Crystalline Structure in Starch

Abstract

Many reviews have been published on the crystalline structure of the starch granule, addressing aspects such as birefringence, crystallinity, and structural models for A- and B-type starches. After a synthetic presentation of the general knowledge on this topic, the present review focuses on a critical description of the main techniques used to investigate the starch crystalline structure, some new data regarding crystalline lamellae, and the most recent models established for the 3D structure of crystalline domains in the granules. Structural and phase transitions occurring during hydrothermal treatments of starch are briefly presented as an introduction to a more detailed description of local order and orientation in amorphous starch materials. Recent results on the structure of amylose complexes which form by heating in the presence of guest molecules are discussed as well. Finally, results regarding the in vitro enzymatic synthesis of amylose, which self-associates into gels or particles, and in vitro enzymatic extension of glycogen external chains are described. They are evaluated as biomimetic systems for a better understanding of the mechanisms involved in starch crystallization during biosynthesis as well as in the different processes used for starch modification.

1 The Crystalline Structure of Native Starch

Starch is biosynthesized as semicrystalline granules whose shape and size depend on the botanical origin (Fig. 3.1a, c). The starch granule presents a hierarchical and multiscale organization with structural length scales ranging from the tenth of nanometer for the glucose monomer up to a few micrometers for the granule size. Intermediate structural levels include the crystalline and amorphous lamellae repeat distance (9–10 nm), the blocklets (30–200 nm), the growth rings (200–600 nm), and other supramolecular arrangements like superhelices (for reviews, see French 1984; Buléon et al. 1998a, b; Tang et al. 2006; Pérez and Bertoft 2010).

Fig. 3.1

Potato starch granules observed by scanning electron microscopy (a), polarized light optical microscopy (b), scanning confocal light microscopy (c). Scale bars: 20 μm (From Buléon et al. 2007 – images courtesy D. Dupeyre, CERMAV). (d, e) Synchrotron X-ray microfocus mapping of a smooth pea starch granules (From Buléon et al. 1998a, b)

1.1 About Chain Orientation in the Starch Granule

When starch granules are observed under polarized light, a characteristic Maltese cross (centered at the hilum) can be seen (Fig. 3.1b), which has led to consider the granules as distorted spherocrystals. The sign of birefringence is positive with respect to the spherocrystal radius (French 1984) which indicates that, theoretically, the average orientation of the polymer chains is radial. The intensity of birefringence strongly depends on the shape and orientation of the granules with respect to the light beam. Therefore, for nonspherical granules, it is more accurate to say that the orientation is perpendicular to the growth rings and to the granule surface (French 1984). This was also confirmed by solid-state light scattering of starch granules (Borch et al. 1972).

Further results on chain orientation in potato and wheat starch granules were obtained by Chanzy et al. (1990) and Helbert and Chanzy (1996), using selected area electron diffraction over 1 μm² on thin sections of partially hydrolyzed starch granules. However, in both studies, the diffractograms were not very well resolved because of the large amount of inelastic electron scattering inherent to this technique.

Significant improvement was achieved using synchrotron microfocus X-ray diffraction with a 1–2 μm beam. Buléon et al. (1997) (Fig. 3.1d, e) and Waigh et al. (1997) showed that the orientation of polymer chains in potato starch was very high at the periphery of the granule and perpendicular to its surface. A lower degree of orientation was found in the internal regions and at the hilum, but the interpretation was more complex since in the center of the granule, the beam averages over regions containing helices pointing forward, backward, and sideways. No specific orientation at a 1–2 μm scale was found for wheat starch granules, either at the periphery or in the center (Buléon et al. 1997), which means that the radial orientation in such granules is weak and limited to smaller domains. Nevertheless, these pioneering works also showed that it was possible to record highly oriented diffraction diagrams, very close to those stemming from recrystallized amylose fibers used to solve the three-dimensional structure of B-type amylose (Buléon et al. 1997). These results show that the corresponding models can be transposed to the crystalline domains in the starch granule.

More recently, many synchrotron microdiffraction experiments confirmed the radial chain orientation in starch granules from various botanical origins (Chanzy et al. 2006). The most recent advances of synchrotrons and the development of the ID13 microfocus beamline at ESRF (Grenoble, France) (Riekel 2000) have significantly improved the microdiffraction method which has now been applied to a number of specimens to study specific properties of individual granules, for instance, the structural changes induced by microdrop hydration (Lemke et al. 2004), high pressure (Gebhardt et al. 2007), or radiation damage propagation (Riekel et al. 2010). X-ray probes with a diameter as small as 0.5 μm are currently available, with the brilliance of the synchrotron, which allows to probe the molecular organization in individual concentric growth rings in the starch granule.

1.2 About Crystallinity

The inner architecture of native starch granules is characterized by “growth rings” (Fig. 3.2a, b) that correspond to concentric semicrystalline 120–400 nm-thick shells separated by amorphous regions (French 1984; Buléon et al. 1998a, b; Donald et al. 2001). The crystalline shells consist of a regular alternation of amorphous and crystalline lamellae with a repeat distance of 9–10 nm (French 1984; Cameron and Donald 1993; Jenkins et al. 1993; Cardoso and Westfahl 2010). Parallel arrays of double helices made of amylopectin short branches form crystalline lamellae (Fig. 3.2c, d). Their 5–7 nm thickness presents some variation depending on the botanical origin. Native starch granules exhibit two main allomorphic types that can be identified using X-ray scattering (Fig. 3.2g) or solid-state ¹³C NMR spectroscopy. The A type mainly occurs in cereal starches and the B type in tubers and amylose-rich starches (Buléon et al. 1984; Zobel 1988). New morphological data on the individual crystalline lamellae of waxy maize starch granules have been reported (Putaux et al. 2003). The mild hydrochloric acid hydrolysis of granules from A-type amylopectin-rich waxy maize leads to an insoluble residue consisting of polydisperse platelet nanocrystals that retain the crystalline A type of the parent granules. They form parallelepipedal blocks with a length of 20–40 nm and a width of 15–30 nm, with characteristic geometrical features such as 60–65° acute angles (Fig. 3.2e, f). Their 5–7 nm thickness is consistent with the presence of double helices made of single strands having a degree of polymerization (DP) between 15 and 18. Such crystalline lamellae could not be as clearly individualized from B-type starch, although lacy networks of acid-resistant units were obtained from low-amylose Dianella potato starch (Wikman et al. 2014).

Fig. 3.2

Schematic representation of the starch granule ultrastructure from growth rings to amylopectin: (a) ultrathin section of a waxy maize starch granule (staining with uranyl acetate and lead citrate (TEM image courtesy I. Paintrand, CERMAV); (b) alternation of amorphous and semicrystalline growth rings; (c) clustered model of amylopectin; (d) detail of one cluster showing the formation of double helices from the short branches of amylopectin; (e) nanocrystals observed after 6 weeks of acid hydrolysis (TEM image after negative staining courtesy of H. Angellier, CERMAV – scale bar: 50 nm); (f) simplified drawing of the crystalline platelets and proposed molecular model in relation to the monoclinic unit cell of the A allomorph; (g) WAXS profiles of A- and B-type starch (From Buléon et al. 2007)

The crystallinity of native starch granules varies from 15 to 45 %, depending on the starch origin, its hydration level, and the characterization method (Buléon et al. 1987; Paris et al. 1999; Zobel 1988; Gernat et al. 1993; Lemke et al. 2004; Lopez-Rubio et al. 2008). The B-starch crystallinity depends more strongly on hydration and increases with water content up to 30–33 % H₂O (dry basis), while dry potato starch does not show any clearly visible diffraction peak (Buléon et al. 1987, 1998a, b; Cleven et al. 1978). Wide-angle X-ray scattering (WAXS), solid-state ¹³C NMR (CP-MAS), differential scanning calorimetry (DSC), and, to a lesser extent, Fourier-transform infrared spectroscopy (FTIR) are the most commonly used techniques to determine the degree of crystallinity. They are based on different physical phenomena which makes the comparison of the collected data very tricky, especially for starch. Indeed, contrary to thermoplastic samples for which the melting enthalpy is directly correlated with the crystallinity index, starch melting enthalpy cannot be used to determine crystallinity since numerous and various processes are involved during melting, including plasticization, swelling in water, competition between melting and dissolution in water, etc. Therefore, the usual relevant interpretation concerns the residual melting enthalpy which varies as the starch structure is disrupted but cannot be quantitatively correlated with the crystallinity. Moreover in case of amylose-lipid complexes, the melting of “isolated” or “amorphous” complexes is very close to that of crystalline ones since the intramolecular hydrogen bonds within the single amylose helix are much stronger than intermolecular H bonds in the crystal (Biliaderis 1992).

When using FTIR, the 1,065−870 cm⁻¹ spectral region is assigned to C-O-C, C-C, and C-H stretching modes, and some peaks have been attributed to ordered (995, 1,047 cm⁻¹) or amorphous (1,022 cm⁻¹) starch (Bernazzani et al. 2000; Capron et al. 2007; Wilson et al. 1987; van Soest et al. 1995). According to Fang and coworkers (2002), the bands located at 1,023 and 1,083 cm⁻¹ could be assigned to the C-O bond stretching in the C-O-C group of the anhydroglucose rings. The peak at 1,023 cm⁻¹ has been shown to be typical of amorphous starch, while that at 1,049 cm⁻¹ increases with increasing crystallinity. However, even when using well-defined crystalline samples, it is difficult to conclude if the FTIR signal is sensitive to short-range order or crystallinity. More recently, the technique was successfully used to determine local order in shape memory starch-based materials (Véchambre et al. 2010).

WAXS and solid-state NMR have proved to be the more reliable techniques to determine the crystallinity and the allomorphic types of starch products. WAXS signals directly stem from the crystalline regions and long-range order. Numerous methods are available to determine the crystallinity from WAXS data. Starch crystallinity was initially calculated by Sterling (1960) and Nara et al. (1978) following the methods developed by Hermans and Weidinger (1948) and Wakelin et al. (1959) for cellulose. They are based on the two-phase concept which assumes that relatively perfect crystalline domains (crystallites) are interspersed with amorphous regions. Solid-state NMR has proved to be a powerful tool to characterize some degree of molecular order such as helicity in the structure of starchy substrates (Gidley and Bociek 1985; Veregin et al. 1986; Horii et al. 1987; Morgan et al. 1995). Gidley and Bociek (1985) demonstrated that C1 and C4 glycosidic sites were more sensitive to conformational changes than the C2, C3, and C5 carbons since C1 and C4 showed higher chemical shift dispersions under various conformations of the glycosidic linkage in α(1,4) glucans. These statements have been confirmed by Veregin et al. (1987) and Gidley and Bociek (1988) who showed correlations between the C1 and C4 chemical shifts of α(1,4) glucans and their torsion angles ϕ and ψ. The multiplicity (stemming from crystallographic constraints in the crystalline regions) and the chemical shifts assigned to C1 atoms have been correlated to the glycosidic conformation and polymorphism of starch (Hewitt et al. 1986; Veregin et al. 1986; Horii et al. 1987; Paris et al. 1999) and other polysaccharides (Jarvis 1994; Isogai et al. 1989). The multiplicity of the C1 resonance is determined by space group equivalent symmetry classes (Veregin et al. 1986; Horii et al. 1987), and for starch, the C1 resonance appears as a triplet for A type and a doublet for B type. Strictly integrating these resonances, directly linked to the crystalline domains, leads to crystallinity values close to those determined by WAXS (Paris et al. 1999).

The influence of hydration on the ordering of starch chains and decrease of the spectral bandwidth was also studied (Tanner et al. 1987; Horii et al. 1987; Paris et al. 1999). The NMR data have been used to identify, and sometimes quantify, the crystalline order of native, hydrolyzed, or gelatinized starches (Willenbucher et al. 1992; Paris et al. 1999), but there is still some debate about the exact level of structure reached with this technique, i.e., the degree of helicity (molecular order) or the fraction of helices packed with respect to crystalline order (Gidley and Bociek, 1985) and the quantitative character of the measurements. While diffraction methods are the main techniques used to quantify crystallinity, shorter-range subcrystalline order can be probed by solid-state NMR spectroscopy which is sensitive to structure at the sub-nanometer level. Starch spectra are interpreted in terms of a combination of amorphous and helical conformations, irrespective of whether helices are present within crystallites or not. Thus, Lopez-Rubio and coworkers (2008) have recently proposed to jointly use WAXS and ¹³C CP-MAS NMR spectroscopy to determine crystallinity, the amounts of single and double helices (Fig. 3.3), amorphous single and double helical components being determined from NMR spectra as described by Tan et al. (2007). In this work, the crystallinity was determined by fitting experimental WAXS curves with theoretical discrete diffraction peaks determined from the 3D structure of A or B starch and an amorphous background.

Fig. 3.3

Combined use of WAXS and ¹³C CP-MAS NMR for the determination of helical order and crystallinity: (a) fit of a A-type WAXS profile using an amorphous background and discrete diffraction peaks calculated from the A-type structure (From Lopez-Rubio et al. 2008); (b) solid-state ¹³C NMR spectra of A- and B-type recrystallized amylose (From Paris et al. 1999); (c) decomposition of the ¹³C NMR spectrum into contributions from the amorphous and ordered phases in potato starch by subtraction at 84 ppm and (d) deconvolution of the ordered subspectrum for potato starch (From Lopez-Rubio et al. 2008)

1.3 The Three-Dimensional Models of Crystalline Domains in Starch

The low crystallinity and complex ultrastructure of the starch granule do not allow direct determination of the three-dimensional shape and distribution of the crystalline domains. Mild acid hydrolysis has widely been used to investigate the structure and properties of starch crystallites by preferential degradation of the amorphous regions (Buléon et al. 1987, 1998a, b). The extent of this preferential erosion depends on the density difference between the crystalline and the amorphous regions (Vermeylen et al. 2004). As previously mentioned, A-type nanocrystals obtained by extended mild hydrolysis of waxy maize starch granules correspond to the crystalline lamellae present in the native granule (Putaux et al. 2003) (Fig. 3.2). However, electron diffraction data could not be collected from individual platelets, because of their small lateral size and thickness.

Contrary to amylopectin which constitutes the backbone of the semicrystalline structure of starch granules, amylose can easily be crystallized from solution by cooling or addition of a precipitant. This property has been used to prepare model crystalline substrates whose diffraction data and thermal properties were analyzed to propose three-dimensional models of starch crystallites and to understand the phase transitions involved in various applications and the processing of starch (Buléon et al. 1987; Whittam et al. 1990). The resulting morphology (aggregates, precipitates, gels, single crystals, spherulites, etc. – Fig. 3.4) and allomorphic type (A, B) depend on factors such as solvent, molar mass, branching degree, concentration, or temperature (Buléon et al. 1984, 2007; Pfannemüller 1987; Gidley and Bulpin 1987). A general rule is that long chains and low recrystallization temperatures favor the B type, whereas high concentrations, high temperatures, and short chains are known to induce A-type crystallization (Buléon et al. 2007). This behavior has some similarities with the crystallization of starch during biosynthesis, since it is well known that (i) B-type starches amylopectin short chains are longer than in A type (Hizukuri 1985); (ii) A type is mostly present in cereal grains, which grow in warmer and dryer conditions than tubers starches; and (iii) an increase of temperature during the plant growth favors A type (Gérard et al. 2000).

Fig. 3.4

Recrystallized amylose: (a) retrograded gel network from diluted long chains, (b) A-type crystals prepared from a narrow fraction of amylose biosynthesized in vitro (DPn = 17.4 – from Montesanti et al. 2010), (c) A-type spherocrystals and (d) corresponding electron diffraction pattern recorded from a ultrathin section of a spherocrystal (From Helbert et al. 1993)

Most molecular models proposed to describe the structure of the crystalline domains in starch granules were built using WAXS data collected from recrystallized fibers (Wu and Sarko 1978a, b; Imberty and Pérez 1988) or lamellar crystals (Imberty et al. 1988). The structural models established from A and B amylose crystals were transposed to crystalline regions of native starch, which exhibit similar but much less resolved diffraction diagrams. WAXS fiber diagrams recorded from native potato and pea starch granules using synchrotron microbeam X-ray mapping confirmed that these A- and B-type models could also be applied to describe native starch crystallites (Waigh et al. 1997; Buléon et al. 1998a, b). In these A and B structures, amylose exhibits a sixfold left-handed double helical conformation with pitches of 2.08 and 2.13 nm, respectively (Imberty et al. 1988; Imberty et al. 1988, Takahashi et al. 2004). In the A structure, these double helices are packed with the B2 space group in a monoclinic unit cell (a = 2.124 nm, b = 1.172 nm, c = 1.069 nm, γ =123.5°) with four water molecules per unit cell (Imberty et al. 1988). In the B structure, double helices are packed with the P6₁ space group in a hexagonal unit cell (a = b = 1.85 nm, c = 1.04 nm, γ =120°) with 36 water molecules per unit cell. The symmetry of the double helices differs in A and B structures, since the repeated unit is a maltotriosyl unit in the A form and a maltosyl unit in the B form, which is in agreement with the triplet and the doublet observed by solid-state NMR for the C1 peak for the A form and the B form, respectively (Fig. 3.3).

5–15 μm-long needlelike A-amylose single crystals have recently been prepared from dilute (0.05 % w/v) solutions of a short-chain amylose fraction (${\overline{\mathrm{DP}}}_n=17.4$) synthesized in vitro by amylosucrase from sucrose (see Sect. 3.3). Their size and perfection allowed to collect WAXS datasets from single crystals, up to a resolution of 0.151 nm, using a microbeam of synchrotron radiation (Fig. 3.5a, b) (Popov et al. 2006, 2009). A total of 57 independent reflections were used to refine the unit cell and confirm the space group proposed by Imberty et al. (1988). However, the high resolution of the diffraction data, which was unique for a crystalline polymer, allowed to resolve important new fine details. Pockets of two intracrystalline water molecules were distributed between the double helices, resulting in helical distortions (Fig. 3.5c). In addition, a tight network of hydrogen bonds involving the primary and secondary hydroxyl groups of the glucosyl moieties stabilized the structure (Fig. 3.5d). The refinement of the new structure indicated a “parallel-down” organization of the amylose molecules within the unit cell as opposed to the previous “parallel-up” model (Imberty et al. 1988). This new feature indicates that within the crystals, the nonreducing ends of the amylose molecules are oriented toward the c-axis direction of the unit cell. The description of this geometry is important to correlate the crystallography of the A-type granules with their ultrastructure and mode of biosynthesis.

Fig. 3.5

(a) SEM image of A-amylose single crystals prepared by crystallizing amylose biosynthesized in vitro; (b) one crystal glued to a borosilicate glass capillary tip; (c, d) projections on the (a, b) and (a, c) planes of the structure determined by crystallographic analysis of the synchrotron X-ray diffraction data collected from such single crystals (From Popov et al. 2009)

The general shape of the starch nanocrystals previously mentioned and obtained by mild acid hydrolysis of waxy maize starch granules can also be described by comparison to the unit cell of the A allomorph. The base-plane projection of the nanoplatelets would be homothetic to the (a, b) plane of the monoclinic unit cell (Fig. 3.2f). The experimental acute angles measured from the nanocrystals (60–65°) are close to the 56.6° acute angle complementary to the γ angle of the unit cell (Fig. 3.2f) (Putaux et al. 2003; Pérez and Bertoft 2010).

2 Phase Transitions and Crystalline Structure of Starch

In most uses of starch, the granule is disrupted, in particular during hydrothermal treatments like extrusion, cooking, high-pressure treatments, etc. Many reviews have been published on hydrothermal treatments of starch and related phenomena such as gelatinization, melting, gelation, retrogradation, structural transition, and amylose complexing (Buléon and Colonna 2007; Colonna and Buléon 2010). Once the granule has been disrupted, the structure and properties of the resulting material are governed by the way amylose and amylopectin rearrange during cooling, which strongly depends on temperature, shear, and water content. Starch glassy materials are obtained by melting of starch in processing conditions yielding a final water content below the glass transition domain. On the contrary, as soon as a sufficient amount of water is present, amylose and amylopectin recrystallize mostly into B type. This starch retrogradation is the basis of many food uses of starch as gelling agent. During hydrothermal treatments, amylose can form semicrystalline single helical inclusion complexes with small molecules (like lipids, alcohols, flavor compounds, and various small hydrophobic molecules). The helical conformation, crystalline packing, and intra- or inter-helical inclusion depend on the complexing molecules, and these parameters govern the thermal stability of the complex and the way the guest molecule is released. Starch melting/retrogradation and starch complexing by lipids and alcohols are well documented (see, e.g., Colonna and Buléon 2010). The following section focuses, on one hand, on local order and orientation in amorphous starch, as a function of the thermomechanical history, and, on the other hand, on the most recent structures determined for amylose complexes.

2.1 Local Order and Orientation in Amorphous and Semicrystalline Starch Materials

Amorphous starch materials present no long-range ordering (i.e., crystallinity), but locally keep some organization and orientation which depend on the process used for the starch structure disruption and the thermomechanical history of the material. This local order can be studied by techniques such as solid-state NMR for short-range order or two-dimensional WAXS and synchrotron radiation (SR) infrared microspectroscopy for local orientation. Paris and coworkers, in the early 2000s, used ¹³C CP-MAS solid-state NMR for the characterization of the different local-range orderings present in amorphous/semicrystalline starch, amylose, or amylopectin, prepared by different techniques like casting, freeze-drying, or 2-propanol precipitation from solution (Paris et al. 2001). All samples were amorphous except reprecipitated ones. A major part of the project consisted in providing a reliable spectral decomposition of the C1 resonance spectrum, which revealed the existence of four to five main types of α(1,4) linkages which were quantified (Paris et al. 2001). Thus, accepting the classical correlation (Gidley and Bociek 1988) between the isotropic chemical shift and the conformational angles (|Φ|+|Ψ|), a specific α(1,4) conformation was associated to each type. The conformations used for the decomposition were extracted from a refined literature analysis on NMR of α-glucans and transposed to the determination of local order within the amorphous samples studied. The bands stemming from the spectral decomposition have the following chemical shift: A (103.4−103.2 ppm), B (102.9 ppm), C (101.4−100.4 ppm), D (98.6−97.1 ppm), and E (94.5−94.4 ppm). As an example, the C1 part of the NMR spectra of casted amylose and amylopectin reprecipitated in 2-propanol, and their corresponding decomposition into four and five components, respectively, is presented in Fig. 3.6c, d. Full spectra are shown in Fig.3.6a, b for reference. A, C, D, and E peaks are present for all types of starchy materials, while the B peak is only present in 2-propanol-reprecipitated samples. The A peak is the most intense (around 50 % of total area) for all samples, except for 2-propanol-reprecipitated samples. Its chemical shift (103.3 ppm) is similar to the classical range of V-type spectra (Horii et al. 1987; Gidley and Bociek 1988) and single sixfold helical conformation. The B peak, which is only observed for 2-propanol-reprecipitated samples presenting a WAXS diagram characteristic of a mixture of amylose V_2-propanol complex and B type, was attributed to V_H-type single helices, since at that time, amylose conformation present in amylose-2-propanol complexes was assumed to be a single sixfold helix similar to that constituting the V_H structure. Considering the new structure recently established for the V_2-propanol crystalline complex (Nishiyama et al. 2010), this B line is more probably associated to a sevenfold single helix. The remaining conformations (C, D, E) are the most sample sensitive and subject to more variation in chemical shift. The C peak appears in the range of double helical conformation and could be associated to paracrystalline bundles stemming from uncompleted melting of starch structure or very local retrogradation. The D peak has been observed in α-cyclodextrins with twisted glycosidic linkage (Gidley and Bociek 1988) and tentatively correlated to α(1,6) linkages by Morgan et al. (1995). The chemical shift and C/D difference observed between linear (amylose) and branched (amylopectin or starch) samples suggest that D conformations are more sensitive to the presence of α(1,6) and could be closer to branching points than the C ones. E line, by far the weakest (2–8 % depending on the sample), is related to energetically unfavorable conformations such as constrained linkages.

Fig. 3.6

¹³C CP-MAS NMR spectra of (a) amorphous amylose (obtained by hot casting) and (b) 2-propanol precipitated amylose. Enlarged C1 resonance for 2-propanol-precipitated (c) and casted (d) amylose, with corresponding deconvolution of the C1 region (From Paris et al. 2001)

Each type of amorphous starchy material has a specific signature with different proportions of each conformation, bringing out their thermomechanical history. Thus, the different local conformations present in amorphous amylose, amylopectin, or starch are directly related to the preparation conditions. For example, for casted samples twisted and/or constrained conformations are less present than in freeze-dried corresponding samples since chains, heated in the presence of water, may relax before complete removal of water. On contrary, during freeze-drying the conformations are frozen before removal of water, and more constrained conformations are created. More generally, E conformations are favored by more drastic methods of preparation. Rehydration and plasticization by water lead to a general decrease in line width and therefore to a more homogeneous distribution of conformation and show a higher sensitivity of more constrained conformations (D, E lines) and freeze-dried samples. Recently, a similar spectral decomposition was successfully performed on cross-linked starch (Thérien-Aubin et al. 2007).

The presence of a local organization in amorphous starch induces very important changes on the properties and phase transition of starch materials. An example is given by recent works showing very efficient shape memory properties of extruded starch materials (Véchambre et al. 2010). Shape memory corresponds to the ability for materials to recover their original shape after being deformed into a temporary shape. In the case of “usual” thermoplastic shape memory materials, the effect is attributed to the presence of two types of domain: one flexible and one rigid. The rigid domains determine the initial shape, while the more flexible domains become oriented in the temporary shape. This orientation remains stable in “constrained” materials, below the glass transition temperature T _g. However, at a temperature above T _g, these domains relax, which results in the recovery of the initial macroscopic shape of the unconstrained sample.

The orientation and physical cross-links and their relation with thermomechanical history have been studied in amorphous extruded starch materials. Such materials are completely transparent when observed using light microscopy, but with polarized light microscopy, constrained samples show typical and clearly visible birefringence fringes, contrary to unconstrained samples which do not present any specific birefringence. This shows that the constrained samples are anisotropic due to residual stress but that anisotropy disappears after shape recovery, due to the relaxation of residual stress. Despite strong birefringence under polarized light, no crystal melting is evidenced by DSC, and WAXS diagrams present no clear diffraction peaks but only a broad amorphous scattering band. The scattering band maximum corresponds to a repeat distance of about 0.5 nm. This is normally considered to arise from the van der Waals (VDW) contact of nonbonded atoms (VDW spacings) (Miller et al. 1984).

WAXS was used to determine the orientation within such constrained samples, usually extruded rods or thermomolded barrels, after uniaxial deformation (Véchambre et al., 2010). The WAXS diagrams were recorded in transmission mode using a two-dimensional detector from specimens lying with their long axis lying parallel to the vertical axis of the detector (Fig. 3.7a). Constrained samples exhibit periodic scattering intensity changes on the maximum of the broad amorphous scattering band, with the azimuthal angle (Fig. 3.7b). The orientation was determined by azimuthal integration between 0.680 and 0.386 nm corresponding to the amorphous scattering band. The presence of two maxima at 90 and 270° shows that orientation is parallel to the axis of deformation of the sample (Fig. 3.7c). It has been shown that orientation increases with the level of deformation. Such orientation is stable as long as moisture content keeps the sample in the glassy state (below T _g). No periodic scattered intensity was observed for the relaxed samples as for unconstrained samples.

Fig. 3.7

Orientation diagrams from WAXS spectra: (a) WAXS setup used and positioning of the specimen, (b) azimuthal scan of the amorphous broad scattering from the transversal sample, (c) evolution of orientation as a function of the deformation rate

Similar measurements revealed that the orientation decreases when the temperature of deformation increases. This decrease in orientation with increasing deformation temperature was attributed to the greater ability of chains to rapidly reorganize at higher temperatures between deformation temperature and T _g at the end of the deformation process (Véchambre et al. 2011). The recovery stress is a collateral effect of shape memory which happens when the sample is stimulated and maintained in its initial shape. It has been shown that the maximum recovery stress increases almost linearly with increasing orientation. This means that the driving force for the starch shape memory is the behavior of oriented amorphous segments that act as “entropic springs” above T _g. From these results, it is expected that amylose, to which the linear structure induces a better ability to be oriented than amylopectin, gives the higher recovery stress.

IR spectroscopy was used to look at the local orientation in constrained samples and at the chemical groups potentially involved. The intensity of the two peaks at 1,023 and 1,049 cm⁻¹, in the C-C, C-O, and C-H stretching and C-O-H bending energy range, has been widely used to determine the amount of ordered and amorphous starch, respectively (Smits et al. 1998; van Soest et al. 1995). The band at 1,153 cm⁻¹ was assigned by Bernazzani and coworkers (2000) to vibrations in the environment of ordered single helices. Finally, the band at 987 cm⁻¹ is very sensitive to hydration and can shift from 987 to 1,003 cm⁻¹ depending on the water content. It was attributed to intramolecular hydrogen bonding of the hydroxyl group at C-6. It was also shown to develop with molecular ordering (Sevenou et al. 2002; Capron et al. 2007). Synchrotron IR microspectroscopy, which yields higher signal-to-noise ratio, smaller recording time, and easier dichroism experiments (the synchrotron beam being naturally polarized), was applied to constrained and unconstrained samples (Véchambre et al., 2010). With this technique, spectra were collected on a surface of 12 μm × 12 μm, and a mapping of samples of about 100 μm² was performed with the synchrotron IR beam parallel (PAR) or perpendicular (PER) to the axis of deformation of the sample (Fig. 3.8a). Small but significant differences, demonstrated by a statistical analysis (PCA), were observed between PAR and PER spectra contrary to unconstrained samples, which confirms the presence of a specific orientation in the constrained samples. The PCA analysis was also used to identify the main FTIR absorbance bands associate to PAR and PER spectra (Fig. 3.8b). From band attribution presented previously, molecular orientation in deformed sample is clearly related to amorphous domains. On the contrary, all the bands assigned to higher order and/or crystallinity are linked to spectra recorded in PER mode. It shows that local orientation is more present in amorphous domains and is limited by ordered domains. The presence of bands linked to hydrogen bonds and α(1,4) linkage could evidence the presence of local order under the form of helical fragments which may be stabilized by intramolecular H bonds and play as springs during the recovery process.

Fig. 3.8

Synchrotron infrared microspectroscopy of shape memory starch-based materials: (a) orientation of the specimen for dichroism experiments, (b) first principal component loading vector determined by the PCA and associated infrared absorptions

2.2 Starch Interactions with Small Molecules and Amylose Complexing

Amylose shows the unique feature to form complexes with a large variety of molecules. When heated in the presence of starch and water, monoacyl lipids and emulsifiers as well as smaller ligands such as alcohols or flavor compounds are able to induce the formation of left-handed amylose single helices. “V amylose” is the generic term used to describe amyloses co-crystallized with such compounds. The resulting helical conformation and crystalline packing depend on the nature of the ligands and conditions for the formation of the complex. In particular, the specific interaction of amylose with lipids or aroma compounds has a strong impact on food quality (Heinemann et al. 2005; Conde-Petit et al. 2006), while the interaction between amylose and some plasticizers impacts the mechanical properties of starch-based materials. A foremost example is the use of fatty acids and monoglycerides as anti-staling agents in bread and biscuits. The incorporation of such additives in the dough induces a slower crystallization (retrogradation) of the amylopectin fraction and therefore retards the staling of bread (Morrison et al. 1993a). Experimental evidence that such complexes might be present in native starches in the amorphous state was given by Morrison et al. (1993b). In that case, amylose chains are involved in either isolated single helices or involved in crystallites too small to be resolved by WAXS.

The crystalline structure of amylose complexes was widely studied by crystallizing amylose from aqueous solutions in the presence of a large variety of small organic and inorganic molecules. Amylose forms inclusion compounds with distinct WAXS signatures that depend on the complexing agent (Takeo and Kuge 1969; Tomasik and Schilling 1998a, b; Putseys et al. 2010). Several generic families of V-amylose have been described (Buléon et al. 2007; Putaux et al. 2011a). Although the knowledge of the molecular structure is important to locate the guest molecules and understand how they are entrapped within the crystal lattice, only a small number of structures have been resolved by crystallographic approaches, and several models are still hypothetical. Besides many studies carried out on polycrystalline powders or oriented fibers (Zobel et al. 1967), electron diffraction of lamellar single crystals, in particular when performed in frozen-solvated conditions (Booy et al. 1979), combined with molecular modeling, played a crucial role in the determination of a number of structures (Yamashita et al. 1973; Helbert 1994; Brisson et al. 1991; Nishiyama et al. 2010).

The so-called V_H type, obtained when amylose is crystallized in the presence of fatty acids (Godet et al. 1993) and some alcohols (Brisson et al. 1991), is well documented. The lamellar single crystals exhibit a characteristic hexagonal shape, and the unit cell has been described by the hexagonal packing of left-handed sixfold single helices (Brisson et al. 1991). In the case of complexes formed with lipids, the aliphatic segment of the guest molecules has been shown to be located inside the helical cavity, while the polar head remained outside (Godet et al. 1993). Other sixfold helical systems have also been proposed to describe complexes formed in the presence of n-butanol (Rundle and Edwards 1943; Helbert and Chanzy 1994), dimethyl sulfoxide^. (French and Zobel 1967), and glycerol (Hulleman et al. 1996).

Bear suggested that bulkier guest molecules should be accommodated in a larger helix with more residues per turn (Bear 1944). By comparing the volume of the unit cells of the complexes formed with tert-butanol and n-butanol, Zaslow obtained a ratio of about 7:6 in favor of the existence of a sevenfold helix for the V _tert-butanol complex (Zaslow 1963), a conclusion supported by Yamashita and Hirai (1966). Later on, based on the electron diffraction data of V_2-propanol single crystals, Buléon et al. (1990) proposed an alternate model: the orthorhombic unit cell would contain sixfold helices separated by a layer of guest molecules. Solving this dilemma was important as it had been observed that a large number of complexing agents (thymol, geraniol, menthone, etc.) promoted the formation of crystals exhibiting similar diffraction patterns (Helbert 1994; Nuessli et al. 2003). Recently, Nishiyama et al. (2010) refined the structure of V_2-propanol crystals using electron diffraction data combined with conformational and packing energy analyses. They proposed an original unit cell based on sevenfold V-amylose helices whose packing was described using two regularly alternating motifs: one is nearly tetragonal and holds a cavity filled with 2-propanol and water molecules, whereas the other corresponds to four close-packed helices with only water molecules located in the inter-helical space (Fig. 3.9a–c).

Fig. 3.9

(a, d) TEM images of amylose V_2-propanol and V_1-naphthol lamellar single crystals, respectively; (b, e) corresponding low temperature base-plane electron diffraction patterns recorded from frozen-wet crystals; (c) axial projection of the molecular model of the amylose V_2-propanol complex. For clarity, the hydrogen atoms have been omitted. The 2-propanol guest molecules are located inside and between the sevenfold amylose single helices (From Nishiyama et al. 2010). (f) Averaged high-resolution TEM lattice image of the crystal structure of amylose V_1-naphthol. The projection of the network of eightfold amylose single helices has been superimposed (From Cardoso et al. 2007)

An eightfold amylose single helix has been proposed to occur in the presence of three complexing molecules: 1-naphthol (Yamashita and Monobe 1971; Helbert 1994), quinoline (Helbert 1994), and salicylic acid (Oguchi et al. 1998). The helices would be organized in a tetragonal unit cell. So far, the strongest experimental evidence of an eightfold helix has been provided by Cardoso et al. (2007) who published a high-resolution transmission electron microscopy (TEM) image of the crystal lattice viewed along the helical axis and showing a repeating ring motif of 8 subunits (Fig. 3.9d-f).

Sequential washing of powdered complexes with ethanol allowed to probe intra- and inter-helical inclusions. Such approach was recently applied to the different crystalline types of amylose alcohol complexes and to aromas like linalool and menthone (Rondeau-Mouro et al. 2004; Biais et al. 2006). High-resolution magic angle spinning (HR-MAS) NMR spectra were also used to compare the chemical shifts of free and bound aroma molecules and allowed to propose hydrogen bonding schemes in amylose complexes. Moreover, free aroma was shown to be completely removed by ethanol washing. Using CP-MAS NMR and X-ray scattering experiments, it was demonstrated that the V_2-propanol type was retained for linalool whatever the treatment used. On the contrary, it shifts toward V_H type for menthone after ethanol washing before the desorption step, reflecting the disappearance of inter-helical associations between menthone and amylose. The stability of the complex prepared with linalool shows that this ligand is more strongly linked to amylose helices. The discrepancies observed in the chemical shifts attributed to carbons C1 and C4 in CP-MAS NMR spectra of V_2-propanol and V_H forms were attributed either to a deformation of the single helix (with possible inclusion of the ligand inside) or to the presence of the ligand between helices (only water molecules are present in the V_H form). As described above, it is now more probably related to the sixfold and sevenfold nature of amylose helices in V_H and V_2-propanol complexes, respectively.

The formation of amylose complexes during in vitro enzymatic synthesis of amylose by phosphorylase in the presence of fatty acids or polymers like polyethers, polylactic acid, polycarbonate, or chemically modified cellulose has also been reported. This direct wrapping of amylose chains around the complexing molecule has been described as “vine-twining polymerization” (Kadokawa et al. 2002) or “parallel enzymatic polymerization system” (Kaneko et al. 2008).

3 In Vitro Enzymatic Synthesis of Starch Building Blocks and Biomimetic Systems

Since the complete separation of amylose and amylopectin from native starch is highly difficult, in vitro enzymatic synthesis of linear or branched α(1,4)-linked glucans has been significantly investigated during the last years, in order to mimic starch biosynthesis and, in particular, to better understand how linear chains self-associate.

3.1 In Vitro Synthesis of Amylose

Three classes of enzymes (glycosyltransferases, glycoside phosphorylases, and transglycosylases) are naturally able to polymerize α(1,4)-linked residues (Fig. 3.10). However, the high-yield in vitro production of amylose based on the use of Leloir glycosyltransferases (GTs) like granule-bound starch synthases (GBSS) cannot be envisaged as these enzymes use an activated sugar nucleotide donor (ADP-glucose) that is too expensive and are not active when they are not embedded in the native starch granule (Ball and Morell 2003). Thus, only glycoside phosphorylases and transglycosylases have been extensively used for in vitro amylose synthesis.

Fig. 3.10

The various enzymatic routes of α(1,4)-glycan synthesis

Bacterial, animal, or plant α-glucan phosphorylases (GPs), which are classified in the GT35 family because of their structural similarity with real GTs, naturally catalyze the breakdown of an α(1,4) glucosidic linkage from amylose or glycogen through a retaining mechanism, with concomitant phosphate glycosylation, to yield a α-d-glucose-1-phosphate (G-1-P) and a shorter α-glucan chain. These enzymes also perform reverse phosphorolysis to form a glycosidic bond between the glycosyl unit originating from the glycosyl phosphate, which acts as the sugar donor, and a carbohydrate acceptor. In this so-called synthetic reaction, the smallest glycosyl acceptor, or primer, is maltotetraose. Phosphorylase-catalyzed enzymatic polymerization is the only method used for production of amylose with low dispersity (${\overline{M}}_w/{\overline{M}}_n<1.2$) and with a weight-average molar mass that can be easily controlled by varying the G-1-P/primer ratio (Kadokawa 2012). However, G-1-P is too expensive for industrial synthesis of amylose, even if two-step amylose synthesis processes were developed, combining actions of phosphorylases to generate G-1-P from sucrose or cellobiose (Ohdan et al. 2006).

Bacterial amylosucrases are thus a highly attractive alternative to mimic in vitro amylose polymerization and to get knowledge on chain self-association during starch synthesis or processing. Indeed, they are the only known enzymes that catalyze, without any primer, the synthesis of an α(1,4)-linked glucan from sucrose, a cheap agroresource used as glycosyl donor, with the concomitant release of fructose. These transglycosylases, classified in family 13 of glycoside hydrolases (André et al. 2010), use a retaining non-processive mechanism to produce an amylose-like polymer of which the average chain length, dispersity, morphology, crystallinity, and chain length involved in crystals can simply be modulated by varying initial sucrose concentration and reaction time (Potocki-Véronèse et al. 2005).

The longest chains (${\overline{\mathrm{DP}}}_w=58$, ${\overline{M}}_w/{\overline{M}}_n=3.0$) produced by Neisseria polysaccharea amylosucrase (NpAS) from 100 mM sucrose entangle into networks similar to those observed by TEM for amylose gels (Fig. 3.11a). These networks contain clusters of semicrystalline 10–15 nm elementary units, formed by association of molecules into parallel double helices, linked by amorphous sections containing loosely organized chains. A synchrotron SAXS study of the amylose conformation during synthesis in such conditions revealed that at an early stage of polymerization, amylose consists of a mixture of wormlike chains and double helical cylindrical structures (Roblin et al. 2013). In a second stage, individual double helices pack into clusters before crystallizing and precipitating. All the dimensions determined for wormlike chains and cylindrical conformations at different times of synthesis were in very good agreement with structural features usually observed on gels of amylose extracted from starch. Furthermore, DSC and WAXS analyses revealed that these short chains arrange into independent B-type crystalline domains, as did the samples produced from 600 mM (${\overline{\mathrm{DP}}}_w=35$, ${\overline{M}}_w/{\overline{M}}_n=2.3$) (Fig. 3.11b). Nevertheless, in these last conditions, the chains suddenly precipitate as polycrystalline aggregates, when the limit chain length and concentration are reached during synthesis (Potocki-Véronèse et al. 2005). Remarkably, this behavior is generally also encountered for plant amylose chains, the longest ones yielding B-type networks upon retrogradation at temperatures between 4 and 30 °C, while shorter chains form polycrystalline precipitates (Buléon et al. 1984).

Fig. 3.11

In situ crystallization of amylose during enzymatic synthesis by amylosucrase from sucrose: (a) gel network (100 mM sucrose), (b) axialitic particles (600 mM sucrose) (From Potocki-Véronèse et al. 2005). (c) WAXS profiles showing the structural transition of the axialitic particles from B to A type after a heat-moisture treatment

In contrast, the crystallinity of the particles formed from the chains synthesized in vitro from 600 mM sucrose is exceptionally high (94 %) considering that they result from a self-aggregation process during enzymatic synthesis, without any optimization of the crystallization conditions (Potocki-Véronèse et al. 2005). The chain organization inside the particles is not known with precision. The analysis of the polarized optical micrographs suggests there is an axial symmetry, hence the term “axialites” used to describe the particles. Such a structure, if it is confirmed, has never been reported before for amylose, but may correspond to the axialitic organization of synthetic polymer chains (Encyclopedia of Polymer Science 1987). In addition, by simply heating these amylose axialites for 5 min at 90 °C, the crystallinity increases further by 30 %, without any change in their external shape. They should thus now be considered as a new standard for the determination of the relative crystallinity of starchy products (Potocki-Véronèse et al. 2005). This ability of short chains to self-associate into a highly organized supramolecular structure is helpful for understanding the crystallization stage during starch biosynthesis. Besides amylose gelation, the behavior of these particles during hydrothermal treatments like annealing in water excess or B- to A-type transition during heat moisture treatment (Fig. 3.11c), both without any change in external morphology, is also very similar to what happens with hydrothermal treatments of starch. The exceptional crystallinity of these particles and the simplicity of the synthesis process were exploited to obtain ¹³C-labeled amylose from 600 mM ¹³C-sucrose and to characterize the conformation of B-type amylose by high-resolution solid-state NMR (Rondeau-Mouro et al. 2006). The assignment of the complete B-type amylose spectrum and correlations between ¹³C-¹³C distances and the atomic positions in the three-dimensional model obtained by low-resolution electron and X-ray diffraction of crystalline B-type fibers (Imberty and Pérez 1988) were established.

The synthesis and fractionation conditions of maltooligosaccharides and short amylose chains produced by NpAS were also optimized in order to prepare fractions with a very low dispersity ($1.003<{\overline{M}}_w/{\overline{M}}_n<1.01$). Several fractions, with a ${\overline{\mathrm{DP}}}_w$ ranging from 10 to 40, were crystallized from dilute solutions in the presence of acetone vapors, resulting in needlelike A-type single crystals. The influence of the molecular parameters of the narrow fractions of amylose chains on the crystallization behavior and on the crystal morphology was investigated (Putaux et al. 2011b). Indeed, a ${\overline{\mathrm{DP}}}_w$ of 15–19 and a low dispersity were necessary to form crystals with a 5–10 μm length, such as those shown in Fig. 3.4b (Montesanti et al. 2010). As previously mentioned in this article, the data collected by synchrotron X-ray microdiffraction from such large model single crystals (Fig. 3.5a) were successfully analyzed using classical crystallography methods, resulting in a revised structure of A amylose (Popov et al. 2009).

3.2 Enzymatic Modification of Branched α(1,4)α(1,6)Glucans

If many advances were done these last years regarding amylose chain self-association, one cannot exclude the key role played by α(1,6) branching points of amylopectin in the elaboration of the starch granule ultrastructure. A large panel of branched starchy macromolecules was synthesized in vitro either by using NpAS, GP, or amylomaltase, alone or in combination with branching and/or debranching enzymes to mimic the activities required for starch biosynthesis. NpAS was also used to modify the structure of native glycogen by elongating the external chains of the macromolecule in the presence of sucrose. In fact, glycogen is particularly efficiently glucosylated by amylosucrase, the extension rate being governed by the relative concentration of sucrose and glycogen (Putaux et al. 2006). Using different sucrose/glycogen mass ratios, original glycodendrimers consisting of a glycogen core and a swollen amylose corona were formed within a few hours (Fig. 3.12a, c). Upon aging or drying, the amylose chains intertwined, resulting in the shrinking of the corona by the formation of crystallites and in a decrease of the particle diameter (Putaux et al. 2006). The resulting morphology depends on the initial sucrose/glycogen mass ratio. For a high ratio (around 340), the synthesized chains are long and reorganize in the form of a random organization of crystallites by a retrogradation process (Fig. 3.12b). For a low ratio (close to 1), small spikelike protrusions are seen on the surface (Fig. 3.12c). The number of glucosyl units added to glycogen external chains was estimated to be around 15. Consequently, the short chains intertwined in double helices can be expected to lie perpendicularly to the particle surface (Fig. 3.12d), like it is the case for the short branches of amylopectin in native starch granules. Such systems can thus be used to develop a biomimetic approach and reproduce in vitro the mechanisms involved in the biosynthesis of amylopectin and the structuring process of starch granules in plants. It shows that there is a close relationship between the length of the biosynthesized chains and the formation of a semicrystalline lamellar organization. Long chains may not result in the formation of lamellae, unless they are used by other biosynthetic enzymes. The influence of a debranching activity may also be studied by adding isoamylase to check the potential role of debranching enzymes during biosynthesis in amylopectin formation/crystallization.

Fig. 3.12

Dendritic particles prepared by extension of glycogen external chains by amylosucrase in the presence of sucrose: (a, c) cryo-TEM images of the particles obtained with initial sucrose/glycogen weight ratios of 342 and 1, respectively; (b, d) schematic representation of the growth and subsequent crystallization of the corona of neosynthesized linear glucan chains (LGC) around the initial glycogen particle (IGP) (From Putaux et al. 2006)

Hyperbranched alpha-glucans have been recently produced by jointly using amylosucrase and a branching enzyme (Grimaud et al. 2013) or phosphorylase and a branching enzyme (Ciric and Loos 2013). This is another approach of amylopectin biosynthesis, but without the use of debranching enzyme, the branching rate remains higher than that of amylopectin and the average chain length too low to properly mimic the typical arborescent branching pattern observed in amylopectin (Rolland-Sabaté et al. 2014). Thus, strictly, to date, amylopectin has never been biosynthesized in vitro or at least mimicked, whatever the enzymatic cocktails used. Two methods have been reported, described as “SP-GP-BE” (based on the use of sucrose phosphorylase (SP), α-glucan phosphorylase (GP), and branching enzyme (BE)) with sucrose and maltotetraose as substrates (Ohdan et al. 2006; Kajiura et al. 2011) and “IAM-BE-AM,” in which the branched linkages of starch were first hydrolyzed by an isoamylase (IAM) to produce short amylose chains that were assembled into spherical particles by a branching enzyme (BE) and an amylomaltase (AM) (Kajiura et al. 2008). Indeed, both methods resulted in the formation of glycogen-like molecules, rather than in a macromolecular architecture typical of amylopectin.