In the real world, odours don't happen one puff at a time. Animals move through, and subsequently distort, plumes of odour molecules that constantly drift, changing direction as the wind disperses them. Now, by exploring how animals smell odours under naturalistic conditions, Rockefeller University scientist Maria Neimark Geffen and her colleagues reveal that the brain encodes these swirling and complex patterns of molecules using surprisingly little neural machinery. The findings suggest a new theory of how animals smell.
In their work, which will appear in the 26 February issue of Neuron, Geffen, a fellow at Rockefeller's Centre for Studies in Physics and Biology, analysed the brain activity of locusts as they smelled plumes of different odours. The plumes were generated by odour molecules released for varying durations and at varying intervals - not in metronome-like pulses as is typically done in odour studies. 'In their habitat, animals don't have the luxury of smelling something for one second and then trying to figure out what it is,' says Geffen. 'They are getting this ever-changing signal. So how does the olfactory system encode the dynamics of that signal?'
The answer, it turns out, is surprisingly simple.
When Geffen and her colleagues from Harvard University and California Institute of Technology initially looked at how the olfactory system responded, the results looked daunting. Even though a small population of neurones was activated in response to each odour, the pattern of activation differed from neurone to neurone. Consider, Geffen says, that each neurone is the source of a flashing light. If the population of neurones all started flashing, each neurone would appear to be flashing in a frenetic and uncoordinated fashion relative to one another. It would be hard to envision a set of rules that coordinated such complex activity. But by looking at how the population of neurones function together, Geffen and her colleagues found that these vastly different responses could be explained by a very simple model.
'Since we know the rules that determine how these neurones behave, we can now create a model that can predict what this network in the locust olfactory system is doing,' says Geffen. 'So, basically, this model summarises, in a very sparse way, how these neurones all work together to process complex odour signals that we encounter everyday. It's evolution at its most elegant.'
Geffen found that her model not only could predict the firing pattern of a neurone from the sequence of odour molecules, but it could also do the reverse: predict the odour's temporal sequence from the neurone's firing pattern. If Geffen used one neurone, the model could predict the odour's time course with 95 percent accuracy. If they used two neurones, the accuracy jumped to 97 percent. 'After two, it doesn't matter,' says Geffen. 'What that means is that even though there is this variety of responses, individual neurones preserve almost the full information about the precise temporal dynamics of the odour.'
The work builds upon two theories on how the animals encode smells. While one theory argues that subsets of neurones carry information about different odours, the other argues that the brain recognises different odours by the firing pattern of a vast number of neurones. 'What this research offers is a sparse solution that merges these two hypotheses,' says Geffen. 'What we are saying is that once you know which neurones to look at, the timing of when and for how long these neurones respond to odours is how the brain encodes smell.'