Involvement of the second extracellular loop (E2) of the neurokinin-1 receptor in the binding of substance P. Photoaffinity labeling and modeling studies.

Substance P (SP) interacts with the neurokinin-1 (NK-1) G-protein-coupled receptor, which has been cloned in several species. In the present study, the domains of the NK-1 receptor involved in the binding of SP and SP-(7-11) C-terminal fragment have been analyzed using two peptide analogs containing the photoreactive amino acid para-benzoylphenylalanine ((p-Bz)Phe) in position 8 of their sequence. This study was carried out with [BAPA-Lys(6),(p-Bz)Phe(8),Pro(9),Met(O(2))(11)]SP-(7-11) and [BAPA(0),(p-Bz)Phe(8)]SP on both rat and human NK-1 receptors expressed in CHO cells. Combined trypsin and endo-GluC enzymatic complete digestions and matrix-assisted laser desorption/ionization time-of-flight mass spectrometry analysis led to the identification of the same domain of covalent interaction, (173)TMPSR(177), for the two photoactivatable peptides. Further digestion of this fragment with carboxypeptidase Y led to the identification of (173)TMP(175) in the second extracellular loop (E2) of the NK-1 receptor as the site of covalent attachment. Models of the conformation of this E2 loop in the human NK-1 receptor were generated using two different strategies, one based on homology with bovine rhodopsin and the other based on the solution conformation preferences of a synthetic peptide corresponding to the E2 loop.

cesses including transmission of pain and inflammation but also depression, epilepsy, and angiogenesis (1)(2)(3). These biological effects are mediated via a G-protein-coupled receptor, the NK-1 receptor, which has been cloned in several species (see Ref. 4 for a review). Binding experiments with SP analogs in tissues and in cells transfected with the NK-1 receptor have shown that two types of non-stoichiometric binding sites with distinct pharmacological profiles are associated with the NK-1 receptor (5)(6)(7)(8)(9)(10)(11)(12). In CHO cells, B max values for the two binding sites were found to be 6000 fmol/mg proteins and 800 fmol/mg proteins for the major binding site ( 11 ]SP- (7)(8)(9)(10)(11), respectively (7). With the same clone it has been shown that in the presence of cholera toxin, the tachykinin NK-1 receptor is uncoupled to G-proteins (7). However, B max values remain unchanged for the two radioligands (7). Differences in the B max values were observed as well with a CHO clone expressing a lower level of NK-1 receptors in membrane homogenates prepared from the clone with the highest expression and in rat submandibular glands (7). Furthermore, the two binding sites internalize differently as observed with radiolabeled peptides (12). Altogether these results suggest that the differences in B max values for the two types of ligands cannot be explained by effects secondary to G-protein interactions. With a plethora of SP analogs in CHO cells expressing high levels of the human NK-1 receptor, the binding affinity for the more abundant binding site could be correlated to the potency to accumulate cAMP, whereas the binding affinity for the less abundant site could be correlated to the potency to activate inositol phosphate production (8). In the CHO clone used herein, the more and less abundant binding sites represent 85 and 15% of the total population of receptors (6 pmol/mg of proteins), respectively (7). Substance P binds the two binding sites with high affinity, whereas some C-terminal fragment analogs of substance P and some substance P- (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11) analogs, as well as the endogenous tachykinin NK-2 ligand neurokinin A, bind only the less abundant one (7)(8)(9)(10)(11)(12). Although several structure-activity relationship studies have been carried out, little is known about the differences in the molecular recognition of these two binding sites (8 -12). Photoaffinity labeling therefore appeared to be a complementary method of studying the interaction of both types of peptide agonists with the NK-1 receptor. Photolabeling of the rat NK-1 receptor with a SP and a NKA photoactivatable analog has been reported recently, and subsequent mapping studies have established that the site of photoinsertion was located in the same segment of the second extracellular loop (E2) of the receptor, probably on an identical residue Met 181 (13). * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
In this study, we have used the same "one-pot" strategy of photolabeling, enzymatic digestion, and purification before matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometry analysis described previously (14), with two photoactivatable analogs of SP designed to screen the two binding sites associated with the NK-1 receptor. Both photoactivatable SP analogs, [BAPA-Lys 6 ,(p-Bz)Phe 8 ,Pro 9 ,-Met(O 2 ) 11 ]SP- (7)(8)(9)(10)(11) and [BAPA 0 ,(p-Bz)Phe 8 ]SP, contain a biotinyl sulfone moiety at the N-terminal end, which is separated from the first amino acid by an aminopentanoic acid flexible spacer (BAPA). After photolabeling and enzymatic digestion(s), this biotinyl sulfone moiety is used to purify the fragment of interest via streptavidin-coated magnetic beads (14,15). The substance P receptor fragment covalent complex is then released from the magnetic beads directly with the matrix solution used for MALDI-TOF mass spectrometry analysis (14). Using this strategy, we have previously shown that the SP analog photoactivatable in position 8, [BAPA 0 ,(p-Bz)Phe 8 ]SP, as well as the constrained selective NK-1 analog [BAPA 0 ,(p-Bz)Phe 8 ,Pro 9 ]SP, specifically interact with Met 174 in the second extracellular loop of the human NK-1 receptor (14). Using a similar photoactivatable SP analog, [[ 125 I]BH (p-Bz)Phe 8 ]SP, Kage et al. (16) identified Met 181 as the site of interaction of this photoreactive amino acid in the rat NK-1 receptor, whereas Li et al. (17) established that domain 173-183 was the site of covalent interaction with the NK-1 receptor on the murine cell line P388D 1 .
In this study, we examined the pharmacological profiles of [BAPA-Lys 6 ,(p-Bz)Phe 8 ,Pro 9 ,Met(O 2 ) 11 ]SP-(7-11) and [BAPA 0 ,(p-Bz)Phe 8 ]SP on the two binding sites associated with the NK-1 receptor and carried out photolabeling studies using the strategy described previously (15) with CHO clones expressing either the human or the rat NK-1 receptor. The results indicate that residue Met 174 in the E2 loop is probably the site of covalent attachment. Models of the conformation of the E2 loop were built using two different strategies, one based on homology with bovine rhodopsin and the other on the solution conformation preferences of a synthetic peptide corresponding to the E2 loop.  11 ]SP-(7-11) (3700 GBq/mmol) was synthesized as described (7).
Cell Culture-CHO cells expressing the human or rat NK-1 receptor were cultured in Ham's F12 medium supplemented with 100 IU/ml penicillin, 100 IU/ml streptomycin, and 10% fetal calf serum. Cultures were kept at 37°C in a humidified atmosphere of 5% CO 2 . Stable transfection was maintained by Geneticin (400 mg/liter). Both CHO clones express similar levels of the human or the rat NK-1 receptor (6 pmol/mg proteins).
Measurements of Inositol Phosphate and cAMP Formation-Inositol phosphate hydrolysis and cAMP accumulation were determined as described (5). All determinations were performed in duplicate in at least three independent experiments.
Photoaffinity Labeling of Membrane Preparations-Membranes (1-2 mg of protein at a concentration of 1 mg/ml protein) from CHO cells expressing the rat or the human NK-1 receptor were incubated for 5 min at room temperature with 100 nM of [BAPA-Lys 6 ,(p-Bz)Phe 8 ,Pro 9 ,Met(O 2 ) 11 ]SP- (7)(8)(9)(10)(11) or [BAPA 0 ,(p-Bz)Phe 8 ]SP in 50 mM Tris-Cl buffer, pH 7.4, containing 1 mM EDTA, 10 mM MgCl 2 , 0.1 mM PMSF, 5 g/ml soybean trypsin inhibitor, and 400 mg/ml bovine serum albumin. The membrane preparation was then irradiated on ice for 40 min using an ultraviolet light at 365 nm (HPR 125-watt lamp) at a distance of 6 to 10 cm. After irradiation, [Pro 9 ]SP (10 M) was added for 10 min to the membrane preparation prior to centrifugation for 2 min at 13,000 rpm (MSE Micro Centaur). The sample was then washed with Tris-Cl buffer and centrifuged again. Finally, photolabeled membranes were incubated for 2 h at room temperature in a denaturation buffer consisting of 17 mM dithiothreitol and 3% SDS in 50 l of Tris-Cl 50 mM, pH 8.0.
Streptavidin-coated Magnetic Bead Purification-The tryptic digested sample was incubated with 100 g of streptavidin-coated magnetic beads (Dynabeads M280, Dynal) for 2 to12 h under gentle agitation. The beads were then washed as described previously (14). The sample was then either submitted to further digestion with endo-GluC or analyzed by MALDI-TOF mass spectrometry.
Endo-GluC Digestion of Tryptic Fragments onto Streptavidin-coated Magnetic Beads-After purification and washing (14), the beads were incubated at 37°C for 15 h with 20 g of endo-GluC (Roche Molecular Biochemicals) in 20 l of 100 mM Tris-Cl, pH 7.8. Digestion was stopped by the addition of 10 l of PMSF (100 mM) and 10 l of soybean trypsin inhibitor (5 g.l Ϫ1 ); 100 g of streptavidin-coated magnetic beads were again added for 2 h. Purification and washing steps were performed as described (14). The sample was then either submitted to further digestion with carboxypeptidase Y or analyzed by MALDI-TOF mass spectrometry.
Carboxypeptidase Y Digestion of Trypsin/Endo-GluC Fragments onto Streptavidin-coated Magnetic Beads-After purification and washing (14), the beads were incubated at 37°C for 2 to 24 h with 0.05 g of carboxypeptidase Y in 20 l of 100 mM Tris-Cl, pH 7.0. After incubation the beads were washed as described (14).
MALDI-TOF Mass Spectrometry Analysis-Peptide fragments were eluted from the magnetic beads with 3 l of MALDI matrix ␣-cyano-4hydroxycinnamic acid in 4:1 (v/v) CH 3 CN/H 2 O (0.1% trifluoroacetic acid). After a 10-min incubation, 1 l of bead-free supernatant was deposited on the sample holder for MALDI-TOF MS analysis. MALDI-TOF mass spectra (averaged over 256 laser shots) were obtained in positive mode on a Voyager Elite (PerSeptive Biosystems) mass spectrometer in the reflector mode. For weak ion signals, a better signal/ noise ratio was obtained by averaging 10 mass spectra. External calibration was applied using standard peptides deposited on the MALDI-TOF target very close to the studied sample. In the following, the measured and indicated m/z values are monoisotopic. Peptide receptor domains corresponding to the mass peaks obtained from MALDI-TOF MS analysis were identified using the Protein Analysis WorkSheet freeware edition (ProteoMics; http://www.proteomics.com) and applied to the NK-1 receptor (human and rat) using the different proteases.
Peptide Synthesis-A peptide encompassing the E2 loop of the NK-1 receptor and named E2 peptide, Ac-YSTTETMPSRVVSMIEWPEH-PNKIYEKVY-NH 2 , was synthesized by solid-phase methodology using t-butyloxycarbonyl (Boc) chemistry on an Applied Biosystems Model 431A synthesizer (7). Synthesis was carried out on an 0.1-mmol scale, starting from a p-methylbenzhydrylamine resin (typical substitution, 0.68 mmol/g of resin). All N-␣-t-butyloxycarbonyl amino acids were assembled using dicyclohexylcarbodiimide and 1-hydroxybenzotriazole as coupling reagents. A 10-fold excess of each amino acid was used. The peptide was cleaved from the resin by anhydrous fluorhydric acid in the presence of scavengers and purified to homogeneity by preparative reverse phase C 8 HPLC using water and acetonitrile with 0.1% trifluoroacetic acid as the solvent system. The peptide purity was determined by analytical reverse phase HPLC, and its structural integrity was confirmed by MALDI-TOF mass spectrometry.
Nuclear Magnetic Resonance Spectroscopy-NMR spectra were recorded at temperatures ranging from 293 to 313 K on 2 mM peptide in 90% H 2 O, 10% 2 H 2 O in the presence of 80 mM SDS-d 25 and in mixed solvents containing either 25 or 50% propionate was used as an internal 1 H chemical shift reference. NMR experiments were collected on a Bruker AM 500 or DMX 500 spectrometers and processed with an Aspect 3000 computer or with Bruker UXNMR software. Solvent suppression was achieved by low power irradiation during the relaxation delay (1.5 s) or with a WATER-GATE sequence prior to acquisition (20). To obtain sequence-specific resonance assignments, the following two-dimensional homonuclear experiments were recorded using conventional pulse sequences: DQF-COSY (21), clean-TOCSY (22) with 30 -80 ms mixing times, and NOESY (23) with 60 -300 ms mixing times. To identify amide protons in slow exchange with solvent, the sample containing 25% TFE was lyophilized and resuspended in 2 H 2 O, and a two-dimensional TOCSY experiment was recorded at 288 K over a period of 4 h. Typical experiments were acquired with 2048 points in t 2 and 400 -600 increments in t 1 over a spectral width of 5000 Hz. Prior to Fourier transformation, the time domain data were multiplied with a sine-bell window shifted by /3 and zero-filled. Base-line distortions were corrected using a fifthorder polynomial.
NMR-derived Constraints and Structure Calculations-Interproton distance restraints were derived from NOESY experiments recorded at 273 K with mixing times of 150 and 300 ms. The NOESY cross-peak intensities were converted into distance ranges of 0.18 to 0.28, 0.18 to 0.38, and 0.18 -0.50 nm corresponding to strong, medium, and weak NOEs, respectively. Equivalent methyl protons, aromatic protons, and nonresolved methylene protons were treated as pseudoatoms, and correction factors were applied to distance limits (24). Backbone dihedral angles were restrained from measurement of 3 J HN-H␣ coupling constants, HN-H␣ NOE intensities, and H ␣ chemical shift deviations. Structure calculations were carried out on Silicon graphics O 2 stations. Fifty structures were calculated by torsion angle dynamics with the DYANA program (25). The 20 structures with the lowest target function were energy-minimized in XPLOR (26) using CHARMM 22 force field (27). The structure quality was evaluated using PROCHECK-NMR (28).
Homology Modeling-Models of the human NK-1 receptor were built on Silicon Graphics O 2 workstations with the Modeler program (29) using the crystallographic structure of bovine rhodopsin as a template (Protein Data Bank code 1F88) (30). The initial models were refined in XPLOR by applying several cycles of conjugate-gradient minimization with decreasing constraints on backbone atom positions. The CHARMM 22 force field was used. Nonbonded interactions were calculated with an 8 Å cutoff. The structures were analyzed with Insight II, and the quality of the models was evaluated using PROCHECK (31).
NMR-based Modeling of the E2 Loop in the NK-1 Receptor-The structure of the E2 loop in the NK-1 receptor was calculated by simulated annealing, using the same set of restraints as used for the NMR structure calculation of the isolated loop in solution. Thirty structures were calculated in XPLOR starting from the homology-built structure. Residues were partitioned into three regions according to their proximity to the E2 loop: a free region corresponding to extracellular loops, a restrained region in which constraints were applied to preserve the backbone structure corresponding to proximal transmembrane segments, and a fixed region encompassing all remaining residues. The topallhdg force field was used in conjunction with a simple quartic repulsion potential (26). In the first stage of the protocol, the structures were submitted to 5 ps of restrained molecular dynamics at 1000 K. The force constants of the distance and dihedral angle restraints were linearly increased in 20 steps, while the nonbonded repulsive potential was kept to a low value. The force constants of the terms maintaining the disulfide bond between Cys 105 and Cys 180 were set to zero to allow a better sampling of the conformational space. In the second stage, the force constants maintaining the covalent geometry of the Cys 105 -Cys 180 disulfide bridge were increased linearly in 20 steps during 10 ps of dynamics. The third stage comprised 10 ps of molecular dynamics during which the nonbonded repulsive potential was gradually increased. The structures were then cooled down from 1000 to 0 K over a period of 20 ps. Finally, the structures were energy-minimized using the CHARMM 22 force field with a Lennard-Jones potential and a distance-dependent dielectric function for the electrostatic term.

Pharmacological Profile of the Photoreactive SP Analogs-
The photoreactive analog [BAPA 0 ,(p-Bz)Phe 8 ]SP binds NK-1M and NK-1m binding sites with nanomolar affinities similar to those of SP for rat or human NK-1 receptor expressed in CHO cells (Table I). This analog also activates with potencies similar to those of SP the phospholipase C and adenylate cyclase second messenger pathways (Table I) 11 ]SP-(7-11) (for NK-1m binding site), and potencies to activate phospholipase C and adenylate cyclase were determined as described under "Experimental Procedures." Data presented are the mean Ϯ S.E. of at least three independent experiments performed in duplicate. PLC, phospholipase C; ND, not determined.

NK-1 species
Peptide analog  11 ]SP- (7)(8)(9)(10)(11) or [BAPA 0 ,(p-Bz)Phe 8 ]SP bound to CHO cells expressing either the rat or the human NK-1 receptor, the covalent complex was digested for 5-24 h with L-1-tosylamide-2-phenyl-ethylchloromethyl ketone-treated trypsin. The ligand-receptor fragment covalent complex was then purified via streptavidin-coated magnetic beads and analyzed by MALDI-TOF MS. Tryptic digestion of the covalent complex after photoinsertion of [BAPA 0 ,(p-Bz)Phe 8 ]SP on the rat or the human NK-1 receptor gave, by MS analysis, two peptides with MH ϩ at m/z 2980.4 and 2817.4 (Fig. 1a). The presence and relative intensity of these two peptides were dependent on the time of trypsin digestion, the higher mass peptide disappearing with a long incubation period, leading solely to the lower mass peptide. In addition, we verified through blank experiments that these two peptides came from the digestion of the covalent peptide-receptor complex and not from nonspecific association with impurities from the streptavidin beads. Because the mass of the photoreactive SP analog, [BAPA 0 ,(p-Bz)Phe 8 ]SP, is (MH ϩ ) ϭ 1808.89, the two peptides identified by MS analysis corresponded to peptide fragments from the receptor with a mass of 1171.5 and 1008.5 atomic mass units. Taking into account the uncertainties on the mass measurements (Ͻ0.1 atomic mass unit) and considering that the ligand-receptor complex was digested both by trypsin and also, as described previously, by trypsin-derived chymotrypsin enzymatic activities (14) (Fig. 2a), respectively. The peptide corresponding to the highest mass was the more abundant at short incubation times with trypsin, whereas after long periods of incubation the peptide with the lowest mass was the only one observed. After subtracting the mass of the photoactivatable peptide, domains Tyr 168 -Arg 177 and Ser 169 -Arg 177 of the NK-1 receptor were again identified as the site of photoinsertion.
Subcleavage on Beads of the Tryptic Fragments by Endo-GluC-After trypsin digestion and purification of the covalent peptide-receptor fragment complex on streptavidin-coated magnetic beads, endo-GluC digestion was performed for 15 h at 22°C. For [BAPA 0 ,(p-Bz)Phe 8 ]SP covalently linked either to the rat or human NK-1 receptor, the ions of the tryptic fragments at m/z 2980.41 and 2817.34 were shifted to a single peak at m/z 2399.1 (Fig. 1, b and c) GluC led to a single peak at m/z 1864.9, which corresponded again to domain Thr 173 -Arg 177 (590.29 atomic mass units measured and expected) of the receptor. It should be mentioned that an oxidized form for the species at m/z 1864.9 was also observed (Fig. 2d, inset) indicating that the receptor fragment contains a residue that can be oxidized, likely the methionine in the sequence Thr 173 -Arg 177 .
Carboxypeptidase Y Digestion on Beads of the Tryptic/Endo-GluC Fragments-Combined tryptic/endo-GluC fragments from the human NK-1 receptor linked to [BAPA 0 ,(p-Bz)Phe 8 ]SP were further submitted to carboxypeptidase Y digestion on beads before MS analysis. Carboxypeptidase Y digestion from 6 to 24 h led to peptides corresponding to the removal of both Arg 177 and Ser 176 from the C terminus of the tryptic fragment Ser 169 -Arg 177 or of the combined tryptic/endo-GluC digest Thr 173 -Arg 177 (Fig. 1, d and e). We have previously established that after CNBr cleavage and MALDI-TOF analysis of the fragment, the methyl of the Met 174 side chain was the site of covalent insertion of p-benzoyl probe from both [BAPA 0 ,(p-Bz)Phe 8 ]SP and [BAPA 0 ,(p-Bz)Phe 8 ,Pro 9 ]SP (15). Altogether, experiments reported in this study support this result and Met 174 is also likely to be the site of covalent attachment of [BAPA-Lys 6 ,(p-Bz)Phe 8 ,Pro 9 ,Met(O 2 ) 11 ]SP- (7)(8)(9)(10)(11).
Molecular Modeling of the Human NK-1 Receptor-This study suggests a spatial proximity in the binding site between Met 174 and residue Phe 8 of substance P, whereas studies from Boyd and colleagues (16) indicated that Met 181 was the major site of photoinsertion. Both methionines belong to the E2 loop connecting transmembrane helices H4 and H5; this E2 loop may be either part of or proximal to the binding site in the NK-1 receptor. To get insight into the molecular basis of SP recognition, a model of the human NK-1 receptor was built to predict the conformation of the E2 loop and the positions of Met 174 and Met 181 . The three-dimensional structure of the human NK-1 receptor was modeled on the basis of its structural similarity with bovine rhodopsin, the only G-proteincoupled receptor for which the structure is known at atomic resolution (30). The sequence alignment of transmembrane helices was based on the structure and analysis of conserved residues in the G-protein-coupled receptor (32). The N-terminal extremity and the cytoplasmic tail beyond helix H8 were not considered in the sequence alignment, because no similarities could be detected between the NK-1 receptor and rhodopsin in these regions. The extracellular and the intracellular loops were included in the alignment. The degree of identity over 296 aligned positions is 23%. The sequences of the E2 loops were aligned by making one deletion in the human NK-1 receptor at the end of helix H4. Analysis of the sequences (Fig. 3) indicated that the E2 loops in both proteins show common properties, including conservation of the disulfide bridge between Cys 180 and Cys 105 in transmembrane helix H3 and similar lengths and distribution of polar and nonpolar residues around Cys 180 (Fig.  3). Furthermore, several algorithms predicted an extended secondary structure in the region around Cys 180 , as observed in the crystal structure of rhodopsin (Fig. 3). These elements suggested that rhodopsin could be used as a template to model not only the transmembrane regions but also the E2 loop. The best homology model of the NK-1 receptor is shown in Fig. 4. The r.m.s. deviation between 296 aligned C ␣ positions of human NK-1 receptor and bovine rhodopsin is 1.7 Å. The model is consistent with mutagenesis analysis based on engineered zinc binding sites (33). The E2 loop adopts a central position on the extracellular face of the receptor, with residues 170 -183 forming a ␤-hairpin. The first ␤-strand dives down into the transmembrane domain, whereas the second ␤-strand is more external. The positions of the two photolabeled Met residues are shown in Fig. 4a. Met 181 and Met 174 lie in the inner and the outer strand of the ␤-hairpin, respectively. In bovine rhodopsin, the inner strand is part of the retinal binding pocket. Interestingly, the two photolabeled residues in the NK-1 receptor, Met 174 and Met 181 , correspond to two residues in rhodopsin (Glu 181 and Gly 188 , respectively), which are part of the retinal binding pocket. In the homology model, the side chain of Met 181 of the NK-1 receptor is buried in the core of transmembrane helix bundle and close to the position occupied by the polyene chain of retinal in rhodopsin. The side chain of Met 174 is also buried, but its more peripheral position on the extracellular face could make it more accessible to a photoreactive probe. The two Met residues are quite close with a distance of about 6 Å between C ⑀ atoms. However the relative inaccessibility of Met residues in the model led us to search for alternative conformations of the E2 loop not based on homology modeling.
Conformational Properties of the Isolated E2 Loop in Solution-To investigate the conformational preferences of the amino acid sequence of the E2 loop, we examined the solution conformation of a synthetic peptide corresponding to residues 168 -196 of the human NK-1 receptor (herein termed E2 peptide). The chosen sequence encompasses the entire E2 loop plus residues in the first turn of transmembrane helix H5. The Cys 180 residue was replaced by a serine to prevent oxidation and aggregation of the peptide in solution. The structure of the peptide was examined in water, in mixed solvents obtained by addition of variable amounts of trifluoroethanol (TFE), and in an aqueous solution of SDS micelles to mimic a water-membrane interface. Complete sequence-specific 1 H assignments were obtained using conventional homonuclear two-dimensional experiments. Two sets of spin systems were observed in the segment from Trp 184 to Asn 189 because of proline isomerization. In the major form (ϳ90%), the observation of strong ing that the C-terminal helix is more stable in the presence of SDS. Because the NMR spectra recorded in 25% TFE exhibited the highest quality in terms of chemical shift dispersion and proton line width, we used these solvent conditions to determine the three-dimensional structure of the peptide. The observation of strong sequential HN-HN NOEs together with medium- range HN i -HN NOE connectivities confirms the presence of helix structures in the central portion (residues 176 -182) and the C-terminal extremity (residues 190 -195). Furthermore, the amide protons of residues 178, 179, and 183 and residues 191-196 are in slow exchange with solvent, indicating that they are engaged in hydrogen bonds. Residues 176, 177, and 180 have weak 3 J HN-H ␣ coupling constants, as expected for helical conformations, but other residues exhibit coupling constants greater than 6 Hz, suggesting local deformations of the helical segments. The N-terminal extremity and the segment connecting the two helices have no propensity for a preferred secondary structure. A total of 281 interproton distance restraints was determined from the analysis of NOESY spectra, 83 of which were medium-range correlations. A family of 20 structures is shown in Fig. 5. The backbone r.m.s. deviation calculated for all residues is large (5.5 Å), indicating that the overall conformation is not well defined. This is mostly due to a poor definition of residues 168 -175 in the N-terminal extremity and residues 183-189 in the segment connecting the two helices (backbone r.m.s. deviations of 2.9 and 1.6 Å, respectively). Conversely, residues 176 -182 and 190 -196 adopt well defined helical conformations (backbone r.m.s. deviations of 0.3 Å).
"NMR-based" Model of the E2 Loop in the NK-1 Receptor-The NMR study indicates that two regions of the 168 -196 fragment of the NK-1 receptor have some helical propensity. The observed helical conformation of the peptide C-terminal extremity is in agreement with the sequence localization in the receptor, as it is expected to be part of transmembrane helix H5. To investigate whether the conformation of the E2 loop observed in solution could be accommodated in the context of the full receptor, we modeled the structure of the E2 loop based on the NMR study. The experimental restraints determined for the isolated E2 loop peptide in solution were incorporated into the homology model of the NK-1 receptor. The structure was submitted to simulated annealing to satisfy the new set of restraints in the E2 loop, with the position of transmembrane helices being held. Our analysis of 30 calculated structures indicates that the E2 loop can adopt two topologies within the receptor differing by the orientation of their central helix (Fig.  4). In the structure seen in Fig. 4b, this helix lies in the center FIG. 3. Sequence alignment of the E2 loops of human NK-1 receptor and bovine rhodopsin. The conformation of each residue in the crystal structure of bovine rhodopsin is indicated above the rhodopsin sequence (e, extended; h, helical conformation). The consensus secondary structure predicted by several methods (npsa-pbil.ibcp.fr), and the solution conformation observed for the synthetic E2 peptide are indicated below the NK-1 receptor sequence.
of the receptor whereas in the alternative conformation in Fig.  4c it is more peripheral. The positions of residues Met 174 and Met 181 are shown in Fig. 4. In the NMR-based conformations seen in Fig. 4b, the two Met residues are more accessible than in the homology model. The photolabeling data support the model in Fig. 4b rather than in Fig. 4c because both Met residues are accessible and in close proximity. DISCUSSION Affinity labeling is a powerful procedure used to establish the spatial proximity between photolabile residues within a ligand and its receptor. The time-consuming step in this strategy is the identification of the residue, or fragment of residues, in the protein that is covalently linked to the photoactivatable ligand. A combination of different strategies is indeed required to identify the site of covalent attachment, i.e. enzymatic and/or chemical digestion(s), immunoprecipitation with antibodies against specific domains of the receptor combined with either HPLC separation or SDS-PAGE analysis, and finally in some cases radiochemical sequencing. With G-protein-coupled receptors this procedure is further hampered by the hydrophobicity of these proteins and their high tendency to aggregate, which precludes working on pure or isolated receptors. Therefore, the benefit of our strategy is to bypass most of the isolation procedures, allowing the direct identification of a fragment or a residue as the site of photoinsertion without the need of radioactivity.
In this study, the use of [BAPA 0 ,(p-Bz)Phe 8 ]SP and [BAPA-Lys 6 ,(p-Bz)Phe 8 ,Pro 9 ,Met(O 2 ) 11 ]SP- (7)(8)(9)(10)(11) led to the unambiguous identification of the tripeptide Thr 173 -Met-Pro 175 in the second extracellular loop of both rat and human NK-1 receptors as the site of interaction of the photoactivatable amino acid p-benzoyl-L-phenylalanine, incorporated in position 8 of both SP analogs. All of these experiments were performed more than 10 times and led unambiguously to this fragment of the NK-1 receptor, whatever the NK-1 receptor species or the photoactivatable SP analog used. The data do not pinpoint any new anchoring point for substance P in the NK-1 receptor. However they indicate that two different photoreactive peptide analogs that differ in their pharmacological profiles and are activatable at the same position (8) in their sequence do interact with the same domain of the NK-1 receptor, whether rat or human species. We had previously established, using cyanogen bromide cleavage and MALDI-TOF analysis, that the covalent attachment is on the methyl of the Met 174 side chain (15). A steric factor induced by this modification on the adjacent methionine probably interrupts herein the cleavage at Pro 175 by carboxypeptidase Y, because digestion of a model peptide, ACTH- , shows that removal of a proline by carboxypeptidase Y may indeed occur, as demonstrated by MALDI-TOF analysis (data not shown).
These two photoreactive analogs were designed to screen the two binding sites described for the NK-1 receptor. The results obtained show that these two analogs interact with the same sequence of the NK-1 receptor within the E2 loop. However the two binding sites of the NK-1 receptor might still be different conformations or protein isoforms.
In all of these experiments we always found Met 174 as the site of photoinsertion, whereas Kage et al. (16)  two-to-one ratio (13). Using [[ 125 I](p-Bz)Phe 7 ]NKA, the sequence 178 -190 is labeled, Met 181 being the amino acid that is attached covalently to the radioactive probe. These results deserve two comments. The first one concerns the relative selectivity of [BAPA-Lys 6 ,(p-Bz)Phe 8 ,Pro 9 ,Met(O 2 ) 11 ]SP-(7-11) (this study) or [[ 125 I](p-Bz)Phe 7 ]NKA (13) to address the question of the molecular recognition by the two binding sites associated with the NK-1 receptor. It should be mentioned that in this study as well as in the one reported by Bremer et al. (13), it is difficult to prove that the photoactivatable probes, [BAPA-Lys 6 ,(p-Bz)Phe 8 ,Pro 9 ,Met(O 2 ) 11 ]SP- (7)(8)(9)(10)(11) or [[ 125 I](p-Bz)-Phe 7 ]NKA, are selective enough to bind only to the less abundant binding site associated with the NK-1 receptor. Indeed, the factor of selectivity of these probes is only 30 to 50 in favor of the less populated receptor conformation, which is 8 to 10 times less abundant than the major binding site NK-1M, whatever the clone expressing either high or low levels of NK-1 receptors (data not shown). Even though both photoactivatable probes were used at concentrations close to their respective affinity (K i ) for the minor binding site NK-1m, this lack of selectivity casts doubt on the labeling by [BAPA-Lys 6 ,(p-Bz)Phe 8 ,Pro 9 ,Met(O 2 ) 11 ]SP- (7)(8)(9)(10)(11) and [BAPA 0 ,(p-Bz)Phe 8 ]SP of the same residue in the minor NK-1m and major NK-1M binding sites. Even though the radical on the methionine side chain may be stabilized by the sulfur atom, it should also be noted that p-benzoylphenylalanine has been shown to be linked covalently to amino acids other than methionine in photolabeling studies with different peptide families (36 -41). The second point concerns the difference in the methionine identified in the two studies, Met 174 (Ref. 14 and this study) versus Met 181 (13,16). This difference does not seem to be related to either the species or the photoactivatable analogs used. Indeed, we always identified Met 174 in whatever human or rat receptor was studied. The photoactivatable analogs used, [BAPA 0 ,(p-Bz)Phe 8 ]SP (Ref. 14 and herein) or [[ 125 I-BH](p-Bz)Phe 8 ]SP (13,16) differ slightly only in the N-terminal part, which is not important for the recognition process (42). Even with [BAPA 0 ,(p-Bz)Phe 8 ,Pro 9 ]SP, a photoactivatable probe modified at a position adjacent to the photoactivatable amino acid, Met 174 , was still identified as the site of photoinsertion (14). The only major difference in the two studies is the concentration ratio between the photoprobe and the receptor. Although it is rather difficult to strictly determine this ratio in the study using [[ 125 I-BH](p-Bz)Phe 8 ]SP (13,16), it can be speculated that the ratio is below the one used in our studies with [BAPA 0 ,(p-Bz)Phe 8 ]SP, which is around 10 to 15. Furthermore, although in their first study Boyd and colleagues (16) identified only Met 181 as the photoinsertion site, both Met 181 and Met 174 were reported in their more recent work (13). Thus, slight variations in the ratio between the photoprobe and the receptor might lead to the preferential detection of one methionine versus the other, as different conformational states of the receptor might be trapped during the photolabeling process. Finally, as we suggested initially (14), Met 174 and Met 181 are spatially very close (this article and Ref. 46), which could explain why both residues may be photolabeled. This latter point is confirmed by the mutants M174A and M181A of the NK-1 receptor, i.e. when one methionine is mutated, the other one serves as the photoinsertion site and vice versa (46).
The identification of Met 174 or Met 181 as sites of covalent attachment of photoreactive agonists indicates a spatial proximity to residue Phe 8 of substance P in the binding site. Mutagenesis data further support a functional role for the E2 loop in tachykinin ligand recognition. For instance mutations at positions 193 and 195, at the junction of the E2 loop and helix H5 affect the binding of neurokinin A but not substance P (43). Interestingly, Ciucci et al. (44) reported that the mutation of Gly 166 at the junction of helix H4 and the E2 loop induces a change in tachykinin ligand selectivity and alters the conformation of the receptor. Furthermore, the binding of substance P is abolished by reducing agents, indicating that the Cys 105 -Cys 180 disulfide bridge is of major importance in maintaining the conformation of the receptor and of the E2 loop in particular (34).
The conformation of the E2 loop has been modeled on the hypothesis of its structural similarity with the E2 loop of bovine rhodopsin. In the calculated homology model, the two Met residues are in close proximity, ϳ6 Å. However the side chains of Met 174 and Met 181 are not likely to be directly accessible to photoreactive probes. This accessibility may depend on the activation-deactivation state of the receptor, and/or the desensitization process. Importantly, the template used for homology modeling is the structure of rhodopsin covalently bound to 11-cis retinal, which therefore corresponds to an inactive state. It seems likely that the homology model of the NK-1 receptor also corresponds to an inactive state. The conformational changes that occur upon agonist binding probably involve movements of transmembrane helices (45), but so far very little is known about the putative conformational changes of transmembrane helices in their outer part or of extracellular loops.
In an attempt to model alternative conformations of the E2 loop, we have analyzed the structural preferences of the amino acid sequence of the E2 loop. NMR studies of a synthetic fragment 168 -196 of human NK-1 receptor showed that the region 176 -182 has a helix propensity in the presence of TFE or SDS. These solvents were used to stabilize secondary structures and mimic the water-membrane interface. MacDonald et al. (46) have applied a similar approach in a synthetic fragment 162-198 of rat NK-1 receptor in phospholipids vesicles. Despite different peptide sequences and solvent conditions, the two structures are very similar. The conformational preferences of the E2 loop in solution were used to generate other models in which the Met residues are more accessible. The structure in Fig. 4b is very similar to the model described by Pellegrini et al. (47) and is in good agreement with photolabeling data.
Two strategies have been used herein to model the E2 loop of the NK-1 receptor, one based on sequence homology to rhodopsin and the other based on the secondary structure of the E2 loop in solution. Homology modeling of the E2 loop yields interesting information, as it indicates that the structure of the extracellular domain of the NK-1 receptor is not completely identical to the one of rhodopsin in the dark state. So far, very few biophysical studies have been carried out to analyze the conformational changes in the extracellular part of G-proteincoupled receptors that occur during activation. Therefore the proposed conformations of the E2 loop remain working models. What can be concluded here from the modeling data is that they are in good agreement with results obtained from the photolabeling study. In the homology model as well as in the NMR-based model, Met 174 and Met 181 are spatially very close, which might explain why these two amino acids can be photolabeled. In the NMR-based model, the two Met residues are well positioned to be cross-linked by photoreactive analogs of substance P. In our view, it will be necessary to identify more anchoring points from photolabeling to go further into the characterization of the interaction between SP and the NK-1 receptor. However, the question of the dynamics within the ligand-receptor complex or during the recognition process will probably limit this approach. Therefore, complementary strategies to mutagenesis and photolabeling studies will be required.