unsuccessful courtship decreases subsequent courtship [1
]. When the initial courtship object (trainer) is a virgin female, suppression has been shown to be the result of formation of an associative memory linking the failure to copulate with volatile stimulatory courtship cues specific to the age of the female trainer [1
]. Exposure to a mated female, on the other hand, results in a suppression of courtship toward all types of females [1
] and is believed to require an aversive pheromone [3
]. The cuticular hydrocarbon profiles of mature and immature females differ significantly [1
], but these types of females also differ behaviorally. Mature virgins are receptive to courtship, while immature virgins and mated females show characteristic rejection behaviors. Immature females kick, fend and run away, while mated females extrude their ovipositors [6
]. To determine if female behavior or appearance had any role in the development of age-specific or general courtship suppression, we trained and tested males with decapitated females in dim red light (). Memory index was expressed as a ratio of the courtship index (CI) during the 10 min test period to the mean CI of a sham-trained males tested with the same type of female. The use of a ratio allows direct comparison of the strength of memory between conditions, with a value of CItest
= 1 indicating no memory.
Courtship suppression learned with a mated female trainer is generalized to all types of females
We find that, consistent with results with mobile trainer females [1
], decapitated virgins provoke an age-specific suppression, while decapitated mated female trainers cause general suppression of courtship. These data indicate that the specificity of learning with different trainer types does not stem from behavioral differences in the trainer female’s response to courtship or from visual cues specific to the trainer type. Generalization of learning with a mated female trainer is therefore the result of chemosensory cues. In all subsequent experiments, decapitated trainers and testers were used, except where noted.
In the previous experiment, males were placed in the same chamber as the trainer female and therefore could obtain both olfactory and gustatory information about that female. To investigate the nature of the generalization cue, we attempted to reconstitute generalized learning with virgin trainers and mated female extracts. Placing a filter containing a hexane extract of mated female in the chamber with either a mature or immature trainer female caused a generalization of learning, as demonstrated by the ability of mature trainers to generate memory against immature testers and vice versa
(Supplemental Table 1
). To determine if the active component of the mated female bouquet was volatile, we used a two compartment courtship chamber () and placed a pheromone source (fly corpse or filter with extract) across a mesh from the side of the chamber containing the male and the trainer female. For both mature and immature trainers, the presence of volatile compounds from either a mated female or a male was sufficient to cause generalization of courtship suppression, although the effects of these pheromones appeared more potent with mature trainers (). In the absence of a courtship object, the presence of a filter with extract (Supplemental Table 1
) or a corpse () did not generate suppression of courtship toward tester females.
The generalizing cue is a volatile pheromone
We next addressed the identity of the generalization cue. The mated female and mature virgin trainers we use are of the same age (4–5 days old), and might be expected to have similar cuticular hydrocarbon profiles, so any compound that differed between these two classes of females might have a role in generalization. We compared hexane washes of 4–5 day old virgins and 4–5 day old mated females that had been mated 24 h before extraction, using gas chromatography-flame ionization detection and mass spectrometry (, top two traces). Qualitatively, the two types of females appear identical with the exception of one peak, cis-
vaccenyl acetate (cVA), which is undetectable in virgins but present at significant levels in mated females. cVA is a major component of mature male cuticular hydrocarbon (, lower trace) and is not synthesized by females [8
]. Its presence in both males and mated females makes it a good candidate for being the generalization cue for courtship learning.
Cuticular hydrocarbons of mature virgins and mated females differ in levels of cis-vaccenyl acetate
Quantitation of mature virgin and mated female hydrocarbon levels shows a significant difference in cVA levels (). There is also a small, but statistically insignificant (P > 0.05) increase in 7-tricosene (7-C23:1). 7-tricosene is a major component of the male cuticular hydrocarbon and believed to have inhibitory effects on male-male courtship [10
]. Transfer of 7-tricosene to females has been shown to occur via cuticular contact during copulation, but it is largely gone by 8 h after mating [11
]. Consistent with this, we see larger amounts of 7-tricosene on virgins that have been courted, but not copulated, when they are extracted immediately after the courtship (, left panel). Mature virgins and mated females that have been aged 24 h after copulation have lower and statistically indistinguishable levels of 7-tricosene. The loss over time (presumably through passive transfer and grooming) of 7-tricosene and the non-volatile nature of this compound make it an unlikely candidate for the generalizing cue. It is also significant to note that we do not see any decreases in mated females of hydrocarbons such as 7,11-heptacosadiene (7,11-nC27:2), 7,11-nonacosadiene (7,11-nC29:2) and 9-pentacosene (9-C25:1) that are believed to be stimulatory pheromones for courtship conditioning [12
]. Thus the only consistent mated female-specific difference in hydrocarbon content we find is in cVA.
How does cVA, a male lipid, become part of the mated female pheromonal profile? Like 7-tricosene, cVA could be transferred directly by contact during courtship and/or copulation. Alternatively, the presence of cVA in the male ejaculatory bulb suggests that it can be transferred with sperm during copulation [8
]. To determine the major mode of cVA transmission, we measured cVA levels on virgin females, virgin females that were courted in a small chamber and extracted immediately, and females that were extracted 24 h after complete (>14 min) copulation or disrupted (≤2 min) copulation. We find that only females that copulated long enough to receive ejaculate [14
] have significant levels of cVA (, right panel). Females that did not copulate and were merely courted by the male had virtually no cVA, even though they had significant amounts of passively acquired 7-tricosene (, left panel). This suggests that transfer of cVA occurs via ejaculate and that mated females store cVA.
These data support a role for cVA as a generalizing cue, but the presence of other volatile compounds in mated female and male extracts might still be required. To test the sufficiency of cVA, we applied varying amounts of purified cVA to filters across the mesh in a two compartment courtship chamber, trained males with either mature or immature virgins, then tested with a virgin of the other age. In both cases, cVA was sufficient to generalize memory (). With mature virgin testers, 0.2 ng of cVA was enough to generalize memory (Supplemental Table 2
). The average amount of cVA present on a mated female 24 h after mating is 9.3 ± 3.5 ng, so the effects of synthetic cVA are occurring in the biologically relevant dose range. In contrast to results with mature virgins, pairing cVA with immature trainers is less effective. Only large amounts of cVA (200 μg) produce generalized learning. cVA alone (with no trainer female) is ineffective, as is cis-
vaccenol (cVOH), a putative metabolite of cVA [15
] with either trainer type (Supplemental Table 2
The circuitry underlying generalization is of great interest for understanding this behavior. As a first step, we sought to identify the olfactory receptor neurons that carry the aversive cVA signal. cVA has been shown to be sensed by a subset of trichoid sensilla [16
] in the Drosophila
antenna, which includes the T1 type sensillum [18
] which expresses Or67d
]. Using the ‘empty neuron’ preparation [20
], which allows the decoding of odor specificity for Drosophila
olfactory receptors (ORs), we found that there is an additional cVA-responsive receptor, Or65a, and that Or65a and Or67d differ in their response to cVOH, with Or67d responding strongly and Or65a not responding (van der Goes van Naters and Carlson, submitted). Or65a is one of the several ORs expressed in neurons of the T3 sensilla [19
Using this information, we investigated the role of the olfactory receptor neurons that express these two receptors in sensing the aversiveness of cVA. Initial courtship levels provide a simple assay for this property of cVA. Naive males show lower levels of courtship toward mated females than toward virgins of the same age [11
]. This effect can be reproduced by addition of a cVA-laced filter across the mesh in the two compartment courtship chamber with a mature virgin courtship object in the upper chamber with the male (). Expression of tetanus toxin (TNT), which blocks synaptic release, under control of Or65a-GAL4
, but not Or67d-GAL4
abolished the ability of cVA to inhibit initial courtship (). Males heterozygous for Or65a-GAL4
or the UAS-TNT
transgene showed cVA-dependent courtship suppression as did males expressing inactive toxin (TNT-VA) under control of Or65a-GAL4
. These results indicated that ORNs expressing Or65a-GAL4
, but not Or67d-GAL4
, are required for sensing cVA as an aversive cue.
Modulation of courtship by cVA requires Or65a-GAL4, but not Or67d-GAL4 olfactory sensory neurons
has been reported to be expressed solely in ORNs that innervated the DL3 glomerulus of the antennal lobe using anti-GFP immunohistochemistry in animals expressing GFP under control of Or65a-GAL4
promoter fusions. [19
]. Our Or65a-GAL4
, while it has strong expression in DL3, shows a somewhat broader pattern, with significant expression in VA1, DC1 and DA4m (Supplemental Figure 1
). Using confocal microscopy to directly visualize GFP from a UAS-mCD8-GFP
transgene in unfixed brains, we compared our GAL4 line to the published Or65a-GAL4
lines. We found that our line was many times stronger than that published by Fishilevich et al. [25
], which has predominant expression in DL3 (Supplemental Figure 2
). GFP fluorescence in the Couto et al. [19
] GAL4 line was barely detectable (data not shown). To determine if the weak, but more DL3-specific Fishilevich et al. driver (which we designate as V-Or65a-GAL4
in ) would also block cVA effects, we used it to express active and inactive tetanus toxin. Consistent with the results using our Or65a-GAL4
, active tetanus toxin significantly abrogated the ability of cVA to suppress initial courtship (), although the effect appeared weaker than with our line. Inactive tetanus toxin had no effect on cVA-mediated suppression. We conclude that the aversive effects of cVA on initial courtship are most likely mediated by ORNs expressing Or65a.