Structural basis of interprotein electron transfer in bacterial sulfite oxidation

Interprotein electron transfer underpins the essential processes of life and relies on the formation of specific, yet transient protein-protein interactions. In biological systems, the detoxification of sulfite is catalyzed by the sulfite-oxidizing enzymes (SOEs), which interact with an electron acceptor for catalytic turnover. Here, we report the structural and functional analyses of the SOE SorT from Sinorhizobium meliloti and its cognate electron acceptor SorU. Kinetic and thermodynamic analyses of the SorT/SorU interaction show the complex is dynamic in solution, and that the proteins interact with Kd = 13.5 ± 0.8 μM. The crystal structures of the oxidized SorT and SorU, both in isolation and in complex, reveal the interface to be remarkably electrostatic, with an unusually large number of direct hydrogen bonding interactions. The assembly of the complex is accompanied by an adjustment in the structure of SorU, and conformational sampling provides a mechanism for dissociation of the SorT/SorU assembly. DOI: http://dx.doi.org/10.7554/eLife.09066.001


Introduction
Although electron transfer reactions are key biochemical events, which underpin fundamental processes, such as respiration and photosynthesis, the study of the molecular details of the interprotein interactions at their core can be largely intractable. In particular, atomic resolution crystal structures of electron transfer complexes are rare, due to their fundamentally transient nature (Antonyuk et al., 2013). Electron transfer pathways are made up of chains of redox proteins, which provide a path for the controlled flow of electrons and rely on efficient docking of protein redox partners through noncovalent, dynamic protein-protein interfaces (Moser et al., 1992). Complementary electrostatic surfaces, hydrophobic interactions and dynamics at the protein-protein interface have all been proposed to contribute to efficient interprotein electron transfer (Leys and Scrutton, 2004), with a strong correlation between the driving force for the reaction, the distance between redox centers and the rate of electron transfer (Moser et al., 1992;Marcus and Sutin, 1985).
Interprotein electron transfer processes are central to the redox conversions of cellular sulfur compounds, which are an evolutionarily ancient type of metabolism that has existed as long as cellular life (Schidlowski, 1979;Kappler et al., 2008). Sulfur-containing compounds mediate many crucial reactions in the cell (for example, in coenzyme A, sulfur containing amino acids or glutathione), but their reactivity also makes them potentially toxic (Kappler, 2011). Sulfite in particular, is a highly reactive sulfur compound that can cause damage to proteins, DNA and lipids, resulting in oxidative stress and irreversible cellular damage (Feng et al., 2007). In most cells, the detoxification of sulfite by oxidation to sulfate (Equation 1) is catalyzed by sulfite oxidizing enzymes (SOEs) (Kisker et al., 1997).
SOEs from plants, higher animals and bacteria have been characterized and they all catalyze the same fundamental reaction. However, their cellular functions, catalytic properties (Kappler, 2011;Feng et al., 2007;Kappler and Wilson, 2009;Hille, 2002;Hänsch et al., 2007) and the identities of their natural electron acceptors vary significantly. Some SOEs transfer electrons to oxygen (Schrader et al., 2003), while others interact with redox proteins such as cytochrome c (Kisker et al., 1997;Kappler and Wilson, 2009;Low et al., 2011;Bailey et al., 2009; or as yet unknown cellular components (Kappler, 2011;Low et al., 2011;Bailey et al., 2009;Wilson and Kappler, 2009). To date, three unique crystal structures of SOEs have been reported: from chicken, plant and bacteria, which differ significantly in their domain architectures and redox cofactor compositions. However, none of these studies show details of a SOE in complex with its external electron acceptor (Kisker et al., 1997;Schrader et al., 2003;Kappler and Bailey, 2005). At present, no structural information on the molecular interactions of any of these enzymes with their respective electron acceptors is available and the determinants that dictate the type of electron acceptor individual SOEs employ, while maintaining the efficiency of the basic enzyme reaction are open questions. (Kappler, 2011;Kappler, 2008) Here, we have investigated an electron transfer complex involving the periplasmic SorT sulfite dehydrogenase from the a-Proteobacterium Sinorhizobium meliloti, which represents a structurally uncharacterized type of SOE, and its electron acceptor, the c-type cytochrome SorU (Low et al., 2011;Wilson and Kappler, 2009). In S. meliloti the SorT sulfite dehydrogenase is part of a sulfite detoxification system that is induced in response to the degradation of sulfur containing substrates such as the organosulfonate taurine . Electrons derived from sulfite oxidation are passed on to the SorU cytochrome, and likely then to cytochrome oxidase, as S. meliloti is capable of sulfite respiration (Low et al., 2011). Here, we report the crystal structures of both the isolated SorT and SorU proteins and the biochemical and structural analyses of the SorT/SorU eLife digest A key feature of many important chemical reactions in cells is the transfer of particles called electrons from one molecule to another. The sulfite oxidizing enzymes (or SOEs) are a group of enzymes that are found in many organisms. These enzymes convert sulfite, which is a very reactive compound that can damage cells, into another compound called sulfate. As part of this process the SOE transfers electrons from sulfite to other molecules, such as oxygen or a protein called cytochrome c. In the past, researchers have described the three-dimensional structure of three SOEs using a technique called X-ray crystallography. However, it has been difficult to study how SOEs pass electrons to other molecules because of the temporary nature of the interactions.
McGrath et al. studied an SOE called SorT, which is found in bacteria. The SorT enzyme passes electrons from sulfite to another protein called SorU. McGrath used X-ray crystallography to determine the three-dimensional structures of versions of these proteins from a bacterium called Sinorhizobium meliloti. This included structures of the proteins on their own, and when they were bound to each other. These structures revealed that a subtle change in the shape of SorU occurs when the proteins interact, which enables an electron to be quickly transferred.
McGrath et al. also found that the interface between the two proteins showed an unexpectedly high number of contact sites. These strengthen the interaction between the two proteins, which helps to make electron transfer more efficient. However, these contact sites do not prevent the two proteins from quickly moving apart after the electrons have been transferred. The next challenge is to find out whether these observations are common to SOEs from other forms of life. electron transfer complex. This is the first time that a crystal structure of a molybdenum enzyme in complex with its external electron acceptor has been solved.

Results
The interactions between SorT and SorU are highly dynamic and efficient in solution Sulfite-oxidizing enzymes, particularly those from bacteria, are known to be highly efficient catalysts (Kappler, 2011;Kappler and Enemark, 2015). Previous work has established that SorT is able to transfer electrons to the SorU cytochrome that is encoded on the same operon; however, no kinetic details of the interaction were reported (Low et al., 2011). With the artificial electron acceptor Figure 1. The crystal structures of the SorT and SorU proteins in isolation. (A) The structure of the SorT homodimer. Molecule A in blue/yellow; molecule B in gray (with transparent surface). For molecule A, the 'SUOXfold domain' and 'dimerization domain' are represented in blue and yellow, respectively. The molybdopterin cofactor is shown as sticks within the SUOX-fold domain. The corresponding domains of the opposing protomer (shown in molecular surface representation), which constitute the dimer interface are colored to highlight the 'head-to-tail' dimer arrangement. INSET: a closer view of the molybdenum binding-site: the molybdenum atom (green sphere) is coordinated by two dithioline ligands from the molybdopterin (yellow spheres), residue Cys 127, an axial oxo ligand and an equatorial hydroxo or water ligand (red spheres). (B) The structure of SorU. The main three helices are labeled and the heme cofactor is shown in red. INSET ferricyanide, SorT was shown to have a turnover number of 338 ± 3 s -1 (Low et al., 2011;Wilson and Kappler, 2009). Employing SorU as the substrate, we analyzed the kinetics of the interaction between SorT and SorU and found the interaction to be fast and highly specific, with a K M (SorU) of 32 ± 5 mM and a k cat of 140 ± 11 s -1 , confirming that SorU is the natural electron acceptor of SorT. Measurement of the thermodynamics of the SorT/SorU interaction by isothermal titration calorimetry (ITC) revealed a dissociation constant of K d = 13.5 ± 0.8 mM with a determined stoichiometry of 0.8 ± 0.2. These values are in the range observed for other electron transfer complexes (Dell'acqua et al., 2008;Pettigrew et al., 2003) and match a model where SorT sequentially transfers two electrons, derived from sulfite oxidation, to two SorU molecules. In other words, the SorT/ SorU complex must form twice (with two different ferric SorU protein molecules) to complete the oxidative half reaction of SorT.
The K M of SorT for SorU is very close to its affinity for sulfite (K M = 15.5 ± 1.9 mM , and the turnover number in the SorU-based assay is~40% of that seen with ferricyanide as the electron acceptor , where no significant reorientation and docking of the electron acceptor is required. These data reveal interesting details about the formation of the SorT/SorU complex, which appears to form with similar affinities between the two proteins when both are oxidized as in the ITC experiments and in a system where SorT constantly undergoes oxidation and reduction (SorU-based enzyme assay). In addition, the similarity between the determined values of K d and K M indicates that the affinity of SorT for SorU is unaffected by the presence of substrate or product. However, the kinetic parameters for this interaction are clearly distinct from those for other sulfite-oxidizing enzymes, such as the bacterial SorAB enzyme or chicken sulfite oxidase (CSO), both of which have much higher affinities for their respective electron accepting cytochromes c (both with K M(Cyt c) ca. 2 mM) (Kappler et al., 2006). The catalytic turnover of CSO is relatively slow (k cat = 47.5 ± 1.9 s -1 ), due to a mechanism requiring internal rearrangement, while turnover of SorAB with its natural electron acceptor (334 ± 11 s -1 ) is significantly faster than that for SorT (Kappler and Enemark, 2015;Kappler et al., 2006).
The crystal structure of the SorT homodimer reveals a head to tail subunit arrangement In order to investigate whether there are any structural reasons for these differences, we solved the crystal structure of SorT by molecular replacement and refined it to 2.4 Å resolution. The structure shows two homodimeric assemblies per asymmetric unit (Table 1), which is in agreement with the quaternary structure as determined by MALLS (Low et al., 2011;Wilson and Kappler, 2009). Unexpectedly however, within the SorT homodimer the protomers are oriented in a head-to-tail orientation ( Figure 1A), a subunit arrangement that has not been previously observed in structures of sulfite-oxidizing enzymes. In keeping with the nomenclature applied to other structurally-characterized SOEs, a 'dimerization' domain typically defines the interface between the two monomers (Kisker et al., 1997;Schrader et al., 2003), but the structure of the SorT dimer does not follow this paradigm.
Nevertheless, the fold of the SorT monomers is similar to those of other SOEs (Kisker et al., 1997;Schrader et al., 2003;Kappler and Bailey, 2005), comprising a central 'SUOX-fold' domain (Workun et al., 2008) that harbors the Mo active site and a 'dimerization' domain ( Figure 1A). The SorT active site has a square-pyramidal geometry seen in all other SOE structures with a five coordinate molybdenum atom and a single tricyclic pyranopterin cofactor ( Figure 1A, Table 2) (George and Pickering, 1999).
Single electron reduction of SorT to its EPR active Mo V form was achieved using a combination of Ti(III)citrate and a suite of organic redox mediators. The Mo V EPR spectrum is similar to the so-called 'high pH' EPR signature of SOEs ( Figure 2A, Table 3, Appendix 1). An additional feature is the presence of superhyperfine coupling between the unpaired electron on the Mo ion and two nearby I = ½ nuclei. This implies that the equatorially coordinated O-donor is an aqua ligand at pH 8 ( Figure 2B (ii)) or that a hydroxido ligand is hydrogen bonded with a water molecule whose proximal H-atom is coupled with the electron spin on Mo ( Figure 2B(i), Appendix 1). Within the SorT dimer, the subunit interface involves both the 'SUOX-fold' and the 'dimerization' domains ( Figure 1A), resulting in a buried surface area of~1280 Å 2 per monomer, which is approximately 9% of the solvent-accessible surface of each monomer. The functional consequences and structural origins of these significant differences among the quaternary structures of SOEs are at this point unknown, as all of these enzymes are highly catalytically active, have similar active site structures and no known kinetic cooperativity that would imply a functional role for the different oligomeric assemblies Hänsch et al., 2007;Wilson and Rajagopalan, 2004).
In complex, SorT and SorU form a SorU/SorT 2 /SorU assembly that reveals a pathway for electron transfer Despite the dynamic nature of the SorT/SorU interaction, it was possible to co-crystallize SorT with SorU, resulting in the crystal structure of the SorT/SorU complex, where a single SorT/SorU entity is present per asymmetric unit, and the application of crystallographic 2-fold symmetry reveals a SorU/ SorT 2 /SorU assembly ( Figure 3A, Table 1). Small-angle X-ray scattering (SAXS) with a sample prepared as a stoichiometric mixture of SorT and SorU ( Figure 3B, Table 4, Appendix 2- Figure 1) confirmed that the structure observed in the crystal is preserved in solution.
The central assembly within the SorU/SorT 2 /SorU complex is the SorT homodimer, which is identical to the structure of the SorT homodimer alone ( Figures 1A and 3A, Table 1). The structure of the SorU protein, both within the SorT/SorU complex and when crystallized in isolation ( Figure 1B, Table 1), is predominantly a-helical with three major a-helices arranged to form a bundle that frames the heme-binding site ( Figure 1B) (C 47 XXC 50 H, axial ligands: His 51 and Met 87).
In the SorT/SorU complex, the SorU protein docks within a pocket adjacent to the SorT active site, with the heme cofactor located at the protein-protein interface ( Figures 3A and 4A). The shortest 'edge-to-edge' distance between the SorT Mo atom and the propionate group from the SorU heme c cofactor is 8.2 Å , which is well within the distance for fast electron transfer through the protein medium (Page et al., 1999). PATHWAY analysis   (Table 5) further indicates that the dominant electron tunneling pathway from SorT to SorU proceeds from the Mo atom, via the coordinating H 2 O/OH -, to the guanidinium group of SorT residue Arg 78 and across the protein-protein interface to the heme propionate group and to the pyrrole ring of heme c to the heme iron ( Figure 4B, Table 5). Conformational change reminiscent of an 'induced fit' mechanism facilitates docking and electron transfer between SorT and SorU Specific structural adaptations of the SorU protein take place when the SorT/SorU assembly is formed ( Figure 5B). A surface loop on SorU (residues 82-93) moves away from the SorT/SorU complex interface, leading to a reorientation of the heme ligand residue Met 87 so that a different Met 87 rotamer coordinates the iron in the SorU structures within and outside of the complex ( Figure 5B). This change in the structure of SorU is required to allow a Mo-heme edge-to-edge distance of 'closest approach' of ca. 8 Å within the SorT/SorU assembly. Without this adjustment (for example, if the SorU residue 82-93 loop structure remained rigid) the closest approach for the redox cofactors would be ca. 10 Å . Changes in the conformation of axial heme ligands are known to alter the orbital interactions between the iron atom and the ligand, which can change the redox properties of the heme group (Tai et al., 2013). However, this does not appear to be the case here as the redox potential of the SorU heme was determined by optical spectroelectrochemistry to be +108(±10) and +111(±10) mV vs. NHE (pH 8.0), respectively, in the presence and absence of SorT (Table 6, Figure 6B). The Mo VI/V and Mo V/IV redox potentials of SorT (+110(±10) mV and -18(±10) mV vs. NHE (pH 8)) were determined by an EPR-monitored redox titration where the initial EPR-silent Mo VI form is reduced to the EPR-active Mo V state (Figure 2A), which then gives way to the EPR-silent Mo IV form Table 4. Data collection and processing parameters for analysis of the SorT/SorU complex in solution by Small Angle X-ray Scattering (SAXS).

Instrument
SAXSess (Anton Paar) Beam geometry 10 mm slit AH, LH (Å -1 ), GNOM beam geometry definition 0.28, 0.12 q-range measured (Å -1 ) 0.01-0.400 Exposure time (min) 60 (4 x 15) SorT 2 SorU 2 concentration range (mg mL -1 ) 2.75-5.5 Temperature (ºC) 10 Structural parameters * R g (Å ), I(0) (cm -1 ) from Guinier (desmeared data) q*R g < 1. Partial specific volume (cm 3 g -1 ) 0.732 Contrast (X-rays) (D x 10 10 cm -2 ) 2.895 Modeling results and validation at low potential, resulting in a bell-shaped curve ( Table 6, Figure 6A). The high Mo VI/V potential at pH 8 matches that of the ferris/ferrous redox couple. (3,4). The structural change in SorU indicates that an 'induced fit' mechanism is responsible for the formation of a productive SorT/SorU electron transfer complex. This type of mechanism has in the past been used to describe electron transfer complexes (involving electron transfer flavoproteins or ferredoxin reductases) where the protein partners include mobile domains, and where conformational change is necessary for the creation of high affinity protein-protein interfaces (Senda et al., 2007;Toogood et al., 2007). However, although facilitating redox interactions, these systems are distinct from the docking mechanism seen for the SorT and SorU proteins which accompanies modifications to the structure of SorU and allows the two redox centers to attain positions of closest approach for fast electron transfer. The SorT/SorU interface features extensive electrostatic interactions The SorT/SorU complex interface shows significant charge complementarity, with the negative charge on SorU correlating with a concentration of positive charges at the SorU binding site on SorT ( Figures 3C,D). Unlike other known cytochromes c that can act as electron acceptors to SOEs Table 5. Electron transfer parameters between SorT (Mo) and SorU (Fe) as calculated by PATHWAYS   C tot where the extinction coefficients refer to the oxidized and reduced forms of the protein and Abs is the absorbance at this same wavelength. c tot is the total protein concentration. The redox potential (E' = +111 mV vs NHE) was obtained by global analysis of all potential dependent spectra across all wavelengths with the program ReactLab Redox (Maeder and King). DOI: 10.7554/eLife.09066.013 (Low et al., 2011;Brody and Hille, 1999;Kappler et al., 2000), the electrostatic surface of SorU has an overall negative charge ( Figures 3C,D). The positive charge on the SorT surface therefore explains the low catalytic activity of SorT with horse heart cytochrome c (7 U/mg, pI~10), the natural electron acceptor for vertebrate SOEs, compared to the high activity observed with SorU (212 U/ mg; pI~4) (Kappler, 2011;Low et al., 2011;Wilson and Kappler, 2009). In its electrostatic nature, the interaction surface between SorT and SorU is unusual. Structures of other cytochrome-containing electron transfer complexes (Nojiri et al., 2009;Axelrod et al., 2002;Pelletier and Kraut, 1992;Solmaz and Hunte, 2008) show binding interfaces characterized by a 'ring' of electrostatic interactions that encompass contact surfaces that are predominantly hydrophobic. In fact, the 'steering' of electron transfer partners by electrostatic interactions, accompanied by 'tuning' via hydrophobic interactions is a dominant observation for protein-protein electron transfer complexes (Nojiri et al., 2009;Axelrod et al., 2002;Pelletier and Kraut, 1992;Solmaz and Hunte, 2008).

An unusually large number of hydrogen bonds and salt bridges characterize the SorT/SorU interface
In addition to the electrostatic interactions that support the formation of the SorT/SorU complex, there are six hydrogen bonds found at the SorT/SorU protein-protein interface, as well as a salt bridge between the SorT active site residue Arg 78 and a propionate group of the SorU heme moiety (Table 7, Figure 7). This is an unusually large number in comparison with structures of other cytochrome-containing electron transfer complexes (Nojiri et al., 2009;Axelrod et al., 2002;Pelletier and Kraut, 1992;Solmaz and Hunte, 2008), which tend to have fewer hydrogen bonds and lack salt bridges. In fact, direct hydrogen bonds between electron transfer proteins are generally considered unfavorable for a transient interaction because of energetically disadvantageous  (Kappler and Bailey, 2005) c Calculated as the average for the protomer of interest divided by the average for the entire complex structure. d (Krissinel and Henrick, 2007) e Taken from (Kappler and Bailey, 2005) f (Lawrence and Colman, 1993) DOI: 10.7554/eLife.09066.015 desolvation (Miyashita et al., 2003). Also significant is the observation that no intermolecular interactions at the interface are mediated by water molecules (Figure 7) (Nojiri et al., 2009;Gray and Winkler, 2005). In fact, very little ordered water is observed in proximity to the interfacing region of the SorT molecule (a total of 2 water molecules only and these are hydrogen bonded to the SorT molecule rather than mediating the SorT/SorU interaction). However, the current analysis is limited by the moderate resolution of the current structure (2.6 Å ), which as a consequence, includes only ca. 0.15 modeled water molecules per residue. Compared with the subunit interface of the SorAB bacterial sulfite dehydrogenase, which contains 30 hydrogen-bonds and 2 salt bridges (Kappler and Bailey, 2005), supporting a permanent, heterodimeric complex of a heme c subunit (SorB) and a Mo cofactor containing (SorA) catalytic subunit (Kappler and Bailey, 2005) (Table 7), the extent of the subunit interactions in the SorT/SorU structure is modest. The difference between the permanent SorAB and the transient SorT/SorU complexes is also illustrated by calculations of the shape complementarity and the buried surface areas between the protomers (Lawrence and Colman, 1993), with the latter being about twice as large for the SorAB assembly than for SorT/SorU (Table 7). Interestingly, much of the additional contact area between molecules in the SorAB structure derives from the SorB N-terminal structure (residues B501-B518, PDB 2BLF), which extends away from the core of the subunit, wraps around the SorA 'SUOX-fold' domain and contributes one salt-bridge and 6 hydrogen bonding interactions ( Table 7) (Kappler and Bailey, 2005). This feature is absent from the SorU structure.

The dynamic SorT/SorU interaction observed in solution is reflected in the crystal
Despite the intricate assembly of interactions at the protein-protein interface, but in agreement with the kinetics and thermodynamics of the SorT/SorU interaction in solution, the SorT/SorU contact is dynamic, as illustrated by a temperature factor analysis of the complex structure. Within the SorT/ SorU complex, the SorU protein shows a significantly increased average atomic temperature factor (reflecting significant flexibility) relative to the structure of the SorT protomer (1.5 versus 0.9 Å 2 , respectively; Table 7). Furthermore, the relative temperature factors per residue for the SorU molecule increase with increasing distance from the SorT/SorU interface ( Figure 5A), indicating that the SorU molecule is dynamic relative to SorT within the crystalline lattice. In contrast, the 'static' SorAB complex shows uniform, low temperature factors for both redox subunits (Table 7, Figure 5A). This observation is an exquisite illustration of 'conformational sampling' within the SorT/SorU electron transfer complex, which results from the conformational flexibility of one protein redox partner relative to the other and both facilitates electron exchange by accessing the optimal orientations of each redox partner and promotes fast dissociation of the complex following transfer (Leys and Scrutton, 2004;van Amsterdam et al., 2002).

Discussion
By describing the structure of the SorT/SorU complex in this work, we report the first example of a structure of an SOE in complex with its external electron acceptor; all previous structures of SOEs being of the enzymes and their internal heme domains or subunits only (Kisker et al., 1997;Schrader et al., 2003;Kappler and Bailey, 2005). The structure of the SorT/SorU complex therefore allows insights into electron transfer in what is thought to be a highly prevalent type of bacterial SOE (Low et al., 2011) and into protein-protein electron transfer in general. While the complex shows dynamic adaptations similar to those demonstrated previously for electron transfer complexes and has a dissociation constant of the right order of magnitude, it also shows some features that have not been seen in electron transfer complexes, namely an interface that is stabilized by a relatively large number of hydrogen bonds and salt bridges, and an induced fit docking mechanism.
The structure of the protein-protein interface in the SorT/SorU structure is particularly intriguing. The relative lack of bound water molecules, mirrors more the observations made of permanent heterodimeric complexes than transient interactions (Gray and Winkler, 2005). Previous investigations into the factors that influence protein-protein docking for electron transfer have shown that the strength of the protein-protein interaction correlates linearly with the product of the total charges on the protein partners (Trana et al., 2012;Xiong et al., 2010). In this way, the affinity between the SorT/SorU proteins (K d = 13.5 ± 0.8 mM) correlates with the fast measured turnover rates (k cat of 140 ± 11 s -1 ) and with the predominantly electrostatic nature of the protein-protein interface.
The turnover number for SorT with SorU as the electron acceptor is also in the range of values measured for the human sulfite oxidase (HSO) and CSO (25.0 ± 1.3 s -1 and 47.5 ± 1.9 s -1 , respectively), but significantly slower than that observed for the permanent SorAB complex (345 ± 11 s -1 ) (Kappler et al., 2006;Wilson and Rajagopalan, 2004;Brody and Hille, 1999). Importantly, in the structure of SorAB, the docking site of the heme subunit (SorB) with the SorA subunit is almost identically positioned to that seen for SorT/SorU, with major differences existing only in the number of hydrogen bonds and salt bridges in the protein-protein interface. For the CSO and HSO enzymes, docking of the mobile heme b domain near the Mo active site has been proposed to be similar to that seen for SorT/SorU (Utesch and Mroginski, 2010). It should be noted, however, that the docking events in CSO (and HSO) and between SorT/SorU serve fundamentally different purposes: for CSO and HSO, domain docking enables intramolecular electron transfer and involves a heme domain that is an intrinsic part of the enzyme. This is a step that precedes interactions with the external electron acceptor for these enzymes. In contrast, and despite the fact that it is occupying a similar docking site to that predicted for CSO and HSO (Utesch and Mroginski, 2010), SorU is the external electron acceptor for SorT and the electron transfer is intermolecular.
The SorT/SorU complex described here thus represents an elegant compromise between the requirements for fast and efficient electron transfer and reaction specificity. It also illustrates new aspects for highly dynamic protein-protein interactions: (i) A relatively large number of hydrogen bonds and salt bridges may be required to form the initial stable protein complex, but this does not preclude a dynamic protein -protein interaction; (ii) relatively subtle structural adjustments in one redox partner (SorU) can facilitate electron transfer by ideally locating the redox active cofactors in close proximity; (iii) the comparatively complex binding interface in SorT/SorU can be counterbalanced by the conformational sampling of one protein relative to the other, which enables the rapid dissolution of the complex following electron exchange.
It remains to be seen whether these principles apply to other SOE -external electron acceptor interactions. Future work should focus on investigating interaction interfaces in currently little studied SOEs where new types of interactions may be present as for many of these enzymes currently no external electron acceptor is known.

Materials and methods
Protein overexpression, purification, data collection and structure solution Recombinant SorT and SorU proteins were overproduced and purified as previously described (Low et al., 2011), with minor modifications. SorT was crystallized by hanging drop vapor diffusion with drops consisting of equal volumes (2 mL) of protein and crystallization solution (0.1 M HEPES pH 7.5, 8% ethylene glycol, 0.1 M manganese (II) chloride tetrahydrate and 17.5% PEG 10,000) at 20˚C. Crystals were cryoprotected in reservoir solution with 30% glycerol before flash-cooling in liquid nitrogen. Small (ca. 20 Â 10 Â 10 m) crystals of SorU were grown in drops containing equal volumes (2 mL) of protein and reservoir solution (1.8 M tri-sodium citrate, pH 5.5, 0.1 M glycine), which were harvested and flash-cooled in liquid nitrogen without additional cryoprotection. Purified SorT (20 mM Tris pH 7.8, 2.5% glycerol) and SorU (20 mM Tris pH 7.8, 150 mM NaCl) were mixed and incubated on ice at a molar ratio of 2:1 (SorU:SorT; total protein concentration 8 mgmL -1 ) before crystallization via hanging-drop vapor diffusion with a reservoir solution containing 0.2 M sodium formate, 0.1 M Bis-Tris propane pH 7.5 and 20% PEG 3350. Crystals grew to a maximum size of ca. 150 x 100 x 20 m in 4 days at 20˚C and were flash-cooled in liquid nitrogen after brief soaking in mother liquor containing 30% glycerol. All diffraction data were collected on an ADSC Quantum 315r detector at the Australian Synchrotron on beamline MX2 at 100 K and were processed with HKL2000 (Otwinowski and Minor, 1997). Unit cell parameters and data collection statistics are presented in Table 1.
The crystal structure of SorT was solved by molecular replacement using PHASER (McCoy et al., 2007) with a search model generated with CHAINSAW (Chainsaw, 2008) from the SorA portion of the SorAB crystal structure (29.0% sequence identity, Protein Data Bank entry 2BLF [Kappler and Bailey, 2005]) as a template (Larkin et al., 2007). The resulting model was refined by iterative cycles of amplitude based twin refinement (using twin operators H, K, L and -H, -K, L with estimated twin fractions of 0.495 and 0.505 respectively) within REFMAC , interspersed with manual inspection and correction against calculated electron density maps using COOT (Emsley and Coot, 2004). The refinement of the model converged with residuals R = 0.208 and R free = 0.239 ( Table 1). The structure of the SorT/SorU complex was solved by molecular replacement using PHASER (McCoy et al., 2007), with the refined SorT structure as a search model. Initial rounds of refinement yielded a difference Fourier electron density map, which clearly showed positive difference density for the location of one molecule of SorU per asymmetric unit, which was manually built using COOT (Emsley and Coot, 2004). Refinement was carried out with REFMAC5  and PHENIX (Adams et al., 2002) and converged with residuals R = 0.211 and R free = 0.260 ( Table 1). The refined SorU model, from the SorT/SorU complex structure, was used as a search model to solve the SorU structure by molecular replacement using PHASER (McCoy et al., 2007). Refinement was carried out with REFMAC5 and PHENIX (Adams et al., 2002) and converged with residuals R= 0.192 and R free = 0.240. All structures were judged to have excellent geometry as determined by MOLPROBITY (Chen et al., 2010) (Table 1).

Small-angle X-ray scattering (SAXS)
SAXS analysis of the SorT/SorU complex was performed in a buffer of 20 mM Tris pH 7.8, 2.5% v/v glycerol. Purified SorU and SorT were mixed and incubated on ice at a molar ratio of 2:1 (SorU: SorT), generating two samples of total protein concentrations 2.75 and 6.25 mgmL -1 , respectively. SAXS data were measured as described previously (Jeffries et al., 2011) with the data collection parameters listed in Table 4. Data were reduced to I(q) vs q (q=4psinl,where q=4sin2 is the scattering angle) using the program SAXSquant that includes corrections for sample absorbance, detector sensitivity, and the slit geometry of the instrument. Intensities were placed on an absolute scale using the known scattering from H 2 O. Protein scattering was obtained by subtraction of the scattering from the matched solvents (20 mM Tris pH 7.8, 2.5% v/v glycerol obtained from the flow-through after protein concentration by centrifugal ultrafiltration). Molecular weight (M r ) estimates for the proteins were made using the equation from Orthaber (Orthaber et al., 2000): Mr=N A I(0)C4M2 where N A is Avogadro's number, C is the protein concentration and 4M=4; where 4 is the protein contrast and the partial specific volume, both of which were determined using the program MULCh (Whitten et al., 2008).
The ATSAS program package (Volkov and Svergun, 2003) was used for data analysis and modeling, with the specific programs used detailed in Table 4, along with the data ranges and results of each of the calculations. Further detail on data interpretation and analysis for these experiments is detailed in Appendix 2.

Electron paramagnetic resonance (EPR) spectroscopy
Continuous-wave X-band (ca. 9 GHz) (CW) electron paramagnetic resonance (EPR) spectra were recorded with a Bruker Elexsys E580 CW/pulsed EPR spectrometer fitted with a super high Q resonator; the microwave frequency and magnetic field were calibrated with a Bruker microwave frequency counter and a Bruker ER 036TM Teslameter, respectively. A microwave power of 20 mW was used and optimal spectral resolution was obtained by keeping the modulation amplitude to a 1/10 of the linewidth. A flow-through cryostat in conjunction with a Eurotherm (B-VT-2000) variable temperature controller provided temperatures of 127-133 K at the sample position in the cavity.
Bruker's Xepr (version 2.6b.45) software was used to control the data acquisition including, spectrometer tuning, signal averaging, temperature control and visualization of the spectra. Computer simulation of the EPR spectra were performed with the following spin Hamiltonian (Equation 2) using the XSophe-Sophe-XeprView (version 1.1.4) computer simulation software suite (Hanson et al., 2004;Hanson et al., 2013) on a personal computer, running the Mandriva Linux v2010.2 operating system. Further detail on data interpretation and analysis for these experiments is detailed in Appendix 1.

Optical spectroelectrochemistry
Spectroelectrochemistry of SorU in isolation and the SorT:SorU complex was performed with a Bioanalytical Systems BAS100B/W potentiostat connected to a Bioanalytical Systems thin layer spectroelectrochemical cell (0.5 or 1 mm pathlength) bearing a transparent Au mini-grid working electrode, a Pt wire counter and Ag/AgCl reference electrode. Redox mediators (all Fe complexes) used in the experiment were employed at concentrations of 50 mM (Figure 8).
None exhibit any significant absorption in the spectral range of interest at micromolar concentrations. The total solution volume was ca. 700 mL. The buffer was 20 mM Tris (pH 8) containing 200 mM NaCl as supporting electrolyte. The SorU concentration was ca. 50 mM, while experiments on the SorU:SorT complex used approximately equal concentrations of both proteins (50 mM). Spectra were acquired within a Belle Technology anaerobic box with an Ocean Optics USB2000 fibre optic spectrophotometer. Initially, the cell potential was poised at ca. -100 mV and the system was allowed to equilibrate until no further spectral changes were apparent (fully reduced SorU). The potentials were then increased in 50 mV increments and the spectrum was measured when no further changes were seen (5-10 min). When the protein was fully oxidized the potential was scanned in the reverse direction in 50 mV intervals to establish reversibility. Data were fitted using the program ReactLab Redox (Maeder and King).

Isothermal titration calorimetry (ITC)
Affinity measurements were conducted using a Microcal ITC200 system (GE Healthcare) at 25˚C using SorT and SorU in buffer (25 mM HEPES pH 7.5, 150 mM NaCl and 2.5% glycerol) at final concentrations of 300 mM and 30 mM, respectively. SorT at a concentration of 300 mM was titrated with eighteen injections (2.0 ml each) of SorU. All affinity measurements were performed in triplicate and fitted using a single site mode. Protein concentrations were estimated using Bicinchoninic acid (BCA) protein assay kit (Thermo Fisher Scientific, Waltham, MA).

Enzyme kinetics
Sulfite dehydrogenase enzyme assays were carried out as described previously (Low et al., 2011;Wilson and Kappler, 2009;Kappler et al., 2000). The reduced -oxidized extinction coefficient for SorU at 550 nm was 17.486 mM -1 cm -1 as determined by spectroelectrochemistry. Data fitting was carried out using Sigmaplot 12 (Systat). on the Macromolecular Crystallography beamline at the Australian Synchrotron (Victoria, Australia) and we thank the beamline staff for their enthusiastic and professional support. A/Prof Matthew Perugini (La Trobe University) is thanked for very helpful discussions. Assistance with EPR sepctroscopy from Dr Jeff Harmer and Dr Chris Noble (Centre for Advanced Imaging, University of Queensland) is also gratefully acknowledged. The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Additional information
Author contributions APM, JT, PVB, GRH, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article; ELL, GPCG, MK, BC, Acquisition of data, Analysis and interpretation of data; JMG, Analysis and interpretation of data, Drafting or revising the article; UK, MJM, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article

Major datasets
The following datasets were generated: