Mouse visual cortex contains a region of enhanced spatial resolution

Abstract

The representation of space in mouse visual cortex was thought to be relatively uniform. Here we reveal, using population receptive-field (pRF) mapping techniques, that mouse visual cortex contains a region in which pRFs are considerably smaller. This region, the "focea," represents a location in space in front of, and slightly above, the mouse. Using two-photon imaging we show that the smaller pRFs are due to lower scatter of receptive-fields at the focea and an over-representation of binocular regions of space. We show that receptive-fields of single-neurons in areas LM and AL are smaller at the focea and that mice have improved visual resolution in this region of space. Furthermore, freely moving mice make compensatory eye-movements to hold this region in front of them. Our results indicate that mice have spatial biases in their visual processing, a finding that has important implications for the use of the mouse model of vision.

Fig. 1. Wide-field calcium imaging reveals a cortical region with small pRF size in mice.

a Calcium signals were imaged through the cleared skull of Thy1-GCaMP6f mice viewing checkerboard bars of different orientations and positions. b The change in fluorescence in response to 31 different bar stimuli from an example pixel (black bars). The predictions of the pRF model are shown as red dots. Pearson’s correlation between the model and the data of the example pixel was 0.99 (p ≤ 0.001, H0: r = 0, two-tailed). c Example cortical maps showing the correlation of the pRF model (left panel), the azimuth of the best-fitting Gaussians (middle panel), and the elevation of the Gaussians (right panel) overlain on the brain imaged through the skull. The maps are thresholded at a correlation coefficient of 0.75. d Maps of pRF size (the full-width at half-maximum of the best-fitting Gaussian). A region of smaller pRF size was observed in left and right visual cortex in all imaged mice. e Average azimuth, elevation, and pRF size maps from 17 mice (only the left hemisphere was imaged in these mice to examine visual area boundaries at higher resolution). The boundaries of V1 and other visual areas were identified using field-sign analysis and are overlaid on the maps as black lines. The maps of individual mice were recentered on V1 and resized by the size of V1 before averaging. The region of small pRF size is centered on the lateral border of V1 and extends into higher lateral visual areas LM and RL. f The relationship between pRF position and size. Azimuth and elevation values from individual pixels were binned. The red/blue lines show data from the right/left hemispheres of 11 mice who were imaged bilaterally. The black line shows the average across mice. g The size of pRFs can be visualized as a 3D surface and is approximately linearly related to the distance from a point in space at 0° azimuth and 20°elevation that we refer to as the focea. The surface was fit to the azimuth and elevation values using linear interpolation. h An example from a single hemisphere showing that pRF size is linearly related to the spherical angle between the pRF center and the focea, which we refer to as r-eccentricity. Linear regression H0: β = 0 (two-tailed). i The distribution of the slope-coefficients, β, across all 22 hemispheres of bilaterally imaged mice, the average slope is shown by the dashed red line. j The average map of cortical magnification factor (CMF) across 17 unilaterally imaged mice. k The relationship between CMF and pRF location in bilaterally imaged mice (22 hemispheres imaged mice). CMF was highest at azimuths close to the vertical meridian (left panel), and increased in regions of azimuth below 30° (i.e., binocular regions). The relationship between CMF and elevation was less clear and variable across animals (right panel).

Fig. 2. Electrophysiological analysis of RF size.

a A total of 28 awake Thy1-5.17-GCaMP mice viewed sparse-noise RF mapping stimuli while multi-unit neural activity was measured across the different layers of V1 using a linear probe. b Four example RFs showing the average change in spiking in response to each check position. The white circle denotes the FWHM of the best-fitting Gaussian. c The azimuth (left panel) and elevation (middle panel) of the RFs across all animals. The binned average (bin size = 5°) is shown as the black line. RF size was relatively constant across different azimuths/elevations with the exception of one penetration at a more negative elevation where RF sizes were larger. Right panel: the relationship between r-eccentricity (angle between the RF center and the focea in a spherical coordinate system) and RF size showed a weak, but significant positive relationship (Linear regression. H0: β = 0 (two-tailed)). Error bars indicate SEM. d The slope of the relationship between r-eccentricity and RF size was significant in the individual layers. The slope did not differ significantly across laminar compartments (ANCOVA, F_2,1786 = 3, p = 0.05).

Fig. 3. Conceptual models of pRF size.

Upper row: two conceptual models of V1 that could account for the influence of r-eccentricity on pRF size in the wide-field data. The top left panel shows the position of model cell bodies in V1. The other panels show models in which RF positions are displaced purely by changes in cortical magnification factor (CMF), with lower magnification at higher eccentricities (middle) or by increased RF scatter at larger eccentricities (right). The black lines connected to each RF illustrate the displacement of the RF. Lower row: two equally sized analysis windows were drawn on the cortex, one at the foceal representation (red) and one in the periphery (blue). The RFs of cells within the analysis window are shown colored in the right panels. An estimate of the pRF can be made by taking the convex hull (shaded region) of the RF positions in space. Both the CMF and scattering model result in smaller pRFs at the focea (compare the areas of the red and blue regions).

Fig. 4. The region of small pRF size is due to higher cortical magnification and decreased RF scatter.

a Tiled two-photon images from an example mouse covering almost the entirety of V1. The mouse viewed a screen placed at an angle of 30° so that the left visual field could be mapped with sparse noise. Cells for which we could reliably measure the RF (r² > 0.33, BVI < 1, see Methods) are shown in color according to their preferred azimuth (left), elevation (middle), and RF size (right). The mean image of cortex is shown in the background. b An example relationship between the azimuth of the RF and the distance of the cell body from the foceal representation. The red line shows the fit of an exponential function. The cortical magnification factor (in mm/deg) can be estimated by the slope of this fit. c CMF estimates in the azimuth and elevation directions. The black line is the average across three mice. d RF scatter was estimated by examining the residuals of the RF positions from the exponential fit. The solid red lines indicate the interquartile range of the residuals and the dashed line the mean residual value in 10° sliding windows. r-eccentricity is the spherical angle between the RF center and the focea. e (Left panel) An example linear regression of the interquartile range of the residuals on r-eccentricity. The shaded region shows ± SEM (right panel). The slopes were significantly positive in all three mice (bootstrap test, one-tailed) indicating increased scatter of the residuals with distance from the focea in visual space. **p < 0.01. f Two example pRFs constructed from the single-cell data. For every cell falling within an analysis window (400 µm radius in this example), the Gaussian RF fit was projected into visual space (gray circles). The convex-hull of the resulting region and its area were used to estimate the pRF and its size. n, number of cells contributing to the pRF. g Example linear regression of pRF size on r-eccentricity in M2. h The slope values (β-coefficients) from three mice as a function of window radius. Asterisks, slopes significantly greater than zero (t-test, p < 0.05, two-tailed), error bars indicate 1 SEM. The slopes determined from individual cell RFs (without computation of aggregate RFs) are shown as square symbols. i There was no significant relationship between r-eccentricity and the number of cells in the analysis window (p > 0.05, linear regression, two-tailed).

Fig. 5. pRFs generated from raw two-photon images show stronger scattering.

a Example maps of azimuth and elevation generated from the (smoothed) raw two-photon images show a clear retinotopic organization in agreement with the RFs measured for the individual cells. The white border indicates the boundary of V1 as determined by field-sign analysis (Methods). The map of RF size (right panel) showed no clear organization at this level of spatial detail. These images form the input into the scatter analysis. b Measures of pRF size obtained from performing the scatter analysis on the retinotopic maps shown in (a), a window size of 400 μm was used to generate this image. The smallest pRFs are in the region representing the focea. c The slope of the regression of pRF size on r-eccentricity for the raw image data approached the values from the wide-field data in Fig. 1i for the larger analysis windows. Example regression fits for mouse M2 are shown in the insets. Asterisks mark significant values, p < 0.05, t-test, two-tailed, and the error bars indicate 1 SEM. d The intercept term of the regression gives the expected pRF size at the focea. This approached the minimum values observed in the wide-field data (approximately 40–50°) only at window sizes of 200–400 μm radius, suggesting that windows of this size best capture the signals that are measured in the wide-field data. The pattern was consistent across mice (colored lines). Error bars indicate 1 SEM. e Summary of the slope (left) and intercept (values) for the different techniques. The results indicate that the small pRFs in the focea are caused by reduced the scatter of RFs across cells. Techniques that measured individual cells or small multi-units (electrophysiology, two-photon cell analysis) did not find strong relationships, whereas those that measured activity pooled over many cells (two-photon scatter analyses and wide-field analysis) found a relationship. The data from the cell+neuropil analysis (i.e., raw) had slope values closest to the wide-field data. The values for the scatter analyses are taken from the 400 μm radius analysis windows. Error bars indicate 1 SEM across animals.

Fig. 6. Receptive fields in three higher visual areas are larger at greater eccentricities.

a If neurons at all retinotopic positions in higher visual areas sample homogeneously from equal-sized regions of V1 (left panel), then neurons sampling from the foceal representation (red cells) will have smaller receptive fields (red cross-hatched region) than cells sampling from more scattered representations in the periphery (blue cells). Alternatively, neurons at different retinotopic positions in higher areas may use different V1 sampling strategies (right panel), which could counteract the reduced scatter at the focea to equalize RF size across eccentricities (compare the red vs blue cross-hatched region). b The relationship between r-eccentricity and single-cell RF size in LM (n = 959 cells) and RL (n = 505 cells) measured using two-photon imaging. c RF size increased with eccentricity in LM and AL. Asterisks indicate regression slopes that were significantly greater than zero: **p < 0.01, ***p < 0.001 (Linear regression, H0: β = 0, two-tailed).

Fig. 7. Visual acuity is higher at the focea.

a Contrast detection task. Mice were trained to lick upon detection of a grating stimulus presented at one of 6 different spatial locations (“go” trials). The locations were grouped into three conditions indicated by the colored squares (not visible to the mouse): focea (red), lateral (yellow), and inferior (blue). On no-go trials no stimulus was presented and mice were trained to refrain from licking. False alarms were punished with timeouts. b Fraction of correct “go” trials for different spatial frequencies of the grating for three different spatial locations for an example mouse. “Focea” indicates stimuli presented in the upper-central visual field at (0° azimuth, 20° elevation), “Lateral” shows responses pooled over the four stimuli at lateral locations (±35° azimuth, −10°/+20° elevation), and “Inferior” for stimuli presented in the lower-central visual field at (0° azimuth, −10° elevation). The vertical line indicates the spatial frequency threshold, estimated by fitting a logistic function. c Spatial frequency thresholds for four mice at the three locations (data from the four lateral locations were similar and is pooled). Spatial frequency thresholds were significantly higher for the focea than the inferior location for all four mice (all p < 0.01, likelihood ratio test, two-tailed).

Fig. 8. Compensatory eye movements in freely moving mice keep the focea ahead of the animal.

a Tracking of head tilt (pitch and roll) and left and right eye positions in a freely moving mouse (inset). Illustration of head pitch and roll axes (relative to the ground), eye torsion (white arrow), and pupil centers (white dots) in angular eye coordinates (blue and red arrows). b The spherical coordinate system used in this study with [azi = 0°, ele = 0°] pointing toward the animal’s nose. The arrows indicate the foceal projection of the right (purple) and left (green) eye. Distribution of angular foceal projections in the same reference frame for an example mouse during spontaneous locomotion in an open field environment (right). Light and dark shading indicate regions of low and high probability, respectively. Dots indicate circular median position of the focea for the left (green) and right (purple) eye. Dashed lines indicate monocular visual fields for the left and right eyes (same color schema; 180° visual collection angle for each eye) showing that the focea fell into the binocular zone. c Thirty-second segment showing elevation of left/right foceas (top), horizontal, and vertical eye position of left and right eye (middle) and head pitch and roll (bottom) for the example mouse in (b). Same conventions as in (a). Control condition (“Focea fixed in head”) in top panel shows what the elevation of the focea would have been in the absence of stabilizing eye movements. d Average circular mean elevation (top) and standard deviation (SD, bottom) of the left/right foceas for the same mice either head-fixed or in three different head-free contexts (open field, social interaction, object tracking). In all three contexts, eye movements counteracted head movements to stabilize the elevation of the focea relative to the ground. Same color schema as in panels a–c. Dark gray bars show means for the focea and light gray bars show means for the control condition (“Focea fixed in head” in panels a–c). Means computed across eyes and mice. Data from four mice. Mean ± SEM. e Optical flow field during locomotion (body speed >10 cm/s) for the left (top) and right (bottom) eye. Optical flow vectors were computed for discrete grid points (spacing 10°) centered at the focea using head and eye positions and the geometry of the environment. Black arrows show average flow vectors across four mice. Green/purple shaded circles illustrate pRF size at the focea. Gray circles, pRF size at an azimuth of 50°.