Peripheral modulation of pheromone response by inhibitory host compound in a beetle

Most essential insect behaviors, such as mate finding and host location, are guided by odors. In nature, insects rarely encounter odors as single compounds. Decisions regarding whether to progress towards an odor source, or to abort, are probably based upon a balance mechanism in which attractive and anti-attractive/repellent inputs in the combined ‘odor bouquet’ are weighed against each other. In the insect olfactory system, specific olfactory receptor neurons (ORNs) constitute separate input channels, each detecting compounds from different chemical and biological categories (Andersson et al., 2009; Bengtsson et al., 2009; Larsson et al., 2001; Mustaparta, 1975; Wibe and Mustaparta, 1996). ORN axons project to the glomeruli of the primary olfactory center, the antennal lobe, where integration of odor input takes place through lateral interactions between glomeruli (Olsen and Wilson, 2008; Shang et al., 2007; Silbering et al., 2008).

However, the organization of the insect peripheral olfactory system allows for integration also at the level of ORNs or sensilla. For instance, ORNs are often excited by some compounds and inhibited by others (e.g. Andersson et al., 2009; Hallem and Carlson, 2006; Said et al., 2003), and responses to compound blends cannot always be predicted based on the responses to the individual blend constituents (Getz and Akers, 1997; Ochieng et al., 2002; Party et al., 2009). Furthermore, insects typically have two or more ORNs co-compartmentalized within a sensillum. The significance of such colocalization is poorly understood, but the pairing rules seem very strict in that specific ORNs are located in the same functional sensillum type (de Bruyne et al., 2001; Ghaninia et al., 2007).

Selective ORN pairing might be an adaptation to refine the perception of odor mixtures; for instance, both the spatiotemporal resolution of volatile stimuli (Fadamiro et al., 1999) and compound-ratio detection would be improved if the olfactory sensors are located at the same point in space. Thus, neurons responding to compounds that together constitute an ecologically important signal should then often be found paired within the same sensillum. Good examples are the pheromone ORNs that are colocalized with ORNs that respond to pheromone antagonists (Cossé et al., 1998; Fadamiro et al., 1999; Larsson et al., 2002; Wojtasek et al., 1998). ORN co-compartmentalization might also provide the means for signal modulation in the periphery, in that responses in neighboring neurons could potentially affect the activity of each other (Getz and Akers, 1994). In fact, a theoretical model predicts the existence of passive electrical interactions between colocalized ORNs (Vermeulen and Rospars, 2004), but the effects of these interactions have not yet been systematically characterized in insect sensilla.

Conifer-feeding bark beetles (Coleoptera: Curculionidae: Scolytinae) constitute excellent models to study peripheral integration of odor mixtures. They are among the most well-studied insects in terms of behavioral responses to odor blends, such as different combinations of ecologically relevant attractants and inhibitors, and much is known about the peripheral detection of these individual compounds (Andersson et al., 2009; Tømmerås, 1985). Mass-attacks on Norway spruce [Picea abies (L.) Karst.] by the European spruce bark beetle (Ips typographus, L.) are induced by release of the male-produced aggregation pheromone, a mixture of (4S)-cis-verbenol (cV) and 2-methyl-3-buten-2-ol (Schlyter et al., 1987). Pheromone attraction is modulated by anti-attractant non-host volatiles (NHVs) from leaves and bark of angiosperm plants (Zhang and Schlyter, 2004). Mechanisms to prevent overcrowding of host trees are thought to involve other semiochemicals, such as verbenone, that appear in later attack phases. Verbenone synergizes the inhibitory effect of NHVs in I. typographus (Zhang and Schlyter, 2003) and is used as a negative cue also by several other bark beetle species (Lindgren and Miller, 2002; Schlyter and Birgersson, 1999). Very little is known about the behavioral relevance of host monoterpenes in I. typographus (Erbilgin et al., 2007) or the host phenolics (Faccoli and Schlyter, 2007) and other less volatile host compounds, such as sesquiterpenes.

It is known from single-sensillum recordings (SSRs) that several ORN classes in I. typographus are tuned to host monoterpenes, including two classes that are highly selective for para-cymene and 1,8-cineole (Ci), respectively (Andersson et al., 2009). All Ci neurons (small-amplitude B cell) were co-compartmentalized with neurons responding to the pheromone component cV (large-amplitude A cell). However, in other sensilla, the cV cell was paired with a different B cell that did not respond to the test odorants. Observations suggested that stimulating with Ci sometimes inhibited the cV cell, and this inhibition appeared mainly in sensilla in which the cV and Ci cells were co-compartmentalized, but the phenomenon was not thoroughly surveyed (Andersson et al., 2009). The organization of cV and Ci neurons in the two functional types of sensilla in I. typographus is excellent for investigations of potential interactions between these ORNs, the colocalization of which we now hypothesize has a functional and ecological relevance. To test this hypothesis, we characterized, by a combination of chemistry, behavior and physiology, a semiochemical system involving both a behaviorally attractive pheromone component and inhibitory plant volatiles.

In this study, we identified, by combined gas chromatographic and electroantennographic detection (GC–EAD) (Zhang et al., 2000), two (among others) EAD-active compounds, Ci and p-cymene, that were abundant in spruce trees heavily attacked by I. typographus, and demonstrated that Ci significantly inhibits pheromone attraction in the field. We also detailed the inhibitory effects of Ci on the cV pheromone neuron, comparing the two types of sensilla by SSR using binary odor mixtures. The plant volatile Ci inhibited the cV neuron primarily when it was paired with a Ci neuron, also when the cV cell was simultaneously activated by its ligand. Thus, coincident responses in colocalized neurons might significantly influence odor perception, suggesting that the two neurons do not act as independent units. Such colocalization effects might sharpen adaptive responses to mixtures of semiochemicals with different ecological origin and significance.

Headspace volatiles from the trunk of a Norway spruce (Picea abies) tree heavily attacked by I. typographus for at least two weeks were sampled in situ within a high-density polyacetate film (Look, Terinex, England) enclosure by a battery-operated pump with an activated charcoal filter tube in the air inlet (Zhang et al., 1999; Zhang et al., 2000). The plastic film around the trunk at a level of 1.1–1.6 m was sealed by a portable heat sealer on the side and tightened to the trunk with tape at both open ends. The distance between the film and bark surface was ca. 1 cm. Volatiles (sampling bark area ca. 0.45 m²) were trapped on Porapak Q (50/80 mesh; 30 mg in Teflon tube: 3 mm×35 mm) for 1.5 h (airflow 300 ml min^–1) and extracted with 300μl diethyl ether (Fluka >99%).

Headspace volatiles from a non-attacked spruce log (25 cm diameter and 30 cm long; freshly cut from a nearby healthy tree) were collected using the same aeration procedure. The film was replaced by a polyacetate cooking bag (35×43 cm, Terinex, sampling area ca. 0.24 m²). All aeration extracts were kept at –20°C before GC–EAD and GC–MS analyses.

A volume of 3μl of aeration samples was injected splitless into an HP 6890 GC (Agilent, Palo Alto, CA, USA) containing a fused silica column (HP-Innowax) with a 1:1 effluent splitter, allowing simultaneous flame ionization detection (FID) and electroantennographic detection (EAD). Hydrogen was used as a carrier gas. The column temperature was 40°C for the initial 2 min, rising to 200°C at 10°C min^–1, and held for 2 min. The outlet for the EAD was inserted into a humidified air-stream (0.5 m s^–1) that passed over an I. typographus antennal preparation. A glass capillary indifferent electrode filled with Beadle–Ephrussi Ringer, and grounded by means of a silver wire, was inserted into the severed head of a beetle. A similar recording electrode, connected to a high-impedance DC amplifier with automatic baseline drift compensation, was placed in contact with the distal end of the antennal club. The antennal signal was stored and analyzed on a PC equipped with an IDAC-card and the program EAD v. 2.3 (Syntech, Kirchzarten, Germany). A repeatable response was defined as a depolarization of the antennal signal at the same retention time in three out of five runs. The headspace samples were also analyzed by GC–MS on an HP 6890 GC with an HP 5973 mass-selective detector (Agilent) using the same type of GC column and conditions as described above. Antennally active volatiles were identified by comparison of retention times and mass spectra of standards.

The behavioral effect of (±)-1,8-cineole (1,3,3-trimethyl-2-oxabicyclo[2,2,2]octane; Ci; >99%, Aldrich) and p-cymene (1-methyl-4-[1-methylethyl]benzene; >99%, Acros) on I. typographus pheromone attraction was investigated using Lindgren multiple-funnel traps (12 funnel size, Pherotech International, Delta, BC, Canada). The minimum distance between traps was 50 m and 15 m between traps and the nearest tree. Insects were collected when at least 30 beetles were caught in a pheromone-baited trap with the lowest catch. Following collection, treatment positions were changed according to a Latin square experimental design (Byers, 1991). The experiment was conducted from June to August 2004 in clear-cuts situated within spruce stands (80–90 years old) in the Polana Mountains in central Slovakia. Traps were placed on SW slopes at altitudes of ca. 750 m above sea level. The standard commercial pheromone dispenser IT-Ecolure was used as an attractant (Fytofarm, Bratislava, Slovakia). IT-Ecolure is a wick-aluminium-foil-protected dispenser (Varkonda, 1996) filled with 3 ml of pheromone mixture. The average release rate is ca. 50 mg/day under field conditions (Fytofarm). The monoterpenes 1,8-cineole and p-cymene were filled in membrane dispensers (Wilhelm Biological Plant Protection, Sachsenheim, Germany) that consisted of a 4 ml glass vial with a twisted cap that was punched (diameter 12 mm) and contained four laminated permeable membranes (0.2 mm thick, diameter 17 mm). Baited vials were positioned with the cap facing downwards, resulting in a uniform evaporation. The estimated average release rate was ca. 50 mg/day under field conditions, as measured by weight loss over time.

Single-sensillum recordings with tungsten microelectrodes were performed using standard equipment (Syntech) and well-established experimental protocols. Details regarding insect rearing, insect preparation, stimuli preparation and the odor delivery system have been described previously (Andersson et al., 2009). Chemicals, diluted in paraffin (product # 1.07162.1000, Merck, Darmstadt, Germany), were applied in aliquots of 10μl on filter papers inside Pasteur pipettes. Two doses (1 and 10 ng) of (4S)-cis-verbenol (cis-4,6,6-trimethylbicyclo[3.1.1]hept-3-en-2-ol; cV; 95%, Borregaard), and two doses (1 and 10μg) of (±)-1,8-cineole (>99%, Aldrich) were tested singly and as the four possible binary combinations. In the latter case, the two compounds were applied on separate filter papers inside the same pipette to prevent any potential compound interactions that could have affected evaporation rates. The cV and Ci cells are indifferent to p-cymene (>99%, Acros), but p-cymene (10μg) was tested in binary combinations with both doses of cV (1 and 10 ng), as control stimuli. In pipettes containing single compounds, a second filter paper loaded with paraffin solvent was also present (i.e. all pipettes contained two filter papers). Stimuli were delivered in random order and prepared before each experimental session. Sensilla with the cV cell paired with a non-responsive B cell (sensillum type I) and sensilla containing the colocalized cV and Ci cells (type II) were both tested for responses. Attempts were made to record from both sensillum types on all individuals, and we succeeded in the majority of cases. Both males and females were used in recordings. Sex separation was based on external morphology (Schlyter and Cederholm, 1981) and confirmed by dissection of genitalia.

For the field test, absolute catch was transformed according to log(catch +1) and used as the variable for statistical processing. Treatments were compared with one-way ANOVA, and, to compensate for unequal variances, Dunnett's T3 was used as a post-hoc test at α=0.05. Sex separation was based on dissection of genitalia and the sex ratio analyzed with 95% binomial confidence intervals (95% CI) (Newcombe, 1998). Three hundred beetles, or all individuals (if <300), per treatment were dissected.

From the SSRs, response curves of high temporal resolution were obtained by counting (in Autospike 3.0, Syntech) the number of spikes (action potentials) in the cV A cell in 200-ms bins, starting 1 s before stimulus onset and ending 2 s after onset. Clear excitatory odor responses were recorded during a time-period of ca. 800 ms. For statistical comparisons of both excitatory and inhibitory responses, the total number of spikes during this period was used, after subtracting the number of spikes fired during stimulation with control. Factorial ANOVA was used to compare the response of the cV cell in the two sensillum types. Regression analysis on log-transformed doses, log(dose +1), was used for comparisons within a sensillum type. t-tests were used in pair-wise comparisons. Cohen's d was used as a measure of standardized effect size. In addition, a d value was calculated from the adjusted R² obtained from regression analyses (Nakagawa and Cuthill, 2007). According to Cohen (Cohen, 1988), a d value above 0.8 is regarded as a large effect. Tests were performed in SPSS 11.0 at α=0.05.

GC–EAD analysis of the aeration samples from both a non-attacked log and an attacked trunk of a spruce tree showed repeatable antennal responses (in at least three out of five runs) to several compounds. In the non-attacked sample, responses were elicited by myrcene, 1,8-cineole/β-phellandrene, α-longipinene, 1-terpineol and 4-allyanisole (Fig. 1A; later-eluting compounds not shown). Responses to the two major monoterpenes, α-pinene and β-pinene, were not repeatable. In the aeration sample from the heavily attacked spruce tree, EAD responses were again recorded to compounds present in the fresh sample (including Ci), but several additional compounds also elicited responses, including p-cymene, isolongifolene, the pheromone component cis-verbenol and several unidentified compounds (Fig. 1B; later-eluting compounds not shown). 1,8-cineole was ca. 20–30 times more abundant than cis-verbenol in the head-space from the attacked tree. A minute amount of the other essential pheromone component, 2-methyl-3-buten-2-ol, was detected in the monoterpene eluting range, but it did not elicit repeatable EAD responses. Again, the antennal responses to the major monoterpenes varied among the antennal preparations.

A total of 32,363 I. typographus were caught in six replicates with a highly significant overall variation among treatments (F_5,35=72.7, P<0.001). The addition of 1,8-cineole to the pheromone strongly reduced trap catch compared with pheromone alone (Fig. 2; 88% reduction, with a very high effect size; Cohen's d=–2.9 based on log-transformed means). The reduction in trap catch by the presence of p-cymene together with pheromone was weaker and not significantly different from pheromone alone (50% reduction, d=–0.80). The proportion of males in the catch ranged from 19 to 33% but did not differ significantly between treatments (data not shown). Catches in traps baited with 1,8-cineole or p-cymene alone did not attract beetles, with close to zero catches like the blank control (Fig. 2).

Sensilla classified as cV sensilla are found in two types. An A neuron responding selectively to cis-verbenol is accompanied either by a non-responding B neuron (sensillum type I) or with a B neuron responding to 1,8-cineole (type II) (Fig. 3, schematic drawing). Dose–response curves for the cV and the Ci cells in type II sensilla have been published previously (Andersson et al., 2009). In type I sensilla, we now show that there was no general inhibition of the cV cell by either dose of Ci. By contrast, the cV cell in type II sensilla was inhibited by both doses of Ci (Fig. 4A). The difference between sensillum types is obvious from the recorded spike trains (Fig. 3). The inhibition in type II sensilla lasted for up to 1.4 s and was significantly stronger in type II than in type I sensilla during the 800 ms time-slot (Fig. 4A).

Stimulating with cis-verbenol at a lower dose (1 ng) elicited a weak response in the cV cells (Fig. 3). The response was significantly stronger in type II than in type I sensilla (t₂₁=–2.12, P<0.05; Cohen's d=–0.94; Fig. 4B). Synchronous stimulation with cV and Ci clearly reduced the cV response in a dose-dependent manner. The reduction was found only in cV cells co-compartmentalized with Ci cells, as demonstrated by factorial ANOVA showing a significant interaction between sensillum types and stimuli (F_2,60=3.43, P<0.05; cV/p-cymene stimulus excluded from analysis; Fig. 4B). In some type II sensilla, the cV cell responded to the cV:Ci (1 ng:10μg) stimulus just as if the cis-verbenol was absent (Fig. 3). Thus, the response of the pheromone cell was in such cases shut down to a level lower than or similar to the blank control. In both types of sensilla, combining 1 ng cV with 10μg p-cymene had no effect on the cV response (Fig. 4B). The B cells in both types of sensilla were, on average, not affected by cV stimulation (the few spikes in the B cells in Fig. 3 upon cV stimulation are due to random fluctuations in background activity).

The higher dose (10 ng) of cV elicited a clear response in both sensillum types (Fig. 3). Similar to the low (1 ng) cV dose, the response to the higher dose also seemed stronger in type II sensilla than in type I, as indicated by the relatively large effect size (Cohen's d=–0.75), but the difference was not significant (t₂₁=–1.81, P>0.05; Fig. 4C). At the even higher screening dose (10μg) of cV, the response did not differ between the sensillum types (average response type I: 157 Hz, N=12; type II: 161 Hz, N=7; P>>0.05).

At the 10 ng cV dose, the binary mixtures of cV and Ci reduced the cV response in a dose-dependent manner. In contrast to the low cV dose, factorial ANOVA showed no interaction between sensillum types and stimuli (F_2,61=0.35, P>0.05; cV/p-cymene stimulus excluded from analysis) and regression analyses demonstrated significant response reductions in both types of sensilla. However, the significance levels from the regression analyses, as well as the effect size measures, differed between sensillum types, indicating a stronger effect in type II sensilla (Fig. 4C). The addition of 10μg p-cymene did not significantly alter the response to 10 ng cV in any type of sensillum, although a tendency was found in sensilla without Ci neurons.

We show that the antennae of I. typographus respond to volatiles from healthy host tree logs and to other volatiles released primarily from heavily attacked trees. More EAD-active compounds were found in the spruce that had been under attack for two weeks, in comparison with a non-attacked log, which indicates an important multi-component ‘unsuitable host signal’ that might regulate bark beetle density on attacked trees. As Ci was present in both attacked and non-attacked samples, its inhibitory properties presumably represent a quantitative rather than a qualitative effect; preliminary results indicate that the amount of Ci increases after attack by I. typographus, as newly attacked trees release approximately four times more Ci than non-attacked trees [N=6 (C. Schiebe, P. Brodelius, G. Birgersson, J. Witzell, P. Krokene, J. Gershenzon, A. Hammerbacher, B. S. Hansson and F.S., personal communication)]. The amount of Ci released from the attacked tree was 20–30 times larger than the amount of cV. In our field test, the release ratio between the pheromone and Ci was 1:1, but, as the pheromone contains 2-methyl-3-buten-2-ol and cV in a ca. 50:1 ratio, the cV:Ci ratio in the field test was similar to the ratio released from the attacked tree. In contrast to earlier reports on the kairomonal activity of mainly (–)-α-pinene on I. typographus (Erbilgin et al., 2007; Hulcr et al., 2006; Jakus and Blazenec, 2003), a mixture of EAD-active host monoterpenes, α-pinene, β-pinene and p-cymene, was unattractive in itself, but interrupted the pheromone response of the closely related larch bark beetle I. subelongatus (Zhang et al., 2007). More work is surely needed to determine the behavioral roles of these antennally active volatiles from healthy and attacked hosts.

The abundance of host monoterpene-selective ORNs on the I. typographus antennae (Andersson et al., 2009) suggests that beetles are able to distinguish spruces based on their odor profiles, which vary among trees of different genotypes and physiological states (Keeling and Bohlmann, 2006). Terpenoid compounds are involved in constitutive and induced defenses of both conifers and angiosperms (Keeling and Bohlmann, 2006; Mumm et al., 2008) – insects feeding on such plants would likely benefit from selecting hosts with low terpenoid content. Indeed, several monoterpenes are toxic and lethal to bark beetles (Everaerts et al., 1988; Raffa and Smalley, 1995) and to their symbiotic fungi (Raffa et al., 1985). Although not vectored by bark beetles, inoculation of Norway spruce with pathogenic fungus Heterobasidion parviporum resulted in an increased content of 1,8-cineole (Zamponi et al., 2007). In addition, p-cymene was the most effective of all tested terpenes in reducing growth of coniferous decay fungi (De Groot, 1972). It is therefore likely that both Ci and p-cymene are involved in the induced defense of conifers against bark beetles and/or their symbiotic fungi.

Because the 1,8-cineole and cis-verbenol neurons are not always colocalized within the same sensilla, we have a unique model for studying intra-sensillum interactions between a pheromone compound and a plant-derived behavioral antagonist. There is much evidence suggesting that insect decision making in host choice is a matter of weighing relative ratios of different plant compounds, possibly from different sources. Compartmentalization of neurons in sensilla might reflect this task (Bruce et al., 2005) as it probably favors coincidence detection, which in turn would improve the accuracy of compound ratio discrimination of odor mixtures. In nature, I. typographus frequently encounters cV and Ci in combination as cV is released when a spruce tree is attacked. Along with our behavioral data, the colocalization of the neurons suggests that the binary mixture constitutes an ecologically significant signal, assuming that selective pairing of ORNs is adaptive and not due to chance. Some male moths have a remarkable ability to distinguish odor filaments from different sources with extremely high spatiotemporal resolution (Fadamiro et al., 1999; Witzgall and Priesner, 1991). It was suggested that such amazing feats depend on ORNs being located within the same sensillum (Fadamiro et al., 1999), which is common among pheromone agonist/antagonist neurons in Lepidoptera (Baker et al., 1998; Larsson et al., 2002). Similar arrangements are found also in Coleoptera (Wojtasek et al., 1998), but, so far, not in bark beetle pheromone neurons (Andersson et al., 2009; Mustaparta et al., 1977; Mustaparta et al., 1980). Consistent with behavioral observations, asynchronous arrival of pheromone constituents disrupts the spiking pattern of projection neurons in the antennal lobe, with resulting effects on the temporal aspects of odor coding (Christensen and Hildebrand, 1997).

The system of cis-verbenol and 1,8-cineole in I. typographus is different from the pheromone agonist/antagonist organization in other insects; the compounds are of animal and plant origins, respectively, and unlikely to be involved in reproductive isolation. It would, rather, be a means to judge the fitness value of an integrated odor stimulus comprised of elements from conspecifics and potential hosts. It would nevertheless be of interest to investigate the effects of spatial separation of pheromone and 1,8-cineole and compare the effect of spacing with that of another anti-attractant compound (e.g. verbenone), detected by an ORN not colocalized with the cV cell. The colocalization of one pheromone-responding ORN with one that responds to a plant odor is also different from that shown by most other insects. Typically, pheromone ORNs are located in specific morphological sensillum types, whereas ORNs responding to plant odors are found in other sensilla (Hansson et al., 1986; Hansson et al., 1999; Larsson et al., 2001) (but see Said et al., 2003). The pairing of host odor- and pheromone-responsive ORNs in I. typographus suggests that host localization, involving both the pheromone and host-derived compounds, is an integrated system that does not segregate between different classes of semiochemicals.

Signal modulation between receptor neurons at the peripheral level is another factor that could influence grouping of neurons in sensilla. Indeed, our results indicate response interactions between neighboring ORNs as the inhibition of the cV cell was much more pronounced when a colocalized Ci cell responded. Neurons within contact chemosensilla of the grasshopper Schistocerca americana have also been shown to interact, in that a single spike in the small-spiking cell resulted in an extended spike interval in the colocalized large-spiking cell (White et al., 1990). Contacts between sensory cells in thermo-/hygroreceptive sensilla styloconica have been found in, for example, the silkmoth Bombyx mori (Steinbrecht, 1989) and were hypothesized to be involved in the antagonistic response characteristics of these cells. However, contacts such as electrical or chemical synapses have not been found between insect ORNs.

A non-synaptic mechanism that could explain the inhibition is the passive electrical interaction that occurs between ORNs that are colocalized in sensilla (Vermeulen and Rospars, 2004). According to the model, excitatory responses in an ORN can hyperpolarize a neighboring cell that is insensitive to the stimulus and in close contact. The model also suggests that pairing of ORNs with similar odor tuning might reduce the amplitude of the receptor potentials and thus lower the sensitivity of the cells (Vermeulen and Rospars, 2004). If signal modulation unavoidably occurs between receptor neurons at the peripheral level, pairing of ORNs with differentiated response profiles would be expected as it would circumvent an overall reduction in olfactory sensitivity. To our knowledge, ORNs with largely overlapping response spectra have rarely been found within the same sensillum (but see Hansen, 1984), although it theoretically could improve discrimination between chemically similar compounds.

In our case, the strong excitatory response in the Ci cell might change the receptor potential of the cV cell so that a higher cV dose is required to produce the same number of action potentials. If so, the inhibitory interaction would be expected to be reciprocal – that is, a strong cV response should inhibit the Ci cell. However, we were unable to investigate this owing to the fact that the small-amplitude spikes of the Ci cell were completely obscured by a strong response of the large-spiking cV cell. In a previous study, we found that, when the methyl-butenol B cell responded, the colocalized A cell was strongly inhibited (Andersson et al., 2009). Unfortunately, no excitatory responses from this A cell were recorded to any odorant tested. In addition, it would be of great interest to identify the ligand(s) for the non-responsive B cell that is paired with the cV cell in type I sensilla, perhaps by means of GC-coupled SSRs (Stensmyr et al., 2001), to investigate whether responses in this cell also inhibit the cV cell. Functional characterization of this ORN in conjunction with behavioral testing of the ligand would be desirable in order to consolidate the relevance of ORN interactions. In fact, inhibition of A cells when B cells respond seems to be a common phenomenon (Blight et al., 1995; Hallem et al., 2004; Larsson et al., 2001), but it has not been systematically addressed.

To our knowledge, our study is the first to investigate interactions between ORNs in sensilla where spikes from the individual neurons could be distinguished. However, the idea of ORN interaction is not novel. In honeybee sensilla placodea, neurons within a sensillum responded to odors in a coordinated manner, suggesting that responses of individual ORNs are not independent (Getz and Akers, 1994). Unfortunately, as these sensilla contain 18–35 ORNs, spikes from individual neurons were not distinguishable but could only be grouped into three or four ‘response units’, which makes the interpretation of the results less straightforward compared with our observations. Other studies have also shown that receptor neuron responses to binary mixtures cannot always be predicted based on the responses to individual constituents, and that the mixture interaction can be both inhibitory and synergistic (Getz and Akers, 1995; Getz and Akers, 1997; Ochieng et al., 2002; Party et al., 2009).

In our case, it was clear that the inhibition of cV neurons in response to binary cV–Ci mixtures occurred exclusively in type II sensilla at the low (1 ng) cV dose. The inhibition in both sensillum types at the higher (10 ng) cV dose suggests that two different mechanisms could be operating, one being the interaction between neurons and the other possibly an effect of a second compound present at a high concentration. An inhibitory interaction would be expected if the rate of molecules entering the cuticular pores is limited or if there is competition for, for example, odorant-binding proteins (OBPs) within the sensillum lymph. Possibly, the different degree of inhibition by Ci could also be explained by inherent differences between the A neurons or between the two types of sensilla. Interestingly, the cV neurons in type II sensilla also demonstrated a greater sensitivity to cV than the ones in type I sensilla. The response to 1 ng cV was almost twice as strong in sensilla containing both the cV and Ci neuron, and a similar, albeit not significant, tendency was found at the 10 ng dose. We do not know why this difference exists, but it might be explained for instance by differences in olfactory receptor gene expression rates or differences in the sensillum environment, such as OBPs or other factors. Alternatively, cis-verbenol-sensitive neurons of type I and type II sensilla might express odorant receptors (ORs) with slightly different response profiles. Molecular characterization of I. typographus ORs is required to distinguish between these hypotheses.

It is uncertain whether the recorded physiological inhibition of Ci is of ecological relevance as the Ci needs to be present at a ca. 1000 times higher dose than the cV in order to elicit detectable inhibition. Our head-space collection from the attacked tree indicated that the ratio between cV and Ci is ca. 1:20–1:30, but our sample size is too small to draw any definite conclusions. In addition, the excitatory inputs from the cV and Ci cells provide the means for central integration, which probably explains some (or most) of the anti-attractant effect of Ci.

In conclusion, the host monoterpene 1,8-cineole inhibits the attraction of I. typographus to its aggregation pheromone, and highly specific ORNs for the compound are present on the antennae. Insects typically group pheromone ORNs in sensilla that are distinct from the ones that contain ORNs for plant compounds, which makes our system deviant from this general rule. In particular, the inhibitory interaction that occurs between the pheromone and host odor ORNs is the first one described for any insect. Although the ecological consequences of the peripheral inhibition in isolation could not be established, it is clear that an inhibitory interaction occurs in the periphery and that the cV cells differ in sensitivity depending on which type of sensilla they are housed in. Our observations thus question the traditional view that ORNs within a sensillum act as independent response units.

We thank Muhammad Binyameen and Elisabeth Marling for assistance in bark beetle rearing. We also thank Sylvia Anton and Michel Renou (INRA, Versailles) for useful input on the manuscript.

 
• Ci

  (±)-1,8-cineole

 
• cV

  (4S)-cis-verbenol

 
• GC–EAD

  gas chromatographic–electroantennographic detection

 
• GC–MS

  gas chromatography–mass spectrometry

 
• NHV

  non-host volatile

 
• OBP

  odorant-binding protein

 
• ORN

  olfactory receptor neuron

 
• SSR

  single-sensillum recordings