Whole cell patch-clamp recordings were obtained from projection neurons and interneurons of the rat basolateral amygdala (BLA) to understand local network interactions in morphologically identified neurons and their modulation by serotonin. Projection neurons and interneurons were characterized morphologically and electrophysiologically according to their intrinsic membrane properties and synaptic characteristics. Synaptic activity in projection neurons was dominated by spontaneous inhibitory postsynaptic currents (IPSCs) that were multiphasic, reached 181 ± 38 pA in amplitude, lasted 296 ± 27 mS, and were blocked by the GABAA receptor antagonist, bicuculline methiodide (30 μM). In interneurons, spontaneous synaptic activity was characterized by a burst-firing discharge patterns (200 ± 40 Hz) that correlated with the occurrence of 6-cyano-7-nitroquinoxaline-2,3-dione-sensitive, high-amplitude (260 ± 42 pA), long-duration (139 ± 19 mS) inward excitatory postsynaptic currents (EPSCs). The interevent interval of 831 ± 344 mS for compound inhibitory postsynaptic potentials (IPSPs), and 916 ± 270 mS for EPSC bursts, suggested that spontaneous IPSP/Cs in projection neurons are driven by burst of action potentials in interneurons. Hence, BLA interneurons may regulate the excitability of projection neurons and thus determine the degree of synchrony within ensembles of BLA neurons. In interneurons 5-hydroxytryptamine oxalate (5-HT) evoked a direct, dose-dependent, membrane depolarization mediated by a 45 ± 6.9 pA inward current, which had a reversal potential of −90 mV. The effect of 5-HT was mimicked by the 5-HT2 receptor agonist, α-methyl-5-hydroxytryptamine (α-methyl-5-HT), but not by the 5-HT1A receptor agonist, (±) 8-hydroxydipropylaminotetralin hydrobromide (8-OH-DPAT), or the 5-HT1B agonist, CGS 12066A. In projection neurons, 5-HT evoked an indirect membrane hyperpolarization (∼2 mV) that was associated with a 75 ± 42 pA outward current and had a reversal potential of −70 mV. The response was independent of 5-HT concentration, blocked by TTX, mimicked by α-methyl-5-HT but not by 8-OH-DPAT. In interneurons, 5-HT reduced the amplitude of the evoked EPSC and in the presence of TTX (0.6 μM) reduced the frequency of miniature EPSCs but not their quantal content. In projection neurons, 5-HT also caused a dose-dependent reduction in the amplitude of stimulus evoked EPSCs and IPSCs. These results suggest that acute serotonin release would directly activate GABAergic interneurons of the BLA, via an activation of 5-HT2 receptors, and increase the frequency of inhibitory synaptic events in projection neurons. Chronic serotonin release, or high levels of serotonin, would reduce the excitatory drive onto interneurons and may act as a feedback mechanism to prevent excess inhibition within the nucleus.
Intracellular recordings from morphologically identified neurons of the basolateral complex of the amygdala (BLA) have demonstrated that projection neurons and interneurons can be distinguished according to their electrophysiological properties (Rainnie et al. 1993; Washburn and Moises 1992). Furthermore several intracellular current-clamp studies have examined synaptic transmission occurring in the basolateral complex in response to activation of intrinsic and/or extrinsic afferent inputs (Gean and Chang 1992; Rainnie et al. 1991a,b; Sugita et al. 1993; Washburn and Moises 1992). Irrespective of the point of stimulation, the most prominent synaptic potentials evoked are a glutamatergic excitatory postsynaptic potential (EPSP) and two GABAergic inhibitory postsynaptic potentials (IPSPs), one fast and one slow. The EPSP is mediated by activation of bothN-methyl-d-aspartate (NMDA) and AMPA/kainate subtypes of glutamate receptor (Gean and Chang 1992;Rainnie et al. 1991a). The fast IPSP is mediated by activation of GABAA receptors, whereas the slow IPSP is mediated by activation of GABAB receptors (Rainnie et al. 1991b).
These in vitro observations have been substantiated further by experiments that have demonstrated that similar excitatory and inhibitory processes occur in the BLA in vivo (Brothers and Finch 1985; Fi et al. 1996; Lang and Paré 1997a,b; Mello et al. 1992;Paré et al. 1995). However, the relatively high resistance of the microelectrodes used in these studies introduced undesirable noise into voltage-clamp records and caused “space-clamp” problems. Whole cell patch-clamp recording has overcome these problems, and this technique now has been applied to the amygdala slice preparation (Keele et al. 1997;Neugebauer et al. 1997; Rainnie 1995;Smith and Dudek 1996). Using this technique,Smith and Dudek (1996) have confirmed the predominance of a glutamatergic and a GABAergic drive onto BLA neurons in juvenile rats. Moreover on the basis of their observations of spontaneous activity in the BLA, these authors have proposed that local feedforward and/or feedback circuitry regulates neuronal activity within this complex.
The dynamic interaction between the relative expression of EPSPs and IPSPs within the basolateral complex can directly influence the input/output relationship for the amygdala. Hence pharmacological manipulations of either glutamatergic or GABAergic transmission in the BLA can markedly alter behavioral responses in vivo (Davis et al. 1994). Consequently if monoamine neuromodulators, such as serotonin, regulate the normal balance between excitatory and inhibitory transmission, they may be expected to alter signal processing within the amygdala and hence modify any behavioral responses dependent on this processing. Support for this hypothesis comes from the observation that the amygdala receives a prominent innervation from serotonergic terminals originating from neurons of the dorsal raphé nucleus (Bobillier et al. 1976;Ma et al. 1991; Sadikot and Parent 1990), stimulation of the dorsal raphé results in the facilitation of firing in a subset of amygdala neurons (Jacobs 1972), iontophoretic application of 5-hydroxytryptamine (5-HT) into the rat amygdala reduced glutamate-induced cell firing in vivo (Mah and Cunningham 1993, Stutzman et al. 1998), and serotonin release in the amygdala is increased during states of stress and anxiety (Fernandes et al. 1994; Gargiulo et al. 1996; Kawahara et al. 1993; Kirby et al. 1995; Rueter and Jacobs 1996).
Serotonin receptors can be grouped into seven classes of receptor, named 5-HT1 through 5-HT7 (see Hoyer 1996). In addition to there being multiple receptor subtypes, individual 5-HT receptors also show a differential distribution within the CNS. Hence the amygdala has moderate to high levels of binding sites for 5-HT1A, 5-HT2, 5-HT3, 5-HT4, and 5-HT6 receptors (Barnes et al. 1989; Eglen et al. 1995; Morales et al. 1996; Morilak et al. 1993; Radja et al. 1991; Ward et al. 1995). Even within the amygdala regional differences exist. Hence 5-HT1A receptors are found predominantly in the central nucleus, whereas 5-HT2, 5-HT3, and 5-HT6 receptors are found predominantly in the basolateral complex.
In many areas of the CNS, the principal receptor mediating postsynaptic 5-HT-induced inhibition is the 5-HT1A receptor subtype (seeSaxena 1995). However, the low levels of 5-HT1A receptor expression observed in the BLA suggested that the 5-HT-mediated inhibition of BLA cell firing (see preceding text) may not result from a direct activation of postsynaptic 5-HT1A receptors but, rather, by an indirect inhibition of glutamate release via activation of presynaptic 5-HT1Areceptors (Bobker and Williams 1989). Moreover, expression of 5-HT3 receptors on interneurons of the BLA (Morales et al. 1996) and 5-HT2 receptors suggested that inhibition of cell firing also may be facilitated by a direct excitation of GABAergic interneurons (Alreja 1996; Sheldon and Aghajanian 1991).
These hypothesis were tested using whole cell patch-clamp recording from morphologically, and electrophysiologically, identified neurons of BLA in vitro. Spontaneous synaptic currents were examined to establish if an intrinsic connectivity exists between projection neurons and interneurons of the BLA slice preparation. Identified neurons then were examined for their intrinsic membrane response to application of serotonin. The effects of serotonin on spontaneous and evoked synaptic inputs to these same neurons also was examined. Finally the response of neurons to specific 5-HT receptor agonists and antagonists were then examined to isolate the 5-HT receptor subtypes mediating these responses.
The basolateral complex, composed of the lateral and basolateral nuclei, was visualized easily in the coronal slice in vitro as it was outlined laterally by the white matter tract of the external capsule (corpus callosum) and medially by the white matter tract of the longitudinal association bundle. Neuronal responses reported in this study were obtained only from neurons located in the basolateral subdivision of the basolateral complex.
Slices (500-μm thick) containing the basolateral complex were obtained from isoflurane anesthetized 24- to 56-day-old male Long-Evans rats and prepared using procedures as previously described (seeRainnie et al. 1991a). Slices were maintained fully submerged in a tissue chamber and continuously perfused with artificial cerebrospinal fluid (ACSF) of the following composition: (in mM) 124 NaCl, 3.75 KCl, 1.25 KH2PO4, 1.3 MgCl2, 3.5 CaCl2, 26 NaHCO3, and 10 glucose. The ACSF was heated to 32 ± 2°C and gassed with a 95%–5% oxygen/carbon dioxide mixture. Whole cell patch-clamp recordings were obtained using the technique of Blanton et al. (1989). Briefly, borosilicate glass patch electrodes (resistance 6–8 MΩ) were pulled on a Flaming/Brown micropipette puller (Model P-87) and filled with (in mM) 120 K-gluconate, 10 phosphocreatinine, 10 KCl, 10 HEPES, 3 MgCl2, 2 MgATP, and 0.2 NaGTP and biocytin 0.45%. Patch solution was buffered to pH 7.2 with KOH and filtered through a 0.2-μm filter (Altech Associates, IL) before use. Final osmolarity of the patch solution was 280 mOsm. Recordings were made with an Axopatch-1D amplifier (Axon Instruments, Burlingame, CA) using pClamp 6.02 software and a Digidata 1200 A-D interface (Axon Instruments). For patch electrodes, a seal resistance was considered acceptable if it was >1.5 GΩ and had a series resistance of <20 MΩ. Afferent fibers were stimulated with a concentric bipolar stimulating electrode (MCE-100; Kopf Instruments), which was placed on the central nucleus of the amygdala through which course the fibers of the stria terminalis.
Whole cell patch-clamp records initially were established in current-clamp mode and then switched to voltage clamp. This procedure allowed the viability of the cell to be determined before voltage-clamp experiments were performed. Only those BLA neurons that showed a stable resting membrane potential more negative than −55 mV for >5 min and an action potential that overshot +5 mV were considered acceptable for further analysis.
All current- and voltage-clamp paradigms were computer controlled using either clampex or fetchex data acquisition programs (Axon Instruments), and data were sampled at a frequency determined by the speed of the response to be measured. Voltage excursions in response to hyperpolarizing current injection were used to determine the membrane input resistance and were sampled at 10 kHz. Voltage excursions in the depolarizing direction were sampled at 20 kHz. This was done to ensure an accurate evaluation of action potential height and duration. Data were filtered at 5 kHz in current clamp and at 2 kHz in voltage clamp. For acquisition of spontaneous PSP/Cs, data were sampled at 20 kHz. Spontaneous PSP/Cs were sampled continuously for 30 s. Miniature spontaneous PSCs were recorded in the presence of TTX (0.6 μM) and also sampled at 20 kHz for 30 s. Only those miniature events that displayed a rise time of ≤0.8 mS were included in the analysis to reduce contamination by events originating in electrotonically distant regions the amplitude of which may not reflect the true synaptic current. Synaptic events were analyzed semi-automatically using the Fetchan analysis package of the pClamp software bundle, and the amplitude and time constant of decay (τ) for EPSCs and IPSCs was calculated for each neuron examined during one continuous 30-s sweep. The accuracy of the analysis was confirmed with an off-line examination of marked traces in Clampfit. Events were averaged, and only those sweeps containing >100 events were included in the analysis. In projection neurons, miniature IPSCs could not be clearly discerned in the presence of TTX, consequently the effects of 5-HT on presynaptic GABA release was not examined. To examine the voltage dependency of the effects of applied drugs, cells were held at −60 mV in voltage clamp and then “ramped” from −100 to −40 mV during a 6-s period. The ramp then was repeated in the presence of the drug, and subtraction of the control current from the drug-induced current allowed the voltage dependency of the pure drug-induced current to be determined.
Means, SD, and SE were calculated for each treatment and two-tailed, paired and unpaired Student’s t-tests or a single factor ANOVA test were performed as necessary. Data are expressed as means ± SE and significance was accepted ifP < 0.05.
Drugs were applied directly to the ACSF using a continuous gravity-fed bath application. This application allowed accurate dose-response relationships to be constructed because the extracellular fluid reached equilibrium with the concentration of applied drug in ∼2 min. Drugs applied were: 5-HT (5–100 μM); α-methyl-5-hydroxytryptamine (α-methyl-5-HT, 50–100 μM); (±) 8-hydroxydipropylaminotetralin hydrobromide (8-OH-DPAT, 1–10 μM); CGS-12066B maleate (20–50 μM); TTX (0.6–1.2 μM); −[2-[4-(2-methoxyphenyl)-1-piperazinyl[ethyl]-N-2-pyridinyl-cyclohexane-carboxamide maleate (WAY 100635, 20–40 nM); 6-cyano-7-nitroquinoxaline-2,3,-dione (CNQX, 10 μM); bicuculline methiodide (Bic, 20–50 μM); and 2-hydroxy-saclofen (2-OH-Sac, 100 μM) from Research Biochemicals International (Natick, MA).
During the course of each experiment, biocytin diffused from the patch electrode into the recorded cell. No current injection protocol was necessary to load the cells with biocytin. At the termination of each experiment, the 500-μm slice containing the loaded cell was removed from the recording chamber and placed in 4% paraformaldehyde overnight. For rapid identification of cell types, slices were washed with 0.1 M phosphate buffered saline (PBS) and then incubated for ≥1 h in PBS containing 0.5% Triton X-100. Slices then were incubated for 2 h in PBS-Triton X-100 supplemented with 40 μL/ml Texas Red-Avidin D conjugate (Vector Laboratories) to visualize the neurons. The slices were washed in PBS and mounted on the chuck of a Vibroslice (Campden Instruments) for further resectioning to 80 μm. Resectioned slices were mounted on gelatin-subbed microscope slides, air dried overnight, mounted in AM 100 fluorescence-free, permanent mounting media (Chemicon; Temecula, CA), and visualized using a Nikon Diaphot microscope equipped with epifluorescence. In addition, 44 cells were characterized electrophysiologically and then shipped to the University of South Carolina School of Medicine for independent morphological identification and determination of location by Dr. A. J. McDonald (see Fig. 2 A, 1–3).
Stable whole cell records were obtained from 205 neurons of the basolateral nucleus, 13% of which were considered to be interneurons because of their morphological and electrophysiological characteristics. Namely, a spine-sparse dendritic arborization, a nonpyramiform cell soma, a high-input resistance (∼121 MΩ), and expression of a high-frequency train of action potentials in response to depolarizing current injection (Fig.1 B) (see also Rainnie et al. 1993). Of the remaining neurons, 86% resembled pyramidal projection neurons (Fig. 1 A) with a prominent apical dendrite, a spine dense dendritic arbor, and low input resistance (∼57 MΩ). One neuron had a neurogliaform-like appearance and firing properties similar to that of a projection neuron. All of the neurons the position of which could be verified anatomically (86) were located in the anterior and posterior subdivisions of the basolateral nucleus between −2.3 and −3.3 mm Bregma, a representative sample of which is shown in Fig.2 A, 1-3. The majority of neurons (80%) were located in the rostral BLA between −2.3 and −2.8 mm Bregma, the remaining 20% were located caudal to Bregma −3.14. Moreover, 76% were located in the anterior subdivision. However, no differences were observed in the electrophysiological characteristics of neurons recorded either from the rostrocaudal aspect or from the anterior or posterior subdivisions of the nucleus.
Spontaneous synaptic currents
The majority of neurons in this study (80%) displayed spontaneous synaptic activity in which, at a membrane potential close to rest (−60 mV), the predominant synaptic activity in projection neurons was low-frequency, long-duration (250–500 ms) IPSP/Cs and high-frequency, short-duration (15–25 mS) EPSP/Cs (Fig.3 A). These results are in agreement with previous sharp electrode studies (see Rainnie et al. 1991b). In current clamp, the waveform of the long-duration spontaneous IPSPs fell into two categories: those that appeared to be “pure” IPSPs (47%) and those IPSPs in which the initial rising phase was corrupted by a barrage of EPSPs (53%). The decay rate of the spontaneous pure IPSPs could be fit by a single exponential with time constants (τ) of 105 ± 36 mS (n = 8). This time constant is more than three times as long as those reported for monosynaptic GABAA-mediated IPSPs in other areas of the CNS (see Buhl et al. 1995; Ropert and Guy 1991), suggesting that the spontaneous IPSPs observed in the BLA were compound IPSPs. Projection neurons also revealed spontaneous “unitary,” or monosynaptic, IPSPs with a peak amplitude of 0.76 ± 0.1 mV (n = 4) at the resting membrane potential.
In voltage clamp, the outward currents mediating the compound IPSPs were typically multiphasic, reached 181 ± 38 pA (n = 16) in amplitude, and lasted 296 ± 27 mS at a holding potential of −60 mV (Fig. 3 B). Similar to current-clamp observations, the initial outward current often was contaminated by multiple inward EPSCs (Fig. 3 C1). However, those neurons in which the IPSC decay appeared monophasic showed a time constant of decay (τ) = 92 ± 10 mS (n = 10). Moreover spontaneous unitary IPSCs were observed in several neurons (Fig. 3 C2, *) with a peak amplitude of 12.5 ± 2.1 pA and a τ of 21.1 ± 2.0 mS (n = 5). In addition, during voltage “ramp” protocols, it was observed that in 53% (16/30) of projection neurons examined the outward current mediating the compound IPSC was biphasic. The initial fast component had a reversal potential of −70 mV, whereas a second, slower, component had a reversal potential close to −85 mV. The remaining 47% of spontaneous IPSCs were monophasic and had a reversal potential close to −70 mV. Interestingly, both the monophasic IPSC and the biphasic IPSC were abolished by application of the GABAA-receptor antagonist, bicuculline methiodide (30 μM, Fig. 3 D; n = 5). Moreover, the slow spontaneous IPSC was insensitive to 2-hydroxy-saclofen (n = 4). These data suggest that for spontaneous inhibitory currents, both the fast and the slow IPSC were mediated by activation of GABAA receptors. Fast EPSCs were observed in all projection neurons and had a mean amplitude of 25 ± 3.5 pA, a τ of 12.5 ± 2.2 mS (n = 10), and were abolished by CNQX (10 μM,n = 5).
In interneurons (Fig. 4), the spontaneous activity consisted primarily of fast high-frequency EPSPs (Fig.4 B1) that were interspersed with bursts of action potentials riding on a depolarizing wave of summated EPSPs (Fig. 4 B2). Within each burst, action potentials fired at a frequency of 200 ± 40 Hz (n = 8). In voltage clamp, the bursts of action potentials were seen to correlate with the occurrence of high-amplitude (260 ± 42 pA, n = 11), long-duration (139 ± 19 mS) inward currents (Fig. 4 C) that had characteristics similar to a synchronous burst of EPSCs (Fig.3 D). Individual EPSCs had a mean amplitude of 46 ± 5 pA and a rapid τ (2.8 ± 0.2 mS; n = 4). Both the fast EPSCs and the synchronous bursts of EPSCs were blocked fully by CNQX (10 μM, not shown).
The low frequency and long duration of the spontaneous, compound IPSP/Cs suggested that they may arise from intrinsic network activity within the amygdaloid complex. Support for this hypothesis came from the observations that raising the external K+ concentration from 5 to 8.5 mM caused a significant increase in the frequency of occurrence of spontaneous IPSCs (Fig.5 A, n = 4) from 0.8 to 1.3 Hz, that compound IPSP/Cs were abolished by exogenous CNQX (10μM, not shown), and that a dorsoventral transection of the slice at the level of the external capsule, to remove reciprocal connectivity of the BLA with the entorhinal and piriform cortex (Fig.2 B), failed to prevent spontaneous inhibitory activity. Furthermore an apparent correlation was observed between the frequency of spontaneous bursts in interneurons and the frequency of spontaneous IPSPs in projection neurons. Interevent-interval histograms generated for either burst firing in interneurons (n = 6) or spontaneous IPSPs in projection neurons (n = 10), supported this hypothesis (Fig. 5 B). When a curve was fit to the data using a Marquardt least-squares analysis fitting protocol, the modal inter-IPSP interval was 831 ± 344 mS, whereas the modal interburst interval was 916 ± 270 mS. These data suggest that the occurrence of IPSPs in projection neurons probably is driven by the occurrence of bursts in interneurons. Furthermore because burst firing rarely is observed in projection neurons, this implies that a synchronized input from two or more projection neurons would be required to drive a single interneuron. Conversely this data also suggests that a single interneuron may innervate multiple projection neurons.
Serotonergic modulation of BLA neuronal properties
Once the electrophysiological characteristics of a cell had been ascertained, the effect of 5-HT on its intrinsic membrane properties was examined. In all cases, exogenous 5-HT was applied to neurons held at −60 mV with DC current injection in current clamp or when voltage-clamped at −60 mV.
The effect of 5-HT on the intrinsic membrane properties of projection neurons was small, but variable, and appeared independent of the dose applied (5–100 μM). Hence 5-HT evoked a reversible, membrane hyperpolarization in 54% of the neurons tested (n = 23) and had no effect in the remaining 46% of the neurons (n = 19). The effects of increasing the concentration of 5-HT in those neurons that did respond is shown in Fig.6 A1 (solid bars). Here, increasing concentrations of 5-HT appear to increase the membrane response in projection neurons; however, the observed difference was not statistically significant (ANOVA, P = 0.66). For example, 20 μM 5-HT evoked a mean hyperpolarization of −2 ± 0.4 mV (n = 4), whereas 100 μM 5-HT evoked a hyperpolarization of only −3 ± 0.5 mV (n = 4). In those neurons that did respond to 5-HT with a membrane hyperpolarization, the membrane input resistance was reduced by 7 ± 3 MΩ (n = 6). However, the reduction of input resistance did not alter the firing properties of BLA projection neurons in response to depolarizing current injection.
In voltage clamp, application of 5-HT (50 μM) evoked an outward current of 36 ± 20 pA in those projection neurons that showed a response to 5-HT (n = 11). A comparison of the membrane conductance evoked during voltage ramp protocols, before and during 5-HT application, revealed a reversal potential for the 5-HT-induced current of −66 ± 1.9 mV (n = 9; Fig.6 A2) and no obvious reversal potential in the remaining two neurons. In a 5-HT-responsive neuron, application of baclofen (2 μM), a selective GABAB receptor agonist, evoked a marked outward current (120 pA) that had a reversal potential at −98 mV, which is close to the reversal potential for K+ ions in this experimental paradigm (Fig. 6 C1). The response of BLA projection neurons to baclofen is mediated by a G-protein-dependent increase in a K+ conductance (see Asprodini et al. 1992). Hence the marked response of all projection neurons tested to baclofen (153 ± 26 pA; n = 4) suggests that G-protein-mediated responses are fully functional in patched BLA neurons. Consequently the small outward current in response to 5-HT may result from either the indirect blockade of a tonic excitatory input, an indirect increase in an inhibitory input, or from a direct, but minimal, effect on postsynaptic 5-HT receptors. Application of TTX (0.6–1.2 μM; n = 6) abolished spontaneous EPSCs and IPSCs in four of six neurons examined and also abolished the effects of exogenous 5-HT application (Fig.7 A, 1 and2). However, in two neurons subsequent application of 5-HT still evoked a small outward current (46 ± 60 pA), although no clear reversal potential was observed (not shown). As a hyperpolarizing response to 5-HT is most often attributable to activation of 5-HT1A receptors in many areas of the CNS (seeSaxena 1995), the effects of the relatively specific 5-HT1A receptor agonist, 8-OH-DPAT and of the specific 5-HT1A receptor antagonist WAY 100635 were investigated. Application of 8-OH-DPAT (1–5 μM) had no effect on the membrane potential of seven of eight neurons tested and caused a small hyperpolarization (−1.5 mV) in the remaining neuron. Moreover, WAY 100635 (20–40 nM) had no intrinsic effects on BLA neurons and was ineffective in blocking the 5-HT-induced hyperpolarization in all neurons examined (n = 5), suggesting that 5-HT1A receptors play little, if any, role in the 5-HT-induced hyperpolarization of projection neurons (Fig. 7,C and D). In contrast, the 5-HT analogue α-methyl-5-HT (100 μM), a more specific 5-HT2 receptor agonist, mimicked the effects of 5-HT and evoked a membrane hyperpolarization in five of six neurons examined (−2.5 ± 1.5 mV) and had no effect on the remaining neuron.
Unlike the variable response found in projection neurons, the effects of 5-HT on interneurons was more consistent. Hence, 5-HT evoked a membrane depolarization in 10 of 13 neurons examined and had no effect in the remaining 3 neurons. Moreover the response to 5-HT was unaffected by prior application of either TTX or CNQX (data not shown), and interneurons showed a dose-response relationship with increasing concentrations of 5-HT applied (see Fig. 6 A1, stippled bars). At a concentration of 20 μM, 5-HT evoked a depolarization of 4.0 ± 0.8 mV (n = 4), whereas at 50 μM 5-HT evoked a significantly larger depolarization of 9.0 ± 1.3 mV (n = 6). ANOVA, at the 95% confidence level used for significance in this study, revealed a statistical significance between responses to 5-HT in interneurons (P = 0.09). A typical response to 5-HT (50 μM) is shown in Fig. 6 B. In current clamp, the depolarization always was associated with a concomitant increase in membrane input resistance. α-Methyl-5-HT (100 μM,n = 5) mimicked the effects of 5-HT in the majority of interneurons: evoking a depolarization (8–10 mV) and an associated increase in membrane input resistance in two neurons and an inward current (59 pA) in one neuron and having no effect in the remaining two neurons. In those interneurons that were responsive to 5-HT, neither 8-OH-DPAT (n = 2) nor CGS 12066A, a specific 5-HT1B-receptor agonist (n = 2), had any effect on the resting membrane potential. In voltage clamp, 5-HT (50 μM) evoked an inward current 45 ± 6.9 pA in four of five interneurons examined, and a comparison of the membrane conductance before and during 5-HT application revealed a reversal potential close to −90 mV (Fig. 6 C2). These data suggest that one of the primary functions of serotonergic transmission within the BLA is to increase the excitability of inhibitory interneurons possibly via activation of 5-HT2 receptors, which, in conjunction with the small inhibition of projection neurons, would result in a dampening of the input-output response of the BLA.
Serotonergic modulation of synaptic transmission
REGULATION OF NETWORK ACTIVITY.
During 5-HT application the most consistent observation was a marked alteration in the occurrence of spontaneous synaptic potentials (n = 47). Irrespective of the intrinsic response of individual BLA neurons, 5-HT caused an alteration in either the frequency, and/or the amplitude, of spontaneous synaptic potentials/currents in all neurons examined. Moreover the effect of 5-HT was both time and concentration dependent.
In all projection neurons examined, continuous bath application of 5-HT (50 μM) evoked a marked reduction in the amplitude of the GABAA-mediated IPSP/Cs that was independent of the effects of 5-HT on membrane potential. Hence in those neurons in which 5-HT evoked a small hyperpolarization, DC current injection to the predrug membrane potential (−60 mV) failed to reverse the effects of 5-HT on the IPSP/C amplitude. On closer examination of the effects of 5-HT on spontaneous IPSP/Cs, it was noted that during the initial 1–1.5 min of perfusion, the rate of spontaneous IPSP/Cs increased significantly from 0.7 ± 0.1 Hz to 1.1 ± 0.1 Hz (Fig.8 A, middle; P < 0.05; n = 11). With continued application (3–4 min), however, the IPSP/C amplitude decreased (Fig.8 A, bottom) until, in some cases, the IPSP/Cs were abolished (n = 4). In addition, α-methyl-5-HT (100μM) mimicked the initial increase in spontaneous IPSP/C frequency (n = 7) and the subsequent decrease in amplitude, whereas in four neurons perfused with 5-HT at 100 μM, the IPSP/Cs were abolished during the initial period of the perfusion.
A similar pattern of initial excitation was observed in the burst firing responses of interneurons (Fig. 8 B). The frequency of spontaneous bursts increased from 0.8 ± 0.2 to 1.4 ± 0.3 Hz (n = 4) during the initial 5-HT application (Fig.8 B, middle), while the duration of the burst decreased. With continued application, bursts of action potentials were abolished in three of four neurons and replaced by high-frequency single action potentials. A similar response pattern was observed in two interneurons voltage-clamped at −60 mV, where compound inward currents were replaced by high-frequency single EPSCs. An increase in burst frequency and decreased duration also was observed with application of α-methyl-5-HT (100 μM) in four neurons examined. In two neurons tested at 100 μM, 5-HT abolished the bursts of EPSP/Cs in the initial perfusion period.
A typical dose-response relationship for the 5-HT-induced blockade of spontaneous IPSCs in projection neurons is shown in Fig.9, A–D. Here, 20 μM 5-HT increased the frequency, but reduced the amplitude, of spontaneous IPSCs. Increasing the 5-HT concentration to 50μM further reduced the amplitude of the IPSCs, whereas 100 μM 5-HT fully blocked the IPSCs. In no instance did 8-OH-DPAT (1–5 μM; n = 5) or CGS 12066A (1–5 μM; n = 3) affect the occurrence of spontaneous synaptic potentials in the BLA. In agreement with data obtained with 8-OH-DPAT, the 5-HT1A receptor antagonist WAY 100635 (20–40 nM) alone had no effects on spontaneous synaptic activity and was ineffective in blocking the increased frequency or decreased amplitude, of spontaneous IPSPs in the presence of 5-HT (50 μM; n = 5).
REGULATION OF TRANSMITTER RELEASE.
In both projection neurons and interneurons application of 5-HT caused a reduction in the amplitude of stimulus evoked EPSP/Cs and IPSP/Cs (see Fig. 10). In projection neurons, 5-HT caused a dose-dependent reduction of the evoked EPSC, whereby the EPSC amplitude was reduced 75 ± 9% by 100 μM 5-HT (n = 8), 43 ± 16% by 50 μM 5-HT (n = 4), and 34 ± 13% by 20 μM 5-HT (n = 4). In contrast, the effect of 5-HT on the evoked IPSP/C was more profound and also more sensitive to lower concentrations of 5-HT. Hence, 100 μM 5-HT reduced the IPSP/C amplitude by 90 ± 10% (n = 8), 50 μM 5-HT by 92 ± 3% (n = 4), and 20 μM 5-HT by 56 ± 23% (n = 4). In addition, in those interneurons in which a distinct EPSC could be evoked (n = 3), 50 μM 5-HT reduced the amplitude by 65% (Fig. 10 B). In those neurons in which a biphasic IPSC was observed, 5-HT (50 μM) caused a reduction in the amplitude of both the evoked and spontaneous fast and slow IPSC (Fig. 10 C).
In most cases, the evoked IPSP/C observed in BLA projection neurons results from activation of a CNQX-sensitive feedforward inhibitory pathway (see Rainnie et al. 1991b). Consequently the marked reduction of the evoked IPSC by 5-HT and the reduction of evoked EPSCs in interneurons suggested that there was a presynaptic locus for serotonergic modulation of glutamatergic transmission within the BLA. Furthermore the response of projection neurons to picospritzed AMPA, a specific agonist at the AMPA/kainate subtype of glutamate receptor, was unaffected by application of 5-HT (not shown), which would exclude the possibility of a 5-HT-induced reduction in affinity of postsynaptic glutamate receptors as has been reported in the spinal cord (Holohean et al. 1995). Examination of the spontaneous EPSCs recorded in interneurons before and during 5-HT application revealed that the propensity for EPSCs to summate was reduced greatly in the presence of 5-HT (not shown). A detailed examination of amplitude-frequency histograms generated from individual EPSCs captured before, and during, 5-HT application showed no significant difference in the distribution of the EPSC amplitudes (Fig.11 A). This probably results from a mixture of local, network, and intrinsic responses contributing to a large variance in the amplitude of EPSCs in control conditions. However, during analysis of the effects of 5-HT on postsynaptic responses, it became clear that in some neurons TTX-resistant miniature mEPSCs could be observed (Fig. 11 D, top).
Several groups have used TTX-resistant mEPSCs and mIPSCs to examine presynaptic regulation of unitary quantal events in the hippocampus (see Edwards et al. 1990; Jonas et al. 1993; Poncer et al. 1995; Scanziani et al. 1992). Consequently TTX-resistant mEPSCs were used to examine the effects of 5-HT on glutamate release in the BLA. A plot of amplitude versus cumulative probability showed no significant difference in the distribution of mEPSCs before or during 5-HT application (Fig. 11 B), suggesting that 5-HT did not reduce quantal content. In contrast, 5-HT markedly reduced the frequency of mEPSCs (Fig. 11 B, inset), suggesting that 5-HT reduces the probability of release from the presynaptic terminal but not the number of quanta that are released. When the peak mEPSC was plotted as a function of frequency (Fig. 11 C), equidistant peaks were observed in the mEPSC histograms, suggesting that some mEPSCs represent multiquantal events. The two clearest peaks corresponded to mEPSCs of 5 and 10 pA, neither peak was changed with 5-HT application. These data suggest that an important function of serotonergic transmission in the amygdala may be to regulate excitatory drive onto inhibitory interneurons.
The results of this study demonstrate that neuronal activity in the BLA is regulated by a complex reciprocal interaction between projection neurons and interneurons, whereby synaptically driven burst firing in interneurons may result in the concomitant expression of compound IPSCs in projection neurons. This feedback inhibition could result in synchronous activity in ensembles of BLA projection neurons. Moreover this interaction, and hence the input-output relationship of the BLA, can be modulated by 5-HT in two distinct ways.
The primary action of exogenous 5-HT was a dose-dependent excitation of interneurons probably via activation of postsynaptic 5-HT2receptors. This excitation resulted in a concomitant indirect inhibition of projection neurons via an increased release of GABA and subsequent activation of postsynaptic GABAA receptors. In a minority of projection neurons (<12%), 5-HT also had a direct inhibitory action via an activation of postsynaptic 5-HT1A-receptors.
With higher concentrations (>50 μM) or prolonged application, 5-HT had an additional indirect effect on neuronal excitability by reducing the probability of glutamate release from presynaptic glutamatergic terminals. Although the 5-HT receptor subtype mediating this response remains to be determined, it is not mediated by either 5-HT1A or 5-HT1B receptors.
It is proposed that a target-specific, serotonergic input from the brain stem raphé nuclei may regulate interneuronal activity and hence create synchronized ensembles of projection neurons. Consequently, serotonergic input to the BLA may act at a subnuclear level to modulate behavioral responses in which the amygdala has been implicated.
Spontaneous synaptic activity
In the adult rat BLA, spontaneous synaptic activity in projection neurons was characterized by fast EPSCs and IPSCs that were interspersed by long-duration, bicuculline-sensitive, compound IPSCs. In contrast in interneurons, spontaneous synaptic activity primarily was characterized by fast EPSCs that were themselves interspersed with bursts of summated EPSCs. However, in a recent BLA study bySmith and Dudek (1996), no compound IPSCs were observed at the resting membrane potential (about −60 mV) of principal neurons. This discrepancy is most probably due to the age difference in the populations of animals used in these two studies. In this study, fully mature animals were used (mean 41 ± 1 days, n = 74), compared with those 8- to 25-days old in the study of Smith and Dudek. Indeed, spontaneous EPSCs also predominate in BLA projection neurons from immature (12 days) Long-Evans rats (personal observation). Hence strain differences are unlikely to account for the discrepancy, and the compound IPSP/C may not be expressed in BLA neurons until the network has reached maturity. Support for this hypothesis comes from the observation that application of the GABAA antagonist bicuculline onto BLA slices from mature animals induces spontaneous epileptiform burst firing activity in all cells tested (see Gean and Chang 1991; Rainnie et al. 1991b, 1992). In contrast, no bicuculline-induced epileptiform activity was reported in the study of Smith and Dudek. Therefore it would appear that disinhibition is much more debilitating in mature BLA circuits than it is in immature circuits. Whether this reflects an alteration in synaptic connectivity, an alteration in GABAA receptor subunit composition, or an alteration of glutamate receptor subunit composition remains to be determined.
Support for an age-related alteration in the strength of synaptic contacts comes from a comparison of the amplitude and time constants of decay (τ) of the mean EPSCs and IPSCs in mature and immature BLA neurons. At a holding of −60 mV, the mean EPSC and IPSC amplitudes recorded from projection neurons were similar to those reported by Smith and Dudek, (−12 pA and 25 pA, respectively). However, the τ for the spontaneous IPSC was four times longer in mature projection neurons (21 vs. 5.8 mS) and the EPSC τ was almost five times longer (12.5 vs. 2.7 mS) than those measured in immature neurons. Several possibilities may account for these differences: developmental alterations in the gating kinetics of the GABAergic and glutamatergic receptor-channels, developmental alterations in transmitter re-uptake kinetics, and the longer τs may reflect frequency-dependent attenuation of the synaptic current due to an imperfect space clamp at electrotonically more distant locations on the dendritic arbor. However, attenuation is unlikely because the rise time of the current also would be prolonged (see Spruston et al. 1994), and PSCs only were accepted with a rise time of ∼ 0.8 mS, which is similar to the rise time of 1.0 mS for EPSCs reported in immature neurons.
In addition, the mean EPSC recorded from interneurons had a larger amplitude and a shorter τ than that recorded from projection neurons. These results suggest that even within the same nucleus projection neurons and interneurons may express different glutamate receptor subtypes. In agreement with Smith and Dudek, the EPSC recorded from both cell types was abolished by CNQX at a holding potential of −60 mV, indicating that they were mediated by AMPA receptor activation. The AMPA receptor is a heterologous receptor-channel complex that is assembled from a combination of four subunits (GluR1–4). Thus the difference in amplitude and τ may reflect a difference in the AMPA receptor subunit composition. Interestingly, GluR2/3 AMPA receptor subunits are expressed in projection neurons, whereas interneurons express the GluR1 subunit but not the GluR2/3 subunits (McDonald 1994, 1996). Indeed the interneuronal EPSC amplitude and τ reported here are similar to those reported by Mahanty and Sah (1998), who recently have demonstrated that EPSCs, recorded in “putative” BLA interneurons, have no NMDA component and are mediated by an inwardly rectifying calcium-permeable AMPA current that is sensitive to external polyamines.
The existence of a reciprocal interaction between projection neurons and interneurons has been discussed in earlier studies (seeGean and Chang 1992; Rainnie et al. 1991a; Smith and Dudek 1996). However, no direct correlation had been established between spontaneous synaptic events occurring in projection neurons with those occurring in interneurons. This study has demonstrated, for the first time, that the occurrence of spontaneous compound IPSCs in projection neurons has a similar interevent interval to the occurrence of spontaneous bursts of EPSCs in interneurons. It is possible, therefore, that compound IPSPs observed in projection neurons may be driven by the occurrence of bursts of action potentials in local interneurons. The local nature of this connectivity was supported by the observation that depolarizing neurons within the slice, by raising external concentrations of potassium in the ACSF, increased the frequency of spontaneous potentials. In addition, spontaneous bursts were blocked by both TTX and CNQX; this implied that an action potential-dependent, glutamatergic drive was necessary to trigger the network. However, spontaneous action potentials rarely are observed at the resting membrane potential of projection neurons in vitro. Consequently it is unlikely that these neurons drive the network. Moreover a dorsoventral transection of the slice that severed the reciprocal connectivity between the basolateral nucleus and the piriform and entorhinal cortex (Krettek and Price 1978; Ottersen 1982) also failed to abolish spontaneous IPSCs or bursts of EPSCs. It is possible, that the excitatory input comes from afferent inputs from the lateral nucleus (see Pitkänen et al. 1997), projection neurons of which demonstrate burst firing, and oscillating firing patterns, both in vitro and in vivo (Pape et al. 1998;Paré and Gaudreau 1996).
Irrespective of the source of the excitatory input, the membrane time constant (τm = 18 mS) in projection neurons and the frequency of action potential firing in BLA interneurons (200 Hz) suggests that one interneuron could generate the compound IPSP/C observed in projection neurons. Even assuming a single release site and a 50% failure rate of any given action potential in an interneuron to evoke transmitter release (see Allen and Stevens 1994), it still would be possible for multiple IPSCs to summate during a 150-mS burst. If each interneuron makes multiple contacts with a single projection neuron (Buhl et al. 1995; Tamás et al. 1997; Thomson et al. 1996), this would increase further the potential for compound IPSPs. If each interneuron innervates more than one projection neuron, this then would facilitate synchronous activity in a micronetwork, or ensemble, of BLA projection neurons.
Two recent studies have reported a slow inhibitory synaptic potential in projection neurons of the lateral nucleus (Danober and Pape 1998; Lang and Paré 1997a) that was bicuculline insensitive and mediated by activation of an intrinsic calcium-dependent potassium conductance. In contrast, bicuculline blocked both the fast and slow components of the spontaneous compound IPSP/C. In a recent dual-cell recording study, a long-duration (>200 mS) IPSP was observed in projection neurons of the hippocampus after activation of a subtype GABAergic interneuron (Thomson et al. 1996). Interestingly, >20 action potentials at >100 Hz were required before the slow IPSP was observed. The similarity between the spontaneous burst firing rate of interneurons in this study and that required to evoked the slow IPSP in the hippocampus suggests that these slow postsynaptic responses may be mediated by a common mechanism.
Serotonergic modulation of intrinsic neuronal properties
Extracellular studies in the amygdala have reported apparently contradictory observations after serotonin application in vivo. Using a strategy of post hoc morphological identification of recorded neurons, this study has revealed that the primary action of exogenous 5-HT within the BLA was a direct excitation of interneurons. Additional indirect evidence came from the observation that 5-HT also caused an increase in the frequency of spontaneous IPSP/Cs in BLA projection neurons. This excitation probably is mediated by activation of 5-HT2 receptors as both the direct and indirect effects could be mimicked by the 5-HT2-receptor agonists, α-methyl-5-HT. However, unlike 5-HT, α-methyl-5-HT did not cause a depolarization in all interneurons examined. These results support the notion that subpopulations of interneurons within the BLA may express different 5-HT receptor subtypes (see Morales and Bloom 1997) and that no one 5-HT receptor subtype is expressed in all interneurons.
The 5-HT-induced depolarization of BLA interneurons was associated with an increase in membrane input resistance and had a reversal potential of −90 mV, suggesting modulation of a potassium conductance. These data are in agreement with previous observations of a 5-HT2-receptor-mediated increase in interneuronal activity in other areas of the CNS (Alreja 1996; Gellman and Aghajanian 1994; McCormick and Wang 1991;Sheldon and Aghajanian 1991). In the piriform cortex and medial septum, this excitation is mediated by activation of 5-HT2A receptors but the underlying ionic mechanism is unclear (Alreja 1996; Marek and Aghajanian 1995). In contrast, the 5-HT-induced depolarization of thalamic reticular neurons appears to be mediated by a decrease in a “leak” potassium conductance (I KL) but the 5-HT receptor subtype has yet to be identified (McCormick and Wang 1991).
Projection neurons of the BLA appear to be particularly unresponsive to 5-HT application. Indeed the small 5-HT-induced hyperpolarization was found to be independent of the concentration of 5-HT applied, blocked by TTX, and has a reversal potential close to −70 mV. These data suggest an indirect response mediated, in part, by an increase in presynaptic GABA release concomitant to the increased excitability of BLA interneurons. The majority of projection neurons do not respond to the 5-HT1A receptor agonist 8-OH-DPAT, and the 5-HT-induced response was insensitive to the 5-HT1A antagonist, WAY 100635. These data support the observation that the BLA has only a low density of 5-HT1A receptors, the highest levels being found in the central nucleus (Pazos and Palacios 1985;Radja et al. 1991).
Serotonergic modulation of synaptic transmission
High concentrations or prolonged application of 5-HT reduced the frequency and amplitude of spontaneous IPSP/Cs. I have shown previously that, in the BLA, evoked and spontaneous IPSPs are reduced by glutamate receptor antagonists (Rainnie et al. 1991b). The reduction of the IPSP/C amplitude by 5-HT provided indirect evidence that an additional action of 5-HT may be to reduce the excitatory drive within the amygdaloid network. This was substantiated by the observation that 5-HT reduced the amplitude of the stimulus evoked EPSP/Cs and IPSP/Cs in projection neurons and reduced the frequency of spontaneous mEPSCs in interneurons. In other areas of the brain, a reduction in glutamatergic and/or GABAergic transmission has been reported to result from activation of presynaptic 5-HT1Aand 5-HT1B receptors (Bobker and Williams 1989; Johnson et al. 1992; Stanford and Lacey 1996). However, the lack of effect of 8-OH-DPAT and CGS-12066A on either spontaneous or evoked EPSP/Cs and IPSP/Cs would suggest that these receptors do not contribute to the presynaptic modulation of transmitter release in the BLA. In contrast, while this manuscript was in preparation, Cheng et al. (1998), using sharp microelectrode recording in the BLA, have reported that the EPSP reduction is probably mediated by 5-HT1A receptors because it can be mimicked partially by 10 μM 8-OH-DPAT. This concentration is ∼1,000 times higher than its receptor binding affinity (pKi) for the 5-HT1A receptor, and the effect should not be considered specific. Experiments are in progress to determine the pharmacological profile of the 5-HT receptor involved in the reduction of glutamate release.
At first glance there is an apparent paradox in the response of the BLA network to 5-HT application. Why would a circuit decrease excitatory drive onto interneurons at a time when the membrane potential of these same interneurons is shifted toward threshold for action potential generation? The answer may lie in the dose-response relationship for each response. As noted above, there is a delay between the expression of increased interneuron excitability and the decreased glutamatergic drive in response to bath applied 5-HT. This could reflect differing affinities of the 5-HT receptors mediating these two responses. Bath application does not instantaneously raise the extracellular concentration of a drug to the desired concentration. Consequently neurons in the slice will receive a time-dependent concentration gradient. If 5-HT receptors located on the postsynaptic membrane of inhibitory interneurons have a high affinity for 5-HT, they may be activated at low 5-HT concentrations, whereas presynaptic 5-HT receptors located on glutamatergic terminals may have a low affinity for 5-HT and hence only may be activated at higher concentrations. This may explain why an increase in spontaneous IPSP/Cs was observed only at 5-HT concentrations <100 μM; above this concentration, the reduction of glutamatergic transmission may occlude the increased excitability of the interneurons. This mechanism could act as a feedback loop to prevent excess inhibition in the presence of prolonged 5-HT release. Moreover the reduced glutamatergic drive would reduce synaptic “noise” and ensure that the nucleus responds only to inputs above a particular intensity.
An alternative explanation for the time-dependent decrease in spontaneous IPSP frequency in response to prolonged 5-HT application may be an activation and subsequent desensitization of 5-HT3 receptors on BLA interneurons. The 5-HT3receptor, unlike the other serotonin receptors, is a ligand gated ion channel and rapidly desensitizes with prolonged activation (Derkach et al. 1989). Moreover, Morales and Bloom (1997) have reported that a small subpopulation of BLA interneurons express 5-HT3 receptors. Two observations suggest that although 5-HT3 receptors may contribute to the 5-HT-mediated response, the time-dependent response probably involves activation of more that one 5-HT receptor subtype: the response always is observed with 5-HT application and the specific 5-HT3receptor agonist, m-chlorophenylbiguanide, (m-CPBG, 500 nM), evoked a brief membrane depolarization in only one of three interneurons examined (personal observation).
Regulation of glutamatergic and GABAergic transmission in the amygdala has been implicated in aspects of fear, anxiety, memory consolidation, and stimulus-reward associations (Brioni et al. 1989; Cador et al. 1989; Davis et al. 1994; Dickinson-Anson and McGaugh 1997;Ferry and Di Scala 1997; Salinas and McGaugh 1996; Wan and Swerdlow 1996). The results of this study suggest that pharmacological manipulations that modulate serotonergic transmission in the BLA would have possible repercussions in each of these behavioral responses. Indeed, Deakin and Graeff (Graeff et al. 1996) have postulated that an ascending 5-HT pathway from the dorsal raphé, which innervates the amygdala and frontal cortex, facilitates conditioned fear. Moreover local infusion of the 5-HT3 antagonist, BRL 46470A, into the amygdala produces an anxiogenic effect (Gargiulo et al. 1996), and acute treatment with the serotonin uptake inhibitor, citalopram, reduces the acquisition and expression of conditioned fear (Inoue et al. 1996). The data reported here would suggest that intra-amygdaloid infusion of specific 5-HT2receptor agonists and/or antagonists also would affect these behaviors. In contrast, infusion of 5-HT1A receptor agonists would be expected to have minimal effects on these behaviors. It is interesting, therefore, that a recent study has reported that local activation of 5-HT1A receptors in the BLA may produce anxiogenic effects (Gonzalez et al. 1996). It is possible that the anxiogenic response observed in this study results from activation 5-HT1A in the adjacent central nucleus due to drug spill-over from the infusion site.
Further experiments are needed to fully elucidate the pharmacological and ionic profiles of the serotonin responses reported here. Moreover, the BLA also has a relatively high density of 5-HT4 and 5-HT6 receptors, and additional experiments are in progress to determine the effects of activation of these receptors on the network properties of the BLA.
The author is indebted to Dr. A. J. McDonald of the University of South Carolina for agreeing to independently determine the morphological characteristics of a representative sample of filled neurons. The author also thanks K. E. McLaughlin and M. J. Mudrick for skillful assistance in preparing this manuscript and Drs. R. W. McCarley, M. Patil, and R. Bergeron for helpful discussion.
This work was supported by National Institute of Mental Health Grant R29 MH-57016-01 and National Alliance for Research on Schizophrenia and Depression Grant NAR97RAIN704-01 to D. G. Rainnie and by the Brockton Veterans Affairs Schizophrenia Center.
Address for reprint requests: Harvard Medical School and Brockton VAMC, Dept. of Psychiatry, Neuroscience Laboratory 151C, 940 Belmont St., Brockton, MA 02301.
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 1999 The American Physiological Society