Structural insights into the electron/proton transfer pathways in the quinol:fumarate reductase from Desulfovibrio gigas

The membrane-embedded quinol:fumarate reductase (QFR) in anaerobic bacteria catalyzes the reduction of fumarate to succinate by quinol in the anaerobic respiratory chain. The electron/proton-transfer pathways in QFRs remain controversial. Here we report the crystal structure of QFR from the anaerobic sulphate-reducing bacterium Desulfovibrio gigas (D. gigas) at 3.6 Å resolution. The structure of the D. gigas QFR is a homo-dimer, each protomer comprising two hydrophilic subunits, A and B, and one transmembrane subunit C, together with six redox cofactors including two b-hemes. One menaquinone molecule is bound near heme bL in the hydrophobic subunit C. This location of the menaquinone-binding site differs from the menaquinol-binding cavity proposed previously for QFR from Wolinella succinogenes. The observed bound menaquinone might serve as an additional redox cofactor to mediate the proton-coupled electron transport across the membrane. Armed with these structural insights, we propose electron/proton-transfer pathways in the quinol reduction of fumarate to succinate in the D. gigas QFR.

subunits are associated with the E. coli and A. suum QFRs, whereas only one membrane-associated subunit is seen in the W. succinogenes enzyme. As anticipated, the structures of the hydrophilic subunits are similar among these QFRs, but major structural variations are found in the membrane-embedded subunits 9  suum. Q D is apparently not redox-active in E. coli QFR 10,11 . Based on these structural findings, electron transfer mechanisms have been proposed to account for the quinol reduction of fumarate to succinate to enable the anaerobic microorganisms to grow on fumarate 4,5,7 . Two protons are consumed in the two-electron reduction of the fumarate, and it has been proposed that the transfer of these protons to the cytoplasm is coupled to the electron transfers, the so-called E pathway 12 , with key residues identified or proposed in the subunit C of W. succinogenes 13,14 .
To clarify the mechanism of QFR, we have cultured large amounts of D. gigas cells under anaerobic conditions and purified the QFR directly from the cell membranes for structural studies. We have determined the crystal structure of QFR from D. gigas at 3.6 Å resolution by X-ray crystallography. All the redox cofactors and a bound menaquinone are revealed in our structure. The menaquinone-binding site is different from that proposed for the QFR structure of W. succinogenes. Detailed comparison of the structures and redox cofactors among the various QFRs has allowed us to propose the pathways of electron and proton transfers in QFR from D. gigas during turnover.

Results
Overall structural fold. Two hetero-trimeric complexes, each comprising subunits A, B and C, form one stable homo-dimer (A 2 B 2 C 2 ) with major contacts between two C subunits (Fig. 1a). There are two homo-dimers 2(A 2 B 2 C 2 ) of the hetero-trimeric complex (ABC) in one asymmetric unit (Fig. 1b), but the contact interface is insufficient to form a stable tetramer in the crystal packing, with averaged buried surface area of only ~382 Å 2 between only one subunit A from each of the two homo-dimers (chains A and I) (Fig. S1), consistent with our observation that the protein elutes as a homo-dimer in size-exclusion chromatography during protein purification. The two membrane-spanning domains of the two homo-dimers 2(A 2 B 2 C 2 ) in the crystal are oriented in opposite directions. Since this could not be the proper scenario for QFR within the cell membrane, the homo-dimer (A 2 B 2 C 2 ) must be the biological functional assembly.
Within each homo-dimer, (A 2 B 2 C 2 ), the averaged buried areas are ~471 Å 2 between the two A subunits, negligible between the two B subunits, and 1340 Å 2 between the two C subunits; the portion of the buried surfaces of the two A subunits and the two C subunits are 2.1 and 9.5%, respectively. The main driving force in the homo-dimer formation thus comes from interactions between the C subunits. In support, the complex formation significance score (CSS) of the two C subunits and the two A subunits calculated by PISA 15 (http://www.ebi. ac.uk/pdbe/prot_int/pistart.html) are ~0.22 and ~0.02 (Fig. S2), respectively, indicating that the formation of the homo-dimer, (A 2 B 2 C 2 ), arises from contact of the two C subunits: the CSS value increases from 0 to 1 as the interface relevant to complex formation increases. The interactions between the two C subunits contain three important hydrogen bonds, two from Arg85-Gln91 and one from Gly151-Thr135 (Fig. S2).
Subunit A (FAD-binding protein). The flavin adenine dinucleotide (FAD)-binding protein (subunit A) of D. gigas QFR is composed by a total of 627 residues, with a FAD-binding domain (A1-260 and A366-437), a capping domain (A261-360), a helical domain (A438-555), and the C-terminal domain (A556-622) containing an antiparallel β-sheet (A564-569 and A578-583). Without interpretable electron density, the structure of the C-terminus A623-627 is too flexible to be built. This structure of subunit A differs from that of W. succinogenes QFR in that the latter contains two additional C-terminal helices, one long and one short.
The domain architecture of subunit A of D. gigas is similar to those of W. succinogenes, E. coli and A. suum. The location of FAD within the FAD-binding domain is similar for all four QFRs. The C8M methyl group of the flavin of FAD binds to the FAD-binding domain through a covalent bond with residue His-A43, which is conserved among these A subunits of QFRs. A fumarate molecule is also observed near the FAD in the A subunit. This position of the fumarate is conserved between D. gigas and W. succinogenes QFRs as well.
During the purification, there was a contamination of a ~60-kD protein after several types of columns (Fig. S3a). The ~60-kDa protein, accounting for about 30% of total proteins, was identified as the subunit A of QFR, lacking the N-terminal fragment, according to analysis by the liquid chromatography-tandem mass spectrometry (LC/MS/MS) (data not shown). The ~60-kDa protein might be a proteolytic fragment of the subunit A even though a protease inhibitor (cOmplete TM EDTA-free protease inhibitor cocktail, Roche) was added in all processes of cell disruption and protein purification. However, we cannot rule out the possibility that there is another isoform of the subunit A of D. gigas QFR without the N-terminal fragment.
Subunit B (iron-sulphur protein). The iron-sulphur protein of D. gigas QFR comprises a total of 264 residues, which can be partitioned into an N-terminal plant-ferredoxin domain (B1-106) and a C-terminal bacterial-ferredoxin domain (B107-240 Subunit C (membrane-embedded protein). This membrane-embedded subunit C of D. gigas QFR comprises a total of 218 residues organized into mainly seven helices, including five transmembrane helices: α1 (C16-41), α3 (C69-83), α5 (C107-134), α6 (C156-173) and α7 (C183-211); one periplasmic helix α2 (C49-56); and one cytoplasmic helix α4 (C91-104). Two b-type hemes are associated with the helix-bundle. The high-potential heme b H is proximal to the hydrophilic A and B subunits, and has as axial ligands His-C79 and His-C166 on helices α3 and α6, respectively. The low-potential heme b L , which is located distal to the hydrophilic A and B subunits, is coordinated to His-C38 and His-C129 on helices α1 and α5, respectively. These four helices, α1, α3, α5 and α6, which coordinate the two hemes, form the so-called the four-helix motif that exists in other di-heme proteins 16,17 . These four histidine residues are conserved in QFRs of D. gigas and W. succinogenes. They are missing for the enzyme in E. coli (Fig. S4), and only two are found in A. suum QFR. Consistent with this primary sequence information, there is no bound heme seen in the membrane-embedded subunit C of E. coli 7 and only one heme is associated with the A. suum enzyme 8 .
Of the remaining α-helices, helix α4 interacts with the iron-sulphur protein (subunit B) at the top near the cytoplasm, and helix α2 is situated at the bottom near the periplasm; helix α7 is also a transmembrane segment, which, together with the four-helix motif-α1, α3, α5 and α6, forms the transmembrane helix bundle sequestered within the inner membrane.
Bound detergents in D. gigas QFR. One n-dodecyl β-D-maltoside (DDM) molecule from the detergent used to solubilize and stabilize the protein is found in subunit C of D. gigas QFR (Fig. S5a). The polar head group of the DDM forms a hydrogen bond with Lys-C95 of QFR; the hydrophobic tail of the DDM interacts with a hydrophobic pocket formed by the hydrophobic residues Ile-C81, Pro-C88, Phe-C89, Trp-C94, Val-C116, Ala-C119 and Ile-C120 (Fig. S5b). A second DDM molecule is bound at the same location in the other subunit C of the homo-dimer. These bound DDM molecules are considered to mimic membrane lipids, indicating that QFR of D. gigas might utilize residue Lys-C95 to interact with the polar head-group in the inner leaflet of the lipid bilayer, and the tail of the lipid molecule might be sequestered by the hydrophobic pocket within the subunit C of QFR.
Structural comparison of QFRs. Superimposed structures of the dimeric QFRs, monomeric QFRs, individual subunits A, B and C from D. gigas with those of the prokaryotic W. succinogenes (PDB code: 2BS2) and E. coli (as a crystallographic dimer, PDB code: 1L0V) show root-mean-square-deviations (RMSD) 1.7/5.6, 1.43/1.92, 1.55/1.53, 0.91/1.59 and 1.38/3.41 Å (Fig. 2b), respectively, using the secondary-structure matching (SSM) 18 tool in Coot 19 . Since the eukaryotic QFR structure from A. suum exists as a monomer in solution and does not exhibit the similar dimeric arrangement as D. gigas QFR in the crystal packing (Fig. S6), only monomer structures between D. gigas and A. suum QFRs (PDB code: 3VR8) are compared: the RMSDs calculated by SSM are 1.90, 1.26, 1.66 and 3.16 Å for monomeric QFR, subunits A, B, and the membrane-embedded subunit C, respectively (Fig. S6). From these comparisons, it is evident that the hydrophilic subunits (subunits A and B) of D. gigas, W. succinogenes, E. coli and A. suum are relatively similar, whereas the hydrophobic subunit(s) (subunit C) exhibit major structural variations (Figs 2 and S6).
As shown in Fig. 2a, superimposition of the three prokaryotic dimeric QFR structures shows that there are minor molecular contacts between the two subunits A of the homodimer from D. gigas and W. succinogenes, but no molecular contact between the two subunits A from E. coli QFR. Moreover, low CSS (~0.05), but high CSS (~0.29) and (~0. 22), are shown between the two membrane-embedded subunits C of E. coli (as a crystallographic dimer), W. succinogenes and D. gigas QFRs, respectively. These observations are consistent with the dimeric form (A 2 B 2 C 2 ) of QFRs of D. gigas and W. succinogenes and the monomeric form (ABC) of QFR of E. coli in solution 7,20,21 .
The structural variations between D. gigas and W. succinogenes QFR are confined mainly to the C-terminal regions of subunits A and a loop between helices 2 and 3 of subunits C (Fig. 2a). There is a pair of long and short helices at the C-terminal region of subunit A in QFR from W. succinogenes; this helix pair is absent in QFR from D. gigas as well as from E. coli. The biological function of this pair of helices in QFR from W. succinogenes is unclear. As for the loop between the helices 2 and 3 in the C subunits, QFR from W. succinogenes contains a longer loop with eight extra residues, compared with QFRs of D. gigas and E. coli. The longer loop with a proposed critical residue Glu-C66 has been implicated in the binding of the menaquinol substrate in QFR from W. succinogenes 13 for the fumarate reduction. Without this long loop, as in the case of QFR from D. gigas, menaquinol binding might be affected here. The structural variation in this region might imply that the menaquinol-binding mode of QFR from D. gigas is different from that of W. succinogenes.
The deformed cavity for the head-group binding of the proposed menaquinol substrate. In an earlier study, individual replacements of Glu-C66 and Glu-C180 with Gln by site-directed mutagenesis of W. succinogenes QFR showed that both of these bacterial mutants could no longer grow under fumarate 13,14 . These results have implicated Glu-C66 and Glu-C180 of W. succinogenes QFR in the proton transfers that must accompany the transport of the electrons across the membrane during the reduction of fumarate by the quinol substrate. As shown in Fig. 3a and b, Glu-C66 is on the loop between helices 2 and 3 of QFR from W. succinogenes. Glu-C66 and a proposed cavity near the loop were suggested to be involved in the binding of the head group of the menaquinol substrate. In this region, there are actually two proposed menaquinol-binding cavities: one is for the head group, and the other, close to the b L heme, is for the hydrophobic tail of the menaquinol 13 .
In D. gigas QFR, however, Glu-C60 (the structural analog of Glu-C66 in W. succinogenes QFR) adapts a structural orientation such that the side chain points away from the proposed menaquinol-binding cavity in W. succinogenes QFR ( Fig. 3a and b). In addition, the loop containing Glu-C60 is shorter by eight residues (Fig. S4). As a result, the corresponding cavity in D. gigas QFR is deformed so that it can no longer accommodate the head group of a menaquinol substrate (Fig. 3). These observations, the different orientation of the Glu-C60 and the deformed cavity for the head group, might be taken to infer that D. gigas QFR possesses a different menaquinol-binding mode, wherein the menaquinol substrate molecule binds more closely to the b L heme to take advantage of the more hydrophilic cavity near this heme in D. gigas QFR compared with that in W. succinogenes QFR. However, Glu-C164 in D. gigas QFR, analog of the other essential residue Glu-C180 in W. succinogenes QFR, exhibits the similar structural conformation as in W. succinogenes QFR (Fig. 3b), indicating that Glu-C164 might play a similar crucial role in proton transfer(s) in D. gigas QFR. Other important residues, which might be involved in the electron-transfer and proton-transfer pathways, are discussed in the following sections.
Bound menaquinone near the low-potential heme (b L ). The menaquinol-oxidation site in D. gigas QFR (henceforth referred to as MQ1) is not known. To explore this issue, we search for potential menaquinol-binding positions based on the electron density maps (SIGMAA-weighted F o − F c and 2F o − F c ). The deformed cavity near Glu-C60 in D. gigas QFR, the binding cavity previously proposed in W. succinogenes, shows no extra density for a menaquinol molecule, in support of a menaquinol-binding location elsewhere on the periplasmic side of subunit C in D. gigas QFR. In any case, we do not expect menaquinol to be bound to the enzyme at the MQ1 site when fumarate is associated with the QFR, as the substrate will be readily oxidized. Neither should menaquinone, the product of the substrate oxidation, be seen in the crystal structure. The quinone should be only weakly associated with the QFR during catalytic turnover of the enzyme. The concentration of menaquinone in the cells is very low under growth conditions in any case.
Nevertheless, we have identified a menaquinol or menaquinone molecule with the appropriate electron density near helix 1 (C-helix1) of one subunit C among the four subunit C molecules in the asymmetric unit (Fig. 4a). We have designated this location in the D. gigas QFR as the MQ2 site, to distinguish this site from the MQ1 site, where substrate oxidation presumably occurs. At this site, the hydrocarbon tail of the menaquinol or menaquinone molecule is sequestered within a hydrophobic pocket formed by Met-C40, Leu-C41, Trp-C36 and Ala-C37 on C-helix1; and the headgroup is directly bound to Tyr-C63 on another short fragment via a hydrogen bond (Fig. 4b). With the limitation of X-ray crystallography, we could not distinguish between a quinol and a quinone occupying the MQ2 site. However, as a "cofactor", it is most certainly oxidized on the basis of chemical considerations, and we henceforth refer to it as a menaquinone (MK). According to the electron density, only the electron densities of two isoprenoid units of this MK were observed. Notably, we also tried to fit a DDM molecule, which was used for crystallization, into this extra electron density. However, the geometry of the head group of DDM could not be fitted appropriately based on the electron density.
Evidently, this MK is readily lost during protein purification as only one MK molecule is discerned per asymmetric unit of the protein crystal. The conformation of this short fragment (a.a. 57-68) containing the Tyr-C63, which is located between helices 2 and 3, is highly variable in the crystal structure of D. gigas QFR (Fig. 5). The four short fragments of the two dimeric QFR in the asymmetric unit exhibit conformations more variable than other regions, indicating that their dynamic property or flexibility might be related to the binding of menaquinones (Fig. S7). The Tyr-C63 that binds MK exhibits the lowest B-factor compared to the three other free Tyr63 residues of the asymmetric subunits C (i.e., G-chain, K-chain and O-chain), which might result from the formation of the hydrogen bond between the side-chain O-H of the C-chain Tyr-C63 and the bound MK. Interestingly, this Tyr-C63 residue, which is specific to D. gigas (Fig. S4), is positioned near residue Glu-C60, the analog of Glu-C66 in W. succinogenes QFR that has been implicated in the proton transfers accompanying the electron transport during the catalytic turnover. Finally, there is a highly variable region (a.a. C147-C152) behind the b L heme, but the dynamic role of this region is unclear (Fig. 5).
Moreover, this MQ2 site, or the bound MK, is located near (~13 Å) the low-potential b L heme (Fig. 5). This observation, together with the conformational flexibility of the short fragment (a.a. 57-68) containing the Tyr-C63 described earlier, has led us to surmise that MK might serve as an electron carrier between heme b L and [3Fe:4S] in the electron-transfer pathway of D. gigas. In other words, MK could be gating the proton-coupled-electron   (Fig. 6a). Reduction of MK gives the semi-menaquinone anion (MKH), which should stabilize the hydrogen bonding of the Tyr-C63 with the cofactor and lower the pK A of this tyrosine further. Upon reoxidation, as the electron is transferred from the MKH downstream to the [3Fe:4S], electrostatic interaction between the electron and the proton of the Tyr-C63 hydrogen-bond provides a linkage to drive the synchronous movement of the proton across the membrane in concert with the electron flow (Fig. 6b). In this manner, proton-coupled electron transfer across the membrane is achieved (Fig. 6c).

Electron-transfer pathway(s).
Based on the arrangement of the redox cofactors in the crystal structure, including the bound MK at the MQ2 site, as well as their redox potentials, it is possible to postulate the E-pathway(s) in QFR of D. gigas (Fig. 7). First, we propose that, upon oxidation of the menaquinol at the MQ1 site, there is bifurcation of the electron flow from the substrate to the high-potential heme b H and the low-potential heme b L . Second, we suggest that heme b L serves as a relay station to transfer one of these electrons to the [3Fe:4S] center via the MK at the MQ2 site, and from the [3Fe:4S] site, the electron passes on to the [4Fe:4S]-[2Fe:2S]-FAD, as mentioned earlier. To replenish the proton of the Tyr-C63 side chain after the proton-coupled electron transfer across the membrane, a proton must be taken up from the periplasm and transported via proton acceptors located within the periplasmic side of subunit C to the side chain of Tyr-C63 as it is known that the hydroxyl group of tyrosine can serve to transfer protons 22 . Potential proton acceptors might include Glu-C60, among others. Similarly, there must be proton acceptors within the transmembrane domain toward the cytoplasmic side to relay the proton that has been translocated across the membrane toward the cytoplasm. Potential proton acceptors here include residues Glu-C164 and Glu-C193 (Fig. 6c), and possibly other acidic residues Asp-C108, Asp-C21 and Glu-C93, which are at or near the cytoplasmic helix α4 (C91-104).
As to the other reducing equivalent, the one transferred to the b H heme in the bifurcation of the electron flow from the menaquinol substrate mentioned earlier, we propose a direct transfer of this electron from heme b H to the [3Fe:4S] (Fig. 7). The high-potential heme b H is in relatively close proximity to the hydrophilic A and B subunits. The redox potentials of heme b H (0 mV), heme b L (−150 mV) and MK (−67 to −74 mV) near pH ~7 are also suggestive of such a possible scenario 23,24 . However, for a proton-coupled electron transfer event to take place between subunit C and subunit B, there will have to be redox linkage between heme b H and the pK A of an amino acid residue within subunit C, as well as conformational gating of a proton transfer across the membrane in concert with the electron transfer. Possible proton donors might include His-C38, which is associated with heme b L . The b H and b L hemes in D. gigas QFR are in sufficient close contact (only ~4.5 Å apart) that we can expect a redox interaction between them.
At this juncture, however, we cannot rule out the possibility of also transferring the electron from the b H heme via heme b L to MK at the MQ2 site for the proton-coupled electron transfer to the [3Fe:4S] center, as we have described earlier for the other reducing equivalent. Indeed, the two hemes are in sufficiently close juxtaposition that the electrons can redistribute readily between these iron porphyrins, provided that their operating redox potentials and their relative orientation, are favorable. Examination of the crystal structure reveals that the two hemes are closely perpendicular to each other, a geometry that is known to be unfavorable toward facile inter-heme electron transfers 25 . Thus, on the basis of this structural information, as well as the relative redox potentials between heme b H and b L (150 mV) 23 , we are inclined to favor different electron transfer pathways for the two reducing equivalents of the menaquinol substrate during the catalytic turnover of D. gigas QFR.

Discussion
Upon oxidation of menaquinol in QFRs, two electrons are transferred across the membrane from the periplasm to the cytoplasm. QFRs usually comprise zero to two hemes in the membrane-embedded subunit, but the role of these hemes has not been fully characterized in the E pathway of QFRs 26 . The two-electron transfer through hemes and Fe-S clusters was previously proposed based on the locations of the redox cofactors in the QFR structure of W. succinogenes 5 . Native SQR of E. coli contains one heme; however, an E. coli SQR mutant lacking this heme retains the activity of ubiquinol reductase 27 , indicating that the exact role of the heme in this electron-transport protein is not fully understood. In the study, we have determined the structure of QFR in D. gigas with all the cofactors at a resolution of 3.6 Å. The overall architecture of this QFR comprises the subunit A (FAD-binding protein), the subunit B (iron-sulfur protein) and the subunit C (membrane-embedded protein). Here, Q P was proposed to be the quinol-oxdizing site 10,28 , and Q D might not be involved in the electron transfer pathway 11 . These observations and implications indicate that D. gigas QFR possesses a different electron transfer pathway from that in E. coli.
The superimposed structures of subunits C from D. gigas QFR and subunits C and D in E. coli QFR indicate that the b L and b H hemes in D. gigas QFR are structurally near Q P and Q D in E. coli QFR (Fig. S8), but the distance of closest approach between the b L and b H hemes (~4.5 Å) is much less than that between Q P and Q D (~25 Å), indicating that a direct electron transfer is more reasonable between the b L and b H hemes of D. gigas QFR than between Q P and Q D in E. coli QFR. If so, Q P and Q D might function more or less independently in mediating proton-coupled electron transfer events across the membrane in E. coli. If the b L heme in D. gigas QFR is involved in the E pathway, the menaquinol substrate-binding site (MQ1) is expected to be near this heme (<14 Å, the edge-to-edge distance for direct electron transfer) 29 . The first possibility for the binding site of this putative menaquinol (MQ1) might be the same position as that previously proposed for QFR in W. succinogenes, near Glu-C60 in D. gigas QFR, but there is no extra electron density in this region. Furthermore, the proposed binding cavity for the head group of this putative menaquinol is deformed, and the other cavity for the hydrophobic tail is more hydrophilic in D. gigas QFR, indicating that D. gigas QFR might have a menaquinol-binding site different from that of W. succinogenes QFR. Instead, the head group of the fully reduced menaquinol substrate might bind at the more hydrophilic cavity near the b L heme without the involvement of Glu-C60 in D. gigas QFR.
Interestingly, in the present study, we have located a menaquinone (MK) near the top of C-helix1. The distance (~13 Å) between this presumably fully oxidized quinone molecule and the b L heme in D. gigas QFR is reasonable to allow a direct electron transfer from the b L heme to this menaquinone, and subsequently to the Fe-S clusters in the subunit B and eventually to the fumarate in subunit A.
Although the distance ~28 Å between MK and [3Fe:4S] is too large for the direct electron transfer, there are many aromatic residues along the path between the two cofactors. These aromatic residues, including Trp-C36, Phe-C171, Phe-C189, Tyr-C172, Phe-C180 and Tyr-C178, might serve as the electron carriers to relay the long-distance electron transfer, as aromatic residues are known conduits for long-range electron transfer 30 .
We propose that MK, the bound menaquinone molecule, might serve as a coupling site to gate the proton-coupled electron transport across the membrane in D. gigas QFR, at least for one of the two reducing equivalents from the menaquinol substrate. When MK accepts the electron from the b L heme, it becomes the semi-oxidized menaquinone (MKH). Subsequently, as an electron is transferred to the [3Fe:4S] cluster, the side-chain O-H proton of the Tyr-C63 that is hydrogen-bonded to the semi-menaquinone anion, can undergo synchronous movement in concert with the electron transfer to the cytoplasmic side of subunit C. Following this proton-coupled electron transport, the Tyr-C63 can acquire a proton from the periplasm via a water channel or proton donors occupying a proton-transfer pathway on the periplasmic side of subunit C, and the fully oxidized menaquinone MK is restored. In principle, the MK can cycle between MK and MKH again, and in two cycles transfers the two reducing equivalents of the menaquinol substrate to subunit B across the membrane via the high and low potential hemes in the QFR of D. gigas. However, based on the X-ray structure, we surmise that it is more likely that one of the two reducing equivalents of the menaquinol substrate is transferred directly from the high-potential heme b H without the involvement of MK (Fig. 7).
The bifurcation of the electron flow, namely the transfer of the two reducing equivalents upon the oxidation of the quinol substrate through two heme centers, is well known and exists in many electron-transport proteins 31,32 . As the edge-to-edge distances relevant for electron transfer here are shorter than 14 Å 29 , the b L heme in D. gigas QFR is far from the [3Fe:4S] cluster (~28 Å). However, it is near the bound MK (~13 Å), indicating that electrons generated by the oxidation of the menaquinol substrate at the MQ1 site can pass through the b L heme. In this scenario, one of the two reducing equivalents of the menaquinol substrate is initially shuttled to the b H heme. This electron could be subsequently transferred to the MK via the b L heme during the second half of the turnover cycle, or it could be transferred directly to the [3Fe:4S] in subunit B.
To avert the buildup of an electrostatic potential across the membrane, the transmembrane electron transfers must be accompanied by coupled proton transfers during the oxidation of the menaquinol substrate and the reduction of fumarate to succinate on opposite sides of the membrane. The proton-transfer pathway(s) are still not clearly elucidated. The proton-active residues, glutamates and histidines, and protein-bound water molecules are considered the most commonly to be involved in the proton transfers 33 . Examination of the QFR structure from D. gigas shows two possible proton acceptors, Glu-C60 and His-C38, near the bound MK molecule, suggesting the possibility of two potential proton-transfer pathways, A and B, in D. gigas QFR (Figs S9 and S10). As shown in Fig. 6, in the proton-transfer pathway A, the distance between the O-H group of Tyr-C63 in the hydrogen bond of the partially oxidized MK, i.e., the semi-menaquinone anion (MKH), and the carboxyl group of Glu-C60 is ~8.8 Å, which is too large for a direct proton transfer (4 to 5 Å) 33 . But the proton transfer might be mediated by water molecules in a protein water channel, despite that no bound waters are observed in our QFR structure. Glu-C60 in D. gigas corresponds to Glu-C66 that was earlier suggested to be involved in the menaquinol binding or in proton transfer in W. succinogenes 13 . Thus, in the proton-transfer pathway A, the proton is transferred from the periplasmic space via Glu-C60 to the side-chain of Tyr-C63 after reoxidation of the MKH. In the proton-transfer pathway B (the transmembrane proton-transfer pathway), the distance between His-C38, the axial ligand of the b L heme, and the MKH in the hydrogen bond with the Tyr-C63 hydroxyl group, is ~12.8 Å. Again, water molecules are thus necessary to mediate the transfer of the proton from the MKH to the His-C38 in this proposed proton transfer pathway. From the analysis of glutamate and histidine residues in the QFR structure of D. gigas (Fig. S10b), the only possible proton acceptor from His-C38 is Glu-C164, so during the proton-coupled electron transfer mediated by MK, the transfer of the proton might be mediated by iron-coordinated His-C38 (and/or the b L ring C-propionate) of heme b L to Glu-C164. From Glu-C164, the subsequent proton acceptor might be Glu-C193 (Fig. S10b), which is a specific proton acceptor unique to D. gigas QFR. This proton acceptor is not found in W. succinogenes and E. coli QFRs (Fig. S4), indicating that D. gigas QFR might adopt a unique pathway, His-C38~Glu-C164~Glu-C193~cytoplasm, to transfer the redox-linked or coupled proton-transfers across the inner membrane to the cytoplasm (Fig. S10b). Note that His-C38 and Glu-C164 in QFR of D. gigas correspond structurally with His-C44 and Glu-C180 in W. succinogenes QFR. These two residues were identified to be critical for proton transfers in W. succinogenes QFR 6,14,34 . The transmembrane proton pathway between His-C38 and Glu-C164 might thus be similar to QFR in W. succinogenes, but the proton transfer between Glu-C164 to Glu-C193 seems to be unique to D. gigas QFR.
In summary, no suitable cavity near Glu-C60 in the C subunit has been identified for binding of the head group of the menaquinol substrate in the QFR structure of D. gigas, in contrast to the head-group binding cavity previously proposed for menaquinol in W. succinogenes QFR. The head group of the fully reduced menaquinol substrate molecule might bind instead to the more hydrophilic cavity close to the b L heme in D. gigas QFR. Since in this study crystals of the QFR are obtained on the enzyme isolated from the D. gigas grown under fumarate, the product menaquinone should not be associated with the enzyme in the protein preparation, and in fact, we have not detected one. Instead, we have identified a new bound menaquinone near helix 1 and heme b L in the QFR structure of D. gigas. We propose that this bound menaquinone (MK) serves as a novel redox cofactor, or a coupling site to gate the proton-coupled electron transport across the membrane in D. gigas QFR. In addition, a newly identified potential proton acceptor, Glu-C193, might be important in the redox-linked proton transfer across the membrane in the D. gigas QFR. On the basis of the QFR structure of D. gigas in a complex with the bound menaquinone, as well as comparison with the QFR structures of W. succinogenes, E. coli and A. suum, we have clarified the roles of the two b hemes in the electron-bifurcation pathway in the menaquinol oxidation and delineated the mechanism of proton-coupled pathways from the periplasm across the membrane during the transfers of the reducing equivalents from the quinol to fumarate in D. gigas QFR (Fig. 7). Finally, in light of this study, we may perhaps understand how the QFR in E. coli can function to mediate electron transfers without any b hemes. With two bound menaquinones, these cofactors can participate in the bifurcation of the electron flow from the quinol substrate and also serve as the coupling sites to gate the proton-coupled electron transport across the membrane exploiting the mechanism that we are proposing here for at least one of the reducing equivalents in D. gigas QFR.

Methods
Cell growth and preparation of cell extracts. Desulfovibrio gigas cells were grown in a fumarate/sulphate medium to obtain a large quantity of cells. The cells were grown over a period of ~24 h (OD 600 ≈ 1.0) at 37 °C with a ten-fold transformation in a 600 or 2000 L fermenter, producing 300 g or 1 kg of cells, respectively. The cultured cells were suspended in a buffer with Tris-HCl (20 mM, pH 7.6), and disrupted twice with a high-pressure homogenizer (AVESTIN ElulsiFlex-C3, 1500 psi) at room temperature. The disrupted cells were collected on ice immediately. After cell breakage, the suspension was centrifuged (KUBOTA 6930, 8000 rpm, 40 min) at 4 °C. The cell extract was concentrated further with an ultra-centrifuge (HITACHI CP70MX, 40,000 rpm, 2 h) at 4 °C. The supernatant was discarded; the pellet was washed (membrane fraction; black), twice with a buffer (volume 10 times) containing Tris-HCl (20 mM)/EDTA (1 mM), and once with 10 volumes of Tris-HCl (20 mM)/glycerol (10%). When the pellet turned from black to brown, the proteins were extracted twice from the pellet with a buffer containing Tris-HCl (20 mM, pH 7.6), glycerol (10%) and DDM (4%) at 4 °C for 2 h. The supernatant separated by ultracentrifugation (40,000 rpm, 2 h) was then ready for further purification.
Purification of QFR from D. gigas. All purification procedures were performed using the AKTA chromatography system (GE Healthcare) in a cold room (4 °C); all buffers used contained 0.1% DDM and were adjusted to pH 7.6. Crude extract was applied to a DEAE column (GE Healthcare, DEAE Sepharose Fast Flow) previously equilibrated with buffer A (Tris-HCl, 10 mM, pH 7.6), glycerol (10%) and DDM (0.1%, w/v). The column was eluted with a linear gradient (0 to 500 mM) of sodium chloride supplemented in buffer A, and the QFR was present in the fraction corresponding to sodium chloride (180 to 200 mM) of the eluted buffer. The QFR fraction was subjected to dialysis to decrease the ionic strength in buffer A. This fraction was then loaded onto a Q-sepharose column (GE Healthcare, Q Sepharose Fast Flow) and the protein faction was again eluted with NaCl (~200 mM). As there were some minor contaminations in this step, we performed a final size-exclusion column chromatography (GE Healthcare, Superdex 200 10/300 GL, equilibrated with 20 mM Tris buffer, pH 7.6, containing 10% glycerol and 0.1% DDM) to remove these contaminants. The protein fraction was concentrated in an ultrafiltration cell (Amicon) with molecular-weight cutoff 100 kDa. The purity of the fraction was examined with SDS-PAGE and the UV-visible spectrum (Fig. S3a).
Crystallization and X-ray data collection. Purified D. gigas QFR was concentrated (10 mg/mL) without glycerol for the crystallization experiment. Initial crystal screenings were performed with a crystallization robot (Mosquito Crystal, TTP Labtech) with the hanging-drop vapor-diffusion method on mixing the protein sample (100 nL) with a reservoir solution (100 nL) with several screen kits. Small crystals were observed in one condition containing PEG 4000 (15%), NaCl (200 mM) in a sodium acetate buffer (100 mM, pH 4.6) at 18 °C after two weeks. Crystals of D. gigas QFR were then transferred to a crystallization buffer containing glycerol (20%) for cryoprotection and were immediately frozen in liquid nitrogen.
The X-ray diffraction experiments were performed with photons of wavelength 0.9 Å on TLS beamlines BL13C, BL15A and TPS 05 A at the NSRRC (Taiwan) and BL44XU and BL12B2 at SPring8 (Japan). The screened crystals initially diffracted to only approximately 10 Å. Protein crystals were then further optimized using hanging drops consisting of equal volumes of protein solution (1 μL) and a reservoir solution (1 μL) equilibrated against a reservoir solution (150 μL) containing NaCl (100 mM), PEG 4000 (28%) and Tris buffer (100 mM, pH 8.5) at 18 °C for two weeks. The improved larger crystals grew to a final size 0.2 × 0.2 × 0.1 mm 3 (Fig. S3b). The best complete data set at resolution 3.6 Å (although some diffractions were observed at 3.5 Å) was collected at BL44XU after a number of crystals were screened simultaneously at BL12B2 of SPring-8. The data were processed with HKL2000. All X-ray data statistics are summarized in Table 1.
Structure determination. The processed data were reduced from averaged intensities to mean amplitudes with TRUNCATE in CCP4 suite 35 . The initial phase was determined with the molecular replacement method using MOLREP 36 in CCP4 suite with the structure of QFR from W. succinogenes (PDB entry 2BS2, sequence identity ~50%) without redox cofactors as the search model. Two homo-dimers formed by four sets of the hetero-trimeric complex (subunits A, B and C) were found in the asymmetric unit. Iterative cycles of model building and refinement using COOT 37 and PHENIX 38 , respectively, were performed. During the refinement, NCS restraints, secondary-structure restraints and the B-group refinement were applied. The redox cofactors, including hemes, iron-sulphur clusters, FAD and fumarate, were built based on electron density maps (2F o − F c and F o − F c ). Detergent (DDM) molecules and the menaquinone (MK) at the MQ2 site were finally fitted into appropriate positive F o − F c electron density maps. Iterative cycles of the model adjustment and refinement were performed in the last step to decrease the final R work /R free factors to 25.22%/30.84% at resolution 3.6 Å. An attempt to extend the resolution of refinement to 3.5 Å was made, which increased both the R work /R free values because of the insufficient data quality beyond 3.6 Å. The refined structure of QFR of D. gigas was validated with MolProbity 39 . The rotamer outlier, clashscore and MolProbity score are 0.37%, 25.45 and 2.45, respectively. The calculations of root-mean-square deviations of the angles, bonds and dihedral and improper angles showed the satisfactory stereochemistry. The figures were prepared with PyMOL 40 .
Accession Codes. The structural factors and coordinates have been deposited in the RCSB Protein Data Bank. The PDB code is 5XMJ.