Mana Mohan Mukherjee1, Peng Xu1, Edwin D Stevens2, Pavol Kováč1. 1. LBC, NIDDK, National Institutes of Health, Bethesda, MD 20892-0815. 2. Department of Chemistry, Western Kentucky University, Bowling Green, KY, 42101.
Abstract
Synthesis of the upstream terminal hexasaccharide part of the lipopolysaccharides (LPS) of Vibrio cholerae O1, serotype Inaba has been improved. The key improvements include but are not limited to optimized conditions for the stereoselectivity of glycosylation reactions involved and fewer number of synthetic steps, compared to previous approaches. Particularly noteworthy is conducting the glycosylation of the very reactive glycosyl acceptor 8-azido-3,6-dioxaoctanol with the fully assembled hexasaccharide trichloroacetimidate under thermodynamic control. It produced the desired α glycoside with an α:β ratio of 7:1, compared with the ratio of 1.1:1, observed when the coupling was conducted conventionally. Several substances, which were previously obtained in purity acceptable only for synthetic intermediates, were now obtained in the analytically pure state and were fully characterized. The structure of the key trisaccharide glycosyl acceptor was confirmed by single-crystal x-ray structure determination.
Synthesis of the upstream terminal hexasaccharide part of the lipopolysaccharides (LPS) of Vibrio cholerae O1, serotype Inaba has been improved. The key improvements include but are not limited to optimized conditions for the stereoselectivity of glycosylation reactions involved and fewer number of synthetic steps, compared to previous approaches. Particularly noteworthy is conducting the glycosylation of the very reactive glycosyl acceptor 8-azido-3,6-dioxaoctanol with the fully assembled hexasaccharide trichloroacetimidate under thermodynamic control. It produced the desired α glycoside with an α:β ratio of 7:1, compared with the ratio of 1.1:1, observed when the coupling was conducted conventionally. Several substances, which were previously obtained in purity acceptable only for synthetic intermediates, were now obtained in the analytically pure state and were fully characterized. The structure of the key trisaccharide glycosyl acceptor was confirmed by single-crystal x-ray structure determination.
The upstream, terminal part of the lipopolysaccharides (LPS) of most Gram-negative bacteria consist of several copies of oligosaccharide, sometimes monosaccharide repeating units. The said macromolecules are termed O-specific polysaccharides (O-SP) and are recognized as protective antigens of these pathogens. O-SPs confer specificity to their homologous antibodies and they, or fragments thereof, are paramount in the development of synthetic/conjugate vaccines. We have been involved in work towards vaccines from synthetic fragments of O-SPs for a number of years. One of our major targets has been a conjugate vaccine for cholera from synthetic fragments of O-SPs of Vibrio cholerae.[1-3]The O-SP of Vibrio cholerae O1, serotype Inaba consists of 12–18 repeats[4] of (1→2)-α-linked perosamine (4-amino-4,6-dideoxy-d-mannose) whose amino group is acylated with 3-deoxy-l-glycero-tetronic acid. We have previously synthesized hexasaccharide fragments of the O-SP and have determined essential immune responses of conjugates therefrom in mice.[5] Chemical structures of O-SPs of the two strains (Inaba and Ogawa) of Vibrio cholerae O1 are the same, except that the terminal, upstream perosamine residue in the Ogawa strain carries a methyl group at O-2. To be able to compare immune responses of the conjugates from the hexasaccharide fragments of the O-SP of both serotypes of Vibrio cholerae O1 with those of similar conjugates from synthetic polymers representing the complete O-SPs, we intend to synthesize glycoconjugates from the analogous octadecasaccharides. Syntheses of such structures are much more involved undertakings. Previous syntheses of the hexasaccharide antigens comprised up to more than 40 linear steps, depending on the individual approach.[1,2] The key intermediates within our strategy towards the octadecasaccharides will be synthons derived from the related hexasaccharides. With the aim to increase the feasibility of a large-scale synthesis required by future immunization studies and decrease the number of synthetic steps involved in making such substances, the objective of this work was to test the practicability and scalability of the current, new synthetic scheme. Thus, we synthesized on large scale the trisaccharide glycosyl donor and acceptor 5 and 4, respectively, and used these to prepare, also on large scale, the related hexasaccharide 3 (Fig. 1). These substances are versatile intermediates, which we intend to use to make considerably larger fragments of the O-SP, up to the complete bacterial O-specific antigen, octadecasaccharide. Using hexasaccharide 3, we proceeded to complete the synthesis of the title hexasaccharide 1. The present pathway is the shortest published to date (33 linear steps). An additional advantage of the approach described here is that the stereoselectivity of the critical glycosylation reaction, which converted the terminal determinant to a conjugation-ready form, was substantially increased by controlling it thermodynamically.[2,6]
Fig. 1
Retrosynthetic analysis of the hexasaccharide 1.
Results and discussion
Synthesis of the title hexasaccharide started with the known[7] methyl 4-azido-2-O-benzoyl-3-O-benzyl-4,6-dideoxy-α-d-mannopyranoside (10). It was treated with Ac2O in presence of H2SO4, and the formed (Scheme 1) 1-O-acetyl-4-azido-2-O-benzoyl-3-O-benzyl-4,6-dideoxy-α,β-d-mannopyranose (11)[7] where the α-anomer largely predominated was resolved by chromatography. Unlike previously, the α-acetate 11a was now obtained crystalline for the first time. We planned to use the latter as a glycosyl donor under BF3·Et2O or TMSOTf catalysis[8,9] to make the corresponding 2-trimethylsilylethyl (SE) glycoside 13a, and perhaps also higher oligosaccharides in this series, thereby improving the economy of the overall synthesis. Unfortunately, this approach was unsuccessful: only the corresponding 1-OH compound was formed. Similarly unsuccessful was the reaction of the corresponding 2-O-acetyl derivative 12 (→13b), prepared conventionally[10] (these reactions are not described in the Experimental). Explanation for these failures was not sought, but in their extensive synthetic study towards this class of substances, Saksena et al.[11] observed that formation of 2-trimethylsilylethyl mannopyranosides is a complex process whose outcome is largely unpredictable. In contrast with the unsuccessful Lewis acid mediated reaction of 11 or 12 with 2-trimethylsilylethanol, the BF3·Et2O-catalyzed reaction of 11 with ethanethiol readily produced ethyl 4-azido-2-O-benzoyl-3-O-benzyl-4,6-dideoxy-1-thio-α,β-d-mannopyranoside (7a and hitherto unknown 7b),[11] which was fully characterized. The latter mixture of anomers was converted to its corresponding SE glycoside 13a as described,[11] and the pure α-anomer 7a was used as glycosyl donor towards oligosaccharide synthesis (Schemes 2 and 3). The initial glycosyl acceptor 8[7] (Scheme 1) was prepared by conventional debenzoylation (Zemplén[12]) from the foregoing 2-O-benzoate 13a.
Scheme 1
Synthesis of the monosaccharide building blocks.
Scheme 2
Synthesis of the disaccharide acceptor 6.
Scheme 3
Synthesis of trisaccharide donor 5 and acceptor 4.
With suitably equipped glycosyl donor and acceptor at hand, we set out for oligosaccharide synthesis on the multigram scale (see Experimental). Accordingly, glycosyl acceptor 8 and thioglycoside donor 7a were coupled using NIS/TMSOTf as promoter, affording disaccharide 14 (94%, Scheme 2). The NMR spectra of the disaccharide showed signals characteristic of the presence of both donor and acceptor moieties (165.3 ppm for the benzoyl carbonyl carbon from the donor and −1.32 ppm for SiMe3 carbon from the acceptor), while the α-configuration of the interglycosidic linkage was confirmed from the corresponding NMR spectra, mainly JC,H at the newly formed glycosidic linkage (173.3 Hz).[13] Zemplén debenzoylation of 14 furnished disaccharide acceptor 6,[7] which was subsequently used for chain elongation (Scheme 3). Coupling of disaccharide acceptor 6 and thioglycoside donor 7a in presence of NIS-TMSOTf at 0 °C produced trisaccharide 15 (93%, Scheme 3). Formation of the desired trisaccharide 15 was confirmed by HRMS and the stereochemistry of the newly formed interglycosidic linkage followed from the 13C–1H coupling constant for the anomeric carbon center at 99.2 ppm (C-1III, JC–H = 173.3 Hz). Zemplén debenzoylation of 15 furnished trisaccharide acceptor 4, which was obtained crystalline (CCDC no. 1939745†). Compound 4 crystallizes with 2 independent molecules in the asymmetric unit of the unit cell. Only one free –OH group is available on each molecule for hydrogen bonding, and in each case, a hydrogen bond is formed to an O5 acceptor on a mannopyranoside ring in an adjacent molecule. A modest number of weaker C–H⋯O and C–H⋯N interactions are also observed. Given the limited number of free hydroxyl groups available for hydrogen bonding, the majority of the intermolecular interactions will be weaker non-polar van der Waals type interactions. In this case, the conformations of the polysaccharide chains of the two independent molecules are more likely to reflect a conformational energy minimum for the chain, then would be the case if extensive hydrogen bonding were present. Significantly, the relative conformations of the trisaccharide chains of both independent molecules are very similar, while the orientations of the substituent groups (especially the benzyl and trimethylsilylethyl groups) show greater variation (Fig. 2). Further analysis of the conformation of X-ray data for compound 4 will be reported in a subsequent communication.
Fig. 2
Stereo projections of the two independent molecules of compound 4 at 120 K. Thermal ellipsoids are plotted at the 50% level, and hydrogen atoms have been omitted for clarity. The projection directions have been chosen to highlight the similarity of the core conformations of the two trisaccharide chains.
Compared to the NMR spectra of parent compound 15, disappearance of the carbonyl carbon signal at 165.3 ppm in the 13C NMR spectrum of 4 and upfield shift of H-2III (from δ 5.59 ppm, dd, J2–3 = 2.9 Hz, J2–1 = 2.0 Hz to δ 3.98 ppm, ddd, J2–3 = 2.8 Hz, J2-OH-2 = 1.7 Hz, J2–1 = 1.3 Hz) in the 1H NMR spectrum confirmed the removal of the benzoyl group and formation of the corresponding 2-hydroxy product 4. Treatment of the foregoing 2-trimethylsilyl ethyl glycoside 15 with trifluoroacetic acid (TFA) produced the corresponding trisaccharide hemiacetal 16 (89%). Absence of signals for the 2-(trimethylsilylethyl) group in the 1H NMR spectrum of 16, together with presence of two anomeric C-1I signals in the 13C NMR spectrum (δ 93.5 and 93.2 ppm) confirmed the successful hydrolysis and formation of the desired hemiacetal 16. Subsequent base-catalyzed reaction of 16 with trichloroacetonitrile and DBU produced the corresponding trichloroacetimidate donor 5 (Scheme 3).Large-scale glycosylation (24.3 g of trichloroacetimidate donor 5 with 19 g of trisaccharide acceptor 4) produced hexasaccharide 3 (92%, Scheme 4) with excellent stereoselectivity (α : β ∼ 34 : 1, as shown from integration of signals at δ 5.32 ppm and 5.59 ppm for H-1V of 3a and H-2VI of 3b, respectively, in the NMR spectrum of the crude reaction mixture). Strong contours for anomeric carbons at 101.2 ppm (C-1II), 101.06 ppm (C-1III), 101.04 ppm (C-1IV), 101.03 ppm (C-1V), 100.3 ppm (C-1VI), 99.16 ppm (C-1I) in the HSQC spectrum confirmed the formation of the hexasaccharide. That the predominant isomer contained the desired α-configuration followed from the 13C–1H coupling constant for the 13C carbon involved in the newly formed interglycosidic linkage. For the major isomer, 13C signal of C-1IV at δ 101.04 ppm showed coupling constant JC-1–H-1 = 174.4 Hz whereas the same for the minor isomer was at δ 96.3 with JC-1–H-1 = 155.5 Hz (Fig. 3).
Scheme 4
Synthesis of the hexasaccharide derivative 3.
Fig. 3
(a) Comparison of 13C-NMR spectra of compounds 3a (150 MHz, C6D6) and 3b (150 MHz, CDCl3); (b) 13C–1H coupled 2D-NMR spectrum of compound 3a (C6D6); (c) 13C–1H coupled 2D-NMR spectrum of compound 3b (CDCl3).
Debenzoylation of hexasaccharide 3a under Zemplén conditions produced the 2-hydroxyl group-free intermediate 17.[14] The six azido groups in the foregoing hexasaccharide were transformed, into amines by H2S reduction (→18,[14] 92%, Scheme 5). The next task was to introduce the N-3-deoxy-l-glycero-tetronoyl groups into the foregoing hexaamine 18. This was performed conventionally[2,6] with 2,4-di-O-acetyl-3-deoxy-l-glycero-tetronic acid[15] and 1-[3-(dimethylamino)propyl]-3-ethylcarbodiimide hydrochloride (EDAC). For easier isolation and next reaction, the crude product was directly acetylated, to give the fully protected hexasaccharide amide 19 (84% over two steps, Scheme 5).
Scheme 5
Synthesis of the title hexasaccharide.
To convert compound 19 into a conjugation-ready form, aglycone in the silyl ethyl glycoside had to be replaced with a suitably equipped linker molecule. Cleavage of 19 with trifluoroacetic acid (TFA), and subsequent conversion of the reducing sugar 20, thus obtained (Scheme 5), to the corresponding trichloroacetimidate donor 21 brought the synthesis of the title antigen to a synthetic step that proved difficult in the past. In most situations across carbohydrate chemistry, synthesis of the 1,2-trans-glycosidic linkage is not problematic because a participating group can be introduced into 2-positions of the glycosyl donors. However, in our case, the 2-O-position of the donor is glycosylated and, therefore, no participating group could be introduced. Furthermore, as we found during our previous syntheses of oligosaccharides within the Vibrio cholerae O1 series, the selectivity of α-mannosylation is impaired when the glycosylation is conducted with donors having the 3-deoxy-l-glycero-tetronic acid side chain already in place.[2,6,16] It is especially noteworthy that when we[7,14] previously glycosylated a fully assembled hexasaccharide Vibrio cholerae O1 donor with a primary hydroxyl group-containing linker molecule, the reaction showed almost no stereoselectivity. This was the reason why more recent syntheses of similar oligosaccharides utilized glycosyl donors containing 4-azido groups,[1,17] unlike in the earlier works where oligosaccharides related to the O-SP of Vibrio cholerae O1 were synthesized using glycosyl donors where the tetronamido side chain was already in place.[18,19] Similarly, in the initial reaction of 21 with 8-azido-3,6-dioxaoctanol (22) under Ogawa's[14] conditions, the α and β glycosides were formed in a ratio of ∼1.1 : 1. When we took advantage of the glycosylation under thermodynamic control developed in this laboratory,[2,6] the stereoselectivity of the same glycosylation increased by many folds (α : β ∼ 7 : 1). Conversion of oligosaccharides into conjugation-ready forms often involves α-mannosylation of very reactive aglycons. Such reactions are characterized by poor stereo selectivity. Performing such reactions under thermodynamic control remarkably increases the stereo selectivity of such conversions, and constitutes a considerable improvement of the synthesis of this and similar conjugation-ready oligosaccharides over existing protocols. Conventional deacetylation (Zemplén[12]) of 2a, followed by hydrogenation/hydrogenolysis (Pd/C) yielded the final glycoside 1 (71%, over two steps). The structure of compound 1 was confirmed by HRMS and NMR spectra.
Conclusions
We have modified existing approaches leading to antigenic determinants of Vibrio cholerae O1 and verified scalability of reactions involved. Intermediates to oligosaccharides in this series, up to and including the hexasaccharide have been successfully prepared on multigram scales. Using such hexasaccharide, the synthesis of the terminal hexasaccharide fragment of the O-specific polysaccharide of Vibrio cholerae O1, serotype Inaba was improved. The number of linear synthetic steps toward the compound was reduced, and the stereoselectivity of the critical 1,2-trans-glycosylation of the very reactive 8-azido-3,6-dioxaoctanol with the fully assembled hexasaccharide trichloroacetimidate was markedly increased, from α : β = 1.1 : 1 to 7 : 1 thereby increasing considerably the yield of the conjugation-ready title compound manifold.
Experimental
Materials and methods, crystallography
Unless specified otherwise, all reagents and solvents were purchased from Sigma-Aldrich Chemical Company and used as supplied. Reactions were monitored by thin-layer chromatography (TLC) on silica gel 60 glass slides. Spots were visualized by charring with H2SO4 in EtOH (5% v/v) and/or UV light. Melting points were determined with a Kofler hot stage. Optical rotations were measured at ambient temperature with a Jasco P-2000 digital polarimeter. NMR spectra were measured at 25 °C for solutions in benzene-d6, methanol-d4 or CDCl3, at 400 MHz, 500 MHz or 600 MHz for 1H, and at 100 MHz, 125 MHz or 150 MHz for 13C with Bruker Avance Spectrometers. Assignments of NMR signals were aided by 1D and 2D experiments (1H–1H homonuclear decoupling, APT, COSY, HSQC, TOCSY and HMBC) run with the software supplied with the spectrometer. Chemical shifts were referenced to that of tetramethylsilane (0 ppm) or signals of residual non-deuterated solvents. Crystals of 4 suitable for X-ray data collection were obtained by slow evaporation of ethanol from ethanolic solution. X-ray intensity measurements were collected at low temperature from a colorless needle-shaped crystal using a Bruker Kappa APEX II 4K CCD diffractometer with MoKα radiation. An Oxford Cryosystems 700 low temperature system was used to generate a stream of cold N2 gas that cooled the sample crystal to 120(2) K during data collection. Data were collected using both ω and ϕ scans with a scan width of 0.50° per frame and a rate of 30 s per frame, with the detector center located 40.0 mm from the crystal at 2θ = 30.00° or 60.00°. The data were processed using the Bruker APEX III software package. The crystal structure was solved and refined using the SHELXL[20] software package. The absolute configuration (Flack) parameter, determined from the X-ray data during the refinement, correctly identified the absolute configuration of the structure, which was also established by the known configuration of the α-d-mannopyranoside rings. The positions of H atoms attached to C and O atoms were calculated using idealized sp2 or sp3 geometry and included as riding atoms in the least-squares refinement. For methyl and –OH hydrogens, the torsion angles about the X-Me or C–OH bond were also optimized during the refinement. Details of the crystal data and structure refinement are given in ESI.† The 7 N solution of NH3 in MeOH was purchased from Sigma-Aldrich. Solutions in organic solvents were dried with anhydrous MgSO4 and concentrated at reduced pressure at <40 °C.
1-O-Acetyl-4-azido-2-O-benzoyl-3-O-benzyl-4,6-dideoxy-α-d-mannopyranose (11a) and 1-O-acetyl-4-azido-2-O-benzoyl-3-O-benzyl-4,6-dideoxy-β-d-mannopyranose (11b)
Attempted use of anomeric acetates 11a and 11b (α/β ∼ 9 : 1), prepared as described above, as glycosyl donors.When compounds 11a and 11b (0.25 g, 0.6 mmol) were treated with 2-(trimethylsilyl) ethanol (0.25 mL, 1.8 mmol) in presence of either boron trifluoride etherate (0.12 mL, 0.9 mmol) or TMSOTf (0.1 mL, 0.6 mmol) as described above for the synthesis of thioglycosides 7a and 7b, TLC showed that a complex mixture was formed where the product of hydrolysis of the anomeric OAc group largely predominated. Optimization of reaction conditions for this approach was not attempted.
To a solution of the linker-equipped hexasaccharide 2a (100 mg, 0.04 mmol) in dry MeOH (10 mL), methanolic NaOMe (1 M, 1 mL) was added with exclusion of moisture and atmospheric CO2, and the mixture was kept at room temperature for 6 hours, when TLC (3 : 2 toluene–acetone) showed that the starting material was consumed and that a much slower moving product was formed (Rf = 0.34 at 4 : 1 DCM–methanol with 1 drop of acetic acid). After neutralization (Dowex 50W H+ resin) and filtration, the solvent was removed, and the residue, showing correct HRMS (ESI-TOF: m/z [M + H+] calcd for C108H152O39N9 2199.0187; found 2199.0188), was used for the next step without further purification.To a solution of the above product in MeOH (2 mL), 10 mg of Pd–C was added, and the mixture was stirred at room temperature under 100 Psi pressure of hydrogen gas for 4 hours. TLC showed complete consumption of starting material and presence of a much more polar product. HRMS analysis confirmed the completion of global reduction. The mixture was filtered over a Celite pad, the solids were washed several times with methanol and the solvent was removed. A solution of the product in MeOH : H2O (2 : 1) was filtered through a 0.2 μm porosity syringe filter and lyophilized to collect pure product as white foam (42 mg, 71% over 2 steps), [α]D +1.8 (c 1.0, CH3OH). 1H NMR (600 MHz, CD3OD): δ 5.16–5.11 (m, 4H, 4× H-1II–V), 4.98 (s, 1H, H-1VI), 4.87 (s, 1H, H-1I), 4.22–4.14 (m, 6H, 6× H-2′), 4.13–4.08 (m, 4H, 4× H-2II–V), 4.05–3.98 (m, 5H, 4× H-3II–V, H-2VI at 4.02), 3.98–3.83 (m, 13H, 2× H-3I,VI, 6× H-4I–VI, 5× H-5II–VI), 3.83–3.78 (m, 3H, H-2I, H-5I, OCHa), 3.77–3.71 (m, 12H, 12× H-4′), 3.71–3.52 (m, 10H, 10× OCH2), 3.13 (t, 1H, J = 4.8 Hz, OCH2CHaNH2), 2.06–1.97 (m, 6H, 6× H-3′a), 1.87–1.78 (m, 6H, 6× H-3′b), 1.22–1.10 (m, 18H, 6× H-6). 13C{1H} NMR (150 MHz, CD3OD): δ 178.0 (5C, 5× NHCO), 177.9 (NHCO), 103.8 (C-1VI), 102.7, 102.5, 105.46, 102.4 (4C, 4× C-1II–V), 100.4 (C-1I), 79.6, 79.23, 79.2 (5C, 5× C-2), 71.6, 71.4, 71.30 (3C, 3× OCH2), 70.9 (C-2), 70.7 (6C, 6× C-2′), 70.0, 69.5, 69.4, 69.3 (11C, 6× C-3, 5× C-5), 68.7 (C-5), 68.1 (OCH2), 67.8 (OCH2), 59.4 (6C, 6× C-4′), 54.8–54.2 (6C, C-4I–VI), 40.7 (CH2NH2), 38.3 (6C, 6× C-3′), 18.4 (2C, CH3, 2× C-6), 18.3 (3C, CH3, 3× C-6), 18.2 (CH3, C-6). HRMS (ESI-TOF): m/z [M + H+] calcd for C66H118O39N7 1632.7465; found 1632.7467.
Authors: N Vijaya Ganesh; Joanna M Sadowska; Susmita Sarkar; Laurence Howells; John McGiven; David R Bundle Journal: J Am Chem Soc Date: 2014-10-09 Impact factor: 15.419
Authors: A Arencibia-Mohar; A Ariosa-Alvarez; O Madrazo-Alonso; E González Abreu; L García-Imia; G Sierra-González; V Verez-Bencomo Journal: Carbohydr Res Date: 1998-01 Impact factor: 2.104
Authors: Michael D Meeks; Rina Saksena; Xingquan Ma; Terri K Wade; Ronald K Taylor; Pavol Kovác; William F Wade Journal: Infect Immun Date: 2004-07 Impact factor: 3.441