In vertebrate nervous systems, vast interconnected networks of neurons underlie brain function. Connectomics is a new field that attempts to map out these networks. At the forefront of this effort is high throughput serial electron microscopy (EM) which can reveal every cell, synapse and organelle in a piece of tissue (Kasthuri et al., 2015: PubMed , Cell, PDF). By extending this approach across large volumes, brain networks become visible.
In “The fuzzy logic of network connectivity in mouse visual thalamus” (Morgan et al., 2016, PDF) we describe a connectomic study of a 100 trillion voxel EM volume within the mouse lateral geniculate nucleus, the primary thalamic relay between retina and the visual cortex. The plan was to characterize the distinct parallel visual pathways as defined by different types of retinal cells that innervate the thalamus. However, we were surprised to find that most of our preconceptions about the way the circuit would be organized were not consistent with what we actually observe. To begin with, the distinct parallel channels of visual information that we planned to characterize are not, in fact, distinct. We find that some thalamocortical cells are innervated by only one kind of retinal ganglion cell axon whereas other thalamocortical cells are innervated by two types. Individual thalamic neurons can, therefore, sample from one or multiple channels of visual information. We further found that individual retinal ganglion cell axons can innervate sets of thalamocortical cells that vary in their dendritic morphology, input number and synaptic structure. A single channel of visual information is, therefore, likely to be processed in different ways by different thalamocortical cells. In short, our connectomic analysis reveals a complicated pattern of synaptic connectivity that does not obey expected cell type boundaries.
Despite the complexity, the network is clearly not just a random set of connections. We observe strong preferences in terms of which particular axons establish synaptic contact with which target cells. The most surprising of these biases operates at the level of individual dendrites. We find that groups of retinal axons that cluster their inputs on a single dendrite of one thalamic neuron sometimes travel as a bundle to connect to a single dendrite of a different thalamic cell thereby reproducing the same input pattern on different neurons. Because of this dendritic specificity, different dendrites of the same thalamic neuron have different network associations.
These results argue that what is going on in the thalamus is more complicated than we thought. Because we find examples of nearly every combination of inputs on different thalamic neurons, the work raises the interesting question of why the system should produce so much variability.
Read more in Harvard Gazette