Mapping of functional domains in the p22(phox) subunit of flavocytochrome b(559) participating in the assembly of the NADPH oxidase complex by "peptide walking".

The superoxide-generating NADPH oxidase complex of phagocytes consists of a membranal heterodimeric flavocytochrome (cytochrome b(559)), composed of gp91(phox) and p22(phox) subunits, and four cytosolic proteins, p47(phox), p67(phox), p40(phox), and the small GTPase Rac (1 or 2). All redox stations involved in electron transport from NADPH to oxygen are located in gp91(phox). NADPH oxidase activation is the consequence of assembly of cytochrome b(559) with cytosolic proteins, a process reproducible in a cell-free system, consisting of phagocyte membranes, and recombinant cytosolic components, activated by an anionic amphiphile. p22(phox) is believed to act as a linker between the cytosolic components and gp91(phox). We applied "peptide walking" to mapping of domains in p22(phox) participating in NADPH oxidase assembly. Ninety one synthetic overlapping pentadecapeptides, spanning the p22(phox) sequence, were tested for the ability to inhibit NADPH oxidase activation in the cell-free system and to bind individual cytosolic NADPH oxidase components. We conclude the following. 1) The p22(phox) subunit of cytochrome b(559) serves as an anchor for both p47(phox) and p67(phox). 2) p47(phox) binds not only to the proline-rich region, located at residues 151-160 in the cytosolic C terminus of p22(phox), but also to a domain (residues 51-63) located on a loop exposed to the cytosol. 3) p67(phox) shares with p47(phox) the ability to bind to the proline-rich region (residues 151-160) and also binds to two additional domains, in the cytosolic loop (residues 81-91) and at the start of the cytosolic tail (residues 111-115). 4) The binding affinity of p67(phox) for p22(phox) peptides is lower than that of p47(phox). 5) Binding of both p47(phox) and p67(phox) to proline-rich p22(phox) peptides occurs in the absence of an anionic amphiphile. A revised membrane topology model of p22(phox) is proposed, the core of which is the presence of a functionally important cytosolic loop (residues 51-91).

Oxygen-derived radicals are important mediators in the destruction of microorganisms by phagocytic cells. All radicals are derived from superoxide (O 2 . ), 1 which is generated by the NADPH-derived one-electron reduction of oxygen, catalyzed by a membrane-associated flavocytochrome (cytochrome b 559 ) (reviewed in Refs. 1 and 2). Cytochrome b 559 is a heterodimer composed of two subunits, a 91-kDa glycoprotein (gp91 phox ) and a 22-kDa protein (p22 phox ) (3). The gp91 phox subunit bears all the redox stations responsible for the transport of electrons from NADPH to oxygen as follows: an NADPH-binding site and noncovalently bound FAD (4,5), and two nonidentical hemes coordinated with two pairs of histidines (6). Stimuli eliciting O 2 . production by phagocytes initiate a signal transduction sequence that, ultimately, leads to the conversion of cytochrome b 559 from a resting to an activated form. This is probably the consequence of a conformational change in gp91 phox and is brought about by the translocation to the membrane environment of some or all of four cytosolic regulatory proteins, p47 phox , p67 phox , p40 phox , and the small GTPase Rac (1 or 2). It is assumed that this leads to the direct interaction of at least some of the cytosolic components with cytochrome b 559 and the formation of a membraneassociated multicomponent complex of yet unknown composition, stability, and reversibility. This process is termed NADPH oxidase assembly, and the multiple protein-protein interactions among the cytosolic components and between some of these and the cytochrome b 559 subunits were and are the subject of intense investigation (reviewed in Refs. [7][8][9]. NADPH oxidase assembly, expressed in NADPH-dependent O 2 . production, can be reproduced in a cell-free system, consisting of phagocyte membranes and cytosol (10,11) or purified cytochrome b 559 and recombinant cytosolic components (12), exposed to a critical concentration of anionic amphiphiles, such as arachidonate or SDS. The identity of the cytosolic component(s) responsible for causing the conformational change in gp91 phox is a still an unsettled issue, as well as the question of whether this is the consequence of direct interaction of the cytosolic component(s) with gp91 phox or an event secondary to its binding to the p22 phox subunit. Despite the fact that the p22 phox subunit does not contain redox centers, its presence is essential for NADPH oxidase assembly, as shown by the lack O 2 . production by phagocytes of patients with chronic granulomatous diseases (CGD), with a mutation in p22 phox (13), and by COS7 cells transfected with cDNA encoding gp91 phox alone (6). This sug-gested that p22 phox might serve as a linker protein between cytosolic components and gp91 phox . This hypothesis is supported by extensive experimental evidence showing that a proline-rich region in the C-terminal region of p22 phox interacts with the N-terminal Src homology 3 (SH3) domain of p47 phox (14,15). This interaction is related to the activation of NADPH oxidase; the accepted view is that phosphorylation of specific serine residues in a cationic C-terminal region of p47 phox (in vivo) or exposure to anionic amphiphile (in vitro) leads to the disruption of intramolecular bonds between the N-terminal SH3 domain and a C-terminal region in p47 phox , allowing the establishment of the intermolecular bond with p22 phox (16,17). So far, no other protein-protein interaction involving p22 phox and a cytosolic protein, other than p47 phox , has been observed.
The possibility that such interactions nevertheless occur is raised by findings showing interactions between p67 phox and Rac and cytochrome b 559 . Thus, the NADPH oxidase can be activated in vitro, in the absence of p47 phox , by p67 phox and Rac, at high concentrations (18,19) or when Rac is in the isoprenylated form (20). These findings are in good agreement with the proposal that p67 phox is the protein directly responsible for the induction of a conformational change in gp91 phox and with the identification of an "activation domain" in p67 phox , responsible for interaction with cytochrome b 559 (21,22). It was implied that p67 phox interacts directly with the gp91 phox subunit (22), a contention for which experimental proof was provided only recently (23). It is of interest and relevance to the results presented in this paper that no direct interaction between p67 phox and p22 phox was ever reported despite the fact p67 phox contains two SH3 domains as follows: one was found to interact with a proline-rich domain in the C terminus of p47 phox (24), and the other was possibly engaged in an intramolecular bond with a proline-rich domain at the center of the molecule (proposed in Ref. 1). It was also suggested that cytochrome b 559 serves as a direct target for Rac. This was based on the finding that maintenance of Rac2 in the membrane, following translocation, depends on its interaction with cytochrome b 559 (25), and it was suggested that the insert region of Rac1 is involved (26). Direct evidence for binding of Rac2 to cytochrome b 559 was recently presented (27). In all these reports it was implied that the target of Rac is gp91 phox ; however, there was no direct evidence for this, and the possibility that the effect on gp91 phox is secondary to the interaction of Rac with p22 phox could not be excluded.
We introduced an approach to the mapping of domains involved in protein-protein interactions between individual components of an enzymatic complex, termed "peptide walking." This is based on breaking down the complete sequence of the protein under investigation in a series of overlapping synthetic peptides and using these peptides as either inhibitors of an enzymatic reaction involving the parent protein or as ligands to which interacting proteins will bind. Peptide walking was used earlier for mapping functional domains in Rac1 (28) and p47 phox (29).
The present report describes the first attempt to apply peptide walking to the analysis of a membrane component, namely to the p22 phox subunit of cytochrome b 559 . Because p22 phox is viewed principally as a linker element between gp91 phox and the cytosolic components of the NADPh oxidase complex and its size was not prohibitive, it seemed particularly suited for this methodology. Peptide walking reveals the existence in p22 phox of three, previously undetected, binding sites for p67 phox , one of which overlaps the known prolinerich region, also binding p47 phox ; it also demonstrates the existence of a second binding region for p47 phox , in addition to that associated with the proline-rich region. These results form the basis of a revised membrane topology model for p22 phox , in which a cytosolic loop serves as an anchor for both p47 phox and p67 phox , distinct from the proline-rich domain localized in the cytosolic C-terminal tail.

EXPERIMENTAL PROCEDURES
Synthetic Peptides-A library of 91 overlapping pentadecapeptides (PepSets), spanning the complete amino acid sequence of p22 phox (30), was synthesized by the multipin synthesis method (31) by Mimotopes (Clayton, Australia). Peptides overlapped by 13 residues and were biotinylated at the N terminus and amidated at the C terminus. The biotin group was attached by the intermediary of a 4-residue spacer, consisting of SGSG. The purity of the peptides ranged from 60 to 70%. Two batches of 91 peptides (1 mol of each peptide) were used in the reported experiments, with identical results. The freeze-dried peptides were dissolved in a mixture of 75 parts dimethyl sulfoxide and 25 parts water (v/v), to a concentration of 1.5 mM, divided in 50-l aliquots, and kept frozen at Ϫ75°C. PepSets were used exclusively for screening experiments; for work with individual peptides found active by initial screening, purified synthetic peptides were used. These were also synthesized by Mimotopes but had a purity of Ն70%, confirmed by reverse phase high pressure liquid chromatography and ion spray mass spectrometry.
Preparation of Phagocyte Membrane Vesicles-Membranes were prepared from guinea pig peritoneal macrophages (10), solubilized by 40 mM n-octyl-␤-D-glucopyranoside and membrane vesicles generated by the removal of detergent by dialysis (31).
Preparation of Recombinant Proteins-p47 phox and p67 phox were prepared in baculovirus-infected Sf9 cells, and nonprenylated Rac1 was produced in Escherichia coli, as described previously (32).
Measurement of Protein Concentration-This was measured by the method of Bradford (33) and modified for use with 96-well plates, as detailed in technical bulletin 1177EC (Bio-Rad). Bovine serum albumin was used as standard.
Cell-free NADPH Oxidase Assay-NADPH oxidase activation was assayed in a semi-recombinant cell-free system consisting of macrophage membrane vesicles and recombinant cytosolic components, with lithium dodecyl sulfate (LiDS), serving as the anionic amphiphilic activator, as described (34). Briefly, the basic system contained membranes, equivalent to 10 nM cytochrome b 559 heme, and p47 phox , p67 phox , and Rac1, exchanged to GTP␥S (as described in Ref. 20 . production. This was quantified by the kinetics of cytochrome c reduction, as described (34). 28 measurements, at 11-s intervals, were executed using a SpectraMax 340 microplate reader (Molecular Devices), fitted with Softmax Pro software. Results were expressed as nanomoles of O 2 . /min produced per well. Specificity of cytochrome c reduction was confirmed by its prevention by superoxide dismutase. In all experiments, each measurement was performed in triplicate. Inhibition of NADPH Oxidase Activation by p22 phox Peptides-The effect p22 phox peptides on NADPH oxidase activation was tested in the system described above. Peptides were diluted from the 1.5 mM stock solution to a concentration of 100 M, and an amount of 10 l was added per well, as the first component of a 100-l reaction mixture (resulting in a final concentration of 10 M peptide). When dose-response experiments were executed, the peptides were diluted to result in final concentrations in the assay varying from 0.6 to 40 M. To control wells, 10 l of assay buffer supplemented with dimethyl sulfoxide were added, to result in a final concentration of solvent identical to that found in the peptide-containing wells. This was followed by the addition of 60 l of assay buffer and 10 l of a mixture of p47 phox , p67 phox , and Rac1-GTP␥S, to reach a final concentration of 50 nM each. The contents of the wells were incubated with mixing for 5 min at room temperature on an orbital shaker, to allow the interaction between peptide and one or more of the cytosolic components. Following this, 10 l of macrophage membrane vesicles, equivalent to a final concentration of cytochrome b 559 heme of 10 nM and 10 l of assay buffer, containing LiDS, to achieve a final concentration of 130 M, were added. The 96-well plate was reincubated for 90 s at room temperature, with mixing and O 2 . production initiated by the addition of 10 l of NADPH, to reach a final concentration of 454 M. The effect of a peptide on NADPH oxidase activation was expressed as percent inhibition of NADPH oxidase activation; this was calculated by considering O 2 . production by control mixtures (in the absence of peptide) as 100%. To assess the effects of peptides not related to NADPH oxidase assembly, we examined the effect of p22 phox peptides added after the completion of NADPH oxidase activation. In these experiments, membranes, p47 phox , p67 phox , and Rac1-GTP␥S were mixed at the concentrations listed above and supplemented with 130 M LiDS. 90-l amounts of the mixture were added to wells of a microplate and incubated for 1.5-5 min, at room temperature, to induce NADPH oxidase assembly. The peptides were then added to a final concentration of 10 M, and after 30 s of incubation, NADPH was added to a final concentration of 454 M, and O 2 . production was measured, as described above.
Assay of Peptide-Protein Interactions by Enzyme-linked Immunosorbent Assay-This assay was first used by us for the mapping of domains on p47 phox interacting with p67 phox (29). In the present experiments we applied a modified version of this methodology, which is described below. The experiments were performed in 96-well microplates, with flat bottom and rounded corners (MaxiSorp C96, product 430341, Nunc). The plates were coated with streptavidin (from Streptomyces avidinii, affinity-purified, product S4762, Sigma), 100 l (16.6 pmol) per well, as described before (29). A number of experiments were also performed with ready-made streptavidin-coated 96-well plates (Combi-plate 8 Streptavidin 200 l, code 95029263; Labsystems), with results indistinguishable from those obtained with plates coated by us. To each well were added 300 l of a blocking solution, consisting of phosphatebuffered saline (10 mM sodium phosphate buffer, pH 7.2, and 140 mM NaCl; PBS) supplemented with 0.1% (v/v) polyoxyethylene sorbitan monolaureate (Tween 20), and 1% sodium caseinate, and the plate kept for 1 h at room temperature. The wells were washed four times with 300 l/well of PBS, containing 0.1% Tween 20 (PBS-T), using an automatic microplate washer (Wellwash Ascent, Labsystems). To groups of three wells, we added 100 l of a 2 M biotinylated peptide solution (200 pmol/well), and the plates were incubated for 1 h at room temperature, with rotary mixing, to allow binding of the peptides to the immobilized streptavidin. Unbound peptide was removed by four washes of 300 l/well of PBS-T. At this stage, one of the three recombinant cytosolic NADPH oxidase components was added to the wells, in 100-l amounts, corresponding to 200 pmol/well of the respective proteins (p47 phox , p67 phox , or Rac-GTP␥S), diluted in NADPH oxidase assay buffer (34), supplemented with 1% sodium caseinate. The components were incubated with the peptides on a orbital shaker for either 1 h at room temperature (p47 phox ) or 18 h at 4°C (p67 phox and Rac1). Unbound components were removed by four washes of 300 l/well of PBS-T, and FIG. 1. List of overlapping p22 phox synthetic pentadecapeptides used for inhibition of NADPH oxidase and for binding of cytosolic components experiments. Numbers at the N and C termini of each peptide indicate the location of the corresponding 15 residues in the amino acid sequence of p22 phox . The four additional residues (SGSG), added at the N terminus as a spacer between biotin and peptide, are not shown. Residues highlighted in yellow represent segments in the sequence of peptides belonging to domains defined by inhibition of NADPH oxidase activation. The boldface italic lowercase letters to the right of the peptides indicate the nomenclature of these domains, as shown in Fig. 3A. Residues in dark red boldface represent segments in the peptide sequence belonging to domains involved in binding of p47 phox ; residues in green boldface represent segments in the peptide sequence involved in binding of p67 phox . The boldface italic uppercase letters to the right of the peptides indicate the nomenclature of these domains (dark red for p47 phox binding and green for p67 phox binding; see Figs. 5A and 6A). the amounts of bound component were quantified by the use of polyclonal goat antisera against purified recombinant p47 phox (diluted 1/1000) and p67 phox (diluted 1/500) (35), kindly provided by Dr. Thomas L. Leto, and of a rabbit antibody against a synthetic peptide corresponding to residues 178 -191 of Rac1 (catalog number sc-217; Santa Cruz Biotechnology) (diluted 1/500). Primary antibodies, diluted in PBS-T, containing 1% sodium caseinate (100 l per well) were added and incubated for 1 h at room temperature, with orbital shaking. The wells were washed four times with PBS-T and covered with 100 l/well of the corresponding secondary antibodies, diluted in PBS-T, supplemented with 1% sodium caseinate. These were peroxidase-conjugated affinitypurified rabbit anti-goat IgG at a dilution of 1/10,000 for anti-p47 phox and 1/5,000 for anti-p67 phox , and goat anti-rabbit IgG, diluted 1/10,000, for anti-Rac1. The plates were incubated for 1 h at room temperature, with orbital shaking, and subjected to four washes with 300 l/well of PBS-T and two washes with 300 l/well of PBS. Detection of peroxidase was performed with a ready-to-use 3,3Ј,5,5Ј-tetramethylbenzidine liquid substrate system (Sigma, product number T8665) by adding 200 l of solution per well and incubating for 30 min at room temperature. The reaction was stopped by the addition of 100 l/well of 0.5 M H 2 SO 4 solution and the absorbance read at 450 nm, against blank wells containing only the peroxidase assay reagent, in a SpectraMax 340 microplate reader.

NADPH Oxidase Activation in Vitro Is Inhibited by Clusters
of Overlapping p22 phox Peptides-Ninety one overlapping pentadecapeptides, spanning the complete amino acid sequence of p22 phox , were tested for the ability to interfere with the process of amphiphile-elicited NADPH oxidase activation in a cell-free system, consisting of membrane vesicles and recombinant p47 phox , p67 phox , and Rac1-GTP␥S. Fig. 1 lists the peptides employed in these experiments and the location of the corresponding sequences within the p22 phox protein. The N-terminal spacer sequence SGSG, to which biotin is attached, does not appear in the list.
This, to the best of our knowledge, is the first time that peptides derived from the sequence of a membrane-spanning protein were tested for a possible inhibitory effect on a biochemical reaction involving the parent protein. Many of these peptides are expected to be hydrophobic. We therefore plotted the degree of hydrophobicity of the individual overlapping peptides, determined as described in Ref. 36 (on the y axis), against the location of the sequences represented by the peptides (the position of the N-terminal residue of each peptide) in the intact protein (on the x axis). As can be seen in Fig. 2A, there are two major groups of hydrophobic peptides. The first ranges from residue 1 to residue 50 and corresponds to a transmembranal region, as proposed by several authors. The second extends from residue 63 to 125 and roughly overlaps a second proposed transmembranal region. A small cluster of hydrophilic peptides, peaking between residues 53 and 59, is of special interest in light of our proposal that this corresponds to the N-terminal half of a cytosolic loop, the existence of which was proposed and which, based on the results to be described, might serve as an anchor for cytosolic components p47 phox and p67 phox (see "Discussion" for earlier proposals of the membrane topology of p22 phox ). Peptides corresponding to the C-terminal region of p22 phox , starting from residue 127, were mildly hydrophobic or hydrophilic, in good agreement with the commonly accepted view that the C terminus forms a cytosolic "tail." Next, the relative net charge of each peptide (on the y axis) was also plotted against the location of the peptide in the intact protein. Each lysine, arginine, or histidine contributed one positive charge unit, whereas each glutamate and aspartate contributed one negative charge unit. As can be seen in Fig. 2B, there was a large cluster of basic peptides, extending from residue 45 to 93. This corresponds to the hydrophilic peptide cluster and the N-terminal half of the second hydrophobic cluster and sets, roughly, the N-and C-terminal limits of the full-length of the cytosolic loop, as suggested by us. A second basic cluster, extending from residue 125 to 155, corresponds to the N-terminal half of the cytosolic tail.
The effect of p22 phox peptides was determined in a semirecombinant cell-free system, as described under "Experimental Procedures," using a fixed concentration of membrane (equivalent to 10 nM cytochrome b 559 heme), and p47 phox , p67 phox , and Rac1-GTP␥S, at identical concentrations (50 nM). This concentration was chosen based on the finding that, in NADPH oxidase activation dose-response experiments, 50 nM was just below the plateau level and that the three cytosolic components are known to translocate to the membrane at equimolar concentrations (37). The design of the experiments was such as to allow preliminary interaction of p22 phox peptides with any of the three cytosolic components, which were looked upon as potential candidates for interaction with p22 phox . Thus, peptides, at a concentration of 10 M, were preincubated for 5 min with a mixture of the three cytosolic components (at a 200-fold lower concentration), and then membranes were added, together with the activating amphiphile. Fig. 3A illus-trates the results of screening all the peptides for inhibition of NADPH oxidase. Four clusters of 3-13 overlapping inhibitory peptides were apparent (labeled b, c, d, and e). An isolated inhibitory peptide (residues 9 -23) was labeled a. Interestingly, all clusters were located in the N-terminal 2/3 of p22 phox ; therefore, most of the C-terminal cytosolic tail did not yield inhibitory peptides.
When the same peptides were added to mixtures of membrane and cytosolic components, preincubated for 90 s with LiDS, no inhibitory effect was found (Fig. 3B). This indicates that p22 phox peptides interfere with the assembly of the  50 values for selected synthetic p22 phox peptides exerting an inhibitory effect on NADPH oxidase activation in vitro Five peptides were chosen, each representing a distinct cluster of inhibitory peptides, corresponding to the five sequence domains in p22 phox , defined by inhibition of NADPH oxidase activation. For each cluster, the most inhibitory peptide was chosen. Purified peptides (purity Ն70%), at concentrations varying from 0.6 to 40 M, were preincubated with a mixture of p47 phox , p67 phox , and Rac1-GTP␥S (50 nM each) for 5 min. This was followed by the addition of membrane vesicles (equivalent to 10   Clusters of overlapping inhibitory peptides were analyzed for the presence of a shared amino acid sequence, which we termed "domain," as described earlier (28,29). Briefly, the sequence associated with the most inhibitory peptide, in a certain cluster, was taken as representing the full-length domain; neighboring peptides, containing truncations of the full-length domain, either at the N or the C terminus, were expected to exhibit a proportionally lesser inhibitory potency. Sometimes exposure of part of a domain at the C-terminal end of a peptide resulted in an inhibitory effect different in amplitude from that caused by a peptide exposing the same sequence at its Nterminal end (see Ref. 28, for a similar situation in the mapping of Rac1 domains). The representation of the NADPH oxidase inhibitory domains, so derived, on the individual peptides is shown in Fig. 1; a summary of the boundaries of the domains is shown in Fig. 8.
For screening purposes, peptides were present in the reaction at a fixed concentration of 10 M, in marked excess over that of the cytosolic components. To get more information on the inhibitory effectiveness of p22 phox peptides, we performed dose-response studies with purified peptides, each representing a different domain. The IC 50 values were derived from typical sigmoid curves, obtained when peptide concentrations were plotted, on a logarithmic scale, against percent inhibition of NADPH oxidase activation values and are shown in Table I. It can be seen that these range from 3 to 5.7 M; peptide 85-99, belonging to domain d, is the most potent inhibitor (as also seen in Fig. 3A).
In further experiments we confirmed the adequacy of the length of time for which peptides were incubated with cytosolic components, by varying the time interval from 0 to 10 min. Four peptides were tested, representative of domains b, d, and e. For all, the inhibitory effect reached a plateau after 5 min of incubation (results not shown). Consequently, exposing the cytosolic components to p22 phox peptides for 5 min in the screening experiments was fully adequate for achieving a maximal effect.
We also investigated whether inhibition of NADPH oxidase activation could be the result of the peptides interacting directly with the anionic amphiphilic activator LiDS. We have shown in the past that both the hydrophobic tail and the anionic head are required for activation by amphiphiles (10,11), and the possibility existed that a hydrophobic and positively charged peptide might bind LiDS and cause a shift to the right of the optimal activating concentration of LiDS (130 -140 M). This eventuality was eliminated by an experiment in which the inhibitory effect of a domain d peptide (residues 85-99; at 10 M) was found to be unaffected by varying the concentration of LiDS from 60 -300 M (results not shown).
We next attempted to identify the specific cytosolic component the interaction of which with p22 phox peptides was responsible for the inhibitory effect. The design of these experiments was identical to that described under "Experimental Procedures," except that peptides were preincubated for 5 min with either p47 phox , p67 phox , or Rac1-GTP␥S instead of the mixture of all three components. This was followed by complementation with membrane, the missing two cytosolic components, and LiDS, incubation for 90 s, and the initiation of the reaction with NADPH. These experiments revealed an inhibition pattern indistinguishable from that found when all three cytosolic components were added together and could, therefore, shed no light on the nature of the peptide-binding component(s) (results not shown). Identifying the components interacting with p22 phox peptides was made possible only by direct binding experiments (see below).
Inhibition of NADPH Oxidase Activation by p22 phox Peptides Is Not Sequence-specific-An inhibitory effect of a peptide, corresponding to a specific segment in the amino acid sequence of a protein, on a biochemical reaction involving that protein is usually interpreted as the consequence of competitive inhibition (see Refs. 29 and 38). This raises the question of whether competition is sequence-specific or involves other elements, such as hydrophobicity and/or charge. Thus, we found that inhibition of NADPH oxidase activation by C-terminal Rac1 peptides was related exclusively to their polybasic character (39). To answer this question, a number of peptides, each of which represented a cluster of inhibitory peptides, were synthesized with scrambled sequences. For scrambling, we used the same algorithm for all peptides; the original 15 residues were placed in the following N to C order: 9, 6, 11, 4, 13, 2, 15, peptides chosen were those which possessed the highest inhibitory activity, in each cluster. The effect of the scrambled peptides was compared with that of the native peptides in parallel dose-response experiments, in which the concentration on peptides was varied from 0.6 to 40 M. As apparent in Fig. 4, there was no significant difference in the inhibitory potency of native and scrambled peptides belonging to all five clusters. On two occasions (peptides 9 -23 and 31-45), the IC 50 values of the scrambled peptides were even lower than those of the native peptides. In a more limited number of experiments, we also investigated the effect of synthetic peptides in which the order of residues was inverse to that found in the native molecule ("retro-peptides"). Thus, peptides 85-99 and 97-101 (both belonging to cluster d) were synthesized in the reversed order, starting with the C-terminal residue and ending with the Nterminal residue. When tested in dose-response experiments, from 0.6 to 40 M peptide, we found that, similarly to scrambled peptides, the inhibitory activity of retro-peptides was indistinguishable from that of the native compounds (results not shown).
Just as was the case for native peptides, scrambled and retro-peptides were only inhibitory when added before NADPH oxidase assembly and exhibited maximal activity when incubated with cytosolic components for 5 min or longer, and the inhibitory effect could not be overcome by increasing the concentration of the amphiphilic activator LiDS.
p47 phox Binds to Surface-immobilized p22 phox Peptides-This approach was first applied to the study of NADPH oxidase assembly when it was demonstrated that p67 phox binds to peptides corresponding to a C-terminal proline-rich domain in p47 phox (29). By using a similar methodology, we investigated the binding of p47 phox to a series of overlapping biotinylated p22 phox peptides, corresponding to the full-length of p22 phox , immobilized on the surface of streptavidin-coated wells. After initial pilot experiments, leading to the identification of a number of p47 phox -binding peptides, the assay was optimized for peptide concentration, p47 phox concentration, length of time, and temperature of incubation (see "Experimental Procedures").
As seen in Fig. 5A, p47 phox binds to two clusters of p22 phox peptides, a large one (cluster A) extending from peptide 45-59 to peptide 61-75 and a smaller one (cluster PR) consisting of peptides 151-165 and 153-167, preceded by two peptides with lesser binding capacity (145-159 and 147-161). A number of peptides C-terminal to cluster A show low binding of p47 phox , with the exception of the isolated peptide 77-91, exhibiting significant binding. Peptides in cluster A share a domain consisting of residues 51-63, and peptides in cluster PR share a domain consisting of residues 151-160 (see Figs. 1 and 8). In mock experiments, in which all steps of the binding assay were executed with the exception of adding p47 phox to the wells, no peroxidase activity was detected (Fig. 5B). There is experimental evidence for the existence in p47 phox of intramolecular bonds between the N-terminal SH3 domain and a C-terminal region (14,15); these bonds can be disrupted by anionic amphiphiles, at NADPH oxidase-activating concentrations (16), allowing the SH3 domain to establish fresh intermolecular interactions with other proline-rich targets. Because domain PR was identical to one of the C-terminal proline-rich domains previously identified in p22 phox (15), we reasoned that a different binding pattern of p47 phox to p22 phox peptides might become apparent in the presence of amphiphile. Therefore, we repeated the binding experiments in the presence of 130 M LiDS added to the p47 phox solution. This had no effect on the extent of binding of p47 phox to previously identified peptide clusters, and no new clusters were identified (results not shown).
p67 phox Binds to Surface-immobilized p22 phox Peptides-The same approach was employed to detect an eventual interaction between p67 phox and p22 phox peptides. In pilot experiments, done using the conditions found optimal for binding of p47 phox , no significant binding of p67 phox was detectable. Only when the time of incubation was extended to 18 h, at a temperature of 4°C, did we find binding of p67 phox to three clusters of peptides as follows: from peptide 73-87 to peptide 81-95 (cluster B), from peptide 99 -113 to peptide 111-125 (cluster C), and to peptides 145-159 and 147-161 (cluster PR) (Fig. 6A). An additional cluster of two p67 phox -binding peptides (37-51 and 39 -53) was also detected (unlabeled). Peptides in cluster B share a domain consisting of residues 81-91; peptides in cluster C share a domain consisting of residues 111-115, and peptides in cluster PR share a domain consisting of residues 151-160 (see Figs. 1 and 8). No domain was attributed to peptides 37-51 and 39 -53, characterized by high hydrophobicity (see Fig. 2) and proposed to correspond to a transmembrane region (see Fig. 9). In mock experiments, in which all steps of the binding assay were executed with the exception of adding p67 phox to the wells, no peroxidase activity was detected (Fig. 6B). As found with p47 phox , performing the p67 phox binding assay in the presence of 130 M LiDS had no effect on the extent of binding of p67 phox to previously identified peptide clusters, and no new clusters were identified.
Rac1 Does Not Bind to Surface-immobilized p22 phox Peptides-Experiments similar to those performed with p47 phox and p67 phox were done with Rac1. We measured binding of Rac1 in both the GTP␥S-and GDP␤S-bound forms. Rac1 bound to only to two isolated peptides, corresponding to residues 43-57 and 109 -123 in the p22 phox sequence (results not shown). Binding was independent of the nucleotide bound to Rac1. The peptides were not associated in a cluster and were of marked hydrophobicity (36). We concluded that there is no evidence for domain-related direct interaction between Rac1 and p22 phox . p47 phox and p67 phox Bind to an Identical Domain on p22 phox Expressed at Different Peptide Termini-In the screening experiments described above, we found that both p47 phox and p67 phox bind to the same C-terminal domain (residues 151-160). However, p47 phox bound preferentially to peptides in which this domain was located at the N terminus, whereas p67 phox bound preferentially to peptides in which this domain Mapping Functional Domains in p22 phox by Peptide Walking was located at the C terminus (see Figs. 5A and 6A). In order to confirm this finding, we repeated the experiments by focusing specifically on the four peptides in question and using purified synthetic peptides (Ն70%). As apparent in Table III, p67 phox bound indeed to peptides 145-159 and 147-161 (with decreasing intensity), exposing domain 151-160 at the C terminus, whereas p47 phox bound to peptides 151-167 and 153-167 (with decreasing intensity), exposing domain 151-160 at the N terminus. Some binding of p47 phox to peptide 147-161 was also evident, a finding which also came up during screening with PepSets peptides.
Binding of p47 phox and p67 phox to p22 phox Peptides Is Sequence-specific-We have shown that inhibition of NADPH oxidase activation by p22 phox peptides was an expression of hydrophobicity and charge, this being true for peptides belonging to all five clusters (Fig. 4). It was, therefore, of interest to find out whether binding of the cytosolic components to p22 phox peptides was specific for the sequences of the individual binding peptides. For this purpose, four peptides known to bind p47 phox (two belonging to cluster A and two to cluster PR) and four peptides known to bind p67 phox (two belonging to cluster B and two to cluster PR) were synthesized in scrambled form, based on the algorithm described above (Table II). The scrambled and corresponding native peptides (Ն70% pure) were tested in parallel for the ability to bind either p47 phox or p67 phox . As a positive and negative control, respectively, we assayed the binding of p67 phox to a pentadecapeptide, corre-sponding to residues 357-371 of p47 phox , and its scrambled form; this peptide is representative of the C-terminal prolinerich region of p47 phox and was found by peptide walking to bind p67 phox with high affinity (29). As an additional negative control, we measured the binding of p47 phox to the same peptide and its scrambled form. As shown in Fig. 7, A and B, binding to PR domain peptides of p47 phox (to peptides 151-165 and 153-167) and p67 phox (to peptides 145-159 and 147-161) is sequence-specific, with no or only minimal binding to the scrambled forms. This pattern follows that of the binding of p67 phox to p47 phox peptide 357-371 (Fig. 7B). As expected, p47 phox did not bind to the autologous p47 phox peptide 357-371 and to its scrambled form. Binding of p47 phox to peptides 47-61 and 49 -63 is only in part sequence-specific, as shown by significant binding to the scrambled forms (Fig. 7A). In view of the hydrophilic character of these peptides and of their positive charge, similar to that of p47 phox , the nature of this nonspecific binding is unknown. Binding of p67 phox to peptide 81-95 (the major p67 phox -binding peptide of cluster B) is sequence-specific (Fig.  7B). We conclude that the dominant, although not the only, element governing binding of p47 phox and p67 phox to p22 phox peptides is the amino acid sequence of the peptides.

DISCUSSION
In this report we describe the application of peptide walking to the identification of sites in the primary sequence of the small cytochrome b 559 subunit p22 phox , participating in the assembly of the NADPH oxidase complex of phagocytes. In the past, identification of such domains rested on the use of blotting and "pulldown" assays, the two-hybrid system, mutagenesis, the use of individual p22 phox peptides, and plasmon resonance methodology. Peptide walking was used previously to identify similar functional domains in the cytosolic NADPH oxidase components Rac1 (28) and p47 phox (29) and led to the identification, in both components, of interaction domains that were not detected by other approaches.
We now subjected to analysis a subunit of a membrane protein, the topology of which was predicted on the basis of hydropathy plots of the primary sequence, combined with the use of antibodies, cell permeabilization, and proteases (40), peptide phage display libraries (41), and antibody imprint analysis (42). Two forms of peptide walking were employed. In the first, we examined the potential of overlapping p22 phox 15-mer peptides to inhibit amphiphile-dependent NADPH oxidase activation in a rigorously standardized semi-recombinant cellfree system. This methodology provides no information on the identity of proteins interacting with p22 phox in the course of NADPH oxidase assembly. In the second approach, p22 phox peptides were used as affinity traps to detect eventual binding of p47 phox , p67 phox , or Rac1 and establish the location of the binding site in the p22 phox sequence. Ideally, the two methodologies should complement and confirm each other.
When we engaged in this project, the dominant view was that p22 phox serves as an adapter protein mediating the interaction between p47 phox and the cytochrome b 559 heterodimer. This was based principally on the finding that a fusion protein consisting of the two SH3 domains of p47 phox binds to a prolinerich region in p22 phox , consisting of residues 151-160 (14,15). A peptide corresponding to residues 149 -162 of p22 phox interfered with binding of p47 phox SH3 domains to the cytoplasmic tail of p22 phox (15). Exchanging Pro-156 to Gln in p22 phox abolished binding of p47 phox . Such a point mutation was described in a CGD patient, who had quantitatively and spectroscopically normal cytochrome b 559 but a nonfunctional NADPH oxidase (13,43). A different approach to the identification of interaction domains in p22 phox was introduced by Kanegasaki and coworkers (44) who showed that a peptide corresponding to res- FIG. 7. Sequence specificity of p22 phox peptides binding p47 phox and p67 phox . A, binding of p47 phox to selected synthetic p22 phox peptides, belonging to clusters A and PR, and to a proline-rich p47 phox peptide, in native and scrambled forms, was measured. B, binding of p67 phox to selected synthetic p22 phox peptides, belonging to clusters B and PR, and to a proline-rich p47 phox peptide, in native and scrambled forms, was measured. All peptides were Ն70% pure. idues 176 -195 of p22 phox inhibited SDS-induced NADPH oxidase activation in a cell-free system, albeit with a rather high IC 50 (36 M); the peptide also bound p47 phox in the presence of SDS. This peptide corresponds to a second proline-rich region, C-terminal to that located at residues 151-160. Random-sequence peptide phage display screening of p47 phox also yielded two proline-rich sequences as potential sites on p22 phox interacting with p47 phox (residues 156 -160 and 177-183) (45). These overlapped the domains described in Refs. 14, 15, and 44; however, an attempt to confirm the inhibition of NADPH oxidase activation by p22 phox peptide 175-195 failed (45), in agreement with the failure of the same peptide to interfere with the binding of p47 phox SH3 domains to p22 phox (15). The considerable difference in the ability of various methodologies to identify interaction domains in p22 phox is also illustrated by the finding that a peptide, corresponding to the consensus domain on p22 phox binding p47 phox (residues 149 -162), was inefficient in inhibiting NADPH oxidase activation and exhibited no change in IC 50 when Pro-156 was replaced by Gln (13).
The possibility that regions other than those located in the C-terminal cytosolic tail of p22 phox might be involved in NADPH oxidase assembly was raised by the finding that a peptide corresponding to residues 82-95 of p22 phox was a potent inhibitor of cell-free NADPH oxidase activation (IC 50 ϭ 13 M) (46). It was originally thought that this region might participate in the binding together of the two cytochrome b 559 subunits, but as a more likely explanation, it was suggested that the region might be exposed to the cytosol and participate in interaction with cytosolic components. The authors proposed that residues 87-89 (YYV) represent the sequence-specific core of this region.
The combined peptide walking approach, used in the present investigation, led us to propose the following 6 points. 1) The p22 phox subunit of cytochrome b 559 serves as an anchor not only for p47 phox but also for p67 phox by direct interaction with the latter component. 2) p47 phox binds not only to the proline-rich region located at residues 151-160 in the C-terminal cytosolic part of p22 phox but also to a domain, consisting of residues 51-63, located on a loop exposed to the cytosol. 3) p67 phox shares with p47 phox the ability to bind to the proline-rich region (residues 151-160) and also binds to two additional domains, in the cytosolic loop (residues 81-91) and at the root of the cytosolic tail (residues 111-115). 4) The affinity of p67 phox for both the cytosolic loop and the cytosolic tail of p22 phox appears to be lower than that of p47 phox . 5). Surprisingly, binding of both p47 phox and p67 phox to proline-rich p22 phox peptides occurred in the absence of an anionic amphiphile and was not enhanced in its presence, suggesting that opening of intramolecular bonds in 47 phox and p67 phox (14,15) is not a prerequisite for binding. 6) Binding of p47 phox and p67 phox to their target domains in the cytosolic loop and at the root of the cytosolic tail is mediated by interactions other than those involving SH3 domains and proline-rich regions, the nature of which remains to be established.
The definition of domains by inhibition of NADPH oxidase activation overlaps only partially that derived by binding of p47 phox and p67 phox . This is evident in Fig. 1, at the level of individual peptides, and in Fig. 8, which presents a synthesis of all our findings. The only major domain revealed by inhibition of activation with no binding equivalent is b. This region is probably exposed on the external surface of the membrane and is not accessible to cytosolic proteins (see Fig.  9); the mechanism of inhibition by corresponding peptides is unclear. Domain c corresponds closely to the p47 phox binding domain A. Domains d and e overlap partially the two p67 phox domains B and C; in both cases the domains defined by inhibition of activation are larger and extend more in the C-terminal direction (Fig. 8). The most striking discrepancy between the two detection systems was, however, the inability of peptides corresponding to the proline-rich domain 151-160, binding both p47 phox and p67 phox , to affect NADPH FIG. 8. Putative functional domains in the amino acid sequence of p22 phox involved in NADPH oxidase assembly. In the top half of the figure, domains in the p22 phox sequence, revealed by inhibition of NADPH oxidase activation, are highlighted in yellow; their nomenclature, in boldface lowercase italic lettering, appears above the sequence. Residues in dark red boldface represent domains involved in binding of p47 phox ; their nomenclature, in dark red boldface uppercase italic lettering, appears below the sequence. Residues in green boldface represent domains involved in binding of p67 phox ; their nomenclature, in green boldface uppercase italic lettering, appears below the sequence. Residues in the proline-rich domain binding both p47 phox and p67 phox are in blue boldface, marked PR, in blue boldface uppercase italics, below the sequence. All domains, whether defined by inhibition of NADPH oxidase activation or binding of p47 phox or p67 phox , are listed and juxtaposed in the bottom half of the figure. with the existence of a second binding site for p67 phox (domain C; residues 111-115). This model allows for the simultaneous binding of p47 phox and p67 phox at more than one site on the cytosolic face of p22 phox . It is also in good agreement with clinical data showing that missense mutations, resulting in complete lack of protein, cause amino acid substitutions within the transmembrane regions, whereas mutations that do not affect the synthesis of p22 phox , as well as polymorphisms, involve residues located outside the membrane (54).