Comparisons of Contact Chemoreception and Food Acceptance by Larvae of Polyphagous Helicoverpa armigera and Oligophagous Bombyx mori

We compared food choice and the initial response to deterrent treated diet between fifth instars of Helicoverpa armigera, a polyphagous generalist pest, and Bombyx mori, an oligophagous specialist beneficial. Bombyx mori was more behaviorally sensitive to salicin than to caffeine. The relative sensitivities were reversed for H. armigera, which was tolerant to the highest levels of salicin found in natural sources but sensitive to caffeine. A single gustatory receptor neuron (GRN) in the medial styloconic sensillum of B. mori was highly sensitive to salicin and caffeine. The styloconic sensilla of H. armigera did not respond consistently to either of the bitter compounds. Phagostimulants also were tested. Myo-inositol and sucrose were detected specifically by two GRNs located in B. mori lateral styloconic sensillum, whereas, in H. armigera, sucrose was sensed by a GRN in the lateral sensillum, and myo-inositol by a GRN in the medial sensillum. Myo-inositol responsiveness in both species occurred at or below 10−3 mM, which is far below the naturally occurring concentration of 1 mM in plants. Larval responses to specific plant secondary compounds appear to have complex determinants that may include host range, metabolic capacity, and gustatory repertoire.


Introduction
Phytophagous insects may be categorised as polyphagous, oligophagous, or monophagous, depending on the range of host plants they will accept. In Lepidoptera, a sophisticated olfactory system helps females navigate towards a host by using volatile chemical cues. The larval gustatory system is adapted to mediate acceptance or rejection of plant tissues following drumming and test bites (Chapman 1982;Devitt and Smith 1985). The mechanism of food choice and preference in caterpillars is of great interest, since many economically important agricultural pests are members of the order Lepidoptera, and most lepidopteran larvae have specialised feeding habits. For example, larvae of the domesticated silkworm, Bombyx mori, feed mainly on mulberry leaves and, only very rarely, on a few other species of Moraceae, Compositae, or Ulmaceae (Legay 1958;Tanaka 1943). In contrast, Helicoverpa armigera, a widely distributed phytophagous lepidopteran, is a typical polyphagous insect, which feeds on at least 200 species from 30 plant families including members of the Solanaceae, Gramineae and Leguminosae (Chin et al. 1962;Xu et al. 1958).
Caterpillars with a restricted host plant range are behaviorally more sensitive to deterrents, bitter plant secondary metabolites, than polyphagous caterpillars Chapman 1987, 1994;Jermy 1964), although this is not necessarily true in other orders (Eichenseer and Mullin 1997). Most behavioral comparisons are based on phylogenetically closely related species with different host ranges Chapman 1987, 2000;. A model has been developed that seeks to explain these differences in terms of central nervous system processing rather than differences in sensory input (Schoonhoven and Blom 1988), and there is experimental support for such a model .
In caterpillars, food evaluation is performed by gustatory organs located on the mouthparts. These are paired sensilla styloconica on the maxillary galea, basiconic sensilla on the top of the maxillary palps, as well as sensilla on the epipharynx (Dethier 1937;Ishikawa and Hirao 1961;Ma 1972;Schoonhoven 1969;Waldbauer and Fraenkel 1961). The two sensilla styloconica seem to play a particularly important role in acceptance of host plants (Schoonhoven 1987). Each styloconic sensillum houses four gustatory receptor neurons (GRNs) with specific response spectra for plant compounds (summarized by Schoonhoven and Van Loon 2002;Shields 2009). Typically, some of the neurons respond to phagostimulants, primary plant metabolites like sugars and amino acids that evoke feeding. Other GRNs are activated by deterrents, secondary plant metabolites commonly bitter to humans that mediate food aversion. Secondary compounds are highly diverse and often characteristic of one or a few plant families. Feeding is not determined by the simple presence or absence of specific compounds, but rather by a balance between deterrent and stimulatory compounds (Chapman 2003).
In the maxillary gustatory system of the silkworm, one GRN in the lateral styloconic sensillum is sensitive to sucrose, and the other three respond to myo-inositol, glucose, or salts, respectively (Ishikawa 1963(Ishikawa , 1966Ishikawa and Hirao 1963). However, one GRN in the medial sensillum is sensitive to bitter compounds, and the others respond to water, salts, acids, ecdysone, or 20-hydroxyecdysone (Descoins and Marion-Poll 1999;Tanaka et al. 1994). A few gustatory sensilla also are found in the maxillary palps, which are involved in food detection and selection (Ishikawa et al. 1969). Synergistic as well as suppressive interactions have been observed at the peripheral level, central nervous system, and behavioral levels (Hsiao and Fraenkel 1968;Ishikawa 1966;Ishikawa et al. 1969;Thorsterinson 1960). Compared with B. mori, H. armigera has two GRNs in each styloconic sensillum that respond to sucrose and alanine separately. The other two GRNs detect deterrents from plant saps (Simmonds and Blaney 1991). A more recent study has reported that one neuron in the medial styloconic sensillum responds to myoinositol rather than sucrose (Tang et al. 2000).
Here, by using identical stimuli presented under the same conditions, we compared larval feeding behavior and the sensory physiology of the sensilla styloconica of fifth instar H. armigera with those of B. mori. As is usual in this field, we restricted our study to fifth instars for experimental convenience. We found that B. mori is approximately two orders of magnitude more behaviorally sensitive than H. armigera to salicin. Surprisingly, however, B. mori was much less sensitive than H. armigera to caffeine. In both cases, the more sensitive species responded in the naturally occurring range for the bitter compounds tested. The styloconic sensilla of B. mori were sensitive to detect both bitter compounds below the behavioral threshold, but detection of these deterrents was not observed in H. armigera. Regarding sugars, we found that larvae of both species could detect the plant sucrose and myo-inositol at or close to physiologically relevant concentrations. Our findings may provide more data for a more species-specific understanding of the physiological and molecular characteristics of food determination in caterpillars.

Methods and Materials
Insects Bombyx mori larvae used in the 24 h feeding inhibition test were reared on mulberry leaves at 25°C in the State Key Laboratory of Silkworm Genome Biology, Southwest University, Chongqing, China (Xia et al. 2004). Larvae used in behavioral studies on initial contact with food and for electrophysiological recording were maintained on mulberry leaves at 25°C in CSIRO Ecosystem Sciences, Canberra, Australia. Helicoverpa armigera larvae were reared on artificial diet, according to Teakle and Jensen (1985), as modified by Mahon et al. (2007), and maintained in CSIRO Ecosystem Sciences. The strain is a combination of lab strains dating back to 1988 that has been regularly outcrossed to "field" material from Australia.
Scanning Electron Microscopy (SEM) Heads were excised from newly molted fifth instars of B. mori and H. armigera and fixed in 2.5 % glutaraldehyde for 3 h. They were dehydrated through a graded alcohol series and air dried at 25°C, before being mounted on a specimen holder with double-sided adhesive tape and sputter-coated with gold. The preparations were examined with a Hitachi H-5700 (Tokyo, Japan) SEM at 15 kV.
Chemicals Salicin and caffeine were chosen as antifeedants. In nature, caffeine concentration varies in the range of 6.8-21 g/kg (35-108 mM) in fresh tea leaves to 8-18 g/kg (41-92 mM) in fresh coffee beans, both naturally rich sources of the compound (Nathanson 1984). Salicin occurs at up to 0.5 g/kg (1.7 mM) in fresh leaves of aspen and similar levels are found in willow (Lindroth et al. 1988;Smiley et al. 1985). These values likely would represent the higher end of the natural range in plants. Sucrose and myo-inositol were chosen as sweet compounds. These chemicals and KCl were purchased from Sigma at ≥99 % purity. All tested compounds were dissolved in 50 mM KCl for electrophysiology, which served as electrolyte in both electrodes, and in ultrapure water for the behavioral tests. The antifeedant and sweet chemicals were selected based on their common use in experiments with lepidopteran insects (Chapman 2003;Schoonhoven and Van Loon 2002). The ranges of concentrations chosen for physiological and behavioral experiments were those found to span the dynamic range of responses, using initial ranging experiments. The concentrations of sucrose and myoinositol tested electrophysiologically were 10 −5 , 10 −3 , 0.1, 1, 50, and 100 mM. Salicin and caffeine were tested at 10 −5 , 10 −3 , 0.1, 1, 5, and 10 mM. For behavioral assays, concentrations of salicin were 0.1, 1, 5, 10, 50, 100, and 500 mM and of caffeine were 0.01, 0.1, 1, 5, 10, 100, and 500 mM.
Behavioral Assay Newly molted fifth instars were starved for 24 h before being tested on artificial diet. Ten B. mori larvae were placed in a dish containing a 2 g disc (~1.5 cm diam.) of artificial diet containing various concentrations of caffeine and salicin (mulberry flour 20 g, sucrose 4.75 g, myo-inositol 1.62 g, agar 1.5 g, X grams of caffeine/salicin, water 100 ml) Ten H. armigera were tested individually to prevent cannibalism as well as on their own diet to avoid maladjustment. The H. armigera artificial diet (Mahon et al. 2007), containing caffeine or salicin, was presented to larvae as for B. mori in 2 cm discs. Controls were offered untreated food discs. After 24 h, discs were removed and weighed. Assays were repeated three times. The 'feeding inhibition index' was calculated as follows: An index close to 100 % indicates a strong deterrent effect, whereas close to 0 % indicates a weak effect.
We also observed the behavior of newly molted fifth instars on their initial encounter with diet containing aversive stimuli (same as above). Fifteen larvae for each treatment starved for 24 h were tested individually. We recorded whether the larvae either tasted the food, which included drumming and biting but not eating, or ate the food. Eating was defined as continuous feeding for at least 10 sec. We also recorded how long larvae spent eating within the first 20 min after encountering food.

Electrophysiological
Recording Tip-recordings were performed on the medial and lateral styloconic sensilla of fifth instar B. mori and H. armigera (aged 2 d after molting), as described by Ishikawa (1963) and van Loon (1990) with some modifications. Larval heads were cut from the thorax, and the reference glass electrode, filled with 50 mM KCl, was inserted into the head until pressure caused the mouthparts to open. A recording glass electrode filled with tastant in 50 mM KCl solution was brought into contact with the tip of a sensillum under a dissecting microscope. Electrophysiological activity was recorded with a Tasteprobe amplifier (Marion-Poll and Van der Pers 1996) linked to a data acquisition controller (IDAC-4: Syntech, The Netherlands) and a computer equipped with the AutoSpike-32 software (Syntech). Responses were recorded from both the medial and lateral sensilla on the same side of the head. Stimuli lasted 2 sec and were separated by an interval of 2 min to allow for recovery and to minimize adaptation. Sensilla were washed between recordings by bringing their tip into contact with a solution of 50 mM KCl, and were repeated twice. The order of stimulation was random except for 50 mM KCl, which was always tested first.
Only recordings from the same side of each insect were used for analysis. Analysis of the spikes was performed by using AutoSpike-32 (Syntech) based on the amplitude. The response intensity (spikes/sec) was determined by automated counting of the number of spikes in the first second. The responses to tested substances were generally quite stable over time, except for sucrose and myo-inositol where the responses decreased noticeably after 1 sec in H. armigera. In addition, as automated counting and classification is not always error free, we revised results by visually counting and distinguishing spikes on the basis of their amplitudes, firing regularity, and shapes (reflecting doublets) (Hiroi et al. 2004). The presence of doublets was used to determine simultaneously occurring spikes, which may appear as a larger spike than usual. Data are presented as means ± standard error of the means from the sensilla of at least five different larvae.
Principal Components Analysis Principal Component Analysis (PCA) was employed to compare the electrophysiological responses of the GRNs in lateral and medial styloconic sensilla to different concentrations of various tastants in B. mori and H. armigera. The correlation matrix of electrophysiological responses of the lateral and medial styloconic sensilla to all concentrations of the tastants (sucrose, myo-inositol, salicin and caffeine) were submitted for PCA analysis by using Unscrambler software (version 9.1, CAMO PRO-CESS AS, Nedre Vollgate, Norway). Correlation matrix was considered so that the responses with the highest variance would not dominate the first principal component. The scores were graphed, and vectors were used to indicate the tastants that characterise differences between the sensilla. (PCA seeks to reduce the number of variables that need to be considered to a smaller number of indices, called principal components. The first principal component accounts for the largest quantum of variance among samples. Subsequent principal components account for successive amounts of the total variance in the data set and are uncorrelated with prior principal components).

Results
Behavioral Sensitivity to Plant Secondary Metabolites Both bitter substances inhibited feeding in both species, with some notable differences in sensitivities ( Fig. 1). (Note that we were only able to obtain a robust dose-response fit for B. mori feeding on salicin. Estimates of the other antifeedant EC 50 values are based on visual inspection of the curve fits). Bombyx mori was highly sensitive to the phenolic glycoside salicin with an EC 50 of 4.9±2.5 mM. Helicoverpa armigera was 40 times more tolerant of salicin than B. mori, requiring concentrations of approximately 200 mM to inhibit 50 % of the feeding. Bombyx mori was moderately behaviorally sensitive (estimated EC 50 ≈100 mM) to the alkaloid caffeine. On the other hand, H. armigera was approximately ten-fold more sensitive to caffeine than B. mori, with an estimated EC 50 of approximately 10 mM.
To exclude the possibility that feeding inhibition by the antifeedants is mediated by post-ingestion toxicity, we characterized the first encounter of larvae with deterrent treated diet (Fig. 2). Approximately 80 % of B. mori larvae ate on initial encounter with diet containing 1 mM salicin and control diet (Fig. 2a). Approximately half the larvae ate diet supplemented with 5 mM salicin, and none ate at 100 mM. Five and 10 mM caffeine resulted in approximately half the larvae eating the food, but the highest concentration tested strongly inhibited eating (Fig. 2b). Over the 20 min of the test, B. mori larvae spent less time eating on diet containing any level of caffeine or high concentrations of salicin than on control diet (Online Resource 1). A video of B. mori tasting and eating behavior is included in Online Resource 2. In contrast, for H. armigera, the presence of salicin did not alter the initial eating behavior (Fig. 2c), although larvae spent less time eating over the 20 min period compared to controls (Online Resource 1). With caffeine, H. armigera had very different initial responses (Fig. 2d). At 5 mM caffeine, approximately half the larvae ate immediately, and the other half only tasted the diet. At 10 mM and 500 mM concentrations of caffeine, H. armigera larvae bit and immediately spat out the food. Less time was spent eating over the 20 min period for all concentrations of caffeine, and effectively nothing was eaten at the higher concentrations (Online Resource 1). A video of H. armigera tasting and eating behavior Fig. 1 Behavioral responses in Bombyx mori and Helicoverpa armigera larvae. Bombyx mori behavioral responses to salicin (a) and to caffeine (b). Helicoverpa armigera behavioral response to salicin (c) and to caffeine (d). Curves indicate the mean feeding inhibition (N=30). Curves were fitted with a variable slope nonlinear fit. Bars represent standard errors of the mean. The orange and green (shaded) areas on the graphs represent the range of concentrations encountered in rich natural plant sources of salicin and caffeine, respectively is included in Online Resource 3. These observations were, thus, highly consistent with the feeding inhibition data, and eliminate the need to invoke post-ingestion toxicity to explain the feeding inhibition observed in the first experiment.
Morphology of the Gustatory Sensilla of B. mori and H. armigera Scanning electron microscopy (Fig. 3) showed that the morphology of H. armigera maxillae is similar to that of B. mori and to the general plan seen in other lepidopteran larvae (Bernays and Chapman 1994;Grimes and Neunzig 1986;Ishikawa 1963). We observed no obvious differences in the locations of either the styloconic sensilla on the maxillary galea or of the basiconic sensilla on the maxillary palp.  Overall Analyses of the Neuronal Responses to Sweet and Bitter Compounds The PCA score plots (Fig. 4) show that the GRNs in the lateral and medial sensilla respond differently to the test compounds in the two species. Compositional loadings plotted as vectors show that GRNs in the lateral styloconic sensillum of B. mori are predominantly excited by sucrose and myo-inositol, whereas those in medial styloconic sensillum are excited by salicin and caffeine (Fig. 4a). In contrast, the plots for H. armigera show that GRNs in lateral sensillum respond to sucrose, while those in the medial sensillum respond primarily to salicin and myo-inositol (Fig. 4b).

Concentration Dependence of Bitter and Sweet Detection in Styloconic Sensilla
In both species, the GRNs in each sensillum responded to at least one compound (Fig. 5). In most cases, the response appeared phasic-tonic, with the highest firing rate for the first 50-100 ms, followed by sustained firing until the end of the two second stimulation (Online Resource 4). In the GRNs of H. armigera lateral sensilla, the spike amplitudes were larger during the first second of stimulation with sucrose, and then dropped noticeably (Online Resource 4). Similar results were observed in medial sensilla GRNs stimulated by myo-inositol. This phenomenon was not observed in B. mori.
The GRNs in the lateral styloconic sensilla of B. mori exhibited concentration-dependent increases in firing rate with sucrose and myo-inositol (Fig. 5a, b). However, the firing rate with 10 mM salicin or caffeine was not distinguishable from that with 50 mM KCl (Fig. 5a). In H. armigera, the spiking rate of GRNs increased with sucrose in the lateral sensillum and with myo-inositol in the medial sensillum (Fig. 5d, e, f). The GRNs in the medial sensilla of B. mori were sensitive to both salicin and caffeine and exhibited concentration-dependent responses to salicin and caffeine (Fisg. 5a, c). In contrast, the neuronal responses to bitter compounds in most H. armigera styloconic sensilla were low and not distinguishable from the control. Therefore, clear concentration-dependence curves could not be generated for H. armigera's responses to salicin or caffeine.
Response of Individual GRN in B. mori and H. armigera Tiprecording aggregates the responses from the four chemosensory neurons in each of the styloconic sensilla. Judicious examination of spike traces allows some conclusions to be drawn about the neuronal basis of responses, although it is difficult to positively identify neuron types across recordings due to the variation among preparations.
In B. mori lateral styloconic sensilla, sucrose or myo-inositol each generated only one spike shape (Figs. 5a and 6a). When the two sugars were presented in a mixture, both spike shapes were observed with superposition of spikes visible, indicating that different neurons are sensitive to sucrose and myo-inositol (Fig. 6a, Online Resource 5). The amplitudes of spikes from the myo-inositol-sensitive neuron appear slightly higher than those from the sucrose-sensitive neuron. The responses of GRNs in B. mori lateral sensillum to salicin and caffeine showed spikes similar to those observed with 50 mM KCl (Fig. 5a, Online Resource 4).
In B. mori medial styloconic sensilla, the spikes generated by sucrose and myo-inositol were similar to those generated by KCl (Fig. 5a). Salicin and caffeine appeared to generate the same type of spikes, unlike the multiple spikes generated by the sugars or KCl (Fig. 5a). Indeed, the binary mixture of bitter compounds elicited only one type of spike, at higher frequency than with the individual components, but with no evidence of superposition, suggesting that in B. mori a single GRN responds to both salicin and caffeine (Fig. 6c, d).
In H. armigera, sucrose generated a distinct uniform spike type in a GRN of the lateral sensillum, and myo-inositol generated a distinct spike type in a GRN of the medial  a Examples of spike traces from B. mori styloconic sensilla over the first 500 ms of stimulation, with spike shape superimposition. b, c Concentrationdependence of GRN responses to sucrose and myo-inositol in the lateral styloconic sensillum (N=13) and to salicin and caffeine in the medial styloconic sensillum (N=10). d Examples of spike traces from H. armigera styloconic sensilla over the first 500 ms of stimulation, with spike shape superimposition. e, f Concentration-dependence of GRN responses to sucrose in the lateral styloconic sensilla of H. armigera to sucrose (N=11) and to myoinositol in the medial styloconic sensilla (N=13). Chemical concentrations were as follows: KCl=50 mM; sucrose=100 mM; myo-inositol=100 mM; salicin=10 mM; and caffeine=10 mM. Dotted lines indicate the GRN responses to 50 mM KCl and bars indicate the standard errors of the mean sensillum (Fig. 5d). The spike amplitude was higher in these recordings than in those where the sensilla were stimulated by the other test compounds or the KCl control. Salicin or caffeine generated similar spike shapes to controls (Fig. 5d).

Discussion
In the present study, we aimed to compare contact chemoreception and food acceptance in two species of Lepidoptera, the oligophagous silkworm B. mori and the polyphagous cotton bollworm H. armigera. Both species are deterred by bitter substances, and the morphology of their maxillary galea is similar. However, the physiology of the gustatory receptor neurons from the styloconic sensilla differs between the two species. We observed that in B. mori, two different neurons within the lateral sensillum respond to sucrose and myo-inositol, while in H. armigera, one neuron in the lateral styloconic sensillum responds to sucrose and one neuron in the medial styloconic sensillum responds to myo-inositol. In B. mori, a single neuron in the medial sensillum likely responds to both salicin and caffeine. However, we did not find any reliable neuronal responses to bitter compounds in either H. armigera lateral or medial styloconic sensilla.
Sugar Responses Sugars are strong phagostimulants in phytophagous insects (Chapman 2003;Schoonhoven and Van Loon 2002). In a previous study on M. sexta, a facultative specialist, it was shown that myo-inositol is detected by two neurons in each styloconic sensillum, and sucrose by a single  (Glendinning et al. 2007). We found the same for sucrose, but not for myoinositol, which is detected by a single neuron in the lateral sensillum in B. mori or in the medial sensillum in H. armigera (Schoonhoven and Van Loon 2002). The finding that in B. mori sucrose and myo-inositol are detected by two different neurons is in agreement with earlier investigations (Ishikawa 1963(Ishikawa , 1967 and leads us to infer the expression of at least two different sugar receptors in the lateral sensillum. Indeed, in B. mori, two sugar receptors, for myo-inositol and for fructose, recently have been characterized (Sato et al. 2011;Zhang et al. 2011). Contrary to B. mori, H. armigera detection of myo-inositol is mediated by a neuron in the medial sensillum. This finding supports that of Tang et al. (2000). Interestingly, Simmonds and Blaney (1991) reported a second sucrose responsive neuron in the medial sensillum, but did not mention the concentrations tested or whether they tested myo-inositol.
In this study, the concentrations of sucrose and myoinositol that generated strong responses bracketed the levels found in mulberry leaves and used in artificial diet (Ito 1960(Ito , 1967. Myo-inositol has been reported to be an essential nutrient for B. mori (Ito 1967), and has been included in media for lepidopteran cell culture since Grace (Grace 1962). In B. mori, myo-inositol elicits prolonged feeding, but does not act as an initiator (Hamamura et al. 1962). This contrasts with the response of Manduca sexta to myo-inositol, which initiates feeding without increasing consumption (Glendinning et al. 2000). In their study on M. sexta host plants, Nelson and Bernays (1998) could not find any link between the levels of myo-inositol on the leaf surface and sugar or protein levels in the leaf tissue. They noted that the highest concentration of myo-inositol in leaves was approximately 1 mM. Here, however, we show neural responses to 10 −3 mM myo-inositol in both species, albeit stronger in H. armigera, bringing them within the physiological range and supporting the hypothesis that myo-inositol plays a role in initiating feeding in nature.
Bitter Responses Food acceptance by phytophagous insects is determined partly by detection of potentially deterrent secondary plant compounds (Schoonhoven et al. 2005). In this study, we used the deterrents salicin and caffeine, both bitter to humans, which are from two different chemical classes (ß-glycoside and alkaloid, respectively). They have been used commonly in experiments with lepidopterans such as B. mori and M. sexta (Blaney and Simmonds 1988;Glendinning et al. 2006;Ishikawa 1966;Ishikawa and Hirao 1963). Both B. mori and H. armigera were inhibited from feeding by caffeine or salicin at naturally occurring concentrations. This is consistent with previous findings that B. mori rejects food containing salicin and has a "bitter substance detector" (Ishikawa 1966). Helicoverpa armigera is much less behaviorally sensitive to salicin, but more sensitive to caffeine than B. mori. The results reported here may suggest a more complex relationship between antifeeding effects and the identities and concentrations of secondary compounds likely to be encountered. The generalization is known that generalists are behaviorally less sensitive to deterrent compounds than specialists . However, our findings could provide new information on the behavioral responses in these two species.
We also found that only H. armigera's behavioral sensitivity to salicin occurred at higher levels than those found in natural rich sources. Helicoverpa armigera fifth instars indeed were able to grow and pupate normally when fed on 10 mM salicin (24/24 survived to the adult stage, data not shown), which is at the upper end of the concentrations found in nature. The result confirms that this species is either highly tolerant of salicin or has potent detoxifying mechanisms for it. As the behavioral test involves exposing larvae to the bitter compounds for 24 hr, we could not exclude the possibility that feeding inhibition is partly mediated through a post-ingestive toxicity mechanism. However, the behavior observed on initial encounter with bitter substances indicated that feeding inhibition is substantially driven by sensory properties of the antifeedants.
Additionally, the previous observations reporting the presence in B. mori of a GRN in the medial styloconic sensillum that responds to salicin, indicates that it may contribute to the behavioral response (Asaoka 2000;Ishikawa 1963Ishikawa , 1966Ishikawa and Hirao 1963). However, 1-10 mM salicin has also been shown to activate a deterrent cell in the epipharyngeal sensilla (Asaoka and Shibuya 1995), so it is possible that this cell in combination with others in the styloconic sensillum (Dethier 1937(Dethier , 1973Descoins and Marion-Poll 1999;Mori 1982;Tanaka et al. 1994) contributes to the behavioral effect. There was a profound difference between the sensory and behavioral responses of B. mori to salicin and caffeine, which could involve as yet non-obvious differences in firing patterns, or implicate the involvement of other sensory mechanisms. In H. armigera styloconic sensilla, we saw no reliable neuronal responses to caffeine and salicin, suggesting that other sensory organs such as the maxillary palps or epipharyngeal sensilla may be involved in the detection of these compounds, as was reported in M. sexta where three neurons located in the epipharyngeal, lateral and medial styloconic sensilla responded to both salicin and caffeine (Glendinning et al. 2006). Interestingly, the initial response of H. armigera to deterrent compounds involves more biting than in B. mori where drumming seemed to predominate (see Online Resource 2, 3).

Relationship Between Gustatory Sensitivity and Host
Range Based on the current study and the spatial pattern of GRNs in styloconic sensilla responding to various taste stimuli in lepidopterans with different feeding habits (Schoonhoven and Van Loon 2002), it is clear that larvae are likely to express several gustatory receptors specific for various sugars and bitter compounds. It also suggested that, at least in this case, species with different feeding preferences express functionally equivalent gustatory receptors in different locations in the medial and lateral styloconic sensilla. These differences may reflect selective pressures to detect or discriminate a different spectrum of plant compounds for each species. Besides, more complex determinants that may include host range, metabolic capacity and gustatory repertoire could multiply mediate larval sensing of specific plant secondary compounds.