ARTICLE IN BRIEF
More than 120,000 volunteers who agreed to play an interactive computer simulation helped a team of researchers to develop a detailed 3-dimensional map of a small portion of the connectome of the mouse retina.
Figure. An online vi...Image Tools
Sometimes, the methodology of a study is as much the story as its results. Or so commentators seem to think is the case for a May 15 paper in Nature that used “crowd-sourcing,” so to speak, to offer new clues on how the retina detects motion.
Among the academic authors listed on the paper are the “EyeWirers” — volunteers from around the world who logged into a website to play a game developed by lead author Sebastian Seung, PhD, formerly of the Massachusetts Institute of Technology but now a professor of computer science at Princeton University's Neuroscience Institute.
The game involved using different colors to fill in cross-sections of neurons from a mouse retina made from high-resolution scanning electron microscopic images by co-author Winfried Denk, PhD, of the Max Planck Institute of Medical Research in Heidelberg, Germany. When all the neurons were filled in by the EyeWire volunteers and combined, they produce a detailed 3-dimensional map of a small portion of the connectome of the mouse retina.
Figure. A 3-D recons...Image Tools
With the information provided by the 3-D view, the authors were able to suggest a solution to a question that has vexed researchers for 50 years — how does the mouse retina detect motion?
The answer they propose involves using their 3-D depiction to trace connections among various subtypes of four broad classes of neurons in the retina – photoreceptors, bipolar cells, starburst amacrine cells, and ganglion cells.
In a series of papers beginning in 1964, British visual neuroscientist Horace B. Barlow, PhD, of the University of Cambridge reported that ganglion cells in the retina detect motion, but he offered no explanation of how.
Using the EyeWire data, Dr. Seung and his colleagues constructed a wiring diagram that traced two pathways through the retina — one through the Type 2 bipolar cell, and the other through the Type 3a bipolar cell. The diagram they created also included the starburst amacrine cell, which has been difficult to reconstruct from electron microscopy images because it is so thin.
Each branch of a starburst amacrine cell displays “direction selectivity” — it responds to motion in a specific direction, and produces direction-specific inhibition of the ganglion cells. But how are Type 2 and Type 3a bipolar cells wired to starburst amacrine cells?
Dr. Seung and his colleagues propose that Type 2 bipolar cells connect with starburst amacrine dendrites close to the cell body, while Type 3a bipolar cells connect farther away from the cell body. In addition, Type 2 bipolar cells respond more slowly than Type 3a bipolar cells, producing a time delay — about 50 milliseconds — that allows signals from the two cell types to arrive simultaneously at the starburst amacrine cell during motion that moves away from the cell body, thereby generating a signal that reveals the direction of motion detected in the visual field.
The map of the retinal connectome was vital to developing this hypothesis, according to Dr. Seung. Nearly 120,000 players have been drawn to the map since EyeWire created it in December 2012. The players so far have completed 2.3 million tiny cubes of retinal tissue — about 2 percent of the retina.
“There's no computer that can do this in a fully automated way,” said Dr. Seung, “People have to do it, so you either incentivize them by paying them, or you incentivize them by inviting them to have fun.”
The explanation of direction selectivity in the mouse retina proposed by the authors amounts to a provocative hypothesis that now must be explored, according to Dr. Seung, author of the 2012 book, Connectome: How the Brain's Wiring Makes Us Who We Are.
“We can't regard this as the truth,” he said. “Now the ball is in the physiologists' court. There are a lot of great experimental tests that physiologists can do.”
Marla B. Feller, PhD, a physiologist at the University of California, Berkeley, agrees that the challenge now will be to demonstrate that these cells are actually connected.
“To me the novelty of this paper doesn't come from the insights into direction selectivity,” said Dr. Feller, professor and head of the Division of Neurobiology. “It comes from the beautiful reconstructions and the fact that they used crowd sourcing as a way to do the reconstructions. I think it's an interesting hypothesis for how the wiring up of these bipolar cells might influence the computation of direction selectivity, but now someone has to do the experiment to see if it's correct.”
Dr. Feller was skeptical of the EyeWire input until she logged on to the site and played the game.
“I questioned this idea of crowd sourcing when I first heard about it,” she said. “Then I tried it myself, and it's fantastic.”
In addition to providing a useful tool for researchers, the computer animations created by assembling the work of EyeWire volunteers also reveal the staggering complexity of the retina and other brain tissue, according to Dr. Feller.
“They make you realize how hard the problem is,” she said. “The tissue is very dense, and there are neurons everywhere. They're extremely useful in thinking about how things are wired up. When we are looking to figure out a potential circuit mechanism we can use that source of data.”
The 3-D map produced with the help of EyeWire data still lacks crucial information, according to Rowland Taylor, PhD, of the Casey Eye Institute at Oregon Health & Science University in Portland, who also studies retinal circuits. For example, the electron microscope images colored in by EyeWire volunteers show no intracellular information — just the outline of the cell membranes. As a result, it's impossible to identify synapses. “So they're not actually counting synapses or connections between the neurons, which is a significant limitation,” Dr. Taylor said.
Also, neurons in the retina are divided into “on” and “off” pathways. “One responds to dark, and one responds to bright things,” said Dr. Taylor. “All the previous analysis has been of the ‘on’ pathway, and this Nature paper focused on the ‘off’ pathway, so it's unclear how applicable one is to the other. I'm not sure this paper really nailed it, but they've advanced a very interesting hypothesis that should spur further research into this fundamental neural computation.”
Understanding how retinal ganglion cells communicate with other cells is the key to working out how the retina encodes visual information, according to David M. Berson, PhD, Sidney A. Fox and Dorothea Doctors Fox professor of ophthalmology and visual science, and professor of medical science at Brown University.
He praised Dr. Seung and his colleagues for developing EyeWire, “which is both a highly creative and effective innovation, and a source of very valuable data,” Dr. Berson said. “What's wonderful about these new methods is that they open up a whole new universe of structural detail.”
Like Dr. Taylor, Dr. Berson considers the lack of intracellular information in the connectome map to be a limitation. Revealing the organelles and other intracellular material of neurons would make image recognition very difficult, but sacrificing that detail to reveal where one cell stops and other begins, as Dr. Seung and his colleagues did, does not reveal the presence of synapses.
“They infer from the amount of surface area whether there's likely to be a synapse, but that's an indirect method,” Dr. Berson said. “The next generation of these methods will allow for more complete staining, and I think we're going to see a revolution as these methods are refined because they allow you to identify cell elements very efficiently without giving up information about synaptic contacts. There's a lot of excitement about this type of connectomic analysis of the nervous system.”
WHAT THE EXPERTS THINK ABOUT EYEWIRE
LINK UP FOR MORE INFORMATION:
Figure. DR. MARLA B....Image Tools
Figure. DR. ROWLAND ...Image Tools
Figure. DR. DAVID M....Image Tools