The membrane-peripheral subunits of transhydrogenase from Entamoeba histolytica are functional only when dimerized.

Unlike their bacterial and mammalian counterparts, the NADP(H)- and NAD(H)-binding components of proton-translocating transhydrogenase from the protozoan parasite Entamoeba histolytica (denoted ehdIII and ehdI, respectively) are tethered by a polypeptide linker. The recombinant tethered fragment, ehdIII-ehdI, was prepared without its membrane-spanning dII component. Dimers of ehdIII-ehdI catalyzed transhydrogenation, but monomers were inactive. The addition of ehdIII to ehdIII-ehdI monomers did not lead to an increase in the rate of transhydrogenation, showing that this inactivity is not the result of an unfavorable topology introduced by the linker. The addition of a bacterial dI to ehdIII-ehdI led to an increase in the rate of transhydrogenation, showing that the linker is flexible. A hybrid protein in which ehdIII is tethered to the bacterial dI (denoted ehdIII-rrdI) more readily formed active dimers. Data from small angle x-ray scattering by the hybrid dimers were fitted to models derived from the high-resolution crystal structure of the bacterial dI(2)dIII(1) complex (Cotton, N. P. J., White, S. A., Peake, S. J., McSweeney, S., and Jackson, J. B. (2001) Structure 9, 165-T176). The results show that the ehdIII-rrdI dimer is asymmetric; one dIII associates with dI, as in the bacterial complex, but the other is displaced. The results provide evidence for the alternating site, binding change model for proton translocation by intact transhydrogenase.

The reaction is driven from left to right by the proton electrochemical gradient generated by the respiratory (or sometimes photosynthetic) electron transport chain. Thus, transhydrogenase is important as a source of NADPH for biosynthesis and glutathione reduction (for protection against free-radical damage) and in the regulation of flux through the tricarboxylic acid cycle (5,6). Genes encoding transhydrogenase have also been found in several protozoan parasites. In Plasmodium falciparum the enzyme is probably located in the mitochondrial membrane, but in Eimeria tenella it is thought to be associated with so-called "refractile bodies" (7) and in Entamoeba histolytica, with "mitosomes" (8), otherwise known as "cryptons" (9). The functions of the Ei. tenella refractile bodies and the En. histolytica mitosomes and the nature of the partner proteins that are coupled to the transhydrogenase in the local chemiosmotic proton circuits are not known. The partner proteins are probably not enzymes normally associated with oxidative phosphorylation (10). All proton-translocating transhydrogenases seem to have a similar structural organization. There is a dII component, which spans the membrane, and dI and dIII components, which protrude from the membrane on the matrix side in mitochondria and on the cytoplasmic side in bacteria. The dI component binds NAD ϩ /NADH, and dIII binds NADP ϩ /NADPH. The manner in which the three components relate to the polypeptide composition varies in different species (Fig. 1). In animal mitochondria, all three transhydrogenase components are on the same polypeptide chain; it runs dI-dII-dIII, N terminus to C terminus. In bacteria, gene sequences show that the dII component is always separated into dIIa and dIIb. In Escherichia coli transhydrogenase (and in the enzyme from some other bacterial species) an ␣ polypeptide comprises dI plus dIIa, and a ␤ polypeptide comprises dIIb plus dIII. In transhydrogenase from Rhodospirillum rubrum (and some other bacterial species) there are three polypeptides; PntAA comprises dI, PntAB comprises dIIa, and PntB comprises dIIb plus dIII. The predicted order of components in the single polypeptide chain of transhydrogenase from the protozoan parasites (from the gene sequences) is dIIb-dIII-dI-dIIa. Thus, the N terminus of the protein corresponds to the N terminus of the bacterial ␤ polypeptide (PntB). An extra segment (38 amino acid residues in both Ei. tenella and En. histolytica) is predicted between dIII and dI and probably serves as a linker.
We recently expressed the dIII-linker-dI fragment of En. histolytica transhydrogenase from cloned DNA in cells of E. coli and purified the protein (henceforth denoted ehdIII-ehdI) 1 in substantial yields (11). Like complexes formed from mixtures of isolated dI and dIII from the transhydrogenases of other species (12)(13)(14)(15)(16), the tethered ehdIII-ehdI catalyzed so-called "cyclic" transhydrogenation at substantial rates and "reverse" transhydrogenation at much lower rates. This shows that, as in other complexes, hydride transfer between nucleotides bound to the dI and dIII components of ehdIII-ehdI is fast, and the release of NADP(H) from dIII is slow (12).
Cross-linking and hydrodynamic studies show that the mammalian transhydrogenase is a homodimer (17,18) and that the E. coli enzyme is a comparable ␣ 2 ␤ 2 tetramer (19); both enzymes are essentially "dimers" of the two "trimeric" units of dI, dII, and dIII. The high-resolution crystal structures of isolated dI (20) and the complex of dI and dIII (21) from R. rubrum transhydrogenase clearly indicate that the intact enzyme from this species has a similar organization. The structure of the R. rubrum complex and its hydrodynamic (22), kinetic (23), and NMR (24) properties reveal another important feature, i.e. only one dIII polypeptide binds to the two polypeptides of the dI dimer. Because hydride transfer across the single dI/dIII interface of the complex is extremely rapid (25), it was proposed that in the complete enzyme the two dI/dIII interfaces must be alternately brought together during turnover. The proposal is supported by earlier studies indicating "half of the sites" reactivity for bovine transhydrogenase (26,27) and by another investigation in which inactivation of the bovine enzyme by high concentrations of Triton X-100 was attributed to the dissociation of dimers into monomers (18).
In this report, we use the tethered complex of ehdIII-ehdI and a tethered hybrid complex of En. histolytica dIII and R. rubrum dI to show the importance of dimeric interactions between dI units in hydride transfer. The unique properties of the parasite transhydrogenase provide further evidence for site alternation during transhydrogenase turnover and begin to reveal the structural basis for this process.
DNA coding for isolated ehdI (amino acid residues Leu 565 -Lys 931 ) (29) was amplified by PCR using oligonucleotide primers made by Alta Bioscience and a construct bearing the complete En. histolytica transhydrogenase gene generously provided by Dr. C. G. Clarke of the London School of Hygiene and Tropical Medicine. The DNA sequence of the "sense" primer was chosen to result in Leu-565 being changed to the N-terminal Met residue of the ehdI protein. DNA coding for ehdIII (amino acids Met 326 -Lys 527 ) was similarly amplified. The PCR products for ehdI and ehdIII were (separately) ligated into pPCR-Script TM SK(ϩ) using the PCR-Script Amp kit from Stratagene. After confirming that the DNA fragments were free of polymerase errors, they were cloned into pET21(d) from Novagen to give pCJW1 (for ehdI) and pCJW4 (for ehdIII). The constructs were transformed into E. coli BL21(DE3) (30). Bacterial cells were then grown to mid-log phase at 25°C in NZCYM medium (see Ref. 11) and induced with 1 mM isopropyl-␤-D-thiogalactoside for 14 h. Growth at higher temperatures led to the aggregation of proteins into insoluble inclusion bodies from which active proteins could not be recovered by conventional refolding techniques. Cells bearing the plasmid coding for the dI construct grew very slowly, and relatively small amounts of recombinant protein were obtained. Cells were harvested by centrifugation, washed in a buffer containing 50 mM Tris-HCl, pH 8.0, 1 mM EDTA, 2 mM dithiothreitol, and stored at Ϫ20°C.
As a preliminary step in the generation of DNA coding for the tethered hybrid ehdIII-rrdI, the internal NdeI site was removed from plasmid pCJW3 by site-directed mutagenesis of T to C at position 1360 using the Stratagene QuikChange kit and oligonucleotide primers from Alta Bioscience. This mutation does not change the amino acid sequence of the transcribed and translated gene. A new NdeI site was then introduced at the 5Ј-end of the DNA coding for the ehdI by mutagenesis of 1793G3 C, 1796T3 A, and 1798A3 G. In the final construct this change results in the substitution of an Asp residue for a His at the extreme C terminus of the linker. DNA sequencing established that no errors were introduced by the polymerase. The NdeI-BamHI fragment (coding for En. histolytica dI) from this construct was replaced by the NdeI-BamHI fragment (coding for R. rubrum dI) from pCD1 using standard cloning procedures. The final construct, designated pCJW7, was transformed into E. coli BL21(DE3), and cells were grown, induced, harvested, washed, and stored as described above.
Protein Purification-Tethered ehdIII-ehdI and isolated rrdI and rrdIII were purified by column chromatography as described (11,12,28). The purification protocols for ehdIII and ehdIII-rrdI were similar to those for ehdIII-ehdI except that, in the case of ehdIII-rrdI, the protein solution was supplemented with 5% (rather than 10%) ammonium sulfate prior to the phenyl-Sepharose HP column, and the reverse gradient was from 5 to 0% saturation. To purify the ehdI protein, a substantial modification of our standard procedures was required. The thawed bacterial extract from 3.2 liters of culture was applied to a 5 ϫ 30-cm column of Q-Sepharose Fast Flow (Amersham Biosciences) preequilibrated with 20 mM Tris-HCl, pH 8.0, and 2 mM dithiothreitol (buffer A). The column was developed with a gradient of 0 -0.4 M NaCl in buffer A. Active fractions were pooled and then brought to 40% saturation with ammonium sulfate and incubated overnight at 4°C. Precipitate was removed by centrifugation, and the supernatant, containing ehdI, was applied to a 2.6 ϫ 25-cm column of butyl-Toyopearl (TosoHaas) pre-equilibrated in buffer A supplemented with 40% saturated ammonium sulfate. The column was developed with a reverse gradient of ammonium sulfate (40 -0% saturation), and recombinant protein was eluted in a final wash of buffer A. Active fractions were pooled, concentrated to ϳ3 ml in Vivascience centrifugal filters (10-kDa cutoff), and subjected to a final purification step on a 2.6 ϫ 62-cm column of Hiload Superdex 200 in buffer A. The protein was stored in 25% glycerol at Ϫ80°C in 2-ml thin-walled cryovials (Nalgene).
During chromatography, proteins were assayed for transhydrogenation activity (see below) either alone or in the presence of partner nucleotide-binding proteins (dI or dIII) added in excess. Protein concentrations were determined by the microtannin procedure (31); all values are given as polypeptide monomers. The final preparations were routinely examined by SDS-PAGE; all were Ͼ95% pure according to staining intensity with Page Blue 83. The amounts of nucleotide associated with the purified recombinant proteins were determined by enzymic assay after appropriate denaturation as described (12).
Analytical Procedures-Prior to all experiments the stored proteins were thawed on ice, concentrated in Vivascience filters (5-kDa cutoff for dIII proteins, 10 kDa for all other proteins), and washed with reaction buffer (see the legends to Figs. 2-6) supplemented with 2 mM dithiothreitol and 4 M NADP ϩ . Reverse transhydrogenation was measured from R. rubrum; ehdIII-rrdI, a hybrid protein in which ehdIII is tethered to the bacterial dI (unless otherwise indicated, the abbreviations above are not intended to represent the oligomeric state of the protein); AcPdAD ϩ , the oxidized form of acetylpyridine adenine dinucleotide. SAXS, small angle x-ray scattering; MOPS, 4-morpholinepropanesulfonic acid. as the reduction of the NAD ϩ analogue, AcPdAD ϩ , by NADPH using the absorbance coefficient 6.07 mM Ϫ1 cm Ϫ1 at 375 nm (32). Cyclic transhydrogenation is the reduction of AcPdAD ϩ by NADH in the presence of NADP ϩ or NADPH (33). It corresponds to the reduction of NADP ϩ by NADH followed by the oxidation of NADPH by AcPdAD ϩ ; the reaction can occur without NADP(H) leaving the enzyme. Steady-state rates of transhydrogenation were measured (at 25°C) in a PerkinElmer Lambda 16 dual wavelength spectrophotometer. Pre-steady-state kinetics were measured in an Applied Photophysics DX-17MV stoppedflow spectrophotometer in its absorbance mode at 25°C. The optical path length was 2 mm, and the slits were set to give 5-nm halfbandwidth. The mixing dead time was 1.31 ms (34).
Gel exclusion chromatography was carried out using a 2.6 ϫ 62-cm column of Hiload Superdex 200 calibrated with the following standards: blue dextran (1 mg ml Ϫ1 ) to give the void volume (V 0 ), alcohol dehydrogenase (5 mg ml Ϫ1 ), bovine serum albumin (1 mg ml Ϫ1 ), carbonic anhydrase (3 mg ml Ϫ1 ), and cytochrome c (1 mg ml Ϫ1 ). The sample and standards were separately applied to the column in a solution containing 20 mM Tris-HCl, pH 8.0, 150 mM NaCl, 5% glycerol, 2 mM dithiothreitol, and 4 M NADP ϩ . The molecular mass was calculated from a plot of log M r versus V e /V 0 , where V e is the volume required to elute the respective protein.
Solution x-ray scattering data were collected following standard procedures on the X33 camera of the European Molecular Biology Laboratory at the Deutsches Elektronen Synchrotron in Hamburg, Germany. Scattering from the ehdIII-rrdI protein (271 M) in a buffer containing 20 mM Hepes, pH 8.0, 10 mM (NH 4 ) 2 SO 4 , 2 mM dithiothreitol, and 4 M NADP ϩ was measured with a sample to a detector distance of 2.5 m. At a wavelength () of 1.5 Å, the data covered the momentum transfer range 0.014 Ͻ s Ͻ 0.251 Å Ϫ1 , where s ϭ (4sin)/ and is half the scattering angle. For molecular mass calibration, bovine serum albumin was used as a standard. Data were normalized to the intensity of the incident beam and corrected for the detector response using the program OTOKO (35). The buffer data were subtracted, and the results scaled for concentration using the program PRIMUS. 2 The maximum diameter of the particle, the forward scattering value I(0), the radius of gyration R g , and the pair-distribution function p(r) were calculated from the scattering curve using the indirect Fourier transform program GNOM (36) with the data truncated to s ϭ 0.1918 Å Ϫ1 . Low-resolution dummy atom models were calculated using the program DAMMIN (37) operating in "slow" mode with a spherical initial search space. The models generated by DAMMIN were subjected to an interaxial alignment step with respect to the crystal structure of the R. rubrum dI 2 dIII 1 complex (PDB 1HZZ) using the program SUPCOMB (38) followed by a round of manual refinement to permit the inclusion of a second dIII polypeptide using the program MASSHA (39). Experimental curves were compared with the scattering curves predicted from the crystal structure of the R. rubrum dI 2 dIII 1 complex and the truncated/extended forms thereof (see below) using the program CRYSOL (40).

In a Mixture, Isolated ehdI and ehdIII Form Catalytically Competent Complexes, but Hybrid Complexes of ehdIII and R. rubrum dI Are Considerably More Active-Recombinant
forms of isolated dI and dIII from En. histolytica transhydrogenase were expressed in E. coli and purified (see "Experimental Procedures"). Analysis by SDS-PAGE indicated that the monomer molecular masses of ehdI and ehdIII were ϳ40 and ϳ28 kDa, respectively (compare 39.7 and 22.0 kDa predicted from amino acid sequence data). Isolated recombinant dIII from transhydrogenases of other organisms also behaves anomalously on SDS-PAGE, and the reasons for this are not clear (12,16). Gel exclusion chromatography under non-denaturing conditions indicated that, at 450 M, recombinant ehdIII is probably monomeric (it eluted in a sharp peak at 29.4 kDa; see "Experimental Procedures"), but the recombinant ehdI protein eluted in a broad peak with molecular mass in the range 46 -64 kDa. As found for other transhydrogenase dIII components (12,14,16), purified ehdIII was associated with tightly bound nucleotide (ϳ0.5 mol NADP ϩ and 0.4 mol NADPH mol Ϫ1 dIII).
Separately, ehdI and ehdIII were incapable of transhydroge-nation, but together in simple mixtures they catalyzed significant rates of reaction. The dependence of the rate of cyclic transhydrogenation on the concentration of ehdI at fixed ehdIII is shown in Fig. 2a. The reaction showed no sign of saturation even at Ͼ4 M ehdI, suggesting that the affinity between the components is quite weak, and the rates were low in comparison with those achieved with tethered ehdI-ehdIII (11). A mixture of isolated R. rubrum dI (rrdI) and En. histolytica dIII gave rise to a very active complex. In the experiments shown in Fig. 2b, the ehdIII was saturated by only moderate concentrations of rrdI (apparent K d Ϸ 500 nM), and the maximum rates were similar to those obtained with mixtures of rrdI and rrdIII (12). In contrast, mixtures of ehdI (for example at 975 nM) and rrdIII (2 M) gave rise to only very low rates of transhydrogenation (0.2 mol AcPdADH mol Ϫ1 ehdIII⅐min Ϫ1 ) (results not shown).
It is concluded that, in its isolated form, ehdIII is a well behaved protein able to bind a partner dI, resulting in a complex capable of good rates of hydride transfer. On the other hand, during expression and/or isolation, ehdI loses integrity (e.g. relative to that in ehdIII-ehdI). An apparently similar situation was noted with the (non-tethered) nucleotide-binding domains of E. coli transhydrogenase; a mixture of ecdI ϩ ecdIII had low activity, and ecdI ϩ rrdIII was inactive, whereas mix-2 P. A. Konarev and D. I. Svergun, unpublished data. tures of rrdI ϩ ecdIII were very active (with rates in the same range as those obtained with rrdI ϩ rrdIII) (14,15). The successful expression of mammalian dI has still not been reported. Thus, the exceptional dI is that from R. rubrum transhydrogenase because, in its isolated state, it maintains the capacity for good reconstitution with its own dIII and those from En. histolytica, E. coli, bovine, and human sources. This could be a consequence of the fact that only in intact R. rubrum transhydrogenase does the dI protein exist as a separate polypeptide (Fig. 1).
Oligomeric Interactions Are Required for Catalytic Activity of Tethered ehdIII-ehdI-At concentrations Ͼ100 M, the mobility of tethered ehdIII-ehdI during gel-exclusion chromatography under non-denaturing conditions at 25°C indicated a molecular mass (155 kDa) somewhat greater than that expected of a dimer (the molecular mass of the monomer is 65.8 kDa) (11), although at 8°C the protein repeatedly ran with a molecular mass of 110 kDa (results not shown). Purified ehdIII-ehdI was associated with a tightly bound nucleotide (ϳ0.5 mol NADP ϩ plus NADPH mol Ϫ1 protein; the proportions of NADP ϩ and NADPH varied slightly with the preparation, but no significant NAD(H) was detected) (data not shown).
Earlier studies showed that the characteristics of the transhydrogenation reaction catalyzed by ehdIII-ehdI (for example, in relation to nucleotide concentration) were similar to those observed in other dI ϩ dIII systems (11) . Fig. 3, however, shows that the specific rates of reverse and cyclic transhydrogenation catalyzed by ehdIII-ehdI were unexpectedly very dependent on protein concentration. The activity tended toward zero at low concentrations and was half-maximal at ϳ1 M ehdIII-ehdI for the reverse reaction and ϳ0.5 M for cyclic. Note that the cyclic reaction (limited by the rate of hydride transfer and by nucleotide occupation of the dI site) was much faster than reverse (limited by the slow release of NADP ϩ from the dIII site) as described and explained for complexes from the transhydrogenase of other species (12,14,41). The results are taken as an indication that ehdIII-ehdI forms catalytically active dimers from inactive monomers with a K d in the order of 1 M. Two possibilities are envisaged. (1) For topological reasons, a dIII and a dI tethered together on the same (monomeric) polypeptide chain cannot approach one another to give a catalytically active complex. In dimers, the dIII of one polypeptide interacts with the dI of another to give the active sites for hydride transfer. (2) The interface between monomeric units in a dimer is required to provide the necessary protein conformation (or the necessary movements within the polypeptides) for hydride transfer.
The experiments shown in Fig. 4 were designed to discriminate between these two hypotheses. Because ehdIII can readily form hydride transfer sites with ehdI (both in mixtures of the isolated proteins, and "intra-complex" sites within tethered ehdIII-ehdI; see above), it was reasoned that if the first hypothesis were correct; the addition of isolated ehdIII to a dilute solution of ehdIII-ehdI (where the latter is predominantly monomeric) would lead to the formation of extra hydride-transfer sites. We should then observe a pronounced increase in the specific rate of transhydrogenation up to that observed with dimers of ehdIII-ehdI. For example, the addition of ehdIII to 127 nM ehdIII-ehdI should stimulate the cyclic reaction from ϳ20 to ϳ200 min Ϫ1 (Fig. 3). In fact, only a very small stimulation was observed (to ϳ25 min Ϫ1 in Fig. 4), and this is taken as firm evidence against the first hypothesis and in favor of the second. The small stimulation that was detected is attributed to an increase in the number of hydride transfer sites due to enhanced association of ehdIII with the low concentration of ehdIII-ehdI dimers in the 127-nM solution (about 10% of the total protein).
The addition of isolated rrdI to a solution of ehdIII-ehdI greatly enhanced its capacity for both reverse (Fig. 5a) and cyclic transhydrogenation (Fig. 5b). The maximal rate of the cyclic reaction in these experiments and the concentration of rrdI needed to reach the half-maximal rate were similar to those observed in the titration of isolated ehdIII with rrdI (Fig.  2b). The results reveal the flexibility of the linker in the En. histolytica complex; the dIII component must be able to move apart from its own dI to interact functionally with the R. rubrum protein and give rise to a highly active rrdI/ehdIII hydride transfer site. Interestingly, for both the cyclic and reverse reactions the degree of stimulation produced by rrdI was more pronounced at low concentrations of ehdIII-ehdI (where the En. histolytica protein is predominantly monomeric) than at high concentrations (where the En. histolytica protein is predominantly dimeric). Thus, the specific rate of transhydrogenation at saturating rrdI was greater at a low ehdIII-ehdI concentration than at high concentration (Fig. 5). This indicates that, when ehdIII-ehdI is in its dimeric form (with moderately active hydride-transfer sites), less of the dIII component is available to form highly active sites with the added rrdI; the dIII is more tightly associated with dI in the ehdIII-ehdI dimer than in the monomer. Although we think that this con- clusion is sound, it is necessary to be cautious in the interpretation of results shown in Fig. 5 because there is a possibility that the proteins can "scramble" on the time scale of the experiment, i.e. mixed dimers of (ehdIII-ehdI) 1 (rrdI) 1 might be formed in significant quantities. The degree of scrambling will depend in a complex way upon the relative stability constants of the homogeneous dimers ((rrdI) 2 and (ehdIII-ehdI) 2 ) and the mixed dimers, and upon the rate constants for dissociation and reassociation.
The addition of isolated ehdI also led to an increase in the transhydrogenation activity of ehdIII-ehdI but, as expected from the experiments described in the previous section, the effect was much less pronounced than with R. rubrum dI. With 200 nM ehdIII-ehdI (predominantly monomeric) the increase was approximately linear with ehdI concentrations up to 5 M (indicating a low affinity interaction), and the rate enhancement of cyclic transhydrogenation was from ϳ10 min Ϫ1 with zero ehdI to only ϳ70 min Ϫ1 with 5 M ehdI (data not shown, and compare Fig. 5). Despite the difference in the affinity of ehdIII-ehdI for rrdI and for ehdI, a similar explanation for the results seems likely.
A Hybrid-tethered ehdIII-rrdI Complex-There is clear evidence that isolated dI of transhydrogenase from several species is dimeric (22, 28, 41, 42). None of the results from a number of different types of hydrodynamic experiments on rrdI gave any suggestion that the protein dissociates significantly into mono-mers even in the micromolar concentration range (22). We have therefore constructed a hybrid of ehdIII and rrdI joined in a single polypeptide by a linker equivalent to that found in ehdIII-ehdI to explore further the importance of a dimeric organization for hydride transfer. DNA coding for ehdI was excised from a plasmid bearing the ehdIII-ehdI gene fragment and replaced by a sequence coding for rrdI (see "Experimental Procedures"). The 38-residue linker from ehdIII-ehdI was retained with only a single substitution (Asp 3 His at the C terminus of the linker region). The hybrid protein, designated ehdIII-rrdI, expressed to high levels in E. coli and was easy to purify by a modification of earlier protocols for the nucleotidebinding components of transhydrogenase.
As with ehdIII-ehdI (see above), the mobility of ehdIII-rrdI during non-denaturing gel-exclusion chromatography indicated a molecular mass (ϳ165 kDa) that was intermediate between that of a dimer and a trimer (the molecular mass of the monomer is 66.4 kDa). However, ehdIII-rrdI proved to be a good subject for structural analysis using SAXS (see below), and the data showed unequivocally that the tethered hybrid is a dimer at protein concentrations in the region of 270 M. It is likely that flexibility in the linker region of both ehdIII-ehdI and ehdIII-rrdI (see below) leads to elevated hydrodynamic radii, causing increased retention times of the proteins in the matrix of the gel-exclusion column and therefore an exaggeration of the apparent molecular masses.
Purified ehdIII-rrdI catalyzed high rates of cyclic and low rates of reverse transhydrogenation (Fig. 6a), the familiar pattern for complexes of dI and dIII proteins (12). The cyclic reaction was faster than that catalyzed by dimeric ehdIII-ehdI, and the reverse reaction was a little slower (compare Fig. 3). Both the cyclic and reverse rates were comparable with those catalyzed by complexes formed from mixtures of isolated rrdI and either rrdIII (12) or ehdIII (see above). However, in marked contrast to the results obtained with ehdIII-ehdI (Fig. 3), the specific activity of ehdIII-rrdI (for both reverse and cyclic) was essentially independent of protein concentration down to Ͻ10 nM, the lowest concentration at which rates could be reliably measured. We suggest that, because of the tight interaction between the two R. rubrum dI polypeptides, active dimers of ehdIII-rrdI form with a much higher affinity (K d Ͻ Ͻ 10 nM) than those of ehdIII-ehdI (K d Ϸ 1 M).
As was observed with ehdIII-ehdI (Fig. 5), the addition of isolated rrdI to a solution of ehdIII-rrdI led to a stimulation of the rate of transhydrogenation, although the effect was not as pronounced (Fig. 6b); for example, with ehdIII-ehdI predominantly in its dimeric form (i.e. at ϳ1 M) rrdI increased the rate of cyclic transhydrogenation maximally by Ͼ18-fold, whereas the effect of rrdI on the reaction catalyzed by ehdIII-rrdI was maximally about 1.8-fold. The concentration of rrdI giving the half-maximal rate was similar to the values observed in the titration of rrdI against isolated ehdIII (Fig. 2b) and of rrdI against ehdIII-ehdI (Fig. 5b). These results again indicate that the linker is sufficiently flexible to allow the ehdIII moiety to move apart from its "own" dI and interact with the supplementary isolated rrdI.
The ability of ehdIII-rrdI to catalyze high rates of cyclic transhydrogenation indicate that the hydride transfer step is rapid. This conclusion was supported by stopped-flow experiments. As with simple mixtures of isolated rrdI and rrdIII (23,34), it was found that the slow steady-state rate of reverse transhydrogenation catalyzed by ehdIII-rrdI was preceded by a rapid burst of reaction (results not shown). This probably arises because the binding of AcPdAD ϩ and hydride transfer are both fast compared with the rate of product NADP ϩ release. As observed with R. rubrum dI 2 dIII 1 complexes, the amount of AcPdAD ϩ reduced during the burst was equivalent to about 50% of the amount of protein (32). The burst comprised two exponential phases of equivalent amplitude, which had apparent first order rate constants of ϳ300 and ϳ20 s Ϫ1 . These rate constants are similar to those measured for the biphasic burst observed with mixtures of rrdI and rrdIII (ϳ550 and ϳ50 s Ϫ1 ). There, it was convincingly established that the fast phase corresponds to hydride transfer and the slow phase to dissociation of dIII from the dI dimer prior to rapid reassociation and further oxidation of NADPH on dIII (23). We suggest that the biphasic burst of transhydrogenation of ehdIII-rrdI arises for similar reasons, but we cannot of course confirm this by independently varying the concentrations of the two components, the strategy we adopted with rrdI ϩ rrdIII.
SAXS Structure of the ehdIII-rrdI Hybrid Complex-SAXS data can provide reliable low-resolution information on the size and shape of macromolecules. Preliminary SAXS analyses of recombinant ehdIII-ehdI were difficult to interpret probably because of the tendency of the molecule to dissociate; the technique is not well suited to solutions that are not monodisperse. Measurements by dynamic light scattering in a range of buffer solutions (not shown) also indicated that solutions of ehdIII-ehdI were polydisperse. However, ehdIII-rrdI did prove to be a good subject for investigation; its SAXS curve is shown in Fig.  7. The molecular mass of the protein calculated from these data, using bovine serum albumin as a reference, was 130 kDa, strongly supporting the notion that ehdIII-rrdI is dimeric (see above). Assuming that the two dI components in the ehdIII-rrdI dimer interact as in the crystal structure of the isolated dI dimer and dI 2 dIII 1 complex from R. rubrum transhydrogenase (see "Discussion"), the results of Fig. 7 were used as a basis for discriminating between three plausible models of the hybrid tethered complex. 1) The two dIII components, held by the linkers, extend from the two dI components approximately along the major axis of the dimer. 2) Two of the dI components and one of the dIII components form a structure similar to that of the R. rubrum dI 2 dIII 1 complex, whereas the second dIII extends from its respective dI. 3) Two of the dI components and one of the dIII components form a structure similar to that of the R. rubrum dI 2 dIII 1 complex, and the second dIII binds close to the symmetrically equivalent location of the first (rotated 180 o about the 2-fold axis running between the two dI polypeptides) (see Fig. 7).
In the first approach toward determining the structural organization of the tethered hybrid complex, a pair-distribution function was generated from the experimental scattering profile using the program, GNOM (36). The function had a smooth profile tending gradually toward zero at a maximum diameter of 150 Å, which is consistent with the second two models but not with the first. Then, low-resolution reconstructions of the particle shape were calculated from the pair-distribution function using the program, DAMMIN (37). The reconstructions from five independent calculations converged as shown in Fig.  8. Only the second model can be fitted satisfactorily into the envelope defined by the dummy atoms of the DAMMIN calculations. Finally, molecular versions of each model were built by manipulating the x-ray structure of the R. rubrum dI 2 dIII 1 complex (together with a second dIII component) using the program MASSHA (Ref. 39, and see Fig. 7), and these were used to generate theoretical scattering profiles with the program CRYSOL (40). Again, the second model gave the best fit to the experimental data. DISCUSSION The results described above provide a further illustration that dI and dIII components of transhydrogenase, even in the absence of membrane-spanning dII, form a complex that is capable of high rates of hydride transfer. Remarkably, hybrid mixtures of dI and dIII from transhydrogenases of widely dif- ferent species are active. Although not the focus of this report, it is important to appreciate that in all of the complexes of dI and dIII that have now been investigated, the rate of cyclic transhydrogenation always greatly exceeds that of the reverse (and forward) reactions. This is a consequence of the fact that NADP ϩ and NADPH binding to dIII is extraordinarily tight in complexes of dI and dIII (12), and it gives strong support to the proposal that, in the complete enzyme, proton translocation through dII drives changes in the binding of these nucleotides (see below and Ref. 4).
The primary objective of the experiments described in this report was directed toward a better understanding of the importance of oligomeric interactions for catalysis by transhydrogenase. The strong dependence of the specific transhydrogenation activity on ehdIII-ehdI concentration, shown in Fig. 3, indicates that oligomeric interactions are indeed vital. The evidence suggests that ehdIII-ehdI is a dimer at high protein concentrations, and we therefore propose that the specific activity dependence in the micromolar range reflects the equilibrium between an inactive monomer and an active dimer. It was concluded that dimeric interactions between the two dI components are required per se for hydride transfer. The idea that dimer formation is necessary because active sites can only form between dI and dIII components from separate polypeptide chains was discounted. Thus, isolated ehdIII (shown to be capable of producing active hydride-transfer sites with dI; see Fig. 2, a and b) was unable to enhance the rate of transhydrogenation when in combination with ehdIII-ehdI monomers (Fig.  4). Note however, that although the experiments establish that dI-dI interactions are essential for transhydrogenation activity, they do not rule out the possibility that active sites are indeed formed between dI and dIII components from separate polypeptide chains, i.e. in a domain-swapped organization.
The crystal structures of isolated dI (20) and the dI 2 dIII 1 complex (21) of R. rubrum transhydrogenase reveal extensive contact surfaces between monomers in the dI dimer; both hydrophobic and H-bond interactions contribute to the dimer interface. As a result of these interactions, the domains designated dI.2(A) and dI.2(B) (of the A and B polypeptides of dI) pack together to form a central rigid core to the dI dimer; they lie "back-to-back" across a 2-fold axis of symmetry. Domains dI.1(A) and dI.1(B) are more peripheral and appear to participate less in dimer formation. There is a high level of sequence homology between R. rubrum and En. histolytica transhydrogenases (43% identity in dI), and therefore similar contact surfaces are likely to be responsible for the formation of the ehdIII-ehdI dimers.
The fact that the specific activity of the hybrid ehdIII-rrdI is approximately constant down to the lowest protein concentrations that gave reliable transhydrogenation rates (Fig. 6a) suggests that it forms very stable dimers; its K d for dimer dissociation must be at least two orders of magnitude lower than that of ehdIII-ehdI (equivalent to Ͼ11.5 kJ mol Ϫ1 ). We do not know which amino acid substitutions are responsible for this difference in stability; in principle, differences in only two or three amino acid residues might be sufficient. Of course, in the intact enzyme, interactions between the dII components will also help to stabilize the overall dimeric structure.
That isolated dIII-dI complexes are active only in their dimeric form is probably a reflection of the process of site alternation thought to take place during turnover of the intact enzyme. It was proposed that, coupled to proton translocation through the membrane-spanning dII, conformational changes drive dIII between an open state (in which bound nucleotides can exchange with those in the solvent) and an occluded state (in which hydride transfer with nucleotides on dI can occur) (4). Within the dimeric unit of the intact transhydrogenase, events in one monomer run 180 o out-of-phase with those in the other. Features in the x-ray structure that mediate conformational changes across the interface of the dI dimer and synchronize the reciprocating alternations were identified (4,21). We suggest that it is the absence of these interactions in monomers of dI-dIII complexes that leads to a failure of hydride transfer. Either dIII fails to engage with dI, or the active site geometry is compromised in a more subtle way. Evidence for the first alternative was obtained from the experiments discussed below.
The stoichiometry in solution (22) and the crystal structure of the dI 2 dIII 1 complex of R. rubrum transhydrogenase show that only one dIII component (in its occluded state) binds to the dI dimer; it is located at the side of the cleft between dI.1(B) and dI.2(B). Modeling studies indicate that the binding of a second occluded dIII to the symmetrically equivalent site at the cleft of dI(A) is prohibited because it would lead to side-chain clashing with the first dIII. It was proposed that in the intact enzyme the second dIII (in its open state) is displaced; site alternation would lead to a switching of states at the next half-cycle. It is interesting, therefore, that the current experiments show that the linker in ehdIII-ehdI allows considerable freedom for relative movement of dIII away from dI, even to the point where supplementary isolated dI (dimer) can bind to ehdIII-ehdI giving additional active hydride-transfer sites (Fig.  5). Furthermore, the results of this experiment show that dI and dIII components interact more strongly with one another in dimeric ehdIII-ehdI than they do in the monomeric form of the protein. Thus, less ehdIII from the tethered protein is available to form hydride-transfer sites (with added rrdI) when it is dimeric than when it is monomeric. This strongly supports the proposal (above) that the very low rate of transhydrogenation catalyzed by monomeric ehdIII-ehdI (Fig. 3) is a consequence of the failure of dIII to engage with dI.
In ehdIII-ehdI (and ehdIII-ehdI with added rrdI) and ehdIII-rrdI, the steady-state rate of cyclic transhydrogenation is two to three orders of magnitude greater than that of the reverse reaction (Figs. 3, 5, and 6). This indicates that most if not all the dIII components in the tethered complexes are in the occluded state; hydride transfer is fast (the cyclic reaction includes two hydride transfer steps, NADH 3 NADP ϩ and NADPH 3 AcPdAD ϩ ), and NADP ϩ release is slow (the ratelimiting step in reverse transhydrogenation). The situation is similar with complexes made from isolated dI and dIII (12). The important question therefore arises as to whether, also in the tethered complexes, only one dIII component is located at FIG. 8. DAMMIN calculations on the SAXS curve of tethered hybrid ehdIII-rrdI. The larger spheres show four different orientations of the overlaid results of five independent DAMMIN calculations on pair-distribution functions derived from a GNOM analysis of the experimental scattering data (see "Results"). The smaller spheres show the crystal structure of the R. rubrum dI 2 dIII 1 complex; the circle/ellipse representing dIII is appended to give an approximation of model 2 (see "Results" and Fig. 7). the proper position for hydride transfer on the dI dimer. The results of SAXS experiments (Figs. 7 and 8) provide evidence that this is indeed the case, at least for the ehdIII-rrdI hybrid. The data are readily consistent with the assumption that the two dI components and one of the dIII components adopt a structure similar to that of the R. rubrum dI 2 dIII 1 complex. Furthermore, several different analyses of the SAXS curve indicate that the second dIII does not bind symmetrically to the first but is significantly displaced from the region of the dI cleft. We believe that this displacement is probably a reflection of what happens in the intact enzyme to the dIII not currently forming a hydride transfer site with dI. During turnover, site alternation will subsequently lead to displacement of the dIbound dIII and a coincident tight association of the other dIII with its dI.
The character of the reaction burst with ehdIII-rrdI revealed in stopped-flow experiments can also be explained by this model. Thus, the fast phase of the burst corresponds to hydride transfer (NADPH 3 AcPdAD ϩ ) at catalytic sites in the interface between the closely associated dI and dIII. The slow phase corresponds to: 1) dissociation of that site; 2) association of dI and dIII at the second site; and 3) hydride transfer at the second site. By analogy with experimental results with mixtures of isolated R. rubrum dI and dIII (where the rate constants are very similar), it is suggested that site dissociation is the rate-limiting step. Thus, in a sense the biphasic burst in the tethered complex would correspond to events occurring in the intact enzyme with first one site reacting and then the other. However, in dI-dIII complexes, because there is no dII (and no proton translocation) the dIII component remains in the occluded state (from which product NADP ϩ (or NADPH) cannot escape), and so further turnover is prevented.
In conclusion, we have used a combination of transhydrogenase activity measurements and biophysical analyses of the subunit organization of tethered complexes of dI and dIII to reveal information that was not available from investigations of complexes formed from the isolated components. The future challenge is to extend the methodology to the intact enzyme and to determine precisely how conformational changes accompanying proton translocation through dII drive the alternating events in the nucleotide-binding subunits.