293 — The logic of single-cell projections from visual cortex

This research looks closely at neural projections from V1 in mouse cortex: Though we’ve known for a long time that V1 projects to many different brain areas, it’s never been clear exactly what the distribution of those proejctions looked like. Perhaps all projections find downstream targets more or less randomly? Maybe different areas of V1 seek out different downstream regions? Or maybe the targets are highly specific, with a single neuron synapsing in one precise brain area — no randomness at all?

To answer this question, the authors looked used MAPseq — a new technique that virally infects individual neurons and, using a process similar to the stochastic-assortment used to endogenously generate antibodies, randomly assigns them a genetic “barcode.”

These barcodes are actively transported to the synaptic terminals of axons, and this makes it possible to sequence parts of the brain in-situ in order to find out which of the infected neurons synapse there.

Based upon this method, the researchers confirmed prior research that stated that the primary conduits of information out of V1 are layers 2/3, 5, and 6. Furthermore, they learned that almost all 2/3 neurons projected out of V1 to other brain regions, and specifically, the locations of the soma in V1 correlated with the target area.

Unsurprising: Neural projections from V1 aren’t random after all. And we can statistically verify that, if you please: If the probability of an axon projecting to area A is random and independent from the probability of the same axon projecting to area B, then $P(A) \times P(B) = P(A\cap B)$. But this does not hold, based upon the MAPseq data. So that’s neat.

In fact, it’s very neat: The actual counts of projection targets per neuron are higher than previous studies have counted, which means that MAPseq is a more reliable method for determining synaptic targets on whole-brain datasets than other methods like retrograde markers.