Conformational States and Dynamics of Rhodopsin in Micelles and Bilayers^†

Construction and Expression of the Mutant Rhodopsin Genes

The expression construct, pMT4-Rho, containing the base mutant ϕ (C140S/C316S) has been described previously (19). Four individual cysteine mutations (250C, 227C, 140C, and 316C) were introduced in the base mutant ϕ, one at a time, according to published procedures (38). Two methods for the transient transfection of the construct in mammalian COS1 cells were employed: the DEAE-dextran-mediated method (38) and the cationic liposome-mediated method (39). The liposome-mediated method was modified: LipofectAMINE (Invitrogen Life Technologies Co.) was substituted by 50 μL of Maxfect (MediaTech, Inc.) and 15 μg of plasmid DNA per plate was used.

Purification of the Rhodopsin Mutants and Derivatization with Spin Labels

Recombinant rhodopsin in intact cells was regenerated with 11-cis-retinal (National Eye Institute of the National Institutes of Health), solubilized in DM (Anatrace, Inc.), purified using a 1D4 monoclonal antibody (University of British Columbia) immobilized on Sepharose (38), and derivatized with the nitroxide methanethiosulfonate reagent (Figure 3) while bound to the 1D4 column (19). The protein was eluted from the column in the appropriate detergent [0.4 mM DM, 34 mM n-octyl β-d-glucopyranoside (Alexis Biochemicals), or 0.16 mM digitonin (Acros Organics)] containing 100 μM peptide corresponding to the last nine C-terminal amino acids of rhodopsin (American Peptide Co., Inc.). Fractions containing rhodopsin were collected and concentrated to 30-50 μM. The final detergent concentration was estimated at ∼12-15 mM DM and ∼5-7 mM digitonin.

Reconstitution of Rhodopsin in Lipids

POPC, POPE, and POPS (Avanti Polar Lipids, Inc.) in the desired molar ratio and buffer were extruded using a LipsoFast membrane extruder (Avestin, Inc.) with a membrane pore diameter of 200 nm. Asymmetric rhodopsin proteoliposomes, with the extracellular side facing the interior of the vesicle, were prepared from the extruded lipids and spin-labeled rhodopsin in n-octyl β-d-glucopyranoside as described (40-42). Detergent was removed by extensive dialysis, and proteoliposomes were separated from protein-free liposomes by sucrose density ultracentrifugation (43). The lipid to protein molar ratio was calculated from the protein and lipid concentrations determined by Bradford (Pierce Biotechnology) and phosphorus assays (44), respectively. Typical lipid to protein ratios were ∼300-400 lipids per rhodopsin. The lipid composition of the proteoliposomes was verified by TLC (45). The diameter of the proteoliposomes was determined to be ∼200 nm by particle light scattering (Particle Sizing Systems, Inc.). Reconstituted samples were used within 24 h of preparation, before significant aggregation of the proteoliposomes occurred.

Optical Spectroscopy

Aliquots of the rhodopsin samples used in EPR experiments were diluted in the appropriate buffer to a final A₅₀₀ of ∼0.2-0.5. UV-vis spectra were collected on a Cary 50 (Varian, Inc.) equipped with a temperature-controlled sample holder. Samples were equilibrated at 20 °C in the dark for 30 min before the spectrum of R was collected, followed by in situ illumination for 30 s (λ > 514 nm) and collection of the R* spectrum. The absence of free retinal after 20 min was verified by lowering the pH to 1.8; in all cases, the absorbance maximum shifted to 440 nm, characteristic of a protonated retinyl Schiff base (46). The light minus dark state difference spectrum was calculated to circumvent the scattering problems in the reconstituted samples. To determine the amount of MI and MII present, the difference spectra were fit, as a weighted sum of reference spectra of the pure components (R, MI, and MII), using a program written in LabVIEW (National Instruments). The results of the general k-dimensional linear least-squares fits were normalized so that the R component had an amplitude of -1 (complete bleaching). The sum of the remaining components was always 1 ± 0.08. The reference spectrum for R at 25 °C was obtained from WT (from rod outer segments) purified in DM, and that for MII was obtained upon illumination of R. The MI reference spectrum was obtained at 0 °C upon illumination of WT (from rod outer segments) purified in digitonin. An example of the analysis is shown in Figure 4 for the difference spectra of 140R1 in DM at pH 6 and 316R1 in PC:PE:PS bilayers at pH 6.

EPR Spectroscopy (A) Continuous Wave EPR Spectroscopy

EPR spectra of samples were measured in a Bruker ELEXSYS 580 fitted with a High Sensitivity resonator in small volume aqueous flat cells (Wilmad Glass Co. Inc.). Spectra were recorded at 100 G scan widths over 30 s with an incident microwave power of 20 mW and modulation amplitude of 4 G. This modulation optimized the signal to noise ratio and was verified to produce minimal spectral distortions. Samples were equilibrated in the dark at 20 or 35 °C for 30 min before the R spectrum was collected, followed by 30 s of illumination in situ (λ > 514 nm) and the collection of the spectrum for R*. Typically, nine scans were signal-averaged per spectrum, and the total collection time did not exceed 5 min. For rhodopsin in bilayers, the EPR spectra are time-dependent due to the decay of MI/MII to Op and/or MIII. On the basis of the decay rate in membranes at 20 °C (12), it can be estimated that the signal-averaged EPR spectrum obtained over a 5 min period represents that for bilayers containing no more than 10% of Op and/or MIII.

(B) Interpretation of EPR Spectra in Terms of Nitroxide Motion

As shown in Figure 2, all of the R1 sensors in the present studies are in helices. At surface sites, where the nitroxide ring does not have interactions with the surrounding protein, the internal motion of R1 due to torsional oscillations about the bonds of the side chain can generally be described in terms of an order parameter (S), measuring the amplitude or order of motion, and a correlation time (τ), the inverse of which measures the rate of motion (47). This internal motion is modulated by fluctuations of the backbone, giving rise to motions with S ≤ 0.5 and τ in the range of 1-3 ns. An example of a weakly ordered state of R1 with S = 0.2 and τ = 2 ns is shown by the spectrum in Figure 5A simulated according to the MOMD model of Freed and co-workers (48). For motions of this kind, changes in the central line width (ΔH₀) serve as a qualitative measure of changes in rate of motion, with increasing line width corresponding to a decrease in rate (47). Interactions of the nitroxide ring with the protein lead to various degrees of immobilization, due either to an increase in order, a decrease in τ, or both. An example of a relatively immobilized state with isotropic motion (S = 0) and τ = 50 ns is shown by the simulated spectrum in Figure 5B. Such a spectrum would arise, for example, for a nitroxide immobilized with respect to the protein and thus having a correlation time determined by protein rotary diffusion. For immobilized states with well-resolved hyperfine extrema, changes in the overall splitting between the extrema (2A_zz’, Figure 2B) provide a measure of changes in mobility; decreases or increases in 2A_zz’ reflect an increase or decrease in mobility, respectively (49).

Often, the EPR spectrum of R1 reflects multiple dynamic states of the side chain, arising either from two rotamers of the side chain, each having different degrees of interaction with the environment (50), or from two states of the protein. An example of such a complex spectrum is shown in Figure 5C, which is a weighted sum of the spectra in Figure 5A,B. Such spectra are typical of those for R1 in rhodopsin. In these cases, changes in 2A_zz’ still reflect changes in mobility of the more immobilized state. For complex spectra arising from states of widely different mobility, as encountered here and illustrated in Figure 5C, the more immobile component dominates spectral intensity in the region labeled α. Likewise, the more mobile component dominates spectral intensity in the region labeled β. Changes in intensity in these two regions will be taken to reflect corresponding changes in the population or dynamics of these two states.

Fitting of single- or multiple-component EPR spectra to the MOMD model of Freed and co-workers is a method of extracting quantitative information on orders and rates of motion for R1 in proteins (48, 49, 51). This model permits the simulation of spectra arising from an isotropic distribution of nitroxides undergoing anisotropic motion, which corresponds to the physical situation of a spin-labeled protein in which the rotary diffusion of the protein is slow compared to the internal motion of R1. In the context of the MOMD model, the motion of the nitroxide in R1 on solvent-exposed sites can often be described by a single order parameter, S, and an effective correlation time, τ_c (47, 52, 53).

For the important mutant 250R1, experimental EPR spectra were fit to a two-component MOMD model to extract information on the dynamics of the populations. Values for the A and g tensors used in simulations of strongly immobilized states were determined by fitting spectra of 250R1 obtained from dark-adapted rhodopsin in the absence of motion at -50 °C. The values are A_xx = 5.1, A_yy = 4.6, A_zz = 35.5, g_xx = 2.0082, g_yy = 2.0065, and g_zz = 2.0022. To constrain the number of parameters, the immobilized state is taken to have isotropic motion. It is further assumed that the more mobile state in two-component spectra is exposed to water, and A and g tensor values previously reported for R1 in an aqueous environment were employed, namely, A_xx = 6.2, A_yy = 5.9, A_zz = 37.0, g_xx = 2.0078, g_yy = 2.0058, and g_zz = 2.0022 (47, 51). The models presented in Results and the good fits to the data justify these assumptions. Fitting procedures followed those previously described (47, 53).

Modeling of the Nitroxide Side Chain in Rhodopsin

The R1 side chain was modeled at residues 140, 227, 250, or 316 in the 1GZM rhodopsin structure using DS Viewer Pro (Accelrys, Inc.), and all molecular graphics were generated using the PyMOL Molecular Graphic System (DeLano Scientific, LLC). The R1 rotamers about X₁ and X₂ (Figure 3) modeled were those commonly found in crystal structures of R1 at helical sites in T4 lysozyme (T4L), namely, {m, m}, {t, p}, and {t, m} (50; Fleissner, Cascio, Sawaya, Hideg, and Hubbell, unpublished results; Guo, Cascio, and Hubbell, unpublished results). The rotamer notation is that of Richardson et al. (54), where m = -60 ± 10, p = +60 ± 10, and t = 180 ± 10. Each structure was optimized by variation of the X₄ and X₅ dihedrals without moving native side chains. Allowed rotamers were determined by manual changes in dihedrals of R1, while avoiding overlap of the nitroxide side chain with other atoms.

Comparison of R and R* in DM and Phospholipid Bilayers

EPR spectra of 140R1, 227R1, 250R1, and 316R1 are presented in Figures 6-9 for R and R* at pH 6 in solutions of DM and reconstituted in POPC:POPE:POPS bilayers having a composition approximating that of the native disk membrane (5:4:1 mol ratio) (55). In each case, data were also collected for the spin-labeled proteins reconstituted in pure POPC. Within experimental error, the corresponding EPR spectra were identical to those shown for POPC:POPE:POPS mixed bilayers.

In DM at 20 °C, R* for each spin-labeled mutant contained 100% MII immediately after photoactivation, as determined from the 380 nm absorbance, consistent with published data on WT rhodopsin (19, 35). All EPR spectra of R* in DM (12-15 mM) were time-independent for at least 2 h.

In phospholipid bilayers both MI and MII are present. The relative amounts were determined within a few minutes after complete photolysis of rhodopsin as described in Methods, and the values are provided below in terms of the apparent (pH-dependent) equilibrium constant K = MII/MI. The MI and MII intermediates (monitored optically) decay slowly to opsin or a mixture of opsin and MIII (12), and the EPR spectra in the light-activated state are also time-dependent. The EPR spectra reported here were acquired within a short interval following activation and correspond primarily to the initial mixture of MI and MII (see Methods). The time-dependent changes in EPR spectra accompanying the decay of MI and MII will be reported elsewhere. The following sections consider the results for each spin-labeled mutant.

(A) 140R1

Residue 140 is located near the contact face of TM3 and TM5 in the headgroup region of the lipid bilayer (Figure 2). Figure 6A[1] shows a comparison of the EPR spectra for 140R1 R at 20 °C in DM and PC:PE:PS bilayers. In DM, the spectrum has two dominant components, reflecting relatively immobilized (α) and mobile (β) states of the side chain. In PC:PE:PS bilayers, there is strong overall immobilization of the nitroxide relative to DM revealed by an increase in the splitting of the outer hyperfine extrema (2A_zz’), an increase in ΔH₀, and a decrease in the spectral intensity in the β region of the spectrum.

Figure 6A[2] compares the corresponding spectra for R*. The small change upon light activation in DM previously reported (19) is reproduced, as shown by the R* - R difference spectrum of Figure 6A[3], and reveals a shift toward a more mobile population. In bilayers, 140R1 R* remains immobilized, and the change upon activation is entirely lost (Figure 6A[4]), although the 140R1 R* population contains both MI and MII with an equilibrium constant K = 1.5, consistent with that for the WT protein reconstituted in a similar lipid mixture at pH 6 (56, 57).

The strong immobilization of the 140R1 side chain in bilayers could mask small changes upon activation; therefore, studies were also carried out at 35 °C, and the corresponding spectra are shown in Figure 6B. The spectral line widths of 140R1 R and R* in DM are decreased due to increased molecular motion, and the change upon activation is similar in shape but much larger than at 20 °C (Figure 6B[1]-[3]). In PC:PE:PS bilayers, the nitroxide remains immobilized, and there is no change upon activation (Figure 6B[4]). The spectra of 140R1 in both R and R* are remarkably temperature-independent in the lipid bilayers.

One interesting possibility for the origin of the two dynamic modes of 140R1 is the presence of two protein conformations in equilibrium, but it is also possible that the two modes arise from two rotamers of the side chain in which one makes immobilizing interactions with the protein (50). The data do not allow these possibilities to be distinguished, so R1 was modeled in R to gain additional insight. As shown in Figure 6C, two commonly observed rotamers of R1, {t, m} and {t, p}, can be modeled at 140 in the 1GZM structure. In the {t, m} rotamer, the nitroxide ring can be positioned close to Y136 in a geometry that favors a CH-O hydrogen bond of the kind observed in crystal structures of R1 in T4 lysozyme (50). This interaction constrains the nitroxide and, together with other local interactions, could account for the more immobilized component in the spectra of 140R1 rhodopsin in DM. The mobile component could arise from the {t, p} rotamer, where the ring is positioned away from Y136 and free to move via torsional oscillations about the X₄ and X₅ dihedrals. Thus, modeling indicates that the two dynamic modes of 140R1 R and R* in DM could be due to a simple rotameric equilibrium but cannot eliminate the possibility of conformational heterogeneity of the protein.

The striking immobilization of 140R1 for both R and R* in bilayers may signal a local change in conformation relative to that in DM. However, another possibility is suggested by the model of Figure 6C. For example, 140R1 projects directly into the highly ordered headgroup domain of the bilayer, an environment which has been found to strongly damp the motion of the R1 side chain on membrane-associated annexin (58). In addition, if a CH-O H-bond is in fact formed, it may be strengthened in the partially hydrophobic environment of the headgroup region, leading to stabilization of the {t, m} state and immobilization. In either case, immobilization of 140R1 leads to loss of sensitivity to structural changes upon activation. Finally, immobilization of 140R1 could result from formation of a rhodopsin dimer in which 140R1 is at the dimer interface. The existence of a rhodopsin dimer remains controversial (59), but one model for a putative dimer has been proposed (60-62). In this model 140R1 is not in the dimer interface, and the presence of the dimer in the reconstituted membranes could not account for the observed immobilization of 140R1.

(B) 227R1

Residue 227 is located in TM5, directly facing the hydrocarbon chains of the bilayer below the headgroup region (Figure 2). Figure 7A[1] compares the spectra of 227R1 R in DM and in lipid bilayers. In DM, the spectrum reflects a dominant component with a relatively high degree of mobility characteristic of an R1 residue on the aqueous-exposed surface of an α-helix in soluble proteins (47), but with broader features likely due to a heterogeneous and more viscous environment of the micelle interior compared to water. In PC:PE:PS bilayers, the spectrum reveals an overall decrease in mobility relative to that in DM, as evidenced by an increase in ΔH₀ [Δ(ΔH₀) = 0.6 G], a corresponding decrease in spectral amplitude in the normalized spectra, and the appearance of a component with partially resolved high- and low-field extrema at the positions of the vertical dotted lines. The decrease in mobility is likely due in part to the increased molecular order of the bilayer environment near the headgroup region.

Figure 7A[2] compares the spectra of 227R1 R* in DM and lipid bilayers. In DM, the salient difference between R and R* is the appearance of a strongly immobilized component at the expense of more mobile states, as shown by the positive and negative features in the R* - R difference spectrum in the α and β regions, respectively (Figure 7A[3]). This previously reported light-dependent change was interpreted to reflect new immobilizing interactions of 227R1 with groups in TM6 as it moves outward (63).

In lipid bilayers, the overall mobility in R* is again decreased relative to DM, but the light-dependent change is retained, although of smaller magnitude (Figure 7A[3],[4]). In the bilayers, 227R1 R* contains both MI and MII (K = 0.7). The smaller magnitude of light-dependent change in the bilayers may be due in part to the presence of MI, absent in DM, which is expected to have a conformation similar to R (64).

For reference, Figure 7B shows the 227R1 side chain modeled in the 1GZM structure. In the unrestrictive environment at this site, all three common rotamers can be modeled without steric clashes. In each, the nitroxide has an environment dominated by direct contact with the hydrocarbon chains of either the micelle or the bilayer. The multiple rotamers and the freedom of motion about X₄ and X₅ are consistent with a dominant component of relatively high mobility observed at this site in R.

(C) 250R1

Residue 250 lies in TM6, projecting into the interior of the helix bundle (Figure 2). Unlike 140R1 and 227R1, 250R1 does not have contacts with the lipid or micelle environment, and the motion reflects only interactions with the protein. Figure 8A[1] compares the EPR spectra of 250R1 for R in DM and in lipid bilayers of PC:PE:PS. In both environments the spectra are dominated by a strongly immobilized state, but spectral intensity in the β region indicates the presence of a more mobile state as well (compare with Figure 5B). The values of ΔH₀ and 2A_zz’ in lipid bilayers are larger than those for DM, indicating a lower mobility for R1 in bilayers.

Figure 8A[2] compares the spectra of R* in DM and lipid bilayers, where the overall similarity of the spectra is evident. The dramatic line shape change upon activation, in both DM and lipid bilayers, indicates a shift in the R1 population toward a more mobile state (Figure 8A[3],[4]). In solutions of DM, this was previously interpreted in terms of an outward movement of TM6 that removes constraints on the motion of 250R1 (30, 63), a hallmark of the receptor activation. The fact that this activating conformational change and the coupled change at 227R1 are preserved in lipid bilayers is an important conclusion of this study.

To quantitatively characterize the motion for R1 at this important site, the spectra were fit to a two-component MOMD model as described in Methods. The fits, shown as dotted traces overlaid on the experimental spectra in Figure 8B, provide values for the order parameters (S), effective correlation times (τ_c), and fractional occupancies (f) for the components (Table 1).

For R in DM, τ_c = 77 ns for the more immobile state corresponding to the rotational diffusion of a spherical object of radius 42 Å, the same as the hydrodynamic radius determined for R in DM micelles by gel permeation chromatography (65). This indicates that the nitroxide population that gives rise to this component has a motion determined by the overall rotational diffusion of the R-DM micellar complex. For R in PC:PE:PS bilayers, τ_c ≈ 200 ns for the immobile component. The rotational diffusion of R in bilayers is slow on the X-band EPR time scale (τ_c ∼ 10-20 μs; 66, 67), and the correlation time for the nitroxide apparently represents slow fluctuation of the nitroxide with respect to the protein.

In both DM and bilayers, the mobile component has weak ordering (S ∼ 0.3) and fast motion (τ_c ∼ 3 ns), corresponding to an R1 side chain having little tertiary interaction with the protein (68). A difference between 250R1 in DM and bilayers lies in the significantly larger proportion of the mobile component in DM.

MOMD fits to the R* spectra in DM and bilayers again reveal the presence of both relatively immobile and mobile components, but with an increased proportion of the mobile state relative to R. However, in both environments the more immobile component has a significantly shorter τ_c relative to R.

In DM, 250R1 R* is pure MII, while in lipid bilayers MI is also present, but at relatively low levels (K = 4). The larger value of K relative to WT under similar conditions (K = 1.5) suggests a mild perturbation of the equilibrium by 250R1 favoring MII.

Figure 8C shows a model of the R1 side chain at site 250 in the 1GZM rhodopsin structure looking down at the cytoplasmic surface. Due to the constrained environment around 250, only a single rotamer can be built without atomic overlaps or altering the structure of the protein. This unusual rotamer is {m, t}, the same as that observed in the crystal structure of T4L 75R1 where the environment is similarly constrained, and is not one of the “relaxed” conformations of the side chain seen at solvent-exposed helix surface sites (50). The nitroxide ring resides in an existing pocket in R, making numerous van der Waals contacts with neighboring atoms that would strongly constrain its motion, accounting for the component in the EPR spectrum corresponding to an immobilized state.

Considering that no other rotamer of the nitroxide can be modeled in R that would give rise to a mobile state of R1, it is likely that the more mobile component represents a second conformational substate of the protein in equilibrium with that corresponding to the crystal structure. This interpretation is strengthened by the fact that this component is reduced in PC:PE:PS bilayers, although the nitroxide itself is well removed from direct contact with the bilayer environment. Models for the putative substates of the protein in both R and R* will be considered below in the Discussion.

(D) 316R1

Residue 316 resides on the solvent-exposed surface of the short helix 8 (H8) that lies perpendicular to the TM helical bundle (Figure 2). Like 250R1, the side chain is not expected to have direct contacts with the micellar or bilayer environment. Figure 9A[1] compares the EPR spectra of 316R1 for R in DM and in PC:PE:PS bilayers. As for 140R1 and 250R1, the spectra each reveal two well-resolved components corresponding to relatively immobilized and mobile states of R1, but with a different population ratio, reflecting a local structural difference in DM and lipid bilayers around 316. As discussed above for 140R1, the two states of R1 could arise from conformational heterogeneity of either R1 or the protein, and these possibilities cannot be distinguished from the spectra alone.

A similar difference between DM and lipid bilayers is seen in R*, although the effect is smaller (Figure 9[2]). The change upon activation in DM corresponds to a small shift in the population toward the immobile state, as reported earlier (Figure 9[3]). The same change with larger amplitude is observed in bilayers (Figure 9[4]), due primarily to the difference in the R state spectra. In DM, 316R1 R* is pure MII; in lipid bilayers, MI is also present (K = 1.5 at pH 6, the same as for the WT protein).

A model of the R1 side chain at 316 is shown in Figure 9B. Due to the crowded environment around this site, only the {m, m} rotamer can be built, but both ±90 configurations of X₃ (the disulfide) can be accommodated. In one (X₃ = -90, shown in Figure 9B), multiple contacts of the R1 side chain with residues H65 and K67 in the TM1-TM2 loop are made, a situation which could give rise to the immobilized state. In the other (X₃ = +90, not shown), the ring projects outward over the H8 helix where no obvious interactions of the ring are possible, allowing freedom of motion about X₄ and X₅. Due to the slow interconversion rate about X₃ on the EPR time scale, this situation could give rise to resolved mobility states of R1. In accounting for the two dynamic states of 316R1, it must also be considered that the C-terminal tail of the molecule, not resolved in the crystal structure, could lie over the H8 helix and make contacts with 316R1.

pH Dependence of R and R* Conformations in DM and Phospholipid Bilayers

The dependence of structure on pH is of particular importance for rhodopsin and other GPCR’s because the population of the active conformation is directly modulated by pH. Changes in pH can produce changes in the motion of R1 due to local or global structural effects. Local effects arise from protonation changes of residues in the immediate environment without consequences for the overall protein structure and complicate interpretation of spectral changes in terms of protein conformational changes. Residues 227R1 and 316R1 were found to show significant spectral changes with pH even in R (data not shown). While these could represent global conformational transitions, the presence of nearby basic residues could have local effects. In the case of 227R1, nearby lysine residues K231 and K248 reside at the bilayer-water interface, where the pK_a could be shifted downward into the pH 6-8 range. For 316R1, the nitroxide ring is in close proximity to H65, which titrates in the relevant pH range. Thus, 227R1 and 316R1 were not considered for further investigation of pH-dependent conformational changes. 140R1 was not employed because light-dependent changes are completely masked in lipid bilayers. On the other hand, the R spectrum of 250R1 shows no pH dependence in the range 6-8, and the dependence of R1 dynamics on protein conformation makes it a suitable choice for exploring the effect of pH.

Parts [1] and [2] of Figure 11A show the effects of pH on the EPR spectra for both R and R* in DM, where only MII is populated. There is little effect of pH on R, but a small reversible pH change is noted in R*. The insets are provided to clearly reveal the changes in the low- and high-field regions, which show that high pH shifts the population toward the more immobilized state, i.e., toward a state more closely resembling the R state. Consequently, the R* - R difference spectra (Figure 10A[3],[4]) show a larger change at pH 6 compared to pH 8. These results are compatible with a pH-dependent equilibrium between conformational substates for the activated receptor in the detergent micelle. Because MII_a and MII_b in DM titrate with a pK_a of 6.75, the substates are tentatively identified with these intermediates. In this case, the spectra of 250R1 at pH 6 and 8 largely reflect the structures of MII_b and MII_a, respectively. For reference, Figure 10B compares the spectrum of R with the spectra assigned to these species.

The corresponding data for 250R1 in POPC:POPE:POPS in Figure 10C show qualitatively similar results, suggesting that MII_a and MII_b states may exist in the bilayer as well as in the DM environment. The EPR spectra for 250R1 in pure POPC bilayers were identical to those shown in Figure 10C. Although qualitatively similar to DM, there is an increase in the immobilized population in the lipid bilayers relative to DM at both pH values, likely due in part to the presence of MI, which is expected to more closely resemble R with an immobilized 250R1 (64). For 250R1 in bilayers of PC:PE:PS, K = 4 at pH 6 and K = 1 at pH 8, with corresponding MI populations of 30% and 50%, respectively. In addition, the detailed dynamics of the MII_a and MII_b substates, and hence the corresponding EPR spectra, could be different in bilayers and DM micelles, just as the dynamics of rhodopsin itself are different in these two environments.

R and R* in Digitonin

Digitonin is unique among micelle-forming amphiphiles in that the hydrophobic core of the micelle is essentially solid compared to the fluid interior of most micelles (69). Figure 11 shows the EPR spectra of R and R* for each of the sensor mutants in digitonin compared to DM and PC:PE:PS bilayers, each at pH 6. Overall, there is a striking immobilization of all R1 residues except for 316R1 in both R and R* in digitonin relative to DM.

For 140R1, the immobilization in digitonin is similar to that in lipid bilayers in both R and R* and likely has a similar origin, namely, direct contact with the solidlike interior of the digitonin micelle and with the highly ordered headgroup region of the bilayers, respectively (Figure 11A[1],[2]). As for the bilayer environment, there is no change upon activation to R* (Figure 11A[3]). For 227R1, there is a progressive decrease in overall mobility in the order DM > lipid bilayers > digitonin in both R and R* (Figure 11B[1],[2]), consistent with the view that this residue directly senses the hydrophobic interior of the respective structures. As for 140R1 in digitonin, there is no change in the EPR spectrum upon activation at 227R1 (Figure 11B[3]).

The EPR spectra of 250R1 R in digitonin and PC:PE:PS bilayers are similar (Figure 11C[1]), both showing a suppression of the more mobile population relative to DM. A dramatic difference is seen between 250R1 R* in digitonin micelles and in DM or lipid bilayers; digitonin effectively blocks the large increase in mobility of R1 and presumably the activating movement of TM6, seen in either of the other two environments (Figure 11C[2],[3]). Nevertheless, in digitonin 250R1 R* consists of approximately 20% MI and 80% MII_a at 20 °C, as determined by our visible absorbance spectra of the 250R1 mutant, which is consistent with the analysis of Szundi et al. on WT rhodopsin (23). This result shows that the movement of TM6 at the cytoplasmic surface detected by 250R1 can be uncoupled from the deprotonation of the K296 Schiff base that accompanies the formation of MII_a. Unlike the other sites, the spectra of 316R1 are qualitatively similar in DM and digitonin both in R and in R*, although the change upon light activation is much smaller in digitonin (Figure 11D).