Adaptation in pheromone-sensitive trichoid sensilla of the hawkmoth Manduca sexta

While the transduction cascade of pheromone-sensitive olfactory receptor neurons (ORNs) in insects has been thoroughly investigated in vitro(Breer et al., 1988, 1990; Boekhoff et al., 1990, 1993; Stengl et al., 1992; Stengl, 1993, 1994; Wegener et al., 1993, 1997), little is known about olfactory adaptation and the mechanisms involved. Adaptation is a universal characteristic of receptor cells of all sensory modalities(Burkhardt, 1961). There are different definitions for the term `adaptation' but implicit in any definition is a decrease in sensitivity due to the influence of a previous (conditioning)stimulus (Zack, 1979).

Moths distinguish pheromone mixtures according to the concentration ratios of the different pheromone components. Thus, differentiation of pheromone concentrations is very crucial for recognition of prospective mates. Turbulences and wind velocities determine the structure of the pheromone filaments that stimulate the antenna of a flying moth. Thus, adapting and non-adapting pheromone stimuli of variable concentrations and stimulus durations reach different parts of the antenna at various time intervals. It is still unknown how the moth can recognize relevant pheromone ratios in various states of adaptation. In our study of pheromone-sensitive ORNs, we examine how adapting pheromone stimuli affect the encoding of different pheromone concentrations (quantity coding) in the intact moth. We distinguish the rapidly (within seconds to minutes) reversible reduction of sensitivity due to prior stimulation (short-term adaptation) from the decline in excitation, as seen during a phasic–tonic response to a stimulus of long duration (desensitization; Zufall and Leinders-Zufall, 2000). Thus, we compare quantity coding in response to short (50 ms) and long (1000 ms) pheromone stimuli, as possibly encountered during flight to the calling female. We do not examine the more slowly occurring (within several minutes to hours) reduction of sensitivity due to previous strong stimulation as seen during long-term (i.e. long-lasting) adaptation (Ziegelberger et al., 1990; Marion-Poll and Tobin, 1992; Boekhoff et al.,1993; Stengl et al.,2001; Dolzer,2002).

Quantity coding in response to adapting pheromone stimuli has been studied in extracellular tip recordings in saturniid moths, in which the different terms `short-term' versus `long-term' adaptation were coined for insect olfaction (Zack, 1979; Zack-Strausfeld and Kaissling,1986; Kaissling et al., 1986, 1987). Concerning ORNs of the hawkmoth Manduca sexta, however, there is only one study about olfactory adaptation in temporal coding of pheromone pulse trains(Marion-Poll and Tobin, 1992)and a few studies on quality odour coding(Kaissling et al., 1989; Kalinová et al.,2001).

Moths detect the pheromones using specialized ORNs, which innervate long multiporous trichoid sensilla on the antenna in pairs(Keil, 1989). It has been shown that in each trichoid sensillum, one of the ORNs (the bombykal cell)responds to bombykal, the main component in the conspecific pheromone blend(Starratt et al., 1979; Tumlinson et al., 1989). The second cell (the non-bombykal cell) is tuned to other different pheromone components in different sensilla. While the study by Kaissling et al.(1989) examined the coding of different odour qualities, not much attention was paid to the coding of odour concentrations. Quantification of pheromone responses in M. sexta was not reliable, because at that time it could not be consistently distinguished whether an action potential originates from the spontaneously active non-bombykal cell or from the bombykal cell. A more recent study of trichoid sensilla in M. sexta could distinguish two different nerve impulse classes, termed the small and the large action potential classes, by amplitude(Dolzer et al., 2001). However, it was still not completely resolved whether the bombykal cell always generates the large or the small action potentials.

Thus, the present study first examines which action potential class the bombykal cell from a specific region of a distal flagellar annulus belongs to,since a clear distinction of both cells is a prerequisite for an unequivocal quantification of the bombykal responses. Then, we examine how a previous strong stimulus (a conditioning stimulus) can adapt and thus change the coding of bombykal concentrations at the level of the sensillar potential and the action potential response and whether these stimuli affect flight behaviour. In dose–response curves to short (short-term adaptation) and long(desensitization) bombykal stimuli, we show that the bombykal cell can resolve four log₁₀-units of pheromone concentrations but is apparently unable to encode stimulus duration with all parameters tested. Finally, we present evidence for the presence of several functionally distinct mechanisms of short-term adaptation, which affect the rising phase of the sensillar potential, its decline and the action potential response. In addition, we show at least one additional mechanism of desensitization. None of these bombykal stimuli affected flight behaviour.

Manduca sexta (Johannson) moths (Lepidoptera: Sphingidae) were raised from eggs, feeding on an artificial diet (modified from Bell and Joachim, 1976). The animals were kept under a long-day photoperiod (17 h:7 h L:D; lights off at 8.00 h) at 24–27°C and 40–60% relative humidity. Male pupae were isolated one day before emergence, gently cleaned with 70% ethanol and allowed to hatch without contact with pheromone. During their second dark phase, the adults were carefully fixed in a teflon holder without the necessity for anaesthesia. The flagellum of the right antenna was immobilized with dental wax (Boxing wax; Sybron/Kerr, Romulus, MI, USA), and the 15–20 most-apical annuli of the flagellum were clipped off. A glass electrode filled with haemolymph Ringer(Kaissling, 1995) was inserted into the flagellar lumen and sealed with electrocardiogram electrode gel (PPG;Hellige, Freiburg, Germany). The tips of long trichoid sensilla from the apical row on the 2nd–10th remaining distal annulus were removed using sharpened forceps. The recording electrode, filled with sensillum lymph Ringer(Kaissling, 1995), was slipped over one sensillum. To minimize contributions of the electroantennogram, we recorded from an annulus close to the tip of the haemolymph electrode. The connection to the amplifier inputs was established with Ag/AgCl wires immersed in the electrolytes. Signals were amplified approximately 200-fold in a custom-built amplifier (0 Hz–2 kHz; input impedance 10¹²Ω) and passed through a 2 kHz anti-aliasing filter (900C/9L8L; Frequency Devices, Haverhill, MA, USA). Flight activity of the animals was monitored using a piezo-electric element (PI Ceramic, Lederhose, Germany) placed at the thorax. For data acquisition, a Digidata 1200 B digitizer (Axon Instruments,Union City, CA, USA) and pCLAMP software (versions 6–8) from the same manufacturer were used. The electrophysiological signal and a highpass-filtered equivalent (cut-off frequency 2 Hz or 5 Hz), as well as the piezo signal, were continuously recorded on a strip chart recorder (WindoGraf or EasyGraf, Gould, Valley View, OH, USA). Sections of the recordings were additionally stored on FM tape (Store 4 D; RACAL, Bergisch Gladbach, Germany)or digital audio tape (DTR-1202, Bio-Logic, Claix, France). Voltage polarity is given with the sensillum lymph electrode in reference to the haemolymph electrode. After the recordings, >90% of the moths were in sufficiently good condition to be returned to the flight cage.

All recordings were performed at room temperature (18–23°C). Charcoal-filtered and moistened air was permanently blown over the preparation through a glass cartridge (13 l min^–1). The airstream could be redirected through cartridges containing a piece of filter paper(approximately 1 cm²) loaded with synthetic bombykal(E,Z-10,12-hexadecadienal) generously provided by T. Christensen (Tucson, AZ,USA). Bombykal doses are always given in log₁₀-unit intervals; for example, in Fig. 9, –6 log₁₀ bombykal dose (μg) means 10^–6 μg bombykal. The cartridges were placed with the outlet in a distance of 4.5–6 cm from the recording site. The airstream was switched between the cartridges using solenoid valves (JFMH-5-PK and MFH-5-1/8; FESTO, Esslingen,Germany) controlled by the computer. Airstream velocity was monitored with a thermistor (BC32L1; Fenwal, Framingham, MA, USA) placed near the recording site and connected to a custom-built anemometer. Limited by the specified switching time of the solenoids (15 ms), and verified by the anemometer recording, the shortest applicable stimulus duration was approximately 50 ms. Doses between 10^–6 μg and 100 μg bombykal dissolved in n-hexane (Merck, Frankfurt, Germany) were applied to the filter papers (10μl or 100 μl per paper), and the solvent was allowed to evaporate. Stimulus intensity is always given in terms of the bombykal dose applied to the filter paper. Stimulus durations between nominally 10 ms and 1000 ms were employed. Due to the switching times of the solenoid valves, stimulus durations below 50 ms are less clearly defined than above this duration and are only included for completeness.

A suction tube of 10 cm diameter was placed below the animal to rapidly remove the pheromone after the stimulations and to avoid uncontrolled stimulation by bombykal leaking out of the stimulus cartridges, which were put in place as short as possible (5–30 s) prior to stimulation. Between the recording sessions, the cartridges were stored at –80°C in individual glass scintillation vials. Control cartridges loaded with hexane alone were prepared and treated the same way. When stored together with the vials containing the bombykal cartridges, the control cartridges elicited weak responses after several recording sessions, indicating a cross contamination that is also likely to have occurred for the lower stimulus doses (Figs 7, 8, 9). This effect is presumably enhanced by molecules of the highly hydrophobic pheromone component adhering to parts of the electrophysiological set-up. These contaminations influence the baseline of the dose–response curves, as well as the peri-stimulus-time histograms at low pheromone doses, but they do not affect the main conclusions of our study. A set of stimulus cartridges was used for 5–20 recording sessions; the control cartridges were replaced more frequently.

While cartridge stimulation, as described above, was used in most of the experiments, a small subset of the sensilla was stimulated locally. In these experiments, the pheromone was applied to a piece of thread, which was then inserted into a glass pipette with an opening of approximately 40 μm(Kaissling, 1995). The stimulation pipette was placed below the recorded sensillum, and the stimulus airstream was redirected through the pipette for stimulation. The only dose tested with local stimulation was 1 μg bombykal; stimulus durations were between 20 ms and 2000 ms. Of the data presented here, only the recordings shown in Fig. 10 were obtained using local stimulation; all other data and results were obtained with cartridge stimulation.

In the beginning of each recording, a series of 5 mV calibration pulses was applied to the haemolymph electrode, which was otherwise grounded. The pheromone responses were recorded in sweeps of 3 s duration at sampling frequencies of 5 kHz and 1.67 kHz (Clampex, episodic stimulation mode with a gear shift after 1000 ms). The solenoids were controlled by a digital output signal switched to `high' 20 ms after the sweep start. The stimulus airstream,as monitored by the anemometer, arrived at the recording site approximately 50 ms after the trigger signal.

The recordings were evaluated using macros in Clampfit 6, Microsoft Excel,versions 7 and 8, and Automate 4 (Unisyn Software; Los Angeles, CA, USA)(Dolzer, 2002). For the analysis of the sensillar potential, the responses were lowpass-filtered at a cut-off frequency of 50 Hz or 70 Hz (Clampfit, Gaussian filter). The evaluated parameters of the sensillar potential, as illustrated in Fig. 1A,B, were: (1) the overall amplitude (SP amplitude), (2) the initial slope between the onset of the sensillar potential and the half-maximal SP amplitude (slope), (3) the half-time of the rising phase (t_{1/2 rise}) and (4) the half-time of the declining phase (t_{1/2 decline}). The half-times of the rising and declining phases were analyzed to compare our results with studies on silkmoths (Antheraea sp.; Zack, 1979; Kaissling et al., 1987; Kodadová, 1993; Kodadová and Kaissling,1996), and the initial slope was analyzed as an important parameter describing the kinetics of the sensillar potential.

For the analysis of the action potentials(Fig. 1C), the lowpass-filtered trace was subtracted from the original response. This pseudo-highpass filtering procedure, in contrast to actual highpass filters, does not distort the shape of the action potentials and therefore allows the analysis of their amplitude and waveform (Dolzer,2002). The action potential response was characterized by: (1) the peak frequency computed from the first five interspike intervals (AP frequency; Fig. 1C) and (2) the latency between the beginning of the sensillar potential and the first action potential (AP latency; Fig. 1B). In the background activity, action potentials of two amplitude classes were recorded (Dolzer et al., 2001). Therefore, when action potentials of two classes,distinct by amplitude or by frequency and latency, were observed during the response, they were analyzed separately. Because responses of trichoid sensilla from different intact moths varied greatly in amplitude(Table 1), it was impossible to quantify these variable data sets without normalization. Thus, for the quantitative analysis, the response parameters were normalized to the highest response during the recording from the same sensillum. The highest response was defined as the largest value of those parameters that were positively correlated to the stimulus intensity and as the smallest value of parameters,which were negatively correlated to the stimulus intensity. Thus, the normalized responses were computed from: Response_norm =Response/Response_max and Response_norm =Response_min/Response, respectively. The sensillar potential amplitude, the initial slope of the SP, t_{1/2 decline} and the action potential response were normalized to the largest value during each recording from the same sensillum. Since t_{1/2 rise} and AP latency were negatively correlated with stimulus strength, both were normalized to the smallest value during one recording. The inverse normalization of the negatively correlated parameters (t_{1/2 rise} and AP latency) allowed us to focus on responses in the physiological range. Direct normalization of, for example, AP latency, either to the largest or smallest value of the data set, exaggerates long latencies in response to low stimulus intensity and obscures differences in the response to physiological stimuli of a higher dose. For the investigation of desensitization, stimuli of 1000 ms duration were applied with 60 s between consecutive stimulations. The action potential response characteristics were analyzed with peri-stimulus-time histograms to evaluate changes in the action potential frequency over time. In time windows (bins) of 10 ms, action potential responses were added up and plotted over time, with t=0 being the start of the DC response.

In total, 1462 stimulations (all doses including controls) were applied to 70 sensilla of 42 animals. Since the experiments were done with a whole-animal preparation, no rundown of the responses was observed, even during recordings of more than 3 h with repetitive stimulation.

Non-adapted responses were obtained by applying either 50 ms stimuli or 1000 ms stimuli of increasing bombykal dose, separated by 60 s (dose ramp). Control stimuli with only hexane on the filter paper were applied before each dose ramp. In some recordings, additional control stimuli were applied. Adapted responses were achieved by conditioning the sensilla with 10 μg stimuli of 250 ms duration, applied 20 s before the test stimuli of 50 ms duration (Fig. 2). For the 1000 ms stimuli, the sensilla were adapted by 10 μg stimuli of 1000 ms, applied 60 s before the test stimuli. To ensure identical conditions, each test stimulus was preceded by its own conditioning stimulus. The sensilla were allowed to recover for at least 10 min between every two stimulus pairs to avoid accumulative adaptation. This interval between the stimulus pairs was chosen because pilot experiments had suggested the complete recovery from adaptation by a stimulus as used in the adaptation protocol after approximately 5 min (Dolzer,1996). To avoid contamination of our short-term adaptation experiments by long-term adaptation (which lasts for hours), we did not employ pheromone doses higher than 10 μg bombykal for the 50 ms stimuli and we used a maximal dose of 100 μg for the stimuli of 1000 ms duration.

In extracellular tip recordings from long trichoid sensilla, we investigated odour quantity coding by measuring the electrical responses to stimulation with bombykal at different doses. Slow, amplitude-modulated bombykal-dependent sensillar potentials with superimposed, fast,frequency-modulated action potentials were quantified. We wanted to know whether the previous exposure to a conditioning stimulus or the duration of bombykal stimuli affect quantity coding and influence flight behaviour. The responses of, in total, 70 sensilla of 42 animals to bombykal doses between 10^–6 μg and 100 μg were characterized and compared in different states of adaptation. Most experiments were only performed on a subset of the sensilla.

Bombykal application causes depolarizing sensillar potentials in olfactory receptor neurons (ORNs), which elicit action potential responses that are carried along the axons of the ORNs to the brain to excite postsynaptic antennal lobe neurons. In extracellular tip recordings, the sensillar potential (SP) in response to pheromone stimulation is measured as a negative deflection of the transepithelial potential. According to previous practice and because the SP is assumed to reflect the depolarizing receptor potential of the ORNs, its increase to maximal amplitude is described as the SP's rising phase. Its repolarization to the steady-state potential is described as its decline. Superimposed on the SP, action potentials can be recorded (Figs 3, 4, 5, 6). The amplitude of action potentials decreased in response to strong stimulation (Figs 1C, 3, 4, 5). When action potentials of two classes occurred during a response (Figs 3B, 4A,B), they are referred to as small and large action potentials according to the amplitude of the first action potential, i.e. before the stimulus-correlated amplitude reduction took place. This terminology was introduced previously in recordings of spontaneously active ORNs (Dolzer et al.,2001). Action potentials of two amplitude classes were also observed between the stimulations. All action potentials, whether spontaneous or in a pheromone response, were recorded with their positive phase first. There was a large variability in both the sensillar potential and action potential responses among individual sensilla from different moths(Table 1). Sixty-nine of the 70 sensilla tested responded to bombykal. The only non-responding sensillum had no spontaneous action potential activity, indicating that it was damaged.

To determine whether stimulation with the different stimuli of the main pheromone component bombykal alone elicits flight activity in the intact,tethered moth, flight activity was recorded with a piezo-electric element placed at the thorax. In addition, since a correlation between flight activity and transepithelial potential oscillations was observed previously(Dolzer et al., 2001), the transepithelial potential was continuously monitored between the stimulations. In 18 of 22 animals analyzed, bombykal stimulation did not elicit flight activity. In three cases, the animals exhibited continuous flight activity. In only one animal, flight activity occurred immediately after strong stimulation, whereas the animal was silent before. Oscillations of the transepithelial potential were observed in the recordings from 16 sensilla. In no case did bombykal stimulation suppress or detectably influence the oscillations.

Because the two ORNs per trichoid sensillum can be clearly distinguished according to the amplitude of their action potentials, we wanted to know whether the bombykal cell always produces the large or the small action potentials. Unless the stimulus dose exceeded 10 μg, only action potentials of the large amplitude class were observed during the bombykal responses. When small action potentials occurred at these dosages, they could not be distinguished from spontaneously occurring action potentials by their frequency or reduction of the amplitude. At doses of 10 μg bombykal,however, the small action potentials occurred at a frequency that could not be considered spontaneous in 22 of 68 stimuli of 50 ms duration(Fig. 3; Table 2). Apparently, the non-bombykal cell was `cross-excited'. In the presence of small action potentials, the SP amplitude was significantly larger (10.85±0.64 mV, N=22) than in responses with only large action potentials(8.75±0.50 mV; means ± S.E.M.; P<0.01, N=46; Student's t-test). The action potential frequency of the small action potentials was significantly lower and the action potential latency was significantly longer than for the large action potentials of the bombykal cell (Table 3).

When cross-excitation occurred during a response to 10 μg bombykal, the stimulus-correlated amplitude reduction of the small action potentials was less prominent than that of the bombykal cells(Fig. 3B). At lower doses, the amplitude and frequency of small action potentials occurring during the response was the same as during spontaneous activity(Fig. 4).

After distinction of the action potential responses of the bombykal and the non-bombykal cells, we measured dose–response relationships of the bombykal cell in response to 50 ms bombykal stimuli to determine the threshold, range of resolution and saturation (Figs 5, 7). The bombykal responses varied considerably among individual sensilla and among individuals(Table 1). Therefore, for a quantitative analysis it was necessary to normalize the response parameters(see Materials and methods; Fig. 7). The SP amplitude, the initial slope of the SP and the half-time of the decay phase (t_{1/2 decline}), as well as the action potential frequency and latency, but not the half-time of the rising phase (t_{1/2 rise}), were dose-dependent (Figs 5, 6, 7). With higher pheromone concentrations, the amplitude of the SP increased, its initial slope became steeper and its decline to baseline became slower(Fig. 7). The action potential frequency increased, and the latency to the first action potential was shortened. The threshold was between 10^–3 μg and 10^–2 μg bombykal for all these variables. The ORN was able to resolve at least four log₁₀-units of odour concentrations. The t_{1/2 rise} exhibited no dose dependence(Fig. 7C).

To evaluate changes in quantity coding after previous pheromone experience,the sensilla were short-term adapted. Short-term adaptation is defined here as a rapidly (within seconds to minutes) reversible reduction of sensitivity due to prior stimulation with a conditioning stimulus. Short-term adaptation is distinguished from long-term adaptation, which lasts for hours and involves rises of intracellular cyclic GMP (cGMP) concentrations(Stengl et al., 2001). When the sensilla were adapted by a strong conditioning stimulus (10 μg bombykal; 250 ms duration) 20 s prior to the test stimulus, the SP amplitude and the action potential frequency were reduced (Figs 6, 7A,E). In addition, the decay of the sensillar potential was accelerated (Figs 6B, 7D), as characterized by a faster t_{1/2 decline}. The dose–response curves of the SP amplitude and the initial slope were shifted to higher stimulus intensities by approximately one log₁₀-unit(Fig. 7A,B), which means that 10 times more pheromone was needed to elicit the same response amplitudes. The t_{1/2 rise} was virtually dose-independent at a stimulus duration of 50 ms and was not significantly altered by the adapting stimulus(Figs 6B, 7C). The dose–response curve of the action potential frequency was shifted by two log₁₀-units, and the shift of the dose–response curve of the latency between the onset of the SP and the first action potential was even larger (Fig. 7E,F). For none of the variables was there an obvious difference in the slopes of the adapted and the non-adapted dose–response curves.

Desensitization is the decline in excitation as seen during a phasic–tonic response to a stimulus of long duration and, thus, is distinct from short-term adaptation (Zack,1979; Zufall and Leinders-Zufall, 2000). To determine whether the stimulus duration affects coding of pheromone quantity in the non-adapted and adapted state, we employed pheromone stimuli of 20 times longer durations. With stimulations of 1000 ms duration in the absence of a preceding adapting stimulus(N=129), the action potential response was phasic–tonic at all doses between 10^–6 μg and 100 μg bombykal(Fig. 8). The phasic component was more prominent with increasing pheromone dose. Thus, this stimulus duration was not reliably encoded by the phasic part of the action potential response. After the phasic peak, the action potential frequency declined to a tonic plateau with a time constant of approximately 150 ms. Thresholds were between 10^–2 μg and 10^–1 μg bombykal for all variables, and approximately four log₁₀-units of bombykal concentrations were resolved (Fig. 9).

In contrast to short stimuli of 50 ms duration, in response to 1000 ms stimuli, the t_{1/2 rise} was dose-dependent at higher doses of bombykal in the non-adapted state as well as in the adapted state(Fig. 9C; Dolzer, 1996). As for short stimuli, after a preceding adapting stimulus of 10 μg bombykal, the amplitude of the sensillar potential was decreased, the initial slope was less steep and the frequency of the action potential response was reduced(Fig. 9A,B,E). In contrast to short stimuli, the t_{1/2 rise} was further prolonged after an adapting stimulus, the decline not further accelerated at all stimulus concentrations and the action potential latency not further increased at all high pheromone concentrations (Fig. 9C,D,F).

After testing responses to different stimulus concentrations, we examined how the bomykal cell can encode different stimulus durations. The stimulus duration of 50 ms was neither encoded by the sensillar potential nor by the duration of the phasic or tonic part of the action potential response (Figs 5, 10). When different stimulus durations from 20 ms to several seconds were tested (at a dose of 10 μg with an interstimulus interval of 60 s), none of the parameters of the sensillar potential reliably encoded all stimulus durations (Figs 10, 11). Only stimulus durations of >100 ms were distinguished by t_{1/2 decline} and by the total number of action potentials, but not by the SP amplitude, t_{1/2 rise} and action potential frequency (i.e. the first five interspike intervals; Fig 11).

In extracellular tip recordings of long trichoid sensilla, we investigated the electrical responses to stimulation with bombykal, the main component in the conspecific pheromone blend. While previous studies on pheromone sensilla in M. sexta focused on the specificity of the olfactory receptor neurons (ORNs; Kaissling et al.,1989; Kalinová et al.,2001) and on temporal coding in the action potential response(Marion-Poll and Tobin, 1992),we characterized and compared the dose dependence of the sensillar potential and action potential responses in different states of adaptation. Because the bombykal cell in each sensillum always produced the action potentials of larger initial amplitude, distinct quantification of bombykal responses was possible. Thus, it could be shown that different mechanisms of adaptation occur in M. sexta at different levels of the transduction cascade: at the generation of sensillar potentials and at the transformation into action potentials. In responses to short and long pheromone stimuli, a previous strong bombykal stimulus decreased the amplitude and the initial slope of the sensillar potential and accelerated its decline. In addition, the initial action potential frequency was reduced even further than the parameters of the sensillar potential, suggesting that an additional adaptation mechanism acts at the level of action potential generation. At least one adaptation mechanism is stimulus-duration-dependent, because for long, but not short, pheromone stimuli the same adapting stimulus slowed t_{1/2 rise} at higher stimulus concentrations. Despite the fact that stimulus duration strongly affects flight behaviour when freely flying M. sexta moths are stimulated with the native pheromone blend(Vickers et al., 2001), none of the bombykal stimulus protocols changed flight activity of the moth. Whether these different adaptation mechanisms have a distinct molecular basis and resemble the mechanisms described in vertebrate olfaction (reviewed in Zufall and Leinders-Zufall,2000) remains to be examined.

In vertebrate olfaction, three stages of adaptation are distinguished in current recordings in vitro by their effects and the underlying mechanisms: short-term adaptation, desensitization and long-lasting adaptation(Zufall and Leinders-Zufall,2000). In vertebrates, short-term adaptation reduces the amplitude of brief odour-dependent currents after previous stimuli, is Ca²⁺-dependent and cGMP-independent, and declines within several seconds. By contrast, desensitization, which reduces the amplitude and slows down the kinetics of the odour response during maintained stimulation, is Ca²⁺-dependent and calcium/calmodulin-dependent protein kinase (CaM kinase)-dependent and lasts longer than short-term adaptation. Long-lasting adaptation, however, is triggered by stronger stimulation, is Ca²⁺-, CO₂- and cGMP-dependent and lasts many minutes. In the following, we will discuss the general characteristics of the bombykal cell in M. sexta and the different mechanisms of adaptation in moths in the context of known olfactory signal transduction pathways, to look for common features between insects and vertebrates and to suggest a testable hypothesis of olfactory adaptation in insects.

As described previously (Dolzer et al.,2001; Kalinová et al.,2001), but in contrast to other studies in M. sexta(Kaissling et al., 1989; Marion-Poll and Tobin, 1992),two classes of action potentials, which can probably be assigned to the two ORNs in each sensillum (Keil,1989), could be distinguished by their amplitude in most of the recordings. Our experiments showed that in M. sexta, as in other moth species investigated, the ORN that fires the action potentials with the larger amplitude responds to the main component of the conspecific pheromone blend(Bombyx mori, Kaissling et al.,1978; Antheraea polyphemus and A. pernyi, Zack, 1979; Mamestra suasa, Lucas and Renou,1989; Mamestra brassicae, Renou and Lucas, 1994). Only at very high pheromone doses was the cell with the smaller action potential amplitude excited as well. Such cross-excitation was also described in A. polyphemus (Kodadová and Kaissling, 1996; Kodadová, 1993).

Interestingly, none of the sensillar potential parameters and neither the phasic nor the tonic part of the action potential response encoded all stimulus durations tested. Only longer stimulus durations were distinguished by the overall action potential frequency but were not reflected by the duration of the action potential response. From analysis of the odour plume and from antennal lobe recordings, it is known that odour intensity and duration vary rapidly in nature and can be resolved by M. sextaantennal lobe neurons on a millisecond time scale (Christensen et al., 1996, 1998; Vickers et al., 2001). Thus,in their natural environment, moths have to resolve rapid changes in stimulus strength and duration but might not need to respond reliably to ongoing odour stimulation over a period of seconds to minutes.

In wind-tunnel assays, Tumlinson et al.(1989) found that both bombykal and (E,E,Z)-10,12,14-hexadecatrienal are required to elicit a sequence of anemotaxis and mating behaviour. They did not test for activation,however. Our results suggest that bombykal alone is not sufficient to activate the moths. Neither did bombykal stimulation elicit flight activity nor suppress oscillations of the transepithelial potential, as would have been expected if activation is assumed to be correlated with octopamine release(Roeder, 1999; Dolzer et al., 2001). Although the experiments were performed during the active phase of the nocturnal moths(reared in an inverse culture), we cannot exclude influences of the preparation procedure or the unnatural situation during the recordings. Thus,this finding needs to be confirmed in experiments with freely behaving moths. Finally, it remains to be tested in behavioural experiments how the different forms of adaptation affect recognition of the species-specific pheromone blend.

In our study, the rising phase of the sensillar potentials (which in the extracellular recording appears as a negative deflection from baseline to the maximal amplitude) was characterized using two different parameters, the initial slope and t_{1/2 rise}. The initial slope, whether determined by a straight line fit (not shown) or computed from t_{1/2 rise} and half the SP amplitude, exhibited a clear dose dependence in correlation with the SP amplitude (Figs 7B, 9B). In contrast to studies with pheromone sensilla of A. polyphemus and Antheraea pernyi (Zack, 1979; Kaissling et al., 1987; Kodadová, 1993; Kodadová and Kaissling,1996; Pophof,1998) and benzoic acid-sensitive sensilla of Bombyx mori(Kodadová, 1993), in our experiments with short pheromone stimuli, t_{1/2 rise}showed virtually no dose dependence and was not influenced by adaptation(Fig. 7C). The major difference in the regimens of stimulation between our study and the studies mentioned above is the stimulus duration (50 ms versus 2 s or 5 s). Thus, we suspect that the longer stimulus duration, which resulted in a larger total pheromone amount applied, produced desensitization and possibly also long-term adaptation, which might be reflected by a change in the kinetics of the rising sensillar potential phase. This assumption is supported by our recordings in M. sexta at stimulus durations of 1000 ms (Figs 8, 9). These recordings showed a dose dependence and prolongation of t_{1/2 rise} after adapting pheromone stimuli (Fig. 9C), as was also found in A. polyphemus(Zack, 1979; Kaissling et al., 1987). This might indicate that a stimulus-duration-dependent mechanism of adaptation exists that occurs during desensitization and possibly also during long-term adaptation, which affects the amplitude and the kinetics of the rising phase of the sensillar potential in different ways.

Based on the assumption that the sensillar potential predominantly reflects the receptor potential of the adequately stimulated ORN, the waveform of the sensillar potential is probably governed by the complex processes of the chemo-electrical transduction cascade (reviewed by Stengl et al., 1998). From patch-clamp studies on cultured ORNs of M. sexta we know that after application of a very low dose of bombykal, inositol (1,4,5)-trisphosphate(IP₃)-dependent Ca²⁺ channels open, causing a very transient rise in the intracellular Ca²⁺ concentration(Stengl, 1994). This rapid increase of intracellular Ca²⁺ then opens Ca²⁺-dependent cation channels, and possibly also Ca²⁺-dependent Cl^– and K⁺ channels, which, as counteracting currents, may determine the kinetics of the rising phase and the amplitude of the sensillar potential (Stengl, 1993, 1994; Stengl et al., 1992, 1998). Because two of these channel types (which share properties with trp- and trpl-like channels) are permeable to Ca²⁺ and because both are closed in a Ca²⁺-dependent manner, it is very likely that intracellular Ca²⁺ rises caused by these channels are involved in short-term adaptation, as has been shown in Drosophila(Störtkuhl et al., 1999; Montell, 2001). Biochemical and physiological evidence suggests the involvement of cGMP(Ziegelberger et al., 1990; Boekhoff et al., 1993; Stengl et al., 2001; Dolzer, 2002) together with Ca²⁺ concentration rises (Stengl, 1993, 1994; Dolzer et al., 1999) in long-term adaptation in moths. Thus, it is likely that Ca²⁺- and cGMP-dependent mechanisms affect the rising phase of the sensillar potential in response to long and very strong pheromone stimuli.

The faster decline of the sensillar potentials in the short-term-adapted state in response to short stimuli, as well as to long stimuli (Figs 6B, 7D, 9D), suggests the presence of an additional adaptation mechanism, acting via stabilization of the resting potential. Thus, we would expect that either K⁺ efflux or Cl^– influx is increased via short-term adaptation,possibly together with a faster closure of depolarizing ion channels. Since preliminary measurements with Ca²⁺-sensitive dyes (M. Stengl and B. Lindemann, unpublished results) suggest that short-term adaptation is caused by prolonged rises in the intracellular Ca²⁺ concentration while long-lasting adaptation is also due to cGMP-dependent mechanisms, it is likely that opening of Ca²⁺-dependent K⁺ channels(Zufall et al., 1991), and possibly also Ca²⁺-dependent Cl^– channels(Dolzer, 2002), is responsible for the faster decline of the sensillar potential. We are currently testing whether large Cl^– channels or non-specific cation channels,which are probably Ca²⁺- and/or cGMP-dependent(Dolzer and Stengl, 1998; Dolzer, 2002), might be responsible for this stabilization of the resting potential.

In accordance with findings in A. polyphemus(Kaissling et al., 1987), in M. sexta the dose–response relationship of the action potential response to short and long bombykal stimuli after an adapting stimulus underwent a larger shift to higher stimulus intensities than did the relationships of the SP amplitude and the initial slope (Figs 7, 9). This suggests the presence of at least one additional mechanism of adaptation, acting on the transformation of receptor potentials into action potentials. This transformation process probably takes place in the soma or axon hillock region, morphologically and electrically remote from the origin of the receptor potentials (de Kramer et al.,1984; Kodadová,1993), which favours the assumption of a functionally distinct adaptation mechanism. We will test whether cGMP-(Zufall et al., 1991; Zufall and Hatt, 1991; Stengl et al., 1992, 2001; Dolzer, 2002) or a Ca²⁺-dependent phosphorylation of ion channels(Zufall and Hatt, 1991; Stengl, 1993) are involved in the adaptation of the action potential response.

In A. polyphemus, a change in the slope of the dose–response curves of the action potential frequency, as well as the SP amplitude and t_{1/2 rise}, was found after adapting stimulation(Kaissling et al., 1987). Responses to weak stimuli were further reduced compared with responses to strong stimuli, leading to a steeper slope of the dose–response relationship in the adapted state. In M. sexta, however, the dose–response curves were only shifted to higher stimulus intensities but were not obviously altered in their slope. Our study aimed to investigate stimuli in a physiological range. Thus, we did not routinely exceed a stimulus dose of 10 μg and did not statistically analyze whether there is a change in the slope. In the study by Kaissling et al.(1987), applied pheromone doses were corrected for a nonlinearity in the ratio of the pheromone dose loaded onto the filter paper and its release, as found by studies on radioactively labeled pheromone(Kaissling, 1995). Because no radioactively labeled bombykal is available to date, we do not know whether the same nonlinearity exists for bombykal. Nevertheless, we assume that the differences in the results are rather due to differences in the stimulus protocols used to adapt the sensilla. Kaissling et al.(1987) used the strongest stimulus of a protocol that resembles our dose ramp as the only adapting stimulus for a sequence of test stimuli that were applied in increasing order successively. Thus, the time interval between the adapting stimulus and the test stimulus was longer for the test stimuli of higher doses, giving the sensillum more time to recover. In our current study, however, the intervals between conditioning and test stimuli were kept constant(Fig. 2).

The phasic–tonic action potential response observed with long stimulations (Fig. 8) indicates the presence of desensitization (Zufall and Leinders-Zufall, 2000). Desensitization quickly stops the response to a stimulus and thus allows for resolving temporally patterned odour stimuli. The phasic–tonic pattern of the action potential response cannot be completely due to a nonlinearity in the stimulus concentration,because the plateau of the sensillar potential lasts longer than the phasic part of the action potential response(Fig. 10; Kaissling and Priesner, 1970). Since rises in the cGMP concentration after adapting pheromone doses are very slow (Boekhoff et al., 1993),it is likely that mostly Ca²⁺- and/or phosphorylation-dependent mechanisms and, to a lesser extent, also cGMP-dependent mechanisms are involved in desensitization. Because after long pheromone stimuli intracellular Ca²⁺ rises closed IP₃-dependent Ca²⁺ channels and Ca²⁺-dependent cation channels but opened protein kinase C-dependent cation channels (Stengl 1993, 1994), it is likely that these Ca²⁺-dependent mechanisms also underlie desensitization. We assume that after long, strong pheromone stimuli the population of the more slowly activating protein kinase C-dependent cation channels, which are less Ca²⁺-permeable and which are not blocked by Ca²⁺,dominates the late phase of the sensillar potential and underlies the tonic depolarization of ORNs. Future studies will examine which ion channels and second messenger cascades are involved in the different states of adaptation to challenge our hypotheses.

The authors would like to thank Thomas Hörbrand, Holger Schmidt,Markus Hammer and Marion Zobel for insect rearing, Thomas Christensen for the generous gift of bombykal, and Blanka Pophof, Karl-Ernst Kaissling, Kai Hansen and Günther Stöckl for help with technical problems and for valuable discussions. Christian Flecke performed the recording shown in Fig. 4. Thanks also to the referees, whose comments helped to considerably improve the manuscript. This work was supported by DFG grants STE 531/5-1 and 10-1,2 to M.S.