A computational model of LGN-V1 dynamics and the emergence of gamma-band rhythms

Gamma oscillations have been hypothesized to play a central role in active vision, contributing to inter-areal information selection and routing. Furthermore, emerging evidence suggests that gamma oscillations play a fundamental role in feedforward neuronal communication. In order to constrain these hypotheses it would be useful to know under what conditions gamma is prominent during active vision. We recently showed (Bastos et al., 2014) that gamma oscillations are an emergent property of the cortex, and are not present in the LGN of alert behaving monkeys. Here, we set out to explain these results in terms of an underlying biophysically realistic network model composed of Hodgkin-Huxley neurons. The prime candidate mechanism to explain our experimental results was recurrent inhibition: the anatomy of the cortex is full of recurrence, allowing excited pyramidal cells to activate inhibitory cells, which can then inhibit a large number of pyramidal cells and set the pace of the gamma oscillation (the classical “PING” mechanism – Pyramidal Interneuron Network Gamma rhythm). However, this recurrence of excitation back onto local inhibitory cells, appears to be lacking within the primate LGN (but perhaps not in the feline LGN) at an anatomical level. In our model of the LGN lacking local recurrence, the circuit did not engage in gamma oscillations even when the Thalamic Reticular Nucleus (TRN), a structure that provides inhibition onto relay cells and is excited by geniculocortical collaterals, was included. However, the cortex, which did contain recurrent inhibition, entered a rhythmic mode despite receiving arrhythmic LGN input. Intriguingly, even when the cortex projected back onto the LGN in a gamma rhythmic mode, this did not rhythmically entrain the LGN. The primary reason for this appears to be the synaptic time constants of corticothalamic synapses, which due to their slow time constants act as a low-pass filter. However, when we did include local recurrent inhibition in the LGN (as a putative model of feline LGN), gamma oscillations did emerge – which may explain interspecies differences between cats and primates. We conclude that the circuitry of the primate LGN is well-suited to ignore cortical gamma rhythms, to focus on a sparse and temporally precise neuronal code, and to maximize the feedforward throughput of information from the retina to the cortex.