Learning by Association

Randy Bruno, Ph.D.
Associate Progressor of Neuroscience
Principal Investigator, Mortimer B. Zuckerman Mind Brain Behavior Institute
Columbia University
Dana Grantee: 2010-2013

At any given moment, your brain is bombarded with an extraordinary amount of sensory information about the world around you. One of neuroscience’s biggest questions is how our brains manage to sort through all those sensory signals to flag what’s most important to you in the moment–and what can safely be ignored.

Randy Bruno, an associate professor of neuroscience at Columbia University, has spent the bulk of his career trying to understand the physiology of neural circuits and the role that these microscopic parts of the brain play in influencing cognition. But with a new foray into behavioral studies in mice, he is shedding new light on how somatosensory neurons, generally thought to be responsible only for our sense of touch, can learn to associate different signals as the result of reward-based learning. Here, Bruno discusses why such associations are important to understanding cognition in both healthy and diseased brains, why apical dendrites may be the seat of such learning, and how his Dana Foundation funding allowed him and his colleagues the room to explore an area they might not have been able to otherwise.

Your work views the brain through a cellular lens. What big-picture questions are your lab trying to answer?

Our current work is driven by questions like: How does your brain learn to associate, or mentally link, different things in the world? How do you know what objects in your environment are most informative and will give you the information you need to decide about something you may or may not what to do? What part of the brain stores those lessons?

We’ve known for a long time that rewards–when something positive happens to us–can potently shape our later behaviors. We are trying to understand how that works. We know that our brains are really good at reward learning, which has been studied quite a bit in the context of the basal ganglia, ancient sub-cortical brain structures responsible for helping us select and modify actions. That’s a good, sensible place to look. “Maybe there are things I can observe in my world that will push me in a certain direction because I might get a reward.” But what we’ve revealed in a recent study is that these rewards can also rewire the evolutionarily newer parts of our brain, like cerebral cortex.

You might imagine that if this reward-based learning was going to reward any part of cerebral cortex, it would be areas that are involved with decision making or planning. But we found that even the areas involved in initial sensory processing, taking in those very basic sensory signals about the world, are being modified and are learning new associations in response to a reward. And that tactile signal that came into somatosensory cortex may even become associated with information that isn’t tactile (or touch-based) in nature.

How so?

Think about any experience introspectively. For example, your body is constantly touching things, right? But we can ignore a lot of it, whether it’s the shoes on your feet or your shirt against your body. You are still being touched. There are physical sensory signals coming in, but you are not always picking up on them because they may not be behaviorally relevant at that particular time. And it’s not just touch. As soon as you think more about it, you realize that, whether you are looking out a window or having a conversation in a crowded room, you pay attention to the things you are interested in and ignore the rest in the general background.

Your recent study in Cell Reports found that reward-based learning rewires cells in somatosensory cortex. Did that surprise you?

Many of our findings were surprises. We were interested in examining how touch and movement were put together in sensory cortex. When you experience touch, you are often moving your hand–many of our tactile experiences occur when we’re moving our fingertips or we’re stroking objects–and we were thinking that somatosensory cortex must combine them. It has to put together the fact that I’m moving my hand in a certain way, in a certain direction, or perhaps gripping an object, with the actual physical signals that are coming from my skin or, in a mouse’s case, its whiskers, to create a model that can help me understand what it is that I’m touching. So, we were interested in that integration of how cortex can jointly analyze movement and touch.

Originally, we weren’t interested in reward or reinforcements at all. We trained a mouse to do this very simple behavior where it actively moves its whiskers to touch an object that is in front of it. The object may or may not actually be there–we varied it during different trials. The mouse is simply scanning for an object, trying to detect an object, and if the object is there, then the animal presses a lever to get a water droplet. And we thought, well, we have the animal moving, and we have the contacts against the object, so we can see how the cortex integrates those things. The first major surprise was that, as we imaged the neurons, not only did we see somatosensory cells lighting up when the animal touches the object, we saw them lighting up a few seconds later when the animal made the choice to pull the lever and got its reward.

At first, it made no sense to us. It was a fascinating feature of the data. We knew we had to dive deeper so we could understand it.

What are the implications of this later activation for our understanding of cognition in healthy individuals? What about for those who may have neurodevelopmental disorders?

If we take an average healthy person, it means the sensory cortex has learned an association between a tactile experience and some other experience. With the mice, it’s the tactile experience and some aspect of getting that water reward. Maybe it’s the pleasure of a drop of water on the tongue, or the sound of the water dropping, or sating thirst. There are a lot of different associations that could be learned. And that’s something we’re working on right now, trying to understand all the different things you could associate.

Can you associate tactile signals with visual signals? What about tactile signals with auditory signals? So many things we do in ordinary life are multimodal sensory experiences – multiple senses are engaged. And, for the healthy individual, you aren’t learning those associations in just the higher order parts of your brain. You’re also learning these things partially in the lower order parts of your brain. That’s an interesting implication.

In terms of disease states, a lot of brain diseases have some sort of sensory component. Whether you’re thinking about schizophrenia or autism, there is some disruption in sensory signaling. One of the things we revealed in our work is a particular cellular location where these changes may be happening. I think the interesting implication for understanding pathological disorders like neurodevelopmental disorders is there may be an important site where learning isn’t happening in the disease or maybe there’s too much plasticity and connections are being changed too much. Either way, you can disrupt this learning of associations.

What are apical dendrites? And what role do they play in this type of learning?

The other cool thing about the Cell Reports paper is that it shows strong evidence that apical dendrites are heavily involved in this associative learning. Apical dendrites are like a big tree sitting on top of the neurons that have connections that go up to the very surface of the brain. And at the surface, you have a lot of long-range connections between different parts of cerebral cortex. These long-range connections allow the frontal part of your cortex to talk to visual cortex and somatosensory cortex. Past studies have shown activity in apical dendrites before, but it was always the same kind of activity that made the neuron fire in the first place.

Our work shows that the apical dendrites are the part of the cell where this associative learning is going on. What’s also important is that this is probably the first study that shows something outside of what a neuron is usually interested in can activate the apical.

The neurons that we study live in the somatosensory cortex. Usually, they fire to a whisker touch. But we’re getting them to fire with things that are not whisker stimuli. It shows that you can get this extrasensory association through learning and reinforcement–and that the apical dendrites are driving that plasticity. So apical dendrites seem to be an important place to start investigating how this associative learning works.

You’ve said that your Dana Foundation grant helped give you the room to try something new and make these new discoveries. How so?

Originally, we were hoping to study how you selectively process different signals. It was part of our original grant: how you process different signals if they are important to you–the ones you really want to attend to versus the unimportant ones that are irrelevant to doing a task. The whole reason this project took so long is because it was the first project my lab ever did where we were looking at behaving animals. Prior to that, everything in my lab was about physiology and looking at neural activity in animals that weren’t doing anything, trying to map out the anatomy and physiology of the networks. We were very focused on the synapses on the cells and on the organization of small circuits.

But, for this, we needed to look at a behavior. So, the fabulous thing was that the Dana Foundation grant gave us the ability to start moving into behavioral studies. Our original goal was to look at behaviors that involve this attentional component where you had to differentiate between important and unimportant sensory stimuli for a particular task. That’s ambitious, even for a monkey lab that’s been doing this kind of stuff for 20 years.

We started thinking, well, maybe instead of doing attention, we should think about sensation and motion. Maybe we should look at reward. The Dana Foundation funding gave us the ability to explore and really think about how to do these behavioral studies. And then we discovered this whole reward story, which is what the project ultimately turned into.

What comes next? How do you plan to follow this study?

We have several follow-up projects in mind. First, we’re trying to understand to what degree a reward or punishment changes the part of the brain that encodes the number one thing it’s interested in. For example, if I start giving you different kind of objects in your hand, and some are more rewarding than others, do you start to represent those objects differently? How did the neurons change their representations of objects that are more important or more rewarding to you? It’s in some ways like what we’ve just done but we still don’t know what caused those big reward responses in the study. Was it the coolness of the water? The taste? We don’t know what the animal was actually learning. We know it’s some kind of association between the whisker stimuli and the water droplets. But not exactly what it is about those water droplets. So, we’re currently working on experiments with controlled stimuli–controlled, visual, and auditory-based–to see what the associations may be.

Think about a mother mouse who nurtures the pup. As she takes care of the pup, it likely feels a certain way to her. She might associate that feeling with the sound the pup makes. And taken together, she builds a stable representation of who the little pup is. So, there are a lot of multimodal questions we want to answer and can have the lab set up so we can more precisely probe and understand them. How do you get plasticity between different senses? Can that plasticity change the way tactile cortex works? Or the visual cortex? Which connections change if you have a visual-tactile or an auditory-tactile association? Those are the critical questions we are working on right now.