BMC Structural Biology BioMed Central

Background: The initial step involved in oxidative hydroxylation of monoaromatic and polyaromatic compounds by the microorganism Sphingobium yanoikuyae strain B1 (B1), previously known as Sphingomonas yanoikuyae strain B1 and Beijerinckia sp. strain B1, is performed by a set of multiple terminal Rieske non-heme iron oxygenases. These enzymes share a single electron donor system consisting of a reductase and a ferredoxin (BPDO-F B1). One of the terminal Rieske oxygenases, biphenyl 2,3-dioxygenase (BPDO-O B1), is responsible for B1's ability to dihydroxylate large aromatic compounds, such as chrysene and benzo[a]pyrene.


Background
Sphingobium yanoikuyae B1 (B1), previously known as Sphingomonas yanoikuyae B1 and Beijerinckia sp. strain B1 [1], was isolated by virtue of its ability to use biphenyl as its sole source of carbon and energy for growth [2]. B1 is capable of using naphthalene, phenanthrene, anthracene, toluene, m-and p-xylene as sole sources of carbon and energy [3]. Other compounds are also oxidized by this microorganism and many of these are converted to cisdihydrodiols. B1 remains one of the only known microbes, along with Mycobacterium vanbaalenii PYR1 [4] and Sphingomonas sp. strain CHY-1 [5][6][7], able to oxidize large aromatic hydrocarbons such as benzo[a]pyrene, benzo[a]anthracene and chrysene [8] to cis-dihydrodiols.
In the B1 genome, at least six sets of putative oxygenase genes are present [9] and are all believed to share a common electron donor system [10]. The genes bphA1A2f, which encode BPDO-O B1 , have been sequenced and found to encode a Rieske oxygenase (RO) [11][12][13] related to naphthalene and biphenyl dioxygenases [14]. BPDO-O B1 is responsible for the oxidation of large aromatic compounds, such as benzo[a]pyrene, by B1. Structural information [15][16][17] and molecular modeling [18] have been used to determine features important for substrate specificity in other biphenyl oxidizing oxygenases; however, most of the effort has been targeted at understanding how biphenyl dioxygenases catalyze the degradation of polychlorinated biphenyls (reviewed in [19][20][21][22]). Previous structural studies of several ROs provide insight into common features of how the terminal component of the RO systems are organized and how substrate interacts with the enzyme to form the hydroxylated product [23]. To date, structures of enzymes that catalyze cis-dihydroxylation of aromatic substrates with more than three rings have not been reported.
Rieske oxygenase systems have multiple components whose function is to transfer electrons from NAD(P)H to active molecular oxygen and ultimately oxidize aromatic hydrocarbon substrates [23]. A single reductase for the multiple Rieske oxygenases is present in the B1 genome [10]. This reductase passes one electron at a time from NAD(P)H to the Rieske ferredoxin, BPDO-F B1 , which in turn passes the electron on to the dioxygenase enzyme [24]. The terminal oxygenase component is responsible for catalyzing the addition of molecular oxygen to the aromatic substrate. This occurs at the mononuclear iron, contained in a large, primarily hydrophobic active site. The residues that form the active site have been shown to control substrate specificity.
Here we report the structures of BPDO-F B1 and BPDO-O B1 . The structure of BPDO-F B1 shows similarities and important differences compared to other known Rieske dioxygenase ferredoxins. Structures of BPDO-O B1 , in both the native form and bound to biphenyl, are presented. These structures demonstrate how BPDO-O B1 binds substrate in the active site. We also discuss the similarities and differences of this terminal oxygenase to other RO terminal oxygenase structures that have been previously determined and how these differences play a role in substrate specificity and regio-and stereoselectivity of product formation.

Ferredoxin structure determination
The asymmetric unit contains two copies of the BPDO-F B1 molecule. The final model contains residues 3 -104 in chains A and B. The structure has been refined to a resolution of 1.9 Å with a final R factor of 19.3% and an R free of 24.0%. The first two N-terminal and last three C-terminal residues could not be modeled into the electron density. The superposition of all Cα atoms in both chains using Lsqman [25] had an RMSD of 0.40 Å. Residues Lys-25, Pro-104, and Glu-95 through Gly-97 had the largest deviations between chains A and B. The surface loop containing Lys-25 and Met-26 had little to no density for their side chains, Pro-104 is the last residue modeled, while Asp-96 assumes two conformations in chain B. The Rieske [2Fe-2S] cluster had isotropic displacement factors of 19.6 and 18.5 Å 2 for chains A and B respectively. Crystallographic statistics are reported in Table 1.

Oxygenase structure determination
The asymmetric unit contains the entire BPDO-O B1 α 3 β 3 hexamer and residues 6 -454 of the α subunits and residues 5 -174 of the β subunits were modeled into the electron density. The structure has been refined to a resolution of 1.7 Å with a final R factor of 18.8 % and an R free of 22.9 %. The loop region located at the entrance of the α subunit active site is disordered, with higher than average Bfactors and low side-chain electron density. This region spanned from residues 220 -240 with residues 235 -239 being the most disordered. Corresponding regions in other ROs are also disordered [15,[26][27][28][29][30]. Lsqkab [25] was used to determine the superposition RMSDs of the three α and β subunits in the asymmetric unit. A few regions in the α subunit, residues 108 -123 and 411 -434, and in the β subunit, residues 77 -85 and 139 -148, had higher than average RMSDs when compared; however, electron density clearly defined the coordinates of these residues. Crystallographic statistics are reported in Table 1.

Structure of the substrate-free oxygenase enzyme
The three copies of the α and β subunits form an α 3 β 3 quaternary structure similar to the quaternary structure observed in all known RO structures. The hexamer forms a mushroom-shaped structure, illustrated in Figure 1, where the three α subunits form the cap and the three β subunits form the stem. The hexamer structure is believed to be the active biological unit, similar to other known ROs.
The α subunit consists of 454 residues, which form two domains. The N-terminal portion consists of a Rieske [2Fe-2S] iron-sulfur cluster domain, defined by residues 40 -140. The Rieske domain consists of β strands and loops, forming an ISP domain fold [31]. Four residues, two histidines and two cysteins, in the α subunit coordinate the Rieske non-heme iron cluster [2Fe-2S]. His-82 and His-103 coordinate one iron, while Cys-80 and Cys-100 coordinate the other. The C-terminal domain is a mix of helices and strands forming a TBP-like or helix-grip fold and is a member of the Bet v1-like superfamily [31]. The structural conservation suggests that the intraprotein electron transport in BPDO-O B1 is similar to the system described for NDO-O 9816-4 and benzoate dioxygenase [32][33][34][35][36][37].
The mononuclear iron is coordinated by two histidines, His-207 and His-212, and one aspartate, Asp-360, at the rear of the active site. This iron has been shown to bind water(s) or dioxygen in other structures. In BPDO-O B1 , a clear bimodal electron density distribution was not observed, but instead an egg-shaped electron density above the mononuclear iron was present ( Figure 2). When a single water/hydroxide molecule was modeled in this position, it resulted in residual positive density on the Fo-Fc electron density maps. Residual positive electron density was not observed on the Fo-Fc electron density maps when modeling molecular oxygen as the fourth iron ligand; however, refinement with relaxed restraints (refinement restraints allowing the oxygen-oxygen bond an RMSD of 0.3Å) placed the oxygen-oxygen bond distance at 0.8 -1.1Å in the three subunits. This is slightly shorter than the average distance of 1.21 Å between two oxygen atoms in O 2 . B-factors for the refined oxygen molecule also suggested that it was less than fully occupied. The final model has both molecular oxygen and water bound to the iron in partially occupied states ( Figure 2). In previous studies, both dioxygen species [15,27,38,39] and water [16,26,28,29,[39][40][41][42] have been observed as ligands in RO crystals grown in atmospheric conditions. The Outer shell values are in parentheses. 2 , where I i (hkl) is the ith measurement of reflection hkl and I(hkl) is the average for that reflection. 3 , where F obs and F calc are the observed and calculated structure factors, respectively. R free is the same, but for a test set of reflections not used in refinement. Top and side view of the BPDO-O B1 hexamer. The structure of biphenyl 2,3-dioxygenase is a mushroom-shaped α 3 β 3 hexamer. This quaternary structure is typical of αβ Rieske oxygenases. This structure allows a mononuclear iron from one α subunit to come within ~15 Å of a Rieske cluster from a neighboring subunit, allowing electron transfer to take place.

R I hkl I hkl
BPDO-O B1 crystals were grown in atmospheric conditions, therefore molecular oxygen or water could constitute the fourth ligand on the mononuclear iron.
The β subunit of BPDO-O B1 has a Cystatin-like protein fold and belongs to the protein superfamily of NTF2-like proteins [31]. The function of the β subunits of ROs is not well understood and reports vary as to whether or not mutations in the β subunit of ROs can influence regioselectivity of product formation (reviewed in [43]). In the case of BPDO-O B1 , the β subunit is not directly involved in creating the topology of the active site; however, the mononuclear iron is within 11 Å of the α-β subunit interface. This distance may allow select β-subunit side chains to indirectly affect the topology of the active site by interacting with residues that directly form the active site.

Structure of the biphenyl bound oxygenase enzyme
B1 has been shown to catalyze the dihydroxylation of biphenyl at positions 2 and 3 of the carbon ring [2,44] and recent studies have confirmed that BPDO-O B1 is responsible for this activity [45]. Previous structural investigations of ROs have demonstrated that substrate bound in the active site is oriented such that the carbon(s) oxidized in the dihydroxylation reaction is (are) closest to the mononuclear iron [16,26,27,30,40,42,46]. This trend is also seen in the structure of BPDO-O B1 bound to biphenyl. The 2 and 3 carbons of biphenyl are positioned closest to the iron, with a water/hydroxide molecule bound to the iron positioned between the substrate and the iron ( Figure 3). Crystals were grown and biphenyl was added in the presence of atmospheric oxygen. The enzyme was oxidized and no electron source was added; therefore, catalysis did not occur. We are unable to determine whether the iron is coordinated to a water molecule (hydroxide ion) or an oxygen molecule at this resolution. No significant changes in the active site main chain were observed near the mononuclear iron compared to the native structure. However, small changes in the mainchain and side-chain positions of the distal portion of the active site were observed. The largest changes were seen in the loop regions that cover the entrance to the active site. The loop main-chain was pushed slightly away (~0.2-0.3 Å) from the substrate in the complex structure, while the main-chain proximal to the mononuclear iron remained static.

Comparison of BPDO-F B1 to other Rieske oxygenase ferredoxins
The structure of the biphenyl dioxygenase system's ferredoxin component, BPDO-F B1 , is the fourth Rieske oxygenase ferredoxin structure to be determined and shares significant sequence and structural homology with these proteins ( Table 2). BPDO-F B1 has structural features similar to other ferredoxins including three stacked beta sheets and a solvent exposed Rieske cluster ( Figure 4) [47]. The protein fold surrounding the Rieske cluster is similar in all Rieske ferredoxins, including the high-potential Rieske ferredoxins found in respiratory electron transport chains such as the bc 1 complex [48] and the Rieske clusters found in dioxygenase enzymes such as BPDO-O B1 .
In addition to the conserved CXH-CXXH motif present in all Rieske ferredoxins, there are two additional conserved residues, Phe-71 and Pro-82, in the dioxygenase ferredoxins [47,49,50]. The phenylalanine is part of the small ferredoxin core near the Rieske cluster. The neighboring side-chains include Thr-46, Leu-52 and Ile-87. These residues are highly conserved with Thr-46 most commonly being a serine or a threonine, Leu-52 existing almost exclusively as a leucine, and Ile-87 being the most varied with leucine as the most common substitution ( Table 3). The conserved proline is located at a hairpin turn at the apex of the ferredoxin and has previously been classified as part of a polyproline loop in Rieske ferredoxins ( Figure  5). Unlike the biphenyl ferredoxin from LB400, which has three consecutive prolines, this ferredoxin has only a single proline in the loop.
The reduction potential of the Rieske cluster in bacterial dioxygenase systems is expected to be approximately -150 mV, similar to that found in BPDO-F LB400 [47] and NDO-F 9816-4 (Lindsay Eltis, personal communicaton). It is believed that the local electrostatic environment, created by charged and hydrogen bonded residues near the cluster, differentiates the reduction potential of the cluster from homologous structures. Thus the low potential oxygenase ferredoxins, with negative reduction potentials, have an electrostatic environment that is more negative Mononuclear iron bound to oxygen and water than the mitochondrial Rieske ferredoxins, with positive reduction potentials. BPDO-F B1 is the ideal protein for exploring these effects; unlike the other Rieske oxygenase ferredoxins, this protein contains a residue which hydrogen bonds to the cluster through the side-chain (Figure 6), as opposed to through the main-chain as present in other ferredoxins. Thus substitution of Cys-83 with alanine, valine, or serine, can probe the effect of local charge and hydrogen bonding on the reduction potential. Interestingly, the conserved Phe-71 in dioxygenase Rieske ferredoxins is exclusively a tyrosine in the Rieske ferredoxins found in mitochondrial and chloroplast electron transport chains. The increased polarity or hydrogen bonding ability of the tyrosine may assist in raising the reduction potential of these ferredoxins. This provides a second rational target for mutational analysis.

Comparison of BPDO-O B1 to other Rieske oxygenases
The X-ray diffraction structural model shows that BPDO-O B1 is structurally similar to other known ROs, as predicted by sequence analysis. Structure and sequence alignment confirm that NDO-O 9816-4 is the most structurally similar, with most of the variation in the α subunit around the active site. Table 2 [16]. Loops covering the active site in the α 3 Rieske monooxygenase 2-oxoquinoline 8-monooxygenase also demonstrate changes upon substrate binding [30]. This loop motion observed in the BPDO-O B1 and other RO structures allows the active site to "breathe" to accommodate ligands and may be one of the key features that allows this class of enzymes to perform catalysis on diverse sets of substrates with respect to overall size and shape. It also provides a framework to interpret the product regioselectivity of BPDO-O B1 and the differences in product regioselectivity compared to other ROs. While the Rieske ferredoxin structure is very similar to other Rieske ferredoxins, the differences in amino acid composition near the cluster in BPDO-F B1 provide a unique opportunity among the Rieske oxygenase ferredoxins to examine the effect of cluster environment and hydrogen bonding on reduction potential.

Protein expression, purification and crystallization
BPDO-O B1 , bphA1fA2f [11,14] and BPDO-F B1 , bphA4 [10], were each cloned from B1 into the protein expression vectors pET101D (Invitrogen, Carlsbad, CA) and pT7-7, respectively, and expressed as described in Yu et al. [45]. Crystallization of the ferredoxin protein was performed using 33 mg/mL of BPDO-F B1 protein in a 50 mM phosphate buffer, pH 6.8, with 0.1 M citric acid and 1.6 M ammonium sulfate as precipitants. Crystallization of the oxygenase protein was performed using purified BPDO-O B1 protein at 20 mg/ml in a 20 mM potassium phosphate buffer, pH 6.8 [45]. Mineral oil was used as a cryoprotectant for the crystals.
Oxygenase crystals were also used for soaking experiments in an attempt to generate the BPDO-O B1 protein-biphenyl complex. Soaking experiments were based on similar experiments previously done with NDO-O 9816-4 to generate crystals of ligand-bound enzyme [27,38,40]. Crystals were moved to fresh drops containing reservoir buffer. Ethanol saturated with biphenyl was added to these drops. The crystals tolerated up to 5% ethanol in the drop. Crystals were allowed to soak for 3-5 days, then were removed from the drop and flash-cooled to 100 K. Mineral oil was used as a cryoprotectant.
Data collection, processing, structure solution and refinement X-ray diffraction data for BPDO-F B1 were collected on beamline X6A at Brookhaven National Laboratory and data for BPDO-O B1 were collected on the IMCA-CAT beamline 17-ID at the Advanced Photon Source in Argonne National Laboratory. Crystallographic statistics are presented in Table 1. d*TREK [53] was used to process the data for BPDO-F B1 to a resolution of 1.60Å. Analyzing systematic absences, the space group was determined to be either P6 1 22 or P6 5 22. Molecular replacement using AMoRe [54] and a polyalanine model based on BPDO-F LB400 (PDB entry 1FQT) produced a solution in space group P6 5 22. Model building using O [55] and Coot [56], density modification using DM, and refinement using Refmac5 [57] from the CCP4-4.0 program suite were assisted by non-crystallographic symmetry between the two monomers in the asymmetric unit. These NCS restraints were loosened as refinement progressed. Cycles of Refmac5 with ARP/wARP [58,59] or the Coot find waters routines were used to identify solvent atoms. Asp-96 in chain B was modeled as having two side chain conformations. A single round of TLS optimization was used at the end of refinement with each protein monomer acting as a TLS group.
Data collected from native BPDO-O B1 crystals was processed using d*TREK [53] to a resolution of 1.70 Å. Analysis of systematic absences suggested that the space group was P3 x 21. Molecular replacement was performed using a polyalanine model based on NDO-O 9816-4 (PDB entry 1NDO) [28]. Molecular replacement using AMoRe [54] gave a clear solution in the space group P3 1 21. Initial refinement of the polyalanine model with the program Refmac5 [57] of the CCP4-5.0.2 [60] suite of programs yielded good starting electron densities. The molecular visualization program O [55] was used for model building. After the bulk of the structure was modeled, refinement was continued with Refmac5 without NCS restraints. Solvent molecules were found using the program Arp/Warp [58,59] and multiple side-chain confor- In the structural superposition, residues marked with an asterisk were more than 3.0 Å from the corresponding residue in BPDO-O B1 .
Residues coordinating Rieske [2Fe-2S] cluster in BPDO-F B1 Figure 6 Residues coordinating Rieske [2Fe-2S] cluster in BPDO-F B1 . All residues interacting with Rieske iron-sulfur cluster or its ligands. Hydrogen bonds are displayed in dashed lines. Gly-48 interacts with His-47 and Ala-50 via hydrophobic interactions. Note the interaction between the side-chain of Cys-83 and the iron-sulfur cluster, a feature not present in other known Rieske oxygenase ferredoxins.
Data from biphenyl soaked oxygenase crystals were processed using d*TREK to a resolution of 2.8 Å. The native BPDO-O B1 structure, with solvent molecules removed, was used as a starting point for the refinement of the ligand bound structure. An energy-minimized structure of the small molecule biphenyl was constructed using the program SYBYL 7.1 [63]. This model was used to create a refinement dictionary for Refmac5 using the ligand sketcher program in CCP4-5.0.2. After initial refinement of the protein model, the biphenyl ligand was modeled into the active site of the enzyme where electron density maps showed un-modeled density in both 2Fo-Fc and Fo-Fc maps. The torsion angle between the two rings of biphenyl was allowed to rotate during refinement and solvent molecules were modeled as appropriate.

Sequence and structural alignments
TCoffee [64] was used to produce a structure-based sequence alignment of the four known RO ferredoxin structures and 19 other RO ferredoxin sequences. Structural alignment of native BPDO-O B1 with structures of other RO oxygenases was performed using the program Indonesia [65]. Structural alignments in Indonesia were done pairwise using the Levitt and Gerstein method with a cut-off value of 3.5 Å. Sequence alignment and comparison was done using full sequences for each of the proteins without structural information. Protein-protein sequence alignments were performed using blastp on the NCBI blast server [66]. All figures showing structures were created using PyMOL 0.98 [67]. Figure 6 was produced using a modified version of LigPlot [68].
Comparison of biphenyl 2,3-dioxygenase active site to other Rieske oxygenaes Additional material