Critical Role of Electrostatic Interactions of Amino Acids at the Cytoplasmic Region of Helices 3 and 6 in Rhodopsin Conformational Properties and Activation*

The cytoplasmic sides of transmembrane helices 3 and 6 of G-protein-coupled receptors are connected by a network of ionic interactions that play an important role in maintaining its inactive conformation. To investigate the role of such a network in rhodopsin structure and function, we have constructed single mutants at position 134 in helix 3 and at positions 247 and 251 in helix 6, as well as combinations of these to obtain double mutants involving the two helices. These mutants have been expressed in COS-1 cells, immunopurified using the rho-1D4 antibody, and studied by UV-visible spectrophotometry. Most of the single mutations did not affect chromophore formation, but double mutants, especially those involving the T251K mutant, resulted in low yield of protein and impaired 11-cis-retinal binding. Single mutants E134Q, E247Q, and E247A showed the ability to activate transducin in the dark, and E134Q and E247A enhanced activation upon illumination, with regard to wild-type rhodopsin. Mutations E247A and T251A (in E134Q/E247A and E134Q/T251A double mutants) resulted in enhanced activation compared with the single E134Q mutant in the dark. A role for Thr251 in this network is proposed for the first time in rhodopsin. As a result of these mutations, alterations in the hydrogen bond interactions between the amino acid side chains at the cytoplasmic region of transmembrane helices 3 and 6 have been observed using molecular dynamics simulations. Our combined experimental and modeling results provide new insights into the details of the structural determinants of the conformational change ensuing photoactivation of rhodopsin.

The cytoplasmic sides of transmembrane helices 3 and 6 of G-protein-coupled receptors are connected by a network of ionic interactions that play an important role in maintaining its inactive conformation. To investigate the role of such a network in rhodopsin structure and function, we have constructed single mutants at position 134 in helix 3 and at positions 247 and 251 in helix 6, as well as combinations of these to obtain double mutants involving the two helices. These mutants have been expressed in COS-1 cells, immunopurified using the rho-1D4 antibody, and studied by UV-visible spectrophotometry. Most of the single mutations did not affect chromophore formation, but double mutants, especially those involving the T251K mutant, resulted in low yield of protein and impaired 11-cisretinal binding. Single mutants E134Q, E247Q, and E247A showed the ability to activate transducin in the dark, and E134Q and E247A enhanced activation upon illumination, with regard to wild-type rhodopsin. Mutations E247A and T251A (in E134Q/E247A and E134Q/T251A double mutants) resulted in enhanced activation compared with the single E134Q mutant in the dark. A role for Thr 251 in this network is proposed for the first time in rhodopsin. As a result of these mutations, alterations in the hydrogen bond interactions between the amino acid side chains at the cytoplasmic region of transmembrane helices 3 and 6 have been observed using molecular dynamics simulations. Our combined experimental and modeling results provide new insights into the details of the structural determinants of the conformational change ensuing photoactivation of rhodopsin.
Rhodopsin is the leading model for the G-protein-coupled receptor (GPCR) 4 superfamily, which includes over 1000 membrane proteins (1). The three-dimensional structure of rhodopsin, characterized by seven transmembrane helices and a conserved disulfide bridge at the extracellular region, has been determined by x-ray crystallography (2)(3)(4). Upon rhodopsin activation by light absorption, the native chromophore 11-cisretinal, bound through a protonated Schiff base linkage to Lys 296 , is isomerized to its all-trans configuration. This photochemical reaction causes a conformational change in the protein that leads to the formation, through a series of short lived photointermediates, of the active state metarhodopsin II. Structures of some of these intermediates have been recently determined (5)(6)(7)(8). Most of these changes are concentrated on the transmembrane domain of the protein involving amino acids in the binding pocket of the retinal chromophore, which are efficiently transmitted to the cytoplasmic domain. The structural connection among several regions of rhodopsin (in the transmembrane and cytoplasmic domains) allows the interaction to its G-protein, transducin (Gt), which is activated, and initiates the visual transduction cascade (9,10). The structural details underlying the signal transmission process, going from retinal isomerization to G-protein activation, have not been determined and still remain for the most part a main scientific question to be answered.
Based on the proposed existence of a common activation mechanism, which has been correlated with the presence of a set of conserved residues among the GPCR superfamily, nomenclature has been established for an easy comparison among their amino acid sequences (11). The highly conserved triplet Glu 134 -Arg 135 -Tyr 136 (corresponding to 3.49 -3.50 -3.51 according to this general numbering code), in the C-terminal part of helix 3, plays a critical role in the rhodopsin photoactivation process (12,13). In particular, the Glu 134 -Arg 135 ionic couple has been shown to be very important for the activation of Gt (14), and any changes in this interaction can alter both the activation of the G-protein and other structural features like the glycosylation of rhodopsin (15)(16)(17). The crystal structures of rhodopsin reveal that several amino acids near the triplet are part of a network of electrostatic interactions between helices 3 and 6 (2,3,18,19). Specifically, there are mainly four residues that are involved in this electrostatic lock in rhodopsin as follows: Glu 134 (3.49), Arg 135 (3.50), Glu 247 (6.30), and Thr 251 (6.34), as shown in Fig. 1. The importance of positions 3.49 and 3.50 has been reported previously, and conformational changes causing constitutive activity of some GPCR have been described upon mutation of these residues. Examples of these receptors are rhodopsin (20 -22), ␣ 1 -adrenergic receptor (23), histamine-H 2 receptor (24), ␤ 2 -adrenergic receptor (25), and the thyrotropin receptor (26). Similarly, these positions have been shown to be involved in the expression, activation, and internalization of the gonadotropin-releasing hormone receptor (27) and the stabilization and expression of the -opioid receptor (28). Specifically protonation of the carboxylic side chain of the amino acid at position 3.49 has been suggested to release the ionic lock that constrains the receptor in its inactive conformation, by means of a network of electrostatic interactions, and to facilitate the movements of helices 3 and 6 necessary for the receptor to reach its active state (15,23,29).
On the other side, positions 6.30 and 6.34 have been studied in other receptors, such as ␤ 2 -adrenergic (30), thyrotropin receptor (31), ␣ 1b -adrenergic (32), human lutropin/choriogonadotropin receptor (33), -opioid receptor (34), and 5-hydroxytryptamine 2A serotonin receptor (35). In rhodopsin, single point mutations at Glu 134 and Glu 247 have been studied because of the importance of the three cytoplasmic loops in Gt activation (14,36), but no combination of mutations at the sites of helices 3 and 6 has been reported so far to address the role of the electrostatic network involving these two helices in the conformational and functional properties of the receptor.
In this work, the study of this electrostatic network, involving positions Glu 134 , Arg 135 , Glu 247 , and Thr 251 , has been undertaken. Specifically, single and double mutants involving Glu 134 in helix 3 and Glu 247 and Thr 251 in helix 6 have been constructed, purified, and characterized spectroscopically, functionally, and through molecular modeling. The results presented here provide experimental evidence for an important role of these amino acids in the functional status of the visual photoreceptor rhodopsin. Some of the double mutants show clear activity in dark conditions, particularly E134Q/T251A, and point to a key role of Thr 251 in the electrostatic network. The results provide experimental evidence, for the first time, for the involvement of Thr 251 in the conformational changes accompanying rhodopsin activation and highlight the importance of this residue in the structure and function of rhodopsin. The results are discussed in terms of the dynamic behavior of the different models of a selected subset of mutants as compared with the behavior of wild-type (WT) rhodopsin through molecular dynamics (MD) simulations.
The complementary study of the experimental and modeling features of the mutants provides new insights into the nature of the electrostatic network that keeps rhodopsin in its inactive conformation, allowing the dissection of the individual role of Glu 134 , Glu 247 , and Thr 251 in the cytoplasmic domain of rhodopsin.

EXPERIMENTAL PROCEDURES
Materials-All buffers and chemicals were purchased from Panreac and Sigma. Oligonucleotides were obtained from Operon (Qiagen). Enzymes for site-directed mutagenesis by PCR were purchased from Stratagene and rho-1D4 antibody from the National Cell Culture Facilities. Cyanogen bromide-4B-Sepharose and phenylmethylsulfonyl fluoride were obtained from Sigma, and all restriction endonucleases were from Amersham Biosciences and New England Biolabs. Dodecyl maltoside (DM) was purchased from Anatrace. 11-cis-Retinal was synthesized at Moscow State University. The nonapeptide corresponding to the last nine residues of the C-terminal region of rhodopsin (9-mer) was synthesized in the Laboratori de Síntesi de Pèptids (Universitat de Barcelona).
Buffers are defined as buffer A (1.8 mM KH 2 PO 4 , 10 mM Na 2 HPO 4 , 137 mM NaCl, 2.7 mM KCl, pH 7.2), buffer B (buffer A ϩ 1%DM ϩ 100 M phenylmethylsulfonyl fluoride), buffer C (buffer A ϩ 0.05%DM), buffer D (2 mM NaH 2 PO 4 ϩ 0.05%DM, FIGURE 1. a, lateral view of bovine rhodopsin in a DPPC membrane; boxes show the retinal pocket and the region where the network of electrostatic residues between helices 3 and 6 are located; lipids are shown in gray, the protein in blue, and retinal and Lys 296 in yellow; water molecules and Na ϩ and Cl Ϫ ions have been omitted for clarity. b, detail of the network of electrostatic residues between helices 3 and 6 extracted from a representative snapshot of the WT simulation. Arg 135 interacts with Glu 134 , Glu 247 , and Thr 251 . pH 6.0), and buffer D ϩ the nonapeptide (buffer D ϩ 100 M 9-mer) for protein elution from the chromatographic column.
Construction, Expression, and Purification of Rhodopsin Mutants-For single mutations, site-directed mutagenesis was performed using cassette-based strategy replacing the PstI/ MluI fragment of pMT4 (36) for an oligonucleotide containing the single mutation E134Q, E247A, and E247Q. PCR was used for obtaining single mutants at position Thr 251 and double mutants. Plasmid DNA was analyzed by restriction analysis (using EcoRI and MluI) and followed by DNA sequencing using the dideoxy chain-terminated method (constructed rhodopsin mutants are shown in Table 1). WT and mutant opsin genes were expressed in transiently transfected COS-1 cells as described (37). Transfected cells were harvested between 60 and 72 h after the addition of DNA and washed twice with buffer A. The cells were then incubated with 20 M 11-cisretinal for 5 h for reconstitution, solubilized in buffer B for 1 h, and finally centrifuged at 35,000 rpm for 35 min. All these procedures were carried out at 4°C. Rhodopsin was purified from the supernatant using rho-1D4-Sepharose and incubated overnight at 4°C. Sepharose was subsequently washed twice with buffer C and twice with buffer D and finally was incubated with buffer D ϩ the nonapeptide for 30 min at 4°C for rhodopsin elution. Rhodopsin concentration was determined by using a molar extinction coefficient value ⑀ 500 of 40,600 M Ϫ1 cm Ϫ1 .
UV-visible (UV-Vis) Absorption Spectroscopic Assays-All measurements were made on a Cary 1E spectrophotometer (Varian, Australia), equipped with a water-jacketed cuvette holder connected to a circulating water bath. Temperature was controlled by a Peltier accessory connected to the spectrophotometer. All spectra were recorded, in the 250 -650 nm range, with a bandwidth of 2 mm, a response time of 0.5 s, and a scan speed of 180 nm/min. For the photobleaching and acidification of rhodopsin, samples were bleached by using a 150-watt power source with a 495 nm cutoff filter. Dark-adapted rhodopsin samples were illuminated for 20 s to ensure complete photoconversion to 380-nm absorbing species. Acidification was carried out by addition of 2 N H 2 SO 4 (1% of the sample volume), immediately after photobleaching, and the absorption spectrum was recorded 2 min later.
Rhodopsin thermal stability in the dark was followed by monitoring the loss of A max in the visible region as a function of time, at constant temperature (45°C). Complete spectra were recorded every 5 min. Spectra were normalized and fitted to single exponential functions using SigmaPlot version 8.02 to derive the t1 ⁄ 2 values. Hydroxylamine treatment was performed by adding hydroxylamine (pH 7.0) to the samples at a final concentration of 30 mM at 20°C.
Gt Activation Assay-Gt activation levels were determined by incorporation of radioactive GTP␥ 35 S in Gt molecules induced by WT rhodopsin or its mutants essentially as described previously (38). Briefly, the reaction mixture, containing 1 M Gt, 20 nM rhodopsin, 10 mM Tris-HCl (pH 7.4), 100 mM NaCl, 5 mM MgCl 2 , 2 mM dithiothreitol, 0.012% DM, and 3 M GTP␥ 35 S, was bleached for 30 s by using a 150-watt power source with a 495 nm cutoff filter. Samples were incubated at room temperature for 1 h, and the reactions were stopped by addition of 10 mM Tris-HCl (pH 7.4), 100 mM NaCl, and 10 mM EDTA. The dark activity of the proteins was determined in exactly the same way but without illumination. The unbound GTP␥ 35 S was removed by microfiltration. The amount of GTP␥ 35 S bound to Gt was determined by Cerenkov counting using a TRI-CARB 2100TR scintillation counter from Packard Instrument Co.
MD Simulations-Atomic coordinates of rhodopsin were retrieved from the Protein Data Bank (entry 1GZM) (18). In contrast to other previously published crystal structures (2)(3)(4), this structure contains the coordinates of all the amino acids of the intradiscal and cytosolic loops of rhodopsin. The N-terminal acetyl group and the two palmitoyl chains at the cytoplasmic end of the receptor that are covalently linked to two consecutive cysteine residues were also included in the present model. In contrast, the C-terminal region after the palmitoylated cysteines, at the end of the short helix 8, was not included. Side chains of all the amino acids were considered in the protonation state they exhibit as free amino acids in water at pH 7, with the exception of Asp 83 and Glu 122 that were treated as protonated and neutral, respectively, according to experimental evidence (39).
The process of protein embedding was performed as reported elsewhere (40). Specifically, the protein was placed in a box containing a mixture of DPPC lipids and water molecules generated and equilibrated according to the procedure described previously (41). The simulation box had an initial size of 10.3 ϫ 8.0 ϫ 10.2 nm 3 (xyz), and organized in such a way that the first two dimensions corresponded to the bilayer plane. The area per lipid was 0.64 nm 2 . Before protein insertion, the box contained 256 lipids and ϳ17,000 water molecules. After the protein was inserted in the center of the box, all water molecules with oxygen atoms closer to 0.40 nm from a non-hydrogen atom of the protein and all lipid molecules with at least one atom closer to 0.25 nm from a non-hydrogen atom of the protein were removed, resulting in a final system that contained 198 lipids and ϳ16,000 water molecules. Removal of these atoms introduces small voids between the protein and water or lipid molecules that are easily removed during the first step of the equilibration process (see below), in which a progressive adjustment of the lipid bilayer and water molecules to the protein takes place. Next, in the system with WT rhodopsin, 113 water molecules were selected randomly and replaced by 57 sodium and 56 chloride ions, providing a neutral system with a concentration of about 0.2 M of both sodium and chloride ions (fairly similar to that found in biological organisms). For the simulations of mutants, single water molecules were converted to ions when necessary, to keep the electroneutrality of the system.
All the computer simulations reported in this study were performed using a parallel version of the GROMACS 3.2 package (42,43). The system was subjected to periodic boundary conditions in the three coordinate directions. Temperature was kept constant at 323 K (well above the gel/liquid crystalline phase transition temperature of 314 K) using separate thermostats for the protein, water, ions, and lipid molecules (44). The time constant for the thermostats was set to 0.1 ps except for water, for which a smaller value of 0.01 ps was used. The pressure in the three coordinate directions was kept at 0.1 MPa using independent Berendsen barostats with a time constant of 1.0 ps. The equations of motion were integrated using the leapfrog algorithm with a time step of 2 fs. All bonds within the protein and lipid molecules were kept frozen using the LINCS algorithm (45). The bonds and the angle of the water molecules were fixed using the analytical SETTLE method. Lennard-Jones interactions were computed with a cutoff of 1.0 nm, and the electrostatic interactions were treated using a twin cutoff of 1.0/1.8 nm, being the interactions in the interval 1-1.8 nm updated every 10 time steps.
The all-atom optimized potential for liquid simulations force field (46) currently implemented in GROMACS was used for all molecules of the system, except for the DPPC molecules, which were modeled using force field parameters previously described in the literature (47). Nonbonded pair interactions were computed as combinations of single atomic parameters to account for protein-lipid interactions. These parameters reproduce the experimental area per lipid of pure DPPC in the liquid crystalline phase (47)(48)(49). Water molecules were modeled using the TIP3P model (50). Fractional charges of retinal atoms were taken from quantum chemical calculations (51) and have already been used in other MD simulations of rhodopsin (52).
The structure of WT rhodopsin was energy-minimized and equilibrated in successive MD simulations at 6.5 ns long. First, 500 ps were run using the NPT ensemble, allowing the system to change the box dimensions, with protein coordinates strongly restrained to the initial structure. During this step, voids between the protein and lipid and water molecules vanished. At that point the simulation was continued for 500 ps in the NVT ensemble with the same positional restraints on the protein. In the last equilibration step, when the bilayer and water around the protein were equilibrated, restraints on the protein were released, and the system was equilibrated for 5.5 ns. The final structure obtained was used, first to continue the compute MD trajectory of the WT up to 10 ns, and second, to construct the different mutants by side-chain replacement and to subsequently perform a 5-ns MD trajectory on each of them.

Expression, Purification, and Spectral Characterization of
Rhodopsin Mutants-The rhodopsin mutants obtained at the cytoplasmic boundaries of helices 3 and 6 ( Table 1) were expressed in COS-1 cells, immunopurified with the monoclonal rho-1D4 antibody, and spectrophotometrically characterized by means of UV-Vis spectroscopy (Fig. 2). The A 280 / A 500 spectral ratios, which reflect chromophore regeneration and protein yield after purification, are shown in Table 2. All mutants studied showed no hydroxylamine sensitivity in the dark, and their photobleaching and acidification properties were very similar to those of WT rhodopsin (data not shown).
Mutations E134Q, E247A, E247Q, E134Q/E247A, and E134Q/E247Q-These mutants were expressed to an extent similar to that of WT rhodopsin according to their absorbance spectra (Fig. 2, A and C). All mutants were regenerated with 11-cis-retinal to form a chromophore with spectral features similar to those of WT ( max at 500 nm). E247A and E247Q showed a chromophore formation similar to WT rhodopsin, whereas E134Q and E134Q/E247A showed some differences judged from the A 280 /A 500 ratios ( Table 2). On the other hand, the double mutant E134Q/E247Q displayed impaired 11-cisretinal binding ( Table 2). This reduction may be interpreted as lack of accessibility of the retinal to the binding pocket as a result of the mutations, to some slight degree of structural misfolding affecting the proper conformation of the ligand-accepting receptor or to a lower stability of the mutant in detergent   WT rhodopsin and mutants were transfected, regenerated with 11-cis-retinal, and purified in Buffer C ϩ 0.05% DM as described (see "Experimental Procedures"). All spectra were recorded at 20°C. solution. In any case, this reduction does not reflect gross conformational perturbations and it may be assumed that the overall conformation of the mutants is similar to WT rhodopsin, as observed for the same maximum absorbance at ϭ 500 nm (Fig. 2). The positions where the mutations have been introduced (3.49 and 6.30 according to a generic numbering system for GPCR (11)) were also studied in the ␤ 2 -adrenergic receptor (30) and proposed to play a critical role in the conformational rearrangement involving helix 3 and helix 6 that is required for rhodopsin activation (12). The results obtained with the ␤ 2 -adrenergic receptor suggested that ionic interactions between 3.49, 3.50, and 6.30 may constitute a common switch governing activation of rhodopsin-like receptors. Furthermore, the features of the mutants, with regard to charge neutralization and changes in electrostatic interactions, showed a good correlation with the extent of constitutive activity observed for the mutants (30). In this study, some degree of dark activity was detected for some of the mutants studied (see "Gt Activation of Rhodopsin Mutants").

Rhodopsin
The thermal stability of these mutants, in the dark, was measured at 45°C (Table 3). The general observed trend is that the mutants are less stable than WT. Lack of hydroxylamine reactivity indicated that the mutants present a compact structure around the Schiff base linkage, as would be expected because of the far away location of the mutations referred to the Schiff base environment (data not shown).
Mutations T251A, T251K, E134Q/T251A, and E134Q/ T251K-We explored for the first time the functional role of the polar amino acid Thr 251 (in helix 6) in the electrostatic interaction with Glu 134 and Glu 247 , in native rhodopsin structure and function, by introducing mutations at this site. The single mutants T251A and T251K were analyzed to determine the involvement of Thr 251 (6.34) in the rhodopsin activation mechanism. Protein yield after purification from COS-1 was about half that of the WT under the same conditions (Fig. 2B). This suggests that the mutant proteins are expressed in the COS-1 cell line to a lower extent than the WT. The mutant rhodopsins show spectroscopic features of the WT concerning the max which is located at 500 nm, but their A 280 /A 500 ratios are slightly higher than those of the WT (Table 2).
Double mutants E134Q/T251A, and E134Q/T251K were constructed and spectroscopically analyzed. These mutants were purified in lower amounts than WT (Fig. 2C). The A 280 / A 500 ratios are different for both mutants as follows: in the case of E134Q/T251A, the ratio is similar to that of WT, whereas E134Q/T251K has a slightly higher ratio when compared with WT rhodopsin (Table 2).
Mutations E247A/T251A, E247A/T251K, E247Q/T251A, and E247Q/T251K-Double mutants replacing amino acids in the same helix 6 (at 247 and 251 positions, which are about one helical turn apart in this helix), E247A/T251A and E247A/ T251K, were obtained and spectrally characterized (Fig. 2D). Both mutants were obtained at lower yields and showed impaired 11-cis-retinal binding with regard to WT rhodopsin. Similar features were also observed for mutants E247Q/T251A and E247Q/T251K (Fig. 2D). Specifically, the mutation E247Q/ T251K produces the most destabilizing effect in the protein resulting in a higher A 280 /A 500 ratio, which is larger than that of mutant E247A/T251K (Table 2). This suggests that there is steric hindrance imposed by the bulkier Gln side chain at position 247, in contrast to the case of Ala substitution in the E247A/T251K, which appears to be better tolerated.
Gt Activation of Rhodopsin Mutants-The spectral characterization shows that most of the mutants studied can be expressed, purified, and regenerated with 11-cis-retinal to an extent not very different from WT rhodopsin. This suggests that the overall conformation of these mutants in the dark state is not dramatically altered. Therefore, it was of great interest to analyze the functional characteristics of these mutants in terms of their ability to activate Gt.
The Gt activation experiments were carried out with detergent-solubilized and purified WT and mutant rhodopsins by means of a classical radioactive assay. Samples were analyzed both under dark and light conditions in all cases, and the data reported correspond to the final activation (after 1 h) detected in the two conditions. The Gt activation data for all the mutants studied are summarized in Table 4.
Single Mutants E134Q, E247A, E247Q, T251A, and T251K- Fig. 3A shows the functional data for the single mutants as well as that of the WT for comparison (in all experiments WT rhodopsin was used as a control for comparison, thereby its light activity was taken as 100% for reference).
The Gt activation for WT rhodopsin in dark conditions, obtained under the current experimental conditions used, is about 10%, as also seen in previous reports (37). No activity was previously detected for purified WT rhodopsin in DM solution

TABLE 4 Gt activation of WT rhodopsin and the mutants determined under dark and light conditions
The Gt activation data were obtained as described under "Experimental Procedures." All values were normalized to light-activated rhodopsin taken as 100%. in dark conditions (53). On the other hand, the activity of mutants in the content of COS-1 cell membranes was found to be lower than when they are purified in detergent solution (54). This may be due to an increased conformational flexibility of the mutants in the detergent-solubilized state that facilitates physical interaction with Gt in the recognition event that takes places in the activation mechanism.

Rhodopsin Gt activation (mol GTP␥S/mol Rho)
In the case of the E134Q mutant, we found a small increase in dark activity as well as increased activity in light conditions (of ϳ30%) which was similar to that previously reported (54,55). For the Glu 247 mutations in helix 6, we found a different behavior for the E247A and E247Q. In the latter case, the light activation is not increased with regard to WT rhodopsin and in the dark appears to be only slightly above the E134Q level. In the case of E247A, the activity upon illumination is similar to that of E134Q, but in dark conditions it is clearly higher. It appears that mutations to Ala result in higher levels of dark activity, as will be discussed below. Mutations at Thr 251 result in a somehow lower activity with regard to WT under light conditions. Double Mutants Involving Amino Acids of Helices 3 and 6 (Interhelical)-In the case of the double mutants ( Fig. 3B and Table 4), a combined effect can also be observed. Addition of E134Q mutation to any of the single mutants in helix 6 (in the E134Q/E247A, E134Q/E247Q, E134Q/T251A, and E134Q/ T251K mutants) resulted in an increase of Gt activation in light conditions, except T251K which showed a clear decrease to about 40% of the WT activity. The E134Q/T251A double mutant shows a clear and definite synergistic effect when compared with the corresponding single mutants both in the dark (50%) and in the light (230%). This is the first time that experimental evidence is provided for the decisive involvement of Thr 251 in rhodopsin helix 6 in the electrostatic network at the cytoplasmic ends of helices 3 and 6.
A synergistic effect was observed previously for mutations in the transmembrane domain like G90D/M257Y (56) or E113Q/ M257Y (57). In the case of the double mutant E134Q/M257A, this was proposed to have a 50% activity in the dark. We observe a similar effect for the E134Q/T251A double mutant. However, in our case this is the first time where mutations at the cytoplasmic boundaries show such dark activity. Mutations at the homologous position of Thr 251 in other GPCR were shown to be constitutively active (31, 58 -60).
In the case of the samples under light conditions, the possibility that different metarhodopsin II decay rates are in part responsible for the observed differences in transducin activation by the different mutants cannot be excluded.
Molecular Modeling Studies-The analysis of the crystal structure of rhodopsin (2-4, 17, 18) reveals that residues Glu 134 , Glu 247 , and Thr 251 are part of a network of hydrogen bonds located at the cytoplasmic sides of helices 3 and 6 as shown in Fig. 1. The use of MD simulations allows a better characterization of these interactions, providing insight into the structural effects caused by the different mutations on WT rhodopsin and their effect on the activation process of the receptor.
The amino acid sequence corresponding to the cytoplasmic side of helices 3 and 6 is particularly abundant in charged residues (Table 5). Thus, it is expected that the structure is sensitive to alterations on the charge of the side chains (Table 6). Hydrogen bond interactions involving the residues on the cytoplasmic sides of helices 3 and 6, as well as on the second and third cytoplasmic loops, are listed in Table 7. A more detailed description of the molecular computations on the WT and mutants follows.
WT and E134Q, E247A, E247Q, and T251K Single Mutants-In native rhodopsin, the positively charged Arg 135 establishes strong interactions with residues Glu 134 , Glu 247 , and Thr 251 along the entire simulation either with one of the negatively charged residues of helix 3 (Glu 134 ) or with one negatively charged and one polar residue of helix 6 (Glu 247 and Thr 251 ). The involvement of Thr 251 is supported by the increased activity exhibited by the Thr 251 mutant reported in this study and the effect of an analogous residue in the ␤ 2 -adrenergic receptor (30). A set of hydrogen bonds among residues at the boundary between helix 6 and the third cytoplasmic loop (involving Glu 239 , Ser 240 , Thr 242 , Thr 243 , Gln 244 , and Lys 248 ) was observed. The removal of the negative charge at position 134 in mutant  Table 4.
E134Q has effects on the competition to form hydrogen bonding interactions between helices 3 and 6. Specifically, the E134Q mutation induces a remarkable preference of Arg 135 for residues located in helix 6. Thus, after a short simulation time, the interaction between residues Gln 134 and Arg 135 is lost, whereas Arg 135 -Glu 247 and Arg 135 -Thr 251 interactions remain stable. Therefore, the interaction between the cytoplasmic side of helices 3 and 6 is stronger than in WT rhodopsin. Glu 134 forms a hydrogen bond with His 152 on helix 4, with the drop of the Glu 134 -Arg 135 interaction, during half of the time of the trajectory analyzed. Additionally, a Glu 239 -Lys 245 interaction, at the boundary between the third cytoplasmic loop and helix 6, can be observed for a small amount of snapshots.
Removal of the negative charge at position 247, in the E247Q mutant, forces Arg 135 to exhibit a more favorable interaction with Glu 134 than with Gln 247 , just the opposite observed in the case of the E134Q mutant. Although the Arg 135 -Gln 247 interaction is maintained during the entire simulation, it is too weak to keep Arg 135 close to Thr 251 . Thus, the Thr 251 -Arg 135 interaction is lost, whereas the distance between Arg 135 and Glu 134 is clearly reduced after a short simulation time, inducing a displacement of Arg 135 side chain toward helix 3 moving away from helix 6. There are also two interactions as follows: Glu 239 -Ser 240 and Ser 240 -Thr 242 at the boundary between the third cytoplasmic loop and helix 6, which can be observed during most part of the trajectory. These interactions are already present in the WT, although to a lower extent.
The T251K mutation provides an additional positive charge at the end of helix 6, a region with a large number of charged residues as follows: three positively charged residues (Lys 245 , Lys 248 , and Arg 252 ) and two negatively charged ones (Glu 247 and Glu 249 ), as shown in Table 5. The presence of Lys 251 forces a rearrangement of the neighboring positively charged residues, driven by electrostatic repulsion. Consequently, the Arg 135 -Glu 247 interaction vanishes after 2.5 ns, with Glu 134 -Arg 135 the only remaining hydrogen bond interaction involving residue Arg 135 . Therefore, the effect of this mutation is similar to that of E247Q, favoring the Arg 135 -Glu 134 interaction. Lys 251 interacts mainly with Glu 239 and Gln 137 on the third cytoplasmic loop, but also with Thr 229 in helix 5. Moreover, Glu 247 exhibits a hydrogen bond interaction with Gln 312 at the short cytoplasmic helix 8 during a long time of the trajectory, not observed in any other simulation. As a consequence of these rearrangements induced by the new positively charged chain at position 251, a set of hydrogen bonds appears to be favored involving residues located either on helix 6 or on the third cytoplasmic loop (Glu 232 , Gln 236 , Gln 237 , Glu 239 , Ser 240 , and Lys 248 ). Nevertheless, the only hydrogen bond detected in WT rhodopsin is Glu 239 -Ser 240 , although for shorter times, as well as in the remainder of the mutants. These new interactions are a consequence of the positive charge density at the end of helix 6, suggesting that T251K can significantly alter the conformation of the cytoplasmic ends of helices 3 and 6 and the third cytoplasmic loop by modifying the hydrogen bond network between them. Because none of the mutants involving this single mutation (including the double ones) exhibited increased activation, neither in the dark nor in the light conditions, it may be regarded as a deleterious one in terms of protein function, probably by constraining the conformation of the receptor in a functionally defective state.
Double Mutants E134Q/E247A, E134Q/T251A, E134Q/ T251K, and E247Q/T251K-The E134Q/E247A double mutation adds two net positive charges with regard to the WT and, more important, removes the two counter-charges of Arg 135 (Glu 134 and Glu 247 ). As shown above, the hydrogen bond interactions with Arg 135 are kept in E247Q and lost in the E134Q single mutant, respectively. Similarly, in the double mutant the interaction Arg 135 -Gln 134 is too weak to be kept without the

Sequence of bovine rhodopsin at the region of the cytoplasmic sides of helices 3 and 6
These parts of the sequence contain a large number of charged residues. The signs (*) and ( ⅐ ) indicate the charged and the polar residues, respectively. The lengths of the secondary motifs have been taken from Ref. 18. H3 and H6 account for helices 3 and 6, respectively; C2 and C3 for the second and the third cytoplasmic loops; Cyt and Int indicate whether the part of the sequence is located on the cytoplasmic or the intradiscal side, respectively.

TABLE 6
The net charge at helices 3 and 6 and their average relative distances in the studied mutants compared to WT The separations are given in nm and related to the value observed in WT.  Gray scale indicates percentage of trajectory present: 0 -20% (white), 20 -40% (light gray), 40 -60% (medium gray), 60 -80% (dark gray), and 80 -100% (black). The limits of helices and loops are taken from Ref. 18. The analysis has been performed on the last 2 ns of each trajectory using a cutoff of 0.45 nm. MAY 11, 2007 • VOLUME 282 • NUMBER 19 negative charge. The E247A mutation only allows the residue to interact with Arg 135 through its backbone carbonyl. As a consequence of this, both interactions vanish, destroying the ionic lock associated with the D(E)RY motif and keeping only the interaction Arg 135 -Thr 251 when compared with WT, although exhibiting shorter residence times. The removal of two negative charges forces hydrogen bond donors on the third loop to be closer to the cytoplasmic end of helix 6 to compensate for the lack of the negative charge, similarly to the effect observed when a positive charge is added in the case of T251K. Therefore, a set of interactions involving residues Glu 239 , Ser 240 , and Thr 242 already observed in the WT are reinforced, specifically those involving Glu 239 (Glu 239 -Thr 243 and Glu 239 -Ser 240 ), at the boundary between the third cytoplasmic loop and helix 6.

Cytoplasmic Regions of Helices 3 and 6 of Rhodopsin
In the case of the E134Q/T251A double mutant, there are two simultaneous effects. On the one hand, as observed for the E134Q single mutant, removal of the negative charge at position 134 favors the Arg 135 -Glu 247 interaction, and on the other hand the T251A mutation removes the hydrogen bond between Arg 135 and residue 251. The present analysis shows a decrease of the fraction of trajectory where Arg 135 and Glu 247 are interacting (about 25%). As a result of all these rearrangements, new hydrogen bonds between Lys 248 on helix 6 and both Thr 229 on helix 5 and Gln 237 on the third cytoplasmic loop are formed and kept during most of the trajectory analyzed. Additionally, interactions involving Ser 240 already present in WT are reinforced. The existence of these new hydrogen bonds and the reinforcement of others already existing in the WT may explain the high level of dark activity observed for this mutant (see Fig. 3 and Table 4).
The E134Q/T251K double mutant implies removal of the negative charge at position 134. This causes impaired interaction with Arg 135 , although this interaction is still observed in about half of the snapshots analyzed. This situation is different from the other mutants containing Gln 134 studied (E134Q and E134Q/T251A), where the interaction is completely lost. The reason for this specific behavior can be associated with the net ϩ2 charge of the mutant, which requires a large number of residues with negative charge density to be closer. In some snapshots, where Gln 134 does not interact with Arg 135 , it is involved in a hydrogen bond with His 152 . Obviously, the absence of Thr 251 results in the loss of the interaction with Arg 135 . Additionally, the presence of Lys 251 weakens the Glu 247 -Arg 135 interaction, because it competes with Arg 135 for Glu 247 , as in the other mutants containing T251K. Here, Lys 251 also interacts with Gln 244 . Although the interactions involving Arg 135 are significantly weakened or lost, when compared with the WT, a network of hydrogen bonds between residues at helix 6 and the third cytoplasmic loop, already present in the WT, is sensibly enhanced as follows: Glu 239 -Ser 240 , Glu 239 -Thr 243 , Ser 240 -Thr 242 , and Ser 240 -Thr 243 .
Finally, the E247Q/T251K double mutant involves mutations of two amino acids in the same helix. In this case, the side-chain replacements introduce a net charge of ϩ2 at helix 6, resulting in five positively charged residues within two helix turns with only a single negatively charged side chain (see Table  5). The suppression of the negative charge at position 247 and the introduction of an additional positively charged residue at position 251 breaks the Arg 135 -Gln 247 interaction, leaving Lys 251 -Gln 247 as the predominant interaction. Additionally, Gln 247 interacts with Glu 239 on the third cytoplasmic loop. The results are somehow similar to those described for the single mutant E247Q, but the changes in the interaction patterns are larger because of the existence of an additional positive charge in the region. Arg 135 and Lys 251 share the same counter-ion (Glu 134 ) with the Arg 135 -Glu 134 interaction being pre-dominant during the trajectory analyzed. The interactions involving Glu 239 , Ser 240 , Thr 242 , and Thr 243 at the boundary between helix 3 and the third loop become favored.
Distance between Helices 3 and 6- Table 6 shows the average relative distances between the cytoplasmic ends of helices 3 and 6 obtained from the simulations. Interestingly, the alterations on the hydrogen bond patterns described in the preceding section do not necessarily imply changes in the relative distances between helices. Specifically, the differences are smaller than 0.1 nm in all cases except for the T251K mutant. The results show that single mutants E134Q, E247Q, and T251K exhibit similar, shorter, or larger distances, respectively, compared with WT. Specifically, because residue 134 in helix 3 does not interact with helix 6, the removal of the negative charge with the E134Q mutation gives the smaller change. The results for E247Q clearly show that the loss of the electrostatic "lock" does not lead to a larger separation of the helices. On the contrary, here they even become slightly closer. T251K exhibits the largest separation between helices 3 and 6 because of the additional volume and/or charge at this position after the mutation. The different results observed for the double mutants, and specifically for those containing T251K, suggest that rhodopsin has mechanisms to rearrange the relative orientations of the helices. This can be done by rigid body motions of helices or helical fragments and lead to long range effects. 5 Overall, the analysis performed here suggests that despite being important for maintaining the receptor in its inactive state, other residues, in addition to Glu 134 , Glu 247 , and Thr 251 , are involved in maintaining the cytoplasmic sides of helices 3 and 6 together. This is in agreement with the observation that Arg 135 is not critical for receptor function, but it is important for stabilizing receptors in the inactive conformation (61).

DISCUSSION
Previous studies have suggested that the cytoplasmic regions of transmembrane helices 3 and 6 of the GPCR are engaged in a network of electrostatic interactions that stabilizes their inactive states. In rhodopsin this would involve Arg 135 in helix 3. Thus, mutation at Arg 135 , as in the R135Q mutant, could affect these interactions leading to a conformation that would have some features of the activated receptor. However, mutations at this site did not result in dark activity of the mutated rhodopsins (21,54,62). This suggested that the network of interactions was more complex than expected and stimulated our interest for other residues, such as Thr 251 in helix 6. Therefore, we have constructed and analyzed rhodopsins with mutations at other residues involved in this network (namely Glu 134 , Glu 247 , and Thr 251 ), but our discussion of the results takes into account the proposed role for Arg 135 in this specific set of interactions (although no mutant at this position was used in this study).
The detailed analysis of the structural and functional properties of mutants at the Glu 247 and Thr 251 positions in helix 6, and combining these with the previously studied E134Q mutant in helix 3, provides experimental evidence for the importance of the electrostatic network around the cytoplasmic boundaries of helices 3 and 6 of rhodopsin in the conformational and functional properties of the receptor. Our results support the existence of such a network and indicate, for the first time, that Thr 251 plays a key role in it. A clear synergistic effect is found for the E134Q/T251A double mutant that shows a significant level of dark activity (as well as a large increase of activity in light condition) in comparison with the corresponding individual single mutants. Previous studies have also shown synergistic effects on Gt activation in rhodopsin, but in that case the amino acids were more deeply buried into the transmembrane domain. This is also the first time where such an effect is reported for amino acids clearly at the cytoplasmic domain in rhodopsin. The dark activity of the E134Q/T251A double mutant suggests that the combination of a neutral nonpolar side chain at position 251 (like in T251A) and a charge at position 247 leads to a rhodopsin conformation that permits activation of Gt in dark conditions and stresses the critical role of amino acid interactions in this region (possibly including residues in the cytoplasmic loop as detected in the MD simulations).
The analysis of the MD simulations performed in this study provides insight into the dynamic nature of the residue interactions within the protein structure, providing a basis for the interpretation of the experimental results of the mutants presented. The analysis outlines important differences, as a result of the side chain replacements, in terms of hydrogen bond patterns between residues in the region around the cytoplasmic sides of helices 3 and 6 and in the third cytoplasmic loop. Specifically, every mutant studied exhibits a different hydrogen bond profile and distinct to that found in WT. Moreover, the profile exhibited by double mutants cannot be predicted from the effects observed for the corresponding single mutants, suggesting that the protein is extremely sensitive to subtle conformational rearrangements. The network of interactions found in WT rhodopsin around the ERY motif becomes weakened in all the mutants studied, and this seems to be partly compensated with new hydrogen bonds involving helix 6 and the third loop.
In summary, the results presented in this study provide experimental evidence that the interaction among Glu 134 , Arg 135 , Glu 247 , and Thr 251 , between helices 3 and 6, respectively, is crucial in keeping rhodopsin in the inactive state. Arg 135 -Glu 134 interaction is important so that Arg 135 adopts the necessary conformation that favors the previous interaction. A clear new finding is that the additional hydrogen bond between helices 3 and 6 with Arg 135 -Thr 251 contributes to the stabilization of the inactive state. Thus, removal of this interaction in the E134Q/T251A double mutant results in high levels of activity in dark and light conditions. Spectral and functional characterization and MD simulations taken together highlight the importance of the so-called "ionic lock" at the cytoplasmic ends of helices 3 and 6 in rhodopsin, and allow the proposal of a key role for Thr 251 in this electrostatic network for the first time. Our results suggest that the mutants can affect the rhodopsin activation process as follows: (i) by alleviating the network of interactions in the region around the ERY motif, favoring rhodopsin activation, and/or (ii) by altering the conformation of the third loop, disrupting the interaction with the G-protein.