Fig 1.
Contrast normalization in early visual system of Drosophila.
(A) Schematic representation of early stages of the motion detection circuit. Highlighted are contrast normalization-exhibiting neurons that provide major input to T4 and T5 cells, the first direction-selective neurons in the ON and OFF pathways, respectively. (B) Contrast normalization experimental protocol and contrast tuning curves for different medulla neurons: Mi1 (n = 20 cells/5 flies), Tm3 (n = 21 cells/8 flies), Tm1 (n = 21 cells/7 flies), and Tm2 (n = 20 cells/6 flies). Shaded areas around the curves show bootstrapped 68% confidence intervals. Adapted from Drews et al. (2020) [19].
Fig 2.
Dynamic surround affects response amplitude and kinetics.
(A) Experimental procedure: (1) white noise stimulus, (2) receptive field (RF) reconstruction from single-neuron calcium signals, (3) experimental stimuli centered on RF. (B) Stimulus protocol. Luminance step in RF center with 4 surround conditions. (C) Average Tm3 response to luminance step in RF center with gray, stationary, moving grating, and stochastic stimulus surround; n = 23 cells/6 flies. Luminance step happened during the gray-shaded period. Left: Amplitudes of cell responses. Shaded areas around the curves show bootstrapped 68% confidence intervals. Right: Kinetics of cell responses. Responses during each condition are normalized to the condition’s maximum. Artificial gap is created between responses to static conditions for easier visualization. See also S1 Fig.
Fig 3.
Contrast normalization affects response amplitude and kinetics.
Average responses of contrast normalization-exhibiting neurons to luminance step in the RF center with gray and moving grating surround. Mi1 (n = 98 cells/25 flies); Tm1 (n = 24 cells/9 flies); Tm2 (n = 22 cells/7 flies), and Tm3 (n = 65 cells/16 flies). Mi1 and Tm3 datasets include the data from Figs 2 and 5. Luminance step happened during the gray-shaded period. (A) Amplitudes of cell responses. Shaded areas around the curves show bootstrapped 68% confidence intervals. (B) Kinetics of cell responses. Responses during each condition are normalized to the condition’s maximum.
Fig 4.
Model with dynamic nonlinearity reproduces dynamic surround effect on response amplitude and kinetics.
(A) Model schematic: input step is sequentially passed through band-pass filter (BP), nonlinearity (NL), and low-pass filter (LP). NL with k = 1 corresponds to moving grating surround condition, k = 0.2 to gray surround condition. (B) Model cascade with stationary nonlinearity. (C) Model cascade with stationary nonlinearity and varying input amplitudes. (D) Model cascade with dynamic nonlinearity and varying input amplitudes. (B-D) For easier visualization, an artificial gap is created between model responses to moving grating and gray surround conditions.
Fig 5.
Luminance step amplitude has no effect on response kinetics.
(A) Spatial and temporal stimulus profile. Left: 3 luminance step amplitudes with gray surround. Right: 3 luminance step amplitudes with moving grating surround. (B) Tm3 responses to 50%, 75%, and 100% luminance steps with gray surround; and to 50%, 75%, and 100% luminance steps with moving grating surround; n = 21 cells/4 flies. Luminance steps happened during the gray-shaded period. Left: Amplitudes of cell responses. Shaded areas around the curves show bootstrapped 68% confidence intervals. Right: Kinetics of cell responses. Responses during each condition are normalized to the condition’s maximum. An artificial gap is created between responses to 100% and 75% steps for easier visualization. (C) Mi1 responses to 50%, 75%, and 100% luminance step with gray surround; and to 50%, 75%, and 100% luminance step with moving grating surround; n = 26 cells/8 flies. Luminance step happened during the gray-shaded period. Left and right as in (B). See also S2 Fig.