Atypical Membrane-embedded Phosphatidylinositol 3,4-Bisphosphate (PI(3,4)P2)-binding Site on p47phox Phox Homology (PX) Domain Revealed by NMR*

Background: The p47phox PX domain binds PI(3,4)P2 during assembly and activation of NADPH oxidase. Results: The PI(3,4)P2-binding site on the p47phox PX domain was identified by NMR. Conclusion: The position of the phosphoinositide-binding site on p47phox PX is unique. Significance: Our results provide insight into a particular role of p47phox PX in the assembly and activation of phagocyte NADPH oxidase. The Phox homology (PX) domain is a functional module that targets membranes through specific interactions with phosphoinositides. The p47phox PX domain preferably binds phosphatidylinositol 3,4-bisphosphate (PI(3,4)P2) and plays a pivotal role in the assembly of phagocyte NADPH oxidase. We describe the PI(3,4)P2 binding mode of the p47phox PX domain as identified by a transferred cross-saturation experiment. The identified PI(3,4)P2-binding site, which includes the residues of helices α1 and α1′ and the following loop up to the distorted left-handed PPII helix, is located at a unique position, as compared with the phosphoinositide-binding sites of all other PX domains characterized thus far. Mutational analyses corroborated the results of the transferred cross-saturation experiments. Moreover, experiments with intact cells demonstrated the importance of this unique binding site for the function of the NADPH oxidase. The low affinity and selectivity of the atypical phosphoinositide-binding site on the p47phox PX domain suggest that different types of phosphoinositides sequentially bind to the p47phox PX domain, allowing the regulation of the multiple events that characterize the assembly and activation of phagocyte NADPH oxidase.

The Phox homology (PX) domain is a functional module that targets membranes through specific interactions with phosphoinositides. The p47 phox PX domain preferably binds phosphatidylinositol 3,4-bisphosphate (PI(3,4)P 2 ) and plays a pivotal role in the assembly of phagocyte NADPH oxidase. We describe the PI(3,4)P 2 binding mode of the p47 phox PX domain as identified by a transferred cross-saturation experiment. The identified PI(3,4)P 2 -binding site, which includes the residues of helices ␣1 and ␣1 and the following loop up to the distorted left-handed PP II helix, is located at a unique position, as compared with the phosphoinositide-binding sites of all other PX domains characterized thus far. Mutational analyses corroborated the results of the transferred cross-saturation experiments. Moreover, experiments with intact cells demonstrated the importance of this unique binding site for the function of the NADPH oxidase. The low affinity and selectivity of the atypical phosphoinositide-binding site on the p47 phox PX domain suggest that different types of phosphoinositides sequentially bind to the p47 phox PX domain, allowing the regulation of the multiple events that characterize the assembly and activation of phagocyte NADPH oxidase.
The phagocyte NADPH oxidase is a multicomponent enzyme that produces reactive oxygen species as a defense against invading microorganisms. Its importance for innate immunity is exemplified by chronic granulomatous disease, a genetic impairment of NADPH oxidase that leads to recurrent diseases and death (1,2). The NADPH oxidase consists of cyto-solic (p47 phox , p40 phox , p67 phox , and small GTPase Rac1 or Rac2) and membrane-integrated components (gp91 phox and p22 phox ). The enzyme switches from an inactive state to an active state and is capable of producing reactive oxygen species when the cytosolic components are recruited to the membrane and interact with the membrane-integrated components to form a functional complex (3)(4)(5)(6).
p47 phox and p40 phox contain PX 2 domains, which are crucial for the recruitment of cytosolic components (7)(8)(9)(10). In addition, the PX domain of p47 phox is important for the activation of the NADPH oxidase, based on the observation that in stimulated cells the presence of a PX-truncated p47 phox mutant led to decreased superoxide production (10).
The PX domain is a functional module that targets membranes through a specific interaction with membrane-embedded phosphoinositides (11,12). Although the p40 phox PX domain specifically binds to PI3P, the p47 phox PX domain reportedly binds to PI(3,4)P 2 , PA, and PtdSer.
The crystal structure of p40 phox in complex with C 4 -PI3P has been solved. In the structure, Arg-60 interacts with the 3-phosphate (through its backbone) and the inositol 2-hydroxyl and 1-phosphate groups (through its side chain) of the bound PI3P. However, Lys-92 interacts strongly with the 1-phosphate group of PI3P. The three-dimensional structures of p47 phox PX alone have been determined by solution NMR (13) and x-ray crystallography (14). However, the structure of p47 phox PX in complex with its ligand has not been reported. The crystal structure of its free form revealed two sulfate molecules bound in two distinct sites (14). One of the sulfates is hydrogen-bonded to Arg-43 and Thr-45 on the C-terminal end of the ␤3 strand, and it coincides with the 3-phosphate of the bound PI3P in the crystal structure of p40 phox PX. The second sulfate is hydrogen-bonded to His-51 and Lys-55 on the ␣1 helix and to Arg-70 on the adjacent loop that leads to the distorted left-handed PP II helix. Based on these results and surface plasmon resonance-based mutational analyses, the first sulfate-binding site was proposed to be the PI(3,4)P 2 -binding site, and the second sulfate-binding site was considered to be the PA-binding site. In the crystal structure of the CISK PX domain, however, one of the two bound sulfate groups was found in a completely unexpected area, which probably represents a nonspecific binding site (15). Therefore, it is unclear whether these sulfate-binding sites of p47 phox PX are the actual ligand-binding sites, and thus the mechanism of the broad ligand specificity of p47 phox PX is still unclear.
Under physiological conditions, phosphoinositides interact with effector proteins in the context of lipid bilayers. In some cases, the consideration of membrane-mimicking conditions can greatly affect the results of in vitro studies. For instance, the ability of the actin regulatory protein N-WASP to sense phosphatidylinositol 4,5-bisphosphate density in the membrane is lost when soluble phosphatidylinositol 4,5-bisphosphate or inositol trisphosphate headgroups are used instead (16). As another example, the actual affinity of the vacuole protein sorting HOPS for PI3P can be revealed only by a liposome assay (17).
NMR is a useful technique for investigations of biological systems within a certain size limit (currently in the area of 800 kDa) (18). To overcome this inherent size limitation in NMR, appropriate NMR methodologies and sample preparations are necessary. This is particularly true in cases involving biomembranes, because a protein bound on a lipid bilayer is a huge biological structure that cannot be observed easily by NMR. Transferred cross-saturation (TCS) is an NMR method that enables the identification of the residues of protein ligands in close proximity to huge (Ͼ100 kDa) complexes (19 -21). Successful applications of TCS for huge and/or inhomogeneous systems have previously been reported (22)(23)(24)(25)(26)(27)(28)(29).
In this study, we employed TCS experiments to identify the PI(3,4)P 2 -containing membrane-interacting interface on the p47 phox PX domain, using liposomes as a membrane-mimicking environment. Mutational analyses verified the results of TCS and revealed that Lys-55 is important for PI(3,4)P 2 recognition. Moreover, we demonstrate that Lys-55 is also important for the proper function of phagocyte NADPH oxidase. Altogether, our data indicate that the p47 phox PX domain binds PI(3,4)P 2 in a unique area, as compared with the phosphoinositide-binding sites of all of the other PX domains characterized thus far (30 -35).

EXPERIMENTAL PROCEDURES
Sample Preparation-The human p47 phox PX domain (residues 1-128), encoded on the pGEX-2T vector (GE Healthcare), was expressed and purified as reported (13). The human p40 phox PX domain (residues 1-167) on the pGEX-2T vector was expressed and purified similarly. The DNA encoding the Vam7p PX domain (residues 1-122) was prepared by total synthesis using the ligation-independent cloning method. The Vam7p PX on the pET43a vector was expressed as an untagged protein and purified by a combination of anion and cation exchange chromatography. Mutants were prepared according to the QuikChange (Stratagene) protocol, and their expression and purification were performed similarly. Their correct folding was confirmed by an inspection of their HSQC spectra. The uniformly double-labeled 2 H-15 N PX domain was prepared by growing cells in 99% D 2 O minimal medium, with 1 g liter Ϫ1 15 NH 4 Cl and 3 g liter Ϫ1 97% [ 2 H]glucose. The cells were grown at 37°C to an A 600 of 1.0 and then were induced by 1 mM isopropyl 1-thio-␤-D-galactopyranoside and incubated overnight at 25°C. The cells were harvested and lysed by sonication, and the supernatant was applied to glutathione-Sepharose 4B beads (GE Healthcare). The GST tag was removed by thrombin, and the isolated 2 H-15 N PX domain was purified according to the reported procedure (13). Large unilamellar vesicles were prepared by extrusion. Briefly, the required amounts of POPC (Avanti), POPE (Avanti), and di-C 16 PI(3,4)P 2 (Echelon),or di-C 16 PI3P (Echelon) were mixed in CHCl 3 , dried over a N 2 stream, and placed under vacuum overnight. The lipid mixture was resuspended in 5 mM [ 2 H]MES, pH 5.5, 90% D 2 O, or 5 mM MES, pH 5.5, 100% H 2 O at 37°C. Hydrated lipids were subjected to extrusion using an Avanti mini-extruder with a 100-nm polycarbonate filter. Phospholipid concentrations were determined by a phosphate assay (36). The size of the prepared liposomes was estimated by Dynamic Light Scattering experiments, using a DynaPro instrument at 25°C with 10% laser intensity. The phosphoinositide-containing and control liposomes gave an average radius of 60.4 and 65.9 nm, respectively, with high homogeneity (low % polydispersity, 100% intensity, and 100% mass). The lamellarity of prepared liposomes was calculated as reported previously (37), and for the phosphoinositide-containing and control liposomes, the lamellarities were estimated to be 0.8 and 0.72, respectively. The stability of the liposomes during the interactions with proteins was evaluated by Dynamic Light Scattering experiments, in which the size of the liposomes was recorded over time. No significant change was observed in the size of the liposomes. The stability of the liposomes was further evaluated by fluorescence measurements of liposomes loaded with carboxyfluorescein. No significant change in self-quenching of carboxyfluorescein was observed over time. For titration experiments, stock solutions of di-C 4 PI(3,4)P 2 (Echelon), di-C 6 PA (Avanti), and di-C 6 PtdSer (Avanti) were prepared in 5 mM MES, 5 mM DTT, pH 5.5. Phosphates (Wako) were mixed to produce a phosphate buffer at pH 5.5.
Fluorescence Assay-Cys-less p47 phox PX (Cys-98 and Cys-111 were substituted with Ala and Ser, respectively) was labeled at position 21 (CPM-(21)p47 phox PX) or alternatively at position 65 (CPM-(65)p47 phox PX) with the fluorescent probe CPM, after the introduction of the required Cys in the labeling site. For labeling, a 10-fold molar excess of CPM was used, in 50 mM Tris, pH 7.4, and 100 mM NaCl, at 4°C overnight. The reaction mixture was cleared by centrifugation, and the excess CPM was removed by passage through a PD-10 column. The purity of CPM-labeled PX was assessed by comparison of a Coomassie Brilliant Blue-stained gel and a UV-transilluminated SDS-polyacrylamide gel for the same sample. The labeling efficiency was calculated from a molar extinction coefficient of 33,000 (M Ϫ1 cm Ϫ1 ) for CPM, and it was Ͼ90%. The correct folding of CPMlabeled p47 phox PX and its ability to interact with C 4 -PI(3,4)P 2 were confirmed by an inspection of its HSQC spectra with and without ligand. For the 1,5-IAEDANS labeling of Cys-less p47 phox PX at position 21 or position 80, a 20-fold molar excess of 1,5-IAEDANS was used in 20 mM Tris, 100 mM KCl, pH 7.5, at 20°C, overnight and gave Ͼ90% labeled product. The stock solution of lipid vesicles and the sample solution of labeled p47 phox PX were prepared in 5 mM MES, pH 5.5. Measurements were conducted on an RF-5300PC Shimadzu spectrofluorophotometer. The CPM-labeled sample was excited at 384 nm, and the emission was detected at 470 nm at 22°C. The IAEDANS-labeled sample was excited at 336 nm, and the emission was detected at 490 nm at 22°C.
NMR Spectroscopy-HSQC spectra were recorded on a Bruker Avance 500 or Bruker Avance 600 spectrometer equipped with a cryo-probe. Assignment data were provided by Dr. Hidekazu Hiroaki, Kobe University. We reconfirmed them by a combination of three-dimensional NMR spectra and selective labeling. The backbone NMR assignment data of p40 phox PX and T45R/P78K/K55E mutant of p47 phox PX were obtained by standard three-dimensional NMR measurements. The previously reported NMR assignment data of Vam7p PX were used (38). For the determination of the lamellarity of prepared liposomes, the one-dimensional 31 P NMR spectra were recorded on a Bruker DRX 400 spectrometer. Unless otherwise stated, measurements were conducted at 25°C, with NMR samples The TCS experiment was performed as described previously (19,22). Selective saturation of the proton content of the liposome components was achieved by using the WURST-2 broadband decoupling scheme. The saturation frequency was set at 1.3 ppm, and the saturation time was 1 s. All spectra were processed with the program nmrPipe, and data analysis was accomplished with the Sparky program (University of California, San Francisco).
Activation System for the Phagocyte NADPH Oxidase-The cDNA for human wild type (WT) p47 phox was prepared as described previously (39). A mutation leading to the K55A substitution in p47 phox was introduced by PCR-mediated site-directed mutageneses. The cDNAs were ligated to a pEF-BOSbased vector for expression of p47 phox (WT) and p47 phox (K55A) as HA-tagged protein in mammalian cells (39). All the constructs were sequenced for confirmation of their identities. The K562 cells stably expressing gp91 phox and p67 phox (K562-gp91 phox /p67 phox cells) were prepared as described previously (40). The cells (5 ϫ 10 6 cells/ml) were electroporated in the presence of 60 g of pEF-BOS-HA-p47 phox (WT) or pEF-BOS-HA-p47 phox (K55A) at 170 V and 1,070 microfarads using a GenePulser (Bio-Rad). After culture for 48 h, the superoxideproducing activity of these transfected cells was determined by superoxide dismutase-inhibitable chemiluminescence with an enhancer-containing luminol-based detection system (DIOGENES, National Diagnostics), as described previously (39,41). After the addition of DIOGENES, cells (1.0 ϫ 10 6 cells/ ml) were preincubated for 5 min at 37°C and subsequently stimulated with 200 ng/ml phorbol 12-myristate 13-acetate. The chemiluminescence was monitored at 37°C using a luminometer (Auto Lumat LB953; EG&G Berthold). For estimation of protein levels of HA-p47 phox , p67 phox , and p22 phox , cell lysates of K562 cells were subjected to SDS-PAGE, transferred to a polyvinylidene difluoride membrane (Millipore), and probed with an anti-HA monoclonal antibody (Covance Research Products), an anti-p67 phox monoclonal antibody (BD Biosciences), or anti-p22 phox polyclonal antibodies (Santa Cruz Biotechnology). The blots were developed using ECL-prime (GE Healthcare) for visualization of the antibodies, as described previously (41).
Interaction of p47 phox PX Domain with an Insoluble Long Lipid Chain, C 16 -PI(3,4)P 2 , Embedded in Liposomes-In this study, we performed TCS experiments to investigate the interactions of C 16 -PI(3,4)P 2 -containing liposomes with the p47 phox PX domain to identify the C 16 -PI(3,4)P 2 -containing membrane-interacting interface on the p47 phox PX domain. An overview of the experiment is shown in Fig. 2A. The TCS experiments are performed with an excess amount of the [U-2 H, 15 N]p47 phox PX domain, relative to the C 16 -PI(3,4)P 2 embedded in the liposomes. In the TCS experiment, the complex is irradiated at a frequency centered at the aliphatic proton resonances, which exclusively affects the liposomes, because almost no aliphatic protons exist in the deuterated PX domain. The saturation in the liposomes is transferred to all of the hydrogen atoms of the liposomes, through intramolecular cross-relaxation (spin diffusion). The saturation is not limited within the liposomes but is transferred to the residues at the interface of the PX domain bound to the liposomes, through intermolecular cross-relaxation (cross-saturation). If the complexes have a sufficiently fast exchange rate between the free and bound states, then the saturation in the interface residues is effectively transferred to the free state of the PX domain. As a result, the irradiation causes a reduction in the intensities of the resonances from the interface residues of the PX domain in the free state.
The efficiency of the TCS effects depends on various experimental conditions, such as the binding constant between ligands and receptors, the molar ratio of the receptors to the ligands, and the molecular weight of the receptors. We investi- The TCS experiment was conducted in the selective saturation mode for both the aliphatic (lipid tail) and nonaliphatic (polar head) protons of the liposome components. The intensity reduction ratio (RI) was calculated as ((I 0 Ϫ I)/I 0 )⅐100, where I 0 represents the intensity per residue of p47 phox PX without saturation, and I is the intensity per residue with saturation. Residues are color-coded according to their intensity reduction, with red and orange for strong (RI Ͼ0.3) and moderate (0.25 Ͻ RI Ͻ 0.3) reduction, respectively. Asterisks mark the position of proline residues. The secondary structure (based on its crystal structure) is shown. The error bars represent the root sum square of (noise level/signal intensity) in the two spectra with and without irradiation. D, mapping of residues displaying strong intensity reductions in the TCS experiment with PI(3,4)P 2 -containing liposomes, on a ribbon representation and on the surface of the crystal structure of p47 phox PX. The residues are color-coded as in C. The bound sulfates are shown in a stick representation. The molecular diagrams were generated by Web Lab Viewer Pro (Molecular Simulations, Inc.). gated the effects of the physical parameters on TCS by a simulation (44), which indicated that the TCS effects would be sufficiently observed under the present experimental condition (supplemental Fig. S1).
To investigate whether the TCS method can accurately identify the membrane-interacting interface on a peripheral membrane protein, the method was first employed on a model protein, the Vam7p PX domain. The location of the phosphoinositide-binding site on this PX domain is well established. The results of the TCS experiment for the interaction of the Vam7p PX domain with C 16 -PI3P-containing liposomes are shown in supplemental Fig. S2. Significant intensity reductions were selectively observed in its reported PI3P binding pocket.
Subsequently, we applied the TCS experiments to identify the C 16 -PI(3,4)P 2 -containing membrane-interacting interface on the p47 phox PX domain. The spectra with and without irradiation are shown in Fig. 2B. Several residues, including Glu-49, Lys-55, and Met-57, are affected by the irradiation, although others are not. As shown in Fig. 2, C and D, residues Glu-49 and Phe-58 of helix ␣1 and Ile-71 of the following loop displayed intensity losses within the 0.3-0.5 range. Moderate 0.25-0.3 intensity losses were exhibited by Tyr-48, Lys-55, and Met-57 in helix ␣1, as well as by Gly-63 in the following loop. All of these residues form a well defined area on the surface of the p47 phox PX domain, outlined by helix ␣1, helix ␣1Ј, and the following loop up to the distorted left-handed PP II helix. The ectopic intensity losses shown by Met-27, which is located far away from the main affected area, might be the result of enhanced spin diffusion, due to the large molecular weight of the liposomes.
To investigate the intramolecular saturation effects induced by the residual protons in the [U-2 H, 15 N]p47 phox PX domain, an identical TCS experiment, using the 2 H-15 N PX domain alone and without liposomes, was performed (supplemental Fig. S3, A and B). As a result, no selective intensity reductions were observed for the residues of the p47 phox PX domain. To examine the effect of the nonspecific binding between the p47 phox PX domain and POPC and/or POPE, we performed the TCS experiments, using liposomes without PI(3,4)P 2 (supple-mental Fig. S3C). As a result, no selective intensity reductions were observed for the residues of the p47 phox PX domain.
Based on the above experiments, we conclude that the C 16 -PI(3,4)P 2 -containing membrane-interacting interface on the p47 phox PX domain is the area on the surface of the protein, formed by helix ␣1, helix ␣1Ј, and the following loop up to the PP II helix. Hereafter, this area and the site that correspond to the PI3P-binding site of p40 phox PX are referred to as "atypical binding site" and "typical binding site," respectively.
Interaction of p47 phox PX with Soluble C 4 -PI(3,4)P 2 -To investigate the interaction between the p47 phox PX domain and the polar headgroup of PI(3,4)P 2 , NMR titration experiments with soluble phosphoinositides were performed. Chemical shifts of main chain amide NH and arginine N⑀H groups, which were observed in these experiments, are affected by both the direct binding effects and the conformational change in each residue (45). As shown in Fig. 3A and supplemental Fig. S4, soluble C 4 -PI(3,4)P 2 induced chemical shift perturbations for specific signals on the HSQC spectrum of the wild type p47 phox PX. Most of these perturbed signals were derived from the residues in the loop between helix ␣1 and the PP II area (Fig. 3B), which is located inside the membrane-interacting interface identified by the TCS experiments (Fig. 2D), although several residues that are outside the membrane-interacting interface and within the site that corresponds to the PI3P-binding site of p40 phox PX were also perturbed. The ⑀NH of Arg-70, which coordinates the sulfate within the identified membrane-interacting interface, displayed a strong chemical shift perturbation (Fig. 3C). From the titration curve, a K d value of 38.7 M was obtained. Further structural analyses using intermolecular NOEs were hampered by the low signal intensities of the complex, which are probably due to exchange broadening.
For the comparison of the binding modes, the titration experiment was also performed with the model proteins the p40 phox PX and Vam7p PX domains. As shown in supplemental Fig. S5, the resonances from the residues in the PI3P binding pocket of the p40 phox PX and Vam7p PX domain were selectively shifted upon the addition of soluble C 8 -PI3P.  Fig. S4. B, mapping of residues displaying strong C 4 -PI(3,4)P 2 -induced chemical shift perturbations onto a ribbon representation of the crystal structure of p47 phox PX. The residues are color-coded as in A. C, chemical shift changes for the Arg-⑀NH signals of p47 phox PX upon interaction with C 4 -PI(3,4)P 2 . p47 phox PX (0.2 mM) was incubated with no C 4 -PI(3,4)P 2 (red) and with 0.2 mM C 4 -PI(3,4)P 2 (blue). Peaks were assigned through mutations.

Interaction of p47 phox PX with Various Soluble Ligands-To
clarify the ligand specificity of the p47 phox PX domain, we further investigated the interaction between the p47 phox PX domain and other soluble phosphoinositides. Although the p47 phox PX shows a preference for PI(3,4)P 2 , it can also bind to other types of phosphoinositides, such as PI3P and PI(4,5)P 2 , with lower affinity. They induced similar pattern of chemical shift perturbation on p47 phox PX and therefore are probably accommodated in the same area as the PI(3,4)P 2 (supplemental Fig. S6).
PA and PtdSer have been reported as putative interacting partners of the p47 phox PX domain (14). We assessed the binding of soluble C 4 -PtdSer or C 6 -PA to the p47 phox PX domain by NMR chemical shift perturbation experiments. The addition of 0.5 mM PtdSer did not induce any chemical shift perturbations of the resonances from the p47 phox PX domain (Fig. 4A). However, the NMR signals from the p47 phox PX domain were perturbed upon the addition of 0.5 mM C 6 -PA (Fig. 4A). To examine whether C 6 -PA binds to the PI(3,4)P 2 -interacting area of p47 phox PX, determined by the TCS experiments, competition experiments were performed for the PA-p47 phox PX domain and PI(3,4)P 2 -p47 phox PX domain interactions. As a result, the chemical shifts of the resonances from the 0.2 mM p47 phox PX domain with both 0.3 mM C 4 -PI(3,4)P 2 and 0.6 mM C 6 -PA were similar to those with 0.3 mM C 4 -PI(3,4)P 2 (Fig. 4B) but different from those with 0.6 mM C 6 -PA (Fig. 4B), suggesting that PA and soluble C 4 -PI(3,4)P 2 competitively bind to the p47 phox PX domain.
To obtain further evidence, the p47 phox PX was titrated with phosphate buffer. At a low concentration, the phosphates induced chemical shift changes in the same area on the surface of p47 phox PX as C 4 -PI(3,4)P 2 . In contrast, Thr-45, a residue that coordinates the sulfate outside the identified phosphoinositide-interacting interface, showed a strong chemical shift change only at high phosphate concentrations, suggesting the lack of a specific interaction (supplemental Fig. S7).
Mutational Analyses to Reveal the Residues Important for Phosphoinositide Binding-To confirm the results of the NMR experiments, a series of p47 phox PX mutants were prepared, and their respective affinities for soluble C 4 -PI(3,4)P 2 were assessed by NMR titration experiments ( Fig. 5 and Table 1). The obtained spectra and titration curves of the alanine mutants of Arg-43 and Thr-45, which formed hydrogen bonds with the sulfate outside the identified phosphoinositide interacting area of p47 phox PX in the crystal structure (Fig. 5A), were similar to those of the wild type (Fig. 5B, supplemental Fig. S4B, and Table  1). In contrast, the mutations inside the phosphoinositide interacting area, as determined by the TCS experiments (R70K, H51Q, and H74A), reduced the affinity. Among them, the resonances from the K55A mutant were not perturbed upon the addition of soluble C 4 -PI(3,4)P 2 ( Fig. 5C and Table 1).
Moreover, to examine whether the K55A mutant also abolishes PI(3,4)P 2 binding in the presence of a membrane, a CPMbased fluorescence assay was conducted. Although the fluorescence of the CPM-labeled T45A mutant was increased upon the addition of C 16 -PI(3,4)P 2 -containing liposomes (gray down triangle in supplemental Fig. S8) the K55A mutant was not perturbed (gray diamonds in supplemental Fig. S8). The fluorescence of another mutant of Lys-55, K55E, was also not affected by the addition of C 16 -PI(3,4)P 2 -containing liposomes ( Table  2). The fluorescence intensities of K55A and K55E were also not perturbed upon the addition of liposomes containing C 16 -PI3P, C 16 -PI(4,5)P 2 , and C 16 -PI(3,5)P 2 (data not shown).
Role for Lys-55 of p47 phox PX Domain in Phagocyte Oxidase Activation-We examined the role of Lys-55 in the activation of the NADPH oxidase in intact cells (Fig. 6). K562-gp91 phox / p67 phox cells, which lack p47 phox but contain the other essential components of the phagocyte oxidase (40), were transfected with cDNA for the wild type and the K55A mutant of the HAtagged p47 phox and stimulated with phorbol 12-myristate 13-acetate. As a result, the amount of superoxide produced in cells that express the K55A mutant was ϳ40% less than that in cells that express the wild type. We confirmed that the expression level of p47 phox , p67 phox , and p22 phox are similar in these cells. Thus, Lys-55 would be important for the oxidase activation.
T45R/P78K Mutations Restore the PI3P-binding Site of p40 phox PX on p47 phox PX-We showed that the Lys-55 mutant of p47 phox PX is unable to interact with phosphoinositides. Guided by the structural similarity between p47 phox PX and p40 phox PX, we attempted to establish a p47 phox mutant that mimics the p40 phox PX-like ligand-binding site and ligand specificity. In the crystal structure of p40 phox PX in complex with   Fig. S4B. D, fold increase in the association constant of the C 4 -PI(3,4)P 2 -p47 phox PX interaction for each of the tested mutants, relative to that of the wild type. Association constant values were obtained from NMR titration curves.

TABLE 2 Values of dissociation constants for the interaction of liposome-embedded C 16 -PI(3,4)P 2 with WT and mutants of p47 phox PX, obtained from the CPM-based fluorescence assay
The numbers in parentheses denote the position of the CPM probe. CPM sites are indicated with arrows in Fig. 5A. Two variants with different positions of CPM were prepared to show that the presence of the probe does not affect the binding affinity. The slightly higher dissociation constant for CPM-(65), as compared with that for CPM- (21), is in accordance with the binding site being closer to site 65 than site 21. ND indicates not detectable. K562-gp91 phox /p67 phox cells were transfected with the cDNA for HA-tagged p47 phox (WT) or p47 phox (K55A). A, transfected cells (1.0 ϫ 10 6 cells/ml) were preincubated for 5 min at 37°C and stimulated with phorbol 12-myristate 13-acetate (PMA) (200 ng/ml). Chemiluminescence change by superoxide produced was continuously monitored with DIOGENED, as described under "Experimental Procedures." The superoxide scavenger superoxide dismutase (SOD) (50 g/ml) was added at the indicated time. B, each graph represents superoxide production estimated by chemiluminescence change (the average ϩ S.D.) from four independent transfections. Protein levels of HA-p47 phox , p67 phox , and p22 phox were analyzed by immunoblot with the anti-HA, anti-p67 phox , and anti-p22 phox antibodies, respectively, as described under "Experimental Procedures." C 4 -PI3P, Arg-60 and Lys-92 of p40 phox PX interact with the phosphates at position 3, 1, and 1 of C 4 -PI3P, respectively, and these residues are not conserved in p47 phox PX. Therefore, we further mutated the K55E p47 phox PX at residues Thr-45 (T45R) and Pro-78 (P78K). The T45R mutation aims to restore the role of Arg-60 in p40 phox PX, and the P78K mutation aims to restore that of Lys-92 in p40 phox PX (Fig. 7). The ligand-binding site and ligand specificity of the mutant were examined by NMR titration experiments with soluble C 4 -PI(3,4)P 2 and C 8 -PI3P. As shown in Fig. 8, the resonances from a site that corresponds to the PI3P-binding site of p40 phox PX were affected by the addition of C 8 -PI3P, whereas those corresponding to the phosphoinositide-binding site of the wild type p47 phox PX, determined by the TCS experiments (Fig. 2D), were not affected. These results suggest that the T45R and P78K mutations restored the PI3P-binding site of p40 phox PX on p47 phox PX. Importantly, most of these resonances were also affected by the addition of C 4 -PI(3,4)P 2 ( Fig. 9 and supplemental Fig. S9), suggesting that the mutation could not restore the absolute PI3P specificity of p40 phox PX.

DISCUSSION
NMR spectra provide structural information about proteins at an atomic resolution. The position of the NMR signal, chemical shift, is directly affected by the microenvironment. In NMR titration experiments with soluble ligands, the binding of a ligand causes the change of the microenvironment in the binding site. In addition, the ligand binding can induce conformational changes in the area located away from the binding site. Therefore, the chemical shift perturbation of the protein signals is useful for the detection of the binding site for the ligands and the site with the induced conformation change. The intensity reduction of the signal in the TCS experiment here indicates the membrane-interacting interface of the protein. Based on the results of the TCS experiment, the C 16 -PI(3,4)P 2 -containing membrane-interacting interface on the p47 phox PX domain includes the residues from helix ␣1, helix ␣1Ј, and the following loop up to the distorted left-handed PP II helix (Fig. 2,  C and D). The mutational analyses verified the results of the TCS experiments and revealed that Lys-55 is indispensable for the phosphoinositide recognition (Fig. 5). Lys-55 is located inside the herein identified membrane-interacting interface and directly contacts the sulfate molecule in the crystal structure of p47 phox PX.  The determined interface differs from those of all of the PX domains elucidated so far, including the p40 phox PX domain. Whereas the loss of the phosphoinositide binding activity of p47 phox PX by the K55E and R70K mutations ( Fig. 5 and Tables 1 and 2) suggests that Lys-55 and Arg-70 are the key residues for phosphoinositide binding, these residues are not conserved in the other PX domains. Therefore, Lys-55 and Arg-70 would be responsible for the unique phosphoinositide-binding site of p47 phox PX.
The chemical shift perturbation area on p47 phox PX, caused by the addition of C 4 -PI(3,4)P 2 , is larger than the phospho-inositide-binding site determined by the TCS experiments and includes residues in the site that corresponds to the PI3P-binding site of p40 phox PX (Fig. 3, A and B). Considering that chemical shift perturbation is induced by both the direct binding effects and the conformational change (45), the perturbation of the residues in the site that corresponds to the PI3P-binding site of p40 phox PX would be a secondary effect, due to a conformational change. The absence of the perturbation on the segment of the ␤ sheet comprising the C terminus of ␤1 and the N terminus of the ␤2 strands, which also corresponds to the PI3Pbinding site of p40 phox PX and was perturbed in the titration experiments on p40 phox and Vam7p (supplemental Fig. S5), supports the conclusion. The phosphoinositide-binding site and the site that corresponds to the PI3P-binding site of p40 phox PX are adjacent, when projected not only on the tertiary structure but also on the secondary structure of p47 phox PX, because they are partly formed by the same structural features, including the ␣1 helix and the long loop that connects helices ␣1 and ␣2. Therefore, a conformational change would easily propagate from one site to the other.
However, in the T45R/P78K/K55E mutant of p47 phox PX, the residues that correspond to the phosphoinositide-binding site on p47 phox PX were not perturbed, and the segment of the ␤ sheet composed of the C terminus of ␤1 and the N terminus of the ␤2 strands was strongly perturbed (Figs. 8 and 9 and supplemental Fig. S9). Therefore, the T45R/P78K/K55E mutant of p47 phox PX would assume the p40 phox -like phosphoinositidebinding site, whereas the phosphoinositide-binding site of the wild type p47 phox PX differs from that of p40 phox PX.
These results suggest that p47 phox PX is unable to accommodate a phosphoinositide in the area corresponding to that in p40 phox PX, because the critical residues, such as Arg-60 and Lys-92 of p40 phox PX, which correspond to Thr-45 and Pro-78 of p47 phox PX, are absent (Fig. 7). We should point out that the P78K mutation not only restored the essential role of Lys-92 in p40 phox PX but also removed the residue Pro-78 that occupies valuable space inside the ligand binding pocket. This might contribute considerably to restoring binding capability (Figs. 8 and 9 and supplemental Fig. S9). It could also explain why the adjacent residue Lys-79, which in the solution structure of p47 phox PX points away from the binding pocket (Fig. 7), cannot assume the role of Lys-92 in p40 phox PX. Therefore, such small but critical variations inside and outside the typical phosphoinositide binding pocket may cause the differences in the ligand-binding sites of these two PX domains.
Interestingly, our results can provide a rationale for the in vivo behaviors of two mutants, those of Pro-73 and Arg-70, that are within the presently identified C 16 -PI(3,4)P 2 -containing membrane-interacting interface. It has been reported that the mutation Pro-73 abrogates phosphoinositide recognition and interferes with the in vitro activity of NADPH oxidase (46), while the mutation Arg-70 prevents membrane translocation in HEK293 cells (47).
Our experiments revealed that PA and soluble C 4 -PI(3,4)P 2 competitively bind to the p47 phox domain (Fig. 4). Therefore, the atypical ligand-binding site on p47 phox recognizes both PA and PI(3,4)P 2 . This low specificity is in stark contrast with the stringent PI3P specificity of the typical binding site on p40 phox . Residues from the typical and atypical binding sites are colored green and blue, respectively. Full scale spectrum is shown in supplemental Fig. S9. B, chemical shift perturbation profile of the T45R/P78K/K55E mutant of p47 phox PX caused by C 4 -PI(3,4)P 2 . Normalized chemical shift changes and their errors were calculated with the same formula as in Fig. 3, but the color code is different. Residues that showed strong (N Ͼ2.9 ppm) normalized chemical shift changes are colored orange, and the broadened out residues are colored red. Note that the scale of the y axis is different. The chemical shift perturbation profile is similar with a reported one for the Vam7p PX (49) in conformity with the presence of a typical binding site in both proteins. C, perturbed signals are mapped on the crystal structure of WT p47 phox PX. The coloring is the same as in Fig. 8B.
Although at this point the physiological meaning of an atypical phosphoinositide-binding site on the p47 phox PX domain is unclear, the role of this atypical binding site should be important for the function of phagocyte NADPH oxidase because mutation of Lys-55, a key residue of this binding site, significantly impedes normal production of superoxides (Fig. 6). In the same biological system, the phagocyte NADPH oxidase, the two PX domains, p47 phox and p40 phox PXs, differ in the locations of the binding sites and in the affinities and selectivities toward phosphoinositides. These different structural properties may hint at distinct functional roles. It has been shown that in the assembled active NADPH oxidase complex, p40 phox PX remains continually bound on membrane-embedded PI3P during reactive oxygen species production (48). We speculate that p47 phox PX is not. As the phagosome matures and PI(3,4)P 2 gives space to PI3P, p47 phox PX may be released from the membrane and reorient itself, relative to the other components of the active enzymatic complex. In this context, the shallowness of the atypical binding site, with its few binding determinants, may be a key structural feature that enables p47 phox PX to interact with multiple phosphoinositide signals, depending on the stage of phagocyte NADPH oxidase assembly and/or activation.
It is worth mentioning that residues corresponding to His-51, Lys-55, and Arg-70 (that contribute to the formation of the atypical binding site) and to Thr-45 and Pro-78 that contribute to the collapse of the typical binding site of human p47 phox PX domain are highly conserved among different species. They are identical among organisms like human, bovine, pig, dog, and mouse. Thus, the atypical phosphoinositide-binding site does not appear to be an exclusive characteristic of the human p47 phox PX domain.
Based on the dodecyl-phosphocholine titration experiments of the Vam7 PX domain, Cheever et al. (49) proposed that the long loop between helices ␣1 and ␣2 functions as a membrane insertion loop. Importantly, the determined binding interface of p47 phox PX does not include the speculated membrane insertion loop in the 77-86-residue region (14). Because the membrane insertion loop can affect the binding affinity, its existence was examined further. It has been shown that the fluorescence of IAEDANS, covalently attached to membrane insertion loops, increases upon membrane insertion (50). The residues at positions 21 and 80, which are located outside and inside the speculated membrane insertion loop region, respectively, were substituted with cysteines and labeled with IAEDANS. Both of the IAEDANS-labeled p47 phox PX domains showed similar and only weak fluorescence increases upon the addition of the PI(3,4)P 2 -containing liposomes, and the magnitude of the fluorescence increase was similar to that induced by control liposomes without PI(3,4)P 2 (supplemental Fig. S10). In addition, the dodecyl-phosphocholine titration experiments with and without soluble C 4 -PI(3,4)P 2 did not reveal an increased propensity of the above region for membrane interactions (data not shown). Based on these results, we conclude that the 77-86residue region of p47 phox PX is not a membrane insertion loop. The 77-86-residue region of p47 phox PX is much shorter and protrudes less than the membrane insertion loop on the Vam7p PX domain (Protein Data Bank code 1KMD). This area of p47 phox PX may nonspecifically contact with the lipid bilayer, as the protein tries to accommodate the polar head of its ligand, but it does not function as a specific enhancer of membrane binding. Herein, the K d value for the interaction of the p47 phox PX domain with liposome-embedded PI(3,4)P 2 was estimated to be 24 M (fluorescence assay), which is comparable with the K d of 38.7 M (Table 1) for the soluble C 4 -PI(3,4)P 2 (NMR titration assay). The similarity of the above values is in agreement with the absence of a membrane insertion loop that could enhance the membrane docking. Glu-49 and Tyr-48, which were affected by the irradiation in the present TCS experiments, are spatially separated from the other affected residues (Fig. 2D). These residues might be important for a productive interaction with the lipid bilayer.
The weak affinity of p47 phox PX for membrane-embedded PI(3,4)P 2 is not necessarily incompatible with accurate membrane targeting. The combination of phosphoinositide recognition by the PX domain and protein-protein interactions involving other domains (such as the interaction of the two tandem SH3 domains with the proline-rich area of the integral membrane protein p22 phox ) can provide for accurate membrane targeting of the full-length p47 phox . Such cooperation between phosphoinositide recognition and protein-protein interactions for membrane targeting has already been demonstrated for other proteins, including Vam7p (51). In the case of p47 phox , it appears that protein-protein interactions are the primary mechanism of membrane targeting, because PX-truncated p47 phox is still able to find its way to the phagosomal membrane (10). However, membrane targeting of isolated p47 phox PX domain in intact cells has not been reported and based on our unpublished observations, 3 isolated p47 phox PX domain invariably fails in membrane localization. This conforms to its herein demonstrated low affinity for phosphoinositides.
The different patterns of phosphoinositide recognition between p47 phox PX and p40 phox PX, despite their similar overall structures, underscore how the substitution of the key residues on the surface of the PX domain can significantly alter its function. The functional alterations can be dramatic. The case of the PX domains from phospholipase D1 and D2, which were identified as a GAP protein and an interacting partner for Dynamin (52), is one of the first reports that challenged the conventional role of PX as merely a membrane interacting domain.
In conclusion, we performed TCS experiments that led to the correct identification of the interacting interface of membraneembedded PI(3,4)P 2 on the p47 phox PX domain. In similar applications, the TCS experiments could provide insights into the phosphoinositide recognition modes of various phosphoinositide-effector proteins, such as the PX, pleckstrin homology, ENTH, and C2 domains.