Mapping brain activity is neuroscience’s lofty new goal
Now that we have the human genome pinned down, understanding the brain should be biology’s next great, bold challenge. So say a group of leading scientists who propose to track the activity of the entire brain, neuron by neuron, millisecond by millisecond.
The Brain Activity Map (BAM) project, first floated in June last year, was also hinted at by US president Barack Obama in his State of the Union Address in February. The fact that it has not yet been formally announced or funded has not stopped researchers expanding their proposal. Writing in Science, they now predict that in 15 years’ time it should be possible for non-invasive technology to observe 1 million neurons in real time. That’s enough coverage to analyse the function of several regions of a mouse’s cerebral cortex. The ultimate goal will be to extend this to the human brain.
Along the way, the hope is that the project will transform the technology of neuroscience – in the same way that the Human Genome Project (HGP) helped take genome-sequencing from pipe dream to everyday reality – and ultimately revolutionise our understanding of brain function. It is the right thing to do at this time, says Paul Alivisatos, director of Lawrence Berkeley National Laboratory in California, the lead author of the proposals.
But just how do you go about mapping a brain? This is a question that two projects with similar lofty goals are already grappling with. The Human Brain Project, which won a billion-euro research prize earlier this year, aims to do it by creating a computer simulation of the entire brain. The Human Connectome Project is using magnetic resonance imaging to track the fibres that connect different regions of the brain on the millimetre scale, giving a rough-grained roadmap of the brain. In contrast, BAM aims to generate the traffic report by getting down to the neuronal level, mapping which neurons fire at which time and how they are synchronised.
To do so, researchers will need to find non-invasive ways to record the firing of individual neurons, because all current methods involve opening the skull and, often, sticking electrodes into brain tissue. “Right now, you’re literally driving posts into the brain. It’s not very sophisticated,” says neurobiologist John Ngai of the University of California, Berkeley.
A few groups have already started working on new approaches. For example, the MindScope project at the Allen Institute in Seattle aims to map the mouse visual cortex. The team identifies where neurons are firing by injecting the brain with dyes or using genetically engineered proteins that bind to calcium molecules. When a neuron fires, calcium flows into the cell and activates the dye or protein.
While powerful and widely used, calcium imaging alone is too slow to generate the kind of real-time map that the BAM project requires, saysMichael Roukes of the California Institute of Technology in Pasadena. A faster alternative would be to record the electrical activity of neurons, but the wires required to do this are invasive and tend to be relatively large. To get around the size issue, Roukes’s lab is creating tiny silicon-based nanowires that are connected to an array of electrodes, recording from multiple neurons at once. This allows the researchers to triangulate the position of any given neuron. Their tiny size means that they are less disruptive than other wires would be but you would still need to undergo an invasive procedure to implant them. Roukes’s team has tested the technology in insects and are now moving on to rats. Eventually, he says, they should be able to locate and record the activity from a million neurons at once.
Decoding the brain
But such an activity map is meaningless if it only shows connections and firing patterns without giving any clue why a circuit fires, says Karl Deisseroth of Stanford University in California. One way to image these cause-and-effect relationships is through optogenetics, which involves genetically engineering mice so that their neurons fire when hit with a beam of light shone through the skull. The firing neurons leave a protein trail, allowing researchers to see which circuits responded to the light or other stimuli.
Other promising technologies come from beyond the realm of biology. Alivisatos’s nanotechnology lab is engineering quantum dots that could be embedded in the cell membranes of neurons. When a neuron grows a new connection, it would stretch the quantum dot particle, causing it to emit light. Similar particles could respond in the same way to changes in the membrane’s voltage. In the lab, the dots are extremely fast and their light does not fade over time, but more work needs to be done to see whether having these dots implanted disrupts the function of the neurons.
The problem plaguing all of these light-based techniques is the brain’s density. It is no good having a technology that tells you that a neuron has fired by giving off a flash of light, if you cannot detect that light. The best microscopes currently available can detect light from 3 to 4 millimetres into the brain, enough to see light signals coming from the cortex of a small animal, but not enough to see deep-seated structures such as the hippocampus. “For this, we will need to redesign the basic concept of the microscope,” says Rafael Yuste of Columbia University in New York.
Despite the many technologies on the horizon, the BAM team is not worried about betting on the wrong one. “The thing right now is to get several ideas tried,” says Alivisatos. Once promising contenders emerge, then the simultaneous mapping of millions of neurons can begin in earnest.
The next issue will then be how to deal with the terabytes of data generated every day. Researchers already have their hands full sorting out the behaviour of just a few hundred neurons at a time – the current state of play. Expanding this to millions will demand the development of better computational and statistical techniques, says Konrad Körding of Northwestern University in Evanston, Illinois.
Moreover, to make sense of these activity records neuroscientists will also need to confront the fact that every brain is different, and changes over time. “If I have your brain and my brain at the level of individual neurons, it would be very difficult to line these brains up and compare them. That is a big challenge,” says Olaf Sporns of Indiana University in Bloomington.
Fortunately, that challenge is likely to diminish over time, as neuroscientists begin to recognise general patterns that emerge as they collect more data. General patterns that represent memories of faces or motor decisions, for example.
Once these patterns begin to emerge, the research possibilities are endless. The HGP bore an entire new field of science in the form of genomics. At this stage it is impossible to predict what “connectomics” will unveil.
Yaser Abu-Mostafa of Caltech expects it will eventually lead to advances in artificial intelligence systems that mimic the brain. “I don’t want to say there will be an artificial brain on your desk in three years, but it will happen,” he says. “This project is the real catalyst.”
Another obvious application is medical: comparing the differences in activity between neurotypical brains and those with conditions such as schizophrenia, clinical depression or autism. Clay Reid of the Allen Institute hopes BAM will develop a technology that can be used to screen for brain differences that may indicate these conditions early. The map could also help researchers understand how they arise and manifest themselves, leading to better treatment.
Bone to pick
Not all neuroscientists are banging the drum for BAM, however. Partha Mitra at Cold Spring Harbor Laboratory in New York says that the technologies currently being discussed are still too far in the realm of imagination and still too invasive to start to think about applying them to humans. You cannot open a human brain to test an invasive technology, he says. “Everyone should be reminded that we have skulls.”
Others worry that despite the project’s far-reaching goals and methods, its approach is too narrow. “The best research combines and looks at multiple levels of detail,” rather than just focusing on the connections between neurons and fibres, says Susan Bookheimer of the University of California, Los Angeles. She says that BAM’s map, while useful, may still not explain phenomena like consciousness and cognitive function, which probably emerge at a broader scale.
If BAM does get the go-ahead – and this is a big if, given the US government’s imminent spending cuts – it will remain to be seen whether the technology will advance as quickly as it did during the HGP. But its proponents are necessarily optimistic. One lesson that we learned from the HGP’s achievements, says Yuste, is “that the predictions were too conservative”.
Journal reference: Science, DOI: 10.1126/science.1236939; Neuron, DOI: 10.1016/j.neuron.2012.06.006