Mechanisms of Intensity Encoding for Non-Image-Forming Visual Functions

Abstract

Light drives many physiological processes that exist outside conscious awareness. These ‘non-image-forming’ visual functions are typically sensitive to the overall environmental light intensity—a parameter which can vary over 8-10 orders of magnitude across the solar day. The majority of the retinal input to these functions comes from a specialized group of photoreceptors, the intrinsically photosensitive retinal ganglion cells (ipRGCs). At the onset of this study, it was unclear how ipRGC output was related to light intensity across the range over which these behaviors are sensitive. My work suggests that the M1 ipRGCs divide their labor to cover a broader dynamic range as a population, and that they have the capacity to collectively encode light conditions in an unexpectedly rich way. I found that the spike output of these ipRGCs is tuned to light intensity. Compared to the overall range of the population, individual ipRGCs were sensitive to narrower intensity ranges, with wide variation in their thresholds. Most were unimodally tuned, with sufficiently bright light suppressing firing. Furthermore, tuning curves shifted position on the irradiance axis depending on light history. A cell’s dark-adapted sensitivity could be used to predict with some certainty whether it would shift its tuning to higher or lower intensities following light adaptation. Perforated-patch recording revealed the cell-autonomous mechanism behind unimodal tuning; in unimodal M1s, the largest steady depolarizations produced by the intrinsic photocurrent silenced action potential firing through inactivation of voltage-gated channels (depolarization block). Decreasing light intensity reversed block and produced rebound firing. The use of depolarization block as a tuning mechanism raised the question of whether signaling fidelity was affected during the approach to block. By recording from the axon and soma of ipRGCs simultaneously, I showed that spike propagation under these conditions was highly reliable—additionally, ambiguous spikelets viewed in somatic recordings typically regenerated into uniform spikes during axonal propagation. Thus, the M1s can use depolarization block as a tuning mechanism without any apparent detriment. These results are followed by an extended discussion of their implications for mechanisms of non-image vision, as well as some important caveats and nuances to the data.