Mitochondrial Energy-linked Nicotinamide Nucleotide Transhydrogenase MEMBRANE TOPOGRAPHY OF THE BOVINE ENZYME*

The mitochondrial energy-linked nicotinamide nu- cleotide transhydrogenase is a homodimer of monomer M, = 109,228. Hydropathy analysis of its cDNA-de-duced amino acid sequence (1043 residues) has indi- cated that the molecule is composed of 3 domains: a 430-residue-long hydrophilic N-terminal domain which binds NAD(H), a 200-residue-long hydrophilic C-terminal domain which binds NADP(H), and a 400-residue-long hydrophobic central domain which ap- pears to be made up mainly of about 14 hydrophobic clusters of -20 residues each. In this study, antibodies were raised to the hydrophilic N- and C-terminal domains cleaved from the isolated transhydrogenase by proteolytic digestion, and to a synthetic, hydrophilic pentadecapeptide, which corresponded to position 640-554 within the central hydrophobic domain. Im-munochemical experiments with mitoplasts (mitochon- dria denuded of outer membrane) and submitochondrial particles (inside-out inner membrane vesicles) as sources of antigens showed that essentially the entire N- and C-terminal hydrophilic domains of the trans- hydrogenase, as well as epitopes from the central pentadecapeptide, protrude from the inner membrane into the mitochondrial matrix, where the N- and C-terminal domains would be expected to come together to form the enzyme’s catalytic site. Treatment of mitoplasts with several proteolytic enzymes indicated that large protease-sensitive masses of the

ing two peptides of 71 and 39 kDa was isolated from mitoplasts and shown to have high transhydrogenase activity.
The mitochondrial energy-linked nicotinamide nucleotide transhydrogenase catalyzes the direct and stereospecific transfer of a hydride ion between the 4A position of NAD(H) and the 4B position of NADP(H) in a reaction that is coupled to transmembrane proton translocation with a H+/H-stoichiometry close to unity (see Equation 1 where HL and H,' are matrix and cytosolic protons, respectively) (1)(2)(3). NADH + NADP + H: + NAD + NADPH + H+, (1) The enzyme from bovine heart mitochondria is a homodimer of monomer M , = 109,228. It is nuclearly encoded, extramitochondrially synthesized, and transported into mitochondria (4, 5). The amino acid sequences of the mature protein (1043 residues) and its signal peptide (43 residues) have been deduced, respectively, from cDNA clones and the mRNA (6, 7). Escherichia coli contains a similar enzyme. It is composed of two subunits, a with M, = 54,000 and / 3 with M , = 48,700 (8). The amino acid sequences of these subunits have also been deduced from the nucleotide sequences of the genes (8). There is considerable sequence identity between the E. coli and the bovine enzymes, especially in the nucleotide binding domains (-65% identity) as determined for the latter enzyme (6).
The bovine transhydrogenase contains a 430-residue-long N-terminal domain, which is hydrophilic and binds NAD(H), and a 200-residue-long C-terminal domain, which is also hydrophilic and binds NADP(H) (6, 9). The central 400residue-long segment of the protein is hydrophobic, and hydropathy analyses have indicated that it is composed mainly of 14 hydrophobic clusters of -20 residues each (6). There are in this central segment only 25 charged residues, 6 Asp, 7 Glu, 6 Lys, and 6 Arg. Thus, one would expect a priori that the hydrophilic N-and C-terminal domains would be extramembranous, protruding into the mitochondrial matrix where they would together form the catalytic site of the enzyme, and the central hydrophobic stretch would intercalate mainly into the membrane and make up the proton channel of the molecule. As will be seen below, the results of our studies on the membrane topography of the bovine transhydrogenase are in full accord with these expectations.
Assay of Transhydrogenase Activity-Transhydrogenation from NADPH to AcPyAD was assayed a t 37°C in a reaction mixture containing 100 mM sodium phosphate, p H 6.5, 4 pg of L-a-lysophosphatidylcholine, and 0.3 mM concentration each of NADPH and AcPyAD. The reaction was started by the addition of enzyme, and the reduction of AcPyAD was followed a t 375 minus 425 nm in an Aminco DW2a dual-wavelength spectrophotometer. Rates were calculated using a value of 6.38 mM cm" for the absorbance difference of AcPyADH and NADPH at the above wavelength pair. When the enzyme activity was measured on SMP or mitoplasts, L-a-lysophosphatidylcholine was omitted from the reaction mixture, and 1 p~ rotenone was added. In the case of mitoplasts, 0.1% Lubrol P X was also added.
Preparation of SMP, Mitochondria, and Mitoplasts-SMP were prepared according to Ref. 12. Intact bovine heart mitochondria for preparation of mitoplasts were prepared by the method of Hatefi et al. (13), and mitoplasts were prepared according to Krebs et al. (14). The activities of monoamine oxidase (outer membrane enzyme) and malate dehydrogenase (matrix enzyme) in the mitoplast preparation were 4 0 % and >95%, respectively, of the total activities that were present in the original sample of mitochondria.
Purification of Transhydrogenase from Proteinase K-treated Mitoplasts-Mitoplasts (280 mg in 28 ml of STA buffer) were digested by the addition of 5.4 mg of proteinase K for 80 min at 23"C, then 560 pl of 0.2 M PMSF in methanol were added to stop further digestion. The residual transhydrogenase activity was 85% as compared to the activity of control mitoplasts not treated with proteinase K. Preparation of SMP from proteinase K-treated mitoplasts and purification of transhydrogenase therefrom followed the procedure described previously (1 1). From 280 mg of mitoplasts, 70 mg of SMP were obtained, and from the SMP 270 pg of transhydrogenase with a specific activity of 24.6 pmol of AcPyAD reduced (min. mg of protein)".
Isolation of the N-terminal 43-kDa and C-terminal 20-kDa fragments-Purified transhydrogenase was digested with trypsin in the presence of 0.4 mM NADPH as described previously (15). The tryptic digest was subjected to SDS-polyacrylamide gel electrophoresis, and the N-terminal 43-kDa peptide was excised and electroeluted from the gel by the method of Hunkapillar et al. (16). Usually -400 pg of peptide was obtained from 2 mg of transhydrogenase. When 4 mg of transhydrogenase in 0.5 ml of 50 mM Tris acetate, p H 7.5, containing 0.01% potassium cholate and 0.4 mM NADPH, were digested with 2 pg of papain a t room temperature for 13 min and the digest was loaded on a Sephadex G-200 column (0.94 X 47 cm), four peaks (A-D ) appeared upon elution of the column with 50 mM Tris acetate, p H 7.5, containing 0.01% potassium cholate (see Fig. 1 below). Peak A contained protein bands of approximately 61,67,82, and 110 kDa. Peak B contained two peptides of 42 and 46 kDa (total -450 pg) whose N-terminal sequences were the same as that of the transhydrogenase itself. Peak C contained a peptide (-200 pg) with an apparent M , of 26,000. The N-terminal sequence of this peptide was determined to be GGKP, as described below, indicating that papain had cleaved the Aa56-G*s6 bo nd near the C terminus. This C-terminal peptide (GR"-K'"') has a molecular mass of 20,392 and will therefore be referred to as the 20-kDa C-terminal fragment. Peak D was mainly due to the absorbance of the added NADPH.
booster injection of 100 pg of peptide in 0.6 ml of the same buffer plus 0.6 ml of Freund's incomplete adjuvant was given subcutaneously. Thereafter, a schedule of one monthly boosting injection was followed. Blood was collected from the ear vein 10 days after each boosting injection.
The peptide DMFKRPTDPPEYNYL, with inclusion of a cysteine residue at the C terminus, was synthesized by Dr. John Tomich (Medical Genetics Division, Children's Hospital, Los Angeles) essentially as reported (18) and was conjugated to the carrier protein, keyhole limpet hemocyanin (KLH), using the hetero-bifunctional reagent succinimidyl 4-(N-maleimidomethyl)cyclohexane-l-carboxylate as described by the manufacturer, Pierce. The peptide (-12 mg) was reduced with dithiothreitol and denatured in 4 M guanidine HCI. Just prior to use, the reducing agent was removed by gel filtration, and the peptide was mixed with 10 mg of the activated carrier. The mixture was allowed to react for 3 h at 25"C, then stored a t 4°C until it was used to inoculate the animals. The KLH-peptide conjugate (-200 pg) in 0.75 ml of PBS was mixed with the same volume of incomplete Freund's adjuvant and, 3 mg of M. butyricum and injected subcutaneously in several places along the back of the rabbit. Fourteen days later, the same solution lacking M. butyricum was similarly injected. Seven days later, 200 pg of KLH-peptide conjugate in 0.6 ml of PBS was mixed well with 0.4 ml of 10 mg/ml Al(OH):, and injected intraperitoneally (19). Ten weeks later, boosting was carried out in the same way as the third injection, and, 14 days later blood was collected.
Control serum was collected from each animal prior to initiation of the injections. The blood was allowed to coagulate a t room temperature, then centrifuged a t 55,000 X g for 15 min. The serum was stored a t -20°C.
Affinity Purification of Antibodies-Purified transhydrogenase (-2 mg of protein) was subjected to SDS-polyacrylamide gel electrophoresis and transferred to a PVDF membrane as described below (20). The membrane was stained with 0.5% Ponceau S in 1% acetic acid for 8 min and destained with water, and the transhydrogenase band was excised. This piece of membrane was used for purification of antibodies raised against the N-terminal 43-kDa and the C-terminal 20-kDa peptides as well as the synthetic peptide described above. Affinity purification of the antisera followed the method of Bisson and Schiavo (21). Preimmune sera were subjected to the same procedure as described above before use.
Treatment of SMP and Mitoplasts with Various Proteases-SMP, 3.8 mg in 0.8 ml of STA containing 0.4 mM NADPH, were treated with 50 pg of papain or 16 pg of trypsin a t 23°C for 40 min, and the digestion was stopped by the addition of 8 p1 of 0.2 M PMSF in methanol and 50 pl of leupeptin (1 mg/ml). Mitoplasts, 1 mg of protein in 100 pl of STA (0-25 M sucrose containing 10 mM Tris acetate, pH 7.5), were treated with 10 pg of various proteases at 23°C for 60 min. Further digestion was stopped by the addition of 2 pl of 0.2 M PMSF. In the case of thermolysin treatment, 1 pl of 0.5 M EDTA was added to stop further digestion. Aliquots containing 25-30 pg of SMP or mitoplast protein from control and protease-treated samples were assayed for transhydrogenase activity. In some cases, the digests were separated into pellet and supernatant fractions using a Beckman Airfuge (100,000 rpm, 5 min) and analyzed as described under "Results." Detection of Peptides by Zmmunoblotting-Protein samples were denatured by the addition of 2 volumes of SDS denaturation buffer (94 mM Tris-HCI, pH 6.8, 15% glycerol, 7.5% P-mercaptoethanol, 3% SDS, and 6 M urea) followed by immersion in boiling water for 4 min, then subjected to 12% SDS-polyacrylamide gel electrophoresis (22).
Peptides were transferred onto PVDF membranes using a Bio-Rad Mini Trans-Blot apparatus. The electrophoretic buffer solution contained 20 mM Tris acetate, p H 8.3, 1 mM EDTA, and 0.2 mM dithiothreitol. Electrotransfer was performed at 30 V for 1 h. After protein transfer, membranes were incubated with 1% skim milk in PBS for 1 h and then for 3 h with the affinity-purified antibodies to the N-terminal 43-kDa or the C-terminal 20-kDa peptide diluted 30fold in 50 mM Tris-HCI, p H 8.0, containing 2 mM CaC12 and 3% bovine serum albumin. These sheets were washed several times with PBS and incubated for 1 h with anti-rabbit IgG-peroxidase conjugate diluted 5000-fold in the dilution buffer described above. They were then washed with PBS three times and stained by immersion in 25 ml of PBS mixed with 5 ml of ethanol containing 15 mg of 4-chloro-1-naphthol and 15 p1 of 29% H20,.
"'II-Protein A Labeling-To each microultracentrifuge tube were added 50 p1 of SMP, mitoplasts, or mitochondria (1 mg/ml in STA), 40 pl of STA, and 10 p1 of control IgG or purified IgG. The tubes were kept overnight a t 4°C and centrifuged for 5 min a t 100,000 rpm in a Beckman Airfuge. The pellets were washed twice by suspension in STA and recentrifuged. Then, 50 p1 of '"1-protein A (6.4 ng, 4.2 X 10" cpm) were added to each tube and incubated for 1 h a t 37°C. The tubes were centrifuged and the pellets were washed three times as before. The radioactivity in the pellets was counted in a Packard Autogramma counter.
Enzyme-linked Immunosorbent Assay (ELZSA)-MicroTest wells were coated with 100 pl of antigen in PBS or STA, and the plates were incubated for 1 h a t 37°C. After the plates were washed with the same buffer, 200 p1 of 2% bovine serum albumin in PBS or STA was added to each well, and the plates were further incubated as before. The wells were then washed several times with the same buffer, and each was treated with 100 pl of control IgG or purified IgG at the dilutions shown in the figures. The plates were incubated for 2 h a t 37"C, followed by thorough washing of the wells with buffer. Antirabbit IgG conjugated to alkaline phosphatase was diluted 2500-fold in PBS or STA, and 100 pl were added to each well. The plates were incubated for 1 h a t 37°C. After the wells were washed, 100 pl of substrate solution (1 mg/ml p-nitrophenyl phosphate dissolved in a solution containing 1 M diethanolamine, 0.5 mM MgC12, and 0.02% NaNs, pH 9.8) were added to each well, the plates were allowed to stand at room temperature for 90 min for color development, and the absorbance of each well was read a t 405 nm in the Titertek Multiskan spectrophotometer.
N-Terminal Sequencing of Peptides-Peptides were separated by SDS-polyacrylamide gel electrophoresis (12% Laemmli gel, Ref. 22) and transferred to PVDF membranes. Peptides were located by staining with a mixed solution of Ponceau S (0.09%) and Coomassie Blue (0.01%) in 10% methanol, then excised from the membranes. They were loaded onto a gas phase sequencer, and the N-terminal amino acids (4-6 residues) of each peptide were determined as reported previously (15). Protein Assays-Protein concentration was measured by the method of Peterson (23), with bovine serum albumin as standard. For mitochondria, mitoplasts, and submitochondrial particles, protein concentration was estimated by the biuret method (24) in the presence of 0.1% sodium deoxycholate.

RESULTS
The strategy used for study of the membrane topography of the transhydrogenase was 2-fold. (a) By controlled proteolytic digestion, the hydrophilic N-and C-terminal domains of the protein were cleaved and isolated from the purified enzyme. Also, an antigenic peptide (DMFKRPTDPPEY-NYL), which in the transhydrogenase is located at position 540-554 within the central hydrophobic domain of the molecule, was synthesized and linked to KLH. Antibodies were raised against these three fragments in the rabbit and tested by ELISA and "'I-protein A binding against mitochondria, mitoplasts (mitochondria denuded of outer membrane), and SMP (inside-out inner membrane vesicles). It might be added that the above pentadecapeptide is not present in the E. coli transhydrogenase, because by comparison, the CY subunit of the E. coli enzyme ends at position 538 of the bovine transhydrogenase, and its p subunit begins at position 570. ( b ) Mitoplasts and SMP were subjected to controlled proteolysis, and antigens to the above N-and C-terminal antibodies that had remained bound to the membrane and released therefrom were characterized.
Isolation of the Nand C-terminal Hydrophilic Fragments- The N-terminal hydrophilic domain of the bovine transhydrogenase was isolated from the purified enzyme after tryptic digestion in the presence of NADPH (15). The isolated Nterminal fragment was shown to be dimeric and capable of binding to NAD-agarose, like the purified transhydrogenase, and exhibited on SDS gels a monomer M , -43,000 (15). A minor band of M , = 41,500 was also present, and both bands had the same N-terminal sequences as the intact transhydrogenase. The 43-kDa peptide was isolated as described under "Experimental Procedures" and used for production of antibodies. The bond cleaved by trypsin to produce the 43-kDa fragment was shown to be K410-T4*' (15). The C-terminal hydrophilic domain was isolated similarly, except that the proteolytic enzyme used in this case was papain. As seen in Fig. 1, digestion of purified transhydrogenase with papain produced three soluble fragments, two had M , values of 46,000 and 42,000 and were eluted from Sephadex G-200 together (peak B of Fig. l), and one exhibited a M , value of 26,000 and was eluted separately from Sephadex G-200 (peak C of Fig.  1). The 46-and 42-kDa fragments had the same N-terminal residues as the intact transhydrogenase and reacted with antibodies raised against the N-terminal 43-kDa fragment described above. The N-terminal sequence of the 26-kDa fragment was GGKP, thus indicating that papain had cleaved off the C-terminal hydrophilic domain of the transhydrogenase at A8s5-G856. These results also indicated that the Cterminal fragment released by papain had 188 amino acid residues and a calculated molecular weight of 20,392. Implicit in this molecular weight calculation is of course the assumption that papain had not removed a few residues from the Cterminal end of this fragment. However, 20,392 is clearly much closer to the molecular weight of the C-terminal fragment than 26,000 as suggested by the mobility of this fragment on SDS gels. Therefore, we shall refer to the C-terminal peptide released by papain as the 20-kDa fragment.
Membrane Sidedness of Epitopes in the Hydrophilic Nand C-terminal Domains and Peptide D540-L554 of the Transhydrogenase-Antibodies raised against the N-terminal 43-kDa fragment, the C-terminal 20-kDa fragment, and the D540-L554 peptide each reacted with the purified transhydrogenase in ELISA and recognized a single polypeptide of M , = 110,000 in immunoblots of SDS gels of whole mitochondria or SMP. Moreover, the reactivity of each antiserum was blocked when it was pretreated with an excess of its respective antigen. When the anti-43-kDa and the anti-20kDa IgG were tested against intact mitochondria, mitoplasts, and SMP, using '*'Iprotein A binding to monitor antigen-antibody interaction, the results with mitochondria and mitoplasts were essentially negative (Fig. 2), the slight positive response a t high antibody concentration being probably due to low levels of mitochondrial fragments in the preparations. By comparison, however, the results with SMP were strongly positive (Fig. 2). These data indicated that epitopes from the N-terminal43-kDa and the C-terminal20-kDa domains of the transhydrogenase were exposed on the matrix, but not on the cytosolic, side of the inner membrane. Anti-43-kDa antiserum partially inhibited the trhshydrogenase activity of SMP, but anti-20-kDa antiserum had no inhibitory effect a t comparable levels (Fig. 3).
Results similar to those of Fig. 2 were obtained with the D540-L5s4 anti-peptide antibody. Using ELISA to monitor antigen-antibody interaction, it was shown that the antipeptide antiserum reacted strongly with the purified enzyme and SMP, but showed marginal reactivity toward mitoplasts (Fig.  4). Again, the low reactivity with mitoplasts could be due to the presence of low levels of fragmented mitoplasts, which is a likely occurrence during the manipulations involved in ELISA. These results indicated, therefore, that epitopes from the D540-L554 peptide are also exposed on the matrix side of the inner membrane.

Extramembranous Segments of the Transhydrogenase Mol-
ecule-While it is highly likely that the hydrophilic D540-F54 pentadecapeptide is largely extramembranous and exposed to the medium on the matrix side of the inner mitochondrial membrane, a similar conclusion does not follow from the results of Fig. 2 for the Nand C-terminal hydrophilic domains. Fig. 2 only shows that epitopes from these segments which are reactive toward our antibodies are exposed on the matrix side. Indeed, hydropathy analyses of these domains,

20.
FIG. 5. Determination of the sizes of the N-and the Cterminal segments of the transhydrogenase that protrude from submitochondrial particles. S M P were treated with papain as described under "Experimental Procedures." Aliquots containing 25 pg of protein from untreated and papain-treated S M P were assayed for transhydrogenase activity (30% of control) and electrophoresed on 12% SDS-polyacrylamide gels. Pellet and supernatant fractions were obtained by centrifuging 400 pl of papain-treated SMP, and equivalent aliquots of these fractions were also electrophoresed. Then, proteins were transferred to PVDF membranes, and the membranes were  (panel B, lanes 2-4) peptides.

terminus in N-terminal 42-and 44-kDa (panel A , lunes 2-4) and C-terminal66-kDa (panel B, lanes 2 a n d 3 ) peptides, and near the C terminus in N-terminal82-kDa (panel A, lanes 2 and 3) and C-terminal 26-kDa
especially of the N-terminal430 residues, show several hydrophobic stretches of amino acids, one or more of which could be intramembranous. Therefore, it was important to find out how much of the transhydrogenase protein mass is extramembranous on the matrix as well as on the cytosolic side of the inner membrane.
For this purpose, SMP and mitoplasts were treated with appropriate proteolytic enzymes, the soluble and membranous fractions were separated by centrifugation, and the sizes of antigens reactive toward the N-terminal anti-43-kDa and the C-terminal anti-20 kDa antisera were determined by SDS-gel electrophoresis of each fraction and immunoblotting. When SMP were treated with papain, the transhydrogenase fragments solubilized were essentially the same as those which were removed by papain from the purified enzyme (Fig. 1, peaks B and C). These results are shown in the immunoblots of Fig. 5, where lanes 1, 2, 3, and 4 A , lane  4 ) , and a smaller peptide with M , (apparent) of 26,000, which blotted to the C-terminal 20- kDa antibodies (panel B, lane   4 ) . The latter peptide was excised from the PVDF membranes, and its N-terminal sequence was determined to be GGKPME, which showed that papain had hydrolyzed the same bond, i.e. AR5S-GRRB, in SMP-bound transhydrogenase as it had done in the purified enzyme. These results indicated, therefore, that essentially the entire 430-residue-long N-terminal hydrophilic domain and the ZOO-residue-long C-terminal hydrophilic domain of the transhydrogenase molecule are extramembranous and bathe in the matrix milieu of mitochondria. This conclusion was confirmed when SMP were treated with trypsin instead of papain. Two N-terminal segments were solubilized with M , values exactly the same as those reported previously for the N-terminal tryptic fragments isolated from the purified enzyme, i.e. about 43,000 and 41,000 (15). The most abundant C-terminal fragment found in the soluble fraction exhibited a M , of 14,000, which in agreement with the primary sequence of the enzyme indicated that the potential tryptic cleavage sites in the C-terminal domain are downstream of the papain cleavage site a t AR""-GR"".
Essentially similar experiments were carried out to investigate the segments of the transhydrogenase molecule exposed on the cytosolic side of the inner mitochondrial membrane, except that in this case mitoplasts (mitochondria denuded of outer membrane) were used instead of SMP. As seen in Fig.  6, treatment of mitoplasts with various proteolytic enzymes produced either no detectable transhydrogenase fragments or split the enzyme into two pieces, an N-terminal piece of M , -71,000 plus a C-terminal piece of M , -39,000. These results suggested that a large transhydrogenase segment is not exposed on the cytosolic side of the inner mitochondrial membrane, because one would have expected such an exposed segment to be susceptible to multiple cuts by the proteases used, thus resulting in N-and C-terminal fragments whose molecular weight would not add up to that of the intact protein. This reasoning is, of course, in excellent agreement with predictions based on the hydropathy analysis of the transhydrogenase primary sequence. It has already been shown above that the hydrophilic N-and C-terminal domains of the enzyme protrude into the matrix. What remains is the central highly hydrophobic domain of the molecule, which one would not expect to be extramembranous to any great extent. The hydropathy analysis of this central segment suggests that it is composed of about 14 hydrophobic clusters of -20 residues each. If one assumes that these clusters span the membrane, then there would have to be links between the clusters that are exposed on the matrix and the cytosolic sides of the inner membrane. These hydrophilic links, as suggested by the hydropathy data, are all small, and Fig. 6 suggests that those exposed on the cytosolic side are resistant to proteolytic attack, except for one link which is broken by proteinase K, subtilisin, thermolysin, and pronase E (a mixture of several proteases) and results in a 71-kDa N-terminal and a 39-kDa C-terminal fragment.
Further characterization of the above fragments was carried out, using proteinase K. First, it was shown that proteinase K treatment of mitoplasts to the extent that essentially all transhydrogenase molecules were split produced only the 71and the 39-kDa bands seen in Fig. 6 (data not shown). Second, it was demonstrated that, after complete conversion of the intact 110-kDa transhydrogenase into the 71-and the 39-kDa fragments, the detergent-solubilized mitoplasts had retained 85-90% transhydrogenase activity (NADPH .--, AcPyAD) as compared to control mitoplasts not treated with proteinase K (data not shown). Third, it was found that the proteinase Knicked enzyme could be isolated from mitoplasts, still retaining >70% activity. These results are shown in Fig. 7. Lane 1 is a Coomassie blue-stained SDS gel of the nicked transhydrogenase isolated from proteinase K-treated mitoplasts. It shows a thin band of the intact transhydrogenase (110 kDa) and two bands of about 71 and 39 kDa, which represent the nicked enzyme. Lanes 2 and 3 are immunoblots of the material of lane 1 blotted with the N-terminal 43-kDa antibodies in lane 2 and with the C-terminal 20-kDa antibodies in lune 3. In a similar experiment, mitoplasts were treated with proteinase K to the extent that the intact transhydrogenase band was about 90% cleaved. Then, the enzyme was purified from the proteinase K-treated and untreated mitoplasts and assayed for activity. The purified intact enzyme had a transhy- drogenase activity of 32 pmol of AcPyAD reduced by NADPH per min per mg of protein; the purified nicked enzyme (containing 510% intact transhydrogenase) exhibited an activity of 24.6. Finally, the 39-kDa C-terminal fragment produced by proteinase K treatment of mitoplast-bound transhydrogenase was isolated, and its N-terminal sequence was determined to be AANLT, which indicated that the bond cleaved by proteinase K is A690-A691. According to our hydropathy data, these residues are located in a hydrophilic segment between hydrophobic clusters 9 and 10 of the transhydrogenase molecule (see "Discussion"). Thus, one may conclude that this small hydrophilic segment is extramembranous and protrudes from the cytosolic side of the inner membrane into the intermembrane space.

DISCUSSION
The hydropathy profile of the bovine mitochondrial transhydrogenase based on the enzyme's predicted amino acid sequence suggested that the transhydrogenase molecule is composed of three domains, a 430-residue-long N-terminal hydrophilic domain, a 400-residue-long central hydrophobic domain, and a 200-residue-long C-terminal hydrophilic domain (6). The N-and the C-terminal domains were shown, respectively, to bind NAD(H) and NADP(H) (9). These domains could be isolated in soluble form after appropriate and limited proteolytic digestion of the purified transhydrogenase, and the purified N-terminal hydrophilic domain (a dimer containing two peptides with M , values mainly of 43,000 and partly of 41,500) was shown to bind NAD with high affinity (15). These findings allowed certain predictions regarding the topography of the transhydrogenase molecule in the mitochondrial inner membrane: (i) that the N-and the C-terminal hydrophilic domains are extramembranous, (ii) that these extramembranous domains protrude into the mitochondrial matrix where together they form the catalytic site of the enzyme, and (iii) that the central hydrophobic domain is largely membrane-intercalated because of its low content of charged amino acids (25 residues only). These considerations also led to the further prediction that there is probably very little transhydrogenase mass protruding from the inner membrane on the cytosolic side, this despite the contrary conclusions of Weis et al. (25) which will be discussed below.
The above predictions were substantiated by the following experiments. The Nand the C-terminal hydrophilic domains (43 and 20 kDa, respectively) were cleaved by appropriate proteolytic digestion from the purified transhydrogenase, purified, identified by N-terminal sequencing of 5 to 6 residues, and injected into rabbits for production of polyclonal antibodies. A hydrophilic pentadecapeptide (D540-L554) located within the central hydrophobic domain, which was expected to be extramembranous and antigenic, was also synthesized and used for production of antipeptide antibodies. The antibodies raised were checked for reactivity against purified transhydrogenase, then used for topography studies against mitoplasts and SMP as sources of antigens. The results showed clearly that epitopes from the isolated N-and C-terminal domains as well as from the central hydrophilic peptide were exposed in SMP, but not in mitoplast preparations. In order to determine how much of the N-and C-terminal hydrophilic domains were protruding from SMP into the medium, the particles were treated in one experiment with papain, and in another with trypsin, then the soluble and the particulate fractions were separated by centrifugation, and the sizes of the solubilized N-and C-terminal antigens were determined after SDS gel electrophoresis of the supernatants and immunoblotting with antibodies to the 43-kDa N-termi-nal and the 20-kDa C-terminal fragments. The results indicated that essentially the entire N-and C-terminal hydrophilic domains are extramembranous and cleavable by appropriate proteolytic digestion of SMP. These findings are summarized in Fig. 8, which shows the N-and the C-terminal hydrophilic domains in abbreviated form and the central hydrophobic domain as interpreted from a Kyte-Doolittle hydropathy analysis with a setting of 9 residues (6, 26). The bonds hydrolyzed by trypsin in the N-terminal domain and by papain in the C-terminal domain are marked, and the antigenic pentadecapeptide (D540-L554) protruding into the matrix space is outlined with dashes.
To determine what parts of the transhydrogenase molecule were exposed on the cytosolic side of the inner membrane, mitoplasts were treated with seven different proteolytic enzymes. Two considerations governed the selection of these proteases. One was to have among them a wide substrate specificity, and another to be able to stop their activity before SDS-gel electrophoresis of the mitoplasts. The results showed that trypsin, @-chymotrypsin, and papain had little or no effect on the transhydrogenase embedded in mitoplasts, while proteinase K, subtilisin (Nagarse), thermolysin, and pronase E (a mixture of several proteases) each cleaved the mitoplastembedded transhydrogenase into a 71-and a 39-kDa fragment. Immunoblotting indicated that the 71-kDa fragment contained the N-terminal hydrophilic domain, while the 39-kDa fragment included the C-terminal hydrophilic domain. Since the M, values of the two fragments added up precisely to the M , of intact transhydrogenase (llO,OOO), these results argued against the cytosolic exposure of large segments of the enzyme, which would be susceptible to digestion by one or another of the seven proteases employed. What appeared more likely was the cytosolic exposure of a single small peptide sensitive to four of the seven proteases mentioned. Only with such an arrangement could one obtain two fragments of approximately 70 and 40 kDa and no fragments of other sizes with the use of several different proteolytic enzymes.
In contrast to the above considerations, Weis et al. (25) have published the following results and conclusions. Using polyclonal antibodies raised against bovine heart transhydrogenase, Weis et al. (25) showed that when rat liver mitoplasts were treated with proteinase K, the antiserum recognized in the pellet an antigen of M, = 75,000 as the final proteolytic product of rat liver transhydrogenase ( M , = 110,000). No antigens were found in the supernatant of mitoplasts treated with proteinase K. When rat liver SMP were treated with proteinase K, a transient antigen of M , = 52,000 was released into the supernatant. The authors concluded from these results that the transhydrogenase in rat liver mitochondria is composed of a 52-kDa domain that protrudes into the matrix, a 23-koa domain (75 -52 = 23) that intercalates into the membrane, and a 35-kDa domain (110 -75 = 35) that protrudes from the cytosolic side of the inner membrane. What Weis et al. (25) did not investigate was the fate of the 35-kDa domain, and this introduces a problem with their conclusions. Assume that in mitoplasts proteinase K had fragmented the rat-liver transhydrogenase into two pieces of 75 and 35 kDa, both membrane-bound. If the antiserum used recognized the 75-kDa fragment, but not the 35-kDa piece, then the immunoblots of the mitoplasts would not show the presence of the 35-kDa piece in the membranes. As a result, it could be assumed that the 35-kDa segment had been cleaved by proteinase K from the mitoplasts and further digested into undetectable fragments. This problem does not complicate our conclusions, because our experiments do not leave any piece of the transhydrogenase unaccounted for. side. Also largely exposed on the M side is the segment D640-L554 wh ich is outlined by a dashed line. The exposed protease-sensitive loop is shown on the cytosolic (C) side connecting the presumed membrane-intercalating clusters 9 and 10. The bonds cleaved by trypsin (TRP), papain, and proteinase K, and the cysteine residue modified by Nethylmaleimide ( N E M ) are marked. For other details, see text.
In order to identify the protease-sensitive segment of the bovine transhydrogenase, which is exposed on the cytosolic side of the mitochondrial inner membrane, the following experiments were performed. Mitoplasts were treated with proteinase K until the transhydrogenase band as checked by immunoblotting had been nearly completely converted into the 71-and the 39-kDa fragments. Assay of the transhydrogenase activity of mitoplasts solubilized with Lubrol showed that, even when the 110-kDa band of intact transhydrogenase had disappeared, the particles still retained 85-90% of their original transhydrogenase activity. This prompted us to isolate the nicked enzyme. The attempt succeeded, and a highly active preparation containing >80% nicked transhydrogenase was isolated. Thereupon, the 39-kDa fragment was excised from SDS gels of the nicked enzyme and subjected to Nterminal sequencing. This sequence, AANLT, is seen in the hypothetical arrangement shown in Fig. 8, where the proteinase K cleavage site is marked. As stated above, Fig. 8 is simply an interpretation of the hydropathy profile of the central hydrophobic domain of the transhydrogenase. It is, therefore, interesting that the proteinase K cleavage site should occur in the largest extramembranous segment on the cytosolic side of this hypothetical picture. Also interesting is that one can rationalize the effects of subtilisin, thermolysin, and pronase E on this same extramembranous loop, in each case resulting in two transhydrogenase fragments with approximate M , values of 70,000 and 40,000. Fig. 8 also shows a trypsin cleavage site on the matrix side in the center of the picture, and an Nethylmaleimide-modifiable cysteine on the cytosolic side. These sites were identified previously in the purified transhydrogenase (15, 27). Their sidedness, as shown in Fig. 8, is not proven, but is simply a consequence of our interpretation of the hydropathy profile. However, in order to alter this arrangement, without changing the positions of the D540-L554 pentadecapeptide on the matrix side and the protease-sensitive loop on the cytosolic side, one would have to pull out of the membrane two of the hypothetical membrane-intercalating clusters (boxes in Fig. 8) that are located between these extramembranous loops. This is, of course, possible, except that a priori the extreme hydrophobicity of the intervening clusters is more suited to an arrangement close to that shown in Fig. 8.
The fact that the proteinase K-nicked transhydrogenase is active and can be purified presents two points of comparison with the E. coli transhydrogenase, which is composed of two subunits with M, values of 54,000 and 48,700 (8). ( a ) It is possible that what holds the two subunits of the E. coli enzyme and the two fragments of the nicked bovine transhydrogenase together is interaction between the extramembranous N-and C-terminal hydrophilic domains. Indeed, the proximity of these domains is necessitated by the fact that hydride ion transfer between NAD(H) located on the N-terminal domain and NADP(H) bound to the C-terminal domain is direct (1)(2)(3). Thus, it is not surprising that the nicked bovine enzyme should retain its transhydrogenase activity, and, like the E. coli enzyme, be purifiable. ( b ) As mentioned earlier, the E. coli transhydrogenase is highly homologous to the bovine enzyme, except that it lacks a stretch of 32 residues where the bovine pentadecapeptide marked in Fig. 8 is located. As a result, the a subunit of the E. coli enzyme contains a short stretch of hydrophobic residues at its C-terminal end, while its p-subunit starts with a long stretch of hydrophobic amino acids at its N-terminal end. By comparison, this situation is reversed in the proteinase K-nicked bovine transhydrogenase. Its 71-kDa fragment, bearing the NAD binding site, contains a long hydrophobic tail, while its 39-kDa piece, bearing the NADP binding site, carries a short hydrophobic stretch. Since proton translocation by the transhydrogenase is driven by the difference in the binding energies of substrates (NADPH + NAD) uersus products (NADP + NADH) and is mediated via protein conformation change by the hydrophobic, membraneintercalcating amino acid residues of the protein, it would be of considerable mechanistic interest to see whether the proteinase K-nicked enzyme is capable of proton translocation.