| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Department of Neuroscience, Albert Einstein College of Medicine, Bronx, New York
Submitted 7 December 2005; accepted in final form 18 January 2006
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSREFERENCES |
|---|
; Connors and Long 2004; Thomson 2000). In fact, studies in many systems, particularly invertebrates, have shown that these synapses might have more specific functions, such as modifying neuronal oscillations (Kepler et al. 1990) and coincidence detection (Edwards et al. 1998). Less clear, however, are the respective roles of electrical and chemical synapses in complex information processing in cases where these two modes of synaptic transmission coexist (i.e., at mixed synapses). The best examples thus far are in invertebrates, where these two components often have opposing excitatory and inhibitory actions (Johnson et al. 1993; Sharp et al. 1992).
In goldfish, the monosynaptic excitatory connection between posterior eighth nerve afferents, called large myelinated club endings (LMCEs), and an identified second-order reticulospinal neuron, the Mauthner (M-) cell, is mediated by mixed electrical and chemical synapses (see Pereda et al. 2004; Zottoli and Faber 2000). The afferents are excited by hair cells in the sacculus that are responsive to acoustic pressure (Furukawa and Ishii 1967). They transmit this auditory information to the M-cell, which in turn triggers the startle response (Zottoli 1977). Because a great deal of information about the basic properties of these synapses has been generated and these afferents are selectively activated by sounds in air, it is an ideal system for defining the different functional roles of their electrical and chemical components during sound reception. However, evidence from paired recordings suggests that >80% of the connections are chemically silent (Lin and Faber 1988a); that is, a presynaptic action potential produces a postsynaptic coupling potential, the sign of electrical transmission, but no chemically mediated excitatory postsynaptic potential (EPSP). It has been shown that glutamatergic transmission is unmasked by modulations of the presynaptic action potential waveform (Lin and Faber 1988b) and might be recruited when a large fraction of the afferent population is coactivated (Pereda et al. 2004). Thus these properties also raise the question of whether, and under what conditions, the chemical synapses contribute to the processing of auditory information.
We report here results of studies that used acoustic stimuli in air comparable in intensity to those that can elicit an M-cell–mediated escape in free-swimming fish under water, specifically abrupt, loud sound pips and longer lasting amplitude- (AM) and frequency-modulated (FM) tone bursts. Applying the stimuli in air involves only the pressure component of sound (Fay and Popper 1999) and therefore selectively activates the input to the M-cell from the posterior VIIIth nerve, including the club endings. Our results demonstrate that the M-cell responses to these stimuli have two components: fast, repetitive EPSPs superimposed on a depolarizing envelope, which we call a slow EPSP. The spatial distribution of the fast EPSPs along the dendrite combined with pharmacological data and results from a multicompartment simulation of the M-cell indicate that they are electrical coupling potentials resulting from impulses in the club endings. In contrast, the slow EPSP is not composed of discrete chemical EPSPs and is generated by synapses on both the soma and proximal dendrite. There is no evidence for a significant glutamatergic component generated at the club ending synapses under the stimulus conditions we used. Power spectral analysis of the responses to the modulated stimuli suggests the electrical synapses encode both stimulus frequency and amplitude in the fast EPSP, whereas the slow EPSP is a smoothly graded function of the envelope of sound amplitude.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSREFERENCES |
|---|
Goldfish (Carassius auratis, length 4–6 in.) were obtained from EECHO Systems (North Kansas City, MO), Hunting Creek Fisheries (Thurmond, MD), and Billy Bland Fisheries (Taylor, AR). They were maintained on a controlled photoperiod of 12 h light/dark at 18°C in recirculating, aerated water (Preuss and Faber 2003). Water was conditioned with NovAqua (0.13 ml/l; Novalek, Hayward, CA), Coppersafe (0.4 ml/l; St. John Laboratories, Harbor City, CA), Instant Ocean (70 mg/l; Aquarium Systems, Mentor, OH), Aquarium Salt (190 mg/l; Jungle Laboratories, Cibolo, TX), and Proper pH (350 mg/l; Aquarium Pharmaceuticals, Chalfont, PA). Fish were initially anesthetized with 100 mg/l 3-aminobenzoic acid ethyl ester (MS-222; Sigma, St. Louis, MO), then further anesthetized in ice water. Fish were then mounted in a recording chamber and respirated with aerated water flowing through the mouth and out over the gills. The water contained 60 mg/l MS-222 and was maintained at 18°C using a Delta Star Chiller (Aqua Logic, San Diego, CA). After exposure of the spinal cord for M-cell antidromic stimulation, and of the midbrain for intracellular recordings (Faber and Korn 1978), the fish was immobilized with d-tubocurarine (1 µg/g body weight; Sigma) injected intramuscularly and the brain was superfused with normal fish saline (in mM: 124.0 NaCl, 5.1 KCl, 2.8 NaH2PO4 · H2O, 0.9 MgSO4, 1.6 CaCl2 · 2H2O, 5.6 glucose, and 20.0 HEPES, pH 7.2). In experiments designed to block specific glutamatergic receptors, saline contained one or more of the following.
-Amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA)/kainate receptor antagonists: 100–150 µM 6-cyano-7-nitroquinoxaline-2,3-dione disodium (CNQX; Sigma, Tocris) and 100 µM 2,3-dioxo-6-nitro-1,2,3,4-tetrahydrobenzo[f]quinoxaline-7-sulfonamide disodium salt (NBQX; Tocris). -methyl-4-carboxyphenylglycine (MCPG; Tocris), Group II (selective): 1 mM (2S)-2-amino-2-[(1S,2S)-2-carboxycycloprop-1-yl]-3-(xanth-9-yl) propanoic acid (LY 341495; Tocris), Group III (selective): 200 µM (RS)--methylserine-O-phosphate (MSOP, Tocris). Electrophysiological recordings
Recording procedures and identification of the M-cell were previously described (Faber and Korn 1978). M-cell intracellular responses to sound as well as to posterior eighth nerve (orthodromic) and spinal cord (antidromic) stimulation were recorded from the soma (50 µm lateral to the axon cap) and distally along the cell's lateral dendrite 
400 µm from the axon cap, using 4- to 7-M
microelectrodes filled with 5 M potassium acetate (KAc). The axon cap is the region surrounding the axon-hillock and initial segment of the axon. All recordings were performed in current clamp (Axoclamp-2B, Axon Instruments, Foster City, CA). In all experiments reported here, resting membrane potential was in the range of –76 to –83 mV in both the soma and dendrite and did not vary by more than 5 mV during a recording session. Similarly, the size of the M-cell's antidromic action potential varied by no more than 5% during control recordings. Because the IPSP equilibrium potential equals the resting membrane potential of the M-cell (Furukawa and Furshpan 1963), inhibitory postsynaptic potentials (IPSPs) were not manifest as frank membrane potential changes unless the cell was significantly depolarized.
Extracellular recording techniques
In a series of experiments, the posterior eighth nerve was exposed surgically and extracellular recordings of evoked multiunit activity were obtained with a 100-µm stainless steel electrode open at its tip (impedance 
40 k) connected to a high-gain AC amplifier (A-M Systems, Carlsborg, WA). These data were collected for comparison with simultaneously collected intracellular responses as a control that the pharmacological agents had minimal effect at the level of the hair cell synapses.
Acoustic stimuli
All stimuli were applied in air (dB re:20 µPa) through three speakers: two tweeters located about 40 cm symmetrically in front and to the left and right of the fish's head, and a subwoofer located about 35 cm above and in front of the fish (CA-3550 "Platinum Series" speakers, Cyber Acoustics, LLC, Vancouver, WA). Because the fish and the speakers were out of the water, only the pressure component of the stimulus was detected by the fish and transmitted to the inner ear from the swim bladder by the Webberian ossicles (Popper and Platt 1993). Custom-designed software for a Macintosh G4 in combination with an AD card (National Instruments, Austin, TX) was used to generate digital sound stimuli, as well as to collect and analyze data.
Two types of acoustic stimuli were used.
Modeling studies
The Neuron simulation package (Yale, New Haven, CT) was used to implement a multicompartment model of the M-cell. The modeled cell possessed 15 segmented compartments, based on earlier morphological studies of the goldfish M-cell (Bodian 1952; Zottoli 1978) and an equivalent circuit model of the cell constructed by Furukawa (1966). The widths of the compartments were tapered to provide realistic continuity between compartments. Because this model was used to compare experimental data with predictions based on the passive membrane properties of the M-cell soma and primary dendrites, the axon was not included in the simulation. Internal resistivity was assumed to be 100 /cm. Previous estimates of the cell's specific membrane capacitance were in the range of 2.5 µF/cm2 with an input resistance of about 100 k (Furukawa 1966). However, more recent measurements suggest values of the latter are 
200 k, with a membrane time constant of <0.5 ms (Preuss and Faber 2003). Therefore we set specific membrane capacitance at 1 µF/cm2, a value consistent with measurements for most biological membranes (Segev and Burke 1998). The uniform passive specific membrane resistance was calculated by iteratively measuring the input resistance and time constant of the M-cell for specific membrane resistance values ranging from 100 to 1,000 /cm2 (Segev and Burke 1998) and by matching the waveform of the antidromic action potential and its experimentally observed passive soma–dendritic decremental conduction. The best fit was obtained with an input resistance of about 250 k and a time constant of about 200 µs in the soma. Synaptic currents were modeled by injecting a dynamic voltage clamp at a single point on the dendrite of the model M-cell, using the experimentally recorded responses as the clamp input. Model responses were calculated using a backward Euler numerical integration method.
Data analysis
All measurements of signal amplitudes and kinetics were made on averaged responses (n 8) using either in-house software, IGOR Pro (Wavemetrics, Lake Oswego, OR) or LabView 7.0 (National Instruments, Austin, TX). To analyze high- and low-frequency signal components separately, averaged recordings were passed through a sixth-order digital Butterworth filter, with a high-frequency cutoff at 60 Hz, or a second-order digital Butterworth filter with a low-frequency cutoff at 60 Hz, respectively. Power spectral densities were calculated with a Fourier transform algorithm, and the signal energy in a joint time–frequency spectrogram (JTFS) was calculated using a short-time Fourier transform with a 40-ms rectangular window shifted in 1-ms increments. To quantify power versus time plots, power was calculated from the JTFS by summating the total power in the 0- to 2,000-Hz band in time increments of 1 ms. The last 25 ms of the sounds and the M-cell responses were poorly represented by the JTFS because of the sliding window algorithm and were not used. The high-pass–filtered response and the unfiltered extracellular recordings from the posterior VIIIth nerve were normalized for DC offset before calculation of root-mean-squared (RMS) values. To further compare the filtered signals with the acoustic stimuli, the amplitude envelopes of the sounds and the high-pass–filtered M-cell responses (HP PSPs) were approximated by finding the successive maxima for each within optimal intervals from 0 to 175 ms. The time course and amplitude of these maxima were fit with sixth-order polynomials. When the power of the sound was plotted instead, it was fit by a sixth-order polynomial at all points. Comparisons between the different response representations and stimulus intensity or power were made after scaling the corresponding fits with an optimal multiplier derived by minimization of the mean square error between the two curves. Linear regression analysis of these relationships was performed by comparing the scaled values of the points only at the time points used for deriving the polynomial fits. All traces shown here represent an average of eight to ten individual responses and values of n refer to the number of fish unless otherwise noted. All statistics represent two-tailed, paired Student's t-test unless otherwise noted. Results are presented as mean ± SE.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSREFERENCES |
|---|
Because sound stimuli were applied in air we were able to isolate their pressure component. Sound pressure is transmitted from the swim bladder to the sacculus, whose hair cells synapse with fibers of the posterior branch of the eighth nerve (VIIIp). The VIIIp fibers, in turn, terminate on the distal lateral dendrite of the M-cell approximately 250–400 µm from the axon cap as large myelinated club endings where they form mixed electrical and chemical synaptic contacts (Bodian 1952; Lin and Faber 1988a; Lin et al. 1983; Tuttle et al. 1986). Estimates based on both morphological and electrophysiological data suggest that about 75 to 100 club endings (average diameter about 10 µm) terminate on the dendrite (about 20 µm in diameter) in this region. As shown in Fig. 1 for a 200-Hz sound "pip," simultaneous extracellular recordings from posterior VIIIth nerve afferents and intracellular recordings from the M-cell suggest that presynaptic spike activity is correlated with high-frequency, relatively synchronous, spikelike transients in the postsynaptic cell, which we call the fast EPSP. These latter events are superimposed on an underlying, depolarizing envelope, the slow EPSP. The experiments described below were designed to identify the origin of sound-evoked EPSPs, focusing first on responses to sound pips.
|
). Thus the fast EPSPs seen in these sound-evoked responses were most likely mediated by current flow across the electrical synapses. The presence of clear peaks in these responses implies a high degree of synchrony in the afferents, as also seen with extracellular nerve recordings (Fig. 1).
|
). This onset latency is within the 2- to 3-ms temporal window during which the M-cell either does or does not trigger an escape (Preuss and Faber 2003). In contrast to P1, the amplitude of P2, measured from trough to peak, appeared to be maximal at low frequencies, but did not appear to be intensity dependent (Fig. 2B2). This apparent lack of intensity dependency might be the consequence of an overlapping feedforward inhibition (Oda et al. 1998; Preuss and Faber 2003) that would shunt P2. Finally, for a given intensity, the interval between P1 and P2 decreased as a function of frequency (n = 5), suggesting a tendency of the fast EPSPs to follow stimulus frequency or its harmonics. This property is analyzed in more detail for responses to long-lasting amplitude- and frequency-modulated sounds (see following text). The pip-evoked EPSP originates on the distal lateral dendrite
Because it is possible to record sequentially from the M-cell soma and identified dendritic loci, the passive spread of the antidromically generated action potential and that of the mixed electrical and chemical EPSP evoked by stimulation of VIIIp have been studied extensively (see Faber and Korn 1978; Preuss and Faber 2003). Both have spatial profiles consistent with their sites of origin and dendritic cable properties (Fig. 3A). If pip-evoked EPSPs in the M-cell arise from activation of the large myelinated club endings, responses should be maximal in amplitude at the postsynaptic region where these afferents terminate, i.e., 250–400 µm distal from the axon cap. Correspondingly, the amplitude of these responses should become smaller at recording sites farther from the club endings and closer to the soma, according to the cable properties of the M-cell dendrite. This prediction was confirmed in four experiments with recordings of the fast sound-evoked peaks at four sites along the lateral dendrite and the soma (e.g., Fig. 3B1). In addition, this spatial distribution was not frequency dependent, as shown in the example of Fig. 3B2. The fact that, in some cases, P1 amplitude was essentially the same for the two distalmost recording sites suggested a distributed input spanning that region (Fig. 3, B1 and B2). These results are consistent with a distal synaptic input in the region innervated by the large club endings and are not suggestive of a tonotopic map.
|
) was quantified using a single exponential fit of the EPSP tail. It averaged 33.43 ± 3.93 ms in the soma (n = 26) and 30.76 ± 3.76 ms in the dendrite [the amplitudes of the corresponding measurement starting points (e.g., * in Fig. 3, B1 and C) were 1.87 ± 0.12 mV in the soma and 2.16 ± 0.23 mV in the dendrite; n = 19; NSD]. The possibility exists that residual sound reverberations contribute to the duration of the tail potential, but the absence of fast EPSPs in the composite M-cell response suggests this effect is minimal. Given that the M-cell membrane time constant is <0.5 ms, the slow decay of the EPSP is not ascribed to passive membrane properties, but instead either involves recruitment of a conductance change mechanism or reflects persistent afferent activity. Pharmacological dissection of electrical and chemical components of EPSPs evoked by sound "pips"
The chemically mediated EPSP evoked in the M-cell by direct stimulation of VIIIp is glutamatergic, with both a small APV-sensitive component and a larger CNQX-sensitive component (Wolszon et al. 1997). We therefore asked whether antagonists to these receptors blocked any component of the sound-evoked EPSP recorded in the dendrite, using their effect on the EPSP evoked by direct nerve stimulation as a control (Fig. 4A1). This question is especially pertinent for the M-cell because the large majority of the club ending mixed synapses, when stimulated individually, are chemically silent (Lin and Faber 1988b). When strong posterior VIIIth nerve stimuli are used, the amplitude of the electrical EPSP recorded in the dendrite can be >25 mV, and then the size of the associated chemical EPSP is approximately one third that of the amplitude (Fig. 4A1). As shown in Fig. 4A2, individual fast EPSPs in the sound-evoked response, which are about 10 mV in amplitude in this case, do not appear to be followed by clear, chemically mediated EPSPs with kinetics and relative amplitudes comparable to the corresponding component in Fig. 4A1. Nevertheless, chemical EPSPs might have been less synchronized than with nerve stimulation and are thus masked by the complex sound response. However, when the brain was superfused with a cocktail of APV and CNQX (100 µM each), the values of P1, P2, and were at best slightly reduced compared with controls (Fig. 4B), and these effects were not significant (P1: P = 0.35; P2: P = 0.08; : P = 0.18; n = 6). This quantitative analysis demonstrated that the first peaks evoked in the dendrite by a sound pip are primarily mediated by distal electrotonic synapses and the decay of the slow EPSP is insensitive to ionotropic glutamate receptor antagonists.
|
> 3 ms) than the glutamatergic EPSP evoked by nerve stimulation ( 1.7 ms). Thus there is a chemical component to the slow EPSP, but its characteristics differ from those of the example in Fig. 4A1. M-cell responses evoked by amplitude- and frequency-modulated acoustic sounds
In natural habitats, fish are likely to be exposed not only to abrupt short sounds, such as a diving bird hitting the water, but also to longer-lasting stimuli. For example, an approaching predator will increase the intensity and frequency of its caudal fin beat during a strike (Wainwright et al. 2001). We thus asked whether a longer-duration ramped stimulus would reveal additional properties of the M-cell system. Specifically, two types of stimuli were generated: 1) the constant frequency AM stimulus and 2) the AM/FM stimulus. These stimuli were usually 50, 100, or 200 ms in duration.
In general, M-cell responses to AM and AM/FM stimuli consisted of a series of fast EPSPs superimposed on a steadily growing slow EPSP that gradually decayed back to baseline after termination of the sound (Fig. 5, A–D). These fast EPSPs were always largest in the dendrite, as indicated by RMS analysis of the high-pass (HP) filtered representation of the EPSP (HP RMS; dendrite: 0.73 ± 0.03 mV, n = 16, 32 responses; soma: 0.31 ± 0.01 mV, n = 19, 28 responses; P < 0.005, unpaired Student's t-test). These values correspond to a 58% attenuation in the fast EPSPs from distal dendrite to soma.In addition, the fast EPSPs appeared to follow the frequency of the sound and increase in amplitude with sound intensity (Fig. 5, A and B). Discrete chemical EPSPs following the fast EPSPs were not obvious (compare Fig. 4A1 vs. Fig. 5, A, right and B, right).
|
), this is unlikely to be the cause of the difference in attenuation in these experiments because the M-cell membrane has a low specific resistance and time constant of about 0.5 ms (see model results below).
|
1 + A2e–t/2), as illustrated in Fig. 6A1 (bottom traces) for an EPSP evoked by a 200-ms AM stimulus. In these experiments, 1 averaged 25.1 ± 6.6 ms (n = 17, 32 responses) and 15.4 ± 2.7 ms (n = 15, 28 responses) in the soma and dendrite, respectively, and 2 was approximately four times slower, averaging 94.2 ± 7.5 ms in the soma and 106.6 ± 15.6 ms in the dendrite. Somatic and dendritic values of both 1 and 2 were not statistically different (1: P = 0.18; 2: P = 0.47). Notably, the slow EPSPs evoked by the modulated stimuli decay appreciably more slowly than those evoked by pips (soma pip, = 33.43 ± 3.9 ms vs. soma modulated, 2 = 84.2 ± 10.1 ms; dendrite pip, = 30.76 ± 3.76 ms vs. dendrite modulated, 2 = 124.2 ± 23.1 ms). Pharmacological dissection of electrical and chemical components of EPSPs evoked by AM and AM/FM sound stimuli
We next sought to determine whether the fast EPSPs arise solely from current flow across gap junctions at electrical synapses, as with the pips, and whether glutamatergic inputs contributed to the slow EPSP. To examine this, somatic and dendritic recordings of M-cell responses to long-lasting modulated sounds were again obtained before and after superfusion with an array of glutamate receptor antagonists. Antagonists of NMDA (200 µM APV, 100 µM CPP) and AMPA/kainate (150 µM CNQX, 100 µM NBQX) receptors were applied together to maximize their effectiveness. The ionotropic antagonists either had no effect on the size of the fast EPSPs (e.g., Fig. 6A) or actually increased them (Fig. 6C, HP RMS). More specifically, the drugs did not have a significant effect on the HP RMS voltage in the soma (1.2 ± 0.4% increase; n = 12, 19 responses; P = 0.49), although they did produce a nonspecific increase in the HP RMS in the dendrite (27.6 ± 9.8% increase; n = 8, 15 responses; P < 0.05). In contrast, the slow EPSP was reduced significantly in the case of somatic recordings, as confirmed with quantifications of the drug effects on the low-pass–filtered representations of the slow EPSP (Fig. 6C, 
LP and LP Max). Specifically, in the soma there was a significant blocking effect of the cocktail on both the integrated low-pass signal (28.0 ± 5.4% reduction; P < 0.05) and its peak amplitude (18.4 ± 3.4% reduction; P < 0.05), whereas the dendritic integrated low-pass signal (1.0 ± 5.2% reduction; P = 0.87) and peak amplitude (5.1 ± 5.5% increase; P = 0.37) were unaffected. This result is consistent with a predominantly proximal source of the glutamatergic component of the slow EPSP.
The data in Fig. 6 are from experiments in which either somatic or dendritic recordings were performed, but not both. To confirm these results, five additional experiments were performed in which recordings were obtained sequentially from the soma and the dendrite in the same preparation. In these experiments, the cocktail of AMPAR and NMDAR antagonists decreased LP max in the soma by 35 ± 4%, from 3.61 ± 0.49 mV (P < 0.05), compared with 17 ± 2% in the dendrite, from 5.17 ± 0.84 mV (P = 0.05). It should be noted that the peak amplitude of the experimentally recorded EPSP was unchanged after drug application (8.98 ± 0.92 vs. 8.52 ± 0.85 mV; P = 0.46; n = 5), resulting from a slight increase in the amplitude of the fast EPSPs that compensated for the reduced slow component. This increase in fast EPSP amplitude was most likely the result of a nonspecific increase in input resistance after drug application.
It is somewhat surprising that the glutamatergic antagonists did not have an effect on the slow EPSP recorded in the distal dendrite because the drug-sensitive component that is generated proximally should contribute to the response recorded distally, albeit with some attenuation arising from the passive cable properties of the dendrite. Ineffectiveness of the antagonists is unlikely, given that data are reported only for experiments in which the nerve-evoked chemical EPSP was blocked. A more likely explanation is that the dendritic representation of the proximally generated slow EPSP is a small fraction of the response recorded in the dendrite and the component removed by the antagonists is therefore relatively small. In confirmation, if experiments in which the integrated area of the low-pass–filtered PSP was reduced in the dendrite are considered separately (five of eight animals), the mean reduction in this measure was 10.8 ± 2.0% (nine responses).
Finally, control experiments demonstrated that a cocktail of AMPA and NMDA antagonists had negligible effects on synaptic transmission at hair cell–VIIIp afferent synapses. If this connection is blocked during drug application, it should result in decreased VIIIth nerve activity and, consequently, decreased fast EPSP amplitude. Therefore simultaneous recordings of extracellular activity in the afferent nerve and intracellular recordings of M-cell PSPs were performed (n = 3, 12 responses). In these experiments, cocktail application resulted in a 51 ± 5% reduction in the integrated area of the low-pass–filtered PSP ( LP: P < 0.0005) and a 42 ± 6% reduction in LP max (P < 0.0005). In contrast, the RMS values of both the high-pass signal and the extracellular recordings were decreased by 19 ± 5% (P < 0.005) and 14 ± 5% (P < 0.05), respectively. Overall, these results indicate a greater effect on the magnitude of the slow EPSP than the fast EPSPs and thus a greater effect of the drugs on the VIIIp–M-cell synapse than the hair cell–VIIIp connection.
We also considered the possibility that a portion of the slow EPSP could be attributable to activation of metabotropic glutamate receptors (mGluRs). However, superfusion of the brain for 3 h with antagonists of Groups I, II, and III mGluRs (1 mM MSOP, 1 mM MCPG, and 200 µM LY 341495) did not reduce any EPSP component (data not shown; n = 12).
To summarize, although the slow EPSP is partially glutamatergic, evidence that the ionotropic antagonists had a proportionally greater effect on the responses recorded in the soma argues against a contribution of chemical synaptic inputs mediated by the large myelinated club endings that terminate on the distal lateral dendrite. Furthermore, discrete chemical EPSPs similar to those seen during direct VIIIp stimulation were not apparent in the sound-evoked responses and the decay kinetics of the slow EPSPs were at least one order of magnitude slower than those of the glutamatergic EPSPs evoked by VIIIp stimulation. Finally, the attenuation from dendrite to soma of this slow response was significantly less than that of the fast EPSPs.
Spatially distributed source of the slow EPSP
Because the difference in the spatial profile of the slow and fast EPSPs might have been a reflection of the frequency-dependent spatial-filtering properties of the dendrite, we implemented a multicompartment model to determine whether this was likely to be the case (see METHODS; Fig. 7, top).
|
In a similar manner, EPSPs recorded in the distal dendrite were injected into the model, and the waveforms of predicted and observed responses were compared at multiple locations. In the example of Fig. 7A2, the composite EPSP evoked by a 300-Hz sound pip recorded 400 µm distally on the lateral dendrite was injected into the model at 350 µm, the rationale being that sound-evoked EPSPs have approximately the same amplitude at 300 and 400 µm. At 200 µm, the model correctly predicted both the attenuation of the overall EPSP amplitude and the alterations in its shape. Although it also accurately predicted the amplitude of P1, i.e., the fast component, and the general shape of the response at 50 µm (soma), it significantly overestimated attenuation of the slow EPSP for all three data sets.
To further explore the origin and attenuation of the slow EPSP, the experimentally obtained average responses to a 100-Hz, 200-ms AM stimulus recorded in the distal dendrite before and after CNQX/APV application were injected into the model. The corresponding results are illustrated in Fig. 7, B1 and B2, respectively. In both cases, the resulting modeled response in the soma underestimated the amplitude of the underlying slow EPSP recorded there (n = 2), but the discrepancy was approximately halved after CNQX/APV application. These results again suggest the existence of a proximal component to the slow EPSP that depends on activation of ionotropic glutamatergic transmission. Also, the component remaining after application of the antagonists does have a spatial profile consistent with a dendritic origin.
Spectral analysis of the sound-evoked EPSPs
The preceding data indicate that modulated acoustic stimuli evoke two types of EPSPs in the M-cell: fast, which are presumably mediated by electrical synapses, and slow, which apparently have a distributed origin. To determine the auditory information transmitted by these two distinct signals, we first asked to what extent the composite EPSPs followed stimulus frequency, and then separated them into their high- and low-frequency components.
The responses of one M-cell to 250- and 300-Hz AM stimuli are illustrated in Fig. 8, A1 and A2, respectively. As already noted, the fast EPSPs grow in amplitude with the stimulus and are more closely spaced at the higher stimulus frequency. The power spectral densities of these EPSPs have, in addition to high power at low frequencies, sharp peaks at both the fundamental frequency and its first harmonic (Fig. 8A3). There was less power in the harmonic, consistent with the observation that in some experiments the amplitude of the fast EPSPs seemed to alternate; i.e., there were two peaks per stimulus cycle with the amplitude of the first at least twice that of the second (e.g., Fig. 5B). The latter is also consistent with previously reported results showing that auditory afferents fire once or twice per cycle depending on fiber type (Furukawa 1967). Overall, high-pass–filtered responses (>60 Hz) to AM stimuli ranging from 100 to 800 Hz were analyzed in 23 experiments, and in no case was power found outside of the range of the fundamental frequency and its harmonics. The half-width of these spectral peaks was about 5 Hz (Fig. 8A3). Thus the high-frequency component of the EPSP accurately transmits information about the frequency of the sound. This was further illustrated by comparing the joint time-frequency spectrograms (JTFS) of the high-pass–filtered representation of the evoked responses and the corresponding sound stimuli, shown superimposed in Fig. 8A4 for a 250-Hz AM stimulus. Another example of M-cell tracking can be seen in a sample response evoked by an AM/FM stimulus, which increases in both frequency and amplitude with time (Fig. 8, B1 and B2). In both cases, superimposing the JTFS of the stimulus on that of the high-pass–filtered EPSP graphically demonstrates that the M-cell follows the fundamental frequency and its first harmonic of complex sounds, including for low intensities.
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSREFERENCES |
|---|
). These results suggest the chemical component of these morphologically mixed synapses was weak or even functionally silent in the present experimental conditions (Lin and Faber 1988a,b). Phasic EPSPs are mediated by electrotonic synapses
Evidence is accumulating that electrical synapses might be more widespread in the vertebrate CNS than previously recognized (Nagy et al. 2004). Our conclusion that fast EPSPs are electrical coupling potentials is based on 1) their kinetics, 2) their insensitivity to AMPAR and NMDAR antagonists, and 3) their spatial profiles along the soma–dendritic cable that are consistent with a distal dendritic origin. These coupling potentials have the same half-width (i.e., time course) as that of the presynaptic spikes in this system because the M-cell membrane time constant is so short. Consequently, the electrical synapses faithfully follow stimulus frequency, up to about 1,000 Hz (data not shown). The notion that the electrotonic synapses act as high-pass filters of auditory signals might be unique to the M-cell system, although such a role was previously observed in the electrosensory system of mormyid electric fish (Bell and Grant 1989; reviewed in Carr and Friedmann 1999). In other systems, such as mammalian cortex and hippocampus, electrotonic coupling between interneurons is very weak and longer membrane time constants slow potentials and distort their waveforms (see Connors and Long 2004). The coupling in these systems seems to aid in synchronizing firing patterns and its role in information processing is less clear. This functional difference would appear to parallel the differences in the type of coupling: cortical and hippocampal coupling is between homologous cell types, and there are no clear pre- and postsynaptic elements, in contrast to the electrical synapses between afferents and identified target neurons in the M-cell case and in the electrosensory lateral line system of Gymnotids.
Distributed origin of the slow EPSP
Previous studies using paired pre- and postsynaptic recordings indicated that at least 80% of M-cell afferent synapses are chemically silent when activated in isolation (Lin and Faber 1988b), but might be unblocked when a population of fibers is coactivated synchronously (Pereda et al. 2004). For example, a clear chemical EPSP is recorded in the dendrite when the preceding coupling potential has an amplitude in the range of 20 mV. Results using sound pips revealed a small glutamatergic component with a longer latency and kinetics slower than those of unitary EPSPs (e.g., Fig. 4C). We used longer-lasting, modulated acoustic stimuli, that is, those that triggered escapes in behavioral studies when applied under water (unpublished observations) and asked whether chemical synaptic transmission might be more prominent in those conditions. These paradigms revealed larger, slow postsynaptic responses that had no significant NMDAR component, despite the fact that the slow EPSP amplitude was in the range where EPSPs evoked by VIIIp stimulation are sensitive to NMDAR blockers (Wolszon et al. 1997). In addition, the slow EPSPs were only partially blocked by the ionotropic glutamate receptor antagonists, and their rates of rise and decay were one to two orders of magnitude slower than those of chemically mediated EPSPs evoked by direct nerve stimulation. Although the spatial distribution of the fast EPSP was consistent with a purely distal dendritic source, this was not the case for the slow EPSP. In some experiments, the slow EPSP amplitude was maximal in the distal dendrite, whereas in others it was the same size in the soma as in the dendrite. The latter case is consistent with a combination of distal and proximal sources. Even when the slow EPSP was maximal distally, its decrement from dendrite to soma was less than that predicted by a multicompartment model of the cell that both reproduced the measured input resistance and time constant of the M-cell, as well as satisfactorily fit the temporal and spatial profile of the fast electrical EPSPs and the antidromic spike. This discrepancy between predicted and observed cable properties was seen only with the attenuation of the slow EPSP from dendrite to soma, especially before superfusion with a cocktail of the ionotropic glutamatergic blockers. Furthermore, the cocktail of antagonists was most effective on responses recorded in the soma and least effective at the dendrite. These observations are consistent with the notion of a proximal glutamatergic source to the slow EPSP and a more distal nonglutamatergic component.
It is important to note that because acoustic stimuli were applied in air, the evoked PSPs were presumably smaller than those that trigger an M-cell spike and an escape behavior when comparable underwater stimuli are used. It is not clear whether the underwater stimulus has a larger pressure component and therefore synchronously activates more posterior VIIIth nerve afferents. Thus that is one possible condition in which the club endings would mediate not only electrical but also chemical PSPs. Furthermore, if the mechanism that unblocks chemical transmission is a result of feedback of M-cell dendritic depolarizations to the presynaptic terminals via the gap junctions (Pereda et al. 2004), an additional source of that depolarization could be PSPs mediated by other eighth nerve or lateral line afferents that are sensitive to different components of an acoustic stimulus.
The most likely source of the proximal input is a mono- or disynaptic input from saccular or lagenar afferents, different from the fibers that give rise to the large myelinated club endings. Previous morphological studies have not provided evidence for such a monosynaptic input (Zottoli 1978), suggesting that a disynaptic pathway is more likely. Indeed, the interposition of interneurons might account for the lack of evidence for discrete time-locked chemical EPSPs.
Potential mechanisms underlying the distal component of the slow EPSP
The source of the residual slow EPSP that persists following application of the ionotropic glutamatergic antagonists is less clear. We considered the possibility that synaptic activation of a metabotropic glutamate receptor (Coutinho and Knopfel 2002; Knopfel and Grandes 2002) might be involved, but the failure to detect any blocking effect with a broad spectrum of mGluR antagonists suggests that is not the case. The most likely alternative is electrotonic transmission of slow potentials generated in the VIIIp afferents. The afferent spikes have depolarizing afterpotentials, as reflected in the isolated coupling potential waveform in Fig. 4A1, and successive depolarizations may summate. The afferents also exhibit a voltage-dependent persistent Na+ current (Curti and Pereda 2004) that might produce a slowly developing depolarization of the afferents, which would in turn spread to the lateral dendrite. However, this nonlinearity might be offset during maintained or repetitive presynaptic depolarization by an opposing action of voltage-dependent K+ channels. Another source of a presynaptic slow potential would be the chemical synapses between sensory hair cells and the afferents. The length constant of the afferents seems to be sufficiently long for such a signal to spread to the M-cell because efferent IPSPs generated at the hair cell synapses can be recorded in the afferents close to the site of entry in the brain (Lin and Faber 1988c). Finally, we consider a subthreshold, voltage-dependent nonlinearity of the dendritic membrane to be less likely. The only nonlinearity described thus far is an instantaneous inward rectification that is manifest as a decreased input conductance at depolarizations 5 to 10 mV above the M-cell resting potential (Faber and Korn 1986). This conductance deactivates more rapidly than the slow EPSP, however, and neither voltage-clamp nor current-clamp experiments have unmasked evidence for a subthreshold inward current with the appropriate kinetics. Regardless, we suggest that, in addition to the proximal glutamatergic input to the M-cell, a major component of the slow EPSP is transmitted to the M-cell via electrical synapses, as is also the case for the fast EPSP.
Functional significance and behavioral relevance of the EPSP components
Both the abrupt "pip" and the more gradual modulated stimuli used in this study can evoke escapes when applied under water. Although the stimuli in air produce the same pressure waves as those in water, they do not generate the other component of sound in water, that is, particle displacement or acceleration (Popper and Fay 1999). This difference most likely explains why the PSPs we recorded were not large enough to trigger an M-cell action potential. We also considered the alternative explanation: that the anesthetic MS-222 reduced the PSP magnitude or altered its waveform. However, data from Palmer and Mensinger (2004) suggests the anesthetic has minimal effect on this afferent pathway at the concentration used. In confirmation, we conducted an institutionally approved pilot study in which two fish were initially anesthetized by immersion in ice water and local anesthetic (20% benzocaine gel, Ultradent) was applied topically to the skull and all contact points before restraining the fish. The EPSP waveforms and magnitude were comparable to those in the experiments described here, and addition of MS-222 (60 µg/l) to the water perfusing the gills had minimal effect on the evoked responses. We thus conclude that the PSPs described here are representative of those that can trigger an escape and that in order to reach threshold for that behavior, these responses summate with those mediated by the other afferent pathways; i.e., by VIIIth nerve afferents excited by the displacement component of underwater acoustic stimuli and, possibly, by lateral line afferents (Mirjany et al. 2005).
In the auditory system of a range of vertebrates, and in fish electrosensory systems, sensory information is separated into two coding channels specialized for phase detection and stimulus amplitude tracking (Carr and Friedman 1999). In the mammalian auditory system, the kinetics of glutamatergic EPSCs allows them to follow high-frequency inputs and therefore have a role in phase detection (Otis et al. 1995; Raman et al. 1994). In contrast, our results suggest that the two coding channels converge at the M-cell level, and that the fast EPSP is mediated by electrical synapses on the M-cell and encodes stimulus frequency and intensity, whereas the slow EPSP converts a phasic input into a tonic signal that is proportional to the envelope of peak stimulus amplitude. These two aspects of a threatening stimulus, intensity and frequency, are both functionally relevant. Intensity and its rate of change are measures of the size of an approaching object and its speed. In the case of frequency and associated phase-related information, it has been postulated that these parameters are important in determining the directionality of the escape behavior, particularly in the case of stimuli directly to one side or the other [i.e., in selecting which M-cell will fire an action potential (Casagrand et al. 1999)]. Thus the M-cell and its afferent network are well suited for retaining information about the dynamic properties of complex stimuli. Indeed, it has been suggested previously that information coding in the CNS is optimized for an efficient representation of dynamic and complex signals from the environment (Barlow 1961; Machens et al. 2005), a principle that apparently is realized for sound-evoked responses in the M-cell. Moreover, we suggest the separation of an incoming signal into two different information channels might be a common function of neural networks that use electrical synapses. Differential coding properties have been suggested for electrical and chemical synapses between proprioceptive afferents and interneuron A in the crayfish (Nagayama et al. 1997), although the functions have not been delineated.
Although stimulation of the eighth nerve also activates a feedforward inhibition of the M-cell (Faber et al. 1989), we have not considered its role here and have focused on the composite EPSPs. Theoretically, however, inhibition could shunt the magnitude of the fast EPSPs and produce frank hyperpolarizations as the slow EPSP reaches its peak. In fact, it might contribute to the initial fast decay (1) of the slow EPSP or to the alternation in the amplitudes of successive fast EPSPs. In addition, it will be important to determine whether inhibition shapes the input–output relations of the fast and slow PSPs, as well as how it influences the correlations between these responses and stimulus intensity or power.
The escape response initiated by an M-cell action potential can be triggered by sound pips as well as by longer-lasting amplitude- and frequency-modulated sounds. These stimuli represent environmental perturbations potentially produced by a variety of predatory activities (Bleckmann et al. 1991). For example, an approaching predator might alter its fin movements and swimming velocity—sometimes even producing suction—during a strike (Wainwright et al. 2001; Webb 1976). These behaviors activate multisensory inputs to the M-cell (Popper and Fay 1999), and although the pressure input, transmitted by the swim bladder to the posterior eighth nerve, is necessary for the escape behavior (Canfield and Eaton 1990), it presumably is insufficient to trigger the escape by itself (Casagrand et al. 1999). In fact, the input from the swim bladder imparts to certain fish, such as goldfish, the designation of hearing specialists; that is, it underlies their particular sensitivity to sound (Fay and Popper 1999). The composite EPSPs evoked by the modulated stimuli used in this study appear to aid the integrative requirements for the escape behavior. Although the phasic component tracks stimulus intensity and frequency, the tonic signal increases the chances that the M-cell reaches threshold when inputs from multiple sources are summed; i.e., the fast EPSP alone accurately follows sound frequency or its first harmonic and therefore favors a threshold mechanism based on temporally coincident, or convergent, input signals. However, the slow EPSP is a more slowly modulated signal that therefore presumably reduces or relaxes this requirement for coincident inputs. That is, the slow EPSP might allow for summation of inputs on a broader timescale than the faster responses, whose waveforms are comparable to those of the presynaptic action potentials. This could be very important in situations where the escape response is triggered by multimodal sensory inputs; for example, a combination of acoustic and visual stimuli.
|
GRANTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSREFERENCES |
|---|
|
FOOTNOTES |
|---|
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.
Address for reprint requests and other correspondence: T. Preuss, Department of Neuroscience, Albert Einstein
