Dendritic mechanisms underlying retinal direction selectivity

The ability to sense motion, both self-motion and motion in the environment, is crucial for animal behavior. In the mammalian visual system, motion selectivity begins in the retina, where nonselective light-evoked signals are converted to the direction-selective responses of a class of ganglion cells. These direction-selective ganglion cells (DSGCs) enable higher-order computations in the brain that are important for behavior in response to motion in the environment. DSGCs, each of which fires action potentials preferentially in response to motion in one of the four cardinal directions, owe their selectivity to direction-selective inhibition from non-spiking interneurons called starburst amacrine cells (SACs). SACs exhibit radial direction selectivity: segments of the dendritic tree depolarize preferentially during motion outward from the cell body to their distal tips. Recent experiments have clarified how radial direction selectivity in SACs is translated to linear direction selectivity in DSGCs. However, the mechanisms that underlie and modulate radial direction selectivity in SACs are unclear. One property that is thought to be crucial for SAC direction selectivity is electrotonic isolation between distinct sectors of the SAC dendritic tree. Here we use calcium imaging from SAC dendrites to show that SAC sectors are not completely isolated during moving visual stimulation of the retina, and that a degree of signal integration between SAC sectors improves the response to outward motion. We further show that reducing electrotonic isolation via pharmacological blockade of metabotropic glutamate receptor 2 (mGluR2) disrupts direction selectivity of SACs and DSGCs by enabling aberrant propagation of depolarization between SAC sectors. Thus, mGluR2 signaling controls direction selectivity by modulating the degree of trans-sector propagation of depolarization. Using pharmacology and patch-clamp recording, we show that mGluR2 signaling occurs through modulation of P/Q- and N-type voltage-gated calcium channels, with no detected effect on voltage-gated potassium channels. Furthermore, using current-clamp recordings from the somata of SACs along with pharmacological and genetic manipulations, we test the effect of mGluR2 signaling, voltage-gated potassium channels, and synaptic inhibition of SACs through GABAA receptors on electrotonic isolation. We find that blockade of mGluR2 signaling improves the propagation of depolarization from distal dendrites to the soma, while blockade of Kv channels prolongs the somatic depolarization in response to motion restricted to a single dendritic sector. Because the somatic depolarization in response to these spatially restricted visual stimuli is indicative of the fidelity of trans-sector propagation, these experiments suggest distinct roles for mGluR2 signaling and Kv channels in controlling the spread of depolarization across a SAC. By contrast, conditional knockout of GABAA receptors on SACs did not affect direction selectivity measured at the soma, suggesting that inhibition by other amacrine cells does not affect direction selectivity of SACs as measured at the soma. Together, our data illustrate mechanisms by which propagation of depolarization along dendrites may be modulated by cell-intrinsic, as well as activity-dependent, cell-extrinsic factors. Because restricted propagation along dendrites is crucial for the function of SACs in direction selectivity, our results directly link molecular mechanisms to the function of motion detection in the mammalian visual system