Purchase this article with an account.
Cameron S Cowan, Jasdeep Sabharwal, Samuel Wu; Space-time codependence of retinal ganglion cells can be explained by novel and separable components of their receptive fields. Invest. Ophthalmol. Vis. Sci. 2015;56(7 ):5864.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Reverse correlation methods such as spike triggered averaging are increasingly utilized to study the light sensitivity of retinal ganglion cells. While the linear receptive fields identified by these methods have a consistently identifiable spatial center, the classic antagonistic surround has proven more elusive. Studies that do identify a center-surround profile in the linear receptive field often rely on models that assume spatial and temporal filtering are independent. This assumption, referred to as space-time separability, has been questioned previously but not tested in this context.
Ganglion cell action potentials were recorded using a multielectrode array and single units were isolated. Binary white noise checkerboards were optically reduced to 50 micrometers per square and presented sequentially at 15Hz. Spike triggered averages were calculated by standard methods, and further data analysis was performed in Matlab (Mathworks, Natick MA). We used regional signal averaging relative to the receptive field center to circumvent common assumptions and improve signal-to-noise ratios.
An antagonistic surround was observed in 754 of 805 mouse GCs across 16 retinas. Importantly, these RFs were frequently codependent on space and time, prescribing against the general assumption of space-time separability when modeling GC linear activity. We studied the nature of this inseparability, and discovered the overall RF can be decomposed into five spatiotemporally distinct subfilters. Each subfilter was individually space-time separable, but their relative strengths differed across space-time allowing them to account for the overall linear RF's inseparability. This led us to propose the sum of separable subfilters (SoSS) model, which successfully accounted for the variance in our data. By leveraging this model, we were able to identify local deviations from a Gaussian profile in the receptive field surround.
The results presented here provide a new and more general model for ganglion cell linear receptive fields that links the spatial and temporal structure in a way that accounts for their observed co-dependencies. In the process we identify new functional components of GC linear receptive fields with distinct spatiotemporal organization, shedding light on the underlying synaptic circuitry and providing tools for its continued study.
This PDF is available to Subscribers Only