Multielectrode array use in insect auditory neuroscience to unravel the spatio-temporal response pattern in the prothoracic ganglion of Mecopoda elongata Free

Auditory information from the ears needs to be processed in the central nervous system to recognize the sound source and its location (Pollack, 2000; Theunissen and Elie, 2014). For acoustically communicating insects, sound signals are crucial for finding and choosing mating partners. For that, two major tasks for the hearing system are: (1) sound frequency and temporal pattern analysis for signal recognition (‘what’) and (2) evaluating directionality of the signal based on binaural time and/or level differences (‘where’). Bushcrickets (katydids), a species-rich family of ensiferan insects, use different acoustic signalling and phonotaxis strategies to find each other (Spooner, 1968; ter Hofstede et al., 2020; Heller et al., 2021). In Mecopoda elongata, the bushcricket species of this study, only the males sing (Nityananda and Balakrishnan, 2006). For sound production, they rhythmically rub their forewings together and thereby produce broadband syllables of approximately 20 ms duration, which increase in sound level during the ∼250 ms long chirps (Hummel et al., 2014). By stereotypically repeating the chirps every few seconds, a group of singing males forms an imperfectly synchronized chorus (Hartbauer et al., 2005, 2006) to attract females. Guided by the acoustic cues, the females phonotactically approach and make their mating choice between the singers (Greenfield, 1994; Siegert et al., 2011; Hartbauer et al., 2015). In this mate-finding system of female phonotaxis towards chorus-singing males, processing of auditory information is crucial for both sexes (Hartbauer et al., 2014; Hartbauer and Römer, 2016).

Sound signals are transduced into auditory neuronal information by sensory cells in the bushcricket's ears located in their foreleg tibia. The receptors in the crista acustica are especially involved in the transduction process of airborne sound (Hummel et al., 2016; Vavakou et al., 2021). In M. elongata, only approximately 45 sensory cells of the crista acustica transduce sound from approximately 4 to 80 kHz in a tonotopic manner (Palghat Udayashankar et al., 2012, 2014; Hummel et al., 2017). In response to sound stimulation, these frequency-specific bipolar sensory cells forward action potentials along their axons into the ventral neuropil of the prothoracic ganglion, the first stage of synaptic processing in the auditory pathway (Schul and Sheridan, 2006; Baden and Hedwig, 2010; Stumpner and Nowotny, 2014). After neural integration of auditory information in the prothoracic ganglion, spectral and temporal signal parameters are encoded by a small number of interneurons (Römer et al., 2002; Stumpner and Molina, 2006; Siegert et al., 2011; Kostarakos and Römer, 2015, 2018) to form specific pathways for further processing of behaviourally important aspects such as binaural timing for directional information, or spectral and temporal pattern discrimination for predator avoidance and mate recognition – similar to other ensiferan insects (Hennig et al., 2004; Pollack, 2015; Schöneich et al., 2015; Clemens et al., 2021).

In bushcrickets, the tonotopically organized primary auditory neuropile is located just underneath the relatively flat ventral surface of the prothoracic ganglion, which allows researchers to reliably record synchronous neuronal spiking (compound action potentials) and sum potentials of synaptic activity (local field potentials) from the outside of the ganglion with the multielectrode array, like electroencephalogram (EEG) recordings of the cerebral cortex in vertebrates (see Buzsáki et al., 2012). The prothoracic ganglion houses three different morphological types of auditory output interneurons with axonal projections into other parts of the central nervous system (CNS). The spike responses of ascending auditory interneurons (ANs) are driven by excitatory inputs from the auditory afferents of the ipsilateral ear, and they also receive inhibitory synaptic inputs from local auditory neurons in the prothoracic ganglion (Stumpner, 1998, 2002). By their ascending axon they forward auditory information to the brain (Stumpner, 1998; Stumpner and Molina, 2006; Kostarakos und Römer, 2018). Their responses to sound were described as tonic spiking without considerable adaptation (Schul, 1997). In Mecopoda, at least two ANs in each body side have been described: one is narrowly tuned to low frequencies and the other shows broadband tuning over a wide frequency range (Kostarakos and Römer, 2015, 2018). Similar to ANs, the T-shaped neurons (TNs) are bilateral mirror-image pairs of prothoracic first-order auditory interneurons that transmit auditory information up to the brain by an ascending axon, but also down to more posterior thoracic and abdominal ganglia by a descending axon. TNs show a broadband frequency tuning and rather phasic spike response that exhibits a strong adaptation effect (Schul et al., 2014). Other studies have found a long-lasting inhibition in the TN spiking (e.g. Siegert et al., 2011). In extracellular recordings of the neck connective with hook-electrodes, especially the TN-1 responses can be reliably detected by their large spike amplitudes (e.g. Faure and Hoy, 2000; ter Hofstede and Fullard, 2008; Siegert et al., 2011). However, a second type of TN with a very narrow low-frequency tuning was also described more recently in M. elongata (Kostarakos and Römer, 2015). The third morphological type of auditory output interneurons of the prothoracic ganglion are the descending interneurons (DNs), which forward their information by a posteriorly projecting axon to the mesothoracic ganglion and beyond. These are described as broadband tuned neurons with mainly tonic response patterns (Stumpner and Nowotny, 2014). However, in M. elongata, at least two different types of DNs that process auditory information are described in each body side: one that is broadband tuned and one with a narrow low-frequency tuning (Kostarakos and Römer, 2015, 2018).

The auditory processing within the bushcricket's prothoracic ganglion is orchestrated by local interneurons such as the omega neuron (ON) pair and a group of dorsal unpaired median (DUM) neurons. In at least two species (Ancistrura nigrovittata and Isophya rossica), additional local auditory neurons, named segmental neurons, have been described in the prothoracic ganglion but have not yet been investigated in detail (for review, see Cillov and Stumpner, 2022). ONs are best studied in field crickets (for reviews, see Pollack and Hedwig, 2017; Schöneich, 2020), which have two pairs in the prothoracic ganglion. ON1 is monosynaptically excited by afferents from the ear ipsilateral to its cell body, whereas ON2 receives inputs from both ears (Wohlers and Huber, 1982). ON1 is mutually inhibitory with its mirror-image counterpart of the other body side (Selverston et al., 1985) and inhibits also the two ascending auditory neurons AN1 and AN2 that receive their excitatory input from the contralateral ear (Horseman and Huber, 1994; Faulkes and Pollack, 2000). Ultrastructural evidence revealed inhibitory synapses onto the axon terminals of auditory afferents (Hirtz and Wiese, 1997; Hardt and Watson, 1999) for presynaptic inhibition (Watson, 1992; Poulet, 2005). Comparative studies suggest a similar functional circuitry such as the ON1 pair in crickets for auditory contrast enhancement by the ON pair in the prothoracic ganglion of bushcrickets (Römer et al., 1988; 2002; Schul, 1997; Molina and Stumpner, 2005; Hartbauer et al., 2010). More recent studies investigated the auditory responses of prothoracic DUM neurons in the bushcricket species Ancistrura nigrovittata (Lefebvre et al., 2018; Stumpner et al., 2019). These auditory DUM neurons are considered to form a ‘filter bank’ for carrier frequency, and inhibit each other as well as other auditory interneurons in the prothoracic ganglion (Stumpner et al., 2020). Auditory DUM neurons that influence the sound-induced spike pattern of ANs and others that modulate the temporal component of the spiking patterns in TNs have been described recently (Lefebvre et al., 2018; Stumpner et al., 2019).

The neuronal elements of the auditory pathway of bushcrickets have been investigated by numerous studies over the last few decades (for reviews, see Hennig et al., 2004; Pollack and Imaizumi, 1999; Stumpner and Nowotny, 2014; Strauß, 2019; Cillov and Stumpner, 2022). The auditory interneurons of the prothoracic ganglion have been individually identified and characterized physiologically and morphologically by intracellular single cell recordings and subsequent dye injection through the recording electrode (e.g. Stumpner, 1997; Baden and Hedwig, 2010; Kostarakos and Römer, 2018; Bayley and Hedwig, 2023), and extracellular hook electrode recordings from the neck connective have been used to analyse the auditory spike responses that are forwarded to the brain (e.g. Rheinlaender et al., 1972; Schul, 1997; ter Hofstede and Fullard, 2008; Siegert et al., 2011). Because every recording method has its specific advantages but also limitations, it is still not fully understood in detail how this rather simple neural network of the prothoracic ganglion processes auditory information for signal recognition and localization. Therefore, here we combined different classical electrophysiological techniques such as intracellular single cell recordings of sensory neurons of the ear and extracellular recordings of axonal spikes in the neck connective. Additionally, to record the auditory input and output of the ganglion, we used a 32-channel multielectrode array that covers the entire ventral surface of the prothoracic ganglion. To our knowledge, this is the first study using this electrophysiological recording technique in an insect (reviewed in Bhavsar et al., 2016) to investigate the spatio-temporal response patterns and network dynamics of the first stage of auditory processing using a more holistic approach.

For our neurophysiological investigations, we used male and female adult individuals of middle-aged bushcrickets (Mecopoda elongata; ‘species S’, Sismondo, 1990, Insecta, Orthoptera, Tettigoniidae). These tropical bushcrickets can be kept under constant temperature (∼24°C) and humidity (∼60%) over the entire year without any seasonal drops in the population size. All aspects of the research complied with protocols that adhered to the legal requirements of animal protection in Germany. Intracellular neuronal recordings in the crista acustica were performed in 27 individuals. In 13 animals, we extracellularly recorded the neuronal activity with a multielectrode array attached to the ventral surface of the prothoracic ganglion and the ipsilateral neck connective with hook electrodes. With the terms ‘ipsilateral’ and ‘contralateral’ we refer to the body side of the animal where the speaker for sound stimulation was located.

At the beginning of the experiments, the animal was briefly (approximately 30 s) anaesthetized with CO₂ for preparation. The wings, hind and middle legs were removed before the animal was fixed to a Y-shaped metal holder with a mixture of beeswax and resin. For intracellular measurements in the crista acustica, the animal was fixed dorsal side up and the tibia of the left foreleg was mounted in a stretched position inside a chamber filled with Ringer's saline for insects (modified after Fielden, 1960: NaCl 7.5 g, KCl 0.1 g, NaHCO₃ 0.2 g, sucrose 3 g, CaCl₂ 0.2 g, Aqua dest. 1 l; pH 7). A small window was carefully cut at the dorsal site of the tibia to gain access to the cap cells of the crista acustica (for more details, see Hummel et al., 2014). Because the tympana were surrounded by fluid during the intracellular recordings, mechanical and consequently also neuronal responses of the ear are less sensitive and the response thresholds for sound stimuli are approximately 15±5 dB SPL higher (Fig. S1). For hook electrode recordings at the neck connective and recordings of the prothoracic ganglion with the multielectrode array, the animals were fixed ventral side up on a plasticine block and a small window was cut into the sternum of the prothorax and the ventral cuticle of the neck to obtain access to the CNS structures of interest. In these experiments, the forelegs were fixed on both sides of the animal holder by small metal clamps, and the response thresholds to sound stimulation reflect the natural conditions as the tympana of the ears were surrounded by air (Fig. S1).

The experiments cover a range of different methods and measurement locations along the auditory pathway. We started by investigating the neural responses to tonal sound pulses in the crista acustica of the ear and then repeated the same stimulus protocol with a multielectrode array attached to the ventral surface of the prothoracic ganglion and hook electrode recordings of the ipsilateral neck connective. Spiking activity of individual sensory neurons in the crista acustica were measured by intracellular cap cell recordings as described earlier (Oldfield, 1982; Scherberich et al., 2016, 2017). Animals were placed dorsal side up in the setup and individual cap cells were pierced with sharp borosilicate glass electrodes (GC100F-10, Harvard Apparatus, Holliston, MA, USA) fashioned by a Sutter electrode puller and filled with 1.5 mol l⁻¹ potassium acetate (electrode resistance: 60–100 MΩ). A silver wire was used as a reference electrode and inserted into the saline in the chamber that holds the foreleg. For extracellular hook and multielectrode recordings, bushcrickets were placed ventral side up in the setup. The flat foil of the multielectrode array (FlexMEA 36-OM+µPA32l, MultiChannel Systems MCS GmbH, Reutlingen, Germany) that holds 32 recording channels, two reference channels and two ground electrode channels was positioned on top of the prothoracic ganglion (Fig. 1). The extracellular measured amplitudes of the neuronal responses are in the range of µV and depend on the contact between the electrode foil and the ganglion. Within an experiment there was some desiccation that led to a trend of increased amplitudes (∼10% in 30 min) because of a better attachment of the foil. There was also a variation in amplitudes between different animals (Fig. 1D, green shadow) that depended mostly on the haemolymph production in response to the dissection. More haemolymph production leads to a weaker contact between electrode foil and ganglion. In this first measurement situation, all neural input and output connections of the ganglion were intact (intact condition: binaural hearing). After the first set of measurements, the input nerve from the contralateral ear to the prothoracic ganglion (T1-N5B; see Strauß, 2017) was cut by cutting off the contralateral leg and the set of measurements was repeated (contralateral-cut condition: monaural hearing). Data were only included when both measurements (intact and contralateral cut) were completed. When all ganglion measurements were finished, the ipsilateral neck connective was hooked onto a silver wire (0.25 mm) and cut anterior to the recording position. After establishing a stable recording, the connection between hook and nerve was insulated with petroleum jelly to prevent tissue desiccation.

Generated by a data acquisition board (DAP 5216a board, Microstar Laboratories, Bellevue, WA, USA), we used pure-tone stimuli from 2 to 62 kHz to induce neuronal activity in the crista acustica and the ascending auditory pathway towards the brain. Combinations of the different sound frequencies and stimulation levels (modified by a programmable attenuator, PA5, TDT PC1, Tucker-Davis Technology, Alachua, FL, USA) from 30 to 80 dB SPL were used to measure the frequency tuning of sensory neurons in the ear and neuronal responses in the prothoracic ganglion. Amplified stimuli (RB-850, Rotel, North Reading, MA, USA) were fed into the calibrated loudspeaker (R2904/700,000, ScanSpeak, Vidbæk, Denmark). Pure-tone stimuli of 20 ms duration (about the duration of a single syllable in the conspecific chirp of M. elongata) or 800 ms (0.5 ms rise and fall time) were randomly presented five times at a stimulus interval of 250 ms for cap cell recordings in the ear. All neuronal data were band-pass filtered to separate spiking activity (300–3000 Hz). Local field potentials were band-pass filtered (1–100 Hz) from the extracellular ganglion recordings with the multielectrode array. Data acquisition and evaluation of neuronal responses were performed using custom-made software in Delphi (Borland, Austin, TX, USA) and MATLAB (v. 7.6.0, The MathWorks, Inc., Natick, MA, USA). To detect action potentials from intracellular recordings in the crista acustica, a threshold was set above the noise level of the filtered recording. Spikes were clustered and sorted using the DBSCAN algorithm (Ester et al., 1996). We assumed that there was no effect of the placement of the electrode inside the cell on spike timing because simultaneous recordings in the cap cell and corresponding sensory cell somata showed no differences in spike timing (Oldfield and Hill, 1986).

For the extracellular recordings (hook electrode and multielectrode recordings) we used a stimulus duration of 20 ms (0.5 ms rise and fall time) and 500 ms stimulus intervals between each stimulus. To collect tuning curves from neck connective recordings, we used randomly presented pure tones from 2 to 62 kHz with a rise and fall time of 4 ms and interstimulus interval of 210 ms. Previous measurements by Triblehorn and Schul (2013) in the bushcricket Neoconocephalus triops showed that this stimulus rate is associated with constant spike numbers of the recorded TN activity. Analysis of frequency tuning for threshold curves included a spike waveform analysis. Thresholds were calculated using 30% of maximal spike activity as a reference point. The sound frequency with the threshold minimum (characteristic frequency) of intracellularly measured cap cell response and extracellular neck connective recording were determined from each individual measurement. Peri-stimulus time histograms (PSTHs) with 0.5 or 1 ms bin widths were used to group spike numbers of 100 to 600 stimulus repetitions.

Triggered by a TTL pulse, the same stimulation was used for the multielectrode recordings. Raw signals of the 32 different recording channels were collected using the ME32-FAI-µPA System (Multi Channel Systems MCS GmbH, Reutlingen, Germany). Local field potentials were used as an indicator of local synaptic activity, pointing to dynamic changes in the electric field around a population of responding neurons. Therefore, it should be noted that the local field potentials depend on multiple factors, mainly on the geometry of the current sources and the distance to the recording side (Spira and Hai, 2013). It is complicated, as currents themselves can add up or cancel out and produce field potentials that are either local or remote (Herreras, 2016). Extracellular measured spiking reflects axonal action potentials of individual interneurons or compound action potentials in case of synchronous spiking of several neurons at the same electrode position. To define a spike, a threshold was set above the noise level of the filtered responses. Timing and amplitude of first spikes were extracted for each repetition from these responses. Data were calculated as medians over 100 repetitions for each channel, and the median over 11–13 animals (N) is shown with interquartile ranges (IQRs) or plotted as contour plots with MATLAB. To determine the onset of local field potentials, a detection threshold was set to RMS+4×s.d. within the averaged trace, and the time point when the voltage trace raised above this threshold was measured within a 25 ms window after stimulus start. The largest absolute voltage deflection (maximum or minimum) within a 30 ms time window after stimulus onset was defined as the local field potential amplitude. It was obtained from the voltage traces that were shifted to zero as a baseline by subtracting the mean value of the first 40 ms. Data were calculated as medians for each channel over 8–13 animals (N) with IQR, or plotted as contour plots with MATLAB. For the contour plots, medians of each channel were calculated from at least five animals for the majority of channels, except for channels 2, 10, 26, 28, 30 and 32 (N=4), channels 21 and 31 (N=3) and channel 11 (N=2) in the intact condition, and channel 1 (N=2) and channel 10 (N=1) for the contralateral-cut condition. Recording channels with signals below the noise threshold (RMS+4×s.d.) are not included and are shown as grey areas. The area under the curve was calculated for 20 and 50 ms bins from the signal-averaged traces that were shifted to zero as a baseline by subtracting the mean value of the first 40 ms in order to correct for the offset. To quantify the amplitude changes over time in more detail, the area under the curve was calculated separately for positive and negative voltage trace parts of the local field potentials. For the calculation of the positive area under the curve, all negative parts of the trace were set to zero and vice versa for negative areas under the curve. Then, medians were calculated over at least five animals for the majority of channels, except for channels 2, 10, 26, 28, 30 and 32 (N=4), channels 21 and 31 (N=3) and channel 11 (N=2) in the intact condition, and channels 1 and 10 (N=1) in the contralateral-cut condition, and plotted as contour plots. Recording channels with signals below the noise threshold (RMS+4×s.d.) are not included and are shown as grey areas.

The sensory cells of the crista acustica project from the foreleg ear into the prothoracic ganglion. Backfill staining of sensory neuron axons from the ear enables us to reconstruct the projection area of auditory afferents inside the prothoracic ganglion. For this, we used a microscalpel to carefully remove the hard cuticle dorsal to the tympanal organ to access and cut the auditory nerve at the point where it is leaving the ear. The proximal nerve trunk was then surrounded by a small well of petroleum jelly to be filled with 2% neurobiotin tracer (Vector Laboratories, Burlingame, CA, USA) dissolved in distilled water to backfill the cut axons of the sensory neurons. After 24–48 h in a humid chamber at 4°C to allow the tracer to diffuse along the leg nerve, the prothoracic ganglion was excised, fixed for 2 h in 4% paraformaldehyde and then further processed following a conventional protocol (Schöneich et al., 2011) to visualize neurobiotin with avidin-Cy2 (Dianova, Hamburg, Germany) in whole-mount preparations using confocal laser-scanning microscopy (TCS SP5, Leica, Wetzlar, Germany).

Data are presented as medians and IQR. Statistical analysis, such as ANOVA, was performed using MATLAB and R (v. 4.0.3) with the ‘dabestr’ package (Data Analysis with Bootstrap Estimation in R, v. 0.3.0). For the comparative analysis of the intact condition versus the contralateral-cut condition, only datasets with at least seven active channels were used, and bootstrapping analysis was performed for each channel separately in R. In the resulting effect plots, the difference between the medians of the intact condition and the contralateral-cut condition defines the effect size, and the measured total effect is interpreted as a significant change at an alpha level of 0.05 when the 95% confidence interval does not overlap with zero.

At their characteristic frequency, the sensory neurons of the crista acustica in the hearing organ typically responded with at least six spikes (22 of 50 recordings) to a 20 ms pure-tone pulse of 80 dB SPL (Fig. 2). As an extracellular monitor of the sensory neuron spiking activity, we intracellularly recorded from the corresponding cap cell on top of the sensory dendrite tip, which inversely reflects the action potentials generated by the dendritic region (Fig. 2A; Hummel et al., 2016). Stimulation with unnaturally long sound pulses (800 ms pulse duration and 600 repetitions) revealed that a very precise spike timing occurred especially for the first six spikes of the response, which usually showed a strong phasic initial part followed by rather tonic spiking until the stimulus stopped (Fig. 2B). This initial burst of approximately six spikes with precise timing is very similar to the typical response to a 20 ms pulse of pure-tone stimulation (Fig. 2A).

Systematic intracellular probing of the neuronal spike pattern along the crista acustica demonstrates a tonotopic distribution of sound-evoked response. Sensory neurons with a low characteristic frequency are located in the proximal part of the crista acustica, and with high characteristic frequency in the distal part of the crista acustica (Fig. 2C, left panel). The most sensitive single cell responses were found in the medial part of the hearing organ. The largest spike numbers with the most precise spike timing in response to pure-tone stimuli pulses at the characteristic frequency (n=600 repetitions; 20 dB SPL above the characteristic frequency level) were also found in the medial crista acustica region (Fig. 2C). The timing of the response onset measured at 20 dB above the threshold at the characteristic frequency did not differ significantly between the sensory neurons along the crista acustica, with proximal 2.2 ms (median, IQR: 2.1 to 2.3 ms, N=11), medial 2.4 ms (median, IQR: 2.2 to 2.5 ms, N=24) and distal 2.3 ms (median, IQR: 2.2 to 2.5 ms, N=15). However, as illustrated in Fig. 2C (right panels), the interspike intervals varied considerably between approximately 2.8 and 6.5 ms (min.–max., N=204). For each recording position along the crista acustica, the median spike interval slightly increased for consecutive spike pairs. The decrease in timing between the fifth and sixth spike is associated with a decrease in the number of spikes (Fig. 2C, right panel). Spike latencies and intervals will be longer near the receptor threshold (Mörchen et al., 1978), but in this study, for timing comparison from the ear towards the brain, the stimulation was always well above threshold.

Auditory information from the ear must pass through the prothoracic ganglion, which is the first stage of synaptic processing (Fig. 3) in the auditory pathway towards the brain. In the ventral neuropil of the prothoracic ganglion, axonal terminals of the sensory cells intermesh and synapse with the dendritic branches of first-order auditory interneurons (Fig. 3A,B). The large-amplitude spikes of the TN interneurons are the most prominent auditory responses in the extracellular connective recordings (Suga and Katsuki, 1961). The spikes of ANs, however, were usually too small in amplitude to be reliably detected in our neck connective recordings. In response to tonal stimulation, the tuning of the TN that projects to the brain covers a wide frequency range, with a characteristic frequency at 16 kHz (inset in Fig. 3A). Overall, the TN responses to sound stimuli were significantly lower in spike number and less precise in timing compared with the sensory neurons (Fig. 3C,D). In response to the 16 kHz (80 dB SPL) sound pulse, the first spike of the TN responses occurred 16.6 ms (IQR: 15.5 to 18.1 ms, N=13) after stimulus onset. We measured a response delay of 18.8 ms (IQR: 16.3 to 20.6 ms, N=13) for the first spike at 5 kHz and 16.8 ms (IQR: 15.8 to 18.7 ms, N=13) at 40 kHz. However, the response latency differences were statistically not significant (ANOVA, P=0.2).

To better understand the signal processing in the prothoracic ganglion, we used extracellular recordings with a 32-channel multielectrode. For stimulation at 16 kHz and 80 dB SPL, the strongest responses of the local field potentials were measured in the projection area of auditory afferents, the anterior medial region of the prothoracic ganglion (Fig. 4A; Movie 1). Local field potentials at these recording positions (ch12 and ch22) were mostly positive (relative to the ground channel), with amplitudes of 54 µV (ch12, IQR: −9 to 68 µV, N=12) on the ipsilateral side and 46 µV (ch22, IQR: −49 to 64 µV, N=13) on the contralateral side of the ganglion midline (Fig. 4A, upper left). The local field potentials at these recording positions also exhibited the earliest response timing with delays of 8.0 ms (ch12, IQR: 7.1 to 8.3 ms, N=12) on the ipsilateral side and a slightly more delayed response on the contralateral side with 8.8 ms (ch22, IQR: 7.5 to 9.1 ms, N=13; Fig. 4A, lower left). The first spike activity occurred also in the anterior medial ganglion region at 11.7 ms on the ipsilateral (ch12, IQR: 11.1 to 12.5 ms, N=13) and 11.9 ms on the contralateral side (ch22, IQR: 11.2 to 12.5 ms, N=13) (Fig. 4B, lower left). This spike activity may be related to the action potentials of first-order auditory interneurons because the local field potential starts 3.7 ms (ch12) and 3.1 ms (ch22) before. The amplitudes of these spikes were 277 µV ipsilateral (ch12, IQR: 197 to 477 µV, N=13) and 279 µV contralateral (ch22, IQR: 237 to 344 µV, N=13) of the ganglion midline (Fig. 4B, upper left).

When the contralateral auditory input to the prothoracic ganglion was removed, the location of the earliest local field potential shifted even more to the ipsilateral side, and at the contralateral side increased the response latency (Fig. 4A, lower right; Movie 2). The earliest local field potential occurred now in channel 5 with an 8.5 ms delay (ch5, IQR: 7.9 to 10.5 ms, N=8). Specifically, in the contralateral-cut condition, our recordings revealed significant differences in the response timing in channel 12 on the ipsilateral side and channel 29 on the contralateral side (bootstrapping analysis, P<0.05, Fig. 4A, lower right, asterisks). There, local field potentials started at 9.3 ms (ch12, IQR: 7.9 to 9.6 ms, N=12) and 11.8 ms (ch29, IQR: 11.6 to 13.9 ms, N=13), which is significantly later than in the intact condition. In the contralateral-cut condition, positive local field potential amplitudes occurred only in the anterior ipsilateral quarter of the ganglion (Fig. 4A, upper right). Between the intact and contralateral-cut condition, there were significant differences in channel 7, where the amplitudes increase to 40 µV (ch7, IQR: 24 to 70 µV, N=12, bootstrapping analysis, P<0.05, Fig. 4A, upper right, asterisk), and in channel 22, where the amplitude decreases to −75.4 µV (ch22, IQR: −118.8 to −37 µV, N=11, bootstrapping analysis, P<0.05; Fig. 4A, upper right, asterisk). A significant decrease in amplitude can also be seen in channels 25, 27 and 29 (bootstrapping analysis, P<0.05; Fig. 4A, upper right, asterisks). So, local field potential amplitudes became more negative on the contralateral side in the anterior part of the prothoracic ganglion, and more positive on the ipsilateral side, resulting in positive values in the anterior contralateral part of the intact–contralateral cut difference plot (Fig. 4C, top). However, the posterior part of the prothoracic ganglion does not seem to be affected by cutting off the contralateral auditory input.

According to the shift of the local field potential timing, the position of the shortest timing of the spikes also shifted in the contralateral-cut condition towards the ipsilateral side (Fig. 4B, lower right). Specifically, the ipsilateral anterior quarter of the ganglion showed the earliest spike timing, with a response delay of 11.6 ms (IQR: 11.0 to 13.1 ms, N=13) in channel 3. Close to the anterior midline of the ganglion, the spike timing was ipsilateral 11.6 ms (ch12, IQR: 11.0 to 12.4 ms, N=13) and contralateral 14.0 ms (ch22, IQR: 11.4 to 15.7 ms, N=13). A significant difference in spike timing between the intact and contralateral-cut conditions can only be seen in the anterior ipsilateral quarter of the array, where the spike response time in channel 6 decreased from 14.9 ms (IQR: 11.9 to 15.4 ms, N=13) to 12.3 ms (IQR: 11.1 to 15.4 ms, N=13, bootstrapping analysis, P<0.05, Fig. 4B, lower right, asterisk). In the contralateral-cut condition, the largest spike amplitudes occurred at the ipsilateral side, but in a smaller area of the anterior medial part of the ganglion (Fig. 4B, upper right). The strongest effect from cutting off the contralateral input can be seen in channel 22, where the amplitude decreased from 280 µV (IQR: 237 to 344 µV, N=13) to 210 µV (IQR: 166 to 279 µV, N=13, Fig. 4B, upper right). Taken together, and similar to the local field potential amplitudes, the spike amplitudes also became smaller at the contralateral side and larger at the ipsilateral side after the contralateral input was cut off. This resulted in positive values in the contralateral anterior part of the ganglion in the difference plot (Fig. 4C, bottom), which indicates the contribution of the contralateral input to the binaural processing.

To analyse the temporal changes in local field potential activity in more detail, we calculated from our original local field potential data the area under curve for consecutive time windows starting from 20 ms before the sound stimulation (Fig. 5). For example, the complex amplitude changes over time in the local field potential traces recorded from channels 12 and 22 differed clearly between the intact and contralateral-cut conditions (Fig. 5A). Before stimulation, the positive as well as the negative areas under the curve were smaller than ±1 µV at all recording channels in both conditions. During the 20 ms of sound stimulation, the local field potential activity started and peaked in the negative area under the curve response (peak in bin 0 to 20 ms; 20 ms bin width) in the middle area of the ganglion before it peaked in the positive area under the curve response (peak in bin 40 to 60 ms) in the area of afferent projections (anterior medial region of the prothoracic ganglion) in the intact condition (Fig. 5B). Most importantly, the positive area under the curve response peaked after the stimulation and the increased positive area under the curve values continued for several hundreds of milliseconds, which is about 10 times longer than the stimulus duration. This long-lasting increase in local field potential amplitude was also visible at different sound pressure levels at 10 kHz stimulation (Fig. S2). In the contralateral-cut condition, the positive area under the curve response also starts later and lasts longer than the negative response (Fig. 5C). First, we compared changes in the positive area under the curve response between intact and contralateral-cut conditions for 20 ms sound stimuli. When intact and contralateral-cut conditions were compared with each other during and shortly after sound stimulation (bin 0 to 20 ms and bin 20 to 40 ms), owing to loss of the contralateral input, positive area under curve increased significantly on the ipsilateral side in channels 6 and 7 (Fig. 5C, asterisks; see Table S1 for ch7 values). The increase in local field potential amplitude is also visible in the example trace of channel 12 in Fig. 5A. However, on the contralateral side, the positive area under the curve decreased significantly owing to loss of contralateral excitation in channels 20, 22, 27 and 29 (Fig. 5C, asterisks; see Table S1 for ch22 values). During the time bin 40 to 60 ms, the positive area under the curve significantly decreased on the ipsilateral side in channels 3, 7 and 14 (Fig. 5C, asterisks; see Table S1 for ch3 values). On the contrary, the positive area under the curve within this time bin increased significantly on the contralateral side in channel 25 (Fig. 5C, asterisk; see Table S1 for ch25 values). Taken together, maximal positive local field potential activity during (bin 0 to 20 ms) and shortly after stimulation (bin 20 to 40 ms) shifted from the medial anterior part to the medial ipsilateral side, and after stimulation (bin 40 to 60 ms) it shifted to the contralateral side. Most important, overall positive activity decreased in the projection area of the prothoracic ganglion in the anterior medial part in the contralateral-cut condition owing to loss of contralateral input.

Next, we compared changes in the negative area under the curve response between the intact and contralateral-cut conditions. Analysis showed a decrease of activity on the ipsilateral side and an increase on the contralateral side during (bin 0 to 20 ms) and shortly after (bin 20 to 40 ms) sound stimulation, respectively. Specifically, the negative area under the curve decreased significantly on the ipsilateral side in channel 13 after stimulation (Fig. 5C, asterisk; see Table S2 for ch13 values). The negative area under the curve values increased significantly in channels 20, 25 and 29 during and 20 ms after stimulation (Fig. 5C, asterisks; see Table S2 for ch29 values). Also, the negative area under the curve increased significantly 20 ms after stimulation in channel 23 (Fig. 5C, asterisk). Taken together, the area of maximal negative local field potential activity enlarged and shifted to the contralateral side, and overall negative local field potential activity increased on the contralateral side and decreased on the ipsilateral side in the contralateral-cut condition.

Using intracellular and extracellular (multielectrode array and hook electrode) recording techniques, we were able to demonstrate the timing of neuronal events in the ear and in the first integration centre, the prothoracic ganglion in bushcrickets, as well as neuronal processing there. In the hearing organ, the crista acustica, sound induces a mechanical response that spreads almost immediately along the <1 mm long crista acustica. With an approximated wave speed of 3–18 m s⁻¹ (Palghat Udayashankar et al., 2012; Sarria-S et al., 2017; Olson and Nowotny, 2019), the mechanical signal reaches the sensory cells with less than 0.3 ms delay between the distal and proximal regions. The spiking of the primary sensory neuron at high sound pressure levels therefore occurs with no significant differences in timing at approximately 2 ms after stimulus onset in all regions of the crista acustica. With intracellular recordings, we found typically four to six precisely timed spikes in response to a 20 ms pure-tone stimulus (80 dB SPL) at the characteristic frequency of the sensory neuron. As shown previously by Baden und Hedwig (2010), long-lasting stimulation leads to a phasic-tonic spiking response that is reminiscent of chopper neurons found in the cochlear nucleus in mammals (Pfeiffer, 1966; Oertel et al., 2011). There, they are supposed to convert phasic excitation from the auditory nerve to tonic firing (Oertel et al., 2011).

The use of multielectrode arrays allow the monitoring of neuronal responses in a network and helps to identify the location of neuronal activity (Buzsáki, 2004). With the planar multielectrode array (32 channel) on top of the prothoracic ganglion as used in this study, we were able to measure spatially restricted extracellular field potentials that correspond to a synchronized neuronal activity in the ganglion. The position of the electrodes that show the maximum of these summed potentials fits with the projection area of primary afferent fibres from the ear at the anterior medial region of the ganglion. Regarding the response timing, we can track the signal processing and detect the start of neuronal responses at approximately 8 ms after stimulus onset. As shown by previous studies, the latency between action potentials in the axonal branches of the sensory neurons in the prothoracic ganglion and the onset of the postsynaptic responses in the connected neurons, such as ON, AN, TN or DUM neurons, are very short and last to approximately 1 ms (Römer et al., 1988; Schul, 1997; Stumpner, 2002). This points to a direct monosynaptic connection. By our multielectrode array measurements, we are not able to assign the local field potentials to either presynaptic or postsynaptic activity based on the amplitude pattern. In general, multielectrode arrays are blind to subthreshold potentials (Spira and Hai, 2013). Here, pharmacological manipulations or intracellular recordings of the different target sides would be necessary to study this in detail. Furthermore, we think that our extracellular measurements from the ganglion outside are not suitable to measure very small and locally restricted presynaptic modulations such as primary afferent depolarizations (Hardt and Watson, 1999; Baden and Hedwig, 2010). Also, the low amplitudes of the field potential (approximately 0.3 mV) point to a low electrical coupling between the active neurons and the array, which is typical for multielectrode recordings (Spira and Hai, 2013). However, we suggest that the measured local field potentials are related to postsynaptic activity, as reported from local field potential recordings in mammals (Bruyns-Haylett et al., 2017). The source of positive and negative local field potentials is still a controversial topic, and the assignment of positive field potentials to passive current loops or active current flow is not possible without single cell recordings (Liu et al., 2012). Beside this difficulty in naming the current source, the response pattern of the local field potentials exhibits a distinct sequence of positive and negative amplitude values that is reminiscent of changes in the graded potential measured in DUM neurons (Lefebvre et al., 2018). The fact that a relatively high number of auditory DUM neurons, at least 15 in the prothoracic ganglion of Ancistrura nigrovittata, show a combination of excitatory and inhibitory postsynaptic potentials in the sound-induced response (Lefebvre et al., 2018) would support our idea for a dominance of the DUM neurons in our measured local field potentials.

We found long-lasting local field potential amplitude increases that lasted more than 100 ms after the end of stimulation. This after-stimulation effect seems not to be related to the DUM neuron responses as no evidence in this direction has been reported so far (Lefebvre et al., 2018; Stumpner et al., 2019, 2020). However, simultaneously recorded responses of both omega neurons (Römer et al., 2002) demonstrate long-lasting effects by suppression of the contralateral cell in the order of 70–130 ms. Especially in M. elongata, the inhibitory activity of these local interneurons is discussed as a main driver in the leader preference of female choice (Römer et al., 2002; Siegert et al., 2011). A short delay in the synchrony between the chirps of two males leads to a strong inhibition of the spiking events measured in the ON in response to the follower signal. In addition, an intrinsic mechanism for activity-dependent long-lasting hyperpolarization that decays back to resting potential with a time course of several seconds has been demonstrated in crickets (Pollack, 1988; Sobel and Tank, 1994; for bushcrickets: Römer and Krusch, 2000). This post-stimulus inhibition by calcium-activated potassium channels closely matches the time course of auditory sensitivity modulation underlying forward masking in response to cricket calling songs, and can be seen as an automatic gain control mechanism in which a loud sound pulse suppresses the response to subsequent sound (Sobel and Tank, 1994).

In our experiments, in which the contralateral leg was cut, we found a decrease in the positive local field potential amplitudes on the contralateral side owing to loss of input from this side. However, at the same time, positive local field potential amplitudes on the ipsilateral side increased after the removal of the sensory input from the contralateral side, which could be explained by a potential loss of contralateral inhibition. Additionally, negative local field potential amplitudes increased on the contralateral side, perhaps because inhibition from the ipsilateral side was still functional. By these local field potential changes, we confirm modulatory processes by local interneurons that are well known in bushcrickets (Römer et al., 2002) and are also identified in other closely related groups such as crickets (Pollack and Hedwig, 2017) and grasshoppers (Hildebrandt et al., 2011). Surprisingly, we found an amplitude increase of positive local field potential amplitudes on the contralateral side after the stimulus presentation ended in cutting experiments. We suggest that this increase in local field potential amplitudes could be related to a late activation of interneurons, such as ON, DUM or segmental neurons (local interneurons that are branching mostly within one hemiganglion).

Spiking activity, extracellularly measured by the multielectrode array, was delayed by approximately 3 ms to the local field potentials, and is therefore related to a postsynaptic activity. There was no difference between the spike occurrence at the ipsilateral and contralateral sides in the intact preparation. We suggest that this is due to the high stimulation amplitude of 80 dB SPL. A previous study on the intensity-dependent latency of AN spiking exhibited differences only up to 65 dB SPL (Stumpner, 2002). However, separation of AN and TN responses by spike amplitudes, as possible in neck recordings (e.g. Clemens et al., 2015), was not possible with the multielectrode array. Therefore, we cannot clearly assign spiking events to specific interneurons. To overcome this uncertainty, future experiments are planned with the multielectrode array below the prothoracic ganglion and intracellular recording from the opposite side to compare the intracellular and extracellular response patterns. Regarding the spike timing, after the cutting of the contralateral input, we measured only small differences in timing that may be related to a change in the local field potential occurrence. Recorded on the neck connectives, spiking activity was largest for the TN sound-induced activity (see Faure and Hoy, 2000). The TN activity in the neck connective occurs at approximately 16 ms after stimulus onset, as found by others (e.g. Krüger et al., 2011). The smaller AN activity was mostly not detectable in our recordings.

To summarize, we were able to show that with the use of a multielectrode array, placed on the top of the prothoracic ganglion of a bushcricket, sound-induced neuronal responses that are temporally and spatially distinct in the 32 different electrodes are measurable. With the great advantage of this technique to simultaneously observe the spatio-temporal dynamics of neuronal activity pattern across the entire ganglion, we monitored changes in the local field potential and spike response pattern to a short tonal pulse. We were able to relate our extracellularly measured response to activities that were previously only measured mainly intracellularly, and were therefore related to single cell responses. Additionally, with this technique, we identified long-lasting amplitude increases of the local field potential response during tonal stimulation, which points to the fact that this rather small neuronal network still includes yet unexplained processing features. Because the communication signal of M. elongata males is a chirp that consists of a sequence of approximately 13 broadband syllables with approximately 20 ms syllable duration and 20 ms syllable intervals, we argue here that during this chirp, the amplitude of the local field potential will be summed and will increase in comparison with the single syllables, presented artificially as single pulses. This feature of the male song structure will be the subject of future projects on neuronal processing in the prothoracic ganglion of bushcrickets.

We thank Steven Abendroth for technical support and Nicole Naumann for help with the histology procedures of the nerve staining preparations.