Nox2/gp91phox (where phox is phagocyte oxidase) is the catalytic membrane subunit of the granulocyte NADPH oxidase complex involved in host defence. The current model of membrane topology of Nox2 is based upon the identification of glycosylation sites, of regions that interact with the regulatory cytosolic factors and of the epitopes recognized by antibodies. So far, the localization of the N-terminus of Nox2 was only speculative. In order to clarify this localization, we raised a polyclonal antiserum against the N-terminal sequence M1GNWVAVNEGL11. Purified antibodies recognize the mature protein as a broad band at 91 kDa (glycosylated form) or a band at 55 kDa after deglycosylation. Immunocytochemistry and flow-cytometry analysis show a strong binding of the anti-N-terminal antibodies to differentiated HL60 cells and neutrophils respectively, after permeabilization only. The N-terminus of Nox2 is therefore present in the mature protein and is located to the cytoplasmic side of the plasma membrane.
- anti-peptide antibody
- confocal microscopy
- flow cytometry
- membrane topology
- NADPH oxidase
Nox2/gp91phox (where phox is phagocyte oxidase) is the catalytic membrane subunit of the NADPH oxidase, an enzyme complex found in phagocytic leucocytes which catalyses the formation of superoxide anion (O2•−), a reactive oxygen species involved in the host defence against pathogens. It is an integral membrane protein of 570 amino acids, which is predicted to contain multiple transmembrane domains [1,2]. The N-terminal part of the protein was shown to be involved in haem binding through histidine residues 101, 115, 209 and 222, located in helices III (101 and 115) and V (209 and 222) . After proteolytic treatment of partially purified cytochrome b558, a spectrally stable fragment comprising the whole N-terminus and ending at amino acids 320 or 363 was recovered . gp91phox has been proposed to contain the binding sites for the NADPH oxidase prosthetic group, FAD, and the substrate, NADPH [5,6]. Based on a weak primary structure similarity between the C-terminal domain of gp91phox and ferredoxin–NADP+ reductase (FNR), a structural model for the nucleotide-binding region of the gp91phox C-terminus was predicted based on the known structure of FNR . As FNR is a soluble protein and the C-terminal region contains the predicted substrate-binding pocket, it is proposed that the C-terminal half of gp91phox is soluble and has a cytoplasmic location. Human gp91phox is a highly glycosylated protein and appears on Western blots as a broad smear from 120 kDa to 60 kDa, average approx. 91 kDa. The glycosylation has previously been demonstrated to be N-linked  and to be attached to residues Asn131 and Asn148 located in the loop between predicted transmembrane helices III and IV (numbering of the helices according to the main current model, Figure 1), and Asn239 between transmembrane helices V and VI of gp91phox . Therefore the current models for the membrane topology of gp91phox place these three glycosylation sites to the exterior of the membrane. Monoclonal antibody (mAb) 7D5 [10,11] and polyclonal anti-L123 (S151YLNFARKRIKNPEGGLYLAVTL173) antibody  have both been reported to bind to intact neutrophils, demonstrating that their epitopes are exposed to the exterior of the cell. On the other hand, mAbs 54.1 , NL7  and polyclonal anti-Lc (I551SNSESGPRGVHFIFNKENF570) antibody  can only bind following permeabilization of the membrane, therefore placing their individual epitopes on the cytosolic side of the membrane. The use of phage-display libraries has mapped the epitopes for these mAbs to: I160KNP163 plus R226IVRG230 for 7D5 , confirming the extracellular location of region 151–173, P383KIAVDGP390 for mAb 54.1 , E498KDVITGL505 for mAb 48  and E498KDVITGLK506 for NL7 . The intracellular location of the epitopes for mAb 54.1, mAb 48, mAb NL7 and anti-Lc antibody provides supporting evidence that the C-terminal domain of gp91phox protrudes into the cytoplasm of the cell. After activation, p47phox, p67phox and Rac, the cytosolic subunits of the NADPH oxidase, interact with the membrane components to form a functional enzyme complex. Using phage-display libraries, three regions of gp91phox which bind p47phox were identified as S86TRVRRQL93, F450EWFADLL457 and E554SGPRGVHFIF564 . Residues 86–93 are between predicted transmembrane domain I and II and contain a high density of positively charged residues. Arg91 and Arg92 from this region were shown to be critical for the activity of cytochrome b558 , presumably by interaction with p47phox. For residues 450–457, their involvement in p47phox binding is currently questioned by the data obtained with mAb NL7 . Recently, a number of homologues of gp91phox, named Nox, have been identified in the human , Drosophila, Caenorhabditis elegans (reviewed in ), Dictyostelium (B. Lardy, M. Bof, L. Aubry, M. H. Paclet, F. Morel, M. Satre and G. Klein, unpublished work), Arabidopsis  and other plant [21,22] genomes. Although the amino acid sequences of the human homologues show varying degrees of identity , they all exhibit a similar hydropathy plot. An alternative to the main current model has recently been proposed by Cheng et al.  based on the suggestion that the N-terminal 30–35 amino acids of Nox1, Nox2, Nox3 and Nox4 act as cleavable signal peptides. In the case of gp91phox/Nox2, the predicted signal peptide would include the 30 amino acids which form the first predicted transmembrane domain. This hypothesis is in direct conflict with the N-terminal sequence data of gp91phox that was originally determined from the protein purified from granulocytes . The amino acid sequence clarified the position of the initiating AUG codon within the identified gene [1,24]. Therefore most models predict that the N-terminal amino acids of gp91phox would protrude on the cytosolic side of the membrane, as shown in Figure 1, but they are not supported by experimental evidence.
To clearly identify the localization of the N-terminus of the protein, we raised an antiserum against the N-terminal sequence of the protein M1GNWAVNEGL10-Y (with a tyrosine residue added to the C-terminal end as indicated). As a control we used an antibody directed against the peptide L153NFARKRIKNPEGGLY168 that had previously been shown to be extracellular . By performing Western blotting, flow cytometry analysis and confocal microscopy with these antibodies, we clearly demonstrated that the N-terminus part of Nox2 is present in the mature neutrophil protein and that it is localized in the cytoplasm. Our study provides the first experimental evidence for the intracellular location of the N-terminus of Nox2.
The following materials were supplied by the indicated companies: N-glycosidase F (recombinant, Escherichia coli), n-octyl glucoside (Roche diagnostics, Meylan, France); ECL® (enhanced chemiluminescence) Western blotting detection reagents (Amersham Biosciences, Orsay, France); anti-rabbit IgG (whole molecule) FITC-conjugate (Sigma-Aldrich, Saint-Quentin-Fallavier, France); Cy™2-labelled anti-rabbit polyclonal antibody (Jackson ImmunoResearch Laboratories, West Grove, PA, U.S.A.); peptides (Immunograde, 80% pure) were obtained from Neosystem (Strasbourg, France). RPMI medium was from Invitrogen (Paisley, Renfrewshire, Scotland, U.K.), Inject ovalbumin from Pierce (Perbio, Brebieres, France), Affi-Gel 10 from Bio-Rad (Marnes-la-Coquette, France), Centricon 30 from Millipore (Saint-Quentin-en-Yvelines, France). All other chemicals were from Sigma.
Production of antipeptide antibodies
For the N-terminal peptide (M1GNWAVNEGLS11-Y) (PNter), a tyrosine residue was added C-terminally to the sequence of the peptide to allow coupling to the carrier protein. The second peptide (P153–168) corresponded to residues L153NFARKRIKNPEGGLY168 with the tyrosine present in the sequence itself. After protection of lysine residues, peptides were coupled to ovalbumin as carrier protein, as described previously , using bis-diazobenzidine. The peptide–ovalbumin (coupled peptide) was used for immunization. Pre-immune serum was taken and five injections of 0.5 ml of coupled peptide were made at 3 week intervals. The rabbits were bled 10 to 15 days after the fifth injection.
Affinity purification of antibodies
A peptide–L-tyrosine resin was made in two steps. First, L-tyrosine was covalently attached to an activated medium (Affi-Gel 10) and residual binding sites were saturated by 1-h incubation with 300 mM ethanolamine. The peptide (2 μmol for 5 ml of resin) was then added and covalently bound to the L-tyrosine resin using bis-diazobenzidine  as above. Approx. 10 ml of serum was incubated with the peptide–L-tyrosine resin (5 ml of dry resin) for 1 h at 4 °C with gentle shaking. The resin was washed first at high ionic strength, i.e. 50 mM Tris/HCl (pH 8.0)/500 mM NaCl, then twice with PBS (pH 7.4; 2.7 mM KCl, 137 mM NaCl, 1.5 mM KH2PO4 and 8.1 mM Na2HPO4). For elution, the resin was poured into a column, washed with PBS and elution was performed by 5 ml of 0.1 M glycine/HCl, pH 1.8, buffer (pH adjusted at 25 °C). The pH was immediately readjusted in eluted fractions with 0.1 volume of 1 M Trizma base. Elution was followed by monitoring the A280. The fractions containing proteins (immunoglobulins) were pooled, and the pH was adjusted to neutral. Antibodies were concentrated on a Centricon 30 device and the buffer was exchanged for PBS. Purified antibodies were supplemented with 50% glycerol and stored at −20 °C until use.
Cytochrome b558 purified from neutrophils (50 pmol) was submitted to SDS/10%-(w/v)-PAGE. Proteins were detected by staining with silver nitrate .
The membrane fraction was isolated from human neutrophils and cytochrome b558 was purified as previously described . Crude membrane (100 μg) or purified cytochrome b558 (8 pmol) was subjected to SDS/PAGE (10% gel) and submitted to Western blotting and probed with 1:1000 dilution of the antibody. The binding of the first antibody was detected following the addition of horseradish peroxidase-conjugated anti-rabbit secondary antibody, using the ECL® detection kit. A control sample was incubated in parallel with pre-immune serum. In some experiments, antibodies were pre-incubated with the peptide used for immunization (50 μg/ml) for 1 h prior to probing the nitro-cellulose membrane.
Purified cytochrome b558 (10 pmol) was denatured by heating at 100 °C for 2 min in presence of 0.2% (w/v) SDS and 1% (v/v) β-mercaptoethanol. At the end of the incubation, 34 mM n-octyl glucoside and 2 units of N-glycosidase F were added for an overnight incubation at 37 °C [28,29].
Culture and differentiation of HL60 cells
HL60 cells, a human promyelocytic cell line , were maintained in RPMI 1640 supplemented with 10% (v/v) foetal calf serum, 50 units/ml penicillin and 50 μg/ml streptomycin, at 37 °C, in the presence of 5% CO2. The cultures of HL60 cell were initiated at 5×105 cells/ml on a weekly basis. Differentiation of HL60 cells was induced by the addition of 1.25% (v/v) DMSO to cells at 5×105 cell/ml. The cells were harvested by centrifugation at 800 g for 10 min, on the fifth day following the addition of DMSO .
Immunocytochemistry on HL60 cells by confocal microscopy
Differentiated and undifferentiated HL60 cells were harvested by centrifugation at 800 g for 10 min. The cells were washed twice in 10 ml PBS, pelleted at 600 g for 10 min, prior to being resuspended in 0.5 ml of PBS. The cells were allowed to settle on round glass coverslips for 10 min at 25 °C and washed twice for 10 min with PBS. Cells for permeabilization were fixed with 4% (v/v) formaldehyde in PBS for 10 min immediately followed by permeabilization with 0.2% (v/v) Triton X-100 in PBS for 2 min. Non-permeabilized cells were washed twice with PBS. Both Triton-X-100-treated and non-treated cells were incubated (three 10 min incubations) with 0.2% (w/v) BSA in PBS (BSA/PBS) prior to 1-h incubation with primary antibody diluted 1:500 in BSA/PBS. The cells were washed (three 10 min washes) with BSA/PBS prior to incubation with Cy™2-labelled anti-rabbit secondary antibody for 1 h and subsequent washes. The location and intensity of the Cy™2 fluorescence were recorded on an inverted Bio-Rad MRC 600 Confocal microscope as described previously .
The differentiation of HL60 cells was confirmed by their ability to generate superoxide in response to activators of the NADPH oxidase . The release of superoxide was measured continuously as the reduction of cytochrome c (550–540 nm) at 37 °C, using a double-beam spectrophotometer. The NADPH oxidase was activated by the addition of 50 nM PMA and inhibited by either 10 μM diphenylene iodonium or 50 μg/ml superoxide dismutase.
Human neutrophils were isolated from citrated venous blood of healthy volunteers using a 33% (v/v) Hypaque–Ficoll gradient. After 20 min centrifugation at 800 g at 20 °C, the pellet was submitted to an hypotonic lysis for 5 to 15 min on ice . After 5 min centrifugation at 350 g at 4 °C, the neutrophil pellet was collected and washed once in PBS. Neutrophils were suspended in PBS/BSA/CaCl2 [PBS containing 0.2% (w/v) BSA and 0.5 mM CaCl2] at the concentration of 107 cells/ml, and fixed on ice by addition of an equal volume of 2% (w/v) paraformaldehyde. After fixation (15 min), cells were centrifuged, washed once in PBS/BSA/CaCl2 and resuspended in saponin buffer [PBS/BSA/CaCl2 containing 0.01% (w/v) saponin] at the concentration of 107 cells/ml for a 10-min incubation on ice. After permeabilization, cells (5×105) were incubated on ice for 30 min with 100 μl of primary rabbit antibody (50 μg/ml of non-immune IgG or anti-peptide IgG in saponin buffer), then washed twice with 500 μl of saponin buffer, and resuspended in 150 μl of FITC-conjugated goat anti-rabbit antibody, diluted 1:200 in saponin buffer. After 30 min incubation on ice, cells were washed twice with 500 μl of saponin buffer before being resuspended in 500 μl of PBS/BSA/CaCl2. Binding of antibodies to intact cells was performed as described above without treatment with saponin buffer. Fluorescence intensity (FL1) of the FITC-labelled neutrophils was measured on a FACScalibur (Becton Dickinson) cytometer.
Immunodetection of gp91phox by Western blot
Although most models place the N-terminus of Nox2 on the cytoplasmic face of the plasma membrane, experimental evidence for this localization was missing. An antiserum was therefore raised to the synthetic peptide M1GNWAVNEGLS12-Y (PNter), the first 11 amino acids encoded by the Nox2 cDNA. The peptide L153NFARKRIKNPEGGLY168 (P153–168), part of the L123 peptide shown to have an extracellular localization , was used for immunization under the same conditions.
On a Western blot, the antibody directed against the N-terminal peptide (anti-PNter) labelled specifically only one broad band at approx. 91 kDa in granulocyte membranes (Figure 2A). This band was also detected by the antibody directed against the peptide corresponding to amino acids 153–168 of gp91phox (anti-P153–168) (Figure 2A), but was absent from the blots probed with non-immune immunoglobulins or pre-incubated with the competitor peptide (Figure 2A). To confirm that the band detected at 91 kDa was gp91phox, a Western blot was performed on cytochrome b558 purified from neutrophils. The purity of the fraction was analysed by SDS/PAGE (Figure 2B, lane 1) and by its specific activity (15–20 nmol of haem b/mg of protein). Both p22phox and gp91phox were present in this fraction (Figure 2B, lane 1), but only gp91phox was detected by the anti-PNter, as shown by the single broad band at 91 kDa that disappeared in presence of the competitor peptide (Figure 2B, lanes G, anti-PNter versus anti-PNter+peptide). A non-specific band at 28 kDa in the membrane fraction was not p22phox, since it was not labelled when purified cytochrome b558 was loaded on to the gel.
gp91phox is a highly glycosylated protein which travels on a gel as a broad band with average molecular mass of 91 kDa. The treatment of purified cytochrome b558 with N-glycosidase F led to a specific decrease in the molecular mass of gp91phox (from 91 kDa to 55 kDa), as reported previously [28,29]. After deglycosylation, the binding of anti-PNter to purified cytochrome b558 was not affected, and the specific labelled band was at 55 kDa, as expected for the deglycosylated form of gp91phox (Figure 2B, lanes D, anti-PNter versus anti-PNter+peptide). The detection of gp91phox by the antibody in both the isolated granulocyte membrane fraction and the purified cytochrome b558 demonstrates that the epitope is present in the mature protein and is not cleaved as the result of the post-translational processing. Therefore the first 30 amino acids of Nox2 are not a cleavable signal peptide.
Immunodetection of gp91phox on human neutrophils by flow cytometry
To check the localization of the N-terminal peptide of gp91phox, we performed flow cytometry on human neutrophils using affinity-purified antibodies to the N-terminal peptide (anti-PNter) and to the 153–168 sequence (anti-P153–168). A negative control with non-immune antibodies gave only a background labelling.
As shown in Figure 3(A), anti-PNter did not stain the cells extracellularly. The signal was similar to that obtained with non-specific antibody (Figure 3A, continuous line versus broken line). After permeabilization by saponin, a clear label was observed (Figure 3C, continuous line versus broken line). So far, only two antibodies, mAb 7D5 and L123, have been shown to recognize an extracellular epitope. We therefore used the polyclonal antibody anti-P153–168 that recognizes part of L123 peptide as a control for an extracellular localization. Binding of anti-P153–168 was clearly above the background for intact cells, compared with anti-PNter (Figure 3B versus 3A), but labelling was slightly increased when neutrophils were permeabilized by saponin (Figure 3D). The fluorescence increase in permeabilized neutrophils could be explained by the binding of anti-P153–168 to the cytochrome b558 molecules inserted in the granule membranes, molecules that were not available in intact cells.
These experiments clearly show that the N-terminus of gp91phox is not accessible to antibodies when neutrophils are intact and that it becomes available when they are permeabilized.
Immunodetection of gp91phox on differentiated HL60 cells by confocal microscopy
The incubation of the promyelocytic HL60 cells with DMSO for 5–7 days induces them to differentiate into neutrophil-like cells. The process is associated with an increase in the ability of these cells to generate superoxide in response to an oxidase activator  and with an increase in expression of the individual subunits of the NADPH oxidase, including gp91phox (Nox2) .
To determine the cellular location of the N-terminal part of Nox2, we incubated both Triton X-100 treated and non-treated HL60 cells with the N-terminal Nox2 polyclonal antibody (Figure 4). The binding of the antibody and its cellular location were determined using a fluorescently labelled secondary antibody and a confocal microscope.
A strong immunofluorescence was observed with differentiated HL60 cells which had been permeabilized with Triton X-100, incubated with anti-PNter (Figure 4A). A much lower intensity of emitted fluorescence was observed from differentiated HL60 cells which had not been treated with Triton X-100 (Figure 4B), and corresponded to the low non-specific binding measured in presence of pre-immune serum in both permeabilized and intact cells (Figures 4E and 4F respectively). In undifferentiated HL60 cells, a low but specific labelling was observed after permeabilization, indicating that a small amount of cytochrome b558 was already present in the membrane of these cells (Figure 4C versus 4D). As predicted, gp91phox antigen shows increased expression upon differentiation of HL60 cells into neutrophil-like cells. In the absence of the anti-PNter antibody, no immunostaining was observed for differentiated HL60 cells either with or without permeabilization (results not shown), indicating that the immunostaining observed in Figure 4(A) is not due to cross-reactivity of the secondary antibody with the HL60 cells.
The failure of the antibody to detect its antigen in the absence of an agent which disrupted the integrity of the plasma membrane confirms that the N-terminal of Nox2 protrudes on the cytosolic face of the plasma membrane. We therefore conclude that the N-terminus has an intracellular location. The membrane topology of gp91phox, including all previously published experimental data (see the Figure for details and references) and the present study, is illustrated in Figure 1.
In the absence of crystallographic studies, topography analysis of membrane proteins relies on several approaches. In the case of gp91phox/Nox2, the number of predicted transmembrane regions ranges from 5 to 8. Many immunological approaches were undertaken, but most antibodies map to the C-terminal region and to the external loop that contains glycosylation sites. Data have accumulated that the C-terminal region of the protein is soluble. It has a cytoplasmic location and it is the site of binding for FAD, NADPH and the cytosolic factors, in particular p47phox. Another ‘hotspot’ for antibody recognition is the extracellular region comprising two loops and three effective glycosylation sites. There were no experimental data to confirm whether the N-terminal region could be placed either on the cytoplasmic or on the external side of the plasma membrane. The lack of information about this region could be explained by the small size, the poor solubility and immunogenicity of this part of the protein. In the present paper, we used for the first time a polyclonal antibody directed against the N-terminal peptide of gp91phox to clearly identify the localization of this region.
The N-terminal 31 amino acids of Nox2 are present in the mature protein and do not act as a cleavable signal peptide as proposed by Cheng et al. . This result is in agreement with the identification by sequencing of the N-terminal 11 amino acids of gp91phox in a fragment purified after endoproteinase treatment of cytochrome b558 . The similarity in the hydropathy plots between gp91phox and its homologues suggests that their N-termini would also protrude to the cytosolic face of the membrane. This information is very important and shows that gp91phox possesses an even number of transmembrane helices (most likely to be 6), as the C- and N-terminal parts of the protein co-localize in the cytosol. Moreover, the protrusion of the N-terminal part of Nox2 (and its homologues) in the cytosol raises the possibility that this small region could be a docking site for proteins involved in the regulation of the NADPH oxidase activity.
The work was supported by a grant number H0604 from the Arthritis Research Campaign, U.K., by grants from the Ministère de l'Enseignement supérieur de la Recherche et Technologie, Paris, the Région Rhône Alpes, programme Mobilité Internationale Rhône Alpes 2001, délégation de Grenoble, the Fondation pour la Recherche Médicale, Isère, and from the French association ‘La Ligue Nationale contre le Cancer’. We thank Helen Kennedy and Robert Meech, Department of Physiology, University of Bristol, for the use of the Bio-Rad MRC 600 Confocal Microscope, and Marie-Claire Joseph for technical assistance.
Abbreviations: ECL, enhanced chemiluminescence; FNR, ferredoxin–NADP+ reductase; mAb, monoclonal antibody; Nox, NADPH oxidase; phox, phagocyte oxidase
- The Biochemical Society, London