INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The strength of a synapse can vary with frequency of activation. This variation, known as short-term plasticity, can be an increase or decrease and can last just tens of milliseconds or as long as several minutes. Synapses in the behaving animal operate within this time frame (Alexander and Crutcher 1990; Chevalier and Deniau 1990; Crutcher and DeLong 1984), so short-term synaptic plasticity is likely an important mechanism in the functioning of biological neural networks.

Short-term plasticity can be due to a number of mechanisms. One is the intrinsic properties of neurotransmitter release. Several types of short-term increases in synaptic strength such as facilitation, augmentation, and posttetanic potentiation are due to the accumulation of residual calcium in the presynaptic terminal as a result of one or more preceding action potentials (Delaney and Tank 1994; Fisher et al. 1997; Kamiya and Zucker 1994; Swandulla et al. 1991). The effects of previous calcium influxes combine nonlinearly with the effects of later calcium influxes, increasing the probability of neurotransmitter release. Sometimes, however, repetitive action potentials deplete the supply of neurotransmitter, thereby negating the effects of residual calcium and causing a short-term decrease in synaptic strength (Dittman and Regehr 1998; Swandulla et al. 1991).

Another mechanism responsible for short-term changes in synaptic strength is activation of neurotransmitter receptors on the presynaptic terminal. For example, activation of GABA_B receptors (Calabresi et al. 1991; Radnikow et al. 1997; Seabrook et al. 1991), muscarinic acetylcholine receptors (Marchi et al. 1990; Sugita et al. 1991), metabotropic glutamate receptors (Stefani et al. 1994), or adenosine receptors (Mori et al. 1996) can decrease synaptic strength at inhibitory synapses in striatum by decreasing the amount of GABA released. Activation of these receptors decreases release either by decreasing the calcium influx into the presynaptic terminal or by interfering with mechanisms of transmitter release (Wu and Saggau 1997). On the other hand, activation of nicotinic acetylcholine receptors can increase the amount of neurotransmitter released (Gray et al. 1996; Lena and Changeux 1997). Finally, synaptic strength can be decreased by the postsynaptic mechanism of receptor desensitization (Jones and Westbrook 1995; Otis et al. 1996).

GABAergic inhibition serves many important functions in the nervous system; it can prevent action potential generation (Bazemore et al. 1957), limit N-methyl-D-aspartate (NMDA) receptor activation (Dingledine et al. 1986; Kanter et al. 1996; Luhmann and Prince 1990; Staley and Mody 1992), hamper the backpropagation of action potentials into the dendritic tree (Kim et al. 1995; Larkum et al. 1999; Tsubokawa and Ross 1996), and synchronize membrane potential oscillations (Cobb et al. 1995). Short-term modulation of the strength of GABAergic synapses will affect all these processes and is thus important for neuronal integration. GABAergic inhibition is particularly important in the striatum as over 95% of striatal neurons are GABAergic (Kemp and Powell 1971). We have therefore studied two forms of short-term plasticity, paired-pulse depression (PPD) and synaptic augmentation, at GABAergic synapses onto medium spiny neurons in the striatum. We have investigated their time course, parametric requirements, pharmacology, and impact on the ability of GABAergic synapses to prevent action potential generation in a situation approximating the in vivo behavior of these cells.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Slice preparation

Adult Fisher 344 rats (2-3 mo) were anesthetized with pentobarbital sodium (60 mg/kg ip) and decapitated. Brains were quickly removed and submerged in cold (3-5°C), aerated (95% O₂-5% CO₂) sucrose artificial cerebral spinal fluid (ACSF). Sucrose ACSF consisted of (in mM) 248 sucrose, 5 KCl, 1.3 MgSO₄, 28 NaHCO₃, 1.25 NaH₂PO₄, and 10 glucose. Hemi-coronal slices (400 µm) were cut on a Vibratome in sucrose ACSF, initially transferred to a mixture of half sucrose ACSF and half normal ACSF, then to normal ACSF, and allowed to warm to room temperature and recover for at least 1 h. Normal ACSF consisted of (in mM) 124 NaCl, 3 KCl, 2.4 CaCl₂, 1.3 MgS0₄, 26 NaHCO₃, 1.25 NaH₂PO₄, and 10 glucose. Slices were transferred individually to a Haas-type interface chamber and perfused with aerated (95% O₂-5% CO₂) normal ACSF at 28-30°C.

In some experiments, we altered the normal ACSF. For Ba²⁺ ACSF, we substituted 0.5-1 mM BaCl₂ for the corresponding concentration of CaCl₂ and substituted MgCl₂ for MgSO₄ to prevent the precipitation of BaSO₄. For Cs⁺ ACSF, we substituted CsCl for KCl.

Intracellular recordings

Intracellular recording electrodes (90-185 M) were pulled with a Flaming/Brown P-87 puller (Sutter Instruments). Medium spiny neurons in the dorsal medial region of the striatum were impaled, and their responses to current injection and synaptic stimulation were recorded. Synaptic responses were evoked using a bipolar electrode (twisted, formvar-coated nichrome wire, 65 µm OD, A-M Systems) placed within the striatum about 100 µm from the recording electrode. Responses were filtered (3 kHz), amplified (Axoclamp 2A, Axon Instruments), digitized (10-20 kHz; Labmaster TL-1 or Digidata 1200B), and stored on computer (pClamp, Axon Instruments). All data were collected with ionotropic glutamate receptors blocked by 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX, 20 µM) and D,L-2-amino-5-phosphonovaleric acid (APV, 50 µM), except for the data in Fig. 1, A and B. 

View larger version (28K): [in this window] [in a new window]   Fig. 1. Pharmacological isolation and voltage dependence of the inhibitory postsynaptic potential (IPSP). A: after a baseline period in the absence of drugs, 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX, 20 µM), D,L-2-amino-5-phosphonovaleric acid (APV, 50 µM), and atropine (1 µM) greatly reduced the synaptic response to intrastriatal stimulation. Increasing stimulation intensity from 90 to 500 µA () increased the amplitude of the nonglutamatergic response. Bicuculline (30 µM) partially blocked the nonglutamatergic response. Traces are averages of 5 consecutive responses. Sample IPSPs (b and c) are depolarizing because the resting membrane potential (90 mV) was hyperpolarized relative to the IPSP reversal potential (65 mV). Electrode was filled with KC2H3O2. Stimulus artifacts are clipped. B, left: application of CNQX and APV reduced the amplitude of the response to intrastriatal stimulation by 85 ± 2% (n = 9). Right: application of bicuculline (in addition to CNQX, APV, and atropine) reduced the response by 61 ± 3% (n = 8). C: the IPSP reversed polarity at 76 ± 1 mV and rectified at hyperpolarized membrane potentials (n = 29). These data were collected using electrodes filled with KCH3SO4. Traces are single responses. In this example, the reversal potential was 72 mV. Electrode was filled with KCH3SO4.

Recording electrodes were filled with methyl potassium sulfate (KCH₃SO₄, 2 M) unless otherwise specified. In some early experiments electrodes were filled with potassium acetate (KC₂H₃O₂, 2 M), and in these cells, the inhibitory postsynaptic potential (IPSP) reversed polarity at more depolarized membrane potentials (64 ± 2 mV; n = 8) than in cells recorded with electrodes containing KCH₃SO₄ (76 ± 1 mV; n = 29; P < 0.001; 2-tailed t-test). This is consistent with the finding that the GABA_A channel is permeable to C₂H₃O (Bormann et al. 1987). All electrodes also contained biocytin (1%) for intracellular staining. Hyperpolarizing current was used to inject cells with biocytin, and biocytin-filled neurons were reacted for horseradish peroxidase histochemistry according to previously published procedures (Dunia et al. 1996). Briefly, slices were fixed after electrophysiological recording for at least 24 h in cold (4°C) 4% paraformaldehyde in phosphate-buffered saline, pH 7.3. The slices were then sectioned (60-75 µm) on a sliding microtome, collected in phosphate-buffered saline, and rinsed several times. Sections were initially incubated in a 1% solution of Triton X-100 for 1 h to permeabilize the cell membranes. Injected cells were labeled by incubating the tissue with an avidin-horseradish peroxidase complex (ABC solution) at a dilution of 1:1000 in phosphate-buffered saline (Vector Labs, Burlingame, CA). After several rinses in phosphate-buffered saline over 1 h, the sections were reacted with diaminobenzidine (Sigma; 0.06%) and H₂O₂ (0.003%) in 0.15 M tris(hydroxymethyl)aminomethane buffer for 10 min. The sections were rinsed and mounted on gelatin-coated slides, which were then air dried, defatted, and coverslipped in Permount. Of the 64 neurons included in this study, 44 were successfully labeled with biocytin and directly identified as medium spiny neurons. The electrophysiological properties of unidentified neurons were similar to those of identified medium spiny neurons, and thus we assume the unidentified neurons were also medium spiny neurons. In some experiments, we bathed the slice in Ba²⁺ or Cs²⁺ to block K⁺ channels. Under these conditions, cell type cannot be determined electrophysiologically because blockade of K⁺ channels alters electrophysiological properties. Of 10 cells recorded in the presence of Ba²⁺ or Cs²⁺, 9 were identified as medium spiny neurons through biocytin staining.

Stimulation and analysis

Stimulus current duration was 100 µs. During drug application, pairs of stimuli with 50-ms interstimulus intervals (ISIs) were applied every 20 s. In tests for paired-pulse modulation, pairs with ISIs from 50 ms to 10 s were applied. The interval between the second impulse of a given pair and the first impulse of the next pair was always 20 s. Each ISI was tested five times consecutively per cell, and pairs were presented in order of increasing ISIs. Paired-pulse modulation was quantified by dividing the maximum amplitude of the second response by the maximum amplitude of the first response, and the result was expressed as a percentage. In tests for augmentation, we applied a conditioning train of 15 stimuli at 20 Hz. This is the same type of stimulation we used to measure augmentation at excitatory synapses onto medium spiny neurons (Ou et al. 1997) and thus allows comparison of the magnitude and time course of augmentation at excitatory and inhibitory synapses. Also, this type of stimulation is physiologically relevant, as it is within the operating range of these synapses in awake animals (Wilson 1993; Wilson and Groves 1981). Test stimuli were applied 2, 5, 10, 20, and 30 s after the end of the conditioning train. This was repeated five times per cell with 2.5 min between conditioning trains. Augmentation was quantified by dividing the maximum amplitude of the response to a test stimulus by the maximum amplitude of the response to a control stimulus applied 15 s before the beginning of the train, and the result was expressed as a percentage. In experiments on the effects of different train lengths, conditioning trains of 1, 3, 5, 7, 9, 11, 13, and 15 stimuli were applied in either ascending or descending order. Recovery time between trains was 1.5 min, and the sequence of trains of different lengths was repeated three times per cell.

Membrane potential was corrected for any offset observed following electrode withdrawal but was not corrected for liquid junction potentials. Input resistance was determined with a 100-pA injection from a membrane potential of 90 mV. Results are expressed as mean ± SE. In the rectification experiments, overall statistical significance was determined with an omnibus F-test ANOVA, and post hoc Scheffe tests were used to identify the significant comparisons underlying the overall significance. Other tests for significance are identified when the results are stated.

The resting membrane potential of medium spiny neurons (87 ± 1 mV, n = 45) was negative with respect to the IPSP reversal potential (Fig. 1C), and thus IPSPs elicited at rest were depolarizing. We did not quantify PPD and augmentation at membrane potentials depolarized to the IPSP reversal potential because we did not want to introduce problems associated with current injection into our analysis. The resistance of high-impedance electrodes changes with current injection, and imposing large depolarizations activates voltage-dependent currents. These problems create recording instability that could limit quantitative comparisons. In addition, there is much evidence that PPD and augmentation are due to presynaptic mechanisms, and thus it is unlikely that sampling cells at their resting potential would affect the plasticity. We did perform a small qualitative sampling of the effects of PPD and augmentation on IPSPs evoked at depolarized membrane potentials when we examined spike probability in Figs. 6 and 9.

Drugs

The following compounds were obtained from Sigma (St. Louis, MO): APV, atropine (sulfate salt), biocytin, cesium methanesulfonate, potassium acetate, dimethyl sulfoxide (DMSO), SCH 23390, sulpiride, and all components of the ACSF. Methyl potassium sulfate was obtained from Pfaltz and Bauer (Waterbury, CT). 6-Cyano-7-nitroquinoxaline-2,3-dione (CNQX) and ()-bicuculline methiodide were obtained from Research Biochemicals (Natick, MA). CNQX was initially dissolved in DMSO, kept refrigerated as a stock solution for no more than 1 mo, and added to ACSF the day of the experiment. The final concentration of DMSO was 0.02%. CGP 35348 was a gift from Novartis Pharma AG.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Properties of the IPSP

Intrastriatal stimulation in control ACSF resulted in a depolarizing postsynaptic potential (Fig. 1A) which could trigger an action potential at higher stimulation intensities. Application of the ionotropic glutamate receptor antagonists CNQX (20 µM) and APV (50 µM) reduced the amplitude of the response by 85 ± 2% (n = 9; Fig. 1B) and revealed an IPSP with a reversal potential of 76 ± 1 mV (n = 29; Fig. 1C). The amplitude of the IPSP varied with stimulation intensity and was sensitive to the GABA_A receptor antagonists bicuculline (30 µM) and picrotoxin (50-100 µM). Sometimes (8 of 32 neurons) the IPSP was multiphasic as previously reported (Seabrook et al. 1991) but only at depolarized membrane potentials.

The amplitude of the IPSP rectified at membrane potentials more hyperpolarized than 85 mV (Fig. 1C). This has also been observed in unitary IPSPs between striatal interneurons and medium spiny neurons (Koos and Tepper 1999). There are at least two possible mechanisms for this. The first is shunting by K⁺ channels. The input resistance of medium spiny neurons decreases at hyperpolarized membrane potentials due to the activation of an inwardly rectifying K⁺ current (Nisenbaum and Wilson 1995). This current has been shown to shunt excitatory responses to intrastriatal stimulation (Calabresi et al. 1990), so it may be responsible for shunting inhibitory responses as well. A second possible mechanism is the intrinsic properties of the GABA_A receptor channel. Rectification of GABAergic responses has been observed in other preparations (Barker and Harrison 1988; Bormann et al. 1987; Collingridge et al. 1984; Weiss et al. 1988) and may be due to the difference between [Cl]_i and [Cl]_o (Barker and Harrison 1988) and to a voltage dependence of the kinetics of the GABA_A receptor channel (Bormann et al. 1987; Dudel et al. 1980; Segal and Barker 1984; Weiss 1988).

To test if the rectification of the IPSP was due to shunting, we blocked K⁺ currents by using either Ba²⁺ ACSF or Cs⁺ ACSF (see METHODS). In four of five experiments with Cs⁺ ACSF, we also applied Cs⁺ internally by filling the intracellular electrode with CsCH₃SO₃ (2 M). The average input resistance of cells bathed in Cs⁺ ACSF (59 ± 3 M; n = 5; P < 0.05) or Ba²⁺ ACSF (97.4 ± 13 M; n = 5; P < 0.001) was higher than the input resistance of cells bathed in normal ACSF (39 ± 2 M; n = 29), indicating that these manipulations did in fact block K⁺ channels (Fig. 2A). On the other hand, the mean IPSP reversal potential of cells bathed in Cs⁺ ACSF (79 ± 4 mV) or Ba²⁺ ACSF (76 ± 1 mV) was similar to that of cells bathed in normal ACSF (76 ± 1 mV), indicating that these manipulations did not affect the transmembrane Cl gradient (Fig. 2C).

View larger version (48K): [in this window] [in a new window]   Fig. 2. Rectification of the IPSP is reduced by Ba2+ or Cs+. A: effect of K+ channel blockade on responses to current injection. Left: normal artificial cerebrospinal fluid (ACSF). Membrane potential was set to 89 mV by constant injection of 40 pA. Lowermost trace is the response to injection of 500 pA for 50 ms. In successive traces, the current injection is incremented by 100 pA. Middle: Ba2+ ACSF. Membrane potential was set to 87 mV by constant injection of 660 pA. Lowermost trace is the response to injection of 250 pA for 300 ms. In successive traces, the current injection is incremented by 100 pA. Right: Cs+ ACSF and CsCH3SO4 (2 M) in electrode. Membrane potential is set to 87 mV by constant injection of 510 pA. Lowermost trace is the response to injection of 250 pA for 300 ms. In successive traces, the current injection is incremented by 100 pA. Longer current injections were used in Ba2+ ACSF and Cs+ ACSF due to the longer membrane time constant and wider action potential. All traces are single responses. B: example of the effect of Cs+ on rectification. Rectification is present shortly after impaling the cell in Cs+ ACSF with an electrode filled with CsCH3SO3 () but not 20-min later (). C: voltage dependence of the IPSP in normal ACSF (; n = 29), Ba2+ ACSF (; n = 5), and Cs+ ACSF (; n = 5). Rectification was reduced in Ba2+ ACSF or Cs+ ACSF. Note that the reversal potential was approximately the same in all 3 conditions. Control data are the same as in Fig. 1C. D: rectification was quantified for each cell by fitting the data points at hyperpolarized membrane potentials (B, left shading) and at depolarized membrane potentials (B, right shading) with separate regression lines and then dividing the slope of the regression line in the hyperpolarized region by the slope of the regression line in the depolarized region. The result is called the rectification ratio. The average rectification ratio was greater for cells bathed in Ba2+ ACSF or Cs+ ACSF than for cells bathed in normal ACSF. E: in the hyperpolarized region, the average slope of the regression line was greater for cells bathed in Ba2+ ACSF or Cs+ ACSF than for cells bathed in normal ACSF. F: in the depolarized region, the average slope of the regression line for cells bathed in Ba2+ ACSF or Cs+ ACSF was not significantly different from for cells bathed in normal ACSF. *P < 0.001.

Bathing cells in Ba²⁺ ACSF or Cs⁺ ACSF reduced the rectification of the IPSP (Fig. 2, B and C). To quantify this effect, we calculated a rectification ratio for the IPSP from each cell. The rectification ratio was determined by performing a linear regression on the reversal data for the IPSP in a hyperpolarized region of membrane potentials (V_m  90 mV) and for a depolarized region of membrane potentials (V_m  72 mV). The rectification ratio is the slope of the regression line from the hyperpolarized region divided by the slope of the regression line from the depolarized region. If there is no rectification, the slopes in the two ranges are equal, and the rectification ratio is unity. If there is complete rectification, the slope in the hyperpolarized range is zero, and the rectification ratio is zero. The average IPSP rectification ratio for cells bathed in Ba²⁺ ACSF (0.45 ± 0.07; n = 5; P < 0.001) or Cs⁺ ACSF (0.82 ± 0.13; n = 5; P < 0.001) was greater than the average IPSP rectification ratio for cells bathed in normal ACSF (0.09 ± 0.02; n = 29; Fig. 2D). This was due to steeper slopes in the hyperpolarized region (Fig. 2E) and not to a difference between slopes in the depolarized region (Fig. 2F). Thus rectification of the IPSP was reduced in the presence of Ba²⁺ or Cs⁺, indicating that it is due at least in part to shunting by K⁺ channels.

PPD of the IPSP

The first form of short-term plasticity we investigated was paired-pulse modulation. Stimulation using pairs of impulses with a wide range of ISIs revealed that the IPSP exhibits PPD (Fig. 3, A and B). PPD was negligible at the shortest ISI tested (50 ms; 94 ± 4%; n = 10) but was present at ISIs from 100 ms to 2 s and was greatest when the ISI was 500 ms (80 ± 2%). (Six of these experiments were conducted with KC₂H₃O₂ in the recording electrode and 4 with KCH₃SO₄. PPD was similar for the 2 groups, so they were combined.) The delayed onset and long duration of PPD suggest that it may be due to the activation of metabotropic receptors on the presynaptic terminals of inhibitory synapses. Several types of metabotropic receptors have been shown to decrease GABA release in striatum and thus may mediate PPD. One type is muscarinic acetylcholine receptors (Marchi et al. 1990; Sugita et al. 1991). We tested whether these receptors are involved in PPD by stimulating with paired pulses in the presence of the muscarinic antagonist atropine (1 µM). Atropine did not block PPD (Fig. 3C; 83 ± 3% at 500 ms ISI; n = 6; P > 0.35, 2-tailed t-test), indicating that PPD is not mediated by muscarinic acetylcholine receptors. Activation of GABA_B receptors has also been shown to reduce IPSPs in striatum (Calabresi et al. 1991; Radnikow et al. 1997; Seabrook et al. 1991). We tested whether GABA_B receptors mediate PPD by measuring PPD in the presence of the GABA_B receptor antagonist CGP 35348 (100 µM). We found that PPD in the presence of CGP 35348 (Fig. 3C, 79 ± 3% at 500 ms ISI; n = 5) is not significantly different from PPD in normal ACSF (Fig. 3C; P > 0.75, 2-tailed t-test). Finally, we tested if activation of D1 or D2 dopamine receptors is responsible for PPD as dopamine receptor agonists have been observed to modulate GABA release in striatum (Harsing and Zigmond 1997; Wang and Johnson 1995) and GABA_A receptor currents in medium spiny neurons (Flores-Hernandez et al. 2000). A combination of the D1 antagonist SCH 23390 (5 µM) and the D2 antagonist sulpiride (20 µM) had no significant effect on PPD (82 ± 5% at 500 ms ISI; n = 3; P > 0.65, 2-tailed t-test).

View larger version (20K): [in this window] [in a new window]   Fig. 3. The IPSP exhibits paired-pulse depression (PPD) over a wide range of interstimulus intervals (ISIs). A: average results of experiments with paired-pulse stimulation. PPD was greatest at ISIs of several hundred milliseconds and persisted for several seconds. For all ISIs n = 10 except for 5 s (n = 6) and 10 s (n = 3). For 6 of these cells, the recording electrode was filled with KC2H3O2; for the other 4 cells, the recording electrode was filled with KCH3SO4. The error bars are smaller than the symbols at some data points. B: responses from a representative paired-pulse experiment. Each trace is the average of 5 responses at the indicated ISI. Recording electrode was filled with KC2H3O2. Stimulation artifacts are clipped. C: PPD is not blocked by atropine (1 µM), CGP 35348 (100 µM), or a combination of SCH 23390 (5 µM) and sulpiride (20 µM). Data shown are for 500 ms ISI. For control, n = 10; for atropine, n = 6; for CGP 35348, n = 5; for SCH 23390 and sulpiride, n = 3.

Augmentation of the IPSP

We next investigated another form of short-term plasticity known as augmentation. Originally described at the frog neuromuscular junction (Magleby and Zengel 1976), augmentation results from the accumulation of calcium in the presynaptic terminal during repetitive activity (Delaney and Tank 1994; Kamiya and Zucker 1994; Swandulla et al. 1991; Tank et al. 1995). It decays with a time constant of several seconds and is often accompanied by facilitation and posttetanic potentiation, two other types of short-term plasticity (Fisher et al. 1997). We have previously shown that augmentation occurs at excitatory synapses onto medium spiny neurons (Ou et al. 1997) but is rather brief, decaying completely after 6 s. To determine if augmentation occurs at inhibitory synapses onto medium spiny neurons as well, we tested the effect of a brief conditioning train (15 stimuli at 20 Hz; Fig. 4, A and B) on the amplitude of the IPSP. We found that 2 s after the end of the train the amplitude of the IPSP was increased to 119 ± 1% of control (Fig. 4, C and D; n = 6; P < 0.005, paired, 2-tailed t-test). The augmentation decayed with a time constant of 10 s. We did not give test pulses at delays shorter than 2 s and thus do not know if the conditioning train also produced facilitation.

View larger version (25K): [in this window] [in a new window]   Fig. 4. The IPSP is augmented by a brief conditioning train. A: augmentation was induced with a conditioning train of 15 stimuli at 20 Hz. Test responses were elicited after delays of 2, 5, 10, 20, and 30 s from the end of the conditioning train. Augmentation was quantified by dividing the amplitude of a test response by the amplitude of a control response elicited 15 s before the beginning of the conditioning train. B: sample response to the conditioning train. Average of 5 trials. Stimulus artifacts are clipped. Resting membrane potential was 93 mV. C: the conditioning train results in augmentation of the IPSP. Barium does not affect the augmentation 2 s after the conditioning train, but 30 s after the conditioning train, the response is depressed. For normal ACSF, n = 9 for all delays except 2 s, for which n = 6. For Ba2+ ACSF, n = 5 for all delays. Line is a single exponential fit to control data: y = 27.4 exp(t/9.9). D: sample responses from an augmentation experiment. Leftmost trace is the control response elicited 15 s before the beginning of the conditioning train. Other traces are test responses elicited after the indicated delays from the end of the conditioning train. Each trace is the average of 5 trials. Stimulus artifacts are clipped.

Studies at the frog neuromuscular junction (Zengel and Magleby 1980), rabbit superior cervical ganglion (Zengel et al. 1980), and chick ciliary ganglion (Poage and Zengel 1993) have demonstrated that Ba²⁺ increases the magnitude of augmentation. We tested whether this was true for IPSPs in medium spiny neurons by stimulating with brief conditioning trains when the slice was bathed in Ba²⁺ ACSF. The presence of Ba²⁺ did not affect augmentation at a delay of 2 s (116 ± 7%; n = 5; P > 0.5; 2-tailed t-test), but augmentation was decreased at longer delays of 5, 10, and 15 s. At the longest delay tested, 30 s, augmentation in control ACSF had completely decayed (100 ± 1%), but in Ba²⁺ ACSF the response was depressed (89 ± 4%; P < 0.05, 2-tailed t-test).

Requirements for generating PPD versus augmentation

With the preceding experiments we have established that inhibitory synapses onto medium spiny neurons exhibit two forms of short-term plasticity. One of these (PPD) weakens inhibition, while the other (augmentation) strengthens it. The requirements for generating the two forms of short-term plasticity differ only in the number of conditioning stimuli; PPD is generated by a single conditioning stimulus, while augmentation is generated by a train of 15 stimuli. To further investigate the relationship between PPD and augmentation, we varied the length of the conditioning train. We used conditioning trains of 1, 3, 5, 7, 9, 11, 13, and 15 stimuli and measured plasticity with a test pulse 2 s after the end of the train (Fig. 5A). We found, as expected, that shorter trains depressed the IPSP and longer trains augmented the IPSP (Fig. 5, B and C). There was a roughly linear transition between the two forms of short-term plasticity, with conditioning trains of 9 and 11 impulses producing no plasticity at all on average.

View larger version (34K): [in this window] [in a new window]   Fig. 5. Increasing the number of conditioning stimuli reveals a gradual transition from PPD to augmentation. A: conditioning trains of 1, 3, 5, 7, 9, 11, 13, and 15 stimuli were applied. Short-term plasticity was assessed with a test pulse 2 s after the end of the train and quantified by dividing the amplitude of the test response by the amplitude of a control response elicited 15 s before the beginning of the train. B: shorter trains produced depression, and longer trains produced augmentation (n = 9). No plasticity was observed in response to trains of 9 and 11 stimuli. C: sample responses illustrating the effects of conditioning trains of different lengths. Conditioning trains of 1 and 5 stimuli depressed the response, while trains of 13 and 15 stimuli augmented the response. The control trace is the average of the control responses preceding all 24 conditioning trains (3 trials each of 8 different train lengths). Other traces are the averages of the test responses from 3 trials of the indicated train length. D: conditioning trains presented in descending order of length (n = 4) produced less depression and more augmentation than conditioning trains presented in ascending order of length (n = 5). Same data as in B.

In these experiments, the longest conditioning trains were the same length as those in our initial augmentation experiments, yet they resulted in less augmentation (105 ± 3 vs. 119 ± 1% with test pulse after 2-s delay). This may be due to methodological differences between the two sets of experiments. In the initial augmentation experiments, we applied a total of five conditioning trains, all of them 15 stimuli in length, and allowed 2.5 min between trains for recovery. In the experiments in which we varied train length, we applied a total of 24 trains (3 trials each of 8 different lengths) and allowed 1.5 min between trains for recovery. To test if such methodological differences might affect the magnitude of augmentation, we separated the data in Fig. 5B by order of train presentation. Trains presented in descending order of length (n = 4) tended to produce less depression and more augmentation than those presented in ascending order (n = 5). Although this difference was not significant (P > 0.15; split plot factorial ANOVA), it indicates that the magnitude of augmentation may be sensitive to methodological parameters. This is further emphasized by results from pilot experiments in which we presented a single test stimulus 2 s after a 15-stimulus conditioning train and allowed 30 s between trains for recovery. This method resulted in less augmentation as well (111 ± 5%; n = 5; not shown).

Effects of PPD and augmentation on action potential generation

One of the purposes of inhibition is to prevent action potential generation. To test if PPD and augmentation change the ability of inhibition to block spikes, we elicited trains of action potentials by constant current injection and compared spike inhibition between control and depressed IPSPs and between control and augmented IPSPs. In four of four cells, a depressed IPSP was less effective in stopping the generation of action potentials, while in three of four cells an augmented IPSP was more effective (Fig. 6). This demonstrates that short-term plasticity of the IPSP can change the output of medium spiny neurons.

View larger version (35K): [in this window] [in a new window]   Fig. 6. PPD and augmentation modify the ability of the IPSP to block action potentials during constant current injection. The neuron generated a train of action potentials in response to constant current injection (1 nA). Top traces: synaptic stimulation elicited an IPSP which temporarily stopped the generation of action potentials. Middle traces: the IPSP was the 2nd in a pair of IPSPs (650 ms ISI; 1st IPSP not shown) and stopped the generation of action potentials for a shorter period due to PPD. Bottom traces: a conditioning train (15 stimuli at 20 Hz; not shown) preceded the onset of current injection by 5 s. In this case, the IPSP stopped the generation of action potentials for a longer period due to augmentation. Duration of current injection and time of synaptic stimulation are indicated below the bottom group of traces. Stimulation intensity and current injection were the same for all 3 conditions. Each group contains 5 superimposed traces. The variable amplitude of the action potentials in the bottom group of traces was due to a slower sampling rate. Note the jitter in the timing of action potentials.

There are two drawbacks to eliciting action potentials with current injections of constant amplitude: the pattern of action potentials varies from trial to trial (Fig. 6) and constant current injection is not physiologically realistic. To avoid these drawbacks, we injected currents of varying amplitude. Such current injections were specified by random sequences of 2,440 or 4,960 values. Each value specified the level of current injection for 100 µs. To generate a sequence, values uniformly distributed between 1 and 1 were convolved with a rectangular window 40-80 values wide. During experiments, sequences were shifted in the positive direction and scaled until the neuron generated the desired number of action potentials. The amount of scaling and shifting was then fixed for the duration of the experiment.

We observed that a given varying current injection produced nearly the same pattern of action potentials in every trial (Fig. 7), replicating the results of a previous study in neocortical pyramidal cells (Mainen and Sejnowski 1995) and demonstrating that medium spiny neurons in vitro are capable of producing a reliable output. Also the response to a varying current better approximates the in vivo behavior of a medium spiny neuron than the response to a constant current (Wilson 1993; Wilson and Groves 1981).

View larger version (28K): [in this window] [in a new window]   Fig. 7. A current of varying amplitude produces a nearly identical pattern of action potentials in every trial. A: response to injection of a varying current. Ten traces superimposed. B: the varying current injected in A. C: raster plot of spike times from the 10 traces in A.

Figure 8 demonstrates that an IPSP can change the response to a varying current. The top trace in Fig. 8 is the response to a varying current. In this case, we injected less current so that the cell would not spike. The middle trace in Fig. 8 is the response to the same varying current with an IPSP elicited 75 ms after current onset. The IPSP decreased the depolarization produced by the varying current, as shown by the difference between the two responses (Fig. 8, bottom trace).

View larger version (21K): [in this window] [in a new window]   Fig. 8. An IPSP can decrease the depolarization produced by injection of a varying current. Top: response to varying current injection alone. The amplitude of the current was low to prevent action potential generation. Middle: an IPSP was elicited 75 ms after onset of the same varying current. , stimulus artifact. Bottom: subtracting the top trace from the middle trace reveals the IPSP. Each trace is the average of 10 trials.

Finally, we tested whether short-term plasticity affects the ability of an IPSP to block action potentials generated by varying current injection, and we found that it does (n = 4 for PPD, n = 2 for augmentation). This is illustrated in Fig. 9. In the top traces (Fig. 9, A1 and B1), the neuron generated six action potentials in response to varying current injection. In the middle traces (Fig. 9, A2 and B2), we elicited an IPSP during the varying current injection, and the IPSP blocked two action potentials. In the bottom left trace (Fig. 9A3), the IPSP elicited during current injection was the second in a pair of IPSPs separated by 350 ms. As a result of PPD, the IPSP blocked the generation of one action potential instead of two. In the bottom right trace (Fig. 9B3), we elicited the IPSP 2 s after a conditioning train. As a result of augmentation, this IPSP blocked the generation of three action potentials instead of two. This demonstrates that PPD and augmentation can modify the output of medium spiny neurons by modifying the ability of inhibitory inputs to block action potentials.

View larger version (21K): [in this window] [in a new window]   Fig. 9. PPD and augmentation modify the ability of the IPSP to block action potentials during varying current injection. A1: injection of a varying current produced 6 action potentials in 10 of 10 trials. A2: eliciting an IPSP during varying current injection blocked 2 action potentials (8 of 9 trials; in the other trial, only 1 action potential was blocked). A3: the IPSP during current injection was the second in a pair of IPSPs (350 ms ISI; first IPSP not shown) and blocked 1 action potential instead of 2 due to PPD (10 of 10 trials). B1: injection of a varying current produced 6 action potentials in 10 of 10 trials. B2: eliciting an IPSP blocked 2 action potentials in 10 of 10 trials. B3: a conditioning train (15 stimuli at 20 Hz; not shown) preceded the onset of current injection by 2 s. The IPSP blocked 3 action potentials instead of 2 due to augmentation (9 of 10 trials; in the other trial, the IPSP blocked 2 action potentials). , action potentials of interest. , stimulus artifacts. Each trace is a single trial. Stimulation intensity in A was 25 µA. Stimulation intensity in B was 20 µA. Current injection was identical for all conditions. Trials of the 3 stimulation conditions in A were alternated. Same for B.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We have investigated the properties and short-term plasticity of an IPSP in medium spiny neurons of the striatum. Our main findings are that the IPSP rectifies at hyperpolarized membrane potentials due in part to shunting by K⁺ channels, the IPSP exhibits augmentation and long-lasting PPD, and augmentation and PPD can modify the output of striatal neurons.

Source of the IPSP

The most likely source of the GABAergic synapses studied here are the medium spiny neurons, as they constitute the vast majority of striatal neurons (Kemp and Powell 1971). The postsynaptic neurons in this study are also medium spiny neurons, and while there is anatomical evidence that medium spiny neurons synapse on each other (Somogyi et al. 1981; Wilson and Groves 1980), attempts to demonstrate this physiologically have not been successful (Jaeger et al. 1994). Two other possible sources are GABAergic striatal interneurons (Kawaguchi et al. 1995; Koos and Tepper 1999) and GABAergic afferents from substantia nigra pars reticulata (Rodriguez and Gonzalez-Hernandez 1999). Given the type of stimulating electrode used and the graded relationship between stimulation intensity and response amplitude, the IPSP described here probably results from the activation of many synapses, and properties of this IPSP are probably average properties of synapses from several different sources.

Rectification of the IPSP

The IPSP rectified at hyperpolarized membrane potentials. This rectification was reduced when the cell was exposed to Ba²⁺ or Cs⁺, demonstrating that K⁺ channels were partially responsible for the rectification. The fraction of rectification insensitive to Ba²⁺ and Cs⁺ may have been due to the voltage dependence of the GABA_A conductance (Bormann et al. 1987; Dudel et al. 1980; Segal and Barker 1984; Weiss 1988) or to the difference between [Cl]_o and [Cl^-]_i (Barker and Harrison 1988).

The finding that K⁺ channels shunt the IPSP is not surprising as K⁺ channels have previously been shown to shunt EPSPs in medium spiny neurons (Calabresi et al. 1990). The activity of medium spiny neurons is determined in part by the K⁺ channels that shunt the IPSP. In vivo intracellular recordings in both anesthetized and awake rats have shown that medium spiny neurons have two membrane potential states a quiet, hyperpolarized state (the "down state") and a noisy, depolarized state (the "up state"). A transition from the down state to the up state can be triggered by a barrage of excitatory input (Wilson 1993; Wilson and Groves 1981). The K⁺ channels are believed to maintain the neuron in the down state by shunting excitatory inputs (Nisenbaum and Wilson 1995).

When the neuron is in the down state, the purpose of inhibition may be to prevent transitions to the up state. Inhibitory inputs produce a transient shunt, which supplements the persistent shunt by K⁺ channels. In the presence of this larger shunt, some excitatory inputs capable of overcoming the shunt by K⁺ channels may not trigger a transition if the neuron receives an inhibitory input at around the same time. Furthermore the shunt provided by inhibitory inputs has a different voltage dependence than the shunt provided by K⁺ channels. The K⁺ channels inactivate with depolarization (Nisenbaum and Wilson 1995), while the conductance of GABA_A channels increases with depolarization (Bormann et al. 1987; Dudel et al. 1980; Segal and Barker 1984; Weiss 1988). This suggests that as the neuron becomes depolarized an increase in shunting by inhibitory inputs could compensate for a decrease in shunting by K⁺ channels. In these ways, inhibitory inputs may act to prevent medium spiny neurons from reaching the up state.

When in the up state, the neuron often generates action potentials (Wilson 1993; Wilson and Groves 1981), and the purpose of inhibition in the up state may be to prevent action potentials. The up state is more depolarized than the IPSP reversal potential, so IPSPs are hyperpolarizing in the up state. Also, in the up state, the amplitude of the IPSP is not restrained by shunting. Therefore an inhibitory input that occurs during the up state will produce a significant hyperpolarization capable of blocking action potential generation.

PPD

The IPSP exhibited PPD over a wide range of ISIs, as previously reported (Radnikow et al. 1997). The time course of PPD, with its slow onset and even slower recovery, suggests it is due to activation of presynaptic metabotropic receptors. Several types of presynaptic metabotropic receptors have been shown to decrease GABA release in striatum. These include GABA_B receptors (Calabresi et al. 1991; Radnikow et al. 1997; Seabrook et al. 1991), muscarinic acetylcholine receptors (Marchi et al. 1990; Sugita et al. 1991), adenosine A_2A receptors (Mori et al. 1996), and metabotropic glutamate receptors (Stefani et al. 1994). We found that neither CGP 35348 nor atropine blocked PPD, indicating that PPD is not mediated by GABA_B receptors (Radnikow et al. 1997) or muscarinic receptors. Activation of dopamine receptors has been reported to have both presynaptic (Harsing and Zigmond 1997; Wang and Johnson 1995) and postsynaptic (Flores-Hernandez et al. 2000) effects on GABAergic synapses in striatum, but D1 and D2 dopamine receptors do not seem to be involved in PPD, as combined application of SCH 23390 and sulpiride had no effect on PPD. This is in accord with the finding that dopamine application does not affect the IPSP (Nicola and Malenka 1998).

If PPD is not due to the activation of presynaptic metabotropic receptors, there are at least two other mechanisms that could be responsible. One is the depletion of neurotransmitter (Takeuchi 1958). Release of neurotransmitter in response to the first stimulus may reduce the number of synaptic vesicles available for release in response to the second stimulus, thereby decreasing the amplitude of the second IPSP. Another possible mechanism is receptor desensitization. Some GABA_A receptors may become desensitized during the first IPSP, rendering them unable to contribute to the second IPSP (Jones and Westbrook 1995; Mellor and Randall 1998; Tia et al. 1996). The time course of PPD does not support either of these mechanisms, however. PPD produced by depletion or desensitization would be greatest at the shortest ISIs and would monotonically recover to baseline, but we observed no PPD at the shortest ISI and increasing PPD until the ISI reached 200 ms. If depletion and desensitization do contribute to the PPD we observed, their effects must be obscured at shorter ISIs by some simultaneous facilitative process.

Augmentation

Repetitive synaptic activity has been shown to temporarily enhance synaptic strength at many synapses across a variety of species (Fisher et al. 1997). After a strong conditioning train, the enhancement in synaptic strength typically decays with fourth-order kinetics. The four time constants of decay define four types of enhancement: Fast-decaying facilitation (F1) has a decay time constant in the range of tens of milliseconds, slow-decaying facilitation (F2) in the range of hundreds of milliseconds, augmentation in the range of seconds, and posttetanic potentiation (PTP) in the range of tens of seconds to minutes. The enhancement of the IPSP that we observed decayed with a time constant of 10 ± 1 s and is therefore an example of augmentation.

The augmentation of the IPSP was unusual in two respects: it was rather small, and it was not accompanied by PTP. Both effects may be due to the relatively weak nature of the conditioning train. Studies that report greater augmentation and co-occurrence of PTP typically employ a conditioning train with 10-50 times as many stimuli, often at a higher frequency (e.g., Bittner and Baxter 1991; Magleby and Zengel 1976; Tanabe and Kijima 1992; Zengel et al. 1980). Moreover, studies employing conditioning trains of varying lengths have shown that longer trains produce greater augmentation and PTP (Magleby and Zengel 1976; Poage and Zengel 1993; Zengel and Magleby 1982).

There is strong evidence that augmentation is due to an increase in presynaptic [Ca²⁺] after repetitive activity (Delaney and Tank 1994; Fischer et al. 1997; Kamiya and Zucker 1994; Swandulla et al. 1991). Studies showing that Ba²⁺ increases augmentation (Poage and Zengel 1993; Zengel and Magleby 1980; Zengel et al. 1980) also suggest that augmentation is dependent on Ca²⁺ as presynaptic Ca²⁺ channels are permeable to Ba²⁺. We tested the effect of Ba²⁺ on augmentation of the IPSP and observed no change at shorter delays but a switch to depression at the longest delay. This suggests that the mechanism responsible for augmentation at inhibitory synapses onto medium spiny neurons may be different from the mechanism at other synapses.

We investigated the relationship between PPD and augmentation by varying the length of the conditioning train. We found that trains of 1 to 7 stimuli depressed the IPSP, trains of 9 or 11 stimuli did not affect the IPSP, and trains of 13 or 15 stimuli augmented the IPSP. This suggests that conditioning trains simultaneously activate both depressing mechanisms and augmenting mechanisms. With shorter trains the depressing mechanisms prevail, with longer trains the augmenting mechanisms prevail, and with trains of intermediate lengths the two mechanisms cancel.

Effects of PPD and augmentation on action potential generation

We have shown that IPSPs can block action potentials generated in response to both constant and varying current injections, thus demonstrating that inhibitory inputs can shape the firing patterns of medium spiny neurons. We have also shown that the effectiveness of the IPSP in blocking action potentials is diminished by PPD and increased by augmentation, indicating that short-term synaptic plasticity at inhibitory synapses can modulate striatal output.

The type of stimulation we used probably activated many inhibitory synapses, and such synchronous inhibitory input may not be physiologically realistic. This raises the possibility that because we have activated the inhibitory inputs in an artificial manner, our results are not physiologically relevant. However, evidence from several other brain areas, including the striatum, indicates that the input of an individual inhibitory neuron is sufficient to change the behavior of its targets. For instance, single inhibitory neurons can alter action potential timing in striatal medium spiny neurons (Koos and Tepper 1999), hippocampal pyramidal cells (Cobb et al. 1995), and cerebellar Purkinje cells (Hausser and Clark 1997) and can alter backpropagating action potentials in neocortical pyramidal cells (Larkum et al. 1999). Furthermore striatal inhibitory interneurons may fire synchronously, as some are electrically coupled (Koos and Tepper 1999). Thus we believe that our findings represent inhibitory mechanisms operating in the behaving animal to modify the output of medium spiny neurons.

The changes in striatal output caused by short-term plasticity of inhibition may propagate through the circuitry of the basal ganglia to thalamus and cortex. This could influence the performance of tasks known to involve the striatum, such as movement (Hauber 1998) and memory (Knowlton et al. 1996). This is likely to be true not just for PPD and augmentation but for any other forms of plasticity, short- and long-term, that exist at these synapses.