Mechanisms of frequency-specific responses of ω neuron 1 in crickets (Teleogryllus oceanicus): a polysynaptic pathway for song?

Crickets use sound to find mates (Alexander, 1960), to compete with rivals (Alexander, 1961) and to detect echolocating bats (Moiseff et al., 1978; Nolen and Hoy, 1986). Because of the importance of sound to the biology of crickets and the well-known advantages of insects for neurobiological studies, crickets have long served as subjects for the study of auditory processing (for a review, see Hoy et al., 1998). A prominent bilateral pair of auditory interneurons in crickets are the omega neurons 1 (ON1) (Casaday and Hoy, 1977; Popov et al., 1978; Wohlers and Huber, 1982). Each ON1 receives excitatory input mainly from one ear (the ear ipsilateral to its soma) and inhibits several auditory interneurons that receive their input from the opposite ear, including the contralateral ON1 and ascending neurons 1 and 2 (AN1, AN2) (Selverston et al., 1985; Faulkes and Pollack, 2000). The resulting inhibition enhances binaural contrast in the auditory pathway, thereby improving the ability to localize sounds (Atkins et al., 1984; Horseman and Huber, 1994; Schildberger and Hörner, 1988).

The neuron ON1 responds to sound over a wide frequency range, but it is most sensitive to the carrier frequency of conspecific calling song (Popov et al., 1978; Wohlers and Huber, 1982; Atkins and Pollack, 1986), which is approximately 4.5 kHz in Teleogryllus oceanicus (Hill et al., 1972; Balakrishnan and Pollack, 1996). Previously, we showed that the latency of ON1 is longer, by up to 10 ms, for 4.5 kHz compared with other frequencies, including ultrasound (Pollack, 1994; Faulkes and Pollack, 1997; Faulkes and Pollack, 2000). We have reported on the functional consequences of this elsewhere (Faulkes and Pollack, 2000), and in the present paper we focus on the mechanisms that might be responsible for the delayed response to cricket-like frequencies.

Fig. 1 illustrates several candidate mechanisms. Pollack (Pollack, 1994) suggested that the conduction velocities of auditory receptors might differ in a frequency-specific manner (Fig. 1A), but Pollack and Faulkes (Pollack and Faulkes, 1998) found no evidence for this. The remaining hypotheses are, briefly, that there is frequency-specific inhibition, either postsynaptic (Fig. 1B) or presynaptic (Fig. 1C), that the integrative properties of ON1 create differences in excitatory postsynaptic potential (EPSP) shape and thus in the timing of spike onset (Fig. 1D) and that there are frequency-specific afferent pathways to ON1 (Fig. 1E).

This work has been published previously in abstract form (Faulkes and Pollack, 1997).

Field crickets, Teleogryllus oceanicus (Le Guillou), were raised in laboratory colonies where food (Purina Cat Chow) and water were available continuously. Unmated female crickets, 1–3 weeks of age after the final moult, were used in all experiments.

Animals were anaesthetised by chilling on ice. The meta- and mesothoracic legs were autotomised, and the wings and antennae were removed. We affixed the animals, ventral side up, to a magnetic base with wax, and then used wax to immobilise the tibia and femur of the prothoracic legs, with the femur positioned horizontally at right angles to the body axis. Two fine Teflon-coated silver wires were inserted into the femur of the leg ipsilateral to the sound source to record sensory compound action potentials (Pollack and Faulkes, 1998). We removed the ventral thoracic cuticle to expose the prothoracic ganglion, which was supported on a metal platform. The ganglion was submerged in physiological saline (140 mmol l⁻¹ NaCl, 10 mmol l⁻¹ KCl, 7 mmol l⁻¹ CaCl₂, 4 mmol l⁻¹ NaHCO₃, 1 mmol l⁻¹ MgCl₂, 5 mmol l⁻¹ Tes, 5 mmol l⁻¹ trehalose; modified from Strausfeld et al., 1983).

We made extracellular recordings of ON1 in the hemiganglion contralateral to the main dendrite receiving auditory input (hereafter designated ‘contralateral’, after Schildberger and Hörner, 1988; the laterality of all auditory interneurons will be described relative to their main source of auditory input) using low-resistance microelectrodes (<10MΩ) filled with 2 mol l⁻¹ NaCl. ON1 was identified by its greater sensitivity to sounds contralateral to the recording electrode (i.e. ipsilateral to input) (Pollack, 1986). We made intracellular recordings from the main auditory dendrite of ON1 using high-resistance (>30MΩ) microelectrodes filled with 2 mol l⁻¹ potassium acetate, 2 mol l⁻¹ lithium chloride or 3 mol l⁻¹ potassium chloride. A chlorided silver-wire indifferent electrode was placed in the abdomen. ON1 was identified by the correspondence between intracellularly recorded spikes and simultaneously recorded extracellular spikes. Intra- and extracellular recordings were amplified using a microelectrode amplifier (Getting 5A or WPI M-707). For some experiments, we made extracellular recordings of ascending neuron 2 (AN2; ‘Int-1’ in Casaday and Hoy, 1977; Wohlers and Huber, 1978) with a polyethylene suction electrode attached to the cervical connective. We identified AN2 by the laterality of its auditory input (i.e. greater sensitivity to sound presented ipsilateral to the recorded connective), large-amplitude spikes and strong response to ultrasound (Wohlers and Huber, 1978; Moiseff and Hoy, 1983; Faulkes and Pollack, 2000).

Stimuli were produced by an analog/digital board (National Instruments AT-MIO-64-F5; 12 bits, update rate 250 kHz) and relayed through a programmable attenuator (Tucker-Davis Technologies PA4) and amplifier (Amcron D150A). The sound stimuli were played through piezoelectric loudspeakers situated to the left and right of the cricket, perpendicular to its longitudinal axis, at a distance of 35 cm. Experiments were performed in a chamber lined with echo-attenuating mineral-wool wedges.

Sound stimuli were played at 2 pulses s⁻¹. The pulse envelope was trapezoidal in shape and, in most experiments, had a duration of 20.4 ms (including 0.2 ms rise and fall times). In some experiments, we used 0.4 ms sound pulses (0.2 ms rise and fall times). As reported previously, the latency of ON1 varies with sound frequency (Faulkes and Pollack, 2000) and is longest at approximately 4.5 kHz. On the basis of this finding, we used 3 kHz, 4.5 kHz (‘low-frequency’), 10 kHz and 30 kHz (‘high-frequency’) stimuli to characterise the response of ON1 in these experiments. Stimulus intensity ranged from 25 to 100 dB SPL (re 2×10⁻⁵ N m⁻² ). Data and event markers were recorded on a frequency-modulated tape recorder (Vetter; model D) and then digitised through an analog/digital board (National Instruments; AT-MIO-64-F5) using the program SWEEPS (Pollack, 1997), at a sampling rate of 10 kHz per channel with 12 bits of analog to digital resolution. The same program was also used for analysing data off-line.

In some experiments, we deafened crickets unilaterally by cutting or crushing the leg nerve contralateral to the sound source. For experiments in which we used 0.1 mmol l⁻¹ picrotoxin (PTX) (Sigma), we cut the anterior and posterior connectives of the prothoracic ganglion to ensure that the drug entered the ganglion quickly. The ganglion remained bathed in PTX for 10 min before data collection commenced.

The latency of ON1 spikes was measured from extracellular recordings as the time between stimulus onset and the succeeding ON1 spike. In most cases, threshold was defined as the lowest sound intensity (±2.5 dB SPL) that generated a mean response of one or more spikes per 100 ms sampling window. In the PTX experiments, threshold was defined in a different manner because PTX tended to increase background activity. Here, spontaneous ON1 activity (spikes per 100 ms in the absence of stimulation) was measured for 3 s to provide a baseline; threshold was defined as the lowest intensity at which the mean spike count per 100 ms sampling window exceeded the baseline count by one or more spikes.

Statistical analyses were performed using the software package Statistica for Windows 5.1 (StatSoft, Inc.). The results for 3 kHz were similar to those for 4.5 kHz, and the results for 10 kHz were similar to those for 30 kHz. The data for 3 and 10 kHz were, therefore, omitted from the figures for clarity and brevity but were included in the statistical analyses.

Postsynaptic inhibition could explain the sound-frequency-specific differences in latency of ON1 if it were appropriately timed, i.e. either preceding sound-evoked excitation or simultaneous with it. Earlier intracellular recordings (e.g. Pollack, 1994) did not reveal inhibitory postsynaptic potentials (IPSPs) preceding excitation, but those recordings were made using microelectrodes containing Cl⁻, while ON1 was strongly hyperpolarised. Such conditions could mask IPSPs, which are typically mediated by Cl⁻ channels and have a reversal potential near the resting potential. We made intracellular recordings using electrodes filled with potassium acetate. No early IPSPs were seen in response to ipsilateral sound stimuli (Fig. 2A,B), either at the normal resting potential of the cell or when ON1 was depolarised with a current of 2–4 nA, which should enhance IPSPs. When depolarised by injected current, ON1 spikes tonically. If there were cryptic inhibition preceding excitation not visible as an IPSP, one might expect (i) that the firing frequency of spontaneous ON1 spiking would be reduced, and (ii) that the height of these current-evoked spikes would be reduced prior to the onset of the auditory-evoked response, as a result of shunting by the conductance increase associated with the IPSP. No reduction in firing rate (Fig. 2C) or in spike height (Fig. 2D) was observed before the sound-evoked response. Previous experiments have shown that ON1 experiences post-excitatory inhibition, which is most prevalent at higher sound intensities and persists after the offset of excitation (Pollack, 1988). This serves as a positive control for the utility of firing frequency and spike-attenuation measurements as a means of revealing inhibition. Current-evoked spikes occurring during the post-excitatory period of inhibition occur at lower frequency (Fig. 2C) and are attenuated (Fig. 2Dii).

Inhibition of ON1 by contralateral sound is well documented and is caused largely, if not exclusively, by the contralateral ON1 (Wiese and Eilts, 1985; Horseman and Huber, 1994; Selverston et al., 1985). We checked for any effect of contralateral inhibition on the latency of ON1 by cutting the leg nerve contralateral to the sound source, thereby eliminating any sound-evoked contralateral influences. Contralateral deafening did not affect the latency of ON1 (repeated-measures analysis of variance, ANOVA, N=5; F_1,125=0.37, P>0.5) (Fig. 3).

γ-Aminobutyric acid (GABA), a common inhibitory neurotransmitter in arthropods, is responsible for presynaptic inhibition in many invertebrates and vertebrates (Watson, 1992), and GABA-immunoreactive profiles terminate both on ON1 (Watson and Hardt, 1996) and on terminals of auditory receptors (Watson and Hardt, 1999). Thus, GABA is a good candidate to mediate presynaptic inhibition of auditory receptors (Fig. 1C) or postsynaptic inhibition of ON1 itself (Fig. 1B). Picrotoxin (PTX) blocks GABA receptors in many invertebrates (Sattelle et al., 1991) and has been shown to block inhibition in the auditory systems of crickets and bush crickets (Harrison et al., 1988; Stumpner, 1998). If GABA-mediated inhibition were responsible for the frequency-specific responses of ON1, PTX should shorten the latency of ON1 specifically to low-frequency sounds. Treatment with 0.1 mmol l⁻¹ PTX caused a small, but statistically significant, reduction in ON1 latency across all frequencies (control, 25.7±4.63 ms; PTX, 24.88±4.52 ms; repeated-measures ANOVA; N=5; F_1,119=20.01, P<0.01; means ± S.D.). Nevertheless, there was no specific effect on responses to low frequencies (Fig. 4; no significant interaction between PTX and sound frequency; repeated-measures ANOVA, N=5; F_3,119=2.02, P=0.12).

As a positive control for the effectiveness of PTX, we also recorded from ascending neuron 2 (AN2). This neuron is strongly excited by ultrasound and receives mixed excitation/ inhibition in response to 4.5 kHz stimuli (Nolen and Hoy, 1987). Previous experiments showed that the low-frequency inhibitory input is PTX-sensitive (Harrison et al., 1988). As expected from the foregoing, PTX treatment enhanced the response of AN2 to 4.5 kHz (Fig. 4A, inset), but had no effect on the response to ultrasound (not shown; interaction between PTX treatment and stimulus frequency on the spike count of AN2, repeated-measures ANOVA: N=8; F_3,168=45.67; P<0.01). Treatment with PTX also tended to increase spike counts in ON1 (control, 6.71±3.40; PTX, 6.90±3.58; repeated-measures ANOVA: N=5, F_1,134=3.91; P=0.05; means ± S.D.).

This increase, however, could be attributed largely to an increase in spontaneous activity (see Materials and methods). When spontaneous activity was subtracted, PTX caused a small, but statistically significant, decrease in spike number (control, 6.23±3.33; PTX, 5.95±3.56; repeated-measures ANOVA: N=5, F_1,134=5.47; P<0.05; means ± S.D.) but, as for latency, this effect was not frequency-specific (repeated-measures ANOVA: N=5, F_3,134=0.88, P>0.05).

The EPSPs evoked by ultrasonic stimuli rise more rapidly than those evoked by low frequencies (Pollack, 1994). Intracellular recordings, however, indicated that the difference in rise times could not be the sole explanation for the longer latency of ON1 to low-frequency sounds. The EPSPs evoked by low-frequency sound also had longer onset latencies (Fig. 5A). The EPSPs also differed in onset latency relative to the first receptor compound action potential (Fig. 5B), demonstrating that the long latency to low-frequency sound is not a reflection of different latencies of receptor neurons.

The last hypothesis (Fig. 1E) makes two separate claims:

(i) that there are monosynaptic connections to ON1, predominantly for high frequencies, and (ii) that there is a polysynaptic pathway to ON1, mainly for low frequencies. There is anatomical evidence (Hirtz and Wiese, 1997) for direct input from receptor neurons to ON1, but the frequency-sensitivity of these receptors was not determined. To provide physiological evidence for monosynaptic input from high-frequency receptors, we made simultaneous recordings of EPSPs in ON1 and extracellular spikes in the leg nerve. Ultrasound stimuli evoke trains of discrete EPSPs in ON1 (Pollack, 1994). Aligning many of these EPSPs in digitised recordings revealed that each was preceded, with constant latency, by a spike in the leg nerve (Fig. 6). The one-to-one, fixed-latency correspondence between EPSPs and receptor spikes is most readily explained by a monosynaptic connection between the receptor and ON1. Low-frequency stimuli, unlike ultrasound stimuli, do not evoke large, discrete EPSPs (Pollack, 1994), so this technique cannot clarify the nature of the connections between low-frequency receptors and ON1.

We sought evidence for a polysynaptic pathway in the shape of compound EPSP waveforms. If all the presynaptic neurons fire once, and simultaneously, they should generate a single EPSP peak if only monosynaptic connections are present. Conversely, the same single, tightly synchronised volley of presynaptic spikes should generate multiple EPSP peaks if a mixture of mono- and polysynaptic connections is present. This assumes that auditory receptors have similar conduction velocities, and we have shown previously that this is the case for the bulk of the population of high- and low-frequency receptors (Pollack and Faulkes, 1998). More recent studies show that the minimum latencies of two different subgroups of low-frequency receptors differ by 2–3 ms, as measured at the entry of the auditory nerve into the prothoracic ganglion (Imaizumi and Pollack, 2001). We applied short (0.1 ms) electrical stimuli to the leg nerve while recording intracellularly from ON1 and extracellularly from the leg nerve. Electrical stimuli evoked compound EPSPs with multiple peaks (Fig. 7). The compound action potentials recorded in the leg nerve also showed multiple peaks, reflecting multiple firings of receptors and/or successive volleys of receptor groups with slightly different conduction velocities. However, successive compound action potentials were separated by only 2–3 ms, which is too short to account for the difference in onsets of approximately 10 ms of the successive compound EPSPs.

Interpreting the results from electrical stimulation is difficult, however. First, electrical stimulation of the entire nerve cannot selectively stimulate different groups of auditory receptors tuned to different frequencies. Second, the ‘auditory’ nerve also carries the axons of non-auditory sensory neurons and of motor neurons (Eibl and Huber, 1979), and additional peaks in the compound EPSP of ON1 might be generated by ortho-or antidromic stimulation of these non-auditory axons. To restrict our stimuli more selectively to auditory receptors, we used short (0.4 ms) sound stimuli to create brief, tightly synchronised receptor responses. In an effort selectively to excite different populations of receptors, we generated these sound pulses at both low and high sound frequencies. These stimuli were amplitude-modulated throughout their entire durations (see Materials and methods) and thus had broader spectra than our standard stimuli, which consisted mainly of a constant-amplitude sine wave. Spectral analysis confirmed, however, that the ‘intended’ frequency (either 5 kHz or 30 kHz; see Materials and methods) dominated the spectra of these brief stimuli, with other components being less intense by at least 30 dB. Nevertheless, each of these stimuli probably excited both high- and low-frequency-tuned receptors when presented at high intensities, both because of their broad spectra and because individual receptors often have additional sensitivity peaks at frequencies other than that to which they are most sensitive (Imaizumi and Pollack, 1999).

Brief ultrasound stimuli always generated a single EPSP peak, regardless of intensity (N=7 individuals; Fig. 8A). Equally brief, low-frequency sounds generated more complex waveforms. At high intensities, these stimuli produced EPSPs with two main peaks (Fig. 8), approximately 10–12 ms apart (Table 1). At lower intensities, the early EPSP became less prominent (often visible as a change in slope in the compound EPSP, rather than as a true peak that rose and fell before rising again) or only a single peak was visible (Fig. 8Aii; Table 1). The mean threshold for any visible EPSP was 66.7 dB SPL (N=6), and the timing of these single EPSP peaks indicates that they correspond to the late EPSPs seen at higher intensities. Two peaks in the EPSP were visible at an average of 78.3 dB SPL (N=6); this should represent the threshold for the early EPSP seen at higher intensities. The latency of the early, low-frequency-evoked EPSP corresponded well with the latency of the ultrasound-evoked EPSP (Fig. 9; Table 2), suggesting that this early peak may result from stimulation, by low-frequency sound, of ultrasound-tuned receptors.

Despite the short duration of the sound stimuli, the recordings of receptor compound action potentials exhibit multiple peaks, indicating repetitive firing of receptor neurons and/or successive volleys from receptors with different conduction velocities or response latencies (Fig. 8A). As in the experiments with electrical stimuli, however, the delay between the compound action potentials, 2–3 ms, is too short to account for the multiple peaks in the compound EPSP, which are separated by approximately 12 ms. Indeed, there is no indication in the EPSP waveform of the arrival of successive, closely spaced waves of afferent input, perhaps because the strong initial depolarisation from the first volley of receptor spikes masks additional EPSPs evoked by subsequent firings of receptors.

We can eliminate several of the proposed hypotheses as mechanisms for the frequency-specific response latency of ON1 (Fig. 1). The conduction velocities of the majority of low- and high-frequency-tuned receptors are not significantly different (Pollack and Faulkes, 1998). More recent work shows that spikes from one subset of low-frequency-tuned receptors arrive at the central nervous system 2–3 ms later, on average, than those of other receptors, including those tuned to ultrasound (Imaizumi and Pollack, 2001), but the magnitude of the delay is too short to account for the difference in the latency of ON1, which may be as much as 10 ms. The longer latency of ON1 to low-frequency sound stimuli is evident not only at the level of spike timing, but also in the onset of EPSPs (Figs 2, 3). We found no evidence for appropriately timed postsynaptic inhibition that might explain the delay in EPSP onset (Fig. 2). Thus, the locus of the delay in the response of ON1 is, at least in part, presynaptic to ON1. PTX-resistant presynaptic inhibition (Fig. 1D) and frequency-specific polysynaptic pathways (Fig. 1E) remain the two most likely possibilities.

It is known that GABA-immunoreactive processes are present in the cricket auditory system (Watson and Hardt, 1996; Watson and Hardt, 1999). Given that PTX is generally an effective blocker of GABA receptors (Sattelle et al., 1991), our PTX experiments (Fig. 4) should have revealed any GABA-mediated inhibition of ON1, whether pre-or postsynaptic. Nonetheless, we cannot rule out the possibility that the long latency of ON1 to low-frequency sound is caused by PTX-resistant inhibition. First, some GABA receptors in insects are PTX-resistant (Amat and Hue, 1997). Second, there is evidence for non-GABAergic inhibition in the cricket auditory system. Specifically, although ON1 is known to be an inhibitory interneuron, it is not immunoreactive for GABA nor does it contain typical GABA synaptic vesicles (Watson and Hardt, 1996). Skiebe et al. (Skiebe et al.. 1990) proposed that histamine is the inhibitory neurotransmitter of ON1 because histamine mimicked the inhibitory effects of ON1 on AN2, and histamine antagonists blocked inhibition of AN2. The latency of ON1 (Fig. 4) and the spike count of ON1 in response to sound are affected only slightly by PTX, suggesting that most of its abundant GABA-immunoreactive input synapses (Watson and Hardt, 1996) are activated by non-auditory input (e.g. subgenual organ, Wiese, 1981; walking motor system, Schildberger et al., 1988).

The balance of evidence suggests that, in addition to a monosynaptic afferent input to ON1 (Hirtz and Wiese, 1997; this study), there is also a polysynaptic path, specific to 4.5 kHz. The most telling evidence for a polysynaptic pathway is that very brief 5 kHz stimuli generate two EPSP peaks, approximately 12 ms apart (Figs 8, 9; Table 1), which is similar to the latency difference of ON1 spikes to low- and high-frequency stimuli (Pollack, 1994; Faulkes and Pollack, 2000). Multiple compound EPSP peaks are difficult to explain if only monosynaptic connections are present and receptor neurons have similar conduction velocities (Pollack and Faulkes, 1998). Brief ultrasound stimuli elicited only a single peak, even though, at high intensity, ultrasound stimuli may excite both low-frequency-tuned receptors and those tuned to ultrasound (Imaizumi and Pollack, 1999), which might be expected to evoke a second, delayed EPSP in ON1. Two factors might explain the absence of this second peak. First, the sensory input that ultrasound provides to low-frequency receptors is weak compared with that provided by low-frequency stimuli and, as a result, relatively few receptors might be expected to respond, causing a relatively small, delayed EPSP. Second, this delayed synaptic input would arrive on the falling phase of a much larger, ultrasound-induced EPSP, when it would be attenuated by both a decreased driving force and an increased membrane conductance. In contrast, the early EPSP elicited by intense low-frequency stimuli is relatively small and does not obscure the delayed peak. The relatively small size of this peak, at least at low-to-moderate sound intensities, may also explain why, despite the ability of low-frequency stimuli to activate a fast pathway to ON1, the latency of ON1 is long at these frequencies: the early EPSP may simply not be large enough to bring ON1 to threshold, at least over the intensity range that we have studied.

The possibility of a polysynaptic pathway to ON1 is consistent with recent findings on the anatomy of receptor neurons. The axon terminals of the most frequently sampled 4.5 kHz receptor neurons overlap little, if at all, with the dendrites of ON1 (Imaizumi and Pollack, 1996; Pollack and Imaizumi, 1999; Imaizumi and Pollack, 2001). If these receptor neurons provide input to ON1, then it seems probable that they do so via a polysynaptic pathway.

An implication of a polysynaptic pathway, of course, is that one or more low-frequency-tuned interneurons are interposed between low-frequency receptors and ON1. We speculate that these putative interneurons are nonspiking. If low-frequency input to ON1 were channelled through nonspiking interneurons, this might account for the difference in shapes of EPSPs evoked in ON1 by low and high sound frequencies. Graded transmitter release from nonspiking interneurons might produce the smooth, graded EPSPs that low-frequency stimuli elicit in ON1, whereas monosynaptic input from high-frequency-tuned receptors results in discrete, unitary EPSPs (Pollack, 1994; Fig. 6). Because they do not fire action potentials, nonspiking interneurons are inconspicuous in both extracellular and intracellular recordings, which might explain why few candidates for these intermediary neurons have yet been identified. Two nonspiking local auditory interneurons are known in the prothoracic ganglion of Acheta domesticus, and both are tuned to low-frequency stimuli (Stiedl et al., 1997).

Financial support was provided by operating grants from the Natural Sciences and Engineering Research Council (NSERC) and the Whitehall Foundation to G.P. We thank Ronald Chase, Michael Landolfa and two anonymous referees for criticising earlier versions of the manuscript.