Identification of the binding sites for ubiquinone and inhibitors in the Na+-pumping NADH-ubiquinone oxidoreductase from Vibrio cholerae by photoaffinity labeling

The Na+-pumping NADH-quinone oxidoreductase (Na+-NQR) is the first enzyme of the respiratory chain and the main ion transporter in many marine and pathogenic bacteria, including Vibrio cholerae. The V. cholerae Na+-NQR has been extensively studied, but its binding sites for ubiquinone and inhibitors remain controversial. Here, using a photoreactive ubiquinone PUQ-3 as well as two aurachin-type inhibitors [125I]PAD-1 and [125I]PAD-2 and photoaffinity labeling experiments on the isolated enzyme, we demonstrate that the ubiquinone ring binds to the NqrA subunit in the regions Leu-32–Met-39 and Phe-131–Lys-138, encompassing the rear wall of a predicted ubiquinone-binding cavity. The quinolone ring and alkyl side chain of aurachin bound to the NqrB subunit in the regions Arg-43–Lys-54 and Trp-23–Gly-89, respectively. These results indicate that the binding sites for ubiquinone and aurachin-type inhibitors are in close proximity but do not overlap one another. Unexpectedly, although the inhibitory effects of PAD-1 and PAD-2 were almost completely abolished by certain mutations in NqrB (i.e. G140A and E144C), the binding reactivities of [125I]PAD-1 and [125I]PAD-2 to the mutated enzymes were unchanged compared with those of the wild-type enzyme. We also found that photoaffinity labeling by [125I]PAD-1 and [125I]PAD-2, rather than being competitively suppressed in the presence of other inhibitors, is enhanced under some experimental conditions. To explain these apparently paradoxical results, we propose models for the catalytic reaction of Na+-NQR and its interactions with inhibitors on the basis of the biochemical and biophysical results reported here and in previous work.

To understand the mechanistic details of respiratory enzymes that use the quinone/quinol molecule as a substrate, identification of the binding site for quinone/quinol at the amino acid level is critical. Recently, by alanine-scanning mutagenesis of aromatic residues in NqrB and molecular docking, Tuz et al. (14) proposed that the binding site of the ubiquinone ring is located at the interface of the NqrB and NqrD subunits. The predicted site is not only inconsistent with the location of the putative cavity in NqrA (13), but is also too far from the riboflavin (ϳ36 Å) for efficient electron transfer. Because the work by Tuz et al. (14) was largely based on steadystate kinetic analysis of mutants, it is difficult to determine whether the critical residues that they identified (NqrB-Phe-211 and -Phe-213) are directly involved in forming the binding pocket for the ubiquinone ring or whether they affect the reaction of ubiquinone indirectly though some long-range conformational change. In addition, to determine the number of binding sites and/or the dissociation constant for short-chain ubiquinone and the inhibitor HQNO, previous studies employed the equilibrium dialysis method with isolated wild-type and mutated Na ϩ -NQR (10,14). However, interpretation of data obtained through this method can be problematic because it may be impossible to distinguish between specific and non-specific binding of the hydrophobic ligands to the enzyme, particularly in the case of an integral membrane protein. Thus, there remains considerable controversy regarding the number and location of quinone-binding site(s) in Na ϩ -NQR.
As specific inhibitors of many respiratory enzymes disturb the enzyme function, in general, by interfering with the redox reaction of quinone/quinol, identification of the binding site for inhibitors is also invaluable. The inhibitors that have been used for previous Na ϩ -NQR research are limited to a few commercially available chemicals such as HQNO (8,9), which has only moderate inhibitory potency (IC 50 or K i value in the micromolar range). Korormicin (Fig. 2), a natural antimicrobial produced by some marine bacteria, was reported by Hayashi et al. (15) to be a very potent inhibitor of Na ϩ -NQR from V. cholerae with a K i about 3000-fold tighter than that of HQNO. On this basis, korormicin can be anticipated to be a useful molecular tool for Na ϩ -NQR research. However, it is no longer possible to acquire korormicin because the Japanese research institute, where it was originally isolated, has been closed. For this study, we made a fresh isolation of korormicin from culture broth of marine bacterium Pseudoalteromonas J010.
Additionally, through extensive screening of potential inhibitors of V. cholerae Na ϩ -NQR from the chemical library of our laboratory, we identified an aurachin derivative (aurachin D-42, Fig. 2) as a very effective inhibitor of the enzyme, with a potency almost identical to that of korormicin. With the two strong inhibitors in hand, we have used a photoaffinity labeling technique to identify the binding sites for ubiquinone and aurachintype inhibitors in isolated V. cholerae Na ϩ -NQR. We also investigated the mechanism of inhibition of aurachin derivatives using the mutated enzymes G140A, G141V, and E144C in the NqrB subunit, in which the effects of these inhibitors are significantly attenuated (8,10,15). Some of the current results were initially difficult to reconcile. For example, there are two mutants in transmembrane helix II of the NqrB subunit (G140A and E144C), in which aurachin derivatives lose almost all inhibitory potency, even though the effectiveness of labeling to the mutated enzymes is essentially unchanged compared with the wild-type enzyme. In an attempt to explain these apparently paradoxical results, we propose models for the catalytic reaction of Na ϩ -NQR and the interaction with inhibitor on the basis of the biochemical and biophysical information obtained in this study and previous works.

Korormicin and aurachin D-42
Korormicin was isolated from the culture broth of the marine bacterium Pseudoalteromonas strain J010 according to the method originally described by Yoshikawa et al. (16). The spectra of our isolated sample completely match those reported previously (16,17), as described under "Experimental procedures." Aurachin D-42 was identified as a very effective inhibitor of V. cholerae Na ϩ -NQR from the chemical library of our laboratory. The inhibitory potencies of korormicin and aurachin D-42 were examined against the NADH-Q 1 oxidoreductase activity of Na ϩ -NQR using the isolated enzyme (0.9 nM) and measuring the reduction of quinone at 282 nm. The IC 50 values of korormicin and aurachin D-42 were 5.0 (Ϯ 0.7) and 2.0 (Ϯ 0.3) nM, respectively. Thus, korormicin and aurachin D-42 are both ϳ1000-fold more potent than commercially available HQNO (an aurachin C-type analog), which has been widely used in previous studies (8,15).

Design of photoreactive ubiquinone and aurachin derivatives
We have synthesized several photoreactive ubiquinone derivatives for previous photoaffinity labeling studies for other membrane proteins (18 -20). However, a preliminary test of these derivatives showed that all are poor electron acceptors for V. cholerae Na ϩ -NQR with catalytic efficiencies ϳ15-30% that of Q 1 . Therefore, we newly synthesized PUQ-3, which has a shorter side chain compared with the previous ubiquinone derivatives. An alkyne group (-C'CH) was attached to the terminal end of the side chain to allow the conjugation of molecular tags, such as fluorophores and biotin, via Cu ϩ -catalyzed click chemistry (i.e. azide-alkyne [3ϩ2] cycloaddition in water (21)) to detect the labeled peptide. PUQ-3 turned out to be an efficient electron acceptor for Na ϩ -NQR with K m and V max values of 8.3 (Ϯ 0.2) M and 14.8 (Ϯ 1.9) mol/min/mg of protein, respectively. Under the same experimental conditions, the K m and V max values of Q 1 were 8.3 (Ϯ 0.3) M and 24.3 (Ϯ 3.1) mol/min/mg of protein, respectively.
Next, we synthesized two photoreactive aurachin derivatives using aurachin D-42 as a molecular template (PAD-1 and PAD-2, see Fig. 2). Both derivatives incorporate photoreactive azido (-NϭN ϩ ϭN Ϫ ) groups; the azido group is on the quinolone ring in PAD-1, although it is attached to the side chain moiety in PAD-2. Introduction of radioactive 125 I label into the two derivatives ([ 125 I]PAD-1 and [ 125 I]PAD-2) makes it possible to work with the lowest possible concentrations, thus minimizing the probability of non-specific labeling.
The inhibitory activities of PAD-1 and PAD-2, in terms of IC 50 value, were 590 (Ϯ45) and 1.6 (Ϯ0.21) nM, respectively. Thus, although PAD-2 retains the potency of the native aurachin, at the single digit nanomolar level, PAD-1 is significantly less effective, presumably because of steric hindrance at the toxophoric quinolone ring, arising from the presence of the azido group. We changed the substitution position of the azido group on the quinolone ring to the 5-, 7-, or 8-position; however, the inhibitory potency did not recover. Nevertheless, we included [ 125 I]PAD-1 in the photoaffinity labeling study because it is still a much stronger inhibitor than HQNO.
In the case of korormicin, our preliminary structure-activity studies revealed that the methyl/ethyl branch and OH group Binding sites for ubiquinone and inhibitors in Na ؉ -NQR located at the 5S and 3ЈR positions, respectively, are critical for inhibitory activity. 3 Because of these complications in structural factors, synthetic production of photoreactive korormicin derivatives, which maintain as much inhibitory potency as possible, is still ongoing in our laboratory.

Photoaffinity labeling of Na ؉ -NQR by PUQ-3
Na ϩ -NQR that has been purified using the non-ionic detergent dodecyl maltoside (DDM) is reported to contain tightly bound ubiquinone-8 (Q 8 ) in substoichiometric quantities that can be removed by washing the enzyme preparation with a buffer containing the zwitterionic detergent lauryldimethylamine N-oxide (LDAO) (7,10). We used LDAO-washed Na ϩ -NQR throughout this study.
Na ϩ -NQR (0.9 M) was incubated with PUQ-3 (5 M) on ice, irradiated with a UV lamp, then denatured with 1% (w/v) SDS, and covalently conjugated with a fluorescent TAMRA-N 3 via Cu ϩ -catalyzed click chemistry. The resulting proteins were separated on a 15% Laemmli-type SDS gel containing 6.0 M urea. Under these analytical conditions, Na ϩ -NQR was separated into six bands corresponding to the six subunits of the enzyme (Fig. 3A). The two bands, which migrated to the region of ϳ50 kDa, were confirmed to be the hydrophilic NqrF subunit, as demonstrated by Western blotting using an antibody against the N-terminal His tag, and the hydrophilic NqrA subunit, as demonstrated by MALDI-TOF MS (13 peptides, 30% coverage). Two other bands, which appeared in the region of ϳ30 kDa, were determined to be the hydrophobic NqrB and NqrC subunits on the basis of their N-terminal amino acid sequences. As these subunits each carry one covalently bound FMN (Fig. 3A, center), the FMN-based fluorescence was used as a loading control for SDS-PAGE throughout this study (22,23). Although the NqrD and NqrE subunits were invisible in this gel, they become visible by heavy staining (data not shown).
As shown in Fig. 3A, a major TAMRA fluorescence was observed at the band corresponding to the NqrA subunit at ϳ50 kDa. The fluorescence intensity incorporated into the NqrA subunit was reduced when labeling was carried out in the presence of an excess of Q 2 or 3-N 3 -Q 2 (100 M each, 20 molarfold) (Fig. 3B), indicating specific binding of PUQ-3 to Na ϩ -NQR. The suppressive effects of aurachin D-42 and korormicin (10 M each, 2 molar-fold) on the labeling were significantly smaller than those of the short-chain quinones (Fig. 3B), even though the binding affinities of the potent inhibitors are at least 2 orders of magnitude stronger than that of PUQ-3. It may be noted that the NqrB subunit was also labeled faintly, but this minor labeling of NqrB was not examined in further detail.

Localization of the labeled site by PUQ-3
To localize the binding site of PUQ-3 in the NqrA subunit, the labeled NqrA band was subjected to in gel partial digestion with V8-protease (Cleveland mapping). As shown in Fig. 4A, this partial digestion reproducibly gave a major fluorescent band at ϳ12 kDa (fragment A) containing the internal sequence corresponding to Gly-7-Lys-24 of NqrA, (confirmed by MALDI-TOF MS, m/z ϭ 1767.96), and an N-terminal sequence, H 2 N-1 MITIKK, as determined by Edman degradation. These results indicate that fragment A is the N-terminal region Met-1-Glu-106 (11.5 kDa) of the subunit (Fig. 4C). The partial digestion also produced an ϳ8-kDa fragment B with weak fluorescence that was identified as the peptide Val-94 -Arg-172 (8.6 kDa) by Edman degradation (N-terminal sequence, H 2 N-94 VAGDD) and MALDI-TOF MS (fragment, Thr-120 -Arg-132, m/z ϭ 1507.80, see Fig. 4C).
Next, the NqrA subunit labeled with PUQ-3 was cleaved using CNBr, producing a major fluorescent band whose apparent molecular mass was ϳ4 kDa (fragment D, Fig. 4B). Unfortunately, we were unable to identify fragment D by Edman degradation because multiple peptides from the CNBr digestion were contained in this region. However, considering the result of the V8 digestion and the theoretical cleavage sites for CNBr in the sequence, fragment D should be the peptide Ile-2-Met-39 (3.9 kDa), which contains a major binding site for PUQ-3 (Fig. 4C).
Another weakly fluorescent band from the CNBr digests of NqrA (fragment C) was assigned as His-44 -Met-154 (12.1 kDa) by MALDI-TOF MS analysis of the tryptic digests (Phe- 105-Arg-115 and Thr-120 -Arg-132, m/z ϭ 1262.68 and 1507.80, respectively). Fragment C shares sequence with fragment B (above) (Fig. 4B). Therefore, the region Val-94 -Met-154, where fragments B and C overlap, may contain a minor binding site of PUQ-3. These two regions, which are likely to be involved in PUQ-3 binding, are highlighted in the structure in Fig. 1B; Ile-2-Met-39 is shown in light and dark blue, and Val-94 -Met-154 is shown in light and dark red, respectively. In the crystallographic model, these two segments of NqrA are in close contact, but it must be remembered that this may not reflect the actual structure present throughout the catalytic cycle. Also, as discussed later, the fact that two regions are labeled by PUQ-3 does not necessarily indicate the presence of two distinct binding sites.

Photoaffinity labeling of Na ؉ -NQR by [ 125 I]PAD-2
A sample of Na ϩ -NQR (0.9 M), which was washed with a buffer containing LDAO, was incubated with [ 125 I]PAD-2 (10 nM) in the presence of NADH and/or Q 1 , irradiated with a UV lamp on ice, resolved on a 15% Laemmli-type SDS gel, and subjected to direct autoradiography. Because the detection sensitivity with 125 I labeling is quite high, we set the concentration of [ 125 I]PAD-2 as low as possible to reduce the probability of non-specific labeling. The radioactivity was exclusively incorporated into an ϳ30-kDa band, which corresponds to the NqrB subunit (Fig. 5A). The amount of radioactivity incorporated was ϳ50% lower in the presence of 100 M NADH (Fig. 5A), but not in the presence of Q 1 , suggesting that reduction of the cofactors may induce a structural change of the NqrB subunit.
The labeling of Na ϩ -NQR (0.9 M) by [ 125 I]PAD-2 (10 nM) was repeated in the presence of other inhibitors as follows: aurachin D-42, korormicin, or PAD-2 (20 M, 2000 mole excess over [ 125 I]PAD-2 in each case). Surprisingly, in the presence of each competitor, the amount of radioactivity incorporated into the NqrB subunit was significantly increased (Fig. 5B). It is important to note that this enhancement was observed in the presence of PAD-2, whose physicochemical properties are completely identical to those of [ 125 I]PAD-2, except for the radioactivity. To exclude the possibility that the enhancement of labeling is merely due to some structural alteration of the enzyme brought about by the high concentrations of hydrophobic chemicals, we conducted the same experiment in the presence of 20 M bullatacin or antimycin A, which are potent and hydrophobic inhibitors of mitochondrial respiratory complexes I and III, respectively. Bullatacin and antimycin A did not affect the labeling yield of the NqrB subunit at all (data not shown). Thus, the enhancement of labeling must be due to specific binding of the competitors to Na ϩ -NQR.
We next changed the concentrations of aurachin D-42 and PAD-2, fixing the concentrations of Na ϩ -NQR (0.9 M) and [ 125 I]PAD-2 (10 nM). The extent of the enhancement of labeling increased as the concentration of the competitors increased but then decreased slightly (Fig. 5C). We did not use concentrations of competitors over 50 M because high concentrations of hydrophobic chemicals can disturb the enzyme structure in a non-specific manner.
To gain further insight into this seemingly peculiar phenomenon, we changed the molar ratio of Na ϩ -NQR to [ 125 I]PAD-2 (or competitor) by lowering the concentration of Na ϩ -NQR from 900 to 0.9 nM, while maintaining the concentrations of [ 125 I]PAD-2 and aurachin D-42 at 10 nM and 10 M, respectively. Interestingly, the extent of enhancement of labeling brought about by aurachin D-42 decreased with a decrease in the concentrations of Na ϩ -NQR (Fig. 5D). The suppressive effect on the labeling became clearer at low concentrations of Na ϩ -NQR (0.9 and 9.0 nM). These results strongly suggest that the molar ratio of Na ϩ -NQR to [ 125 I]PAD-2 (or Na ϩ -NQR to competitor) is a critical factor determining the total amount of labeled enzyme. These effects of competitors on the photoaffinity labeling will be discussed below in the context of a proposed kinetic model.

Photoaffinity labeling of Na ؉ -NQR by [ 125 I]PAD-1
We also carried out photoaffinity labeling of Na ϩ -NQR (0.9 M) using [ 125 I]PAD-1 (2.6 nM). As in the case of labeling with , conjugated with TAMRA-N 3 via Cu ϩ -catalyzed click chemistry, followed by the separation on a 15% Laemmli-type SDS gel containing 6.0 M urea. The CBB-stained gel fragment containing the NqrA subunit (ϳ5 g of protein) was further subjected to in-gel partial digestion with V8-protease, followed by the separation on a 15% Tris-EDTA mapping gel. The N-terminal sequences of the CBB spots corresponding to the fluorescent spots were determined by Edman degradation. B, CNBr cleavage of the NqrA subunit labeled by PUQ-3. The NqrA labeled by PUQ-3 was conjugated with TAMRA-N 3 and cleaved with CNBr in 70% formic acid. The digests, which are equivalent to 10 g of protein, were resolved by Schägger-type SDS gel (16% T, 6% C, containing 6 M urea). C, schematic presentation of the digestion of the NqrA subunit. The predicted cleavage sites are denoted by arrows and marked with their residue numbers in the sequences of the V. cholerae NqrA subunit (SwissProt entry Q9KPS1).

Binding sites for ubiquinone and inhibitors in Na ؉ -NQR
[ 125 I]PAD-2 (above), the radioactivity was exclusively incorporated into the NqrB subunit (Fig. 6A), and the yield was slightly lower in the presence of 100 M NADH. It is important to note that although the inhibitory potency of [ 125 I]PAD-1, in terms of IC 50 (0.9 nM Na ϩ -NQR), is much weaker than that of [ 125 I]PAD-2, the effectiveness of the labeling by [ 125 I]PAD-1 was comparable with that of [ 125 I]PAD-2. This was due to a much higher concentration of the enzyme (0.9 M) used for the photoaffinity labeling experiment. To examine whether the unusual effect of competitors is also seen with [ 125 I]PAD-1, the photoaffinity labeling was repeated in the presence of an excess of aurachin D-42, korormicin, or PAD-1 (10 M each, 4000 molefold of [ 125 I]PAD-1). In contrast to the case of [ 125 I]PAD-2, there was no clear enhancement of the labeling; but no significant competitive suppression was observed (Fig. 6B). It is noteworthy that the suppressive effect of aurachin D-42 became clearer at low concentrations of Na ϩ -NQR (Fig. 6C). Viewed in this light, the unusual effect of competitors on the pho- toaffinity labeling is a phenomenon common to [ 125 I]PAD-1 and [ 125 I]PAD-2, whereas the extent of the competitive suppression and the enhancement of the labeling differ between the two inhibitors because of their different chemical structures.
The Asp-N digestion of NqrB labeled by [ 125 I]PAD-1 or [ 125 I]PAD-2 afforded the same radioactive band with a molecular mass of ϳ9 kDa (A1) (Fig. 7A), which contains no FMN. Along with the results from the Lys-C digestion, this indicates that the ϳ9-kDa bands may be region Asp-9 -Gly-89 (9.2 kDa). Thus, [ 125 I]PAD-1 and [ 125 I]PAD-2 label the N-terminal regions Arg-43-Lys-54 and Trp-23-Gly-89 (Fig. 7B), respectively, a part of the structure that protrudes from the membrane and forms a long stretch that anchors NqrA to the trans-membrane subunits (Fig. 1C). We cannot rule out the possibility that [ 125 I]PAD-2 labels two (or more) positions in the region Trp-23-Gly-89 because the side chain moiety of aurachin flexibly binds to the enzyme.

Inhibitor sensitivities of mutated Na ؉ -NQR
Hayashi et al. (15) discovered that a spontaneous mutation in the NqrB subunit of Vibrio alginolyticus Na ϩ -NQR, replacing Gly-140 with Val (corresponding to NqrB-Gly-141 in V. cholerae Na ϩ -NQR), confers a remarkable resistance against korormicin compared with the wild-type enzyme (approximately 5 orders of magnitude). In subsequent studies (8,10), working with the V. cholerae Na ϩ -NQR, Barquera and co-workers demonstrated important functional roles for NqrB-Gly-140 and -Gly-141 in the electron transfer to ubiquinone. However, the effects of mutations at these positions on the inhibition of Na ϩ - NQR by korormicin were not studied at that time because the inhibitor was not available.
We have now examined changes in the inhibitor sensitivity of five mutant enzymes, three with substitutions in helix II of NqrB, near the spontaneous mutation identified by Hayashi et al. (15) (G140A, G141V, and E144C in the NqrB subunit), and two at the rear wall of the putative ubiquinone-binding cavity (Y36A and G38V in the NqrA subunit), as described above, using korormicin, aurachin D-42, PAD-1, PAD-2, and HQNO ( Table 1). The inhibitor sensitivity, in terms of IC 50 value, of NqrA-Y36A and -G38V were essentially unchanged compared with the wild-type enzyme. In contrast, NqrB-G140A and -E144C exhibited remarkable resistance against all inhibitors tested (at least 1000-fold); the inhibitors lost their effectiveness almost completely. For NqrB-G141V, the inhibitor sensitivities to aurachin D-42, PAD-2, and HQNO were very similar to those of the wild-type enzyme, whereas the sensitivities to PAD-1 and korormicin were reduced by a factor of ϳ10 and ϳ100, respectively. These results indicate that NqrB-Gly-140 and -Glu-144, in transmembrane helix II of NqrB, are critical for the interference of the enzyme function by these inhibitors, whereas Gly-141 appears to play a lesser role.

Photoaffinity labeling of Na ؉ -NQR mutants by [ 125 I]PAD-2
The remarkable resistance of the NqrB-G140A and -E144C mutants against the inhibitors would typically indicate that the inhibitors are no longer able to bind to the mutated enzymes.
To check this, we carried out the photoaffinity labeling on the mutants using [ 125 I]PAD-2. Unexpectedly, the labeling yields did not decrease compared with the wild-type enzyme, but rather they remained the same or, in the case of NqrB-G140A, increased (Fig. 8A). The fact that the mutated enzymes were labeled in apparently the same manner as the wild-type enzyme indicates that the sites that bind the inhibitors in the wild type have not been destroyed by the mutations, although the inhibitory effect has disappeared. It is noteworthy that the presence of aurachin D-42 or korormicin significantly enhanced the labeling of NqrB-E144C by [ 125 I]PAD-2 (Fig. 8B), as observed for the wild-type enzyme (Fig. 5B). For the G140A mutant, labeling was not affected by the presence of aurachin D-42 or korormicin. The labeling yields with the NqrA-Y36A and -G38V mutants were unchanged compared with the wild-type enzyme (Fig. 8A).

Discussion
Na ϩ -NQR from V. cholerae has been extensively investigated using a wide variety of techniques; however, the positions of the binding sites for ubiquinone and inhibitors in this enzyme remain controversial. Steuber et al. (13) identified a deep solvent-accessible cavity, in the NqrA subunit, in their crystallographic structure, and pointed out that it could accommodate a ubiquinone molecule. Now, to more precisely locate the binding sites for ubiquinone and inhibitors, we have carried out photoaffinity labeling experiments with a photoreactive ubiquinone (PUQ-3) and two different aurachin-type inhibitors ([ 125 I]PAD-1 and [ 125 I]PAD-2) using the isolated Na ϩ -NQR.
The photoaffinity labeling experiment using PUQ-3 showed that the ubiquinone ring binds primarily in the region Ile-2-Met-39 of NqrA and secondarily in the region Val-94 -Met-154 of the same subunit (Fig. 4). We were unable to pinpoint the labeled residue(s) exactly because of low labeling yield. Nevertheless, some parts of these regions can be ruled out as candidates for the binding site of the ubiquinone ring because they lie on the exterior surface of NqrA, on the opposite side of the subunit from the likely position of the riboflavin cofactor in NqrB, and thus too far away for efficient electron transfer; these include the sequence stretches Ile-2-Leu-31 and Val-94 -Phe-130, thus leaving Leu-32-Met-39 and Phe-131-Met-154 as the remaining candidates. The fact that there is labeling in two different stretches of the sequence does not necessarily require the existence of two different binding sites for ubiquinone. In the crystallographic structure, the folding of NqrA brings the two labeled stretches together, and ubiquinone could adopt two slightly different bound conformations at the same locus; the rod-like azido (ϪNϭN ϩ ϭN Ϫ ) group would then cross-link in different places reflecting the two conformations. The two labeled stretches come into closest contact in the region of Leu-32-Met-39 and Phe-131-Lys-138 of NqrA (orange circle in Fig.  1B), which together form a highly plausible binding location for the ubiquinone ring. It is important to note that, although these stretches of polypeptide form the rear wall of the putative ubiquinone-binding cavity in the crystallographic structure (13), the actual size, shape, and position of the cavity may not be the same at all times throughout enzyme turnover, as mentioned in the Introduction. The labeling by PUQ-3 was effectively suppressed by short-chain quinones (Q 2 and 3-N 3 -Q 2 ) confirming the likely role of this location as the catalytic quinone reduction site. However, labeling was not suppressed by the potent inhibitors aurachin and korormicin (Fig. 3B), indicating that the binding position of ubiquinone does not overlap those of the inhibitors. A previous photoaffinity labeling study using a different photoreactive ubiquinone also showed that the ubiquinone ring binds to NqrA (7); however, in that case the labeled region was not further defined.
Photoaffinity labeling with [ 125 I]PAD-1 and [ 125 I]PAD-2 showed that the toxophoric quinolone ring of aurachin and its alkyl side chain both bind to NqrB, but in two different stretches of polypeptide (Arg-43-Lys-54 and Trp-23-Gly-89, respectively) within the N-terminal portion of the subunit. These two segments are part of a region that protrudes from the membrane and forms a long stretch that attaches to the NqrA subunit, anchoring it to the integral membrane subunits of the enzyme (Fig. 1C). Although the structure of the region Met-1-Pro-37 is not included in the model of Steuber et al. (13), this Table 1 The

inhibitory activities of different inhibitors with the wild type and mutated Na ؉ -NQR
The inhibitory activity, in terms of the IC 50 value (nM), was examined with the NADH-Q 1 oxidoreductase activity using isolated Na ϩ -NQR (0.9 nM). When 50% inhibition was not observed, the residual enzyme activities (%) at the indicated concentrations are shown in parentheses. Data shown are mean values Ϯ S.D. (n ϭ 3-4).  N-terminal region would apparently form a part of the wall of the putative quinone-binding pocket identified above, because Gly-38 -Leu-53, which follows immediately in the sequence, lies in a groove neighboring the cavity in NqrA. Together, the results from the photoaffinity labeling with PUQ-3, [ 125 I]PAD-1, and [ 125 I]PAD-2 indicate that the binding positions of ubiquinone and aurachins do not overlap one another, even though aurachin-type inhibitors resemble the ubiquinone molecule in structure. It is thus likely that the inhibitors perturb the reaction of ubiquinone indirectly by binding to a site or sites in the N-terminal region of NqrB and causing a conformational change. This notion is consistent with the results obtained in earlier steady-state kinetic studies (11,24,25), which showed that HQNO and korormicin do not compete directly with ubiquinone. It is generally thought that "quinone-analog" inhibitors of respiratory enzymes occupy the quinone-binding cavity in a manner competitive with quinone. However, this is not necessarily true of all respiratory inhibitors. For example, piericidin A is extremely similar to the ubiquinone molecule but is a highly selective inhibitor of mitochondrial complex I and has no effect on other enzymes, including those that interact with quinone/ quinol. Moreover, there are a number of inhibitors that work by binding to the quinone-binding cavity of various enzymes but lack any strong structural resemblance to ubiquinone, for example antimycin A analogs, inhibitors of Qi site in complex III (26), and various inhibitors of Qo site in complex III (e.g. myxothiazol and famoxadone) (27). Thus, the modes of action of respiratory inhibitors cannot always be inferred from their structural similarity to ubiquinone.

HQNO
The results obtained in this study raise two interesting questions. First, in light of the fact that PAD-1 and PAD-2 lose their inhibitory activities almost completely against the NqrB-G140A and -E144C mutants (Table 1) (Figs. 5D and 6C)? It is obvious that these results are not consistent with a simple scenario in which the different inhibitors share a common binding pocket, and where NqrB-Gly-140 and -Glu-144 lie within this pocket and interact directly with the inhibitors. Instead, we must assume that there is a dynamic interaction between the adjacent subunits NqrA and NqrB, which regulates the reaction of ubiquinone and inhibitors with the enzyme. In this scenario, NqrB-Gly-140 and -Glu-144 would play key roles in the interaction of the two subunits.
To address the first question, we will present models for the catalytic reaction of the wild-type Na ϩ -NQR (Fig. 9A) and the mutants (Fig. 9B) in the presence and absence of inhibitor. These models are based on the results of our previous kinetic study of Na ϩ -NQR showing that ubiquinone binds to Na ϩ -NQR only after the enzyme becomes reduced by accepting electrons from NADH (11). In addition, we will adopt the working hypothesis that a large conformational change in the NqrB sub-unit takes place during enzyme turnover, shortening the distance between the ubiquinone-binding pocket in NqrA and the riboflavin in NqrB, thus facilitating efficient electron transfer (supposition i). Because the existence of a high-affinity bound quinone in isolated Na ϩ -NQR (in substoichiometric amounts) and its catalytic role have not yet been definitively established, we will not discriminate between bound and catalytic quinones in this model.
To assess validity of supposition i, we again consider the structural information obtained from the crystallographic study of V. cholerae Na ϩ -NQR (13). The distances between several pairs of redox centers in the proposed electron transfer pathway are too long for physiologically relevant electron transfer: for example, between the [2Fe-2S] in NqrF and the iron in NqrD-E and between the FMN in NqrB and the riboflavin in NqrB. Facile electron transfer between these redox centers cannot be explained without the assumption that large conformational changes take place during enzyme turnover and that these changes have the effect of decreasing each one of these electron transfer distances at some point during the catalytic cycle (13). The same is true for the distance between the "crystallographic" positions of the putative ubiquinonebinding cavity in NqrA and the riboflavin in NqrB. According to our model, the catalytic cycle of wild-type enzyme would proceed as shown in Fig. 9A. Although the ubiquinone-binding pocket is initially located far from the riboflavin in the oxidized form of Na ϩ -NQR, a conformational change in the NqrB subunit (or at the interface between NqrA and NqrB), triggered by reduction of one or more cofactors, including the riboflavin in NqrB, causes the spatial gap between the riboflavin and the ubiquinone ring to decrease, thus facilitating efficient electron transfer. The upward movement of the NqrB subunit shown in Fig. 9A represents this conformational change.
The bound inhibitor could serve as a "wedge" at the interface between NqrA and NqrB, which prevents the conformational change in NqrB required for efficient electron transfer.
Our model also accounts for the effects of mutations on the same steps in the catalytic cycle, as described above. Our earlier FTIR spectroelectrochemical study demonstrated that the NqrB-G140A mutation does not alter the midpoint potentials of any of the redox cofactors or the overall structure of the enzyme, compared with the wild type. The mutant enzyme has a much lower V max for steady-state turnover, but the K d value for quinone is not significantly changed from wild type. Consistent with this, mutation markedly decreases the rates of both the forward electron transfer from the riboflavin cofactor to ubiquinol (10) and reverse reaction (8). Taken together, these results suggest that a change in the size of the residue at position 140 in the NqrB sequence has an indirect effect on the reaction of ubiquinone in some way but also that this residue is not part of the domain that makes up the actual binding site for ubiquinone. This is illustrated in Fig. 9B, where the distance between the riboflavin and the quinone site has been slightly increased relative to the wild-type enzyme (Fig. 9A) to reflect a perturbed but still functioning electron transfer step between the two sites (10). It is worth noting that the same FTIR spectroelectrochemical study also found evidence for structural differences in the environment surrounding a bound inhibitor (HQNO) between the wild-type enzyme and the mutants (NqrB-G140A and -G141V) (10). Thus, it appears likely that in the mutants the structural rearrangement of NqrB, due to the reduction of cofactors, may be able to take place even in the presence of bound inhibitor. This is illustrated in our scheme (Fig. 9, A and  B); in the wild-type enzyme, the inhibitor serves as a wedge to block the normal conformational change (shown as upward shift of NqrB), preventing a decrease in the distance between the riboflavin cofactor and ubiquinone and thus impeding fast electron transfer (Fig. 9A). In the mutants, a structural alteration of the pocket surrounding the inhibitor creates sufficient additional space or flexibility that the inhibitor is no longer able to prevent the conformation change, and thus facile electron transfer from riboflavin to ubiquinone takes place Figure 9. Models for the catalytic reaction of the wild-type Na ϩ -NQR (A) and the G140A mutant (B). The reduction of cofactors by NADH induces a conformational change of NqrB (indicated as an upward movement of the subunit), which decreases the gap between riboflavin and the ubiquinone-binding pocket, facilitating electron transfer. The bound inhibitor serves as a "wedge" and interferes with the conformational change of NqrB required for the electron transfer. The G140A mutant may be able to undergo the conformational change even in the presence of inhibitor because of structural alteration of the environment surrounding the inhibitor. C, model of the photoaffinity labeling with the wild-type Na ϩ -NQR in the presence of competitor. The model postulates two inhibitor-binding sites in Na ϩ -NQR to give three inhibitor-bound states (states "OI," "IO," and "II"), which are defined by three association constants K 1 , K 2 , and K 3 . We assume that K 3 is negligibly small compared with K 1 and K 2 and, crucially, that the sum of the yields of labeling reaction at the both sites in state II is considerably higher than the yield in state IO. (Fig. 9B). Thus, our scheme may answer the first question as to why aurachin derivatives maintain high binding affinities to the mutated enzymes (G140A, G141V, and E144C) without interfering with electron transfer from the riboflavin to ubiquinone.
Next, we consider the second question. Why does the effect of competitors on the photoaffinity labeling by [ 125 I]PAD-1 and [ 125 I]PAD-2 change in opposite directions (suppression versus enhancement) depending on the concentrations of the enzyme and competitor? It should be emphasized that the same effect on labeling was also observed using cold ligands (PAD-1 and PAD-2) and hot ligands ([ 125 I]PAD-1 and [ 125 I]PAD-2) together and also that the enhancing effect of competitors manifested itself at much higher inhibitor concentrations than those required for complete inhibition of the enzyme activity. We tried using a number of different kinetic models to simulate the total amount of labeled Na ϩ -NQR under different experimental conditions. However, it became clear that this phenomenon cannot be explained unless our model includes two distinct inhibitor-bound states in which the yields of the labeling reaction are considerably different. Furthermore, the equilibrium between the two inhibitor-bound states must be controlled by the inhibitor concentration, meaning that there needs to be two binding sites for the inhibitor that have different affinities. An example of such a model is shown in Fig. 9C, where there are two inhibitor-bound states (IO and II), and the yield of labeling in state II (the sum of the yields at the both sites) is considerably higher than in IO (supposition ii). It would not be surprising if the reactivity of the aryl azido (-NϭN ϩ ϭN Ϫ ) groups in PAD-1 and PAD-2 (and their 125 I-labeled derivatives) changes with variations in the binding environment. In our scheme, the putative second molecule of inhibitor is provisionally shown at the interface of the NqrA and B subunits.
The scheme in Fig. 9C includes free inhibitor (I), free enzyme (state OO), and three inhibitor-enzyme complexes (states OI, IO, and II), which leads to the following equilibrium Equations 1-3, Equations 1-6 could then be numerically solved using Mathematica (Wolfram). For this analysis, we further assume that the inhibitor does not bind to the second site unless the first site is already occupied by a molecule of the inhibitor (i.e. state OI scarcely exists: K 1 , K 2 , Ͼ Ͼ K 3 ), as discussed below. Thus, for the current simulation we set K 1 , K 2 , and K 3 to 1000, 800, and 1 mM Ϫ1 , respectively. Based on these assumptions, the total amount of Na ϩ -NQR labeled by [ 125 I]PAD-2 and PAD-2 is a sigmoid function of the total inhibitor concentration. This relationship is shown for four different enzyme concentrations in Fig. 10A. Fig. 10B shows the relative amount of radioactivity incorporated into Na ϩ -NQR by [ 125 I]PAD-2 (10 nM) as a function of the concentration of added PAD-2, normalized so that the value at zero-added PAD-2 is one, as calculated from the data in Fig.  10A. Values below one indicate competitive suppression of labeling, whereas values above one indicate enhancement. It should be noted that our model (Fig. 9C) can account for the consecutive changes of the effects of competitor (from suppression to enhancement and back to suppression) as the concentration of the competitor increases relative to that of the enzyme. Considering that the extent of suppressive or enhancing effects vary depending on the parameters (K 1 , K 2 , and the reaction yields in the states IO and II) that are determined by the individual chemical structures of the inhibitor, the small difference in the effects of competitors between [ 125 I]PAD-1 and [ 125 I]PAD-2 can easily be accounted for (Figs. 5 versus 6).
To determine the binding site for the second molecule of inhibitor, we examined the Lys-C digestion pattern of the NqrB subunit labeled by [ 125 I]PAD-2 (10 nM) in the presence of aurachin D-42 (10 M). As shown in Fig. 11, the resulting digestion pattern was almost identical to that observed in the absence of aurachin D-42, but the labeling intensity at the ϳ8-kDa band (Arg-55-Lys-121) relative to other bands was significantly higher, suggesting that a second molecule of the inhibitor binds to this region. The structure around this region may change in the presence of the first molecule of inhibitor to allow the binding of the second molecule of inhibitor. The binding of the second molecule may, in turn, affect the bound state of the first molecule of inhibitor, resulting in an increase in the reaction yield in state II.
Although we cannot rule out other scenarios that might explain these apparently paradoxical results, our model gives a possible explanation for them. The model is also able to explain the effects of competitor molecules on the photoaffinity labeling of the mutants. The labeling yield for NqrB-G140A mutant was significantly higher even in the absence of a competitor compared with the wild-type, G141V, and E144C mutants (Fig.  8A) and was not enhanced in the presence of a competitor. This unique behavior of the NqrB-G140A mutant can also be reproduced in the model if we assume K 1 Ͼ Ͼ K 2 (e.g. 1000:100). It should be remembered that the inhibitor sensitivities are considerably different between G141V and G140A (or E144C) ( Table 1). Taken together, the mode of interaction between the ligands and Na ϩ -NQR may vary in a complicated way depending on the structural changes in the NqrB subunit (or interfacial regions between NqrB and NqrA) induced by the different mutations. Because of this complexity, it is difficult to provide a complete scenario at present, which can explain all the details of the results obtained in this study. Gly-140, Gly-141, and Glu-144 are located in a transmembrane helix II of the NqrB subunit and do not make a direct contact with other subunits (13). This study revealed that these residues do not participate directly in the binding sites of ubiquinone and inhibitor. As there is currently no information about structural changes brought about by these mutations, the roles of the residues in the regulation of structural changes of the NqrB subunit (or interfacial regions between NqrA and NqrB subunits) remain to be investigated.
In conclusion, this study demonstrates that the solvent-accessible cavity in the NqrA subunit, revealed by the recent X-ray crystallographic study and predicted to accommodate ubiquinone (13), does in fact include a site for ubiquinone binding. However, the position, size, and shape of this cavity in the crystallographic model may not be the ones maintained throughout actual enzyme turnover. Aurachin-type inhibitors (and presumably also korormicin) interfere indirectly with the electron transfer from the riboflavin cofactor to ubiquinone, presumably via blocking a conformational change involving the NqrB subunit by binding to the long N-terminal stretch of the subunit. The base part of this N-terminal stretch closest to the transmembrane helices of the subunit is able to accommodate the second molecule of inhibitor in high concentration conditions. NqrB-Gly-141, Gly-140, and Glu-144, located in transmembrane helix II, may confer resistance against aurachin (and korormicin), apparently by interfering with dynamic conformational changes in NqrB or at the interface between NqrA and NqrB.  The labeled NqrB subunit was isolated and digested by Lys-C according to the same procedures as those described in the Fig. 7, followed by the SDS-PAGE using a 16% Schägger-type SDS gel (16% C, 6% T containing 6.0 M urea). The autoradiogram "control" is the same as that used in Fig. 7A.

Materials
Ubiquinone-1 (Q 1 ) was a kind gift from Eisai (Tokyo, Japan). 3-N 3 -Q 2 was synthesized as described previously (28). Protein standards (Precision Plus Protein Standards) for SDS-PAGE were purchased from Bio-Rad. The Click-iT reaction buffer kit and TAMRA-azido (18 -20) were purchased from Life Technologies, Inc. [ 125 I]NaI was purchased from PerkinElmer Life Sciences. DDM and LDAO were purchased from Dojin (Kumamoto, Japan) and Sigma. Other reagents were all of analytical grade.

Isolation of korormicin
Korormicin was isolated from the marine bacterium Pseudoalteromonas strain J010 cells (a kind gift from Tillman Harder and Jan Tebben) according to the reported methods with some modifications (16,17). Cells were grown in fermenters in 30liter batches in marine broth (Difco) at 28°C with constant agitation (300 rpm) and aeration (20 liters/min). Cells were harvested in late stationary phase (48 h), washed with 50 mM Tris-HCl, pH 8, 300 mM NaCl, and frozen at Ϫ80°C until used.
The cell pellet of Pseudoalteromonas strain J010 (50 g) was extracted with ethanol for 2 h at room temperature; the extraction was repeated three times. The combined extracts were centrifuged at 2500 rpm for 10 min at 4°C, and then the supernatant was collected and concentrated under reduced pressure to afford the extract residue (3.25 g). After the residue (1.56 g) was partitioned with diethyl ether (3 ϫ 30 ml) and water (40 ml), the fraction extracted with diethyl ether was dried over anhydrous MgSO 4 and concentrated to afford the diethyl ether fraction (336 mg). The diethyl ether fraction was chromatographed on a silica gel column (15 ϫ 120 mm, Wako Gel C-200) and eluted with 5, 10, 30, 50, and 100% ethyl acetate/nhexane. Only the fraction eluted with 50% ethyl acetate/nhexane elicited the inhibition of the NADH-Q 1 oxidoreductase activity of Na ϩ -NQR. This 50% fraction (12 mg) was further purified by reversed phase HPLC (20 ϫ 250 mm, COSMOSIL 5C 18 -MS-II, 80% methanol, 0.1% formic acid, detected at 254 nm) with a flow rate of 5.0 ml/min at 30°C to provide korormicin as a colorless oil (5.6 mg, 12 mol, retention time is 54.2 min). The following spectral data completely corresponded to those reported previously (16,17)

Expression and purification of wild-type and mutants of Na ؉ -NQR
Site-directed mutants were obtained using the QuikChange Lightning mutagenesis kit (Agilent) as reported before (29). The forward primers used for the mutants in NqrA are as follows: NqrA-Y36A, GCTTGGCGAAGAGGCCGTTGGCAT-GCGTC; NqrA-G38V, GGCGAAGAGTACGTTGTCATGC-GTCCTACTATG.
The primers for the mutants in NqrB were reported before (29). Recombinant wild-type Na ϩ -NQR and mutants were grown in LB medium as reported before (8) in 30-liter fermenters. The expression of the nqr operon was induced by adding arabinose. Cells were harvested, washed, and broken in 50 mM Tris-HCl, pH 8, 300 mM NaCl, 10 mM MgCl 2 in the presence of DNase and protease inhibitors. Membranes were obtained by ultracentrifugation (100,000 ϫ g) and washed with 5 mM imidazole, 300 mM NaCl, 0.05% glycerol. Wild-type and mutant proteins were purified by affinity chromatography as reported before (29).

Measurement of the electron transfer activity of Na ؉ -NQR
Steady-state NADH-quinone oxidoreductase activity of purified Na ϩ -NQR was measured by following the reduction of quinone at 282 nm (⑀ ϭ 14.5 mM Ϫ1 cm Ϫ1 ) with a Shimadzu UV-3000 instrument at 30°C (30). The reaction medium (2.5 ml) contained 5% glycerol, 0.05% DDM, 1 mM EDTA, 100 mM NaCl, and 50 mM Tris-HCl, pH 8.0. The final protein concentration was set to 0.9 nM. The reaction was started by adding 100 M NADH after the incubation of the enzyme with inhibitor and quinone for 5 and 1 min, respectively. The kinetic parameters of quinone reduction were determined by fitting the experimental data to the Michaelis-Menten equation using Prism (version 6, GraphPad, La Jolla, CA).

Photoaffinity labeling of Na ؉ -NQR by PUQ-3
DDM-purified or LDAO-washed Na ϩ -NQR was suspended in the reaction buffer (100 mM NaCl, 5% glycerol, and 50 mM Tris-HCl, pH 8.0) containing 0.05% DDM or 0.05% LDAO, respectively. The Na ϩ -NQR solution (0.9 M, 30 -100 l) was incubated in a 1.5-ml Eppendorf tube with PUQ-3 (5-10 M) for 10 min at room temperature. Then, the mixture was irradiated with a long wavelength UV lamp (Black Ray model B-100A, UVP, Upland, CA) for 10 min on ice, positioned 10 cm from the light source. The labeled enzyme sample was denatured with 1% (w/v) SDS at 40°C for 1 h and conjugated with a fluorescent TAMRA-N 3 tag via Cu ϩ -catalyzed click chemistry using the Click-iT Protein reaction buffer kit (Life Technologies, Inc.) according to the manufacturer's instructions (19,20).
The TAMRA-attached proteins were separated on a Laemmli-type 15% SDS gel containing 6.0 M urea. The migration pattern of fluorescent protein (TAMRA-attached) was visualized using one of two available scanners: a model FLA-5100 (Fuji Film, Tokyo, Japan) or a Typhoon FLA9500 (GE Healthcare, Buckinghamshire, UK) bio-imaging analyzer, in both cases using a 532 nm light source and an LPG emission filter (575 nm). The proteins containing covalently bound FMN (NqrB and NqrC) were visualized using a 473 nm light source and an LPB emission filter (510 nm). MultiGauge (Fuji Film) software was used for data processing. ImageQuant (GE Healthcare) software was used for quantification of fluorescence.

Photoaffinity labeling of Na ؉ -NQR by [ 125 I]PAD-1 or [ 125 I]PAD-2
The Na ϩ -NQR solution (0.9 M, 30 l) was incubated with [ 125 I]PAD-1 or [ 125 I]PAD and photoirradiated with a UV lamp under the same conditions as those used for the labeling by PUQ-3. The reaction was terminated by addition of an appropriate volume of 4ϫ Laemmli's sample buffer, followed by the separation on a 15% Laemmli-type SDS gel (31) containing 6.0 M urea. The gels were stained with CBB, dried, exposed to an imaging plate (BAS-MS2040, Fuji Film) for 12-24 h, and visualized with the FLA-5100 Bio-Imaging analyzer. The radioactivity of each band was quantified from the digitized data using Multi-Gauge (Fuji Film) or directly from the gel slices using a ␥-counting system (COBLA II, Packard), as described elsewhere (32).

Proteomic analysis
To partially digest the NqrA subunit labeled by PUQ-3, the CBB-stained NqrA band was treated with V8-protease (Roche Applied Science, Penzberg, Germany) in a 15% Tris-EDTA mapping gel using according to the procedures described previously (32)(33)(34). The partial digests were characterized by mass spectrometry or N-terminal sequence analysis. For exhaustive digestion, the TAMRA-conjugated NqrA subunit was cleaved by CNBr in 70% aqueous formic acid at 37°C for 16 h (18). The digests were resolved on a Schägger-type SDS gel (16.5% T, 6% C containing 6.0 M urea, see Ref. 35).
For the analysis of the NqrB subunit labeled by [ 125 I]PAD-1 or [ 125 I]PAD-2, the NqrB subunit was partially isolated by SDS-PAGE and electroelution (19) in a buffer containing 10 mM Tris-HCl, pH 8.0, and 0.025% (w/v) SDS using a model 422 Electro-Eluter (Bio-Rad). The purified subunit was digested with lysyl-endopeptidase (Lys-C, Wako Pure Chemicals, Osaka, Japan) or endoprotease Asp-N (Roche Applied Science) in 50 mM Tris-HCl buffer containing 0.1% SDS or 50 mM NaP i buffer containing 0.01% SDS, respectively. These digests were separated on a Schägger-type SDS gel.
For the mass spectrometric analysis of proteins, CBB-stained protein bands were digested in gel with trypsin (Promega, Fitchburg, WI) in buffer containing 25 mM NH 4 HCO 3 at 37°C overnight. The tryptic digests were desalted with ZipTip (Merck-Millipore) and spotted onto the target plate using ␣cyano-4-hydroxycinnamic acid as a matrix (36). A mass spectrometric analysis was conducted using a Bruker Autoflex III Smartbeam MALDI-TOF/TOF instrument (Bruker Daltonics). The mass spectra obtained were analyzed according to previously published procedures (36). N-terminal amino acid residues were examined with a Procise 494 HT protein sequencing system (Applied Life Sciences, Foster City, CA) at the APRO Life Science Institute, Inc. (Tokushima, Japan).