Skip to main content

Circuits of Visual Attention

Periodic Reporting for period 3 - Daphne (Circuits of Visual Attention)

Reporting period: 2020-12-01 to 2022-05-31

Due to the limited neuronal resources of any nervous system, extracting relevant sensory information from cluttered natural environments is critical to allocating computational power correctly. In this proposal, we explore the neuronal mechanisms used by the nervous system to attend visual cues and, thus, enable appropriate behaviors. To do so, we study the visual transformations of the mouse Superior Colliculus (SC), an evolutionarily conserved midbrain area known to process sensorimotor transformations and to be involved in the allocation of attention. By understanding the principles underlying visual and sensorimotor transformation, our work will contribute to building a framework to study attention in health and disease. The overall objectives of our ERC project are to (i) provide a detailed description of visual representation in the SC, focusing on understanding how defined retinal information streams, like motion and color, contribute to these properties; (ii) understand the relationship between motor instructions and sensory coding; and (iii) explore how higher brain areas can modulate sensory transformations within the SC.
I. Collective neuronal dynamics of the SC

As reported previously, we developed a paradigm that allows us to image collicular activity in awake behaving animals immersed in a virtual reality arena. While using this system, we unexpectedly observed that neural population activity switches between periods dominated by spatially localized “ordered” hotspots and globally distributed “disordered” patterns (Fig. 1). These switches occur in the presence and absence of visual stimuli. Given that perception results from a dynamic interplay between feed-forward processing of sensory stimuli and spontaneous neural activity. These spontaneous dynamics can be presumed to reflect prior expectations, task demands, and attentional focus. We have established operant training paradigms to study these dynamics during discrimination tasks, which strengthens the idea that these dynamics are a dynamical representation of attention, work currently being finalized.
Fig. 1. Overview of intrinsic dynamics.

II. vLGN/IGL modulation of visual processing.

It is indispensable to clearly describe the inputs the SC integrates to understand how visual computations emerge. One poorly described input is the vLGN/IGL complex, a retino-recipient area that has avoided deep scrutiny for decades. We have found, contrary to what was suggested previously, that some vLGN/IGL projection is highly specific to the superficial retino-recipient layers (Fig. 2A-B), and are strongly inhibitory influence collicular dynamics ex vivo (Fig. 2C-D) and in vivo. Finally, we could show, using awake-behaving electrophysiology and Calcium imaging of axonal buttons, that information relayed from the vLGN to the superficial SC carries is determined by animals’ locomotion speed and forward translational optic flow. Thus, vLGN inhibitions seems to act as a context-specific gain control of visual signals, allowing the animals to suppress predicted increases of visual dynamics due to self-motion.
Fig. 2. vLGN/IGL projections to the superficial SC strongly modulate SC activity

III. Panoramic view of retinal processing

We developed a novel system to image large regions of retinal explants (Gupta et al., 2020). Our method allows simultaneous imagining from up to 1.6 mm2, roughly 20 times larger field of view than done in the past, for a tenth of the price of current multi-photon approaches. This system allowed us to explore the functional inhomogeneities in retinal receptive field structures. Interestingly, our results match our theoretical predictive coding framework, indicating that the eye’s computations adapted to the natural statistics of the visual environment (Fig. 3).
Fig. 3 Experiments and theoretical prediction from natural scene statistics

IV. Visuomotor deficits in models of ASD.
Individuals with autism spectrum disorder (ASD) exhibit cognitive difficulties together with impairments in sensory processing. These conditions are believed to arise during cortical neurodevelopment; however, a comprehensive description of neuronal mechanisms liking sensation with behavior remains unclear. Here we show that a subcortical mechanism of sensory processing required to initiate efficient threat responses are disrupted in models of ASD. Although mutant animals can quickly detect visual threats, on average, they require longer to evaluate and respond to them compared to their WT siblings (Fig. 4). We could determine that the dorsal periaqueductal gray (dPAG), a key node in initiating this response has a hypo-excitability phenotype due mis-regulation of Kv1.1 channel, suggesting a disruption in homeostatic mechanisms. Overall, our results provide a mechanistic model of cognitive dysfunction from gene to behavior (MS in submission).
Fig. 4. Loom avoidance in ASD model and WT mice.
We have established several independent but complementary approaches to understand sensory processing in the SC. Within these projects, we have shown to be capable of recording large-scale population dynamics in extended fields of view in retinas and in the SC, allowing us to differentially assess the contributions of (i) genetically-defined cell types, their projections, (ii) population dynamics, and (iii) their relationship with animal behavior. In the retina, we have finalized a project that merges theoretical and experimental approaches to show how natural scene statistics sculpture receptive field structures (MS being revised). SC also gets input from other brain areas; however, their role remains largely unexplored. We have been focusing on the feed-forward pathway from the vLGN to the SC, where we found a more specific projection than previously reported. We have shown that this feed-forward di-synaptic inhibition appears to relay contextual information, improving the animals’ ability to “see” while moving around (MS in preparation). In a different project, we are now exploring different aspects of collicular population dynamics. By doing so, we recently identified a new type of salient, population-level phenomenon in the SC, which is intriguing by itself and calls for a detailed description. Due to the novel dynamics, we have been collaborating closely with a theory group at IST Austria (Gasper Tkacik) to develop new tools and approaches to properly describe it. In addition, we have implemented operant conditioning experiments that allow us to combine imaging experiments with behavioral attention readouts. These operant condition experiments a point that the observed dynamics are directly related to an attentional process (project is being finalized). Finally, since the SC is a brain region known to be involved in the control of attention and goal-directed behaviors such as eye movements, and that these characteristics are key to diagnose attention deficit disorders, we explored the possibility that visually-driven behaviors might also be affected by specific linked to ASD. We found clear, reproducible, and robust behavioral changes that are in line with our starting hypothesis and have dissected the pathway, showing that subcortical circuit have an important role in ASD.
Overview of intrinsic dynamics
Loom avoidance in ASD model and WT mice
vLGN/IGL projections to the superficial SC strongly modulate SC activity
Experiments and theoretical prediction from natural scene statistics