Solid-state 2H NMR relaxation illuminates functionaldynamics of retinal cofactor in membraneactivation of rhodopsinAndrey V. Strutsa,b, Gilmar F. J. Salgadoc, and Michael F. Browna,d,1

aDepartment of Chemistry, University of Arizona, Tucson, AZ 85721; bDepartment of Physics, St. Petersburg State University, St. Petersburg 198904,Russia; cLaboratoire ARNA, Institut National de la Santé et de la Recherche Médicale (INSERM) U869, Institut Européen de Chemie et Biologie (IECB),Université de Bordeaux, F-33000 Bordeaux, France; and dDepartment of Physics, University of Arizona, Tucson, AZ 85721

Edited by Adriaan Bax, National Institutes of Health, Bethesda, MD, and approved March 3, 2011 (received for review October 6, 2010)

Rhodopsin is a canonical member of the family of G protein-coupled receptors, which transmit signals across cellular mem-branes and are linked to many drug interventions in humans. Herewe show that solid-state 2H NMR relaxation allows investigation oflight-induced changes in local ps–ns time scale motions of retinalbound to rhodopsin. Site-specific 2H labels were introduced intomethyl groups of the retinal ligand that are essential to the activa-tion process. We conducted solid-state 2H NMR relaxation (spin-lattice, T1Z , and quadrupolar-order, T1Q) experiments in the dark,Meta I, and Meta II states of the photoreceptor. Surprisingly, wefind the retinylidene methyl groups exhibit site-specific differencesin dynamics that change upon light excitation—evenmore striking,the C9-methyl group is a dynamical hotspot that corresponds to acrucial functional hotspot of rhodopsin. Following 11-cis to transisomerization, the 2H NMR data suggest the β-ionone ring remainsin its hydrophobic binding pocket in all three states of the protein.We propose a multiscale activation mechanism with a complexenergy landscape, whereby the photonic energy is directed againstthe E2 loop by the C13-methyl group, and toward helices H3 and H5by the C5-methyl of the β-ionone ring. Changes in retinal structureand dynamics initiate activating fluctuations of transmembranehelices H5 and H6 in the Meta I–Meta II equilibrium of rhodopsin.Our proposals challenge the Standard Model whereby a singlelight-activated receptor conformation yields the visual response—rather an ensemble of substates is present, due to the entropygain produced by photolysis of the inhibitory retinal lock.

GPCR ∣ solid-state NMR ∣ generalized model-free analysis

G protein-coupled receptors (GPCRs) (1–3) are the largestfamily of integral membrane proteins in the human genome,

and they are the targets of about 30% of all human pharmaceu-ticals. At present the 3D structures of several GPCRs—includingrhodopsin (4–8), the β1- and β2-adrenergic receptors (9, 10), andthe adenosine A2A receptor (11)—have been established in var-ious functional states (2). Yet the mechanisms of GPCR activa-tion remain elusive, as they are membrane-embedded moleculesthat are often recalcitrant to crystallization, and push the size lim-its of solution NMR spectroscopy. Rhodopsin is the quintessen-tial prototype for Family AGPCRs, because previous X-ray (5, 6)and electron diffraction (12) studies have yielded crystal struc-tures of its dark state (4–6) and several photointermediates(13, 14). Notably, the retinal ligand is buried deeply within a7-transmembrane (TM) helical bundle, where it locks the recep-tor in the inactive state, with negligible basal (constitutive) activ-ity (4). Upon light absorption, rhodopsin catalyzes the rapid andhighly selective 11-cis to trans isomerization of retinal (15, 16)switching it from an inverse agonist to an agonist. In the Metar-hodopsin I–Metarhodopsin II equilibrium, recognition sites areexposed for a heterotrimeric G protein (transducin or Gt), initi-ating the amplified visual response (17, 18). However, the light-activated Meta II state is transient and crystal deformation occursgiving low-resolution data (14). Crystal structures of ligand-free

opsin (19) with a bound synthetic Gt peptide (20) show anelongation of the protein (21, 22) due to helical displacements(19)—but opsin does not contain the activating all-trans-retinalligand, and its activity does not match rhodopsin (20).

Spectroscopic approaches including spin-label EPR (23),13C NMR (18, 24–26), and Fourier transform infrared (FTIR)(17, 27) studies are thus needed to establish the activation me-chanism of the photoreceptor as it underlies the visual process.Solid-state NMR spectroscopy (28) is particularly important, as itgives knowledge of both protein structure and dynamics in amembrane lipid environment (29). We show here that 2H NMRrelaxation (30, 31) delivers crucial missing information that can-not be obtained by any other method. Crystallography presents uswith only a static structure of the protein, while spin-label studiesof rhodopsin address conformational changes of the protein andnot the cofactor. From temperature-dependent relaxation times,we extracted the amplitudes, activation energies, and preexpo-nential factors for ligand motions through key states of the acti-vation pathway. Combining NMR data with the crystal structure,an activation mechanism is put forth whereby retinal triggerslarge-scale (millisecond) helical fluctuations of the receptor ina membrane lipid bilayer.

ResultsWe applied solid-state 2HNMR relaxation methods (30) to studyretinal bound to rhodopsin in the inactive dark state, as well asthe Meta I and Meta II states over a wide temperature range(≈150 K). Rhodopsin was regenerated with 11-cis-retinal 2H-labeled at the C5-, C9-, or C13-Me groups, and recombined withphospholipids followed by membrane alignment (32). Deletion ofany of the methyl substituents transforms all-trans-retinal from afull agonist to a very weak partial agonist (1, 33, 34). The X-raystructure of rhodopsin in the dark state (5) is compared inFig. 1A to ligand-free opsin (19), showing the activating move-ments of TM helices H5 and H6. However, the structure ofrhodopsin in the active Meta II state has never been determined.In Fig. 1B we show the chemical structure of retinal with itsnumbering scheme. Examples of solid-state 2H NMR spectra aredepicted in Fig. 1 C–E for rhodopsin in 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) recombinant membranes. Re-sidual quadrupolar couplings (RQCs) of hχQi ¼ 50.7–52.0 kHzare obtained from the motion-averaged 2H NMR line shapes,and evince rapid threefold spinning of the retinylidene methylgroups, with rotational correlation times <10−5 s down to at least

Author contributions: A.V.S., G.F.J.S, and M.F.B. designed research; A.V.S. and G.F.J.S.performed research; A.V.S. and G.F.J.S. analyzed data; and A.V.S. and M.F.B. wrotethe paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1To whom correspondence should be addressed. E-mail: mfbrown@u.arizona.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1014692108/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1014692108 PNAS ∣ May 17, 2011 ∣ vol. 108 ∣ no. 20 ∣ 8263–8268

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

−160 °C (SI Text). From the RQCs, methyl order parameters ofSC3

≈ 0.9 are calculated directly, corresponding to off-axis fluc-tuations of ≈15°. The dynamics include methyl libration withrespect to the unsaturated polyene, as well as reorientations ofthe polyene chain and β-ionone ring within the binding cavity.

Solid-State 2H NMR Spectroscopy Illuminates Retinal Mobility WithinBinding Pocket of Rhodopsin. To explore how the retinal dynamicschange during visual function, we measured 2H NMR relaxationtimes (spin-lattice or Zeeman, T1Z, and quadrupolar-order, T1Q)(30) for rhodopsin in POPC or POPC/1,2-dioleoyl-sn-glycero-3-phosphoethanolamine (DOPE) (3∶1) membrane bilayers.Exponential relaxationwas observed at all temperatures, with pro-nounced variations among the different methyl sites. All 2HNMRmeasurements were conducted below the melting temperature(TM) of the lipid bilayer, from−30 to−160 °C.Rotational diffusionand large-scale protein motions were thereby suppressed to revealthe internal (local) dynamics of the retinal cofactor. If only 2HNMR spectra were recorded (Fig. 1 C–E) we might conclude that

few differences exist among the retinylidene methyl groups—yetsite-specific variations in retinal mobility are clearly evident (29).Notably, the 2H NMR relaxation rates differ by up to an order ofmagnitude, due to the internal dynamics of the retinylidene ligand.

In Fig. 2Awe summarize our experimental 2HNMR relaxationexperiments for rhodopsin in the dark state. Arrhenius-type plotsof both the T1Z times and T1Q times are shown vs. inverse tem-perature for methyl substituents of retinal crucial to its function(SI Text). Structural fluctuations of retinal in the dark state areunaffected by lipid phase transitions or freezing of water, andthe relaxation data can be smoothly extrapolated to physiologicaltemperature. For the C5-Me group, in Fig. 2A a clear minimum inT1Z is evident at −120 °C, and for T1Q at −135 °C. By contrast, theC9- and C13-Me groups have T1Z and T1Q minima at lower tem-peratures, outside the measured range. Note that the T1Z andT1Q minima occur at higher temperature (to the left) for slowermotions, and conversely at lower temperature (to the right) forfaster motions (29). At a given minimum, the effective correla-tion time τc is related to the Larmor (resonance) frequency ω0

by τc ≈ 1∕ω0 ¼ 2 ns. Smaller correlation times are obtained athigher temperatures (ps). The relatively short T1Z and T1Q re-laxation times for the C5-Me group indicate a mainly 6-s-cis con-formation of the β-ionone ring (32), rather than 6-s-trans. (For a6-s-cis conformation the 1–7 interactions with hydrogen H8would give slower spinning of the C5-Me vs. 1–6 interactions withhydrogen H7 for 6-s-trans; see below.) By contrast, the polyeneC9- and C13-Me groups have longer T1Z times, showing greatermobility due to relatively weak noncovalent interactions withinthe retinal ligand, and with the surrounding amino acid residues.

Generalized Model-Free Analysis Explains Retinal Structure andDynamics.At this level our approach (30) is relatively model free,because we directly observe the RQCs and derived C–C2H3 orderparameters, together with the corresponding R1Z relaxationrates. We next applied Redfield theory to treat the nuclear spindynamics due to molecular motions (30). Experimental spin-lattice relaxation rates depend on thermal fluctuations of theC–2H bond orientation (electric field gradient, EFG tensor) nearω0, and are expressed by (30):

R1Z ¼ 1∕T1Z ¼ 3

4π2χ2Q½J1ðω0Þ þ 4J2ð2ω0Þ� [1]

and

Fig. 1. Site-specific 2H NMR spectroscopy illuminates retinal mobility withinthe binding pocket of rhodopsin. (A) Structure of dark-state rhodopsin(1U19) (5) compared to opsin (1GZM) (19) showing helical movements dueto release of inhibitory cofactor. (B) Chemical structure of 11-cis-retinal ligandwhere correspondence to crystallographic numbering scheme is C5-Me ¼C18, C9-Me ¼ C19, and C13-Me ¼ C20. (C–E) Solid-state 2H NMR spectrafor rhodopsin in dark state with 11-cis-retinal 2H-labeled at C5-, C9-, orC13-methyl (Me) groups, respectively, in POPCmembranes (1∶50molar ratio).Experimental 2H NMR line shapes for random (powder-type) membrane dis-persions were fit to theoretical Pake doublet (40), where Δνpowder

Q ¼ 34 hχQi is

the frequency separation between the two largest peaks. The residual quad-rupolar couplings hχQi directly yield order parameters for the rapidly spin-ning methyl groups (40).

Fig. 2. Solid-state 2H NMR relaxation characterizes molecular dynamics of retinal ligand bound to rhodopsin. (A) Spin-lattice (T1Z ) and quadrupolar-order(T1Q) relaxation times (filled and open symbols, respectively) are plotted against inverse temperature in dark state. Experimental data for retinal 2H-labeled atthe C5-Me (filled diamond, empty diamond), C9-Me (filled circle, open circle), or C13-Me (filled square, open square) groups are indicated. The T1Z and T1Q

times (solid and dashed lines, respectively) were simultaneously fit either with an axial threefold jump model (rate constant k) or rotational diffusion model(coefficients Djj and D⊥, where D⊥ ¼ 0). (B) Theoretical analysis of rotational dynamics of retinylidene methyl groups within binding cavity of rhodopsin. Eulerangles Ωij ≡ ðαij ;βij ;γijÞ transform among coordinate systems fi;jg≡ fP;I;M;D;Lg. The principal axis system (PAS) of EFG tensor of 2H nucleus is denoted by P(z-axis parallel to C–2H bond); I is intermediate frame for methyl rotation (z-axis is C–C2H3 bond; not shown); M is methyl coordinate system (z-axis is averageC–C2H3 bond direction); D is membrane system (z-axis is surface normal n0), and L is the laboratory system (z-axis along external magnetic field B0) (31).

8264 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1014692108 Struts et al.

R1Q ¼ 1∕T1Q ¼ 9

4π2χ2QJ1ðω0Þ: [2]

Here, T1Z is the relaxation time for Zeeman order and T1Q forquadrupolar order, χQ is the quadrupolar coupling constant, andω0 is the nuclear resonance (Larmor) frequency. A simple phy-sical picture is that matching of the power spectrum of the fluc-tuations to the Zeeman energy level gap drives transitionsbetween the nuclear spin angular momentum states (30). Basedon generalized model-free (GMF) analysis (30), the spectral den-sities JmðωÞ depend on the average methyl angle to the magneticfield, together with the mean-square amplitudes and rates ofmotions. The spectral densities are given by:

JmðωÞ ¼ Σr;qjDð2Þ0r ðΩPIÞj2½hjDð2Þ

rq ðΩIMÞj2i− jhDð2Þ

rq ðΩIMÞij2δr0δq0� jð2Þrq ðωÞjDð2ÞqmðΩMLÞj2; [3]

where ω ¼ mω0 andm ¼ 1, 2 (31). In the above formula Dð2ÞðΩijÞindicates the second-rank Wigner rotation matrix, andΩij ≡ ðαij;βij;γijÞ are the Euler angles (30) for relative orienta-tions of coordinate systems fi;jg≡ fP;I;M;D;Lg (see Fig. 2B).The mean-square amplitudes are the quantities in the angularbrackets and indicate a time-ensemble average, while the reducedspectral densities are denoted by jð2Þrq ðωÞ ¼ 2τrq∕ð1þ ω2τ2rqÞ,where τrq are the rotational correlation times.

Now for biomolecules such as membrane proteins (28) or li-pids (31), with the GMF approach the mean-square amplitudescorrespond to the segmental order parameters, and the rates tothe correlation times of the motions (30). E.g., a methyl grouphas ΩPI ¼ ð0;70.5°;0Þ giving an effective coupling constant ofχeffQ ¼ χQD

ð2Þ00 ðΩPIÞ ¼ − 1

3χQ. The order parameter of the methyl

threefold symmetry axis describes the off-axial fluctuations, as de-fined by SC3

¼ 12h3 cos2 βIM − 1i, where βIM is the angle between

the instantaneous methyl orientation and its average value (videsupra). Further theoretical analysis requires a motional model(30, 35) to interpret the fluctuations. For N-fold hops with a rateconstant k about a single axis, the correlation times are (31, 35)1∕τrq → 1∕τr ¼ 4k sin2ðπr∕NÞ giving 1∕τr ¼ 3k for a methylgroup. Alternatively, for continuous diffusion in a potential ofmean torque (36), the rotational correlation times are:

1∕τrq ¼ μrqD⊥∕½hjDð2Þrq ðΩIMÞj2i − jhDð2Þ

rq ðΩIMÞij2δr0δq0þ ðD∥ −D⊥Þr2� [4]

The symbols D∥ and D⊥ indicate the axial and off-axial diffu-sion coefficients for methyl rotation, and the moments μqn andmean-square moduli hjDð2Þ

qn ðΩMDÞj2i depend on both second-and fourth-rank order parameters (36). For a strong-collisionapproximation, the correlation times are just those of a rigid rotor(30): 1∕τrq → 1∕τr ¼ 6D⊥ þ ðD∥ − D⊥Þr2. Last, transition statetheory tells us the correlation times are inversely related toA expð−Ea∕RTÞ, where A is the preexponential factor (k0 or D0),and Ea is the activation energy (barrier) for methyl rotation. For acontinuous rotational diffusion model D∥ ¼ D0∥ expð−Ea∥∕RTÞand D⊥ ¼ D0⊥ expð−Ea⊥∕RTÞ and analogously for threefoldjumps. More detailed description is given in refs. 31, 36.

Retinal Dynamics Indicate Loss of Molecular Strain upon Illumination.Next we applied Eqs. (1–4) to numerically fit the T1Z and T1Qrelaxation times for the retinylidene C5-, C9-, and C13-Megroups of rhodopsin (Fig. 2A). At temperatures well above theT1Z or T1Q minimum, the fast-motion limit applies (30). Correla-tion times in the range of 3–45 ps were obtained at −60 °C and2–12 ps at 30 °C (SI Text), indicating rapidly spinning methylgroups (37). At lower temperatures, closer to the T1Z and T1Q

minima, the correlation times are greater (vide supra). Light-induced changes in the T1Z relaxation times for the dark, Meta I,and Meta II states are summarized in Fig. 3 A–C. For theC5-Me group of the β-ionone ring (Fig. 3A), the T1Z minimumfor the inactive dark state is at ≈ − 120 °C, as noted above. Uponlight absorption, in the preactivated Meta I state a further T1Zminimum may be displaced to lower temperature (≈ − 160 °C),and in the active Meta II state a broad minimum occurs at≈ − 140 °C (Fig. 3A). The short T1Z relaxation times for theC5-Me suggest the β-ionone ring maintains its predominantlytwisted 6-s-cis conformation (32)—it is little affected by transitionsamong the dark, Meta I, and Meta II states.

Perhaps most striking (Fig. 3B), for the polyene C9-Me groupthe T1Z times are longer, with the absence of a T1Z minimumdown to at least −160 °C. In Meta I there is a clear reduction

Fig. 3. Solid-state 2H NMR relaxation uncovers changes in ps–ns dynamics ofretinal during light activation of rhodopsin. Relaxation times (T1Z ) for (A) theC5-Me group, (B) the C9-Me group, and (C) the C13-Me group are plotted vs.inverse temperature in the dark state (——), the Meta I state (– – –), and theMeta II state (∙ ∙ ∙ ∙). Theoretical fits include both an axial threefold jump mod-el (rate constant k) and a rotational diffusion model (axial coefficient Djj withD⊥ ¼ 0). For the C5-Me group it is assumed D⊥ ≠ 0 in Meta I. Fittingparameters include order parameters for methyl threefold axis, activationenergies, and preexponential factors (SI Text).

Struts et al. PNAS ∣ May 17, 2011 ∣ vol. 108 ∣ no. 20 ∣ 8265

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

of the T1Z times at all temperatures, followed by an increase inthe activeMeta II state. Moreover, for the polyene C13-Me group(Fig. 3C) in the dark state the T1Z values fall between those forthe C5-Me and the C9-Me groups, with a T1Z minimum below−160 °C. Despite that three methyl environments are found forthe retinal ligand in the dark state, upon 11-cis to trans isomer-ization ≈ two methyl sites are evident. Such a loss of retinal strainis consistent with two major planes of unsaturation following lightabsorption (1, 34, 38–40). Application of transition state theorythen allows us to connect the T1Z results with the energetic para-meters for retinal bound to rhodopsin (Table 1). For all threestates the Ea barrier for the C5-Me group is largest, while Eafor the C9-Me becomes greater than for the C13-Me in MetaII. It is likely that the β-ionone ring remains in its hydrophobicbinding pocket up to the Meta II state.

DiscussionSolid-state NMR spectroscopy (28) is a powerful adjunct to X-raycrystallography, as it gives knowledge of membrane proteins in anative-like environment where function is preserved (17, 21). Inthe case of rhodopsin, solid-state NMR clearly demonstrates mo-bility of the retinylidene ligand—despite its crucial biological roleas an inverse agonist—even at cryogenic temperatures, as typi-cally used in X-ray crystallography. Deleting any of the methylgroups transforms all-trans-retinal from a full agonist to a weakpartial agonist (34, 41), due to back shifting of the metarhodopsinequilibrium (33, 34, 42). Hence, the retinylidene methyl substitu-ents serve not only as 2HNMR probes of the ligand conformationand dynamics, but they also are directly implicated in the receptoractivation mechanism (1). By combining NMR spectroscopy withX-ray studies (19, 20), one can gather a more complete picture ofthe GPCR dynamics than with either method alone.

Solid-State NMR Relaxation Shows Changes in Retinal Dynamics Dur-ing Light Activation of Rhodopsin. In Fig. 4 we summarize our in-terpretation of the experimental 2H NMR data for retinal boundto rhodopsin together with crystallographic analysis (5, 6, 12, 20).At one end of retinal (Fig. 4A) the β-ionone ring is positionedwithin a hydrophobic cavity, and the C5-Me is in contact withGlu122. At the opposite end, the protonated Schiff base (PSB)interacts electrostatically with its primary Glu113 counterion (43).Our analysis (Fig. 4B) considers the order parameters SC3

for theretinylidene methyl groups, plus the preexponential factors (k0 orD0) and energy barriers (Ea) for rotation (Table 1). A simpleexplanation of why the C9-Me barrier is smaller than occurs intypical quantum mechanical calculations is provided in Fig. 4C.Here intraretinal 1–6 interactions with hydrogen atoms H7 andH11 of the polyene chain are approximated by a threefold

potential. Due to periodic eclipsing interactions, the rotationalpotentials are offset by ≈180° giving a shallow resultant curve.Interactions within the binding pocket are relatively weak, be-cause the C9-Me occupies a slot between Tyr268 and Thr118.By contrast, for the C13-Me group near the PSB end of retinal(Fig. 4D) and the C5-Me of the β-ionone ring (Fig. 4E) themobility is less due to nonbonded 1–7 interactions with hydrogensH10 and H8, respectively. The Ea barrier for the C13-Me groupmay reflect steric interactions with retinal hydrogen H10, and po-tentially Tyr268 as well. Upon light absorption, the barrier for theC9-Me group is increased about twofold; its rotational dynamicsbecome slower (Table 1). Due to 11-cis to trans isomerization, the1–6 interactions of the C13-Me group occur with both polyenehydrogens H11 and H15; hence the C13-Me barrier resemblesthe C9-Me. The most pronounced influences of light-inducedisomerization entail the polyene chain of retinal rather thanthe β-ionone ring (Fig. 4).

Table 1. Structural and dynamical parameters for methyl groups of retinal within rhodopsin binding pocket *

Methyl group Dark Meta I Meta II

C5-MeSCD

† 0.90� 0.05 0.90� 0.05 0.90� 0.05Ea∕ kJ mol−1 11.4� 0.3 10.0� 0.3 ‡ 15.0� 0.4 ‡ 10.3� 0.3k0ðD0Þ∕1011 s−1 § 17� 4 ð34� 8Þ 40� 9 ð80� 20Þ ‡ 100� 20 ð200� 40Þ ‡ 15� 3 ð31� 6Þ

C9-MeSCD

† 0.90� 0.05 0.90� 0.05 0.90� 0.05Ea∕ kJ mol−1 1.4� 0.1 3.7� 0.1 4.3� 0.1k0ðD0Þ∕1011 s−1 § 0.9� 0.2 ð1.8� 0.4Þ 0.6� 0.1 ð1.2� 0.3Þ 1.5� 0.3 ð3.1� 0.6Þ

C13-MeSCD

† 0.90� 0.5 0.90� 0.05 0.90� 0.05Ea∕ kJ mol−1 7.3� 0.2 4.7� 0.2 2.8� 0.1k0ðD0Þ∕1011 s−1 § 11� 2 ð21� 5Þ 1.3� 0.3 ð2.6� 0.6Þ 0.30� 0.01 ð0.6� 0.1Þ

*Errors were calculated from curvature of χ2 hypersurface for individual fitting parameters.†Order parameters were nearly independent of temperature from −30 to −150 °C.‡Assumes two-conformer model (see text).§Preexponential factors assume either a threefold hop model (k0) for methyl rotation or a rotational diffusion model (D0) (see text).

Fig. 4. Illustration of how solid-state 2H NMR relaxation probes local energylandscape of retinal bound to rhodopsin. (A) Solid-state NMR structure ofretinal in dark state (32) (red) compared to NMR structure of Meta I (gray)(40) inserted into binding pocket of rhodopsin (1U19) (5). (B) Rotationaldynamics of methyl groups characterize mobility of retinal within the bind-ing cavity (see text). (C) Schematic methyl torsional potentials in dark stateshow low activation barrier for C9-Me due to antiphase 1–6 interactions withhydrogen atoms H7 and H11. (D, E) Potential surfaces for C13- and C5-Megroups include nonbonded 1–7 interactions with hydrogen atoms H10 andH8, respectively.

8266 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1014692108 Struts et al.

Changes in Local Retinal Structure and Dynamics Initiate CollectiveFluctuations of Transmembrane Helices of Rhodopsin. We next de-scribe a multiscale activation mechanism for how light-inducedchanges in local ps–ns motions of retinal underlie TM helix fluc-tuations in the Meta I–Meta II equilibrium (23, 44). By trappingthe conformational states and probing their local dynamics, weexplore the energy landscape for the receptor activation. Accord-ing to our molecular dynamics simulations (45), helices H1–H4make up the TM core of the receptor (4, 17). The rhodopsin coreis stabilized by two ionic locks involving the PSB and the E(D)RYmotif with their counterions, plus three hydrogen-bonded net-works and several conserved microdomains (27). Light-inducedretinal isomerization destabilizes the inactive receptor conforma-tion (17, 18). Activation entails large-scale movements of helicesH5 and H6 together with H7 (19, 23), exposing transducin (Gt)recognition sites on the C2 and C3 cytoplasmic loops (22).

A current picture (17, 27, 44) is that photon absorption leads tothe major reactions (17, 44): Meta I ⇌ Meta IIa ⇌ Meta IIbþH3Oþ ⇌ Meta IIbHþ, where Meta IIb and Meta IIbHþ are con-formational substates that activate transducin (Gt). In Fig. 5we combine the 2H NMR relaxation data with our FTIR studies(17, 27) and other measurements (18, 19, 23, 40, 45) of μs–msfunctional protein movements. In the Meta I state, the C9-Meacts as a hinge point that prevents roll of the entire ligand. TheC13-Me moves into close proximity to the β4 strand of the secondextracellular loop E2, as suggested by the 2H NMR structure ofretinal in Meta I (40). As a result of 11-cis to trans isomerization,the steric force is transmitted via the C13-Me group to the extra-cellular E2 loop (Fig. 5). A key hydrogen-bonding network is

disrupted about E2 connecting the extracellular ends of helicesH4, H5, and H6 (26). At the opposite end of retinal, the C5-Megroup of the β-ionone ring stays in close proximity to Glu122 ofTM helix H3 and Trp265 of helix H6. The β-ionone ring occupies ahydrophobic pocket that encompasses Met207, Phe208, His211, andPhe212 of H5—it is another key anchor point for retinal (33). Thehigh Ea barrier for the C5-Me in Meta I (Table 1) suggests theisomerization force (strain energy) is transmitted via the β-iononering towards a second hydrogen-bonding network connectinghelices H5 and H3 (40). Expulsion of the β-ionone ring fromits hydrophobic cavity (46) or a large movement of retinal arerendered unlikely (25). Dipolar-assisted rotational resonance13C NMR studies (47) indicate a translation of retinal towardMet207 and Phe208 of helix H5 (18), which is in line with our pro-posals (40).

The resulting destabilization of the retinylidene PSB yieldsdeprotonation and breaking the ionic lock with the complexcounterion (43) due to Glu113 of H3 and Glu181 of E2. In theMeta I–Meta II equilibrium, internal proton transfer occurs fromthe PSB to Glu113, yielding the Meta IIa substate that occurs priorto helix motion (17, 44). Straightening of the distorted chromo-phore gives a potential steric clash of the midportion of the poly-ene chain with Trp265 of H6 (40). Hence, Trp265 moves towardshelix H5, accompanied by a concerted tilting of helix H6 awayfrom the H1–H4 helical core (23) (Fig. 5). Displacement of theβ-ionone ring away from Trp265 of H6 and towards Glu122 of H3yields rearrangement of a hydrogen-bonded network aboutGlu122 and His211 connecting helices H3 and H5. In the Meta IIbsubstate, allosteric constraints on helix H5 are released, allowingit to rotate its cytoplasmic end towards helix H6. The chargeadduct due to Glu134∕Arg135 of the E(D)RY sequence at thecytoplasmic end of helix H3 with Glu247 of helix H6 becomes lessstable than Arg135-Tyr223 and Glu247-Lys231 interactions as inopsin (19) (Fig. 5), breaking a second ionic lock. Reversiblefluctuations of both helices H5 and H6 alter the cytoplasmicloops, transiently exposing a cavity for binding of transducin(20). Last, the Meta IIbHþ substate is formed by protonationof Glu134 giving a pH-dependent thermodynamic driving forcefor activation (17).

Collective Protein Fluctuations Within the Membrane Lipid BilayerUnderlie Activation of Rhodopsin. To connect changes in local ret-inal mobility with large-scale protein fluctuations, we introducethe concept of a dynamically active receptor. Induced fit of retinalto the binding pocket enjoins a highly twisted conformation uponthe ligand (32). Light-induced isomerization entails changes inps–ns scale mobility and methyl group dynamics within theligand-binding cavity. Such changes in local retinal dynamics(Fig. 3) underlie the influences of retinylidene methyl groupson the Meta I–Meta II equilibrium (34) within the pressure fieldof the lipid bilayer (22). In the currency of protein dynamics,large-scale functional motions involve displacements of manyatoms within the protein. Significant amounts of strain energy—what we call frustration (21)—are captured in the dark state,and then dissipated in the structural relaxation. We have pro-posed that rhodopsin is elongated (21, 48) and partially unfoldedin the Meta II state (21, 49), so an ensemble of conformers isexpected. Reversible TM helical fluctuations of H5 and H6together with H7 increase the configurational degrees of freedominherent to the GPCR dynamics. Large-scale protein fluctuationsentail a gain in partial molar volume (50) together with an in-crease in entropy (21, 49), thus providing exposure of recognitionsites for transducin (Gt) (19).

Our interpretation differs from the StandardModel, whereby asingle light-activated receptor (R*) conformation yields the visualresponse. As an alternative, we suggest an ensemble of substatesoccurs due to the entropy gain produced by release of the inhi-bitory retinal lock. An ensemble activation model is put forth,

Fig. 5. Changes in retinal structure and dynamics initiate fluctuations oftransmembrane helices in rhodopsin activation. Extracellular (e) and cyto-plasmic (c) sides of membrane are indicated. Rhodopsin in dark state(1U19) (5) is stabilized by two ionic locks (17) plus hydrogen-bondingnetworks about Glu122 and about the second extracellular loop E2. TheC9-Me group acts as a hinge point for light-induced isomerization of retinal.The polyene chain proximal to the C11═C12 double bond including the C13-Me and C ¼ NHþ– of the PSB rotates towards E2; by contrast, the β-iononering with its C5-Me group remains in its hydrophobic pocket (40). Allostericprotein movements of E2 loop and helices H5 and H6 expose transducin (Gt )recognition sites involving cytoplasmic C2 and C3 loops. Release of the inhi-bitory retinal lock yields a volumetric expansion of the protein within themembrane bilayer.

Struts et al. PNAS ∣ May 17, 2011 ∣ vol. 108 ∣ no. 20 ∣ 8267

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

whereby a manifold of conformational substates leads to rhodop-sin activation on a high-dimensional energy landscape. ForGPCRs like rhodopsin, their specificity of function masks a subtlebalance between stability (enthalpy) and flexibility (conforma-tional entropy) (2, 45, 49). Photon absorption overcomes the bar-riers between the major conformational states, allowing access todifferent tiers of the energy landscape. Release of the inhibitoryconstraints biases the ensemble in favor of substates that exposerecognition elements for transducin (Gt). Variables that reducethe entropy gain—for example, rhodopsin mutants, retinal ana-logues (1, 34, 42), or membrane lipids (21, 22, 49)—yield lossof function through back shifting the Meta I–Meta II equili-brium. Solid-state NMR plays an important role because it isexquisitely sensitive to local order and dynamics. Changes inretinal structure and dynamics thus initiate a transmembraneexpansion of rhodopsin, with an increase in entropy that opensthe doors of visual perception.

MethodsBovine rhodopsin was regenerated with retinal 2H-labeled at the C5-, C9-, orC13-Me groups (32) and was recombined with phospholipids by detergentdialysis (22). Membranes were aligned on planar glass substrates (40) or wereinvestigated as randomly oriented suspensions. For the dark and Meta Istates, rhodopsin was recombinedwith POPC (1∶50molar ratio). Quantitativetrapping of the Meta II state utilized a DOPE/POPC (1∶3) mixture (protein/lipid molar ratio of 1∶75). All 2H NMR spectra were acquired with a quadru-polar-echo pulse sequence (40). Relaxation of Zeeman order hIZi was mea-sured by an inversion-recovery pulse sequence, and 2H longitudinalquadrupolar-order (31) hI2Z − I2∕3i was measured with a broadband Jeener-Broekaert pulse sequence (SI Text). Molecular graphics were created withPyMOL visualization software [http://www.pymol.org/]. SI Text accompaniesthis paper.

ACKNOWLEDGMENTS. We thank T.A. Cross, H. Frauenfelder, K.P. Hofmann,M. Hong, W.L. Hubbell, L.E. Kay, J.W. Lewis, H. Oschkinat, S.O. Smith,R.W. Pastor, and R. Vogel for discussions. Retinal was generously provided byR. Crouch, S. Krane, K. Tanaka, and K. Nakanishi. Financial support of thisresearch by the US National Institutes of Health (NIH) is gratefullyacknowledged.

1. Nakanishi K (2000) Recent bioorganic studies on rhodopsin and visual transduction.Chem Pharm Bull 48:1399–1409.

2. Rosenbaum DM, Rasmussen SGF, Kobilka BK (2009) The structure and function ofG-protein-coupled receptors. Nature 459:356–363.

3. Nygaard R, Frimurer TM, Holst B, Rosenkilde MM, Schwartz TW (2009) Ligand bindingand micro-switches in 7TM receptor structures. Trends Pharmacol Sci 30:249–259.

4. Palczewski K, et al. (2000) Crystal structure of rhodopsin: a G protein-coupled receptor.Science 289:739–745.

5. Okada T, et al. (2004) The retinal conformation and its environment in rhodopsin inlight of a new 2.2 Å crystal structure. J Mol Biol 342:571–583.

6. Li J, Edwards PC, Burghammer M, Villa C, Schertler GFX (2004) Structure of bovinerhodopsin in a trigonal crystal form. J Mol Biol 343:1409–1438.

7. Standfuss J, et al. (2007) Crystal structure of a thermally stable rhodopsin mutant.J Mol Biol 372:1179–1188.

8. Murakami M, Kouyama T (2008) Crystal structure of squid rhodopsin. Nature453:363–367.

9. Rasmussen SGF, et al. (2011) Structure of a nanobody-stabilized active state of theβ2 adrenoceptor. Nature 469:175–189.

10. Warne T, et al. (2011) The structural basis for agonist and partial agonist action on aβ1-adrenergic receptor. Nature 469:241–244.

11. Jaakola V-P, et al. (2008) The 2.6 Angstrom crystal structure of a human A2A adenosinereceptor bound to an antagonist. Science 322:1211–1217.

12. Ruprecht JJ, Mielke T, Vogel R, Villa C, Schertler GFX (2004) Electron crystallographyreveals the structure of metarhodopsin I. EMBO J 23:3609–3620.

13. Nakamichi H, Okada T (2006) Local peptide movement in the photoreactionintermediate of rhodopsin. Proc Natl Acad Sci USA 103:12729–12734.

14. SalomD, et al. (2006) Crystal structure of a photoactivated deprotonated intermediateof rhodopsin. Proc Natl Acad Sci USA 103:16123–16128.

15. Lewis JW, Kliger DS (2000) Absorption spectroscopy in studies of visual pigments: spec-tral and kinetic characterization of intermediates. Method Enzymol 315:164–178.

16. Kukura P, McCamant DW, Yoon S, Wandschneider DB, Mathies RA (2005) Structuralobservation of the primary isomerization in vision with femtosecond-stimulatedRaman. Science 310:1006–1009.

17. Mahalingam M, Martínez-Mayorga K, Brown MF, Vogel R (2008) Two protonationswitches control rhodopsin activation in membranes. Proc Natl Acad Sci USA105:17795–17800.

18. Ahuja S, et al. (2009) Location of the retinal chromophore in the activated state ofrhodopsin. J Biol Chem 284:10190–10201.

19. Park JH, Scheerer P, Hofmann KP, Choe H-W, Ernst OP (2008) Crystal structure of theligand-free G-protein-coupled receptor opsin. Nature 454:183–188.

20. Scheerer P, et al. (2008) Crystal structure of opsin in its G-protein-interacting confor-mation. Nature 455:497–502.

21. Brown MF (1994) Modulation of rhodopsin function by properties of the membranebilayer. Chem Phys Lipids 73:159–180.

22. Botelho AV, Huber T, Sakmar TP, Brown MF (2006) Curvature and hydrophobic forcesdrive oligomerization and modulate activity of rhodopsin in membranes. Biophys J91:4464–4477.

23. Altenbach C, Kusnetzow AK, Ernst OP, Hofmann KP, Hubbell WL (2008) High-resolu-tion distance mapping in rhodopsin reveals the pattern of helix movement due toactivation. Proc Natl Acad Sci USA 105:7439–7444.

24. Creemers AFL, et al. (2002) 1H and 13C MAS NMR evidence for pronounced ligand-protein interactions involving the ionone ring of the retinylidene chromophore inrhodopsin. Proc Natl Acad Sci USA 99:9101–9106.

25. Spooner PJR, et al. (2004) The ring of the rhodopsin chromophore in a hydrophobicactivation switch within the binding pocket. J Mol Biol 343:719–730.

26. Ahuja S, et al. (2009) Helix movement is coupled to displacement of the secondextracellular loop in rhodopsin activation. Nat Struct Mol Biol 16:168–175.

27. Zaitseva E, Brown MF, Vogel R (2010) Sequential rearrangement of interhelicalnetworks upon rhodopsin activation in membranes: the Meta IIa conformationalsubstate. J Am Chem Soc 132:4815–4821.

28. McDermott A (2009) Structure and dynamics of membrane proteins by magic anglespinning solid-state NMR. Ann Rev Biophys 38:385–403.

29. Struts AV, Salgado GFJ, Martínez-Mayorga K, Brown MF (2011) Retinal dynamicsunderlie its switch from inverse agonist to agonist during rhodopsin activation.Nat Struct Mol Biol 18:392–394.

30. Brown MF (1982) Theory of spin-lattice relaxation in lipid bilayers and biologicalmembranes. 2H and 14N quadrupolar relaxation. J Chem Phys 77:1576–1599.

31. Nevzorov AA, Trouard TP, Brown MF (1998) Lipid bilayer dynamics from simultaneousanalysis of orientation and frequency dependence of deuterium spin-lattice and quad-rupolar order relaxation. Phys Rev E 58:2259–2281.

32. Salgado GFJ, et al. (2004) Deuterium NMR structure of retinal in the ground state ofrhodopsin. Biochemistry 43:12819–12828.

33. Vogel R, Siebert F, Lüdeke S, Hirshfeld A, Sheves M (2005) Agonists and partial agonistsof rhodopsin: retinals with ring modifications. Biochemistry 44:11684–11699.

34. Vogel R, et al. (2006) Agonists and partial agonists of rhodopsin: retinal polyenemethylation affects receptor activation. Biochemistry 45:1640–1652.

35. Torchia DA, Szabo A (1982) Spin-lattice relaxation in solids. J Magn Reson 49:107–121.36. Trouard TP, Alam TM, BrownMF (1994) Angular dependence of deuterium spin-lattice

relaxation rates of macroscopically oriented dilaurylphosphatidylcholine in the liquid-crystalline state. J Chem Phys 101:5229–5261.

37. Copié V, et al. (1994) Deuterium solid-state NMR studies of methyl group dynamicsin bacteriorhodopsin and retinal model compounds: evidence for a 6-s-trans chromo-phore in the protein. Biochemistry 33:3280–3286.

38. Kochendoerfer GG, Verdegem PJE, van der Hoef I, Lugtenburg J, Mathies RA (1996)Retinal analog study of the role of steric interactions in the excited state isomerizationdynamics of rhodopsin. Biochemistry 35:16230–16240.

39. Salgado GFJ, et al. (2006) Solid-state 2H NMR structure of retinal in metarhodopsin I.J Am Chem Soc 128:11067–11071.

40. Struts AV, et al. (2007) Structural analysis and dynamics of retinal chromophore in darkand meta I states of rhodopsin from 2H NMR of aligned membranes. J Mol Biol372:50–66.

41. Bartl FJ, et al. (2005) Partial agonism in a G protein-coupled receptor. Role of theretinal ring structure in rhodopsin activation. J Biol Chem 280:34259–34267.

42. Knierim B, Hofmann KP, Gärtner W, Hubbell WL, Ernst OP (2008) Rhodopsin and9-demethyl-retinal analog: effect of a partial agonist on displacement of transmem-brane helix 6 in class A G protein-coupled receptors. J Biol Chem 283:4967–4974.

43. Martínez-Mayorga K, Pitman MC, Grossfield A, Feller SE, Brown MF (2006) Retinalcounterion switch mechanism in vision evaluated by molecular simulations. J AmChem Soc 128:16502–16503.

44. Knierim B, Hofmann KP, Ernst OP, Hubbell WL (2007) Sequence of late molecularevents in the activation of rhodopsin. Proc Natl Acad Sci USA 104:20290–20295.

45. Huber T, Botelho AV, Beyer K, Brown MF (2004) Membrane model for the GPCRprototype rhodopsin: hydrophobic interface and dynamical structure. Biophys J86:2078–2100.

46. Borhan B, Souto ML, Imai H, Shichida Y, Nakanishi K (2000) Movement of retinal alongthe visual transduction path. Science 288:2209–2212.

47. Patel AB, et al. (2004) Coupling of retinal isomerization to the activation of rhodopsin.Proc Natl Acad Sci USA 101:10048–10053.

48. Salamon Z, Brown MF, Tollin G (1999) Plasmon resonance spectroscopy: probingmolecular interactions within membranes. Trends Biochem Sci 24:213–219.

49. Brown MF (1997) Influence of non-lamellar forming lipids on rhodopsin. Curr TopMembr 44:285–356.

50. Attwood PV, Gutfreund H (1980) The application of pressure relaxation to the study ofthe equilibrium between metarhodopsin I and II from bovine retinas. FEBS Lett119:323–326.

8268 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1014692108 Struts et al.