Available online at ~DIRI=CT SCIENCE. Structure of Bovine Rhodopsin in a Trigonal Crystal Form

Transcription

1 doi:l /j.jmb J. MoL BioL (2004) 343, jib Available online at SCIENCE ~DIRI=CT ELSEVIER Structure of Bovine Rhodopsin in a Trigonal Crystal Form Jade Li 1., Patricia C. Edwards 1, Manfred Burghammer 2, Claudio Villa 1 and Gebhard F. X. Schertler 1. 1Medical Research Council Laboratory of Molecular Biology Hills Road Cambridge CB2 2QH, UK 2European Synchrotron Radiation Facility, BP Grenoble, France *Corresponding authors We have determined the structure of bovine rhodopsin at 2.65 A resolution using untwinned native crystals in the space group P31, by molecular replacement from the 2.8 A model (1F88) solved in space group P41. The new structure reveals mechanistically important details unresolved previously, which are considered in the membrane context by docking the structure into a cryo-electron microscopy map of 2D crystals. Kinks in the transmembrane helices facilitate inter-helical polar interactions. Ordered water molecules extend the hydrogen bonding networks, linking Trp265 in the retinal binding pocket to the NPxxY motif near the cytoplasmic boundary, and the Glu113 counterion for the protonated Schiff base to the extracellular surface. Glu113 forms a complex with a water molecule hydrogen bonded between its main chain and sidechain oxygen atoms. This can be expected to stabilise the salt-bridge with the protonated Schiff base linking the 11-cis-retinal to Lys296. The cytoplasmic ends of helices H5 and H6 have been extended by one turn. The G-protein interaction sites mapped to the cytoplasmic ends of H5 and H6 and a spiral extension of H5 are elevated above the bilayer. There is a surface cavity next to the conserved Glu134-Arg135 ion pair. The cytoplasmic loops have the highest temperature factors in the structure, indicative of their flexibility when not interacting with G protein or regulatory proteins. An ordered detergent molecule is seen wrapped around the kink in H6, stabilising the structure around the potential hinge in H6. These findings provide further explanation for the stability of the dark state structure. They support a mechanism for the activation, initiated by photo-isomerisation of the chromophore to its all-trans form, that involves pivoting movements of kinked helices, which, while maintaining hydrophobic contacts in the membrane interior, can be coupled to amplified translation of the helix ends near the membrane surfaces Elsevier Ltd. All rights reserved. Keywords: G protein-coupled receptor; G protein activation; ligand binding pocket; membrane protein structure; visual pigments Abbreviations used: C8E4, n-octyltetraoxyethylene; C1, cytoplasmic loop connecting helices 1 and 2; C2, cytoplasmic loop connecting helices 3 and 4; C3, cytoplasmic loop connecting helices 5 and 6; cgmp, 3~,5~-cyclic guanosine monophosphate; cryo-em, electron cryomicroscopy; El, extracellular loop connecting helices 2 and 3; E2, extracellular loop connecting helices 4 and 5; E3, extracellular loop connecting helices 6 and 7; EM, electron microscopy; EMTS, ethyl mercurithiosalicylate; FTIR, Fourier transform infrared spectroscopy; GPCR, G protein-coupled receptor; G-protein, heterotrimeric guanine nucleotide binding protein; Gt, transducin; Gtc~, transducin ~-subunit; Gt~ , C-terminal peptide from transducin ~z-subunit; LDAO, N,N-dimethyldodecylamine-N-oxide; MI and MII, metarhodopsin I and II; NCS, non-crystallographic symmetry; PDE, phosphodiesterase. addresses of the corresponding authors: jl@mrc-lmb.cam.ac.uk; gfx@mrc-lmb.cam.ac.uk /$ - see front matter 2004 Elsevier Ltd. All rights reserved.

2 1410 Rhodopsin Structure in Trigonal Crystals Introduction G protein-coupled receptors (GPCRs) constitute the largest superfamily of transmembrane signalling proteins in the eukaryotic kingdom. They share a common core structure comprising seven transmembrane helices, as implied by a conserved pattern of amino acids at key positions along each of the seven hydrophobic sequence segments. 1 These proteins capture external signals, including light, odorants, hormones and neurotransmitters, and transmit the stimuli across the plasma membrane by selectively activating heterotrimeric guanine nucleotide binding proteins (G proteins) on the cytoplasmic surface. Activation of G protein amplifies the signal and leads to activation of effector enzymes or molecules, which elicit the cell's signalling response indirectly by altering second messenger concentrations or by direct control of ion channel activities. Thus, the GPCRs play a central role in regulating many physiological processes. They are consequently major targets for drug design. Rhodopsin, the photoreceptor protein in retina rod cells, is a prototypical GPCR. It contains a lightsensitive ligand, the 11-cis-retinal chromophore, bound covalently to the apoprotein opsin via a protonated Schiff base with Lys296 on helix 7. Absorption of a photon at a wavelength of about nm isomerises the retinal to all-trans within picoseconds and with a quantum efficiency of This event initiates the formation of a series of photointermediates 3 with conformational changes in the opsin. 4 The biochemically active conformation R* is attained within one millisecond at physiological temperature. It is spectroscopically identified with the metarhodopsin II (MII) intermediate, which has an absorption maximum at 380 nm due to deprotonation of the Schiff base of all-trans-retinal. R binds and activates the heterotrimeric G-protein transducin (Gt) at the cytoplasmic surface, to catalyse GDP/GTP exchange on the Gt ~-subunit and dissociation of the Gt heterotrimer. The GTP-bound ~-subunit then activates the effector enzyme, phosphodiesterase (PDE). Hydrolysis of cyclic guanine monophosphate (cgmp), the second messenger, by PDE leads to closure of the cgmp-gated cation channel in the plasma membrane, causing hyperpolarisation and initiation of nerve impulse in the retina. 5 In contrast to the efficient and rapid photoisomerisation and activation, the rate of thermal isomerisation is very low, about one per 400 years per rhodopsin molecule at 37 C. 6 The low thermal noise coupled with large amplification via transducin activation underpins the sensitivity of single photon detection operating in dim light vision. An atomic structure of rhodopsin provides a model for all the visual pigments as well as the majority of the GPCRs. Crystallisation in two-dimensional (2D) lattices with endogenous membrane lipids, followed by electron cryo-microscopy (cryo-em) 7-11 at resolution limits up to 5 A in the membrane plane and 13.5 A normal to it, have produced images of bovine, frog and squid rhodopsins in a membranelike environment, which are in essence similar. Using packing constraints derived by extensive sequence comparisons across the GPCR superfamily, the density peaks in cryo-em maps were assigned to hydrophobic sequences, leading eventually to a C a model for the seven transmembrane helices in the rhodopsin-like GPCR fairly. 1"12'13 Remarkably, this model came within 2.3 A rms deviation of C a coordinates determined by X-ray crystallography subsequently 14"15 and it provided a framework for mutagenesis and biophysical studies in the absence of an atomic structure. To determine the atomic structure of rhodopsin, we have obtained untwinned three-dimensional crystals of bovine rhodopsin in the trigonal space group P31 that diffract X-rays to 2.65 A resolution, and prepared an ethylmercury derivative that showed anomalous scattering. 16 However, significant non-isomorphism prevented experimental phasing. While this work was in progress, a structure of bovine rhodopsin in a tetragonal P41 space group at 2.8 A resolution was reported. 14'17 Using these coordinates as search model, we have determined the structure of bovine rhodopsin to 2.65 A resolution in the P31 lattice by molecular replacement followed by multiple crystal averaging among the non-isomorphous data sets. Here, we describe the refined structure of bovine rhodopsin in the trigonal crystal form, and compare it with previous structural reports of the tetragonal crystal form. 14"18"19 In addit'ion, we relate the crystal structures to the membrane environment by docking them into a cryo-em map 1 of 2D crystals. Structural implications for the stability of the "dark" state and the potential for light-induced conformational change, leading to transducin activation on the cytoplasmic surface, are discussed. Results and Discussion Structure determination Bovine rhodopsin was crystallised in the trigonal space group P31 with two protein molecules per asymmetric unit, from a detergent mixture of C8E4 and LDAO as described in the accompanying paper. 16 Data sets were obtained from native and mercury-derivatised crystals prepared and frozen under dim red light; however, they showed significant non-isomorphism with one another. Data collection and refinement statistics are given in Table 1. Using the protein part only of the rhodopsin coordinates solved in a tetragonal P41 space group (PDB code 1F88) 14'17 at 2.8 A resolution as search model, molecular replacement solutions were calculated in parallel for three sets of amplitudes (Native-2, EMTS-1 and EMTS-2). The resulting 2Fo--F~ maps showed no density in three cytoplasmic regions of 1F88 coordinates, and the

3 Rhodopsin Structure in Trigonal Crystals 1411 Table 1. Data collection and final refinement statistics: space group P31 A. Data collection Native-2 a EMTS-1 EMTS-2 Native-3 X-ray source o ESRF ID13 ESRF ID13 ESRF ID14-4 ESRF ID13 Wavelength (A) Unit cell a, c (A) 104.2, , , , 76.6 No. of crystals Mosaicity (deg.) Twin fraction Resolution (A) 3.2 ( ) 3.6 ( ) 3.4 ( ) 2.65 ( ) amerge b (0.322) (0.464) (0.426) (0.434) I/o 7.5 (2.3) 7.6 (2.2) 17.2 (4.7) 11.0 (1.4) Unique reflections Completeness (%) (0.894) (0.180) (0.172) (0.861) Multiplicity 3.1 (1.6) 5.3 (2.9) 11.3 (9.5) 4.4 (1.6) Wilson B-factor (~2) B. Refinement against Native-3 Reflections in working set Reflections in test set Resolution range (A) Rc~ys~ R free Luzzati coordinate error ( ~) SigmaA coordinate error ( A) tins-deviation from ideal geometry Bond lengths (A) Bond angles (deg.) Dihedral angles (deg.) Improper rotations (deg.) Ramachandran plot Residues in most favoured regions (%) Residues in additional allowed regions (%) Residues in generously allowed regions (%) Residues in disallowed regions (%) Average B-factor (~2) B rmsd for bonded main-chain atoms ~2) B rmsd for bonded side-chain atoms IA 2) B rmsd for angle main-chain atoms ~A 2) B rmsd for angle side-chain atoms (A 2) (2165) Protein chains (130) Protein residues ( ) Palmitoyl chains (0.312) N-linked carbohydrate chains (0.315) Carbohydrate residues A LDAO ,~ C8E4 12 Phospholipid Water Ions Values in parentheses apply to the outer resolution shell. a Names of data sets conform to Table I of the accompanying paper. 16 Native-1 is not referred to here, because it was not used in the structure determination. b Rmerge = Y'~hkt Y'~i ]Ii --(I)]/~hkl Y'~i Ii, where Ii is the intensity of an individual reflection and <I> is the mean intensity obtained from multiple observations of symmetry-related reflections taken from one or more crystals. c Rcryst y~hkl(]fo(hkl) _ Fc(hkl)D/y~hkl(]Fo(hkl)])" calculated for reflections in the working set. a Rfre e y~hk1(ifo(hki) _Fc(hkl)[)/y~hkl(iFo(hkl)b, calculated for a random 5% of the data placed in the test set. corresponding Fo-Fc maps showed negative peaks in the same regions. These regions were: the C2 loop connecting transmembrane helices H3 and H4; the C3 loop connecting H5 and H6; and the C-terminal tail of residues By contrast, density was present in the 2Fo-Fc maps over positions of the retinal chromophore, even though, as a control, we had omitted the retinal from the 1F88 coordinates used for molecular replacement. No negative peaks were seen in the retinal-binding pocket. Therefore our amplitude data disagreed with the starting model in the cytoplasmic regions. Exploiting the large non-isomorphism, we carried out crosscrystal averaging among these data sets. The averaged map for Native-2 showed densities for the C2 and C3 loops at positions different from the starting model, enabling them to be rebuilt and the cytoplasmic end of helix H6 to be extended from residue 247 in the 1F88 coordinates to residue 244. The middle section of the C3 loop (residues in molecule A and in molecule B) and the C terminus still had no interpretable densities, so they were left out of the model. The Schiff base linked 11-cis-retinal chromophore was built from the small molecule crystal struc- tures 2 '21 by adjusting torsion angles along the polyene according to the averaged density. A new native data set (Native-3) was then collected from four untwinned crystals to 2.65 resolution, and used to refine the atomic model, using CNS 22 under 2-fold non-crystallographic symmetry restraints with simulated annealing and restrained individual B-factors. Bound water molecules were located using the water_pick task of CNS 22 and confirmed in maps calculated using EDEN. 23 Tightly bound phospholipids and

4 1412 Rhodopsin Structure in Trigonal Crystals detergent molecules were added. Including the bound water and amphiphile molecules in the refinement further improved the map, so that the cytoplasmic ends of both H5 and H6 were extended and the C3 loop was completed. The most significant differences between our structure and all the published coordinates from the P41 crystal are found in this region. Quality of the structure The final crystallographic R-factor is with a free-r-factor of 0.235, and there is close agreement with ideal geometry (Table 1B). The Ramachandran plot contains 90.6% of the protein residues in the most favoured regions, compared with values of % for the published rhodopsin structures in the tetragonal crystal form. The ~A-weighted 24 2Fo-F~ map showed an average real-space correlation of 0.95 for all residues including the nonprotein residues. These statistics demonstrate that the structure was well determined, and bias toward the starting model was minimal. The refined model (Table 1B) comprises amino acid residues in both molecules, compared with the complete sequence of 348 residues. Post-translational modifications have been defined on both molecules, including N-acetylation of Met1, N-glycosylation of Asn2 and Asn15, palmitoylation of Cys322 and Cys323, and the Schiff base linkage of 11-cis-retinal to Lys296. In addition, there are 20 ordered water molecules bound to each protein chain and a number of tightly bound lipid and detergent molecules. The covalent structure is shown as a ribbon diagram in Figure l(a). The residue ranges of the seven transmembrane helices are: H1, 34-64; H2, ; H3, ; H4, ; H5, ; H6, and H7, Residues in the peripheral helix H8 are , and those making up the B-strands are: [31, 4-6; B2, 9-11; ~3, and f~4, Portions of the transmembrane helices, H4 between residues 170 and 173 and H7 between residues 296 and 299, are in the 310 conformation. Comparison with the tetragonal crystal form Packing differences Both the P3l and P41 crystal forms contain two rhodopsin molecules per asymmetric unit. In the P41 form, they are related by a non-crystallographic rotation of about an axis nearest to a and are oriented obliquely from the P41 axis. They make contacts in the middle of helix 1 but mainly over the hydrophilic extra-membrane segments. In particular, the C terminus of molecule A from the cytoplasmic domain forms a contact with the extracellular domain of molecule A in another asymmetric unit. In the P31 form, the non-crystallographic symmetry (NCS) is a nearly perfect 2-fold about an axis parallel with the 31 axis and passing through (x , y= 0.167). Both helix bundles are tilted by 108 from c and make antiparallel contacts along the entire length of helix 5. The antiparallel dimers are stacked according to the 31 screw symmetry, forming two protein columns in the unit cell, which are centred on two of the three 3-fold positions of the trigonal lattice, and leaving a solvent channel centred on the remaining 3-fold position. Between layers of stacked dimers the contacts involve mainly the sides of helices I and 4. The other helices participate in crystal contacts to a lesser extent or indirectly via interleaved palmitoylate chains or the C8E4 molecule (see subsection titled Bound lipid and detergent molecules). Thus, uncommonly among crystals of membrane proteins, the inter-molecular contacts in the P31 form are predominantly hydrophobic. The hydrophilic extra-membrane segments point into the solvent channel and are not responsible for ordered crystal contacts (Figure l(b)). Consequently, the P31 crystal packing has retained an amphipathic molecular environment characteristic of the native membrane, permitting the hydrophilic segments to display conformations present in a membrane environment unhindered by packing interactions. Coordinate differences between NCS copies and between crystal forms The 1L9H coordinates 19 at 2.6.~ resolution are used to represent the P41 form. Here, the rms C = distance between NCS copies is A for 301 matched residues, and the rms B-factor difference is ~2 with molecule A being the more ordered copy. In the P31 form with closer NCS, the two molecules show rms differences of only 0.13 A and 2.216/~2 for 326 matched residues. Therefore, molecules A from the 1L9H coordinates and the present structure are used to compare the crystal forms. Figure 2(a) shows the coordinate difference per residue as a function of sequence position, between NCS copies in each crystal form and between the molecules A of different crystal forms. Both crystal forms showed significant main chain NCS differences, above 1 A, in the C2 and C3 loops and the C-terminal tail following helix 8. These regions also have the highest B-factors and disorder in each crystal form (Figure 2(b)). Part of the C3 loop is unresolved in both molecules of the P41 form. The C-terminal tail in the P31 form is located in the solvent channel and disordered, while in the P41 form it is partially resolved only in molecule A, where it is stabilised through contacts with the extracellular domain of a neighbouring molecule. Finding significant NCS differences in the same surface segments indicates an intrinsic conformational variability or flexibility of those segments, and the high B-factors of the segments support this. These segments are known to contain residues that interact with transducin (Gt) and the regulatory 2~31 proteins rhodopsin kinase and arrestin; their

5 Rhodopsin Structure in Trigonal Crystals 1413 (a) (b) Surf~ resic~ crystal ~ither c lia ord~ and d~ mol~ N-linked ~ ~ oligosaccharides Figure 1. The bovine rhodopsin molecule in the transmembrane orientation with cytoplasmic surface at the top and extracellular surface at the bottom. (a) Ribbon representation. The seven transmembrane and one peripheral helices are coloured in the rainbow order: H1 (residues 34-64), dark blue; H2 (71-100), light blue; H3 ( ), blue-green; H4 ( ), yellow-green; H5 ( ), yellow; H6 ( ), orange; H7 ( ), red; H8 ( ), magenta. The 13- strands are coloured in cyan. They are arranged in the order of 131 (4-6), 132 (9-11), f~3 ( ), and J~4 ( ), from the extracellular surface into the retinal-binding pocket. The transmembrane helices are kinked around the 11-cis-retinal to form the binding pocket. Also shown are the N-linked oligosaccharides attached to Asn2 (near 131) and Asn15, and the palmitoylate chains attached to Cys322 and Cys323 after H8. This colour scheme for the secondary structures is used throughout. (b) Space filling representation. Residues coloured in yellow form the inter-molecular contacts in the P31 crystal form, directly or via ordered lipid and detergent molecules. They are all located in the transmembrane helices. The oligosaccharides are coloured pink, and the C = trace is shown in grey for comparison with (a). Figures 1, 2(b), 4, 5, 6, 7(b) and 9(a) were prepared using MOLSCRIPT; 146 Figures 1, 2(b) and 9(a) also using Raster3D. 147 flexibility in the absence of the binding partners may have a functional role. If main-chain atoms in the molecules A of the two crystal forms are superimposed by aligning the more ordered regions selected by a B-factor cut-off of 90,~ 2, then the rms coordinate difference is 0.36 A, which is slightly less than the mean coordinate error for all atoms in either structure. Therefore, the two structures are in good agreement in the ordered core of the molecule, which comprised residues 1-137, and , and excluded only the neighbourhood of the C2 and C3 loops, and the C-terminal tail. The backbone differences are concentrated in the same regions shown by the NCS comparisons to be intrinsically more flexible. Figure 2(a) shows that, for the C2 loop, the difference between crystal forms is on the same scale as the NCS difference within the P41 form, so they are accountable by local flexibility and packing contacts. However, around the C3 loop the differences between the two crystal forms are much larger than the NCS differences in either form. Therefore, they indicate statistically significant disagreement. Location of major coordinate differences between crystal forms Figure 2(b) shows the C ~ trace of the molecules A from the two crystal forms side-by-side, coloured

6 1414 Rhodopsin Structure in Trigonal Crystals (a) 20 A "" 16 o= C ==.=:_= 12 "0 e" m 8 "0 L JI, I P3(1) vs P4(I), main chain P3(I) vs P4(1), side chain P3(I) ncs, main chain P4(1) ncs, main chain --N c2t~ Sequence position (b) C-terminal fragment P ~O /O ~ 1 IU I.SU 13U B-factor scale (~) Figure 2. (legend opposite)

7 Rhodopsin Structure in Trigonal Crystals 1415 by the mean B-factor in each residue. The colour scheme shows that the new structure from the P31 crystal form has the lower B-factors everywhere except in the C3 loop, which is absent in the other structure, and the major differences lie in the regions of higher B-factors. The C2 loop has a similar "L-shape"14 in both structures but with a different orientation relative to the helix bundle. The difference may be described as a hinge movement about the junctions with the cytoplasmic ends of H3 and H4. The C3 loop and the cytoplasmic ends of H5 and H6 show the most striking difference between the two structures. In molecule A of the P41 form, H5 terminates at residue 226, following which the polypeptide chain re-enters the bilayer ~nd appears disordered between residues 236 and 240, while H6 terminates at residue 244. In the P31 form, H5 extends to residue 230, then the C a trace continues in a helix-like spiral path away from the membrane to residue 236, where it changes direction to run parallel with the membrane and joins the cytoplasmic end of H6 at residue 241. The cytoplasmic tip of H6 is slightly bent at residue 244 towards H5. The C-terminal tail between residues in the P41 structure adopts an extended conformation supported by contacts with a neighbouring molecule. In the P31 structure this segment is disordered, except for a di-peptide suggested by the density in contact with the cytoplasmic loop 1 (loop C1). The disorder can be expected, since the fragment, without disulphide bridges, is too small to form an independently folded domain, and it is located in a solvent channel in this crystal form. The structure of the cytoplasmic domain of the P31 form is discussed at greater length in the context of explaining the surface accessibility and mobility observed in site-specific spin labelling, chemical 32 35" labelling and cross-linking studies, - as well as the relationship to G protein activation, in the section Cytoplasmic domain. Other differences between the crystal forms The P31 structure shows mechanistically important differences from the 1L9H coordinates in the conformation of the covalently linked 11-cis-retinal chromophore (see subsection titled Conformation and environment of the chromophore). All the internal water molecules have now been located, thus leaving no cavity TM in the protein. The eight newly identified water molecules include one in the vicinity of Trp265, and they augment the inter-helical H-bonding network (see subsection titled Transmembrane domain). The N-linked carbohydrate chains do not form ordered crystal contacts in the P31 crystals, unlike in the P41 crystals (see subsection titled Extracellular domain). Placing the X-ray structure in the membrane context Before coordinates from the tetragonal 3D crystals became available, we were able to determine the orientation of rhodopsin molecules in the P31 crystals by rotation search at 5 A resolution, 16 using as search model the density for a single molecule masked from a cryo-em map of 2D crystals 1 (see Methods). Conversely, by docking the refined X-ray coordinates into this cryo-em map, 1 we placed the atomic structure in the membrane context of the 2D crystal. The height of rhodopsin relative to the bilayer was therefore ascertained, since the p22121 symmetry of those 2D crystals requires symmetrical insertion of molecules from both faces of the bilayer and therefore the plane defined by c = 0 is equivalent to the midplane of the bilayer. Figure 3(a) shows the mean residue height in the bilayer as a function of sequence position, overlaid by the mean residue B-factors of the refined P31 structure. The sequence ranges of the ~ and secondary structure elements are shown as vertical bars. Small-angle X-ray and neutron scattering have shown that phosphate peaks in the retinal disc membranes are separated by 40 A across the bilayer, 4 '41 and lipid head groups in model membranes occupy a layer about a 10 A thick centred on the phosphate peaks. 42 Therefore in Figure 3 the hydrophobic zone extends from approximately - 15 A to + 15 A, and the lipid head group layers extend from there to approximately -25 A and + 25 A. On this scale, the 33 N-terminal residues preceding H1 fall entirely outside the hydrophobic zone. They form the extracellular domain. However only the residues 1-5 and 13-21, which carry the N-linked carbohydrate chains on Asn2 and Asn15, reach into the intra-discal space. On the extracellular face all transmembrane helices terminate in the lipid head group layer, except H4, which terminates within the hydrophobic zone. The strand fk3 lies at the interface between hydrophobic zone and head group layer, while ~4 is contained within the hydrophobic zone. On the cytoplasmic face, helices H1, H2, H4 and H7 Figure 2. Comparison between the P31 and P41 crystal forms. (a) Coordinate difference as a function of sequence position. The main chain and side-chain differences between the molecules A from different crystal forms are superimposed on the main chain difference between NCS copies within each crystal form. Significant differences exist only around the C2 loop (residues ) connecting H3 and H4 and the C3 loop (residues ) connecting H5 and H6, and in the C-terminal tail following HS. (b) C ~ trace of the molecules A from the two crystal forms coloured by their mean residue B-factors. The two structures differ in the orientation of the C2 loop, the lengths of H5 and H6, the conformation of the C3 loop and the C-terminal fragment. Regions of maximal structural differences also show the highest B-factors.

8 141 6 Rhodopsin Structure in Trigonal Crystals (a) 180 Cytoplasmic O,-.= fie d~ A 20 '~.0_ lo "~ 0 13_ 0 ~.Q -I0 6 t - I Residue Number I ml~ Helices Y~II: Strands - - B-factor --~ Trans-bilayer Posiffon I ExtraceUular (b) 200 &-. % (D o9 0 rv '.'. O. m ":: "'.i....o o"...- ~.. ~.,...,,.... # "'~ l" " :..;' ~.~.'. t.: : o,0 ;:. ::.i Mean residue distance from mid-plane of bilayer (A) Figure 3. Relation of the P31 crystal structure to the membrane environment (a) Mean residue height relative to the bilayer derived by docking the X-ray coordinates in the cryo-em map of 2D crystals, with mean residue B-factors superimposed The mid-plane of the bilayer is defined by the c = 0 plane of the p D lattice. The hydrophobic zone is estimated to lie in a 30 A band about this plane, and the lipid headgroups in two 10 A wide layers outside that. The secondary structure assignments are represented by vertical bars, with notches in the bars for H4 and H7 indicating the location of 310 segments. The curve of residue heights follows a straight path over sections of H4 and H6 that are perpendicular to the bilayer, but shows a zigzag pattern for the tiffed sections. Surperimposed on this curve is the mean B-factor per residue as a function of the sequence. (b) Mean residue B-factor as a function of transmembrane position. The B-factors are well clustered between the lipid phosphate positions at 20 A. They increase from the membrane interior towards the surface, but the lowest B-factors are found just to the extracellular side of the bilayer mid-plane, at the depth where the retinal is bound. The sharp rise of B-factors on the cytoplasmic surface is due to conformational variability in several regions.

9 Rhodopsin Structure in Trigonal Crystals 1417 Table 2. Tilt and kinks in the transmernbrane helices Tilt from membrane Azimuthal orientation c Kink angle Helix Straight segments a Residue range a normal b (deg.) (deg.) H1 N-segment C-segment H2 N-segment C-segment H3 N-segment Midsegment C-segment H4 N-segment C-segment H5 N-segment C-segment H6 N-segment C-segment H7 N-segment Midsegment C-segment H a Residue range of the straight segments were determined by fitting cylinders of 2.8 A radius over the C a positions. All segments have a-helical conformation, except for the C-segment of helix 4 and residues within the mid-segment of helix 7, which have the 310 conformation. b The membrane normal was taken from the c-axis of the 2D crystal in the two-sided plane group p Positive tilt angle means the helix runs from the extracellular side towards the cytoplasmic side as the sequence number increases. c Angle from the a-axis of the p22121 lattice to the projection of helix segments in the plane of the 2D crystals, by a right-handed rotation about the membrane nornial. terminate in the head group layer, whereas H3, H5 and H6 reach into the cytoplasmic space. Residues 66, , and from the three cytoplasmic loops are well exposed above the lipid head group layer (Figure 3(a)) and available for interactions with cytoplasmic proteins. 3 '38'39 The peripheral helix H8 is partitioned between the head group layer and the cytoplasm and is slightly inclined with its N terminus more exposed to the cytoplasm than its C terminus. The polypeptide following H8 turns to the cytoplasm, but the palmitoylate chains on Cys322 and Cys323 are inserted into the hydrophobic zone. Transmembrane helices in rhodopsin are kinked (see subsection Transmembrane domain), and only the N-terminal segments of H4 and H6 are oriented roughly normal to the membrane (Table 2). Figure 3(b) shows that the mean residue B-factors are strongly correlated with the depth in the bilayer, irrespective of the helix tilt angles. The lowest B-factors, on average, are found not at the midplane of the bilayer midplane but in a 4 A band to the extracellular side of it. Therefore the dark state rhodopsin structure is most ordered at that depth in the membrane, which coincides with where the ionone ring of the 11-cis-retinal chromophore is located. Beyond the phosphate head group positions (+20 A), the B-factors rise distinctly more rapidly on the cytoplasmic surface than in the extracellular domain, which is due to conformational variability in several regions. Extracellular domain The extraceuular domain includes the short ~1-B2 hairpin, which together with the ~3-B4 hairpin in the transmembrane domain (Figure l(a)) has been described as forming a "plug" for the chromophorebinding pocket. 14 The two B-hairpins are crosslinked by water-mediated H-bonds between mainchain atoms of Tyrl0 in 132 and Pro180 in ~3 (Figure 4). Three water molecules (3, 5 and 13) are responsible for this, of which one (13) was identified in the P41 structure. 19 These water molecules form additional H-bonds to Gly182 and Gln184 in the ~3-f~4 hairpin and to Tyr192 near 134, further linking the hairpin to the transmembrane domain and securing the plug. Between Pro12 at the end of 132 and Pro23, the polypeptide forms an outward "horn" (Figures l(a) and 4), which is rigid with a turn (residues 15-18) at the apex and several sidechain to main-chain H-bonds. The horn was suggested to act as a spacer between opposite faces of the retinal disc, 43 but the N-linked oligosaccharides on Asn2 and Asn15 may also contribute in this role (Figure l(a)). In our structure there is density for only three residues in each of the oligosaccharide chains on Asn2 and Asn15:44 GlcNAc-(131,4)-GlcNAc-(~l,4)-mannose. The distal carbohydrate residues cannot be resolved, probably because they adopt heterogeneous orientations within the solvent channel. The orientations of the oligosaccharide chains relative to the protein differ from those in the P41 structure, in which they were found to form intermolecular contacts. The remainder of this domain contains two turns (residues 22-25, 28-32) in the lipid head group layer adjacent to the transmembrane domain. Many H-bonds and van der Waals interactions exist between the extracellular domain and the base of the transmembrane domain, lending credence to the view that the N-terminal domain may fold first

10 1418 Rhodopsin Structure in Trigonal Crystals 1-/6 #4 H3 / ;. YI02 G280 F9 GIcNAc GIcNAc Figure 4. Interface between the extracellular and transmembrane domains. Loops in the two domains are coloured blue and yellow, respectively. Note the water-mediated H-bonds linking 132 and 133, the stacking of Pro12 with Pro285 and Pro27 with Tyrl02, and the environment of Pro23. Amino acids are designated by the one-letter code for clarity in Figures and provide a template to assist the assembly of the helical domain. 45 Figure 4 shows that Asn2 is in contact with Gly280 and Asp282 in the E3 (H6-H7) loop, and Pro12 is stacked with Pro285 in the same loop. Pro27 is stacked with Tyrl02 in the E1 (H2-H3) loop, which is completely conserved among vertebrate opsins. Pro23 and Gin28 form mainchain H-bonds with Glyl01 and Tyrl02. Pro23, the most conserved (99%) residue in the extracellular domain of vertebrate opsins, is in a buried turn and in contact with Phe9 and Va111 in 132, and Gln28, as well as Tyrl02 and Phel03 in the E1 loop and Pro180 and Gln184 in the E2 loop. Mutations of Pro23 can prevent correct folding of both domains. Pro23His, the most frequent mutation in North American patients with retinitis pigmentosa, has been shown to cause aggregation of misfolded rhodopsin, which is targeted for degradation by the ubiquitin proteosome pathway in cultured cells but also impairs the degradation pathway, and that impairment may underlie the neurodegenerative phenotype. 46 Sequences of visual pigments are highly conserved and, relative to vertebrate opsins, the invertebrate opsins show few insertions (GPCRDB data baser). The longest insertion, of 13 residues, is located in the turn of the hairpin; the others, of two, one and five residues each, are found after residue 332. Within the 131 and 132 strands visual pigments share around 30% sequence identity. At this level of similarity the invertebrate opsins can be expected 47 to contain a similar B-hairpin and accommodate the insertion on the extracellular surface (Figure 3(a)). Transmembrane domain Fitting cylinders of 2.8 A radius to the C ~ trace in 0 48 divided the transmembrane helices into two or three straight segments each, with kink angles of 8 to 36 in-between (Table 2). Having kinks allows the tilted transmembrane helices to pack more intimately than otherwise possible and to enfold the chromophore in a tight binding pocket. The kinks in H1, H4, H5, H6 and H7 are caused by proline t

11 Rhodopsin Structure in Trigonal Crystals 1419 residues at positions 53, , 215, 267 and 302, respectively. Except for Pro53 these proline residues are among the most conserved residues in rhodopsin-like GPCRs, 1"49 so this set of kinks is expected at equivalent sequence positions throughout this family, although the kink angles may vary in a sequence-dependent manner. On the other hand, H2 is distorted at the consecutive glycine residues Gly89-Gly90, which allows the side-chain of Thr92 to H-bond to the main chain at residue 88. In H3 the kinks occur at Glu113, the counterion 5 -s2 for the protonated Schiff base in the dark state, and next to the ionone ring of the chromophore. In H7 a second kink occurs near Lys296, which is at the start of the 3w segment and the site of chromophore attachment. These kinks in H2, H3 and H7 are in contact with the chromophore and would be shared by the visual pigments. Helix kinks have been noted in rhodopsin structures determined in the tetragonal form of 3D crystals and in the p22121 form of 2D crystals; w'ls our definition (Table 2) differs slightly from earlier descriptions regarding location and angle of the kinks. The kinks are also important as sites of lateral polar interactions, where rotation of the peptide planes enables formation of inter-helical H-bonds and water-mediated H-bonds. Thus the helices are linked by three-dimensional H-bonding networks, ls'19 which in our refined structure are more extensive, laterally and perpendicular to the membrane, than previously reported. Figure 5 shows one of the H-bonding networks that involves Asn55 at the kink in H1, and also Asp83 in H2, Gly120 in H3, Met257 and Trp265 in H6, Ser298, Ala299, Tyr301 and Asn302 in H7, and four water molecules including the hitherto unidentified Watl0. This network extends from Trp265 in the chromophore binding pocket via Watl0 all the way to the highly conserved NPxxY motif s3 (Asn302-Tyr306 in bovine rhodopsin) in H7 close to the cytoplasmic limit of the hydrophobic zone (see Figure 3(a)). Asn55 and Asp83 in this network are respectively the most conserved residues in H1 and H2 of rhodopsin-like GPCRs. 54 Trp265, Ser298, Ala299 and Asn302 are all widely conserved across this family, 1 and Met257 is conserved in vertebrate opsins. The proximity and interactions of these residues are critical to the stability and function of GPCRs and of rhodopsin in particular. The H-bonds between Asn55, Asp83, between Asn55 and the peptide O of Ala299 in the kink of H7 at the C terminus of its 310 segment, and the watermediated H-bonds between Asp83 and Asn302 on the other side of this H7 kink can be expected to stabilise the packing of H1, H2 and H7 in the dark state. H-bonding between these conserved residues underlies the apparent interaction amon~ corresponding residues in related GPCRs. 5s-59 For example, the thyrotropin-releasing hormone receptor was completely inactivated by mutation to a non-h-bonding residue at the position of Asp83 or at both positions of Asn55 and Asn In several receptors an Asp-Asn pair at positions of Asp83 and Asn302 was required for activation of G proteins, but their positions can be interchanged without loss of this activity The examples suggest that different H-bonding interactions of these residues are required for stabilising the dark state and for attaining the activated state. In bovine rhodopsin Asp83 is not in contact with the chromophore (Figure 5), but its replacement by Asn causes a slight blue shift of the absorption maximum, indicating an alteration of the Schiff base environment. 6 Low-temperature Fourier transform infrared (FTIR) difference spectroscopy showed structural changes in two internal water molecules and peptide groups during the conversion to bathorhodopsin, 61 the first intermediate state following photo-isomerisation of the retinal, which would have been reached within picoseconds of light absorption had it occurred at physiological temperature. 3"62"63 The spectral change due to one of the water molecules is abolished when the Glu113 counterion is replaced by Gin, but the change due to the second water molecule persists with a frequency shift. 61 The latter was thought to be vibrationally linked to the Schiff base region through the peptide backbone and at a location affected by replacement of Asp83 and Gly '64 This interpretation is borne out by the water-mediated H-bonds between Asp83 and the peptide backbone of H3 and H7 (Figure 5), and Wat12 can be identified as that second water molecule. Other FTIR studies showed that Asp83 is protonated in the dark state and remains so in the biochemically active metarhodopsin II (MII) state, but it undergoes a change or increase of H-bonding on entering MII. " Note that the internal water molecules in the surrounding of Asp83 (Figure 5) could, by small movements, facilitate rapid exchange of H-bonds without requiring concerted movement of the protein. Trp265 in this H-bonding network was one of the first chromophore contacts identified. 66'67 Its side-chain is in a U-shaped bend formed by the 11-cis-retinal and the side-chain of Lys296, with the indole plane perpendicular to the membrane and in contact with the ionone ring (Figures 5 and 6). Resonance Raman data suggest that its microenvironment becomes less hydrophobic within picoseconds of photon absorption, which was attributed to changing interaction with the ionone ring. 6s On MII formation, UV absorbance changes also indicate a weakening of the indole H-bond in Trp265 and Trp In the dark state structure the side-chain of Trp265 is oriented with the NH group pointing to the cytoplasmic side and forming an H-bond with Watl0 (Figure 5), which is located in-line with the NH group. The UV difference in MII might indicate a net displacement of the Trp265 side-chain towards the extracellular side. The newly identified water molecule Watl0 is in close contact with the side-chain of Ala124 in H3 (Figure 5). In rhodopsin-like GPCRs the residue at position 124 is highly conserved, but the predomii nant amino acid is Ser. From this we infer that

12 1420 Rhodopsin Structure in Trigonal Crystals El.?4 T62 L H3 L 13 L76 H7 M257 H1,o303 L79 ~ : 'i: ~,~ H6 [953 N55 e Y3oi w9 " A299 /,#,-,$298 Gf20 ~ "... W265 _ G89 0 G90 K296 Figure 5. A transmembrane slice showing H-bonding networks and hydrophobic contacts between the retinal-binding pocket and the cytoplasmic surface. Watl0 is a surrogate for the Ser OH in providing an H-bond to the indole NH of Trp265, and furthermore that H-bonding between this pair of sidechains in H3 and H6 is a conserved feature of GPCR structure in the inactive state. The side-chain of Ser298 is adjacent to Trp265, but it is H-bonded instead to the main chain O at residue 295 (Figure 5). This side-chain to mainchain H-bond compensates for the elongation of helix pitch at the start of the 310 segment (residues ). In rhodopsin-like GPCRs, position 298 is highly conserved as Ser or Asn," both being residues capable of forming side-chain to main chain H-bonds. Therefore it is likely, that H7 in the wider family also contains a 310 segment at this level in the membrane. Asp83 and Trp265 are both linked by H-bonds through ordered water molecules to Asn302 and Met257. This extended H-bonding network is sealed from the cytoplasmic compartment by a hydrophobic barrier made of six residues from helices H2 (Leu76, Leu79), H3 (Leu128, Leu131) and H6 (Met253, Met257). Five of those residues are highly conserved in rhodopsin-like GPCRs, 1 and the sixth, Met253, is 100% conserved in vertebrate visual pigments. 49 They are arranged with methionine residues on one side and leucine residues on the other (Figure 5) and could facilitate the relative movement of H3 and H64 following photoactivation while retaining van der Waals contacts. Replacement of Met257 by all other amino acids except Leu caused bovine opsin to activate Gt

13 Rhodopsin Structure in Trigonal Crystals 1421 ~~ H6 Figure 6. Environment of the 11-cis-refinal. (a) Environment of the protonated Schiff base and its counterion Glu113. (b) Environment of the ionone ring and the kinks in H6 and H7 cross-linked by a water molecule. constitutively. The level of constitutive activity is well correlated with the affinity of mutant opsins for all-trans-retinal, suggesting that the mutations disposed the apoprotein towards the active conformation normally attained only following light-induced chromophore isomerisation. 7 By implication, Met257 in wild-type rhodopsin contributes a critical packing interaction. It is at the junction between the H-bonding network and the hydrophobic barrier, is in contact with Asn302 of the NPxxY motif in H7, and prevents the helix movement until this is triggered by chromophore isomerisation. Tyr306, at the cytoplasmic end of the NPxxY motif in H7, is at the height of the lipid headgroup layer (Figure 3(a)) where it is H-bonded via Wat7 to Thr62 in H1 and directly to Asn73 in H2 (Figure 5), two residues which are conserved in vertebrate visual pigments and rhodopsin-like GPCRs, respectively. Tyr306 is stacked against Phe313 in H8, and similarly Ile307 in H7 is in contact with Met317. Such interactions can relay movements of transmembrane helices to parts of the cytoplasmic surface, for example realigning H8 after the photoisomerisation. Spin-label data on Tyr306, Phe313 and Cys316 showed light-dark structural differences, 71 and double spin labelling showed movements by 2-4 A of the cytoplasmic end of H7 relative to that of H1, 72 and of H2 relative to H8. 73 The proper realignment of H8 appears to be important in the interaction between the cytoplasmic surface of rhodopsin and the C-terminal peptide of Gta, because the inhibitory mutations either map to the structurally sensitive connecting region between the transmembrane H7 and peripheral H8 (Asn310, Lys311 and Gln312), or form buried contacts between H7 and H8 (Tyr306, Ile307, Phe313 and Met317). 39"74-77 The kinks in H3 and H5 are intimately associated with the lining of the chromophore pocket. At Glu113 in H3, a water molecule (Wat16) is H-bonded between the peptide carbonyl and sidechain carboxyl oxygen atoms. Consequently, the peptide oxygen atom of Glu113 is not H-bonded along the helix axis, hence the kink (Figure 6(a)). Wat16 is most likely the water molecule responsible for that low-temperature FTIR difference feature between the dark and batho states, that was abolished when Glu113 was replaced by Gin. 61 OE1 of Glu113 is H-bonded to the peptide N of Cys187 in ~4 (Figure 6(a)), so the Schiff base counterion is in contact with an H-bonding network that leads from the chromophore-binding pocket, through the plug, to the extracellular surface. This network encompasses Tyr268 in H6, and residues Glu181, Gly182, Gln184, Tyr191 and Tyr192 from the [ and [34-H5 loops (Figure 6(a)), as well as Tyrl0 from ~2 and five water molecules. The H-bonds that cross-link the ~1-B2 and ~3-[34 hairpins (Figure 4) belong to this network. Tyr268, whose OH group is 3.6 A from the 11,12-cis bond of the retinal in the dark state (see Table 4 below) was implicated by FTIR and Raman studies to participate in the initial conformational change at the batho stage and in the transition to the MII state. 68"78'79 The second kink in H3 is in contact with the ionone ring of the chromophore, where H3 and H5 are linked due to a triangular set of H-bonds on the side of the ionone ring away from Trp265 (Figure 6(b)). The Trp126 indole NH forms a pair of bifurcated H-bonds with OE1 of Glu122 and ND1 of His211, while OE2 of Glu122 is H-bonded to the backbone O of His211, which is so oriented because of the kink in H5 at His211 due to Pro215. Glu122 is known to be protonated (neutral) in the dark state and in MII. 6 These interactions account for the

14 1422 Rhodopsin Structure in Trigonal Crystals (a) (b) Figure 7. The non-conventional H-bond between a ring hydrogen of Phe203 in H5 and the backbone O atom in H4. (a) Electron density map around Phe203. (b) Hydrogen bonds between residues in H4, H5 and the several loops connecting these helices and the B3, f~4 strands. Four newly identified water molecules are located in this region. Besides Wat8 and Wat11 shown here, Wat14 is H-bonded to Pro194 O, Glu196 OE1 and Glu201 N, and Wat17 is H-bonded to Tyr192 O and to Gly280 N in the H6-H7 loop. For clarity, the side-chains of Tyr191 and Tyr192 have been omitted. Figures 7(a), 8(a) and 10 were prepared using FTIR observation that mutations of His211 influence Glu122 through a third residue, which is itself affected by mutation of Glu122 to Ala but not to Gin. s Figure 6(b) suggests that Trp126 is the third residue. These residues can be expected to show sensitivity to movements of the ionone ring, with which Glu122 is in direct contact. FTIR difference suggested stronger H-bonding of Glu122 on tran sition to MI, ' whereas UV absorption difference indicated reduced hydrophobicity of the microenvironment for Trp126 as well as weaker H-bonding of its indole NH caused by a general conformational change in the formation of MII. 69 The kinks in H6 and H7 are only 4.8 A apart between C ~ atoms of Tyr268 and Pro291. A water molecule (Watl), H-bonded to the backbone at Tyr268, Pro291 and Ala295, bridges the kinks (Figure 6(b)). H6 shows the most pronounced kink among the rhodopsin helices (Table 2) and contains the conserved Pro267-Tyr268 pair at this kink. Similarity in the light versus dark FTIR difference between [2H]tyrosine-labelled rhodopsin and bacteriorhodopsin has led to the proposal 79 that during photoactivation Pro267-Tyr268 in H6 of rhodopsin may play an analogous role to Tyr185-Pr0186 in helix F of bacteriorhodopsin. 82 That role is to provide a pivot for the outward tilt of the cytoplasmic segment of H6 during the transition 4 to MII as indicated by spin-label studies. The structures of dark-state bacteriorhodopsin and a triple mutant mimic of its M-state (analogous to the MII state in rhodopsin) were shown to differ mainly by an outward tilt of the mutant's helix F by a maximum of 3.5,~ at the cytoplasmic end, which can be described as a hinge movement extrapolating to a pivot around the Tyr-Pro pair. s2 If a similar movement occurs in the transition of rhodopsin to its MII state, then the water molecule between the H6 and H7 kinks would appear to be securing the potential pivot in the configuration of the dark state. At the kink in H4 near its extracellular end, the 2Fo-Fc map (Figure 7(a)) reveals a non-conventional H-bond, with an aromatic ring hydrogen atom of Phe203 in H5 acting as donor s3 to the carbonyl O atom of Cys167 as acceptoro(figure 7(b)). The donor-acceptor distance of 3.10 A indicates a relatively strong H-bond. Residue 203 is Tyr in the consensus sequence of rhodopsin-like GPCRs, and it is Tyr in 76% of the visual pigment sequences but Phe in the rest s4 such as bovine rhodopsin. While

15 178 Rhodopsin Structure in Trigonal Crystals 1423 Table 3. Dihedral angles along the conjugated double bond system of the ll-cis-refinal chromophore Observed values (deg.) Refinement restraints Dihedral angle Molecule A Molecule B Average (error a) (deg.) Target (deg.) Weight (kcal/mol) - C4, C5, C6, C (2) C5, C6, C7, C (2) None none C6, C7, C8, C (2) C7, C8, C9, C (0) C8, C9, C10, Cll (0) C9, C10, Cll, C (2) C10, Cll, C12, C (2) 0 50 Cll, C12, C13, C (2) C12, C13, C14, C (0) C13, C14, C15, N~ (1) C14, C15, N~, Ca (0) a Error: deviation of observed values from their mean in the absence of noncrystallographic symmetry restraints on the retinal moiety. Tyr can be expected to form an H-bond at this position, our observation demonstrates that in the low dielectric medium of the membrane interior, Phe can replace Tyr in donating a stabilising H-bond to the conserved kink in H4. From the Pro170 at the kink in H4 to Asp190 at the end of the hairpin, visual pigments contain nine highly conserved residues, attesting to the importance of this region to the structure and function of rhodopsin, s4 Figure 7(b) shows some of the H-bonds from the conserved Pro170, Pro171, Gly174, Trp175, Tyr178 and Asp190, which fasten the B-hairpin to the helix bundle. The hairpin is anchored to H3 by the disulphide bond between Cys110 and Cys187, and its first and last residues, Arg177 and Asp190, form a salt-bridge (Figure 7(b)) whose removal causes release of retinal from rhodopsin in the dark without apparently increasing the accessibility of the Schiff base to bulk solventy This destabilisation is probably due to perturbation of the H-bonding network (Figure 6(b)) that helps to position 134 at the "floor" of the retinal-binding pocket. Conformation and environment of the chromophore Table 3 lists the torsion angles along the conjugated double bonds of the 11-cis-retinal and its imine linkage to Lys296. Uncertainties were estimated by comparing values in the two NCS copies after one cycle of mock refinement in the absence of NCS restraints on the retinal moiety. The chromophore is confirmed as containing three planar segments, 2 with deviations from the planarity of the n-bonding system about the 6s-cis, and 11-cis, 12s-trans bonds because of the steric interactions between 5-CH3 and 8-H, and between 13-CH3 and 10-H, respectively. The conformation of the chromophore in our structure is similar to that in the 1F88 coordinates for the tetragonal crystal form, and disagrees with the 1HZX and 1L9H 19 coordinates. For example, 1L9H shows a planar 11-cis bond but significantly non-planar C8-C9 and C14-C15 bonds. It should be pointed out that when we used the 1F88 coordinates for the molecular replacement model, we excluded the retinal as a test against model bias, and we have built the initial model of the chromophore from the small molecule crystal structure of 11-cis-retinal 2 by adjusting its torsion angles to fit the cross-crystal averaged electron density map (see Methods), which was consistent with three planar segments. Refinement was carried out with planarity restraints imposed on the chromophore, as was appropriate at the medium resolution. However, the weights of the restraints (Table 3) were set to be 5-12 times smaller than those on peptide bonds or aromatic rings, so they could not have induced a planar conformation had it not been rooted in the X-ray data. Thus our structure provides a credible description of the retinal-protein interactions in the dark state. The torsion angle about the 11-cis bond is determined to be -13(+2) (Table 3 and Figure 6). This value is compatible with the calculation, based on analysis of resonance Raman intensities, that the Cl1=C12 torsion angle increases, within 50 fs after excitation, from -15 in the dark state towards the 90 transition state for the cis-to-trans isomerisation. 86 Both the rapidity and stereospecificity of the initial torsional dynamics are driven by the strain on this non-planar segment, s6"87 The chirality and pucker of the ionone ring are also in agreement with NMR data. s8 Figure 8(a) shows the electron density map around the retinal. Figure 8(b) is a schematic diagram of the retinal-binding pocket, showing H-bonds of Lys296 and Glu113, and the mostly hydrophobic contacts with the retinal within a distance of 4 A. Those contact distances are given in Table 4. The elongated 11-cis-retinal buries about 220/~2 of surface area on a total of 22 residues that form the binding pocket. Roughly one-third of this area is contributed by Trp265 and Tyr268 in H6, another third is from Thr118, Glu122 in H3 and Met207, Phe212 in H5, and the rest is made up by 16 residues from helices H3 to H7 and the plug 14 (Ala117, Lys296, Cys187, Phe261, Ala269, Phe208, Ala292, Tyr191, Gly121, His211, Gly188, Cys167, Ile189, Glu181, Ser186 and Ala272 in descending order of contact area). The large number of contact residues in addition to its covalent linkage

16 1424 Rhodopsin Structure in Trigonal Crystals (a) (b) C~Z ('El Phe 293 ~ c o ~ CF2 ~ N Retinal Binding Pocket CA Thr 297 Me~ (X;IcG2( 3,25/,,, / ~ 3.~(")" Ala 295 Glu 122 ~ ~ Ala 117./ c5, c3 ~212 ceu 47~ Lys 296 Key o Ligand bond G)~113 Non-ligand bond O... Hydrogen bond and its length c)el~ ; ~ I... ~-o~ x d 2.!8 WAT 16,..~d././.. N~ 2.8~ Cys 187 Gly 188 c,o ~ c'~6 ~Met 207 Phe 91 Non-ligand residues involved in hydrophobic "##~" contact(s) Thr 25 Non-ligand residues in contact using polar ")z~ " atoms CI7 ~lle 189 Corresponding ligand atoms involved in contact(s) Figure 8. The retinal-binding pocket. (a) Electron density map around the 11-cis-retinal. (b) Schematic drawing of the binding pocket showing all residues within 4 A of the retinal and Lys296. This drawing was modified from the output of LIGPLOT) 48

17 ,.. o 0 o (~ CO U t~ C',I (~ I I e,1 t~ P~ U U P1 i~ (',1 U u~ 0 U ~0 r-- o'1 c,~ CO r-- U t,1 i U ~D f,1 M 0"1 o :U c,1 ~0 r~ 0 o"1 t~l ~D io P1 ~D ol o'1 I 0,< 0 0 o co n~ o U U ~ U U U U U 0 0 U 0 L) (J Z 0 U 0 0 U Z U 0 U ~ U L) L) U L;, U rju L) UU U ~ U 0 0 U ~ U ~ o ~.~Y~,-'4 ~ r.tl ~ [-.t ~ 0 I 0 u

18 1426 Rhodopsin Structure in Trigonal Crystals contribute to a favourable binding enthalpy of -11 kcal mol In our structure, the seven residues with the lowest B-factors, Trp265, Ala295, Ala269, Thr118, Phe91, Ala117 and Tyr268, are all located around the retinal-binding pocket (Figure 8), which attests to the stabilising interaction between the chromophore and its binding pocket. This stabilisation is responsible for the "inverse agonist" effect of the 11-cis-retinal that keeps the basal activity of rhodopsin towards Gt in the dark very low, below that of the ligand-free opsin, which is itself 106-fold less active than the MII state with all-trans-retinal bound. 9 The three aromatic residues showing the largest chromophore contact areas, Trp265, Tyr268 and Phe212, are highly. conserved throughout rhodopsin-like GPCRs," suggesting common elements in the mechanism for stabilising the inactive receptor conformation from the boundaries of the ligand-binding pocket. Electrostatic interactions of the binding pocket with the chromophore affect the visible absorption maximum Three residues near the ionone ring correspond to the major colour ttming positions in red/green cone pigments identified by primate genetics. 95 Among these Phe261 and Ala269 are direct contacts of the ionone ring (Figure 6(b)), and mutagenesis to Tyr and Thr, respectively, reproduced the predicted red shift, 91 which is consistent with a dipolar stabilisation of the positive charge redistribution in the excited state chromophore towards the ionone ring. 94"96 The third residue, Ala164, is a second shell contact of the ionone ring via Glu122. In rhodopsin, its substitution by Ser produced only a minor red shift, 91 but this can be explained by the dipole contribution of Ser being overshadowed by the intervening Glu122, 97 since replacing Glu122 by Asp was known to cause a blue shift. 66 In the red/green cone pigments, where the conserved residue at the position 122 is Ile, a strong tuning effect can be expected. The structure of the binding pocket is illustrated in Figure 6 with the Schiff base (Figure 6(a)) or the ionone ring (Figure 6(b)) in the foreground. In the Schiff base region, the Lys296 side-chain is fully extended and held in a channel between Met44, Leu47 and Phe91 from H1 and H2 and Phe293 from H7, as previously reported. 14 The Glu113 side-chain is in an unusual rotamer because Wat16 is H-bonded between its peptide O and carboxyl OE2 atoms. The OE2 atom is thus placed in-line with the NH bond of the protonated Schiff base. Thr94 and Gly90 from H2 make close contact with Glu113 and Wat16 bound to it (see below). OE1 of Glu113 is H-bonded to the backbone of Cys187, which is in a disulphide bond with Cys110 in H3. Thus the geometry of the Lys296, Glu113 ion pair is stringently defined. The structure suggests that a salt-bridge between them can also exist in the apoprotein, where it is required for maintaining the inactive state. 98 In Figure 8(b), Wat16 forms a shorter H-bond to the negatively charged carboxyl OE2 than to the neutral peptide O, showing that it is strongly bound to the anion. The Wat16-mediated H-bonds delocalise the negative charge on the Glu113 carboxylate group and lower its pka. Stabilisation of the anion, aided by the precise geometry of the ion pair, in turn stabilises the proton on the Schiff base, elevating its pka in the dark state to > Therefore spontaneous deprotonation is kept at a low rate, which is essential for dim light vision. 1 Change of geometry in the Schiff base region after the photoisomerisation can destabilise the proton and pave the way for the deprotonation of Schiff base in MII. The distance between OE2 of Glu113 and the Schiff base nitrogen atom is 3.2 A (Table 4) on both NCS copies, and is in agreement with the 3.1 A distanceoin the P41 crystal form. 19 This is less than the 4.3 A distance implied by a similarity of 15N chemical shift of the protonated Schiff base in rhodopsin and that in model retinylidene with iodide as counterion. 1 1 However, delocalisation of the negative charge in Glu113 effectively moves the centre of the counterion further away from the protonated Schiff base by about 1 A, thus weakening their electrostatic coupling to permit greater delocalisation of the positive charge from the Schiff base into the polyene of the retinal, 1 2 which was indicated by the nitrogen chemical shift. Therefore, Glu113 and Wat16 form a complex counterion in the dark state, which both stabilises the proton on the Schiff base and determines the wavelength of the visible absorption maximum. Mutating Glu113 to the smaller Asp results in a red shift without dependence on external chloride ions is also consistent with an increase of the ion pair distance. 1 3 Recently, because the Glu181 to Gln mutation was found to lower the pka of the Schiff base under MI-like conditions, the counterion was proposed to switch in MI to Glu181,1 4'1 s accompanied by reorganisation of the E2 loop coupled to movements of H3. l s The dark state structure suggests that the proposed backbone movements are not required for the Glu181 side-chain to get close enough to the Schiff base to measurably influence its electrostatic environment. The Schiff base NH bond in the all-trans retinal may be tilted more towards Ser186 and Glu181 on the H6, H7 side of the polyene backbone (out of the page in Figure 6(a)), instead of tilting towards Glu113 on the H3 side as in the dark state structure. 1 6 The NMR result, that the ionone ring and the adjacent segment of the retinal in MI despite being more relaxed remain largely in the dark state conformation, 1 7 is consistent with the retinal isomerisation up to this stage involving rotations mainly on the Schiff base side of the isomerised bond. There is also evidence that the retinal isomerisation impinges on Tyr "78 Its effect on the H-bonding network (Figure 6(b)) might allow the Glu181 side-chain to rotate towards the Schiff base. Such localised rearrangements from the known dark state structure may apply to the invertebrate rhodopsins, which lack an acidic residue at position 113 and use a conserved residue

19 Rhodopsin Structure in Trigonal Crystals 1427 corresponding to Glu181 as the counterion, 1 8 and to UV-absorbing pigments, which have a neutral Glu113 in the dark state and switch to Glu181 for the counterion in MI. 109"110 Steric interactions with the binding pocket allow relaxation of the photo-isomerised retinal to trigger conformational changes that result in Gt activation in MII. The ionone ring is under steric restraints from residues in helices H3, H5 and H6 (Table 4; Figure 6(b)) up to the MI state./ 7 Evidence for a change in the microenvironment of adjacent Trp26569 in the formation of MII may indicate a movement of the ionone ring. A ring movement will impinge on Trp265, which is the most ordered residue in the dark state structure, the residue with the largest contact area with the retinal, and is widely conserved among GPCRs. It may also affect Phe261, a conserved aromatic residue in visual pigments which is in contact with the C3, C4 edge of the ring and located in the cytoplasmic segment of H6 one helical turn above Trp265 and one below Met257 (Figure 5). The residue at position 261 in the glycoprotein hormone receptors has been shown to play a similar role to Met257 in rhodopsin in forming a stabilising helix packing interaction with Asn302 in H7. 7 '111 Thus retinal interactions with the aromatic residues following photoisomerisation could precipitate a release of the packing interactions between H3 and H6 (see previous subsection on Transmembrane domain), and between H6 and H7, 7 that stabilise the inactive state. As shown in Figure 5, and in Figure 9(a), when the dark state structure is placed in the membrane frame, the ionone ring appears vertically below the conserved Glu134, Arg135 charge pair at the cytoplasmic border of H3, where light-induced protonation of Glu134 is a prerequisite for rhodopsin to bind Gt The H-bonding networks and van der Waals contacts described in the transmembrane domain serve as conduits for vectorial transmission of the conformational changes across this distance, in details that are still unknown. Deletion of the 19-CH3 attached to C9 of the polyene chain was shown to result in a looser fit of the chromophore analogue in the bindin~ pocket, consequently reducing the entropz 115", 89 gain that drives the conversion from MI to MII. Therefore the 19-CH3 in the native structure, wedged between Thr118 from H3, and Ile189 and Tyr191 from the plug (Figure 6(a)), appears to provide a grip against the floor of the binding pocket and restrict the relaxation of the chromophore so that it is directed into a productive activation. Thus steric interactions of all parts of the chromophore are important for the activation. Cytoplasmic domain Having positioned the X-ray coordinates in the membrane frame (Figure 3(a)), we regard all residues above the mean level of phosphate groups (beyond 20 A from the mid-plane of the bilayer) as belonging to the cytoplasmic domain. They include the three cytoplasmic loops (C1, C2, C3) and adjoining ends of transmembrane helices (residues Gln64-Pro71, Glu134-His152 and Gln225-Arg252), the peripheral helix H8 and adjoining loops (residues Asn310-Asn326), and a di-peptide fragment from the C-terminal tail (Figure 9). Cl loop and C-terminal fragment The C1 loop (residues 65-69) is very similar to that in the P41 form. 14 It is well ordered (Figure 3(a)) and comprised of basic and hydrophobic residues (His65, Lys66, Lys67, Leu68, Arg69), which except for His65 are more than 97% conserved among vertebrate opsins, more conserved than the ends of H1 and H2. Leu68 and Arg69 point inward, making van der Waals and H-bonding contacts with residues in H1, H2 and H8; the other residues point outward to form a basic patch. The 2Fo-F c map next to the C1 loop shows a strong density feature with the flatness of a peptide plane in contact with the side-chains of His65 and Lys67, and the main chain from His65 to Lys67. Its aqueous location and shortness make it rather unlikely to belong to a detergent molecule. We have interpreted this density as a di-peptide from the disordered C-terminal tail and tentatively assigned it to Asp330 and Asp331, because of the basic contact site and because the distance from the last ordered residue, Asn326, can be spanned by three or more residues with reasonable main chain torsion angles. The C terminus of the putative di-peptide points toward H6. No other interpretable density was found for the C-terminal tail in the P31 crystal form. The presence of a thermolysin cleavage site between Pro327 and Leu and an Asp-N endoproteinase cleavage site between Gly329 and Asp33016 is consistent with solvent exposure of the unobserved segment. Site-directed spin labelling of rhodopsin in dodecylmaltoside micelles showed the C-terminal tail to be in equilibrium between an ordered and disordered conformation, with the disordered component increasing dramatically beyond Asp331. '* The density adjacent to loop C1 probably corresponds to the partially ordered component. In the P41 structure, loop C1 also forms contact with the C-terminal tail, but the chain orientations (Figure 2(b)) and residue assignments differ. Our tentative assignment of Asp330 and Asp331 pointing towards H6 is reconcilable with the report that the Cys mutant at position 338 in the C-terminal tail does not form a disulphide with the Cys mutant at position 65 in the C1 loop, but it does cross-link with Cys mutants at positions 242, 245 and 246 in H6, rapidly in the dark and slowly in the light. 117 C2 loop and cytoplasmic ends of I-t3 and H4 The C2 loop (residues ) is L-shaped in both crystal forms, but lies more parallel with the membrane surface in the present structure (Figure 2(b)). It is as if the two conformations

20 1428 Rhodopsin Structure in Trigonal Crystals (a) H4 R147 C2 Loop C1 Loop A153 H2 P71 F148 V137 J Y136 ~k~o ~ -..." : T229 H1 t Y306 L72 E24; ~ / A H8 M R314 M309 E249 t ~ T242 T243 K248 T251 A234 C3 Loop (b) I.tAt Figure 9. The cytoplasmic domain. (a) Residues implicated in interaction with the heterotrimeric G protein transducin. The main chain is coloured in the four segments that project above the lipid phosphate groups and form the cytoplasmic domain: C1 loop region (residues 64-71), C2 loop region ( ), C3 loop region ( ), and H8 and C-terminal tail ( ). The side-chain colours indicate: green, mutations causing constitutive Gt activation; red, residues cross-linked to Gt~; pink, mutations inhibiting stabilisation of MII by Gtc~ ; yellow, mutations reducing Gt activation; greenish-yellow, gain-of-function for Gt activation upon reverse mutation from Ala; blue, Arg135 of the conserved E(D)RY motif; blue-grey, other residues discussed in the text. Residues falling in more than one category are coloured according to the former category. Long broken chains joining pairs of C a atoms marked by backbone spheres indicate inter h '" -~ -~ v. = - ehcal cross-links that inhibit activation. ' The footprint of the 11-cis-retinal in the membrane plane (light pink) is shown in the background to illustrate its vertical alignment with key residues on the cytoplasmic surface. (b) Topography of the cytoplasmic surface showing a cavity adjacent to H3, with Leu131 at its bottom, and the Glu134- Arg135 pair forming a salt-bridge in its walls. Note the close contact of Arg135 with H6. Part Co) was prepared using PYMOL (

21 Rhodopsin Structure in Trigonal Crystals 1429 were related through a hinge movement, and comparisons with the NCS differences showed that it is most likely caused by different packing contacts in the two crystal forms. The coordinate differences are above background level for residues (Figure 2(a)), starting at the point where the highly tilted H3 becomes accessible from the surface (Figure 9). Thus the cytoplasmic tip of H3 appears to bend with C2 as the loop is re-oriented. The middle of C2, where B-factors exceed 100 A 2 in both crystal forms, is probably rather flexible. Peptides corresponding to the C2 and C3 loops, and H8, have been shown to interact with Gt. n18 Residues important for Gt activation (Figure 9(a)) are found in the C terminus of H3 and in the C2 loop. Glu134, Arg135 and Tyr136 form the highly conserved E(D)RY motif found in H3 of all rhodopsin-like GPCRs. In the dark state, Glu134 is negatively charged and forms a salt-bridge with Arg135. The latter faces H6 and forms van der Waals contacts with Va1250, Thr251 and Va1254 in H6. Protonation of Glu134, in addition to deprotonation of the Schiff base, is required for rhodopsin to activate Gt. 114 The cytoplasmic surface of rhodopsin shows a cavity next to H3 (Figure 9(b)). Leu131 from the hydrophobic cluster that seals the inter-helical H-bonding network (Figure 5) from the cytoplasm is visible from the opening. But the cavity extends like a tunnel under Pro71 and Arg147 to reach Phe148 and Ala153, and may provide a site for transducin to interact with rhodopsin. The Glu134, Arg135 charge pair is sequestered in the wall of this cavity in the dark state, so that they can form a salt-bridge, but Glu134 also has access to the aqueous phase to take up a proton. Protonation is thought to cause Glu134 to move to a non-polar environment and favours binding to Gt. 113 Loss of the salt-bridge partner might also cause Arg135 to transfer to the aqueous phase, thus breaking its van der Waals contacts with the residues along H6 (Figure 9(a)). This would favour a parting of the cytoplasmic ends of H3 and H6, which is believed to occur in the activation of rhodopsin as indicated by studies using spin labelling, 4 metal chelation, 119 and disulphide cross-linking (Figure 9(a)). 35 The Gln134 mutant, which mimics the protonation, favours binding of rhodopsin to Gt 113 and causes constitutive activity 112 of opsin. This mutant shows increased mobility of spin label attached to Cys140 in H3 or Cys316 in H8 in the dark, 12 which is otherwise associated with photoactivation Eliminating the charge on Arg135, the other half of the salt-bridge, does not cause the same mobility change. 12 The binding of Gt or the C-terminal peptide of its ~-subunit (Gt~ ) stabilises MII. 124'125 Replacing the Tyr136-Va1139 sequence reduced Gt activation by ~ 80% and abolished stabilisation of MII by Gt~ Tyr136, which points outwards and does not interact with the rest of the protein, is an important determinant for the stabilisation of MII by Gt~ , but Va1137-Va1139 may also contribute stabilising interactions for Gt binding, as Va1138 is adjacent to Glu134 and Arg135 (Figure 9(a)). Mutating individual residues of Tyr136-Va1139 to cysteine also caused partial loss of Gt activation. 27 A spin label attached to Lys140 showed a rise and a fall of mobility with time constants corresponding to the formation and decay of the MII state, la2 Therefore, the cytoplasmic end of H3 is involved in the formation of the active state and binding of Gt. In the C2 loop (residues ), mutating Lys141 to Cys reduced Gt activation. 126 Reagents attached to this residue cross-linked photoactivated rhodopsin to Gt~. 38 Replacing the sequence Lys141-Arg147 caused partial loss of Gt activation but did not prevent binding of Gt~ After mutating Lys141-Phe148 to polyalanine abolished activation, restoring only two residues from each end, namely Cys140, Lys141, Arg147 and Phe148, restored 50% of the activity, but the middle residues also contribute to activation, possibly by influencing the conformation of the loop.3gthe flexible middle section might become more ordered when MII binds to Gt and help to increase the affinity of Gt binding. C3 loop and cytoplasmic ends of H5 and 1-16 Helices H5 (residues ) and H6 (residues ) in our structure are both longer by one helical turn at the cytoplasmic boundary in the P41 form. 19 Therefore the C3 loop is elevated above the membrane surface (Figure 2(b)). This loop has the highest B-factors in the whole rhodopsin molecule (Figure 3(a)). However, these B-factors are authentic indicators of main chain flexibilities, since the relative scale of B-factors in the C3 loop compared to those in the C2 loop is similar to the relative mobilities of these surface loops measured by spin-label studies in detergent micelles. 43 During model building, main chain fragments selected from a data base of well-refined, high-resolution structures by the program O 48 for their similarity to the densities of the cytoplasmic ends of H5 and H6 and the loop were tightly clustered and all showed helix-like sections. After refinement with the best fitting fragment inserted into the C3 loop density, the real space correlation was 0.72 in the C3 loop, whereas the average for the whole structttre was The density was poor for residues Ala233-Ala235 and Gln238, which have the highest B-factors. Coordinates for the C3 loop represent the most probable conformation in a population of overlapping, variable conformations. However, the extra helical turn at the ends of H5 and H6 compared with the P41 coordinates is unambiguous. From residue 231 to 236 the C3 loop rises from the membrane like an extension of H5, but with the axis tilted from H5 towards the cytoplasmic tip of H3; from residue 236 onwards, the loop follows another spiral path running parallel with the membrane until it joins H6. The tip of H6 from residue 244 to

22 1430 Rhodopsin Structure in Trigonal Crystals 241 is bent towards H5 to meet with the loop. The secondary structure of loop C3 consists of hydrogen-bonded turns and not regular 0~-helices, 33 but the periodicity of accessibilities detected by a previous spin-label study 33 is accounted for. Therefore the P31 structure and the spin label study both show the C3 loop to be well elevated from the bilayer and mobile but not unstructured. Residues around the C3 loop that were shown by mutagenesis to affect Gt activation map to the cytoplasmic end of H5 (Leu226, Thr229 and Va1230) 127 and the first turn of H6 (Thr242, Thr243 and Gln244), 39"127 which are ordered in the dark state, and to the spiral extension of H5 (Ala233, Ala234), 29"127 which is highly mobile. All three regions project well above the cytoplasmic surface in our structure (Figures 2(b) and 3(a)). Ser240 next to the N terminus of H6 is the residue most elevated above the membrane. Reagents attached to the Ser240Cys side-chain cross-linked the lightactivated rhodopsin mutant to Gto~. 37"38'128 The cross-linked sites were located within the Gt~ peptides 19-28, and ,128 which overlap respectively with the three rhodopsinbinding regions of Gt~ mapped with synthetic peptides, 8-23, and I25 The mutants Lys141Cys and Lys248Cys can also be cross-linked to Gt~z. 38 Replacing Glu247-Thr251 in H6 by unrelated sequence reduced Gt activation tenfold and abolished Gt0t binding. 28 Restoring the sequence of Glu247-Glu249 restored Gt0~ binding, but only part of the Gt activation. 28 Tyr136-Va1139 and Glu247-Glu249 were proposed to form a binding pocket for Gt= In the dark state structure they map to H3 and H6 at similar heights from the membrane (Figure 3(a)). The effects of the mutations may have resulted from interactions with other sites, since Va1250 and Thr251 are in contact with Arg135 in H3, Thr251 with Leu226 in H5 (Figure 9(a)), and Glu249 with residues in H7 and H8 (see next subsection on H8). In chimeras of the Gt-activating bovine rhodopsin and Go-activating scallop rhodopsin, replacing the C2 loop with the scallop sequence reduced activation of both G proteins, but replacing the C3 loop with the scallop sequence enhanced activation of Go fourfold while depressing activation of Gt by 60%. 129 Therefore the C3 loop contains sites for specific activation of G protein. The ability of the C3 loop of bovine rhodopsin to discriminate between different Gt~ sequences may be a mechanism to ensure specificity of coupling. 129 H8 and adjoining loops The structure of H8 (residues ) is very similar in both crystal forms. Asn310 is the only non-helical residue between H7 and H8. Several side-chain to main chain H-bonds stabilise the H7/H8 corner, such as from Glu249 to Met309 and Lys311, from Asn310 to Phe313 and from Arg314 to Ile307. In addition, H8 is anchored at the membrane-aqueous interface by buried sidechain contacts with residues in H1 and H7 underneath, such as the contacts of Phe313 with Thr58, Va161, Thr62 and Tyr306, and of Met317 with Leu57, Va161 and Ile307. The presence of Asn310 and Lys311 are required for correct folding of the rhodopsin. 39 If these two residues are kept, then replacing all other residues in H8 except Phe313 and Met317 by Ala, including replacing all outwardpointing residues, produced a mutant pigment with a potency for Gt activation equal to the wild-type rhodopsm. The contacts of these four residues just described suggest that they are required for the correct tertiary structure and orientation of H8. Deleterious effects of mutations in the N-terminal part of H8 are all interpretable in terms of perturbation of the interaction between H8 and the transmembrane domain It has been shown that trtmcating rhodopsin following Asn315 does not diminish Gt activation. 13 Therefore, although H8 participates in binding of Gt, its role in activation of Gt is secondary to the C2 and C3 loop regions. Expected location of G protein interaction sites in other GPCRs In the C2 loop, residues important to activation have been mapped in the conserved E(D)RY motif or just C-terminal to this. The C3 loop varies widely in length among GPCRs. As seen in the bovine rhodopsin structure, however, residues believed to participate in Gt binding and activation are clustered around the cytoplasmic ends of H5 and H6, and the spiral extension immediately above H5, while residues shown by alanine scanning mutagenesis to have no effect on Gt activation29"map to the C-terminal section of C3, which runs across the gap between H5 and H6. We may expect that, in the C3 region of other GPCRs, the Gt interaction sites will be similarly clustered with respect to the transmembrane helices. In the cannabinoid receptor, the C3 loop is about 30 residues long and a small helix was detected by NMR adjacent to the c y to p lasrmc " end of H5, *131 which is analogous to the H-bonded spiral extension of H5 in bovine rhodopsin, even though there is no sequence homology in this region. In the muscarinic acetylcholine receptors, the C3 loops are very large, but deletions of over 100 residues to reduce it to the length of C3 in bovine rhodopsin affected only the desensitisation process and left the G proteindependent functions intact, such as agonist and antagonist binding and phospholipase acti vanon. ' Therefore it is likely that these residues mapped in the C2 and C3 regions of bovine rhodopsin are the common G protein-binding sites in all GPCRs. Bound lipid and detergent molecules The presence of a bound phospholipid molecule near the cytoplasmic ends of H6 and H7

23 " Rhodopsin Structure in Trigonal Crystals 1431 (a) (b) ~". : '-~:~:~..\ 9-.~: '-dj',,~? ( {~: '"->4"' ~(" 1't1~2u4 \.. I't1~212 /. " v'.:l" ;,.:~.~.~ " r,.,,_',,s ~ :s > -7' _-..4: t>th'j',t',~i~,/!l'li% %~T,, - >..- " :;~ '.3- s's,<:,. L ] ". :X "/ I'r,,2Ul 0 ~" "" <,. \,- i~,\, # ).,,- (c), :~,J ".:-<-C~ ~ ~:~ j I,,..,., 5- -~" ' l ~Y "~"': /,4 / / ~ ~ >,i //"!.1.~ ~" tu/isn; Ilalh9. I ll/. llll ttql~tgl [ _~,, ~ +-. Figure 10. Bound lipid and detergent molecules. (a) Two acyl chains of a phospholipid molecule bound to H6 and H7 in the cytoplasmic half of rhodopsin. The cytoplasmic segment of H6 is orientated roughly along the membrane normal. The acyl chains are inclined relative to it. Met309 is the C-terminal residue of H7. (b) An ordered LDAO molecule bound to H5, H6 and H7just to the extracellular side of the kinks in H6 and H7. Pro267 is one of the contact residues. (c) A C8E4 molecule bound in the crevice between stacked helix bundles of the P31 crystals.

24 1432 Rhodopsin Structure in Trigonal Crystals (Figure 10(a)) is indicated by densities for a pair of acyl chains with a mean separation of 4.15(+0.02)A, which is characteristic of closepacked hydrocarbon tails in an ordered bilayer. The lipid acyl chains form contacts with residues in H6, Ile256, Ala260 and I1e263 in molecule A and Ile256, Ala260 in molecule B, and in H7, Thr297, Tyr301, Va1304, Ile305, Met309 in molecule A and Thr297, Ty301, Ile305 and Met309 in molecule B. The identity of this phospholipid is unknown, because no density was observed for the head group, which is located in the solvent channel and is apparently disordered. The addition of 0.05% LDAO to CsE4-solubilised rhodopsin was the key to obtaining isotropically diffracting crystals of this form. 16 The structure shows a single ordered LDAO molecule in addition to several partially ordered C8E4 molecules bound to each rhodopsin molecule. The LDAO is bound just to the extracellular side of the H6 kink (Figure 10(b)) and is itself kinked at the C5 atom, with the amine oxide arm pointing roughly along H6 towards the extracellular face and the hydrocarbon arm running from H6 towards H5. The contacts are with H6 mainly (Leu266, Pro267, Gly270, Va1271, Phe273, Tyr274), but also with H5 (Phe208 and Phe212) and H7 (Phe287). In wrapping around these helices, the LDAO may be re-enforcing the more ordered molecular structure originating from the extracellular leaflet (see Figure 3(b)), therefore stabilising a particular protein conformation resulting in better ordered crystals. Pro267 in the binding site of LDAO has been proposed as a possible hinge in H643"79 for the outward tilt of its cytoplasmic segment upon photoactivation. 4'3sm9 The presence of a hinge midway along H6 is supported by disulphide linking experiments showing a lack of relative movement between the extracellular ends of H6 and H5 during activation. TM Note that in the dark state structure H6 is kinked in a plane tangential to the helix bundle. In the activated form the cytoplasmic segment of H6 is kinked about the same place but in a radial plane instead. For this change to occur, the structure of the hinge region may have to undergo a rotation first. Recently, cryo-em of rhodopsin 2D crystals highly enriched for the MI state showed a local density change next 135"_ to the dark state position of Trp265, which would be consistent with mobilisation of the hinge prior to the larger conformational changes required to enter the active MII state. Thus the LDAO molecule wrapped around the kink in H6 appears to stabilise the inactive conformation around a potential hinge. The most completely modelled CsE4 molecule shows the octyl chain and three ordered oxyethylene groups out of the total four. The elongated detergent molecule is intercalated in a tapered crevice between rhodopsin helix bundles, as they are stacked along the 31-screw axis. Together with the palmitoylate chains linked to Cys322 and Cys323 it fills this crevice (Figure 10(c)). There are hydrophobic-to-hydrophobic and hydrophilic- to-hydrophilic contacts between the C8E4 and two protein molecules, one of them presenting the cytoplasmic end of H7 and H8 and the other presenting the extracellular end of H4 and H5 (Figure 10(c)). The choice of C8E4 as the detergent and its interactions in this crevice most likely precipitated the P31 crystal form. Methods Data collection and processing Crystallisation and preparation of heavy atom derivatives are described in the ~companying paper) 6 Complete data sets were obtained:~6~ combining wedges of collected from different positions on one or more crystals. Data were processed using programs in the CCP4 suite. 136 Intensities were ii~tegrated to anisotropic resolution limits using MOSFLM, and merged using SCALA. A consistent indexing regime was maintained, by re-indexing each wedge under the four indexing regimes of point group 3, and selecting the one leading to the lowest R-factor on amplitudes (Rderiv), calculated using SCALEIT, relative to a previously merged reference data set or partial data set. Using the same procedure, data sets from non-isomorphous crystals were placed under a common indexing regime prior to cross-crystal averaging. The intensities were scaled in two rounds. The first round was done in the "batch" mode, with the batch showing the least negative B-factor serving as reference for B-factor normalisation; batches found with relative B-factors above 13, indicating significant radiation damage, were rejected from further scaling. The second round used smooth scale and B-factors, in which secondary beam corrections in the form of spherical harmonic functions were determined for each crystal. A "tails" correction for contributions from diffuse scattering was applied during both batch and smooth scaling rounds. Twinning was detected by inspecting the cumulative intensity distribution given by TRUNCATE. When it was present, the twin fraction was estimated and intensities were de-twinned using the program DETWIN. TM Molecular replacement from the 1 F88 coordinates and refinement Molecular replacement was done using the program AMoRe 137 and the rhodopsin coordinates in the P41 crystal form at 2.8 A resolution (PDB code 1F8814"17) as search model. The two rhodopsin molecules in the P41 asymmetric unit were superimposed to form one search model, but only the more ordered and more complete of the two molecules (chain A) was used to calculate initial phases. Non-protein atoms were omitted from the search model and initial phase calculation. In the calculated maps, the density for the retinal moiety, which was omitted from the model, was used to assess the quality of the phasing. Molecular replacement was done in parallel against each data set in Table 1A, and each solution was subjected to rigid body refinement, using first one molecule, then one helix, per rigid group. Cross-crystal averaging was carried out among Native-2, EMTS-1 and EMTS-2, together with solvent flattening, 2-fold NCS averaging and histogram matching, in DMMULTI. 138 The atomic

25 Rhodopsin Structure in Trigonal Crystals 1433 model was rebuilt using 0 48 into the averaged map of Native-2. The model was refined using CNS 22 with simulated annealing and group B-factors. NCS restraints were applied between the two molecules in the asymmetric unit, but the restraints were relaxed in the loop regions. This was because in simulated annealing trials, the loop residues showed systematically larger inter-molecular coordinate differences than residues in the helices, regardless of whether the NCS restraints have been imposed over the loops, whereas excluding the loop regions from NCS restraints was found to reduce the free-r-factor and improve overall stereochemistry. Native-3 was not averaged with the other data sets, but was subjected to 20 cycles of solvent flattening (46% solvent) and 2-fold NCS averaging, during which the averaging radius was reduced and the resolution extended from 2.9 A to 2.65 A. The resulting map showed clear densities for the retinal and carbohydrate residues, the palmitoyl modifications to Cys322 and Cys323, as well as a few tightly bound water and detergent molecules. The 1F88 model used for molecular replacement was extensively rebuilt, especially the side-chain conformations. Coordinates for the retinal and carbohydrate moieties and cytoplasmic loops were taken from the partially refined Native-2 model and rebuilt. Refinement against Native-3 data was done using simulated annealing and individual B-factors, under NCS restraints imposed over the helical segments. When the free- R-factor decreased to 24.5%, ordered water molecules were located in the difference map using the water pick task file in CNS. After another round of refinement with individual B-factors, NCS copies of the bound water molecules were located in the aa-weighted 2F o- Fc map and confirmed in the solvent-flattened EDEN 23 map (kindly calculated by Dr Luca Jovine). Detergent and phospholipid molecules were added during refinement. Geometric parameters for the 11-cis-retinal were taken from the small molecule structure, 2 except the 12-s-cis bond was made trans, and parameters for the protonated Schiff base were taken from the retinylidene structure. 21 In early rounds of the refinement, dihedral angles in the polyene section of retinal were tightly constrained to the plane using energy constants for the tryptophan ring (500 kcal/mol), but for dihedrals about the C6-C7, Cll-C12 and C12-C13 bonds the energy constants were reduced tenfold. After addition of water molecules, the tight dihedral restraints along the polyene were relaxed to 100 kcal/mol. Initial models for the detergent molecules LDAO and C8E4, and the hydrocarbon chains of phospholipids were taken from structures in the Protein Data Bank as provided by the HIC-UP web site% Topology parameters for these groups and for the thioester linkage of palmitoylated cysteine were derived from structures in the Cambridge Small Molecule Database. Molecular replacement from an EM model From 2D crystals of bovine rhodopsin imaged by electron cryomicroscopy, a three-dimensional map has been calculated to 5 A resolution in the plane and 13.5 A normal to it. 1 The seven transmembrane helices appear as density rods of unknown azimuthal orientations. To reduce the distortion caused by the missing cone in data collected from 2D crystals, this map was subjected to one t cycle of solvent-flattening using the program dm 139 and a solvent content of 75%, which was estimated from the volume per molecule in the P31 crystal 16 and a unit cell thickness of 200 A in the 2D crystal. The solvent-flattened EM map was re-calculated to 6 resolution for skeletonisafion, r4 The bones within the boundary of a single rhodopsin molecule were edited in O to follow the high-density ridges along the helix axes, and they were annotated in O for the helices represented, in accordance with the model of helix arrangement proposed by Baldwin et al. 12 Using the bone_mask option in O and a 6 A radius for the bones atoms, a molecular mask was formed. The mask was edited using MAMA TM to remove internal cavities, and converted to the CCP4 format on a 1 A sampling grid using MAMA2ccp4.136 The solvent-flattened EM map was interpolated onto the 1A grid inside the molecular mask using MAPROT, 1~2 to yield the masked density of a single rhodopsin molecule. The centre of gravity for the masked region was shifted to the origin of a P1 unit cell that has 150 A cell edges and 90 cell angles, by editing the map header using the program 1MEDIT in the MRC image processing suite.14 The asymmetric unit was filled with zeros outside the mask, using the program MAPMASK. TM This density of a single rhodopsin molecule, cut out from the EM map, and placed in a large P1 cell was the "EM model" for molecular replacement. Structure factors for the EM model were calculated using SFALL. TM Cross-rotation function was calculated using GLRF, TM in which the best signal-to-noise ratio was obtained by using a 28 A radius of Patterson integration and a resolution limit of 9-5 A. Using MAPROT14Zagain, the EM model was rotated about its centre of gravity to the cross-rotation angles, and re-interpolated into the actual P31 unit cell at 1 A sampling intervals. Translation searches were conducted using TFFC 145 in both P31 and 16 P32. The molecular orientation found was subsequently confirmed by molecular replacement from the 1F88 coordinates, once these became available. The molecular orientations found using the EM model agreed with those found subsequently from the atomic model to within 2. However, the map of translation search from the EM model showed only streaks rather than discrete peaks. Docking X-ray coordinates to the EM map The C a coordinates of the 1F88 model and of the refined structure from this work were first manually aligned with the "bones" for the EM map, which trace out the helix axes. Then using the RSR_rigid option in O, in the convolution mode, the fit of all atoms of the protein molecule to the EM density was optimised. Protein Data Bank accession numbers Coordinates and structure factors have been released through the RCSB Protein Data Bank (code 1GZM, 1GZM-SF). Acknowledgements We thank the ESRF, SPring-8, APS and Daresbury Laboratory for providing synchrotron radiation facilities and staff support during data collection,

26 1434 Rhodopsin Structure in Trigonal Crystals and in particular the ESRF for a long-term award to G.F.X. for use of the microfocus beamline ID13. We are grateful to Drs Richard Henderson, Paul Hargrave, Joyce Baldwin, Harry Powell and Alexei Murzin for helpful discussions; Luca Jovine for calculating the EDEN map from partially refined coordinates; Richard Henderson, Phil Evans, Andrew Leslie and Dan Oprian for comments on the manuscript; colleagues at the MRC Laboratory of Molecular Biology for encouragement. We acknowledge the use of the GPCRDB data base. References 1. Baldwin, J. M. (1993). The probable arrangement of the helices in G protein-coupled receptors. EMBO J. 12, Hubbard, R. & Kropf, A. (1958). The action of light on rhodopsin. Proc. Natl Acad. Sci. USA, 44, Wald, G. (1968). Molecular basis of visual excitation. Science, 162, Farrens, D. L., Altenbach, C., Yang, K., Hubbell, W. L. & Khorana, H. G. (1996). Requirement of rigid-body motion of transmembrane helices for light activation of rhodopsin. Science, 274, Sakmar, T. E (1998). Rhodopsin: a prototypical G protein-coupled receptor. Prog. Nucl. Acid Res. Mol. Biol. 59, Baylor, D. A., Matthews, G. & Yau, K. W. (1980). Two components of electrical dark noise in toad retinal rod outer segments. J. Physiol. 309, Scherfler, G. F., Villa, C. & Henderson, R. (1993). Projection structure of rhodopsin. Nature, 362, Scherfler, G. E & Hargrave, P. A. (1995). Projection structure of frog rhodopsin in two crystal forms. Proc. Natl Acad. Sci. USA, 92, Unger, V. M. & Schertler, G. F. (1995). Low resolution structure of bovine rhodopsin determined by electron cryo-microscopy. Biophys. J. 68, Krebs, A., Edwards, P. C., Villa, C., Li, J. & Schertler, G. F. (2003). The three-dimensional structure of bovine rhodopsin determined by electron cryomicroscopy. J. Biol. Chem. 278, Davies, A., Gowen, B. E., Krebs, A. M., Schertler, G. E & Saibil, H. R. (2001). Three-dimensional structure of an invertebrate rhodopsin and basis for ordered alignment in the photoreceptor membrane. J. Mol. Biol. 314, Baldwin, J. M., Schertler, G. F. & Unger, V. M. (1997). An alpha-carbon template for the transmembrane helices in the rhodopsin family of G-protein-coupled receptors. J. Mol. Biol. 272, Unger, V. M., Hargrave, P. A., Baldwin, J. M. & Schertler, G. F. (1997). Arrangement of rhodopsin transmembrane alpha-helices. Nature, 389, Palczewski, K., Kumasaka, T., Hori, T., Behnke, C. A., Motoshima, H., Fox, B. A. et al. (2000). Crystal structure of rhodopsin: a G protein-coupled receptor. Science, 289, Bourne, H. R. & Meng, E. C. (2000). Structure. Rhodopsin sees the light. Science, 289, Edwards, P. C., Li, J., Burghammer, M., McDowell, J. H., Villa, C., Hargrave, P. A. & Schertler, G. F. X. (2004). Crystals of native and modified bovine rhodopsins and their heavy metal derivatives. J. Mol. Biol. jmb56619 accompanying. 17. Okada, T., Le Trong, I., Fox, B. A., Behnke, C. A., Stenkamp, R. E. & Palczewski, K. (2000). X-ray diffraction analysis of three-dimensional crystals of bovine rhodopsin obtained from mixed micelles. J. Struct. Biol. 130, Teller, D. C., Okada, T., Behnke, C. A., Palczewski, K. & Stenkamp, R. E. (2001). Advances in determination of a high-resolution three-dimensional structure of rhodopsin, a model of G-protein-coupled receptors (GPCRs). Biochemistry, 40, Okada, T., Fujiyoshi, Y., Silow, M., Navarro, J., Landau, E. M. & Shichida, Y. (2002). Functional role of internal water molecules in rhodopsin revealed by X-ray crystallography. Proc. Natl Acad. Sci. USA, 99, Gilardi, R. D., Karle, I. L. & Karle, J. (1972). The crystal and molecular structure of 11-cis-retinal. Acta Crystallog. sect. B, 28, Santarsiero, B. D. & James, M. N. G. (1990). Crystal structure of N-methyl-N-phenylretinal iminium perchlorate: a structural model for the bacteriorhodopsin chromophore. J. Am. Chem. Soc. 112, Briinger, A., Adams, P., Clore, G., DeLano, W., Gros, P., Grosse-Kunstleve, R. et al. (1998). Crystallography & NMR system: a new software suite for macromolecular structure determination. Acta Crystallog. sect. D, 54, Sz6ke, A. (1993). Holographic methods in X-ray crystallography. II. Detailed theory and connections to other methods of crystallography. Acta Crystallog. sect. A, 49, Read, R. (1986). Improved Fourier coefficients for maps using phases from partial structures with errors. Acta Crystallog. sect. A, 42, K6nig, B., Welte, W. & Hofmann, K. P. (1989). Photoactivation of rhodopsin and interaction with transducin in detergent micelles-effect of doping with steroid molecules. FEBS Letters, 257, Franke, R. R., Konig, B., Sakmar, T. P., Khorana, H. G. & Hofmann, K. P. (1990). Rhodopsin mutants that bind but fail to activate transducin. Science, 250, Ridge, K. D., Zhang, C. & Khorana, H. G. (1995). Mapping of the amino acids in the cytoplasmic loop connecting helices C and D in rhodopsin. Chemical reactivity in the dark state following single cysteine replacements. Biochemistry, 34, Acharya, S., Saad, Y. & Karnik, S. S. (1997). Transducin-alpha C-terminal peptide binding site consists of C-D and E-F loops of rhodopsin. J. Biol. Chem. 272, Shi, W., Osawa, S., Dickerson, C. D. & Weiss, E. R. (1995). Rhodopsin mutants discriminate sites important for the activation of rhodopsin kinase and Gt. J. Biol. Chem. 270, Thurmond, R. L., Creuzenet, C., Reeves, P. J. & Khorana, H. G. (1997). Structure and function in rhodopsin: peptide sequences in the cytoplasmic loops of rhodopsin are intimately involved in interaction with rhodopsin kinase. Proc. Natl Acad. Sci. USA, 94, Kisselev, O. G., Downs, M. A., McDowell, J. H. & Hargrave, P. A. (2004). Conformational changes in the phosphorylated C-terminal domain of rhodopsin during rhodopsin arrestin interactions. J. Biol. Chem. In the press. 32. McDowell, J. H., Ubel, A., Brown, R. A. & Hargrave,

27 Rhodopsin Structure in Trigonal Crystals 1435 P. A. (1986). Transglutaminase modification of rhodopsin in retinal rod outer segment disk membranes. Arch. Biochem. Biophys. 249, Altenbach, C., Yang, K., Farrens, D. L., Farahbakhsh, Z. T., Khorana, H. G. & Hubbell, W. L. (1996). Structural features and light-dependent changes in the cytoplasmic interhelical E-F loop region of rhodopsin: a site-directed spin-labeling study. Biochemistry, 35, Langen, R., Cai, K., Altenbach, C., Khorana, H. G. & Hubbell, W. L. (1999). Structural features of the C-terminal domain of bovine rhodopsin: a site-directed spin-labeling study. Biochemistry, 38, Cai, K., Klein-Seetharaman, J., Hwa, J., Hubbell, W. L. & Khorana, H. G. (1999). Structure and function in rhodopsin: effects of disulfide cross-links in the cytoplasmic face of rhodopsin on transducin activation and phosphorylation by rhodopsin kinase. Biochemistry, 38, McDowell, J. H., Nawrocki, J. P. & Hargrave, P. A. (1993). Phosphorylation sites in bovine rhodopsin. Biochemistry, 32, Resek, J. F., Farrens, D. & Khorana, H. G. (1994). Structure and function in rhodopsin: covalent crosslinking of the rhodopsin (metarhodopsin II)- transducin complex-the rhodopsin cytoplasmic face links to the transducin alpha subunit. Proc. Natl Acad. Sci. USA, 91, Itoh, Y., Cai, K. & Khorana, H. G. (2001). Mapping of contact sites in complex formation between lightactivated rhodopsin and transducin by covalent crosslinking: use of a chemically preactivated reagent. Proc. Natl Acad. Sci. USA, 24, Natochin, M., Gasimov, K. G., Moussaif, M. & Artemyev, N. O. (2003). Rhodopsin determinants for transducin activation: a gain-of-function approach. J. Biol. Chem. 278, Blaurock, A. E. & Wilkins, M. H. F. (1972). Structure of retinal photoreceptor membranes. Nature, 236, Saibil, H., Chabre, M. & Worcester, D. (1976). Neutron diffraction studies of retinal rod outer segment membranes. Nature, 262, Lewis, B. A. & Engelman, D. M. (1983). Lipid bilayer thickness varies linearly with acyl chain length in fluid phosphatidylcholine vesicles. J. Mol. Biol. 166, Hubbell, W. L., Altenbach, C., Hubbell, C. M. & Khorana, H. G. (2003). Rhodopsin structure, dynamics, and activation: a perspective from crystallography, site-directed spin labeling, sulfhydryl reactivity, and disulfide cross-linking. Advan. Protein Chem. 63, Fukuda, M. N., Papermaster, D. S. & Hargrave, P. A. (1979). Rhodopsin carbohydrate. Structure of small oligosaccharides attached at two sites near the NH2 terminus. J. Biol. Chem. 254, Doi, T., Molday, R. S. & Khorana, H. G. (1990). Role of the intradiscal domain in rhodopsin assembly and function. Proc. Natl Acad. Sci. USA, 87, Illing, M. E., Rajan, R. S., Bence, N. F. & Kopito, R. R. (2002). A rhodopsin mutant linked to autosomal dominant retinitis pigmentosa is prone to aggregate and interacts with the ubiquitin proteasome system. J. Biol. Chem. 277, Chothia, C. & Lesk, A. M. (1986). The relation between the divergence of sequence and structure in proteins. EMBO J. 5, Jones, T. A., Zou, J. Y., Cowan, S. W. & Kjeldgaard, M. (1991). Improved methods for building protein models in electron density maps and the location of errors in these models. Acta Crystallog. sect. A, 47, Horn, F., Bettler, E., Oliveira, L., Campagne, F., Cohen, F. E. & Vriend, G. (2003). GPCRDB information system for G protein-coupled receptors. Nucl. Acids Res. 31, Sakmar, T. P., Franke, R. R. & Khorana, H. G. (1989). Glutamic acid-113 serves as the retinylidene Schiff base counterion in bovine rhodopsin. Proc. Natl Acad. Sci. USA, 86, Zhukovsky, E. A. & Oprian, D. D. (1989). Effect of carboxylic acid side chains on the absorption maximum of visual pigments. Science, 246, Nathans, J. (1990). Determinants of visual pigment absorbance: identification of the retinylidene Schiff's base counterion in bovine rhodopsin.1. Biochemistry, 16, Probst, W. C., Snyder, L. A., Schuster, D. I., Brosius, J. & Sealfon, S. C. (1992). Sequence alignment of the G-protein coupled receptor superfamily. DNA Cell Biol. 11, Ballesteros, J. A. & Weinstein, H. (1995). Integrated methods for the construction of three dimensional models and computational probing of structurefunction relations in G-protein coupled receptors. Methods Neurosci. 25, Perlman, J. H., Colson, A. O., Wang, W., Bence, K., Osman, R. & Gershengorn, M. C. (1997). Interactions between conserved residues in transmembrane helices 1, 2, and 7 of the thyrotropin-releasing hormone receptor. J. Biol. Chem. 272, Flanagan, C. A., Zhou, W., Chi, L., Yuen, T., Rodic, V., Robertson, D. et al. (1999). The functional microdomain in transmembrane helices 2 and 7 regulates expression, activation, and coupling pathways of the gonadotropin-releasing hormone receptor. J. Biol. Chem. 274, Zhou, W., Flanagan, C., Ballesteros, J. A., Konvicka, K., Davidson, J. S., Weinstein, H. et al. (1994). A reciprocal mutation supports helix 2 and helix 7 proximity in the gonadotropin-releasing hormone receptor. Mol. Pharmacol. 45, Sealfon, S. C., Chi, L., Ebersole, B. J., Rodic, V., Zhang, D., Ballesteros, J. A. & Weinstein, H. (1995). Related contribution of specific helix 2 and 7 residues to conformational activation of the serotonin 5-HT2A receptor. J. Biol. Chem. 270, Donnelly, D., Maudsley, S., Gent, J. P., Moser, R. N., Hurrell, C. R. & Findlay, J. B. (1999). Conserved polar residues in the transmembrane domain of the human tachykinin NK2 receptor: functional roles and structural implications. Biochem. J. 339, Fahmy, K., Jager, F., Beck, M., Zvyaga, T. A., Sakmar, T. P. & Siebert, F. (1993). Protonation states of membrane-embedded carboxylic acid groups in rhodopsin and metarhodopsin II: a Fouriertransform infrared spectroscopy study of site-directed mutants. Proc. Natl Acad. Sci. USA, 90, Nagata, T., Terakita, A., Kandori, H., Kojima, D., Shichida, Y. & Maeda, A. (1997). Water and peptide backbone structure in the active center of bovine rhodopsin. Biochemistry, 36, Wang, Q., Schoenlein, R. W., Peteanu, L. A., Mathies,

28 1436 Rhodopsin Structure in Trigonal Crystals R. A. & Shank, C. V. (1994). Vibrationally coherent photochemistry in the femtosecond primary event of vision. Science, 266, Kandori, H., Shichida, Y. & Yoshizawa, T. (2001). Photoisomerization in rhodopsin. Biochemistry (Mosc), 66, Nagata, T., Terakita, A., Kandori, H., Shichida, Y. & Maeda, A. (1998). The hydrogen-bonding network of water molecules and the peptide backbone in the region connecting Asp83, Gly120, and Glu113 in bovine rhodopsin. Biochemistry, 37, Rath, P., DeCaluwe, L. L., Bovee-Geurts, P. H., DeGrip, W. J. & Rothschild, K. J. (1993). Fourier transform infrared difference spectroscopy of rhodopsin mutants: light activation of rhodopsin causes hydrogen-bonding change in residue aspartic acid-83 during meta II formation. Biochemistry, 32, Nakayama, T. A. & Khorana, H. G. (1991). Mapping of the amino acids in membrane-embedded helices that interact with the retinal chromophore in bovine rhodopsin. J. Biol. Chem. 266, Nakayama, T. A. & Khorana, H. G. (1990). Orientation of retinal in bovine rhodopsin determined by cross-linking using a photoactivatable analog of 11-cis-retinal. J. Biol. Chem. 265, Kim, J. E., Pan, D. & Mathies, R. A. (2003). Picosecond dynamics of G-protein coupled receptor activation in rhodopsin from time-resolved UV resonance Raman spectroscopy. Biochemistry, 42, Lin, S. W. & Sakmar, T. P. (1996). Specific tryptophan UV-absorbance changes are probes of the transition of rhodopsin to its active state. Biochemistry, 35, Han, M., Smith, S. O. & Sakrnar, T. P. (1998). Constitutive activation of opsin by mutation of methionine 257 on transmembrane helix 6. Biochemistry, 37, Altenbach, C., Cai, K., Khorana, H. G. & Hubbell, W. L. (1999). Structural features and light-dependent changes in the sequence extending from helix VII to the palmitoylation sites in rhodopsin: a site-directed spin-labeling study. Biochemistry, 38, Altenbach, C., Cai, K., Klein-Seetharaman, J., Khorana, H. G. & Hubbell, W. L. (2001). Structure and function in rhodopsin: mapping lightdependent changes in distance between residue 65 in helix TM1 and residues in the sequence at the cytoplasmic end of helix TM7 and in helix H8. Biochemistry, 40, Altenbach, C., Klein-Seetharaman, J., Cai, K., Khorana, H. G. & Hubbell, W. L. (2001). Structure and function in rhodopsin: mapping lightdependent changes in distance between residue 316 in helix 8 and residues in the sequence 60-75, covering the cytoplasmic end of helices TM1 and TM2 and their connection loop CL1. Biochemistry, 40, Fritze, O., Filipek, S., Kuksa, V., Palczewski, K., Hofmann, K. P. & Ernst, O. P. (2003). Role of the conserved NPxxY(x)5,6F motif in the rhodopsin ground state and during activation. Proc. Natl Acad. Sci. USA, 100, Marin, E. P., Krishna, A. G., Zvyaga, T. A., Isele, J., Siebert, F. & Sakmar, T. P. (2000). The amino terminus of the fourth cytoplasmic loop of rhodopsin modulates rhodopsin-transducin interaction. ]. Biol. Chem. 275, Ernst, O. P., Meyer, C. K., Marin, E. P., Henklein, P., Fu, W. Y., Sakmar, T. P. & Hofmann, K. P. (2000). Mutation of the fourth cytoplasmic loop of rhodopsin affects binding of transducin and peptides derived from the carboxyl-terminal sequences of transducin alpha and gamma subunits. J. Biol. Chem. 275, Cai, K., Klein-Seetharaman, J., Farrens, D., Zhang, C., Altenbach, C., Hubbell, W. L. & Khorana, H. G. (1999). Single-cysteine substitution mutants at amino acid positions in rhodopsin, the sequence between the cytoplasmic end of helix VII and the palmitoylation sites: sulfhydryl reactivity and transducin activation reveal a tertiary structure. Biochemistry, 38, DeGrip, W. J., DeLange, F., Klaassen, C. H., Verdegem, P. J., Wallace-Williams, S., Creemers, A. F. et al. (1999). Photoactivation of rhodopsin: interplay between protein and chromophore. In Novartis Found Symposium (Yoshizawa, T., ed.), vol. 224, pp , Wiley, Chichester. 79. DeLange, F., Klaassen, C. H., Wallace-Williams, S. E., Bovee-Geurts, P. H., Liu, X. M., DeGrip, W. J. & Rothschild, K. J. (1998). Tyrosine structural changes detected during the photoactivation of rhodopsin. J. Biol. Chem. 273, Beck, M., Sakmar, T. P. & Siebert, E (1998). Spectroscopic evidence for interaction between transmembrane helices 3 and 5 in rhodopsin. Biochemistry, 37, Furutani, Y., Kandori, H. & Shichida, Y. (2003). Structural changes in lumirhodopsin and metarhodopsin I studied by their photoreactions at 77 K. Biochemistry, 42, Subramaniam, S. & Henderson, R. (2000). Molecular mechanism of vectorial proton translocation by bacteriorhodopsin. Nature, 406, Levitt, M. & Perutz, M. E (1988). Aromatic rings act as hydrogen bond acceptors. J. Mol. Biol. 201, Horn, F., Bettler, E., Oliveira, L., Campagne, F., Cohen, E E. & Vriend, G. (2003). GPCRDB information system for G protein-coupled receptors. Nucl. Acids Res. 31, Janz, J. M., Fay, J. E & Farrens, D. L. (2003). Stability of dark-state rhodopsin is mediated by a conserved ion-pair in intradiscal loop E-2. J. Biol. Chem. 278, Mathies, R. A. (1999). Photons, femtoseconds and dipolar interactions: a molecular picture of the primary events in vision. In Rhodopsins and Phototransduction (Yoshizawa, T., ed.), vol. 224, pp , Wiley, Chichester. 87. Kochendoerfer, G. G., Verdegem, P. J., van der Hoef, I., Lugtenburg, J. & Mathies, R. A. (1996). Retinal analog study of the role of steric interactions in the excited state isomerization dynamics of rhodopsin. Biochemistry, 35, Creemers, A. E, Kiihne, S., Bovee-Geurts, P. H., DeGrip, W. J., Lugtenburg, J. & de Groot, H. J. (2002). (1)H and (13)C MAS NMR evidence for pronounced ligand-protein interactions involving the ionone ring of the retinylidene chromophore in rhodopsin. Proc. Natl Acad. Sci. USA, 99, Cooper, A. & Converse, C. A. (1976). Energetics of

29 Rhodopsin Structure in Trigonal Crystals 1437 primary processes in visual excitation: photocalorimetry of rhodopsin in rod outer segment membranes. Biochemistry, 15, Melia, T. J., Cowan, C., Angleson, J. & Wensel, T. (1997). A comparison of the efficiency of G protein activation by ligand-free and light-activated forms of rhodopsin. Biophys. J. 93, Chan, T., Lee, M. & Sakmar, T. P. (1992). Introduction of hydroxyl-bearing amino acids causes bathochromic spectral shifts in rhodopsin. Amino acid substitutions responsible for red-green color pigment spectral tuning. J. Biol. Chem. 267, Lin, S. W., Kochendoerfer, G. G., Carroll, K. S., Wang, D., Mathies, R. A. & Sakmar, T. P. (1998). Mechanisms of spectral tuning in blue cone visual pigments. Visible and Raman spectroscopy of blue-shifted rhodopsin mutants. J. Biol. Chem. 273, Kochendoerfer, G. G., Lin, S. W., Sakmar, T. P. & Mathies, R. A. (1999). How color visual pigments are tuned. Trends Biochem. Sci. 24, Fasick, J. I., Lee, N. & Oprian, D. D. (1999). Spectral tuning in the human blue cone pigment. Biochemistry, 38, Neitz, M., Neitz, J. & Jacobs, G. H. (1991). Spectral tuning of pigments underlying red-green color vision. Science, 252, Birge, R. R. & Hubbard, L. M. (1980). Molecular dynamics of cis-trans isomerization in rhodopsin. J. Am. Chem. Soc. 102, Imai, H., Kojima, D., Oura, T., Tachibanaki, S., Terakita, A. & Shichida, Y. (1997). Single amino acid residue as a h~nctional determinant of rod and cone visual pigments. Proc. Natl Acad. Sci. USA, 94, Rao, V. R. & Oprian, D. D. (1996). Activating mutations of rhodopsin and other G protein-coupled receptors. Annu. Rev. Biophys. Biomol. Struct. 25, Steinberg, G., Ottolenghi, M. & Sheves, M. (1993). pka of the protonated Schiff base of bovine rhodopsin. A study with artificial pigments. Biophys. J. 64, Hecht, S., Schlaer, S. & Pirenne, M. H. (1942). Energy, quanta, and vision. J. Gen. Physiol. 25, 819~ Creemers, A. F. L., Klaassen, C. H. W., Bovee-Geurts, P. H. M., Kelle, R., Kragl, U., Raap, J. et al. (1999). Solid state 15N NMR evidence for a complex Schiff base counterion in the visual G-protein-coupled receptor rhodopsin. Biochemistry, 38, Irving, C. S., Byers, G. W. & Leermakers, P. A. (1970). Spectroscopic model for the visual pigments. Influence of microenvironmental polarizability. Biochemistry, 9, Sakmar, T. P., Franke, R. R. & Khorana, H. G. (1991). The role of the retinylidene Schiff base counterion in rhodopsin in determining wavelength absorbance and Schiff base pka. Proc. Natl Acad. Sci. USA, 88, Yan, E. C., Kazmi, M. A., De, S., Chang, B. S., Seibert, C., Marin, E. P. et al. (2002). Function of extracellular loop 2 in rhodopsin: glutamic acid 181 modulates stability and absorption wavelength of metarhodopsin II. Biochemistry, 41, Yan, E. C., Kazmi, M. A., Ganim, Z., Hou, J. M., Pan, D., Chang, B. S. et al. (2003). Retinal counterion switch in the photoactivation of the G proteincoupled receptor rhodopsin. Proc. Natl Acad. Sci. USA, 100, Pan, D., Ganim, Z., Kim, J. E., Verhoeven, M. A., Lugtenburg, J. & Mathies, R. A. (2002). Timeresolved resonance Raman analysis of chromophore structural changes in the formation and decay of rhodopsin's BSI intermediate. J. Am. Chem. Soc. 124, Spooner, P. J., Sharpies, J. M., Goodall, S. C., Seedorf, H., Verhoeven, M. A., Lugtenburg, J. et al. (2003). Conformational similarities in the beta-ionone ring region of the rhodopsin chromophore in its ground state and after photoactivation to the metarhodopsin-i intermediate. Biochemistry, 42, Terakita, A., Yamashita, T. & Shichida, Y. (2000). Highly conserved glutamic acid in the extracellular IV-V loop in rhodopsins acts as the counterion in retinochrome, a member of the rhodopsin family. Proc. Natl Acad. Sci. USA, 97, Birge, R. R. & Knox, B. E. (2003). Perspectives on the counterion switch-induced photoactivation of the G protein-coupled receptor rhodopsin. Proc. Natl Acad. Sci. USA, 100, Kusnetzow, A. K., Dukkipati, A., Babu, K. R., Ramos, L., Knox, B. E. & Birge, R. R. (2004). Vertebrate ultraviolet visual pigments: protonation of the retinylidene Schiff base and a counterion switch during photoactivation. Proc. Natl Acad. Sci. USA, 101, Govaerts, C., Lefort, A., Costagliola, S., Wodak, S. J., Ballesteros, J. A., Van Sande, J. et al. (2001). A conserved Asn in transmembrane helix 7 is an on/off switch in the activation of the thyrotropin receptor. J. Biol. Chem. 276, Cohen, G. B., Yang, T., Robinson, P. R. & Oprian, D. D. (1993). Constitutive activation of opsin: influence of charge at position 134 and size at position 296. Biochemistry, 32, Fahmy, K. & Sakmar, T. P. (1993). Regulation of the rhodopsin-transducin interaction by a highly conserved carboxylic acid group. Biochemistry, 32, Arnis, S. & Hofmann, K. P. (1993). Two different forms of metarhodopsin II: Schiff base deprotonation precedes proton uptake and signaling state. Proc. Natl Acad. Sci. USA, 90, Vogel, R., Fan, G. B., Sheves, M. & Siebert, F. (2000). The molecular origin of the inhibition of transducin activation in rhodopsin lacking the 9-methyl group of the retinal chromophore: a UV-vis and FTIR spectroscopic study. Biochemistry, 39, Bayramashvili, D. I., Drachev, A. L., Drachev, L A., Kaulen, A. D., Kudelin, A. B., Martynov, V. I. & Skulachev, V. P. (1984). Proteinase-treated photoreceptor discs. Photoelectric activity of the partiallydigested rhodopsin and membrane orientation. Eur. J. Biochem. 142, Cai, K., Langen, R., Hubbell, W. L. & Khorana, H. G. (1997). Structure and function in rhodopsin: topology of the C-terminal polypeptide chain in relation to the cytoplasmic loops. Proc. Natl Acad. Sci. USA, 94, K6nig, B., Arendt, A., McDowell, J. H., Kahlert, M., Hargrave, P. A. & Hofmann, K. P. (1989). Three cytoplasmic loops of rhodopsin interact with transducin. Proc. Natl Acad. Sci. USA, 86, Sheikh, S. P., Zvyaga, T. A., Lichtarge, O., Sakmar, T. P. & Bourne, H. R. (1996). Rhodopsin activation blocked by metal-ion-binding sites linking transmembrane helices C and F. Nature, 383, Kim, J. M., Altenbach, C., Thurmond, R. L., Khorana,

30 1438 Rhodopsin Structure in Trigonal Crystals H. G. & Hubbell, W. L. (1997). Structure and function in rhodopsin: rhodopsin mutants with a neutral amino acid at E134 have a partially activated conformation in the dark state. Proc. Natl Acad. Sci. USA, 94, Resek, J. F., Farahbakhsh, Z. T., Hubbell, W. L. & Khorana, H. G. (1993). Formation of the meta II photointermediate is accompanied by conformational changes in the cytoplasmic surface of rhodopsin. Biochemistry, 32, Farahbakhsh, Z. T., Hideg, K. & Hubbell, W. L. (1993). Photoactivated conformational changes in rhodopsin: a time-resolved spin label study. Science, 262, Farahbakhsh, Z. T., Ridge, K. D., Khorana, H. G. & Hubbell, W. L. (1995). Mapping light-dependent structural changes in the cytoplasmic loop connecting helices C and D in rhodopsin: a site-directed spin labeling study. Biochemistry, 34, Emeis, D., Kuhn, H., Reichert, J. & Hofmann, K. P. (1982). Complex formation between metarhodopsin II and GTP-binding protein in bovine photoreceptor membranes leads to a shift of the photoproduct equilibrium. FEBS Letters, 143, Hamm, H. E., Deretic, D., Arendt, A., Hargrave, P. A., Koenig, B. & Hofmann, K. P. (1988). Site of G protein binding to rhodopsin mapped with synthetic peptides from the alpha subunit. Science, 241, Ridge, K. D., Lu, Z., Liu, X. & Khorana, H. G. (1995). Structure and function in rhodopsin. Separation and characterization of the correctly folded and misfolded opsins produced on expression of an opsin mutant gene containing only the native intradiscal cysteine codons. Biochemistry, 34, Yang, K., Farrens, D. L., Hubbell, W. L. & Khorana, H. G. (1996). Structure and function in rhodopsin. Single cysteine substitution mutants in the cytoplasmic interhetical E-F loop region show positionspecific effects in transducin activation. Biochemistry, 35, Cai, K., Itoh, Y. & Khorana, H. G. (2001). Mapping of contact sites in complex formation between transducin and light-activated rhodopsin by covalent crosslinking: use of a photoactivatable reagent. Proc. Natl Acad. Sci. USA, 98, Terakita, A., Yamashita, T., Nimbari, N., Kojima, D. & Shichida, Y. (2002). Functional interaction between bovine rhodopsin and G protein transducin. J. Biol. Chem. 277, Weiss, E. R., Osawa, S., Shi, W. & Dickerson, C. D. (1994). Effects of carboxyl-terminal truncation on the stability and G protein-coupling activity of bovine rhodopsin. Biochemistry, 33, Ulfers, A. L., McMurry, J. L., Kendall, D. A. & Mierke, D. F. (2002). Structure of the third intracellular loop of the human cannabinoid 1 receptor. Biochemistry, 41, Moro, O., Lameh, J. & Sadee, W. (1993). Serine- and threonine-rich domain regulates internalization of muscarinic cholinergic receptors. J. Biol. Chem. 268, Hulme, E. C. & Curtis, C. A. (1998). Purification of recombinant M1 muscarinic acetylcholine receptor. Biochem. Soc. Trans. 26, $ Struthers, M., Yu, H. & Oprian, D. D. (2000). G protein-coupled receptor activation: analysis of a highly constrained "straitjacketed" rhodopsin. Biochemistry, 39, Ruprecht, J. J., Mielke, T., Vogel, R., Villa, C. & Schertler, G. F. X. (2004). Electron crystallography reveals the structure of metarhodopsin I. EMBO J. 23, The Collaborative Computational Project, Number 4. (1994). The CCP4 suite: programs for protein crystallography. Acta Crystallog. sect. D, 50, Navaza, J. (2001). Implementation of molecular replacement in AMoRe. Acta Crystallog. sect. D, 57, Cowtan, K. (1994). "dm": an automated procedure for phase improvement by density modification. Joint CCP4 ESRF-EABM Newsletter Protein Crystallog. 31, Cowtan, K. & Main, P. (1998). Miscellaneous algorithms for density modification. Acta Crystallog. sect. D, 54, Greer, J., Kleywegt, G. J. & Jones, T. A. (1996). xdlmapman and xdldataman-programs for reformatting, analysis and manipulation of biomacromolecular eletron-density maps and reflection data sets. Acta Crystallog. sect. D, 52, Kleywegt, G. J. & Jones, T. A. (1999). Software for handling macromolecular envelopes. Acta Crystallog. sect. D, 55, Stein, P. E., Boodhoo, A., Tyrrell, G. J., Brunton, J. L. & Read, R. J. (1992). Crystal structure of the cellbinding B oligomer of verotoxin-1 from E. coli. Nature, 355, Crowther, R. A., Henderson, R. & Smith, J. M. (1996). MRC image processing programs. J. Struct. Biol. 116, Tong, L. A. & Rossmann, M. G. (1990). The locked rotation function. Acta Crystallog. sect. A, 46, Tickle, I. J. (1992). Fast Fourier translation functions. In Molecular Replacement (Dodson, E. J., Gover, S. & Wolf, W., ed.s), vol. DL/SCI/R33, pp , SERC, Daresbury Laboratory, Warrington, UK Kraulis, P. J. (1991). MOLSCRIPT: a program to produce both detailed and schematic plots of protein structures. J. Appl. Crystallog. 24, Merritt, E. A. & Bacon, D. J. (1997). Raster3D: photorealistic molecular graphics. Methods Enzymol. 277, Wallace, A. C., Laskowski, R. A. & Thornton, J. M. (1995). LIGPLOT: a program to generate schematic diagrams of protein-ligand interactions. Protein Eng. 8, Edited by R. Huber (Received 18 June 2004; received in revised form 26 August 2004; accepted 27 August 2004)