Division of labor in transhydrogenase by alternating proton translocation and hydride transfer

See allHide authors and affiliations

Science  09 Jan 2015:
Vol. 347, Issue 6218, pp. 178-181
DOI: 10.1126/science.1260451

Dueling dimers serve dual purposes

Both bacteria and mitochrondria produce NADPH for amino acid biosynthesis and to remove reactive oxygen species. The enzyme that makes NADPH must translocate a proton across the membrane and transfer a hydride from NADH to NADP+—processes that happen some 40 Å apart. To understand this complex geometry, Leung et al. solved the structures of the entire transhydrogenase enzyme and the membrane domain from the bacterium Thermus thermophilus (see the Perspective by Krengel and Törnroth-Horsefield). The entire enzyme exists as a dimer, with the two membrane domains in alternate orientations. One of the membrane domains interacts with the membrane component for proton translocation, whereas the other domain exchanges hydride with NAD(H) in another large soluble domain.

Science, this issue p. 178; see also p. 125


NADPH/NADP+ (the reduced form of NADP+/nicotinamide adenine dinucleotide phosphate) homeostasis is critical for countering oxidative stress in cells. Nicotinamide nucleotide transhydrogenase (TH), a membrane enzyme present in both bacteria and mitochondria, couples the proton motive force to the generation of NADPH. We present the 2.8 Å crystal structure of the transmembrane proton channel domain of TH from Thermus thermophilus and the 6.9 Å crystal structure of the entire enzyme (holo-TH). The membrane domain crystallized as a symmetric dimer, with each protomer containing a putative proton channel. The holo-TH is a highly asymmetric dimer with the NADP(H)–binding domain (dIII) in two different orientations. This unusual arrangement suggests a catalytic mechanism in which the two copies of dIII alternatively function in proton translocation and hydride transfer.

Nicotinamide nucleotide transhydrogenase (TH) is an integral membrane protein that can utilize the proton motive force for hydride transfer from NADH (the reduced form of NAD+) to NADP+ (nicotinamide adenine dinucleotide phosphate), forming NADPH (1): NADH + NADP+ + H+out ↔ NAD+ + NADPH + H+in, where “out” denotes the space outside the prokaryotic plasma membrane and “in” denotes the cytosol (or mitochondrial intermembrane space and matrix in eukaryotes). The resultant NADPH is used for amino acid biosynthesis and by glutathione reductase and glutathione peroxidase to remove reactive oxygen species (2). Mutations in Nnt genes, which encode for TH, are correlated with familial glucocorticoid deficiency of the adrenal cortex (3). Mice with mutations in the Nnt genes exhibit glucose intolerance and impaired secretion of insulin, as in type 2 diabetes (4, 5).

Functional TH exists as a dimer (6), where each protomer contains three domains: a hydrophilic NAD(H)-binding domain (dI), a transmembrane domain (dII), and a hydrophilic NADP(H)-binding domain (dIII) (Fig. 1A). Despite topological variations of TH subunits across species (fig. S1), dIII is always a C-terminal extension of dII. Hydride transfer between NAD(H) and NADP(H) occurs across the dI-dIII interface; proton translocation is across dII. One proton is translocated across the membrane per hydride transferred between nucleotides (7).

Fig. 1 TH function and structure.

(A) Schematic representation of TH function. The NAD(H)-binding domain (dI) and the NADP(H)-binding domain (dIII) catalyze hydride transfer. Proton translocation occurs through the transmembrane domain (dII). “In” represents the prokaryotic cytosol or mitochondrial matrix, and “out” is the space outside the prokaryotic plasma membrane or mitochondrial intermembrane space. (B) A Tt TH protomer is encoded by three genes: α1 encodes dI; α2 encodes three transmembrane (TM) helices of dII; and the β gene encodes the remaining nine helices of dII, the hinge region, and dIII. TM1 and TM5 of dII are missing in Tt TH.

Structures of hydrophilic dI and dIII were previously solved for TH isolated from bovine (8) and human (9) mitochondria as well as from Rhodospirillum rubrum (10) and Escherichia coli (11). Co-crystallization of dI with dIII results in the asymmetric heterotrimer (dI)2:dIII (12, 13). We determined the structures of the dimeric (dI)2 and the heterotrimeric (dI)2:dIII from Thermus thermophilus (Tt) TH (fig. S2A and table S1). In each of the heterotrimer structures, the NADP(H)-binding site of dIII is near the NAD(H)-binding site of one copy of dI, poised for direct hydride transfer. If a second copy of dIII were present and obeyed local twofold symmetry, there would be a steric clash between the two copies of dIII (12, 13) (fig. S2B). The structures therefore present a fundamental puzzle: how the second copy of dIII is oriented in the intact enzyme (14, 15) and how hydride transfer at the dI-dIII interface is coupled to proton translocation across dII, approximately 40 Å away (6, 16). Solving the structures of dII and the intact enzyme from Tt TH addresses these questions.

Tt TH is encoded by three genes (α1, α2, and β): α1 encodes dI; α2 encodes the first three transmembrane (TM) helices of dII; and β encodes nine TM helices of dII linked with dIII (Fig. 1B). Based on sequence alignments, two helices (TM1 and TM5) present in TH from some species are missing in Tt TH dII (fig. S3). To determine the crystal structure of dII, we made a construct that expresses the α2 gene and a truncated version of the β gene, in which dIII is replaced by a histidine tag (Fig. 2A). The resulting dII contains 12 TM spans and crystalizes as a symmetric dimer with a twofold axis normal to the membrane plane (Fig. 2B and table S2). The topology is consistent with previous studies of TH from different species, so the fold of dII observed in Tt TH is probably universal (17) (fig. S3).

Fig. 2 Structural details of the proton channel.

(A) Schematic diagram of the Tt dII construct used for crystallization, labeling TM helices, cytoplasmic loops (CL), and periplasmic loops (PL). The Tt β gene was truncated at CL8 by a histidine tag (red H). Residues in orange form hydrogen bonds within the proton pathway. Residues in green form a conserved salt bridge. Residues in gray outlines are disordered in the crystal structure. (B and C) Atomic model of the dII dimer viewed from (B) within the membrane and (C) from the periplasm. TM3, TM9-10, and TM13 form the proton channel. TM2 is at the dimer interface. (D) Residues α2N39 (TM3), βN89 (TM9), βS132 (TM10), and βN211 (TM13) form a hydrogen bond network (black dotted lines). (E) The Asp-Arg salt bridge formed by βD202 (TM13) and βR254 (CL8) is located on the membrane surface adjacent to the proton channel opening (red star).

Each dII protomer constitutes a proton pathway where helices TM3-4, TM9-10, and TM13-14 arrange as a hexagram (Fig. 2C). Conserved motifs located at a level corresponding to the middle of the membrane are found within the interior of the hexagram on TM3 (NXXH/S), TM9 (NXXG or HXXV), and TM13 (NXXT/S) (fig. S4). In the Tt dII structure, α2N39 (TM3), βN89 (TM9), βS132 (TM10), and βN211 (TM13) form a hydrogen bond network (Fig. 2D). Of specific interest is βN89 (TM9); this residue is at the equivalent position as E. coli βH91, which has been identified as important for proton translocation by mutagenesis studies (2, 18, 19). Two other residues that have been implicated in proton translocation in E. coli TH are βS139 (20) and βN222 (18), which are equivalent to βS132 (TM10) and βN211 (TM13) in Tt TH. No ordered water molecules are observed in the structure; hence, this may correspond to an occluded conformation of the proton channel.

On the basis of previous cysteine crosslinking studies, a salt bridge between E. coli βD213 and βR265 has been postulated and implicated as important for both proton translocation in dII and NADP(H) binding in dIII (18, 20, 21). The structure shows that this Asp-Arg salt bridge is formed between the equivalent residues in Tt TH, βD202 (TM13), and βR254 (the “hinge region” connecting dII and dIII) and is positioned on the membrane surface adjacent to the entrance of the proton channel (Fig. 2E).

Using the structures of the dI dimer, the dII dimer, and the dIII subunits, we solved the crystal structure of Tt holo-TH at 6.9 Å resolution by molecular replacement in Phaser (22) (Fig. 3A and table S3). In the 200-kD holo-TH dimer, two copies of dIII are positioned between the dI and dII dimers in different orientations (fig. S5A). One copy of dIII (dIII-A) has its NADP(H)-binding site “face up” to interact with the NAD(H)-binding site of dI-A as in the (dI)2:dIII heterotrimer structures (fig. S2A). The second copy of dIII (dIII-B) occupies the space between dI-B and dII-B, but it is flipped over 180° with its NADP(H)-binding site “face down” toward dII-B. Although the structures of the hinge region between dII and dIII are unknown, both orientations of dIII can be accommodated by the full-length linker in all species (fig. S5B).

Fig. 3 Structural asymmetry of holo-TH.

(A) The crystal structure shows two orientations of dIII: face-up dIII-A (green) and face-down dIII-B (magenta) with respect to their NAD(P)H-binding sites. Nucleotides, NAD(H) in dI, and NADP(H) in dIII, are modeled as black spheres based on the heterotrimer structure (11, 12). Subunits sandwiching dIII-A and dIII-B are respectively labeled as A and B. Underlined labels mark the mutations used for crosslinking experiments. (B) SDS-PAGE shows the crosslinking of dIII in two orientations. The table on top describes samples in lanes 1 to 10. For the wild-type enzyme (lane 10), subunits α1 and β each run with the same apparent molecular weight (MW) of about 45 kD; the α2 subunit shows a faint band at 10 kD. βD284C on dIII can crosslink with either α1T164C in dI (lanes 1 and 3), forming α1+β; or with α2V28C on dII (lanes 1 and 5), forming α2+β. The presence of β-Me largely eliminates the crosslinking (lanes 2, 4, and 6). (C) The cryo-EM structure reveals an asymmetric dimer: One copy of dIII exhibits weaker density than the other.

To confirm the existence of the two dIII orientations in holo-TH, pairs of cysteine residues were inserted into holo-TH for crosslinking experiments using the purified enzyme solubilized in detergent. The structure predicts that residue βD284 (dIII) should be at the interface with dII in the face-up conformation, but at the interface with dI in the face-down conformation. Residue βD284 was replaced by a cysteine (βD284C) in combination with a cysteine placed in dI (α1T164C) and/or dII (α2V28C). SDS–polyacrylamide gel electrophoresis (PAGE) analysis (Fig. 3B) showed that dIII can crosslink either to dII, forming the α2+β product, or to dI, forming the α1+β product. The crosslinks were largely reversed by β-mercaptoethanol (β-Me). Activity measurements showed that hydride transfer from NADH to NADP+ and then back to an NAD+ analog (so-called cyclic activity), which is not coupled to proton translocation (23), is minimally influenced by the crosslinking (table S4). In contrast, hydride transfer from NADPH directly to the NAD+ analog, which is coupled to proton translocation (24), is decreased by crosslinking and is increased after disrupting the crosslinks. Although the data are not quantitative, it is clear that the face-down conformation of dIII is present in the active form of holo-TH and is consistent with the importance of domain dynamics for catalytic function (21).

In addition to crystallography, we performed single-particle cryo–electron microscopy (cryo-EM) imaging of detergent-solubilized Tt holo-TH, yielding a structure at 18 Å resolution (Fig. 3C and fig. S6). The structure shows that the enzyme is an asymmetric dimer: One copy of dIII is more disordered than the other.

Previous biochemical data have shown that mutations in a conserved Asp residue in the NADP(H)-binding site of dIII, E. coli βD392, abolish both hydride transfer and proton translocation (16); and the rate-limiting step in forward catalysis is the release of the product NADP(H) from dIII (23). The holo-TH structure provides mechanistic implications about how dIII can be involved in proton translocation: The face-down conformation of dIII brings Tt βD378 (equivalent to E. coli βD392) (16) and regions of dIII that regulate NADP(H) binding, including loops D and E (14, 25), close to the surface of the proton channel and the Asp-Arg salt bridge (Fig. 4A). Hence, it is reasonable that the face-down conformation of dIII is competent for proton translocation, whereas its face-up conformation allows hydride transfer (Fig. 4B). The division-of-labor assignation works into the binding-change model of the catalytic cycle proposed by Jackson and colleagues (26). The model is consistent with previous biochemical determinations: (i) TH exhibits half-of-the-sites reactivity, so that inhibition of one protomer within the dimeric enzyme inhibits the second protomer (23, 24, 27); (ii) There are “open” and “occluded” conformational states of dIII via its loops D and E, summarized in (26). The model assumes that dIII is mobile, which is a reasonable hypothesis because previous cysteine mutations of the Asp-Arg salt bridge (Fig. 2E) lead to crosslinking and inhibition of both proton translocation in dII and NADP(H) binding in dIII (21). The membrane domain is proposed to have a single buried protonatable site (in most cases, a histidine in TM3 or TM9), and this residue can either be (i) deprotonated and occluded, (ii) accessible to protonation by an “outward-facing” conformation, or (iii) accessible to protonation by an “inward-facing” conformation. We speculate that changes in dIII, which include the protonation state of Tt βD378 (E. coli βD392) (28), the conformation of loops D and E (14, 25), and the redox state of NADP(H), are coupled to conformational/protonation changes in the membrane domain coupled to proton translocation (Fig. 4C).

Fig. 4 Possible role of face-down dIII in proton translocation.

(A) Regions of dIII that regulate NADP(H) binding, namely loop D, loop E, and Tt βD378 (green star), are close to the proton channel opening (red star) and the Asp-Arg salt bridge (blue stars). The surface area of dII is shown as translucent. NADP(H) is modeled as black spheres. (B) Division of labor of enzymatic activities within a TH dimer: One half of the dimer carries out hydride transfer and the other translocates proton. (C) Coordination between dII and dIII in one half of the TH dimer for the forward reaction. A proton is represented by a red cross. Two halves of the TH dimer cycle out of phase; for simplicity of representation, the other half is shown in gray and remains unchanged.

In general terms, we postulate that the two halves of the TH dimer cycle out of phase between the occluded and open states of dIII, each of which is coordinated to the inward-facing or outward-facing conformation of the proton channel. After the hydride transfer reaction, dIII becomes occluded in the face-up orientation, whereupon it flips to the face-down orientation without NADP(H) exchange with the solvent pool of NADP(H). For the forward reaction (NADPH product with a proton translocated inward), the interaction between the occluded dIII and the membrane domain converts dII from the deprotonated occluded state to become outward-facing, followed by the conversion of dIII to the open state, switching dII to inward-facing, and eventually the flipping up of dIII. Dissociations of both NADPH from dIII and the proton from dII to the solvent are required for the catalytic cycle to proceed. For hydride transfer in the reverse direction, the sequence of events is similar, except that dII proceeds from proton-occluded→inward-facing→outward-facing (fig. S7). The proton motive force is coupled to the dissociation of NADP(H) and therefore to the equilibrium for hydride transfer between NAD(H) and NADP(H).

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S12

Tables S1 to S4

References (2955)

References and Notes

  1. Acknowledgments: We thank Y. Hatefi for commitment and inspiration on TH; M. Yeager, H. Heaslet, J. Chartron, and R. Akhouri for their work on E. coli holo-TH; M. Saier and Z. Zhang for providing the bacterial strain; M. Soltis and D. Terrell for the Gryphon robot; V. Cherezov and M. Caffrey for helpful comments and suggestions; and W. Liu for help with using the RockImager. We thank the staff of the Stanford Synchrotron Radiation Lightsource and the staff of the Northeastern Collaborative Access Team at the Advanced Photon Source for assistance with data collection. This research was funded by National Institutes of Health (NIH) grant 5R01GM061545 2004 – 2008 and by National Institute of General Medical Sciences (NIGMS) grant 1R01GM103838 - 01A1, GM095600 (to R.B.G) PA is the Program Announcement number of NIH Research Project Grant. “Structural Biology of Membrane Proteins” is the title of the program. It was a mistake to include it in acknowledgements. The work was performed at the National Resource for Automated Molecular Microscopy, which is supported by a grant from the NIH NIGMS (P41GM103310). A. Moeller is supported by NIH grant GM073197. Atomic coordinates and diffraction data for the structures reported here are deposited in the RCSB Protein Data Bank under accession codes 4O93 and 4O9U. The authors declare no competing financial interests. For the crystallography research, J.H.L. and C.D.S. designed experiments; J.H.L., M.Y., and C.D.S. performed experiments; and J.H.L. and C.D.S. analyzed the data. For the cryo-EM research, A.M. and J.A.S. designed and performed experiments and analyzed data. L.A.S. and R.B.G. contributed the Tt TH operon for expression and initial purification. L.A.S., J.H.L., and R.B.G. created holo-TH constructs and expressed and assayed mutants. J.H.L., R.B.G., and C.D.S. wrote the manuscript.
View Abstract

Navigate This Article