INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Three types of AMPA receptor modification have been advanced as potential substrates of long-term potentiation (LTP): increased numbers (Hayashi et al. 2000; Isaac et al. 1995; Lledo et al. 1998; Liao et al. 1995, 1999; Lynch and Baudry 1984), clustering (Xie et al. 1997), or changes in biophysical rate constants (Ambros-Ingerson et al. 1993). In accord with the first idea are reports that high-frequency afferent stimulation increases the number of AMPA receptors in adult hippocampus (Maren et al. 1993) or causes receptors to become incorporated into synapses in immature cultures (Hayashi et al. 2000; Shi et al. 1999). However, there is also evidence for induction of robust LTP in adult slices without detectable changes in receptor density (Kessler et al. 1991). The clustering and biophysical hypotheses were prompted by evidence that LTP is accompanied by subtle changes in the waveform of the AMPA receptor-mediated component of the postsynaptic response (Ambros-Ingerson et al. 1991, 1993; Kolta et al. 1998; Stricker et al. 1996; Xie et al. 1997). Simulation work showed that this effect, along with increases in response size, can be reproduced by tighter packing of a fixed population of AMPA receptors, thereby altering the interaction between a transmitter packet and the receptor pool (Xie et al. 1997). However, other models revealed that simply increasing the opening/closing rates of the receptor channel generates waveform and amplitude effects of the kind seen with LTP (Ambros-Ingerson et al. 1993).

Changes in the rate constants governing a multi-state, nonlinear system such as AMPA receptors, should interact. Accordingly, the kinetic hypothesis for LTP predicts that AMPA receptor modulators will have different effects on control versus potentiated synapses. Evidence in support of this was obtained in experiments using aniracetam, a drug that slows desensitization and deactivation of AMPA receptors (Arai et al. 1995; Isaacson and Nicoll et al. 1991; Ito et al. 1990; Tang et al. 1991). LTP modestly reduced the effect of aniracetam on the amplitude of the AMPA receptor-mediated component of excitatory postsynaptic potentials (EPSPs) (Staubli et al. 1992; Xiao et al. 1991) while increasing its effect on the decay time constant (Kolta et al. 1998). However, other work did not find a statistically reliable interaction between LTP and cyclothiazide (CTZ), a drug with potent effects on AMPA receptor desensitization (Yamada and Tang 1993), leading the authors to argue against changes in AMPA receptor kinetics as possible mechanisms for LTP expression (Rammes et al. 1999). In the 1999 study, LTP was induced via tetanic stimulation, and experiments were conducted at room temperature. In contrast, the above-cited studies involving LTP and aniracetam were conducted at physiological temperatures, and LTP was induced by theta burst stimulation, a method reported to be optimal for induction of LTP (Larson et al. 1986).

It has been proposed that different types of lasting potentiation are elicited by different experimental protocols (Abraham and Huggett 1997; McNaughton et al. 1994), and this might explain why induction of LTP would change the effects of a modulatory drug in one experiment and not in a second. The present study re-examined the effects of LTP on CTZ using induction conditions comparable to those employed in the aniracetam experiments. Additionally, a kinetic model of the AMPA receptor was employed to simulate experimental findings and test the hypothesis that LTP expression is associated with changes in AMPA receptor kinetics. The results provide new, quantitative constraints on hypotheses about the nature of the receptor changes responsible for the expression of potentiation.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Preparation and recording

Hippocampal slices (400 µm) were prepared from 15- to 25-day-old rats and placed in a holding chamber for at least 1 h before being transferred to a recording chamber. The slices were submerged in oxygenated artificial cerebrospinal fluid (ACSF) infused at 1.2 ml/min. The composition of the ACSF was (in mM) 124 NaCl, 3 KCl, 1.25 KH₂PO₄, 3.4 CaCl₂, 2.5 MgSO₄, 26 NaHCO₃, and 10 D-glucose, equilibrated with 95% O₂-5%CO₂ (pH 7.3). All experiments were carried out at 32°C.

CA1b pyramidal neurons were visualized with an infrared microscope (Olympus BX50WI, Olympus, Melville, NY) with DIC configuration, and whole cell recordings were made with 3-5 M recording pipettes containing (in mM) 130 Cs gluconate, 10 CsCl, 0.2 EGTA, 8 NaCl, 2 ATP, 0.3 GTP, and 10 HEPES (pH 7.35, 290-300 mosM). The liquid junction potential of the pipette solution was 6 mV with respect to the Ringer solution. Holding potentials were 70 mV after correcting for the junction potential. Excitatory postsynaptic currents (EPSCs) were recorded with a patch amplifier (AxoPatch-1D, Axon Instruments, Burlingame, CA) with a 4-pole low-pass Bessel filter at 2 kHz and digitized at 10 kHz. Two bipolar stimulating electrodes were placed in the stratum radiatum, one on each side of the recording locus. Synaptic responses were generated by stimulating the Schaffer collateral/commissural projections every 20 s. The stimulation intensity was adjusted so as to obtain <30% of the maximum amplitude. Input and series resistances were continuously monitored and recordings discarded if significant changes occurred.

Baseline responses were collected for 20 min after which a train of 10 "theta" bursts was applied to one of the two pathways. Each burst contained four pulses at 100 Hz, and the bursts themselves were separated by 200 ms; stimulation intensity was not increased during the bursts. The potentiated and control EPSCs were recorded for 30 min to ensure stability after which CTZ was added to the infusion line. The drug was dissolved in DMSO at 100 mM and diluted 1,000-fold before every experiment (100 µM, 0.1% DMSO in perfusion). Paired-pulse experiments were conducted under the same conditions using four paired-pulse protocols having interpulse intervals of 50, 80, 100, and 200 ms.

Similar methods were employed in experiments in which picrotoxin (50 µM) and 3-amino-propyl (diethoxymethyl)-phosphonic acid (CGP 35348, 100 µM) were used to block hyperpolarizing GABA_A and GABA_B currents, respectively. Picrotoxin and CGP 35348 were dissolved in ACSF and infused throughout the entire recording period. Lidocaine N-ethyl bromide (QX-314; 5 µM) was intracellularly applied to block voltage-gated sodium channels. An incision was routinely made between CA1 and CA3 in disinhibited slice preparations.

Miniature EPSCs were recorded and averaged to provide an estimate of a single synaptic excitatory event. mEPSCs were recorded using whole cell clamp techniques similar to those described in the preceding text. Additionally, tetrodotoxin (TTX, 1 µM) and picrotoxin (50 µM) were added to the ACSF to suppress action potential-dependent transmitter release and GABA_A inhibitory synaptic currents, respectively. TTX was dissolved in a 5 mM stock solution in 100% DMSO before being diluted to its final concentration in ACSF. mEPSCs were first extracted from the raw data using an automatic detection algorithm and then averaged. To do this, the raw data were smoothed using a moving average filter window consisting of 21 points. The first and second derivatives of the smoothed data were computed using time windows chosen to best represent the dynamics of the mEPSC events; the first derivative was computed considering a window of 1 ms, while the second derivative was computed considering a window of 20 ms. An mEPSC event was detected as a positive peak in the second derivative with a magnitude >8 pA/ms², the selected threshold. The detected events were then averaged to obtain the representative mEPSC waveform. Results of automatic selection/extraction were visually checked for agreement with expected shape and magnitude of mEPSCs.

2-Amino-5-phosphonopentanoic acid (AP5), QX-314, and CTZ were purchased from RBI (Natick, MA); [³H]6-cyano-7-nitro-quinoxaline-2,3-dione (CNQX) was purchased from NEN/DuPont (Boston, MA); and TTX and picrotoxin were purchased from Sigma (St. Louis, MO). CGP 35348 was provided free of charge by Novartis (Basel). Sprague-Dawley rats were obtained from Charles River (Wilmington, MA); the animals were housed, cared for, and killed according to an accredited protocol and guidelines set forth by the National Institutes of Health.

Statistics and estimations

Paired two-tailed t-tests were employed to test for significant effects in the first series of experiments, unless indicated otherwise. In the replications involving GABA antagonists, one-tailed t-tests were used in lieu of two-tailed tests because effects were predicted beforehand from the previous results.

Decay time constants were estimated from individual EPSCs. Records were examined to ensure that responses that were not monophasic, such as those with latent spikes, were excluded from calculation of decay time constants. Additionally, responses were required to meet least squares fitting criteria described in the following text.

The decay time constant was computed as  = 1/b, where a + bt is the linear least-squares fit to the data set
and I_i is the recorded current at time t_i, and t_p and I_p are the time and current at peak amplitude, respectively. This is equivalent to fitting a single exponential to the selected decaying portion, corresponding to 90-30% of the peak current. Decay time constants were not calculated for responses that did not meet the least squares fitting criteria (r² > 0.95).

Kinetic model of AMPA receptors

Simulations of the interaction between the effects of LTP and CTZ were carried out using a variant of previous kinetic models of the AMPA receptor (Ambros-Ingerson and Lynch 1993; Clements et al. 1998; Tour et al. 2000). Figure 1 depicts the states used in the current model and the associated rate constants linking the states. Various arguments suggest that each of the homologous subunits in the AMPA receptor possesses a binding site for glutamate, raising the possibility that the behavior of the receptor will depend on the number of agonist molecules it binds during a given release event. Experimental work indicates that at least two binding events are required to open the channel (Clements et al. 1998; Tour et al. 2000), and it has been proposed that distinct conductance states are associated with the number of bound molecules (Rosenmund et al. 1998). Previous simulations have shown that an AMPA receptor model with two independent binding sites better fits responses following application of high-affinity glutamate agonists than models incorporating additional binding sites (Clements et al. 1998). Thus the present simulation follows Tour et al. (2000) in simplifying the AMPA receptor model to consist of one and two ligand bound states (RL₁ and RL₂) with the latter leading to an open state (RL₂open). Desensitization, which is unusually rapid for AMPA receptors, occurs in the model from RL₁, RL₂, and RL₂open, as illustrated in Fig. 1A. The idea of desensitization proceeding from both a closed and an open state of the receptor is based on evidence regarding the kinetics of desensitization at various glutamate concentrations (Tour et al. 2000; Trussell and Fischbach 1989).

View larger version (23K): [in this window] [in a new window]   Fig. 1. Kinetic model of AMPA receptor. A: 7-state model in which R represents receptor and L represents ligand. R, RL1, and RL2 all represent closed states; RL1desens and RL2desens represent the 2 closed, desensitized states; and RL2open and RL2open desens represent the open state of the channel and the open but desensitized state, respectively. Note that the binding constants (k1, k2, and k7) are multiplied by the concentration of the ligand (10 mM). Decreases in the desensitization rate constants (k4, k5, and k6) that were implemented to model the effects of CTZ (see RESULTS section) are depicted as dashed arrows with smaller arrowheads, while increases in the rate constants leading to the open states (k3 and k8) that were implemented to simulate the effects of LTP (see RESULTS section) are depicted as dashed arrows with larger arrowheads. B: final values of rate constants that best fit the data following Nelder-Mead optimization with certain constraints. Note that k2 is half the value of k1 and k-2 is twice the value of k-1, consistent with the probability of binding when independence between the 2 binding sites is assumed.

Figure 1B lists the values of the rate constants used in the simulations reported here, while Table 1 lists comparable rate constant values used in earlier models. Experimental results from several techniques are available regarding association and dissociation rates of glutamate from AMPA receptors (k₁ and k_-1); the starting values chosen here are based on binding results with synaptosomal membranes from rat cortex and hippocampus [K_d for L-glutamate in the range 30-100 µM (Honore et al. 1989); K_d for L-glutamate of 200 nM at 1 binding site and 1 µM at a second binding site, (Werling et al. 1983)]. The latter study suggested that the two sites do not bind ligands with the same affinity; however, the present model assumes noncompetitive binding at the two sites, as in Tour et al. (2000). Channel opening and closing rates are based on average values from single channel studies of AMPA receptors (Dingledine et al. 1999, for review). There is reasonable agreement across experimental techniques regarding net desensitization and resensitization rates for the AMPA receptor; specifically, desensitization occurs more rapidly than resensitization (Dingledine et al. 1999). Thus the desensitization/resensitization kinetics in the present model are constrained to comply with this property. The model was further constrained to have a desensitization rate of ~8 ms, consistent with values reported for hippocampal cultures (7.5 ± 2.0 ms) (Vyklicky et al. 1991) and field CA1 in acute slices (9.3 ± 2.8 ms) (Colquhoun et al. 1992) and to follow a recovery rate from desensitization of 58 ms (Colquhoun et al. 1992). Most of the starting values were taken from an earlier model (Ambros-Ingerson and Lynch 1993) and are consistent with data involving repetitive millisecond agonist pulses applied to patches taken from cultured hippocampal slices (Arai and Lynch 1998). Initial values for additional rate constants that were not present in the reduced-state Ambros-Ingerson and Lynch model (Ambrose-Ingerson and Lynch 1993) were taken from Tour et al. (2000). Final values for the rate constants were obtained using a Nelder-Mead optimization algorithm, while satisfying the constraint of microreversibility and assuming a step-shaped glutamate release event lasting 1 ms at a concentration of 10 mM. Simulated responses were required to fit normalized amplitudes and decay time constants of empirical data with a sum of squared error of <0.5. Computer simulations were performed using software developed by Dr. Brücher in MATLAB (MathWorks, Natick, MA).

Summation of EPSCs from scattered dendritic sites, as occurs on the spatially extended pyramidal neurons in adult hippocampus, results in a whole cell response with a waveform that is necessarily distorted from the shape of the EPSC at individual synapses. Asynchronous activation, arising from collaterals of unequal length as well as nonuniform delays in release, is likely to be an important contributor to the shape of the summated current (Diamond and Jahr 1995). Accordingly, a first step in extrapolating from receptor kinetics to whole cell recordings is to obtain an estimate of the temporal dispersion of synaptic activation. This has been done in the present report using two data sets from hippocampal CA1 neurons: mEPSCs and whole cell clamp responses to activation of the Schaffer-commissural projections. A function that describes the probability over time that a synapse will become active was estimated through deconvolution from the averaged mEPSC, representing the "synaptic" response, and the whole cell clamp response. The results of the kinetic model for individual receptors combined with the convolution-based process for simulating whole cell responses provide a means for estimating how changes in AMPA receptor rate constants will affect whole cell responses.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Excitatory currents measured in this study resulted from activation of AMPA receptors (Fig. 2). Responses clamped at 70 mV and elicited by modest stimulation did not contain significant contributions from N-methyl-D-aspartate (NMDA)-type glutamate receptors, as evidenced by the absence of any effect of NMDA receptor antagonists on the amplitude or duration of the EPSC (Fig. 2A). On the other hand, the EPSC was eliminated by the AMPA receptor antagonist CNQX (Fig. 2B).

View larger version (15K): [in this window] [in a new window]   Fig. 2. Sample responses illustrating the effect of glutamate receptor antagonists. The N-methyl-D-aspartate (NMDA) antagonist, 2-amino-5-phosphonovaleric acid (AP5), did not alter control excitatory postsynaptic currents (EPSCs, A). Aa: representative control response. Ab: response after infusion of 50 µM AP5. Ac: superposition of the 2 responses. Calibration: 30 ms, 40 pA. The AMPA receptor antagonist, 6-cyano-7-nitroquinoxalene-2,3-dione (CNQX), eliminated the EPSC (B). Ba: representative control response. Bb: representative response after 10 µM CNQX infusion. Bc: superposition of the 2 responses. Calibration: 30 ms, 20 pA.

Effects on duration of evoked responses were assessed by estimating decay time constants. Figure 3A illustrates the 90-30% segment of the EPSC over which the decay time constant was calculated. Estimates were obtained from responses that satisfied correlational criteria for satisfactory curve fitting. Decay tau estimates did not co-vary with EPSC amplitude as indicated by the scatter plots in Fig. 3, B and C. There was no significant correlation between decay tau and amplitude for the experimental pathway (r = 0.21) nor for the control pathway (r = 0.08). Shown are the results for all control (Fig. 3C) and experimental (Fig. 3B) responses collected for the 10 min prior to delivering theta bursts. Averaging the responses for each pathway and then testing for relationships across cells did not improve the correlation between amplitudes and decay tau (control: r = 0.003; experimental: r = 0.15, n = 14 in each case).

View larger version (25K): [in this window] [in a new window]   Fig. 3. Characteristics of EPSCs prior to theta burst stimulation and drug infusion. A: estimation of decay time constants. Decay time constants were estimated for individual EPSCs that met curve-fitting criteria. A representative trace from the control period is shown. , the single exponential fit used to calculate tau. - - -, the selected decaying portion of the response (90-30% of peak amplitude) used in the calculation. Note the close correspondence between the recorded EPSC and the fitted single exponential. The individual control EPSC depicted had an estimated decay time constant of 5.75 ms, and the exponential fit to the data surpassed criteria (r2 > 0.95). Calibration: 20 ms, 50 pA. B: decay time constants of EPSCs plotted against corresponding peak amplitudes for the experimental pathway during the control period. , 1 time point. There is no relationship between the 2 variables apparent in the plot. C: the same variables were plotted in the same manner as in B for control pathway EPSCs during the control period. The random pattern of dots suggests that the 2 variables are independent of each other.

Interactions between LTP and CTZ

Figure 4, A and B, summarizes the over-time effects of theta bursts and CTZ on amplitudes and decay time constants across all cells (n = 14). Average mean amplitudes (Fig. 4A) for the control period (t = 10 to 0 min) were 124 ± 39 (SD) pA for control EPSCs and 110 ± 28 pA for experimental EPSCs, and amplitudes for the two inputs were correlated (r = 0.81), indicating that similar sized responses were used in individual experiments. After 10 min of baseline responses, TBS was delivered to the experimental pathway (Fig. 4, t = 0 min), resulting in an immediate increase in peak amplitude only in the potentiated pathway that was sustained for the entire 30-min drug period. Potentiation was evident from the first post-theta pulse, decayed slightly to a stable level over 5-10 min and did not change in a systematic fashion over the remainder of the recording period (Fig. 4A). The increases were large and restricted to the stimulated pathway: mean EPSC amplitude in the test pathways increased by 99 ± 76%. Examination of the summary plot for EPSC decay tau (Fig. 4B) shows that theta bursts also caused an immediate and persistent reduction in decay time constants, an effect in accord with earlier field recording studies (Ambros-Ingerson et al. 1993). For all slices (Fig. 4B), tau decreased from 8.08 ± 1.3 ms for the 10 min prior to theta burst stimulation to 6.39 ± 1.1 ms for the 30-min period following theta burst stimulation (P < 0.0001). (See following text for a description of the results for a subset of slices with particularly stable responses.)

View larger version (37K): [in this window] [in a new window]   Fig. 4. Effects of theta burst stimulation (TBS) and cyclothiazide (CTZ) on peak amplitude and decay tau. A: peak amplitudes (means ± SE) for 14 experiments are plotted across time for control pathway () and experimental pathway (). At t = 30 min, 100 µM CTZ infusion began, and its effects on peak amplitude were apparent ~10 min and maximal ~15 min after drug infusion began. CTZ was infused for a 20-min period (indicated by bar). Peak amplitudes decreased in both pathways during the washout period. Representative traces from the control pathway are shown (inset), from left to right: response during control period; response following delivery of TBS to the experimental pathway. Note that the control pathway response does not change; Response during CTZ infusion increases both peak amplitude and decay tau of control EPSC; superposition of the 3 responses. Calibration: 20 ms, 50 pA. B: decay time constants (means ± SE) for 14 experiments are plotted across time for control pathway () and experimental pathway (). There were no differences between decay tau for the 2 pathways during the baseline period. Delivery of TBS pulses to the experimental pathway (t = 0 minutes) resulted in a rapid decrease in decay tau for the experimental pathway only that was sustained for the entire 30-min period following TBS and prior to drug infusion. Infusion of 100 µM CTZ gradually increased decay tau in both pathways, although to a larger extent in the experimental pathway than the control pathway. Values of decay tau for the control pathway were larger than values of decay tau for the experimental pathway during the period following TBS and prior to drug infusion, whereas the relationship shifted as CTZ took effect (t = ~40 min) such that values of decay tau for potentiated EPSCs became significantly larger than those of control EPSCs. Sample traces for the potentiated pathway are illustrated (inset), from left to right: response during control period; response following TBS; response during CTZ infusion. The drug increased peak amplitude and strongly increased the decay time of the response; Superposition of the 3 responses. Calibration: 20 ms, 100 pA. C: peak amplitude (mean ± SE) 5 min prior to and 25 min during CTZ infusion for a subset of particularly stable cases (n = 8). CTZ effects were maximal during the 40- to 55-min time window, and absolute amplitude increases due to CTZ were greater for the potentiated pathway than the control pathway during this time period. D: decay time constants for the same time window and for the same subset of cases shown in C. Following CTZ infusion, the effect on potentiated decay taus was reversed such that potentiated decay tau values surpassed control decay tau values.

CTZ had differential effects on control versus potentiated pathways. Figure 4A shows that while amplitude increased in both pathways, the absolute effect of the drug beginning 10 min following the start of infusion was greater in potentiated responses (+9.4 ± 38 pA for control pathway vs. +78.2 ± 74 pA for test pathway, P < 0.0005). CTZ also had unequal decay time effects on control versus potentiated responses. The drug increased the decay tau for both pathways (Fig. 4B), but, as with amplitude, the change was greater for the potentiated responses (4.24 ± 3.5 ms for control pathway tau vs. 7.00 ± 3.7 ms for experimental pathway tau, P < 0.0002). In all, decay taus were significantly shorter for potentiated responses prior to drug infusion and longer afterward: the direction of the decay tau change was opposite for drug and LTP. Data were collected from 11 cells during washout and indicated that the effects of CTZ on both pathways were fully reversible.

Figure 4, A and B, insets, shows results from individual experiments. CTZ, in accord with earlier reports (Isaacson and Walmsley 1996; Rammes et al. 1996; Yamada and Tang 1993), caused a marked increase in EPSC decay tau of control responses (Fig. 4A, inset) without changing the initial slope of the EPSC; the resultant waveform distortion is evident in the superimposed traces. Theta bursts applied to the second input (Fig. 4B, inset) caused an immediate increase in the slope and amplitude of the EPSC without noticeably changing its waveform. Potentiation was not accompanied by increases in the control EPSCs (Fig. 4A, inset). The increases in peak amplitude and decay tau produced by CTZ in potentiated responses are evident in the figure.

Figure 4, C and D, provides a detailed description of the transition periods for the control and potentiated responses using the 8 (of 14) slices with the most stable potentiated and control EPSCs during the 5-min period leading up to the transition. Figure 4C highlights the differential effects of CTZ on amplitude during the 25 min after drug infusion began. The 5-min period prior to drug infusion (t = 25-30 min) is shown, and it is clear that the potentiated and control pathway amplitudes remained significantly different (P < 0.0001), indicating that LTP effects on amplitude were still in place. CTZ statistics were computed for the time period 10 min after the start of drug infusion. Predrug amplitudes in the two pathways were correlated to a similar degree as predrug decay tau (r = 0.70). As can be seen in the figure, CTZ increased the amplitude of the potentiated responses (+97 ± 94 pA) to a greater degree than it did the control EPSCs (36 ± 42 pA, P < 0.04).

Figure 4D illustrates the decay time constant changes elicited by CTZ during the same transition period shown in Fig. 4C. Drug effects on decay tau paralleled those of amplitude, although the two variables were not significantly correlated [r = 0.03 for the potentiated pathway and r = 0.53 for the control pathway (P >0.05)]. Note that decay tau in the control and potentiated pathways was significantly different (7.7 ± 0.7 ms in the control pathway and 6.9 ± 1.0 ms in the potentiated pathway, P < 0.016, paired t-test, 1 tail) in the 5 min prior to drug infusion. CTZ resulted in a significant increase in decay tau for both potentiated responses (6.9 ± 1.0 ms increased to 15.4 ± 3.5 ms) and control responses (7.7 ± 0.7 ms increased to 13.3 ± 3.4 ms). Following drug infusion, control tau and potentiated tau were significantly different from each other (P < 0.03) but in the reverse direction (potentiated > control) from the pre-CTZ values (control > potentiated). As anticipated, absolute increases in decay tau following CTZ infusion were greater for potentiated responses (8.5 ± 3.2 ms) than for control responses (5.6 ± 3.1 ms, P < 0.025).

Figure 5 summarizes two final tests for interactions between LTP and CTZ. The effect of the drug on the amplitude of the potentiated response over the entire perfusion period was correlated with percent LTP (r = 0.53, P < 0.05). As can be seen in Fig. 5A, the weakness of the relationship reflected two cases in which the drug caused large increases in slices in which potentiation was relatively weak; in the remaining 12 slices, CTZ's influence on amplitude appeared to be strongly related to percent LTP (Fig. 5A, dotted line, r = 0.87). Specifically, CTZ tended to have a greater effect on amplitude when the increase in amplitude produced by LTP was greater. A repeated measures ANOVA was performed to test for an interaction between drug effects and LTP effects on decay tau in the data summarized in Fig. 4D. Absolute changes in decay tau produced by experimental manipulations, namely induction of LTP and infusion of CTZ, on potentiated and control pathways are plotted in Fig. 5B. The change in decay tau following delivery of TBS to the experimental pathway was 0.3 ± 2 ms for the control pathway and 1.6 ± 0.3 (SD) ms for the potentiated pathway (n = 8). The absolute change in decay tau following infusion of CTZ was 5.6 ± 1.1 ms for the control pathway and 8.5 ± 1.1 ms for the experimental pathway (n = 8). The interaction term of the analysis proved to be significant (P < 0.02), indicating that the increased effect of CTZ on decay rates of potentiated responses is not equivalent to simply adding the individual decay tau effects of potentiation and the drug.

View larger version (10K): [in this window] [in a new window]   Fig. 5. Relationships between LTP and CTZ. A: correlation between degree of amplitude increase produced by LTP and degree of amplitude increase produced by CTZ on potentiated EPSCs. A correlation coefficient of 0.80 was calculated for 12 representative cases (), discarding 2 outlying points () that deviated considerably from the average effects of LTP and CTZ. When all points were used (n = 14), the correlation was decreased, yet still significant. B: Interaction effect of LTP and CTZ. Decay tau averages (mean ± SE) from 8 slices with pre-CTZ means () and post-CTZ means (). The nonparallel nature of the lines depict the presence of a significant interaction effect between the change in decay tau due to LTP vs. the change in decay tau due to CTZ.

Effects of LTP and CTZ in the presence of a GABA receptor antagonist

In a separate set of experiments, the GABA_A receptor antagonist picrotoxin and GABA_B antagonist CGP 35348 were infused at 50 and 100 µM, respectively, to block possible effects on decay time constants of di-synaptic feed-forward IPSCs. The mean amplitude of the EPSCs for slices (n = 7) in which LTP was induced was 151 ± 42 pA. This was close to, and not significantly different from, the mean pre-LTP value for the slices tested without GABA antagonists (110 ± 28 pA, n = 14). The mean decay tau was 6.7 ± 1.0 ms, a value that was not significantly different from that recorded in the first experiments. In all, the size and waveform of the EPSCs (see Fig. 6A) did not differ in any obvious way from those obtained without picrotoxin. LTP was if anything less robust in the presence of the antagonists (Fig. 6B). The mean degree of potentiation for the last 5 min of the 30-min test period was 53 ± 37% relative to the pre-theta burst baseline; the comparable value in the first experiment was 99 ± 76% (P = 0.07, Mann-Whitney U test). There were no increases in the control inputs to the same neurons.

View larger version (37K): [in this window] [in a new window]   Fig. 6. Effects of TBS and CTZ on peak amplitude and decay tau in the presence of 50 µM picrotoxin and 100 µM 3-amino-propyl (diethoxymethyl)-phosphonic acid (CGP 35348). A: representative traces from the control (top) and potentiated (bottom) pathways. Shown from left to right are baseline responses, responses following theta burst stimulation to the experimental pathway, responses following CTZ infusion (100 µM), and superposition of the 3 cases. The horizontal calibration bar is 30 ms for the traces from both pathways, but the vertical calibration bar was increased from 50 pA (control) to 100 pA (test) for illustration purposes. B: Normalized amplitudes (mean ± SE) for 7 experiments are plotted across time for control pathway (open circles) and potentiated pathway (filled circles). TBS was delivered to the experimental pathway (indicated by arrow) following 10 min of baseline recordings, resulting in an increase in percent amplitude that remained significant throughout the entire pre-CTZ period. CTZ was infused for 20 min (indicated by horizontal bar) starting at t = 40 min. Note that although amplitude did appear to increase in the control pathway, the amplitudes of the potentiated responses were significantly more affected by the drug. C: decay time constants (mean ± SE, n = 7) plotted across time as percent of baseline responses. Following delivery of TBS to the experimental pathway, the mean decay time constant tended to be lower in the potentiated pathway (filled circles). During the time period in which LTP was maximal (i.e., t = 15 to 25 min), the decrease in decay tau for the potentiated responses reached statistical significance. Infusion of CTZ produced the same result as in the previous experiments; decay time constants increased in both pathways but did so to a significantly higher degree in the potentiated pathway. D: absolute differences in decay tau (mean ± SE, n = 7) between control and experimental pathways. During the 10 min of baseline recording, the differences between decay time constants were close to 0. Following LTP induction, the difference between experimental and control decay time constants (potentiated tau minus control tau) dropped <0 for most time points, indicating that decay tau tended to be reduced for a potentiated response compared with a control response. Application of CTZ reversed the relationship between the 2 pathways, such that decay time constants of potentiated responses were significantly greater than decay time constants of control responses in the presence of the drug.

Figure 6C summarizes the percent changes in decay time constants over the course of the experiment. The potentiated pathway exhibited a small decrease that did not reach statistical significance (e.g., 0.5 ± 0.7 pA at 5-10 min after theta bursts). Within-cell comparisons of post-theta burst changes in control and potentiated synapses provide a sensitive test for LTP-related shifts in decay tau. Potentiation reduced tau relative to control inputs during the time periods in which percent LTP was maximal; i.e., for the period running from 5 to 15 min post-theta burst stimulation, the decay tau in the potentiated pathway was reduced by 7.7 ± 5.6 ms relative to the control synapses (P < 0.0015, paired t-test, 1 tail). Figure 6D describes the differences between decay taus in the two pathways over the entire course of the experiment.

CTZ had a variable and generally modest effect on peak amplitude (Fig. 6A) in control pathways (28 ± 41 pA) that did not reach statistical significance (P < 0.06). It caused a rapidly developing and more robust effect on peak amplitudes in the potentiated synapses (+94 ± 26 pA, P < 0.0001). The within-cell difference in absolute drug effect for potentiated versus control synapses was highly significant (P < 0.0001), and the responses in the two pathways were correlated (r = 0.87). In percentage terms, CTZ increased the peak amplitude of the control responses by 39 ± 53% and that of the potentiated responses by 76 ± 42% (P < 0.039, Fig. 6B). Figure 6C shows that CTZ caused a rapid and pronounced increase in the decay time constants of the responses, again in the same manner as in the first study. The decay tau increased by 6.7 ± 2.3 ms in the control pathway and by 10.2 ± 2.9 ms in the potentiated synapses, a difference that was significant (P < 0.009). In percentage terms, the effects on decay tau were larger and more reliable than those on amplitude; the control time constants were increased by 100 ± 34% and those for the potentiated synapses by 161 ± 65% (P < 0.004).

Effects of CTZ on presynaptic transmitter release

To evaluate whether the above-reported interaction between CTZ and LTP was in part due to effects of CTZ on transmitter release, responses to paired pulses were compared in the absence and presence of CTZ. Figure 7 depicts effects of CTZ on paired-pulse response amplitude and decay tau. Four paired-pulse protocols were used with 50-, 80-, 100-, and 200-ms interpulse intervals. Responses to the second pulse in the pair are facilitated, as expected, but facilitation in the control condition and drug condition did not significantly differ from one another, indicating that CTZ was not significantly facilitating transmitter release. The amplitude of the response to the second pulse increased by 211 ± 54% relative to the response to the first pulse in the control condition versus 238 ± 84% with CTZ for the 50-ms interpulse interval (ISI), 168 ± 41% in the control condition versus 153 ± 42% with CTZ for the 80-ms ISI, 143 ± 24% in the control condition versus 165 ± 39% with CTZ for the 100-ms ISI, and 124 ± 18% in the control condition versus 112 ± 18% with CTZ for the 200-ms ISI. These values were not statistically different from one another (P > 0.7 for 50-ms ISI, P > 0.8 for 80-ms ISI, P > 0.6 for 100-ms ISI, and P > 0.6 for 200-ms ISI, n = 2). There were also no significant changes in decay tau of the response to the second pulse relative to the response to the first pulse for all pulse protocols. There was a significant increase in decay tau following CTZ infusion, indicating that the drug was exerting its normal effect. The average decay tau for the first EPSC in the pair was 10.5 ms during the control period and 17.7 ms after CTZ infusion (P < 0.02) for 50-ms ISI, 11.2 ms during the control period and 16.9 ms after CTZ infusion (P < 0.01) for 80-ms ISI, 10.3 ms during the control period and 16.3 ms after CTZ infusion (P < 0.004) for 100-ms ISI, and 10.6 ms during the control period and 16.8 ms after CTZ infusion (P < 0.02) for 200-ms ISI.

View larger version (21K): [in this window] [in a new window]   Fig. 7. Effects of CTZ on amplitude and decay tau of facilitated responses. Each symbol represents mean ± SE. The horizontal bar indicates the period of CTZ infusion (t = 25-45 min). Percent increase in amplitude of the response to the second pulse of paired stimulation is shown for interpulse intervals of 50 ms (A), 80 ms (B), 100 ms (C), and 200 ms (D). No significant differences were found for control vs. drug facilitated response amplitudes. Decay time constants are shown for responses to the 1st pulse (filled circles) and 2nd pulse (open circles) for interpulse intervals of 50 ms (E), 80 ms (F), 100 ms (G), and 200 ms (H). There were no significant changes in decay tau of the response to the 2nd pulse relative to the 1st pulse either with or without the drug. However, note that CTZ did significantly increase decay tau of responses, indicating that the drug was effective.

Modeling of responses

Figure 8A illustrates a simulated response obtained using the above-described model. The "synaptic response" (dashed line) is the probability over time that a receptor will be in the open state; the whole cell response (solid line) is the convolution of the synaptic response with the above-described probability over time function for activating a synapse. The rise time and decay tau of the composite response closely resemble recorded EPSCs; this is as expected given that the kinetic model output approximates single synaptic events and the convolution function is derived from mEPSCs and whole cell EPSCs. The decay tau (3.3 ms) of the modeled response is comparable to that found in excised patches from mature cultured hippocampal slices after a millisecond agonist pulse (Arai et al. 1996) and is in reasonable agreement with the average mEPSC presented here, which had a decay rate of 2.7 ms. Maximum open probability of the simulated control response reached ~0.58, in close agreement with the value reported for CA1 excised patches (0.57) (Spruston et al. 1995).

View larger version (6K): [in this window] [in a new window]   Fig. 8. Simulated responses. Dashed traces indicate probability that the AMPA receptor will be in the open state (`synaptic' response), and solid traces depict composite responses resulting from the convolution process described in "Methods" and are intended to represent the whole cell response. Responses were normalized by peak amplitude of control to more easily observe waveform alterations. Dotted horizontal and vertical lines indicate zero axes. Calibration: 10 ms, 100%. A: Simulated responses under baseline conditions. These responses were generated using the constants listed in Fig. 1B. Note that the whole cell simulated response is not merely proportional to the synaptic simulated response. B: Effects of LTP on simulated responses. Modeling the effects of LTP as an increase in opening rates produced large increases on slope and amplitude of simulated synaptic and whole cell responses. C: Effects of CTZ on simulated responses. Effects of CTZ were modeled as decreases in desensitization rate parameters. Note that CTZ increased the normalized amplitude of the `synaptic' response to a lesser degree than the `whole cell' response, corresponding to experimental findings. D: Interaction of LTP and CTZ. Changes in rate parameters used to model the effects of LTP were combined with those changes used to model the effects of CTZ. The CTZ-induced changes in amplitude and decay time of the composite EPSC more than doubled when applied against a background of the LTP-related changes in channel opening rates.

The effects of LTP on the size and waveform of the EPSC were simulated in an earlier model by simply increasing the opening and closing rates of the AMPA receptor (Ambros-Ingerson and Lynch 1993). The present model, using the parameters described in the preceding text, reproduced the qualitative and quantitative effects of LTP on amplitude and waveform of the composite EPSC by accelerating the channel opening time from both sensitized and desensitized states. This was accomplished by multiplying the channel opening rate constants (k₃ and k₈, Fig. 1) by 5. Figure 8B shows a simulation in which these changes were implemented, producing large and equivalent increases in the slope and amplitude of the composite response.

The effects of CTZ were simulated as a decrease in desensitization rates k₄, k₅, and k₆ to zero, serving to eliminate transition into the desensitized states (see Fig. 1A). CTZ holds steady-state AMPA receptor currents in excised patches to near peak values rather than simply slowing the rate at which they decay toward baseline; this effect is most easily explained by the assumption that the drug is decreasing the probability that the receptor will be in the desensitized state (Arai et al. 1996). The consequences of these shifts in desensitization rate constants are shown in Fig. 8C. CTZ had essentially no effect on the initial slope of the current of the synaptic response (dashed line), increased peak amplitude by ~30%, but markedly extended the decay time of the response. These effects are consistent with those seen in excised patch studies using long pulses of glutamate where the drug has a small effect on peak amplitude while greatly increasing steady state current (Arai et al. 1996; Yamada and Tang 1993). The predicted effects of CTZ, as implemented in the kinetic model, on whole cell responses are also described in Fig. 8C (solid line). As is evident, the drug had a much more pronounced effect on the amplitude of the composite response (88%) than expected from its effects on the amplitude of single EPSCs. This disparity in drug effect across levels of analysis corresponds closely to experimental results; i.e., CTZ has a much larger effect on amplitude in whole cell responses than in excised patches.

Combining the LTP related kinetic changes with the shifts in desensitization used to reproduce CTZ's effects produced the results shown in Fig. 8D. The interaction produced by the combined effects resulted in an increase in decay tau of the composite response of 130% above control, compared with a 50% above control increase produced by CTZ effects alone. Likewise, amplitude was increased by 270% above control values when effects of CTZ and LTP were combined and 80% above control values for CTZ alone.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Induction of LTP does not disturb transient forms of synaptic facilitation elicited by repetitive afferent stimulation (Grover 1998; Muller et al. 1989) or other manipulations (Muller et al. 1988b) that enhance transmitter release. Related experiments showed that LTP could be produced without substantially changing the currents passed by postsynaptic NMDA receptors (Kauer et al. 1988; Muller and Lynch 1988; Muller et al. 1988a). These findings are not compatible with the idea that increased release is responsible for LTP. Experimental work also failed to confirm predictions from the hypothesis that LTP arises from shifts in the resistance of dendritic spines (Jung et al. 1991; Larson and Lynch 1991). By exclusion, then, the changes responsible for expression appear to be located between release and spines and therefore are likely to involve the AMPA-type glutamate receptors that mediate fast EPSCs. As predicted by this hypothesis, changing the composition of the receptors by reducing the concentration of their GluR-1 subunits has a profound effect on the expression of LTP (Vanderklish et al. 1992; Zamanillo et al. 1999).

Also in accord with the receptor hypothesis are reports that LTP changes the waveform of synaptic responses (Ambros-Ingerson et al. 1991, 1993; Kolta et al. 1998; Stricker et al. 1996; Xie et al. 1997) as well as the effects of a drug that modulates AMPA receptor kinetics (Kolta et al. 1998; Staubli et al. 1992; Xiao et al. 1991). These LTP-induced effects did not detectably change over a 75-min test period and thus appear to have the stability required of an expression mechanism (Kolta et al. 1998). The present studies show that LTP recorded under membrane-clamp conditions changes the effects of a structurally and functionally different type of AMPA receptor modulator. Aniracetam is reported to affect both receptor deactivation and receptor desensitization (Isaacson and Nicoll 1991; Tang et al. 1991; Vyklicky et al. 1991), while CTZ more specifically and potently targets the desensitization state (Arai et al. 1996; Partin et al. 1996; Patneau et al. 1993; Yamada and Tang 1993). The LTP-drug interaction found in the present study was much larger than that obtained in earlier work with aniracetam (Staubli et al. 1992), yet both interaction results lend support to the hypothesis that LTP changes AMPA receptor dynamics. The interaction between LTP and CTZ was again obtained in a second set of experiments using GABA antagonists to block hyperpolarizing currents. In addition to providing a replication, this result rules out the possibility that the shunting effects of fast feed-forward inhibition could have distorted the manner in which LTP modifies CTZ's effects on decay tau. The interaction results discussed in the preceding text provide a constraint on hypotheses regarding the nature of biophysical changes that accompany LTP. That is, hypotheses should include an explanation of how LTP can alter the effects of drugs that act on AMPA receptor kinetics.

Several studies have reported presynaptic effects of CTZ; namely, an increase in transmitter release probability (Barnes-Davies and Forsythe 1995; Bellingham and Walmsley 1999; Diamond and Jahr 1995; Isaacson and Walmsley 1996; Ishikawa and Takahashi 2001). In the present study, paired-pulse facilitation was used to test for a presynaptic effect of CTZ because effects of this type could cloud interpretation of results with respect to the site of LTP expression. Previous studies reporting significant presynaptic effects of CTZ involved various experimental preparations that were different from that used in the present study, including rat auditory brain stem slices (Barnes-Davies and Forsythe 1995; Bellingham and Walmsley 1999; Isaacson and Walmsley 1996; Ishikawa and Takahashi 2001) and immature cultures of dissociated CA1 neurons (Diamond and Jahr 1995). The results presented here suggest that presynaptic effects of CTZ on field CA1 of acute hippocampal slices from juvenile rats are negligible.

The present studies also found that induction of LTP is accompanied by a small decrease in the decay time constant of the EPSC measured before infusion of CTZ. The size of this effect was not greatly different from that (3-10%) recorded in earlier work using pharmacologically isolated AMPA receptor-mediated field EPSPs (Kolta et al. 1998). An LTP-related reduction in decay tau was also detected in the picrotoxin experiment using within-cell comparisons of control and potentiated responses. That similar qualitative and quantitative effects have been obtained with whole cell recording and field potentials strengthens the conclusion that potentiation causes a measurable alteration to the waveform of synaptic responses.

As discussed, an interaction between drug and potentiation is expected from the broad hypothesis that the latter effect involves a shift in the rate constants that describe transitions between receptor states. The particular nature of the interaction provides constraints on more specific arguments about which constants are affected by LTP. However, attempts to account for the characteristic features of LTP with kinetic changes face the problem of extrapolating from receptor behavior to the whole cell EPSCs used to measure potentiation. The present work addressed the problem by estimating the convolution process that occurs between the synaptic current, as estimated from averaged mEPSCs, and the whole cell EPSCs recorded in hippocampal slices. The empirically derived convolution process was then used to construct composite EPSCs from the output of the AMPA receptor model. The base model by itself reasonably well reproduced the waveforms obtained from mEPSCs, and the convolution process generated whole cell responses that approximated those obtained in slice experiments. A less obvious result was obtained in the case of CTZ. The drug has little effect on peak currents in patches (Arai et al. 1996) but causes sizeable amplitude increases in whole cell recordings as demonstrated in the present study. Modeling CTZ as a reduction in desensitization reproduced this discrepancy along with changes in the decay time constant of the EPSC that were comparable to those seen in slice experiments.

We postulate that the effect of LTP is to increase the channel opening rate. Thus using the same base model as in the preceding text, the kinetic constants leading to the open state (k₃ and k₈, Fig. 1) were increased equivalently in an attempt to generate a receptor that would, after the same convolution process used for control and CTZ simulations, yield an EPSC corresponding to that found in potentiated synapses. The acceleration in channel opening resulted in a response that had equivalent increases in slope and amplitude. Faster opening generates a steeper initial slope and a higher probability of the channel open state (and thus increased amplitude). This gives more importance to desensitization from the open state and ultimately produces the differential effects reported.

Combining the rate constants used to implement LTP with those for CTZ produced, after convolution, a whole cell EPSC with the characteristics found in the experimental work. The effects of CTZ on amplitude and decay tau of simulated responses were significantly greater for potentiated responses to about the degree found in experimental studies. Thus the hypothesized kinetic changes of LTP produced the interaction effects seen in the physiological data and suggest a novel explanation for LTP. In summary, the hypothesis put forth in this report is supported by both empirical data and modeling results.

Several studies have shown that phosphorylation modifies the biophysical properties of AMPA receptors (Banke et al. 2000; Greengard et al. 1991; Lee et al. 2000; Roche et al. 1996; Traynelis and Wahl 1997) and that two sites on the cytoplasmic tail of the GluR-1 subunit are of particular importance in this regard (Roche et al. 1996). As noted, the presence of GluR-1 within the AMPA receptor may be essential to LTP expression (Vanderklish et al. 1992). Certain of the reported phosphorylation effects could potentially result in EPSC changes of the types seen with LTP (e.g., Banke et al. 2000), and evidence of GluR-1 C-terminal phosphorylation after LTP has been described (Barria et al. 1997; Lee et al. 2000). Interestingly, the same rate change that was predicted by the kinetic model to occur following induction of LTP, an acceleration of channel opening rate, has been reported to occur in response to phosphorylation of Glu-R6 homomeric receptors (Traynelis and Wahl 1997). Studies are currently in progress to determine if phosphorylation changes the effects of CTZ on AMPA receptor kinetics in a manner consonant with that found after LTP.