This upward motion–detecting ganglion cell (red with yellow center) from the mouse retina helps rodents see movement in the dark.

Yao et al., Neuron 10.1016 (2018)

How humans—and other mammals—might have gotten their night vision

On a moonless night, the light that reaches Earth is a trillion–fold less than on a sunny day. Yet most mammals still see well enough to get around just fine—even without the special light-boosting membranes in the eyes of cats and other nocturnal animals. A new study in mice hints at how this natural night vision works: Motion-sensing nerve cells in the retina temporarily change how they wire to each other in dark conditions. The findings might one day help visually impaired humans, researchers say.

Scientists already knew a bit about how night vision works in rabbits, mice, humans, and other mammals. Mammalian retinas can respond to “ridiculously small” numbers of photons, says Joshua Singer, a neuroscientist at the University of Maryland in College Park who was not involved in the new study. A single photon can activate a light-sensitive cell known as a rod cell in the retina, which sends a minute electrical signal to the brain through a ganglion cell.  

One kind of ganglion cell specializes in motion detection—a vital function if you’re a mouse being hunted by an owl, or a person darting to avoid oncoming traffic. Some of these direction-selective ganglion cells (DSGCs) get excited only when an object is moving upward. Others fire only when objects are moving down, or to the left or right. Together, the cells decide where an object is headed and relay that information to the brain, which decides how to act.

DSGCs “stand out as one of the few places in the brain” where neuroscientists feel pretty confident they know what neurons are doing, Singer says. But the cells behave in surprising ways when the lights go down.

To find out how DSGCs adapt to the dark, neuroscientist Greg Field and colleagues at Duke University in Durham, North Carolina, examined slices of mouse retinas by laying them on tiny glass plates embedded with an electrode array. Each array includes about 500 electrodes, but is so small that it spans just a half-millimeter, Field says. Bathed in an oxygenated solution, the mouse retinas can still function and “see” while the array records electrical activity from hundreds of neurons.

The team showed the dissected retinas a simple movie—bands moving across a contrasting background—then turned the light down by a factor of 10,000, going from typical office-level lighting to a more moonlit scene. Three of the four directional DSGCs remained “rock solid” in their response to the motion when the lights went down, Field says. But the fourth type, which usually responds to upward motion, now responded to a much broader range of motion, including down and sideways, they report today in Neuron.

Field and his colleagues then analyzed why the “up” cells were acting oddly. Using a computer model of all four directional cells’ activity, they concluded that when the “up” cells sacrificed some of their preference for one direction, they improved the performance of the group as a whole, boosting DSGCs’ ability to detect motion in low light.

To find out how the “up cells” had switched their function, scientists genetically engineered mice that lacked intracellular connections called gap junctions in their upsensing neurons. Such protein channels allow chemical signals to pass from one neuron to another and have previously been linked to night vision. Field’s team found that in retinal tissue from mice without the gap junctions, upsensing cells didn’t adapt to the dark. That means that at least some of the “up” cells’ ability to boost motion detection in low light depends on gap junctions, the authors say.

Whether this holds true in people as well is unclear, but the rodent insight might still be applied to artificial vision efforts. Even though DSGCs make up just 4% of ganglion cells in humans, compared with about 20% in mice, many new retinal prosthetics for visually impaired people rely in large part on electrically stimulating ganglion cells. Studies like this could help fine-tune those technologies, Field says. “If you’re going to stimulate ganglion cells, you need to get them to send the right signals to the brain.”