(1/n) What circuit generates selectivity in L2/3 cortical neurons? We answer this question in our paper:

Spatial connectivity matches direction selectivity in visual cortex

out today in @nature with @kennethd_harris and @MatteoCarandini. https://www.nature.com/articles/s41586-020-2894-4

(2/n) Primary visual cortex (V1) is renowned for its selectivity for orientation and direction. Neurons in L4 obtain this selectivity by summing thalamic inputs displaced in space and time.

https://doi.org/10.1113/jphysiol.1962.sp006837
https://doi.org/10.1038/s41586-018-0148-5

(3/n) How about downstream cortical neurons, in L2/3? The classical view is that they inherit selectivity from co-tuned L4 inputs, and amplify it by recurrent connectivity with co-tuned L3 neurons.

(4/n) To test this view, we used rabies to trace the excitatory and inhibitory presynaptic neurons connecting to an individual L2/3 postsynaptic neuron. We used multispectral 2p imaging to record the activity flowing through this circuit during visual stimulation.

(5/n) The tuning of the excitatory presynaptic ensemble matched the selectivity of the postsynaptic neuron for orientation but not for direction. The inhibitory presynaptic ensemble was not tuned for direction either.

(6/n) With this kind of inputs, how can an L2/3 neuron be selective for direction?
To solve this mystery, we looked at the spatial organization of presynaptic inputs.

(7/n) When we plotted the data in visual space, aligned by the postsynaptic preferred direction, a clear pattern emerged! Excitatory neurons aligned to preferred orientation and favoured the region opposite of preferred direction. Inhibitory neurons were mostly on the other side.

(8/n) The offset between excitatory and inhibitory presynaptic ensembles correctly predicted the direction preference of each postsynaptic neuron.

(9/n) This success suggests that L2/3 neurons could compute direction selectivity from spatially and temporally offset excitatory and inhibitory inputs.

https://doi.org/10.1016/s0896-6273(00)80991-5
https://doi.org/10.1038/nature11665

(10/n) This model makes a prediction: neurons whose excitatory and inhibitory inputs are more spatially offset should show stronger direction selectivity. Sure enough, our data confirmed this prediction.

(13/n) In conclusion, we discovered that L2/3 of the cortex computes directions selectivity de novo, by summing from offset excitatory and inhibitory neurons.

(14/n) A similar trick is used by the retina. We would have loved to be able to tell Horace Barlow about this discovery.

https://doi.org/10.1113/jphysiol.1962.sp006837
https://www.theguardian.com/science/2020/aug/23/horace-barlow-obituary