Lipid Protein Interactions Couple Protonation to Conformation in a Conserved Cytosolic Domain of G Protein-coupled Receptors*

The visual photoreceptor rhodopsin is a prototypical class I (rhodopsin-like) G protein-coupled receptor. Photoisomerization of the covalently bound ligand 11-cis-retinal leads to restructuring of the cytosolic face of rhodopsin. The ensuing protonation of Glu-134 in the class-conserved D(E)RY motif at the C-terminal end of transmembrane helix-3 promotes the formation of the G protein-activating state. Using transmembrane segments derived from helix-3 of bovine rhodopsin, we show that lipid protein interactions play a key role in this cytosolic “proton switch.” Infrared and fluorescence spectroscopic pKa determinations reveal that the D(E)RY motif is an autonomous functional module coupling side chain neutralization to conformation and helix positioning as evidenced by side chain to lipid headgroup Foerster resonance energy transfer. The free enthalpies of helix stabilization and hydrophobic burial of the neutral carboxyl shift the side chain pKa into the range typical of Glu-134 in photoactivated rhodopsin. The lipid-mediated coupling mechanism is independent of interhelical contacts allowing its conservation without interference with the diversity of ligand-specific interactions in class I G protein-coupled receptors.

The visual photoreceptor rhodopsin is a prototypical class I (rhodopsin-like) G protein-coupled receptor. Photoisomerization of the covalently bound ligand 11-cis-retinal leads to restructuring of the cytosolic face of rhodopsin. The ensuing protonation of Glu-134 in the class-conserved D(E)RY motif at the C-terminal end of transmembrane helix-3 promotes the formation of the G protein-activating state. Using transmembrane segments derived from helix-3 of bovine rhodopsin, we show that lipid protein interactions play a key role in this cytosolic "proton switch." Infrared and fluorescence spectroscopic pK a determinations reveal that the D(E)RY motif is an autonomous functional module coupling side chain neutralization to conformation and helix positioning as evidenced by side chain to lipid headgroup Foerster resonance energy transfer. The free enthalpies of helix stabilization and hydrophobic burial of the neutral carboxyl shift the side chain pK a into the range typical of Glu-134 in photoactivated rhodopsin. The lipid-mediated coupling mechanism is independent of interhelical contacts allowing its conservation without interference with the diversity of ligandspecific interactions in class I G protein-coupled receptors.
G protein-coupled receptors (GPCRs) 2 are hepta-helical membrane proteins that couple a large variety of extracellular signals to cell-specific responses via activation of G proteins. In the visual photoreceptor rhodopsin, a prototypical class I GPCR (1,2), molecular activation processes can be monitored in real time by spectroscopic assays and analyzed in the context of several crystal structures (3)(4)(5)(6)(7)(8). The primary signal for rhodopsin is the 11-cis to all-trans photoisomerization of retinal covalently bound to the apoprotein opsin through a protonated Schiff base to Lys 296 . Current models converge toward a picture in which "microdomains" act as conformational switches that are coupled to different degrees to the primary activation process. Two activating "proton switches" have been identified (9) as follows: breakage of an intramolecular salt bridge (10) by transfer of the Schiff base proton to its counter ion Glu-113 (11) is followed by movement of helix-6 (H6) (12,13) in the metarhodopsin II a (MII a ) to MII b transition. The MII b state takes up a proton at Glu-134 (14) in the class-conserved D(E)RY motif at the C-terminal end of helix-3 (H3) leading to the MII b H ϩ intermediate (15,16), which activates transducin (G t ), the G protein of the photoreceptor cell. Glu-134 regulates the pH sensitivity of receptor signaling (17) in membranes as reviewed previously (18), and in complex with G t the protonated state of the carboxyl group becomes stabilized (19). This charge alteration is linked to the release of an "ionic lock," originally described for the ␤ 2 -adrenergic receptor (20), which also in rhodopsin stabilizes the inactive state (16) through interactions between the cytosolic ends of H3 and H6 (21).
In the absence of a lipidic bilayer, proton uptake and H6 movement become uncoupled (15). Lipidic composition affects MII formation, rhodopsin structure, and oligomerization (22)(23)(24) and differs at the rhodopsin membrane interface from the bulk lipidic phase (25). Likewise, MII formation specifically affects lipid structure (26). Although of fundamental importance for GPCR activation, the potential implication of lipid protein interactions in "proton switching" is not clear. A functional role of Glu-134 in lipid interactions has been originally derived from IR spectra where E134Q replacement abolished changes of lipid headgroup vibrations in the MIIG t complex (19). Computational approaches emphasized the "strategic" location of the D(E)RY motif (27), and the Glu-134 carboxyl pK a may critically depend on the lipid protein interface (28). However, the implications for proton switching are not evident, and the theoretical interest is contrasted by the lack of experimental data addressing the effect of the lipidic phase on side chain protonation, secondary structure, and membrane topology of the D(E)RY motif.
We have studied the coupling between conformation and protonation in single transmembrane segments derived from H3 of bovine rhodopsin. We have assessed the "modular" function of the D(E)RY motif by determining parameters not evident from the crystal structures, i.e. the pK a of the conserved carboxyl, its linkage to helical structure, and the effect of protonation on side chain to lipid headgroup distance. We show that the D(E)RY motif encodes an autonomous "proton switch" controlling side chain exposure and helix formation in the low dielectric of a lipidic phase. The data ascribe a functional role to lipid protein interactions that couple the chemical potential of protons to an activity-promoting GPCR conformation in a ligand-independent manner.
Fourier Transform IR Spectroscopy and Circular Dichroism-Fourier transform IR (FTIR) spectra were obtained with peptides (ϳ10 mg/ml) in 5% DM, 100 mM sodium phosphate buffer, using a vector 22 spectrometer (Bruker, Ettlingen, Germany) at 2 cm Ϫ1 resolution. 30 l of the sample were transferred into a Bio-ATR-II cell (Bruker, Ettlingen, Germany) and the pH changed from 8.8 to 3 by dialysis (100 mM sodium phosphate). Pure buffer spectra were used as spectral references to calibrate the pH-dependent absorption strength and frequency of the phosphate stretching modes. Therefore, a very sensitive real time monitor of pH changes could be implemented. Circular dichroism spectra were recorded at room temperature with a J-815 instrument (Jasco, Gross-Umstadt, Germany) at ϳ4 liters/min N 2 flow rate from 200 to 260 nm in 0.1-cm cuvettes on DM-solubilized peptides (0.2-0.3 mg/ml). Ellipticity (⍜) was recorded in millidegrees.
Fluorescence Spectroscopy-Fluorescence measurements were performed with an LS55 spectrometer (PerkinElmer Life Sciences) in 200-l cuvettes at 285 nm for Trp excitation and emission recorded from 290 to 450 nm (slit width of 5 nm). Peptide-containing micelles were prepared by solubilization of 1-2 mg of peptide in ϳ200 l of 5% DM followed by 30-fold dilution in 100 mM phosphate buffer of different pH. "Lipiddoped" peptide-containing micelles were prepared by dissolving ϳ2 mg of peptide in 5% DM (ϳ600 l) and mixing with the appropriate vacuum-dried amounts of dansyl-PE and PS in a molar ratio of 1:2.5:1, respectively. 20 l aliquots were diluted 30-fold in a series of phosphate buffers (100 mM) allowing duplicate recording of emission spectra at each pH. Peptidecontaining vesicles were prepared by mixing 1-2 mg of peptide solubilized in ϳ 800 l of 5% OG with PS, dansyl-PE, and PC in the ratio 1:2.5:2.5, respectively, with a final total lipid:peptide ratio of 50 -80. The detergent was removed by 15 h of flow cell dialysis, and the suspension was sonicated and freeze-thawed 10 times and diluted in 100 mM phosphate buffer. Cholesterolcontaining vesicles were prepared in parallel from the same stock by supplementing an aliquot of the OG-solubilized mixture with cholesterol to a final total lipid:cholesterol ratio of 1:0.2.
Synthetic Peptides-Peptides were synthesized and high pressure liquid chromatography-purified (free of trifluoroacetate) by ThermoFisher (Ulm, Germany) with C and N termini amidated and acetylated, respectively. Peptides of the following sequences were derived from amino acids 108 -138 in ; native amino acids in parentheses were replaced by the preceding ones. The helicity of the peptides was assessed by infrared spectroscopy and circular dichroism measurements. As shown in Fig. 1, the degree of helicity depends strongly on even very subtle changes in the sequence. In the context of the most native-like sequence in peptide H3 E , the additional shift of the tryptophan from position 19 to 31 as well as the further replacement of the Glu-134 homologue by glutamine does not interfere with a predominantly helical secondary structure (inset). Spectral decomposition of the integral IR absorption in the amide I range into gaussian/lorentzian curves shows 70 -80% ␣-helical structure for H3 E and H3 W19F/V31W/E27Q with peak frequencies at 1654 -1656 cm Ϫ1 . The CD spectra of the latter peptides ( Fig.  1, inset) were converted to mean residue ellipticity in FIGURE 1. Secondary structure of H3-derived peptides. The infrared spectrum of solubilized H3 E shows a predominant helical secondary structure exhibiting an amide I absorption at 1655 cm Ϫ1 in agreement with previous data of its lipid-reconstituted form (48). Replacement of E27 by either Asp or Gln leads to mis-folding of the corresponding peptides H3 E27D and H3 E27Q evident from weak amide absorption at 1652-1656 (two lower traces) and featureless CD spectra (inset). However, the E27Q replacement is not detrimental to helix formation in the background of the additional W19F replacement, resulting in the peptide H3 W19F/V31W/E27Q with a clear spectral signature of helical secondary structure in both IR and CD data. Helical content is 75-80% (from spectral decomposition of IR absorption) and 70% (from CD, see "Experimental Procedures"). mAU, milliabsorbance units. degrees⅐dmol Ϫ1 ⅐cm Ϫ2 as ⍜ res ϭ 100⅐⍜/(c⅐l⅐n), where c is the molar concentration of the peptide; l is optical path length (0.1 cm), and n is the number of amino acids in the peptide (here 31). The fraction p of helical structure was calculated as p ϭ Ϫ⍜ res222 /(39,500⅐(1-2.57/)) as described (29). Consistency requires that the average number of peptide bonds in helical conformation complies with /31 ϭ p, which was obtained for ϭ 22 and p ϭ 0.71. Thus, ϳ22 out of the total of 31 amino acids in each of the two peptides are on average in a helical conformation that is on the lower side of the range estimated by FTIR spectroscopy and suggests that 4 -5 disordered amino acids form the C-and N-terminal helical ends. In contrast to the largely helical peptides H3 W19F/V31W and H3 W19F/V31W/E27Q , the related pair of sequences H3 E and H3 E27Q shows helicity only for H3 E . Three independent synthesis rounds using different technologies failed to produce peptides H3 E27Q and H3 E27D with helical structures upon detergent solubilization. The CD spectra of these peptides are rather featureless (Fig. 1, inset), and their IR absorption shows only a small fraction of helical structure (1656 cm Ϫ1 ) relative to a predominating amide I mode at low frequency indicative of aggregation. Thus, under the conditions of the FTIR and CD experiments, the replacement of Glu-27 by Gln allows helix formation only in the presence of the additional W19F replacement. Therefore, the proton-induced conformational switch was studied by FTIR in the least modified sequence, i.e. H3 E , whereas the negative control was provided by H3 W19F/V31W/E27Q . For ICL2, AIE RYV W(V)VC KPM SNF RFG, this peptide was derived from amino acids 132-149, containing the C-terminal end of H3 and its cytosolic extension.

Coupling of Protonation and Conformation by D(E)RY Motif of Rhodopsin-
The linkage between conformation and carboxyl protonation in the D(E)RY motif was studied in the peptide H3 E corresponding to amino acids 108 to 138 of rhodopsin. The native residues Glu-113 and Glu-122 were replaced by alanines, rendering the carboxylate in the Glu-134 homologue (i.e. Glu-27 in the peptide) the only carboxyl in H3 E . The symmetric carboxylate stretching mode of Glu-27 (1401 cm Ϫ1 ) and the narrow amide I absorption at 1656 cm Ϫ1 (typical of the helical secondary structure of the peptide backbone) were used to monitor by FTIR spectroscopy the side chain protonation and secondary structure, respectively. Fig. 2A shows the absolute IR absorption and the pH-induced difference spectra, obtained by dialysis-coupled attenuated total reflectance-FTIR difference spectroscopy (30). Spectra recorded at different pH values were subtracted from a reference spectrum acquired at pH 8.8, where the Glu-134 homologue is fully ionized (maximal absorption at 1401 cm Ϫ1 ). The stretching vibrations of the buffer phosphates between 1200 and 800 cm Ϫ1 served as a pH monitor. The macroscopic pK a of Glu-27 was identified by the spectrum with half-maximal intensity of the 1401 cm Ϫ1 band. The latter was obtained at pH 5.9, i.e. 1-2 pH units above the pK of a glutamate side chain in aqueous solution. Also the peptide backbone exhibited pH-dependent transitions, giving rise to changes of the amide II band around 1550 cm Ϫ1 (overlapping with the antisymmetric carboxylate stretching mode at ϳ1560 cm Ϫ1 ) and the amide I absorption at 1656 cm Ϫ1 . The structural change was evaluated using the titration spectra in Fig. 2B. The inset in Fig. 2B shows that the 1656/1638 cm Ϫ1 absorption change correlates linearly with the carboxylate absorption at 1401 cm Ϫ1 . The secondary structural change is coupled to side chain protonation with a common pK a of 5.9, with a more helical structure favored at acidic pH. The integral intensity change at 1656 cm Ϫ1 corresponds to 3-5% of the total amide I absorption of H3 E ( Fig. 2A), indicating that side chain protonation extends the helical structure by an average of 1-2 peptide bonds. This process was abolished upon E27Q replacement in the peptide H3 W19F/V31W/E27Q (readily adopting a helical structure in contrast to the alternative negative control sequence H3 E27Q , see "Experimental Procedures"), which did not exhibit the pH-sensitive 1656/1638 cm Ϫ1 difference band (Fig. 2B). Thus, the H3 secondary structure is controlled by the protonation state of the carboxyl in the D(E)RY motif. The functionality of the proton-dependent rearrangement of the C-terminal structure in the H3 W19F/V31W background containing the Glu-134 homologue is shown below and proves its independence from the altered location of the Trp.

Protonation of the D(E)RY Motif Alters the C-terminal
Structure-The upshift of the pK a relative to that of a glutamate in water indicates that side chain protonation at the H3 C terminus is stabilized by the concomitant helix extension. To prove the localization of the structural transition, a Trp was C-terminally attached, replacing the Val-138 homologue of rhodopsin, and for the purpose of specificity, Trp-19 (corresponding to the native Trp-126) was replaced by Phe, resulting in the construct H3 W19F/V31W , which shows a pH-sensitive emission in the 300 -400 nm range (Fig. 3). Trp-31 is two residues away from the tyrosine of the D(E)RY motif. It acts as an energy acceptor for the excited state of tyrosine, which shows appreciable emission only in the absence of Trp-31 or in the presence of the more distant Trp-19 (Fig. 3, inset A). The Trpdominated emission in H3 W19F/V31W provides a sensitive fluorescence monitor of structural transitions in the D(E)RY motif due to both the geometrical constraints for FRET and the hydrophobicity dependence of Trp emission itself. The pH-induced emission difference spectra show that in the ionized state of Glu-27 (positive lobe at 365 nm) the emission of Trp-31 is shifted to longer wavelengths versus the protonated state (negative lobe at 320 nm). This could reflect reduced quenching of tyrosine as well as exposure of Trp to a more hydrophobic environment at pH Ͻ6. The half-maximal emission change from the D(E)RYVW sequence is interpolated to pH 5.6 and is lost upon E27Q replacement. This parallels the infrared results and supports the localization of the detected secondary structure formation within the direct vicinity of the D(E)RY motif. The protonation-induced blue shift of the Trp emission maximum is also observed with H3 W19F/V31W in DOPC vesicles (Fig. 3), although with an ϳ50% reduced amplitude that is probably caused by the presence of zwitterionic (i.e. more hydrophilic) headgroups as compared with the neutral phase boundary of the DM micelle. Due to the strong light scatter in the near-UV light caused by the vesicles, the signal to noise ratio is also ϳ5-fold reduced. In contrast to the systematic pH dependence of Trp emission shown for the TM3-derived peptides, an unsystematic pH dependence was obtained from the peptide ICL2 (Fig. 3, inset B) containing the D(E)RY motif and the ensuing second cytosolic loop of rhodopsin (Ala-132 to Gly-149 with V138W replacement).
Protonation-dependent Repositioning of the D(E)RY Motif at the Hydrophilic/Hydrophobic Phase Boundary-Side chain neutralization and helix extension in the vicinity of the D(E)RY motif is expected to alter the hydrophobic length of the H3 segment. To test whether these coupled processes affect the distance between the H3 C terminus and the phase boundary in a micellar or lipidic environment, the C-terminal Trp of the peptide H3 W19F/V31W was employed as a donor fluorphore whose emission is quenched by a dansyl group (via FRET) in measurements, including dansyl-labeled phosphatidylethanolamine (dansyl-PE). Fig. 4 (inset) shows that excitation at 285 nm of "dansyl-PEdoped" peptide-containing micelles evokes emission from Trp and dansyl at 345 and 524 nm, respectively. Trp emission was typically 10 -30% of that in the absence of dansyl-PE. The 524 nm emission increased upon acidification, whereas Trp emission decreased, indicating a more efficient energy transfer when Glu-27 is protonated. dansyl absorption and emission are intrinsically pH-independent in the pH 4 -10 range (31,32). To correct for the different amounts of dansyl-PElabeled micelles or vesicles in the different samples, however, the residual Trp emission was normalized with respect to the fluorescence efficiency upon direct excitation of dansyl-PE at 336 nm in these systems. Therefore, the residual Trp emission from samples derived from the same stock of a given lipidic composition but transferred to different buffers is scaled to the same total concentration of vesicles in each buffer. The evaluation is restricted to the quenching of the donor fluorescence, i.e. to the residual Trp emission at 345 nm. This allows comparing the pH sensitivity of dansyl-mediated quenching of Trp-31 emission in micelles and vesicles consisting of PC, PS, and cholesterol at different pH values (Fig. 4) and independently of the determination of absolute FRET efficiency. For better comparison of the apparent pK a values obtained from fits to Henderson-Hasselbach curves, the traces (and original data) were scaled to the fitted signal level at pH 9 for each sample. The quenching of Trp-31 exhibited an apparent pK a of 6 Ϯ 0.3, which was independent of the assessed lipid composition and of the presence of cholesterol within the accuracy of the experiment. Trp fluorescence was more efficiently quenched at acidic pH indicating that upon neutralization of Glu-27 the C terminus moves from an aqueous to a more hydrophobic environment near the lipid headgroups. This agrees with the slight blue shift of the Trp emission (causing a negligible change of the overlap integral between dansyl absorption and the Trp emission of Ͻ0.5%) and would favor the deactivation of the excited state of Trp-31 upon shortening of the averaged donor to acceptor distance. Alternatively, the C-terminal Trp may become rotationally constrained if it gets partially immersed in the lipidic phase. This could also reduce Trp emission if a more favorable alignment with the dansyl electronic transition moment is achieved. The same mechanistic conclusion would apply; the H3 W19F/V31W C terminus moves from an exposed to a more lipid-immersed state with Glu-27 protonated, whereas the opposite transition contradicts the data. No attempt was made to determine and compare absolute FRET efficiencies in the different lipidic compositions by further evaluation of the dansyl emission at 524 nm in relation to Trp quenching. The unknown effect on the rotational freedom of both the C-terminal Trp and the dansyl group in the different lipidic compositions would render a more detailed analysis ambiguous. ii-vii, difference spectra generated by subtraction of the absorption at pH 3 (Glu-27 side chain protonated) from the spectra recorded during acidification (color-coded blue to red) show the decreasing intensity of the symmetric COO Ϫ stretching vibration of Glu-27 at 1401 cm Ϫ1 . Absorption changes of the buffer phosphates (1200 to 800 cm Ϫ1 ) provide an internal pH reference. Blue traces were generated by subtracting the "pH 8.8 minus pH 5.5" IR difference spectrum of calibration buffers from the dialysis-induced difference bands in the peptide sample. Inflection of the residual phosphate bands in (v and vi) shows that half-maximal absorption at 1401 cm Ϫ1 is obtained by a change from pH 8.8 to pH Ͼ5.5. B, linear relation between carboxyl side chain protonation and helix extension. Acidification causes reduction of the carboxylate absorptions at 1401 and 1560 cm Ϫ1 (antisymmetric) and an increase of ␣-helical structure (negative band at 1656 cm Ϫ1 ). This pH sensitivity is lost upon E27Q replacement in H3 W19F/V31W/E27Q (black line, scaled to the same amount of peptide) showing essentially only the residual pH-dependent absorption by the micellar solution. Inset, plot of the absorption at 1656/1638 cm Ϫ1 versus the integral 1401 cm Ϫ1 band reveals a common pK a of 5.9 for side chain protonation and helix formation (vertical and horizontal lines). Upper and right axis show the fraction of protonated peptide and the corresponding pH, respectively, related by the sigmoidal curve (gray) in a four-state model (see text and Fig. 5A). mAU, milliabsorbance units.   OCTOBER 16, 2009 • VOLUME 284 • NUMBER 42

JOURNAL OF BIOLOGICAL CHEMISTRY 28805
Hydrophobicity Links Protonation to Conformation-The high pK a of carboxyl protonation in micelles and vesicles implies that the protonated state of the Glu-134 couples to a free enthalpy-delivering reaction. The coupling depends on the ␣-helix preceding the D(E)RY motif in a hydrophobic phase as it is not seen in ICL2 (Fig. 3). The low dielectric constant in vesicle membranes and in the interior of a micelle could promote both stabilization of the protonated side chain and of helical secondary structure. Therefore, we have asked whether a hydrophobicity-mediated proton switch is consistent with thermodynamic estimates of helix stability and charge-dependent partitioning of a carboxyl side chain at a phase boundary. The coupling of conformation to protonation is described by the microscopic equilibria defined in Fig. 5A. Peptide confor-mations with an exposed (E) or buried (B) state of the ionized carboxylate of Glu-27 interconvert with the equilibrium constant K C and their protonated counterparts (EH and BH) with the equilibrium constant K CH . State function properties require that the free enthalpy difference for the conformational transition in the protonated versus the ionized state (i.e. ⌬⌬G ϭ ϪRT ln K CH /K C ) equals that of transferring a neutral versus an ionized glutamic acid side chain from an aqueous to a hydrophobic medium (i.e. ⌬⌬G ϭ ϪRT ln K B /K E ). Free enthalpies ⌬G tr have been determined for water to octanol transfers (33). The reported ⌬⌬G of 14.7 kJ (3.52 kcal) for glutamic acid provides an estimate of the enthalpy available to link protonation to conformation by a purely hydrophobicity-driven mechanism. These transfer enthalpies do indeed reproduce the measured pK a of 5.9 for protonation/helix extension with a ⌬G c of 6 kJ for the transition of the ionized state from the exposed to the buried conformation. The resulting free energy surface of the system is shown in Fig. 5B. Upon acidification, the system moves along the "valley" from less than 10% protonation with Ͼ80% of the carboxylates exposed to the aqueous phase at pH 7 (microscopic pK B of the buried state) to more than 90% protonation with Ͼ80% of the neutral side chains buried at pH 4.4 (microscopic pK E of the exposed state). The equilibrium states trace out a trajectory that runs roughly diagonally through the conformation/ protonation plane reproducing the linearity between the measured amide I and carboxylate absorption changes.

DISCUSSION
Breakage of the ionic lock at the cytosolic H3/H6 interface in the transition to the G t-activating state of rhodopsin is linked to protonation of Glu-134 in the class-conserved D(E)RY motif in H3. This constitutes one of the two proton switches that control rhodopsin activation (34) and is likely to operate in other class I GPCRs (20,(35)(36)(37)(38). Here we have shown that the D(E)RY motif represents an autonomous proton switch that is not per se dependent on interhelical contacts. Instead, local lipid protein interactions alter the pK a of the side chain carboxylate by providing a medium of low dielectricity that stabilizes the proto- FIGURE 5. Thermodynamic parameters and structural model of the hydrophobicity-mediated coupling between carboxyl side chain protonation and peptide conformation. A, definition of microscopic equilibrium constants. Protonation occurs in conformations with the H3 glutamic acid side chain either exposed to water (E) or buried in the hydrophobic phase of a membrane or micelle (B). The free enthalpies of protonation in these two states define the association constants K E and K B and differ by 14.7 kJ, i.e. the difference in water to octanol transfer enthalpies (33). The same ⌬⌬G applies for the exposed to buried conformational transitions (equilibrium constants K C and K CH , with ⌬G values of 6.0 and Ϫ8.7 kJ, respectively). B, free enthalpy G sys of the coupled system. G sys describes a potential surface as a function of the proton saturation f H and the fraction (fract.) f E of exposed carboxyl side chains. G sys is the minimized sum of the individual chemical potentials i ϭ i°ϩ RT ln c i of all states at a given proton saturation (exposed ionized state chosen as reference, i.e. E°ϭ 0 kJ) and plotted from 5 to 95% saturation along both coordinates. The trajectory of minimal enthalpy (gray line) reproduces the observed linear relation between carboxyl and amide absorption (squares). The arrow delineates the amplitude of the conformational transition induced by a pH change from 4.4 (microscopic pK E of the exposed state) to 7.0 (microscopic pK B of the buried state). C, rhodopsin dark state structure (Protein Data Bank ID code 1LH9). The dotted line defines the putative membrane border. Tyr-136 of the D(E)RY motif is "displaced" toward the cytosol. Color code: red and blue, formal negative and positive, respectively; green, neutral; white, hydrophobic. D, structural interpretation of the four states of the D(E)RY motif. The Glu-134 side chain switches from a gauche in E to a trans rotamer in B, and Tyr-136 aligns with the other cytosolic tyrosines (data not shown) in opsin (Protein Data Bank ID code 3CAP) by sliding about one peptide bond deeper into the bilayer. Similar to H3 E , helicity extends further to the C-terminal end in opsin (symbolized by the ribbon thickness). In the realm of the late rhodopsin intermediates described previously (15), the E state corresponds to dark rhodopsin, and we assign the B and BH states to the MII b and MII b H ϩ intermediates, respectively. The EH state is probably not accessible by dark rhodopsin (44), where interhelical contacts prevent exploration of active-like substates and suppress dark noise to a much larger extent than in other GPCRs. Structure graphics were created with VMD (version 1.8.3). nated state, thereby coupling protonation to transmembrane positioning and helix extension. Qualitatively, the switching mechanism can be understood on the basis that carboxyl side chains are the most potent residues in defining transmembrane helix terminations (39). Quantitatively, switching in the D(E)RY motif is surprisingly efficient exhibiting a large and "symmetric" pH modulation of the conformation, which is ideally realized when ⌬G c ϭ 1 ⁄ 2⌬⌬G. Switching originates in the protonation-induced drop of the free enthalpy of carboxyl transfer by ⌬⌬G ϭ 14.7 kJ. In the coupled system (Fig. 5A), the measured pK a of 5.9 correlates with a ⌬G c of 6 kJ, which is indeed close to the above condition. However, ⌬G c is different from ⌬G tr coo-of 15.2 kJ (33) for the hydrophobic burial of an ionized glutamate. Therefore, the E 3 B transition cannot consist of the mere "swing out" of the carboxylate into the hydrophobic phase. The demonstrated formation of ϳ2 helical peptide bonds fully accounts for the reduction by 8 -9 kJ because of the stabilizing effect of additional intramolecular H-bond formation (40). Proton switching and reorganization of the lipid peptide interface, shown by the "tryptophan to lipid" FRET measurements, are thus in excellent agreement with the energetics of helix stability and side chain partitioning. The relation of ⌬G c ϳ 1 ⁄2⌬⌬G is a robust feature of the IR signature of the D(E)RY motif for coupling enthalpies that shift the microscopic pK of the buried glutamate to 6.7-8.0 (in our evaluation pK ϭ 7), i.e. close to the pK estimated for Glu-134 in MII b (34).
How can a lipid-mediated coupling mechanism be understood in the context of crystal structures of rhodopsin? The intrinsic properties of the D(E)RY motif favor the equilibration of substates such that 1) the Glu-134 side chain becomes protonated in the hydrophobic phase, 2) the helical propensity of the C-terminal end of H3 increases, and 3) the H3 C-terminal end approaches the membrane. Fig. 5D shows the H3 structure in dark rhodopsin and opsin, where the ionic lock is broken (7). In contrast to the dark state structure (Fig. 5C), Tyr-136 of the D(E)RY motif in opsin aligns with the other cytosolic tyrosines (not shown) indicative of a net shift of the H3 C terminus toward the membrane surface similar to the protonation response of the isolated H3 E segment. The most striking feature, however, is the transition from a 2 gauche state in dark rhodopsin to an extended rotamer of the Glu-134 side chain in opsin also seen in the "partially active" ␤ 2 -adrenergic structure (36,41). A lipid bilayer imposes further hydrophobic constraints not present in the crystals. Although the C ␣ atom of Glu-134 points to the putatively lipid-facing side already in the dark (42), the ionic lock prevents it from "partitioning" between exposed and buried states. However, once "liberated" after activation, it can be stabilized in the protonated, activity-promoting extended conformation by hydrophobic burial. In the absence of a lipidic phase, packing of Glu-134 against the helix-2 helix-4 interface in the crystal (7) appears to provide the low dielectric environment accommodating the (putatively neutral) side chain. The crystal structure supports such a hydrophobicity-controlled rotamer state as there are no specific interactions that favor the extended conformation (7). The partially active state-like features in opsin suggest that upon receptor activation in a bilayer, H3 becomes liberated from interheli-cal constraints such that it can insert the C-terminal end slightly more into the membrane. However, this occurs efficiently only upon protonation and lipidic burial of the extended Glu-134 side chain, i.e. by forming the MII b H ϩ species. Therefore, hydrophobic matching is achieved for the MII b transmembrane configuration of H3, for which the partially active opsin structure could serve as a model. In the native receptor, the cytosolic chemical potential of protons thus overcomes the geometric constraints on transmembrane position imposed on the MII a to MII b transition by the planar bilayer. The latter disfavors the burial of the strongly hydrophilic patch placed by the D(E)RY motif between the strongly hydrophobic flanking regions (Fig. 5D) but stabilizes it in the MII b H ϩ state.
Also in the absence of a planar lipidic phase, the same coupling of protonation to local carboxyl burial and helix stabilization occurs, as shown here in micelles. For the full-length delipidated receptor, however, the flexible nonplanar micellar environment can easily adjust to the hydrophobic profile of the protein, irrespective of protonation-dependent increase of hydrophobicity of the D(E)RY motif. After the release of dark state constraints, the repositioning of H3 relative to the rest of the hepta-helical bundle can therefore proceed without the need of the protonation-induced formation of a more contiguous hydrophobic profile across the interspersed D(E)RY motif. The micelle adopts to the MII b conformation without requiring MII b H ϩ formation. The latter occurs as an uncoupled local pH-dependent transition similar to that seen with the synthetic peptides that are "uncoupled" from the holoprotein. The reason for this is the lack of a defined hydrophobic thickness relative to which side chain rotamers or surface protonation of the receptor would be constrained in the detergent micelle. Therefore, a lipid-mediated proton switch is in full agreement with the pH insensitivity of the total amount of MII (i.e. the sum of states carrying an unprotonated Schiff Base) formed and with the uncoupling of H6 movement from proton uptake in detergentsolubilized rhodopsin (15). The intrinsic properties of the D(E)RY motif are thus sufficient to explain key features of the cytosolic proton switch in membranes and in delipidated rhodopsin.
The proposed autonomous local proton-dependent equilibration between activity-promoting and activity-impeding conformations of the D(E)RY motif at the phase boundary independent of specific ligand interactions further agrees with recent MD simulations of the ␤ 2 -adrenergic receptor (43). Here the ionic lock appears to fluctuate between the open and closed state without corresponding "global" conformational switching of the entire inactive-like receptor structure to an active conformation. Although proton exchange reactions were not modeled in the MD calculations, our data support the notion that receptor activation originates in the shift of locally defined structural equilibria. Fluctuations of the ionic lock imply a continuous "exploration" of its physical environment. In rhodopsin, this may lead to an efficient shift of the local conformational equilibrium once dark state constraints are lifted in MII or opsin. In rhodopsin, however, proper function relies on efficient minimization of dark noise, and there is no indication that protonation couples to conformation already in the dark (44).
Such a stringent selection for noise reduction has not occurred for the ␤ 2 -adrenergic receptor. Although the frequent formation of the ionic lock appears to be a typical trait of the inactive receptor, sporadic breakage of the ionic lock even in the presence of inverse agonists parallels the biochemical phenotype of a basal activity in ␤-adrenergic receptors (36, 45) not found in dark rhodopsin. Importantly, however, the conformational shift, whether in the inactive (␤ 2 -adrenergic receptor) or the active receptor state (rhodopsin), does not require a corresponding switch in the ligand-binding site and is shown here to operate as an isolated proton switch module even in the absence of the hepta-helical bundle. It is rather the stabilization of one of the substates (in rhodopsin the membraneburied protonated state of the Glu-134 side chain) that adds a ⌬G to the population of the other accessible substates, thereby shifting the equilibrium to an activity-promoting ensemble of conformations. Here we have shown that the lipidic phase contributes to the energetics of the functionally relevant conformational substates.
In a broader sense, the regulation of the cytosolic GPCR structure by re-equilibration of lipid protein interactions after removal of dark state constraints appears to originate in small stretches of class-conserved amino acids. Their physical properties are preserved in corresponding peptides demonstrated here for the isolated transmembrane segment H3 as a part of the ionic lock and were shown previously for the amphipathic helix-8 of rhodopsin (46), which participates in signal transfer to the cytosolic face along the transmembrane H-bond network of rhodopsin (47).
Conclusions-Similar to recent results on lipid-dependent secondary structure in rhodopsin (24), our data show that the lipidic phase constrains the structure of the D(E)RY motif, rather than adopting to it. We attribute a functional relevance to these constraints in linking protonation to GPCR conformation via a proton-driven structural re-equilibration at the protein-lipid interface which is independent of ligand-specific interactions. The evolutionary diversity underlying ligand specificity can thus be reconciled with the conservation of a cytosolic "proton switch" that is adapted to the general physical constraints of a lipidic bilayer.