An Array of Descending Visual Interneurons Encoding Self-Motion in Drosophila

Abstract

The means by which brains transform sensory information into coherent motor actions is poorly understood. In flies, a relatively small set of descending interneurons are responsible for conveying sensory information and higher-order commands from the brain to motor circuits in the ventral nerve cord. Here, we describe three pairs of genetically identified descending interneurons that integrate information from wide-field visual interneurons and project directly to motor centers controlling flight behavior. We measured the physiological responses of these three cells during flight and found that they respond maximally to visual movement corresponding to rotation around three distinct body axes. After characterizing the tuning properties of an array of nine putative upstream visual interneurons, we show that simple linear combinations of their outputs can predict the responses of the three descending cells. Last, we developed a machine vision-tracking system that allows us to monitor multiple motor systems simultaneously and found that each visual descending interneuron class is correlated with a discrete set of motor programs.

Significance statement: Most animals possess specialized sensory systems for encoding body rotation, which they use for stabilizing posture and regulating motor actions. In flies and other insects, the visual system contains an array of specialized neurons that integrate local optic flow to estimate body rotation during locomotion. However, the manner in which the output of these cells is transformed by the downstream neurons that innervate motor centers is poorly understood. We have identified a set of three visual descending neurons that integrate the output of nine large-field visual interneurons and project directly to flight motor centers. Our results provide new insight into how the sensory information that encodes body motion is transformed into a code that is appropriate for motor actions.

Keywords: ethology; flight; self-motion estimation; vision.

Figure 1.

Photoactivation of visual interneuron outputs. A, GFP signal after photoactivation of PA-GFP in VS cell output terminals (green) in the right hemisphere of the central brain. The left of the image is approximately the midline of the brain. Red signal shows residual tdTomato expression driven by the driver DB331-Gal4 used to target the visual interneurons initially. The approximate center of activation is marked by the pink asterisk. Activation at VS output terminals led to robust activation in the VS cells (arrow points toward cell bodies and dendritic arbors), as well as a number of putative descending neuron cell bodies (arrowheads). B, GFP signal after photoactivation of PA-GFP in the thoracic ganglia in the same brain as in A. Red signal shows tdTomato expression driven by the driver DB331-Gal4 used to target the visual interneurons in the brain initially. A large output process of a descending neuron in the neck motor neuropil (arrow) and a second prominent descending process (arrowhead) are identified. Scale bar, 50 μm in both A and B.

Figure 2.

Anatomy of the descending neurons and gap junction-coupled synaptic partners. A, DNOVS1 (green) and DNOVS2 (red) are traced in the same brain and shown with an anti-nc82 neuropil stain (blue). We traced GFP expression in DNOVS1 and an intracellular fill in DNOVS2 (biocytin later conjugated to an Alexa Fluor dye). The neuropil signal is a maximum-intensity projection. B, DNHS1 (pink) was traced using an intracellular fill and is shown against a neuropil stain in the same brain (blue). C, Processes of DNHS1 traced in the thoracic ganglia indicate putative output processes in neck and haltere motor regions. D, Magnified maximum intensity projection of DNOVS1 and DNOVS2 processes in the brain. GFP expression in DNOVS1 appears green. The processes of DNOVS2, which are also labeled by GFP, are also labeled with an Alexa Fluor dye conjugated to biocytin (red; overlap appears yellow-orange). The cell body of DNOVS2 was extracted during removal of the recording electrode, but that of DNOVS1 remains. E, DNHS1 processes overlap with HS terminals (arrows). DNHS1 and HS cells express GFP (green) and DNHS1 is also labeled with an intracellular biocytin fill (red; overlap appears yellow-orange). DNOVS1 is also labeled by this driver line and appears in this image (arrowhead). F, DNOVS1 output terminals in the neck motor region. DNOVS1 is labeled both by GFP (green) and an intracellular dye fill (red) and appears yellow. Dye-coupled synaptic partners with DNOVS1, the frontal neck motor neurons, appear in red (dendrites are indicated by the arrow and cell bodies by the arrowhead). G, DNOVS1 and dye-coupled synaptic partners VS cells 4–6 are labeled with a biocytin fill (red). DNOVS1's axon is indicated by the arrow and cell bodies of VS 4–6 by the arrowhead. GFP expression is shown in green and anti-nc82 neuropil stain is in blue. VS 1–3 are not dye coupled to DNOVS1, but are also labeled by this Gal4-driver line (green) and appear next to the red VS 4–6 cell dendrites. H, DNOVS2 (green) is dye coupled to VS2 (red). Both neurons were traced using the same biocytin stain after filling DNOVS2, but are colored differently. Neuropil appears blue. I, DNOVS2 (green) is dye coupled to VS3 (red). Neuropil appears blue. J, DNHS1 (green) is dye coupled to HSN and HSE (red; HSE is faintly filled in this image and marked with an arrow). Neuropil stain appears blue. K, DNOVS2 (axon indicated by arrow) is labeled by GFP (green) and filled with biocytin (red signal; overlap with GFP signal results in yellow appearance of DNOVS2). A dye-coupled neuron in the protocerebral bridge appears red (asterisk). DNHS1 is also labeled by GFP in this driver line (green; arrowhead). Scale bars in D and E are 20 μm; all others are 50 μm. We used the following driver lines to label neurons in this figure: for A, D, and K: 94H09-Gal4; for B, C, and F: R15C02-Gal4; and for E and G: R20C05-Gal4. In all figures, the brain is oriented so anterior is upward.

Figure 3.

The three descending neurons are tuned to distinct axes of rotation. A, Schematic of the electrophysiology–behavior preparation (not to scale). B, Icon depicting how rotational stimuli are represented in the present study. Direction of rotation is presented according to the right-hand rule. For instance, rotation at an angle of zero corresponds to roll of the fly down to the right around its longitudinal axis. C, Membrane potential traces from DNOVS1. In the leftmost column, an entire motion stimulus is plotted. Response during quiescence is shown in black and the trace from during flight is shown in red. In the second-most column, 100 ms of raw membrane potential (from the left trace, after motion) is shown. DNOVS1 responds with purely graded changes in membrane potential. The right two columns show average membrane potential responses for DNOVS1 for eight individuals (lighter thin lines) and across flies (thick lines). The light gray region indicates the 2 s when the rotational stimulus was presented. Motion corresponding to clockwise and counterclockwise rotations of the fly at the preferred axis is indicated by the curved arrows. Baseline membrane potentials during quiescence are shown. D, DNOVS2 raw (left two columns) and average (right two columns) responses to motion. Blue indicates responses during flight and black indicates responses during quiescence. Yellow dots indicate spikes in the expanded traces. Averages across and single traces for 11 flies are shown. DNOVS2 produces small spikes on top of graded changes in membrane potential during flight and motion. E, DNHS1 raw and average responses to motion for two flies (same format as C and D). Responses during flight are in green and during quiescence are in black. DNHS1, similar to DNOVS2, produces small spikes, indicated by the yellow dots in the second-most column. The traces in the left column in C, D, and E are all plotted on the same y-scale; traces in the second column of C, D, and E are equal but slightly expanded relative to the left column. F, Average membrane potential versus spike rate during motion for DNOVS2 and DNHS1. DNOVS2 responses during flight and quiescence are plotted in dark and light blue, respectively. DNHS1 responses during flight and quiescence are plotted in dark and light green, respectively. Averages at each azimuthal rotation axis is plotted as a dot. Line indicates least-squares fit for each set of responses. The slope of the least-squares fit for DNOVS2 flight and quiescence data are 8.2 and 8.8 Hz/mV, respectively. The slope of the least-squares fit for DNHS1 flight and quiescent data are 5.9 and 3.9 Hz/mV, respectively. G, Average tuning curves for the three descending neurons. Average responses to motion are shown for the entire range of rotation axes along the x-axis (every 15°) for each cell. DNOVS1 averages are shown in red, DNOVS2 in blue, and DNHS1 in green. The nonspiking DNOVS1 averages are plotted in terms of membrane potential (red y-axis) and DNOVS2 and DNHS1 averages are plotted as spike rates (black y-axis). The solid and dashed lines indicate average tuning responses measured during flight and quiescence, respectively. For DNOVS1 and DNOVS2, averages for flight and quiescence are derived from the same set of flies (n = 8 and n = 11, respectively). For DNHS1, tuning during quiescence was measured in four flies and during flight for two of these four flies. H, Same data as in G but normalized to the maximum response for each neuron during flight and quiescence and plotted in polar coordinates. The arrowhead indicates the peak tuning of each neuron determined by a sine least-squares linear model.

Figure 4.

Physiological responses of all neurons to a suite of motion stimuli during nonflight. Each neuron's responses are plotted in a single row (name indicated to the right of the traces). The icons at the top of each column indicate motion of the visual stimulus and these include rotation in the azimuthal plane at the peak response axis during nonflight (indicated to the left for each neuron), progressive motion, regressive motion, rightward yaw, leftward yaw, 1 Hz downward- and upward- moving sine-wave gratings, and 1 Hz rightward- and leftward- moving sine wave gratings. The stimulus directions are presented relative to the fly's perspective and all recordings were made from cells on the right side of the brain. The light-gray-shaded region indicates when the visual stimulus was in motion (2 s for azimuthal rotation, yaw, and translation and 4 s for 1 Hz vertically and horizontally moving gratings). Gray traces indicate single fly averages and the thick black lines indicate averages across flies. Responses for DNOVS2 and DNHS1 are plotted as spike rates (in Hertz) and all other responses are plotted on the same scale as membrane potential (scale bar on top row). The numbers of flies plotted for each neuron are as follows: DNOVS1, n = 8; DNOVS2, n = 10; VS1, n = 6; VS2, n = 13; VS3, n = 6; VS4, n = 5; VS5, n = 4; VS6, n = 7; DNHS1, n = 4; HSN, n = 3; HSE, n = 2; and HSS, n = 6.

Figure 5.

Rotation tuning of VS and HS cells. A, Membrane potential responses of the six VS cells to rotational stimuli centered at the cell's preferred axis during flight. Responses to motion corresponding to clockwise (preferred) rotation of the fly are plotted followed by responses to counterclockwise (nonpreferred) rotation. The light-gray-shaded regions indicate when the 2 s rotational visual stimulus was in motion. Averages across flies and for single flies are plotted as thick and (lighter) thin lines, respectively. Black traces show responses during quiescence and responses during flight are plotted in blue. Baseline membrane potential during quiescence is indicated. B, Membrane potential responses of the three HS cells to azimuthal rotational stimuli. Averages across flies and for single flies are plotted as thick and (lighter) thin lines, respectively. The gray-shaded regions indicate when the 2 s rotational visual stimulus was in motion. B has same scale as A. C, Rotation tuning curves during flight for the VS and HS cells across the entire range of azimuthal rotation axes measured (every 15°). Each neuron is plotted in a different color. This color, and the number of flies contributing to the average for each cell, is indicated in the legend to the right of the tuning curves. Normalized mean responses to rotational stimuli are plotted. D, Same data as in C, normalized and plotted in polar coordinates. The arrowheads indicate the peak tuning of each neuron determined by a sine least-squares linear model.

Figure 6.

Model and schematic of visual interneuron and visual descending neuron connectivity. A, Summary of known direct (black arrows), indirect (solid gray arrows), and hypothesized (dashed gray arrows) gap junction connections among VS cells, HS cells, and the three descending neurons. Thicker black arrows indicate the single best LPTC predictor for each descending neuron from the linear model in B. B, Linear prediction model of descending neuron rotation responses built using varying numbers of visual interneuron responses. We model DNOVS1 and DNOVS2 responses (Fig. 3G) as a linear sum of VS cell rotation tuning responses (Fig. 5C). DNHS1 responses (Fig. 3G) are modeled by a linear sum of HS cell responses (Fig. 5C). RSS values for the model built with responses during flight are connected by solid lines, quiescence by the dashed line. Including more than two neurons in each descending neuron's model adds only marginal gains in its predictive quality. C, Schematic diagram of the Drosophila central brain and thoracic ganglion showing the three descending neurons and VS and HS cells. Thoracic nerves associated with flight motor systems (neck, wing, and haltere) are shown and the pro-thoracic, meso-thoracic, and meta-thoracic neuromeres (PN, MN, and MtN, respectively) are labeled.

Figure 7.

Behavioral responses to rotational stimuli. A, Right wing amplitude. Below the fly icon, the peak response trace to motion (determined by a sine least-squares linear fit to the response across all rotation axes) and to rotation in the opposite direction are shown. Thick traces indicate the average response of the right wing amplitude across flies and the gray region shows SEM. The thin black bar indicates when the 2 s rotational visual stimulus was presented. On the lower right, the average responses to azimuthal rotational motion across all axes (every 15°) are plotted. Above, the same data are plotted in a normalized polar histogram and the peak axis of rotation is indicated by the arrowhead. The average descending neuron tuning responses are plotted in light colors behind the behavioral tuning curve for comparison. All data are plotted using the same right-hand rule convention as Figures 3 and 5. B, Left wing amplitude, plotted in the same manner as in A. C, Right wing deviation. Deviation of the leading tip of the wing is indicated by the red dot. D, Left wing deviation. E, Head yaw. Peak azimuthal rotation responses (“max” and “min”) are plotted to the left of responses to pure yaw motion (visual motion corresponding to the fly yawing to the left and right are indicated by arrows). F, Abdomen ruddering. Yaw responses are plotted to the right of maximum azimuthal rotation responses. G, Head pitch. H, Abdomen flexion. Note that the vertical scale for E through H is different from that for A through D. Averages for A–H were obtained in the same 56 flies for all traces. r² values for the sine fits in A–H are 0.94, 0.95, 0.93, 0.95, 0.73, 0.59, 0.61, and 0.60, respectively.