Neuroscientists today know a lot about how individual neurons operate but remarkably little about how large numbers of them work together to produce thoughts, feelings and behavior. What is needed is a wiring diagram for the brain—known as a connectome—to identify the circuits that underlie brain functions. The challenge is dizzying: There are around 100 billion neurons in the human brain, which can each make thousands of connections, or synapses, making potentially hundreds of trillions of connections.
So far, researchers have typically used microscopes to visualize neural connections, but this is laborious and expensive work. Now in a paper published March 28 in Nature, an innovative brain-mapping technique developed at Cold Spring Harbor Laboratory (CSHL) has been used to trace the connections emanating from hundreds of neurons in the main visual area of the mouse cortex, the brain’s outer layer. The technique, which exploits the advancing speed and plummeting cost of genetic sequencing, is more efficient than current methods, allowing the team to produce a more detailed picture than previously possible at unprecedented speed. Once the technology matures it could be used to provide clues to the nature of neuro-developmental disorders such as autism that are thought to involve differences in brain wiring.
The team, led by Anthony Zador at CSHL and neuroscientist Thomas Mrsic-Flogel of the University of Basel in Switzerland, verified their method by comparing it with a previous gold-standard means of identifying connections among nerve cells—a technique called fluorescent single neuron tracing. This involves introducing into cells genes that produce proteins that fluoresce with a greenish glow, so they and their axons (neurons’ output wires) can be visualized with light microscopy. The team traced connections from 31 neurons using the more traditional method, moving from the mice’s primary visual cortex up to seven other cortical regions. But whereas this work took three years to complete, the team was able to map connections from 591 neurons in just three weeks using the new technique.
The method, called MAPseq (Multiplexed Analysis of Projections by Sequencing), works by tagging cells with genetic “bar codes.” The researchers inject viruses containing random RNA sequences into mice brains. Once inside a cell each virus expresses a unique 30-letter, or nucleotide, RNA sequence (a bar code consisting of the letters G, A, T, C), along with a protein that cells naturally transport along axons. This protein is engineered to bind to the RNA bar code so that the tag, too, is dragged along the axon. The researchers later sacrificed the mice, dissected their brains and separated them into target regions, which they then sequenced, enabling them to see which tagged neurons connected to which regions. The more connections a neuron has to a target region, the more of that neuron’s bar code will be present in that region’s sequencing data.
The team found, contrary to much previous thinking, neurons in the mouse primary visual cortex (which is the first cortical region to receive and process signals from the eyes before passing information to other areas) typically send outputs to multiple other visual regions “A key finding was that individual neurons in the cortex connect to many brain areas, and this connectivity is not random,” says neuroscientist Botond Roska of the Institute of Molecular and Clinical Ophthalmology in Switzerland, who was not involved in the work. He adds this is this first description of how long-range connections are organized in the cortex.
The team obtained the same results using both methods, validating the new technique. But because many more cells were mapped using the new method, it also revealed nearly three quarters of the tagged cells fell into six groups, based on how many, and which, regions they connect to. This implies there are subtypes of neurons in the mouse’s main visual area that presumably perform distinct functions. “Because we have so many neurons we can do statistics and start understanding the patterns we see,” says Justus Kebschull, a collaborator with Zador at CSHL and co-lead author of the study. “That’s why MAPseq is so useful,” he adds. The researchers speculate these connections to specific subsets of regions may coordinate those regions’ activity and link visual information across the brain to help form complex visual, or even multisensory, perceptions.
The idea for MAPseq was inspired by a fluorescent-tracing method that uses multiple colors, nicknamed “brainbow.” This involves introducing genes that produce different-colored fluorescent proteins (typically red, green and blue) into cells. A mechanism that randomly disrupts some genes then ensures each cell produces a different, random mixture of colors, effectively “tagging” them with distinct hues. As many as 200 colors can be produced, but that amount is still far less than the number of neurons in a mouse brain, or region. MAPseq overcomes this limitation because 30 nucleotides can generate 1018 different sequences. The technique’s speed and ability to accommodate so many combinations makes it a good match for the complexity of connectome mapping. “I predict that as this technology matures it will be a key way we analyze brain connectivity,” Rosko says. “It is single-cell resolution, simple, high-throughput, cheap and there’s no limitation on the number of cells that can be bar coded since there are many more possible bar codes than neurons in the brain.”
Up to now models of neural computation have been little more than guesswork based on behavioral data and the properties of neurons, but detailed anatomical knowledge could constrain such theories. “Hopefully progress in systems neuroscience is going to be faster, thanks to knowing underlying circuit structures,” Kebschull says. “If you have the anatomy, you can quickly weed out hypotheses that can’t possibly be true.” Comparing findings across species will also be useful. “We can start to try to understand what is conserved, and then specializations in different species,” he says. “That gives you deeper insight into brain function, rather than just stamp collecting.”
Currently, the method only identifies which individual cell a connection originates from, and the overall brain region where they end. But future developments may allow barcoding of both the start and finish of a connection. Other layers of information such as cell-type and -shape, can also be added. “All the pieces are there to do that now,” Kebschull says.
Perhaps most important, the technology may help in understanding neuro-developmental disorders like autism and schizophrenia. Some theories assume brain wiring goes awry during development in these conditions, especially in autism. Whereas MAPseq cannot be used in humans (who do not like to be genetically tinkered with), numerous animal models could be studied. Many such models of autism reproduce symptoms by mimicking genetic mutations linked to the human condition. One possibility is that the huge diversity of autism-linked mutations may all point to common wiring defects. “If we could find that, we might have a handle to understand the condition and maybe do something about it genetically,” Kebschull says. “We’re starting to work on that in the lab now.”