Regioselective Glycosylation Strategies for the Synthesis of Group Ia and Ib Streptococcus Related Glycans Enable Elucidating Unique Conformations of the Capsular Polysaccharides

Abstract Group B Streptococcus serotypes Ia and Ib capsular polysaccharides are key targets for vaccine development. In spite of their immunospecifity these polysaccharides share high structural similarity. Both are composed of the same monosaccharide residues and differ only in the connection of the Neu5Acα2‐3Gal side chain to the GlcNAc unit, which is a β1‐4 linkage in serotype Ia and a β1‐3 linkage in serotype Ib. The development of efficient regioselective routes for GlcNAcβ1‐3[Glcβ1‐4]Gal synthons is described, which give access to different group B Streptococcus (GBS) Ia and Ib repeating unit frameshifts. These glycans were used to probe the conformation and molecular dynamics of the two polysaccharides, highlighting the different presentation of the protruding Neu5Acα2‐3Gal moieties on the polysaccharide backbones and a higher flexibility of Ib polymer relative to Ia, which can impact epitope exposure.


Introduction
Group BS treptococcus (GBS) is al eading cause of pneumonia, sepsis,m eningitis, and death in neonates. [1] It hasa lso been associated with high rates of invasived iseases in the elderly. [1] On the basis of variation in polysaccharide composition, the GBS sialic acid-rich capsular polysaccharides (CPSs) are divided into ten serotypes (Ia, Ib, andI I-IX). [2] GBS CPSs are key virulence factors and consideredt he prime vaccine candidate to combat GBS infections. [3] Monovalent conjugatev accinesp repared with GBS type-specific polysaccharides representing the most frequent disease-causings erotypes (Ia, Ib, II, III, and V), as well as at rivalent combination (Ia, Ib, III), have been tested in phase I/II clinical trials [4] with the ultimate goal of developing a maternal vaccination strategy. [1,5] Multivalent formulations with six different serotypes are currently under clinical testing. [6] GBS serotypes Ia, Ib, and III account for the majority of GBS related diseases. [7] CPS Ia and Ib are structurally very similar. Both are composed of the same monosaccharide residues andd iffer only in the linkage between the Neu5Aca2-3Gal side chain and the GlcNAc unit:ab1-4 linkagei nt ype Ia and a b1-3 linkage in type Ib. [8] This difference is criticali nd eterminingt he immunospecificity ( Figure 1). [3,9] The repeating units of CPS Ia and Ib can be described by the branched 1, 3 andl inear 2, 4 frameshifts depicted in Figure 1. Intriguingly,t he latter pentasaccharides 3 and 4 share identicalm onosaccharide composition with milk oligosaccharides, which have recently been proposed to inhibit GBS colonization. [10] The availability of well-definedG BS CPS glycansi s key to explore interactions with serotype-specific monoclonal antibodies in order to identify relevant glycoepitopes for elucidating the mechanism of action of the polysaccharide conjugates and for the development of synthetic carbohydratebased vaccines. [11] The most studied of GBS polysaccharides is type III. This CPS is knownt of orm ah elical structure, [12] and this feature has an impact on epitope exposure. [13] Our group has recently synthesized CPSIII oligosaccharides that were used along with fragments obtained from CPS depolymerization to map as ialylated structural epitope spanning two repeating units. [14] Considering that neither chemical nor enzymatic depolymerization reactions are available for CPS Ia and Ib, chemical synthesis is the only approach to obtain homogeneous oligosaccharides from the CPS. Although the synthesis of the GBS CPS Ia repeating unit has been reported, [15] the preparationo ft he pentasaccharider epeating unit of GBS CPS Ib has not been achieved. When approaching the synthesis of CPS Ia and Ib fragments, we envisaged the formation of the disaccharideG lcNAcb1-3Gal motif as akey step to enableconvergent syntheses of avariety of structures depicted in Figure 1. Typically,i nstallation of the GlcNAcb1-3Gal disaccharide within more complex glycans has been achieved with the 4-hydroxyl group of the Gal acceptor either protected [16,17] or already engaged in ag lycosidic linkage. [18] Particularly,i nt he preparation of CPS Ia repeating units [15] a4 ,6-O-benzylidene-protected Gal acceptor was used for glycosylationw ith ag lucosamine trichloroacetimidate donor,a nd subsequent regioselective ring opening before further glycosylation of position 4f or the construction of the trisaccharide GlcNAcb1-3[Glcb1-4]Gal could take place. Therei s need of expeditiousp rocedures for the construction of complex glycans, and regio-a nd stereoselective reactions distinguishing among diverse deprotected hydroxyls are highly desirable to simplify the oligosaccharide assembly. [19] We reasoned that regioselective glycosylation of Gal 3-OH would be the key for accelerating the synthesis of the GlcNAcb1-3Gal disaccharide and rendering the 4-OH availablef or further glycosylation without the need of tedious protection/deprotection sequences. [20] Herein, we report tactics to achiever egioselective syntheses of protected GlcNAcb1-3Gal buildingb locks and the use of these key synthons in convergent routes towards as eries of fragments from CPS Ia and Ib repeating units with ab uilt-in aminopropyl linker amenablef or future conjugation to carrier proteins ( Figure 1). Furthermore, combination of NMR data from the synthetic GBS CPS Ia and Ib repeating units in their branchedf orm 1 and 3,r espectively,a nd moleculard ynamics simulation allowed to shed light on how the variation of a single sugar connection dramatically affects the conformational properties of CPS Ia and Ib polysaccharides,a nd hence exposition of potential epitopes for antibodyr ecognition.

Results and Discussion
Optimization of regioselective glycosylation of galactose According to our retrosynthetic design (Figure 1), the target glycans 1-4 can be obtained througha[2+ +3] convergent strategyb ased on the glycosylation of as uitable trisaccharide acceptorw ith aN eu5Aca2-3Gal donor.T hisa pproach envisages the challenging stereoselective a sialylationo ft he upstream galactose at an early stage of the synthesis. [21] Alternative use of aG al donor would enablet he synthesis of 5.I nt his design, faster and efficient access to aG lcNAcb1-3Gal disaccharide buildingb lock plays ac entral role to obtain the trisaccharide acceptor withoutatemporary protection at position 4f or further assembly of GBS CPS Ia fragments. To achieve its regioselective synthesis, we investigated the effect of arming benzyl and disarming benzoyl groups [22,23] at position 2a nd 6o ft he Gal acceptors in tuning the reactivity of the 3-and 4-OH, respectively,i nc ombination with variousp rotecting and leaving groups in the glucosamine donors. Despite the expected higher reactivityo ft he equatorial 3-OH versust he axial 4-OH, regioselective glycosylation of position 3h as been shown not to be trivial. [16] Accordingly,w es ynthesized as eries of glucosamine thioglycoside and trichloroacetimidate donors with the amine protectedb yt he participating phthalimido (Phth) or trichloroethyl carbamate (Troc) group (experimentalp rocedures are provided as Supporting Information).
Levulinoyl (Lev) and fluorenylmethyloxycarbonyl (Fmoc) were selected for temporary protection of either position 3o r 4. Alternatively,a4,6-O-benzylidene was used to lock the 4and 6-hydroxyls to be subjected to regioselective ring opening delivering the 4-OH at al ater stage of the synthesis (Scheme 1). The prepared donors and acceptors were then coupled under severalg lycosylation conditions (Table 1a nd Scheme1)t oo ptimize the synthesis of the GlcNAcb1-3Gal buildingb lock. The most efficient routes provedt ob et he combination of the 2,6-di-O-benzoyl acceptor 11 with both donor 6 or 7 under N-iodosuccinimide (NIS)/AgOTf-mediated activation ( Table 1, entries 7a nd 8), which gave 14 a and 15 a in yields of 53 and 65 %, respectively,o rt he imidate 9 and acceptor 11 (Table1,e ntry 9), which enabled the attainment of 14 a in 77 %y ield.
Similarly,c onditions fort he preparation of aG lcNAcb1-3Gal synthon with at emporary group at its C3'-OH, to allow the ensuing assembly of GBS CPSIb fragments, weree xplored (Table 2a nd Scheme 1). The glycosylation of di-O-benzyl acceptor 10 with donor 16 by using NIS with either TfOH or AgOTf as co-promotersg ave variable mixtures of the b1-3 21 a and b1-4 21 b disaccharides (Table 2, entries 1a nd 2). Again, the di-O-benzoyla cceptor 11 in the presence of NIS/AgOTf activation at À30 8Ca llowed achieving ay ield of 68 % ( Table 2, entry 4), which confirms the improved capacity of the benzoyl substituents to govern the regioselectivity of the reaction compared with benzyls ubstituents. These conditions werea lso efficient for the GlcNTroc donor 17,w hich gave 23 a in 65 % yield (Table 2, entry 6). When the trichloroacetimidate 18 was used, the yield was increased up to 70 %( Ta ble 2, entry 7), which corroborates the potential of this type of donor for the regioselective control of the reaction. Finally,t rifluoroacetimidate glucosamine 20 bearinga4,6-O-silylidene protection in the presence of TMSOTf as promoter afforded the target disac-charide 25 a in 62 %y ield. [24] The slighty higher flexibility or lower hindering effect of the silylidene relative to that of the benzylidene group favored the reaction. Overall,t heser esults indicatet hat the regioselectivity of the glycosylation benefits from the decreased nucleophilicity of the axial 4-hydroxyl, which is intrinsically less reactivethan the 3-hydroxyl group, induced by the electron-withdrawing effect of the 2,6-O-benzoyl as comparedw ith 2,6-di-O-benzyl substituents in the Gal acceptor.
In addition, mild activation conditions (NIS/AgOTf)f or the thioglycoside donor or the torsional disarming effect of the benzylidene/silylidene group for the imidate donors appearsto favor the regioselectivity of glycosylation at position3 .

Synthesis of GBS CPS Ia linearand branched repeatingu nits
Having identified the two glycosylation partnersg iving the GlcNAcb1-3Gal motif in ar egioselective fashion, we elongated the disaccharide buildingb lock to assemble the pentasaccharide repeating unit of GBS CPS Ia. To this end, reactions of glucose donor 26 [25] with disaccharide donors 12 a and 14 a were performed to furnish trisaccharides 27 a and 27 b in 75 and 68 %y ield, respectively (Scheme 2). The newly formed glycosidic bond was in b configuration, as expected by the presence of ap articipatingg roup.
Despite the deactivating effect of the 6-O-benzoyl ester relative to that of the 6-O-benzyl ether,t he reactionp roceeded with almosti dentical efficiency (Scheme 2), whereas ap eracetylatedt richloroacetimidateglucose donor with TMSOTf activation was ineffective for glycosylation of the 4-OH. Considering the higher regioselectivity and yield achieved in synthesizing disaccharide 14 a,t he resulting trisaccharide 27 b was advancedi nt he GBS CPS Ia repeating unit construction and subjected to regioselective openingo ft he 4,6-O-benzylidene acetal with BF 3 ·Et 2 Oa nd Me 3 N·BH 3 to provide the acceptor 28 (70 %).
In order to complete the pentasaccharidec onstruction, the sialo-galactosyl trifluoroacetimidate donor 29 [14a, 26] andt hioglycoside 30 [27] weret ested. Of these two disaccharides, 30 can be prepared with ah igher a stereoselectivity,w hereas 29 is easily accessible from ac ommerciald isaccharidep recursor. [14a] Glycosylation of trisaccharide 28 with 29 under TMSOTf activation gave the protected pentasaccharide 31 in 75 %y ield, and the use of disaccharide 30 in the presence of NIS/TfOH led to the protected pentasaccharide 32 in as imilar yield (73 %). Compound 30 was deprotectedb yafour-step procedure, [18] including 1) saponificationo ft he methyl ester of Neu5Acw ith lithium iodide in pyridine;2 )reaction with ethylenediamine in ethanol heatedt or eflux for concomitant removal of the O-acetates and the NPhth protecting group;3 )reacetylation with acetic anhydride/pyridine to install the acetamide group of the GlcNAcr esidue along with acetyl esters;4 )methanolysis and final catalytic hydrogenation over Pd/charcoal to provide the targetb ranched pentasaccharide 1.
Pentasaccharide 32 was first subjected to saponification with NaOH in THF heated to reflux, followed by amine reacetylation with a2:3 acetic anhydride/methanol mixture. [a] CH 2 Cl 2 was the solvent in all tested conditions. Hydrogenation over Pd/charcoal afforded the target branched pentasaccharide 1 equipped with the aminopropyl linker suitable for conjugation. After purification by size exclusion chromatography,t he final compound was obtained in 40 %o verall yield from 31 and 45 %o verall yield from 32,r espectively (Scheme 2).
Next, we extended the same regioselective approach to the synthesis of the linear frameshift 2 of the serotype Ia repeating . In this case, the benzoylated lactose 33 and the glucosamine donor 8 were chosen as glycosylation partners affordingt he linear trisaccharide acceptor 34 in 68 %y ield with complete regioselectivity.F ollowing benzylidene opening, the trisaccharide acceptor 35 wasg lycosylated with the two donors 29 and 30.T he first glycosylation promoted by TMSOTf at 0 8Ca fforded the target linear pentasaccharide 36 in 65 % yield, with b stereo-and regioselectivity at C-4 of GlcNAc over the C-4 of Gal. The presence of the free galactose 4-OH throughout all stages of the synthesis, from trisaccharide 34 to pentasaccharide 36,w as monitored by followingt he signal of the Gal H-4, which appeared at 3.97 ppm (d, J = 2.7 Hz) in the 1 HNMR and HSQC spectra of all synthetic intermediates.T his confirmedt he regioselectivity of the two glycosylations performed. Unexpectedly,r eaction of 35 with the tolyl thioglycoside 30 under NIS/TfOH activation at À40 8Cy ieldedo nly traces of the corresponding pentasaccharide, whereas mainly decomposition of the glycosyl donor was observed, asr evealed by LC-MS analysis. The linear pentasaccharide 36 was subjected to the five-step deprotectionp rotocol previously described for compound 31.T he target oligosaccharide 2 was purified by size exclusion chromatography and obtainedi n 33 %o verall yield (Scheme 3). NMR data of the synthesized fragments were in excellent agreement with those of the CPS Ia samples. [8] From acceptor 35 ad esialylated CPS Ia linear fragment for future mapping studies wasa lso obtained by glycosylation (72 %y ield) with the trifluoroacetimidate 38,p repared from the known 1-OH compound 37. [18] After global deprotection tetrasaccharide 5 was obtained in 42 %yield (Scheme 3).

Synthesis of GBS CPS Ib linear and branched repeating unit
Differently than the GBS CPS Ia pentasaccharides, the two Ib frameshifts 3 and 4 required ag lucosamine buildingb lock bearingatemporary protecting group at its C3-OH and the creation of the Galb1-3GlcNAc linkage,w hich had as trong impact on ours ynthetic design. Initiala ttempts to preparet he branchedp entasaccharide 3 by using aN Phth-protected trisaccharidea cceptor,s imilarly as done for the CPS Ia branched unit, were unsuccessful ( Supporting Information, SchemeS9).
The C3-OH of the glucosamine appeared significantly less reactive than the C4-OH, which is likely due to the presence of the bulky NPhth group that could hindert he glycosylationr eaction at the C3ÀOH. We anticipated that its replacement with aT roc protection would result in ah ighern ucleophilicityo f the vicinal hydroxyl. Disaccharides 23 a and 25 a,w hich differ only in the cyclic protecting group blockingt he glucosamine C4,6ÀOH groups,w ere selected to be elongated to the branched pentasaccharide 3 (Scheme 4). Glycosylation of the two acceptors with the armed Glc donor 26 under TMSOTf activation at 0 8Ca fforded the trisaccharides 40 and 41 in 63 and 70 %y ield, respectively,a sb anomers. After Fmoc removal by treatment with 10 %p iperidine in CH 2 Cl 2 (92 %), glycosylation with the sialogalactoside donor 29 of the two acceptors 42 and 43 was tested.
Reaction of the 4,6-O-benzylidene trisaccharide 42 and 29 with TMSOTf as ap romoter failed to afford the target pentasaccharide, leading to complete recovery of the unreacted acceptor.I nc ontrast, reaction of acceptor 43,b earing the more flexible 4,6-O-silylidene ketal, with 29 in the presence of TMSOTf gave the target pentasaccharide 44 in 80 %y ield (Scheme 4). This result suggests that the glycosylation of 42 was prevented by the steric and torsional constrain of the 4,6-O-benzylidene ring. Trisaccharide 43 was also efficiently b-glycosylated with disaccharide donor 30 by NIS/TfOH activation, affording 45 in 65 %y ield (Scheme 4). Despite as lightly lower yield in this step, the overall efficiency of the synthesis of the GBS serotype Ib branched repeating unit was superior when using the thioglycoside 30 with respect to the imidate 29 because of the better a stereoselectivity of the glycosylation leadingt o30. [26,27] Pentasaccharides 44 and 45 were then deprotected by af our-step protocol (Scheme4): 1) desilylation by treatment with HF·pyridine, 2) saponification with NaOH in THF heated to reflux, for concomitanth ydrolysis of the acyl esters, the Troc group, and the 5-N,4-O-oxazolidinone protecting group and Neu5Acm ethyl ester,3 )reacetylation of the amines by a2 :3 acetic anhydride/methanol mixture, 4) hydrogenationo verP d/charcoal.T he target branched pentasaccharide 3 was obtained in 40 %y ield.

Conformational analysis
The conformationalp roperties of the CPS Ia and Ib branched repeating unit pentasaccharides 1 and 3 were studied by a combination of NMR spectroscopy and modeling tools, [28] and compared with those of the corresponding polysaccharides. Interglycosidic interproton distances for 1 and 3 were estimated from ROESY spectra. The obtained experimentald istances were comparedw ith those derived from a2 00 ns molecular dynamics (MD) simulation. Ta ble 3g athers the results for the CPS Ia pentasaccharide 1.T he comparison reflectsagood agreement between the NMR-a nd the MD-derived distances for the glycosidicl inkages GlcNAcb1-3Gal and Glcb1-4Gal (defined by the interprotond istances H1GlcNAc-H3Gal and H1Glc-H4Gal, respectively). The F/Y population analysisf rom the MD simulation showedasingle population for F fulfillingt he exoanomeric effect (exo-syn-F), [29,30] and two populations around y for both linkages (Supporting Information, Figure S1).
For the Neu5Aca2-3Gal linkage, MD simulations predict three different populations, 1808/À308, À608/À208,a nd À608/ À508.T he interglycosidic interproton distances for each population are gathered in Table 3. There is ar emarkable difference for the H3Gal-H3axNeuNAc distance, being shorter according to NMR spectroscopy, which indicates that the MD simulation has ab ias for the conformational ensemble towards exo-syn-F populations. Indeed, according the NOE-derivedd istance, the exo-anti-F population should be the major one, representing around7 5% of the total ensemble. Twor epresentativec onformations for the CPS Ia pentasaccharide 1 are shown in the Supporting Information, differing in the Neu5Aca2-3Gal linkage ( Figure S3). The analysis for the CPS Ib pentasaccharide 3 yielded similar results( Supporting Information, Ta bleS3), althought he GlcNAcb1-3Gal linkagec ould not be fully characterized because of the overlapping between the H1GlcNAc and H3Gal protons. The linkageb etween Gal and GlcNAc, now b1-3 insteado fb1-4, populates am inimum aroundt he exosyn-F/syn(À)-Y conformation (Supporting Information , Figure S4). Twor epresentative conformationsf or the CPS Ib pentasaccharide 3 are shown in the SupportingI nformation, which also differ in the orientation around the Neu5Aca2-3Gal F torsion (SupportingI nformation, Figure S5). As uperimposition of representative 3D structures for the CPS Ia and Ib pentasaccharides 1 and 3,w ith the major conformation exo-anti-F aroundt he Neu5Aca2-3Gal linkage is shown in Figure 2. The conformational behavior of the polysaccharides was then analyzed following as imilarp rotocol. Am odel for the polysaccharide was built with 10 repeating units (50 monosaccharides) and MD simulations were runfor 2.5 ms.
The analysis of the glycosidic linkages wasc arriedo ut for the 49 glycosidicb onds, revealing that the behavior for every glycosidicb ondt ype is reproducible along the polysaccharide ( Figure 3).
These populations are comparable to those of the corresponding pentasaccharide fore very glycosidicl inkage, and thus, the resulting interglycosidic interproton distances are very similar (Supporting Information, Table S1-2). Remarkably, for both GBSs erotypeI aa nd Ib, the HSQC spectra of the polysaccharide and the pentasaccharide were very similar,w ith the obvious exceptionf or the Glcb1-4 linked moiety (E), which is not glycosylateda tO 4i nt he pentasaccharides (Figure 4). The analysiso ft he interglycosidic NOE (from NOESY spectra at 20 ms mixingt ime) was consistent with the MD-derived populations.T he only discrepanciesa rose again for the Neu5Aca2-3Gal linkages. Interestingly,f or the Ia polysaccharide the NOEderived distance for H3axNeu5Ac-H3Gal is 2.4 ,s horter than that in the pentasaccharide. At the same time, there is ac lear NOE between H3eqNeu5Ac-H3Gal, not observed for the pentasaccharide. On the contrary,f or the Ib polysaccharide the distance H3axNeu5Ac-H3Gal is longer, 3.3 ,w hereas the NOE between H3eqNeu5Ac-H3Gal does not exist ( Figure 5A,B). At the same time, the distance H8Neu5Ac-H3Gal is slightly shorter for the Ib than that for the Ia polysaccharide ( Figure 5C,D). These data suggest that for the Ia polysaccharide, the major conformation around the Neu5Aca2-3Gal fragment is the exo-anti-F (ca. 85 %), whereas for the Ib polysaccharide, there is al arger flexibility,w ith am ajor exo-syn-F form (ca. 55 %). The model structures for the polysaccharides with all Neu5Aca2-3Gal link-    [14b] The favored presentationo ft he different epitopes for the major conformation is rather different. However,g iven their intrinsic flexibility,e specially for Ib, both molecules could be accommodated to interactw ith the monoclonal binding pockets without am ajor entropyp enalty. [31] Conclusions To have fast access to homogeneous oligosaccharide antigens from GBS serotypes Ia and Ib and to gain insights into the conformational difference among these structurally similarp olymers, we developed ah ighly convergent synthetic strategy based on the regioselective glycosylation of ag alactose C3,4diol to obtain GlcNAcb1-3Gal disaccharide building blocks. Investigation of the different reactivities of the C3-and C4-hydroxyls allowed us to reduce the number of protectiveg roup manipulations and synthetic steps to the final fragments, therefore simplifying the overall synthetic design.
Particularly,t he use of a2 ,6-O-benzoyl galactose diol resulted in improved regioselectively relative to that of the 2,6-di-Obenzylc ounterpart. In addition, mild activation conditions (NIS/AgOTf)f or the glucosamine thiol donors or the torsional disarming effect of the benzylidene group for the trichloroacetimidate donors appear to favor the glycosylationr eaction. The regioselective glucosamine incorporation was successfully applied to the synthesis of GBS CPS Ia and Ib branched repeating units (1 and 2). Their linear frameshifts (3 and 4)a nd an on-sialylated CPS Ia form (5)w ere also synthesized to achievea na dditional regioselective glycosylation of the Gal C3-OH over the C4-OH residue.
These results support the general applicability of the methodt oavariety of medically relevant glycans. Importantly, the structures synthesized through regioselective glycosylation appear extendible at the 4-OH positiono ft he Gal residue, thus potentially enabling the synthesis of longera nd more complex GBS oligosaccharide structures.
Conformation analysis studies of the prepared oligosaccharides by NMR spectroscopy and MD simulations showed the impact of the GlcNAcb1-3Gal versusG lcNAcb1-4Gal connectivity in the orientation of the Neu5Aca2-3Gal branching. The model,e stablished from the single synthetic pentasaccharide repeating units,w as used to study the conformational behavior of the GBS Ia and Ib polysaccharides, showing ad ifferent preferential shape for each polysaccharide with the Neu5Aca2-3Gal linkages in exo-anti-F for Ia and exo-syn-F for Ib. These unique structuralf eatures are expected to influence antibody recognition and immunospecificity.S tudies are ongoing to map the relevant glycoepitopes. Moreover, all glycanswere designed with ac hemicalh andle for conjugation to carrier proteins for immunological evaluation. Results on structurala nd immunogenic studies will be reported in due course.