A catalytically versatile benzoyl-CoA reductase, key enzyme in the degradation of methyl- and halobenzoates in denitrifying bacteria

Class I benzoyl-CoA (BzCoA) reductases (BCRs) are key enzymes in the anaerobic degradation of aromatic compounds. They catalyze the ATP-dependent reduction of the central BzCoA intermediate and analogues of it to conjugated cyclic 1,5-dienoyl-CoAs probably by a radical-based, Birch-like reduction mechanism. Discovered in 1995, the enzyme from the denitrifying bacterium Thauera aromatica (BCRTar) has so far remained the only isolated and biochemically accessible BCR, mainly because BCRs are extremely labile, and their heterologous production has largely failed so far. Here, we describe a platform for the heterologous expression of the four structural genes encoding a designated 3-methylbenzoyl-CoA reductase from the related denitrifying species Thauera chlorobenzoica (MBRTcl) in Escherichia coli. This reductase represents the prototype of a distinct subclass of ATP-dependent BCRs that were proposed to be involved in the degradation of methyl-substituted BzCoA analogues. The recombinant MBRTcl had an αβγδ-subunit architecture, contained three low-potential [4Fe-4S] clusters, and was highly oxygen-labile. It catalyzed the ATP-dependent reductive dearomatization of BzCoA with 2.3–2.8 ATPs hydrolyzed per two electrons transferred and preferentially dearomatized methyl- and chloro-substituted analogues in meta- and para-positions. NMR analyses revealed that 3-methylbenzoyl-CoA is regioselectively reduced to 3-methyl-1,5-dienoyl-CoA. The unprecedented reductive dechlorination of 4-chloro-BzCoA to BzCoA probably via HCl elimination from a reduced intermediate allowed for the previously unreported growth of T. chlorobenzoica on 4-chlorobenzoate. The heterologous expression platform established in this work enables the production, isolation, and characterization of bacterial and archaeal BCR and BCR-like radical enzymes, for many of which the function has remained unknown.

Aromatic compounds are the second most abundant class of naturally occurring organic molecules that are predominantly produced by plants and microorganisms. Aromatic hydrocar-bons of fossil oil reservoirs serve as solvents and precursors of plastics, dyes, resins, insecticides, herbicides, and pharmaceuticals, many of which are harmful to the environment and human health (1). The complete degradation of aromatic compounds by aerobic microorganisms heavily depends on oxygenases and has been studied in detail for more than six decades (2,3). Such a strategy is not an option at anoxic sites, e.g. at marine or freshwater sediments, aquifers with high carbon loads, or rice fields. In these environments, anaerobic bacteria channel monocyclic aromatic growth substrates into the central intermediate benzoyl-CoA (BzCoA). 2 It serves as substrate for dearomatizing benzoyl-CoA reductases (BCRs) that dearomatize their substrate to cyclohexa-1,5-diene-1-carboxyl-CoA (1,5-dienoyl-CoA; see Fig. 1) (4). The redox potential of the BCR substrate/product couple (EЈ 0 ϭ Ϫ622 mV) is among the lowest in biology, and electron transfer from any physiological reductant to BzCoA, such as reduced ferredoxins (EЈ Ϸ Ϫ500 mV), has to be coupled to an exergonic reaction (5).
The analogous reaction in organic synthesis, referred to as the Birch reduction, involves the formation of radical intermediates at extremely low redox potentials with the first electron transfer yielding a radical anion being rate-limiting (EЈ 0 ϭ Ϫ1.9 V for a benzoic acid thioester (6)). The Birch reduction uses solvated electrons as electron donors that are generated by dissolving alkali metals in ammonia; an alcohol serves as proton source (7). Kinetic data with BCR favor a Birch-like mechanism (8,9).
There are two totally nonrelated classes of BCRs, suggesting they have independently evolved in nature. Class I BCRs predominantly occur in facultative anaerobic bacteria and couple BzCoA reduction to a stoichiometric ATP hydrolysis (two ATPs hydrolyzed to ADP ϩ P i per BzCoA reduced). So far, a class I BCR has exclusively been isolated from the denitrifying Thauera aromatica (BCR Tar ) more than 20 years ago (10). BCR Tar is a 170-kDa heterotetramer with an ␣␤␥␦-architecture, encoded by the bcrABCD genes (see Fig. 1) (11). Class II BCRs are found in obligate anaerobes (12,13). The active site of This work was supported by the German Research Foundation, Deutsche Forschungsgemeinschaft Project BO 1565/14-1. The authors declare that they have no conflicts of interest with the contents of this article. This article contains Figs. S1-S6 and Tables S1-S3. 1  class II BCRs harbors a tungstopterin cofactor (14), and endergonic electron transfer to the aromatic ring is suggested to be driven by a flavin-based electron bifurcation (4). Class I BCRs belong to the BCR/2-hydroxyacyl-CoA dehydratase (HAD) radical enzyme family, which are all composed of two functional modules (16). The dimeric electron-activating module (BcrAD or ␣␦-subunits in BCR Tar ) harbors an ATPbinding site in each subunit and contains a bridging [4Fe-4S] cluster (6). After reduction of the cluster by a reduced ferredoxin, ATP hydrolysis results in conformational changes initiating low-potential electron transfer to the two [4Fe-4S] clusters of the CoA ester-binding module (see Fig. 1). The CoA ester-binding module (BcrBC or ␤␥-subunits in BCR Tar ) catalyzes electron transfer to BzCoA. Although class I BCRs transfer two electrons and protons to the aromatic ring of BzCoA yielding the reduced 1,5-dienoyl-CoA product (Fig. 1), electron transfer in the redox-neutral dehydration of 2-hydroxyacyl-CoAs to enoyl-CoAs by HADs is only catalytically required (17,18). The crystal structure of a 2-hydroxyacyl-CoA dehydratase revealed that the CoA ester substrate is directly ligated to an active-site cluster via the carbonyl oxygen atom of the thioester substrate, suggesting an inner-sphere electron transfer (19). A similar mode of binding to an active-site [4Fe-4S] cluster seems plausible in the catalytic BcrB subunit.
Based on amino acid sequence similarities and subunit sizes, the Thauera and Azoarcus subclasses of class I BCRs are distinguished. Recently, a phylogenetically distinct third BCR sub-class has been proposed that comprises BCRs that are up-regulated during complete anaerobic degradation of 4-and 3-methylbenzoates (20,21). Although no BCR of this third subclass has been isolated and characterized so far, it was suggested that MBR-like enzymes are especially optimized for the conversion of para-and meta-substituted BzCoA analogues. In contrast, BCR Tar exhibits a decreased activity with meta-positioned substrates (10,22,23), whereas analogues with parasubstituents generally do not serve as substrates for BCR Tar (22). The only recently reported exception is BCR-catalyzed defluorination of 4-fluoro-BzCoA to BzCoA (24).
The BCR/HAD family comprises a large number of related enzymes of which only a few can unambiguously be assigned to true BCRs or HADs (16). Heterologous production allowed for the study of a number of HADs, promoting the elucidation of their structure and function (19,25). Attempts to establish a comparable tool for BCRs were less successful. A first promising step was achieved with the BCR from the hyperthermophilic euryarchaeon Ferroglobus placidus. Heterologous production of all four subunits in Escherichia coli yielded a recombinant BCR (BCR Fpl ) that exhibited BzCoA reducing activity, albeit 4 orders of magnitude below that of WT BCR Tar . This low activity was assigned to the Ͼ99% loss of the activesite subunits (26). Recently, functional heterologous expression of the ATP-binding module of BCR Fpl was achieved in the absence of the active-site subunits (27).
In this work, we developed a general platform for the heterologous production of the four subunits of class I BCRs and related enzymes. Using this tool, we produced recombinant BCR Tar as a proof of principle. Moreover, an archetypical MBR was heterologously produced and characterized as a class I BCR with a largely extended substrate preference. This enzyme allowed for the previously unreported growth of Thauera chlorobenzoica on 4-chlorobenzoate under denitrifying conditions.

Phylogenetic analysis of BCR/HAD family enzymes
The recent discovery of a new subclass of ATP-dependent BCRs putatively specific for the conversion of 3-or 4-methyl-BzCoA (20, 21) motivated us for an updated phylogenetic analysis of the designated active-site subunits of class I BCRs (referred to as BcrB or BzdO). Among the candidates identified in databases, only those that derived from gene clusters containing all four structural genes of BCRs were considered. We assessed the diversity by aligning the amino acid sequences from 64 putative BCRs and two HADs (Table S1). In agreement with earlier studies, the computation of evolutionary distances revealed that MBR-like enzymes do not affiliate with Thauera and Azoarcus subclass BCRs (20,21). Instead, they group with a separated cluster of class I BCRs from ␣,␤,␦-proteobacteria but also from a number of distinct phyla ( Fig. 2 and Table S1); this phylogenetic cluster is henceforth referred to as the MBR subclass of ATP-dependent BCRs.
The analysis identified a putative MBR from T. chlorobenzoica next to the putative 4-methylbenzoyl-CoA reductase from Magnetospirillum sp. pMbN1 (20) and the 3-methylbenzoyl-CoA reductase from Azoarcus sp. CIB (21). Similar to The electron-activating module is depicted in gray, and the BzCoA-reducing module is depicted in white. The electron transfer from the donor-reduced ferredoxin (15) to BzCoA is schematically indicated by arrows. Similarities with 2-hydroxyacyl-CoA dehydratases suggest that the cluster in BcrB binds BzCoA. AH, proton donor for BzCoA reduction.

Heterologous production and isolation of BCR Tar and MBR Tcl
For the development of a heterologous expression platform, the BCR Tar -encoding genes bcrABCD and the MBR Tcl -encoding genes mbrONPQ (also referred to as mbdONPQ in Azoarcus sp. CIB (21)) were chosen. The corresponding DNA sequences were cloned into the midcopy plasmid pOT1 constructed in a recent work (28). This plasmid permits inducible gene expression at moderate levels in E. coli to minimize misfolding of proteins. A C-terminal Strep-Tag II was fused to the individual ␦-subunits. After gene expression in E. coli under anaerobic growth conditions, the proteins produced were largely enriched by affinity chromatography (Fig. 3). The yields were 0.4 -0.6 mg of protein/g of cells.
The four subunits of MBR Tcl had an almost perfect 1:1:1:1 stoichiometry (Fig. 3A), whereas in preparations of BCR Tar the ␤and ␥-subunits of the CoA ester-binding module were frequently less represented than the ␣and ␦-subunits of the electron-activating module (Fig. 3B). This finding may result from stronger contacts between the modules in MBR Tcl than in BCR Tar under the enrichment conditions.

ATP-dependent BzCoA dearomatizing activities of recombinant BCR Tar and MBR Tcl
Heterologously produced BCR Tar and MBR Tcl both catalyzed the Ti(III) citrate-dependent (5 mM) reduction of BzCoA to 1,5-dienoyl-CoA as evidenced by coelution with authentic standards and typical UV/visible spectra. BzCoA reduction of both enzymes strictly depended on the presence of MgATP (5 mM); representative diagrams from ultraperformance LC (UPLC)-based assays for MBR Tcl are shown in Fig. 4. BCR Tar dearomatized BzCoA to 1,5-dienoyl-CoA at a rate of 93 milliunits min Ϫ1 mg Ϫ1 (1 milliunit ϭ 1 nmol min Ϫ1 mg Ϫ1 ). This value is 3-4 times slower than that reported for the WT enzyme (10), which corroborates the corresponding loss of the active subunit during the enrichment. The specific activity of MBR Tcl (212 milliunits mg Ϫ1 ) was substantially higher, fitting to the almost stoichiometric presence of all four subunits. Reduced methyl viologen, which routinely served as electron donor for WT BCR Tar in continuous spectrophotometric assays, was also used by recombinant BCR Tar at rates comparable with that of Ti(III) citrate. However, it barely served as electron donor for MBR Tcl (1.5% of the rate with Ti(III) citrate). Addition of 0.5 mM methyl viologen to an assay with 5 mM Ti(III) citrate as electron donor enhanced BzCoA reduction rate by 10%, indicating that methyl viologen acts as an electron carrier between Ti(III) citrate and MBR Tcl . Sodium dithionite (5 mM) served as an alternative electron donor at 100 (BCR Tar ) and 53% (MBR Tcl ) of the rate observed with Ti(III) citrate. With dithionite as electron donor, UPLC analyses revealed a different product pattern, an observation that was also reported for the conversion of BzCoA by WT BCR Tar (29). This altered pattern results from artificial effects of dithionite and its oxidized product sulfite, yielding physiologically irrelevant CoA ester products that were not further analyzed in this work. NAD(P)H did not serve as electron donor for MBR Tcl . Based on the results obtained, Ti(III) citrate was routinely used as electron donor in UPLC-based, discontinuous MBR Tcl assays.

Native molecular mass, cofactor content, and spectral properties of MBR Tcl
The partial loss of the ␤and ␥-subunits precluded a detailed analysis of the molecular and kinetic properties of BCR Tar , and they have already been described in detail for WT BCR Tar (10). The molecular mass of MBR Tcl was 140 Ϯ 1 kDa (mean Ϯ S.D.  Table  S1. The asterisk marks MBR from T. chlorobenzoica.

Catalytically versatile methylbenzoyl-CoA reductase
of three biological replicates) as determined by size exclusion chromatography, suggesting a heterotetrameric composition. This value is only slightly lower than that deduced from the amino acid sequence (152 kDa) (25).
The iron content was 11.6 Ϯ 1.0 irons per MBR Tcl as determined by a spectrophotometric assay, suggesting the presence of three [4Fe-4S] clusters as described for BCR Tar (30). Supernatants of acid-precipitated MBR Tcl were analyzed by UPLC coupled to UV/visible absorbance detection. The results obtained excluded the presence of a flavin or other organic cofactors to a significant extent (Յ0.02 mol/mol of enzyme).
The UV/visible absorbance spectrum of oxidized MBR Tcl showed a major maximum at 407 nm and a minor maximum at 317 nm (Fig. 5). The spectrum was bleached in the visible region upon reduction by sodium dithionite at pH 8.3 where EЈ 0 of dithionite is below Ϫ600 mV (31). Titration of MBR Tcl with dithionite at this pH revealed an almost linear course of absorbance decrease at 407 nm up to the addition of one electron equivalent, suggesting that one [4Fe-4S] 2ϩ/ϩ cluster was readily reduced (Fig. 5C). In this range, the difference spectrum of oxidized minus dithionite-reduced enzyme revealed a difference absorbance coefficient of ⌬⑀ 407 Ϸ 5.0 mM Ϫ1 cm Ϫ1 (Fig.  5B). Further addition of dithionite did not result in a further stoichiometric reduction of the spectrum at 407 nm, indicating that a redox equilibrium was approached (Fig. 5C). Upon the addition of a 300-fold excess of dithionite versus enzyme, the maximal difference in absorbance was reached with ⌬⑀ 407max ϭ 9.6 mM Ϫ1 cm Ϫ1 , indicating the reduction of two of the three [4Fe-4S] 2ϩ/ϩ clusters. The difference absorbance at 317 nm could not be resolved accurately due to the absorbance of accumulating dithionite.
Similar to BCR Tar , MBR Tcl activity was highest at pH Ϸ7.3 (Fig. S1A). Whereas BCR Tar activity was only measurable in a rather narrow pH range between 6 and 9 (10), MBR Tcl still exhibited 68% residual activity at pH 6.3 and 88% at pH 8.8.

Sensitivity of MBR Tcl toward oxygen
MBR Tcl was incubated in air for different time spans before the reaction was started with BzCoA under anaerobic conditions. The enzyme was severely susceptible to loss of activity upon contact with oxygen with a half-life time in air of around 30 s (Fig. S1B). This value is in the same range as half-life times of WT BCR Tar and the activator module of HAD from Acidaminococcus fermentans, which are 20 and 10 s, respectively (10,32).

Substrate preference of MBR Tcl
The substrate preference of MBR Tcl was investigated and compared with reported values for WT BCR Tar in UPLC-based discontinuous assays at 0.25 mM thioester concentrations (Table 1). For selected substrates, the K m , k cat , and k cat /K m values were determined. Most products were additionally subjected to electrospray ionization quadrupole (ESI-Q) TOF MS analysis (Figs. S2-S4).
MBR Tcl preferentially dearomatized meta-substituted BzCoA analogues containing methyl-, chloro-, or hydroxy-functionalities as indicated by the highest specificity constants determined. With these substrates, relative specific activities compared with BzCoA were substantially higher with MBR Tcl than with BCR Tar . Remarkably, MBR Tcl also converted para-substituted halo-and methyl-BzCoA analogues that were not converted by BCR Tar . Exceptions were 3-fluoro-and 4-fluoro-benzoyl-CoA that served as substrates for both enzymes. In summary, MBR Tcl appears to be less sensitive to steric effects of meta-and para-positioned substituents than BCR Tar ; even 4-bromo-BzCoA was readily converted. Neither of the enzyme converted 4-hydroxy-BzCoA. In contrast, orthosubstituted BzCoA analogues were generally accepted by both enzymes with BCR Tar showing higher relative activities than MBR Tcl . Neither of the two enzymes reduced heterocyclic nicotinoyl-CoA.

Catalytically versatile methylbenzoyl-CoA reductase
For further analysis of the product formed from 3-methyl-BzCoA, the preferred substrate of MBR Tcl , the enzyme was incubated with its substrate in D 2 O, and the product was analyzed by 1 H and 13 C NMR techniques. The NMR data obtained are summarized in Table S2; chemical shift maps for the two possible isomers formed are shown in Figs. S5 and S6. The NMR data clearly indicate that 3-methyl-1,5-dienoyl-CoA was specifically formed by MBR Tcl .

ATPase activity of MBR Tcl and stoichiometry of ATP hydrolysis
Reduction of BzCoA and analogues of it by MBR Tcl strictly depended on hydrolysis of ATP to ADP ϩ P i . In the presence of Ti(III) citrate and concurrent absence of a thioester substrate, the enzyme also exhibited ATPase activity, albeit at only ϳ17% of the rate coupled to substrate reduction (Fig. 6). The stoichiometry of ATP hydrolysis and BzCoA reduction was determined by taking samples from a running assay, and the concentrations of ATP and BzCoA were analyzed by UPLC analysis coupled to diode array detection in four biological replicates. Using this setup, 2.8 Ϯ 0.2 (mean Ϯ S.D.) ATPs were hydrolyzed to ADP ϩ P i per BzCoA reduced to 1,5-dienoyl-CoA. Assuming that BzCoA-dependent and -independent ATPase activities run in parallel, the stoichiometry between ATP hydrolysis and BzCoA reduction would be 2.3 Ϯ 0.2.
Oxygen-sensitive MBR Tcl was routinely stored in the presence of 50 M dithionite. Under anoxic conditions, the enzyme

Table 1 Substrate preference of MBR Tcl and BCR Tar and identification of the products formed
The relative activities were derived from specific activities determined in assays with 0.25 mM CoA ester substrate and are referenced to those obtained with BzCoA as substrate (100%). Specific activity of MBR Tcl with BzCoA was 212 milliunits mg Ϫ1 . Activity values of recombinant BCR Tar were determined in this work or are referred to wildtype BCR Tar taken from the references indicated in parentheses after the values. Reaction products were identified by UPLC analysis and, in the case of unknown retention times, and/or UV-visible absorbance spectra coupled to ESI-Q-TOF-MS detection (Figs. S2-S4). After prolonged incubation with Ti(III) citrate, 1,5-dienoyl-CoA and the fluorinated and methylated analogues were further reduced to corresponding monoenoyl-CoAs as reported earlier for BCR Tar (29

Catalytically versatile methylbenzoyl-CoA reductase
could reversibly be oxidized with a 10-fold excess of thionine with virtually no loss of activity. In the thionine-oxidized state, CoA ester-independent ATPase activity of MBR Tcl was greatly diminished to Ͻ1% of the rate observed in the presence of BzCoA and Ti(III) citrate. Re-reduction of thionine-oxidized enzyme by excess Ti(III) citrate fully restored ATPase activity to the level before thionine oxidation. This redox dependence of the thioester substrate-independent ATPase activity was also observed for BCR Tar but not for the isolated electron-activating module of the Azoarcus subclass BCR from F. placidus (27,33,34).

Complete degradation of 3-methyl-and 4-chlorobenzoate by T. chlorobenzoica
The MBR subclass of ATP-dependent BCRs was originally deduced from up-regulated bcr-like genes during anaerobic growth of Magnetospirillum sp. pMbN1 with 4-methylbenzoate (20) and Azoarcus sp. CIB with 3-methylbenzoate (21). The dearomatizing activities of MBR Tcl with 3-or 4-methyl-/ chloro-BzCoA analogues motivated us to test whether T. chlorobenzoica is capable of using the respective benzoate analogues as carbon and, together with nitrate, as energy sources.
After three passages, T. chlorobenzoica readily used 3-methylbenzoate and 3-chlorobenzoate as growth substrates under denitrifying conditions (Fig. 7). Doubling time with the former was 3.2 h and only slightly increased versus that observed with benzoate (2.8 h), whereas that with 3-chlorobenzoate was clearly higher (6.5 h). No growth with 4-methlybenzoate was observed even after incubation for several weeks. In contrast to previous reports (35,36), T. chlorobenzoica used 4-chlorobenzoate as a growth substrate under denitrifying conditions with a doubling time of around 18.6 h (Fig. 7). In all cases, an increase of optical density at 578 nm was observed until the substrate was completely consumed.

Discussion
By establishing the first heterologous production platform for functional class I BCRs, the recombinant MBR prototype of an only recently defined BCR subclass was isolated and biochemically characterized in this work. The properties are summarized and compared with the only other described BCR Tar isolated more than two decades ago ( Table 2). General properties, such as subunit architecture, cofactor content, ATP depen-dence/ATPase activity, and oxygen sensitivity, are similar in members of different BCR subclasses.
However, there are marked differences between the two enzymes with regard to substrate specificity. Although BCR Tar predominantly converts BzCoA and ortho-substituted analogues, MBR Tcl exhibits a clearly extended substrate preference toward meta-and para-substituted BzCoA analogues with methyl, chlorine, and bromine substituents. In particular, the para-position has previously been suggested as a critical position for class I BCRs (22), and so far, only the sterically irrelevant fluorine atom was accepted as a para-substituent by BCR Tar , albeit at drastically reduced rates (24).
The observed differential substrate preference of the BCR and MBR subclasses suggests structural variations with regard to the electronic settings of the benzene-binding pocket and the nature, position, and pK a values of amino acid residues involved in C3 and C4 protonation. The proposed radical anion intermediate exhibits highest electron density at the para-position (22). For this reason, any substituent that adds additional electron density at this position should have a counterproductive effect on BCR activity as demonstrated by the inability of all BCRs to reduce 4-hydroxy-BzCoA containing a para-substituent having a marked positive mesomeric effect. The positive inductive effect of alkyl substituents is less pronounced, and as a result MBR Tcl dearomatizes 4-methyl-BzCoA, albeit at a clearly reduced rate compared with BzCoA. In contrast, BCR Tar shows no activity with 4-methyl-BzCoA, which is most likely due to steric effects of the relatively bulky methyl moiety. The proposed steric hindrance of BCR Tar by bulky para-substituents is further corroborated by the ability to defluorinate 4-fluoro-BzCoAbut not the bulky 4-chloro-and 4-bromo-Bz-CoA analogues with good leaving groups. In contrast, MBR Tcl converts para-substituted BzCoA analogues in the order of bromo Ͼ chloro Ͼ fluoro as expected from the order of bond dissociation energies (37).
The reductive dehalogenation of 4-chloro-/4-bromo-BzCoA to BzCoA and HCl has not been reported previously for a BCR and accounts for further ATP-dependent dehalogenation reaction of BCRs in addition to the dehalogenation of 3-chloro-/3-bromo-BzCoA (23) and 4-fluoro-BzCoA (24). The dehalogenation reactions most likely proceed via reduced 4-chloro-/4-bromo-1,5-dienoyl-CoA intermediates that, simi-  Catalytically versatile methylbenzoyl-CoA reductase lar to 3-chloro-1,5-dienoyl-CoA, will spontaneously eliminate HCl/HBr, driven by rearomatization to BzCoA. This assumption is supported by the determination of traces of a compound by ESI-Q-TOF-MS with ion spectra fitting to a chloro-1,5-dienoyl-CoA intermediate. As an alternative, dehalogenation may occur on the level of the one electron-reduced aryl radical anion intermediate. Such an S RN 1 mechanism has recently been proposed for aminofutalosine synthase during debromination of a substrate analogue (38).
The preferred substrate of MBR Tcl , 3-methyl-BzCoA, was predominantly reduced to the 3-methyl-1,5-dienoyl-CoA isomer, demonstrating the regioselectivity of the enzyme toward the meta-positioned substituent. This finding is in full accordance with the observed complete reductive dehalogenation of 3-chloro-BzCoA to BzCoA and HCl in both MBR Tcl and BCR Tar (23). The latter reaction can only occur via a 3-chloro-1,5-dienoyl-CoA but not via a 5-chloro-1,5-dienyol-CoA intermediate. No regioselectivity was observed for the conversion of 3-fluoro-BzCoA where both 3-and 5-fluoro-1,5-dienyol-CoA products were identified in a recent study (28). Therefore, obviously steric effects govern the regioselectivity of BCRs toward meta-positioned functionalities.
This work demonstrates that the previously unknown capability of a BCR to reduce 4-chloro-BzCoA to BzCoA and HCl allows for growth with 4-chlorobenzoate under denitrifying conditions. Growth with this halobenzoate in aerobic bacteria has been demonstrated already 25 years ago (39 -42). Interestingly, the aerobic degradation also proceeds via 4-chloro-Bz-CoA, but in this case, dehalogenation is achieved by hydrolysis to 4-hydroxy-BzCoA rather than by reduction, as demonstrated in this work. The 4-hydroxy-BzCoA dehalogenase involved catalyzes a nucleophilic substitution reaction via a Meisenheimer complex, and its structure and function has been studied on the molecular level (43)(44)(45). Under anaerobic conditions, complete mineralization of 4-chloro-benzoate has so far only been described for enrichment cultures, often comprising ␤-proteobacteria (36, 46 -49). Thus, T. chlorobenzoica appears to be the first pure culture described with the capability to use 4-chlorobenzoate anaerobically as a growth substrate.
Dechlorination of 4-chloro-BzCoA via ATP-dependent MBR, instead of using a 4-chloro-BzCoA hydrolase, appears at first sight as an unnecessary consumption of ATP. However, MBR converts 4-chloro-BzCoA directly to the central interme-diate BzCoA and circumvents the biosynthesis of 4-hydroxy-BzCoA reductase that would be required for reductive dihydroxylation. This complex enzyme of the xanthine oxidase family depends on numerous redox cofactors, such as molybdopterin, FAD, and three Fe-S clusters (11, 50 -52). Finally, the expense of two ATPs for the reductive dehalogenation appears to be rather marginal relative to the overall high ATP yield achieved by the oxidation of benzoate analogues under denitrifying conditions (53).
The capability of MBR Tcl to reduce 4-methyl-BzCoA suggests that T. chlorobenzoica is also capable of using 4-methylbenzoate as a growth substrate as reported for Magnetospirillum sp. pMbN1 (20). However, growth of T. chlorobenzoica with 4-methylbenzoate was not observed, probably due to the lack of a 4-methylbenzoate-CoA ligase and/or downstream enzymes of the 4-methylbenzoate degradation pathway.
In contrast to BCR Tar , MBR Tcl did not accept reduced methyl viologen as an electron donor, which prevented the use of a continuous spectrophotometric assay monitoring the oxidation of colored reduced methyl viologen. The rationale for this finding could be that the redox potential of reduced viologen (EЈ 0 ϭ Ϫ446 mV (31)) might be too positive to donate electrons to MBR Tcl . Methyl viologen did not act as an inhibitor as the reaction catalyzed by MBR Tcl with Ti(III) citrate as electron donor was stimulated by 0.5 mM methyl viologen.
The stoichiometry of ATP hydrolyzed per electrons transferred is controversial for members of the BCR/HAD family. For 2-hydroxyacyl-CoA dehydratases, a stoichiometry of two ATP hydrolyzed per one electron transferred has been suggested (17). In contrast, for BCR Tar , stoichiometries of 2.1-2.2 ATPs hydrolyzed per BzCoA reduced were determined (33), and for MBR Tcl this study revealed a slightly higher stoichiometry (2.3-2.8 ATPs hydrolyzed per BzCoA reduced). Taken together, experimental data point to one rather than two ATPs hydrolyzed per electron transferred in class I BCRs, assuming that each electron is transferred in an ATP hydrolysis-dependent manner. The question arises, why the stoichiometries determined for BCRs are always slightly higher than two ATPs per BzCoA? One possible explanation could be that ATP hydrolysis is not 100% coupled to BzCoA reduction and that a parallel futile background ATP hydrolysis activity exists. Such a scenario is in line with the first electron transfer to the substrate being rate determining in the overall BCR Tar reaction rather than ATP hydrolysis (22); it would theoretically allow for a noncoupled ATP hydrolysis to some extent. As an alternative explanation, a portion of the active-site module of BCR might be inactive, e.g. due to partial cluster degradation. In such a scenario, the apparent ratio of ATP hydrolysis uncoupled versus coupled to BzCoA reduction would be artificially increased.
The expression platform for genes encoding BCRs established in this work paves the way for the heterologous production of many other BCR and BCR-like enzymes within the BCR/ HAD radical enzyme family. Many BCR-like proteins that do not grow with aromatic compounds under anaerobic conditions are present in bacteria (16), and their function has remained unknown. Very recently, the crystal structure of an ATP-dependent, BCR-like enzyme has been solved that contained two [4Fe-4S] clusters bridged by an inorganic sulfur Catalytically versatile methylbenzoyl-CoA reductase ligand. The enzyme did not contain a CoA ester-binding pocket but instead reduced a number of small molecules, such as acetylene and azide (54). Thus, the catalytic versatility of BCRs and related enzymes may be much larger than originally anticipated.

Phylogenetic analyses
Amino acid sequences of 66 BcrB Tar homologues of the BCR/ HAD family (Table S1) were aligned using MUSCLE (55) in MEGA 6.0.6 software. The evolutionary history was inferred by using the maximum likelihood method based on the JTT matrix-based model (56). The tree with the highest log likelihood (Ϫ16,445.6661) was depicted. The initial tree(s) for the heuristic search were obtained automatically by applying Neighbor-Join and BioNJ algorithms to a matrix of pairwise distances estimated using a JTT model and then selecting the topology with superior log likelihood value. The tree was drawn to scale with branch lengths measured in the number of substitutions per site. No branch swap filter was applied. All positions containing gaps and missing data were eliminated. Evolutionary analyses were conducted in MEGA 7.0 software.

Growth of bacterial cells
T. chlorobenzoica was cultivated, and the concentrations of growth substrates were determined as described for T. aromatica in a recent work (24). The initial concentrations of growth substrates and nitrate were 1.3 and 6.0 mM, respectively.

Heterologous gene expression in E. coli
Primers for the amplification of the bcrABCD and mbrONQP gene clusters were derived from partial genome sequences of T. aromatica and T. chlorobenzoica (gi numbers 19571177 and 1341820151, respectively) (Table S3). Reverse primers were elongated for C-terminal fusions of the ␦-subunits with Strep-Tag II (SA-WSHPQFEK). Cloning of DNA sequences using pOT1 expression vector was conducted as described recently (28).
Gene expression was carried out in E. coli MC4100 (57) growing anaerobically in 2 liters of TB medium (89 mM KH 2 PO 4 /K 2 HPO 4 at pH 7.5, 12 g liter Ϫ1 Tryptone, 24 g liter Ϫ1 yeast extract, 0.4% (v/v) glycerol, 5 mM NaNO 3 , 50 mM fumarate, 0.2 g liter Ϫ1 Fe(III) citrate, 75 g ml Ϫ1 kanamycin A). After inoculation, flasks were sealed with rubber stoppers, and air in the headspace was replaced by flushing with N 2 gas. Cells were incubated at 37°C until reaching an optical density of 0.4 -0.6 whereupon gene expression was induced with 1.0 mM isopropyl ␤-D-1-thiogalactopyranoside, and growth medium was supplemented with 1 ml liter Ϫ1 vitamin solution VL-7 (58), trace element solution SL9 (59), and 0.8 mM MgSO 4 . Cells were incubated for another 10 -12 h at 30°C with agitation at 120 rpm. The pH was monitored and frequently adjusted to values Ͼ7.0. Cells were harvested anaerobically by centrifugation at 4,500 ϫ g (4°C) and immediately further processed or stored in liquid nitrogen until use.

Purification of recombinant BCRs
Purification procedures were carried out under anaerobic conditions (95% N 2 , 5% H 2 ) at 4°C. Cells were resuspended (1.5 ml of buffer/g of cells) in a 20 mM HEPES/NaOH standard buffer (pH 7.8) containing 200 mM KCl, 4 mM MgCl 2 , 50 mM L-arginine, 50 mM L-glutamate, 10% (v/v) glycerol, 1 mM dithioerythritol, 200 M dithionite, and ϳ0.1 mg liter Ϫ1 DNase I. Cell suspensions were passed twice through a French pressure cell at 120 megapascals. The cell lysate was centrifuged at 150,000 ϫ g for 1 h prior to applying the supernatant onto a Strep-Tactin Superflow High Capacity column (IBA) according to the manufacturer's instructions. After washing with DNasefree buffer containing 100 mM KCl and 50 M dithionite, proteins were eluted at pH 7.3 with 3 mM desthiobiotin. Proteins were concentrated using Vivaspin Turbo 4 centrifugal concentrators (30,000 molecular weight cutoff) (Sartorius) to final concentrations of 5-30 mg ml Ϫ1 and stored anaerobically at Ϫ80°C.

Estimation of the molecular mass of MBR Tcl
The molecular mass was estimated by size exclusion chromatography applying 100 l of a 3 mg ml Ϫ1 protein solution to a 24-ml Superdex 200 Increase 10/300 GL column (GE Healthcare) in 20 mM HEPES buffer with 150 mM NaCl, 4 mM MgCl 2 , 1 mM dithioerythritol, and 50 M dithionite. Calibration was performed using 100 l of standard solutions containing thyroglobulin (bovine thyroid) (M r ϭ 669,000), alcohol dehydrogenase (Saccharomyces cerevisiae) (M r ϭ 150,000), carbonic anhydrase (M r ϭ 29,000), and cytochrome c (equine heart) (M r ϭ 12,400), each at 3 mg ml Ϫ1 .

Synthesis of CoA thioesters
BzCoA was synthesized from benzoic acid anhydride and CoA as described (60). Substituted BzCoA analogues were synthesized from the corresponding acids via their succinimidyl esters as described (61).

Determination of protein and iron content
Protein concentrations were routinely determined by the method of Bradford (63) using BSA solutions as a standard. The iron content was determined as described recently (27). Amounts of the heterotetrameric protein applied for determination were 0.5-3 nmol.

UV/visible spectroscopy
UV/visible spectra were recorded under anaerobic conditions in a glove box. MBR Tcl as isolated was pretreated as a mixture of 25.5 M enzyme plus 125 M thionine acetate as mild oxidant in 20 mM HEPES buffer containing 4 mM MgCl 2 and 10% (v/v) glycerol. Thionine was then removed by applying to a PD MiniTrap G-25 desalting column (GE Healthcare) according to the manufacturer's instructions. During the recording of UV/visible spectra, 1 ml of MBR Tcl diluted to 4.2 M was reduced by addition of dithionite from a 0.5 mM stock solution in 2.5-l steps. After the addition of 22.5 l, the steps were raised to 10 l, and after the total addition of 72.5 l, a 5 mM stock solution was used instead.

Catalytically versatile methylbenzoyl-CoA reductase Determination of BCR/MBR activities
Enzyme activities were determined, and CoA ester intermediates were identified in a discontinuous HPLC/UPLC (Waters)-based assay as described earlier routinely using 5 mM Ti(III) citrate as artificial electron donor (24,29). Dithionite, NADH, and dithionite-reduced methyl viologen as alternative electron donors were applied at concentrations of 5, 1, and 1 mM, respectively. For the determination of the pH dependence of enzyme activities, MOPS as standard reaction buffer was substituted by MES (pH 6.3), TES (pH 7.8), or TAPS (pH 8.3/ 8.8). Studies of oxygen sensitivity were conducted by aerobic incubation of enzyme solutions for different periods in shaking 1.5-ml Eppendorf tubes before the immediate start of reactions under anoxic conditions. Reactions were stopped by 2 volumes of methanol, and centrifuged supernatants were subjected to C 18 reversed-phase UPLC (Waters). For the detection of lower CoA ester concentrations during K m value determinations, reactions were stopped by the addition of 0.08 volume of 1% H 2 SO 4 , instead. For determination of catalytic constants, the initial rates determined at different substrate concentrations were fitted to Michaelis-Menten curves using the Prism 6 software package (GraphPad). Activity of BCR Tar was also determined in a continuous spectrophotometric assay with dithionite-reduced methyl viologen serving as the electron donor as described before (10).

Determination of ATPase activities and stoichiometry of ATP hydrolysis/BzCoA reduction
MBR Tcl , oxidized as described above for UV/visible spectroscopy, was incubated at a final concentration of 2 M with 5 mM ATP and 15 mM MgCl 2 in 100 mM MOPS buffer (pH 7.3) at 30°C. Samples were taken at different time points from the running assays and stopped by the addition of 2 volumes of a 1 M HCl solution containing 10% acetonitrile. For determination of the stoichiometry of ATP hydrolysis and BzCoA reduction, BzCoA was present in the assays. Samples for BzCoA determination were stopped in 66% methanol (v/v, final concentration).
Analysis of ATP/ADP and BzCoA by UPLC was performed using a Waters C 18 HSS T3 column (2.1 ϫ 100 mm, 1.8-m particle size) at 25°C. ATP/ADP were separated isocratically in 10 mM potassium phosphate buffer (pH 6.5) as the sole component in the mobile phase at a flow rate of 0.2 ml min Ϫ1 for 5 min. For detection of BzCoA, the acetonitrile content in 10 mM potassium phosphate buffer (pH 7) was gradually increased at a flow rate of 0.21 ml min Ϫ1 from 2 to 10% within 1.5 min followed by an increase from 20 up to 30% within 1.6 min. Products were identified by comparing retention times and UV/visible absorption spectra with standards and/or additionally by subsequent MS analysis (see below).

Analysis of CoA esters by MS
LC/MS analysis of CoA esters was performed with a Waters Acquity I class UPLC using a Waters C 18 HSS T3 column (2.1 ϫ 100 mm, 1.8-m particle size) coupled to a Waters Synapt G2-Si HDMS ESI/Q-TOF system. For separation, a 6-min linear gradient from 2 to 30% acetonitrile in 10 mM ammonium acetate (pH 6.8) at a flow rate of 0.35 ml min Ϫ1 was applied. The mass spectrometer was operated in MS negative or positive mode with a capillary voltage of 2.5 or 3.0 kV, 120 or 150°C source temperature, 400 or 450°C desolvation temperature, and 750 or 1,000 liters h Ϫ1 N 2 desolvation gas flow, respectively. In MS negative mode, additionally cone gas flow was applied at a rate of 100 liters h Ϫ1 .

Analysis of products formed from 3-methylbenzoyl-CoA by NMR spectroscopy
The assay for NMR spectroscopic analyses comprised 0.5 mM 3-methyl-BzCoA, 5 mM Ti(III) citrate, 5 mM MgATP, and 2.5 M MBR Tcl in 100 mM MOPS buffer (pH 7.3) with 15 mM MgCl 2 in 95% D 2 O, and the reaction mixture had a total volume of 22.7 ml. Reactions were stopped by addition of 66% methanol (v/v, final content), which was removed after subsequent centrifugation from the supernatant by flash evaporation at 40°C and before freeze-drying the residual solution. CoA thioesters were purified by reversed-phase HPLC following procedures described previously (61). Purified compounds were desalted by solid-phase extraction as described earlier (62).
Prior to analysis, the compounds were dissolved in 0.5 ml of deuterated water. 1 H NMR and 13 C NMR spectra were recorded at 500 and 126 MHz, respectively, with Avance-HD 500 spectrometers operating at 27°C. 1 H-detecting experiments, including two-dimensional COSY, NOESY, heteronuclear single quantum coherence, and heteronuclear multiple bond correlation, were measured with an inverse 1 H/ 13 C probe head; direct 13 C measurements were performed with a quattro nucleus 13 C/ 31 P/ 29 Si/ 19 F/ 1 H cryoprobe. All experiments were done in full automation using standard parameter sets of the TOPSPIN software package (Bruker). 13 C NMR spectra were recorded in proton-decoupled mode. Data processing was typically done with MestreNova software.