ELECTROPHYSIOLOGICAL AND OPTICAL METHODS FOR MEASURING INPUTS TO INDIVIDUAL NEURONS

Abstract

Understanding neural circuits requires being able to explain the mechanisms that result in the tuning functions of neurons during behavior. Such mechanistic explanations typically involve hypotheses about the tuning of both the postsynaptic neuron and its presynaptic partners. Despite the prevalence of these types of models, testing them systematically remains technically challenging.

Here we first illustrate one strategy for model evaluation that exploits the fact that, in special cases, different circuit architectures proposed by different model families make distinct predictions about the membrane potential of neurons during behavior. Using whole cell recordings in mice navigating in virtual reality, we evaluate two model families explaining the computation of grid fields in the medial entorhinal cortex (MEC) and show that the membrane potential is consistent with predictions made by attractor models and inconsistent with interference models. We identify grid cells with very large intracellular theta oscillations but show that, across neurons, slow ramps of depolarization are the likely explanation for field formation.

We next develop strategies that build towards a general method for measuring the tuning of neurons and their inputs during behavior by using spine calcium imaging. By combining sparse GCaMP6s labeling, mouse screening for low brain motion, and a large number of environmental traversals with specialized methods for motion correction and analysis, we measure place fields from CA1 somas and individual dendritic spines in both strata oriens and radiatum (these are predominantly from Schaffer collaterals). Place cells receive place inputs that span the available environment. Nevertheless, there is a bias in the density and/or strength of inputs agreeing with the postsynaptic somatic tuning. These data suggest that place cells maintain their flexibility by remaining synaptically connected with other place cells with unrelated tuning, and yet form precise spatial fields in part by having a slight connectivity bias with similarly-tuned presynaptic partners. These data are consistent with the emerging view that, instead of grid cells forming place cells by themselves, place fields are computed, sharpened or maintained through a collaboration between the different types of inputs received by the hippocampus.