Neuropsin (OPN5)-mediated photoentrainment of local circadian oscillators in mammalian retina and cornea

Significance

The behavioral circadian rhythms of mammals are synchronized to light/dark cycles through rods, cones, and melanopsin (OPN4)-expressing, intrinsically photosensitive ganglion cells in the retina. The molecular circadian rhythms in the mammalian retina are themselves synchronized to light/dark signals. We show here that this retinal photoentrainment, in an ex vivo setting, requires neuropsin (OPN5), an orphan opsin in mammals. Remarkably, the circadian clocks in the cornea are also photoentrained ex vivo in an OPN5-dependent manner.

Abstract

The molecular circadian clocks in the mammalian retina are locally synchronized by environmental light cycles independent of the suprachiasmatic nuclei (SCN) in the brain. Unexpectedly, this entrainment does not require rods, cones, or melanopsin (OPN4), possibly suggesting the involvement of another retinal photopigment. Here, we show that the ex vivo mouse retinal rhythm is most sensitive to short-wavelength light but that this photoentrainment requires neither the short-wavelength–sensitive cone pigment [S-pigment or cone opsin (OPN1SW)] nor encephalopsin (OPN3). However, retinas lacking neuropsin (OPN5) fail to photoentrain, even though other visual functions appear largely normal. Initial evidence suggests that OPN5 is expressed in select retinal ganglion cells. Remarkably, the mouse corneal circadian rhythm is also photoentrainable ex vivo, and this photoentrainment likewise requires OPN5. Our findings reveal a light-sensing function for mammalian OPN5, until now an orphan opsin.

Most mammalian tissues contain autonomous circadian clocks that are synchronized by the suprachiasmatic nuclei (SCN) in the brain (1). The SCN clock itself is entrained by external light cycles through retinal rods, cones, and melanopsin (OPN4)-expressing, intrinsically photosensitive retinal ganglion cells (ipRGCs) (2, 3). The retina also manifests a local circadian clock, which regulates many important functions, such as photoreceptor disk shedding, photoreceptor gap-junction coupling, and neurotransmitter release (4–6). Surprisingly, local retinal photoentrainment does not require the SCN, and it also does not require rods, cones, or ipRGCs (7, 8). Thus, the rd1/rd1;Opn4^−/− mouse retina, which has lost essentially all rods and cones due to degeneration and also has an ablated Opn4 gene (3), remains synchronized to light/dark cycles both in vivo and ex vivo (7).

To determine the photopigment(s) responsible for local circadian entrainment in the retina, we took a candidate gene approach. Because some cone nuclei may persist in degenerate rd1/rd1 retinas (9), and murine short-wavelength–sensitive cone opsin (OPN1SW) has been reported to be present in the ganglion cell layer (10), we tested the necessity of this pigment for local retinal circadian photoentrainment. We also tested for the involvement of two orphan pigments, encephalopsin (OPN3) (11) and neuropsin (OPN5) (12), both of which are expressed in mammalian retina and, when expressed heterologously, form light-sensitive pigments that activate G proteins (13–17). The function of OPN3 in mammals is unknown despite its widespread expression in neural tissues (18). OPN5 appears to be a deep-brain photopigment in the hypothalamus of birds and is thought to contribute to seasonal reproduction (19–22); it has been immunolocalized to the mammalian inner retina (13, 16) (SI Text); however, to date, no retinal function for this mammalian pigment has been identified. We did not examine two other pigments, retinal G protein-coupled receptor (RGR) opsin (23) and peropsin (RRH) (24). RGR opsin participates in retinoid turnover (25, 26), whereas RRH is expressed principally in the retinal pigment epithelium (24), a cell layer absent in the photoentrainable ex vivo retina preparation (7).

SI Text

Antibodies Against OPN5.

An antibody against the N terminus of mouse OPN5 has been reported to label inner retinal neurons by immunohistochemistry on retinal sections (13). We obtained an initial aliquot of the antibody and verified the expression of OPN5 in the majority of cells in the ganglion cell layer, as well as in the inner part of the inner nuclear layer of WT retinas. The immunosignals were absent in Opn5^−/− line 1 retinal sections that were processed in parallel. To confirm our results, we subsequently obtained another aliquot (owing to the exhaustion of the initial aliquot) of the same antibody, but from a separately purified stock. With identical immunohistochemical procedures, however, we observed a much weaker labeling of inner retinal neurons and also enhanced nonspecific staining of, for example, photoreceptors and the outer portion of the inner nuclear layer. More importantly, these immunosignals persisted even in Opn5^−/− line 1 retinas, in contrast to the earlier immunohistochemical results described above and to the absence of the 5′ Opn5 transcript in Opn5^−/− line 1 eyes as demonstrated by RT-PCR experiments (SI Materials and Methods and Fig. S1E). Because of the inexplicable inconsistency in these immunostaining results, we did not include them in the current paper.

Fig. S1.

Generation of Opn3^−/− and Opn5^−/− mice. (A) Schematic of the Opn3^flox/flox targeting construct (modified from the International Knockout Mouse Consortium). Through homologous recombination at the 5′ and 3′ arms, a flippase recognition target (FRT)-lacZ cassette is inserted before exon 2 of the Opn3 gene, which is also flanked by loxP sites. Cre-dependent recombination at these sites leads to the deletion of exon 2, which encodes three transmembrane domains of the OPN3 protein. Dark bars indicate exons. En2SA, mouse Engrailed2 intron splice acceptor; hActbP, human β-actin promoter; IRES, internal ribosome entry site; lacZ, β-gal cDNA; neo, neomycin-resistance gene; pA, SV40 polyadenylation site. (B) Schematic of the Opn5^−/− line 1 targeting construct. Homologous recombination at the 5′ and 3′ arms replaces part of exon 3 and the entire exon 4 of the Opn5 gene with a tau-lacZ-neo cassette, which disrupts at least two transmembrane domains of the OPN5 protein. hsvP, herpes simplex virus promoter; pA, polyadenylation site; pgkP, mouse phosphoglycerate kinase 1 promoter; tau, cDNA of microtubule-associated protein tau; tk, thymidine kinase. (C) Schematic of the Opn5 ^flox/flox targeting construct (modified from the International Knockout Mouse Consortium) for generating Opn5^−/− line 2. Through homologous recombination at the 5′ and 3′ arms, an FRT-lacZ cassette is inserted before exon 4 of the Opn5 gene, which is also flanked by loxP sites. Cre-dependent recombination at these sites leads to the deletion of exon 4, which encodes two transmembrane domains of the OPN5 protein. pA, SV40 polyadenylation site. (D) RT-PCR experiment on total eye RNA to confirm the absence of Opn3 transcripts in Opn3^−/− mice. The control (Ctrl) is pure water. (E) RT-PCR experiment on total eye RNA to confirm the absence of Opn5 transcripts in Opn5^−/− line 1 mice. PCR-spanning exons 1–4 detects two Opn5 isoforms as previously reported (12): a protein-coding isoform (lower band) and a noncoding isoform (upper band) that contains an unspliced exon. A 3′ transcript of Opn5 (exons 5–7, band is dim in WT) is up-regulated in Opn5^−/−, possibly due to enhancement by the knocked-in promoter (SI Materials and Methods). The housekeeping gene β-actin serves as a positive control for RT. The negative control for PCR is from the RNA extraction buffer processed in identical ways but without tissues. (F) RT-PCR experiment on total eye RNA to confirm the absence of Opn5 transcripts in Opn5^−/− line 2 mice. PCR-spanning exons 1–4 detect two Opn5 isoforms as in E. A 3′ transcript of Opn5 (exons 5–7) is still present in Opn5^−/− line 2, possibly due to the splicing of exon 3 into exon 5 (SI Materials and Methods). The housekeeping gene β-actin serves as a positive control for RT. The control is pure water.

Separately, we have obtained another OPN5 antibody, generated against a different N-terminal epitope (16). This antibody has been shown to label fewer cells in the ganglion cell layer and the inner nuclear layer (16). Nevertheless, we were unable to observe any specific immunosignal on WT and Opn5^−/− line 1 retinal sections with the immunohistochemical conditions specified by the provider. Therefore, once again, we did not include the results in the current paper.

Results

Wavelength Dependence of ex Vivo Retinal Circadian Photoentrainment.

We first examined the wavelength dependence of retinal circadian photoentrainment with Per2::Luciferase mouse retinas (27). The luciferase luminescence from these retinas comes predominantly from the inner retina (28, 29), although autonomous rhythms have also been found in the outer retina (28). Pairs of retinas from these mice were exposed ex vivo to 9-h/15-h light/dark cycles for 4 d, with the two retinas in each pair subjected to opposite-phase light/dark entrainment (7) (SI Materials and Methods). Immediately after cyclic light exposure, the luciferase luminescence in each retina was measured for 4 d in continuous darkness to determine the circadian phase. In previous work with white light (7), paired retinas under these conditions showed opposite phases, with peak luminescence occurring at roughly 4 h after the respective light-to-dark transition. To determine wavelength dependence, we replaced white light with monochromatic light-emitting–diode light at 370 nm, 417 nm, 475 nm, 530 nm, or 628 nm; peak photon flux at each of the last four wavelengths was identical (9 × 10¹³ photons cm⁻²⋅s⁻¹) but was 10-fold less (9 × 10¹² photons cm⁻²⋅s⁻¹) at 370 nm to avoid tissue toxicity at higher UV intensities. After 4 d of entrainment, the paired retinas exposed to 370-nm light/dark cycles had stable, nearly opposite entrainment phases, as did retinas exposed to 417-nm light/dark cycles (Fig. 1A). In contrast, the paired retinas exposed to 475-nm light/dark cycles showed only a moderate, ∼6-h difference in their phases. Paired retinas exposed to 530-nm or 628-nm light/dark cycles did not photoentrain, with each having phases identical to retinas kept in continuous darkness (Fig. 1A, rightmost vertical bar). Thus, circadian photoentrainment in ex vivo retinas is most sensitive to UV-A and violet light.

Fig. 1.

Sensitivity of retinal circadian photoentrainment to short-wavelength light and lack of effect of the Opn1sw^−/− genotype. (A) Wavelength dependence of ex vivo photoentrainment. Pairs of retinas from Per2::Luciferase mice were cultured in light/dark cycles (vertical colored bars) or in continuous darkness (full gray bar) for 4 d. Points represent mean phase (peak of bioluminescence) on the day following light/dark exposure ± 1 SEM for 370-nm and 417-nm light (n = 6 retinal pairs each; P < 0.001 for 0°/180° comparison, 0°/dark control comparison, and 180°/dark control comparison, one-way ANOVA, Tukey post hoc), 475-nm light (n = 7 pairs; P = 0.038 for 0°/180° comparison, P = 0.007 for 180°/dark comparison, and P > 0.5 for 0°/dark comparison), and 530-nm and 628-nm light (n = 5 pairs each; all comparisons not significant in post hoc analyses). *The 370-nm light was at 9 × 10¹² photons cm⁻²⋅s⁻¹, but all other wavelengths were at 9 × 10¹³ photons cm⁻²⋅s⁻¹. (B) Phase-response relation of Per2::Luciferase retinal rhythms after exposure to a 3-h pulse (1.5 × 10¹⁵ photons cm⁻²⋅s⁻¹) of 417-nm light (purple) or 475-nm light (blue), together with dark-handling controls (gray). “Pulse phase” denotes the phase of the retinas at the time of the pulse. Points represent mean ± SEM and are connected by straight lines for a given wavelength (n ≥ 5 for each point). (C, Upper) Dark-recorded bioluminescence traces (with background luminescence subtracted) from a pair of WT;Per2::Luciferase or Opn1sw^−/−;Per2::Luciferase retinas after 4 d of culturing at the 0° (blue) or 180° (red) position of the light/dark photoentrainment apparatus (white, 5 W⋅m⁻²). (C, Lower) Time of peak bioluminescence of retinas as in A. Bars indicate periods of darkness (gray) and light exposure (white) in entrainment cycles. WT, n = 6 retinal pairs; Opn1sw^−/−, n = 7 retinal pairs. (D) Intensity-response relation of WT;Per2::Luciferase (black) and Opn1sw^−/−;Per2::Luciferase (red) retinal rhythms after exposure to a 3-h pulse of 417-nm light of various intensities at CT 9 and CT 20. Points represent mean ± SEM and are connected by straight lines for a given genotype (n ≥ 5 for each point).

To confirm this spectral sensitivity, we tested retinas for acute phase shifts under different wavelengths. Per2::Luciferase retinas were cultured in continuous darkness. At different phases during their free-running circadian rhythms, we administered equal quanta of a 3-h light pulse at 417 nm or 475 nm and measured the resulting phase delay or phase advance in the rhythm to generate a phase-response curve. At all phases, 417-nm light triggered phase shifts that were consistently larger than those phase shifts triggered by equal quanta of 475-nm light (Fig. 1B), in agreement with the photoentrainment experiment.

OPN1SW Pigment Is Not Required for ex Vivo Retinal Photoentrainment.

Because photoentrainment was most sensitive to UV-A and violet light, OPN1SW [wavelength of maximal absorption (λ_max) = 360 nm] could be responsible. We therefore tested retinas from mice lacking OPN1SW (Opn1sw^−/−;Per2::Luciferase) (30). Retinas from these animals photoentrained to white light just as WT (i.e., Per2::Luciferase) did (Fig. 1C). Furthermore, in response to an acute violet light pulse (3 h, 417 nm) of various intensities administered at circadian time (CT) 9 or CT 20, Opn1sw^−/−;Per2::Luciferase retinas phase-shifted with sensitivity identical to the sensitivity of WT retinas (Fig. 1D). The quantitatively similar photosensitivities exhibited by WT and Opn1sw^−/− retinas to violet light demonstrate that OPN1SW is not required for retinal photoentrainment ex vivo.

OPN5, but Not OPN3, Is Required for ex Vivo Retinal Photoentrainment.

Mouse OPN5 has a λ_max in the UV-A region (380 nm) when heterologously expressed (13, 16). For OPN3, heterologously expressed fish and insect homologs have a λ_max of 460 nm and 500 nm, respectively (14), but the λ_max of mammalian OPN3 is still unknown. We examined both OPN3 and OPN5 with the gene KO approach. We generated Opn3^−/− and Opn5^−/− (“Opn5^−/− line 1”) mice by homologous recombination (SI Materials and Methods and Fig. S1 A and B). We bred these lines into the Per2::Luciferase line to perform ex vivo photoentrainment experiments (retinas cultured for 4 d under 9-h/15-h white light/dark cycles). We found that the Opn3^−/−;Per2::Luciferase retinas photoentrained normally (despite a substantially weaker circadian luminescence amplitude), but Opn5^−/−;Per2::Luciferase retinas completely failed to entrain, maintaining phases identical to WT controls kept in continuous darkness (Fig. 2). We also generated an Opn5^flox/flox mouse line (SI Materials and Methods and Fig. S1C), which allowed us to ablate OPN5 in all tissues (“Opn5^−/− line 2”) after crossing it with a ROSA-FLP line and an EIIa-Cre line. Retinas from transheterozygotes of the two Opn5^−/− lines, in a Per2::Luciferase background, similarly failed to entrain to light/dark cycles. However, retinas from heterozygous Opn5^+/−;Per2::Luciferase littermates of the two Opn5^−/− lines entrained normally (Fig. S2). Thus, both Opn5^−/− lines appeared to behave identically. All experiments below (except for X-Gal labeling) were done only with Opn5^−/− line 1.

Fig. 2.

Disruption of photoentrainment in Opn5^−/−, but not Opn3^−/−, retinas. (A) Dark-recorded bioluminescence traces (with background luminescence subtracted) from a pair of WT;Per2::Luciferase, Opn3^−/−;Per2::Luciferase or Opn5^−/−;Per2::Luciferase retinas after 4 d of culturing at the 0° (blue) or 180° (red) position of the light/dark photoentrainment apparatus. The WT data are from the same experiment as shown in Fig. 1C. (B) Time of peak bioluminescence of retinas (points, mean phase ± SEM) photoentrained at 0° (blue) and 180° (red) positions. Controls (DARK) are retinas maintained in continuous darkness. Bars indicate periods of darkness (gray) and light exposure (white; 5 W⋅m⁻²) in entrainment cycles. WT, n = 7 retinal pairs; Opn3^−/−, n = 6 retinal pairs; Opn5^−/− line 1 (main text), n = 7 retinal pairs.

Fig. S2.

Effect of Opn5 genotypes on retinal and corneal photoentrainment. (Left) Dark-recorded bioluminescence traces from a pair of transheterozygous Opn5^−/−;Per2::Luciferase retinas (Top) and corneas (Bottom) after 4 d of culturing at the 0° (blue) or 180° (red) position of the light/dark photoentrainment apparatus (n = 4 each). Background luminescence is already subtracted. (Middle) Similar recordings, but with Opn5^+/−;Per2::Luciferase line 1 retinas and corneas. (Right) Similar recordings, but with Opn5^+/−;Per2::Luciferase line 2 retinas and corneas (n = 2 for individual heterozygous lines).

To determine if the nonphotoentrainable phenotype arose secondarily from possibly widespread retinal dysfunction, we performed basic histological, electrophysiological, and visual function tests (SI Materials and Methods) on Opn5^−/−;Per2::Luciferase animals. Their retinas showed normal histology, with normal patterns of expression of rod/cone opsins and OPN4 (Fig. S3). Opn5^−/− and WT mice showed similar electroretinogram responses to identical light flashes in both dark- and light-adapted states (Fig. 3A) (when assayed between 6 and 12 h after lights-on in a 12-h/12-h light/dark cycle; we did not perform measurements at other CTs). Opn5^−/−;Per2::Luciferase mice also had a normal optokinetic tracking reflex to a rotating grating of different spatial frequencies, unlike rd1/rd1;Opn4^−/− mice, which showed a near-complete lack of visual tracking (Fig. 3B). Finally, the circadian rhythm of wheel-running activity of Opn5^−/− mice showed entrainment to 12-h/12-h light/dark cycles, and reentrained to a phase advance in the external light/dark cycles, although more slowly than WT mice (Fig. 3C). When kept in constant darkness, Opn5^−/− and WT mice had virtually identical free-running periods (Opn5^−/−: 23.89 ± 0.04 h, n = 7; WT: 23.91 ± 0.06 h, n = 6; P = 0.762, Student’s t test). Thus, overall, no substantial retinal or circadian dysfunction was observed.

Fig. S3.

Normal expression of rod/cone opsins and OPN4 in Opn5^−/− retina. (A–F) Retinal sections of WT and Opn5^−/− line 1 mice immunostained (green) for rhodopsin, OPN1SW, and medium-wavelength–sensitive cone opsin (OPN1MW), with DAPI (blue) indicating nuclear layers. GCL, ganglion cell layer; INL, inner nuclear layer; IPL, inner plexiform layer; IS, inner segment layer, ONL, outer nuclear layer; OPL, outer plexiform layer; OS, outer segment layer. Immunolabeling of the above pigments in the flat-mount retina similarly did not reveal any ectopic expression of pigments across the entire retina. Similar results were observed in at least three animals. (G and H) Flat-mounts of WT and Opn5^−/− line 1 retinas immunostained for OPN4. Images are stacked confocal images. Similar results were observed in at least three animals.

Fig. 3.

Normal visual functions in Opn5^−/− mice. (A) Dark- and light-adapted electroretinograms (ERGs) on anesthetized WT (blue) and Opn5^−/− (red) mice. For dark-adapted ERGs, a flash intensity of 3 cd⋅m⁻² was administered over a dark background. Light-adapted ERGs consisted of 900-cd⋅m⁻² white light over a background of 340-cd⋅m⁻² white light. Each trace is the average of ∼150 trials. (B) Optokinetic tracking responses of WT (blue), Opn5^−/− (red), and rd1/rd1;Opn4^−/− (black) mice to rotating gratings with different spatial frequencies (0.05–0.5 cycles per degree). Scores were given based on the head movements of the mice, with 0 = no visual tracking, 1 = ambiguous tracking, and 2 = obvious visual tracking. WT, n = 6; Opn5^−/−, n = 7; rd1/rd1;Opn4^−/−, n = 4. (C) Representative actograms of individual WT and Opn5^−/− mice. Actograms are double-plotted so that each horizontal trace represents two consecutive days, with the second day replotted as the first day on the trace beneath it. Periods of darkness are indicated by shaded areas. Black marks represent the number of wheel revolutions in 5-min bins. All Opn5^−/− animals were from line 1.

OPN5-Dependent ex Vivo Corneal Photoentrainment.

The mammalian cornea has a strong circadian rhythm ex vivo (27). We also detected Opn5 transcripts by RT-PCR in both fresh and cultured WT corneas (Fig. S4). Therefore, we tested whether the corneal circadian luciferase rhythm is light-entrainable ex vivo and, indeed, found this to be the case (Fig. 4); the corneal luciferase rhythm was nearly opposite in phase to the luciferase rhythm of the retina (i.e., peaking at subjective dawn rather than subjective dusk). Furthermore, OPN5 is essential for this corneal photoentrainment because this function was abolished ex vivo (Fig. 4) in two different strains of OPN5-KO mice (Opn5^−/−;Per2::Luciferase line 1 and transheterozygous Opn5^−/−;Per2::Luciferase mice obtained from crossing Opn5^−/− lines 1 and 2; Fig. S2). By contrast, corneal photoentrainment persisted in mice lacking OPN1SW (Opn1sw^−/−;Per2::Luciferase) or OPN3 (Opn3^−/−;Per2::Luciferase) (Fig. 4). For comparison, we tested the pituitary gland (another strongly rhythmic tissue) for photoentrainment ex vivo. The circadian rhythm of cultured pituitary glands was not sensitive to light, and we did not detect Opn5 transcripts in this tissue (Fig. S5).

Fig. S4.

Expression of Opn5 transcripts in cornea. Detection of Opn5 transcripts in fresh or ex vivo (10 d in culture) corneas and retinas by RT-PCR. The housekeeping gene Gapdh serves as a positive control for RT. Water is used as a negative control for PCR.

Fig. 4.

Loss of photoentrainment in Opn5^−/− cornea. (A) Dark-recorded bioluminescence traces from a pair of WT, Opn1sw^−/−, Opn3^−/−, or Opn5^−/− line 1 corneas (all in a Per2::Luciferase background) after 4 d of culturing at the 0° (blue) or 180° (red) position of the light/dark photoentrainment apparatus. (B) Time of peak bioluminescence of corneas (points, mean phase ± SEM) photoentrained at 0° (blue) and 180° (red) positions. Controls (DARK) are corneas maintained in continuous darkness. Bars indicate periods of darkness (gray) and light exposure (white, 5 W⋅m⁻²) in entrainment cycles. WT, n = 7 corneal pairs; Opn1sw^−/−, n = 6 corneal pairs; Opn3^−/−, n = 5 corneal pairs; Opn5^−/− line 1, n = 6 corneal pairs.

Fig. S5.

Lack of photoentrainment of pituitary rhythms. (A, Top) Dark-recorded bioluminescence traces from a pair of WT, Opn3^−/−, or Opn5^−/− line 1 (all in Per2::Luciferase background) pituitaries after 4 d of culturing at the 0° (blue) or 180° (red) position of the light/dark photoentrainment apparatus. Background luminescence is already subtracted. (A, Bottom) Collective data are shown, with points representing mean phase (peak of bioluminescence) on the day following light/dark exposure ± 1 SEM. (B, Left) Detection of Opn5 in fresh retina, cornea, pituitary, and liver. (B, Right) Relative abundance of Opn5 RNA compared with Gapdh based on quantitative RT-PCR from fresh mouse retina, cornea, pituitary, and liver (n = 3).

Identity of Cells Expressing OPN5.

In the absence of any validated specific antibody against OPN5 (SI Text), we used Opn5-promoter activity to perform initial localization of Opn5 expression in the retina. The Opn5^−/− line 1 mice carry a knock-in tau-lacZ fusion gene in the Opn5 gene locus (Fig. S1B), allowing us, in principle, to reveal Opn5-expressing cells by X-Gal staining. Puncta of X-Gal precipitate were found in about 4,800 cells in the retina, evenly spaced in the ganglion cell layer (Fig. 5A, Right). Colocalization experiments comparing X-Gal labeling and immunohistochemistry with a ganglion-cell marker, retinal binding protein with multiple splicing (RBPMS) (31, 32) (Fig. 5B), suggested that the X-Gal–positive neurons constitute a small subset of ganglion cells rather than displaced amacrine cells. The punctate X-Gal signal in Opn5^−/− retinas was typically confined to the periphery of the labeled somata, unlike, for example, the diffuse X-Gal signal observed in the somata, proximal dendrites, and axons of ipRGCs in Opn4^−/− retina (33). Such punctate X-Gal labeling has been reported (34), particularly in situations of low expression of β-gal coded for by lacZ (34), with confinement to subcellular organelles such as the endoplasmic reticulum (35); in our case, this pattern could be caused by weak Opn5-promoter activity. The Opn5^−/− line 2 mice likewise carry a knock-in lacZ gene in the Opn5 gene locus (Fig. S1C). It gave a similar punctate-labeling pattern with X-Gal (Fig. 5C), totaling ∼5,200 cells per retina, broadly similar to line 1. To reveal the Opn5-expressing neurons better, we tried staining Opn5^−/− line 1 retinas with a fluorescent substrate of β-gal [5-chloromethylfluorescein di-β-d-galactopyranoside (CMFDG)] and then injecting Alexa Fluor 555/488 hydrazides into a small number of the labeled neurons. Unfortunately, the CMFDG labeling appeared to be cytotoxic, causing membrane leakage and rapid dissipation of the injected dyes (SI Materials and Methods). Only four of 20 injected cells successfully retained enough dye to show discernable cell bodies and primary dendrites, with three cells showing a well-defined axon heading toward the optic disk, confirming them as retinal ganglion cells (Fig. S6). We also tried several commercial antibodies against the β-gal coded for by lacZ but failed to detect any labeling in the retina. Finally, we could not observe any X-Gal labeling or anti–β-gal immunolabeling in the cornea, possibly also due to its very low Opn5-promoter activity [a situation not unlike the absence of X-Gal labeling in non–M1-subtype ipRGCs in an Opn4-tau-lacZ knock-in mouse line, which also have low Opn4-promoter activity (2, 33, 36–38)].

Fig. 5.

Apparent Opn5 expression in retinal ganglion cells. (A) X-Gal staining in flat-mount adult WT and Opn5^−/− line 1 retinas. Blue puncta in Opn5^−/− retinas are X-Gal precipitate. (B, Left) Colocalization of X-Gal (blue puncta) and immunosignal from an antibody against retinal binding protein with multiple splicing (RBPMS; brown), a retinal ganglion cell marker, in a small subset of cells in a flat-mount Opn5^−/− retina. White arrowheads show examples of colocalization. (B, Right) Color-inverted and contrast-adjusted version of the image on the left for distinguishing the X-Gal labeling and immunosignal more easily. (C, Left and Middle) X-Gal staining of postnatal day 8 (P8) WT and Opn5^−/− line 2 retinas at about one-third of the distance from the optic nerve head to the retinal periphery. (Inset) Two positive cells enlarged. (C, Right) Ten-micrometer cryosection of an X-Gal–stained P10 Opn5^−/− line 2 retina, showing positive cells situated in ganglion cell layer (GCL). X-Gal staining was not detected in the inner plexiform layer (IPL), inner nuclear layer (INL), outer plexiform layer (OPL), outer nuclear layer (ONL), or photoreceptor outer segments (POS). (D) Lack of photoentrainment in ex vivo Math5^−/− retinas. (Top) Dark-recorded bioluminescence traces from a pair of Per2::Luciferase retinas (Left) or Math5^−/−;Per2::Luciferase (Right) retinas after 4 d of culturing at the 0° (blue) or 180° (red) position of the light/dark photoentrainment apparatus. (Bottom) Time of peak bioluminescence of Per2::Luciferase or Math5^−/−;Per2::Luciferase retinas (points, mean phase ± SEM) photoentrained at 0° (blue) and 180° (red) positions (n = 4 each). (E) Relative levels of Opn5 transcript as assayed by quantitative RT-PCR on total retinal or liver RNA. Opn5 transcript levels from retinas of Per2::Luciferase, rd1/rd1;Per2::Luciferase, or Math5^−/−;Per2::Luciferase mice were normalized to the Gapdh levels of each respective sample and then compared with the normalized Opn5 transcript level in Per2::Luciferase liver using the delta-delta threshold cycle (2^−ΔΔCt) method (SI Materials and Methods).

Fig. S6.

CMFDG-labeled neurons in Opn5^−/− line 1 retinas injected with Alexa Fluor 555 hydrazide. Three of four cells that have retained the Alexa Fluor dye showed an obvious axon (arrowheads). The cell body and the proximal dendrites are not clearly outlined because the images were overexposed to reveal the full morphology of the axons.

To confirm the necessity of retinal ganglion cells for local retinal circadian photoentrainment, we performed ex vivo photoentrainment experiments on Math5^−/− retinas, in which >80% of ganglion cells are absent developmentally (39). We found that the intrinsic rhythmicity of these retinas was intact and of normal amplitude, but their photoentrainment had largely disappeared (Fig. 5D). Opn5 transcript expression was also greatly reduced in these retinas (Fig. 5E). Taking all results together, we tentatively conclude that OPN5 is expressed primarily in a small subset of Math5-dependent retinal ganglion cells.

SI Materials and Methods

Animals.

All animal experiments were carried out according to protocols approved by the Institutional Animal Care and Use Committee at the University of Washington, Johns Hopkins University, and Cincinnati Children’s Hospital Medical Center. The rd1/rd1;Opn4^−/− mice were generated by crossing the rd1/rd1 line (47) (C57BL/6J background from The Jackson Laboratory) and Opn4^−/− line (3) (backcrossed to C57BL/6J background for at least seven generations). Opn1sw^−/− mice were from the laboratories of Jay and Maureen Neitz (University of Washington, Seattle, WA) (30). The Opn3^−/− line, as well as Opn5^−/− lines 1 and 2, was generated by homologous recombination as described below. For monitoring the retinal rhythm, the respective mouse lines were bred into the Per2::Luciferase background (27).

Generation of Opn3^−/−, Opn5^−/− Line 1, and Opn5^−/− Line 2 Mouse Lines.

The Opn3^−/− mouse line was generated by crossing an Opn3^flox/flox mouse line with a Sox2-Cre transgenic line (The Jackson Laboratory), in which Cre recombinase was expressed globally in early embryos. The Opn3^flox/flox line was derived from ES cell clones (Clone EPD0197_3_E01; Knockout Mouse Project Repository) that harbor a targeting construct for inserting both a flippase recognition target-flanked lacZ cassette upstream of, and a pair of loxP sites across, exon 2 of the Opn3 gene (Fig. S1A). The ES cells were injected into 129/SvJ1 blastocysts to produce chimeric mice, which were bred to C57BL/6J mice to give Opn3^flox/+ heterozygous mice and, subsequently, homozygous animals. A set of three primers was used to genotype the Opn3 alleles: OPN3WT forward (For): 5′-TGT ACC GTG GAC TGG AGA TCC AAG-3′, OPN3KOFor: 5′-TTA TGG CCC ACA CCA GTG GC-3′, and OPN3 reverse (Rev): 5′-GTT CCC ACA CAC GAC CTG CTC-3′. A combination of OPN3WTFor and OPN3Rev gave a 530-bp band for the WT allele, whereas a combination of OPN3KOFor and OPN3Rev gave a 670-bp band for the Opn3-KO allele. Absence of Opn3 transcripts was confirmed by RT-PCR experiments on total eye RNA (Fig. S1D and below). The Opn3^−/− line was backcrossed to the C57BL/6J background for five generations before mating with Per2::Luciferase mice to produce Opn3^−/−;Per2::Luciferase mice for experiments.

Opn5^−/− line 1 was generated by homologous recombination with a targeting construct to replace part of exon 3 and the entire exon 4 of the Opn5 gene with a tau-lacZ-neo cassette (Fig. S1B). The targeting construct was linearized and electroporated into ES cells derived from 129/SvJ1 mice. G418-resistant colonies were selected and genotyped by PCR to identify positive colonies for expansion and subsequent injection into C57BL/6J blastocytes. The resulting chimeric mice were bred to give heterozygous, and then homozygous, animals. A set of three primers was used to genotype the Opn5 alleles: OPN5WTFor: 5′-TGC TTT ACC ATG CCC AGC TAA GC-3′, OPN5KOFor: 5′-GCA GCC TCT GTT CCA CAT ACA CTT C-3′, and OPN5Rev: 5′-GCC TCT CTG ACC TTA CCT TC-3′. A combination of OPN5WTFor and OPN5Rev gave a 182-bp band for the WT allele, whereas a combination of OPN5KOFor and OPN5Rev gave a 253-bp band for the Opn5-KO allele. Absence of Opn5 5′ transcripts (both coding and noncoding isoforms) was confirmed by RT-PCR experiments on total eye RNA (Fig. S1E and below). In the Opn5^−/−;Per2::Luciferase animals that were tested, the neo cassette was retained. A 3′ transcript of Opn5, fused to an upstream neo transcript, was up-regulated in these mice (Fig. S1E), possibly due to the knocked-in mouse phosphoglycerate kinase 1 (PGK) promoter. However, this transcript should not result in OPN5 protein because of the presence of a STOP codon in the neo transcript.

Opn5^−/− line 2 was generated by using an ES cell clone from the International Knockout Mouse Consortium. The gene-targeting construct (Fig. S1C) is of the “knockout first” design, in which Cre-loxP–mediated recombination would delete the neo positive selectable marker but leave an IRES-lacZ in place. We used conventional methods to generate an Opn5^flox/flox mouse line from these ES cells, which was then crossed to ROSA-FLP mice (stock no. 003946; The Jackson Laboratory) and, subsequently, EIIa-Cre mice (stock no. 003314; The Jackson Laboratory; with Cre expression in germ line) to give Opn5^−/− line 2. Primers used for genotyping were F1: 5′-CAC AGT ATG TGT GAC AAC CT-3′, R1: 5′-GTG GAC AGA TTA ACT GAA GC-3′, F2: 5′-ACT ATC CCG ACC GCC TTA CT-3′, and R2: 5′-GAA CTG ATG GCG AGC TCA GA-3′. Absence of Opn5 5′ transcripts (both coding and noncoding isoforms) was again confirmed by RT-PCR experiments on total eye RNA (Fig. S1F). A 3′ transcript of Opn5 still existed in these mice (Fig. S1F), possibly due to the splicing of exon 5 into exon 3 after exon 4 was removed. However, because the splicing introduces a frame shift, and thus a premature STOP codon, the resulting transcript should not be translated into functional OPN5 protein.

RT-PCR.

Eyes and other tissues were freshly harvested into TRIzol reagent (Life Technologies) containing polyacryl-carrier (Molecular Research Center) and homogenized with autoclaved pestles. RNA was extracted by phase separation and precipitated by addition of isopropanol. After resuspension in diethylpyrocarbonate-treated water, RNA was reverse-transcribed by the SuperScript III First-Strand Synthesis System (Life Technologies). The resulting cDNA was amplified by PCR with the following primer sets: Fig. S1D, Opn3 (5′-TGG CTC TAC TCC TTG GCA TGG-3′ and 5′-ACT GGG TTG TAC ACA GTG CTC G-3′); Fig. S1 E and F, Opn5 Ex1-Ex4 (5′-GCC CCA CTA TCT TCG AGA CGA GG-3′ and 5′-TCC AGG GTG CAT GAG GTT CCG A-3′), Opn5 Ex5-Ex7 (5′-CCA TAC AGC TCT CCG TGG TG-3′ and 5′-TTC TCG GCC TCA GAA CAC AG-3′); Fig. S1 D–F, β-actin (5′-AAA GAG AAG CTG TGC TAT GTT G-3′ and 5′-CAT AGA GGT CTT TAC GGA TGT C-3′); and Fig. S4, Opn5 (5′-AGC TTT TGG AAG GCC AGA C-3′ and 5′-CAG CAC AGC AGA AGA CTT CC-3′), and Gadph (5′-GAC TTC AAC AGC AAC TCC CA-3′ and 5′-ATT GTG AGG GAG ATG CTC AGT-3′). PCR products were visualized by electrophoresis on agarose gels.

Photoentrainment of Cultured Tissues and Light Pulse-Induced Phase-Shift Experiments.

Immediately after CO₂ asphyxiation, the retinas, corneas, and pituitary glands were harvested from euthanized mice into cold HBSS (Gibco). Corneal and pituitary explants were placed directly into cell culture inserts (PICM0RG50; Millipore) in dishes containing DMEM supplemented with B-27 supplement (Life Technologies), 352.5 μg/mL NaHCO₃, 10 mM Hepes (Life Technologies), 25 units/mL penicillin, 25 μg/mL streptomycin (Life Technologies), and 0.1 mM luciferin potassium salt (Biosynth). Retinas were cultured in the same cell culture inserts, but initially in Neurobasal A medium (Cellgro) containing B-27 serum-free supplement (Life Technologies), 25 units/mL penicillin, 25 μg/mL streptomycin, and 2 mM l-Gln (Life Technologies) incubated under 5% CO₂/95% O₂ overnight, before being transferred to dishes containing the same DMEM/luciferin medium described above. All culture dishes were sealed with vacuum grease and maintained at 36 °C in incubators without CO₂. After culturing, the circadian clocks in the tissues were reset apparently due to media shock.

To subject pairs of tissues from an animal to opposite light/dark cycles, the tissues were cultured in dishes located at diametrically opposed positions (designated 0° and 180°) as described previously (7). Above the cultured tissues, a motor rotated a solid black disk with a pie-shaped transparent window at 24 h⋅revolution⁻¹ such that each of the paired tissues was illuminated for 9 h in every 24-h cycle, but in antiphase to each other. In initial experiments, we found full photoentrainment of retinal rhythms after 4 d of culturing under light/dark cycles ranging from 9 h/15 h to 14 h/10 h, but maximum phase separation between the two diametrically opposed retinas was attained by using the 9-h/15-h light/dark cycle; we therefore chose the latter light/dark setting. Light came from an array of five light-emitting diode (LED) sets within the incubator, with peak wavelengths at 370 nm, 417 nm, 475 nm, 530 nm, and 628 nm, respectively. White light was produced by simultaneously switching on the 417-nm, 475-nm, and 530-nm LEDs. All radiometric measurements were made with a Macam Q203 quantum radiometer (Macam Photometrics), and spectral irradiance was measured with a SpectroCal spectral radiometer (Cambridge Research Systems).

After 4 d of light/dark cycles, the cultured tissues were placed in darkness in a Lumicycle luminometer (Actimetrics), in which bioluminescence was measured continuously by four photomultiplier tubes. All bioluminescence data were detrended by using a polynomial fit with an order of 1 to remove the steady decline of background bioluminescence and then subjected to a best-fit sine-wave analysis for measuring the period of the oscillations (Lumicycle Analysis). Phases were determined from the sine wave best-fitted to at least 3 d of the oscillations.

For phase-shift experiments, the initial phase of a cultured tissue was first measured with the Lumicycle luminometer, and the tissue was then transferred to the incubator in a light-proof, insulated chamber (to maintain constant culture temperature) and exposed to a 3-h light pulse. Afterward, the tissue was returned to the Lumicycle luminometer and the phase measurement was resumed. Phase shift was defined as the difference between the time of peak luminescence observed after the light pulse and the time of peak luminescence expected from the phase of the ongoing clock if no light pulse were administered.

Electroretinogram.

Mice were given an i.p. injection of 110.25 mg/kg ketamine and 11.02 mg/kg xylazine. Once full anesthesia was achieved, one eye was dilated with one drop of atropine sulfate (1%) and the animal was placed on a water-heated stage. The active lead was a silver-impregnated nylon fiber (“Dawson, Trick, and Litzkow” electrode) placed over the cornea and held in place by a transparent contact lens. The reference and ground leads were platinum needle electrodes placed under the skin above the cranium (reference) and under the skin of the dorsal abdomen (ground). The mouse’s head was then advanced into a Roland Consult Q450 Ganzfeld electroretinogram (ERG) system. LEDs with peak spectral output between 594 nm and 405 nm were used simultaneously to produce light of a broad spectrum for the spectral range of mouse photoreceptors.

For dark-adapted ERGs, mice were left in complete darkness for at least 2 h before receiving the anesthetics under dim red light. Flashes of 3 cd⋅m⁻² were administered at 5-s intervals over a dark background. For light-adapted ERGs, mice received 900 cd⋅m⁻² flashes of white light over a background of 340 cd⋅m⁻² white fluorescent room light. Traces were averaged for ∼150 trials until 60-Hz noise was reduced. A bandpass filter between 1 and 100 Hz was applied to scotopic traces, and a bandpass filter between 20 and 100 Hz was applied to photopic traces.

Optokinetic Tracking.

Mice were placed on an elevated platform at the center of an arena surrounded by four computer monitors on the sides plus a reflective floor and ceiling (Cerebral Mechanics, Inc.). The unrestrained mice were allowed to acclimate to the environment, with a uniform gray field displayed on the monitors. During the test, the monitors displayed continuous high-contrast gratings of some randomly chosen spatial frequencies (0.05–0.5 cycles per degree), rotating at a constant speed in either a clockwise or counterclockwise direction. As controls, the experimental sessions were punctuated randomly with sessions showing an equiluminant gray field. Head movements of the mice in response to the rotating bars were captured by an overhead video camera and were scored by an observer (blinded to the conditions) as 0 = no visual tracking, 1 = ambiguous visual tracking, or 2 = obvious visual tracking.

Behavioral Wheel-Running Analysis.

Mice were housed individually in cages equipped with running wheels. The activation of a microswitch by each revolution of the wheel was registered in a computer. The number of revolutions per day was counted, and the χ²-period measurements (mean ± SEM) were performed using ClockLab software (Actimetrics). Fluorescent bulbs (1-W⋅m⁻² white light) were controlled by a timer to give 12-h/12-h light/dark cycles.

Staining for β-Gal and Dye Injection.

For X-Gal staining of Opn5^−/− line 1 retina, anesthetized mice were transcardially perfused with PBS, followed by freshly made X-Gal fixative (0.2% glutaraldehyde and 2 mM MgCl₂ in PBS). Retinas were isolated and postfixed with X-Gal fixative for another 30 min at room temperature. Afterward, the retinas were rinsed three times with detergent buffer (2 mM MgCl₂, 0.01% sodium deoxycholate, and 0.02% Nonidet P-40 in PBS). Staining was done by incubating the retinas in detergent buffer containing 5 mM potassium ferricyanide, 5 mM potassium ferrocyanide, and 1 mg/mL X-Gal. For Opn5^−/− retinas, it usually took about 24 h of incubation at room temperature to obtain reasonable staining intensities. Subsequently, retinas were rinsed with detergent buffer and processed for imaging or immunohistochemistry as described below.

For X-Gal labeling of Opn5^−/− line 2 retinas, the tissue samples were washed twice, each time for 15 min in PBS/0.02% Nonidet P-40 solution at 4 °C, and then fixed for 45 min in X-Gal fixative (1% formaldehyde, 0.2% glutaraldehyde, 2 mM MgCl₂, 5 mM EGTA, and 0.01% Nonidet P-40). They were washed twice again in the above PBS solution for 15 min each time at 4 °C and then stained overnight in X-Gal staining solution [5 mM K₃Fe(CN)6, 5 mM K₄Fe(CN)6, 2 mM MgCl₂, 0.01% Nonidet P-40, and 1 mg/mL X-Gal], and mounted afterward under coverslips. To provide a cross-sectional view, X-Gal–stained retinas were cryoprotected in 30% (wt/vol) sucrose and sectioned at a thickness of 10 μm.

To label neurons expressing β-gal for dye injection, freshly dissected retinas were incubated in oxygenated Ames’ Medium (Sigma–Aldrich) containing 100 μM CMFDG (D2920; Invitrogen) at room temperature for 5 min. CMFDG labeled neurons’ cell bodies and proximal processes. Based on control experiments on Opn4^−/− retinas, CMFDG only labeled a selective fraction of cells expressing β-gal, perhaps because of some problems with penetration into certain cell types. In addition, CMFDG signals did not persist after fixation and CMFDG did not stain fixed retinas. Thus, we relied on the injection of a fixable Alexa Fluor dye to reveal the morphology of a CMFDG-labeled neuron. We also found CMFDG to be cytotoxic, rendering the cell membrane very leaky within minutes. Hence, we performed all dye injections immediately after the staining procedures.

Dye injections were performed with sharp electrodes pulled from thick-walled borosilicate glass capillaries (GC150F-10; Harvard Apparatus). The tip of an electrode was filled with a small volume of 4 mM Alexa Fluor 555 and 1 mM Alexa Fluor 488 hydrazides dissolved in 1.5 M KCl, and the back was filled with 3 M KCl. The resistance of the electrodes was typically 50–150 MΩ. The targeted CMFDG-labeled cell and the pipette tip were visualized by green fluorescent light. The dye was injected by negative current (−500 to −1,000 pA, 3 min). Successful injection was confirmed by the retention of Alexa Fluor 555 dye in the soma and primary dendrites. After dye injection, retinas were fixed in 4% (wt/vol) paraformaldehyde (PFA) at 4 °C overnight and washed for 45 min in PBS. The injected cells were imaged by a Zeiss confocal microscope (LSM-510), and Z-stack projection was applied.

Immunohistochemistry of Retinal Sections and Flat-Mount Retinas.

For section immunohistochemistry, anesthetized mice were transcardially perfused with PBS, followed by freshly made 4% PFA, to remove endogenous mouse IgG antibodies in the bloodstream that might contribute to a high staining background. Eyeballs were postfixed with alcohol-based zinc-formalin solution (Z-fix; Anatech) overnight at room temperature, dehydrated through a series of increasing concentrations of ethanol, and finally embedded in paraffin (processed by Johns Hopkins Medical Laboratories). Five-micrometer sections at the level of optic nerve emergence were collected on glass slides. They were then deparaffinized with xylenes (Sigma) and rehydrated through a series of decreasing concentrations of ethanol. Antigen retrieval was performed with citrate buffer (Sigma) for 20 min in a 98 °C water bath. After cooling to room temperature, the sections were blocked with 10% (vol/vol) newborn calf serum in 0.1% Triton X-100 in PBS (PBST-0.1%) for 1 h at room temperature and incubated with the mouse monoclonal 1D4 antibody against rhodopsin (a gift from Robert Molday, University of British Columbia, Vancouver, BC, Canada) or a rabbit polyclonal antibody against OPN1SW or medium-wave length–sensitive cone opsin at 1:50, 1:500, and 1:500 dilutions, respectively, in the same blocking solution overnight at 4 °C. The latter two antibodies were generated by Jason Chen (Baylor College of Medicine, Houston, TX) against the same epitope as in the method used by Applebury et al. (48). After several washes with PBST-0.1%, the sections were incubated with Alexa Fluor 568 goat anti-mouse IgG antibody or Alexa Fluor 488 goat anti-rabbit IgG antibody (Molecular Probes) at a 1:500 dilution in blocking solution for 1 h at room temperature. Finally, sections were washed with PBST-0.1%, mounted with antifade medium (Vector Laboratories) containing DAPI, and cover-slipped.

For immunohistochemistry on flat-mount retinas, the retinas were first dissected from fresh mouse eyeballs and fixed with 4% PFA for 30 min at room temperature. They were then blocked with 10% (vol/vol) normal goat serum in 0.5% Triton X-100 in PBS (PBST-0.5%) overnight at 4 °C. Subsequently, the retinas were incubated with a rabbit polyclonal antibody against OPN4 (Advanced Targeting Systems) at a 1:1,000 dilution in blocking solution for 5 d at 4 °C. Afterward, the retinas were washed with PBST-0.5% and incubated with Alexa Fluor 488 goat anti-rabbit IgG antibody (Molecular Probes) at a 1:500 dilution in blocking solution overnight at 4 °C. Finally, the retinas were washed with PBST-0.5%, mounted with the above DAPI-containing antifade medium, and cover-slipped. For immunostaining following X-Gal labeling, the primary and secondary antibodies were replaced by a rabbit polyclonal antibody against RNA binding protein with multiple splicing (Abcam) at a 1:100 dilution and an HRP-conjugated anti-rabbit IgG antibody (Molecular Probes) at a 1:500 dilution, respectively. Chromogenic signals were developed by using the Metal Enhanced DAB Substrate Kit (Life Technologies).

Quantitative RT-PCR.

RNA was isolated from a tissue as described above. Quantitative RT-PCR was performed with Absolute Blue SYBR Low ROX Master Mix (Thermo Scientific) in an Applied Biosystems 7500 Real-Time PCR system. The delta-delta threshold cycle (2^-ΔΔCt) method was used for quantifying the relative abundance of RNA. Specifically, the ΔΔC_t was given by ΔC_t (Opn5_retina − Gapdh_retina) − ΔC_t (Opn5_liver − Gapdh_liver), where Opn5_retina and Gapdh_retina denote the C_t for Opn5 and Gapdh, respectively, in retinal samples, whereas Opn5_liver and Gapdh_liver denote the C_t for Opn5 and Gapdh, respectively, in liver samples serving as reference tissue. The final mRNA relative abundance was calculated by fold change in 2^-ΔΔCt. Primer sets for quantitative RT-PCR were Opn5 (5′-AGC TTT TGG AAG GCC AGA C-3′ and 5′-CAG CAC AGC AGA AGA CTT CC-3′) and Gadph (5′-GAC TTC AAC AGC AAC TCC CA-3′ and 5′-ATT GTG AGG GAG ATG CTC AGT-3′).

Discussion

We show in this work that OPN5 is required for the photoentrainment of the murine retinal and corneal circadian clocks ex vivo. The concordance between the wavelength dependence of the retinal entrainment and the absorption spectrum of OPN5 (13, 16) lends further support to the notion that OPN5 functions as a photopigment for these clocks. As such, we have identified a function for mammalian OPN5, previously an orphan opsin. It remains to be confirmed whether OPN5 is also required for photoentrainment of the retinal rhythm in vivo. Mouse cornea and lens, unlike their human counterparts, transmit UV-A (40), a property exploited by the abundant blue-cone pigment, OPN1SW, in the mouse retina, and presumably also by OPN5. Even in humans, there is likely enough transmitted blue light in full sunlight for OPN5 to signal, because OPN5’s sensitivity at the human blue cone pigment’s λ_max of 430 nm is only one log-unit lower than at its own λ_max of 380 nm [based on heterologous expression (13, 16)]. The cis-retinal required by OPN5 as a chromophore (13, 16) could potentially be accessed in the inner retina in the same way as by ipRGCs, possibly from the Müller glial cells (41). For corneal OPN5, interphotoreceptor retinoid-binding protein in the aqueous and vitreous humors (42) could potentially transport cis-retinal via diffusion from the retina. Because OPN5 appears to be bistable (13, 16), similar to bistable/tristable OPN4 (43–45), this property may explain its sustained photosensitivity under extended culture conditions ex vivo.

It is remarkable that the four mouse retinal pigments (rhodopsin, S/M-cone opsins, and OPN4) already known, which cooperatively photoentrain the SCN, nonetheless appear not to entrain the mouse retinal circadian clock. Instead, the latter function is left to OPN5. Conversely, so far, OPN5 does not appear to have an essential role in SCN photoentrainment (2, 3), although a more subtle role by this pigment still cannot be ruled out, as implicated by the slower reentrainment of Opn5^−/− mice to a phase advance in the experiment illustrated in Fig. 3C. The division of labor between OPN5 and the other retinal pigments may suggest a selective advantage in maintaining segregation of the light-signaling pathway for local retinal photoentrainment and the light-signaling pathway for SCN photoentrainment, for reasons presently unknown. Given that OPN5 appears to be expressed in a subset of ganglion cells (albeit of unknown subtype) and that OPN5 and S-cone opsin (at least in the mouse) both have a λ_max in the UV-A spectral window, it is not immediately clear how OPN5-expressing cells manage to avoid commingling photosignals from their intrinsic OPN5 pathway and those photosignals from the synaptically conveyed rhodopsin-cone pigments/OPN4 pathway. One solution to this problem would be that the OPN5-expressing cells, via unique circuitry, are functionally isolated from the mainstream light signaling in the retinal network. Another possibility would be that the two light pathways use distinct signal codes, such as electrical activities with different temporal properties that can be distinguished from each other by their respective decoders. Alternatively, the nature of the light signal downstream of OPN5 for local retinal photoentrainment may be entirely different from the conventional electrical signals (i.e., action potentials) sent to the SCN for photoentraining these nuclei. Thus, for example, OPN5-triggered signal transduction could lead to an electrically silent biochemical reaction rather than a change in membrane potential. If so, this characteristic may also imply that even though OPN5 is present in retinal ganglion cells, its signaling will not necessarily be transmitted to the brain. In contrast, ipRGCs make use of intrinsic OPN4-generated signals and also synaptically driven rod/cone signals (both electrical in nature), either synergistically or individually, for phase-shifting the SCN clock in the brain (2, 3).

The identity of the rhythmic cells in the retina remains unclear. In our experiments, Math5^−/− retinas showed a luciferase rhythm of largely unattenuated amplitude despite the loss of most Opn5 transcripts and the loss of photoentrainment (Fig. 5C). The simplest interpretation of these findings is that the majority of the rhythmic cells in the retina do not express OPN5, although their exact locations are still unclear (28, 29). If OPN5 cells are indeed distinct from the rhythmic cells, the important question arises of how light signals from OPN5 are transmitted to these rhythmic cells, especially if the OPN5-generated signal is nonelectrical downstream. Separately, it appears that the ex vivo retinal rhythm has a lower amplitude in the Opn3^−/− genotype than in WT (Fig. 2A), raising the possibility that although OPN3 is not necessary for local photoentrainment of most inner retinal cells, it nonetheless may have a role in influencing the intrinsic retinal rhythmic activity, perhaps by synchronizing the rhythmic cells with a mechanism that is still unknown. For example, we cannot rule out the possibility that OPN3 is required for the photoentrainment of rhythmic outer retinal neurons, which contribute a minority of the overall Per2::Luciferase signal in the retina (28, 29).

The photoentrainment of the cornea ex vivo provides the first evidence, to our knowledge, for photosensory function in the mammalian cornea. Although insects, reptiles, amphibians, and birds have long been known to use extraretinal photoreceptors, this corneal photosensitivity via OPN5 is a rare example of opsin-dependent, extraretinal photoreception in mammals [with another being the OPN4-mediated intrinsic photosensitivity of some mammalian irises (46)]. The fact that photoentrainment of the cornea likewise requires OPN5 instead of the other known retinal pigments may suggest an overall concerted ocular (i.e., not just retinal) photoentrainment that is segregated from SCN photoentrainment. We currently know nothing about the OPN5-expressing cell type or the nature of OPN5 signaling in the cornea, although the latter may, in principle, be similar to OPN5 signaling in the retina. The identity of the corneal rhythmic cells is likewise unknown.

OPN4 was initially identified by virtue of its role in entrainment of the SCN clock, but it is now known to be involved in a broad range of functions, including pupillary light reflex, SCN photoentrainment, visual irradiance coding, retinal vascular development, and negative phototaxis during mammalian development (reviewed in refs. 36–38). By the same token, OPN5 may have other functions besides circadian photoentrainment of the retina and cornea.

Materials and Methods

Opn3^−/− and Opn5^−/− mouse lines were generated by homologous recombination. Photoentrainment assays were carried out by culturing pairs of tissues in specified medium under a 9-h/15-h light/dark cycle at antiphase for 4 d and then recording the rhythms of luciferase bioluminescence in darkness. Light-induced phase shifts were determined from cultured retinas by measuring the bioluminescence rhythms before and after exposure of the retinas to a 3-h light pulse at selected phases of their rhythms. X-Gal labeling, immunohistochemistry, and quantitative RT-PCR were done with standard protocols. Detailed methods are described in SI Materials and Methods.