,2006). Fluorescence intensities were background corrected and expressed as percentage changes from baseline. Image processing and analysis were conducted using open source imaging software ImageJ3.5. Experiments with drug applications were interleaved with control measurements in the absence of drug. Control measurements are presented as one group.
Drugs, including AP5 (50 μM; Du et al.,2009), roscovitine (5 μM), anisomycin (30 μM; Sigma-Aldrich, Shanghai, China), MG132 (25 μM; Calbiochem, La Jolla, CA) (Cai et al.,2010; Craiu et al.,1997; Karpova et al.,2006) were bath-applied at concentrations as indicated. Stock solutions of MG132 in dimethyl sulfoxide (DMSO, Sigma-Aldrich, Shanghai, China) were stored at -20°C. Stock solutions were diluted prior to use to the desired working concentration in ACSF. To avoid precipitation of MG132, the stock solution was diluted slowly in freshly carbogenated ACSF and kept protected from light. The final DMSO concentration did not exceed 0.1%.
Data acquisition and analyses were carried out using Clampex 10.2 and Clampfit 10.2 (Molecular Devices, Shanghai, China) and SPSS that allow measurements of fEPSP-slope and the calculation of descriptive statistics. Data gained from electrophysiological or imaging experiments were normalized to baseline values and expressed in percentage as Mean ± SEM%. Comparisons of groups were done by using Mann–Whitney U-test. P values of ≤0.05 were considered as statistically significant. Numbers of slices corresponds to numbers of animals studied. Drug experiments were interleaved with drug-free controls.
To investigate LTP-related SPAR degradation and to monitor its activity-dependent degradation, SFV-based particles encoding eGFP-SPAR were injected into the hippocampus of 2–3 week old rats. Particles encoding eGFP served as a control. Fluorescence microscopic inspection of hippocampal slices from injected animals revealed substantial expression of eGFP-SPAR or eGFP within 20–24 h post injection. Moreover, eGFP-SPAR was enriched in spine heads whereas fluorescence of eGFP alone was uniformly distributed within spines and dendrites (Fig. 1B).
Synaptic activity-dependent attenuation of eGFP-SPAR fluorescence
To monitor the time course of activity-dependent SPAR degradation, we acquired the fluorescence of dendrites (including their spines) every 3 min (Fig. 2B). Over a period of 45 min eGFP-SPAR fluorescence decreased by about 20% (9th min: 95.2 ± 6.78%; 42nd min: 78.3 ± 4.46%). We attribute this decrease to photobleaching of eGFP. In contrast, eGFP-SPAR decreased by 70% in slices that were tetanized three times every 6 min to induce LTP (9th min: 91.7 ± 4.59%; 42nd min: 38.7 ± 5.55%; Fig. 2C).
To test if this additional decrease in eGFP-SPAR fluorescence is mediated through activation of the proteasome we bath-applied the proteasome inhibitor MG132 starting 20 min before tetanization (Karpova et al.,2006). Bath-application of MG132 reduced the fluorescence decrease significantly in comparison to drug free experiments (P ≤ 0.05; 9th min: 99.4 ± 3.33%; 42nd min: 60.3 ± 5.76%; Fig. 3A). We therefore ascribed the tetanization-induced additional decline in eGFP-SPAR fluorescence to activity and UPS-dependent SPAR degradation.
It has been suggested that SPAR degradation requires protein synthesis. To study if protein synthesis is indeed required for SPAR degradation we inhibited RNA translation using anisomycin. Application of anisomycin (30 μM) significantly reduced the decrease of fluorescence after tetanization (P ≤ 0.05; 9th min: 94.4 ± 1.2%; 42nd min: 59.9 ± 3.56%; Figs. 3B1 and 3B2). Comparisons of drug-free and anisomycin-treated experiments revealed a significant difference 10 min after tetanization (Fig. 3B).
Previously it was suggested that CDK5 phosphorylates SPAR (Seeburg et al.,2008), however its involvement in tetanization-induced SPAR degradation in hippocampal slices has not yet been established. To clarify if phosphorylation by CDK5 is required for activity-dependent SPAR degradation we bath-applied the CDK5 inhibitor roscovitine during image acquisition. Bath application of roscovitine (5 μM) during tetanization reduced the eGFP-SPAR degradation (9th min: 95.9 ± 1.44%; 42nd min: 63.5 ± 5.41%). Differences in comparison to drug-free experiments reached significance as soon as 5 min after first tetanization (P ≤ 0.05, Fig. 3C).
Protein synthesis-independent eGFP-SPAR degradation
The results described so far showed that SPAR degradation depends, at least in part, on protein synthesis and CDK5-mediated phosphorylation. However, in the presence of protein synthesis inhibitor the fluorescence decay of eGFP-SPAR still differed from that of eGFP (P ≤ 0.05; eGFP: 9th min: 96.8 ± 0.99% and 42nd min: 79.1 ± 3.40%; Fig. 4A). This suggests that strong synaptic activity could enhance SPAR degradation in addition via a protein-synthesis independent signaling pathway. In line with this possibility, we observed that in the presence of anisomycin, bath application of the NMDA receptor blocker APV over the whole time of imaging caused significantly less degradation of eGFP-SPAR (9th min: 97.3 ± 0.62%; 42nd min: 81.37 ± 4.39%) than seen in APV-free experiments (P ≤ 0.05; Fig. 4B).
Importantly, when protein synthesis was blocked, the CDK5 inhibitor lost its effect on SPAR degradation (9th min: 98.5 ± 1.97%; 42nd min: 67.6 ± 9.01%; Fig. 4C) whereas inhibition of the proteasome still significantly delayed the fluorescence decrease after tetanization in comparison to the degradation under anisomycin (P ≤ 0.05; MG132: 9th min: 94.3 ± 1.2%; 42nd min: 74.82 ± 5.93%; Fig. 4D). These findings are consistent with the Plk2-CDK5 pathway being dependent on de novo protein synthesis but also imply the existence of a second, protein synthesis-independent pathway involved in UPS-mediated SPAR degradation.
The present results revealed that induction of LTP in acute hippocampal slices triggers a UPS-dependent degradation of SPAR via two complementary signaling pathways: (i) a protein synthesis-dependent route involving CDK5, which corresponds to a previously demonstrated mode of homeostatic plasticity during chronically elevated activity (Seeburg et al.,2008) and (ii) a protein-synthesis independent pathway that requires activation of NMDA receptors.
In a previous study we found that LTP induction triggered de novo synthesis and a net increase in the level of d2eGFP, a degradable reporter sensitive to UPS activity. This effect was reverted to a MG132-sensitive net decrease when protein synthesis was inhibited, demonstrating that in this case synthesis exceeds degradation (Karpova et al.,2006).
On a global level, LTP-related synaptic remodeling requires the well-balanced synthesis and degradation of proteins (Fonseca et al.,2006). Focusing on the turn-over of an individual synaptic protein, however, both processes may even act in synergy. This is particularly obvious in the case of SPAR. In the present study we found that even in the absence of protein synthesis inhibition, LTP did not increase but rather decreased eGFP-SPAR levels. Therefore, in the case of SPAR, degradation exceeds any potential increase in translational activity triggered by LTP induction. Moreover, SPAR degradation itself depends on de novo synthesis of a cofactor as indicated by the effect of anisomycin. This factor most likely is Plk2, a member of the Polo-like kinase family (Pak and Sheng,2003; Seeburg et al.,2005).
For example, it was found that Plk2 transcription was enhanced in response to prolonged epileptiform activity of neurons and mediated activity-dependent reduction in membrane excitability (Kauselmann et al.,1999). Expression of a dominant-negative Plk2 or suppression of Plk2 by RNA interference not only prevented down-regulation of membrane excitability during epileptiform activity, but also unmasked a potentiation in synaptic strength that prevented further induction of long-term potentiation (Seeburg and Sheng,2008). Moreover, it was shown that SPAR degradation induced by Plk2 is an effective homeostatic mechanism to maintain synaptic plasticity.
In addition, it was suggested that interaction of SPAR with other proteins requires phosphorylation. In particular, phosphorylation of SPAR on ser-1328 by CDK5 was shown to prime the binding of Plk2 to its substrate SPAR. Moreover, it was shown that inhibition of CDK5 by roscovitine increased SPAR protein levels in neurons (Seeburg et al.,2008; Seeburg and Sheng,2008). Accordingly, in our experiments inhibition of CDK5-mediated phosphorylation by roscovitine prevented activity-dependent SPAR proteolysis. Similar results were described in cell culture studies where CDK5 inhibition also prevented the decay. Moreover, CDK5 was shown to be involved in LTP induction. This implies that SPAR degradation requires a precise sequence of steps including phosphorylation and protein synthesis to become a target of proteasome mediated protein degradation.
The CDK5 and protein synthesis-independent component of tetanization-induced SPAR degradation has previously not been described. Interestingly, this mechanism required NMDA receptor activation and, hence, most likely a calcium-dependent signaling pathway. Our results are consistent with the idea, that a tight control of SPAR expression is crucial for learning and memory (Lu et al.,2009; Spilker and Kreutz,2010).
The present findings strongly suggest that the integrity of SPAR mediated homeostatic plasticity is required for normal LTP. More generally, our study supports the notion that LTP must be associated with some form of homeostatic plasticity to secure long-term balance and stability within a synaptic network.