Insects navigating toward an odor can compare signals from their two antennae, and generally turn in the direction of the stronger signal. Confounding this strategy (by crossing and fixing the antennae, thereby spatially reversing them) impairs olfactory navigation in bees, ants, and locusts6-8
resemble other insects in using this strategy3-5
, but they are unlike other insects in having mainly bilateral ORN projections. In the Drosophila
brain, both ipsi- and contralateral ORN axons synapse onto antennal lobe projection neurons (PNs)9
, and ipsi- and contralateral synapses have approximately the same strength10
. A minority of Drosophila
olfactory glomeruli receive unilateral ORN projections1,11
, raising the question of whether these unilateral glomeruli are what enable lateralization.
To investigate lateralization behavior, we built a spherical treadmill to measure olfactory behavior in walking Drosophila
( and Supplementary Fig. 1
). To prevent the head from moving, it was glued to the body, and the two antennae were independently stimulated with odor. When we delivered fermented peach volatiles to one antenna and clean air to the other, flies on the treadmill made a fictive turn toward the odor (). If odor or clean air was delivered symmetrically to the two antennae, the fly continued to walk straight ().
Figure 1 Drosophila can lateralize odors based on bilateral receptor neuron input to a single pair of glomeruli. a. Schematic diagram of a fly on a spherical treadmill, with odor tubes directed at each antenna. b. Representative running trajectories from a single (more ...)
Fruit volatiles typically activate multiple ORN types12
. Therefore, we next used a monomolecular odor (pentanoic acid) to more specifically target a glomerulus that receives bilateral ORN input. This odor has just one known high-affinity receptor in the antenna11,12
. This receptor corresponds to glomerulus DM6, which receives bilateral ORN innervation2,9
. Pentanoic acid elicited turning behavior as robust as that elicited by peach volatiles (). Turning was significantly reduced by a mutation (Orco2
) which silences bilateral ORNs, including the DM6 ORNs (). The ORN types that project unilaterally do not express the Orco
, and so this result implies that asymmetric input to strictly bilateral glomeruli can produce turning.
To target a single bilateral ORN type more selectively, we expressed channelrhodopsin-2 (ChR2) in the ORNs that express the olfactory receptor Or42b and project to glomerulus DM1 (). These ORNs fired a burst of 3 or 4 spikes during a 50-ms light pulse directed at the antenna with a fine fiber optic filament (). In the same fly, light had no effect on adjacent ORNs that did not express ChR2 (). Illuminating one antenna produced a turn toward the stimulus (; Supplementary Figure 2
). Conversely, light offset produced a compensatory turn in the other direction (, Supplementary Information Footnote 1
). The turns evoked by ChR2 in one ORN type were as large as when ChR2 was expressed in most ORN types (under the control of the Orco-Gal4
line; ). Thus, bilateral ORNs can support lateralization behavior, and turning can be elicited by just a few spikes per neuron in one ORN type.
The temporal control enabled by the optogenetic approach allowed us to precisely determine the latency of the behavioral response. Surprisingly, turning began within ~ 70 ms of the ORN response onset (). This is faster than the fly's stride period14
(about 100 ms, Supplementary Fig. 1
). Thus, there must be a rapid mechanism in the brain for extracting lateralized information from sensory neurons.
We next asked how asymmetric odor stimuli are encoded at the level of PN spikes. We removed one antenna in order to lateralize the odor stimulus, and we made simultaneous cell-attached recordings from PNs ipsi- and contralateral to the intact antenna (). We recorded from pairs of PNs in the same glomerulus on different sides of the brain (“sister PNs”), using GFP to target our electrodes to PNs in glomerulus DM6 or DM1. We used odors that preferentially activate either DM6 ORNs or DM1 ORNs (pentanoic acid or dilute ethyl acetate12
). In these experiments, we found a small but consistent asymmetry in PN odor responses. This was apparent in the latency to the first odor-evoked spike: the first ipsilateral spike occurred 2.47±0.70 ms earlier than the first contralateral spike in DM6 PNs, and 1.01±0.41 ms earlier in DM1 PNs (n
=16 and 6, ). The latency difference was statistically significant for DM6 PN pairs, although it fell short of significance for DM1 PN pairs (p
< 0.005 and p
= 0.06, Wilcoxon signed-rank tests). In addition, we found an asymmetry in odour-evoked firing rates: ipsilateral firing rates were on average about 50% higher than contralateral firing rates for DM6 PNs, with a larger asymmetry for DM1 PNs (). The asymmetry was observed even during the spontaneous firing of these cells, and it was proportionately similar for all odour concentrations (). The asymmetry in spikes must be due to an asymmetry in synaptic currents. Indeed, asymmetric stimulation of the antennae produced systematically larger inward currents in ipsi- versus contralateral PNs (Supplementary Fig. 3
Figure 2 Lateralized odors produce an asymmetry in spike latency and spike rate in antennal lobe projection neurons. a. Schematic of simultaneous cell-attached recording configuration. An ORN axon innervates a glomerulus on each side of the midline, where it synapses (more ...)
Synaptic currents in PNs reflect the combined effects of ORNs and local neurons (LNs), most of which are GABAergic15-17
. We first asked if the asymmetry in PN activity requires GABAergic inhibition, by bath-applying GABAA
receptor antagonists. These antagonists elevated PN firing rates, but the difference between ipsi- and contralateral PN firing rates was not significantly altered ().
Figure 3 The asymmetry arises at the level of ORN-to-PN synapses. a. Mean DM6 PN responses to pentanoic acid (10-2) in normal saline. b. Same as (a) after adding the GABAA receptor antagonist picrotoxin (5 μM) and the GABAB receptor antagonist CGP54626 (more ...)
Our results suggest that the asymmetry originates at the level of ORN input to PNs. To compare ORN-to-PN synapses in ipsi- versus contralateral PNs, we made simultaneous whole-cell recordings of spontaneous excitatory postsynaptic currents (sEPSCs) in pairs of sister PNs. These sEPSCs are known to arise from ORN-to-PN synapses9,10
. Each ORN-to-PN synapse consists of many release sites with high release probability, and so spike-evoked synaptic events are large and reliable10
. In recordings from sister PNs in glomerulus DM6, virtually all sEPSCs occurred in a paired fashion (). Almost 100% of the sEPSCs in one cell had a corresponding paired event in the other cell, and this percentage was not significantly different in ipsi- versus contralateral PNs (98.3 ± 0.6% versus 96.4 ± 1.3%, n
= 15 pairs, p
= 0.11, Wilcoxon signed-rank test). This result indicates that the number of unitary ORN-to-PN synaptic connections is essentially identical on the ipsi- and contralateral sides. This result also argues that action potential conduction failures do not occur to any appreciable degree in the contralateral axon branches of ORNs, although we cannot completely exclude the idea that this occurs at a low rate.
These dual whole-cell recordings allowed us to measure the mean difference in ipsi- versus contralateral sEPSC arrival times (0.80 ± 0.51 ms), which represents the delay imposed by axonal conduction between sister glomeruli. This delay is less than the contralateral lag in first spike latency in glomerulus DM6 (2.47 ms, see above). This makes sense, because the lag in first spike latency is the result of two delays: the delay imposed by axonal conduction, and the delay that results from less net synaptic excitation in the contralateral PN (Supplementary Fig. 3
) which means that the contralateral PN requires more integration time to reach spike threshold.
Comparing the amplitude of paired sEPSCs on an event-by-event basis, we found considerable variation in the relative amplitude of ipsi- versus contralateral EPSCs (). This is expected, given that synaptic vesicle release is a stochastic process occurring independently at ipsi- and contralateral synapses. When many sEPSCs were considered together, there was a consistent asymmetry in the amplitude of sEPSCs (). Averaged across all experiments, ipsilateral sEPSCs were 39% larger than their contralateral counterparts (). Thus, although ORN spikes reliably invade both ipsi- and contralateral axon branches, each ORN spike typically has a larger effect on ipsilateral PNs.
In principle, the asymmetry in sEPSC amplitudes could have either a pre- or postsynaptic locus. To determine if we can observe this asymmetry at the level of presynaptic release sites, we expressed synaptobrevin:GFP (a marker of presynaptic release sites) in DM6 ORNs. We removed one antenna and allowed three days for the cut ORN axons to degenerate, leaving only the axons from the intact antenna. We found that total synaptobrevin fluorescence was on average 41 ± 16% higher on the ipsilateral side, and the ipsi/contra ratio of synaptobrevin fluorescence was significantly greater than 1 (p
< 0.05, n
= 5, t
-test, ). This suggests an asymmetry in the number or size of neurotransmitter release sites. This is consistent with a previous study reporting that a plasma membrane marker was more abundant on the ipsilateral side18
. That study also severed one antenna and waited three days before imaging. Functional remodeling can occur over that time period18,19
, so both findings should be interpreted cautiously.
Figure 4 The asymmetry in ORN-to-PN synapses has a presynaptic origin. a. Synaptobrevin:GFP (green) was expressed in DM6 ORNs to label neurotransmitter release sites, and one antenna was removed 3 days prior to allow the cut axons to degenerate. The brain is viewed (more ...)
Next, to visualize presynaptic calcium, we expressed the genetically-encoded calcium indicator GCaMP3.0 in ORNs (using the pebbled-Gal4
line) and used two-photon microscopy to visualize calcium signals in ORN axon terminals. To lateralize the odor, we removed one antenna immediately prior to the experiment (). We used pentanoic acid to evoke a fluorescence increase preferentially in glomerulus DM6 (). We found that the size of the calcium response was significantly larger on the side of the brain ipsilateral to the intact antenna (). This asymmetry did not require feedback from central circuits, because it persisted after washing in mecamylamine (to block nicotinic acetylcholine receptors, which mediate ORN-to-PN synaptic transmission10
) along with picrotoxin and CGP54626 (). In separate experiments, we saw similarly asymmetric calcium signals in glomerulus DL5, using an odor stimulus which is relatively selective for DL5 ORNs20
dilution, data not shown). We also saw a similar asymmetry in presynaptic currents, using simultaneous bilateral field potential recordings (Supplementary Figure 4
). Together, these results demonstrate that the asymmetry in EPSC amplitudes has a presynaptic origin.
Interestingly, we found a roughly equal asymmetry (~40%) in the amplitude of sEPSCs, the level of synaptobrevin fluorescence, and the odour-evoked GCaMP3 fluorescence change. The simplest explanation for these results is that the ipsilateral arbor is 40% larger than the contralateral arbor. If everything else is equal, then this mechanism should produce a proportional change in all these measurements (Supplementary Information Note 2
Our results are consistent with the finding that, in flies with one antenna removed, 2-deoxyglucose uptake is higher on the side ipsilateral to the intact antenna21
. A previous study using calcium imaging has reported that asymmetric ORN stimulation can produce stronger signals in ipsilateral PNs than in contralateral PNs22
, which is also broadly consistent with our results. In contrast to our results, however, this asymmetry was restricted to just a few glomeruli. This finding is puzzling, given that a variety of stimuli can produce turning behavior ( and refs 3-5
). The same study also concluded that GABAergic inhibition mediates the ipsi/contra asymmetry in PNs22
. These discrepancies may reflect limitations of the imaging techniques used in that study, as well as differences between stimuli or glomeruli.
Another previous study failed to find any significant ipsi-contra differences in the strength of ORN-PN synapses10
. Those measurements involved sequential recordings from sister PNs, rather than simultaneous recordings. Also, that study measured EPSCs arising from a single ORN per PN. Here, we have better statistical power because we made simultaneous recordings from sister PNs, and because we sampled sEPSCs arising from many ORNs. Our finding that ipsi- and contralateral synapses differ only modestly explains why this difference has been difficult to resolve previously.
It is well-known that a single axon can form neurotransmitter release sites with different properties onto different target cell types (e.g., principal neurons versus interneurons)23
. Here, we have shown that ORN axons discriminate between two classes of target cells (left and right) that are morphologically and functionally identical, and that share the same lineage and birth dates24
. This is reminiscent of the circuitry of the leech midbody ganglion, where individual mechanoreceptor axons make stronger synapses onto ipsi- versus contralateral sister cells that are otherwise functionally identical25
. How are these target cells distinguished? Right and left sister neurons might express different molecular tags26
, thereby allowing a given axon to recognize them as ipsi- versus contralateral to itself. Alternatively, an axon might form more release sites at proximal locations as compared to distal locations; this cell-intrinsic mechanism would suffice because contralateral sites are always more distal than ipsilateral sites.
Our results reveal that even small signals in the Drosophila
nervous system can be behaviorally relevant. The stimuli we used in our optogenetic experiments produced only a slight fractional difference in input to the right and left sides of the brain. This difference amounted to 5% over the 50-ms stimulus period (Supplementary Information Footnote 3
). The finding that this incremental difference is relevant for behavior should motivate the continued development of sensitive methods for monitoring neural activity in the fly brain.