Electron crystallography of linear polysaccharides

JOURNAL OF ELECTRON MICROSCOPY TECHNIQUE 11:280-285 (1989)

Electron Crystallography of Linear Polysaccharides SERGE PfiREZ AND HENRI CHANZY Laboratoire de Physicochimie des Macromolicules, Znstitut National de la Recherche Agronomique, Nantes (S.P.), and Centre de Recherches sur les Macromoltkules Vegitales, C.N.R.S., Saint Martin d'H&es, (H.C.), France

KEY WORDS Electron diffraction, Molecular modeling, Polysaccharides, Crystal structure

ABSTRACT The present article illustrates how the use of electron diffraction data coupled with realistic molecular modeling can yield to unambiguous structural elucidation of linear polysaccharides in the crystalline state. The series of the original ab initio quantitative crystal structure analyses is presented, along with a description of the methods. Pertinent examples showing the unique ability and utility of electron crystallography to yield new structural insights for biologically interesting macromolecules are given.

INTRODUCTION

Carbohydrate molecules are ubiquitous in nature, where they have an essential role in promoting structure and texture or establishing storage. Also, their implications in biological recognition are now well established. Such a spectrum of roles for carbohydrates arises from their ability to generate a large number of highly specific structures from only a small number of monomeric residues. Optimized storage, as in the starch granule, or the making up of cell-wall structural elements seem to be achieved through native semicrys- talline arrangements of macromolecules. In order to understand the molecular basis of such native arrangements, the elucidation, at the atomic level, of the three-dimensional structures must be performed. The most important method for structure determination of crystalline polymers is x-ray fiber diffraction. It has been observed that linear polysaccharides prefer to be long helices rather than more complexly folded structures. After dissolution, one usually can produce samples in which such helical molecules are aligned with their long axis parallel (or antiparallel). Further lateral organization may occur, but rarely to the degree of a three-dimensionally ordered single crystal. Fibrous structures typically provide diffraction data of low resolution. Many helical polysaccharides yield no more than 50 independent x-ray reflections that can be used to determine the molecular geometry of the crystallographic asymmetric unit. Other shortcomings of this approach are the difficulties in assigning the unit-cell parameters and the ambiguities regarding the choice of the space group. Advances in biomolecular structures of polysaccharides have been more recent than parallel devlopments in the nucleic acid and the protein fields. Nevertheless, progress has been made in the development of methodologies, both on the experimental and on the computational levels. Insights into the fascinat- ing world of polysaccharide architectures have been gained. However, despite efforts of the fiber diffraction community, it has not always been possible to mimic classical crystallographic studies so as to reach credible solutions of structures using noncontroversial methods of elucidation. It is the aim of the present work to illustrate how the use of electron diffraction data

coupled with realistic molecular modeling can yield, in a quantitative fashion, to unambiguous structural elucidations.

METHODS Monosaccharides are the simplest building units. In

polysaccharides, most of the sugar residues are hex- oses, which exist in the pyranose form and belong primarily to the D series. In theory, the possibility of rotations about single bonds of the pyranose ring can generate a number of conformers. However, it has been established from x-ray structural studies and from nuclear magnetic resonance (NMR) studies that the pyranose ring in monosaccharides, oligosaccharides, and polysaccharides exists in the chair conformation. Theoretical studies also indicate that the chair is slightly flexible and that small alterations of the ring torsional angles (up to 10") and of the ring-bond angles (up to 3") are possible. When two monosaccharide units are joined, they are free to rotate about the interglyco- sidic junctions (Fig. 1). The resulting disaccharide can assume a number of different conformations. The torsional angles of rotation about the glycosidic linkage are designated by @ and V; in principle they can take any value between - 180" and 180". However, the range of values is limited due to steric reasons. Sets of limiting distances for different pairs of atoms to be used in the construction of steric maps have been proposed. On such maps appear the allowed and disal- lowed conformations for a particular disaccharide unit (Fig. 2). More precise information about the potential energy of a given conformation can be gained by using "empirical methods." The potential energy Etot is par- titioned into a number of discrete contributors:

Etot = E n b + EM, + E h b + Eexo + . . . where Enb is the nonbonded interaction energy; Etor is

Received May 20, 1988; accepted in revised form June 21, 1988. Address reprint requests to Dr. Serge Perez, Laboratoire de Physicochimie des

MacromolBcules, Institut National de la Recherche Agronomique, Rue de la araudibre. Nantes, F-44072, France.

0 1989 ALAN R. LISS. INC

ELECTRON CRYSTALLOGRAPHY OF POLYSACCHARIDES 281

Y /

"A -\ U

Fig. 1. Molecular representation of a disaccharide unit: 0-p- D-mannopyranosyl(l44)-a D-mannopyranose, along with the glycosidic torsion angles @ and V.

- 60

180

60

I I I 180 - 60 6 0 6

Fig. 2. Conformational analysis about the glycosidic torsion angles @ and .\lr of O-p-D-mannopyranosyl (1-+4)-a-D mannopyranosyl disaccharide, repeating unit of the mannan polysaccharide. External contours of the shaded area correspond to the 10 Kcal/mol contour, expressed relative to the calculated energy minimum. Helical param- eter: iso-n = 2 (---I, iso-n = 3 (AAA) iso-n = 4 (...I, with the sign associated to the appropriate helix chirality; iso-h contours (h = 0.4, 0.5 and 0.5135 nm). The indicates the (a, Y) conformation giving rise to the observed set of helical parameters.

the energy due to the torsional strain about glycosidic bonds, Ehb is the energy due the formation of hydrogen bond, and E,,, is the energy due to the exoanomeric effect. Detailed indications about these types of functions have been reported recently by Tvaroska and Perez (1986). When subjected to the constraints im- posed by the helical symmetry of macromolecular chains, equivalent monomeric units should occupy equivalent positions about the molecular axis. This is

achieved when the (@, q) angles are the same at every linkage; the secondary structure is described in terms of a set of helical parameters n and h, when n is the number of residues per turn of the helix, and h is the translation of the corresponding residue along the helix axis. In practice, these parameters are readily observable on the fiber diffraction pattern, where the spacing of the layer lines provides the pitch of the helical structure, whereas the axial projection of the repeating unit and the helix type are deduced from the position of the meridional reflections. The helical parameters are plotted as iso-n and iso-h contours on the same map as a function of torsional angles (@,TI (Fig. 2). Values of the torsional angles consistent with the observed parameters are found at the intersection of the corresponding iso-n and iso-h contours. Discrim- ination between possible solutions is performed on the basis of the magnitude of the potential energy. Provid- ing that the data set is of sufficient quality and/or the unit-cell dimensions and space group symmetry are well assigned, the final stage of elucidation involves a complete structural determination of the unit-cell con- tent. In order to do so, a linked-atom description similar to that reported by Smith and Arnott (1978) or Zugenmaier and Sarko (1980) may be used. Optimiza- tion procedures seek to fit observed and calculated structure amplitudes with simultaneous optimization of the nonbonded interactions and preservation of helix pitch and symmetry as well as ring closure. Inter- atomic energy functions, mimicking intermolecular nonbonded interactions are used extensively for this purpose (Williams, 1969).

Once dissolved and recrystallized, simple linear polysaccharides can yield single crystals. As with other polymer crystals, growth depends on proper nucleation followed by crystal growth. Usually, materials having a low molecular degree of polymerization (DP) and narrow molecular weight distribution give the best results. Crystallization is achieved from dilute solution by temperature modification or by the addition of nonsolvent. Crystallization temperatures as high as possible and compatible with the system under inves- tigation are preferred. Typical temperatures range from 100" to 2OO"C, and the experiments are performed in pressure vessels. Polysaccharides often crystallize with the incorporation of water or solvent molecules. In most cases, a well-defined morphology is obtained, the most popular being plate-like. These thin lamellae have lateral dimensions of several micrometers for only one to a few 10s of nanometers in thickness (the polymer chain axis lies normal to the lamellae surface). Crystalline domains of such dimensions are well suited to examination by transmission electron microscopy in both imaging and diffraction modes.

The recording of the diffraction is performed in two dimensions. The accelerating voltage is at least 100 kV; the wavelength of the electron is quite small. As a result, the radius of the Ewald diffraction sphere is quite large; it is possible to record many diffraction maxima simultaneously. As a rule, the polymer chain axes (c) are perpendicular (or nearly so) to the platelet face. Therefore, the hkO reflections can be readily collected from crystals aligned perpendicular to the

282 s. PEREZ AND H. CHANZY

I I

t ,"R.D,.TION

Fig. 3. Schematic representation of several electron-diffraction reciprocal planes which can be recorded by proper rotations about the reciprocal axis (taken from Guizard, 1981).

electron beam. Not only the a*, b*, and c* unit cell parameters are available, but also the two-dimensional symmetry of the base plane. Depending on the perfec- tion of the crystalline domains under study, information to a resolution of 0.1 nm, can be obtained. Because typical unit-cell dimensions of crystalline linear polysaccharides are 1-3 nm, one can expect to collect between 50 and 100 F(hk0) amplitudes. With the availability of accurately controllable electron micro- scope goniometer stages, insights into the third dimen- sion of the unit cell have become possible. Determina- tion of the remaining crystallographic parameters can be made after tilting the crystal around a* and b* (rotations about any of the reciprocal axis are also possible). Three-dimensional symmetry can be assigned by comparing the diagrams obtained when the crystals are rotated by an angle pa about either a* orb* (Fig. 3) (Guizard, 1981). Preliminary trials have been made to incorporate the intensities from higher-layer reflections into the data set. Despite the small number of layers that can be expected to contribute to scatter- ing from the tilted specimens, there are no detectable subsidiary maxima, as might be expected to occur from small crystallites. For lamellar single crystals containing no ((heavy" atoms, the kinematic approximation is assumed. Within such a scheme, structure amplitudes which are the square-root of the observed intensities can be used in a satisfactory manner. Nevertheless, some precautions and corrections may be applied.

The correction for beam damage by measurement of intensities a t various exposure times and their extrap- olation to zero time is certainly critical (Fig. 4). The small dimensions of natural polymer microcrystals introduce some complications. Diffraction maxima are only several silver grains in width, yielding a very granular diffractogram. The microcrystals are poor electrical conductors so the beam damage is rapid. Exposures must be made quickly, resulting in consid- erable noise, both in the background and in the diffraction maxima, Methods and computer programs for

Fig. 4. Influence of the exposure times on the relative intensities of (hk0) reflections in the high-temperature polymorph of dextran (taken from Guizard, 1981).

evaluating electron diffraction data films have become available (Miller et al., 1986).

RESULTS AND DISCUSSION The joint use of molecular modeling and electron

diffraction data has been invaluable in quantitative elucidations of the crystal and molecular structures of six linear polysaccharides: cellulose triacetate I1 (Roche et al., 1978), anhydrous nigeran (Perez et al., 1979), anhydrous dextran (Guizard et al., 1984), hydrated dextran (Guizard et al., 19851, mannan I (Chanzy et al., 19871, and A-amylose (Imberty et al., 1988). This set of structures was resolved over a 10-year period, during which most of the methodolog- ical aspects described above were established and used. Well-documented description of these structures can be found in the original reports. Some of these polysaccharide structures were solved using combined electron and x-ray diffraction techniques. In that process, it has been determined that the base-plane electron diffraction intensities constitute reliable data for crystal structure analysis, even for a structure containing water. Particularly good results were obtained when the diffraction intensities were corrected from beam damage.

The crystal and molecular structures of mannan I (Fig, 5) were determined by a constrained linked-atom least-squares refinement utilizing stereochemical constraints and intensities derived from electron diffraction. Mannan chains crystallize on an orthorhombic

ELECTRON CRYSTALLOGRAPHY OF POLYSACCHARIDES

D

C

0 6'

E

283

Fig. 5 . A A mannan I crystal; B: Its electron diffractogram properly oriented with respect to the crystal as in A; C: Digitized electron diffractogram as in B after noise removal; D: Projection of

the mannan I structure; E: The mannobiose repeat unit viewed along the mannan chain axis.

space lattice, the systematic absences being consistent with the space group P212121. A density of 1.57 re- quires four mannose residues per unit cell. Intensities

were measured from diffraction patterns produced by specimens arranged with c* either parallel to the electron beam or rotated about b* and a*. The best

284 S. PEREZ AND H. CHANZY

E

Fig. 6. A Typical crystal of high-temperature polymorph of dextran; insert: a corresponding electron diffraetogram with proper orientation showing the a* b* projection; B: As in the insert in A, but tilted by about +31" around b*; C: As in the insert in A, but tilted by

about -31" around b*; D: Projection of the dextran structure perpendicular to the chain direction; E: A molecule of dextran projected along the chain axis and its crystalline arrangement.

model using the base-plane data coupled with stereochemical refinement model yielded R = 0.24. Structure amplitudes derived from the tilted specimens were scaled to the (hkO) data by varying only the electron

scale factor. For the (h k k) data obtained at 31" and (h k h) data obtained at 41" value of R was 0.36 and 0.37, respectively. The systematic use of data derived from tilted samples is still a matter of contention. Additional

ELECTRON CRYSTALLOGRAPHY OF POLYSACCHARIDES 285

factors that may be inadequately accounted for in our methodology include: 1) differences in the interaction of a bent crystal with the electron beam as a function of tilt angle; 2) changes in the linear absorption coeffi- cient with orientation; 3) changes in the sampled surface area of the crystal with orientation; 4) volume fraction corresponding to each diffracted beam may not be the same; and 5) crystal damage effects may be more extensive in the data derived from tilted crystals.

In all the cases studied, electron-diffraction data provided highly reliable lattice constants, test for symmetry elements, and base-plane structure amplitudes. A very enlightening example was found in the course of elucidating the structure of the anhydrous form of dextran. It was observed that the base-plane diffraction pattern (a*, b*) displayed a mm symmetry with y* = 90" (Fig. 6). However, it was found that the diagrams obtained from a given crystal rotated clockwise showed different intensities compared with the diagrams obtained from the same crystal in a counter- clockwise tilting of the same magnitude. A calibration of the tilted and untilted diagrams yielded a monoclinic unit cell characterized by p = go", space group P21. Therefore, the a-b plane of the cell was not at right angles to the chain axis. Dextran was the first polysaccharide studied so far to exhibit such a feature. Be- cause the P-angle at 91.3" differs very little from that a t go", it is doubtful that its value would have been determined correctly from a fiber x-ray diagram, had the latter been available.

Another illustration was provided by the structural elucidation of A-amylose (Imberty et al., 19881, for which an approximate volume and a space-group symmetry were suggested by x-ray fiber (Wu and Sarko, 1977) and x-ray powder diffraction experiments. An orthorhombic-like unit cell, having a = b = c = 90" was highly probable. Molecular modelling coupled with experimental values of helical parameters indicated the presence of a double-helical structure. Electron diffraction data provided some hints about the symmetry elements. Throughout the experimentally accessi- ble reciprocal lattice, the electron diffractogram exhib- ited systematic absences of reflections with indices h + k + 1 = 2n + 1, indicating a body-centered lattice. It was rationalized that the possible orthorhombic space groups had too many requirements for accommodating two such chiral double helices in the unit cell. The only remaining solution was the face-centered monoclinic space group B2 (with the fiber axis c as the unique axis) after adequate transformation of the cell axis a, b, and c. Later, electron-diffraction data provided proof of monoclinic symmetry. By sequential tilting about an axis perpendicular to the crystal axis, some hkl and -hkl (in the orthorhombic referential) reflections intensities could be compared and were found to be unequal, contrary to expectations for an orthorhombic space group.

CONCLUSIONS Owing to the development of the predictive power of

molecular modeling, it is fair to say that, apart from

the knowledge of the primary structure, the only experimental data required for a crystal structure elucidation are accurate unit-cell parameters and an unambiguous determination of the space group. From this information follow all the structural details. This is a particularly easy task to solve in the case of polymers, where the number of packing parameters is restricted. Basically, the above-described methodology may be applied to a whole range of polymeric structures. Extension also can be made to the case of large oligomeric structures, which are not likely to yield crystals good enough for conventional crystallographic work (Henrissat et al., 1987). Providing that micron- sized crystals are obtained, a structural elucidation can be envisaged successfully. Other avenues also can be explored in which electron-diffraction data are used to index powder diffraction patterns accurately and help resolve partially overlapped lines into individual intensities.

REFERENCES

Chanzy, H., Perez, S., Miller, D.P., Paradosi, G., and Winter, W.T. (1987) An electron diffraction study of mannan I crystal and molecular structure. Macromolecules, 20:2407-2413.

Guizard, C. (1981) "Cristallisation et Polymorphisme des Glucanes lies (1-6). Etude de la Structure Tridimensionnelle des Cristaux par Microsocopie Eleetronique." DSc Thesis. Universite Scienti- fique et Medicale de Grenoble, France.

Guizard, C. , Chanzy, H., and Sarko, A. (1984) Molecular and crystal Structure of Dextrans. A Combined Electron and x-ray Diffraction Study. 1: The Anhydrous High Temperature Polymorph. Macromol- ecules, 17:lOO-107.

Guizard, C., Chanzy, H. and Sarko, A. (1985) The molecular and crystal structure of dextrans. A combined electron and x-ray diffraction study. 2: A low temperature, hydrated polymorph. J. Mol. Biol., 183:397-408.

Henrissat, B., PBrez, S., Tvaroska, I., and Winter, W.T. (1987) Mul- tidiscplinary approaches to the structures of model compounds for cellulose 11. American Chemical Society Advances Series: Solid State Characterization of Cellulose, Paper and Wood, 340:38-67.

Imberty, A., Chanzy, H., Perez, S., Buleon, A., and "ran, V. (1988) The double-helical nature of the crystalline part of a starch. J . Mol. Biol., 201:365-378.

Miller, D., Chanzy, H., and Paradossi, G. (1986) Evaluation of diffraction data from electron diffraction patterns of natural polymer microcrystals. J Physique, 46:lOl-107.

Perez, S., Roux, M., Revol, J.F., and Marchessault, R.H. (1979) Dehydration of nigeran crystals: Crystal structure and morpholog- ical aspects. J. Mol. Biol., 129:113-133.

Roche, E., Chanzy, H., Boudeulle, M., Marchessault, R.H., and Sundararajan, P. (1978) Three-dimensional crystalline structure of cellulose triacetate 11. Macromolecules, 11:86-98.

Smith, P.J.C., and Arnott, S. (1978) LALS: A linked-atom least- squares reciprocal space refinement system incorporating stereochemical restraints to supplement sparse diffraction data. Acta Crystallogr., A34:3-11.

Tvaroska, I., and Perez, S. (1986) Conformational energy calculations for oligosaccharides: A comparison of methods and a strategy of calculation. Carbohydr. Res., 149:389-410.

Williams, D.J. (1969) A method of calculating molecular crystal structures. Acta Crystallogr., A25:464-470.

Wu, H.C.H., and Sarko, A. (1977) The double-helical molecular structure of crystalline A-amylase. Carbohydr. Res., 61:27-40.

Zugenmaier, P., and Sarko, A. (1980) The variable virtual hand modelling techniques for solving polymer crystal structures. Amer- ican Chemical Society Symposium Series 141: Fiber Diffraction Methods, pp. 225-237.

Documents

Electron crystallography of linear polysaccharides