Role of Bulk Water in Hydrolysis of the Rhodopsin Chromophore*

Rhodopsin (Rho) is a prototypical G protein-coupled receptor that changes from an inactive conformational state to a G protein-activating state as a consequence of its retinal chromophore isomerization, 11-cis-retinal → all-trans-retinal. The photoisomerized chromophore covalently linked to Lys296 by a Schiff base is subsequently hydrolyzed, but little is known about this reaction. Recent research indicates a significant role for tightly bound transmembrane water molecules in the Rho activation process. Atomic structures of Rho and hydroxyl radical footprinting reveal ordered waters within Rho transmembrane helices that are located close to highly conserved and functionally important receptor residues, forming a hydrogen bond network. Using 18O-labeled H2O, we now report that water from bulk solvent, but not tightly bound water, is involved in the hydrolytic release of chromophore upon Rho activation by light. Moreover, small molecules (and presumably, water) enter the Rho structure from the cytoplasmic side of the membrane. Thus, this work indicates two distinct origins of water vital for Rho function.

Rhodopsin (Rho), 3 the visual pigment in retinal rod photoreceptors, is a G protein-coupled receptor responsible for dim light vision (1,2). Rho is composed of its apoprotein, opsin, and a retinylidene chromophore in an 11-cis-conformation covalently linked by a Schiff base to Lys 296 of the opsin (Fig. 1A). Photoactivation of Rho is initiated by isomerization of the retinylidene ligand to its all-trans-configuration (3), allowing the photoreceptor after deprotonation of the Schiff base to adopt the active Meta II state required for G protein (transducin) activation. The atomic structure of Rho led to the identification of crystallographically ordered water molecules adjacent to functionally important conserved protein residues (Fig. 1A) (4), supporting the conclusion that these waters are essential not just for structural stabilization of the receptor but also for the activation process. One of these tightly bound waters posi-tioned close to the chromophore-binding pocket has been postulated to play a key role in the counterion switch between ground state (dark; Glu 113 ) and activated states (Glu 181 ) of Rho (5,6). Moreover, Glu 113 and Glu 181 are also likely involved in the hydrolytic process via the carbinol ammonium ion and in a regeneration reaction between 11-cis-retinal and opsin. To form the Schiff base, Lys 296 must be deprotonated, the carbonyl group must be polarized, and water must be accommodated within the chromophore-binding site (1). Hydroxyl radical footprinting revealed local conformational changes in the Rho structure following its photoactivation, presumably mediated by the dynamics of both ordered water molecules and the protein (7,8). Moreover, protein footprinting combined with rapid H 2 18 O mixing methodology and deuterium-hydrogen exchange on C 2 His residues indicated that these tightly bound internal waters do not exchange with bulk solvent in ground state Rho, Meta II, or opsin (7,8). Thus, ordered waters function as noncovalent cofactors that actively participate in transmitting the activation signal from the retinylidene-binding pocket to the cytoplasmic face of Rho, where binding of transducin occurs. These individual waters are observed in high resolution structures of Rho as well as other G protein-coupled receptors, and a high level of conservation of polar residues in close proximity to these waters is also found in all G proteincoupled receptor sequences (9,10).
Rho does not react with hydroxylamine (NH 2 OH) in the dark, but changes in the Rho structure upon photoactivation allowing this small nucleophile to penetrate inside the protein and accelerate chromophore hydrolysis ( Fig. 1B) (11). Molecular dynamic simulations and 1 H magnetic spinning NMR indicate a dramatic increase in interior Rho hydration after 11-cis-retinylidene isomerization (12). This large increase in intratransmembrane region hydration is characteristic of the Meta I state and appears to be directly related to the Schiff base counterion interaction that occurs immediately after breaking the Glu 113 -Schiff base salt bridge. Water influx is not accompanied by a significant change in the overall structure of this receptor, but most likely, it plays a functional role in the transition of dark state Rho to its Meta I intermediate (13).
As a consequence of photoactivation and relaxation processes in opsin, the chromophore-Schiff base linkage is hydrolyzed, releasing all-trans-retinal (at-RAL) (Fig. 1B), but it is unclear if ordered water or bulk water is responsible for retinylidene Schiff base hydrolysis. Although a role of extracellular water in this process had been suggested (14), we addressed this issue by using 18 O and 15 N labeling methodologies combined with modern mass spectrometry techniques. This approach allowed the experimental determination of which side of the photoreceptor NH 2 OH (and presumably, bulk water) enters the chromophore-binding pocket upon Rho photoactivation.

EXPERIMENTAL PROCEDURES
Materials-Fresh bovine eyes were obtained from a local slaughterhouse (Mahan's Packing, Bristolville, OH). WT and Lrat-deficient (Lrat Ϫ/Ϫ ) mice (15)   The water network extends from the region of visual chromophore binding (red) to the N and C termini. A magnified view of the chromophore-binding site reveals the presence of oriented water molecules near the retinylidene Schiff base. B, schematic representation of retinylidene Schiff base chemical reactivity in the ground and photoactivated states of Rho. As a consequence of photoreceptor activation by light, the retinylidene-Lys covalent bond is hydrolyzed. In this process, a water molecule donates its oxygen atom to form a new at-RAL molecule. The activated state of Rho allows a small nucleophile (NH 2 OH) or reducing agents such as NaBH 4 to penetrate inside the photoreceptor and chemically modify the Schiff base. Additionally, NaBH 3 CN was shown to be effective in reducing retinylidene-Lys bound in the ground state of Rho (38). twice with the same buffer, followed by centrifugation at 1000 ϫ g to remove contaminants. Bovine ROS membranes were prepared from frozen retinas under dim red light according to the procedure of Papermaster (16). Isolated ROS membranes were washed five times with hypotonic buffer (5 mM Bistris propane, pH 7.5) to dispose of membrane-associated proteins. Washed ROS were stored at Ϫ80°C or used immediately for Rho extraction. Rho was solubilized in n-nonylglucopyranoside and purified from ROS by ZnCl 2 /opsin precipitation (17). ZnCl 2 was removed by dialysis in the presence of 0.5% n-nonylglucopyranoside. Rho concentrations were determined with a Cary 50 UV-visible spectrophotometer (Varian, Palo Alto, CA) and quantified by absorption at 500 nm using the absorption coefficient ⑀ ϭ 40,600 M Ϫ1 cm Ϫ1 (18).
Preparation of Bovine Rho Proteoliposomes-Proteoliposomes were prepared by the procedure described by Niu et al. (19). A solution of 1% asolectin (commercially available soybean lipids, Sigma) in buffer composed of 10 mM MES, pH 5.8, 100 mM NaCl, and 1% n-nonylglucopyranoside was mixed with purified Rho. The molar ratio of phospholipids to Rho was 250:1. This solution was dialyzed against buffer composed of 10 mM MES, pH 5.8, and 100 mM NaCl in the dark at 4°C for 48 h with four changes of buffer totaling 8 liters. The proteoliposome suspension was pelleted by centrifugation at 100,000 ϫ g for 1 h at 4°C. To prepare vesicles with enclosed NH 2 OH, the above procedure was performed with buffers containing 50 mM NH 2 OH. Pelleted proteoliposomes were washed with 10 mM MES, pH 5.8, and 100 mM NaCl and centrifuged again at 100,000 ϫ g for 1 h at 4°C. Proteoliposomes were resuspended in the same buffer to achieve a final Rho concentration of ϳ2 mg/ml. The efficiency of Rho incorporation into lipid vesicles was determined by measuring the absorption spectra of the supernatant and pellet recorded after the first ultracentrifugation in a Cary 50 UV-visible spectrophotometer.
Orientation of Rho in Proteoliposomes-A proteolysis assay with Asp-N endoprotease, which specifically cleaves Rho between Gly 329 and Asp 330 in the C terminus (20), was used to determine the orientation of Rho in asolectin vesicles. A total of 20 l of proteoliposomes containing 30 g of Rho in buffer composed of 10 mM Bistris propane, pH 7.4, and 100 mM NaCl was subjected to proteolysis. The sample was incubated at room temperature for 20 h, and an aliquot containing 5 g of Rho was run on 10% SDS-polyacrylamide gel to resolve the cleaved fragments of Rho. The efficiency of digestion was determined by densitometric analysis with ImageJ software.
Chromophore Hydrolysis and Extraction-Bovine ROS membranes or mouse retinas were washed twice with 0.5 ml of buffer containing 10 mM Bistris propane, pH 7.0, and 100 mM NaCl prepared in either H 2 16 O or H 2 18 O and resuspended in the same buffer. The final concentration of Rho in ROS was 1.5 mg/ml. Each experimental tissue sample from mice contained the equivalent of two retinas. To initiate chromophore hydrolysis, 50 l of Rho in ROS membranes was illuminated with a 150-watt bulb for 5 min, during which time the samples were placed on ice to prevent overheating. Retinoids were then immediately extracted with 0.3 ml of hexane. The organic phase was separated by centrifugation at 16,000 ϫ g for 1 min, and retinoid composition was analyzed promptly by LC/MS (see below). In experiments with proteoliposomes, 30 g of Rho incorporated into liposomes and resuspended in 0.2 ml of 10 mM MES, pH 5.8, and 100 mM NaCl was illuminated with a 150-watt bulb for 5 min. Both NH 2 OH-loaded and "empty" proteoliposomes were exposed to light in the presence or absence of 10 mM NH 2 OH. Depending on the experiment, the sample extraction procedure involved addition of either 0.3 ml of hexane or 0.2 ml of methanol prior to the hexane, followed by vigorous shaking for 2 min and centrifugation at 16,000 ϫ g for 1 min. To control the time during which at-RAL reacted with NH 2 OH and to minimize derivatization upon extraction, samples were supplemented with 40 mM formaldehyde just prior to hexane addition. The efficiency of rhodopsin photoactivation in the above-described light conditions was ϳ98% as deduced from the 11-cis/all-trans-retinal oxime ratio after retinoid extraction in the presence of hydroxylamine and methanol.
LC/MS of Retinoids-Retinoid composition was analyzed with an Agilent 1100 series HPLC system attached to an LXQ mass spectrometer (Thermo Scientific). Retinoids extracted with hexane were injected onto a normal phase HPLC column (5-m ZORBAX SIL, 4.6 ϫ 250 mm, Agilent Technology) equilibrated with 10% ethyl acetate in hexane. Retinals were separated by isocratic elution with the equilibration solvent at a flow rate of 1.4 ml/min. The eluate was directed into the mass spectrometer through a diode array detector, followed by an atmospheric pressure chemical ionization source working in the positive mode. Data were recorded and analyzed with Xcalibur 2.0.7 software (Thermo Scientific). Elution times for at-RAL and its oximes were determined based on their characteristic UV-visible spectra with absorbance maxima at 368 and 357 nm, respectively. The ratio between at-RAL and at-[ 18

RESULTS
Carbonyl Oxygen Exchange in at-RAL-Aldehydes readily exchange their carbonyl oxygen with the oxygen atom of water (22). This replacement proceeds via a hydration-dehydration mechanism that depends on the electrophilicity of the carbonyl carbon and is accelerated by both acidic and basic conditions (23). This characteristic of aldehydes, which can lead to loss of the initial oxygen label, represents the main obstacle to interpreting experimental studies involving the mechanism of aldehyde formation (24). To counteract the potential effect of oxygen back-exchange, the rates of 18 (Fig. 2, A and B). This phenomenon can be explained by transient formation of Schiff base adducts of at-[ 18 O]RAL with available primary amino groups present in proteins or phospholipids. Reversibility of this reaction in the presence of water (hydrolysis) facilitates exchange of the 18 O label with bulk solvent. Therefore, samples containing protein and lipids should be employed as appropriate controls for oxygen back-exchange in studies of at-RAL formation upon Rho photoactivation in ROS.
Origin of the Oxygen Atom in Newly Formed at-RAL Released from Light-activated Rho-To investigate the potential role of internal structural water molecules in retinylidene Schiff base hydrolysis, we washed isolated bovine ROS with buffers composed of water isotopically labeled with 18 O. Because the tightly bound water of Rho does not exchange with bulk solvent in the ground state, unlabeled water molecules are preserved within the transmembrane helical bundle (7). Prepared experimental materials were exposed to light for 5 min and directly extracted with hexane. The amount of extracted at-RAL represented ϳ10% of the total at-RAL found in a methanol-treated sample (protein denaturing conditions) and corresponded closely to the amount predicted by Meta II decay kinetics (25). Photoactivation of Rho in the presence of H 2 18 O followed by rapid organic extraction revealed a pool of at-RAL that was highly enriched in the 18 O isotope (Fig. 2C). The isotopic composition of oxygen in at-RAL closely imitated that in the bulk water solvent and was much higher than the isotopic content of unlabeled at-RAL exposed to H 2 18 O predicted from oxygen exchange rates in control experiments ( Fig. 2A). Thus, the contribution of unlabeled internal protein-bound water molecules to retinylidene Schiff base hydrolysis appears to be insignificant. Because experimental conditions might influence the rate of oxygen back-exchange, we examined retinas isolated from Lrat Ϫ/Ϫ mice to approximate the environment present in a biological sample. Although the lipid and protein composition and the retinal morphology are preserved in these animals, the retinoid pool is dramatically reduced due to metabolic blockade of the uptake, storage, and production of visual chromophore (15). Thus, the isotopic composition of externally added at-RAL cannot be diluted by internal retinoids. Experimentally determined oxygen back-exchange in retinas isolated from Lrat Ϫ/Ϫ mice that were supplemented with synthetic at-RAL was comparable to previous observations made in the presence of soybean lipids and indicated an ϳ12% loss of label during the 5-min incubation (Fig. 2D). Consistent with the data obtained from bovine ROS, at-RAL liberated from WT mouse retinas resuspended in buffer containing H 2 18 O and exposed to light contained predominantly 18 O (Fig. 2D). The above observations indicate that external water molecules penetrate the Rho photoreceptor upon light activation and contribute extensively to retinylidene Schiff base hydrolysis.

Light Activation of Rho Allows Small Molecules to Reach the Chromophore-binding Pocket from the Cytoplasmic Side-
The N and C termini of Rho are oriented on opposite sides of the lipid bilayer and differ in terms of function (1). To investigate the potential asymmetrical role of these termini in the transduction of water molecules into the chromophore-binding pocket, we prepared proteoliposomes that contained preferentially oriented Rho with N termini located inside the liposomal lumen (Fig. 3A). The functional integrity of the preparation was examined spectrophotometrically by record-ing Meta II state formation in response to a light stimulus (Fig.  3B). The orientation of photoreceptors inserted into liposomes, as determined by proteolytic digestion with Asp-N protease, indicated that ϳ85% of the protein was appropriately and uniformly oriented. Moreover, at-RAL released from rhodopsin incorporated into proteoliposomes was enriched in 18 O to the same extent as was observed in the experiments on bovine ROS (data not shown). Notably, photoreceptor orientation in the liposomes was preserved during proteoliposome washes (Fig. 3,  C and D). Therefore, it was possible to deplete the external NH 2 OH concentration without changing the lumenal pool of solvent and generate proteoliposomes loaded with NH 2 OH (Fig. 3A). Under our experimental conditions, Rho exposed to light is rapidly converted into the Meta II state, which is relatively stable at acidic pH. Moreover, hydrolysis of the retinylidene Schiff base in Meta II can be accelerated by added nucleophiles such as NH 2 OH. Consequently, extraction of proteoliposomes under nondenaturing conditions in the absence of NH 2 OH revealed the presence of only small amounts of at-RAL. Interestingly, the presence of NH 2 OH inside the liposomes did not provoke massive at-RAL oxime formation under these conditions (Fig. 4A, panel a). The small amount of oximes seen in chromatograms represented Ͻ10% of the total extractable at-RAL oximes and may have been derived from the small fraction of Rho molecules with a reversed orientation. However, significant amounts of at-RAL oximes were detected after liberation of NH 2 OH trapped inside the vesicles by addition of either methanol or detergent (10 mM diheptanoylphosphatidylcholine) following light exposure (Fig. 4A, panel b, data for methanol treatment shown). In this case, the extent of at-RAL derivatization depended on the amount of NH 2 OH released from inside the vesicles. The striking difference between these two groups of samples indicates that NH 2 OH was indeed trapped within the liposomal lumen and did not diffuse across the lipid membrane, at least within the time frame of this experiment. In contrast, the presence of NH 2 OH added externally to proteoliposomes led to the rapid breakdown of the retinylidene Schiff base, resulting in quantitative formation of at-RAL oximes in both types of proteoliposomes (Fig. 4A, panel c). Thus, NH 2 OH molecules involved in Schiff base decomposition came from outside the vesicles. Considering the orientation of this protein in liposomes, we conclude that small molecules enter the chromophore-binding pocket of activated Rho from its C-terminal side.
For improved discrimination between internal and external pools of NH 2 OH, proteoliposomes that contained NH 2 OH inside the vesicles were exposed to light in the presence of 10 mM 15 N isotope-labeled reagent ( 15 NH 2 OH) added externally during the bleaching procedure (Fig. 5A). Similar to previous experiments and to avoid potential cross-contamination during extraction, the excess of unreacted NH 2 OH was depleted by addition of 40 mM formaldehyde after the incubation period but just prior to adding the organic solvent. In this case, the isotopic composition of at-RAL oximes examined by MS revealed a predominant peak at m/z 301.17 ([M ϩ H] ϩ ) corresponding to [ 15 N]RAL oxime (Fig. 5B). The contribution of the m/z 300.25 peak corresponding to unlabeled at-RAL oxime in the mass spectrum of extracted retinoid was slightly higher than that in FIGURE 3. Characterization of Rho proteoliposomes. A, schematic representation of Rho orientation within proteoliposomes. B, functional characterization of incorporated Rho. Spectrophotometric analyses of changes in absorption spectra observed after proteoliposomes were exposed to light indicate Meta II state formation. The difference spectrum was calculated by subtracting the spectrum recorded after and before light exposure. AU, absorbance units. C, determination of Rho molecular orientation by proteolytic digestion with Asp-N endoprotease, which specifically cleaves the C terminus of Rho. Lanes 1 and 3 represent washed intact proteoliposomes prepared in the presence or absence of NH 2 OH. Lanes 2 and 4 (protease-treated samples) reveal dominant protein bands corresponding to the Rho proteolytic fragment. D, densitometric quantification of protein bands shown in C indicate that at least 85% of Rho molecules preferentially adopt an orientation in which the N termini face the lumen of proteoliposomes.
the spectrum of the synthetic [ 15 N]RAL standard. However, it did not exceed 10% of the m/z 301.17 ion intensity. Again, this result could be a consequence of a small pool of Rho molecules with a reversed membrane orientation (Fig. 3). Together, these data indicate that only externally added 15 NH 2 OH contributed to retinylidene Schiff base cleavage.

DISCUSSION
Photoisomerization of the chromophore bound to Rho leads to activation of this G protein-coupled receptor. Consequently, the retinylidene Schiff base is hydrolyzed, followed by dissociation of at-RAL from the chromophore-binding pocket. We investigated this process in more detail and arrived at two significant conclusions: first, bulk solvent provides the source of water molecules utilized for chromophore hydrolysis following Rho activation, and second, a flux of small molecules (and presumably, water) enters into the retinal-binding pocket from the cytoplasmic side of this photoreceptor upon attainment of the activated state.
Increased intramembranous accessibility to external small molecules and water is required for Rho activation to advance beyond the Meta I state (26,27). Moreover, the retinylidene chromophore becomes accessible to reagents such as NH 2 OH or NaBH 4 only upon attainment of the Meta II state (11,28). Results obtained by NMR and hydrogen-deuterium exchange also suggest that the hydration state of Rho increases upon photoactivation (12,29). The process of retinylidene hydrolysis is complex, involving multiple protonation/deprotonation and proton transfer events. It most likely occurs in dark state Rho by employing the internal water network to mediate cleavage of the Schiff base linkage (30), but due to structural constraints provided by this state of the receptor, 11-cis-retinal release is hindered, and the Schiff base reforms. Subtle structural relaxation at the retinal-binding site rather than major conformational changes upon photoactivation can be envisioned as the FIGURE 4. Side of NH 2 OH entry into light-activated Rho. Two types of Rhocontaining proteoliposomes (prepared in either the absence or presence of NH 2 OH) were exposed to light while they were maintained in NH 2 OH-free buffers. Extracted retinoids were analyzed by HPLC. A, chromatograms show HPLC separation of at-RAL (peak 1) and its oxime (syn) (peak 2). Red traces correspond to proteoliposomes loaded with NH 2 OH, and blue traces represent data obtained from the empty vesicles. Panel a, extraction with hexane; panel b, disruption of proteoliposomes with methanol prior to extraction; panel c, incubation with NH 2 OH added after light exposure and 5 min prior to extraction. B, identification of detected retinoids. Peaks corresponding to at-RAL (left panel) and its oxime (syn) (right panel) were assigned based on the characteristic shapes and absorbance maxima of their UV-visible spectra. mAU, milliabsorbance units. FIGURE 5. Isotopic composition of at-RAL oximes. Proteoliposomes containing NH 2 OH were exposed to light in the presence of 15 NH 2 OH added to the surrounding buffer. After extraction, the isotopic composition of at-RAL oxime (syn) liberated from Rho was examined by MS. A, schematic representation of the experimental setup. B, isotope distribution of the at-RAL oxime ([M ϩ H] ϩ ) charged state. Upper and middle panels, unlabeled and 15 N isotope-containing synthetic standards, respectively; lower panel, isotopic composition of at-RAL oxime (syn) extracted from proteoliposomes exposed to light. Robust incorporation of the 15 N isotope into at-RAL oxime released from photoactivated Rho indicates that 15 NH 2 OH derived from the C-terminal side of Rho is solely involved in retinylidene Schiff base cleavage. driving force for this protein's transitions to new thermodynamically stable subconfigurations (31). The concomitant influx of water molecules upon photoactivation appears to remodel the network of internal water molecules relative to the ground state. This increase in hydration elevates the rate of deuterium exchange with the protonated Schiff base by several orders of magnitude (32). Such rapid proton exchange is enabled by an effective increase in water concentration near the retinal Schiff base. In D 2 O isotopic effect studies, slower rates of Schiff base hydrolysis were measured during Meta II decay, providing evidence that external D 2 O rather than internal water was involved in this process (30). Furthermore, exchange of the carbonyl oxygen of at-RAL released from Rho was visualized by FTIR spectroscopy (14). This evidence, along with our observations of 18 O incorporation into at-RAL, indicates a role for bulk solvent in the transition from Meta II to opsin and hydrolyzed dissociated at-RAL.
Interestingly, increased hydration per se does not promote rapid replacement of tightly bound intertransmembrane waters as judged by radiolytic footprinting observations (7). Because the hydrolysis reaction requires a protonated Schiff base for efficacy, it is tempting to speculate that ordered water activated by the Glu 113 side chain contributes to the protonation event, whereas the abundance of bulk solvent molecules ensures rapid and spontaneous hydrolysis of the retinylidene-Lys 296 linkage. Thus, transient reprotonation of the Schiff base should be the rate-limiting step in overall Meta II decay, and the hydrogen bond network surrounding the chromophore-binding site would play an essential role in this process.
Ordered water molecules lie in a specific channel connecting the N-and C-terminal sides of Rho, which most likely acts as a conduit for the activation signal through the membrane. Increased hydration of Rho upon light exposure poses the question as to from which side of the membrane this incoming bulk water originates. To investigate this problem, we prepared proteoliposomes containing Rho preferentially oriented with its N termini facing the liposomal lumen, filled with unlabeled NH 2 OH, and its C termini, located on the liposome exterior, exposed to solvent enriched in 15 N-labeled NH 2 OH. This experiment revealed that the NH 2 OH involved in Schiff base breakdown arose exclusively from the extravesicular space. Why would small polar molecules and water take such a long detour deep into the retinal-binding pocket if the chromophore is located closer to the lumen of the disk membrane and N terminus of Rho? The answer may relate to the fact that transmembrane helices III and IV, located on the extracellular side of Rho, are connected by an antiparallel ␤-sheet that forms a "plug," completely masking the retinal-binding pocket and retinal from extracellular bulk solvent (33). This plug also hinders chromophore release from dark state Rho. Mutations in this region have been shown to perturb photoreceptor function, leading to retinitis pigmentosa (34). Minor structural changes in the same location have been identified by FTIR spectroscopy after Rho photoactivation. Therefore, it appears that changes in the N-terminal face of Rho that occur after light activation are not large enough to allow massive water influx from the extracellular space. Marked changes do occur at the cytoplasmic surface of Rho, opening its structure to the entry of bulk water.
In fact, crystallographic structures of opsin and opsin complexed with the C-terminal peptide of G␣ t display formation of a solvent-accessible cavity at the cytoplasmic site (35). Moreover, none of these structures reveal any openings on the extracellular side of this photoreceptor. However, in opsin, transmembrane bundles transiently open into the hydrophobic membrane region through two holes located between helices I and VII and helices V and VI (35). A connecting channel between these openings might provide the means for retinal passage after Schiff base hydrolysis, confirming earlier predictions based on random acceleration molecular dynamics (36,37).