Fruit flies are attracted to some chemicals and repelled by others, but rather than being in-built, these preferences can be influenced by experience and learning. The insects can be easily trained to become averse to a specified odor, by the method of classical conditioning with is commonly used in other experimental animals such as rats and mice. For example, when an odour is repeatedly paired with an electric shock, a fly will learn the association between the two, and actively avoid that odour when it next encounters it. If it is then exposed to the odour in the absence of the shock, the aversive memory will be extinguished.

Encoding of these olfactory memories is thought to require the actions of the neurotransmitter dopamine on neurons called Kenyon cells, which are located in a brain structure called the mushroom body. The Kenyon cells also receive olfactory information from the antennae, so are presumed to involved in encoding and storing associative memories. But although the fruit fly's brain contains 200-300 dopamine-producing cells, the exact source of the dopamine signal during aversive reinforcement has not been identified.

Gero Miesenböck and his colleagues used genetic methods to generate a strain of fruit flies which express an ATP receptor called P2X₂ in specified subsets of dopamine-producing cells in the brain. The flies were then injected with a form of ATP that remains inactive until a light-sensitive bond which attaches it to a carrier molecule is broken. The flies were then placed into a narrow chamber in which two air streams, each carrying a different chemical odour, converged. When the flies were given small electric shocks while exposed to one of the streams, they quickly learned a negative association with the odour in the stream, and subsequently avoided it.

The researchers found that they could elicit the same avoidance behaviour by using pulses of light to evoke the release of dopamine. The laser pulses, triggered by a fly's entry into one of the streams, caused the "caged" ATP to be released, leading to P2X₂ receptor activation, and dopamine release. This was sufficient to induce the formation of aversive memories, leading to avoidance behaviour that was no different from that of flies conditioned with electric shocks. This confirmed that olfactory-driven aversive associations are indeed mediated by the actions of dopamine, but did not identify where the actions take place.

To determine the source of the dopamine signal, the researchers used another genetic technique to label four different subsets of dopamine-producing neurons, and to selectively silence each of them one by one. Silencing one of these subsets was found to reduce the movements of the flies, and to eliminate aversive reinforcement altogether. Microscopic examination revealed that a cluster of just 12 cells provides the dopamine signal required for memory formation. These cells, called PPL1 neurons, sit to the side of the mushroom body, and send elaborately branched processes into it.

The mechanism of aversive memory formation is still unclear, but this study gives a clearer picture of how it might work. A memory would be written when olfactory information from the antennae and the dopamine signal from PPL1 neurons converge on Kenyon cells. This would strengthen the synapses in the circuit (by a process known as synaptic plasticity), and the information would be stored in the weight of the connections. Once the memory trace had been laid down, it could be read when the fly re-encounters the learned odour, which would cause Kenyon cells to re-activate the circuit. Finally, the output neurons would compute the weight of the synapses and initiate the appropriate behaviour.

In primates, dopamine-producing midbrain cells also transmit information regarding movement and associative learning, but they encode reward rather than aversion. Learning is thought to be driven by changes in the expectations about rewards, and the resulting changes in synaptic strength are believed to reflect prediction errors rather than the size of the rewards. Behaviour changes accordingly following a mismatch between prediction and outcome, and learning stops when the prediction error is zero. A comparison of how PPL1 neurons respond to predicatable and unpredicatable aversive stimuli could therefore establish whether this type of learning in the fruit fly is driven by prediction errors, and elucidating the fine structure of the cells' braching patterns is likely to provide more information about how aversive memories are written, stored and read.

• Read more about memory
• Subscribe to my RSS feed
• Follow me on Twitter
  

Claridge-Chang, A., et al. (2009). Writing Memories with Light-Addressable Reinforcement Circuitry. Cell 139: 405-415. DOI: 10.1016/j.cell.2009.08.034.

Lima, S. Q. & Meisenbock, G. (2005). Remote Control of Behavior through Genetically Targeted Photostimulation of Neurons. Cell 121: 141-152. [PDF]