Ancestral circuits for vertebrate colour vision emerge at the first retinal synapse

Abstract

For colour vision, retinal circuits must separate information about intensity and wavelength. This requires circuit level comparison of at least two spectrally distinct photoreceptors. However, many vertebrates use four or more, and in those cases the nature and implementation of this computation remains poorly understood. Here, we establish the complete circuit architecture and function of outer retinal circuits underlying colour processing in the tetrachromatic larval zebrafish. Our findings reveal that the specific spectral tunings of the four cone types near optimally rotate the encoding of natural daylight in a principal component analysis (PCA)−like manner to yield one primary achromatic axis, two colour opponent axes as well as a secondary UV-achromatic axis for prey capture. We note that fruit flies − the only other tetrachromat species where comparable circuit-level information is available − use essentially the same strategy to extract spectral information from their relatively blue-shifted terrestrial visual world. Together, our results suggest that rotating colour space into achromatic and chromatic axes at the eye′s first synapse may be a fundamental principle of colour vision when using more than two spectrally well separated photoreceptor types.

Distinct Synaptic Transfer Functions in Same-Type Photoreceptors

“…2a,b, Methods). UV-cones were unambiguously identified based on their robust responses only to UV-light (Methods, (Yoshimatsu et al, 2020a)). We then concurrently recorded red and green fluorescence signals under two-photon during presentation of 100% contrast widefield flashes of UV-light (3 s On, 3 s Off), starting from a 50% contrast background (Fig.…”

“…To unequivocally identify UV-cones, responses to the ‘red’, ‘green’, ‘blue’ and ‘UV’-flashes were always recorded (always 1 s flash, 1 s darkness). Only cones that preferentially responded to the UV-LED were kept for further analysis (Yoshimatsu et al, 2020a).…”

“…We applied the Sequential Neural Posterior Estimation method described in (Lueckmann et al, 2017) (code available at https://github.com/mackelab/delfi) (also called SNPE-B) with some modifications which were also applied in (Oesterle et al, 2020; Yoshimatsu et al, 2020a). In brief, SNPE-B draws over several rounds samples { θi} iϵl from a prior and evaluates the model for these parameters.…”

“…To establish possible structural differences amongst larval zebrafish UV-cone ribbon synapses, we obtained volumetric electron microscopy datasets of the outer retina taken from three different regions: Acute zone (AZ), Nasal (N) and Dorsal (D). For each region, we anatomically identified (Yoshimatsu et al, 2020a) and 3Dreconstructed UV-cone pedicles (nAZ, N, D = 6, 6, 6) including their full complement of ribbons and surrounding vesicles Methods). This revealed that dorsal ribbons were smaller ( Fig.…”

“…Parameter Inference. We applied the Sequential Neural Posterior Estimation method described in (Lueckmann et al, 2017) (code available at https://github.com/mackelab/delfi) (also called SNPE-B) with some modifications which were also applied in (Oesterle et al, 2020;Yoshimatsu et al, 2020a We took the same approach for setting an adaptive bandwidth for the kernel. As an additional post-hoc verification of the posteriors, we took as final posterior distributions the posterior of the round with the smallest median discrepancy of its samples ("early stopping").…”