As a graduate student in the Gabbiani lab at Baylor College of Medicine, I studied how locust brains processed sensory information about an approaching predator and gave rise to timely escape behaviors. I simulated approaching predators as dark disks expanding on a light background and watched locusts jump, while measuring their brain activity. The challenge with those experiments was that locusts stopped responding to looming stimuli after a few trials and I could no longer use the ‘habituated’ individual for my experiments. When I joined the Engert lab as a postdoc, my aim was to address this exact challenge, i.e. to understand what are the neural underpinnings of habituation to sensory stimuli that are deemed non-threatening based on experience.
Habituation to repeating irrelevant sensory stimulation is considered the simplest form of learning and is ubiquitous in animal kingdom. How do animals learn that a simulated predator is safe to ignore? What are the changes that happen in the brain as the animal stops responding?
Just like locusts, larval zebrafish respond to looming stimuli, this time with a quick tail flick, followed by swimming away from the direction of the stimulus. These fish provide a great system to study brain-wide changes that occurs in a vertebrate brain: they are transparent, and it is possible to measure their whole brain activity using non-invasive functional imaging in transgenic fish whose neurons glow when they are activated. All of this can be done in a larva whose head is kept immobile for imaging, while their tails are free to allow simultaneous measurements of escape attempts. Using this system, I showed looming stimuli repeatedly to the larvae and measured their brain activity as they learned to ignore these innately threatening stimuli.
One hypothesis we had about changes that could occur in the brain, was that the activity in all cells, will simply decline and therefore, excitation to motor networks will decrease. A second hypothesis was that some neuronal populations might become potentiated with stimulus repetition and inhibit the ones that drive the escape response. Looking into the fish brain, we measured the activity in two major population of neurons that were previously known to responded to a dark looming stimulus. Ones that were sensitive to the ‘expansion’ aspect, which we called looming sensitive (LS) and others that were tuned to the overall dimming of the visual field that occurs during the expansion (dimming sensitive, DS). We found that the LS neurons decreased their response amplitude with repeated stimulation. Consistent with our second hypothesis, we discovered a certain population of DS neurons that potentiated their response with repeated stimuli. Through a series of behavioral and imaging experiments we showed that these neurons are in fact critical for habituation learning, and that they likely inhibit the activity of LS neurons, which in turn drive the motor response.
Notably, dimming stimuli on their own are not threatening to the fish, i.e. the fish do not generate an escape response to dimming. Our findings indicate that the brain can implement habituation to threatening stimuli by selectively potentiating pathways that respond to non-threatening constituents of those stimuli, in this case dimming.
In summary, our data allow us to propose a realistic circuit model where a subset of inhibitory, DS neurons are incrementally potentiated by repetitive stimulation and where they serve to locally depress the looming selective relay pathway. As such, this model generates a series of testable predictions about behavior, neural response properties and synaptic connectivity which are all eminently testable and straightforward to validate – or invalidate – in future experiments.