INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The basal ganglia are a set of interconnected structures critically involved in operations as diverse as motor activation (Graybiel 1995; Marsden 1982) and reward-related learning (Reynolds et al. 2001). Disorders of these structures underlie major neurological and psychiatric conditions such as Parkinson's disease and attention-deficit hyperactivity disorder (Hynd et al. 1993). The striatum is the principal input structure of the basal ganglia and is made up of a diverse population of different types of neurons, the great majority of which are GABAergic spiny projection neurons (Bennett and Bolam 1993; Luk and Sadikot 2001; Oorschot 1996; Oorschot et al. 1999; West et al. 1996). The spiny projection neurons are the principal output neurons of the striatum and also the sites at which cortical inputs terminate (Somogyi et al. 1981). Thus they play a crucial role in the input-output operations of the striatum. These neurons not only project to other basal ganglia nuclei but also give rise to an extensive local plexus of collateral branches (Grofova 1975; Preston et al. 1980; Wilson and Groves 1980). Extensive overlap occurs between the axon collaterals and the dendritic trees of adjacent spiny projection neurons, suggesting that synaptic connectivity within the extent of local axonal spread is probable. Ultrastructural evidence has also consistently supported the existence of synaptic connections between spiny neurons (Kitai and Wilson 1982; Somogyi et al. 1981; Wilson and Groves 1980). However, up to now there has been no direct physiological evidence for functional inhibitory interactions between spiny projection neurons.

Early extracellular studies suggested that spiny projection neurons activated by stimulation of their axons may inhibit neighboring neurons (Katayama et al. 1981). However, in these studies there was uncertainty over the identity of the cells from which records were obtained. Another study using axonal stimulation suggested the presence of weak interactions that were effective in reducing synaptic input from presumed distal dendrites (Park et al. 1980). However, these effects may have been mediated by a small population of feedforward interneurons activated by collaterals of corticofugal axons (Koos and Tepper 1999). More recent dual recording studies (Jaeger et al. 1994; Stern et al. 1998) found no direct evidence for inhibitory synaptic interactions among spiny projection neurons. The lack of direct physiological evidence for functional interactions between spiny neurons remains puzzling in the light of the anatomical structure of the striatum.

We made simultaneous intracellular recordings from pairs of spiny projection neurons to test whether inhibitory interactions occur among spiny projection neurons in a striatal slice preparation. The chance of detecting an interaction was optimized by preparing slices using procedures previously shown to maximize the likelihood of detecting synaptic interactions (Thomson et al. 1996) and by making spike-triggered averages of postsynaptic responses to enhance the signal-to-noise ratio. Our experiments have revealed for the first time the existence of GABA_A receptor-mediated synaptic interactions between striatal spiny projection neurons. Preliminary results have been presented in abstract form (Tunstall et al. 2001).

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

In vitro slice preparation

Experiments were conducted in acutely prepared brain slices of 24- to 28-day-old (65-120 g) male Wistar rats. The exception was one acutely prepared brain slice of a 45-day-old (196 g) male Wistar rat. Each rat was anesthetized by an intraperitoneal injection of 120 mg/kg pentobarbital sodium (Nembutal, Virbac, New Zealand). They were then perfused transcardially for 2 min with ice-cold artificial cerebrospinal fluid (ACSF) of composition (in mM) 124 NaCl, 2.5 KCl, 2 MgSO₄, 2.5 CaCl₂, 1.25 NaH₂PO₄, 26 NaHCO₃, and 11 glucose. This procedure is reported to increase the probability of obtaining functional connectivity (Thomson et al. 1996). The brain was quickly removed and chilled further in ACSF for 3 min. The hemispheres were divided, and one hemisphere placed medial surface down on a chilled surface. A block was prepared by making a parahorizontal cut at 45° to the horizontal plane, midway between the anterior and posterior poles. The cut surface of the block was glued to a Teflon stage and submerged in chilled, oxygenated ACSF. Slices (400 µm thick) were cut from the cortical surface to the ventral striatum using a Campden vibroslice. Prepared slices were maintained submerged in a continuous flow of ACSF at 35 ± 0.5°C under humidified gas (95% O₂-5% CO₂). The slices were held down by a U-shaped platinum wire supporting an array of nylon threads in a customized chamber mounted on the fixed stage of a microscope (Olympus BX50WI).

Intracellular recording from pairs of projection neurons

Simultaneous intracellular recordings were made from pairs of striatal spiny projection neurons. Conventional sharp microelectrodes (90-120 M) were drawn from glass capillaries and filled with 2 M potassium acetate. Microelectrodes were connected to the headstages of an Axoclamp 2B intracellular recording amplifier. The V2 channel was connected to a 10× band-pass amplifier (Haer). Data was low-pass filtered at between 3 (Axoclamp 2B) and 5 kHz (Haer amplifier). The voltage and current outputs from each channel of the microelectrode amplifier were digitized at a sampling rate of 10 kHz per channel and recorded to disk using a Digidata 1200 in conjunction with pClamp 6.0 software (Axon Instruments).

Each microelectrode was advanced into the slice using a separate microdrive (Burleigh Inchworm, LSO). The microdrives for each microelectrode were positioned so that the tips could be brought into close juxtaposition within the brain slice. The first electrode was advanced until a spiny projection neuron (identified by its electrophysiological properties, see following text) was stably impaled. The second electrode was then advanced toward the predetermined position of close juxtaposition until a second neuron was impaled. The final distance between the electrode tips during the electrophysiological experiments was estimated geometrically by measuring the distance each microelectrode had advanced into the slice, the initial points of penetration on the surface of the slice, and the angle of each axis of penetration.

Measurements of cellular properties (resting membrane potential, input resistance and action potential characteristics) were made from all neurons in the sample. Input resistance was determined from the slope of a regression line fitted to the membrane potentials produced by a series of subthreshold depolarizing current pulses. Threshold for action potential firing was defined as the point on the voltage trajectory at which the rate of depolarization exceeded 8 mV/ms. Action potential amplitude was defined as the difference between threshold and the peak of the action potential waveform.

Detection of inhibitory interactions

Following stable impalement of a pair of adjacent spiny projection neurons, interactions were tested for by injecting continuous current into one neuron (the "postsynaptic" neuron) to hold it at a membrane potential positive to the chloride reversal potential (i.e., between 49 and 51 mV) while evoking one or two action potentials in the other neuron (the "presynaptic" neuron) by means of a single suprathreshold depolarizing square current pulse of 200-ms duration. The acquisition interval was 15 s. To optimize the chances of detecting interactions, the spike-triggered average of 200 responses recorded from the postsynaptic neuron was obtained off-line using a software program written in C⁺⁺ (Borland) by Dr Tunstall. This averaging enabled responses of the order of 100 µV peak amplitude to be detected. In cases where more than one spike occurred in the presynaptic neuron, only the postsynaptic response to the first spike was used.

Estimates of the peak amplitude of the IPSP and the percentage of failures of single action potentials to evoke a unitary postsynaptic response were obtained by comparing the distributions of synaptic amplitudes for individual responses, and the distribution of noise. The procedure involved two steps. In step 1, the equation for a single Gaussian distribution was fitted to the amplitude distribution of the noise. The parameters of this distribution (the mean and standard deviation of the noise) were recorded. In step 2, the weighted sum of two Gaussian distributions (Eq. 1) was fitted to the distribution of the peak amplitudes of the postsynaptic potentials (v)

(1)

In step 2 of the curve-fitting procedure, the values for the mean and standard deviation of the first component (µ₁ and ₁ in Eq. 1) were fixed to the values obtained in step 1. The resulting values of the remaining free parameters obtained after fitting Eq. 1 provided estimates of the mean amplitude of the postsynaptic potential (µ₂), its standard deviation (₂) and the overall failure rate (). The best fits in both cases were found using the Levenberg-Marquardt method in Clampfit (Axon Instruments). The latency and the duration of the response were determined from the best-fit line "by eye" through the spike-triggered average. The duration of the response was measured at half the peak amplitude.

Characterization of inhibitory interactions

The reversal potential of responses was obtained by holding postsynaptic neurons at different membrane potentials and measuring the peak amplitude of the spike-triggered average of the 200 responses evoked at each holding potential. The range of holding potentials chosen was sufficient to cause a reversal in the polarity of the averaged response. Peak amplitudes were plotted as a function of holding potential, and the reversal potential (the holding potential at which the response peak amplitude was 0) determined by interpolation from the equation of the best-fit line through the data points. Curve fitting was performed using the standard simplex algorithm implemented in Sigma Plot (Jandel Scientific). Pharmacological properties of responses were determined using the GABA_A receptor antagonist bicuculline methiodide (100 µM, Sigma-RBI). The effects of this agent were measured by comparing the spike-triggered average of 200 responses evoked under control, treatment, and recovery conditions.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Characteristics of recorded cells

Dual intracellular recordings were obtained from 45 pairs of neurons in slices from 45 animals. All the neurons in the sample (n = 90) exhibited the characteristic electrophysiological properties of spiny projection neurons (Fig. 1) as previously described (Kawaguchi et al. 1989; Kita et al. 1984, 1985; Nisenbaum et al. 1994; Tepper et al. 1993; Wickens and Wilson 1998). These included a resting membrane potential of 86.2 ± 7.6 (mean ± SD) mV, input resistance of 71.1 ± 5.2 M, evidence of inward rectification in the hyperpolarizing direction, an action potential threshold at 48.1 ± 1.6 mV above rest, and an action potential amplitude of 75.4 ± 4.6 mV.

View larger version (32K): [in this window] [in a new window]   Fig. 1. Cellular properties of recorded neurons. Both presynaptic (top) and postsynaptic (bottom) neurons exhibited voltage-time (left) and current-voltage (right) properties characteristic of spiny projection neurons. In particular note the depolarizing ramp to spike-threshold, overshooting spikes and pronounced inward rectification. Traces show individual responses to a series of current pulses (superimposed) for each neuron of a simultaneously recorded pair.

Incidence of interactions between striatal projection neurons

Unitary postsynaptic responses were detected in averages of 200 traces that were triggered by a single presynaptic action potential. Examples of these responses from different pairs are shown in Figs. 2, A-F, 3B, 4A, and 5A. Unidirectional interactions were detected in 9 of the 45 pairs of neurons recorded. On the basis of this data, the probability of a connection existing between a neighboring pair of neurons is 9/90 or approximately 1 in 10 because in all cases, interactions were tested in both directions. Bidirectional or mutual interactions were not detected in any of the pairs tested. By chance alone, if the probability of a connection in one direction is 0.1, then mutual interactions would be expected in approximately 1 in 100 pairs. Thus failure to find bidirectional interactions does not rule out mutual connectivity, although it does suggest that there is not a strong bias in favor of it.

View larger version (24K): [in this window] [in a new window]   Fig. 2. Detection of inhibitory postsynaptic potentials (IPSPs) between different pairs of spiny projection neurons. A-F: averages of 200 traces that were triggered by a single presynaptic action potential. Each trace is from a different pair. The lower noise level in A was obtained after relocating the laboratory to another building. In this and subsequent figures, the voltages stated are the membrane potentials at which the postsynaptic neurons were held. The vertical arrows indicate the timing of the spike peak in the presynaptic neuron that was used to trigger the acquisition of the spike-triggered averages.

View larger version (35K): [in this window] [in a new window]   Fig. 3. Unitary IPSPs evoked by a single presynaptic action potential varied in amplitude. A: 2 successive traces from the same cell showing a unitary IPSP (top) and a failure (bottom). B: averaged IPSP showing, superimposed, the average of 200 successive traces, and the average when failures were excluded (116 traces). C: the percentage of failures of single action potentials to evoke a unitary postsynaptic response was estimated by comparing the distributions of peak amplitudes for individual responses and the distribution of noise recorded 5 ms prior to the presynaptic action potential (n = 200 observations in both cases). Left histogram: the single Gaussian curve fitted to the amplitude distribution of the noise. Right histogram: the peak amplitudes of the individual responses, to which the weighted sum of 2 Gaussian distributions has been fitted. For the example shown, the average IPSP amplitude was 293 µV with an estimated failure rate of 42%.

View larger version (23K): [in this window] [in a new window]   Fig. 4. Reversal potential of IPSPs. A: effect of holding potential on amplitude of IPSPs. Changing the holding potential from 51 to 56 mV gave rise to a reduction in the peak amplitude of the IPSPs, and at 70 mV the IPSPs had reversed polarity. Each trace is an average of 200 successive responses. B: a plot of the peak amplitude of each of the responses in A vs. the holding potential shows that the reversal potential of the IPSPs in this case was 62.8 mV.

View larger version (21K): [in this window] [in a new window]   Fig. 5. IPSPs are mediated by GABAA receptors. A: IPSPs recorded under control conditions (top), 5 min after washing on bicuculline (middle), and 15 min after washing off bicuculline (bottom). It can be seen that bicuculline reversibly blocked the IPSPs. Each trace is an average of 200 successive responses. B: current-voltage response of the postsynaptic neuron during the control and bicuculline conditions illustrated in A. Note that bicuculline was without effect on the membrane resistance of the postsynaptic neuron. The current-voltage response of the postsynaptic neuron during the wash-off condition was similar to that of the control and has been omitted for clarity.

The cellular properties of the neurons constituting pairs in which interactions were detected were not significantly different from those of noninteracting pairs in the sample (Table 1). There were no significant differences in cellular properties between pre- and postsynaptic neurons of interacting pairs or between these groups and neurons from noninteracting pairs. This indicates that the detection of interactions was not dependent on distinctive cellular characteristics of the pre- or postsynaptic neurons. Interactions were also detected across the full age range of the animals in the sample (24-45 days).

Although the distance between the neurons could not be systematically varied, there was variation in the final distance between the tips of the recording electrodes. This occurred because the second cell of the pair was usually obtained at a different depth from the first. Successful impalements were made at depths ranging from 100 to 250 µm below the slice surface. The estimated distance between interacting pairs ranged from 153 to 445 µm (mean: 264 ± 101 µm, n = 9). The distance between noninteracting pairs ranged from 212 to 509 µm (mean: 335 ± 90 µm, n = 36). The difference between these groups was statistically significant (P < 0.05, unpaired Student's t-test). The method used to estimate the distance between the tips provided an upper bound on the inter-tip distance and did not allow for bending of the tapered shafts of the electrodes, which would have brought the tips closer. However, such an error would not produce a systematic bias. Thus it appears that closer units are more likely to be functionally connected.

Properties of interactions between projection neurons

In the raw data, the peak amplitude of the unitary IPSPs fluctuated randomly from trial to trial (Fig. 3A). However, noise levels in single traces precluded quantal analysis. Thus the data provided no indication of the number of synaptic sites involved in each functional interconnection. However, the percentage of failures of single action potentials to evoke a unitary postsynaptic response could be estimated by comparing the distributions of synaptic amplitudes for individual responses and the distribution of noise recorded immediately prior to the presynaptic action potential (Fig. 3C). A curve-fitting procedure (described in METHODS) was used to find the parameters of Eq. 1 (the weighted sum of 2 Gaussians) that gave the best fit to the distribution of the amplitudes of the peak postsynaptic potentials. The parameters µ₁ and ₁ were set to the mean and SD of the noise distribution and held fixed (see METHODS). Using this method the correlation coefficient computed for fitted data points ranged from 0.45 to 0.54, based on the difference between the observed data values and the corresponding expected value for each data point as computed by the fitted function.

The resulting parameters of the fitted equation provided reproducible estimates of the average IPSP amplitude (µ₂) while the weighting factor () gave a measure of the failure rate. For the example pair shown in Fig. 3, the IPSP amplitude was 293 µV and the failure rate was 42%. Across the population of interacting pairs, the failure rate ranged from 9 to 63% (mean: 38 ± 14%, n = 9). The peak amplitude of the unitary IPSPs evoked by a single presynaptic action potential ranged from 157 to 319 µV (mean: 277 ± 46 µV, n = 9).

The duration of the unitary IPSPs at half-amplitude ranged from 9.3 to 17.7 ms (mean: 12.7 ± 2.3 ms, n = 9) and the latency ranged from 3.2 to 8.6 ms (mean: 5.4 ± 1.9 ms, n = 9).

Figure 4 shows the reversal of the IPSP observed at depolarized membrane potentials. The group average value for the reversal potential of the IPSP was 62.4 ± 0.7 mV (n = 5). This is similar to the reversal potential of GABA receptor-mediated IPSPs reported by others (Fitzpatrick et al. 2001; Koos and Tepper 1999; Tepper et al. 1993). Consistent with this, the GABA_A receptor antagonist bicuculline reversibly blocked responses in all interacting pairs tested (n = 4). Blockade of the IPSPs with bicuculline did not change the input resistance of the postsynaptic neuron at the membrane potential from which the responses were elicited (Fig. 5).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The experiments of the present study provide direct electrophysiological evidence for the existence of local interactions between the spiny projection neurons of the rat striatum. The neurons in the sample were identified by their electrophysiological properties, which were characteristic of spiny projection neurons. The reversal of the postsynaptic potential and its blockade by bicuculline show that the interactions between the spiny projection neurons are GABA_A receptor-mediated IPSPs. To the best of our knowledge, this is the first time that such inhibitory interactions have been described between identified spiny projection neurons in rat striatal slices.

Probability of a connection

The present study provides the first experimental measure of the probability of inhibitory interactions between neighboring pairs of spiny projection neurons (i.e., 1 in every 10 connections tested). A similar probability (also 1 in every 10) of finding an interacting pair in dual recordings from layer 5 pyramidal neurons has been reported in brain slices of rat somatosensory cortex (Markram et al. 1997). Similarly, inhibitory connections between fast spiking interneurons and pyramidal neurons in slices of adult rat neocortex were found, on average, in 1 in 15 pairs (Thomson et al. 1996). However, Thomson et al. (1996) also noted that both slice thickness and preparation procedure appeared to affect this proportion. Using thick (500 µm) slices prepared from animals that had been perfused transcardially with ice-cold sucrose-containing ACSF, this probability rose to 7 in 32 pairs. The detection of any inhibitory interactions in the present study is, however, in contrast to an earlier study in which inhibitory interactions were not detected (Jaeger et al. 1994). Apart from technical factors, interactions may have previously gone undetected because of the relatively high failure rate of unitary IPSPs that we observed (mean, 38%). Even though the postsynaptic response was sometimes detected in a single sweep, it was only reliably detected by averaging 200 sweeps. The findings of the present study thus help to resolve an apparent contradiction between the known anatomy and physiology of striatal projection neurons. The physiology now correlates very well with the anatomical evidence for local synaptic connectivity between spiny projection neurons (Kitai and Wilson 1982; Somogyi et al. 1981; Wilson and Groves 1980).

Properties of the unitary IPSP

The observed amplitude of the IPSP evoked by a single action potential in a presynaptic spiny projection neuron is less than that evoked by presynaptic action potential firing of the GABAergic interneurons of the striatum. Koos and Tepper (1999) report that the IPSP evoked in a postsynaptic spiny projection neuron by a single presynaptic action potential in a GABAergic interneuron ranged from 330 to 2,130 µV (mean: 1,060 ± 220 µV, n = 7). This compares with a range of 157-319 µV (mean: 277 ± 46 µV, n = 9) for the IPSP evoked by a single presynaptic action potential in a spiny projection neuron in the present study. The higher amplitude of the IPSP evoked by GABAergic interneurons correlates with the anatomy of their terminals, which are large and probably make multiple terminations on each spiny projection neuron they innervate (Kita 1993). The amplitude of the unitary IPSP is also compatible with those reported in other brain areas. In the cerebral cortex, single axon IPSPs elicited in pyramidal cells by interneurons have been reported to range from 200 to 3,500 µV in amplitude (Thomson et al. 1996). In the latter study, the connections involving several boutons resulted in the smaller IPSPs recorded, and some connections involving two boutons went undetected. The largest IPSPs recorded might have resulted from 12 to 20 boutons (Thomson et al. 1996). When placed in this context, the amplitude of the unitary IPSP evoked by a single action potential in a presynaptic spiny projection neuron is consistent with the apparently smaller number of boutons involved in the synaptic connection between spiny projection neurons (Somogyi et al. 1982; Wilson and Groves 1980).

Previous studies have used intrastriatal stimulation to characterize inhibitory interactions between spiny projection neurons. Application of focal bipolar electrical stimulation evokes a GABA-mediated IPSP with two components exhibiting a differential sensitivity to GABA_B agonists (Seabrook et al. 1991). It has been suggested that the fibers that evoke GABA_B agonist-insensitive synaptic potentials originate from the recurrent collaterals of spiny projection neurons (Radnikow et al. 1997). However, in studies using intrastriatal stimulation, it is difficult to separate the effects of spiny projection neurons from the GABA interneurons, which also produce strong inhibitory responses in the spiny projection neurons (Koos and Tepper 1999). Nevertheless, studies using intrastriatal stimulation have shown that, as in other systems, GABA synapses in the striatum are highly modifiable and display short-term activity-dependent plasticity. Both paired-pulse depression of the IPSP evoked by intrastriatal stimulation (Radnikow et al. 1997) and synaptic augmentation (Fitzpatrick et al. 2001) have been described. Thus it should be noted that the characteristics of the synaptic response described in the present study are based on the physiological characteristics of an IPSP evoked by a single action potential, and the response might be increased or decreased by the repetitive firing patterns of the presynaptic neurons.

Implications for current and future theoretical models of striatal function

The demonstration of synaptic interactions supports previous proposals that inhibitory interactions between spiny projection neurons may be a key determinant of the signal processing operations performed in the striatum (Beiser and Houk 1998; Wickens et al. 1991; Wilson and Groves 1980). One of these proposals is the "winner takes all" model, which is based on the premise that spiny projection neurons are mutually connected (Bar-Gad and Bergman 2001; Gillies and Arbuthnott 2000; Wickens and Oorschot 2000). In the present study, however, the connections found were in one direction only, and no mutually inhibitory interactions were detected. The absence of mutually inhibitory interactions in the current sample does not rule these out as an infrequent chance event, which may have functional significance. However, as the findings indicate no bias toward mutual interactions, such interactions are unlikely to dominate the striatal dynamics.

Many theoretical studies have emerged regarding the role of the basal ganglia in brain function (Amos 2000; Berns and Sejnowski 1998; Redgrave et al. 1999; Suri and Schultz 1999). In these theoretical models, the functional importance of the striatum is paramount, although the principles underlying its operation have so far been based on uncertain anatomical and physiological data. We anticipate that the definitive physiological data on the functional properties of the striatum that we present here will serve to spark a new and influential generation of theoretical models of basal ganglia operation.