Thioredoxin reductase from Bacillus cereus exhibits distinct reduction and NADPH‐binding properties

Low‐molecular‐weight (low M r) thioredoxin reductases (TrxRs) are homodimeric NADPH‐dependent dithiol flavoenzymes that reduce thioredoxins (Trxs) or Trx‐like proteins involved in the activation networks of enzymes, such as the bacterial class Ib ribonucleotide reductase (RNR). During the last few decades, TrxR‐like ferredoxin/flavodoxin NADP+ oxidoreductases (FNRs) have been discovered and characterized in several types of bacteria, including those not encoding the canonical plant‐type FNR. In Bacillus cereus, a TrxR‐like FNR has been shown to reduce the flavodoxin‐like protein NrdI in the activation of class Ib RNR. However, some species only encode TrxR and lack the homologous TrxR‐like FNR. Due to the structural similarity between TrxRs and TrxR‐like FNRs, as well as variations in their occurrence in different microorganisms, we hypothesized that low M r TrxR may be able to replace TrxR‐like FNR in, for example, the reduction of NrdI. In this study, characterization of TrxR from B. cereus has revealed a weak FNR activity toward NrdI reduction. Additionally, the crystal structure shows that only one out of two binding sites of the B. cereus TrxR homodimer is occupied with NADPH, indicating a possible asymmetric co‐substrate binding in TrxR.

Low-molecular-weight (low M r ) thioredoxin reductases (TrxRs) are homodimeric NADPH-dependent dithiol flavoenzymes that reduce thioredoxins (Trxs) or Trx-like proteins involved in the activation networks of enzymes, such as the bacterial class Ib ribonucleotide reductase (RNR). During the last few decades, TrxR-like ferredoxin/flavodoxin NADP + oxidoreductases (FNRs) have been discovered and characterized in several types of bacteria, including those not encoding the canonical plant-type FNR. In Bacillus cereus, a TrxR-like FNR has been shown to reduce the flavodoxin-like protein NrdI in the activation of class Ib RNR. However, some species only encode TrxR and lack the homologous TrxR-like FNR. Due to the structural similarity between TrxRs and TrxR-like FNRs, as well as variations in their occurrence in different microorganisms, we hypothesized that low M r TrxR may be able to replace TrxR-like FNR in, for example, the reduction of NrdI. In this study, characterization of TrxR from B. cereus has revealed a weak FNR activity toward NrdI reduction. Additionally, the crystal structure shows that only one out of two binding sites of the B. cereus TrxR homodimer is occupied with NADPH, indicating a possible asymmetric co-substrate binding in TrxR.
Thioredoxin reductases (TrxRs) are members of the pyridine nucleotide-disulfide oxidoreductase family of flavoenzymes [1,2]. As part of the thioredoxin (Trx) system, comprising nicotinamide adenine dinucleotide phosphate (NADPH), TrxR, and Trx, TrxR is a member of a major disulfide reductase system which can provide electrons to a large range of enzymes. The TrxR-Trx system participates in multiple important biological processes. It plays a preventive role in oxidative stress in bacteria, archaea, and eukarya, and in regulation of DNA synthesis, gene transcription, cell growth, and apoptosis in bacteria, yeast, and mammals [3][4][5][6]. TrxR contains a redox-active disulfide/dithiol adjacent to a flavin adenine dinucleotide (FAD) cofactor [7] and reduces its substrate Trx through this highly conserved cysteine pair (CXXC). The regeneration is aided by the reduction of the enzyme-bound FAD group by NADPH [8,9].
In low M r TrxRs, both the NADPH-binding site and the CXXC motif are positioned on the re-face of FAD and hence cannot simultaneously be positioned close to the FAD isoalloxazine ring. Therefore, a˜67°r otation of the NADPH-binding domain relative to the FAD-binding domain is essential for sequential and efficient electron transfer from NADPH to FAD and from FADH 2 to the redox-active disulfide. Hence, the NADPH-and FAD-binding domains are found in two conformational states: one where the FAD cofactor is reduced by NADPH (FR, flavin-reducing conformation) and one where the FAD cofactor is reoxidized by the active-site disulfide (FO, flavinoxidizing conformation), interchanged by the˜67°r otation [14][15][16]. The rotation between the two domains also allows the redox-active cysteines to interact with its substrate, Trx, and the dithiol form of Trx can further reduce its target substrate. Trx was originally discovered to be a reducing substrate of ribonucleotide reductase (RNR) [17]. This essential enzyme catalyzes the de novo synthesis of 2 0deoxyribonucleotides from their corresponding ribonucleotides and is therefore crucial for DNA repair and replication in all DNA-based living organisms [4,5,18,19]. The reducing equivalents provided by the TrxR-Trx system are essential for the catalytic activity of oxygen-dependent class I and II RNRs.
In addition to the thiol-based redox pathway, the bacterial class Ib RNR has been shown to rely on another redox partner, the small flavodoxin (Fld)-like protein NrdI [20][21][22]   The TrxR-like FNRs are widely found in firmicutes, but can also be found in other bacterial phyla encoding the class Ib RNR and NrdI. However, for example, Escherichia coli (Ec) does not contain any TrxRlike FNRs, Deinococcus radiodurans (Dr) only contains a Bdr and no FNR1 or FNR2 homologs, whereas the firmicute S. aureus does not contain a B. cereus FNR2 homolog. Due to the lack of TrxR-like FNRs (in particular homologs of B. cereus FNR2) in some bacteria encoding class Ib RNR, and due to the structural similarities between FNRs and TrxR, it should be explored whether TrxR could possibly serve as an alternative redox partner to NrdI and hence activation of the class Ib RNRs. Furthermore, to understand the differences between TrxR and TrxR-like FNRs in terms of redox partner selectivity, it is of interest to investigate which structural features could account for variations in substrate specificity.
In this work, we tried to answer whether TrxR could potentially replace TrxR-like FNRs by monitoring their activity as reductants of NrdI. Also, by introducing mutations to TrxR, we aimed to map which structural variations in these two enzymes could explain the differences in activity toward the NrdI and Trx substrates for FNR and TrxR, respectively, and if the TrxR mutant, designed to resemble an FNR, would reveal an increased reduction rate toward NrdI as a redox partner. By choosing B. cereus as a model organism, encoding both TrxR and three TrxR-like FNRs, a direct comparison of the reduction rates between these different enzymes could be made. By solving the crystal structure of B. cereus TrxR, we have also observed structural features indicating a possible half-of-the-sites reactivity with respect to NADPH in B. cereus TrxR.

Cloning, protein expression, and purification
TrxR and TrxR mutant (TrxR wt and TrxR mutant ) were expressed and purified as described previously for the B. cereus FNRs [24], with some modifications. In brief, genes for TrxR wt and TrxR mutant (locus tag BC5159, B. cereus ATCC 14579, GenScript) were synthesized and cloned into the pET-22b(+) plasmid using restriction enzymes NdeI and HindIII (GenScript), and transformed into competent E. coli BL21 (DE3) cells (Novagen). For the TrxR mutant , mutations were introduced by GenScript according to the description in the 'Results and Discussion' section. Cells were grown in Terrific Broth (TB) medium containing 100 lgÁmL À1 ampicillin. Protein expression was induced by adding isopropyl b-D-1-thiogalactopyranoside (IPTG, Thermo Scientific, Oslo, Norway) to a final concentration of 0.5 mM, and the cultures were incubated for 12-15 h at 20°C with vigorous shaking before harvesting. The frozen cell paste was lysed either using an X-press [32] or by sonication (1 : 5 ratio of cell wet weight-to-buffer volume), dissolved in 100 mM Tris/HCl, pH 7.5, 2 mM DTT, 5 lgÁmL À1 DNase, protein inhibitor cocktail tablet (Roche, Oslo, Norway), and cleared from nucleic acids by streptomycin sulfate (2.5%). Proteins were precipitated with 0.45-0.6 gÁmL À1 ammonium sulfate, dissolved in 50 mM Tris/ HCl, pH 7.5, and desalted using a HiTrap Desalting column (GE Healthcare, Oslo, Norway). Desalted protein was applied on a HiTrap HP Q column (GE Healthcare), eluted with a 0.1-0.4 M KCl gradient, desalted as described previously, and applied on a Source 15Q column (GE Healthcare), eluted with a 0-0.3 M KCl gradient. Proteins were concentrated, aliquoted, flash-frozen in liquid N 2 , and stored at À80°C. Protein concentration was determined using a molar extinction coefficient of 11.3 mM À1 Ácm À1 for the protein-bound flavin at 450 nm [33].

Crystal data collection and refinement
Diffraction data were collected at beamline ID23-1 at the European Synchrotron Radiation Facility (ESRF), Grenoble, France. The dataset was processed with MOSFLM [34], scaled, and merged using AIMLESS [35,36], and the space group confirmed with POINTLESS through the CCP4 software suite [37]. The TrxR wt structure was solved by molecular replacement using PHASER [38]. A homolog model of B. cereus TrxR wt modeled with SWISS-MODEL [39,40] from the S. aureus TrxR Protein Data Bank identification code (PDBid: 4GCM) was used as a search model in PHASER to solve the B. cereus TrxR wt structure. Initial refinement was done using restrained refinement in REFMAC5 [41] followed by several cycles of refinement with phenix.refine [42] in the Phenix suite [43] and model building in Coot [44]. All structure figures were prepared using PyMOL (Schr€ odinger, LLC, New York, NY, USA).
The standard Phenix restraints used for the FAD cofactor were modified to take into account potential X-ray radiation-induced reduction of the FAD cofactor making the isoalloxazine ring free to bend along the N5-N10 axis  . Similarly, the disulfide bond restrains were loosened to take into account potential reduction. The angle of the butterfly bend of the isoalloxazine ring was calculated with the psico module in PyMOL, by calculating the angle between the two planes defined by atoms N5, C4X, C4, N3, C2, C10, N10, and N5, C5X, C3, C7, C8, C9, C9A, and N10. A slight butterfly bend of 3.1°and 4.8°w as observed for the isoalloxazine rings, consistent with expected radiation-induced reduction. The lengths of the disulfide bonds were refined to 2.5 and 2.3 A (for monomer 2 and 1, respectively), indicating only a slight elongation from a disulfide bond of 2.04 A due to radiation-induced reduction [45,46]. The largely intact disulfide bond can also be seen from the clear electron density between the sulfur atoms with no significant electron difference density. The absorbed X-ray dose was calculated with the program RADDOSE-3D [47].

Steady-state kinetic measurements
To measure the ability of TrxR wt and the TrxR mutant to reduce NrdI, steady-state kinetic measurements were performed. The experimental procedure and data analysis were performed essentially as described previously [23,24]. In brief, the reactions were carried out with 200 lM NADPH, 0.5 lM TrxR wt or TrxR mutant , and various concentrations of NrdI (2.5-20 lM). NADPH and TrxR were incubated and stirred for 15 min, and NrdI was added to start the reaction. The reduction of NrdI was monitored by the disappearance of the NrdI ox state at k max = 447 nm and the appearance of the NrdI sq state at k max = 575 nm. Due to NrdI's high reactivity with dioxygen, the steadystate experiments were carried out in a glove box (Plas Labs 855-AC) under strict anaerobic conditions (91% N 2 , 9% H 2 , AGA) using an Agilent 8453 UV-visible spectrophotometer.

Bioinformatic analysis
Selected bacterial TrxR and TrxR-like FNR protein sequences were obtained from the National Center for Biotechnology Information resources and used for multiple sequence alignments. The selected bacteria were from firmicutes containing both TrxR and TrxR-like FNRs: Bc, Bacillus subtilis (Bs), Sa, and Lactococcus lactis (Ll), in addition to TrxR from Ec, Helicobacter pylori (Hp), and Dr. The two latter organisms have TrxR structures in the PDB that contain either two or no NADPH molecules bound per homodimer, respectively. Multiple sequence alignments were performed with Clustal Omega and Phylogenetic tree calculated with Jalview with average distances using the BLOSUM62 matrix on the sequence alignment in Fig. 2. phylogenetic tree analysis with average distances using the BLOSUM62 matrix. The figures were generated in JalView, and the sequence alignments were colored by % identity [48].

Results and Discussion
TrxR and FNR differences with respect to sequence and structure In an initial attempt to understand which residues are most important in differentiating the TrxR and FNR functions as reductases of Trx and Fld/NrdI, respectively, B. cereus TrxR was mutated and designed to resemble an FNR. As noted above, TrxR undergoes a domain rotation for successive positioning of the CXXC motif and NADPH close to the FAD isoalloxazine ring, while FNR undergoes a rotation to alter between a close positioning between NADPH and FAD, versus exposing the binding site close to the FAD isoalloxazine ring for NrdI/Fld. Hence, the requirement of a rotation is different for TrxR and TrxR-like FNRs. Therefore, the sequences of selected TrxRs and TrxR-like FNRs were compared to investigate particular differences in the regions most likely involved in the rotation. A sequence alignment and phylogenetic tree of selected TrxRs and TrxR-like FNRs can be seen in Figs 2 and 3. The CXXC motif and the common sequence motifs HRRXXXR for the 2 0 phosphate group of NADPH and GXGXXA/G and GXGXXG for the pyrophosphate groups of NAD(P) H and FAD, respectively, are also highlighted [26] Three extra residues in one of the hinge regions are conserved among FNRs and absent in TrxRs. This extended linker is hypothesized to play a role in the rotation of the NADPH-domain in FNRs and could be important for access and reduction of NrdI/Fld. By introducing the linker in TrxR, the ability of TrxR to bypass the disulfide active site and shuffle electrons directly from FAD to the flavin mononucleotide (FMN) cofactor of the NrdI substrate can be examined. The changes introduced to the sequence of B. cereus TrxR mutant , also depicted in Fig. 2, include the following: (a) insertion of an extra linker between the FAD and NADPH-binding domains named 'GAF', consisting of tree amino acids; glycine, alanine, and phenylalanine (conserved in FNRs including B. cereus FNR2 and Bacillus subtilis FNR/YumC); (b) mutation of the redox-active cysteines to serines; and (c) mutation of residue 320 (glutamate to serine), which is shown to form a hydrogen bond with the N5 atom on FAD in FNRs. In the latter, this residue is

Kinetics of NrdI reduction
To investigate TrxR-NrdI as an electron transfer pair, the rates of reduction of NrdI by TrxR wt and TrxR mutant were measured and compared to the published reduction rates of NrdI by B. cereus FNR1, FNR2, and Bdr (FNR3) [24] ( Table 1 and Fig. 4). The results showed that TrxR wt is able to reduce NrdI; however, the steady-state kinetic measurements reveal that NrdI achieves a similar turnover rate with TrxR wt as with the least efficient FNR in B. cereus, namely Bdr, with a k cat of 3.2 min À1 versus k cat of 2.94 min À1 , respectively [24]. Although TrxR is able to reduce NrdI, B. cereus FNR2 has been shown to reduce NrdI at a significantly higher rate,˜30-fold higher than for the TrxR-NrdI electron transfer pair investigated in this study (Fig. 4). Regardless, the weak turnover rate seen for the reduction of NrdI by TrxR may imply that TrxR could possibly function as a reductase of NrdI, and hence* in the activation of class Ib RNR under specific conditions or salvage modes. Further studies are needed to explore and unveil alternative and more efficient NrdI reductases in bacteria encoding the class Ib RNR, however, lacking the TrxR-like FNR2 homolog. Moreover, the mutations introduced in the TrxR mutant in this study show no significant increase in the NrdI reduction rate, with a k cat of 3.9 min À1 versus k cat of 3.2 min À1 for the TrxR mutant and TrxR wt , respectively. Therefore, other aspects associated with the FNR/TrxR structures could be more crucial and should be considered in order to elucidate important features responsible for the substrate specificity in TrxRs and FNRs and for efficient FNR-NrdI reduction.
Although the turnover rate of NrdI reduction presented in this work is low with TrxR, previous studies have confirmed the role of B. cereus TrxR as a reductase of the Trx-like redoxin NrdH (BC3987) from B. cereus (Fig. 1) and verified the role of these redox partners in the activation of class Ib RNR from B. cereus [51]. Bacillus anthracis NrdH, as well as Trx1, have also been proven to serve as redox partners for B. anthracis TrxR (sequence identities with B. cereus NrdH, Trx, and TrxR are 97%, 100%, and 98%, respectively) [51,52].
The overall structure of B. cereus TrxR and presence of NADPH The B. cereus TrxR structure was solved and refined to a resolution of 2.25 A with one TrxR homodimer in the asymmetric unit (Table 2). TrxR has crystallized in the FO conformation, where the disulfide active site is located on the re-face of the flavin cofactor, suitable for electron transfer from FADH 2 to the redox-active disulfide, which will require the rotation of the NADPH-binding domain for reduction of FAD by NADPH and reaction with the Trx substrate in the FR conformation (Fig. 5A,B). As seen in Fig. 5A (inset), a well-defined electron density is observed for the FAD, disulfide bond, and NADPH/NADP + (monomer   NADP + in monomer 1 as seen from the OMIT map in Fig. 6A, while there is no electron density for NADPH/NADP + observed in the binding site in monomer 2 (Fig. 6B). A survey of the published low M r TrxR structures in the Protein Data Bank showed that some TrxRs have crystallized with one monomer in the asymmetric unit and hence, containing either 0 or 1 NADPH/NADP + molecule bound. Others have crystallized with the dimer in the asymmetric unit containing either 0 or 2 NADPH/NADP + bound. The only exception is the structure of TrxR from the firmicute S. aureus, which is the TrxR most homologous to B. cereus TrxR (sequence identity 71%), which only has NADPH/NADP + bound in one of the monomers (Fig. 6E,F). Overlaying the two NADPH-binding sites in monomers 1 and 2 in B. cereus TrxR shows that the loop on one side of the NADPH-binding site containing the HRRXXXR motif (right side in Fig. 6C) is closing in on the NADPH cofactor in monomer 1, while the loop creates a wider opening in monomer 2. This structural change on one side of the NADPH-binding site is also seen in the overlay of the two B. cereus TrxR monomers in Fig. 6D, while no other structural changes are observed. In these two pathogenic firmicutes (B. cereus and S. aureus), only half of the sites contain NADPH/NADP + fully bound, and a structural change is observed. This raises the question if the binding of the NADPH leads to the protein closing in on NADPH in an induced fit manner, or if the binding of NADPH to one site leads to a structural change so that the binding site in the other monomer opens up more and consequently increases the dissociation constant for NADPH binding to that site. A comparison to the NADPH-binding sites in other low M r TrxRs with and without NADPH bound shows that these structures have the same conformation of the NADPH-binding site as B. cereus TrxR with NADPH bound, in respect to the beforementioned loop, regardless if these structures have NADPH bound or not (Fig. 7). This may point toward an opening of the loop in monomer 2 upon binding of NADPH in monomer 1, although it cannot be ruled out that the asymmetry of the two sites already exists in the NADPH-apo form [53]. Together, these observations may possibly imply a so-called half-of-the-sites reactivity, which could be a case of extreme negative cooperativity [53,54]. It can, however, not be ruled out that the pure half-of-the-site reactivity is a crystallographically induced effect of a possible weak negative cooperativity in solution or a preexisting asymmetry, thereby not exhibiting cooperativity [53,54]. Nevertheless, half-site reactivity or subunit asymmetry is observed in several other pyridine nucleotide-disulfide oxidoreductases [55]. The two conformational states of TrxR, FO and FR, are interchanged by the˜67°rotation and could also lead to alternating NADPH binding and catalysis within the TrxR dimer, functioning in a sequential mode of action. These structural observations must be further confirmed by binding studies and kinetic experiments to prove whether this TrxR has an asymmetric and sequential reactivity where NADPH binds to one of the monomers at a time.

Conclusions
We have, by using B. cereus as a model system encoding TrxR and three TrxR-like FNRs, shown that TrxR is able to reduce NrdI, a flavoprotein crucial for the activation of class Ib RNR, however, at a very low turnover rate as compared to the suggested endogenous NrdI reductase in B. cereus. Therefore, TrxR does not seem to be a probable endogenous redox partner for NrdI, indicating that a search for more efficient redox partners for NrdI in bacteria lacking  This could indicate that certain pathogenic firmicutes have adopted to an asymmetric mode of enzyme activation and successive substrate reduction by TrxR.