Periodic Reporting for period 1 - FeaGatSu (Feature-gating in superior colliculus)
Reporting period: 2018-06-01 to 2020-05-31
Mice, like humans, use vision to interact with their environment, e.g. a sudden, expanding shadow above their head, mimicking an approaching predator, triggers innate freezing or escape behavior. Mice also rely on vision when hunting. These visually-guided innate behaviors are mediated by the activation of neural circuits going through the superior colliculus (Fig. 1). Wide-field neurons of the superior colliculus are known to be involved in innate fear and hunting behaviors and respond to stimuli mimicking an approaching or passing-by predator or prey. They receive direct input from the retina and integrate the visual information along their massive dendritic trees.
We investigated how wide-field neurons of the mouse superior colliculus combine signals from the retina to detect salient features that are important for the animal’s survival in three steps. First, we characterized the visual features wide-field neurons respond to. Second, we identified the visual features relayed by the retinal ganglion cells to wide-field neurons. Third, we determined the computational rules that allow wide-field neurons to effectively extract visual features that represent prey or predator from the retinal signals. This was achieved by labeling and recording from wide-field neurons and their retinal inputs using a combination of genetic tracing and two-photon microscopy of a fluorescent indicator that lights up when cells respond to a stimulus.
We found that 10 types of retinal ganglion cells provide input to wide-field neurons, with each anatomical type (Fig. 2, left) providing a unique signal in response to a visual stimulus (Fig. 2, right). These responses are combined along the dendrites of wide-field neurons. We found that signals in the dendritic tips can be represented by the sum of the retinal inputs. However, at the soma, such a simple summation of the retinal inputs fails to represent the output of wide-field neurons (Fig. 3). Further investigation of this complex signal integration will advance our understanding of signal processing in central neurons and eventually aid intervention in neural processing disorders.
We investigated how wide-field neurons of the mouse superior colliculus combine signals from the retina to detect salient features that are important for the animal’s survival in three steps. First, we characterized the visual features wide-field neurons respond to. Second, we identified the visual features relayed by the retinal ganglion cells to wide-field neurons. Third, we determined the computational rules that allow wide-field neurons to effectively extract visual features that represent prey or predator from the retinal signals. This was achieved by labeling and recording from wide-field neurons and their retinal inputs using a combination of genetic tracing and two-photon microscopy of a fluorescent indicator that lights up when cells respond to a stimulus.
We found that 10 types of retinal ganglion cells provide input to wide-field neurons, with each anatomical type (Fig. 2, left) providing a unique signal in response to a visual stimulus (Fig. 2, right). These responses are combined along the dendrites of wide-field neurons. We found that signals in the dendritic tips can be represented by the sum of the retinal inputs. However, at the soma, such a simple summation of the retinal inputs fails to represent the output of wide-field neurons (Fig. 3). Further investigation of this complex signal integration will advance our understanding of signal processing in central neurons and eventually aid intervention in neural processing disorders.
First, we characterized the responses of wide-field neurons to a comparative set of visual stimuli. We found that wide-field neurons strongly respond to dark and bright moving edges but not to ambient brightness changes. Responses of wide-field neurons to dark expanding discs at different locations and expansion speeds, did not depend on the specific location but rather on the size of the stimulus. This invariance to stimulus location is typical for higher visual brain areas but was not observed in the retina or the superficial superior colliculus before. Further, we found, that neurons looking at a region above the horizon, were more selective to dark expanding discs than other stimuli. Neurons looking at the bottom responded to a more diverse set of stimuli. The increased selectivity to dark expanding discs in the upper visual field could provide a better detection of approaching aerial predators, important for the animal’s survival.
Second, we characterized the responses of retinal ganglion cell types that target wide-field neurons to the same set of visual stimuli used for wide-field neurons. A recent publication from the lab could identify 10 different types based on anatomical cues and molecular markers. We found that responses within an anatomical type were very similar to each other whereas responses between types were very diverse (Fig. 2).
Third, we were able to predict some of the responses of wide-field neurons by summing the retinal inputs (Fig. 3), however, for many stimuli, there was a large discrepancy in magnitude and timing. By measuring the inputs to wide-field neurons at different depths of the dendritic tree, we found that summing the retinal responses was successful in explaining the responses in the dendritic tips, but progressively failed closer to the soma. This hints to a more complex processing in wide-field neurons than previously thought.
The results of this project were communicated to the expert audience at several international neuroscience conferences, as the FENS meeting 2018 (Berlin, Germany) and the European Visual Cortex Meeting 2019 (Leiden, The Netherlands). An invited talk at the European Retina Meeting 2019 (Helsinki, Finland) increased the visibility of the project further. Thus far, there is one open-access scientific publication associated to the project. Three more manuscripts for disseminating results of the project are in preparation, one publication on the processing of retinal signals in wide-field neuron dendrites, one on the distribution of retinal inputs to superior colliculus across visual space and one on how biases in the responses of wide-field neurons across visual space relate to behavior. The project activities are communicated to the general public via press releases, social media outlets, and on the project webpage.
Second, we characterized the responses of retinal ganglion cell types that target wide-field neurons to the same set of visual stimuli used for wide-field neurons. A recent publication from the lab could identify 10 different types based on anatomical cues and molecular markers. We found that responses within an anatomical type were very similar to each other whereas responses between types were very diverse (Fig. 2).
Third, we were able to predict some of the responses of wide-field neurons by summing the retinal inputs (Fig. 3), however, for many stimuli, there was a large discrepancy in magnitude and timing. By measuring the inputs to wide-field neurons at different depths of the dendritic tree, we found that summing the retinal responses was successful in explaining the responses in the dendritic tips, but progressively failed closer to the soma. This hints to a more complex processing in wide-field neurons than previously thought.
The results of this project were communicated to the expert audience at several international neuroscience conferences, as the FENS meeting 2018 (Berlin, Germany) and the European Visual Cortex Meeting 2019 (Leiden, The Netherlands). An invited talk at the European Retina Meeting 2019 (Helsinki, Finland) increased the visibility of the project further. Thus far, there is one open-access scientific publication associated to the project. Three more manuscripts for disseminating results of the project are in preparation, one publication on the processing of retinal signals in wide-field neuron dendrites, one on the distribution of retinal inputs to superior colliculus across visual space and one on how biases in the responses of wide-field neurons across visual space relate to behavior. The project activities are communicated to the general public via press releases, social media outlets, and on the project webpage.
This project advanced our understanding of early visual processing in three different ways. First, we found that central neurons combine the outputs of the retina in more complex ways than previously thought, such that a simple summation of the retinal signals is not enough to describe their responses (Fig. 3). Wide-field neurons of the superficial superior colliculus showed many response characteristics that were thought to be reserved to neurons in higher visual areas, as the visual cortex and the deep superior colliculus. In particular, the responses of wide-field neurons to moving edges did not depend on the specific location but rather on the size of the stimulus. This has been previously described in neurons of the lateral pulvinar, a higher visual area targeted by wide-field neurons. Our findings suggest that this complex processing already happens in the superior colliculus and information is then passed on to higher order processing centers.
Second, in the retina, anatomy and molecular signature of a neuron are a strong indicator of a neuron’s visual response. However, consistent datasets combining anatomy and visual responses are scarce. Here, we collected responses from 10 anatomically identified retinal ganglion cell types to a set of parametric and ecologically relevant stimuli. This unique dataset constitutes an important building block for reconstructing higher brain function from traced and anatomically identified retinal ganglion cells and will give mechanistic insights into how the outputs of the retina are used in the brain.
Third, we furthered the understanding of how visual information extracted by the retina is locally distributed in the brain. We found that the signals coming from the retina, as well as the responses of wide-field neurons in the superior colliculus, show regional specializations. These specializations could explain why animals respond differently to a passing-by object above their head compared to when it is shown on the ground and hints to a regional adaption of the visual system to the ecological needs of an animal.
These advances in understanding sensory processing in the healthy brain could aid in intervening with neural processing disorders as attention deficit hyperactivity disorder where the processing of stimuli that catch our attention is perturbed. Further, our results can inspire hardware solutions for early warning systems in cars to fast and efficiently detect potential hazards.
Second, in the retina, anatomy and molecular signature of a neuron are a strong indicator of a neuron’s visual response. However, consistent datasets combining anatomy and visual responses are scarce. Here, we collected responses from 10 anatomically identified retinal ganglion cell types to a set of parametric and ecologically relevant stimuli. This unique dataset constitutes an important building block for reconstructing higher brain function from traced and anatomically identified retinal ganglion cells and will give mechanistic insights into how the outputs of the retina are used in the brain.
Third, we furthered the understanding of how visual information extracted by the retina is locally distributed in the brain. We found that the signals coming from the retina, as well as the responses of wide-field neurons in the superior colliculus, show regional specializations. These specializations could explain why animals respond differently to a passing-by object above their head compared to when it is shown on the ground and hints to a regional adaption of the visual system to the ecological needs of an animal.
These advances in understanding sensory processing in the healthy brain could aid in intervening with neural processing disorders as attention deficit hyperactivity disorder where the processing of stimuli that catch our attention is perturbed. Further, our results can inspire hardware solutions for early warning systems in cars to fast and efficiently detect potential hazards.