We made loose-patch recordings of action potentials from 133 ganglion cells by targeting the largest cell bodies in the ganglion cell layer (~20 μm diameter). The targeting of large somas biases the recordings to one of three cell types: ON alpha/transient cell, OFF alpha/transient cell or OFF delta/sustained cell (Pang et al., 2003
; Murphy and Rieke, 2006
; Margolis and Detwiler, 2007
; van Wyk et al., 2009
). All cells could be unambiguously classified as either ON or OFF type based on the response to brief flashes or white-noise stimuli. In separate experiments, 3D reconstruction from confocal microscope images showed that all cells with large somas stratified in one of three distinct strata of the inner plexiform layer, similar to the corresponding cell types in the guinea pig retina (Manookin et al., 2008
; see also Margolis and Detwiler, 2007
; van Wyk et al., 2009
). For the OFF cells recorded here, response properties were relatively uniform, and therefore we distinguish only ON from OFF cells below where relevant.
Rod-mediated responses generate empirical estimates of rhodopsin sensitivity to green and ultraviolet light
The following experiments aim to distinguish rod- from cone-mediated responses in wild-type retina. Most mouse cones show peak sensitivity to UV light, suggesting that UV stimulation will be useful for studying cone-mediated vision (Jacobs et al., 1991
; Nikonov et al., 2006
). However, rods are also sensitive to UV light due to rhodopsin’s ‘beta-band’ of absorption in the UV range (Govardovskii et al., 2000
; ). Thus, to distinguish rod- from cone-mediated responses to UV light requires a quantitative estimate of the rod’s relative sensitivity to UV and visible wavelengths. The template for rhodopsin’s spectral sensitivity predicts that rods should be 27% as sensitive to our UV LED stimulus as to our green LED stimulus (; Govardovskii et al., 2000
), and we tested this prediction by measuring ganglion cell responses in the Gnat2cpfl3
retina (Chang et al., 2006
). This retina lacks cone function, and thus the response should be mediated by rods. In a ‘balancing experiment,’ a green light stimulus (1.8 R*/rod/s) turned off as a UV light turned on to different intensities. When the UV light matches the green light in R*/rod/s, there should be no response at the transition; whereas when the UV light drives higher or lower R* rates, there should be ‘on’ or ‘off’ responses (). For each cell, we determined the ‘balance point’ when both the onset and offset of the UV light evoked no response. The balance point across cells suggested that the rhodopsin sensitivity to UV light, relative to the green light, was 52 ± 3% (mean ± SEM; n = 13) higher than predicted by the standard template ().
Rod-mediated ganglion cell responses in the Gnat2cpfl3 retina show higher than expected sensitivity to UV light
As a second test of rhodopsin’s sensitivity to UV light, we recorded responses to brief flashes of green or UV stimuli, at several intensities (). The responses at the two wavelengths should match when equated for R*/rod. Consistent with the result above, the ganglion cells showed 48 ± 8% (n = 9) higher sensitivity to UV light than predicted by the standard template. We thus conclude that rhodopsin sensitivity to the UV light stimulus is ~50% higher than predicted [i.e., rods are 41% (27% × 1.5) as sensitive to our UV LED as to our green LED]. This relatively high sensitivity to UV light is consistent with previous in vivo
ERG recordings of rod-mediated responses (Lyubarsky et al., 1999
; ). This enhanced UV sensitivity was taken into account below when calculating R* rates for rhodopsin.
To assess the absolute sensitivity of Gnat2cpfl3
cells in our preparation, we re-plotted the flash response data on a modified R*/rod axis, taking into account the estimated 50% elevation in UV sensitivity described above. On average (n = 9 cells), the response to UV and green light now overlapped (). Furthermore, the absolute sensitivity of the rod-mediated ganglion cell responses was similar to previous measurements in the wild-type retina (Dunn et al., 2006
). Thus, the Gnat2 cpfl3
rod-mediated ganglion cell responses are apparently similar in sensitivity to the wild type retina, consistent with ERG recordings (Chang et al., 2006
Functional estimate of cone opsin distribution across the mouse retina
Most mouse cones (~95%) co-express both M and S opsins in a dorsal-ventral gradient (), and the total M:S ratio across the retina is ~1:3 to ~1:5 (Rohlich et al.,1994
; Lyubarsky et al., 1999
; Applebury et al., 2000
; Jacobs et al., 2004
). Thus, for most of the retina, cone-mediated responses should show strong sensitivity to UV light (). To measure the relative percentage of the M and S opsins along the dorsal-ventral axis, we recorded ganglion cell responses to the green and UV light stimuli in two strains with rod dysfunction: Rho−/−
(Humphries et al., 1997
) and Gnat1−/−
(Calvert et al., 2000
). For each ganglion cell, we presented 200-ms flashes of either green or UV light at several intensities and fit each curve with a Naka-Rushton equation; the relative sensitivity to the two lights was determined by the difference in the half-saturation intensity for each light (see Materials and Methods).
For both the Gnat1−/− and Rho−/− retinas, there was a dramatic shift in spectral sensitivity across the retina. Cells in the dorsal retina showed stronger sensitivity to green light, whereas those in the ventral retina showed stronger sensitivity to UV light (). Based on the relative sensitivities to green and UV light, we calculated the percentage of M opsin across the cone population driving the ganglion cell’s response and plotted this percentage against the ganglion cell’s position along the dorsal-ventral axis (see Materials and Methods; ). In both strains, the M percentage was ~70% in the dorsal retina (2 mm dorsal to the optic disc) but dropped to less than ~5% in the ventral retina (2 mm ventral to the disc), with a steep decline in M percentage beginning at ~1 mm dorsal to the optic disc (). Thus, the cones in the majority of the ventral retina apparently express >95% S opsin.
Estimated ratio of M:S opsin expression as a function of retinal position
We fit a modified Naka-Rushton equation to describe the percentage of M opsin (M%
) as a function of dorsal-ventral position (p
, in mm, starting in the ventral retina, at –2mm from the disc):
is the maximum M percentage minus the minimum M percentage, Mσ
is a half-saturation value, Mn
is the exponent describing the slope of the function, and Mmin
is the minimum M percentage. The best fitting parameters were: Mmax
= 80; Mσ
= 3.2; Mn
= 6.4; Mmin
= 0.8. This equation was fit to the Gnat1−/−
cells, where the dorsal-ventral positions were recorded with relatively high accuracy (see Materials and Methods). However, the general pattern was very similar in the Rho−/−
cells. Notably, of the 35 total cells recorded in the ventral retina, all expressed <10% M opsin, and 86% (30/35) expressed <5% M opsin. In the following experiments, we used the fitted curve to estimate R*/cone/s based on the dorsal-ventral position of each cell.
Rod-mediated responses support band-pass temporal filtering and show light adaptation
The mouse strains described above allow us to assess the temporal properties of isolated rod- or cone-mediated ganglion cell responses. We started by characterizing responses in Gnat2cpfl3 cells using white-noise stimulation and a linear-nonlinear (L–N) cascade analysis. In this analysis, the cell’s response is modeled by a temporal filter and a static nonlinearity (). The filter describes the cell’s temporal sensitivity to the stimulus, and the nonlinearity describes how the filtered stimulus (i.e., a linear model) is converted into a firing rate. The nonlinearity captures the threshold and saturation in the firing response (see Materials and Methods). The L–N model is useful because it provides a compact functional description that captures most of the variance in the response (). We modeled the effect of increasing mean luminance as a change in the linear filter followed by a constant nonlinearity (see Materials and Methods).
Rod-mediated responses showed biphasic filters, indicating band-pass temporal frequency tuning at both levels of mean luminance (; Zaghloul et al., 2005
). The response adapted at the higher mean luminance by becoming faster, which we quantified by the filter’s zero-cross time (). Across cells, the 10-fold increase in mean luminance shortened the zero-cross time from 105 ± 5 ms (mean ± SEM) to 91 ± 4 ms (difference of 14 ± 4 ms; p < 0.01, n = 12). We also plotted the Fourier transform of the linear filter to generate a temporal frequency tuning curve (). The peak amplitude shifted from 5.1 ± 0.3 Hz to 6.9 ± 0.3 Hz with the increase in mean luminance (). Thus, the temporal tuning of the light-adapted, rod-mediated response was sufficient to explain the temporal tuning of downstream circuits and behavior shown previously (see Introduction). The response at the light levels tested (52, 520 R*/rod/s) likely depend on both the rod bipolar pathway and additional pathways for rod signaling (Murphy and Rieke, 2006
): rod synapses with certain types of cone bipolar cells (Soucy et al., 1998
; Tsukamoto et al., 2001
; Li et al., 2010
); and rod gap junctions with cones, which then signal through the cone bipolar circuits (Deans et al., 2002
; Abd-El-Barr et al., 2009
). A previous study of the Gnat2cpfl3
retina also suggested that rod-cone gap junctions were functional despite the lack of cone phototransduction (Altimus et al., 2010
Rod-mediated responses in Gnat2cpfl3 ganglion cells show band-pass temporal tuning and adapt to changes in mean luminance
Rod-mediated responses were nearly saturated at a mean luminance of 5,200 R*/rod/s (see Materials and Methods; ). However, stimulating at this mean luminance for several minutes did not cause substantial bleaching of rhodopsin, as the responses at lower mean luminance could be subsequently re-measured. Rod-mediated responses could be bleached by exposing the tissue to a green LED stimulus that generated ~1.6 × 106
R*/rod/s for two minutes (, pink line). As expected, light responses never recovered following the bleach (measured up to one hour following bleach; n = 10) (Wang and Kefalov, 2009
Cone-mediated responses show high temporal frequency tuning
We measured the temporal properties of pure cone-mediated responses in the Gnat1−/−retina. White-noise responses were generated using a UV LED stimulus in the ventral retina, where most cones express >95% S opsin (). Responses could be measured with a mean luminance of 140 R*/cone/s (). These responses were relatively slow, with a zero-cross time of 122 ± 5 ms (mean ± SEM; n = 4). Increasing the mean luminance to brighter levels (2,000 and 12,000 R*/cone/s) shortened the zero-cross time substantially, to 81± 6 ms at the highest mean (). Filters were biphasic in time and showed band-pass tuning in the frequency domain ().
Cone-mediated responses in ventral Gnat1−/−retina are resistant to bleaching with green light
We tested the effect of the rod bleaching stimulus used above (i.e., bright green light) on the primarily S-cone mediated response of ventral Gnat1−/−
cells. The bleaching light had only a small impact on the cone-mediated response (). The primary effect was a lengthening of the zero-cross time (; increase of 11 ± 3, 9 ± 2, 5 ± 2 ms at the low, middle and high mean luminance, n = 4). This may be caused by a bleaching of the small percentage of M opsin expressed by the co-expressing cones in ventral retina (Lyubarsky et al., 1999
; Nikonov et al., 2006
). These results suggest that the bleaching stimulus could be used in the ventral wild-type retina to bleach rods and isolate an unbleached cone-mediated responses driven by S-opsin stimulation.
Cone-mediated responses in the wild-type retina are fast and robust
In the ventral wild-type retina, where the cones express primarily S opsin, we studied temporal properties of ganglion cell responses across three levels of mean luminance. White-noise modulation of the UV stimulus at the two lower light levels should generate a mixed rod- and cone-mediated response, whereas modulation at the brightest level should saturate the rods and generate a pure cone-mediated response (). The response became faster with increasing mean luminance, as indicated by a shorter zero-cross time and a higher TFpeak (). Across cells, the filter’s zero-cross time decreased to 53 ± 4 ms (n = 6) at the highest mean luminance, with an average TFpeak of 10.7 ± 2.3 Hz (mean ± SEM, n=6) (). Some individual cells showed a TFpeak above 10 Hz ().
Isolated cone-mediated responses in ventral wild-type retina show fast temporal kinetics
The rods were bleached using the bright green stimulus described earlier (), and the cone-mediated responses were studied in isolation. Responses at the lowest mean luminance were suppressed following the bleach, suggesting a strong contribution from rods in the initial, unbleached condition. At the intermediate level, the response became faster, reflecting the cone contribution, whereas at the highest level, the response was largely unaffected by the bleach (). Thus, cone-mediated responses in the ventral wild-type retina can be routinely isolated from the rod-mediated response, in vitro, by a bleaching green light. The isolated cone-mediated response in the wild-type retina showed a TFpeak of 10 ± 1 Hz (n = 10). The amplitude dropped to half the peak (TF50) at 22 ± 2 Hz (). At 30 Hz, the response was, on average, still within a log10 unit of the peak amplitude (). Thus, mouse cones show substantial responsiveness up to 30 Hz, so long as cones are stimulated sufficiently given the local opsin distribution (, ).
Rod and cone systems generate a smooth transition in ganglion cell temporal kinetics across light levels with similarly robust responses
We summarize the above results on the linear filter temporal properties by plotting the filter’s zero-cross time as a function of the R* rate in the photoreceptors driving the response. There is a smooth transition in the temporal response across light levels, as reflected in the Gnat2cpfl3
and wild-type recordings (). The zero-cross time is halved from ~100 ms to ~50 ms across ~2.5 orders of R* rates. The Gnat1−/−
filters showed relatively longer zero-cross times, compared to wild-type. This property of the Gnat1−/−
ganglion cell recordings could be explained by the relatively slow kinetics of the Gnat1−/−
cones, as shown by single cell recordings (Nikonov et al., 2006
Rod and cone systems generate a smooth transition in ganglion cell temporal kinetics across light levels and drive similarly robust responses
We compared the absolute level of rod- and cone-mediated responses by plotting firing rate as a function of the R* rate of the photoreceptors driving the response. Firing rate was quantified as the peak of the nonlinear function minus the rate at f(x)=0 (i.e., the maximum rate minus the estimated rate at 0% contrast). The maximum firing rate was ~150–300 spikes/s across the ~2.5 order of R* rates (). Furthermore, the firing rate increased slightly with mean luminance. Thus, despite having only ~3% cone photoreceptors in the retina, cone-mediated responses in mouse ganglion cells are robust.