Action and learning shape the activity of neuronal circuits in the visual cortex

Abstract

Nonsensory variables strongly influence neuronal activity in the adult mouse primary visual cortex. Neuronal responses to visual stimuli are modulated by behavioural state, such as arousal and motor activity, and are shaped by experience. This dynamic process leads to neural representations in the visual cortex that reflect stimulus familiarity, expectations of reward and object location, and mismatch between self-motion and visual-flow. The recent development of genetic tools and recording techniques in awake behaving mice has enabled the investigation of the circuit mechanisms underlying state-dependent and experience-dependent neuronal representations in primary visual cortex. These neuronal circuits involve neuromodulatory, top-down cortico-cortical and thalamocortical pathways. The functions of nonsensory signals at this early stage of visual information processing are now beginning to be unravelled.

Figure 1

Experimental procedures for recording neuronal activity and correlated behavioural parameters in head-fixed awake behaving rodents. (a) Schematic of a cranial window preparation above primary visual cortex (V1) that allows chronic two-photon imaging of neurons labelled with a genetically encoded calcium indicator (GCaMP6). (b) GCaMP6-labelled neurons (green) can be imaged over multiple days and even weeks (example field of view from imaging Day 1 and 9). Relative changes in fluorescence (ΔF/F₀) over time are used as a proxy read-out of neuronal activity. Signals from genetically defined subpopulations of cells can be isolated via a second fluorescent marker (e.g. tdTomato expression in somatostatin expressing inhibitory interneurons, shown here in red). (c) Head-fixed rodents can freely move on an air-supported styrofoam ball that acts as a spherical treadmill while calcium imaging and/or electrophysiological recordings are performed. Optical computer mice are used to assess running speed. Pupil diameter and eye-tracking can be recorded with cameras. In this configuration, animals can navigate an open virtual reality environment or view a visual stimulus where left/right behavioural choices can be made with motor movements. (d) Neuronal activity can be correlated with running speed and changes in pupil diameter, used as a measure of arousal. GCaMP6 signal from a V1 VIP expressing interneuron is shown in green and animal's speed in black. As traditionally done in anaesthetised animals, neuronal activity in V1 can be correlated with visual stimulation in the form of passively viewed stimuli, such as drifting gratings displayed on a screen. Bottom panel shows single trials (grey) and average response (black) of a GCaMP6-labelled neuron to different oriented gratings. Polar plot shows the amplitude of calcium transients in response to each orientation, normalised to the maximum response. (e) Head-fixed rodents can be placed in a virtual environment: animals run as if on a linear treadmill and, with surrounding screens, can navigate virtual corridors with defined wall patterns. Note that the spherical treadmill can be replaced by a cylindrical wheel where an optical encoder attached to the central axle records speed. (f) Using a virtual reality environment, the visual flow experienced by the animal can be measured and manipulated. An animal navigating along a virtual corridor creates visual flow: the experimenter can manipulate this coupling to create a mismatch between the visual flow and the animal's movement. (g) Experience-dependent neuronal changes in V1 can be studied using head-fixed animals learning visually guided tasks. Water deprived animals learn to lick during key experimental cues and goal-directed behaviours in order to receive water rewards through a spout. (h) Licking behaviour is monitored by a lick sensor on the spout during task performance. Example of a task in which the animal must lick at a certain reward location demarcated by a visual cue (oriented grating) along a virtual track. Each dot represents a lick: prereward licking (black dots), rewarded-licking at the right visual cue (blue dots) and postreward licks (grey dots). With learning, licking behaviour becomes tightly coupled to the location of the rewarded visual cue along the track.

Figure 2

Neuronal activity in V1 is shaped by behavioural-state and experience-dependent processes, mediated through the integration of nonvisual inputs. (a) Schematic representing the increased gain in neuronal responses to oriented gratings during locomotion (orange trace) versus stationary periods (black trace). Both additive (as illustrated) and multiplicative gain modulations were reported in V1 excitatory neurons [22^•]. Illustration is based on Ref. [10]. (b) Schematic of representations of reward timing in V1. After the learning of a task associating a visual cue (light bulb) and a reward (blue drop), neuronal responses in V1 predict the timing of reward events by sustaining either an increase or decrease in activity after a visual stimulus onset (grey dotted line) or peaking at the expected reward time (blue dotted line). Illustration is based on Refs. [37, 38]. (c) Schematic of V1 responses during the learning of a visually guided task. Example of a go/no-go task with two oriented gratings and only one grating is rewarded (blue drop). A schematic of the responses of a single neuron to the two presented stimuli show that neuronal discriminability between the rewarded (orange) and nonrewarded (grey) stimuli increases with task learning. On the population level (bottom panel), a higher proportion of neurons in V1 show increased selectivity to task-relevant gratings after learning (purple). Illustrations are based on Ref. [43]. (d) Schematic of responses to spatial expectation of visual stimuli in V1. Top panel: schematic of a paradigm where animals are presented with a sequence of visual cues along a virtual track. Traces illustrate neuronal responses to visual cues, before (black) and after (orange) repeated exposure to the same sequence. A population of V1 neurons show specific responses to a given visual stimulus (e.g. vertical grating) but also specific responses for a given stimulus at a particular spatial location (response to vertical grating at B2 location larger than B1). With experience, a population of neurons develop predictive responses, shifting the onset of their response to before the appearance of their preferred stimulus (orange trace). Bottom panel illustrates the effect of omitting an expected stimulus in a trained sequence. On the population level, when the stimulus is present there is an evoked response to the stimulus (black trace), but when the stimulus is un-expectantly omitted (orange trace), there is a large and delayed increase in activity. A subpopulation of neurons respond selectively to these omission events, and not to the initially expected stimulus. Illustration is based on Ref. [28^•]. (e) Schematic of the major cortico-cortical inputs to V1, including top-down influences from higher visual areas (V2), the retrosplenial cortex (RSC) and secondary motor regions (A24b/M2) as well as inputs from other sensory modalities such as the primary auditory cortex (A1) and the somatosensory cortex (SS). Inputs from higher visual areas (V2) include connections from lateral, medial and mediolateral secondary visual areas. (f) Schematic of neuromodulatory and thalamocortical inputs to V1 that have been shown to influence V1 activity in awake behaving mice. LP, Lateral posterior nucleus; dLGN, dorsal lateral geniculate nucleus; MLR, mesencephalic locomotor region.