Linear and nonlinear chromatic integration in the mouse retina

Mohammad H. Khani, Tim Gollisch

TL;DR

This paper investigates how mouse retinal ganglion cells integrate color (chromatic) information, revealing that this process is often nonlinear, especially in the ventral retina. Unlike spatial integration, nonlinear chromatic integration in "On" cells leads to response suppression, not activation, and appears crucial for detecting chromatic boundaries like skylines.

What Problem Does This Paper Address?

While temporal and spatial integration in the retina have been extensively studied, chromatic integration—how different color signals are combined—remains largely unexplored. Understanding this is crucial because photoreceptors separate light into different color channels, and downstream neurons must integrate these signals to form a coherent visual percept. This paper aims to characterize the linearity and nonlinearity of chromatic integration in mouse retinal ganglion cells, analogous to how spatial integration was studied in the past.

Methods

The researchers developed a novel "chromatic-integration stimulus" that presents opposing contrasts of UV and green light. By systematically varying these contrasts and observing the ganglion cell responses, they could identify a "balance point" where the signals from the two color channels cancel each other out. If the cell's response is suppressed or enhanced at this balance point, it indicates nonlinear integration. They also used:

  • Multielectrode arrays: To record from thousands of retinal ganglion cells simultaneously.
  • Opsin-isolating stimuli: To specifically target S-cones (UV-sensitive) and M-cones/rods (green-sensitive) without co-activating other photoreceptors.
  • Receptive field mapping: Using spatiotemporal white noise to characterize the spatial extent of each cell's receptive field.
  • Pharmacological manipulations: Using blockers for GABAergic and glycinergic inhibition to probe the role of inhibitory circuits.
  • Light level variations: To investigate the contribution of the rod pathway.
  • Local vs. full-field stimulation: To differentiate between center and surround contributions to chromatic integration.
  • Natural scene modeling: To explore the functional consequences of nonlinear chromatic integration for detecting chromatic boundaries.

Datasets

The study used natural images from existing UV/Green Image Databases for their modeling experiments. No new benchmark datasets were introduced.

Key Insights and Ideas

  • Nonlinear chromatic integration is common: A significant portion of retinal ganglion cells exhibit nonlinear chromatic integration, meaning their responses are not simply additive combinations of color inputs.
  • "On" cells suppress, "Off" cells activate: A striking difference emerged: nonlinear "On" cells suppress their firing at the chromatic balance point, while nonlinear "Off" and "On-Off" cells show increased firing. This "suppression" response in "On" cells is a novel finding.
  • Independence from spatial integration: Nonlinearity in chromatic integration is largely independent of nonlinearity in spatial integration, suggesting distinct underlying mechanisms.
  • Surround inhibition is key: Nonlinear chromatic integration is strongly influenced by signals from the receptive field surround, particularly inhibitory inputs. Restricting stimulation to the receptive field center linearized responses.
  • Rod pathway involvement: The rod pathway plays a significant role in nonlinear chromatic integration, especially in the ventral retina where M-cones are scarce.
  • Detecting chromatic boundaries: Nonlinear chromatic integration may be important for detecting chromatic boundaries, such as the skyline, where UV and green light contrast can oppose each other.

Key Results

  • Approximately one-third of recorded cells showed nonlinear chromatic integration based on their criteria.
  • Nonlinear "On" cells consistently showed response suppression at the balance point, while nonlinear "Off" and "On-Off" cells showed activation.
  • Pharmacological blockade of GABAergic and glycinergic inhibition significantly linearized chromatic integration in many nonlinear cells.
  • Nonlinear chromatic integration was significantly reduced when stimulation was confined to the receptive field center, highlighting the role of the surround.
  • Nonlinear cells were predominantly found in the ventral retina, an area with higher UV sensitivity.
  • Modeling natural scenes suggested that nonlinear chromatic integration could help detect chromatic boundaries like skylines.

Limitations and Future Work

  • Cell-type specificity: While differences were observed between "On," "Off," and "On-Off" cells, a more detailed cell-type-specific analysis is challenging due to the limitations of multielectrode array recordings in definitively identifying all cell types. Future work could explore this further, potentially by combining electrophysiology with genetic markers or advanced imaging.
  • Surround contribution: The exact circuit mechanisms by which the inhibitory surround contributes to nonlinear chromatic integration, especially the excitatory component in "Off" cells, require further investigation.
  • Natural scene relevance: While modeling suggests a role in detecting chromatic boundaries, further experimental validation of this functional consequence in vivo is needed.
  • Contrast range: The limited range of contrast values tested might not have fully activated the surround in all cases, potentially masking some nonlinearities.
  • Model simplicity: The phenomenological model used for natural scenes captures some aspects of nonlinear integration but doesn't specify the exact circuit mechanisms. Future work could explore more detailed circuit models.

Summary updated

Abstract

The computations performed by a neural circuit depend on how it integrates its input signals into an output of its own. In the retina, ganglion cells integrate visual information over time, space, and chromatic channels. Unlike the former two, chromatic integration is largely unexplored. Analogous to classical studies of spatial integration, we here study chromatic integration in mouse retina by identifying chromatic stimuli for which activation from the green or UV color channel is maximally balanced by deactivation through the other color channel. This reveals nonlinear chromatic integration in subsets of On, Off, and On-Off ganglion cells. Unlike the latter two, nonlinear On cells display response suppression rather than activation under balanced chromatic stimulation. Furthermore, nonlinear chromatic integration occurs independently of nonlinear spatial integration, depends on contributions from the rod pathway and on surround inhibition, and may provide information about chromatic boundaries, such as the skyline in natural scenes.
BibTeX
@inproceedings{khani2021linear,
  author = {Mohammad H. Khani and Tim Gollisch},
  title = {{Linear and nonlinear chromatic integration in the mouse retina}},
  booktitle = {Nature Communications},
  year = {2021},
  doi = {10.1038/s41467-021-22042-1},
}
Figure