Phage Display Epitope Mapping of Human Neutrophil Flavocytochromeb558

Despite extensive experimental and clinical evidence demonstrating the critical role of flavocytochromeb558 (Cyt b) in the NADPH-dependent oxidase, there is a paucity of direct structural data defining its topology in the phagocyte membrane. Unlike other Cyt b-specific monoclonal antibodies, 7D5 binds exclusively to an extracellular domain, and identification of its epitope should provide novel insight into the membrane topology of Cytb. To that end, we examined biochemical features of 7D5-Cytb binding and used the J404 phage display nonapeptide library to identify the bound epitope. 7D5 precipitated only heterodimeric gp91-p22 phox and not individual or denatured Cyt b subunits from detergent extracts of human neutrophils and promyelocytic leukemia cells (gp91-PLB). Moreover, 7D5 precipitated precursor gp65-p22 phox complexes from detergent extracts of the biosynthetically active gp91-PLB cells, demonstrating that complex carbohydrates were not required for epitope recognition. Epitope mimetics selected from the J404 phage display library by 7D5 demonstrated that 226RIVRG230 and160IKNP163 regions of gp91 phox were both bound by 7D5. These studies reveal specific information about Cytb membrane topology and structure, namely that gp91 phox residues 226RIVRG230 and160IKNP163 are closely juxtaposed on extracytoplasmic domains and that predicted helices containing residues Gly165–Ile190 and Ser200–Glu225 are adjacent to each other in the membrane.

The phagocyte NADPH oxidase is a plasma membrane redox system that produces superoxide anion (O 2 . ), an essential precursor for other reactive oxygen metabolites critical for oxygendependent microbicidal activity (1)(2)(3). A genetic lesion affecting any one of four of the oxidase components, gp91 phox , p22 phox , p47 phox , or p67 phox , results in defective oxidase activity and the inability of phagocytes to kill pathogenic microorganisms, a disorder clinically recognized as chronic granulomatous disease (CGD) 1 (2, 4 -9). Human neutrophil flavocytochrome b 558 (Cyt b) is a heme-containing, heterodimeric integral membrane protein composed of subunits gp91 phox and p22 phox (10). Cyt b is the electron transferase of the NADPH oxidase, relaying electrons from bound NADPH within the cell to an oxygen acceptor region of Cyt b on the exterior aspect of the cell membrane where O 2 . is formed. In this functional capacity, Cyt b has been established as an essential component of the respiratory burst oxidase, although little published experimental data describe its topology in the membrane. Determination of structural and functional aspects of epitopes bound by specific antibodies can provide information about the protein against which the antibody is directed (11). Antipeptide and antisubunit polyclonal antibodies against regions of Cyt b have been used to locate the corresponding epitopes (12,13) and information derived from identification of epitope mimetics has led to the description of anti-Cyt b "antibody imprints" (14). We continue to elucidate the structure of Cyt b domains recognized by monoclonal antibodies to better define its transmembrane topology and gain insight into its functional organization.
Reported epitope mapping data for monoclonal antibodies (mAbs) specific for Cyt b indicate that they bind cytosolic aspects of the protein (15)(16)(17)(18)(19). However, it has been reported that mAb 7D5 (19) binds an extracellular Cyt b epitope on intact neutrophils derived from normal but not CGD patients that lack Cyt b (20). Thus, 7D5 has proven useful in the determination of Cyt b up-regulation as an indicator of neutrophil activation and granule exocytosis (21) and for the identification of individuals deficient in Cyt b (20,22,23). However, neither the subunit location nor chemical nature of the 7D5 epitope has been elucidated.
In our current studies, we have used phage display and immunological analyses to identify the 7D5 epitope on Cyt b. Although we confirmed the inability of 7D5 to recognize Cyt b on immunoblots, we found that 7D5 immunoprecipitated detergent-solubilized Cyt b heterodimer containing its fully processed 91kDa form of gp91 phox and its 65-kDa precursor. Additionally, it precipitated the deglycosylated gp91 phox core protein, 2 suggesting that neither the mature nor high man-nose-containing carbohydrate contributes significantly to the epitope. Furthermore, our results indicate that the 226 RIVRG 230 and 160 IKNP 163 segments of gp91 phox form the 7D5 epitope and therefore must be exposed on the cell surface. These sequences of gp91 phox were not bound by 7D5 in the absence of p22 phox nor under conditions that disrupted the heterodimer, suggesting that, although 7D5 binding is confined to nonlinear but contiguous regions of gp91 phox , it depends on associated p22 phox for its conformational integrity. In combination, these data provide direct evidence for the identity of two adjacent transmembrane helices in gp91 phox .

EXPERIMENTAL PROCEDURES
Chemicals, Reagents, and Materials-Prestained protein molecular weight standards were purchased from Life Technologies, Inc. Reagents for buffers and bacteriological media and Nunc Maxisorb flat bottom plates for ELISA were purchased from Fisher. Cyanogen bromideactivated Sepharose CL-4B and GammaBind Sepharose were purchased from Amersham Pharmacia Biotech. Sequencing data were obtained using a Sequenase version 2.0 sequencing kit purchased from U.S. Biochemical Corp. Unless specified, all other reagents were purchased from Sigma.
Neutrophil Isolation and Flow Cytometry-Heparinized, venous blood was obtained from healthy individuals (or from patients with CGD) in accordance with a protocol approved by the Institutional Review Board for Human Subjects at the University of Iowa, and neutrophils were isolated as described previously using Hypaque-Ficoll gradients after dextran sedimentation (24). Genetic analyses of the individuals with either X-linked deficiency of gp91 phox or autosomal deficiency of p22 phox were performed by Paul G. Heyworth (The Scripps Research Institute, La Jolla, CA) and John T. Curnutte (Genentech, Inc.). The individual with autosomal deficiency of p22 phox (A22 0 ) had a nucleotide replacement at position C 354 (C 354 3 A), which resulted in the replacement of Ser 118 with Arg. The individual with deficiency of gp91 phox (X91 0 ) had a frameshift in exon 11, resulting in the replacement of nine nucleotides with eight. Surface-expressed Cyt b was detected using 7D5 by flow cytometry as described previously (21). Samples were analyzed on a FACscan flow cytometer (Becton Dickinson, San Jose, CA) at the University of Iowa Core Flow Cytometry facility. For most experiments, unfixed neutrophils were used, and a single live gate eliminated debris and contaminating cells. However, when patient cells were shipped overnight, it was sometimes necessary to use propidium iodide staining in combination with an additional gate to further exclude dead cells from the analysis.
Cyt b ELISA-35 l of relax buffer (10 mM Hepes, 100 mM KCl, 10 mM NaCl, pH 7.4) containing 8 pmol of heparin-purified Cyt b (determined by spectral absorbance at 414 nm, extension coefficient ϭ 21.6 mM Ϫ1 cm Ϫ1 ) (25), in 2% octyl glucoside was used to coat each well of a 96-well Corning Maxisorb ELISA plate overnight at 4°C. Rinsing and blocking were performed as described (14), except a different blocking buffer (Hanks' balanced salt buffer with 10 mM HEPES, pH 7.4, 0.5% bovine serum albumin) was used. Cyt b was probed for 1 h at 25°C with 80 l of the indicated mAb at a concentration of 3 g/ml. To measure the ability of the intact phage clones to block the binding of 7D5 to immobilized Cyt b, the antibody was diluted to 3 g/ml and exposed to 2.5 ϫ 10 12 plaque-forming units for 30 min 25°C. Pretreated 7D5 was then exposed to the immobilized Cyt b for 4 h. The reactivities of the mAbs for the Cyt b were determined by probing the wells with goat antimouse secondary antibody conjugated with horseradish peroxidase as described above.
Antibody and Phage Display Epitope Mapping-The generation of monoclonal antibody 7D5 has been described previously (19), and the production of the J404 nonapeptide phage display library was reported (28). 7D5 was purified from spent RPMI 1640 media of a hybridoma cell line using GammaBind Sepharose (Amersham Pharmacia Biotech) according to the manufacturer's instructions, and purity was assessed by SDS-PAGE. Mapping of 7D5 with the phage display library and plaque lift analyses were carried out as described (16), except the affinityselected clones were amplified following each round of selection by replicating as plaques on a lawn of K91 cells (instead of as K91 colonies on LB agar containing 75 g/ml kanamycin).
Immunoblots of Phage-displayed Sequences-5 ϫ 10 10 plaque-forming units of phage produced as described above were disrupted with SDS loading buffer at 100°C for 5 min and loaded onto a 5-20% SDS-PAGE gel (29) to separate capsid proteins. Following transfer to nitrocellulose, the immunoblot was probed with 5 g/ml 7D5 and detected by goat anti-mouse alkaline phosphatase-conjugated secondary (Bio-Rad) in combination with chromagen reagent (Kirkegaard and Perry Laboratories, Inc., Gaithersburg, MD) as described (16).

RESULTS
Specificity of mAb 7D5-7D5 has been previously reported to recognize Cyt b expressed at the plasma membrane of neutrophils but not on cells from X-linked CGD individuals (19,30). The absence of either gp91 phox or p22 phox in X-linked or autosomal recessive CGD, respectively, results in the absence of both proteins on the neutrophil membrane (5,31). Thus, to rule out the possibility that 7D5 binding is dependent on transcriptional regulation of only gp91 phox and not p22 phox , we compared its binding to cells derived from homozygous and heterozygous individuals with autosomal and X-linked inheritance of the disease. Intact neutrophils derived from individuals deficient in gp91 phox (X91 0 ) or p22 phox (A22 0 ) were probed for surface expression of Cyt b using 7D5 in flow cytometry (Fig. 1, A and  B). In contrast to neutrophils from normal individuals, 7D5 did not stain the cell surface of neutrophils deficient in either p22 phox or gp91 phox (Fig. 1, A and B). Neutrophils from the mother and female siblings of an individual with X-linked CGD displayed bimodal fluorescence, demonstrating that they have both normal and Cyt b-deficient neutrophils. The identification of heterogeneous Cyt b expression in their neutrophils indicates that they possess one copy of a mutant CYBB allele and therefore are carriers for X-linked CGD (Fig. 1B). Approximately 87% of the neutrophils from the mother and 80% from the daughter shown in Fig. 1B 1B). These findings indicate that 7D5-epitope recognition required surface expression of both gp91 phox and p22 phox . Moreover, these results illustrate how 7D5 can be used to identify Cyt b-deficient individuals or those who are carriers for Xlinked CGD.
To demonstrate further the specificity of immunoreactivity of 7D5 to Cyt b, human neutrophil Cyt b was purified from cell membranes as described (38) and bound to 96-well plates. Probing the immobilized Cyt b with 7D5, anti-p22 phox mAb 44.1 (␣p22) (14), anti-gp91 phox mAb 54.1 (␣gp91) (14), and an irrelevant mAb suggested that the epitope bound by 7D5 was intact on the detergent-solubilized protein (Fig. 2). Although the reactivity of 7D5 in the ELISA was less than that of ␣p22 or ␣gp91, all three mAbs specifically recognized Cyt b. It is possible that the binding of mAb 7D5 to Cyt b requires a more native or membrane-resident conformation of the protein than does either ␣p22 or ␣gp91, although we were unable to test this directly using our ELISA.
7D5 Recognizes Native Cyt b Heterodimers-To gain insight into the structural nature of the 7D5 epitope, we compared nondenaturing with denaturing conditions for immunoprecipitation of Cyt b using 7D5. Using nondenaturing conditions, 7D5, ␣gp91, and ␣p22 precipitated both gp91 phox and p22 phox from human neutrophil detergent lysates (Fig. 3A). When the lysates were denatured with heat and 1% SDS, ␣gp91 and ␣p22, binding linear regions of the protein, precipitated their respective subunits alone. However, 7D5 precipitated neither subunit after denaturation (Fig. 3A).
To determine whether 7D5 bound complex carbohydrates on gp91 phox , we examined the ability of 7D5 to precipitate Cyt b using biosynthetically active gp91-PLB cells as a source of both precursor and mature Cyt b (Fig. 3B). Using nondenaturing conditions, 7D5 precipitated gp91 phox , p22 phox , and a small amount of gp65, the gp91 phox precursor that has exclusively high mannose oligosaccharides (Fig. 3B). Since 7D5 precipitated gp65, the complex carbohydrates of gp91 phox were not required for 7D5 binding to Cyt b. These findings were confirmed by the ability of 7D5 to precipitate the heterodimeric complex consisting of 55-58-kDa core gp91 phox protein-p22 phox synthesized in the presence of tunicamycin 2 and also bind heterodimer following digestion with PNGase F (data not shown). Consistent with the neutrophil experiments, when gp91-PLB lysates were denatured with 1% SDS and heat, 7D5 precipitated neither mature subunit nor gp65. In contrast, ␣gp91 and ␣p22 precipitated their respective subunits following denaturation (Fig. 3B). These findings suggest that 7D5 recognized the Cyt b peptide backbone in its native form only but do not elucidate whether the antibody bound to an epitope shared by both subunits or if it associated with individual gp91 phox or p22 phox subunits in their native conformation.
Our previous studies on the biosynthesis of Cyt b demonstrated that pools of uncomplexed gp65 and p22 phox accumulate for a limited time following synthesis in gp91-PLB cells (26). Therefore, these cells provide a source of native, monomeric Cyt b subunits from which to test whether 7D5 recognizes FIG. 2. Analysis of 7D5-Cyt b interaction by ELISA. One g of heparin-purified human neutrophil Cyt b was used to coat wells in an ELISA plate so that binding by each of three anti-Cyt b mAbs could be determined. ␣gp91 and ␣p22 were previously shown to be Cyt b-specific (14) and were used as positive controls, and the anti-rhodopsin mAb K42.41 (60) served as a negative control. The results indicate specific binding of 7D5 to the detergent-solubilized form of the protein. These data are typical of five separate analyses.
FIG. 1. Identification of female carriers of gp91 phox deficiency or individuals deficient in gp91 phox or p22 phox by flow cytometry. A, neutrophils (10 6 ) from healthy subjects (father, mother, and normal daughter as indicated) or from an individual with autosomal deficiency of p22 phox (A22 0 ) were probed with 7D5 and analyzed by flow cytometry as described under "Experimental Procedures." B, a similar analysis was performed on neutrophils from an individual with X-linked CGD (X91 0 ) or from the mother and female sibling of an individual with X-linked CGD (mother and daughter carriers as indicated) and were compared with a healthy individual (normal as indicated). Dashed histograms represent staining with IgG1, an isotype control antibody.
individual gp65 or p22 phox subunits. Sequential immunoprecipitations were performed to determine whether 7D5 was capable of depleting lysates of Cyt b heterodimer and/or individual subunits (Fig. 4A). Although 7D5 completely removed gp91-p22 phox heterodimer from lysates after three rounds of immunoprecipitation, monomeric gp65 could be subsequently precipitated from those lysates using ␣gp91 (Fig. 4A, arrowheads). Since neither gp91 phox nor p22 phox coprecipitated with gp65 using ␣gp91 following three rounds of precipitation with 7D5, gp65 was precipitated free of associated p22 phox . The finding that gp65 but neither gp91 phox nor p22 phox coprecipitated with ␣gp91 after the immunodepletions with 7D5 suggests that heterodimeric gp91 phox -p22 phox complexes were completely removed by 7D5 (Fig. 4A). The inability to precipitate and detect by immunoblotting monomeric p22 phox following the three immunodepletions with 7D5 or those with 7D5 followed by a single depletion with ␣gp91 may be due to the rapid degradation of uncomplexed p22 phox or its rapid processing to heterodimeric form (26) (Fig. 4A). It is likely that the amount of newly synthesized, monomeric p22 phox present in the depleted lysates may have been below the limits of detection by immunoblotting. Therefore, we pulse-labeled gp91-PLB cells and subsequently chased for 2 h to allow for partial processing of gp65 to gp91 phox and also to allow the formation of gp91 phox -p22 phox complexes (Fig. 4B). As with the unlabeled immunodepletion experiment, three rounds of precipitation of radiolabeled Cyt b using 7D5 completely removed gp91 phox -p22 phox complexes from lysates (Fig. 4B). In contrast, following depletion with 7D5, subsequent precipitation with either ␣gp91 or ␣p22 revealed that both gp65 and p22 phox remained in depleted lysates free of their complementary subunit (Fig. 4B). These results suggest that 7D5 precipitated only gp91 phox -p22 phox or gp65-p22 phox heterodimers.
Epitope Mapping Using Phage Display-To identify whether the 7D5 epitope was located on gp91 phox , p22 phox , or both subunits, we selected phage display library clones using a 7D5 immunoaffinity Sepharose bead matrix. Limiting dilutions were performed on each of three successive eluate samples, to determine the titer and to provide isolated plaques for plaque lift analysis. A six-log increase in the number of adherent clones was observed between the first and third round of selection, suggesting strong enrichment for peptide sequences by the antibody (data not shown). About 15% of the plaques from the second round and 95% of the plaques from the third round elution gave strong signals when probed by 7D5 in a plaque lift analysis (14), compared with an irrelevant monoclonal (data not shown). Isolated plaques were then selected for nucleotide sequence analysis as described (16) (Table I).
Immunoaffinity selection of peptides presented on phage display clones produced 29 unique amino acid sequences, many of which appeared on several different phage (Table I). The first phage sequence listed in Table I showed a five-residue match to the 226 RIVRG 230 segment of gp91 phox , a region predicted to be extracytoplasmic by hydropathy analysis and by its proximity to the Asn 240 putative glycosylation site. The recovery of this single RIVRGVGGI peptide thus provided important evidence supporting the 226 RIVRG 230 segment of gp91 phox as being part of the epitope (Table I,  Results shown are representative of three separate experiments. B, gp91-PLB or X-CGD PLB cells (2 ϫ 10 6 ) were solubilized in radioimmune precipitation buffer, and Cyt b was precipitated from lysates using mAb 7D5 or ␣gp91 and ␣p22 as indicated. Immunoprecipitations were carried out using nondenaturing conditions or after lysates were heated to 100°C in the presence of 1% SDS to denature all proteins. Following SDS-PAGE, proteins were transferred to nitrocellulose, and immunoblots were probed with a combination of ␣gp91 and ␣p22. Results are representative of three separate experiments. (Table I, clones B, C, D, and F). The 161 KNP 163 region of gp91 phox lies above a predicted transmembrane region 12 residues from Asn 149 , another possible glycosylation site, and 63 residues from the 226 RIVRG 230 segment, just above the fifth predicted transmembrane-spanning domain (12,39). Modest matches in several other selected clones (Table I) also reflected the 161 KNP 163 segment of gp91 phox . More than 50% of the selected phage peptides contained an aliphatic or hydrophobic residue, including Leu, Ile, Val, or Tyr, which when aligned with the 161 KNP 163 and 226 RIVRG 230 segments of gp91 phox , corresponded to Ile 160 (Table I). Most impressively, these two regions of gp91 phox were represented by four phage peptides with five-residue identities and additional conservatively substituted or shifted residues (clones A-D, Table I). In our previous studies, mapping antibody epitopes or protein-protein interactions, such extended matches were rare (16,40). When phage peptides were aligned with the gp91 phox segments 161 KNP 163 and 226 RIVRG 230 , Trp was represented in the same position in greater than 91% of the total number of phage isolates (Table I). Although no residue in the identified gp91 phox sequences fits with this selected residue, Trp 125 immediately outside of the third transmembrane region of gp91 phox is a possible candidate, since all of these transmembrane regions are likely to be in close proximity. Trp 251 could also be represented by the phage-selected sequences, yet we were unable to block the binding of 7D5 in flow cytometry with another antibody that binds this residue (data not shown). It is also possible that Trp 68 of p22 phox contributes the tryptophan in the epitope identified in the phage display mapping and could therefore provide some rationale for the heterodimer requirement for epitope conformation. Another possibility is that the Trp selected by the phage clones represented a hydrophobic "pocket" or "spacer," which bridged the gap between the extracellular transmembrane loops containing 161 KNP 163 and 226 RIVRG 230 sequences.
Several phage sequences also contained an RGD tripeptide motif, which aligned well within the gp91 phox residues 226 RIVRGQ 231 , if flexibility is provided to allow for a polar residue substitution at Gln 231 . The unexpected identification of RGD in several selected sequences suggested that this surfaceaccessible region of Cyt b might interact with integrins in an RGD-dependent manner (41), yet our attempts to confirm such an interaction were unsuccessful (data not shown).
Our mapping data indicate that the minimal epitope bound by 7D5 consists of five mapped residues in the 226 RIVRG 230 segment, and four more matching the 160 IKNP 163 region. These two regions are likely to be extracellular based on hydropathy analysis that suggests they are located immediately adjacent to the extracellular aspect of two putative membrane-spanning helices (39) (see Fig. 6 for putative epitope location). The com-

FIG. 4.
Immunodepletion of gp91 phox -p22 phox and gp65-p22 phox using 7D5. A, gp91-PLB cells were solubilized in radioimmune precipitation buffer, and lysates were immunodepleted using three successive 7D5 precipitations and subsequently subjected to precipitation using either ␣gp91 or ␣p22 as indicated. Following SDS-PAGE, proteins were transferred to nitrocellulose, and immunoblots were probed with a combination of anti-gp91 phox and anti-p22 phox mAbs. The arrowheads indicate gp65. B, alternatively, gp91-PLB cells were pulse-labeled with [ 35 S]methionine for 1 h and then chased 2 h with unlabeled methionine. Cells were solubilized in radioimmune precipitation buffer, and lysates were immunodepleted using three successive 7D5 precipitations followed by either ␣gp91 or ␣p22 as indicated. The arrowhead indicates gp65, and the line at the left indicates a protein of 60 kDa unrelated to gp65. We have previously demonstrated that precipitation of radiolabeled p22 phox by ␣p22 results in uninterpretable signal due to high background above 50 kDa on autoradiograms (26). However, this background signal is not due to gp65 or gp91 phox . The panels at the far right in both A and B illustrate immunoprecipitation (IP) of gp65/91 phox and p22 phox by ␣gp91 and ␣p22, respectively. Following SDS-PAGE, gels were processed for autoradiography. Results are representative of two or three separate experiments. bination of these two regions constitutes a logical target for binding by 7D5, based on our findings from the biochemical assays, i.e. the ability of the antibody to identify an accessible epitope on the plasma membrane of intact neutrophils combined with its inability to bind denatured protein.
The binding of 7D5 to the selected peptides on the denatured pIII display protein (42) from 5 ϫ 10 10 plaque-forming units was examined by SDS-PAGE and immunoblotting (Fig. 5, immunoblot). These signal intensities varied significantly with the displayed sequence, with YPGWGRNDA and YPGW-PRKDL sequences (clones Z and BB, respectively) producing the strongest signals, the clones containing cysteine pairs (clones U and W) showing weak signals, and the rest showing intermediate staining. An irrelevant clone not selected by 7D5 gave no detectable signal (Fig. 5B). These findings demonstrate that the selected peptides exhibited varied degrees of binding to 7D5, and the binding presumably represents the extent to which each denatured clone was able to represent the Cyt b epitope. The binding by 7D5 to clones A and B, which have five residues identical to the putative epitope, were bound less by 7D5 than at least three other clones with fewer identical residues (Fig. 5, immunoblot; compare clones A and B with clones Z, BB, and M). This result suggested that some structure in the displayed epitope mimetic was lost by denaturation.
Since 7D5 required Cyt b in its native conformation to bind, the binding of 7D5 to SDS-denatured phage clones may have represented conditions less than optimal to determine which was the best epitope mimetic. Therefore, we analyzed the ability of various intact 7D5-binding phage clones to block the interaction of 7D5 with purified Cyt b in an ELISA (Fig. 5, bar  graph). Of the eight phage representatives tested, clones M and A (LNTKWLRGD and RIVRGVGGI, respectively) were most effective at blocking the interaction of 7D5 with Cyt b in this assay. Both clones have much greater linear homology to the putative gp91 phox epitope than do clones Z and BB (YPG-WGRNDA and YPGWPRKDL, respectively), which had greater reactivity with 7D5 on the immunoblot (Fig. 5). Moreover, clone B, which had relatively weak reactivity with 7D5 on the immunoblot despite its high similarity to the putative epitope, showed much greater interaction with 7D5 in the ELISA. The difference in immunoreactivity between the immunoblot and the ELISA probably represents the difference in the way each clone is presented to 7D5; the immunoblot presents denatured protein to 7D5, whereas the ELISA promotes interactions that reflect those between native proteins. Our finding that 7D5 required native Cyt b for binding is reflected most by the ELISA data, which demonstrated that clones of higher similarity reacted with 7D5 better in native conformation than in denatured form. DISCUSSION In response to difficulties in predicting the structure of membrane proteins, epitope mapping of antibodies that bind native protein provides an alternative means of assigning localized structure to individual protein domains, as well as elucidating certain membrane topology. The apparent complex nature of the 7D5 epitope suggested that its characterization could provide information about Cyt b structure beyond localization of

I -K-N -P/R -(X)-V-R-G-Q
7D5 was screened with a phage display library as described under "Experimental Procedures," and phage peptides were sequenced and aligned with gp91 phox amino acid residues. Boldface and underlined residues represent identities; italicized and underlined residues, conservative substitutions; residues in boldface only, residues shifted one position; double underlined residues, polar residue conservation; dotted underlined residues, hydrophobic residue conservation. regions outside the plasma membrane. mAb 7D5 has been previously utilized in reports describing molecular and genetic analysis of CGD and the NADPH oxidase, yet the region(s) of Cyt b bound by 7D5 have remained undefined. To provide a better view of the 7D5 epitope and to gain structural information about Cyt b membrane topology, we investigated the 7D5-Cyt b interaction to identify which regions of Cyt b are recognized by the antibody.
Our analysis of 7D5 binding first established its specificity for the Cyt b protein. 7D5 bound neither to neutrophils from patients with autosomal deficiency of p22 phox nor to those from patients with X-linked CGD. Flow cytometric 7D5 staining of neutrophils from female carriers of X-linked CGD was bimodal, consistent with the mosaic expression of Cyt b in such subjects (43). The use of mAb 7D5 to identify carriers of CGD as well as affected individuals constitutes an important yet simple screening test for the genetic types associated with impaired Cyt b expression. Specificity of 7D5 for Cyt b was further supported by the reactivity of 7D5 to partially purified Cyt b in an ELISA.
7D5 precipitated heterodimeric gp91-p22 phox under nondenaturing conditions but recognized neither individual subunit when Cyt b was denatured. Anti-p22 phox and anti-gp91 phox mAbs, ␣p22 and ␣gp91, respectively, which bind linear epitopes (16), also precipitated heterodimeric gp91-p22 phox . In contrast to 7D5, ␣p22 and ␣gp91 precipitated individual p22 phox and gp91 phox subunits after Cyt b had been denatured. The inability of 7D5 to precipitate either individual subunit following denaturation of Cyt b is consistent with its failure to recognize either on an immunoblot. Furthermore, 7D5 recognized neither native monomers of gp65 nor p22 phox , suggesting that individual subunit conformation alone was not sufficient for binding. These results suggest two possible explanations for the requirements of 7D5 binding: 1) that both subunits contribute to the epitope or 2) that the epitope conformation is dependent upon heterodimer assembly, although the residues constituting the epitope reside on only one of the subunits. In addition to gp91 phox and p22 phox , 7D5 precipitated gp65, the ER-resident precursor containing only high mannose carbohydrate (Figs. 3A and 4, A and B) (44), which is later modified in the Golgi to include complex carbohydrates to form gp91 phox (45). Moreover, Yamauchi et al. 2 have demonstrated that the unglycosylated gp91 phox core protein can be precipitated by 7D5 from gp91-PLB cells cultured in the presence of tunicamycin, and our previous studies indicate that the unglycosylated gp91 phox core protein associates with p22 phox to form an unglycosylated heterodimer (26). Together these data indicate that carbohydrate did not contribute significantly to the 7D5 epitope.
Because 7D5 appeared to bind a polypeptide region on Cyt b, we used phage display epitope mapping to identify residues of Cyt b involved in 7D5 binding. The phage-displayed sequences show similarity to both 160 IKNP 163 and 226 RIVRG 230 regions of gp91 phox , yet no matches to p22 phox sequences could be identified. The diversity of the 29 selected sequences suggests that the 7D5 epitope involves a nonlinear or conformational epitope, consistent with the inability of 7D5 to recognize denatured protein on immunoblots. Several phage clones, A-D from Table  I, gave five-residue identities to the discontinuous region spanning 160 IKNP 163 and 226 RIVRG 230 of gp91 phox . Such strong similarity provides credible support to the identification of this region as the 7D5 epitope, especially because these epitope mimetics were recovered from a randomly generated peptide library (14,16,40). Moreover, clone A, one of the clones with five residues identical to the gp91 phox sequence, interacted with 7D5 in the native ELISA better than all clones tested except one (Fig. 5). Tryptophan was also recovered in nearly every phage clone (Table I), although this residue does not appear in either of the two gp91 phox regions identified. It is possible that this tryptophan represents a residue from another membranespanning helix of gp91 phox or a hydrophobic pocket, potentially bridging 160 IKNP 163 and 226 RIVRG 230 of gp91 phox . A subset of four phage clones (Z, AA, BB, and CC, each beginning with YPGW; Table I) are listed in the reverse orientation (carboxyl to amino, left to right, respectively) compared with the other clones in Table I. These clones are unique because 1) they fit FIG. 5. Immunoreactivity of 7D5 with selected phage-displayed sequences by ELISA and immunoblotting. Intact phage display clones showed variable blocking of the binding of 7D5 to immobilized Cyt b by ELISA. Each well of the ELISA plate was coated with Cyt b as described for Fig. 2. Following the pretreatment of 7D5 with the indicated phage clones bearing the unique sequences listed below as described under "Experimental Procedures," the binding of 7D5 to the wells was measured. For these phage clones, immunoblotting (lower section of the figure) was used to illustrate the immunoreactivity of 7D5 for the denatured form of the phage pIII fusion protein bearing the selected peptide sequence. Each lane in the immunoblot represents 5 ϫ 10 10 plaque-forming units of phage clones following separation in the reducing and denaturing gel conditions. Sequences displayed on the phage clones for both the ELISA and immunoblot are as follows: clone A (RIVRGVGGI), clone X (GWIKYRLEG), clone Z (YPGWGRNDA), clone BB (YPGWPRKDL), clone B (YKNPWIRGM), clone M (LNTKWLRGD), clone U (FRCSWCRGE), clone W (GECRWCKGD), unselected clone (unsel.), or none. The pIII capsid protein fused to the unique peptides typically migrates as a 68-kDa protein in these gel conditions. The blot was probed with 3 g/ml 7D5, and immunoreactive phage peptides were detected using alkalinephosphatase conjugated secondary antibody. the consensus best if written in reverse order, 2) clones Z and BB were most strongly recognized by 7D5 in immunoblot (Fig.  5), and 3) a synthetic peptide analog comprising part of clones Z and BB (NH 2 -ADNRPWGPYGP-CONH 2 ) was the only synthetic peptide sequence found to compete with the binding of 7D5 to immobilized Cyt b in ELISA, albeit only at an EC 50 of 1 mM (data not shown). The ability of the clones to display differential reactivity in the immunoblot versus the ELISA is probably due to the complex nature of the epitope; i.e., the sequences displayed by the phage are the best linear representatives of an epitope that requires tertiary structure.
This selection of apparent reverse sequences by mAb 7D5 suggests that this antibody can select epitope mimetics that are not always in the same orientation as natural epitope. We previously observed the recovery of some peptides from the J404 phage display library that fit the consensus if listed in reverse order (40, 46, 47), and the cross-reactivity of antibodies with retropeptides has been specifically addressed (48). We also observed that the reverse sequence of a bioactive peptide shows significantly greater effects than a peptide bearing a randomly chosen sequence (49). An interaction between the bacterial ribonuclease barnase and its natural inhibitor barstar, are also reported to be "strong and relatively insensitive to ideal geometry" when compared with the interaction between the enzyme and the nucleotide substrate (50). These findings support the rationale for our observations that antibodies can interact with peptides by identifying critical side chain charge distribution and hydrophobicity, in a way that is not entirely dependent on peptide backbone orientation.
The findings from the immunological analyses and phage display are summarized in a schematic model of the topology and identity of the 7D5 epitope contained within two separate extracellular regions of gp91 phox (Fig. 6). Previous attempts to elucidate Cyt b structure, including mapping of extracellular versus intracellular domains have been based on hydropathy analyses and enzymatic cleavage experiments (12), structural homology to known proteins in combination with computer modeling (39,51,52), and identification of the binding regions of p47 phox (40, [53][54][55][56] and NADPH and FAD (25,54,57,58)). Imajoh-Ohmi et al. (12) found that antibodies raised against gp91 phox residues 150 -172 stained intact neutrophils, leading the authors to conclude that that region is exposed on the cell surface. Our data not only extend those findings but demonstrate that gp91 phox residues 160 IKNP 163 and 226 RIVRG 230 are juxtaposed on the extracytoplasmic face of the plasma membrane (Fig. 6), providing direct evidence for the identity and positioning of two previously predicted transmembrane regions (12,39). Moreover, these results are compatible with the findings of Wallach et al., which demonstrated that gp91 phox residues Asn 132 , Asn 149 , and Asn 240 are glycosylated and, therefore, exposed on the cell surface (59).
Our mapping data suggest that the epitope bound by 7D5 was located entirely on gp91 phox , but the presence of p22 phox in the gp91-p22 heterodimer was required for 7D5 binding on human myeloid cells. It should also be noted that heme insertion is a prerequisite for heterodimer formation (26) that in turn is essential for recognition by 7D5. To the extent that heterodimer formation requires heme coordination by one or both subunits, the 7D5 epitope is at the least indirectly hemesensitive. Thus, p22 phox , possibly in combination with heme, appears to impart a structural constraint to gp91 phox that is essential for recognition by 7D5. Further structural analysis will be necessary to elucidate the influence of p22 phox on the membrane topology and processing of gp91 phox and the possible role of heme in this process.
FIG. 6. A hypothetical model for the 7D5 epitope and membrane topology of Cyt b. The epitope was mapped to two juxtaposed extracytoplasmic regions of gp91 phox , residues 160 IKNP 163 and 226 RIVRG 230 , represented by blue and green spheres, respectively. These residues are predicted by hydropathy analysis of the protein to exist immediately extracellular to the fourth and fifth membrane-spanning helices. Residues identified in yellow lettering were represented by the sequences selected from the phage display library and thus accessible to 7D5 on the surface of the cell. The underlined glutamine residue represents polar residue conservation for the epitope in that position although it was not represented by selected sequences. Although tryptophan was almost always found between sequences representing 160 INKP 163 or 226 RIVRGQ 231 in individual phage clones, a tryptophan residue does not exist in this vicinity of the protein. Therefore, the black W represents the position of this residue as selected by phage clones.