Automated glycan assembly of Lewis type I and II oligosaccharide antigens

Lewis antigens are fucosylated oligosaccharides that play crucial roles in various biological processes. Here, we illustrate how automated glycan assembly (AGA) provides quick access to a series of more than ten defined Lewis type-I and type-II antigens.

While the importance of blood group antigens for blood transfusions is established, their involvement in infectious diseases and cancer development is still emerging. Lewis b (Le b ) 3 expressed on gastric epithelium is the receptor for Helicobacter pylori, the cause chronic gastritis and peptic ulcers. 6 Type-II chain analogue Lewis y (Le y ) 7 is a tumor-associated carbohydrate antigen (TACA), that is overexpressed on the cell surface of several types of cancer. [12][13][14] Le b and Le y vary minimally in the regiochemistry of the glycosidic linkage in the terminal Nacetyllactosamine subunit of the Lc 4 (4)/nLc 4 (8) core (b(1-3) vs. b (1)(2)(3)(4), respectively). Extended chain versions of Lewis antigens such as Le x dimer (Le x -Le x , 9) and KH-1 (Le x -Le y , 10) are attractive targets for tumor immunotherapy (Fig. 1B). These TACAs are overexpressed in colorectal cancer and overcome the low immunogenicity observed in human trials when using shorter antigens such as Le y . [15][16][17][18][19][20] Synthetic access to oligosaccharide antigens is essential as isolation of useful amounts of pure glycans from biological sources is difficult. 21,22 Numerous syntheses of Lewis antigens that serve as tools e.g. for immunotherapy 20,23,24 have been reported, including solution-phase, solid-phase and automated syntheses. 19,[25][26][27][28] Despite the structural similarity of Lewis antigens, typically total syntheses produce single structures. [29][30][31] Here, we present a general method to assemble the entire class of Lewis antigens. The lacto-and neolacto-series target molecules differ in three structural aspects: the presence or absence of fucose on the terminal galactose; the presence or absence of fucose on GlcNAc, and the b(1-3) or b(1-4) linkage in the lactosamine subunit. The logic of automated glycan assembly (AGA) is based on the selection of a minimum set of monosaccharide building blocks to assemble all targeted glycans via a linear glycosylation and deprotection sequence. 32,33 AGA reduces the synthesis time considerably 34,35 and facilitates the assembly of large, complex antigens such as 9 or 10. 36 Access to lacto-and neolacto-series variants that carry a linker for conjugation allows them to be used as tools for diverse biological applications, in particular those where minor structural differences between type-I-and type-II-chain antigens, 37 are concerned.

Results and discussion
"Approved building blocks", referring to monomers that can be prepared in large amounts, are shelf stable and upon activation result in reliable and selective glycosidic bond formation, are key to AGA. 35 The ve monosaccharide building blocks 11-15, which suffice to access all Le type-II blood group related antigen targets (Fig. 2), are either commercially available or were synthesized in multi-gram scale in a few steps starting from commercial intermediates (see ESI † for details). Hydroxyl groups of building blocks 11-14 that engage in chain elongation are temporarily protected as 9-uorenylmethoxycarbonyl (Fmoc) carbonates that are stable under the acidic glycosylation conditions, but can be easily removed under mild basic conditions. Levulinoyl ester (Lev) marks the C3-hydroxyl in glucosamine (GlcN) 13 as orthogonal temporary protecting group as illustrated for AGA before. 38,39 Benzyl ether (Bn) and benzoyl esters (Bz), when C2 participation is needed, are used as permanent protecting groups. The amine of GlcN 13 was protected as an N-trichloroacetyl (TCA) group to ensure b-selectivity. The permanent protecting groups should be removed aer AGA and light-induced cleavage of the solid support by methanolysis (Bz) and hydrogenolysis (TCA and Bn). Ready synthetic access and high stability, prompted us to use thioglycosides as anomeric leaving groups of all building blocks except galactose 14. The C2-OFmoc in 14 resulted in low stereoselectivity for the thioglycoside, while the dibutyl phosphate building block ensured excellent stereoselectivity (see ESI †). The selectivity differences may be a result of a change in solvent used during the glycosylation, since dioxane is added to ensure solubility of NIS/TfOH required for thioglycoside activation. Dioxane coordinates the b-face of the oxocarbenium ion that forms during glycosylation and favors the formation of the aglycosylation product. 40 Dibutyl phosphates can be activated by TMSOTf in DCM, hence solvent coordination does not inuence the reaction.
Before synthesizing the target molecules, the ve building blocks 11-15 were tested for their performance in AGA by establishing optimal glycosylation conditions in the context of  dimer or trimer syntheses ( Table 1). All AGA syntheses were executed using an oligosaccharide synthesizer and Merrield resin equipped with photolabile linker 16 as solid support. 26,39,41,42 Aer each glycosylation, the temporary protecting group was removed to allow for further chain elongation. In the case of Fmoc, 10 min incubation with 20% piperidine in DMF was sufficient; Lev was selectively cleaved by treatment with a hydrazine acetate solution (0.15 M), liberating a hydroxyl group for the subsequent glycosylation step. Aer complete assembly, the desired oligosaccharide was cleaved from the solid support in a continuous ow photoreactor and analyzed by HPLC. 39 Full conversion and excellent stereoselectivity were achieved for thioglycosides 11 and 12 employing eight equivalents of building block and a glycosylation time of 20 min at 0 C, aer a short incubation (incubation time, t 1 ) at lower temperatures (Table 1). Glycosylation reagents are added dropwise to the reaction vessel at a controlled incubation temperature (T 1 ). T 1 is typically 20 C below the glycosylation temperature (T 2 ), to minimize reactivity before the reagent delivery process is completed. Aerwards the reaction vessel is warmed up to T 2 to perform the coupling. Highly reactive perbenzylated fucose 15 was activated at À20 C to avoid hydrolysis. Glucosamine 13 proved less reactive than the other building blocks and required 40 min glycosylation time to achieve full conversion. Dibutyl phosphate 14 was coupled at À15 C for 30 min using ve equivalents of building block.
Linear hexasaccharide nLc 6 17, the unbranched backbone of KH-1 and Le x -dimer, served as a rst test for the optimized conditions. HPLC analysis of the crude products aer AGA and light-induced cleavage showed that just one product was formed (ESI †) before protected nLc 6 was isolated in 55% yield by preparative HPLC (Fig. 3A).
Next, two branching strategies for Le type-II antigens Le y , Le x , Le x -dimer and KH-1 were evaluated. Aer introduction of building block 13 into an oligosaccharide, the Fmoc group can be removed and a galactose can be attached at the C4 hydroxyl group of GlcN before the Lev at the C3 hydroxyl is cleaved and a fucose is attached; subsequently chain elongation continues ('strategy 1'). Alternatively, this process can be inversed ('strategy 2', Fig. 3B). Both branching strategies were tested  using 18 as a model to ensure that all glycosylation cycles result in full conversion and excellent stereoselectivity. Aer lightinduced cleavage and preparative HPLC, 18 was isolated in 39% yield using strategy 1. Strategy 2 resulted in an improved result as judged by analytical HPLC and 51% isolated yield. With a set of building blocks in hand, a series of protected Lewis type-II-chain blood group related antigens 19-26 (28-65% yield) was synthesized via strategy 2 with excellent stereoselectivity (Fig. 4). The syntheses of 19-24 yielded no signicant amounts of deletion sequences. The AGA protocol was adjusted slightly for the extended chain antigens 25 and 26 to achieve full conversion in every glycosylation cycle. Two glycosylation cycles were performed when introducing the second GlcN building block since the GlcN building block is less reactive and the branched acceptor is sterically more hindered (Fig. 4). Antigens 17-26 were assembled using ve building blocks. Le x and Hantigen were assembled with the initial lactose unit (21 and 22) or without (19 and 20) to provide access to all variants of these structures for biological studies. 43 Lewis type-chain ceramides on the surface of human cells carry the initial lactose unit, 44 while many biological studies only consider the terminal fucosylated epitope. 45,46 Streamlined coupling cycles rendered the assembly of KH-1 nonasaccharide 26 (15 h) signicantly faster than an earlier AGA version that required 23 h. 25 Aer preparing all Lewis type-II-chain antigens, we set out to assemble the Lewis type-I-chain antigens Lewis a (Le a , 2) and Le b 3. Both contain a b(1-3) instead of a b(1-4) linkage between GlcN and the terminal Gal, with the C4 GlcN fucosylated (Fig. 1). Since type-I-and type-II-chain antigens differ only in the substituents attached at the non-reducing end of the GlcN unit, AGA of type-I-chain structures can rely on the sequential cycles developed for type-II chains, as simply the order of Fmoc and Lev cleavage in the GlcN unit has to be inverted. Protected Lc 4 tetrasaccharide 27 was obtained in a yield comparable to that for type-II-chain analogue nLc 4 (Fig. 5). However, the approach failed to afford branched structures since under these conditions, aer C4 fucosylation of GlcN, Lev was not properly cleaved from the C3 hydroxyl. Therefore, assembly strategies where galactosylation of GlcN precedes Fmoc deprotection and fucosylation were used for the AGA of Le b 28 and Le a 29 (Fig. 5).
Replacement of 14 by 30 that only bears permanent protecting groups allowed for the assembly of Le a 29. A difucosylation strategy, where the Fmoc groups at C2 of galactose 14 and C4 of the GlcN 13 were removed simultaneously furnished 28 (Fig. 5).
Following AGA and photo-induced cleavage from the solid support, all permanent protecting groups have to be removed, to furnish the desired antigens. Global deprotection relied on two steps. Benzoyl groups were cleaved with sodium methoxide in MeOH/DCM as reaction progress was monitored by MALDI. In the second step, benzyl ethers, TCA and the carboxybenzyl group at the amino linker were removed using hydrogenolysis catalyzed by Pd/C to afford the deprotected, N-acetylated oligosaccharides. Reversed phase HPLC yielded a series of fully deprotected Lewis antigens 31-38 that carry a C5-amino linker at their reducing end for immobilization on glycan array surfaces or conjugation to carrier proteins (Fig. 6). 47,48 Rather low isolated yields over the two deprotection steps (17-54%) may be a result of the poor solubility of partially and fully deprotected glycans during and aer deprotection. 49 In addition, partial cleavage of fucose and TCA was observed. For the extended chain antigens 25 and 26 the combination of methanolysis and hydrogenolysis did not lead to the desired products Le x dimer (39) and KH-1 (40). Severe solubility issues for the partially-deptrotected oligosaccharides formed during the deprotection process prevented complete hydrogenolysis. Birch reduction followed by peracetylation to facilitate purication was the endgame during the total synthesis of KH-1. 30 This procedure was not applicable as it would result in the irreversible acetylation of the C5-amino linker. Finally, compounds 25 and 26 were deprotected using sodium in lique-ed ammonia gas followed by methanolysis. The nal oligosaccharides were puried using a Sephadex-G25 column to yield 39 (13%) and 40 (19%). The bottleneck of chemical glycan synthesis is the removal of protective groups from the nal molecules and is common to AGA and other modes of glycan construction. Alternatives such as a combination of AGA with enzymatic glycosylations may circumvent deprotection issues but these methods have not yet been established for complex, branched oligosaccharides such as KH-1. 50,51 Conclusions A set of Lewis type-I and type-II-chain antigens was synthesized via automated glycan assembly in a fast, reliable and reproducible fashion. Five building blocks were sufficient to synthesize all Lewis type-II antigens including H-antigen II, Le x , Le y and KH-1 and the Lewis type-I-chain-related structures Lc 4 and Le b . For the assembly of Le a an additional building block was necessary, since the temporary Lev protecting group was difficult to cleave. Optimized glycosylation cycles led to full conversion and excellent stereoselectivity during AGA and minimized the formation of side products. Global deprotection of the target molecules provided sufficient quantities of the target glycans for biological studies.

Conflicts of interest
There are no conicts to declare.