Not all sugars are created equal: some mask aversive tastes better than others in an herbivorous insect

Most insects show immediate appetitive responses to sugars. For instance, the probability that butterflies, bees and flies will extend their proboscis in response to taste stimulation increases with sugar concentration (Dethier, 1976; Schoonhoven and van Loon, 2002; Omura and Honda, 2003; de Brito Sanchez et al., 2005; Dahanukar et al., 2007; Gordon and Scott, 2009; Zhang et al., 2010). Likewise, the rate at which lepidopteran caterpillars either bite from disks (Ma, 1972) or swallow solutions (Sasaki and Asaoka, 2006) increases with sugar concentration. Manduca sexta caterpillars (Sphingidae; Lepidoptera) are unusual because they exhibit a strong peripheral taste response to sugars, but nevertheless fail to show immediate appetitive responses to them – e.g. they bite at the same rate from sugar-treated disks as from water-treated disks (Glendinning et al., 2007). Here, we examined the functional significance of the peripheral gustatory response to sugars in M. sexta.

The solanaceous host plants of M. sexta are laden with noxious compounds (Harborne, 1986; Pomilio, 2008). We hypothesized that the primary function of the peripheral gustatory response to sugars is to mask the aversive taste of these noxious compounds. It is important to note that ingestion of noxious (and potentially toxic) compounds in solanaceous foliage does not typically pose a serious mortality risk to M. sexta caterpillars because they possess an array of mechanisms for coping with toxic compounds. These include: (1) a constitutive insensitivity to some solanaceous poisons (Morris, 1984; Murray et al., 1994); (2) an ability to induce cytochrome P450 detoxification enzymes in response to ingestion of solanaceous compounds (Snyder and Glendinning, 1996; Stevens et al., 2000); (3) post-oral chemosensory mechanisms that permit caterpillars to sense ingested poisons rapidly (Glendinning, 1996), and subsequently regulate their intake in proportion to the extent of P450 enzyme induction (Snyder and Glendinning, 1996); and (4) multidrug transporters that pump alkaloids from the central nervous system to the hemolymph, and then from the hemolymph into the lumen of the malpighian tubules for excretion (Murray et al., 1994; Gaertner et al., 1998).

Manduca sexta caterpillars have eight bilateral pairs of taste sensilla associated with their mouthparts. Each of these sensilla contains three to four gustatory receptor neurons (GRNs), the axons of which project directly to the central nervous system (Schoonhoven and van Loon, 2002). There is a sugar-sensitive GRN in both the lateral and medial styloconic sensilla, which responds to glucose (Glu) and inositol (Ino), and a second sugar-sensitive GRN in the lateral (but not the medial) styloconic sensilla, which responds to sucrose (Suc) (Table 1). Glu, Ino and Suc are all present in the solanaceous foliage that M. sexta caterpillars ingest (Nelson and Bernays, 1998). There is also a bitter-sensitive GRN in both the lateral and medial styloconic sensilla, which responds to a variety of noxious compounds, including caffeine (Caf) and aristolochic acid (AA) (Table 1). Although Caf and AA do not occur in solanaceous foliage, they can rapidly inhibit biting in M. sexta caterpillars by activating the bitter-sensitive GRNs in the lateral and medial styloconic sensilla (Glendinning et al., 1999).

Most models of taste processing (e.g. labeled-line coding) treat each GRN as an independent processing unit (i.e. ‘silo’), which relays its responses to the central gustatory system with high fidelity (de Brito Sanchez and Giurfa, 2011). Although this perspective may be relevant to single-component taste stimuli (e.g. 0.3 mol l^–1 Suc), it is not relevant to complex mixtures of taste stimuli (e.g. what is typically present in foods). This is because inhibitory interactions occur between GRNs within the same sensillum. For instance, when two taste stimuli (e.g. S₁ and S₂) that activate different GRNs (e.g. GRN₁ and GRN₂) are presented in a mixture, insect taste sensilla usually exhibit mixture suppression (i.e. generate fewer action potentials than would be predicted based on the sum of the action potentials elicited by each stimulus alone). Two different types of mixture suppression have been reported. In one type (reciprocal inhibition), S₁ stimulates GRN₁ and S₂ stimulates GRN₂, but when S₁ and S₂ are presented together, the excitatory responses of both GRN₁ and GRN₂ are diminished (Ishikawa, 1963; White et al., 1990; Shields and Mitchell, 1995a; Bernays and Chapman, 2001; Glendinning et al., 2007; Wright et al., 2010). In the other type of mixture suppression (non-reciprocal inhibition), S₁ stimulates GRN₁ and S₂ fails to stimulate any GRN; nevertheless, when S₁ and S₂ are presented together, the excitatory response of GRN₁ is diminished (Dethier and Bowdan, 1989; Bernays and Chapman, 2000; Bernays and Chapman, 2001; de Brito Sanchez et al., 2005).

Previously, we reported that the aversive taste of Caf could be masked by the presence of Ino, but not Glu, in M. sexta caterpillars (Glendinning et al., 2000). Here, we re-examined this prior work based on studies indicating that noxious compounds are differentially sensitive to peripheral mixture suppression (Hiroi et al., 2004), and that binary mixtures of sugars can produce more mixture suppression than single sugars (Shields and Mitchell, 1995b). We tested three sugars (Suc, Glu and Ino) and two noxious compounds (Caf and AA). In Experiment 1, we examined the extent to which the sugars masked the aversive taste of Caf and AA. In Experiment 2, we asked whether any of the observed instances of taste masking could be explained (at least in part) by peripheral mixture suppression.

We maintained a colony of tobacco hornworms [Manduca sexta (Linnaeus 1763); Sphingidae] in our laboratory. This colony was derived from eggs that were donated by the Manduca rearing facility at the University of Arizona, Tucson, AZ, USA. The caterpillars were reared on a wheat-germ-based artificial diet (Bell and Joachim, 1976), and were maintained in an environmental chamber with a 16 h:8 h light:dark cycle at 25°C. The experiments were conducted during the first or second day of the fifth larval growth stage (instar). All caterpillars were naive to the taste stimuli prior to testing. To control for differences between caterpillars from different egg batches, individuals from each batch were interspersed randomly across treatment levels, according to a blind procedure. We provide sample sizes in the figure legends.

We used (1) 200 mmol l^–1 Suc and Glu because it is the lowest concentration of each sugar that elicits a maximal excitatory response in both the lateral and medial styloconic sensilla; (2) 1 mmol l^–1 Ino because it elicits response similar in magnitude to 200 mmol l^–1 Suc in the lateral styloconic sensillum; and (3) 0.1 mmol l^–1 AA (sodium salt) and 5 mmol l^–1 Caf because they elicit robust aversive responses in M. sexta (Glendinning et al., 2006). We purchased all chemicals from Sigma-Aldrich (St Louis, MO, USA).

During the behavioral and electrophysiological tests, we presented each caterpillar with one of two test series. For the AA test series, we included a negative control stimulus (water), a positive control stimulus (AA alone), AA plus one of three sugars (Suc, Glu and Ino), and AA plus binary mixtures of the three sugars (Suc+Glu, Suc+Ino or Glu+Ino). The Caf test series was the same in all respects, except that we used Caf instead of AA. For the electrophysiological tests, we also tested Suc+Glu, Suc+Ino and Glu+Ino.

For the biting assay, the chemical stimuli were dissolved in deionized water. For the electrophysiological recordings, they were dissolved in an electrolyte solution (i.e. 100 mmol l^–1 KCl) and presented at room temperature (i.e. 22–24°C).

We measured the number of bites taken from the experimental disk during the initial 10 s of the meal. This disk consisted of a 4.25 cm diameter glass-fiber disk (Whatman GF/A) treated with 400 μl of each test solution. We ran the behavioral test immediately after applying the test solution so as to minimize evaporative water loss. By focusing on the initial 10 s of biting, we minimized the contribution of any post-ingestive effects of the test solutions. To avoid experiential effects, we ran each caterpillar through a single biting assay.

The assay consisted of four stages. (1) We placed a caterpillar in the ‘food-deprivation arena’, which consisted of a clean (inverted) Petri dish covered by a clear plastic cylinder (7.5 cm in diameter, 10 cm tall). We fasted the caterpillar in this arena for 30 min to standardize its ‘hunger’ state. (2) Then, we transferred the caterpillar to the ‘test arena’, which was identical to the food-deprivation arena in all respects except that it contained a piece of cork in the center, containing a disk pre-moistened with 400 μl experimental solution. The observer was kept blind to the identity of the test solution. (3) Next, we positioned the caterpillar on the edge of the glass-fiber disk so that its legs and prolegs grasped the edge. (4) Finally, we recorded the number of bites taken from the disk over the first 10 s of feeding.

To ensure that the observer could record the number of bites accurately and without bias, we positioned the test arena on a turntable-like device, which permitted us to rotate the turntable and keep the caterpillar’s mandibles clearly visible as it fed.

To compare the number of bites taken from each experimental solution, we used a two-way factorial ANOVA. One factor was test series (AA or Caf test series) and the second factor was test solution (water, noxious compound alone, or noxious compound plus sugars). We used a post hoc Tukey’s test to determine the relative number of bites taken from each test solution, separately for the AA or Caf test series. We inferred that an experimental disk (e.g. a disk treated with 5 mmol l^–1 Caf) was less acceptable than the control disk (i.e. a disk treated with water alone) if the caterpillars took significantly fewer bites from the experimental disk. For each factorial ANOVA, we confirmed that each dependent measure met the normality assumption (Kolmogorov–Smirnov test), and that there was homogeneity of variance across treatment levels (Levene’s test). In all comparisons, we set the alpha level at 0.05.

We focused on the lateral and medial styloconic sensilla, each of which responds to sugars and noxious plant compounds (see Table 1 for the molecular receptive range of the each GRN). Our goal was to determine how the presence of the sugars influenced peripheral taste responses to Caf or AA.

For each caterpillar, we recorded neural responses of a lateral and medial styloconic sensilla to the AA or Caf test series, using a previously described tip recording technique (Glendinning et al., 2007). Neural recordings were pre-amplified (10×) with a TasteProbe (Syntech, Hilversum, The Netherlands), filtered (100–1200 Hz), digitized (IDAC-4, Syntech) and analyzed with Autospike software (Syntech).

We tallied the total number of action potentials during each successive 100 ms bin across the initial 1000 ms of each recording. Because many of the recordings contained responses from multiple GRNs with temporally variable spike amplitudes, it was impossible to assign spikes based solely on amplitude and shape. As a result, we incorporated the distinct temporal pattern of firing that each stimulus elicited into our spike assignments.

We used repeated-measures ANOVAs to ascertain whether the instantaneous firing rate changed systematically with time. We confirmed equality of variance across treatment levels of a repeated-measures factor with Mauchly’s sphericity test. We compared the total number of spikes that the stimulus elicited over 1000 ms with that elicited by the KCl solution, using a paired t-test. We inferred that the test solution elicited a detectable response when total number of spikes differed significantly from than that elicited by the electrolyte solution. In comparisons involving multiple t-tests, we used the Bonferroni correction.

To test for mixture suppression, we first compared the number of the spikes elicited by the stimulus mixture (i.e. mixed condition) with the total number of spikes elicited by each component of the stimulus mixture separately (i.e. unmixed condition). We made no attempt to discriminate between spikes from different GRNs; we simply counted the total number of spikes that occurred during the initial 1000 ms of response. Second, we compared the magnitude of mixture suppression across the different stimulus mixtures with a repeated-measures ANOVA and a Tukey-type post hoc test (corrected for a within-subject design). We calculated percent mixture suppression as follows: 100×[1–(number of spikes during the mixed condition/number of spikes during the unmixed condition)]. We confirmed equality of variance across treatment levels of the repeated-measures factor with Mauchly’s sphericity test. Third, we used mixed-model ANOVA and unpaired t-tests to compare % mixture suppression in the AA test series with that in the Caf test series. Finally, we conducted a fine-grained analysis of mixture suppression. To this end, we examined changes in instantaneous firing rates and spike amplitude over time. We also compared the total number of spikes (over the initial 1000 ms of response) elicited by each binary (or ternary) mixture with that elicited by each of its component stimuli, using a repeated-measures ANOVA and a post hoc Tukey’s test. We used these analyses to assign spikes to individual cells (whenever possible), and to draw inferences into the nature of the mixture suppression (e.g. relative contribution of reciprocal versus non-reciprocal inhibition).

The caterpillars exhibited nearly continuous biting activity during the trials with water alone (producing approximately 20 bites), but highly disrupted (or inhibited) biting activity during the trials with AA alone or Caf alone (producing ≤8 bites; Fig. 1). The addition of sugars to the AA (or Caf) solution inhibited biting to varying degrees. There was a significant main effect of test solution series (F_1,311=86.2, P≤0.05), revealing that the caterpillars generally took fewer bites from stimulus mixtures in the AA test series than in the Caf test series. There was also a significant main effect of stimulus mixture (F_7,311=46.4, P≤0.05) and a significant interaction of test solution series × stimulus mixture (F_7,311=4.3, P≤0.05).

For the AA test series, the relative acceptability was: water=AA+Suc+Ino>AA+Suc+Glu=AA+Glu+Ino=AA+Suc>AA alone (Fig. 1, left panel). The number of bites taken for AA+Glu and AA+Ino could not be distinguished from that taken for AA alone. AA+Suc+Ino was the only solution that was as acceptable as water. Although AA+Suc+Glu, AA+Glu+Ino and AA+Suc were more acceptable than AA alone, they were significantly less acceptable than water.

For the Caf test series, the relative acceptability was: water = Caf+Suc+Ino = Caf+Suc+Glu > Caf+Suc = Caf+Ino > Caf alone (Fig. 1, right panel). The number of bites taken for Caf+Glu could not be distinguished from that taken for Caf alone. Caf+Suc+Ino and Caf+Suc+Glu were the only solutions that were as acceptable as water. Even though Caf+Glu+Ino, Caf+Suc and Caf+Ino were more acceptable than Caf alone, they were still less acceptable than water.

The neural response to AA in both sensilla was unique in two respects: both spike amplitudes and instantaneous firing rates increased over time (Fig. 2). In contrast, the spike amplitudes elicited by Caf and the sugars either remained constant or decreased over time in both sensilla. The firing rates generated by KCl did not change systematically over time in either sensillum.

In Fig. 3, we compare total spikes elicited by each plant chemical with that elicited by KCl alone (the reference stimulus). In the lateral styloconic sensillum, the total number of spikes elicited by Caf, AA, Suc, Ino and Glu were all significantly greater than that elicited by KCl alone. In the medial styloconic sensillum, the total number of spikes elicited by AA, Ino and Glu (but not Caf and Suc) was significantly greater than that elicited by KCl alone.

All sugar solutions, except Ino, inhibited the response of the lateral and medial styloconic sensilla to both AA and Caf (Fig. 4A). The magnitude of mixture suppression, however, varied greatly across the test solutions (Fig. 4B). For solutions containing AA, the magnitude of mixture suppression was greatest for AA+Suc+Ino in the lateral styloconic sensillum, and for AA+Suc+Ino and AA+Suc+Glu in the medial styloconic sensillum. For solutions containing Caf, the magnitude of mixture suppression was greatest for Caf+Suc+Ino and Caf+Suc+Glu in the lateral styloconic sensillum. There were no significant differences in magnitude of mixture suppression across the stimulus mixtures in the medial styloconic sensillum. Finally, the magnitude of mixture suppression in the lateral styloconic sensillum was significantly greater for the AA test series than for the Caf test series (main effect of noxious compound: F_1,78=12.3, P≤0.05); no such difference was observed in the medial styloconic sensillum (main effect of noxious compound: F_1,78=0.2, P>0.05; Fig. 4C).

Mixture suppression also occurred in test solutions containing sugars alone (i.e. Suc+Ino, Suc+Glu and Ino+Glu; Fig. 5). The magnitude of mixture suppression was greatest for Glu+Suc in the lateral styloconic sensillum, and for Ino+Suc in the medial styloconic sensillum.

The presence of Suc altered the peripheral response to AA and Caf in different ways (Fig. 6, supplementary material Fig. S1). The instantaneous firing rate for AA+Suc increased significantly with time in both sensilla (F_9,69>6.1, P≤0.05), and resembled the response to AA alone (Fig. 6A, top trace). However, the total number of spikes elicited over the initial 1000 ms was significantly less than that elicited by AA alone (supplementary material Fig. S1A). These observations indicate that Suc partially inhibited the response to AA in both sensilla.

Caf+Suc elicited an instantaneous firing rate that decreased significantly with time in both sensilla (F_9,79>3.1, P≤0.05). However, because of the weak responses of the medial styloconic sensilla to Caf, Suc and Caf+Suc (supplementary material Fig. S1B), it is difficult to draw any inferences about mixture interactions. In contrast, there were two distinctive features of the response of the lateral styloconic sensilla to Caf+Suc: it contained significantly more total spikes than did the response to Caf or Suc alone, and multiple GRNs fired out of phase with one another (Fig. 6B). This indicates that Suc produced only limited inhibition of the response to Caf.

The instantaneous firing rate elicited by AA+Glu decreased over time in both sensilla (F_9,69>3.8, P≤0.05; supplementary material Fig. S2A). The high firing rates during the initial 200 ms were virtually identical in intensity and duration to those elicited by Glu alone. Further, the firing rates during the final 800 ms were more intense than those elicited by Glu alone, implicating a contribution of AA.

Caf+Glu elicited an instantaneous firing rate that decreased over time in both sensilla (F_9,79>12.7, P≤0.05), and resembled the temporal pattern of firing elicited by Glu alone (supplementary material Fig. S2B). The total number of spikes generated by Caf+Glu was significantly greater than that generated by Glu alone in the lateral but not the medial styloconic sensillum (supplementary material Fig. S2B).

The instantaneous firing rate elicited by AA+Ino changed significantly over time in both sensilla (F_9,69>6.7, P≤0.05; supplementary material Fig. S3A). In the lateral styloconic sensillum, the temporal pattern of firing appears to reflect a composite of the response to both AA and Ino. For example, in Fig. 6A (second trace from top), the initial 200 ms resembles the phasic response to Ino, and the final 700 ms resembles the increasing firing rate to AA. Although the instantaneous firing rate in the medial styloconic sensillum followed a typical phasic–tonic pattern, it also appears to contain a mixture of the responses to both AA and Ino.

The neural responses to Caf+Ino in both sensilla resembled a composite of that to Caf and Ino alone (supplementary material Fig. S3B). There are elements of both the phasic response to Ino and the tonic response to Caf (Fig. 6B).

In response to AA+Ino+Glu, the instantaneous firing rate decreased over time in the medial (F_9,69=10.6, P≤0.05) but not in the lateral styloconic sensilla (F_9,69=0.7, P>0.05; supplementary material Fig. S4A, left panels). Further, the total number of spikes elicited by AA+Ino+Glu was significantly greater than that elicited by AA alone in the medial but not the lateral styloconic sensillum (supplementary material Fig. S4A, right panels). These findings indicate that the binary mixture of Ino+Glu largely eliminated the AA response in the lateral styloconic sensillum, but failed to do so in the medial styloconic sensillum.

Caf+Ino+Glu elicited an instantaneous firing rate that decreased significantly over time in both sensilla (F_9,79>25.3, P≤0.05; supplementary material Fig. S4B, left panels). These decelerating firing rates presumably stem from strong phasic responses to Glu and Ino.

The instantaneous firing rate elicited by AA+Suc+Glu changed slightly but significantly over time in both sensilla (F_9,69>2.3, P≤0.05; supplementary material Fig. S5A). These temporal changes included a small phasic response (during the initial 300 ms) followed by an increasing instantaneous firing rate over the remaining 700 ms. The small phasic response to AA+Suc+Glu probably reflects the inhibitory effect of Suc on the initial phasic response to Glu, which is apparent particularly in the medial sensillum (supplementary material Fig. S6).

Caf+Suc+Glu elicited an instantaneous firing that decreased significantly over time in both sensilla (F_9,79>7.0, P≤0.05; supplementary material Fig. S5B). In the lateral styloconic sensillum, Caf appears to have contributed to the vigorous response to Caf+Suc+Glu. This is based on the observation that Suc+Glu generated the same number of total spikes as Suc or Glu alone (supplementary material Fig. S6), whereas Caf+Suc+Glu generated nearly three times more spikes than did Suc+Glu (supplementary material Fig. S5B). This suggests that Suc+Glu did not effectively inhibit the response of the lateral styloconic sensillum to Caf.

The instantaneous firing rate elicited by AA+Ino+Suc did not change significantly over time in either sensilla (F_9,69<1.4, P>0.05; supplementary material Fig. S7A). Further, the total number of spikes generated by AA+Ino+Suc was significantly less than that generated by AA alone in the lateral sensillum. These results indicate that the response to AA was completely inhibited in the ternary mixture of AA+Ino+Suc.

Caf+Ino+Suc elicited an instantaneous firing that decreased significantly over time in both sensilla (F_9,79>2.1, P≤0.05; supplementary material Fig. S7B). In the lateral styloconic sensillum, Caf likely contributed to the vigorous response to Caf+Ino+Suc. This is based on the observation that Ino+Suc generated the same number of total spikes as did Ino or Suc alone (supplementary material Fig. S8), whereas Caf+Ino+Suc generated nearly two times more spikes than did Ino+Suc (supplementary material Fig. S7B). This suggests that Ino+Suc did not effectively inhibit the response of the lateral styloconic sensillum to Caf.

In many species of insect, sugars cause immediate stimulation of feeding. Sugars do not have this effect in M. sexta, despite causing strong peripheral taste responses (Glendinning et al., 2007). To explain this conundrum, we hypothesized that the primary function of the peripheral taste response to sugars is to mask the taste of aversive compounds. We found strong support for this hypothesis in our short-term biting assay. Some, but not all, of the sugars masked the taste of AA and Caf. The most effective masking agent was Suc+Ino; it completely masked the taste of both AA and Caf. Although Suc+Glu completely masked the taste of Caf, it only partially masked that of AA. Suc was the only single-component sugar stimulus that partially masked the taste of both AA and Caf. Glu, in contrast, was the only single-component sugar stimulus that failed to even partially mask the taste of AA or Caf.

To assess the contribution of peripheral mixture suppression to the behavioral results, we examined how the medial and lateral styloconic sensilla responded to binary and ternary mixtures of the sugars and noxious compounds. We observed extensive mixture suppression, both between sugars and noxious compounds and between the different sugars (Figs 4, 6).

The extent of mixture suppression differed greatly across the binary and ternary mixtures. At one extreme was Ino. Despite eliciting a vigorous response in the lateral and medial styloconic sensilla, Ino did not produce any mixture suppression when mixed with AA or Caf. Indeed, the neural responses to AA+Ino or Caf+Ino appeared to reflect a composite of the response to each stimulus component alone (supplementary material Fig. S3). Given that Ino stimulates a different GRN than both AA and Caf (Table 1), it appears that each GRN was able to respond to its own ligand without inhibiting the response of the neighboring GRNs. At the other extreme was Ino+Suc, which produced the most extensive mixture suppression. Based on the fact that the neural responses to AA+Ino+Suc in both sensilla lacked any of the features typical of the AA response (i.e. spike amplitudes and instantaneous firing rates that increase over time; supplementary material Fig. S6), it appears that Ino+Suc strongly inhibited the response of the AA-sensitive GRN in both the lateral and medial styloconic sensillum.

It is notable that the magnitude of mixture suppression was, in most cases, significantly greater for solutions containing AA than for those containing Caf in both the medial and lateral styloconic sensillum (Fig. 4C). This finding is even more remarkable given that Caf and AA are thought to stimulate the same GRN in the lateral styloconic sensillum (and perhaps in the medial styloconic sensillum as well) (Glendinning and Hills, 1997). One explanation for this result stems from the observation that Caf and AA stimulate different signaling pathways within the same bitter-sensitive GRN, and that these pathways generate distinct temporal patterns of spiking (Glendinning and Hills, 1997). The peripheral taste response to AA has a relatively long latency of activation (i.e. >200 ms; Figs 2, 3), which could make it more vulnerable to inhibitory effects of neighboring sugar-sensitive GRNs, which reach their maximal firing rates more quickly (i.e. <100 ms; e.g. Fig. 2). This explanation could be tested by evaluating mixture suppression with noxious compounds that produce temporal patterns of spiking that vary in latency of activation. Accordingly, mixture suppression should increase with the latency of activation.

There was no obvious relationship between the firing rate elicited by a sugar-sensitive GRN and the magnitude of mixture suppression it caused. For instance, when Suc was presented alone, it elicited a feeble excitatory response in the medial styloconic sensillum, but when presented in a binary mixture (i.e. AA+Suc, Suc+Ino or Suc+Glu), it produced robust mixture suppression (i.e. >40%). In contrast, when Ino was presented alone, it elicited a vigorous excitatory response in the medial styloconic sensillum, but when presented in a binary mixture (i.e. AA+Ino, AA+Glu and Ino+Glu), it produced modest mixture suppression (i.e. <40%).

Our findings indicate that mixture suppression in the lateral styloconic sensillum was mediated largely by reciprocal inhibition. This is because when each of the plant chemicals was presented individually, it elicited a strong excitatory response in a single GRN. However, when the same chemicals were presented in binary or ternary mixtures, they caused varying degrees of inhibition in multiple GRNs (Fig. 3). The only exceptions to this pattern were the binary mixtures of AA+Ino and Caf+Ino, which failed to produce any mixture suppression. Although little is known about the mechanistic basis of mixture suppression in insect taste sensilla, the most parsimonious explanation is that activation of one GRN stimulates paracrine release into the sensillar lumen, which modulates the response of neighboring GRNs. There is widespread evidence for these types of modulatory signaling networks in mammalian taste buds (Herness and Zhao, 2009).

In the medial styloconic sensillum, mixture suppression was mediated by reciprocal and non-reciprocal inhibition. For instance, AA, Glu and Ino each elicited a strong excitatory response in a single GRN when presented individually, but caused varying degrees of reciprocal inhibition in multiple GRNs when presented in binary or ternary mixtures. In contrast, Suc produced a feeble excitatory response when presented alone (Fig. 3), but caused non-reciprocal inhibition of the excitatory responses of GRNs to the other plant chemicals (Figs 4, 5, 6). One possible explanation for this latter case of non-reciprocal inhibition is that Suc directly antagonized the signaling pathways for AA, Glu and Ino (see Talavera et al., 2008).

Peripheral mixture suppression was also observed between sugars (Fig. 5). This finding is notable because it indicates that some of the mixture suppression produced by the ternary mixtures (Fig. 4) was due to inhibitory interactions among the sugars. It is important to note, however, that the existence of peripheral mixture suppression between sugars is tangential to the question of whether sugars inhibited the peripheral gustatory response to AA and Caf. Indeed, we were able to establish specific inhibitory effects of the sugars on responses of the bitter-sensitive GRNs.

Before discussing the neural basis of the biting responses, it is necessary to highlight four caveats about our experimental approach. First, we did not record from the epipharyngeal sensilla because they do not generate excitatory responses to sugars (de Boer et al., 1977; Glendinning et al., 2000). Although prior reports indicate that the excitatory response of epipharyngeal sensilla to Caf is suppressed by Glu but not Ino (Glendinning et al., 2000), we are not certain how this class of sensillum contributed to our behavioral results. Second, because little is known about the actual sugar concentrations that M. sexta caterpillars encounter while ingesting solanaceous foliage, we selected concentrations that elicit maximal responses in the sugar-sensitive GRNs (Glendinning et al., 2007). We reasoned that these high concentrations would provide the clearest evidence of peripheral mixture suppression. Third, we used Caf and AA as model noxious plant compounds because they strongly activate the peripheral taste system of M. sexta caterpillars. Additional studies are needed to determine whether there are noxious compounds in solanaceous foliage that activate the same GRNs as Caf and AA, and how sugars influence taste responses to these compounds. Fourth, because consumption of host-plant foliage can alter the responses of the peripheral taste system of M. sexta (del Campo et al., 2001; Glendinning et al., 2009), it is possible that the nature of the mixture suppression differs between caterpillars reared on artificial diet and those reared on host-plant tissue. Notwithstanding these caveats, we discuss below the extent to which peripheral mixture suppression contributed to the masking phenomenon.

For the AA test series, the extent to which each sugar masked the taste of AA increased with the magnitude of peripheral mixture suppression. For instance, the one sugar stimulus that completely masked the aversive taste of AA (Suc+Ino) also strongly inhibited the peripheral taste of both sensilla to AA. The three sugar stimuli that partially masked the aversive taste of AA (Suc, Glu+Ino and Suc+Glu) also partially blocked the response of both sensilla to AA. Finally, the two sugar stimuli that failed to mask the aversive taste of AA (Glu and Ino) also provided the weakest suppression of the peripheral response to AA. The strong correlation between the acceptability of the AA-containing solutions and the magnitude of peripheral taste suppression points to a causal connection between these two phenomena.

For the Caf test series, there was no clear relationship between taste acceptability of each stimulus mixture and magnitude of peripheral taste suppression. For instance, even though Suc+Glu and Suc+Ino completely masked the aversive taste of Caf, they only partially suppressed the peripheral response to Caf. Likewise, even though Ino partially masked the aversive taste of Caf, it failed to produce any significant suppression of the peripheral taste response to Caf. These findings, together with those of an earlier report (Glendinning et al., 2000), indicate that the ability of the sugars to mask the aversive taste of Caf cannot be explained by peripheral taste suppression alone. We propose, instead, that sugars masked the taste of Caf by activating inhibitory mechanisms in both the peripheral and central taste systems. A similar explanation has been proposed to explain how sugars mask the aversive taste of quinine in mammals (Kroeze and Bartoshuk, 1985; Formaker and Frank, 1996).

This study, together with prior reports (Glendinning et al., 2000; Glendinning et al., 2007), indicates that sugars can modulate immediate appetitive responses of M. sexta caterpillars, but only when they are presented together with noxious plant compounds. It is important to emphasize that this result does not preclude a role of sugars in the stimulation of feeding. For instance, some sugar solutions (e.g. Suc+Ino) can increase meal length in M. sexta caterpillars (Glendinning et al., 2007). However, it took more than 6 min of feeding before the effect of Suc+Ino on meal length manifested itself. This indicates that the stimulation of feeding was mediated primarily by a post-oral mechanism (Burke and Waddell, 2011; Dus et al., 2011; Fujita and Tanimura, 2011).

Although all sugars modulated taste responses to AA and Caf, Suc had the greatest impact. For instance, the only binary sugar mixtures that completely masked the taste of AA and/or Caf contained Suc (i.e. Suc+Ino and Suc+Glu). Likewise, the only single-component sugar that partially blocked the taste of AA was Suc. For the peripheral taste recordings, the sugar stimuli that produced the greatest mixture suppression almost always contained Suc. This finding is remarkable because M. sexta possesses only one bilateral pair of GRNs that responds to Suc, and two bilateral pairs of taste sensilla that respond to Glu and Ino (Table 1). If the ability of sugars to mask the taste of an aversive compound was determined simply by the relative amount of taste input from sugar-sensitive GRNs, then one would expect that mixtures of Glu and Ino would be more effective than Suc (Figs 3, 5). The fact that we observed the opposite result indicates that the taste system of M. sexta caterpillars pays particular attention to input from the bilateral pair of sucrose-sensitive GRNs. More information about the composition of host-plant foliage is needed to explain the functional significance of Suc (or input from the sucrose-sensitive GRN) to the feeding behavior of M. sexta caterpillars.