Amylose Dimerization in Solution Can Be Studied Using a Model System

Amylose, the linear polymer of α‐1,4‐linked glucopyranose units, is known to crystallize as a parallel double helix, but evidence of this duplex forming in solution has remained elusive for decades. We show how the dimerization of short amylose chains can be detected in solution using NMR spectroscopy when the glucans are labeled at the reducing‐end with an aromatic moiety that overcomes chemical shift degeneracy leading to distinct signals for the single‐stranded and duplex amylose. A set of α‐1,4 glucans with varying lengths of 6, 12, 18, and 22 glucose units and a 4‐aminobenzamide label were synthesized, enabling the first systematic thermodynamic study of the association of amylose in solution. The dimerization is enthalpically driven, entropically unfavorable and beyond a minimum length of 12, each additional pair of glucose residues stabilizes the duplex by 0.85 kJ mol−1. This fundamental knowledge provides a basis for a quantitative understanding of starch structure, gelation and enzymatic digestion, and lays the foundations for the strategic use of α‐1,4‐glucans in the development of self‐assembled materials.


Introduction
Glucose-based polysaccharides are the most abundant biomaterials in Nature.Their functions and physicochemical properties are directly linked to both the nature of their glycosidic linkages and the consequent possibilities for defined secondary structure and higher-order assemblies. [1]For example, in the solid state, α-1,4-glucans form left-handed double helices (amylose, Figure 1a), [2][3][4] β-1,4-glucans align in sheets (cellulose) and β-1,3glucans form right-handed triple helices (curdlan). [1]α-1,4glucans can also form single left-handed helices around hydrophobic guests (known as V-amylose), both in the solid state , [5,6] and in aqueous solution: , [7,8] a classic example is the binding of polyiodide to starch (Figure 1b). [9] The structure of starch has been the subject of extensive research over several decades, and yet the solution behavior of amylose has remained poorly understood.While it is well-established that amylose crystallizes in a parallel left-handed double helix, [2][3][4] evidence of this structure in solution has remained elusive.
The solid state double helical structure of amylose is found in two allomorphs: the closely packed A-amylose , [2,3] and the more openly packed B-amylose. [4]Both consist of two lefthanded parallel helices with six glucose units per turn (Figure 1a) and can be distinguished using solid-state 13 C NMR spectroscopy. [12]A minimum degree of polymerization (DP) of 10 is required for crystallization, [13] suggesting that almost two turns are needed to form a stable helix.Double helices within amylose gels have also been reported based on analysis by Xray diffractometry, [14] 13 C CP/MAS NMR spectroscopy [15] and CryoTEM. [16]Cylindrical structures compatible with short double helices were detected during the enzymatic synthesis of amylose using SAXS. [19]Computational modelling of short amylose chains in solution suggests that the parallel double helix is the preferred conformation. [20]xperimental investigation of the solution structure and dynamics of α-1,4-glucans is hampered by the severe NMR chemical shift degeneracy for homooligosaccharides in solution, [21] as well as the challenge of synthesizing or isolating individual oligosaccharides of specific length. [22]Seeberger and co-workers recently described the total synthesis of maltohexadecaose (G16) using automated glycan assembly. [23]Instead, we chose an enzymatic synthesis to obtain a mixture of α-1,4glucans and successfully separated these chromatographically. , [24,25] We synthesized a set of four different length labeled α-1,4-glucan oligosaccharides.Herein, we report our discovery that these short labeled amylose chains can dimerize in solution, allowing, for the first time, the quantification of the binding interaction between associating amylose chains in water.
Upon characterization of these isolated labeled maltooligosaccharides using NMR spectroscopy, we observed two sets of signals for G12-AB, G18-AB and G22-AB in aqueous solution  [3] and V-amylose with 1-butanol guest omitted. [6]b) Schematic illustration of the solution state structures of amylose: random coils can assembly into guest-included single helices or dimerize, as shown herein.(Figure 3).Moreover, the relative intensity of the additional set of signals increased as the length of the glycan increased for samples at the same concentration.
We speculated that the second set of signals could result from dimerization, and thus performed dilution experiments.Figure 4a shows 1 H NMR spectra for the dilution series of G18-AB (see Figure S14 for full spectra).The ratio between the proposed duplex (signals highlighted in red) and monomer (signals highlighted in blue) decreased upon dilution and the duplex was not detectable at 0.1 mM.The duplex signals exhibited a larger linewidth indicating a larger molecular weight (Figure S12) and a DOSY experiment showed two distinct species (Figure S21).A slightly smaller diffusion coefficient was found for the proposed duplex, suggesting that it must have a compacted structure.Similar dilution experiments were performed for G12-AB and G22-AB and showed the same trend (Figures S12-S15).
A quantitative analysis of the NMR dilution study was necessary to distinguish dimerization from possible aggregation.By integration of the aromatic signal HI, at ~7.56 ppm for the monomer and at ~7.53 ppm for the proposed duplex, the concentrations of the duplex ([(Gn-AB) 2 ]) and monomer ([Gn-AB]) could be obtained, taking into consideration that [Gn-AB] T = [Gn-AB] + 2[Gn-AB] 2 , where [Gn-AB] T is the total concentration of the oligosaccharide.In the case of dimerization, the equilibrium constant can be obtained from 2 were made for G12-AB, G18-AB and G22-AB (Figure 4b).In each case, a linear correlation was observed, confirming that the additional set of signals in the NMR spectra were due to dimerization.From the gradients, dimerization constants (K dim ) of 230 M À 1 , 1500 M À 1  and 7000 M À 1 were obtained for G12-AB, G18-AB and G22-AB, respectively (Table 1).
A plot of ~G°against degree of polymerization (DP revealed a linear relationship between the length of the α-1,4-glucan and the Gibbs free energy change upon dimerization ( ~G°= -0.85×DP-3.5kJmol À 1 ) (Figure 5a).From the gradient we could compute that each additional pair of glucose residues stabilizes the duplex by 0.85 kJmol À 1 .Extrapolating backwards to shorter α-1,4-glucans, an association constant of 29 M À 1 would be predicted for G6-AB.However, dimerization of G6-AB was not observed even at 29 mM, at which concentration 6.8 mM duplex and 15.2 mM monomer would be predicted if the linear trend continued (Figure S16).It must therefore be inferred that, as seen in the solid state, [13] there is a minimum length requirement for the formation of amylose duplexes in solution, and, in the case of these labeled α-1,4-glucans, that minimum length occurs between DP6 and DP12.
To obtain thermodynamic parameters for the dimerization process, 1 H NMR spectra were recorded for G12-AB, G18-AB and G22-AB at different temperatures (303-333 K) (Figure S17-S20), and a van't Hoff analysis was made (Figure 5b).The dependence of ~G°, ~H°and ~S°on the length of the α-1,4glucans is shown in Figure 5a and Table 1.Duplex formation in solution is enthalpically favourable and entropically unfavourable.[4] The formation of such a network of intermolecular hydrogen bonds could provide an enthalpic benefit also in solution, whereas an entropic penalty would be expected based on a loss of conformational flexibility and translational freedom upon dimerization.The observed thermodynamic signature could also indicate that a non-classical hydrophobic effect is driving the dimerization process, although this phenomenon is more typically associated with complexation within macrocycles. [28]uplex formation is increasingly enthalpically favourable and Table 1.Binding parameters for the dimerization of the linear α-1,4-glucans.increasingly entropically unfavourable as the length of the α-1,4-glucan increases.However, the entropic cost per glucose unit decreases as the glucan length increases, and thus duplex formation in solution is increasingly favourable as the α-1,4glucan length increases.It was expected that the aminobenzamide (AB) label could play a role in stabilizing the duplexes.To investigate this, we obtained a commercial sample of unlabeled maltoundecaose (G11).The NMR spectrum of G11 showed only a single set of signals, and the spectrum was unchanged upon dilution and temperature variation (Figures S22 and S23).On the other hand, ITC experiments did indicate that an endothermic dissociation process took place upon dilution of G11 but not G6 (Figures S24 and S25).Evidently, the AB label plays a key role in overcoming the chemical shift degeneracy in the NMR spectra of these α-1,4-glucans such that signals from the monomer and the duplex become distinct from one another.It is also likely that the AB label contributes to the stabilization of these duplexes, but as G6-AB does not dimerize even at high concentration, the stabilizing influence of the AB label must be weak, and a specific interaction between the oligosaccharide chains of sufficient length is required for duplex formation.The 'per glucose' stabilization energy of 0.85 kJmol À 1 , calculated for glucans of at least DP12 (Figure 5a), is independent of the AB label, and can, thus, be used to predict the self-association of native amylose, and shorter α-1,4-glucans, in solution.
[4] Structural assignment of the duplex in solution suggests that this arrangement is maintained in solution, at least, for these model compounds.A full assignment of signals in the 1 H- 13 C HSQC spectra (Figures S6-S11) of the monomer/ duplex mixtures for G12-AB, G18-AB and G22-AB was performed (Tables S4-S6).The partial 1 H- 13 C HSQC spectrum of G18-AB is shown in Figure 5c, wherein distinct signals from the monomer are colored blue, distinct signals from the duplex are colored red, and signals where the monomer and duplex overlap are colored in black.For the protons on the sugar residues A and B, which are closest to the AB label (signals H1 A -H6 A and H1 B -H5 B ), two distinct sets of signals are observed for the monomer and duplex, while the signals for the protons on the non-reducing end glucose (H1 R -H6 R ) are completely overlapped.This suggests that the two oligosaccharides in the duplex are parallel.In an antiparallel orientation, a change in the chemical shift for the protons at the non-reducing end would be expected upon duplex formation, as this would bring the non-reducing end glucose unit near the aromatic AB-label and some effect of the aromatic ring current would be expected.The same observations were made for G12-AB and G22-AB, and these experimental results are consistent with reported solution state simulations suggesting that parallel duplex formation is favored. [20]

Conclusions
In conclusion, we have identified a model system whereby the dimerization of short amylose chains can be observed and studied in solution.By installation of a 2-aminobenzamide label at the reducing-end of a set of α-1,4-glucans, the dimerization could be visualized using NMR spectroscopy.Quantitative information on polysaccharide interactions is generally very difficult to obtain and therefore often considered only as average or global values.While model peptides have been extensively used to help understand the thermodynamics and structural basis for folding in proteins, [29][30][31] the use of model oligosaccharides to explore the conformation and assembly of polysaccharides has, until recently, been very limited. [22,32]The synthesis and isolation of this set of individual α-1,4-glucans of different lengths (DP 6, 12, 18, and 22) as model compounds has allowed for a systematic study and quantification of the 'per glucose' interactions between chains of this important biopolymer.Specifically, the dimerization of amylose is enthalpically driven, entropically unfavourable, and beyond a length of 12 glucose units there is a linear correlation between the stability of the duplex and the α-1,4-glucan length: each additional pair of glucose residues stabilizes the duplex by 0.85 kJ mol À 1 .This fundamental knowledge is critical to understanding the physicochemical properties, structure, biosynthesis and enzymatic digestion of starch, and can provide a quantitative basis for the design and self-assembly of α-1,4glucan-based materials.

Figure 1 .
Figure1.a) Double and single helical solid-state structures of amylose: A-amylose[3] and V-amylose with 1-butanol guest omitted.[6]b) Schematic illustration of the solution state structures of amylose: random coils can assembly into guest-included single helices or dimerize, as shown herein.

Figure 4 .
Figure 4. Dilution studies enabled determination of the dimerization constants for different length α-1,4-glucans.a) 1 H NMR spectra of G18-AB at different total concentrations in D 2 O at 303 K. b) Plot of [(Gn-AB) 2 ] versus [Gn-AB] 2 at varying total concentrations used to determine the dimerization constants K dim .

Figure 5 .
Figure 5. Dimerization in enthalpically driven, linearly dependent on the glycan length and occurs with the glycans oriented parallel.a) Plot of ~G°, ~H°and T~S°for the dimerization as a function of the degree of polymerization (DP) of the α-1,4-glucans.b) Van 't Hoff plot for the dimerization of the linear α-1,4glucans at 303-333 K. c) Partial 1 H-13 C HSQC NMR spectrum of G18-AB (0.9 mM) in D 2 O at 303 K.The signals corresponding to the monomer and the duplex are shown in blue and red, respectively, whereas overlapping signals are shown in black.