Antibody Imprint of a Membrane Protein Surface

Structural features of the integral membrane protein flavocytochrome b (Cyt b) were discovered using an antibody “imprint” of the Cyt bsurface. Amino acid sequences were selected from a random nonapeptide phage-display library by their affinity for the monoclonal antibody 44.1 binding site, which recognizes the native conformation of the p22 subunit of Cyt b. Transferred nuclear Overhauser effect spectroscopy and rotating frame Overhauser effect spectroscopy NMR were used to study the antibody-bound conformation of a synthetic peptide derived from phage-displayed sequences. The NMR data supported the phage-display analysis suggesting the existence of a complex epitope and allowed the modeling of the close spatial proximity of the epitope components 29TAGRF33 and183PQVNPI188 from discontinuous regions of p22. Although these regions are separated by two putative membrane-spanning domains and are 150 residues apart in the sequence, they appear to combine to form a complex epitope on the cytosolic surface of the transmembrane protein. NMR constraints, measured from the antibody-bound conformation of a composite peptide mimetic of the Cyt b epitope, and one constraint inferred from the phage-display results, were used to demonstrate the close proximity of these two regions. This information provides a low resolution view of the tertiary structure of the native discontinuous epitope on the Cytb surface. Given additional antibodies, such imprint analysis has the potential for producing structural constraints to help support molecular modeling of this and other low abundance or noncrystallizable proteins.

Structural features of the integral membrane protein flavocytochrome b (Cyt b) were discovered using an antibody "imprint" of the Cyt b surface. Amino acid sequences were selected from a random nonapeptide phage-display library by their affinity for the monoclonal antibody 44.1 binding site, which recognizes the native conformation of the p22 phox subunit of Cyt b. Transferred nuclear Overhauser effect spectroscopy and rotating frame Overhauser effect spectroscopy NMR were used to study the antibody-bound conformation of a synthetic peptide derived from phage-displayed sequences. The NMR data supported the phagedisplay analysis suggesting the existence of a complex epitope and allowed the modeling of the close spatial proximity of the epitope components 29 TAGRF 33 and 183 PQVNPI 188 from discontinuous regions of p22 phox . Although these regions are separated by two putative membrane-spanning domains and are 150 residues apart in the sequence, they appear to combine to form a complex epitope on the cytosolic surface of the transmembrane protein. NMR constraints, measured from the antibody-bound conformation of a composite peptide mimetic of the Cyt b epitope, and one constraint inferred from the phage-display results, were used to demonstrate the close proximity of these two regions. This information provides a low resolution view of the tertiary structure of the native discontinuous epitope on the Cyt b surface. Given additional antibodies, such imprint analysis has the potential for producing structural constraints to help support molecular modeling of this and other low abundance or noncrystallizable proteins.
Human phagocyte flavocytochrome b (Cyt b) 1 is an electron transferase that directs metabolic electrons across the plasma membrane to reduce molecular oxygen and form superoxide (O 2 . ). The production of superoxide by phagocytes, which involves modulation of Cyt b structure, is essential for antimicrobial host defense (1), plays a central role in inflammatory tissue injury (2), and may play a more general role in cellular regulation (3). Individuals with a defective Cyt b have insufficient superoxide production, suffer from chronic granulomatous disease (4), and sustain repeated life-threatening infections (5). Despite its cardinal role in many inflammatory processes and possibly in growth regulation, few experimental methods have been able to provide information on the structure of Cyt b. Therefore, we have developed an alternative method to describe topological features of the Cyt b surface. Because antibodies and their cognate antigens form complementary surfaces of interaction (6), we have sought structural information about Cyt b from an antibody binding site, specific for the Cyt b surface. We recently reported mapping of monoclonal antibody epitopes and functional sites of Cyt b, using random sequence phage-display peptide libraries (7;8). mAb 44.1 recognizes Cyt b in situ in permeabilized human neutrophils and thus contains information about the three-dimensional structure of the protein surface. Phage bound by mAb 44.1 display sequences that implicate the 183 PQVNPI 188 region of p22 phox as at least part of the epitope bound by mAb 44.1 (7). Our current results, obtained using a more sensitive colony-lift method of immunological screening of phage sequences, suggest that mAb 44.1 also recognizes the 29 TAGRF 33 segment of p22 phox . By including both peptide segments into one contiguous synthetic construct separated by flexible glycine residues, as suggested by phage mapping, we were able to determine the conformation of the bound structure at low resolution by transferred NOESY and transferred ROESY NMR. Our results show the combined application of phage-display random peptide epitope mapping and NMR analysis of specific epitope peptides bound to the antibody recognition sites can provide new structural information about conformational epitopes of large proteins.

EXPERIMENTAL PROCEDURES
Colony-lift Screening-Colony-lift screening is the replica blotting of bacterial colonies onto nitrocellulose and developing the replica in a similar fashion as a Western blot. This procedure was carried out by infecting K91 cells with affinity-purified J404 phage library members and growing them as isolated colonies on Luria broth/kanamycin agar. The colonies were then blotted to nitrocellulose for 5 min, the blots were rinsed in phosphate-buffered saline (pH 7.4), and probed with 2 g/ml mAb 44.1 for 2 h at room temperature, washed, and developed with secondary antibody and chromogen (9). Signals on the blots were used to locate corresponding colonies that were used to inoculate individual cultures for nucleotide sequence analysis as described (10).
Mutant Phage Construction-A mutant phage, which expressed the ATAGRFGGG peptide sequence on the amino terminus of the pIII protein, was constructed by using two complementary oligonucleotides, 5Ј-ctctcactccgctaccgcgggccgttttggcggcggcgtggaaagttgt-3Ј, and 5Ј-ctttc-  1 The abbreviations used are: Cyt b, flavocytochrome b; mAb, monoclonal antibody; Tr-NOESY, transferred nuclear Overhauser effect spectroscopy; Tr-ROESY, transferred rotating frame Overhauser effect spectroscopy; ELISA, enzyme-linked immunosorbent assay; phox, phagocyte oxidase. cacgccgccgccaaaacggcccgcggtagcggagtgagagtaga-3Ј. The oligonucleotides were phosphorylated and allowed to anneal under conditions previously described (11). The insert was then ligated to the M13KBst (12) vector in a molar ratio of 1:2.5 for vector:insert, respectively. The ligation product was then transfected into electrocompetent MC1061 cells and plated on a lawn of K91 cells to produce isolated plaques. The relevant nucleotide sequence of resulting clones was determined (13) to verify the intended mutations.
ELISA-ELISA analysis was used to determine the ability of synthetic peptides to mimic the putative epitope regions of p22 phox when bound by mAb 44.1. Seventy-five l of a wheat germ agglutinin:Nacetylglucosamine solution (0.4 mg/ml:2 mM) was adsorbed to a Dynex Immulon 2 plate for 3 h at 4°C. Unbound protein was removed with four rinses of wash buffer (2.7 mM KCl, 137 mM NaCl, 1.5 mM KH 2 PO 4 , 8 mM Na 2 HPO 4 , 0.9 mM CaCl 2 , 0.5 mM MgCl 2 (pH 7.4)). Human neutrophil membranes served as a source of Cyt b, which were extracted with octyl glucoside according to a previous protocol (14), except the membrane resuspension buffer lacked dithiothreitol and EDTA. Thirty l of octyl glucoside extract (containing ϳ5 ng of Cyt b) was then added to each rinsed well and incubated overnight at 4°C, prior to washing as above, except 0.1% Triton X-100 was included in the wash. Seventy-five l of mAb 44.1 (final concentration 200 ng/ml) was then exposed to the immobilized Cyt b in the presence of various dilutions of synthetic peptides as shown in Fig. 2. The antibody remaining after four rinses of wash buffer containing 0.1% Triton X-100 was measured with an antimouse antibody conjugated with horseradish peroxidase. The signals were developed with a standard ABTS-hydrogen peroxide reagent and measured at 405 nm.
NMR Analysis-NMR spectra were obtained on a Bruker DRX 500 instrument at 5 and 25°C, in 50 mM phosphate buffer (pH 6.5 or 5.0), in the presence of 5% D 2 O, as described in the legend to Fig. 3. The TAGRFGGGQVGPP peptide was 1 mM, and the mAb 44.1 was either 47 or 84 M. The interproton distances in the peptide complexed with mAb 44.1 were estimated by comparison with the cross-peak intensity of the F5:H␤ protons, which are 1.8 Å apart.

RESULTS
To obtain structural imprint information about Cyt b from mAb 44.1, we extended phage-display epitope mapping of the mAb recognition site (7). From a pool of mAb 44.1-immunoreactive phage-displayed peptides, colony-lift screening (see "Experimental Procedures") was used to select avidly binding sequences that most closely resemble the epitope bound by mAb 44.1. Of approximately 500 colonies of Escherichia coli infected by different mAb 44.1-selected phage clones, 22 (4%) were found to produce distinctly darker staining replica signals (represented by signals in Fig. 1A; I ϭ darker, and II ϭ lighter). The unique peptide displayed on phage associated with each colony type was deduced by nucleotide sequence analysis. Peptides expressed by the light-staining II-type phage were found to belong to the PX 1 VX 2 P motif (where X 2 R if X 1 ϭ Q) and contained from two to four residues matching the 183 PQVNPI 188 sequence of p22 phox , as shown previously (7). Peptides associated with the more strongly staining I-type colonies segregated into two groups according to their sequences (Table I). Sequences 1-3 in Table I (AQPQVRPIG, NMPQVRPID, and DRPQVRPIL) each contained the six-residue sequence, PQVRPI, 2 matching five of six residues of the 183 PQVNPI 188 epitope. The second group of phage-expressed peptides giving strongly staining colony-lift signals (sequences 4 -7 in Table I) included 19 clones bearing four unique sequences and suggested the new consensus, TAGRF-GGGQV(GPP) (sequence 10 in Table I). The GPP tether residues in this sequence are not part of the variable region of these clones (12), but are shown because they may also be recognized in part by mAb 44.1. This new consensus epitope contains the TAGRF and GGGQV(GPP) sequence segments, which are similar to the 29 TAGRF 33 and 183 PQVNPI 188 regions of p22 phox , respectively (Table I). Two of these individual phage peptides: 4, EGRFGGGQV(GPP) and 5, ISRFGGGQV(GPP), were identical in seven of the nine randomized positions and sequence 4 was found in 14 clones (64%) selected in this way.
The immunoreactivity of mAb 44.1 with phage-expressed peptides resembling each of the two different putative epitope moieties of p22 phox was supported using semiquantitative Western blots probed by mAb 44.1. All lanes of the Western blot in Fig. 1B contain ϳ10 10 phage plaque-forming units differing only by the sequence of their displayed peptides. In lanes b and c, strong signals were observed at the molecular weight corresponding to the phage pIII protein bearing either the AQPQVRPIG or EGRFGGGQV(GPP) peptides (1 and 4, respectively, Table I). These sequences were derived from darkly staining type I colonies. Phage displaying peptides that mimicked the putative 29 TAGRF 33 epitope moiety alone were not selected from the library, presumably because this sequence has relatively low affinity for mAb 44.1 when it occurs without the PQVNPI region of the epitope. To determine whether mAb 44.1 can indeed recognize the TAGRF sequence expressed on the pIII protein of the M13 phage, a clone displaying the ATAGRFGGG sequence (sequence 15, Table I) (7) were used to infect 1 ml of mid-log phase K91 cells. Dilutions of the infected cells were performed to obtain isolated colonies when plated on Luria broth agar containing kanamycin at 75 g/ml. The colonies resulting after overnight incubation at 37°C were blotted on nitrocellulose discs and probed with mAb 44.1, producing signals of variable intensity that corresponded to the immunoreactivity of the phage-displayed peptides. I and II represent strong and light-staining classes of colonies, respectively. B, Western blot analysis of putative epitope regions recognized by mAb 44.1. To determine whether mAb 44.1 was able to recognize the epitope-mimicking regions expressed on the pIII protein of phage, 100 ml of Luria broth with 75 g/ml kanamycin was inoculated with K91 cells infected with phage expressing specific peptide sequences. Phage were harvested, the proteins were separated by SDS-polyacrylamide gel electrophoresis, and Western blots were probed with mAb 44.1 as described previously (7) The ability of phage-discovered peptides to substitute for the native cytochrome b epitope was quantitated by their ability to compete for the antibody combining site in an ELISA shown in Fig. 2. Component peptides examined in ELISA analyses (ac-PQVNPI-amide and ac-ATAGRFTQW-amide, 11 and 12, respectively, in Table I) were identical to the 183 PQVNPI 188 and 28 ATAGRFTQW 36 regions of p22 phox . The composite peptide (ac-ATAGRFGGPQVNPI-amide, sequence 9, Table I) was identical to a combination of these two regions, representing both in a contiguous linear sequence (Table I). The consensus peptide sequence 10 (ac-TAGRFGGGQVGPP-amide), suggested by colony-lift-selected phage clones 4 -7, was tested as was the ac-PQVRPI-amide sequence (sequence 8) that was strongly represented by clones 1-3. The peptides tested all demonstrated well behaved competition curves suggestive of specific interaction up through the millimolar concentrations needed for the NMR experiments described below. The half-maximal inhibitory concentrations of these peptides in the ELISA are listed in Table I. The composite sequence of p22 phox ATAGRFGGPQVNPI (sequence 9 in Table I), representing both epitope regions, was greater than 100-fold more inhibitory toward mAb 44.1 binding to Cyt b than either half of the epitope (peptides 11 and 12 in Table I). The strikingly lower IC 50 for the composite peptide relative to the component epitope peptides strongly supports the participation of both regions of p22 phox in the binding site for mAb 44.1 on Cyt b. Interestingly, the non-native peptide sequence PQVRPI, where arginine replaces the native asparagine 186, demonstrated an even higher (580 nM) avidity for the mAb, indicating the importance of arginine in this region of the epitope.
The consensus peptide (TAGRFGGGQVGPP, 10 in Table I) from phage-display has a binding affinity (K d ϭ 150 M) for mAb 44.1, which is in a suitable range for study of the antibody-bound conformation of the peptide by transferred-NOESY (Tr-NOESY) NMR (15,16). The free peptide evidently has little or no persistent structure in solution in the absence of the antibody, as evidenced by two-dimensional NOESY NMR shown in part in Fig. 3A, showing cross-peaks only within amino acid side chains and between nearest neighbor residues. Upon addition of antibody, the bound peptide develops NOESY signals while it is bound, which persist when the peptide exchanges into solution, as shown in the section of the Tr-NOESY NMR spectrum in Fig. 3B at pH 6.5. The intensity of the Tr-NOESY signals depends on the amount of antibody added and the fraction of peptide bound to the antibody, supporting specificity of binding at the high peptide concentrations needed (data not shown). In all, 68 NMR constraints were observed in the bound peptide, eight of which are long range, involving residues separated in the primary structure by three to four other residues (Table II). Of the remaining 60 NOEs (data not shown), 21 involved adjacent residues and 39 were within residue NOEs, many of which provided information on side chain conformation.
Upon addition of antibody 44.1, the Tr-NOESY shows increases in cross-peak intensities involving residues Arg 4 , Phe 5 , Gln 9 , and Val 10 , indicating that these residues are immobilized upon binding to the antibody. These same residues, corre-  sponding to 32 RF 33 and 184 QV 185 of p22 phox , are abundantly represented in the phage-displayed peptides 1-7 in Table I. Fig. 3B and Table II show long range cross-peaks among im-mobilized residues, which are widely separated in the primary sequence of the peptide, supporting a folded conformation of the peptide when bound to mAb 44.1. Long range cross-peaks, which indicate the association between F5:ring and V10:H ␥ protons in the peptide when bound to mAb 44.1, are denoted with an asterisk in the Tr-NOESY spectrum (Fig. 3B). Within the resolution limits of the Tr-NOESY experiments, no medium range i to i ϩ 2, or i to i ϩ 3 were observed (data not shown), arguing against helices or tight turns in the modeled structure. The short range i to i ϩ 1 cross-peaks were included in simulated annealing calculations modeling the bound structure that suggested a folded peptide with a broad open bend centered around Gly 7 and Gly 8 and a close proximity (3-5 Å) between Phe 5 and Val 10 . When the additional phage display constraint suggesting juxtaposition of Asn 186 and Arg 32 side chains of p22 phox was included in the calculation, the peptide maintained its general folded conformation with minimal structural violations. A higher resolution model of this epitope will be the subject of a future report. FIG. 3. A, a portion of the two-dimensional NOESY NMR spectrum of the free TAGRFGGGQVGPP synthetic peptide. B, a portion of the Tr-NOESY NMR spectrum of the peptide in exchange with the mAb 44.1-bound form at pH 6.5 and 5°C. C, a portion of the two-dimensional Tr-ROESY NMR spectrum of the peptide in exchange with mAb 44.1-bound form at pH 6.5 and 5°C. Spectra A and B were collected using a standard two-dimensional NOESY pulse sequence with a watergate solvent suppression and a 300 ms mixing time. Spectrum C was collected using two-dimensional ROESY pulse sequence with 150-ms spin lock of 2.3 KHz and watergate solvent suppression. The spectra were processed to give 2K ϫ 2K real points and are displayed at identical contour levels. In spectrum A of the free peptide, only cross-peaks of protons within an amino acid or between adjacent residues are observed and are labeled on the figure. In spectrum B, new cross-peaks are observed and labeled and arise from the peptide binding to the antibody. In spectrum C, magnetized transfers mediated by other protons subtract from the cross-peak intensities in the Tr-ROESY NMR experiments, as described in the text. The long range NOEs that support the bound conformation of the folding of the peptide onto itself are weakened compared with the corresponding peaks the Tr-NOESY, but are still present in the Tr-ROESY (spectrum C), and so these cross-peaks largely result from direct cross-relaxation between the bound peptide protons as discussed under "Results." a Indicates that pseudoatoms were used in the structure determination, since the two or three protons at the sites denoted were not resolved or not resolved separately.
When ligands are bound to macromolecules it is possible that Tr-NOESY cross-peaks between ligand protons can be mediated by macromolecule protons that are in intimate contact with the ligand (17)(18)(19)(20). Rotating frame Overhauser enhancement (Tr-ROESY) experiments can be used to distinguish direct ligand-ligand proton interactions (two spin interactions) from interactions that are mediated indirectly by third protons (three spin interactions), since the three spin contributions have opposite signs in NOESY and ROESY (21). Cross-peaks indicating long range interactions between peptide protons in Tr-NOESY (Fig. 3B) are weaker, but still present in Tr-ROESY (Fig. 3C), and so the major portion of the Tr-NOESY intensity is due to direct interaction between ligand protons.
The Tr-NOESY cross-peak intensities observed are due to a sum of the bound peptide and free peptide signals when the exchange rate of the ligand off the macromolecule is fast, relative to the fastest cross-relaxation rate of the bound ligand (22). At pH 6.5 and 5°C the consensus peptide was found to be in relatively slow exchange by the spin-lock power dependence of the T1 relaxation (23), and the one-dimensional NMR spectrum of the peptide was not substantially broadened by kinetic exchange with the antibody (data not shown). To increase the exchange rate of the peptide with the antibody, the pH was lowered to 5, and the temperature was raised to 25°C. The proton NMR line widths of the free peptide (Fig. 4A) increased in the presence of the antibody (Fig. 4B), indicating faster exchange of the peptide with the antibody binding site. The shifts of peptide peaks upon binding to the antibody and the increase in line width indicate that the approximate off rate is about 70 s Ϫ1 at pH 5 and 25°C, which is over 10 times faster than the fastest bound cross-relaxation rate (ϳ5.4 s Ϫ1 ). All the cross-peaks previously identified at pH 6.5 are present under the faster exchange rate conditions. DISCUSSION Antibodies that bind native proteins contain topological information about the protein surface structure, in the form of a three-dimensional imprint. Because many epitopes bound by antibodies involve residues that are surface-accessible, but not necessarily contiguous in the polypeptide chain, their resolution can reveal important structural data about the protein.
Phage-display peptide libraries have proven a useful source of epitope analogs (12) and therefore provide a tool in antibody binding site analysis. The ability to apply NMR to confirm structures of phage-derived amino acid sequences, when bound by the antibody, enables structural models of protein epitopes to be constructed.
Our colony-lift data suggested two widely separated regions of the p22 phox polypeptide might both participate in the epitope bound by mAb 44.1, implying the antibody recognizes an epitope in the folded protein. The mAb 44.1 binds to Cyt b in permeabilized cells (7) and so recognizes the native Cyt b structure. The mAb 44.1 does not inhibit oxidase function (data not shown) and so does not perturb its structure in functionally important ways. ELISA analysis was used to measure the ability of synthetic peptides to compete with mAb 44.1 binding to octyl glucoside-solubilized Cyt b. This detergent-solubilized form of Cyt b retains its native heme spectrum (24) and can be reconstituted with exogenous FAD to produce a functional protein capable of generating superoxide by itself (25) or in complex with other neutrophil proteins (26). Our ELISA data provide evidence for synergistic binding affinities of mAb 44.1 to the composite peptide sequence (ATAGRFGGPQVNPI) relative to the component sequences (ATAGRFTQW and PQVNPI), supporting the existence of both protein segments in the mAb 44.1 epitope (Fig. 5).
Tr-NOESY (16) (15,27,28) and Tr-ROESY (29) NMR experiments were performed on the consensus TAGRFGGGQVGPP (sequence 10) peptide to study its molecular conformation when bound to mAb 44.1 and to test conclusions drawn from the phage-display analysis. Tr-NOESY analysis could not be carried out on peptide 9 (ATAGRFGGPQVNPI), which more closely resembled both regions of Cyt b, because under the conditions investigated it was bound too tightly by mAb 44.1. The consensus peptide (peptide 10) showed both sufficient affinity and off-rate for NMR analysis, under the conditions used. The long range NOEs in Fig. 3 and in Table II indicate that upon binding to mAb 44.1, a globally folded structure is conferred upon the TAGRFGGGQVGPP peptide. The similarities between the Tr-NOESY and the Tr-ROESY (Fig. 3, B and C,   FIG. 4. A, one-dimensional spectrum of the 4.8 -0.5 ppm range of the TAGRFGGGQVGPP synthetic peptide at pH 5.0 and 25°C. B, onedimensional spectrum of the peptide in the same region of the spectrum under the same conditions, after the addition of mAb 44.1. Spectrum B shows broadening of the peptide signals due to exchange with the antibody. At pH 6.5 and 5°C, there was very little exchange broadening, and therefore the exchange rate of the peptide off the antibody is increased due to the lowering of the pH and increasing the temperature.  Table I), discovered by phage display and ELISA epitope analysis of mAb 44.1, can mimic the proposed p22 phox fold in A. By removing the underlined Ala 1 amino acid and replacing the underlined Pro 9 with Gly, Asn 12 with Gly, and Ile 14 with Pro to produce the consensus ac-TAGRF-GGGQVGPP-amide (peptide 10, Table I), the affinity of the antibody for the peptide was low enough to use Tr-NOESY NMR to measure the proximity of the FR and QVG segments. The NMR data support the folded conformation of the bound peptide and the close proximity between the two regions of p22 phox . Long range NOEs are represented by dotted lines between the residues indicated and are numbered to correspond to the respective NOEs listed in Table II. respectively) show that mediation of cross-relaxation between peptide protons by antibody protons is not the major source of cross-peak intensity in the peptide Tr-NOESY. A decrease in the Tr-ROESY cross-peak intensity between two protons relative to the TR-NOESY cross-peak is likely to be due to antibody protons mediating NOESY cross-relaxation of the peptide protons (29). The ability to adjust the off-rate of the peptide with the antibody by lowering the pH and increasing the temperature allows the adjustment of the off-rate (Fig. 4, A and B) so that a structure with additional reliability can be determined. Such studies are under way using a fluorine-substituted peptide to obtain more accurate off-rate measurements (30).
Our results imply that the ATAGRF segment of the consensus ATAGRFGGPQVNPI sequence is folded back upon the PQVNPI region to bring Phe 5 and Val 10 to less than 5 Å of one another, and hence place the positively charged arginine residue in close proximity to the carboxyl half of the peptide. The importance of an arginine in this vicinity would explain the repeated recovery of phage-expressed peptides containing an arginine in the consensus, PQVRPI, and the failure to obtain phage peptides containing an asparagine in this position, as it exists in the 183 PQVNPI 188 p22 phox segment (7). The importance of arginine in this region is further demonstrated by ELISA, in the 600-fold higher binding effectiveness of mAb 44.1 to PQVRPI compared with the PQVNPI peptide (Fig. 2). Because arginine is not found in the native 183 PQVNPI 188 segment of p22 phox , the occurrence of this residue in the epitope could well be satisfied by the arginine in the 29 TAGRF 33 region of the protein if it were located close to Asn 186 in the native folded protein, as suggested by the phage-display mapping. The Tr-NOESY and Tr-ROESY data on the TAGRFGGGQVGPP peptide bound to mAb 44.1 (Fig. 3B) corroborates this concept by indicating the arginine side chain of the bound peptide is oriented close to the QV residues, perhaps across a cleft of the folded peptide.
The potential to determine the approximate positions of noncontiguous amino acids on the surfaces of folded proteins appears to be a capability of peptide combinatorial libraries and thus a strength of phage-display peptide library analysis (31). The approach we present, where mAbs directed against surface structures of integral membrane proteins in situ are used to probe random peptide libraries and the mAb-bound structures of the selected peptide sequences are studied by Tr-NOESY and Tr-ROESY NMR, provides visualization of an imprint of the surface structure of the membrane protein antigen. Such imprint analysis may provide a new approach for studying the surfaces of proteins and thus aid in the modeling of intermolecular interactions for which traditional structural data are unavailable.
In summary, the data presented in this report indicate that regions of an integral membrane protein, separated by membrane-spanning domains, can be shown to be associated in an epitope that exists in the native form of the protein. Future applications of this approach promise to provide more detailed three-dimensional information about native conformations of integral membrane proteins.