Visual system plasticity
We study the cellular and circuit mechanisms that enable the visual cortex to adapt to changes in the visual environment and to take part in the storage of visual memories. To address these questions at the functional and at the structural level, we use a number of imaging techniques, such as two-photon microscopy and intrinsic signal imaging.
In mouse visual cortex, plasticity can be readily induced by closing one eye for a couple of days. This intervention, termed monocular deprivation, shifts the balance between the representation of the two eyes in the visual cortex, such that inputs from the deprived eye are weakened, while open eye inputs gain influence. In order to assess how individual cells in the visual cortex change their responses after monocular deprivation, we used two-photon imaging of neurons loaded with a calcium indicator dye. While many neurons shifted their eye preference towards the non-deprived eye (see Figure 1), a small number of cells resisted this overall trend, and, instead, responded more vigorously to the deprived eye. We hypothesize that the change in response properties of these cells is driven by a homeostatic mechanism, which increases the cell’s responsiveness following a drop in afferent drive. Mrsic-Flogel et al., Neuron 2007.
Which mechanisms might underlie this homeostatic plasticity? To address this question, we removed all sensory input to the visual cortex by permanently lesioning the retina in both eyes and repeatedly imaged cortical neurons in awake, behaving mice. In fact, after an initial drop, neuronal activity gradually increased again over the following two days (see Figure 2). In parallel to the restoration of cortical activity, two-photon imaging revealed an increase in the size of dendritic spines – sites of synaptic contacts on cortical neurons – suggesting that a general increase in synaptic strength contributes to homeostatic plasticity in the visual cortex. Keck et al., Neuron 2013.
When the eye’s retina is lesioned only partially, another aspect of visual cortex plasticity becomes apparent: The corresponding, initially silenced cortical region regains responsiveness for visual stimuli in the weeks and months following the retinal lesion (see Figure 3, top panels). This functional reorganization is paralleled by a massive restructuring of cortical circuitry, as we could demonstrate by repeatedly imaging synaptic structures of excitatory neurons in the visual cortex (see Figure 3, bottom panels). The rate at which dendritic spines disappeared and reappeared was strongly increased following lesioning of the retina, suggesting that the restructuring of excitatory neuronal connections within the visual cortex allows for functional reorganization after a loss of sensory inputs. Keck et al., Nature Neuroscience 2008.
So far we have focused on the structural plasticity of excitatory neurons. Do inhibitory neurons take part in the reorganization process, too? We addressed this question by monitoring the fine structure of inhibitory cells after lesioning the retina. Shortly after the lesion, both, synaptic input and output structures were reduced in number (see Figure 4), indicating an overall lowered level of inhibition after removal of the sensory input. We believe that it is this rapid decrease in inhibition which sets the stage for the subsequent changes in excitatory connectivity that form the basis of the functional reorganization observed in the study above. Keck et al., Neuron 2011.
These studies illustrate the visual cortex’ capacity to alter its structure and function in response to (partial) removal of the sensory input. Does the visual environment also play an instructive role in shaping the stimulus selectivity of neurons in the visual cortex? We reasoned that the best way to answer this question is interfering with the most prominent response property of cells in the visual cortex, orientation selectivity. We raised mice for several weeks with special goggles, limiting the animals’ visual experience to contours of only one orientation. Two-photon calcium imaging revealed that this orientation was overrepresented among neurons in the visual cortex (see Figure 5), and that at least some cells must have changed their stimulus selectivity during the altered experience. Thus, neurons in the visual cortex adapt their tuning to the specific statistics of the visual environment. Kreile et al., Journal of Neuroscience 2011.
In the studies described above, adaptive changes in the visual cortex were induced by passive exposure of the mice to an altered visual environment. We next asked how responses in the visual cortex change when certain stimuli become behaviorally meaningful to the animal. To this end we train mice to distinguish between two visual stimuli, for example two gratings of different orientations, only one of which is associated with a food reward. We then monitor changes in stimulus selectivity of individual neurons over time (see Figure 6). These experiments show, for example, that in mice that rapidly master the task, a higher proportion of neurons in the visual cortex become selective for stimulus orientation during training. Thus, learning of stimulus-reward associations is reflected in changes in neuronal response properties, even at the level of the primary visual cortex.
We are currently expanding this experimental approach to more complex learning tasks, which require mice to discriminate between categories of simple visual objects. In order to search for neuronal correlates of such category learning, we are turning our attention to higher areas of the mouse visual cortex.
The primary visual cortex is frequently characterized as a purely sensory area, acting as the first cortical stage of a feed-forward processing hierarchy. More and more evidence suggests, however, that behavioral state and other variables can modulate the responses of neurons in visual cortex. In order to assess the functional significance of this modulation, we imaged neuronal activity in awake mice during locomotion in a virtual visual environment. Using a combination of open and closed loop configurations, we found that locomotion alone can strongly drive responses in the primary visual cortex, even in the absence of any visual input (see Figure 7 and movie to the right). Moreover, mismatch between actual and expected visual feedback elicited strong signals in many neurons. These data suggest that processing in visual cortex may be based on predictive coding strategies that use motor-related and visual input to detect mismatches between predicted and actual visual feedback. Keller et al., Neuron 2012.