To appreciate the beauty and specificity of Nirenberg's approach, it is important to realize that there is not just a single homogeneous group of retinal ganglion cells leaving the eye. Rather about 20 distinct types of cells exist, each one specialized for a different task. Some ganglion cells respond only to the onset of light but not when it ceases (“on” cells), whereas a second set signals the reverse—they respond with spikes when light is turned off (“off” cells) but are silent when they see a bright region. If a microelectrode array simultaneously stimulated both “on” and “off” cells—as would happen with an all-electronic strategy—it would confuse the visual brain because it would appear that light had just been turned both on and off simultaneously!
Other populations of ganglion cells carry information relating to a specific wavelength (involved in color vision), whereas still others convey information about things moving downward or sideways, and so on. In a sense, all of us have 20 different views of the world, emphasizing varying aspects of the visual environment. How these fractionated and disparate views are unified to yield the coherent picture of the world that we perceive consciously remains deeply puzzling.
Fortuitously, it looks as if each of these cell types has its own distinct molecular bar code. This knowledge can be used to restrict the expression of the optogenetic molecules to just those cells and then to target the artificial stimulation appropriately. That is, if we knew the retinal code of “on” cells—the way they convert visual information into electrical pulses—as well as their molecular signature, these cells (or any other group) could be selectively targeted.
Nirenberg and Pandarinath accomplished this targeted approach in blind mice by making them carry a mutated version of a gene needed for photoreceptors whose ganglion cells also express ChR2. An encoder takes an image captured by a digital camera and converts it into a train of spikes appropriate to a particular group of ganglion neurons, for instance, “on” cells. It does this conversion from images into the retinal code by training and comparing its response with those actually recorded from “on” retinal ganglion cells. Thus, as a simple example, if a bright light had just moved into the field of view, the encoder should generate a burst of pulses. These signals are turned into pulses of blue light that drive the “on” retinal ganglion cells to fire a similar sequence of pulses. To the neurons in the brain proper that are the recipients of these “on” retinal ganglion cells, these pulses convey the datum that something luminous has just made its appearance. Exploiting the same code as used by a healthy retina should help these blind mice see.
How well this device reconstructs pictures is shown in the images at the lower left. If the baby picture at the far left is sent through the device, the brain could, in principle, reconstruct the image at the near left. Far from perfect, but clearly the image of a toddler.
In a field test, actual mice outfitted with this retinal prosthetic could reliably detect motion to the left or to the right.
The true measure of performance, injecting a blind person's eye with viruses that express ChR2 in retinal ganglion cells and giving the patient a set of glasses that carries the encoder and light stimulator, is within reach. The fantastic marriage of molecular biology, optics and electronics that is optogenetics will soon bear fruit and help people regain their eyesight. Stay tuned.
This article was originally published with the title Cracking the Retinal Code.