An olfactory circuit increases the fidelity of visual behavior

Abstract

Multimodal integration allows neural circuits to be activated in a behaviorally context-specific manner. In the case of odor plume tracking by Drosophila, an attractive odorant increases the influence of yaw-optic flow on steering behavior in flight, which enhances visual stability reflexes, resulting in straighter flight trajectories within an odor plume. However, it is not well understood whether context-specific changes in optomotor behavior are the result of an increased sensitivity to motion inputs (e.g., through increased visual attention) or direct scaling of motor outputs (i.e., increased steering gain). We address this question by examining the optomotor behavior of Drosophila melanogaster in a tethered flight assay and demonstrate that whereas olfactory cues decrease the gain of the optomotor response to sideslip optic flow, they concomitantly increase the gain of the yaw optomotor response by enhancing the animal's ability to follow transient visual perturbations. Furthermore, ablating the mushroom bodies (MBs) of the fly brain via larval hydroxyurea (HU) treatment results in a loss of olfaction-dependent increase in yaw optomotor fidelity. By expressing either tetanus toxin light chain or diphtheria toxin in gal4-defined neural circuits, we were able to replicate the loss of function observed in the HU treatment within the lines expressing broadly in the mushroom bodies, but not within specific mushroom body lobes. Finally, we were able to genetically separate the yaw responses and sideslip responses in our behavioral assay. Together, our results implicate the MBs in a fast-acting, memory-independent olfactory modification of a visual reflex that is critical for flight control.

Figure 1.

Experimental apparatus, stimulus, and analysis procedure. A, B, An electronic visual flight simulator displays either wide-field yaw motion (A) or sideward translational motion (B) to a tethered fly. C, An odor nozzle was positioned in the headspace of the animal during the beginning of each experiment, allowing the delivery of either water vapor or apple cider vinegar. D, Cross-correlating the fly's behavior y(t) with the white noise stimulus f(t) produces a linear filter h(t) that describes the temporal dynamics of the fly's response to a single motion impulse (top panel). Convolving h(t) with any arbitrary stimulus s(t) produces a prediction of the fly's response to that stimulus. We therefore were able to quickly and robustly measure the optomotor phenotype of each strain through correlation with a pseudo-white noise sequence. E, The space–time plot (left) maps the angular displacements of a single row of pixels due to the cumulative motion impulses of the m-sequence (right). F, G, The resultant impulse response may be convolved with the visual motion stimulus (F) to predict the fly's behavior (G). H, By plotting the prediction against the actual data, we may estimate the amount of nonlinearity present in the fly's response (see Materials and Methods for more detail).

Figure 2.

Olfactory cues modify the gain and reliability of the optomotor response. A, B, The yaw optomotor filter increases in magnitude (no odor N = 51, odor N = 58) and in the prediction–response correlation (inset) due to the influence of odor (no odor gray, odor black) (A), whereas the sideslip filter decreases in height but does not change in predictive power (no odor N = 66, odor N = 71) (B). These changes in magnitude reflect changes in the gain of the filters in linear systems theory. C, The measured changes in height at the peak of each filter (∼236 ms for yaw and 109 ms for sideslip) differed significantly due to the influence of odor (paired t test). D–F, In contrast, replacing the odor port with a second water vapor tube did not result in any qualitative changes in either the yaw (no odor N₁ = 57, no odor N₂ = 62) (D) or sideslip (no odor N₁ = 75, no odor N₂ = 72) (E) filters or statistical differences in their peak heights (F). G, H, Olfactory cues did not significantly increase the variance of the yaw steering response (G) but did significantly decrease the variance of the sideslip steering response (H). These changes in variance imply that the observed increase in filter gain due to odor is due to an increased steering effort correlated to the white noise stimulus (*p < 0.05, **p < 0.01, ***p < 0.001).

Figure 3.

The prediction of the fly's steering response to a sinusoidal stimulus matches observed responses. A prediction was obtained by convolving the estimated filter with the stimulus velocity signal (see inset). A, B, In yaw, the prediction captures a previously observed increase in steering amplitude in response to odor (A), and in sideslip, the prediction also matches a previously observed, but smaller, decrease in steering amplitude (B). However, the filters are not able to replicate the observed high-frequency cutoff (near middle of stimulus). Plotting a random sample of the predicted response against the real data reveals the linearity of the phase relation between the two traces. C, Odor improves the phase correlation for the yaw prediction as measured by Pearson's correlation coefficient, r, to a linear fit. D, However, the phase correlation of the sideslip response is already largely independent of olfactory context.

Figure 4.

A circuit ablated by larval hydroxyurea treatment is required for yaw olfactory-mediated optomotor modification. Wild-type animals increase wingbeat frequency in the presence of an attractive odor, apple-cider vinegar (ACV), but not due to a switch to water stimuli (W). A, B, Similar to wild-type animals, HU-treated flies significantly increase their WBF in the presence of odor (A) and decrease the variance of sideslip responses in the presence of odor (B). C, Treatment with HU results in a loss of olfaction-dependent changes in gain in both yaw and sideslip filters (yaw no-odor N = 49, yaw odor N = 47, sideslip no-odor N = 57, sideslip odor N = 59), although the sideslip height shows a decreasing trend (p = 0.25). D, Neither filter shows olfaction-dependent changes in prediction–response correlation. E, Similar to wild-type animals, HU-treated flies can localize an odor source hidden in the floor in a walking trap assay, albeit with a slower time course than wild-type animals. F, The predicted yaw response to the sinusoidal stimulus in the presence of odor has much lower amplitude than the actual wild-type odor response. *p < 0.05, **p < 0.01, ***p < 0.001.

Figure 5.

Transgenic yaw filters. A, B, Control lines UAS-TnT and PCF exhibit wild-type yaw OMOR (A), whereas MB lines UAS-TnT^+/−; OK107-gal4^+/− and c320-gal4/UAS-TnT show a deficit in olfaction-dependent modulation of the yaw filters (B). C, Expressing TeTxLc in the gal4-lines targeting single MB lobes, such as c739-gal4 (α/β), 17d-gal4 (α/β), c305a-gal4(α′/β′), and H24-gal4 (γ), does not have a significant impact on OMOR. D, Expressing DT with c739-gal4 and 17d-gal4 results in animals with highly reduced yaw filters. E, Lines with wider MB expression, simultaneously targeting the α/β and γ lobes, do not result in a loss of OMOR in the case of MB247-gal4 but do result in a loss of OMOR in the case of D52H-gal4. F, The line GH298-gal4^+/−; UAS-DTI^+/− exhibits robust OMOR, indicating that the antennal lobe local interneurons are not required. Similarly, expressing DT in the olfactory receptor neurons labeled by OR22a-gal4 and OR43b-gal4 has no effect on OMOR. G, The line OR42b-gal4^+/−; UAS-DTI^+/− exhibits a loss of OMOR. H, Phosphodiesterase mutant dunce and adenylate cyclase mutant rutabaga both have normal yaw OMOR, as does the rover variant of the for allele. I, The sitter variant has very little OMR in the absence of odor and the S2 variant exhibits a large decrease in the magnitude of the yaw filter and a loss of OMOR. *p < 0.05, **p < 0.01, ***p < 0.001. An asterisk over the filters indicates a significant difference in peak height. The numbers listed after the name of the line indicates the N in the no-odor and odor conditions, respectively.

Figure 6.

Transgenic sideslip filters. A, A statistically significant decrease in sideslip filter gain was not apparent in the control line UAS-TnT, although such a decrease was detected in the control line PCF. Both control sideslip filters exhibited wild-type dynamics. In contrast, the MB line UAS-TnT^+/−; OK107-gal4^+/− had reduced sideslip filter with slower dynamics. B, Since the MB line c320-gal4/UAS-TnT had a wild-type sideslip filter, the altered dynamics in the OK107-gal4 line may be due to expression outside the MBs. Expressing TeTxLc in single MB lobes, such as in c739-gal4 (α/β), 17d-gal4 (α/β), c305a-gal4 (α′/β′), and H24-gal4 (γ), did not much alter the overall dynamics of the sideslip response. C, The line 17d-gal4/UAS-TnT appeared to have an enhanced suppression of sideslip filter gain due to odor, and the line H24-gal4 seemed to exhibit an inverted response to odor—an increase of sideslip filter gain. D, Expressing DT with c739-gal4 and 17d-gal4 did not much affect the sideslip filter, by comparison to its effects on the respective yaw filters. E, However, the α/β and γ lobes line UAS-TnT^+/−; MB247-gal4^+/−had a greatly decreased sideslip filter, perhaps due to expression in the optic ganglia, while the line D52H-gal4; UAS-TnT was wild type. F, Neither the antennal lobe interneuron line GH298-gal4^+/−; UAS-DTI^+/− or lines expressing DT in OR lines 22a or 42b had abnormal sideslip filters. G, The line OR43b-gal4^+/−; UAS-DTI^+/− inexplicably had reduced filter gain. H, I, cAMP mutants rut1 and dnc1 both had wild-type sideslip filters (H), while for lines R, S, and S2 all had highly reduced sideslip filters (I). *p < 0.05, **p < 0.01, ***p < 0.001. An asterisk over the filters indicates a significant difference in peak height. The numbers listed after the name of the line indicates the N in the no odor and odor conditions, respectively.

Figure 7.

Selected filters show genetic separation of yaw and sideslip optic flow processing. A–D, Although the lines c730-gal4/UAS-TnT (A) and 201y-gal4/UAS-TnT (B) were wild type, expression of DT with c739-gal4 (C) and 201y-gal4 (D) resulted in highly disrupted yaw filters without much affecting the sideslip filters. E, F, Crossing these lines with a fly homozygous for MBGal80 partially rescued the yaw optomotor filters in both lines, which might indicate that the output of the MBs impinges on the yaw optomotor circuit downstream of the optic ganglia. *p < 0.05, **p < 0.01, ***p < 0.001. An asterisk over the filters indicates a significant difference in peak height. The numbers listed after the name of the line indicates the N in the no odor and odor conditions, respectively.

Figure 8.

A hypothesized circuit for olfactory modification of the optomotor reflex. A, We posit that yaw and sideslip optomotor responses rely on different networks of electrically coupled LPTCs, or at least functionally separated parallel visual–motor control channels. Olfactory cues influence the two optomotor channels disparately, increasing gain in the yaw motor transformation and reducing gain in sideslip. Furthermore, HU ablation defines the olfactory circuit responsible for olfaction-dependent optomotor modification, suggesting the involvement of the MBs. B, Olfactory information acts as a switch between a search state that allows greater flexibility in the yaw optomotor channel and tracking state that requires greater yaw fidelity to track odor plumes.