Carnitine metabolism in the human gut: characterization of the two-component carnitine monooxygenase CntAB from Acinetobacter baumannii

Bacterial formation of trimethylamine (TMA) from carnitine in the gut microbiome has been linked to cardiovascular disease. During this process, the two-component carnitine monooxygenase (CntAB) catalyzes the oxygen-dependent cleavage of carnitine to TMA and malic semialdehyde. Individual redox states of the reductase CntB and the catalytic component CntA were investigated based on mutagenesis and electron paramagnetic resonance (EPR) spectroscopic approaches. Protein ligands of the flavin mononucleotide (FMN) and the plant-type [2Fe-2S] cluster of CntB and also of the Rieske-type [2Fe-2S] cluster and the mononuclear [Fe] center of CntA were identified. EPR spectroscopy of variant CntA proteins suggested a hierarchical metallocenter maturation, Rieske [2Fe-2S] followed by the mononuclear [Fe] center. NADH-dependent electron transfer via the redox components of CntB and within the trimeric CntA complex for the activation of molecular oxygen was investigated. EPR experiments indicated that the two electrons from NADH were allocated to the plant-type [2Fe-2S] cluster and to FMN in the form of a flavin semiquinone radical. Single-turnover experiments of this reduced CntB species indicated the translocation of the first electron onto the [Fe] center and the second electron onto the Rieske-type [2Fe-2S] cluster of CntA to finally allow for oxygen activation as a basis for carnitine cleavage. EPR spectroscopic investigation of CntA variants indicated an unusual intermolecular electron transfer between the subunits of the CntA trimer via the “ bridging ” residue Glu-205. On the basis of these data, a redox catalytic cycle for carnitine monooxygenase was proposed. suggest hierarchical metallocenter assembly of CntA: Rieske [2Fe-2S]

Bacterial formation of trimethylamine (TMA) from carnitine in the gut microbiome has been linked to cardiovascular disease. During this process, the two-component carnitine monooxygenase (CntAB) catalyzes the oxygen-dependent cleavage of carnitine to TMA and malic semialdehyde. Individual redox states of the reductase CntB and the catalytic component CntA were investigated based on mutagenesis and electron paramagnetic resonance (EPR) spectroscopic approaches. Protein ligands of the flavin mononucleotide (FMN) and the plant-type [2Fe-2S] cluster of CntB and also of the Rieske-type [2Fe-2S] cluster and the mononuclear [Fe] center of CntA were identified. EPR spectroscopy of variant CntA proteins suggested a hierarchical metallocenter maturation, Rieske [2Fe-2S] followed by the mononuclear [Fe] center. NADH-dependent electron transfer via the redox components of CntB and within the trimeric CntA complex for the activation of molecular oxygen was investigated. EPR experiments indicated that the two electrons from NADH were allocated to the plant-type [2Fe-2S] cluster and to FMN in the form of a flavin semiquinone radical. Single-turnover experiments of this reduced CntB species indicated the translocation of the first electron onto the [Fe] center and the second electron onto the Rieske-type [2Fe-2S] cluster of CntA to finally allow for oxygen activation as a basis for carnitine cleavage. EPR spectroscopic investigation of CntA variants indicated an unusual intermolecular electron transfer between the subunits of the CntA trimer via the "bridging" residue Glu-205. On the basis of these data, a redox catalytic cycle for carnitine monooxygenase was proposed.
The human gut contains trillions of microbes supporting to breakdown otherwise indigestible food constituents (1). There is growing evidence that the human microbiota plays an important role in health and disease (2)(3)(4)(5). The gut microbiome was related to obesity (3,6), to the development of the adaptive immune system (7), and to the emergence of atherosclerosis (8). Accordingly, manipulation of the gut microbiota was considered a future therapeutic strategy to combat disease and improve health (4).
Dietary nutrients containing a trimethylamine (TMA) moiety, inter alia L-carnitine or choline, were related to the development of atherosclerotic heart disease (9). The C-N bond of these compounds is cleaved by bacterial carnitine monooxygenase (CntAB) or choline trimethylamine-lyase (CutCD), leading to the volatile compound TMA, which is further metabolized in the human liver by flavin monooxygenases (FMOs) (Fig. 1). The resulting trimethylamine N-oxide molecule (TMAO) was shown to enhance atherosclerosis development in animal models (10,11). In multiple human studies, elevated TMAO levels have been associated with prevalent cardiovascular diseases and the risks for myocardial infarction and stroke (12,13). Only recently, inhibition of the choline lyase CutCD using the substrate analog 3,3-dimethyl-1-butanol was exemplified as an efficient strategy to attenuate choline diet-enhanced atherosclerosis in a mouse model (14) (Fig. 1).
The two-component carnitine monooxygenase CntAB (EC 1.14.13.239) catalyzes the oxidative cleavage of L-carnitine into TMA and malic semialdehyde ( Fig. 2A). The latter compound serves as a carbon and energy source and is channeled into the tricarboxylic acid cycle by virtue of malic semialdehyde dehydrogenase and malate dehydrogenase (15). Despite the structural similarity of the substrates L-carnitine and choline, CntAB catalysis is completely unrelated to CutCD, which utilizes an unusual glycyl radical enzyme mechanism (16).
Only recently, the clustering of genes cntA and cntB in bacterial reference genomes of the human microbiota has been demonstrated (17,18). Theoretical sequence analyses for CntB revealed a potential nucleotide binding sequence, a flavin binding domain, and a motif characteristic of the presence of a plant-type [2Fe-2S] cluster. The sequence of CntA shows substantial homology to the family of Rieske-type oxygenases ( Fig.  2B) (19). These iron-dependent enzyme systems typically comprise an oxygenase, a specific reductase, and eventually an additional ferredoxin component (20). Numerous Rieske-type oxygenases have been described in the literature, e.g. the first discovered toluene dioxygenase and naphthalene dioxygenase (21,22), other biotechnologically relevant systems that can break aromatic ring structures (23,24), and Rieske-type oxygenases that oxidize compounds such as cholesterol (25) or the amine groups of 4-aminobenzylamine, stachydrine, or caffeine (26)(27)(28). All highlighted oxygenases share a highly conserved Rieske sequence (-CXHX [15][16][17] CXXH-), providing two cysteine and two histidine ligands for the characteristic [2Fe-2S] cluster. Apart from this, CntA also possess a mononuclear iron-binding region. The related [Fe] center of Rieske proteins, typically ligated by two histidine residues and one carboxylate ligand, plays a fundamental role in the sophisticated activation of molecular oxygen (Fig. 2). The oligomeric architecture of Rieske oxygenases is of central importance, because the electron transfer onto the [Fe] center from the Rieske center is facilitated via an intersubunit rather than an intrasubunit redox process (29) (Fig. 2C). As judged from three-dimensional protein structures, a highly conserved aspartate residue (e.g. Asp-205 in naphthalene 1,2-dioxygenase, PDB entry 1NDO) is substantial for the electron transfer between the adjacent subunits (29). Interestingly, this "bridging" aspartate is replaced by a glutamate in all available sequences of CntA.
Recently, carnitine monooxygenase activity of the Acinetobacter baumannii enzyme was reconstituted, and the functional relevance of the abovementioned Glu-205 has been demonstrated by site-directed mutagenesis (18). Furthermore, a closely related substrate specificity of orthologous CntA proteins was demonstrated in vitro. High activity in the presence of L-carnitine and gamma-butyrobetaine, medium activity toward glycine betaine, and very low activity in the presence of choline was shown (14,18). Apart from this, the underlying bio-inorganic redox chemistry of the unusual Rieske-type carnitine monooxygenase has not been investigated.
Here, we present the first biochemical characterization of the redox-active components of CntB and CntA from A. baumannii. Site-directed mutagenesis in combination with kinetic and electron paramagnetic resonance (EPR) experiments allowed us to resolve individual electron transfer steps required for the oxygenation of the carnitine substrate.

Enzymatic activity of CntAB
Components CntB and CntA of the carnitine monooxygenase from A. baumannii were individually overproduced and purified from Escherichia coli cells. The N-terminally Histagged CntB protein was affinity purified using Co 21 -loaded chelating Sepharose. Approximately 5 mg purified CntB with a relative molecular weight of ;45,000 (calculated molecular mass, 37,151 Da) was obtained from a 1-liter cell culture (SDS-PAGE) (Fig. 3A, lane 3).
The catalytic component CntA was overproduced as an Nterminal GST fusion protein and purified using chromatography on GSH-Sepharose. The untagged target protein was eluted after on-column cleavage with PreScission protease. Overall, 6 mg CntA was purified from a 1-liter culture volume. SDS-PAGE analysis (Fig. 3A, lane 6) revealed a relative molecular weight of ;43,000 (calculated molecular mass, 43,155 Da).
Oxygenation of the carnitine substrate must be considered a key step of CntAB catalysis, and molecular oxygen was proposed to be a source of the carbonyl oxygen of the malic semialdehyde reaction product (18). Only recently, in vivo experiments demonstrated higher TMA production in more aerobic environments. However, verification of molecular oxygen as a cosubstrate is still pending.
Accordingly, components CntB and CntA were recombinantly purified under strict anaerobic conditions. A subsequent carnitine depletion assay or an NADH depletion assay (performed in an anaerobic chamber) did not reveal detectable CntAB activity ( Fig. 3B and D), in agreement with the reaction scheme depicted in Fig. 2A. Enzymatic activity was effectively restored upon oxygen exposure.

CntB is an FMN-dependent reductase
CntB acts as a carnitine monooxygenase reductase, transferring reduction equivalents from NADH onto CntA. Analytical size exclusion chromatography revealed a native molecular mass of 43,000 Da, indicating that CntB is functional as a monomer. The purified protein showed an orange/yellow color, and UV-visible spectroscopy revealed absorption maxima at 272 nm, 340 nm, 410 nm, and 463 nm, which might be indicative of a prosthetic FAD or FMN cofactor and/or the presence of an iron-sulfur cluster (30) (Fig. 4A).
Further experimental evidence for the presence of a flavin cofactor was obtained from fluorescence spectroscopy. In Fig.   Figure 1. Gut microbiota and metabolism of dietary nutrients associated with atherosclerosis. The gut microbiota makes use of dietary nutrients Lcarnitine and choline as carbon and energy sources. The C-N bond of these compounds is cleaved by TMA lyases. The resulting TMA molecule is further metabolized in the human liver by flavin monooxygenases (FMOs) of the host. The synthesized TMAO molecule was shown to enhance atherosclerosis development. In multiple human studies, elevated TMAO levels have been associated with cardiovascular diseases and the risks for myocardial infarction and stroke. For the TMA lyase CutCD, an efficient targeted inhibitor was identified that prevented the formation of TMA in mouse models.
4A, inset, the characteristic emission of the supernatant from a heat-denatured CntB sample at 526 nm (370 nm excitation) indicates the presence of a noncovalently bound flavin. Comparative HPLC analyses in the presence of authentic FAD/ FMN samples clearly identified FMN as a cofactor (Fig. 4B). As judged from the supernatant of thermal denaturation experiments, an FMN content of 0.6 6 0.1 mol per mol CntB was determined.
FMN binding was further explored using structure-guided mutagenesis. Functional relevance of amino acid residues Asp-75 and Ser-82 in the FMN binding pocket of CntB was proposed after assessment of the three-dimensional structure of the orthologous protein benzoate dioxygenase reductase (PDB entry 1KRH, 23% sequence identity). Purified CntB variant proteins D75K and S82A revealed a characteristic decrease of the absorption at 340 and 463 nm and the parallel reduction of the fluorescence signal at 526 nm ( Fig.  4A and Table 1). In agreement with a significant decrease of the cofactor content (78% 6 5% and 15% 6 4% FMN), residual activities of 70% 6 8% and 35% 6 6% were observed compared with the WT protein (Table 1). From these data, we conclude that Asp-75 and Ser-82 contribute to the tight binding of FMN. NADH-dependent electron transfer of both variants was efficiently restored when the flavin cofactor was externally supplied to the assay.

CntB carries a plant-type [2Fe-2S] cluster
EPR spectroscopy allows for the characterization of individual redox states of iron-sulfur cluster-containing proteins. The aerobically purified CntB protein (;1 mM) was EPR silent. However, upon reduction with 2 mM sodium dithionite, a complete loss of color, a drastic reduction of the characteristic optical absorption maxima, and an S = 1/2 signal with g values of g 1 = 2.037, g 2 = 1.943, and g 3 = 1.900, located in a region characteristic for a [2Fe-2S] 1 cluster, were observed ( Fig. 4C and D) (31). Concurrently, an iron content of 2.16 6 0.18 mol Fe per mol CntB and a sulfur content of 2.07 6 0.39 mol S per mol CntB was determined.
Theoretical sequence analysis indicated an overall five cysteine residues as potential ligands of the redox-active [2Fe-2S] cluster of CntB. Accordingly, the respective amino acids were individually mutagenized into alanine and the purified variants C265A, C267A, C272A, C275A, and C305A were characterized. Mutant C265A revealed WT-like enzymatic activity in the carnitine and NADH depletion assay. UV-visible experiments, flavin quantification, and EPR spectroscopy indicated the presence of FMN and the [2Fe-2S] cofactor by analogy to the WT protein (compare Fig. 4E and Table 1). Clear differences for variants C267A, C272A, C275A, and C305A were observed. Eluate fractions always contained substantially lower amounts of the target protein (;1/10) contaminated with significant amounts of CntB degradation products. Attempts to further purify these variants were impeded because of the instability of the respective target proteins. Accordingly, freshly prepared protein samples were directly analyzed after affinity purification. Variants C267A, C272A, C275A, and C305A did not reveal detectable enzymatic activity, and EPR measurements of the related samples indicated the complete loss of the [2Fe-2S] center (;200 mM sample concentration). From these data, we propose residues Cys-267, Cys-272, Cys-275, and Cys-305 as ligands of the plant-type [2Fe-2S] cluster of CntB. Apart from functioning as a redox-active component of carnitine monooxygenase, we conclude that this oxygen-insensitive [2Fe-2S] cluster is of central importance for the structural integrity of CntB.
EPR spectroscopy was also used to analyze the FMN-binding variants D75K and S82A (Fig. 4E). The dithionite-reduced samples indicated identical g values and approximately the same signal intensity as that observed for the WT protein. Obviously, these mutations in the FMN binding pocket do not impair the geometry or the maturation of the plant-type [2Fe-2S] cluster of CntB.
The redox midpoint potential of CntB was accessed by employing UV-visible spectro-electrochemistry in a custombuilt cell described previously (32). Because of the limited sta-bility of the protein during these experiments, only approximate values for the redox midpoint potential could be obtained ( Fig. S1 and S2, left). For CntB, a midpoint potential of 2162 6 10 mV was established.

CntA is a functional trimer
Intersubunit electron transfer plays a central role for the redox catalysis of Rieske-type oxygenases. Accordingly, the oligomeric architecture of CntA was assessed by different experimental strategies. Analytical size exclusion chromatography revealed a native molecular weight of 139,000 Da, indicative of a functional CntA homotrimer (calculated monomer mass, 43,156 Da) (Fig. 5A).
Subsequently, small-angle X-ray scattering (SAXS) experiments were performed to verify the proposed trimer. A dilute protein solution was used to determine the experimental . Lanes M, molecular mass marker; relative molecular masses (31,000) are indicated. B, L-carnitine depletion assay. A standard assay containing 5 mM purified CntA, 20 mM purified CntB, and 300 mM L-carnitine at 37°C was initiated by the addition of 2 mM the electron donor NADH (black trace) or NADPH (blue trace). Samples were prepared at least in triplicates and collected after 15, 30, 60, and 90 s, and L-carnitine concentrations were determined colorimetrically in a coupled carnitine acetyltransferase assay (see Experimental procedures). Control reactions in the absence of CntA (red trace), electron donor (green trace), or O 2 (orange trace; proteins purified under strict anaerobic conditions) were performed. C, TMA analysis by GC. Top, a standard assay in the presence of 2 mM L-carnitine and 5 mM NADH was incubated at 37°C for 2 min. The reaction was stopped in the presence of 1 M perchloric acid and subjected to GC analyses. Control reactions in the absence of CntB (middle) or NADH (bottom) were performed. D, NADH depletion assay in the presence of 2 mM purified CntA and CntB, 300 mM L-carnitine, and 200 mM NADH (black trace). The absorption at 340 nm was continuously monitored (black trace). Control reactions in the absence of O 2 (orange trace), CntB (purple trace), and CntA (red trace) were performed.
CntA scattering curve. In Fig. 5B, the comparison of the experimental scattering curve (black dots) with the theoretical scattering curves for a hypothetical monomer (blue) or trimer (red) is shown. A significantly higher quality of fit was obtained for the trimer (x 2 = 4.95) compared with the monomer (x 2 = 99.27). Alternatively, processing of the SAXS raw data using the SAXSMoW 2.0 server (33) revealed a molecular weight of 123.7 kDa, in full agreement with the proposed trimer.
Trimeric Rieske-type oxygenases comprise two evolutionarily conserved contact regions that are centered by a bridging aspartate residue (34,35) that might be replaced by a glutamate in CntA (Glu-205).
The Phyre2 server (36) was used for the three-dimensional modeling of the hypothetical structure of the CntA trimer (monomer templates 3VCA, 3N0Q, 2B1X, 1ULJ, and 1WQL, trimer superimposed to 3N0Q), and the sequence conservation among orthologous CntA sequences was indicated as a color code. As shown in Fig. 5C, the highest degree of surface conservation is found for the interface regions I and II of the depicted trimer interface (Shannon entropy H, shown from 0 to 4.322 bit, maximum conservation, top). These two regions further indicated an opposed electrostatic surface potential (f, from 2257 mV to 1257 mV, middle), arguing for a tight monomer-monomer interaction that is not based on hydrophobic interactions (compare  Table summarizes biochemical and spectroscopic properties of CntB variants. The FMN content (;0.6 6 0.1 mol per mol CntB), the specific activity of the L-carnitine depletion assay (771 6 67 nmol min 21 mg 21 ), enzymatic TMA formation (determined by gas chromatography), and NADH depletion activity (407 6 11 nmol min 21 mg 21 ) of the wild type were set as 100% (compared to experimental procedures). -, experiment not performed. ,5 or ,3, enzymatic activity below the detection limit of the respective assay. ND, an EPR signal was not detectable. hydropathy from 24.5 to 14.5, bottom). The theoretical position of Glu-205 at the trimer interface is indicated by an asterisk. In summary, these experiments allowed us to clearly assign a functional CntA trimer that is tightly connected by polar protein interfaces.

CntA contains a [2Fe-2S] cluster and a mononuclear iron center
Aerobically purified CntA protein fractions showed a brownish color, and UV-visible spectroscopy revealed absorption maxima at 463 nm and 570 nm, indicative of an iron-sulfur protein. An iron content of 3.05 6 0.21 mol Fe per mol CntA and a sulfur content of 2.11 6 0.43 mol S per mol CntA was determined. A concentrated protein sample (;1 mM) without reducing agent was subjected to EPR measurements (Fig. 6B). A strong S = 5/2 signal with a g value of 4.27 was observed, which is characteristic for the presence of a mononuclear iron (III) center. Conversion of such an [Fe] center into the EPR silent iron (II) state can be accomplished upon chemical reduction (37).
Treatment of a related CntA sample with 10 mM sodium dithionite resulted in the bleaching of the brownish color, as indicated by drastically reduced absorption maxima at 463 nm and 570 nm, with the appearance of a new maximum at 520 nm, which is in agreement with previously reported spectroscopic data of related Rieske oxygenases (38) (Fig. 6A, inset). Subsequent EPR experiments indicated complete loss of the [Fe] center signal with the parallel appearance of an S = 1/2 signal with g values of g 1 = 2.016, g 2 = 1.921, and g 3 = 1.759 (Fig.  6B). The obtained values are characteristic of a [2Fe-2S] 1 center of the Rieske type (30).
These findings might be indicative for the artificial two-electron reduction of the CntA protein in the presence of dithionite. A first electron leads to the conversion of the mononuclear iron (III) center into the EPR silent iron (II) state. A subsequent electron is required to render the [2Fe-2S] 21 cluster into the EPR active [2Fe-2S] 1 state. Within the overall EPR study, exclusive reduction of the [2Fe-2S] 21 cluster of CntA was not observed.
UV-visible spectro-electrochemical experiments of CntA were hampered, as a redox-dependent sample precipitation was observed. A lower limit for the midpoint potential of 1300 mV could be estimated (Fig. S2, right).

Ligands of the Rieske [2Fe-2S] cluster and the mononuclear iron center of CntA
The protein sequence of CntA contains a motif that is typical for the ligation of a Rieske-type [2Fe-2S] cluster: CXHX 17 CX 2 H, with residues Cys-86, His-88, Cys-106, and His-109 as potential cluster ligands (in boldface). Mutant proteins C86A, H88A, C106A, and H109A were purified in close analogy to the WT protein. They revealed a substantial decrease of the iron and sulfur content (Fig. 3A, lanes 7-10, and Table 2). Related EPR experiments after dithionite reduction indicated complete loss of the respective [2Fe-2S] cluster signal ( Table 2). No detectable L-carnitine monooxygenase activity was determined in related enzyme activity assays. These findings clearly indicate that the Rieske-type [2Fe-2S] cluster of CntA is coordinated by residues Cys-86, His-88, Cys-106, and His-109 ( Table 2).
The four mutant proteins were also subjected to EPR spectroscopy in the nonreduced state to assess the integrity of the [Fe] center of CntA. Variants H88A and C106A indicated a barely detectable S = 5/2 signal, whereas proteins C86A and H109A were completely devoid of the [Fe] center (Fig. 7, left). These findings indicate that the presence of the Rieske [2Fe-2S] cluster is a prerequisite for the subsequent incorporation of the mononuclear iron.
The related [Fe] center of Rieske proteins is typically ligated by two histidine residues and an aspartate or glutamate residue, forming a so-called 2-His-1-caboxylate triad (39). Residues His-208, His-213, and Asp-323 of CntA were proposed as candidate ligands of the [Fe] center of CntA (18). Compared with the WT, the purified variants H208A, H213A, and D323 indicated partial reduction of the iron content. The nonreduced EPR samples of these mutants revealed a complete loss of the [Fe] center. Proteins H208A, H213A, and D323 did not support any enzymatic activity (Table 2). These findings argue for a mononuclear iron center of CntA that is coordinated by protein ligands His-208, His-213, and Asp-323.  These three variants were also subjected to EPR experiments under reduced conditions. In all cases, a clear signal for the presence of the [2Fe-2S] 1 cluster of CntA was determined (Fig. 7,  right). Obviously, the presence of the mononuclear iron center is not a prerequisite for the maturation of the [2Fe-2S] cluster of CntA. The results of the overall mutagenesis study suggest a hierarchical metallocenter assembly of CntA: Rieske [2Fe-2S] cluster followed by the [Fe] center. However, intact Rieske cluster ligation might be a prerequisite for the spatial positioning of the [Fe] center ligands. According to this, the mutagenesis of residues Cys-86 and His-109 would result in the destabilization of the [Fe] center.

Mutagenesis of the bridging Glu-205
The importance of residue Glu-205 was recently demonstrated based on kinetic experiments of mutant CntA proteins (18). Involvement of this residue in the electron transfer from the Rieske-type [2Fe-2S] cluster to the [Fe] center of the adjacent subunit was proposed.
EPR spectroscopy of variants E205Q and E205D was used to further specify the functional role of this residue. Neither variant did revealed detectable enzymatic activity and a moderately reduced Fe and sulfur content ( Table 2). Reduced EPR samples showed a clear signal of the Rieske-type [2Fe-2S] cluster. In contrast, the nonreduced samples showed almost complete loss of the [Fe] center signal (;90% reduction) (Fig. 8). Obviously, mutagenesis of residue Glu-205 might hamper the maturation or the spatial ligation of the [Fe] center. In this context, variant E205Q indicates that an intact carboxylic group of Glu-205 is essential for the proposed intersubunit electron transfer of CntA. By analogy to related Rieske-type oxygenases, a bridging carboxylate between the [2Fe-2S] cluster and the [Fe] center of the adjacent CntA subunit was concluded. Variant E205D aims to "restore" the bridging aspartate that has been described for various Rieske proteins on the structural and functional level. Obviously, the slightly smaller aspartate residue is not capable of substituting for the bridging Glu-205 of CntA. We conclude that Glu-205 is important for the intersubunit electron transfer between the metallocenters of CntA (18) and/or the subsequent activation of molecular oxygen.

Electron transfer pathway of carnitine monooxygenase
Components CntB and CntA catalyze carnitine monooxygenation (with subsequent TMA, maleic semialdehyde, and water formation, Fig. 2A) using one molecule of dioxygen and two electrons according to the following equation: carnitine 1 2 e 2 1 O 2 ! monooxygenated carnitine 1 H 2 O (substrate oxidized by 2 e 2 , O 2 reduced by 4 e 2 , and 2 external reduction equivalents required).
Both electrons derived from NADH are transferred via the electron transport protein CntB onto the catalytic component CntA. This intercomponent electron transfer is not based on a tight interaction of CntB and CntA. Parallel overproduction of both components in E. coli followed by affinity chromatography of the GST-tagged component CntA did not result in the co-purification of detectable amounts of CntB. In vitro protein-protein interaction experiments using immobilized CntA (or CntB) did not allow for the trapping of CntB (or CntA). Accordingly, transient interaction of the monomeric CntB with the CntA trimer was concluded.
The presented EPR spectroscopic investigation of CntB and CntA already reflected individual redox states that are relevant for the understanding of the electron transfer pathway of carnitine monooxygenase. However, caution should be exercised for the interpretation of chemically reduced samples (inter alia because of the strong negative redox potential of dithionite [40]). Accordingly, we made use of the natural two-electron donor NADH to exemplify intermediate states of carnitine monooxygenase catalysis. All subsequent experiments were performed as single-turnover experiments in the anaerobic workstation to strictly prevent O 2 -dependent reaction(s) prior to EPR measurements.
An immobilized CntB sample (;0.5 mM on Co 21 -loaded chelating Sepharose) was incubated in the presence of 1 mM NADH before the cofactor was removed by an excessive washing step. The resulting reduced elution fraction indicated a substantially differing EPR spectrum when related to the dithionite-reduced spectrum (Fig. 9A, gray trace, versus Fig. 4D). Compared with the dithionite reduced state, the characteristic signal of the plant-type [2Fe-2S] 1 cluster remains almost unaltered (g 1 = 2.037, g 2 = 1.943, and g 3 = 1.900), but a new dominant signal with g = 2.007 in the characteristic region for the semiquinone form of FMN was observed. The temperature-dependent spin relaxation behavior of both signals was further elucidated. Increasing the sample temperature (from 15 to 85 K) did not alter the semiquinone signal but led to an almost complete loss of the plant-type [2Fe-2S] 1 signal (Fig. 9A, compare gray and black dashed traces). To further identify the slow relaxing g = 2 species, we performed a pulsed Q-band electron nuclear double resonance (ENDOR) experiment at 30 K (Fig. S3). The observed hyperfine splittings in the range of 1-24 MHz are consistent with a mostly protonated FMN radical (41). From these experiments, we conclude that the NADH-dependent two-electron reduction locates one electron to the FMN cofactor (in the form of a neutral FMNH* radical) and a second electron to the plant-type [2Fe-2S] cluster of CntB (42). The proposed CntB FMNH*/[2Fe-2S] 1 species (I) is indicated in Fig. 10.
Subsequently, we wanted to figure out if the NADH-reduced CntB species (I) facilitates the intercomponent electron transfer onto CntA. Therefore, a CntB sample was immobilized on Co 21 -loaded chelating Sepharose, reduced by NADH, and the physiological reducing agent was removed by extensive washing. An equimolar amount of CntA then was incubated in the presence of the reduced electron donor protein, and the resulting CntA protein was subsequently eluted in the absence of CntB.
The immobilized CntB was specifically eluted and subjected to EPR measurements. An EPR-silent sample was observed (Fig. 9B) that might indicate electron translocation onto CntA, which in turn would leave the reductase CntB in the FMN/ [2Fe-2S] 21 state (II).
Subsequent EPR measurements of the eluted CntA protein revealed the two-electron reduction of the Rieske protein, as indicated by g 1 = 2.016, g 2 = 1.921, and g 3 = 1.762 (Fig. 9C)

Discussion
TMA formation in the gut microbiome came into focus in recent years, since the related metabolic processes were linked to the development of cardiovascular disease. Bacterial conversion of dietary L-carnitine into TMA and malic semialdehyde is catalyzed by an unusual Rieske-type monooxygenase, which is composed of the reductase component CntB and the catalytic unit CntA. The best-studied Rieske-type oxygenases play an important role in the degradation of aromatic compounds (23,24), but the overall spectrum of reactions ranges from cis-hydroxylations, desaturations, and Oor N-dealkylations to the formation of chiral sulfoxides (22,43). Protein crystallographic studies enabled the understanding of the substrate specificity of many Rieske-type proteins (44). These oxygenases share a conserved domain structure, but their sequences are highly divergent (45). From an evolutionary perspective, CntA was recently ascribed to a novel group of Rieske-type oxygenases (46,47). The enzymes of this group V facilitate the degradation of quaternary ammonium compounds in N-dealkylation reactions. Glycine betaine monooxygenase is another member of this group and has been initially characterized with respect to the reductase BmoB and the oxygenase BmoA. BmoB is an NADH-dependent reductase that carries a noncovalently bound FAD cofactor. The Rieske-type unit BmoA enables the oxygen-dependent degradation of glycine betaine into dimethylglycine and formaldehyde (47).
In the present study, the multistep electron transfer of carnitine monooxygenase was elucidated based on biochemical and spectroscopic experiments. A chain of closely spaced redox relays enables the NAD(P)H-dependent activation of O 2 with subsequent formation of TMA, malic semialdehyde, and water. Biochemical and spectroscopic approaches, in combination with site-directed mutagenesis and single-turnover experiments, allowed us to postulate a catalytic redox cycle of carnitine monooxygenase, which is schematically depicted in Fig. 10. All experimentally characterized redox states of this cycle were highlighted boldface (I-IV).
NADH facilitates the initial reduction of the flavin by two electrons. Subsequently, single-electron transfer from FMNH 2 Figure 9. EPR analysis of single-turnover experiments. A-C, intermediate states of the redox catalysis of CntB and CntA were spectroscopically characterized. All experiments performed under strict anaerobic conditions. A, an immobilized CntB sample was incubated in the presence of 2 mM NADH, the cofactor was removed by extensive washing, and the eluted protein was subjected to low-temperature EPR measurements at 15 K. A [2Fe-2S] 1 cluster signal (g 1 = 2.037, g 2 = 1.945, and g 3 = 1.900) and a semiquinone signal (g = 2.007) was observed (gray trace). Identical measurements at higher temperature (85 K) revealed the specific depletion of the g = 1.  According to this, CntA is an unusual member of the Rieske-type protein family, because most other Rieske oxygenases make use of a fully conserved aspartate residue instead (29,35). The reductive activation of O 2 allows for the monooxygenation of L-carnitine paralleled by the formation of one molecule of H 2 O. Heterolytic cleavage of the C-N bond of this reaction intermediate then results in the formation of malic semialdehyde and TMA ( Fig. 2A).
Future experiments will focus on the mechanism for the activation of molecular oxygen at the mononuclear [Fe] site of CntA and the related recognition of the L-carnitine substrate. The further understanding of this unusual Rieske-type monooxygenase will be of great interest for the targeted inhibition of carnitine monooxygenase as a future therapeutic strategy to combat cardiovascular disease.

Production and purification of CntA and CntB
Codon-optimized genes for the production of A. baumannii ATCC 19606 CntA and CntB were synthesized by GeneArt (Thermo Fisher). cntA and cntB genes were cloned into vectors pGEX-6P-1 and pACYCDuet-1 using the In-Fusion cloning system from Takara with the following primers: GGGCCCCTG-GGATCCATGAGCGCAGTTGAAAAACTGC and GGCCG-CTCGAGTCGACTTACTGGTGATACTGTGCAACC (for construct pGEX-6P-1-cntA) and ACCACAGCCAGATCCG-ATGGCCAGCCACTATGAAATG and CCGCAAGCTTGTC-GACTTACAGATCCAGAACCAGTTTTTTGC (for construct pACYCDuet-1-cntB). The recombinant CntA fusion protein with an N-terminal GSH S-transferase tag or the N-terminally His-tagged CntB protein was produced in E. coli Tuner (DE3) cells. An overnight culture was used to inoculate 500 ml of LB medium containing 100 mg ml 21 ampicillin (pGEX-6P-1-cntA) or 34 mg ml 21 chloramphenicol (pACYCDuet-1-cntB). After ;2 h at 37°C and 200 rpm, an optical density at 578 nm of 0.4 was reached. The cell cultures were incubated for 15 min at 17°C without agitation. Subsequently, the production of the target proteins was initiated by addition of 50 mM isopropyl-b-D-thiogalactopyranoside (IPTG), 1 mM Fe(III)-citrate, and 1 mM L-cysteine hydrochloride (to foster iron center maturation). After 16 h of cultivation at 17°C and 180 rpm, the cells were harvested by centrifugation at 4,000 3 g for 15 min at 4°C, resuspended in 6 ml buffer 1 (100 mM HEPES-NaOH, pH 7.5, 150 mM NaCl), and disrupted by a single passage through a French press at 14,500 psi. Lysates were clarified by centrifugation at 112,000 3 g for 60 min at 4°C, and the supernatant was applied to 1 ml GSH-Sepharose (Macherey-Nagel) for the CntA fusion or 1 ml of a Co 21 -loaded chelating Sepharose (Clontech) for His-tagged CntB. Affinity columns were subsequently washed with 18 ml buffer 1 to remove unbound proteins. Proteolytic on-column cleavage of the CntA fusion was performed with PreScission protease (400 units, 16 h at 4°C) and the untagged CntA protein was eluted in 33 1 ml buffer 1. CntB was eluted with 43 1 ml buffer 1 containing 15 mM imidazole. Target protein fractions were identified by SDS-PAGE.
Purified protein samples were concentrated to ;1 M using Amicon Ultra-0.5 centrifugal filter units (Millipore) equipped with a 30,000-Da cutoff membrane.

Determination of protein concentration
The concentration of purified proteins was determined using Bradford reagent (Sigma-Aldrich) according to the manufacturer's instructions with bovine serum albumin as a standard.

UV-visible light absorption spectroscopy
UV-visible light spectra of purified recombinant CntA and CntB were recorded at room temperature using a V-650 UV-Vis spectrophotometer (Jasco).

Iron determination method
Protein-bound iron was determined colorimetrically with ophenanthroline after acid denaturation of purified CntA or CntB (48).

Sulfur determination method
Protein-bound sulfur was determined as described elsewhere (32).

UV-visible spectro-electrochemical measurements
Redox titrations were performed as described previously (49), using a 25-ml protein sample (0.5 mM) in 0.1 M HEPES-NaOH, pH 7.5, containing 150 mM NaCl. The potential was altered in steps of 10 mV with 10 min of equilibration before spectra were recorded. The redox reaction was followed by the absorbance change at 462 nm, and redox potentials were calculated by fitting the data points to the Nernst equation.

Standard L-carnitine depletion assay
A typical 100-ml standard L-carnitine depletion assay was performed in buffer 1 containing 5 mM purified CntA, 20 mM purified CntB, and 300 mM L-carnitine at 37°C. Reactions were initiated by the addition of 2 mM NADH (or NADPH). At defined time points (0, 10, 30, 60, and 90 s), 100-ml samples were heat inactivated (99°C, 10 min) and centrifuged for 10 min at 12,000 3 g, and the L-carnitine concentration of the resulting supernatant was determined by a modified method according to reference 50. Samples of 50 ml were supplemented with 7.5 ml 2 mM 5,59-dithiobis-(2-nitrobenzoic acid) (DTNB) (Sigma-Aldrich), 15 ml 2 mM acetyl-CoA, and 0.2 U carnitine acetyltransferase (Sigma-Aldrich) in a total volume of 300 ml. The carnitine acetyltransferase reaction is accompanied by the formation of CoA, which then leads to the cleavage of DTNB. The resulting 2-nitro-5-thiobenzoic acid was spectroscopically quantified at 412 nm using an extinction coefficient of e 412 = 14.15 mM 21 cm 21 . A standard curve was determined using defined L-carnitine concentrations. The specific activity was determined by calculation of the slope in the linear range of the experiment. All experiments were performed in triplicate and completed by kinetic measurements in the absence of electron donor or CntA. The L-carnitine depletion assay was also performed under strict anaerobic conditions. Therefore, the purification of components CntB and CntA and the activity measurements were conducted in an anaerobic chamber (Coy Laboratories) using N 2 saturated buffers. Sample manipulation procedures always resulted in an initial lack phase of 10 s. Relative L-carnitine depletion activities of CntB variant proteins were conducted at a reduced CntB concentration of 5 mM.

TMA quantification by GC
Enzymatic TMA formation was determined by GC according to established procedures (51, 52) (IBEN). Assay mixtures (1 ml) containing 5 mM CntA, 20 mM CntB, 2 mM L-carnitine, and 5 mM NADH were incubated for 2 min at 37°C. Reactions were stopped by the addition of 1 M perchloric acid. Variant proteins of CntA or CntB or control experiments in the absence of CntB or NADH were processed accordingly.

NADH depletion assay
A colorimetric NADH depletion assay was performed at 37°C in the presence of 2 mM CntA, 2 mM CntB, 300 mM L-carnitine, and 200 mM NADH using an extinction coefficient of e 340 = 6.2 mM 21 cm 21 for NADH (53). The specific activity was determined by calculation of the slope in the linear range of the assay. Control reactions in the absence of CntA or CntB or under strict anaerobic conditions (see above) were performed.

Flavin cofactor determination by HPLC
A purified CntB sample (50 mM) in a total volume of 300 ml was heat denatured at 99°C for 10 min. The liberated cofactor was separated from the denatured protein by centrifugation (16,000 3 g, 10 min at 22°C). 10 ml of the supernatant was analyzed on a Reprosil 100 C 18 column (Techlab) on an HPLC system equipped with a pump (PU-1580, Jasco), a column oven (X-LC, Jasco), a multiwavelength detector (MD-1515, Jasco), and a fluorescence detector (FP-1520, Jasco). Isocratic separation was performed at a flow rate of 0.5 ml min 21 at 30°C using the following mobile phase: 14% (v/v) acetonitrile, 1.5% (v/v) TFA, and 0.09% (v/v) of a 75% (w/w) phosphoric acid in water. Authentic FAD or FMN samples (Sigma-Aldrich) were used as a standard, and the related elution profiles were normalized according to maximum signal of the CntB sample.

FMN cofactor quantification
The FMN content of purified CntB samples was quantified by UV-visible absorption spectroscopy using an extinction coefficient of e 450 = 12.2 mM 21 cm 21 for free FMN on a V-650 UV-Vis spectrophotometer (Jasco). FMN was liberated from the protein by heat denaturation as described above.

Analytical size exclusion chromatography
The oligomeric state of CntA and CntB was determined by analytical size exclusion chromatography using a Superdex 200 increase 5/150 GL column (GE Healthcare). The column was calibrated with protein standards (molecular weight marker kit MWGF 1000; Sigma) in the presence of buffer 1 at a flow rate of 0.45 ml min 21 . Samples of CntB (20 ml) at a concentration of 350 mM or of CntA (20 ml) at a concentration of 280 mM were injected. Elution was monitored at 280 and 320 nm, respectively.

SAXS experiments
SAXS experiments were carried out on beamline BM29 (54) of the European Synchrotron Radiation Facility. A CntA sample with a concentration of 430 mM was exposed to an X-ray energy of 12,500 eV at a temperature of 293 K while steadily flowing through a quartz capillary. The scattering intensities were recorded with a Pilatus 1M hybrid photon counting detector within a momentum transfer range of 0.035-5 nm 21 . A total of ten scattering images were collected with an exposure time of 0.5 s per image. The scattering data were processed with programs from the ATSAS suite (55). The scattering images were appraised for apparent effects of radiation damage and subsequently normalized, averaged, and baseline corrected through the subtraction of buffer scattering with the program PRIMUS (56). Theoretical scattering intensities for the trimer and the monomer of CntA were calculated and fitted to the experimental data with the program CRYSOL (57). CRYSOL reports x 2 values as an indicator for the general agreement of theoretical and experimental curves, which are calculated by with N being the number of individual readout data points. For a more thorough assessment of the quality of fit throughout the resolution range, we calculated the residual DI=sI ¼ I obs À I calc ð Þ =sI obs which was plotted as a function of momentum transfer/resolution. The best agreement is reflected by the smallest x 2 value

Preparation of EPR samples for WT and variant CntA and CntB proteins
Purified proteins were concentrated up to 1 mM using an Amicon Ultra-0.5 centrifugal filter unit (Millipore) equipped with a 30,000-Da cutoff membrane. Sodium dithionite was added as indicated. After 10 min of incubation, samples were transferred into 3-mm EPR tubes and frozen in liquid nitrogen.

Interaction of CntA and CntB
Theoretical interaction of CntA and CntB was explored using the existing pGEX-6P-1-cntA protein production vector in combination with the cntB gene with an additional ribosome binding site. Therefore, cntB was cloned into vector pGEX-6P-1-cntA using the primers CACCAGTAAGTCGACTTTCA-CACAGGAAACAGTATTCATGGCCAGCCAC (including ribosome binding site, indicated in boldface) and CAGTC-ACGATGCGGCCGCTTACAGATCCAGAACCAG. Parallel overproduction of CntA and CntB and the purification of the bait protein CntA was accomplished as described above. SDS-PAGE analysis was used to indicate potential copurification of CntA and CntB.
Alternatively, purified CntA or CntB was immobilized on the respective affinity material (0.3 ml resin). The bait proteins were incubated with the potential pray proteins (purified CntB or CntA) at a concentration of ;50 mM for 5 min. After a subsequent washing step (5 ml buffer 1), the respective target proteins were eluted (as described above) and the related fractions were subjected to SDS-PAGE analysis.
EPR spectroscopy X-Band EPR spectra were recorded on a Bruker Elexsys E-500 CW X-band spectrometer. The sample tubes of 3-mm outer diameter were placed in a standard TE102 resonator. Low-temperature measurements were obtained using an Oxford ESR 900 helium flow cryostat (3-300 K). Baseline corrections were performed by subtracting a background spectrum, obtained under the same experimental conditions, from an empty tube.
Q-band experiments were performed on a Bruker ELEXYS E580 spectrometer with a SuperQ-FT microwave bridge (5-W output power) using a home-built resonator described earlier (58). Cryogenic temperatures (10-30 K) were obtained by a closed-cycle helium flow cryostat (Cryogenic Ltd.). ENDOR experiments were performed using a 300-W ENI 300 L RF amplifier.

Single-turnover experiments
Single-turnover experiments were performed under strict anaerobic conditions in an anaerobic chamber (Coy Laboratories) with an oxygen partial pressure below 1 ppm (oxygen detector, Coy Laboratories). Buffers were N 2 saturated prior to use. CntB was purified according to the standard procedure described above. The on-column immobilized protein (;1 mM) was incubated in the presence of 2 mM NADH for 10 min which resulted in a clear bleaching of the immobilized sample. After an excessive washing step (6 ml buffer 1), the reduced protein was eluted in 2 ml buffer 1 containing 50 mM imidazole. The resulting protein fraction was concentrated (;1 mM, 20-30 ml), transferred into 3 mm EPR tubes, and frozen in liquid nitrogen.
In related experiments, on-column immobilized CntB (after NADH reduction and extensive washing) was incubated in the presence of 1 mM CntA for 10 min. Subsequently, both proteins were specifically eluted. CntA was recovered in the presence of 2 ml buffer 1. Subsequently, CntB was liberated with 2 ml buffer 1 containing 50 mM imidazole. Both fractions were concentrated (;1 mM, 20-30 ml), transferred into 3-mm EPR tubes, and frozen in liquid nitrogen.

Data availability
All data are contained within the present manuscript. original draft; M. M., E. R., J. K., W. L., D. J., and J. M. project administration.
Funding and additional information-E.R. and W.L. thank the Max Planck Society for its continuous financial support.