Melanopsin Regulates Both Sleep-Promoting and Arousal-Promoting Responses to Light

Previous studies have shown that sleep induction in melanopsin knockout mice (Opn4^-/-) in response to nocturnal white light stimuli is attenuated [4–6]. These data suggest a critical role for melanopsin in the regulation of sleep. To investigate this further, we studied behaviourally-defined sleep induction in response to different wavelengths in C57BL/6 mice. To prevent confounds from animals already being asleep, studies were performed at ZT14 when sleep pressure is low and animals show highest activity levels. In mice, the maximum absorption of rod opsin peaks at 498 nm [22] with M-cone opsin at 508 nm [23], S-cone opsin around 360 nm [24], and melanopsin around 480 nm [3]. We studied sleep onset in response to a 1 h isoquantal light exposure at ZT14 in wildtype C57BL/6 mice. We used three different wavelength light stimuli—violet (405 nm), blue (470 nm), and green (530 nm) (S1A Fig) to produce differential activation of UVS cones, melanopsin, and rods/MWS cones (S1B Fig). Given existing data on the role of melanopsin in sleep regulation, we predicted that blue light would result in the most rapid sleep induction, as this wavelength most closely corresponds to the peak sensitivity of melanopsin (~480 nm).

All three wavelengths of light resulted in sleep induction at ZT14, but with different latencies (Fig 1A and 1B). Green light produced a very rapid sleep onset, which was observed almost immediately after light onset (2 ± 1.33 min). By contrast, both violet and blue light showed delayed sleep onset (Fig 1B). Contrary to our prediction, sleep onset in response to blue light was significantly delayed (17. 5 ± 1.64 min) compared to either violet (7.5 ± 2.5 min) or green light exposure (S1, S2, S3 and S4 Videos). This delayed sleep induction under blue light resulted in a reduction of total sleep duration during the 1 h light pulse compared to either violet or green stimuli (Fig 1C). To confirm that this effect was consistent at different times, sleep induction in response to blue and green light was also studied at ZT22. A comparable delay in sleep induction in response to blue light was also observed at this later time (S2 Fig), although differences were not as dramatic. This is most likely due to preceding differences in sleep/wake behaviour during the nocturnal active phase.

To investigate whether the different effects of blue light on sleep and light aversion are mediated via melanopsin, we then performed the same experiments in melanopsin knockout (Opn4^-/-) mice, on a congenic C57BL/6 background, using violet, blue and green light (Fig 2A–2C). When compared with wildtype responses, sleep onset in Opn4^-/- mice was significantly advanced under blue light (Fig 2D), whereas under violet and green light it was significantly delayed. Sleep duration was reduced in response to violet and green light in Opn4^-/- mice, but unaffected in response to blue light (Fig 2E). Consistent with previous findings, delayed sleep induction and total sleep duration in response to white light were observed—comparable to the green light condition (S3A–S3C Fig). Overall, when responses to different wavelengths were compared in Opn4^-/- mice, no significant differences were evident (S3D–S3F Fig). These results show that melanopsin is necessary for the acute wavelength-dependent effects of light on sleep.

To determine why sleep onset was delayed in response to blue light, we studied behavioural light aversion to investigate if different wavelengths of light were associated with increased anxiety. Behavioural light aversion was measured using the light/dark box at ZT14, comparing the time spent in the hidden (dark) zone of the apparatus when the transparent zone of the box was illuminated with different wavelengths (Fig 3A). Data were analysed based upon the initial response to the test arena (first entry to dark zone and time in dark zone during the first minute of the trial) and over the whole course of the trial (time in the dark zone from 2–10 minutes). A control condition in which both sides of the apparatus were unlit (dark) was also used.

The latency from placing the mouse into the box until the first entry to the dark zone was significantly shorter under blue light (10.98 s ± 1.68) compared to green light (46.52 s ± 4.99), violet (46.47 s ± 4.42), and control (22.95 s ± 6.45) conditions (S4 Fig). During the first minute of testing, mice under blue light spent more time in the hidden zone than under other conditions (Fig 3B). By contrast, during the first minute of testing, mice under green or violet light remained in the illuminated area significantly longer than either the blue light or control group, suggesting that their exploration of this novel environment was not inhibited by light. Over the whole trial, mice spent significantly more time (87.5 ± 2.8%) in the hidden zone under blue light compared to either green or violet light, as well as compared to control conditions (Fig 3C). Under violet or green light, mice spent slightly more than 50% of their time in the dark zone (green 64.15 ± 1.76%, violet 51.91 ± 1.59%) and did not differ from that of control condition where the apparatus was unlit (Fig 3C; control 66.54 ± 2.68%). Overall, these data demonstrate that mice find blue light aversive, both immediately within the first minute and over the 10 min timecourse of the experiment. This aversive effect of light may relate to the delay in sleep onset observed under blue light. By contrast, violet or green light does not evoke an aversive response, and over the first minute of the trial may result in novelty-evoked exploratory behaviour. In the home cage setting, this may in turn relate to the rapid sleep induction in response to green light.

We subsequently compared behavioural light aversion in Opn4^-/- mice under blue and green light conditions to determine if melanopsin plays a role in these responses. As we previously found no difference in light aversion between violet and green light (Fig 3B and 3C), violet light was not studied. We found that in comparison to wildtype controls, Opn4^-/- mice showed reduced light aversion to blue light over the whole trial and no difference to green light except in the first minute of the test (Fig 3D and 3E). In the first minute of light exposure (Fig 3D), in comparison to wildtype animals, Opn4^-/- mice spent less time in the hidden zone under blue light, but more time in the hidden zone under green light. Rather than an increase in anxiety, this may simply reflect no preference for either zone (as this was not significantly different to chance), unlike the increased preference for the lit zone seen in wildtype animals. Over the remaining test period (Fig 3E), the reduced aversion to blue light in Opn4^-/- mice was found to persist. However, the difference between wildtype and Opn4^-/- mice in response to green light was not sustained over this period. When light aversion responses to blue and green light were compared in Opn4^-/- mice, no effect of wavelength was observed (Fig 3D and 3E). These data are consistent with the effects of light on sleep in Opn4^-/- mice. The reduced aversive response to blue light in melanopsin-deficient animals is consistent with an advanced sleep onset, whereas the enhanced aversive response to green light at the beginning of the stimulus is consistent with delayed sleep onset.

To determine if the different wavelengths of light result in different effects at a molecular level, we measured the expression of the immediate early gene Fos, and the light-regulated clock genes Per1 and Per2 following a 1 h light pulse in the SCN and adrenal gland (Fig 4A). Consistent with previous studies, all three wavelengths evoked increased expression of Fos, Per1, and Per2 in the SCN. Different wavelengths resulted in differential levels of induction of all three genes in the SCN, with blue light evoking significantly greater increases compared to either violet or green light. Violet and green light resulted in comparable levels of gene induction in the SCN.

Nocturnal light exposure has previously been shown to result in induction of immediate early genes in the adrenal gland [19,25]. We reasoned that if light input is relayed from the retina to the adrenal gland via the SCN, light-induced changes in the adrenal should reflect those observed in the SCN. In the adrenal gland, we found that responses to blue light were significantly greater than to either green or violet light. Both blue and green light induced Per1 expression, with significantly greater Per1 induction compared to either green or violet light. Fos and Per2 expression were also induced by blue light, but not violet or green light.

Given the high-amplitude changes in gene expression in response to blue light, we next investigated whether these responses were melanopsin-dependent. We compared gene expression in the SCN and adrenal gland between wildtype and Opn4^-/- mice exposed to blue light (Fig 4B). As the previously observed molecular responses to violet and green light were of much lower amplitude, these were not studied, as it would be extremely difficult to attain sufficient statistical power to resolve any difference. In response to blue light, Opn4^-/- mice showed light-induced gene expression for all three genes in the SCN. However, these responses were attenuated compared to wildtype animals. In the adrenal gland, only Per1 and Per2 were increased in comparison to dark controls in response to light. All three genes showed attenuated responses in comparison to wildtype mice. Together, these results show that molecular responses to blue light are impaired in Opn4^-/- mice at the level of both SCN and adrenal gland. The effects of light on SCN and adrenal, and the high levels of gene induction in response to blue light in particular, suggested that the behavioural responses to blue light may be accompanied by changes in plasma corticosterone.

Previous studies have shown that acute white light exposure at ZT16 produces an increase in plasma corticosterone levels related to light-induced adrenal Per1 expression [19,20,25]. To determine if the observed sleep, behavioural, and molecular responses to blue light are accompanied by a rise in plasma corticosterone, we exposed wildtype C57BL/6 mice to 1 h of violet, blue or green light. All three wavelengths produced significantly elevated plasma corticosterone levels compared with dark-exposed controls (Fig 5A). However, blue light resulted in significantly higher corticosterone levels compared to violet, green or dark conditions. Responses to green light were significantly lower compared to either violet or blue light.

To investigate whether the effects of different wavelengths of light on Opn4^-/- mice are reflected at the level of plasma corticosterone, we measured plasma corticosterone levels in Opn4^-/- mice exposed to either blue or green light (Fig 5B). Opn4^-/- mice still show an elevation of plasma corticosterone levels in response to light, but responded differently in comparison to wildtype mice. Responses to blue light were significantly attenuated in Opn4^-/- animals, although responses to green light were enhanced. As a result of these changes, no significant differences in corticosterone levels between blue and green stimuli were apparent in Opn4^-/- mice. This finding is consistent with the loss of wavelength-dependent effects of light on sleep induction and behavioural light aversion in melanopsin-deficient animals. Surprisingly, baseline levels of corticosterone at ZT14 were slightly elevated in Opn4^-/- mice compared to wildtype controls.

To determine if blue and green light exert different effects at the level of retinorecipient targets, we studied responses to these wavelengths at the level of the SCN and VLPO (Fig 6). Mice were exposed to blue and green light stimuli and after 30 min SCN and VLPO were collected for gene expression analysis using quantitative PCR (qPCR). Consistent with our previous findings (Fig 4A), we found that blue light produced a larger Fos induction in the SCN than green light. However, in the VLPO we found a greater response to green light than blue light. As the sleep active neurons of the VLPO are galaninergic [26], we also studied expression of Gal in both SCN and VLPO. In the VLPO (but not the SCN), we found that Fos induction was accompanied by increased Gal expression. These data suggest that the different behavioural effects of blue and green light are mirrored by molecular responses at the level of the SCN and VLPO, suggesting that different pathways may mediate the wake-promoting and sleep-promoting effects of light on behaviour.

It is unclear whether the inhibitory effects of blue light on sleep are mediated via elevated plasma corticosterone or whether elevated corticosterone levels are a downstream consequence of the heightened state of arousal produced by this stimulus. To investigate this question, we blocked the effects of corticosterone via systemic injection of the glucocorticoid receptor antagonist mifepristone (RU-486) in wildtype C57BL/6 mice at ZT12. These mice were then exposed to a blue light pulse at ZT14. In comparison to vehicle-treated animals, mifepristone-treated mice showed elevated basal sleep (S5 Fig). However, they did show significantly enhanced sleep onset in comparison to vehicle treated animals (two-way ANOVA for treatment and time, treatment F_(1.98) = 16.473, p ≤ 0.001, posthoc Tukey drug versus control first 10 min pulse p = 0.003, 10–20 min p ≤ 0.001, 20–30 min pulse p = 0.01). These findings show that the effects of blue light on sleep induction can be attenuated by blocking the effects of elevated plasma corticosterone.

All experiments were conducted in accordance with the Home Office (UK) regulations, under the Animals (Scientific Procedures) Act of 1986. All procedures were reviewed by the clinical medicine animal care and ethical review body (AWERB), and were conducted under PPL 30/2812, under PILs IF656800A and I869292DB. Wild type male C57BL/6 mice (3/7 mo old) and melanopsin knockout (Opn4^-/-) on a C57BL/6 background (ten generations of backcrossing) were used throughout. Before all experiments, mice were singly housed under a 12:12 light dark cycle with white light 250 lux when measured at the bottom of the cage, with food and water available ad libitum. Animals were housed in specific pathogen free (SPF) conditions, and the only reported positives on health screening over the entire time course of these studies were for Helicobacter hepaticus and Entamoeba sp (Envigo, Alconbury UK).

To assess behavioural responses (including activity, sleep induction/duration and light aversion) to different wavelengths of light we monitored home cage behaviour with miniature near infrared (NIR) video cameras (Sentient Mini night vision CCTV camera, Maplin, UK) mounted above each cage. Behaviour was tracked for 3 h and analysed in real time using ANY-maze (Version 4.98, Stoelting, US). Sleep onset was determined from episodes of immobility of 40 s or longer, and data were grouped into 10 min bins. This method has previously been validated, showing that the extended immobility correlates very highly (correlation coefficient = 0.95) with EEG/EMG under both baseline conditions as well as following administration of pharmacological agents. Moreover, we have also shown that this method is capable of accurately identifying the dose-dependent effects of sedatives, stimulants and light [52].

To estimate the effect of monochromatic light pulses on immobility induction we housed Opn4^-/- and WT mice under a LD cycle (white light 250 lux). During video tracking, all mice have been kept in polypropylene cages, and the top of the cage was open so that the entire cage could be used as open field arena. The water bottle as well as the feeder (a drawer) were installed outside of the cage. White light as well as monochromatic light (530 nm, 470 nm, and 405 nm) were provided using cool white LEDs (Luxeon Star, Quadica Developments Inc, Brantford, Ontario). These wavelengths were used to produce spectral separation between stimuli, due to the overlapping photopigment absorption curves. The photon flux of all three monochromatic LEDs was isoquantal at 14.9 log quanta when measured at the bottom in the middle of the cage. The wavelength as well as the photon flux were measured using a calibrated Ocean Optics spectrometer. For acclimatisation in the new cages mice were kept for 5 days under a white light LD cycle (~250 lux). A 1 h monochromatic light pulse was administered at ZT14. Mice which exhibited behaviourally-defined sleep prior to the light pulse were excluded from the analysis, as light was found to result in waking (rather than sleep) under such conditions. As such, unlike circadian responses to light, the behavioural state of animals prior to light exposure was found to be critical when considering sleep induction. Therefore, only mice which showed activity 10 min before light pulse were considered for sleep analysis. Moreover, as baseline sleep is lowest at the start of the dark phase, sleep induction light pulses were administered at ZT14 (rather than later in the dark phase) to coincide with this period of greatest waking. Sleep latency for each animal was determined as the time during which the first bout of behaviourally-defined sleep occurred. Total sleep was defined as the total period of behaviourally-defined sleep during the 1 h light pulse.

Light aversion was measured in naive mice using a dark chamber light chamber paradigm as previously reported [53]. Behavioural testing was performed for 10 min at ZT14 and analysed using ANYmaze software. To estimate the difference in aversion behaviour induced by monochromatic light, the testing was performed with 470 nm, 405 nm, and 530 nm. Control studies were performed with no light. Each mouse was gently placed into the illuminated chamber of the box facing the portal to the dark chamber. A video tracking system (see experimental setup for behavioural test) was used to monitor behaviour and to quantify the total time that mice spent in the illuminated versus the dark chamber during a 10 min trial. The chambers were thoroughly cleaned between animals.

For ex vivo phase analysis we exposed wildtype and Opn4^-/- to a 1 h light pulse and collected trunk blood after 30–40 min, the time of peak corticosterone induction [19,25]. Dark-treated mice were used as a control. A parallel set of mice was equally treated with light pulse, and one hour after light exposure (the expected peak of Fos and Per1) SCN and adrenal gland were collected and used for Per1, Per2, and Fos gene expression analysis using qPCR.

After 1 hr light pulse, animals were killed under dim light by cervical dislocation. To prevent any photic stimulation to the SCN, eyes were immediately removed. Removed brains were placed into a brain matrix (Kent Scientific, Torrington CT, US). Two skin graft blades (Kent Scientific, Torrington CT, US) one positioned at Bregma −0.10 mm and the second at Bregma −1.10 caudal from the first, were used to cut a 1 mm thick brain slice. SCN punches were collected from the brain slices by using a sample corer (1 mm internal diameter, Fine Science Tools GmbH, Heidelberg, Germany). SCN punches were stored at −80°C prior to RNA extraction.

Wildtype mice were exposed to a 1hr green light pulse before SCN and VLPO punches were collected 30–40 min after the end of the light pulse. This time point is characterised by a peak of Fos in the SCN [5,54,55]. To prevent any photic stimulation to the SCN eyes were immediately removed. Three skin graft blades were used (Kent Scientific, Torrington CT, USA) one positioned at Bregma −0.10mm and the second at Bregma −1.10 caudal from the first for the SCN (as described above), and the third 1 mm before 1.10 mm for VLPO. Two 1 mm thick brain slices were then dissected. SCN as well VLPO punches were collected as described previously [5,55] using a sample corer (1mm internal diameter, Fine Science Tools GmbH, Heidelberg, Germany). Samples were stored at −80°C prior RNA extraction.

Total RNA of SCN and VLPO punches was extracted using the microRNeasy column method (QIAGEN, Hilden, Germany). Total RNA of adrenal gland was extracted using the miniRNeasy column method (QIAGEN, Hilden, Germany). The quantity of RNA was estimated using Nanodrop1000 (Thermo Fisher Scientific, Waltham, MA USA).

cDNA was synthesised from RNA samples with a qScript cDNA synthesis kit (Quanta Biosciences, Gaithersburg, MD). qPCR was performed with Sybr green and SDS7700 thermal cycler (Applied Biosystems, Foster City, CA). Relative quantification of transcript levels was conducted as described previously [56]. The geometric mean of three housekeeping genes Actb, Gapdh, and Arbp was used for normalisation. Galanin was used as a positive control as described in previous studies [6]. Expression levels were normalised to control (dark) values. Per1, Per2, Fos and Gal expression were estimated using following primers: Per1 forward AGTTCCTGACCAAGCCTCGTTAG, Per1 reverse CCTGCCCTCTGCTTGTCATC, Per2 forward GGGGTGAGATTCGTCATTGAACTTG, Per2 reverse AGGACATTGGCACACTGGAAAGAG, Fos forward ATCGGCAGAAGGGGAAAGTAG, Fos reverse GCAACGCAGACTTCTCATCTTCAAG, Gal forward ATGCCTGCAAAGGAGAAGAGAGGT, Gal reverse TCTGTGGTTGTCAATGGCATGTGG.

Trunk blood was collected in microfuge tube containing anticoagulant EDTA (10 ul for 500 ul blood), kept for 10 min on ice and centrifuged. Obtained plasma was stored at −20°C until corticosterone measurement. Corticosterone was measured from 1:10 diluted samples using an ELISA kit (Assaypro LLC St. Charles, MO) according to manufacturer’s instructions.

To block glucocorticoid receptor mediated effects, we injected mice with the glucocorticoid antagonist RU-486 (Sigma Aldrich) at ZT12 and started measuring behaviour from ZT13 until ZT15. RU-486 was administered as described previously [57]. Briefly, RU-486 was dissolved in 20% dimethyl sulfoxide (DMSO)/80% polythelyne glucol (PEG)-300. Drug dosage (100 mg/kg body weight) was based on the previous studies [57,58]. As control the same amount of (1 ml/kg body weight) of vehicle (saline diluted in 20%DMSO/80%(PEG)-300) was injected.

To determine the light available to the different photoreceptors of the mouse retina for the different wavelength stimuli used in this study as well as white light sources, the Rodent Toolbox v1.0 was used [15]. For the 405 nm, 470 nm, and 530nm LEDs used in this study, spectral power distributions were measured using a calibrated Ocean Optics spectrometer. For white light sources, standard illuminant functions were used from the Toolbox. For daylight and twilight data, previously published data were used [59]. Comparable results for daylight were obtained using the D65 standard illuminant function. Data were expressed in alpha-opic lux. However, as it is not possible to directly compare different photoreceptor channels, these were expressed as relative to the total alpha-opic lux across all four channels.

Results are presented as mean ± SEM. Statistical analysis was performed using SPSS Statistics 22.0. All data for statistical analysis were first checked for normal distribution using Kolmogorov Smirnov test. Non-normally distributed data were log transformed prior to statistical analysis to ensure normality and homogeneity of variance. Statistical significance of group comparisons was tested with one-way ANOVA and posthoc Tukey. Multivariate experiments such as genotype and treatment were analysed using two-way ANOVA with Tukey’s posthoc analysis. Time course experiments such as across multiple hours or min were analysed using rep. ANOVA. Assessment of relative photoreceptor activity under the 405 nm, 470 nm, and 530 nm wavelengths has been performed using the previously published rodent toolbox [15].

All relevant data are within the paper and its Supporting Information files.