Feature encoding: How back-to-front motion guides the polite fly

Motion of a visual image from back-to-front across a visual field can provide an early-stage cue for impending collisions. A new study reveals visual feature encoding neurons that drive behavioral responses to back-to-front motion in the fly Drosophila melanogaster. Motion of a visual image from back-to-front across a visual field can provide an early-stage cue for impending collisions. A new study reveals visual feature encoding neurons that drive behavioral responses to back-to-front motion in the fly Drosophila melanogaster. We all try to avoid collisions, both because they can be deleterious and because they can be rude — bumping into someone at the supermarket rarely elicits a warm reaction. When we think of collision avoidance behavior, however, we often think of the rapid, explosive movements animals from invertebrates to humans employ to propel themselves away from an impending threat1Branco T. Redgrave P. The neural basis of escape behavior in vertebrates.Annu. Rev. Neurosci. 2020; 43: 417-439Crossref PubMed Scopus (27) Google Scholar,2Peek M.Y. Card G.M. Comparative approaches to escape.Curr. Opin. Neurobiol. 2016; 41: 167-173Crossref PubMed Scopus (42) Google Scholar. Sometimes avoiding a collision does not require such energetically expensive acrobatics: in a locomoting animal, collisions can be avoided by merely stopping or slowing down. Observing fruit flies in a small dish, Zabala et al.3Zabala F. Polidoro P. Robie A. Branson K. Perona P. Dickinson M.H. A simple strategy for detecting moving objects during locomotion revealed by animal-robot interactions.Curr. Biol. 2012; 22: 1344-1350Abstract Full Text Full Text PDF PubMed Scopus (42) Google Scholar noticed particular interactions that would lead to stopping behavior. If two flies were moving with intersecting trajectories, the faster fly would often continue on its path while the slower fly stopped, as a human would when making a polite gesture to let someone else pass by. If two flies were moving in parallel but at different velocities, the fly moving at a slower velocity would often stop. To determine the visual cues that may instigate these behaviors, Zabala et al.3Zabala F. Polidoro P. Robie A. Branson K. Perona P. Dickinson M.H. A simple strategy for detecting moving objects during locomotion revealed by animal-robot interactions.Curr. Biol. 2012; 22: 1344-1350Abstract Full Text Full Text PDF PubMed Scopus (42) Google Scholar developed a robot, the Flyatar: a fly-sized magnet whose trajectories were controlled by servomotors in closed loop. From the interactions between the Flyatar and individual flies, Zabala et al.3Zabala F. Polidoro P. Robie A. Branson K. Perona P. Dickinson M.H. A simple strategy for detecting moving objects during locomotion revealed by animal-robot interactions.Curr. Biol. 2012; 22: 1344-1350Abstract Full Text Full Text PDF PubMed Scopus (42) Google Scholar concluded the fly stopping behavior was caused by the image of the Flyatar moving from back-to-front across the fly’s visual field. As they report in this issue of Current Biology, Tanaka and Clark4Tanaka R. Clark D.A. Neural mechanisms to exploit positional geometry for collision avoidance.Curr. Biol. 2022; 32: 2357-2374Abstract Full Text Full Text PDF PubMed Scopus (9) Google Scholar have now investigated the neural underpinnings of this behavior. They began by performing a virtual version of the Flyatar experiment, simulating a moving observer surrounded by objects moving at different velocities and trajectories. Their simulation revealed that two object trajectories are indicative of high collision risks. The first was an object approaching on a direct collision course. Here, the object expands uniformly across an animal’s retina as the center of mass of the object remains fixed in one location of the visual field. These stimuli are well known to elicit brief or sustained freezing behaviors5Card G. Dickinson M.H. Visually mediated motor planning in the escape response of Drosophila.Curr. Biol. 2008; 18: 1300-1307Abstract Full Text Full Text PDF PubMed Scopus (179) Google Scholar, 6Zacarias R. Namiki S. Card G.M. Vasconcelos M.L. Moita M.A. Speed dependent descending control of freezing behavior in Drosophila melanogaster.Nat. Commun. 2018; 9: 3697Crossref PubMed Scopus (39) Google Scholar, 7Yilmaz M. Meister M. Rapid innate defensive responses of mice to looming visual stimuli.Curr. Biol. 2013; 23: 2011-2015Abstract Full Text Full Text PDF PubMed Scopus (296) Google Scholar. The other high-risk trajectory came from laterally located objects moving back-to-front across the observer’s visual field, the same motion cues in Flyatar trajectories causing stopping behavior in freely moving flies3Zabala F. Polidoro P. Robie A. Branson K. Perona P. Dickinson M.H. A simple strategy for detecting moving objects during locomotion revealed by animal-robot interactions.Curr. Biol. 2012; 22: 1344-1350Abstract Full Text Full Text PDF PubMed Scopus (42) Google Scholar. Why would animals be prone to stop when witnessing back-to-front motion in their lateral visual field? If you are moving forward, your own motion creates front-to-back optic flow from stationary objects in your visual field; if you witness an object with back-to-front motion, the object itself must be moving3Zabala F. Polidoro P. Robie A. Branson K. Perona P. Dickinson M.H. A simple strategy for detecting moving objects during locomotion revealed by animal-robot interactions.Curr. Biol. 2012; 22: 1344-1350Abstract Full Text Full Text PDF PubMed Scopus (42) Google Scholar. Back-to-front motion also signals that the object is moving faster than you and may intersect your trajectory or reach your destination before you do — an early prediction of a potential collision3Zabala F. Polidoro P. Robie A. Branson K. Perona P. Dickinson M.H. A simple strategy for detecting moving objects during locomotion revealed by animal-robot interactions.Curr. Biol. 2012; 22: 1344-1350Abstract Full Text Full Text PDF PubMed Scopus (42) Google Scholar (Figure 1A). How back-to-front motion is detected and drives stopping behavior are not well understood. To determine the neural substrates that drive responses to back-to-front motion, Tanaka and Clark4Tanaka R. Clark D.A. Neural mechanisms to exploit positional geometry for collision avoidance.Curr. Biol. 2022; 32: 2357-2374Abstract Full Text Full Text PDF PubMed Scopus (9) Google Scholar presented small moving objects to tethered flies walking on a ball. As predicted by the Flyatar and simulation experiments, they found objects moving parallel to the fly decreased the average fly walking speed in the absence of significant expansion cues. Interestingly, although front-to-back and back-to-front motion both decreased walking speed, a significantly larger effect was elicited with back-to-front motion. In the fly, motion selectivity first emerges in T4/T5 neurons within the optic lobes. Each T4/T5 neuron has a small receptive field and encodes one of the four cardinal directions of motion within that receptive field — up, down, front-to-back and back-to-front8Maisak M.S. Haag J. Ammer G. Serbe E. Meier M. Leonhardt A. Schilling T. Bahl A. Rubin G.M. Nern A. et al.A directional tuning map of Drosophila elementary motion detectors.Nature. 2013; 500: 212-216Crossref PubMed Scopus (226) Google Scholar. Each directionally selective T4/T5 neuron sends axon terminals to a dedicated layer in the lobula plate of the fly’s optic lobe8Maisak M.S. Haag J. Ammer G. Serbe E. Meier M. Leonhardt A. Schilling T. Bahl A. Rubin G.M. Nern A. et al.A directional tuning map of Drosophila elementary motion detectors.Nature. 2013; 500: 212-216Crossref PubMed Scopus (226) Google Scholar. Motion feature selectivity can then be established in downstream visual projection neurons by their layer-specific dendrite locations. For example, lobula plate lobula columnar neurons type 2 (LPLC2), the dendrites of which project in a radial manner to all four T4/T5 directional layers, are selective for uniformly expanding objects approaching on a direct collision course9Klapoetke N.C. Nern A. Peek M.Y. Rogers E.M. Breads P. Rubin G.M. Reiser M.B. Card G.M. Ultra-selective looming detection from radial motion opponency.Nature. 2017; 551: 237-241Crossref PubMed Scopus (73) Google Scholar. Back-to-front motion selectivity could then be established in a neuron with dendrites that occupy the back-to-front motion layer, but do not occupy the front-to-back motion layer. The lobula plate lobula columnar neuron type 1 (LPLC1) has this dendrite configuration10Wu M. Nern A. Williamson W.R. Morimoto M.M. Reiser M.B. Card G.M. Rubin G.M. Visual projection neurons in the Drosophila lobula link feature detection to distinct behavioral programs.eLife. 2016; 5: e21022Crossref PubMed Scopus (136) Google Scholar, making it a candidate for preferential tuning to back-to-front motion. Tanaka and Clark4Tanaka R. Clark D.A. Neural mechanisms to exploit positional geometry for collision avoidance.Curr. Biol. 2022; 32: 2357-2374Abstract Full Text Full Text PDF PubMed Scopus (9) Google Scholar identified inputs to LPLC1 in an electron microscopy dataset of the fly hemibrain11Scheffer L.K. Xu C.S. Januszewski M. Lu Z. Takemura S.Y. Hayworth K.J. Huang G.B. Shinomiya K. Maitlin-Shepard J. Berg S. et al.A connectome and analysis of the adult Drosophila central brain.eLife. 2020; 9: e57443Crossref PubMed Google Scholar and confirmed LPLC1 receive substantial inputs from back-to-front, but not front-to-back, T5 neurons. Pinpointing LPLC1 as their candidate cell type, they transitioned back to the fly on the ball assay and found that silencing synaptic transmission in LPLC1 removes the back-to-front behavioral bias to moving objects. Employing optogenetics, they demonstrated that activating LPLC1 decreases average walking speed similar to that elicited with visual stimuli. But what appeared to be straightforward evidence for LPLC1 having a role in the detection of back-to-front motion became muddled when Tanaka and Clark4Tanaka R. Clark D.A. Neural mechanisms to exploit positional geometry for collision avoidance.Curr. Biol. 2022; 32: 2357-2374Abstract Full Text Full Text PDF PubMed Scopus (9) Google Scholar investigated LPLC1 visual response tuning. Imaging Ca2+ activity from the bundle of LPLC1 axon terminals, they found LPLC1 population responses are unlike those of many directionally selective neurons, where a stimulus moving in a preferred direction excites the neuron and a stimulus moving in a nonpreferred direction inhibits it12Monier C. Chavane F. Baudot P. Graham L.J. Fregnac Y. Orientation and direction selectivity of synaptic inputs in visual cortical neurons: a diversity of combinations produces spike tuning.Neuron. 2003; 37: 663-680Abstract Full Text Full Text PDF PubMed Scopus (296) Google Scholar, 13Hausen K. Motion sensitive interneurons in the optomotor system of the fly.Biol. Cybern. 1982; 45: 143-156Crossref Scopus (237) Google Scholar, 14Wilson D.E. Scholl B. Fitzpatrick D. Differential tuning of excitation and inhibition shapes direction selectivity in ferret visual cortex.Nature. 2018; 560: 97-101Crossref PubMed Scopus (44) Google Scholar. Instead, LPLC1 are excited by both back-to-front and front-to-back stimuli, with the difference in response magnitude seemingly small when compared to the difference in behavior. Previously documented LPLC1 responses are smaller than other visual projection neurons tuned to vertical bar or small object motion15Stadele C. Keles M.F. Mongeau J.M. Frye M.A. Non-canonical receptive field properties and neuromodulation of feature-detecting neurons in flies.Curr. Biol. 2020; 30: 2508-2519Abstract Full Text Full Text PDF PubMed Scopus (17) Google Scholar,16Klapoetke N.C. Nern A. Rogers E.M. Rubin G.M. Reiser M.B. Card G.M. A functionally ordered visual feature map in the Drosophila brain.Neuron. 2022; 110: 1700-1711.e6Abstract Full Text Full Text PDF PubMed Scopus (14) Google Scholar, and a directional bias in LPLC1 motion responses had not been observed16Klapoetke N.C. Nern A. Rogers E.M. Rubin G.M. Reiser M.B. Card G.M. A functionally ordered visual feature map in the Drosophila brain.Neuron. 2022; 110: 1700-1711.e6Abstract Full Text Full Text PDF PubMed Scopus (14) Google Scholar. Overall, the functional imaging results for LPLC1 across multiple labs are somewhat surprising given the motion direction inputs to LPLC1 and the apparent necessity of LPLC1 for the witnessed behavioral responses to back-to-front motion. How do you get such a strong bias in the behavior response with a mild bias in the neuronal response? Here, Tanaka and Clark4Tanaka R. Clark D.A. Neural mechanisms to exploit positional geometry for collision avoidance.Curr. Biol. 2022; 32: 2357-2374Abstract Full Text Full Text PDF PubMed Scopus (9) Google Scholar suggest that each LPLC1 neuron may not have the same back-to-front versus front-to-back tuning. When recording from dendrites of individual LPLC1, they found that direction selectivity depends on the azimuthal location of the LPLC1 neuron’s receptive field. The authors also report evidence that inhibition suppresses LPLC1’s back-to-front bias. By expressing the glutamate indicator iGluSnFR in LPLC1 to monitor glutamate release around the LPLC1 lobula plate dendrites — glutamate is an inhibitory neurotransmitter in the fly — they observed larger responses to back-to-front stimuli. Similarly, when they decreased inhibition by knocking down glutamatergic receptors in LPLC1, they found a significant increase in back-to-front direction selectivity. On the basis of these data, Tanaka and Clark4Tanaka R. Clark D.A. Neural mechanisms to exploit positional geometry for collision avoidance.Curr. Biol. 2022; 32: 2357-2374Abstract Full Text Full Text PDF PubMed Scopus (9) Google Scholar propose a model in which inhibition is distributed as a gradient across the LPLC1 population to restrict the back-to-front bias to the lateral region of the fly’s visual field (Figure 1B) — the area where back-to-front motion is most predictive of a future collision. This is of broad interest as within cell type variations in tuning have been reported across visual systems17Heukamp A.S. Warwick R.A. Rivlin-Etzion M. Topographic variations in retinal encoding of visual space.Annu. Rev. Vis. Sci. 2020; 6: 237-259Crossref PubMed Scopus (14) Google Scholar. Tanaka and Clark4Tanaka R. Clark D.A. Neural mechanisms to exploit positional geometry for collision avoidance.Curr. Biol. 2022; 32: 2357-2374Abstract Full Text Full Text PDF PubMed Scopus (9) Google Scholar thus provide behavioral relevance for differential tuning and a testable model to uncover mechanisms for direction selectivity12Monier C. Chavane F. Baudot P. Graham L.J. Fregnac Y. Orientation and direction selectivity of synaptic inputs in visual cortical neurons: a diversity of combinations produces spike tuning.Neuron. 2003; 37: 663-680Abstract Full Text Full Text PDF PubMed Scopus (296) Google Scholar. Tanaka and Clark4Tanaka R. Clark D.A. Neural mechanisms to exploit positional geometry for collision avoidance.Curr. Biol. 2022; 32: 2357-2374Abstract Full Text Full Text PDF PubMed Scopus (9) Google Scholar note that LPLC1 cells respond robustly to expanding object stimuli containing no translational motion. A recent comparison by Klapoetke et al.16Klapoetke N.C. Nern A. Rogers E.M. Rubin G.M. Reiser M.B. Card G.M. A functionally ordered visual feature map in the Drosophila brain.Neuron. 2022; 110: 1700-1711.e6Abstract Full Text Full Text PDF PubMed Scopus (14) Google Scholar of responses across different visual projection neurons actually categorized LPLC1 as looming-responsive based on this tuning and the location of its axons adjacent to other looming-responsive visual projection neurons. Similarly, a strong optogenetic activation of LPLC1 causes takeoff escapes similar to those elicited with looming stimuli10Wu M. Nern A. Williamson W.R. Morimoto M.M. Reiser M.B. Card G.M. Rubin G.M. Visual projection neurons in the Drosophila lobula link feature detection to distinct behavioral programs.eLife. 2016; 5: e21022Crossref PubMed Scopus (136) Google Scholar. How (and why) do inputs to LPLC1 in other lobula and lobula plate layers enable both expanding object and moving object responses? And similarly, how are the more broadly tuned LPLC1 visual responses reconciled in downstream neurons that drive behaviors? Tanaka and Clark4Tanaka R. Clark D.A. Neural mechanisms to exploit positional geometry for collision avoidance.Curr. Biol. 2022; 32: 2357-2374Abstract Full Text Full Text PDF PubMed Scopus (9) Google Scholar begin to investigate this question in major postsynaptic partners of LPLC1, identified in the electron microscopy dataset. Out of the partners probed, they find one neural cell type shows looming and small object motion sensitivity that follows the LPLC1 population response. When optogenetically activated, these neurons caused slowing behavior; however, silencing experiments determined they were not essential for the back-to-front slowing responses. These data suggest other, yet to be identified neurons must be involved in the behavior, perhaps neurons that show an enhanced back-to-front bias by pooling responses from LPLC1 neurons from the lateral, back-to-front selective part of the fly’s visual field. Spatial mapping appears to be lost in the seemingly disorganized axon terminals of most visual projection neurons10Wu M. Nern A. Williamson W.R. Morimoto M.M. Reiser M.B. Card G.M. Rubin G.M. Visual projection neurons in the Drosophila lobula link feature detection to distinct behavioral programs.eLife. 2016; 5: e21022Crossref PubMed Scopus (136) Google Scholar, however, emerging evidence suggests spatial information can be reassembled in postsynaptic partners18Morimoto M.M. Nern A. Zhao A. Rogers E.M. Wong A.M. Isaacson M.D. Bock D.D. Rubin G.M. Reiser M.B. Spatial readout of visual looming in the central brain of Drosophila.eLife. 2020; 9: e57685Crossref PubMed Scopus (15) Google Scholar. In all, the extensive experiments by Tanaka and Clark4Tanaka R. Clark D.A. Neural mechanisms to exploit positional geometry for collision avoidance.Curr. Biol. 2022; 32: 2357-2374Abstract Full Text Full Text PDF PubMed Scopus (9) Google Scholar highlight the importance of examining seemingly subtle differences in visual feature encoding that may have significant behavioral relevance, and provide insight on how innate politeness emerges through collision avoidance circuits. The author declares no competing interests. Neural mechanisms to exploit positional geometry for collision avoidanceTanaka et al.Current BiologyMay 3, 2022In BriefVisual motion contains rich information about space, but how brains decode spatial information to guide specific behaviors remains poorly understood. Tanaka and Clark show how Drosophila LPLC1 neurons implement a selective collision avoidance behavior by pooling outputs of motion and object detectors, as well as spatially biased inhibition. Full-Text PDF