8) (Figure 1A). Emission was captured both by the objective and a substage oil condenser (Olympus,

NA = 1.4), through GFP emission filters (HQ 535/50, Chroma Technology) before detection with photomultiplier tubes (Hamamatsu). Laser scanning and image acquisition were controlled using Antidiabetic Compound Library research buy ScanImage v. 3.0 (Pologruto et al., 2003). Light stimuli were generated by an amber LED and 600/10 BP filter and delivered through a light guide placed close to the eye of the fish. Stimulation was synchronized to image acquisition through Igor Pro v. 4.01 software. The mean intensity of the stimulus was controlled by neutral density filters and modulations around this mean by a custom-built LED driver which switched the driving current Osimertinib molecular weight at 10 kHz while adjusting the duty cycle. The unattenuated stimulus was ∼5.5 × 105 photons/μm2/s, and a period of 40 s dark adaptation was interleaved between each presentation of a stimulus. Data were obtained from 42 fish. Movies were analyzed using SARFIA, a custom-written suite of macros running in Igor Pro (Dorostkar et al., 2010). First, movies were registered to correct small lateral movements but were rejected if the plane of focus altered significantly. Next, images were transformed using a Laplace

operator and segmented by applying a threshold. The ability of this algorithm to define ROIs corresponding to individual terminals is shown in Figure 3 of Dorostkar et al. (2010). The cross-sectional area of each ROI was measured and the sypHy or SyGCaMP2 signals quantified as the average fluorescence per unit area, after background subtraction. In some terminals, a small linear correction for bleaching

was applied, as shown in Figure S2. Terminals were only used all for analysis if the response to a step of bright light occurred with a SNR > 4 when imaging at 4 Hz, or SNR > 2 when imaging at 8 Hz. Measurements using sypHy were carried out on a total of 1021 ON and 1995 OFF terminals. Measurements using SyGCaMP2 were carried out on a total of 60 ON and 132 OFF terminals. Calculations of release rates involved differentiation of the sypHy signal (Equation 1) resulting in an amplification of noise. We therefore calculated the initial rate of release simply by fitting a line to the first 2 or 4 s of the response to a step of light or contrast respectively. For ON terminals, the value of Fmin for each terminal was calculated in the dark, and for OFF terminals it was calculated over the last 10 s of a 40 s step at ND 1 (see Figure 3D, top graph). To assess the degree to which the luminance tuning curves were linear, the Hill equation was fit to the relation between luminance and the initial rate of release at light onset.