Biochemical properties and crystal structure of the flavin reductase FerA from Paracoccus denitrificans

Microbiological Research 188 (2016) 9–22

Contents lists available at ScienceDirect

Microbiological Research journal homepage: www.elsevier.com/locate/micres

Biochemical properties and crystal structure of the flavin reductase FerA from Paracoccus denitrificans Vojtˇech Sedláˇcek a , Tomáˇs Klumpler b , Jaromír Marek b , Igor Kuˇcera a,∗ a b

Department of Biochemistry, Faculty of Science, Masaryk University, Brno, Czech Republic X-ray Diffraction and Bio-SAXS Core Facility, Central European Institute of Technology, Masaryk University, Brno, Czech Republic

a r t i c l e

i n f o

Article history: Received 11 January 2016 Received in revised form 15 April 2016 Accepted 20 April 2016 Available online 27 April 2016 Keywords: FerA FMN FAD Flavin reductase family Oxidative stress Paracoccus denitrificans

a b s t r a c t The Pden 2689 gene encoding FerA, an NADH:flavin oxidoreductase required for growth of Paracoccus denitrificans under iron limitation, was cloned and overexpressed as a C-terminally His6-tagged derivative. The binding of substrates and products was detected and quantified by isothermal titration calorimetry and fluorometric titration. FerA binds FMN and FAD with comparable affinity in an enthalpically driven, entropically opposed process. The reduced flavin is bound more loosely than the oxidized one, which was confirmed by a negative shift in the redox potential of FMN after addition of FerA. Initial velocity and substrate analogs inhibition studies showed that FerA follows a random-ordered sequence of substrate (NADH and FMN) binding. The primary kinetic isotope effects from stereospecifically deuterated nicotinamide nucleotides demonstrated that hydride transfer occurs from the pro-S position and contributes to rate limitation for the overall reaction. The crystal structure of FerA revealed a twisted seven-stranded antiparallel ␤-barrel similar to that of other short chain flavin reductases. Only minor structural changes around Arg106 took place upon FMN binding. The solution structure FerA derived from small angle X-ray scattering (SAXS) matched the dimer assembly predicted from the crystal structure. Site-directed mutagenesis pinpointed a role of Arg106 and His146 in binding of flavin and NADH, respectively. Pull down experiments performed with cytoplasmic extracts resulted in a negative outcome indicating that FerA might physiologically act without association with other proteins. Rapid kinetics experiments provided evidence for a stabilizing effect of another P. denitrificans protein, the NAD(P)H:acceptor oxidoreducase FerB, against spontaneous oxidation of the FerA-produced dihydroflavin. © 2016 Elsevier GmbH. All rights reserved.

1. Introduction The common mechanisms of iron assimilation in microorganisms include the reduction of ferric (Fe(III)) to ferrous (Fe(II)) ion (Miethke, 2013). We have previously demonstrated that the cytoplasm of the soil bacterium Paracoccus denitrificans constitutively contains two enzymes capable of reducing a number of Fe(III) complexes by NADH which we have named ferric reductase A and B (FerA and FerB) (Mazoch et al., 2004). Based on experiments in vitro and in vivo using a FerA-deficient mutant strain, a dual role for

Abbreviations: FAD, flavin adenine dinucleotide; FerA, ferric reductase A from Paracoccus denitrificans; FerB, ferric reductase B from Paracoccus denitrificans; FMN, flavin mononucleotide; ITC, isothermal titration calorimetry; KIE, kinetic isotope effect; SOD, superoxide dismutase. ∗ Corresponding author at: Department of Biochemistry, Faculty of Science, Masaryk University, Kotláˇrská 2, 61137 Brno, Czech Republic. E-mail address: [email protected] (I. Kuˇcera). http://dx.doi.org/10.1016/j.micres.2016.04.006 0944-5013/© 2016 Elsevier GmbH. All rights reserved.

FerA was suggested with the idea that it both participates in reductive release of siderophore-bound Fe(III) and promotes siderophore biosynthesis (Sedlacek et al., 2009). The biological function of FerB remains unclear, although recent work has indicated its involvement in scavenging excessive superoxide anion radicals (O2 • - ) (Sedlacek et al., 2015). Substrate analysis of the purified FerA enzyme showed it to act primarily as a flavin reductase and the enzymatically formed dihydroflavin was validated as the actual reducing agent for Fe(III) (Mazoch et al., 2004). Flavin reductases (FRs) catalyze the reduction of oxidized flavin (FMN, FAD or riboflavin) by NADH or NADPH. They function either alone or in conjunction with other proteins, most notably with flavin-dependent monooxygenases (Huijbers et al., 2014). The designations FRD, FRP and FRG, respectively, are used for the NADH-preferring FRs, the NADPH-preferring FRs and the general FRs, which utilize NADH and NADPH with similar efficiencies. Furthermore, FRs are assigned to class I (flavoproteins) or II (non-flavoproteins) depending on whether they contain or

10

V. Sedláˇcek et al. / Microbiological Research 188 (2016) 9–22

lack a permanently bound flavin cofactor (Tu, 2001). The class I FRs typically follow a ping-pong mechanism in which reduction of the cofactor by NAD(P)H and release of NAD(P)+ precedes electron transfer from the reduced cofactor to the flavin substrate. The members of class II bind both NAD(P)H and flavin forming a ternary complex which then decomposes to release the products, NAD(P)+ and reduced flavin (for specific examples of enzymes, see Ref. (Ellis, 2010)). In the present study, we have conducted a detailed ligand binding, kinetic and structural characterization of FerA to obtain molecular-level insight into the catalytic mechanism employed by this enzyme. Furthermore, by studying the FerA-catalyzed reaction in the presence of FerB, we uncovered a novel possibility of stabilizing the produced dihydroflavin towards spontaneous oxidation by O2 . 2. Materials and methods 2.1. Gene cloning, protein expression and purification The coding sequence of FerA was amplified by PCR from the genomic DNA of P. denitrificans 1222. The primers were FerAU1 (5 -ACAGGGCGCGCATATGAGCCGTCTT-3 ) and FerAL3 (5 ATGCGCGGCTAGCCCCTCGAGGCCGGCATC-3 ) with restrictase sites for NdeI and XhoI (in italics). The amplified gene was cloned into the expression plasmid pET21a (Novagen) and subsequently transformed into the One Shot BL21(DE3)pLysS Escherichia coli strain (Invitrogen). Overnight culture in rich LB (Luria–Bertani) medium inoculated from a single colony of the transformants and verified by PCR using the primers for ferA gene and T7 promoter, were diluted 1:25 in an M9 medium containing 0.4% (w/v) glucose, 100 ␮g ml−1 ampicillin and 37 ␮g ml−1 chloramphenicol and grown to OD600 of 0.3–0.4 at 30 ◦ C. The expression of a C-terminal hexahistidine tagged recombinant protein was induced by addition of 1 mM isopropyl ␤-d-thiogalactoside (Duchefa). The overnight grown cells were collected by centrifugation and FerA were purified as described previously (Tesarik et al., 2009). 2.2. Molecular mass determination One-dimensional SDS gel electrophoresis of FerA was carried out in 12% polyacrylamide gel according to (Laemmli, 1970) using a Mini Protean III (Bio-Rad) vertical electrophoretic system. The mass spectra of FerA were recorded on a Bruker Reflex IV mass spectrometer in a MALDI-TOF mode. Prior analysis, the purified protein was desalted to a final 50 mM sodium phosphate buffer (pH 8.0) by PD10 column (GE Healthcare). The desalted sample was mixed with a matrix solution in a ratio 1:2 and dried. The spectra were taken in positive-ion linear mode. 2.3. Determination of protein concentration Protein concentration was determined by QuantiPro BCA Assay Kit (Sigma-Aldrich) using bovine serum albumin as the standard. Proteins eluted from chromatographic columns during purification were also monitored at 280 nm. 2.4. Isothermal titration calorimetry Binding of FMN or FAD by FerA was assessed via ITC measurements using the Auto-iTC 200 isothermal titration calorimeter (Malvern). Experiments were carried out in a buffer containing 50 mM sodium phosphate, 300 mM NaCl and 250 mM imidazole (pH 8.0). During the titration, the reaction mixture was continuously stirred at 750 rpm and at 25 ◦ C. For each step of the titration, the binding heat was determined as the difference between the heat

change generated after the injection of the flavins into the protein solution and the corresponding background, which was obtained by injection of the flavins into the sample cell with the buffer. Thermodynamic parameters for the binding were determined from ITC results using Microcal Origin version 8.0 (OriginLab) with the “ITC custom” add-on installed. 2.5. Fluorescence measurement of dissociation constants The binding of FMN, FMNH2 , FAD, FADH2 , NAD+ , NADH, AMP, riboflavin and lumichrome by FerA was assessed by protein fluorescence quenching measurements carried out at 30 ◦ C using Luminiscence Spectrometer LS-50B (Perkin-Elmer). A fixed amount of FerA was titrated with aliquots of ligands mentioned above of a known concentration in 50 mM phosphate buffer (pH 7.0) containing 10 mM EDTA. Excitation was at 280 nm, and the emission was recorded at 330 nm. Fluorescence intensities were read 3 min after each addition of quencher. Where appropriate, reduction of flavins was achieved by sodium dithionite and their reduced forms were kept under nitrogen atmosphere for subsequent titration which was also performed under strict anaerobic conditions. Titrations by FerA of FMN and FAD as fluorophores were done in a similar manner but at the excitation and emission wavelength of 450 and 517 nm, respectively. The fluorescence anisotropy of FAD was measured in a FluoroMax-4 spectrofluorometer (Horiba Scientific) equipped with polarizers operated by FluorEssence software version 3.5 and calculated from the intensities at four mutually perpendicular polarizer settings (IVV , IVH , IHV , and IHH ) as previously described (Lakowicz, 2006). The dissociation constants Kd were obtained by nonlinear regression analysis according to a 1:1 binding model (Lostao et al., 2000). 2.6. Determination of the standard redox potential The standard redox potentials of FMN and FMN with FerA were determined at 25 ◦ C according to the method of Massey (Massey, 1990). The reaction mixture contained in 1-ml cuvette 50 ␮M FMN, 20 ␮M phenosafranine, 0.4 mM xanthine, 5 ␮M benzyl viologen in a 50 mM sodium phosphate buffer with 0.1 mM EDTA (pH 7.0). After flushing by argon (99.9999%, v/v), a catalytic amount of a milk xanthine oxidase (Sigma-Aldrich) suspension was added to a final concentration 0.013 mg ml−1 still under strict anaerobic conditions. The concentration of FerA was 0.2 mM. Absorbance of the dye and FMN was recorded by UV–vis spectrophotometer UltraSpec 2000 (GE Healthcare). The absorbance readings at 450 nm (FMN) and 521 nm (phenosafranine) were corrected for the contribution of the other component. 2.7. Enzyme kinetics The bi-substrate kinetic analysis was performed by an UltraSpec 2000 spectrophotometer with a temperature-controlled cuvette holder (30 ◦ C) in 25 mM Tris–HCl (pH 7.4). NADH oxidation was measured at 340 nm; initial velocities were calculated from the linear slope of the progress curves obtained, using a molar absorption coefficient of 6.22 mM−1 cm−1 . The kinetic parameters were calculated by the Marquardt-Levenberg nonlinear fit algorithm included in the Microcal Origin software. Data conforming to a sequential mechanism were fitted to Eq. (1). Data for competitive and mixed inhibition were fitted to Eqs. (2) and (3), respectively.

v = V [A][B]/{K a [B] + K b [A] + [A][B] + K ia K b }

(1)

v = V [A]/{K a (1 + [I]/K is ) + [A]}

(2)

v = V [A]/{K a (1 + [I]/K is ) + [A](1 + [I]/K ii )}

(3)

V. Sedláˇcek et al. / Microbiological Research 188 (2016) 9–22

in Eqs. (1)–(3), v and V represent initial and maximum velocities, Ka and Kb are Km values for the substrates A and B, Kia is the dissociation constant for the binary complex EA, Kis and Kii are slope and intercept inhibition constants, and A, B, and I represent reactant and inhibitor concentrations. The catalytic rate constant (kcat ) was calculated as the ratio of V to the enzyme concentration used. 2.8. Preparation of stereospecifically C-4-deuterated NADH samples [4R-2 H]NADH and [4S-2 H]NADH were produced from NAD+ using the NAD+ -linked enzymes horse liver alcohol dehydrogenase (pro-R/A side-specific) and Leuconostoc mesenteroides glucose-6phosphate dehydrogenase (pro-S/B side-specific) essentially as before (Sedlacek et al., 2014). 2.9. Crystallization, X-ray data collection and structure determination Preliminary screening of crystallization conditions for FerA was carried out in sitting drops (100 nl protein solution mixed with 100 nl reservoir solution equilibrated against 100 ␮l reservoir solution) at 293 K in 96-well Innovaplate SD-2 plates using an automated nanoliter liquid-handling system (Mosquito, TTP Labtech) and commercially available screening sets. Promising microcrystals were obtained after a few days from condition A5 [0.1 M HEPES (pH 7.4), 10% (v/v) 2-propanol, 20% (w/v) PEG 4000] of The Classics Suite (Qiagen). Optimization of the initial conditions with an aim to increase the size of the crystals and to improve their diffraction quality was performed in EasyXtal 15-well plates (Qiagen) by the hanging-drop vapor-diffusion method, in which drops containing 1 ␮l of protein solution mixed with 1 ␮l of reservoir solution were equilibrated against 1 ml reservoir solution at 293 K. Using optimized conditions consisted of 0.1 M HEPES (pH 7.4), 10% (v/v) 2-propanol, 10% (v/v) PEG 400, 17–20% (w/v) PEG 4000, hexagonal prism-like crystals with dimensions a ∼150 ␮m and h ∼ 30 ␮m were grown in 7–10 days. Co-crystals of FerA and flavin mononucleotide (FMN) were prepared by addition of 200 nl of 0.18 M FMN water solution into drops containing FerA crystals prepared as described above. Both, FerA and FerA-FMN diffraction data were collected at the beamline X13 of the DORIS-III storage ring at EMBL/DESY (Hamburg, Germany). Prior to the diffraction experiment, the crystals were soaked in the cryoprotectant Paratone-N (Molecular Dimensions, Newmarket, UK) and then flash-cooled to 100 K in a nitrogen stream. All diffraction data were processed and merged using the XDS system (Kabsch, 2010). The structure of FerA was determined by the molecular replacement method with Phaser (Mccoy et al., 2007) using 1RZ0 (van den Heuvel et al., 2004) as a search model. The quality of the electron density maps allowed us the successful application of the auto-build regime of ARP/wARP (Cohen et al., 2008). The structure of FerA in complex with FMN was determined by molecular replacement using native FerA as a search model. Refinement of models of FerA and FerA in complex with FMN was performed using restrained refinement with the maximum likelihood method of REFMAC5 (Murshudov et al., 2011). Manual fine-tuning of both structures was performed with Coot (Emsley et al., 2010). Resulting models of FerA and FerA in complex with FMN were re-refined against 1.53 Å (resp. 1.8 Å) diffraction data and rebuilt using PDB-REDO algorithm (Joosten et al., 2014). Final models were deposited in the Protein Data Bank under accession codes 4XHY (FerA) and 4XJ2 (FerA in a complex with FMN). Data collection and refinement statistics for both structures are summarized in Table 1. Molecular graphics images were produced using UCSF CHIMERA v1.10.1 (Pettersen et al., 2004). Energy minimization of FerA in complex with FMN and NAD+ was performed with Amber14, using the ff14SB parameter set for the protein (Case et al.,

11

Table 1 Data collection and refinement statistics. Parameter

FerA

FerA with FMN

PDB code Data collection Wavelength (Å) Space group [No] Cell dimensions a; c (Å) Resolution range (Å) [last resolution shell] Rmerge (all/observed) (%) [last resolution shell] I/␴ (I) (all/observed) [last resolution shell] Completeness (all/observed) (%) [last resolution shell] Number of unique reflections (all/observed)

4XHY

4XJ2

0.8123 P31 21 [152] 43.91; 154.66 1.53–51.55 [1.53 −1.57] 3.1 (2.9) [23.7 (10.3)] 45.59 (50.75) [8.44 (12.34)] 99.7 (90.6) [99.9 (73.1)] 27064/24577

0.8123 P31 21 [152] 43.92; 154.48 1.8–15 [1.8–1.85] 4.1 (3.5) [37.4 (10.5)] 21.43 (28.73) [3.07 (8.08)] 98.5 (74.0) [98.5 (35.5)] 16627/12465

1.53–51.55 25671/1393

1.8–15.0 15793/834

0.158/0.205

0.166/0.202

1193

1201 31 208

Refinement Resolution (Å) Number of reflections (refined/for Rfree ) Rwork /Rfree Number of atoms Protein Ligand/ion Water B factors (Å) Protein Ligand Water Root-mean-square deviation Bond lengths (Å) Bond angles (◦ )

257 25.859 25.157

43.387 56.797 54.446

0.019 1.987

0.020 2.062

2014). The parameter set used for the cofactors FMN (Schneider and Suhnel, 1999) and NAD+ (Walker et al., 2002) were taken from the literature. All atom energy minimization were performed using 5000 steps steepest descent and 5000 steps of the conjugate gradient method in vacuum.

2.10. Small angle X-ray scattering (SAXS) The SAXS data were collected using the BioSAXS-1000, Rigaku at CEITEC (Brno, Czech Republic). Data was collected at 290 K with focused (confocal Max-Flux SAXS optic, Rigaku) Cu K␣ X-ray (1.54 Å). Sample to detector (PILATUS 100 K, Dectris) distance was 0.4 m covering a scattering vector (q = 4␲sin(␪)/␭) range from 0.009 to 0.65 Å−1 . The buffer was identical to the buffer of the last step of the protein purification. The FerA sample was measured at three concentrations: ∼6; 2; and 0.7 mg ml−1 . Scattering data from lowest concentration sample contained unacceptable level of noise and was omitted from further analysis. For buffer and sample one twodimensional image was collected with an exposure time of 60 min per image. Radial averaging, data reduction and the buffer subtraction were performed using SAXSLab3.0.0r1, Rigaku. Subtracted scattering data were truncated to 0.3 Å−1 limit. Integral structural parameters were determined using PRIMUS/qt (r3709) from ATSAS v.2.6.1 (Konarev et al., 2003;Petoukhov et al., 2012). Evaluation of the solution scattering of the atomic models and the fitting to experimental data was performed by CRYSOL v2.8.3 (Svergun et al., 1995) using atomic structure of most probable dimeric assembly proposed by PDBePISA (Xu et al., 2008). Data collection and scattering-derived parameters are summarized in Table 2. Ab initio model (for 6 mg ml−1 data) was reconstructed using DAMMIN 5.3 (Svergun, 1999), where the mode was set to ‘slow’, while all other parameters were kept default.

12

V. Sedláˇcek et al. / Microbiological Research 188 (2016) 9–22

Table 2 Data collection from SAXS.

3. Results 3.1. The recombinant FerA

Data-collection parameters Instrument Wavelength (Å) q range (Å−1 ) Exposure time (min) Temperature (K) Concentration of FerA (mg·ml−1 ) Structural parameters I(0) (A.U.) [from Guinier] Rg (Å) [from Guinier] I(0) (A.U.) [from P(r)] Rg (Å) [from P(r)] Dmax (Å) Vporod volume estimate (Å3 ) Software employed Primary data reduction Data processing

BioSAXS-1000 1.5418 0.009–0.3 60 290 6.0

2.0

0.05 22.97 0.05 21.67 61.94 57435

0.05 21.39 0.05 21.23 61.33 59027

SAXSLab3.0.0r1 PRIMUS/qt (r3709)

2.11. Site-directed mutagenesis Site-directed mutagenesis was carried as specified in the QuickChange II Site-Directed Mutagenesis Kit (Stratagene) with a minor modification of replacing XL1-supercompetent cells by NEB 5-alpha High Efficiency Competent E. coli (New England Biolabs). The pET21 plasmid with the ferA gene served as a PCR template. The primers are listed in Supplementary Table S1. Expression and purification of the recombinant protein mutants were performed as before (Sedlacek et al., 2014).

2.12. In vitro interaction pull-down assay His pull-down assay was performed with a ProFound Pull-Down PolyHis Protein:Protein Interaction Kit (Thermo Fisher Scientific) according to the manufacturer’s instructions. FerA with a Cterminal hexahistidine extension was produced in E. coli BL21 cells, and used as a bait protein bound to the cobalt-chelate matrix. A fresh lysate of the aerobically grown P. denitrificans cells was incubated for 60 min at 4 ◦ C with the FerA bait together with or without FMN equimolar to FerA, the resin was then washed five times with 10 mM imidazole and eluted at 300 mM imidazole. The eluted proteins were separated by SDS-PAGE and visualized with Coomassie blue or silver staining. The possibility of false positives caused by nonspecific binding of proteins from the P. denitrificans lysate was checked by using a nontreated resin lacking the bait protein.

2.13. Stopped-flow experiments Rapid kinetics measurements were performed on a SFM-3000 (Bio-Logic, Claix, France) stopped-flow system equipped with a 0.8× 0.8-mm cuvette (FC-08). The temperature was kept constant at 10 ◦ C. Six measurements were averaged for each sample. A solution in the syringe 1 contained 40 ␮M FMN and 200 ␮M NADH in 25 mM Tris–HCl buffer (pH 7.4) was mixed with a solution in the syringe 2 containing the same buffer with 40 ␮M FMN, 1.37 ␮M FerA and various concentrations superoxide dismutase from bovine erythrocytes (Sigma-Aldrich, catalog number S7571) in a ratio 1:1 to a final volume of 0.15 ml. The reaction was monitored at 450 nm. A flow speed of 7 ml s−1 resulted in a dead time of 0.4 ms. Data collected from the experiments were processed using the Bio-Kine software (Bio-Logic).

The gene Pden 2689 nucleotide sequence coding for FerA was used to produce a C-terminally hexahistidine-tagged fusion protein in E. coli BL21(DE3)pLysS, which was then purified by His-tag affinity chromatography (see Section 2.1). The final preparation yielded a single band of 20 kDa by polyacrylamide gel electrophoresis under denaturing conditions and a major sharp peak at 19936.4 ± 1.9 Da in MALDI-TOF MS analysis. Both of these quantities are in reasonable agreement with the predicted value of 20080.8 Da. The UV–VIS spectrum of the purified protein lacked a distinct peak centered at about 450 nm that would be indicative of a bound oxidized flavin. The turnover rate for NADH oxidation initiated by the addition of FMN (∼89 s−1 ) was comparable to that for the native, untagged FerA, indicating that the His tag does not interfere with catalysis. Consequently, all experiments reported below were conducted directly with the His-tagged derivative. 3.2. Binding studies Several complementary approaches were used to probe the interaction of FerA with its substrates. Isothermal titration calorimetry (ITC) has the advantage of allowing simultaneous determination of the enthalpy H◦ , association constant Ka and stoichiometry N from a single titration. Fig. 1 shows representative calorimetric titrations of FMN (A) or FAD (B) into the solution of FerA at 25 ◦ C. Each peak in the binding isotherm (upper panels) corresponds to a single injection of flavin. The negative deviations of the signal from the baseline on addition of flavin indicate that heat was released (an exothermic process, H < 0). The heat evolved per injection of ligand was plotted versus the flavin/FerA molar ratio (lower panels), and the best parameter estimates for the binding model were obtained by non-linear regression. The results demonstrate that FerA binds FMN and FAD with similar affinity, with the dissociation constants lying in the micromolar range (Kd FMN = 3.7 ± 0.1 ␮M; Kd FAD = 6.6 ± 0.2 ␮M). For binding of flavins to FerA, the negative value of enthalpy change (−H◦ ) outweighed the negative Gibbs energy change (−G◦ ). Because G◦ = H◦ − TS◦ , it follows that the entropy term is also negative resulting in an unfavorable contribution to the Gibbs energy. The binding of FAD was associated with a greater entropic loss compared to FMN possibly due to a more extensive reduction in the degrees of freedom of the former compound. The dominating role of exothermic enthalpy suggests the formation of a number of hydrogen bonds in the flavin-FerA complexes. Insights from the crystal structure of the FMN-FerA complex (see below) support this notion. In ITC measurements with reduced forms of flavins (FMNH2 and FADH2 ) a difficulty may arise because extra heat evolves during their auto-oxidation if oxygen is not fully purged from the instrument. Instead of ITC, a simpler alternative turned out to be to measure the changes in the intrinsic fluorescence of FerA upon flavin binding. With an excitation wavelength of 280 nm, the emission spectrum of FerA had a peak at 330 nm. Titration with a flavin produced a reduction in fluorescence emission, which could be fitted to the single binding site equilibrium model (Fig. 2). The dissociation constants obtained from these experiments (Kd FMN = 3.0 ± 0.3 ␮M, Kd FMNH2 = 12.2 ± 1.0 ␮M, Kd FAD = 8.9 ± 0.5 ␮M, Kd FADH2 = 31.1 ± 2.0 ␮M) indicate that the oxidized flavins bind to FerA more tightly than the reduced ones. Additional data in support of this conclusion came from redox titration experiments using the xanthine/xanthine oxidase system as the electron source and phenosafranine as the internal redox indicator. From the results presented in Fig. 3 it follows that the midpoint potential of FMN underwent a negative shift of 19 mV in

V. Sedláˇcek et al. / Microbiological Research 188 (2016) 9–22

13

Fig. 1. Isothermal calorimetry. ITC measurements of binding of FMN (panel A and B) and FAD (panel C and D) by FerA in 50 mM sodium phosphate containing 300 mM NaCl and 0.25 mM imidazole (pH 8.0) at 25 ◦ C. Panels A and C show the raw data, generated by titration of 0.1 mM FerA by 20 × 1 ␮l injections of 2 mM FMN or 1.6 mM FAD. The area under each peak was integrated and plotted against the molar ratio of FMN or FAD to the protein in panels B or D. The inset graph in panels B and D show changes in the standard Gibbs energy, enthalpy and entropy for FMN or FAD binding to FerA.

Fig. 2. Quenching of FerA fluorescence by flavins. Concentrated stock solutions of FMN (solid squares), FMNH2 (open squares), FAD (solid circles), and FADH2 (open circles) were gradually added to 914 nM FerA in 50 mM sodium phosphate buffer (pH 7.4) with 10 mM EDTA at 30 ◦ C. After each addition, the sample was allowed to equilibrate for 3 min prior to reading the fluorescence at 280 nm/330 nm. The inset shows excitation and emission spectra of FerA.

the presence of a surplus of FerA. This can be readily explained by the preferential complex formation of FerA with the oxidized flavin. By applying an equation relating the ratio of dissociation constants (Kd FMN /Kd FMNH2 ) to the alteration in midpoint potential (Em ) KdFMN /KdFMNH2 = 10nFEm /2.3RT (see, for example, (Lambeth and Kamin, 1976)), Kd FMN /Kd FMNH2 comes out as 0.23 which is quite comparable with the ratio of values estimated by fluorometric titration (3.0/12.2 = 0.25). Since riboflavin (dephosphorylated FMN) and lumichrome (an FMN analogue completely lacking the ribityl side chain) still showed significant binding (Kd riboflavin = 11.7 ± 0.3 ␮M, Kd lumichrome = 53 ± 2 ␮M) it became evident that the isoalloxazine ring itself is essential for the interaction of flavins with FerA. The fluorometric method also proved to be useful in evaluating the binding affinity of FerA to NADH and to NAD+ , its oxidation product. We found that the complexes between FerA and NADH or NAD+ can

Fig. 3. Redox titration of FMN and its complex with FerA. 50 ␮M FMN without (solid squares) or with 0.2 mM FerA (solid circles) was incubated with 20 ␮M phenosafranine (Em 0 = –252 mV) as a redox indicator under strict anaerobic condition in the xanthine/xanthine oxidase system. The slopes 1.21 for FMN and 1.09 for FMN with FerA are close to the theoretical value of 1 as expected for two electron transfer between the dye and flavin. The standard redox potential of FMN was calculated from the y-axis intercept.

form, but are less stable than the complexes of FerA with flavins (Kd NADH = 68 ± 2 ␮M; Kd NAD = 182 ± 5 ␮M). An even weaker binding was observed with AMP, which represents a part of the NADH molecule (Kd AMP = 31 ± 1 mM). We also investigated whether the fluorescence signal of free flavins at 517 nm (excited at 450 nm) would change upon binding to FerA. As depicted in Fig. 4, fluorescence of FMN was quenched by increasing concentrations of FerA, showing a typical saturation behavior. Fitting of the titration curve to the 1:1 binding model gave rise to a dissociation constant of 3.9 ± 0.1 ␮M. When FAD was titrated with FerA, the fluorescence intensity decreased monotonically without a clear indication of saturation, but the fluorescence anisotropy increased in a saturable manner from 0.05 to 0.32, in line with that is expected for binding of a small-size fluorophore (MW = 786) to the larger protein (MW = 40 160). The dissociation constant of FAD was calculated from the fluorescence anisotropy data and found to be 11.7 ± 0.4 ␮M. Overall, the

14

V. Sedláˇcek et al. / Microbiological Research 188 (2016) 9–22

Fig. 4. Effect of FerA on the fluorescence of flavins. Fluorescence during titration of 343 nM FMN (solid squares) or 540 nM FAD (solid circles) with FerA was measured at 450 nm/517 nm in 50 mM sodium phosphate buffer (pH 7.4). Open circles denote fluorescence anisotropy of FAD under the same conditions.

dissociation constants derived from flavin fluorescence measurements are close to those obtained with the other two approaches reported above. 3.3. Kinetic analysis To elucidate the kinetic mechanism, a two-substrate steadystate kinetic analysis was performed (Leskovac, 2003). Initial reaction velocities for NADH oxidation were measured with varying concentrations of NADH at several fixed concentrations of FMN and vice versa. In both cases, the plots of the inverse initial velocity as a function of the inverse of the variable substrate concentration gave rise to a set of lines intersecting

to the left of the ordinate and on the abscissa axis (Fig. 5). This kinetic pattern ruled out ping–pong (parallel lines), steadystate random (non-linear reciprocal plots), and rapid-equilibrium ordered (lines intersected on the ordinate for the second substrate that binds) mechanisms, leaving equilibrium random and steady-state compulsory ordered mechanisms as possibilities. We distinguished between them through the use of lumichrome and AMP as dead-end inhibitors that structurally mimic the FMN and NADH substrates. The inhibition by lumichrome was competitive vs. FMN (Ki = 49 ± 4 ␮M) and mixed noncompetitive vs. NADH (Kis = 110 ± 10 ␮M, Kii = 290 ± 20 ␮M), whereas AMP inhibited mixed noncompetitively (Kis = 29 ± 5 mM, Kii = 270 ± 50 mM) vs. FMN and competitively vs. NADH (Ki = 35 ± 2 mM). These findings suggest a random-ordered sequence in which either FMN or NADH may be the first substrate. If a steady-state compulsory ordered mechanism were operative, then a structural analogue of the second-binding substrate should have been an uncompetitive-type inhibitor with respect to the first-binding substrate. Accordingly, the original data from bisubstrate kinetics were fitted to the equation for a sequential mechanism (Eq. (1)), yielding the following values for the kinetic parameters: the kcat of 89 ± 9 s−1 , the Km for FMN and NADH (Ka , Kb ) of 3.8 ± 0.5 ␮M and 60 ± 10 ␮M, and the dissociation constants (Ki ) of FMN and NADH of 2.7 ± 0.7 ␮M and 59 ± 9 ␮M. From the fact that the dissociation constants do not differ much from the Km values it follows that there is no significant mutual influence in substrate binding. This lack of interaction between substrates is also demonstrated graphically in the primary plots (Fig. 5) through the position of the lines´ı intersection points on the horizontal axes. Bisubstrate kinetics with FAD and NADH followed a similar kinetic pattern with kcat = 76 ± 3 s−1 ; Ka = 6 ± 1 ␮M, Kia = 20 ± 4 ␮M for FAD and

Fig. 5. Initial-velocity kinetics of the NADH:FMN oxidoreductase reaction. Measurements were performed at 30 ◦ C with 137 nM enzyme in 25 mM Tris–HCl buffer (pH 7.4). Reciprocal initial velocity is plotted against the reciprocals of NADH (panel A, C, E) or FMN (panel B, D, F) concentration at a series of fixed concentrations of FMN equal to 40 (squares), 16 (circles), 4 (triangles) and 1.6 (rhombs) ␮M (panel A) or NADH equal to 250 (squares), 125 (circles), 40 (triangles) and 20 (rhombs) ␮M (panel B). Inhibition study with lumichrome or AMP was arranged analogically with a series of fixed concentrations of lumichrome equal to 0 (squares), 40 (circles), 120 (triangles) and 200 (rhombs) ␮M of lumichrome at 0.05 mM FMN (panel C) or 0.2 mM NADH (panel D), or of AMP equal to 0 (squares), 15 (circles), 100 (triangles) and 200 (rhombs) mM at 0.05 mM FMN (panel E) or 0.2 mM NADH (panel F).

V. Sedláˇcek et al. / Microbiological Research 188 (2016) 9–22

Fig. 6. Overall structure of FerA from P. denitrificans. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) The two protein subunits are shown in blue and green ribbon representation. The secondary structural elements (three ␣-helices, ten ␤-sheets) are designated ␣ and ␤ and numbered as they appear along the polypeptide chain. The image was generated with Chimera 1.10.1 [22] on the basis of the FerA crystal structure (4XHY).

Kb = 35 ± 6 ␮M, Kib = 70 ± 20 ␮M for NADH. In this case a slight negative cooperativity between substrates was apparent (Ka /Kia = 0.3; Kb /Kib = 0.5) and the common point of intersection was situated below the abscissa in the primary plots (results not shown). Additional useful information with respect to the mechanism of the FerA-catalyzed reaction could be acquired by determining the primary deuterium kinetic isotope effects (KIEs). Both [4S-2 H] and [4R-2 H]NADH were synthesized and utilized in place of the corresponding [1 H]NADH substrate. The assays were performed by varying the concentration of [1 H]NADH, [4S-2 H]NADH or [4R2 H]NADH at a saturating concentration of FMN (50 ␮M), as well as by varying the concentration of FMN at a saturating concentration of [1 H]NADH, [4S-2 H]NADH or [4R-2 H]NADH (150 ␮M). The kinetic parameters V and Km of each data set were obtained from fitting the data to the Michaelis–Menten equation and used to calculate D V = V /V and D K = (K ) /(K ) for both substrates. A substanm m H m D H D tial KIE on V was observed when [4S-2 H]NADH was employed (D V = 3.74 ± 0.02), whereas little or no KIE was seen with [4R2 H]NADH (D V = 1.2 ± 0.1). The alteration in V by the 4S but not by the 4R analogue indicates that the pro-S hydride of NADH is transferred to the flavin and that this transfer step is significantly rate-determining in catalysis. There was no KIE (within the experimental error) from [4S-2 H]NADH on the Michaelis constants (D Km NADH = 1.1 ± 0.1; D K FMN = 1.0 ± 0.1). Equal KIE on the two V/K m m values is consistent with the random mechanism (Leskovac, 2003). 3.4. Crystal and solution structure of FerA The amino acid sequence of FerA contains two consensus motifs, S/T/CXXPP and GDH, that are characteristic of the HpaC-like subfamily of the non-flavoprotein (class II) flavin reductases (Galan et al., 2000). A search against the Pfam database (Finn et al., 2010) placed FerA into the protein family of flavin reductases (PF01613) within the FMN-binding split barrel superfamily (CL0336). Experimental results from X-ray crystallography (Fig. 6) are in agreement with the prediction. The crystal structure of FerA (PDB ID: 4XHY) was determined by the molecular replacement method. The crystallographic and model statistics for the structures are summarized in Table 1. Crystals belong to space group P31 21 with one protein molecule in the asymmetric unit. The overall structure is predicted to be a dimer of two identical subunits (see also below). Each consists of three ␣-helices (␣1–␣3) and 10 ␤-strands (␤1–␤10),

15

Fig. 7. SAXS scattering curves. X-ray scattering of FerA (open circles) was compared with calculated scattering of the monomer (dotted line) and dimer (solid line) computed by CRYSOL with good overall fit (2 = 1.87) for concentration 6.0 mg ml−1 .

connected in the order ␤1-␣1-␤2-␤3-␤4-␤5-␣2-␤6-␣3-␤7-␤8␤9-␤10. The core of each subunit contains a twisted ␤-barrel made of seven antiparallel ␤-strands arranged in the sequence with order 3-2-6-7-8-5-4. The monomers are joined together by their terminal ␤-strands, ␤1 and ␤10, which interact with ␤8 and ␤10 of the other monomer creating three two-stranded inter-subunit antiparallel ␤-sheets, ␤1↑↓ ␤8 , ␤1 ↑↓ ␤8, and ␤10↑↓␤10 . The N-terminal ␣1 helix extends over the neighboring subunit where it caps one end of the ␤-barrel. Native FerA, isolated from P. denitrificans, was initially estimated as monomeric in dilute solutions (Mazoch et al., 2004). The availability of large amounts of recombinant protein enabled a detailed SAXS study on its oligomerization state. Data collection parameters and integral structural parameters are summarized in Table 2. Radius of gyration (Rg ) calculated using Guinier approximation was 22.97 Å for c = 6 mg ml−1 and 21.39 Å for c = 2 mg ml−1 . The Porod volume was determined as 5.74 × 103 Å3 for c = 6 mg ml−1 and 5.90 × 103 Å3 for c = 2 mg ml−1 . Values of Rg and Porod volume are well comparable for both concentrations, indicating no changes in oligomerization state of FerA over the concentration range analyzed. Evaluation of theoretical scattering of the crystal structure of FerA and fitting to experimental SAXS data was performed by CRYSOL. Monomeric model FerA (asymmetric unit) gave poor fit with discrepancy 2 > 20. The most probable dimeric FerA assembly suggested by PISA resulted in a good overall fit with 2 = 1.87 for c = 6 mg ml−1 and 2 = 1.01 for c = 2 mg ml−1 (Fig. 7). Higher 2 value could be caused by a substantial number of residues missing in the crystal structure (54 of 374 in dimeric FerA). Ab initio model reconstructed by DAMMIN agrees well in size and shape with the dimeric form of FerA (Fig. 7). Thus, the current SAXS data suggest that the predicted FerA dimer is really present in the solution. 3.5. Structure of complexes of FerA with substrates Although the aforementioned titration experiments suggested the existence of binary complexes of FerA with both FMN and FAD, we only managed to obtain co-crystals of FerA with FMN. Their space group remained P31 21 as for FerA and the unit cell dimensions were only slightly affected (Table 1). The crystal structure of the complex (PDB ID: 4XJ2) shows FMN to be located in a cavity at the interface between the subunits of the homodimer. Electron density is well defined for the isoalloxazine ring but less well ordered for the ribityl moiety (Fig. 8A). The O4 atom of the isoalloxazine forms a hydrogen bond with the hydroxyl group of Ser57 while O2, O4, N3 and N5 of isoalloxazine and O2 and O4 of the ribityl side chain bind to backbone atoms on the loops ␤3–␤4,

16

V. Sedláˇcek et al. / Microbiological Research 188 (2016) 9–22

Fig. 8. Structure of the active site in FerA. Residues of amino acids interacting directly with FMN and hydrogen bonds formed between these residues and FMN are shown in panel A. Hydrogen bonds and amino acids that interact with NAD+ derived from PheA2 (1RZ1) are shown in panel B. Panel C shows a comparison of tertiary structure of FerA without (yellow, 4XHY) and with FMN (blue, 4XJ2) in the region of the ␣3–␤7 loop in the vicinity of the Arg106 residue. The images were generated with Chimera 1.10.1 [22].

␤5–␣2, and ␣3–␤7. The protein structure of FerA in the complex is very close to that of the flavin-free protein, with a root mean square deviation (RMSD) of 0.285 Å between the corresponding 161C␣ atoms pairs. A superposition of the two structures (Fig. 8C) reveals that the sole significant conformational change induced by flavin binding is a movement of the ␣3–␤7 loop in the vicinity of the Arg106 residue, the side chain CZ atom of which becomes placed 5 Å away from the phosphorus atom of FMN. Since FerA displays sequential kinetics, the existence of a ternary complex of FerA with FMN and NAD(H) can also be expected. However, all the attempts to prepare appropriate crystals were unsuccessful. A comparison with existing high resolution structures using PDBeFold server (Krissinel and Henrick, 2004) (Table 3) identified the closest structural homolog of FerA as the flavin reductase PheA2 from Geobacillus (formerly Bacillus) thermoglucosidasius (van den Heuvel et al., 2004), with a Z-score of 13.8 and an RMSD

over the aligned C␣ atoms of 1.46 Å. The published PDB structure 1RZ1 contains NAD in a folded conformation in which the re-face (A-side) face of the nicotinamide ring lies against the adenine ring, leaving the si-face (B-side) available for interaction with the isoalloxazine moiety of flavin. The hydride transfer stereochemistry thus parallels that observed for FerA. Using the PheA2 structure as a guide, we manually placed NAD into the active site of FerA with bound FMN and then energy-minimized by means of the Amber force field (Fig. 8B). The residues Arg7, Asn34, His123, and Ser38 (in the neighbor subunit) were predicted to be involved in hydrogen bonds to NAD in PheA2 (van den Heuvel et al., 2004). They all have homologous counterparts in FerA (Arg29, Asn56, His146, and Ser60 ) that occupy similar spatial positions and thus may fulfill the same functions. We also examined the structure of the Archaeoglobus fulgidus ferric reductase FeR (PDB ID 1I0S) as an example of an FerA homolog that binds NADP(H) in an

V. Sedláˇcek et al. / Microbiological Research 188 (2016) 9–22

17

Table 3 Sequence and structural comparisons of FerA with a flavin reductase like domain protein family (PF01613). PDB Code

UniProt ID

Protein name

Organism

Z-score (PDBeFold)

Sequence Identity (%)

RMSD (Å) (Number of residues)

3NFW

E5Q9D7

NmoB

13.8

37

1.59 (154)

1RZ0

Q9LAG2

PheA2

13.8

32

1.46 (146)

2ECR

Q5SJP7

HpaC

13.3

32

1.30 (137)

4F07

O33495

SmoB

13.3

30

1.59 (143)

2D36 3K87

Q72HI0 O87008

HpaC TftC

12.5 10.9

25 28

1.76 (148) 1.60 (150)

3CB0 1I0R

Q8YHT7 O29428

CobR FeR

Mycobacterium thermoresistibile Bacillus thermoglucosidasius Thermus thermophilus Pseudomonas putida Sulfolobus tokodaii Burkholderia cepacia Brucella melitensis Archaeoglobus fulgidus

10.9 9.7

28 24

1.67 (145) 1.94 (139)

extended conformation (Chiu et al., 2001). Among the amino acid residues participating in hydrogen bonds with NADP in FeR (Thr31, His126, Tyr147, Tyr150, Lys154) only histidine is conserved in FerA (His146). It therefore emerged as improbable that FerA would bind NAD(H) in a FeR-like fashion. 3.6. Effect of mutational alteration of Arg29, Asn56, Arg106 and His146 to alanine The amino acid residues implicated in the interaction with substrates were changed to alanine by site-directed mutagenesis. Alanine was chosen as a substitution residue of choice as it does not possess the side chain beyond the ␤-carbon, thus it neither alters the main-chain conformation nor introduces electrostatic or steric effects within the protein. The purified mutant proteins were checked by running CD spectra in order to confirm proper folding (data not shown) and subjected to affinity and kinetic measurements under conditions already described above, except for the fact that a higher amount of the enzyme was generally needed to achieve an appreciable rate of reaction. We found that the mutants still bind substrates, retain a residual enzymatic activity and follow the same basic mechanism and stereochemistry as the parent FerA, albeit with different parameter sets (Table 4). R29A, N56A and H146A substitutions led to at least tenfold reduction in kcat while R106A had only about a threefold effect. Mutating Arg106 and His146 mildly impaired the binding of FMN and NADH, respectively, which is indicated by the increased Km and Kd values for these substrates. On the other hand, Km and Kd for NADH were not affected by the mutation of Arg29 or Asn56. It is therefore unlikely that the interactions between the latter residues and NADH inferred from the crystal structures are energetic enough to play a significant role in NADH binding. 3.7. The fate of dihydroflavins Dihydroflavins FMNH2 and FADH2 generated in the FerAcatalyzed reaction are unstable to oxidation with molecular oxygen. One mechanism of their protection from an oxygen attack might involve their direct delivery from FerA to other proteins through formation of specific protein–protein interactions. With the aim to identify potential FerA binding partners, we set up a pull-down assay in which proteins from cell lysates were captured by FerA coupled to beads with FMN either present or absent, and analyzed by SDS-PAGE and mass spectrometry. These analyses, however, provided no conclusive evidence for the presence of a specific FerA binder. Therefore we chose to examine another possibility that some antioxidant enzymes, which are normally

present in an in vivo environment, may confer increased stability to dihydroflavins. Fig. 9 shows representative stopped flow traces recorded at 450 nm, the maximum of the oxidized flavin absorbance peak, after rapidly mixing a catalytic amount of FerA with FMN and NADH under aerobic conditions. It can be seen that FMN was first reduced to FMNH2 and then, when the remaining NADH could no longer sustain the reaction, rapid back-oxidation of FMNH2 by O2 begun. Introduction of superoxide dismutase into the reaction mixture caused a concentration dependent prolongation of the time interval between flavin reduction and oxidation. This is consistent with previous reports of inhibition by superoxide dismutase of FMNH2 autooxidation (Kemal et al., 1977; Sucharitakul et al., 2007). A similar effect was achieved by the addition of the P. denitrificans flavoprotein FerB (NAD(P)H:acceptor oxidoreductase). The time lag was increased up to tenfold in the presence of the latter enzyme. FerB also catalyzes reduction of FMN, but at a rate two orders of magnitude slower than FerA (Sedlacek et al., 2014). Such a small activity cannot in itself account for the lag period changes elicited by FerB. A more plausible explanation lies in the recently discovered superoxide scavenging by this enzyme (Sedlacek et al., 2015). Almost similar kinetic traces as in Fig. 9 were seen when FerA was omitted and FMN was anaerobically reduced by titration with dithionite before mixing with oxygenated assay buffer (results not shown). This apparently rules out direct interactions of FerA with SOD or FerB as a basis for the observed lags.

4. Discussion The present work identifies a novel member of the family of short-chain flavin reductases (Galan et al., 2000). It has a lot in common with the other members, but there are also significant differences. A valuable insight can be gained from a comparison of FerA with the prototype enzyme PheA2 from G. thermoglucosidasius A7. Although both proteins have strikingly similar tertiary structures, they basically differ in catalytic mechanism, thus belonging to different classes of flavin reductases in the Tu’s classification system. Whereas the reaction of FerA proceeds according to a random ternary complex mechanism (this work), PheA2 contains a bound FAD cofactor that accepts electrons from NADH and then donates them to another flavin molecule (a flavin substrate) with bi–bi pingpong (double-displacement) kinetics (Kirchner et al., 2003). As seen in Fig. 10, a distinguishing feature of the crystal structures of FerA and PheA2 is the position of a loop extending from the ␣3 helix. In FerA it adopts a conformation that does not allow the adenosine part of FAD to fit easily into the enzyme because of steric hindrance. This could help explain why the dissociation constants of FAD and FMN complexes with FerA are similar to each other,

18

V. Sedláˇcek et al. / Microbiological Research 188 (2016) 9–22

Table 4 Kinetic parameters of FerA for its mutant variants.

WT R106A R29A N56A H146A

kcat (s−1 )

Km FMN (␮M)

89 ± 9 30 ± 3 3.3 ± 0.2 6.3 ± 0.7 2.7 ± 0.1

3.8 16 2.9 14 3.6

± ± ± ± ±

0.5 3 0.3 6 0.4

Km NADH (␮M) 60 53 48 50 76

± ± ± ± ±

10 9 7 10 5

Kd FMN (␮M) 2.7 10 2.8 15 2.5

± ± ± ± ±

0.7 1 0.2 1 0.2

Kd NADH (␮M) 59 59 58 53 67

± ± ± ± ±

9 4 4 6 7

The kinetic parameters were obtained by the initial velocity kinetic analysis at a constant FMN (0.05 mM) or NADH (0.2 mM) and various concentrations of NADH or FMN. Measurements were performed at 30 ◦ C with 5.2 ␮M FerA-R29A, 1.2 ␮M FerA-N56A, 1.0 ␮M FerA-R106A, or 2.0 ␮M FerA-H146A in 25 mM Tris–HCl buffer (pH 7.4).

Fig. 9. Rapid kinetics of FMN oxidation. Kinetic traces of FerA-catalyzed reaction of NADH with FMN in the presence of superoxide dismutase (panel A) or FerB (panel B) were monitored at 450 nm at the stopped-flow apparatus. 100 ␮M NADH and 40 ␮M FMN in a final volume of 0.15 mL of 25 mM Tris–HCl buffer (pH 7.4) at 10 ◦ C were mixed with 1.4 ␮M FerA and traces were recorded for various concentrations of superoxide dismutase equal to 0 (1), 0.6 (2), 1.6 (3), 3.1 (4), 6.3 (5), and 18.8 ␮M (6) or FerB equal to 0 (1), 0.1 (2), 0.25 (3), 0.5 (4), 1.0 (5), and 3.0 ␮M (6).

and are about three orders of magnitude higher than the value of Kd FAD = 9.8 ± 0.2 nM determined previously by a fluorometric titration of the PheA2 apoprotein with FAD (van den Heuvel et al., 2004). These results support and extend the hypothesis that variations in the affinities for FAD in various flavin reductases may be attributed to difference in the interaction between the AMP moiety of FAD and a less conserved loop region which shows structural divergence (Okai et al., 2006; Kim et al., 2008). Comparative analysis of the structures shows that the FMN binding site of FerA corresponds to a part of the FAD cofactor binding site of PheA2 (Fig. 10). However, as evidenced from kinetic data in the current study, only the first, reductive, half-reaction of the catalytic cycle of PheA2 takes place in FerA because the reduced flavin formed is no longer retained at the binding site and dissociates from the enzyme due to a drop in affinity (Figs. 2 and 3). Details on interaction of FerA with FAD are less clear than is the case with FMN due to the absence of structural data for an FerA-FAD complex. In aqueous solution, the FAD molecule is predominantly in a compact, bent conformation in which the isoalloxazine and adenine ring systems stack coplanarly. As a result, the fluorescence of the flavin is efficiently quenched (Weber, 1950; van den Berg et al., 2002). The fluorescence intensity of PheA2-bound FAD is significantly higher when compared to free FAD (van den Heuvel et al., 2004), which correlates with the structurally governed ability of PheA2 to stabilize FAD in its extended, unstacked state. An opposite behavior of FerA, i.e., a drop in fluorescence upon FAD binding (Fig. 4), might indicate a preference for the bent conformation of FAD over the extended one, although other mechanisms of quenching probably also contribute. Similar FAD fluorescence quenching effect was observed previously for some other flavin reductases (Lee and Zhao, 2007; Chakraborty et al., 2010; Groning et al., 2014). FAD in bent conformation occurs naturally in enzymes (Dym and Eisenberg, 2001) and flavin binding proteins (Grininger et al., 2006). Its specific presence in bacterial ferredoxin-NADP(H) reductases is considered to underlie a lower catalytic activity of bacterial enzymes with respect to plastidic counterparts containing

extended FAD (Ceccarelli et al., 2004). Two molecules FAD, one in extended and another in folded conformation, were recently identified in the cofactor and substrate binding sites of one of the FerA homologs, the corrin reductase (CobR), when it crystallized with an excess of FAD (Lawrence et al., 2014). The authors claimed that the substrate FAD is configured in a curled fashion extremely similar to the conformation of NADH in PheA2. CobR binds FAD stronger than FerA does, having a one-order smaller Kd value, 230 ± 27 nM (Lawrence et al., 2008). We assume that FAD interacts with FerA through its isoalloxazine part in analogous manner to FMN (Fig. 10) and that the extended form of FAD cannot be stabilized as it occurs in PheA2 or CobR. The active site of FerA is apparently spacious enough to allow sufficient flexibility of the adenosine end of the FAD molecule such that the isoalloxazine ring can become exposed for the reaction with NADH. Our kinetic data show that FAD is a slightly worse substrate for FerA than FMN and that some steric hindrance between the FAD-NADH substrate pair may occur, as opposed to the less space-consuming FMN-NADH pair. Crystal structure analysis and direct mutagenesis experiments have led to identification of two amino acid residues, specifically involved in recognition of either flavin (Arg106) or NADH (His146). The interaction between the phosphate tail of FMN and the guanidinium moiety of Arg106 seems to be specific for FerA since this amino acid is not conserved across other FRs (Fig. 11). Even though not being in the correct position for the formation of hydrogen bonds, both groups can still interact electrostatically, thus contributing to the stability of the complex. Evidence for this can be deduced from a reduced affinity of R106A for FMN (Table 4), as well as weakened binding of FMN analogs with truncated side chain (riboflavin, lumichrome) to the wild-type FerA. From a simple Coulomb’s Law standpoint, elementary charges separated by 5 Å in water have an interaction energy of about 3.5 kJ mol−1 . This represents a near fourfold change in the dissociation constant, which is very close to what we observed experimentally. His146 is part of a conserved sequence in many FRs, the GDH (glycine, aspartic acid, histidine) motif. The importance of this histidine residue in

V. Sedláˇcek et al. / Microbiological Research 188 (2016) 9–22

19

Fig. 10. Comparison of FerA from P. denitrificans with PheA2 from B. thermoglucosidasius. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) Surfaced model of FerA (green) and PheA2 (brown) are shown in panel A and B. Detailed models in the same coloured arrangement are shown in panel C and D.

NADH binding and utilization was first demonstrated for the FR from Aminobacter aminovorans (Russell and Tu, 2004). His146 of P. denitrificans FerA is suitably located to make a hydrogen bond with the C-3 carboxamide moiety of NADH. It is therefore likely to assists in fixing the dihydropyridine ring of NADH in a manner critical for an efficient hydride transfer from the C-4 atom. Its mutation to alanine probably makes the ring unstable in regards to rotation, which leads to the observed reduction in kcat . Free rotation of the ring is evidently not possible, however, since the stereochemistry of hydride transfer is unaffected by the H146A mutation. Most of the flavin reductases with high structural similarity to FerA (Table 3) act as small components of two component monooxygenase systems in which they supply reduced flavins to the large component monooxygenases involved in oxidative degradation of aromatic or hydrocarbon compounds. The genes for these enzymes are clustered together on the bacterial chromosomes. FerA occupies a distinct position in that it probably does not cooperate with a single defined flavin acceptor. This can be inferred because (i) a PmbA-like peptidase and an inositol monophosphatase, the putative products of the Pden 2687 and Pden 2688 genes neighboring the Pden 2689 gene coding for FerA, are unlikely to participate in any flavin-requiring redox processes, and (ii) pull-

down assays reported here did not reveal any evidence for FerA being capable of directly associating with a specific cellular protein. Given a body of experimental evidence in support for flavin transfer between proteins via free diffusion (Sucharitakul et al., 2014), it is conceivable that FerA takes part in formation of an intracellular pool of reduced flavins serving as enzyme substrates and chemical reducing agents (e.g., for reductive release of iron from its complex with a siderophore, (Sedlacek et al., 2009)). Reduced flavins (FlH2 ) are susceptible to spontaneous nonenzymatic oxidation by molecular oxygen (O2 ) leading to, as final products, oxidized flavin (Fl) and hydrogen peroxide. There is an induction period at the start of the reaction, connected with the formation of two autocatalytic agents, Fl and superoxide (O2 - ). Fl comproportionates with unreacted FlH2 to give two molecules of flavin semiquinone (FlH). FlH is also generated in one-electron oxidations of FlH2 by O2 - (rapid) or O2 (slow). The anionic form of FlH, Fl- , then undergoes a rapid reaction with molecular oxygen, producing Fl and O2 - and closing in this way a positive feedback (autocatalytic) loop (Kemal et al., 1977). In view of the foregoing, one would expect an increase in FlH2 stability at the concerted action of a flavin reductase and an enzyme able to eliminate superoxide. This is indeed what we observed experimentally not

20

V. Sedláˇcek et al. / Microbiological Research 188 (2016) 9–22

Fig. 11. Structure-based sequence alignment of selected structural homology proteins. Strictly conserved residues are shown in black, dark gray shades indicate a high degree of conservation, while light grays indicate various degrees of low conservation. Two consensus motifs, S/T/CXXPP and GDH that are characteristic of the HpaC-like subfamily of the non-flavoprotein are outlined by solid black lines. The mutated residues are numbered. Generally, shaded residues indicate that a conservation of similar residues persists across at least 70% of the alignment. The alignment was generated using ClustalW ver. 1.81 with a BLOSUM62 matrix and default parameters and processed with BoxShade 3.3.1. Organisms and ORF designations from the corresponding genomic sequence: P. denitrificans FerA (P84468, gi: 62286691), M. thermoresistibile NmoB (E5Q9D7, gi: 299689362), G. thermoglucosidasius PheA2 (Q9LAG2, gi: 47168900), T. thermophilus HpaC (Q5SJP7, gi: 165760831), P. putida SmoB (O33495, gi: 491668269), S. tokodaii HpaC (Q72HI0, gi: 109157431), B. cepacia TftC (O87008, gi: 266618800), B. melitensis CobR (Q8YHT7, gi: 169791887) and A. fulgidus FeR (O29428, gi: 14278202).

only with superoxide dismutase but also with the ferric reductase B (FerB) enzyme (Fig. 9). FerB is a FMN-containing NAD(P)Hdependent oxidoreductase which among other substrates reduces superoxide (Sedlacek et al., 2015). Its Km for NADH is one order of magnitude lower compared to FerA, so that it can work effectively also at NADH concentrations subsaturating for FerA. FerB was recently shown to protect P. denitrificans cells from the oxidative stress created by exposure to methyl viologen (Pernikarova et al.,

2015). The present finding of FMNH2 stabilization by FerB thus adds further support to the proposed role of FerB as an important endogenous enzymatic antioxidant. The presence of antioxidative milieu within the cells has to be taken into account before making any generalization from in vitro experiments on dihydroflavin transfer between the purified enzymes.

V. Sedláˇcek et al. / Microbiological Research 188 (2016) 9–22

Author contributions VS produced and purified the recombinant protein and its mutants, performed majority of experimental work, made data analysis and contributed to manuscript preparation. TK and JM performed crystallization, X-ray and SAXS data collection and structure determination. IK conceived and supervised the project, made data analysis and wrote the main body of the manuscript. Acknowledgments This research was supported by the Czech Science Foundation project No. GAP503/12/0369 to I.K. The X-ray crystallography and SAXS part (T.K.) of the work was also supported by the project CZ.1.07/2.3.00/30.0037. We thank the staff of ITC, supported by open access project LM2011020 of CEITEC (Central European Institute of Technology). We wish to thank the EMBL/DESY Hamburg for providing us with synchrotron facilities and M. Groves for his assistance with data collection. We are grateful to Sushil Kumar Mishra (CEITEC) for performing energy minimization and to Zbynek Prokop and Jiri Damborsky for access to the stopped-flow apparatus. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.micres.2016.04. 006. References Case, D.A., Berryman, J.T., Betz, R.M., Cai, Q., Cerutti, D.S., Cheatham III, T.E., Darden, T.A., Duke, R.E., Gohlke, H., Goetz, A.W., Gusarov, S., Homeyer, N., Janowski, P., Kaus, J., Kolossva´ıry, I., Kovalenko, A., Lee, T.S., LeGrand, S., Luchko, T., Luo, R., Madej, B., Merz, K.M., Paesani, F., Roe, D.R., Roitberg, A., Sagui, C., Salomon-Ferrer, R., Seabra, G., Simmerling, C.L., Smith, W., Swails, J., Walker, R.C., Wang, J., Wolf, R.M., Wu, X., Kollman, P.A., 2014. Amber 14. University of California, San Francisco. Ceccarelli, E.A., Arakaki, A.K., Cortez, N., Carrillo, N., 2004. Functional plasticity and catalytic efficiency in plant and bacterial ferredoxin-NADP(H) reductases. Biochem. Biophys. Acta 1698 (2), 155–165. Chakraborty, S., Ortiz-Maldonado, M., Entsch, B., Ballou, D.P., 2010. Studies on the mechanism of p-hydroxyphenyl acetate 3-hydroxylase from Pseudomonas aeruginosa: a system composed of a small flavin reductase and a large flavin-dependent oxygenase. Biochemistry 49 (2), 372–385. Chiu, H.J., Johnson, E., Schroder, I., Rees, D.C., 2001. Crystal structures of a novel ferric reductase from the hyperthermophilic archaeon Archaeoglobus fulgidus and its complex with NADP(+). Structure 9 (4), 311–319. Cohen, S.X., Ben Jelloul, M., Long, F., Vagin, A., Knipscheer, P., Lebbink, J., Sixma, T.K., Lamzin, V.S., Murshudov, G.N., Perrakis, A., 2008. ARP/wARP and molecular replacement: the next generation. Acta Crystallogr. D: Biol. Crystallogr. 64, 49–60. Dym, O., Eisenberg, D., 2001. Sequence-structure analysis of FAD-containing proteins. Protein Sci. 10 (9), 1712–1728. Ellis, H.R., 2010. The FMN-dependent two-component monooxygenase systems. Arch. Biochem. Biophys. 497 (1–2), 1–12. Emsley, P., Lohkamp, B., Scott, W.G., Cowtan, K., 2010. Features and development of Coot. Acta Crystallogr. D: Biol. Crystallogr. 66, 486–501. Finn, R.D., Mistry, J., Tate, J., Coggill, P., Heger, A., Pollington, J.E., Gavin, O.L., Gunasekaran, P., Ceric, G., Forslund, K., Holm, L., Sonnhammer, E.L.L., Eddy, S.R., Bateman, A., 2010. The Pfam protein families database. Nucleic Acids Res. 38, D211–D222. Galan, B., Diaz, E., Prieto, M.A., Garcia, J.L., 2000. Functional analysis of the small component of the 4-hydroxyphenylacetate 3-monooxygenase of Escherichia coli W: a prototype of a new flavin: nAD(P)H reductase subfamily. J. Bacteriol. 182 (3), 627–636. Grininger, M., Seiler, F., Zeth, K., Oesterhelt, D., 2006. Dodecin sequesters FAD in closed conformation from the aqueous solution. J. Mol. Biol. 364 (4), 561–566. Groning, J.A.D., Kaschabek, S.R., Schlomann, M., Tischler, D., 2014. A mechanistic study on SMOB-ADP1: an NADH:flavin oxidoreductase of the two-component styrene monooxygenase of Acinetobacter baylyi ADP1. Arch. Microbiol. 196 (12), 829–845. Huijbers, M.M., Montersino, S., Westphal, A.H., Tischler, D., van Berkel, W.J., 2014. Flavin dependent monooxygenases. Arch. Biochem. Biophys. 544, 2–17. Joosten, R.P., Long, F., Murshudov, G.N., Perrakis, A., 2014. The PDB REDO server for macromolecular structure model optimization. IUCrJ 1, 213–220. Kabsch, W., 2010. Integration, scaling, space-group assignment and post-refinement. Acta Crystallogr. D: Biol. Crystallogr. 66, 133–144.

21

Kemal, C., Chan, T.W., Bruice, T.C., 1977. Reaction of 3 O2 with dihydroflavins 1. N3,5-dimethyl-1,5-dihydrolumiflavin and 1,5-dihydroisoalloxazines. J. Am. Chem. Soc. 99 (22), 7272–7286. Kim, S.H., Hisano, T., Iwasaki, W., Ebihara, A., Miki, K., 2008. Crystal structure of the flavin reductase component (HpaC) of 4-hydroxyphenylacetate 3-monooxygenase from Thermus thermophilus HB8: structural basis for the flavin affinity. Proteins 70 (3), 718–730. Kirchner, U., Westphal, A.H., Muller, R., van Berkel, W.J.H., 2003. Phenol hydroxylase from Bacillus thermoglucosidasius A7, a two-protein component monooxygenase with a dual role for FAD. J. Biol. Chem. 278 (48), 47545–47553. Konarev, P.V., Volkov, V.V., Sokolova, A.V., Koch, M.H.J., Svergun, D.I., 2003. PRIMUS: a Windows PC-based system for small-angle scattering data analysis. J. Appl. Crystallogr. 36, 1277–1282. Krissinel, E., Henrick, K., 2004. Secondary-structure matching (SSM), a new tool for fast protein structure alignment in three dimensions. Acta Crystallogr. D Biol. Crystallogr. 60 (Pt. 12 Pt. 1), 2256–2268. Laemmli, U.K., 1970. Cleavage of structural proteins during assembly of head of bacteriophage-T4. Nature 227 (5259), 680–685. Lakowicz, J.R., 2006. Principles of Fluorescence Spectroscopy, 3rd ed. Springer, New York. Lambeth, J.D., Kamin, H., 1976. Adrenodoxin reductase—properties of complexes of reduced enzyme with NADP+ and NADPH. J. Biol. Chem. 251 (14), 4299–4306. Lawrence, A.D., Deery, E., McLean, K.J., Munro, A.W., Pickersgill, R.W., Rigby, S.E.J., Warren, M.J., 2008. Identification, characterization, and structure/function analysis of a corrin reductase involved in adenosylcobalamin biosynthesis. J. Biol. Chem. 283 (16), 10813–10821. Lawrence, A.D., Taylor, S.L., Scott, A., Rowe, M.L., Johnson, C.M., Rigby, S.E.J., Geeves, M.A., Pickersgill, R.W., Howard, M.J., Warren, M.J., 2014. FAD binding, cobinamide binding and active site communication in the corrin reductase (CobR). Biosci. Rep. 34, 345–355. Lee, J.K., Zhao, H.M., 2007. Identification and characterization of the flavin: NADH reductase (PrnF) involved in a novel two-component arylamine oxygenase. J. Bacteriol. 189 (23), 8556–8563. Leskovac, V., 2003. Comprehensive Enzyme Kinetics. Kluwer Academic/Plenum Pub, New York. Lostao, A., El Harrous, M., Daoudi, F., Romero, A., Parody-Morreale, A., Sancho, J., 2000. Dissecting the energetics of the apoflavodoxin-FMN complex. J. Biol. Chem. 275 (13), 9518–9526. Massey, V., 1990. A simple method for the determination of redox potentials. In: Curti, B., Ronchi, S., Zanetti, G. (Eds.), Flavins and Flavoproteins. de Gruyter W, Berlin, pp. 59–66. Mazoch, J., Tesarik, R., Sedlacek, V., Kucera, I., Turanek, J., 2004. Isolation and biochemical characterization of two soluble iron(III) reductases from Paracoccus denitrificans. Eur. J. Biochem. 271 (3), 553–562. Mccoy, A.J., Grosse-Kunstleve, R.W., Adams, P.D., Winn, M.D., Storoni, L.C., Read, R.J., 2007. Phaser crystallographic software. J. Appl. Crystallogr. 40, 658–674. Miethke, M., 2013. Molecular strategies of microbial iron assimilation: from high-affinity complexes to cofactor assembly systems. Metallomics 5 (1), 15–28. Murshudov, G.N., Skubak, P., Lebedev, A.A., Pannu, N.S., Steiner, R.A., Nicholls, R.A., Winn, M.D., Long, F., Vagin, A.A., 2011. REFMAC5 for the refinement of macromolecular crystal structures. Acta Crystallogr. D: Biol. Crystallogr. 67, 355–367. Okai, M., Kudo, N., Lee, W.C., Kamo, M., Nagata, K., Tanokura, M., 2006. Crystal structures of the short-chain flavin reductase HpaC from Sulfolobus tokodaii strain 7 in its three states: NAD(P)(+)-free, NAD(+)-bound, and NADP(+)-bound. Biochemistry 45 (16), 5103–5110. Pernikarova, V., Sedlacek, V., Potesil, D., Prochazkova, I., Zdrahal, Z., Bouchal, P., Kucera, I., 2015. Proteomic responses to a methyl viologen-induced oxidative stress in the wild type and FerB mutant strains of Paracoccus denitrificans. J. Proteom. 125, 68–75. Petoukhov, M.V., Franke, D., Shkumatov, A.V., Tria, G., Kikhney, A.G., Gajda, M., Gorba, C., Mertens, H.D., Konarev, P.V., Svergun, D.I., 2012. New developments in the program package for small-angle scattering data analysis. J. Appl. Crystallogr. 45 (Pt. 2), 342–350. Pettersen, E.F., Goddard, T.D., Huang, C.C., Couch, G.S., Greenblatt, D.M., Meng, E.C., Ferrin, T.E., 2004. UCSF chimera—a visualization system for exploratory research and analysis. J. Comput. Chem. 25 (13), 1605–1612. Russell, T.R., Tu, S.C., 2004. Aminobacter aminovorans NADH:flavin oxidoreductase His140: a highly conserved residue critical for NADH binding and utilization. Biochemistry 43 (40), 12887–12893. Schneider, C., Suhnel, J., 1999. A molecular dynamics simulation of the flavin mononucleotide-RNA aptamer complex. Biopolymers 50 (3), 287–302. Sedlacek, V., van Spanning, R.J.M., Kucera, I., 2009. Ferric reductase A is essential for effective iron acquisition in Paracoccus denitrificans. Microbiology 155, 1294–1301. Sedlacek, V., Klumpler, T., Marek, J., Kucera, I., 2014. The structural and functional basis of catalysis mediated by NAD(P)H:acceptor oxidoreductase (FerB) of Paracoccus denitrificans. PLoS One 9 (5), e96262. Sedlacek, V., Ptackova, N., Rejmontova, P., Kucera, I., 2015. The flavoprotein FerB of Paracoccus denitrificans binds to membranes, reduces ubiquinone and superoxide, and acts as an in vivo antioxidant. FEBS J. 282 (2), 283–296. Sucharitakul, J., Phongsak, T., Entsch, B., Svasti, J., Chaiyen, P., Ballou, D.P., 2007. Kinetics of a two-component p-hydroxyphenylacetate hydroxylase explain how reduced flavin is transferred from the reductase to the oxygenase. Biochemistry 46 (29), 8611–8623.

22

V. Sedláˇcek et al. / Microbiological Research 188 (2016) 9–22

Sucharitakul, J., Tinikul, R., Chaiyen, P., 2014. Mechanisms of reduced flavin transfer in the two-component flavin-dependent monooxygenases. Arch. Biochem. Biophys. 555, 33–46. Svergun, D., Barberato, C., Koch, M.H.J., 1995. CRYSOL—a program to evaluate X-ray solution scattering of biological macromolecules from atomic coordinates. J. Appl. Crystallogr. 28, 768–773. Svergun, D.I., 1999. Restoring low resolution structure of biological macromolecules from solution scattering using simulated annealing. Biophys. J. 76 (6), 2879–2886. Tesarik, R., Sedlacek, V., Plockova, J., Wimmerova, M., Turanek, J., Kucera, I., 2009. Heterologous expression and molecular characterization of the NAD(P)H:acceptor oxidoreductase (FerB) of Paracoccus denitrificans. Protein Expres. Purif. 68 (2), 233–238. Tu, S.C., 2001. Reduced flavin: donor and acceptor enzymes and mechanisms of channeling. Antioxid. Redox Sign. 3 (5), 881–897. Walker, R.C., de Souza, M.M., Mercer, I.P., Gould, I.R., Klug, D.R., 2002. Large and fast relaxations inside a protein: calculation and measurement of reorganization

energies in alcohol dehydrogenase. J. Phys. Chem. B 106 (44), 11658–11665. Weber, G., 1950. Fluorescence of riboflavin and flavin-adenine dinucleotide. Biochem. J. 47 (1), 114–121. Xu, Q., Canutescu, A.A., Wang, G., Shapovalov, M., Obradovic, Z., Dunbrack Jr., R.L., 2008. Statistical analysis of interface similarity in crystals of homologous proteins. J. Mol. Biol. 381 (2), 487–507. van den Berg, P.A.W., Feenstra, K.A., Mark, A.E., Berendsen, H.J.C., Visser, A.J.W.G., 2002. Dynamic conformations of flavin adenine dinucleotide: simulated molecular dynamics of the flavin cofactor related to the time-resolved fluorescence characteristics. J. Phys. Chem. B 106 (34), 8858–8869. van den Heuvel, R.H.H., Westphal, A.H., Heck, A.J.R., Walsh, M.A., Rovida, S., van Berkel, W.J.H., Mattevi, A., 2004. Structural studies on flavin reductase PheA2 reveal binding of NAD in an unusual folded conformation and support novel mechanism of action. J. Biol. Chem. 279 (13), 12860–12867.