6). Regions of interest (ROI) extraction, background subtraction, and

brightness normalization (ΔF/F0) were performed in Igor Pro 6.2 and facilitated by SARFIA analysis routines (Dorostkar et al., 2010). Fluorescence traces were then sorted and analyzed by custom-made scripts and NeuroMatic. The detection of active ROI in the IPL high throughput screening assay was based on the thresholding of the Laplacian Transform of the two-photon recordings. In this way, responding bipolar cell terminals and active areas of the ganglion cell dendrites were identified in ribeye::SyGCamp2 and eno2::GCamp3.5 fish, respectively. The responses to light of bipolar cell terminals and retinal ganglion cell dendrites were characterized according to their response amplitude, i.e., the variation in fluorescence during stimulation in comparison to baseline (ΔF/F0). Responses to CH5424802 light were plotted in full, as in Figure 1B, left, or in stimulus versus amplitude plots (e.g., Figure 1B, right). In the case of traces representing single terminals (e.g., Figure 1B), the error curve (gray shadow in Figure 1B) represents the SEM of the four trials employed to assess the terminal responsiveness (see

Stimulation Protocols). In the case of traces representing whole populations of terminals (e.g., Figure 1D), the error curve represents the standard error of all the responses employed to generate the final average. As described in the stimulation protocols section, a stimulus could be light intensity, contrast, or frequency. Intensity versus amplitude plots were obtained by averaging amplitude values over 300 ms long time windows

around the maximum response occurring during the stimulation time (e.g., Figure 1B, right). Contrast versus amplitude and frequency versus amplitude plots were obtained by averaging amplitude values over the whole stimulation period (e.g., Figures 2B and S2D, respectively). The intensity versus amplitude plots were fitted with Hill curves, in the form A = Ih/Ih + I1/2h, A being the response amplitude, I the stimulation intensity, h the Hill coefficient, and I1/2 Thymidine kinase the sensitivity at half maximum, i.e., the stimulation intensity that elicits half of the maximum response. I1/2 has been used as a metric for the sensitivity of each intensity versus amplitude curve. Contrast versus amplitude plots were fitted with power functions, in the form A = k × Cα being A the response amplitude, k a constant, C the stimulation contrast, and α the power exponent. The sensitivity shift induced by olfactory stimulation for each individual terminal (e.g., Figure 1F) was measured by comparing the values of the lowest light intensity eliciting a statistically significant response before and after methionine administration. The statistical significance of a response was assessed by comparing (t test) the average calcium level during light stimulation with a threshold defined as three times the SD of a baseline epoch.