Teaching Bacteria to Behave

Continued from page 1

The gene products pA and pB would persist in the cell and therefore act as a memory that lasts for a long time once they have been produced. Their concentrations are the equivalents of the synaptic weights in the Pavlovian-dog model. Only in conjunction with these molecules can iA and iB (the analogs of the smell and the bell) have their effects. By the researchers pairing the iA and iB, the bacteria is able to respond to iB, whereas before it only responded to iA. This means that the bacterium has been "trained" to respond to iB, says Fernando.

Eva Jablonka, a theoretical biologist at Tel-Aviv University and a leading researcher in the field, agrees. "This is conceptually a bit difficult," she says, "but if you look at the definition of learning--because of something happening, you have some kind of physical traces, and this changes the threshold of the response in the future--then this is what you have here." She adds, "I think that it is a good and potentially very useful paper, and I think they do demonstrate associative learning."

The model is based on the assumption that such a chemical-genetic circuit could be created and planted into a bacterium such as E. coli. "It seems to me quite possible at the theoretical level, and I don't see great obvious hurdles for the construction of the suggested vectors," says Jablonka, who published a paper on conditioning in single-celled organisms this month.

Significantly, Fernando estimates that the changes induced in the bacteria could easily persist for the 30-minute life cycle of an E. coli bacterium. This would make the changes, or "learning," heritable. This is an especially important point when it comes to medical applications for trained bacterium. "After all, diseases or drug doses are going to last longer than 30 minutes," notes Jablonka.

The trick would be to train bacteria to recognize chemical processes in the body that are associated with danger. This might be an adverse and dangerous reaction to a drug, or to the presence of tumor cells, indicating that a medicine in the system needs to be activated in certain tissues.

Research on genetically engineering remote-controlled bacteria to release drugs is already under way. In 2005, for example, a team from the National Institutes of Health proposed genetically engineering naturally occurring bacteria to release antiviral treatments for HIV. The realization that such bacteria might be trained to do this work more effectively could bring a whole new dimension to the field.