Heme histidine ligands within gp91(phox) modulate proton conduction by the phagocyte NADPH oxidase.

The membrane subunit of the phagocyte NADPH oxidase, gp91(phox), possesses a H(+) channel motif formed by membrane-spanning histidines postulated to coordinate the two heme groups forming the redox center of the flavocytochrome. To study the role of heme-binding histidines on proton conduction, we stably expressed the gp91(phox) cytochrome in human embryonic kidney 293 cells and measured proton currents with the patch clamp technique. Similar to its shorter homologue, NADPH oxidase homologue 1, which is predicted not to bind heme, gp91(phox) generated voltage-activated, pH-dependent, H(+)-selective currents that were reversibly blocked by Zn(2+). The gp91(phox) currents, however, activated faster, deactivated more slowly, and were markedly affected by the inhibition of heme synthesis. Upon heme removal, the currents had larger amplitude, activated faster and at lower voltages, and became sensitive to the histidine reagent diethylpyrocarbonate. Mutation of the His-115 residue to leucine abolished both the gp91(phox) characteristic 558-nm absorbance peak and voltage-activated currents, indicating that His-115 is involved in both heme ligation and proton conduction. These results indicate that the gp91(phox) proton channel is activated upon release of heme from its His-115 ligand. During activation of the oxidase complex, changes in heme coordination within the cytochrome might increase the mobility of histidine ligands, thereby coupling electron and proton transport.

The NADPH oxidase catalyzes the one-electron reduction of molecular oxygen to superoxide, the precursor of a variety of toxic oxygen radicals generated by neutrophils, eosinophils, and macrophages, to kill invading microorganisms (1,2). This enzyme is crucial in the host defense against microbial pathogens, and patients with chronic granulomatous disease (CGD) 1 , who fail to assemble a functional oxidase, suffer from severe recurrent infections (3,4). The NADPH oxidase is a multicomponent enzyme composed of at least three cytosolic subunits, p47 phox , p67 phox , and p40 phox , and two membrane-associated subunits, p22 phox and gp91 phox (1,5). Upon stimulation, the cytosolic components associate with the membrane-bound subunits, resulting in a functional oxidase complex that transfers electrons from cytosolic NADPH to extracellular oxygen (6,7). The transfer of electrons across the plasma membrane generates a massive depolarization, that, if uncompensated, would prevent further electron transfer and the associated production of superoxide (8,9). Because large amounts of acid equivalents are released in the cytosol during the hydrolysis of NADPH and its resynthesis by the hexose monophosphate shunt, the oxidase has been proposed to act as a H ϩ channel to preserve electroneutrality and allow the extrusion of the intracellular acid (10,11). Accordingly, both the depolarization and the cytosolic acidification generated by the oxidase are potentiated by Zn 2ϩ , an inhibitor of H ϩ channels (9,12).
The proton channel function of gp91 phox is still debated, although patch clamp studies in activated phagocytes confirmed the close coupling between H ϩ channels and the NADPH oxidase (9,13). Activation of the NADPH oxidase induced the appearance of a new type of H ϩ currents in human eosinophils stimulated with GTP␥S or Ca 2ϩ (9). The oxidaseassociated H ϩ currents had an unusually low threshold of voltage activation, enabling the entry of H ϩ ions and a cytosolic acidification. In addition, the new currents activated faster, deactivated more slowly, were more sensitive to Zn 2ϩ , and were blocked by the histidine-reactive agent DEPC (9). These unusual H ϩ currents were not strictly coupled to electron transport, as they persisted in the presence of diphenyliodinium (DPI), an inhibitor of the NADPH oxidase, or under anoxic conditions. However, they were absent in cells from CGD patients lacking either the gp91 phox or the p47 subunit (9), indicating that channel activation requires a functional oxidase. Similar currents were induced by PMA in neutrophils voltageclamped using the perforated patch configuration to preserve cellular integrity. Upon PMA stimulation, the H ϩ current amplitude increased progressively and did not correlate with the amplitude of the electron currents but instead strongly correlated with the amplitude of resting H ϩ currents (13). Block of the oxidase with DPI had no effect on the maximal H ϩ conductance but reversed the slowing of current deactivation, suggesting a complex interaction between oxidase activity and H ϩ channels.
Different conclusions were drawn from these diverging, but not contradictive, observations. We concluded that phagocytes express two types of H ϩ channels and that gp91 phox is probably the low threshold channel associated with an active oxidase (9). In contrast, DeCoursey et al. (13) concluded that PMA modulates a pre-existing H ϩ channel that is probably not gp91 phox (13). Consistent with gp91 phox being the H ϩ channel itself, expression of full-length or N-truncated gp91 phox generates H ϩ currents and pH changes in CHO cells (14 -16). Moreover, the recently cloned NOX-1S protein, generated by alternative splicing of the NADPH oxidase homologue gene NOX-1, also catalyzes voltage-gated H ϩ currents when expressed into HEK-293 cells (17). The NOX-1S currents had a high threshold of voltage activation, unlike the H ϩ currents observed in phagocytes with an active oxidase. Thus, gp91 phox and NOX-1S are either two distinct H ϩ channels, or, alternatively, they both function as channel modulators. If gp91 phox is indeed the H ϩ channel of phagocytes, its membrane expression should be sufficient to generate phagocyte-like H ϩ currents in non-phagocytic cells. In contrast, if gp91 phox is a channel modulator, the phagocytic currents are not likely to be reproduced by the heterologous expression of gp91 phox in non-phagocytic cells, which display distinct ion channels and cannot assemble a functional oxidase.
In this study, we show that the stable expression of gp91 phox at the plasma membrane of HEK-293 cells is associated with voltage-gated H ϩ currents. Like all H ϩ currents, the gp91 phoxassociated currents were pH-dependent, H ϩ -selective, and reversibly blocked by micromolar concentrations of Zn 2ϩ . However, they had faster activation kinetics than the currents observed in the same recipient cell line stably expressing the shorter oxidase homologue NOX-1S. Inhibition of heme synthesis did not affect the expression of gp91 phox at the plasma membrane but dramatically modulated the H ϩ currents. The currents observed in heme-depleted gp91 phox cells closely resembled the H ϩ currents observed in phagocytes with an active oxidase. They had faster activation and slower deactivation kinetics, larger amplitude, and a much lower threshold of voltage activation that allowed H ϩ influx at voltages negative to H ϩ equilibrium potential. The currents induced by heme depletion were blocked by the histidine-reactive agent DEPC, indicating that removal of the heme moiety exposed a critical histidine residue. Replacement of the His-115 residue by leucine abolished voltage-activated currents, confirming that this residue is critical for proton conduction, as reported previously (16,21). Spectral analysis of membranes prepared from H115L cell lines revealed that the characteristic 558-nm absorbance peak of the gp91 phox cytochrome was absent in the H115L mutant, indicating that His-115 is also a heme ligand. These results suggest that the gp91 phox cytochrome indeed acts as a H ϩ channel but that its H ϩ conductive properties depend on the binding of heme to the His-115 residue. Changes in heme coordination and/or spin state during the activation of the oxidase complex might thereby functionally couple electron and proton transport.
EXPERIMENTAL PROCEDURES gp91 phox Cell Lines and Mutagenesis-The full-length human gp91 phox sequence was inserted into the pSCT vector as described (18). HEK-293 and COS-7 cells were transfected with the pSCT-gp91 phox construct (4 g/ml) using the calcium phosphate precipitation method. Stable gp91 phox cell lines were generated by co-transfecting an empty pcDNA 3 vector containing the neomycin resistance gene with the pSCT-gp91 phox construct at 1/10 proportion. 3 days after transfection, cells were exposed to 500 g/ml G418 for 20 days, and surviving clones were replaced in 96-well plates and cultured with 100 g/ml G418. Individual gp91 phox clones were screened for the presence of voltage-gated currents by patch clamp measurements, and clones (gp91 phox -2 and gp91 phox -3) that showed intermediate current densities were selected for further experiments.
For site-directed mutagenesis, the gp91 phox cDNA was extracted from the pSCT vector by BamHI digestion and inserted into pcDNA 3 . The H115L mutant was synthesized with the QuickChange TM kit according to the manufacturer's instructions (Stratagene, Bâ le, Switzerland), using forward primer 5Ј-CTT CAC TCT GCG ATT CTG ACC ATT GCA CAT C-3Ј and reverse primer 5Ј-G ATG TGC AAT GGT CAG ATT CGC AGA GTG AAG-3Ј. Stable H115L cells were generated by transfecting HEK-293 cells with 1.06 g/ml H115L-gp91 phox using the TransFast TM reagent (Promega, Wadisellen, Switzerland). After selection with neomycin, individual clones were screened for the presence of the gp91 phox protein by immunofluorescence and Western blot with the mAb48 antibody. A clone (B4) was selected as it showed a constant expression with both techniques. Cells were cultured in Dulbecco's modified Eagle's medium supplemented with 200 nM L-glutamine, 10% fetal calf serum, 100 units/ml penicillin, and 100 g/ml streptomycin (Life Technologies, Inc.) at 37°C in an incubator with 5% CO 2 . For patch clamp measurements cells were plated on 25-mm glass coverslips 2 days before experiments.
Patch Clamp Measurements-Whole cell patch clamp recordings were performed as described previously (9), using an Axopatch 1D amplifier (Axon Instruments, Foster City, CA) in the voltage clamp mode. Patch pipettes were pulled from borosilicate glass with a P-87 Flaming/Brown micropipette puller (Sutter Instrument Company, Novato, CA). Pipette resistance ranged between 3 and 7 megohms, seal resistance between 2 and 20 gigaohms, and the mean access resistance varied between 10 and 30 megohms. Cell capacitance averaged 16.83 Ϯ 0.6 picofarads (n ϭ 111). Cells were voltage-clamped at a holding potential of Ϫ60 mV and depolarized to various test potentials as indicated. The currents were filtered at 1 KHz and sampled at 200 Hz using pClamp 6.0 software. Leak currents were small compared with the whole cell H ϩ currents and were subtracted only to allow calculation of the whole cell conductance. All recordings were performed Ͼ3 min after achieving the whole cell configuration to allow the equilibration of the cytosol with the pipette solution.
Unless Data Analysis-Data analysis was performed using Origin software (MicroCal, Northampton, MA). For exponential fits, the first 5 ms of the activating or deactivating (tail) currents were not considered to avoid capacitance artifacts. Computation required ϳ20 -200 iterations to reach a stable condition with a level of confidence of 1%, as assessed by the nonlinear least squares regression method.
Immunofluorescence-The gp91 phox antibodies used were as follows: a rabbit polyclonal antibody directed against amino acid 562-569 of the C-terminal domain (kindly provided by Dr. F. Morel, Grenoble, France; see Figs. 1 and 4), the c-terminal mAb48 monoclonal antibody (kindly provided by Dr. Dirk Roos, Amsterdam, The Netherlands; see Fig. 6), and the 7D5 monoclonal antibody directed against an external epitope of gp91 phox (kindly provided by Dr. K. Nemet, Budapest, Hungary). A mouse IgG1 (anti-Myc) was used as isotype-matched control for the two monoclonal antibodies.
Cells plated in 12-mm coverslips were washed two times with phosphate-buffered saline (PBS), fixed with 3.7% paraformaldehyde, permeabilized with 0.3% Triton X-100, and incubated for 20 min in PBS supplemented with 1% BSA. Cells were then exposed to the rabbit peptide antibody (1/800), mAb48 (1/600), or mAb7D5 (1/800) for 1 h on a shaker in 0.2% Tween and 1% BSA. After three successive washes, unspecific labeling was blocked with 10% non-immune goat serum in PBS for 10 min. Cells were then immediately exposed to a Texas Red secondary antibody, goat anti-mouse (1/800), or goat anti-rabbit (1/600) for 1 h in 0.2% Tween and 1% BSA PBS. Cells were then washed two times in 0.2% Tween and 1% BSA PBS and two times in PBS and finally mounted on glass. Fluorescence images were acquired using a Zeiss LSM 510 confocal microscope (see Fig. 4) or with a cooled CCD camera mounted on an inverted Zeiss Axiovert S100 TV microscope (see Figs. 1 and 6).
Spectral Analysis of gp91 phox -Total cellular membranes were prepared from stable gp91 phox clones, control HEK-293 cells, and human neutrophils as described (19). Briefly, 10 8 cells/ml in PBS ϩ 2 mM EGTA were incubated for 20 min at 0°C under gentle agitation in lysis buffer containing 1% Triton X-100, 10 mM HEPES, 3.5 mM MgCl 2 , pH 7.4, and 2 tablets of Complete TM protease inhibitor mixture (Roche Molecular Biochemicals). Cells were then centrifuged at 12000 ϫ g for 30 min at 4°C. The resulted supernatant was collected, and the protein concentration measured by the Bradford method. Absorbance spectra of the membrane preparations were recorded between 400 and 600 nm on a Bio-Tek spectrometer (Kontron) before (oxidized) and after (reduced) addition of a few grains of sodium dithionite. The differential spectra were obtained by subtracting the oxidized from the reduced measurement.
Western Blot-20 g of membrane proteins solubilized in sample buffer (0.187 M Tris-HCl, pH 6.8, 7% SDS, 0.03% bromphenol blue) were loaded on an 8% SDS-polyacrylamide gel electrophoresis gel (4% stacking gel) and transferred to a polyvinylidene difluoride membrane in semi-dry conditions at 200 milliamperes for 45 min at room temperature using Trans-blot SD (Bio-Rad). The polyvinylidene difluoride membrane was reversibly stained with red ponceau and incubated for 1 h in 0.1% Tween Tris-buffered saline (TBS; 20 mM Tris-HCl, 500 mM NaCl, pH 7.5) supplemented with 5% milk. The membrane was then exposed overnight at 4°C to the mAb48 (1/1000) in 0.1% Tween TBS plus 5% milk, washed three times for 5 min in 0.1% Tween TBS, and exposed to a peroxidase-conjugated goat anti-mouse secondary antibody (1/20000; Bd Transduction Laboratories, Basel, Switzerland). The blot was revealed with enhanced chemiluminescence reactifs (LumiGlo reagent and Peroxide reagent; BioConcept, Cell Signaling Switzerland, Luzern, Switzerland) and images taken with the luminescent image analyzer LAS-1000 plus (Fujifilm) using Aida 2.3 acquisition software.

gp91 phox Expression Induces Voltage-activated Outward
Currents-To investigate whether gp91 phox functions as a proton channel, the membrane cytochrome was expressed in HEK-293 cells, the cell line used previously to characterize the H ϩ currents generated by the oxidase homologue NOX-1S (17). Cells were stably transfected with the cDNA coding for the fulllength gp91 phox cytochrome (18), and expression of the protein was assessed by immunofluorescence using a polyclonal antibody directed against a C-terminal gp91 phox epitope (20). As shown in Fig. 1, a bright fluorescence staining was observed in gp91 phox transfectants, whereas no fluorescence was observed in non-transfected cells. The staining pattern was consistent with a predominant plasma membrane localization of the epitope in the gp91 phox cell line, as reported previously in CHO cells expressing full-length or truncated gp91 phox (16,21). A strong fluorescence signal was also observed in non-permeabilized cells stained with the 7D5 monoclonal antibody, directed against an extracellular loop of gp91 phox , demonstrating that the protein is delivered to the plasma membrane and inserted in the correct orientation (data not shown). This indicates that the gp91 phox cytochrome can be expressed in non-phagocytic cells and remains properly targeted to the plasma membrane.
The presence of H ϩ currents was then assessed with the whole cell patch clamp technique, using acidic pipette solutions to impose an outward pH gradient across the plasma membrane (100 mM MES, pH i -5.7, pH o -7.5). Cells were allowed to equilibrate for 3 min with the pipette solution to ensure a good pH clamp before recording the membrane currents. As shown in Fig. 1A, slowly activating outward currents were observed at voltages ϾϪ40 mV in the gp91 phox cell line, whereas small background currents were observed in non-transfected HEK-293 cells or in cells transfected with the empty pcDNA 3 vector. The background HEK-293 currents were shown previously to be unaffected by the expression of other channel proteins (17) and were only observed at very acidic pipette solutions. To ensure that gp91 phox did not up-regulate these endogenous HEK-293 currents, we expressed the protein in COS-7 cells, which were found to be devoid of background currents (Fig. 1A, bottom). The transient expression of gp91 phox induced even larger currents in COS-7 cells, suggesting that gp91 phox indeed acts as a voltage-gated channel. The current densities, however, were much more variable in the transiently transfected COS-7 cells (Fig. 1B, bottom panel), and the HEK-293 cell line was used to characterize the currents associated with the expression of the gp91 phox protein.
The gp91 phox -associated Current Is H ϩ -selective, pH-dependent, and Zn 2ϩ -sensitive-Voltage-gated proton channels are extremely H ϩ -selective, strongly modulated by the pH gradient, and reversibly inhibited by polyvalent metal cations (22). To assess the selectivity of the gp91 phox currents, we measured the reversal potential of the tail currents, E rev . Outward currents were activated by depolarizing cells to 60mV for 2 s, and deactivating tail currents were recorded during repolarization to voltages ranging from Ϫ100 to 0 mV. Fig. 2 shows superimposed tail currents recorded at pH i :pH o , 5.7:7.5 in symmetric CsAsp solutions (inset). An exponential fit to the tail currents was performed to extract the instantaneous current/voltage relationship and calculate the reversal potential, which averaged Ϫ72 Ϯ 3 mV in these conditions (n ϭ 15). This negative reversal potential indicates that the current is unlikely to be carried by Cs or Asp, whose equilibrium potential is close to zero, but rather by H ϩ , whose equilibrium potential is Ϫ105 mV. Accordingly, E rev was not affected by the replacement of Cs ϩ by NMG ϩ or of Asp Ϫ by Cl Ϫ but changed to Ϫ44.6 Ϯ 2.4 (n ϭ 9) and to Ϫ7.3 Ϯ 5.9 mV (n ϭ 6) when the pH value of the pipette was changed to 6.3 and 7.3, respectively ( Fig. 2A). This corresponds to a change of 42 mV per pH unit, whereas the Nernst equation predicts a shift of 58 mV. This deviation most likely reflects the imperfect intracellular pH clamp and the increased H ϩ depletion occurring at acidic pipette pH, as consistently reported in previous H ϩ channel studies (9, 23, 24). However, even assuming that the deviation is because of a finite permeability of the conductance to other ions, the relative H ϩ FIG. 1. gp91 phox membrane expression is associated with voltage-gated currents. A, voltage-gated currents in HEK-293 and COS-7 cells stably or transiently transfected with gp91 phox . Cells were voltageclamped in the whole cell configuration of the patch clamp technique, using acidic pipette solutions to impose an outward pH gradient (pH i : pH o , 5.7:7.5; inset). Currents were activated by 3-s depolarizing steps to voltages ranging from Ϫ40 to 80 mV. WT, wild-type cells. B, immunolocalization of gp91 phox in stable HEK-293 transfectants. Cells were fixed with 3.7% paraformaldehyde, permeabilized with 0.3% Triton X-100, and exposed to a polyclonal antibody directed against the Cterminal tail of gp91 phox (kindly provided by Dr. F. Morel, Grenoble, France). Images are representative of three independent experiments. C, current densities of wild-type cells and of cells transfected with the empty pcDNA 3 vector or with gp91 phox . The currents measured at the end of a 3-s depolarizing pulse to 60 mV were divided by the cell capacitance.
permeability p H /p Cs calculated according to the Goldman-Hodgkin-Katz equation is Ͼ10 6 given the very low concentration of H ϩ compared with the other ions. Thus, the gp91 phox currents are very selective for H ϩ .
Consistent with the known pH dependence of H ϩ channels, increasing the pH gradient by perfusing cells with more acidic pipettes increased the amplitude of the gp91 phox current and shifted its threshold of voltage dependence to more negative voltages (Fig. 2B). Decreasing the pH gradient had the opposite effect, the threshold for voltage activation remaining above the H ϩ equilibrium potential, and only outward currents were observed in all conditions (Fig. 2B). As expected, addition of the divalent metal cation Zn 2ϩ (3 mM free [Zn 2ϩ ] calculated to be ϳ200 M) markedly reduced the outward currents (Fig. 2C). The block was rapid and was fully reversible upon wash of Zn 2ϩ from the bath solution, with a slight overshoot reflecting the build up of H ϩ ions near the plasma membrane during the block, as H ϩ extrusion is blunted under these conditions. The predominant effect of Zn 2ϩ was to shift the voltage dependence of the current activation by 60 mV toward more positive voltages and to cause a marked slowing of the kinetics of activation (not shown). This is consistent with the known effects of Zn 2ϩ , which modulates H ϩ channel gating but not its instantaneous current-voltage relationship (25), an effect attributed to the binding of Zn 2ϩ to external sites on the channel protein. Thus, the H ϩ currents observed in stable gp91 phox transfectants have properties that are typical of voltage-gated H ϩ channels.
Comparison of gp91 phox and NOX-1S H ϩ Currents-The gp91 phox -associated H ϩ currents appear very similar to the H ϩ currents associated with the expression of NOX-1S (17). This was somewhat surprising, as the NOX-1 gene is expressed mostly in colon, whereas gp91 phox expression is restricted to phagocytes. Because H ϩ channels in different tissues exhibit very distinct kinetics of voltage activation (22), these kinetic differences would be expected to persist in the HEK-293 background if the two proteins indeed function as H ϩ channels. If, on the other hand, the proteins act as channel modulators, as has been proposed (13), the kinetic properties of the currents would reflect the endogenous channel of the recipient cell line. To test this possibility, we analyzed in detail the kinetics of activation of the gp91 phox and NOX-1S currents. To minimize the effects due to differences in expression levels, stable gp91 phox and NOX-1S clones displaying similar current densities were chosen for comparison. Fig. 3 shows that the H ϩ The time for half-maximal activation ( act ) was estimated by fitting exponential curves to the currents recorded at different voltages. Data are means Ϯ S.E. C, kinetics of tail current deactivation during repolarization to Ϫ50 mV. Currents of similar amplitude were elicited in gp91 phox and NOX-1S cells by a 2-s pulse to 60 mV (inset). The reversal potential was identical in both conditions. D, voltage dependence of tail current deactivation kinetics in gp91 phox (squares; n ϭ 15) and NOX-1S cells (circles; n ϭ 12). The time for half-maximal deactivation ( tail ) was estimated by fitting exponential curves to the tail currents. *, p Ͻ 0.001 versus NOX-1S, from unpaired t test.
currents activated faster in gp91 phox than in NOX-1S cells. An exponential fit to the current activated by a 60-mV pulse yielded a time constant for activation ( act ) of 295 Ϯ 18 ms for gp91 phox cells (n ϭ 24) and 370 Ϯ 27 ms for NOX-1S cells (n ϭ 19, p Ͻ 0.02). Faster activation was observed in gp91 phox cells at all voltages but became significant only at voltages exceeding 60 mV (Fig. 3B). In addition, the H ϩ currents deactivated more slowly in gp91 phox cells than in NOX-1S cells (Fig. 3C). The time constant of deactivation ( tail ) was determined by fitting an exponential curve to the tail currents measured at different repolarizing voltages following a pulse to 60 mV (Fig.  3D). At Ϫ50 mV, tail was 379 Ϯ 38 ms for the gp91 phox cells (n ϭ 15) and 237 Ϯ 27 ms for the NOX-1S cells (n ϭ 12, p Ͻ 0.02). Thus, proton currents activated faster and deactivated more slowly in gp91 phox than in NOX-1S cells, suggesting that the underlying channel proteins have distinct properties.
Heme Depletion Modulates gp91 phox but Not NOX-1S H ϩ Currents-Unlike NOX-1S, gp91 phox contains two non-identical heme groups embedded within the membrane that mediate the final steps of electron transfer to molecular oxygen. Both hemes are coordinated non-covalently by histidines in the axial positions, the most probable ligand pairs being His-101-His-209 and His-115-His-222 (26). His-209 and His-222 are located in the fifth transmembrane domain missing from the NOX-1S splice variant, which is thus predicted not to bind heme. In contrast, His-101 and His-115 are located in the third transmembrane domain, within the putative H ϩ transport motif shared by gp91 phox and NOX-1S. Importantly, His-115 is the central residue within the chain of histidines thought to conduct H ϩ ions and has been shown by mutagenesis to be the residue most critical for H ϩ channel activity (16,21). The binding of heme to this residue within the full-length cytochrome might explain the kinetic differences between the gp91 phox and NOX-1S currents.
To assess whether the presence of heme within the gp91 phox cytochrome affects H ϩ currents, we cultured the cells in the presence of succinyl acetone (SA), an inhibitor of heme synthesis. SA specifically inhibits the enzyme 5-aminolevulinic acid dehydratase that catalyzes the formation of porphobilinogen from 5-aminolevulinate, thereby preventing the synthesis of heme (27). Exposure of HEK-293 cells to 10 g/ml SA for 4 days had no effects on the fluorescence intensity and staining pattern obtained with the gp91 phox antibody, indicating that heme depletion does not alter the expression of gp91 phox at the plasma membrane (Fig. 4B). The lack of heme incorporation was verified by spectroscopy, by measuring the reduced minus oxidized difference absorption spectra from membrane prepa-rations of gp91 phox cells treated with SA (see Fig. 6). The specific ␣-band absorbance peak at 558 nm characteristic of neutrophils (orange line) was clearly visible in untreated gp91 phox cells (red) but was almost completely abolished in SA-treated cells (green), indicating that the heme was not incorporated in the presence of SA. Consistent with the conserved membrane expression of gp91 phox , no differences were observed between control and heme-depleted cells when H ϩ currents were measured with acidic pipette solutions (current densities at 40 mV were as follows: 3.1 Ϯ 0.3 versus 2.7 Ϯ 0.2 pA/picofarad for control and SA-treated cells, respectively; n ϭ 31 and 6; pH i :pH o , 5.7:7.5). However, in conditions favoring H ϩ influx (pH i :pH o , 8.0:7.5), the currents were markedly altered in heme-depleted gp91 phox cells (Fig. 4A). The currents in SAtreated gp91 phox cells activated faster and had larger amplitude, slower deactivation kinetics, and a much lower threshold of voltage activation ( Fig. 4B; n ϭ 18). The threshold for voltage activation was shifted by ϳ50 mV to more negative voltages, allowing inward steady-state H ϩ currents at negative potentials (Fig. 4B). The currents reversed sign around 5 mV; i.e. ϳ15 mV below H ϩ equilibrium potential, a deviation likely reflecting the drop in pH below the plasma membrane caused by the inward H ϩ currents. As expected, the H ϩ currents in cells expressing the heme-devoid protein NOX-1S were not affected by SA, ruling out nonspecific effects of the heme depletion protocol (Fig. 4A). Thus, heme depletion modulates H ϩ currents in gp91 phox but not in NOX-1S cells. The modulated currents closely resemble the low threshold H ϩ currents observed in activated phagocytes (9,13), strongly suggesting that gp91 phox is the oxidase-associated H ϩ channel of phagocytes.
Block by DEPC-The large inward H ϩ currents observed upon heme depletion suggest that removal of the heme molecules increases the mobility of protons flowing from the extracellular side. The heme coordinated by the His-115 residue is located near the extracellular side and transfers electrons to an external oxygen binding pocket where O 2 Ϫ is formed (26). To test whether removal of heme from the cytochrome increases the accessibility of the His-115 residue to external protons, we assessed the sensitivity of H ϩ currents to the histidine-reactive agent DEPC. As shown in Fig. 5A, 1.2 mM DEPC completely blocked the SA-induced increase in H ϩ currents but had only minimal effects in cells that had not been exposed previously to SA. The effects of DEPC were rapid and progressive, and the decrease in steady-state current amplitude was paralleled by a decrease in the amplitude of tail currents (Fig. 5C). The 4-fold increase in current amplitude observed upon SA induction at alkaline pH i was completely reversed by DEPC (Fig. 5B), sug- FIG. 4. Heme removal modulates gp91 phox but not NOX-1S currents. A, effect of heme depletion on H ϩ currents in gp91 phox and NOX-1S cells. Cells were incubated with 10 g/ml SA for 4 -5 days to inhibit the synthesis of heme, and H ϩ currents were recorded using alkaline pipette solutions to favor H ϩ influx (inset). B, effect of SA on gp91 phox immunostaining. C, current-voltage relationship measured at a pipette pH value of 8.0 in control (circles; n ϭ 10) and SA-treated gp91 phox cells (squares; n ϭ 18). Data are means Ϯ S.E. Inset, voltage dependence of H ϩ current activation kinetics ( act ) in control (circles) and SA-treated gp91 phox cells (squares). *, p Ͻ 0.05 versus ϪSA, from unpaired t test.
gesting that histidine residues mediate most of the effects of the heme depletion protocol. This suggests that the binding of heme to a critical histidine residue, likely His-115, modulates the conductive properties of the gp91 phox H ϩ channel. The changes in heme coordination occurring within gp91 phox during oxidase activation might thus explain the appearance of low threshold H ϩ currents during activation of phagocytes.
Role of His-115-To verify that the His-115 residue mediates both H ϩ conduction and heme ligation, this residue was replaced by leucine, and the mutated gp91 phox was stably expressed in HEK-293 cells. The H115L mutation did not affect the expression of gp91 phox , as verified by immunofluorescence and Western blotting (Fig. 6A). The H115L gp91 phox mutant was then tested for H ϩ conduction by the patch clamp technique as in Fig. 1. No voltage-activated currents could be elicited by depolarizing steps up to 80 mV (Fig. 6B), the current density being comparable with control, non-transfected cells (0.78 Ϯ 0.14 pA/picofarad; n ϭ 10). This confirmed that the His-115 residue is critical for proton conduction as reported previously (16,21). The ability of the H115L mutant to bind heme was then assessed by spectroscopy (Fig. 6C). The specific ␣-band absorbance peak at 558 nm present in neutrophils (orange) and gp91 phox cells (red) was completely absent in the H115L gp91 phox mutant (black). Similar findings were obtained recently in COS-7 cells expressing gp91 phox mutants, 2 in agreement with the proposal that the four heme-coordinating histi-dines are contained within gp91 phox (26). The His-115 residue thus has a dual role of H ϩ conduction and heme ligation. DISCUSSION Our results confirm that gp91 phox , the membrane-associated subunit of the NADPH oxidase, functions as a voltage-gated H ϩ channel. Upon expression in non-phagocytic cells, gp91 phox generated H ϩ currents similar to the currents generated by the oxidase homologue NOX-1S. The currents were strongly pHdependent, extremely H ϩ selective, and reversibly blocked by micromolar concentrations of Zn 2ϩ . The gp91 phox currents had slow kinetics of activation, resembling the currents of resting phagocytes but activated slightly faster and deactivated more slowly than the NOX-1S currents. Unlike NOX-1S currents, the gp91 phox currents were markedly affected by the inhibition of heme synthesis, consistent with the different heme content of the two proteins. Upon heme depletion, the gp91 phox currents 2 M. Dinauer, personal communication.  Fig. 1A. Traces are representative of 10 cells. C, differential absorption spectra (dithionite-reduced minus oxidized) of membranes prepared from neutrophils (polymorphonuclear; orange), gp91 phox cells (red), gp91 phox cells treated with SA (green), H115L cells (black), and control HEK-293 cells (blue). The absorbance peak at 558 nm specific of gp91 phox hemes is shown. Traces are representative of three experiments from different membrane preparations. activated at lower voltages, had larger amplitude and distinct kinetics of activation and deactivation, and were blocked by the histidine-reactive agent DEPC. Thus, the presence of heme modulates the H ϩ channel properties of the gp91 phox cytochrome.
Heme insertion is critical for the formation of stable heterodimers between gp91 phox and the non-glycosylated subunit p22 phox , a process that is required for the insertion of the flavocytochrome in the plasma membrane of phagocytes (28,29). Accordingly, heme depletion prevented the formation of stable heterodimers (29,30) and strongly decreased the surface expression of gp91 phox in PLB-985 cells, an effect that was reversible and prevented by the addition of exogenous heme (28). Heme removal did not affect the biogenesis of p65, the high mannose precursor of gp91 phox , but the protein failed to mature to its fully glycosylated state, and p65 monomers were rapidly degraded by the cytosolic proteasome (29). Non-phagocytic cell lines appear more tolerant in allowing the expression and membrane insertion of gp91 phox monomers, possibly reflecting difference in the proteolytic environment. Stable gp91 phox cell lines have been generated in COS-7 and 3T3 cells (28), as well as in CHO (14) and HEK-293 cells (this study). In all cases, the expression of gp91 phox was confirmed by Western blots or immunofluorescence. The predominant species detected in immunoblots from COS-7 membranes appears to be the 58-or 65-kDa precursor. Despite the apparent lack of maturation of the gp91 phox protein, the cells retained the characteristic spectral and redox properties of neutrophil flavocytochrome b558 (26). The heme spectral signature was also clearly visible in our gp91 phox HEK-293 transfectants, but the size of the expressed protein was closer to 91 kDa even in the H115L mutant, which is unable to bind heme (Fig. 6, A and C). This suggests that, in the HEK-293 cell line, the gp91 phox protein is able to mature to its fully glycosylated state even in the complete absence of heme insertion. A role of the p22 phox subunit in the maturation process cannot be ruled out, as expression of p22 phox could be detected in the HEK-293 cell line by reverse transcriptase polymerase chain reaction (data not shown). However, the lack of effect of heme depletion or of the His-115 mutation on the plasma membrane staining indicates that, regardless of its glycosylation state, gp91 phox is stably expressed in HEK-293 cells in the absence of heme.
Whereas gp91 phox expression was sufficient to reconstitute the currents observed in resting phagocytes, removal of the heme molecules produced very distinct currents. The currents generated by heme-depleted gp91 phox mimicked the H ϩ currents observed in phagocytes during the activation of the NADPH oxidase (9,13), suggesting that gp91 phox is the channel activated by PMA or GTP␥S in phagocytes. In neutrophils, a clear correlation was observed between the density of H ϩ currents measured before and after stimulation with PMA, suggesting that most of the activated channels are already in the membrane (13). However, gp91 phox is clearly not the only channel of phagocytes, because eosinophils from (X91°), CGD patients, who completely lack the gp91 phox subunit, had resting current densities similar to control cells (9). The other channel had properties similar to NOX-1S and did not generate low threshold H ϩ currents upon stimulation with GTP␥S (9). This indicates that phagocytes express not only gp91 phox but probably also NOX-1S, consistent with the presence of NOX-1S mRNA in HL-60 cells (17). The relative contribution of the NOX-1S and gp91 phox channels in resting and activated phagocytes, however, is difficult to assess for several reasons. 1) In the absence of stimulation, the gp91 phox currents are undistinguishable from the currents generated by NOX-1S. 2) Although it did not display a shift in voltage activation, the channel of CGD cell was also activated by GTP␥S (9), making it difficult to attribute the increase in current amplitude to a specific channel. 3) CGD cells might compensate for the lack of gp91 phox by expressing increased levels of NOX-1S or of other channels. 4) PMA and GTP␥S might activate the two channels in a different manner. The effects of PMA are largely mediated by the generation of arachidonic acid, which directly binds to the gp91 phox protein (31) and induces its H ϩ channel activity (15,32). Accordingly, cells lacking cytosolic phospholipase A2 do not respond to PMA but respond to exogenous arachidonic acid (33). The effects of GTP␥S might be more complex, as it induces a massive exocytosis and also promotes the translocation of the cytosolic oxidase subunits (34,35).
The shift in voltage activation of the gp91 phox channel clearly does not require the redox function of the oxidase, as it can be induced in cells unable to mount a functional oxidase. Thus, H ϩ fluxes are not strictly coupled to the flux of electrons, and the lack of correlation between electron and proton currents cannot be used as an argument against the H ϩ channel function of gp91 phox . Indeed, earlier observations in a series of CGD variants suggested that the H ϩ channel function of the oxidase required its assembly but not its redox function (36). Accordingly, the H ϩ currents activated by GTP␥S in eosinophils persisted in the presence of DPI or in anoxic conditions (9). The gp91 phox channel can be directly activated by arachidonic acid (16), which binds to the cytochrome and induces the transition of the heme iron from a low spin hexacoordinated state to a high spin pentacoordinated state (37). This change in heme coordination increases the affinity of the heme for O 2 and appears to reproduce several of the effects of heme depletion. In CHO cells expressing full-length or truncated gp91 phox , arachidonic acid increased the current amplitude and caused a 23-mV shift in the current-voltage relationship, allowing steady-state  26. A proton wire is formed by the histidine chain within the gp91 phox third transmembrane domain. Bottom, the "hopping" of protons is limited by the reduced mobility of the His-115 residue, which is a heme ligand. The mobility of this residue is increased upon heme depletion or during oxidase activation, as the heme undergoes a transition from a low spin hexacoordinated state to a high spin pentacoordinated state. The change in heme coordination, by facilitating proton conduction through gp91 phox , functionally couples electron and proton transport.
inward H ϩ currents (16). However, the currents in CHO transfectants activated much faster than in our HEK-293 cells, and arachidonic acid had no further effects on the kinetics of activation or deactivation. This might reflect the higher expression levels achieved, as the average currents were ϳ50-fold higher in CHO than in HEK-293 transfectants (3.4 -4.2 nA at 80 mV, pH i :pH o , 6.5:8.0 (16), versus 70 pA at 80 mV, pH i :pH o , 5.7:7.5). These high levels might have saturated the heme synthesis pathway, leading to the production of gp91 phox proteins lacking heme. The absence of heme in a fraction of the expressed gp91 phox proteins would explain the increased current amplitude and faster activation observed in approximately half the cells (16), as well as the partial effects of arachidonic acid. Taken together, however, the similar effects of arachidonic acid and heme depletion indicate that residues at or near the hemebinding site mediate the voltage-dependent H ϩ fluxes.
Strong evidence suggests that the residues His-111, His-115, and His-118 form a proton-conducting channel within gp91 phox . This motif is retained in all known proteins that function as voltage-gated proton channels, NOX-1S (17), NOX-1L, 3 gp91 phox (see Ref. 16 and this study), and a gp91 phox truncation mutant (16). These histidine residues are aligned along the axis of an ␣-helix (Fig. 6) and might function as a "proton wire" by allowing the hopping of protons between pairs of hydrogenbonded donors and acceptor residues (38). Water molecules might extend deep into the wire, allowing the same histidine to be alternatively exposed to the intracellullar and extracellular side, as shown in a mutated Shaker channel (39). Regardless of the mechanism, the reorientation or "turning" step of the donor component requires a certain rotational mobility of the histidines residues that form the wire. The central His-115 residue appears critical, as the gp91 phox currents were nearly abolished when His-115 was mutated to leucine (see Fig. 6B and Ref. 16). As His-115 functions both as a heme ligand and as a proton donor/acceptor (Fig. 6), its mobility likely depends on the presence of heme (Fig. 7). Accordingly, in the absence of heme the currents in our gp91 phox transfectants had larger amplitude, faster activation, and increased voltage dependence and were blocked by DEPC. This indicates that the removal of heme from its histidine ligands increases proton conduction, as well as the accessibility of these histidines to the external solvent. The binding of heme to His-115 thus determines the H ϩ conductive properties of the oxidase and underlies the transitions between the H ϩ currents observed in resting and activated phagocytes (Fig. 6). This model has important implication for the function of the oxidase. It implies that the protons pass close to the heme-iron center of the flavocytochrome and that His-115 is a labile heme ligand. The shifting accessibility of the His-115 residue during oxidase activation would thus functionally link electron and proton transport.