The flavocytochrome b558 of the phagocyte NADPH oxidase complex comprises two membrane proteins, a glycosylated gp91phox and a non-glycosylated p22phox. Gp91phox contains all of the redox carriers necessary to reduce molecular oxygen to superoxide using NADPH. The capacity of gp91phox to produce superoxide in the absence of its membrane partner p22phox has been little studied. In the present study, we have generated in Pichia pastoris for the first time an active form of bovine gp91phox able to carry out the entire NADPH oxidase activity in the absence of p22phox. Collected information on the maturation and the activity of the recombinant gp91phox and the participation of individual cytosolic subunits in the active complex allowed us to propose, in the absence of p22phox, an unconventional stabilized complex compared with the heterodimer.
The NADPH oxidase complex generates ROS (reactive oxygen species) by transferring electrons from NADPH across biological membranes to dioxygen. These reactions occur through the membrane catalytic subunit which presents conserved structures in all members of the NADPH oxidase family [1,2]. It harbours an NADPH- and a FAD-binding site in the cytosolic C-terminus region and two non-identical b-type haem groups [3,4] localized in the six-transmembrane domain that mediate the final steps of electron transfer to molecular oxygen . In phagocytes, NOX2 (NADPH oxidase 2), also known as gp91phox, is a highly glycosylated protein  and is localized in plasma membranes in close association with its membrane partner p22phox [7–9], together forming the so-called cytb558 (flavocytochrome b558). NOX2 appears to be the most widely distributed among the NOX isoforms and has been localized in a large number of cells and tissues (phagocytes, neurons, cardiomyocytes, endothelial cells etc.) . The stimulus-dependent activation of gp91phox occurs through multiple protein–protein interactions with cytosolic factors (p67phox, p47phox p40phox and a small GTP-binding protein, Rac1/Rac2). Previous studies demonstrate that phagocytes from p22phox-deficient patients have no detectable gp91phox [11,–13]. In transfected CHO (Chinese-hamster ovary) cells, Dinauer and colleagues demonstrated that p22phox-dependent maturation of gp91phox carbohydrates, cell surface expression of gp91phox and the enzymatic function of cytb558 are relatively correlated. In these cells, low levels of gp91phox are expressed at the cell membrane in the absence of p22phox co-expression  and it was concluded that when p22phox is absent, no oxidase activity was possible . Until now, all studies and interpretations of the phagocyte NADPH oxidase function have agreed that the presence of the p22phox membrane subunit is crucial for the activity of gp91phox [14,16]. Nevertheless, many studies demonstrated that the C-terminally truncated forms of gp91phox obtained in heterologous expression systems can perform diaphorase activity by electron and protons transfer from NADPH to FAD in the absence of p22phox [17,18]. However, until now, no heterologous expression of an active form of the monomer gp91phox has been described.
Using a yeast heterologous expression system, we aimed at producing gp91phox and p22phox in distinct cells in order to better understand their relationship. To our surprise, we engineered an active recombinant monomeric gp91phox. Moreover when reconstituted in liposomes, it had NADPH oxidase activity with no prerequisite for other protein partners. Our findings also shed a new light on the protein–protein interactions essential for their membrane recruitment.
MATERIALS AND METHODS
Expression of bovine gp91phox in Pichia pastoris cells
We created the transgenic yeast strains X33/gp91, X33/p22 and X33/gp91-p22 by the integration of recombinant vectors in the genomic DNA of Pichia pastoris cells strain X33 (Life Technologies). The coding sequences of the bovine gp91phox and p22phox were subcloned separately into pPICZαA vectors at the XhoI and XbaI restriction sites, in-frame with the α-factor signal peptide (to enhance the membrane addressing of the recombinant proteins) under the control of the AOX1 (alcohol oxidase 1) promoter. Then the expression cassette from the recombinant pPICZαA/p22phox was cut out using BamHI/BglII restriction enzymes and then inserted at the BamHI site of the recombinant pPICZαA/His-gp91phox to obtain the chimaeric vector containing the two expression cassettes (containing their own AOX1 promoter) in the same vector. Transgenic strains were confirmed by PCR amplification using specific primer sets described previously . Expression of recombinant proteins was carried out by growing the transgenic yeast clones under methanol induction for 72 h as described previously . A negative control X33/° strain was generated by transforming X33 cells with the empty vector pPICZαA. After culture, harvested cells were stored at −80°C.
Membrane preparation and protein analysis
Cell pellets were thawed, broken using glass beads and membrane fractions were collected by centrifugation at 100000 g for 120 min at 4°C. The pelleted membrane fractions were resuspended in 50 mM Tris/HCl (pH 8), 120 mM NaCl, 10% glycerol and 1 mM PMSF and loaded on to a discontinuous sucrose gradient (60%, 40% and 20% sucrose) as described previously . After overnight centrifugation at 110000 g, the essential pure plasma membranes were removed from the 60% and 40% interface, diluted 4-fold and pelleted at 130000 g for 90 min. The cytb558-enriched plasma membrane fraction were resuspended in 1 mM EDTA, 30% sucrose and 20 mM Tris/HCl (pH 8.0) and stored at −20°C. Reduced-minus-oxidized spectra were performed using a double beam Uvikon 943 spectrophotometer (Kontron Instruments). Excess dithionite was added to the cuvette to reduce the sample before obtaining the spectra. The absorbance difference between the peak at 428 nm and the trough at 411 nm was used to determine the cytb558 concentration using the molar absorption coefficient ε428–411=200 mM−1·cm−1 . Deglycosylation was performed with the PNGase F (peptide N-glycosidase F; New England Biolabs) treatment as described by the manufacturer.
For the Western blotting assay, anti-His antibody conjugated to HRP (horseradish peroxidase; Clontech), rabbit anti-gp91 (54.1) and anti-p22 (FL-195) antibodies (Santa Cruz Biotechnology) were used for specific identification at a dilution of 1:1000. Anti-rabbit (NA934V; GE Healthcare) and anti-mouse (NA931; GE Healthcare) IgG monoclonal antibodies were used at a 1:15000 dilution as secondary antibodies to detect the anti-gp91 and anti-p22 primary antibodies respectively. Signal development was performed by the ECL Plus Advance Western Blotting Detection Kit (GE Healthcare).
NADPH oxidase activity in cell-free assay
Plasma membranes isolated on sucrose gradient or PLs (proteoliposomes) were used for the oxidase activity assays. Unless indicated, 2.24 nM gp91phox was mixed with recombinant cytosolic factors [p47phox (0.89 μM), p67phox (0.42 μM) and Rac1Q61L (0.34 μM)] and AA (arachidonic acid; Sigma–Aldrich) at the required concentrations in a total reaction volume of 500 μl. Cytosolic proteins were prepared as described in . After 5 min of incubation at 25°C, the superoxide production was initiated by the addition of 400 μM NADPH and followed by the SOD (superoxide dismutase)-inhibitable reduction of ferricytochrome c (100 μM). The rate of generated superoxide was calculated using a value of Δε550nm (cytc) of 21.1 mM−1·cm−1 . The plasma membrane fraction isolated from the Pichia strain transformed by the empty vector was used as control for the cell-free oxidase activity under the same experimental conditions as described above.
Detection of cytosolic proteins interacting with gp91phox membrane
The cytosolic proteins were incubated combined or separately for 5 min at 25°C with 3 μg of membrane fraction containing gp91phox (2.24 nM, 0.16 μg) with 650 μM AA. Cytosolic proteins were used at the following concentrations: p47phox (0.89 μM), p67phox (0.42 μM) and Rac1Q61L (0.34 μM). The protein mixture was loaded on to a sucrose gradient (20–60%). After overnight centrifugation at 110000 g, the membrane-assembled and non-assembled fractions were separated. Co-localized cytosolic subunits with membrane were found in the pellet and soluble (unassembled) proteins remained mostly in the supernatant. The pellet and supernatant were analysed by Western blotting using mouse anti-human p47phox, mouse anti-human p67phox (kindly provided by Dr Florence Lederer) or mouse anti-Rac1 (ARC03, Cytoskeleton) antibodies.
Purification of the gp91phox and PL reconstitution
Gp91phox-containing membranes were solubilized with 39 mM DDM (n-dodecyl-β-D-maltoside; 2%). The extract was purified using a column of Ni-NTA (Ni2+-nitrilotriacetate) superflow Sepharose (GE Healthcare) followed by gel-filtration chromatography using S200 Sephadex resin (GE Healthcare) in TBS buffer [20 mM Tris/HCl (pH 7.5) and 50 mM NaCl] supplemented with 0.49 mM (0.025%) DDM. The purity of the protein was checked by SDS/PAGE (10% gel).
To reconstitute the purified protein into PLs, a stock solution of DOPC (1,2-dioleoyl-sn-glycero-3-phosphocholine; P6354; Sigma–Aldrich; purity ≥99%) was diluted in PBS (pH 7.4) and sonicated to induce the formation of lipid vesicles. Purified His–gp91phox (200 μg) was mixed with vesicles to obtain a 5:1 lipid to protein ratio (w/w). Detergent was removed by incubating the proteoliposome mixture with 1% (w/v) Bio-Beads SM2 (Bio-Rad Laboratories) under rotating condition over 2 h at room temperature.
Gp91phox maturation is partially determined by the presence of p22phox
On the basis of the successful expression of bovine cytb558 in P. pastoris , we intended to improve the expression level of recombinant proteins by using another Pichia strain (X33). X33/gp91, X33/p22 and X33/gp91-p22 transgenic clones were generated by integration of the coding sequences of the bovine gp91phox and p22phox, separately or simultaneously, into the genomic DNA of X33 cells using the pPICZαA vector (Figure 1A). For X33/gp91 and X33/gp91-p22 cell membranes, we have clearly identified the characteristic peaks of the neutrophil cytb558 in the reduced-minus-oxidized difference spectra; the Soret band at ~428 nm and α band at 558 nm indicating haem incorporation in gp91phox proteins (Supplementary Figure S1 at http://www.biochemj.org/bj/462/bj4620337add.htm). Unfortunately these peaks were not easily distinguishable from similar, but smaller, peaks measured in the membrane spectrum of the negative control strain containing the empty vector (X33/°) and which have been attributed to the endogenous FRE1 (ferric reductase 1) protein described in yeast cells . Interestingly, we observed that the contribution of FRE1 to the spectra was largely decreased in X33/p22 cell membranes suggesting that heterologous expression of membrane proteins that are targeted to the yeast plasma membranes may reduce the presence of FRE1.
Constructs in X33 Pichia strains and analysis of the expression and glycosylation of recombinant gp91phox
Analysis of transgenic cell membrane contents showed that the expression of gp91phox protein occurred under different forms depending on the presence or absence of p22phox protein (Figure 1B). After 72 h of methanol induction, when co-expressed with p22phox, gp91phox was predominantly in two N-glycosylation states (~80 and heavily glycosylated >100 kDa) as shown previously . However, without p22phox gp91phox was found with molecular masses of approximately 65 and 80 kDa (Figure 1B). The 80 kDa form is likely to correspond to the above mentioned glycosylated form, whereas the 65 kDa form might correspond to the theoretically expected size (68 kDa) of the gp91phox protein without post-translational modifications. In support of this hypothesis, PNGase F treatment followed by Western blot analysis using anti-gp91 monoclonal antibodies (Figure 1C) gave rise to the major 65 kDa deglycosylated form.
These findings demonstrate that an incomplete glycosylation of gp91phox subunit correlated with the absence of the p22phox protein.
Functional activity of gp91phox in the absence of p22phox
In the literature it has been proposed that gp91phox is not functional without p22phox . To find out whether this is also the case for Nox2 expressed in P. pastoris, the capacity to produce superoxide by the recombinant gp91phox monomer was tested and compared with the X33/gp91–p22 heterodimer under the same cell-free assay conditions, i.e. incubation with an anionic amphiphile AA (commonly used as an NADPH oxidase activator in cell-free systems) and the cytosolic partners (p67phox, p47phox and Rac). Surprisingly, a significant rate of superoxide was measured with X33/gp91 membranes, whereas the control X33/° membranes, also tested for superoxide production under the same experimental conditions, showed no SOD-inhibitable NADPH oxidase activity. We therefore investigated more thoroughly the behaviour of X33/gp91 membranes and their activation by AA. The X33/gp91phox membranes displayed the AA-dependent bell-shaped activation curve (Figure 2A) commonly described for neutrophil NADPH oxidase, except that a lesser amount of AA, compared with X33/gp91–p22 membranes, was needed for optimal oxidase activity.
Functional properties of the recombinant gp91phox
To further characterize the recombinant form of gp91phox, we determined its kinetic parameters deduced from the Michaelis–Menten fitting of the NADPH concentration dependence of oxidase activity (Figure 2B). We obtained a Km value of ~20 μM and an average turnover of ~181 mol of superoxide/s per mol of gp91phox. These values are in the range of those obtained for cytb558 neutrophils [25,26] or recombinant heterodimer (H. Souabni and L. Baciou, unpublished work) indicating that the absence of p22phox did not affect the apparent substrate affinity and the superoxide-generation capacity. Known as a good in vitro inhibitor of NADPH oxidase, DPI (diphenyliodonium) showed an inhibition effect on the oxidase activity when added in the activity assay mixture with the X33/gp91 membrane (Figure 2C), although at a higher concentration than for neutrophil cytb558. It is likely that midpoint redox potentials of FAD (Flox/Flred) might be different in gp91phox without p22phox and thus might have modified the kinetics of inhibition.
How does gp91phox interact as a monomer with the cytosolic partners?
The existence of an active monomeric gp91phox in Pichia membranes brings, for the first time, the opportunity to investigate its interactions with the cytosolic proteins individually omitting the interactions related to p22phox. For that purpose, the activation of gp91phox by each cytosolic partner singly or by pairs was measured (Figure 3A). In the absence of the in vitro stimulant AA, but in combined presence of all cytosolic partners, almost no activity was detected confirming the prerequisite of the presence of AA for optimal oxidase activity. AA, itself, leads to a basal activity of approximately 20% compared with the entire complex although statistical analysis indicates no significant difference with the activity in the presence of only cytosolic partners (~2%).
Activation of gp91phox by the cytosolic subunits and co-sedimentation analysis with membranes expressing gp91phox
In the presence of AA, when p47phox or Rac1 were added as the sole cytosolic partner, the oxidase activity reached approximately 40% and 60% compared with the entire complex respectively. In contrast, the presence of p67phox alone had no significant effect suggesting an AA-inactive form of p67phox. The total activity was recovered only when p67phox and Rac1 partners were present together, although the couples of p47phox/p67phox or Rac1/p47phox both restored high oxidase activities.
To correlate gp91phox activation by individual interactions with cytosolic proteins, we performed co-localization experiments (Figure 3B). In these experiments, Western blots were performed to identify if the cytosolic subunits individually or together translocate to X33/gp91phox or X33/° membranes. The p47phox, p67phox and Rac proteins were incubated either individually or together with the membrane fractions in the presence of AA. Then, the membrane-translocated cytosolic proteins were separated from unassembled proteins on a sucrose gradient. The membrane co-migrating proteins (assembled fraction) were found in the pellet, whereas the unbound proteins remained in the supernatant. Rac, partially p47phox and to a less extent p67phox co-localized with the membranes of X33/gp91phox, whereas in the same experiments using X33/° membranes p47phox and p67phox do not translocate with the P. pastoris membranes. In contrast Rac interacts with X33/° membranes probably through unspecific polybasic domain interaction. When all cytosolic proteins (p47phox, p67phox and Rac) are added together, p47phox and p67phox co-localized more efficiently with the X33/gp91phox membranes. The absence of co-localization with the X33/° membrane indicates that the interactions of the cytosolic subunits with the X33/gp91phox membrane were specific to the presence of gp91phox.
These findings corroborate the activation process of individual cytosolic protein described above, since the low activity detected in the presence of AA-activated p67phox with gp91phox derived from its low capacity to develop protein–protein interactions with gp91phox. In the absence of p22phox, the active complex interactions might have been modified in which Rac appeared to be a good stabilizer of the complex.
Is the recombinant gp91phox still active after detergent extraction and relipidation?
To assess the stability of gp91phox in the presence of commonly employed detergent, the protein was solubilized in DDM, purified and the absorption spectrum of the purified protein was recorded. The purified protein has been identified by Western blot analysis using anti-gp91phox monoclonal antibodies (Figure 4).
Purification and identification of the recombinant gp91phox subunit
To measure the oxidase activity of gp91phox following purification, different experimental conditions were performed. The purified gp91phox in detergent showed almost no activity (Table 1). However, when complemented with the X33/° and X33/p22 membranes (which are inactive and have no capacity to produce any superoxide anions) by incubation for 30 min with DDM–gp91phox, 60% and 75% of the NADPH oxidase activity was restored respectively (Table 1). This finding indicates a successful rescue of the oxidase activity of the non-active DDM–gp91phox by its membrane re-insertion. Moreover, the complementation with p22phox-containing membranes leading to higher oxidase activity suggests possible interactions of both membrane proteins and, furthermore, underlines the putative effect of p22phox on gp91phox activity.
|Type of membrane tested||DDM–gp91phox||Mol of O2•−/s per mol of gp91phox||Activity (%)|
|Type of membrane tested||DDM–gp91phox||Mol of O2•−/s per mol of gp91phox||Activity (%)|
The purified DDM–gp91phox protein was also incorporated into DOPC lipid bilayers previously shown to allow NADPH oxidase activity . The liposome reconstitution restored the SOD-sensitive NADPH oxidase activity and the similar AA-dependent bell-shaped activity as for X33/gp91 membranes, except that no cytosolic proteins were necessary (Figure 5). Incubation with or without cytosolic proteins did not improve the NADPH oxidase activity. Our results show that the superoxide production detected in yeast X33/gp91 membranes does originate from the monomeric gp91phox, which displays kinetic properties very similar to that of neutrophils or recombinant heterodimer p22phox–gp91phox and can be activated in a reconstituted lipid environment by AA alone. These results stressed the importance of the membrane environment for the stability and activity of gp91phox.
NADPH oxidase activity of the reconstituted gp91phox in DOPC liposomes
In phagocytes, cytb558 is the key component of the NADPH oxidase complex since it contains all redox intermediates for electron transfer and can produce superoxide in the absence of the other cytosolic partners [27–29]. However, there was no evidence to date that gp91phox can produce superoxide alone independently of p22phox association. Rather it was postulated that the heterodimerization was essential for functional cytb558. The heterologous expression of gp91phox reported to date produced only truncated forms containing the C-terminal domain of gp91phox that exhibits diaphorase activity [17,18]. On the basis of a new expression of cytb558 in the methylotrophic yeast P. pastoris , we produced separately gp91phox and p22phox and provided a model for the study of the activation process involving only the catalytic subunit gp91phox.
p22phox is not required but contributes to gp91phox glycosylation
The impact of the presence of p22phox on the expression and maturation of gp91phox was studied by immunoblot analysis. As shown in Figure 1, the co-expressed p22phox and gp91phox in yeast results in a mature glycosylated (mannosylation) form of gp91phox that migrates as a 100 kDa protein as previously described in . The absence of p22phox failed to promote the maturation of gp91phox carbohydrate at the same level. Instead the mature glycosylated form of gp91phox migrates as an 80 kDa protein that is shifted to a focused 65 kDa band after deglycosylation. The heterodimer formation is not required to initiate the maturation and to target the protein to the plasma membrane, but increases significantly the abundance of the mature gp91phox glycoprotein [30–33]. In contrast with what was observed in CHO gp91 cells [31,34], in Pichia the absence of p22phox does not prevent the glycosylation, but affects its pattern (80 kDa instead of 100 kDa) and does not reduce the expression of the mature gp91phox and its targeting to plasma membranes. A possible explanation could be that, in P. pastoris, a p22phox-like protein could replace p22phox in the maturation process of gp91phox. We have performed bioinformatics analyses (BLAST) to identify a putative p22phox-like protein in the P. pastoris genome; however, no significant homologies have been found. In addition Western blots using anti-p22phox monoclonal antibody have been performed with the P. pastoris membrane that does not express p22phox (X33/° control membranes). Under those experimental conditions, no signals were detected indicating, at the least, no recognition of p22phox epitopes. We also observed for gp91phox, but not for the heterodimer, a decrease in thermal stability and duration of the oxidase activity, but also enhancement of sensitivity to proteolysis. These observations are consistent with the idea that, in the absence of heterodimerization in COS cells, the low expression of the mature form of gp91phox (91 kDa) is due to protein instability .
Although the glycosylation pattern associated with mammalian proteins from P. pastoris is different from that of higher mammalian proteins, the question remains whether the different glycosylation pattern are dependent of the presence of p22phox. Three glycosylation sites (Asn131, Asn148 and Asn239) have been identified in two different extracellular loops of gp91phox (loops C and E) . An additional glycosylation site in loop C was discovered for patients with variant CGD (chronic granulomatous disease) in which p22phox was absent . We can speculate that the locations of glycosylation (i.e. external loops) point to structural elements that are functionally linked to the gp91phox–p22phox association processes. This hypothesis is supported by modelling approaches that identified a large extracellular loop in p22phox as an extensive interacting region with gp91phox . We propose that the proteins expressed in yeast benefit heterodimerization processes which adapt to the evolving glycan shield.
p22phox is not required for superoxide production
Previous studies have shown that p22phox is functionally important for enzyme activity and that particular regions, such as the N-terminal 11 amino acids and the C-terminal proline-rich-region of p22phox, are required . We have studied the impact of the absence of p22phox on the capacity of gp91phox to produce superoxide. In contrast with all expectations, gp91phox exhibited a classical NADPH oxidase activity. When expressed in Pichia in the absence of p22phox, gp91phox showed kinetic properties (Km value of ~20 μM and turnover of ~180 s−1) similar to what is measured in neutrophils  or with recombinant heterodimer (H. Souabni and L. Baciou, unpublished work), showed also sensitivity to DPI and was activated by the cytosolic proteins and AA. Similarly, the AA activation follows a characteristic bell-shape.
The active form of gp91phox in the absence of p22phox might be correlated with structural changes mediated by different glycosylation levels. As reviewed by Imperiali and O’Connor , glycosylation can affect the local secondary structure of proteins by facilitating the formation of segments of secondary structure and can play a crucial role in directing the protein folding pathway. In turn, glycosylation can increase the stability of the residues in the vicinity of the glycosylation site. The modulation of local structure might serve to enhance the overall stability of gp91phox or might enable gp91phox to perform some required function at given sites, such as NADPH or cytosolic protein binding. Hence, this can explain the ability of this new form of gp91phox to produce superoxide in the absence of p22phox.
Alternatively, it has been reported in the literature that the presence of anionic phospholipids promotes NADPH oxidase activity [27,40,41]. Comparison of anionic phospholipid content between plasma membranes indicates a higher content of negatively charged phospholipids in yeast [42–44] than in neutrophils . This would suggest that P. pastoris membranes provide, naturally, a remarkable environment for gp91phox activation and may reflect the phagocyte-activated state of gp91phox observed within mammal cells, which have been shown to change its lipid composition during neutrophil activation.
Following detergent extraction, the relipidation with zwitterionic DOPC rescued NADPH oxidase activity. However, this enzyme activity depends only on the presence of the AA as activator. This result suggests that, in the liposomes, anionic phospholipid contaminants might be present arising either from impure DOPC preparations, as was previously shown by Koshkin and Pick  for relipidated cytb558 in PC (phosphatidylcholine) vesicles, or from P. pastoris lipids co-purified with gp91phox. The mixing assay of DDM–gp91phox purified with X33/° and X33/p22 membranes is suitable for proper assembly of a functional heterodimer and leads to superoxide production. The restored NADPH oxidase activity was higher with membrane containing p22phox, but still did not reach X33/gp91–p22 or X33/gp91membrane activity levels. This result raises the question of a structural role for p22phox and of the membrane allowing a restructuration of gp91phox that was inactive in detergent. It has been shown that structural changes in cytb558 are observed by cis-AA activation [46,–48], but not by the isomer trans-AA , underlining important structural determinant in the activation process. The present study provides strong evidence that AA can directly target and activate gp91phox.
Interaction pattern of gp91phox with oxidase cytosolic factors
Because p22phox possesses a C-terminal proline-rich region known to provide a high affinity for the SH3 (Src homology 3) domains of p47phox during NADPH oxidase assembly [49,50], we studied to what extent the interaction of monomeric gp91phox with the cytosolic partners has been modified. This behaviour was important to clarify since we provide for the first time the opportunity to highlight the possible interactions of the cytosolic partners with gp91phox in the absence of p22phox. Our data show that in the Pichia membrane, p67phox alone cannot activate gp91phox (absence of a membrane attachment signal), but are required to be tethered by Rac. This is consistent with in vitro experiments and the proposed model in which Rac must associate simultaneously both with p67phox and the membrane to activate NADPH oxidase [51,52]. However, our data also showed that the presence of p47phox, and in particular Rac, alone enhances substantially the production of superoxide anion by gp91phox expressed in P. pastoris membranes. We also observed that gp91phox in the P. pastoris membrane can be activated with quasi-equal potency by AA combined with Rac and p47phox or p47phox and p67phox. We can conclude that, in the absence of p22phox, in the P. pastoris membranes Rac and p47phox are the essential elements for the stabilization of an active complex.
In P. pastoris membranes, the interactions are likely to be different compared with the scheme currently proposed when p22phox is associated to gp91phox. As shown in Figure 6, we propose a model in which gp91phox acts as the central docking component interacting directly with the cytosolic proteins p47phox, p67phox and Rac, independently of the presence of p22phox. Despite being unconventional, this assembly leads to active NADPH oxidase. This indicates that the presence of p22phox offers interacting domains to stabilize p67phox on gp91phox. In turn, the Rac protein serves as a membrane tether for the juxtaposition of p67phox to gp91phox protein. Thus, in addition to its active form, this recombinant gp91phox conserves and consolidates the potential interactions with their partners in the oxidase complex. The proposed protein–protein interaction might help to extend our understanding of the molecular mechanisms that govern oxidase regulation in particular for Nox proteins such as Nox5, where no p22phox, was identified or Nox4, where the presence of p22phox is not necessary.
Schematic representation of the active gp91phox subunit and potential interactions with cytosolic proteins in the absence of p22phox in P. pastoris membranes
alcohol oxidase 1
ferric reductase 1
NADPH oxidase 2
F, peptide N-glycosidase F
Src homology 3
Aymen Ezzine and Laura Baciou designed the research; Aymen Ezzine and Hager Souabni performed the research; Aymen Ezzine, Hager Souabni, Tania Bizouarn and Laura Baciou analysed the data; and Aymen Ezzine and Laura Baciou wrote the paper.
We thank Florence Lederer for stimulating discussions and careful reading of the paper before submission.
This work was supported by the Agence Nationale de la Recherche [project number ANR-2010-BLAN-1536-01].