C2 dominates the expression pattern of driver lines with deficits in motion vision upon neuronal silencing.

a-c) Four examples of InSITE lines with behavioral deficits to OFF motion stimuli expressing the columnar feedback neuron C2. Shown are the full expression pattern (a, scale bar = 20µm, 10µm), single cell clones (b, scale bar = 10µm) and the gad1 intersection pattern (c, scale bar = 20µm), where the InSITE and gad1 intersection (InSITE-Gal4 UAS-LexADBD, Gad1-p65AD, lexAop-GFP) is shown in green (GFP), and the neuropil is marked with nc82 (blue). The full InSITE-Gal4 expression pattern is additionally visualized with RFP. d) Three examples of InSITE lines with behavioral deficits to ON motion stimuli screened for neurons within the gad1 intersection pattern. e) Drawing of the two GABAergic neurons C2 (magenta) and C3 (green) in the fly visual system including the four neuropiles: lamina (La), medulla (M), lobula (Lo) and lobula plate (Lp). The local motion detectors T4 and T5 are additionally visualized in grey.

C2 and C3 connect to circuitry upstream of motion-sensitive neurons.

a) Confocal images of C2 cells labelled with the dendritic marker DenMark and a GFP-tagged Synaptotagmin (syt::EGFP) to label pre-synapses in C2 (scale bar = 20µm). b) Counts of synapses with neurons postsynaptic and presynaptic to C2 or C3. EM data. c) Illustration of the C2 and C3 circuitry within the motion detection pathway. Data in (b-c) were extracted from the FAFB EM dataset and Flywire connectome (Dorkenwald et al. 2024).

Response properties of C2 and C3 to visual stimulation.

a) Calcium responses of C2 (magenta, N=16 flies, 181 cells) and C3 (green, N=12 flies, 149 cells) neurons in medulla layer M1 to full field ON and OFF flashes. b) Polar plots showing response amplitude of C2 (N=6 flies, 85cells) and C3 (N=8 flies, 77cells) to ON and OFF edges moving onto eight different directions. c) Average aligned spatiotemporal receptive fields (STRFs) of C2 and C3 cells extracted from responses to horizontal or vertical ternary noise bars. Red and blue color indicate a positive and negative correlation with the stimulus, respectively. Sample size (N) equals number of cells. d) Temporal filter extracted from averaging single STRFs along the time axis of the horizontal and vertical STRFs. e) Timing of the ON peak of the temporal filter (left) and the full width half maximum (FWHM) of a Gaussian fit extracted along the spatial dimension of maximal response of single STRFs (right). Shown are mean ± standard error. Significances are based on Wilcoxon rank sum test (p≤0.05 *, p≤0.01**, p≤0.001***).

C2 shapes physiological properties of Mi1 neurons upstream of T4.

a) Calcium responses of Mi1 axon terminals for control condition (grey, N=5 flies, 24 cells) or while genetically silencing C2 using shibirets (magenta, N=3 flies, 20 cells). b) Left: Average aligned spatiotemporal receptive fields (STRFs) of Mi1 cells extracted from responses to horizontal or vertical ternary noise bars. Sample size (N) equals number of cells. Right: Temporal filter extracted from averaging single STRFs along the time axis of the horizontal and vertical STRFs. Timing of the ON peak of the temporal filter (top) and the full width half maximum (FWHM) of a gaussian fit extracted along the spatial dimension of maximal response of single STRFs (bottom). Box plots show median, the interquartile range, and whiskers show min and max of the data. Significances are based on Wilcoxon rank sum test (p≤0.05 *, p≤0.01**, p≤0.001***).

C2 and C3 disinhibit responses of motion detectors T4 and T5 to full field flash stimuli.

a) Schematic of visual circuitry illustrating C2 and C3 block by expressing Kir 2.1. In vivo calcium responses were recorded from axon terminals of T4 and T5 neurons (rectangles) in response to full field ON and OFF light flashes for controls (grey, N=9 flies), upon C2 block (magenta, N=8 flies) or upon C3 block (green, N=10 flies). b) Calcium response of T4 neurons to the onset of light (ON step) and T5 neurons to the offset of light (OFF step) quantified from data in (a). c) Time to peak of the ON response of T4 upon C2 and C3 block (a) and decay rate quantified from fitting an exponential function to the decay of the ON response. Box plots show median, the interquartile range, and whiskers show min and max of the data. Significances are based on Kruskal-Wallis Test (p≤0.05 *, p≤0.01**, p≤0.001***, + Bonferroni correction for multiple testing).

C2 and C3 are required for direction-selective responses of T4 and T5 cells.

a) Compass plots showing direction tuning of T4 and T5 neurons extracted from responses to ON and OFF bars moving into eight different directions for the control (UAS-Kir2.1), C2 block (C2>>Kir2.1), C3 block (C3>>Kir2.1) and C2/C3 double block (C2&C3>>Kir2.1) conditions. Vector length corresponds to the strength of DS tuning. b) Direction selectivity averaged across cells and flies from all layers for control condition (grey, N=7 flies, 370 T4 cells, 196 T5 cells), C2 block (magenta, N=8 flies 170 T4 cells, 196 T5 cells), C3 block (green, N=6 flies, 188 T4 cells, 117 T5 cells) or double C2/C3 block (blue, N=7 flies, 132 T4 cells, 78 T5 cells). Box plots show median, the interquartile range, and whiskers show min and max of the data. c) Response amplitude (dF/F) of T4 and T5 cells in relation to the angular distance of stimulus motion direction to the neurons PD. Shown are mean ± standard error. Significances are based on ANOVA (p≤0.05 *, p≤0.01**, p≤0.001***) (b) or Kruskal Wallis test (c) and with Bonferroni correction for multiple testing. In (c) significant comparisons to the control condition are indicated by asterisks color-coded by genotype (p≤0.05 *).

C2 is required for higher temporal resolution of the behavioral responses to moving ON edges.

a) Schematic of the stimulus. Single moving ON edges were presented for 0.75s, interleaved by 1.5s of darkness. b) Time traces of control and C2-silenced flies, to the stimulus in (a). c) Deceleration of the turning velocity in the post-peak interval (0.45s-0.75s) of the time traces in (b). *p < 0.05, two-tailed Student’s t tests against both controls. d) A stimulus epoch comprised of two moving ON edges of 0.75s duration each, presented one after the other without delay, and with 1.5 s darkness between two epochs. The ON edge luminance was 9806.3 photons s-1 receptor-1. e) Time traces of control and C2-silenced flies, to the stimulus in (d). f) Percent recovery from the turning response elicited by the first of the moving edge pair. **p < 0.01, two-tailed Student’s t tests against both controls. g) A stimulus epoch comprised of two moving ON edges of 0.5s duration each, presented one after the other without any delay, and with 1.5 s darkness between two epochs. h) Time traces of control and C2-silenced flies, to the stimulus in (g). i) Percent recovery from the turning response elicited by the first of the moving edge pair. **p < 0.01, two-tailed Student’s t tests against both controls. j-k) Stimuli structured similar to the one used in (d) (in j) or in (g) (in k) were presented at two dimmer ambient light conditions (ON edge luminance of 1225.8 and 153.2 photons s-1 receptor-1, respectively). The left and the middle panels show percent recovery from the turning response elicited by the first of the moving edge pair in the dimmer and the dimmest conditions. The right panel compares the recoveries across all three light conditions. *p < 0.05, **p < 0.01, ns: non-significant, two-tailed Student’s t tests against both controls. All data show mean ± SEM. For (b-c), n = 15 (UAS-shits/+), n =18 (C2-Gal4/+), n =12 (C2>>shits) flies. For E-F and H-I, n = 10 flies each genotype. For (j, left), n = 8 (UAS-shits/+), n = 9 (C2-Gal4/+), n = 8 (C2>>shits) flies. For (j, middle), n = 7 (UAS-shits/+), n = 9 (C2-Gal4/+), n = 9 (C2>>shits) flies. For (k, left), n = 7 (UAS-shits/+), n = 9 (C2-Gal4/+), n = 9 (C2>>shits) flies. For (k, middle), n = 7 (UAS-shits/+), n = 9 (C2-Gal4/+), n = 9 (C2>>shits) flies. Gray patches in each stimulus schematics mark the corresponding motion durations in the response-time traces in the following panel.

Key resources table

List of primary and secondary antibodies used

Response properties of C2 and C3.

Statistical summary of Wilcoxon rank sum test. Sample Size (N) is given in number of cells (C).

Response properties of Mi1.

Statistical summary of Wilcoxon rank sum test. Sample Size (N) is given in number of cells (C).

Response properties of T4 and T5 to full field flashes.

Statistical summary of ANOVA.

Flash responses in T4/T5 upon blocking C2 or C3.

Statistical summary of multi comparisons with Bonferroni correction. Sample Size (N) is given in number of flies (F).

Direction tuning T4 and T5.

Statistical summary of ANOVA. Related to Fig. 6.

Statistical summary of multi comparisons with Bonferroni correction.

Sample Size (N) is given in number of cells (C) or flies (F).

Statistical summary of two-tailed Students’s t-tests for Figure 7 and Supp Fig 5.

Screen for behaviorally relevant GABAergic neurons.

a) Schematic illustrating the procedure of screening for behaviorally relevant GABAergic neurons in the visual system of Drosophila. InSITE Gal4 driver lines were selected based on a reduced rotation behavior to moving OFF stimuli when the targeted neurons were silenced. GABAergic neurons within the expression pattern were identified by screening for colocalization of the InSITE expression pattern (GFP, green) with anti-GABA (magenta) staining. b) Same examples of InSITE lines as in Fig. 1a-c with behavioral deficits to OFF motion stimuli screened for GABA (magenta) colocalization with the InSITE expression pattern labeled with GFP (green) (scale bar = 20µm, 10µm) c) InSITE Gal4 driver lines with a deficit for ON motion behavior were further screened for GABAergic neurons using a gad1 intersection strategy. GABAergic neurons within the expression pattern were labelled with GFP (green) and the whole expression pattern with RFP (red). d) Examples of InSITE lines with behavioral deficits to ON motion stimuli screened for neurons within the gad1 intersection pattern. The neuropil is marked with nc82 (blue), the InSITE expression pattern is marked with RFP (red) and the GABAergic neurons are additionally labeled with GFP (green), (scale bar = 20µm). Besides C2 we identified a large tangential cell (LT) located in the lobula and a medulla neuron, with a branching pattern reminiscent of the medulla intrinsic neuron Mi1new (Supplementary Table 1, (Raghu et al. 2011).

C2 and C3 responses are weaker in proximal medulla layers.

a) Confocal image of C2 and C3 split gal4 driver lines expressing GFP (green). The neuropil is marked with nc82 (blue). Scale bar = 20µm. b) Calcium responses of C2 (magenta) and C3 (green) neurons in different medulla layers M1, M5, M8, M9 and the cell bodies (CB) to full field ON and OFF flashes. Sample Size (N) refers to number of flies.

C2 and C3 suppress flash responses in T4 and T5 neurons from different lobula plate layers

a) ROIs of single T4 and T5 axon terminals extracted mathematically from calcium imaging data. b) Number of ROIs (cells) extracted for all flies for control condition (grey, N=9 flies), C2 block (magenta, N=9 flies), C3 block (green, N=10 flies) and C2/C3 double block (blue, N=10 flies). c) Calcium responses of T4 and T5 neurons to full field ON and OFF light flashes for C2/C3 double block. d) Response change of T4 neurons to the onset of light (ON step) and T5 neurons to the offset of light (OFF step) quantified from (c). e) Response change of T5 neurons to the onset of light (ON step). f) Calcium responses of T4 and T5 neurons extracted from different layer of the lobula plate to full field ON and OFF light flashes. Shown are mean ± standard error. Significances are based on ANOVA (p≤0.05 *, p≤0.01**, p≤0.001***, + Bonferroni correction for multiple testing).

C2 and C3 are required for direction-selective responses of T4 and T5 cells from all lobula plate layer.

Direction selectivity averaged across cells and flies, for different layer of the lobula plate. Control condition (grey, N=7 flies), C2 block (magenta, N=8 flies), C3 block (green, N=6 flies) and double C2/C3 block (blue, N=7 flies). Significances are based on KKW (p≤0.05 *, p≤0.01**, p≤0.001***, + Bonferroni correction for multiple testing).

C2 temporally sharpens optomotor responses to moving ON edges in bright light conditions.

a) Schematic of the stimulus. Single moving ON edges (9806.3 photons s-1 receptor-1) were presented for 0.75s, interleaved by 1.5s of darkness (same as Fig. 7a). b) Time traces of control and C3-silenced flies, to the stimulus in (a). c) Deceleration of the turning velocity in the post-peak interval (0.45s-0.75s) of the time traces in (b). ns: non-significant, two-tailed Student’s t tests against both controls. d) Schematic of the stimulus. A stimulus epoch comprised two moving ON edges of 0.75s duration each, presented one after the other without any delay (same as in Fig. 7d, but with a dimmer ON edge luminance of 1225.8 photons s-1 receptor-1). The interval between two epochs lasted 1.5s. e) Time traces of control and C2-silenced flies, to the stimulus in d. f) Percent recovery from the turning response elicited by the first of the moving edge pair illustrated in (d). *p < 0.05, two-tailed Student’s t tests against both controls. g-i) Same as in (d-f), in the dimmest light condition (the ON edge luminance of 153.2 photons s-1 receptor-1). ns: non-significant, two-tailed Student’s t tests against both controls. j) A stimulus epoch comprised two moving ON edges of 0.5s duration each, presented one after the other without any delay (same as in Fig 7g, but with the ON edge luminance same as in d). The interval between two epochs was still 1.5s long, with dark screen. k) Time traces of control and C2-silenced flies, to the stimulus in (g). l) Percent recovery from the turning response elicited by the first of the moving edge pair illustrated in j. ns: non-significant, two-tailed Student’s t tests against both controls. m-o) Same as in (j-i), in the dimmest light condition (the ON edge luminance same as in g). ns: non-significant, two-tailed Student’s t tests against both controls. All data show mean ± SEM. For (b-c), n = 15 (UAS-shits/+), n = 13 (C3-Gal4/+), n = 16 (C3>>shits) flies. For (e-f), n = 8 (UAS-shits/+), n = 9 (C2-Gal4/+), n = 8 (C2>>shits) flies. For (h-i), n = 7 (UAS-shits/+), n = 9 (C2-Gal4/+), n = 9 (C2>>shits) flies. For (k-l), n = 7 (UAS-shits/+), n = 9 (C2-Gal4/+), n = 9 (C2>>shits) flies. For (n-o), n = 7 (UAS-shits/+), n = 9 (C2-Gal4/+), n = 9 (C2>>shits) flies. Gray patches in each stimulus schematics mark the corresponding motion durations in the response-time traces in the following panel.

List of behaviorally relevant neurons identified from the expression pattern of InSITE lines with behavioral deficits to either OFF- or ON edge motion stimuli.

Neurons were identified based on either colocalization of the InSITE expression pattern with a GABA antibody, followed by single cell Flp-Out experiments, or a InSITE-Gal4-Gad1-intersection strategy.