CORDIS - EU research results

Decision-Making from Visual Cortex to Midbrain

Final Report Summary - VISION AND DECISION (Decision-Making from Visual Cortex to Midbrain)

Every day, we make a myriad of movements executing decisions that are based on visual information. We want to understand how these decisions are formed in the brain, and how information is processed along the hierarchy of visual areas and in superior colliculus (SC), a motor structure of the midbrain. Cortical visual areas progressively encode more complex stimulus features. But signals that reflect the percept in addition to the stimulus are already present at early stages of visual processing. Activity in SC controls eye and head movements, but might also have more abstract roles in decision-making reflecting task difficulty and performance.
The neural circuit underlying visual decision processes is not yet understood. To unravel these circuits requires that the role of various brain areas of the mouse in decision-making is solidly understood. To that end, we trained mice to perform visual decision tasks of different complexity and recorded the activity of large neuronal populations in cortical visual areas and SC. Based on these recordings, we can quantify how strongly single neurons and the neuronal population are driven by visual stimuli in contrast to the mouse’s decisions and performance. To probe the involvement of different visual areas in the task, we quantify the mouse’s performance while inactivating those areas using optogenetic techniques.
The SC of the adult mouse is partly covered by the cortex and partly by a large blood vessel. Studying the SC therefore poses several challenges methodologically, and we decided to focus our attention on this area first. During the first months of this project, we established a surgical protocol allowing us to image part of the SC by permanently displacing the blood vessel while leaving the cortex intact. Several problems had to be overcome, such as developing the necessary surgical skill necessary, developing a suitable implant that holds the vessel in place while allowing imaging at the same time, and reaching a high level of mechanical stability over time periods of months. Using this method, we are able to record neural activity of large volumes in the superficial layers of the SC.
The superficial layers of the SC receive direct input from the retina and from visual cortex and are mainly concerned with the processing of visual information. To understand the function of the SC in vision, we measured in detail the stimulus properties, such as location, size, orientation, and direction of movement, that elicit responses in the neurons, termed receptive field (RF). Our results will contribute to resolving a current debate on the existence of orientation maps in the SC of the mouse (Ahmadlou & Heimel, 2015; Feinberg & Meister, 2014; Inayat et al., 2015), i.e. whether neighbouring neurons prefer similar orientations, a feature that is found in visual cortex of higher mammals but not in visual cortex of the mouse. Furthermore, we are collaborating with Dr. Samuel Solomon, a Senior Lecturer at UCL, to study surround suppression in the SC, i.e. how visual responses of neurons are modulated by visual stimuli outside their RF. The data from all these experiments led to the identification of a new cell type, which is suppressed by the visual stimulus. Using genetic markers, we could show that this behaviour is almost exclusively exhibited by inhibitory neurons.
During the presentation of the various visual stimuli, the mice in our experiments were free to run on a floating ball. This paradigm led to a further important insight: the behaviour of the animal, i.e. the speed of running, and the state of its alertness (measured by pupil dilation) modulate the activity of a large portion of neurons in the superficial SC. About one half of these neurons increase their activity during running and states of high alertness, whereas the other half decreases activity. Preferred orientations of the neurons are constant across the two states. These results have been presented at the Annual Meeting of the Society for Neuroscience (SfN) in October 2015 and at the UCL symposium in June 2016. We are currently assessing how behavioural state affects the processing of visual stimuli in SC neurons.
A similar modulation by running and alertness has already been established for neural activity in the visual cortex of the mouse. The modulation we observe in the SC could therefore be inherited from V1. Another possibility is that due to changes in alertness, and thus due to changes in pupil diameter, activity in the retina changes because different amounts of light enter the eye. This could successively be reflected in SC activity. We are currently following two approaches to test these inheritance hypotheses: (1) We will image the inputs coming from visual cortex or retina within the SC to investigate whether they show the same modulations as neurons in the SC. To this end, we have developed a new virus that infects retinal ganglion cells and expresses the calcium dependent fluorescence in the synapses of these cells. Within the next months, we will image retinal activity in the SC. (2) We have recorded activity in the SC while simultaneously inactivating visual cortex in order to test whether the missing input from visual cortex extinguishes modulation through behaviour and alertness in SC.
The results of the previously described experiments, i.e. tuning properties of SC neurons, the modulation of activity by behaviour, and measuring the input to the SC, are crucial steps towards understanding the role of SC in decision tasks that involve vision.
To study the role of SC in visual decision tasks directly, we have developed several decision tasks of varying complexity (detecting the location of a stimulus, and discriminating stimuli of different contrast). We have successfully trained mice in these tasks and have imaged neural activity in SC during performance of the detection task and are about to record neural activity during the performance of the more difficult discrimination task. These data will reveal whether superficial SC represents decision signals in addition to purely visual signals. The data that were collected in primary visual cortex (by Dr. Christopher Burgess) showed that neural activity in this area only reflects stimulus features but does not represent any decision signals.
The training of the mice on these decision tasks typically takes several weeks before a good performance is reached. To boost the efficiency of the training procedure, we established a method to optogenetically stimulate dopaminergic neurons in the ventral tegmental area, which are involved in the signalling of rewarding stimuli. With the help of another member of the lab (Dr. Amin Lak), we successfully used this method to significantly increase the number of trials the animals perform per session, which effectively shortens the time to reach good performance and leads to larger and therefore more meaningful datasets showing the relationship between task performance and neural activity. These results have been presented on a poster at the Annual Meeting of SfN 2015.
At last, we proposed to determine the necessity of various visual areas for the performance of the decision task by optogenetically inhibiting them when the animal is performing the task. This part of the project was performed by two members of the lab (Dr. Nicholas Steinmetz and Peter Zatka-Haas). The results show that inactivation of any region within visual cortex leads to a marked decrease in task performance. Inactivation acts in the same way as eliminating or reducing the perception of visual stimuli processed by the inactivated brain region. A very similar procedure can be applied to inhibit the SC, thus probing its role in the decision task.
In summary, all proposed objectives have been implemented and will be finalized in the near future. The timeline of the objectives has changed as the focus was set on the SC first, instead of cortical areas, and in the course of studying the SC new scientific interests and questions developed.

Ahmadlou, M., & Heimel, J. A. (2015). Preference for concentric orientations in the mouse superior colliculus. Nature Communications, 6, 6773. doi:10.1038/ncomms7773
Feinberg, E. H., & Meister, M. (2014). Orientation columns in the mouse superior colliculus. Nature. doi:10.1038/nature14103
Inayat, S., Barchini, J., Chen, H., Feng, L., Liu, X., & Cang, J. (2015). Neurons in the Most Superficial Lamina of the Mouse Superior Colliculus Are Highly Selective for Stimulus Direction. Journal of Neuroscience, 35(20), 7992–8003. doi:10.1523/JNEUROSCI.0173-15.2015