Flavonoids enhance rod opsin stability, folding, and self-association by directly binding to ligand-free opsin and modulating its conformation

Rhodopsin (Rho) is a visual G protein–coupled receptor expressed in the rod photoreceptors of the eye, where it mediates transmission of a light signal into a cell and converts this signal into a nerve impulse. More than 100 mutations in Rho are linked to various ocular impairments, including retinitis pigmentosa (RP). Accordingly, much effort has been directed toward developing ligands that target Rho and improve its folding and stability. Natural compounds may provide another viable approach to such drug discovery efforts. The dietary polyphenol compounds, ubiquitously present in fruits and vegetables, have beneficial effects in several eye diseases. However, the underlying mechanism of their activity is not fully understood. In this study, we used a combination of computational methods, biochemical and biophysical approaches, including bioluminescence resonance energy transfer, and mammalian cell expression systems to clarify the effects of four common bioactive flavonoids (quercetin, myricetin, and their mono-glycosylated forms quercetin-3-rhamnoside and myricetrin) on rod opsin stability, function, and membrane organization. We observed that by directly interacting with ligand-free opsin, flavonoids modulate its conformation, thereby causing faster entry of the retinal chromophore into its binding pocket. Moreover, flavonoids significantly increased opsin stability, most likely by introducing structural rigidity and promoting receptor self-association within the biological membranes. Of note, the binding of flavonoids to an RP-linked P23H opsin variant partially restored its normal cellular trafficking. Together, our results suggest that flavonoids could be utilized as lead compounds in the development of effective nonretinoid therapeutics for managing RP-related retinopathies.

G protein-coupled receptors (GPCRs) 2 are the largest group of membrane receptors that transmit signals across the plasma membrane into the cell (1,2). GPCRs are activated by multiple ligands such as small organic molecules, protons, ions, peptides, lipids, and light (3). Binding of the ligand triggers a conformational change in the receptor, allowing for coupling of the specific heterotrimeric G protein that further amplifies the signal and activates downstream effectors leading to the biological responses (1,4).
GPCRs are expressed in most tissues and contribute to almost all physiological processes in the human body. They are also important targets of about 30 -40% of all medications available on the market (5). Structural information available for many GPCRs enables the designing of more specific drugs targeting these particular receptors. However, often due to high homology of the orthosteric ligand-binding pockets among GPCRs within the specific class, selectivity is limited and can lead to unwanted side effects. Thus, more work is needed to develop novel and more explicit therapeutic compounds. GPCR activity can also be modulated allosterically, either by endogenous modulators or exogenous natural products and synthetic molecules. Discovering new allosteric modulators of GPCRs is another promising avenue to explore for achieving higher drug-receptor specificity. In fact, several such modulators are already Food and Drug Administration-approved medications (6 -8).
Rhodopsin (Rho) is a light-sensing GPCR. Its major function is the absorption of light photons and transduction of their absorption into a neural impulse (9,10). Over 100 mutations identified in Rho are linked with various ocular impairments, including congenital stationary night blindness and retinitis pigmentosa (RP) (11,12). There is a growing interest in finding novel ligand molecules that would improve the folding and stability of particular Rho mutants (13,14). However, small molecule synthetic drugs can often cause toxicity despite their specific modulatory effects. Thus, utilization of natural compounds could be an alternative approach to drug discovery efforts.
The medical use of plants or plant extracts to prevent or treat various diseases has a long history and is still commonly practiced in developing countries (15)(16)(17)(18). The natural dietary polyphenol compounds, in particular flavonoids, ubiquitously present in fruits and vegetables, improve sight in several eye-related diseases, including age-related macular degeneration, glaucoma, RP, and diabetic retinopathy, most likely due to their anti-oxidant, anti-inflammatory, or anti-apoptotic properties (19 -23). Only limited studies indicate that flavonoid compounds can interact with GPCRs. For example, myricetin can act as an agonist of the glucagon-like peptide-1 (GLP-1) receptor, which plays a critical role in the regulation of glucose homeostasis (24). Also, iso-flavone and flavone compounds are able to inhibit the histamine H3 receptor that is expressed mainly in the central nervous system (25,26). In the case of Rho, as found in this study and demonstrated previously, flavonoids can interact with the opsin apoprotein resulting in the structural changes that enhance the access of native retinal to the ligandbinding pocket (27)(28)(29). Moreover, flavonoids interacting with GPCRs can modify their membrane organization through the formation of aggregates (30). However, the effect of flavonoids on the stability, function, and oligomeric membrane organization of Rho is not entirely clear. Therefore, here we describe studies of the effects of four common bioactive flavonoids, quercetin 2-(3,4-dihydroxyphenyl)-3,5,7-trihydroxychromen-4-one and myricetin 3,5,7-trihydroxy-2-(3,4,5-trihydroxyphenyl)chromen-4-one and their mono-glycosylated forms (31,32) on rod opsin stability, its function, and membranous supramolecular assembly in vitro and in living cells.

Computational analyses of the interaction between opsin and flavonoids
The crystal structure of bovine rod opsin (PDB code 3CAP (33)) was used as a template to identify the possible interaction sites of quercetin, myricetin, quercetin-3-rhamnoside, and myricetrin with the ligand-free opsin, and the pocket analysis was performed with the CASTp software (Fig. 1a). Three potential binding pockets were identified, of which one was the orthosteric retinal-binding pocket and two were located on the surface of opsin. The extracellular binding pocket 1 was present between transmembrane helix (TM) 5, TM6, and extracellular loop (ECL) 2, and the extracellular binding pocket 2 was located between TM2, TM3, and ECL1 (Fig. 1a). Blind docking (34) of quercetin (as a model flavonoid) to opsin structure co-validated the results obtained with the CASTp software and defined the coordinates of the binding sites (Table 1). Quercetin was accommodated best within the orthosteric-binding site with the lowest binding free energy of Ϫ9.3 kcal/mol. The calculated binding energies of quercetin to the external binding sites were higher as compared with the orthosteric site: Ϫ7.9 and Ϫ6.5 kcal/mol for pockets 1 and 2, respectively. Based on the best binding free energies and the conformation of the pocket, the orthosteric site and pocket 1 were selected to develop the further analyses. Molecular docking of myricetin, quercetin-3rhamnoside, and myricetrin, into the opsin structure, revealed that all these flavonoid compounds, comparable to quercetin, accommodated in the orthosteric-binding pocket with lower binding energy than to the external binding sites. Additionally, the glycosylated flavonoids displayed a change in the isoflavone group orientation within the orthosteric-binding pocket as compared with their aglycone forms. The aglycone compounds quercetin and myricetin, but not the glycosylated quercetin-3rhamnoside and myricetrin, could also fit into the TM5-TM6 -ECL2 external binding site (pocket 1), and their isoflavone group protruded outside the binding site ( Fig. 1, b and c). The enlarged size of the glycosylated flavonoids most likely precluded their proper accommodation into the external binding site. Only about 40% of the glycosylated flavonoid molecules occupied pocket 1, whereas 60% was exposed to the solvent, thus suggesting low stability of the interaction between protein and the compound over time. All flavonoid compounds formed mainly hydrogen bonds andinteractions with the protein side chains, within both the orthosteric site and the external binding pocket 1. However, fewer such interactions were detected in the external binding pocket as compared with the orthosteric-binding site. The details of the interactions between flavonoids and opsin molecule are shown in Fig. 2 and are listed in Table 1.

Effects of flavonoids on the stability of opsin
Ligand-free opsin is highly unstable in vitro, and in vivo, its excess accelerates retinal degeneration (35)(36)(37). The binding of the natural ligand or its analogs significantly increases the stability of opsin (38). The melting temperature of opsin in the rod outer segment (ROS) disk membranes was reported at 55.9°C, whereas a 16°higher temperature of 71.9°C is required to denature Rho (38). Based on our computational analyses, flavonoids could interact with the opsin apoprotein by either binding into the orthosteric site or the spatially distant extracellular vestibule of the receptor, suggesting that flavonoids potentially could have stabilizing effects on opsin similar to those of the retinal chromophore. Interestingly, treatment of ROS membranes containing opsin with aglycone flavonoids resulted in an increase of opsin's melting temperature from 55.4 Ϯ 0.4 to 61.0 Ϯ 0.2°C in the presence of quercetin and to 58.9 Ϯ 0.4°C in the presence of myricetin ( Fig. 3a and Table 2). An increasing dose-dependent stabilizing effect of quercetin was detected at a 0.1-10 M concentration range with a half-maximal effective dose, EC 50 ϭ 0.95 Ϯ 0.05 M. The effect of quercetin was inversely proportional at higher (100 -500 M) concentrations. In the presence of myricetin, the stabilization effect was dosedependent and observed within the 100 -500 M concentration range, reaching the highest melting temperature at the highest concentration evaluated. The calculated EC 50 of myricetin was 180 Ϯ 8.9 M. Despite this stabilizing effect of both quercetin and myricetin, neither of them could reach the level of stabilization produced by the 9-cis-retinal isochromophore. Surprisingly, glycosylated flavonoids, quercetin-3-rhamnoside and myricetrin, showed no beneficial effects on opsin stability (Fig.  3a and Table 2). Moreover, binding of quercetin and myricetin to opsin prior to its incubation with 9-cis-retinal and regeneration of isoRho revealed cooperative effects with an increase of melting temperature to nearly 80°C at the highest doses for both flavonoids, whereas the glycosylated flavonoids showed no beneficial effect ( Fig. 3b and Table 3). Interestingly, none of these tested flavonoids showed any stabilizing effects for retinal-bound isoRho (Fig. 3b, black).
Prolonged incubation of opsin at room temperature results in loss of its ability to bind the retinal chromophore. After 4 h at room temperature, only 50% of opsin could bind 9-cis-retinal

Effects of flavonoids on rhodopsin properties
and form isoRho when compared with opsin kept at 0°C. However, in the presence of flavonoids, due to their stabilizing effects, the same 50% regeneration of isoRho could be reached after extended incubation times. The regeneration half-time of 8 and 5 h was observed for 1 M and 100 M quercetin, respectively. In the case of myricetin, a low 1 M concentration did not improve opsin's half-life. However, at 100 M concentration, the regeneration half-time extended to about 14 h (Fig. 3c). These results correlate with the changes of the opsin's melting temperature detected in the presence of quercetin and myricetin.

Effects of flavonoids on the binding of retinal chromophore
To examine whether flavonoids change the rate of 9-cis-retinal binding to rod opsin, the pigment regeneration after treatment of opsin membranes with each flavonoid was monitored by UV-visible spectroscopy (Fig. 4). The maximum absorption at 487 nm that appeared after incubation of opsin with 9-cisretinal is due to the formation of the Schiff base linkage. This was used to quantify the regenerated isoRho and to calculate the rate of its regeneration. As noted, treatment with flavonoids increased the rate of regeneration in the presence of each tested flavonoid, in a concentration-dependent manner from about 4 to 2 min at the highest (500 M) flavonoid concentrations (see Table 4 for details). This finding suggests that the binding of flavonoids results in a conformational change within the opsin structure permitting better accommodation of the retinal chromophore, and thus its faster binding as compared with nontreated opsin.
The binding of flavonoids to the orthosteric retinal-binding pocket was further confirmed by using the Trp fluorescence to monitor the changes within the chromophore-binding pocket in bovine opsin. In Rho, the intrinsic Trp fluorescence at 330 nm is quenched by the natural ligand 11-cis-retinal and increases upon its light-induced release from the retinal-binding pocket. This increased Trp fluorescence could be quenched by the retinal analogs, including 9-cis-retinal. Molecular docking revealed the close proximity of flavonoid accommodated in the orthosteric-binding pocket to Trp-265 (Fig. 2). Interestingly, the quenching of opsin's Trp fluorescence at 330 nm was observed in the presence of each tested flavonoid in a concentration-dependent manner. Almost full quenching of the Trp fluorescence was reached at the highest flavonoid concentration. Thus, this result suggests that flavonoids can either bind to the chromophore-binding pocket of opsin or their binding to the allosteric site changes the conformation of the orthosteric ligand-binding site. None of the flavonoids prohibited the binding of 9-cis-retinal. Addition of 9-cis-retinal to the flavonoid-bound opsin samples promoted a further decrease of the Trp fluorescence and resulted in the full quenching of Trp fluorescence independent of the flavonoid concentration, suggesting that both flavonoid and 9-cis-retinal could accommodate the opsin molecule (Fig. 5).

Effects of flavonoids on the spectral properties of isoRho
The UV-visible absorption spectra of flavonoid-bound, immunoaffinity-purified isoRho revealed the clear effect of aglycone compounds on its spectral properties, whereas glycosylated flavonoids had no significant effect. Titration with an increasing concentration of both quercetin and myricetin caused about 20% decrease of the absorption maximum and an increase in the ratio of absorption at 280 -487 nm at the highest concentration. This finding suggested incomplete binding of 9-cis-retinal, presumably due to the presence of the flavonoids within the chromophore-binding pocket or a conformational change, introduced by an allosteric modulation, affecting the total amount of the regenerated isoRho (Fig. 6a). Additionally, an increase in absorbance at about 360 nm was observed, which is characteristic for flavonoids and related to increasing flavonoid concentrations. A similar increase in the absorbance at 360 nm was also observed in the immunoaffinity-purified, ligand-free opsin samples treated with these flavonoids (Fig. 6b). No peak within the visible light wavelengths was detected in these flavonoid-bound opsin samples, indicating that flavonoids do not act as visual chromophore analogs when bound to opsin. Figure 1. Bioinformatic analysis of the interaction between bovine rod opsin and the flavonoids. a, assessment of the potential flavonoid-binding sites within the bovine rod opsin structure (PDB code 3CAP) performed with the CASTp 3.0 software is shown on the left. These binding sites were co-validated using a blind docking approach. Three potential binding sites were found: 1) the orthosteric site or retinal-binding site (shown in red); 2) the external binding site between TM5, TM6, and ECL2 pocket 1 (shown in purple); and 3) the external binding site between TM2, TM3, and ECL1, pocket 2 (shown in yellow). The binding energies for each binding site calculated were as follows: 1) Ϫ9.3 kcal/mol; 2) Ϫ7.9 kcal/mol; and 3) Ϫ6.5 kcal/mol. Chemical structures of quercetin, quercetin-3-rhamnoside, myricetin, and myricetrin are shown on the right. b, molecular docking of quercetin into the orthosteric-binding pocket (left) and pocket 2 (middle), and docking of quercetin-3-rhamnoside into the orthosteric-binding pocket (right) of rod opsin. c, molecular docking of myricetin into the orthostericbinding pocket (left) and pocket 2 (middle), and docking of myricetrin into the orthosteric-binding pocket (right) of rod opsin. Quercetin and myricetin could accommodate the retinal-binding pocket and the external binding pocket located between TM5, TM6, and EC L2, whereas their glycosylated forms quercetin-3-rhamnoside and myricetrin could accommodate only into the retinal-binding pocket. Table 1 The details of the interactions between flavonoids and opsin protein side chains within the binding pockets The stable binding of quercetin and myricetin to opsin was confirmed by HPLC analysis. Both quercetin and myricetin were detected in flavonoid-bound immunoaffinity-purified opsin and isoRho samples, with the specific retention time at 19.6 min for quercetin and 8.5 min for myricetin (Fig. 6c). The small change in the retention time of quercetin extracted from the protein samples as compared with the standard (18.5 min) could be due to the oxidation of this molecule during samples preparation. Pure quercetin and myricetin were used as standards.

Effects of flavonoids on function of isoRho
To evaluate the effect of the aglycone flavonoids on the function of the visual receptor, G-protein activation in vitro and in

Effects of flavonoids on rhodopsin properties
situ as well as chromophore release (Meta II decay) assays were performed. The rates of light-induced G t activation by flavonoid-bound purified isoRho were measured in vitro as a Trp fluorescence change due to the G t ␣ dissociation from the isoRho-G t complex upon uptake of GTP␥S. Only a slight increase (3-5%) in the activation rates was detected for both quercetin-and myricetin-bound isoRho treated with low (1-10 M) flavonoid concentrations when compared with isoRho. For isoRho samples treated with higher (100 -500 M) flavonoid concentrations, the G t activation rates were either similar to those of isoRho for quercetin or significantly (30 -60%) decreased at 250 -500 M myricetin, possibly due to receptor structural alterations caused by the flavonoid compounds ( Fig.  7a and Table 5). Interestingly, in the absence of G t , isoRho also

Effects of flavonoids on rhodopsin properties
couples to G i in the heterologous expression systems to induce the G i/o signaling cascade. Thus, we evaluated the effect of quercetin and myricetin on cAMP accumulation levels in response to light stimulation in HEK-293 cells stably expressing opsin. Cells were treated with quercetin or myricetin at two nontoxic (1 or 100 M) concentrations ( Fig. 7b) followed by their regeneration with 9-cis-retinal (Fig. 7c). The levels of cAMP were significantly reduced in cells treated either with 9-cis-retinal alone or 9-cis-retinal followed by the flavonoid treatment, which were exposed to light. However, no difference between these samples was detected. The cellular levels of cAMP were not changed by flavonoids alone either in the dark or after light stimulation. Thus together, these findings suggested no major effect of flavonoids on the Rho-signaling activation within the physiological relevant range of concentrations. The presence of flavonoids had only a mild effect on the rates of light-induced chromophore release (Meta II decay), for both quercetin-and myricetin-bound isoRho treated with low (1-10 M) flavonoid concentrations, and they were 3-10% slower as compared with the decay rates of isoRho ( Fig. 7d and Table 6). However, in samples treated with higher (100 -500 M) flavonoid concentrations, chromophore release was 13-25% faster than in isoRho, most likely due to the structural changes introduced by the bound flavonoid compounds.
Binding of flavonoids had only a minor effect on chromophore release induced by the conformational changes in the chromophore-binding pocket. Flavonoid-bound purified After that, the temperature of melting was determined by using a fluorescence BFC probe. Opsin membranes not incubated with flavonoids, and opsin membranes incubated with 9-cis-retinal were used as controls. The values of fluorescence were plotted as a function of temperature, and the melting temperature was calculated using Prism GraphPad 7.04 software. The derived melting temperature for each experimental condition is shown as a function of concentration. The values of the melting temperatures derived for nontreated controls (opsin and isoRho) were 55.4 Ϯ 0.4 and 72.1 Ϯ 1.2°C, respectively. This experiment was repeated three times. Error bars represent standard deviation (S.D.). All values of the melting temperatures derived from these experiments and statistical significance of the effect of flavonoids are shown in Table 2. b, opsin membranes were incubated with different concentrations (1, 10, 100, 250, and 500 M) of quercetin, quercetin-3rhamnoside, myricetin, or myricetrin for 1 h at room temperature and then regenerated with 10 M 9-cis-retinal (color), or isoRho was regenerated first, and then membranes were incubated with the flavonoid compounds (black). The melting temperature was determined in these samples by using a BFC fluorescence probe. These experiments were performed in triplicate. Error bars represent S.D. All values of the melting temperatures derived from these experiments and statistical significance of the effect of flavonoids are shown in Table 3. c, regeneration of isoRho from aged opsin. Opsin-containing membranes were incubated with quercetin or myricetin for 1 h on ice, solubilized with DDM, and then kept at room temperature for 2, 4, 6, or 24 h before regeneration with 10

Table 2 Melting temperature of opsin in the presence of flavonoids
The melting temperature was determined for nontreated (NT) controls: opsin, isoRho, and opsin incubated with the flavonoid compound at 0.01-500 M concentrations before the measurement. Each flavonoid-treated opsin sample was compared with the nontreated opsin sample. Statistical significance (p) was calculated with the one-way ANOVA test and Dunnett's post-test for multiple comparisons (GraphPad Prism 7 software). Statistically significant effect of flavonoid was denoted with asterisks: ***, p Յ 0.0001; **, p Յ 0.001.

Quercetin Quercetin-3-rhamnoside Myricetin Myricetrin
Opsin ( Table 3 Melting temperature of isoRho in the presence of flavonoids The melting temperature was determined for isoRho and opsin incubated with the flavonoid compound at 1-500 M concentrations followed by its regeneration with 9-cis-retinal before the measurement. Each flavonoid-treated sample was compared with the nontreated isoRho sample. Statistical significance (p) was calculated with the one-way ANOVA test and Dunnett's post-test for multiple comparisons (GraphPad Prism 7 software). Statistically significant effect of flavonoid was denoted with asterisks: ***, p Յ 0.0001; **, p Յ 0.001.

Effects of flavonoids on rhodopsin properties
isoRho released chromophore at rates comparable with those of purified isoRho when incubated at a high constant temperature of 55°C ( Fig. 8 and Table 6).

Effects of flavonoids on the expression and distribution of opsin
Although flavonoids increased opsin stability, the following question arose: do they also improve the folding and membrane targeting of misfolded opsin mutants? To answer this question, the effects of quercetin and myricetin on the expression, glycosylation pattern, and membrane localization of opsin was tested in NIH-3T3 cells stably expressing a P23H rod opsin mutant.
The results were compared with WT rod opsin. In humans, the P23H Rho mutation associates with autosomal dominant retinitis pigmentosa due to the structural instability of opsin. In mammalian cells, P23H rod opsin features immature glycosylation and impaired membrane trafficking, which could be corrected by 9-cis-retinal isochromophore added to the cell culture during opsin biosynthesis (Figs. 9b and 10, a and b) or a recently discovered nonretinoid chaperone YC-001 (13). To test whether a comparable effect could be induced by flavonoids, cells expressing either WT or P23H rod opsin were treated with either 1 or 100 M quercetin or myricetin (concentrations within nontoxic range (Fig. 9, a and e)), and then mem-  Table 4.

Table 4 Regeneration rates or isoRho in the presence of flavonoids
Regeneration of isoRho with 10 M 9-cis-retinal after opsin incubation with the flavonoid compound at 1-500 M concentrations. The isoRho regeneration half-time (t1 ⁄ 2 ) of each flavonoid-treated sample was compared with the nontreated (NT) control sample. Statistical significance (p) was calculated with the one-way ANOVA test and Dunnett's post-test for multiple comparisons (GraphPad Prism 7 software). Statistically significant effect of flavonoid was denoted with asterisks: ***, p Յ 0.0001; **, p Յ 0.001; *, p Յ 0.05.

Effects of flavonoids on rhodopsin properties
brane localization of opsin was determined in nonpermeabilized cells by fluorescent immunostaining and high-content image analysis. Interestingly, treatment with flavonoids (100 M quercetin and 1 M myricetin) resulted in the noticeable movement of the P23H rod opsin mutant to the cell surface as compared with the nontreated cells ( Fig. 9, b-d). Moreover, the relative ratio of the cell surface to the total protein expression also was significantly increased in these conditions (p Յ 0.0001). As noted, 9-cis-retinal resulted also in an increased membrane fluorescence of WT opsin as compared with the nontreated cells (Fig. 9f). Moreover, both quercetin and myricetin at either 1 or 100 M concentration resulted in similar, although slightly lesser effects as compared with 9-cis-retinaltreated cells (Fig. 9f). Quantification of the fluorescence at the plasma membrane in nonpermeabilized cells and the total opsin expression in the permeabilized cells indicated increased membrane localization of WT opsin related to the increased total opsin levels upon treatment with 9-cis-retinal and both flavonoids ( Fig. 9, g and h). However, the relative ratio of the cell surface to the total protein expression for WT opsin was similar in all the cells regardless of the applied treatment ( Fig. 9, g and h). Together these findings indicated that flavonoids could enhance biosynthesis of both WT opsin and the P23H RP-linked opsin mutant and improve their membrane targeting in the absence of retinal chromophore. The SDS-polyacrylamide gel and immunoblotting analyses showed that flavonoids did not affect the glycosylation pattern of WT and P23H rod opsin mutant (Fig. 10). Opsin deglycosylation resulted in a change of opsin migration within the SDSpolyacrylamide gel from about 72 to 55 kDa for P23H rod opsin and 50 to 26 kDa for WT rod opsin. Interestingly, higher concentrations of quercetin led to the formation of WT rod opsin oligomers or aggregates appearing at about 100 and 140 kDa, which were resistant to the treatment with PNGaseF deglycosylase. Thus, this finding suggested the potential role of quercetin in modulation of rod opsin dimeric/oligomeric organization.

Effect of flavonoids on the membrane oligomeric organization
Rod opsin has the ability to self-associate, forming dimers and/or higher-ordered oligomers within the membranes (39,40). To understand whether flavonoids could modulate the oligomeric state of rod opsin in live-cell membranes, the bioluminescence resonance energy transfer (BRET) assay was performed in HEK-293 cells stably expressing both opsin⅐Rluc (donor) and opsin⅐Venus (acceptor). Likewise for HEK-293S

Effects of flavonoids on rhodopsin properties
GnTI

Effects of flavonoids on rhodopsin properties
detected in the nontreated control cells after 16 h, which is likely due to the increased cell number (Fig. 11a, left panel). However, no change in the BRET signal was found at earlier time points (2, 4, and 8 h) in the nontreated cells (Fig. 11a, left panel, black bars). Thus, these results suggested the stimulating effect of flavonoids on opsin dimerization/oligomerization. The treatment with quercetin for 2 h resulted in ϳ30% elevation of the BRET signal, which increased to ϳ45% after 16 h of incubation as compared with the nontreated control cells (Fig.  11a, left panel, blue bars). The effect of myricetin was slightly less prominent with ϳ18 and ϳ30% increase of the BRET signal after 2 and 16 h of incubation, respectively (Fig. 11a, left panel, red bars). These flavonoid-stimulated dimers/oligomers were sensitive to the detergent and could be disrupted upon addition of DDM, resulting in a decrease in the BRET signal (Fig. 11a, middle and right panels). Interestingly, neither quercetin nor myricetin had an effect on dimerization/oligomerization of isoRho in the cells regenerated with 9-cis-retinal prior to the treatment with flavonoids, thus supporting the in vitro results and indicating that flavonoids could affect properties of ligandfree opsin but not chromophore-bound isoRho (Fig. 11b). Furthermore, to confirm these results, the formation of flavonoidstimulated rod opsin dimer/oligomers was also determined in the transiently transfected HEK-293 cells with a 1:4 donor to *** *** *** *** *** acceptor ratio. As determined previously (41,42) and in this study (Fig. 11c), the BRET signal in cells co-expressing opsin⅐Rluc and opsin⅐Venus in a 1:4 ratio was maintained within a linear range. Interestingly, quercetin and myricetin enhanced BRET at all examined donor to acceptor ratios (1:2, 1:4, 1:6, and 1:8) with the highest effect at the 1:2 and 1:4 ratios. The effect of flavonoids was less prominent at the opsin expression levels, resulting in the saturation of the BRET signal (Fig.  11c). Likewise, in the cells stably expressing opsin⅐Rluc and opsin⅐Venus and in the cells transiently co-expressing these proteins at a 1:4 ratio, a significant (p Յ 0.0001) time-dependent increase of the BRET signal upon incubation with either quercetin or myricetin (100 M) was detected. A small increase of BRET was detected in the nontreated control cells after 16 h, which is likely due to the increased cell number (Fig. 11d, left panel, black bars). However, no significant change in the BRET signal was found at earlier time points (2, 4, and 8 h) in nontreated cells. The opsin oligomers were DDM-sensitive and could be disrupted by the detergent resulting in a decrease of the BRET signal (Fig. 11d, right panel) consistent with the results obtained in cells stably expressing rod opsin constructs.
To ensure the specificity of the flavonoid-associated effect on rod opsin dimerization, we performed an additional control experiment, in which HEK-293 cells were transiently co-trans-  Table 6.

Table 5 Effects of flavonoids in G t activation rates
The G t activation rates (k) were determined for purified isoRho samples, either nontreated (NT) or treated with the flavonoid compound at 1-500 M concentrations prior to the regeneration with 10 M 9-cis-retinal. Statistical significance (p) was calculated with the one-way ANOVA test and Dunnett's post-test for multiple comparisons (GraphPad Prism 7 software). Statistically significant effect of flavonoid was denoted with asterisks: ***, p Յ 0.0001; **, p Յ 0.001.   The changes in the absorbance maximum were calculated as a percentage of residual pigment assuming the absorbance at the initial point as 100%. The changes were plotted as a function of time, and the half-lives (t1 ⁄2 ) of chromophore release were calculated using these plots. These measurements were performed in triplicate. Error bars represent S.D.

Effects of flavonoids on rhodopsin properties
fected with opsin⅐Rluc (donor) and an unrelated but localized to the plasma membrane protein Kras⅐Venus (acceptor) at a 1:4 donor to acceptor ratio. Confirming our previous observation (41,42), the BRET signal detected in cells co-transfected with these constructs was due to co-localization of opsin and Kras at the plasma membrane; however, it was significantly lower than

Effects of flavonoids on rhodopsin properties
the BRET signal detected in cells co-expressing opsin⅐Rluc and opsin⅐Venus. Interestingly, treatment with either quercetin or myricetin had no effect on the opsin⅐Rluc/Kras⅐Venus BRET pair signal, justifying specificity of the flavonoid-mediated increase in rod opsin dimerization (Fig. 11e).

Discussion
Steady biosynthesis and folding of Rho are critical for the formation of ROS in the retinal tissue of the eye, its correct morphology, and its function. Imbalanced biogenesis or misfolding of this visual receptor associates with retinal degeneration and visual impairments in several pathologies, including RP (43). Pharmacological chaperones could preclude these pathological effects by enhancing Rho stability, improving folding, and restoring its function. Thus, development of novel analogs of the native retinal chromophore, 11-cis-retinal, gained appreciable interest within the past decade (44 -48). Multiple investigations showed the therapeutic potential of 9-cis-retinal

Effects of flavonoids on rhodopsin properties
and 11-cis-6-membered ring-retinal analogs against retinopathies based on in vitro and in vivo studies in animal models of retinal diseases (49,50). Supplementation with a precursor of 11-cis-retinal chromophore, vitamin A, also displayed beneficial effects in patients with RP (51). However, prolonged retinoid therapy could disturb the retinoid homeostasis and cause toxicity (52). To overcome this limitation, the development of nonretinoid pharmacological chaperones is indispensable. A discovery of a novel nonretinoid modulator of rod opsin, YC-001, exhibiting promise in treating retinal disorders linked with disrupted Rho homeostasis was recently reported (13). However, further studies developing a specific retina-targeting delivery system of YC-001, enabling steady release of this compound in the eye, is still required due to its fast clearance rates. An alternative promising approach for preventing or delaying retinal degeneration could be the use of natural products as a source of the bioactive compounds such as flavonoids. Indeed, these polyphenolic plant compounds could improve sight in several eye-related diseases (19 -23). However, the mechanism of action of these compounds is still not fully understood. As suggested previously, flavonoids could potentially interact with rod opsin and act as allosteric modulators of its function (27)(28)(29)53). Yet, a comprehensive study deciphering the underlying mechanism enhancing vision by this class of compounds is lacking. Thus, to clarify the role of flavonoids in modulating rod opsin properties, we studied the effect of four flavonoids: quercetin, myricetin, and their glycosylated forms quercetin-3rhamnoside and myricetrin on opsin stability, function, and its oligomeric membrane organization by using a combination of computational, biochemical, and biophysical approaches. Previous molecular docking studies predicted prospective binding sites on the surface of rod opsin (27,53) in addition to the retinal-binding pocket. Our bioinformatic analysis of the interaction site(s) between flavonoids and ligand-free opsin revealed two potential binding regions with the lowest binding energy: located within the helical bundle at an orthosteric chromophore-binding pocket, and an external binding pocket positioned between TM5, TM6, and ECL2. Interestingly, the external binding pocket could only be found in the ligand-free opsin but not in the retinal-bound Rho structure, most likely due to the structural rearrangements involving movements of TM5 and TM6 occurring in response to the light stimulus. Followed by these computational analyses, first we investigated the effect of flavonoids on the stability of rod opsin. Ligand-free rod opsin is highly unstable. However, its stability greatly enhances upon binding of either the native 11-cis-retinal or 9-cis-retinal isochromophore (38). Interesting, aglycone flavonoids, quercetin, and myricetin, but not their glycosylated forms, both displayed enhancement of the opsin stability effects, presumably due to an increased rigidity of the protein structure by the bound flavonoid. The increased chemical stability in the presence of quercetin was also reported for the G90V RP-linked Rho mutant associated with reduced protein stability (27). Interestingly, the opsin stabilizing effect was achieved at different concentration ranges of quercetin and myricetin. Quercetin positively modulated opsin stability at low micromolar concentrations, but at higher concentrations it displayed an inverted effect. In contrast, myricetin showed a dose-dependent opsin-stabilizing effect with its maximum at the highest concentration evaluated. This difference in the dose response can suggest different binding modes of quercetin and myricetin to the opsin structure. Although the dual effect of quercetin is likely related to its binding to both the orthosteric-binding site and the allosteric-binding site, the effect of myricetin is rather associated with its binding to the orthosteric-binding pocket. Furthermore, flavonoids and 9-cis-retinal displayed a cooperative effect, resulting in an even higher melting temperature, as compared with isoRho. This finding signified additional tightening of the opsin structure by bound flavonoids. In contrast, the presence of flavonoids did not affect the stability of isoRho, indicating that specific flavonoid-binding site(s) emerge only upon the conformational rearrangements associated with chromophore release from the retinal-binding pocket. In fact, potential allosteric binding sites emerge only in the opsin conformation. Moreover, the chromophore-binding pocket is vacant only in opsin, and it could be occupied by the flavonoid compound.
The bound flavonoid did not prevent regeneration of isoRho with 9-cis-retinal but rather it enhanced the rates of retinal binding to opsin in agreement with the previously reported effects of quercetin (27) and the anthocyanin, cyanidin-3-glucoside (28), on the regeneration of the visual pigment. Based on the 1 H NMR study reported for cyanidin-3-glucoside, this enhanced pigment regeneration was mediated by the direct interaction of flavonoid with opsin, resulting in the modulation of the structural conformation and dynamics of opsin by restricting the mobility of some residues (53). Interestingly, both quercetin and myricetin resulted in a similar dose-dependent increase in the rate of pigment regeneration. Thus, the allosteric modulation of opsin properties by myricetin cannot be excluded.
Despite the fact that flavonoids can directly interact with opsin, modulating its structure, increasing its stability, and enhancing the regeneration of the visual pigment, their effects on receptor function are rather modest. At physiological concentrations, flavonoids tested in this study had no significant impact on the rates of G t activation, the activation of the signaling cascade in cultured cells, or the light-stimulated chromophore release. Although myricetin at higher (250 and 500 M) concentrations resulted in the inhibition of the G t activation rates, these high doses of myricetin are unlikely to be considered for a therapeutic application due to its high cellular toxicity at concentrations above 100 M. Furthermore, the effect of flavonoids on the function of the specific RP-linked Rho mutants still needs to be elucidated.
Both an increase of rod opsin stability and enhancement of retinal binding mediated by flavonoids potentially contribute to vision-improving effects in the RP-linked diseases (22,54). A large group of RP-causing Rho mutations is accompanied by misfolding and arrest in the secretory pathway of the inner segments of rod photoreceptor cells, preventing proper transport of Rho to the ROS, and thus leading to the photoreceptor cells' death. The P23H RP Rho mutant is associated with decreased protein stability and misfolding. However, these defects could be mitigated by pharmacological treatments not only with 9-cis-retinal isochromophore but also with the nonretinoid

Effects of flavonoids on rhodopsin properties
modulator YC-001 (13). Both compounds increase the stability of P23H rod opsin mutant, its folding, and membrane targeting. Flavonoids, due to their opsin-stabilizing effects, could potentially provide analogous benefits to this RP-linked Rho mutant. Indeed, incubation of cells stably expressing P23H rod opsin with flavonoid compounds improved its mobility to the plasma membrane as compared with nontreated control cells. However, although treatment with 9-cis-retinal or YC-001 improved the glycosylation pattern of P23H opsin, this effect was not detected for flavonoids. Nevertheless, as noted in this study, flavonoids enhanced the expression level of opsin, both in WT and the P23H opsin mutant. Thus, the mechanism of increased membrane mobility mediated by flavonoids potentially relates to the increased expression and higher cellular levels of this protein. In the native tissue, such enhanced expression of Rho with increased ability of retinal binding could be helpful in maintaining the proper balance of functional protein. Indeed, enhanced expression of Rho was observed in a mouse model of endotoxin-induced uveitis upon treatment with anthocyanin-rich bilberry extract and resulted in suppression of the shortening of ROS in photoreceptor cells, and thus the preservation of retinal degeneration and vision loss (54).
Enhanced rod opsin stability facilitated by flavonoids could be related to opsin's increased self-association and formation of the condensed oligomers within the cell membrane. Accumulated research on the Rho supramolecular membrane organization documented that Rho forms dimers organized in higherordered oligomers tightly packed in the native ROS membranes of photoreceptor cells (39,40,55,56). In the heterologous expressionsystem,opsinalsoexistsinequilibriumbetweenmonomers and dimers that could be shifted toward increased dimer population by manipulating the opsin expression level (57). Rho within the isolated ROS membranes or in a form of the extracted oligomers is much more stable than monomeric Rho (40,42,55). Additionally, exogenous phospholipids increase the stability of detergent-solubilized Rho (58). In the presence of quercetin, oligomers of opsin were detected in SDS-PAGE, suggesting that flavonoids, especially quercetin, stimulated either aggregation or oligomerization of opsin. In fact, small organic molecules, as well as natural products, could promote protein aggregation, often resulting in false inhibition or activation of these proteins, including GPCRs (30). Thus, to discriminate between nonspecific flavonoid-stimulated aggregation and oligomerization of opsin, a BRET assay, commonly used in studying GPCR oligomerization, was performed in live cells either stably or transiently expressing opsin. Indeed, quercetin and to a lesser extent myricetin promoted oligomerization of opsin in a time-dependent manner. Interestingly, no effects on oligomer formation were observed in the cells regenerated with 9-cis-retinal prior to the treatment with flavonoids, thus supporting other results indicating that flavonoids bind to and modulate the properties of the ligand-free opsin only. Moreover, the flavonoid-stimulated increased BRET signal occurred due to the formation of opsin oligomers rather than aggregates as they were disrupted with the DDM detergent, in opposition to the RP-linked opsin aggregates that are not sensitive to such treatment (59). Thus, this finding suggests that increased opsin stability could be related to the formation of tightly packed oligomers within the biological membranes.
Altogether, this study suggests that flavonoid compounds enhance rod opsin stability through direct binding to the ligand-free opsin molecule and modulate its conformation by rigidifying its structure, and promote opsin self-association within the phospholipid bilayer. Different patterns of responses to the treatment with quercetin and myricetin suggest differential binding of these flavonoids to opsin. Although the effects of quercetin are likely associated with its dual effect through the binding to both the orthosteric-binding pocket and allosteric site, the effect of myricetin is rather related to its accommodation in the orthosteric-binding pocket. The difference in the interaction pattern of quercetin and myricetin with Trp-265 within the binding pocket also emerged from the molecular docking. Although quercetin interacted with Trp-265 via van der Waals interactions, myricetin formed much strongerstacking interactions. However, more advanced structural studies, including crystallography, molecular dynamic simulations, and/or MS-based techniques, are required to learn about flavonoid-opsin interactions in more detail. Moreover, this work supports the therapeutic potential of flavonoids as lead compounds to discover novel nonretinoid medications for treating or delaying retinal degeneration associated with disturbed stability and imbalanced homeostasis of the rod photoreceptors. However, more studies are necessary to fully understand the pharmacological potential and the underlying mechanism of flavonoids in the prevention of retinopathies associated with Rho misfolding.

Molecular docking
The coordinates of the ligand-free bovine opsin at 2.9 Å resolution were obtained from PDB code 3CAP (33). One monomeric unit of the protein was selected, and the co-crystallized molecules and all crystallographic water molecules were removed from the coordinate set; hydrogen atoms were added,

Effects of flavonoids on rhodopsin properties
and partial charges were assigned to all atoms. The protein was then submitted to the restrained molecular mechanics refinement with NAMD 2.12 software (60), using the CHARMM22 force field (61). After further energy minimization, the protein was suitable for bioinformatic analysis performed with CASTp 3.0 software using the http://sts.bioe.uic.edu/castp/server 3 (76). The potential ligand-binding pockets located in the extracellular part of the protein and the orthosteric ligand-binding site were selected and then verified by "a priori" docking approach with quercetin using the Achilles Blind Docking server. The 3D structures of the flavonoid compounds used in a docking study were obtained from the PubChem database. The molecular docking of these compounds into the opsin structure was performed with VINA/Vegas 3.1.0.21 software (62) with 30 iterations conducted for each compound. The results were prioritized according to the predicted binding free energy in kilocalories/mol. The results obtained from the docking simulation were visualized with the Biovia Discovery Studio Visualizer 17.2.0 software.

Preparation of opsin membranes
ROS membranes were isolated from frozen bovine retinas under dim red light as described previously (63). The buffer is composed of 10 mM sodium phosphate, pH 7.0, and 50 mM hydroxylamine was used to resuspend ROS membranes to Rho concentrations of ϳ3 mg/ml. The membranes then were exposed to white light with a 150-watt bulb for 30 min at 0°C. These membranes were then pelleted by centrifugation at 16,000 ϫ g for 10 min. The supernatant was discarded, and the membrane pellet was washed twice with 10 mM sodium phosphate, pH 7.0, and 2% BSA followed by four washes with 10 mM sodium phosphate, pH 7.0, and two washes with 20 mM BTP, pH 7.5, and 100 mM NaCl. To pellet the membranes after each wash, centrifugation at 16,000 ϫ g for 10 min was applied.

Thermal shift assay
The BFC probe was diluted to a stock concentration of 10 mM in DMSO. A final working concentration of 2 mM BFC in 20 mM BTP, pH 7.5, and 100 mM NaCl was used in all experiments. The opsin membranes at a concentration of 0.01 mg/ml were loaded onto a 96-well plate (Applied Biosystem). Then, specific flavonoid at 1, 10, 100, 250, and 500 M concentration was added, and the plate was incubated for 1 h at 4°C. Alternatively, 30 min prior to the addition of flavonoid, the regeneration of isoRho with 9-cis-retinal was performed. Next, 5 l of the BFC probe was added to each well, which contained 20 l of opsin membrane with or without the flavonoid compound. The plate was sealed with a ClearSeal film (HR4-521) from Hampton Research and incubated for 10 min on ice prior to measurement of their fluorescence. All measurements were performed with a StepOnePlus Real-Time PCR System (Applied Biosystems), and melting curve experiments were recorded using StepOne software version 2.3. The fluorescence in the SYBR, FAM, and ROX channels was recorded for each sample. The run was set to cool the plate to 4°C within 10 s, kept at 4°C for 1 min, and then increased 1°C per min in a step-and-hold manner up to 99.9°C. The multicomponent data were exported to a Microsoft Excel sheet and analyzed with Prism 6.0. The melting temperatures of bovine Rho and opsin within the membranes were 71.9 and 55.9°C, respectively (38).

Pigment reconstitution and purification by 1D4 immunoaffinity chromatography
Bovine opsin membranes were resuspended in 20 mM BTP containing 120 mM NaCl, pH 7.5, and incubated with a specific flavonoid at 1, 10, 100, 250, and 500 M concentration for 1 h at 4°C. Then, opsin was regenerated with 9-cis-retinal, which was added to the membrane suspension from a DMSO stock solution to a final concentration of 10 M followed by an incubation in the dark for 1 h at 4°C on a nutator. To solubilize the membranes, DDM was added to 20 mM final concentration and incubated for 1 h at 4°C on a nutator. The lysate was centrifuged at 16,000 ϫ g for 1 h at 4°C, and isoRho was purified from the supernatant by 1D4 immunoaffinity chromatography (using anti-Rho C-terminal 1D4 antibody (a generous gift from Dr. K. Palczewski, University of California-Irvine, CA) immobilized on cyanogen bromide (CNBr)-activated agarose) (64). Three hundred l of 6 mg of 1D4/ml agarose beads were added to the supernatant and incubated for 1 h at 4°C on a nutator. The resin was then transferred to a column and washed with 10 ml of 20 mM BTP, 120 mM NaCl, and 2 mM DDM, pH 7.5. The isoRho was eluted with the same buffer supplemented with 0.6 mg/ml of the 1D4 peptide (TETSQVAPA).
Alternatively, the ability to stabilize rod opsin over a longer period of time was tested. Opsin membranes (2.5 M) were incubated with or without 1 or 100 M flavonoid for 1 h on ice, followed by their solubilization in 20 mM DDM. Solubilized opsin samples were incubated at room temperature for 0, 2, 4, 6, and 24 h before incubation with 10 M 9-cis-retinal in the dark for isoRho regeneration. Then UV-visible spectra of these samples were measured.

UV-visible spectroscopy of opsin and isoRho
The concentration of Rho or opsin within membranes was measured after membrane solubilization with 20 mM DDM and pelleting insoluble material by centrifugation at 16,000 ϫ g for 15 min at 4°C. A UV-visible spectrophotometer (Cary 50, Varian, Palo Alto, CA) and the absorption coefficients ⑀ 500 nm ϭ 40,600 M Ϫ1 cm Ϫ1 and ⑀ 280 nm ϭ 81,200 M Ϫ1 cm Ϫ1 (65) were used for Rho and opsin, respectively. The concentration of isoRho was measured and quantified in freshly purified samples by using a UV-visible spectrophotometer and the absorption coefficient ⑀ 485 nm ϭ 43,600 M Ϫ1 cm Ϫ1 (66).

HPLC flavonoid detection
To detect the flavonoids bound to opsin, opsin membranes were incubated with quercetin or myricetin at a concentration of 100 M for 1 h at 4°C. These samples were divided, and half was regenerated with 9-cis-retinal. The excess flavonoid was removed by ID4 immunoaffinity purification of opsin and isoRho. Protein-bound flavonoids were extracted from the purified samples in 1 ml of methanol/acetone (1:1, v/v) by vigorous shaking followed by centrifugation at 3,220 ϫ g for 15 min at 4°C to separate the precipitated material. The supernatants were filtrated through a 0.22-m filter to glass vials and dried in a Savant speedvac concentrator (Thermo Fisher Scientific, Waltham, MA). Then, these samples were dissolved in 300 l of methanol, and 100 l of each was injected into an Agilent 1260 Infinity II HPLC system. Flavonoids were separated on a reverse-phase analytical HPLC column (Gemini RPC18 5 m, 4.6 ϫ 250 mm) equilibrated with 50% (v/v) methanol in water (0.1% H 3 PO 4 ) with a gradient elution from 50 to 80% methanol for 15 min and then an isocratic elution at a flow rate of 0.4 ml/min for 10 min. The absorbance signals at 258 and 360 nm were collected, and the flavonoids were identified by a comparison with an elution profile of authentic standards.

Fluorescence spectroscopy
The binding of flavonoid into the opsin's chromophorebinding pocket was determined by measuring the quenching of the intrinsic Trp fluorescence after the addition of increasing concentrations of the flavonoid compounds. Then 9-cis-retinal was added, and the measurements were repeated. The emission spectra were recorded with an L55 luminescence spectrophotometer (PerkinElmer Life Sciences) at 20°C between 300 and 450 nm after excitation at 295 nm. The excitation and emission slit bands were set at 5 and 10 nm, respectively. Changes in the intrinsic Trp fluorescence at 330 nm were plotted as a function of the flavonoid concentration. All experimental data were corrected for the samples' background and self-absorption at excitation and emission wavelengths (inner filter effect correction).

G t activation
G t was extracted from ROS membranes isolated from 200 dark-adapted bovine retinas followed by its purification as described previously (67,68). Activation properties of the isoRho and flavonoid-bound isoRho were tested in a Trp fluorescence G t activation assay. Briefly, a mixture of G t and isoRho samples at a 10:1 ratio (250 nM G t and 25 nM isoRho) diluted in a buffer consisting of 10 mM BTP, 120 mM NaCl, and 1 mM DDM, pH 7.0, was illuminated for 30 s with a Fiber-Light illuminator through a band-pass wavelength filter (480 -520 nm) followed by a 5-min incubation with continuous low-speed stirring. G t activation was recorded as the intrinsic fluorescence increase from G t ␣ upon addition of 5 M GTP␥S. Measurements were performed with an L55 luminescence spectrophotometer (PerkinElmer Life Sciences). Excitation and emission wavelengths were set at 300 and 345 nm, respectively (69 -71). In the control experiments, no signals from isoRho without G t were detected.

Chromophore release
Changes in the intrinsic Trp fluorescence were measured with 50 nM purified isoRho or flavonoid-bound isoRho samples diluted in a buffer consisting of 10 mM BTP, 100 mM NaCl, and 1 mM DDM, pH 6.0, after their illumination with a Fiber-Light illuminator through a 420 -520-nm bandpass filter for 15 s. Light exposure was conducted at a distance of 15 cm. Changes in the intrinsic Trp fluorescence were recorded for 60 min, and they correlate with the decrease in the protonated Schiff base concentration (72). These measurements were performed with an L55 fluorescence spectrometer (PerkinElmer Life Sciences) at 20°C. Spectrofluorometer slit settings were 8 nm at 295 nm for excitation and 10 nm at 330 nm for emission collection.

Thermal stability
Purified isoRho or flavonoid-bound isoRho samples diluted in a final volume 0.4 ml of 20 mM BTP, 120 mM NaCl, and 1 mM DDM, pH 7.5, were incubated at 55°C in the dark, and their spectra were recorded every 2 min for 1 h. Absorbance at the maximum wavelength at the initial time point was assumed to be 100%. The percentages of remaining pigments were normalized to their initial concentrations and then plotted as a function of time. From these plots, the half-lives (t1 ⁄ 2 ) of chromophore release were calculated.

Cell culture
NIH-3T3 cells stably expressing WT mouse opsin, HEK-293S GnTI Ϫ cells stably expressing WT mouse opsin, and HEK-293 cells stably expressing both WT mouse opsin⅐Rluc and opsin⅐Venus were cultured in DMEM with 10% FBS (Hyclone, Logan, UT) and 1 unit/ml penicillin with 1 g/ml streptomycin (Life Technologies, Inc.) at 37°C under 5% CO 2 according to the instructions from the ATCC Animal Cell Culture Guide.

Cytotoxicity assay
The above-described cells were seeded in 96-well plates at a density of 3 ϫ 10 4 cells/well. The next day, different concentrations of the flavonoid compounds were added to the cells, and 24 h later, the cell viability was evaluated by using the MTT cell proliferation assay (Sigma) (73). The percentage of dead cells under tested conditions was calculated.

cAMP detection
HEK-293S GnTI Ϫ cells stably expressing WT mouse opsin were plated in two 96-well plates at a density of 90,000 cells per well in 85 l of DMEM containing 10% FBS and antibiotics. The cells were treated with flavonoids at a different concentration for 16 h. Next, 9-cis-retinal was added for 2 h to regenerate isoRho in cells either treated or not with the flavonoid compounds. After that, forskolin (5 M) and phosphodiesterase inhibitor (0.1 M) were added to maximize the concentration of total cAMP within the cells. One plate then was kept in the dark, whereas the second plate was exposed to bright (150 watts) light for 15 min from a 10-cm distance. Levels of accumulated cAMP were detected with the cAMP Direct Biotrak EIA kit (GE Healthcare) following the manufacturer's protocol and the absorbance readout at 630 nm by the Tecan M1100 plate reader (Tecan Life Sciences) at the final step of the assay, according to the protocol provided.

Membrane localization of rod opsins in HEK-293 cells
NIH-3T3 cells expressing WT opsin or P23H mutant opsin together with GFP (a generous gift from Dr. K. Palczewski, University of California-Irvine) were plated in a 96-well plate at a density of 2.0 ϫ 10 5 cells/well and cultured overnight. The next day, the cells were treated with the flavonoids at different concentrations for 16 h. For the regeneration of isoRho, the cells were incubated with 9-cis-retinal for 2 h. The cell growth Effects of flavonoids on rhodopsin properties medium was removed from each well, and the cells were washed three times with PBS. The cells were fixed with 3% formaldehyde freshly prepared in PBS for 20 min at room temperature and then they were washed two times with PBS. Next, the cells were incubated in 10% normal goat serum in PBS for 1 h at 37°C. To detect opsin, cells were incubated with the B6-30 anti-Rho antibody that recognizes the N-terminal epitope (a generous gift from Dr. K. Palczewski, University of California-Irvine) for 3 h at room temperature. Then, cells were washed three times with PBS for 10 min at room temperature. To detect total opsin, cells were permeabilized with Triton X-100 before immunostaining. Opsin immunostaining was visualized by incubating the cells with an anti-mouse antibody conjugated with Alexa Fluor 594 (Thermo Fisher Scientific) at a 1:100 dilution for 1 h at room temperature. After that, the cells were washed three times with PBS for 10 min at room temperature. The cell nuclei were stained with DAPI, following the manufacturer's protocol. The plate was sealed with a transparent film, and the cells were imaged with the Operetta High Content Imager (PerkinElmer Life Sciences) using a ϫ20 long objective. Eight fields were taken of each well for cell images with four channels, including bright field, GFP, Alexa Fluor 594, and DAPI. Images were analyzed with the Columbus storage and analysis system (PerkinElmer Life Sciences). DAPI fluorescence images were used to define nuclei and count cells. Bright field images and GFP fluorescence were employed to define cells and select populations of intact cell images. The plasma membrane was defined within Ϯ5% of the cell border.

BRET assay
Stable HEK-293 cells expressing both opsin⅐Rluc (donor) and opsin⅐Venus (acceptor) were plated at a density of 8 ϫ 10 4 cells/ml into a 96-well plate. The next day, these cells were treated with flavonoid compounds at a 100 M final concentration for 2, 4, 8, or 16 h. During this treatment, cells were cultured at 37°C with 5% CO 2 and 90% humidity. Next, 9-cisretinal at a final concentration of 10 M was added, and the plates were wrapped with aluminum foil and incubated for 2 h at 37°C with 5% CO 2 and 90% humidity. Then, the culture medium was aspirated and replaced with 200 l of PBS/well. Cells were resuspended and transferred from the 96-well cell culture plate to a white-walled opaque 96-well plate (Corning Life Sciences). DDM in PBS at a final concentration of 5 or 0.1 mM was added to the wells (59). Coelenterazine h was diluted to 25 M in PBS from a 2.5 mM stock solution. Each well of the 96-well plate was injected with 25 l of diluted coelenterazine h followed by dual luminescence readings at 480 and 530 nm. The measurements were performed with a Tecan M1100 plate reader (Tecan Life Sciences). The BRET signal was calculated as the ratio of emission at 530 and 480 nm. The average of three independent experiments was used to obtain the final plot.
Alternatively, HEK-293 cells were plated into two 12-well plates at ϳ25 ϫ 10 4 cells/ml. The plates were cultured at 37°C with 5% CO 2 and 90% humidity. The next day, cells were transiently transfected either with both opsin⅐Rluc (donor) and opsin⅐Venus (acceptor) at a 1:4 earlier employed ratio of donor to acceptor constructs (41) or opsin⅐Rluc construct only using polyethyleneimine (74,75). Constructs of mouse opsin fused to Venus (opsin⅐Venus) and Renilla luciferase (opsin⅐Rluc) in the pcDNA3.1Zeo vector were a generous gift from Dr. N. A. Lambert (Georgia Regents University, Augusta, GA).
The cells were treated with the flavonoid compounds as described above for the stable cells. On the following day, 48 h post-transfection, the culture medium was aspirated and replaced with 500 l of PBS/well. Cells were resuspended, and 200 l of this suspension was transferred to a white-walled opaque 96-well plate (Corning, NY). DDM in PBS at a final concentration of 5 mM or 0.1 mM was added to the assigned wells (59). The measurement of the BRET signal was performed as described above for the stable cell line.

Statistical analyses
The cytotoxicity assays, image-based analyses with the Operetta High Content Imager, quantification of cAMP levels, and the BRET analyses included three biological replicates and were performed at least twice. Positive and negative controls were included in each assay. Effects of the tested flavonoid compounds were analyzed in a dose-dependent manner. The thermal stability, opsin binding, regeneration of isorhodopsin, thermal stability at 55°C, and G t activation assays for each experiment were repeated three times. The parameters derived from these experiments were averaged, and standard deviations (S.D.) were calculated. The effect of each compound was either plotted in a dose-dependent or time-dependent manner as compared with controls. Statistical analyses were performed with GraphPad Prism 7 software. For two group comparisons, the one-way ANOVA test with Dunnett's multiple comparisons post-test was employed. For the means of two variables of more than two groups, the two-way ANOVA with Tukey's or Sidak's multiple comparisons post-test was used. Differences were considered statistically significant at a p value of Ͻ0.05 (*, p Ͻ 0.05; **, p Ͻ 0.001; ***, p Ͻ 0.0001).