Evidence That the Transfer of Hydride Ion Equivalents between Nucleotides by Proton-translocating Transhydrogenase Is Direct*

The molecular masses of the purified, recombinant nucleotide-binding domains (domains I and III) of transhydrogenase from Rhodospirillum rubrum were determined by electrospray mass spectrometry. The values obtained, 40,273 and 21,469 Da, for domains I and III, respectively, are similar to those estimated from the amino acid sequences of the proteins. Evidently, there are no prosthetic groups or metal centers that can serve as reducible intermediates in hydride transfer between nucleotides bound to these proteins. The transient-state kinetics of hydride transfer catalyzed by mixtures of recombinant domains I and III were studied by stopped-flow spectrophotometry. The data indicate that oxidation of NADPH, bound to domain III, and reduction of acetylpyridine adenine dinucleotide (an NAD+ analogue), bound to domain I, are simultaneous and very fast. The transient-state reaction proceeds as a biphasic burst of hydride transfer before establishment of a steady state, which is limited by slow release of NADP+. Hydride transfer between the nucleotides is evidently direct. This conclusion indicates that the nicotinamide rings of the nucleotides are in close apposition during the hydride transfer reaction, and it imposes firm constraints on the mechanism by which transhydrogenation is linked to proton translocation.

Transhydrogenase is found in the inner membranes of animal mitochondria and the cytoplasmic membranes of some bacteria. It couples the transfer of hydride ion equivalents between NAD(H) and NADP(H) to the translocation of protons across the membrane. The net reaction is as follows.
For many years, the question as to whether hydride transfer between the nucleotides is direct or indirect has been a matter of controversy. It is central to our understanding of the energycoupling reactions. Transhydrogenase comprises three domains. Domains I and III protrude from the membrane (on the matrix side in mitochondria and on the cytoplasmic side in bacteria). Domain II spans the membrane. There are separate sites on the enzyme for NAD(H) and for NADP(H); the former is located on domain I, and the latter on domain III (for reviews, see Refs. [1][2][3]. The results of some early experiments on transhydrogenases from mitochondria and from Rhodospirillum rubrum were interpreted as evidence for the existence of a stable, reducedenzyme intermediate (4 -6). It was implied that a functional group on the enzyme, presumably either an amino acid residue or an unidentified prosthetic group, can serve alternately as a hydride acceptor and hydride donor. For example, where E(H) represents the reduced-enzyme intermediate. However, in subsequent work other plausible explanations were found for the earlier data (7,8). Moreover, the conclusions from steady-state kinetic analysis of transhydrogenase from various sources (7, 9 -11) have been interpreted as evidence that the reaction proceeds through the formation of a ternary complex of enzyme and nucleotide substrates. The addition of nucleotides is random, and fast, relative to the rate of a subsequent step in turnover. The reaction does not appear to take place via a substituted enzyme mechanism, and therefore the existence of a reduced enzyme intermediate, which is stable in the absence of nucleotide, is unlikely (viz. reactions exemplified by Equations 2 and 3). However, the steady-state data do not rule out the possible existence of a reduced enzyme intermediate within the ternary complex, that is a reaction of the following type.
The possibility that Cys residues in the polypeptide chain might serve as reducible intermediates in hydride transfer (see, for example, Refs. 1 and 12) has been eliminated by amino acid sequence comparisons, there are no conserved Cys residues in transhydrogenases from different species, and by the fact that complete Cys replacement has only a minimal effect on transhydrogenation activity (13). It is unlikely that other amino acid residues have redox potentials in the appropriate range to serve as intermediates in transhydrogenation between NAD(H) and NADP(H). It has sometimes been stated in the literature that transhydrogenase is devoid of prosthetic groups that might be involved in the hydride transfer pathway, although our survey indicates that studies are incomplete. 1) It was established many years ago that there is no detectable flavin fluorescence from purified preparations of the mitochondrial enzyme (14). 2) Analyses of amino acid sequences of transhydrogenases do not reveal the existence of metal-binding motifs. 3) It has been shown that concentrated solutions of the highly purified recombinant domains I and III (which together are catalytically active, see below) do not have any absorbance that might be attributable to chromophoric groups in the proteins (15,16). However, we are unaware of studies on transhydrogenase which rule out the presence of weakly, or nonabsorb-ing, prosthetic groups. By analogy with dehydrogenases, a number of possible prosthetic groups or metal ions might be envisaged to have a role as an intermediate in hydride transfer. For example, in view of the strong sequence similarity between transhydrogenase and the soluble enzyme alanine dehydrogenase (17), covalently bound pyruvate might be considered as a potential hydride acceptor.
A mixture of the isolated recombinant forms of domains I and III of R. rubrum transhydrogenase catalyzes the so-called "cyclic reaction" at a rate approaching that observed with the complete, membrane-located enzyme (16,18). Evidently the complex of domains I and III is capable of rapid rates of hydride transfer, even in the absence of membrane-spanning domain II, and thus the apparatus for hydride transfer is located entirely within the two peripheral domains. We report below on the use of electrospray mass spectrometry to determine accurately the molecular masses of these peripheral domains with a view to establishing whether or not they possess covalently bound groups that might participate in the hydride transfer reaction. We also describe an experiment in which we examine the pre-steady-state kinetics of transhydrogenation catalyzed by a mixture of recombinant domains I and III using stopped-flow spectroscopy. In contrast to steady-state kinetic analysis, this procedure can reveal the presence of reaction intermediates. There are no other published descriptions of the pre-steadystate kinetics of reactions catalyzed by transhydrogenase. As hydride donor (binding to domain III), we use NADPH, and as hydride acceptor (binding to domain I), we use the NAD ϩ analogue, AcPdAD ϩ . 1 The difference in the absorbance spectra between the reduced forms of the two nucleotides enables us to measure the rate of oxidation of NADPH, and the rate of reduction of AcPdAD ϩ , in real time.

MATERIALS AND METHODS
Recombinant forms of domain I and domain III of R. rubrum transhydrogenase were expressed in E. coli C600 from plasmids pCD1 (15) and pCD2 (16), respectively, and purified by column chromatography (15)(16)(17). The purity was confirmed by SDS-polyacrylamide gel electrophoresis (7), and the protein concentrations were determined using the microtannin assay (19). As normally prepared, the domain III protein is associated with tightly bound NADP ϩ (typically 0.1-0.4 mol⅐mol Ϫ1 ) and NADPH (about 0.5 mol⅐mol Ϫ1 ) (16). For stopped flow experiments the bound NADP ϩ was replaced by incubating domain III protein (75 M) in 10 mM (NH 4 ) 2 SO 4 , 20 mM Hepes buffer, pH 8.0, with 105 M NADPH at 4°C for 2 h (16).
Stopped-flow experiments were performed with an Applied Photophysics DX-17MV in its absorbance mode (2-mm optical path length). The instrument dead time, determined from the reaction of L-ascorbic acid with 2,6-dichlorophenolindophenol, as described (20), was 1.31 ms. The monochromator slit widths were set to 5 nm. The absorbance coefficient at 375 nm for AcPdAD ϩ reduction (corrected for the contribution from accompanying NADPH oxidation) was 6.1 mM Ϫ1 ⅐cm Ϫ1 (21) and that at 320 nm for NADPH oxidation (corrected for the contribution from accompanying AcPdAD ϩ reduction) was 3.01 mM Ϫ1 ⅐cm Ϫ1 (calculated from data given (22)). Rate constants and amplitudes were calculated from the raw data using the instrument software, making due correction for the apparatus dead time.
Samples of domains I and III were prepared for analysis by electrospray mass spectrometry by dialyzing samples (25 M) against water. The examination of domain I was performed in collaboration with Dr. C. Robinson, University of Oxford, using a Micromass Platform II and Nanoflow electrospray ionization. Data were acquired using MassLynx software. The domain III analysis was carried out in collaboration with Dr. P. R. Ashton, University of Birmingham, using a VG-Prospec in electrospray ionization mode. In both cases the carrier matrix was 20% methanol. The precision of both instruments is better than Ϯ0.01%. Molecular masses were estimated from the amino acid sequences of domains I and III using the Genetics Computer Group program (University of Wisconsin). Table I shows the results of an analysis by electrospray mass spectrometry of the molecular masses of the purifed, recombinant domains I and III of R. rubrum transhydrogenase. In both cases only a single distinct charge envelope was evident in the raw data. The similarity between the measured molecular masses of domains I and III, and the values calculated from the amino acid sequences (23), eliminates the possibility of a bound prosthetic group or metal center.

The Molecular Weights of Domains I and III of R. rubrum Transhydrogenase Determined by Electrospray Mass Spectrometry-
Direct Transfer of Hydride from NADPH, Bound to Domain III, and AcPdAD ϩ , Bound to Domain I, as Revealed by Stoppedflow Spectrophotometry-The experiment shown in Fig. 1 was set up to monitor hydride transfer from NADPH to AcPdAD ϩ in a mixture of the nucleotide-binding domains of R. rubrum transhydrogenase under pre-steady-state conditions. The first drive syringe of the stopped flow apparatus was filled with equimolar domain I and NADPH-loaded domain III proteins. The second drive syringe was filled with AcPdAD ϩ . Upon mixing, the kinetics of AcPdAD ϩ reduction (Fig. 1A) were, within the resolution of the instrument, identical to the kinetics of NADPH oxidation (Fig. 1B), thus illustrating the direct hydride transfer between nucleotides. In both cases the reaction had the character of a "burst" before the onset of steady state. The steady-state rate is extremely slow, equivalent to approximately 0.025 A⅐s Ϫ1 at 375 nm in the conditions of the experiment and therefore negligible on the time scale displayed in Fig. 1; it is profoundly limited by the release of NADP ϩ from domain III (k off Ϸ 0.03 s Ϫ1 (16)). AcPdAD ϩ ϩ E⅐NADPH 3 AcPdADH ϩ E⅐NADP ϩ (Eq. 5) A good fit to the burst was obtained by the sum of two exponentials. For AcPdAD ϩ reduction, the rate constants of the two phases were 490 and 8.6 s Ϫ1 , and for NADPH oxidation, 511 s Ϫ1 and 8.3 s Ϫ1 . These values are similar within the experimental error.
While it is clear that the product NADP ϩ remains on domain III during the burst (see above), we are not yet certain of the time scale of dissociation of the product AcPdADH. If the nucleotides do remain bound to the protein, then their absorbance coefficients might be somewhat distorted from those in free solution (for example, see Ref. 24). Nevertheless, using solution absorbance coefficients for the reduced nucleotides (see "Materials and Methods") the amplitudes of the fast and slow phases for AcPdAD ϩ reduction were 11.7 and 5.6 M, respectively, and for NADPH oxidation, they were 9.1 and 4.1 M, respectively. The minimal conclusion is that the proportion of fast and slow phases was similar for AcPdAD ϩ reduction and NADPH oxidation. The sum of the fast and slow phases corresponded to approximately 60% of the concentration of domain I and domain III in the reaction mixture (25 M), indicating that the burst corresponds to a single turnover event. Although there will be some perturbation of the redox potential of the nucleotides in the protein binding sites (25), the reaction is expected to go substantially to completion (the E o Ј values of NADP(H) and of AcPdAD(H) in solution are Ϫ320 and Ϫ248 mV, respectively (26)).  Fig. 2 shows the wavelength dependence of the fast and slow phases of the burst. Both components have a similar spectrum to that obtained by subtracting the molar absorbance spectra of NADPH and AcPdADH, though the possibility that the spectra of the reduced nucleotides on the enzyme are distorted by a few nanometers (see Ref. 24) cannot be excluded. No other intermediates can be identified in the spectrum.
Consequences for the Mechanism of Action of Transhydrogenase-It is clear from this work that hydride transfer between nucleotides bound to domains I and III of transhydrogenase proceeds directly and not by way of a reduced enzyme intermediate. This observation indicates that, during catalysis, the peripheral, nucleotide-binding domains must bring the C-4 atoms of the nicotinamide rings of NAD(H) and NADP(H) into juxtaposition. It rules out a number of models in which proton translocation was assumed to arise from chemical interactions between nucleotides and a putative hydride-accepting intermediate (reviewed (1)), and it complements our recent evidence that the hydride transfer step itself is not coupled to proton translocation (27). The indications that the coupling reactions involve changes in the nucleotide-binding energy, as suggested by a number of workers (3,28,29), are now very strong. We developed the view that proton translocation is coupled to events that are consequent on the binding and release of NADP ϩ and NADPH (28), and this was supported by the finding that NADP ϩ and NADPH release from domain III are promoted by interactions with the membrane-spanning domain II (16). The present observations, consistent with the interpretation that hydride transfer in I:III complexes is extremely rapid, and NADP ϩ release is very slow, lend further weight to the hypothesis.
Factors giving rise to the biphasic kinetics of the pre-steadystate burst remain to be investigated. Our preliminary interpretation is that the fast phase (k Ϸ 500 s Ϫ1 ) corresponds to hydride transfer in preformed domain I⅐III complexes appropriately loaded with AcPdAD ϩ and NADPH, it therefore reflects the true intramolecular rate constant for the reaction, whereas the slow phase (k Ϸ 8 s Ϫ1 at 8°C) results from protein rearrangements and/or AcPdADH release. The steady-state cyclic reaction catalyzed by the R. rubrum domain I⅐III complex (the alternate reduction/oxidation of protein-bound NADP ϩ / NADPH by NADH/AcPdAD ϩ ), with a k cat in the region of 100 -200 s Ϫ1 at 30°C (30), is expected to include contributions from both components.