Melanopsin (Opn4) utilizes Gα_i and Gβγ as major signal transducers

Non-mammalian vertebrates employ Opn4, a G-protein-coupled receptor (GPCR) for extraocular photoreception (Beaulé et al., 2003). Photo-isomerization of chromophore-retinal in Opn4 induces its activation and governs downstream signaling by activating heterotrimeric G proteins. Opn4 was discovered in the melanophores of Xenopus laevis (African clawed frog) and later in tissues such as the eyes and deep brain of non-mammalian vertebrates (Beaulé et al., 2003; Panda et al., 2003). Two Opn4 isoforms, Opn4L and Opn4S, have been identified in mammals (Bellingham et al., 2006; Pires et al., 2007). In humans, Opn4 colocalizes with the circadian neuromodulator pituitary adenylate cyclase-activating polypeptide (ADCYAP1) in intrinsically photosensitive retinal ganglion cells (ipRGCs) (Hughes et al., 2016). They form part of the retinohypothalamic tract, which transmits photic information from the retina to suprachiasmatic nucleus to generate circadian rhythms. Opn4 is also present in other non-image-forming visual pathway structures, including the intergeniculate leaflet and olivary pretectal (Beaulé et al., 2003).

Opn4 functions as a photoreceptor by detecting brightness levels and discriminating visual signals (Brown et al., 2012). Opn4 also allows visual circuits to acquire optimal settings and perform light adaptation by measuring irradiance levels (Allen et al., 2014; Hankins and Hughes, 2014). In addition to its visual roles, Opn4 contributes to a broad range of non-visual responses, including control of the circadian clock, pupillary constriction, pineal melatonin suppression, axon regeneration after nerve system injury and sleep regulation (Li et al., 2016; Panda et al., 2003; Pilorz et al., 2016). Opn4 also influences plasma corticosterone levels, causing behavioral arousal and delaying sleep onset when exposed to blue light (Pilorz et al., 2016).

Opsins, including Opn4, belong to class A G-protein-coupled receptors (GPCRs), the largest subfamily, representing 85% of GPCR genes (Raible et al., 2006). Most opsins reported in the literature are G_i/o-coupled GPCRs (Masseck et al., 2014; Terakita, 2005), several, including Opn4, are considered G_q-coupled, whereas the only known G_s-coupled opsin to date is jellyfish opsin (Koyanagi et al., 2008). It has been suggested that Opn4 may also activate the G_i pathway (Bailes and Lucas, 2013). However, no direct evidence supports Opn4-induced Gβγ signaling. The present study shows that, in addition to its activation of the canonical G_q pathway, Opn4 also efficiently activates Gβγ signaling, which may provide a molecular explanation for its functional diversity. Gβγ controls the activity of several major signal transducers, including: (1) PI3Kγ (Kolsch et al., 2008), (2) adenylyl cyclase (AC) isoforms (activation of AC2, AC4 and AC7; inhibition of AC1 and AC5) (Sunahara and Taussig, 2002; Tang and Gilman, 1991), (3) inwardly rectifying potassium channels (Kawano et al., 1999; Nakajima et al., 1996), (4) PLC isoforms (PLCβ₂, PLCβ₃) (Camps et al., 1992; Smrcka and Sternweis, 1993), (5) Ca²⁺ channels (N, P/Q type) (Herlitze et al., 1996; Ikeda, 1996), (6) GPCR kinase 2 (Pitcher et al., 1992), and (7) GEFs, including Rac, Cdc42 GEF (FLJ00018) and p114-GEF (Mayeenuddin et al., 2006; Niu et al., 2003; Ueda et al., 2008). The ability of Gβγ to control such a large number of signaling regulators should allow cells to utilize Gβγ signaling via Opn4 to perform their designated functions.

Our goal was to show that Opn4 possesses distinct heterotrimer activation abilities compared with other G_q-coupled GPCRs. Gα_qGTP generated upon G_q-coupled GPCR (including Opn4) activation stimulates PLCβ, which hydrolyzes the plasma membrane (PM) phospholipid, PIP₂, and generates IP₃ and DAG. HeLa cells expressing Opn4, mCherry-γ₉ and Venus-PH exhibited robust translocation of mCherry-γ₉ from the PM to inner membranes (IMs), with simultaneous PIP₂ hydrolysis (Fig. 1A, left, Movies 1 and 2) when Opn4 is photoactivated at 515 nm in the presence of 50 µM 11-cis-retinal. Mean fluorescence changes in IMs (white arrow) and cytosol represent the rate and extent of Gγ₉ translocation and PIP₂ hydrolysis, respectively. Despite its previously shown promiscuity towards G_i heterotrimers (Offermanns et al., 1994; van Unen et al., 2016), G_q-coupled M3-muscarinic receptor (M3R) activation in HeLa cells only exhibited minor translocation of mCherry-γ₉ (Fig. 1A, middle, Movie 3). The G_q-coupled gastrin-releasing peptide receptor (GRPR) responded similarly (Movie 4). Nevertheless, Opn4, M3R and GRPR activation exhibited comparable PIP₂ hydrolysis responses (Fig. 1A). These data indicate that, like G_i-coupled GPCRs, Opn4 is more efficient at translocating free Gβγ than both M3R and GRPR (Fig. 1B); however, different expression levels of receptors could alter the extent of signaling. Control cells lacking receptor expression did not exhibit PIP₂ hydrolysis upon exposure to carbachol, bombesin or 515 nm light (Fig. S1A), thus the Gβγ translocation observed is a result of αGTP and free Gβγ generated by activated GPCRs (Akgoz et al., 2004, 2006; Bondar and Lazar, 2014).

We compared activated Opn4, G_i-coupled α2-AR and CXCR4-induced Gβγ translocations (Movie 5). The broad absorption spectrum of Opn4 (λ_max∼480 nm) (Enezi et al., 2011; van Oosterhout et al., 2012), allowed imaging of the fluorescence sensor upon photoactivation at 445, 488 and 515 nm (Fig. S1B). Activated Opn4, α₂AR and CXCR4 exhibited comparable Gγ₉ translocations (Fig. S2). To demonstrate the distinct ability of Opn4 to induce Gβγ translocation compared with G_q-coupled GRPR, sequential activation of these receptors was examined in the same cells. HeLa cells expressing both Opn4 and GRPR together with mCherry-γ₉ and Venus-PH were imaged using 515 nm and 595 nm excitation wavelengths. At 40 s, GRPRs were activated using 1 µM bombesin (Fig. 1C). Although an intense PIP₂ hydrolysis and robust recovery were observed, mCherry-γ₉ translocation was not detected (Fig. 1C, graph). After PIP₂ recovery, adding 11-cis-retinal at 50 s activated Opn4 in the same cell. Both PIP₂ hydrolysis and robust mCherry-γ₉ translocation were observed (Fig. 1D). This demonstrates the ability of activated Opn4, like α2-AR and CXCR4, to efficiently generate Gβγ (Senarath et al., 2016).

Our next goal was to decipher the origin of free Gβγ generated by activated Opn4. Gβγ₉ translocation and PIP₂ hydrolysis were examined in the presence of pertussis toxin (PTx). PTx catalyzes ADP-ribosylation of the Gα_i subunit and prevents coupling of heterotrimers to GPCR (Ribeiro-Neto and Rodbell, 1989). HeLa cells expressing Opn4, mCherry-γ₉ and YFP-PH were incubated with 100 nM PTx for 5 h at 37°C. Opn4 activation did not induce Gβγ₉ translocation (Fig. 2A, top). Interestingly, the same cells exhibited PIP₂ hydrolysis (Fig. 2A, bottom). This demonstrates that Opn4 induces translocation of Gβγ₉ generated through the G_i pathway. Whether the observed Gβγ translocation is a result of G_q heterotrimer activation was examined. HeLa cells were incubated with G_q inhibitor YM-254890 (1 µM) for 5 min at 37°C, to block G_q heterotrimer dissociation by preventing GDP-GTP exchange (Takasaki et al., 2004). With YM-254890, Opn4 activation still induced efficient Gβγ₉ translocation (Fig. 2B, top) from the PM (yellow arrows) to IMs (white arrow), while G_q-mediated PIP₂ hydrolysis was abolished (Fig. 2B, bottom). Similarly, HeLa cells treated with only vehicle solvent used for PTx (Fig. S3A) and YM-254890 (Fig. S3B) exhibited both PIP₂ hydrolysis and Gγ₉ translocation. Collectively, these data suggest that Opn4 can simultaneously activate Gα_i and Gα_q signaling in the same cells to a comparable extent, and Gα_i-heterotrimer activation induces Gβγ translocation. Whether Opn4 activates translocation of other Gγ subunits was also examined. HeLa cells transiently expressing Opn4 and either YFP-γ₁ or YFP-γ₁₁ and pre-treated with 11-cis-retinal exhibited significant Gγ₁ and Gγ₁₁ translocations from the PM (yellow arrows) to IMs (white arrows) upon Opn4 activation (Fig. 2C). Similarly, Opn4 activation also induced Gβ₁-YFP translocation (Fig. 2C, bottom). This clearly demonstrates that Opn4 activates G protein heterotrimers, regardless of Gγ subtype. Gβ₁ translocation indicates that it translocates as a Gβγ dimer, as we have showed previously (Senarath et al., 2018).

Dimerization of certain GPCR combinations, activating distinct heterotrimer types has been demonstrated (Terrillon and Bouvier, 2004). To test the likelihood of Opn4 forming a heterodimer with a specific G_i-coupled GPCR in HeLa cells, activation of G_i signaling upon optical activation of Opn4 was examined using H125 lung carcinoma cells. The different origins of HeLa and H125 cells makes it highly unlikely that both cell types express the same GPCR, which can dimerize with Opn4 to induce its G_i activity. H125 cells expressing Opn4, mCherry-γ₉ and Venus-PH, pre-incubated with 11-cis-retinal, also exhibited both mCherry-γ₉ translocation (yellow and white arrows) and PIP₂ hydrolysis upon Opn4 activation (Fig. 2D). This indicates that Opn4 induces mCherry-γ₉ translocation.

The next objective was to examine concurrent activation of Gα_q and Gβγ effectors by activated Opn4. Both GαqGTP and Gβγ activate PLCβ, hydrolyzing PIP₂ to IP₃ and DAG (Park et al., 1993; Wu et al., 1993). DAG stays bound to the PM (Kamp and Hell, 2000; Park et al., 1993). Generation of DAG in HeLa cells was measured using translocation of an mCherry-conjugated DAG binding domain (DBD) sensor from the cytosol to the PM (Gallegos et al., 2006). 11-cis-retinal was added at 50 s to HeLa cells expressing mCherry-DBD and Opn4-YFP. Upon Opn4 activation, cytosolic DAG sensor translocated to the PM (Fig. 3A, top). While the receptor remained in an active state, the DBD sensor reverse-translocated to the cytosol, probably due to adaptation of PIP₂ hydrolysis (Ferguson and Caron, 1998; Gainetdinov et al., 2004). Cells treated with G_q inhibitor did not exhibit any detectable mCherry-DBD translocation (Fig. 3A, bottom). Calcium mobilization elicited by Opn4-generated IP₃ was measured using HeLa cells pre-incubated with the calcium indicator Fluo-4 AM (2.28 μM). Cells were imaged using 488 nm excitation and 515 nm emission to capture Fluo-4 fluorescence. Opn4-expressing cells exhibited increases in Ca²⁺ upon addition of 11-cis-retinal, which was transient and gradually disappeared (Fig. 3B, top). Cells incubated with G_q inhibitor did not exhibit any Ca²⁺ increase on Opn4 activation (Fig. 3B, bottom). These data suggest that G_q-pathway activation is primarily responsible for Opn4-regulated Ca²⁺ and DAG signaling. G_i pathway activation by Opn4 was measured in HeLa cells using a bi-molecular translocation-based cAMP sensor, PKA-δRIIβ: PKA-Cα-YFP (Dyachok et al., 2006; Siripurapu et al., 2017). Briefly, HeLa cells expressing Opn4, cAMP sensor, and mCherry-PH were exposed to 10 µM forskolin in IBMX (3-isobutyl-1-methylxanthine). Cells exhibited complete translocation of PKA-Cα-YFP from the PM to the cytosol, indicating cAMP generation (Fig. 3C, top middle). After ∼3 min, 11-cis-retinal was added, activating Opn4, and cells exhibited a reverse translocation of PKA-Cα-YFP. This indicates Gα_i-mediated inhibition of AC and cAMP production. A simultaneous PIP₂ hydrolysis upon Opn4 activation was also observed (Fig. 3C, bottom).

Gβγ is a crucial signal transducer for cellular homeostasis and it controls a cohort of essential effectors in the GPCR pathway. We demonstrated that only cells expressing high PM affinity Gγ subtypes (γ₂, γ₃) activate PM-bound effectors of Gβγ, including PI3K and PLBβ (Senarath et al., 2018). HeLa cells lack γ₂ and γ₃, and exhibit only minor PIP₃ production upon G_i pathway activation, whereas γ₃-expressing RAW264.7 cells exhibit PIP₃ production (Senarath et al., 2018). We also showed that localized G_i pathway activation and Gβγ generation in RAW264.7 cells induces directional migration, accompanied by PIP₃ generation at the leading edge (Senarath et al., 2018; Siripurapu et al., 2017). Thus, Opn4-induced PIP₃ generation and cell migration were examined using RAW264.7 cells. RAW264.7 cells expressing Opn4-GFP and Akt-PH-mCherry (PIP₃ sensor) and treated with 11-cis-retinal were exposed to a spatially restricted blue light stimulus (Fig. 3D, white box). Time-lapse images clearly indicate PIP₃ production at the optically activated leading edge. The orthogonal slice shows accumulation of PIP₃ at the PM (white arrow) (Fig. 3D). This Opn4-induced PIP₃ production was abolished by PTx (Fig. 3E), as well as by the Gβγ inhibitor Gallein (Lehmann et al., 2008) (Fig. S3C). This indicates that Opn4-induced PIP₃ production is independent of the G_q pathway and is governed by free Gβγ. Additionally, activation of Opn4 in the presence of G_q inhibitor resulted in not only PIP₃ production but also a noticeable cell migration (Fig. 3F). Generated PIP₃ alone can activate many signaling pathways, including Akt/PKB, several GEFs (Tiam1, P-Rex1), tyrosine kinase (Btk), and growth factor receptor-bound protein 2 (Grb2) to control cellular functions ranging from migration to gene transcription (Chen et al., 2012; Karunarathne et al., 2013; Weiner, 2002). Since Opn4 induces PIP₃ generation, in addition to its signaling through Gα_i, it should control PIP₃-regulated pathways, thus affecting diverse cellular functions.

Since Opn4 short (Opn4S) and long (Opn4L) splice variants in ipRGCs are anticipated to deliver diverse cellular functions (Hughes et al., 2012; Pires et al., 2009), we examined whether they possess distinct signaling activities. We generated Opn4S by truncating the C-terminus (Ct) of Opn4L (or Opn4). When expressed in HeLa cells, Opn4S and Opn4L exhibited similar signaling properties (Fig. 1A, left, Fig. S4A). Overall, these data suggest that G_i and G_q pathway selectivity in Opn4 is independent of its Ct domain.

To govern both G_i and G_q signaling, Opn4 should be able to interact with both G_i and G_q heterotrimers. GPCRs recognize heterotrimers with appropriate Gα types, primarily using intracellular loop 3 (IL3) (Itoh et al., 2001). We examined IL3 regions of G_i- and G_q-coupled GPCRs and compared them with the IL3 sequence of Opn4 using MAFFT multiple sequence alignment software (Fig. 4A,B) (Katoh et al., 2002; Katoh and Standley, 2013). This analysis showed that IL3 of Opn4 clustered with the IL3 regions of G_i-coupled complement 5a receptor (C5a) and serotonin receptor (5-HT-1) (Fig. 4A), and the IL3 region of Opn4 clustered with G_q-coupled calcitonin receptor (Fig. 4B). Interestingly, the left side of the IL3 region of Opn4 exhibited G_q-coupled (Fig. S4B) whereas the right side showed G_i-coupled GPCR-like characteristics (Fig. 4C). Left side mutations of IL3 – R262A, E263A and T264A (Fig. S4B, box) – resulted in completely non-responsive Opn4. This result suggests either that G_q-Opn4 coupling prior to G_i heterotrimer formation is crucial for their activation, or that mutations in the left side of IL3 may alter Opn4 conformation, rendering it non-functional. Among right side mutations, mutant R279A, R280A Opn4 (Fig. 4C, box) exhibited a 4-fold lower level of mCherry-γ₉ translocation compared with wild-type Opn4 (Fig. 4D,E). Its ability to induce PIP₂ hydrolysis remained intact (Fig. 4D,F). This further suggests that Opn4 interacts with and activates both G_q and G_i heterotrimers.

Since Opn4 can activate Gα_i and regulate cAMP, it can control axonal regeneration and sleep regulation (Hannila and Filbin, 2008; Spencer and Filbin, 2004). G_i-coupled melatonin 2 receptor controls circadian rhythms by regulating both Gα_i and Gβγ (Dubocovich and Markowska, 2005; Dubocovich et al., 2003). Muscarinic receptor-induced G_i signaling has been shown to regulate pupillary constriction and dilation in human (Abrams et al., 2006), serum corticosterone levels and the sleep-wake cycle in mice (Hemrick-Luecke et al., 2002). Therefore, Opn4 can regulate these functions by activating G_i pathway. Considering the broad range of effectors and cellular functions controlled by the G_i pathway, especially through Gβγ, the functional diversity of Opn4 and its ability to signal through multiple pathways are likely to be linked.

YFP-PH cDNA has been previously described (Gallegos et al., 2006). Bi-cistronic Opn4-GFP, mCherry-γ₉, mCherry-PH, M3R, mCherry-DBD, Venus-PH and Akt-PH-mCherry cDNAs were kindly provided by Professor Narasimhan Gautam's lab (Washington University, St Louis, MO). GRPR was a kind gift from the lab of Zhou-Feng Chen (Washington University, St Louis, MO). Opn4 PCR product was inserted into KpnI and XbaI sites in pcDNA 3.1 to create untagged Opn4. Opn4S was created by adding a stop codon to Opn4L after aligning splice variants (Opn4L and Opn4S). Overlap PCR and Gibson assembly were performed to generate Opn4 R279A, R280A mutant based on multiple sequence alignment data described. Reagent sources are as follows: pertussis toxin (PTx) and norepinephrine (Sigma-Aldrich), Gibson assembly master mix and competent cells (NEB, Ipswich, MA, USA), YM-254890 (Focus Biomolecules, Plymouth Meeting, PA, USA), Fluo-4 AM (Molecular Probes, Eugene, OR, USA), Bombesin (Tocris, Park Ellisville, MO, USA), Carbachol (Fisher Scientific, Pittsburgh, PA, USA), SDF-1α (PeproTech, Rocky Hill, NJ, USA), 11-cis-retinal (National Eye Institute, Bethesda, MD, USA), PolyJet (SignaGen, Rockville, MD, USA), IBMX and Forskolin (Cayman Chemical, Ann Arbor, MI, USA). All the reagents were dissolved in the appropriate solvent according to the manufacturers' instructions and diluted in 1% HBSS supplemented with NaHCO₃ or regular cell culture medium before adding to cells. Cell lines (HeLa, RAW264.7) were originally purchased from the American Tissue Culture Collection (ATCC, Manassas, VA, USA) and authenticated using a commercial kit to amplify 9 unique STR loci. NCI-H125 cell line was kindly provided by Dr Randall Ruch (University of Toledo, OH, USA).

RAW264.7 mouse macrophage cell line (ATCC) was cultured in RPMI 1640 (10-041-CV; CORNING, Manassas, VA, USA) with 10% fetal bovine serum (Atlanta Biologicals, Flowery Branch, GA, USA), L-glutamine and 1% penicillin-streptomycin in a humidified incubator at 37°C and 5% CO₂. At 75% confluency, cells were lifted after incubation with versene-EDTA (CellGro) for 3 min at 37°C, centrifuged at 1000 g for 3 min and versene-EDTA was aspirated before resuspending cells in RPMI 1640 with 10% dialyzed fetal bovine serum at 8×10⁴/ml cell density. One day before transfection of DNA into cells, 8×10⁴ cells were seeded on 35 mm glass bottom dishes (In Vitro Scientific). HeLa cells (ATCC) were cultured in minimum essential medium (CellGro) containing 10% fetal bovine serum (Atlanta Biologicals), in the presence of 1% penicillin-streptomycin in 60 mm tissue culture dishes in a humidified incubator at 37°C, 5% CO₂. The remaining cell culture procedures for HeLa cells were similar to the protocols described above for RAW264.7 cells. Transfection was performed using PolyJet transfection reagent (SignaGen, Rockville, MD, USA) according to the manufacturer's protocol and imaging was performed 12-14 h after transfection. Before imaging, cell culture medium was replaced with Hank's Balanced Salt Solution (HBSS) that was equilibrated for 1 h at 37°C and 5% CO₂.

G_i- and G_q-coupled GPCRs were identified (Alexander et al., 1999, 2017). The IL3 regions of G_i- and G_q-coupled receptors were examined by HMMTOP transmembrane prediction software (http://www.enzim.hu/hmmtop/). GenomeNet bioinformatics tools (http://www.genome.jp/en/gn_tools.html) were utilized to perform sequence analysis and tree generation. Sequences of IL3 regions of G_i- and G_q-coupled receptors were aligned using MAFFT (Katoh and Standley, 2013; https://mafft.cbrc.jp/alignment/software/). Aligned sequence phylogenies were subjected to unweighted pair-group method using arithmetic averages (UPGMA) clustering to generate phylogenetic trees.

A Nikon-Andor spinning disk confocal imaging system, with a Nikon Ti-R/B inverted microscope, a Yokogawa CSU-X1 spinning disk unit (5000 rpm) and iXon ULTRA 897BV back illuminated deep-cooled EMCCD camera was used. Photo-activation of selected regions of Opn4 was conducted using an Andor Fluorescence Recovery After Photo bleach and Photo Activation (FRAP-PA) unit. All DNA constructs were transiently transfected and cells were imaged after overnight transfection. Bi-cistronic vectors carrying Opn4 and GFP, Opn4 untagged and Opn-4GFP constructs were used to transiently express Opn4 and their expression was confirmed by using cytosolic GFP, PIP₂ hydrolysis or Gβγ translocation on activation or using the GFP tag, respectively. Imaging of cells was carried out using a 60×, 1.4 NA oil objective employing 50 mW 445, 488, 515, 595 nm solid-state lasers. Sensors were imaged using the following settings: GFP and Fluo-4 AM, 488 nm at 56 µW/515 nm; YFP and Venus, 515 nm at 22 µW/540 nm; mCherry, 594 nm at 20 µW/630 nm (excitation/emission). Cells were pre-incubated with 50 µM 11-cis-retinal unless otherwise specified. Optical activation was performed using a FRAP-PA device by drawing a desired light exposure region in the field of vision and exposing those areas to 445 nm blue light pulsed at 1 Hz at 0.22 µW. Imaging of sensors for PIP₂, PIP₃, DAG and Gγ₉ translocation were also performed at 1 Hz; since Opn4 has a broad absorption spectra, imaging YFP activates it. Digital image analysis was performed using Andor iQ 3.1 software and fluorescence intensity obtained from regions of interest (PM, IMs and cytosol) was normalized to initial value (baseline). Normalized data were then plotted using Origin pro (OriginLab Corporation). Results of all quantitative assays (Gβγ translocation and PIP₂ hydrolysis, Ca²⁺ responses) are expressed as standard error of mean (s.e.m.) from n number of cells (indicated in the figure legends) on multiple days, using cells with a different passage number.

For cytosolic Ca²⁺ measurements, HeLa cells were cultured on glass-bottomed dishes in MEM/DFBS at 37°C with 5% CO₂ as described above. Experiments were performed 12-24 h after transfection of DNA. Opn4-transfected cells were washed twice with 1% (v/v) HBSS supplemented with NaHCO₃ and Ca²⁺, adjusted to pH 7.2, and incubated for 30 min at room temperature with the fluorescent Ca²⁺ indicator, Fluo-4 AM (2.28 µM) in dark conditions. Before imaging, cells were again washed twice with 1% HBSS supplemented with NaHCO₃ and Ca²⁺ to remove excess Fluo-4 AM. Processed dishes were mounted and fluorescence intensity of Fluo-4 AM (488 nm) was continuously imaged at 1 s intervals using 488 nm excitation, 515 nm emission for confocal microscopy. Fluo-4 AM fluorescence intensity obtained from regions of interest was normalized to initial value.

Results of all quantitative assays (PIP₂ hydrolysis, PIP₃ generation, Gγ translocation, DAG formation, calcium, cAMP production) were computed as mean±s.e.m. from ≥10 cells in multiple independent experiments (>3) unless otherwise specified. Statistical analysis of the extent of Gγ9 translocation induced by wild-type and mutant Opn4 was performed using two-tailed unpaired t-test. The two means were significantly different at P<0.0001.

We thank Prof. Narasimhan Gautam for cDNA for G proteins, GPCRs and sensors; Dr Matthew Toomey for R script generating opsin spectra; Dr Anders Tengholm for providing bi-molecular cAMP probe; Dr John L. Payton, Mithila Tennakoon, Kanishka Senarath and Zehra Fasih for experimental assistance, discussions and comments. We acknowledge the National Eye Institute for providing 11-cis-retinal.