ATP-dependent hydroxylation of an unactivated primary carbon with water

Enzymatic hydroxylation of unactivated primary carbons is generally associated with the use of molecular oxygen as co-substrate for monooxygenases. However, in anaerobic cholesterol-degrading bacteria such as Sterolibacterium denitrificans the primary carbon of the isoprenoid side chain is oxidised to a carboxylate in the absence of oxygen. Here, we identify an enzymatic reaction sequence comprising two molybdenum-dependent hydroxylases and one ATP-dependent dehydratase that accomplish the hydroxylation of unactivated primary C26 methyl group of cholesterol with water: (i) hydroxylation of C25 to a tertiary alcohol, (ii) ATP-dependent dehydration to an alkene via a phosphorylated intermediate, (iii) hydroxylation of C26 to an allylic alcohol that is subsequently oxidised to the carboxylate. The three-step enzymatic reaction cascade divides the high activation energy barrier of primary C–H bond cleavage into three biologically feasible steps. This finding expands our knowledge of biological C–H activations beyond canonical oxygenase-dependent reactions.

T he selective oxidation of unactivated C-H bonds at alkyl functionalities to alcohols is of great importance for a plethora of synthetic processes. Enzymatic solutions for these challenging reactions have continuously motivated organic chemists for developing bioinspired strategies [1][2][3][4][5] . Biocatalytic C-H bond activations of alkyls via hydroxylation have generally been associated with metal-dependent monooxygenases, peroxidases or peroxygenases [6][7][8] . As an example, the well-studied cytochrome P450 monooxygenases reduce the dioxygen cosubstrate to the formal oxidation state of a hydroxyl radical, which allows for the hydroxylation of unactivated primary carbons to alcohols 9 . While cytochrome P450 enzymes require an auxiliary electron donor system, peroxygenases and peroxidases depend on a balanced supply with its reactive co-substrate 6,8 .
The use of oxygenases or peroxygenases is not an option for C-H bond oxidations in anaerobic hydrocarbon-degrading microorganisms, for which a few oxygen-independent enzymatic solutions for C-H bond hydroxylation at activated positions have been discovered in the past decades 10 . Prototypical examples are the water-dependent hydroxylations of the benzylic carbons of p-cresol 11,12 , p-cymene 13 , and ethylbenzene 14,15 , as well as of the allylic carbon of limonene 16 . Here, C-H bond activation proceeds via hydride transfer to FAD or Mo(VI) = O cofactors ( Fig. 1a-d). In all these reactions the resulting carbocation intermediates are stabilised by the possibility of forming multiple resonance structures. However, the water-dependent hydroxylation of unactivated primary carbons of alkanes or isoprenoids has not been reported so far in enzyme catalysis.
The ubiquitous, biologically active steroids are composed of a tetracyclic steran system and an isoprenoid side chain resulting in a low water-solubility and persistence in the environment. Bacterial degradation of steroids is of global importance for biomass decomposition, removal of environmental pollutants and for intracellular survival of Mycobacterium tuberculosis and other pathogens [17][18][19] . In aerobic cholesterol degrading bacteria, side chain oxidation is initiated by the hydroxylation of the unactivated primary C26 (or equivalent C27) to a primary alcohol by cytochrome P450 enzymes that is further oxidised to a C26carboxylate 20,21 . The isoprenoid side chain of the latter is then converted to acetyl-CoA and propionyl-CoA units via modified β-oxidation 22,23 . In denitrifying bacteria, oxygen-independent cholesterol degradation also proceeds via β-oxidation of the 26carboxylate intermediate involving highly similar enzymes [24][25][26] . However, the oxidation of the primary, unactivated C26 to the 26-carboxylate must proceed in an oxygen-independent manner.
In the denitrifying β-proteobacterial model organism Sterolibacterium denitrificans Chol-1S, anaerobic cholesterol degradation is initiated by the periplasmic isomerization/ dehydrogenation of ring A to cholest-4-en-3-one (CEO) by the AcmA gene product 27 , that may be further dehydrogenated to cholesta-1,4-dien-3-one (CDO) by AcmB 25 . Both intermediates serve as substrate for steroid C25 dehydrogenases (S25DH) that hydroxylates tertiary C25 of the isoprenoid side chain with water to 25-OH-CEO/-CDO in the presence of cytochrome c or an artificial electron acceptor 28 (Fig. 1e). Periplasmic S25DH belongs to the DMSO reductase family of Mo-cofactor containing enzymes and is composed of the catalytic α-subunit harbouring the Mo-cofactor and the FeS-clusters/heme b containing electron transferring βand γ-subunits 29 . This enzyme is proposed to abstract a hydride from tertiary C25 to the Mo(VI) = O species yielding the Mo(IV)-OH intermediate 14  cholesterol and its analogues to tertiary alcohols 30,31 . The rationale for the initial formation of the tertiary alcohol 25-OH-CDO, that cannot be directly further oxidised, during anaerobic cholesterol catabolism has remained enigmatic. Though previous high resolution mass spectrometry (HRMS)-based analyses suggested the primary alcohol 26-OH-CDO as an intermediate 24,32 , the underlying conversion of a tertiary alcohol to a primary alcohol involves an unknown enzymology.
Here, we aim to identify the enzymatic steps involved in the water-dependent oxidation of a primary carbon atom during anaerobic cholesterol degradation. We demonstrate that the formation of a tertiary alcohol at the cholesterol side chain initiates a three-step enzymatic reaction cascade finally resulting in the oxidation of unactivated C26 to a carboxylate.

Results
Enzymatic dehydration of 25-OH-cholesta-1,4-diene-3-one. To elucidate the enzymology responsible for production of CDO-26oate, we synthesised 25-OH-CDO from CEO using overproduced S25DH and AcmB. Cell-free extracts of S. denitrificans grown with cholesterol under denitrifying condition were anoxically reacted with 25-OH-CDO in the presence of a multitude of natural and artificial electron acceptors (e.g., NAD[P] + , K 3 [Fe (CN) 6 , or 2,6-dichlorophenolindophenol at 0.5 mM, respectively). In no case, conversion of 25-OH-CDO to a product was observed.
In further trials, electron acceptors were omitted in the reaction assays and a number of potential co-substrates were tested. In the presence of ATP and MgCl 2 (5 mM each), the time-dependent conversion of 25-OH-CDO 1 to a minor, more polar intermediate 2 was observed using ultra-fast performance liquid chromatography (UPLC) analyses, which was readily further converted to a second less polar product 3, (Fig. 2). After ultracentrifugation of cell-free extracts this conversion was found to occur only in the soluble protein fraction. UPLC-HRMS analysis of 2 did not allow a clear assignment to a molecular mass, whereas analysis of compound 3 revealed a [M + H] + ion with m/z = 381.3161 ± 0.4 Da suggesting the loss of H 2 O from the substrate (m/z = 399.3274 ± 1.1; for MS data of all steroid intermediates analysed in this work, see Supplementary Table 1). The product 3 was purified by preparative HPLC and identified by 1 H, 13 C, and 2D NMR analyses as desmost-1,4-diene-3-one (DDO) containing an allylic Δ24 double bond ( Supplementary Figs. 1-3). Thus, the hydroxyl functionality introduced by S25DH was subsequently eliminated to an alkene by an ATP-dependent 25-OH-CDO dehydration activity.
We tested whether product 2 represents a phosphorylated intermediate during the ATP-dependent dehydration of 25-OH-CDO. For this purpose we chemically synthesised a 25-phospho-CDO standard from enzymatically prepared 25-OH-CDO. We opted for a P(III)-amidite approach due to the steric hindrance of the tertiary OH, using a bis fluorenylmethyl (Fm) protected For the determination of cofactor specificity and stoichiometry, the ATP-dependent 25-OH-CDO dehydration activity was enriched via ammonium sulphate precipitation (activity retained in the suspended 50% saturation pellet). During formation of DDO from 25-OH-CDO no other nucleoside triphosphate than ATP was accepted as co-substrate; Mg 2+ could substitute for Mn 2+ albeit at only 5% activity. UPLC analyses of adenosine nucleotides revealed that 1.2 ± 0.1 mol ATP were hydrolysed to 0.9 ± 0.1 mol ADP and 0.3 ± 0.1 mol AMP per mol 25-OH-CDO consumed (mean value ± standard deviation in three independent determinations); ATP hydrolysis was negligible in the absence of the steroid.
Oxidation of the allylic methyl group. For analysing the further conversion of DDO, we enzymatically synthesised DDO from 25-OH-CDO using cell-free extracts from S. denitrificans. Strictly depending on K 3 [Fe(CN) 6 ] as electron acceptor, cell-free extracts of S. denitrificans converted DDO to product 4 that was subsequently converted to 5 in assays performed anaerobically (Fig. 3).
In summary, DDO 3 formed by ATP-dependent dehydration of the tertiary alcohol was hydroxylated first to the allylic primary alcohol 4 with water, followed by the oxidation to the aldehyde 5. Specific activities in cell-free extracts were 2 ± 0.1 nmol min -1 mg -1 for 26-OH-DDO formation and 1.2 ± 0.2 nmol min -1 mg -1 for 26-OH-DDO dehydrogenation (mean value ± standard deviation in three biological replicates of cell extract preparation).
To test whether one of the DH 5-7 enzymes may be involved in the water-dependent hydroxylation of DDO, we heterologously expressed the genes potentially encoding the active site αβγsubunits of DH 5-7 and an assumed δ-chaperone in the βproteobacterium Thauera aromatica as described for S25DH 1-4 30 . Using cell-free extracts from T. aromatica producing DH 5-7 , only with DH 5 we observed the K 3 [Fe(CN) 6 ] dependent conversion of DDO to 26-OH-DDO; we therefore refer this enzyme to S26DH 1 . Heterologously produced S26DH 1 was enriched from T. aromatica extracts by three chromatographic steps under anaerobic conditions using a modified protocol established for S25DH 1-4 30 .
SDS-PAGE analysis of enriched S26DH 1 revealed four major protein bands (Fig. 4b) that were excised, tryptically digested, and analysed by electrospray ionisation quadrupole time-of-flight mass spectrometry (ESI-Q-TOF-MS). The masses of the ions obtained from tryptic peptides confirmed the presence of the α,β, γ-subunits of S26DH 1 from S. denitrificans, those obtained from the band eluting at 43 kDa were assigned to a co-purified 6oxocyclohex-1-ene-1-carboxyl-CoA hydrolase from T. aromatica (Supplementary Table 2). The latter cofactor-free enzyme plays a role in the catabolic benzoyl-CoA degradation pathway 34 and is unlikely to be associated with steroid hydroxylation. Amino acid sequence similarities of S26DH 1 with characterised S25DHs 30 suggested the presence of one Mo-cofactor, four [4Fe-4S] clusters, one [3Fe-4S] cluster, and a heme b. The metal content of S26DH 1 determined by colorimetric procedures was 0.81 ± 0.03 mol Mo, 19.7 ± 0.4 mol Fe, and 0.89 ± 0.01 mol heme b per mol S26DH 1 , respectively (mean value ± standard deviation in three different enzyme preparations). These values are in good agreement with the expected cofactor content and with the values determined for other S25DHs 30 .
Unexpectedly, enriched S26DH 1 catalysed the hydroxylation of DDO to 26-OH-DDO 4 and the subsequent dehydrogenation of the latter to DDO-26-al 5 with K 3 [Fe(CN) 6 ] as electron acceptor ( Supplementary Fig. 13). As a control, T. aromatica extracts containing the plasmid without the S26DH 1 genes did not catalyse such a reaction. Both partial reactions of S26DH 1 dehydrogenase followed Michaelis-Menten kinetics with V max = 219 ± 11 nmol min -1 mg -1 /K m = 123 ± 25 µM for DDO hydroxylation, and V max = 73 ± 3 nmol min -1 mg -1 /K m = 83 ± 12 µM for 26-OH-DDO dehydrogenation (mean values ± standard deviations of enzyme activity measurements at various substrate concentrations, for the data points used see Supplementary Fig. 6c, d). No further oxidation to a carboxylate was observed by enriched S26DH 1 .
Oxidation of 26-OH-DDO to DDO-26-carboxylate. Enriched S26DH 1 used K 3 [Fe(CN) 6 ] but not NAD + as electron acceptor for the oxidation of the allylic alcohol 4 to the aldehyde 5, which is in agreement with its proposed periplasmic location (see Discussion). We tested whether this activity is relevant for the cholesterol catabolic pathway or whether an additional cytoplasmic alcohol dehydrogenase is involved. Using NAD + as electron acceptor, extracts from S. denitrificans grown with cholesterol showed virtually no conversion of 26-OH-DDO 4 to DDO-26-al 5 (<0.1 nmol min -1 mg -1 ), whereas it served as acceptor for the ready oxidation of DDO-26-al 5 to DDO-26-carboxylate (5 ± 0.2 nmol min -1 mg -1 ) (mean value ± standard deviation in three replicates). In this assay, formation of an additional product was observed that after ESI-Q-TOF-MS analyses was assigned to 26-OH-DDO (Supplementary Table 1, Fig. 14). Obviously, the NADH formed during DDO-26-al oxidation served as donor for a promiscuous dehydrogenase that reduced the C26 aldehyde to the C26 alcohol.
Recently, a cholesterol induced gene was identified in S. denitrificans (WP_154716401) that was hypothesised to encode an aldehyde dehydrogenase (C26-ALDH) involved in cholesterol C26 oxidation 25 . We heterologously expressed this gene in E. coli, and extracts from cells producing the recombinant enzyme The previously assigned DH 5 is now referred to as S26DH 1 (marked in red). b SDS-PAGE of S26DH 1 heterologously produced in T. aromatica; the numbers refer to the molecular masses (kDa) of a standard. The band marked with an asterisk was identified as co-purified 6-oxocyclohex-1-ene-1carboxyl-CoA hydrolase from T. aromatica; the minor band eluting slightly above the 55 kDa is a degradation product of the α 5 -subunit (Supplementary Table 2). c Cluster of genes encoding the αβγ-subunits and the gene encoding the δ-chaperone of S26DH 1 (accession numbers: γβα-subunits, WP_154715926-8; δ-chaperone, WP_154716737). Source data are provided as a Source Data file.  Fig. 15b).

Discussion
In this work we identified an enzymatic reaction sequence that allows for the oxidation of an unactivated primary carbon to a carboxylate using water as only hydroxylating agent. In this enzyme cascade the high activation energy barrier for C-H bond hydroxylation of a primary carbon is divided into three individual steps, each of which being energetically and mechanistically plausible under cellular conditions. The C-H bond dissociation energies to carbocations and hydrides relevant for the hydroxylation of tertiary C25 in cholesterol and allylic C26 in desmosterol are 142 kJ mol -1 and 117 kJ mol -1 lower than that for the primary C26 of cholesterol, respectively (based on gas phase calculations for the isopentane and 2-methyl-2-butene analogues) 35 . Thus, the rationale for the initial formation of a tertiary alcohol during anaerobic cholesterol catabolism 28,29 is to enable the ATPdependent dehydration to a trisubstituted alkene that is crucial for the subsequent water-dependent hydroxylation of the allylic carbon via a relatively stable allylic carbocation intermediate. The major challenge of oxygen-independent primary carbon activation represents the dehydration of the tertiary alcohol that is achieved by coupling to exergonic ATP hydrolysis (ΔG' ≈ -50 kJ mol −1 ). The phosphate eliminated from the phosphoester intermediate is a much better leaving group than water. A related reaction is known from isoprenoid biosynthesis via the mevalonate pathway: (R)-mevalonate-5-diphosphate is ATPdependently decarboxylated to isopentenyl diphosphate 36,37 .
Here, phosphate elimination from an assumed phosphorylated intermediate is accompanied by decarboxylation, which additionally drives the elimination reaction forward ( Supplementary  Fig. 16). ATP-dependent dehydration has also been reported for the recycling of spontaneously formed NAD(P)H hydrates 38 , however it is unclear whether it proceeds via a phosphorylated intermediate. While a reaction mechanism via tertiary and allylic carbocation intermediates is plausible for the two Mo-dependent hydroxylases S25DH 1 /S26DH 1 , it remains uncertain whether this is the case for phosphate elimination from 25-phospho-DDO. If the phosphate elimination would proceed via the identical tertiary carbocation as proposed for S25DH catalysis, the question rises why it is not directly deprotonated by S25DH to the alkene in a single step? Probably, the rebound of the hydroxyl-functionality at the assumed Mo(IV)-OH intermediate to the carbocation is much faster than a competing deprotonation at C24 to an alkene by a putative base.
The reaction cascade identified in this work allows in principle for any enzymatic water-dependent hydroxylation at a primary carbon next to a tertiary one. The genome of S. denitrificans contains two further copies of S26DH-like enzymes and they may represent isoenzymes specifically involved in the hydroxylation of allylic methyl groups of intermediates during catabolism of steroids with modified isoprenoid side chains such as β-sitosterol or ergosterol. Hence, a pair of specific Mo-dependent S25DH and S26DH appears to be required for the hydroxylation of primary carbons at the individual isoprenoid chains. The reaction cascade may also be involved in the degradation of non-steroidal isoprenoids or tertiary alcohols. E.g., the anaerobic degradation of the fuel oxygenate methyl tert-butyl ether, an environmental pollutant of global concern, has frequently been described at anoxic environments 39,40 . However, the anoxic degradation of the tert-butanol intermediate is unknown but could in principle be accomplished via ATP-dependent dehydration to isobutene, analogous to the described pathway.
Cholesterol degradation in anaerobic bacteria is initiated in the periplasm by AcmA, S25DH 1 , and probably AcmB dependent conversion into 25-OH-CDO/25-OH-DDO (oxidation to the diene in ring A may also occur later in the pathway). However, subsequent alkene formation is ATP-dependent and therefore has to occur in the cytoplasm. In contrast, subsequent C26 hydroxylation and oxidation to the C26 aldehyde will again take place in the periplasm as evidenced by the N-terminal twin-arginine translocation (TAT) sequence present in S25DH and S26DH (Supplementary Table S3). Thus, initial steps of anaerobic cholesterol degradation involve enzymes alternately accessing their substrates from the periplasm and the cytoplasm (Fig. 5). Though flip-flop of cholesterol between the two leaflets of biological membranes is generally considered fast with rate constants in the 10 4 s -1 range 41-43 , it is 5.5-fold faster for 25-OH-cholesterol vs cholesterol 43 . Thus, initial side chain-hydroxylation facilitates the accessibility of cholesterol-derived intermediates from both sides of the cytoplasmic membrane, which appears to be crucial for the reaction cascade involved in water-dependent primary carbon oxidation.
Heterologous production of AcmB in E. coli BL21. The gene encoding AcmB was amplified using the primers listed in Supplementary Table S4. The resulting 1.8-kb DNA fragment was SacI/HindIII double digested and cloned into pASK-IBA15plus before transformed into E. coli BL21. Induction of AcmB was carried out in 2 YT medium at 20°C, supplemented with 100 µg mL -1 ampicillin and 20 µg mL -1 anhydrotetracycline. Cells were harvested in the late exponential phase (16 hours induction) and used for further experiments.  Fig. 5 Proposed subcellular localisation of initial anaerobic cholesterol degradation steps. For easier presentation, reduction of Δ24 double bond is omitted. The dotted arrows indicate that hydrophobic side chain needs to be bound by S25DH 1 and S26DH 1 . The assignment of S25DH 1 and S26DH 1 to the periplasm are based on their TAT sequence, that of S25 dehydratase and aldehyde dehydrogenase (ALDH) to the cytoplasm on their cytoplasmic cosubstrates. Hydroxylations at C25 and C26 abolish polarity of CDO/DDO and improves access from both sides of the cytoplasmic membrane.
Heterologous production of DDO-26-al dehydrogenase (C26-ALDH) in E. coli BL21. The C26-ALDH gene (WP_154716401) was amplified using the primers listed in Supplementary Table S4. The resulting 1.5-kb DNA Fragment was digested by NheI/NcoI and cloned into pASK-IBA15plus and transformed into E. coli BL21. Induction of C26-ALDH was carried out as described for the AcmB gene.
Enrichment of S26DH 1 after heterologous production in T. aromatica K172. All steps were performed under anaerobic conditions in an anaerobic glove box (95% N 2 , 5% H 2 , by vol.; O 2 < 2 ppm). The buffers and reagents used were degassed using alternating (20 cycles) N 2 (0.5 bar) and vacuum (> -0.9 bar) to reach anaerobicity. Anaerobically harvested cells were lysed with a French pressure cell at 137 MPa using two volumes (w/v) of buffer A (20 mM Tris/HCl pH 7.0, 0.02% [w/v] Tween 20, 0.5 mM dithioertythritol [DTE]). Solubilisation of lyzed cells was carried out for 16 h at 4°C using 1% (v/v) Tween 20. After ultra-centrifugation at 150,000 × g for 1.5 h, the supernatant was used for enzyme enrichment. The soluble proteins were applied to a DEAE-Sepharose column (75 mL; GE Healthcare) at 7. Spectrophotometric determination of Mo, Fe and heme b. Iron was determined with o-phenantroline using a modified protocol as described 45 . Enriched proteins (5 µL) were added to 245 µL H 2 O and acidified with 7.5 µL 25% HCl, mixed, incubated at 80°C for 10 min, followed by centrifugation at 10,000 × g for 10 min. The supernatant was transferred and 750 µL H 2 O, 50 µL 10% [w/v] hydroxylamine and 0.1% [w/v] o-phenanthroline were added, mixed and incubated at room temperature for 30 min. The absorbance was measured at 512 nm and compared with a (NH 4 ) 2 Fe(SO 4 ) 2 standard.
Mo was determined colorimetrically with dithiol using a modified protocol 46 . 20 µL proteins were incubated at 60°C until dryness was reached. The dried product was solved in 250 µL 4 M HCl and heated to 90°C for 30 min. 100 µL reducing solution (15% [w/v] ascorbic acid and 2% [w/v] citric acid in H 2 O) were added, mixed and incubated for 5 min at room temperature. 300 µL H 2 O and 100 µL dithiol solution (0.1 g zinc dithiol and 0.4 g NaOH diluted in 600 µL ethanol, 31 mL H 2 O and 200 µL thioglycollic acid) were added, mixed and incubated for 5 min. Then, 500 µL isoamyl acetate were added and the mixture was shaken vigorously for 1 min. The absorbance was measured in the organic phase at 680 nm and compared with a sodium molybdate dihydrate standard.
The heme b content was determined spectrophotometrically at 556 nm after the anaerobically conducted complete reduction of the enzyme with sodium dithionite. The heme-content was calculated using the extinction coefficient of reduced heme b (ε 556 = 34.64 mM −1 cm −1 ) 47 .
Protein identification by mass spectrometry of peptides. Proteins were identified by excising the bands of interest from SDS-PAGE gels. After in-gel digestion with trypsin, the resulting peptides were separated by UPLC and identified using a Synapt G2-Si high-resolution mass spectrometry (HRMS) electrospray ionisation quadrupole time-of-flight (ESI-Q-TOF) mass spectrometry system (Waters); the system has been described in detail previously 48 .
LC-ESI-MS analyses of steroid compounds. Metabolites were analysed by an Acquity I-class UPLC system (Waters) using an Acquity UPLC CSH C18 (1.7 µm, 2.1 × 100 mm) column coupled to a Synapt G2-Si HRMS ESI-Q-TOF device (Waters). For separation, an aqueous 10 mM NH 4 OAc/acetonitrile gradient was applied. Samples were measured in positive mode with a capillary voltage of 2 kV, 100°C source temperature, 450°C desolvation temperature, 1000 L min -1 N 2 desolvation gas flow, and 30 L min -1 N 2 cone gas flow. Evaluation of LC-MS metabolites was performed using MassLynx (Waters). Metabolites were identified by their retention times, UV-vis spectra, m/z values.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/ licenses/by/4.0/.