2A inhibition induces cerebellar long-term depression and declustering of synaptic AMPA receptor

T. Launey*, S. Endo, R. Sakai, J. Harano, and M. Ito

Laboratory for Memory and Learning, RIKEN Brain Science Institute, 2-1 Hirosawa, Wako, Saitama 351-0198, Japan

Edited by Roger A. Nicoll, University of California, San Francisco, CA, and approved November 13, 2003 (received for review May 15, 2003) Phosphorylation of synaptic (RS)-␣-amino-3-hydroxy-5-methyl-4- tion in mature cerebellum (26, 27), we undertook to clarify the isoxazolepropionic acid) (AMPA) receptors (AMPARs) is an essen- relative influence of PP-2A, PP-1, and PP-5 on AMPAR regu- tial component of cerebellar long-term depression (LTD), a form of lation at the GC–PC synapse and as their interactions with the synaptic plasticity involved in motor learning. Here, we report that molecular mechanisms of pairing-induced LTD. protein phosphatase 2A (PP-2A) plays a specific role in controlling synaptic strength and clustering of AMPARs at synapses between Materials and Methods granule cells and Purkinje cells. In 22- to 35-day cerebellar cultures, Cerebellar-Dissociated Culture. The experimental procedures and specific inhibition of postsynaptic PP-2A by fostriecin (100 nM) or housing conditions for animals were approved by the RIKEN cytostatin (10–60 ␮M) induced a gradual and use-dependent Institute’s Animal Experiment Committee. Dissociated rat cer- decrease of synaptic current evoked by the stimulation of a single ebellar cultures were prepared as described (28) with minor granule cell, without altering receptor kinetics nor passive electri- modifications (see Supporting Text, which is published as sup- cal properties. By contrast, PP-2A inhibition had no effect on porting information on the PNAS web site). immature Purkinje cells (12–15 days). Concurrent PP-2A inhibition and AMPAR stimulation induced a reduction of miniature synaptic Reagents and Drugs. Microcystin-LR and fostriecin were pur- currents and a reduction of AMPAR density at synapses. Either chased from Calbiochem, and AMPA, D(-)-2-amino-5- PP-2A inhibitor alone or AMPA stimulation alone had no significant phosphonopentanoic acid, and 7-(hydroxyimino)-cyclopropa- effect. Inhibition of PP-1 by inhibitor 1 (10–27 units͞␮l) had no chromen-1a-carboxylate ethyl ester were from Tocris Cookson effect on synaptic current. Synaptic depression induced by PP-2A (Bristol, U.K.). Cytostatin and dephosphocytostatin were gen- inhibition occluded subsequent induction of LTD by conjunctive erous gifts from T. Takeuchi (Institute for Chemotherapy, stimulation and was abolished by a calcium chelator or a protein Shizuoka, Japan). The noncompetitive AMPAR antagonist kinase inhibitor, suggesting a shared molecular pathway and (Ϯ)1-(4-aminophenyl)-3-methylcarbamyl-7,8-methylenedioxy- involvement of PP-2A in LTD induction. 3,4-dihydro-5H-2,3-benzodiazepin (GYKI-53655) was gener- ously provided by I. Tarnawa (Gedeon Richter, Budapest). cerebellum ͉ Purkinje ͉ plasticity ͉ synapse Thiophosphorylated inhibitor 1 was prepared as described (16).

erebellar long-term depression (LTD) is a long-lasting at- Electrophysiological Recording. Whole-cell voltage-clamp record- Ctenuation of the synaptic transmission between granule cells ings were made from cultured PCs that were visually identified (GCs) and Purkinje cells (PCs), which has been thought to be a by using phase-contrast optics, occasionally confirmed by filling major mechanism of motor learning (1, 2). Complex signal with fluorescent dye during recording (Fig. 1A). The extracel- transduction mechanisms underlie this unique form of synaptic lular solution contained 140 mM NaCl, 3 mM KCl, 3 mM CaCl2, plasticity (reviewed in refs. 3–5) with several pathways eventually 1 mM MgSO4, 0.5 mM Na2HPO4,10mMD-glucose, 10 mM converging to alter (RS)-␣-amino-3-hydroxy-5-methyl-4- Hepes, 3 mM pyruvate, 0.03 mM picrotoxin, and 0.03 mM isoxazolepropionic acid) (AMPA)-selective glutamate receptors D(-)-2-amino-5-phosphonopentanoic acid (pH 7.35, 330 Ϯ 5 (AMPARs), hence reducing GC–PC synaptic strength. Evidence mOsm, 33°C). Pipette-filling solution contained 60 mM K glu- indicate that phosphorylation of the AMPAR in its C-terminal conate, 60 mM K methanesulfonate, 20 mM KCl, 3 mM MgCl2, domain plays a key role in this process (6–9), apparently 4mMNa2ATP, 0.4 mM NaGTP, 15 mM Hepes, 2 mM EGTA, Ϯ triggering endocytotic removal of these receptors from the 0.8 mM CaCl2, and 1 mM reduced glutathione (pH 7.4, 310 postsynaptic membrane (7, 10). Accordingly, several protein 5 mOsm, pCa ϭ 7.3). We waited Ͼ10 min after patch rupture to kinase-dependent signaling cascades have been shown to be allow diffusion of the PP inhibitors into PC dendrites before essential for cerebellar LTD induction (reviewed in ref. 5). starting experimental protocols. Seal resistance was 2–16 G⍀. Although protein (PPs) share with protein kinases Series resistance was 7–20 M⍀ and was not compensated. the task of regulating protein phosphorylation level, they have Recording was terminated if unrecoverable changes of series or long been considered mere housekeepers, with poorly under- input resistance exceeded 25%. For evoked excitatory postsyn- stood substrate specificity and loose regulation. In recent years, aptic current (eEPSC) a single GC was stimulated (0.3–4 V, 500 however, studies using phosphatase inhibitors revealed that ␮s) through a saline-filled, fire-polished glass pipette gently these exert a major influence on synaptic plasticity in the cerebellum (11–15) and other brain structures (16, 17). In the cerebellum, investigation of AMPAR regulation by PP at the This paper was submitted directly (Track II) to the PNAS office. GC–PC synapse represents a challenge as PCs abundantly Abbreviations: AMPA, (RS)-␣-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid; AM- ͞ PAR, AMPA receptor; LTD, long-term depression; PP, protein phosphatase; GC, granule cell; express at least five major types of PPs: PP-1, PC, Purkinje cell; EPSC, excitatory postsynaptic current; eEPSC, evoked EPSC; mEPSC, min- PP-2A, PP-2B, PP-2C, and PP-5 (18–22). In addition, PCs have iature EPSC; DIV, days in vitro; CS, conjunctive stimulation; GluR, glutamate receptor; the distinctive property of expressing G substrate, a PP-1, PP-2A GYKI-53655, (Ϯ)1-(4-aminophenyl)-3-methylcarbamyl-7,8-methylenedioxy-3,4-dihydro- (23–25), and possibly PP-5 inhibitory protein. Because G sub- 5H-2,3-benzodiazepin. strate is activated through the nitric oxide͞guanylyl cyclase͞ *To whom correspondence should be addressed. E-mail: t࿝[email protected]. protein kinase G pathway known to be involved in LTD induc- © 2003 by The National Academy of Sciences of the USA

676–681 ͉ PNAS ͉ January 13, 2004 ͉ vol. 101 ͉ no. 2 www.pnas.org͞cgi͞doi͞10.1073͞pnas.0302914101 Downloaded by guest on September 25, 2021 pressed against its soma. To record miniature EPSCs (mEPSCs), 1 ␮M tetrodotoxin was added to the extracellular solution, and potassium was replaced by cesium in the internal solution. The mEPSCs were analyzed by using a custom semiautomatic algo- rithm (ref. 29 and see Supporting Text). When data from several experiments were combined, EPSC amplitudes for each exper- iment were normalized relative to the average amplitude at the beginning of the recording (first 12 stimulations). The eEPSC failures were not included in averaged traces. Statistical com- parisons were done by using t test or Mann–Whitney U test when appropriate. All values are mean Ϯ SEM. The different exper- imental protocols were interleaved to avoid the influence of any heterogeneity between culture batches. The data presented were obtained from voltage-clamp recording of 161 PCs from 113 cultures (40 batches).

Quantification of AMPAR Density. Expression of AMPAR at the surface of PCs was assessed by immunofluorescence, using an antibody that recognizes the N-terminal extracellular domain of its glutamate receptor 2 (GluR2) subunit (7), a gift from H. Hirai (Kobe University, Kobe, Japan). Briefly, after pharmacological treatment, fixation, and immunostaining, the cultures were imaged by using a charge-coupled device camera (Coolsnap, Photometrics, Tucson, AZ). We chose PCs for which most of the dendrites were within the focal plane (ϫ100 objective, 1.4 numerical aperture). Digital images were high-pass filtered by convolution with a 5-by-5 pixels Laplace filter and thresholded

to isolate receptor clusters. This mask was then applied to the NEUROSCIENCE original image. For each detected cluster, the mean fluorescence intensity and area were recorded. All images were acquired with the same exposure time and analyzed blindly with the same segmentation threshold. The data presented are from 36 cultures and 197 images. A minimum of 20,000 spines were analyzed for each experimental conditions.

AMPA C-Terminal Domain Dephosphorylation Assay. Dephosphory- lation of the cytoplasmic domain of AMPAR GluR2 subunit by 0.01 unit of PP-2A and PP-1 catalytic subunits was measured by using either the 51-aa C-terminal fragment of the protein (GluR2-CT) or a fusion protein made of this fragment together with GST (GST-GluR2-CT), as described (30). The procedure and unit definition have been reported in detail (31) and are provided in Supporting Text. Results Synaptic activation was obtained by the stimulation of a single GC located 50–300 ␮m from the recorded PC soma while PP inhibitors were infused into the recorded PC, allowing contin- uous monitoring of AMPAR alteration at functional synapses. The average amplitude of eEPSC was 139.2 Ϯ 6.2 pA, probably reflecting the presence of more than one synaptic contact between the stimulated GC and the PC (32), owing to the coplanar growth of GC axons and PC dendrites. These eEPSCs had brief and constant latencies of 2.16 Ϯ 0.11 msec (n ϭ 91). Fig. 1. Effect of PP inhibitors on GC–PC synapse. (A)(Left) A single GC is electrically stimulated, and eEPSCs are recorded in the postsynaptic PC infused The latency and amplitude did not change noticeably when the with PP inhibitor. (Right) A PC infused with a fluorophore, from a 27-day stimulus intensity was varied up to three times the threshold for cerebellar culture. (Scale bar: 15 ␮m.) (B1) Time course of eEPSC depression eight GC–PC pairs examined in detail. The eEPSCs occurred in induced by 100 nM fostriecin. Synaptic stimulation was started at least 10 min an all-or-none fashion (see failures in Fig. 1B1), the failures after establishment of the whole-cell configuration to allow for drug diffusion probably arising from spontaneous fluctuation of the GC soma into distal dendrites. (B2)(Upper) eEPSCs from the experiment shown in B1,at excitability or axon refractoriness after spontaneous action the indicated time points a and b (average of 10 consecutive traces). The potentials (33). vertical transient preceding the eEPSC corresponds to the GC stimulation artifact. (Lower) Depressed eEPSC (b) peak-scaled to original amplitude. Note PP-2A Inhibition Induces Synaptic Depression. Infusion of the spe- that the eEPSC time course did not change, nor did the passive electrical response to a 3-mV voltage pulse. (C) Average decay time course in control cific PP-2A inhibitor fostriecin (100 nM) in PCs at 22–35 days in condition (n ϭ 14) and in the presence of fostriecin (100 nM, culture 12–15 DIV, vitro (DIV) produced a gradual depression of eEPSC amplitude n ϭ 8, and 22–35 DIV, n ϭ 14), cytostatin (10–60 ␮M, n ϭ 12), microcystin-LR over 30–40 min (Fig. 1B1). The passive electrical properties of (10 ␮M, n ϭ 13), thiophosphorylated inhibitor 1 (10–27 units͞␮l, n ϭ 11), and the recorded PC and the apparent kinetics of the stimulated dephosphocytostatin (30 ␮M, n ϭ 4). AMPARs were unaffected (Fig. 1B2). On average, the depres-

Launey et al. PNAS ͉ January 13, 2004 ͉ vol. 101 ͉ no. 2 ͉ 677 Downloaded by guest on September 25, 2021 sion produced by 100 nM fostriecin and 10–60 ␮M cytostatin (23–28 DIV), another specific PP-2A inhibitor, reached 35 Ϯ 5% and 46 Ϯ 9% of control amplitude, at 50 min (n ϭ 14 and 12, respectively, Fig. 1C). No significant depression was observed in the absence of phosphatase inhibitor, with vehicle alone (0.1% DMSO) in the pipette (97 Ϯ 5%, n ϭ 14, 14–32 DIV). Dephosphocytostatin (30 ␮M), an inactive form of cytostatin, did not produce any marked effect on eEPSC amplitude (91 Ϯ 3%, n ϭ 4). In contrast, infusion of 100 nM fostriecin into immature PCs (12–15 DIV) did not depress eEPSCs (93 Ϯ 5%, n ϭ 8), confirming a recent report using young cerebellar cultures from mouse (15). All subsequent experiments were done on PCs aged 22–35 DIV (average 25.7 DIV). Being membrane permeable, both fostriecin and cytostatin may diffuse transsynaptically into GC axon terminals and affect glutamate release (34). Thus, the experiment was repeated with the membrane-impermeable PP inhibitor microcystin-LR (10 ␮M) to confirm that the depression only resulted from postsyn- aptic inhibition of PP. Intracellular infusion microcystin-LR into PCs reproduced both the time course and magnitude (37 Ϯ 8% of control, at 50 min, n ϭ 11) of the depression observed with fostriecin and cytostatin (Fig. 1C), whereas bath application had no effect (96 Ϯ 6%, n ϭ 4, data not shown). To verify that PP-2A inhibitors were used at saturating concentration, 10, 30, and 60 ␮M cytostatin was tested, and no differences were observed in time course or magnitude of depression (data not shown). At these concentrations, both fostriecin and cytostatin block PP-2A Ͻ Ͻ [EC50 5nMand 100 nM, respectively (35, 36), see also Fig. 6, which is published as supporting information on the PNAS web site], but do not affect PP-1, PP-5, PP-2B, or PP-2C. Intracellular infusion of thiophosphorylated inhibitor 1, a PP- 1-specific inhibitor, at 10 or 27 units͞␮l, did not significantly affect eEPSCs (90 Ϯ 7% of control, n ϭ 11, P Ͼ 0.5). To ascertain the diffusion of this large peptide into the PC, some experiments were done with a fluorescent conjugate, showing diffusion to the distal dendrites within 15 min of infusion (data not shown).

Occlusion with Classical LTD. To establish the relationship between PP inhibitor-induced depression and the classical LTD induced by conjunctive stimulation (CS), we performed an occlusion experiment (Fig. 2A). Once the depression induced by PP-2A Fig. 2. Synaptic depression induced by PP-2A inhibitor and LTD share a inhibitor had reached steady state, further reduction of eEPSC common molecular pathway. (A) Occlusion experiment. Once depression induced by a PP-2A inhibitor (100 nM fostriecin) reached steady state, CS amplitude by CS was attempted. Although the CS protocol used (vertical bar at 38 min) failed to induce further eEPSC depression. A gap was produces robust LTD at naive synapses (Fig. 2A Inset), here no created to align the CS onsets, and eEPSC amplitudes were normalized to further depression was observed (n ϭ 4). This finding suggests average amplitude within the 5-min period preceding the CS (n ϭ 4). (Left that both forms of synaptic depression share at least part of a Inset) eEPSC from such an experiment (average of 30 traces each) at 2 (a), 27 common molecular pathway. This absence of depression after (b), and 57 (c) min. (Right Inset) LTD induced by CS in the absence of PP-2A CS does not result from PC dialysis nor from the prior reduction inhibitor, with prior reduction of eEPSC amplitude by a glutamate antagonist, of eEPSC amplitude because comparable partial reduction of GYKI-53655 (500–800 nM, n ϭ 6). (B) Replacement of intracellular calcium by 2 mM 1,2-bis(2-aminophenoxy)ethane-N,N,NЈ,NЈ-tetraacetate acid (BAPTA) eEPSC amplitude by a noncompetitive AMPAR antagonist ϭ (GYKI-53655) perfused for 20 min did not prevent CS-induced (n 11) prevents depression whereas bath application of the metabotropic GluR inhibitor 7-(hydroxyimino)-cyclopropachromen-1a-carboxylate ethyl es- LTD (Fig. 2A Inset), as recently shown (37). Note that we ter (CPCCOet, 100 ␮M, n ϭ 6) has no effect on microcystin-LR-induced depres- selected cell pairs for which the eEPSC amplitude was still at sion (10 ␮M, n ϭ 13). (C) Requirement of protein kinase activity. Coinfusion of least 35 pA after the depression induced by fostriecin͞GYKI, to the broad-spectrum protein kinase inhibitor staurosporin (1 ␮M) and fostrie- allow better detection of any further depression. cin (n ϭ 4) abolishes synaptic depression induced by fostriecin (100 nM) alone (n ϭ 14, same data set as in Fig. 1C). (D) Use-dependent depression. Time Requirement for Ca2؉ and Protein Kinase but Not Metabotropic course of synaptic depression for GC stimulation intervals of 3 and 30 s. Thin Receptors. We next examined whether the signaling cascades lines represent exponential fit to the data with decay time constants of 5.1 min ϭ ϭ underling classical LTD are shared by PP-2A inhibitor-induced (n 5) and 28.9 min (n 7), respectively. (E) Decay time constant as a function depression. Replacement of calcium in the internal solution of stimulation interval. The numbers in brackets indicate the number of data points (E). by the high-affinity calcium chelator 1,2-bis(2-aminophe- noxy)ethane-N,N,NЈ,NЈ-tetraacetate acid (BAPTA) (2 mM) abolished the depression induced by microcystin-LR (94 Ϯ 7% of control at 50 min, n ϭ 11, Fig. 2B), whereas superfusion of the suggests that PP-2A inhibitors may cut into LTD signal trans- metabotropic GluR (mGluR) antagonist 7-(hydroxyimino)- duction downstream of mGluRs. To ascertain that the depres- cyclopropachromen-1a-carboxylate ethyl ester (100 ␮M) had no sion induced by fostriecin͞cytostatin was specifically mediated effect on depression (44 Ϯ 14% at 50 min, n ϭ 6). This finding by an alteration of the balance between phosphorylation and

678 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0302914101 Launey et al. Downloaded by guest on September 25, 2021 dephosphorylation, we examined whether a protein kinase in- hibitor would prevent fostriecin-induced depression. A broad- spectrum kinase inhibitor was used because at least five different protein kinases families (PKC, mitogen-activated protein kinase, cGMP-dependent protein kinase, tyrosine kinase, and Ca2ϩ͞ calmodulin-dependent protein kinase) are essential for LTD induction (5). Coinfusion of 1 ␮M staurosporin with 100 nM fostriecin abolished the depression induced by fostriecin alone (97 Ϯ 13% and 46 Ϯ 9% of control at 50 min, n ϭ 14 and n ϭ 4, respectively). Staurosporin alone produced a small, nonsig- nificant potentiation (105 Ϯ 6% of control at 50 min, n ϭ 4, P Ͼ 0.5, Fig. 2C).

Synaptic Depression Is Use-Dependent. It has been previously reported that depression induced by PP-1͞PP-2A inhibitors requires repetitive stimulation (12, 15). We extended this ob- servation by using the PP-2A-specific cytostatin and further analyzed the effect of various stimulation frequencies on the depression time course (Fig. 2 D and E). Indeed, reducing the stimulation interval from 5 to3sresulted in a faster depression (time constant of 12.0 Ϯ 3.2 min and 5.9 Ϯ 0.9 min, respectively). Conversely, increasing the stimulation interval to 30 s resulted in Fig. 3. Depression of mEPSCs induced by PP-2A inhibitor requires activation Ϯ of AMPARs. (A) Drug application protocol, control mEPSCs were recorded an increase of rate constant to 33.8 4.5 min (Fig. 2D). Plotting during the first 15 min after patch rupture and compared to mEPSCs recorded decay rate constant for various stimulation intervals revealed a at least 15 min after the end of AMPA perfusion. (B) mEPSCs at the beginning near-linear relationship, suggesting that the fractional depres- of the recording (Upper) and after perfusion of cytostatin plus AMPA (Lower). sion per stimulation is independent of the stimulation interval Superposition of five consecutive traces. (C) Average of mEPSCs before (488 within the 3- to 30-s range (Fig. 2E). On average, this fractional traces, Left) and after AMPA plus cytostatin (429 traces, Right). The dashed

depression was 0.92 Ϯ 0.11% per stimulation and may corre- blue line represents the average of the fitting function. Overlapping mEPSCs NEUROSCIENCE spond to the fraction of synaptic AMPARs rendered nonfunc- were not included in these averages. (Inset) Response to a 3-mV test pulse. (D) tional after each synaptic stimulation. Trace from C, with the second average peak-scaled (dashed red line) to the average of the control mEPSC. (E) Cumulative distribution of mEPSC amplitude and decay time constant from the cell shown in B.(F) Change of mEPSC Depression Requires Coincident PP-2A Inhibition and AMPAR Stimu- amplitude under the indicated experimental conditions, normalized to the lation. Because PP-2A inhibitor-induced depression requires median amplitude during the control period. Group size is indicated above synaptic stimulation but is independent of metabotropic GluR each condition. ***, P Ͻ 0.001. activation, we hypothesized that among the various synaptic signals, the postsynaptic action on AMPAR is the major com- ponent. Thus, we combined PP-2A inhibition and AMPA bath decay time constant between these different experimental con- application. To avoid GC firing and consequent synaptic acti- ditions did not reveal any significant difference between control vation, we added 1 ␮M tetrodotoxin. With all eEPSCs blocked, (4.9 Ϯ 0.7 ms) and cytostatin alone (5.5 Ϯ 1.0 ms), AMPA alone we turned to mEPSCs to evaluate alteration of synaptic trans- (3.9 Ϯ 0.8 ms), AMPA ϩ cytostatin (4.4 Ϯ 1.1 ms), or 45-min mission. After recording control mEPSCs during 15 min, we recording without drug (4.7 Ϯ 1.0 ms), with P Ͼ 0.2 in all cases ␮ bath-applied the membrane permeable cytostatin (10 M) for 30 (U test). Thus, combined PP-2A inhibition and AMPAR stim- ␮ min with or without AMPA stimulation (10 M, 5 min). A ulation are necessary and sufficient for inducing depression. second sample was then recorded after 15 min of AMPA ϭ wash-out (Fig. 3A). For each recorded PC (n 19, 21–24 DIV), PP-2A Activity Controls AMPAR Clustering. The comparable mag- a minimum of 500 mEPSCs was analyzed by fitting to a template nitude of the cytostatin-induced depression between eEPSCs function to extract amplitude and shape parameters (29). Com- (53.1%) and mEPSCs (46.3%) suggests that site of expression of bination of cytostatin and AMPA resulted in a prominent the depression is mostly postsynaptic. Thus, we examined the decrease of average mEPSC amplitude and an apparent decrease effect of a PP-2A inhibitor on AMPAR synaptic clustering of frequency (Fig. 3B). Because reduction of amplitude below because LTD and declustering have been shown to be correlated noise level may masquerade as a decrease of frequency, this (7). To estimate the density of AMPARs at the surface of PCs parameter was not further analyzed. For the representative PC (20.2 Ϯ 0.2 DIV, n ϭ 197), we used an antibody recognizing the shown in Fig. 3 B–D, comparison of amplitude and decay time ͞ ϩ N-terminal extracellular domain of GluR2 3 (GluR2-N), as in constant distribution between control and cytostatin AMPA ͞ conditions revealed that although the amplitude decreased (14.9 ref. 7. Cytostatin and or AMPA were added to the culture medium by using the same incubation times as for our mEPSC to 9.1 pA, Fig. 3E), there was no conspicuous change of decay ␮ time constant (3.2 vs. 3.8 ms, Fig. 3 C and D). Because amplitude study, but with the cytostatin concentration raised to 60 Mto distributions were often skewed (14) because of the imperfect counteract potential interaction with serum in the culture me- space clamp and the heterogeneous population of synapses, we dium. Spontaneous action potentials were blocked by addition of ␮ normalized all mEPSC amplitudes to the median amplitude 1 M tetrodotoxin. After incubation, the cultures were fixed and measured during the control period. Perfusion of cytostatin stained for GluR2-N. Control cultures not exposed to chemical alone (n ϭ 5), AMPA alone (n ϭ 4), or saline without drug (n ϭ stimulation exhibited bright, small fluorescence spots (Fig. 4A), 5) did not produce any significant alteration of mEPSC ampli- previously shown to correspond to synaptic sites (7). Because tude or shape after 45 min of recording, with median amplitudes PCs exhibited very high densities of GluR2͞3 compared to other at 87.5%, 103.2%, and 89.7% of control, respectively (P Ͼ 0.1 in cell types in the cerebellar cortex (38) and a very characteristic all cases, U test). When both AMPA and cytostatin were applied, dendritic morphology, no PC-specific counterstain was neces- mEPSC median amplitude was reduced to 53.7% of control sary. After exposure to a combination of cytostatin (60 ␮M, 30 amplitude (P Ͻ 0.001, n ϭ 5, U test). Comparison of average min) and AMPA (10 ␮M, 5 min), we observed a decrease in the

Launey et al. PNAS ͉ January 13, 2004 ͉ vol. 101 ͉ no. 2 ͉ 679 Downloaded by guest on September 25, 2021 Fig. 5. In vitro assay of AMPA GluR2 C-terminal domain dephosphorylation by PP-2A (see Materials and Methods). Both GST-GluR2-CT and the thrombin cleaved GluR2-CT fragment are dephosphorylated. Thin line shows exponen- tial fit to data points with indicated time constant. Average is from triplicate.

intensity, only puncta remaining ‘‘bright enough’’ after depres- sion can be detected and quantified. Thus, the decrease of cluster intensity is probably underestimated. An apparent decrease of the number of clusters per unit of dendritic surface was also observed after AMPA plus cytostatin treatment but this reduc- tion could not be reliably quantified. Although it would have been interesting to examine declustering in young cultures, most PCs before 15 DIV showed only short and spineless dendrites, precluding morphological analysis.

PP-2A Regulates Phosphorylation of GluR2 Cytoplasmic Domain. Cer- ebellar LTD correlates with an increased phosphorylation of the AMPAR GluR2 subunit cytoplasmic domain (7–9). Because PP-2A is known to dephosphorylate several membrane ion channels (39) and a GluR (40) there is a possibility that PP-2A may directly control AMPAR dephosphorylation. To examine this hypothesis, we conducted an in vitro assay with the PKC- phosphorylated 32P-GluR2 C-terminal domain as a substrate and revealed that PP-2A dephosphorylated this substrate quickly Fig. 4. Coapplication of AMPA and PP-2A inhibitor induces a declustering of (time constant: Ͻ9.2 min) and efficiently (Fig. 5). Under the GluR2͞3. (A)(Left) Digital fluoromicrography of a PC dendrite immuno- ͞ same conditions, dephosphorylation by PP-1 was also observed labeled with an antibody binding the extracellular domain of GluR2 3. (Up- Ͼ per) Control. Bright small puncta correspond to GluR2͞3 clustering at GC–PC albeit with slower kinetic (time constant 15.8 min, data not synapses. (Lower) In a sister culture, cytostatin (60 ␮M) plus AMPA (10 ␮M) shown). treatment caused a reduction of density, area, and fluorescence intensity of GluR2͞3 clusters. (Scale bar: 10 ␮m.) (Right) Pseudocolored 3D elevation of Discussion pixel intensity. Scale values correspond to camera analog͞digital conversion Our results demonstrate that selective inhibition of PP-2A in 22- units (12 bits). (B) Segmentation of cluster regions from fluorescence images. to 35-DIV cerebellar cultures produces a depression of GC–PC (Upper) Inverted fluorescence image. (Lower) Result of segmentation algo- synaptic current resembling LTD induced by CS. This depression rithm (see Materials and Methods). (C) Comparison of cluster alteration under various experimental conditions. A minimum of 25,000 clusters was analyzed requires activation of AMPA-gated channels, protein kinase per condition, and experiments were done in triplicate. *, P Ͻ 0.02; ***, P Ͻ activity, and the presence of free cytoplasmic calcium (the 0.001. estimated concentration in the pipette was Ϸ50 nM). Combined PP-2A inhibition and AMPAR activation was sufficient to induce a depression of mEPSCs and a declustering of synaptic number of clusters per dendrite and a decrease in spot intensity AMPA GluR2͞3 subunits. A selective PP-1 inhibitor did not (Fig. 4A). For quantification, digital fluoromicroscopic images induce any notable synaptic depression. By contrast, in immature were acquired and analyzed blind to detect receptor clusters PCs (12–15 DIV) a PP-2A inhibitor had no effect. This finding (Fig. 4B). Cluster fluorescence intensity and size were reduced suggests that sustained PP-2A activity in mature PC contribute after coapplication of cytostatin and AMPA to 78.7 Ϯ 2.8% and to the regulation AMPARs at GC–PC synapses (see below). The 69.0 Ϯ 4.2% of control, respectively (P Ͻ 0.05 and P Ͻ 0.001, Fig. similar reduction of eEPSCs and mEPSCs amplitude, without 4C). By contrast, incubation with cytostatin alone, AMPA plus alteration of kinetics, suggests that this depression is mostly dephosphocytostatin, or AMPA plus cytostatin plus GYKI- postsynaptic, as described for LTD in similar preparations (41, 53655 (10 ␮M) did not produce any significant alteration of 42). This notion is further confirmed by our receptor decluster- clusters. Cluster intensities were 99.8 Ϯ 2.7%, 105.5 Ϯ 1.9%, and ing experiments showing that the same stimulation protocol that 103.6 Ϯ 3.9% of control for each condition, and areas were induces mEPSC depression also induces a declustering of syn- 104.5 Ϯ 5.2%, 97.3 Ϯ 5.8%, and 105.7 Ϯ 5.5% of control, aptic AMPARs. Taken together, these results suggest that respectively (Fig. 4C). The results suggest that PP-2A inhibition PP-2A may regulate AMPAR density at synapses through a per se does not induce receptor declustering and that coactivation use-dependent mechanism involving protein kinase(s) activity, of AMPARs is required. Similar results were obtained when calcium signaling, and repetitive or sustained activation of cytostatin was replaced by fostriecin (data not shown). It should synaptic AMPARs. Because cerebellar LTD depends on phos- be noted that because cluster detection is based on fluorescence phorylation of GluR2 at Ser-880 (6–9), an attractive hypothesis

680 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0302914101 Launey et al. Downloaded by guest on September 25, 2021 is that PP-2A directly regulates the phosphorylation level of this synaptic remodeling (disappearance of functional N-methyl-D- subunit, as shown for N-methyl-D-aspartate receptors in the aspartate receptors and of multiple climbing fiber innervation) hippocampus (40) and as suggested by our in vitro dephosphor- and changes in electrical properties resulting from alteration of ylation result (Fig. 5). In addition, PP-2A is known as a key voltage-dependent channel populations (45, 46). The signaling regulator of several types of protein kinases (reviewed in ref. 43), cascade underlying cerebellar LTD also appears to undergo including some involved in LTD such as PKC, mitogen-activated considerable maturation during this period, with neither IP nor ϩ 3 protein kinase, and Ca2 ͞calmodulin-dependent kinase (3–5). NO required in immature PCs (47, 48), whereas these signals are Reduction of PP-2A activity by an endogenous inhibitor would necessary in the mature cerebellum (5). Furthermore, this period thus result in increased protein kinase activity, enabling LTD also corresponds to a drastic inversion of the relative abundance ␤ ␥ induction. Indeed, one such inhibitor is G substrate, which is of PP-2AB and PP-2AB in the brain, two PP-2A regulatory interesting in the context of cerebellar LTD as it is expressed subunits highly expressed in PCs and involved in PP-2A targeting almost exclusively by PCs (23), it is phosphorylated through the (21). Our results may provide a functional explanation for the NO-cGMP-protein kinase G signaling cascade known to be recent clinical report that some forms of human spinocerebellar involved in LTD induction in cerebellar slices (4, 5), and its ataxia are associated with an alteration of the gene encoding the phosphorylation is increased after LTD-inducing stimuli (44). ␤ PP-2A regulatory subunit PP-2AB (49). The phosphorylated form is a potent inhibitor of PP-2A and to In conclusion, our experiments shows that PP-2A plays a a lesser extent of PP-1 (refs. 24 and 44, but see also ref. 25). prominent role in the regulation of GC–PC synaptic efficacy. Our study completes the demonstration by Eto et al. (15) that, Unraveling the full extent of its involvement in cerebellar LTD in immature mouse PCs (9–16 DIV), a complex with PP-1 at its will require a better understanding of its interactions with core, plays a major role in cerebellar LTD, whereas PP-2A protein kinases and other of the postsynaptic complex, inhibitors had little effect at this age (ref. 15 and this study). This probably through its regulatory subunits. finding raises the interesting hypothesis that the relative contri- bution of PP-2A and other phosphatases might change during We thank the late Dr. R. T. Kado for his helpful advice; Drs. T. Takeuchi, cerebellar maturation, with PP-1-related phosphatase regulating H. Hirai, I. Tarnawa, and M. Chinkers for their generous gifts of synaptogenesis and plasticity at early developmental stages, plasmids, antibodies, and reagents; Dr. N. Murphy for editorial com- whereas PP-2A would play a dominant role only at more mature ments; and Mr. T. Torashima for technical help. This work was supported stages. It should be noted that the period between the second in part by Japan Society for the Promotion of Science Grants-in-Aid

and third postnatal week in vivo corresponds to an extensive 12780612 (to T.L.) and 12680790 and 13041064 (to S.E.). NEUROSCIENCE

1. Ito, M. (1989) Annu. Rev. Neurosci. 12, 85–102. 26. Lev-Ram, V., Jiang, T., Wood, J., Lawrence, D. S. & Tsien, R. Y. (1997) Neuron 2. Lisberger, S. G. (1998) Cell 92, 701–704. 18, 1025–1038. 3. Linden, D. J. & Connor, J. A. (1995) Annu. Rev. Neurosci. 18, 319–357. 27. Hartell, N. A. (1994) NeuroReport 5, 833–836. 4. Levenes, C., Daniel, H. & Crepel, F. (1998) Prog. Neurobiol. 55, 79–91. 28. Furuya, S., Makino, A. & Hirabayashi, Y. (1998) Brain Res. Brain Res. Protoc. 5. Ito, M. (2001) Physiol Rev. 81, 1143–1195. 3, 192–198. 6. Matsuda, S., Mikawa, S. & Hirai, H. (1999) J. Neurochem. 73, 1765–1768. 29. Larkum, M. E., Launey, T., Dityatev, A. & Luscher, H. R. (1998) J. Neuro- 7. Matsuda, S., Launey, T., Mikawa, S. & Hirai, H. (2000) EMBO J. 19, 2765–2774. physiol. 80, 924–935. 8. Xia, J., Chung, H. J., Wihler, C., Huganir, R. L. & Linden, D. J. (2000) Neuron 30. Hirai, H., Yoshioka, K. & Yamada, K. (2000) Brain Res. Brain Res. Protoc. 5, 28, 499–510. 182–189. 9. Chung, H. J., Steinberg, J. P., Huganir, R. L. & Linden, D. J. (2003) Science 31. Shenolikar, S. & Ingebritsen, T. S. (1984) Methods Enzymol. 107, 102–129. 300, 1751–1755. 32. Hirano, T. & Kasono, K. (1993) J. Neurophysiol. 70, 1316–1325. 10. Wang, Y. T. & Linden, D. J. (2000) Neuron 25, 635–647. 33. Isope, P. & Barbour, B. (2002) J. Neurosci. 22, 9668–9678. 11. Ito, M. & Karachot, L. (1992) Neurosci. Res 14, 27–38. 34. Turner, K. M., Burgoyne, R. D. & Morgan, A. (1999) Trends Neurosci. 22, 12. Ajima, A. & Ito, M. (1995) NeuroReport 6, 297–300. 459–464. 13. Kasahara, J. & Sugiyama, H. (1998) Neurosci. Lett. 247, 139–142. 35. Kawada, M., Amemiya, M., Ishizuka, M. & Takeuchi, T. (1999) Biochim. 14. Fujii, H. & Hirano, T. (2002) Eur. J. Neurosci. 16, 1777–1788. Biophys. Acta 1452, 209–217. 15. Eto, M., Bock, R., Brautigan, D. L. & Linden, D. J. (2002) Neuron 36, 36. Walsh, A. H., Cheng, A. & Honkanen, R. E. (1997) FEBS Lett. 416, 230–234. 1145–1158. 37. Hirai, H., Launey, T., Mikawa, S., Torashima, T., Yanagihara, D., Kasaura, T., 16. Mulkey, R. M., Endo, S., Shenolikar, S. & Malenka, R. C. (1994) Nature 369, Miyamoto, A. & Yuzaki, M. (2003) Nat. Neurosci. 6, 869–876. 486–488. 38. Petralia, R. S., Wang, Y. X., Mayat, E. & Wenthold, R. J. (1997) J. Comp. 17. Winder, D. G. & Sweatt, J. D. (2001) Nat. Rev. Neurosci. 2, 461–474. Neurol. 385, 456–476. 18. Abe, H., Tamura, S. & Kondo, H. (1992) Brain Res. Mol. Brain Res. 13, 283–288. 39. Herzig, S. & Neumann, J. (2000) Physiol Rev. 80, 173–210. 19. Hashikawa, T., Nakazawa, K., Mikawa, S., Shima, H. & Nagao, M. (1995) 40. Chan, S. F. & Sucher, N. J. (2001) J. Neurosci. 21, 7985–7992. Neurosci. Res. 22, 133–136. 41. Hirano, T. (1990) Neurosci. Lett. 119, 141–144. 20. Usuda, N., Arai, H., Sasaki, H., Hanai, T., Nagata, T., Muramatsu, T., Kincaid, 42. Linden, D. J. (2001) Proc. Natl. Acad. Sci. USA 98, 14066–14071. R. L. & Higuchi, S. (1996) J. Histochem. Cytochem. 44, 13–18. 43. Millward, T. A., Zolnierowicz, S. & Hemmings, B. A. (1999) Trends Biochem. 21. Strack, S., Zaucha, J. A., Ebner, F. F., Colbran, R. J. & Wadzinski, B. E. (1998) Sci. 24, 186–191. J. Comp. Neurol. 392, 515–527. 44. Endo, S., Nairn, A. C., Greengard, P. & Ito, M. (2003) Neurosci. Res. 45, 79–89. 22. Bahl, R., Bradley, K. C., Thompson, K. J., Swain, R. A., Rossie, S. & Meisel, 45. Muller, Y. L. & Yool, A. J. (1998) Brain Res. Dev. Brain Res. 108, 193–203. R. L. (2001) Brain Res. Mol. Brain Res. 90, 101–109. 46. Liljelund, P., Netzeband, J. G. & Gruol, D. L. (2000) J. Neurosci. 20, 7394–7403. 23. Detre, J. A., Nairn, A. C., Aswad, D. W. & Greengard, P. (1984) J. Neurosci. 47. Narasimhan, K., Pessah, I. N. & Linden, D. J. (1998) J. Neurophysiol. 80, 4, 2843–2849. 2963–2974. 24. Endo, S., Suzuki, M., Sumi, M., Nairn, A. C., Morita, R., Yamakawa, K., 48. Linden, D. J. & Connor, J. A. (1992) Eur. J. Neurosci. 4, 10–15. Greengard, P. & Ito, M. (1999) Proc. Natl. Acad. Sci. USA 96, 2467–2472. 49. Holmes, S. E., O’Hearn, E. E., McInnis, M. G., Gorelick-Feldman, D. A., 25. Hall, K. U., Collins, S. P., Gamm, D. M., Massa, E., DePaoli-Roach, A. A. & Kleiderlein, J. J., Callahan, C., Kwak, N. G., Ingersoll-Ashworth, R. G., Sherr, Uhler, M. D. (1999) J. Biol. Chem. 274, 3485–3495. M., Sumner, A. J., et al. (1999) Nat. Genet. 23, 391–392.

Launey et al. PNAS ͉ January 13, 2004 ͉ vol. 101 ͉ no. 2 ͉ 681 Downloaded by guest on September 25, 2021