Structural and spectroscopic characterization of a HdrA‐like subunit from Hyphomicrobium denitrificans

Many bacteria and archaea employ a novel pathway of sulfur oxidation involving an enzyme complex that is related to the heterodisulfide reductase (Hdr or HdrABC) of methanogens. As a first step in the biochemical characterization of Hdr‐like proteins from sulfur oxidizers (sHdr), we structurally analyzed the recombinant sHdrA protein from the Alphaproteobacterium Hyphomicrobium denitrificans at 1.4 Å resolution. The sHdrA core structure is similar to that of methanogenic HdrA (mHdrA) which binds the electron‐bifurcating flavin adenine dinucleotide (FAD), the heart of the HdrABC‐[NiFe]‐hydrogenase catalyzed reaction. Each sHdrA homodimer carries two FADs and two [4Fe–4S] clusters being linked by electron conductivity. Redox titrations monitored by electron paramagnetic resonance and visible spectroscopy revealed a redox potential between −203 and −188 mV for the [4Fe–4S] center. The potentials for the FADH•/FADH− and FAD/FADH• pairs reside between −174 and −156 mV and between −81 and −19 mV, respectively. The resulting stable semiquinone FADH• species already detectable in the visible and electron paramagnetic resonance spectra of the as‐isolated state of sHdrA is incompatible with basic principles of flavin‐based electron bifurcation such that the sHdr complex does not apply this new mode of energy coupling. The inverted one‐electron FAD redox potentials of sHdr and mHdr are clearly reflected in the different FAD‐polypeptide interactions. According to this finding and the assumption that the sHdr complex forms an asymmetric HdrAA′B1C1B2C2 hexamer, we tentatively propose a mechanism that links protein‐bound sulfane oxidation to sulfite on HdrB1 with NAD+ reduction via lipoamide disulfide reduction on HdrB2. The FAD of HdrA thereby serves as an electron storage unit.

Many bacteria and archaea employ a novel pathway of sulfur oxidation involving an enzyme complex that is related to the heterodisulfide reductase (Hdr or HdrABC) of methanogens. As a first step in the biochemical characterization of Hdr-like proteins from sulfur oxidizers (sHdr), we structurally analyzed the recombinant sHdrA protein from the Alphaproteobacterium Hyphomicrobium denitrificans at 1.4Å resolution. The sHdrA core structure is similar to that of methanogenic HdrA (mHdrA) which binds the electron-bifurcating flavin adenine dinucleotide (FAD), the heart of the HdrABC-[NiFe]-hydrogenase catalyzed reaction. Each sHdrA homodimer carries two FADs and two [4Fe-4S] clusters being linked by electron conductivity. Redox titrations monitored by electron paramagnetic resonance and visible spectroscopy revealed a redox potential between −203 and −188 mV for the [4Fe-4S] center. The potentials for the FADH•/FADH − and FAD/FADH• pairs reside between −174 and −156 mV and between −81 and −19 mV, respectively. The resulting stable semiquinone FADH• species already detectable in the visible and electron paramagnetic resonance spectra of the as-isolated state of sHdrA is incompatible with basic principles of flavin-based electron bifurcation such that the sHdr complex does not apply this new mode of energy coupling. The inverted one-electron FAD redox potentials of sHdr and mHdr are clearly reflected in the different FAD-polypeptide interactions. According to this finding and the assumption that the sHdr complex forms an asymmetric HdrAA 0 B1C1B2C2 hexamer, we tentatively propose a mechanism that links protein-bound sulfane oxidation to sulfite on HdrB1 with NAD + reduction via lipoamide disulfide reduction on HdrB2. The FAD of HdrA thereby serves as an electron storage unit.

Introduction
Sulfur is a very versatile element and undergoes permanent biogeochemical cycling in terrestrial as well as in marine environments. Dissimilatory sulfate-reducing prokaryotes use sulfate as a respiratory electron acceptor instead of oxygen or nitrate [1]. On the other hand, hydrogen sulfide, elemental sulfur, and other reduced sulfur compounds serve as electron donors for a huge array of chemo-and photolithotrophic prokaryotes [2][3][4][5][6][7][8][9]. While a considerable portion of these organisms employs the comparatively well-characterized dissimilatory sulfite reductase (Dsr) pathway of sulfur oxidation [10,11], a novel pathway involving a new lipoate-binding protein (LbpA) and a heterodisulfide reductase (Hdr)-like complex resembling HdrABC from methanogens [12] has recently been identified in the Alphaproteobacterium Hyphomicrobium denitrificans by reverse genetics [13,14]. The relevant genes hdrC1B1AhyphdrC2B2hyp-lbpA also occur in a large number of other sulfur-oxidizing bacteria and archaea (e.g., species of the genera Acidithiobacillus, Thioalkalivibrio, Aquifex, Sulfolobus or Metallosphaera). Strong responses of these genes and the encoded proteins to the presence of reduced sulfur compounds have been reported in numerous transcriptomic and proteomic studies [3,[14][15][16][17][18].
Biochemical information on Hdr-like complexes from sulfur oxidizers is extremely scarce [13,14,28]. Co-purification of sHdrAB1B2C1C2 from membranes of the hyperthermophilic bacterium Aquifex aeolicus provided a first milestone, revealing the formation of a large protein complex consisting of several different sHdr subunits [28]. However, a functional role or enzymatic activity for this complex could not be assigned. We started filling this knowledge gap by determining the crystal structure of recombinant sHdrA from the model organism H. denitrificans and the redox potentials of the [4Fe-4S]-cluster and the FAD cofactor via electron paramagnetic resonance (EPR) and UV-visible spectroscopy. First insights are gained into the structure and possible catalytic processes performed by Hdr-like complexes of sulfur oxidizers.

sHdrA is an iron-sulfur flavoprotein
Recombinant sHdrA from H. denitrificans carrying a Strep-tag at its carboxy-terminus was produced as a soluble protein in Escherichia coli BL21 ΔiscR [29] grown anaerobically on fumarate [30]. One liter E. coli culture routinely yielded 7-8 mg of pure protein [14]. 2.5-3.5 mol iron and sulfur per mol sHdrA monomer were determined for different preparations, fully in line with one [4Fe-4S] cluster predicted from the sequence. The protein displayed spectral properties typical for iron-sulfur flavoproteins.
Overall structure of sHdrA The structure of sHdrA was determined at 1.4Å resolution with R/R free factors of 16.2%/19.1% (Table 1) by using the single anomalous dispersion method for phase determination. sHdrA adopts a thioredoxin reductase (TrxR) fold ( Fig. 1) [31] composed of two classical dinucleotide-binding (Rossmann) domains as originally found for glutathione reductase [32]. The FAD-binding domain (2-115, 271-341) binds the flavin cofactor at the C-terminal end of the central βsheet in the classical manner, the isoalloxazine ring being packed between the two domains (Fig. 1A). The [4Fe-4S] cluster-binding domain (116-270) has embedded the [4Fe-4S] cluster between the C-terminal end of the central β-sheet and the N-terminal side of α-helix 180 : 195. The edge-to-edge distance between two redox centers is 9.4Å (Fig. 1A), which allows physiologically relevant electron transfer rates between them.
The nearest structurally known relative of sHdrA is mHdrA, the core subunit of the methanogenic Hdr complex (5odc) which binds FAD and the [4Fe-4S] cluster at equivalent positions [12]. Their overall rms (root mean square) deviation is 1.5Å (337 of 653 residues, sequence identity 32%). In addition, sHdrA shares a high structural similarity to (low-molecularmass) TrxR. The lowest rmsd of 3.4Å (287 of 305; 23%) and 3.4Å (287 of 303; 19%) were calculated for the Thermoplasma acidophilum (3cty) and Lactococcus lactis (5mh4) enzymes, respectively [33,34]. TrxR does not contain a [4Fe-4S] cluster but instead binds NADP (Fig. 1B). An exception is the T. acidophilum enzyme [33] whose electron donor is unknown. For (low-molecular-mass) TrxR two structural forms are distinguished depending on the orientation of the FAD and [4Fe-4S] cluster (NADP) domains. In the first conformational state, FADH − is oxidized by hydride transfer to the adjacent disulfide, while in the second state FAD is reduced by NADPH [31,35]. sHdrA is present in the second state.
Just as mHdrA, H. denitrificans sHdrA is present as a homodimer characterized by an extended interface of 2470Å 2 built up between the two FAD domains and between the two [4Fe-4S] cluster domains (Fig. 1A). In sHdrA, the edge-to-edge distances for the two isoalloxazine rings, the two [4Fe-4S] clusters, and the FAD and the [4Fe-4S] cluster of the counter subunits are 18.8, 11.5 and 17.8Å, respectively. TrxR also oligomerizes to a homodimer, however, the two monomers are oriented differently compared to HdrA; their two-fold axes are rotated by 90°relative to each other (Fig. 1B). In TrxR, the interface is essentially formed by the FAD domains resulting in a smaller contact area of 1640Å 2 .

FAD-and [4Fe-4S]-cluster binding sites
Although the basic architecture of the FAD-binding site is shared by all members of the TrxR family [31], the conformation of the polypeptide segments and of relevant side chains contacting the isoalloxazine ring are substantially changed in HdrA proteins (Fig. 2). Most notable are an elongated segment following strand 26:33 and segment 131-141 linking the FADand [4Fe-4S] cluster-binding domains. In both sHdrA and mHdrA, the isoalloxazine is essentially buried inside the polypeptide scaffold, whereas especially in the second NADP-bound form of TrxR, the isoalloxazine is accessible from two sides. In sHdrA, the isoalloxazine is in a slightly bent conformation mainly provided by hydrophobic side chains such as Val183, Ile179, Ile322, Tyr45, Leu48, and Val49 pointing to the si-and re-faces ( Fig. 2A,B). In mHdrA, FAD is not in a butterfly conformation.
The rather hydrophobic character but not individual residues of the isoalloxazine binding site are conserved between sHdrA and mHdrA (Fig. 2C). A hydrogen bond ( Fig. 2A) is formed between the functionally important N5 and a water molecule, which is hydrogen-bonded to Lys47-O, Glu137-OE1 and via a second water molecule to Glu190-OE2. In mHdrA, however, N5 and O4 are hydrogen-bonded with Lys409-NZ anchored in its position by Glu356-OE2 (Fig. 2C). For TrxR enzymes no direct N5-polypeptide interactions are reported but variable water-mediated hydrogen bonds to nonconserved polar residues (Fig. 2D). Hydrogen bonds invariant among TrxR family members are formed between O2 and Ile322-NH positioned at the N-terminal end of helix 322 : 340 and between the N3-C4 = O4 groups and main chain peptide nitrogen and carbonyl groups or asparagine side chains (Fig. 2). A notable difference is a further water-mediated hydrogen bond between O2 and Asp323 in sHdrA and between O2 and Lys524' in mHdrA ( Fig. 2A-C). Lys524' points from the counter mHdrA (marked by an apostrophe) toward the isoalloxazine ring, which is enabled by the different conformation of the elongated segment following strand 26:32 of sHdrA compared to mHdrA. The [4Fe-4S] cluster-binding is well conserved in sHdrA and mHdrA. The four irons are coordinated with Cys163, Cys176, Cys180, and Cys181 (sHdrA numbering) and all sulfurs interact with at least one proton donor (Fig. 3). Arg167, Arg208, and the partially positively charged N-terminal end of helices 180 : 195 and 212 0 : 225 0 create a positively charged [4Fe-4S] cluster binding pocket.

UV-vis spectroscopical characterization
Just as expected for a protein carrying FADs and iron-sulfur clusters in a 1 : 1 ratio, the UV-vis spectrum for sHdrA in the 'as-isolated' state is dominated by the absorption of FAD. This was apparent in spectra taken at pH 8.0 by absorption maxima at 360 and 455 nm with shoulders at 433 and 485 nm (Fig. 4A). A broad absorption band observed at 590-650 nm in the H. denitrificans sHdrA spectra is a characteristic feature for a stable FAD semiquinone radical in a neutral state (FADH•) [36][37][38]. The relatively pronounced absorption at 360 nm indicated that a part of the FAD semiquinone population was present in the anionic (FAD• − ) state at pH 8.0. By shifting the pH to 7.0, the absorbance at 590-650 nm further increased, while the band at 360 nm decreased simultaneously (Fig. 4B). This finding corresponds exactly to the type of behavior expected. The equilibrium is shifted from a partially anionic FAD• − toward a neutral FADH• state upon protonation [38]. Under the crystallization conditions at pH 6.5, the FAD of sHdrA is presumably present in a partial neutral semi-reduced state.
Further evidence for the presence of a stable FADH• in sHdrA was obtained by stepwise oxidation of 50 µM prereduced protein with potassium ferricyanide (E 00 = +0.452 V in 0.2 M Tris/HCl, pH 7.0, [39]; Fig. 4C). Reduced sHdrA exhibited low absorption at 610 nm (Fig. 4C, dark blue spectrum), while the absorption value increased by gradual oxidation of the protein (Fig. 4C, blue to green spectra). A maximum was reached after addition of an equimolar amount of ferricyanide (50 µM) when one electron is withdrawn from each reduced FAD cofactor. Upon further oxidation, absorption at 610 nm decreased again. After its disappearance, the FAD of sHdrA is completely oxidized accompanied with a strong increase of absorption at 456 nm. Corresponding results were obtained by titration in the reverse direction, that is, by stepwise reduction of pre-oxidized sHdrA with titanium citrate (E 00 = −0.48 mV at pH 7.0 [40]; Fig. 4D).

Potentiometric titration of FAD
In general, flavins can have a wide range of redox potentials (E m , +150 to −500 mV [41]) and these are not predictable from the primary structure of the protein. The reduction potential of the sHdrA flavin was determined by a redox titration monitored by recording UV/vis spectra. As-isolated sHdrA (50 µM) was titrated stepwise with sodium dithionite, which has a redox potential below −550 mV [42], and allows complete reduction of the flavin cofactor of the protein. Complete oxidation was achieved in a separate measurement by stepwise addition of 2,6-dichloroindophenol (E 0 = +330 mV at pH 6.0 [43]) as oxidant. This experiment allowed the assignment of the redox potentials of the FAD cofactor (Fig. 5). The ratio of oxidized/reduced FAD species was calculated from the UV-visible spectra at 456 nm and 610 nm, respectively, and plotted against the observed potential (Fig. 5A). A stable FADH• species appeared between −104 to −70 mV recognizable by a plateau in the titration curve at 456 nm. The FAD absorption at 590-650 nm was where E is the observed potential. When E is plotted against the logarithm of the concentration ratio of oxidized (c ox ) and the reduced species (c red ), linear functions can be derived from the data and statistically analyzed ( Fig. 5B-E). E°0 is taken from the plots at c ox = c red , and 59 divided by the slope gives n, the number of electrons involved in the reaction. Redox potentials were determined by following the absorbance at 456 and 610 nm. The calculations at 610 nm are, however, based on the predominant presence of the flavin in the FADH• state at around −100 mV which makes them less reliable. The values of the midpoint potential calculated for the FADH•/FADH − couple were E 00 = −168 AE 5 mV (Fig. 5B) and −160 AE 4 mV (Fig. 5C), respectively. In both cases, data fitted well to the Nernst equation (R 2 = 0.95 and 0.97, respectively) and indicated a one-electron transition (n = 0.97 AE 0.07 for 456 nm and n = 0.99 AE 0.06 for 610 nm). When holding n fixed to 1, the potentials amount to −169 AE 5 mV (R 2 = 0.95) and −160 AE 4 mV (R 2 = 0.97), respectively ( Fig. 5A-C).
The analysis of the FAD/FADH• couple is not as unambiguous as that for FADH•/FADH. The data obtained at 456 nm for the FAD/FADH• couple yielded E 00 = −24 AE 4 mV and n = 2.25 AE 0.30 (Fig. 5  D). The data points could not be fitted with a one-electron but with a two-electron Nernst equation shown in Fig. 5D and in the upper part of Fig. 5A. For n = 2, the midpoint potential was −23 AE 4 mV (R 2 = 0.86). At 610 nm the fit of the titration points with the Nernst equation was comparably poor (R 2 = 0.44) and gave E 00 = −59 AE 10 mV and n = 1.58 AE 0.70 for the numbers of electrons involved (Fig. 5E). Since only an integer number of electrons can be transmitted, the titration curve was simulated for n = 1 and n = 2 resulting in midpoint potentials of −70 AE 11 mV (R 2 = 0.29) and −54 AE 10 mV (R 2 = 0.41). The reason for the shape of the FAD/ FADH• titration curve that rather reflects, in particular at 456 nm, a two-electron transition and the origin of a potential second electron remains obscure at this point. Nevertheless, the stable FADH• peak around −100 mV at 456 and 610 nm, respectively, and its increase/decrease upon further oxidation indicates that definitively one electron stems from FADH• (Fig. 5A).
In summary, the presented visible titration data are in accordance with the formation of a stable FADH• species and indicating two temporally separated oneelectron steps for the FAD/FADH − transition. Although very good agreement was obtained for the midpoint potential of the FADH•/FADH − couple at both wavelengths analyzed, some uncertainty remains regarding the value for the FAD/FADH• pair, albeit it is clearly more positive. More accurate determination of the actual potentials will have to await determination of the spin concentration of the FADH• radical as has been outlined for iron-sulfur clusters by others [44]. The X-band EPR spectrum of the as-purified sHdrA exhibited a typical broad signal centered at g = 2.002 characteristic of a flavin radical, that is, a oneelectron-reduced FADH• (Fig. 6A) fully in line with UV-vis spectroscopic detection of this species.

Potentiometric titration of the [4Fe-4S] cluster
A redox titration of sHdrA using sodium dithionite as reductant was performed inside the anaerobic chamber, and followed by EPR to further characterize the [4Fe-4S] center of sHdrA (Fig. 6B,C). The flavin radical signal is masked by the presence of redox mediators and is not titratable by EPR. In the EPR spectrum of fully reduced HdrA a rhombic signal with g max = 2.053, g med = 1.961, and g min = 1.926 is observed (Fig. 6B), which is characteristic for a single [4Fe-4S] center in agreement with the primary sequence and crystal structure data. The height of the g max or g med of the rhombic signal was followed during the titration (Fig. 6B) in order to determine the redox potential of the center. Both g values gave rise to the same midpoint redox potentials within experimental error, namely approximately −197 mV (g max ) and −192 (g med ), and fit a one-electron process (Fig. 6C).

Properties and function of sHdrA
The performed structural and spectroscopic characterization of sHdrA indicated substantial functional differences to those TrxR family members acting as hydride-transferring enzymes but also to mHdrA operating as a FBEB module. (a) HdrA and TrxR differ in their monomer-monomer arrangement. In HdrA, the side-by-side attachment of the FAD and [4Fe-4S] domains hold the two monomers together and shields both cofactors from one side (Fig. 1). In addition, the distance between the two [4Fe-4S] clusters is sufficiently short for a rapid electron transfer [45] (Fig. 2) thereby forming an electron-conducting and -storing unit between the two FAD and the two [4Fe-4S] clusters. In contrast, the TrxR dimer interface only consists of two FAD domains (Fig. 1B) and thus enables a 66°rotation of the NADP domain to cycle between FAD-reducing and -oxidizing forms [35]. (b) In contrast to TrxR, the isoalloxazine ring of the HdrAs is buried inside the protein implicating no access for an external hydride-transferring compound. (c) The NADP binding site in TrxR is converted into a [4Fe-4S] cluster binding site in HdrAs, predominantly by prolonging helix 180 : 195 by one turn and by redesigning its preceding loop (Fig. 3). (d) sHdrA cannot operate as FBEB enzyme as on one hand the ferredoxin insertion/extension domains essential for bifurcation in mHdrA are lacking in sHdrA [12]. On the other hand, the redox properties of FAD in sHdrA (E 00 FAD/FADH• > E 00 FADH•/FADH −) indicate a stable FADH• semiquinone (Figs 5 and 6), which is incompatible with the current concept of FBEB. This mode of energy coupling requires an unstable, energy-rich semiquinone with inverse redox properties (E 00 FAD/ FADH• ≪ E 00 FADH•/FADH −) to donate an electron to the low-potential electron donor ferredoxin [27,46]. Recently, an unstable semiquinone with a half-life of only 10 ps was observed for the FBEB enzyme NADH-dependent ferredoxin-NADP + oxidoreductase [47]. The described different polypeptide environment of the N5 and the N2-C1=O1 groups in sHdrA and mHdrA is in line with the determined inverted redox potentials. (e) The determined midpoint reduction potential of −203 to −188 mV of the [4Fe-4S] cluster is relatively high. Values for [4Fe-4S] +/2+ clusters are mostly below −300 mV [48][49][50] and those for TrxR from T. acidophilum and E. coli are −305 mV (at pH 7, 30°C) and −243 mV (at pH 7, 12°C and in a disulfidic state of TrxR) [51,52], respectively. The [4Fe-4S] cluster of sHdrA is located inside a pocket characterized by a positive electrostatic surface potential, which preferentially stabilizes the reduced [2 Fe 2+ , 2 Fe 2.5+ ] state and thus explains the high redox potential. The redox potential of the FAD/FADH• and FADH•/ FADH − pairs is also relatively high in sHdrA probably due to the stabilization of protonated relative to an unprotonated N5 (Fig. 2B). The redox potential of the persulfide/sulfite pair is unknown but should be lower than or similar to that of the proposed terminal electron acceptor nicotinamide adenine dinucleotide (NAD + ; E 00 = −320 AE 50 mV). (6) The different redox

The Hdr complex of sulfur oxidizers (sHdr)
The genes encoding sHdrA, sHdrB1, sHdrB2, sHdrC1 and sHdrC2 are equivalently arranged in the same operon in all sulfur-oxidizing microorganisms containing shdr genes [8,11,13,14]. Based on the heterohexameric architecture of the recently reported methanogenic Hdr(ABC) 2 complex [12], the postulation of an asymmetric sHdr(AA 0 B 1 B 2 C 1 C 2 ) complex appears to be plausible (Fig. 7). For model building, the structures of the individual subunits were calculated with the SwissModel server [53] and inserted into the mHdr hexamer instead of the corresponding mHdr subunits (Fig. 7). The HdrA-HdrC and HdrB-HdrC interfaces are highly similar between the sulfur-oxidizing and methanogenic Hdr hexamers supporting the reliability of the model. In addition, the predicted asymmetric sHdr(AA 0 B 1 B 2 C 1 C 2 ) complex coincides with the reported subunit composition of the purified sHdr complex from A. aeolicus [28]. Its molecular mass of 240 kDa derived from a Blue native gel [28] is similar to 225.3 kDa calculated for the H. denitrificans heterohexamer on the basis of the genome sequence. The assumed hexameric sHdr architecture assigns sHdrC1/C2, as mHdrC, a function as redox relay between sHdrA and sHdrB1/B2. Like mHdrC, sHdrC1/C2 each contain two [4Fe-4S] clusters deduced from eight strictly conserved cysteines. The peripheral sHdrB1/B2 subunits are attributed a function as catalytic subunits like mHdrB. For CoM-S--S-CoB reduction mHdrB remarkably uses two unique noncubane [4Fe-4S] clusters both spectroscopically and structurally analyzed [54]. The nonmethanogen H. denitrificans lacks the classical CoM-S-S-CoB heterodisulfide and sHdrB1/B2 must possess other Fig. 6. EPR spectra of the as-isolated Hyphomicrobium denitrificans sHdrA. (A, B) EPR spectra taken during the redox titration and (C) Redox titration curve following the g value of 2.05. Linearization of the plot and statistical analysis were performed as described for the absorbance data in the visible range (Fig. 5). Best simulation of the experimental data (black diamonds) was achieved by assuming reduction of one Fe-S center (n = 0.96 AE 0.09, R 2 = 0.96) with an E 00 of −197 mV. Data acquired at the g value of 1.926 yielded comparable results (E 00 = −192 mV, n = 0.92 AE 0.07, R 2 = 0.98) and are not shown here. Temperature of 5 K for all spectra except for spectra in A (18 K). enzymatic activities. sHdrB2 can be reliably assigned as a disulfide reductase due to the conservation of all 10 cysteines ligating the two noncubane [4Fe-4S] clusters in mHdrB [12,14]. In each cluster, four cysteines ligate the four iron atoms and a fifth provides the bridging sulfur of a [3Fe-4S] subcluster. The specific substrate of sHdrB2 remains, however, unknown. The model of sHdrB2 suggests an increase in the size of the substrate binding pocket compared with mHdrB because the loop prior to helix 223 : 236 is shortened in sHdrB2. The functional annotation of sHdrB1 is even more complicated because of significant sequence deviations from the classical active site noncubane iron-sulfur cluster-binding motif of mHdrB [14]. Four of the five cysteines in the proximal iron-sulfur cluster of mHdrB [12] are present in sHdrB1 and the fifth is replaced by an aspartate, a residue known as iron ligand [55,56]. Three of the five cysteines ligating the second, distal noncubane iron-sulfur cluster in mHdrB [12] are conserved in sHdrB1 and two of them (Cys-41 and Cys-81 in mHdrB) are replaced by serine, a residue with reported iron-ligating capacity [57]. Therefore, sHdrB1 may also contain one or two functional Fe/S clusters, albeit not necessarily noncubane [4Fe-4S] clusters. The function of the sHdrC1-sHdrB1 branch and the concrete substrate is unknown, however, no participation in the catalytic process is unlikely. Notably, the binding pocket for CoM-SH is significantly reduced in size in sHdrB1 compared to mHdrB. No information is available about the binding of a redox protein to sHdrA in a suitable electron transfer distance to FAD. Therefore, the postulated catalytic mechanism (see below) does not consider FAD as electron relay.

The reaction of the sHdr complex
The sHdr pathway runs in the cytoplasm of sulfur-oxidizing prokaryotes where reduced sulfur is never processed in a free form, but rather in a protein-bound persulfidic form [11,58]. The sulfur ultimately stems from an oxidized inorganic sulfur substrate; for example, thiosulfate in the case of H. denitrificans and is delivered to the sHdr complex in the form of a protein persulfide via a cascade of sulfur transfer reactions [11,58]. The sHdr complex catalyzes the oxidation of the protein-bound sulfane sulfur to sulfite, the most likely end product [14]. The four released electrons are possibly shuttled to an oxidized lipoamide bound to the sHdr-specific LbpA [13] and the formed dihydrolipoamide is regenerated for the next reaction cycle by NAD + -reducing dihydrolipoamide dehydrogenase (DLDH). The limited data basis allows several preliminary mechanistic proposals. We will, however, only present one scenario in more detail that is compatible with the currently available data (Fig. 8). Accordingly, the formation of sulfite (SO 2À 3 ) from a protein-bound sulfane sulfur, R 1 -CysS-S − is completely executed at the sHdrB1 active site. The sulfane sulfur and the cysteine sulfur may each be bound to an iron of the two postulated iron-sulfur centers of sHdrB1. Subsequently, the Fe-ligated sulfane sulfur is oxidized to the Fe-ligated sulfenate (-S-OH) and/or sulfinate (ÀS À O À 2 ) intermediates en route to sulfite, reminiscent of the reactions on the siroheme iron [59]. Then, the released electrons could flow from sHdrB1 to Hdr-like complex (sHdrAA 0 B1C1B2C2). For model building of the sHdr hexamer, the structures of sHdrA (gray), sHdrA 0 (blue), sHdrC1 (lime), sHdrB1 (orange), sHdrC2 (green), and sHdrB2 (salmon) subunits were calculated by the SwissProt server based on the template of the mHdrBC. The sequence identities between sHdrC1(C2) and mHdrC are 18.7 (34.7) %, and between sHdrB1(B2) and mHdrB 21.5 (32.4) %. Subsequently, the mHdrA, mHdrB, and mHdrC subunits were exchanged by the corresponding sHdr subunits to obtain the structural model of the sHdr hexamer. The sHdrA-sHdrC and mHdrA-mHdrC interfaces are highly similar and do not involve the ferredoxin insertion/extension domains of mHdrA truncated in sHdrA as contact area [12]. Except for the differences in the sequence, sHdrB1 contains a prolonged N-and C-terminal extensions of ca. 100 and 50 residues compared to mHdrB and sHdrB2. While sHdrC1 was confidently modeled with mHdrC as template, the modeling of sHdrC2 for the last 70 amino acids failed. The terminal 70 residues are adjacent to the active site of sHdrB and might be crucial for substrate binding and specificity. We assume that sHdrC2 and sHdrB2 form one branch due to their closer relationship to mHdrB. The figure was generated with PYMOL. sHdrB2 via the two [4Fe-4S] clusters of the sHdrA dimer which conducts electrons between the two sHdrBC branches. The two reduced noncubane [4Fe-4S] clusters of sHdrB2 would reduce oxidized lipoamide disulfide analogous to those of mHdrB that reduce the CoM-S-S-CoB heterodisulfide [12]. Dihydrolipoamide is finally reoxidized by NAD + reduction. In this mechanism, the two flavins have only a function as an electron sink in order to partly accept the four electrons rapidly released from the sulfur during oxidation at HdrB1 (Fig. 8). As pointed out, other mechanisms cannot be excluded yet. For example, the oxidation of the protein-bound sulfane sulfur to sulfite may not occur at one active site as described (Fig. 8) but instead in a concerted fashion at the active sites of sHdrB1 and sHdrB2. Further experiments, in particular, with the entire sHdr complex are required to shed light on the structure of the Fe/S clusters and on the mode of action of the sHdr complex.

Production of recombinant sHdrA
The H. denitrificans hdrA gene with an additional carboxyterminal Strep-tag encoding sequence was cloned in plasmid pET22b and overexpressed in E. coli BL21 (DE3) ΔiscR. One liter batches of lysogeny broth medium containing 100 mM 3-(N-morpholino)propanesulfonic acid (MOPS) buffer pH 7.4, 25 mM glucose and 2 mM iron ammonium citrate as well as ampicillin and kanamycin were inoculated with 5% (v/v) E. coli precultures hosting plasmid pET-22bHdhdrA [14] and cultivated in 2-L flasks at 37°C and 180 r.p.m. until an OD600 of 0.4-0.6 was reached. Cultures were then moved into an anaerobic chamber (Coy Laboratory Products, Grass Lake, MI, USA) containing 98% N 2 and 2% H 2 . Cysteine (0.5 mM), sodium fumarate (25 mM) and IPTG (0.1 mM) were added. Cultures were then transferred into completely filled and tightly closed 500 mL bottles, incubated for 48-72 h at 16°C and harvested by centrifugation (11 000 g, 12 min). Cells were lysed by sonication in the anaerobic chamber. After removal of insoluble cell material by centrifugation (16 100 g for 30 min at 4°C) protein was purified under strictly anaerobic conditions by Strep-Tactin affinity chromatography according to the manufacturer's instructions (IBA Lifesciences, Göttingen, Germany) followed by concentration to a final volume of < 2 mL via Amicon Ultracel-10K filters (Merck Millipore, Tullagreen, Ireland). The protein was stored under anaerobic conditions at −70°C.
One liter E. coli culture routinely yielded 7-8 mg of sHdrA. Purity was assessed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis.

Quantification of iron and sulfur
Quantification of non-heme iron and acid-labile sulfur was conducted as described by Fish [60] and King and Morris [61], respectively.

UV-visible titration of sHdrA
UV-visible absorbance spectroscopy was carried out at 20°C on a Specord 210 UV/Vis-spectrophotometer (Analytik Jena, Jena, Germany). The protein sample (50 µM) was prepared in The proposed scenario of the reaction from a cysteine-bound persulfide (R 1 -CysS-S − ) to sulfite is based on the assumption that sHdrB2 performs a disulfide/thiolate redox reaction, sHdrB1 acts as redox carrier/active site and sHdrA as redox carrier/storage unit but not as hydride donor for a disulfide-containing molecule or NADH. The sHdrB1/B2, HdrC1/C2, and HdrA subunits are depicted in orange, blue, and green, respectively. The following prosthetic groups are indicated: orange boxes, [4Fe-4S] clusters; dark red diamonds, noncubane iron-sulfur clusters; light red diamonds, iron-sulfur clusters of unclear nature; and yellow oval, FAD. Thiol groups and persulfides are shown in the protonated or ionized state according to their supposed pK a values of around 8.5 [73] and 6.2 [74], respectively. Due to high reactivity, measuring the pK a for a sulfenic acid is complicated and, as far as we are aware, estimated in a range of 6-10 or a value of 7.6 AE 0.3 for cysteine sulfenic acid [75,76]. The sulfinic acid functional group has a pK a value at around 2 and is always deprotonated under physiological conditions [77]. We assume that R 1 is a sulfur carrier protein encoded in the shdr gene cluster or in its immediate environment, that is, TusA or DsrE [14]. The depicted scenario suggests that the sulfane sulfur is completely oxidized at the postulated iron-sulfur cluster(s) of HdrB1 yielding sulfite and a regenerated sulfur carrier protein R 1 -CysSH. 500 µL of 200 mM ammonium acetate buffer (pH 6.0) and assembled in a quartz glass cuvette (Hellma Analytics, Müllheim, Germany) in the Coy anaerobic chamber. The cuvette was sealed with airtight septa, and titanium (III)-citrate [40] as reductant and potassium ferricyanide as oxidant were added via a gas-tight Hamilton syringe. All spectra were normalized to their absorption at 750 nm.

EPR spectroscopy and potentiometric redox titrations via EPR
Electron paramagnetic resonance spectra at X-band were obtained using a Bruker EMX spectrometer equipped with an ESR-900 continuous flow of helium cryostat from Oxford Instruments (Abingdon, UK). Spectra were recorded under the following conditions: microwave frequency, 9.39 GHz; microwave power, 2.01 mW; modulation frequency, 100 kHz; modulation amplitude, 1 mT. The EPR-based potentiometric titration was performed inside an anaerobic chamber at 25°C using 110 μM of HdrA and 180 μM of the same mixture of redox mediators in 100 mM MOPS pH 7.5, 5 mM EDTA. Samples were transferred to EPR tubes under anaerobic conditions, capped, and immediately frozen in liquid nitrogen upon removal from the chamber.

Crystallization and structure determination
Crystallization trials were performed using the sitting-drop method at 18°C in an anaerobic chamber (Coy Laboratory Products, Grass Lake, MI, USA) with a gas phase of N 2 /H 2 (95 : 5 [v/v]) equipped with an OryxNano crystallization robot (Douglas Instruments Ltd., Hungerford, UK). The applied sHdrA solution and the crystallization conditions found with the JBScreen Pentaerythritol (Jena Bioscience) are given in Table 1. Yellow-brownish crystals diffracted to ca. 1.4Å at the beamline PXII at the Swiss-Light-Source in Villigen (Switzerland). A dataset collected at a wavelength of 1Å was processed with XDS [62]. A second dataset was measured at the iron edge (1.739Å) at 2.2Å resolution ( Table 1). The iron positions of the two [4Fe-4S] clusters in the asymmetric unit were identified by SHELXD [63]. Phases were determined by using SHARP [64] and improved by SOLOMON [65]. The model was essentially built automatically by using ARP/wARP [66] and manually completed by using COOT [67]. Refinement was carried out with REFMAC5 [68], PHENIX.REFINE [69] and BUSTER (Phaser; Global Phasing Ltd., Cambridge, UK). The quality of the model was evaluated by COOT and MOLPROBITY [70]. Structure comparison calculations were performed with DALI [71] and surface area calculations by the PISA server [72]. Figures 1, 2A, 3A, and 7 were generated with PYMOL (Schrödinger, LLC, New York, NY, USA). The atomic coordinates and structure factors of sHdrA have been deposited in the Protein Data Bank, www.pdb.org with ID code 6TJR.