Rod cells help reset the eye’s internal clock in response to light in mammals

Although bioluminescence monitoring of PER2::Luc retinal explants has been used in several studies [23–25,44,45], a standardized procedure to determine the circadian phase of the retinal clock in vitro is still lacking. In photobiology, this is an essential prerequisite to enable meaningful comparisons between findings from different studies and to replicate experimental conditions. Our observations showed that the trough and the peak of the first PER2::Luc oscillation consistently occurred around circadian time (CT) 8 and CT20 of the circadian cycle (respectively, CT 7.65 ± 1.33 and CT 19.94 ± 1.55; mean ± SD; n = 42; S1 Fig). However, when explants were removed from the incubator for exposure to light (the method generally employed for light exposures), this induced random, robust advances or delays of the phase for each individual retinal explant (S2 Fig). Similar problems related to displacement have previously been shown for other in vitro cultures [46]. To avoid biases due to these artifactually induced phase shifts resulting from physical displacement, we developed a new light-emitting diode (LED)-based light delivery apparatus embedded within the Lumicycle (see Materials and methods). This procedure allowed for an accurate, artifact-free standard protocol to assess the photic dose-response properties (duration, irradiance) of the retinal clock.

Phase-shift properties of PER2::Luc wild-type (WT) retinas were first analyzed using 465 nm monochromatic light of different durations (0.25, 0.5, 1, or 3 h), at a constant irradiance (1 x 10¹⁵ photons/cm²/s), and subsequently at different irradiances for a fixed duration (0.5 h) at CT16. We observed that exposures to 465 nm light from 15 min to 3 h are sufficient to induce significant phase delays (15 min: −1.62 ± 0.30 h; 30 min: −2.05 ± 0.29 h; 1 h: −2.17 ± 0.28 h and 3 h: −2.67 ± 0.17 h; P < 0.001) in comparison to dark control retinas (dark control [DC]: −0.13 ± 0.13 h; Fig 1A, left panel, and Fig 1B). The slope of the stimulus 4-parameters curve (Naka–Rushton fit, Fig 1B) is steep, resulting in a narrow stimulus-duration range with a half maximum response at 0.51 h. In order to verify that the duration of culture and of light exposure did not affect the amounts of photopigments at the end of the experiment, levels of opsin mRNAs (MW and SW opsins, rhodopsin, melanopsin, and neuropsin) were quantified from the retinal explants (S3 Fig). We observed no differences in the relative expressions of the opsins between stimulated and DC retinas.

Using a 30-min duration (half-saturation value) of 465-nm light exposures to establish an irradiance-response curve, we found that 1 x 10¹³ photons/cm²/s was insufficient to induce a significant delay (−0.55 ± 0.45 h; P = 0.22; Fig 1A, right panel, and 1C), whereas higher irradiances (10¹⁴–10¹⁵ photons/cm²/s) produced significant phase delays (respectively, −1.73 ± 0.22 h and −2.05 ± 0.29 h; P < 0.01; Fig 1C). The slope of the irradiance-response curve again was steep, with an irradiance of 2.38 x 10¹³ photons/cm²/s necessary to induce a half maximum phase shift. A classical property of the circadian system is the ability to integrate photon number over time [47–50]. For the retina, plotting response amplitude in relation to the total number of photons yields a coherent photon dose-response function with a half maximum response at 5.53 x 10¹⁷ photons/cm² and saturation of the response above 3.5 x 10¹⁹ photons/cm² (Fig 1D).

Based on the optimal parameters of duration and irradiance, PER2::Luc retinal explants were exposed to equal quanta of monochromatic light (30 min, 1 x 10¹⁴ photons/cm²/s) of different wavelengths (395, 465, and 520 nm) at CT16. The choice of the three wavelengths was based on the peak photoreceptors sensitivities in the WT mouse (SW-cone opsin, λ_max = 360 nm; OPN5, λ_max = 370 nm, melanopsin, λ_max = 479 nm; rhodopsin, λ_max = 498 nm and MW-cone opsin, λ_max = 508 nm; S4A Fig; [37,42,51]). Significant phase delays of PER2::Luc are observed at 395 nm (−2.13 ± 0.62 h), 465 nm (−1.73 ± 0.22 h), and at 520 nm (−1.63 ± 0.38 h) compared to the DC (−0.13 ± 0.13 h, P < 0.001; Fig 2). The phase delays induced at 395, 465, and 520 nm were not significantly different from each other, suggesting that the irradiances used were at saturating levels. In addition, since the stimulation at 520 nm corresponds to a more than 5 log unit decrease in SW opsin and OPN5 sensitivities (S4B Fig), it appears unlikely that OPN5 alone can account for the light-induced phase shifts of the retinal clock in the visible range of the spectrum, suggesting a putative role of rods, MW cones, and/or melanopsin.

To determine the photoreceptors involved in the phase-shift response of the retinal clock, we used 520-nm stimulations to rule out possible contributions of SW cones and OPN5. However, at this wavelength, MW cones, rods, and ipRGCs have relatively similar spectral sensibilities, precluding determination of their relative contributions in the WT mouse. We thus backcrossed mouse models that are deficient for each of these photoreceptors with the Per2^Luc mice to obtain rodless (Nrl^−/−::Per2^Luc), MW coneless (TRβ^−/−::Per2^Luc), and melanopsin knockout (Opn4^−/−::Per2^Luc) models. A “rod-only” (Opn4^−/−::TRβ^−/−::Per2^Luc) model was obtained by crossing Opn4^−/−::Per2^Luc and TRβ^−/−::Per2^Luc mice. Opn4^−/−::Per2^Luc, TRβ^−/−::Per2^Luc, and Opn4^−/−::TRβ^−/−::Per2^Luc mice show a similar phase shift (respectively, −1.33 ± 0.47 h; −1.63 ± 0.23 h and −1.44 ± 0.28 h; Fig 3A and 3B) following 30 min exposure to 1 x 10¹⁴ photons/cm²/s at 520 nm compared to WT Per2^Luc mice (−1.46 ± 0.30 h; P = 0.38, P = 0.51, and P = 0.34). In contrast, the absence of rods in the Nrl^−/−::Per2^Luc model totally abolished the light-induced phase shift at 520 nm (−0.18 ± 0.21 h, P < 0.01). Furthermore, the absence of cones, rods, and melanopsin in the Opn4^−/−::rd/rd::Per2^Luc mice prevented any light-induced phase shift of the retinal clock (−0.05 ± 0.56 h, P < 0.01). To eliminate the possibility that the differences in the phase shift could be related to a light-induced change in the period, we compared periods before and after the stimulation in retinas and in the DC and found no differences between genotypes (Fig 3C). The same experiments have been performed on heterozygous mice of each genotype, and similar results were obtained (S5A and S5B Fig). Taken together, these results suggest that rod input alone is required to shift the retinal clock in the visible part of the spectrum. We then assessed whether a phase shift of Nrl^−/−::Per2^Luc retinal explants could be obtained in the UV part of the spectrum. At 1 x 10¹³ photons/cm²/s (30 min, 395 nm), we found a significant and similar phase delay in both WT Per2^Luc and Nrl^−/−::Per2^Luc retinal explants (respectively, −1.29 ± 0.09 h and −1.09 ± 0.45 h) by comparison with the DC retinas (P < 0.01; Fig 3D). In the WT mouse, the phase delay obtained at 395 nm is significantly increased from the DC retinas, whereas the same equal quanta stimulation at 465 nm did not induce a significant phase shift (Fig 1C). Using an increased irradiance at 395 nm (1 x 10¹⁴ photons/cm²/s; 30 min), we found a significantly reduced phase shift (−0.98 ± 0.24 h) in the Nrl^−/−::Per2^Luc compared to that of WT Per2^Luc retinal explants (−2.13 ± 0.62 h; P < 0.05). The residual phase shift in the UV suggests a possible involvement of SW cones and/or OPN5 in addition to rods.

To ascertain if the Nrl^−/−::Per2^Luc mouse presents a global deficit in the light response of the circadian system, we examined the amplitude of locomotor activity phase shifts induced by a 15-min pulse of 530 nm monochromatic light at two different irradiances (2.8 x 10¹² and 2.8 x 10¹⁴ photons/cm²/s; Fig 3E). No significant differences in the magnitude of the phase shift and the pattern of locomotor activity were observed between Nrl^−/−::Per2^Luc and WT mouse at both irradiance levels used. By contrast, the light-induced phase shift is dramatically reduced at both irradiances in the Opn4^−/−::TRβ^−/−::Per2^Luc mice, which otherwise retained light-induced phase shift of the retinal clock (Fig 3B and 3E).

Finally, we assessed whether invalidation of photopigments in the different mouse strains altered the endogenous functioning of the retinal clock. Our results show a role of melanopsin and/or MW cones since a significant shortening of the endogenous period was observed in Opn4^−/−::Per2^Luc (23.93 ± 0.15 h; P < 0.001), TRβ^−/−::Per2^Luc (23.71 ± 0.09 h; P < 0.001), Opn4^−/−::TRβ^−/−::Per2^Luc (23.82 ± 0.01 h; P < 0.001), and Opn4^−/−::rd/rd::Per2^Luc (24.21 ± 0.2 h; P < 0.05) mice compared to WT Per2^Luc (24.69 ± 0.08 h; S5C Fig).

Comparison of photoreceptor spectral absorptions with the relative sensitivity of evoked responses is a critical strategy for identification of the photopigments mediating non-image forming (NIF) responses to light in rodents [52]. Light entrainment of the retinal clock is gated in a phase-specific manner, as in the SCN, with maximum phase delays occurring at CT16 and phase advances during the late subjective night [23–25].

The dose-response function for eliciting a phase shift and reciprocity, core properties of the circadian system, have not been previously evaluated for the retinal clock. These properties translate the ability to integrate photic input over a relatively long period of time, ranging from a few seconds to several hours, and to respond proportionally to the total energy of the stimulus [47–50]. Compared to previous studies in the retina [23–25], we find that relatively shorter duration exposures (15 min) at lower irradiance levels (1 x 10¹⁴ photons/cm²/s) are sufficient to induce a phase delay of PER2::Luc signal. In particular, the studies by Buhr and colleagues used 3-h light exposures at 1 x 10¹⁵ photons/cm²/s [24, 25]. This value in terms of total photon number is 2 log units higher than the amount required to induce a phase shift in our study and is in the range of saturating amounts. The stimulus irradiance threshold for eliciting a retinal phase shift is nevertheless relatively high (>10¹³ photons/cm²/s, present study) compared to the energy required for a behavioral phase shift (approximately 10¹⁰–10¹¹ photons/cm²/s, [18,20,53,54]). Fig 4A shows the differences in sensitivity for phase-shifting and entrainment responses of retinal and SCN clocks to light between 465–520 nm from our and other laboratories [18,20,22,49,53–56]. Furthermore, the retinal clock appears unable to integrate light energy for durations longer than 30 min. This response resembles the pattern of light-induced FOS expression in the retina that increases sharply at a relatively high level of irradiance and long duration (Fig 4A, [49]). The difference in light sensitivity between retinal and SCN clocks suggests clock-specific tuning of light responses in each system that may be related to different integration and/or feedback mechanisms in these two clocks, downstream from the photoreceptor level.

A potential role of rods, cones, and ipRGCs in the light response of the retinal clock has recently been challenged by two studies from Buhr and colleagues claiming that none of these photoreceptors are involved in entrainment of the retinal clock and that OPN5, a UV-sensitive retinal opsin, is the sole photopigment involved [24,25]. Our findings show phase delays of PER2::Luc oscillation using similar quanta of 395, 465, or 520 nm light, demonstrating that the retinal clock is capable to respond across a broad range of visible wavelengths. Since the spectral sensitivity of OPN5 is attenuated by more than 5 log units at 520 nm (S4B Fig), a robust response at this wavelength strongly argues against an exclusive mediation of light-induced phase shifts by OPN5. Furthermore, even at the short wavelength (417 nm) employed by Buhr and colleagues [25], it is difficult to rule out the possibility that the long-duration and high-irradiance exposures applied at this wavelength [25] could also activate rods, MW cones, and/or ipRGCs.

Since MW cones, rods, and ipRGCs have largely overlapping spectral sensitivities and are thus difficult to completely isolate in the WT mice using spectral stimulation strategies, we used Per2^Luc mouse models deficient in different photoreceptor classes (Nrl^−/−::Per2^Luc; TRβ^−/−::Per2^Luc, Opn4^−/−::Per2^Luc, Opn4^−/−::TRβ^−/−::Per2^Luc, and Opn4^−/−::rd/rd::Per2^Luc). The Nrl and TRβ genes are essential for photoreceptor development. In TRβ^−/− knockout mouse, MW opsin is not expressed, and all cones express SW opsin [55,57]. Nrl encodes a transcription factor that is essential for rod development [58–60]. When Nrl gene is knocked out, a complete absence of rods is observed, as revealed by histology, immunocytochemistry, electrophysiology, and gene expression analysis. Rod progenitor cells differentiate into SW cones [59,61,62], and this mouse is widely used as a cone-only model in vision [63]. We found no change in the relative expression of OPN5 and a slight increase in melanopsin and MW opsin mRNAs (S6 Fig). In addition, behavioral phase shifts in WT and Nrl^−/− mice are similar, indicating a dichotomy of the light response between retinal and central clocks. The use of these models suggests that rods but neither MW cones or melanopsin are required for in vitro light-induced phase shift of the mouse retinal clock. In agreement, Buhr and colleagues [24,25] showed in WT mice (in which rods are conserved) a large phase shift of more than 3 h using 475-nm stimulus in the rod sensitivity region of the visible light spectrum.

The involvement of rods at the relatively high irradiance levels employed here appear paradoxical compared to their sensitivity range for visual responses at low scotopic light levels. However, rods have a wide response range and can drive dopaminergic cell responses in the retina, pupillary constriction, and behavioral entrainment not only under dim light levels but also at higher light levels within the sensitivity range of cones [64–66]. This response appears to be mediated through rod–cone pathways, involving gap junctions between rods and cones [65]. As a preliminary observation, retinal explants from Per2^Luc mice that were exposed to 520 nm light in the presence of carbenoxolone (CBX), a general gap junction blocker, fail to exhibit a phase shift compared to DC (CBX: 0.5 ± 0.04 h; n = 4; P = 0.21). These data are thus consistent with the idea that rods have the capacity to elicit phase shifts to light at high irradiances for both retinal and SCN clocks (Fig 4B).

Since rods are involved, the relatively high threshold necessary to obtain a phase shift of the retinal clock is unexpected. Indeed, a 30-min exposure of 1 x 10¹³ photons/cm²/s of 465 nm light is insufficient to induce a phase shift. Compared to rod-driven visual responses, thresholds for NIF responses are, in general, relatively high and can vary in sensitivity by several log units, depending on the response (Fig 4A, [18,20,22,49,53–56]). Even in WT mice, the threshold to elicit a behavioral phase shift is in the mesopic range (approximately 10¹⁰–10¹¹ photons/cm²/s [18,20,53–56], Fig 4A), well above the sensitivity of rods [67]. In support of a difference in light sensitivity between retinal and SCN clocks, we also found differences in their phase-shifting response in the Nrl^−/−::Per2^Luc and the Opn4^−/−::TRβ^−/−::Per2^Luc, suggesting a clear dichotomy in their light response.

The discrepancy between the current study and those of Buhr and colleagues may be related to methodology and the responses studied [25]. A first issue is related to the variable artifactual phase shift caused by physical displacement of the retinal explants, an effect also previously shown in cultured retinal pigment epithelium (RPE) [46]. The use of long light durations (3 h) by Buhr and colleagues [24,25] may, to some extent, mask the effect of the physical displacement on the phase of the retinal clock. Secondly, these authors mainly used an entrainment paradigm (9L:15D, [24,25]), whereas we (present study) and Ruan and colleagues examined the acute effect of short-duration light pulses on the phase of the retinal clock [23]. Moreover, the failure of WT retinal explants to entrain to a light:dark (L:D) cycle at 530 nm [25] is puzzling, since significant phase shifts have been demonstrated using, respectively, 520 nm or broadband while light (present study, [23]).

Our data show a residual sensitivity to light at 395 nm in rodless mice, suggesting that OPN5 and/or SW cones may contribute to the phase shift (Fig 3D). Buhr and colleagues attribute the phase shift of the retinal clock in SW knockout mice at 417 nm to OPN5 [25]. However, the use of long-duration and high-irradiance light does not exclude a rod contribution since the spectral sensitivity of rods is only slightly reduced compared to that of OPN5 at this wavelength. High light levels in in vitro explant cultures that lack photoprotective RPE and light absorption by the lens can also provoke phototoxic cell damage and other deleterious effects [68,69]. The RPE plays an important role in the maintenance of the health and function of photoreceptors. In particular, rhodopsin and cone opsin pigments require a continuous supply of visual chromophore to maintain photosensitivity in bright light. This is carried out by multistep enzyme pathways in the RPE and Müller cells. We designed our retinal explant culture without the RPE for three main reasons: firstly, the RPE also contains a circadian clock with a rhythmic expression of PER2::Luc [46,70–72] that may interfere with the signal emitted by the retina alone. Secondly, Kaylor and colleagues recently demonstrated in vivo and in vitro that rhodopsin may regenerate in photoreceptor membranes in the absence of the RPE [73]. Finally, we were careful in our strategy to use a single light stimulation in each retinal explant to avoid issues of photopigment depletion (S3 Fig).

The effect of photoreceptor/photopigment loss on the endogenous period of the retinal clock may have different explanations. The retinal clock is composed of a network of multiple strongly coupled circadian oscillators located within distinct cellular layers [16,30,45,74]. Photoreceptors (cones and ipRGCs) have also been shown to contain a cellular circadian clock (for review, see [2]), with the notable exception of rods. All of the mouse models examined in the present study (Opn4^−/−::Per2^Luc, TRβ^−/−::Per2^Luc, Opn4^−/−::TRβ^−/−::Per2^Luc, and Opn4^−/−::rd/rd::Per2^Luc) exhibit shortening of the endogenous period, except the Nrl^−/−::Per2^Luc mouse model. This effect on the endogenous period of the retinal clock may be related to the absence of one of these cellular oscillators from a complex tightly coupled network leading to a modification of the endogenous period at the level of the entire retina. In agreement with this hypothesis, we did not observe any impact of the absence of rods on the period of the retinal clock in the Nrl^−/−::Per2^Luc mice. In addition, it cannot be excluded that the period changes might be secondary to developmental effects induced by the absence of a photoreceptor/photopigment.

In conclusion, our findings reveal that the absence of rods but not of melanopsin or MW cones totally prevented a light-induced phase shift in the visible spectrum and suggest additional recruitment of SW cones and/or OPN5 at shorter UV wavelengths. A putative role of OPN5 in humans and nonhuman primates is, however, questionable since, in contrast to UV transmittance of the mouse lens, UV wavelengths are effectively blocked out by the human/primate lens [68,75–81]. Furthermore, while the first study reported the expression of OPN5 in human retinas [36], recent works do not detect this opsin both in human [37] and in nonhuman primate retinas [70]. Finally, we show a clear dichotomy in the light response between retinal and SCN clocks, highlighting the need for a comprehensive understanding of the neural circuits involved in the light response of the retinal clock.

All animal procedures were in strict accordance with current national and international regulations on animal care, housing, breeding, and experimentation and were approved by the regional ethics committee CELYNE (C2EA42-13-02-0402-005). All efforts were made to minimize suffering.

Mice were housed in a temperature-controlled room (23 ± 1 °C) under 12 h light/12 h dark cycle (12L/12D, light intensity around 200 lux), with food and water ad libitum. Per2^Luc mice [82] and several photoreceptor-deficient mice were used: TRβ^−/− lacking MW opsin [55,83], Opn₄^−/− knockout for melanopsin [20], and Nrl^−/−, characterized by the complete loss of rods and an increased number of SW cones [59]. All photoreceptor-deficient mice were bred with the Per2^Luc mice to obtain TRβ^−/−::Per2^Luc, Opn₄^−/−::Per2^Luc, and Nrl^−/−::Per2^Luc mice. TRβ^−/−::Per2^Luc and Opn₄^−/−::Per2^Luc were then crossed together to obtain Opn₄^−/−::TRβ^−/−::Per2^Luc mice. The Opn4^−/−::rd/rd::Per2^Luc mice were obtained from Dr. Van Gelder and Dr. Buhr. All lines were maintained on a C57BL/6J background. We used female and male mice in all experiments. All mice were used between 2–4 months old, except the Nrl^−/−::Per2^Luc model, which were used at 4 weeks before the onset of apoptotic degeneration [84]. The Opn4^−/−::rd/rd::Per2^Luc mice were aged >200 days. To verify that photoreceptor degeneration (rods and cones) was complete, light entrainment of the rhythm of locomotor activity of each animal was analyzed. Only animals exhibiting a free-running locomotor rhythm were used.

Mice were killed by cervical dislocation 1 h before light offset (Zeitgeber Time 11 or ZT11), except the Opn4^−/−::rd/rd::Per2^Luc mice that were euthanized at CT11 according to their behavioral locomotor rhythm. Eyes were enucleated and placed in Hank’s balanced salt solution (HBSS; Invitrogen) on ice. Retinas were gently isolated from the rest of the eye cup and flattened ganglion cell layer upon a semipermeable (Millicell) membrane in a 35-mm culture dish (Nunclon) containing 1.2 mL Neurobasal-A (Life Technologies) with 2% B27 (Gibco), 2 mM L-Glutamine (Life Technologies), and 25 U/mL antibiotics (Penicillin/Streptomycin, Sigma), incubated at 37 °C in 5% CO₂ for 24 h. From this step on, all manipulations of explants were performed under dim red light. After 24 h, at the projected ZT12, retinas were transferred to 1.2 ml of 199 medium (Sigma), supplemented by 4 mM sodium bicarbonate (Sigma), 20 mM D-glucose (Sigma), 2% B27, 0.7 mM L-Glutamine, 25 U/mL antibiotics (Penicillin/Streptomycin, Sigma), and 0.1 mM Luciferin (Perkin). Culture dishes were sealed and then placed in a Lumicycle (Actimetrics, Wilmette, IL, United States of America) to record the global emitted bioluminescence. For blocking gap junction, 100 μM CBX was added to 199 medium. PER2::Luc bioluminescence was analyzed using Lumicycle Analysis software (Actimetrics, Wilmette, IL, USA).

All retinal explants were dissected at ZT11 and cultured just before light offset (ZT12). The projected ZT12, at which medium was changed and recording started, was then considered as CT12 and used as a time reference (). The time of occurrence of the trough and the peak of the first complete PER2::Luc oscillation were determined by using this CT12 reference and by correcting time for the endogenous period (). The phase of the peak was then used to calculate the timing of the following CT16 to apply light stimulation or physical displacement. S1A Fig S1B Fig

The classical procedure commonly employed to assess light-induced phase shifts involves the transfer of the cultured tissue from the Lumicycle into a light-stimulation chamber. To evaluate the effects of displacement of retinal culture dishes on the phase of PER2::Luc, we first established a 4-day baseline bioluminescence signal for each sample in the Lumicycle (Actimetrics), and retinal explants were then cautiously transferred to a nearby incubator in a light-proof, insulated chamber at CT16. Subsequently, the tissue was returned to the Lumicycle, and the phase shift was measured. Each retinal explant was submitted to three successive movements, each separated by a medium change.

To avoid effects of displacement, we developed a new light-delivery apparatus embedded within the Lumicycle. This device is composed of an opaque matrix that fits to the shape of the five exposed dishes on the turntable and by a black cylinder reflective white inside containing the LEDs (Super Bright LEDs). To avoid any light diffusion to the photomultiplier tubes during the light stimulation, the bottom edge of the cylinder was sealed to the contours of the matrix with light impermeable seals inside the Lumicycle. Temperature was monitored by placing a temperature data logger (HOBO data logger, ONSET) inside the light delivery apparatus, at the level of the retinal explants. The temperature change occurring during light stimulation is 0.57 ± 0.01 °C. Retinas placed on the opposite side of the Lumicycle during the light stimulations were used as DCs. To exclude that the temperature change in the incubator due to the heat from the light source is responsible for the phase shift, an additional control was included. Using the same apparatus, we applied a similar light stimulation (520 nm, 30 minutes, 10¹⁴ photons/cm²/s) with an opaque filter inserted between the LEDs and the retinal explants and completely blocks any light transmission to the retinal cultures. The change in temperature (0.59 °C ± 0.01 °C) is identical to that observed without the opaque filter (0.57 °C ± 0.01 °C). Under these conditions, no significant phase delay of PER2::Luc oscillations of the retinal explants was observed (−0.52 ± 0.19 h, n = 4, P = 0.13).

Raw bioluminescence data were detrended using a 25-h running average and smoothed with a 3-h running average method. The phase and the period were determined by using best-fit sine wave function (sin fit) of the Lumicycle Analysis software. Using this function, the phases (pre- and poststimulation) are determined as the maximum of the oscillation. The phase of the prestimulation rhythm (phase of the third oscillation before the light stimulation) was used to predict the phase of the PER2::Luc oscillation (phase of the third oscillation before the light stimulation plus two values of the endogenous period) if no light stimulation (or mechanical displacement) was applied (predicted phase). The calculation of the phase shift is based on the difference between the predicted phase and the observed phase (phase of the first complete oscillation after the light stimulation/displacement) in the same retinal explant. The observed phase is also determined by the best sine wave function (sin fit function) on three complete oscillations after light stimulation/displacement. The circadian period (pre- and poststimulation) are determined on the 3 days, respectively, before and after light stimulation. The phase shifts were calculated by the experimenter and by a person blind to the different light/genotype conditions. The goodness of fit was determined and was between 93.3% and 97.5%, with an average of 95.93% ± 0.29%.

According to the experiment, retinal explants are exposed to different durations (0.25, 0.5, 1, and 3 h) and irradiances (10¹³, 10¹⁴, and 10¹⁵ photons/cm²/s) of 465 nm monochromatic light. Light stimulations were done at CT16 since previous data has shown that light maximally phase delays the retinal clock at this CT time [25]. Thereafter, we used constant irradiance and duration (10¹⁴ photons/cm²/s, 30 min) to study the effect of different wavelengths (395, 465, and 520 nm) using bright LED light sources (Superbrightleds; S4C Fig). Data from the irradiance, duration, and total number of photons were fit with a four-parameter logistic equation using a modified form of the Naka–Rushton equation previously described [49,85]. Radiometric measurements were made by using an International Light model IL1700 photometer (International Light Technologies) and a spectrophotometer (Specbos 1211, JETI).

The relative expression of the different opsins was quantified in cultured WT Per2^Luc retinas (light stimulated and DC) and collected at the end of the duration experiment and in retinas from 4-week-old WT and Nrl^−/− mice. Cultured retinas were snap frozen at the trough after 10 days in culture, during which they had been exposed to 0.5, 1, or 3 h of light or no light (DC). Total RNA was extracted using Trizol reagent (Invitrogen) and reverse transcribed using random primers and MMLV Reverse Transcriptase (Invitrogen). Real-time reverse transcription PCR (RT-PCR) was then performed on a LightCycler system (Roche Diagnostics) using the light Cycler-DNA Master SYBR Green I mix. Hprt was used for internal standardization of target gene expression. The efficiency and the specificity of the amplification were controlled by generating standard curves and carrying out melting curves. Relative transcript levels of each gene were calculated using the second derivative maximum values from the linear regression of cycle number versus log concentration of the amplified gene. Primer sequences were Hprt sens ATCAGTCAACGGGGGACATA and reverse AGAGGTCCTTTTCACCAGCA, SW opsin sens CAGCCTTCATGGGATTTG and reverse GTGCATGCTTGGAGTTGA, MW opsin sens GCTGCATCTTCCCACTCAG and reverse GACCATCACCACCACCAT, Rhodopsin sens GCCACCACTCAGAAGGCAG and reverse GATGGAAGAGCTCTTAGCAAAG, Melanopsin sens TGCGAGTTCTATGCCTTCTG and reverse GGCACGTAGGCACTCCAAC, Neuropsin sens ACTATGCACCTGAGCCCTTC and reverse TGGCTGCTATGGATTCGACT, and Per2 sens CCACACCTTGCCTCCGAAAATA and reverse ACTGCCTCTGGACTGGAAGA.

Singly housed male WT Per2^Luc, Nrl^−/−::Per2^Luc, and Opn₄^−/−::TRβ^−/−::Per2^Luc mice (WT: n = 6; Nrl^−/−::Per2^Luc and Opn₄^−/−::TRβ^−/−::Per2^Luc: n = 3–5) were first entrained in a 12L/12D cycle for 20 days. Subsequently, animals were maintained in constant darkness (DD) to examine the free-running period calculated by periodogram analysis using ClockLab software (Actimetrics). Phase shifts were studied using a single 15-min monochromatic light pulse (530 nm, half-bandwidth, 10 nm) at 2.8 x 10¹² photons/cm²/s and 2.8 x 10¹⁴ photons/cm²/s applied at CT16. The stimulator (light source and chamber) has been described previously [49,55]. After the light pulse, animals were returned to their home cages, and activity was monitored in DD for an additional 15 days before the next light pulse. The magnitude of a light-induced phase shift was determined from the difference between the regression lines of the activity onsets before and after the light stimulation, extrapolated to the day following the light pulse. The transient responses on the 3–4 days immediately after the pulse were discounted [86].

Four-week-old WT and Nrl^−/− mice were killed by CO₂ inhalation and decapitation. Whole eyes were dissected and immersion fixed in 4% paraformaldehyde in phosphate-buffered saline (PBS) overnight at 4 °C. Eyes were frozen, cut at 20-μm thickness, and mounted on glass slides. Retinal sections were hydrated in graded ethanol solutions (95%, 75%, 50%) for 30 s each. The retinal sections were then stained with 1% cresyl violet and dehydrated in graded ethanol solutions (50%, 75%, 95%, 100%).

Normal distribution of the data and homogeneity of the variance were respectively tested using Shapiro–Wilk and the Levene’s test. Statistical analyses were performed using one-way ANOVA followed, when significant, (P < 0.05) by Fisher’s LSD posthoc test. Results are expressed as mean ± SEM.