Active site analysis of yeast flavohemoglobin based on its structure with a small ligand or econazole

Authors


Correspondence

L. Baciou, Laboratoire de Chimie Physique, Bat. 350, CNRS – Université de Paris Sud, UMR8000, F-91405 Orsay, France

Fax: +33 (0) 1 69 15 6188

Tel: +33 (0) 1 69 15 7711

E-mail: laura.baciou@u-psud.fr

U. Ermler, Max-Planck-Institut für Biophysik, Frankfurt, Germany

Fax: +49 (0) 69 6303 1002

Tel: +49 (0) 69 6303 1054

E-mail: ulrich.ermler@biophys.mpg.de

Abstract

Flavohemoglobins (flavoHbs) serve various microorganisms as the major protective enzymes against NO˙-mediated toxicity. FlavoHbs dominantly function as an NO˙ dioxygenase (inline image), the required electron being shuttled from NAD(P)H via FAD to the heme iron. The X-ray structures of the flavoHb from Saccharomyces cerevisae presented in complex with an unknown small ligand (Yhb) and with econazole (YhbE) at 2.1 and 3.0 Å resolutions, respectively, reveal a high architectural accordance between prokaryotic and eukaryotic family members. The active site is characterized by a proximal heme side with a strictly conserved histidine, glutamate and tyrosine triad and a highly variable distal heme side with helix shifts up to 10 Å mainly dependent on the presence/absence and size of the bound ligand. In yeast flavoHb, the small heme iron ligand adjusts a catalytically productive active site geometry that reliably suggests the NO and O2 binding site. O2 is activated by its ligation to an electron-rich heme iron and a hydrogen bond to Tyr29 and Gln53. High active site similarities between eukaryotic Yhb and bacterial single-domain globins argue for identical biochemical reactions. Binding of the bulky econazole implies a large-scale induced-fit process concerning, in particular, an outwards shift of helices B and E to increase the active site pocket. Yeast Yhb and Ralstonia eutropha flavoHb both structurally studied in complex with econazole indicate conformational differences between the inhibitors and the polypeptide primarily caused by stable binding of a phospholipid to the latter and by distinct loop D structures.

Database

Structural data and final coordinates of Yhb and Yhb-econazole are available in the Protein Data Bank under the accession numbers 4G1V and 4G1B

Abbreviations
FHP

flavohemoglobin from Ralstonia eutropha

flavoHb

flavohemoglobin

Hmp

flavohemoglobin from Escherichia coli

NOD

NO˙ dioxygenase

Yhb

flavohemoglobin from Saccharomyces cerevisae

Yhbeco

flavohemoglobin from Saccharomyces cerevisae complexed with econazole

Introduction

While globins are present in all kingdoms of life, flavohemoglobins (flavoHbs) occur in bacteria, yeast and filamentous fungi [1, 2] but not in archaea, plants and animals. FlavoHbs are monomeric proteins with a molecular mass of about 45 kDa composed of the globin domain carrying heme b and the flavoprotein ferredoxin reductase built up of an FAD and NADH binding domain. The globin domain represents a distinct phylogenetic group within the hemoglobin gene family with no significant sequence identity to the well-known eukaryotic globins like myoglobin and hemoglobin. Sequence comparisons among the flavoHb family members show amino acid identities of ~ 50%, ~ 38% and ~ 33% for the globin, FAD and NADH binding domains, respectively. Up to now, the three-dimensional structures of the bacterial flavoHbs of Ralstonia eutropha (FHP) [3] and Escherichia coli (Hmp) [4] have been solved. These proteins adopt open and closed states dependent on the absence, presence and chemical nature of the heme ligand [5]. The open-to-closed transition proceeds by a rigid body movement of the NAD binding domain accompanying a large rearrangement of the C-terminal segment from the FAD binding domain towards the crevice at the coincidence point of all three domains.

FlavoHbs catalyze several biochemical reactions and might have different physiological functions depending on the organism and the environmental conditions. It is well established that microbial flavoHbs have an important function in NO˙ signaling modulation and in protection against NO˙ toxicity [1, 6-10]. The latter function is crucial for (pathogenic) microorganisms as invaded mammals and plants defend themselves by massive release of reactive oxygen and nitrogen species. Exposure to NO˙ usually leads to flavoHb expression in these microorganisms. In the presence of O2, bacterial flavoHbs predominantly function as NO˙ dioxygenase (NOD) by converting NO˙ into nitrate (inline image) [7, 11]. In the absence of air, they function as an NO˙ reductase that generates nitrous oxide [12] and, in addition, as alkylhydroperoxide reductase [6, 13]. As flavoHbs reduce the NO˙ concentration during the host–pathogen interaction and thereby suppress the antimicrobial effects of NO˙, the search for potential NOD inhibitors accompanied by a better understanding of the catalytic mechanism is an attractive research subject. It was shown that several azole-based heme iron ligating compounds, in particular antifungal imidazole derivatives, effectively inhibit the flavoHb-catalyzed NOD activity [14]. Recently, the structural basis for binding imidazole derivatives to flavoHbs was established by determination of the X-ray structures of the R. eutropha enzyme in complex with miconazole, econazole and ketoconazole, respectively [5].

This protective role of flavoHbs against nitrosative stress appears also to be used in eukaryotic microorganisms such as Dictyostelium discoideum [15] and Aspergillus oryzae [16] and in human major fungal pathogens Candida albicans [17] and Cryptococcus neoformans [18] where homologous genes were detected. Evidence has been provided that flavoHb of Saccharomyces cerevisae (Yhb) (discovered nearly 60 years ago in this organism [19]) consumes NO˙ [20] and Yhb-deleted strains accumulate NO˙ and reactive nitrogen species [21]. Yhb is encoded by the nuclear gene Yhb1 and is located in both the cytosol and the mitochondrial matrix of normoxic cells but exclusively in the latter in the absence of oxygen [22]. Localization of Yhb in mitochondrial matrix suggests a role for Yhb as an NO detoxifying system under hypoxic conditions by controlling levels of NO˙ that suppress respiration by inhibiting cytochrome c oxidase [19]. Moreover, Yhb expression is activated through the transcription factor Fzflp when exposed to exogenously supplied NO˙ [23]. The expression of yeast flavoHbs is regulated during the cellular growth phase [24] and is induced under aerobic conditions. Interestingly, oxygen has an opposite effect on the regulation of the expression of flavoHbs in most bacteria suggesting that yeast and bacterial flavoHbs may have functional differences [20, 25].

In this paper, we report the crystal structure of the flavoHb from S. cerevisae in native-like (Yhb) and econazole-bound (YhbE) forms obtained at resolutions of 2.1 Å and 3.0 Å respectively. The global structure and the binding of heme, FAD and econazole of Yhb are described and compared with those of R. eutropha and E. coli.

Results and Discussion

The overall structure of Yhb

The Yhb structure was determined at 2.1 Å resolution by the single wavelength anomalous dispersion method using the heme iron as the anomalous scatterer. Data collection and refinement statistics are summarized in Table 1. The X-ray structure of the YhbE complex was established at 3.0 Å resolution (Table 1).

Table 1. Data collection, structure determination and refinement
Data setYhbYhb (Fe edge)YhbE
Data collection
Wavelength (Å)0.97911.73160.992
Space groupR32R32P21
Unit cell parameter a, b, c (Å) (°)172.6, 82.3172.6, 82.3

103.9, 93.8, 111.0

108.2

Number of molecules per asymmetric unit114
Resolution range (Å) (highest shell)30.00–2.10 (2.14–2.1)30.0–2.9 (2.95–2.9)30.00–3.0 (3.08–3.0)
Redundancy3.9 (4.0)10.3 (9.9)3.0 (3.2)
Completeness (%)99.1 (99.8)99.8 (100.0)94.6 (99.1)
Rmerge (%)6.2 (42.7)6.9 (26.1)11.5 (42.9)
I/σ(I)25.9 (3.0)38.7 (9.7)10.3 (5.7)
Anomalous signal (Å) 2.0 (30.0–5.8) 
Figure of merit (Å) 0.61 (30.0–5.8) 
Phasing power (Å) 2.4 (30.0–5.8) 
Refinement
Resolution limit (Å)2.1 30.0–3.0 (3.08–3.0)
Rwork/Rfree (%)17.7/24.7 (20.7/29.0) 23.5/27.6 (29.2/32.0)
Rmsd bond lengths (Å)0.016 0.012
Rmsd bond angles (°)1.57 1.40
Average B2)32.7 30.4

The structure of monomeric Yhb is subdivided into an N-terminal globain domain containing heme b (residues 1–144) with a classical three-on-three α-helical fold, an FAD binding domain (residues 149–260) and a C-terminal NADH binding domain (residues 272–399) (Fig. 1). According to structural alignments (Fig. S1), Yhb and YhbE are both structurally characterized in a closed conformation, the rmsd between them and FHPE or Hmp – also present in a closed state – being between 2.0 and 2.5 Å [3-5]. There are no global structural differences between eukaryotic and prokaryotic representatives of the flavoHb family; however, a few notable deviations exist.

Figure 1.

Overall fold of S. cerevisae flavoHbs. The three domains are the globin (pink), the flavin binding (light green) and the NADH binding domains (green). The cofactors, FAD and heme, are indicated in red.

The most striking differences include the positions of various helices of the globin domains which, however, are determined besides sequence variations by the compound bound at the distal heme side. While the positions of helices F and H are essentially maintained, helix E moves up to 10 Å, helices A and B up to 6 Å and helix G up to 5 Å demonstrating the enormous conformational flexibility of the globin fold (Fig. 2A). Of particular importance is the segment between helices C and E termed loop D. It consists of a rather irregular conformation in FHP and Hmp but, interestingly, in a short α-helical segment in Yhb (Fig. 2A). Obviously, loop D represents a variable segment that balances different distances between helices C and E. Therefore, loop D forms different interactions to heme b, the FAD domain and the adenine base of FAD (Fig. 2B). The distinct structure of loop D might also be influenced by the different position of the AMP part of FAD in Yhb, FHP and Hmp (highlighted in Fig. S2). The D helix serves as a structural signature of the vertebrate β-globin fold but was recently also found in a microbial single-domain globin [26] which is highly related to the globin domain of flavoHbs.

Figure 2.

Structural comparison of globin domains of flavoHbs. (A) Superimposition of the globin domains of E. coli (blue), S. cerevisiae (green) and R. eutropha (light grey) flavoHbs. Loop D in Hmp (orange) and FHP (yellow) is irregular and in Yhb (red) α-helical. In ligand-free Hmp, helix E sits in front of the distal heme b side thereby occupying the substrate binding pocket, whereas in Yhb the N-terminal side of helix E is displaced ~ 2.5 Å away from heme b. FHP contains the voluminous ligand in the distal heme pocket and helix E moves a further 3.5 Å away from the distal heme side. (B) Loop D (gold) connects heme b and FAD without being directly involved in the electron transfer process. In Yhb the helical loop D is stabilized by a hydrogen bond between its Asn47 and the pyrrole A propionate, whereas in Hmp Asn47 is about 13 Å apart from heme b and in FHP the equivalent His47 is oriented towards heme b without contacting it (except in the econazole bound state containing a more helical loop D) [5].

Heme and FAD binding sites

The canonical heme b binding site of the globin domain in flavoHbs is coated by mostly hydrophobic residues that are essentially conserved (highlighted in Fig. S3). However, global conformational differences in the distal heme pocket especially those of helix E and loop D described above substantially induce different side-chain positions/conformations of amino acids and thus also different heme–polypeptide interactions (see Fig. 2B). For example, in Yhb and Hmp, the three side chains Gln53, Leu57 and Val61 (Ile61 in Hmp) of helix E point towards heme b. However, Leu57 interacts with the heme iron in Hmp and with the porphyrin scaffold in Yhb, whereas Gln53 points to the iron ligation site in Yhb and is hydrogen bonded to the pyrrole B propionate in Hmp (Fig. 2B). In FHP, solely Ala60 of helix E forms weak interactions to heme b as a phospholipid is bound between heme b and helix E [3]. In addition, the C-terminal arms of flavoHbs interact differently with the pyrrole B propionates which also change their conformations.

The heme b porphyrin rings of the structurally known flavoHbs slightly deviate from planarity, particularly noticeably in Yhb where rings A and B and rings C and D are related by a kink. The pyrrole rings are sandwiched between conserved and hydrophobic residues (Leu57, Ile81, Phe133, Phe43, Val98, Leu88 and Ile90) which, as mentioned, are partly differently positioned and might induce the specific porphyrin deviations. However, the UV–visible absorption spectrum of Yhb is similar to those previously reported for other flavoHbs [12, 27, 28] and we assume similar porphyrin conformations under identical conditions. The spectra of the purified Yhb were recorded in ferric (Fe3+) and NADH-reduced (Fe2+-O2) forms (Fig. 3A,B) resulting in a non-symmetric Soret peak at 402 nm, a broad shoulder at around 540 nm and a charge transfer band at 646 nm in the oxidized state and in a Soret peak at 431 nm and a broad β-band maximally absorbing at 558 nm in the reduced state (Table 2).

Table 2. UV–visible spectral features of Fe3+ (methemoglobin) forms of flavoHbs
FlavoHbsSoret (nm) (±0.3 nm)λ/λ (nm) (±0.3 nm)References
Yhb

402

397–403

488/648

485/645

[28]

Our work

Hmp404491/622 [34, 50]
FHP

400

395

485/645

480/645

[28]

[5]

Figure 3.

Absorption spectra of purified Yhb flavoHb in various oxidation and ligand forms in the UV (A) and visible (B) regions at different scales. Yhb (0.7 μm) was suspended in 20 mm Tris/HCl and 10 mm KCl buffer. The oxidized spectrum is drawn as a thick black line, supplemented with 13 μm miconazole as a grey line and the reduced spectrum after addition of 1 mm NADH as a thin black line. (C) Spectrophotometric titration of ferric Yhb with econazole. Spectral changes accompanying the titration of the ferric protein are shown in the visible and the Soret region. The spectrum of the unligated derivative is rendered as a solid black line whereas the solid grey line pertains to the econazole saturated spectrum. (D) Ligand binding titration of ferric Yhb with econazole (circles) and miconazole (triangles). Titrations were carried out as described in Materials and methods.

In Yhb, the heme iron is hexagonally ligated and virtually sits in the plane of the porphyrin ring which is reminiscent of the findings in the FHP-azole [5] and single-domain globins [26, 29]. From the proximal side the invariant His85 protrudes towards the porphyrin plane and its Nε2 atom axially ligates to the iron (2.1 Å). His85 together with Glu137 and Tyr95 constitutes the strictly conserved catalytic triad. His85-Nδ1 is hydrogen bonded with Glu137-Oε1 (2.6 Å) and Tyr95-Oη (3.3 Å), and Glu137-Oε2 with Tyr95-Oη (2.6 Å). Surprisingly, the distal heme side of Yhb – the site of the substrate binding pocket – is occupied by an unknown compound that is coordinated to the iron. Its chemical nature could only be estimated on the basis of the electron density profile (Fig. 4) which is most compatible with a three-atom bent compound. In Yhb, the Fe-N(His85) bond and the bond between Fe and the ligating atom of the unknown compound are linear. The three-atom distal ligand is sandwiched between Phe43 and Leu57. One of its protruding atoms (probably an oxygen) is in hydrogen-bond distance to Gln53-Nε1 and Tyr29-Oη both pointing towards the porphyrin plane.

Figure 4.

Heme and FAD binding sites. Electron density map with a sigma cut-off of 1.5 (σ = 1.5). The empty density in the distal heme pocket side indicates the presence of a small ligand. Lys84 sits between the heme propionates and the isoalloxazine methyl groups.

In the FAD binding domain, the binding site of the isoalloxazine ring of FAD (located in front of the C-terminal end of the central β sheet) and its planar conformation are well conserved among the flavoHb family members (Fig. S4). Phe390 is in van der Waals contact with the isoalloxazine ring and thereby shields it against NADH binding as observed for all flavoHbs in the closed state. Despite the high degree of conservation of the heme and FAD binding mode, the positions of the residues between the prosthetic group differ in flavoHbs (Fig. S4).

Catalytic reaction

FlavoHbs confer protection against NO˙ and nitrosative stress by consumption of NO˙ [1] via an NO˙ oxidoreductase, primarily via the NOD reaction. According to biochemical and kinetic studies on NOD [28], O2 is first ligated to the heme iron in the Fe2+ state, thereby activated and then attacked by NO˙ forming nitrate [7]. Enzymatic catalysis of flavoHbs is based on specific structural features on the proximal and distal side of the heme iron that are reminiscent of those of cytochrome P450, peroxidases, cytochrome c oxidases and catalases [30]. Their proximal sides are characterized by the strictly conserved residues His85, Tyr95 and Glu137 (Glu135 in Hmp) that also structurally superimpose almost exactly. Strong hydrogen-bond interactions between them increase the imidazolate character of His85 and thereby stabilize high-valent iron oxidation states and electron-poor Fe-oxygen/nitrogen intermediates [31]. This type of catalytic triad is also found in single-domain globins [26, 32], and in a highly related manner in peroxidases [33].

In contrast, the architecture of the distal heme side in flavoHbs was found to be highly diverse primarily due to the absence/presence and the nature of the chemical compound bound. In Hmp, helix E is attached to the porphyrin ring in front of the iron such that no substrate binding pocket exists [4]. FHP structures are only known in complex with bulky phospholipid or azole inhibitors which bias the size/shape of the substrate binding pocket [3, 5]. The structure of Vitreoscilla single-domain globin in complex with cyanide or water as distal iron ligands draws a more informative picture of the active site with respect to a mechanistic interpretation. However, the flexibility of the N-terminal side of helix E is suggested to be biologically rather irrelevant [32]. The observed geometry of the distal substrate pocket of Yhb in front of the heme iron appears to optimally reflect the active site geometry during the reaction cycle. It largely corresponds to that of the recently reported single-domain globin of Campylobacter jejuni [26] which strongly suggests the catalysis of identical reactions perhaps performed in relatic metabolic contexts. In Yhb, the active site geometry is determined by the unknown distal heme ligand (Fig. 4). The ligand appears to adjust the geometry of the active site after O2 and NO binding and allows an estimation of their binding positions. Assuming a bent triatomic molecule as Fe ligand, the two-atom fragment adopts the position of the substrate O2 and the third atom points to an adjacent hydrophobic pocket – the putative NO˙ binding site (Fig. 5A). The pocket formed by Phe28, Met32, Phe43 and Val98 provides space for fitting a two-atom compound like NO in van der Waals contact to both O atoms of the modeled O2 (Fig. 5A). The terminal oxygen atom of O2 is hydrogen bonded with both Tyr29-OH and Gln53-NE1, which are also hydrogen bonded with each other. Tyr29 and Gln53 are both suitable for stabilizing negative or partially negative charges and their catalytic competence was substantiated by site-directed mutagenesis studies [34, 35]. They are strictly conserved in flavoHbs, Campylobacter globin [26], Vitreoscilla hemoglobin [32] and in ‘truncated’ hemoglobins [36] although the glutamine is partly directed away from the active site. The conserved tyrosine is further found in several non-vertebrate hemoglobins including those from Mycobacterium tuberculosis [37] and Ascaris suum [38].

Figure 5.

Catalytic reaction of Yhb. (A) Potential O2 and NO binding sites. Based on the geometry of the active site and the electron density profile of the unknown covalent heme iron adduct the position of O2 and NO could be estimated. In this model, the diatomic O2 (yellow) is oriented towards Gln53 and Tyr29. The NO˙ binding site is positioned in the hydrophobic cavity surrounded by Phe28, Tyr29, Leu57, Val98, Leu102 and Tyr126. (B) Proposed catalytic mechanism. The postulated mechanism combines previous data with the structural information of the Yhb structure.

The catalytic mechanism of O2 cleavage of many hemoenzymes is based on the ‘push–pull’ process originally postulated for peroxidases [30] in which the electron-rich proximal ligand (for flavoHbs the histidine imidazolate) pushes electrons to the heme Fe(II)-O2 species and hydrogen-bond donors of the polypeptide stabilize the terminal oxygen [39]. Both effects weaken the oxygen–oxygen bond. The postulated catalytic mechanism outlined in Fig. 5B is initiated by transferring one electron (two sequentially) from NADH via FAD to heme b. In the generated heme-Fe(II) state, O2 can be strongly bound and subsequently activated to a superoxo Fe(III)-O-O˙ species as demonstrated for several hemoenzymes. Spectroscopic investigations on Yhb in solution after addition of NADH under aerobic conditions (Fig. 3A,B) give rise to a Soret peak at 414 nm characteristic for an Fe(II)-O2 adduct and to α- and β-bands at 545 and 580 nm, respectively, indicating that heme reduction occurs via FAD. Interestingly, the Gln53 amide oxygen is linked via a solvent molecule to Lys84-Nζ that bridges FAD and heme b as described. A coupling between the electron transfer and the catalytic process is therefore in principle feasible (Fig. S5). After O2 ligation, NO˙ binds to the predicted binding site (Fig. 5A), and reacts with the activated Fe(II)–O2 species to nitrate most likely via a peroxynitrite (OONO) intermediate. An Fe(III)–OONO intermediate experimentally detected in myoglobin and hemoglobin [40, 41] is isomerized to nitrate by a homolytic cleavage via radical intermediates inline image or by a heterolytic cleavage via ionic intermediates (inline image) or by a concerted internal O-atom rearrangement via bending and vibration of the Fe(III)–O–ONO2 bond [7]. According to the geometric constraints of the substrate binding pocket the nitrogen atom of NO˙ is located between the two oxygen atoms of the Fe(II)–O2 species which especially attract the concerted mechanism.

Econazole binding site and conformational change of Yhb upon econazole binding

The X-ray structure of Yhb in complex with the econazole inhibitor (YhbE; see Table 1) has been determined on the basis of co-crystallization experiments between Yhb and econazole. Econazole consists of a central asymmetric carbon atom linked to the aromatic substituents imidazole, monochlorophenoxy and dichlorophenyl. YhbE contains one econazole per flavoHb molecule embedded in the substrate binding pocket (Fig. 6A). The imidazole ring of econazole placed in the pocket center is perpendicularly oriented to the porphyrin plane. The imidazole nitrogen is axially ligated to the heme iron (bond length 2.1 Å) and linearly prolongs the Fe–His85–Nε1 bond as was previously reported for FHP-azole structures [5]. Consistently, the UV–visible spectra recorded in solution from Yhb agree with those from Hmp and FHP [4, 5, 14] indicating the formation of a hexacoordinated low spin heme iron due to imidazole ligation. Spectroscopic experiments on Yhb were performed by titrating the ferric heme with increasing concentrations of econazole (and miconazole). The absorption differences between the ligand-free (403 nm) and the ligated (414 nm) heme are followed at their Soret band (Fig. 3C).

Figure 6.

(A) Econazole binding to Yhb. The superimposed globin domain in the ligand-free (grey) and econazole-bound (light blue) Yhb indicates large-scale conformational changes upon binding of the bulky econazole (orange). The electron density at econazole is shown beyond 1σ. (B) Differences in econazole binding in Yhb and FHP. Superposition of the econazole binding site indicates different conformations of the inhibitor although the contacting residues are largely conserved. The major differences are the selective presence of a lipid fragment and the more expanded outer section of FHPE (grey) compared with YhbE (light blue).

A comparison between the Yhb and YhbE structures reveals that binding of the bulky econazole leads to an increase of the substrate binding pocket (Fig. 6A) affecting several helices of the globin domain (but not the ferredoxin reductase part). The large-scale induced-fit process implies a shift of helix B of ~ 4 Å away from heme b to evade the dichlorophenyl group pointing to it. This shift is propagated to the C-terminal end of helix G that also moves ~ 2 Å. Helix E evades econazole in the same direction and order as helix B and thereby becomes highly flexible but, interestingly, without inducing any significant changes in the helical conformation of loop D. Helices A, F and H are only slightly affected by econazole binding. As a result Phe28 and Leu57 in van der Waals contact with each other in Yhb (3.6 Å) sandwich the dichlorophenyl ring in YhbE, the distance between them increasing to 7.3 Å. The partly preformed binding site of the monochlorophenoxy ring is enlarged by conformational changes of side chains of Val61, Leu102, Trp122 and Tyr126 in an order of 2.5 Å. The catalytic residues Tyr29 and Gln53 are displaced about 2.1 Å and 3.8 Å, respectively, away from the substrate binding pocket.

Structural determinants for differential econazole binding in YhbE and FHPE

As can be seen from Fig. 6B, the binding of econazole is substantially different between Yhb and FHP. In YhbE, the monochlorophenoxy substituent of econazole sits at the bottom of the inner section at the same position as the dichlorophenoxy substituent in the FHP–miconazole complex [5]; in FHPE this position is occupied by a lipid fragment (Fig. 6B).The monochlorophenoxy group in FHPE moves to the upper subsite of the inner section which is occupied by the dichlorophenyl group in YhbE. The angle between the chlorinated aromatic rings is ~ 70° in YhbE and ~ 110° in FHPE. Indiscernible in the absorption spectra, the imidazole rings in YhbE and FHPE structures are rotated relative to each other around an axis vertical to the heme plane by ~ 45°. The different conformations of econazole in Yhb and FHP demonstrate that its conformational flexibility is exploited for binding optimization. The structures of the substrate binding pocket of YhbE and FHPE substantially differ, which is unexpected as the same inhibitor is bound into a pocket coated by highly conserved residues. Obviously, its size/shape is significantly influenced by less conserved regions far away from it. One distinct feature is the loop D adopting different conformations in YhbE and FHPE despite similar primary structures (Fig. S1). However, a few amino acid exchanges might be important. For example, the substitution of Ala52 and Pro54 in Yhb to glutamines in FHP at the N-terminal side of helix E influences the conformation of loop D; Ala52 of YhbE would directly interfere with His47 of loop D of FHPE. Because of the more compact arrangement of loop D in Yhb than in FHP, its protruding side chains and helix E are positioned closer to heme b. The substrate binding pocket shrinks in this region and is consequently not occupied by econazole in YhbE. A second crucial feature is the binding of a phospholipid into FHPE but not YhbE verified by thin layer chromatographic experiments (Fig. S5 and [42]). The affinity for lipid binding appears to be significantly higher in FHP than in Yhb. Tyr64 and Val77 in FHP promote lipid binding whereas Thr60, Leu74 and Met394 in Yhb are directed towards the lipid binding site and prevent its binding. The binding of the phospholipid in FHPE and the monochlorophenoxy substituent in YhbE results in different positions of the C-terminal side of helix E that, in turn, influence the N-terminal side of helix E and the loop D conformation.

The organism-specific structural differences among the flavoHbs rationalize the different apparent inhibition constants of azole compounds. Econazole has a significantly weaker inhibition for Yhb (Ki ~ 30 000 nm) than for HMP (Ki ~ 550 nm) and FHP (Ki ~ 100–2000 nm) [14]. The difference in inhibition potency is even more pronounced for miconazole (Yhb, Ki ~ 12 000 nm; HMP, Ki ~ 80 nm; FHP, Ki ~ 5–650 nm). However, these differences are not reflected in the Kd values determined from the titration curves of the ferric heme from Yhb with increasing concentration of azole derivatives. The sigmoid shape of econazole and miconazole saturation curves were best fitted to the Hill equation with Kd values of 1.56 ± 0.13 μm and 2.56 ± 0.28 μm in Yhb (Fig. 3D) which is also not far from the Kd value determined for miconazole (1.3 ± 0.1 μm) in FHP [5]. Moreover the observed structural differences open the opportunity to develop specific inhibitors for flavoHbs from different organisms. Because the profile of the substrate binding pocket is not predominantly determined by local amino acid exchanges their design is a challenging task.

Materials and methods

Flavohemoglobin expression and purification

Yhb was overproduced in E. coli strain BL21 (DE3) harboring the plasmid pRSET 6A as described previously [42]. 100 mL precultures of the E. coli BL21 strain were grown aerobically at 37 °C in Luria–Bertani medium supplemented with ampicillin (100 μg·mL−1) until cells reached an optimal density at 600 nm of 0.5. These cultures were used to inoculate 2 L of Terrific Broth containing 100 mg·mL−1 ampicillin in 4 L flasks (1/100 v/v). Bacteria were grown at 30 °C overnight without induction with isopropyl thio-β-d-galactoside, and shaken continuously at 150 rpm.

The cells were harvested by centrifugation (6200 g for 15 min at 4 °C) and resuspended in 20 mm potassium phosphate pH 7.5 (buffer A) with DNase. After cell disruption, the soluble extract was separated by ultracentrifugation at 75 000 g for 50 min at 4 °C and submitted to a first precipitation with 25% ammonium sulfate. The supernatant containing the Yhb was submitted to a second precipitation with 12% ammonium sulfate. The pellet was resuspended in a minimum volume of buffer A, dialyzed overnight at 4 °C and then loaded onto an anion-exchange column of DEAE-Sepharose previously equilibrated with buffer A. The elution was performed with 500 mL of a linear gradient from 0% to 30% KCl 1 m. Yhb fractions were pooled, concentrated and applied onto a gel filtration column of Superdex 75 equilibrated with buffer B (Tris/HCl 20 mm, KCl 10 mm, pH 7.5). Fractions of pure Yhb were pooled and subjected to 10% bistris Nupage SDS gel electrophoresis stained with Coomassie Brilliant Blue. Protein concentration was assayed by the bicinchoninic acid method [43] with BSA as standard and spectrophotometrically using an extinction coefficient at 395 nm of 63 000 m−1·cm−1, experimentally determined.

Spectrophotometric titration of the ligand binding

The formation of econazole–Yhb and miconazole–Yhb complexes was assessed by monitoring the increase of the high spin form of the protein upon the addition of the azole inhibitors using visible absorption spectroscopy. The measurements were performed at 21 °C on a double beam Uvikon 943 spectrophotometer (Kontron Instruments, Milan, Italy) using, routinely, a solution of 7 μm Yhb in buffer A filled in a 1.0 cm light-path quartz cuvette. Increasing amounts of the azole compound dissolved in dimethylsulfoxide were added sequentially to the sample cuvette. Spectra were measured after 3 min incubation time. Dimethylsulfoxide was present in the cuvette at a final concentration of < 1% (v/v) as described by Helmick et al. [14]. The reference cuvette contained an equivalent amount of the added drug in the same buffer to avoid a diffusion effect. The binding mode of the drug to Yhb was analyzed in terms of Hill plots [44] as previously described [5]. All reagents were purchased from Sigma Aldrich unless specifically mentioned.

Crystallization of Yhb and Yhb–econazole complex

Crystals of Yhb were grown by the sitting drop method at 18 °C. Droplets were prepared by mixing 2 μL of reservoir solution and 2 μL of protein solution consisting of 15 mg·mL−1 Yhb. The reservoir solution contained 2 m ammonium sulfate and 0.1 m Tris/HCl pH 8.5. To determine the structure of the Yhb–econazole complex, protein and inhibitor were co-crystallized with the sitting drop method at 4 °C. The optimal crystallization condition contained 2 μL of 40 mg·mL−1 Yhb and 0.5 mm econazole mixed with 2 μL of precipitant solution containing 16% PEG 4000, 0.1 m Tris/HCl pH 8.5 and 0.2 m lithium sulfate. Within 2 weeks, brownish colored crystals formed reproducibly and were immersed in mother liquid supplemented with 25% glycerol (w/v) as cryoprotectant, before being mounted on a nylon loop and flash-cooled in liquid nitrogen.

Data collection and refinement

Data for Yhb were collected from a single crystal at the European Synchrotron Radiation Facility-ID29 beamline, Grenoble, and processed with hkl [45]. Phases were determined with the single anomalous dispersion method by using heme iron as anomalous scatterer. The position of the iron was calculated with shelxd [46], the phases with sharp (sharp-develop@GlobalPhasing.com). Data for YhbE were measured at the Swiss-Light-Source-PXII beamline (Villigen, Switzerland) and processed with xds [47]. Phases were determined with the molecular replacement program epmr [39] using Yhb as a search model. Structure refinement was performed with refmac of the ccp4 program suite [48]. The model was built with coot [49]. Data collection and final refinement statistics are given in Table 1. Structural images were generated using pymol software (www.pymol.org). Structure factors and final coordinates of Yhb and Yhb–econazole are deposited in the PDB with code 4G1V and 4G1B.

Acknowledgements

We thank Hartmut Michel for continuous support and the staff at the ESRF and SLS for help during data collection for help during data collection. This work was supported by the collaborative French and Tunisian grant (PHC-CMCU Utique ref: 08G0801), CNRS, University Paris-Sud and the Max-Planck-Society.

Ancillary