Fig 1.
A high-resolution, high-throughput assay for measuring Drosophila locomotor patterns.
(A) Schematic of planar behavioral arenas. Laminar flow of air or odor (10% acetic acid) is presented to either or both halves of the arena. Colored arrows indicate flow inlets and outlets (green/blue and red, respectively). (B) Camera-view of five experimental arenas and behavior tracking. Each fly is represented as a colored triangle. A colored line represents a fly’s location during the previous 10 s. (C) Schematic time-course of the behavioral experiment. We studied basal and odor-evoked locomotion as well as odor aversion for each fly. While not shown, during minutes 3–4.5, air flowed throughout the arena. (D) Speed time-series (black) were transformed into binary ‘Walking’ or ‘Stationary’ time-series (red). (E) Representative locomotor traces for five Canton-S strain flies during the odor impulse experiment. Flies were exposed to 60 s of air flow, 30 s of odor throughout the arena, and then 90 s of air flow. Behavior for each fly is shown in red. High values indicate walking while low values indicate stationary periods. (F) Locomotor traces averaged across 225 Canton-S flies during the odor impulse experiment. Prior to odor stimulation (grey bar) there is basal locomotion (green) followed by decay in locomotor frequency (cyan) to a reduced level of basal locomotion (magenta).
Fig 2.
Locomotor patterns for genetically distinct Drosophila strains.
(A) Locomotor frequency for 98 strains from the Drosophila melanogaster Genetic Reference Panel (DGRP) during the odor impulse experiment. Strains are sorted by average pre-odor basal locomotor frequency. (B) The duration of locomotor decay following odor removal for all DGRP strains. Strains are ordered as in panel A. (C) The ratio of post-odor to pre-odor basal locomotion for all DGRP strains. Strains are ordered as in panel A. The black dashed line indicates no change in basal locomotion. Values below this line represent reduced post-odor basal locomotor frequency. (D) The basal locomotor frequency for 65 randomly sampled flies (50% of flies for the strain with the smallest sample size: 130 flies) from each strain. The mean (light gray boxes) and standard deviation (black error bars) of 100 random samplings for each strain are shown. Strains are ordered as in panel A. (E) The correlation (R2) between odor-evoked locomotion time-series for groups of 65 randomly sampled flies taken from either the same strain (red) or from different Drosophila strains (blue). Strains are ordered as in panel A. The mean of 100 correlation measurements is shown. (F) Walking trajectories (black lines) along the long axis of the arena during an odor aversion experiment for 201 flies of the DGRP strain 78 (RAL85). Red bars indicate the half of the arena filled with odor. A histogram of odor aversion values for these flies is shown below. For each fly, odor aversion was calculated as the time spent in the odor zone subtracted from the time spent in air zone, divided by the total time of the odor aversion experiment. The median for this population of flies is indicated (black arrowhead). (G) A scatter plot showing the correlation between mean pre-odor basal locomotion and median odor aversion across all 98 strains (Pearson’s correlation coefficient R = 0.65, P < 10−4). A black dashed line indicates the best linear fit.
Fig 3.
Workflow for generating and analyzing neural network models.
(A) Basal locomotor patterns (filtered as in Fig 1D) are shown for two of ten Canton-S flies recorded for 30 min each. A stationary interval is highlighted for fly number ten. From these data, walking and stationary interval durations are aggregated and represented using weighted variable-width histograms for walking (top) and stationary (bottom) intervals. (B) A Continuous-Time Recurrent Neural Network (CTRNN) modeling framework used to investigate AS network dynamics. Models included up to five neurons, Gaussian noise inputs representing ongoing activity fluctuations, reciprocal and recurrent connections, odor inputs, and an output threshold to define the locomotor state of the virtual fly. (C) In the first step, a stochastic optimization approach is used to generate models that best reproduce Drosophila basal locomotor statistics. Locomotor statistics generated by a given model for 100 trials (virtual flies) are aggregated and compared to Canton-S strain data. The parameters for this model are then adjusted to reduce the difference from Canton-S data. This process is performed iteratively. (D) In the second step, odor inputs are added to the best previously generated models. For each model the strength of these inputs and the output threshold are adjusted iteratively to reduce the Root-Mean-Square Error (RMSE) between the model’s odor impulse locomotor pattern averaged across 200 trials (virtual flies) and a Drosophila strain’s odor impulse locomotor pattern averaged across ~200 flies. (E) In the third step, models that could match both Canton-S basal walking statistics and DGRP strain odor impulse locomotor patterns are characterized by the number of times specific neural activity levels are observed (‘Trajectory density’) color-coded from very frequent (dark red) to very rare (dark blue), and the tendency of neural activity, in the absence of fluctuations, to move toward (‘Stable’, cyan) or away from (‘Unstable’, orange) equilibrium activity levels (‘Phase portrait’). In each Trajectory density plot and Phase portrait, the bottom-left corner represents the lowest neural activity values.
Fig 4.
Small, fluctuation-driven models reproduce Drosophila locomotor patterns.
(A) The capacity of discovered models without (left) or with (right) fluctuating inputs (n = 250 models for each condition with 50 models of each size) to reproduce the basal locomotor statistics of Canton-S flies. (B) Fluctuation-driven models from panel A, dashed box, separated as a function of network size (n = 50 models for each size). (C) A dendrogram illustrating the similarity of odor-evoked locomotor patterns across 98 DGRP strains. Hierarchical clustering distance was based on the Pearson’s correlation coefficient between odor-response time-series for each strain. The three strains chosen for further analysis are color-coded cyan (strain A—RAL57), orange (strain B—RAL790), and red (strain C—RAL707). (D) A graph representation of the best model overall in panel B. This model was chosen for all subsequent analysis. Recurrent and reciprocal connection strengths are color-coded. The tau value for each neuron is shown in grey-scale. (E) Odor impulse locomotor patterns for the model in panel D (purple) optimized to match the odor impulse locomotor patterns of DGRP strains A (RAL57), B (RAL790), and C (RAL707). Locomotor frequency time-series for each strain are color-coded cyan, orange, and red, respectively.
Fig 5.
Dynamical mechanisms for reproducing Drosophila locomotor patterns.
(A) Network activity, locomotion, and population-averaged locomotor patterns in the absence (top) or presence (bottom) of fluctuations. ‘Individual network activity’ plots show neural activity trajectories during odor stimulation starting from ten color-coded initial conditions. Solid circles indicate the starting points of each trajectory. ‘Individual locomotion’ plots show locomotor patterns for two representative virtual flies in the absence (top) or presence (bottom) of fluctuations. ‘Population average’ shows the locomotor frequency averaged across 200 virtual flies in the absence (top) or presence (bottom) of fluctuations. (B) Odor-evoked locomotor pattern of the best fluctuation-driven model (Class 1) tuned to match the locomotor pattern of Drosophila strain A. Color-coded are periods of pre-odor basal locomotion (green), odor impulse (blue), locomotor decay (cyan) and post-odor reduced basal locomotion (magenta). (C) Trajectory density plots and (D) phase portraits for this model during each time period indicated in panel B. In trajectory density plots, arrowheads highlight increased neural activity not observed during pre-odor basal locomotion. These are further labeled as being above (white) or below (red) the threshold for walking. In all phase portraits, grey lines with arrows are trajectories that indicate the direction of flow in phase space.
Fig 6.
Fluctuations linearize the relationship between sensory drive and locomotor response.
(A) Odor-evoked locomotor patterns as a function of odor stimulus amplitude for models without (top), or with ongoing fluctuations (bottom). (B) Maximum locomotor frequencies observed for models without (red), or with (blue) ongoing fluctuations at different odor stimulus amplitudes. Shown are the mean and standard deviations for 5 repetitions of each experiment.