Using Silent Substitution to Track the Mesopic Transition From Rod- to Cone-Based Vision in Mice

Abstract

Purpose: To describe the activity of rods and cones in visually intact mice in mesopic conditions, and establish the relative importance of each photoreceptor type in defining the transition from rod to cone vision.

Methods: Using mice (Opn1mw^R) carrying a red-shifted cone opsin, we applied silent substitution methods to record light-adapted ERGs to flash stimuli visible only to rods or cones across a range of light levels (corneal irradiance 10⁹–10¹⁴ photons/cm²/s; ∼10⁰–10⁶ photoisomerisations/rod/s). We tested the impact of selectively changing the background light intensity as experienced by cones on the rod ERG (and vice versa) by adjusting the spectral composition of stimuli. The ERG parameters (b-wave amplitude and implicit time, oscillatory power) were extracted, and their relationship to background intensity and the effective irradiance for cones versus rods/melanopsin was established. We also attempted to record a melanopsin ERG by using modifications of the rod-isolating stimuli.

Results: We saw the predicted decay and increase in rod- and cone-ERG amplitude, respectively, as a function of background intensity. There was only a single irradiance (10¹³ photons/cm²/s) at which both ERGs had high amplitude. Adjustments in the effective irradiance for rods/melanopsin did not impact the cone ERG except at the brightest backgrounds at which there was a melanopsin-dependent suppression of b-wave amplitude. Increasing effective irradiance for cones suppressed rod b-wave amplitude across all background intensities. In addition, we were unable to record a melanopsin ERG.

Conclusions: The cone measure of irradiance was particularly important in driving the transition from rod- to cone-based vision across mesopic light levels.

Whether in starlight or bright daylight, the visual system is capable of responding to the increments and decrements in light levels that occur across space and time. This wide sensitivity range is achieved in part by shifting from rod- to cone-dominated vision. Rods, which can respond to just a single photon, support vision in the dimmest (scotopic) conditions but become increasingly saturated at higher irradiances. Cones are at least 100-fold less sensitive than rods but retain the ability to detect small changes in light intensity under even the brightest (photopic) conditions. 

The transition from rod- to cone-based vision across intermediate (mesopic) irradiances could then in theory be a passive reflection of their different sensitivity ranges. In fact, studies in mice have shown that the sensitivity range of rods and cones overlaps substantially,^1,2 and there is abundant evidence that the relative importance of the two photoreceptors as origins for visual responses is also defined by the behavior of downstream elements in the retinal circuitry (see Bloomfield and Volgyi³ and Witkovsky⁴). Rod and cone signals converge at a number of points as they pass through the retina.^5–7 Saturation and adaptation in these various retinal pathways can therefore provide an additional control over the degree to which the retinal output reflects the activity of rods versus cones. 

One interesting implication of this network-level regulation is that the relative importance of rods may in principle be determined by light intensity as experienced by cones, and vice versa.^8–10 Here we set out to use the method of silent substitution in combination with electroretinography to describe the mesopic shift from rod- to cone-based vision in mice and the relative importance of light intensity as experienced by these two photoreceptor classes in defining it. 

Materials and Methods

Electroretinography

Electroretinograms were recorded from 12 Opn1mw^R, 6 Opn4^−/−;Opn1mw^R, and 4 Cnga3^−/− male mice (aged 3–6 months). Mice were from a mixed C57/BL6;127sv strain. Opn1mw^R refers to the transgenic allele originally generated by Smallwood et al.¹¹ and termed “R” by them. Opn4^−/− mice contain an insertion of tau-lacZ into the melanopsin gene locus, rendering mice “melanopsin-knockout.”¹² Cnga3^−/− animals lack cone function owing to the deletion of a cone-specific subunit (Cnga3) of the cGMP-gated channel.¹³ All animal care was in accordance with the Animals, Scientific Procedures, Act of 1986 (UK) and approved by the local (University of Manchester) ethics committee. All experiments complied with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. Animals were kept in a 12-hour dark/light cycle at a temperature of 22°C with food and water available ad libitum. 

Anesthesia was induced with an intraperitoneal injection of urethane (1.6 g/kg, 30% w/v; Sigma-Aldrich, Gillingham, Dorset, UK). A topical mydriatic (tropicamide 1%; Chauvin Pharmaceuticals, Kingston-upon-Thames, Surrey, UK) and mineral oil (Sigma-Aldrich) were applied to the recording eye before placement of a corneal contact-lens–type electrode. Mice were placed into a stereotaxic frame to keep a fixed head position; a bite bar was also used for head support and acted as a ground. A needle reference electrode (Ambu Neuroline; Ambu, Ballerup, Denmark) was inserted approximately 5 mm from the base of the contralateral eye. Electrodes were connected to a Windows PC (Microsoft, Redmond, WA, USA) via a signal conditioner (model 1902 Mark III, Cambridge Electronic Design; Cambridge, UK) that differentially amplified (X3000) and filtered (band-pass filter cutoff of 0.5–200 Hz) the signal, and a digitizer (model 1401, Cambridge Electronic Design). Core body temperature was maintained at 37°C throughout recording with a homeothermic heat mat (Harvard Apparatus, Cambridge, UK). The ERG a-waves were not readily measurable; b-wave amplitudes and implicit times were measured relative to baseline values (time of flash onset). Oscillatory potentials were extracted with a band-pass filter (50–300 Hz). The peaks of the first four oscillatory potentials (OPs; relative to baseline) were measured and summed to generate the total OP amplitude. Repeated measures (RM) 2-way ANOVAs were used to test for differences in amplitude or latency, with post hoc Bonferroni's multiple comparisons test to compare responses at different background intensities. For b-wave amplitudes, when possible, sigmoidal dose-response curves were fitted to irradiance-response relationships. An F-test comparison was then used to determine whether normalized irradiance-response functions were best with the same or two separate sigmoidal curve(s). To confirm the absence of ERG recordings in certain conditions, a 99% confidence interval based on the preflash potential was calculated. The ERGs were classed as absent if the mean waveform did not cross this threshold. 

Visual Stimuli

Light Calibration.

Stimuli were measured at the corneal plane by using a spectroradiometer (Bentham Instruments Ltd., London, UK) between 300 and 700 nm. The effective photon flux for each photopigment was then calculated by weighting spectral irradiance according to pigment spectral efficiency function (derived from a visual pigment template¹⁴ and λ_max values of 365, 480, 498, and 556 nm for short-wavelength sensitive (SWS) cone opsin, melanopsin, rod opsin, and the introduced long-wavelength sensitive (LWS) cone opsin, respectively) mutliplied by an in vivo measurement of spectral lens transmission.¹⁵ The approach is equivalent to that described in Lucas et al.,¹⁶ using spectral efficiency functions (available in the public domain at http://lucasgroup.lab.ls.manchester.ac.uk/research/measuringmelanopicilluminance). To calculate photoisomerisations (R*)/photoreceptor/cm², the effective photon flux for rod/cone opsins was converted to retinal irradiance by multiplying by 0.18 (see Lyubarsky et al.¹⁷) and multiplied the effective collecting area of a single photoreceptor (0.5 μm² for rods,¹⁸ 1 μm² for cones¹). 

Full Field Visual Stimuli.

All visual stimuli were generated by using a custom-made light source (Cairn Research, Faversham, Kent, UK) containing three independently controlled LEDs (λ_max at 365, 460, and 600 nm). Light from LEDs was combined by a series of dichroic mirrors, passed through a filter-wheel containing neutral-density (ND) filters and focused onto opal diffusing glass (5-mm diameter; Edmund Optics, Inc., York, UK) positioned <1 mm from the eye. The LED intensities and the filter-wheel position were controlled with a PC running LabView 8.6 (National Instruments Ltd., Newbury, Berkshire, UK). 

The LEDs were combined to generate four spectra (A–D), which are summarized in Figure 1 and the Table. Stimuli were designed to recreate the ratio of excitation for SWS and LWS opsins produced by natural daylight (see Allen et al.¹⁹ for detailed description). Transitions between pairs of stimuli (from A to B, or from C to D) were designed to be apparent for rods/melanopsin and to have negligible contrast for LWS and SWS opsins (84% and 87% Michelson contrast for rod and melanopsin, respectively; intensity of [unattenuated] background spectrum = 14.03 rod and melanopsin effective photons/cm²/s). Although both stimulus pairs presented negligible contrast for LWS and SWS cone opsins, they resulted in ∼3.8-fold difference in intensity for these opsins (unattenuated background spectrum of A and B = 14.63 and 13.95 log LWS and SWS effective photons/cm²/s, respectively; unattenuated background spectrum of C and D = 15.21 and 14.53 log LWS and SWS effective photons/cm²/s, respectively). Transitions between pairs of stimuli (A and C, or B and D) were designed to be apparent for cones expressing LWS and/or SWS opsins but provide no change in the excitation of rods and melanopsin (58% Michelson contrast for LWS and SWS opsins; intensity of [unattenuated] background spectrum = 14.63 and 13.95 log LWS and SWS effective photons/cm²/s, respectively). Although both stimuli resulted in negligible contrast for rods and melanopsin, these stimulus pairs resulted in ∼11.7- and ∼14.7-fold difference in background intensity for these opsins, respectively (unattenuated background spectrum of A and C = 14.03 log rod and melanopsin effective photons/cm²/s; unattenuated background spectrum of B and D = 15.10 and 15.20 log rod and melanopsin effective photons/cm²/s, respectively). 

Figure 1

 

Stimulus design and quantification. The output of a three-primary LED light source (peak emission at 354, 460, and 600 nm) was used to generate four spectra, with precise excitation of melanopsin, rod, SWS, and LWS opsins. The figure shows spectral power distribution for these four spectra, labeled A though D.

Table

 

Stimulus Design and Quantification

 

For describing rod contrast response relationship, variations on spectra A and B were produced by symmetrical changes across all wavelengths such that when B was presented against a background of A, predicted rod contrast varied from the extreme 84% (described above) to 4.5%. 

Results

To isolate rod- or cone-driven ERGs, we opted for an experimental approach that we have previously used with success.^19,20 We used Opn1mw^R mice, in which the native mouse green cone opsin is replaced with the human LWS opsin, rendering mouse cones anomalously sensitive to longer wavelengths.¹¹ With a clear divergence in spectral sensitivity between rod opsin and the two cone opsins, we were able to use a three-primary illumination system (λ_max: 365, 460, 630 nm) to deliver carefully calibrated pairs of stimuli, designed to provide significant contrast for only one class of photoreceptor (see Methods and Fig. 1). 

Rod ERGs

We first recorded ERG responses in Opn1mw^R mice to 200 × 50 ms transitions from spectrum A to B (interstimulus interval = 1 second; Fig. 1) across a 6-decade range of background light levels (produced with ND filters). Based on our calculations, this paradigm should appear as a train of flashes for rods while being (close to) invisible for cones (Table). Neutral density filters were placed in the light path so as to reduce light intensities in 1-log unit steps, allowing us to track the rod responses. In support of this prediction, we found that ERG traces were apparent over a range of dimmer irradiances, but undetectable at high light levels (Figs. 2A, 2D). In all cases, the trace was dominated by a single positive deflection (b-wave). The amplitude of this b-wave was stable across lower irradiances (ND5 until ND2, equivalent to 9.83–12.83 log photons/cm²/s, or 10⁰∼10⁴ R*/rod/s; Fig. 2D). This was accompanied by irradiance-dependent decreases in the b-wave latency (implicit time) and increases in the amplitude of higher-frequency OPs (Figs. 2E, 2F). 

Figure 2

 

Rod and cone ERGs over mesopic irradiances. (A) Mean rod ERG traces from a representative mouse, elicited across a range of irradiances by presenting 200 repeats of 50 ms of spectrum B (at time of arrow) against a background of stimulus A. Scale bar: 200 ms (x-axis), 20 μV (y-axis). Figures on the right are background irradiance in log photons/cm²/s. (B) Mean cone ERG traces from a representative mouse, elicited across a range of irradiances by presenting 200 repeats of 50 ms of spectrum D (at time of arrow) against a background of stimulus C. Scale bar: 200 ms (x-axis), 20 μV (y-axis). Figures on the right are background irradiance in log photons/cm²/s. (C) Oscillatory potentials were readily extracted by using a band-pass filter (50–300 Hz); the peaks of four OPs (relative to baseline) were measured and summed to generate the total OP amplitude. Oscillatory potentials were extracted from cone ERGs in (B) with band-pass filter. Scale bar: 200 ms (x-axis), 2 μV (y-axis). (D–F) Quantification of the b-wave amplitudes (D), b-wave implicit times (E), and OP amplitudes (F) of rod (filled circles) and cone (open circles) ERGs across intensities (N = 11 and N = 7, respectively). Lower x-axis plots irradiance in terms of photons/cm²/s; upper x-axis is number of ND filters used to produce the condition. (G–J) Because the rod-isolating stimulus also presented significant contrast for melanopsin, two further stimulus conditions were tested at the brightest intensity to determine whether there was a detectable melanopsin ERG. First, 5-second steps were presented every 60 seconds (lower panel in [G]) and second, three repeats of a 1-Hz sinusoid were presented every 5 seconds (lower panel in [H]). (G, H) Electroretinograms recorded from seven Opn1mw^R mice (gray fine lines), with mean response shown in bold black line; stimulus onset is shown below. (I, J) Electroretinograms recorded from six Opn1mw^R;Opn4^−/− mice (gray fine lines), with mean response shown in bold black line. Scale bars: 1 second (x-axis), 10 μV (y-axis). No repeatable or reproducible deflection was detected in the ERG waveform for either stimulus in either genotype.

Cone ERGs

Our second stimulus (50-ms transitions from spectrum A to C; Fig. 1) was designed to drive a substantial modulation in the effective excitation of both SWS and LWS cone opsins (58.7% Michelson contrast for both cone opsins; Table), but with negligible contrast for rods and melanopsin (<1%). We indeed found that the ERG evoked by this stimulus was strongly positively correlated with background irradiance as expected for a cone-evoked event (Figs. 2B, 2D). At the three dimmest backgrounds (ND5–ND3; 9.8–11.8 log photons/cm²/s, or 10¹–10³ R*/cone/s), this stimulus evoked negligible responses. Above this light level, a b-wave became apparent, the amplitude of which built up over the range 12.8 to 13.8 log photons/cm²/s, or 10⁴ to 10⁵ R*/cone/s to then stabilize at the highest light levels. As with the rod ERG, b-wave implicit time decreased and the amplitude of OPs increased as a function of irradiance (Figs. 2C, 2D, 2E). 

Rod and cone ERGs had equivalent amplitudes at ND2 (2-way ANOVA testing for differences in amplitude with rod or cone stimuli, P = 0.024; post hoc Bonferroni's multiple comparisons test at ND2, P > 0.05). Interestingly, there was no significant difference in implicit time between rod and cone ERGs at this background (2-way ANOVA testing for differences in implicit time with rod or cone stimuli, P = 0.17; post hoc Bonferroni's multiple comparisons test at ND2, P > 0.05), which presumably is important in ensuring that the merged visual signal is coherent. At the 10-fold lower background, the cone ERG was nearly absent, while at the 10-fold brighter background the same was true for the rod ERG. This result indicates that the light-adapted ERG has significant contributions from both rods and cones over only a narrow range of irradiances (∼12.8 log photons/cm²/s, or ∼10⁴ R*/rod/s). 

Validation of Stimuli Using Cnga3^−/− Mice

As a control for our stimuli, we recorded a parallel data set in Cnga3^−/− mice lacking cone function.¹³ These provided an opportunity to validate our method of estimating effective photon flux for each photoreceptor class. Thus, if our cone-isolating stimuli were sufficiently rod silent then we expect no ERG in the absence of cones. Accordingly, the cone-isolating stimuli elicited no significant ERG in Cnga3^−/− mice (ERG trace remained within 99% confidence interval from baseline voltage throughout 550 ms after stimulus presentation in each of four mice tested) at either a rod-favoring background or one at which such stimuli elicited large responses in Opn1mw^R mice (11.8 log photons/cm²/s [10³ R*/rod/s] and 14.8 log photons/cm²/s, or 10⁶ R*/rod/s, respectively; Figs. 3A, 3B). Conversely, Cnga3^−/− mice showed good ERG responses to our rod-isolating stimuli (Figs. 3C, 3D). We next described a rod-contrast response function for the ERG in these animals (Fig. 3E). This revealed a measurable ERG down to 9% Michelson contrast (voltage in 550 ms after stimulus crossed 99% confidence interval for baseline in three out of four mice tested). On this basis, we conclude that the effective rod contrast of our cone-isolating stimuli falls below 9% and below our ability to detect rod-driven responses in these experiments. 

Figure 3

 

Response to rod- and cone-isolating stimuli in Cnga3^−/− mice. (A) Electroretinogram traces in response to our cone-isolating stimuli from four Cnga3^−/− mice recorded at 11.8 and 14.8 log photons/cm²/s. Arrow depicts flash onset. Scale bar: 200 ms (x-axis), 20 μV (y-axis). Numbers to right are irradiance of background in log photons/cm²/s. (B) Electroretinogram traces in response to our cone-isolating stimuli at two intensities (11.8 and 14.8 log photons/cm²/s) from a representative Cnga3^−/− mouse. Dotted lines show 99% confidence intervals based on preflash recordings. Both traces remain below confidence intervals after the flash. Arrow depicts flash onset. Scale bar: 200 ms (x-axis), 20 μV (y-axis). Numbers to right are irradiance of background in log photons/cm²/s. (C) Electroretinogram traces in response to our rod-isolating stimuli from four Cnga3^−/− mice recorded at 11.8 log photons/cm²/s. Arrow depicts flash onset. Scale bar: 200 ms (x-axis), 20 μV (y-axis). (D) Electroretinogram traces in response to rod-isolating stimuli at 84% and 9% Michelson contrast for rod opsin, from the representative Cnga3^−/− mouse in (B). Dotted lines show 99% confidence intervals based on preflash recordings. Both representative responses exceed confidence intervals. Arrow depicts flash onset. Scale bar: 200 ms (x-axis), 20 μV (y-axis). Numbers to right are the effective Michelson contrasts for rods. (E) Mean (±SEM; n = 4) normalized ERG b-wave amplitudes (E) for cone-silent stimuli varying in effective rod contrast from Cnga3^−/− mice. Data are fitted with exponential curve (dotted line).

Isolated Influence of Melanopsin on the ERG Waveform

We have previously used the method of silent substitution to isolate responses evoked by the ganglion cell photopigment melanopsin in the mouse brain.^19–21 It has been suggested that when applied to humans this method can also reveal a melanopsin-derived ERG.²² As spectra A and B differ substantially in effective irradiance for melanopsin, we asked here also whether we could use these to record a melanopsin ERG in mice. The absence of a measurable deflection at light intensities within the melanopsin-sensitivity range (Figs. 2A, 2D) indicates that the 50-ms flash paradigm cannot be used to evoke a melanopsin ERG. We continued to look for a response to two further stimulus presentations: (1) a “sustained” 5-second stimulus presented every minute, with a steady ramp from baseline to stimulus at onset and offset (matching the sort of stimulus that produces melanopsin-evoked responses at the level of retinal ganglion cells and the brain); and (2) a replica of a stimulus (three repeats of a 1-Hz sinusoidal modulation from baseline to stimulus, followed by 2 seconds at an intermediate spectrum) previously reported to evoke a melanopsin ERG in humans.²² Neither stimulus elicited a robust response in either Opn1mw^R or control melanopsin knockout (Opn1mw^R;Opn4^−/−) mice (Figs. 2G–J). 

Rod and Melanopsin Influences on the Cone ERG

The spectra designed to isolate rod and cone ERGs also allowed us to examine the effect of equivalent cone stimuli presented against backgrounds that differed substantially in effective photon flux for rods and melanopsin. Thus, because spectra A and B provide negligible contrast for cones, as do spectra C and D, transitions from A to C and B to D should represent an equivalent stimulus for cones (Figs. 1, 4A, 4B). On the other hand, effective photon flux for rods and melanopsin is low for spectra A and C, compared to spectra B and D. In this way, comparing ERG responses to 50-ms transitions from A to C (“rod/mel-low”) with those from B to D (“rod/mel-high”) allowed us to address the question of how changes in the effective irradiance for rods and melanopsin influences the cone ERG. Using ND filters allowed us to undertake this comparison over a range of backgrounds. 

Figure 4

 

Rod and melanopsin influences on the cone ERG. (A) Spectral power densities of spectra used to generate cone-isolating stimuli against a background with relatively low effective irradiance for rods and melanopsin (“rod/mel-low” condition; background [black] and stimulus [gray] are spectra A and C from Fig. 1). (B) Spectral power densities of spectra used to generate cone-isolating stimuli against a background whose effective irradiance is 12 and 14 times greater for rods and melanopsin, respectively (“rod/mel-high” condition; background [black] and stimulus [gray] are spectra B and D from Fig. 1). (C) Light-adapted cone ERGs from a representative Opn1mw^R mouse under rod/mel-low (black trace) and rod/mel-high (gray). Arrow depicts flash onset. Scale bar: 200 ms (x-axis), 20 μV (y-axis). Numbers to right are effective irradiance of background for an average cone log effective photons/cm²/s. (D–F) Mean (±SEM; n = 6) normalized b-wave amplitudes (D), implicit times (E), and OP amplitudes (F) for light-adapted cone ERGs in Opn1mw^R mice for pairs of rod-divergent stimuli (black filled circles are rod/mel-low and gray open circles are rod/mel-high). The b-wave and OP amplitudes were well fit with sigmoidal dose-response functions; an F-test comparison reveals that rod/mel-low and rod/mel-high data can be fit with a single sigmoidal curve (P = 0.183 and P = 0.97 for b-wave amplitude and OP amplitude, respectively). Repeated measures 2-way ANOVAs revealed significant influence of rod background on cone ERG b-wave amplitude (P < 0.02), but not implicit time (P = 0.61) or OP amplitude (P = 0.38). Bonferroni's comparison between rod/mel-low and rod/mel-high b-wave amplitudes at each irradiance revealed significant difference between backgrounds only at 13.4 log average cone effective photons/cm²/s; (*P < 0.05). Lower x-axis plots irradiance in terms of effective photons/cm²/s; upper x-axis is number of ND filters. (G–I) Replotting of data in (D–F) but with stimulus intensity quantified in terms of rod effective photons/cm²/s. An F-test comparison reveals that b-wave amplitude and OP amplitude data are best fit with separate sigmoidal curves (P < 0.001).

We were able to record reliable cone ERGs at irradiances above ND3 (∼11.4 log cone-opsin effective photons/cm²/s) by using both rod/mel-low and rod/mel-high stimuli (Fig. 4C). To determine whether the amplitude of the cone ERG was determined by the irradiance as experienced by rods/melanopsin and/or cones, this parameter (which was well-fitted with a sigmoidal dose-response curve from ND4–ND1; Fig. 4D) was plotted as a function of background intensity expressed in effective photon flux for either cone or rod photoreceptors. We found that cone ERGs were indistinguishable in rod/mel-low and rod/mel-high conditions across most intensities when expressed as a function of cone effective irradiance (RM 2-way ANOVA testing for differences in b-wave amplitude between with rod/mel-low or rod/mel-high stimuli, P = 0.03; post hoc Bonferroni's multiple comparisons test at ND4–ND2, P < 0.05). The exception was at ND1 (13.4 log cone-opsin effective photons/cm²/s), at which b-wave amplitude was significantly lower in the rod/mel-high condition (post hoc Bonferroni's multiple comparisons test at ND1, P < 0.05). We have recently published evidence of melanopsin-driven adaptation in retinal circuits at high light levels that have the effect of suppressing cone ERG amplitudes.¹⁹ To determine whether this was a likely origin for the difference in b-wave amplitude at the brightest background tested in this study, we repeated the experiments in mice lacking melanopsin (Opn1mw^R;Opn4^−/−). We found no difference between responses to rod/mel-high and rod/mel-low stimuli at any background in this genotype (RM 2-way ANOVA testing for differences in b-wave amplitude between with rod/mel-low or rod/mel-high stimuli, P = 0.11; post hoc Bonferroni's multiple comparisons test at ND4-1, P < 0.05; Figs. 5A–E). 

Figure 5

 

Cone ERGs in Opn1mw^R;Opn4^−/− mice. (A) Light-adapted cone ERG traces from a representative Opn1mw^R;Opn4^−/− mouse, for pairs of rod/mel-divergent stimuli: black traces are rod/mel-low and gray traces are rod/mel-high. Arrow depicts flash onset. Scale bar: 200 ms (x-axis), 20 μV (y-axis). Numbers to right are stimulus irradiance in log average cone effective photons/cm²/s. (B, C) Mean (±SEM; n = 6) normalized b-wave amplitudes (B) and implicit times (C) for light-adapted cone ERGs in Opn1mw^R;Opn4^−/− mice for pairs of rod-divergent stimuli (black filled circles are rod/mel-low and gray open circles are rod/mel-high). Repeated measures 2-way ANOVAs were used to test whether b-wave amplitudes and implicit times were influenced by rod background. No significant effect of rod background was found for cone ERG b-wave amplitude (P = 0.11) or implicit time (P = 0.81). The b-wave amplitudes were well fit with sigmoidal dose-response functions; an F-test comparison reveals rod/mel-low and rod/mel-high data are best fit with a single sigmoidal curve (P = 0.83). Lower x-axis plots irradiance in terms of log cone effective photons/cm²/s; upper x-axis is number of ND filters. (D, E) Replotting of data in (B, C) but with stimulus intensity quantified in terms of rod effective photons/cm²/s. An F-test comparison reveals rod/mel-low and rod/mel-high data are best fit with a separate sigmoidal curve (P < 0.001).

The similarity between rod/mel-high and rod/mel-low ERGs across lower irradiances implies that the effective irradiance as experienced by rods has no detectable effect on the mesopic build-up of cone ERGs. Nor were there any differences apparent in the b-wave implicit times, or OP amplitudes between the two conditions (Figs. 4E, 4F). As a further exploration of this we replotted these data as a function of the effective irradiance for rods. In this case, there was a clear distinction between responses to the two conditions (Figs. 4G–I). For b-wave amplitudes, we applied an F-test comparison to determine whether normalized irradiance-response functions could be fit with the same sigmoidal curve; a single curve was appropriate when expressed in cone (P = 0.18), but not rod effective photons (P < 0.001). This was also the case in animals lacking melanopsin (P = 0.83 and P < 0.001 for cone and rod effective photons, respectively; Fig. 5D). Our data are thus consistent with the view that the increase in the cone ERG across the mesopic transition is entirely dependent upon the background light intensity as experienced by cones, but that once within the photopic regimen it comes under the influence of network adaptation driven by melanopsin. 

Cone Influences on the Rod ERG

We finally used these stimuli to probe cone influences on the rod ERG. Thus, transitions from A to B and C to D should present identical stimuli for rods but against backgrounds that differ in cone effective photon flux (“cone-low” and “cone-high,” respectively; Figs. 6A, 6B). Both sets of stimuli elicited robust ERG responses over the lower irradiances tested (ND5–ND2; 9.07–12.07 log rod-opsin effective photons/cm²/s), but not at ND1 or above (≥13.07 log rod-opsin effective photons/cm²/s; Figs. 6C, 6D). Across the intensity range tested, b-wave amplitude was reliably lower in the cone-high condition, indicating that this parameter was not defined solely by rod-based assessments of irradiance (RM 2-way ANOVA testing for differences in amplitude with cone-low or cone-high stimuli, P = 0.02). Rather, the simplest interpretation of these data is that cones drive a progressive inhibition of rod b-wave amplitude as irradiance increases. However, the pattern of ERG b-wave modulation in the two conditions was not obviously normalized by expressing background irradiance in cone effective photon flux (Fig. 6G). This would argue that the irradiance as experienced by rods is also an important determinant of rod b-wave amplitude. 

Figure 6

 

Cone influences on the rod ERG. (A) Spectral power densities of spectra used to generate rod-isolating stimuli against a background with relatively low effective irradiance for cones (“cone-low” condition; background [black] and stimulus [gray] are spectra A and B from Fig. 1). (B) Spectral power densities of spectra used to generate rod-isolating stimuli against a background whose effective irradiance is four times greater for cones (“cone-high” condition; background [black] and stimulus [gray] are spectra B and D from Fig. 1). (C) Light-adapted rod ERGs from a representative Opn1mw^R mouse under cone-low (black trace) and cone-high (gray). Arrow depicts flash onset. Scale bar: 200 ms (x-axis), 20 μV (y-axis). Numbers to right are rod effective irradiance of background in log effective photons/cm²/s. (D–F) Mean (±SEM; n = 5) normalized b-wave amplitudes (D), implicit times (E), and OP amplitudes (F) for light-adapted rod ERGs in Opn1mw^R mice for pairs of cone-divergent stimuli (black filled circles are cone-low and gray open circles are cone-high). Repeated measures 2-way ANOVAs were used to test whether b-wave amplitudes, implicit times, or OP amplitudes were influenced by cone background. A significant influence of cone background on rod ERG b-wave amplitude (*P < 0.05 [0.019]), but not implicit time (P = 0.39) or OP amplitude (P = 0.70). Lower x-axis plots irradiance in log rod effective photons/cm²/s; upper x-axis is number of ND filters. (G–I) Replotting of data in (D–F) but with stimulus intensity quantified in terms of log average cone effective photons/cm²/s.

The final abrupt reduction in b-wave amplitude between ND2 and ND1 (12.07 and 13.07 log rod-opsin effective photons/cm²/s) occurred simultaneously in cone-low and cone-high conditions (Fig. 6D), implying an event intrinsic to rods. However, there is equally no clear discontinuity in this event when these data were plotted as a function of cone effective irradiance (Fig. 6G). Consequently, additional stimuli providing a greater difference in cone excitation are required to resolve whether the final disappearance of the rod ERG is a product of irradiance as experienced by rods and/or cones. 

Other parameters of the rod ERG were indistinguishable between cone-low and cone-high stimuli across all NDs (RM 2-way ANOVA testing for differences in b-wave implicit time and OP amplitude between with cone-low or cone-high stimuli, P = 0.08 and P = 0.71, respectively; Figs. 6E, 6F), and showed a clear separation when replotted as a function of the effective irradiance for cones (Figs. 6H, 6I). Our data thus provided no support for the hypothesis that cones contribute to the irradiance-dependent modulation of these elements of the rod ERG. 

Discussion

Using silent substitution we have been able to separately record rod and cone ERGs to a simple flash stimulus under equivalent light-adapted conditions. Because this method relies on the difference in spectral sensitivity between rod and cone opsins, it can be applied over a much wider range of conditions than alternative approaches based upon differences in the fundamental sensitivity, rate of bleach recovery, or temporal resolution of the two photoreceptor classes. In theory, silent substitution is applicable to any species in which there is sufficient separation in the spectral sensitivity of photopigments. In practice, this strategy becomes increasingly achievable with larger difference in spectral sensitivity between photopigments in order to achieve reasonable contrast for the target receptor class. As mouse rods and medium-wavelength sensitive (MWS) cone opsin actually have similar λ_max, we have here adopted a transgenic line (Opn1mw^R) in which the native mouse MWS opsin is replaced by the human LWS opsin. As previously reported for this mouse (and for a related strain [see Tsai et al.²³]), this allows generation of stimuli presenting substantial contrast for rods but negligible contrast for cones, and vice versa.^19,20 

A further consideration in applying silent substitution to mice is that most mouse cones coexpress SWS and MWS opsins (or in Opn1mw^R, SWS and LWS opsins¹¹). As the relative expression of the two pigments varies widely across the cone population,^18,24,25 so too does the spectral sensitivity of individual photoreceptors.²⁶ Nonetheless, provided that stimuli designed to be “cone-silent” provide negligible contrast for theoretical photoreceptors containing only SWS or only LWS opsin, they will also be “silent” for any cone expressing both pigments. Our stimuli meet that condition. A related concern is that it is possible to generate “rod-silent” stimuli presenting widely differing contrast for SWS and LWS opsin. This can be useful when one is interested in the separate activity of photoreceptors dominated by one or other pigment.²⁷ However, here our objective was to record activity across the whole cone population and we have therefore also ensured that the effective contrast of our rod-silent transitions is equivalent for cones expressing either or both pigments. Finally, the divergent spectral sensitivity means that SWS and LWS opsins could potentially be exposed to quite distinct mean irradiance levels. To avoid any such issue, our stimuli have also been designed to recreate the ratio of excitation for SWS and LWS opsins produced by natural daylight.¹⁹ In this case, the effective photon flux for individual cones is relatively little impacted by relative expression of SWS and LWS opsins (e.g., for the extreme cases of cones expressing either only SWS or LWS opsin, the maximum difference in irradiance at ND0 is 13.7–14.6 log effective photons/cm²/s). 

As expected, our data revealed an irradiance-dependent transition from rod to cone vision. Thus, rod ERGs are apparent at the lowest irradiance but decay substantially between ND2 and ND1 to be essentially absent at ND0. Conversely, cone ERGs are absent at dim conditions but build up across the range ND3 to ND1. It is only at a single background (ND2) that both rod and cone ERGs have substantial amplitude. In relating our data to the literature, care must be exercised in drawing direct comparisons between thresholds for electrophysiologic and behavioral studies, and allowances made for the possibility that remaining visual pathways are not functioning entirely normally in previously published work with knockout mice. Nevertheless, our data are broadly consistent with previous reports of a rather abrupt transition from rod- to cone-dominated vision.^2,28 The threshold irradiance for eliciting a detectable cone ERG in this study (∼10¹¹ log cone effective photons/cm²/s; estimated to produce ∼700 R*/cone/s) is somewhat higher than reports of thresholds for behavioral responses in mice with defective rod photoreception. Working with the optokinetic response, Umino et al.²⁸ report responses at backgrounds greater than −2 cd/m²/s (1.4 R*/cone/s) in mice lacking rod transducin (Gnat1^−/−), while others^2,29 have reported successful maze navigation in this genotype at intensities greater than ∼10¹⁰ photons/cm²/s (∼50–70 R*/cone/s). Allowing for pupil constriction (which was blocked in our experiments) these data sets thus report functional cone vision at irradiances below that at which we could measure a detectable cone ERG. The reason for this discrepancy is unclear. It is possible that those behavioral assays are better able to reveal small cone responses at low intensities. Alternatively, the presence of functional rods in the Opn1mw^R mice used here could act to inhibit cone responses at dim backgrounds. Interestingly, Naarendorp et al.,¹ working with visually intact animals, suggest that cones define threshold contrast sensitivity from approximately 10³ R*/cone/s, which is much closer to our estimates of cone threshold. 

We found that the rod ERG becomes undetectable at irradiances >10¹² photons/cm²/s (10⁴ rod R*/rod/s). This value is similar to that reported by Nathan et al.² for loss of maze navigation in cone-deficient mice (>10¹² photons/cm²/s), although Umino et al.²⁸ report loss of optokinetic responses in cone-less animals at much lower light levels (>10⁻² cd/m²/s or 10⁰ R*/rod/s). A detailed psychophysical assessment of rod saturation (using cone-deficient mice) has found that while there is a significant loss in contrast sensitivity at backgrounds >10³ R*/rod/s, rods can support visual discrimination up to at least 10⁴ R*/rod/s.¹ That last figure is consistent with other reports of residual rod activity even at the brightest backgrounds (our own unpublished observations^30–32) and highlights the important point that while our failure to record a significant rod ERG at brighter backgrounds is certainly consistent with the view that vision becomes cone dominated, other methods of investigation may reveal rod contributions to vision at higher irradiance. 

The main objective of this study was then to explore how rods and cones each shape the transition from rod to cone vision across mesopic light intensities. By presenting equivalent cone-isolating stimuli against backgrounds that differed substantially in effective photon flux for rods and melanopsin, we were able to confirm our previous finding of a melanopsin-driven modulation of the cone ERG at bright backgrounds. However, we found no evidence that rods influence the build-up in cone ERG amplitude across mesopic intensities. Conversely, we did find a robust difference in the rod ERG when presented under cone-low versus cone-high backgrounds, indicative of a significant influence of cones on the rod ERG. 

However they pass through the retina, rod signals must travel through cone bipolar cells to reach retinal ganglion cells. There are currently three known routes by which this may occur: (1) a chemical synapse from rod bipolar cells to AII amacrine cells, which in turn transmit the signal to cone ON and OFF bipolar cells via gap junction and glycinergic synapses, respectively^33–35; (2) gap junction connectivity between rod and cone photoreceptors⁷; and (3) a direct synapse from rods to cone (OFF) bipolar cells.^36,37 The ERG b–wave primarily reflects the activity of ON bipolar cells and under these conditions—in which the rod bipolar is likely largely saturated (e.g., as reported by Sampath et al.³⁸)—primarily cone ON bipolar cells. The difference in rod-ERG amplitude between cone-low and cone-high backgrounds therefore implies a cone-dependent regulation of the responsiveness of cone ON bipolar cells to signals originating in rods. This could simply be because the impact of rod signals passing through cones or transmitted directly to cone ON bipolar cells becomes increasingly small compared to steady-state cone activity. Alternatively, cones may drive adjustments in coupling between rod and cone pathways according to background light intensity.^39–42 A decoupling of rod/cone pathways at any retinal level would ultimately serve to reduce rod-driven depolarization of cone ON bipolar cells, resulting in a dampened b-wave amplitude. 

If cone signals alone defined the irradiance-dependent changes in rod b-wave amplitude, one would expect the b-wave amplitudes to be equivalent in cone-low and cone-high conditions when expressed as a function of cone effective photon flux. This was not the case (Fig. 6). Our data therefore argue firstly that as irradiance increases across the mesopic range, a progressive inhibition of rod b-wave amplitude associated with cone activity is counteracted by a rod-dependent increase in amplitude. The net effect is to maintain b-wave amplitude at a relatively steady level from ND5 to ND2. Between ND2 and ND1 the rod b-wave amplitude collapses; our data do not allow us to determine whether this is caused by an increase in the inhibitory influence of cones, and/or an irradiance-dependent decay in the signal generated by rods. 

A final point of discussion from our data was our failure to record a measurable modulation in field potential associated with melanopsin photoreception. On the one hand this is unsurprising because melanopsin is expressed in retinal ganglion cells, whose activity in isolation is not reflected in conventional flash ERG traces. Moreover, only a small fraction of retinal ganglion cells express melanopsin.⁴³ However, there is evidence that melanopsin-expressing retinal ganglion cells modulate the activity of cell types (especially amacrine cells^44–46), which could in theory contribute to ERG waveforms. Moreover, it has been suggested that a melanopsin ERG can be recorded from human subjects by using techniques similar to those reported here.²² Our failure to reproduce that finding in mice (even though we were able to generate melanopsin contrasts significantly greater than what is achievable in human subjects) raises questions regarding the general utility of that approach as a method of assaying melanopsin function. 

Acknowledgments

Supported by grants from the European Research Council (268970) and the Biotechnology and Biological Sciences Research Council (BB/I007296/1) to RJL. 

Disclosure: A.E. Allen, None; R.J. Lucas, None