Gustatory synergism in ants mediates a species-specific symbiosis with lycaenid butterflies
- First Online:
- Cite this article as:
- Hojo, M.K., Wada-Katsumata, A., Ozaki, M. et al. J Comp Physiol A (2008) 194: 1043. doi:10.1007/s00359-008-0375-6
Here we show that larvae of the lycaenid butterfly Niphanda fusca secrete droplets containing trehalose and glycine. These droplets attract the larva’s host ants Camponotus japonicus, which collect and protect the larvae. We comparatively investigated gustatory preference for trehalose, glycine or a mixture of the two between host (C. japonicus) and non-host (Camponotus obscuripes) species of ants in behavioral and electrophysiological experiments. Glycine itself induced no taste sensation in either host or non-host ants. The mixture of trehalose plus glycine was chosen as much as pure trehalose by non-host ants. However, the host ants clearly preferred the mixture of trehalose plus glycine to trehalose alone. When we used sucrose instead of trehalose, the mixture of sucrose plus glycine was chosen as much as sucrose alone, in both species. These behavioral data are supported by the electrophysiological responsiveness to sugars and/or glycine in the sugar-taste receptor cells of the ants. Considering that lycaenid butterflies’ secretions have species-specific compositions of sugar and amino acid; our results clearly showed that such species-specific compositions of larval secretions are precisely tuned to the feeding preferences of their host ant species, in which the feeding preferences are synergistically enhanced by amino acid.
Dorsal nectary organ
The majority of lycaenid butterflies have interactions with ants (Pierce 1987). These interactions are divided into facultative and obligate interactions (Pierce et al. 2002). In facultative interactions, lycaenid larvae intermittently interact with numerous ant species (Jordano and Thomas 1992; Wagner 1993; Fraser et al. 2001). By contrast, obligate interactions are highly specific. The larvae of obligate species constantly interact with only a single species or genus of ant and cannot survive without certain ant species (Pierce et al. 2002). Accordingly, in obligate interactions, lycaenids may have evolved signals to select a certain ant species (Fiedler et al. 1996).
Lycaenid larvae secrete nutritious droplets for ants from specialized exocrine gland called the dorsal nectary organ (DNO). The DNO secretions, which attract ant partners (Agrawal and Fordyce 2000), play a critical role in the maintenance of lycaenid–ant interactions (Leimar and Axén 1993; Axén et al. 1996). The secretions from the DNO appear to be mainly composed of sugars and amino acids (Maschwitz et al. 1975; Cushman et al. 1994; Pierce and Nash 1999; Daniels et al. 2005). The concentration of amino acids has shown to be relatively higher than those of aphid honeydew (Yao and Akimoto 2002), or floral and extrafloral nectars (Blüthgen et al. 2004), and these nitrogen-rich secretions may be important for lycaenid–ant interactions (Pierce 1985; Daniels et al. 2005).
Most ants prefer mixed solutions containing sugar and amino acid to sugar alone (Lanza 1991; Wada et al. 2001; Blüthgen and Fiedler 2004b). Different ant species show different feeding preferences to the mixtures (Blüthgen and Fiedler 2004b; Kay 2004). Thus, it is expected that lycaenid larvae secrete nutritious droplets adapted to the feeding preferences of their specific ant partners (Fiedler et al. 1996).
The lycaenid butterfly Niphanda fusca has a parasitic interaction with ants Camponotus japonicus (Nagayama 1950). After hatching, larvae of N. fusca initially feed on the honeydew of aphids, which are tended by C. japonicus workers. The butterfly larvae start to produce secretions from the DNO at the third-instar stage and are then adopted by C. japonicus workers. In the ant nest, the larvae are fed by the workers via trophallaxis until they have pupated. N. fusca cannot survive without its host ant, C. japonicus, so this species-specific interaction is a typical obligate interaction. The larvae can continually offer DNO secretions for ants until pupation. The amino acid content of the DNO secretions of N. fusca is dominated by one compound, glycine. Until the present paper, the dominant sugar in the secretion was thought to be glucose (Nomura et al. 1992). Thus, in a previous paper (Wada et al. 2001), the feeding preferences of C. japonicus workers were investigated using d-glucose and d-fructose but not trehalose and sucrose. Host ant workers preferred a mixture of d-glucose and glycine to pure d-glucose, and they preferred a mixture of d-fructose and glycine to pure d-fructose. In any case, glycine itself induced no taste sensation in C. japonicus (Wada et al. 2001).
We began our study by reanalyzing the DNO secretions of N. fusca, and found that the major sugar component was trehalose rather than glucose. Trehalose is a major sugar in insect hemolymph and, unlike most other butterfly larvae, N. fusca do not feed on plants but indirectly on insects. Based on these findings, we hypothesized that glycine is the key component in the species-specific symbiosis between N. fusca and C. japonicus, and that the synergistic effect of glycine would therefore be observed in the host ant, but not in a non-host ant species. In the present study, we compared the gustatory synergism between sugar and glycine for workers of the host ant, C. japonicus and a closely related non-host ant, Camponotus obscuripes, using both behavioral and electrophysiological experiments.
Materials and methods
Final instar larvae of Niphanda fusca were collected from four colonies of Camponotus japonicus with workers, males, alate females, ant larvae and eggs in the city of Fujinomiya, Shizuoka prefecture, Japan. In those colonies, the N. fusca larvae were reared with C. japonicus. For behavioral and electrophysiological experiments, we used two ant species, C. japonicus and Camponotus obscuripes, as a host and a non-host ant, respectively. C. obscuripes were chosen because the species is closely related to C. japonicus. We collected worker ants, ant larvae and eggs from several colonies of C. japonicus and C. obscuripes in Kyoto prefecture, Japan. These colonies had not been parasitized by N. fusca, so the workers of both species had probably never fed on the DNO secretions of N. fusca. The collected ants were reared at room temperature in a plastic box (35.0 × 25.0 × 6.0 cm) serving as a foraging arena, in which a wet plaster nest box (11.0 × 7.5 × 3.0 cm) had been placed. They were continuously given 10% sucrose solution and mealworms twice per week.
Chemical analysis of the DNO secretions
A final instar larva of N. fusca and a worker of C. japonicus were introduced to a petri dish (6 cm in diameter). When the worker tapped around the DNO, a droplet was secreted. These droplets were collected using microcapillaries (MICROCAPS; Drummond, Broomall, PA, USA). After 2–4 droplets had been collected, the larva was replaced by another. A total of 0.48–1.24 μl of DNO secretions was collected from 3 to 5 larvae. Using the four parasitized colonies, the collection was replicated seven times in total. Immediately after collection, all samples in the microcapillaries were stored at −30°C until chemical analysis.
For chemical analysis, samples were added to 15 μl of Milli-Q-Water. Ten and 5-μl aliquots of the mixture were used for sugar and amino acid analysis, respectively.
For sugar analysis, samples were analyzed by high performance liquid chromatography using a 5NH2-MS packed column (4.6 × 150 mm; Cosmosil, Nacalai Tesque, Kyoto, Japan) at room temperature. The mobile phase was 80% acetonitrile (Wako Chemicals, Tokyo, Japan) and the flow rate was 1 ml min −1. The samples were injected directly onto the column. Peak sizes for the sugars present in the samples were calculated directly by a refractive index detector (RID6A; Shimadzu, Kyoto, Japan) and used to calculate the concentrations of sugars in the DNO secretions. The sugars in the DNO secretions were identified by comparison of the retention times with those of standard sugar solutions (d-xylose, d-fructose, d-glucose, d-galactose, sucrose, turanose, maltose, trehalose, lactose, melibiose, melezitose, and raffinose; all from Nacalai Tesque, Kyoto, Japan).
For amino acid analysis, each sample was adjusted with 0.02 N HCl (Nacalai Tesque, Kyoto, Japan) to a final volume of 100 μl, and analyzed using an automated amino acid analyzer L-8800 (Hitachi, Tokyo, Japan).
Two-choice feeding experiments
To examine feeding preference, we conducted a two-choice preference test (Tanimura et al. 1982) modified by Wada et al. (2001). In order to check for phagostimulative effects of trehalose, sucrose or glycine, we tested feeding preferences between distilled water and trehalose, between distilled water and sucrose and between distilled water and glycine, respectively, in host and non-host ants. Thus, the pair of solutions prepared for each two-choice test was distilled water and 400 mmol l−1 trehalose, distilled water and 50 mmol l−1 sucrose, or distilled water and 50 mmol l−1 glycine. Feeding preferences between trehalose and trehalose mixed with glycine, and between sucrose and sucrose mixed with glycine were then investigated. For each of these two-choice test, the following pairs of solutions were prepared: 400 mmol l−1 trehalose and 400 mmol l−1 trehalose plus 50 mmol l−1 glycine; 50 mmol l−1 trehalose and 50 mmol l−1 trehalose plus 50 mmol l−1 glycine; and 50 mmol l−1 sucrose and 50 mmol l−1 sucrose plus 50 mmol l−1 glycine.
In each two-choice test, we used three 60-well (10 μl each) test plates (Nunc Mini Trays; Nun™, Roskilde, Denmark), plates 1–3. In each plate, the wells were alternately filled with two different kinds of solutions, A or B. In plate 1, solution A was stained blue (0.01% brilliant blue FCF, Wako Chemicals, Tokyo, Japan) and solution B was colorless. In plate 2, solution A was colorless and solution B was stained blue. In plate 3, both A and B solutions were colorless. We placed four workers that had been starved for 6 days and only provided water onto each of the plates 1–3 and let them drink freely for 3 h in the dark. The workers were then allowed to select the more preferable solution ad libitum. After that, workers were frozen at −30°C and their crops were dissected out. Immediately after dissection, the crops of the four workers from each plate were homogenized and extracted with 1 ml of 50% ethanol (Wako Chemicals, Tokyo, Japan). The absorbances of the extracts were measured at 630 nm to evaluate the intake of the blue-colored solutions. The absorbance of plate 3 was subtracted as a blank value from the absorbances of plate 1 or plate 2. Using corrective absorbances of plate 1 (Abs. 1) and plate 2 (Abs. 2), relative preference rates for solutions A and B were expressed as Abs.1/(Abs.1 + Abs.2) and Abs.2/(Abs.1 + Abs.2), respectively. Thus, the relative preference rates were all between 0 and 1. In the case of no difference in preference between solutions A and B, Abs.1/(Abs.1 + Abs.2) = Abs.2/(Abs.1 + Abs.2) = 0.5. If the workers preferred A to B, Abs.1/(Abs.1 + Abs.2) was higher than Abs.2/(Abs.1 + Abs.2) and vice versa. Each two-choice test with three plates, using four workers from the same colony per plate, was replicated 6 times. A total of 72 workers from 2 to 3 colonies were used. The relative preference rates were statistically compared using a Wilcoxon signed-rank test (SPSS version 11.0; SPSS Inc., Chicago, IL, USA).
For electrophysiological experiments, we used foragers. We have observed that foragers, as well as nurses, feed on the DNO secretions of N. fusca. Under an optical microscope, an isolated head of an ant connected to a platinum indifferent electrode was surrounded by a moist pad. The electrophysiological responses were then recorded from the tip of a long-type chemosensillum located on the third segment of the labial palps, according to the tip-recording method (Hodgson et al. 1955). There are several types of sensilla on the labial palps, classified on the basis of their lengths, but only this type of chemosensillum had previously been shown to be sensitive to sugars and was predicted to contain water, salt and sugar receptor cells by Wada et al. (2001). Several long-type sensilla are present in the segment, and two or three long-type sensilla on each labial palp were randomly selected and subjected to the electrophysiological recordings. The tip of the sensillum was stimulated by capping with a glass capillary recording electrode containing trehalose and/or glycine dissolved in 1 mmol l−1 KCl or sucrose and/or glycine dissolved in 1 mmol l−1 KCl, respectively. Each stimulation lasted for 5 s with an interstimulus interval of 4–5 min to avoid any adaptation. The recording electrode was connected to the taste probe and then to an amplifier (USB IDAC; Syntech, Hilversum, Netherlands). For each stimulation, impulses were recorded from five to eight sensilla in two to three different individual ants and the data were stored on a computer and analyzed using the software Auto Spike v. 4.0 (Syntech, Hilversum, Netherlands). The types of impulses were sorted on the basis of their amplitudes. In order to evaluate the effect of glycine on trehalose reception, the stimuli of 50 mmol l−1 trehalose, 50 mmol l−1 glycine and 50 mmol l−1 trehalose mixed with 50 mmol l−1 glycine were randomly given to the same chemosensilla and the impulses were recorded. This concentration of glycine (50 mmol l−1) was chosen because it is the mean concentration in the DNO secretions of N. fusca. The concentration of sugars (50 mmol l−1) fell within the natural range of the secretion, because impulse amplitude evoked by average sugar concentration (400 mmol l−1) was too small to count correctly. We confirmed that glycine also enhances feeding behavior toward 50 mmol l−1 trehalose (Fig. 2e). Using sucrose instead of trehalose, we also investigated the effect of glycine on sucrose reception in another set of workers. Then, the stimuli of 50 mmol l−1 sucrose, 50 mmol l−1 glycine and 50 mmol l−1 sucrose mixed with 50 mmol l−1 glycine were randomly applied to the same chemosensilla. In any experiments, the impulses from the sugar receptor cell generated for 1 s starting from the beginning of stimulation were counted and the numbers of impulses s−1 were statistically compared, using linear mixed models, with stimulus as a fixed factor and chemosensillum nested within forager as random factors (SPSS version 11.0; SPSS Inc., Chicago, IL, USA).
Scanning electron microscopy
The isolated heads of the worker ants were washed in 70% ethanol and dried in a desiccator. Specimens were coated with a gold layer and observed with a scanning electron microscope (VE-7800; Keyence, Osaka, Japan).
Chemical composition of the DNO secretion
Sugar composition of the DNO secretions of Niphanda fusca (N = 7)
Concentrationa (mmol l−1 ± SEM)
Number of samples in which detected
15.22 and 18.30
379.84 ± 59.32
390.73 ± 58.81
Amino acid composition of the DNO secretions of Niphanda fusca (N = 7)
Concentrationa (mmol l−1 ± SEM)
Number of samples in which detected
0.19 ± 0.10
7.63 ± 1.67
13.32 ± 2.45
0.83 ± 0.22
62.39 ± 13.28
0.79 ± 0.24
2.13 ± 0.42
0.63 and 0.12
1.88 ± 0.35
0.38 ± 0.15
1.54 ± 0.93
Total amino acids
91.25 ± 15.25
Effect of glycine on feeding preference to sugar
Taste response of sugar receptor cell in host and non-host ants
Effect of glycine on sugar receptor response in host ants
When stimulated with 50 mmol l−1 trehalose, the chemosensillum on the labial palps showed impulses from the sugar receptor cell. By contrast, 50 mmol l−1 glycine did not elicit any impulses from the sugar receptor cell (Fig. 6a, c). The mixture of 50 mmol l−1 trehalose and 50 mmol l−1 glycine elicited impulses from the sugar receptor cell and the impulse frequency was significantly larger than that elicited by 50 mmol l−1 trehalose alone (Fig. 6e, N = 6 sensilla from two foragers, F1,5 = 19.286, P < 0.01).
We also recorded responses to 50 mmol l−1 sucrose or 50 mmol l−1 sucrose mixed with 50 mmol l−1 glycine. Both stimulants elicited impulses from sugar receptor cells (Fig. 6c), but the impulse frequencies were almost the same regardless of the presence of glycine (Fig. 6f, N = 8 sensilla from three foragers, F1,7 = 1.880, P = 0.213).
Effect of glycine on sugar reception in non-host ants
In non-host ants as well as host ants, impulses from the sugar receptor cells were not observed when the chemosensilla were stimulated with 50 mmol l−1 glycine (Fig. 6b, d). The mixture of 50 mmol l−1 trehalose and 50 mmol l−1 glycine elicited impulses from the sugar receptor cell, but, in contrast to host ants, the impulse frequency did not significantly differ from that evoked by 50 mmol l−1 trehalose alone (Fig. 6 g, N = 5 sensilla from two foragers, F1,4 = 6.438, P = 0.064). When we used sucrose instead of trehalose, the impulse frequency evoked by the mixture of 50 mmol l−1 sucrose and 50 mmol l−1 glycine was not significantly different from that evoked by 50 mmol l−1 sucrose alone (Fig. 6 h, N = 8 sensilla from three foragers, F1,7 = 3.69, P = 0.093). Glycine did not affect either trehalose or sucrose reception in this case.
The effect of glycine on behavioral and sensory responsiveness to trehalose and sucrose
C. japonicus (host ant)
C. obscuripes (non-host ant)
It is known that feeding behavior and sensory responsiveness are modified by dietary experience. If animals are exposed to certain chemicals, the responsiveness to these chemicals is reduced (Glendinning et al. 1999; Bernays et al. 2003) or increased (Chapman et al. 2003; Del Campo and Miles 2003). In our study, however, the workers of both host and non-host ant species had no experience of feeding on the DNO secretions of N. fusca, because their colonies had not been parasitized by N. fusca. Thus, the gustatory synergism of glycine for sugar reception may not be induced by dietary experience, but rather be an innate gustatory event.
Synergistic effects on gustatory reception are well known in the umami taste of mammals. When two or more umami substances (monosodium l-glutamate and inosine 5-monophosphate) were mixed together, a synergistic enhancement of the response occurred. As the umami taste receptor, three G protein-coupled receptors (GPCRs) have been proposed in mammals (Chaudhari et al. 2000; Nelson et al. 2002; San Gabriel et al. 2005). In ants, sugar receptor genes are unknown, but a large gene family of GPCRs expressed in taste organs (Gr genes) has been reported in some insect species (Clyne et al. 2000; Dunipace et al. 2001; Scott et al. 2001; Hill et al. 2002; Robertson and Wanner 2006). In Drosophila melanogaster, Dahanukar et al. (2001) showed that a gustatory receptor encoded by Gr5a gene is highly tuned to respond to trehalose, but not to other disaccharides such as sucrose, and argued that other receptor(s) mediate the response to other sugars (Chyb et al. 2003). We found that the synergistic effects of glycine on sugar reception were altered either by sugars or ant species (Fig. 6; Table 3). Thus, our result suggested two possibilities: (1) there are at least two gustatory receptor molecules, one for trehalose and one for sucrose, and the synergistic effect of glycine on sugar taste would be a receptor molecule-specific event occurring in a sugar receptor cell and (2) the receptor molecules affected by glycine are specifically determined among ant species. Further studies focused on receptor molecules are necessary to reveal the molecular mechanism underlying amino acid-sugar gustatory synergism in ants.
Nomura et al. (1992) reported that glucose is the major sugar component of the DNO secretions of N. fusca and trehalose was not detected. It is suggested that the chemical composition of DNO secretion of lycaenid butterfly was affected by their diets (Daniels et al. 2005). Because N. fusca larvae are fed by workers through trophallaxis, the difference in sugar composition may be derived from host ant nutrition. Nevertheless, the dominant amino acid component of N. fusca secretions was glycine in both studies (Table 2; Nomura et al. 1992), so the amino acid component of N. fusca secretions is not affected by their diet.
Many ant species preferred sugar solutions containing a complex mixture of amino acids over sugar solutions. Such nutritious solution is similarly preferred across the ant species and monopolized by dominant ant species (Blüthgen and Fiedler 2004a). By contrast, gustatory preferences for sugar solutions mixed with different single amino acids varied among ant species (Blüthgen and Fiedler 2004b). These differences may be linked to variable nutrient requirements (Kay 2004) or physiological constraints (Wada et al. 2001). Therefore, in the species-specific interactions, lycaenid larvae can attract specific ant partners by designing nectars composed of sugar solutions mixed with single amino acids. In lycaenid butterflies, ant interactions are highly correlated with the consumption of nitrogen-fixing host plants (Pierce 1985). Intimate and host-specific lycaenids will secrete more nitrogen-biased nectar from the DNO (Pierce 1987; Daniels et al. 2005). The chemical compositions of the DNO secretions of several species of myrmecophilous lycaenids have been analyzed (Maschwitz et al. 1975; Cushman et al. 1994; Pierce and Nash 1999; Daniels et al. 2005), and the amino acid compositions of the DNO secretions were typically occupied by a single dominant substance, which varied among species. Therefore, the chemical composition of the DNO secretions would have been adaptively set for specific ant partners in various lycaenid–ant interactions.
However, despite the fact that species-specificity has only been observed in obligate interactions (Pierce et al. 2002), single dominant amino acids have also been reported in facultative lycaenids, albeit at much lower concentrations (Maschwitz et al. 1975; Daniels et al. 2005). This raises the question why facultative species, which can interact with numerous ant species, secrete single dominant amino acids. Previous studies reported that even in facultative interactions, actually only one or few species provide a clear benefit to the lycaenid (Wagner 1993; Fraser et al. 2001). For example, when the larvae of the facultative lycaenid butterfly Glaucopsyche lygdamus were tended by four ant species, only one ant species (Formica podzolica) reduced larval parasitism, whereas another ant (Formica obscuripes) significantly increased larval disappearance (Fraser et al. 2001). Therefore, even in facultative interactions, it is expected that the chemical composition of the DNO secretions tends to be tuned to the feeding preference of certain ant species, so as to provide clear benefits for lycaenids.
It is thought that highly specific obligate lycaenid–ant interactions have evolved from loose facultative mutualism (Fiedler et al. 1996; Fiedler 1998; Als et al. 2004). Further studies comparing the gustatory systems of ants for tasting DNO secretions of facultative and obligate lycaenid species will greatly contribute to clarifying the evolutionary aspects of the chemosensory factors mediating the species-specificity of lycaenid–ant interactions.
The larvae of many lycaenid species that parasitize ant nests and are fed by the workers through trophallaxis possess a DNO, and the function of the DNO in these species has long puzzled (Heath and Claassens 2003). Our results show that the synergistic gustatory response of the host ants to the DNO secretions provides a basis for observed species-specificity. However, it is possible that there are other substances in the secretions from the DNO, and they might also have a vital role in adoption by the host ants as well as appeasement of ants (Fiedler 1998; Pierce et al. 2002). More experimental work is required to investigate additional role of the DNO secretions, and the related questions, such as the cost in producing the DNO secretions as rewards for the ants.
We are grateful to David H. Hembry for helpful comments on the manuscript and language revision and two anonymous referees for their valuable comments. We thank Yasuhisa Endo for his help in scanning electron microscope, Hiroshi Oyama and Mayuko Shiraga for their help in chemical analysis, and Takeshi Takeda, Yuji Satoji, Daisuke Imaeda, Hayato Inui and Daisuke Umemoto for their help with collecting our study animals. All work presented here complied with the ‘‘Principles of animal care’’, publication No. 86–23, revised 1985 of the National Institute of Health, and also with the current laws of Japan, where these experiments were performed.