November 1, 2014 (Vol. 34, No. 19)

Agnes Shanley

For the past few decades, scientists have made great progress decoding the human genome. An even greater challenge will be understanding the wiring of the human brain and how it stores and uses information.

Connectomics, the term coined nine years ago by Olaf Sporns, Ph.D., a brain science professor at the University of Indiana, to describe this field of research, examines the connections made by the 90 billion neurons and 160 trillion synapses in the human brain. The goal is to discover how humans learn, how they remember, and what leads to neurological and mental illness.

As Princeton University’s Sebastian Seung, Ph.D., has noted, the connectome is really “where nature meets nurture.” Knowledge gained through this research could help prevent some diseases, speed recovery from stroke or other disorders, and optimize conditions to improve learning.

Ambitious mapping projects are underway, with large-scale programs such as the Human Connectome Project, which involves a consortia of universities (Harvard, Washington University in St. Louis, and the University of Minnesota). It received $40 million in funding from the NIH in 2009. Other initiatives include the European Union’s $1-billion Human Brain Project, and the NIH’s BRAIN initiative, launched in June, as well as smaller scale efforts such as the $55-million brain atlas being developed by the Allen Institute.

The overarching reason for connectomics to exist is our need to understand what is stored in our brains, says Jeffrey Lichtman, Ph.D., professor at Harvard University’s Center for Brain Science and Department of Molecular and Cellular Biology.

“Most [of our] behavioral repertoire is stably maintained information that is stored in the physical structure of the brain,” he says. “Our goal is to see what the wiring diagram looks like, and how the human brain changes as a human being goes from [being a] baby to an adult full of information.”


Curiously Complex Babies’ Brains

Dr. Lichtman began working on simpler wiring diagrams as a graduate student, focusing on salivary secretion and the parasympathetic nervous system “What we saw was the opposite of what you might expect,” he points out. “Babies’ brains are actually much more complex than adults, so it seems that we start out with a wiring diagram that’s capable of being almost anything. Then, environment and experience come into play and we choose to be a particular thing.” Connections are pruned away as the individual adapts to his or her environment.

Connectomics promises to help explain situations where the brain’s wiring has gone awry and the wrong things get removed, as for example is the case with learning disabilities, or when particular behavior patterns take root such as obsessive-compulsive disorder or Tourette’s syndrome.

Dr. Sporns has been applying a multidisciplinary approach to neuroscience for the past 25 years, using abstract computational tools and collaborating with engineers, physicists, and biologists. He believes that network science will be important in understanding dynamic brain activity.

“We are applying relatively simple modelling approaches to connectome networks to predict how they become active and how nodes interact,” he explains.

One project he is working on is a model for Drosophila, which has about 120,000 neurons. Another of his projects, involving colleagues at the University of San Diego and the University of Taiwan, is mapping the cerebral cortex of the rat brain.


So far, this work has underscored the importance of modules (sets of neuronal regions that talk more to each other than to the rest of the brain) and hubs  (traffic centers of neurons that can control network nodes). “We find these in Drosophila, in the rodent, and in [a] primate, [the] macaque,” Dr. Sporns says.

Most connectomics research is using functional and perfusion magnetic resonance imaging (MRI) and electron microscopy. Research has benefitted from improvements in temporal and spatial resolution and innovation in both hardware and software. “A resolution increase of 2.5 to 3, or 1 mm, may not sound like much, but it is huge,” emphasizes Dr. Sporns. “Incremental advances have been important.”

Paul Laurienti, Ph.D., director of the laboratory for complex brain networks and professor of radiology at Wake Forest Baptist Medical Center, notes that magneto-encephalography (MEG) has potential. “Its temporal resolution is good, but it is challenging to work with the data. You record all these magnetic fields on the scalp and try to figure out where they came from in the brain.” Many researchers are working with beamformers, signal processing techniques that are at a exploratory stage, he says.


Human brain network from imaging data shows the functional connectivity of the resting human brain. Each node in the network represents a small piece of brain tissue. The brain areas are connected if they have highly synchronized brain activity. The more connections a node has, the larger it is shown in the network image. The colors reveal functional neighborhoods in the brain that are all highly interconnected to one another. [Dr. Paul Laurienti/Wake Forest Baptist Medical Center]

Range of Approaches

Each research group is taking a somewhat different approach to connectome studies. Many are not focusing on animal models.

“We are working with humans and don’t use animal studies, except for some work we’ve done with primates,” says Dr. Laurienti. “The biggest challenge for animal models is getting really good data. Since primate and rodent brains are small, imaging them is much more difficult. Another challenge is that most of the software used for analyzing [image] data was developed for human data, and relies on templates of the human brain.”

At Harvard, Dr. Lichtman’s group has used mouse models to study autism and is developing wiring diagrams for acquiring and analyzing data, using fluoresence and electron microscopy. The researchers use genetically engineered Brainbow mice and viruses that make the mice neurons glow. As a result, each neuron can be visualized in its own color.

In the central nervous system, the density of connections is high; in the cerebral cortex, so many cells make it difficult to use electron microscopy. Researchers use an automatic tape-collecting lathe ultramicrotome (ATCLU). Within this device, a diamond knife cuts plastic-embedded, stained tissue slides to a thickness of 30 nm, 1/1000th the thickness of a human hair.

Thin slicing allows a cross-section of all the wires, 10,000 sections in a series, to appear in a film strip on tape substrate. This way, each frame shows a successive depth of brain tissue. The tissue is put on flat silicon wafers until a library of the image is built.

Connectome research is generating a mind-boggling amount of data, researchers agree. In Dr. Lichtman’s electron microscope wiring diagrams, each cubic mm generates 2,000 terabtyes of information. With available imaging systems, the choice, he says, is forest or trees: You either get high resolution or the ability to process large volumes of data, but not both.


New Mathematical Models Needed

Connectomics researchers agree on the need for new mathematical models. “The biggest advances will be found using dynamic networks to understand cognition,” predicts Danielle Bassett, Ph.D., assistant professor of innovation at the University of Pennsylvania’s department of bioengineering. Her team is applying concepts from network and communications theory.

Like Dr. Seung, Dr. Bassett is a physicist by training. Starting out in nursing, she soon gravitated to physics and math, getting a B.S. in physics from Penn State and a Ph.D. in physics from Cambridge University, with postgraduate work in neuroscience at the University of California at Santa Barbara. Dr. Bassett won a Macarthur grant for her work this year.

“My interest in the field began in graduate school in 2004, when I was interested in understanding the difference between the connectomes of healthy individuals compared to those with conditions such as schizophrenia,” she says. “In my postdoctoral work, around 2009, I saw that we were simplifying too much. The brain has communication patterns that change rapidly throughout the day, and I wanted to understand the dynamics of these connections.”

She decided to focus on learning, hypothesizing that brain communication patterns would change if behavior changed. Ultimately, she said, her work suggested that individuals with flexible networks can learn much faster than those with rigid networks.

“Brain reconfiguration patterns matter at an individual level, and predict what an individual can learn,” she adds, with potential implications for education, rehabilitation, and relearning lost skills.

Dr. Laurienti, who has made a special study of processing in the aging brain, discovered the importance of network theory and complex systems almost by accident, after a colleague heard a presentation on applying chaos theory and complex systems to cardiology. In 2007, they built a functional brain network model from functional MRI data, which focused on how brain areas interacted dynamically.

“Since 2009, the field of network science has become incredibly popular in brain imaging journals, and roughly one third the papers in the field are on this topic,” according to Dr. Laurienti. “However, many people aren’t applying network science from a complex systems perspective Traditional thinking focuses on certain regions of the brain and brain cells. Reductionist approaches model every synapse and cell, but they fail to take into account the dynamics of the system. It’s the interactions between parts that matter. A more dynamic approach is needed to study the entire system.”


Healthy Brain Aging

Dr. Laurienti’s team hopes to identify features of successful aging. So far, it has shown that exercise alters brain connectivity with a specific increase in hippocampal connections. The team is also studying eating and obesity, attempting to understand and recognize a certain fingerprint of connectivity associated with successful weight loss.

“The mathematical tools we need haven’t been developed yet,” he says. “One of our biggest challenges is comparing groups of networks without focusing on single brain regions.” His team is collaborating with biostatisticians who work in such areas as exponential random graph multivariate models, which would give a whole-system perspective.

Dr. Bassett’s team, meanwhile, is working with such concepts as multilayer network constraints, which are mathematicaly represented as adjancency tensors, and tools for network central theory, such as the controllability Gramian. Dr. Bassett’s group has been working on a new toolbox that will be openly available to researchers. This toolbox will be for dynamic networks. The first such toolbox, for static networks, was developed by Dr. Sporns.

There are questions as to how much progress can be reasonably expected in a few years, considering the complexity of the connectome project.

“We’re dealing in the world of information and big data, and it has been a tough adjustment for scientists of my generation,” says Dr. Lichtman. “We were not impeded by information. We focused on big ideas. Now, big ideas and big data are on a seesaw. As big data becomes available, it becomes harder to have a general idea or to turn that into something that humans can understand, given the massive amounts of data involved.”

Dr. Seung has estimated that at the rate things are going now, it could take 100,000 years to finish tracing all the neuron paths in the human connectome, an eternity that could shrink to 1,000 years if artificial intelligence were used. He has started “citizen scientist” initiatives, crowdsourcing work in projects such as Eyewire (which traces the retinal connectome) and Wired Differently (which uses a computer gaming-like interface to trace the wires of the brain).

Despite the challenges, connectomics is clearly getting more attention and more funding. The next few years promise to shed more light on the most complex and least understood organ in the human body.


























This site uses Akismet to reduce spam. Learn how your comment data is processed.