How Do Inositol and Glucose Modulate Feeding in Manduca Sexta Caterpillars

Sugars play a critical role in determining the palatability of foods in vertebrates and invertebrates, and are often considered to be universal feeding stimulants (Dethier, 1976; Ramirez, 1990). With few exceptions (see, for example, Martinez del Rio et al., 1988), increasing the concentration of sucrose, glucose or fructose in foods causes animals to feed more vigorously and for longer periods. These feeding responses are mediated by sensory input from sugar-sensitive taste cells and post-ingestive response mechanisms (Mook, 1963; Davis, 1973; Jakinovich and Sugarman, 1989; Bernays and Chapman, 1994). There is another class of carbohydrate, the sugar alcohols, which can also enhance the palatability of foods. Sugar alcohols evoke a sweet sensation in humans and other vertebrates (Moskowitz, 1971; Jakinovich and Sugarman, 1989) and stimulate feeding in many species of insect (Dethier, 1976; Bernays and Chapman, 1994). Like sugars, they are widespread in the plant kingdom (Plouvier, 1963; Harborne and Baxter, 1993) and represent a rich source of calories. However, in contrast to sugars, comparatively little is known about the neural response mechanisms that mediate feeding responses to sugar alcohols (see, for example, Jakinovich and Sugarman, 1989).

Much of our knowledge about how animals detect sugar alcohols has been obtained from a single order of insect (Lepidoptera) and a single class of sugar alcohol (inositol). Many species of caterpillar possess identified taste cells that respond vigorously to inositol (e.g. Dethier and Kuch, 1971; Schoonhoven, 1972). In some species, the inositol-sensitive taste cells respond to sugars and several sugar alcohols, whereas in other species, they respond exclusively to specific isomers of inositol (Jakinovich and Agranoff, 1971; Schoonhoven, 1974; den Otter, 1992). In most inositol-sensitive taste cells, the dynamic range for inositol (i.e. the steepest portion of the concentration/response curve; Jakinovich and Agranoff, 1971; Blom, 1978; den Otter, 1992; Asaoka and Shibuya, 1995; Bernays et al., 1998) overlaps with the known range of inositol concentrations in plant tissues (i.e. 0.5–10 mmol l⁻¹; Schoonhoven, 1969a; Nelson and Bernays, 1998). This latter observation indicates that the inositol-sensitive taste cells would be ideally suited for detecting small differences in inositol concentration among different leaves. Given that inositol is an essential nutrient in some caterpillar species (Horie et al., 1966) and a utilizable, but non-essential, carbohydrate in others (Nelson, 1996), caterpillars should benefit by feeding on plant tissues containing these compounds. To date, however, little is known about how sensory input from these inositol-sensitive taste cells modulates feeding.

Our current understanding of the effects of dietary inositol and sugars on feeding in caterpillars stems primarily from long-term feeding studies. These studies revealed that inositols and sugars can (i) cause a concentration-dependent increase in consumption (Blom, 1978; Städler and Hanson, 1978; but see Ma, 1976, for a contradictory finding), (ii) synergize the phagostimulatory effect of plant nutrients (Hamamura et al., 1962; Ishikawa and Hirao, 1966) and (iii) make diets containing aversive plant compounds more acceptable (Schoonhoven, 1969a; Shields and Mitchell, 1995a). Our ability to relate these behavioral findings to sensory input from taste cells is limited, however, by two methodological features of the above-cited studies. First, they were run over a period of 3 h or more, making it impossible to determine the relative contribution of gustatory input versus post-ingestive feedback to the behavioral responses. For instance, post-ingestive feedback from sugars and toxic compounds can modulate ingestive behavior within seconds or minutes of the initiation of feeding by caterpillars (Timmins and Reynolds, 1992; Glendinning and Slansky Jr, 1994; Glendinning, 1996). Second, the long-term feeding studies measured the outcome of behavior (e.g. the total amount of diet eaten or frass produced), rather than the behavior itself (e.g. the temporal pattern of biting). This makes it impossible to ascertain how the inositols or sugars actually modulated the ingestive process. For instance, compounds that act via gustatory input usually alter a variety feeding parameters (e.g. the probability of initiating feeding, biting rate, bite size and/or feeding duration) within seconds of initiating feeding (Beck, 1965; Ma, 1972; Glendinning et al., 1999b).

In the present study, we examined how inositol and sugars modulate the feeding responses of Manduca sexta caterpillars (Sphingidae). Several laboratories have reported that M. sexta exhibits phagostimulatory responses to diets containing inositol or sugars (Yamamoto and Fraenkel, 1960; Städler and Hanson, 1978; Bowdan, 1995) and that it feeds readily on non-host plant tissues that have been treated with inositol (Schoonhoven, 1969a). However, the relative contribution of gustatory input versus post-ingestive feedback to these behavioral responses was not assessed. The possibility that these feeding responses were mediated by gustatory input stems from the observation that this caterpillar possesses separate inositol- and sugar-sensitive taste cells in each of its medial and lateral styloconic sensilla (Schoonhoven, 1972; Frazier, 1986; Lam and Frazier, 1991). M. sexta may contain additional inositol- and/or sugar-sensitive taste cells in its other two classes of gustatory sensilla (i.e. the epipharyngeal sensilla and maxillary palp sensilla), but few studies have examined the response properties of taste cells in the maxillary palp and epipharyngeal sensilla. One group of investigators specifically looked for sugar-sensitive (but not inositol-sensitive) taste cells in the epipharyngeal sensilla of M. sexta (de Boer et al., 1977). Even though they never located taste cells that exhibited excitatory responses to sugars, they did locate a bitter-sensitive taste cell that was inhibited by sugars.

Our study addressed four interrelated issues. First, to identify the entire population of inositol- and sugar-sensitive taste cells, we determined whether the epipharyngeal or maxillary palp sensilla contained taste cells that exhibit excitatory responses to myo-inositol or glucose. Second, to examine the molecular receptive range of the inositol- and sugar-sensitive taste cells, we stimulated each with a variety of sugars, amino acids, sugar alcohols and a flavenoid glycoside, rutin (a non-nutritive feeding stimulant). Third, we studied how myo-inositol and glucose alter long-term and short-term feeding responses to various taste stimuli, in both early and late instar caterpillars. Finally, we asked whether the behavioral responses to taste stimuli in binary mixtures could be explained by peripheral interactions among neighboring taste cells. There is evidence from other caterpillar species that excitatory responses of inositol- and sugar-sensitive taste cells can inhibit the responsiveness of the so-called bitter-sensitive taste cells and thereby render diets containing ‘bitter’ taste stimuli more acceptable (see, for example, Shields and Mitchell, 1995c).

We reared Manduca sexta caterpillars from eggs on a wheat-germ-based artificial diet (Bell and Joachim, 1976). We maintained the caterpillars in an environmental chamber on a 16 h:8 h light:dark cycle (27.5 °C:25.5 °C). We used caterpillars 1–2 days into their fifth stadium in most experiments (see below for details). All caterpillars were naive to the taste stimuli prior to testing. To control for any potential differences among caterpillars from different egg batches, we interspersed individuals from each batch across experimental treatments. The number of caterpillars tested in each experiment is indicated in the figure legends.

The goal of this experiment (experiment 1) was to identify the entire population of taste cells that respond in an excitatory manner to myo-inositol or glucose. Once we had identified these taste cells, we determined concentration/response functions for each compound. We felt that this information would provide critical baseline information for interpreting the behavioral data that we present in subsequent experiments.

We recorded the neural responses of the taste cells using a non-invasive extracellular tip-recording technique (Gothilf and Hanson, 1994; Asaoka and Shibuya, 1995; Glendinning et al., 1998). In brief, this involved placing a glass electrode (containing a specific taste stimulus) over the tip of a chemosensory sensillum (or multiple sensilla in the case of the maxillary palp tip) and recording excitatory responses of taste cells. Using this technique, we could specifically stimulate individual epipharyngeal, lateral and medial maxillary styloconic sensilla, but not individual maxillary palp sensilla because of their diminutive size.

We dissolved all taste stimuli in an electrolyte solution (100 mmol l⁻¹ KCl) for conductivity. We processed the neural recordings with a high-impedance preamplifier using a baseline-restoring circuit (Frazier and Hanson, 1986) and an amplifier-filter with a bandwidth set at 130–1200 Hz. We digitized and stored neural recordings directly onto a computer with SAPID tools (Smith et al., 1990).

For each neural recording, we stimulated a sensillum (or entire maxillary palp tip) for approximately 1500 ms, and quantified the number of action potentials generated 10–1010 ms after contact with the sensillum (or entire maxillary palp tip). We paused for at least 3 min between successive stimulations of a sensillum. To minimize the effects of solvent evaporation at the tip of the recording/stimulating electrode, we drew fluid from the tip with a piece of filter paper 7–10 s before each stimulation. We tested only one member of each bilateral pair of gustatory sensilla per caterpillar.

We initially stimulated each of the gustatory sensilla with two concentrations of myo-inositol (10 and 100 mmol l⁻¹) and two concentrations of glucose (100 and 250 mmol l⁻¹) (Sigma). These concentrations of inositol and sugar were selected because they have been found to produce maximal excitatory responses of taste cells in all other species of caterpillar studied (for references, see Introduction). Because our myo-inositol sample was derived from corn, it probably consisted predominantly of the natural cis configuration. Even though the primary goal of this experiment was to determine whether the epipharyngeal or maxillary palp sensilla contained inositol-or sugar-sensitive taste cells, we also tested the lateral and medial styloconic sensilla as positive controls. When we discovered a sensillum containing a taste cell that responded vigorously to myo-inositol (or glucose), we subsequently stimulated that taste cell with an ascending series of myo-inositol (0.1, 1.0, 10 and 50 mmol l⁻¹) or D-glucose (10, 50, 100 and 250 mmol l⁻¹) concentrations (Sigma). We could readily discriminate action potentials from the inositol- and sugar-sensitive taste cells (within the same sensillum) on the basis of spike amplitude and the temporal pattern of firing. Spikes from the inositol-sensitive taste cell have a larger amplitude and a higher and more protracted firing rate during the initial phasic portion of the neural response (see, for example, Gothilf and Hanson, 1994; Glendinning and Hills, 1997; Stitt et al., 1998).

A previous study of M. sexta concluded that the inositol-sensitive taste cells in the lateral and medial sensilla respond exclusively to myo-inositol, and that the sugar-sensitive taste cells (in the same two sensilla) respond exclusively to sugars (Schoonhoven, 1969a). This study did not indicate, however, which taste stimuli were tested to substantiate this conclusion. Several species of caterpillar (Schoonhoven, 1974; Asaoka and Akai, 1991) and fly (Dethier, 1976) are known to possess individual taste cells that also respond to a structurally diverse range of sugar alcohols, sugars and, in some instances, amino acids. In this experiment (experiment 2), we determined whether the inositol-or sugar-sensitive taste cells in the lateral and medial styloconic sensilla of M. sexta also respond to sugar alcohols (other than inositol), amino acids or rutin (a feeding stimulant).

We chose L-glutamine, L-asparagine and L-proline (Sigma) because they are abundant as free amino acids in many plant tissues (Harborne and Baxter, 1993) and are known to stimulate taste cells in a variety of caterpillar species (Dethier, 1973; van Loon and van Eeuwijk, 1989; Panzuto and Albert, 1998). We selected rutin (Sigma), a glycoside abundant in many plant tissues (Harborne and Baxter, 1993), because it stimulates taste cells in several species of caterpillar (Dethier, 1973) and is a phagostimulant for some insects (Bernays et al., 1991). We also tested two straight-chain sugar alcohols, D-mannitol (Mallickrodt) and D-sorbitol (Sigma), and three cyclic sugar alcohols, myo-inositol, L-quebrachitol (Aldrich) and D-pinitol (courtesy of Richard Jensen, University of Arizona). Of these sugar alcohols, only myo-inositol has been reported in solanaceous plants (Yamamoto and Fraenkel, 1960; Nelson and Bernays, 1998). Nevertheless, all five are individually abundant in many other species of plant (Harborne and Baxter, 1993) and are known to stimulate taste cells and/or elicit feeding (or egg-laying) in particular insect species (Ishikawa, 1963; Hewitt et al., 1969; Dethier, 1973; Schoonhoven, 1974; Papaj et al., 1992; Roessingh et al., 1999). We did not test glucose because this was done in experiment 1.

Using the neural recording technique described in experiment 1, we determined the response of all taste sensilla containing an inositol-sensitive taste cell to the taste stimuli listed above. We tested all taste stimuli at 10 mmol l⁻¹ because this concentration produces maximal neural responses to amino acids and sugar alcohols in all other insect species studied (e.g. Jakinovich and Agranoff, 1971; van Loon and van Eeuwijk, 1989; Roessingh et al., 1999). All taste stimuli were dissolved in an electrolyte solution (100 mmol l⁻¹ KCl). As a control, we also stimulated each taste sensillum with the electrolyte solution alone.

To determine whether the two styloconic sensilla responded to the taste stimuli, we compared neural responses to each taste stimulus in electrolyte solution with those to the electrolyte solution alone. We tallied the total number of spikes that occurred between 10 and 1010 ms after contacting the taste sensillum and compared this total with that elicited by the electrolyte solution alone. Because of the small sample size (N=4 caterpillars), we did not run statistical analyses.

In this first behavioral experiment (experiment 3), we examined the long-term feeding responses of M. sexta to myo-inositol by addressing the following questions. Does the behavioral response to myo-inositol increase with increasing concentrations of myo-inositol? If so, does the dynamic range of the concentration/response curve from the behavioral experiments correspond with that from the inositol-sensitive taste cells? A positive answer to both these questions would provide support for a direct role of sensory input from the inositol-sensitive taste cell in the behavioral response to myo-inositol.

We also asked whether the behavioral responses of fifth-instar caterpillars to myo-inositol were qualitatively similar to those from third-instar caterpillars. All previous studies of how M. sexta responds to myo-inositol have used fifth-instar caterpillars exclusively and thus could not determine whether the results obtained were specific to the final instar.

The feeding assay involved two steps. First, we placed a caterpillar (24–48 h after molting) in the ‘food-deprivation arena’, which consisted of an inverted clear plastic cup (160 ml volume) attached to a flat plastic surface. The cup reduced evaporative water loss from the glass-fiber disk used to deliver the test substance by maintaining an elevated relative humidity. Caterpillar in this arena were not fed for 30–45 min to standardize their ‘hunger’ state. Because the intermeal interval for these caterpillars usually ranges between 15 and 30 min (Reynolds et al., 1986), it is unlikely that this food-deprivation period created an extreme state of hunger. Next, we transferred the caterpillar to the ‘test arena’, which was identical to the food-deprivation arena in all respects except that a piece of cork (1 cm in diameter, 3–4 mm high) had been taped to the middle of the inverted Petri dish. Immediately before each test session, we pinned a glass-fiber disk (Whatman GF/A, 4.25 cm diameter) to the piece of cork, and then moistened it with 400 μl of distilled water containing one of the following concentrations of myo-inositol: 0, 0.5, 1, 5, 10 or 100 mmol l⁻¹. A caterpillar was then placed in the test arena and permitted to feed ad libitum for the duration of the test session (180 min for third-instar and 30 min for fifth-instar caterpillars). The caterpillar was then removed, and the area of glass-fiber disk eaten was quantified. For approximately 90 % of the tests, we calculated the disk area eaten by digitizing the two-dimensional area removed from the disk and then calculating the size of this ‘meal’ (in mm²) using SigmaScan (SPSS, Chicago, IL, USA). For the remaining 10 % of the tests, we calculated disk area eaten using visual estimates.

We ran two types of analysis to determine whether the caterpillars responded differently to the control (i.e. water-treated) and myo-inositol-treated disks. First, we asked whether more caterpillars initiated feeding (i.e. ate ?:5 % of the glass-fiber disk area) on the control versus myo-inositol-treated disks. We used the G-test to compare the percentage of caterpillars initiating feeding on each type of disk, separately for each myo-inositol concentration. In this and all subsequent tests, we controlled for the use of multiple pairwise comparisons on the same data set by setting the alpha level at 0.01. Second, among those caterpillars that initiated feeding, we asked whether the caterpillar ingested more diet from the control or myo-inositol-treated disks. To this end, we used the Mann–Whitney U-test to compare total consumption from both types of disk, separately for each myo-inositol concentration (α=0.01).

In this experiment (experiment 4), we asked whether myo-inositol alters the feeding responses of caterpillars to three types of plant chemical: sugars (glucose and sucrose), amino acids (alanine and proline) and feeding inhibitors (caffeine and salicin). Sugars were tested because a previous feeding study reported that myo-inositol synergizes the phagostimulatory effects of sugars in Bombyx mori (Ishikawa and Hirao, 1966). Amino acids were tested because a previous study indicated that, when offered a choice between disks treated with sucrose alone versus sucrose plus alanine (or proline), M. sexta fed disproportionately on the disks treated with sucrose alone (Städler and Hanson, 1978). Finally, caffeine and salicin (both of which inhibit feeding in M. sexta; Glendinning et al., 1999b) were tested because M. sexta is reported to ingest greater quantities of unpalatable plant tissues when inositol is added to them (Schoonhoven, 1969b).

We employed a commonly used two-choice protocol (Städler and Hanson, 1978). A covered Petri dish (150 mm in diameter) served as the testing arena. Within each test arena, we placed three control and three test glass-fiber disks (Whatman GF/A, 15 mm diameter) in a circular pattern, 25 mm from the edge of Petri dish; the control and test disks were arranged in an alternating pattern. We attached each disk to the end of a pin that protruded from the bottom of the test arena. To prevent the disk from touching the bottom of the test arena, we slid a piece of plastic tubing (5 mm high, 3 mm in diameter) over the shaft of the pin, and then placed the disk on top of the pin. The disks were moistened completely with 75 μl of the appropriate control or test solution (see below for details), and a caterpillar was then placed in the center of the test arena and permitted to feed for 180 min. At the end of the test session, the caterpillar was removed, and the area of each glass-fiber disk it had eaten was determined using the digitization procedure described above.

In all cases, the control disks contained one of the primary taste stimuli (see below), and the test disks contained the same primary taste stimulus plus 10 mmol l⁻¹myo-inositol. The primary taste stimuli were (in mmol l⁻¹): distilled water, sucrose (100), alanine (10), proline (10), salicin (10) and caffeine (10). Because of the relatively long duration of the test sessions, water evaporation undoubtedly increased the concentration of the taste stimuli over the course of each test session. Thus, the concentration of myo-inositol presented in the Results section (for this and the next experiment) indicates the initial concentration.

We ran two types of analysis to determine whether the caterpillars treated the control and test disks differently. First, we asked whether one type of disk was more effective at inciting feeding. To this end, we calculated the percentage of caterpillars that fed at all on each type of disk (e.g. the percentage of caterpillars that ate ?:5 % of the total glass-fiber disk area from each type of treated disk; α=0.05). Then, we used the G-test to compare the percentage of caterpillars that fed on the control versus test disks. Second, among the caterpillars that initiated feeding, we asked whether one type of disk stimulated more sustained feeding. To this end, we used the Mann–Whitney U-test to compare total consumption from both types of disk (α=0.05). (Note that we excluded caterpillars that failed to feed from this latter analysis.)

Because the previous two behavioral experiments were conducted over a period of at least 30 min, we could not determine the relative contribution of pre-versus post-ingestive input to the amount of each diet consumed. In this experiment (experiment 5), therefore, we used a brief-access biting assay to increase the likelihood that the ingestive responses were mediated exclusively by orosensory input.

Our brief-access biting assay involved seven steps. (i) We placed a caterpillar (day 2, fifth-instar) 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. (ii) We then transferred the caterpillar to the ‘test arena’, which was identical to the food-deprivation arena in all respects except that a piece of cork (1 cm in diameter, 3–4 mm high) had been taped to the middle of the inverted Petri dish. Immediately before each test session, we pinned a glass-fiber disk (Whatman GF/A, 4.25 cm diameter) to the piece of cork, and then moistened it with 400 μl of control or test solution (see below for details). (iii) Next, we placed the caterpillar on the edge of the glass-fiber disk, positioning it so that its legs and prolegs grasped the edge of the glass-fiber disk securely. (iv) Once the caterpillar had brought its mouthparts into contact with the glass-fiber disk and appeared to be actively tasting it (i.e. drumming the surface of the glass-fiber disk with its chemosensilla), we began the 10 min trial. To be included in the experiment, the caterpillar had to initiate feeding on the disk (defined as taking a single bite) within 8 min of the beginning of the trial. (v) Once feeding had commenced, we recorded the timing of each bite with a software-based event recorder over a 2 min period. Because the test arena was positioned on a turntable-like device, we could rotate the caterpillar slowly and keep its mandibles clearly visible as it fed; such rotation did not disrupt the feeding of the caterpillar. (vi) At the end of the 2 min feeding period, we removed the caterpillar from the glass-fiber disk (taking care to prevent the caterpillar from tearing the edge of the disk) and transferred it to a 160 ml plastic cup for 30 min. While in this cup, the caterpillar had ad libitum access to its wheat-germ-based rearing diet. (vii) Next, we transferred the caterpillar to the food-deprivation arena for 30 min, and then ran it through steps i–vi again, but with a different taste stimulus.

To ensure that each caterpillar would eat during our biting assay, we initially ran them through the biting assay with the control taste stimulus (i.e. distilled water). If a caterpillar failed (i) to initiate feeding within 8 min of contacting the control disk with its mouthparts, and (ii) to take at least 50 bites over the 2 min biting assay, it was excluded from further testing; this amounted to 6 % of the caterpillars.

Caterpillars that passed the two screening criteria were subjected to five additional brief-access biting assays with different taste stimuli. We used three single-component taste stimuli (in mmol l⁻¹: myo-inositol, 10; glucose, 100; caffeine, 10) and two two-component taste stimuli (myo-inositol plus caffeine and glucose plus caffeine). We used a within-animal design to evaluate the response to all five taste stimuli; that is, we ran each caterpillar through steps i–vii of the biting assay five times. To control for order effects, we presented these five taste stimuli in randomized sequences, and to control for any observer bias, we kept him/her blind with respect to the identity of the taste stimulus.

We analyzed five aspects of caterpillar behavior during the 10 min trial. We determined (i) the latency to initiate feeding (i.e. the time that elapsed between initially tasting the surface of the glass-fiber disk and initiating feeding), (ii) the total amount of disk area eaten during the 2 min biting assay, using the digitization procedure described above, (iii) the total number of bites taken over the 2 min biting assay, (iv) bite size by dividing the total area of disk eaten by the total number of bites (mm² per bite) and (v) biting rate by partitioning the entire feeding test into 12 sequential 10 s bins and then tabulating the total number of bites exhibited during each bin.

To ascertain whether either taste stimulus altered the latency to initiate feeding, total intake or bite size, we made pairwise comparisons (separately for each response variable) between the response to the control disk and that to the taste stimulus. We made these comparisons using the Wilcoxon matched-pairs signed-rank tests (two-tailed; α=0.01).

To determine the latency to altered biting activity, we determined the earliest 10 s bin (during the 2 min brief-access biting assay) in which caterpillars exhibited significantly fewer (or more) bites on the test disk than on the control disk. To make these comparisons, we used the Wilcoxon matched-pairs signed-rank test (two-tailed; α=0.01).

In this final experiment (experiment 6), we compared the response of taste cells within the lateral and epipharyngeal sensilla to single taste stimuli (in mmol l⁻¹: glucose, 100; myo-inositol, 10; caffeine, 10) versus binary mixtures of the same stimuli (glucose plus caffeine or myo-inositol plus caffeine). We focused on the lateral and epipharyngeal sensilla because they alone contain a bitter-sensitive taste cell that is sufficient to mediate the inhibited feeding response to caffeine (Glendinning et al., 1999b). Our goals were to determine (i) whether the responses of each taste cell to the binary mixture differed from those to the individual components and if so (ii) whether these mixture effects could explain the behavioral results observed in experiments 4 and 5.

We used the technique described in experiment 1 to record from taste cells in the lateral sensilla and techniques described in Glendinning et al. (1999b) to record from taste cells in the epipharyngeal sensilla. We stimulated a single sensillum with all five taste stimuli (in mmol l⁻¹) in the following order: glucose (100), myo-inositol (10), caffeine (10), glucose (100) plus caffeine (10), and myo-inositol (10) plus caffeine (10). To minimize any carry-over effects, we paused for at least 3 min between successive stimulations.

To monitor responses of specific taste cells, we exploited the fact that each taste cell responds to its respective best stimulus with a characteristic temporal pattern of firing: that from the salt-sensitive taste cell is temporally irregular, that from the inositol-sensitive taste cell is strongly phasic–tonic; that from the sugar-sensitive taste cell is less strongly phasic–tonic; and that from the bitter-sensitive taste cell is predominantly tonic, with a variable latency of onset (Peterson et al., 1993; Glendinning and Hills, 1997). These distinctive temporal patterns of firing enabled us to discriminate reliably between spikes from different taste cells. Recordings from the binary mixtures were often complex because several different taste cells were firing in and out of phase with one another. This was particularly evident in the traces involving binary mixtures, which contained many examples of spike superpositions, in which spikes occurring nearly simultaneously appear as a single waveform with increased amplitude and duration; there were also several examples of partially overlapping spikes. However, we employed standard techniques for addressing these analysis challenges (see, for example, Schnuch and Hansen, 1990; White et al., 1990).

Because our primary goal was to study potential interactions between taste cells that are firing simultaneously, it was essential that we used sensilla that exhibited relatively strong responses to each single component taste stimulus. To this end, we established the following screening criteria for the lateral sensilla: the bitter-sensitive taste cells had to produce at least 60 impulses s⁻¹ in response to the caffeine solution; the inositol-sensitive taste cell had to produce at least 50 impulses s⁻¹ in response to the inositol solution; and the sugar-sensitive taste cell had to produce at least 18 impulses s⁻¹ in response to the glucose stimulus. On the basis of these screening criteria, we rejected only one of the 16 caterpillars (fifth-instar, day 2) tested.

Because previous work had indicated that the epipharyngeal sensilla do not contain taste cells that respond to glucose (de Boer et al., 1977) or myo-inositol (J. Glendinning and K. Asaoka, unpublished data), we could not use the same screening criteria for the epipharyngeal sensilla recordings as we used for the medial and lateral sensilla recordings. As a result, we decided a priori that the bitter-sensitive taste cell had to produce at least 60 impulses s⁻¹ in response to the caffeine solution to be included. Four out of the five sensilla that we tested met this screening criterion.

We found that the medial and lateral styloconic sensilla each contained one taste cell that was excited by myo-inositol and a second taste cell that was excited by glucose (for representative traces, see Fig. 1A). In contrast, neither the epipharyngeal nor maxillary palp sensilla contained taste cells excited by myo-inositol or glucose, and the neural responses of these sensilla to myo-inositol, glucose or electrolyte solution alone were indistinguishable (see Fig. 1A). These results indicate that only the medial and lateral styloconic sensilla contain taste cells that are excited by myo-inositol or glucose.

We subsequently obtained concentration/response functions for the inositol- and sugar-sensitive taste cells present in the lateral and medial styloconic sensilla. These concentration/ response functions revealed that the inositol-sensitive taste cell in both sensilla responded vigorously to myo-inositol (median firing rate >45 impulses s⁻¹) at concentrations as low as 1 mmol l⁻¹ (Fig. 1B). It is worth noting, however, that the maximal response of the inositol-sensitive taste cell in the medial sensillum was almost twice as great as that in the lateral sensillum. In contrast, the sugar-sensitive taste cell in both the medial and lateral sensilla exhibited a comparatively weak response to glucose (Fig. 1B); the concentration/response curve was relatively flat, with a median maximal firing rate of less than 40 impulses s⁻¹ even at high concentrations (i.e. >100 mmol l⁻¹).

The response of four lateral and four medial styloconic sensilla exposed to rutin, several sugar alcohols (mannitol, sorbitol, pinitol, quebrachitol and myo-inositol) and several amino acids (proline, asparagine and glutamine) is shown in Fig. 2. Apart from myo-inositol, these substances induced very low levels of response produced by several taste cells firing slowly and in a temporally inconsistent manner. This pattern of response was qualitatively and quantitatively similar to that elicited by the electrolyte solution alone. However, myo-inositol elicited clear, consistent and robust excitatory responses in a single taste cell (Fig. 2). These results, together with those in the previous experiment, indicate (i) that myo-inositol is one of the few nutrients that strongly stimulates the peripheral taste system of M. sexta, (ii) that the response to myo-inositol is mediated by a single, inositol-sensitive taste cell in each of the lateral and medial sensilla, and (iii) that the inositol-sensitive taste cells appear to respond selectively to myo-inositol, while the sugar-sensitive taste cells appear to respond selectively to sugars.

We found that myo-inositol significantly increased the chances that both third- and fifth-instar caterpillars would initiate feeding; this effect became more pronounced with increasing concentrations of myo-inositol (Fig. 3A,B). Once the caterpillars had initiated feeding, however, there was no clear effect of myo-inositol on the total amount of disk eaten; the only exception was that the fifth-instar caterpillars ate slightly, but significantly, more of the disks treated with 100 mmol l⁻¹myo-inositol (Fig. 3C,D). We should note, however, that plants do not appear to contain inositols at such high concentrations. Thus, these results indicate that myo-inositol is a potent feeding incitant, but an ineffective feeding stimulant for sustained feeding at ecologically relevant concentrations (i.e. between 0.5 and 10 mmol l⁻¹). Moreover, the fact that the dynamic range for behavioral responsiveness to myo-inositol (i.e. feeding incitation) overlaps with that for gustatory responsiveness to myo-inositol (see Fig. 1) provides support for a causal connection between sensory input from inositol-sensitive taste cells and feeding incitation.

There were some small differences in how the third- and fifth-instar caterpillars responded to myo-inositol (Fig. 3). First, a greater percentage of the fifth-instar caterpillars initiated feeding on the 100 mmol l⁻¹myo-inositol disks than did the third-instar caterpillars. Second, the fifth-(but not third-) instar caterpillars ate significantly more of the 100 mmol l⁻¹myo-inositol disks than the control disks. Third, when one focuses on the percentage of caterpillars that initiated feeding, it is apparent that each instar responded differently to increasing concentrations of myo-inositol. The concentration/ response curve for third-instar caterpillars reached a plateau at 10 mmol l⁻¹, whereas that for the fifth-instar caterpillars continued to rise at higher concentrations. The ecological significance of these differences across instars is unclear, however, given that myo-inositol does not appear to occur in plant tissues at concentrations greater than 10 mmol l⁻¹ (Nelson and Bernays, 1998).

When water was the primary (i.e. control) taste stimulus, the caterpillars initiated feeding significantly more often on the test inositol disks (Fig. 4A). However, once the caterpillars had initiated feeding, they ingested statistically indistinguishable amounts of disk material from the control and test disks (Fig. 4D). Thus, in this test, myo-inositol functioned exclusively as a feeding incitant.

When glucose, sucrose, alanine or proline was the primary taste stimulus, the caterpillars initiated feeding significantly more often on the test disks containing myo-inositol and nutrient (Fig. 4B). However, a comparison of the results in Fig. 4A and Fig. 4B reveals that the percentage of caterpillars that initiated feeding on the control and test disks was quite similar in both tests. This indicates that the caterpillars initiated feeding more often on the test disks simply because they contained myo-inositol; that is to say, the nutrients themselves had little effect. Once the caterpillars had initiated feeding, they ingested statistically equivalent amounts of disk material from the control and test disks (Fig. 4E). Thus, myo-inositol did not synergize with any of the other nutrients (i.e. cause an unexpectedly strong feeding response). Parenthetically, even though glucose alone was not a strong feeding initiator, it nevertheless had a pronounced phagostimulatory effect on the small percentage of caterpillars that initiated feeding on the glucose-treated disks.

When either caffeine or salicin, which normally have a deterrent effect on feeding, was the primary taste stimulus, the caterpillars initiated feeding significantly more often on the test disks (Fig. 4C). In addition, once the caterpillars had initiated feeding, they ingested significantly more disk material from the test disks (Fig. 4F). These results indicate that myo-inositol counteracted the inhibitory effects of caffeine and salicin on feeding.

Despite extensive individual variation, the latency to initiate feeding was significantly shorter on the disks treated with myo-inositol alone or with myo-inositol plus caffeine than it was on the control disk containing water alone (Fig. 5A). In contrast, the latency to initiate feeding on the disks treated with glucose alone, caffeine alone or caffeine plus glucose was statistically equivalent to that on the control disks. These results are consistent with the idea that myo-inositol causes M. sexta to initiate feeding.

Over the course of the 2 min biting assay, the caterpillars ingested approximately 80 mm² of control disk and a statistically equivalent amount from disks treated with myo-inositol alone, glucose alone or myo-inositol plus caffeine (Fig. 5B). However, they ingested significantly less of the disks treated with caffeine alone or with caffeine plus glucose. They ate less of these latter disks by taking significantly smaller bites (Fig. 5C) and by biting significantly more slowly during the first and all subsequent 10 s time intervals (Fig. 6A,B). These results demonstrate that, once feeding had been initiated, neither myo-inositol alone nor glucose alone modulated the feeding behavior of the caterpillars (e.g. increasing the rate of biting, the size of bites or the total intake). The only way that either compound modulated feeding behavior was when it was in binary mixture with an aversive taste stimulus (e.g. caffeine). Under such conditions, myo-inositol totally counteracted the inhibitory effect of caffeine on total intake, bite size and biting rate. In contrast, glucose partially counteracted the inhibitory effect of caffeine on biting rate, but had no effect on total intake or bite size.

We obtained no evidence for peripheral mixture interactions among taste cells in the lateral sensillum. Instead, we found (i) that the bitter-sensitive taste cells fired at the same rate, irrespective of whether we stimulated it with caffeine alone or with a binary mixture of caffeine plus myo-inositol or caffeine plus glucose (Fig. 7A,B); (ii) that the inositol-sensitive taste cell fired at the same rate, irrespective of whether we stimulated it with myo-inositol alone or with a binary mixture of myo-inositol plus caffeine (Fig. 7A,C); and (iii) that the sugar-sensitive taste cell fired at the same rate, irrespective of whether we stimulated it with glucose alone or with a binary mixture of glucose plus caffeine (Fig. 7A,D).

In contrast, we obtained clear evidence for peripheral mixture interactions in the epipharyngeal sensilla, which were localized to the bitter-sensitive taste cell (see Glendinning et al., 1999a,b). We found that the responsiveness of this taste cell to caffeine was strongly inhibited by glucose, but not myo-inositol, in all the epipharyngeal sensilla that we tested (e.g. Fig. 8).

We found that Manduca sexta has only two pairs of inositol-sensitive taste cells: one in the lateral and one in the medial styloconic sensilla. Both pairs of inositol-sensitive taste cell exhibited vigorous excitatory responses to low millimolar concentrations of myo-inositol. In contrast, the sugar-sensitive taste cells exhibited a relatively weak excitatory response to glucose, even at concentrations as high as 250 mmol l⁻¹. The fact that these sugar- and inositol-sensitive taste cells failed to respond to a variety of amino acids, to rutin and to other sugar alcohols suggests that they are narrowly tuned to a small subset of the available nutrients in plants (i.e. sugars or inositol). Our feeding studies revealed that myo-inositol (i) causes the initiation but not the maintenance of feeding in third- and fifth-instar caterpillars, and (ii) counteracts the inhibitory effect of 10 mmol l⁻¹ caffeine on feeding. Sugars, in contrast, failed to incite feeding, to promote increased consumption or to counteract the inhibitory effect of caffeine on feeding.

Several lines of evidence indicate that it was sensory input from the inositol-sensitive taste cells (and not, for instance, post-ingestive feedback) that caused the observed feeding responses to myo-inositol. Two of the behavioral effects of myo-inositol were apparent almost instantaneously. First, the caterpillars initiated feeding almost immediately (i.e. in less than 10 s) when presented with disks containing myo-inositol, but took significantly longer to do so (approximately 25 s) when presented with control disks containing water alone (see Fig. 5A). Second, the caterpillars exhibited a significantly reduced biting rate on the disks containing caffeine alone during the first and all subsequent 10 s time bins, but did not exhibit any evidence of a reduced biting rate on the disks containing caffeine plus myo-inositol (see Fig. 6). It is unlikely that myo-inositol could have acted through a post-ingestive response mechanism in such a short period. Finally, the dynamic range for behavioral responsiveness to myo-inositol overlapped with that for neural responsiveness of the inositol-sensitive taste cells to myo-inositol.

Because the ability of myo-inositol to counteract the inhibited feeding response was not associated with any peripheral interactions among taste cells, it is likely that this effect was mediated within the central nervous system. One likely site for these inhibitory interactions would be the neuropil of the subesophageal ganglion. This neuropil is the primary projection site for most taste cell axons (Kent and Hildebrand, 1987) and also contains the cell bodies for the motor neurons that insert onto the mandibular muscles (Griss et al., 1991; Rohrbacher, 1994).

In contrast to myo-inositol, glucose was a relatively ineffective feeding modulator. In the one instance in which glucose modulated feeding, we could not determine the relative contribution of pre-versus post-ingestive response mechanisms: the caterpillars tended to bite more rapidly on the disks containing both glucose and caffeine than they did on the disks containing caffeine alone, but this did not occur until 40–60 s after initiating feeding (Fig. 6). If this effect was mediated exclusively by a pre-ingestive mechanism (e.g. the inhibiting effect of glucose on the response of the bitter-sensitive taste cell in the epipharyngeal sensillum), why did it not manifest itself sooner? A study by Timmins and Reynolds (1992) established that post-ingestive feedback plays an important role in regulating meal size in M. sexta caterpillars.

Our results indicate that myo-inositol was substantially more effective than glucose at modulating feeding in M. sexta. We have developed two hypothetical models to explain this unexpected finding. The first model proposes that the inositol- and sugar-sensitive taste cells constitute functionally distinct ‘labeled lines’, and that the central nervous system discriminates between sensory input from each of these labeled lines. Accordingly, as sensory input from the inositol-sensitive taste increases, there would be a proportionally stronger activation of the feeding control mechanisms that incite feeding and counteract the inhibited feeding response. A critical assumption of this model is that sensory input from the sugar-sensitive taste cells would be incapable of producing these behavioral effects, even if they fired at rates comparable with those observed in the inositol-sensitive taste cells.

The second model proposes that the central nervous system of M. sexta does not actually discriminate between sensory input from the inositol- and sugar-sensitive taste cells. Instead, it treats them as functionally equivalent because they both indicate the relative concentration of carbohydrates in food. Accordingly, any taste stimulus that strongly activates one or both classes of taste cell would produce the behavioral effects described above. Because myo-inositol was the only compound that strongly activated either of these nutrient-sensitive taste cells, it alone produced the behavioral effects.

One could test these two models by using compounds that stimulate the sugar-sensitive taste cell more vigorously than does glucose (e.g. 5-thio-D-glucose; Lam and Frazier, 1991). According to the second model, compounds such as 5-thio-D-glucose should produce behavioral effects similar to those elicited by myo-inositol, whereas according to the first model, compounds such as 5-thio-D-glucose should produce behavioral effects similar to those elicited by glucose.

The finding that a plant compound incites feeding, but fails to promote sustained feeding, is not novel. Two other plant compounds, flavonoids and sitosterol, have been found to produce a similar effect in Bombyx mori (Hamamura et al., 1962; Hamamura, 1970). However, the finding that myo-inositol incites feeding is novel. Previous studies with B. mori have reported that inositol promotes sustained feeding, but plays little role in initiating feeding (Hamamura et al., 1962; Hamamura, 1970). These findings, together with our own, reveal two important insights into the feeding control mechanisms of caterpillars: (i) that there are functionally independent neural mechanisms controlling the initiation of feeding and the promotion of sustained feeding, and (ii) that sensory input from the inositol-sensitive taste cell does not necessarily activate the same neural mechanism in all caterpillar species.

Even though our results offer limited insight into the central mechanisms underlying the initiation of feeding, they nevertheless provide evidence against two potential mechanisms. For example, it is conceivable that sensory input from the inositol-sensitive taste cells provides some kind of arousal or excitatory effect, which causes the caterpillars to initiate feeding (Dethier et al., 1965). However, given that elevated central excitation immediately prior to a meal appears both to incite feeding and to promote sustained feeding in other species of insect (Simpson, 1995), it seems that this mechanism alone could not explain the observed effects of myo-inositol in M. sexta. Alternatively, it is conceivable that the inositol-sensitive taste cells could become progressively more adapted to myo-inositol over the course of the meal, resulting in ever-diminishing phagostimulatory input. However, this explanation is unlikely because inositol-sensitive taste cells in M. sexta (particularly the one in the medial sensilla) are remarkably resistant to adaptation (see, for example, Fig. 1; Bernays et al., 1998).

At face value, the finding that myo-inositol failed to promote sustained feeding appears to contradict previous feeding studies with inositol in M. sexta (see, for example, Yamamoto and Fraenkel, 1960; Städler and Hanson, 1978). In these studies, the authors offered caterpillars a choice between equal numbers of water- and inositol-treated disks over several hours, and then determined the total consumption from each type of disk. They found that caterpillars ingested significantly more total disk area from the inositol-treated disks, indicating that inositol was a feeding stimulant. If we had used a similar calculation in Fig. 4D, we too would have found that the caterpillars ate significantly more of the inositol-treated disks than of the control disks (median 11 % versus 0 %, respectively; P<0.05; Wilcoxon matched-pairs signed-rank test). Thus, our results do not contradict these earlier studies. Nevertheless, we believe that our more thorough data analysis procedure (i.e. breaking down the results of long-term feeding assays into two response variables, the percentage of caterpillars that initiated feeding and the total consumption once feeding had been initiated) provides deeper insight into the nature of the feeding response.

From an ecological standpoint, it is unclear why inositol is such an effective feeding incitant for M. sexta since the concentration of inositol on the surface of at least two of its preferred host plants (tomato and tobacco plants) does not correlate with the concentration of sugars, inositol or proteins within the leaf tissues (Nelson and Bernays, 1998). Thus, inciting feeding on leaves with high surface concentrations of inositol would be an ineffective behavioral mechanism for identifying the most nutritious leaves. It may be, however, that myo-inositol plays a different role for plant-feeding insects such as M. sexta. For instance, myo-inositol could serve as a generalized sign for plant tissues and thereby help insects to discriminate between plant and nonplant tissues. In addition, the fact that myo-inositol initiated feeding so effectively indicates that it elevated the central excitatory state of the caterpillars. This arousal effect of myo-inositol may cause M. sexta to become more attentive to other important cues in plant materials (e.g. host-plant-specific chemicals).

We found that sensory input from the inositol-sensitive taste cell counteracts the inhibitory effect of caffeine (and perhaps salicin; Figs 4–6) on feeding. A similar result was obtained independently by Frank Hanson (unpublished data). The caterpillars not only ingested equal amounts of disk material treated with water or myo-inositol plus caffeine, but they also ingested both types of disk in a similar manner (i.e. used similar biting rates and bite sizes). Thus, it appears that, even though 10 mmol l⁻¹ caffeine stimulates the bitter-sensitive taste cells vigorously when it is in a binary mixture with myo-inositol, the caterpillars nevertheless fail to exhibit any apparent behavioral response to caffeine when it is in a binary mixture with myo-inositol.

There is evidence that carbohydrates can also counteract the inhibitory effects of aversive taste stimuli in several other species of caterpillar. For instance, meso-inositol and sucrose (either alone or in a binary mixture) counteract the inhibitory effects of low, but not of high, concentrations of sinigrin in two species of crucifer-feeding caterpillar, Mamestra configurata and Trichopulsia ni (Shields and Mitchell, 1995b,c). The authors attributed these concentration-dependent effects of carbohydrates to a gustatory mechanism. However, given that the authors used an end-point feeding assay that lasted 3–4 h, it is also possible that the high sinigrin concentrations inhibited feeding through a post-ingestive toxicity mechanism. In an analogous study with B. mori, Hirao and Arai (1991) reported that a mixture of sugars and inositol failed to counteract the inhibitory effects of salicin (another aversive taste stimulus) on long-term feeding responses.

As a caveat, we should note that the ability of carbohydrates to counteract the inhibitory effects of aversive taste stimuli on feeding may diminish as the complexity and/or concentration of the carbohydrate mixture increases. For example, one study of M. sexta reported that a binary mixture of 300 mmol l⁻¹ glucose and 100 mmol l⁻¹myo-inositol failed to counteract the inhibitory effects of caffeine during a short-term biting assay (Frazier, 1986). However, given that these concentrations of glucose and inositol are substantially higher than those found in the host plants of M. sexta (Nelson and Bernays, 1998), the ecological relevance of this latter result is unclear.

Prolonged ingestion of diets containing secondary compounds such as caffeine could be toxic to M. sexta (see, for example, Nathanson, 1984), so one might question the adaptive significance of a central mechanism that counteracts the inhibitory effects of such compounds on feeding. One resolution to this apparent paradox is that, even though caffeine may be toxic to M. sexta, there are many other relatively harmless plant compounds that also stimulate its bitter-sensitive taste cells and inhibit feeding (e.g. salicin and grindelanes; Glendinning et al., 1998, 1999b). If M. sexta encountered these latter compounds in a highly nutritious plant, it would clearly benefit from a central mechanism that enables it to ingest this plant readily. In those cases where the presence of inositol causes M. sexta to ingest toxic plant compounds unwittingly, it would still have at its disposal several post-ingestive response mechanisms for inhibiting feeding (see, for example, Glendinning, 1996). Thus, inositol may enable M. sexta to feed on host plants containing relatively harmless compounds that inhibit feeding.

Even though glucose partially counteracted the inhibitory effects of caffeine on feeding, the magnitude of this effect was small compared with that produced by myo-inositol. The observation that sugars alone do not modulate biting rate in M. sexta has been reported elsewhere (Bowdan, 1995). On the basis of studies with another species of caterpillar, Pieris brassicae, we expected the sugars to have a greater impact on feeding. In this latter species, sugars dramatically increased biting rate in a concentration-dependent manner (Ma, 1972). These interspecific differences may reflect the fact that the sugar-sensitive taste cells in M. sexta respond to sugars much less vigorously than do those in P. brassicae.

Myo-inositol, in contrast, had several robust effects on feeding in M. sexta. The presence of myo-inositol significantly increased the probability of initiating feeding on both palatable and unpalatable diets. However, once feeding had been initiated on a palatable diet, the presence of myo-inositol alone did not appear to influence biting rate or bite size. The situation was quite different with unpalatable diets. Once feeding had been initiated on this latter type of diet, the presence of myo-inositol completely counteracted the inhibitory effects of aversive taste stimuli. Further work is needed to identify the central mechanisms that mediate these effects.

We acknowledge the help of Betty Estesen, Stacy Brenner, Marci Tarre and Christian Roche in carrying out some of the experiments. We thank Bruce O’Gara for editorial comments and Richard Jensen (University of Arizona) for providing the pinitol sample. The work was funded in part by a Plant–Insect Training Grant to the University of Arizona from NSF/USDA/DOE (to E.A.B.) and in part by research grant number 5 R29 DC 02416 from the National Institute on Deafness and Other Communication Disorders, National Institutes of Health (to J.I.G.).