Melanopsin-driven light adaptation in mouse vision

Abstract

Background: In bright light, mammals use a distinct photopigment (melanopsin) to measure irradiance for centrally mediated responses such as circadian entrainment. We aimed to determine whether the information generated by melanopsin is also used by the visual system as a signal for light adaptation. To this end, we compared retinal and thalamic responses to a range of artificial and natural visual stimuli presented using spectral compositions that either approximate the mouse's experience of natural daylight ("daylight") or are selectively depleted of wavelengths to which melanopsin is most sensitive ("mel-low").

Results: We found reproducible and reversible changes in the flash electroretinogram between daylight and mel-low. Simultaneous recording in the dorsal lateral geniculate nucleus (dLGN) revealed that these reflect changes in feature selectivity of visual circuits in both temporal and spatial dimensions. A substantial fraction of units preferred finer spatial patterns in the daylight condition, while the population of direction-sensitive units became tuned to faster motion. The dLGN contained a richer, more reliable encoding of natural scenes in the daylight condition. These effects were absent in mice lacking melanopsin.

Conclusions: The feature selectivity of many neurons in the mouse dLGN is adjusted according to a melanopsin-dependent measure of environmental brightness. These changes originate, at least in part, within the retina. Melanopsin performs a role analogous to a photographer's light meter, providing an independent measure of irradiance that determines optimal setting for visual circuits.

Figure 1

Polyspectral Stimuli Allowing Independent Control of Irradiance as Experienced by Cones versus Melanopsin (A–H) To ensure that some of our stimuli approximated the mouse’s experience of natural light, we recorded spectral irradiance in an urban scene (A) in Manchester over a dusk transition. The resultant spectral power distributions (B, black line shows data for solar angle 11.2°) were multiplied by spectral efficiencies of mouse SWS opsin, melanopsin, rod opsin, and MWS opsin (purple, blue, gray, and green lines, respectively) to calculate effective corneal irradiance in photons/cm²/s for each photopigment over a range of solar angles (C). To allow independent modulation of cone opsins versus melanopsin, these experiments employed Opn1mw^R mice in which mouse MWS opsin is replaced by red-shifted human LWS cone opsin (red dotted line in B). A three primary LED light source (peak emission at 365, 460, and 600 nm) produced four spectrally distinct stimuli shown in (D) with log₁₀ effective photon fluxes for each photopigment in inset. Spectra 1 and 2 approximated the mouse’s experience of natural light at solar angle +8°, while spectra 3 and 4 were selectively denuded of those wavelengths to which melanopsin is most sensitive. The individual elements of these daylight and mel-low stimulus pairs were designed to be rod and melanopsin isoluminant but to differ in effective irradiance for SWS opsin and LWS opsin. By contrast, spectra 1 and 3 and spectra 2 and 4 were designed to be cone isoluminant but to differ substantially in effective photon flux for melanopsin and rods. As a result, switching from either spectrum 1 to 2 or spectrum 3 to 4 produced a 58% Michelson contrast step to cones presented against backgrounds differing substantially in effective photon flux for rods and melanopsin. This was validated by measuring ERG responses to 200 repeats of 1 Hz, 50 ms transitions from either spectrum 1 to 2 and back again (daylight) or spectrum 3 to 4 and back (mel-low). Two control ERG measurements were made in response to daylight and mel-low stimuli: (1) in Opn1mw^R mice at a moderate intensity (100-fold dimmer than maximum and hence with reduced melanopsin excitation; E and F) and (2) in Opn4^−/−;Opn1mw^R mice at the maximum intensity (G and H). (E) and (G) show representative traces in daylight (black traces) and mel-low (blue traces); arrow indicates time of flash. Scale bars, 100 ms (x); 40 μV (y). Population response amplitudes are plotted in (F) and (H). Data were compared with paired t tests. In each control condition, responses to daylight and mel-low had equivalent amplitudes (p > 0.05).

Figure 2

Melanopsin-Dependent Changes in the Cone ERG (A and B) ERG traces (A) from a representative Opn1mw^R mouse to 200 repeats of 1 Hz, 50 ms daylight (black traces) or mel-low (blue traces) flashes at maximum intensity (irradiances as in Figure 1D). Arrow indicates time of flash. Scale bars, 100 ms (x); 40 μV (y). The increase in b-wave amplitude apparent in the mel-low condition in this representative trace was observed in all seven Opn1mw^R mice for which data were recorded (B, black symbols represent b-wave amplitude in daylight condition and blue circles in mel-low condition; paired t test, p > 0.001). (C) The change in ERG b-wave occurred gradually following the switch from mel-low to daylight conditions. Data show mean ± SEM of b-wave amplitude to 180 repeats of the 1 Hz flash stimulus in seven mice. These data were collected in parallel with those for Opn4^−/−;Opn1mw^R data shown in Figure S2A and compared by two-way ANOVA (significant effects of genotype and genotype-time interactions [p < 0.05]; post hoc Bonferroni multiple comparisons tests against Opn1mw^R at time zero revealed significant differences for daylight recordings after approximately 9 min [^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001]). (D and E) Transitions from mel-low to daylight and then back again confirmed that the changes in response were reversible; shown for ERG traces from a representative mouse (D) and for normalized b-wave amplitude from six mice (E; mean ± SEM; one-way ANOVA with a post hoc Bonferroni multiple comparisons test: mel-low versus daylight, p < 0.05; mel-low versus mel-low [recovery], p > 0.05). (F and G) The contrast response relationship differed between daylight and mel-low conditions. ERG traces (F) from a representative mouse exposed to variants of daylight and mel-low stimuli in which the increase in effective cone excitation of the flash (spectra 2 and 4) was altered (Michelson contrasts provided at right) while maintaining rod and melanopsin isoluminance. Scale bars, 100 ms (x); 20 μV (y). Population contrast response relationships (G) were produced by plotting mean ± SEM b-wave amplitude (normalized to 1 = the maximum recorded for that mouse under any condition; n = 6) against flash cone contrast revealed differences between conditions (F-test comparisons of sigmoidal fits to data; curves for mel-low and daylight conditions are significantly different [p < 0.05], but those for mel-low and mel-low [recovery] are not [p = 0.93]).

Figure 3

Melanopsin-Dependent Changes in the dLGN Response to Flash Stimuli (A and B) Responses to daylight and mel-low flashes (50 ms flash at 1 Hz, 200 repeats) in contralateral dLGN were recorded using multielectrode probes. (A) Top: targeted region in sagittal (left) and coronal (right) views, with projected recording site positions superimposed in orange. Bottom: representative multiunit responses recorded in the dLGN of an Opn1mw^R mouse. Histograms show mean changes in firing rate detected at each of 32 recording sites (shown in gray). (B) Changes in multiunit activity averaged across multiple recording sites for a representative Opn1mw^R mouse in daylight and mel-low conditions (black and blue lines, respectively), at full (ND0) and 100 times reduced (ND2) irradiance. Scale bars, 250 ms (x); 10 spikes/s (y). (C) Mean change in multiunit firing rate over the 200 ms following flash onset for recording sites across the dLGN of six Opn1mw^R mice was significantly different under daylight (black) and mel-low (blue) at ND0 but not ND2 (paired t tests). See also Figure S2D. Responses of representative single units reveal that the change in global response amplitude between mel-low and daylight conditions at high irradiance (ND0) are associated with alterations in response properties for many single units. At left are peri-event rasters for 200 repeats, with associated peri-event time histograms to the right (flash onset at time zero). (D) Responses of representative single units reveal that the change in global response amplitude between mel-low and daylight conditions at high irradiance (ND0) are associated with alterations in response properties for many single units. At left are peri-event rasters for 200 repeats, with associated peri-event time histograms to the right (flash onset at time zero). (E–H) The distribution of response latencies for single units (E) reveals a slight change in timing between conditions but no significant alteration in response synchrony between mel-low and daylight conditions (p > 0.05 for F-test comparison of Gaussians fitted to the two distributions, R² > 0.8). Distributions of differences in response amplitude (F) and trial-to-trial reproducibility (G; quantified as Pearson’s correlation coefficient) for single units between conditions (mel-low − daylight) reveal that a large fraction of units showed increases in response amplitude and reliability in the mel-low condition. There was a significant correlation between these two features (Pearson’s correlation coefficient R² = 0.16, p < 0.0001) across the population of units (H). Data shown are from 161 single units isolated in six mice.

Figure 4

Spatial Tuning Properties Are Modulated According to Melanopsin Excitation (A–C) Spatial receptive field mapping. RFs single dLGN units were mapped under both daylight and mel-low conditions using a white noise presentation of black-and-white horizontal and vertical bars upon a gray background. (A) The location of spatial RFs for 17 cells from 4 Opn1mw^R mice (circles show half-maximum width of Gaussian fits of spike-triggered averages in each dimension). (B) RF diameter of these units (half-maximum bandwidth of Gaussian) ranged from 5.6° to 23°. (C) Spatial RFs mapped under both daylight and mel-low conditions for a representative unit. Heatmaps (scale is normalized response) show response to bars in each orientation under each condition, derived spike triggered averages plotted above and to the left. Note the similarity in RF position in each condition. (D–G) Many units that failed to respond to the full-field flash under the daylight spectrum responded well to spatially structured stimuli. (D) Mean peri-event histograms of baseline-subtracted firing rate to 200 presentations of full-field flashes (left) and 40 presentations of inverted gratings (0.035 cpd) in a representative unit under daylight (black) and mel-low (blue) conditions. Flashes occurred at time 0 and 1 and grating inversions every 0.5 s starting at zero. Spatial frequency tuning was examined by recording responses to 1 Hz inverting gratings at four orientations, at five different spatial frequencies (0.035–0.56 cpd) in mel-low and daylight conditions. (E) Mean modulation in firing rate for individual cells when the grating in preferred orientation revealed a shift in the spatial frequency tuning between mel-low and daylight condition. Responses to full-field stimuli plotted for reference (50 ms flashes at 1 Hz). (F) Distribution of preferred spatial frequency (maximum response) for single units in each condition; circle width scaled to represent number of cells in each group (circle size for 10 cells shown above for comparison). Blue circle show cells with higher preferred spatial frequency in mel-low; gray cells show higher preferred spatial frequency in daylight, and lilac cells with no change. (G) The mean response at different spatial frequencies from (E) is replotted for those cells showing a higher preferred spatial frequency in daylight (gray circles in F; top graph), or the remaining cells (lilac and blue circles in F; bottom graph). Data in (E) and (G) are fitted with partial Gaussian curves. Data shown indicate mean ± SEM.

Figure 5

Changes in Temporal Frequency Tuning among Direction-Sensitive Units (A) Temporal frequency tuning was assessed by recording responses to drifting gratings (0.035 cpd) moving in eight orientations, at four temporal frequencies (0.2–2 Hz) in both daylight and mel-low conditions. Across all temporal frequencies, response amplitude (mean ± SEM peak firing rate; for all light-responsive units, n = 117 from five mice) was higher in the daylight condition, but there was no change in the relative effectiveness of different frequencies between daylight and mel-low. Two-way repeated-measures ANOVA, significant effects of frequency (p < 0.0001) and condition (p = 0.036), but not of the frequency-condition interaction (p > 0.05). (B–E) A subset of cells showed direction sensitivity at one or more temporal frequencies. (B) The number of direction-sensitive cells as a function of grating temporal frequency in daylight (gray) and mel-low (blue) conditions, revealing a switch to more abundant direction selectivity for higher temporal frequencies in the daylight condition. This was apparent also at the single-unit level, with examples of cells switching the temporal frequency at which they were most direction sensitive. (C) The DS index was compared for all cells that were classed as direction sensitive in any condition, at any frequency (data shown indicate mean ± SEM). This again revealed a significant change in the temporal frequency tuning of cell responses in mel-low and daylight conditions, with repeated-measures two-way ANOVA revealing significant increase in DS index at 1 Hz in the daylight condition (p < 0.001). (D and E) Representatives of such behavior, with peristimulus rate histograms and associated polar plots depicting response amplitude for movement in each of eight directions differing markedly under daylight (black) and mel-low (blue) conditions. The temporal frequency (TF) at which the grating was presented is provided above. (D) A cell with selectivity for direction of slow moving gratings (0.2 Hz) under the mel-low but not day-light condition. (E) A cell showing direction sensitivity at higher frequencies in daylight but not mel-low.

Figure 6

Responses to Natural Scenes in Daylight and Mel-Low Conditions (A) Example frames from a short natural movie (mice moving around an open arena in horizontal view) presented to Opn1mw^R and Opn4^−/−;Opn1mw^R mice. (B) Raster plots for four representative units from Opn1mw^R mice exposed to 50 repeats of the movie under either daylight (black) or mel-low (blue) spectra. (C) The trial-to-trial reproducibility of firing patterns of individual units (estimated by Pearson’s correlation coefficient; autocorrelation) was significantly higher in the daylight condition (paired t test, p < 0.01; n = 28 from four Opn1mw^R mice). (D) The similarity in firing pattern between pairs of units from the same mouse under mel-low and daylight conditions is depicted as signal correlation (Pearson’s correlation coefficient; 28 pairs from four Opn1mw^R mice). Signal correlation was significantly reduced under the daylight spectrum (paired t test, p < 0.001). This was especially apparent for those pairs of units showing high correlation (>0.25) in either or both conditions, the pink data points. (E–G) The response of melanopsin knockout mice to the movie was similar in daylight and mel-low spectra. (E) Raster plot of firing of a representative Opn4^−/−;Opn1mw^R unit in response to the movie under daylight (black) and mel-low (blue) conditions. At the population level (33 single units from three mice), there was no difference in autocorrelation (F) or signal correlation (G) between daylight and mel-lo conditions. Data shown indicate mean ± SEM.