Abstract
Purpose :
Direction-selective ganglion cells (DSGCs) respond with robust spiking when objects move through their receptive fields in a ‘preferred’ direction, but spike minimally when they move in the opposite or ‘null’ direction. Underlying these responses are excitatory (E) and inhibitory (I) synaptic inputs, the former being non-directional, and the latter being highly directional. How exactly DSGCs integrate these opposing signals is poorly understood. Recent studies suggest that the dendrites of DSGCs are active –i.e. can generate action potentials. This endows dendritic arbors with the capability of integrating E/I on a sub-cellular scale. The increased computational power this bestows, however, comes at a cost: it requires E/I to be balanced across space and time, as even temporarily unopposed excitation could initiate spiking activity. Here, using a computational modeling approach, we test how E/I balance across space and time impacts direction coding by DSGCs.
Methods :
Based on known morphological and biophysical properties, a DSGC model was built in NEURON. E/I synaptic inputs were driven by a simulated moving light bar, which activated synapses probabilistically in different regions as it swept across the dendritic tree. Directionality was simulated by increasing release probability of inhibitory synapses and by modulating the temporal offset between E/I. Temporal and spatial balance are modelled by the degree to which event timing and success are correlated within each synapse across the DSGC’s dendritic arbour.
Results :
As expected, the average macroscopic E/I currents measured at the DSGC’s soma were minimally affected by the degree to which E/I was correlated in space-time across the DSGCs dendritic arbours. Interestingly, we found the degree of correlations dramatically impacted the directional tuning properties of the model DSGC. The direction selectivity index (DSi; calculated from a vector sum of responses evoked in 8 directions) more than doubled, while the variability of the computed preferred angle halved, between uncorrelated and well-correlated states. Additionally, we found that the effects of space and time correlations on spiking are synergistic, as their separate DSi impacts (1.4x and 1.2x) underestimated the full effect (2.3x).
Conclusions :
A tight balance of excitation and inhibition in space and time significantly improves direction coding by DSGCs.
This is an abstract that was submitted for the 2018 ARVO Annual Meeting, held in Honolulu, Hawaii, April 29 - May 3, 2018.