Cephalopods are highly visual animals with camera-type eyes, large brains, and a rich repertoire of visually guided behaviors. However, the cephalopod brain evolved independently from that of other highly visual species, such as vertebrates, and therefore the neural circuits that process sensory information are profoundly different. It is largely unknown how their powerful but unique visual system functions, since there have been no direct neural measurements of visual responses in the cephalopod brain. In this study, we used two-photon calcium imaging to record visually evoked responses in the primary visual processing center of the octopus central brain, the optic lobe, to determine how basic features of the visual scene are represented and organized. We found spatially localized receptive fields for light (ON) and dark (OFF) stimuli, which were retinotopically organized across the optic lobe, demonstrating a hallmark of visual system organization shared across many species. Examination of these responses revealed transformations of the visual representation across the layers of the optic lobe, including the emergence of the OFF pathway and increased size selectivity. We also identified asymmetries in the spatial processing of ON and OFF stimuli, which suggest unique circuit mechanisms for form processing that may have evolved to suit the specific demands of processing an underwater visual scene. This study provides insight into the neural processing and functional organization of the octopus visual system, highlighting both shared and unique aspects, and lays a foundation for future studies of the neural circuits that mediate visual processing and behavior in cephalopods.

Calcium imaging was performed with a two-photon microscope (Neurolabware Inc.), using a 16X Nikon CFI75 LWD objective, via the Scanbox software package for Matlab (MATHWORKS). Data were acquired at a 10Hz framerate, with an 800x800μm (796x796 pixel) field of view. Recordings were taken at 90 to 170μm depths from the dorsal surface of the optic lobe.  Custom generated visual stimuli, rendered using the PsychToolbox package for Matlab72, were displayed with a pico LCD projector (AAXA Technologies) onto the diffusing glass on the side of the recording chamber. To avoid light from the stimulus entering the two-photon detection pathway, the projected light was passed through a 450/50 bandpass filter (Chroma Technology Corporation), avoiding overlap with the emission spectrum of the Cal-520 calcium dye and the bandpass 525/50 emission filter of the microscope. The stimulus bandpass filter also restricted the stimulus light to be within the known absorption spectrum of cephalopod photopigments26. Stimuli were gamma-corrected in software and presented at 60FPS.  Full RF mapping was performed using a sparse noise stimulus, consisting of white and black spots (radius = 3, 6, 12 deg; density = 10%) on a gray (50% luminance) background, along with full-field white or black on 2% of frames. Each stimulus frame was presented for 1sec in a randomized order for a total duration of 10min.

Data analysis was performed using custom software in MATLAB. We applied a rigid alignment of imaging data using the sbxalign function in Scanbox (Neurolabware, Inc.). In order to detect large movements that were not corrected by the alignment algorithm, for each frame we calculated the pixel-wise correlation coefficient to the mean image. Frames with less that 90% correlation were discarded from further analyses. 

We calculated the fluorescence activity (dF/F0) at each pixel as the standard (F(t) - F0) / F0, where F(t) is the fluorescence intensity of the pixel on each frame and F0 is the median fluorescence intensity of the pixel across the recording. To analyze local responses, we defined “units” as a 20μmx20μm wide square window, centered on local peaks within the mean fluorescence that were above the background fluorescence, to ensure that only areas with sufficient dye loading were analyzed. dF/F0 for each unit was calculated as the mean dF/F0 across pixels within the unit. Units were manually assigned to anatomical layers (OGL, IGL, Plex, and Med) based on location within the mean fluorescence image from the recording session.

Data files are in Matlab .mat format, which is based on HDF5 and is readable across languages.