Discrimination between edible and contaminated foods is crucial for the survival of animals. In Drosophila, a family of gustatory receptors (GRs) expressed in taste neurons is thought to mediate the recognition of sugars and bitter compounds, thereby controlling feeding behavior. The expression of eight Gr genes in the labial palps, the fly's main taste organ, has been characterized in detail. These genes fall into two distinct groups: seven of them, including Gr66a, are expressed in 22 or fewer taste neurons in each labial palp. Additional experiments show that many of these genes are coexpressed in partially overlapping sets of neurons. In contrast, Gr5a, which encodes a receptor for trehalose, is expressed in a distinct and larger set of taste neurons associated with most chemosensory sensilla, including taste pegs. Mapping the axonal targets of cells expressing Gr66a and Gr5a reveals distinct projection patterns for these two groups of neurons in the brain. Moreover, tetanus toxin-mediated inactivation of Gr66a- or Gr5a-expressing cells shows that these two sets of neurons mediate distinct taste modalities -- the perception of bitter (caffeine) and sweet (trehalose) taste, respectively. It is concluded that discrimination between two taste modalities -- sweet and bitter -- requires specific sets of gustatory receptor neurons that express different Gr genes. Unlike the Drosophila olfactory system, where each neuron expresses a single olfactory receptor gene, taste neurons can express multiple receptors and do so in a complex Gr gene code that is unique for small sets of neurons (Thorne, 2004).

The labellum, considered to be the main taste organ in Drosophila, has approximately 62 chemosensory bristles (sensilla) that are arranged in a stereotyped pattern. These sensilla are morphologically identified as short (S), intermediate (I), and long (L). S and L bristles house dendrites of four chemosensory neurons, whereas I bristles are associated with two chemosensory neurons. To determine expression of Gr genes in these chemosensory neurons, the Gal4/UAS system has been used. This indirect method of expression analysis has proven far superior to RNA in situ hybridization due to low levels of Gr transcripts per cell and the wide distribution of taste neurons in tissues not amenable to sectioning procedures. The Gal4/UAS analyses revealed that a given Gr gene is expressed in a small number of chemosensory neurons per labial palp and, in each case, in only one neuron per chemosensory bristle. Hiroi (2002) demonstrated an association of specific Gr genes with certain bristles of the labellum. The majority of receptors examined are expressed in one of the four neurons of S type sensilla. For example, several Gr genes are strongly expressed in a single neuron associated with three S type sensilla (S1, S3, and S6) (Thorne, 2004).

Several issues with broad implications for taste coding remain to be elucidated. For example, it is still not known whether some Gr genes are coexpressed in the same neurons and, if so, to what extent. Similarly, it is not known what kind of taste properties are mediated by GRNs expressing these receptors. Finally, experiments to visualize axonal targets in the CNS of neurons expressing individual Gr genes have not been performed in any detail. To further advance understanding of Drosophila taste perception, these questions were addressed: the number of neurons expressing novel and previously characterized Gr genes was addressed, their extent of coexpression was investigated, the projection patterns of GRNs expressing these genes was visualized, and taste perception of flies lacking specific sets of GRNs was determined (Thorne, 2004).

Gal4 drivers (p[Gr]-Gal4) for eight Gr genes, Gr5a, Gr22b, Gr22e, Gr22f, Gr28be, Gr32a, Gr59b, and Gr66a, were combined with a UAS-nucGFP reporter gene encoding a green fluorescent protein tagged with a nuclear localization signal and images of optical sections through the entire labellum were collected by using confocal microscopy after anti-GFP antibody staining. By using the map generated by Hiroi (2002) as a guide, detailed analysis of confocal stacks allowed the organization and number of neurons expressing each of these genes to be more accurately determined. The expression patterns fell into two broad groups: Gr5a (representing the first group) was expressed in a large number of neurons throughout the entire labial palp, whereas the other Gr genes had restricted expression to relatively few neurons (Thorne, 2004).

Of the second group, Gr66a was expressed in the largest number (n = 22 ± 1)of cells per palp. Significantly, only a single neuron per S and I type sensillum stained positive for this driver. The neurons associated with S type bristles, which are located more medially, appeared larger in size compared to more laterally located neurons of I type sensilla. Gr22b, Gr22e, Gr22f, Gr28be, Gr32a, and Gr59b were expressed in fewer neurons than Gr66a. Expression of these receptors appears more restricted to larger neurons associated mostly with S type bristles. These expression studies provided the groundwork necessary to determine whether two or more Gr genes are actually coexpressed in the same neuron associated with an S type bristle (Thorne, 2004).

Ideally, coexpression of Gr genes may be addressed by labeling individual Gr gene-specific probes with different markers. However, expression levels of these genes are too low for reliable detection of transcripts by RNA in situ hybridization. Attempts were made to use the Gal4/UAS system along with a second reporter system, the tetracycline transactivator/tet-O reporter system. The sensitivity of this system, however, was too low to obtain reliable cell staining in taste neurons. Therefore, the issue of coexpression was addressed by quantification of labeled cells using the Gal4/UAS system, an approach that seemed feasible given the relatively low number of cells in which each receptor is expressed. Transgenic fly lines were made expressing UAS-nucGFP under the control of two different Gal4 drivers and then the number of labeled neurons was counted and compared to that of flies containing each driver alone. Surprisingly, in all cases where such double-driver experiments were carried out, the number of labeled cells expressing two drivers was close or equal to the number of labeled cells of flies containing the single driver with the higher cell count. For example, in flies that express either the p[Gr66a]-Gal4 or p[Gr22e]-Gal4 driver, an average of 22 and 14 neurons/labial palp are labeled, respectively. In flies that express both drivers, again approximately 22 neurons are detected per palp, which indicates that most if not all cells that express Gr22e also express Gr66a (Thorne, 2004).

A crucial determinant for discerning chemical cues present in the environment is embedded in the peripheral expression pattern of cell surface receptors in sensory epithelia. In the olfactory systems of Drosophila and mice, each olfactory receptor neuron expresses only one of 60 or one of approximately 1000 Or genes, respectively, enabling these animals to discriminate between hundreds or thousands of different odors. In contrast, taste cells of the tongue allow mammals to distinguish only a few taste qualities: bitter, sweet, umami, salty, and acidic taste. Lack of discrimination between the hundreds of diverse chemical compounds -- all perceived as bitter -- is thought to be caused by coexpression of the approximately 40 T2R receptors in a single set of taste cells (Chandrashekar, 2000; Zhang, 2003). Therefore, activation of the bitter taste cells by any one of the T2Rs is likely to generate a single activation pattern in taste centers of the brain, leading to a similar, repulsive behavioral output. Associating primary taste centers in the mammalian brain with specific taste modalities has, as of yet, proved challenging (Thorne, 2004).

Insect taste is still rather poorly understood, especially at the molecular level. Drosophila, which exhibits remarkably similar taste preferences with humans, is the only insect for which candidate receptors have been characterized experimentally. The investigations presented here provide significant new insights into insect taste perception (Thorne, 2004).

Initial expression studies suggested that the fly gustatory receptors are not simply coexpressed in three sets of cells dedicated to bitter, sweet, and umami taste like the T2Rs, T1R2/T1R3, and T1R1/T1R3 receptors of mammals. Instead,these experiments suggested that they either are expressed according to the one receptor (the 'one neuron' hypothesis well established for insect and mammalian olfactory systems) or they are expressed in partially overlapping sets of neurons. The current analysis supports the latter of these possibilities. Most labellar Gr genes (seven out of eight) are expressed in a single neuron of mostly S and some I type bristles. Most interestingly, coexpression studies provide evidence that individual neurons express anywhere from one to six receptors. In this way, S bristle-associated neurons are defined by unique receptor gene codes, thereby outfitting the labellum with an array of sensory assemblies that may exhibit distinct, albeit overlapping, ligand specificities (Thorne, 2004).

The functional implications of distinct neuronal receptor codes on taste perception are currently unclear and will require analysis of mutations of individual Gr genes. However, a general role for these neurons in feeding inhibition ('avoidance neurons') can be inferred from experiments presented in this study and supported through analogy with the mammalian taste system/receptors. (1) Avoidance neurons express the majority of analyzed Gr genes -- and by extension -- the majority of the genes in the entire Gr gene family. In mammals, bitter taste receptors far outnumber the sweet taste receptors (40:3). (2) Avoidance neurons associated with S type bristles do not express the receptors for the sugar trehalose encoded by the Gr5a gene. In fact, avoidance neurons associated with S type bristles have a distinct appearance compared to neurons expressing the Gr5a gene. In mammals, the sweet/umami taste receptors and the bitter taste receptors are expressed in distinct group of cells (Chandrashekar, 2000; Zhang, 2003). (3) Inactivation of avoidance neurons has no effect on sucrose or trehalose sensitivity in flies but significantly reduces their sensitivity to caffeine. (4) Avoidance neurons and Gr5a-expressing neurons have distinct targets in the subesophageal ganglion (SOG), a feature consistent with the detection of different taste qualities by these neurons (Thorne, 2004).

If the avoidance neurons have a general function in the detection of toxic or otherwise undesirable chemicals, what is the rationale for a complex and distinct Gr gene code among different groups of such neurons? It is proposed that the receptor code allows a fly to discriminate among different chemicals, which are in general avoided but might have distinct consequences on their health if ingested. According to such a proposal, a fly encountering a food source rich in nutrients (sugars) but contaminated with toxic chemicals may choose between feeding and avoidance, depending on the impact the particular toxic compound may have on its health. There is indirect evidence from feeding studies in Maduca sexta larvae that discrimination between the bitter substrates caffeine and aristolochic acid does occur in insects, even though actual taste preference, adaptation, or both may contribute to this phenomenon (Glendinning, 2001). Thus, discrimination among toxic/bitter-tasting compounds might be possible in insects including Drosophila (Thorne, 2004).

It was somewhat surprising that the sensitivity to other compounds known to be avoided by insects -- denatonium benzoate, quinine hydrochloride, and berberine -- was not affected in animals lacking Gr66a-expressing neurons. This may simply be explained by the presence of additional neurons expressing receptors that recognize these particular substrates. Alternatively, one or a few neurons coexpressing Gr66a along with a receptor for one (of these) ligand(s) might not have been completely inactivated by TNT. Finally, studies in rodents indicate that caffeine may directly affect neurons in the brain, circumventing activation of taste cells altogether. This is not likely to be the case in the current experiments, because none of the Gr genes examined is expressed in the CNS (Thorne, 2004).

Relatively few studies have investigated bitter taste sensitivity in insects, particularly Drosophila. Electrophysiological studies have identified bristles in the legs, but not the labellum of Drosophila, that respond to bitter-tasting chemicals (Meunier, 2003). However, S type sensilla are notoriously difficult to record from, because their bristles are extremely difficult to access for this type of experiment (Thorne, 2004).

Gr5a-expressing neurons represent more than half of chemosensory cells in the labellum and appear to be associated with all sensilla types, including the taste pegs. In fact, association of Gr5a with taste pegs provides the best evidence yet that these sensilla have a specific chemosensory function in the detection of trehalose. Significantly, Gr5a-expressing neurons define a largely distinct set of neurons from the avoidance neurons. This observation is consistent with the results from behavioral investigations of flies lacking the function of specific sets of neurons. Specifically, inactivation of Gr5a-expressing neurons leads to a reduction in trehalose sensitivity, but the sensitivity to any bitter substrate tested was unaffected. These flies did not exhibit reduced sucrose sensitivity, another nutrient-relevant sugar for Drosophila. This result is somewhat unexpected, since electrophysiological investigations have led to the proposal that a single neuron in L, I, and S bristles is responsive to several sugars including trehalose and sucrose (the 'sugar' neuron) (Dahanukar, 2001; Hiroi, 2002; Rodrigues, 1981; Dethier, 1976). According to these studies, sugar neurons may express a single, broadly tuned sugar receptor, or more likely, they may coexpress several distinct sugar receptors, each of which recognizes a specific sugar (i.e., sucrose, trehalose, fructose, etc.). This latter possibility is favored from genetic studies, which have shown that mutation in the Gr5a gene reduces the sensitivity of flies to trehalose, but not to sucrose. However, the proposition of a single sugar neuron per bristle is also not consistent with expression studies, which show that two to three neurons within a bristle can express Gr5a. The possibility that the p[Gr5a]-Gal4 drivers do not represent endogenous Gr5a expression cannot be excluded, but this is unlikely to be the case for two reasons: (1) several lines with p[Gr5a]-Gal4 show the same expression, and (2) the p[Gr5a_C]-Gal4 driver containing a much larger promoter fragment produces a similar expression profile, with many clusters of Gr5a-expressing neurons associated with the same bristle (Thorne, 2004).

In order to realign the electrophysiological data with expression analysis, another explanation is proposed: the 'sugar neuron' identified in electrophysiological studies expresses many (possibly all) distinct sugar receptors, including GR5a. However, one or two additional neurons per bristle express only a fraction, or possibly just one, of the sugar receptors present in the sugar neuron. Worth noting in this context is the fact that electrophysiological recordings are carried out at significantly higher substrate concentrations (up to 100 mM for sucrose and trehalose) than behavioral experiments (2 mM for sucrose and 25 mM for trehalose). The model is also more consistent with recent experiments (Hiroi, 2002) that noted different electrophysiological sugar responses among labellar sensilla (Thorne, 2004).

Approximately 45 labellar neurons have not yet been associated with any Gr gene, and some of these neurons might express putative candidate receptors for sucrose or additional sugars. These genes are likely to be encoded by members of the Gr64 gene cluster, which share much higher sequence similarity with Gr5a than any other Gr genes. Gal4 drivers for two of these genes (Gr64a and Gr64e) were analyzed and found to be expressed in the pharyngeal taste organs, but not in the labellum. Whether these two receptors are indeed involved in sugar detection remains to be seen, but it is predicted that other Gr genes for sugars like sucrose and fructose are be expressed more broadly and in taste neurons of labellar bristles and pegs (Thorne, 2004).

In summary, expression and behavioral studies suggest two fundamentally different roles for neurons expressing nonoverlapping groups of Gr genes in the detection of substrates that lead to feeding or avoidance behavior. According to this new model, S and I bristles on the labial palps contain one avoidance neuron and one or more feeding neurons (depending on the number of neurons associated with the particular bristle). The avoidance neuron expresses multiple Gr genes, and avoidance neurons of different bristles express these Gr genes in different combinations (the Gr gene code). The feeding neurons, which appear morphologically smaller than the avoidance neurons, express an entirely different set of receptors that includes Gr5a and possibly Gr genes encoding receptors for other sugars, amino acids, and peptides (Thorne, 2004).

The different functions for GRNs expressing Gr66a and Gr5a are also supported by their different projection patterns in the brain. Neurons expressing Gr66a or any of the partially coexpressed receptors target similar regions in the SOG/tritocerebrum, though the number of termini differs significantly depending on the number of peripheral sensory neurons the Gr is expressed in. For example, Gr66a-expressing neurons show a robust array of termini in the SOG/tritocerebrum, whereas the termini of Gr59b- and Gr22f-expressing neurons are significantly less numerous. In all cases, dense, contralaterally projecting fibers provide extensive innervation of both halves of the SOG by labellar neurons, as demonstrated by labial palp ablation experiments (Thorne, 2004).

An entirely different projection pattern is observed for feeding neurons that express Gr5a. Most strikingly, the axon termini of these neurons are distributed over a very large area of the SOG and extend into regions not innervated by avoidance neurons. A second striking difference is the poorly established contralateral connective between the two halves of the SOG, suggesting that neurons located in the right labial palp preferentially terminate in the right half of the SOG. This idea was tested and confirmed through ablation studies and has interesting implications, namely that spatially restricted activation of neurons in one palp will preferentially stimulate the same side of the SOG; this could potentially allow for spatial discrimination of taste input in the brain. This feature might allow the fly to orient its labellum in the direction of a food source, identifying regions with high concentrations of trehalose or other sugars (Thorne, 2004).

The distinct pattern of axon termini in the SOG of neurons required for feeding and avoidance suggests that these behaviors are mediated through different neuronal pathways. Anatomical studies in honeybees have identified second-order neurons that mediate synaptic activity of primary taste neurons to higher brain centers. It will be interesting to see whether second-order neurons contacting synapses of avoidance and feeding neurons define different target regions in these higher brain centers (Thorne, 2004).

Taste is an ancient sense, which exists in bacteria in the form of chemotaxis. Neuroanatomical and molecular comparison of taste systems between mammals and insects imply that this sense has evolved independently in these phyla. In mammals, for example, taste ligands are perceived through sensory epithelial cells in the lingual epithelium of the tongue. These cells then activate secondary neurons that innervate taste centers in the brain. In insects, tastants are detected by primary sensory neurons that directly innervate the CNS. Moreover, insects have multiple taste organs (legs, wings, and in some cases, the female genitalia) for which no counterparts exist in mammals. Finally, sequence comparison of the Gr and T1R/T2R genes has failed to reveal any direct kinship between mammalian and insect taste receptors (Thorne, 2004).

However, a remarkable convergence of anatomical as well as molecular features of gustatory systems between mammals and insects (Drosophila) appears to emerge from these studies. The functional taste units, the taste buds in the tongue and the taste bristles of the labellum, are composed of 30 to 100 taste cells and two to four chemosensory neurons, respectively. Individual taste cells in each taste bud are dedicated to the perception of sweet, umami, or bitter taste sensation based on the T1R or T2R receptors they express. Similarly, the data indicate that taste bristles of the labellum contain neurons that either respond to repulsive or attractive stimuli, properties that are likely determined by the specific (set of) taste receptors they express (Thorne, 2004).

Despite the sequence divergence of mammalian and insect taste receptors, it is believed there are intriguing similarities at the molecular level as well. The number of taste receptors in mammals and Drosophila is very similar. The eight genes described in this study probably encode a significant number of the functional labellar taste receptors. Some of the 60 Gr genes are likely to encode taste receptors only expressed in the pharynx, legs, and wings or might only be expressed in the larva. Other Gr genes are likely to function as pheromone receptors, or might recognize internal ligands based on their restricted expression in the CNS. Considering these alternative functions for some Gr genes, it is estimated that the fly has about 30 to 45 labellar taste receptors, a number close to the total number of T1Rs and T2Rs (30 in humans and 45 in mice) (Thorne, 2004).

In addition to the similar size of the Gr and T1R/T2R gene families, taste receptors of mammals and Drosophila fall into similar functional groups. Only three mammalian T1R receptors are thought to be dedicated to the detection of attractive stimuli (sugars and amino acids/proteins), whereas the large majority -- the T2Rs -- are thought to be exclusively involved in the detection of repulsive (bitter) ligands. If the current expression analysis is more or less representative of the entire Gr gene family, it might be expected that 25 to 40 Gr genes will be expressed in the avoidance neurons, whereas just three to six are expected to be expressed in feeding neurons. Identification and analysis of Gr genes encoding receptors for known ligands, combined with biochemical analyses, should reveal whether additional molecular features are shared between the GRs and T1Rs and T2Rs, such as whether Drosophila also possess a specific receptor for amino acids and whether some receptors also function as multimers, as is proposed for mammalian T1Rs (Thorne, 2004).