Systematic Hydrogen‐Bond Manipulations To Establish Polysaccharide Structure–Property Correlations

Abstract A dense hydrogen‐bond network is responsible for the mechanical and structural properties of polysaccharides. Random derivatization alters the properties of the bulk material by disrupting the hydrogen bonds, but obstructs detailed structure–function correlations. We have prepared well‐defined unnatural oligosaccharides including methylated, deoxygenated, deoxyfluorinated, as well as carboxymethylated cellulose and chitin analogues with full control over the degree and pattern of substitution. Molecular dynamics simulations and crystallographic analysis show how distinct hydrogen‐bond modifications drastically affect the solubility, aggregation behavior, and crystallinity of carbohydrate materials. This systematic approach to establishing detailed structure–property correlations will guide the synthesis of novel, tailor‐made carbohydrate materials.


Introduction
Cellulose,t he most dominant polysaccharide on earth, with 700 billion tons produced annually,i sa ni mportant material in the textile,f ood, paper,a nd pharmaceutical industries. [1] Thes tability,c rystallinity,a nd poor water sol-ubility of cellulose are the result of adense network of interand intramolecular hydrogen bonds that create allomorphs with different properties (Figure 1). [2] Theh ydrogen bond between OH(3) and O(5) of the ring stabilizes the cellobiose repeating unit, with additional stabilization gained from intraand intermolecular interactions (chain stacking) involving OH(6) and OH (2). [3] Hydrophobic interactions between the CH-rich, apolar faces of the glucose units as well as van der Waals forces also play an important role. [4] Although different levels of cellulose organization have been studied in detail, not all the allomorphs have been described. [5] Chemical modification alters the organization of cellulose and creates new materials with enhanced water solubility or ionic character. [6] Non-regioselective derivatization results in polydisperse materials with respect to the length and modification patterns,w hich do not allow for proper structure-function correlations. [6d] Thel ack of standards and experimental data has hampered in silico modeling studies.Molecular dynamics (MD) simulations capture some structural changes, [7] but ad etailed structural description is often lacking due to the flexibility of the carbohydrates.
Thes ynthesis of carbohydrate materials by polymerization [8] or enzymatic reactions [9] provides an attractive alternative to the modification of natural polysaccharides, but has been limited by poor product solubility and narrow substrate scope.L aborious procedures for the synthesis of oligosaccharides have been overcome by automated glycan assembly (AGA), which enables rapid access to synthetic polysaccharides as long as 50-mers. [10] Well-defined natural and unnatural glycans served as useful probes for systematic structural investigations,w hich revealed that even hexasaccharides adopt distinct secondary structures. [11] Here,weuse tailor-made cellulose derivatives,designed to selectively disrupt hydrogen-bond networks and/or alter the electronic properties,t oestablish as tructure-property relationship ( Figure 1). [12] Methylated, deoxygenated, and deoxyfluorinated cellulose,i na ddition to well-defined carboxymethyl cellulose (CMC) and chitin analogues,a re prepared with full control over the length, pattern, and degree of substitution. MD simulations guided the synthesis,b yc orrelating the disruption of the hydrogen-bonding network with the increased flexibility of the modified oligosaccharides.
All the unnatural derivatives are highly water soluble as they aggregate less.A nalogues with the same degree of substitution, but different substitution patterns, show dramatic differences in the conformation and aggregation. Crystallographic (XRD) and solubility analysis confirmed the in silico prediction, strongly supporting that fine-tuning the natural polysaccharide backbone greatly influences the macroscopic material properties.T hese insights will guide the development of novel, high-performance biomaterials.

Results and Discussion
Automated glycan assembly (AGA) increases the efficiency of oligosaccharide synthesis by iteratively combining monosaccharide building blocks (BBs) on as olid support, thus replacing laborious purification processes with simple washing steps. [10a,13] Each BB is equipped with areactive thioglycoside leaving group and atemporary Fmoc protecting group that is easily removed after glycosylation to release af ree hydroxy group that serves as the new glycosyl acceptor in the next coupling cycle.Iterative glycosylation and deprotection cycles allow for the step-wise elongation of polysaccharides and the insertion of specific modifications in defined positions of the chain. Thef ully protected glycan target with afree reducing end is released from the solid support upon cleavage of the UV-labile linker 1a ( Figure 2). [14] To overcome glycan decomposition during the basic methanolysis of the ester protecting groups in the presence of the free reducing end, as olid-phase methanolysis was developed. Subsequent photocleavage and hydrogenolysis then afforded the desired oligosaccharides.A lternatively, cleavage of linker 1b liberates the desired glycan equipped with the 4-hydroxymethylbenzyl group at the reducing end, thereby allowing for solution-phase methanolysis and subsequent cleavage during hydrogenolysis.Acollection of welldefined cellulose derivatives was prepared by AGA ( Figure 2). Tw on atural cellulose oligomers (hexamer A 6 and dodecamer A 12 )a nd one chitin analogue (N 6 )s erved as standards for the structural analysis.U nnatural analogues with defined substitution patterns were prepared to tune the conformation and properties of the material. Regioselective functionalization was achieved with five "unnatural" monosaccharide BBs 3-7 ( Figure 2). Global deprotection afforded oligosaccharide derivatives with complete control over the length, pattern, and degree of functionalization.
Methylation effectively alters the solubility and gelation properties of cellulose by influencing the intra-and intermolecular hydrogen bonding.M ethylcellulose is widely used in the food and pharmaceutical industries. [15] Six hexamers and four dodecamers with different methylation patterns were synthesized using BBs 3 and 4,w hich contain 3-methyl and 3,6-dimethyl motifs,r espectively ( Figure 2). Thep osition of the substituents was chosen to selectively disrupt hydrogen bonds that play af undamental role in the rigidity of the cellulose.Methylation of OH(3) impedes the hydrogen bond between O(5) and OH(3), while 6-methylation hinders the inter-a nd intrachain stabilization offered by OH(6). Structures with aregular methylation pattern (e.g.(AB) 3 ), diblock analogues (e.g. A 3 B 3 ), as well as irregularly functionalized structures (e.g.(ABA) 2 )were assembled to assess the effect of methylation patterns on the overall cellulose conformation. 3-Deoxygenated BB 7 prevents the formation of hydrogen bonds between O(5) and OH (3), while 3-deoxyfluorinated BB 5 not only prevents hydrogen-bond formation, but is expected to affect the overall dipole of the oligomer as ar esult of the strong electron-withdrawing nature of fluorine. [12b,16] Three CMC derivatives were prepared using BB 6 to evaluate the effect of negative charges on the resulting material. Lastly,o ne hybrid cellulose-chitin derivative (ANA) 2 was assembled.
Thesynthesis of A 6 was low yielding (18 %) due to the low solubility of the oligosaccharide product. Methylation and carboxymethylation drastically improved the product solubility,w hich is reflected in higher yields for the unnatural hexasaccharide analogues (26-73 %o verall yield). Similar results were observed for the 12-mer syntheses ( Figure 2) with the insertion of BBs 3, 4,a nd 6 with higher yields observed than for A 12 (2 %o verall yield). 3-Deoxyfluorination and 3deoxygenation also improved the product solubility,b ut the reduced reactivity of BBs 5 and 7 as ag lycosyl acceptor resulted in moderate yields (9-25 %overall).
Theperturbation of the 3D shape of the oligosaccharides as aresult of single-site substitutions was modeled using MD simulations,e mploying am odified version of the GLY-CAM06 carbohydrate force field. [17] Theeffect of substitution of the neighboring monomer on the torsion angles (w, Y, F) was monitored and compared with the unsubstituted analogue A 6 ( Figure 3). Particular attention was paid to the changes in the population of Y,which is directly related to the presence of ah ydrogen bond. To monitor the overall conformation of the hexamers,t he end-to-end distance ( Figure 4) and the radius of gyration (RoG;F igure S4) were calculated as af unction of time.C ellulose A 6 and chitin N 6 oligomers showed af airly rigid backbone core with low conformational variability (average end-to-end distance 2.76 AE 0.17 nm for A 6 ;F igure 4). Both structures tend to adopt an extended helical conformation (Figure 2). To examine how specific modifications affect such organized structures,t he series of methylated analogues was studied ( Figure 4). Ap reliminary analysis of Cremer-Pople parameters [18] showed surprisingly frequent 4 C 1 -to-1 C 4 interconversions for all methylated analogues (B and C)d uring the simulation time (even at the monomer level). However,t his tendency was disproved by NMR analysis of the 1 J C1H1 and 3 J H1H2 values (see the Supporting Information), and thus dihedral restraints were applied to these monomers (B and C)inthe simulations to prevent the "flipping-chair" artifact.
Aregular alternating substitution pattern, as in the case of (AB) 3 ,r evealed am oderate,b ut important, decrease in the population of Y at negative degrees (À278 8). This results from the increased distance between OMe(3) and O(5) because of the decreased tendency to form hydrogen bonds and the increased steric bulk (Figure 4). Thes ame degree of methylation with ab lock distribution A 3 B 3 resulted in dramatic changes.Asignificantly more flexible bent shape (Figure 2) with an end-to-end distance of 2.65 AE 0.26 nm was observed for most of the simulation time.S urprisingly,t he OH(3)···O-(5) hydrogen bond between the first two glucose monomers was detected for most of the simulation time,thus suggesting the coexistence of ar igid rod block (A 3 )a nd av ery flexible counterpart (B 3 ;F igure 4).
Methylation at the 3-and 6-positions (C), aimed to reduce inter-a nd intramolecular hydrogen bonds,d isrupts the "standard" dihedral values,thereby resulting in acompletely new geometry (Figure 4). Ramachandran plots of the dodecamers ( Figure S16) confirmed that increased length enhances the resistance to deformation, since the cooperativity of intramolecular hydrogen-bonding interactions stabilizes the overall structure.N evertheless,anoticeable deviation from the main population of A 12 was observed for all the substituted analogues.A ni rregular substitution pattern appears to be important to drastically change the cellulose conformation (e.g. (ABA) 2 A 3 B 3 ). Ar egular substitution pattern such as (ABA) 4 maintains more cellulose character while improving the water solubility.
Similar to methylation, deoxyfluorination and deoxygenation prevent the formation of hydrogen bonds between O(5) and OH(3). In addition, these substitutions influence the electron density along the chain (electronegative F) and the steric hindrance (deoxygenation). Since dipoles are key to the stability of cellulose,the replacement of OH(3) by the isosteric electronwithdrawing Fisexpected to greatly influence the conformation of the resulting material. Thec alculated mean RoG for (AFA) 2 shows al arge dispersion and the average end-to-end distance is among the lowest (2.61 AE 0.34 nm), indicative of avery flexible system (Figure 4) with alower population at negative degrees of Y 1 and Y 4 .This effect extends beyond the single AF glycosidic bond, with significant variation of Y 3 (Figure 4). 3-Deoxygenation had an even bigger effect on the Y distribution for (AEA) 2 ,asreduced steric hindrance allows for more conformational freedom. Theinsertion of acarboxylic group (e.g. (ADA) 2 )r esulted in ah ighly flexible,m ostly linear conformation (Figure 2). Moreover,t he carboxylate can engage in additional hydrogen bonds,a sobserved between COO À and OH(2) of the same residue,a sw ell as between COO À and OH(6) of the adjacent previous sugar residue (OH(6)···COO À ···OH (2)).
Theb ehavior of the oligosaccharides in ac rowded environment was studied and correlated to the crystallinity and solubility of the materials.L ong MD simulations (1 ms production run) of concentrated experiments (see the Supporting Information) aimed to elucidate the molecular interactions.Radial distribution functions (RDFs) were used to characterize the spatial correlations in the systems (Figure 5). TheR DF for A 6 shows three sharp signals at small distances and remains large for distances up to 1.5 nm, which indicates high aggregation tendencies of such oligosaccharides.T he more soluble methylated analogue (AB) 3 shows some tendencytoaggregate at high concentrations.However, as ignificantly decreased signal at 0.5 nm indicates the lower probability of finding two chains in proximity,c ompared to cellulose oligomers.R DF peaks are only found at shorter distances,t hus revealing al ower tendencyf or cluster formation and al ess organized structure,w ith ah omogeneous distribution of molecules beyond the nearest neighbors.N o aggregation was detected for A 3 B 3 ,asexpected from the high flexibility of such compounds,w hich should prevent chain stacking ( Figure 5). X-ray diffraction and solubility data ( Figure 5) support the calculations.A sa nticipated, A 6 and A 12 are very poorly soluble in water (less than 1mgmL À1 ), due to the formation of cellulose-like aggregates.P owder XRD measurements of both A 6 and A 12 gave sharp peaks (Figures 5a nd 6) that are distinctive for cellulose II, thus indicating that short oligomers adopt the same aggregation pattern and the same hydrogenbonding arrangement as cellulose.The flat XRD profile of the diblock analogue A 3 B 3 indicates the absence of any structural organization, as predicted by the theoretical model ( Figure 5). Thea lternating methylation pattern of (AB) 3 renders the material more sensitive to the X-ray beam angle and, while the XRD peaks are still broad, they resemble the cellulose II structure,a sp redicted by MD simulations (Figure 5). This trend is confirmed by the longer oligomers,w here more intense,y et broad, XRD profiles are observed for the regularly substituted analogues (Figure 6). No cellulose-like character is detected for randomly functionalized structures. Similart oc ellulose,t he XRD profileo fc hitina nalogueN 6 is identicalt ot hat of naturalc hitin ( Figures5 andS 2),a si ti s poorly soluble( 13-17mgmL À1 )a nd tendst of ormg elsa t higher concentrations.S urprisingly, theh ybridc ellulose-chitin (ANA) 2 ,ismuchmoresoluble (> 50 mg mL À1 )withnoordered supramolecular structures (Figure6). Allt he functionalized cellulosea nalogues are, in contrast to then atural derivatives, highly water-soluble( > 50 mg mL À1 )a nd form amorphous solids (Figure6). Interestingly, although remaininghighlywatersoluble, thed eoxy series (E)a dopts acellulose-likecharacter in thesolid statewithtwo broad, butnoticeable, peaksinthe XRDprofile ( Figure 6) and as imilar peak structure in the RDF (Supplementary Information).

Conclusion
Tailor-made cellulose oligosaccharide analogues,p repared by sequential addition of monomeric BBs using AGA, allow for control over the length and substitution patterns.S even BBs were prepared bearing modifications to disrupt specific hydrogen bonds and tune the three-dimensional shapes and properties of the materials.Methylation blocked the hydrogen bond between OH(3) and O(5), thereby resulting in an increased flexibility of the chain, as observed by MD simulations.Adetailed dihedral analysis depicted how each glyosidic bond is affected by the modifications,a nd the consequences for the overall structure,s uch as fluctuation of the end-to-end distance during the simulation time.Compounds with the same degree of methylation, but different substitution patterns,b ehave drastically different. Regular substitution patterns result in quasilinear structures,w hereas more bent geometries are observed with ab lock arrangement. These structural features control the aggregation process,w hich is expressed by high crystallinity  for the natural compound and amorphous organization for irregular or block-substituted analogues.Amore significant disruption of the "standard" dihedral values was observed with methylation at the OH(3) and OH(6) positions,aswell as for the deoxy derivatives (E). Interestingly,u pon drying, the highly water-soluble deoxy derivatives show at endency for cellulose-like packing. Replacement of the OH(3) group by the isosteric electron-withdrawing Fatom resulted in compact analogues (shortest end-to-end distance) and amorphous organization. Carboxylates (D)o ra mides (N)d erivatives made new conformations accessible thanks to the formation of additional hydrogen bonds.All the unnatural analogues are drastically more soluble,d ue to the more flexible backbone. Novel biomaterials with tuned properties that could be engineered depending on the nature and pattern of the substituents can be envisioned. Thec ollection of unnatural compounds will be available to evaluate enzymatic degradation and substrate specificity.