Something close to home. The Scientist and New York Times reports that Nobel Laureate Linda Buck retracted two papers recently. This follows a retraction of a Nature paper from 2001. All three papers featured work done by a post-doc, Zhihua Zou. The retracted papers have no bearing on the work Buck and Richard Axel had performed in identifying the family of G-protein coupled receptors, which won them the Prize.
Rather than focusing on the work in the retracted papers, I would like to explain why, fortunately enough, the retractions do not substantially alter our view of how olfaction works.
First and foremost, independent researchers, using independent means, have found similar results presented in the retracted papers. The main points from the three retracted papers are,
1) Using a genetically encoded neuronal tracer, the paper purported to show that neurons that express the same olfactory receptor connect to neurons that wire to the same brain regions responsible for olfactory processing (Zou et al, 2001).
2) That a marker of neuronal activity, c-fos expression, showed that activity patterns in olfactory cortex is reproducible across animals and are typical for a given smell, a molecule of which is termed “odorant” (Zou et al., 2005).
3) That mixtures of smells activate neurons that respond to the components individually (the pattern of activation is a summation the patterns evoked by single odor components.) (Zou and Buck, 2006).
The peripheral olfactory system can be described as follows. The primary, sensory neurons are situated in the nasal cavity and are responsible for detecting odor molecules. These neurons form connections with neurons of the olfactory bulb. In turn, OB neurons project to “higher olfactory centers”, which includes the piriform cortex.
For the first retracted point, there already exists research showing that connections into the piriform cortex from the olfactory bulb is both convergent and divergent. This can be shown by labeling small groups of neurons in the olfactory bulb, and then watching where the labels wind up. In the piriform cortex, one can see the label over large areas even if the label started out in a confined area in the olfactory bulb, thus showing divergence. A small location in the piriform cortex also receives neurons from all over the bulb (i.e. convergence) when using labels that travel from the cortex to the bulb.
The specific detail offered by the Buck group is that the neurons connecting to piriform cortex share a common origin. That is, the sensory neurons in the nose connect to olfactory bulb neurons that in turn connect in to clusters of physically near neurons within the piriform cortex. The groupings at this level suggest that the piriform cortex could be built from many such groups of neurons. Thus, when an odor molecule activates receptor neurons in the nose, eventually, clusters of activity could be found in the piriform cortex. These spatial patterns may result from the sums of all the receptor neurons that were activated (both these points were covered in Zou et al., 2005 and Zou and Buck, 2006). The spatial organization may reflect an (still unclear) advantage or need for neural processing.
That is a very rough sketch of some basic ideas in mammalian olfaction. As noted, tracing experiments, performed in separate labs with different methods show that there is some structure in where neurons form projections. Whether one can make specific statements linking a response and/or connection to some neurons in the nose is at issue.
One should note that the retractions from Buck lab do not indicate misconduct (yet) – i.e. doctoring and faking of data. The problems could have arisen in analysis. Indeed, Illig and Haberly, in 2003, using the same c-fos methodology to indicate activity, found that the piriform cortex had widespread activation in response to odorant exposure. A “pattern” of activity was absent. Using both electrophysical recordings and optical indicators of neural activity, similarly wide-spread activity was also observed in mice and rats. Even in zebrafish, wide-spread activity within the olfactory cortex analogue was observed. The clustering seen by the Buck group could have arisen by chance clustering of the c-fos signal and could have been enhanced by the analytical techniques they used. I do not know how the data was analyzed to lead to the results published in the paper, but there are mistakes one can commit, without any malfeasance intended.
A note on the methods: there are some significant differences in the way activity is reported with c-fos when compared to electrophysiological and optical recordings. Expression of c-fos is linked to calcium influx in activated neurons. The usual method in evoking this response, to create enough signal against a background, is to expose the animal for 30-60 minutes, to a single odorant. In contrast, the other techniques show responses lasting less then a second (and at millisecond precision) in response to exposure to smells. Further, the chain of events leading from activated neuron to c-fos expression is unknown. For example, how many electrical impulses (i.e. action potentials) in neurons correspond to a given level of c-fos expression? While useful as a gross measure to identify areas of interest, the c-fos technique ultimately lacks some of the advantages researchers need to make definitive statements about smell processing at the actual time scales relevant to brain function.
That is an important distinction: techniques that are similar to those Linda Buck’s lab used have worked in other labs. The key point is that we can no longer use the tracing results that purported to show connections at 3 structures in the olfactory system. Although we no longer know the specific identities of the connected neurons, the general principle of convergent/divergent connections remain. As for her other conclusions in subsequent papers, there is enough evidence to suggest that the organization of activity in the piriform cortex occurs in the timing of neural responses and not necessarily by their physical locations. Further, the technique her group used to assess activity has disadvantages in assessing neural activity at millisecond time scales (which is the regime where neurons work. Thus the findings themselves, although retracted, do not alter at a deep level what researchers think about olfaction.