Harvard University COVID-19 updates

Department News



(L to R) Yifeng Zhang, Markus Meister, Joshua R. Sanes, In-Jung Kim

What does the eye tell the brain? Two very different answers have been proposed. In one account, the retina’s purpose is to quickly transmit the pixels of the visual image to the brain. There, a vastly greater number of neurons and circuits analyze the image. A contrasting account has suggested that the retina already extracts the specific information needed for certain visual behaviors and transmits a highly processed feature set upon which the brain can act more directly. This might explain why the retina’s output includes more than 20 different types of retinal ganglion cells (RGCs), many selective for specific visual features, for example, motion in a specific direction. In this view, the retina discards much of the raw visual information in favor of certain salient visual features that allow the creature to, for example, evade predators or capture prey. Of course both hypotheses could be true in part.

In the present work, Zhang and colleagues considered these two hypotheses for the retina of the mouse. Contrary to the propaganda spread by children’s songs, mice have a fully functional visual system, and they are increasingly popular in vision research owing to the ease of manipulating their genome. Here, Zhang et al. focused on a type of RGC called W3 that has high density and small receptive fields and thus would be a strong candidate for a generic pixel detector that transmits a high-resolution image to the brain. The responses of W3 cells were recorded to movies of visual scenes that a mouse encounters in its natural environment. Most RGCs responded to these movies, but surprisingly, W3 cells remained completely silent. Therefore, W3 cells—counter to expectation—do not act as generic encoders of the visual scene.

This intriguing observation led Zhang and colleagues to conduct a systematic survey of patterned stimuli. They learned that W3 cells are exquisitely sensitive to moving stimuli within a small center region of the receptive field. However, the firing of W3 cells was strongly suppressed in response to pattern motion in the surrounding region. These aspects of the W3 response account for the lack of signaling during the animal’s natural locomotion when the entire visual field experiences image flow. What natural visual input would present motion in a tiny region with no motion in the background? Zhang et al., reasoned that a likely scenario might result from an attack from aerial predators, viewed against a near featureless sky. Indeed, when the retina was exposed to images of birds flying overhead, the W3 cells fired strongly. Furthermore, the extreme selectivity of the response seems adapted to a role as alarm neurons. Therefore, the most numerous class of retinal ganglion cells in the mouse may serve as a highly selective feature detector for stimuli that threaten the animal’s survival.

Read the paper in PNAS

Download pdf

Listen to Professor Meister on the BBC Science in Action Podcast at minute 10:50–slightly past the halfway point of this 18 minute MP3 file.  (Note, this podcast is titled “16 Aug 12: Brain machine interfaces and was available on the BBC website until September 16, 2012.)