AKAP150-anchored PKA activity is important for LTD during its induction phase

Abstract

Protein kinase A (PKA) is thought to tonically maintain an enhanced level of postsynaptic AMPA receptor responses. Injection of PKA inhibitory peptides leads to a run-down of AMPA receptor responses and prevents long-term depression (LTD). This run-down of AMPA receptor activity was proposed to occlude a further reduction that would otherwise constitute LTD. PKA is recruited to postsynaptic sites by the A kinase anchor protein AKAP150. We found that LTD was strongly impaired in acute hippocampal slices from 2-week-old mice in which the PKA binding site on AKAP150 had been genetically deleted (D36 mice). However, basal postsynaptic AMPA and NMDA receptor activity was indistinguishable between D36 and WT mice. During extracellular recordings of field EPSPs and during intracellular recording of EPSCs from hippocampal slices from WT mice, H-89 and KT5720, two structurally different PKA inhibitors, inhibited LTD by more than 70% without affecting basal synaptic transmission or basal phosphorylation of serine 845 on GluR1. Collectively our data indicate that AKAP150-anchored PKA activity is required to induce LTD and not merely to maintain a tonically heightened activity level of AMPA receptors as proposed earlier.

Numerous findings indicate that LTP and LTD underlie learning and memory (Martin et al. 2000; Collingridge et al. 2004; Whitlock et al. 2006). Yet the precise molecular mechanisms of LTP and LTD remain unclear. The protein phosphatases PP1 and PP2B are critical for LTD (Mulkey et al. 1994). PP2B dephosphorylates the PKA phosphorylation site of inhibitor-1, which in turn relieves PP1 inhibition by phosphorylated inhibitor-1 (Mulkey et al. 1994). However, PP1 is not sufficient for LTD. Injection of PP1 into CA1 pyramidal cells does not alter basal synaptic transmission (Morishita et al. 2001) despite being competent to modulate postsynaptic function as it increased LTD following a weak LTD induction protocol in parallel experiments (Morishita et al. 2001). These findings raise the question of which other regulatory factors are required for LTD.

LTD and GluR1 internalization, which contributes to LTD, are triggered by Ca²⁺ influx through the NMDAR and are to some degree due to dephosphorylation of GluR1 on S845, a major PKA phosphorylation site (Kameyama et al. 1998; Ehlers, 2000; Lee et al. 2000, 2003; Hu et al. 2007). Postsynaptic injection of highly specific PKA inhibitory PKI peptide or Ht31 peptide, which generically displaces PKA from the different AKAPs, leads to a run-down of AMPAR responses (Rosenmund et al. 1994; Kameyama et al. 1998; Snyder et al. 2005). Subsequently, LTD cannot be induced perhaps because the AMPAR run-down occludes LTD by sharing the same mechanism (Kameyama et al. 1998; Snyder et al. 2005), possibly dephosphorylation of critical PKA sites such as S845 on GluR1 (Lee et al. 2000, 2003).

S845 phosphorylation promotes surface expression of GluR1 (Swayze et al. 2004; Sun et al. 2005; Gao et al. 2006; Oh et al. 2006). It is important for functional expression of GluR1-containing AMPAR at postsynaptic sites during LTP (Esteban et al. 2003; Oh et al. 2006; Hu et al. 2007). Furthermore, PKA decreases internalization of GluR1 and increases its recycling back to the plasma membrane (Ehlers, 2000; Sun et al. 2005; Man et al. 2007). However, under basal conditions only 10–15% of GluR1s are phosphorylated on S845 (Oh et al. 2006). This low level of basal S845 phosphorylation is further supported by the observation that massive activation of adenylyl cyclases by forskolin induces a nearly 10-fold increase in total S845 phosphorylation in hippocampal slices (Boehm et al. 2006; Oh et al. 2006). Dephosphorylation of S845 and the resulting loss of postsynaptic AMPAR might thus account only for a portion of LTD.

It is conceivable that the PKI and Ht31 peptides block LTD not by an occlusion mechanism that prevents further decreases. We found that LTD was inhibited although the amplitude of AMPAR mEPSC was undiminished in AKAP150 D36 mice, in which the PKA binding site of AKAP150 was deleted. This mutation displaces more than 70% of PKA from postsynaptic sites (Lu et al. 2007). We further demonstrate that two different membrane-permeant inhibitors of PKA inhibit LTD even though they do not cause a run-down of basal synaptic transmission during extracellular recordings, nor do they lead to decreased S845 phosphorylation under basal conditions. Accordingly, PKA activity is required for LTD to trigger molecular changes that actively induce LTD in parallel with PP1 and PP2B.

Methods

Animals

All mice were decapitated with an appropriate guillotine without anaesthesia before collection of brains and production of hippocampal slices. All animal procedures had been approved by the University of Iowa Animal Care and Use Committee and followed NIH guidelines.

The generation of AKAP150 mutant mice and their genotyping is described in (Lu et al. 2007). Briefly, TCTTAA in the mouse AKAP150 gene (GenBank locus XM138063 position 2126–2131) was mutated to TCTAGA to introduce a stop mutation (underlined). The neomycin phosphotransferase gene (positive selection) was flanked by loxP sites and introduced into an Hpa1 site that is about 0.8 kB downstream of the Xba1 site that marks the stop codon defining the truncated AKAP150 coding region. This selectable marker was then removed by breeding mutants to MORE mice also known as MEOX2-Cre mice (Tallquist & Soriano, 2000). Mice were back-crossed 8 times with WT C57BL/6 (Taconic Farms) to obtain a nominally > 99% genetic C57BL/6 background. Heterozygous male and female mice were bred to obtain homozygous control and D36 mice. For genotyping, tail clips were digested with RNase A and proteinase K, DNA strands isolated and the 3′ coding region of AKAP150 amplified with primers correspond to sequences upstream and downstream of the new stop codon (5′-CCCACAGATACAGAGAAACCGAG-3′ and 5′-GGAAACGAAGTCACTGGAACAGCG-3′). The 400 bp product was purified and digested with Xba I to test for the newly created Xba1 restriction site in mice with D36 mutation, which gave cleavage products of 285 bp and 115 bp.

Hippocampal slice preparation

Slices were prepared as in Lu et al. (2007). In short, brains were rapidly sectioned in ice-cold slicing buffer (in mm: 127 NaCl, 26 NaHCO₃, 1.2 KH₂PO₄, 1.9 KCl, 1 CaCl₂, 2 MgSO₄, 10 dextrose, saturated with 95% O₂ and 5% CO₂) with a vibrating microtome (VT1000S, Leica Microsystems, Nussloch, Germany), kept in this buffer for 30 min at 34°C and for another 30 min at 22°C, and transferred to a submersion-type recording chamber perfused with 32°C oxygenated artificial cerebrospinal fluid (ACSF; in mm: 127 NaCl, 26 NaHCO₃, 1.2 KH₂PO₄, 1.9 KCl, 2.2 CaCl₂, 1 MgSO₄, 10 dextrose) at a rate of 2 ml min⁻¹.

Electrophysiology

Only 10- to 14-day-old male homozygous D36 or litter-matched WT control mice were compared. The PKA inhibitor experiments were performed with C57black/6 from Taconic Farms (Hudson, NY, USA). For extracellular recording ACSF-filled glass electrodes (resistance < 1 MΩ) were localized in the CA1 stratum radiatum. Schaffer collaterals were stimulated with 0.1 ms pulses with a bipolar tungsten electrode (WPI Inc., Sarasota, FL, USA) once every 15 s (Lim et al. 2003). The stimulation intensity was titrated to define the maximal field excitatory postsynaptic potential (fEPSP) slope and then reduced to yield 40–60% of the maximal response. If maximal fEPSPs were less than 0.5 mV or if fibre volleys showed substantial changes, experiments were rejected. After a stable baseline response was obtained for 15–25 min, LTD was induced by 900 pulses delivered with a frequency of 1 Hz. For depotentiation, LTP was induced by one stimulus train at 100 Hz for 1 s followed by one train of 900 stimuli at 1 Hz applied 5 min later. The following PKA inhibitors were added to the perfusion buffer if indicated: N-[2-(p-bromocinnamylamino)ethyl]-5-isoquinolinesulphonamide (H-89; Sigma; stock solution: 80 mm in DMSO; final concentration: 20 μm); KT5720 (A.G. Scientific, San Diego, CA, USA; stock solution: 1 mm in DMSO; final concentration: 1 μm). The fEPSPs were recorded with an AxoClamp 2B amplifier (Axon Instruments, Union City, CA, USA), filtered at 1 kHz, digitized at 10 kHz via Axon Digidata 1200, and analysed with Clampfit 9 in a Windows XP environment. The initial slopes of the fEPSP were calculated for each experiment and expressed as a percentage of the baseline average. Each point in the summary graphs reflects the average of four consecutive responses. The time-matched, normalized data were averaged for all experiments and expressed as means ±s.e.m. For statistical analysis, the last 5 min in each measurement were averaged and compared with the corresponding controls (t test; n equals the number of slices; each experimental group contains multiple slices from at least 3 different mice).

Miniature EPSCs (mEPSCs) were monitored by whole-cell recordings from pyramidal neurons in the hippocampal CA1 region at a holding potential of −70 mV. Patch pipettes (3–6 MΩ) were pulled from KG-33 glass capillaries (1.1 mm i.d., 1.7 mm o.d., Garner Glass Company, Claremont, CA, USA) on a Flaming–Brown electrode puller (P-97, Sutter Instruments Co., Novato, CA, USA) and filled with intracellular solution (in mm: 125 potassium gluconate, 20 KCl, 10 NaCl, 2 Mg-ATP, 0.3 Na-GTP, 2.5 QX314, 10 Pipes, 0.2 EGTA, pH 7.3 adjusted with KOH). Slices were perfused with ACSF. Fifty micromolar picrotoxin, 10 μm APV, and 1 μm tetrodotoxin (TTX, all Sigma) were included to block currents from GABA_A receptors, NMDA receptors, and voltage-gated Na⁺ channels, respectively. Recordings were obtained with an Axopatch 200B amplifier (Axon Instruments), filtered at 1 kHz, digitized at 10 kHz via an Axon Digidata 1322A, and analysed under a Windows XP environment. The threshold for mEPSCs was 5 pA.

Like mEPSCs, evoked EPSCs (eEPSCs) were recorded in the whole-cell configuration at a holding potential of −70 mV to monitor AMPA receptor currents. Slices were perfused with ACSF without any blocker. The stimulation intensity was titrated to define the maximal eEPSC amplitude and then reduced to yield 40–60% of the maximal response. LTD was induced by 900 pulses delivered with a frequency of 1 Hz. H-89 (20 μm) and KT5720 (1 μm) were added as indicated.

Cell culture

Primary cultures of hippocampal neurons were prepared as previously described (Fedulova et al. 2000). In brief, hippocampi from newborn mice were dissected, treated with trypsin (0.5 mg ml⁻¹; 15 min, 22°C), washed (modified Eagle's medium (MEM; Invitrogen) containing 5% Horse serum (HS) and 5% fetal bovine serum (FBS)), and triturated with fire-polished Pasteur pipettes. The cells were plated onto poly l-ornithine–laminin-coated 25 mm coverslips and cultured in MEM–5% HS–5% FBS at 37°C, 5% CO₂ for 2–3 weeks before use. In order to suppress growth of non-neuronal cells, 3 days after plating cultures were treated with 5 mm of cytosine-A–d-arabino-furanoside for 24 h. One-third of the medium was replaced every week.

[Ca²⁺]_i imaging

Hippocampal neurons were placed in a flow-through chamber on an inverted microscope IX-71 equipped with a 60× objective (NA = 1.40; Olympus). The Ca²⁺-sensitive fluorescent indicator Oregon Green 488 BAPTA-1 (OGB-1) was loaded via a patch pipette (filling in mm: 125 potassium gluconate, 10 KCl, 3 Mg-ATP, 1 MgCl₂, 10 Hepes, 0.15 OGB-1, pH 7.25 with KOH, 290 mosmol kg⁻¹ with sucrose) in the whole-cell configuration for 5 min before withdrawal of the pipette. OGB-1 is widely used to monitor Ca²⁺ signalling in dendritic spines because of its high quantum yield, resistance to photobleaching, and relatively high signal even at resting [Ca²⁺]_i levels (Mainen et al. 1999; Kovalchuk et al. 2000; Lambe & Aghajanian, 2003; Nevian & Sakmann, 2004). Recordings began 30–40 min after withdrawal of the patch pipette. Na⁺ channels were blocked with TTX (0.5 μm) and AMPA receptors with CNQX (10 μm). Mg²⁺ was removed for optimal NMDA receptor activity. OGB-1 fluorescence was excited at 475 nm (10 nm bandpass) and imaged at 530 nm (25 nm bandpass) at a sampling frequency of 10 Hz using a CCD-camera-based imaging system (Till Photonics, Graefelfing, Germany). Fluorescence was corrected for background and expressed as ΔF/F= (F–F₀)/F₀, where F was fluorescence intensity over time and F₀ was fluorescence intensity at t= 0. Data analysis was performed with TILLvisION 4.0 software (Till Photonics).

Results

LTD but not depotentiation is reduced in D36 mice

PKA-mediated regulation of various ion channels requires anchoring of PKA by specific AKAPs (Wong & Scott, 2004). AKAP150 (AKAP5) is the main AKAP at glutamatergic postsynaptic sites (Lu et al. 2007). We showed earlier that the postsynaptic scaffolding protein SAP97 directly interacts with both the AMPAR GluR1 subunit (Leonard et al. 1998) and AKAP150 to connect the PKA-AKAP150 complex to GluR1 for S845 phosphorylation and thereby up-regulation of GluR1 channel activity (Tavalin et al. 2002; Wong & Scott, 2004) and surface expression (Ehlers, 2000; Swayze et al. 2004; Sun et al. 2005; Oh et al. 2006). Like SAP97, PSD-95 also recruits the AKAP150–PKA complex to postsynaptic sites via binding AKAP150 with its SH3 domain (Colledge et al. 2000; Xu et al. 2008).

To investigate whether PKA anchoring by AKAP150 is necessary for LTD, we truncated the C-terminal residues of AKAP150 by introducing a stop codon in the mouse genome at the respective position. These 36 residues interact with the dimers that are formed by type II regulatory R subunits of PKA, each of which binds one catalytic C subunit (Wong & Scott, 2004). D36 deletion reduces postsynaptic PKA by more than 70% without changing expression or postsynaptic localization of AKAP150 (Lu et al. 2007) or affecting binding sites for other AKAP150-associated molecules including F-actin, cadherin, calmodulin, PKC, phosphatidylinositol-4,5-bisphosphate, and calcineurin (Gomez et al. 2002; Gorski et al. 2005).

LTD was strongly impaired in acute hippocampal slices from D36 mice compared to litter-matched WT mice at 10–14 days of age when LTD is most obvious (Fig. 1A and B). Specifically, the initial fEPSP slope was reduced to 62 ± 3% in WT mice and 90 ± 2% in D36 mice 60 min after LTD induction.

Figure 1. LTD but not depotentiation is inhibited in 2 week old D36 mice.

A, example of fEPSP recordings before (dashed lines) and 60 min after LTD induction (continuous lines) in the CA1 area of acutely prepared hippocampal slices from 10- to 14-day-old homozygous D36 or litter-matched WT mice. B, LTD was induced by a stimulus train of 1 Hz for 15 min (bar). Plotted are averages of the initial slope of the fEPSPs of multiple recordings for each time point normalized for each recording (100% equals the average of the recorded fEPSP slopes during the 15 min interval immediately preceding LTD induction). LTD is substantially decreased in D36 mice (90 ± 2% mean ± s.e.m. at 70–75 min, ○) versus littermate-matched WT mice (62 ± 3% at 70–75 min, •). This decrease is highly significant (P < 0.0001, t test). C, LTP was induced in slices from 3-week-old D36 or litter-matched WT mice by one train of 100 stimuli at 100 Hz (arrow). Five minutes later depotentiation was induced by one train of 900 stimuli at 1 Hz (black bar). The initial potentiation and the subsequent depotentiation were comparable in D36 and littermate-matched WT mice (D36 mice: 130 ± 3% at 5 min; 101 ± 3% at 75–80 min, ○; WT mice: 134 ± 4% at 5 min; 101 ± 4% 75–80 min, •). The 100 Hz tetanus reliably induces LTP in 3-week-old mice in our hands (Lu et al. 2007).

Unlike LTD, depotentiation of LTP correlates with dephosphorylation of phosphorylation sites for CaMKII rather than for PKA (Barria et al. 1997; Lee et al. 2000). To test whether depotentiation is affected in D36 mice we first applied one 1 s/100 Hz tetanus, which reliably induces LTP (Lu et al. 2007), before a 1 Hz/900 stimulus train to reverse LTP. This reversal was indistinguishable between WT and D36 mice (Fig. 1C). Accordingly, PKA anchoring by AKAP150 is important for LTD but not depotentiation.

Because S845 phosphorylation increases GluR1 surface expression and activity, D36 deletion could cause a reduction in postsynaptic AMPAR activity, which in turn could occlude LTD induction. However, GluR1 was undiminished in subcellular postsynaptic density fractions from D36 mice as compared to WT mice (Lu et al. 2007). Furthermore, plots of the initial slope of the field EPSP (fEPSP) reflecting the strength of the postsynaptic response against the fibre volley amplitude reflecting the degree of presynaptic activation did not show any difference between D36 and WT mice (Fig. 2A). We also monitored mEPSC amplitudes of AMPAR by whole cell patch recording from CA1 pyramidal neurons. There was no change in amplitude or decay time constants of mEPSCs (Fig. 2B–D). Amplitudes of NMDAR-mediated spontaneous Ca²⁺ transients in dendritic spines of primary hippocampal cultures 14–16 DIV were also indistinguishable between D36 and WT mice (Fig. 3A, B and E). Similarly, NMDAR-mediated mEPSCs were unaltered in neuronal cultures from D36 mice (Fig. 3C, D and F). These findings indicate that basal synaptic transmission and specifically AMPAR and NMDAR activity are unaltered in D36 versus WT mice.

Figure 2. Postsynaptic AMPAR responses are normal in D36 mice.

A, indistinguishable input–output curves were obtained by plotting the initial slope of the fEPSPs as a function of the presynaptic fibre volley amplitude for D36 and litter-matched WT mice. B, examples of AMPA receptor mEPSC recordings from litter-matched WT (a) and D36 mice (b). C and D, mEPSC amplitudes (C) and decay times (D) are unaltered in 2-week-old D36 versus litter-matched WT mice as illustrated by cumulative fraction plots (Ca and Da), histogram distributions (inserts) and overall averages (Cb and Db).

Figure 3. Postsynaptic NMDAR responses are normal in D36 mice.

NMDAR mediated spontaneous [Ca²⁺]_i transients and mEPSCs were recorded from cultured hippocampal neurons (14–16DIV) from WT (A and C) and D36 (B and D) littermates in the absence of Mg²⁺ with TTX (0.2 μm) and CNQX (10 μm) present. A and B, traces of [Ca²⁺]_i transients from OGB-1 loaded neurons from two (A) and three (B) spines and corresponding dendritic shafts (left images) are shown on the same time and signal scales (e.g. s1 and d1). Spine signals were highly localized without spreading to dendritic shafts or neighbouring spines (e.g. s1 and s2 in B). The NMDAR antagonist dl-AP5 (50 μm) completely abolished the [Ca²⁺]_i transients (right traces; n= 17, 5 independent experiments). C and D, NMDAR mEPSCs were recorded with GABA_A receptors blocked by 10 μm bicuculline. E, amplitude distributions of [Ca²⁺]_i minispikes were virtually identical in D36 (31 spines, 7 cells) and WT mice (29 spines, 6 cells; P > 0.05, Kolmogorov–Smirnov test). The corresponding mean amplitudes were 15 ± 2%ΔF/F in WT and 17 ± 3%ΔF/F in D36 mice (P= 0.35; Student's t test). F, cumulative probability distributions of NMDAR mEPSC amplitudes were virtually identical in D36 (4 cells) and WT mice (6 cells; P > 0.05, Kolmogorov–Smirnov test).

Paired pulse facilitation refers to the phenomenon that in CA1 pyramidal neurons the postsynaptic response to a second presynaptic stimulus is increased if the time passed since the first stimulus is less than 1 s. Alterations in presynaptic neurotransmitter release properties would lead to changes in paired pulse facilitation (Zalutsky & Nicoll, 1990; Schulz et al. 1994; Jensen et al. 2003; Han et al. 2006). However, paired pulse facilitation was unaffected in D36 versus WT mice over the whole range of interpulse intervals (Fig. 4). These data suggest that the LTD deficit is due to a postsynaptic rather than presynaptic mechanism.

Figure 4. Paired-pulse facilitation is normal in D36 mice.

Paired pulse facilitation was determined in acute hippocampal slices from 2-week-old D36 and litter-matched WT mice for interstimulus intervals ranging from 10 to 1000 ms without any obvious statistically significant difference.

The finding that LTD was reduced despite undiminished postsynaptic AMPA receptor responses in D36 versus WT mice indicates that the role of PKA in LTD is not restricted to maintaining an elevated level of basal postsynaptic AMPAR activity until induction of LTD. In other words, PKA is not just there to keep phosphorylation levels of certain PKA sites such as S845 at an increased level under basal conditions. We hypothesized that PKA might play an additional role in LTD by regulating yet unknown signalling pathways at the postsynaptic site during LTD induction. Application of two different membrane-permeant PKA inhibitors, H-89 and KT5720, to hippocampal slices from WT mice did not reduce basal field EPSPs even during extended incubation periods (Fig. 5A). Despite the lack of inhibition of basal synaptic transmission, both PKA inhibitors markedly reduced LTD (Fig. 5B and C). Specifically, in these experiments postsynaptic responses were reduced to 65 ± 1% (vehicle control for H-89 experiments) and 67 ± 2% (vehicle control for KT5720 experiments) 55–60 min after LTD induction under control conditions. In slices treated with H-89 or KT5720 the initial fEPSP slope was only reduced to 91 ± 4% and 88 ± 3%, respectively, 55–60 min after LTD induction.

Figure 5. Bath-applied PKA inhibitors reduce LTD of fEPSPs in 2-week-old WT mice.

A, H-89 (20 μm in DMSO) or KT5720 (1 μm in DMSO) does not affect baseline recordings (96 ± 5% and 95 ± 2%, respectively, at 55–60 min). B, H-89 inhibits LTD (91 ± 4%versus 65 ± 1% for vehicle control at 70–75 min: P < 0.0001; t test). C, KT5720 inhibits LTD (88 ± 3%versus 67 ± 2% for vehicle control at 70–75 min: P < 0.0001; t test). Open bars: 1 Hz/900 s stimulations; filled bars: drug applications.

These observations contrast with earlier findings that injection of PKI peptide, a specific PKA blocker, or Ht31 peptide, which removes PKA from AKAPs, causes run-down in EPSCs as the peptide diffuses throughout the dendrites and blocks LTD induction (Rosenmund et al. 1994; Kameyama et al. 1998; Snyder et al. 2005). It is conceivable that the whole cell recordings used in these earlier studies might have promoted a negative effect of PKA inhibition on postsynaptic responses by washing out essential cellular factors. Alternatively, bath application of H89 or KT5720 might inhibit PKA to a lesser extent than intracellular injection of those two peptides. To evaluate these alternatives and to provide further evidence for a role of PKA during and not only prior to the induction of LTD, we monitored EPSCs and LTD induction during bath application of H89 and KT5720 (Fig. 6). Neither inhibitor affected basal EPSCs over a time period of 20 min yet both inhibited LTD induction. This inhibition confirms that PKA activity is required for induction of LTD. The absence of an effect by either PKA inhibitor on baseline activity is corroborated by the fact that EPSC magnitudes returned to original levels after a temporary reduction during LTD induction. Accordingly, bath application of H89 or KT5720 does not affect basal AMPAR activity even during whole cell patch recordings, contrasting intracellular injection of PKI or Ht31 peptide.

Figure 6. Bath-applied PKA inhibitors reduce LTD of EPSCs in 2-week-old-WT mice.

EPSCs and input (R_i) and series (R_s) resistances were monitored by whole cell patch recordings. A, LTD was readily induced with vehicle (DMSO) present (61 ± 3% at 70–75 min). B, H-89 (20 μm in DMSO) blocks LTD (101 ± 10% at 70–75 min; mean ±s.e.m.; P < 0.002; t test). C, KT5720 (1 μm in DMSO) blocks LTD (102 ± 3% at 70–75 mi; P < 0.000004; t test). Open bars: 1 Hz/900 s stimulations; black bars: drug applications.

To test whether bath application of H89 or KT5720 affected basal phosphorylation of GluR1 on S845, acute slices were treated for 40 min with either inhibitor before extraction with Triton X-100, immunoprecipitation with anti-GluR1, and immunoblotting with our phosphospecific antibody against S845, followed by reprobing with our general GluR1 antibody, as described (Lu et al. 2007). Although there was a tendency to reduced phosphorylation of S845, this tendency was small and statistically not significant (Fig. 7). It thus appears that bath application of H89 or KT5720 does not affect S845 phosphorylation in intact neurons.

Figure 7. Bath-applied PKA inhibitors do not reduce basal phosphorylation of GluR1 on serine 845.

A, acute slices from 13- to 17-day-old WT C57BL/6 mice were equilibrated in oxygenated ACSF at 30°C, treated with vehicle control (DMSO), H89 (20 μm in DMSO), or KT5720 (1 μm in DMSO) for 40 min, extracted with ice-cold 1% Triton X-100 buffer, and centrifuged for 30 min at 50 000 g. GluR1 was immunoprecipitated before immunoblotting with a phosphospecific antibody against S845, followed by reprobing with a general GluR1 antibody. The figure shows a representative blot from a total of 5 experiments. B, immunoblot signals were quantified after digitalization. Phospho-S845 signals were corrected with respect to total GluR1 signals and normalized with control (DMSO) equaling 100%. Shown are means ±s.e.m. from 5 experiments.

Discussion

PKA maintains heightened AMPAR responses under basal conditions, which is reversed upon injection of a PKI or Ht31 peptide, potentially occluding LTD (Rosenmund et al. 1994; Kameyama et al. 1998; Snyder et al. 2005). In this context PKA might act at least in part by phosphorylating GluR1 on S845, which is important for GluR1 accumulation at the neuronal cell surface and specifically at the synapse during LTP (Esteban et al. 2003; Swayze et al. 2004; Sun et al. 2005; Gao et al. 2006; Oh et al. 2006; Hu et al. 2007). However, only 10–15% of GluR1 are phosphorylated on S845 under basal conditions (Boehm et al. 2006; Oh et al. 2006). Furthermore, basal surface expression of GluR1 is not affected in S831/845A knock-in mice suggesting that under basal conditions surface expression is not regulated by phosphorylation of S845 (Lee et al. 2003; see also Ehlers, 2000; Sun et al. 2005; Man et al. 2007). Therefore, it appears unlikely that inhibition of basal S845 phosphorylation alone is responsible for LTD, which typically leads to a 40–50% reduction in postsynaptic responses.

Perfusions with H-89 or KT5720 did not affect basal synaptic transmission even over an extended period of time (Figs 5 and 6) indicating that neither inhibitor affects postsynaptic responses over the duration of the LTD experiments when bath applied. We showed earlier that H-89 and KT5720 block LTP induction in slices from 8-week-old mice (Lu et al. 2007). These LTP experiments and the LTD experiments presented here were performed in parallel. Accordingly, bath application of both inhibitors reduces PKA activity to a degree that is sufficient to inhibit LTP and LTD but not basal postsynaptic AMPAR responses in our hands. In fact bath application of either inhibitor has only a minimal if any effect on S845 phosphorylation (Fig. 7). Perhaps injections of PKI or Ht31 peptide inhibits PKA in complex with GluR1 more effectively with respect to S845 phosphorylation than bath applied H89 or KT5720. The whole cell patch clamp configuration might enhance efficacy of PKA blockers. In fact, block of LTD by H89 and KT5720 was complete during intracellular EPSC recording but incompletely during extracellular fEPSP recording (compare Fig. 5 and 6), possibly due to wash-out of factors important for phosphorylation.

Basal synaptic transmission is not affected in D36 mice. Consistent with this finding, GluR1 phosphorylation on S845 is not reduced in D36 mice under basal conditions (Lu et al. 2007). It is conceivable that compensatory mechanisms in D36 mice lead to readjustment of synaptic strength and thereby reinstate S845 phosphorylation even in the absence of PKA anchoring by AKAP150. Unanchored cytosolic PKA might phosphorylate S845 more effectively in D36 than WT mice and S845 phosphorylation might occur in part inside neurons rather than at the synapse. In fact, homeostatic mechanisms that are not well defined on a molecular level counteract chronic reduction of postsynaptic activity by increasing EPSCs via pre- and postsynaptic alterations (Turrigiano, 2007).

Be that as it may, our results clearly indicate an active role for PKA in LTD rather than exclusively maintaining a certain level of postsynaptic AMPAR response. However, neither H-89 nor KT5720 completely blocked LTD as measured by fEPSP recordings but rather inhibited about 75% of it. The same is true for the D36 mutation in AKAP150. Perhaps PKA contributes to certain portions or forms of LTD but is not absolutely critical for other portions. It is conceivable that the remaining 25% of LTD is in part due to reversal of PKA-mediated S845 phosphorylation, which is not accomplished in the D36 mice or by bath application of H-89 or KT5720.

Several pieces of evidence indicate that LTD is at least in part mediated by endocytosis of AMPARs (Beattie et al. 2000; Ehlers, 2000; Lin et al. 2000; Snyder et al. 2001). As discussed above, PKA-mediated S845 phosphorylation on GluR1 supports AMPAR expression at postsynaptic sites. At the first glance it thus appears counter-intuitive that PKA would help to down-regulate synaptic AMPAR function. However, LTD in dopaminergic neurons in the ventral tegmental area (VTA) also requires PKA (Gutlerner et al. 2002). In fact, in these neurons activation of PKA is sufficient to induce LTD (Gutlerner et al. 2002). Similarly, LTD in the cerebellum at the parallel fibre–Purkinje cell synapse is mediated by AMPAR internalization that requires PKC (Chung et al. 2003; Leitges et al. 2004) and also CaMKII (Hansel et al. 2006). Although the molecular mechanisms for the PKA-mediated LTD in the VTA and the PKC- and CaMKII-dependent LTD in the cerebellum are unknown it is now clear that in both cases one or more kinases are important during the induction phase for the down-regulation of the postsynaptic AMPA receptor response. An analogous mechanism might mediate a substantial portion of LTD at the most commonly studied CA3–CA1 synapse.

In summary our results demonstrate that the requirement for intact PKA signalling in LTD is not restricted to the maintenance of basal phosphorylation levels of postsynaptic substrates such as GluR1, as proposed earlier. We rather conclude that the activity of AKAP150-anchored PKA actively contributes to the induction of LTD. However, how PKA does so is currently unknown.