1. Mushroom-bodies regulate habit formation in Drosophila
Björn Brembs
Freie Universität Berlin, Inst. Biol. – Neurobiologie

2. Decisions, Decisions Spontaneous choice behavior Monday, July 14, 2008
FENS-Forum 2008: S28 – The neurobiology of choice and decision-making.

3. Drosophila at the Torque Meter
Spontaneous choice in a constant environment
How flies use re-afferent feedback to modulate this spontaneous choice behavior
Monday, July 14, 2008
FENS-Forum 2008: S28 – The neurobiology of choice and decision-making.

4. FENS-Forum 2008: S28 – The neurobiology of choice and decision-making.
Feedback Stimuli
Fly drawing courtesy of Reinhard Wolf
Monday, July 14, 2008
FENS-Forum 2008: S28 – The neurobiology of choice and decision-making.

5. FENS-Forum 2008: S28 – The neurobiology of choice and decision-making.
Associative Learning
The dichotomy of operant and classical conditioning
classical ?
operant ?
Monday, July 14, 2008
FENS-Forum 2008: S28 – The neurobiology of choice and decision-making.

6. Dissecting Operant Learning
Classical alone
Operant alone
Operant + Classical
Brembs and Plendl, Curr. Biol. (2008, in press)
Monday, July 14, 2008
FENS-Forum 2008: S28 – The neurobiology of choice and decision-making.

7. Multiple Memory Systems
Operant and classical components interact hierarchically ? ?
classical
operant
classical +
Brembs & Heisenberg (2000)
Monday, July 14, 2008
FENS-Forum 2008: S28 – The neurobiology of choice and decision-making.

8. Isolating the Components
Test:
Classical + operant components 
Training: operant + classical 1 2
Operant component alone  3
One reason for a dominant fact-learning process suppressing skill-learning may be to render our fact-knowledge independent from the behavior we acquired it with. In other words, keeping facts isolated from the means by which we learned them may help generalizing our fact-knowledge to different situations. This can be tested in flies as well. First, we train the flies in composite conditioning as before. The red bar shows the successful conditioning. After training we now test the flies in a completely different situation. Similar to the T-patterns before, the flies now control the angular position of four identical stripes. Flying towards one pair of opposing stripes, for example, switches the color to green, flying towards the other pair produces blue illumination. Here, in fs-mode, the flies have to fly straight to keep one color, while in sw-mode, they need to constantly turn. The situations are very different. Testing the flies in this fs-mode after training, does not reveal any color preference. However, if they receive a short 60s familiarization or reminder training, which alone is not sufficient to produce a significant learning score, they show that they have transferred their fact knowledge to a different behavior in a generalization experiment. This transfer effectively replaces the training behavior with a new behavior, testing the classical component in isolation. We now have the tools to isolate both components in composite conditioning and can proceed to screen for the biological substrate mediating their hierarchical interactions.
Classical component alone 
Brembs & Heisenberg (2000)
Monday, July 14, 2008
FENS-Forum 2008: S28 – The neurobiology of choice and decision-making.

9. Mushroom-Bodies central
…for keeping classical memories flexible for generalization
Mushroom-bodies mediate inhibition of operant learning
Mushroom-bodies enable flexible generalization
Mushroom-bodies delay acquisition of stereotyped, habitual responses   
"classical" (fact)
The mushroom bodies are a prominent insect neuropil. They are depicted in green in this 3D rendering of the Drosophila brain. It is known that the mushroom bodies are necessary for a different form of generalization, context generalization. Because of this previous evidence, we blocked mushroom-body output by expressing tetanus toxin in mushroom-body Kenyon cells. These flies were then tested in the paradigms explained before: they were trained in composite sw-mode and then tested either with both predictors present, shown in red, or with the two predictors isolated. Green bars depict the PI of the operant component and blue bars the PI of the classical component. As you can see, in contrast to the wildtype results, flies with blocked mushroom body output do not suppress the operant component and cannot generalize the classical component. The genetic control strains reproduced wildtype behavior by suppressing the operant component and showing generalization. These results are consistent with the hypothesis that you need to suppress behavioral learning in order to keep your knowledge about your environment flexible. Taking this idea one step further, one may speculate that behavioral learning limits your behavioral repertoire and makes it more stereotyped. Habits are a form of stereotyped behavior and can be very persistent. Let's say you take the freeway to work every day. You take exit 5, turn left and end up at your workplace a few minutes later. On a Sunday, you plan to go to the nearby park, but instead of taking exit 6, which would lead you straight to the park, out of your habit, you take exit 5 and end up at your workplace instead of the park. William James in his famous 1890 "principles of psychology" writes that “very absent-minded persons in going in their bedroom to dress for dinner have been known to take off one garment after another and finally to get into bed, merely because that was the habitual issue of the first few movements when performed at a late hour”. Repetition is what is usually required to form a habit. If the operant component in these experiments is akin to habit formation, increasing the amount of training may lead to behavioral learning which is normally suppressed. To test this hypothesis, we doubled the amount of training and indeed what we find is habit formation: the behavior becomes stereotyped - even if the colors are missing, the behavioral bias is kept. Mimicking the results from the transgenic flies with blocked mushroom body output, the behavioral learning prevents generalization. This experiment constitutes the first invertebrate model for habit formation and parallels similar vertebrate experiments.
"operant"
(skill)
Brembs (subm.)
Monday, July 14, 2008
FENS-Forum 2008: S28 – The neurobiology of choice and decision-making.

10. Complex, Hierarchical Interactions
between multiple memory systems instead of a simple dichotomy
Monday, July 14, 2008
FENS-Forum 2008: S28 – The neurobiology of choice and decision-making.