C-terminal Periplasmic Domain of Escherichia coliQuinoprotein Glucose Dehydrogenase Transfers Electrons to Ubiquinone*

Membrane-bound quinoprotein glucose dehydrogenase (GDH) in Escherichia coli donates electrons directly to ubiquinone during the oxidation of d-glucose as a substrate, and these electrons are subsequently transferred to ubiquinol oxidase in the respiratory chain. To determine whether the specific ubiquinone-reacting site of GDH resides in the N-terminal transmembrane domain or in the large C-terminal periplasmic catalytic domain (cGDH), we constructed a fusion protein between the signal sequence of β-lactamase and cGDH. This truncated GDH was found to complement a GDH gene-disrupted strain in vivo. The signal sequence of the fused protein was shown to be cleaved off, and the remaining cGDH was shown to be recovered in the membrane fraction, suggesting that cGDH has a membrane-interacting site that is responsible for binding to membrane, like peripheral proteins. Kinetic analysis and reconstitution experiments revealed that cGDH has ubiquinone reductase activity nearly equivalent to that of the wild-type GDH. Thus, it is likely that the C-terminal periplasmic domain of GDH possesses a ubiquinone-reacting site and transfers electrons directly to ubiquinone.

Membrane-bound GDH 1 in Escherichia coli is a PQQ-containing quinoprotein that catalyzes a direct oxidation of Dglucose to D-gluconate in the periplasm and concomitantly transfers electrons to ubiquinol oxidase through ubiquinone in the respiratory chain (1)(2)(3). GDH is an 88-kDa monomeric protein with five transmembrane segments at the N-terminal portion (residues 1-140), which ensure a strong anchorage of the protein to the inner membrane (4,5). The remaining large C-terminal portion (residues 141-796) has a catalytic domain including PQQ- (6,7) and Ca 2ϩ or Mg 2ϩ -binding sites (8,9) that is located in the periplasmic side. A model structure of GDH based on the x-ray crystallographic structure of the ␣-subunit of MDH in Methylobacterium extorquens has been proposed (10), and the putative structure of the PQQ-binding catalytic site has been further confirmed and characterized by mutagenic analysis of several amino acid residues around PQQ (11)(12)(13)(14)(15).
The ubiquinone-reacting site in GDH has also been analyzed. Friedrich et al. (16) proposed that the ubiquinone-reacting site may be located at the N-terminal transmembrane domain of Acinetobacter calcoaceticus GDH in which Arg-91 and Asp-93 may be involved in interaction with ubiquinone. The topological model of the N-terminal transmembrane domain of E. coli GDH has shown that the corresponding amino acid residues, Arg-93 and Asp-95, are located near the membrane surface of the periplasmic side (5). Furthermore, using depth-dependent fluorescent ubiquinone analogues, Miyoshi et al. (17) demonstrated that the ubiquinone-reduction site of GDH is located close to the membrane surface rather than in the hydrophobic interior. X-ray crystallographic structures of cytochrome bo in E. coli (18,19) and cytochrome bc 1 complex (Q o and Q i centers) in bovine heart mitochondria have recently been determined (20,21), and it has been indicated that their ubiquinone-binding sites may be close to the membrane surface. Thus, it seems reasonable that the ubiquinone-reacting site of GDH is located near the membrane surface. However, in quinohemoprotein alcohol dehydrogenase of acetic acid bacteria despite the absence of a transmembrane domain, subunit II appears to be embedded in the cytoplasmic membrane and contain the ubiquinone-reduction site (22,23). In addition, in the case of E. coli GDH, the mutations in the possible ubiquinone-reacting sites, Arg-93 and Asp-95, in the transmembrane domain have been shown to have no effect on the ubiquinone reductase activity. 2 Thus, we cannot exclude the possibility that the Cterminal periplasmic domain of GDH, named cGDH, contains the ubiquinone-reacting site.
In efforts to clarify the role of the cGDH in reaction with ubiquinone, we constructed a fusion protein that was composed of the ␤-lactamase signal sequence (residues 1-25) and the cGDH (residues 142-796). The signal sequence was found to export the cGDH through the inner membrane to the periplasm, and the truncated GDH successfully complemented a GDH gene-disrupted strain in vivo and in vitro. In this paper, we also discuss the evolutionary relationship between GDH in E. coli and SLDH in Gluconobacter suboxydans or MDH in M. extorquens.

EXPERIMENTAL PROCEDURES
Materials-Restriction enzymes, T 4 DNA ligase, and Taq DNA polymerase were purchased from Takara Shuzo (Kyoto, Japan) and New England Biolabs (Hertfordshire, United Kingdom). Oligonucleotide primers were purchased from Sawady Technology (Tokyo, Japan). Q-2 * This work was supported in part by a grant-in-aid from the Ministry of Education, Science, and Culture of Japan. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Construction of pBLAGCD Encoding a Bla-cGDH Fusion Protein-Conventional recombinant DNA techniques were applied (27). To produce a Bla-cGDH fusion protein, the N-terminal transmembrane domain of GDH was substituted with the ␤-lactamase signal sequence as shown in Fig. 1A. A DNA fragment encoding the ␤-lactamase signal sequence (residues 1-25) was amplified by PCR with a set of primers, lacZ3 (5Ј-TTCTGGTGCCGGAAACCAGGCAAAG-3Ј) and bla-sg1 (5Ј-A-TAGGATCCGGGTGAGCAAAAACAGG-3Ј) containing a BamHI site and pMC1396 DNA as a template. A DNA fragment encoding the Cterminal periplasmic domain of GDH (residues 142-796) was amplified by PCR with a set of primers gcd-sg2 (5Ј-CTTCTGGATCCGCAGGAG-ATCAA-3Ј) containing a BamHI site and PQ-2 (5Ј-AAGCTTGCATGC-CTGCAGGTC-3Ј) containing a PstI site and pUCGCD1 DNA as a template. PCR of 30 cycles was carried out, each of which consisted of denaturation at 94°C for 1 min, annealing at 60°C for 2 min, and extension at 72°C for 5 min. Both PCR products were recovered by ethanol precipitation after phenol extraction. The PCR product for the ␤-lactamase signal sequence was digested with EcoRI and BamHI, and the C-terminal periplasmic domain of GDH was digested with BamHI and PstI. Both digested products were recovered from 6% polyacrylamide gel after electrophoresis. Both fragments were then inserted into the EcoRI and PstI sites of the vector pTTQ18, generating pBLAGCD. The insertion was confirmed by restriction mapping and by DNA sequencing (28) using a thermo sequenase cycle sequencing kit (Amersham Biosciences, Inc.).
Complementation Test of cGDH to YU423 Strain-For the complementation test in liquid medium, a single colony of PPA322-harboring pTTQ18, YU423-harboring pBLAGCD, or YU423-harboring pTTQ18 was inoculated into 2 ml of minimal medium supplemented with 0.4% glycerol and incubated for 48 h at 30°C while being shaken. The cells were diluted to a turbidity corresponding to an optical density at 600 nm (A 600 ) of 1.0. 100 l of the diluted preculture was then inoculated into 50 ml of minimal medium supplemented with 0.2% glucose, 200 nM PQQ, and 0.1 mM IPTG and incubated at 30°C. Cell growth was followed by monitoring the turbidity at A 600 . For the complementation test on agar plates, a single colony of these three strains was streaked first on minimal agar plates supplemented with 0.4% glycerol and incubated for 48 h at 30°C. Single colonies of each strain grown on the plates were then streaked on the same agar plate of minimal medium supplemented with 0.2% glucose, 200 nM PQQ, and 0.1 mM IPTG and incubated for 60 h at 30°C. Cell growth of these strains was then compared.
Preparation of Membrane and Periplasmic Fractions-For preparation of the membrane fraction, cells grown as described in the following enzyme purification procedure were harvested by centrifugation and quickly chilled on ice. All of the subsequent steps were carried out at 4°C. The cells were washed twice with 0.85% NaCl and suspended in 10 mM KPB, pH 7.0. The membrane fraction was then prepared as described previously (5) and homogenized to a final protein concentration of 10 mg/ml in the same buffer containing 1 mM MgCl 2 .
For preparation of the periplasmic fraction, cells grown as described in the following enzyme purification procedure were washed with 10 mM Tris-HCl, pH 7.0, containing 30 mM NaCl. The cells were resuspended in 33 mM Tris-HCl, pH 7.0, containing 0.1 mM EDTA and 20% sucrose and gently stirred at room temperature for 10 min. After centrifugation, the pellet was resuspended in 20 ml of ice-chilled 0.5 mM MgCl 2 and gently stirred for 10 min in an ice bath. The suspension was then centrifuged at 16,000 ϫ g for 10 min to separate the supernatant and precipitate. The latter was resuspended in 10 mM KPB, pH 7.0, and treated by a French pressure cell press. In both samples, the activity of periplasmic ␤-lactamase as a control was measured as described previ-ously (29), and 80 and 20% activities of ␤-lactamase were recovered in the supernatant and precipitate, respectively. Thus, the supernatant was used as the periplasmic fraction.
Purification of Wild-type GDH and cGDH-YU423 harboring the wild-type GDH-encoded plasmid pUCGCD1 was grown in LB medium for 8 h at 30°C. The wild-type GDH was then purified from the membrane fraction according to the procedure described previously (5).
YU423 harboring pBLAGCD was grown at 25°C in LB medium supplemented with 0.2% glucose and 200 nM PQQ. IPTG was added at the final concentration of 0.1 mM into the culture when A 600 had reached ϳ0.3 and the cells were grown for an additional 4 h. All purification steps were then carried out at 4°C. The membrane fraction prepared as described above was washed with 1 M KCl by stirring for 30 min and then centrifuged at 86,000 ϫ g for 90 min. The pellet was homogenized to a final protein concentration of 10 mg/ml in 10 mM KPB, pH 7.0, containing 1 mM MgCl 2 and stirred in the presence of 0.3% Triton X-100 (w/v), 100 nM PQQ, and 100 mM KCl for 30 min for solubilization of the enzyme from the membrane fraction. The suspension was centrifuged at 86,000 ϫ g for 90 min, and the supernatant obtained was dialyzed against the same buffer containing 0.1% Triton X-100. The dialyzate was applied onto a DEAE-Toyopearl column (1-ml bed volume/ϳ10 mg of protein) equilibrated with 10 mM KPB, pH 7.0, containing 1% Triton X-100. The column was washed with 10 bed volumes of the same buffer and successively with 10 bed volumes of 10 mM KPB, pH 7.0, containing 0.1% Triton X-100 and 40 mM KCl. The enzyme was eluted by a linear gradient composed of 4 bed volumes of the same buffer and 4 bed volumes of 10 mM KPB, pH 7.0, containing 0.1% Triton X-100 and 120 mM KCl. Active fractions that came out at approximately 90 mM KCl were pooled and dialyzed against 1 mM KPB, pH 6.8, containing 0.1% Triton X-100. The dialyzate was applied onto a ceramic hydroxyapatite column (1-ml bed volume/ϳ1 mg of protein) equilibrated with 1 mM KPB, pH 6.8, containing 0.1% Triton X-100. The column was washed with 10 bed volumes of 2 mM KPB, pH 6.8, containing 0.1% Triton X-100. The enzyme was eluted by 10 bed volumes of 5 mM KPB, pH 6.8, containing 0.1% Triton X-100. Active fractions were concentrated by a DEAE-Toyopearl column (1-ml bed volume/ϳ10 mg of protein) in which the enzyme absorbed was eluted with a small volume of 10 mM KPB, pH 7.0, containing 0.1% Triton X-100 and 150 mM KCl. The concentrated enzyme was used as the purified cGDH, which was found to have a homogeniety of ϳ50% as judged from the results of SDS-polyacrylamide gel electrophoresis.
Sequencing of cGDH from the N Terminus-Purified cGDH was electrophoresed on SDS-12% polyacrylamide gel and transferred onto a polyvinylidene difluoride membrane. After staining with 0.1% Ponceau S in 3% trichloroacetic acid and 3% sulfosalicylic acid for 1 min, the membrane was washed subsequently with 50% methanol and H 2 O. After it had dried, the stained band was excised from the membrane, and the sample was analyzed on an automatic gas phase sequenator (PPSQ-21A, Shimadzu).
Treatment of Membrane Fraction with a High Salt or Mild Base-To examine the binding characteristics of cGDH to membrane, treatments with a high salt or mild base were carried out. KCl at a final concentration of 1 M or NaOH at a final concentration of 0.025, 0.05, or 0.1 M was added into the suspension of membrane fraction (10 mg/ml) prepared as described above and stirred gently for 30 min at 4°C. The suspension was then centrifuged at 86,000 ϫ g for 90 min to separate the supernatant and precipitate. The latter was resuspended in 10 mM KPB, pH 7.0, containing 5 mM MgCl 2 . Both samples were then subjected to the measurement of PMS reductase activity and Western blot analysis.
Measurement of Protein and Enzyme Activity-Expression of the wild-type GDH or cGDH was examined by SDS-7% polyacrylamide gel electrophoresis followed by Western blotting using a polyclonal antibody raised against E. coli wild-type GDH as described previously (30). Protein content was determined according to the Dulley and Grieve method (31) using bovine serum albumin as a standard.
Holo-enzyme formation was performed by incubating the membrane fraction, periplasmic fraction, or purified enzyme in 10 mM MOPS, pH 7.0, containing 30 M PQQ and 1 mM MgCl 2 for 30 min at 25°C. Using the holo-enzyme thus prepared, the following enzyme activities were measured. PMS reductase activity was measured spectrophotometrically (U-2000A, Hitachi) with PMS and dichloroindophenol as an electron mediator and acceptor, respectively, as described previously (13,32). Glucose oxidase activity of membrane fractions (coupling ability of GDH to an electron transport chain in the membrane) was determined using an oxygen electrode as described previously (13,32). Q-2 reductase activity was measured spectrophotometrically as described previously (13). One unit of PMS reductase or Q-2 reductase activity is defined as 1 mol of dichloroindophenol or Q-2, respectively, reduced/ min, both of which correspond to 1 mol of glucose oxidized/min. One unit of glucose oxidase activity is defined as 1 microatom of oxygen consumed/min, which is also equivalent to 1 mol of glucose oxidized/ min. The K m value was estimated on the basis of a Lineweaver-Burk plot.
Reconstitution of Purified Wild-type GDH or cGDH with the YU423 Membrane Fraction-The YU423 membrane fraction was suspended in 10 mM KPB, pH 7.0, at a final protein concentration of ϳ30 mg/ml. For the reconstitution experiments, Triton X-100 in the purified wild-type GDH and cGDH enzyme solution was replaced with 1% ␤-octyl-Dglucopyranoside using a DEAE-Toyopearl column as described previously (33). The membrane suspension (4 mg of protein) was mixed with the purified wild-type GDH (15 units/50 g) or cGDH (15 units/100 g) at a final concentration of 1% ␤-octyl-D-glucopyranoside in a final volume of 0.5 ml of 10 mM KPB, pH 7.0. The mixture was then dialyzed three times against 500 ml of 10 mM KPB, pH 7.0, for 4 h each. The dialyzed sample was centrifuged (RP100 AT rotor, Hitachi) at 128,000 ϫ g for 90 min. The resultant precipitate was suspended in 0.5 ml of 10 mM MOPS, pH 7.0, at a protein concentration of ϳ8 mg/ml and subjected to holo-enzyme formation. This sample was used for PMS reductase and glucose oxidase assay.
Homology Search-A homology search was performed using FASTA and BLAST in the GenBank data base. The comparison and alignment of amino acid sequences, construction of a hydrophobicity plot by the Kyte and Doolittle procedure (34), and amphiphilic segment prediction by the Edmundson helical wheel plot procedure (35) were conducted using GENETYX (Software Development, Tokyo, Japan).

Construction and Localization of a Bla-cGDH Fusion
Protein-To explore the function of the N-terminal transmembrane domain or the C-terminal periplasmic domain of GDH in transferring electrons to ubiquinone, a Bla-cGDH fusion protein was constructed as shown in Fig. 1. Because ␤-lactamase is a periplasmic protein, we speculated that its signal sequence might be able to export cGDH through the inner membrane to the periplasm. If the signal sequence works successfully as expected, cGDH could be recovered in the periplasmic fraction or in the membrane fraction. In the latter case, cGDH would bind to the membrane through the uncleaved signal sequence or through its specific segment, even though the signal se-quence is cleaved off. To detect the cellular localization of Bla-cGDH, the membrane and periplasmic fractions were prepared and subjected to Western blotting using a GDH antibody ( Fig. 2A). Protein bands were detected only in the membrane fraction (lane 4), and the main and largest band was estimated to be approximately 75 kDa, which agrees with the size estimated from the nucleotide sequence. Some other bands, which seemed to be degradation products of cGDH, were also detected. These results support the latter possibility as described above.
Purification of cGDH and Amino Acid Sequence of Its N Terminus-We purified cGDH from the membrane fraction of YU423 harboring pBLAGCD grown in LB medium as shown in Table I. The membrane fraction was treated with 1 M KCl to obtain the washed membrane in which all the activity of cGDH was recovered. The enzyme was solubilized from the washed membrane with Triton X-100 and successively purified by two column chromatographies. The enzyme was purified 1500-fold from the membrane fraction with an overall recovery of 20%, and the homogeneity was approximately 50% (Fig. 2B). Its molecular size was estimated to be approximately 75 kDa, which is consistent with that of the largest band observed in the membrane fraction by Western blotting (Fig. 2A, lane 4). Although two other minor bands of approximately 45 kDa still remained after ceramic hydroxyapatite column chromatography, we used this sample for enzymatic analysis or reconstitution experiments.
Because cGDH was purified from the membrane fraction, it was unclear whether the signal sequence was cleaved from cGDH or not. The N-terminal amino acid residues were thus sequenced with the 75-kDa protein band excised from the blotted membrane. The sequence of the first nine amino acids obtained was His-Pro-Asp-Pro-Gln-Glu-Ile-Asn-Gly, which fully corresponds to the sequence of the fusion protein shortly after the ␤-lactamase signal cleavage site as shown in Fig. 1B. Therefore, it is likely that the ␤-lactamase signal sequence of the fusion protein is proteolytically cleaved during secretion, and that the cleaved cGDH binds to the membrane without the signal sequence.
Characteristics of cGDH-To characterize cGDH and compare it with the wild-type GDH, kinetic analysis was performed with membrane fractions and both purified enzymes (Tables II  and III). Glucose oxidase activity in the membrane fraction reflects the ability of the intermolecular electron transfer from cGDH to membrane ubiquinone and finally to terminal oxidase. Surprisingly, we found that the membrane fraction prepared from YU423-harboring pBLAGCD possessed significant glucose oxidase activity as well as PMS reductase activity. On one hand, these relative activities are comparable to those of PPA322 containing the wild-type GDH gene on the genome and harboring the vector plasmid pTTQ18, which are calculated based on their relative GDH contents in the membrane fraction estimated by Western blot analysis ( Fig. 2A). On the other hand, no detectable activity was observed in a GDH genedisrupted strain, YU423, harboring pTTQ18. Therefore, it is likely that cGDH is able to transfer electrons to ubiquinone. Notably, cGDH contains a catalytic domain including PQQand Ca 2ϩ or Mg 2ϩ -binding sites (6 -9). We thus compared PMS reductase and Q-2 reductase activities in purified cGDH and wild-type GDH (Table III). The purified cGDH was found to have almost equivalent activities of both reductases to those of the wild-type GDH. This outcome is because the purity of cGDH and that of wild-type GDH was 50% (Fig. 2B) and 95%, respectively. On the basis of these activities after purification, approximately half of cGDH was assumed to be inactive in membrane fraction, which was removed during purification. This is because the relative PMS reductase and glucose oxidase activities of cGDH in membrane fraction was 50 and 63%, respectively, compared with those of wild-type GDH (Table II).
To further examine whether the deletion of the N-terminal hydrophobic domain had any effect on the affinity for PQQ, glucose, or Q-2, the K m values of cGDH were compared with those of wild-type GDH. The K m values for PQQ and glucose of cGDH in the membrane fraction were found to be nearly the same as those estimated in wild-type GDH (Table II). The K m values for PQQ, glucose, and Q-2 of the purified cGDH were also found to be almost equivalent to those of the purified wild-type GDH (Table III). Thus, the deletion of the N-terminal hydrophobic domain appears to have no significant effect on the affinity for PQQ, glucose, or Q-2, and cGDH seems to possess the ubiquinone-reduction site in its sequence. Furthermore, we examined the substrate specificity of the purified  2. Cellular localization of cGDH (A) and SDS-12% polyacrylamide gel electrophoresis of purified cGDH sample (B). A, membrane fractions, purified enzymes, and periplasmic fraction were prepared and subjected to SDS-7% polyacrylamide gel electrophoresis and transferred to polyvinylidene difluoride membrane. The same membrane fractions and purified enzymes were used for measuring enzyme activities in Tables II and III. Both cGDH and wild-type GDH were visualized using a polyclonal antibody against the E. coli wild-type GDH. The relative amount of GDH proteins was densitometrically estimated by using Bio-Rad molecular imager.     The relative amount of GDH protein in the membrane fractions was estimated from the results of the Western blot shown in Fig. 2A and expressed as a percentage of that from the wild type.
c Values of PMS reductase and glucose oxidase activities were corrected by the relative amount of GDH protein in the membrane fractions. Values in parentheses are percentages of activity of the wild type. cGDH for various substrates, such as D-fucose, D-galactose, D-xylose, D-mannose, D-fructose, and maltose (data not shown). cGDH showed a wide range of substrate specificity similar to that of wild-type GDH, and the relative PMS reductase activities of cGDH with different substrates were also nearly the same as those measured in wild-type GDH.
cGDH Can Complement a GDH Gene-disrupted Strain in Vivo-To examine whether cGDH is also functional in vivo or not, pBLAGCD or its control vector was introduced into a GDH gene-disrupted strain, YU423, or its isogenic parent strain, PPA322, and the growth of these transformants was tested in minimal medium or on minimal agar plates containing 0.2% glucose as a carbon source. YU423 harboring pBLAGCD showed a nearly similar growth curve (Fig. 3A) and similar growth in colony size (Fig. 3B) to those of PPA322-harboring pTTQ18. On the other hand, YU423 harboring pTTQ18 showed no growth under both conditions. These results suggest that cGDH can complement a GDH gene-disrupted strain in vivo. Considering these results and the fact that cGDH showed significant glucose oxidase activity in the membrane fraction (Table II), it seems that electron transfer reactions occur from glucose to ubiquinol oxidase through cGDH and ubiquinone to form membrane potentials for cell growth in vivo.
Regeneration of Glucose Oxidase Activity in the YU423 Membrane with Purified cGDH-To examine whether cGDH is able to interact with the cytoplasmic membrane and whether it can donate electrons directly to the intrinsic ubiquinone, purified cGDH or wild-type GDH was mixed with the YU423 membrane fraction in the presence of ␤-octyl-D-glucopyranoside. After removing the detergent from the samples by dialysis, membranes were collected by ultracentrifugation. As seen in Table IV, 55 and 60% activities of PMS reductase were recovered in cGDHreconstituted and wild-type GDH-reconstituted membranes, respectively. Moreover, PMS reductase and glucose oxidase activities in the cGDH-reconstituted membrane were found to be comparable to those in the wild-type GDH-reconstituted membrane. These results suggest that cGDH interacts with the cytoplasmic membrane similar to the wild-type GDH and is able to donate electrons directly to ubiquinone.

DISCUSSION
Ubiquinone-reacting Site of GDH-The following findings obtained in this study clearly indicate that cGDH is bound to the cytoplasmic membrane. First, the signal sequence of ␤-lactamase was found to be absent from purified cGDH. Second, cGDH was recovered in the membrane fraction ( Fig. 2A). Third, Triton X-100 was required for solubilization of the protein from the membrane fraction. Finally, cGDH is functionally reconstituted into the membrane fraction by the octylglucoside dialysis method. From these findings, it is believed that cGDH may have some amphiphilic segment responsible for binding to the membrane. The binding mode appears to be similar to those of peripheral proteins because no membrane-spanning segment exists in cGDH. GDH has such transmembrane segments in the N-terminal portion but not in the C-terminal portion (5), i.e. cGDH. This idea of cGDH being a peripheral protein was supported by results of experiments with a mild base. Treatment with 0.025 M NaOH caused a release of 60 -70% of cGDH protein from the membrane fraction, whereas the same treatment could not release the wild-type GDH protein.
cGDH was found to show glucose oxidase activity equivalent to that of wild-type GDH in membrane fractions (Table II) as well as in reconstituted membranes (Table IV). Moreover, purified cGDH showed a significant Q-2 reductase activity, and the removal of the N-terminal hydrophobic portion had no influence on the affinity for Q-2. Judging from these findings, it seems that the ubiquinone-reacting site resides in the C-terminal periplasmic domain of GDH and that the site is close to the membrane surface, as supported by the evidence provided by Miyoshi et al. (17). Likewise, alcohol dehydrogenase of acetic acid bacteria may have a ubiquinone-reacting site in the peripheral cytochrome subunit (22). Originally, the ubiquinonereacting site was proposed to be located in the N-terminal hydrophobic domain, especially Arg-91 and Asp-93, of A. calcoaceticus GDH based on the sequence homology to that of mitochondrial NADH (16). However, mutants R93A, R93D, D95A, and D95N of Arg-93 and Asp-95 in E. coli GDH (5) corresponding to those in A. calcoaceticus GDH showed no significant effect on PMS reductase and glucose oxidase activities and also on affinity for Q-2. 3 Therefore, the possible involvement of Arg-93 and Asp-95 in interaction with ubiquinone was excluded. Thus, the ubiquinone-reacting site of GDH 3   seems to be present in the C-terminal periplasmic domain but not in the N-terminal transmembrane domain. Possible Evolutionary Relationship between GDH and SLDH or MDH-The C-terminal periplasmic domain of E. coli GDH shares a sequence similarity with that of the ␣-subunit of M. extorquens MDH, and a model structure of the domain equivalent to cGDH has been proposed on the basis of the x-ray crystallographic structure of the MDH ␣-subunit (10). By the homology search, a significant sequence similarity was also found between E. coli GDH and G. suboxydans SLDH (Fig. 4). SLDH is a membrane-bound quinoprotein that was isolated as an 80-kDa protein. 4 However, Miyazaki et al. 5 found that SLDH is encoded by two genes, small and large ORFs. The small ORF encodes a small hydrophobic protein that appears to have four transmembrane segments, whereas the large ORF encodes a large hydrophilic protein having homology to quinoproteins. The first four N-terminal transmembrane segments (residues 1-120) of GDH showed a 33% sequence identity to and a similar pattern in a hydropathy plot to the small protein ( Fig. 4 and data not shown). The C-terminal periplasmic domain of GDH (residues 170 -796) showed a 35% sequence identity to the large protein (residues 230 -867) and also a 22% identity to the ␣-subunit of MDH (residues 1-626). Moreover, there is a putative common segment between cGDH (495-575) and SLDH (525-607), which is absent in MDH, that transfers electrons to cytochrome C L but not to ubiquinone (36). The common domain consists of 80 amino acid residues including an amphiphilic segment. Therefore, this common domain is possibly related to ubiquinone binding because SLDH also has Q-2 reductase activity. 6 These sequence homologies and differences allow us to consider the evolutionary relationship among these proteins. We propose that the MDH ␣-subunit is significantly more similar to the common evolutionary origin than is GDH or SLDH, and that the common primordial protein had acquired a ubiquinone-binding site and a subunit containing four transmembrane segments as in SLDH, which had then fused to become a single protein like GDH.
At present, with the exception of the likelihood that the N-terminal transmembrane domain ensures a strong anchorage of GDH to the inner membrane since it is likely that the large C-terminal periplasmic domain possesses the catalytic site and ubiquinone-reacting site, which thus donates electrons directly to membrane ubiquinone, the role of the N-terminal domain in the electron transfer from glucose to ubiquinone is not clear. Considering the existence of multiple transmembrane segments in GDH, the hydrophobic domain may interact with the C-terminal domain and therefore somehow control its activity. FIG. 4. Schematic representation of amino acid sequence homology between GDH and SLDH or MDH. GDH is a single protein (796 amino acids) with five transmembrane segments (dotted boxes) at its N terminus. SLDH (GenBank TM accession number E32232) seems to be encoded by two ORFs; the small one encodes a protein (120 amino acids) containing four putative transmembrane segments (dotted boxes), and the large one encodes a protein (867 amino acids) containing a signal sequence (SS). The MDH ␣-subunit consists of 626 amino acids including a signal sequence. Stripped boxes in the large protein of SLDH and in cGDH represent segments including putative membranebinding and ubiquinone-reacting sites. Numbers indicated are the amino acid residues. Percentages show similarities between the small protein of SLDH and the first four N-terminal transmembrane segments of GDH between the C-terminal domain (residues 230 -867) of the large protein of SLDH and the C-terminal domain (residues 170 -796) of GDH and between the C-terminal domain (residues 121-495 and 575-796) of GDH and the entire MDH ␣-subunit.