A Multisensory Centrifugal Neuron in the OlfactoryPathway of Heliothine Moths

Xin-Cheng Zhao,1 Gerit Pfuhl,1 Annemarie Surlykke,2 Jan Tro,3 and Bente G. Berg1*1Department of Psychology, Neuroscience Unit, Norwegian University of Science and Technology, 7491 Trondheim, Norway2Sound and Behavior Group, Institute of Biology, University of Southern Denmark, 5230 Odense M., Denmark3Acoustics Group, Department of Electronics and Telecommunications, Norwegian University of Science and Technology,

7491 Trondheim, Norway

ABSTRACTWe have characterized, by intracellular recording and

staining, a unique type of centrifugal neuron in the

brain olfactory center of two heliothine moth species;

one in Heliothis virescens and one in Helicoverpa armi-

gera. This unilateral neuron, which is not previously

described in any moth, has fine processes in the dorso-

medial region of the protocerebrum and extensive neu-

ronal branches with blebby terminals in all glomeruli of

the antennal lobe. Its soma is located dorsally of the

central body close to the brain midline. Mass-fills of

antennal-lobe connections with protocerebral regions

showed that the centrifugal neuron is, in each brain

hemisphere, one within a small group of neurons having

their somata clustered. In both species the neuron was

excited during application of non-odorant airborne sig-

nals, including transient sound pulses of broad band-

width and air velocity changes. Additional responses to

odors were recorded from the neuron in Heliothis vires-

cens. The putative biological significance of the centrif-

ugal antennal-lobe neuron is discussed with regard to

its morphological and physiological properties. In partic-

ular, a possible role in multisensory processes underly-

ing the moth’s ability to adapt its odor-guided behaviors

according to the sound of an echo-locating bat is con-

sidered. J. Comp. Neurol. 521:152–168, 2013.

VC 2012 Wiley Periodicals, Inc.

INDEXING TERMS: modulatory neuron; antennal lobe; olfactory pathway; sound sensitive; heliothine moth; multimodal

responses

Odor-guided behaviors are particularly prominent in

insects. Olfactory cues are essential for vital tasks such

as finding food, seeking a mate, and localizing a suitable

host plant for oviposition. In order to respond optimally

for the survival of the individual and preservation of the

species, the odor-induced behaviors need to be adapted

to both endogenous and exogenous conditions. For

instance mating status, age, and previous sensory experi-

ence are reported to affect olfactory neuronal responses

(Gadenne et al., 2001; Anderson et al., 2007; Anton et al.,

2007; Arenas et al., 2009; Denker et al., 2010). Chemo-

sensory neurons also seem to change their sensitivities

according to the circadian rhythm (Linn et al., 1996). At

the behavioral level, associative learning mechanisms

involving pairing of stimuli from different sensory modal-

ities are documented to be of vital importance for estab-

lishing appropriate adaptations (Menzel, 2001). Thus, pre-

vious olfactory exposure associated with rewarding or

aversive taste cues is reported to form memory for dis-

tinct odors (Fan et al., 1997; Daly et al., 2004; Jørgensen

et al., 2007). Furthermore, the combination of visual and

olfactory stimuli is shown to promote nectar feeding

behavior (Raguso and Willis, 2002). Bees even intensify

their foraging activity when being exposed to particular

colors formerly associated with a conditioned odor

(Giurfa et al., 1994).

Grant sponsor: the Norwegian Research Council; Grant number:1141434; Grant sponsor: the Royal Norwegian Society of Sciences andLetters (I. K. Lykke’s Foundation, 2008); Grant sponsor: the DanishCouncil for Independent Research/Natural Sciences; Grant numbers: 272-08-0386 and 09-062407.

*CORRESPONDENCE TO: Bente G. Berg, Department of Psychology,Neuroscience Unit, MTFS, Olav Kyrresgt. 9, Norwegian University ofScience and Technology (NTNU), N-7491 Trondheim, Norway.E-mail: bente.berg@ntnu.no

VC 2012 Wiley Periodicals, Inc.

Received August 19, 2011; Revised December 21, 2011; Accepted June 5,2012

DOI 10.1002/cne.23166

Published online June 11, 2012 in Wiley Online Library (wileyonlinelibrary.com)

152 The Journal of Comparative Neurology |Research in Systems Neuroscience 521:152–168 (2013)

RESEARCH ARTICLE

Modulation of the various odor-guided behaviors pre-

sumes activation of distinct mechanisms, which implies

that the principle of plasticity has to be built in at the neu-

ronal level. The central olfactory pathway of insects

includes three main neuropils: the primary olfactory

center, i.e., the antennal lobe, and two higher integration

centers, the mushroom body calyces and the lateral pro-

tocerebrum (Boeckh et al., 1984; Homberg et al., 1988).

The antennal lobe is the final destination of the numerous

olfactory sensory neurons located mainly on the anten-

nae. Here, the sensory axon terminals make synapses

with antennal-lobe neurons in spherical structures called

glomeruli. The second-order neurons comprise two cate-

gories, local interneurons, which are confined to the

antennal lobe, and projection neurons carrying the olfac-

tory information to the higher integration centers in the

protocerebrum. An additional and relatively small group

of antennal-lobe processing units is comprised of the cen-

trifugal neurons, which are presumed to play a key role in

modification of olfactory information (reviewed by Anton

and Homberg, 1999; Galizia and R€ossler, 2010).

The general morphology of this neuron category

includes neuronal processes in various regions of the

nervous system and a fiber projecting into the antennal

lobe, often innervating all glomeruli. The centrifugal neu-

rons seem to receive input from various sensory channels

and to forward these signals to the antennal lobe, pre-

sumably for appropriate adjustment of the olfactory infor-

mation (Homberg and Muller, 1999). In addition to com-

parable branching patterns, a common hallmark of this

category of neurons is their release of biogenic amines

(e.g., serotonin, octopamine, and dopamine) or neuropep-

tides (reviewed by Galizia and R€ossler, 2010). A main por-

tion of the modulatory neurons identified so far has there-

fore been revealed by immunocytochemical studies

(reviewed by Homberg and Muller, 1999; Schachtner

et al., 2005).

One centrifugal neuron that is particularly well

described in several moth species, as well as other

insects, is the so-called serotonin-immunoreactive (SI)

antennal-lobe neuron (Kent et al., 1987; Salecker and

Distler, 1990; Hill et al., 2002; Dacks et al., 2006; Zhao

and Berg, 2009). In Lepidoptera species, the SI neuron

has wide-field arborizations in both protocerebral hemi-

spheres and extensive terminal projections in all glomer-

uli of the antennal lobe contralateral to that containing

the soma (reviewed by Kloppenburg and Mercer, 2008).

The localization of input synapses primarily in the proto-

cerebrum and output synapses in the antennal lobe con-

firms the assumption that the neuron is a descending ele-

ment (Sun et al., 1993). Concerning functional properties

of the SI-neuron, pharmacological investigations have

reported that serotonin increases pheromone responses

of antennal-lobe neurons (Kloppenburg et al., 1999; Klop-

penburg and Heinbockel, 2000; Hill et al., 2003). A role of

the SI-neuron in adjusting pheromone-guided responses

according to the circadian rhythm has been suggested

(Gatellier et al., 2004). However, the main function of this

widespread neuron type is not yet fully understood. Elec-

trophysiological recordings from the neuron itself have

been obtained a few times, and they include responses to

mechanical and odor stimuli (Hill et al., 2003; Zhao and

Berg, 2009).

Another intensively studied category of centrifugal

antennal-lobe neurons is the so-called midline neurons of

the subesophageal ganglion (reviewed by Anton and

Homberg, 1999). This kind of octopaminergic antennal-

lobe neuron has been identified in the fruit fly (Stocker

et al., 1990), the locust (Br€aunig, 1991), and the honey-

bee (Hammer et al., 1993). In a sequence of well-

designed experiments carried out in the last mentioned

species, Martin Hammer (1993) showed that one particu-

lar neuron of this category, the VUMmx1 neuron, is sensi-

tive to sucrose and mediates information about the

unconditioned taste stimulus in associative olfactory

learning. The morphology of the VUMmx1 neuron, having

dendritic arborizations in the suboesophageal ganglion,

i.e., the primary gustatory center, and extensive projec-

tions in the antennal lobe and the calyces of both brain

hemispheres, corroborates its function. Octopaminergic

projections in the antennal lobe, possibly originating from

the same type of neuron, have also been reported in the

moth (Dacks et al., 2005).

Because most centrifugal neurons have been identified

through immunocytochemical rather than electrophysio-

logical methods, our understanding of their functions and

encoding characteristics is, in general, still limited. The in-

tracellular recording and staining technique, which pro-

vides first-hand information from the individual process-

ing units, is suitable for obtaining knowledge about

physiological and morphological features characterizing

these neuron categories—even though the data materials

may be quantitatively restricted. Based on the current

methodological approach, we here present the morphol-

ogy and, in addition, physiological properties of a unique

category of antennal-lobe centrifugal neurons in the moth

brain that has previously not been described.

MATERIALS AND METHODS

The current analyses are based on intracellular record-

ing and staining of two neurons, plus data from mass

staining experiments. Even though the number of

recorded neurons is small, the morphological similarity

between the successful stainings in two heliothine spe-

cies confirms the presence of this hitherto undiscovered

neuron type in moth brains.

A multisensory centrifugal neuron

The Journal of Comparative Neurology |Research in Systems Neuroscience 153

Insects and preparationPupae of Heliothis virescens and Helicoverpa armigera

were imported from laboratory cultures (H. virescens

from Syngenta, Basel, Switzerland; and H. armigera from

Henan University of Science and Technology, Henan,

China). Male and female pupae were separated and kept

in climate chambers on reversed photoperiod light-dark

14:10 hours at 22�C. The adults were fed a 5% sucrose

solution. Experiments were performed on adult males 2–

5 days after ecdysis, as previously described by Zhao and

Berg (2009). The moth was restrained inside a plastic

tube with the head and antennae exposed. The head was

immobilized with wax (Kerr, Romulus, MI), and the anten-

nae were lifted up by needles. The brain was exposed by

opening the head capsule and removing the mouth parts,

the muscle tissue, and major trachea. The sheath of the

antennal lobe was removed by fine forceps in order to

facilitate microelectrode insertion into the tissue. Once

the head capsule was opened, the brain was supplied

with Ringer’s solution (in mM: 150 NaCl, 3 CaCl2, 3 KCl,

25 sucrose, and 10 N-tris (hydroxymethyl)-methyl-2-

amino-ethanesulfonic acid, pH 6.9). The whole prepara-

tion was positioned so that the antennal lobes were fac-

ing upward and were thus accessible for intracellular

recordings.

Intracellular recording and stainingThe intracellular recordings from the antennal-lobe

neurons were carried out as previously described (Zhao

and Berg, 2009). Recording electrodes were made by

pulling glass capillaries (borosilicate glass capillaries; Hil-

genberg, Malsfeld, Germany; OD: 1 mm, ID: 0.75 mm) on

a horizontal puller (P97; Sutter Instruments, Novarto, CA).

The tip was filled with a fluorescent dye (4% tetramethylr-

hodamine dextran with biotin; Micro-Ruby, Molecular

Probes/Invitrogen, Eugene, OR; in 0.2 M Kþ-acetate),

and the glass capillary was backfilled with 0.2 M Kþ-ace-

tate. A chloridized silver wire inserted into the eye served

as the indifferent electrode. The recording electrode,

which had a resistance of 150–400 MX, was lowered

carefully into the brain by means of a micromanipulator

(Leica, Bensheim, Germany). The recording site was cho-

sen within the dorsolateral region of the antennal lobe.

Neuronal spike activity was amplified (AxoClamp 2B,

Axon Instruments, Union, CA) and monitored continu-

ously by oscilloscope and loudspeaker. The recordings

were stored by using the Spike2 6.02 software (Cam-

bridge Electronic Design, Cambridge, England).

After physiological characterization of responses to the

test stimuli, the neurons were iontophoretically stained

by applying 2–5-nA depolarizing current pulses with 200

ms duration at 1 Hz for about 2–10 minutes via the glass

capillary electrode. In order to allow neuronal transporta-

tion of the dye, the preparation was kept for 2 hours at

room temperature. The brain was then dissected from the

head capsule. After being fixed in 4% paraformaldehyde

for 1 hour at room temperature, the brain was rinsed with

phosphate-buffered saline (PBS; in mM: 684 NaCl, 13

KCl, 50.7 Na2HPO4, 5 KH2PO4, pH 7.4). The staining was

then intensified by incubating the brain in fluorescent

conjugate streptavidin-Cy3 (Jackson ImmunoResearch,

West Grove; PA, diluted 1:200 in PBS), which binds to bio-

tin, for 2 hours. Incubation was followed by rinsing with

PBS and dehydration in an ascending ethanol series (50%,

70%, 90%, 96%, 2 � 100%; 10 minutes each). Finally, the

brain was cleared and mounted in methylsalicylate.

Mass stainingsMass staining experiments were performed on five H.

virescens males. A region located centrally in the anten-

nal lobe was cut with a razor blade. Crystals of fluores-

cent dye (Micro-Ruby, Molecular Probes) picked up by

the fine tip of a micro needle were then applied into the

antennal lobe by hand. The brain was subsequently sup-

plied with Ringer’s solution and kept for 2 hours at room

temperature for transportation of the dye. The following

procedure was as described above.

Odor and air puff stimulationThe odor delivery system for the intracellular recording

consisted of two glass cartridges (ID 0.4 cm) placed side

by side, both pointing toward the antenna at a distance of

2 cm. One replaceable cartridge contained a piece of fil-

ter paper onto which a particular odor stimulus was

applied. The other cartridge contained a clean filter pa-

per. An air flow (500 ml/min) led through the odorless

cartridge was continuously blown over the antenna. Dur-

ing each stimulus period, which lasted for 400 ms, the air

flow was switched by a valve system from the odorless to

the odor-bearing cartridge. As olfactory stimuli we used

the two-component pheromone blends of the two spe-

cies, i.e., cis-11-hexadecenal (Z11–16:AL) and cis-9-tetra-

decenal (Z9–14:AL) in a ratio of 94:6 for H. virescens and

Z11–16:AL and cis-9-hexadecenal (Z9–16:AL) in a ratio of

95:5 for H. armigera (pheromone chemicals, Plant

Research International, Pherobank, Wageningen, Nether-

lands), plus the plant oil ylang-ylang (Dragoco, Totowa,

NJ). Odor compounds diluted in hexane were applied onto

a small filter paper. The hexane was allowed to evaporate

before the filter paper was wrapped up and placed in the

cartridge. All stimuli were prepared so that the filter pa-

per contained 10 ng of the binary pheromone blend, and

100 lg of the plant oil. A cartridge containing a clean fil-

ter paper was used as control. The odor stimuli were reg-

ularly renewed during the experimental period.

Zhao et al.

154 The Journal of Comparative Neurology |Research in Systems Neuroscience

Sound stimulationThe sound stimuli originated from the solenoid valve

system that controls the air flow during odor stimulation.

The two valves (2-way Direct Lift Solenoid Valves, 01540;

Cole-Parmer, Vernon Hills, IL) are mounted on one of the

walls of the room, 2 m away from the insect. Valve 1

(type normally open) regulates the continuous airflow and

valve 2 (type normally closed) the odor/air puff. When

the odor stimulation system is activated, the two valves

produce four separate sound pulses; the first pulse is

caused by valve 1 being closed (first event); the second

pulse, which appears 1 second later, by valve 2 being

opened (second event); the third pulse occurs 0.4 second

thereafter by valve 2 being closed (third event); and the

forth pulse, which appears after another 1 second, by

valve 1 being reopened (fourth event; Fig. 1). The second

and the third event thus correspond with the odor stimu-

lus’s onset and offset, respectively. Activation of the two

valves was recorded in the Spike program via separate

channels.

We recorded the sound pulses created by opening and

closing the valves using a 1=4-inch microphone and pream-

plifier (types 4138 and 4939; Bruel & Kjær, Nærum, Den-

mark), plus a front end amplifier (type 1201/30517; Nor-

sonic, Lierskogen, Norway); the pulses were then

digitized (sample frequency 192 kHz), and the signals

Figure 1. Measurements of the sound pulses and air flow during the stimulation sequence. A: Showing how the changes in airflow suc-

ceed the sound pulses. The arrows on the top point to when the four sound clicks are produced by the valve system as shown in the

time signal below. B: Expanded time view of sound click 1, showing the time signal and the spectrogram of the pulse recorded outside

and inside the plastic. C: Amplitude spectra of clicks 1 and 2 (left) and 3 and 4 (right) when recorded inside the plastic. The spectra show

that all four clicks have energy above 20 kHz.

A multisensory centrifugal neuron

The Journal of Comparative Neurology |Research in Systems Neuroscience 155

were stored on a signal analyzer, WinMLS. The sounds

were measured with and without air flow both at the posi-

tion of the insect’s ear in air as well as inside the plastic

tube wherein the moth was restrained. The recorded sig-

nals were analyzed by using commercial software, Bat-

Sound Pro (Pettersson Elektronik, Uppsala, Sweden) (Fig.

1). The pulses were broad band with energy from audible

up to ultrasonic frequencies of 40–50 kHz. The main

effect of the plastic was to dampen frequencies above

�25 kHz (Fig. 1: compare spectrogram of ‘‘click 1’’ with

‘‘click 1 in plastic’’). Inside the plastic the spectra of all

clicks had energy up to above 20 kHz (Fig. 1, lower

panel). 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. Due to the decreasing amplitude

it was 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 �100 ms (Fig. 1). 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.

Measurement of air flowWe 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;

TSI, Shoreview, MN). The air flow was measured at the

site of the insect antenna over an interval of 11 second,

starting 4 seconds before the closing of valve 1 and end-

ing 4.6 seconds after the reopening of valve 1. The length

of the Teflon tubes (ID 2 mm) transporting the air from

the valves to the glass cartridges is approximately 4 m.

The changes in air flow during the relevant time period

are shown in Figure 1; the initial event, the closing of

valve 1, caused a relatively rapid decrease in air flow. The

two subsequent events, opening and closing of valve 2,

induced a transient air flow increase and decrease,

respectively. The fourth air flow change, caused by valve

1 being reopened, differed from the three preceding ones

by a pronounced slower time course. As the exact onsets

of the air-flow changes at the position of the insect could

not be plotted via the Spike2 program, the curve describ-

ing the air flow is aligned with the time signal of the sound

pulses according to the minimum delay possible (Fig. 1).

In order to simplify the terminology, the rapid increase in

air velocity caused by valve 2 being opened is termed air

puff stimulation in the subsequent text.

ImmunocytochemistryAntibody characterization

In order to identify antennal-lobe glomeruli and proto-

cerebral neuropil structures, immunostaining with anti-

bodies marking synaptic regions was performed on the la-

beled brains. Two monoclonal antibodies shown to detect

proteins associated with synaptic terminals were used,

both developed in mouse (Table 1). The anti-SYNORF1

was raised against fusion proteins composed of glutathi-

one-S-transferase and the Drosophila SYN1 protein (SYN-

ORF1; Klagges et al., 1996). The specificity of the anti-

SYNORF1 has been described previously; in Drosophila at

least four, probably five, isoforms were recognized by the

monoclonal antibody in Western blots of head homoge-

nates (three bands at 70, 74, and 80 kDa and one or two

bands at �143 kDa; Klagges et al., 1996). In Western

blots of H. virescens brains, the current antibody labels

five bands, two at 74 kDa and three at 55–60 kDa (Rø

et al., 2007). The anti SYNORF1 was delivered by Devel-

opmental Studies Hybridoma Bank (University of Iowa,

Iowa City, IA) (Table 1). The other antibody, nc46, which

was raised against homogenized Drosophila heads (Hof-

bauer et al., 2009), detects an epitope of a protein that is

connected with synaptic vesicles, the so-called synapse-

associated protein of 47 kDa (SAP47; Hofbauer et al.,

2009; Saumweber et al., 2011). In Western blots of ho-

mogenates from isolated Drosophila brains, the antibody

is reported to label a prominent band at 47 kDa (Reich-

muth et al., 1995). The nc46 antibody was kindly donated

by E. Buchner (Table 1).

TABLE 1.

Primary Antibodies Used

Antigen Immunogen

Manufacturer, species antibody was raised in,

mono- vs. polyclonal, cat. no. Dilution

Synapsin Fusion protein of glutathione-S-transferase(GST) and the Drosophila SYN1 protein

Developmental Studies Hybridoma Bank, Universityof Iowa (Iowa City, IA; developed by Erich Buchner,Wurtzburg University, Wurzburg, Germany), mouse;monoclonal; 3C11

1:10

Synapse-associatedprotein of 47 kDa(SAP47)

Homogenized Drosophila heads Erich Buchner, Wurzburg University,Wurzburg, Germany

1:40

Zhao et al.

156 The Journal of Comparative Neurology |Research in Systems Neuroscience

Immunostaining for identification of glomeruliand brain structures

After analysis of the iontophoretically stained neuron

by confocal laser scanning microscopy, the brain was

rehydrated through a decreasing ethanol series (10

minutes each) and rinsed in PBS. To minimize nonspecific

staining, the brain was submerged in 5% normal goat se-

rum (NGS; Sigma, St. Louis, MO) in PBS containing 0.5%

Triton X-100 (PBSX; 0.1 M, pH 7.4) for 3 hours at room

temperature. The preparation was then incubated for 2

days at 4�C in the primary antibodies, anti-SYNORF1 and

nc46 (dilutions 1:10 and 1:40 in PBSX containing 5%

NGS, respectively; Table 1). After rinsing in PBS 6 � 20

minutes at room temperature, the brain was incubated

with Cy2-conjugated anti-mouse secondary antibody

(Invitrogen, Eugene, OR; dilution 1:500 in PBSX) for 2

days at 4�C. Finally, we rinsed, dehydrated, cleared, and

mounted the brain in methylsalicylate.

Confocal image acquisitionWe created serial optical images by using a confocal

laser scanning microscope (LSM 510, META Zeiss, Jena,

Germany) with a 10� (C-Achroplan 10�/0.45 W), 20�(Plan-Neofluar 20�/0.5), and 40� objective (C-Achro-

plan 40�/0.8W). The intracellular staining, obtained

from the fluorescence of rhodamine/Cy3 (Exmax 550 nm,

Emmax 570 nm), was excited by the 543-nm line of a

HeNe1 laser and the immunostaining, obtained from the

Cy2 (Exmax 490 nm, Emmax 508 nm), was excited by the

488-nm line of an Argon laser. The distance between

each section was 8 lm for the 10� objective and 2 lm

for the 20� and 40� objectives. The pinhole size was 1

and the resolution 1,024 � 1,024 pixels. Optical sections

from the confocal stacks were reconstructed by means of

the LSM 510 projection tool.

Digital 3D reconstructionsThe centrifugal neuron and its surrounding brain struc-

tures were manually reconstructed in subsequent confo-

cal slices by means of the visualization software AMIRA

4.1 (Visage Imaging, Furth, Germany); the neuron was

reconstructed by using the skeleton module of the soft-

ware, and the brain structures by using the segmentation

editor. Thus, the neuron was traced so that a surface

model built by cylinders of particular lengths and thick-

nesses was created, whereas each brain structure was

outlined based on its gray value according to the back-

ground so that a polygonal surface model could be cre-

ated. The neuronal structures and the neuropil regions

were reconstructed from the same confocal image

stacks.

Data analyses and image processingThe electrophysiological recordings were stored and

analyzed by using the Spike2 software. We counted the

numbers of spikes within intervals of 100 ms during a

total period of 5 seconds. This period started 1 second

before the onset of the odor/air stimulus, close to the

onset of the first auditory pulse, and ended 2.5 seconds

after the onset of the last auditory pulse. Based on the

spike frequencies, we made histograms to visualize the

neuronal activity during the sequence of sound, air veloc-

ity, and odor/air puff stimuli for each recording. The

images were adjusted in Photoshop CS5 (Adobe Systems,

San Jose, CA) by means of the auto-contrast tool before

the final figures were edited in 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).

RESULTS

In males of both species we identified a unique proto-

cerebral neuron with extensive processes in the antennal

lobe, which was comparable not only morphologically but

also physiologically, one in an H. virescens male (Fig. 2)

and one in an H. armigera male (Fig. 3). For simplicity,

these neurons are named the vir neuron and the arm neu-

ron, respectively.

Morphology of the centrifugalantennal-lobe neurons

The morphologies of the two neurons, characterized by

processes in the medial protocerebrum of one hemi-

sphere and extensive innervations in the ipsilateral anten-

nal lobe, showed a high degree of similarity. The soma,

which had a diameter of 20 lm, was located dorsally of

the central body and close to the brain midline (Figs. 2,

3). From the soma, the primary neurite projected anteri-

orly along the medial margin of the brain and bifurcated

adjacent to an area where the mushroom body lobes

merge. One thin branch projected further anteriorly along

the medial edge and entered into the antennal lobe. Here

it gave off relatively solid branches with blebby terminals

apparently in all glomeruli (Figs. 2, 3). The fiber projected

outside the antennocerebral tracts. The other branch,

being considerably thicker, ran laterally for a short dis-

tance and then turned posteriorly along the medial border

of the mushroom body lobes, where numerous fine arbori-

zations extended in a restricted region of the medial pro-

tocerebrum (Figs. 2B, 3C). This area was located anterior

to the median calyx, dorsolateral to the central body, dor-

somedial to the pedunculus, ventromedial to the a-lobe

(vertical lobe), and posterior to the b-lobe (medial lobe).

One small distinction between the neurons in the two

preparations was observed; whereas the arm neuron

A multisensory centrifugal neuron

The Journal of Comparative Neurology |Research in Systems Neuroscience 157

targeted the mushroom body lobes via a few protocere-

bral branches (Fig. 3D), the vir neuron did not innervate

this structure (Fig. 2C). Whereas the centrifugal neuron

was the only neuron stained in H. virescens, an additional

antennal-lobe local interneuron was stained in the H.

armigera preparation. Figure 3B thus includes branches

of this neuron. Its soma, however, is not visible in the cur-

rent confocal stack.

Mass staining experiments (n ¼ 5), including applica-

tion of dye into one or both antennal lobes of H. virescens,

showed (in four of the five preparations) a cluster of two

or three adjacently located cell bodies positioned close

to the brain midline in each hemisphere, dorsal to the

central body (Fig. 4). This site precisely corresponds with

the location of labeled cell bodies in the two iontophoreti-

cally stained preparations. The somata visualized via

mass stainings had diameters identical to those labeled

by iontophoresis, i.e., 20 lm. The characteristic branch-

ing pattern in the dorsomedial protocerebrum obtained

by mass staining of the antennal lobe also corresponds

well with that of the neuron type labeled by the single cell

technique (Figs. 2–4).

The stained centrifugal neuron in the H. virescens prep-

aration (the vir neuron) permitted detailed analysis of the

glomerular branching pattern. Eight sub-branches were

given off from the main axon in the antennal lobe, each

targeting a group of adjacently located glomeruli. This is

illustrated in Figure 5, in which reconstructed glomeruli

are given distinct colors according to the sub-branch they

are connected to. The number of glomeruli targeted by

each branch varied in a range from 2 to 19. The projec-

tion patterns within the glomeruli differed slightly, some

having more dense innervations than others (Fig. 5E,F).

Whereas most glomeruli were targeted by one sub-branch

only, some received terminals from two (Fig. 5G). One of

the sub-branches targeted the macroglomerular complex

(MGC) and some of the ordinary glomeruli located adja-

cently (Fig. 5A,B). By adding up the numbers of glomeruli

innervated by each sub-branch, we counted 71 units. This

number includes four posterior glomeruli that were not

incorporated in the previous anatomical study by Berg

et al. (2002). Taking into account the fact that a few units

also received projections from different sub-branches,

the number of innervated glomeruli seems to correspond

Figure 2. Morphology of the antennal-lobe centrifugal neuron identified in the H. virescens male, the so-called vir neuron. A: Confocal

reconstruction of one brain hemisphere showing the whole neuron with its cell body located close to the brain midline (arrow), numerous

thin arborizations in the medial protocerebrum (PC), and more blebby processes in the ipsilateral antennal lobe (AL; put together by three

confocal stacks: 72 sections for the protocerebral part, i.e., 144 lm; 58 sections for the middle part containing the soma, i.e., 116 lm;

97 sections for the antennal lobe, i.e., 194 lm). B: Digital reconstruction of one brain hemisphere, made by the AMIRA software. The dor-

sally oriented model shows, in addition to the dense antennal-lobe processes, the location of the protocerebral arborizations according to

the central body (CB), the pedunculus (Pe), the a- and b-lobes, and the calyces (Ca). C,D: Part of the digital model showing the protocere-

bral branches in a ventral (C) and a sagittal view (D). a a-lobe; b b-lobe; M, medial; P, posterior. Scale bar ¼ 100 lm in A–D.

Zhao et al.

158 The Journal of Comparative Neurology |Research in Systems Neuroscience

well with the total number identified in this moth species,

i.e., 65 (Berg et al., 2002).

Physiological properties of the centrifugalantennal-lobe neurons

The two neurons also showed notable similarities

regarding their physiological properties. Both displayed

spikes with amplitudes of approximately 40 mV and spike

durations of �4 ms. Furthermore, their spontaneous

activities were similar at the beginning of the recordings,

showing a frequency of approximately 20 Hz (Figs. 6, 7).

The vir neuron, tested for stimuli several more times than

the arm neuron, showed a decreased spontaneous activ-

ity during the time course of recording, however (Fig. 6).

Both neurons displayed excitatory responses to distinct

non-odorant stimuli produced by the valve system (Figs.

6, 7). Also, the characteristic transient time courses of

these responses were similar. The non-odor activations of

the two neurons occurred at different time points of the

stimulation sequence, however. The vir neuron responded

consistently during the last event (Fig. 6), which involved

an acoustic pulse and a slow increase in air velocity (Fig.

1). The arm neuron, on the other hand, repeatedly

showed an increased firing rate during the first event (Fig.

7), which involved an acoustic pulse and a decrease in air

velocity (Fig. 1). Odor stimuli were delivered with long

pauses, so adaptation during the testing series seems

unlikely, but the repetition rate of non-odorant stimuli

within each sequence may explain the clear response

only during the first event in the arm neuron. As shown in

Figure 7, the response in the arm neuron occurred before

the change in air flow. It is therefore reasonable to

assume that the excitation was induced by the sound

pulse specifically. In contrast, the stimulus giving rise to

the non-odorant response in the vir neuron is not obvious;

as shown in Figure 6, both the sound pulse and the air

flow change occurred before the onset of the response.

The vir neuron also showed a strong excitatory

response at the onset of odor/air puff stimulation, which

overlapped in time with the second acoustic pulse

(Fig. 6). The different strengths of the responses elicited

Figure 3. Morphology of the antennal-lobe centrifugal neuron identified in the H. armigera male, the so-called arm neuron. A: Confocal

reconstruction of one brain hemisphere showing the arborizations in the medial protocerebrum (confocal stack containing 25 sections,

i.e., 50 lm). The site of the cell body is indicated by an arrow (the primary neurite is missing due to a slight damage of the preparation).

B: Confocal reconstruction showing the axon passing from the medial protocerebrum (PC) to the ipsilateral antennal lobe where it gives

off extensive neuronal branches (confocal stack containing 72 sections, i.e., 144 lm). An additional local antennal-lobe neuron was

stained in this preparation, but its soma is not visible in the image. C: Digital model, made by the AMIRA software, of the protocerebral

arborizations located between the mushroom body lobes, the pedunculus (Pe), and the central body (CB), plus the axon projecting into

the antennal lobe (AL). D,E: Part of the digital model showing the protocerebral branches in a ventral (D) and a sagittal (E) perspective. As

indicated by arrows in D, two of the ventrally located dendrites innervated the mushroom body lobes. a a-lobe; b b-lobe; OL, optic lobe;

P, posterior; M, medial. Scale bar ¼ 100 lm in A–E.

A multisensory centrifugal neuron

The Journal of Comparative Neurology |Research in Systems Neuroscience 159

during pheromone, plant odor, and pure air puff stimula-

tion, and also the relatively longer response durations

compared with those occurring at the last event, at least

during the initial stimulation sequences, indicate the

presence of odor responses from the current neuron. The

restricted repetitions of each test stimulus and the

change in spontaneous spiking activity make statistical

calculations difficult, but the mean responses of 79.74

spikes per second to the pheromone mixture (n ¼ 2, SD

¼ 67.05), 43.61 spikes per second to the plant odor (n

¼ 2, SD ¼ 68.81), and 35.72 spikes per second to the

air puff (n ¼ 3, SD ¼ 612.29) suggest the presence of

specific odor responses as well as an additional and

somewhat weaker air puff response. The arm neuron

showed no increase in spike firing rate at the time of the

air/odor puff (Fig. 7). The durations of the excitatory

responses seemed to depend on the modality. On the

whole, the responses elicited during the second event,

which included application of odor/air puff stimuli, had

longer durations, i.e., up to 200 ms, compared with those

elicited by the non-odor stimuli, which lasted about 50–

100 ms. The response delays varied considerably; in the

vir neuron the delay of the odor/air puff responses was

measured to 130 ms whereas the subsequent response,

occurring at the fourth event, started approximately 90

ms after the onset of the air flow change and 270 ms af-

ter the sound pulse. In the arm neuron, the response

started 100 ms after the sound pulse. The sequence of

recordings as listed in Figures 6 and 7 corresponds to the

order of stimuli applied during the experiments.

The recording from the arm neuron, which was stained

together with a local interneuron, probably originated

from the centrifugal neuron and not from the local one; in

addition to comparable response properties, the duration

and amplitude of the spikes were similar in the two

recordings, as mentioned above, and therefore indicate

that they originated from the same neuron category.

DISCUSSION

The main result of our study is the identification of a

central neuron type in the moth brain, which responded

regularly to distinct non-odorant airborne stimuli compris-

ing transient broad band sound pulses and air velocity

changes, as well as in some cases to odors. Its morphol-

ogy, characterized by extensive ramifications in the proto-

cebrum and the antennal lobe, plus a cell body located

close to the protocerebral branches, suggests a function

in modulating odor information. Furthermore, we found

that the soma of the neuron is, in each brain hemisphere,

located within a small cell cluster including one or two

additional somata of antennal-lobe neurons.

Morphological propertiesThe morphology of the two neurons, which is charac-

terized by fine branches in the protocerebrum and consid-

erably thicker and partly blebby structures in the antennal

lobe, plus a cell body located in the protocerebrum, indi-

cates the presence of a particular type of centrifugal

antennal-lobe neuron. This kind of neuron has not been

found previously in any moth species or other insects,

although there might be one exception. A particular cen-

trifugal neuron type formerly identified in the honeybee is

to some extent morphologically comparable to the cate-

gory presented here (Iwama and Shibuya, 1998; Kirsch-

ner et al., 2006). This type, which was called the feed-

back neuron 1 (ALF-1) by Kirschner and colleagues, also

connects parts of one protocerebral hemisphere to the ip-

silateral antennal lobe. Furthermore, the smooth neuronal

branches in the protocerebrum and terminals with bleb-

like structures in the antennal lobe match our findings in

the heliothine moths. Differences, however, include the

distinct target regions of the protocerebral processes and

their branching pattern, which is considerably less exten-

sive in the honeybee than in the moth. The position of the

honeybee’s cell body adjacent to the a-lobe (vertical

lobe) of the mushroom bodies also differs from that of the

moth. Another distinction is that the ALF-1 is reported to

Figure 4. Labeling of somata and neural branches in the proto-

cererum (PC) obtained by mass staining of the antennal lobes

(AL) of a H. virescens male. The labeled processes occupying the

dorsomedial protocerebrum (PC; arrows) are in each hemisphere

connected to two somata located close to the brain midline

(arrowheads in the magnified image). Ca, calyces; P, posterior; A,

anterior, Scale bar ¼ 100 lm; 50 lm in inset.

Zhao et al.

160 The Journal of Comparative Neurology |Research in Systems Neuroscience

Figure 5. Glomerular innervation pattern of the centrifugal neuron in H. virescens (digital model of the vir neuron made by the AMIRA

software). A: Frontal view of the eight neuronal sub-branches extending from the main axon in the antennal lobe, each of which is given a

distinct color. B,C: Frontal and posterior view, respectively, of the glomerular clusters targeted by the eight sub-branches. Each cluster is

given a color corresponding to that of the innervating sub-branch. D: Neuronal tree showing the branching patterns of the eight axonal

extensions and, in addition, one example of glomerular innervation from each category. E–G: Three examples of glomerular innervation pat-

terns, two showing the typical pattern, i.e., terminals from one sub-branch only, one with relatively dense innervations (E) and the other

with scarce innervations (F) whereas the third example shows the unusual innervation pattern, i.e., terminals originating from two different

sub-branches (G). Scale bar ¼ 50 lm in A–C.

A multisensory centrifugal neuron

The Journal of Comparative Neurology |Research in Systems Neuroscience 161

be unique in each hemisphere whereas the centrifugal

type of the moth seems to be one in a particular subcate-

gory. It should be mentioned, however, that in spite of the

finding of a small cell cluster in the heliohine moths, the

antennal-lobe neurons to which they connect may have

different morphological and physiological properties.

Figure 6. Physiological characteristics of the neuron found in the H. virescens male, i.e., the so-called vir neuron. A: The initial spontane-

ous activity of the neuron (�20 Hz) decreased considerably during the sequence of recording. The spiking activity demonstrates excitatory

responses at the last event that included a sound pulse plus an air velocity increase and at the second event implying the presence of an

odor in addition to the non-odorant stimuli. An odor-specific response is indicated by the higher spike frequency rate appearing during

pheromone and plant odor application compared with that displayed when the pure air puff was delivered. The occurrences of acoustic

signals are, at all four events, indicated by a closed arrowhead below the final recording, whereas the estimated start of air flow changes

is shown at the first and last event by an open arrowhead. The onset/offset of the odor/air puff stimulus is indicated by a horizontal bar

below each recording. Vertical bar, 10 mV; horizontal bar, 400 ms. B: Histograms made by counting the number of spikes every 100 ms.

Significant responses are indicated by stars. Horizontal bar, 400 ms; vertical bar, 30 Hz.

Zhao et al.

162 The Journal of Comparative Neurology |Research in Systems Neuroscience

Physiological propertiesThe non-odor stimulations of the preparation was an

unplanned side effect of the activated valve system, and

was thus not studied as systematically as we would have

liked. However, the results show that both neurons

responded consistently to non-odor cues occurring at

particular events during each stimulation sequence.

These air-borne signals included two subsequent constit-

uents, sound pulses followed by air velocity changes,

both produced by the activated valves. Because the

sound pulses contained ultrasound frequencies at a level

that is above the threshold for the thoracic tympanal

organ of the moth, as demonstrated in Figure 1, there is

no doubt that the insect was exposed to audible sound.

From the data obtained, it is reasonable to assume

that the arm neuron was activated by the sound pulse.

The lack of synchrony between the change in air flow and

the response in the arm neuron renders the possibility of

an alternative mechanical response very unlikely (Fig. 7).

The response from the vir neuron, in contrast, might have

been a result of either stimulus, i.e., the sound pulse or

the air flow change (Fig. 6). Mechanical responses to air

puffs have previously been recorded from the moth brain,

usually from the main categories of antennal-lobe neu-

rons (Kanzaki and Shibuya, 1986; Han et al., 2005; Zhao

and Berg, 2010), but in one case also from a centrifugal

antennal-lobe neuron (Hill et al., 2005). Taking the

response pattern of both neurons into account, it seems

relatively unlikely that the two individuals of the unique

neuron type presented here, showing such an extent of

morphological similarity, were activated by totally differ-

ent sensory channels.

Interestingly, we have repeatedly measured responses

to the sound pulses produced by the solenoid valves from

a population of morphologically identified ventral-cord

neurons in various species of heliothine moths, H. vires-

cens and H. armigera included. Figure 8 demonstrates

one typical example from an H. virescens male—a ventral-

cord neuron that responded not only in synchrony with all

four events linked to the activated vales, but also when it

was exposed to a well-known ultrasound source, namely,

a bunch of jingling keys. The neuron had innervations in

the ventrolateral protocerebrum and an axon projecting

in the ventral cord (the location of the soma outside the

brain indicates the presence of an ascending neuron).

Unfortunately, we did not test the keys on the centrifugal

neurons presented here. However, the characteristic

transient responses during non-odor stimulation as

shown in both Figures 6 and 7 are quite comparable to

the responses of the ventral-cord neuron (Fig. 8).

Taking these data together, we can conclude that both

centrifugal neurons were activated by non-odorant air-

borne stimuli produced by the valve system. The sound

response as displayed by the arm neuron, combined with

the previous findings of sound responses from ventral-

cord neurons in the heliothine species, which were

Figure 7. Physiological characterization of the neuron found in the H. armigera male, i.e., the so-called arm neuron. A: The spontaneous

activity of the neuron was �20 Hz. The spiking activity demonstrates an excitatory response to the first of the four auditory pulses. The

occurrences of acoustic signals are, at all four events, indicated by a closed arrowhead below the final recording, whereas the estimated

start of air flow changes is shown at the first and last event by an open arrowhead. The onsets and offsets of the stimuli delivered via the

odor/air puff system are indicated by horizontal bars below each recording. Horizontal bar, 400 ms; vertical bar, 10 mV. B: Histograms

made by counting the number of spikes every 100 ms. Significant responses are indicated by stars. Horizontal bar, 400 ms; vertical bar,

30 Hz.

A multisensory centrifugal neuron

The Journal of Comparative Neurology |Research in Systems Neuroscience 163

Figure 8. Response properties of a ventral-cord neuron innervating the brain. A: Four excitatory and pronounced transient responses

occurring in synchrony with the four sound pulses produced by the valves can be seen. As shown, the neuron also responded strongly to

a well-known ultrasound source, i.e., a bunch of jingling keys. Horizontal bar, 400 ms; vertical bar, 10 mV. B: Confocal reconstruction of

the whole brain in a dorsal orientation showing the neuronal branches innervating the ventrolateral protocerebrum (PC) close to the border

of the antennal lobe (arrow; AL). C: Higher magnitude image in a frontal view showing the neuronal processes in the ventral part of the lat-

eral protocerebrum (arrow). D,E: AMIRA reconstructions of the neuronal branches in the brain and the subesophageal ganglion (SOG) in a

dorsal (D) and sagittal (E) orientation. The preparation was cut below the SOG and the axonal connection to the ventral cord is demon-

strated in the sagittal reconstruction (E). The location of the cell body outside the brain indicates the presence of an ascending ventral-

cord neuron. CB, central body; Ca, calyces; OL, optic lobe; P, posterior; A, anterior; D, dorsal; M, medial; Scale bar ¼ 100 lm in B–E.

Zhao et al.

164 The Journal of Comparative Neurology |Research in Systems Neuroscience

induced by the identical valve system, suggests the gen-

eral significance of acoustic stimuli related to this centrif-

ugal neuron type. Based on the stimulus conditions for

the non-odor response of the vir neuron, we cannot

exclude the possibility that air flow changes may also

influence the activity of the current neuron. However, no

matter whether the vir neuron was excited by sound or by

mechano-stimuli from air flow changes, the responses to

different stimulus modalities, as shown in Figure 6, dem-

onstrate for the first time a multimodal reaction from an

identified antennal-lobe centrifugal neuron in a moth

brain.

The fact that the two neurons responded to different

events in the stimulus sequence is difficult to interpret,

owing to the small dataset. As already mentioned, we find

it rather unlikely that two morphologically similar centrifu-

gal neurons would respond to completely dissimilar sen-

sory modalities. However, we found two to three closely

located somata, and the different physiological response

patterns of the neurons can thus be due to the possibility

that they are of distinct subtypes. Alternatively, the differ-

ence may reflect the fact that the two individuals pos-

sessing this neuron type belong to allopatric species liv-

ing on different continents. The behavioral context in

which the auditory sense has evolved has strongly deter-

mined the design and properties of the underlying neural

pathways in various insect species (reviewed by Stumper

and Helversen 2001). Some moths use ultrasound for

intraspecific acoustic communication in courtship, for

instance (Hoy and Robert, 1996; Miller and Surlykke,

2001), which is a newer evolutionary adaptation than that

ensuring a defense mechanism against a predator.

Regarding heliothine moths, it is not known whether H.

virescens males produce sounds, but the sympatric spe-

cies Helicoverpa zea, does (Kay 1969)—whereas H. armi-

gera is silent (Nakano et al., 2009). Thus, the different

responses may have to do with species-specific proper-

ties concerning threshold curves, adaption rates, or other

physiological distinctions characterizing these centrifugal

neurons. However, it might also be that the general

response pattern depends on previous neuronal activity,

because the spontaneous firing rate in the vir neuron

changed dramatically during application of the various

stimuli.

Functions of the centrifugalantennal-lobe neurons

The discovery of a centrifugal neuron type that

responds to non-odorant air-borne stimuli and targets the

brain olfactory center in two allopatric species of helio-

thines suggests its general significance for this subfamily

of noctuid moths. The finding of distinct responses that

are likely induced by acoustic stimuli is indeed interest-

ing, particularly when taking into account the number of

studies reporting that moths adjust their behavior accord-

ing to sensory input through the two main modalities,

odor and sound, in order to find a mate and to avoid echo-

locating bats (Baker and Card�e, 1977; Acharya and

McNeil, 1998; Surlykke et al., 1999; Skals et al., 2003b,

2005; Svensson et al., 2007; Anton et al., 2011).

Recently, Anton et al. (2011) reported data that may point

to central neuronal mechanisms responsible for this kind

of cross-modal integration. Their results showed, in addi-

tion to behavioral changes, long-term sensitization to ol-

factory stimuli of antennal-lobe projection neurons in the

noctuid moth, Spodoptera littoralis, based on pre-expo-

sure to pulsed ultrasonic bat calls.

This finding suggests that the sensitization depends on

modulatory protocerebral neurons. Both olfactory subsys-

tems, the plant odor and the pheromone system, were

influenced by bat sounds in the study of Anton et al.

(2011). Also, a previous study by Skals et al. (2003a),

reporting the influence of ultrasound from an odor

sprayer on the flight behavior of moths, points to a central

integration of information about odor and sound. Even

though the sound stimuli in our study were a side effect

of the odor stimulus setup, their frequency range overlaps

with the frequencies to which moth auditory organs are

most sensitive. The simple noctuid moth ear contains two

acoustic sensory neurons specialized for sensing echolo-

cating bats, both responding to frequencies between 10

and 60 kHz (Roeder, 1966; Surlykke et al., 1999). The

threshold for the most sensitive auditory sensory cell, A1,

is around 30–35 dB SPL in this frequency range in most

noctuid moths, with the other sensory cell, A2, being

around 20 dB less sensitive (Surlykke and Miller, 1982,

Surlykke et al., 1999). Thus, all four clicks from the valves

very likely exceeded the threshold for both sensory cells

in the ears of the moths.

It is also known from several moth species—H. vires-

cens included—that both sensory neurons are connected

to the brain, one directly and indirectly via interneurons,

and the other via interneurons only (Surlykke and Miller,

1982; Boyan and Fullard, 1986; Boyan et al., 1990). The

brain regions targeted by the afferent auditory projec-

tions are not yet mapped, however. The fact that the type

of neuron we have identified here responds to non-odor

air-borne stimuli, including sounds that are biologically

relevant for the moth ear in terms of sound pressure as

well as frequencies, in combination with its extensive pro-

jections into all glomeruli of the brain olfactory center,

makes it a possible candidate for providing integration

between the auditory and the olfactory system. Interest-

ingly, an early synaptic level of the mammalian olfactory

pathway, namely, the olfactory tubercle, has recently

A multisensory centrifugal neuron

The Journal of Comparative Neurology |Research in Systems Neuroscience 165

been reported to receive auditory cross-modal informa-

tion (Wesson and Wilson, 2010).

Various insect species with particular adaptations to

distinct ecological niches have often served as suitable

models for exploration of general neuronal principles.

Their nervous system, comprising such unique and identi-

fiable elements as the centrifugal neurons, offers the op-

portunity of revealing not only the functional circuits

underlying innate response patterns but also more com-

plex networks linked to learning and behavioral adapta-

tions. The finding of the antennal-lobe centrifugal neuron

presented here may be a key to understanding the neuro-

nal basis for moth behavior in response to multimodal

sensory stimulation. In order to determine in detail the

distinct physiological properties of the neuron, an experi-

mental approach similar to that employed here should be

utilized, dedicated, however, to specifically disentangling

the responses to different modalities.

ACKNOWLEDGMENTS

We thank Dr. Pal Kvello and Dr. Bjarte Bye Løfaldli,

Department of Biology, NTNU, for assistance with the

AMIRA reconstructions, PhD fellow Yan Jiang and Tim

Cato Netland, Department of Electronics and Telecommu-

nications, NTNU, for helping us with measurement of the

acoustic signals, and Jørgen Indahl Skoglund, Department

of Psychology, NTNU, for support with the air flow meas-

urements. We are also grateful to Syngenta (Basel, Swit-

zerland) and to Dr. Jun-Feng Dong, Henan University of

Science and Technology (Henan, China), for sending

insect pupae. Finally, we appreciate the constructive sug-

gestions from the two anonymous reviewers.

CONFLICT OF INTEREST STATEMENT

All authors confirm that there is no identified conflict of

interest including any financial, personal, or other relation-

ships with other people or organizations within three years

of beginning the submitted work that can inappropriately

influence, or be perceived to influence, their work.

ROLE OF AUTHORS

All authors had full access to all the data in the study and

take responsibility for the integrity of the data and the

accuracy of the data analysis. In addition, all authors have

contributed significantly to the elaboration of the paper.

Study concept and design: B.G. Berg and X.C. Zhao.

Acquisition of data: X.C. Zhao and G. Pfuhl; Analysis and

interpretation of data: X.C. Zhao, G. Pfuhl, A. Surlykke, J.

Tro, and B.G. Berg; Drafting of the manuscript: B.G. Berg

and X.C. Zhao; Critical revision of the manuscript for

important intellectual content: B.G. Berg, A. Surlykke, and

G. Pfuhl; Statistical analysis: X.C. Zhao; Obtained funding:

B.G. Berg, A. Surlykke, and G. Pfuhl; Administrative,

technical, and material support: B.G. Berg, A. Surlykke,

and J. Tro; Study supervision: B.G. Berg.

LITERATURE CITEDAcharya L, McNeil JN. 1998. Predation risk and mating behav-

ior: the responses of moths to bat-like untrasound. BehavEcol 6:552–558.

Anderson P, Hansson BS, Nilsson U, Han Q, Sj€oholm M, SkalsN, Anton S. 2007. Increased behavioral and neuronal sen-sitivity to sex pheromone after brief odor experience in amoth. Chem Sens 32:483–491.

Anton S, Homberg U. 1999. Antennal lobe structure. In: Hans-son BS, editor. Insect olfaction. Berlin: Springer. p97–124.

Anton S, Dufour MC, Gadenne C. 2007. Plasticity of olfactory-guided behaviour and its neurobiological basis: lessonsfrom moths and locusts. Entomol Exp Applicata 123:1–11.

Anton S, Evengaard K, Barrozo RB, Anderson P, Skals N.2011. Brief predator sound exposure elicits behavioral andneuronal long-term sensitization in the olfactory system ofan insect. Proc Natl Acad Sci U S A 108:3401–3405.

Arenas A, Giurfa M, Farina WM, Sandoz JC. 2009. Early olfac-tory experience modifies neural activity in the antennallobe of a social insect at the adult stage. Eur J Neurosci30:1498–1508.

Baker TC, Card�e R. 1977. Disruption of gypsy moth male sexpheromone behavior by high frequency sound. EnvironEntomol 7:45–52.

Boeckh J, Ernst KD, Sass H, Waldow U. 1984. Anatomical andphysiological characteristics of individual neurons in thecentral antennal pathway of insects. J Insect Physiol 30:15–26.

Boyan GS, Fullard JH. 1986. Interneurons responding to soundin the tobacco budworm moth Heliothis virescens (Noctui-dae): morphological and physiological characteristics.J Comp Physiol A 158:391–404.

Boyan G, Willias L, Fullard J. 1990. Organization in the tho-racic ganglia of noctuid moths. J Comp Neurol 295:248–267.

Br€aunig P. 1991. Subesophageal DUM neurons innervate theprincipal neuropils of the locust brain. Philos Trans R SocLond B 332:221–240.

Christensen TC, Hildebrand JG. 1987. Male-specific, sex-phero-mone-selective projection neurons in the antennal lobes ofthe moth Manduca sexta. J Comp Physiol A 160:553–569.

Dacks AM, Christensen TA, Agricola H-J, Wollweber L, Hilde-brand JG. 2005. Octopamine-immunoreactive neurons inthe brain and subesophageal ganglion of the hawkmothManduca sexta. J Comp Neurol 488:255–268.

Dacks AM, Christensen TA, Hildebrand JG. 2006. Phylogeny ofa serotonin-immunoreactive neuron in the primary olfactorycenter of the insect brain. J Comp Neurol 498:727–746.

Daly KC, Christensen TA, Lei H, Smith BH, Hildebrand JG.2004. Learning modulates the ensemble representationsfor odors in primary olfactory networks. Proc Natl Acad SciU S A 101:10476–10481.

Denker M, Finke R, Schaupp F, Grun S, Menzel R. 2010. Neu-ral correlates of odor learning in the honeybee antennallobe. Eur J Neurosci. 31:119–133.

Fan RJ, Anderson P, Hansson BS. 1997. Behavioural analysisof olfactory conditioning in the moth Spodoptera littoralis(Boisd.) (Lepidoptera: Noctuidae). J Exp Biol 200:2969–2976.

Gadenne C, Dufour M-C, Anton S. 2001. Transient post-mat-ing inhibition of behavioural and central nervous responsesto sex pheromone in an insect. Proc R Soc Lond B 268:1631–1635.

Zhao et al.

166 The Journal of Comparative Neurology |Research in Systems Neuroscience

Galizia CG, R€ossler W. 2010. Parallel olfactory systems ininsects: anatomy and function. Annu Rev Entomol 55:399–420.

Gatellier L, Nagao T, Kanzaki R. 2004. Serotonin modifies thesensitivity of the male silkmoth to pheromone. J Exp Biol207:2487–2496.

Giurfa M, N�unez J, Backhaus W. 1994. Odour and colour infor-mation in the foraging choice behaviour of the honeybee. JComp Physiol A 175:773–779.

Hammer M. 1993. An identified neuron mediates the uncondi-tioned stimulus in associative olfactory learning in honey-bees. Nature 366:59–63.

Han Q, Hansson B, Anton S. 2005. Interactions of mechanicalstimuli and sex pheromone information in antennal lobeneurons of a male moth, Spodoptera littoralis. J CompPhysiol A 191:521–528.

Hill ES, Iwano M, Gatellier L, Kanzaki R. 2002. Morphologyand physiology of the serotonin-immunoreactive putativeantennal lobe feedback neuron in the male silkmothBombyx mori. Chem Sens 2:475–483.

Hill ES, Okada K, Kanzaki R. 2003. Visualization of modulatoryeffects of serotonin in the silkmoth antennal lobe. J ExpBiol 206:345–352.

Hillier NK, Vickers NJ. 1987. Physiology and antennal lobeprojections of olfactory receptor neurons from sexually iso-morphic sensilla on male Heliothis virescens. J Comp Phys-iol A 193:649–663.

Hofbauer A, Ebel T, Waltenspiel W, Oswald P, Chen Y-C,Halder P, Biskup S, Ewandrowski U, Winkler C, SickmannA, Buchner S, Buchner E. 2009. The Wuerzburg hybridomalibrary against Drosophila brain. J Neurogenetics 23:78–91.

Homberg U, Muller U. 1999. Neuroactive substances in theantennal lobe In: Hansson BS, editor. Insect olfaction. Ber-lin: Springer. p181–206.

Homberg U, Montague RA, Hildebrand JG. 1988. Anatomy ofantenno-cerebral pathways in the brain of the sphinx mothManduca sexta. Cell Tissue Res 254:255–281.

Hoy RR, Robert D. 1996. Tympanal hearing in insects. AnnuRev Entomol 41:433–450.

Iwama A, Shibya T. 1998. Physiology and morphology of olfac-tory neurons associating with the protocerebral lobe of thehoneybee brain. J Insect Physiol 44:1191–1204.

Jørgensen K, Stranden M, Sandoz J-C, Menzel R, MustapartaH. 2007. Effects of two bitter substances on olfactory con-ditioning in the moth Heliothis virescens. Exp Biol 210:2563–2573.

Kanzaki R, Shibuya T. 1986. Identification of the deutocere-bral neurons responding to the sexual pheromone in themale silkworm moth brain. Zool Sci 3:409–418.

Kay RE. 1969. Acoustic signalling and its possible relationshipto assembling and navigation in the moth, Heliothis zea. JInsect Physiol 15:989–1001.

Kent KS, Hoskins SG, Hildebrand JG. 1987. A novel serotonin-immunoreactive neuron in the antennal lobe of the sphinxmoth Manduca sexta persists throughout postembryoniclife. J Neurobiol 18:451–465.

Kirschner S, Kleineidam CJ, Zube C, Rybak J, Grunewald B,R€ossler W. 2006. Dual olfactory pathway in the honeybee,Apis mellifera. J Comp Neurol 499:933–952.

Klagges BR, Heimbeck G, Godenschwege TA, Hofbauer A,Pflugfelder GO, Reifegerste R, Reisch D, Schaupp M, Buch-ner S, Buchner E. 1996. Invertebrate synapsins: a singlegene codes for several isoforms in Drosophila. J Neurosci16:3154–3165.

Kloppenburg P, Heinbockel T. 2000. 5-Hydroxytryptaminemodulates pheromone-evoked local field potentials in mac-roglomerular complex of the sphinx moth Manduca sexta. JExp Biol 203:1701–1709.

Kloppenburg P, Mercer AR. 2008. Serotonin modulation ofmoth central olfactory neurons. Annu Rev Entomol 53:179–190.

Kloppenburg P, Ferns D, Mercer AR. 1999. Serotonin enhan-ces central olfactory neuron responses to female sex pher-omone in the male sphinx moth Manduca sexta. J Neurosci19:8172–8181.

Linn CE, Campbell MG, Poole KR, Wu W-Q, Roelofs WL. 1996.Effects of photoperiod on the circadian timing of phero-mone response in male Trichoplusia ni: relationship to themodulatory action of octopamine. J Insect Physiol 42:881–891.

Menzel R. 2001. Searching for the memory trace in a mini-brain, the honeybee. Learn Mem 8:53–62.

Miller LA, Surlykke A. 2001. How some insects detect andavoid being eaten by bats: tactics and countertactics ofprey and predator. BioScience 51:571–582.

Nakano R, Takanishi T, Fujii T, Skal N, Surlykke A, Ishikawa Y.2009. Moths are not silent, but whisper ultrasonic court-ship songs. J Exp Biol 212:4072–4078.

Raguso RA, Willis MA. 2002. Synergy between visual and ol-factory cues in nectar feeding by naıve hawkmoths, Man-duca sexta. Anim Behav 64:685–695.

Reichmuth C, Becker S, Benz M, Debel K, Reisch D, HeimbeckG, Hofbauer A, Klagges B, Pflugfelder GO, Buchner E.1995. The sap47 gene of Drosophila melanogaster codesfor a novel conserved neuronal protein associated withsynaptic terminals. Mol Brain Res 32:45–54.

Rø H, Muller D, Mustaparta H. 2007. Anatomical organizationof antennal lobe projection neurons in the moth Heliothisvirescens. J Comp Neurol 500:658–675.

Roeder KD. 1966. Auditory system of noctuid moths. Science154:1515–1521.

Salecker I, Distler P. 1990. Serotonin-immunoreactive neuronsin the antennal lobes of the American cockroach Peripla-neta americana: light- and electronmicroscopic observa-tions. Histochemistry 94:463–473.

Saumweber T, Weyhersmueller A, Hallermann S, DiegelmannS, Michels B, Bucher D, Funk N, Reisch D, Krohne G,Wegener S, Buchner E, Gerber B. 2011. Behavioral andsynaptic plasticity are impaired upon lack of the synapticprotein SAP47. J Neurosci 31:3508–18.

Schachtner J, Schmidt M, Homberg U. 2005. Organization andevolutionary trends of primary olfactory brain centers inTetraconata (Crustacea þ Hexapoda). Arthropod StructDev 34:257–299.

Skals N, Plepys D, El-Sayed AM, L€ofstedt C, Surlykke A.2003a. Quantitative analysis of the effects of ultrasoundfrom an odor sprayer on moth flight behavior. J Chem Ecol29:71–82.

Skals N, Plepys D, Lofstedt C. 2003b. Foraging and mate-find-ing in the silver Y moth, Autographa gamma (Lepidoptera:Noctuidae) under the risk of predation. OIKOS 102:351–357.

Skals N, Andersen P, Kanneworff M, L€ofstedt C, Surlykke A.2005. Her odors make him deaf: crossmodal modulationof olfaction and hearing in a male moth. J Exp Biol 208:595–601.

Stocker RF, Lienhard MC, Borst A, Fischbach K-F. 1990. Neu-ronal architecture of the antennal lobe in Drosophila mela-nogaster. Cell Tissue Res 262:9–34.

Stumper A, Helversen D. 2001. Evolution and function of auditorysystems in insects. Natrurwissenschaften 88:159–170.

Sun XJ, Tolbert LP, Hildebrand JG. 1993. Ramification patternand ultrastructural charateristics of the serotonin-immuno-reactive neuron in the antennal lobe of the moth Manducasexta: a laser scanning confocal and electron microscopicstudy. J Comp Neurol 338:5–16.

A multisensory centrifugal neuron

The Journal of Comparative Neurology |Research in Systems Neuroscience 167

Surlykke A. 1988. Interaction between echolocating bats andtheir prey. In: Nachtigall PE, Moore PWB, editors. Americansonar: processes and performance. New York: PlenumPress. p 551–566.

Surlykke A, Miller LA. 1982. Central branches of three sen-sory axons from a moth ear (Agrotis segetum, Noctuidae).J Insect Physiol 28:357–364.

Surlykke A, Filskov M, Fullard JH, Forrest E. 1999. Auditoryrelationships to size in noctuid moths: bigger is better.Natrurwissenschaften 86:238–241.

Svensson GP, Lofstedt C, Skals N. 2007. Listening in pher-omone plumes: disruption of olfactory-guided mate

attraction in a moth by a bat-like ultrasound. J InsectSci 7:59.

Wesson DW, Wilson DA. 2010. Smelling sounds: olfactory-auditory sensory convergence in the olfactory tubercle.J Neurosci 30:3013–3021.

Zeiner R, Tichy H. 1998. Combined effects of olfactory andmechanical inputs in antennal lobe neurons of the cock-roach. J Comp Physiol A 182:467–473.

Zhao XC, Berg BG. 2009. Morphological and physiologicalcharacteristics of the serotonin-immunoreactive neuron inthe antennal lobe of the male oriental tobacco budworm,Helicoverpa assulta. Chem Sens 34:363–372.

Zhao et al.

168 The Journal of Comparative Neurology |Research in Systems Neuroscience