An efficient regioconvergent synthesis of 3-aza-obeticholic acid

Bile acids (BAs) are steroidal molecules that play important roles in nutrient absorption, distribution, and excretion. They also act on specific receptors implicated in various metabolic and inflammatory diseases demonstrating their importance as potential drug candidates. Accordingly, there has been a concerted effort to develop new BA derivatives to probe structure – activity relationships with the goal of discovering BA analogues with enhanced pharmacological properties. Among the many steroidal derivatisations reported, the formation of endocyclic azasteroids appeals due to their potential to deliver altered biological responses with minimal change to the steroidal superstructure. Here, we report the synthesis of 3-aza-obeticholic acid ( 6 ) via a regioconvergent route. Ammoniolysis of lactones, formed from an m -CPBA-mediated Baeyer-Villiger reaction on a 3-keto-OCA derivative, furnished protected intermediate amido-alcohols which were separately elaborated to amino-alcohols via Hofmann degradation with BAIB. Upon individual N -Boc-protection, these underwent annulation to the 3-aza-A-ring when subjected to either mesylation or a Dess-Martin oxidation/hydrogenation sequence. Global deprotection of the 3-aza-intermediate delivered 3-aza-OCA in ten steps and an overall yield of up to 19%.


Introduction
Bile acids (BAs) are steroidal molecules that are produced from cholesterol in the liver.They are best known for their role in the enterohepatic circulation where they enable the absorption of fats and other fat-soluble nutrients (e.g.vitamins) in the intestine [1,2].In addition to their digestive role, BAs also operate as signalling molecules that modulate various cellular processes [2,3].Most notable is their interaction with the Farnesoid X nuclear receptor (FXR) and the Takeda G-protein-coupled bile acid receptor 5 (TGR5, GPBAR1) [4,5].FXR is highly expressed in the liver and intestine and is an important regulator of BA homeostasis as well as lipid and glucose metabolism [6][7][8].While a range of FXR agonists are known, including the potent endogenous activator chenodeoxycholic acid (CDCA, 1), the semisynthetic BA obeticholic acid (OCA, 2) is the only agonist approved for clinical use [9,10].However, drawbacks in the use of OCA include side effects such as pruritus, drug-induced liver injury, and dyslipidemia [11,12].TGR5 is expressed in a wider range of cells and tissues where its activation elicits a range of processes including energy expenditure in muscle and brown adipose tissue; the secretion of glucagon-like peptide (GLP)-1 and insulin from endocrine cells; cytoprotective responses in cholangiocytes and gallbladder epithelial cells; and driving anti-inflammatory activity in neurons and immune cells.The most potent agonists of TGR5 are the secondary bile acids lithocholic acid (LCA, 3) and deoxycholic acid (DCA, 4) [2,3].Over the past two decades, both FXR and TGR5 receptors have received substantial attention as attractive pharmacological targets for the treatment of liver and metabolic disorders, such as diabetes, obesity, metabolic syndrome, atherosclerosis, colitis, primary biliary cholangitis (PBC) or non-alcoholic steatohepatitis (NASH) [13][14][15][16][17][18][19][20].Despite numerous drug leads, undesirable off-target responses due to non-selective receptor binding, or poor tissue selectivity has meant that the search for receptor-specific drugs remains at the forefront of BA research (Fig. 1).
Amongst the many explored variations to BA structures [21][22][23], Di Leva et al. investigated the impact of replacing the 3-hydroxy group with a 3-amino group in LCA derivatives and identified the 3α-amino bile alcohol 5 as a selective, low micromolar TGR5 agonist (EC 50 = 6.8 µM) [24].Taking inspiration from this study, we became interested in synthesising the 3-aza-derivative of OCA (6) in which an amino function is inserted into position 3 of the A-ring of the steroid scaffold.Indeed, the synthesis of endocyclic azasteroids has previously been pursued as a promising strategy to generate new steroids with minimal structural change but with altered biological properties.As such, azasteroids have displayed a wide range of biological activities including antiatherogenicity, anti-carcinogenicity, antifungal, antilipidemic, cytotoxicity, local anaesthetic, neuromuscular blocking activity, and inhibition of steroidal reductases [25][26][27].
In another six-step sequence, Rasmusson and co-workers employed a Fig. 1.Natural bile acids and semi-synthetic OCA (2), 3α-amino bile alcohol 5 as well as the synthetic target of this study: 3-aza-OCA (6).
ruthenium tetroxide-mediated ring cleavage and a modified Hunsdiecker reaction to enable the synthesis of lactam 32 (Fig. 3C) [31].The authors, however, were uncertain about the exact configuration of the C-5 stereocenter, but tentatively assigned a 5α-orientation based on the 1 H NMR shift of the 19-methyl group.Despite these various methods to install a nitrogen atom in the A ring of steroidal structures, many of these involve lengthy synthetic routes and require reagents that are not desirable at practicable scales.Moreover, no synthesis of 3-aza-BA analogues of therapeutic interest has been reported previously [37].
To address this, we sought to devise a flexible and scalable route to 3aza-BAs as a foundation for subsequent drug discovery projects.In this study, we report on our efforts to develop a mild and efficient synthesis of 3-aza-obeticholic acid (6).

Results and discussion
Our route to 3-aza-obeticholic acid began with the preparation of the tert-butyl ester (33) of OCA (2) which was obtained in 88% yield (Scheme 1) [38].Treatment of 33 with TEMPO/bis-acetoxy-iodobenzene (BAIB) in dichloromethane at room temperature enabled selective oxidation of the 3-hydroxy group [39].The resulting 3-keto compound 34 was then subjected to a meta-chloroperbenzoic acid (m-CPBA)-mediated Baeyer-Villiger (BV) oxidation [40], which produced a separable 4:1 mixture of the regiomeric lactones 35a and 35b.Because the BV reaction is substrate-dependent and various studies to date have failed to optimise the regioselectivity of the reaction we elected to pursue elaboration of both regiomers to final 3-aza-BAs, thus enabling access to such BAs regardless of differing regioselectivities for different substrates in the BV reaction.Ammonolysis of the lactone rings was attempted by heating a solution of the individual isomers in 7 N methanolic ammonia in a sealed tube at 90 • C [41].While the major lactone regiomer 35a gave amide 36 in 72% yield, the minor lactone regiomer 35b did not undergo the desired reaction and instead resulted in an intramolecular translactonisation forming lactone 38.As regiomer 35b was the minor product from the BV reaction, we opted to not pursue further attempts to ammonolyse lactone 38.Finally, amide 36 was subjected to a BAIB-promoted Hofmann degradation at room temperature to afford the amino alcohol 37 in 66% yield [42].
With amino-alcohol 37 in hand, we then attempted its cyclisation to generate the 3-aza-A-ring (Scheme 2).Initially, we employed a Mitsunobu reaction [43], but under these reaction conditions we once again found that participation of the 7α-OH group favoured the exclusive formation of cyclic ether 39 in 80% yield.The same cyclisation occurred when we tried to deprotect the tert-butyl ester moiety of 37 under acidic reaction conditions or when we sought to effect the desired cyclisation via mesylation of Cbz-protected amine 41 [44].From these results it was clear that the proximity of the activated primary hydroxy to the 7α-OH group stereoelectronically favours its substitution.
To disable participation of the 7α-OH group in the cyclisation step, this position was BOM-protected giving 44 (Scheme 3) [45].The BOM group was chosen for its ease of installation and its potential optional deprotection by acid or catalytic hydrogenation, which would allow orthogonality to other functional groups within subsequent intermediates.The subsequent BV oxidation reaction gave a 3.6:1 mixture of regiomeric lactones 45a/45b (as determined by 1 H NMR analysis).An unforeseen consequence of BOM protection was that lactone separation by chromatography was challenging.Therefore, the ammonolysis step was conducted on the regiomeric mixture yielding the corresponding amido-alcohol regiomers 46a/46b which were isolated as individual isomers in 78% overall yield.Pleasingly, both amido-alcohols (46a and 46b) underwent Hofmann degradation to give the amino-alcohols 47a and 47b, respectively.Although the separate regiomers were characterised by NMR spectroscopy, the absolute stereochemistry of regiomer 47b was determined by X-ray crystallography (Scheme 3).
Next, we attempted to cyclise the BOM-protected amino-alcohol 47b to establish if this isomer could produce the 3-aza-A-ring target.Our initial approach to employ Mitsunobu reaction conditions furnished the desired cyclised product 48 albeit in a low 21% yield (Scheme 4).Alternatively, Cbz-protection of the amine in 47b followed by mesylation of the resulting alcohol 51 was found to lead directly to the desired 3-aza-derivative 49 via facile intramolecular cyclisation (52).To confirm both cyclisation methods gave the same product, compound 48 was Cbz-protected to afford a product with identical NMR spectra to those of 49.Final cleavage of the protecting groups via catalytic hydrogenolysis and acid-mediated ester hydrolysis delivered the desired 3-aza-obeticholic acid (6⋅HCl).This step, however, proved problematic as only a trace amount of product 6 was detectable, and instead an Nmethyl species 50 was formed as the major product.We attributed the formation of 50 to an extremely efficient reductive amination reaction resulting from the one equivalent of formaldehyde released from hydrogenolysis of the BOM group.Adding aldehyde-scavenging reagents (i.e.ammonia/hydrazine) to the catalytic hydrogenolysis did not prevent this unwanted side-reaction.However, the reductive amination (49 → 50) could be overcome by performing the catalytic hydrogenation of intermediate 49 in the presence di-tert-butyl dicarbonate, which gave the N-Boc-protected alcohol 53 in 84% yield.Furthermore, the intermediate 53 could be prepared in a similar yield by direct Boc-protection of amine 47b followed by the mesylation-cyclisation sequence giving rise to intermediate 54, and subsequent catalytic hydrogenation.Final deprotection of the Boc and tert-butyl ester groups from precursor 53 was accomplished by exposure to aqueous hydrochloric acid in THF, furnishing the desired 3-aza-OCA 6 which crystallised from the reaction mixture.X-ray crystallographic analysis of suitable crystals of 6 Scheme 1. Synthesis of amino-alcohol 37 from OCA (2).Scheme 2. Attempts to cyclise amino-alcohol 37. confirmed its structure.
While BOM-protection of the 7α-OH proved to be a successful strategy in converting regiomer 47b (formed from the minor lactone isomer 45b) to the targeted 3-aza-BA 6, subjecting regiomer 47a (derived from the major lactone isomer 45a) to the same reaction conditions revealed that the method was not transferable.Here, we observed the formation of the undesired cyclic ethers 39 and 43, respectively, when either the Mitsunobu or the mesylation strategies were employed (Scheme 5).
To mitigate this issue, we replaced the previous mesylation step with a Dess-Martin oxidation and deployed a tandem oxidative cyclisation-dehydration method developed by Ishizaki et al. [44a] Here, the intermediate aldehyde underwent spontaneous and rapid cyclisation to afford the Boc-protected enamine 57 in excellent yield (Scheme 6).
During our experiments we did not observe the intermediate hemiaminal 56.Despite a sluggish catalytic hydrogenation, the desired Scheme 3. Synthesis of the BOM-protected amino alcohols 47a and 47b.Scheme 4. Cyclisations of the BOM-protected amino-alcohol 47b and elaboration to 3-aza-OCA (6⋅HCl).
stereoselectivity in the reduction afforded the previously characterised N-Boc-protected 3-aza-OCA derivative 53 in 71% yield.Screening 3-aza-OCA (6⋅HCl) at a concentration of 10 µM for either agonistic or antagonistic activity at FXR and TGR5 revealed that this compound displays no activity.To attempt to rationalise these results, we evaluated 3-aza-OCA (6) in comparison with the 3-amino-bile alcohol 5 by molecular modelling for their binding interactions at TGR5 and FXR.These efforts revealed, that in the TGR5 receptor, 3-aza-OCA did not bind deep enough into the orthosteric site.Thus, the compound was not able to interact with Tyr240, which has been shown previously to be important for sensing the agonistic signal [46].In contrast, the amino group attached to C3 of compound 5 extends further and can interact with Tyr240 (Fig. 4A).
When modelled to the FXR binding site, we found that both compounds caused a change of the side chain conformation of His444, which likely disrupts the pi-pi stacking and pi-cation interactions between His444 and Trp466 which is essential for FXR activation (Fig. 4B) [47].
Despite the inactivity of 3-aza-OCA (6), employing the chemical methodologies described here enabled the preparation of other 3-azabile acid derivatives [37b], which emerged as promising leads for the treatment of neurodegenerative disease such as Parkinson's and Alzheimer's diseases and will be reported in due course.

Conclusions
In conclusion, we have developed a mild synthesis of 3-aza-OCA.Selective oxidation of the bile acid 3-hydroxy group, followed by BV oxidation of the resultant 3-ketone gave a mixture of regiomeric lactones.Although we have not optimised the BV regioselectivity, this has been investigated previously in the literature and the regioselectivity is largely substrate dependant [40].Thus, we have established methods to elaborate both regiomers of the BV oxidation to 3-aza-BAs.Ammonolysis of the lactones to the corresponding regiomeric amido-alcohols was achieved when a tert-butyl ester protecting group was adopted for the side-chain carboxylic acid.Subsequently, a mild Hofmann degradation afforded the respective amino-alcohols.In the obeticholic acid system, with a 6α-ethyl group and a 7α-hydroxyl, annulation of the aminoalcohol regiomers required separate approaches.The minor regiomer (47b) could easily be cyclised by carbamate-protection of the amine group followed by mesylation of the primary alcohol.However, the proximity and orientation of the 7α-OH prohibited application of this method to the major amino-alcohol regioisomer (47a), as a previously undescribed and facile etherification reaction occurred when the proximal primary alcohol was activated.The major regiomer (47a) required carbamate-protection of the amine group followed by Dess-Martin oxidation of the alcohol to form enamine 57 in a tandem oxidative cyclisation-dehydration sequence.This protected enamine could be stereoseletively hydrogenated to give the protected 3-aza-bile acid precursor (53) and converged the synthetic sequence for both regiomers.Overall, our synthetic approach is scalable and applicable to the synthesis of a diverse range of 3-aza-BAs [37].
While biological assessment of 3-aza-OCA (6) did not display agonistic or antagonistic activity at the target receptors FXR and TGR5, respectively, our subsequent synthetic efforts have enabled the identification of promising lead compounds for the treatment of neurodegenerative disease such as Parkinson's and Alzheimer's diseases which we will report in due course [37b].

General synthetic methods
Proton ( 1 H) and carbon ( 13 C) NMR-spectra were recorded on Bruker Avance (III)-500.Chemical shifts are reported in ppm relative to Me 4 Si (TMS, δ 0), or residual solvent peaks as an internal standard set to δ 7.26 and 77.00 (CDCl 3 ), or δ 3.31 and 49.00 (CD 3 OD).NMR data is reported as follows: chemical shift in ppm, multiplicity (br = broad, s = singlet, d = doublet, t = triplet, dd = doublet of doublets, td = triplet of doublets, dt = doublet of triplets, m = multiplet), coupling constant in Hz, integration.Electrospray ionization (ESI) mass spectrometry (MS) experiments were performed on a quadrupole time-of-flight (QTOF) Premier mass spectrometer (Micromass, UK) under normal conditions.Sodium formate solution was used as calibrant for high resolution mass spectra (HRMS) measurements.Crystallographic data were collected on an Agilent SuperNova diffractometer fitted with an EOS S2 detector.Specific optical rotations were acquired on a Rudolph Autopol IV Automatic polarimeter at ambient temperature (20 • C), unless otherwise stated, λ = 589 nm and concentration (g/100 mL) in the solvent indicated, using a cell of path length 100 mm.All reactions were monitored by thin layer chromatography (TLC) using 0.2 µm silica gel (Merck Kieselgel 60 F 254 ) precoated aluminium plates, using UV light, ammonium molybdate, or potassium permanganate staining solution to visualise.Flash column chromatography was performed on Davisil® silica gel (60, particle size 0.040-0.063mm), or using Reveleris® silica or C-18 reversed phase flash cartridges on a Grace Reveleris® automated flash system with continuous gradient facility.Solvents for reactions and chromatography were analytical grade and were used as supplied unless otherwise stated.Petroleum ether (PE) refers to the fraction boiling at 60-80 • C, throughout.The starting material, OCA (2), was provided by New Zealand Pharmaceuticals (NZP).Following a literature procedure for the preparation of tert-butyl cholate [38]: to a solution of obeticholic acid 2 (OCA; 7.92 g, 18.8 mmol) in tetrahydrofuran (60 mL) was added trifluoroacetic anhydride (TFAA; 10 mL, 71.2 mmol) dropwise at 0 • C. The mixture was allowed to warm to room temperature.After 15 min the mixture was treated with tert-butanol (28.7 mL, 315 mmol) and left to stir overnight (o/n).The mixture was cooled to 0 • C and saturated aqueous ammonia (25 mL) was added slowly.The mixture was left to warm to room temperature and stirred for 5 h before being diluted with diethyl ether and successively washed with a 1 M aqueous solution of sodium hydroxide (100 mL) and water (2 × 100 mL).The combined organic layers were washed with brine, dried over MgSO 4 and concentrated to give 8 g (88%) of the tertbutyl ester 33 which was used in the next step without further purification.An analytical sample was purified by automated column chromatography on silica gel gradient eluting with ethyl acetate/petroleum ether 0-30%.The compound was obtained as a colorless oil.R f = 0.55
Method B: Compound 43 (48 mg, 0.081 mmol) was dissolved in methanol and 10% palladium on charcoal (20 mg) was added.The atmosphere was exchanged with hydrogen and the reaction mixture was stirred for 20 h.After complete hydrogenation (TLC analysis) the reaction was filtered through celite and filtrate concentrated to afford compound 39 as an oil (quant.).
Method C: Following the procedure as described for Method A, amino-alcohol 47a (110 mg, 0.183 mmol) gave 30 mg (28%) of compound 39 as an oil.
Method B: Following the procedure as described under Method A, compound 55 (292 mg, 0.40 mmol) gave rise to 190 mg (80%) of 43 as a gum.
Method B: Compound 49 (39 mg, 15%) could be isolated from the attempted preparation of mesylate 52 (below, 4.2.19).Mesylate 52 could also be dissolved in pyridine and heated (70 • C) to be completely converted quantitatively to cyclized product 49.
Method B: To a solution of the Boc-protected compound 54 (688 mg, 1.01 mmol) in methanol (25 mL) was added 10% palladium on charcoal (96 mg).Then, the atmosphere was exchanged with hydrogen and the mixture was stirred overnight.After complete hydrogenation (TLC analysis), the reaction was filtered through celite and was concentrated in vacuo.The crude was purified by column chromatography over silica gel on automated column chromatography (silica gel, ethyl acetate/ petroleum ether 0-20%) to yield 499 mg (88%) of compound 53 as a foam.

Fig. 4 .
Fig. 4. (A) Modelled poses of 3-aza-OCA (6, carbon skeleton in green) and the 3-amino-bile alcohol 5 (carbon skeleton in yellow in human TGR5 receptor.(B) Modelled poses of 3-aza-OCA (6, carbon skeleton in green) and compound 5 (carbon skeleton in yellow) in rat FXR receptor, superimposed with the crystal structure of rat FXR receptor with OCA (2, carbon skeleton in magenta) bound (PDB 1OSV).Carbon atoms of His444 and Trp466 are in the same colour as the corresponding BA in the binding site.(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) , 11.74 (C18), 11.48; HRMS (ESI) m/z calcd for C 30 H 50 O 4 Na + 497.3601, found 497.3610.