Hydrogen peroxide‐mediated conversion of coproheme to heme b by HemQ—lessons from the first crystal structure and kinetic studies

Heme biosynthesis in Gram‐positive bacteria follows a recently described coproporphyrin‐dependent pathway with HemQ catalyzing the decarboxylation of coproheme to heme b. Here we present the first crystal structure of a HemQ (homopentameric coproheme‐HemQ from Listeria monocytogenes) at 1.69 Å resolution and the conversion of coproheme to heme b followed by UV‐vis and resonance Raman spectroscopy as well as mass spectrometry. The ferric five‐coordinated coproheme iron of HemQ is weakly bound by a neutral proximal histidine H174. In the crystal structure of the resting state, the distal Q187 (conserved in Firmicutes HemQ) is H‐bonded with propionate p2 and the hydrophobic distal cavity lacks solvent water molecules. Two H2O2 molecules are shown to be necessary for decarboxylation of the propionates p2 and p4, thereby forming the corresponding vinyl groups of heme b. The overall reaction is relatively slow (k cat/KM = 1.8 × 102 m −1·s−1 at pH 7.0) and occurs in a stepwise manner with a three‐propionate intermediate. We present the noncovalent interactions between coproheme and the protein and propose a two‐step reaction mechanism. Furthermore, the structure of coproheme‐HemQ is compared to that of the phylogenetically related heme b‐containing chlorite dismutases. Database Structural data are available in the PDB under the accession number 5LOQ.


Introduction
Heme biosynthesis in monoderm (Gram-positive) bacteria follows a novel coproporphyrin-dependent pathway described recently by Dailey and coworkers [1]. The missing link in understanding this noncanonical bacterial heme biosynthesis pathway was HemQ which was shown to catalyze the decarboxylation of two propionate groups at positions 2 (p2) and 4 (p4) of coproporphyrin III (coproheme), thereby producing protoporphyrin IX (heme b) [2]. Coproheme seems to act as both substrate and redox-active cofactor as its conversion is triggered by the addition of FMN, hydrogen peroxide, or peracetic acid [1,3,4]. However, no structural data of coproheme-HemQ are available and the mechanism of the two decarboxylation steps is unknown.
Many Firmicutes and Actinobacteria are pathogens and a major threat to public health, due to the development of multiple resistance to common antibiotics [5,6]. Their steady-state iron uptake, their heme biosynthesis, heme degradation, and iron release is crucial for the viability of these bacterial organisms and depend strongly on HemQ [7], the latter might be a promising therapeutic target in the future. However, this requires a detailed knowledge of structure-function relationships that at present is lacking; only apostructures of HemQs are available in the Protein Data Bank (www.pdb.org) [8]. The interaction of apo-HemQ from Listeria monocytogenes with coproheme was previously evaluated by means of molecular dynamics simulations of a coproheme-bound model of LmHemQ, based on the apo-structure (4WWS), suggesting hydrogen bonds between Y147 and S225 and p4 as well as between Q187 and p2 [9]. Coproheme binding to apo-HemQ was shown to be biphasic resulting in the formation of mainly high-spin ferric protein with a standard reduction potential of the Fe (III)/Fe(II) couple of À205 mV at pH 7.0 [9]. Celis et al. and Lobo et al. reported that ferric coproheme-HemQ needs an excess of hydrogen peroxide to mediate the two decarboxylation steps of coproheme to heme b and that under such conditions, heme bleaching and destruction occurred [3,4]. Additionally, it was demonstrated that the conversion proceeds stepwise yielding an intermediate three-propionate monovinyl, monopropionate deuteroheme isomer III [3].
Herein we report the first coproheme-bound HemQ structure at 1. 69 A resolution, which gives valuable information about the interaction of the substrate and redox-cofactor coproheme with the active site, including coproheme coordination and noncovalent interactions of the four propionates with the protein matrix.
Comparison between apo-LmHemQ and coproheme-LmHemQ structures (HemQ from L. monocytogenes) allows identification of the amino acid residues likely involved in catalysis as well as in a gating function in the course of substrate binding. Additionally, by UVvis and electronic circular dichroism (ECD) spectroscopy as well as mass spectrometry, we demonstrate a clear two-step kinetics of the conversion of coproheme to heme b as well as a H 2 O 2 to coproheme ratio of 2 : 1. The presence of excess of H 2 O 2 leads to the irreversible formation of heme-protein linkages and heme bleaching. The data will be discussed with regard to the proposed physiological role of HemQ in heme biosynthesis of Gram-positive bacteria and possible reaction mechanism(s). In addition, they are compared to the phylogenetically related chlorite dismutases (Clds) and dye-decolorizing peroxidases (DyPs) [10][11][12] which share a similar overall fold but exhibit completely different enzymatic properties and display distinct differences in the architecture and amino acid composition of the proximal and distal sides of the heme cavity [13].

Crystal structure of coproheme-LmHemQ
In this work, we heterologously expressed and purified apo-HemQ from L. monocytogenes (LmHemQ) as described previously [8]. The enzyme was reconstituted by the addition of equimolar concentrations of coproheme. Coproheme-LmHemQ (7.5 mgÁmL À1 ) was crystallized in 0.1 M bicine pH 9.0, 10% w/v PEG6000, and diffracted to a resolution of 1.69 A. Crystals were pale red.
Coproheme-LmHemQ crystallized as a pentamer in space group P12 1 1, with unit cell dimensions of 77.69 A 9 129.37 A 9 77.91 A. The final model (PDB code: 5LOQ) consists of 9801 protein atoms, 713 water molecules, has good stereochemistry, and Rfactors ( Table 1). The Wilson temperature factor is 26.8 A 2 . A clear electron density was observed for residues 4-251 of subunit A; for subunit B, 6-111 and 125-251; for subunit C, from 3 to 113 and 126 to 251; for subunit D, from 2 to 251; and for subunit E, from 6 to 113 and 125 to 251. The missing amino acids are part of a loop from S111 to N123. The weak electron density in this region indicates flexibility of this loop. In chain A and D, the loop is fully resolved due to crystal packing contacts.
Apo-LmHemQ (PDB code: 4WWS) crystallizes in the same oligomeric state and space group with similar unit cell dimensions (Fig. 1A). Subunit A of apo- LmHemQ [8] and coproheme-LmHemQ (with flexible loop S111-N123 resolved in both structures) superimpose with a root-mean-square deviation (RMSD) of 0.223 A over 225 Ca atoms. A noteworthy difference between the structures of apo-LmHemQ and coproheme-LmHemQ concerns the orientation of two arginine side chains (R133, R179) at the substrate access channel. Arginine 133 of apo-LmHemQ and copro-hemeQ exhibit the same conformation in chain A but different conformations in the other subunits. The flexibility of this side chain is also supported by weak electron densities. R179 shows different conformations in the five subunits and thus no distinct electron density can be assigned (due to crystal contacts, only R179 of chain B shows a distinct electron density) (Fig. 1B,C).
Additionally, the oligomeric assembly of coproheme-LmHemQ shows further structural heterogeneities of heme cavity residues and the substrate/cofactor within the five subunits. All coprohemes are anchored by a proximal histidine residue (H174) with coproheme iron distances between 2.4 and 3.0 A. Comparison with the structures of the phylogenetically related wild-type and mutant chlorite dismutases (Clds) shows that the distances between the N e of the imidazole of the proximal His and the iron ion are shorter (Table 2). Both the longer bond length and the high variability of distances in coproheme-HemQ suggest that the proximal histidine is not part of a hydrogen bond network as seen in the related Clds and DyPs. Note that the a-helix (M164-Y181) harbors both H174 as well as the (flexible) R179 residue (Fig. 1C).
The amino acids of the distal side of the coproheme-LmHemQ complex are Q187 (which aligns with the catalytic arginine of chlorite dismutases), I189, and V202; distances to the heme iron are 5.2 A (Q187), 4.3 A (I189), and 6.4 A (V202). Q187 is oriented toward the protein surface, pointing away from the coproheme iron in coproheme-LmHemQ ( Fig. 2A). Interestingly, no electron densities of potential solvent molecules are present on the distal side (Fig. 2B).
Comparing the coprohemes of all five subunits clearly demonstrates higher flexibility of those propionates that are closer to the protein surface, namely p2 and p7. The conformational freedom of p4 and p6 is much more restricted (Fig. 3). Note that upon conversion of coproheme to heme b, p2 and p4 are eliminated. H-bond analyses of the catalytically relevant p2 and p4 clearly identify Y147, M149, S225 as H-bonding partners for p4, and G178, V185, Q187 as H-bonding partners for p2 (Fig. 3). The interactions of Y147 and S225 with p4 and Q187 with p2 were previously correctly predicted in silico using MD simulations [9].
The interaction of p6 and p7 (which are not cleaved off by HemQ and therefore are also present in the heme b of structurally related chlorite dismutases) with the protein moiety is shown in Fig. 4A, where residues within a distance of 3 A to p6 or p7 are highlighted, not necessarily building H-bonds with the propionates (Fig. 4B,C). In LmHemQ and Clds, conserved residues interact with p6 and p7 (Y113 for p7; K151, W159 for p6), and also residues unique for HemQ are in close proximity to the two propionates. The most prominent one is R133, which potentially forms H-bonds with p7 and is located at the substrate entrance channel; this arginine aligns with alanine, leucine, or valine of Clds.
As described above, in the structure of coproheme-LmHemQ, only in chain A and chain D, a flexible loop between L114 and Y125 could be completely built, because of reliable electron densities due to crystal packing contacts. In the apo-structure of LmHemQ (PDB code: 4WWS) [8], the loop was only resolved in chain A and E, exhibiting the same conformation as in coproheme-LmHemQ, probably because this conformation was stabilized by the crystal packing contacts. MD simulations confirmed the flexibility of this loop [9]. Interestingly, this loop is only present in HemQs from Firmicutes and not from Actinobacteria or in Cld/HemQ-like proteins from Archaea (alignment shows the absence of 10 residues) [14].
UV-vis and resonance Raman spectroscopy of ferric and ferrous coproheme-LmHemQ and heme b-LmHemQ The spectroscopic signatures of LmHemQ are comparable to those reported for HemQ from Staphylococcus aureus (SaHemQ) [3]. The UV-vis spectrum of ferric coproheme-LmHemQ shows a Soret band at 395 nm (shoulder at 375 nm) and a very strong CT1 band at 630 nm typical of high-spin (HS) heme (Fig. 5A). Considering the overall blue-shift of the spectrum compared to that of other heme proteins due to the lack of vinyl conjugation, the electronic absorption spectrum appears typically five-coordinated (5c). The UV-vis spectrum of ferric heme b-LmHemQ (produced by addition of hydrogen peroxide to coproheme-HemQ) shows the Soret maximum at 410 nm and a weak CT1 at 640 nm, typical of a six-coordinated lowspin (6cLS) heme with a minor HS component (Fig. 5A). Comparison of the low frequency region of the RR spectra of coproheme-LmHemQ and heme b-LmHemQ did not show any significant intensity change in the propionyl bending modes, assigned to the broad bands at 376 and 392 cm À1 (Fig. 5B). Also in case of HemQ from S. aureus (SaHemQ) [3], changes in intensity of the propionyl bending modes (at 378 and 390 cm À1 ) are evident only by inspection of the difference spectra between heme b-LmHemQ and coproheme-LmHemQ. By contrast, the formation of the vinyl substituents can be clearly observed. The spectra in polarized light allowed us to assign the m (C=C) stretching modes (polarized) at 1621 and 1632 cm À1 (Fig. 4B, inset). Moreover, in the low frequency region, the d(C b C a C b ) vinyl modes are observed as a single band at 418 cm À1 (Fig. 5B), similar to that observed at 416 cm À1 for SaHemQ [3].
The UV-vis spectrum of ferrous coproheme-LmHemQ is characteristic of a predominant 5cHS species in the absence of vinyl conjugation (Soret at 423 nm and maximum at 548 nm), whereas ferrous heme b-LmHemQ is a mixture of 6cLS (Soret at 423 nm) and 5cHS (Soret at 437 nm) species (Fig. 5C). Upon conversion of coproheme-LmHemQ to heme b-LmHemQ, two new bands at 530 and 555 nm appear (at the expense of the band at 548 nm), consistent with the significantly increased amount of LS species. For both proteins, the m(Fe-Im) stretching mode is observed at 214 cm À1 (Fig. 5D). This is identical to that reported for SaHemQ [3], suggesting that the N e of the imidazole of the proximal His (H174) is weakly bonded to the heme iron in agreement with the crystal structure and the interatomic distances (2.4-3.0 A). This finding clearly demonstrates that conversion of p2 and p4 to the respective vinyl groups has no impact on the proximal heme coordination and architecture.

Kinetics of conversion of coproheme to heme b
Upon addition of hydrogen peroxide to coproheme-LmHemQ, decarboxylation of the propionate substituents of coproheme leads to the formation of heme b-LmHemQ. During this reaction, the original spectrum of coproheme-LmHemQ (Soret maximum: 395 nm) converts to a spectrum with the characteristics of heme b-HemQ (Soret maximum: 410 nm) with an isosbestic point at 400 nm ( Fig. 6A) [3,8]. When a 10-fold excess of H 2 O 2 is used (5 lM coproheme-HemQ, 50 lM H 2 O 2 ), the reaction is completed after 140 s (Fig. 6A, inset).
To obtain catalytic parameters for this H 2 O 2mediated reaction, initial velocities (v 0 ) were obtained by monitoring the absorbance increase at 410 nm. The observed kinetic behavior is shown in Fig. 6 and the apparent K M value for H 2 O 2 was calculated to be 37.1 lM (Fig. 6B). The turnover number is very low (~0.0065 s À1 ) resulting in a k cat /K M value of approximately 1.8 9 10 2 M À1 Ás À1 (Fig. 6B) at pH 7.0. In contrast to SaHemQ, heme bleaching is not observed within the investigated initial reaction phase (~50 s) ( Fig. 6A) [3]. Coproheme concentration could not be varied, as free coproheme (if present in excess) is bleached by H 2 O 2 .

Stoichiometry of the H 2 O 2 -driven decarboxylation reactions
The absence of heme bleaching allowed the determination of the H 2 O 2 to coproheme stoichiometry for this reaction. Coproheme-LmHemQ was titrated with subequimolar concentrations of H 2 O 2 followed by ECD spectroscopy (Fig. 7A), UV-vis spectroscopy (Fig. 8A, B), and mass spectrometry (Fig. 8C,D).
The ECD spectrum of coproheme-LmHemQ exhibits a minimum at 388 nm and a maximum at around 417 nm (Fig. 7A, black line). Upon reaction with 0.3 eq H 2 O 2 , the ellipticity at 417 nm decreases and a prominent minimum at 391 nm is formed (Fig. 7A, bold green line). At increasing hydrogen peroxide, the ellipticity of this minimum continuously decreases and red-shifts. At 2 eq H 2 O 2 , the final spectrum with a  (Fig. 7A, inset). The ellipticities initially decreased and subsequently increased again showing turning points just below 0.5 equivalents (388 nm) and 1.0 equivalents (410 nm) of H 2 O 2 to coproheme-LmHemQ. The reaction was completed after reaching a twofold stoichiometric excess of hydrogen peroxide. The stepwise conversion of coproheme to heme b is also supported by following the change in ellipticity at 395 nm with time (Fig. 7B). Time traces exhibit a clear biphasic behavior, where the first (single exponential) phase represents the formation of the intermediate and the second, slower (single exponential) phase represents the formation of the product heme b. Apparent bimolecular rates for the first and the second transitions were approximated to be 7.2 AE 0.6 9 10 2 M À1 Ás À1 and 1.6 AE 0.2 9 10 2 M À1 Ás À1 , respectively (Fig. 7B). These values are in good agreement with the determined steady-state parameter k cat /K M of approximately 1.8 9 10 2 M À1 Ás À1 at pH 7.0 (Fig. 6B) for the overall reaction.
Next, the stoichiometry of the reaction was followed by UV-vis spectroscopy (Fig. 8A). The normalized absorbances at 375 and 395 nm (Soret maximum of coproheme-HemQ, shoulder at 375 nm) as well as 410 nm (Soret maximum of heme b-HemQ) from these studies were plotted against the stoichiometric ratio of H 2 O 2 over LmHemQ. Fitting with a sigmoidal function reveals midpoints of conversion at [H 2 O 2 ]/ LmHemQ ratios between 0.6854 and 0.7761. The reaction is complete after reaching a twofold stoichiometric excess of H 2 O 2 (Fig. 8B).
Finally, during the UV-vis titration, small aliquots were taken for mass spectrometric analysis at each     (Fig. 8D). This clearly demonstrates that the reaction occurs in a stepwise manner with a three-propionate species as intermediate. LmHemQ numberi ng  100  110  120  130  135 pdb-code Chain 5LOQ Listeria monocytogenes Listeria monocytogenes  However, the mass spectrometric data did not allow to distinguish between monovinyl, monopropionate deuteroheme isomer III or IV (same molar mass).
Mass spectrometric analysis of the titration of coproheme-LmHemQ with H 2 O 2 also included whole protein analysis of the samples. The obtained data clearly showed cross-linking of the heme group with the protein, starting at approximately equimolar H 2 O 2 concentrations (Fig. 8E,F). After addition of 40 lM H 2 O 2 to 18 lM coproheme-LmHemQ, 23.8% of the protein was covalently cross-linked to a modified prosthetic group (Fig. 8E,F).

Discussion
It has been recently discovered that HemQ [2] is a key element in the heme biosynthetic pathway of monoderm bacteria [1]. Coproheme was identified as the substrate which is decarboxylated at p2 and p4 to form heme b. While the physiological role of HemQ now seems to be elucidated (although several other roles have been proposed), the knowledge regarding structure-function relationships was limited until now. The catalytic efficiency of the proposed reaction is rather low but this is reasonable in a biological context. The apparent k cat /K M value for HemQ from L. monocytogenes for H 2 O 2 was determined to be 1.8 9 10 2 M À1 Ás À1 , very similar to that reported for HemQ from S. aureus (SaHemQ) [3]. In vivo in the Gram-positive bacterial cell, H 2 O 2 most likely is not present in (high) micromolar concentrations, but is continuously provided by HemY and consumed by coproheme-HemQ [1,14]. This consequently means that the conditions used to determine Michaelis-Menten parameters (Fig. 6) do not reflect in vivo conditions. This assumption is also supported by the fact that coproheme-HemQ is very susceptible to irreversible inhibition by modification of the substrate/cofactor and the protein in the presence of a stoichiometric excess of H 2 O 2 . Consequently, the H 2 O 2 -mediated decarboxylation reaction was studied under subequimolar concentrations of H 2 O 2 to minimize side reactions. In this way, it was possible to demonstrate that complete conversion of 1 mol coproheme to 1 mol heme b requires 2 mol of hydrogen peroxide. The reaction occurs in a stepwise manner with the formation of a threepropionate intermediate (Figs 7 and 8) and needs Fe-coproheme for catalysis, as Ni-coproheme-HemQ displays no enzymatic activity [1]. This observation suggests that coproheme acts as both substrate and redox-cofactor. Furthermore, it suggests that hydrogen peroxide mediates the oxidation of ferric coproheme-HemQ to a reactive (electron-deficient) compound I-like state, as observed in heme peroxidases and catalases [i.e., oxoiron(IV) in combination with either a porphyryl radical (Por •+ ), (i.e., Compound I) or with an amino acid radical (aa •+ ), (i.e., Compound I*]. This redox intermediate initiates a radical mechanism which restores the Fe(III) state and converts the propionyl to the vinyl substituent thereby releasing carbon dioxide. To affect the decarboxylation of p2 and p4, two reaction cycles and thus two H 2 O 2 molecules are necessary. The detection of the three-propionate intermediate, as well as the spectral changes observed during turnover by ECD spectroscopy (which show biphasic behavior, including a turning point in ellipticity) suggest that these two decarboxylation steps occur consecutively (Figs 7  and 8).
Both the UV-vis and RR spectra as well as the standard reduction potential of the Fe(III)/Fe(II) couple of coproheme-LmHemQ of À205 mV [9] demonstrate that the enzyme is in the ferric state under physiological conditions. The crystal structure of coproheme-LmHemQ together with the RR data suggest a relatively weak interaction between the heme iron and N e of the proximal ligand H174. Due to the absence of an H-bonding partner, the anionic character of H174 must be significantly less pronounced compared to the phylogenetically related Clds. Another important difference between Cld and coproheme-HemQ concerns the coordination of the heme iron at the distal side. Interestingly, the crystal structure of coproheme-LmHemQ (in contrast to Clds) shows no electron densities of potential solvent molecules, which is also reflected by temperature-dependent investigations of the standard reduction potential of the Fe(III)/Fe(II) couple of coproheme-SaHemQ [9]. This study revealed a highly negative DS°0 rc,solv value, which indicates a reduction-induced increase in solvent ordering, which is typical for a system with a buried metal site and limited solvent accessibility to the distal heme cavity [9]. By contrast, in Clds reduction of the ferric to the ferrous state is entropically favored [15]. Limited accessibility is also reflected by cyanide binding rates that are two orders of magnitude slower compared to Clds [9], and it is reasonable to assume that this restricted access to the distal heme cavity concerns also the substrate hydrogen peroxide. How H 2 O 2 oxidizes coproheme-HemQ for production of a redox intermediate that initiates the oxidative decarboxylation reaction is unknown. In contrast to typical heme peroxidases of the peroxidase-catalase and peroxidase-cyclooxygenase superfamilies [10] and catalases [16], which typically provide a distal His-Arg or His-Asn pair for heterolytic cleavage of H 2 O 2 and Compound I formation, in coproheme-HemQ comparable catalytic residues are absent. The distal amino acids within 6 A distance from the heme iron are Q187, I189, and V202, respectively. Glutamine 187, which aligns with the catalytic arginine in Clds, is fully conserved in HemQs from Firmicutes but is exchanged by alanine in (putative) HemQs from Actinobacteria.
In the crystal structure of coproheme-LmHemQ, Q187 is in the 'out' conformation (distance between Fe and amide nitrogen: 5.2 A) making a hydrogen bond with the carboxyl group of p2. According to MD simulations [9], this H-bond is formed during~80% of the 30 ns simulation time [9]. By contrast in the structure  of apo-LmHemQ, Q187 is in the 'in conformation' (distance between Fe and amide nitrogen: 3.5 A). This suggests that in heme b-LmHemQ, Q187 the 'in' conformation is possible (due to the absence of p2) which might contribute to the very pronounced LS character of heme b-LmHemQ compared to coproheme-LmHemQ.
The hydrophobic nature of the distal heme cavity, the absence of a catalytic base for deprotonation of H 2 O 2 , the absence of distal water molecules, and thus of an obvious pathway for proton delivery or elimination, suggest that formation of the initial reaction product [i.e., enzyme-peroxide complex, Fe(III)-O-O-H] must be extremely slow compared to heme peroxidases or heme catalases and could be the rate-limiting step of the enzyme turnover. In the classic heme biosynthesis pathway of diderm (Gram-negative) bacteria, HemN decarboxylates coproporphyrinogen III (no iron inocorporated) to protoporphyrinogen IX. It utilizes a [4Fe-4S] cluster cofactor and SAM-dependent mechanism to cleave off propionates at positions 2 and 4 [17]. In this reaction, SAM is cleaved to methionine and a highly oxidizing 5 0 -deoxyadenosyl radical, which subsequently abstracts a hydrogen atom from the b-C atom of the propionate side chain. The formation of the propionyl radical initiates the decarboxylation reaction and, finally, after a second one-electron oxidation step, the vinyl group is formed [17]. Except for the formation of the propionyl radical, coproheme-HemQ could follow the same mechanism. The conversion of 1 mol coproheme to 1 mol heme b starts with the oxidation of the resting state by H 2 O 2 . Compound I*, [Fe(IV) = O CoproPorpropionate • ) is built by one of the two homolytic pathways described above (Reaction 1). Finally, by intramolecular transfer of the unpaired electron to the iron and decarboxylation, the ferric state is restored and the vinyl group is formed (Reaction 2 The involvement of a (reactive) Compound I-like intermediate in this reaction is also supported by the fact that in the presence of an excess of H 2 O 2 (after complete conversion of p2 and p4 to the respective vinyl substituents) formation of a crosslink between the cofactor (i.e., reaction product heme b) and the protein occurs (Fig. 8E,F). Formation of covalent heme-protein bonds is known from mammalian [18] and bacterial heme peroxidases [19,20] and typically follows an autocatalytic radical mechanism. This observation is most probably an in vitro artifact and not relevant in vivo, where heme b is easily released and taken up by an apo-protein [9]. Nevertheless, it underlines that the oxidized state of cofactor-bound HemQ is highly reactive.
The open question remains why p2 and p4 (and not p6 and p7) of coproheme are decarboxylated. The first crystal structure of coproheme-HemQ of L. monocytogenes now allows to critically investigate the environment of these substituents. Interestingly, both p6 and p7 form H-bonds with the positively charged amino acids K151 and R133 (Fig. 4), which impairs the formation of propionyl radicals. By contrast, p2 interacts with (distal side) Q187 (conserved in all Firmicutes, but is exchanged by alanine in Actinobacteria), V185 (mostly conserved in both phylae), and G178 (conserved in Firmicutes, but exchanged by alanine in Actinobacteria) [9]. Residues interacting with p4 are Y147, M149, and S225, with Y147 being fully conserved in all Firmicutes and Actinobacteria. Methionine 149 is present in Firmicutes and aligns with phenylalanine in Actinobacteria. Serine 225 is mostly conserved in Firmicutes, but interestingly is exchanged in HemQ from S. aureus by a threonine. In Actinobacteria, an arginine is found at the respective position [9]. The impact of these amino acid residues (e.g., Q187, Y147, and M149) on Reactions 1 and 2 as well as the observed differences between HemQs in Firmicutes and Actinobacteria must be investigated in the future. To date, a HemQ from Actinobacteria has not been investigated on a protein level.
In conclusion, HemQ is a homopentameric enzyme that uses coproheme as substrate and redox-cofactor for the conversion of coproheme to heme b. For the relatively slow and stepwise conversion of coproheme to heme b, two H 2 O 2 molecules are necessary. The heme cavity architecture together with biochemical/ physical data suggest homolytic cleavage of hydrogen peroxide and a radical mechanism that includes two one-electron oxidations of the propionate substituent (p2 and p4). As a consequence of interaction with basic amino acids, the propionates p6 and p7 seem to be protected from oxidation.

Expression and purification of LmHemQ
LmHemQ was subcloned into a modified version of the pET21(+) expression vector with an N-terminal StrepII-tag, cleavable TEV protease, expressed in Escherichia coli Tuner (DE3) cells (Merck/Novagen) and purified via a StrepTrap HP 5 mL column (GE Healthcare) as described in previously [8]. The enzyme was reconstituted by the addition of equimolar concentrations of free Fe(III) coproporphyrin III chloride (coproheme; Frontiers Scientific, Logan, UT, USA) (in 50 mM phosphate buffer, pH 7.0, predissolved in 0.5 M NaOH) to apo-LmHemQ dissolved in 50 mM phosphate buffer, pH 7.0. In order to determine the extinction coefficient of coproheme, various concentrations of coproheme solutions in 50 mM phosphate buffer, pH 7.0, were prepared and UV-vis absorption spectra were recorded with an Agilent 8453 diode array spectrophotometer (Hewlett-Packard, Palo Alto, CA, USA). The absorbance at 390 nm was plotted against the coproheme concentrations and from the slope, the extinction coefficient for coproheme at 390 nm was obtained (128 800 M À1 Ácm À1 ).

Enzymatic activity of LmHemQ
The conversion of coproheme to heme b by LmHemQ was followed spectroscopically at 410 nm (e 410 = 76 600 M À1 Ácm À1 ) [9]. Solution of 2 lM coproheme-LmHemQ in 50 mM phosphate buffer, pH 7.0, was used and the reactions were started by the addition of 10-100 lM H 2 O 2 (Sigma, Vienna, Austria). The H 2 O 2 concentration was checked spectrophotometrically prior to each measurement (e 240 = 39.4 M À1 Ácm À1 ) [21]. The initial linear phase was used to determine the Michaelis-Menten parameter. Production rates of heme b (lM s À1 ) were plotted against H 2 O 2 concentrations, and K M and v max were calculated using a Hanes plot [extraction of catalytic parameters from the slope (1/v max ) and the intercept (K M /v max )].
The activity of LmHemQ was followed spectroscopically between 250 and 700 nm using the cyclic measurement mode (every 2.2 s, a spectrum was recorded for 200 s, Specord 10 photodiodearray photometer from Zeiss, Jena, Germany). In this experiment, 5-50 lM H 2 O 2 was added to 5 lM coproheme-LmHemQ, in 50 mM phosphate buffer, pH 7.0.
Reaction of H 2 O 2 with coproheme-LmHemQ followed by electronic circular dichroism spectroscopy Electronic circular dichroism spectroscopy was performed using Chirascan (Applied Photophysics, Leatherhead, UK). The instrument was flushed with a nitrogen flow of 5 LÁmin À1 and allowed simultaneous monitoring of UV-vis electronic absorption and circular dichroism. Coproheme-LmHemQ was analyzed in the near UV and visible region (260-500 nm; 5 s nm À1 scan speed, 1 nm bandwidth, 10 mm path length, stirred cuvette). Temperature was controlled by a Peltier element and kept at 25°C. All measurements were performed in 50 mM phosphate buffer, pH 7.0. During titrations, subequimolar aliquots of H 2 O 2 were added to the enzyme solution under constant stirring.

Determination of the reaction stoichiometrymass spectrometry
In order to determine the stoichiometry of H 2 O 2 to coproheme, titrations were performed, where small (subequimolar) aliquots of H 2 O 2 were added to coproheme-LmHemQ. In a typical experiment, H 2 O 2 was added in 2 lM aliquots to 18 lM coproheme-LmHemQ. To ensure complete reaction of the H 2 O 2 , an interval of at least 15 min was set between each peroxide addition. The titrations were performed in a stirred cuvette (1 mL sample volume) at room temperature and monitored by UV-vis absorption spectroscopy between 250 and 700 nm, using a scanning photometer (Hitachi U-3900, Tokyo, Japan). About 20 lL of the sample was taken after each addition of H 2 O 2 and analyzed by mass spectrometry.
For mass spectrometric analysis, 5 lL of each sample was analyzed using a Dionex Ultimate 3000 system directly linked to a QTOF mass spectrometer (maXis 4G ETD; Bruker, Rheinstetten, Germany) equipped with the standard ESI source of the positive ion mode. MS scans were recorded within a range from 300 to 3750 m/z and the instrument was tuned to detect both the rather small, free heme derivatives and intact proteins in a single run. Instrument calibration was performed using ESI calibration mixture (Agilent, Santa Clara, CA, USA). For separation of the analytes, a Thermo ProSwift TM RP-4H Analytical separation column (250 9 0.2 mm) was used. A gradient from 99% solvent A and 1% solvent B (solvent A: 0.05% TFA, B: 80% ACCN and 20% solvent A) to 65% B in 11 min was applied, followed by a 2-min gradient from 65% B to 95% B, at a flow rate of 8 lL min À1 and at 75°C. A blank run (5 lL H 2 O) was performed after each sample to minimize carryover effects.  Harvesting and X-ray data collection, structure determination, and refinement Prior to data collection, the crystal was soaked with mother liquor supplemented with 25% MPD, harvested using cryoloops, and flash-vitrified in liquid nitrogen. Datasets were collected at beam-line ID29 [22] of European Synchrotron Radiation Facility (ESRF, Grenoble, France) at 100 K using a DECTRIS PILATUS 6M detector. The dataset was processed with XDS and symmetry equivalent reflections merged with XDSCONV [23]. Intensities were not converted to amplitudes. The high-resolution cutoff was based on a CC1/2* criterion [24]. The phase problem was solved by molecular replacement using Phaser-MR [25] taking PDB structure 4WWS, apo-HemQ from L. monocytogenes [8]. The model was further improved by iterative cycles of manual model building using COOT [26] and maximum likelihood refinement using PHENIX-Refine [27]. PHENIX-refine converted intensities into amplitudes using the French and Wilson algorithm [28]. Restraints for coproheme (Ligand ID RM9) were generated using eLBOW taking a sdf file as input and applying the final-geometry option. Final stages of refinement included Translation Liberation Screw (TLS) parameters, isotropic B-factor model, automated addition of hydrogens and water molecules, optimization of X-ray/ADP weight, and optimization of X-ray/stereochemistry weight. The model was validated with MolProbity [29]. Figures were prepared with PYMOL (http://www.pymol.org).

Sample preparation
Ferric coproheme-LmHemQ samples were prepared by adding free coproheme (in 50 mM phosphate buffer, pH 7.0, predissolved in 0.5 M NaOH) to apo-LmHemQ dissolved in 50 mM phosphate buffer, pH 7.0. Ferric heme b-LmHemQ samples were prepared by adding small aliquots of a concentrated solution of H 2 O 2 in 50 mM phosphate buffer, pH 7.0 to coproheme-LmHemQ. Ferrous samples were prepared by addition of a freshly prepared sodium dithionite (10 mgÁmL À1 ) solution to the ferric forms previously degassed with nitrogen.

Electronic absorption measurements
Electronic absorption measurements were recorded using a 5 mm NMR tube (300 nmÁmin À1 scan rate) or a 1 cm cuvette (600 nmÁmin À1 scan rate) at 25°C by means of a Cary 60 spectrophotometer (Agilent Technologies) with a resolution of 1.5 nm. Absorption spectra were measured both prior and after RR measurements to ensure that no degradation occurred under the experimental conditions used.

Resonance Raman measurements
The resonance Raman (RR) spectra were obtained at 25°C using a 5 mm NMR tube by excitation with the 406.7 nm line of a Kr + laser (Coherent, Innova 300 C; Coherent, Santa Clara, CA, USA) and with the 441.6 nm line of a He-Cd laser (Kimmon IK4121R-G). Backscattered light from a slowly rotating NMR tube was collected and focused into a triple spectrometer (consisting of two Acton Research SpectraPro 2300i instruments and a SpectraPro 2500i instrument in the final stage with gratings of 3600 and 1800 groovesÁmm À1 ) working in the subtractive mode, equipped with a liquid nitrogen-cooled CCD detector. A spectral resolution of 1 cm À1 was calculated theoretically on the basis of the optical properties of the spectrometer for the 3600 grating. The RR spectra were calibrated with indene, n-pentane, and carbon tetrachloride as standards to an accuracy of 1 cm À1 for intense isolated bands. All RR measurements were repeated several times (about 20 different samples) under the same conditions to ensure reproducibility. No changes have been observed within 30 min of irradiation with 5 mW (ferric) and 10 mW (ferrous) for laser excitation in the Soret band. To improve the signal-to-noise ratio, a number of spectra were accumulated and summed only if no spectral differences were noted (see caption of Fig. 5). All spectra were baseline-corrected.
in the refinement process. This project was supported by the Austrian Science Fund, FWF [Doctoral Program BioToP -Molecular Technology of Proteins (W1224) and the projects P25270, P29099, I2429] and Federal Ministry of Economy, Family and Youth through the initiative 'Laura Bassi Centres of Expertise', funding the Center of Optimized Structural Studies, No. 253275.

Author contributions
SH, DP, IS, GM performed crystallization experiments. DP, GM refined the structure. SH, IS performed and analyzed activity assays. DM performed mass spectrometry analysis. LM, GS performed and analyzed resonance Raman spectroscopy experiments. The study was designed by SH, GS, PGF, KDC, CO. The paper was written by SH, GM, GS, PGF, KDC, CO. All authors reviewed the results and approved the final version of the manuscript.