Learning of Drosophila at the Flight Simulator: Classically Conditioned Visual Pattern Discrimination (mushroom bodies and cAMP signaling)
|Help SiteNavigator Index Where Am I? Search Page Status||Contact Us|
Reinhard Wolf, Marcus Dill, Dirk Eyding, Tobias
Wittig and Martin Heisenberg
Theodor-Boveri-Institut für Biowissenschaften der Universität Würzburg, Lehrstuhl für Genetik, Am Hubland (Biozentrum) 97074 Würzburg, Germany
In the Drosophila flight simulator, single flies learn to associate the occurance of a life threatening heat beam with their flight direction in an artificial visual panorama (Fig.1). Our experiments reveal the following capabilities of the fly:
During training in the flight simulator, the mode of conditioning is ambiguous:
However: Evidence for operant conditioning is provided by the 'replay experiments' (Fig.3).
The role of the mushroom bodies in learning: Flies with genetic or ontogenetic mushroom body (MB) lesions learn as well as normal WT-flies in the operant paradigm (Fig.4). It had been shown earlier (Heisenberg et al., 1985; deBelle and Heisenberg, 1994) that the MBs are required for odour discrimination learning. Besides the different conditioned stimuli (visual patterns vs. odours) and the context a major difference between the present and the earlier work is the operant and the classical conditioning. Hence, the question:
Are the mushroom bodies necessary for classical learning?
Replay training is classical conditioning, by definition. However, during replay training with an average avoidance index of PI=0.7 the heat-associated pattern orientations are presented to the fly for only 15% of the time. Therefore, we designed a new training procedure for classical conditioning (Fig.5) and tested normal WT-flies as well as flies without MBs:
Both, flies with and without MBs, are able to learn just as well in the classical, non-olfactory (visual) learning paradigm. Thus, it is the visual modality that does not require the MBs.
Learning mutants dnc and rut: The genes dunce (dnc) and rutabaga (rut) are components of the adenosine 3',5'-monophosphate (cAMP) 2nd messenger cascade, are required for olfactory learning and are preferentially expressed in the MBs (review: Davis 1993). These genes have therefore been discussed as additional evidence for a role of the MBs in olfactory learning (e.g. Davis 1993). We show here (Fig.6) that pattern discrimination learning in the flight simulator does require dnc and rut. Thus, even outside the MBs the two genes are involved in associative conditioning.
In a variant of the procedure shown in Fig.5 the new classical conditioning method reveals that flies can store fully stabilized images (Fig.7). Motion detection is not a necessary step in pattern recognition. Moreover, flies can retrieve the memory template of a stationary image while the actual image is continuously moving.
Fig.1: A single fly, glued to a small hook of silver wire and attached to a torque meter, is flying stationarily in the center of a cylindrical panorama (arena). For all experiments described here, four equally spaced, T-shaped black patterns, two of them upside-down, are used as visual landmarks. In the flight simulator mode (closed loop), the rotational speed of the arena is made proportional to the fly's recorded yaw torque around its vertical body axis. This enables the fly to stabilize the rotational movements of the panorama (i.e. to fly straight) and to adjust certain flight directions with respect to the visual landmarks. Yaw torque and flight direction of the fly are recorded continuously and stored in the computer memory (sampling frequency 20 Hz). An infrared beam heating the fly is used for reinforcement (Wolf and Heisenberg, 1991).
Fig.2: No spontaneous preference for the subsequently 'heated' pattern orientation is apparent during the 3x2min preference test (bright yellow bars). During the following 4 minutes of training (orange bars) the flies switch off the heat for about 83% of the time. Already after this short training period, they continue for the next two minutes to avoid the pattern that previously was combined with the heat (first yellow bar). After another 4 min training period, a solid learning effect is obvious from the continuous preference for the flight direction towards the 'non-heated' pattern (last four yellow bars).
The block diagram in the upper part of the figure depicts the operant training (and test) paradigm.
For simplification, in the following figures (except Figs.5 and 7) the PI's for preference test, training and learning tests are summarized to only give one bar each.
Fig.3: In the lower left diagram, 4 minutes of operant training (master experiment; orange bar) are followed by 4 min of learning test (yellow bar). In the right diagram, heat and flight directions of the master experiment are replayed to 16 other, unexperienced flies (dark red bar) which then are tested in the same way as their predecessors. Whereas learning is indicated from the positive PI in the left diagram, no learning appears after the replay training. Since both groups of flies are trained with exactly the same stimulation procedures, the different result must be due to the only difference between the two training paradigms: In the right one the fly has no control of the stimuli. Thus, during master training in closed loop, the flies obviously learn from the consequences of their own behaviour (= operant learning).
Fig.4: The mutant mushroom body miniature (mbm1) and, as control, Wildtype Canton S were tested in the operant paradigm. Both of them, mbm1 as well as Canton S, show positive learning after the operant training. Even flies, whose mushroombodies are ablated using the Hydroxy-Urea method (HU; deBelle and Heisenberg, 1994), do not have a significant reduction in operant avoidance and learning if compared with the respective WT (Berlin) controls. In the experimental animals brains are inspected post mortem. Only animals without MBs are included in the behavioural data. From both results: mbm1 and HU, it is evident, that the MBs are not involved in this learning task.
Fig.5: Classical conditioning paradigm (see also in summary). After a preceeding preference test of 4 minutes, the flight simulator is switched to open loop conditions with one of the patterns kept stationarily in front of the fly. Every 3 seconds, the computer program switches the panorama to the next pattern position (+ or -90° in 220ms). With one of the patterns the heat is on and with the other it is off. This procedure is repeated periodically during the whole training period of 4 minutes. During Test (closed loop), PI is evaluated as defined above, except that th and tc indicate the time during which flight is directed towards a narrow sector of ±3° around the center positions of the patterns. In contrast to the (also classical) replay training, with this procedure normal and HU-flies learn to avoid the 'hot' patterns.
Fig.6: The biochemical learning mutants dnc1 and rut1 do not learn after operant training (control: Canton S). For further details and conclusions see summary.
Fig.7: Any motion can be eliminated during training if light is turned off while pattern positions are switched. For further details and conclusions see summary.
Heisenberg, M., Borst, A., Wagner, S. and Byers, D. (1985): Drosophila mushroom body mutants are deficient in olfactory learning. J. Neurogenet. 2: 1-30
deBelle, J.S. and Heisenberg, M. (1994): Associative Odor Learning in Drosophila Abolished by Chemical Ablation of Mushroom Bodies. Science 263: 692-695
Davis, R.L. (1993): Mushroom bodies and Drosophila learning. Neuron 11: 1-14
Wolf R, Heisenberg, M. (1991) Basic organization of operant behavior as revealed in Drosophila flight orientation. J Comp Physiol A 169: 699-705
Go to Table Of Contents | Go to Explanations | Go to Top of this page | Go to List of Posters
Add comment to poster
|Flybrain Poster Sessions Roles of the E(spl) Genes|
All contents are copyright © 1995-2000 Flybrain or their original publication (as noted).
Copyright and use policy
PP00005Page last modified: July 25, 2000 by Managers.