|Home | About | Journals | Submit | Contact Us | Français|
Homeostatic synaptic scaling calibrates neuronal excitability by adjusting synaptic strengths during prolonged changes in synaptic activity. The molecular mechanisms that regulate the trafficking of AMPA receptors (AMPARs) during synaptic scaling are largely unknown. Here we show that chronic activity blockade reduces PICK1 protein level on a time scale that coincides with the accumulation of surface AMPARs. PICK1 loss of function alters the subunit composition and the abundance of GluA2-containing AMPARs. Due to aberrant trafficking of these receptors, the increase in synaptic strength in response to synaptic inactivity is occluded in neurons generated from PICK1 knockout mouse. In agreement with electrophysiological recordings, no defect of AMPAR trafficking is observed in PICK1 knockout neurons in response to elevated neuronal activity. Overall, our data reveal an important role of PICK1 in inactivity-induced synaptic scaling by regulating the subunit composition, abundance and trafficking of GluA2-containing AMPARs.
The AMPA-type glutamate receptor mediates the majority of fast excitatory synaptic transmission and its trafficking into and out of the synapse plays a critical role in synaptic plasticity (Shepherd and Huganir, 2007). Hebbian forms of synaptic plasticity, such as long-term potentiation (LTP) and long-term depression (LTD), have been proposed as a physiological correlate of learning and memory. However, this type of plasticity exerts a powerful long-term destabilizing influence on neural circuits. Homeostatic plasticity maintains a stable neuronal circuit by adjusting synaptic properties to keep activity close to the internal target firing range (Turrigiano and Nelson, 2004; Davis, 2006). One such adaptation involves a cell-wide adjustment of postsynaptic AMPA receptors (AMPARs), a process termed synaptic scaling (Turrigiano and Nelson, 2004; Turrigiano, 2008). Synaptic scaling has been observed in a variety of central synapses upon chronic decrease or increase in activity in dissociated neuronal cultures (O’Brien et al., 1998; Turrigiano et al., 1998), as well as in intact animals in response to developmental or activity-dependent changes in sensory inputs (Desai et al., 2002; Maffei et al., 2006; Goel and Lee, 2007). Recently, several molecules such as TNF-α (Stellwagen and Malenka, 2006), Arc (Shepherd et al., 2006), β3-integrin (Cingolani et al., 2008), Plk2 (Seeburg et al., 2008) and retinoic acid (Aoto et al., 2008) have been shown to influence synaptic scaling. However, the molecular mechanisms underlying the trafficking of AMPARs during synaptic scaling remain elusive.
The BAR (Bin/amphiphysin/Rvs) and PDZ (PSD-95/Dlg/ZO1) domain containing protein PICK1 (protein interacting with C-kinase 1) directly binds to GluA2 subunit of AMPARs (Dev et al., 1999; Xia et al., 1999; Xu and Xia, 2006). PICK1 plays important roles in AMPAR surface expression, trafficking and synaptic targeting (Xu and Xia, 2006; Hanley, 2008). Several forms of Hebbian plasticity, such as hippocampal and cerebellar LTD are abolished in PICK1 knockout animals, most likely due to impaired internalization, recycling or intracellular retention of GluA2-containing AMPARs (Steinberg et al., 2006; Terashima et al., 2008; Volk et al., 2010). In contrast to the established role of PICK1 in Hebbian plasticity, the role of PICK1 in homeostatic plasticity is currently unknown.
Here, we tested whether PICK1 participates in homeostatic plasticity by investigating how loss of PICK1 function affects bidirectional synaptic scaling induced by chronic tetrodotoxin (TTX) or bicuculline treatment. We found that TTX treatment reduces PICK1 protein levels on a time scale that coincides with the accumulation of surface AMPARs. Electrophysiological recordings of AMPAR-mediated miniature excitatory postsynaptic current (mEPSC) reveal that TTX-induced synaptic scaling is occluded in PICK1 knockout or PICK1 knockdown neurons due to altered subunit composition and aberrant trafficking of GluA2-containing AMPARs. Interestingly, bicuculline-induced synaptic scaling is normal in PICK1 knockout neurons. Overall, our data reveal an important role of PICK1 in inactivity-induced synaptic scaling by regulating the subunit composition, abundance and trafficking of GluA2-containing AMPARs.
PICK1 shRNAs were cloned into bicistronic pSuper-Venus construct (Hayashi-Takagi et al., 2010), where Venus and shRNA expression was driven by the CMV and the H1 RNA polymerase III promoters, respectively. The PICK1 shRNA targeting sequences were as follows: sh#1, 5′-GCCTCACCATCAAGAAGTACC-3′ and sh#2, 5′-AAAGTACTATAATGACTGCTAT-3′. The PICK1 shRNA#2 sequence was reported previously (Rocca et al., 2008). The efficiency of these shRNA constructs was tested in HEK cells overexpressing Myc-PICK1. The best sequence (sh#1) was subsequently cloned into FuGW vector for lentiviral production as described previously and was the one used throughout the study (Takamiya et al., 2008).
Homozygote PICK1 knockout mice (Gardner et al., 2005) were obtained by mating heterozygote parents. High-density cortical neurons were prepared from PICK1 knockout pups and WT littermates (both males and females) at postnatal day 0 and plated onto poly-L-lysine coated culture dish. Neurons were maintained in glial-conditioned Neurobasal medium supplemented with 1% horse serum, 2% B-27, 2 mM Glutamax, and fed twice a week. Young neurons between days in vitro (DIV) 11–13 were used throughout experiments to minimize the contribution of presynaptic homeostatic plasticity (Wierenga et al., 2006; Han and Stevens, 2009). To induce synaptic scaling, neurons at DIV 9–11 were treated with 2 μM TTX or 40 μM bicuculline for 48 h, unless otherwise noted. To determine the pathway by which PICK1 is degraded, neurons were treated with 2 μM TTX in the presence or absence of the proteosome inhibitor lactacystin (0.5 μM) or the lysosomal inhibitors leupeptin (50 μg/mL) for 48 h. For PICK1 knockdown experiments, cortical neurons were prepared from E18 C57BL/6 mouse embryos, transfected at DIV 8 and treated with 2 μM TTX or 40 μM bicuculline at DIV 9 for 48 h. All neuronal culture reagents and drugs were obtained from Invitrogen and Tocris, respectively.
Neurons were transfected at DIV 8 with either pSuper-Venus or pSuper-Venus-PICK1 sh#1 using Lipofectamine 2000 (Invitrogen) for 72 h. Neurons were fixed with 4% paraformaldehyde/4% sucrose in PBS for 15 min, permeabilized with 0.25% Triton X-100 in PBS and incubated with 10% BSA for 1 h. Neurons were then incubated with anti-PICK1 primary antibody (JH2906) followed by Alexa-568-conjugated goat-anti-rabbit secondary antibody (Invitrogen). Images were collected with a 63X oil-immersion objective on a Zeiss LSM510 confocal microscope for both green (Venus) and red (PICK1) channels. Series of optical sections were collected at 0.38 μm intervals, and maximal intensity projection was shown.
Neurons were washed twice with ACSF (in mM, 25 HEPES, 120 NaCl, 5 KCl, 2 CaCl2, 2 MgCl2, 30 D-glucose, pH 7.4) and incubated with 1 mg/mL Sulfo-NHS-SS-Biotin (Pierce) for 30 min on ice. Free biotin was quenched by washing cells twice with ice-cold 50 mM glycine (pH 7.4 in ACSF). Cultures were lysed and sonicated in RIPA buffer and incubated with Neutravidin beads (Pierce) for 3 h at 4°C. Beads were washed three times, eluted with 2X SDS sample buffer followed by Western blotting analyses. Specific antibodies against GluA1 (4.9D), GluA2 (JH4297 or 6A), GluA3 (JH4300) and PICK1 (JH2906) were generated in-house. Monoclonal antibodies against α-tubulin and β-actin were purchased from Sigma.
Neurons were incubated with 1 mg/mL Sulfo-NHS-SS-Biotin (Pierce) for 30 min on ice and free biotin was quenched by washing neurons with ice-cold TBS (in mM, 50 Tris-HCl, 150 NaCl, pH 7.4). Neurons were lysed in RIPA buffer and the total protein concentration was determined by BCA assay. Two-hundred microgram of neuronal lysate was incubated overnight with 10 μg of anti-GluA2/3 (JH3724) or anti-GluA1 (JH4297) antibodies. The unbound fraction was subjected to another round of immunoprecipitation with the same antibodies for another 6 h at 4°C. After the second immunoprecipitation, the unbound fraction was incubated with Neutravidin beads (Pierce) at 4°C for 6 h. Beads were washed three times in RIPA buffer, eluted with 2X SDS sample buffer followed by SDS gel electrophoresis and Western blotting analyses. Quantitative analyses of Western blots were performed by determining the intensity of each band on the developed films with Image J software. The percentage of total and surface AMPAR subunits remaining in the unbound fraction was calculated by normalizing band intensities to the total input values.
Pyramidal neurons were targeted for whole-cell recording with borosilicate electrodes (3–6 MΩ) at room temperature in normal ACSF (in mM, 150 NaCl, 10 HEPES, 3 KCl, 10 D- glucose, 2 CaCl2, 2 MgSO4, pH 7.4) containing 100 μM picrotoxin, 100 μM DL-APV and 1 μM TTX. Electrode internal solution contained (in mM): 130 cesium methanesulfonate, 10 HEPES, 0.5 EGTA, 8 CsCl, 5 TEA-Cl, 1 QX-314, 10 Na phosphocreatine, 0.5 Na-GTP, 4 Mg-ATP and Alexa-594 (Invitrogen). Data were acquired with a Multiclamp 700B and PCLAMP 10 software (Molecular Devices) at 10 KHz. Current traces were low-pass filtered at 1 KHz and events having amplitude of >2X RMS noise were detected using MiniAnalysis (Synaptosoft). Kinetic measurements were performed on scaled, mean EPSC traces using a monoexponential decay function.
It is well established that chronic changes in synaptic activity result in global remodelling and reorganization of postsynaptic proteins (Ehlers, 2003). However, the effect of chronic changes in neuronal activity on PICK1 expression has not been examined. Bidirectional homeostatic scaling can be modelled in primary cultured neurons wherein the synaptic strength is scaled up by prolonged suppression of neuronal activity, or scaled down by elevated neuronal activity (O’Brien et al., 1998; Turrigiano et al., 1998). We found that 48 hours of TTX treatment, which blocks all evoked neuronal activity, significantly reduced PICK1 expression (73.1 ± 6.8% of control, Fig. 1A,B). On the other hand, chronic bicuculline treatment, which elevates neuronal firing, had no effect of PICK1 expression (93.2 ± 5.4% of control, Fig. 1A,B). The reduction of PICK1 protein level was apparent 24 hours after TTX incubation and persisted for as long as 72 hours. Interestingly, short-term treatment with TTX (up to 6 hours) did not have any significant effect on PICK1 protein level (Fig. 1C,D). The time scale by which PICK1 protein level is down-regulated by chronic TTX treatment coincides with the accumulation of surface AMPAR in response to decreased synaptic activity (O’Brien et al., 1998; Turrigiano et al., 1998; Sutton et al., 2006), suggesting a potential role for PICK1 in regulating the trafficking of AMPARs during synaptic scaling.
To gain insight into the mechanism behind reduced PICK1 level following chronic synaptic inactivity, cultured cortical neurons were incubated with TTX for 48 hours in the presence or absence of pharmacological agents that block the proteosomal and the lysosomal degradation pathways. As expected, cortical neurons treated with TTX in the presence of 0.1% DMSO (vehicle) led to a significant decrease in PICK1 protein level (53.9 ± 5.9% of control, Fig. 1E,F). This effect was largely prevented by co-incubation with the lysosomal inhibitor leupeptin (50 μg/mL, 85.6 ± 2.9% of control, Fig. 1E,F), but not by the proteosomal inhibitor lactacystin (0.5 μM, 59.4 ± 7.0% of control, Fig. 1E,F). These data suggest that the degradation of PICK1 during activity deprivation is mediated by the lysosomal degradation pathway.
To examine the role of PICK1 in synaptic scaling, we treated cultured cortical neurons derived from PICK1 knockout and WT littermates with TTX or bicuculine for 48 hours prior to mEPSC recording. WT neurons exhibited normal synaptic scaling up and scaling down of mEPSC amplitudes upon TTX and bicuculline treatment, respectively (Ctrl = 14.45 ± 0.82 pA, TTX = 23.96 ± 2.57 pA, Bic = 9.89 ± 0.69 pA, Fig. 2A-C). Strikingly, PICK1 knockout neurons showed significantly higher basal mEPSC amplitudes and thus, TTX-induced synaptic scaling appeared to be occluded (Ctrl = 20.13 ± 1.43 pA, TTX = 21.39 ± 2.25 pA, Fig. 2A-C). In contrast, PICK1 knockout neurons showed a robust synaptic scaling down of mEPSC amplitudes following chronic bicuculline treatment (Ctrl = 20.13 ± 1.43 pA, Bic = 10.84 ± 1.12 pA, Fig. 2A-C). As expected, neither TTX nor bicuculline treatment changed mEPSC frequency in PICK1 knockout or WT neurons (WT: Ctrl = 2.20 ± 0.42 Hz, TTX = 3.00 ± 0.88 Hz, Bic = 2.29 ± 0.50 Hz; KO: Ctrl = 2.66 ± 0.75 Hz, TTX = 4.22 ± 0.88 Hz, Bic = 3.48 ± 1.18 Hz, Fig. 2A,D). In addition, we did not observe any change in mEPSC decay kinetics in any condition (WT: Ctrl = 5.18 ± 0.35 ms, TTX = 4.78 ± 0.62 ms, Bic = 6.38 ± 0.73 ms; KO: Ctrl = 5.32 ± 0.60 ms, TTX = 4.74 ± 0.49 ms, Bic = 5.73 ± 0.45 ms), indicating that bidirectional synaptic scaling involves the accumulation and removal of GluA2-containing AMPARs on synapses. Together, these data suggest that PICK is involved in the unidirectional scaling up of AMPAR-mediated mEPSC upon chronic synaptic inactivity.
To ascertain that the effects we observed in PICK1 knockout neurons were not due to the loss of PICK1 function during neuronal development or the disruption of PICK1 presynaptic function, which may lead to abnormal network activity, we transiently transfected PICK1 shRNA construct to acutely knockdown PICK1 expression in a sparse population of cortical neurons. The PICK1 shRNA#1 efficiently reduced myc-PICK1 expression in heterologous cells and endogenous PICK1 expression in neurons after 3 days of expression (7 days in the case of lentiviral-mediated knockdown, Fig. 3A-C). We further characterized PICK1 shRNA#1 using two independent assays. Knockdown of PICK1 in cultured hippocampal neurons accelerated GluA2 recycling following NMDA receptor activation, resembling the phenotype seen in PICK1 knockout neurons (Lin and Huganir, 2007) (Supplemental Fig. 1). In addition, PICK1 knockdown in young cortical neurons resulted in significantly reduced number of distal dendritic processes, consistent with result from a previous study (Rocca et al., 2008) (Supplemental Fig. 2).
To directly investigate the cell autonomous effect of PICK1 knockdown in synaptic scaling, we treated cultured cortical neurons that had been transfected with either pSuper-Venus or pSuper-Venus-PICK1 shRNA#1 for 24 hours with TTX or bicuculine for 48 hours prior to mEPSC recording. Neurons transfected with pSuper-Venus exhibited normal synaptic scaling up and scaling down of mEPSC amplitudes upon TTX and bicuculline treatment, respectively (Ctrl = 12.57 ± 1.05 pA, TTX = 16.58 ± 1.10 pA, Bic = 8.54 ± 0.41 pA, Fig. 3D-F). Consistent with our findings in PICK1 knockout neurons, PICK1 knockdown also resulted in significantly higher basal mEPSC amplitudes and occluded TTX-induced synaptic scaling (Ctrl = 16.13 ± 0.80 pA, TTX = 15.99 ± 1.00 pA, Fig. 3D-F). In addition, PICK1 knockdown neurons showed a robust synaptic scaling down of mEPSC amplitudes following chronic bicuculline treatment (Ctrl = 16.13 ± 0.80 pA, Bic = 9.17 ± 0.52 pA, Fig. 3D-F). Again, we did not observe any significant changes in mEPSC frequency following TTX and bicuculline treatment in pSuper or PICK1 sh#1 transfected neurons (pSuper: Ctrl = 1.85 ± 0.42 Hz, TTX = 1.62 ± 0.30 Hz, Bic = 0.99 ± 0.27 Hz; sh#1: Ctrl = 2.19 ± 0.46 Hz, TTX = 1.35 ± 0.36 Hz, Bic = 1.08 ± 0.25 Hz, Fig. 3D,G). Moreover, we did not observe any change in mEPSC decay kinetics in any condition (pSuper: Ctrl = 4.51 ± 0.33 ms, TTX = 4.46 ± 0.23 ms, Bic = 5.52 ± 0.52 ms; sh#1: Ctrl = 4.21 ± 0.23 ms, TTX = 4.36 ± 0.22 ms, Bic = 4.67 ± 0.33 ms), again indicating that bidirectional synaptic scaling involves the accumulation and removal of GluA2-containing AMPARs at synapses. Together, the lack of TTX-induced synaptic scaling in both the PICK1 knockdown and PICK1 knockout neurons provides important complementary results establishing the involvement of PICK1 cell autonomously in postsynaptic neurons in unidirectional scaling up of AMPAR-mediated mEPSC.
To assess the contribution of PICK1 to the trafficking of AMPARs during synaptic scaling, we compared the surface expression of GluA1, GluA2 and GluA3 in PICK1 knockout neurons using a biotinylation assay. Under basal conditions, we observed a change in surface AMPAR levels in PICK1 knockout neurons. The levels of surface GluA2 and surface GluA3 in PICK1 knockout neurons were significantly higher than in neurons derived from WT littermates (surface/total ratio; GluA2 = 203 ± 31% of WT, GluA3 = 171 ± 24% of WT Fig. 4A,B). In addition, we also observed a slight but significant decrease in the level of surface GluA1 expression (surface/total ratio: 74 ± 11% WT, Fig. 4A,B). The decrease in surface GluA1 may be due to a compensatory mechanism since a reciprocal increase in surface GluA1 expression has also been observed in neurons overexpressing GluA2 siRNA (Gainey et al., 2009). These altered surface AMPAR levels are most likely due to impaired receptor trafficking events rather than protein synthesis or degradation defects as there was no significant difference in the total expression levels of GluA1, GluA2 and GluA3 subunits in PICK1 knockout neurons (GluA1 = 94 ± 7% of WT, GluA2 = 109 ± 14% of WT, GluA3 = 85 ± 7% of WT, Fig. 4A,C). These data indicate that the increase in surface GluA2 and GluA3 levels may account for the increase in basal synaptic transmission observed in PICK1 knockout neurons (Fig. 2C).
Next, we examined the surface expression of GluA1, GluA2 and GluA3 subunits following TTX and bicuculline treatment. Consistent with electrophysiological recordings, WT neurons exhibited robust increase and decrease in both surface GluA1 and GluA2 expression after TTX and bicuculline treatment, respectively (TTX: GluA1 = 156 ± 13% of control, GluA2 = 133 ± 9% of control; Bic: GluA1 = 64 ± 5% of control, GluA2 = 65 ± 8% of control, Fig. 5A,B). Strikingly, adaptation of surface GluA2 expression, but not GluA1, was impaired in PICK1 knockout neurons in response to TTX treatment (GluA1 = 146 ± 10% of control, GluA2 = 96 ± 8% of control, Fig. 5A,B). The inability of GluA2 subunit to scale up is likely due to an occlusion effect since basal surface GluA2 expression was already elevated in PICK1 knockout neurons (Fig. 4A,B). The normal increase in surface GluA1 expression upon TTX treatment indicates a crucial role of PICK1 in selectively controlling GluA2 trafficking. This is consistent with a recent finding that demonstrates a requirement of AMPAR GluA2 subunit for TTX-induced synaptic scaling (Gainey et al., 2009). In agreement with our electrophysiological recording, no defect of AMPAR trafficking was observed in PICK1 knockout neurons in response to bicuculline treatment (GluA1 = 57 ± 6% of control, GluA2=67 ± 6% of control, Fig. 5A,B).
To determine whether the lack of an increase in GluA2 surface expression was due to changes in total protein levels, we next examined the total expression of AMPAR subunits following TTX and bicuculline treatment. In WT neurons, chronic TTX treatment selectively increased the total GluA1 level, but not the total expression of GluA2 subunit (GluA1 = 139 ± 13% of control, GluA2 = 91 ± 8% of control, Fig. 5A,C). Conversely, bicuculline treatment led to significant reduction in total AMPAR expression (GluA1 = 67 ± 6% of control, GluA2 = 70 ± 8% of control, Fig. 5A,C). We found no significant difference between PICK1 knockout and WT neurons in changes in the total expression of AMPAR subunits upon TTX and bicuculline treatment (PICK1 KO, TTX: GluA1 = 118 ± 5% of control, GluA2 = 91 ± 6% of control; Bic: GluA1 = 63 ± 6% of control, GluA2 = 78 ± 5% of control, Fig. 5A,C). Moreover, surface to total ratio analyses revealed a distinct mechanism of regulation for GluA1 and GluA2 during TTX-induced synaptic scaling. In both PICK1 knockout and WT neurons, GluA1 total and surface levels change at the similar rate in response to TTX treatment (WT = 121 ± 9% of control, KO = 126 ± 10% of control, Fig. 5D). In contrast, the surface to total GluA2 ratio becomes significantly higher upon TTX treatment in WT but not in PICK1 KO neurons (WT = 156 ± 9% of control, KO = 115 ± 13% of control, Fig. 5D). Altogether, these data suggest that loss of PICK1 function occludes TTX-induced synaptic scaling by specifically impairing the trafficking of GluA2-containing AMPARs.
Unlike surface GluA2, the increase in surface GluA1 induced by TTX treatment is not impaired in PICK1 knockout neurons. Surprisingly, however, mEPSC amplitudes failed to scale up in PICK1 knockout neurons. This result along with a lack of change in mEPSC decay kinetics combined with the lack of surface GluA2 increases suggests that these newly inserted GluA1-containing receptors are most likely extrasynaptic Ca2+-permeable GluA1 homomers. To test this hypothesis, we combined surface biotinylation and GluA2/3 immunodepletion assays from neuronal lysates and examined the fraction of total and surface homomeric GluA1 AMPARs remaining. Two-rounds of immunoprecipitation effectively pulled down more than 98% of GluA2 and GluA3 leaving an approximately 15% of total GluA1 in the unbound fraction in all conditions (Fig. 6A,B). As expected, we did not observe any significant changes in the amount of surface GluA1 homomers following TTX treatment in WT neurons (Ctrl = 3.9 ± 0.4%, TTX = 3.4 ± 0.6%, Fig. 6A,C), suggesting that synaptic inactivity drives incorporation of GluA1/2 heteromers to the plasma membrane. Under basal condition, the levels of surface GluA1 homomers were significantly lower in the PICK1 knockout neurons (WT = 3.9 ± 0.4%, KO = 2.7 ± 0.3%, Fig. 6A,C), consistent with our previous study (Clem et al., 2010). Surprisingly, the level of surface GluA1 homomers was not altered following activity deprivation in PICK1 knockout neurons (Ctrl = 2.7 ± 0.3%, TTX = 2.0 ± 0.4%, Fig. 6A,C). This suggests that the TTX-induced increase in surface GluA1 in PICK1 knockout neurons represents an increase in surface GluA1/2 heteromers.
How can we account for the lack of change in surface GluA2 subunit in the PICK1 knockout neurons following chronic activity deprivation? Since PICK1 knockout neurons have increased levels of surface GluA2/3 AMPAR complexes, we hypothesize that during chronic activity deprivation, the increase in surface GluA1/2 heteromers is counterbalanced by the removal of heteromeric GluA2/3 receptors from the plasma membrane. Since nearly all of GluA2 subunit forms heteromeric complexes with either GluA1 and GluA3 subunits (Wenthold et al., 1996; Lu et al., 2009), we performed GluA1 immunodepletion assays from neuronal lysates and examined the fraction of total and surface heteromeric GluA2/3 AMPARs remaining following chronic activity blockade in both PICK1 WT and knockout neurons. Two-rounds of immunoprecipitation effectively depleted more than 98% of total GluA1 from neuronal lysates in all conditions (Fig. 7A,B). Interestingly, under basal condition total GluA2 and GluA3 in the unbound fraction following GluA1 immunodepletion were significantly higher in the PICK1 knockout neurons compared to the neurons prepared from the WT littermates (WT: GluA2 = 47.4 ± 3.0%, GluA3 = 56.3 ± 3.0%; KO: GluA2 = 59.6 ± 1.4%, GluA3 = 70.8 ± 2.1%, Fig. 7A,B), suggesting a possible role of PICK1 in AMPAR assembly or stability. Consistent with our previous finding, levels of surface GluA2 and GluA3 subunits were also significantly higher in the PICK1 knockout neurons (WT: GluA2 = 3.9 ± 0.5%, GluA3 = 16.1 ± 2.2%; KO: GluA2 = 6.4 ± 0.4%, GluA3 = 34.9 ± 5.0%, Fig. 7A,C). We found no significant differences in surface GluA2 or GluA3 levels following activity deprivation in WT neurons, indicating the absence of heteromeric GluA2/3 AMPARs insertion to the plasma membrane (Ctrl: GluA2 = 3.9 ± 0.5%, GluA3 = 16.1 ± 2.2%; TTX: GluA2 = 4.9 ± 0.6%, GluA3 = 21.6 ± 1.9%, Fig. 7A,C). In PICK1 knockout neurons, however, the level of surface GluA2/3 heteromers decreased significantly, suggesting internalization of these receptors following synaptic inactivity (Ctrl: GluA2 = 6.4 ± 0.4%, GluA3 = 34.9 ± 5.0%; TTX: GluA2 = 3.9 ± 1.0%, GluA3 = 19.6 ± 4.7%, Fig. 7A,C). Altogether, our data suggest that loss of PICK1 function leads to a complex disruption of proper trafficking of heteromeric GluA2/3 AMPARs, which underlies the occlusion of TTX-induced synaptic scaling in cultured neurons.
Synaptic scaling is a form of homeostatic plasticity by which a neuron adjusts it synaptic strength to maintain stable neuronal output during changes in network activity (Turrigiano, 2008). This mechanism can be achieved through a cell-wide adjustment of AMPARs at excitatory synapses (O’Brien et al., 1998; Turrigiano et al., 1998; Wierenga et al., 2005; Shepherd et al., 2006). However, the subunit composition of newly inserted or removed receptors during synaptic scaling is complex and currently controversial. Some studies have reported that both GluA1 and GluA2 subunits contribute to synaptic scaling (O’Brien et al., 1998; Wierenga et al., 2005; Cingolani et al., 2008; Gainey et al., 2009; Sun and Wolf, 2009), but others have reported that synaptic scaling operates mainly on the GluA1 subunit (Thiagarajan et al., 2005; Sutton et al., 2006; Aoto et al., 2008). This discrepancy depends on factors such as basal activity or developmental stage of particular types of neuronal culture, choices of pharmacological treatment as well as durations of the treatment. In this study, we chose to induce bidirectional synaptic scaling in young cultured cortical neurons with prolonged TTX or bicuculline treatment, as this system has been shown to target GluA1 and GluA2 heteromeric receptors (Wierenga et al., 2005; Wierenga et al., 2006; Ibata et al., 2008; Gainey et al., 2009). Consistent with previous studies, we observed bidirectional changes in both surface GluA1 and GluA2 levels that led to synaptic scaling of AMPAR-mediated mEPSC amplitudes, but not mEPSC frequency, suggesting an exclusive postsynaptic locus of expression. In addition, we did not observe any change in mEPSC decay kinetics, nor did we see any increase in surface homomeric GluA1 receptors following prolonged activity blockade, further confirming the involvement of GluA2-containing AMPARs during homeostatic synaptic scaling in our system.
Several studies have shown that PICK1 regulates the surface expression of GluA2-containing AMPARs either by facilitating their activity-dependent removal from the plasma membrane or by stabilizing them in intracellular pools, both of which result in a net decrease in surface GluA2 level (Kim et al., 2001; Perez et al., 2001; Terashima et al., 2004; Lin and Huganir, 2007; Rocca et al., 2008). Consistent with the proposed role of PICK1, we saw an increase in basal surface GluA2 and GluA3 accompanied by a slight compensatory decrease of surface GluA1 in PICK1 knockout neurons. Interestingly, we also saw an increase in total heteromeric GluA2/3 receptors in PICK1 knockout neurons, indicating a possible role for PICK1 in regulating the assembly of AMPAR in the biosynthesis pathway (Greger et al., 2002). Alternatively, PICK1 may also be involved in determining the stability of GluA2-containing AMPARs. In addition, we observed elevated basal synaptic transmission, suggesting that the increase in GluA2-containing AMPARs, most likely in the form of GluA2/3 heteromers, occurred at synapses. Indeed, previous studies have reported synaptic potentiation of basal transmission in hippocampal slices when GluA2-EVKI peptides, which disrupt PICK1-GluA2 interaction, are included in the patch pipette (Kim et al., 2001; Yao et al., 2008). Conversely, overexpression of PICK1 leads to a decrease in surface GluA2 level but not surface GluA1 in hippocampal neurons (Terashima et al., 2004).
In the present study, we found compelling evidence that PICK1 plays a cell autonomous role in TTX-induced synaptic scaling of postsynaptic neurons. Chronic synaptic inactivity reduces PICK1 protein levels in a time course that coincides with the accumulation of surface AMPARs. Consistent with the role for PICK1 in stabilizing intracellular pools of GluA2-containing AMPARs, we propose a model whereby the reduction in PICK1 level during prolonged synaptic inactivity relieves the intracellular retention, thus promoting the forward trafficking and accumulation of GluA2-containing receptors to the plasma membrane (Fig. 8A). Based on our model, one would expect that loss of PICK1 function would mimic the TTX-induced scaling of surface AMPARs and mEPSC amplitudes (Fig. 8B). Strikingly, TTX-induced accumulation of surface GluA1 subunit was normal, while the adaptation of GluA2 subunits was impaired in the PICK1 knockout neurons. Our GluA2/3 immunodepletion experiments suggest that these newly inserted GluA1-containing receptors are not Ca2+-permeable GluA1 homomers, instead they accumulate on the extrasynaptic plasma membrane as GluA1/2 heteromers. Moreover, GluA1 immunodepletion experiments in PICK1 KO neurons revealed an abnormal removal of surface GluA2/3 heteromers upon TTX treatment. These two pieces of evidence may account for the lack of a net increase in surface GluA2 levels upon TTX treatment. Thus, alteration in AMPAR subunit composition and aberrant trafficking GluA2-containing AMPARs underlie the occlusion of AMPAR-mediated mEPSC increase following activity deprivation in PICK1 KO neurons.
A prominent current theory of homeostatic plasticity holds that a global negative feedback mechanism compensates for long-term perturbations of network activity. But, how might these seemingly opposing types of plasticity coexist in a neuron? A recent study demonstrated a differential requirement for GluA1 and GluA2 subunits in LTP and synaptic scaling, respectively; implying that these two types of plasticity may involve two molecularly distinct pathways for AMPAR trafficking (Gainey et al., 2009). Here, we present yet another piece of evidence to support this notion. Previous studies have demonstrated the crucial role of PICK1 in hippocampal and cerebellar LTD (Steinberg et al., 2006; Terashima et al., 2008; Volk et al., 2010). By analogy, both LTD and bicuculline-induced synaptic scaling require the removal of AMPARs from synapses. Surprisingly, we found that the bicuculline-induced synaptic scaling was intact in PICK1 knockout neurons as shown by a robust reduction in surface AMPARs and mEPSC amplitudes. Our data argue against the role of PICK1 in AMPAR internalization from the plasma membrane, a common feature for both types of plasticity. Instead we favor a role for PICK1 in stabilizing internalized GluA2-containing AMPARs in intracellular pools, which may explain the LTD deficit in PICK1 knockout mouse (Steinberg et al., 2006; Terashima et al., 2008; Volk et al., 2010). In the case of bicuculline-induced synaptic scaling, the internalized receptors are destined for degradation, either by the proteosomal or lysosomal pathways as shown by the decrease in total GluA1 and GluA2 protein levels (Fig. 5A,C). Together, our data strongly support the idea of a differential requirement of two molecularly distinct AMPAR trafficking pathways in Hebbian and homeostatic plasticity. These mechanisms may ensure that the two types of plasticity may coexist in neurons without interference.
We thank Dr. Akira Sawa (Johns Hopkins University) for providing pSuper-Venus construct. We are grateful to Drs. Gareth Thomas and Lenora Volk for critical reading of the manuscript. We thank Min Dai, Yilin Yu, Monica Coulter and Richard Johnson for technical support. This work was supported by grants from the National Institute of Health and the Howard Hughes Medical Institute (to R.L.H.). V.A. is supported by fellowships from the International Human Frontier Science Program (LT00399/2008-L) and the Australian National Health and Medical Research Council (ID. 477108). R.L.C. is supported by a fellowship from National Research Service Award from the National Institute of Health (F32 MH087037-01).
Competing Financial Interests Statement:
Under a licensing agreement between Millipore Corporation and The Johns Hopkins University, R.L.H. is entitled to a share of royalties received by the University on sales of products described in this article. R.L.H. is a paid consultant to Millipore Corporation. The terms of this arrangement are being managed by The Johns Hopkins University in accordance with its conflict–of–interest policies.