|Home | About | Journals | Submit | Contact Us | Français|
Fruit flies make their living on the fly in search of attractive food odors. To maintain forward flight, flies balance the strength of self-induced bilateral visual motion  and bilateral wind cues , but it is unknown whether they use bilateral olfactory cues to track odors in flight. Tracking an odor gradient requires comparisons across two spatially separated chemosensory organs and has been observed in several walking insects [3–5], including Drosophila . The olfactory antennae are separated by a fraction of a millimeter, and most sensory neurons project bilaterally and symmetrically activate the first-order olfactory relay [7, 8], both of which would seem to constrain the capacity for bilateral sensory comparisons. Are fruit flies nonetheless able to track an odor gradient during flight? Using a modified flight simulator that enables maneuvers in the yaw axis , we found that flies readily steer directly toward a laterally positioned odor plume. This capability is abolished by occluding sensory input to one antenna. Mechanosensory input from the Johnston’s organ and olfactory input from the third antennal segment cooperate to direct small angle yaw turns up the plume gradient. We additionally show that sensory signals from the left antenna contribute disproportionately more to odor tracking than the right, providing further evidence of sensory lateralization in invertebrates [10–13].
In a wind tunnel, freely flying Drosophila surge upwind after contacting an attractive odor plume, and then show small amplitude oscillations in flight heading [14, 15] suggesting fine-scale steering control to remain within the plume boundaries. We therefore postulated that flies use bilateral comparisons of odor intensity to actively orient up a spatial odor gradient during flight. We used a magnetic tether system described previously that allows a fly to yaw freely and orient toward a narrow odor plume located on one side of a circular arena [9, 16]. Flies were visually “dragged” into an odor plume by oscillating a vertical stripe at the position of the plume, eliciting a robust visual fixation reflex [9, 17] (see methods). We then presented a high contrast panoramic visual stimulus to enhance plume tracking  (Fig. 1A) and recorded the fly’s movements relative to the plume. Fine scale inspection of the resulting flight trajectories reveals that flies perform frequent yaw deviations often lasting longer than one second (Fig. 1B, black arrows). While the mechanism for their initiation is unknown, an interesting feature of these deviations is that they are corrected – a fly may steer out of the plume but its subsequent turn is typically directed back into the plume.
Consistent with our previous findings , histograms of flight heading revealed that flies remained tightly centered within the plume (Fig. 1B) and despite the small yaw deviations, or saccades, showed an odor mediated reduction in mean saccade frequency (Fig. 1C). An analysis of the ratio of left turns to right turns (mean turning ratio, Fig. 1D, see methods) is consistent with the observation that yaw deviations are typically corrected - turning ratio equals zero indicating equal proportions of left and right turns (Fig. D). Finally, odor tracking is characterized by a significantly reduced deviation from the plume (Fig. 1E), measured as the total angular displacement from the plume during the duration of the trial . This measurement distinguishes between a fly that never leaves the plume (plume deviation = 0) and one that makes frequent yaw deviations (plume deviation > 0).
To test whether flies actively orient toward an odor stimulus, we visually “dragged” flies to the right (+90°, blue arrows) and left (−90°, green arrows) of an odor plume with an oscillating stripe, presented a high contrast visual stimulus, and recorded the resultant turning behavior (Fig. 1F). In each case, flies steered the shortest angle when orienting to the plume (Fig. 1G). Analysis of the mean turning ratio of saccades executed before entering the plume (see methods) confirmed that saccades were oriented directly toward the plume, regardless of the fly’s initial heading (Fig. 1H). We calculated the average time it took for the fly to enter the plume, or the time to plume acquisition (see methods), at roughly four seconds (Fig. 1I) and 100% of the tested flies acquired the plume from both directions (Fig. 1J). These results confirm that flies actively track an odor plume during flight, and do so by orienting saccades directly up an odor gradient.
To test the hypothesis that bilateral antennal comparisons mediate gradient tracking, we unilaterally and bilaterally occluded the olfactory sensilla on the 3rd antennal segment (a3) with non-toxic UV-cured glue (Fig. 2 A–C). The unilateral occlusion presumably precludes odor detection thereby eliminating bilateral spatial comparisons. Like control flies, animals with a single intact a3 executed frequent yaw deviations (Fig. 2A, B, lower right, black arrows). However, unlike control flies, these paired saccades frequently occurred outside the plume. Furthermore, odor presentation resulted in a significant bias in flight heading toward the intact a3 (Fig. 2A, B, lower left). Although right occluded flies exhibited an odor mediated reduction in saccade frequency, indicating that they indeed detected the odor (Fig. 2D), their turning ratio revealed that in the presence of odor, a majority of turns were leftward (Fig. 2E), resulting in a mean plume deviation that was statistically indistinguishable from the no odor control and higher than that of control flies (Fig. 2F). Unlike right occluded flies, left occluded flies did not show a significant reduction in saccade frequency in the odor treatment (Fig. 2D) but did exhibit a slight rightward turning bias toward the intact a3 (Fig. 2E). Although flies with an intact right a3 showed a significant odor mediated reduction in mean plume deviation (Fig. 2F), it was not reduced to the level of control flies. These results suggest that sensory input from a single antenna is insufficient for either spatial or temporal odor plume tracking, largely because odor activates turns away from the plume. Unilaterally occluded flies could neither maintain their spatial orientation within the plume, nor temporally integrate unilateral signals to locate the plume.
Flies with bilateral a3 occlusions (Fig. 2C) showed normal capacity to follow the visual object into the odor plume (Fig. 2C, lower left), but these flies did not track the odor plume any better than the no-odor controls (Fig. 2C, lower right, Fig. S2). When positioned in the plume, bilaterally-occluded flies showed no changes in saccade frequency (Fig. 2D), turning ratio (Fig. 2E), or mean plume deviation (Fig. 2F) indicating that these animals were completely insensitive to the odor stimulus. This suggests that bilateral a3 occlusions specifically silence odor detection, without generally perturbing flight behavior.
To assess the olfactory contribution to active plume tracking under more challenging conditions we positioned occluded flies +/−90° away from the odor plume. Both right and left occluded flies failed to orient toward the odor plume ipsilateral to the occluded a3 (Fig. 2G, H) as evidenced by a loss of saccades biased toward the odor stimulus (Fig. 2J), an increase in mean time to plume acquisition (Fig. 2K), and a decrease in the probability of plume acquisition (Fig. 2L). However, occluded flies were successful in orienting toward a plume ipsilateral to the intact a3 (Fig. 2G, H) resulting in control values of mean turning ratio (Fig. 2J), mean time to plume acquisition (Fig. 2K) and acquisition probability (Fig. 2L). As expected, bilaterally occluded flies failed to locate the plume from either direction (Fig. 2I), showed un-directed mean turning ratios (Fig. 2J), a significant delay in the mean time to plume acquisition (Fig. 2K), and reduced probability of acquiring the plume (Fig. 2L).
We can be confident that occlusion of a3 abolishes olfactory perception because bilaterally occluded animals failed to show any odor mediated changes in behavior. When presented with an odor (and only then), unilaterally occluded animals turn in the direction of the intact a3 (Fig. 2E) and fail to orient toward an odor ipsilateral to the occlusion (Fig. 2G, H, J). These results indicate that that bilateral olfactory comparisons facilitate plume re-acquisition during fly flight. Similarly, unilaterally and bilaterally antennectomized crayfish are unable to track a turbulent odor plume in water  and both humans  and Drosophila larvae  show enhanced olfactory tracking with bilateral olfactory input by comparison to unilateral input.
Control flies exhibit saccades directed toward a laterally positioned odor stimulus (Fig. 1G) and unilaterally occluded flies exhibit biased turning only in the presence of odor (Fig. 2E). Together these results preclude the possibility that the observed steering bias is a nonspecific artifact of the glue itself. However, occluding a3 may compromise the normal movement of this segment relative to the 2nd antennal segment (a2), and thereby perturb the normal activity of the Johnston’s Organ (JO), which encodes the rotation of a3 relative to a2 [20, 21]. Active upwind orientation requires the JO in fruit flies and is enhanced in the presence of odor , thus a unilateral inactivation of JO might impose a turning bias independent of the olfactory system.
To test this idea we next fixed a3 to a2 with UV cured glue to abolish mechanosensory input to the JO . Fixing the right (Fig. 3A) and left (Fig. 3B) JO did not abolish frequent yaw deviations from the odor plume (Fig. 3A, B, lower left, black arrows) However, the distribution of heading was significantly biased in the direction of the fixed JO - opposite the directional bias generated by the respective antennal occlusion (Fig. 3A, B, lower right, compare to Fig. 2A, B). Flies with a fixed left JO failed to exhibit an odor dependent reduction in saccade frequency (Fig. 1D) and the mean turning ratio was biased strongly to the left, ipsilateral to the fixed JO (again, opposite to that of the respective antennal occlusions, Fig. 3E, compare to Fig. 2E). Furthermore, all flies with a single fixed JO showed odor mediated reduction in mean plume deviation that did not reach the level of intact flies (Fig. 3F). Bilaterally fixed JO flies (Fig. 3C) showed markedly reduced yaw deviations (Fig. 3C, lower left, black arrows), an overall reduction in turning frequency independent of odor stimulation (Fig. 3D), a small but insignificant leftward turning bias (Fig. 3E), and an odor mediated reduction in mean plume deviation that also did not reach that of control flies (Fig. 2F). These results show that fixing a single JO results in a steering bias in the direction ipsilateral to the intact JO, and this bias is enhanced by an attractive odor.
To assess the contribution of JO to active plume tracking under more challenging conditions we positioned flies +/−90° lateral to the odor plume. Unilaterally fixing both the right and left JO resulted in a loss of olfactory orientation when the odor plume was located contralateral to the fixed JO (Fig. 3G, H). This is evident in a turning ratio biased toward the fixed JO rather than toward the intact a3 (Fig. 3J), an increased mean time to plume acquisition (Fig. 3K) and a decreased probability of plume acquisition (Fig. 3L). However, unilaterally JO-fixed flies were successful in orienting toward a plume ipsilateral to the fixed JO (Fig. 3G, H) resulting in a mean turning ratio (Fig. 3J), mean time to plume acquisition (Fig. 3K) and acquisition probability (Fig. 3L) that were not significantly different from control flies. These results suggest that turns are activated by the contralateral JO and are consistent with a study showing that Drosophila with unilaterally fixed JOs fail to orient upwind ipsilateral to the intact JO . Consistent with the unilateral manipulations, bilaterally JO-fixed flies did not orient toward the odor plume (Fig. 3I), and exhibited an equal distribution of right and left turns (Fig. 3J), increased time to plume acquisition (Fig. 3K), and reduced probability of plume acquisition (Fig. 3L).
Unilaterally fixing the JO has the opposite effect on steering behavior from unilaterally occluding a3 suggesting that the observed turning biases cannot be explained by non-specific effects of the glue but rather are specific to the sensory manipulation (Fig 2G,H, Fig 3G,H). Does gradient detection by the olfactory system require synergistic olfactory and mechanosensory feedback? Or rather are these cooperative but separate sensory responses? To test these ideas, we controlled odor stimulation of the antennae by using a flight arena equipped with a fixed head-on plume in which steering responses are measured by optically tracking wing kinematics of rigidly fixed flies (Fig. S1B). In response to a head-on odor stimulus, control flies exhibited no net steering bias (Fig. 4A). By contrast, unilateral a3 occlusions generated tonically biased turning in the direction of the intact a3 which scaled depending on the length of the odor pulse (Fig. 4B, 4B inset). However, this did not occur when we unilaterally fixed the JOs (Fig. 4C). In the magnetic tether, directional odor tracking requires JO input contralateral to the a3 mediating the turn (Fig. 3G, H). We therefore unilaterally occluded the a3 in addition to its contralateral JO and found that the tonic steering bias persisted in the absence of contralateral JO input. Likewise, unilateral occlusions of a3 and its ipsilateral JO also had little effect on the observed olfactory mediated steering bias.
Although a3 alone is sufficient to tonically direct steering in the direction of higher odor intensity, it would appear that JO input is involved in initiating real turns during flight independent of odor. The JO is crucial for detecting and responding to wind, gravity and sound [22–24]. A branched moment arm protruding from a3 called the arista amplify mechanical rotation about the a2/a3 JO joint , thus clipping the aristae results in diminished physiological activation of JO sensory neurons and diminished behavioral responses to sound .
We surgically removed the aristae (Fig. S3A), which had no detectable influence over corrected yaw deviations and active plume tracking (Fig. S3A, black arrows), odor reduced saccade frequency (Fig. 3D), turning ratio (Fig. 3E), or mean plume deviation (Fig. 3F). Aristae-clipped flies showed near normal responses to 90° displacements (Fig. S3B) and a normal capacity for saccades directed toward an odor stimulus (Fig. 3J). Although flies with clipped aristae took somewhat longer to acquire the plume (Fig. 3K) and showed a reduced probability of success (Fig. 3L), their overall tracking ability (Fig. 3F) mediated by directed turns (Fig. 3J) suggests that the aristae do not transduce sensory feedback required for tracking an odor gradient. However, the antennae of many insects are not purely sensory organs but rather are under muscular control [27–30] and in Drosophila can be observed to twitch during tethered flight (unpublished observation), a feed-forward behavior which is presumably independent of the aristae.
Results of our free-yaw magnetic tether experiments suggest that that JO mechanosensory feedback is required for the proper directional orientation of saccades. Results of our rigid-tether experiments in which mechanosensory feedback is disabled, suggest that the a3 olfactory sensors alone are sufficient for encoding an odor gradient and generating a tonic bias in steering. Taken together, these results lead to the hypothesis that asymmetrical olfactory cues generate a tonic bias in steering which stimulates active contralateral antennal movements to direct steering maneuvers into either an odor gradient, when one is present (current study), or into an odorless headwind . While the underlying neural mechanisms of these interactions are as yet undetermined, it is clear that flies must use bilateral olfactory comparisons to facilitate gradient tracking, possibly through the fraction of olfactory sensory neurons which project unilaterally, or through as yet unidentified local circuits. Also, stable olfactory tracking requires cooperative interactions among both the olfactory and mechanosensory systems further highlighting the importance of multi-sensory interactions in flies .
In addition to the cooperative influence of olfactory and mechanosensory systems, our results reveal a consistent asymmetry in antennal-mediated flight control. In general, occluding the left antenna has a stronger effect than occluding the right antenna. For example, input to the left a3 is sufficient to mediate a significant odor mediated decrease in saccade frequency (Fig. 2D), and input from the left a3 generates a higher proportion of left turns than the right a3 in response to odor (Fig. 2E, J). Additionally, an intact left JO is sufficient to facilitate a significant odor mediated decrease in saccade frequency (Fig. 3D) and a mean plume deviation near control levels (Fig. 3F), and fixing the left JO had a more significant impact on rightward plume tracking than fixing the right JO had on leftward plume tracking (Fig. 3K, L). These results suggest that both the left a3 and the left JO contribute disproportionately more to behavioral responses than their counterparts. The lateralization effects also appear in the rigid tether flight assay, such that the occluding the right a3 with the left JO produces a stronger right steering bias than does occluding the left a3 and the right JO (Fig. 4D, E). One potential adaptive function of lateralized sensitivity is that the stronger left a3 signals cooperate with weaker right JO signals during left-directed turns, and weaker right a3 signals cooperate with stronger left JO signals during right- directed turns. The hierarchical cooperation of asymmetrical cross-modal olfactory and mechanosensory signals thereby facilitates stable odor tracking in complex multisensory environments, and likely also increases the efficiency of search behavior.
Adult female Drosophila 3–5 days post eclosion were cold-anesthetized on a Peltier plate. To remove olfactory input to the antennae a small drop of UV dental glue (Cas-Ker, Cincinnati OH, USA) was spread over the third antennal segment and after annealing cured with a 10-second burst of UV light (ELC-410, Electro-Lite, Bethel CT, USA). To remove mechanosensory input to the Johnston’s organs, a small drop of UV glue was applied medially at the a2/a3 joint and cured with UV light. Flies were then prepared for behavioral experiments as follows.
The olfactory magnetic tether apparatus has been described previously [16, 17]. Briefly, a fly is glued to a minutien pin (Fine Science Tools) and placed between two rare-earth magnets allowing the fly to rotate freely in the yaw axis (Fig. S1A). Mass-flow regulated air was passed through a switchable gas multiplexer (Sable Systems) at 7 liters per min, saturated by bubbling through water or apple cider vinegar, and delivered to the fly through a double-barrel nozzle placed 4mm dorsal and 3mm anterior to the fly’s head. Air was drawn downward over the fly through a glass vacuum tube and into a clear acrylic vacuum chamber at a rate of 13 L/min (flow regulator, Cole Parmer Instruments) to create a spatially discreet odor plume. We digitized images of the fly from below at 30Hz using an infrared Firewire camera (Fire-i). Body heading was analyzed from the resulting images using custom software routines written in MATLAB (The Mathworks). The magnet arena was outfitted with an electronic visual display that fully surrounded the fly in azimuth, and subtended 60 degrees in zenith.
Before each experimental condition, we revolved a 30°-wide vertical stripe several times in the arena to ensure that each fly was able to orient along any arbitrary heading. We then simultaneously switched on either a water or odor vapor plume and oscillated the same vertical stripe ±22.5 degrees at 1.6 Hz for 8.5 seconds to visually “drag” the flies to a set arena heading relative to the odor plume. Finally, we presented a stationary wide-field grating with a spatial wavelength of 30 degrees and a periodic contrast of 93% at roughly 70 cd m−2 and recorded the resulting behavior.
The following experimental conditions were presented in random order: Note that the location of the odor port is fixed at 0 degrees for all conditions. (1) Flies were positioned at 0° with water vapor emanating from the odor port and then presented high contrast wide-field visual cues. (2) Flies were positioned at 0° with vinegar vapor emanating from the odor port and then presented high contrast wide-field visual cues (Fig. 1A). (3) Flies were positioned at +90° with vinegar vapor emanating from the odor port and then presented high contrast visual cues (Fig. 1F, blue arrows). (4) Flies were positioned at −90° with vinegar vapor emanating from the odor port and then presented high contrast visual cues (Fig. 1F, green arrows). We only analyzed flies with a heading at time=0 within +/− 45° of the specified start position. A saccade was defined by a deviation in heading with an angular velocity between 150 and 1500 degrees per second. Turning ratio was calculated as the number of left turns minus the number of right turns, divided by the number of left turns plus the number of right turns (L−R/L+R). We estimated the width of the plume from the distribution of flight heading, which had a full with at half maximum of 20° (Fig. 1B). For figure 1H, ,2J2J & 3J, turning ratio was calculated for all turns executed before the fly’s heading entered the plume window (+/− 10° from 0°). Likewise, the time to acquisition (Fig. 1I, ,2K2K & 3K) was the amount of time a fly took to reach the plume window.
The visual flight arena, optical wing beat analyzer, and odor delivery system have been described previously . Briefly, cold anesthetized flies tethered to a tungsten pin using UV-activated glue and placed in the geometric center of a cylinder composed of 8×8 computer-controlled LED panels. An infrared diode projects light onto the dorsal surface of the fly to cast a shadow of the beating wings onto an optical sensor. During flight, the overall wing beat frequency and amplitude for each wing is extracted from the time-varying signal using custom hardware (U. of Chicago, JFI Electronics). The difference in voltage between the left and right wing beat amplitude is fed back into the display in real time to simulate active control by the fly over the visual panorama. The visual arena was equipped with an olfactory stimulus system described previously . Briefly, stimuli are delivered to the fly by bubbling mass-flow regulated (MFC-4, Sable Systems, Las Vegas NV) room air at 20 mL/min through test tubes containing water or a strip of filter paper soaked in pure apple cider vinegar. The headspace from the two stimulus tubes converge upon a single nozzle positioned anterior to the fly and in-line with a vacuum system positioned posterior to the fly. A computer-controlled solenoid valve permitted switching between the vinegar and water stimulus.
During the course of the experiment, flies were allowed closed-loop control over a wide-field rotating high contrast checkerboard pattern. Flies were presented with a continuous water vapor stimulus, which was interrupted with either a five second pulse of apple cider vinegar, or a 0.1 second pulse to provide the mechanosensory equivalent of switching to odor while minimizing the impact of the odor stimulus. Steering responses were quantified as the mean difference in the voltage signal that encodes the left and right wing beat amplitude during the 3 seconds after stimulus onset.
Visual flight simulators equipped with odor plumes. Both apparatus are explained in more detail elsewhere [9, 31]. (A) Female fly is fixed to a steel pin with UV glue. The pin is suspended vertically by a magnetic field, enabling the flying animal to steer in the horizontal (yaw) plane. An odor nozzle delivers either water vapor or apple cider vinegar vapor in a narrow plume, approximately 20° wide (see methods), that is drawn away by a vertical suction. Infrared floodlights illuminate the fly for a video camera viewing the fly from beneath. The angular orientation of the body is tracked with off-line software routines in Matlab. The arena is surrounded by a high-contrast pattern of vertical stripes projected by an array of light emitting diodes. (B) Female fly is rigidly fixed within an infrared beam that casts shadows of the beating wings on an optical sensor. Attempts to steer are recorded as changes in wing beat amplitude. A nozzle delivers a plume that is drawn away by a regulated suction. The visual panorama rotates with a velocity proportional to the asymmetry in wing beat amplitude, but the odor plume is fixed in place. As such the fly can actively control the visual panorama but does not leave the odor plume.
Flies show no odor tracking in the absence of odor. Flies were visually dragged to 0° and presented with a high contrast visual panorama (Fig. 1A) in the absence of odor. Four representative flight trajectories for (A) Control flies, (B) Flies with the right 3rd antennal segment (a3) occluded with UV glue (yellow for a3 occlusions), (C) left a3 occluded, (D) both a3s occluded, (E) flies with the right Johnston’s organ (JO) fixed with UV glue (red for JO manipulations), (F) left JO fixed, (G) both JO fixed, and (H) flies with both aristae clipped.
Flies with bilaterally clipped aristae (A, top) show stable plume tracking and slightly impaired orientation behavior. (A, lower left) Representative flight trajectories and (A, lower right) histograms of flight heading as in Fig. 1B where red dashed line indicates mean flight heading in the presence of odor. Mean heading = −6°. (C left) Representative flight trajectories and (B, right) smoothed mean heading as in Fig. 1G. Accompanying behavioral measures can be found in Fig. 3G, H, I, M, N, and O (N=26).
M.A.F. is an HHMI Early Career Scientist. This study was supported by the National Science Foundation (IOS-0718325) to M.A.F., by NIH NRSA T32 (GM065823) to D.M.C., and the Gresser Fellowship in Life Sciences to B.J.D.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.