Using ERG, the electrical activity of the retina in response to a light stimulus is measured. The ERG arises from currents generated by the retinal neurons and glia. The animals were kept in total darkness in their home cage for at least 1 hour before scotopic ERG measurements and were anesthetized with an intraperitoneal injection of a mixture of ketamine (22 mg/kg for animals <4 weeks of age; 65 mg/kg for animals ≥4 weeks of age) and xylazine (2.2 mg/kg for animals <4 weeks of age; 7.5 mg/kg for animals ≥4 weeks of age) diluted in 0.9% NaCl. The eyes were locally anesthetized using tetracaine-hydrochloride drops (1% w/v) and were dilated using tropicamide (0.5% w/v), and atropine (1% w/v) drops. Hylocomod drops were applied to maintain corneal hydration at all times. The animals were placed in the RETImap full flash Ganzfeld (Roland Consult, Brandenburg an der Havel, DE) using a carrier table, which was kept at 37 °C. Body temperature was carefully monitored during all measurements. ERGs were recorded using gold electrodes, which were placed on the corneas of both eyes. Another gold electrode was placed in the animal’s mouth serving as a reference for both eyes simultaneously. A needle was placed subcutaneously near the tail, which served as a ground electrode. See Supplementary Table S1 for the light intensities, the number of flashes used for averaging, and the flashes’ interval. ERG traces were 350 ms long, utilizing 512 data points.

All ERG data were systematically analyzed, without human intervention, using a custom-made Matlab script. The data was zero-centered by averaging the signal before the stimulus (<20 ms) and subtracting the resultant from the entire trace. A low-pass filter (4th order, 30 Hz (for the b-wave) and 235 Hz (for the a-wave)) was applied in both the forward and backward direction to remove noise and the oscillatory potentials (OPs). 30 Hz is well below the minimum expected frequency, and 235 Hz resembles the expected maximum frequency of OPs in rats [37]. The findpeaks function in Matlab was used to find the latencies of the a- and b-waves in the filtered data. The magnitudes of the unfiltered signal at the selected latencies were characterized as the values for b-wave and the absolute a-wave amplitude. The absolute a-wave was subtracted from the value of the b-wave amplitude to calculate the absolute b-wave amplitude. Flicker properties were determined from the original, unfiltered trace. The time to the first peak (P1) and the second peak amplitude (P2) were identified. The (absolute) b-wave, a-wave, and flicker properties of each group, at each age, were averaged and normalized to the corresponding 30 cd· s/m2 response from the wildtype control group (Supplementary Figure S5).