Synaptic scaling has been proposed as a form of plasticity that may contribute to drug addiction but it has not been previously demonstrated in the nucleus accumbens (NAc), a critical region for addiction. Here we demonstrate bidirectional synaptic scaling in postnatal rat NAc neurons that were co-cultured with prefrontal cortical neurons to restore excitatory input. Prolonged activity blockade (1–3 days) with an AMPA receptor antagonist increased cell surface (synaptic and extrasynaptic) glutamate receptor 1 (GluR1) and GluR2 but not GluR3, as well as GluR1/2 co-localization on the cell surface and total GluR1 and GluR2 protein levels. A prolonged increase in activity (bicuculline, 48 h) produced opposite effects. These results suggest that GluR1/2-containing AMPA receptors undergo synaptic scaling in NAc neurons. GluR1 and GluR2 surface expression was also increased by tetrodotoxin alone or in combination with an N-methyl-d-aspartate receptor or AMPA receptor antagonist but not by the l-type Ca2+ channel antagonist nifedipine. A cobalt-quenching assay confirmed the immunocytochemical results indicating that synaptic scaling after activity blockade did not involve a change in abundance of GluR2-lacking AMPA receptors. Increased AMPA receptor surface expression after activity blockade required protein synthesis and was occluded by inhibition of the ubiquitin-proteasome system. Repeated dopamine (DA) treatment, which leads to upregulation of surface GluR1 and GluR2, occluded activity blockade-induced synaptic scaling. These latter results indicate an interaction between cellular mechanisms involved in synaptic scaling and adaptive mechanisms triggered by repeated DA receptor stimulation, suggesting that synaptic scaling may not function normally after exposure to DA-releasing drugs such as cocaine.
Synaptic scaling is a form of homeostatic plasticity in which prolonged activity blockade leads to augmented excitatory synaptic transmission, whereas prolonged increases in activity produce opposite effects (Turrigiano & Nelson, 2004; Thiagarajan et al., 2007; Turrigiano, 2008). This stabilizes the activity of neurons and neuronal circuits against a backdrop of ongoing perturbations including Hebbian plasticity. For example, synaptic scaling may prevent synapses from becoming unresponsive due to ceiling effects after repeated long-term potentiation or prevent their run-down after repeated long-term depression. Although many mechanisms contribute to the expression of synaptic scaling, the major postsynaptic component involves AMPA receptor (AMPAR) accumulation after chronic activity blockade and removal after enhanced activity (Lissin et al., 1998; O’Brien et al., 1998;Turrigiano et al., 1998; Wierenga et al., 2005).
We have proposed that synaptic scaling contributes to plasticity in animal models of cocaine addiction (Boudreau & Wolf, 2005). Specifically, the surface and synaptic expression of glutamate receptor 1 (GluR1)- and GluR2-containing AMPARs increases in the nucleus accumbens (NAc) after withdrawal from repeated cocaine injections (Boudreau & Wolf, 2005; Boudreau et al., 2007, 2009; Kourrich et al., 2007), whereas GluR2-lacking AMPARs are added to NAc synapses after prolonged withdrawal from long-access cocaine self-administration (Conrad et al., 2008). Imaging studies in humans and primates have demonstrated persistent metabolic hypoactivity after cocaine exposure in cortical areas that send glutamate projections to NAc neurons (Goldstein & Volkow, 2002; Porrino et al., 2007). We therefore hypothesized that decreased excitatory transmission after cocaine withdrawal leads to synaptic scaling in NAc neurons, resulting in the addition of AMPARs to NAc synapses (Boudreau & Wolf, 2005).
Synaptic scaling has been demonstrated in spinal, cortical and hippocampal neurons but not in NAc or other basal ganglia neurons. Our goal was to determine if synaptic scaling occurs in medium spiny neurons, the main cell type in the NAc, and, if so, to investigate its properties. We used postnatal rat NAc neurons co-cultured with prefrontal cortex (PFC) neurons from enhanced cyan fluorescent protein (ECFP) mice. PFC neurons restore excitatory synaptic input to NAc neurons but are distinguishable based on fluorescence (Sun et al., 2008). Our results demonstrate that bidirectional synaptic scaling occurs in medium spiny neurons and that activity blockade scales up GluR1/2-containing AMPARs through mechanisms involving protein synthesis and the ubiquitin-proteasome system (UPS). Finally, we found that repeated dopamine (DA) treatment, mimicking repeated exposure to DA-releasing psychostimulants such as cocaine, occludes synaptic scaling after activity blockade.
Materials and methods
All animal use procedures were in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and were approved by the Institutional Animal Care and Use Committee of the Rosalind Franklin University of Medicine and Science. Pregnant Sprague–Dawley rats (Harlan, Indianapolis, IN, USA; Zivic Miller, Pittsburgh, PA, USA), obtained at 18–20 days of gestation, were housed individually in breeding cages. One-day-old [postnatal day (P)1] rat offspring were decapitated and used to obtain NAc neurons. PFC cells were obtained from P1 offspring of homozygous ECFP-expressing mice [strain B6.129(ICR)-Tg(ACTB-ECFP)1Nagy/J; Jackson Laboratory, Bar Harbor, ME, USA]. The homozygous ECFP transgenic mouse strain was maintained by mating ECFP male and female mice in house. All offspring expressed ECFP.
This co-culture system has previously been described in detail (Sun et al., 2008). Briefly, the medial PFC of ECFP-expressing P1 mice was isolated and dissociated with papain (20–25 U/mL; Worthington Biochemical, Lakewood, NJ, USA) at 37°C for 30 min. The PFC cells were plated at a density of 20 000 cells/well onto coverslips coated with poly-d-lysine (100 μg/mL; Sigma-Aldrich, St Louis, MO, USA) in 24-well culture plates. Two to three days later, the NAc from P1 rats was isolated and dissociated with papain (20–25 U/mL) at 37°C for 25 min. NAc cells were plated at a density of 30 000 cells/well with the PFC cells described above. The NAc/PFC co-cultures were grown in Neurobasal medium (Invitrogen, Carlsbad, CA, USA) supplemented with 2 mm GlutaMAX, 0.5% gentamicin and 2% B27 (Invitrogen). Half of the medium was replaced with this Neurobasal growth medium every 4 days. Cultures were used for experiments between weeks 2 and 3.
Co-cultures were either kept in control medium (vehicle) or treated with tetrodotoxin (TTX) (2 μm), 6-cyano-7-nitroquinoxaline-2,3-dione disodium salt (CNQX) (20 μm), d(−)-2-amino-5-phosphonopentanoic acid (APV) (50 μm), (−)bicuculline methiodide (20 μm), MG 132 (10 μm), lactacystin (5 μm), anisomycin (40 μm) or nifedipine (10 μm) for 24–72 h. Drug treatments were refreshed every 24 h if the incubation lasted more than 24 h. All drugs were from Sigma-Aldrich, except bicuculline, which was from Tocris Bioscience (Ellisville, MD, USA). None of these drug treatments significantly affected cell viability (data not shown) assessed with the Live/Dead/Viability/Cytotoxicity Assay (Invitrogen, OR, USA).
Analysis was restricted to processes of NAc medium spiny neurons, which can be distinguished from NAc interneurons by morphology and from ECFP mouse PFC neurons based on fluorescence (Sun et al., 2008). For cell surface GluR1 and GluR2 double immunostaining, live neurons were incubated with polyclonal antibody to the extracellular N-terminal domain of GluR1 (PC246, amino acids 271–285, 1 : 15; Calbiochem, La Jolla, CA, USA) and mouse monoclonal antibody to the extracellular N-terminal domain of GluR2 (MAB397, amino acids 175–430, 1 : 20; Millipore, Billerica, MA, USA) in NeuroBasal media at 37°C for 15 min. Cells were then fixed with 4% paraformaldehyde in phosphate-buffered saline for 15 min at room temperature (RT; 21°C), blocked with 5% donkey serum in phosphate-buffered saline for 1 h and incubated with donkey anti-rabbit secondary antibody conjugated to Cy3 (1 : 500; Jackson ImmunoResearch, West Grove, PA, USA) and donkey anti-mouse secondary antibody conjugated to Alexa 488 (1 : 1000; Invitrogen) for 1 h at RT under non-permeabilized conditions. For cell surface GluR1 and GluR3 double immunostaining, mouse monoclonal antibody to the extracellular N-terminal domain of GluR3 (MAB5416, 1 : 20; Millipore) was used instead of GluR2 antibody with the same protocol.
To assess the synaptic localization of AMPAR subunits, synaptobrevin/vesicle-associated membrane protein 2 (synaptobrevin) and synaptophysin were used as synaptic markers for GluR1 and GluR2 experiments, respectively. Cell surface GluR1 was first labeled with rabbit polyclonal antibody to N-GluR1 (1 : 15; Calbiochem) in NeuroBasal media at 37°C for 15 min. Cells were then fixed with 4% paraformaldehyde, blocked with 5% donkey serum and incubated with donkey anti-rabbit secondary antibody conjugated to Cy3 (1 : 500; Jackson ImmunoResearch) at RT for 1 h under non-permeabilized conditions. Cells were then permeabilized with 0.1% Triton X-100 in phosphate-buffered saline for 15 min, blocked with 5% donkey serum in phosphate-buffered saline for 1 h and incubated with mouse monoclonal antibody to synaptobrevin (1 : 1000; Cat. no. 104 001, Synaptic Systems, Göttingen, Germany) overnight at 4°C followed by donkey anti-mouse secondary antibody conjugated to Alexa 488 (1 : 1000; Invitrogen) for 1 h at RT. For cell surface GluR2 and synaptophysin double immunostaining, mouse monoclonal antibody to N-GluR2 (1 : 20; Millipore) and rabbit polyclonal antibody to synaptophysin (Cat. no. 08-0130, 1 : 1000; Zymed, Carlsbad, CA, USA) [paired with donkey anti-mouse secondary antibody conjugated to Alexa 488 (1 : 1000; Millipore) and donkey anti-rabbit secondary antibody conjugated to Cy3 (1 : 500; Jackson ImmunoResearch), respectively] were used with the same protocol described above for surface GluR1 and synaptobrevin double immunostaining. The specificity of GluR1, GluR2, GluR3, synaptophysin and synaptobrevin antibodies has been confirmed in previous studies (GluR1: Molnár et al., 1993; GluR2: Vissavajjhala et al., 1996; GluR3: Moga et al., 2003; synaptophysin: Jahn et al., 1985; synaptobrevin: Edelmann et al., 1995). We also used additional controls for the immunocytochemical experiments, which consisted of: (i) omitting the primary antibodies and applying the secondary antibodies alone and (ii) comparing double immunostaining with sequential single immunostaining with each antibody. These experiments confirmed that the fluorochromes used in the double-staining experiments did not introduce artifactual fluorescent labeling and there was no cross-reaction between ‘unmatched’ primary and secondary antibodies.
A cobalt-quenching assay using calcein imaging was performed according to methods modified from Sutton et al. (2006). For these experiments, co-cultures were grown on glass-bottomed culture dishes (MatTech, Acworth, GA, USA). Calcein loading and imaging were performed in HEPES-buffered saline (Invitrogen). Neurons were first washed once and then incubated with calcein-AM (10 μm; Invitrogen) for 10 min at RT. Neurons were then washed three times with HEPES-buffered saline to remove residual extracellular dye and incubated in HEPES-buffered saline for 10 min prior to imaging. Calcein fluorescence images were acquired for medium spiny neurons, selected based on morphology, with a Nikon inverted microscope using a 40 × oil objective, an ORCA-ER digital camera and MetaMorph software (Universal Imaging, Downingtown, PA, USA). Immediately after acquisition of the first image, an excess of CoCl2 (2.5 mm; Sigma-Aldrich) was added to the bath and a second image was acquired exactly 5 min later. The images were acquired with identical parameters across all experimental groups. The average intensity of calcein fluorescence before and after cobalt addition was measured and compared in the processes of medium spiny neurons. It should be noted that this method measures quenching in dendritic segments, not selectively at synapses. In pilot studies, we observed that calcein fluorescence is extremely stable over the time period required for our experiments. Previous studies have established that calcein fluorescence is very stable, insensitive to pH and can be quenched rapidly and stoichiometrically by divalent metals such as Fe2+ and Co2+ but not by Cd2+ and Mg2+ (Breuer et al., 1995). To minimize bleaching effects (Sutton et al., 2006), we used a higher concentration of calcein, bright-field optics for selecting neurons and imaged with reduced power of excitation light and minimal exposure time.
Quantification and statistical analysis
Images were acquired with the same system described above for calcein imaging and analysed as described previously (Sun et al., 2005, 2008). All experimental groups to be compared were from the same culture preparation and were processed simultaneously. For each experimental group, cells from at least four different wells were used and approximately 4–6 cells from each well were analysed. Medium spiny neurons adjacent to PFC neurons were selected based on morphology (Sun et al., 2008) under phase contrast imaging to avoid investigator bias based on the intensity of fluorescence staining. For each image, the total area of AMPAR surface staining or the number of AMPAR-immunoreactive puncta in a fixed length of process (15 μm), located at least one soma diameter away from the soma, was measured using a threshold set at least two times higher than average background fluorescence in processes of untreated cells. The soma was excluded for all measurements. The same approach was used to define the area of synaptic marker (synaptobrevin or synaptophysin) staining. The synaptic GluR1 or GluR2 area was defined as the area of subunit staining that overlapped with staining for its respective synaptic marker. The non-synaptic AMPAR subunit area was defined as the area of subunit staining that did not overlap with the synaptic marker. The area of co-localization of GluR1/GluR2 or GluR1/GluR3 was defined using the same approach. The synapses analysed in our study represent PFC–NAc excitatory synapses. We showed previously that culturing PFC neurons with NAc neurons results in the appearance of excitatory synapses onto NAc medium spiny neurons and increases spine formation (Sun et al., 2008). Although NAc–NAc connections are present in the co-cultures, they are excluded from the analysis because they are not glutamatergic. All values are presented as mean ± SEM. For statistical analysis, raw data were analysed with a one-way anova on ranks to compare multiple groups using SigmaStat software (version 2.0; Systat Software Inc., San Jose, CA, USA). When a significant group effect was found, post hoc comparisons were performed using a Dunn’s test to examine specific group differences. Independent group t-tests were used for comparing two groups. The criterion for significance was set at P < 0.05 (n, number of cells analysed). Graphs were produced using SigmaPlot (version 8.0; Systat Software Inc.).
Bidirectional regulation of GluR1 and GluR2 surface expression by synaptic activity
The AMPAR antagonist CNQX and the GABAA receptor antagonist bicuculline have been among the most commonly used agents in previous studies of synaptic scaling (Turrigiano et al., 1998; O’Brien et al., 1998; Liao et al., 1999; Watt et al., 2000; Leslie et al., 2001; Thiagarajan et al., 2002, 2005; Shepherd et al., 2006; Ibata et al., 2008). CNQX has been used to produce long-term inhibition of excitatory synaptic activity, whereas bicuculline has been used to produce the opposite effect by blocking inhibitory GABA transmission. We therefore selected CNQX and bicuculline for our initial studies, although other actions of both drugs should be noted. Bicuculline salts block Ca2+-activated K+ channels (reviewed by Seutin & Johnson, 1999), which could enhance the excitatory effects produced by GABAA receptor blockade. CNQX, in addition to potent antagonistic effects at AMPARs, can reduce N-methyl-d-aspartate receptor (NMDAR) transmission by competing with glycine for its binding site on the NMDAR (Birch et al., 1988; Lester et al., 1989). However, this is unlikely to have contributed to the effects of CNQX in our experiments because blockade of NMDARs did not produce these same effects (see fourth section of Results).
To measure changes in AMPAR surface expression on the processes of medium spiny NAc neurons after incubation with CNQX or bicuculline, we used live-cell labeling with antibodies directed against the extracellular N-terminal regions of GluR1 and GluR2. All experiments were conducted in NAc–PFC co-cultures. We were able to selectively analyse NAc neurons because PFC neurons, but not NAc neurons, express ECFP (see Introduction). Incubation of NAc–PFC co-cultures with CNQX (20 μm, 48 h) significantly increased surface GluR1 expression on the processes of medium spiny NAc neurons (Fig. 1A and B; one-way anova on ranks followed by Dunn’s test, P < 0.05). Conversely, increasing excitatory synaptic activity with bicuculline (20 μm, 48 h) significantly decreased surface GluR1 expression (Fig. 1A and B). Similar effects were observed for surface GluR2 expression (Fig. 1A and B; Dunn’s test, P < 0.05) and the area of GluR1/GluR2 co-localization (Fig. 1A and C; Dunn’s test, P < 0.05). Furthermore, total protein levels of GluR1 and GluR2 subunits in neuronal processes (determined in permeabilized neurons) increased after CNQX and decreased after bicuculline (Fig. 2; one-way anova on ranks followed by Dunn’s test, P < 0.05). CNQX and bicuculline had no effect on cell viability (data not shown; see Materials and methods). These results indicate that medium spiny neurons exhibit bidirectional synaptic scaling of surface and total AMPAR levels in NAc/PFC co-cultures. In subsequent studies, we focused on mechanisms underlying AMPAR upregulation after activity blockade because we hypothesize that this is relevant to AMPAR plasticity after cocaine withdrawal (see Introduction).
AMPA receptor subunit composition after chronic AMPA receptor blockade
The GluR1/2 double-immunostaining results in Fig. 1A–C suggest that GluR1/2-containing AMPARs scale up in NAc neurons but they do not rule out a contribution of GluR2-lacking AMPARs. To address this, we conducted additional analyses and experiments. First, we quantified GluR1-containing/GluR2-lacking receptors and found that neither CNQX nor bicuculline significantly altered their surface expression (Fig. 1D; one-way anova on ranks, P > 0.05). Second, we investigated the effect of activity blockade on GluR3. We did not study GluR4 because it does not contribute to AMPARs in medium spiny neurons (Bernard et al., 1997; Stefani et al., 1998; Conrad et al., 2008). Using live-cell double immunostaining of GluR1 and GluR3, we found that CNQX increased surface GluR1 expression (Fig. 3A and B; t-test, P < 0.05), confirming the findings in Fig. 1A and B. However, CNQX did not increase surface GluR3 expression (Fig. 3A and B; t-test, puncta number, P = 0.24; puncta area, P = 0.28) or the area of GluR1/GluR3 co-localization (Fig. 3C; t-test, puncta number, P = 0.41; puncta area, P = 0.25). Furthermore, total protein expression of GluR3 in the processes of permeabilized neurons was not altered by CNQX or bicuculline (Fig. 2; one-way anova on ranks, P = 0.70). Although these results suggest that GluR2-lacking AMPARs do not scale up after CNQX, two factors limit the ability to deduce subunit composition from double-labeling experiments. First, because antibodies differ in affinity for their target, staining with different antibodies cannot be directly compared. Second, we do not know how many GluR molecules must be present in a puncta for it to exceed our threshold for detection; thus it is possible that we failed to detect some GluR subunits.
To address these concerns, we used a cobalt-quenching assay (Dravid & Murray, 2003; Sutton et al., 2006) to further examine whether chronic AMPAR blockade increases surface GluR2-lacking receptors. This assay exploits the ability of Co2+ to stoichiometrically quench the fluorescent dye calcein (Breuer et al., 1995). It enables the detection of GluR2-lacking AMPARs because they are permeable to Ca2+ (Cull-Candy et al., 2006; Liu & Zukin, 2007; Isaac et al., 2007) and, unlike other Ca2+-permeable channels such as NMDARs or voltage-gated Ca2+ channels, they exhibit high permeability to Co2+ (Pruss et al., 1991). Previous studies provided direct evidence that Co2+ quenching is blocked by acute application of the AMPAR antagonist 2,3-dioxo-6-nitro-1,2,3,4-tetrahydrobenzo[f]quinoxaline-7-sulfonamide (NBQX; Dravid & Murray, 2003) or 1-naphthylacetylsperimine, a selective antagonist of GluR2-lacking AMPARs (Sutton et al., 2006). By measuring the quenching of intracellular calcein fluorescence in neuronal processes, we examined whether CNQX pretreatment (20 μm, 48 h) increased Co2+ influx through GluR2-lacking receptors. A baseline image was acquired after a brief period of calcein dye loading. Immediately afterwards, an excess of extracellular CoCl2 was added to the bath and a second image was acquired exactly 5 min later. There was a small decrease (6 ± 0.6%) in calcein fluorescence in the control group, indicating a basal level of Co2+ influx. This suggests the presence of some GluR2-lacking AMPARs in NAc neurons under control conditions. After CNQX treatment, the decrease in calcein fluorescence at 5 min after Co2+ addition (7 ± 0.8%) was similar in magnitude to that observed in the control group (Fig. 4A and B; t-test, P = 0.36), suggesting that CNQX did not increase the number of GluR2-lacking AMPARs. Additional support for this conclusion was obtained with immunocytochemical methods. Under basal conditions, 43 ± 3% of surface GluR1 was not co-localized with GluR2 and 84 ± 6% of surface GluR1 was not co-localized with GluR3, suggesting that a significant portion of GluR1 exists in homomeric GluR1 receptors (n = 30 neurons). After CNQX treatment, about the same fraction of surface GluR1 was not co-localized with GluR2 (35 ± 4%) or with GluR3 (87 ± 4%) (n = 26 neurons). Combined with our findings that surface GluR1, GluR2 and GluR1/2 co-localization increases after CNQX (Fig. 1A–C), whereas GluR1-containing/GluR2-lacking receptors (Fig. 1D) and GluR1/3 co-localization (Fig. 3) do not, these results indicate that GluR2-lacking AMPARs exist under basal conditions but do not contribute significantly to the increase in AMPAR transmission produced by CNQX. Instead, CNQX induces increased surface expression of GluR1/2-containing AMPARs.
6-Cyano-7-nitroquinoxaline-2,3-dione (CNQX) increases GluR1 and GluR2 synaptic expression
Having measured GluR1 and GluR2 surface expression (Fig. 1), we next examined their synaptic expression. Live-cell labeling of either GluR1 or GluR2 was performed, followed by fixation and immunostaining for a synaptic marker under permeabilized conditions. The synaptic markers synpatobrevin and synaptophysin were used for GluR1 and GluR2, respectively. CNQX (20 μm, 48 h) significantly increased both synaptic and non-synaptic GluR1 (Fig. 5A and C; t-test, P < 0.05), as well as synaptic and non-synaptic GluR2 (Fig. 5B and D; t-test, P < 0.05). CNQX treatment did not significantly alter the expression of synaptobrevin or synaptophysin, although trends towards increased puncta size were observed (data not shown).
We could not evaluate AMPAR subunit co-localization within synapses by using triple staining because all available fluorophores were used (cyan for PFC cells, red and green for labeling AMPAR subunit pairs). However, the results presented in Figs 1–4 indicate that GluR1/2-containing AMPARs are added to synapses.
Upregulation of surface GluR1 and GluR2 by prolonged activity blockade is independent of N-methyl-d-aspartate receptors and l-type Ca2+ channels
Two studies in hippocampal neurons reported that the combination of TTX and the NMDAR antagonist APV increased the expression of GluR2-lacking AMPARs, although the time-course differed between these studies (Ju et al., 2004; Sutton et al., 2006). Therefore, we tested the effects of TTX and APV in NAc neurons using live-cell double immunostaining of GluR1 and GluR2. As observed for CNQX, TTX (2 μm, 72 h) increased the surface expression of GluR1 and GluR2 (Fig. 6A; one-way anova on ranks followed by Dunn’s test, P < 0.05). TTX + CNQX exerted a similar effect (Fig. 6A; Dunn’s test, P < 0.05 compared with control, P > 0.05 compared with TTX group). Time-course studies showed that significant increases in GluR1 and GluR2 surface expression were first observed after 24 h of TTX treatment but not after shorter incubations (1 or 4 h) (data not shown). Unlike CNQX or TTX, APV (50 μm, 72 h) did not alter GluR1 or GluR2 surface expression (Fig. 6A; Dunn’s test, P > 0.05 compared with control). Furthermore, co-incubation with APV did not influence the ability of TTX to induce scaling up of surface GluR1 and GluR2 (Fig. 6A; Dunn’s test, P > 0.05 TTX + APV compared with TTX group). No neurotoxic effects were observed after incubation with TTX or APV (data not shown). Thus, prolonged activity blockade leads to an NMDAR-independent increase in surface GluR1 and GluR2 and, in NAc neurons, combining TTX with NMDAR blockade does not result in scaling up of GluR2-lacking AMPARs. Furthermore, the fact that AMPAR scaling was produced by CNQX but not APV demonstrates that this effect of CNQX was due to blockade of AMPARs, rather than inhibition of NMDAR transmission through interaction with the glycine modulatory site (see first section of Results). This is consistent with evidence that 20 μm CNQX acts primarily on AMPARs in NAc medium spiny neurons recorded from postnatal rats (Zhang & Warren, 2008).
Considerable evidence implicates Ca2+ as the sensor that links neuronal activity to synaptic scaling (Ibata et al., 2008; Turrigiano, 2008). In hippocampal cultures, l-type Ca2+ channel blockade with nifedipine (10 μm) mimicked the effect of activity blockade on AMPAR transmission (Thiagarajan et al., 2005). We found that the same concentration of nifedipine did not alter surface GluR1 and GluR2 expression (Fig. 6B; one-way anova on ranks followed by Dunn’s test, P > 0.05 compared with control) or the increased GluR1 and GluR2 surface expression induced by AMPAR blockade with CNQX (Fig. 6B; Dunn’s test, P > 0.05, CNQX group compared with nifedipine + CNQX group). This suggests that decreased activation of l-type Ca2+ channels is not a major mechanism underlying effects of activity blockade in NAc neurons. It should be noted that the effects of nifedipine on l-type Ca2+ channels include a component of use-dependent block (e.g. Shen et al., 2000), so its efficacy in different culture systems may be influenced by differences in the level of spontaneous activity. Extensive future studies will be required to thoroughly address the role of Ca2+ and its downstream signaling pathways in synaptic scaling in NAc neurons (see Discussion).
Upregulation of surface GluR1 and GluR2 by prolonged activity blockade involves protein synthesis and the ubiquitin-proteasome system
Activity blockade has been reported to enhance the synthesis of AMPAR subunits in hippocampal and cortical neurons (Ju et al., 2004; Sutton et al., 2006; Ibata et al., 2008). To evaluate the role of protein synthesis in effects of activity blockade on NAc neurons, NAc/PFC co-cultures were treated with the protein synthesis inhibitor anisomycin (40 μm, 24 h) alone or with CNQX. Anisomycin dramatically decreased basal surface GluR1 and GluR2 expression, suggesting that protein synthesis is required to maintain the surface pool of AMPARs (Fig. 7A; one-way anova on ranks followed by Dunn’s test, P < 0.05). In combination with CNQX, anisomycin blocked the effect of CNQX and decreased surface GluR1 and GluR2 expression to the same extent as observed in cells treated with anisomycin alone (Fig. 7A; Dunn’s test, P < 0.05 anisomycin + CNQX compared with CNQX group, P > 0.05 anisomycin + CNQX compared with anisomycin group). These results support the idea that protein synthesis is necessary for increased AMPAR surface expression after activity blockade, although we did not determine the location of new protein synthesis.
Next, we examined the possibility that the UPS also contributes to increased AMPAR surface expression after activity blockade. Supporting this possibility, synaptic activity controls the molecular composition of the postsynaptic density and its signaling output via the UPS; activity blockade is associated with decreased ubiquitin conjugation (Ehlers, 2003). Combined with evidence that AMPAR internalization is dependent on UPS activity (Patrick et al., 2003; Colledge et al., 2003; Bingol & Schuman, 2004), these results suggest that activity blockade, by reducing UPS activity, may prevent AMPAR internalization and thus lead to increased surface AMPAR expression. We tested this by treating NAc/PFC co-cultures with either of two structurally distinct proteasome inhibitors, MG 132 (10 μm, 24 h) and lactacystin (5 μm, 24 h). Both inhibitors increased surface GluR1 and GluR2 expression to the same degree as CNQX (Fig. 7B; one-way anova on ranks followed by Dunn’s test, P < 0.05 compared with control group, P > 0.05 compared with CNQX group). In addition, the effects on AMPAR surface expression induced by proteasome inhibition and CNQX occluded one another (Fig. 7B; Dunn’s test, P > 0.05 MG 132 + CNQX or lactacystin + CNQX compared with either drug alone). Together with the anisomycin results (Fig. 7A), this suggests that activity blockade may increase AMPAR surface expression through a mechanism that requires both ongoing protein synthesis and inhibition of UPS-mediated AMPAR endocytosis (see Discussion).
Another possible contributor to increased AMPAR surface expression after activity blockade is decreased lysosomal degradation. To test this, NAc/PFC co-cultures were treated with a lysosomal inhibitor, leupeptin (10 μm, 24 h). Leupeptin had no significant effect on surface AMPAR subunit expression on its own and failed to alter the effect of CNQX (Fig. 7C; one-way anova on ranks followed by Dunn’s test, P > 0.05 compared with control and P < 0.05, Leup + CNQX group compared with control group). These results do not support a contribution of lysosome-mediated degradation to synaptic scaling.
Repeated dopamine treatment increases surface GluR1 and GluR2 expression with or without activity blockade
We hypothesized that AMPAR upregulation after cocaine withdrawal may occur in response to hypoactivity of excitatory inputs to NAc neurons during withdrawal, leading to synaptic scaling (see Introduction). In the incubation model of cocaine addiction, upregulation of GluR1/2-containing AMPARs is observed (Conrad et al., 2008). However, the present results indicate that these receptors do not participate in synaptic scaling in postnatal NAc neurons. This prompted us to wonder if prior cocaine exposure, known to produce profound adaptations in NAc neurons (Robinson & Kolb, 2004; Kalivas & Hu, 2006; Peoples et al., 2007; McClung & Nestler, 2008), may alter the properties of synaptic scaling in NAc neurons to promote participation of GluR2-lacking AMPARs.
To test this, we treated NAc/PFC co-cultures repeatedly with vehicle or DA (1 μm for 30 min) on days 7, 9 and 11 in vitro. Then, beginning on day 12, cells were incubated in media containing vehicle, TTX (2 μm) or CNQX (20 μm). After ∼36 h of incubation with these drugs, GluR1 and GluR2 surface expression were measured on day 15. Thus, four groups were formed: vehicle/vehicle, DA/vehicle, DA/TTX and DA/CNQX. The DA/vehicle group was intended to model repeated cocaine treatment followed by withdrawal in the presence of normal synaptic activity. The DA/TTX and DA/CNQX groups were intended to model repeated cocaine treatment followed by hypoactivity of excitatory inputs during withdrawal. An obvious caveat is that in-vitro treatments are limited in their ability to reproduce the complex circuitry and longer duration of in-vivo models of addiction.
In the DA/vehicle group, we observed increased surface expression of both GluR1 and GluR2 compared with vehicle/vehicle controls, indicating upregulation of GluR1/2-containing AMPARs (Fig. 8; one-way anova on ranks followed by Dunn’s test, P < 0.05). The GluR1 result replicated a previous study, in which we also showed that increased GluR1 surface expression on day 15 was not a persistent response to the final DA exposure on day 11 but instead required Ca2+/calmodulin-dependent protein kinase (CaMK) activation during the 3 day withdrawal period (Sun et al., 2008). Thus, this phenomenon is distinct from the acute increase in GluR1 surface expression on NAc medium spiny neurons observed immediately after brief (5–15 min) D1 receptor stimulation (Chao et al., 2002; Mangiavacchi & Wolf, 2004), although it could be related to the ability of stronger D1 receptor stimulation to increase GluR1 surface expression via a protein synthesis-dependent mechanism in higher density hippocampal cultures (Smith et al., 2005).
We predicted that prior DA treatment would alter the expression mechanisms of TTX- or CNQX-induced synaptic scaling, perhaps by enabling participation of GluR2-lacking AMPARs. This outcome would support the hypothesis that synaptic scaling contributes to upregulation of GluR2-lacking AMPARs in the incubation model (Conrad et al., 2008). Instead, the DA/TTX and DA/CNQX groups exhibited no further elevation in AMPAR subunit surface expression compared with the DA/vehicle group (Fig. 8; one-way anova on ranks followed by Dunn’s test, P > 0.05). One possible explanation is that repeated DA treatment caused AMPAR surface expression to reach a ceiling level. However, we showed previously that the same repeated DA paradigm does not prevent subsequent increases in AMPAR surface expression in response to protein kinase A activation (Sun et al., 2008), thus ruling out this explanation. We conclude that there is an interaction between cellular events triggered by repeated DA treatment and those that mediate synaptic scaling.
Using NAc/PFC co-cultures, we show that medium spiny GABAergic neurons of the NAc exhibit bidirectional synaptic scaling, assessed by monitoring postsynaptic AMPAR levels. Chronic activity manipulation has been demonstrated to alter postsynaptic AMPAR accumulation in cultured spinal neurons (O’Brien et al., 1998), cortical neurons (Turrigiano et al., 1998; Wierenga et al., 2005; Gainey et al., 2009) and hippocampal neurons (Lissin et al., 1998; Liao et al., 1999; Ju et al., 2004; Thiagarajan et al., 2005). In NAc neurons, chronic activity blockade (24–72 h) increased surface levels (synaptic and extrasynaptic) of GluR1 and GluR2 and increased GluR1/2 co-localization on the cell surface. Total protein levels of GluR1 and GluR2 also increased. Conversely, increasing activity with bicuculline (48 h) decreased surface and total expression of GluR1/2-containing receptors. GluR3 expression was not altered by activity manipulations. These results indicate that GluR1/2-containing AMPARs undergo synaptic scaling in NAc neurons.
Role of protein synthesis and the ubiquitin-proteasome system
Increased GluR1 and GluR2 levels in NAc neurons after chronic activity blockade were dependent on protein synthesis, as reported previously in other cell types (Ju et al., 2004; Sutton et al., 2006; Ibata et al., 2008). In addition, UPS inhibition occluded the activity block-induced increase in surface GluR1 and GluR2, consistent with studies in spinal neurons showing that increased synaptic incorporation of GluR1 after chronic activity blockade was due to an increase in the metabolic half-life of GluR1 (O’Brien et al., 1998). More generally, activity blockade has been shown to decrease UPS activity and thereby regulate the composition of the postsynaptic density (Ehlers, 2003). These studies, as well as our own work, suggest that both protein synthesis and decreased UPS activity contribute to AMPAR upregulation after CNQX. This is consistent with accumulating evidence for complicated interplay between protein synthesis and the UPS pathway under both physiological and pathological conditions (Ding et al., 2007). For example, proteasome inhibition enhances the induction of late-phase long-term potentiation through a mechanism that may involve stabilizing newly translated proteins, as the enhancing effect of proteasome inhibition is reversed by anisomycin (Dong et al., 2008). Further complicating the matter, the identity of the UPS target(s) that regulates AMPAR endocytosis is unclear (Patrick et al., 2003; Colledge et al., 2003; Bingol & Schuman, 2004) and it remains possible that direct ubiquitination of AMPARs is involved (see Zhang et al., 2009 for discussion). Additional studies are needed to determine how protein synthesis and the UPS interact during synaptic scaling. Finally, as shown in hippocampal and cortical neurons (Ju et al., 2004; Ehlers, 2003), lysosomal degradation does not contribute to AMPAR scaling in NAc neurons.
AMPA receptor subunit composition
The subunit composition of AMPARs undergoing synaptic scaling was of interest to us based on a role for GluR2-lacking AMPARs in the incubation model of cocaine addiction (see Introduction). Unlike the prevalent GluR1/2 or GluR2/3 receptors, GluR2-lacking AMPARs exhibit permeability to Ca2+ resulting in higher conductance, so their synaptic incorporation changes the properties of excitatory transmission and subsequent plasticity (Cull-Candy et al., 2006; Thiagarajan et al., 2007; Liu & Zukin, 2007; Isaac et al., 2007).
In the NAc of adult rats (P60–120), our biochemical results indicate that GluR2-lacking receptors comprise ∼10% of the total AMPARs (Boudreau et al., 2007; Reimers et al., 2007; Conrad et al., 2008), whereas electrophysiological results indicate that they make a very small contribution to synaptic currents in medium spiny neurons (∼5% of evoked excitatory postsynaptic current) (Conrad et al., 2008). No detectable contribution to synaptic currents was observed in NAc medium spiny neurons of mice (P24–28) (Kourrich et al., 2007). It is possible that GluR2-lacking AMPARs contribute more significantly to Ca2+ signaling than excitatory currents. At least in young rats (P15–19), GluR2-lacking AMPARs provide an important source of Ca2+ entry into dorsal striatal medium spiny neurons (Carter & Sabatini, 2004).
In the postnatal NAc neurons used herein, we detected basal expression of GluR2-lacking AMPARs using both immunocytochemistry and a cobalt-quenching assay. From immunocytochemical results, we estimate that as much as one-third of surface-expressed GluR1 may be present in GluR1 homomers. The significant basal expression of GluR2-lacking AMPARs in cultured neurons prepared from very young brains is consistent with a previous functional study of embryonic mouse striatal cultures (Perkinton et al., 1999) and more generally with results indicating a higher prevalence of GluR2-lacking AMPARs in principal neurons early in development (e.g. Kumar et al., 2002; Ho et al., 2007).
Despite the fact that GluR2-lacking AMPARs were detected in NAc neurons under basal conditions, activity blockade did not affect their surface expression but instead increased GluR1/2-containing AMPARs (see above). Scaling of both GluR1 and GluR2 was similarly observed in cultured spinal and cortical neurons (O’Brien et al., 1998; Wierenga et al., 2005), whereas scaling up of GluR2-lacking AMPARs was found in hippocampal cultures (Ju et al., 2004; Thiagarajan et al., 2005; Sutton et al., 2006). This may reflect a difference between brain regions. Alternatively, some results suggest that GluR1/2 receptors scale up when postsynaptic firing is blocked (AMPAR antagonist or TTX), whereas GluR2-lacking AMPARs scale up if this is combined with NMDAR blockade (Sutton et al., 2006; Turrigiano, 2008; Gainey et al., 2009). However, this scheme does not fit all results. Sutton et al. (2006) found that TTX + APV increased GluR2-lacking AMPARs after short incubations (3–12 h) but these were replaced by GluR2-containing AMPARs after 24 h in TTX + APV. Thiagarajan et al. (2005) detected scaling of GluR2-lacking AMPARs after 20–30 h in 2,3-dioxo-6-nitro-1,2,3,4-tetrahydrobenzo[f]quinoxaline-7-sulfonamide (NBQX) alone, although TTX was present in the recording solution. We observed scaling up of GluR1/2-containing AMPARs regardless of whether activity blockade was achieved with CNQX, TTX, CNQX + TTX or APV + TTX. Interestingly, although GluR3 did not undergo synaptic scaling in NAc neurons, its surface expression increased in cortical neurons during synaptic scaling, suggesting that GluR3 contributes to synaptic scaling in these cells, although to a lesser degree than GluR1 or GluR2 (Gainey et al., 2009). Many factors may influence the AMPAR population that undergoes synaptic scaling.
Relationship between synaptic scaling and cocaine action
We were interested in the possibility that synaptic scaling might contribute to the AMPAR upregulation observed in two animal models of cocaine addiction. In the first model, rodents receive repeated intraperitoneal injections of cocaine leading to behavioral sensitization; when sensitized animals are killed after 1–3 weeks of withdrawal from cocaine, they exhibit increased surface and synaptic expression of GluR1/2-containing AMPARs (Boudreau & Wolf, 2005; Boudreau et al., 2007; Kourrich et al., 2007; Boudreau et al., 2009). Small increases in total GluR1 levels have also been reported (Churchill et al., 1999; Scheggi et al., 2002). In the incubation model, rats self-administer cocaine for 6 h/day for 10 days and each injection is paired with a cue. Cue-induced cocaine seeking intensifies or ‘incubates’ over the first weeks of withdrawal (Grimm et al., 2001) in concert with the addition of GluR2-lacking AMPARs to NAc synapses (Conrad et al., 2008).
The present results, showing scaling of GluR1/2-containing AMPARs in NAc/PFC co-cultures, are consistent with a possible role for synaptic scaling in the behavioral sensitization model. Supporting this, one study has demonstrated decreased metabolic activity of brain regions sending excitatory projections to the NAc (and the NAc itself) after experimenter-administered cocaine (a rat study; Hammer & Cooke, 1994). However, all other evidence for hypoactivity has been obtained from studies of cocaine self-administration (humans: Volkow et al., 1992; Goldstein & Volkow, 2002; primates: Porrino et al., 2004; Beveridge et al., 2006; Porrino et al., 2007; rats: Hammer et al., 1993; Macey et al., 2004; Sun & Rebec, 2006). These latter results suggest that synaptic scaling is a potential explanation for the upregulation of GluR2-lacking AMPARs in the incubation model, which utilizes cocaine self-administration (Conrad et al., 2008). However, only GluR1/2-containing AMPARs were found to undergo synaptic scaling in the present study. It is possible that postnatal NAc neurons (present study) and adult NAc neurons (incubation model; Conrad et al., 2008) differ in the expression mechanisms of synaptic scaling, with only the latter involving GluR2-lacking AMPARs. There is a precedent for developmental changes in expression mechanisms (Wierenga et al., 2006). Furthermore, in-vivo scaling of GluR2-lacking AMPARs occurs in the sensory cortices of 4–5-week-old rats (Goel et al., 2006). Another possibility is that neuroadaptations produced in the NAc by the long-access cocaine self-administration regimen that leads to incubation alter the expression mechanisms of synaptic scaling to favor synaptic accumulation of GluR2-lacking AMPARs. For example, NAc levels of brain-derived neurotrophic factor (BDNF) are increased after the same cocaine regimen that produces incubation of cocaine craving and GluR2-lacking AMPARs (Grimm et al., 2003; see also Graham et al., 2007). In other cell types, BDNF promotes plasticity at glutamatergic synapses through multiple mechanisms, including increased GluR1 synthesis and synaptic delivery of homomeric GluR1 receptors (Caldeira et al., 2007; Li & Keifer, 2008). By increasing the abundance of GluR2-lacking AMPARs, BDNF may favor their involvement in synaptic scaling. This proposed modulatory role is distinct from BDNF’s role as a mediator of synaptic scaling in the cortex, where pyramidal neurons produce and release BDNF in an activity-dependent manner and decreased BDNF after activity blockade is responsible for scaling up (Rutherford et al., 1998).
We conducted experiments to test the idea that repeated DA exposure prior to activity blockade might alter the properties of synaptic scaling. NAc/PFC co-cultures received three DA treatments (30 min on days 7, 9 and 11 in vitro) followed by a 3 day ‘withdrawal period’ (days 12–15). On day 15, we observed an increase in the levels of GluR1 and GluR2 on the surface of NAc medium spiny neurons. Additional cultures received the same repeated DA treatment, followed by CNQX or TTX on days 12–15 to mimic the hypoactivity of excitatory inputs to NAc neurons after cocaine withdrawal. We predicted additional upregulation of AMPARs due to synaptic scaling, perhaps involving GluR2-lacking AMPARs. Contrary to our hypothesis, the increased surface expression of GluR1/2-containing receptors produced by repeated DA treatment occluded the increase normally observed after chronic activity blockade. This could mean that they share a common mechanism. Alternatively, repeated DA treatment may produce cellular changes that preclude synaptic scaling. The occlusion between repeated DA treatment and synaptic scaling observed in vitro does not rule out a role for synaptic scaling in AMPAR upregulation after in-vivo cocaine treatment. Repeated DA treatment in vitro cannot possibly reproduce the many circuit-level changes elicited by repeated systemic cocaine administration. Even if occlusion occurs in vivo, it may be relieved at longer withdrawal times.
The major significance of these in-vitro findings is that they demonstrate that there is an interaction between expression mechanisms of synaptic scaling and mechanisms engaged by repeated DA receptor stimulation, which implies that synaptic scaling may not function normally after cocaine exposure. It is possible that this interaction is mediated by Ca2+ signaling. In a previous study using the same DA treatment regimen, we also demonstrated increased GluR1 surface expression on medium spiny neurons on day 15 (GluR2 was not examined). Furthermore, we showed that this was associated with increased CaMKII phosphorylation and was blocked by inclusion of the CaMK inhibitor KN-93 in the media during ‘withdrawal’ (days 12–15; Sun et al., 2008). CaMK inhibition also prevented activity blockade from increasing excitatory transmission in hippocampal neurons (Thiagarajan et al., 2002). A recent study by Ibata et al. (2008) suggested that CaMKIV, rather than CaMKII, is the Ca2+ sensor for synaptic scaling. They found that AMPAR synaptic accumulation after activity blockade resulted from decreased somatic Ca2+ influx, leading to decreased CaMKIV activation and a resultant decrease in transcription (Ibata et al., 2008). In contrast, CaMK inhibition did not prevent increased synaptic insertion of AMPARs at chronically inhibited single hippocampal synapses (Hou et al., 2008). Future experiments should compare the roles of different CaMKs in regulating AMPAR expression after cocaine withdrawal in vivo and after activity blockade in NAc cultures, and investigate additional mechanisms implicated in synaptic scaling in other brain regions (e.g. Stellwagen & Malenka, 2006; Shepherd et al., 2006;Cingolani et al., 2008; Hou et al., 2008).
Rapidly induced forms of plasticity like long-term potentiation do not provide a satisfying explanation for slowly developing plasticity during drug withdrawal. Here we show for the first time that neurons of the NAc, a critical component of the reward circuit, undergo synaptic scaling in response to prolonged changes in the level of excitatory drive. Furthermore, the mechanisms underlying synaptic scaling interact with adaptive mechanisms triggered by repeated DA receptor stimulation. This suggests that repeated exposure to psychomotor stimulants may alter synaptic scaling in the NAc. Indeed, a different form of homeostatic plasticity in the NAc was recently shown to be prevented by prior cocaine exposure (Ishikawa et al., 2009).
Note added in proof
After the present manuscript was accepted for publication, a paper was published showing that 10 daily cocaine injections to young mice (P16–35) followed by 35 days of withdrawal resulted in the presence of G1uR2-lacking AMPARs in the NAc shell (Mameli et al., 2009). This contrasts with the addition of G1uR2-containing AMPARs at earlier withdrawal times (7–21 days) in the NAc of cocaine-sensitized young mice (Kourrich et al., 2007; injections began P24–28) and adult rats (Boudreau & Wolf, 2005; Boudreau et al., 2007; injections began P60-70). It will be important to determine if cocaine-sensitized rats express GIuR2-lacking AMPARs after longer withdrawal periods, if this occurs when cocaine is administered in adulthood, and whether these receptors are also added in the NAc core.
This work was supported by USPHS grants DA015835 and DA00453 (M.E.W.). We thank Dr Michael A. Sutton for advice on conducting cobalt-quenching assays.
brain-derived neurotrophic factor
Ca2+/calmodulin-dependent protein kinase
6-cyano-7-nitroquinoxaline-2,3-dione disodium salt