For senses like sight, touch, and hearing, there are ways to map grids of neurons in cortical regions that make sense in terms of the external reality. But for smell there is seemingly no such map. So how are cortical neurons distributed based on their odorant input?
Three recent papers (Sosulski et al, Ghosh et al, and Miyamichi et al) used different techniques, all in mice, to attack this question from different angles. It’s interesting to look at the differences in their methods, because they made some very similar discoveries.
Sosulski et al labeled mitral cells with TMR-dextran, which allows for retrograde tracing from individual (GFP-expressing) glomerular cells all the way to their axon termini in different cortical regions. This produces beautiful confocal images. See, for example, their fig 2b (if you have access), in which the projections from a single glomerulus are shown across all olfactory areas in a flattened sagittal slice.
Ghosh et al labeled clusters of mitral cells with attenuated Sindbis virus, which was altered to lead to the expression of GFP (making cells green), RFP (making cells red), or both (making cells yellow). They also looked at mitral cells specifically coming from specific GFP-expressing glomeruli. They then used confocal microscopy and 3d reconstruction to analyze the distribution of these projections within various cortical regions.
Miyamichi et al used a dual viral system to trace the afferent axons of the anterior olfactory nucleus, piriform cortex, and cortical amygdala. Their first virus contained genes necessary for the expression of their second virus, which was modified to contain a fluorescent marker, and which they injected two weeks later. Their second virus was rabies, so it can cross synapses, but once in the pre-synaptic neuron it won’t have access to proteins it needs to replicate and so it won’t be able to spread further. Therefore, their system allows for long-distance single trans-synapse tracing, which they visualized with confocal microscopy and reconstructed the images of in 3d.
So, all of the three papers are looking at the distribution of axons from mitral neurons to cortical regions, but the first two studies are tracing from the olfactory bulb to the cortex, whereas the third study is tracing back (in the context of info flow) from the cortex to the olfactory bulb.
All three studies found that the distribution of connections from the mitral cells to the piriform cortex was uniform, meaning that there was no spatial bias or tendency for axon termini to be located in a particular subregion.
In contrast, the studies found that connections from mitral cells to the cortical amygdala and anterior olfactory nucleus had particular patterns. Sosulski et al showed that the correlograms between glomerular inputs and cortical amygdaloid axon termini were patchy, indicative of a stereotyped spatial bias. Similarly, Ghosh et al showed that axons traveling from the olfactory bulb to the anterior olfactory nucleus pars externa retained their spatial segregation.
Replicating both of these, Miyamichi et al used their 3d reconstructions, transformations of inputs along the different axes, and permutation tests to show that 1) there is a dorsal bias in input from the olfactory bulb to the cortical amygdala and 2) there is a dorsal–ventral axis correlation between the olfactory bulb inputs and the anterior olfactory nucleus (see their figure 3e).
The Ghosh et al study looked at the projections coming from “homotypic” cells which have inputs from the same glomerulus, and showed repeatedly that they do not have significant overlap. This indicates that mitral neurons containing the same odorant info do not branch in similar locations. This sets smell apart from other sensory systems.
One of the major differences in their findings is how the authors interpret them. Sosulski et al and Miyamichi et al emphasize the similarities of their findings to the Drosophila olfactory system, because in both systems, piriform cortical inputs are more random and allow for learning and plasticity, while amygdala cortical inputs are more determined and innate.
On the other hand, Ghosh et al emphasize the differences between their findings and the Drosophila system, where projections from the neurons homologous to mitral cells that contain the same info are more stereotyped. The diversity in mice might allow for more combinatorial control in odor perception, allowing for different time courses and concentrations of odors to be recognized as distinct. But it’s difficult to understand how the diversity of homotypic mitral cell projections could be accounted for by a simple pattern of neural activities or molecular guidance cues. Developmental mechanisms are far from fully elucidated, but developmental biologists could use this as a challenging finding to attempt to explain. Perhaps they will discover that there is a method to the organizational madness after all.
Miyamichi K, et al. 2011 Cortical representations of olfactory input by trans-synaptic tracing. Nature doi:10.1038/nature09714
Ghosh S, et al. 2011 Sensory maps in the olfactory cortex defined by long-range viral tracing of single neurons. Nature doi:10.1038/nature09945
Sosulski D, et al. 2011 Distinct representations of olfactory information in different cortical centres. Nature doi:10.1038/nature09868