Synthesis of (+)-ribostamycin by catalytic, enantioselective hydroamination of benzene

Abstract

Aminoglycosides (AGs) represent a large group of pseudoglycoside natural products, in which several different sugar moieties are harnessed to an aminocyclitol core. AGs constitute a major class of antibiotics that target the prokaryotic ribosome of many problematic pathogens. Hundreds of AGs have been isolated to date, with 1,3-diaminocyclohexanetriol, known as 2-deoxystreptamine (2-DOS), being the most abundant aglycon core. However, owning to their diverse and complex architecture, all AG-based drugs are either natural substances or analogues prepared by late-stage modifications. Synthetic approaches to AGs are rare and lengthy; most studies involve semi-synthetic reunion of modified fragments. Here we report a bottom-up chemical synthesis of the 2-DOS-based AG antibiotic ribostamycin, which proceeds in ten linear operations from benzene. A key enabling transformation involves a Cu-catalyzed, enantioselective, dearomative hydroamination, which set the stage for the rapid and selective introduction of the remaining 2-DOS heteroatom functionality. This work demonstrates how the combination of a tailored, dearomative logic and strategic use of subsequent olefin functionalizations can provide practical and concise access to the AG class of compounds.

Graphical Abstract

Since their initial discovery in 1943¹, aminoglycosides (AGs) have become an esteemed class of carbohydrate-derived antibiotics obtained predominantly from actinomycetes; several being registered on the World Health Organization’s List of Essential Medicines². The most extensively investigated and clinically employed AGs contain a central 1,3-diaminocyclohexanetriol scaffold (Figure 1a), known as 2-deoxystreptamine (2-DOS, 1)³, which is adorned with a variety of sugars, as exemplified with ribostamycin (2)⁴, neomycin B (3)⁵, and sisomicin (4)⁶. Their broad-spectrum antibiotic activity derives from the suppression of protein synthesis by binding to prokaryotic rRNA^7,8, and they are particularly useful against several important Gram-negative bacteria as well as Mycobacterium tuberculosis. Despite their impressive anti-infective properties, AGs often suffer from the evolution of resistant bacterial strains^9,10, as well as exhibiting undesired side effects, such as acute oto- and nephrotoxicity in patients¹¹, restricting the prescription of AGs primarily to last resort treatment of life-threatening infections. On the other hand, synthetically modified analogues have shown improved safety and efficacy within this class of antibiotic. However, owing to their stereochemical complexity and numerous functional groups, such modifications are limited only to simple functionalizations of primary alcohols and amines on the periphery of natural products¹². Nevertheless, late-stage derivatization has provided several clinically important drugs; plazomicin (5) being the latest in this series¹³ gaining FDA approval in 2018.

Fig. 1 |. Aminoglycosides and synthetic considerations.

a, 2-Deoxystreptamine (2-DOS, 1) is a central aminocyclitol present in natural AG antibiotics, such as ribostamycin (2), neomycin B (3), sisomicin (4), as well as semisynthetic plazomicin (5)³. b, State of the art asymmetric approaches to 2-DOS aminocyclitols involving ring construction (6→7)¹⁹, sugar modification (8→9)²⁰, and sequential substitution (10→11)²¹. c, Olefin functionalization approach to 4,5-diglycosylated 2-DOS AG 12 reveals asymmetric dearomative hydroamination of benzene (14→13) as a suitable entry into synthesis of AGs. Previously reported hydroamination of benzene provided low yields, and racemic 1,2-hydroaminated product (±)-15 as a inseparable mixture alongside 16 and 17²⁴.

Bottom-up synthetic approaches could potentially provide more efficient access to AG analogues and enable further investigations into this biologically important class of compounds. However, the inherent structural complexity has placed significant practical limitations on new synthetic routes. Only a few total syntheses have been reported to date^14–17, requiring numerous steps and unselective glycosylations which prevented their translation into medicinal chemistry studies. On the other hand, asymmetric syntheses of differentiated 2-DOS derivatives from different precursors are more established (Fig. 1b)¹⁸; ranging from ring construction strategies (6→7)¹⁹, elaboration of carbohydrate chiral pool precursors (8→9)²⁰, and to functional group manipulation of carbocycles (10→11)²¹. However, due to the structural issues associated with the 2-DOS framework, these approaches rely on the extensive use of different modifications of pre-existing functionality to change each group one by one. Such an elaborate collection of functional group interconversions (FGIs) ultimately results in lengthy syntheses, regardless of the nature of the starting material, which have not been further elaborated into glycosylated AGs. Guided by this analysis and previous challenges, we realized that a global olefin functionalization approach could reduce FGIs as it could directly install two desired functional groups per olefin in a stereoselective manner (Fig. 1c). According to this plan, the most logical disconnection of the 2-DOS AGs of type 12 can be traced back to reduced aniline intermediate 13, such that hydroamination of benzene (14) could serve as an ideal template for introduction of five contiguous heteroatom substituents²². Although hydroamination of olefins has been extensively developed within the last two decades²³, the corresponding dearomative transformation involving arenes is virtually nonexistent. Only one report documents the UV light-mediated addition of amines to benzene (Fig. 1c, inset)²⁴, delivering the 1,2-hydroaminated product 15 as a racemate in low yield as an inseparable mixture alongside the corresponding constitutional isomer 16 and rearomatized aniline 17. We imagined that the development of an efficient dearomative hydroamination would be an empowering method and also a formidable challenge in view of the shortcoming apparent from the reported example. The work reported herein features the development of a catalytic, enantioselective, dearomative hydroamination of benzene, which enables the synthesis of (+)-ribostamycin (2) in ten synthetic operations.

Results and discussion

With the foregoing analysis in mind, the bottom-up synthesis of (+)-ribostamycin (2) was initiated with the methodological challenge of a dearomative hydroamination of benzene. This approach was inspired by previous dearomative processes involving photoactivatable arenophiles²⁵, as they have successfully served in the formation of densely functionalized, heteroatom-rich motifs. However, these examples could not be adapted to the 2-DOS core, as it is distinct from other structures owing to the presence of a methylene group within the aminocyclitol framework. We hypothesized that the combination of visible-light-promoted para-cycloaddition with arenophile N-methyl-1,2,4-triazoline-3,5-dione (MTAD, 18) and copper-promoted ring opening could provide the requisite amine and methylene, formally resulting in dearomative 1,2-hydroamination (Fig. 2a). Mechanistically, we envisioned this catalytic process could proceed through two distinct, yet convergent pathways, involving hydrocupration²⁶ or allylic substitution²⁷ (Fig. 2b). In a hydrocupration scenario, a ligated copper(I) hydride species (L_nCuH) undergoes π-complexation to MTAD-benzene cycloadduct 19 followed by hydrometallation anti to the arenophile moiety (19→I→II). The resulting organocopper intermediate II is poised for β-elimination of bridgehead arenophile motif (urazole) to deliver diene III. An additional equivalent of silane mandates concomitant regeneration of the copper hydride species and release of product 13. On the other side, copper π-complex I could also undergo oxidative addition, forming organocopper species IV, which upon reductive elimination delivers diene III. Finally, because of the symmetrical nature of MTAD-benzene adduct 19, a suitable chiral ligand bound to the Cu center could enable stereodifferentiation, forming the desired product through an enantioselective fashion in either mechanistic manifold (i.e., I→II or I→IV).

Fig. 2 |. Design of Cu-catalysed 1,2-hydroamination of benzene.

a, Cycloaddition between benzene (14) and MTAD (18) provides cycloadduct 19, which upon exposure to catalytically-generated copper hydride delivers hydroaminated product 13. b, Plausible catalytic cycle for this reaction that proceeds through Cu-olefin ligated complex I, involving either hydrocupration (I→II) and β-elimination (II→III), or an oxidative addition (I→IV) and reductive elimination (IV→III) pathway.

Armed with this plan, we began our investigations by conducting a series of orienting studies for the desired dearomative hydroamination (Fig. 3a, see Supplementary Tables 1–3). The optimized reaction conditions involved formation of the cycloadduct between the benzene (14) and MTAD (18), followed by addition of a premixed solution of Cu salt (CuOAc or CuTC), ligand, and potassium tert-butoxide with subsequent introduction of diethoxymethylsilane. Several privileged chiral ligands were viable for this asymmetric dearomative strategy, ranging from NHC-type ligand (20, 74% yield and 89:11 e.r.) to P,P-bidentate BINAP (21, 74% yield and 79:21 e.r.) and TaniaPhos (22, 69%, 96:4 e.r.). To gain preliminary mechanistic insight, benzene-d₆ (14-d₆) was employed as the substrate to effectively track the fate of the introduced hydride (Fig. 3b). This experiment revealed the exclusive formation of 13-Me-d₆, providing definitive proof of high site- and stereoselectivity, as well as supporting the mechanisms proposed above (Fig. 2b).

Fig. 3 |. Enantioselective Cu-catalysed 1,2-hydroamination of benzene.

a, Optimized reaction conditions and selected results with ligands 20–22. Yields were determined by ¹H NMR integration relative to the internal standard (MeNO₂). Methylation of product 13 (quench with Me₂SO₄/K₂CO₃ to provide 13-Me) was necessary to facilitate determination of enantiopurity by HPLC. b, Dearomatization of benzene-d₆ (14-d₆) as a preliminary mechanistic probe.

Density functional theory calculations were performed to investigate the mechanism and the origin of enantioselectivity of the dearomative hydroamination (Fig. 4). The DFT calculations indicated that the hydrocupration pathway is more favourable than the allylic substitution pathway (see Supplementary Fig. 6). Upon dissociation of the TaniaPhos-supported dimeric CuH catalyst²⁸ (23) to monomeric LCuH (24), hydrocupration of the MTAD-benzene adduct 19 requires a low activation free energy of 21.9 kcal/mol (TS1a) with respect to the dimeric CuH resting state 23. The resulting alkyl Cu species (25) undergoes facile anti-β-elimination (TS2) to form π-alkene Cu complex 26, which then isomerizes to the more stable product complex 27. Because the hydrocupration is exothermic and irreversible, this step desymmetrizes adduct 19 and determines the product enantioselectivity. Among eight possible stereoisomers of the hydrocupration transition state (see Supplementary Fig. 4), TS1a is the most favourable and leads to the (S)-hydroamination product after β-elimination. This is consistent with the (S)-product observed experimentally (Fig. 3a). The enantioselectivity is controlled by steric repulsions in the hydrocupration transition state²⁹—in the lowest-energy transition state TS1a, adduct 19 is placed in the least occupied quadrant (III) of the (S,S_P)-TaniaPhos-supported CuH catalyst, whereas in TS1b, the substrate is located in the more occupied quadrant (IV), causing greater steric repulsions with the TaniaPhos ligand.

Fig. 4 |. Computational studies.

a, Computed reaction energy profile indicating a facile hydrocupration (TS1a)/β-elimination (TS2) process. b, Origin of enantioselectivity. In the hydrocupration transition state leading to the major enantiomeric product (TS1a), the MTAD-benzene adduct 19 is placed in the least occupied quadrant of the TaniaPhos-supported Cu catalyst. DFT calculations were performed at the M06/SDD–6–311+G(d,p)/SMD(CH₂Cl₂)//B3LYP/SDD–6–31G(d) level of theory

With dearomative hydroamination developed and key intermediate 13 in hand, we set our attention toward the synthesis of ribostamycin (Fig. 5). We were able to conduct the hydroamination step on 100 mmol scale, which provided a reliable supply to decagram quantities of the key intermediate; albeit erosion of enantioselectivity was observed due to challenges associated with temperature control on a larger scale. Moreover, due to sensitive nature of diene 13, this product was directly subjected to a one-pot tert-butylcarbonate (Boc) protection and epoxidation sequence yielded allylic epoxide 28 as a single constitutional and diastereoisomer. The Boc protection proved crucial for obtaining the desired anti-configuration between both functionalities. Treatment of 28 under basic conditions (Na₂CO₃ in methanol) allowed for Boc deprotection with concurrent 5-exo-tet cyclization of the urazole moiety. Solvent exchange and exposure of the resulting secondary allylic alcohol to benzyl bromide and NaH afforded benzylic ether 29. Importantly, the installation of the bridgehead urazole served as a strategic cornerstone on several levels: (1) it installed both nitrogen atoms in a required syn-1,3-relationship; (2) provided a unique and robust protecting group for the diamine throughout the synthesis and, (3) served as a conformational directing group, guiding downstream selectivity of an incoming oxygen nucleophile (vide infra).

Fig. 5 |. Synthesis of (+)-ribostamycin (2) from benzene (14).

Reagents and conditions: (2) Boc₂O (1.5 equiv.), DMAP (0.15 equiv.), CH₂Cl₂, 0 to 23 °C, 30 min; then NaHCO₃ (1.4 equiv.), mCPBA (1.3 equiv.), MeOH, 65 °C, 8 h, 45%; (3) Na₂CO₃ (3.0 equiv.), MeOH, 65 °C, 24 h; then removal of solvents, NaH (2.0 equiv.), BnBr (2.0 equiv.), DMF, 0 to 23 °C, 24 h, 54%; (4) NBS (2.0 equiv.), I₂ (0.2 equiv.), MeCN/H₂O, 0 to 23 °C, 16 h; then K₂CO₃, MeOH, 65 °C, 8 h, 70%; (5) acetophenone oxime (2.0 equiv.), nBuLi (1.9 equiv.), toluene, 0 °C, 10 min; then 30 in DMF, 80 °C, 12 h, 69%; (6) Lawesson’s reagent (1.0 equiv.), AgClO₄ (2.0 equiv.), 32 (3.0 equiv.), CH₂Cl₂, 23 °C, 12 h, 66%; (7) Zn (10 equiv.), AcOH/MeCN/H₂O, 0 to 80 °C, 4 h, 90%; (8) 35 (3.0 equiv.), NIS (3.0 equiv.), TfOH (6.6 equiv.), −50 to −20 °C, 3 h, 58% (75% based on recovered 34); (9) KOH (20 equiv.), iPrOH, 80 °C, 24 h; then neutralization with AcOH, CuCl₂ (2.0 equiv.), 23 °C, 12 h, 60%; (10) Pd(OH)₂ (20 wt %), H₂ (130 psi), EtOAc/MeOH/H₂O, 23 °C, 48 h; then filtration and solvent evaporation, Pd(OH)₂ (20 wt %), H₂ (130 psi), AcOH/H₂O, 23 °C, 6 days, 91% (isolated as an tetraacetate salt). The Fürst-Plattner analysis reveals that the nucleophilic opening of epoxide 30 occurs with desired chemoselectivity only in case when nitrogen atoms are tethered. In the open form, the attack at the same position would proceed through the unfavored twist boat-like transition state and is unfavourable.

A final olefin functionalization, this time an epoxidation³⁰ proceeding through a bromohydrin followed by treatment with base, furnished bicyclic epoxide 30 as a single diastereoisomer. This intermediate is one step away, involving an epoxide opening with suitable O-nucleophile, from a fully assembled 2-DOS aminocyclitol core. The regioselectivity of this process was controlled by the conformationally locked syn-1,3-tethered urazole, permitting an otherwise unfavorable kinetic differentiation during the epoxide opening (see comparative Fürst–Plattner analysis, Fig. 5, bottom inset). Nevertheless, incorporation of the final oxygen atom to the core proved nontrivial, as most oxygen nucleophiles were either unreactive or incompatible with the substrate under more forcing conditions. Gratifyingly, we found the nucleophilic character of lithiated acetophenone ketoxime to be perfectly tempered³¹, as evidenced by the exclusive formation of product 31. Notably, the newly installed oxime ether served as a complementary protecting group to the benzyl ether. Finally, according to our dearomative retrosynthetic logic, this step completes the short functionalization sequence that converts benzene into the fully decorated 2-DOS core, installs all five contiguous stereocenters, and sets the stage for the selective glycosylations.

To avoid extensive manipulation of protecting groups, we chose D-ribose and D-neosamine glycoside donors 32 and 35 (Fig. 5, bottom inset), tailored for the global deprotection with contemporaneous azide reductions. After an exhaustive screening campaign, glycosylation of 31 with benzylated ribose 32 and an in-situ-generated activator from Lawesson’s reagent and silver perchlorate³², delivered pseudodisaccharide 33 with exclusive β-selectivity. Thereafter, reduction with zinc in acetic acid chemoselectively removed the oxime group and revealed the secondary alcohol 34. A second, this time α-selective glycosylation using thioglycoside donor 35 in the presence of NIS and triflic acid³³, afforded fully decorated pseudotrisaccharide 36 as a single diastereoisomer. To secure the natural product through a global reduction, the urazole motif was first converted to cyclic diazene 37 by exposure to base, followed by oxidation of the corresponding cyclic hydrazine with copper chloride³⁴. The final reduction, resulting in removal of six benzyl groups and conversion of two azides and the bridgehead diazine to the corresponding amines, required Pearlman’s catalyst in a mixture of water, methanol, and ethyl acetate at 130 psi of hydrogen, followed by solvent swap to aqueous acetic acid and addition of a second portion of the catalyst³⁵. Filtration of the reaction mixture and solvent removal provided (+)-ribostamycin (2) as a tetraacetate salt, which matched the authentic sample in all physical and spectroscopic aspects. Notably, this two-stage reductive protocol was essential in obtaining a pure product, obviating the use of size exclusion chromatography, a commonly used purification technique for this highly polar class of compounds.

The culmination of this work represents development of dearomative hydroamination (Figs. 2 and 3) and a concise bottom-up approach to the 2-DOS AG (+)-ribostamycin from benzene (Fig. 5). Additional salient features include the strategic application of an arenophile motif (urazole), which served as a nitrogen source for the syn-1,3-diamine moiety, and as a controlling element during the selective introduction of oxygen functionality. Moreover, selective glycosylation introduced both carbohydrates with pertinent functionality amenable to global reduction, delivering the final product in ten operation from benzene. Most of the synthetic manipulations were conducted on a gram scale and this route already provided several hundred milligrams of the natural product. The described synthetic platform constitutes a practical and rapid preparation of a variety of 2-DOS AGs and their analogues, enabling explorations toward the development of more effective and safe antibiotics. Similarly, the dearomative hydroamination should also provide access to tailored aminocyclitols that would be challenging to prepare using conventional chemistry.

Methods (including Data Availability, Code Availability and Statistics subsections where relevant):

In the procedure for enantioselective dearomative hydroamination, MTAD (18, 2.83 g, 25 mmol, 1.0 eq.) was dissolved in anhydrous dichloromethane (167 mL) and degassed benzene (14, 11.2 mL, 125 mmol, 5.0 eq.) was added. The solution was cooled to −78 °C and the flask was irradiated with white LEDs at −78 °C until the pink color disappeared. In a separate flask, a catalyst solution was prepared from CuOAc (0.138 g, 1.13 mmol, 4.5 mol%), Taniaphos (22, 0.860 g, 1.25 mmol, 5.0 mol%), and KOtBu (3.37 g, 30 mmol, 1.2 eq.), dissolved in dry, degassed toluene (50 mL) and stirred for 30 minutes. Note: the solids should be ground to a fine powder to ensure that the cannulation process is not clogged at the low temperature. The LEDs for the reaction were turned off, and the cooled (−78 °C) catalyst solution was then cannulated (14 gauge cannula) into the reaction along the wall of the flask. After catalyst addition was complete, diethoxymethylsilane (10 mL, 62.6 mmol, 2.5 eq.) was added dropwise, and reaction mixure was allowed to slowly warm to −50 °C and stirred at this temperature for 12 hours. Then the reaction was warmed to −20 °C and quenched with water (150 mL). After warming to room temperature with vigorous stirring, the reaction mixture was poured into a separatory funnel and the organic phase was drained and discarded. Then, the aqueous layer was transferred to an Erlenmeyer flask and carefully stirred and acidified to pH 4.5 using citric acid (0.2 M, aq. sol., 30 mL). The organic phases were extracted with CH₂Cl₂ (6 × 200 mL), dried over anhydrous MgSO₄, filtered, and concentrated under reduced pressure to afford diene 13 as a light-yellow solid (3.13 g, 16.2 mmol, 65%).

Data Availability

The experimental data as well as the characterization data for all the compounds prepared during these studies are provided in the Supplementary Information. Cartesian coordinates of all optimized geometries are provided in a separate file in the .xyz format. Crystallographic data for the structures reported in this article have been deposited at the Cambridge Crystallographic Data Centre, under deposition number CCDC 2113321 (30). Copies of the data can be obtained free of charge via https://www.ccdc.cam.ac.uk/structures.