Dopamine, Reward, Motivation, Lateral Hypothalamus, Feeding Behavior & Eating Disorders | Melissa Sharpe | #164

About the guest: Melissa Sharpe, PhD is a neuroscientist. Her lab at the University of Sydney studies reinforcement and reward learning in rodents.

Episode summary: Nick and Dr. Sharpe discuss: dopamine and its association with reward learning and motivation; reinforcement learning & the brain; the lateral hypothalamus and feeding behavior; and more.Related episodes:

* Drugs, Addiction & Neuroplasticity: Psychedelics, MDMA, Opioids, Cocaine, Amphetamine (Adderall), Nicotine, Marijuana & Alcohol | Robert Malenka | #162

* Neuroscience of Pleasure, Reward, Liking vs. Wanting, Motivation, Food vs. Drug Addiction & Emotion | Kent Berridge | #94

*This content is never meant to serve as medical advice.

* Support M&M if you find value in this content.

* Full audio only version: [Apple Podcasts] [Spotify] [Elsewhere]

* Full video version: [YouTube] [Odysee]

* Episode transcript below.

Full AI-generated transcript below. Beware of typos & mistranslations!

Melissa Sharpe 2:28

Yeah, so I'm an Australian that's recently moved back from the US after about eight and a half years, I grew up in Sydney. So it's really exciting to be home and doing what I love. And generally consider myself a behavioral or cognitive neuroscientist. We work with rats. So that generally means you have to say behavioral instead of cognitive that though I don't necessarily agree with that. And what we're trying to do is to understand how different parts of the brain contribute to our understanding of our environment, and help us learn about different contingencies in our environment. So we can respond appropriately, and get the things we need and avoid the things that we don't.

We like using different techniques like optogenetics, which allows us to manipulate neuronal populations with light with a lot of temporal precision, because we're really interested in the very specific ways in which different neuronal populations contribute to learning. And we also record neural activity using things like fiber photometry, that allows us to record for example, dopamine release for dopamine centers and these sorts of things, again, in a way that relates to learning and specific things that are happening for the rat in their environment.

Nick Jikomes 3:41

What, what is the difference between cognition and behavior? Let me just ask you that. So

Melissa Sharpe 3:49

I guess there's kind of a long standing argument in the field as to whether rats are really capable of cognition. And a lot of people often I guess, human or primate researchers generally would consider rats not to have cognitive abilities, but rather exhibit behaviors that might mimic some of the behaviors that we mimic, but don't necessarily have the same cognitive processes that are underlying that.

Nick Jikomes 4:16

or so. So what? So how exactly would you define cognition?

Melissa Sharpe 4:24

He sure so I guess I would define cognition as the ability to relate new bits of information together to essentially make novel inferences about the state of the world in a way that hasn't necessarily been told to you or a way that you haven't directly experienced in your environment before.

Nick Jikomes 4:45

And so completely, it's completely uncontroversial and obvious that primates like us have different mental and behavioral capabilities than rats and other rodents do. Can you What would be? What would be can you still make the argument that rodents don't have true cognition before we maybe get you to provide counter counter arguments to that.

Melissa Sharpe 5:13

So I guess there's an argument as to whether they really have a prefrontal cortex. And so people would often think that a prefrontal cortex is what allows you to make these kinds of novel inferences and perform sophisticated behaviors that might reach the benchmark of cognition. And so there's an argument that they don't have the same kind of prefrontal cortex that primates

Nick Jikomes 5:36

do I mean, I'm guessing you don't find that compelling.

Melissa Sharpe 5:40

I think that primate prefrontal cortex is definitely different from a rodent prefrontal cortex. But there's a lot of work that, you know, we are the people that I've worked with have done that show prefrontal cortices and the rat do very similar things to what they do in the primate. So for example, you can isolate, you know, the stroop effect or something like that in parts of the prefrontal cortex in the same way you can in humans. And I think that's really great evidence for inference and also for cognition, higher order cognition that's dictated by prefrontal cortices in similar ways, across primates and rodents. So

Nick Jikomes 6:17

what are some of the things that you guys study in the lab in rats, that that might be examples of cognitive behavior?

Melissa Sharpe 6:23

Yeah, sure. So one of the big ones is probably sensory preconditioning. And so that's essentially tasks that's generally used to assess what we call model based inference, which is just making an inference from information that you've learned about that you haven't had to piece together in a novel way before. And what that involves is essentially, for example, teaching a rat, although this has been shown in, in primates and, and also in mice, teaching around, for example, that there's two neutral stimuli that are related. So for example, one term goes on for 10 seconds. And then an auditory clip will come on for 10 seconds. And the animals learned that those two auditory stimuli are related in time. And so that they learned that they're essentially linked, but they don't mean anything to the rat, so the rats not really doing anything about them. But you can reveal what they have learned about them, by them presenting one of those cues, for example, to click with food separately. So on different days, they'll experience a click that comes on for 10 seconds, and then the food is delivered. And they learn that there's a relationship between the click and the food across many days and many trials. And then what you can do is to present back that auditory tone that they first heard that predicted the clip, in a test by itself, and what will do is to exhibit an expectation for food as though they expect food is about to be delivered. And we think that's due to a model based in France, which is that they know the tone and the click are related together, they know that the click leads to food. And so they're making this novel length during that test where you're presenting the tone by itself for the first time that it's also going to lead to food. And so that's what we essentially call sensory preconditioning. And, and model based inference. I

Nick Jikomes 8:11

see. So you're so associated to arbitrary sensory stimuli. Together with a rat. So you present one you present the other, they go together in time, they don't mean anything. So the rat doesn't associate it with a shock or something bad happening or a treat or something good happening. And then later on you, you pair one of those stimuli with a reward. And the rat automatically associates the other one with a reward.

Melissa Sharpe 8:38

Well, so there's a debate as to when it is that they associate the other cue with food. And to some people would argue that it's taking place during the learning about the Polycom food. And that's something called mediated conditioning. So the idea would be, okay, you've got this clicker that comes on, it makes me think of the tone that it was associated with, and therefore the tone and the food are going to become paired together. Because it's a mediated thing, where you're just thinking about the tone, at the time that the food comes, we have evidence that that's not the case. And so what we think is happening is that at the what we call the protests, where you present that tone, for the first time, they're going, Oh, hey, the tone that leads to the click, and that click leads to food. So I think it's likely that this tone is going to lead to food, and so they go to the football

Nick Jikomes 9:26

naturally, but before we dive into some of this in a little more, a little bit more detail, maybe we should go over some more basic things. So you know, a lot of what you do would be considered reinforcement learning, and and associative learning and some of these concepts. What's Um, can you just kind of briefly describe for people what is reinforcement learning and what is Association and how are those things typically measured and eroded?

Melissa Sharpe 9:51

Yeah, so we typically think of reinforcement learning is you know, you're learning about information in your environment because it's paired with something important and That means that it's really important that you learn about it. And there's a whole bunch of rules that people have made and in different models about how you associate those things together and how it weights up different things like how far away a cue might be from the important thing like food, how salient the queue is, all those sorts of things. It's essentially, Reinforcement learning is essentially a lot of maths to describe the way in which we learn about things that are important to us. And an association would be a little bit different in that associative learning theory is really thinking about how we explicitly make associations between representations of things in our environment. So the difference, I guess, is in the representation, in some ways, although it's not always as disparate. And that's thinking more about how you link things together. But there are also mathematical models of how you do that, that can be very similar to reinforcement learning and computational reinforcement learning.

Nick Jikomes 10:58

And so like when we think about rodents, studying rodents, and setting learning and memory in rodents, oftentimes, the first thing people learn about is Pavlov's dogs. So this would be a simple form of association. Can you kind of give people a sense for like the different types of learning that we we tend to see in rodents that we tend to study simple associations versus things that are contingent on behavior, and then maybe up to some more elaborate things that perhaps you or others have worked on?

Melissa Sharpe 11:29

Yeah, so I mean, really just like Pavlov's dog, where the dog is learning that there's a bell that is associated with food delivery in it, it has a conditioned response, whereby they show salivating to the bell that's paired with food, even if you don't bring the food out, you can get wrapped to learn these sorts of associations between like a tone and food. So they'll exhibit an expectation for food when the tone comes on, they'll go and search for food when the time comes on. They also can exhibit things like taste reactivities. So if you can add some kind of flavor, or even an auditory stimulus with the taster, you'll actually see that they exhibit the kind of Oral Facial activities that were associated with the testing itself. So these are all kind of conditioned responses, whereby cues are kind of choosing the environment being flavors or being tones or lights or something like that can can take on some of the properties that the actual food itself has, which has probably been conditioning,

Nick Jikomes 12:30

and so on. Like your description of your research. On your website, it says that you guys are looking at revisiting current theories that are often taken as dogma, but may not be incomplete. Can you give us a sense for what some of those are, we're a sort of the field right now in terms of things that maybe people are taking for granted that we're starting to learn don't work the way we thought they did, given some some of the detail we have now, with new technologies that allow us to probe the brain more deeply.

Melissa Sharpe 13:02

Yeah, so I think obviously, dopamine is a really great example of that. And we've done a lot of work with dopamine. I think there's lots of other areas of the brain where we thought we know what they do. But actually, they're doing something more complicated. But dopamine is probably one of the most obvious ones that were we've really changed the way that we're thinking about how that neurotransmitter and how dopamine is that dopamine neurons. So neurons that are releasing dopamine, are contributing to learning and cognition.

Nick Jikomes 13:29

So I think, you know, some of the stereotypes that are out there, especially in, you know, the wider culture. But even even among scientists, to some extent, I think, you know, dopamine is strongly associated with pleasure with hedonic value with drug addiction and things like that. Over the past, I don't know, 510 years or so, since you've been in this field. How has thinking among neuroscientists who study dopamine and related things? How is it sorted to change? And what are some of the misconceptions about it that are out there?

Melissa Sharpe 14:01

Yeah, so I think it's still hotly debated as to what dopamine does. And I think more and more, we're understanding that dopamine does a lot of different things. One of the ways in which we've contributed to the literature is that we've shown that dopamine can help you learn associations between things rather than just serving as this chemical that might endow food or queues paired with food with a kind of hedonic pleasure or something like that. So one of our papers, for example, showed that we could drive learning in sensory preconditioning, which is essentially learning between two neutral sounds in the environment, by stimulating dopamine optogenetically as as a prediction error, which is how dopamine neurons usually fire during learning. And we showed that you can essentially get animals to learn about the associations between those neutral stimuli in a way that that couldn't be understood by just assigning value or pleasure or something to those cues. So that suggested that dopamine contributes in More complex ways to reinforcement learning, and really is helping us to understand our environment and the complexities within it over and above a role for it and assigning pleasure or some kind of hedonic valence towards different stimuli or or food itself.

Nick Jikomes 15:16

I mean, does it even make sense to you that we should be trying to talk about, on the one hand, there's a challenge here from like an educational perspective. You know, if you're a scientist studying this stuff, at some point, the information has to get out to the wider world. And we have to obviously use words to do that and simplify things. But on the other hand, you know, a lot of misconceptions can pop up. And and we can sort of get things very wrong, if we're not careful with how we speak about about some of these things. With dopamine being equated with things like pleasure and things like that. You know, you often hear that about other neurotransmitters as well, people will talk about serotonin as being for mood or emotion, as norepinephrine as being, you know, for attention or about attention or something. Is it useful to equate neurotransmitters and molecules like this with individual calm concepts as long as we're careful? Or is it just fundamentally, too simplistic? And we should avoid thinking like that?

Melissa Sharpe 16:18

I think that's a really interesting question. And I don't really know what the answer is. I mean, I think in a lot of ways, it was really smart of people that came before to be talking about dopamine as being involved in reward or hedonic pleasure, because there was evidence that, you know, dopamine neurons are just firing to things in the environment that predict something good, which is explicitly food or reward. And so in some ways, that was really useful way to communicate to, you know, the general audience about what it is that dopamine is doing. I don't think that it now captures what we understand dopamine to be doing. And I guess now we have to think about a way to communicate more effectively, in simple terms, what it is that dopamine might be doing to help us learn about our environment and represent the different associations within that. I

Nick Jikomes 17:10

often hear people now talk about dopamine as being involved in motivation in a more general way than just being narrowly focused on pleasure and reward. Would you say that's accurate slash? How would you? You know, given the state of the art today, how would you talk about our understanding of dopamine and what it does?

Melissa Sharpe 17:27

Yeah, I mean, I think dopamine does a lot of different things. I'm just gonna close the window over there for a second. Oh, yeah. Notice?

Nick Jikomes 17:40

One moment, folks.

Melissa Sharpe 17:46

Lots of kids going to school. Yeah, so I think dopamine is definitely involved in motivation. And the magnitude of the dopamine response is going to influence how much for example, a rat responds to a cue that's paired with food. A lot of the work that we've done, we actually have a paper that's coming out soon shows that you can dissociate the magnitude of the dopamine response in somewhere like the nucleus accumbens, from responding itself. So I think dopamine is reflecting the extent of Lent associations or the magnitude to which you've associated, for example, a cue that might be paired with food, and that influences responding. But what we're finding is that you can dissociate that from the actual response itself. But we know that dopamine and particularly in the striatum is really important for how you respond and when you respond. And whether that is the cause of the response or associated with the response itself. I think we still haven't really dissociated. What.

Nick Jikomes 18:54

So when we talk about dopamine, we often talk about, people will say, the reward circuitry, the mesolimbic dopamine reward system, they'll talk about places like the nucleus accumbens that you just mentioned, the ventral tegmental area that the synapses where dopamine is released, that are really important for things like reward learning, really important for things like addiction and other things. But is that the whole story? Or is there are we sort of learning and piecing together a wider and more elaborate circuitry that involves dopamine? And if so, what are some of the maybe key circuits that you've been working on or that we've discovered more recently, that connect to but aren't, you know, just that classic VTA do accumbens connection?

Melissa Sharpe 19:35

That's a great question. And so this is what I'm really excited about in our lab and like generally in the field. So I think now that we're understanding that dopamine is doing something more complicated, we now know that the prediction error signals and other types of signals that are coming from dopamine neurons in the ventral tegmental area, which is kind of the source of these learning signals, right, it's going to be doing something really different in a lot of different circuits because dopamine neurons project all throughout the brain, right. And so one of the things we've been doing, for example, like many in the field is to try to understand how these different circuits are contributing in different ways to cognition. And I kind of thought this was, you know, not a boring question, but I just thought that we'd be recording prediction errors all over the brain in different circuits. And then you kind of figure out how it is that they contribute in different ways to learning, which is probably like, very similar to what we think the brain region that's receiving the signal is contributing to cognition, right. But actually, what we're seeing is something that's really different in our lab, and we're seeing really different kinds of dopamine release and dopamine signals in different circuits that really don't have a lot in common with one another. And so it doesn't just look like these areas of the brain are receiving prediction errors in the way that we think about it. phasic prediction error that comes from VTA dopamine neurons and get sent to places like nucleus accumbens, a signal looks really different. And I think that's really exciting, because it's telling us a lot more about how it is that dopamine is contributing to cognition. And, and really, I think showing that dopamine is is extremely influential in the way in which we're learning about our environment by dictating what is learned downstream is the way that I'm thinking about it.

Nick Jikomes 21:18

I see. So it kind of sounds to me like, so there's this area called the VTA, the ventral tegmental area, in the hindbrain, basically, you know, deep deep at the bottom, sort of primitive part of the brain, and it's famous for, you know, it has the neurons that release dopamine, they release it in the nucleus accumbens, and the so called classic reward circuitry, but they project all over the brain. And they're also famous for doing something called a reward prediction error, which is, I don't know how you would say it, sort of like a kind of subtraction that neurons can do that scales with how surprising an outcome is. And I guess what you were saying, If I heard you correctly, is we know that there's this war prediction error coming from these dopamine neurons in this classic reward circuitry, synapse, and the VTA projects there and a bunch of other places. So it's probably just sending the same type of prediction or everywhere else. But it sounds like we're starting to discover that's not the case IT projects to other places. And there's other types of computations probably happening. Yeah, exactly.

Melissa Sharpe 22:16

It looks really different. And we do know that there's a lot of heterogeneity in dopamine neurons themselves and the VTA in the way that they respond. And I think it's circuit defined which many other people in the field have shown in different ways. And now we're starting to understand the extent of that. So the extent of heterogeneity in the response from VTA dopamine neurons, that might be kind of dictating what is learned downstream in many different regions. But even in the nucleus accumbens, itself, where we think of this as really receiving and being a proxy for the reward prediction error that you were talking about. We've shown now that in the nucleus accumbens core, for example, that's a particular type of learning that's been reflected there, which is not the whole error, right? It's not the whole component of what is learned about cues and rewards. So even that is this much more specific thing than what we've previously thought about. And I think that is really illustrative of what we're going to find downstream in other regions is they're doing something much more specific than what we thought they were in terms of how dopamine is contributing to that sort of learning. So

Nick Jikomes 23:29

So these dopamine neurons that live that were their cell bodies are in the ventral tegmental area in the hind brain, they project to multiple areas. Can you give us a sense for where is it truly all over the brain is several regions can can you give us a picture of that? Yeah, I

Melissa Sharpe 23:44

mean, it's it's a lot of different regions, right. So it's ventral striatum, which is nucleus accumbens, its dorsal striatum, dorsolateral, striatum, and also medial striatum. So basically, the whole stray unknown, which is, we often think of as being important for responding and things like that. It's also areas we've revealed like the lateral hypothalamus. So we've characterized a novel circuit, there were dopamine neurons or projecting to the lateral hypothalamus to facilitate learning and unique ways. Also, recently, we're looking into projections to the ventral hippocampus as well. But dopamine neurons also project to the dorsal hippocampus. And the hippocampus is often thought to be you know, really important for cognition and, and forming associations between things that are related in your environment. So that's a really important projection because it really is showcasing the complexity of what dopamine can do. Dopamine also projects of course, to prefrontal cortex, many different regions in prefrontal cortex and what it's doing there has been known. Well, it's been known for a long time it's doing something influential and I think it'd be really interesting to understand exactly what it is that those terminals are doing.

Nick Jikomes 24:56

Yeah, I want to dig into that a little bit more. But earlier, you mentioned something is called model based learning. Can you contrast that with model free learning? I know this is something that you've worked on. But what are those two things? And how do you study those in rats, for example?

Melissa Sharpe 25:10

Yeah, so model based learning is essentially where you are learning about how things are related in your environment in a way that allows you to make novel inferences when you're faced with something new, right. And the way in which we test that in rats, for example, is in sensory preconditioning, where we teach them that two neutral cues are related. And then we teach them that one of those cues leads to food. And the animal will make the novel based inference model based inference that the other cue that hasn't been paired with food is likely to lead to food. And that would be an example of a model based inference, because they're using information that they've learned in the past to predict something new model, free learning is where we're essentially assigning a cash value to cues or other things in the environment like contexts that have been associated with something of value like food reward. And essentially, that cash value is thought to have back propagated from the food itself. So there's a kind of number that's associated with that food, like a seven out of 10. And that seven out of 10 attaches to the queue. So it's not teaching you about the relationship between the cue and the food, right? It's just making that cue valuable in and of itself, so that you want it. But you don't necessarily know why.

Nick Jikomes 26:29

So, I mean, off the top of my head, if to give people an intuitive sense for this, if we were thinking about something like a food reward, or an addictive drug, something, something naturally rewarding, that you really want a tasty dessert or something. If you're sort of thinking about that in a model free way, I don't even know if I should say thinking there, if you're looking at that, in a model free way. That would be like if someone had like a compulsive addiction or orientation to this, like, Okay, I want that piece of cake right now. And it's just sort of like, boom, I see it, I want it. I know, it's tasty. Whereas if you had a model based way of learning about that, that would be a situation where you're thinking, Alright, there's a piece of chocolate cake there, but I just ate a little while ago, I'm trying to lose weight. Do I really want to eat the tasty thing? Or do I want to hit my weight goal? It's the more deliberative cognitive way of making a decision. Yeah,

Melissa Sharpe 27:25

I think that's a really good way of describing it. And that's often how we describe it. In our papers, as as we kind of attach it to examples in our environment. The idea is that model free kind of agent, as we call them, a model, free purchase person, or rat or whatever, essentially, isn't thinking about their actions, they're just kind of attracted to these things. And they're not really thinking about the consequences of that. And often, it's thought that in addiction, these processes are heightened in some way. So people are not able to kind of avoid drug peds stimuli and stuff like that, because it has this very high model free cash value. Whereas if your model based agent and you're performing more on the basis of model based associations and inference, then you're going to be able to say, Well, no, I really don't want to do that, because it's going to have negative implications on my life. And so I start,

Nick Jikomes 28:16

I mean, this immediately makes me think about things like drug addiction, and how they tie into development. You know, so you could imagine, if you form associations between a stimulus and and a strong natural reward, early in development, before someone's even gotten to the point where they have model based learning, you could imagine that would have very different consequences from introducing a reward later in life. So maybe there's an interaction between the strength of reward and someone's ability to react to it non compulsively based on where and their cognitive development, those associations were formed.

Melissa Sharpe 28:55

Yeah, I mean, I think that's a really interesting way to think about it. I haven't really thought about it like that before. I guess, in part, because I don't necessarily buy into the very strict model free model based argument of addiction, I guess. But I think that would be a really interesting test of that. So we think, you know, that rats develop model based learning and that primates develop model based learning kind of around adolescence, when they develop a prefrontal cortex right properly and that whole circuitry becomes defined. So if they experience some kind of drug que Association, would that mean that they are they're more likely to learn about those types of associations in a model three way? Yeah. Yeah,

Nick Jikomes 29:42

I guess you know, partly what I was thinking there too. And I'm just thinking off the cuff here. You know, when you see something like a compulsion develops a like a drug addiction. Is that because there is this just super strong sensory reward Association? That just sort of is loud Are there any competing Association? Or is it somehow like preventing model based learning from even happening?

Melissa Sharpe 30:09

Yeah, I think that's really interesting. So the, you know, traditional view on this is that we have model free and model based processes that a constantly going on in parallel, right, yeah. And for some things, they're weighted differently. So for example, when you start driving, like, you're gonna be acting in this very model based way, because you need to think about everything that you're doing and the consequences of that. But later on, you become really efficient and really good at this. So you're going to become more model free and develop the kind of habit which is what we call, essentially a well established model free association. And the idea is that in addiction, taking the drug itself is essentially waiting your brain more towards the model, free learning and the model free behavior, which means that these associations with reward and the value that these cues can have, are having more control over your behavior than a more deliberative process that allows you to think about the consequences of your actions. The problem is, it's getting more complicated, you know, there's always these really nice theories that that have a lot of evidence for them. And they're kind of easy to convey, in a way like the kind of action habit, drug addiction argument, which is the same as the model three model based drug addiction argument. But there's now a lot of evidence that a lot of drug directed behaviors are actually the result of a model based agent. So a model based process, and a lot of these cues that are associated with drugs are essentially able to control behavior in a model based way. And that still has this huge effect on how much they're going to seek out the drug and and how much they're able to stop trying to seek out that drug. It's getting more complicated.

Nick Jikomes 31:57

Yeah. And I think, you know, if I just think about this off the cuff, you can imagine, you know, if someone's going for the, the reward, let's just call it a drug reward. There's two ways at least, they could take the reward, one could just be don't even think about it, just boom, boom Association take it just completely like zombie behavior. But the other one could be, they think about it, they don't, I don't know if I should take it because there's all these negative consequences. And yet, the they're still compelled to take it. After deliberation. We've probably all been in those situations, right? Where we're thinking about doing something, and we do think about it. And yet, you know, we go with the reward anyway. And it sounds like you're just sort of saying the second thing is been observed. Yeah,

Melissa Sharpe 32:39

I think that's a really, really great way of describing it. Right? That there are two sort of ways that we can go about this. And it does look more and more as though people for example, that are struggling to stop taking drugs are thinking about it, and they are thinking about the consequences. But something else wins out, right, the power of the cue Drug Association went out. And I did my postdoc at the National Institute on Drug Abuse, where people are really thinking about and have been thinking about this in those times for a long way, for a long time. And one of the examples that they often bring up, or that I often heard, when I was training there is, you know, people that have severe drug addictions, right that for going jobs and family and all these sorts of things to keep taking a drug, they go to really complicated lengths to try to get the drug. Right, there are crazy stories of things that people are doing to try to get hold of these drugs. And that doesn't

Nick Jikomes 33:35

mean meaning there's a level of planning and a lot of cognition going into it. It's not just simplistic. Exactly. The behavior.

Melissa Sharpe 33:43

Yeah, exactly. Exactly. And when you think about it, you know, in the way that addiction looks like in the real world, it doesn't sound like people are just like picking something up and taking it right, there are all these things and all these plants that are going into it to be able to put you have that drug, and it's not easy in our environment. Yeah. Which I thought was a really interesting way to think about it. And our data are consistent with that.

Nick Jikomes 34:09

So stepping back for just a second. I so I had a discussion with someone about a different subject the other day, we were talking about, we were talking about animal models, and how well like animal models of depression, actually model human depression and the translatability there. And the opinion of that person was basically like, well, you know, we've had so many animal models of depression as a role model for depression for so long, that you know, we should be on like ninth generation SSRIs by now, if they're, they were modeling depression and humans very well. So his opinion was, you know, wrote a model to depression, you know, are not modeling human depression as well as we would like. I'll put it conservatively like that. And obviously, in the literature, people are careful to say things like anxiety like behavior, depressed, depressive, like behavior, and in the addiction and reward, you know, there's an analogous thing I could ask about addiction reward and how well rodent models capture human addiction. But just at a basic level, how, how well do some of these rodent models actually capture the phenomenon that we're interested in from a sort of human translational perspective?

Melissa Sharpe 35:14

Yeah, I mean, I think that's a really complicated question with a complicated answer. I'm not sure I'm the best person in the field to answer it, I guess the way that I would think about it, is that there are some things the models are able to capture and some things they're not. So for example, this change in goal directed habitual behavior, which is like model based and model free behavior seems to come out in humans and rodents that have been exposed to drugs. So there are definitely changes in cognition that look really similar in a rodent model of addiction to humans that have addictions to different substances, right. And I think that's probably our best evidence for translation, in terms of how an animal responds for drug, actually, rats often don't really like drugs. And there's usually just a subset of rats, of course, that will go on to exhibit what looks like addictive behavior. And people have captured this in a number of different ways. I think it's like, you know, 15% of rats or something. And people say that actually, that's quite like human.

Nick Jikomes 36:27

Yeah, yeah, I've talked to Christian Lucia, for example, on the podcast, and in his head, this is a very reproducible result. It's just going to be specific. It's it's rats and cocaine I'm talking about here. But I assume that this generalizes to some extent, basically, you know, roughly 30% of rats will become addicted to cocaine if you give it to them over and over again. And they're not too enriched environment. And that apparently is like the number for humans as well, it's actually remarkably similar, despite the fact that you know, rats and mice are inbred populations, and cetera, et cetera. And so when, with that in mind, so when you say, rats don't really like drugs, do you mean that only a minority will proceed to compulsive drug seeking? Or do you mean that, like, they often don't find drugs rewarding that we do?

Melissa Sharpe 37:15

I think it's a little bit of both, you definitely have to work harder to get rats to take drugs right, then. And I don't know whether that's because, you know, obviously, we can't do this with humans, and maybe humans, you know, 85% of humans will show the same kind of thing. But a lot of the time, people are doing things in the lab with rats to get them to press a lever because they won't press as they will suffered more something like that. And yes, definitely, yes. Sorry.

Nick Jikomes 37:44

I'm just gonna say like, so. I think there's a bit of a misconception out there. You know, people probably think it's probably pretty easy to if you just get cocaine in there, they'll just start pressing it forever and ever automatic. Yeah, that's not the case. No.

Melissa Sharpe 37:57

Um, so yeah, it's quite hard to get them to do it. There's a lot of kinds of ways that you can get rats to behave, right. And the people in the addiction field often use those to get them to start taking the drug, right? Because it's quite hard. But then, you know, as you're saying, Christian said, and lots of the people in the field have found that there are about 15%, or something of rats that really like it and will very happily take it and will forego food for the drug and will also, you know, undergo punishment. So get shocked sometimes for pressing to take the drug and things like that. So, maybe that is actually a good model for what humans are also like, but, yeah.

Nick Jikomes 38:41

So, VTA, it's where these dopamine neurons live, that we've talked about a little bit, they project famously to the nucleus accumbens. That's where we see this reward prediction error signal that's been fairly well studied. That's really important for reward learning. And it's part of the classic sort of reward circuitry, the muse, a limbic dopamine system. But as you mentioned, these VTA dopamine neurons project project to all sorts of places. One of those places is the lateral hypothalamus. And that's a place that you've worked on. So before we get to some of the specifics there in terms of the dopamine circuit, can you give everyone just a brief, concise crash course on the hypothalamus and the lateral hypothalamus? How do we normally see these things talked about in textbooks say,

Melissa Sharpe 39:25

Yeah, so the lateral hypothalamus is my favorite topic to talk about. Lots of people like talking about dopamine. And people come to the lab and they want to do stuff with dopamine. But the lateral hypothalamus is like really where my jam is at right now. Essentially, we thought that the lateral hypothalamus is the feeding center of the brain. So the idea is if I was to turn on your lateral hypothalamus right now you'd like to start eating whatever is in front of you, and maybe even start snoring on your microphone or something like that. It's like both to just drive this innate motivation to consume food and approach food like subs senses. But during my postdoc, I got really interested in how a role for lateral hypothalamus in approaching food might actually reflect a really basic role. So this area in learning about the predictors of food. So, you know, for us feeding is all about learning, right? We're not born with an innate understanding of what food is and how it reduces our hunger in the same way, we actually have to learn that water is rewarding, and it reduces. So we're not actually born with that. It's all about learning. And so we ran a series of experiments that tried to understand how it is that the lateral hypothalamus is important for learning about the predictors of food, for example, even the sight of or smell, or taste of food as being predictive of the reduction in your hunger and the intake of calories. And that sort of put us on a trajectory of a bunch of different experiments that revealed some really surprising things about what this area is doing. So we've shown now that the lateral hypothalamus is really critical for learning about choosing the environment that are really close to food. So it really wants to learn about anything that is close to food, and the association between those things. But actually, it really wants to stop you learning about anything that's not close to food. So what we found is that it seems to be opposing learning about, for example, neutral cues. And we do that using things like sensory preconditioning, it seems to try to stop you learning about the association between those neutral things. And the model that we've now developed is that the lateral hypothalamus is really important for kind of getting you to learn about everything that's proximal to what you need, which includes food, and to ignore everything else that isn't close to food. Right. And so we think that this is an area of the brain that's essentially trying to bias you to learn about primary reinforcement to survive

Nick Jikomes 42:03

it. And this is probably why I'm gonna ask you to betray my background a little bit. This is probably why classically, you know, decades ago, the lateral hypothalamus was always associated with things like the orienting response, you know, which way what is the animal paying attention to? How's it pointing at sensory organs? It sounds like behaviorally what's been known classically about this area of fits with what you just described? Yeah,

Melissa Sharpe 42:25

I'm actually not aware of those studies, I would love to see studies looking at lateral hypothalamus and orienting to different cues that are paired with food and stuff that doesn't surprise me that fits with that. I think that's great. Yeah.

Nick Jikomes 42:36

And so before we dig into this a little bit more, is it fair to say so so you said, you know, we're not we don't, we're not born actually knowing that you know, water is going to quench your thirst or that specific things are going to quench our hunger? Is it fair to say the animals are born with hunger and with thirst, they're born of the affective states that upset us, you know, when we're deprived of, of those things. But we have to actually learn through association or through other means, what are the stimuli in the environment that will actually get rid of those feelings that we're born with?

Melissa Sharpe 43:08

Exactly. So that's why we have all these innate things in our body that make us do stuff when we're born, like go towards our mom's breast, right? That's so that we learn that association. So you don't have an innately what you have is this response to do that with your mother when you're born. And that's how you learn about those associations to start with. Yeah,

Nick Jikomes 43:29

and then so the hypothalamus generally, so it's hypothalamus below the thalamus, it's sort of at the top of the brainstem classically, is like, classically, historically, people would normally talk about the hypothalamus as being a quote unquote, primitive or ancient part of the brain. It's a neuro endocrine organ, and it's doing. It's involved in relatively simple things. It's involved in automatic or innate behaviors, it's not involved or thought to be involved in thinking stuff or cognitive stuff. Is that an accurate way of describing how people have historically thought about it? And is, you know, it sounds like in fact, what you're discovering is, is it is doing some more sophisticated things? Yeah,

Melissa Sharpe 44:09

exactly. That's exactly the case. So that's what we always thought hypothalamus was doing. And this kind of circles back to this idea about prefrontal cortex and that being important for cognition, and how rats can't have cognition in the same way because their prefrontal cortex is a bit different. And prefrontal cortex is doing all the thinking, right, it's this real top down approach to the way that we learn about other our environment and the way that we behave. But actually, even what the subcortical regions are doing like VTA, like the lateral hypothalamus, it's really complicated. And it has cognition. Yeah,

Nick Jikomes 44:44

yeah. It's, I mean, it's amazing to me that I can understand historically why so many people started to think in terms of like, okay, lower brain regions. Yeah, so called, they're not doing cognitive cognitive thinking stuff. Neocore Text is, but that just seems to have completely persisted over time, even though we know that many subcortical regions are doing quote unquote, cognitive things that's quite clear. And we know that there's many animals that don't have homologous structures at all. And they are doing those things as well. Yeah,

Melissa Sharpe 45:14

yeah, no, I agree with that.

Nick Jikomes 45:18

So can you describe some of these experiments that you've done in a little bit more detail with respect to this dopamine, lateral hypothalamus circuit? What does the circuit look like? Are there are there projection, dopamine dopaminergic predictions going to the lateral hypothalamus? Is it bi directional? What's sort of the structure of this circuit?

Melissa Sharpe 45:35

Yeah, so it's a bidirectional circuit. So we've known for a long time that lateral hypothalamus provides input to dopamine neurons in the VTA. And we've done some experiments to show that that's really critical for the learning phenomenon that we find with lateral hypothalamus. Now, we've revealed that VTA dopamine neurons also project to the lateral hypothalamus. And they have terminals there and they're regulating learning in the lateral hypothalamus as well. So the way that we think about it is this bidirectional learning circuit, where VTA dopamine neurons are sending their teaching signals to the lateral hypothalamus so that the lateral hypothalamus biases learning towards cues that are proximal to food, and then relays that expectation back to VTA dopamine neurons so that VTA dopamine neurons can do their subtraction of the prediction error that essentially allows it to calibrate How surprising something is, and whether it's going to send out a teaching signal to keep getting you to learn or whether learning is done.

Nick Jikomes 46:37

I see and so on. Is this prediction doing something like the classic reward prediction error computation? Have you recorded that? Yeah,

Melissa Sharpe 46:46

so Well, we've recorded dopamine release in the lateral hypothalamus across learning. We've also optogenetically inhibited VTA dopamine terminals in the lateral hypothalamus during learning. And we find that the optogenetic experiments reveal that lateral hypothalamus is learning associations between cues and rewards. And this is facilitated by a dopamine era. We don't know exactly what that dopamine era looks like yet. But what we've been doing is to record dopamine release during cue reward learning in the lateral hypothalamus. And what we see is a really unique signature of dopamine release, that doesn't look like what we expected it to.