Spatial integration of visual signals in ganglion cells of the mouse retina
Final Report Abstract
Processing of visual information begins in the retina, a layered neural network at the back of the eye. Traditionally, the retina had been thought of as providing spatiotemporal filtering of the incoming stimuli for efficient representation of visual information in the transmission along the optic nerve. In recent years, however, it has become increasingly clear that many retinal pathways provide specific visual functions, such as the detection of different types of motion, and that the implementation of these functions relies on nonlinear integration of visual signals by the retinal network. In this project, we analyzed the characteristics of stimulus integration (in particular integration of visual signals over space) in the mouse retina. To do so, we extended the approach of measuring iso-response stimuli to recordings from retinal ganglion cells in isolated mouse retina. In this approach, the goal is to identify such combinations of visual contrast at different locations inside the receptive field of a ganglion cell so that the cell’s response is always at a fixed, predefined level. The distribution of these contrast combinations inside the space of all contrast combinations then reveals whether the ganglion cell integrates these contrast contributions linearly or what type of nonlinear transformations take place. We developed methods for measuring iso-response curves both in single-cell recordings, using automated measurements of receptive fields and adaptive tuning of the required contrast levels, as well as in multielectrode-array recordings where many ganglion cells can be investigated simultaneously. The results showed that ganglion cells of the mouse display a considerable variety of nonlinear integration types, which may thus help differentiate different types of ganglion cells according to their functional characteristics. In particular, we found that there is a systematic difference between On and Off ganglion cells, with the former showing more pronounced nonlinearities of spatial integration. This is particularly interesting in the light of previous findings indicating that Off cells rather than On cells have stronger nonlinearities when considering how changes in visual contrast are represented by the cells’ activity. For one particular type of ganglion cells, so-called transient Off alpha cells, the identified nonlinear spatial integration characteristics led to a computational model of how these cells respond to saccade-like transitions of images. This model provided an explanation of the curious observation that these cells respond to a newly fixated images particularly strongly when the spatial pattern inside the receptive field is identical to the pattern prior to the transition. Finally, we extended our approach of identifying nonlinearities in stimulus integration in two ways. First, we developed a technique for identifying the spatial layout of nonlinear signal integration by detecting spatial subunits inside receptive field centers through computational analyses of responses to finely structured stimulation. Second, we generalized the analysis of signal integration to the spectral domain, that is, we investigated how different wavelengths of light are integrated by the retinal network. Specifically, we tested whether signals emerging from different types of cone photoreceptors, with different spectral sensitivities, are combined linearly or nonlinearly by retinal ganglion cells. Here, we found that the vast majority of mouse ganglion cells integrate linearly over signals from different cone types; yet a small set of specific ganglion cells showed distinct nonlinear integration. Together, the results of this project demonstrate how nonlinearities of stimulus integration over space or other dimensions can be systematically studied in electrophysiological recordings and how the identification of specific nonlinearities can be used to study the structure and function of neural circuits.
Publications
- (2017) Inference of neuronal functional circuitry with spike-triggered non-negative matrix factorization. Nature communications 8 (1) 149
Liu, Jian K.; Schreyer, Helene M.; Onken, Arno; Rozenblit, Fernando; Khani, Mohammad H.; Krishnamoorthy, Vidhyasankar; Panzeri, Stefano; Gollisch, Tim
(See online at https://doi.org/10.1038/s41467-017-00156-9) - Nonlinear spatial integration in the receptive field surround of retinal ganglion cells. Journal of Neuroscience 34:7548-7561 (2014)
Takeshita D, Gollisch T
(See online at https://doi.org/10.1523/JNEUROSCI.0413-14.2014) - Spike-triggered covariance analysis reveals phenomenological diversity of contrast adaptation in the retina. PLoS Computational Biology 11:e1004425 (2015)
Liu JK, Gollisch T
(See online at https://doi.org/10.1371/journal.pcbi.1004425) - Joint encoding of object motion and motion direction in the salamander retina. Journal of Neuroscience 36:12203-12216 (2016)
Kühn N, Gollisch T
(See online at https://doi.org/10.1523/JNEUROSCI.1971-16.2016) - Using matrix and tensor factorization for the single-trial analysis of population spike trains. PLoS Computational Biology 12:e1005189 (2016)
Onken A, Liu JK, Karunasekara PPCR, Delis I, Gollisch T, Panzeri S
(See online at https://doi.org/10.1371/journal.pcbi.1005189) - Sensitivity to image recurrence across eyemovement-like image transitions through local serial inhibition in the retina. eLife 6:e22431 (2017)
Krishnamoorthy V, Weick M, Gollisch T
(See online at https://doi.org/10.7554/eLife.22431)
