Presynaptic inhibition by 5-HT_1B receptors of glutamatergic synaptic inputs onto serotonergic caudal raphe neurones in rat

Abstract

1. Autonomous, pacemaker-like activity of serotonergic raphe neurones and its autoregulation by somatodendritic 5-HT_1A receptors are well described, but little is known of synaptic inputs onto raphe neurones or their modulation. Therefore, we recorded unitary excitatory postsynaptic currents (EPSCs) in caudal raphe neurones (raphe obscurus and pallidus) following local electrical stimulation in a neonatal rat brainstem slice preparation; most neurones (79 %; n = 72/91) recovered following post hoc immunohistochemistry were tryptophan hydroxylase-immunoreactive, indicating that they were serotonergic.

2. Evoked EPSCs occurred at relatively constant latency with variable amplitude and apparent ‘failures’ at fixed suprathreshold stimulus intensity. At −60 mV, EPSCs were wholly due to CNQX-sensitive, non-NMDA glutamate receptors; at depolarized potentials, a small AP-5-sensitive NMDA component was often observed.

3. EPSCs were potently and reversibly inhibited by 5-HT with an EC₅₀ of 0.1 μM. This effect was mimicked by 5-HT_1B agonists (CP-93,129 and anpirtoline), but not by a 5-HT_1A agonist (8-OH-DPAT), indicating that 5-HT_1B receptors mediate the inhibition of EPSCs.

4. Multiple lines of evidence indicate that inhibition of EPSCs by 5-HT was mediated presynaptically. First, currents evoked by exogenous glutamate application were unaffected by 5-HT and/or 5-HT_1B agonists. In addition, the frequency of spontaneous glutamatergic miniature EPSCs was diminished by CP-93,129 and paired-pulse facilitation of EPSCs was enhanced by 5-HT. Finally, the 5-HT_1B receptor agonists that blocked synaptic transmission had no effect on resting membrane properties of raphe neurones.

5. These data indicate that serotonergic caudal raphe neurones receive glutamatergic inputs that are inhibited by presynaptic 5-HT_1B receptors; inhibition of excitatory synapses onto raphe cells may represent a novel mechanism for autoregulation of serotonergic neuronal activity by 5-HT.

Serotonergic raphe neurones are located in the mid-line region of the brainstem and project widely throughout the neuraxis, where they influence numerous functions. Raphe neurones display a regular, tonic firing pattern in vivo that is highly state dependent, characterized by elevated levels of activity during active waking states that decrease progressively as the animal moves through slow wave sleep stages to reach a nadir during rapid eye movement (REM) sleep (reviewed in Jacobs & Azmitia, 1992). In addition, the activity of neurones in the caudal raphe nuclei, which have a prominent projection to motor nuclei (Skagerberg & Bjorklund, 1985), is also highly correlated with certain rhythmic motor outputs (e.g. breathing, locomotion; Jacobs & Azmitia, 1992; Veasey, Fornal, Metzler & Jacobs, 1995).

The mechanisms that contribute to these characteristic patterns of activity in serotonergic raphe neurones are incompletely understood. Much evidence suggests that the regular, ‘pacemaker-like’ firing pattern is generated autonomously (although a noradrenergic input may also contribute) and that it reflects the interplay of membrane properties intrinsic to the neurones (reviewed in Aghajanian, Sprouse & Rasmussen, 1987; Jacobs & Azmitia, 1992). Superimposed on the basic tonic firing pattern is an autoregulatory mechanism in which activity-dependent release of 5-HT within the raphe nuclei acts to inhibit neuronal activity via 5-HT_1A somatodendritic autoreceptors expressed on the serotonergic raphe neurones themselves (Wang & Aghajanian, 1978; Williams, Colmers & Pan, 1988; Pan, Colmers & Williams, 1989; Pan, Wessendorf & Williams, 1993; Fornal, Metzler, Gallegos, Veasey, McCreary & Jacobs, 1996; Bayliss, Li & Talley, 1997a; reviewed in Jacobs & Azmitia, 1992). Although these two mechanisms - intrinsic ‘pacemaker-like’ activity and feedback autoinhibition by 5-HT - are well described, it is unlikely that they can account for all aspects of raphe neuronal behaviour. In this regard, there is currently only sparse information regarding the nature and modulation of fast synaptic inputs onto raphe neurones in general (Pan et al. 1989; Pan & Williams, 1989), and none specifically regarding caudal raphe neurones. This may be particularly relevant given recent demonstrations that the behaviour of caudal raphe neurones does not simply reflect sleep-wake state, but can be highly correlated with specific motor activities (Veasey et al. 1995).

In order to examine directly the synaptic inputs onto caudal raphe neurones, we used local electrical stimulation to evoke whole-cell synaptic currents from neurones in raphe obscurus and raphe pallidus using a neonatal rat brainstem slice preparation. We also employed an immunohistochemical technique that allowed us to determine if the recorded neurones were serotonergic. In short, we showed that serotonergic neurones received prominent glutamatergic inputs that were inhibited by 5-HT_1B receptor activation. Inhibition of excitatory inputs onto raphe neurones by presynaptic 5-HT_1B receptors located on glutamatergic terminals provides a substrate for autoinhibitory control of raphe neuronal activity, which could complement previously described autoregulatory mechanisms. Some of these results have been presented in preliminary form (Bayliss & Li, 1997; Li & Bayliss, 1998).

METHODS

Brain slice preparation

The procedures for preparing brain slices were as described previously (Bayliss et al. 1997a). In brief, neonatal Sprague-Dawley rats (3-12 days postnatal) were anaesthetized by hypothermia and rapidly decapitated. The brainstem was quickly removed and immersed in an ice-cold Ringer solution (see below for composition of solutions) that was bubbled with 95 % O₂-5 % CO₂. Transverse slices (150-200 μm) of the medulla oblongata were cut using a vibrating microslicer (DSK-1500E, Dosaka Co., Japan) and immediately transferred to a holding chamber filled with a Ringer solution bubbled with 95 % O₂-5 % CO₂. After incubating at 37°C for 1 h, slices were maintained at room temperature (22-24°C) in Ringer solution.

Solutions

The solution used for preparing slices contained (mM): 130 NaCl, 26 NaHCO₃, 1.25 NaHPO₄, 3 KCl, 1 CaCl₂, 5 MgCl₂ and 10 glucose. For incubation of slices, the solution was the same except that CaCl₂ and MgCl₂ were both at 2 mM, and lactic acid was added to improve slice viability (4 mM; Takahashi, 1992). The external solution used for recording evoked synaptic currents contained (mM): 140 NaCl, 3 KCl, 2 CaCl₂, 1 MgCl₂, 10 N-(2-hydroxyethyl-piperazine)-N'-(2-ethanesulphonic acid) (Hepes), and 10 glucose (pH = 7.3 using NaOH). In most experiments, bicuculline (10 μM) and strychnine (10 μM) were added to block GABA_A and glycine receptor-mediated inhibitory synaptic currents. Miniature synaptic currents (mEPSCs) were recorded with TTX (1 μM) in the external solution to block action potential-evoked transmitter release. The internal solution for recording EPSCs contained (mM): 110 CsCH₃O₃S, 4 NaCl, 10 CsCl, 4 MgCl₂, 0.5 CaCl₂, 10 Hepes Cs, 10 EGTA Cs, 3 ATP-Mg, 0.3 GTP-Tris (pH = 7.3 using CsOH). In some experiments, QX-314 (N-(2,6-dimethylphenylcarbamoylmethyl)-triethylammonium bromide; 5 mM) was added to the internal solution to block voltage-activated Na⁺ channels in the postsynaptic neurone. To record inwardly rectifying potassium currents, the external solution was the same as above; the internal solution contained (mM): 17.5 KCl, 122.5 potassium gluconate, 9 NaCl, 1 MgCl₂, 10 Hepes, 0.2 EGTA, 3 ATP-Mg and 0.3 GTP-Tris (pH = 7.2 using KOH).

Drugs

All drugs were prepared as stock solutions and stored at -20°C; they were thawed and diluted to final working concentrations on the experiment day. Monosodium L-glutamate (1 mM) and 5-HT (0.001-5 μM) were from Sigma. AP-5 ((±)-2-amino-5-phosphonopentanoic acid; 30-100 μM), CNQX (6-cyano-7-nitroquinoxaline-2,3-dione; 10 μM), 8-OH-DPAT (R-(+)-8-hydroxydipropylaminotetraline hydrobromide; 1 μM), bicuculline methchloride (10 μM), strychnine hydrochloride (30 μM), anpirtoline (2-chloro-6-(4-piperidinylthio)-pyridine hydrochloride; 1 μM) and QX-314 (100 μM) were from Research Biochemicals International. Tetrodotoxin (TTX; 1 μM) was from Calbiochem. CP-93,129 (3-(1,2,5,6-tetrahydropyrid-4-yl)pyrrolo[3,2-b]pyrid-5-one; 1 μM) was a gift from Pfizer Central Research (Groton, CT, USA). Lucifer Yellow (0.02 %) was from Molecular Probes.

Identification, recording and orthodromic stimulation of caudal raphe neurones

All experiments were performed at room temperature. Slices were submerged in a chamber mounted on a fixed-stage microscope (Axioskop FS; Zeiss) equipped with Nomarski optics and a × 40 water-immersion objective. The chamber was continuously perfused with external solution. The criteria for visually identifying caudal raphe neurones in the slice were the same as in our previous studies (Bayliss et al. 1997a; Bayliss, Li & Talley, 1997b); we targeted large neurones in raphe obscurus and pallidus based on their location (on the mid-line in the medulla oblongata) and their morphology (shape, size and dendritic orientation; Steinbusch & Nieuwenhuys, 1983).

Recording and stimulating pipettes were pulled from borosilicate glass capillaries (Clark Electromedical) on a two-stage puller (P97; Sutter). Recording electrodes had a DC resistance of 4-6 MΩ and were coated with Sylgard 184 (Dow Corning) before use. Whole-cell voltage clamp recordings were made using an Axopatch 200A patch-clamp amplifier and pCLAMP software (Axon Instruments). Series resistance was usually < 20 MΩ and compensated by ∼70 %. Records were not corrected for a small junction potential (< 2 mV). Currents were filtered at 2 kHz and digitized at 5-10 kHz. Stimulating pipettes had a DC resistance of ∼5 MΩ and were filled with standard external solution that also contained Lucifer Yellow (0.02 %). Stimulus pulses were generated using a Grass stimulator (S48). A raphe neurone was identified 20-100 μm away from the recorded neurone and the tip of the stimulating pipette was placed onto its surface. Although it was possible occasionally to evoke synaptic currents by stimulating in the nearby neuropil (i.e. not obviously near any cell body), it was our experience that EPSCs were evoked with a higher probability using lower stimulus intensities when the stimulating pipette was placed directly onto a neuronal soma in the raphe nuclei. Although this approach provided an effective means to evoke EPSCs, it is not at all certain that the neurone targeted for stimulation actually represents the source of the synaptic input. Stimulus pulses 0.1 ms in duration were applied at a frequency of 0.7 Hz. Stimulus amplitude was gradually increased until a stimulus-evoked synaptic current was recorded. Once a threshold for evoking synaptic currents was reached, the stimulation strength was set at 1.5-2 times the threshold voltage. The threshold for evoking EPSCs ranged from 5 to 30 V (corresponding to ∼1-6 μA); these levels of current injection were similar to other reports in which unitary synaptic currents were studied (Stern, Edwards & Sakmann, 1992; Takahashi, 1992; Jonas, Major & Sakmann, 1993). In some experiments, glutamate currents were evoked by pressure application (Picospritzer; General Valve, Fairfield, NJ, USA) of 1 mM L-glutamate from a glass pipette placed 5-10 μm from the recorded neurones.

Data acquisition and analysis

Data were acquired and analysed using pCLAMP 6 software. Latency, peak amplitude, 10-90 % rise time and decay time constant of evoked synaptic currents were determined. Latency was defined as the time between the stimulus artifact and the beginning of the EPSC. Decay time constants of the EPSCs were obtained by fitting a single exponential equation of the form: y = a[1 - exp(-t /τ)] + c, where τ is the decay time constant from peak (a + c) to baseline (c). The reversal potential of evoked EPSCs was calculated for each neurone using linear regression. Transmission ‘failures’ were defined as events with a peak amplitude < 2 pA. Unless otherwise stated, sample traces of synaptic currents represent averages from five to thirty consecutive events. Miniature EPSCs (mEPSCs) were analysed with MEPP software (courtesy of Drs W. E. Satterthwaite and A. J. Berger, University of Washington, Seattle, WA, USA). Events were accepted for analysis if they had amplitude > 2 pA, a monotonic rising phase and decayed to baseline in an approximately exponential fashion in < 10 ms.

Data are expressed as means ± s.e.m. Statistical analyses were performed using Student's t tests or one-way analysis of variance (ANOVA) followed by t tests with Bonferroni correction. Cumulative probability distributions were compared statistically using a Kolmogorov-Smirnov test. In all cases, P < 0.05 was accepted as statistically significant.

Immunohistochemical staining for tryptophan hydroxylase

Lucifer Yellow (0.02 %) was included in the patch pipette to label the recorded neurone and a post hoc immunohistochemical technique based on tryptophan hydroxylase (TPH) immunoreactivity (IR) was employed to determine if the labelled neurones were serotonergic. After each recording, the location of the labelled neurone with respect to relevant landmarks (e.g. mid-line, inferior olive) was noted and a freehand sketch was made of its shape and orientation. The slices were immersed in 4 % paraformaldehyde solution for 2-4 days at 4°C before they were processed for immunohistochemical detection of TPH. The procedures for TPH immunohistochemistry were similar to those described previously (Bayliss et al. 1997a). In brief, following rinses in 0.1 M phosphate buffer (PB, pH 7.4) and 1 % sodium borohydride in PB, slices were incubated (1-3 days at 4°C) in mouse monoclonal antibody to TPH (1: 500, Sigma) in Tris-saline solution (50 mM Tris, 150 mM NaCl, pH 7.4) containing 5 % normal goat serum and 0.5 % Triton X-100 (TS-NGS-TX). Following rinses in TS-NGS-TX, sections were incubated for 1 h at room temperature with biotinylated rabbit anti-mouse IgG₃ antisera (1:100; Zymed, San Francisco, CA, USA) in TS-NGS-TX and then incubated for 1 h at room temperature in TS with avidin-Texas Red (1: 200; Molecular Probes) or avidin-Cy3 conjugates (1:1000; Jackson Immunoresearch, West Grove, PA, USA). After rinses in Tris, sections were mounted, dried and coverslipped in Krystalon (EM Science, Fort Washington, PA, USA). Following immunohistochemical processing, recorded neurones were identified and photographed with a Zeiss fluorescence microscope.

RESULTS

Excitatory synaptic inputs onto serotonergic neurones in the caudal raphe nuclei

We recorded synaptic currents evoked by local extracellular stimulation from visualized caudal raphe neurones in a thin brainstem slice preparation in order to characterize synaptic inputs onto serotonergic caudal raphe neurones. To determine if recorded neurones were serotonergic, we filled the cells with Lucifer Yellow (Fig. 1A) and processed the slices for TPH immunoreactivity after the experiments (Fig. 1B). The recorded neurone shown in Fig. 1 was serotonergic (i.e. TPH immunoreactive; asterisk) and an example of synaptic current evoked in that cell is shown in Fig. 1C. The synaptic current was inward at a holding potential of −60 mV, with a latency of ∼2 ms and a peak amplitude of ∼-45 pA. Under current clamp conditions, action potentials could be triggered in the postsynaptic cell by the corresponding evoked synaptic potentials (data not shown; see Li & Bayliss, 1998).

Figure 1. Excitatory postsynaptic currents in a serotonergic caudal raphe neurone.

A, neurones were recorded in the caudal raphe nuclei and labelled with Lucifer Yellow (asterisk). B, slices were stained immunohistochemically using an antibody to tryptophan hydroxylase (TPH), the 5-HT synthesizing enzyme. Fluorescence photomicrographs reveal that the recorded neurone was TPH immunoreactive (i.e. serotonergic); other serotonergic neurones are also evident (arrows in B). C, synaptic currents evoked at −60 mV in the cell of A and B. The arrowhead points to the stimulation artifact. Traces were leak-subtracted and averaged. Scale bar in A represents 20 μm and applies also to B.

In total, EPSCs were recorded from 171 caudal raphe neurones. A high percentage of the neurones sampled were serotonergic; of 91 caudal raphe neurones recovered following histology, 72 were TPH immunoreactive (TPH-IR; 79 %). In a sample of positively identified serotonergic neurones in which they were analysed in detail (n = 9), excitatory postsynaptic currents (EPSCs) were evoked at a latency of 2.3 ± 0.1 ms with an amplitude of 61 ± 10 pA; rise times (10-90 %) and decay time constants were 1.1 ± 0.3 and 3.2 ± 0.3 ms, respectively; EPSC properties of serotonergic neurones were essentially identical to those of either unidentified neurones (n = 14) or cells that were not TPH-IR (n = 9). All EPSCs were blocked by TTX (1 μM, n = 5) or a nominally Ca²⁺-free external solution (0 mM Ca²⁺-5 mM Mg²⁺, n = 4), indicating that transmission was action potential and Ca²⁺ dependent.

Evoked EPSCs are glutamatergic

Under our recording conditions, inward PSCs could be mediated by Cl⁻ conductances. However, we found that synaptic currents were unaffected by the GABA_A antagonist, bicuculline (10 μM), or the glycine receptor antagonist, strychnine (30 μM), suggesting that GABA- and glycine-gated Cl⁻ channels did not contribute to evoked synaptic currents (data not shown); to avoid any possible contamination from those channels in our experiments, we routinely added bicuculline and strychnine to the external solution. By contrast, evoked synaptic currents recorded at −60 mV were completely blocked or significantly attenuated by either 1 mM kynurenate (Fig. 2A; n = 11), a wide-spectrum glutamatergic receptor antagonist, or 30 μM CNQX (Fig. 2C; n = 26), a non-NMDA receptor antagonist. Thus, synaptic currents evoked in caudal raphe neurones by local electrical stimulation were glutamatergic and, at least near resting membrane potential, were mediated exclusively by postsynaptic α-amino-3-hydroxy-5-methylisoxazole-4-propionate/kainate (AMPA/KA) receptors.

Figure 2. Synaptic currents are mediated by glutamate receptors.

A, synaptic currents were evoked at −60 mV in a raphe neurone by local stimulation in the absence and presence of kynurenate (1 mM), a broad-spectrum glutamatergic receptor antagonist. The peak amplitude of individual synaptic events was plotted as a function of time. Kynurenate reversibly inhibited synaptic currents. Inset: sample records of synaptic currents evoked at −60 mV from the control period and during kynurenate application (averages of 20 events). B, the peak amplitudes of evoked EPSCs were recorded at different holding potentials in a group of neurones (n = 14 cells); the grouped data were averaged and plotted as a function of holding potential. The averaged peak I-V relationship was linear and reversed at ≈0 mV. Inset: sample records of EPSCs evoked at holding potentials ranging from −60 to +40 mV from a single cell are superimposed (traces are leak-subtracted averages of 10 events). Note the slow component of the EPSC at +40 mV (arrow). C, EPSCs were recorded at −60 and +60 mV holding potential in control and in the presence of AP-5 (an NMDA antagonist) and CNQX (an AMPA/KA antagonist). AP-5 (30 μM) was without effect on the peak amplitude of EPSCs at either potential, but it reduced the slow component of EPSCs evoked at +60 mV. Subsequent addition of CNQX (10 μM) into the bath solution abolished the remaining EPSCs. Traces at −60 and +60 mV are offset for clarity; they are leak-subtracted averages of 20 events.

To determine if NMDA receptors contribute to evoked EPSCs at more depolarized potentials, we recorded EPSCs while the postsynaptic neurone was held at different membrane potentials. The EPSCs became proportionately smaller as the membrane potential was depolarized from −60 mV, eventually reversing between -20 and 20 mV (Fig. 2B, inset) and the averaged I-V relationship of the peak EPSC revealed an essentially linear profile with a reversal potential at ∼0 mV (Fig. 2B; n = 14). Although the I-V relationship of the peak EPSC amplitude was linear, a small slow component appeared in the EPSC decay at depolarized potentials, intimating that NMDA receptors might contribute to the EPSC at those potentials. Therefore, we tested if the slow and fast components of EPSCs were sensitive to NMDA and non-NMDA glutamate receptor antagonists. As shown in Fig. 2C, the EPSC evoked at +60 mV had a clear slow component that was not evident at −60 mV. During application of 30 μM AP-5, a selective NMDA receptor antagonist, the slow component was diminished but the peak currents at −60 and +60 mV were not affected. Subsequent addition of CNQX completely blocked the remaining fast component of EPSCs. In a group of seventeen raphe cells (including 6 that were TPH-IR), ten had a slow EPSC component at +60 mV which was sensitive to AP-5 (30-100 μM); the remaining seven neurones had no clear AP-5-sensitive component. The variability in sensitivity to AP-5 was unrelated to the TPH-IR of the recorded neurone. These experiments suggest that near resting membrane potential, evoked EPSCs in caudal raphe neurones were mediated exclusively by AMPA/KA receptors whereas at more depolarized potentials, NMDA receptors may contribute slightly to EPSCs.

Evoked EPSCs are unitary

Several lines of evidence suggest that synaptic currents resulted from stimulation of a single presynaptic element (i.e. they were unitary events). First, EPSCs displayed a characteristic all-or-none increase in mean amplitude as stimulus intensity was increased (Stern et al. 1992; Takahashi, 1992; Jonas et al. 1993; Umemiya & Berger, 1995). This is shown in Fig. 3A, in which the averaged EPSC amplitude (including failures) is plotted as a function of stimulus intensity. In this neurone, no response could be detected when the stimulus strength was below 16 V. The threshold for evoking EPSCs appeared to be at a stimulus intensity of ∼18 V and the mean amplitude of the EPSCs remained constant at all levels of stimulus strength above 20 V. The intermediate size of the averaged EPSC at threshold is expected, as it reflects a greater number of failures at that point (i.e. sum of stimulus and transmission failures); when failures were excluded, the averaged EPSC amplitude was not different at threshold and suprathreshold stimulus intensities. We also observed that slight displacements of the stimulating electrode (2-5 μm) resulted in either disappearance of the EPSCs or an increase in stimulation threshold for evoking EPSCs.

Figure 3. Kinetic properties of EPSCs evoked in caudal raphe neurones.

A, synaptic currents were evoked in a raphe cell using different stimulus intensities and the stimulus-response curve plotted. Each data point represents the averaged peak amplitude of 20-30 events (including failures). A threshold voltage for evoking EPSCs was clearly evident at 18 V. Inset: sample records of synaptic currents evoked at the indicated stimulus intensities (traces are averages of 20-30 events, including failures). B, frequency distribution histogram of EPSC latencies shows that EPSCs were evoked at short and relatively constant latency. Inset: five consecutive EPSC traces demonstrate the short, but slightly varying latency as well as the fast rise and decay time course of EPSCs. The time of stimulation (vertical dashed line) and the onset of the EPSCs (arrows) are indicated; individual traces were leak subtracted and offset for clarity. C, a frequency distribution histogram shows that EPSC amplitudes (excluding failures) evoked in a raphe neurone at constant suprathreshold stimulus intensity varied in amplitude. Inset: sample records of three successive, leak-subtracted EPSCs and superimposed single exponential fit to the EPSC decay (thick dashed line). D, plot of rise time (10-90 %) (upper graph) and decay-time constant (τ_decay; lower graph) as a function of peak EPSC amplitude. The rise time and decay time constants were not correlated with the peak amplitude (same cell as C).

Other properties of the synaptic currents were also consistent with those expected for a unitary synaptic event. Single evoked synaptic currents occurred at relatively consistent latency following stimulation (Fig. 3B, inset); this is reflected in the corresponding frequency distribution of the EPSC latencies (Fig. 3B). The range of mean EPSC latencies we observed (1-3 ms) was generally similar to those of unitary synaptic currents recorded using similar preparations (Stern et al. 1992; Takahashi, 1992). Moreover, the peak amplitude of EPSCs evoked at a constant supra-threshold intensity fluctuated between trials, with many stimuli failing to evoke EPSCs (transmission failures; see control period in Fig. 2A). The frequency distribution of the peak EPSC amplitudes, excluding failures, is plotted for one cell in Fig. 3C; the distribution had a range of 10-150 pA with a median at ∼40 pA. Despite the variable amplitudes of evoked EPSCs recorded in a given cell, their kinetic properties were unrelated to amplitude, as illustrated in the scatterplots of Fig. 3D. Thus, neither the 10-90 % rise time (Fig. 3D, upper graph) nor the decay time constant (Fig. 3D, lower graph) of the EPSCs (derived from a single exponential fit, as shown in Fig. 3C, inset) were correlated with the EPSC amplitude. This suggests that EPSCs did not arise from multiple inputs distributed widely over the somatodendritic membrane of the neurone.

Effect of 5-HT on evoked EPSCs

We demonstrated previously that 5-HT has potent autoinhibitory effects on caudal raphe neurones mediated by 5-HT_1A receptors; it produces membrane hyperpolarization via activation of inwardly rectifying K⁺ channels and modulates the firing behaviour of these neurones by inhibition of Ca²⁺ channels (Bayliss et al. 1997a, b). Here, we tested whether 5-HT also modulates glutamatergic synaptic transmission onto caudal raphe neurones. As shown in Fig. 4A, we found that 5-HT significantly inhibited EPSCs evoked in caudal raphe neurones. In this neurone, the averaged peak amplitude of EPSCs under control conditions was ∼35 pA and was reduced by ∼95 % during application of 1 μM 5-HT. The peak amplitudes of individual EPSCs were plotted as a function of time in the same neurone (Fig. 4B); 5-HT reversibly decreased the amplitude of the EPSCs and increased the frequency of synaptic transmission failures. The inhibitory effect of 5-HT on EPSCs was concentration dependent, as shown in Fig. 4C and D. At 0.1 μM and 1.0 μM, 5-HT inhibited the averaged peak amplitude of EPSCs by ∼31% and ∼73 %, respectively; at 5 μM 5-HT the EPSC was virtually abolished (Fig. 4C). Figure 4D shows the concentration-dependent inhibition of EPSC amplitude by 5-HT. The averaged data were fitted with a logistic equation which predicted an EC₅₀ for 5-HT of 0.1 μM and ∼80 % inhibition at 1 μM.

Figure 4. 5-HT inhibits evoked EPSCs in caudal raphe neurones.

A, averaged traces of EPSCs in control and in the presence of 1 μM 5-HT. B, the peak amplitude of individual EPSCs from the same neurone (in A) was plotted as a function of time. 5-HT reversibly inhibited EPSCs. Note that 5-HT reduced the current amplitude and increased the frequency of apparent transmission failures. C, averaged traces of EPSCs in control and in the presence of increasing concentrations of 5-HT. The inhibition of EPSCs by 5-HT was concentration dependent. D, the averaged inhibition of EPSC amplitude (percentage of control) was plotted as a function of 5-HT concentration. Concentration-response data were fitted with a logistic equation (continuous line) that predicted an EC₅₀ of 0.1 μM. Only neurones tested with at least two concentrations of 5-HT are included; each point represents data from 4-12 neurones.

We tested the effect of 5-HT (0.1-5 μM) on EPSCs in a total of thirty-two raphe neurones. EPSC amplitude was significantly inhibited in twenty-six of those thirty-two neurones (81 %) and unaffected in the remaining six neurones. We also examined whether the effect of 5-HT was correlated with serotonergic phenotype. Of the twenty-one neurones recovered, seventeen were TPH-IR; of those seventeen TPH-IR cells, EPSCs were inhibited by 5-HT in fifteen (88 %). Of the four unidentified cells, 5-HT inhibited the EPSC in three. These data indicate that 5-HT inhibits excitatory synaptic transmission onto neurones in the caudal raphe nuclei, including serotonergic raphe cells.

In addition to its effects on the EPSC, we noticed that 5-HT caused a small inward current in some neurones tested (see also Umemiya & Berger, 1995). The nature of that current was not studied in detail, but it appeared to be unrelated to the inhibitory effect of 5-HT on EPSCs; the averaged inhibition of EPSC was not different between neurones in which 5-HT caused an inward current and those in which 5-HT had no effect on membrane current (79 ± 5 % vs. 76 ± 3 %, P > 0.6).

5-HT_1B receptors mediate the effects of 5-HT on evoked EPSCs

In other neuronal systems, inhibition of synaptic transmission by 5-HT has been attributed to activation of either 5-HT_1A or 5-HT_1B receptors (Bobker & Williams, 1989; Schmitz, Empson & Heinemann, 1995a, b;Umemiya & Berger, 1995; Singer, Bellingham & Berger, 1996). Therefore, we performed pharmacological experiments to determine the receptor subtype(s) that mediates inhibition by 5-HT of EPSCs onto raphe neurones. Specifically, we tested the effects of 8-OH-DPAT, a 5-HT_1A receptor agonist, and two 5-HT_1B receptor agonists, CP-93,129 and anpirtoline (Schlicker et al. 1992; Zifa & Fillion, 1992; Hoyer et al. 1994). As shown in Fig. 5A and B, CP-93,129 (1 μM) reversibly inhibited the EPSC evoked in a caudal raphe neurone and increased the frequency of apparent transmission failures. An example of the effect of both the 5-HT_1A agonist 8-OH-DPAT and the 5-HT_1B agonist, anpirtoline on EPSCs recorded in a different neurone is shown in Fig. 5C and D. Note that whereas 8-OH-DPAT (1 μM) had very little effect on the mean EPSC amplitude (∼5 % inhibition), anpirtoline (1 μM) nearly completely abolished the EPSC and increased the frequency of apparent transmission failures. Averaged data from experiments using these agonists are presented in Fig. 5E. The inhibitory effect on evoked EPSCs of CP-93,129 (73 ± 8.3 %, n = 7) and anpirtoline (76 ± 9.8 %, n = 6) was significantly greater than that of 8-OH-DPAT (16 ± 11 %, n = 6), indicating that 5-HT_1B, but not 5-HT_1A, receptors mediate the inhibition of evoked EPSCs by 5-HT.

Figure 5. 5-HT_1B receptors mediate inhibition of evoked EPSCs.

A, averaged EPSC traces in control and in the presence of 1 μM CP-93,129, a 5-HT_1B receptor agonist. B, the peak amplitudes of individual EPSCs recorded in the same neurone (as A) were plotted as a function of time. C, averaged EPSC traces in control and in the presence of anpirtoline, a partially selective 5-HT_1B receptor agonist, and 8-OH-DPAT, a 5-HT_1A receptor agonist (both at 1 μM). D, the peak amplitudes of individual EPSCs recorded in the same neurone (as C) were plotted as a function of time. Note that whereas both CP-93,129 and anpirtoline increased the frequency of the transmission failures and decreased the peak amplitudes of EPSCs, 8-OH-DPAT had little effect. E, averaged data showing inhibition of EPSC amplitude (percentage of control) by CP-93,129, anpirtoline and 8-OH-DPAT (each at 1 μM; n = 7, 5 and 6, respectively). * Significantly different from CP-93,129 and anpirtoline (P < 0.05).

Effects of 5-HT and 5-HT_1B agonists are mediated presynaptically

The data presented to this point indicate that 5-HT inhibition of glutamatergic synaptic inputs onto serotonergic caudal raphe neurones is mediated by 5-HT_1B receptors. To determine if the effect of 5-HT is mediated by modulating glutamate release from presynaptic neurones (i.e. presynaptic inhibition) or by modulating postsynaptic responses to glutamate-evoked current, we performed a number of experiments. First, we tested if 5-HT (or 5-HT_1B agonists) altered the postsynaptic response to glutamate, as shown in Fig. 6. We applied L-glutamate (1 mM) to caudal raphe neurones by pressure-ejection from a pipette located near the recorded neurones under conditions identical to those used for recording evoked EPSCs. Glutamate evoked an inward current that showed little desensitization following repeated applications. As shown in Fig. 6A, the glutamate current evoked under control conditions (∼0.90 nA) was not substantially reduced during bath application of 1 μM 5-HT, CP-93,129 or anpirtoline (0.88, 0.90 and 0.87 nA, respectively). Averaged data from similar experiments presented in Fig. 6B revealed that neither 5-HT nor the two 5-HT_1B agonists had any effect on glutamate-evoked currents. Thus, these compounds which inhibited glutamatergic EPSCs were without any effect on postsynaptic glutamate currents.

Figure 6. 5-HT and 5-HT_1B agonists do not alter postsynaptic glutamate sensitivity.

A, inward currents were evoked by pressure-ejection of 1 mM l-glutamate in control and in the presence of 5-HT, CP-93,129 or anpirtoline (each at 1 μM). Each trace shown is the average of three trials. Glutamate-evoked currents were not affected by any of the three agonists. B, averaged data showing that currents evoked by exogenously applied glutamate were unaffected by 5-HT, CP-93,129 or anpirtoline (each at 1 μM; n = 8, 7 and 8, respectively).

We next tested the effect of CP-93,129 on the frequency and amplitude of spontaneous miniature EPSCs (mEPSCs) recorded in caudal raphe neurones. Agonist-induced changes in the frequency of miniature synaptic events, independent of changes in amplitude, provide strong evidence for a presynaptic effect (Katz, 1969; Thompson, Capogna & Scanziani, 1993). Glutamatergic mEPSCs were recorded in the presence of TTX (1 μM) to block action potential-dependent synaptic currents, and with bicuculline (10 μM) and strychnine (10 μM) in the perfusate to block fast inhibitory synaptic currents. All mEPSCs recorded under these conditions were blocked by CNQX. The effects of CP-93,129 on mEPSCs recorded in a caudal raphe neurone are illustrated in Fig. 7. A number of spontaneous synaptic events are evident in the sample traces from the control period (Fig. 7A, left). In the presence of CP-93,129, the frequency of those events was dramatically diminished (Fig. 7A, middle), and spontaneous synaptic currents were abolished in the presence of CNQX indicating that they indeed represented glutamatergic mEPSCs (Fig. 7A, right). Quantitative analysis of the spontaneous mEPSCs from this cell are presented in Fig. 7B and C, which plots the cumulative probability distributions (and frequency distributions; insets) for the effects of CP-93,129 on mEPSC inter-event intervals (Fig. 7B) and amplitude (Fig. 7C). As is clear from these plots, CP-93,129 caused a shift in the inter-event interval distribution towards longer intervals (i.e. lower frequencies; P < 0.01 by Kolmogorov-Smirnov test), with no significant change in the amplitude distribution (P > 0.2). The mean frequency of mEPSCs decreased from 1.6 ± 0.4 Hz to 0.64 ± 0.02 Hz in the presence of CP-93,129. By contrast, CP-93,129 had no effect on the mean peak amplitude of the mEPSCs (control, -10.5 ± 0.2 pA; CP-93,129, -11.4 ± 0.2 pA). CP-93,129 had no effect on the rise time or decay kinetics of mEPSCs, as shown in the traces of digitally averaged mEPSCs in Fig. 7D, suggesting that the compound did not substantially alter neuronal passive properties or preferentially interfere with synaptic events located at distinct sites. Similar experiments were performed in six caudal raphe cells; CP-93,129 significantly shifted the inter-event interval distribution towards longer intervals in all but one neurone (P < 0.05, by the Kolmogorov-Smirnov test). In the five neurones that responded to CP-93,129 (all TPH-IR), the frequency of mEPSCs decreased from 1.4 ± 0.2 Hz to 0.6 ± 0.1 Hz (P < 0.05), whereas the mean peak amplitude was not affected (control, -13.5 ± 3.5 pA; CP-93,129, -13.8 ± 3.2 pA; P > 0.8). Thus, CP-93,129 decreased the frequency of mEPSCs recorded in serotonergic raphe neurones, consistent with a presynaptic inhibitory effect of 5-HT_1B receptors.

Figure 7. CP-93,129 decreases miniature EPSC frequency.

Miniature glutamate EPSCs (mEPSCs) were recorded in the presence of TTX (1 μM), strychnine (30 μM) and bicuculline (10 μM). A, consecutive current traces recorded in control and during application of CP-93,129 (1 μM) and CNQX (10 μM). B and C, cumulative probability distribution of inter-event intervals (B) and peak amplitudes (C) of mEPSCs in control and during application of CP-93,129 in the same neurone (as A). The inter-event interval was increased (i.e. mEPSC frequency was decreased), but the amplitude distribution was unaffected, by CP-93,129. Insets: frequency distributions of mEPSC inter-event intervals and amplitudes. D, averaged traces of mEPSCs (65 events each) recorded in control and during application of CP-93,129 (same neurone as A). Note the lack of effect of CP-93,129 on mEPSC amplitude and kinetics. Thus, CP-93,129 decreased the frequency of CNQX-sensitive mEPSCs but had no effect on their amplitude or kinetics.

Although inhibition of spontaneous mEPSC frequency by CP-93,129 strongly suggests that the 5-HT_1B receptor agonist acts presynaptically to reduce glutamate release, the origin of the glutamatergic inputs that contribute to those spontaneous mEPSCs is unknown. Therefore, we used another approach to test if the inhibitory effects of 5-HT on evoked EPSCs were also mediated presynaptically. Paired stimuli delivered to the presynaptic element can evoke paired EPSCs in the postsynaptic cell in which, on average, the second EPSC is smaller than the first (paired-pulse depression, PPD) or in which the second EPSC is larger than the first (paired-pulse facilitation, PPF). PPD and PPF are thought to reflect differences in the release probability at a presynaptic site, where PPF is observed at sites with lower release probability (Manabe, Wyllie, Perkel & Nicoll, 1993; Debanne, Guerineau, Gahwiler & Thompson, 1996). Therefore, presynaptic inhibition (i.e. a decrease in presynaptic release probability) is often associated with a shift towards PPF; changes in the sensitivity of the postsynaptic cell, by contrast, would be expected to decrease the first and second EPSCs proportionately and have no effect on the paired-pulse ratio (Manabe et al. 1993; Debanne et al. 1996).

Accordingly, we measured the peak amplitude of synaptic currents evoked by a pair of electrical pulses with an inter-stimulus interval of 30-50 ms in eleven caudal raphe neurones and calculated the amplitude ratio of the second to the first EPSC in each pair. We found that the second EPSC was larger than the first in ten neurones (i.e. PPF) and that the corresponding amplitude ratio averaged 1.5 ± 0.1. In the remaining cell, the ratio of the second to the first EPSC was 0.7 (i.e. PPD). The effect of 5-HT on a cell which displayed PPF is shown in Fig. 8A. During application of 1 μM 5-HT, the amplitude of both the first and second EPSCs in the pair were reduced. However, the inhibition of the first EPSC was greater than that of the second, and therefore the PPF ratio increased from 1.3 to 1.7. The increase in the PPF was even more evident in the sample records when the traces from control and 5-HT were scaled to match the amplitudes of the first EPSC in the pairs (Fig. 8A inset); the second EPSC in these scaled records was clearly larger in the presence of 5-HT. Averaged data from four neurones tested with 5-HT are illustrated in Fig. 8B (includes two TPH-IR neurones); 5-HT significantly increased the PPF ratio from 1.5 ± 0.1 to 1.8 ± 0.1 (P < 0.05). In the additional neurone that displayed PPD under control conditions, 5-HT actually induced a change from PPD to PPF (ratio increased from 0.7 to 1.3). Data from these experiments provide further evidence indicating that 5-HT reduces EPSCs in caudal raphe neurones by acting at a presynaptic site.

Figure 8. 5-HT enhances paired-pulse facilitation of evoked EPSCs.

A, effect of 5-HT (1 μM) on EPSCs in a caudal raphe neurone evoked by a paired-pulse stimulation (60 ms interpulse interval). Although the amplitude of both the first and second EPSC was reduced by 5-HT, the amplitude ratio of the second to the first EPSC in the pair was enhanced by 5-HT (1.3 in control; 1.7 in 5-HT). This is seen more clearly when the EPSCs in control and 5-HT were scaled to match the amplitudes of the first EPSC in each pair (inset). Sample traces represent the digital average of 20 events. B, averaged data show the absolute current amplitudes of the first and second EPSCs in the pair, and the amplitude ratio of the second to the first EPSC, under control conditions and in the presence of 5-HT (1 μM). (n = 5 cells; * P < 0.05.)

We have demonstrated previously that 5-HT, via 5-HT_1A receptors, activates an inwardly rectifying potassium conductance in serotonergic caudal raphe neurones (Bayliss et al. 1997a). It was therefore possible that the inhibitory effects of 5-HT on EPSCs could have resulted from a change in the membrane properties of the postsynaptic neurone that affected our ability to accurately detect the EPSCs. However, this is unlikely because we found that CP-93,129 and anpirtoline, which potently inhibited evoked EPSCs, had no effect on membrane currents in raphe neurones (over a voltage range between -40 and −120 mV), even in cells that showed a robust response to 5-HT (data not shown). Thus, although 5-HT can activate somatic inwardly rectifying K⁺ currents via 5-HT_1A receptors in caudal raphe neurones (see also Bayliss et al. 1997a), the 5-HT_1B agonists which inhibited synaptic transmission had very little effect on those currents in raphe cells.

DISCUSSION

We found that serotonergic neurones of the brainstem raphe nuclei, specifically those in raphe obscurus and pallidus, receive a prominent excitatory glutamatergic input. In addition, we showed that 5-HT_1B receptors modulate those excitatory connections via a presynaptic inhibitory effect. This presynaptic inhibition may provide a previously undescribed mechanism for autoregulation of raphe neurones by 5-HT. The prevailing view of autoregulation by 5-HT of activity-dependent serotonin release holds that two mechanisms primarily account for autoregulatory effects: (1) within the raphe nuclei, release of 5-HT by serotonergic raphe neurones causes feedback autoinhibition of action potential discharge via 5-HT_1A receptors expressed on the somatodendritic membrane of raphe neurones; and (2) at their terminals, 5-HT released by serotonergic raphe neurones activates presynaptic 5-HT_1B receptors located on those same terminals, suppressing further 5-HT release. According to this view, then, activity-dependent release of 5-HT is attenuated by 5-HT_1A receptor-mediated inhibition of raphe neuronal activity at the soma and by 5-HT_1B receptor-mediated inhibition of release at the terminal. Our results suggest an additional autoinhibitory role for 5-HT_1B receptors; release of 5-HT within raphe nuclei, via presynaptic inhibition of glutamatergic synapses, could reduce excitatory inputs onto serotonergic neurones and thereby also contribute to inhibition of raphe neuronal activity. The mechanism we propose is distinctly different from classic autoregulatory mechanisms in that the effects of 5-HT we describe are not necessarily mediated directly on serotonergic neurones. We nevertheless consider it ‘autoregulatory’ in the sense that it provides a means, albeit indirect, by which 5-HT released from serotonergic neurones can inhibit the activity of those serotonergic cells.

Immunohistochemical identification of serotonergic raphe neurones

All neurones studied were located in raphe obscurus and pallidus. However, because many raphe neurones are not serotonergic (Moore, 1981; Steinbusch & Nieuwenhuys, 1983; Jacobs & Azmitia, 1992), we employed a technique for identifying and immunostaining the recorded neurones, as described previously (Bayliss et al. 1997a). Using this approach, we found that the majority of neurones sampled was serotonergic. This was despite the fact that more than half of the neurones in the caudal raphe nuclei of rats may be non-serotonergic (Moore, 1981) and indicates that our sampling procedures were biased towards recording serotonergic cells, as intended. It is difficult to interpret the results from the neurones that were not identified as TPH-IR. It is possible that some non-TPH-IR cells were misidentified as non-serotonergic (a ‘false-negative’ result) as we suggested in our earlier work (Bayliss et al. 1997a); however, it is also likely that some of those non-TPH-IR cells represent the non-serotonergic neurones that are found in the raphe nuclei. In this respect, it is noteworthy that EPSC properties and their modulation by 5-HT were similar in serotonergic and non-TPH-IR cells, suggesting that a common 5-HT_1B mechanism may modulate glutamatergic inputs impinging on both serotonergic and non-serotonergic caudal raphe neurones.

Unitary glutamatergic synaptic inputs onto raphe neurones

We characterized the properties of unitary EPSCs evoked in neonatal serotonergic caudal raphe neurones. As expected, EPSCs were blocked in external solutions containing TTX or nominally zero Ca²⁺, indicating that they required action potential-dependent influx of Ca²⁺ into the presynaptic neurone. EPSCs were primarily mediated by AMPA/KA-type, non-NMDA receptors at all potentials, although a small and variable NMDA component was often observed at more depolarized potentials. A number of lines of evidence support our conclusion that EPSCs evoked in raphe neurones resulted from stimulation of a single presynaptic element (i.e. the EPSCs were unitary). For example, the synaptic response disappeared if the tip of the stimulation pipette was moved only a short distance (2-5 μm). More compelling, however, was the fact that we could demonstrate a clear stimulus threshold for evoking EPSCs beyond which the mean EPSC amplitude did not increase despite increasing stimulus strength. Such all-or-none behaviour is strong evidence for a unitary event (Stern et al. 1992; Takahashi, 1992; Jonas et al. 1993; Umemiya & Berger, 1995). Other EPSC characteristics (short and relatively constant latency, variable amplitude with occasional ‘failures’ at fixed suprathreshold stimulus intensity, activation and decay kinetics that were independent of EPSC amplitude) provide further evidence, albeit indirect, consistent with the suggestion that EPSCs represent unitary monosynaptic events (Stern et al. 1992; Takahashi, 1992; Jonas et al. 1993).

The source of the glutamatergic inputs onto raphe neurones was not determined. In this respect, we consistently obtained glutamatergic EPSPs when the stimulating pipette was placed directly onto the somata of neighbouring raphe cells, suggesting that glutamate release may derive from the serotonergic neurones themselves. This is consistent with the demonstration that serotonergic raphe neurones contain glutamate-like immunoreactivity (Nicholas, Pieribone, Arvidsson & Hökfelt, 1992) and that glutamatergic synaptic connections are formed between serotonergic raphe neurones in culture (Johnson, 1994a, b; Johnson & Yee, 1995). However, unequivocal demonstration that serotonergic raphe neurones release glutamate onto neighbouring raphe cells will require dual intracellular recording of pairs of serotonergic raphe neurones, an approach that to date has not supported this possibility (Y.-W. Li & D. A. Bayliss, unpublished observations). Despite some uncertainty regarding their derivation and unitary nature, it is nevertheless clear that serotonergic raphe neurones are well-endowed with excitatory glutamatergic inputs.

We recorded glutamatergic EPSCs and their modulation by 5-HT in caudal raphe neurones from neonatal rats. We have not tested whether similar mechanisms obtain in caudal raphe neurones of older animals because more extensive myelination of fibres near the mid-line medullary raphe obscures the cells. However, it has been possible to duplicate some of these findings in dorsal raphe slices from adult rat. Specifically, we found that glutamatergic EPSCs could be evoked in adult dorsal raphe neurones by local electrical stimulation and that those evoked EPSCs were inhibited by the 5-HT_1B agonist, CP-93,129 (E. M. Talley & D. A. Bayliss, unpublished observations). A fast glutamatergic EPSP has also been recorded previously in adult dorsal raphe neurones following local electrical stimulation in a slice preparation (Pan et al. 1989). Together, this suggests that mechanisms we have described in neonatal caudal raphe neurones are also present in raphe cells of older animals.

5-HT inhibits glutamatergic synaptic transmission via presynaptic 5-HT_1B receptors

We found that 5-HT inhibits glutamatergic synaptic transmission onto serotonergic caudal raphe neurones via 5-HT_1B receptors. A presynaptic site of action for this 5-HT_1B-mediated effect was supported by multiple lines of evidence. First, 5-HT and the 5-HT_1B agonists, CP-93,129 and anpirtoline, did not alter the responses of caudal raphe neurones to exogenously applied L-glutamate, indicating that 5-HT had no effect on postsynaptic glutamate sensitivity. In addition, CP-93,129 decreased miniature glutamate EPSC frequency and 5-HT enhanced paired-pulse facilitation of evoked glutamate EPSCs recorded in caudal raphe neurones, both indications of a presynaptic site of action (Thompson et al. 1993; Debanne et al. 1996). Finally, a change in the passive properties of the postsynaptic neurone could not account for inhibition of EPSCs, since there was a clear pharmacological dissociation between 5-HT receptors that modulated raphe neuronal properties (i.e. 5-HT_1A; see Bayliss et al. 1997a, b) and those that mediated inhibition of evoked synaptic currents (i.e. 5-HT_1B). The cellular mechanism by which 5-HT inhibits synaptic transmission remains to be determined. In this regard, the decrease in mEPSC frequency by CP-93,129 suggests that 5-HT_1B receptors may inhibit transmitter release independent of changes in voltage-dependent currents or altered calcium entry (Thompson et al. 1993), perhaps via effects on the transmitter release mechanism itself (Thompson et al. 1993; Umemiya & Berger, 1995; Singer et al. 1996).

Functional significance of glutamatergic EPSCs and their inhibition by 5-HT_1B receptors

Autoregulation mediated by presynaptic 5-HT_1B receptors has traditionally been associated with serotonergic terminals: 5-HT released from serotonergic raphe terminals feeds back to activate 5-HT_1B receptors on those same terminals, inhibiting further release of 5-HT (reviewed in Zifa & Fillion, 1992; Hoyer et al. 1994). This view has strong support from work indicating that neurones in raphe nuclei express 5-HT_1B receptor mRNA (Voigt, Laurie, Seeburg & Bach, 1991; Doucet, Pohl, Fattaccini, Adrien, Mestikawy & Hamon, 1995) and that 5-HT_1B receptors inhibit K⁺-evoked [³H]5-HT release (Engel, Gothert, Hoyer, Schlicker & Hillenbrand, 1986). However, in addition to these homologous 5-HT_1B receptors (i.e. receptors localized on terminals of serotonergic neurones) there is substantial evidence that 5-HT_1B receptors are found on heterologous terminals throughout the CNS (i.e. on terminals of non-serotonergic neurones), where they inhibit synaptic transmission (Bobker & Williams, 1989; Tanaka & North, 1993; Mooney, Shi & Rhoades, 1994; Umemiya & Berger, 1995; Singer et al. 1996; reviewed in Zifa & Fillion, 1992; Hoyer et al. 1994). Our results show that 5-HT_1B receptors are expressed on glutamatergic terminals contacting raphe cells, providing a substrate for inhibition by 5-HT of excitatory transmitter release onto serotonergic neurones themselves. Thus, in addition to the conventional view of autoregulation by 5-HT_1B receptors, in which homologous receptors inhibit 5-HT release from serotonergic terminals, 5-HT_1B receptor-mediated autoinhibition may also result from effects of receptors located on glutamatergic excitatory synapses that impinge on serotonergic neurones in the raphe nuclei.

The scenario we propose for autoinhibition of raphe neuronal activity by 5-HT_1B receptor-mediated inhibition of glutamatergic inputs requires that those excitatory synapses influence raphe neuronal activity. Although this would not seem to be an unreasonable presupposition, it is generally believed that raphe neurones display a spontaneous tonic ‘pacemaker-like’ activity that results largely from intrinsic properties of the neurones (Aghajanian et al. 1987; Jacobs & Azmitia, 1992). Indeed, we have found that the spontaneous firing of caudal raphe neurones in vitro was unaffected by CNQX (D. A. Bayliss, unpublished observations), suggesting that glutamatergic inputs are not required for that activity. Moreover, the activity of raphe neurones recorded in vivo does not seem to be strongly influenced by 5-HT_1B receptors (Sprouse & Aghajanian, 1987), suggesting that they have little influence on the regular pacemaker-like firing of raphe neurones. However, it is important to point out that those experiments could not test the response of raphe cells over the full range of behaviours possible and, under certain conditions, glutamatergic excitatory inputs and their modulation by 5-HT_1B receptors may be more important. For example, increased dorsal raphe activity in awake cats invoked by auditory clicks is kynurenate sensitive, although spontaneous activity is not, suggesting a role for glutamatergic inputs in mediating responses of dorsal raphe neurones to phasic sensory stimuli (Levine & Jacobs, 1992). In the caudal raphe, it was recently demonstrated that the activity of serotonergic neurones correlates well with certain motor behaviours (e.g. locomotion and respiration) and under those conditions of heightened activity, the regular tonic discharge can be transformed into a more phasic pattern (Veasey et al. 1995). The synaptic interactions that link caudal raphe neuronal activity with motor behaviours have not been determined. We speculate that glutamatergic inputs, such as those we have described, may provide such a link - and that under such conditions, the 5-HT_1B receptor-mediated autoinhibitory mechanism we have proposed may influence raphe neuronal firing patterns more robustly than has been shown to date.