Synthesis of ω‐Muricholic Acid by One‐Pot Enzymatic Mitsunobu Inversion using Hydroxysteroid Dehydrogenases

The biocatalyzed conversion of hyocholic acid (3α,6α,7α‐trihydroxy‐5β‐cholan‐24‐oic acid) into ω‐muricholic acid (3α,6α,7β‐trihydroxy‐5β‐cholan‐24‐oic acid) has been obtained exploiting a small library of 7α‐ and 7β‐HSDHs (hydroxysteroid dehydrogenases). The process has been optimized and performed avoiding the isolation of the 7‐oxo intermediate using the appropriate coupled enzymes for the in situ cofactor regeneration. Moreover, the biocatalyzed reduction of 6,7‐dioxolithocholic acid (3α‐hydroxy‐6,7‐dioxo‐5β‐cholan‐24‐oic acid) was also investigated.


Introduction
Hyocholic acid (HCA, 1, 3α,6α,7α-trihydroxy-5β-cholan-24-oic acid) is one of the major components of pig bile and is the only trihydroxylated bile acid that can be isolated from it. [1] Its structural characterization was described several decades ago, during the golden age of steroid chemistry, [2] and its trivial name as well as those of its cognate bile acids has been codified later on by Hofmann in a seminal paper specifically devoted to the nomenclature of these compounds. [3] Muricholic acid is the trivial name for hyocholic acid isomers at C-6 (6β-OH, 7α-OH, α-muricholic acid) or at C-7 (6α-OH, 7β-OH, ω-muricholic acid) or both at C-6 and at C-7 (6β-OH, 7β-OH, β-muricholic acid). They account for~35 % of the total bile acid pool in mice, but both their isolation in pure form and their chemical syntheses are challenging, as it is reflected by their commercial costs from specialized suppliers. This is particularly the case for ω-muricholic acid (CAS number 6830-03-1), which is available at more than 250 USD for 1 mg (e. g., see in MolPort: https://www.molport.com/).
HCA and its derivatives can also be detected in human blood and urine and these molecules have been recently proposed as biomarkers for the early detection of severe metabolic disorders, such as type 2 diabetes mellitus. [4] Moreover, 1 might be considered as an interesting alternative starting material for the preparation of ursodeoxycholic acid (UDCA, 2), presently synthesized on multiton scale per year from cholic acid (CA, 3, 3α,7α,12α-trihydroxy-5β-cholan-24-oic acid) via chenodeoxy-cholic acid (CDCA, 4, 3α,7α-dihydroxy-5βcholan-24-oic acid). [5] In this respect, one of the synthetic requirements is an efficient protocol for the α/β epimerization of the hydroxyl group at C-7.
Hydroxysteroid dehydrogenases (HSDHs) are a group of NAD(P)(H)-dependent dehydrogenases specifically devoted to the regioselective oxidation as well as to the regio-and stereoselective reduction of polyhydroxylated/ketonic steroids. [6] Their synthetic potential for the selective transformation of bile acids has been demonstrated in a series of papers published since the early eighties. [7] Specifically, almost 30 years ago we described the α/β inversion of the C-7 hydroxyl of several bile acids, either in water (the free acids) or in biphasic systems (the corresponding methyl esters). [8] As it is shown in Scheme 1 for the acids, it was a two-step procedure based on the HSDH-catalyzed oxidation of the C-7α OHs, the isolation of the corresponding 7-oxo derivatives, and their subsequent HSDH-catalyzed reduction to the expected C-7β OHs. To achieve these results, three HSDHs were employed: two 7α-HSDHs, the commercially available NAD(H)-dependent enzyme from Escherichia coli and the NADP(H)-dependent 7αand 7β-HSDHs isolated by us from a Clostridium strain. When this protocol was applied to hyocholic acid (1), the expected 7oxo derivative (5) was obtained, isolated, and characterized. However, compound 5 was the only 7-oxo bile acid intermediate that the 7β-HSDH available for us at that time failed to reduce and the 7β-OH derivative 6, that is ω-muricholic acid, could not be isolated.
In recent years, also exploiting metagenomic approaches, several new HSDHs became available to us, among which three novel 7α-HSDHs and four 7β-HSDHs (Table 1). [9] We thought that some of these new enzymes might have helped us to overcome the limitation that we encountered years ago and in the following we report on the obtained results.

Results and Discussion
Hyocholic acid (1) was submitted to the action of five different 7α-HSDHs (the two previously available [8] and three new enzymes, Table 1) with a suitable system for the in situ enzymatic regeneration of the oxidized cofactor NAD(P) + . As expected, all these 7α-HSDHs were able to regioselectively and quantitatively oxidize 1. The reactions were scaled up to 500 mg scale and the 7-oxo derivative 5 was isolated and characterized by NMR. Specifically, the location at carbon C-7 of the carbonyl group was shown by the proton and carbon correlation experiments, which allowed to recognize the CH(6)-CH(5)-C(10)-CH3(19) chain. Moreover, the value of 6.0 Hz of the vicinal coupling constant J H5-H6 , similar to that of hyocholic acid, indicated that the stereochemistry at C(5) and C(6) remained unchanged [8,10] (for more details, see the Experimental Part and the Supplementary Materials).
A similar preliminary screening of the enzymatic reduction of 5 with the available 7β-HSDHs (see Table 1) gave very positive results, as in all cases the formation of a more polar product was observed by TLC (data not shown). The reactions were thus scaled up to a 100-200 mg scale (see Experimental section) either with the NADP(H)-dependent 7β-HSDH from Collinsella aerofaciens (Cae7β-HSDH) or with the NAD(H)dependent 7β-HSDH from Stanieria cyanosphaera (Sc7β-HSDH) and in both cases the quantitative conversion of 5 into a more polar product was observed, which could be isolated as a white powder simply by acidification of the water solution followed by filtration of the precipitate. Its structure was confirmed to be the expected 7β-OH derivative 6 (3α,6α,7β-trihydroxy-5β-cholan-24-oic acid) by NMR analysis. In fact, the configurational assignment of its H-6, H-7, and H-8 could be determined from the vicinal coupling constants J H6-H7 and J H7-H8 . In hyocholic acid (1) the values of these coupling constants were 3.5 and 2.8 Hz, respectively, typical of an axial (H-6), equatorial (H-7) and axial (H-8) arrangement in rigid hydroxylated six-membered rings in a chair conformation. [10] In the product 6 the values of such coupling constants rose to 9.4 Hz, diagnostic for an axial orientation of all the three protons H-6, H-7 and H-8, thus confirming the β orientation of OH-7.
Looking for further improvements of this process, we considered the possibility of performing the 7α/β inversion of 1 into 6 avoiding the isolation of the 7-oxo derivatives 5. The one-pot enzymatic epimerization of the C-7 OH of cholic acid (3) and/or chenodeoxycholic acid (4) has been studied since a long time. However, as it has been confirmed in detailed recent research reports, [11] the inversion could not be quantitative when using 7α-HSDHs and 7β-HSDHs with the same cofactor specificities (either both NAD(H)-or both NADP(H)-dependent enzymes) and the approximately 1 to 9 mixtures of 7α-and 7βisomers were isolated.
This "thermodynamic trap" could be overcome by exploiting enzymes with different cofactor specificities, as we reported years ago working on cholic acid. [12] By applying the same protocol to obtain the quantitative one pot stereoinversion of 1 into 6, we considered both the options shown in Scheme 2, that is either catalyzing the oxidation of 1 with a NAD(H)dependent 7α-HSDH and the reduction of the intermediate 5 with a NADP(H)-dependent 7β-HSDH (part A) or, on the contrary, the oxidation of 1 with a NADP(H)-dependent 7α-HSDH and the reduction of the intermediate 5 with a NAD(H)dependent 7β-HSDH (part B). The thermodynamic driving force was given by the coupled enzymatic regeneration systems, Scheme 1. Enzymatic stereoinversion of C-7 OH in a Mitsunobu-like approach. Table 1. 7α-and 7β-hydroxysteroid dehydrogenases employed in this study. [9] Enzyme Activity Cofactor Source The small-scale reactions (12.5 mM of the starting substrate 1, total volume 1 mL) were monitored by TLC and, following derivatization, by GC-MS (see time evolution in GC-MS of onepot 2B in the Supporting information). We were pleased to observe, in both cases, a complete conversion of 1 into 6, without residual starting substrate 1 or ketonic intermediate 5.
The biotransformation shown in Scheme 2A was scaled up to 100 mg scale and the product was isolated again by acidification and filtration/extraction (see Experimental and/or Supplementary Materials for details). In this way an efficient and simple access to the rare ω-muricholic acid 6 [13] was optimized.
As a side investigation, in order to get useful information on the performances of our enzymes with compounds that might be obtained as intermediates en route of the conversion of hyocholic acid 1 into ursodeoxycholic acid (2), we submitted to the action of 7β-HSDHs the commercially available 6,7-dioxo derivative (7) (Scheme 3). Once again, to our satisfaction, all the five 7β-HSDHs were active on 7. The reaction was scaled up to a 40 mg scale using Cae7β-HSDH and product 8 was isolated and characterized. NMR analysis confirmed the 7β-configuration of the obtained alcohol (J H7-H8 of 9.5 Hz), but, to our surprise, the A and B ring showed a trans junction instead of the original cis configuration, i. e., the hydrogen atom at C(5) occurred anti with respect to the CH 3 (19). A possible explanation was given by the NMR analysis of the starting diketone 7, which, at least in DMSO solution, exists in the ketoenolic form 7 a (C(5) = C(OH)(6)-CO(7)-CH (8)). The double bond occurs between carbons C(5) and C(6), as indicated by the observation of the proton-carbon long range coupling constants CH(3)/C(5) and CH 3 (19)/C(5) in the HMBC experiment. The enzymatic reaction reducing the carbonyl at C-7 to the correspondent hydroxyl derivative, allows the enolic double bond rearrangement to the C-6 oxo form, resulting in the formation of the more stable trans ring fusion of the cyclohexane rings (Scheme 3 and Figure 1). The conformational instability of the A-B ring fusion of 6-oxo steroids has been previously observed, as it has been reported, for instance in a recent patent. [14]

Conclusion
The one-pot biocatalyzed conversion of hyocholic acid (1, 3α,6α,7α-trihydroxy-5β-cholan-24-oic acid) into the rare ω-muricholic acid (6, 3α,6α,7β-trihydroxy-5β-cholan-24-oic acid) has been obtained by exploiting a small library of new 7α-and 7β-HSDHs. It is an additional example of concomitant enzymatic oxidations and reductions running together and, thanks to the appropriate choice of the coupled enzymes for the in situ cofactor regeneration, performing a quantitative Mitsunobu-like stereoinversion of the C-7 OH of hyocholic acid.

HSDHs expression and purification
Expressions and purifications of the HSDHs were carried out as previously described. [9] Activity assays Dehydrogenase activities of HSDHs and FDHs were determined spectrophotometrically by measuring the reduction of NAD(P) + at 340 nm (ɛ: 6.22 mM À 1 cm À 1 ), while the activity of RmLDH was measured by following the oxidation of NAD(P)H at the same wavelength. The activity of GlutDH was indicated by Merck and thus not tested. Assays were carried out in polyethylene cuvettes at room temperature by adding the opportune amount of the purified dehydrogenase to the following assay mixtures (1 mL final volume): HSDH assay: 2.5 mM substrate (cholic acid for 7α-HSDHs, ursocholic acid for 7β-HSDHs); 50 mM potassium phosphate buffer, pH 9.0; 0.20 mM NAD(P) + . One unit (U) is defined as the enzyme activity that reduces/oxidizes 1 μmol of NAD(P)(H) per min under the assay conditions described above.

Analytical methods
Usually, TLC analyses were sufficient to determine whether a reaction on a bile acid was concluded or not. However, all the tested eluent mixtures were not capable to separate 1 and 6. Therefore, at scheduled times, samples from the one-pot reactions (100 μL) were lyophilized, resuspended in MeOH to 1 mM concentration, and treated overnight with catalytic acetyl chloride to obtain the corresponding methyl esters. The derivatized samples were then submitted to GC-MS analysis. The samples were dissolved in deuterated dimethylsulfoxide, and their NMR spectra were acquired on a Bruker AV 400 or AV 500 instruments at 305 K. Since the spectra were quite complex only a partial analysis was performed. The assignments reported in the description of the spectra for each compound were based on the proton homonuclear experiments COSY and NOESY and on the proton-carbon heteronuclear experiments HSQC and HMBC. The spectra were also acquired after addition of D 2 O to the solution to remove the hydroxyl protons thus allowing a more precise measurement of the coupling constants along the fragment C5-C6-C7-C8. The multiplicity of the carbon nuclei was determined by the DEPT 135 experiment which allows to distinguish among quaternary (s), methine (d), methylene (t) and methyl (q) carbons. ChemCatChem Full Papers doi.org/10.1002/cctc.202101307 ESI-MS spectra were recorded on a Bruker Esquire 3000 PLUS instrument (ESI Ion Trap LC/MSn System), equipped with an ESI source and a quadrupole ion trap detector (QIT). The samples were dissolved in methanol to 1-2 g L À 1 and then directly syringed in the ESI-MS at 4 μL min À 1 rate. The analyses were performed in positive mode. The acquisition parameters were optimized as such: 4.5 kV needle voltage, 10 L h À 1 N 2 flow rate, 40 V cone voltage, trap drive set to 46, 115.8 V capillary exit, 13000 (m/z) s À 1 scan resolution over the 35-900 m/z mass/charge range, source temperature 250°C.

Screening of 7α-HSDHs for the regioselective oxidation of 1 to 5
The reactions catalyzed by 7α-HSDHs were coupled with a lactate/ LDH system to regenerate NADH and to α-ketoglutarate/GlutDH to regenerate NADPH. For the initial activity screening of the whole library of 7α-HSDHs, the general protocol was as follows: 12.5 mM 1; 50 mM sodium pyruvate/α-ketoglutarate; 0.5 U mL À 1 LDH/GlutDH; 0.4 mM NAD(P) + ; 3.4 U mL À 1 HSDH; 50 mM potassium phosphate buffer, pH 8.0 (total volume: 1 mL). Additionally, the reaction catalyzed by Hh7α-HSDH was performed in the presence of 0.4 M NaCl. The mixtures were shaken at 25°C and 100 rpm for 24 to 48 h and monitored over time via TLC (eluent CHCl 3 : MeOH: AcOH = 9 : 1 : 0.02). Reaction conversions were evaluated by TLC analyses, showing that the reactions were mostly complete in 1 h.

Scale-up and isolation of 6
The reactions catalyzed by Cae7β-HSDH and Sc7β-HSDH were scaled-up to 100-200 mg following the aforementioned general protocol till full conversion was shown by TLC monitoring (eluent CHCl 3 : MeOH : AcOH = 9 : 1 : 0.02). The product was recovered via precipitation by acidification of the reaction mixture and subsequent filtration to afford a white amorphous powder in quantitative yield. 1