Fast hydride transfer in proton-translocating transhydrogenase revealed in a rapid mixing continuous flow device.

Transhydrogenase couples the redox reaction between NAD(H) and NADP(H) to proton translocation across a membrane. Coupling is achieved through changes in protein conformation. Upon mixing, the isolated nucleotide-binding components of transhydrogenase (dI, which binds NAD(H), and dIII, which binds NADP(H)) form a catalytic dI(2).dIII(1) complex, the structure of which was recently solved by x-ray crystallography. The fluorescence from an engineered Trp in dIII changes when bound NADP(+) is reduced. Using a continuous flow device, we have measured the Trp fluorescence change when dI(2).dIII(1) complexes catalyze reduction of NADP(+) by NADH on a sub-millisecond scale. At elevated NADH concentrations, the first-order rate constant of the reaction approaches 21,200 s(-1), which is larger than that measured for redox reactions of nicotinamide nucleotides in other, soluble enzymes. Rather high concentrations of NADH are required to saturate the reaction. The deuterium isotope effect is small. Comparison with the rate of the reverse reaction (oxidation of NADPH by NAD(+)) reveals that the equilibrium constant for the redox reaction on the complex is >36. This high value might be important in ensuring high turnover rates in the intact enzyme.

Transhydrogenase is found in the inner membranes of animal mitochondria and in the cytoplasmic membranes of many bacteria. It couples the redox reaction between NAD(H) and NADP(H) to inward translocation of protons across the membrane (Eq. 1).
The enzyme provides NADPH for biosynthesis and glutathione reduction, and in mitochondria, it also helps to control flux through the tricarboxylic acid cycle (1,2). The relative simplicity of transhydrogenase, the emergence of methods for determining its rate of reaction in real time (3)(4)(5)(6), and recent high resolution structural information (7)(8)(9)(10)(11)(12) make it a good model for understanding the general principles of operation of conformationally coupled ion translocators. The enzyme is composed of three components. dI and dIII, which bind NAD(H) and NADP(H), respectively, protrude into the mitochondrial matrix (or the bacterial cytoplasm), and dII spans the membrane. The intact enzyme is effectively a dimer of two dI⅐dII⅐dIII "trimers" (13,14), though there are species variations in the way the polypeptide chains are joined. The findings that the transfer of hydride-ion equivalents between the bound nucleotides on transhydrogenase is direct (3,5) and that there is no exchange of the transferred hydride with water protons (15) together establish that coupling to proton translocation does not occur at the redox step. We proposed an NADP(H) binding change model in which NADP ϩ (or NADPH) from the solvent can only bind to (or leave from) an "open" state of the dIII component of the protein and in which the redox reaction can only take place in an "occluded" state. Association and dissociation of protons during translocation, gated by the redox state of the NADP(H), drives the protein between the open and occluded states (16,7). Recent observations on pronounced structural asymmetries in transhydrogenase suggest that the two dI⅐dII⅐dIII trimers of the complete enzyme undergo reciprocating alternations of conformation during turnover (12). Thus, as the dIII in one trimer enters the open state to permit product release and substrate binding, the other enters the occluded state to permit hydride transfer; the two trimers run 180 o out of phase. In general, a mechanism of coupling involving changes in NADP(H) binding is also favored by other authors (17)(18)(19).
Isolated recombinant dI and dIII readily form a complex. That from Rhodospirillum rubrum transhydrogenase has been studied in most detail (3-6, 12, 20) (see also Refs. 21 and 22). It is a dI 2 ⅐dIII 1 heterotrimer in both the crystalline state and in solution (12,23). The two characteristic properties, (a) a capacity to catalyze a rapid redox reaction between NAD(H) and NADP(H) (or their analogues) and (b) an extremely slow rate of release of bound NADP ϩ and NADPH, suggest that the isolated complex adopts a conformation similar to that of the occluded state in the intact enzyme. These properties of the dI 2 ⅐dIII 1 complex can be partly rationalized in terms of features seen in the x-ray structure of the protein. The redox reaction between NADP(H) and AcPdAD(H) 1 (an NAD(H) analogue) on the dI 2 ⅐dIII 1 complex was studied by stopped-flow spectrophotometry (3)(4)(5). A rapid, single turnover burst of hydride transfer precedes the slow steady-state reaction that is limited by product NADP(H) release. A E155W mutant of dIII was subsequently isolated to study these processes with physiological nucleotides from changes in protein fluorescence, but the forward reaction was too fast to measure by stopped-flow (6). In this report we describe measurements of the forward reaction with physiological nucleotides using a newly developed continuous flow device. The reaction is faster than predicted from the stopped-flow experiments, and thus the on-enzyme equilibrium constant is greater than expected.

MATERIALS AND METHODS
The expression and purification of the isolated wild-type dI (and both the wild-type and the E155W mutant of dIII) from R. rubrum transhydrogenase were carried out as described (3)(4)(5)(6). The dIII component was prepared in either its NADP ϩ or NADPH form, as required. Protein concentrations were determined using the microtannin assay (24). In both stopped-flow and continuous flow experiments, the dI and dIII proteins were premixed in a 2:1 molar ratio to give the dI 2 ⅐dIII 1 complex and loaded into in the appropriate drive syringe. Given the low K d and rate of dissociation of the complex, this ensures that the redox reaction proceeds as a monoexponential reaction (3)(4)(5)(6). Experiments were performed in 10 mM (NH 4 ) 2 SO 4 and 20 mM Hepes, pH 8.0, at 20°C.
Sub-millisecond kinetic measurements were performed in a homebuilt continuous flow instrument based on that described by Shastry et al. (25). Our instrument uses their capillary mixer coupled to a 0.25 ϫ 0.25 ϫ 20-mm quartz flow cell (Hellma Cells, Inc.). The dI 2 ⅐dIII 1 -NADP ϩ solution and the NADH solution were delivered to the mixer in a 1:10 volumetric ratio using a stepper-motor drive unit developed by Hi-Tech Ltd. (Salisbury, United Kingdom). A 100-watt Hg arc lamp (Osram, Germany), a monochromator, and a cylindrical lens were used to excite fluorescence at 291 nm (4-nm bandwidth), illuminating the 20-mm axis of the flow cell along a 250-m path length. Fluorescence emission (320 -400 nm) was selected with a combination of WG320 and UG11 filters (Schott) and was recorded by imaging the flow (total length of image, 16.5 mm) using a digital camera system (Micromax, Roper Scientific) equipped with a SenSys:1401E CCD chip with an array of 1317 ϫ 1035 pixels. Images were converted into intensity profiles along the direction of flow by averaging the rows of pixels across the width of the channel. Relative fluorescence intensities were determined by comparison with images derived from experiments in which NADP ϩ -loaded dI 2 ⅐dIII 1 complex was mixed with buffer in the absence of NADH. Kinetic traces were calculated as described before (25) and are an average of two independent experiments of 5.2-s acquisition time at a flow rate of 0.83 ml s Ϫ1 using a dI 2 ⅐dIII 1 concentration in the observation flow cell of 5 M. Observed rate constants were obtained by nonlinear least squares analysis. The dead time of the instrument, measured by quenching tryptophan fluorescence with N-bromosuccinimide (26), was 150 s.
Stopped flow experiments were performed on an Applied Photophysics DX-17MV operating in its fluorescence mode with a 280-nm excitation light and Ͼ305-nm emission (selected with a WG305 cut-off filter) and with both monochromator slits set to a bandwidth of 9.3 nm. In each experiment, seven data sets were averaged and analyzed using the instrument software. For further details, see Refs. 3-6.
Deuterated NAD 2 H (labeled in the NC4A position) and NADP 2 H (labeled in the NC4B position) were prepared from uniformly deuterated ethanol using yeast alcohol dehydrogenase and from D-glucose-1-d using yeast hexokinase plus yeast glucose-6-phosphate dehydrogenase, respectively, as described (4, 27-29). The preparative enzymes were removed by centrifugation through Vivaspin filters (5-kDa cut-off), and the deuterated nucleotides were purified by ion-exchange chromatography. For both nucleotides, A 260 /A 340 Ͻ 2.25 (where Յ2.3 is analyti-cally pure). In all measurements of the kinetic isotope effect, control experiments were performed with undeuterated reduced nucleotides prepared by similar procedures.

RESULTS
The isolated dIII component of R. rubrum transhydrogenase, as prepared in our laboratory, is associated with tightly bound nucleotide, either NADP ϩ or NADPH (3)(4)(5)(6)20). The E155W mutant of the dIII protein has a single tryptophan residue. The mutation does not significantly affect either the NADP(H) binding properties or the catalytic properties of the protein (6,30). However, the fluorescence emission of the introduced Trp residue is sensitive to the redox state of the bound nucleotide; it is about 25% less intense when the nucleotide is reduced. For the equivalent fluorescence change in Escherichia coli dIII, it was proposed that resonance-energy transfer is responsible for the difference, but effects due to conformational changes cannot be eliminated (22). In the experiments shown in Fig. 1, dIII⅐E155W in its NADP ϩ form was premixed with R. rubrum dI (nucleotide-free) at protein concentrations that gave the dI 2 ⅐dIII 1 heterotrimer. The change in Trp-155 fluorescence was recorded after the rapid mixing of this solution with NADH in the continuous flow instrument. In each trace, the fluorescence rapidly decreased with approximately monoexponential kinetics as the bound NADP ϩ was reduced. Note that product NADPH is released only extremely slowly from the protein (k off Ͻ 3 ϫ 10 Ϫ4 s Ϫ1 (23)), and therefore the redox reaction is limited to only one turnover (4). The dependence of the apparent firstorder rate constant of the fluorescence change on the concentration of NADH is shown in Fig. 2. At saturation, this rate constant approached a very high value, k app ϭ 21,200 s Ϫ1 , close to the limit of resolution of the instrument.
We can now go some way toward estimating the microscopic rate constants and the equilibrium constant for the redox reaction involving the physiological substrates on transhydrogenase. Scheme 1 describes the forward reaction (above), and Scheme 2 describes the corresponding reverse reaction; that is the oxidation of dI 2 ⅐dIII 1 ⅐NADPH by NAD ϩ , as studied earlier in the stopped-flow instrument (6). Because the k app for the reaction in Scheme 1 (k app(f) ϭ 21,200 s Ϫ1 ) is much greater than that for Scheme 2 (k app(r) ϭ 590 s Ϫ1 (6)), the former goes almost to completion and k app(f) Ϸ k f (the first-order rate constant of the redox step in the forward direction). However, the reaction in Scheme 2 is reversible, and k app(r) has contributions from both k f and k r : in the limit k app(r) ϭ k f ϩ k r (for example, see Ref.  31). Therefore, k r Ͻ 590 s Ϫ1 , and the equilibrium constant of the redox reaction on the protein is K eq ϭ k f /k r Ͼ 36. In fact earlier experiments show that the reaction of Scheme 2 is readily reversible (6), and thus k r is quite considerably less than 590 s Ϫ1 and K eq is quite considerably greater than 36.
Isolated dI 2 binds two NADH/mole of dimer, both with a K d ϳ20 M (32). However, the dI 2 ⅐dIII 1 complex (in its NADPH form) binds one NADH with a K d ϳ20 M and one with a K d ϳ300 M (23). It was concluded that the dI protomer whose interdomain cleft is associated with dIII⅐NADPH (as indicated by the crystal structure (12)) has a decreased affinity for NADH, whereas the dI protomer with the unoccupied cleft has an unchanged affinity for NADH. In Fig. 2, the concentration of NADH giving the half maximal value of k app was ϳ500 M, similar to the K d of the low affinity site of the dI 2 ⅐dIII 1 complex. This suggests that the nucleotide-binding reactions are fast enough to reach equilibrium and that the high K d for NADH is the functionally relevant parameter in the steps leading to hydride transfer in the dI 2 ⅐dIII 1 complex. Another more convoluted explanation is that NADH binding is slow relative to the redox reaction and that the high K d measured for NADH on dI in the dI 2 ⅐dIII 1 complex (23) results from the fact that, in those experiments, NADPH (rather than NADP ϩ ) was bound on dIII (the increased K d might represent a device for minimizing the formation of catalytically dead-end complexes). That is, the effective K d during catalysis is 20 M.
There is only a small primary deuterium isotope effect (KIE Ϸ 2) on the rate constant for reduction of AcPdAD ϩ by NADPH catalyzed by dI 2 ⅐dIII 1 complexes at 25°C (4), but the temperature dependence displays non-classical behavior (as defined in Ref. 33), and this might indicate a contribution from quantum mechanical tunneling at the hydride-transfer step. There is the usual caveat in experiments of this kind, that it is difficult to eliminate the possibility that kinetic complexities cause the unusual temperature dependences (see Ref. 33). A number of enzymes have been shown to have KIEs substantially larger than the limit of ϳ7 set by transition-state theory, and this has also been taken as evidence for hydrogen tunneling (34,35). In view of the extremely large rate constant of the reaction (see above), it was of interest to look for a KIE on the reduction of NADP ϩ by NADH on dI 2 ⅐dIII 1 complexes. Experiments were performed at high concentrations of NADH and NAD 2 H to saturate the binding reaction. Although this meant that we were working close to the limit of resolution of the continuous flow instrument, it was clear from the data that any KIE could not have been greater than 2-3-fold. In an alternative strategy, we compared rates of reduction of NAD ϩ by NADPH and by NADP 2 H; this reaction is slow enough to measure by stopped-flow instrumentation (6). The KIE on the k app was 2.6 ( Fig. 3). Because this reaction very readily reverses, there is a much larger contribution from k f to the measured k app than that from k r (see above). Thus, the measured KIE largely corresponds to an effect on k f . It is a significant value, confirming the view that the reaction is (at least partly) limited by hydride transfer, but it is well within the limit set by semiclassical analysis. DISCUSSION A reliable lower limit of the equilibrium constant for the physiological nucleotides at the redox step of transhydrogenase (K eq Ͼ 36) was determined from the data described above. In aqueous solution, the standard redox potential of NAD(H) is On the other hand, the elevated equilibrium constant of the redox reaction (a reflection of how the nucleotide binding energies are expressed) will have a kinetic effect. In the framework of the model, the elevated equilibrium constant will increase the turnover rate of the enzyme in the forward direction. Thus, the redox reaction takes place in the occluded state.
Conversion from the open state to the occluded state precedes and re-conversion to the open state follows the redox reaction. These conversions are driven by appropriately switched protonation/deprotonation reactions associated with proton translocation and will be limited by the proton electrochemical potentials of the bacterial cytoplasm and periplasm. Through an elevation of the equilibrium constant of the redox reaction, the concentration of the occluded intermediate bearing NADH/ NADP ϩ will be lowered, and the concentration of that bearing NAD ϩ /NADPH will be raised, increasing the rates of interconversion of the open and occluded states.
The apparent K d for NADH during hydride transfer in dI 2 ⅐dIII 1 complexes is rather high (ϳ500 M, Fig. 2) relative to the expected concentration of NADH in the bacterial cell and compared with the K m for NADH in the intact enzyme in membrane preparations (ϳ10 M). This might indicate that during turnover of the complete enzyme, NADH binds predominantly to the dI associated with open dIII (K d for NADH, ϳ30 M). As this dIII (with bound NADP ϩ ) is converted to the occluded state, its partner dI-NADH site is shifted to the high K d form (ϳ300 M), but NADH does not significantly dissociate because the hydride transfer step is very fast and proceeds almost to completion (see above). The different conformations of NAD ϩ seen in each of the four polypeptides of the asymmetric unit of the crystal structure of isolated dI (11) probably reflect these events in the complete enzyme. NAD(H) binding changes are not directly coupled to proton translocation (and in isolated dI 2 ⅐dIII 1 complexes, NADH binding has to proceed through the "wrong" protein conformation), but they are advantageous (a) to keep the nicotinamide rings apart to prevent hydride transfer in the open state (11,12) and (b) to permit the relative stabilization of NAD ϩ during hydride transfer in the occluded state.
The k f for the reduction of NADP ϩ by NADH on dI 2 ⅐dIII 1 complexes of R. rubrum transhydrogenase is very large (ϳ21,200 s Ϫ1 ); its measurement required the use of a novel mixing device and continuous flow observation (25). It is much greater than rate constants reported for redox reactions catalyzed by other enzymes utilizing nicotinamide nucleotide coenzymes, including the soluble dehydrogenases and flavoproteins; typically, these are rather less than 1000 s Ϫ1 . It is not clear what factors are responsible for the very high rate. A K eq of 36 does not represent a very large driving force (⌬G ϭ Ϫ8.9 kJ mol Ϫ1 ), although judging from comparisons of rate constants measured for NAD(H)7NADP(H) with those for AcPdAD(H)7NADP(H), the driving force does have a large effect on the reaction rate (6). At this stage a tunneling mechanism cannot be ruled out but as yet, there is no compelling evidence to indicate non-classical behavior.