Neural Roots of Curiosity Explored

Posted on April 25th, 2016

Credit: Allan Zepeda

Simons Society Junior Fellow Jennifer Bussell is designing experiments to study the neural roots of curiosity.

Jennifer Bussell is curious about curiosity. A basic desire to learn about the environment confers an evolutionary advantage on many species, but we humans also seek out information for its own sake. What is it that drives us to know something, just for the satisfaction of knowing it? “That’s a fundamental question that we know very little about,” says Bussell, a postdoctoral research scientist at Columbia University and a Simons Society Junior Fellow. Bussell is just beginning experiments in mice to investigate the neural underpinnings of our desire to find things out. Recently, I spoke with her about her work. An edited version of the interview follows.

Anyone who has had dogs or cats knows that they will investigate a new toy or other object in the room, as a way to make sense of the environment. But how do you scientifically test whether an animal is curious?

In most animals, the drive to get information, the drive to explore, to know or seek information, evolved because usually it’s useful to do that in the environment. But to study this drive — call it curiosity — we have to isolate it from other rewards and artificially make the information gained useless.

My collaborator, Columbia professor Ethan Bromberg-Martin, has come up with a wonderful way to measure monkeys’ desire to gain information independent of other rewards. In one experiment, thirsty monkeys can get a drink of water by moving their eyes to choose a symbolic target on the left or right of a computer screen. The monkey has a 50-50 chance of getting a large water reward whatever its choice. On the right, it sees one of two symbols: One symbol is associated with getting a lot of water, and the other is associated with getting less. On the left, the monkey sees one of two other symbols, neither of which means anything — the symbols on the left are not correlated with the amount of water. So the monkey can choose whether to have information in advance, but its choice has no effect on its water reward.

Once the monkey learns that it can look to the right and know ahead of time whether it will get the larger water amount, it almost always chooses to know. What’s even more amazing is that the same reward-encoding neurons that fire when the monkey gets water also fire when it sees the symbol that gives it information. We’re trying to design a similar experiment in mice.

How can you test information seeking in mice?

Smell is at the center of a mouse’s world. The researchers in the lab I’m in [run by Nobel laureate Richard Axel] know a lot about how the identity of a smell is represented in the brain. In the experiments I am setting up, we will test thirsty mice in a box with holes, or ports, where they can receive water. We will offer the mice information in the form of odors rather than visual cues, and they can indicate their choices by entering different ports. We expect that curious mice will choose the ports with informative odors, and we are interested in how those odors are represented in the brain.

How do you determine whether neurons involved in recognizing a smell are also involved in or correlated with curiosity of that smell and whether it holds information?

We can identify which neurons are activated by an odor using microscope images of a particular fluorescent protein inserted into neurons. Because we can see which neurons fire in response to a smell, we can ask whether different cells are activated in the mouse’s cerebral cortex when the smell signals the possibility of information versus when it does not.

We can also silence or activate the particular brain pathways we think might be involved in driving this curiosity and ask if those pathways play a causal role in the choice of information.

How did your own curiosity lead you to neuroscience?

As an undergraduate, I worked in a lab where we wanted to understand, genetically, what makes humans unique. We looked for brain-specific genetic differences between humans and other primates, so I’ve always been fascinated by the question of how a physical object embodies a mind and consciousness and a drive to understand the world.

In graduate school, I set out to study molecular biology, but then I learned about all of the discoveries being made in neuroscience. It seemed like such an exciting field, and I wanted to be a part of that. That was the first time that I took formal neuroscience classes, and I switched my training rotations to neuroscience. I remember watching fruit flies’ courtship under a microscope for the first time and thinking about how we knew and could, in a way, control the 2,000 neurons that make the insects do that complicated behavior. That was really amazing.

Before going to graduate school, you worked as a management consultant in the biotechnology industry. How has that experience shaped your career?

My graduate school adviser likes to say that, unlike most scientists who spend their entire lives in the captivity of academia, I’ve been out in the wild. Working in the ‘real world’ was really helpful in terms of learning how to work in a team. But mostly it made me appreciate how lucky I am to be an academic scientist. I have so much gratitude toward the taxpayers and other funders for allowing me to think deeply about the brain and how it works, and I really want to be able to do something meaningful with the opportunity.

What questions about the brain do you hope to see answered during your career?

So much about the brain is still mysterious, but one of the big questions is how information is transformed within a neural circuit. We know there are electrical and molecular signals, but what’s the code? Knowing that would be the first step toward understanding the brain in the same way that we understand an electrical circuit or a computer, where we actually know how information processing is accomplished.

Personally, I would also love to know more about the extent to which seeking information is the motivation for animals to do things. Curiosity is starting to seem like an important motivation for learning. If we can understand more about how it works, maybe we can encourage it.

How Do Different Brain Regions Interact to Enhance Function?

Posted on March 3rd, 2016

The neuron-level recordings of a stimulated larval zebrafish brain. Courtesy of Misha Ahrens.

The neuron-level recordings of a stimulated larval zebrafish brain. Courtesy of Misha Ahrens.

Textbooks have long divided the brain into dozens of discrete areas, each with its own set of functions. Students learn that the hippocampus is crucial for learning and memory. The cerebellum coordinates movements. The prefrontal cortex is the seat of higher cognition, handling processes such as decision-making.

This tidy picture might work well for introductory classes, but in reality, even the simplest thought or action involves many brain regions, each containing thousands or millions of neurons. Surprisingly little is known about how brain areas interact and cooperate to interpret sensations and generate behaviors. To tackle this puzzle, the Simons Collaboration on the Global Brain (SCGB) held its first Multiregional Models of Population Coding workshop on January 19, bringing together leading experimental and theoretical neuroscientists from around the world to discuss how brain regions work together. Increasing investment in neurotechnology — including that by the U.S. BRAIN initiative — have finally given scientists the tools they need to observe large swaths of the brain in living animals in unprecedented detail.

Having the tools is one thing. The challenge now is to apply these tools in ways that will lead to a better understanding of the brain. “We are still in the dark about basic facts concerning multiregional interactions and function,” said workshop co-organizer Larry Abbott, an SCGB investigator and co-director of Columbia University’s Center for Theoretical Neuroscience. “How do multiple brain regions divide up the components of a task? What assures that the proper information is conveyed from one region to another?” Fortunately, he said, “we are finally getting the appropriate data to begin to figure this out.”

Misha Ahrens, an SCGB investigator and a group leader at the Howard Hughes Medical Institute’s Janelia Research Campus, presented a spectacular example of what the new technology can accomplish. On a monitor screen, thousands of tiny specks of colored light twinkled in bursts and waves, circumscribed by the faint outline of the body of a larval zebra fish. Over the past several years, Ahrens and his colleagues have pioneered a technique that monitors the activity of almost all the neurons in the larval zebra fish brain, and this display showed what tens of thousands of neurons were doing.

Ahrens and his collaborators genetically modified the animal so that the florescence of molecules in the neurons changes whenever calcium ions enter them, which is (a process that serves as a proxy for electrochemical activity. Because the animal is naturally transparent, the researchers were able to apply sophisticated microscopy techniques to see the neurons. Doing so enabled Ahrens to observe interactions among brain regions in unusual detail. With this setup, he can ask more in-depth questions. “Simple analyses are dominated by large populations of neurons doing similar things,” he said. His approach can capture “more complex, nonlinear interactions among neurons across the brain, or interactions depending on just a few neurons.” In new work he presented at the workshop, he showed that a small number of neurons that release serotonin react only when sensory information coincides with motor information. Ahrens surmised that these neurons could detect when the fish’s own movements alter its sensory input — an important skill to distinguish self-generated sensory stimulation from environmental input.

Other presenters at the meeting described work based on data similarly recorded from many neurons at once, spanning organisms from the tiny worm C. elegans, with its 302 neurons, to the human brain, with its nearly 100 billion. David Van Essen, a professor of neuroscience at Washington University in St. Louis, Mo., presented some of the early discoveries from the Human Connectome Project, a five-year, $30-million effort by the National Institutes of Health to map large-scale connections in the human brain. A consortium of over 100 researchers at 10 institutions are employing cutting-edge brain imaging techniques — including functional magnetic resonance imaging — to observe activity and trace the pathways of fiber bundles. In his presentation, Van Essen described the analysis of over 1,000 brains. Researchers used a combination of information from many imaging modalities to generate a much-improved map of the cerebral cortex, which will enable neuroscientists to better navigate the human brain.

The kind of data collected by Ahrens and Van Essen, said Xiao-Jing Wang, a workshop co-organizer and SCGB investigator who co-directs New York University’s Sloan-Swartz Center for Theoretical Visual Neuroscience, will allow neuroscience to move into new territory. “Neuroscientists have been mostly focused on local circuits,” he said, “but really important questions depend on interactions between brain areas.” Wang has personally invested in this approach, as his lab recently shifted focus from local cortical circuits in the prefrontal cortex to large-scale circuits across cortical brain regions. “A really important, fundamental issue,” he said, “is how information flows from area to area.”

Cora Ames, a postdoctoral researcher at Columbia University and an SCGB fellow studying with SCGB investigators Mark Churchland and Larry Abbott, is addressing that very issue using a combined experimental and theoretical approach. Her question is simple: What happens to the motor cortices on the two sides of our brains when we move our hands independently, as opposed to when we move them in sync? “Sometimes we want our hands coordinated, but not all the time,” Ames said. “It’s the same problem as rubbing your tummy and patting your head at the same time.” Because each arm is controlled by the motor cortex in the opposite hemisphere of the brain, these two brain regions must communicate differently depending on how coordinated the movements need to be. “Yet no one really knows much about how that information flows or is gated,” Ames said.

Ames is currently training monkeys to turn pedals with their hands, either in or out of sync. She then records the electrical activity of several neurons at once in the motor cortex on each side of the brain. She has also built a computer model of those regions to simulate the coordinated brain activity. The model generates a sinusoidal output that is analogous to signals in the monkey’s muscles when it turns the pedals. She can vary the strength and complexity of the connections between the two regions in the model and determine how strongly they coordinate the output. The modeling will ultimately enable her to predict what she will observe in the monkeys’ actual brains during the task.

Neuroscience, according to Larry Abbott, is long overdue for research questions that get scientists to think outside of their comfort zone. “When asked what we study,” he said, “most of us answer by identifying a region of the brain. We’ve compartmentalized the brain, but that’s only a crude approximation of the truth.” He hopes this workshop will prompt researchers to “go back to their labs, and think about the data a little differently, build models a little differently.” It could also result in more direct cross-pollination. “I would be surprised if new collaborations did not come out of this meeting.”

Such collaborations will be necessary for researchers to take the next big steps in brain research. Most neuroscientists agree that conceptual breakthroughs have not kept pace with the explosion of technical advances in the past decade. That’s why Cori Bargmann, a neuroscientist at Rockefeller University and co-chair of the committee whose report led to the creation of the BRAIN Initiative, was quick to point out that conferences such as this one are a crucial way of complementing worldwide efforts to develop neurotechnology. “There is a huge need for science to use the technologies being created by the BRAIN Initiative intelligently,” she said. The work presented at this conference, she believes, fits the bill. “This is not about technology, it’s about science.”

Theoretical Brain Studies Highlighted in Special Issue of Nature

Posted on February 24th, 2016

nature neuroscienceOn February 23, Nature Neuroscience published a special issue focusing on neural computation and theory and highlighting recent advances. The issue consists of a collection of reviews and perspectives written primarily by investigators from the Simons Collaboration on the Global Brain (SCGB). In their commentary, SCGB investigators Anne Churchland of Cold Spring Harbor Laboratory and Larry Abbott of Columbia University argue that the need for theoretical approaches in neuroscience research has never been greater. They note that this need has been driven by the “explosion of technologies” used for measuring and manipulating neurons and by developments within theoretical neuroscience itself. They review some of the challenges of data analysis and highlight the diversity of approaches to modeling brain function. Global understanding of the brain, they write, will be based on loosely stitching together highly diverse approaches.

In the related reviews and perspectives, several SCGB investigators provide more details about the advances and challenges of computation- and theory-driven approaches to modeling brain function. Abbott and his co-authors review methods for developing more realistic spiking network models. James DiCarlo of the Massachusetts Institute of Technology and his co-author describe advances in using hierarchical convolutional neural networks and outline how these networks can provide more insight into sensory cortical processing. Alexandre Pouget of the University of Geneva and his co-authors discuss the probabilistic computations of uncertainty and strategies for studying the neural codes underlying confidence and certainty, explaining how these investigations are essential to understanding the roles of cortical areas in decision-making. Brent Doiron of the University of Pittsburgh and his co-authors examine recent theoretical results related to the analysis of simultaneous recordings from large neural populations, findings that they say are important for statistical analyses of high-dimensional neural data. Ila Fiete of the University of Texas at Austin and her co-author review the computational principles of memory. Two other papers, written by investigators outside of SCGB, round out the issue by describing experimental and theoretical studies of cortical inhibition and computational psychiatry.

The special issue is available here.