Introduction

Many moths, including most Noctuoidea, Geometroidea and Pyraloidea, possess paired ultrasound-sensitive ears located on the dorsolateral part of the thorax, at the junction with the abdomen (Noctuoidea), or in the abdomen (Geometroidea and Pyraloidea). Hearing in moths is assumed to have evolved as a specific defense against the predation pressure of insectivorous bats (Roeder 1969). With their auditory sensory neurons tuned to frequencies between 10 and 60 kHz (Roeder 1966; Surlykke et al. 1999; Minet and Surlykke 2003), the moth ear seems to have evolved for the specific purpose of detecting short ultrasonic pulses emitted by echolocating bats (for a review, see Fullard 1998). In spite of the simple moth ear, which has only one or two sensory cells, hearing in this nocturnal insect group aids in controlling complex behavior linked to multi-modal sensory input. For example, bat calls are reported to modulate not only odor-driven behavior (Skals et al. 2005) but also responses from antennal-lobe projection neurons (Anton et al. 2011).

Moth ears have evolved independently several times in different families. However, the hearing organ displays a similar basic structure in most species (Hoy and Robert 1996). The sensory neurons attach directly to a thin tympanic membrane that is stretched over an air-filled cavity and are capable of detecting ultrasound and responding to it. In noctuidea, the ear has two auditory sensory neurons, the A cells and one non-auditory mechano-sensitive B cell (Stumpner and von Helversen 2001; Yager 1999). The auditory A-neurons are broadly tuned to the same frequency range (10–60 kHz) but have different sensitivities. Whereas the A1 cell has a threshold of 30–40 dB SPL at its best frequency, i.e., the frequency with the lowest threshold, the A2 cell requires a stimulus sound pressure of ca. 20 dB higher (Surlykke and Miller 1982; Surlykke et al. 1999). The sensory axons run from the ear to the central nervous system via the IIIN1 nerve. The A2 cell has projections mainly confined to the metathoracic ganglion, whereas the A1 projects upstream and downstream of the ventral cord. Both auditory receptor neurons have connections to the brain via interneurons (Surlykke and Miller 1982; Boyan and Fullard 1986; Boyan et al. 1990). Boyan and Fullard (1986) found six types of auditory interneurons ascending at least to the prothoracic ganglion. However, whether any of these interneuron types target the brain is as yet unknown (Roeder 1969).

While recording intracellularly within the lateral protocerebrum of three heliothine moth species in order to identify olfactory neurons, we fortuitously discovered neurons that responded to transient acoustic pulses of broad bandwidth (Zhao et al. 2013). Iontophoretic staining of the neurons suggested the ventro-lateral protocerebrum as a significant brain region for processing acoustic information, since this was the main area of branching. As the sound stimulus was an un-planned event caused by the valve system controlling the application of odor stimuli, the design of the current study had its obvious limitations. However, we analyzed the valve sound pulses post-experimentally and determined their acoustic features, which revealed the presence of ultrasound frequencies. We also stimulated the neurons with a well-known ultrasound source, namely jingling keys (Fig. 1c) in order to confirm the presence of auditory responses.

Fig. 1
figure 1

Measurement of the airflow and sound pulses during the stimulus sequence. a Changes in airflow succeeding the sound pulses (arrows). b Oscillogram showing that the first click (valve 1 being closed) occurs 1 s before the second pulse (opening of valve 2). The third click follows 0.4 s later (valve 2 closing). Finally, re-opening of valve 1 produces the fourth click after a further 1 s. c The spectra taken from a recording of the sound clicks inside the plastic tube show more energy at the first, second and fourth clicks (above 20 kHz) than at the third click. d Amplitude-frequency curves of the four acoustic pulses (measured inside the plastic tube). e Spectrum of jingling keys. Most energy is in the frequency range 15–35 kHz

Full size image

Materials and methods

Insects

All experiments were carried out on male heliothine moths of three species: the tobacco budworm moth, Heliothis virescens, the cotton bollworm moth, Heliothis armigera and the oriental budworm moth, Heliocoverpa assulta (Lepidoptera: Noctuidae). H. virescens pupae originated from a laboratory culture in Basel, Switzerland (Novartis Crop Protection), whereas the H. armigera and the H. assulta pupae came from a laboratory culture of Dr. Jun Feng Dong (Henan University of Science and Technology, China). Emerging moths were placed in separate plexiglass cylinders (height 20 cm, diameter 10 cm) covered by a perforated lid and with a reservoir of 0.15 M sucrose solution. Animals were studied 3–8 days after eclosion.

Preparation

The animal was fixed into a small cylindrical plastic tube with the head exposed and was immobilized by means of utility wax (Kerr). The scales were removed and a window was made in the head cuticle in order to expose the brain. The head was fastened in a position that allowed recordings to be made from the ventro-lateral protocerebral region of the brain. To keep the preparation alive, Ringer’s solution (in mmol l−1: 130 NaCl, 6 KCl, 4MgCl2, 5 CaCl2, 160 sucrose, 25 glucose, 10 HEPES, pH 6.7, osmolarity 500 mOsm; all chemicals from Sigma) was applied regularly. The brain was desheated around the area posterior of the antennal lobes in order to facilitate insertion of the recording electrode into the neural tissue.

Intracellular recording and staining

Glass electrodes consisting of borosilicate glass capillaries with a filament (Hilgenberg, Germany) and pulled with Sutter Instrument Model P-97 (Sutter, USA) were filled with the fluorescent dye dextran tetramethyl-rhodamine, MW 3000 (Molecular Probes, http://www.probes.com) and 0.2 M potassium acetate, resulting in a resistance of 150–600 MOhm. The reference electrode was a chloridized silver wire immersed into the saline solution. The recording electrode was inserted dorsally into an area posterior of the antennal lobe. Movement of the microelectrode was controlled with a Leitz micromanipulator (Leica, Bensheim, Germany). The electrical signals of the neurons were amplified (Axoprobe-1A) before being monitored on an oscilloscope and a computer. The software programme Spike2 4.02 (updated to 6.03 during the experiments; Cambridge Electronic Design 1998–2001) was used for analyzing and storing the data. The sampling rate was 20,000 Hz. After recording, single neurons were iontophoretically stained by passing 10 nA depolarizing current pulses (200 ms duration at 1 Hz) through the recording electrode for 2–7 min. In order to allow neuronal transportation of the dye, the preparation was kept for 2 h at room temperature. The brain was then pre-fixed with 4 % paraformaldehyde in phosphate-buffered saline (PBS; in mM: 684 NaCl, 13 KCl, 50.7 Na2HPO4, 5 KH2PO4, pH 7.4). After 10 min, the brain was dissected from the head-capsule and further fixed in 4 % paraformaldehyde for 1–2 h at room temperature or overnight at 4 °C. After fixation, the brain was dehydrated in increasing alcohol concentrations (50 %, 70 %, 90 %, 96 %, 2 × 100 %; 10 min each) and cleared in methyl salicylate. For confocal scanning, the brains were put on a metal plate (see also Zhao and Berg 2009).

Mass-staining

Mass-staining experiments were performed on three H. virescens males. The moths were pinned out ventral side up. The cuticle surrounding the neck was removed and the exposed ventral nerve cord was cut with scissors. Crystals of a fluorescent dye, a biotinylated substance also known as dextran (Micro-Ruby, Molecular Probes) were picked up on the fine tip of a micro needle and were then applied to the cut end of the connectives. Ringer’s solution was supplied and the preparation was left overnight in a refrigerator (4 °C) for the transportation of the dye. The subsequent procedure of dissection and fixation was as described above. After being fixed in 4 % paraformaldehyde for 1 h at room temperature, the brain was rinsed with PBS. Staining was then intensified by incubating the brain for 2 h in the fluorescent conjugate streptavidin-Cy3 (Jackson ImmunoResearch, West Grove, Pa., USA; diluted 1:200 in PBS), which binds to biotin. Incubation was followed by rinses with PBS and dehydration in an ascending ethanol series (50 %, 70 %, 90 %, 96 %, 2 × 100 %; 10 min each). Finally, the brain was cleared and mounted in methylsalicylate.

Sound stimulation

The sound stimuli originated from two solenoid valves controlling the air flow during odor stimulation. The valves (2-way Direct Lift Solenoid Valves, 01540; Cole-Parmer, Vernon Hills, Il., USA) were mounted on one of the walls of the laboratory, 2 m away from the insect. Valve 1 (normally open) regulated the continuous airflow and valve 2 (normally closed) the odor/air puff. When the odor stimulation system was activated, the two valves produced four separate sound pulses; the first pulse was caused by valve 1 being closed; the second pulse, which appeared 1 s later, by valve 2 being opened; the third pulse occurred 0.4 s thereafter by valve 2 being closed; and the fourth pulse, which appeared after another 1 s, by valve 1 being reopened (Fig. 1; see also Zhao et al. 2013). Activation of the two valves was recorded in the Spike program on a separate channel. We recorded the sound pulses created by opening and closing the valves by using a ¼″ microphone (Bruel & Kjær, type 4138) and preamplifier (Bruel & Kjær type 4939), plus a front end amplifier (Norsonic, Type 1201/30517); we digitized (sample frequency 192 kHz) and stored the signals on a signal analyzer, WINMLS. The sounds were measured with and without air flow, both adjacent to the plastic tube and inside the plastic tube in which the moth was restrained. The recorded sound signals (clicks) were analyzed by using commercial software (BatSound Pro, Pettersson Electronics; Fig. 1). The pulses were of broad bandwidth with energy from audible up to ultrasonic frequencies of 40–50 kHz. The main effect of the plastic was to dampen frequencies above ca. 25 kHz (Fig. 1). Inside the plastic tube, the spectra of all clicks had energy up to above 20 kHz. Thus, the frequency range of the clicks overlapped with the frequency range of 10–60 kHz, to which noctuid moths are most sensitive. The initial pulse amplitude was high, falling off quickly. Because of the decreasing amplitude, we found it difficult to measure pulse length accurately. The duration for which the amplitude was above 50 % of the maximum was 30–40 ms for all pulses. Pulse durations measured as the time containing 95 % of the total energy were ca. 100 ms. The sound pressures of the four pulses were between 71 and 89 dB SPL (peak–peak), which is well above typical moth hearing thresholds of 35–40 dB around 15–30 kHz (Boyan and Fullard 1986; Surlykke and Miller 1982). A bunch of jingling keys was used as an ultrasound source to check the response of the neurons to sound stimuli during the recordings. When the keys were used, the valve system was turned off.

Odor and air puff stimulation

The odor delivery system was as described in Zhao et al. (2013). Two glass cartridges placed side by side were directed at the insect antenna. One cartridge contained a pure filter paper and provided a constant air flow (500 ml/min) blown over the antenna. The other cartridge, which was replaceable, contained a piece of filter paper onto which a particular odor stimulus was applied. As olfactory stimuli, we used the two-component pheromone blends specific for each of the three species (Pheromone chemicals, Plant Research International, Pherobank, Wageningen, Netherlands), plus the plant oil ylang-ylang (Dragoco, Totowa, N.J., USA; for details, see Zhao and Berg 2010; Zhao et al. 2013). We measured the changes in air flow during the time period when the odor stimulation system was activated by using an air velocity meter (VelociCalc model 8355WS), as described in Zhao et al. (2013). The change in air velocity always occurred after the acoustic stimulation (Fig. 1).

Immunocytochemistry

In order to identify protocerebral neuropil structures, immunostaining with antibodies to proteins associated with synaptic terminals was performed on the labeled brains. After analysis of the iontophoretically stained neuron by confocal laser scanning microscopy, the brain was rehydrated through a decreased ethanol series (10 min each) and rinsed in PBS. To minimize non-specific staining, the brain was submerged in 5 % normal goat serum (NGS; Sigma, St. Louis, Mo., USA) in PBS containing 0.5 % Triton X-100 (PBSX; 0.1 M, pH 7.4) for 3 h at room temperature. The preparation was then incubated for 2 days at 4 °C in the primary antibodies, namely anti-SYNORF1 and nc46 (dilution 1:10 and 1:40 in PBSX containing 5 % NGS, respectively). The anti-SYNORF1 was obtained from the Developmental Studies Hybridoma Bank, University of Iowa. The other antibody, nc46, was kindly donated by E. Buchner. After the incubation period, the brain was rinsed in PBS (6 × 20 min) at room temperature before being stained with Cy2-conjugated anti-mouse secondary antibody (Invitrogen, Eugene, Ore., USA; dilution 1:500 in PBSX) for 2 days at 4 °C. Finally, we rinsed, dehydrated, cleared and mounted the brain in methylsalicylate.

Confocal laser scanning microscopy

Whole-mount preparations were viewed with a confocal laser scanning microscope (LSM 510, Carl Zeiss). The intracellular staining, seen as fluorescence of rhodamine/Cy3 (Exmax 550 nm, Emmax 570 nm), was excited by the 543-nm line of a HeNe1 laser, whereas the immunostaining, seen as Cy2 (Exmax 490 nm, Emmax 508 nm), was excited by the 488-nm line of an Argon laser. Objectives used were: 10× water, numeric aperture (NA) 0.45; 20× air, NA 0.5; 40× water, NA 0.8; 63× oil, NA 1.2. We made serial optical sections with an inter-slice distance of 2–4 μm and images were obtained by using an average of four scans per slice in order to reduce noise. Image resolution was set at 1024 × 1024 pixels. We generated three-dimensional (3D) models from the stacks of two-dimensional images.

Digital 3D-reconstructions

The confocal image stacks were carefully examined section by section and stained neurons of particular interest, plus their surrounding brain structures, were manually reconstructed in subsequent confocal slices by means of the visualization software AMIRA 4.1 (Visage Imaging, Fürth, Germany). Thus, the neurons were reconstructed by using the skeleton module of the software and the brain structures by using the segmentation editor. Thus, the neuron was traced to create a surface model built by cylinders of particular lengths and thicknesses, whereas each brain and ventral-cord structure was outlined based on its gray value relative to the background to create a polygonal surface model. With regard to the single-stained neurons, we reconstructed the neuronal structures and the neuropil regions from the same confocal image stacks. In the figure showing mass-staining (see below), the single-stained neurons were transformed into a reference model. Registration of the neurons into the brain atlas followed the same procedure as described by Kvello et al. (2009). When processing the images, brightness and contrast were adjusted by using Photoshop CS (version 8.0; Adobe, San Jose, Calif., USA). The final figures were edited by using Adobe Illustrator CS5. The orientation of all brain structures is indicated relative to the body axis of the insect, as in Homberg et al. (1988).

Analysis of physiological data

The electrophysiological recordings were stored and analyzed by using Spike2 software. During each stimulation, neuronal activity was displayed over a period of 5 s, starting simultaneously with the onset of the first acoustic pulse and ending 2.5 s after the onset of the last acoustic pulse (Fig. 1). A neuron was scored as “sound sensitive” when it displayed brief excitation bursts coinciding with one or several acoustic pulses from the valves. For each neuron, we measured the latency to the first sound pulse from the onset of the stimulus to the peak of the first spike in the response. The means and standard deviations based on the number of recordings are presented in Table 1. Spontaneous activity was set as the spike rate during the third to fifth second (see Fig. 1). The sound response was defined as the spike rate measured for the duration of the acoustic stimulation (i.e., 50 ms). The mean response and standard error to sound pulses were based on the spike rates to the four acoustic stimulations over all recordings and repetitions, respectively. Most neurons responded to the first, second and fourth acoustic stimulation, i.e., calculations are based on three repetitions. Exceptions are neuron vir5, which responded to all four acoustic stimulations (calculation based on all four stimulations and four recordings) and neuron vir9 for which the calculation is based on three recordings including a response to the first acoustic stimulation only. The response strength is the difference between the mean spike frequency during sound stimulation and the mean spontaneous activity. The relative increase in response strength is the ratio of the mean frequency during sound stimulation and the mean spontaneous activity. All physiological data are presented in Table 1. With a few exceptions, we verified the sound-sensitive nature of the neuron by making sure that it responded to the shaking of a bundle of keys. Neurons that did not respond to the keys but to the occurrence of the second sound pulse from the valves, which coincided with the air/odor puff, were considered mechano-sensory/chemosensory and were excluded from further analysis.

Table 1 Physiological and morphological properties of the sound-responding neurons
Full size table

Results

We recorded from approximately 400 neurons in the brains of the three heliothine moth species and found 17 neurons that showed an unambiguous increased firing rate to the acoustic stimuli. Seven of these neurons were successfully stained, six of which were projection neurons ascending from the ventral cord and the seventh a brain interneuron. All the stained neurons had processes in an overlapping region of the ventro-lateral protocerebrum of the brain. In Table 1, which presents an overview of the physiological and morphological data, the individual neurons are named according to the species from which they were recorded.

Morphology

All six examples of ventral-cord neurons that were stained lacked a soma in the brain and were thus determined as ascending neurons. For one of the neurons from H. virescens (vir 1), the soma was found in the posterior region of the contralateral part of the metathoraric ganglion (Fig. 2). In addition to the projections in the ventro-lateral protocerebrum, this neuron had many ramifications in the metathorax and a few in the meso- and prothorax. Four of the remaining ascending neurons, namely three from H. virescens (vir1-3, Figs. 3, 4) and one from H. armigera (arm1, Fig. 5), had similar morphological characteristics in the brain, projecting in the anterior region of the ventro-lateral protocerebrum in which the dense neuronal branches formed a unique structure resembling a fan (Figs. 3, 4, 5, Supplementary movie). Two of the neurons, namely one from H. virescens and one from H. armigera, extended an additional assembly of short processes into the subesophageal ganglion (Figs. 2, 4, 5). The two remaining neurons of this category, both from H. virescens, had none or only weakly visible processes in the subesophageal ganglion. Two neurons ascended from the ventral cord and displayed different morphologies. One of these, vir4, had relatively few branches in the ventro-lateral protocerebrum and additional and more numerous processes in a more medially located brain area (Fig. 6). The other ventral-cord neuron, vir5, innervated the lateral accessory lobe and the ventro-lateral protocerebrum. Because of their weak staining, further details of their neuronal morphology were inaccessible.

Fig. 2
figure 2

Morphological and physiological characteristics of the vir1 neuron. a AMIRA reconstruction showing the soma in the metathoracic ganglion (arrow), plus short neural branches in the ganglia, the subesophageal ganglion (SOG) and the ventro-lateral protocerebrum. b Confocal reconstruction of the whole neuron in a dorsal orientation. c, d Higher magnification images of the neural processes in the SOG (c) and the metathoracic ganglion (d). Note the faint neural processes in the prothoracic ganglion and mesothoracic ganglion (OL optic lobe). Bars 50 μm. e Responses of the neuron to repeated key shaking, plus spontaneous activity shown below (horizontal bar 1 s, vertical bar 10 mV)

Full size image
Fig. 3
figure 3

Morphological and physiological characteristics of the vir2 neuron. a Confocal reconstruction of the whole brain in a dorsal orientation showing the neuronal branches innervating the ventro-lateral protocerebrum (PC) close to the border of the antennal lobe (white line; AL). b Higher magnification image of the neural processes in the ventro-lateral protocerebrum in a frontal view (white line). c–e AMIRA reconstructions showing the neuronal branches in the brain and the axon passing through the subesophageal ganglion (SOG). c Dorsal orientation. d Frontal orientation. e Sagittal orientation (CB central body, Ca calyces, O esophagus, OL optic lobe, P posterior, A anterior, D dorsal, M medial). Bars 100 μm. f Response traces of the neuron showing four excitatory and especially transient responses occurring synchronously with the four acoustic pulses emitted from the valves, indicated by vertical lines below the second recording. The onset/offset of the odor/air puff stimulus is indicated by a horizontal bar below each recording. The neuron also responded strongly to ultrasound from a bunch of jingling keys (horizontal bar 400 ms, vertical bar 10 mV; data from this neuron have previously been published in Zhao et al. 2013)

Full size image
Fig. 4
figure 4

Morphological and physiological characteristics of the vir3 neuron. a The morphology of vir3 is similar to that of vir1 and vir2 in Figs. 2, 3, respectively. b Higher magnification image of the neural processes in the ventro-lateral protocerebrum illustrating their characteristic dense ramifications (white line). c, d AMIRA reconstructions in a frontal (c) and a sagittal (d) orientation. The black line indicates the neural processes in the ventro-lateral protocerebrum (OL optic lobe, O esophagus, SOG subesophageal ganglion, PC protocerebrum, D dorsal, M medial, A anterior). Bars 100 μm. e Response traces of the neuron showing three excitatory and pronounced transient responses occurring synchronously with the first, the second and the fourth acoustic pulse produced by the valves (verticals lines below the recording traces). The onset/offset of the odor/air puff stimulus is indicated by a horizontal bar below each recording (horizontal bar 400 ms, vertical bar 10 mV). f Response to the fourth acoustic stimulation (indicated by the asterisk in e) on a larger time scale (horizontal bar 20 ms, vertical bar 10 mV, Bottom row Interstimulus time histogram (INTH) showing that most spikes occur in quick succession

Full size image
Fig. 5
figure 5

Morphological and physiological characteristics of the arm1 neuron. a Confocal reconstruction of the whole brain in a frontal orientation showing the neuronal branches innervating the ventro-lateral protocerebrum (PC) close to the border of the antennal lobe (white line). b Higher magnification image of the neural branches in a frontal view (white line). c Higher magnification image of the axon (white line) projecting through the suboesophageal ganglion (SOG). d, e AMIRA reconstructions in a frontal (d) and a sagittal (e) orientation (CB central body, OL optic lobe, O esophagus, D dorsal, M medial, A anterior). Bars 100 μm. f Response traces of the neuron demonstrating three or possibly four excitatory and transient responses occurring in synchrony with the acoustic pulses produced by the valves (vertical lines below the second recording). The onset/offset of the odor/air puff stimulus is indicated by a horizontal bar below each recording. The neuron also responded to jingling keys. The start of and duration of jingling are indicated by the thin horizontal bar (horizontal bar 400 ms, vertical bar 10 mV)

Full size image
Fig. 6
figure 6

Morphological and physiological characteristics of the vir4 neuron. a Confocal reconstruction of the whole brain in a dorsal orientation showing the neural processes innervating the medial and the ventro-lateral region of the protocerebrum (the preparation is tilted slightly posteriorly). Inset Higher magnified image showing part of the ventral cord including the stained axon. b Higher magnification of the neural processes in the medial protocerebrum. c Response traces of the neuron. All three stimulation sequences elicited an excitatory and transient response to the first acoustic pulse produced by the valves, indicated by the vertical lines below the final recording. The onset/offset of the odor/air puff stimulus is indicated by a horizontal bar below each recording (horizontal bar 400 ms, vertical bar 10 mV). d AMIRA reconstruction showing the stained processes in a dorsal view (PC protocerebrum, AL antennal lobe, CB central body, OL optic lobe, P posterior, A anterior, D dorsal, M medial). Bars 100 μm

Full size image

In addition to the ascending ventral-cord neurons, we identified one sound-sensitive brain interneuron with its soma located in the posterior protocerebrum, near the Kenyon cell cluster (vir6; Fig. 7). This neuron also ramified densely within the anterior part of the ventro-lateral protocerebrum. As demonstrated in the AMIRA reconstruction in Fig. 7, this area precisely overlapped with that of the terminal projections of the stained neurons ascending from the ventral cord. The brain interneuron also had a few scattered branches in a more posterio-dorsal region of the protocerebrum (Fig. 7).

Fig. 7
figure 7

Morphological and physiological characteristics of the vir5 neuron. a Confocal reconstruction of the unilateral brain interneuron having its cell body located nearby the calyces of the mushroom bodies (Ca) and extending relatively dense processes in the ventro-lateral region of the protocerebrum (PC, white line). b, c AMIRA reconstructions. b Dorsal orientation. c Sagittal orientation (CB central body, P posterior, M medial, D dorsal, A anterior). Bars 100 μm. d Response properties of the neuron showing a few spikes occurring in synchrony with the first, the second and the fourth acoustic pulse (indicated by vertical lines below the third recording). The onset/offset of the odor/air puff stimulus is indicated by a horizontal bar below each recording. The neuron also responded to jingling keys. The start and duration of jingling are indicated by the thin horizontal bar (horizontal bar 400 ms, vertical bar 10 mV). e Response to the first pulse on a smaller time scale (horizontal bar 20 ms, vertical bar 5 mV). f Interstimulus time histogram (INTH) showing that most spikes occur in quick succession

Full size image

Mass-staining of ventral-cord neurons

Mass-staining of ventral-cord neurons in three H. virescens preparations indicated that the anterior part of the ventro-lateral protocerebrum is a main synaptic region for this category of neuron (Fig. 8). Even though we could not discriminate between ascending and descending neurons by this method or sort out the sound-sensitive neurons from other neuron categories, the finding that one of the strongly labeled brain areas obtained by this kind of mass-staining perfectly overlapped with the region of the dense ramifications of single-stained ventral-cord neurons indicated that this particular area of the ventro-lateral protocerebrum was a general target region for auditory neurons ascending from the ventral cord (Figs. 2, 3, 4, 5, 8).

Fig. 8
figure 8

Labeling of brain regions obtained by mass-staining of ventral-cord neurons in H. virescens. a Confocal reconstruction of the whole brain in a dorsal orientation demonstrating several regions that are strongly stained. In both hemispheres, the white lines indicate a region located in the ventro-lateral protocerebrum, close to the border of the antennal lobe (AL). These stained neuropils are characterized by dense processes forming a circular structure. b AMIRA reconstruction of the whole brain including prominent structures such as the ALs, the mushroom body lobes (MBL), the calyces of the mushroom bodies (Ca) and the central body (CB). In addition, the location of the sphere-like structures located in the anterior ventro-lateral protocerebrum (a-v LP) is indicated. c–e Images of the model brain used as a framework in which reconstructed neurons from various individuals are registered. The neuron in red is the vir1 neuron and the neuron in yellow is the vir 5 neuron. The neural processes of the two neuron categories intermingle within the circular region appearing during mass-staining experiments (white line). c Dorsal orientation. d Frontal orientation. e Sagittal orientation (P posterior, A anterior, D dorsal, V ventral, SOG subesophageal ganglion). Bars 100 μm

Full size image

Physiology

The 17 neurons (9 from H. virescens, 5 from H. armigera and 3 from H. assulta; Table 1) responded to the sound pulses from the valve system with a distinct increase in spike frequency. The responses were in general transient; all the ventral-cord neurons, for example, responded for less than 50 ms (Figs. 3, 4, 5). Figure 9a shows the excitation of one auditory neuron recorded in H. assulta and demonstrating a response pattern similar to those of the ventral-cord neurons stained in the two other species. Most of the responses, including those displayed by the unstained neurons, were elicited by pulse one, two and four (Figs. 3, 4, 5, 6, 7). The vir2 neuron with the funnel-shaped terminations was excited by all four pulses (Fig. 3). This was probably because pulse three was of significantly lower amplitude than the others (Fig. 1). Jingling keys also elicited an increased firing in the neurons (Table 1, Figs. 2, 3, 5, 7, 9). The longer duration of this stimulus, as compared to that of each valve click, generally elicited more persistent neuronal responses.

Fig. 9
figure 9

Response traces of two sound-sensitive neurons. a Neuron from the brain of H. assulta responding transiently to the first, second and fourth sound pulse (second row). A “true” baseline was recorded when the valve system was switched off (top row). In addition, the neuron (ass1) was activated by jingling keys. The start and duration of jingling are indicated by the thin horizontal bar. b Neuron from H. virescens responding not only to the sound pulses but also to the odor/air puff. The persistent response during this kind of stimulation clearly differs from the transient responses induced by the pure sound pulses. The presence of a synaptic potential indicates that the recording was performed close to the dendritic region of the neuron. The four acoustic signals are indicated by vertical lines below the second recording in a and the final recording in b. The onset/offset of the odor/air puff stimulus is indicated by a horizontal bar below each recording (horizontal bar 400 ms, vertical bar 10 mV)

Full size image

We observed some differences in response properties that seemed correlated to the morphological type of the neuron. The ventral-cord neurons all displayed similar responses, characterized by an increased spike frequency occurring at the onset of the acoustic stimuli. The response lasted 20 to 80 ms and never outlasted the acoustic stimulus. The brain interneuron, vir6, on the other hand, fired only three spikes at the corresponding time points (Fig. 7). Moreover, a synaptic potential can be seen in this recording. Furthermore, the ventral-cord neurons with major ramifications in the ventro-lateral protocerebrum all showed response latencies close to 50 ms, whereas the brain interneuron (vir6) had a longer response delay, i.e., ca. 60 ms. The ventral-cord neuron lacking the dense ramifications in the ventro-lateral protocerebrum, namely vir4, which was activated by the first pulse only, showed a response delay of ca. 65 ms (Table 1). In addition to those neurons, which were all unimodal, one neuron (vir9) responded not only to the sound pulses but also to odor/air-puff stimulation when tested with the pheromone mixture (Fig. 9). This neuron showed the longest response latency. Since the control was not tested in the current experiment, we cannot conclude whether the response was caused by the odor or the mechano-sensory stimulus.

The majority of neurons had a relatively low spontaneous activity of less than 10 Hz. Only two neurons fired at frequencies up to 30 Hz (for details, see Table 1).

Discussion

The main finding in the current study is the demonstration of the profuse branching of sound-sensitive neurons in the ventro-lateral protocerebrum of the moth brain, indicating a key role of this brain area in processing acoustic information. The dense processes of the sound-sensitive ventral-cord neurons in the anterior region of the ventro-lateral protocerebrum suggest the presence of an anatomically distinctive synaptic region for auditory processing. We found the same neuron category in two heliothine species that are geographically isolated, i.e., H. virescens and H. armigera. Based on previous measurements of the air flow changes induced by an odor stimulation set-up, we can exclude that these neurons are mechanosensory, because of the absence of their response to the air puff stimulation (see Fig. 1 in Zhao et al. 2013).

Morphology

The ventro-lateral protocerebrum was the target region of all the stained neurons, including the six ascending ventral-cord neurons and the brain interneuron. The vir1 neuron had an axon that was contralateral to the soma and was similar in morphology (within the pterothoracic ganglion) to the interneuron 506 in Boyan and Fullard (1986). Most of the other five ascending neurons were probably interneurons because only two auditory sensory neurons occur in the moth ear (Surlykke and Miller 1982; Boyan and Fullard 1986; Agee and Orona 1988).

In most insect taxa (see also below), the receptor neurons synapse onto ascending neurons and some of these have projections in the protocerebrum (Stumpner and von Helversen 2001). Our finding of an ascending ventral-cord neuron category characterized by dense ramifications in the ventro-lateral protocerebrum is supported by data from the mass-staining experiments demonstrating a prominent sphere-like structure in the same brain region (Fig. 8). We have identified the particular type of neuron in the two allopatric species, H. virescens and H. armigera. Given their similar response pattern to that of the unstained neurons recorded in the brain of H. assulta (Fig. 9a), we expect that the same neuron type is also present in this closely related species.

The location of an auditory brain center as presented here is similar to previous findings in other insects. In the locust brain, auditory neurons ascending from the ventral cord are reported to project into the lower lateral protocerebrum (Boyan 1983). In bushcrickets (Ancistrura nigrovittata, Decticus albifrons and D. verrucivorus) ascending auditory neurons also innervate a ventral region of the lateral protocerebrum (Nebeling 2000; Ostrowski and Stumpner 2010; Stumpner and Molina 2006). Interestingly, the additional ramifications of ascending auditory neurons in the subeosophagal ganglion, reported for bushcrickets, resemble those in our data (Fig. 5). Furthermore, in A. nigrovittata, local brain interneurons target densely the medial and ventral region of the lateral protocerebrum (Ostrowski and Stumpner 2010, 2013). A dissimilar projection pattern has been reported in the well-studied cricket, Gryllus bimaculatus, in which the auditory ventral-cord interneurons seem to target a more frontally and medially located region of the protocerebrum (Zorović and Hedwig 2011). Nevetheless, the discovery of auditory projections targeting a distinct brain region in the ventro-lateral protocerebrum, as found in the present study, also corresponds to data reported in a species lacking a tympanic ear, namely, the fruit fly, Drosophila melanogaster, which uses its Johnston’s organ for sound detection. In spite of the different architecture and location of the external hearing organ in the fruit fly compared with that of the moth, the Drosophila brain includes a small third-order synaptic region in its auditory circuit, which seems to be located in a brain region corresponding to that of the auditory projections presented here, i.e., the ventral-most part of the lateral protocerebrum (Kamikouchi et al. 2006; Lai et al. 2012).

The brain interneuron, vir6, also innervates the ventro-lateral protocerebrum. The intermingled branches of the interneuron and one ventral-cord neuron in this particular region, as shown in the AMIRA model (Fig. 8), indicate a possible synaptic connection between the two neuron categories. The brain interneuron also extends some processes in a more medially located region of the protocerebrum. Interestingly, this area generally overlaps with processes of a newly identified antennal-lobe centrifugal neuron that responds to sound and that has been identified in two of the heliothine species included in the present investigation, namely H. virescens and H. armigera (Zhao et al. 2013). In addition to the ventro-lateral protocerebrum, other higher integration centers of the moth brain might also receive auditory information. Here, we have found that one of the seven stained neurons, vir4, also has processes in the lateral accessory lobe. Generally, this is a region of the insect brain receiving input from many descending neurons, some of which respond to acoustic stimulation (Olberg and Willis 1990).

Previous studies of a variety of insect taxa have found that the lateral protocerebrum contains neuropils for specialized sensory processing. For example, visual information of movement and color are processed in non-overlapping areas of the lateral protocerebrum in the bee brain (Paulk et al. 2009). Furthermore, in the silk moth, antennal-lobe projection neurons for bombykol target a specific area in the inferior lateral protocerebrum, whereas neurons tuned to bombykol are relayed close to but outside this area (Seki et al. 2005). In bush crickets, neurons carrying information about conspecifics project medially in the lateral protocerebrum, whereas neurons responding to ultrasound innervate more ventral areas (Ostrowski and Stumpner 2010, 2013). Thus, the lateral protocerebrum is a region including higher sensory centers linked to distinct modalities.

With respect to the function of the ramifications seen in the subesophageal ganglion (SOG), we can only speculate about whether taste information and feeding are also modulated by predation risk as is pheromone tracking (Skals et al. 2005).

Physiology

The finding of sound stimuli originating from the activated valves was an unplanned event and therefore not part of a systematic stimulus paradigm. Thus, we cannot discuss details of the physiological properties of the neurons; however, the analysis of the four acoustic pulses (measured at the position of the ear) showed that they contain ultrasound frequencies within the range that would activate the thoracic tympanal organ of moths (Fig. 1) and sound pressure levels (71–89 dB SPL) exceeding the threshold of moth ears. Thus, the insect is clearly exposed to sound from the valve pulses. The consistent neuronal activation by jingling keys confirms the classification of the responses as acoustic. Moreover, a newly identified centrifugal antennal-lobe neuron type is activated by particular sound pulses from the same valves (Zhao et al. 2013).

The “funnel-shaped” neurons showed, in general, similar response patterns including a transient increase in spike rate, most often to sound pulse one, two and four and short latencies. The response characteristics match the “repeater type” neurons that have been reported by Roeder (1966) and that have also been found in H. virescens (Boyan and Fullard 1986). The vir4 and vir6 neurons in the present study, which show a somewhat different response pattern by firing only a few spikes during sound stimulation, might correspond to the “pulse coder” type of Roeder (1966). Alternatively, the low firing rate might be attributable to a lower sensitivity of the neurons. In order to explore whether characteristic differences occur in the morphology of “repeater type” and “pulse coder” type neurons, as might be suggested from the current study, more data have to be obtained. The finding of responses to sound and odor/air-puff stimuli in the same recording, as shown in Fig. 9b, demonstrates the multimodal nature of the response and indicates that two sensory modalities are integrated. Olberg and Willis (1990) found, in male gypsy moths (Lymantria dispar), descending neurons whose visual response is modulated by pheromones. Interestingly, the response of one of these neurons is also modified by a click sound inherent in the stimulus set-up. In the locust brain, individual neurons (presumably of the descending type attributable to the absence of a cell body in the brain) have been reported to respond to odor and light (Gupta and Stopfer 2012).

Concluding remarks

Heliothine moths have sound-sensitive neurons with branches in the ventro-lateral protocerebrum of the brain, thus indicating that this confined area is a higher order neuropil dedicated to auditory processing. The ascending neurons also have ramifications in the SOG, whereas the brain interneuron is connected only to other protocerebral areas. Cross-modal integration, found in one of the unstained neurons, also suggests the importance of the lateral protocerebrum for perceptual integration and for the control of moth behavior in the complex natural environment.