Proton-translocating Nicotinamide Nucleotide Transhydrogenase RECONSTITUTION OF THE EXTRAMEMBRANOUS NUCLEOTIDE-BINDING DOMAINS*

The nicotinamide nucleotide transhydrogenase of bo- vine mitochondria is a homodimer of monomer M r (cid:53) 109,065. The monomer is composed of three domains, an NH 2 -terminal 430-residue-long hydrophilic domain I that binds NAD(H), a central 400-residue-long hydrophobic domain II that is largely membrane intercalated and carries the enzyme’s proton channel, and a COOH-ter-minal 200-residue-long hydrophilic domain III that binds NADP(H). Domains I and III protrude into the mitochondrial matrix, where they presumably come to- gether to form the enzyme’s catalytic site. The two-sub-unit transhydrogenase of Escherichia coli and the three- subunit transhydrogenase of Rhodospirillum rubrum have each the same overall tridomain hydropathy pro- file as the bovine enzyme. Domain I of the R. rubrum enzyme (the (cid:97) 1 subunit) is water soluble and easily removed from the chromatophore membranes. We have isolated domain I of the bovine transhydrogenase after controlled trypsinolysis of the purified enzyme and have expressed in E. coli and purified therefrom domain III of this enzyme. This paper shows that an active bidomain transhydrogenase lacking domain II can be reconstituted by the combination of purified bovine domains I plus III or R. rubrum domain I plus bovine domain III. The proton-translocating nicotinamide nucleotide transhydrogenases of mammalian mitochondria and bacteria are integral membrane proteins that catalyze the reaction shown in Equation 1,

The nicotinamide nucleotide transhydrogenase of bovine mitochondria is a homodimer of monomer M r ‫؍‬ 109,065. The monomer is composed of three domains, an NH 2 -terminal 430-residue-long hydrophilic domain I that binds NAD(H), a central 400-residue-long hydrophobic domain II that is largely membrane intercalated and carries the enzyme's proton channel, and a COOH-terminal 200-residue-long hydrophilic domain III that binds NADP(H). Domains I and III protrude into the mitochondrial matrix, where they presumably come together to form the enzyme's catalytic site. The two-subunit transhydrogenase of Escherichia coli and the threesubunit transhydrogenase of Rhodospirillum rubrum have each the same overall tridomain hydropathy profile as the bovine enzyme. Domain I of the R. rubrum enzyme (the ␣1 subunit) is water soluble and easily removed from the chromatophore membranes. We have isolated domain I of the bovine transhydrogenase after controlled trypsinolysis of the purified enzyme and have expressed in E. coli and purified therefrom domain III of this enzyme. This paper shows that an active bidomain transhydrogenase lacking domain II can be reconstituted by the combination of purified bovine domains I plus III or R. rubrum domain I plus bovine domain III.
The proton-translocating nicotinamide nucleotide transhydrogenases of mammalian mitochondria and bacteria are integral membrane proteins that catalyze the reaction shown in Equation 1, NADH ϩ NADP ϩ nH out ϩ º NAD ϩ NADPH ϩ nH in ϩ (Eq. 1) where n has been determined to be close to unity (Fisher and Earle, 1982;Rydström et al., 1987). The bovine mitochondrial enzyme is a homodimer of monomer M r ϭ 109,065 (Yamaguchi et al., 1988;Hatefi and Yamaguchi, 1992). The Escherichia coli and the Rhodobacter capsulatus transhydrogenases have two unlike subunits each, ␣ with M r ϭ ϳ54,000 and ␤ with M r ϭ ϳ49,000 (Clarke et al., 1986;Lever et al., 1991), and the former has been shown to be an ␣ 2 ␤ 2 heterotetramer (Hou et al., 1990). The transhydrogenase of Rhodospirillum rubrum has three unlike subunits, ␣1, ␣2, and ␤, with respective molecular masses of 40.3, 14.9, and 47.8 kDa (Yamaguchi and Hatefi, 1994;Williams et al., 1994). Subunit ␣1 is water soluble and easily removed from the R. rubrum chromatophore mem-branes; the other two subunits are integral membrane proteins. The bovine transhydrogenase monomer is composed of three domains, an NH 2 -terminal 430-residue-long hydrophilic domain I that binds NAD(H), a central 400-residue-long hydrophobic domain II that is largely membrane-intercalated and bears the enzyme's proton channel, and a COOH-terminal 200residue-long hydrophilic domain III that binds NADP(H). In mitochondria, the two nucleotide-binding domains I and III protrude into the matrix (Yamaguchi and Hatefi, 1991), where they presumably come together to form the enzyme's catalytic site in which direct hydride ion transfer takes place between the 4-A position of NAD(H) and the 4-B position of NADP(H). Together, the two subunits of the E. coli or the three subunits of the R. rubrum enzyme also display the same tridomain hydropathy profile, with similar nucleotide binding characteristics in domains I and III (Clarke et al., 1986;Hatefi, 1993, 1994;Diggle et al., 1995). The purpose of this communication is to show that unanchored to the membrane-intercalated domain II, the isolated, soluble domains I and III readily interact to reconstitute a structure capable of catalyzing transhydrogenation between NAD(H) and NADP(H).

EXPERIMENTAL PROCEDURES
Materials-NADH, NADPH, and NADP were from Calbiochem. AcPyAD, 1 AcPyADP, Reactive Green 19-agarose, palmitoyl coenzyme A, FSBA, and NEM were from Sigma. DEAE-Bio-gel agarose and Macro-Prep High Q Support were from Bio-Rad. Expression vector pET-16b and its host strain pLysS were from Novagen. Restriction enzymes were from Stratagene and New England Biolabs. NAD-agarose (Type I) was prepared by the method of Mosbach et al. (1972).
Assays of Transhydrogenase Activities-Transhydrogenations from NADPH to AcPyAD, from NADH to AcPyADP, and from NADH to AcPyAD catalyzed by bovine transhydrogenase domains I plus III or by R. rubrum domain I plus bovine domain III were assayed spectrophotometrically at 375 nm at 37°C in a reaction mixture containing 100 mM sodium phosphate, pH 6.5, and 0.2 mM of each nucleotide indicated. Reactions were started by the addition of enzyme. An extinction coefficient of 6.1 mM Ϫ1 cm Ϫ1 was used to calculate rates. During purification of R. rubrum ␣1 subunit (domain I), activity assay was carried out by adding aliquots of fractions to a reaction mixture containing chromatophore membranes depleted of ␣1 subunit (Fisher and Guillory, 1971). The reaction mixture contained 125 mM sucrose, 44 mM Tris/HCl, pH 8.0, 2.7 mM MgCl 2 , 0.2 mM NADPH, 0.2 mM AcPyAD, a fixed amount of chromatophores, and an appropriate amount of ␣1 subunit.
Preparation of the ␣1 Subunit of R. rubrum-R. rubrum cells were grown photosynthetically (Ormerod et al., 1961), and chromatophores were prepared according to Fisher and Guillory (1971). The cell-free extract obtained after centrifugation of chromatophores and containing the soluble ␣1 subunit of the transhydrogenase was treated according to Konings and Guillory (1973). The protein fraction precipitating between 0.45 and 0.80 saturation with ammonium sulfate was suspended in a * This work was supported by United States Public Health Service Grant GM24887. This is publication 9538-MEM from the Scripps Research Institute. 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.
‡ To whom correspondence should be addressed. Tel.: 619-554-8092; Fax: 619-554-6838. buffer containing 10 mM Tris/HCl, pH 8.0, 1% sucrose, 1 mM dithiothreitol, 5 mM (NH 4 ) 2 SO 4 , and 1 mM phenylmethylsulfonyl fluoride and dialyzed overnight against the same buffer. The crude ␣1 was divided into two portions, and each was applied to a column (1.5 ϫ 14 cm) of Macro-Prep High Q Support equilibrated with a buffer consisting of 10 mM Tris/HCl, pH 8.0, 10 mM (NH 4 ) 2 SO 4 , 1 mM dithiothreitol, and 0.5 mM phenylmethylsulfonyl fluoride. The ␣1 subunit was eluted with the same buffer, and fractions containing ␣1 activity were pooled and concentrated 3-fold. Then the sample was applied to a column (1.5 ϫ 14 cm) of Reactive Green 19-agarose equilibrated with a buffer containing 10 mM MES/NaOH, pH 6.5, 10 mM (NH 4 ) 2 SO 4 , 1 mM dithiothreitol, and 0.5 mM phenylmethylsulfonyl fluoride (buffer A). The ␣1 subunit was eluted with a linear gradient formed with 100 ml of buffer A and 100 ml of 500 mM (NH 4 ) 2 SO 4 in buffer A. Fractions showing ␣1 activity were pooled and concentrated to about 7 ml. The sample was dialyzed against 1 liter of 10 mM Tris/HCl, pH 7.4, containing 5 mM (NH 4 ) 2 SO 4 and 1 mM dithiothreitol and applied to a column (2.5 ϫ 10 cm) of NAD-agarose equilibrated with the same buffer. After washing the column with 100 ml of the same buffer, ␣1 was eluted with the same buffer containing 0.5 mM NADH. The active fractions were combined and concentrated by Centricon-30 concentrator. By this procedure the ␣1 subunit was purified 1900-fold from the cell-free extract and was homogeneous when examined by SDS-polyacrylamide gel electrophoresis.
Expression and Purification of the Bovine Domain III-For expression of the bovine domain III (Met 860 -Lys 1043 ; molecular mass ϭ 20,053 Daltons), the NcoI-BamHI 1.5-kilobase pair DNA fragment was cut out from TH36 -1 (Yamaguchi et al., 1988) and inserted into the NcoI-BamHI site of expression vector pET-16b. E. coli strain pLysS was transformed with the constructed plasmid. The transformed cells were cultured in 500 ml of LB medium containing ampicillin (100 g/ml). When the absorbance of bacterial culture at 600 nm reached 0.6, isopropyl ␤-D-thiogalactopyranoside was added to 1 mM, and 3 h later E. coli cells were harvested.
E. coli cells (1.6 g) were suspended in 32 ml of 50 mM Tris/HCl, pH 7.8, containing 1 mM EDTA and 1 mM dithiothreitol (buffer C) and disrupted by sonication. The undisrupted cells were removed by centrifugation at 12,000 rpm for 10 min (Sorvall, GSA rotor), and the supernatant was centrifuged at 39,000 rpm for 45 min (Beckman 42Ti rotor). Most of the expressed peptide was present in the supernatant fraction. The supernatant obtained was loaded onto DEAE-Bio-gel Aagarose column (1.5 ϫ 14 cm) equilibrated with buffer C, and peptides were eluted by a linear (0 -0.5 M) gradient of NaCl (total volume, 200 ml). The elution position of domain III was located by SDS-polyacrylamide gel electrophoresis using aliquots of the fractions. The fractions containing domain III were combined and concentrated to ϳ1 ml using a Centricon-10 concentrator. Then the sample was loaded on an Ultrogel AcA34 column (1.5 ϫ 68 cm) equilibrated with buffer C containing 0.1 M NaCl, and 0.9-ml fractions were collected. The fractions containing domain III were combined and concentrated to ϳ1.5 ml by a Centricon-10 concentrator. Approximately 10 mg of homogeneous domain III were obtained by this procedure from 500 ml of the culture medium.
Isolation of Domain I of the Bovine Transhydrogenase-Bovine transhydrogenase and its NH 2 -terminal 43 and 41.5 kDa tryptic fragments (domain I) were prepared as reported previously .
Other Methods-Protein concentration was determined by the modified Lowry method (Peterson, 1977) using bovine serum albumin as a standard. SDS-polyacrylamide gel electrophoresis was carried out according to Laemmli (1970).

RESULTS AND DISCUSSION
Experimental results on the bovine transhydrogenase together with theoretical considerations have indicated that outward proton translocation coupled to NADPH 3 NAD transhydrogenation (reverse of Equation 1) is driven via protein conformation change mainly by the difference in the binding energies of NADPH and NADP (Hatefi and Yamaguchi, 1992). Recent results from this laboratory have further supported this view and have provided evidence that the NADP(H)-binding domain of the E. coli transhydrogenase is in communication with a region of the membrane-intercalating domain II that appears to be concerned with proton translocation (Yamaguchi and Hatefi, 1995). These findings made it particularly desirable to obtain structural information about the NADP(H)-binding domain III and persuaded us to express domain III of the bovine enzyme in E. coli. This was done (see "Experimental Procedures"), and the expressed domain III of the bovine transhydrogenase was isolated in good yield. However, it was necessary to know whether this expressed and isolated domain III possessed the conformation it has in the native enzyme, even though preliminary experiments had shown that it binds [ 14 C]NADPH.
Because hydride ion transfer between NAD(H) and NADP(H) is direct in the transhydrogenase, one would expect that the nicotinamide rings of the two nucleotides would be only a few Angströms apart and the nucleotide-binding regions of domains I and III would share complementary surfaces and attractive forces to allow such close approximation. Therefore, it was thought possible that isolated domains I and III might interact and catalyze transhydrogenation. As will be seen below, this expectation proved correct not only for the interaction of isolated domains I and III of the bovine transhydrogenase but also for cross-interaction of domain I from the R. rubrum (␣1 subunit) and the expressed domain III from the bovine enzymes (Fig. 1). . Domain I of the bovine enzyme was isolated after controlled digestion of the purified enzyme with trypsin. As was shown previously , trypsin cleaves the bovine enzyme at Lys 410 -Thr 411 as well as at Lys 428 -Thr 429 , resulting in two fragments that copurify with respective molecular mass values of 41.5 and 43 kDa. The molecular masses of the bovine domain III and the R. rubrum ␣1 subunit are 20 and 40.3 kDa, respectively, but the former displays on SDS-gels an anomalous mobility corresponding to an molecular mass of about 29 kDa.
Data for reconstitution of transhydrogenase activity involving the 20-kDa domain III of the bovine enzyme with domain I from the same source (43 and 41.5 kDa) or from R. rubrum (␣1 subunit, 40.3 kDa) are shown in Table I. It is seen that reconstitution of NADPH 3 AcPyAD transhydrogenation activity was much more effective between domain I from the R. rubrum and domain III from the bovine enzymes than between these two domains from the latter source and that each domain added alone to the assay medium exhibited little or no activity. In these experiments, the protein fragments were placed on a glass rod and added together to the assay mixture. Preincubation of the fragments together at high concentration or their

FIG. 1. Schematic representation of the tridomain composition of the bovine and the R. rubrum transhydrogenases and the bidomain composition of the reconstituted transhydrogenases.
The bovine enzyme is a homodimer in which domain I is hydrophilic and binds NAD(H), domain II is hydrophobic and carries the enzyme's proton channel, and domain III is hydrophilic and binds NADP(H) (Hatefi and Yamaguchi, 1992;Yamaguchi and Hatefi, 1993 separate addition to the assay medium did not result in very different activities. The effect of pH on the rate of NADPH 3 AcPyAD transhydrogenation catalyzed by R. rubrum domain I plus bovine domain III was similar to that of the intact enzymes. The activity increased slightly in going from pH 8.5 to 7.5 and by Ͼ1.5-fold in going from pH 7.5 to 6.5. Table I also shows data for the transhydrogenation reactions NADH 3 AcPyADP and NADH 3 AcPyAD as catalyzed by the R. rubrum domain I (␣1 subunit) plus the bovine domain III. It is seen that relative to the rate of NADPH 3 AcPyAD reaction, the rate of the forward transhydrogenation from NADH to AcPyADP is also quite substantial. In the intact enzyme, this reaction is very much slower than NADPH 3 AcPyAD transhydrogenation and is accelerated about 10-fold in the presence of a proton motive force. The fact that the rates of the forward and reverse transhydrogenation are not very different as cat-alyzed by the protein fragments lacking the enzyme's proton translocation domain II may indicate that in the intact enzyme this domain inhibits NADH 3 AcPyADP transhydrogenation in the absence of a proton motive force. The last entry in Table  I shows transhydrogenation from NADH to AcPyAD, an activity that the intact enzyme does not exhibit unless under special conditions in the presence of NADP(H) (Wu et al., 1981;Fisher and Earle, 1982;Hutton et al., 1994). However, we have observed that the E. coli transhydrogenase mutant ␤H91K also catalyzes NADH to AcPyAD transhydrogenation in the absence of added NADP(H). 2 Glavas et al. (1995) state in a recent paper that their ␤H91K mutant contains bound NADP, but the data were not presented.
In the reaction NADPH 3 AcPyAD catalyzed by the R. rubrum domain I (␣1 subunit) plus the expressed bovine domain III, apparent K m values, as determined from Lineweaver-Burk double reciprocal plots, were 10.9 M for AcPyAD and 3.2 M for NADPH. The respective values for the R. rubrum and the bovine transhydrogenases are 26 and 20 M (this study and Yamaguchi et al., 1990). In the E. coli enzyme mutation of Asp ␤213 to Ile, which inhibits proton translocation by 90%, lowered the apparent K m NADPH by 3.5-fold and K d NADPH by 3.3fold (Yamaguchi and Hatefi, 1995). It is possible that the absence of the membrane-intercalating domain II in the system reconstituted from domains I and III has a similar effect on the affinity of domain III for NADPH. Table II contains data on the inhibitory effects of FSBA and NEM. In the bovine transhydrogenase, FSBA modifies Tyr 245 in the NAD(H)-binding domain I and very slowly modifies Tyr 1006 in the NADP(H)-binding domain III (Wakabayashi and Hatefi, 1987). As seen in Table II, pretreatment with FSBA had only a small inhibitory effect on the R. rubrum ␣1 subunit but nearly completely inactivated the bovine domain III. In five transhydrogenases whose predicted amino acid sequences are known, the residue corresponding to the bovine Tyr 245 is conserved, and in the R. rubrum ␣1 subunit this conserved Tyr is in a glycine-rich region 33 residues downstream of a ␤␣␤ fold and is flanked by residues GGYAKEM, which are the same in the bovine enzyme (Yamaguchi and Hatefi, 1994). It is therefore surprising that in the bovine transhydrogenase Tyr 245 readily reacts with FSBA, but in the R. rubrum ␣1 subunit the corresponding Tyr does not.
The data for the effect of NEM were the same as would be expected from the behavior of the bovine enzyme toward NEM, in which this reagent alkylates Cys 893 in domain III.   Table I for NADPH 3 AcPyAD transhydrogenation. Where indicated, purified R. rubrum domain I (␣1 subunit) or expressed bovine domain III was pretreated for 30 min with 2 mM FSBA or for 2 min with 2 mM NEM in the absence or presence of 0.5 mM NADPH or NADH. The concentration of domain III was 10 g throughout. The concentration of domain I was 1.9 g in the first three experiments and 1.84 g in the others. Specific activity is expressed as in Table I   . This preparation contains two peptides of molecular masses 43 and 41.5 kDa. Lane B, domain III of the bovine transhydrogenase expressed in E. coli and isolated therefrom. Lane C, domain I of R. rubrum transhydrogenase isolated from chromatophore membranes. SDS-polyacrylamide gel electrophoresis was carried out according to Laemmli (1970), and to each lane 1 g of the preparation indicated was added. The gels were stained with Coomassie Blue and destained in 10% acetic acid. The lane to the extreme left shows protein standards with their molecular masses indicated in kDa.

TABLE I
Reconstitution of transhydrogenase activity with the soluble, nucleotide-binding domains I and III of the enzyme Assay conditions were the same as described under "Experimental Procedures." Where indicated, 7 g of bovine domain I, 10 g of expressed bovine domain III, and 1.84 g of R. rubrum domain I were added. Specific activity (mol AcPyAD or AcPyADP reduced/min/mg protein) is expressed per mg of bovine or R. rubrum domain I where one or the other was present. The NADPH 3 AcPyAD activity of purified bovine transhydrogenase is 30 -32 mol AcPyAD reduced (min ⅐ mg) Ϫ1 (Yamaguchi and Hatefi, 1993 chi and Hatefi, 1989). As seen in Table II, NEM had no inhibitory effect on the R. rubrum ␣1 subunit but inhibited the bovine domain III. Also, similar to the intact bovine enzyme, addition of NADPH but not NADH stimulated the NEM inhibition. In the bovine transhydrogenase, it was shown that the reason for this augmentation of the NEM inactivation rate is that in the presence of NADPH, the effective pK a of Cys 893 is lowered from 9.1 to 8.7 (Yamaguchi and Hatefi, 1989). Not shown in Table II is the inhibitory effect of palmitoyl-CoA (Rydström, 1972). In the R. rubrum-bovine cross-reconstituted system of Table II, palmitoyl-CoA was a competitive inhibitor with respect to NADPH, with K i ϭ 3.36 M. The reconstituted NADPH 3 AcPyAD transhydrogenase activities shown in Table I are considerably less than the activity of the intact enzyme purified from bovine mitochondria. This may be related to the fact that in the intact transhydrogenase the attractive forces between domains I and III do not appear to be strong. In R. rubrum, the ␣1 subunit (domain I) is readily washed away from the chromatophore membranes, and in submitochondrial particles, domains I and III of the bovine enzyme can each be easily separated once its link to the membraneintercalated domain II is severed by controlled proteolysis (Yamaguchi and Hatefi, 1991). Nevertheless, the reconstituted activities shown above indicate that domains I and III interact physically to form a catalytic unit. Whether the presence of bound nucleotides strengthens or weakens this physical association remains to be investigated. What is known is that in the intact bovine enzyme NADPH binding at domain III makes domain I more susceptible to attack by dicyclohexylcarbodiimide or trypsin (Phelps and Hatefi, 1984;Yamaguchi et al., 1990).
The transhydrogenases of E. coli and bovine mitochondria, whose amino acid sequences are known and have been purified in the presence of detergents, are somewhat unstable in the isolated state and tend to aggregate when concentrated. These problems complicate attempts at crystallization for structure studies. By comparison, the purified domain I of the R. rubrum and the expressed and purified domain III of the bovine enzymes are water soluble and relatively stable. Furthermore, the R. rubrum domain I (␣1 subunit) has been expressed in E.
coli and can also be obtained in high yield (Diggle et al., 1995). Therefore, these preparations provide an opportunity for structure studies. It may even be possible to reconstitute and cocrystallize them and solve the structure of the catalytic sector of the transhydrogenase.