Studies of the membrane‐bound flavocytochrome MsrQ flavin mononucleotide (FMN)‐binding site reveal an unexpected ubiquinone cofactor

Abstract

The methionine sulfoxide reductase PQ system (MsrPQ) is a newly identified type of bacterial methionine sulfoxide reductase (Msr) involved in the repair of periplasmic methionine residues that have been oxidized by hypochlorous acid. MsrP, which carries out the Msr activity, is a molybdoenzyme located in the periplasm, whereas MsrQ, an integral membrane‐bound flavohemoprotein, specifically transfers electrons to MsrP to drive catalysis. MsrQ belongs to an important superfamily of heme‐containing membrane‐bound proteins, which includes the eukaryotic NADPH oxidases (NOX) and six‐transmembrane epithelial antigen of the prostate (STEAP) ferric reductases. Like STEAP, and in addition to a b‐type heme, MsrQ contains a flavin cofactor [flavin mononucleotide (FMN)], which mediates electron transfer from a cytosolic NADH oxidoreductase to the heme, and subsequently to MsrP. In this study, we characterized the FMN‐binding site of MsrQ using an AlphaFold model, identifying R77 and R78 residues as potentially critical for FMN stabilization. The R77A and R78A mutations result in the complete loss of the FMN cofactor, showing that both residues are essential for FMN binding. Surprisingly, electron paramagnetic resonance (EPR) spectroscopy and biochemical analysis of the mutants revealed the presence of a ubiquinone (UQ) cofactor associated with MsrQ, independently of the binding of FMN. The mid‐point redox potentials of the MsrQ heme and FMN cofactors, measured through redox titration and cyclic voltammetry experiments, contradict the previous assumption that UQ serves as the electron donor for MsrQ. Instead, our data suggest that UQ may function as an electron acceptor for the reduced form of MsrQ. We propose that UQ bound to MsrQ could act as a protective mechanism when MsrP substrate is limiting.

The membrane‐bound heme flavoprotein MsrQ is the specific electron donor for the periplasmic methionine sulfoxide reductase MsrP. Analysis of its flavin environment revealed that ubiquinone (UQ) is a cofactor for MsrQ. However, redox studies demonstrate that UQ cannot function as an electron donor for the MsrPQ system. Instead, UQ likely serves as an electron acceptor for reduced MsrQ, potentially providing a protective mechanism when the Met‐SO substrate is limiting.

Abbreviations

3D

three‐dimensional

CV

cyclic voltammetry

DMK

demethylmenaquinone

DUOX

dual oxidase

EPR

electron paramagnetic resonance

FAD

flavin adenine dinucleotide

FMN

flavin mononucleotide

FRD

ferric reductase domain

Fre

NAD(P)H‐flavin reductase

HALS

highly anisotropic low spin

HPLC

high‐performance liquid chromatography

HPLC‐ECD‐MS

HPLC coupled to electrochemical detection and mass spectrometry

LMNG

lauryl maltose neopentyl glycol

Met‐SO

methionine sulfoxide

MK

menaquinone

Mo‐MPT

molybdenum‐molybdopterin cofactor

MsrP

methionine sulfoxide reductase periplasmic catalytic component

MsrPQ

methionine sulfoxide reductase PQ system

MsrQ

methionine sulfoxide reductase membrane‐bound component

NAD(P)H

reduced nicotinamide adenine dinucleotide (phosphate)

NHE

normal hydrogen electrode

Ni‐NTA

nickel‐nitrilotriacetic acid

NOX

NADPH oxidase

RMSD

root mean square deviation

ROS

reduced oxygen species

STEAP

six‐transmembrane epithelial antigen of the prostate

TEV

tobacco Etch virus protease

UQ

ubiquinone

WT

wild‐type

Introduction

Reactive oxygen and chlorine species are major issues for cellular systems. Protein methionine residues are particularly susceptible to oxidation in the presence of these species, forming methionine sulfoxide (Met‐SO), which dramatically affects protein structures and functions [1]. It has been known for many years that cells contain methionine sulfoxide reductase enzymes (Msr) to repair oxidized methionine by reducing it to methionine, thus playing an important role in the integrity of prokaryotic and eukaryotic cells [2, 3].

Recently, it was discovered that some bacteria can express a novel type of Msr system, called MsrPQ, that appears to be specialized to counteract the toxicity of hypochlorous acid HOCl [4, 5]. HOCl is the main component of bleach and is produced as a first line of defense by the innate immune system to kill pathogens, such as bacteria, for which it is highly toxic [6]. Within cells, HOCl primarily oxidizes protein methionine residues to Met‐SO, leading to the dysfunction of essential proteins [7]. MsrPQ repairs these damaged residues [4] and may also provide a protection/virulence system for certain pathogens [8].

MsrPQ is unusual among the Msr class of enzymes in that it is a two‐component system with the metalloproteins MsrP and MsrQ encoded within the same operon [4]. While most Msr described to date lack metals in their active sites and act through catalytic cysteine residues [9], MsrP, formally known as YedY, is a periplasmic soluble protein that harbors a molybdopterin cofactor in its active site, resembling enzymes of the sulfite oxidase molybdenum family [10, 11]. MsrP catalyzes the two‐electron reduction of periplasmic proteins Met‐SO back to methionine and to be functional in vivo, it has to be associated with MsrQ [4, 12]. MsrQ, previously named YedZ [13, 14], is an integral membrane heme protein that acts as the specific electron donor for MsrP [4]. While MsrP structure and function have been well investigated [10, 11, 12, 15, 16, 17], MsrQ has not been as extensively studied, and in particular, no experimental three‐dimensional (3D) structure is available.

Phylogenetic studies have assigned MsrQ to an important superfamily of proteins with a heme‐containing membrane domain, named ferric reductase domain (FRD), present in bacteria and eukaryotes, including NADPH oxidases (NOX) [18]. The FRD domain consists of six‐transmembrane segments in which generally two bis‐histidine residue motifs each chelate a b‐type heme, one located on the cytosolic side of the membrane, the other on the extracellular side. Within this family, MsrQ and the eukaryotic ferric reductases STEAP1‐4 (STEAP for six‐transmembrane epithelial antigen of the prostate) form a special group in which one heme‐histidine ligand is not conserved, suggesting that there is only one transmembrane heme in these proteins [18]. The presence of only one heme localized on the extracellular side of the membrane has been confirmed in the cryo‐EM structures of STEAPs 1, 2, 3, and 4 [19, 20, 21] and in MsrQ by biochemical characterizations [22].

Instead of the heme localized on the cytoplasmic side of the membrane in most members of the FRD family, a flavin cofactor is found in STEAP and MsrQ. Specifically, STEAP contains a flavin adenine dinucleotide (FAD) [20, 21, 23, 24], while MsrQ contains a flavin mononucleotide (FMN) [22]. The reason for the presence of a flavin instead of a heme on the cytoplasmic side of the membrane in STEAP and MsrQ is still under investigation [20, 22, 23, 24]. Nevertheless, the presence of two redox cofactors in the FRD domain family clearly allows electron transfer across the membrane by coupling a cytosolic reductase system to an electron acceptor on the extracellular/periplasmic side of the membrane [18]. The electron acceptor is MsrP for MsrQ and ferric complexes for STEAP. In the case of STEAP 2–4 proteins, a cytosolic NADH dehydrogenase serves as electron donor and is part of the entire polypeptide chain, forming a soluble N‐terminal domain that remains in the cytoplasm [25]. Conversely, the MsrQ sequence lacks such a soluble cytosolic dehydrogenase domain. Nevertheless, biochemical experiments have shown that the cytosolic NAD(P)H‐flavin reductase Fre catalyzes efficient electron transfer from NADH to the FMN cofactor of MsrQ, which subsequently leads to heme reduction [22, 26]. Other work based on genetic and phenotypic approaches has suggested that the initial electron donor for MsrQ is the membrane‐bound quinone pool, allowing reduction of MsrQ heme and subsequent electron transfer to MsrP to catalyze Met‐SO reduction [4]. However, the ability of membrane‐bound quinones to reduce MsrQ heme and/or FMN cofactors has not been clearly established and requires further investigations.

In this work, we characterized the FMN site of MsrQ using an AlphaFold model and identified the Arg77 and Arg78 residues as involved in FMN binding. Unexpectedly, spectroscopic and biochemical characterization of the R77A, R78A, and H151A mutants revealed the presence of a ubiquinone (UQ)/menaquinone (MK) cofactor associated with MsrQ, independent of the presence of the FMN cofactor. By measuring the redox potentials of the MsrQ heme and FMN cofactors, we determined that UQ cannot be an electron donor for MsrQ. On the contrary, we suggest that UQ may function as an electron acceptor for reduced MsrQ.

Results and discussion

AlphaFold model of MsrQ

Our previous work has highlighted the presence of an FMN cofactor in MsrQ, which as described for STEAP, could occupy the site where a second b‐type heme binds in other members of the FRD superfamily [22]. These data strongly suggest that the domain containing the two cofactors in MsrQ, a b‐type heme and a FMN, has a topology similar to that of other FRD members [18], such as STEAP [19, 20, 21], NOX [27, 28], and dual oxidase (DUOX) [29, 30], for which 3D structures are available.

In the absence of an experimental 3D structure for MsrQ, we used the AlphaFold 3 program with b‐type heme and flavin cofactors to model the structural arrangement of the residues lining up the cavity that binds the FMN in MsrQ (Fig. 1A). The predicted template modeling indicator (pTM) of 0.89 shows a high level of confidence of the overall model and of its complexation with the heme and FMN cofactors (Fig. S1). The local distance difference test (lDDT), which evaluates the local quality per residue, shows that the structure of the functional transmembrane domain with a four alpha helix bundle (h2‐h5) that binds the two cofactors in the FRD family [31] is properly predicted (Fig. S1). The structural positioning of the transmembrane helices h1 and h6, specific to each FRD protein, is also well‐predicted in MsrQ. Finally, the positioning of the soluble C terminus h7 helix appears less well‐defined (Fig. S1). In order to better position the side chains of the residues binding the cofactors, the AlphaFold MsrQ structural model was aligned with the experimental structures of STEAP [19, 20, 21] NOX [27, 28] and DUOX [29, 30] (Fig. S2). Table S1 summarizes the Cα‐Cα root mean square deviation (RMSD) of the alignments and shows that only 65–75% of the MsrQ residues were considered for the structural alignments. These residues are mostly localized in the four alpha helices h2‐h5. The alignments confirm that the MsrQ transmembrane h2‐h5 core is more similar to that of STEAPs (RMSD ~2.3–2.5 Å), holding a cytosolic flavin cofactor, instead a heme for NOX (RMSD ~1.9–3.5 Å) and DUOX (RMSD ~2.9–4.2 Å) (Table S1). In MsrQ, the heme and FMN pockets are directly exposed to the periplasm and cytoplasm spaces, respectively (Fig. 1B; Fig. S3). The distribution of the MsrQ surface electrostatic potential shows that the environment of the periplasmic heme is slightly negative, while that of the cytoplasmic FMN, which is flanked by several basic residues, is strongly positive (Fig. 1B). This electrostatic charge difference between the flavin and heme environments, negative around the FMN and positive around the heme, could favor an electron transfer from the cytoplasm to the periplasm. The heme in MsrQ is positioned at the same location as its counterpart in STEAPs (Fig. S2A), with the two histidines ligands in STEAP fully superimposing the corresponding His in MsrQ, His91, and His164 (Fig. S2A). The isoalloxazine ring and the ribityl phosphate moieties of the FMN cofactor in MsrQ (Fig. 1D) superimpose very well with the corresponding moieties of the FAD in STEAP (Fig. S2A). Interestingly, in STEAP4 an electronic density in a groove of the hydrophobic protein surface near the flavin‐binding site has been associated with a structural phospholipid, potentially stabilizing the nearby basic residues [21]. In the MsrQ structural model, a similar phospholipid molecule could be positioned at the corresponding location, including potential interactions with equivalent basic residues (Fig. 1C; Fig. S2A).

Fig. 1.

AlphaFold 3 structural model of MsrQ with its b‐type heme and flavin mononucleotide (FMN) cofactors. (A) Overall structure of MsrQ, with transmembrane helices h2‐h5 (in green) forming the four alpha helix bundle holding the cofactors in the ferric reductase domain family (FRD), while h1 and h6 (in blue and orange, respectively) are specific to each protein in this family. The b‐type heme and FMN cofactors point toward the periplasmic and cytoplasmic sides of the membrane, respectively. Short connecting loops lie in the periplasm and cytosol. (B) MsrQ electrostatic charge surface (in red and blue for negative and positive charges, respectively), highlighting the heme (top) and the FMN (bottom) electrostatic environments. (C) Close view of the MsrQ FMN‐binding site, the phospholipid molecule in cyan. (D) Structure of the FMN molecule. The protein structure images were generated using PYMOL.

Structural alignments of MsrQ with STEAP have identified equivalent residues at the same position on either side of the isoalloxazine ring, forming a pocket that stabilizes the flavin cofactor (Figs S2A, S4A,B). In MsrQ, such residues are His151, Trp147, Leu131, and Ser135 located on one side of the isoalloxazine ring, and Arg77 and Arg78 on the other side (Fig. 1C; Fig. S4A, Table S2). Interestingly, Arg77 is strictly conserved in MsrQ and STEAP proteins and points its guanidinium group toward the ribityl phosphate group of the flavin (Fig. S4A,B). In NOX and DUOX, Arg77 is replaced by an essential histidine that coordinates the heme cofactor (Fig. S4C,D). Arg78, conserved in MsrQ and NOX but replaced by a lysine in STEAP and DUOX, points its guanidinium group toward the structural phospholipid in such a way that its aliphatic side chain is in van der Waals contact with the FMN isoalloxazine plane (Fig. 1C; Fig. S4A,B). Arg77 and Arg78 appear to close the MsrQ FMN pocket which is otherwise fairly open to the cytoplasm (Fig. 1B,C; Fig. S4A). These two residues could therefore be involved in the stabilization of the FMN cofactor. Leu131, His151, and Trp147 could play a similar role, forming a vault above the FMN in van der Waals contact with the upper side of the isoalloxazine plane (Fig. 1C). In STEAP, the leucine is conserved while the tryptophan is replaced by a phenylalanine (Fig. S4B). Importantly, His151, whose mutation to alanine has been shown to destabilize the FMN in MsrQ [22], is positioned identically to the histidine that coordinates the second b‐type heme in NOX/DUOX (Fig. S4). This residue is replaced by a Gln in STEAP (Fig. S4B), which could be involved in a hydrogen bond with the 2′OH ribityl group of the flavin (Fig. S4B) [21]. In MsrQ, His151 is positioned just above the central part of the isoalloxazine ring and could form a π‐π stacking interaction with it (Fig. 1C; Fig. S4A). This is consistent with the involvement of His151 in FMN stabilization, as highlighted above [22]. These differences between Gln in STEAP and His151 in MsrQ in their potential interactions with the redox part of the flavin could have functional implications, possibly with respect to the redox potential. Finally, Ser135, which is well conserved in STEAP, could bind the N5 atom of the FMN (Fig. 1C; Fig. S4A,B) and possibly modulates the redox process.

In conclusion, in addition to the already characterized His151 [22], the present MsrQ structural model suggests that Arg77 and Arg78 may also be important for FMN cofactor stabilization.

Characterization of MsrQ R77A and R78A mutants

To investigate the role of Arg77 and Arg78 residues in FMN binding, MsrQ R77A and R78A mutants were constructed. Both mutants were overexpressed in E. coli, solubilized from the membranes by the amphiphilic detergent lauryl maltose neopentyl glycol (LMNG) and purified using the same protocol as for the wild‐type (WT) protein [22]. The UV–visible spectra of both purified MsrQ mutants were similar to that of the WT [22], with typical absorption bands derived from oxidized b‐type heme, a Soret peak at 412–413 nm and broad α/β bands centered around 560 and 530 nm, respectively (Fig. S5). Upon reduction with dithionite, the reduced minus oxidized difference spectra of both MsrQ mutants showed absorption peaks at 558, 528, and 426–427 nm, corresponding to the α, β, and Soret reduced heme bands, respectively (Fig. S5). The extinction coefficients for the heme absorption bands of both mutants are similar to those reported for the MsrQ WT (Table S3) [22]. As shown in Table S3, the heme/protein ratios for the R77A and R78A mutants indicate that, as reported for MsrQ WT [22], a significant amount of apoprotein is present in the R77A preparation, with an even higher proportion in the R78A sample. Hereafter, as reported in [22], the concentrations of MsrQs are related to their holoprotein forms based on their heme content.

The FMN content of the MsrQ mutants was analyzed by high‐performance liquid chromatography (HPLC) in comparison with the WT protein (Fig. 2). As previously reported, the WT has a near‐stoichiometric amount of FMN with respect to its heme content [22]. For both MsrQ R77A and R78A mutants, no flavin cofactor was detected in the purified proteins (Fig. 2). The addition of FMN to the MsrQ mutant solutions prior to extraction and HPLC analysis allowed the detection of an amount of FMN equal to that added to the protein solutions (Fig. 2). Thus, the absence of FMN in both mutants is not due to its degradation during the analysis. These data show that the R77A and R78A mutations cause the complete loss of the FMN cofactor in MsrQ.

Fig. 2.

HPLC analysis of the flavin mononucleotide (FMN) content of MsrQ wild‐type (WT), R77A and R78A mutant forms. The elution of the C18 column was followed at 266 nm. The data were reproduced on 3 independent protein preparations. (A) Commercial FMN solution at 1 μm. (B) MsrQ WT solution at 1 μm. The concentrations of MsrQs refer to the concentration of their holoprotein forms, based on their heme content. The peak eluted at 12.5 min corresponds to an FMN at a concentration of 0.90 μm. (C) MsrQ R77A solution at 1 μm. (D) MsrQ R77A plus FMN solution, 1 μm each. The FMN peak at 12.5 min corresponds a concentration of 0.85 μm. (E) MsrQ R78A solution at 1 μm. (F) MsrQ R78A plus FMN solution, 1 μm each. The FMN peak at 12.5 min corresponds to a concentration of 0.90 μm. No FAD or riboflavin, which in these conditions elute at 8 and 15 min, respectively, were detected in the MsrQ samples.

Electronic paramagnetic resonance (EPR) of MsrQ mutants

To characterize the heme cofactor in the MsrQ mutants, EPR spectroscopy was performed on the oxidized proteins and analyzed based on previous work on WT and H151A proteins [22]. The spectrum of the R77A mutant shows signatures characteristic of the two different forms of low‐spin b‐type hemes, as observed in the WT protein (Fig. 3A,B). They correspond to a major highly anisotropic low‐spin (HALS) heme species (S = 1/2) with a signal at g = 3.62, attributed to a highly anisotropic large g _z contribution [22, 32, 33] and a minor conventional low‐spin heme (S = 1/2) with a rhombic signal g _z = 2.95, g _y = 2.25, g _x = 1.5 [22, 32]. These two low‐spin contributions reflect the presence of two different relative orientations for the imidazole plane of the axial histidine‐heme ligands in MsrQ. The rhombic signal is associated with roughly parallel histidine ligand planes, whereas the HALS one is related to nearly perpendicular histidine ligand planes [32, 33, 34]. These EPR data provide an additional information compared to the AlphaFold model, predicting only one conformation for the His ligands (Fig. S2A). The g = 5.8 signal is characteristic of a high‐spin penta‐coordinated heme (S = 5/2), which, in spite of its apparent intensity, represents only a very small fraction of the heme in MsrQ [22, 35]. Also, nonspecific ferric iron is identified at g = 4.3. The MsrQ R78A mutant shows EPR signals that are similar to those of the R77A mutant, except that the intensity of the low‐spin HALS signal (g = 3.6) was strongly decreased and that of the low‐spin rhombic species was increased (Fig. 3C). This suggests that the b‐type heme conformation with the two histidine ligand planes parallel is favored in this mutant compared to R77A and WT proteins.

Fig. 3.

X‐band electron paramagnetic resonance (EPR) spectra of the ferric wild‐type (WT) MsrQ and R77A and R78A mutants, in 50 mm Tris/HCl pH 7.6, LMNG 0.004%. Spectra were reproduced on 2 independent protein preparations. (A) WT MsrQ, 302 μm (black trace), (B) MsrQ R77A, 274 μm (blue trace), (C) MsrQ R78A, 208 μm (red trace). Experimental conditions: temperature, 15 K; microwave power, 2 mW at 9480 GHz; modulation amplitude, 2 mT at 100 kHz. Experimental spectra were obtained from one scan and are only weakly saturated at 2 mW (less than 10%), a power value which enables to improve the signal to noise ratio. The spectra result from subtraction with the signal of the buffer alone, without baseline correction. In (A) the g‐values are labeled for each signal.

Unexpectedly, while the flavin cofactor is absent in the R77A and R78A mutants (Fig. 2), the g = 2.00 signal indicates that an organic radical is still present in these mutants (Fig. 3). This observation suggests that an unidentified organic cofactor that stabilizes a radical species other than a flavin is present in these MsrQ mutants.

The g = 2.00 signals were further studied at a higher temperature (60 K) at X‐band and at a higher microwave frequency at Q‐band (Fig. 4). The H151A mutant, which has also been previously shown to lack the flavin and exhibit a weak EPR g = 2.00 signal [22], was also examined for comparison. In the three mutants, the radical signal exhibits at X‐band the same isotropic line at g = 2.004 with a peak to peak linewidth of about 1 mT and no resolved hyperfine structures (Fig. 4, upper panel). At higher frequency (Q‐band), the anisotropy of the g‐tensor of the radical is partially resolved and with similar values (g _┴ = 2.005 and g _// = 2.002) for the three mutants (Fig. 4, lower panel). These values are identical to those measured for the radical signal in the WT MsrQ [22] and close to those typically measured with flavin radicals [36] or semiquinone radicals in membrane‐bound respiratory enzymes [37] and references therein. It is worth noting that radicals based on amino acid side chains exhibit EPR signals with significant differences: Their g‐tensor anisotropy is much higher for Tyr^• (2.0073 to 2.0021) [38, 39] and Cys^• (2.29 to 2.006) [40] and they show well‐resolved hyperfine structures at X‐band with major hyperfine splitting ranging between 1.5 and 2 mT for Tyr^•, Cys^•, Trp^•, Gly^• [38, 39, 40, 41]. These differences led us to rule out the assignment of the g = 2.00 MsrQ signal to a radical based on amino acid. As the FMN group is no longer present in the mutants, it suggests that a quinone species may be bound to the mutated MsrQ.

Fig. 4.

X‐band (upper panel) and Q‐band (lower panel) electron paramagnetic resonance (EPR) spectra of the radical signal in MsrQ R77A, R78A and H151A mutant proteins. (A) R77A (blue traces), (B) R78A (black traces), (C) H151A (green traces), (D) buffer alone (purple trace). Experimental conditions: temperature, 60 K; microwave power, 0.4 mW at 9.4688 GHz (X‐band) or 0.1 mW at 34.019 GHz (Q‐band); modulation amplitude, 1 mT (X‐band) or 0.5 mT (Q‐band) at 100 kHz; averaged scans, 10 (X‐band) or 23 (Q‐band). The lines marked with stars at Q‐band arise from the signal of a weak amount of contaminating Mn²⁺.

Ubiquinone and menaquinone as cofactors of MsrQ

The quinone content of different purified MsrQ forms was analyzed by HPLC coupled to electrochemical detection and mass spectrometry (HPLC‐ECD‐MS). As shown in Table 1, all the different purified MsrQ proteins tested here, WT, H151A, R77A, and R78A, contain quinones, mainly ubiquinone 8 (UQ8) and minor quantities of menaquinone 8 (MK8). For the WT protein, quinone analysis was performed on seven independent MsrQ preparations purified either by a single nickel‐nitrilotriacetic acid (Ni‐NTA) column or by a three‐step protocol, which consists of a Ni‐NTA column followed by tobacco Etch virus protease (TEV) cleavage to remove the C‐terminal His tag and a final Mono‐Q anion‐exchange chromatography, as previously reported [22]. HPLC‐ECD‐MS analysis of the different WT protein preparations revealed comparable levels of UQ8 and MK8, with an average of 0.4–0.5 mole of ubiquinone per mole of holo‐MsrQ and about 5 times less of menaquinone (Table 1). Importantly, demethylmenaquinone 8 (DMK8) was not detected in any of the purified MsrQ proteins, while it is present in E. coli cells overexpressing MsrQ (Fig. S6). This shows that the low levels of MK8 detected in the different MsrQ preparations, and a fortiori the more important levels of UQ8, result from specific interactions of these quinones with MsrQ. In all cases, the quinones/MsrQ stoichiometry is less than 1 (Table 1), suggesting a single quinone binding site on MsrQ.

Table 1.

Quinone content ^a of purified MsrQ wild‐type (WT) and mutant forms.

MsrQ form	UQ8/MsrQ b , c	MK8/MsrQ b , d	UQ8/MK8
WT	0.45 ± 0.06	0.090 ± 0.020	5
H151A	0.39 ± 0.10	0.038 ± 0.012	10.2
R78A	0.26 ± 0.04	0.028 ± 0.010	9.3
R77A	0.12 ± 0.01	0.009 ± 0.002	13.3

^a Quinones were identified and quantified by HPLC coupled to electrochemical detection. Data were shown as mean ± SD from 7 preparations of MsrQ WT, 6 preparations of MsrQ H151A, 4 preparations of MsrQ R77A, and 4 preparations of MsrQ R78A.

^b The concentrations of MsrQs refer to the concentration of their holoprotein forms based on their heme content and calculated using the epsilon values from Table S1.

^c Mole of UQ8 per mole of MsrQ.

^d Mole of MK8 per mole of MsrQ.

These data are fully consistent with the presence of the organic radical detected by EPR at g = 2.00 in both mutants and originating from a small fraction of ubiquinone/menaquinone stabilized under a one‐electron reduced semiquinone radical state in the presence of air (Fig. 4). By extension, the EPR organic radical observed for the WT could also come from a stabilized semiquinone form of UQ8 and/or MK8 rather than a flavin semiquinone, as previously proposed [22]. As highlighted below, taking into account the redox potential of the FMN cofactor of −340 mV vs. normal hydrogen electrode (NHE), stabilization of a flavin semiquinone radical in the presence of air is unlikely.

It should be noted that the quinone content of the R78A mutant, and even more for R77A, is significantly lower than that of the WT, whereas that of H151A is similar (Table 1). This observation suggests that R77 and possibly R78 may somehow be involved in stabilizing the quinone in its binding site. However, since H151A, R77A, and R78A all lack the FMN cofactor (see above), binding of the quinone is unlikely to be directly correlated with the presence of FMN, suggesting that FMN and ubiquinone have independent binding sites on MsrQ.

Unfortunately, no consensus amino acid sequences have been defined for quinone binding sites on membrane‐bound proteins. Therefore, in the absence of an experimental 3D structure, it is not possible to localize the quinone cofactor on the MsrQ model shown in Fig. 1.

Potentiometric redox titration and cyclic voltammetry on MsrQ

The mid‐point potentials of the MsrQ cofactors were investigated by redox titration and cyclic voltammetry (CV) experiments.

Potentiometric redox titrations of purified MsrQ WT were monitored by optical absorption spectroscopy at pH 7.6 using the Soret heme bands at 413 and 426 nm in their oxidized and reduced forms (Fig. 5). The dye mediators were chosen to achieve a redox equilibrium between the protein and the electrode over the range of potentials studied (+200 to −370 mV vs. NHE). At 426 nm, the spectral contribution of these mediators was negligible compared to that of heme and this wavelength was favored to follow redox titrations. The redox dependence of the heme absorption spectrum was determined by successive addition of sodium dithionite (Fig. 5). As shown in the inset of Fig. 5, reductive redox titrations yield a mid‐point potential for the MsrQ WT heme of −242 mV ± 2 mV vs. NHE, with the Nernst fit consistent with a one‐electron reduction process. The reversibility of the redox process was verified by successive addition of K₃Fe(CN)₆ to the fully reduced protein and was observed in all cases (data not shown). It should be noted that the FMN cofactor of MsrQ, which has a very weak contribution to the visible absorption spectrum of MsrQ compared to that of heme [22], could not be studied with this experimental protocol.

Fig. 5.

Potentiometric redox titration of the MsrQ wild‐type heme cofactor. The reduction of MsrQ, 10 μm in 50 mm Tris/HCl pH 7.6, LMNG 0.004%, was achieved by successive addition of sodium dithionite and monitored by following the Soret absorptions of ferric and ferrous heme at 413 and 426 nm, respectively. The arrows show the evolution of the different heme absorbance bands during the reduction process. The asterisk (*) indicates absorbance bands associated with oxidized redox mediators, which disappear upon reduction. The inset shows the fraction of the reduced Soret heme absorption band at 426 nm plotted as a function of the redox potential. The data were analyzed using a Levenberg–Marquardt algorithm from the KaleidaGraph^® software package (Synergy Software). The solid red line displays the theoretical Nernst fit of the data points for a one‐electron reduction process. A mid‐point redox potential value of −242 mV ± 2 mV vs. normal hydrogen electrode (NHE) was obtained from experiments performed in triplicate and expressed as the mean ± SD.

The mid‐point potential of the MsrQ WT cofactors was also investigated by CV, using an edge plane pyrolytic graphite electrode. MsrQ was deposited as a protein film on the surface of the electrode, which was then placed on the electrochemical cell buffer for analysis. As shown in Fig. 6A, CV showed two broad quasi‐reversible processes, which were highlighted after correction for the nonfaradaic current (Fig. 6B). The first process displays a mid‐point potential of −256 mV vs. NHE (peak cathodic potential, Epc = −275 mV; peak anodic potential, Epa = −238 mV, ΔEp = 37 mV), while the second one exhibits a more negative mid‐point potential of −340 mV vs. NHE (peak cathodic potential, Epc = −367 mV; peak anodic potential, Epa −314 mV, ΔEp = 53 mV).

Fig. 6.

Cyclic voltammetry (CV) on MsrQ wild‐type (WT) and H151A mutant proteins. (A) Cyclic voltammograms of MsrQ WT (red line) and H151A mutant (black line). Data were recorded under anaerobic conditions using a pyrolytic graphite electrode in 50 mm Tris/HCl pH 7.6, LMNG 0.004%, scan rate 150 mV·s⁻¹. CV experiments were replicated on 2 different samples, with each trace resulting from an average of 10 individual CVs. Values expressed as the mean ± SD. For MsrQ WT, two quasi‐reversible redox systems are observed with mid‐point potentials ((Epc + Epa)/2) of −340 mV ± 5 mV and ‐256 mV ± 5 mV, vs normal hydrogen electrode (NHE), respectively. For MsrQ H151A, a single quasi‐reversible redox system is observed with a mid‐point potential of −256 mV ± 5 mV, vs NHE. (B) Electrochemical signals from (A) after subtracting nonfaradaic current using a spline function.

CV experiments were performed under the same conditions on one of the MsrQ mutants that specifically lacks the FMN cofactor. The H151A mutant was chosen because its quinone content is similar to that of the WT (Table 1). In CV experiments, the H151A mutant shows only a single broad quasi‐reversible process (Fig. 6A,B), superimposed on the first reduction wave observed with the WT protein (mid‐point potential of −256 mV vs. NHE, Epc = −275 mV; Epa = −238 mV, ΔEp = 37 mV). The second process at ‐340 mV vs. NHE observed in the WT totally disappears in the H151A mutant (Fig. 6A,B). These data strongly suggest that the reduction process at ‐340 mV originates from the MsrQ FMN cofactor. The first reduction process at −256 mV, present in the WT and H151A mutant, can be attributed to the heme, which is in good agreement with the mid‐point potential of the MsrQ heme measured by redox titration (−242 mV, Fig. 5).

Notably, a small additional redox process around −65 mV vs. NHE, mainly observable after correction for the nonfaradaic current, was detected in the WT but not for H151A (Fig. 6B). The origin of this redox process remains difficult to attribute. However, it does not appear to be associated with the quinone bound to MsrQ, which is similarly present in both WT and H151A (Table 1). Apparently, the quinone bound to MsrQ could not be investigated by CV.

Redox titration experiments using EPR spectroscopy on E. coli membranes overproducing MsrQ suggested a much more positive mid‐point potential for the MsrQ heme, with a value of −8 mV vs. NHE [10]. However, our observations indicate that overexpressed MsrQ represents only a small fraction of the total membrane‐bound heme proteins in E. coli membranes. Therefore, it is possible that the highly anisotropic low‐spin (HALS) EPR signal used in [10] for redox titration is not primarily derived from MsrQ, but rather from other more abundant b‐type heme‐containing proteins present in the E. coli membrane.

Overall, potentiometric redox titration and cyclic voltammetry on MsrQ allowed the determination of the mid‐point potentials for the heme and FMN cofactors, which are around −250 mV and −340 mV vs. NHE, respectively.

Interestingly, the redox potential of the MsrQ heme measured in this study is in the range of that of its periplasmic partner MsrP (YedY) (Fig. 7), whose active molybdenum‐molybdopterin cofactor (Mo‐MPT) was shown to exhibit a mid‐point potential of −250 mV vs. NHE [15, 42]. Thus, our data support that MsrQ can be an effective electron donor for MsrP, in agreement with the fact that they act as specific physiological partners, as shown by genetic and phenotypic experiments in E. coli [4]. These data also suggest that the low redox potential of the MsrQ heme may be an important aspect of the specificity of MsrP for MsrQ. Indeed, most membrane‐bound hemoproteins tend to have much more positive redox potentials than MsrQ, which are not thermodynamically compatible with electron transfer to MsrP.

Fig. 7.

Scheme of the electron transfer chain associated with MsrPQ activity. The couple NADH/NAD(P)H‐flavin reductase (Fre) is presented as the cytosolic electron source and methionine sulfoxide (Met‐SO) as the final periplasmic electron acceptor. The mid‐point redox potentials, vs. normal hydrogen electrode (NHE), of the different cofactors (flavin mononucleotide (FMN), heme, ubiquinone (UQ), molybdenum‐molybdopterin cofactor (Mo‐MPT)/substrates (NADH, Met‐SO) are noted versus NHE. ^a Present work. ^b From reference [15].

Previously, it was proposed that the E. coli membrane‐bound quinone pool could be the initial electron donor for the MsrPQ system [4]. The model was that quinones may directly reduce the MsrQ heme, which would then transfer the electrons to MsrP in the periplasm. However, our current data strongly suggest that the quinone pool cannot be the cellular reductive source for the MsrPQ system. Indeed, the usual mid‐point redox potential values of membrane‐bound quinone have been reported to be +110 mV for ubiquinone, −7 mV for demethylmenaquinone, and −70 mV for menaquinone (vs. NHE) [43, 44]. Obviously, these values are too positive to consider any of these quinones as competent MsrPQ electron donors, since the mid‐point potentials of the MsrQ heme and the MsrP Mo‐MPT active site are both around −250 mV, as pointed above. As a consequence, the original proposal that the quinone pool of the E. coli membrane could be the electron donor for MsrPQ [4] has to be reconsidered.

Taken together, the data presented in this paper are consistent with our previous proposal that the electron source of MsrPQ could be the cytosolic NAD(P)H‐flavin reductase Fre (Fig. 7) [22, 26]. Fre catalyzes the reduction of the MsrQ heme through an unconventional electron transfer process, utilizing the FMN cofactor of MsrQ as a cosubstrate alongside NADH [22]. The proposed mechanism suggests that FMN transfers from its binding pocket in MsrQ to the active site of Fre, where it is reduced by NADH before returning to its MsrQ binding site to subsequently reduce the heme. Our current data may provide additional insights into this mechanism. We have demonstrated that residues Arg77 and Arg78, located on the cytosolic side of MsrQ, are critical for maintaining the stable and tight binding of FMN in MsrQ. These residues may play a central role in the proposed mechanism. We hypothesized that conformational changes involving Arg77 and/or Arg78 could facilitate the release of FMN, allowing it to move from MsrQ's pocket to the active site of Fre. Such conformational changes may be triggered by the formation of the specific complex between Fre and MsrQ, as previously characterized [22].

Finally, given our present results, the function of the quinone bound to MsrQ remains unclear. As stated above, it is obvious that quinones cannot act as electron donors for MsrPQ. The different mid‐point redox potentials measured here suggest that electrons could flow in the opposite direction, that is, from the reduced MsrQ heme and FMN cofactors to the oxidized quinone, which would therefore act as an electron acceptor (Fig. 7). The reasons why such MsrQ‐associated quinone could accept electrons from the heme and FMN cofactors clearly deserves further investigation. Nevertheless, the presence of a quinone cofactor in MsrQ, as reported here, should be placed in parallel with the genetic and phenotypic experiments in E. coli showing that deletion of genes involved in quinone biosynthesis results in loss of MsrPQ activity [4]. These data revealed an essential role of the quinone somehow associated with the MsPQ activity.

In light of the current data, we hypothesize that the MsrQ quinone cofactor might be involved in a protective mechanism that allows the recycling of electrons from MsrQ to the respiratory chain when MsrP substrate is limited (Fig. 7). Indeed, at low concentrations of Met‐SO, MsrP turnover slows down, resulting in the accumulation of electrons on MsrQ. The presence of oxygen could then lead to MsrQ autoxidation with the formation of harmful reduced oxygen species (ROS) such as superoxide radical and/or hydrogen peroxide. The presence of a ubiquinone cofactor could favor a redirection of the excess electrons on MsrQ to the membrane‐bound respiratory chain, avoiding detrimental oxygen reduction processes and the formation of ROS.

Conclusions

MsrQ was previously described as a membrane‐bound flavohemoprotein, homologous to the eukaryotic STEAP proteins, with its heme and FMN cofactors located on the periplasmic and cytoplasmic sides of the membrane, respectively [22]. However, our present findings reveal that MsrQ is a more intricate electron transfer protein, incorporating a third cofactor, an ubiquinone–menaquinone, tightly bound to its polypeptide chain.

The mid‐point redox potentials of the MsrQ heme and FMN cofactors reported here suggest that they could transfer electrons to the quinone cofactor, while the reverse is highly unlikely. These data strongly suggest that the membrane‐bound quinone pool could not function as an electron donor system for MsrPQ, as previously proposed [4]. Instead, we hypothesize that the quinone cofactor in MsrQ may play a protective role, potentially preventing the formation of ROS during MsrPQ turnover.

Another objective of this study was to gain a deeper understanding of the properties of the FMN‐binding site in MsrQ. Our results, based on AlphaFold modeling and site‐directed mutagenesis, indicate that residues Arg77 and Arg78 located on the cytosolic side of MsrQ are essential for maintaining FMN bound within its pocket in MsrQ. These data suggest that any conformational changes affecting these residues could result in the release of FMN from its active site. This could constitute a key process for the activity of the NAD(P)H‐flavin reductase Fre, which has been proposed as a cytosolic electron source for MsrPQ [22], by enabling the exchange of the FMN cofactor between the active site of MsrQ and that of Fre to transfer electrons from NADH to the heme.

Materials and methods

Chemicals and reagents

LMNG (lauryl maltose neopentyl glycol) was obtained from Anatrace. All other chemicals were from Sigma‐Aldrich.

Expression and purification of MsrQ forms

The msrQ‐Tag gene cloned into a pET28 plasmid was used for site‐directed mutagenesis (Quick Change mutagenesis kit from Stratagene) and overexpression of the E. coli MG 1655 MsrQ, with an eight histidine noncleavable tag added at its C‐terminal side [22]. For the R77A mutation, the following oligonucleotides were used, primer 1: 5′ CGTTATTGATACGCACTGCGCGCCTGTTAGGATTATG 3′; primer 2: 5′ CATAATCCTAACAGGCGCGCAGTGCGTATCAATAACG 3′. For R78A, primer 1: 5′ CGTTATTGATACGCACTCGCGCGCTGTTAGGATTATG 3′; primer 2: 5′ CATAATCCTAACAGCGCGCGAGTGCGTATCAATAACG 3′. DNA sequences were verified on both strands. Recombinant MsrQ proteins were overexpressed and purified using a Ni‐NTA column, as described previously [22]. Purified proteins were equilibrated in 50 mm Tris/HCl pH 7.6, 0.004% LMNG, using a PD10 column.

Heme extinction coefficients

The heme content of MsrQ was measured by the pyridine hemochromogen assay, using the α peak of the reduced pyridine hemochromogen with an extinction coefficient at 556 nm of 34.7 mm ⁻¹·cm⁻¹ [22, 45].

AlphaFold modelization

The MsrQ structure (UniProt P76343 sequence from E. coli K12) was modeled by the AlphaFold 3 program with one b‐type heme and one FAD as ligands. The adenine part of FAD was further deleted from the prediction. A phospholipid molecule (phosphatidic acid with a 5 carbons fatty‐acid chain) was added in the model at a same position than in the STEAP4 structure (PDB: 6hcy), after superposition of both structures.

Flavin content

Identification and quantification of the MsrQ flavin content were carried out by HPLC analysis [22]. All the solutions were protected from light with aluminum foils. MsrQ proteins were diluted to 0.5–10 μm in 5 mm ammonium acetate, pH 6.5 (200 μL final volume), incubated at 95 °C for 5 min, and then cooled down 10 min in ice. The denatured proteins were eliminated by centrifugation, 15 000  g for 10 min at 4 °C, and 100 μL of the supernatant solution were injected onto an HPLC analytical reverse phase C18 column (4.6 × 250 mm) kept at 35 °C, using an Agilent 1260 Infinity device. The HPLC column was equilibrated at a flow rate of 1 mL·min⁻¹ with 85% 5 mm ammonium acetate, pH 6.5 and 15% methanol. The flavins were eluted with a linear gradient from 15% to 75% methanol for 20 min at 1 mL·min⁻¹. Calibration curves were obtained by injection of 100 μL of 0.2 to 2 μm commercial FMN solutions. The flavin peaks were identified by their UV–visible spectra and integrated from their absorbance at 266 nm or at 450 nm.

EPR and electronic absorption spectroscopies

Low‐temperature (15 K) X‐band EPR spectra were recorded on a Bruker EMX spectrometer equipped with an Oxford Instrument ESR 910 cryostat. The microwave frequency was calibrated with a frequency counter and the magnetic field with a NMR gaussmeter. High temperature (60 K) radical spectra were obtained on a Bruker Elexsys E500 spectrometer equipped at X‐band with an ER4102ST standard rectangular cavity fitted to an Oxford Instruments ESR 900 cryostat and at Q‐band with a ER6106 QT resonator fitted to an Oxford Instruments CF935 cryostat. The presence of the Mn²⁺ signal visible at Q‐band enabled an accurate calibration of the magnetic field (±0.1 mT) in order to determine precisely the g‐values of the radical.

Aerobic optical absorbance measurements were performed using a Varian Cary 50 spectrophotometer (0.25 nm bandwidth) with 1 cm path length cuvettes.

Quinone extraction and analysis

Quinones were extracted from purified MsrQ solutions using methanol and petroleum ether [46] and quantified by HPLC coupled to electrochemical detection and mass spectrometry [47]. About 50 μL of purified MsrQ (200 picomoles) was completed with water up to 200 μL into glass tubes. Then, 50 μL KCl 3 m, 3 mL methanol and 80 pmoles of UQ10 (used as internal standard to correct for potential losses during extraction [46]) were added and vortexed for 1 min. About 2 mL petroleum ether (40–60° boiling range) was added and vortex for 1 min. After centrifugation at 700 rpm for 1 min at 22 °C, the upper phase was collected and the methanol phase was extracted again with 2 mL petroleum ether. The two petroleum ether phases were pooled and dried under a nitrogen flow. The extracts were resuspended in 100 μL ethanol and fractions corresponding to 5 and 20 μg of protein were analyzed by HPLC coupled to electrochemical detection and mass spectrometry, as described in [47]. The probe temperature was 400 °C, the cone voltage was 80 V, and the MS spectra were recorded between m/z 500 and 900 with a scan time of 0.5 s. A solution of commercial MK7 (Ɛ_{248 nm} = 18 900 m ⁻¹·cm⁻¹) was used for a standard curve (2 to 100 pmoles), in order to quantify the MK8 peaks from the protein samples. UQ8 was quantified based on a UQ10 commercial standard as in [47].

Potentiometric redox titration

Redox titrations were performed at 20 °C under anaerobiosis on a glove box (Jacomex B210), filled with a nitrogen atmosphere containing less than 2 ppm O₂. The glove box was equipped with a UV–visible cell coupled to a Cary 60 spectrophotometer by optical fibers (Photonetics).

The protein solution (1.5 mL) was placed in a plastic tube (Ø 1.4 cm) with a small magnetic stir bar inside the glove box. The redox potentials were measured directly in the protein solution with a combined Pt‐Ag/AgCl/KCl (3 m) microelectrode (InLab® Redox Micro, Mettler‐Toledo International Inc., Columbus, OH, USA) and given versus a standard hydrogen electrode (NHE). The electrode was calibrated with a commercial redox buffer solution (Mettler‐Toledo International Inc.). MsrQ was at 10 μm, in 50 mm Tris/HCl pH 7.6, LMNG 0.004%, in the presence of the following mediators at 0.6 μm each: 2,5‐dimethyl benzoquinone (+180 mV), phenazine methosulfate (+80 mV), methylene blue (+11 mV), resorufin (−55 mV), indigo disulfonate (−80 mV), 2‐hydroxy 1–4 naphtoquinone (−145 mV), phenosafranin (−255 mV), and neutral red (−325 mV). The protein solution was first fully reduced and then re‐oxidized by stepwise additions of small quantities of 10 mm of sodium dithionite or K₃Fe(CN)₆, respectively. After each addition of sodium dithionite or K₃Fe(CN)₆, the solution was stirred for 10 min to allow thermodynamic equilibrium to be reached, with the redox potential value, measured by the microelectrode, to be stabilized. Then, 100 μL of the solution was transferred in a Hellma quartz ultra‐micro cell (path length 10 mm) and a UV–Vis spectrum of the solution was recorded.

Electrochemistry

Protein‐film voltammetry experiments (CV) were recorded under anaerobic conditions in a Jacomex glovebox (O₂ < 2 ppm) using a potentiostatic three‐electrode assembly and a computer‐controlled BioLogic potentiostat PS 200. The working electrode was a 3‐mm diameter disc of edge plane pyrolytic graphite edge (EPPG) electrode (Biologic, Seyssinet‐Pariset, France), the reference electrode was an Ag/AgCl electrode (RE1b from BioLogic), while the counter‐electrode was a platinum wire (BioLogic). For each experiment, electrodes were placed into a setup combining a small volume analytical cell and the sample holder of the SVC‐2 kit (Biologic), as described previously [48]. The electrode was polished with sandpaper 1200 followed by 1 μm alumina polishing, and then, baseline measurements were collected by placing the EPPG electrode into the electrochemical buffer cell solution (50 mm Tris/HCl pH 7.6, LNMG 0.004%). A 5 μL droplet of MsrQ (180 μm) in 50 mm Tris/HCl pH 7.6, LNMG 0.004% was deposited to the polished EPPG electrode positioned upside for 15 to 20 min until to obtain a protein film before being rinsed with 200 μL of the buffer cell solution. Then, the electrode was immediately placed back in the cell buffer solution for measurements. Mid‐point redox potentials of MsrQ WT and H151A mutant were determined through CV at room temperature at a scan rate of 150 mV/s. Potentials versus the normal hydrogen electrode (NHE) were obtained by adding 0.210 V. CV control experiments in the presence of the buffer alone were performed. A direct subtraction of the CV of MsrQ by the CV of the buffer is not possible due to the alteration of the capacitive response caused by the absorption of the protein on the electrode. Rather, a spline polynomial interpolation was used to subtract the baseline due to the nonfaradaic component from the raw data using the SOAS package [49].

Conflict of interest

The authors declare no conflict of interest.

Author contributions

PC performed the AlphaFold structural analysis and proposed the mutations. FP coordinated and analyzed quinone contents. ND performed and interpreted CV experiments. BG performed and interpreted EPR spectra. HC and LF analyzed quinone contents. CB purified proteins. CC carried out site‐directed mutagenesis. ST recorded EPR spectra. VN conceived and coordinated the study, analyzed and interpreted the data throughout the work, performed experiments and wrote the paper with the contributions from all the authors.

Supporting information

Fig. S1. Confidence metrics of the AlphaFold 3 structural model of MsrQ, hosting the b‐type heme and FMN cofactors.

Fig. S2. Superimposition of the predicted MsrQ structure with the FRD superfamily structures in the PDB.

Fig. S3. Computation of the cavities in the predicted structure of MsrQ.

Fig. S4. Comparison of the cytosolic cofactor environment of predicted MsrQ (FMN) with that of STEAP (FAD), NOX (heme) and DUOX (heme).

Fig. S5. UV–visible spectral characterization of MsrQ R77A and R78A mutants.

Fig. S6. Quinone content of cells overexpressing MsrQ and of purified MsrQ.

Table S1. Metrics of superimposition between MsrQ and the other structures of the FRD superfamily in the PDB.

Table S2. Correspondence between residues in the cytosolic cofactor environment of MsrQ and those of other proteins in the FDR superfamily.

Table S3. Extinction coefficient values of the oxidized Soret and reduced α heme bands for the purified MsrQ wild‐type, R77A and R78A mutants.

Data availability statement

The data that support the findings of this study are available from the corresponding author on reasonable request.