N. Scrutton, Faculty of Life Sciences, University of Manchester, 131 Princess Street, Manchester M1 7DN, UK Fax: + 44 1613065201 Tel: + 44 1613065152 E-mail: email@example.com Website: http://www.mib.manchester.ac.uk
Electron transferring flavoproteins (ETFs) are soluble heterodimeric FAD-containing proteins that function primarily as soluble electron carriers between various flavoprotein dehydrogenases. ETF is positioned at a key metabolic branch point, responsible for transferring electrons from up to 10 primary dehydrogenases to the membrane-bound respiratory chain. Clinical mutations of ETF result in the often fatal disease glutaric aciduria type II. Structural and biophysical studies of ETF in complex with partner proteins have shown that ETF partitions the functions of partner binding and electron transfer between (a) a ‘recognition loop’, which acts as a static anchor at the ETF–partner interface, and (b) a highly mobile redox-active FAD domain. Together, this enables the FAD domain of ETF to sample a range of conformations, some compatible with fast interprotein electron transfer. This ‘conformational sampling’ enables ETF to recognize structurally distinct partners, whilst also maintaining a degree of specificity. Complex formation triggers mobility of the FAD domain, an ‘induced disorder’ mechanism contrasting with the more generally accepted models of protein–protein interaction by induced fit mechanisms. We discuss the implications of the highly dynamic nature of ETFs in biological interprotein electron transfer. ETF complexes point to mechanisms of electron transfer in which ‘dynamics drive function’, a feature that is probably widespread in biology given the modular assembly and flexible nature of biological electron transfer systems.
electron transferring flavoprotein ubiquinone oxidoreductase
ferricenium ion (oxidized)
glutaric acidaemia/aciduria type II
medium-chain acyl-CoA dehydrogenase
small-angle X-ray solution scattering
Electron transferring flavoprotein (ETF) is positioned at a key metabolic branch point, and is responsible for transferring electrons from up to 10 primary dehydrogenases to the membrane-bound respiratory chain, the nature and diversity of which vary between organisms . ETFs are highly dynamic and engage in novel mechanisms of interprotein electron transfer, which is dependent on large-scale conformational sampling to explore optimal configurations to maximize electronic coupling. Sampling mechanisms enable efficient communication with structurally distinct redox partners , but require additional mechanisms for complex assembly to impart specificity in the protein–protein interaction.
ETFs are soluble heterodimeric FAD-containing proteins that are found in all kingdoms of life. They contain a second nucleotide-binding site which is usually occupied by an AMP molecule . In bacteria and eukaryotes, ETFs function primarily as soluble one- or two-electron carriers between various flavoprotein-containing dehydrogenases. Electrons are accepted or donated to ETF via the formation of transient complexes with their partners . Almost all ETFs are mobile carriers containing a flexible domain essential for function . ETFs need to balance promiscuity with specificity in their interactions with protein donors and acceptors, in keeping with their function in respiratory pathways. In this review, we discuss new aspects of the structure and mechanism of ‘typical’ ETFs, and explore the diversity in function and structure of ETFs across kingdoms. Finally, we analyse, in the context of new structural information, the role of clinical mutations in human ETFs and their partner proteins that give rise to severe metabolic diseases.
ETFs across kingdoms interact with a variety of electron donors/acceptors that are involved in diverse metabolic pathways. ETFs belong to the same families of α/β-heterodimeric FAD-containing proteins [5–7]. Members of these families can be divided roughly into three groups based on sequence homology and functional types.
Group I ETFs are a well-studied group of electron carriers, typically found in mammals and a few bacteria. Mammalian ETFs are physiological electron acceptors for at least nine mitochondrial matrix flavoprotein dehydrogenases [4,8]. These dehydrogenases include the chain length-specific acyl-CoA dehydrogenases (e.g. medium-chain acyl-CoA dehydrogenase, MCAD) involved in fatty acid β-oxidation, isovaleryl-CoA dehydrogenase, 2-methyl branched-chain acyl-CoA dehydrogenase, glutaryl-CoA dehydrogenase involved in amino acid oxidation, as well as dimethylglycine and sarcosine dehydrogenases involved in choline metabolism [4,8]. Electrons are passed from these primary dehydrogenases through ETF to membrane-bound ETF ubiquinone oxidoreductase (ETFQO) [9,10].
Another well-studied group I ETF is from the bacterium Paracoccus denitrificans[11–13]. It is capable of accepting electrons from P. denitrificans glutaryl-CoA dehydrogenase, in addition to the butyryl-CoA and octanoyl-CoA dehydrogenases from pig liver. The physiological electron acceptor for ETF has been found to be ETFQO .
Group II ETFs are homologous to the proteins FixB and FixA, equivalent to α-ETF and β-ETF, respectively, which are found in nitrogen-fixing and diazotrophic bacteria . These ETFs are often electron donors to enzymes such as butyryl-CoA dehydrogenase, and may also accept electrons from donors such as ferredoxin and NADH . No ETF-dependent activity has been observed with the membrane-bound respiratory enzymes in nitrogen-fixing bacteria, and so it is thought that the electron transfer pathway from ETF to dinitrogen is via the enzymes ETF:ferredoxin oxidoreductase, ferredoxin, nitrogenase reductase and nitrogenase .
A well-studied group II ETF is from the bacterium Methylophilus methylotrophus strain W3A1, which contains only one known dehydrogenase partner, namely trimethylamine dehydrogenase (TMADH) [3,16]. FixB/FixA proteins have been characterized from the microaerobic Azorhizobium caulinodans, which is known to accept electrons from pyruvate dehydrogenase under aerobic conditions . The nitrogen-fixing organism Bradyrhizobium japonicum contains two sets of ETF-like genes: one with high homology to group I ETFs (etfSL), and the other very similar to group II FixB/FixA proteins . Under aerobic conditions, only the etfSL genes are expressed, whereas the reverse is true for anaerobic growth, as nitrogen fixation only occurs anaerobically .
One ETF from the anaerobe Megasphaera elsdenii (formerly Peptostreptococcus elsdenii) is unusual, as it contains two FAD-binding sites per ETF molecule, and so does not bind AMP [6,15,18,19]. This ETF serves as an electron donor to butyryl-CoA dehydrogenase via its NADH dehydrogenase activity , and is an electron acceptor for d-lactate dehydrogenase . It has also been shown to contain a low percentage of the modified flavins 6-OH-FAD and 8-OH-FAD .
Group III ETFs include a pair of putative proteins, YaaQ and YaaR, located adjacent to the cai operon, which encodes carnitine-inducible proteins in Escherichia coli. Group III members will not be discussed further in this review.
An examination of the databases of genomic sequences shows organisms containing multiple ETF-like genes as well as ETFs fused with other proteins (Pedant; http://pedant.gsf.de). The genome of the eubacterium Fusobacterium nucleatum ssp. nucleatum (ATCC 25586) suggests the presence of two complete ETF molecules, each positioned upstream of an acyl-CoA dehydrogenase. The genome also contains a large ORF (GI:19704756; Pedant; http://pedant.gsf.de) containing a fusion of three proteins comprising an N-terminal short-chain acyl-CoA dehydrogenase, followed by the α-subunit only of ETF and a C-terminal rubredoxin (Fig. 1). As no functional studies of this enzyme have been published, it is presumed that the absence of the β-ETF subunit is a result of its role as a ‘fixed’ electron carrier, although flexibility within the multidomain complex may be possible.
Another example of an organism with multiple ETF content is the iron-reducing, nitrogen-fixing bacterium Geobacter metallireducens (Pedant; http://pedant.gsf.de). At least three of the sets of ETF genes are unusual (e.g. ORF4) as the N-terminal portion of the α-ETF subunit contains the gene sequence encoding a [4Fe−4S]2+/+ ferredoxin domain (Fig. 1). These ETFs are found upstream of genes such as putative Fe–S oxidoreductases (Pedant; http://pedant.gsf.de). At least nine other putative [4Fe−4S]2+/+ ferredoxin-containing ETFs have been identified (NCBI blast; http://www.ncbi.nlm.nih.gov/BLAST).
Many archaea contain ETF- or FixB/A-like sequences, such as Archaeoglobus fulgidus DSM 4304, Pyrobaculum aerophilum st. IM2, Aeropyrum pernix and Thermoplasma volcanium st. GSS1, but these are absent in methanogens (Pedant; http://pedant.gsf.de). Several genera, such as Thermoplasma and Sulfolobus, contain multiple ETF genes, including a fusion protein of the two subunits, with the β-subunit at the N-terminus (βα-ETF). In Sulfolobus solfataricus, βα-ETF is found in an operon-like cluster of genes containing the primary dehydrogenase 2-oxoacid ferredoxin oxidoreductase, a putative ferredoxin-like protein and a FixC-like protein, homologous to the membrane-bound ETF ferredoxin oxidoreductase in nitrogen-fixing organisms .
A blast search of the structurally equivalent N-terminal (non-FAD-binding) α-ETF and β-ETF sequences against known ORFs showed homology with a variety of adenosine nucleotide-binding enzymes (NCBI blast; http://www.ncbi.nlm.nih.gov). Such enzymes include members of the adenosine nucleotide α-hydrolase superfamily from Oryza sativa, which contains an ATP-binding fold . The thiamine biosynthesis-like protein from three Leishmania species contains β-ETF and aminotransferase components at the N- and C-termini, respectively . This class of enzyme is known to bind ATP. Other ATP-binding enzymes with homology to β-ETF in the database (NCBI blast; http://www.ncbi.nlm.nih.gov) include adenylyl-sulfate kinase from Anaeromyxobacter sp. Fw109-5 (GI:121539501), the predicted glutamate-dependent NAD(+) synthase from Strongylocentrotus purpuratus (GI:115971088) and the asparagine synthase from Desulfovibrio vulgaris ssp. vulgaris DP9 (GI:120564303). As β-ETF typically binds AMP, homology to domains of other enzymes known to bind adenosine nucleotides is not surprising.
Sequence homology of ETFs
An alignment of α- and β-ETFs from all kingdoms of life (Fig. 2) shows that, within the α-ETF family, the overall sequence homology is low, although high sequence homology is found in the C-terminal region. By contrast, in the β-ETF family, there is a similar degree of sequence similarity throughout the length of the protein. Group I ETFs align better than group II ETFs, although both groups contain significant sequence similarity in conserved regions.
The C-terminal portion of α-ETF contains a highly conserved region, known as the β1αβ2 region of FAD enzymes, which binds the adenosine pyrophosphoryl moiety of FAD . Within this region is the α-ETF consensus sequence of PX[L,I,V]Y[L,I,V]AXGISGX[L,I,V]QHX2G , similar to the consensus sequence for FAD-binding dehydrogenases of GXGXXGX15[E/D]. The β-ETF family contains a conserved signature sequence of VXRX2[E,D]X3[E,Q]X[L,I,V]X3LP[C,A][L,I,V]2 which is used to identify members of the β-ETF family . Adjacent to this signature sequence, group I β-ETFs also show the highly conserved region of DLRLNEPRYA[S/T]LPNIMKAKKK (residues 184–204; human numbering), containing the recognition loop and the highly conserved L195 necessary for partner binding in humans . The group II β-ETF from M. methylotrophus also contains a recognition loop and the highly conserved L193 partner binding to TMADH . Other group II members appear not to contain a significant group I-like recognition loop, suggesting a different mode of partner binding.
Structure of ETF
Domains of ETF
The three-dimensional structures of group I ETFs have been solved from humans (Fig. 3A)  and P. denitrificans, and group II ETF from M. methylotrophus (W3A1; Fig. 3B) . The structure of the P. denitrificans ETF is nearly identical to human ETF, with the major difference being a random loop between residues β90–96 which is an α-helix in humans . All three structures can be divided into three distinct domains. Domain I is composed of mostly the α-subunit, whereas domain III is made up entirely of the β-subunit . These domains share nearly identical polypeptide folds related by a pseudo-twofold axis, in spite of a lack of sequence similarity. Both domains I and III are composed of a core of a seven-stranded parallel β-sheet, flanked by solvent-exposed α-helices. These domains also contain a three-stranded antiparallel β-sheet with a fourth strand coming from the opposite domain. Together these two domains form a shallow bowl shape, and make up the ‘rigid’ or more static part of the molecule upon which domain II rests. Domain III contains a deeply buried AMP molecule which plays a purely structural role .
Domain II is the FAD-binding domain, and is attached to domains I and III by flexible linker regions (Fig. 3) . Domain II can be subdivided into two domains, IIα and IIβ, which are composed of the C-terminal portions of the α- and β-subunits, respectively. Domain IIα is the larger of the two, folds in a manner similar to bacterial flavodoxins  and forms most of the region that binds FAD. This is the region of high sequence similarity within the α-subunit. This fold consists of a core of a five-stranded parallel β-sheet surrounded by alternating α-helices . A sixth strand of the β-sheet is provided by the β-subunit. FAD is bound in an orientation in which the isoalloxazine ring is situated in a crevice between domains II and III, with the xylene portion pointed towards the β-subunit. By contrast, domain IIβ does not interact with FAD, but instead wraps around the lower portion of domain IIα near domains I and III .
Despite the low sequence similarity between the two groups of ETF, the overall folding of the structures is very similar, with the exception of the orientation of the flavin-binding domain. Domain II of W3A1 ETF is rotated by about 40° relative to the human and P. denitrificans flavin domains, with Vα190 and Pβ235 (W3A1 numbering) serving as hinge points . In human ETF, the conserved Eβ165 of domain III interacts with Nα259, which is located near the conserved Rα249 (Rα237 in W3A1) and FAD (Fig. 4A). There are also hydrophobic interactions between the C7- and C8-methyl groups of the isoalloxazine ring of FAD and residues Fβ41 and Yβ16, respectively, of domain III . These interactions are likely to transiently stabilize the flavin domain in this position . Sequence alignments show that Eβ165 (human numbering, Fig. 1) is highly conserved amongst mostly group I ETFs, including P. denitrificans ETF (Eβ162), which also contains the flavin domain in the same position as humans. This suggests that this may be a common orientation of the flavin domain amongst group I members.
As a result of the change in orientation of the flavin domain in W3A1 ETF, Eβ163 (equivalent to human Eβ165) interacts instead with the conserved Rα237 via a bifurcated salt bridge (Fig. 4B) . This arginine residue also forms a single salt bridge with Dα241 of domain II. A second interaction between these two domains is seen in the low-resolution W3A1 ETF structure , between residues Rα211 and Eβ37. In humans, the equivalent arginine residue, Rα223, interacts directly with the flavin and is over 8 Å from domain III .
Solution structure of free ETF
Small-angle X-ray solution scattering (SAXS) studies carried out on human, P. denitrificans and W3A1 ETFs have shown that the solvent envelopes of each ETF are almost identical, in spite of the different conformations of domain II . A superimposition of the solvent envelope of W3A1 ETF onto the structure of its free ETF shows that, although domains I and III fit well, the envelope around domain II shows the existence of multiple conformations in solution (Fig. 4C) . These conformations appear to arise from domain II rotating about 30–50° with respect to domains I and III via two flexible hinge regions. This corresponds to a shift in position of domain II from the W3A1 position to the human/P. denitrificans position. The lack of an appropriate shoulder in the intermediate angle range, which can be associated with the static lobed domain structures, suggests that all three ETFs possess similar domain arrangements in solution, with the flavin domain sampling a range of conformational states. These states are likely to include multiple discrete, but transient states. A superimposition of W3A1 ETFs with different flavin domain positions, modelled by weighted masses molecular dynamics, has shown that these conformations are consistent with the solvent envelope of ETF . The solvent envelopes of both oxidized and reduced W3A1 ETF are essentially identical, suggesting that no large conformational change occurs as a result of changing the redox state . The conformations seen crystallographically may have arisen from the trapping of a particular discrete state as a result of crystal packing constraints, but may also reflect differences in the proportions of the discrete states between the different ETFs .
The isoalloxazine rings of FAD from human and W3A1 ETFs are sandwiched between several conserved residues that make distinct, but structurally equivalent, interactions (Fig. 5A) [1,3]. A key characteristic of ETF FAD-binding domains is the ‘bent’ conformation of the ribityl chain of FAD as a result of 4′OH hydrogen bonding with N1 of the isoalloxazine ring . It is thought that the 4′OH group helps to stabilize the semiquinone/dihydroquinone couple, and may be involved in electron transfer to ETFQO. Another characteristic feature is the absence of aromatic residues that stack parallel to the ring. One or two aromatic residues (Yβ16 and Fβ41 in humans) are within hydrophobic interaction distance, but the rings are not oriented towards FAD. In its place the guanidinium portion of the side chain of the conserved Rα249 is perpendicular to the xylene portion of the isoalloxazine ring, which may function by stabilizing the anionic reduced FAD , and also by conferring a kinetic block on full reduction to the dihydroquinone . Other key interactions include the N1 residue of Hα268 with O2 of the isoalloxazine ring, which may also function in stabilizing the anionic semiquinone . The hydroxyl group of Tα266 interacts with N5 of FAD, which may aid in modulating the redox potential. The ADP moiety of FAD is solvent exposed, more so in W3A1 ETF . Stabilization of the negative charge imposed by the phosphates is achieved through interactions with residues such as Sα248 and Sα281 .
The AMP-binding sites of all three ETF structures are very similar, both in terms of the position and types of interaction between AMP and β-ETF. AMP is buried deeply within domain III and is thought to play a purely structural role (Fig. 5B) . These interactions are mostly backbone interactions; thus, although there is a high degree of conservation of position of the interacting residues, there is often a low sequence conservation (Fig. 2; blue residues). The phosphate moiety of AMP from humans forms hydrogen bonds with the residues Aβ126, Dβ29, Nβ32, Qβ33 and Tβ34, as well as a water molecule. A few hydrogen bonds are found to anchor the rest of the AMP molecule, including backbone interactions with Cβ66 and Aβ9 and two water molecules . It is thought that AMP binding may be a structural remnant of a NADP-binding site, which is a known electron donor of the group II ETF from Megasphaera elsdenii, which does not bind AMP .
Structure of ETF–partner complexes
Methylophilus methylotrophus TMADH:ETF
The first structure of an ETF in complex with its partner protein was solved between TMADH and ETF from M. methylotrophus W3A1 . The structure of the free TMADH dimer had been solved previously, and was shown to contain the redox-active cofactors 6-S-cysteinyl FMN and [4Fe−4S]2+/+ (electron donor to ETF), as well as a purely structural ADP molecule (Fig. 6A) [26,27]. Two crystal forms were obtained for the wild-type complexes, which were found to be virtually identical, suggesting that the structure is largely independent of crystal packing contacts. The total buried interfacial surface visible in the structures was elongated in shape and covered 1750 Å2, with 10% and 8% of the surface contributed by ETF and TMADH, respectively . Surprisingly, there was a complete absence of density for the mobile flavin domain of ETF, in spite of SDS-PAGE analysis of the TMADH:ETF crystals showing its presence .
The structures showed that there was an interaction site between the two proteins, which was distinct from the predicted location of the flavin-binding domain of ETF . This consists of a hydrophobic interaction between a surface patch in the ADP-binding domain of TMADH and a loop in ETF domain III (residues Pβ189–Iβ197), termed the ‘recognition loop’ (Fig. 6B). This loop consists of the N-terminal portion of an α-helix and part of the preceding loop. A residue key to this interaction is the ETF residue Lβ194 (red sphere in Fig. 3), which is buried within this hydrophobic patch of TMADH. Other hydrophobic residues of ETF interacting with TMADH are Yβ191, Iβ197 and Sβ193, the latter of which stabilizes the initial turn of the α-helix in the recognition loop. These residues are highly conserved, in particular within group I ETFs (Fig. 1). Several residues preceding Yβ191 which do not contact TMADH are also conserved, including Lβ186, Nβ187, Pβ189 and Rβ190. The recognition loop is stabilized by both the close packing of these residues and a bifurcating salt bridge between Rβ190 and residues Eβ44 and Eβ51. Several other residues involved in complex formation include a salt bridge between the N-terminus of TMADH and Dβ16 of ETF, and a number of direct or water-mediated hydrogen bonds. This relatively small number of interactions helps to explain why the dissociation constant (∼ 5 µm) of TMADH:ETF is weak [3,28].
In free ETF, the recognition loop is more flexible and is oriented slightly differently, with Pβ189 and Pβ204 serving as hinge points . Limited trypsin proteolysis, which removed the recognition loop, produced an ETF whose structure and redox capabilities with dithionite were virtually identical to native ETF, yet it had lost its ability to accept electrons from TMADH. This shows the pivotal role of the recognition loop in complex formation, and serves as an ‘anchor’ distant to the redox centres . This anchor may serve as a means of recognizing specific redox partners, as all that would be required would be a suitably placed hydrophobic patch to interact with the recognition loop .
The absence of density for the flavin domain of ETF occurs after residues Vα190 and Pβ235, which serve as hinge points . This total lack of density was initially surprising, as the free ETF structure showed clear density for the flavin domain, in spite of the known flexibility of the molecule in solution from SAXS studies . This suggests that either the flavin domain has an increased mobility within the complex, or packing constraints with the free ETF structure lock the domain in one position. This mobility of the flavin domain within the complex lends support to the transient nature of the electron transfer-competent state, as predicted from kinetics and other studies [4,25].
Several mutant TMADH:ETF complexes were designed which altered the interactions between the flavin domain and domain III of ETF, as well as its interaction with TMADH (see ‘Human MCAD:ETF’ section below). At least two of each of the mutant complex structures were determined, TMADH WT:ETF Eβ37Q and TMADH Y442F:ETF WT, including two structures in a new space group (H. S. Toogood, D. Leys & N. S. Scrutton, unpublished results). All structures were virtually identical to the wild-type complex, including the absence of the flavin domain, highlighting the rapid mobility of this domain.
Modelling studies in which the flavin domain of ETF was docked into the TMADH:ETF complex, based on its position in free ETF, showed that the flavin domain had to undergo a significant conformational change to prevent clashes with TMADH [3,4]. This is supported by the detection of structural changes on complex formation by observing spectral changes during difference spectroscopy studies of TMADH:ETF . Shifting the domain into a human-like conformation would allow the domain to fit within the allowable space. The ‘empty volume’ observed between TMADH and ETF is of sufficient size and shape to allow the flavin domain of ETF to undergo a ‘ball-in-socket’ type of motion , suggesting that multiple (> 2) conformations are possible. This suggests an ‘induced fit’ model for partner association, with electron transfer likely to be possible from an ensemble of thermodynamically metastable complexes rather than one discrete species .
Kinetics studies have shown that, in the electron transfer-competent state, the flavin of ETF is likely to be close to a surface groove of TMADH close to residues V344 and Y442 . Molecular dynamics calculations were performed on the flavin domain of free ETF superimposed onto the complex to determine potential electron transfer-competent states . A model of one of the putative ‘active’ conformations between the [4Fe−4S]2+/+ centre of TMADH and the flavin domain of ETF gives an intercofactor distance of less than 14 Å (Fig. 6C) . In this state, the guanidinium ion of the conserved Rα237 is located close to the aromatic ring and hydroxyl group of Y442 of TMADH. Cross-linking studies using bismaleimidohexane with TMADH Y442C and ETF Rα237C mutants led to the rapid formation of a cross-linked complex, establishing the close contact of these residues in the complex. Also, difference spectroscopy studies with TMADH and the ETF mutant Rα237A showed that electron transfer was severely compromised as a result of a change in the rate of rearrangement of ETF to form the electron transfer-competent state, rather than a change in the intrinsic rate of electron transfer . However, any interactions between TMADH and the flavin domain of ETF are likely to be fleeting, and simply increase the half-life of the electron transfer-competent states to allow fast electron transfer .
To investigate the way in which ETF can interact with its structurally distinct partners, the structure of human ETF with its partner MCAD was determined . The structure of free MCAD had been solved previously, and was shown to be a homotetramer of 43 kDa monomers (dimer of dimers) containing one FAD per monomer . The first structure of the complex between MCAD and ETF was found to contain a tetramer of MCAD with one ETF molecule . The total buried interfacial surface visible in the structures (excluding the ETF flavin domain) was elongated in shape and covered 536 Å2, with 3.2% and 4.3% of the surface contributed by ETF and MCAD, respectively. In this structure, the flavin domain of ETF was barely visible in the density .
Four mutant MCAD:ETF complexes were designed which altered the interactions between the flavin domain and domain III of ETF (MCAD:ETF Eβ165A), as well as its interaction with MCAD (MCA D:ETF Rα249A; MCAD E212A:ETF; MCAD E359A:ETF) . The aim was to alter the ratio of the different conformational states sufficiently to trap discrete flavin domain positions. Kinetic studies of these complexes showed a reduction in electron transfer rates [when using 2,6-dichloroindophenol as the terminal electron acceptor], except for the MCAD:ETF Eβ165A complex, which showed both a dramatic increase in rate and decrease in the apparent Km value. Crystal structures of all four mutant complexes were obtained (Fig. 7A; last three: H. Toogood, A. van Thiel, D. Leys & N. S. Scrutton, unpublished work), which showed an increase in density for the flavin domain to about 70% occupancy (except for MCAD:ETF Rα249A), with the flavin domain in the same position as in the wild-type structure. In these structures, ETF is interacting with a dimer of MCAD .
As with the TMADH:ETF structures, human ETF contains a recognition loop (Pβ190–Iβ198), including the highly conserved residue Lβ195, which interacts with a hydrophobic pocket on MCAD (Fig. 7B) . The recognition loop interacts with the MCAD surface in such a way that causes an extension of α-helix C of MCAD , with a nearly perfect alignment of the axes and corresponding dipoles of both helices . The side chain of Lβ195 is buried within a hydrophobic pocket formed by α-helices A, C and D of MCAD, and is lined by residues such as F23, L61, L73 and I83. ETF residues which also interact with this pocket include Yβ192, Pβ197, Iβ198 and Mβ199 .
A comparison of the free and complex crystal structures reveals that, although MCAD adopts a nearly identical conformation in both structures, ETF adopts a slightly different backbone conformation with more extensive side chain rearrangements, including Lβ195 . The structure of the free ETF mutant Lβ195A does not show any significant rearrangements of the recognition loop, yet kinetic studies with both MCAD, isovaleryl-CoA dehydrogenase and the structurally distinct partner dimethylglycine dehydrogenase show a severe decrease in electron transfer rates (A. van Thiel, H. Toogood, H. L. Messiha, D. Leys & N. S. Scrutton, unpublished work). Mutations of MCAD, such as L61M, L73W and L75Y, which were designed to ‘fill in’ the binding pocket, were all severely impaired in electron transfer rates with ETF . Microelectrospray ionization mass spectrometry and surface plasma resonance studies showed competitive binding of ETF to acyl-CoA dehydrogenases and dimethylglycine dehydrogenase, suggesting similar or closely overlapping binding sites for each . Cross-linking experiments with ETFQO showed that it preferentially interacts with the β-subunit of ETF . These results suggest a similar mode of interaction between ETF and its structurally distinct partners .
An alignment of MCAD-like partners shows very little sequence conservation of the residues interacting with the recognition loop . However, the amino acid substitutions tend to retain their hydrophobic or hydrogen-bonding ability, suggesting that ETF does not have to recognize an exact binding pocket, but a structurally equivalent one. The high conservation of the recognition loop, particularly in group I ETFs, suggests that ETFs across kingdoms may also interact with their partners in a similar manner via a recognition loop .
The orientation of the flavin domain within the MCAD:ETF complex is dramatically different from its position in any of the free ETF structures (Fig. 7C) . The contact surface between MCAD and the flavin domain is about 330 Å2, with a shape complementarity value of 0.56, suggesting that the interaction is weak and of a transient nature . Within this interface, Rα249 of the flavin domain forms a salt bridge with E212 of MCAD, as well as interacting with E359 via a bridging water molecule. This is in agreement with chemical modification studies, which show that an arginine residue in ETF and carboxylates on MCAD are involved in complex formation . Other interactions between ETF and MCAD include direct hydrogen bonds between Qα285/N354, Qα265/E359 and a phosphate of ETF FAD/Q163, respectively . The smallest distance between the isoalloxazine rings of the two FAD molecules is 9.7 Å, suggesting that this is an electron transfer-competent state. The indole group of MCAD W166 is positioned between the isoalloxazine rings, and is within van der Waals' contact with both the C7 and C8 methyl groups of ETF FAD .
The complex structure shows that electrostatic interactions are essentially absent from the interface, yet it is known that the electron transfer rate decreases with increasing ionic strength . These observations could be a result of the destabilization of the protein–protein interaction between E212 and Argα249. Alternatively, these results may arise from enhanced hydrophobic interaction at high ionic strength involving the hydrophobic patch/recognition loop. The concomitant decrease in the rate of complex dissociation following electron transfer might lead to the observed reduction in steady-state turnover .
Although there are no structural similarities between TMADH and MCAD, ETF interacts in a similar manner with both proteins . This is a result of the recognition loop interacting with distinct, but structurally equivalent, hydrophobic patches on the partners, which creates a near-identical volume and shape of the space occupied by the flavin domain of ETF. The relative positions of the docking sites for the leucine anchoring residue within the recognition loop between the two complexes are very similar. However, the two partner proteins interact with ETF via different redox cofactors, with the electron-donating cofactors in different relative positions within the two complex structures. This highlights the need for the flavin domain to sample the available conformational space to find an electron transfer-competent state, as seen by the lack of density for the flavin domain in both wild-type structures. These conformations are transiently stabilized through key interactions between conserved residues specific to each dehydrogenase type . As both the [4Fe−4S]2+/+ and FAD cofactors of TMADH and MCAD, respectively, are located within a 10 Å radius of the ETF FAD, this suggests that a similar conformation of ETF in both complexes is possible for fast interprotein electron transfer.
Kinetics of electron transfer between ETF and partners
Methylophilus methylotrophus TMADH:ETF
TMADH is a 166 kDa homodimeric iron–sulfur flavoprotein which catalyses the oxidative demethylation of trimethylamine (TMA) to form dimethylamine and formaldehyde (Eqn 1). Substrate oxidation is accompanied by the transfer of reducing equivalents, first to the covalently bound cofactor 6-S-cysteinyl FMN , followed by reduction of a ferredoxin-like [4Fe−4S]2+/+ located approximately 4–6 Å from the 8-α-methyl group of FMN . The physiological terminal electron acceptor of TMADH from M. methylotrophus is ETF, with electron transfer from the [4Fe−4S]2+/+ centre occurring via quantum electron tunnelling [37,38]. Stopped-flow kinetics studies of the reductive half-reaction shows that it occurs in three kinetic phases. The fast phase represents the two-electron reduction of 6-S-cysteinyl FMN, followed by intermediate and slow phases which reflect the transfer of one electron from the dihydroquinone of flavin to the [4Fe−4S]2+/+ centre, and the formation of a spin-interacting state between the flavin semiquinone and the reduced [4Fe−4S]2+/+. This latter state is formed after the binding of a second substrate molecule, which induces the ionization of Y169 located close to the pyrimidine ring of 6-S-cysteinyl FMN . This state is distinguished by a complex EPR signal centred near g ∼ 2 with an unusually intense half-field g ∼ 4 signal . However, the kinetics are further complicated as the extent of the biphasic nature changes with both substrate concentration and pH . Detailed kinetic and mechanistic analyses of the reductive half-reaction have been studied extensively, and readers are referred to papers such as Scrutton et al. , Scrutton and Sutcliffe , Roberts et al., Basran et al. [42–46], and references cited therein.
TMADH:ETF oxidative half-reaction
The oxidative half-reaction of TMADH involves the transfer of two electrons through [4Fe−4S]2+/+ to the ETF FAD in two single electron transfer steps. The midpoint reduction potential of the oxidized flavin/semiquinone couple of ETF (Eox/sq) is unusually positive (+ 153 mV) , but is consistent with the need to accept electrons from the [4Fe−4S]2+/+ centre of TMADH, which has a 4Fe−4S2+/4Fe−4S+ potential of + 102 mV . This highly positive redox potential of ETF suggests exceptional stabilization of the anionic semiquinone, most probably because of the location of the guanidinium group of the conserved Rα237 over the si face of the flavin isoalloxazine ring. Conversion of FAD to the dihydroquinone form is incomplete, with a midpoint potential of the semiquinone/dihydroquinone couple of less than − 250 mV, as a result of the presence of both kinetic and thermodynamic blocks on full reduction of FAD .
Recent mutagenic studies have shown the importance of Sα254 as a hydrogen bond donor to the N(5) atom in the oxidized state of the flavin [48,49]. Mutation of Sα254 to threonine and cysteine abolished the kinetic barrier to dihydroquinone formation, as well as significantly decreasing the Eox/sq midpoint potential. Changes in the observed Kd values showed that, although the mutations destabilized the oxidized state of the flavin, the anionic semiquinone state was also destabilized to a much greater extent. Thus, Sα254 plays a key role in establishing the high Eox/sq value and the unusually high stability of the anionic semiquinone of ETF .
Further studies have suggested that the redox properties of Eox/sq of ETF may be perturbed on complex formation with TMADH . In the presence of TMADH, ETF can be fully reduced to the dihydroquinone species by dithionite, with estimated redox potentials of + 196 mV and 0 mV for the Eox/sq and Esq/hq couples, respectively. This change in redox potentials of ETF is thought to be a result of a conformational change in the flavin-binding domain on complex formation .
Steady-state kinetic parameters for the electron transfer between the [4Fe−4S]2+/+ centre of TMADH and ETF flavin give kcat and Km values of 16.8 ± 0.5 s−1 and 14.8 ± 1.2 µm, respectively, at 25 °C . Modelling studies of the complex between TMADH and ETF, as well as kinetics studies of TMADH mutants, revealed the existence of a small surface groove on TMADH which may accommodate the isoalloxazine ring of FAD bound to ETF . These studies revealed the existence of two possible routes of electron transfer from the [4Fe−4S]2+/+ centre to an external electron acceptor. The shortest pathway extends from C345, a ligand on the [4Fe−4S]2+/+ centre, to V344, which is located at the bottom of a small groove on the surface of TMADH. The second pathway extends from C345 to E439 and finally to Y442, the latter of which forms one side of the groove on the surface of the enzyme .
The steady-state kinetic parameters for five mutations of V344 and four mutations of Y442 of TMADH with ETF were determined . The Michaelis constant Km was largely unaffected by the mutations, except for Y442G, which showed a fivefold increase, and the effect on kcat was minimal for V344 mutants (at most twofold for V344I). By contrast, Y442 substitutions had a more noticeable effect, particularly with smaller and less bulky Y442C and Y442G mutations, which resulted in 19- and 31-fold reductions in turnover number, respectively. The steady-state kinetic parameters of these TMADH mutations were also determined with ferricenium ion (Fc+) as the electron acceptor. In this case, mutations of V344 had quite dramatic effects, with substitutions of V344 to cysteine, alanine or glycine causing a significant reduction in the apparent Km value and increase in kcat/Km, but only a modest overall increase in kcat. The kinetics of the reductive half-reaction of these mutants showed only small changes in the rate of intramolecular electron transfer from 6-S-cysteinyl FMN to the [4Fe−4S]2+/+ centre, which may reflect minor structural changes around the [4Fe−4S]2+/+ centre .
Stopped-flow studies on the kinetics of transfer of electrons from two-electron-reduced TMADH to oxidized ETF revealed complex multiphasic kinetics . To simplify these studies, the 6-S-cysteinyl FMN cofactor of TMADH was inactivated by phenylhydrazine, rendering it inert to reduction/oxidation. This allows TMADH to be reduced anaerobically to the one-electron state, via titration with dithionite, with the electron located on the [4Fe−4S]2+/+ centre . Fast reaction studies of this one-electron-reduced modified TMADH with ETF eliminates the complications arising from internal electron transfer in TMADH . Initial studies of the kinetics of the oxidative half-reaction with ETF were carried out at 25 °C, and showed a hyperbolic dependence on ETF concentration, which exhibited saturation behaviour . However, more recent rigorous studies of the reaction kinetics carried out at 5 °C, slowing the reaction sufficiently to obtain more precise data, demonstrated the biphasic nature of the transients and a linear dependence of ETF concentration on the rate . A recent attempt was made to reinstate the saturation behaviour of the electron transfer rate on ETF concentration. However, this study omitted to show data for reactions carried out at 5 °C, and there was no rigorous attempt to analyse changes in the nature of reaction transients at different temperatures .
Both wild-type TMADH and mutants of V344 showed a linear dependence on ETF concentration at 5 °C, with a second-order rate constant of 1.44 × 106 M−1·s−1 for the native enzyme . By contrast, mutants of Y442 displayed saturation behaviour on ETF concentration, with saturation behaviour with respect to ETF detected with mutants Y442F, Y442L and Y442G . The dissociation constants calculated for these Y442 mutant complexes varied from 6 to 46 µm, compared with 3–7 µm for the native complex, the latter determined by analytical ultracentrifugation . By contrast, the reactions of the mutants and native TMADH with Fc+ all showed a linear dependence on Fc+ concentration at 25 °C. Val344 substitutions to cysteine, alanine or glycine showed a moderate increase in second-order rate constants, whereas the opposite was true for the bulkier substitutions tyrosine and isoleucine, with the latter showing a nearly 20-fold reduction. This could be the result of a change in the length of the electron transfer distance and/or changes in packing density. These mutations were found to have little or no effect on the binding or limiting rate constant (klim) for oxidation of the substrate. These stopped-flow and steady-state results suggest that electron transfer to ETF proceeds via the longer pathway through Y442, whereas electron transfer to Fc+ is via the shorter route through V344, as Fc+ is likely to penetrate the groove more fully than the ETF flavin. Substitutions at Y344 showed that shortening the length of the side chain increased the electron transfer rates to Fc+, presumably by reducing the pathway length. Mutations at Y442 possibly disrupt electron transfer by perturbing the interaction geometry of TMADH and ETF in the complex, leading to less efficient coupling between the [4Fe−4S]2+/+ centre and FAD [30,53].
Intrinsic rate of electron transfer to ETF
Given the highly dynamic nature of ETF, a kinetic model of intermolecular electron transfer incorporating intermediate state(s) representing flavin domain motion is needed. These intermediate states include the large change in position of domain II on complex formation  and small-scale conformational changes in the formation of electron transfer-competent state(s). A simplified kinetic scheme for such a system, where A is one-electron-reduced TMADH (4Fe−4S+) and B is oxidized ETF, is shown in Scheme 1. In this scheme, kr (and k–r) refer to the reversible rate of reorganization of the flavin domain to form an electron transfer-competent state:
( (Scheme 1))
As native and V344 mutants of TMADH do not display saturation kinetics with ETF at 5 °C, this model predicts that complex formation is rate limiting . Thus, both keT and any possible conformational reorganization of the ETF flavin domain following complex formation are predicted to be fast. As the Y442 mutants display saturation kinetic behaviour, this suggests that either keT or the rate of conformational reorganization has been dramatically reduced. Given that the observed limiting rates of electron transfer for Y442 mutants are relatively slow, the latter is possibly more likely .
Simulations have shown that, for keT values above 103 s−1, there is essentially a linear relationship between kobs and ETF concentration . The values of kobs obtained from the simulations were very similar to those obtained experimentally for native and Y442 mutant enzymes. The switch to saturation behaviour was seen only when keT was less than ∼ 103 s−1. Such low predicted keT values for Y442 mutants are not likely to correspond to intrinsic electron transfer rates for transfers occurring over 11–12 Å, and so rate limitation is likely to be the result of impaired structural reorganization during complex assembly. In native TMADH, Y442 may enhance the rates of reorganization of the electron transfer complex by a direct interaction with ETF via its phenolic hydroxyl group. Disruption of favourable interactions by Y442 mutants could thereby alter the nature of the electron transfer-competent state, such as a change in the [4Fe−4S]2+/+ to ETF FAD distance, leading to a dramatically reduced keT value .
Kinetic scheme of intra- and interprotein electron transfer
A branching kinetic steady-state scheme has been proposed for intra- and interprotein electron transfer of TMADH (Fig. 8) . TMADH is unusual as it shows substrate inhibition at high TMA concentrations. Figure 8 presents a branching kinetic scheme in which TMADH can utilize two alternative catalytic cycles: a 0/2 cycle, in which it cycles between an oxidized and two-electron-reduced enzyme, and a 1/3 cycle, in which it cycles between a one- and three-electron-reduced enzyme . This situation exists because, although the substrate can donate two electrons at a time, the terminal electron acceptor ETF (or phenazine methosulfate) can only accept one electron at a time, and as a result of the presence of both a 6-S-cysteinyl FMN and a [4Fe−4S]2+/+ centre, TMADH can take up as many as three electrons .
Stopped-flow studies of the reaction of TMADH with TMA using diode array detection showed four characteristic spectra of the reductive half-reaction . The states distinguishable are the oxidized enzyme, two-electron-reduced flavin dihydroquinone, two-electron-reduced anionic flavin semiquinone plus reduced [4Fe−4S]2+/+ and, finally, the so-called spin-interacting state. The one- and three-electron-reduced forms are not observed in single turnover studies, although they are likely to be present in steady-state reactions. Direct evidence for the presence of alternative redox cycles was detected with enzyme-monitored turnover experiments with TMA concentrations of 20 µm to 2 mm. At high TMA concentrations, the spectrum indicates that the predominant species at steady state is the one-electron-reduced flavin semiquinone/oxidized [4Fe−4S]2+/+, consistent with the species predicted to accumulate in the 1/3 cycle. At low TMA concentrations, the predominant species is oxidized TMADH, with only a small quantity of flavin semiquinone. Thus, at low substrate concentrations, the 0/2 cycle is predominant and substrate inhibition does not occur. Interestingly, at low Fc+ concentrations (oxidizing substrate), the switch between the 1/3 and 0/2 cycles with a decrease in TMA concentration does not occur. Thus, as the ratio of reducing to oxidizing substrate increases, the level of steady-state enzyme reduction increases .
It is thought that the 1/3 cycle is slower than the 0/2 cycle, indicating that it would be predominant only at high TMA concentrations or low ETF/phenazine methosulfate concentrations. This is because substrate binding stabilizes the semiquinone form of the flavin in the one-electron-reduced enzyme . The binding of substrate to the one-electron-reduced [4Fe−4S]2+/+ centre of TMADH, which is likely to accumulate under the high substrate concentrations of the steady-state condition, may result in the redistribution of the reducing equivalents, leading to the formation of a flavin semiquinone (boxed reaction in Fig. 8). In this case, the substrate is unable to be oxidized, as the flavin is unable to accept two reducing equivalents because of the unfavourable equilibrium between the two redox centres in the substrate-bound, one-electron-reduced state. This would cause an apparent substrate inhibition of the reaction without the need for a second inhibitory substrate-binding site .
This model also predicts that the partitioning of the two redox cycles is not dependent on the rate of 6-S-cysteinyl FMN reduction . This is supported by stopped-flow studies with triethylamine, which has been shown to be a poor substrate, yet still displays a clear steady-state inhibition. However, studies with the substrate n-butyldimethylamine (DMButA) show a substantial compromise in flavin reduction, as well as a reduction in substrate inhibition, although it has the tightest binding affinity to oxidized TMADH. It is thought that this lack of substrate inhibition is caused by a weaker binding of DMButA to the reduced enzyme, which could account for the observed low accumulation of the anionic semiquinone form of TMADH .
Alternative stable conformations of ETF
Compounding the problems associated with understanding the dynamics of the interface between TMADH and ETF is the issue of ‘structural imprinting’— a slow conformational change in ETF that is catalysed by interaction with TMADH. These ill-defined structural changes in ETF give rise to an increase in fluorescence emission of the FAD cofactor, and suggest a more ‘open’ structure for ETF in which FAD is more solvent exposed . These relatively slow structural changes associated with imprinting are not to be confused with ‘conformational sampling’. Thus, unlike the structural change imparted through imprinting of ETF by TMADH, conformational sampling is an integral part of the electron transfer mechanism. Others  have incorrectly challenged our interpretation of structural imprinting by suggesting that we have inferred that both structural imprinting and conformational sampling are the same process. This is not the case. The timescales for structural imprinting are far too slow to be associated with electron transfer from TMADH to ETF as observed in stopped-flow studies. Conformational sampling is an intrinsically rapid motion of the FAD domain in the complex that allows the FAD domain to search out electron transfer-competent conformations. The structurally imprinted form of ETF accumulates over an extended time course (typically hours) when as-purified ETF is incubated with TMADH. The precise structural change(s) that occurs during the imprinting reaction is not known, but both fluorescence emission and anisotropy analysis indicate that there is a slow structural change in ETF when incubated with TMADH over extended time periods (H. Messiha, S. E. Burgess, D. Leys & N. S. Scrutton, unpublished results).
MCAD is a 172 kDa homotetrameric flavoprotein that catalyses the α,β-dehydrogenation of acyl-CoA thioesters to their corresponding trans-2,3-enoyl-CoA . Substrate oxidation is accompanied by the transfer of reducing equivalents to the covalently bound cofactor FAD, followed by electron transfer to its physiological terminal electron acceptor ETF. The kinetic scheme of MCAD is very complex and involves several intermediates. In the reductive half-reaction, after formation of the initial Michaelis complex, there are multiple steps incorporating an isomerization step, H+ uptake by the carboxyl group of catalytic base E376, and multiple steps of flavin reduction resulting in a charge transfer complex . This is followed by the oxidative half-reaction, in which electrons are transferred from the enoyl-CoA product–(reduced) enzyme complex to the flavin of ETF in two single-electron steps . Detailed kinetic and mechanistic analyses of the reductive half-reaction have been studied extensively and are beyond the scope of this review, and so readers are referred to a recent review by Ghisla and Thorpe  and references cited therein.
MCAD:ETF oxidative half-reaction
A minimal kinetic scheme for the oxidative half-reaction of MCAD with substrate and ETF has been proposed, which is an example of the general mechanism of acyl-CoA dehydrogenases (ACADs; Fig. 9) [59,60]. Species 1 is the tight binding complex between the enoyl-CoA product and dihydroquinone-reduced MCAD, known as the charge transfer complex. This intermediate is distinct as it has an absorbance maximum of around 570 nm . In the absence of an external electron acceptor, species 1 is converted into a further complex (species 2), which can lead to a slow product release (species 5). The free reduced MCAD can then bind excess substrate (species 6) which does not have charge transfer transitions . Either species 1 or 2 can transfer electrons to form the oxidized MCAD:product complex by two single-electron reductions to two oxidized ETF molecules. Reversible product release completes the catalytic cycle .
Electron transfer between reduced MCAD and oxidized ETF occurs in the presence of bound product for several reasons. Firstly, acyl-CoA thioesters are relatively weak thermodynamic reductants, and so the equilibrium is shifted towards product formation by preferential binding of enoyl-CoA product to the reduced enzyme . A consequence of this is that the product must remain bound until MCAD is reoxidized by ETF. The product-bound complex also has a higher kinetic, but not thermodynamic, reductant ability than free reduced MCAD, as product binding reduces the pK value of the reduced flavin species . Finally, the presence of bound product dramatically reduces the oxidase activity of the enzyme, which prevents the loss of reducing equivalents to non-ATP-generating processes .
The oxidative half-reaction between product-bound pig ACAD and ETF at 3 °C is multiphasic, composed of two rapid phases (t1/2 ∼ 20 and 50 ms) and two slower phases (t1/2 ∼ 1 and 20 s) (inset, Fig. 9) . The first and second phases correspond to the reoxidation of the ACAD dihydroquinone in two successive one-electron steps requiring two molecules of oxidized ETF. A slow decline in absorbance at 370 nm is attributed to further reduction of the semiquinone of ETF by the one- and two-electron-reduced product–ACAD complex. The ETF semiquinone undergoes disproportionation by ACAD in the presence of the product, which contributes to the final attainment of the equilibrium state. In the absence of the bound product, the reduction of oxidized ETF by two-electron-reduced ACAD proceeds much more slowly, with the generation of a blue semiquinone in ACAD, as opposed to a red semiquinone in the presence of the bound product . The increased electron transfer rates in the presence of the bound product is possibly caused by the stabilization of the red anionic ACAD FAD (and possibly the dihydroquinone form also ), or it may aid in the structural alignment of the flavin centres within the complex .
The kinetics of the oxidative half-reaction between human reduced MCAD and oxidized ETF at 3 °C show only a biphasic reaction because of difficulties inherent with these reactions . The fast phase corresponds to the two rapid phases of successive one-electron reduction steps by reduced MCAD to two molecules of oxidized ETF. The slower phase corresponds to the reduction of ETF semiquinone by the two-electron-reduced MCAD–product complex. Studies of human ETF with 2′-deoxy-FAD, instead of unmodified FAD, show that, although the binding constants and overall two-electron redox potential of the two FADs are similar, the potential of the oxidized/semiquinone couple is 116 mV lower than with unmodified FAD (+ 37 mV for the unmodified oxidized/semiquinone couple ). This suggests that the 4′-hydroxy-N(1) hydrogen bond stabilizes the anionic semiquinone by delocalizing the electron over the N(1)–C(2)O region. The turnover rate of ETF with MCAD is significantly reduced with 2′-deoxy-FAD, a reflection of the decreased potential of the oxidized/semiquinone couple .
The role of ACAD:ETF complex formation and reorganization has not been investigated thoroughly in kinetics studies of mammalian systems. However, complex formation is known to be transient, as shown by the Kd values of 2 and 5 µm with dimethylglycine dehydrogenase and short-chain acyl-CoA dehydrogenases, respectively . Mutagenesis of the conserved ETF residue Rα249 to alanine or lysine resulted in less than 10% of the activity with MCAD remaining, with a decreased potential of the oxidized/semiquinone couple to − 39 mV [23,65]. These changes are most probably caused by a decrease in complex formation , as well as a reduction in semiquinone stabilization, as a result of the change of the delocalized positive charge of arginine to a point charge in lysine . Mutations of other residues known to be involved in complex formation, such as MCAD residues E212A and L75Y and ETF mutation Lβ195A, show dramatic losses in electron transfer rates [23,25]. However, it is thought that complex dissociation may not be the rate-limiting step in electron transfer [23,25].
Dynamics drive function: conformational sampling mechanisms for electron transfer
Biological electron transfer complexes are characterized by their transient nature and fast rates of complex dissociation that support rapid interprotein electron transfer . Rapid complex dissociation can occur by having no more than a few weak interactions between the partners, or by having one of the redox cofactors situated in a mobile domain. Both strategies are employed in ETF–partner protein complexes.
The structures of complexes between ETF and its partners show a dual mode of interaction [3,23,25]. The recognition loop binds to a hydrophobic patch on ETF partners, which provides a weak anchoring site between the two partners. This creates a suitable interfacial cavity for the flavin domain to sample a large range of conformational interactions, some of which are compatible with fast electron transfer. Transient stabilization of electron transfer-competent states is achieved through interactions between the two partners, including interactions with the conserved Rα237 (W3A1 numbering) .
This separation of the partner recognition site (recognition loop) from the electron transfer site (flavin domain) is critical in understanding how ETF can interact with specific, yet structurally distinct, partners . The observed specificity of ETF to its partners can be understood by the latter needing only to provide a suitable hydrophobic patch in the correct position to ensure interaction with the recognition loop of ETF. Once a complex is established, the flavin domain is promiscuous in searching out suitable transient electron transfer-competent states that place the redox cofactors within 14 Å of one another, the maximum distance allowed for biologically relevant electron transfer . This flexibility of the ETF flavin domain, combined with the potential of each ETF–partner complex to employ a different strategy to transiently stabilize the electron transfer-competent states, can explain how one ETF can interact with structurally distinct partners .
The known flexibility of ETF, detected both crystallographically and in solution studies, suggests that each ETF flavin domain exists in a population of at least three rapidly interconverting states (Fig. 10) . In state I, Eβ165 interacts with Nα259, which puts the conserved Rα249 in a close position to Lβ185 of domain III, as seen in the free human ETF structure (Fig. 3A). State II ETF is in a position in which this arginine residue interacts with Eβ165, as seen in the free W3A1 ETF structure (Fig. 3B). On complex formation, ETF can also occupy state III, collectively the electron transfer-competent states. Possibly states I and II can also exist within the complex, although a slight reorientation of the latter state is needed. The relative population of these and other potential states is likely to be species dependent, as the range of stabilizing interactions between domain III and the flavin domain varies between ETFs .
States I and II are clearly ‘inactive’ states, although a slightly modified state I can occupy the conformational space within the complex without clashes with its partner . ETF domain III residues, such as Eβ165 (human numbering), are thus stabilizing inactive conformations of the ETF flavin domain. Mutation of Eβ165 to alanine could lead to a loss of a stabilizing interaction in both states I and II, which could potentially shift the equilibrium in favour of state III, the active conformation(s). This could explain the increase in density for the flavin domain in the MCAD:ETF Eβ165A structure, as well as the enhanced electron transfer rates in the presence of 2,6-dichloroindophenol (which is independent of complex dissociation rates). The presence of an equilibrium distribution of conformations of the FAD domain, including stabilization of inactive states, ensures that the electron transfer-competent state (state III) is relatively short lived, thus favouring fast interprotein electron transfer rates .
Inborn errors of metabolism
Because of their role in the catabolism of fatty acids, several amino acids and choline, mutations in mammalian α- and β-ETF, as well as ETFQO, result in the often fatal disease glutaric acidaemia/aciduria type II (GAII), also known as multiple acyl-CoA dehydrogenase dysfunctional disease (MADD) . This disorder differs from glutaric aciduria type I, which arises from defects in glutaryl-CoA dehydrogenase, as this disease results in the large excretion of compounds such as butyric and isovaleric acids . GAII is an autosomal recessively inherited disorder subdivided into IIA, IIB and IIC, depending on which of the three respective genes contains mutations . The mutations can lead to a range from mild to severe cases, with variable presentation times, depending on the location and nature of the mutation. The neonatal-onset forms are usually fatal and are characterized by symptoms such as severe nonketotic hypoglycaemia, metabolic acidosis and excretion of large amounts of fatty acid- and amino acid-derived metabolites. Late-onset GAII symptoms include lethargy, vomiting, hypoglycaemia, metabolic acidosis and hepatomegaly, which tend to be periodic and often preceded by metabolic stress [68,69].
The types of known clinical mutation of α- and β-ETF include single amino acid substitutions, deletions or insertions of bases resulting in the loss of amino acids, early termination or frameshifts, and variable gene products resulting from incorrect splicing (Table 1) [68,70–73]. In some cases, the mutations lead to a decrease in the amount of mRNA transcript and/or protein levels in the cell. Mutations which destabilize either subunit could lead to a reduction in the levels of correctly folded protein in the cell. Some patients completely lack β-ETF transcript or ETF protein, presumably as a result of mutation(s) in the regulatory sequences of the gene that affect expression and/or turnover of the corresponding mRNAs .
Table 1. Clinical mutations of α- and β-ETF. GA, glutaric acidaemia or glutaric aciduria.
a The mutated base is in bold. All base numbering is for cDNA sequences with the initiating Met codon, except for gen which refers to genomic DNA numbering. The number refers to the a base number at the beginning of the codon or the b residue number. c Number refers to the codon number range, with lower case bases the deletions. Additional information was obtained from the Human Gene Mutation Database (archive.uwcm.ac.uk/uwcm/mg/hgmd0.html). d This mutation results in the skipping of exon 3 or in the creation of a downstream cryptic splice site and the insertion of the nine 5′proximal nucleotides of intron 3.
In some cases, the effects of the mutations on ETF have been determined, or can be inferred by the structures of free or complexed human ETF. Expression of the cloned form of the clinically mild ETF mutant Dβ128N is significantly reduced at physiological temperatures because of its lower thermostability . This residue is highly conserved and is located in a cavity near the AMP-binding site, suggesting that it is likely to be important in protein folding. The Tα171I mutant also shows a decreased thermostability, but, when combined with a clinically mild mutant of very long-chain acyl-CoA dehydrogenase, the risk of clinical disease is significantly reduced . Substitutions of the highly conserved glycine residues Gα255V and Gα267R are presumed to affect local folding of the protein, whereas the mutant Gα116R is known to fold into a catalytically inert form [72,74].
The most frequent clinical mutation detected is Tα266M, which forms two interactions with FAD . This mutant shows an altered flavin environment, with a 10-fold increase in stability of the semiquinone form. Thus, although the mutation has little effect on the reaction with acyl-CoA dehydrogenases, the rate of disproportionation of the semiquinone, catalysed by ETFQO, is reduced 33-fold . The mutation Cβ42R probably interferes with the interaction with AMP (O3*-SG), which may influence overall protein folding as a result of the importance of AMP in the overall folding of β-ETF . A mutation likely to affect complex formation between ETF and MCAD is the deletion mutation Kβ204. This residue is close to the recognition loop, and may affect the local structure to the extent that complex formation is impaired but not abolished, as indicated by the mild form of the disease .
In recent years, detailed biophysical analysis, coupled with the determination of the structures of ETF–partner protein complexes, has revealed a novel mode of interprotein electron transfer. Complex formation triggers mobility of the FAD domain, an ‘induced disorder’ mechanism contrasting with the more generally accepted models of protein–protein interaction by induced fit mechanisms. The subsequent interfacial motion of the FAD domain is the basis for the interaction of ETF with structurally diverse protein partners. This motion seeks out optimal geometries and distances for interprotein electron transfer, a mechanism termed ‘conformational sampling’. Given the modular nature of redox proteins, this might be a more general feature of intra- and interprotein electron transfer in biological systems. Similar mechanisms have been proposed for intraprotein electron transfer in the multidomain nitric oxide synthases . In addition, crystal structures of the cytochrome b6f complex have identified a similar, yet distinct, motion of the Rieske iron–sulfur domain compared with that observed for the cytochrome bc1 complex .
The mechanism of conformational sampling identified within the ETF systems contrasts with other modes of protein–protein interaction in which partner binding is to several structurally well-defined and distinct partner proteins. In the case of complexes formed with calmodulin , members of the POU family of DNA-binding proteins , the peptidyl-prolyl cis/trans-isomerase Pin1  and the translocation domain of colicin E9 protein toxin , these proteins have been found to contain highly flexible regions or domains in the uncomplexed state. In these cases, the proteins sample a range of conformations prior to complex formation to enable the recognition of structurally distinct partner proteins. Following complex formation, the mobile regions of these proteins become rigid, and the protein is effectively locked into a single conformational (active) state. This is in essence an induced fit mechanism, which contrasts with the ETF system which requires conformational sampling after complex formation to seek out the electron transfer-competent state.
Work in the authors' laboratory was funded by the UK Biotechnology and Biological Sciences Research Council (BBSRC). NSS is a BBSRC Professorial Research Fellow. DL is a Royal Society University Research Fellow.