|Home | About | Journals | Submit | Contact Us | Français|
Users may view, print, copy, download and text and data- mine the content in such documents, for the purposes of academic research, subject always to the full Conditions of use: http://www.nature.com/authors/editorial_policies/license.html#terms
The mosquito Anopheles gambiae is the major vector of malaria in sub-Saharan Africa. It locates its human hosts primarily through olfaction, but little is known about the molecular basis of this process. Here we functionally characterize the Anopheles gambiae Odourant Receptor (AgOr) repertoire. We identify receptors that respond strongly to components of human odour and that may act in the process of human recognition. Some of these receptors are narrowly tuned, and some salient odourants elicit strong responses from only one or a few receptors, suggesting a central role for specific transmission channels in human host-seeking behavior. This analysis of the Anopheles gambiae receptors permits a comparison with the corresponding Drosophila melanogaster odourant receptor repertoire. We find that odourants are differentially encoded by the two species in ways consistent with their ecological needs. Our analysis of the Anopheles gambiae repertoire identifies receptors that may be useful targets for controlling the transmission of malaria.
Mosquitoes transmit many diseases, including malaria, which afflicts hundreds of millions of people each year1. The malaria burden is heaviest in sub-Saharan Africa, where the An. gambiae mosquito is the major vector. An. gambiae relies heavily on olfactory cues to identify its human hosts2–4, but the molecular basis of host-seeking behavior is unknown.
Insects detect odours via olfactory receptor neurons (ORNs). The odourant specificities of many ORNs are conferred by the expression of individual odourant receptor genes5. A family of 79 AgOr (Anopheles gambiae Odourant receptor) genes has been identified bioinformatically in An. gambiae6, 7. Two of these receptors have been characterized functionally8 using an in vivo heterologous expression system, the “empty neuron” system9, which has also been used to decode the D. melanogaster odourant receptor repertoire10–12. These results invited a systematic, functional characterization of the AgOr repertoire and a comparison between the receptor repertoires of these two species, which exhibit different olfactory-driven behaviors. D. melanogaster consumes fruit and is considered a generalist. An. gambiae has evolved an anthropophilic host-seeking olfactory response that allows it to find human bloodmeals4. Little is known about how the odourant receptor repertoires of these species have adapted to meet their distinct ecological requirements.
To investigate the molecular basis of odour reception in An. gambiae, we amplified the coding regions of 72 AgOr genes from olfactory organ cDNA of adult mosquitoes. We then expressed each AgOr in the “empty neuron,” a mutant ORN in D. melanogaster that lacks its endogenous odourant receptor9. Fifty of the 72 cloned AgOr receptors were functional in the empty neuron, conferring a regular, characteristic, spontaneous firing rate and exhibiting excitatory and/or inhibitory responses to odourant stimuli (Figure 1a). This success rate (69%) is comparable to that for D. melanogaster antennal Or genes (77%) expressed in the empty neuron11.
The empty neuron system previously was demonstrated to be a high-fidelity expression platform for the D. melanogaster Or genes11, referred to here as DmOrs. Since An. gambiae and D. melanogaster are separated by 250 million years of evolution13, we wanted to determine whether the empty neuron is also a faithful expression system for AgOr genes. One of the few AgOrs that has been unequivocally mapped to a specific ORN in the mosquito is AgOr8, which in its endogenous neuron responded to seven- and eight-carbon-chain compounds among a panel of tested odourants14. We expressed AgOr8 in the empty neuron and found that its response profile closely resembled that of the endogenous neuron (Figure 1b). We also generated dose-response curves for two ligands of AgOr8, 1-octen-3-ol and 1-hepten-3-ol (Supplementary Figure 1), and found that the differential sensitivity to these ligands observed in the endogenous neuron14 was maintained in the empty neuron. These results validate the empty neuron as a faithful heterologous expression system for AgOrs.
The 50 functional AgOrs were tested against a chemically diverse panel of 110 odourants, including components of human emanations and oviposition site volatiles (Supplementary Table 1). Fifty-three of the 110 odourants were previously tested against the D. melanogaster antennal receptor repertoire in the empty neuron system12, permitting functional comparisons between the odourant receptor repertoires of the fruit fly and the mosquito.
We tested each of the functional AgOrs against the 110-odourant panel, generating a data set of 5500 odourant-receptor combinations, with each combination tested n≥5 times (Figure 1c and Supplementary Table 2a–d). We found that individual receptors respond to subsets of odourants, and individual odourants activate subsets of receptors, consistent with a combinatorial model of odour coding12, 15, 16. Some receptors gave strong responses (defined as ≥100 spikes/second) to many odourants, while other receptors are more selective (Supplementary Table 2a–d). These differences were visualized by generating a tuning curve for each receptor (Figure 2). The breadths of the tuning curves were quantified according to their kurtosis value, a measure of the “peakedness” of the distribution (Supplementary Tables 3a, b). We found a continuum of AgOr tuning breadths ranging from broad to narrow, consistent with analysis of the D. melanogaster Or repertoire12. We then considered the receptors at each extreme for insight into the molecular basis of odour recognition.
Narrowly tuned receptors have been suggested to be specialist channels that carry information about odourants of high biological relevance17. Consistent with this hypothesis, the most narrowly tuned AgOrs are robustly excited by odourants with high biological salience. Among the receptors that respond strongly to at least one odourant of the panel, the most narrowly tuned are AgOr2, AgOr8, AgOr5, and AgOr65. AgOr2 is narrowly tuned to a small set of aromatics including indole, which was found to constitute nearly 30% of the volatile headspace of human sweat18. AgOr8 responds strongly to 1-octen-3-ol, a human volatile that is a strong attractant for several species of mosquito4. AgOr5 is tuned to 2,3-butanedione, is a metabolic byproduct of human skin microflora19. AgOr65 responds strongly to 2-ethylphenol, which is found in the urine of many animals20, 21. We found no mosquito receptors narrowly tuned to esters or aldehydes, odourants that dominate the headspace of many fruits22, 23. By contrast, among the most narrowly tuned receptors in the fruit fly, the strongest responses are in most cases to esters (DmOr85a, DmOr59b, DmOr67c) or to a terpene that contains an ester group (DmOr82a)12 (Supplementary Table 3b).
Some of the narrowly tuned AgOrs, in addition to responding strongly to odorants with high biological salience, respond with high sensitivity to these odorants. AgOr8 and AgOr2 respond to concentrations of 1-octen-3-ol and indole that range over more than four orders of magnitude and have response thresholds that lie between a 10−7 and a 10−6 dilution (Supplementary Figure 2).
Broadly tuned receptors lie at the other end of the AgOr distribution. It is possible that these receptors act in signaling the presence of odourants but not in specifically identifying or discriminating among them. We found that in An. gambiae, all strong responses to esters and aldehydes are conferred by broadly tuned AgOrs (all have kurtosis values less than the mean; Supplementary Table 2b and Supplementary Table 3a).
“Odourant tuning curves,” the reciprocal of receptor tuning curves, were also generated (Figure 3, Supplementary Table 3c, Supplementary Figure 3). For each odourant we plotted the responses of the 50 receptors along the X-axis, placing the strongest response at the center. Interestingly, the five odourants with the most narrow response distributions are all highly relevant to mosquito ecology. 3-methylindole is an oviposition site volatile that induces egg-laying and ORN responses in Culex mosquitoes24, 25. Indole is another oviposition site volatile26, in addition to being a major component of human emanations that is found both in sweat18 and human breath27. Geranyl acetate and citronellal are emitted from plants that are repellant to An. gambiae28, 29. Dimethylsulfide is emitted in human breath30 and is attractive to the Aedes aegypti mosquito31.
D. melanogaster odourant tuning curves were constructed from earlier data12 to identify odourants that likewise strongly activate a small number of receptors (Supplementary Table 3d). In contrast to the An. gambiae tuning curves, many of the odourants with high kurtosis values are esters, the dominant chemical class in fruit emanations23. Together, the mosquito and fruit fly odourant tuning curves support a model in which odourants of particular biological relevance are coded via a small number of channels.
Taken together, these results suggest that odourant and receptor tuning analyses provide complementary avenues to identify receptors and odourants that are important for innate insect behavioral responses. We note with special interest that in An. gambiae, one of the “narrowly tuned” odourants activates one of the narrowly tuned receptors. AgOr2 is strongly activated by indole, the odourant that constitutes almost 30% of the volatile headspace of human sweat18.
DmOrs were previously shown to yield responses that vary widely in their temporal characteristics, and the temporal dynamics were specific to the odourant-receptor combination11,12. To investigate the temporal dynamics of responses from AgOrs, we generated peristimulus time histograms for a number of odourant-receptor combinations (Supplementary Figure 4). As was observed with DmOrs, we find a diversity of temporal dynamics. Some odourants, including the human volatiles 1-octen-3-ol and indole, were capable of generating tonic responses that persisted throughout the three-second analysis period. Linalool oxide also generated a prolonged response. These odourants generated phasic responses from other receptors. The diversity of these responses suggests that temporal features may be a rich source of information about odourant identity.
How is the chemical world represented by the AgOrs? Excitatory activity is not distributed evenly across the odourant panel (Figure 1c, Supplementary Table 2a). For example, 28% of the strong responses were generated by heterocyclics, which constitute only 8% of the odourants in the panel. However, chemical class is only one descriptor of molecular identity. To represent the structural diversity among odourants more fully we adapted a recently developed odourant metric32. This metric is based on an optimized set of 32 molecular descriptors, including functional group, carbon chain length, and other physicochemical properties, which provide the basis of a 32-dimensional coordinate system. Each odourant can be mapped to a unique location in this multidimensional space. Odourants that are structurally similar lie close together in the space, while odourants that are structurally dissimilar lie far apart. To visualize this space we applied principal components analysis (PCA) to project it into two dimensions (Supplementary Figure 5a). As shown by the aromatics, odourants can be of the same chemical class yet map far apart.
Having mapped the odourants of the panel into this chemically defined odour space, we then asked how the AgOr repertoire covers the space. To illustrate the responses elicited by each odourant we generated a bubble plot, in which the location of each bubble indicates odourant identity, and the size of each bubble represents the magnitude of the response to that odourant, summed across all receptors and measured in total spikes/second (Figure 4a).
The AgOr repertoire is sensitive to a broad region of the odour space. However, the responses are not of uniform magnitude across the space. The odourants in the region of the space that is occupied by heterocyclics (purple), for example, elicit greater responses than the odourants in the region that is home to esters (dark green), and much greater responses than those in the region inhabited by carboxylic acids (pink). The AgOr repertoire may have evolved particular sensitivity to certain regions of odour space, such as the region containing aromatics, some of which are major components in human emanations 18, 33–39, and heterocyclics, a chemical class that includes volatiles proposed to promote oviposition behavior40. Such enhanced sensitivity could reflect the insect’s investment in detecting and discriminating among chemicals of these classes.
We also investigated the distribution of inhibitory responses across the An. gambiae receptor repertoire, which are not visualized in the odour space described above. As was observed across the D. melanogaster receptor repertoire12, most odourants elicit at least one inhibitory response, and most receptors are inhibited by at least one odourant (Supplementary Table 2a).
We next asked whether the An. gambiae and D. melanogaster Or repertoires differ in their coverage of odour space. As an initial means of addressing this issue, we considered the 53 odourants that were tested against both the AgOr and the DmOr repertoires12 and constructed odour spaces of the type described above for each receptor repertoire.
The two species differed in their relative coverage of odour space. The mosquito allocated greater relative coverage to the aromatics (Figure 4b; dark blue in lower right quadrant). By contrast, the fly devoted greater relative activity to some of the esters (dark green) and one of the two aldehydes that were compared (gray) (Figure 4c, Supplementary Figure 5b). We then compared the two species with respect to the strong responses (≥100 spikes/second). In the mosquito, 15% of the receptor-aromatic combinations yielded strong responses, compared to 7% of the receptor-ester combinations. By contrast, in the fly, 9% of the receptor-aromatic combinations yielded strong responses, compared to 20% of receptor-ester combinations (Figure 4d).
We next considered whether the species differed with respect to another kind of odour space, a biological odour space that relates odourants based on the primary sensory signals they generate. We created a space in which each axis represents the response magnitude (in spikes/second) for one odourant receptor, as described previously12. Odourants that elicit similar patterns of activity across the receptor repertoire map close together. Odourants that are close may be similar in their perceptual qualities, and may be more difficult for the animal to discriminate because they generate similar patterns of ORN activity. Experiments conducted with Drosophila larvae have provided evidence to support a relationship between such odour space distances and perception10.
We constructed such spaces for the mosquito and the fly and depicted them in three dimensions by applying PCA (Figure 5a). Odourants of the same chemical class tend to cluster together (Figure 5a), as previously observed in the fly12. However, some chemical classes were differentially distributed in the odour spaces of the two insects. Esters were more widely distributed in D. melanogaster space than in An. gambiae space, while aromatics were more widely distributed in An. gambiae space (Figures 5b and 5c). To quantify these differences, we calculated the Euclidean distance for every pair of odourants with the same functional group, within the odour space of each species. We found that the mean inter-odourant distance for esters is significantly higher for D. melanogaster than An. gambiae, while the mean inter-odourant distance for aromatics is higher for An. gambiae (p<0.001 for esters, p=0.01 for aromatics, Mann-Whitney); there were no differences in distances among alcohols or ketones, the other groups that could be compared. These results suggest that mosquitoes may be better able than fruit flies to discriminate among aromatics, while fruit flies may be better able to discriminate among esters, perhaps reflecting the biological relevance of these classes of compounds to the animals.
Are the functional differences between the mosquito and fruit fly Or repertoires due to a particular clade of mosquito or fly receptors? When the AgOr family was first identified, phylogenetic analysis revealed a clade of An. gambiae odourant receptors with no close D. melanogaster relatives, and a clade of Drosophila odourant receptors with no close An. gambiae relatives6, 7 (see also Supplementary Figure 6). Do these species-specific clades of receptors respond to odourants of a particular kind? We used matrix analysis to generate a dendrogram of receptors (Supplementary Figure 7), and PCA to create a “receptor space,” based on the responses of the receptors to odorants (Supplementary Figure 8). In neither case did we observe clustering of receptors of the species-specific clades. Thus, the observed functional differences between the An. gambiae and D. melanogaster odourant receptor repertoires may reflect evolutionary changes distributed across the receptor repertoires rather than concentrated in a specific branch of the phylogenetic tree.
We note finally that no AgOr showed even a modest response of 50 spikes/second to any carboxylic acid or to any amine in our system. By contrast, some DmOrs respond strongly to certain carboxylic acids12, 41. Recently, a set of variant-ionotropic glutamate receptors that respond to amines and carboxylic acids have been identified in D. melanogaster42. Receptors of this class may detect these compounds in An. gambiae.
Here we have functionally characterized the AgOr repertoire of odourant receptors from the mosquito An. gambiae. The olfactory system of An. gambiae allows the insect to locate human blood-meal hosts, thereby facilitating the transmission of malaria. We have identified individual receptors that respond robustly to human volatiles and may be central to the process by which the mosquito identifies its human hosts.
Strikingly, some receptors that respond to human odourants are narrowly tuned and highly sensitive. Reciprocally, some notable odourants elicit strong responses from one or a few receptors. These highly focused relationships between certain odourants and certain receptors suggest a role for specific transmission channels in guiding the animal’s behavior. A recent study of the D. melanogaster antennal lobe documented lateral inhibitory interactions that increased with greater total ORN input43. Odourants that excite few receptors in the mosquito may produce signals that suffer less inhibition and enjoy more saliency. Another study found that ablation of either of two narrowly tuned ORN classes impaired behavioral attraction to their cognate odourants44.
Since our analysis was conducted using the same expression system as our previous study of D. melanogaster odourant receptors, it provided a unique opportunity to compare the Or repertoires of two species that belong to the same order but that exhibit different olfactory-guided behaviors. An outstanding question in the field of olfaction is how an organism’s ecology shapes the function of its odourant receptor repertoire. A full answer to this question requires the functional characterization of entire receptor repertoires. We found that the two species show different coverage of a chemically defined odour space. Certain classes of odourants are differentially distributed in a biologically defined odour space of each species. These differences suggest the evolution of olfactory acuity and discriminatory power consistent with the ecological needs of each species. These evolutionary changes appear to have occurred over the odourant receptor repertoire as a whole, as opposed to having been effected by the emergence of a species-specific receptor clade.
The results may have implications for the control of malaria, one of the world’s most devastating diseases. Screens for activators and inhibitors of selected receptors may identify compounds that attract mosquitoes into traps, interfere with their navigation, or repel them.
AgOr cDNAs were cloned by standard procedures and expressed in the empty neuron as described previously8, 14. Extracellular single-unit physiology was performed as described previously9, 11. Physicochemical odor space was constructed using the set of 32 optimized DRAGON descriptors32, which were normalized.
Five-day-old laboratory-maintained An. gambiae mosquitoes of the Suakoko strain were cold anesthetized and the antennae, maxillary palpae, and proboscises were dissected by hand on a chill table. RNA was prepared with RNeasy (QIAGEN) according to the manufacturer’s instructions. The RNA preparation was used for oligo dT-primed cDNA synthesis with Superscript II Reverse Transcriptase (Invitrogen) for the generation of templates for subsequent PCR reactions. Negative control samples with no reverse transcriptase were included in each cDNA synthesis and subsequent PCR analysis. PCR was performed with a Mastercycler Gradient (Eppendorf) under the following conditions: 94°C for 5 min; 40 cycles of 94°C for 30 s, 55°C to 60°C for 30 s (annealing temperature varied depending on primer pair), 72°C for 60 s; and 72°C for 7 min. PCR amplification products were separated on a 1.0% agarose gel and were cloned into the Gateway pENTR entry vector (Invitrogen) or pGEM-TEasy (Promega) and verified by sequencing. Sequences of at least two independent clones were obtained for each Or and compared to verify polymorphisms as such rather than PCR errors. Sequence discrepancies were resolved by a second PCR reaction. We amplified 72 AgOrs, in addition to AgOr7, the An. gambiae ortholog of the atypical D. melanogaster odourant receptor gene Or83b45. Four (AgOr37, AgOr40, AgOr52, AgOr58) of the remaining six AgOr genes that were not amplified from adult tissue are expressed in the larval stage46, and the other two may be artifacts of gene annotation.
The ab3A mutant flies and Or22a-GAL4 constructs were described previously 9, 11. To generate UAS-Or constructs, the pUAST vector (C. G. Warr) was adapted to generate a Gateway (Invitrogen) compatible destination vector. LR recombination reaction was performed with the Gateway (Invitrogen) pENTR entry clones and the modified pUAST vector. For those amplification products cloned into pGEM-TEasy, subcloning into unadapted pUAST was performed using the restriction enzymes BglII, KpnI, and NotI (New England Biolabs). An exception was AgOr53, which was cloned into pNmyc-UAST vector (C. G. Warr) in frame with the start codon and three copies of the myc tag coding sequence. The resulting protein had an N-terminal myc tag. No other receptors had epitope tags.
Extracellular single-unit recordings were performed essentially as described previously 11. Odourant stimuli were prepared in Pasteur pipettes as described previously 11. Chemicals were of the highest purity available (Sigma-Aldrich). Seven-octenoic acid and cis/trans-3-methyl-2-hexenoic acid were synthesized by Richman Chemical (Lower Gywnedd, PA). All chiral chemicals were racemic mixtures with the exception of (+)-carvone, (−)-carvone, (+)-fenchone, (−)-fenchone, and L(+)-lactic acid. Ammonia, cadaverine, putrescine, acetic acid, propanoic acid, butanoic acid, isobutyric acid, isovaleric acid, L(+)-lactic acid, and 2-oxohexenoic acid were diluted in H2O. Hexadecanoic acid, octadecanoic acid, 5α-androsten-3α-ol, were diluted in ethanol. All other odourants were diluted in paraffin oil (Fluka). Liquid odourants were diluted in solvent to a concentration of 10−2 volume/volume, and solid odourants were dissolved, 50 mg in 5 ml solvent.
Stimuli were presented by placing the tip of the pipette through a hole in a tube carrying a purified air stream (32 ml/s) directed at the fly and administering a pulse of charcoal-filtered air (3.2 ml/s) through the pipette containing the odourant. Pulse duration was 500 ms. Stimuli were used for a maximum of four presentations. Responses were quantified by subtracting the number of impulses in 500 ms of unstimulated activity from the number of impulses in the 500 ms following odourant stimulation, after a 150 ms delay to allow the odourant to travel down the airstream. Responses to diluents were also subtracted. For each odourant, each recording was from a separate sensillum, with no more than three sensilla analyzed per fly. Recordings were obtained from flies between 4 and 14 days old.
A panel of six odourants previously tested by E. Hallem against the ab3B neuron were retested in this study to control for possible differences between electrophysiology rigs; none of the responses were significantly different (p ≥0.12 in all cases, t-test).
Principal component analysis (PCA), and hierarchical cluster analysis were performed using PAST, a statistics program (http://folk.uio.no/ohammer/past/) as described previously 12. Physicochemical odour space was constructed using the set of 32 optimized DRAGON descriptors32 (Talete, srl, DRAGON for Windows, version 5.5, 2007, http://www.talete.mi.it/). Descriptors were normalized. The twelve odourants that generate net negative (inhibitory) responses (1-chlorododecane, 3-methyl-2-hexenoic acid, cadaverine, cis-9-octadecanoic acid, delta-decalactone, hexadecanoic acid, nonanoic acid, octadecanoic acid, octanal, octanoic acid, putrescine, and tridecanoic acid) are not shown in the bubble plots of Figure 4. To generate odour spaces and for cluster analyses, we removed from analysis receptors that did not respond to any odourant on the panel with a response ≥50 spikes/second and odourants that did not elicit any responses ≥50 spikes/second (before solvent responses were subtracted) unless otherwise noted. Error bars represent SEM, unless otherwise noted. Phylogenetic analysis was performed with MEGA 4.0.2, using a Neighbor-joining algorithm and 500 replications. Peristimulus time histograms were generated using IGOR Pro 6.0 (Wavemetrics).
We thank W. van der Goes van Naters and C. Yao for help with electrophysiology, E. Hallem, S. Kreher, J. Salzman and T. Emonet for assistance with data analyses, and T.-W. Koh for comments on the manuscript. We thank P. Graham, Z. Berman, A. Rabin, M. Dillon and E. Kelley-Smith for technical assistance. We thank Y.-T. Qiu for assistance generating Figure 1b and Supplementary Figure 1. This work was funded in part by grants from the Foundation for the National Institutes of Health (NIH) through the Grand Challenges in Global Health Initiative to L.J.Z. and from the NIH to L.Z. and J.R.C.
Electrophysiology and computational analysis were performed by A.C. Molecular cloning was performed by A.C., G.W., and C.-Y.S. A.C. and J.C. wrote the manuscript. All authors contributed to the design and interpretation of the study.
The authors declare no competing financial interests.