Structure of the N-terminal Regulatory Domain of a Plant NADPH Oxidase and Its Functional Implications*

Plant NADPH oxidases (Rboh, for respiratory burst oxidase homolog) produce reactive oxygen species that are key regulators of various cellular events including plant innate immunity. Rbohs possess a highly conserved cytoplasmic N-terminal region containing two EF-hand motifs that regulate Rboh activity. Rice (Oryza sativa) RbohB (OsRbohB) is regulated by the direct binding of a small GTPase (Rac1) to this regulatory region as well as by Ca2+ binding to the EF-hands. Here, we present the atomic structure of the N-terminal region of OsRbohB. The structure reveals that OsRbohB forms a unique dimer stabilized by swapping the EF-hand motifs. We identified two additional EF-hand-like motifs that were not predicted from sequence data so far. These EF-hand-like motifs together with the swapped EF-hands form a structure similar to that found in calcineurin B. We observed conformational changes mediated by Ca2+ binding to only one EF-hand. Structure-based in vitro pulldown assays and NMR titration experiments defined the OsRac1 binding interface within the coiled-coil region created by swapping the EF-hands. In addition, we demonstrate a direct intramolecular interaction between the N and C terminus, and that the complete N-terminal cytoplasmic region is required for this interaction. The structural features and intramolecular interactions characterized here might be common elements shared by Rbohs that contribute to the regulation of reactive oxygen species production.

The rapid generation of reactive oxygen species (ROS) 2 such as superoxide (O 2 . ) and hydrogen peroxide (H 2 O 2 ) is considered to be a component of the resistance response of plants to pathogens. ROS play multiple roles as signals that mediate responses to biotic and abiotic stresses, developmental cues, and pro-grammed cell death. ROS accumulation during early stages of plant defense reponse is known as the oxidative burst (1). The oxidative burst can further trigger the collapse of challenged host cells at the onset of the hypersensitive response and generate signals for defense gene induction. ROS are generated by plant NADPH oxidases and genes termed rboh (respiratory burst oxidase homolog) that encode homologs of the mammalian NADPH oxidase catalytic subunit gp91 phox , have been isolated from many plant species including rice, Arabidopsis, tobacco, and potato (2)(3)(4)(5)(6)(7).
The phagocyte enzymatic complex of NADPH oxidase (known as flavocytochrome b 558 ) consists of two membrane proteins, gp91 phox (Nox2) and p22 phox . Upon cell activation in mammals, the three cytosolic regulatory proteins, p47 phox , p67 phox , and p40 phox , and the small GTPase Rac translocate from the cytosol, assemble, and activate flavocytochrome b 558 at the cell membrane. However, the mammalian cytosolic regulatory proteins (p47 phox , p67 phox , and p40 phox ) and p22 phox are missing in plants (8). Furthermore, unlike the mammalian gp91 phox , all plant Rboh proteins carry a cytosolic N-terminal region with two EF-hands that bind Ca 2ϩ (Fig. 1A) (3,8). Although these motifs are absent from the mammalian gp91 phox , they are present in non-phagocytic NADPH oxidases such as Nox5, Duox1, and Duox2 (9,10).
Rboh is stimulated directly by Ca 2ϩ , likely through the N-terminal region (11). Recently, it has been reported that Ca 2ϩ activated the Arabidopsis NADPH oxidase, RHD2 (AtRbohC), to produce ROS at the growing tip of root hairs. Moreover, Ca 2ϩ binding to EF-hand motifs is required for activation of ROS production by RHD2 (12). Recent work has also demonstrated that calcium-dependent protein kinases activate potato RbohB (StRbohB) protein by phosphorylation of the N-terminal region (13), and that Ca 2ϩ binding and phosphorylation synergistically activate ROS production by AtRbohD (14). These results demonstrate that the extended N-terminal region plays a key role in the regulation of Rboh.
OsRac1, a Rac GTPase located in the plasma membrane, functions as a positive regulator of NADPH oxidase activity and is involved in defense responses in rice (15,16). In fact, constitutively active and dominant-negative forms of OsRac1 were found to activate and suppress ROS production, respectively. Direct interaction between the constitutively active OsRac1 and the N-terminal region of OsRbohB was verified by yeast two-hybrid analysis, in vitro pulldown assays, NMR titration experiments, and in vivo fluorescence resonance energy trans-fer microscopy (17). OsRac1 has also been shown to activate cinnamoyl-CoA reductase, a key enzyme in lignin biosynthesis, thus regulating the biosynthesis of lignin precursors (monolignols) that can be polymerized by H 2 O 2 -consuming peroxidases in the cell wall (18). These studies indicated that Rac GTPases interact coordinately with downstream pathways through regulation of multiple effector proteins. The elucidation of the regulatory mechanisms should be of great interest in plant signal transduction research.
Here we present the first crystal structure of the N-terminal region of OsRbohB, revealing that two molecules form a dimer with swapped EF-hands and that OsRbohB contains two additional EF-hand-like motifs so far not predicted from sequence analysis. We show that Ca 2ϩ binding to the EF-hands is necessary for the Ca 2ϩ -mediated conformational change. Structurebased mutagenesis and NMR studies revealed that key residues for the interaction of OsRbohB and OsRac1 are located in the coiled-coil region created by EF-hand swapping. Here we demonstrate for the first time a direct intramolecular interaction between the N terminus and C terminus for which the complete N-terminal cytoplasmic region is required. The highly conserved amino acid sequences of the N-and C-terminal regions suggest that the structural features and intramolecular interactions characterized here might be common elements shared by Rbohs that contribute to the regulation of ROS production.
Structure Determination and Refinement-The structure of OsRbohB-(138 -313) was solved by multiwavelength anomalous diffraction using the SeMet-labeled OsRbohB-(138 -313) crystal. Experimental phases were calculated up to 3.0-Å resolution with SOLVE (21) and improved by solvent-flattening with RESOLVE (22). An initial model was built by ARP/wARP (23), followed by COOT (24), and refined with CNS (25). The model was refined using REFMAC (26) with no NCS restraints and then cycled with rebuilding in COOT. TLS refinement was incorporated into the later stages of the refinement process with each chain as a separate TLS group. The model finally converged, resulting in a crystallographic R value of 23.4% and a free R value of 26.6% for all diffraction data up to 2.4-Å resolution. The final model quality was assessed using PROCHECK (27) and MolProbity (45). The Ramachandran plot of the final model, containing 333 amino acid residues, 2 calcium ions, and 77 water molecules, shows that all amino acid residues were in the most favored and allowed region as defined by PRO-CHECK. The final refinement statistics are summarized in Table 1 and the output of Molprobity is shown in supplemental Table S1. The figures were generated by PyMOL (28). The surface area was calculated by PROTORP (29). Coordinates for OsRbohB-(138 -313) are being deposited in the Protein Data Bank of the Research Collaboratory for Structural Bioinformatics (Protein Data Bank code 3A8R).
ESI-Mass Spectrometry (ESI-MS)-ESI mass spectra were acquired using a Q-Tof2 (Waters) with a nano-ESI ion source. Mass spectra from m/z 2000 to 6000 or 8000 were obtained by calibration with (CsI)nCsϩ. MassLynx version 3.5 software (Waters) was used for data processing and peak integration. The parameters for ESI-mass spectrometry were similar to those previously reported (45). The temperature of the ion source was set to 80°C. An aliquot of 4 l of sample solution was deposited in a nanospray glass tip (Waters) and introduced into a nano-ESI source. To observe protein complexes, the pressure in the quadrupole ion guide of the Q-Tof2 was maintained at 8 ϫ 10 Ϫ3 pascals by throttling down the speedivalve fitted to the rotary pump for the ion source region.
Small Angle X-ray Scattering (SAXS)-SAXS measurements were carried out at the RIKEN structural biology beam-line I (BL45XU) at SPring-8 (Hyogo, Japan). SAXS data were collected at 20°C on an imaging plate using Rigaku R-AXIS IV ϩϩ . The sample to detector distance was determined to be 2004 mm by silver behenate as a standard sample in the SAXS experiment. Three successive measurements were made for each solution with an exposure time of 100 s and wavelength of 1.0000 Å, and a circular average was made to obtain one-dimensional intensity data. The SAXS profiles of Ca 2ϩ -bound and Ca 2ϩ -free forms of OsRbohB-(138 -313) (0.9 mM) containing buffer A was compared. To detect structural changes induced by Ca 2ϩ binding, the middle-and high-angle ranges of the scattering curves, which contain information on secondary structures, were compared.

RESULTS
Overall Structure-The N-terminal region of OsRbohB containing residues 138 -313 (hereafter referred to as OsRbohB-(138 -313)) has a structure composed of 8 ␣-helices (␣A-␣H) and 2 short ␤-strands (␤1, ␤2). The amino acid sequence of the molecule, together with assignments of the secondary structure elements, is shown in Fig. 1B. Two molecules (molecules A and B) are present in the asymmetric crystal unit (Fig. 1C). Residues 304 -313 in molecule A and residues 305-313 in molecule B were disordered. Surprisingly, the C-terminal region (residues 263-303) of one molecule is swapped into the core domain (residues 138 -262) of the other molecule. As described below, the EF-hand of one molecule is packed together with the corresponding EF-hand of the other molecule (Fig. 2B, left). An intermolecular coiled-coil structure (residues 251-285) connects the two molecules (Fig. 1C). With the exception of the swapped region, the core domains were superposed well with root mean square deviations of 0.81 Å for C␣ atoms, whereas the superposition of the whole structures yielded root mean square deviations of 3.0 Å for C␣ atoms, indicating that orientation of the C-terminal region is different between the two molecules.
In the swapped dimer, there are two kinds of interfaces, one being the region between the swapped EF-hands and the other being a coiled-coil region formed by the swapping. The contact area of the swapped EF-hands is ϳ1133 Å 2 , and contains hydrophobic contacts as well as a few hydrogen bonds. The region that forms the coiled-coil structure (residues 260 -274) exhibits a loose packing with a contact area of 268 Å 2 , suggesting that the dimer is mainly stabilized by the swapped EF-hands and that the coiled-coil structure contributes to stabilization merely to a minor extent.
The assembled protomer structure composed of the core domain and the swapped C-terminal region has a three-dimensional structure similar to calcineurin B and recoverin (Fig. 2C), which contain four EF-hands in a highly compact, globular structure, in contrast to the dumbbell arrangement of calmodulin. In fact, a structural similarity search using DALI (30) revealed that the protomer structure of OsRbohB-(138 -313) shares structural similarity with calcineurin B (PDB accession code 1AUI) with a Z-score of 11.1 and a sequence identity of 16% (135 equivalent residues). It also revealed similarity with recoverin (PDB accession code 2d8n) with a Z-score of 8.4 and 14% sequence identity.
EF-hand-containing proteins such as calcineurin B and recoverin have a hydrophobic pocket on the side of the molecule opposite the EF-hands. In these proteins, the conformation of the C-terminal ␣-helix varies and affects the degree to which the hydrophobic pocket is exposed. OsRbohB-(138 -313) also possesses a hydrophobic pocket. Interestingly, the N-terminal amino acids 139 -156 of OsRbohB-(138 -313) adopt a helical conformation and are located in the hydrophobic pocket ( Number of Ca 2ϩ Ions in the Protein-One peak appeared on the difference Fourier map with a contour level higher than 5 in EF-1 ( Fig. 2A), although the OsRbohB amino acid sequence suggests the presence of two EF-hand motifs (Fig. 1A, EF-1 and EF-2). This single peak was scrutinized using an anomalous Fourier map calculated with FFT program in the CCP4 suite (31) based on the diffraction data collected at ϭ 1.6000 Å. The anomalous Fourier map showed a significant peak with contour levels above 7 corresponding to the single peak. We also confirmed the number of calcium ions coordinated in OsRbohB by ESI-MS. Fig. 3A 6 and 19,888.7). Following addition of EDTA, the Ca 2ϩ -bound species disappeared from the ESI mass spectrum (data not shown). This is consistent with the fact that only one peak was observed in the electron density map. Therefore, the protein binds one calcium mole per mol of protein. In contrast, the E253A mutant, in which Ca 2ϩ binding is expected to be disturbed because of the substitution of Glu, which is important for Ca 2ϩ coordination in the first EF-hand motif, gave protonated molecules of only one species in the ESI mass spectrum (Fig. 3A). The observed molecular mass of the E253A mutant was 19,793.6, well in line with the theoretical value of 19,792.6, corroborating that the E253A mutant does not bind Ca 2ϩ .
Two Predicted EF-hands Motifs and Two Unpredicted EFhand Motifs-The canonical Ca 2ϩ -binding EF-hand (EF-1) forms a conventional helix-loop-helix EF motif (␣F-␣G) and the calcium ion is coordinated in the typical pentagonal-bipyramidal geometry ( Fig. 2A, EF-1). In EF-1 of OsRbohB, the side chains of Asp 242 , Asn 244 , Asp 246 , and Glu 253 coordinate a Ca 2ϩ ion with Glu 253 as a bidentate substrate. Arg 248 ligates the Ca 2ϩ ion through its carbonyl oxygen and also coordinates a water molecule that ligates Ca 2ϩ . In contrast, the sequence of EF-2 (␣G-␣H) significantly deviates from the canonical one and no bound Ca 2ϩ was identified. The structure of EF-2 is stabilized through internal interactions. Asp 286 interacts with main chain atoms of Asn 289 , Leu 290 , and Gly 291 , and the main chain atom of Glu 294 is hydrogen-bonded to Asp 297 ( Fig. 2A, EF-2).
Generally, proteins containing EF-hand motifs have a domain formed by a pair of EF-hands. It should be noted that the EF-1 motif of one OsRbohB molecule makes extensive hydrophobic and hydrogen-bond contact with the EF-2 motif of its partner, thereby forming a domain in a similar way as conventional EF-hand pairs (Fig. 2B, left).
Sequence analysis suggested no EF-hands other than EF-1 and EF-2, but the structure definitely shows two additional EF-hands, EF-like 1 (␣B-␣C) and EF-like 2 (␣D-␣E; Fig. 2A). In general, the structural integrity of the EF-hand domain is maintained by the backbone hydrogen bonds in a short anti-parallel ␤-sheet, and by numerous hydrophobic contacts between the helices. The domain formed by EF-like 1 and EF-like 2 preserved this structural feature of the EF-hand, although the sequences of EF-like 1 and 2 are completely different from the canonical EF-hand motif. In fact, Phe 168 , Leu 177 , and Phe 182 of EF-like 1 are in close proximity to Ile 208 , Leu 213 , Leu 218 , and Phe 221 of EF-like 2 and form a hydrophobic core. Leu 177 of EF-like 1 and Leu 213 of EF-like 2 interact through backbone hydrogen bonds in such a way that they form an anti-parallel ␤-sheet (Fig. 2B, right).
Ca 2ϩ -mediated Conformational Change-We measured far-UV CD to examine the effect of Ca 2ϩ binding. Consistent with the threedimensional structure, the CD spectrum of OsRbohB-(138 -313) showed two negative minimum at around 208 and 222 nm, which is typical of ␣-helical proteins. The CD spectra differed between the Ca 2ϩ -bound and Ca 2ϩ -free forms (Fig. 3C, left). In contrast, the E253A mutant underwent no obvious conformational change following addition of Ca 2ϩ (Fig.  3C, right). This result suggests that OsRbohB changes its conformation in a calcium-dependent manner and may function as a molecular switch.
To gain further insight into the behavior of the protein in solution, we measured SAXS profiles of the wild type and E253A mutant of OsRbohB-(138 -313). The SAXS profile of OsRbohB-(138 -313) was modified by Ca 2ϩ in the q range from 0.2 to 0.35 Å Ϫ1 (Fig. 3D, left), whereas no such effect was observed in the E253A mutant (Fig. 3D, right). We performed in vitro pulldown assays using the wild type and E253A mutant of OsRbohB-(138 -313) to investigate the Ca 2ϩ dependence of the binding to OsRac1. Both proteins preferentially bound to OsRac1 in the GTP-bound form with a similar affinity, indicating that OsRac1 binding is not affected by Ca 2ϩ (Fig. 3E).
Interaction with OsRac1-We performed in vitro pulldown assays using OsRbohB-(138 -313) mutants to identify residues mediating the interaction of OsRbohB with OsRac1 (Fig. 4A). K147A in the N-terminal helix, E193A in ␣D, and E283A in EF-2 did not affect the binding affinity. The triple (V173D/ V176D/L178D) and quadruplex (Q170R/R181D/K184E/ D192K) mutations near the EF-like 1 motif maintained the binding affinity, and so did the triple (E252K/K255E/E256K) mutation in EF-1. These results suggested that the core domain of OsRbohB-(138 -313) is not part of the binding interface for OsRac1 (Fig. 4A). Similarly, the affinity did not change following truncation of 7 residues (⌬307-313) in the C terminus where the protein structure was disordered.
The deletion of 9 residues (⌬263-271) in the central coiledcoil region had little effect on interaction with OsRac1, but a 15-residue deletion (⌬260 -274) completely abolished OsRac1 binding, suggesting that the flanking region of the central coiled-coil region is essential for binding. Next, we mutated single residues in the flanking region. The single mutation of Arg 273 to Gln diminished the binding affinity. The Y277Q mutation and double mutations of Arg 273 and Tyr 277 markedly reduced the binding affinity. We concluded that the flat surface of the flanking region of the coiled-coil structure served as an OsRac1 binding interface.
Ca 2ϩ -independent Dimerization Necessary for Rac Binding-We then checked whether Ca 2ϩ binding affects the oligomeric state of OsRbohB. The purified protein was eluted from a gel-filtration column at a position corresponding to the homodimer (molecular mass ϳ39.3 kDa) both in the Ca 2ϩbound and the Ca 2ϩ -free forms (Fig. 3B), which were produced by treating the Ca 2ϩ -bound form with EDTA. The E253A  mutant lacking Ca 2ϩ binding affinity was also eluted from the same position. In addition, we measured SAXS profiles of the wild type and E253A mutant. In the presence of Ca 2ϩ , the molar masses of wild type and E253A mutant proteins were estimated to be 41 and 45 kDa, respectively, corresponding to dimers. The ab initio shape analysis of protein in solution by SAXS must be done using the SAXS data at infinite dilution to avoid an interparticle interference effect. Such SAXS data are obtained from SAXS measurements at various protein concentrations. OsR-bohB is, however, in equilibrium of dimer-monomer in solution at low protein concentrations, thus indicating that OsRbohB is unsuitable for ab initio shape analysis by SAXS. The shape analysis was therefore not carried out in the present study. Selfassociation was also confirmed for the wild type and E253A mutant by ESI-MS (data not shown). These data show that selfassociation occurs in a Ca 2ϩ -independent manner.
The elution profile of OsRbohB-(138 -313) after loading at a concentration of 1 mM onto a gel-filtration column showed a single but asymmetric peak, implying that the monomeric form was contaminated (Fig. 3B). In fact, a concentration dependence of dimer formation was observed as the protein was eluted as a monomer at a low concentration (40 M). These data suggest that OsRbohB-(138 -313) exists in a dynamic equilibrium between homodimers and monomers, and that the homodimer predominates at higher protein concentrations.
The OsRbohB-(138 -313)⅐OsRac1 complex was eluted from a gel filtration column at a position distinct from that of the OsRbohB-(138 -313) dimer. A 25:1 mixture of OsRbohB-(138 -313) and OsRac1 (1 mM:40 M) yielded a peak corresponding to the complex (Fig. 5A). However, a 1:25 mixture (40 M:1 mM) did not show the complex peak (Fig. 5B), probably because OsRbohB-(138 -313) did not form dimers. Likewise, a 1:1 mixture (40 M:40 M) at low total concentration provided no evidence for complex formation (Fig.  5C). These data strongly suggested that OsRbohB dimers are the functional units for Rac binding. Moreover, a 1:1 mixture at high total concentration (1 mM:1 mM) gave a large peak corresponding to the complex and also contained a peak representing unbound OsRac1 (Fig. 5D). This residual peak can be explained by assuming that OsRbohB binds to OsRac1 in a molar ratio of 2:1.
The OsRbohB-OsRac1 interaction was further analyzed by ESI-MS measurements. Peaks of 2:1 and 2:2 OsRbohB-(138 -313)⅐OsRac1 complexes were observed in the mass spectra in addition to the equimolar, free OsRbohB-(138 -313) and OsRac1 species. OsRbohB-(138 -313)⅐OsRac1 complexes may partly dissociate during ESI-MS analysis because of instability in the gas phase. ESI-MS shows that the OsRbohB-(138 -313) homodimer binds to one or two molecules of OsRac1, but does not support an unambiguous stoichiometry.
The Complete N Terminus Is Required for Binding to the C Terminus-In the case of Nox5, which possesses four EF-hand motifs in the N-terminal region, binding of Ca 2ϩ to EF-hands causes a conformational change and leads to intramolecular interaction of the EF-hand domain with the C-terminal domain, resulting in Nox5 activation (32). Therefore, we investigated whether the N-terminal region binds to the C-terminal domain in OsRbohB. Surprisingly, pulldown assays using OsRbohB-(138 -313) showed a very faint band (Fig. 6B), suggesting the EF-hand domain has a very weak affinity for the C-terminal region. When using OsRbohB-(1-137), the bound protein was somewhat more abundant, but the affinity between OsRbohB-(1-137) and C terminus still appeared weak. In contrast, OsRbohB-(1-355) including the complete N-terminal cytoplasmic region exhibited a strong affinity for the C-terminal region (Fig. 6B). OsRbohB-(1-306) showed a similar affinity as OsRbohB-(1-355), suggesting that residues 307-355 were not involved in the interaction. Unexpectedly, Ca 2ϩ seemed to have little effect on the interaction (Fig. 6C). We also checked whether OsRbohB-(1-355) had a binding affinity for the cytosolic loops connecting transmembrane helices 2 and 3, and helices 4 and 5; the results were negative (Fig. 6D). Taken together, intramolecular interaction between with the N and C terminus occurs in Rboh as observed in Nox5, but it is Ca 2ϩ -independent and requires the N-terminal region preceding the EF-hand domain in Rboh, in contrast to Nox5.

DISCUSSION
In the present study, we determined the first crystal structure of the N-terminal region that is important for regulation of Rboh. Our structural study revealed that the N-terminal cytoplasmic region contained four EF-hands (one functional EFhand, one disabled EF-hand, and two EF-like motifs) although sequence analyses had predicted two EF-hands. EF-1 of one molecule interacts with EF-2 of another one in a domain-swapping manner to form an EF-hand domain, and vice versa. Several EF-hand proteins, such as S100 proteins, calpain, and polcalcins, naturally occur as dimers. The structural unit of S100 proteins is a symmetric dimer comprised of two EF-hand motifs, which are organized into an eight-helix bundle (33). Calpain, the penta EF-hand family, dimerizes through an interaction between their fifth EF-hand motif (EF5) (34,35). Some polcalcins form a compact structure by domain-swapping between pairs of EF-hands (36). The dimer of RbohB exhibits an elongated shape and a coiled-coil region that forms a binding interface for Rac1. This unique structural feature is quite distinct from those in S100 proteins, calpain, or polcalcins.
The structural study also revealed that OsRbohB contains two unpredicted EF-hand-like domains, with sequences that differ markedly from those of classical EF-hands. Considering the conservation of the sequence of the N-terminal cytoplasmic region of plant NADPH oxidases (supplemental Fig. S2), they likely share a common structure featuring swapped EF-hands and unpredicted EF-hands. It is reported that enzymes of the fungal NoxC subfamily contain a single EFhand in the N-terminal cytoplasmic region (37). Based on our results, it appears possible that NoxC might possess an EF-hand domain comprising predicted EF-hands and unpredicted EF-hands.
Rac Binding Motif-To date, a number of Rac effectors have been identified and several Rac binding motifs have been classified. The CRIB (Cdc42/Rac-interactive binding) motif consisting of about 15 residues is a common sequence motif of Rac and Cdc42 effectors. Structural studies have shown that the CRIB motif forms an intermolecular ␤-sheet interaction with ␤-strand of Rac/Cdc42 (38,39). In addition, p67 phox , one of the subunits of phagocytic NADPH oxidases, contains tetratrico-peptide repeat domains that were identified as binding domains for Rac. The tetratrico-peptide repeat domain includes repeated ␣-helical motifs and ␤-hairpins. ␤-Hairpins and loops create the binding surface for Rac (40). Rac binds to the coiled-coil region of Arfaptin and the pleckstrin homology domain of phospholipase C-␤2 (41). In RbohB, residues important for Rac1 binding are located in the flanking region of the coiled-coil region. This binding region is dissimilar to those previously identified as Rac binding motifs. Thus, OsRac1 interacts with OsRbohB in a manner distinct from known interactions between Rac and its target proteins.
Dimer in Solution-Our structural and physicochemical studies showed that Rboh dimerizes in a concentrationdependent by domain swapping, and that the dimers probably represent functional units. In tobacco, Rboh has been shown to be localized in lipid rafts in the plasma membrane (42), implying that the local concentrations might be high enough for dimer formation due to macromolecular crowding effects. Recently the crystal structure of Steap3 was reported (46), showing the dimerization of the F 420 H 2 :NADP ϩ oxidoreductase domain. The F 420 H 2 :NADP ϩ oxidoreductase domain of Steap3 differs significantly from the ferredoxin-NADP ϩ reductase domain, which is a C-terminal cytoplasmic domain of NADPH oxidase. The dimerization of ferredoxin-NADP ϩ reductase in NADPH oxidases has not been reported. Rboh and Steap3 would function as a dimer, but it remains unknown whether dimerization is common for NADPH oxidases and other membrane proteins bearing the NADPH-and FAD-binding domains.
Proposed Regulation of Rboh by Ca 2ϩ and Rac1-We showed that Ca 2ϩ binding to the EF-hand in the N terminus of OsR-bohB induced a conformational change. This Ca 2ϩ -mediated change regulates Rboh activity, because Ca 2ϩ binding to the EF-hands is required for ROS production by Rboh (12,14). In fact, the E277Q mutation of AtRbohD, corresponding to the E253A mutation of OsRbohB, completely abolished ROS production (14). Similarly, the E250A mutation of RHD2, corresponding to the E253A mutation of OsRbohB, causes a 50 -60% decrease in activity (12). As we demonstrated, the mutation at this position (E253A) causes no Ca 2ϩ -mediated conformational change, ROS production would be expected to be abolished or decreased. Although the intramolecular interaction between the N and C terminus observed here was Ca 2ϩ independent, it is evident that Ca 2ϩ binding is essential for activation of OsRbohB. Probably, Ca 2ϩ -mediated conformational changes lead to a precise orientation of the cytoplasmic domains that are necessary for activation through interactions of the N and C domains. It remains to be determined how phosphorylation by Ca 2ϩ -dependent protein kinases as reported from potato (13) affects the regulatory mechanism in rice.
Rac1 is a component of the NADPH oxidase complex together with RACK1, RAR1, SGT1, HSP90, and HSP70 (43,44). The precise architecture of this Rac1 immune complex probably plays a crucial role in the production of ROS (supplemental Fig. S3). Further studies will reveal whether Ca 2ϩ -mediated conformational changes in Rboh contribute to Rac1 immune complex formation.