Visual response properties of neurons in the superficial layers of the superior colliculus of awake mouse

Abstract

Key points: In rodents, including mice, the superior colliculus is the major target of the retina, but its visual response is not well characterized. In the present study, extracellular recordings from single nerve cells in the superficial layers of the superior colliculus were made in awake, head-restrained mice, and their responses to visual stimuli were measured. It was found that these neurons show brisk, highly sensitive and short latency visual responses, a preference for black over white stimuli, and diverse responses to moving patterns. At least five broad classes can be defined by variation in functional properties among units. The results of the present study demonstrate that eye movements have a measurable impact on visual responses in awake animals and show how they may be mitigated in analyses.

Abstract: The mouse is an increasingly important animal model of visual function in health and disease. In mice, most retinal signals are routed through the superficial layers of the midbrain superior colliculus, and it is well established that much of the visual behaviour of mice relies on activity in the superior colliculus. The functional organization of visual signals in the mouse superior colliculus is, however, not well established in awake animals. We therefore made extracellular recordings from the superficial layers of the superior colliculus in awake mice, while the animals were viewing visual stimuli including flashed spots and drifting gratings. We find that neurons in the superficial layers of the superior colliculus of awake mouse generally show short latency, brisk responses. Receptive fields are usually 'ON-OFF' with a preference for black stimuli, and are weakly non-linear in response to gratings and other forms of luminance modulation. Population responses to drifting gratings are highly contrast sensitive, with a robust response to spatial frequencies above 0.3 cycles degree^-1 and temporal frequencies above 15 Hz. The receptive fields are also often speed-tuned or direction-selective. Analysis of the response across multiple stimulus dimensions reveals at least five functionally distinct groups of units. We also find that eye movements affect measurements of receptive field properties in awake animals, and show how these may be mitigated in analyses. Qualitatively similar responses were obtained in urethane-anaesthetized animals, although receptive fields in awake animals had higher contrast sensitivity, shorter visual latency and a stronger response to high temporal frequencies.

Keywords: anaesthesia; functional properties; receptive field; tectum; vision.

Figure 1. Visual responses of a representative unit recorded in superficial layers of the superior colliculus in awake mouse

Each graph shows the mean response of the unit to one of the tested dimensions, and the best‐fitting prediction of the relevant model. All responses were obtained from a single unit in a single session (4‐160705‐6). The error bars are ±1 SEM over trials and dashed horizontal lines show the mean spontaneous activity measured from interleaved presentations of a grey screen. A, receptive field maps for flashed white or black squares (‘sparse noise’). Black circle shows 1 SD of best‐fitting Gaussian function. B, size‐tuning for flashed white (open symbols) or black (closed symbols) discs. The lines show the best‐fitting difference‐of‐Gaussians function in each case (solid: white; dashed: black). C, direction tuning for drifting sinusoidal gratings. Line shows the best fitting modified von Mises function. D, spatial frequency tuning (at 3.8 Hz). Line shows the best fitting difference‐of‐Gaussians function. Arrows show preferred frequency (left) and the spatial resolution (fc, right). E, temporal frequency tuning (at 0.04 cycles degree⁻¹). Line shows the best fitting difference‐of‐exponentials function. Arrow shows preferred temporal frequency. F, contrast response function. Line the shows best fitting modified Naka‐Rushton function. Arrow shows contrast at half the fitted maximum response (C ₅₀). G, visual field locations of units recorded from awake animals (n = 144). Azimuth of 0° represents directly in front of the animal; elevation is relative to eye height.

Figure 2. Responses to flashed white and black stimuli in awake animals

Aa, response of a unit to white and black squares (0.2 s in duration, size 15°) presented against a grey background. The peristimulus time histogram (PSTH) at each of 81 spatial positions (separation 7.5°) is shown. Ab, PSTHs of the same unit as in (Aa), to white (W) or black (B) circular discs (0.5 s in duration) of logarithmically increasing size and centred on the receptive field. Only the first 0.2 s of the response are shown. Ba and Bb, response of another example unit. Conventions as in (Aa) and (Ab), Scale bars in (Bb) apply to (A) and (B). C, scatterplot showing the relative strength of responses to white and black stimuli, obtained from sparse noise (n = 219). The response ratio is (Rw – Rb)/(Rw + Rb), where Rw is the maximum response to a white stimulus and Rb is the maximum response to a black stimulus. A value of −1 (or 1) indicates a response only to black (or white) stimuli, and a value of 0 indicates an equal response to both. Here and subsequently, uniformly random offsets have been applied to the x‐axis to allow better visibility. The central mark in the boxplot indicates the median, and the edges indicate the 25th and 75th percentiles. The whiskers are 1.5 times the interquartile distance (99% coverage). D, receptive field radii (at 1 SD) obtained from best‐fitting Gaussian functions (as in Fig. 1 A) to white (W, n = 46) or black (B, n = 110) responses. Here and subsequently, model predictions are only shown for units where the normalized log‐likelihood of the model (see Methods) was at least 0.5. E, latency of visual responses from onset of white (W, n = 102) or black (B, n = 180) flashes in the preferred location, for all responsive units. [Color figure can be viewed at wileyonlinelibrary.com]

Figure 3. Size‐tuning for flashed white and black stimuli in awake animals

A, population size‐tuning for a flashed white disc (duration 0.5 s) centred on the receptive field. Each row of the image shows the best‐fitting difference‐of‐Gaussians function obtained for a single unit (as in Fig. 1 B), normalized to its maximum response. Only units in which the normalized log‐likelihood of the model was at least 0.5 are shown. The units are ordered, from bottom‐to‐top, by the preferred size. Here and elsewhere, the colour bar applies in all cases. B, mean size‐tuning for a white disc, obtained by averaging across the rows in (A). Dashed lines show ±1 SEM. Dashed horizontal line shows the spontaneous rate, normalized to the unit's maximum visual response before averaging. C and D, same as (A) and (B) but for black discs. E, radius of the preferred white (W) or black (B) stimulus. F, percentage reduction in response from a stimulus of optimal size, to the largest tested (‘suppression index’). G and H, radius of the centre (G) and surround (H) components of the receptive field in the difference‐of‐Gaussians function. [Color figure can be viewed at wileyonlinelibrary.com]

Figure 4. Linearity of spatial summation

A and B, raster plots showing responses of a linear (A) and a non‐linear (B) unit to a stationary grating, which flickered (counterphase modulated) at 2 Hz. The schematic at top shows the temporal profile of the grating. Each row shows the spiking activity of the unit on one trial. The tick marks on the y‐axis separate groups of trials of the same spatial phase (also indicated by the colour of the raster). A, response of a unit showing linear spatial summation. Responses are modulated at the temporal frequency of stimulation. B, response of a non‐linear unit. Responses at all spatial phases are modulated at twice the temporal frequency of stimulation. C–E, amplitude of mean (F0) and modulated (F1) response as a function of the spatial frequency of a drifting grating. Horizontal lines show the F0 (dashed) and F1 (solid) for spontaneous activity. C, same unit as in (A). D; same unit as in (B). E, another unit. F, non‐linearity index (F2/F1 ratio) calculated from counterphase modulated gratings, in awake animals (n = 48). Values greater than 1 indicate non‐linear responses; values less than 1 indicate linear responses. G, linearity index (F1/F0 ratio), as calculated from drifting gratings of spatial frequency optimal for the F0, in awake animals (n = 227). Values larger than 1 indicate linear responses; values less than 1 indicate non‐linear responses. H, comparison of spatial frequency resolution for F0 and F1 responses in awake animals (n = 191). Values substantially above the unity line (dashed line) indicate the F0 resolves higher spatial frequencies than the F1. I–K, same as (F) to (H) but for units recorded in anaesthetized animals (I, n = 37; J, n = 97; K, n = 93). [Color figure can be viewed at wileyonlinelibrary.com]

Figure 5. Impact of eye movements in awake animals

A, response of a single unit to drifting gratings of varying spatial frequency (temporal frequency 4 Hz). Aa, each row shows the spiking activity of the unit on one trial. The tick marks on the y‐axis separate groups of trials of the same spatial frequency. Colours determined by average eye position in the trial (shown in Ac). Ab, cycle‐averaged PSTHs of each trial at 0.08 cycles degree⁻¹, arranged by the horizontal (‘X’) eye position on that trial, with the most positive estimates at the top. The colours help indicate the horizontal eye position in each trial (shown in Ac). The schematic above shows one cycle of a sinusoid for comparison. Ac, top: image of the eye obtained during the recording session. Bottom: average pupil position on each trial, in degrees of visual angle, relative to the average of all eye positions across all trials. Larger symbols indicate trials at 0.08 cycles degree⁻¹. Ad, spatial frequency tuning for the F1 response obtained by subjecting trial averaged PSTHs to Fourier analysis (‘F1psth’; closed symbols) or applying Fourier analysis to individual trials and averaging their amplitudes (‘F1’; open symbols). Ba, comparison of spatial frequency resolution for the F1 response obtained by analysing individual trials (y‐axis) or trial‐averaged PSTHs (x‐axis). Points above the line (i.e. F1/F1psth ratios greater than 1) indicate that trial averaging reduced the estimate of spatial resolution, consistent with presence of eye movements. Shown are both non‐linear (n = 53) and linear (n = 73) units, as defined by the F1/F0 ratio. Bb, impact of trial averaging (F1/F1psth) on estimates of spatial resolution, for linear units only. Units with small receptive fields are most affected. C, relationship between variability in eye position and impact of trial averaging, for linear units only. Measurements of resolution in sessions with low SD of eye position (i.e. few eye movements) are less affected by trial averaging. [Color figure can be viewed at wileyonlinelibrary.com]

Figure 6. Population responses to drifting gratings in awake animals

Aa, contrast response functions. Each row shows the predictions of a modified Naka‐Rushton function for a single unit. The units are ordered, from bottom‐to‐top, by the contrast at half‐maximal response (semi‐saturation constant, C ₅₀). Conventions as in Fig. 3 A. Ab, mean contrast response, obtained by averaging across the rows in (Aa). Conventions as in Fig. 3 B. Ba, orientation/direction‐tuning functions. Each row shows the predictions of a modified von Mises function for a single unit. The units are ordered by their preferred direction within each of three subgroups: neither direction, nor orientation selective (a; gOSI and gDSI < 0.1), orientation but not direction selective (b; gOSI > 0.1 and gDSI < 0.1) and remaining units (c; gDSI > 0.1). Bb, mean direction tuning, after aligning responses to the preferred direction. C–D, spatial and temporal frequency tuning. Each row in (Ca), shows the predictions of a difference‐of‐Gaussians function for a single unit and each row in (Cb), shows the predictions of a difference‐of‐exponentials function. The units are ordered by their preferred spatial or temporal frequency. Cb, mean spatial frequency response, obtained by averaging across the rows in (Ca). Db, mean temporal frequency response, obtained by averaging across the rows in (Da). [Color figure can be viewed at wileyonlinelibrary.com]

Figure 7. Parametric descriptions of responses to drifting gratings in awake and anaesthetized animals

Units recorded in awake animals (AW) are shown to the left in orange, and those for anaesthetized animals (AN) are shown to the right in blue. A, parameters estimated from contrast response function fits to the tuning curves. Aa, semi‐saturation constant (C ₅₀, AW: 71 units; AN: 70 units). Ab, exponent for the expansive non‐linearity in the same units. B, orientation/direction‐tuning functions. Ba, direction‐selectivity (DSI) and global direction selectivity indices (gDSI) (AW: 151 units; AN: 86 units). Bb, orientation‐selectivity (OSI) and global orientation selectivity (gOSI) indices for the same units. Bc, bandwidth of a modified von Mises function fit to the tuning curves (AW: 135 units; AN: 81 units). C, parameters estimated from spatial frequency tuning fits (AW: 210 units; AN: 98 units) for linear (L) and non‐linear (N) units. Ca, preferred spatial frequency. Cb, radius of the receptive field centre. Cc, percentage reduction in response from preferred spatial frequency to modulation of a uniform field. D, parameters estimated from temporal frequency tuning fits (AW: 146 units; AN: 95 units). Da, preferred temporal frequency. Db, excitatory time constant (s). Dc, percentage reduction in response from preferred temporal frequency to 0.5 Hz. [Color figure can be viewed at wileyonlinelibrary.com]

Figure 8. Tuning for visual speed

A, three representative units from awake animals showing responses to a matrix of spatial and temporal frequencies. Red ellipses indicate the best fitting models described in the text. Ba, average spatio‐temporal tuning for units from awake animals (n = 56). Ellipses obtained as in (A) are semi‐transparent and overlaid. Bb, same as (Ba) but for units from anaesthetized animals (n = 62). Ca, relationship between preferred speed and preferred spatial frequency in awake animals. Cb, same as (Ca) but for units from anaesthetized animals. Da, population temporal frequency tuning functions in awake animals, for drifting gratings of low spatial frequency (0.01 cycles degree⁻¹; n = 80). Db, same but for mid‐range spatial frequencies (0.04 cycles degree⁻¹; n = 78). Dc, same but for high spatial frequencies (0.15 cycles degree⁻¹; n = 75). E, same as D but for anaesthetized animals (n = 60, 63 and 50, respectively). Fa, comparison of preferred temporal frequency at mid‐range spatial frequency with that at low (open symbols, n = 68) or high (filled symbols, n = 73) spatial frequency, for units from awake animals. Fb, same as (Fa) but for units from anaesthetized animals (n = 50 and 60, respectively). G, distribution of speed tuning index in awake (n = 56) and anaesthetized (n = 62) animals. The speed tuning index is effectively the slope of the ellipses shown in (A), where 0 indicates that the ellipse is parallel to the temporal frequency axis, positive values indicate ellipses similar to that in (Aa) (0.70) and (Ab) (0.45) and negative values indicate ellipses similar to that in (Ac) (−0.38). [Color figure can be viewed at wileyonlinelibrary.com]

Figure 9. Functional subclasses of units in the sSC of awake mouse

The boxplots show the tuning of five functional groupings (Groups A–E, plotted from left to right) that were identified by fuzzy k‐means clustering performed on measurements from 227 sSC units along 10 dimensions of response variation (see Results). The number of observations in each boxplot can vary. The white–black response ratio, index of sustained response, F1/F0 ratio and global direction selectivity indices were omitted to improve clarity. The letters and filled boxplots in some plots indicate dimension(s) of response that most distinguished each group. A, response latency for flashed black stimuli. B, global orientation selectivity index (gOSI) for drifting gratings. C, excitatory time constant (Ca) and temporal frequency tuning index (Cb) for the response to drifting gratings. D, receptive field centre size (Da) and spatial frequency tuning index (Db) for drifting gratings. E, comparison of spatial (as in Db) and temporal (as in Cb) tuning indices. F, average tuning curves for each of the groups. Left–right: response to flashed black spots (0.2 s in duration); response to drifting gratings of varying orientation and motion direction after aligning responses to the preferred direction; response to drifting gratings of varying spatial frequency; response to drifting gratings of varying temporal frequency. Response to flashed spots was simply averaged across units. Other responses were normalized to the unit's maximum response before averaging. A horizontal dashed line indicates the relative level of spontaneous activity. Other dashed lines show ±1 SEM across units. [Color figure can be viewed at wileyonlinelibrary.com]

Figure 10. Size‐tuning for white and black stimuli in anaesthetized animals

Same conventions as Fig. 3. [Color figure can be viewed at wileyonlinelibrary.com]

Figure 11. Population responses to drifting gratings in anaesthetized animals

Same conventions as Fig. 6. Grey lines in (Ab) to (Db), replot the average population responses from awake animals in Fig. 6. [Color figure can be viewed at wileyonlinelibrary.com]

Figure 12. Comparison of responses to flashed and modulated stimuli

A, response of a representative sSC unit recorded in an anaesthetized animal. Aa, mean response to flashed white or black discs (0.5 s in duration) of varying size. Ab, response to sinusoidal modulation of a uniform field of varying size, at the frequency of modulation (2 Hz; ‘F1’) or twice that frequency (‘F2’). Ac, mean (F0) and modulated (F1) response to large drifting gratings of varying spatial frequency (drifting at 3.8 Hz). B, comparison of preferred size for modulated fields (‘M’, as obtained in Ab) and flashed stimuli (‘F’, as obtained in Aa). Measurements from white (n = 54) and black flashes (n = 59) are overlaid and most units contribute to both datasets. Dashed line indicates unity line. C, comparison of preferred size for modulated fields (‘M’) and preferred spatial frequency for drifting gratings (‘G’, as obtained in Ac) (n = 66). Dashed line is a linear regression in logarithmic co‐ordinates. D, comparison of response purity for flashed stimuli, with index of non‐linearity for modulated fields (F2/F1 ratio; n = 59). Purity indices of 1 indicate response only to white or black, whereas 0 indicates equal response to both. Purity index and F2/F1 ratio are both derived from responses to the preferred size, obtained from modulated fields. E, comparison of response purity for flashed stimuli and index of linearity for drifting gratings (F1/F0 ratio, calculated at the optimal spatial frequency for the F0; n = 60). [Color figure can be viewed at wileyonlinelibrary.com]