Research in insect olfaction has produced valuable insightsinto neural processes involved in sensory coding andprocessing. The fruitfly Drosophila melanogaster has emergedas a highly suitable model organism for olfactory research. Ithas an anatomically relatively simple and morphologicallywell-characterized olfactory system (Stocker, 2001).Additionally, as the complete genomic sequence is available(Adams et al., 2000; Rubin et al., 2000), there are extensivepossibilities for genetic manipulations.

The recent identification of the first insect odorant receptors(OR) in Drosophila (Clyne et al., 1999; Gao and Chess, 1999;Vosshall et al., 1999) has opened up new possibilities forresearch regarding olfactory coding in insects. The ≥60DrosophilaOR genes (DORs; Vosshall et al., 2001) encode afamily of seven transmembrane G-protein-coupled receptors,whose function is to recognize specific odorant molecules(Störtkuhl and Kettler, 2001; Wetzel et al., 2001). The DORsare expressed in dendrites of olfactory receptor neurones(ORNs) housed in sensilla located on the antennae and themaxillary palps, the two olfactory organs of Drosophila.

Although Drosophila is currently a favoured model for

olfactory research, information regarding odour ligands isscarce. With the cloning of the DOR family, this shortcomingis highlighted, as much related future work will rely on theavailability of key stimuli of the system. Drawing conclusionsconcerning, for example, ligand–OR interactions is difficult ifthe relevance of the ligands at hand is questionable. Moreinformation regarding biologically significant odour ligands istherefore sorely needed. Investigations into odour detection inDrosophilaORNs are surprisingly few considering the amountof work done on other aspects of the system. The thoroughstudies by de Bruyne et al. (1999, 2001) report several odorantseliciting responses from antennal and maxillary palp ORNs.Other studies report odorants of various potency as ligands(Siddiqi, 1991; Clyne et al., 1997). These studies have all reliedon synthetic stimulus sets. A problem when screening withsynthetic odorants is that only a fraction of all possiblyimportant stimuli can be tested. The number of potentiallybiologically relevant odorants for a polyphagous species likeDrosophila is in the range of thousands. Thus, in the studiesso far conducted it is likely that many key ligands have beenoverlooked.

715The Journal of Experimental Biology 206, 715-724© 2003 The Company of Biologists Ltddoi:10.1242/jeb.00143

Due to its well-defined genome, the fruitfly Drosophilamelanogaster has become a very important modelorganism in olfactory research. Despite all the researchinvested, few natural odour ligands have been identified.By using a combined gas chromatographic–single receptorneurone recording technique (GC–SC), we set out toidentify active odour molecules in head space-collectedvolatiles from preferred food sources, i.e. differentoverripe or rotting fruit. In total, we performed 101GC–SC experiments on 85 contacted sensilla. UsingGC–mass spectrometry, we identified 24 activecompounds. Synthetic samples of these compounds wereused to establish dose–response curves for several of thereceptor neurone types encountered. The responsepatterns of individual neurones were repeatable, andneurones were found to reside in stereotyped pairs. In

total, we identified eight distinct sensillum types based onresponse profiles of 12 olfactory receptor neurone types.In most recordings, a single GC peak would produce astrong response, whereas a few other, often chemicallyrelated, compounds would produce weaker responses. TheGC–SC recordings revealed that the olfactory receptorneurones investigated were often selective and could bedivided into distinct functional types with discretecharacteristics. Dose–response investigations revealedvery low response thresholds to the tested compounds. Sixof the novel ligands were also tested for their behaviouraleffect in a T-maze set up. Of these, five elicited attractionand one elicited repulsion.

Key words: Drosophila, olfaction, insect, antenna, food odour,receptor neurone, ligand, odour coding.

Summary

Introduction

Novel natural ligands for Drosophilaolfactory receptor neurones

Marcus C. Stensmyr1, Elena Giordano2, Annalisa Balloi2, Anna-Maria Angioy2 andBill S. Hansson1,*

1Division of Chemical Ecology, Department of Crop Science, Swedish University of Agricultural Sciences,PO Box 44, SE-23053 Alnarp, Sweden and 2Department of Experimental Biology, University of Cagliari,

S.S. 554 Km, 4500 Monserrato, Italy*Author for correspondence (e-mail: bill.hansson@vv.slu.se)

Accepted 13 November 2002

716

Linking electrophysiology with gas chromatography (GC)(Arn et al., 1975; Wadhams, 1982) can solve this shortcoming.GC–electrophysiology, in particular GC linked with recordingsfrom single ORNs, so called GC–single cell (GC–SC), hasproven itself a potent method for identification of odour ligands(Wibe et al., 1997; Stensmyr et al., 2001). The GC–SCtechnique enables screening of single odours present in extractsof favourable food resources that are likely to contain theodorants that the ORs have evolved to detect. Here, we reportphysiological responses from antennal ORNs using a GC–SCstimulation technique. We screened a large number of potentialligands from favoured food resources and characterized theolfactory tuning of 12 ORN types housed in eightphysiologically distinct types of sensilla. In addition, wepresent dose–response functions as well as behaviouralresponses to several of the identified odorants.

Materials and methodsRecordings from receptor neurones

Throughout this study, we used the wild-type Oregon R stockof Drosophila melanogaster L. For the electrophysiologicalrecordings, a fly was mounted in a truncated pipette tip with theantennae protruding from the narrow end. The pipette tip wasfixed with wax on a microscope slide, and the antennae gentlyplaced on a cover slip and stabilized with a glass micropipette(method modified from Clyne et al., 1997). To establishcontact with the ORNs, we used tungsten microelectrodes,electrolytically sharpened in KNO2 solution (Hubel, 1957). Wepositioned the recording electrode and the indifferent electrode(inserted into the head capsule) using a preparation microscopeand a recording unit with combined joystick micromanipulatorsand amplifier (Syntech INR-02, Hilversum, The Netherlands).Action potentials from contacted ORNs were visualized on anoscilloscope (Tekscope, Tektronix, Beaverton, USA) linked toa loudspeaker for audio monitoring. Recordings of actionpotentials were stored on a computer and analysed with thesoftware Auto Spike v. 4.0 (Syntech, Hilversum, TheNetherlands). The activity of co-located ORNs in single sensillawas based on differences in spike amplitude. The ORN with thelargest spike amplitude was termed A; the second largest B andso forth. As spike amplitudes sometimes changed duringextensive firing, we had to complement the static template usedby the software for spike sorting with manual sorting, whereattention was also paid to the shape of the spikes. Contactswhere problems arose in assigning identity to ORNs wereexcluded from further analysis. The position of the recordingelectrode was noted on template drawings of the Drosophilaantennae (based on Shanbhag et al., 1999). Magnification (upto 220×) was too small to accurately pinpoint the morphologicaltypes of sensilla contacted but was, however, sufficient toaccurately note the general area recorded from.

Odour stimuli

We chose six different fruit types for extraction of volatilecomponents. The fruits chosen, on the basis of them being

attractive to Drosophila, were banana, litchi, mango, papaya,passionfruit and pineapple. In addition, we also prepared ayeast extract. From these sources, we prepared nine extracts,three from banana (from various stages of ripeness) and onefrom each of the others. Prior to volatile extraction, the fruitwas left to ripen and the yeast was mixed with sucrose. Thevolatile contents were collected using a closed loop strippingsetup, as described by Stensmyr et al. (2001). We also used 22synthetic compounds: acetoin, butyl acetate, butyl butyrate,cyclohexanol, ethyl butyrate, ethyl hexanoate, ethyl 3-hydroxyhexanoate, furfural, hexyl acetate, hexyl n-butyrate,isoamyl acetate, isoamyl alcohol, isovaleric acid, methylhexanoate, phenylacetonitrile (Aldrich, purity >98%),phenylacetaldehyde (Aldrich, purity >90%), ethylideneacetone (Aldrich, purity >70%), acetyl furan, 2,3-butanediol,ethyl 3-hydroxybutyrate, 1-hexanol (Fluka, purity >97%) andphenylethyl alcohol (Sigma, purity >98%). Neat compoundswere diluted in redestilled hexane or CH2Cl2 in decadic stepsfrom a concentration of 10µgµl–1 down to 100 pgµl–1. Fromthe extracts and from the synthetic compounds, 10µl waspipetted onto small pieces of filter paper (approximately10 mm×15 mm; Munksjöpapp, Grycksbo, Sweden) placed inPasteur pipettes. Blank cartridges, containing only filter paperplus solvent, were also prepared. We frequently refilled orrenewed the test cartridges.

Odour stimulation

Once contact with ORNs was established, the testcartridges containing odours were screened (presented inrandom order) for physiological activity, i.e. if the odourselicited a change in the action potential firing frequency. Aglass tube, with its outlet approximately 10 mm from theantenna, delivered a constant flow of charcoal-filtered andhumidified air at a velocity of 0.5 m s–1 over the preparation.Stimulation was performed by inserting the tip of the testcartridge into a hole in the glass tube, approximately 15 cmbefore the outlet. The test cartridge was connected to astimulus controller (Syntech CS-02, Hilversum, TheNetherlands) that generated air puffs (2.5 ml in 0.5 s) throughthe cartridge into the constant air stream in the glass tube.ORNs responding to cartridge-delivered fruit extract odourswere further examined by stimulation of the active extract(s)via GC. A sample of approximately 2–3µl of the extract tobe tested was injected onto the GC column for separation. Atthe end of the GC column, a cross split was installed, leadinghalf of the effluent to the flame ionization detector (FID) andhalf out of the GC oven into the glass tube carrying theconstant airflow to the antenna, enabling simultaneousrecordings of activity from ORNs and FID. The GCseparations were performed on a 30 m×0.25 mm i.d. polarcapillary column (HP-innowax) fitted in a Hewlett Packard5890 GC Plus. Carrier gas was hydrogen at 0.5 m s–1 linearvelocity. Detector temperature was set at 250°C and injectortemperature at 225°C. Oven temperature was maintained at40°C for 2 min, then programmed to rise to 230°C at either20°C min–1 or 10°C min–1.

M. C. Stensmyr and others

717Natural olfactory ligands in Drosophila

During the GC–SC recordings, an increased firing rate ofaction potentials of at least twice the spontaneous activity wasinterpreted as a response (no inhibitions were observed; seeResults). We quantified the response strength as the maximumspike frequency (spikes s–1; counted over 200 ms intervals)during the period of increased neurone activity followingstimulation.

Dose–response relationships

We also performed dose–response trials with syntheticcompounds in order to establish the response thresholds forthose odorants eliciting the strongest response. These werepresented in increasing decadic dosages from 100 pg to 10µgfrom Pasteur pipettes using the stimulus controller describedabove. The response strength was calculated by subtracting thenumber of spikes 1 s after stimulation with the number ofspikes in the preceding 1 s period. The net response was thensubtracted by the net response to blank in order to avoid anynon-specific responses.

Chemical analysis

Active compounds in the extracts were identified throughcoupled GC–mass spectroscopy (GC–MS). 1–2µl extract wasinjected onto a Hewlett Packard 5890 GC Plus equipped witha 30 m×0.25 mm ID polar capillary column (HP-innowax).Helium was used as carrier gas, set at 0.44 m s–1. A gradient of10°C min–1 (alternatively 5°C min–1) from 40°C to 230°C witha starting temperature hold of 6 min was utilized. Separationsthen passed through an HP5972 mass spectrometer in scanmode with an electron ion source and quadropole massfilter. Active FID peaks were identified by their massspectra. These spectra were compared with referencespectra from a NIST/EPA/NIH 75K electronic database(http://www.nist.gov/srd/nist1a.htm). The mass spectraidentification was confirmed viaco-injection of the syntheticcompound with the extract itself.

Behavioural analysis

To screen the behavioural effect of identified odorants weused a T-maze setup (Tully and Quinn, 1985; Helfand andCarlson, 1989). Tested odorants were diluted in paraffin oil(except acetoin, which was diluted in water) in decadic stepsranging from 10 pgµl–1 to 10µgµl–1. 10µl of the test odorantwas applied to a 1 cm2 filter paper; as a control, we usedidentical filter papers with 10µl of the solvent. 20–30 flies ofthe same sex, 3–8 days old, were introduced into the setup andexposed to an airstream (1 l min–1) passing through the system.The flies were allowed 30 s to choose between the odour sideand the control side, thereafter the flies were gathered andcounted. Flies were only tested once and all tests wereperformed in the dark to exclude visual effects. The responseindex (RI) was calculated as the number of flies choosing odour(O) subtracted by the number of flies in the control (C), dividedby the sum of all choosing flies (O+C); i.e. RI=(O–C)/(O+C).An RI of 1.0 equals full attraction, whereas an RI of –1.0 equalsfull avoidance. Indifference to the odour is indicated by an RI

of zero. Flies not making any choice were excluded fromfurther analysis.

ResultsDrosophila ORNs respond selectively to odours in the large

set of stimuli screened

We recorded neural activity of ORNs from single olfactorysensilla on the third antennal segment (Fig. 1). In total, weperformed 101 GC–SC experiments on 85 contacted sensillafrom both male (46 sensilla from 17 individuals) and female(39 sensilla from 15 individuals) Drosophila. An example ofa simultaneous GC–SC recording is shown in Fig. 2. Nodifference in response profile between the sexes was observed,and thus the data were pooled. Responses to one or several ofthe screened fruit odorants were observed in 69 recordings. Ofthese, 12 were excluded from analysis due to bad contact orcontact being lost during the experiment.

Individual ORNs responded in a very selective manner tothe screened odorants. Typically, an ORN would only respondto a few of the GC-separated extract components, and allresponses observed were excitatory. The maximum number ofresponses from an individual ORN to any of the extracts wasfrom a banana sample, which contained seven activecomponents. In most recordings, a single peak would producea strong response, whereas a few other compounds wouldproduce weaker responses (Fig. 2). As we used extracts, wewere not able to correct for dosages. The components werepresent in the ratio that they occur in nature. However, thecompounds eliciting response were not necessarily those that

Fig. 1. Simultaneously recorded activity of two co-located olfactoryreceptor neurones. Differences in action potential (spike) amplitudeallow for separation of activity. Stimulation with a blank results inno increased activity from any of the neurones. Stimulation withethyl 3-hydroxybutyrate elicits a strong response from the smallerspiking neurone, indicated by ‘B’ (neurone later classified as S2B),whereas the larger spiking neurone, indicated by ‘A’ is unaffected bythe stimulus. Boxed area shows an expanded part of the recording.Horizontal bar indicates stimulus duration, 0.5 s.

ABlank

Ethyl 3-hydroxybutyrate

B

BA

718

occurred in the highest amounts. On the contrary, many of theactive compounds were only present in very low amounts,some of them barely visible on the FID trace.

Using GC–MS, we identified the active compounds. In total,we observed responses to 35 FID peaks in the fruit extracts.We found 31 of these peaks to correspond to 23 differentcompounds, of which four occurred in more than one extract.The remaining four peaks either occurred in too low amountsor did not produce clear mass spectra. These componentsremain unidentified. The identified compounds are presentedin Table 1.

Eight physiological types of sensilla were identified

Our GC–SC recordings revealed that the ORNs could bedivided into functional types with discrete characteristics

(Fig. 3). The response patterns from individual neurones wererepeatable, and neurones were found to reside in stereotypedconstellations. In total, we identified eight physiologicallydistinct sensillum types based on the response profiles of theirORNs (Fig. 4). The topographical distribution on the antennaeof these sensilla is shown in Fig. 5. In addition, we alsoscreened the most potent stimuli for each ORN type assynthetics in known concentration ranges to verify theneurones’ response characteristics and detection thresholds(Fig. 6; see below).

From the proximo-medial region of the antennae, whichhouses the large s. basiconica (Shanbhag et al., 1999), weidentified three types of sensilla, which we denoted S1(sensillum type 1), S2 and S3 (see Fig. 4). The S1 type, whichhoused four ORNs, was the most frequently encountered(N=32). Of these four ORNs, the A ORN did not respond toany fruit compounds. The C ORN responded strongly andselectively to CO2, whereas the B ORN responded moststrongly to acetoin; weaker responses were elicited by threeother compounds. The D ORN was found to fire moderateresponses to furfural. The S2 type (N=13) housed twoneurones, of which only the B ORN responded to thescreened odorants. The primary ligand was ethyl 3-hydroxybutyrate, although these neurones also responded,less strongly to five other compounds. Three of these remainunidentified. One of the identified compounds, ethyl butyrate,shared structural properties with the primary ligand, while the

M. C. Stensmyr and others

Table 1.Physiologically active compounds identified fromfruit extracts

Compound CAS no.

Acetoin 513-86-0Acetyl furan 1192-62-72,3-butanediol 513-85-9Butyl acetate 123-86-4Butyl butyrate 109-21-7Cyclohexanol 108-93-0Ethyl 3-hydroxybutyrate 5405-41-4Ethyl 3-hydroxyhexanoate 2305-25-1Ethyl butyrate 105-54-4Ethyl hexanoate 123-66-0Ethylidene acetone 625-33-2Furfural 98-01-11-Hexanol 111-27-3Hexyl acetate 142-92-7Hexyl n-butyrate 2639-63-6Isoamyl acetate 123-92-2Isoamyl alcohol 123-51-3Isovaleric acid 504-72-2Methyl hexanoate 106-70-7Phenylacetaldehyde 122-78-1Phenylacetonitrile 140-29-4Phenylethyl alcohol 60-12-8Sec-amyl acetate 626-38-0

CAS no., Chemical Abstract Service number.Spike amplitude

S5A

No.

spi

kes S5B

S5A

S5B

GC

SC

SC

A

B

41

25

3

Fig. 2. A linked gas chromatography–single cell recording froman antennal sensillum (S5) containing two olfactory receptorneurones (ORNs). (A) Gas chromatogram (GC) of a papaya head-space sample and the corresponding single-cell recording (SC),presented in histogram form (spikes s–1; vertical bar, 25 spikes s–1;horizontal bar, 1 min). The two ORNs present are sensitive todifferent components in the papaya volatile collection. The S5Aneurone responds primarily to butyl butyrate (5), but also to 1-hexanol (2) and isovaleric acid (4). The S5B neurone respondsstrongly to isoamyl acetate (3) and moderately to butyl acetate (1).(B) Action potential amplitudes recorded during the aboveexperiment. The distribution of amplitudes from the recordingshows the clear physiological separation of the two ORNs (S5Aand S5B).

719Natural olfactory ligands in Drosophila

third most-potent stimulus was cyclohexanol, a radicallydifferent molecule. From the third sensillum type in thisregion, the S3 type (N=24), we only observed increasedactivity from the A neurone, which responded most stronglyto ethyl hexanoate and, in decreasing strength, to six othercomponents of structural proximity. These compounds wereall similarly sized (6–8-carbon length) straight-chainedesters, some only differing slightly from each other. Inaddition, a compound that remains unidentified also elicitedmoderate responses. The distribution of these three sensillumtypes within this antennal region showed slight differencesbetween them (Fig. 5). We found all three types along themedial section. However, the S1 and S2 types were also

found on the posterior face, in proximity to the sacculusopening.

The other identified sensillum types were not foundclustered in specific regions of the antenna but had a morescattered distribution. The S6 (N=1) and S7 (N=3) types bothresponded selectively to single compounds: sec-amyl acetateand isoamyl alcohol, respectively. The S6 sensillum wasfound lateral to the sacculus opening, a region that exclusivelyhouses s. coeloconica (Shanbhag et al., 1999). The S8sensillum (N=4) was distinctive in that it housed threeneurones. The B ORN responded to three similar phenoliccompounds of which phenylacetonitrile triggered thestrongest response. The presence of three ORNs in thissensillum suggests that it was an s. coeloconicum (Shanbhaget al., 1999). The S4 (N=4) and S5 (N=6) types had similar,partly overlapping, response spectra. The most efficientligand, triggering strong responses for the S5A ORNs, was 1-hexanol, which also elicited weak responses from the S4AORNs. The S4A ORNs did, however, respond slightly morestrongly to butyl butyrate. The B ORN of both sensillum typesresponded to isoamyl acetate. However, S5B responded morestrongly to this compound. In addition, the S5B ORNsresponded to three additional compounds that elicited noactivity from the S4B ORNs whatsoever. In addition, wefound seven sensilla for which the key stimuli remainunidentified.

Testing the identified key ligands as synthetics, we obtainedrecordings from both male (51 sensilla from seven individuals)and female (45 sensilla from five individuals) flies. Nodifference in response patterns between sexes was observed,thus data were pooled. We relocated and obtained dose–responserelationships for five of the identified ORN types from fivedifferent sensillum types. The S1B and S2B ORNs were bothhighly sensitive, capable of detecting their key ligands at 100pgdoses (Fig. 6B,C), whereas the S5A and S8B ORNs wereslightly less sensitive to their respective ligands, requiring 1ngdoses to elicit response (Fig. 6D,E). For the S3A ORN, we testedthe four most potent stimuli as synthetics. Ethyl hexanoate and

GC

SC

#19

Passion fruit

Banana

Pineapple

A

B

C

#52

#31

#52

41

253

87

6

7 4

Fig. 3. Physiological sensillum types were identified by the responsespectra of their olfactory receptor neurones (ORNs) when challengedwith volatiles collected from different fruits. (A) An S3A ORN(number 52) stimulated with a volatile collection from passion fruit.A strong response is elicited by ethyl hexanoate [peak 4 in the gaschromatogram (GC)], while smaller responses are elicited by ethylbutyrate (1), ethyl 3-hydroxybutyrate (2), ethyl 3-hydroxyhexanoate(5) and an unidentified compound (3). The physiological response isrepresented in the histogram displaying the single cell (SC) responseover time (spikes s–1; vertical bar, 25 spikes s–1; horizontal bar,1 min). (B) The same neurone challenged with a volatile collectionfrom banana reveals three smaller responses to isoamyl acetate (6),methyl hexanoate (7) and butyl butyrate (8). (C) Two S3A ORNs(numbers 19 and 31) show the repeatability of the responses of acertain ORN type to a particular extract, here a pineapple volatilecollection, where ethyl hexanoate (4) is again the major stimulus,while methyl hexanoate (7) elicits a weaker response.

720

methyl hexanoate produced highly similar response curves, bothcompounds already eliciting responses at 100pg doses, whereasethyl butyrate and butyl butyrate required a 100-fold and a10000-fold increase, respectively, in dose to produce anyresponse (Fig. 6A). We also confirmed the S4 and S7 classes;however, we were not able to obtain reliable dose–responsecurves for any of their neurones.

Behavioural effect of the identified ligandsWe screened the behavioural response to a set of the

identified odorants over a range of concentrations in a T-mazesetup (Fig. 7). From the identified ligands, we chose to testacetoin, butyl butyrate, ethyl hexanoate, ethyl 3-hydroxybutyrate, 1-hexanol and phenylacetonitrile, allpotential key ligands for six of the identified ORN classes.

M. C. Stensmyr and others

A NR

B

Acetoin 2,3-butanediol112 67 65

C CO2

D

FurfuralN=32 34

S2 A NR

B

Ethyl 3-hydroxybutyrate Ethyl butyrateN=13 150 90

S3 A

Ethyl hexanoate Methyl hexanoate Isoamyl acetate134 58 30

N=24 B NR

S4 A

Butyl butyrate 1-Hexanol56 30

B

Isoamyl acetate Butyl acetateN=4 104 25

S5 A

1-hexanol170

B

Isoamyl acetate Sec-amyl acetateN=6 142 107

A NR

B

Sec-amyl acetateN=1 58

S7 A

Isoamyl alcohol128

N=3 B NR

A NR

B

Phenylacetonitrile Phenylacetaldehyde Phenylethyl alcohol58 42 36

N=4 C NR

S1

Ethylidene acetone34

Cyclohexanol

80

Isovaleric acid17

S6

S8

Hexyl acetate102

O

O

OH

O

OOH

O

O

O

O

OH

O

O

O

O

O

O

O

O

O

O

O O

OH

OH

OH

OH

O

O

O

O

O

O

O

O

OH

OH

O

O

O

O

O

O

O

OOH

O

O

O

O

HO

OO

H

O

N OH

O

Ethyl butyrate Butyl butyrate Ethyl 3-hydroxybutyrate

35

Ethyl 3-hydroxyhexanoate30

Acetyl furan

52

Isoamyl acetate8

Hexyl N-butyrate74

45

Fig. 4. Identified classes of antennal sensilla based on physiological characteristics of olfactory receptor neurones (ORNs) present. Eachsensillum type (S1–S8) is presented together with the response spectra of the respective ORNs (A, B, C and D) housed within each sensillumtype. Beneath the sensillum type the number (N) encountered is indicated. Beneath the molecular configuration of each ligand type themaximum response is indicated (spikes s–1). NR, no response to any tested stimulus.

721Natural olfactory ligands in Drosophila

Although not identified in our study, ethyl acetate was alsoincluded. T-mazes vary individually, as well as the conditionsunder which they are operated (e.g. the method of applying thestimuli and the stimulus amount and temperature), thus weincluded ethyl acetate as a reference for the RIfunctions of thenovel ligands, as the behavioural effect of this odorant in T-mazes is well known (Ayyub et al., 1990; Alcorta, 1991;Acebes and Ferrús, 2001).

In total, we examined the behavioural response of 7120 fliesin a total of 392 experiments. We did not observe anydifference between the sexes, thus the data were pooled. All ofthe tested odorants were attractive, except 1-hexanol, whichhad an increasingly negative RI throughout the wholeconcentration range tested. The attractive odorants were onlyattractive over certain concentration ranges, and the overallresponse pattern followed the same trend with indifference tolow concentrations, attraction to intermediate concentrationsand repulsion to high concentrations. Butyl butyrate, ethyl 3-hydroxybutyrate and phenylacetonitrile had a very narrowconcentration range in which the RIwas positive, whereas theother attractive compounds were attractive over a wider range.The peak RIvalues for the attractive compounds are in thesame range as we found for ethyl acetate and suggest that thesenovel ligands are equally potent Drosophilaattractants as ethylacetate.

DiscussionThe fruitfly D. melanogaster antenna houses highly sensitive

ORNs physiologically tuned to detect odours emitted bypotential food sources. By using a novel technique, combininga chemical and a biological detector, a number of novel odourligands were identified. These ligands were subsequentlyshown to elicit attraction or repulsion in the fly. The ORNs caninform the animal that food is present in the vicinity and theycan also signal the quality of the food.

In several of the model systems used in olfactory researchtoday, very few biologically relevant key odour stimuli areknown. Investigations often rely on synthetic odourcompounds chosen randomly from the laboratory stock. Anadvantage for researchers investigating insects has traditionallybeen that behaviourally relevant odour ligands have indeedbeen identified and used. Unfortunately, this has not been thecase for one of the main insect model systems, the fruitfly D.melanogaster. In the present investigation, we set out toidentify natural odour ligands for this species. By collectingodours from potential food sources and using the insect itselfas a detector, we found ligands that were singled out by ORNsamong a very large selection of molecules extracted. In extractstypically containing hundreds of components, responses weregenerally recorded to a few compounds, and never to more thaneight. The fly thus seems to single out key compounds in the

LB

Med

ial

Posterior Anterior

ST

SA

ST

A B

S3 (24)

S1 (32)

S2 (13) S5 (6)

S4 (4)

S8 (4)

S6 (1) Unidentified (7)

S7 (3)

LB

Fig. 5. Antennal distribution of the sensillum types identified. (A) Frontal view of a Drosophilahead. The olfactory organs, antennae (above)and maxillary palps (below), are marked in green. (B) Schematic drawing of the Drosophila third antennal segment’s posterior and anteriorface. LB, large s. basiconica region; A, arista; S, sacculus entrance; ST, s. trichodea region. The locations of the different sensillum typesinvestigated are indicated by coloured circles (corresponding to the colours in Fig. 4). The total number of each sensillum type encountered isindicated in parentheses. Circles with a black dot in the centre represent investigation with GC–SC methodology, while open circles werecharacterised by stimulation with synthetic compounds.

722

extracts. Subsequent screening with the novel ligands assynthetic compounds showed that the neurones were capableof detecting key ligands at very low concentrations.Furthermore, the compounds were behaviourally active.

The odours identified reflect the food preference ofDrosophila, namely rotten fruit. We found, for example, ORNsdetecting microbial volatiles (e.g. acetoin and isoamyl alcohol)and ORNs detecting typical fruit volatiles (e.g. ethyl hexanoateand isoamyl acetate). Accordingly, of the six novel ligandstested for behavioural effect, five were attractive. The onlyrepellent odour, 1-hexanol, is a so-called green leaf volatile,i.e. characteristic of green plant tissue and unripe fruit. Being

an indicator of unsuitable food sources, this may explain whyfruitflies avoid 1-hexanol. Considering both the highlysensitive receptor neurones, physiologically tuned to detect thecompounds identified here, and the behavioural responsesrecorded, we conclude that the odours identified are indeed keysensory stimuli used to identify and locate suitable foodsources of the fruitfly.

In the dataset presented here for Drosophila, different degreesof specificity are observed. A general problem in investigationsof olfactory functions is the fact that you can never test thecomplete odour set. However huge the stimulus set used, anatural comment will be: “how can you be sure that you have

not missed the key stimulus?” The potential stimulusspectrum for an ORN is unlimited so the answer tothe question must be that a logically selectedstimulus spectrum was used. Using the GC–SCmethod we can approach the ‘real’ world as we areactually testing hundreds of odours at their naturallyoccurring proportions. In exchange, we have tosacrifice the absolute control of concentration as allcompounds will occur in the amounts found in theextracted stimulus medium, in this case fruit. Toremedy the lack of threshold information anddose–response curves, we performed stationaryinvestigations of a large number of ORNs to establishthresholds and dose dependencies. From these it wasclear that the ORNs investigated are indeed highlysensitive to the ligands identified.

How do our results correlate with previousinvestigations of Drosophila olfaction? de Bruyne etal. (2001) characterized the physiological specificityof ORNs housed in antennal basiconic sensilla usingsynthetic stimuli. Based on the response profiles ofthe ORNs, they reported seven distinct types ofsensilla. The three types of sensilla we found in theproximo-medial region display similarities indistribution as well as in ORN tuning with the larges. basiconica described by de Bruyne et al. However,using natural stimuli, we report other, albeit similar,key ligands. For example, we found acetoin was aprimary ligand for the S1B neurone, whereas thecorresponding ORN in the de Bruyne et al. studyresponded to 2,3-butanedione, a structurally similarcompound. Acetoin is a good key-ligand candidatefor this ORN type. It is effective at lowconcentrations (Fig. 6) and makes ecological senseas a ligand. The primary stimulus we identified forthe S2B ORN, ethyl 3-hydroxybutyrate, is also agood key-ligand candidate. It produced strongresponses and was effective at very lowconcentrations (Fig. 6B), whereas hexanol and ethylbutyrate only elicited moderate responses from whatprobably constitutes the corresponding ORN type inthe de Bruyne et al. study.

How do insects code odours? Pheromonedetection is widely accepted to rely on highly

M. C. Stensmyr and others

N=100

20

40

60

80

100

120

140

160 Ethyl hexanoateMethyl hexanoateEthyl butyrateButyl butyrate

N=30

20

40

60

80

100

120

140

160

–1 0 3

Ethyl 3-hydroxybutyrate

N=20

20

40

60

80

100

120

140

160

Stimulus load (log ng)

PhenylacetonitrileE

N=40

20

40

60

80

100

120

140

160 1-Hexanol

Stimulus load (log ng)

N=60

20

40

60

80

100

120

140

160 AcetoinC

A

Res

pons

e (s

pikes

s–1)

Res

pons

e (s

pikes

s–1)

Res

pons

e (s

pikes

s–1)

D

B

21 4–1 0 321 4

–1 0 321 4–1 0 321 4

–1 0 321 4

Fig. 6. Dose–response relationshipsfor five of the identified olfactoryreceptor neurone (ORN) classes.Odorants were presented inincreasing dosages from 0.1 ng to10µg (x-axis). ORN response to astimulus is given as the net numberof action potentials (number ofspikes during 1 s after stimulationminus the number of spikes during1 s prior to stimulation) minus thenet blank response (y-axis).(A) Dose–response curves for the

S3A neurone type, (B) S2B type, (C) S1B type, (D) S5A type and (E) S8Btype. For details on each ORN type, see Fig. 4. Vertical bars indicate the S.E.M.

723Natural olfactory ligands in Drosophila

selective and sensitive ORNs. In recent investigations it has alsobeen shown that plant odour-detecting ORNs can matchpheromone ORNs with respect to selectivity and sensitivity (e.g.Dickens, 1990; Anderson et al., 1995; Hansson et al., 1999;

Stensmyr et al., 2001). Both these types of neurones will alsorespond to compounds of structural similarity but generally at ahigher concentration than that required to elicit the samemagnitude of response by the key ligand(s). Thus, pheromonesand general odours in insects are likely to be coded along similarprinciples, i.e. the ORNs are primarily tuned to one, or maybe ahandful, of chemically similar key ligands that elicit responsesat very low concentrations. However, the ORs present on theORNs are not only capable of binding these key ligands, as theORN will also respond to other compounds if these are presentedin high enough dosages and if they are of structural proximity.No ORN has so far been identified that will respond to a singlecompound only, irrespective of concentration, but the responsethreshold will differ for the compounds depending on theirinteractions with the receptor proteins of the ORN. In mothpheromone-detecting ORNs, the decrease in activity whenmoving from the key ligand to a structurally similar one has beenshown to be directly proportional to the conformational energyneeded to fold the ‘suboptimal’ ligand into a conformation mostclosely mimicking the key molecule (Gustafsson et al., 1997).

How does Drosophilacode odours? We challenged the ORNswith a very large number of odorants (probably in the range of1000) extracted from natural fly resources, and out of all theseodorants, only 27 compounds, a fraction of all screened, eliciteda response. Five ORN types responded only to singlecompounds. ORNs that did respond to several compounds wereprimarily stimulated by compounds that shared structuralproperties. The S3A, S4B, S5B and S8B ORN classes were eachtriggered by molecules sharing a functional group as well as

–0.75

–0.5

–0.25

0

0.25

0.5

0.75

–2 –1 0 2 4

Acetoin–0.75

–0.5

–0.25

0

0.25

0.5

0.75

Butyl butyrate

0

Ethyl hexanoate

Ethyl 3-hydroxybutyrate 1-Hexanol

0

Phenylacetonitrile

RI

Ethyl acetate

A C

D

G

*

*

** *

**

****

*

*

*

**

*

***

1 3 5

–0.75

–0.5

–0.25

0

0.25

0.5

0.75

–0.75

–0.5

–0.25

0

0.25

0.5

0.75

–0.75

–0.5

–0.25

0

0.25

0.5

0.75

Stimulus load (log ng)

–2 –1 0 2 41 3 5

–2 –1 0 2 41 3 5

–2 –1 0 2 41 3 5

–2 –1 0 2 41 3 5

–2 –1 0 2 41 3 5

–2 –1 0 2 41 3 5

Stimulus load (log ng) Stimulus load (log ng)

–0.75

–0.5

–0.25

0.25

0.5

0.75

–0.75

–0.5

–0.25

0.25

0.5

0.75

F

B

E

RI

RI

Fig. 7. Behavioural responses to a set of the identified ligands. Flybehaviour was tested in a T-maze setup and odorants were presentedover a wide concentration range. The behavioural response functionsare based on 392 experiments with a total of 7120 flies. A responseindex (RI) of 1.0 equals full attraction, whereas an RI of –1.0 equalsfull avoidance. Indifference to the odour is indicated by an RI of 0.(A) RI curve for acetoin (N=89, number of experiments used tocreate the RI curve), (B) butyl butyrate (N=43), (C) ethyl hexanoate(N=57), (D) ethyl 3-hydroxybutyrate (N=52), (E) 1-hexanol (N=51),(F) phenylacetonitrile (N=46), (G) ethyl acetate (N=48). Asterisksindicate RI values significantly different from zero (P<0.05, χ2 test).Vertical bars indicate the S.E.M.

724

being similar in the overall hydrocarbon structure. Of the ORNsthat responded to multiple odorants, in the majority of cases (seeORNs S1B, S2B, S3A, S4B and S5B) one of the ligandsproduced a significantly stronger response than the alternateligands. Naturally, the response magnitude as well as the ligandspectrum of the ORNs is affected by the stimulus amountapplied. However, in many cases the ligand eliciting thestrongest response in an extract did not occur in the highestamounts. The response curve of the S3A ORN obtained throughscreening with synthetic compounds over a wide concentrationrange shows that these ORNs are primarily configured for twovery similar esters, and that alternate ligands require muchhigher doses in order to elicit response. Generally, thedose–response relationships show a very low detection level forseveral of the ligands and indicate a high degree of sensitivity.

The fruitfly thus seems to code odour molecules in a fashionthat is presently emerging for several different insect types. Atlow concentrations, detection is performed with an arguably highselectivity. When concentrations are increased the specificity isdecreased. The specificity of the system might thus vary withdistance to the source. When considering the range ofconcentrations that a fruitfly will move through, such a systemcan be envisaged as highly appropriate. Flies need to detect foodsources at a distance, i.e. they need to detect some key odourswith very high sensitivity. As the fly moves towards the foodsource it will experience a concentration gradient from very lowconcentrations to more or less saturation close to or on thesubstrate. Most likely key ligands are detected first, at a distance,and as the fly moves closer and closer the threshold for severalother molecules is reached and the ‘distance-specialist’ becomesa ‘close-range-generalist’. An initial attraction by selective ORNsis gradually transformed into an across-fibre coded, detailedodour image. Discussions of specificity will always be a matterof concentration. As mentioned above, not even pheromonereceptors display an absolute specificity (Peterlin et al., 2002).

We would like to express our gratitude to Dr ChristerLöfstedt for support and help on chemistry as well as to DrsJohn Carlson and Marien de Bruyne for help on Drosophilaelectrophysiology. This work was supported by grants fromthe Swedish Research Council. E.G.’s and A.B.’s visit inSweden was sponsored by a European Union SOCRATESexchange grant.

ReferencesAcebes, A. and Ferrús, A.(2001). Increasing the number of synapses

modifies olfactory perception in Drosophila. J. Neurosci.21, 6264-6273.Adams, M. D., Celniker, S. E., Holt, R. A., Evans, C. A., Gocayne, J. D.,

Amantides, P. G., Scherer, S. E., Li, P. I., Hoskins, R. A. and Gaile, R.F. (2000). The genome sequence of Drosophila melanogaster. Science287,2185-2196.

Alcorta, E. (1991). Charcterization of the electroantennogram in Drosophilamelanogaster and its use for identifying olfactory capture and transductionmutants. J. Neurophysiol.65, 702-714.

Andersson, P., Hansson, B. S. and Löfqvist, C.(1995). Plant-odour-specificreceptor neurones on the antennae of female and male Spodoptera littoralis.Physiol. Entomol. 20, 189-198.

Arn, H., Städler, E. and Rauscher, S.(1975). The electroantennographicdetector: a selective and sensitive tool in the gas chromatographic analysisof insect pheromones. Z. Naturforsch.30, 722-725.

Ayyub, C., Paranjape, J., Rodrigues, V. and Siddiqi, O.(1990). Genetic ofolfactory behavior in Drosophila melanogaster. J. Neurogenet.6, 243-262.

Clyne, P., Grant, A. J., O’Connel, R. J. and Carlson, J. R.(1997). Odorantresponse of individual sensilla on the Drosophilaantenna. Invert. Neurosci.3, 127-135.

Clyne, P. J., Warr, C. G., Freeman, M. R., Lessing, D., Kim, J. andCarlson, J. R. (1999). A novel family of divergent seven-transmembraneproteins: candidate odorant receptors in Drosophila. Neuron22, 327-338.

de Bruyne, M., Clyne, P. J. and Carlson, J. R.(1999). Odor coding in amodel olfactory organ: the Drosophilamaxillary palp. J. Neurosci.19,4520-4532.

de Bruyne, M., Foster, K. and Carlson, J. R. (2001). Odor coding in theDrosophilaantenna. Neuron30, 537-552.

Gao, Q. and Chess, A.(1999). Identification of candidate Drosophilaolfactory receptors from genomic DNA sequence. Genomics60, 31-39.

Gustafsson, A.-L., Tuvesson, M., Larsson, M. C., Wenqi, W., Hansson, B.S. and Liljefors, T. (1997). Bioisosteric approach to elucidation of bindingof the acetate group of a moth sex pheromone component to its receptor. J.Chem. Ecol.23, 2755-2775.

Hansson, B. S., Larsson, M. L. and Leal, W. S.(1999). Green leaf volatile-detecting olfactory receptor neurones display very high sensitivity andspecificity in a scarab beetle. Physiol. Entomol. 24, 121-126.

Helfand, S. and Carlson, J.(1989). Isolation and characterization of anolfactory mutant in Drosophilawith a chemically specific defect. Proc. Natl.Acad. Sci. USA 86, 2908-2912.

Hubel, D. H. (1957). Tungsten microelectrode for recording from single units.Science125, 549-550.

Peterlin, Z. A., Rogers, M. E., Chesler, A. T. and Firestein, S. J.(2002).Dose dependent responses of Manducta sexta“pheromone-specific”olfactory neurons to general odorants. AChemS XXIV (Abstr.). Sarasota,USA: Association for Chemoreceptor Sciences.

Rubin, G. M., Yandell, M. D., Wortman, J. R., Gabor Miklos, G. L.,Nelson, C. R., Hariharan, I. K., Fortini, M. E., Li, P. W., Apweiler, R.and Fleischmann, W. (2000). Comparative genomics of the eukaryotes.Science287, 2204-2216.

Shanbhag, S. R., Müller, B. and Steinbrecht, R. A.(1999). Atlas of olfactoryorgans of Drosophila melanogaster1. Types, external organization,innervation and distribution of olfactory sensilla.Int. J. Insect Morphol.Embryol.28, 377-397.

Siddiqi, O. (1991). Olfaction in Drosophila. In Chemical Senses, vol. 3,Genetics of Perception and Communications (ed. C. J. Wysocki and M. R.Kare), pp. 79-96. New York: Dekker.

Stensmyr, M. C., Larsson, M. C., Bice, S. and Hansson, B. S.(2001).Detection of fruit and flower-emitted volatiles by olfactory receptor neuronsin the polyphagous fruit chafer Pachnoda marginata(Coleoptera:Cetoniinae). J. Comp. Physiol. A187, 509-519.

Stocker, R. F.(2001). Drosophilaas a focus in olfactory research: mappingof olfactory sensilla by fine structure, odour specificity, odorant receptorexpression, and central connectivity. Microsc. Res. Techniq.55, 284-296.

Störtkuhl, K. F. and Kettler, R. (2001). Functional analysis of an olfactoryreceptor in Drosophila melanogaster. Proc. Natl. Acad. Sci. USA 98, 9381-9385.

Tully, T. and Quinn, W. G. (1985). Classical conditioning and retention innormal and mutant Drosophila melanogaster. J. Comp. Physiol. A 157, 263-277.

Vosshall, L. B.(2001). The molecular logic of olfaction in Drosophila. Chem.Senses26, 207-213.

Vosshall, L. B., Amrein, H., Morozov, P. S., Rzhetsky, A. and Axel, R.(1999). A spatial map of olfactory receptor expression in the Drosophilaantenna. Cell 96, 725-736.

Wadhams, L. J.(1982). Coupled gas chromatography – single cell recording:a new technique for use in the analysis of insect pheromones. Z.Naturforsch.37C, 947-952.

Wetzel, C. H., Behrendt, H.-J., Gisselmann, G., Störtkuhl, K. F.,Hovemann, B. and Hatt, H. (2001). Functional expression andcharacterization of a Drosophilaodorant receptor in a heterologous cellsystem. Proc. Natl. Acad. Sci. USA 98, 9377-9380.

Wibe, A., Borg-Karlson, A. K., Norin, T. and Mustaparta, H. (1997).Identification of plant volatiles activating single receptor neurons in the pineweevil (Hylobius abietis). J. Comp. Physiol. A180, 585-595.

M. C. Stensmyr and others