Biomolecular Total syntheses of disulphated glycosphingolipid SB1a and the related monosulphated SM1a †

Total syntheses of two natural sulphoglycolipids, disulphated glycosphingolipid SB1a and the structurally related monosulphated SM1a, are described. They have common glycan sequences and ceramide moieties and are associated with human epithelial carcinomas. The syntheses featured e ﬃ cient glycan assembly and the glucosyl ceramide cassette as a versatile building block. The binding of the synthetic sulphoglycolipids by the carcinoma-speci ﬁ c monoclonal antibody AE3 was investigated using carbohydrate microarray technology.


Introduction
Sulphoglycolipids are acidic glycosphingolipids that contain one or two sulphate (HSO 3 − ) groups, and are components of cell membranes. 1 To date, several sulphoglycolipids have been isolated and characterised. 1,2 Interestingly, many of them have the same glycan sequence as ganglio-series 3 or globo-series gangliosides 4 but they contain sulphate groups instead of sialic acid residues. These sulphated glycolipids with a ganglio-series core framework are found in mammalian nervous tissue and in the kidney, where they may play a role in osmotic adaptation. 5 In addition, many of these sulphated glycolipids are known to be associated with various carcinomas. 6 SB1a, for instance, a di-sulphated analogue of the sialoglycolipid GD1a, has been found to be abundantly expressed on the surface of the human hepatocellular carcinoma cell line PLC/ PRF/15 but not in glycolipid extracts from a cirrhotic liver or normal liver. 6a Although association has been suggested between sulphoglycolipid accumulation and cancerous behaviours of epithelial cancer cells, 6b-d, 7 the precise roles of these sulphoglycolipids in carcinomas and other biological systems has not been clarified. To elucidate these roles, structurally homogeneous forms of sulphoglycolipids are in great demand. In the course of searching for carbohydrate sequences recognized by the epithelial cancer-specific monoclonal antibody (mAb) AE3, microarray analyses were performed with approximately 500 sequence-defined glycan probes. These revealed an unpredicted binding of the antibody to SM1a, 8 the monosulphated analogue of the sialoglycolipid GM1. The mAb AE3 was bound relatively weakly to the three structurally-related gangliosides asialo-GM1, GM1 and GM1 (Gc). There was no binding to the di-sulphated analogue SB1a. As the SM1a and SB1a included in the screening analyses were from natural sources, it was important to rule out the presence of any minor contaminants which could not be detected by mass spectrometry. The structures of SM1a (1) and SB1a (2), which were first isolated from rat kidney, 9 were assigned as shown in Fig. 1. 9,2a Characteristically, their ceramide compositions were mainly C24 non-hydroxylated fatty acids (tetracosanoic acid) and C18 4-hydroxysphinganine (D-ribo-phytosphingosine). Very recently, Li and co-workers reported the first total synthesis of SM1a; however, its ceramide part consisted of unnatural unsaturated sphingosine. 10 Previously, our group accomplished the syntheses of a large number of sialoglycans, including gangliosides 11 and their sulphated analogues, 12 using a chemical approach. Here we report the first total syntheses of the disulphated glycolipid SB1a and the monosulphated SM1a, each with a ceramide tail as found in natural products. The binding of the synthesized sulphoglycolipids by mAb AE3 was investigated using carbohydrate microarray technology. Results and discussion Fig. 2 presents the retrosynthetic analyses of the target compounds. Obviously, the systematic synthesis of SB1a and SM1a is possible because they have a gangliotetraose skeleton and a ceramide part. The key disconnection toward total synthesis is performed between the inner galactose and the reducing end glucose residues to produce key fragments, i.e. the non-reducing end trisaccharide unit and the glucosyl ceramide (GlcCer) cassette. Each C3 position of the galactose residues should be protected by the different selectively removable protecting groups because they are potential sites to introduce the sulphate group. To date, we have succeeded in the efficient total synthesis of natural gangliosides such as GQ1b, 13 GalNAc-GD1a, 14 GalNAc-GM1b, 15 X2, 16 LLG-3 17 and GAA-7, 18 using the GlcCer cassette approach. 19 This approach has been proven to be able to overcome a long-standing issue in glycolipid synthesis, namely that coupling of the flexible, poor reactive ceramide acceptor and the oligosaccharyl donor results in low yield and loss of the valuable oligosaccharide unit. Encouraged by previous successes, we decided to apply the GlcCer cassette approach to the present total syntheses. The present GlcCer cassette 4 was designed as shown in Fig. 2 and could be divided into the known glucose 7 and ceramide 8 derivatives. The common trisaccharyl donor 3 could be assembled from the known monosaccharide derivatives 5 and 6. First, following the above synthetic strategy, the common trisaccharyl donor 3 was prepared from the known galactose derivative 5 20 and N-Troc-protected galactosamine derivative 6. 21 As shown in Scheme 1, selective introduction of the benzyl group at the C3 position of 5 was performed in a one-pot fashion with two steps, tin acetal formation and subsequent benzylation, giving the 4-OH galactosyl acceptor 9 in 93% yield. The benzyl group was extremely stable under both acidic and basic reaction conditions, and it could be chemoselectively removed by hydrogenolysis prior to sulphonation.
Thereafter, the non-reducing end galactose unit was also prepared from 5. The 3-OH of 5 was selectively protected as a levulinoyl ester by treatment with levulinic acid, EDC·HCl and DMAP in CH 2 Cl 2 at −20°C, affording 10 in 88% yield. Regioselectivity was produced on the basis of the difference in reactivity between equatorial 3-OH and axial 4-OH. The selectively removable Lev group in the presence of other acyl protecting groups was installed for SB1a. After acetylation of 4-OH (giving 11), the p-methoxyphenyl (pMP) group at the anomeric position was removed with CAN and H 2 O in an optimised mixed solvent system 22 to give hemiacetal 12 in 81% yield. Finally, the introduction of the trichloroacetimidoyl group 23 yielded the nonreducing end galactosyl donor 13 in a nearly quantitative yield.
Scheme 2 illustrates the synthesis of the inner disaccharide GalNβ(1 → 4)Gal acceptor 16. The glycosylation of 9 with the known galactosaminyl donor 6 equipped with the 2,2,2-trichloroethoxycarbonyl (Troc) group at C2 as a β-stereo-directing element was performed in the presence of NIS and TfOH 24 in CH 2 Cl 2 at −40°C, giving disaccharide 14 as a single β-isomer in 76% yield. The reductive removal of the Troc group by treatment with zinc and acetic acid in MeCN gave the amino derivative 15 in good yield. MeCN was selected as a solvent to obtain the pure amino derivative before the following acetyl migration step because our previous study on the synthesis of a ganglioside found that MeCN effectively suppressed random acetyl group migration. 17 To rapidly convert 15 into the acceptor form 16, acetyl migration from 3-O to 2-N was exerted. Although this reaction was attempted several times under different conditions, the undesired formation of 17, which was the over-migrated product from 4-O to 3-O, could not be avoided. The results are summarised in Table 1. As shown in entry 1, the reaction was performed in 1,4-dioxane under acidic conditions at 60°C. The migration of the acetyl group was very sluggish and prolonged reaction time increased the generation of the over-migrated side-product 17. Moreover, 19% of 15 were recovered. In entry 2, to facilitate acetyl migration, the reaction temperature was increased up to 90°C. However, the reaction was an abysmal failure, providing a complex mixture after 16 h. Changing the solvent to DMF did not affect the outcome significantly (entry 3). However, in AcOH/DMF 1 : 4 at 90°C (entry 4), the reaction time decreased and the best result (56% of 16) was obtained. Although basic conditions using triethylamine were also examined, the yield of 16 was not improved (entry 5).
As shown in Scheme 3, access to the trisaccharyl donor 3 as a pivotal common unit began with glycosylation of 16 with 13 in the presence of TMSOTf in CH 2 Cl 2 at 0°C, giving trisaccharide 18 in 77% yield. The selective removal of the p-methoxyphenyl group with CAN and H 2 O followed by the introduction of the trichloroacetimidoyl group afforded the common trisaccharyl donor 3 in 80% yield over two steps.
Scheme 4 shows the assembly of the targeted tetrasaccharyl ceramide framework. First, the GlcCer cassette 4 was constructed by the coupling of the known glucosyl donor 7 16 and ceramide acceptor 8 25 mediated by a NIS-TfOH promoter system in CH 2 Cl 2 at 0°C to afford 20 in 71% yield. A previous study on the GlcCer cassette approach has shown the importance of the protection of hydroxyl groups in the Glc residue to increase the reactivity of the 4-OH of the GlcCer cassette. The best combination for protection was the presence of electrondonating groups on both O4 and O6 and an electron-withdrawing group on O2. [13][14][15] Furthermore, the TBDMS protection of the C4 hydroxy group on a Glc donor is also important for efficient coupling of glucose and ceramide. 17 Thus, we used the Glc donor 7 in this study. The removal of the TBDMS group in 20 by exposure to TBAF gave an 87% yield of the GlcCer cassette 4, which was the coupling partner for the trisaccharyl donor 3 in the final glycosylation.
The key glycosylation reaction performed between the equimolar amounts of trisaccharyl donor 3 and GlcCer cassette 4 was allowed to proceed in the presence of a catalytic amount of TMSOTf in CH 2 Cl 2 at 0°C, providing the tetrasaccharyl ceramide framework 21 in 72% yield. Following this, the p-methoxybenzyl groups in 21 were removed under acidic conditions (giving 22). Subsequent acetylation afforded 23, which was used for the following reaction sequences as the core framework of both target compounds. Scheme 3 Synthesis of common trisaccharide donor 3.
Scheme 4 Synthesis of GlcCer cassette 4 followed by glycosylation with 3.
The selective removal of the benzyl group in 23 by hydrogenation was examined. The results are summarised in Table 2. The Pearlman's catalyst (Pd(OH) 2 -C) was chosen as a Pd catalyst and the reaction was performed in EtOAc under ambient pressure of H 2 gas. Consequently, only a poor yield (5%) of the desired product 24 was obtained and most of the starting material 23 was recovered (entry 1). Although increasing the Pd catalyst amount gave a marginally better result (9%), the effect was very limited (entry 2). As entry 3 shows, the reaction solvent was changed to EtOH, which has been often used as a solvent for hydrogenolysis. After stirring for 72 h under ambient pressure, the reaction was subsequently subjected to a work-up process, providing 24 in 50% yield. Further prolonged stirring produced unwanted by-products, one of which was identified as a product in which the benzene ring of the benzoyl groups was reduced. The separation of the crude mixture of 24 and by-products by silica gel column chromatography was a laborious task. We speculated that steric hindrance surrounding the benzyl group made access of the benzyl group onto the surface of palladium carbon difficult.
Toward solving the problem, the reaction was carried out under a H 2 atmosphere at higher pressure (0.3 MPa) (entry 4). As anticipated, the reaction rate became faster and the yield of 24 was increased up to 77%. Although reduction of the benzene ring of the benzoyl group was observed, the limited reaction time (<24 h) minimised the formation of the byproduct. The hydroxyl group of the obtained 24 was available for sulphonation toward SM1a.
Scheme 5 shows the final steps to mono-sulphated glycosphingolipid SM1a. For sulphonation of the hydroxyl group at the C3 position of the inner galactose residue, 24 was reacted with SO 3 ·Pyr in pyridine at 80°C, yielding 25 in 91% yield. Finally, saponification of whole acyl protecting groups using 0.5 M aq. NaOH in MeOH was performed to furnish one of the target compounds SM1a (1) as an Na form in good yield.
SB1a was also derived from 24 via three steps (Scheme 6). First, the levulinoyl group was removed by treatment with NH 2 NH 2 ·AcOH in THF, giving diol 26 in excellent yield.
Sulphonation of two hydroxyl groups was performed using the same procedure as that of SM1a, affording the fully protected SB1a 27 in 96% yield. Finally, cleavage of acyl groups under saponification conditions successfully delivered the targeted SB1a (2) in excellent yield.
The synthetic sulphoglycolipids SM1a and SB1a were printed in a small focused microarray together with the naturally occurring SM1a. The array was then probed with mAb AE3 to compare the binding signals elicited (Fig. 3). The synthetic and the naturally occurring SM1a were equally well Scheme 5 Sulphonation followed by global deprotection toward SM1a.
Scheme 6 Conversion of 24 into SB1a through sulphonation followed by deprotection. bound by mAb AE3, whereas there were no binding signals with the synthetic disulphated glycolipid SB1a. This result corroborated the previous finding in the screening array analysis that AE3 binding was restricted to the monosulphated glycolipid SM1a with an unsubstituted terminal galactose residue. Sulphation of this galactose, as in SB1a, obviated the AE3 binding.

Conclusions
The total synthesis of disulphated glycosphingolipid SB1a was accomplished for the first time, as well as synthesis of the related monosulphated SM1a with a natural ceramide moiety. These glycolipids were systematically synthesised by efficient glycan assembly using common building blocks and the GlcCer cassette approach. The success of the total synthesis of sulphoglycolipids demonstrates the broad utility of the GlcCer cassette approach for glycolipid synthesis. Furthermore, binding analysis using the two synthetic sulphoglycolipids corroborated the previous observations on the recognition of the SM1a glycolipid by mAb AE3, although the antibody was originally raised against a cancer-associated mucin-type glycoprotein. These findings raise the possibility that sulphated O-glycan analogs of SM1a occur on cancer-associated mucins.