An interview with Dr Andreas Lüthi, Friedrich Miescher Institute for Biomedical Research.

Nearly everything we do in daily life is driven by a goal. The brain sustains a motivational state that guides our actions to achieve desired outcomes — whether it’s preparing lunch or choosing which show to watch on Netflix.

Our motivation is shaped by a combination of internal and external factors. Internally, physiological needs like hunger, thirst, survival, and reproduction play a central role. Externally, environmental stimuli and past experiences influence our emotional and motivational states. Together, these factors determine the range of behaviours we exhibit in any given situation.

Dr Andreas Lüthi, from the Friedrich Miescher Institute for Biomedical Research, is investigating the brain circuits involved in processing emotional, metabolic, and social cues, and how these shape behaviour.

In a recent SWC Seminar, he shared his research on the mouse amygdala, where he has identified populations of neurons that drive experience-dependent behavioural changes. We spoke to Andreas about his latest work.

What first got you interested in neuroscience and in particular the cellular basis of learning and memory?

During my undergraduate studies, I became more and more interested in learning and memory. It was a very attractive question - what is the biological basis that allows us to learn and remember something? 

That was at the time when people started studying synaptic plasticity, and how synapses change in an activity or experience-dependent manner. People were studying this in the context of the hippocampus, but it was difficult to relate changes in synaptic transmission to learning at the behavioural level. However, in the amygdala, people had described very simple associative conditioning paradigms, where one stimulus predicts another stimulus, like a particular sound predicts a mild electric shock. 

That's what brought me to study the amygdala and the cellular mechanisms there which represent learning. 

Can you tell us about your work in the basolateral amygdala (BLA), and what you have found about motivation?

In mice, we have identified large groups or ‘ensembles’ of neurons that are active when animals are engaged or disengaged in different behaviours. Two ensembles inversely correlate with each other for any given activity.

These are not motor neurons, we aren’t seeing a population of cells directing or responding to a certain movement. These ensembles are representations of what the animal expects. 

For example, after an animal learns a task to receive a food reward, two ensembles become active during subsequent repeats of the task. They switch between one another, and together, their patterns of activity represent the whole task-action-reward sequence. This representation includes the identity of the reward – like milk vs sugar water, the value of the reward in terms of amount, as well as the outcomes the animal experiences. 

We’ve shown that representations are dynamically updated according to both the reward context, as well as the state of the animal. So, for the ensembles that relate to food, they are only active when the animal is hungry. If it's not hungry, the system doesn’t care. 

This shows us that there is a state of the amygdala that represents internal physiological states. We’ve shown this not just for hunger, but for metabolic, thirst, and social states too. Many brain areas can make predictions - if I do that, this will happen. But importantly, what we see in the BLA is that prediction is embedded in this representation of the animal’s state. 

We think the ensembles could be an important interface that tell the animal what to focus on. If you’re hungry you should focus on predictions that relate to food. And it will allow animals then to adapt and learn according to their particular needs. 

Most likely the amygdala doesn't perform that in isolation, as it's connected to other brain areas, including the prefrontal cortex. This state information is bidirectional.

Could you expand on the nature of these different states – how are these represented in the BLA and how do you test that?

The ensembles for each of these different behaviours – be it feeding or social interaction – come from the same pool of neurons. I think these neurons are most likely multitasking, but there might be some populations that specialise more in one thing or the other.

We do have some evidence for prioritisation, which might have something to do with an animal’s state. We see neurons that are more active during social interactions, even though before they cared about something else, like spatial exploration. It seems social stimulus overrides some of the other functions. I think it is also important to consider body-wide states like sickness – this could be a state that modifies behaviour. 

We can study these neurons at single-cell resolution using calcium imaging – but we are facing a problem when it comes to manipulation. Everyone in the field has the same issue – it's very difficult to get enough specificity to manipulate these cells because they're all the same type of neurons. There doesn’t seem to be any difference, metabolically or genetically, between a neuron that represents hunger vs. one that represents social interaction. We would like to test hypotheses by manipulating them, but it's not straightforward.

Are there any other challenges with this type of research?

There are multiple levels of challenge. Probably one of the most significant challenges is the data analysis. There are huge amounts of data. I'm not only talking about the neuronal activity but also behavioural data. 

In the past few years, we have had more and more unbiased behavioural tracking, and we study freely moving animals in naturalistic environments. This means they learn faster, and they have a rich behavioural repertoire. But, it makes it harder to study because it is more difficult to create controls and harder to segment the data. 

I do expect new analysis approaches will emerge in the next few years though. We have a lot of work to do, going back and forth between behavioural data and neuronal data to look for correlates.

Another difficulty is the definition of cell types. It’s currently very fuzzy, at least genetically. People are coming up with more precise and more precise definitions, but how that relates to function and anatomy is not always clear – I’m not sure how that is going to be addressed. 

The amygdala is a very old structure, evolutionarily, and present in all mammals.

How similar do you think what you've seen in mice will be to what happens in other animals, including humans?

Other species, including humans, face very different challenges compared to mice, obviously. And another thing to consider is that the amygdala is also about risk-taking.

A mouse must use different strategies to survive, compared to humans. In humans and other primates, taking risks is probably a fundamental factor that drives survival because you have to explore to survive. But in mice, exploring too much can be a deadly decision. 

I think probably there are similar principles covering the neuronal activity in all mammalian amygdalas. The neurons are similar, if not the same, the plasticity mechanisms will be the same, and the circuits will be very similar. So what they can compute is probably not that much different, but the inputs they get and the outputs they create are going to be very different. The way the computations are used to drive defensive or survival behaviour – that’s going to be completely different.

I think understanding the role of the amygdala, or any brain structure, is important. The amygdala is just one node in the complex system of the brain. I think understanding that node and how it interacts with other nodes might be key to understanding how larger brain networks function, both in health and disease.

About Dr Andreas Lüthi

Andreas Lüthi obtained his PhD in Neurobiology at the University of Basel, Switzerland. After postdoctoral stays in Bristol, UK and in Zurich, Switzerland, he established his own research group in 2000, initially at the Biozentrum of the University of Basel and then at the Friedrich Miescher Institute for Biomedical Research. His lab addresses how neuronal circuits can generate behaviour with a particular emphasis on synaptic, cellular and circuit mechanisms underlying learning and memory using classical conditioning as a model system. Recently, his lab became interested in the neuronal network mechanisms underlying the generation of emotional states.