Scientists may be able to improve treatments for neurodegenerative conditions, and even make faster computers, by figuring out how the human brain learns and remembers.
Putting people into brain scanners to discover which brain regions light up when they think about playing tennis or eating chocolate has become a standard research tool and a source of great stories in the media. But the resulting images are the equivalent of a satellite map of Britain. Neuroscientists can discover there's a lot of traffic in, say, Manchester on Tuesdays, but precisely what that has to do with racket sports or food, they can only guess at.
Mapping the brain's exact micro-circuitry in order to get much clearer answers is an undertaking on a completely different scale, not least because the human brain has 100 billion neurons, each one linked to around 10,000 others. And yet such detail is becoming tantalisingly within reach as scientists in a number of major projects are beginning to make progress in understanding, modelling and simulating the brain down to nanometre resolutions.
Microsoft co-founder Paul Allen has put $500m into reverse-engineering the brain at the Allen Institute for Brain Science in Seattle. The Institute started out developing a publicly available online map of where and when different genes are switched on and off in the brains of mice, and more recently has followed this with a similar chart for humans.
In Europe, there is the virtual brain project Blue Brain at EPFL (Ecole Polytechnique F'd'rale de Lausanne) run by Henry Markram. Blue Brain began in 2005 with an agreement between the EPFL and IBM, which supplied the BlueGene supercomputer to build the brain. Now Markram is leading a further proposal called the Human Brain Project (HBP), which has been submitted for funding under the EU's Future Emerging Technologies Flagship program. If the grant is awarded, Europe's best neuroscientists, doctors, physicists, mathematicians, computer engineers and ethicists will get '1bn over 10 years to build on the Blue Brain work and develop a unified understanding of the brain that spans multiple levels of organisation from genes to cognition and behaviour. The grand plan is to eventually store everything we know about the brain in massive databases and detailed computer models.
"What's new is that we have the computing resources, the experimental equipment and the techniques to get the complete circuit diagram," says Winfried Denk, director at the Max Planck Institute for Medical Research in Heidelberg in Germany. "The completeness has huge advantages as it allows you to validate theories. You could also speculate that if we understand the basic circuit mechanisms in the brain, we have a better chance of seeing what might be wrong in cases of disease or illness and fixing the problem."
Mapping the Connectome
Denk is one of the leading scientists involved in mapping the intricate network of connections between neurons, which is the basis for all information processing in the brain. Mapping this dense wiring diagram (also known as the 'connectome') should provide the data needed to distinguish between models of neural computation. Arguably it is the most fundamental way of reverse-engineering the brain, says Denk, whose work gives an insight into the immense challenges involved in human brain research.
Known for pioneering techniques for imaging cells at high resolution, Denk has spent the last decade developing a way of automatically acquiring 3D images of tissue at a resolution of a few nanometres. Prior to this, the only way of getting a detailed 3D image of brain tissue was a time-consuming and error-prone method of finely slicing the sample, imaging each slice separately, and then trying to reconstruct the tissue from multiple, often quite distorted, images. In Denk's serial block face scanning electron microscopy, however, an entire piece of tissue is put into the microscope and the surface scanned, and only then is a thin section cut and the layer below scanned. This new automated method can take data around the clock seven days a week and reduces the amount of distortion caused by slicing, so is less error-prone.
A couple of years ago, Denk and his team demonstrated using this approach to reconstruct the connections within a 300 x 300 x 60μm piece of mouse brain tissue, which they published in Nature. More recently, the researchers have taken a big step towards obtaining the wiring diagram of a complete mouse brain. One major challenge was to treat a large piece of tissue so that it is evenly fixed and stained right through to the inside.
In an initial analysis, scientists followed the axons of 50 randomly selected neurons, marked them by hand and found that the axon paths could be clearly reconstructed using the process. It would take far too long to trace all of the neurons in this way as a mouse brain consists of around 75 million neurons. Therefore, the image evaluation must be automated – something to keep computer scientists busy for a few years.
"Our goal with the mouse brain is to eventually describe it completely at a neuron-to-neuron [cell-to-cell] connection level. It means when you have a signal in a particular cell and that cell fires, you can trace all its inputs and outputs," says Denk. If all goes well, Denk predicts we should have the full mouse connectome within ten years. Image data for a whole mouse brain at 20nm resolution could be stored in a petabyte or so (a small room full of hard drives).
The human brain is 3,000 times bigger than a mouse brain and, as such, a much greater challenge. "We do not have the storage capacity, microscopy equipment, or methods to prepare such a large piece of tissue as the human brain at the moment, " says Denk. "But what might be possible is to do small pieces to confirm the structure of local circuitry. So, for instance, how is a human cortex different from a mouse cortex? Are there different circuit motifs? We can take advantage of these techniques as we progress with the mouse brain."
Making accurate predictions
Blue Brain (and HBP if it is funded) will benefit from such data in their simulations, although Markram's methods are partially predictive, which is faster than waiting for detailed measurements: "We believe that if we can get information about the shapes of neurons and all the geometrical constraints of how neurons are put together inside the brain, then we can predict a lot," explains Richard Walker,'senior science writer, EPFL, in a clip from a documentary currently being made by Noah Hutton. "But it is a model describing biology so we have to check at every stage. Then the data from other people is very valuable. It's only by checking between model data and biological data that we can find things wrong with our model," adds Walker.
But there are some who are not convinced by this approach. In Hutton's film, Sebastian Seung, professor of computational neuroscience at MIT's Department of Brain and Cognitive Sciences, is critical of the Blue Brain project. Seung, who is the author of a book called 'Connnectome: How the Brain's Wiring Makes Us Who We Are', is interested in constructing the 3D structure of brain connections using precisely imaged brain slice data, including some from Denk's group. Seung says that without detailed and reliable maps of neuron wiring connections, brain simulation is bound to fail.
Blue Brain scientists have confidence in their methods. So far they have successfully built and accurately simulated the neocortical column of the somatosensory cortex of the young rat – the part of the brain thought to be responsible for higher functions such as conscious thought. But there is some way to go. A rat's brain has about 100,000 columns in the order of 10,000 neurons each. A human cortex may have as many as two million columns, each having around 100,000 neurons each. Each simulated neuron requires the equivalent of a laptop computer.
Currently the team is taking the first steps towards getting the virtual neocortical column to interact as if it had a body. One interesting experiment that Markram showed at the 2012 FENS neuroscience conference in Barcelona is teaching the virtual neocortical column to balance a virtual ball on a virtual tray by contracting four virtual muscles.
According to Markram, it shows that the brain slice is already embodying physics. "The exciting thing is you can look at any neuron you want, any synaptic connection, any pathway. When you let them fire you can see how the information is moving across the layers and between cells, how many cells are recruited into a task."
Changing our understanding of the brain
Those working on reverse-engineering the brain have already made some interesting discoveries. The Allen Institute's human brain gene map, for example, was derived in the main from the brains of a 24-year-old man and a 39-year-old man. Both showed a surprisingly similar pattern of gene activity. Work is now underway on donated tissues from both sexes to see if there is a common gene expression profile in all healthy human brains.
On this side of the Atlantic, Markram's team has gained insight into the role of nature and nurture in brain development by examining small clusters of around 50 pyramidal neurons in the neocortex of different rats.
It is well known that neuronal circuits become established and get reinforced through experience but it turns out that these small pyramidal neuron clusters share similar characteristics and connect according to simple rules.
The conclusion is that neurons make connections independently of a subject's experience, suggesting we are born pre-programmed with some innate knowledge. This could explain why we all share similar perceptions of physical reality, while our memories reflect our individual experience, according to Markram.
Understanding how we think, remember, learn and feel is a massive challenge. One hesitates to use the word mind-boggling but it does seem strangely apt. By gaining insights into exactly how brains work, the hope is that we can understand what happens when things go wrong, and (perhaps) in the future fix damaged tissue by wiring in replacement circuitry. We might get smarter computers too.
Watch the video of Noah Hutton’s Bluebrain documentary