Four Molybdenum-Dependent Steroid C-25 Hydroxylases: Heterologous Overproduction, Role in Steroid Degradation, and Application for 25-Hydroxyvitamin D3 Synthesis

ABSTRACT Side chain-containing steroids are ubiquitous constituents of biological membranes that are persistent to biodegradation. Aerobic, steroid-degrading bacteria employ oxygenases for isoprenoid side chain and tetracyclic steran ring cleavage. In contrast, a Mo-containing steroid C-25 dehydrogenase (S25DH) of the dimethyl sulfoxide (DMSO) reductase family catalyzes the oxygen-independent hydroxylation of tertiary C-25 in the anaerobic, cholesterol-degrading bacterium Sterolibacterium denitrificans. Its genome contains eight paralogous genes encoding active site α-subunits of putative S25DH-like proteins. The difficult enrichment of labile, oxygen-sensitive S25DH from the wild-type bacteria and the inability of its active heterologous production have largely hampered the study of S25DH-like gene products. Here we established a heterologous expression platform for the three structural genes of S25DH subunits together with an essential chaperone in the denitrifying betaproteobacterium Thauera aromatica K172. Using this system, S25DH1 and three isoenzymes (S25DH2, S25DH3, and S25DH4) were overproduced in a soluble, active form allowing a straightforward purification of nontagged αβγ complexes. All S25DHs contained molybdenum, four [4Fe-4S] clusters, one [3Fe-4S] cluster, and heme B and catalyzed the specific, water-dependent C-25 hydroxylations of various 4-en-3-one forms of phytosterols and zoosterols. Crude extracts from T. aromatica expressing genes encoding S25DH1 catalyzed the hydroxylation of vitamin D3 (VD3) to the clinically relevant 25-OH-VD3 with >95% yield at a rate 6.5-fold higher than that of wild-type bacterial extracts; the specific activity of recombinant S25DH1 was twofold higher than that of wild-type enzyme. These results demonstrate the potential application of the established expression platform for 25-OH-VD3 synthesis and pave the way for the characterization of previously genetically inaccessible S25DH-like Mo enzymes of the DMSO reductase family.

A S25DH was initially isolated and characterized by Dermer and Fuchs (11) as a molybdenum cofactor (MoCo)-containing enzyme of the dimethyl sulfoxide reductase (DMSOR) family of Mo enzymes. The enzyme has an ␣␤␥ architecture comprising a 108-kDa (␣), 38-kDa (␤), and 23-kDa (␥) subunit. The catalytically active ␣-subunit contains a molybdo-bis-pyranopterin guanine dinucleotide (Mo-bisPGD) cofactor and a [4Fe-4S] cluster; the ␤-subunit harbors four additional FeS clusters, and the ␥-subunit contains a b-type heme. Like other members of the DMSOR family, S25DH is proposed to be located in the periplasm. It was proposed that a chaperone (SdhD) is involved in proper folding and probably in MoCo insertion (11). S25DH together with ethylbenzene dehydrogenase (EbDH) (17,18) and p-cymene dehydrogenase (19) forms a phylogenetic subcluster within class II of the DMSOR family, that hydroxylate alkyl side chains of steroids or aromatic compounds with water (see Fig. S1 in the supplemental material) (20). Attempts for heterologous/homologous production of S25DH/EbDH subclass members have failed so far (21), which prevented an easy enrichment procedure, as well as the generation of molecular variants.
A recent study revealed that S25DH is capable of hydroxylating vitamin D 3 (VD 3 ) to the clinically relevant 25-OH-VD 3 (Fig. 1B) (22). This activity was dependent on 2-hydroxypropyl-␤-cyclodextrin that is known to promote the isomerization to pre-VD 3 (23), the assumed actual substrate of S25DH. In contrast to cytochrome P450dependent hydroxylation of VD 3 , S25DH-dependent catalysis is independent of an electron donor system and requires only the electrochemically regenerative K 3 [Fe(CN) 6 ] (ferricyanide) as an electron acceptor (22). Though S25DH serves as a promising catalyst for the synthesis of 25-OH-VD 3 , there is a high demand for heterologous expression and for improving the enrichment of S25DH-like enzymes with regard to activities and stabilities.
Sterolibacterium denitrificans is capable of degrading phyto-and mycosterols such as ␤-sitosterol, stigmasterol, or ergosterol with modifications in the isoprenoid side chain (for structures, see Table 1), but the only cholest-4-en-3-one-converting S25DH studied so far is unable to convert any of the 4-en-3-one analogues of these growth substrates (8,11). In addition to the gene encoding the active site ␣-subunit of this S25DH (henceforth referred to as ␣ 1 subunit of S25DH 1 , gene accession number SDENCHOL_20805), the genome contains seven paralogous genes encoding putative S25DH-like enzymes, all affiliating with the class II DMSOR family (␣ 2 -␣ 8 ) (8,11). In particular, the predicted active site ␣ 2-4 (amino sequence identities to ␣ 1 of 72 to 82%) have been hypothesized to represent the active site subunits of S25DH 2 , S25DH 3 , and S25DH 4 involved in C-25 hydroxylation of steroids with modified isoprenoid side chains ( Fig. 2A). This assumption is based on their differential abundance during growth on different steroids such as ␤-sitosterol or ergosterol (8); the role of the other four putative S25DHs (S25DH 5 , S25DH 6 , S25DH 7 , and S25DH 8 ) is unclear (8). Notably, there are fewer genes encoding the ␤␥-subunit components than for the ␣-subunits in the genome of S. denitrificans ( Fig. 2A), suggesting that S25DHs with different ␣-subunits share common ␤␥-subunit components. S25DH 1 from S. denitrificans is composed of the ␣ 1 ␤ 3 ␥ 3 -subunits ( Fig. 2A). Enriched S25DH 1 always contained impurities of other ␣-subunits, which made a clear assignment of activities to individual ␣-subunits problematic (11,22).
In this work, we aimed to elucidate the unknown function of Mo-containing S25DH isoenzymes in S. denitrificans in anaerobic steroid degradation and to explore their potential use for 25-OH-VD 3 synthesis. For this purpose, an expression platform for S25DH and related enzymes was established in the denitrifying betaproteobacterium Thauera aromatica yielding S25DHs with high yields and specific activities. With this tool, four recombinant S25DHs were isolated and characterized; their applicability as catalysts for 25-OH-VD 3 synthesis was probed.

RESULTS
Heterologous production of S25DH 1 . The previously established five-step enrichment of the oxygen-sensitive S25DH 1 (␣ 1 ␤ 3 ␥ 3 complex) from wild-type Sterolibacterium denitrificans grown with cholesterol always gave low yields, low specific activities, and a partially degraded ␣ 1 -subunit (11,22,24). Moreover, the enriched enzyme frequently contained impurities from other S25DH ␣-subunits next to the ␣ 1subunit, which did not allow unambiguous assignment of activities to individual gene products (11,22). These findings motivated us to establish a platform for actively expressing the genes encoding S25DH 1 and related enzymes. For this reason, the genes encoding the ␣ 1 ␤ 3 ␥ 3 -subunits, together with the putative chaperone (henceforth referred to as SdhD) (Fig. 2) were cloned into the broad-host-range plasmid pIZ1016. This construct contained the twin-arginine translocation (TAT) secretion sequence at the N terminus of the ␣ 1 -subunit; to avoid any possible negative effect on ␣␤␥ complex formation, we avoided the use of a tagged subunit. Heterologous production of the resulting ␣ 1 ␤ 3 ␥ 3 SdhD construct was tested in Escherichia coli strains BL21 and Top 10, Azoarcus sp. strain CIB, and Thauera aromatica K172.
Heterologous production of the ␣ 1 -subunit alone in the presence or absence of SdhD was not monitored in this work, as preliminary experiments indicated that such constructs did not result in the formation of soluble/active proteins. Expression of the ␣ 1 ␤ 3 ␥ 3 SdhD-encoding genes did not give soluble gene products in either of the two E. coli strains, and consequently, virtually no formation of 25-hydroxy-cholest-4-en-3one was observed. However, after heterologous production of ␣ 1 ␤ 3 ␥ 3 SdhD in Azoarcus sp. CIB (25) and T. aromatica (26), anaerobically prepared cell extracts from both species showed the conversion of 0.5 mM cholest-4-en-3-one (0.5 mM) to 25-hydroxy-cholest-4-en-3-one; this conversion was dependent on time, protein, K 3 [Fe(CN) 6 ] (5 mM), and 2-hydroxypropyl-␤-cyclodextrin (HPCD) (9% wt/vol) and was observed only with constructs containing all four ␣ 1 ␤ 3 ␥ 3 SdhD components. The specific activities of recombinant S25DH 1 were maximally 3.3 nmol min Ϫ1 mg Ϫ1 in cell extracts from Azoarcus sp. CIB and 12 nmol min Ϫ1 mg Ϫ1 in cell extracts from T. aromatica. Remarkably, both

Mo-Dependent Steroid Hydroxylases
® specific activities were 176% and more than 640% of the activity in cell extracts of cholesterol-grown S. denitrificans (1.87 nmol min Ϫ1 mg Ϫ1 ), respectively, demonstrating overproduction of S25DH 1 in the Azoarcus and Thauera species (Table 1). Under anoxic conditions, loss of recombinant S25DH 1 activity was less than 10% for 1 week at 4°C. In the presence of 5 mM K 3 [Fe(CN) 6 ], activity in T. aromatica crude extracts was stable in air at 30°C for 5 to 8 h but decreased to 50% after 24 h.
T. aromatica crude extracts producing S25DH 1 catalyzed the conversion of VD 3 to 25-OH-VD 3 at 2.8 nmol min Ϫ1 mg Ϫ1 that was 6.5-fold higher compared to wild-type extract (0.43 nmol min Ϫ1 mg Ϫ1 ). This increased VD 3 conversion rate shortened the conversion of 1 mM VD 3 correspondingly (90% conversion in 2 h by 12 mg ml Ϫ1 crude extracts) (Fig. 3), allowing 25-OH-VD 3 synthesis even in air without loss of activity. The absence of 25-OH-VD 3 -converting enzymes in T. aromatica made the addition of AgNO 3 for blocking follow-up enzymes of the cholesterol degradation pathway dispensable (22). Recombinant S25DH also converted 7-dehydrocholesterol to its 25-OH form, as had also been reported for the wild-type enzyme (11); no analogues with modifications in the isoprenoid side chain were converted (Table 1).
Heterologous production of S25DH 2 , S25DH 3 , and S25DH 4 isoenzymes in T. aromatica. The results obtained indicated that T. aromatica K172 represents the most suitable host for the heterologous production of S25DH-like enzymes. For this reason, we used this strain for the production of three additional S25DH-like proteins that contain the ␣ 2-4 active site subunits, respectively. The constructs were cloned into pIZ1016 in a manner to give S25DH 2 , S25DH 3 , and S25DH 4 with the compositions ␣ 2 ␤ 3 ␥ 3 , ␣ 3 ␤ 3 ␥ 3 , and ␣ 4 ␤ 4 ␥ 4 referred to as S25DH 2 , S25DH 3 and S25DH 4 (Fig. 2). Notably, the ␤ 3 ␥ 3 and ␤ 4 ␥ 4 subunits are almost identical (99% sequence identity), and the coexpression of their genes with those encoding individual ␣-subunits was chosen to facilitate the cloning procedure. In all cases, SdhD was coproduced with the individual S25DH isoenzymes.
Using cell extracts of T. aromatica producing S25DH 2 , the rate of 25-OH-7dehydrocholesterol formation was slightly higher than conversion of cholest-4-en-3one. This extract converted VD 3 to 25-OH-VD 3 at a higher rate than S. denitrificans extracts grown with cholesterol, albeit the activity was only around 20% of recombinant S25DH 1 (Table 1). Previous differential proteome analysis showed that the ␣ 3 - subunit of an S25DH (SDENCHOL_20460) was most abundant during growth with ergosterol containing a Δ22 double bond in the side chain (8). Unexpectedly, T. aromatica expressing the S25DH 3 formed only traces of 25-OH-ergosterol from ergosterol. Likewise, brassica-4-en-3-one, an ergosterol analogue with AB rings identical to those in cholest-4-en-3-one was virtually not converted. Surprisingly, campest-4-en-3-one, which lacks the double bond in the side chain but contains an additional methyl branch at (R)-configured C-24, was the only steroid substrate tested that was converted by S25DH 3 to its 25-OH-form in significant amounts ( Table 1). The ␣ 4 -subunit was found upregulated in S. denitrificans cells grown with ␤-sitosterol, a cholesterol analogue with an additional ethyl branch at C-24 (8). In full agreement, T. aromatica cell extracts producing S25DH 4 converted ␤-sitost-4-en-3-one and the structurally related campest-4-en-3-one, both with an (R)-configured tertiary C-24; cholest-4-en-3-one lacking an additional alkyl substituent at this position was also converted at a slightly higher rate (Table 1). Enrichment, activity, and composition of recombinant S25DH 2 , S25DH 3 , and S25DH 4 . The results obtained so far indicated a higher heterologous production of S25DH 1 and probably other S25DHs in Azoarcus sp. CIB and T. aromatica K172 than in wild-type S. denitrificans. As a result, only two out of five chromatographic enrichment steps described for purification from the wild-type bacteria were necessary. The two steps were DEAE anion-exchange chromatography and affinity chromatography on Reactive Red agarose (11,22); we initially tested the purification of the recombinant S25DH 1 from both strains.
With Azoarcus sp. CIB extracts, five protein bands were obtained after the two-step purification procedure (Fig. S3). The bands migrating at 110, 40, and 25 kDa were clearly assigned to the ␣ 1 ␤ 3 ␥ 3 -subunits, whereas those migrating between 50 and 55 kDa represent truncated ␣ 1 degradation products, as observed during purification from S. denitrificans (11,22). The two degradation products were estimated to make more than 80% of the total amount of the ␣ 1 -subunit. This finding suggests that expression in Azoarcus sp. CIB produced predominantly degraded S25DH 1 (see Fig. S3 in the supplemental material).
Purification of S25DH 1 from T. aromatica extracts revealed a highly enriched (purity Ͼ95%) ␣ 1 ␤ 3 ␥ 3 complex with an almost perfect 1:1:1 ratio of the three subunits (Fig. 4). ESI-QTOF-MS analysis of tryptic digestion products of the three excised protein bands identified the expected ␣ 1 ␤ 3 ␥ 3 subunits (see Table S1 in the supplemental material). Most importantly, when prepared from T. aromatica K172, ␣ 1 -degradation products were negligible. These findings explain why the specific activity of recombinant S25DH 1 Mo-Dependent Steroid Hydroxylases ® in Azoarcus sp. CIB extracts was-though still higher than in S. denitrificans-significantly lower than in T. aromatica.
The subunit architecture and native molecular masses of the S25DHs were determined by size exclusion chromatography as follows: 167 Ϯ 5 kDa for S25DH, S25DH 2 , and S25DH 3 and 160 Ϯ 5 kDa for S25DH 4 . These values clearly point to an ␣␤␥ composition of all four heterologously produced S25DHs.
Metal content of recombinant S25DHs. On the basis of the results of previous metal analyses with wild-type S25DH 1 enzyme and on the conserved binding motifs of the individual cofactors, the purified recombinant S25DH 1 , S25DH 2 , S25DH 3 , and S25DH 4 were expected to bind a Mo-bisPGD, four [4Fe-4S] clusters, one [3Fe-4S] cluster, and a heme b, giving one Mo atom and 20 Fe atoms per ␣␤␥ trimer, respectively (11). The Fe content was analyzed by inductively coupled plasma atomic emission spectroscopy (ICP-AES), and the Mo content was analyzed by inductively coupled plasma mass spectroscopy (ICP-MS). The metal content of the four S25DHs varied between 0.5 and 0.8 Mo atoms and between 13.2 and 15.3 Fe atoms per ␣␤␥ trimer, which is in the range of the values determined for the wild-type enzyme (0.7 Mo atom and 16.5 Fe atoms per enzyme) (11) ( Table 3).
Kinetic parameters of recombinant S25DH 1 , S25DH 2 , S25DH 3 , and S25DH 4 . The K m and k cat values of each recombinant S25DH were determined using their individual preferred substrates. Notably, the K m values determined have to be regarded as apparent values, as the nearly insoluble steroid substrates were converted only in the presence of high concentrations of the solubilizing 2-hydroxypropyl-␤-cyclodextrin (9%, wt/vol). The kinetic parameters are summarized in Table 3 with S25DH 1 , S25DH 2 , and S25DH 4 showing the highest k cat /K m values for cholest-4-en-3-one, 7-dehydrocholesterol, and ␤-sitost-4-en-3-one, respectively, suggesting that the respective steroids do indeed represent the preferred substrates. In the case of S25DH 3 , the low catalytic number together with the high K m determined for campest-4-en-3-one indicates that the enzyme is specific for a so far unknown steroid, with campest-4-en-3-one representing the only analogue that is converted to some extent. Enrichment and characterization of S25DH 4 from S. denitrificans grown with ␤-sitosterol. To ascertain whether S25DH 4 has an extended substrate spectrum toward ␤-sitost-4-en-3-one and campest-4-en-3-one, it was isolated from the wild-type bacteria. For this purpose, S. denitrificans was grown in a 200-liter fermenter with 3 mM ␤-sitosterol in a fed-batch culture, yielding 184 g cells (wet weight) (Fig. S4). Cell extracts of S. denitrificans grown with ␤-sitosterol converted ␤-sitost-4-en-3-one to the corresponding 25-OH-form at a specific activity of 0.6 nmol min Ϫ1 mg Ϫ1 (17% of the activity in recombinant T. aromatica).

DISCUSSION
So far, a number of DMSOR family members have been produced by homologously or heterologously expressing their corresponding genes (27)(28)(29)(30)(31)(32)(33)(34). In contrast, attempts to produce the phylogenetically related, complex heterotrimeric molybdenum-ironsulfur/heme b-containing alkyl chain hydroxylases of the class II DMSOR family in an active form have failed so far. In this work we now provide a tool for their recombinant  production in highly active forms as demonstrated for the example of four S25DH isoenzymes with differing substrate specificities. The expression platform opens the door for studying other alkyl chain-hydroxylating members of the DMSOR family comprising enzymes specifically forming tertiary (S25DHs), secondary (ethylbenzene DH) (17,18), and primary (p-cymene DH) (19) alcohols, and many others of unknown function. Notably, production of highly active enzymes in the presence of SdhD required the expression of all three functional genes. Production in Thauera aromatica yielded crude extract activities that in the case of S25DH 1 and S25DH 4 were sixfold higher than in cell extracts from the parental Sterolibacterium denitrificans after growth with the preferred substrates cholesterol (S25DH 1 ) and ␤-sitosterol (S25DH 4 ). The apparent overexpression largely facilitated and shortened the enrichment procedure by using only two chromatographic steps, which in turn greatly reduced to a minimum the generally observed degradation of the active site ␣-subunit. As a further advantage, the specific activities of the isolated enzymes of S25DH 1 and S25DH 4 were twofold higher than after purification from the wild type. Finally, while purification from the wild-type bacteria always yielded mixtures of different isoenzymes, the T. aromatica expression platform produced a preparation of a single S25DH. In summary, production in T. aromatica largely enhanced yield, specific activity, and specificity.
The results obtained now allow for the unambiguous assignment of substrate specificities to individual S25DH isoenzymes. The prototypical S25DH 1 , originally isolated from S. denitrificans cells grown with cholesterol (11), used cholest-4-en-3-one as the preferred substrate. The S25DH 2 has an extended substrate spectrum toward 7-dehydrocholesterol and S25DH 4 for the phytosterol-derived ␤-sitost-4-en-3-one and campest-4-en-3-one with (R)-configured ethyl and methyl branches at C-24, respectively. The function of S25DH 3 appears to be less clear, as it showed only minor activity with campest-4-en-3-one indicating that the natural substrate is still at issue. None of the four S25DHs showed significant activity with sterols containing a Δ22,23 double bond in the side chain such as ergosterol, brassica-4-en-3-one, or stigmast-4-en-3-one, although ergosterol and stigmasterol are growth substrates. This finding suggests that either one of the four remaining S25DHs is involved in their conversion, or a more likely alternative is that the double bond needs to be reduced or otherwise converted prior to C-25 hydroxylation. As the abundance of S25DH 3 in cells grown with ergosterol was clearly increased (8), it is likely that ergosterol is first converted to a so far unknown intermediate that then serves as the substrate for S25DH 3 . However, reduction of the nonactivated Δ22,23 double bond can hardly be achieved in the absence of oxygen with physiological electron donors. The function of S25DH 5 , S25DH 6 , S25DH 7 , and S25DH 8 needs to be elucidated in the future using the system established in this work. S25DH 7 has recently been proposed to be involved in the conversion of 25-to 26-OH-cholest-4-en-3-one. It is striking that the ␣-subunit of putative S25DH 8 is more closely related to that of p-cymene dehydrogenase and to an S25DH-like enzyme from Thauera terpenica (see Fig. S1 in the supplemental material), suggesting that it may play a role in hydroxylating a nonsteroidal isoprenoid compound.
This work demonstrated that of the four S25DH isoenzymes investigated in this work, S25DH 1 is the optimal catalyst for VD 3 hydroxylation to the clinically relevant 25-OH-VD 3 (22). This reaction is of considerable biotechnological potential, as it has several advantages in comparison to multistep chemical (35) or oxygenase-and electron donor-dependent 25-OH-VD 3 synthesis procedures (36)(37)(38)(39)(40). The established expression platform overcomes previously identified limitations of S25DH-catalyzed 25-OH-VD 3 synthesis, which have largely prevented biotechnological application so far. First, the easy and rapid enrichment procedure minimizes degradation of the ␣-subunit as always observed during purification from the wild-type strain. As a result, the specific activities of the enriched S25DHs are around twofold higher in the recombinant than in the wild-type strain. Second, S25DH 1 was produced to a 6.4-fold-higher extent in T. aromatica than in the wild-type bacteria, resulting in the corresponding increase of the VD 3 conversion rate in crude extracts. As a consequence, enzymatic 25-OH-VD 3 synthesis by crude extracts is drastically shortened and can now be accomplished even under aerobic conditions without a significant loss of activity. Third, due to the lack of enzymes catalyzing downstream reactions of anaerobic steroid degradation in T. aromatica, the addition of AgNO 3 , the established inhibitor of these reactions, is now dispensable when using crude extracts for VD 3 conversion.
In summary, the expression platform established in this work provides not only easy access to previously nonstudied members of the alkyl chain-hydroxylating DMSOR family members, it also allows for previously hardly achievable mechanistic studies (e.g., site-directed mutagenesis) and applied studies (25-OH-VitD 3 synthesis) of a catalytically versatile class of molybdenum enzymes.
Culture conditions and preparation of cell extracts. S. denitrificans was cultivated under denitrifying conditions in mineral medium with steroid substrates as described previously (7). Cells were harvested in the late exponential growth phase by centrifugation (8,000 ϫ g, 20 min, 4°C). During large-scale cultivation with ␤-sitosterol (3 mM) in a 200-liter fermenter, nitrate was discontinuously added in 10 mM steps. T. aromatica K172 and Azoarcus sp. CIB expressing genes encoding S25DHs were cultivated under denitrifying conditions using a phosphate-buffered medium with benzoate as the carbon source at 30°C as described previously (41). Cells were harvested anaerobically by centrifugation (8,