Buléon, 1998. Starch Granules Structure and Biosynthesis

International Journal of Biological Macromolecules

23 (1998) 85112

Mini review

Starch granules: structure and biosynthesis

A. Buleon a,*, P. Colonna a, V. Planchot a, S. Ball b

a INRA, BP 71627, 44316 Nantes Cedex 3, Franceb CNRS, UMR 111, Cite Scientifique, 59655 Villeneu6e dAscq, France

Received 5 January 1998; accepted 19 February 1998

Abstract

The emphasis of this review is on starch structure and its biosynthesis. Improvements in understanding have been brought aboutduring the last decade through the development of new physicochemical and biological techniques, leading to real scientificprogress. All this literature needs to be kept inside the general literature about biopolymers, despite some confusing results ordiscrepancies arising from the biological variability of starch. However, a coherent picture of starch over all the different structurallevels can be presented, in order to obtain some generalizations about its structure. In this review we will focus first on our presentunderstanding of the structures of amylose and amylopectin and their organization within the granule, and we will then giveinsights on the biosynthetic mechanisms explaining the biogenesis of starch in plants. 1998 Elsevier Science B.V. All rightsreserved.

Keywords: Biosynthesis; Chemical structure; Starch

1. Chemical structure: molecular andmacromolecular constituents

In most common types of cereal endospermstarches, the relative weight percentages of amy-lose and amylopectin range between 72 and 82%amylopectin, and 18 and 33% amylose (Table 1).However, some mutant genotypes of maize (Zeamays), barley (Hordeum 6ulgare), and rice (Oryzasati6a) contain as much as 70%, amylose whereasother genotypes, called waxy, contain less than 1%(maize, barley, rice, sorghum). Recently, a waxywheat (Triticum aesti6um) starch has been ob-tained by classical breeding [1].

1.1. Amylose

Amylose is defined as a linear molecule of (14) linked a-D-glucopyranosyl units, but it is todaywell established that some molecules are slightlybranched by (16)-a-linkages.

The oldest criteria for linearity consisted in thesusceptibility of the molecule to complete hydroly-sis by b-amylase. This enzyme splits the (14)bonds from the non-reducing end of a chain re-leasing b-maltosyl units, but cannot cleave the(16) bonds. When degraded by pure b-amylase,linear macromolecules are completely convertedinto maltose, whereas branched chains give alsoone b-limit dextrin consisting of the remaininginner core polysaccharide structure with its outerchains recessed [2].

* Corresponding author. Tel.: 33 2 40675000; fax: 33 240675043; e-mail: [email protected]

0141-8130:98:$19.00 1998 Elsevier Science B.V. All rights reserved.

PII S0141-8130(98)00040-3

A. Buleon et al. : International Journal of Biological Macromolecules 23 (1998) 8511286

Table 1Number-average MW (M( n), weight-average MW (M( W) ofpotato and wheat amyloses

M( n105 ReferencesStarch M( W105

10.3Potato [2] 7.0 [5]

1.3 [6]9.04.6 8.7 [7]

2.1Wheat [2] 5.1 [5]

0.6 3.9 [6]2.8 5.8 [7]

amylose branching involves measuring the volumeoccupied by the polymer in solution. This volumemay be represented by the radius of gyration, R( G,which is dependent upon the molecular weight(MW). The molecular weight dependence of R( Gcan be described by the power law, RiMin,where the exponent n is a kind of MarkHouwinkcoefficient, and Mi and Ri are, respectively, themolecular weight and the radius of the componenti of a series of particles of the same architecturebut different molecular weights.

The coupling of HPSEC (high-performance size-exclusion chromatography) with a detector basedupon light scattering or viscosimetry leads to RiMin relations over the whole range of fractionatedmolecules Mi. The purpose of this approach is todetermine up to which point amylose chains insolution behave differently from strictly linearchains. Amylose fractions extracted by aqueousleaching of corn starch using 5C steps over atemperature range of 6595C [5] were studied bythe aforementioned techniques. Molecular weightdependences of macromolecular sizes were consis-tent with the behaviour of expanded chains ingood solvent. Indeed, the hydrodynamic coeffi-cient n was estimated within the 0.60.7 range forthe RGMn relation, suggesting therefore an ex-tended linear random coil in a large range of MW(31059106 g:mol).

We therefore conclude that the presence ofbranches does not alter significantly the solutionbehaviour of amylose chains, which remains iden-tical to that of strictly linear chains.

Measurements of both molecular weight distri-butions and average molecular weights have beenperformed on a large number of starch samples ofdifferent botanical origins. In contrast to proteinswhich are genetically coded, polysaccharides occurwith a molecular weight distribution, usually rep-resented by the average molecular weights: num-ber-average MW (M( n), weight-average MW (M( W),or z-average MW (M( z). Important discrepanciesare observed in the literature due to: (i) the biolog-ical origin of amylose, leading to uncontrolledvariations in the biosynthesis mechanisms; and (ii)molecular degradation occurring during amylosefractionation [6]. The different values (Table 2)quoted in the literature for potato and wheatamyloses illustrate these variations. Intrinsic vis-cosity remains a basic technique to calculate theviscosity-average MW (Mv), using the MarkHouwink coefficients measured on different sol-vents [2].

Takeda et al. [3] developed an extensive study ofamylose fine structure, based upon the successiveuse of b-amylase and isomylase. No effectivemethods for the separation of linear and branchedamyloses are known, so all results concerningamylose branching have been obtained by assum-ing the existence of two quite distinct populations,one strictly linear and the second one character-ized by a 40% b-amylolysis limit [3]. This lowvalue suggests that the branch linkages are fre-quently located rather near the reducing terminalend and:or they have multiple branched sidechains. b-Limit dextrins of branched amylosesalso show properties (iodine binding capacity,molecular weight) close to those of the corre-sponding original amyloses, while remaining com-pletely different from those of amylopectin. Thisobservation confirms the relevance of the distinc-tion between amylose and amylopectin. Indeed, nostructural continuum is observed between thesetwo types of a-glucans.

The branched molecule amounts range from 25to 55% on a molecular basis [3,4]. Using fraction-ated amyloses, this level of branching was shownto increase continuously as a function of themolecular weight [2]. The average number of sec-ondary chains attached through occasional (16)branch points to the branched molecules are in thewide range 418. Hizukuri et al. [4] found anaverage of two to eight branch points permolecule, the side chains from four to over 100glucosyl units in length. By further fractionationof b-limit dextrins, they concluded that branchedamylose has some tiny clusters of short chainsrather than a pure comb structure. The majordrawback of the aforementioned approach is togive averaged information.

Another useful approach to this problem of

A. Buleon et al. : International Journal of Biological Macromolecules 23 (1998) 85112 87

Table 2Radius of gyration, R( G, and weight-average MW (M( W) ofmaize and wheat amylopectins

Source M( W106R( G (nm) References

10Maize [2]10560 19.5 [13]

22 [15]229333260 [34]10 [35]10205 [36]

13 60 [37]75100205 [38]

160Potato 195 [2]47 [37]14444 [38]

195 1436 [39]

[15]. Using a specific optimization algorithm, ex-perimental molecular weight distributions (MWD)obtained by SECMALLS offers the possibility offitting them to mathematical MWD such as nor-mal distribution, most probable (MP) distribu-tion, and lognormal (LN) distribution. Goodagreements were obtained using either a sum ofoverlapping Gaussian curves [13] or a most prob-able model [5].

It is interesting to calculate the number of amy-lose chains present inside a single starch granule.Assuming an amylose content of 25% and a gran-ule density of 1.5, with an average relative molecu-lar weight of 500000, for 20-mm-diameter granulesthe number could be 1.8109.

Flexibility of polymeric chains is considered onthe basis of the dimensionless quantity C (thecharacteristic ratio), generally defined as the ratiobetween the dimensions of real and freely jointedchains. An equivalent characteristic is the persis-tence length a, defined as the average projection ofan infinitely long chain on the initial tangent ofthe chain. Amylose chains in solution (a1.71nm) are more flexible [5] than those of modifiedcellulose (cellulose diacetate a4.87.2 nm; car-boxymethyl cellulose a8.012.0 nm), but stifferthan those of pullulan (a1.21.9 nm). There-fore, all of these polysaccharides belong to theclass of loosely jointed polysaccharides, in contrastto stiff polysaccharides [16] such as xanthan (a310940 nm) and scleroglucan (a180930 nm).This conformational feature explains why amylosehas a low intrinsic viscosity value compared toother polysaccharides.

Another specific feature of interest concerningamylose is its capacity to bind iodine. The exis-tence of I3 and I5 was checked using Ramanspectral measurements and UV:Vis, coupled withtheoretical analyses. Yu et al. [17] demonstratedthat the four dominant polyiodide chains whichcoexist are longer species such as I3

9 , I3

11 , I3

13 andI3

15 . An interesting result is the demonstration ofthe absence of participation of I2 in the polyiodidechain.

1.2. Amylopectin

Amylopectin is the highly branched componentof starch: it is formed through chains of a-D-glu-copyranosyl residues linked together mainly by(14) linkages but with 56% of (16) bonds atthe branch points.

Polydispersity depends on the botanical originbut also on extraction procedures: high tempera-tures of leaching give values up to 10 [8]. Themolecular sizes of the branched molecules are1.53.0-fold higher than those of linear molecules.No significant differences are observed betweenaverage molecular sizes of cereal amyloses on theone hand, and those of tuber and rhizome starcheson the other hand.

Nowadays, amylose solutions can be easilycharacterized by size-exclusion chromatographycoupled on-line to multi-angle laser light scattering(SECMALLS). However, the hydrodynamic dif-ferences between amylose (RG 722 nm; [8]) andamylopectin (RG 2175 nm; [9]) are not in favourof a complete resolution between both compo-nents, and the use of combined detectors shouldprovide direct access to the complete molecularweight distribution of starch rather than eachcomponent separately.

Different solvents were proposed in the litera-ture: DMSO, aqueous alkaline solution, acidicconditions or DMAc:LiCl. Some authors reportedquantitative problems due to incomplete recoveryafter SEC [10]. However, this problem should notbe misinterpreted by reducing the difficulties tosolvent quality. The system solventsize exclusioncolumntemperature has to be considered as awhole. Its overall efficiency will be assessed con-sidering both chromatographic recovery and linearcalibration, demonstrating an exclusion effect.Efficient and reproducible fractionation of amy-lose and whole starch have been reported by Yuand Rollings [11,12], Fishman and Hoagland [13],Roger and Colonna [5,14] and Bello-Perez et al.


Amylopectin is a branched polysaccharide com-posed of hundreds of short (14)-a-glucanchains, which are interlinked by (16)-a-linkages.The multiplicity in branching is a common featureof both amylopectin and glycogen. The basic orga-nization of the chains is described in terms of theA, B and C chains as defined by Peat et al. [18].Thus, the outer chains (A) are glycosidically linkedat their potential reducing group through C6 of aglucose residue to an inner chain (B); such chainsare in turn defined as chains bearing other chainsas branches. The single C chain per moleculelikewise carries other chains as branches but con-tains the sole reducing terminal residue. The ratioof A-chains to B-chains, which is also referred toas the degree of multiple branching, is an impor-tant parameter. Using careful enzymatic experi-ments [19], the A:B ratio has been shown to be0.82.2 on a molar basis and 0.41.0 on a weightbasis. Nevertheless, a proportion of B-chains mustcarry more than one A-chain. Manners [20] re-ported A:B ratios ranging between 1.11.5, withA:B potato amylopectin at 1.2, whereas Hizukuri[21] assigns a 0.8 A:B ratio to potato amylopectin.This discrepancy appears to be due both to accu-mulated errors while harvesting experimental dataand to the very nature of the methods used. Thegeneral rule is that amylopectins have rather moreA-chains than B-chains, with ratios ranging from1.0:0 to 1.5:1. These values are consistent with thecluster and Meyers structures, but not with thoseof Haworth and Staudinger, as described in theManners review [20].

Very significant information is gained fromchain distributions profiles. The debranching en-zymes, isoamylase and pullulanase, specifically hy-drolyze the branch linkages and produce the shortlinear chains. SEC [21,22] and high-performanceanion-exchange chromatography with pulsed am-perometric detection [2224] are the two basictechniques used to estimate the chain length distri-butions. Amylopectin is built with three types ofchains: short chains, S, consisting of both outer(A) or inner (B) chains with a mean polymeriza-tion degree (DP) ranging between 14 and 18, inner(B) long chains, L, of DP 4555, and a fewB-chains of DP above 60. Hizukuri [21] evendetected a third peak as a weak shoulder on the Lpopulation for wheat, tapioca and tulip amy-lopectins. Differences related to the botanical spe-cies were mainly concerned with the L:S ratioexpressed on a molar basis. This ratio was esti-

mated at 5 for amylopectins from B-crystallinetype (potato) starch and at 810 for normal cerealamylopectins from A-crystalline type (normalgenotypes) granules.

Polymodal chain distributions have been re-ported by Hizukuri et al. [22,23] and Koizumi etal. [25]. On 11 amylopectins from different species,chromatograms exhibited periodic waves of DP12, except edible canna and yam amylopectins,where the period might be DP 15. Therefore, thechains were fractionated into four different frac-tions with DP ranging in the intervals 612, 1324, 2536 and more than 37. Amylopectins withhigh and low amounts of the DP 612 fractionrespectively, show A- and B-type X-ray diffrac-tions of starch granules respectively. Hence,Hizukuri et al. [4] suggested that these S-chainswith DP between 6 and 12 determined the starchcrystalline allomorph. The DP 612 fractionshould play an important role in the determinationof starch crystallite polymorphs. Nevertheless,Ong et al. [6] considered that there was limitedevidence to assess this influence of S-chains on theallomorphism since important variabilities of theseratios are observed inside the same botanical spe-cies. The involvement of the amylopectin modelremains, however, based upon correlation. Therelative amounts of each of the polymodal chainpopulations is genetically coded through the selec-tivity of the starch biosynthetic enzymes. In addi-tion, physical effects due to short chaincrystallization during synthesis also have to betaken into account.

Without doubt, the most important feature ofthis branched molecule is that S-chains are foundin discrete clusters. Nowadays, the cluster struc-ture concept originally proposed by Robin et al.[26] and by French [27] has found wide accep-tance. Several lines of evidence supporting thisstructure have accumulated (Fig. 1). The localiza-tion of the a-(16) linkages has been definedonly by using a sequential enzymatic method. Thisanalysis has been performed both on native amy-lopectins together with their corresponding b-limitdextrins [2830], lintners [26,31] and Naegeli [32]amylodextrins from granular starches. Thesechains are not linked together randomly, as in theMeyer and Bernfeld model but, rather, are orga-nized in a cluster structure. The two aforemen-tioned L-chain populations represent thebackbone supporting the S-chains-bearing clusters.It was for the first time refined into a statistical


Fig. 1. Diagrams of the molecular structure of amylopectin as proposed by (1) Haworth, (2) Staudinger, (3) Meyer and (4) Meyerredrawn as a cluster-type architecture.

model by Burchard and Thurn [35]. Branchingpoints are arranged in clusters which are not dis-tributed randomly throughout the macromolecule.An average chain length of 2023 units intercon-nects two clusters. The success of the cluster-typemodel proposed by Robin et al. [26,31] andFrench [27] is explained by its ability to accountfor the higher viscosity of amylopectin when com-pared to glycogen, for the involvement of amy-lopectin chain in crystallinity for the degradationpattern displayed by a-amylases. A-chains can beconsidered as the limiting factor defining crystal-lite thickness. Assuming classical models for A-and B-crystallites, involvement of S-chains in crys-tallite thickness leads to an helicoidal length ofabout 5.7 nm, with 16 glucosyl units, each onegiving a repeating distance of 0.35 nm per glucose.The lateral assembly of clusters could explain theformation of crystallites. However, the exactglobal conformation of amylopectin inside a gran-ule remains unknown.

Another enzymatic approach to the problem ofamylopectin structure is obtained through a-amy-lolysis. Zhu and Bertoft [33] observed a non-ran-dom hydrolysis of potato amylopectin with theformation of intermediate products presenting rel-ative molecular weights above 30000. These dex-trins resulted from the preferential hydrolysis ofintercluster linkages. However, dextrins containedmolecules with several clusters, therefore compli-cating their interpretation.

Alternative experimental strategies, such asthose based on the combined use of static and

dynamic light scattering [34,35], rely directly onanalyses performed on the native molecule withoutany molecular weight reduction. Different modelsfor amylopectin were examined through compari-son of calculated and experimentally determinedparticle scattering functions from combined staticand dynamic light-scattering studies. The dataprove that amylopectin is an heterogeneouslybranched polymer, in agreement with the pro-posals of French [27] and Robin et al. [26,31]. Formaize amylopectin, these authors concluded thateach L-chain has 1.4 clusters made of 3.22 S-chains on average, whereas Robin et al. [26] as-sumed exactly two clusters per L-chain. Thedistance between two clusters on the same B-chainis 22 glucosyl units on average. This modellingbased upon the cascade branching theory shouldbe renewed in the light of the last refinements ofHizukuri and his colleagues [2123].

This branched character based upon shortchains explains two main features of amylopectin.Thus, 100 g of amylopectin bind less than 1 mg ofiodine, giving a lmax around 540550 nm. The lowiodine binding capacity is based upon the forma-tion of an arrangement of four iodine atoms moreor less arranged linearly within the cavity of thehelix structure of 11 glucosyl units [36]. This fitsnicely with the abundant S-chain populations. Theb-amylolysis susceptibility of all amylopectin islow (5458%), whatever its botanical origin orpurification procedure [2].

Amylopectin has one of the largest relativemolecular weights (107109), but mostly in excess


of 108 (Table 2). However, there is a lack ofknowledge about the molecular weight distribu-tion, as all commercial chromatographic phasesare unable to fractionate this polymer on a molec-ular weight basis. For degraded amylopectins[11,12], the quantitative branching characteriza-tion via size-exclusion chromatography and on-line low-angle light scattering detection was ingood agreement with theoretical predictions ob-tained by branching modelling. Due to itsbranched character, amylopectins have low intrin-sic viscosity (120190 ml:g) despite huge molecu-lar weights.

As for amylose, it is interesting to calculate howmany amylopectin chains are present inside a sin-gle starch granule. Assuming an amylopectin con-tent of 75%, with a weight-average relativemolecular weight of 50000000, for a 20-mm di-ameter granule and a density of 1.5, the numbercould be 5.4107.

Atypical amylopectins have been observed fromhigh-amylose genotypes from maize [40] and pea[41], where they represent up to 20% of totalstarch. Called intermediate material because oftheir medium iodine binding capacities, they canbe distinguished from normal amylopectins by (i)longer L-chains with 515 glucosyl units more,and by (ii) low-molecular-weight amylopectin witha lower S:L ratio.

Despite intensive investigations, the fine struc-ture of the wild-type amylopectin clusters and therelative proportions of the side chains still need tobe determined. Here the main limitation is theinability of all actual analytical systems to fractionamylopectin according to size. Transmission elec-tron microscopy (TEM) [42] of starch moleculesdeposited from dilute solution and rotary shad-owed with platinum enables us to probe the het-erogeneity within the amylopectin family. Waxymaize amylopectin could be described as about28% circular space filling patches containingbranched clusters and 72% asymmetric linear con-taining branched. This structural problem stillcontinues to elude scientists, as the molecular ar-chitecture of this polymer is of importance toexplain its physicochemical properties. A detailedknowledge of the internal molecular structurewould help to describe mechanisms of crystalliza-tion, biosynthesis and organization of amylopectinwithin starch granules.

1.3. Minor components

Minor components associated with starches cor-respond to three categories of materials: (i) partic-ulate material, composed mainly of cell-wallfragments; (ii) surface components, removable byextraction procedures; and (iii) internal compo-nents. Lipids represent the most important frac-tion associated with the starch granules. Highcontents are generally observed for cereal starches:0.81.2 and 0.60.8% for wheat and normalmaize, respectively.

The main constituents of surface componentsare proteins, enzymes, amino acids and nucleicacids. Some components can be extracted withoutgranule disruption: approximately 10% of proteinsand 1015% of lipids. A number of proteins [43]associated with wheat starch granules isolated byaqueous extraction from grain or flour have re-ceived the most attention. Some appear to beintegral components of the granule structure,whereas others appear to be associated with thegranule. One of the starch granule proteins, fri-abilin, was studied in detail because of its associa-tion with changes in wheat grain endospermtexture, from soft wheat to hard wheat [44]. Itconsists of several different polypeptides, in whichpuroindoline polypeptides are mostly represented.The reason for the differential binding of theseproteins to the surfaces of hard and soft wheatstarch granules remains to be found [45]. Associa-tion of friabilin with starch could be considered asartefactual. Being located at the starch granulesperipheries, the puroindoline-b-polypeptideswould presumably become accessible for adsorp-tion to the granule surface immediately after flourwetting. Implication of lipid binding in the mecha-nism by which friabilin polypeptides could be re-lated either on a direct bridge of friabilin on lipidzones or a protein conformational change offriabilin.

Triglycerides represent a major fraction of sur-face lipids of maize and wheat. Glycolipids andphospholipids would correspond to amyloplastmembrane remains [46]. The location of the lipidsat the surface of starch granule is still unknown.

In contrast, internal components are composedmainly of lipids. Proteins, including granule boundstarch synthase, are minor. Extraction procedureshave been optimized by Morrison and coworkers[4750] on the basis of water-saturated butan-1-olat 95100C. The presence of internal lipids is acharacteristic of cereal starches (Table 3).


In contrast to tubers and legume starches, cerealstarches are characterized by the presence ofmonoacyl lipids (free fatty acids (FFA) andlysophospholipids (LPL)) in amounts positivelycorrelated to amylose content [47]. Wheat, barley,rye, and other triticale starches contain almostexclusively LPL, whereas other cereals containmainly free fatty acids together with a minority oflysophospholipids (Table 3). Lysophosphatidyl-choline is the major lipid found for both wheatand maize, with palmitic and linolenic acids. Forbarley starches, the fatty acid composition of thelipids becomes progressively more unsaturated aslipid content increases, but this pattern is lessconsistent in starches from maize [48]. Most waxystarches have negligible lipids. In the family ofmaize starches, by comparing different mutants,lipid content appears most directly correlated withlong-chain linear a-1,4-glucan (i.e. the backboneof amylose) revealed by enzymatic debranching. Instarch from wheat and barley harvested at variousstages of grain development, both amylose andLPL contents increase with maturity. The picturethat is thus emerging is that of a large A-typestarch granule displaying a gradient (from thehilum to the periphery) of increasing amylose andLPL [49].

The presence of LPL and FFA is still a surpris-ing finding from a biological point of view.Monoacyl lipids are usually associated with lipoly-sis, and lysophospholipids are potentially harmfulbecause they can cause cell membrane lysis. Hy-drolysis of amyloplast lipids should have led to thepresence of monoacylgalactosylglycerides.

Numerous studies have observed a correlationbetween these monoacyl lipids and the functionalproperties of barley [50], oat [51] and wheat [52]starches. Monoacyl lipids will induce the forma-tion of amyloselipid complexes during gelatiniza-tion. They will restrict swelling, dispersion of thestarch granules and solubilization of amylose, thusgenerating opaque pastes with reduced viscosityand increased pasting temperatures.

Mineral fractions are negligible in cerealstarches in contrast to tuber starches. Phosphorusis the most important one, which can easilytracked using nuclear magnetic resonance (31P-NMR) [53]. Cereal starches contain phosphorusthat is mainly in the form of phospholipids. Rootand tuber starches are unique in that they containphosphate monoesters, with an exceptionally highlevel in potato (0.089%). Phosphate monoesters ofall these starches were located mainly on the pri-mary C-6 alcohol function and less on the sec-ondary C-3 function of the glucosyl unit.

2. From the chain to the crystal

Starch is biosynthesized as semi-crystalline gran-ules with varying polymorphic types and degreesof crystallinity. As powder diffraction patterns ofnative starch are difficult to interpret because ofboth poor crystallinity and molecular structurecomplexity, many structural models of starch wereestablished from recrystallized amylose. The dif-ferent conformations of amylose and amylopectinwhich can be present at the crystalline or amor-phous state are described.

2.1. A- or B-crystals and double helices

A- and B-type fibrillar or lamellar crystals canbe prepared by deacetylation of amylose triacetatefibers [54,55] or precipitation from dilute solutionsof low polymerization degree amylose [56]. Therecent models for A- and B-type structures arebased on parallel double-stranded helices, right-handed [54,55] or left-handed [57,58] and packed

Table 3Free fatty acids and lysophospholipids present in purifiedcereal starches

Source Free fatty acid Lysophospholipid con-tent rangecontent range

Barley0.030.04Waxy 0.120.750.030.05 0.471.14Normal0.050.09High amy- 0.861.36

lose

Maize0.010.05Waxy 0.010.030.300.53 0.160.35Normal0.380.67High amy- 0.260.61

lose

Rice0.220.50Normal 0.410.86

Wheat0Normal 0.781.190 0.070.17Waxy

Values represent % total dry wt.; data quoted in Morrisonand coworkers [49,50].


Fig. 2. Helical conformations present in starch and starch components: (A) 6-fold single helix (pitch 0.805 nm); (B) 6-foldleft-handed double helix (pitch 2.13 nm); (C) double helix between two short chains in amylopectin. When looking at solidrendering views at the top of the figure, it is clear that the single helix presents a central cavity, contrary to the double helix.

antiparallel [54,55] or parallel [57,58] in the unitcell. The left-handed form (Fig. 2B) is energeticallypreferred to the right-handed form [57], but Mulleret al. [59] showed, using computer simulations ofX-ray scattering curves of amylose gels, that theydo not scatter in a significantly different way.Schulz et al. [60] from a crystallographic study ofa polyiodidemaltohexaose complex, proposed anamylose antiparallel double helix with a centralcavity which can adopt, upon removing the com-plexing molecules, a conformation very close to theparallel stranded one (six residues:turn and a pitchheight of about 2.1 nm). The antiparallel packingis strongly debated since it seems incompatiblewith the cluster model of amylopectin and thebiosynthetic pathway. The most recent models forA and B amylose structures are based upon 6-foldleft-handed double helices with a pitch height of2.082.38 nm [57,58]. In the A-structure [57,58],these double helices are packed with the spacegroup B2 in a monoclinic unit cell (a2.124 nm,b1.172 nm, c1.069 nm, g123.5) with eightwater molecules per unit cell (Fig. 3A). In theB-type structure [58], double helices are packedwith the space group P61 in a hexagonal unit cell(ab1.85 nm, c1.04 nm) with 36 watermolecules per unit cell (Fig. 3B). The symmetry of

the double helices differs in A and B structures,since the repeated unit is a maltotriosyl unit in theA form and a maltosyl unit in the B form [58].Independent evidence for the individuality of eachglucosyl residue in maltosyl and maltotriosyl unitscomes from solid-state 13C-NMR. The C1 peak inA-form spectra is a triplet while it is a doublet inspectra of the B form [61,62].

Fig. 3. Crystalline packing of double helices in A-type (A)and B-type (B) amylose. Projection of the structure onto the(a, b) plane.


Fig. 4. Molecular modelling representation of amylosefatty acid complexes showing the inclusion of the aliphatic part (C12) ofthe fatty acid inside the hydrophobic cavity of the amylose single helix.

2.2. The amylopectin conformation

Amylopectin is usually assumed to support theframework of the crystalline regions in the starchgranule. Two independent studies [63,64] showedthat the branch point does not induce extensivedefects in the double helical structure (Fig. 2C).Only dihedral angles F and C of residues adjacentto the (16) linkage were modified slightly withregard to their values in the double helix. Thecalculated conformation differs from that deter-mined by crystallography for the trisaccharidepanose (aGlc(16)aGlc(14)aGlc), which has aprimary structure equivalent to that of the branchpoint of amylopectin but contains a specific hydro-gen bond. The conformation of chain segmentsbetween branching points is not well known up tonow.

2.3. Amylose as single helices

V-amylose is a generic term for amyloses ob-tained as single helices co-crystallized with com-pounds such as iodine, DMSO, alcohols or fattyacids. Although such compounds are required forformation of the V-type structure, they are notsystematically included in the amylose helix. Thecharacteristic Vh structure obtained with linear

alcohols and fatty acids has been most extensivelystudied [6567]. The chain conformation consistsin a left-handed six residues per turn helix (Fig.2A) with a pitch height of 0.7920.805 nm (i.e. arise per monomer of between 0.132 and 0.136 nm).In the case of amyloselipid complexes, it is as-sumed that the aliphatic part of the lipid is in-cluded inside the amylose helix (Fig. 4), while thepolar group lies outside, being too large to beincluded [6870]. In the Vh-type, the most com-mon form obtained by complexation of amylosewith lipids, the single helices are packed in anorthorhombic unit cell (a1.37 nm, b2.37 nm,c0.805 nm) with the space group P212121 and 16water molecules within the unit cell. The amyloselipid complexes can be crystalline or amorphousdepending on the temperature at which they form[67,71]. The two forms cannot be evidenced bydifferential scanning calorimetry (DSC) since theyyield very similar melting:decomplexing enthalpyand temperature values. The latter form consists inindividual complexed single helices which are notinvolved in a crystalline packing.

The double helical conformation is still ques-tionable since many authors reported some Vh toB transition upon rehydration [72,73]. It is un-likely that any drastic conformational change suchas defolding:refolding can be provoked by such


Fig. 5. Schematic representation of the different structural levels of the starch granule and the involvement of amylose andamylopectin.

mild treatment as rehydration, and Saito et al. [72]proposed that the helical conformations ascribedto both Vh- and B-types are not very different.

2.4. The conformation of amylose or amylopectinin the dry amorphous state

Very few reports are available concerning theconformation of amorphous amylose and amy-lopectin. Trommsdorf and Tomka [74] calculateda detailed model for dry amorphous starch (amy-lose) with different densities (including 1.5 g:cm3)and compared the differential radial distributionfunctions evaluated from X-ray scattering mea-surements and from their model structures. Theirmodel allowed a molecular view of the specificinteractions in amorphous starch which is stronglynetworked by hydrogen bonds but no experimen-tal evidence was reported by the authors.

3. From the crystal to the starch granule

The starch granule organization is very compli-cated and depends strongly on the botanicalorigin. Despite several decades of investigation onthe crystalline ultrastructure of starch, many ques-tions remain unresolved such as the respectivecontribution of amylose and amylopectin to crys-tallinity, the distribution of ordered and unorderedareas in the granule, the size distribution of crys-talline areas or the organization of mixed A- andB-type granules. We present here the most com-monly accepted data about the different levels of

organization within the starch granule. A sche-matic representation of the granule architecture isgiven in Fig. 5.

3.1. Size and shape of starch granules

In plants, starch accumulates in many differentphotosynthetic or non-photosynthetic tissues. Oneof the most abundant and universally distributedforms of the storage polysaccharide concerns tran-sient starch, which is accumulated during the dayin plant leaf cells and broken down at night toachieve a more or less constant supply of sucroseto the non-photosynthetic tissues. However, verylittle information is available concerning either thestructure or the morphology of such granules.

Major attention has been devoted to the long-term form of storage present in endosperm ofcereals, parenchyms of tubers, cotyledons oflegume seeds. These differences are noteworthy asstarch biosynthesis will occur in different geneticploidy backgrounds: 3n for wheat endosperm, 2nfor tubers and cotyledons (assuming a diploidplant in each case which is not accounted for byhexaploid wheat or tetraploid potato). Sizes varyfrom 1 mm up to 100 mm. The most interestingfeature remains the granule shape (Table 4). Thisdemonstrates that the biosynthetic pathway canshape the granule in accordance with the physicalconstraints encountered during cellular starch de-position due to the presence of other subcellularcompounds such as the shape assumed by theamyloplast membrane.


Table 4Morphological features of starch granules and amylose content in major plant sources

Size Average (mm)Source ShapeAmylose content (% total starch)

2124 A-granuleBarley normal 20(wild)

B-granule 232529 A-granuleWheat normal 30 Discs

(wild)B-granule 23 Perfect spheres

Wheat waxy 1.22.0 One type2528 One type 30Maize normal Polyhedral and rounded

(wild)One type 15Maize waxy 0.5

6073 One type 525Maize high amy- Highly elongated irregular filamentlose

Oat A-granule 15 CompoundB-granule 23 Oval

1821 One typePotato normal 40 Large oval(wild)

Potato amylose free 1 One type 40 Large ovalOne type 303336 OvalPea RR (wild)One type 50Pea rr Compound6672One type 202332 RoundPea rbrbOne typePea rr rbrb Compound49

Some sources contain compound granules, cor-responding to the fusing of different granules de-veloping simultaneously within a singleamyloplast. Other special forms occurs in highamylose types where non-symmetrical shapes areobserved, more especially with high amylose con-tent (above 50%). This emphasizes further theimportance of amylopectin as an order parameteracting during starch granule morphogenesis.

Some cereals including oat, wheat and barleypresent large (A-granule) and small (B-granule)granules, whose biosynthesis occurs at two differ-ent stages of development. Contradictory resultshave been published on the respective amylose andlipids contents of A- and B-granules. The basicproblem is to explain how these different popula-tions are deposited in the plant tissue. No satisfac-tory explanation has yet emerged frombiosynthesis studies with respect to the origin ofthis dual population.

3.2. Surface of starch granules

Starches of corn, sorghum, millet, large granulesof wheat, rye and barley observed by scanningelectron microscopy (SEM) were found to havepores [75]. However, it appears that some granulescontain many pores, others a few, and some none

on the surface observed. Hall and Sayre [76],making similar observations, identified the pres-ence of artefacts produced during granule purifica-tion. Gallant and Guilbot [77] also identifiedchemical or physicochemical events which couldoccur during the preparation and observation ofsamples. Moreover, all these experiments do nottake into account the propensity of the granule toswell when water is absorbed, inducing higherporosity inside the amorphous matrix by inducinghigher molecular mobility (plasticizing role of wa-ter). In fact, starch granules, as it has been demon-strated for gels [78], can be considered as porousmaterial exhibiting both external and internal sur-face area. The external surface area is determinedby the shape and size of the particle which can bedetermined using microscopy or light scattering.The internal surface area is defined by the capil-lary structure of the hydrated particle and the sizeof the penetrating reagent.

Surfaces of wheat and potato starch granules,including surface morphology and minor compo-nent composition, have been observed usingatomic force microscopy (AFM) [79]. The authorobserved small protusions (e.g. 1050 nm) onwheat granules and larger ones (200500 nm) onpotato starch granules. Because of the novelty ofthis approach, both a wider panel of starches and


milder granule purification conditions are awaitedto confirm these results.

3.3. Macromolecules orientation

When the starch granules are observed underpolarized light, a characteristic dark cross (centredat the hilum) is seen which has led to the granulesbeing considered as distorted spherocrystals [80].The sign of birefringence is positive with respect tothe spherocrystal radius (nen00.015) whichtheoretically indicates that the average orientationof polymer chains is radial. The intensity of bire-fringence strongly depends on the shape and onthe orientation of the granules with regard to thelight beam. Therefore, for non-spherical granules,it is more accurate to say that the orientation isperpendicular to the growth rings and to the sur-face of the granule [80]. This was also confirmedby solid-state light scattering of starch granules[81]. More recently, Buleon et al. [82] and Waighet al. [83] using microfocus X-ray diffraction witha 2-mm beam, showed that the orientation ofpolymer chains in potato starch was very strongon the edges of the granule and perpendicular tothe surface of the granule. A lower level of orien-tation was found in the more internal regions andat the hilum, but the interpretation was morecomplicated since in the centre of the granule, thebeam probably averages over helices pointing for-ward, backward and sideways. No specific orienta-tion at a 2-mm scale was found for wheat starchgranules, either on the edges or in the centre [82],which means that the radial orientation in suchgranules is weak and limited to very small do-mains. Similar results were obtained on potatoand wheat starch by Chanzy et al. [84] and Helbertand Chanzy [85], respectively, using electron dif-fraction (1 mm2 area) on thin sections of partlyhydrolyzed starch. However, in both studies, thediffractograms were not as well resolved becauseof the great amount of inelastic scattering inherentto this technique.

3.4. The alternating ultrastructure of starch

Research, to date, has been concentrated oncrystalline areas because of the paucity of analyti-cal means for detailed studies of the amorphouszones. Detailed knowledge regarding the structure,the organization and the arrangement of the gran-ule may be obtained studying remnant residues

after acid or enzymatic hydrolysis (Fig. 6). Cou-pled to the introduction of improved analyticaland microscopic techniques, the different levels ofthe starch granule organization have begun to bevisualized.

The distribution of crystalline and amorphousareas of starch has been approached through acidhydrolysis. Under these conditions, granulesweaken, crack and show a lamellar organization[86]. The layered concentric shell structure ofstarch granules has been observed by SEM ortransmission electron microscope (TEM) after acidhydrolysis [87]. The more resistant lamellae arebelieved to represent the crystalline part of thegranule. Amorphous areas which are supposed tobe more susceptible to acid hydrolysis would behydrolyzed first. Starch granules are usually be-lieved to consist of alternating 120400 nm thickamorphous and semi-crystalline layers [80,88]. Theorganization of semi-crystalline shells has beenwidely studied by electron microscopy of thin sec-tions of granules with silver staining [89,90] orfragments of granules [88,91]. It was also assessedon entire granules using small-angle X-ray [92,93]or neutron scattering [94]. Considerable evidencenow supports the finding that the crystalline shellsconsist of a regular succession of alternating amor-phous and crystalline lamellae as reviewed byFrench [80]. The sum of the sizes of one amor-phous and one crystalline lamella ranges from 9 to10 nm (Fig. 5). Nevertheless, detailed knowledgeregarding the structure, organization and arrange-ment of lamellae is still limited. No precise deter-mination of the lateral (tangential) dimensions ofthe elementary crystallites inside the crystallinelamellae has been obtained up to now. From theobservation of the 1.6-nm reflection in potatostarch X-ray diffractograms, Hizukuri and Nikuni[95] propose a crystallite size of 14.7 nm. Thisreflection arises from the 100 plane and thereforeyields some information about the number oftimes the inter-helical distance is contained in thecrystallites. Sterling [96] from observations per-formed on partly gelatinized potato starch gran-ules showed the presence of radially orientedfibrillar elements of 1020 mm in length and 0.31mm in diameter. Oostergetel and van Bruggen [97],from electron diffraction data and silver stainingof potato starch fragments, concluded that thesemi-crystalline domains form a network of left-handed superhelices (diameter 18 nm, pitch 10nm), which could be a well-ordered skeleton for


Fig. 6. Electron micrographs of maize starch granules after hydrolysis by Aspergillus fumigatus (K-27) alpha-amylase: (a, c, e)SEM, and (b, d, f) TEM. Micrographs (a) and (b) show high-amylose maize with an extent of hydrolysis of 50%; (c) and (d) shownormal maize with an extent of degradation of 15%; (e) and (f) show waxy maize with an extent of degradation of 22%. Barscorrespond to 1 mm [100].

the starch granule (Fig. 7). Lastly, on the basis ofelectron and atomic force microscopy observa-tions, Gallant et al. [98] and Baldwin [79] pro-posed that lamellae are organized in sphericalblocklets which range in diameter from 20 to 500nm, depending on the botanical origin of starchand their location within the granule. These con-clusions are based on observations made on thinsilver stained sections of hydrolyzed granules andon protusions observed by AFM at the surface ofpotato and wheat starch granules. Thomson et al.[99] also observed by AFM the surface of wheatstarch granules, and saw features between 50 and450 nm in diameter. Such features compare wellwith the surface protusions seen by Baldwin [79].Unfortunately, Thomson et al. believed them to bea contaminant, possibly protein; thus, further

confirmations are needed for to validate the block-let structure concept.

The crystalline material, as defined by its vol-ume fraction, morphology, distribution and crys-talline type, is considered to be an importantfactor in defining the rate and extent of enzymatichydrolysis. Investigation of starch amylolysis is away to refine our understanding of starch granulestructure. Indeed the crystalline material be-haviour during hydrolysis shows susceptibility dif-ferences, reflecting some structural features. Someauthors claim that during a-amylolysis the lesscrystalline parts of the starch granules are moreeasily digested than the crystalline parts. Althoughenzymatic hydrolysis of starch can sometimes leadto a layered and concentric shell structure, as wellas for acid hydrolysis, no work has reported an


increase in crystallinity after amylolysis of nativestarch granule. Moreover, previous experiments[100] have shown that the botanical type deter-mines the composition of the resistant structuresrecovered. These structures consist either of highlydegraded and pitted granules (even for a hydroly-sis extent as low as 1525%) or fragmented exter-nal shells (Fig. 6), suggesting that the structures donot strictly correspond to the alternative crys-talline and amorphous shells proposed by Jenkinset al. [101]. Results could be interpreted on thebasis of the fact that arrangement of crystallitesinside the granule is responsible for the enzymaticsusceptibility, whatever the crystallite type. It isdifficult to design experiments which providemeaningful analysis of the intricate relation exist-ing between different structural levels and directamylolytic susceptibilities. Starches which are veryresistant towards amylolysis are of the B type,such as potato or cereal with high amylose contentstarches. One of the factors limiting their hydroly-sis could be the nature of the granule surface withrespect both to crystallinity [82] and to the pres-ence of adsorbed non-starch material [102]. Bothwould impede enzyme action. The new techniques,e.g. synchrotron radiation or AFM, would allowthese factors to be investigated.

3.5. The crystalline nature of starch

Native starch granules exhibit two main types ofX-ray diffraction diagrams (Fig. 8), the A type for

Fig. 8. X-ray diffraction diagrams of A-, B- and Vh-typestarch.

cereal starches and the B type for tuber and amy-lose-rich starches [103105]. Another C-type dif-fraction diagram, which has been shown to be amixture of A- and B-type diagrams [71], is charac-teristic of most legume starches [30], and also fromcereals grown in specific conditions of temperatureand hydration [106]. The crystalline V-form char-acteristic of amylose complexed with fatty acidsand monoglycerides, which appears upon gela-tinization of starch, is rarely detected in nativestarches [107], although it has been proved, usingsolid-state NMR, that amorphous amyloselipidcomplexes are present in native maize, rice and oatstarches [70].

The structural models established from A and Bamylose crystals were transposed to crystallineregions of starch, since the main reflections con-tained in the powder diffraction diagrams of na-tive starch were present in the diffractograms ofthese crystals. More recently, Buleon et al. [82]have recorded some fibre-like diffraction patternswithin single potato starch granules using microfo-cus (2 mm) synchrotron analysis. These diagramshave proved to be very similar to those obtainedfrom the B-type amylose fibrillar crystals obtainedby Wu and Sarko [54], which shows that theexperimental data used for solving the B-typestructure of amylose can be directly transposed toB-type starch.

As mentioned above, amylopectin is usually as-sumed to support the framework of the crystallineregions in the starch granule. Hence, the shortchains with polymerization degrees ranging be-

Fig. 7. Schematic model for the arrangement of amylopectinin potato starch in a super-helical organization (from Oost-ergetel and van Bruggen [97]).


tween 15 and 18 probably have a double helicalconformation (Fig. 5) compatible with the modelsdescribed above. Moreover, the same short chainswere used, after extraction from native starch bymild acid hydrolysis, for the preparation of the Aand B lamellar crystals mentioned above. In anycase, the lamellae are believed to represent thecrystalline (side-chain clusters) and the amorphousregions (branching regions) of the amylopectinmolecule according to the model of Robin et al.[26]. A major uncertainty remains concerning therelative content of amorphous lamellae in the dif-ferent models proposed. The density of the starchgranule is very high (about 1.5 depending on thewater content [108]) and slightly greater than thatof lamellar crystals of A-amylose [57]. In the 9-nmrepeat (crystallineamorphous lamellae), thecrystalline region represents 56 nm, assuming thedouble-helical model of Imberty and Perez [63]and a polymerization degree of 15 for amylopectinshort chains. It is unlikely that only the branchpoints of amylopectin are present within the amor-phous lamellae. In that case a density of 1.5 couldnot easily be reached.

The conformation of free amylose in the nativestarch granule is not known. It could also bepartly involved in double helices with amylopectinshort chains in the crystalline regions [109]. Kasen-suwan and Jane [110], using NMR on an insolublefraction of cross-linked starch, showed that amy-lose molecules are randomly interspersed amongamylopectin and must be located in close proxim-ity with amylopectin for them to be cross-linked.They concluded it is plausible that some largemolecules participate in double helices with amy-lopectin, especially since amylose can only be com-pletely extracted at temperatures above 90C.Moreover, amylose chains could be involved inamyloselipid complexes. V-crystallinity is rarelydetected in native starches and most native cerealstarches containing fatty acid and:or monoglyce-rides yield no diffraction diagram characteristic ofthe Vh structure. Gernat et al. [107] determinedsome amount ranging from 15 to 25% of the totalcrystalline phase for amylose-rich starches frommaize and wrinkled pea. High-amylose starchesare known to contain greater amounts of lipids;the extraction, purification and drying of starchprocesses may easily induce amyloselipid com-plexes upon heating. Nevertheless, experimentalevidence was given by Morrison et al. [70] thatsuch complexes may be present in native starches

in the amorphous state. In that case amylosechains are involved in separate (isolated singlehelices) inclusion compounds or arrangements oftoo limited an extent to show the Vh-type diffrac-tion diagram. Such structures can reorganize in adetectable crystalline packing (propagation) in thepresence of heat and moisture [71]. Morrison [111]determined amounts of lipid-complexed amyloseranging from 1520% of total amylose for maizeand wheat to 3759% for waxy barley.

The crystalline nature of starch depends proba-bly on both genetic control and climatic condi-tions during the plant growth. The chain lengthinvolved in the crystalline phase and the branchingpattern in amylopectin molecules influence thecrystallinity and the crystalline type [4]. However,temperature and hydration conditions during theplant growth may also induce some importantchanges in the A:B composition [106]. As forpreparation of amylose crystals [56], the A type isobtained preferentially in dry and warm condi-tions as opposed to the B type (wet and coldconditions). In any case, it is still a real challengefor anyone working on starch biosynthesis andstructure to elucidate the mechanism of starchcrystallization and granule formation.

3.6. Starch and degree of crystallinity

A variety of techniques has been used to deter-mine the absolute crystallinity of native starch[80,109] but the values obtained are very depen-dent on the technique used (Table 5). NMR andinfrared spectroscopy yield something close to he-licity while the easily degradable fractions drawnfrom acid or enzymatic hydrolysis curves mayinclude a part of crystalline regions. Lastly, inmethods based on moisture regain, porosity andsurface of the granules have a strong influence.The major effort has been devoted to the use ofX-ray diffraction. The determination of crys-tallinity of native starch is tricky because of thesmall crystal size and the role of hydration. Twomain types of techniques are commonly usedbased on internal and external comparison, respec-tively. In the first ones, the areas corresponding tothe respective contributions of amorphous andcrystalline scattering are evaluated, as it was firstdescribed by Hermans and Weidinger [112]. In thesecond type of technique, experimental diagramsare compared to 100% crystallinity and completelyamorphous diagrams [113]. Such a technique may


Table 5Degree of crystallinity (%) of different starches determined by acid hydrolysis, X-ray diffraction and solid-state NMR [26,115,116]

Starch X-ray diffractionAcid hydrolysis 13C-NMR

A-Type3843Normal maize 424318.127.0384819.728 4853Waxy maize25 38High amylose maize 18.1363920.027.4 39Wheat3839 49Rice

B-TypePotato 254018.124.0 4050

24 4424.0Cassava

be very useful to determine relative crystallinitiesand compare some starches in the native state orduring processes, but may only yield absolutecrystallinity if the amorphous and crystalline stan-dards have a true 0 and 100% crystallinity. Themethods based on internal reference are used fordetermination of absolute crystallinity but thereare such discrepancies in the published values thatit is difficult to rely on such absolute crystallinityvalues. The values vary from 15 to 45% dependingnot only on the origin and the hydration of starchbut also on the technique used [104,108,109]. Inany case, the crystallinity has to be determined ata well defined hydration since it has been provedto depend strongly on it. For example, the crys-tallinity of B-type starches increases from 8 to 33%H2O [108,114]. Gidley and colleagues [115,116], bycomparison of solid-state NMR and X-ray diffrac-tion data, observed that the double helix contentin native starches (which is the detected order byNMR) is considerably higher than the crystallinitydetermined from X-ray diffraction. It was there-fore concluded that molecular ordered regionscontaining double helices not involved in crys-talline arrays were present within the granule dueto either amylose or amylopectin.

Acid hydrolysis supposed to yield the crystallineresistant part of the granule is used to determineamorphous and crystalline parts. When the per-centage of carbohydrate solubilized by acid isplotted versus time, the curve showed two differ-ent steps (Fig. 9). The first high rate step is at-tributed to the amorphous part of the granuleswhile the second step tends towards a plateau [26].Acid hydrolysis of starch granules might be decon-voluted into two first-order kinetics distinguishingthe hydrolysis behaviour of two main structuralorganizations. Plotting 100:(100X) (where X

represents the hydrolysis extent in percent) versustime, makes these two phases conspicuous. Thesecond part of the curve may be extrapolated attime zero, allowing the part of the granule suscep-tible to acid hydrolysis to be estimated, which isbelieved to correspond to the amorphous part.However, this degradable fraction seems to bemore linked to amylose content than to crys-tallinity of the granule. Increasing amylose contentappears to decrease the starch granule degradationby acid [117]. High amylose content mutants ofmaize or pea starches show a hydrolysis rate 2-fold less important than the normal or high amy-lopectin content equivalent starches. Remnantresidues after acid hydrolysis of high-amylosestarches show a polymerization degree higher thanthose obtained for other starches. While preferen-tial acid hydrolysis fits with current observationsand thinking regarding the possible organizationof amorphous and crystalline zones, the validity ofthe acid hydrolysis method to determine contentand organization of the crystalline part should beexamined carefully. However, this approach is use-ful to isolate more crystalline residues and there-fore to assess the organization of the starchgranule crystalline regions. For instance, Oost-ergetel and van Bruggen [97], on the basis ofobservations using lintnerized starches, have beenable to propose a three-dimensional reconstructionof the residual crystallites.

4. From precursors to macromolecules: anoverview of starch biosynthesis

The presence of only one type of sugar residueand two distinct chemical linkages in starch gran-ules have mislead biologists to believe that biosyn-


Fig. 9. Hydrolysis of waxy, normal and high-amylose-content maize starches in 2.2 N HCl at 35C. Percentage of hydrolyzedstarch (X) vs. time of hydolysis.

thesis of amylose and amylopectin consisted of alogical economical and exceedingly simple path-way. Such pre-conceived ideas were largely re-sponsible for a long lasting lack of interest instarch biosynthesis. Similarly, such apparent andmisleading simplicity discouraged many structuralbiologists from studying the structure of thesepolysaccharides, thus favouring investigationsdealing with proteins, nucleic acids and complexoligo- or polysaccharides. This situation has beenslowly changing and the picture of starch emergingtoday is that of a highly organized structure re-quiring sequential action of many different biosyn-thetic steps for its biogenesis.

4.1. The synthesis of the glucosylnucleotidesubstrate

An overview of starch biosynthesis is given inFig. 10. ADPglucose synthesis through ADPglucose pyrophosphorylase occurs in all starch-ac-cumulating plastids examined to date. Ghosh andPreiss [118] established that plant ADPglucosepyrophosphorylases were regulated through al-losteric activation by 3-PGA and inhibition byorthophosphate. Levi and Preiss [119] subse-quently found that cyanobacterial glycogen syn-thesis through AGPase (ADPglucosepyrophosphorylase) was regulated in a similarfashion. The in vivo concentrations reported forthe effectors and substrates of AGPase fully sup-port the hypothesis that the 3-PGA:Pi ratio regu-

lates the rate of starch biosynthesis in plant leaves(reviewed by Preiss [120]). In storage tissues, AG-Pases with similar regulatory properties werepurified from potato tubers, cereal endospermsand pea embryos (reviewed by Preiss [121]). Inaddition to these enzymes, evidence for the exis-tence in endosperm of cereals of a cytosolic formof AGPase is accumulating. In barley, both thepurified and recombinant starchy endosperm en-zymes display a lowered sensitivity to 3-PGA acti-vation [122]. Since the other enzymes of the starchpathway are located in the plastid, this observa-tion implies the existence of an import mechanism.The adenylate translocator encoded by the maizeBT1 gene could fulfil this function [123]. Indeedthe phenotype due to the maize bt1 mutations

Fig. 10. General scheme for starch biosynthesis: (1) phospho-glucomutase; (2) ADPglucose pyrophosphorylase; (3) gran-ule-bound and soluble starch synthases; (4) branchingenzymes; (5) starch phosphorylase; (6) amylases, debranchingenzymes, maltases; (7) hexokinase.


consists of a dramatic decrease in starch amountaccompanied by a substantial increase in ADPglucose concentration [124], both of which are tobe expected in the absence of the putative translo-cator. It must be stressed that these results do notquestion either the importance of AGPase as arate-limiting step or that of allosteric regulationsof the enzyme. The existence of the cytosolic en-zyme can be envisioned as a late refinement of thebiosynthetic pathway particularly suited to theproduction of storage starch from sucrose incereals.

Ever since the initial purification of the spinachleaf enzyme by Ribereau-Gayon and Preiss [125],it appeared that all plant enzymes purified to date,including those from unicellular green algae, areheterotetramers (a2b2) containing two so-calledlarge and small subunits of related primary struc-tures and similar molecular mass within the 5055kDa range [121]. Cyanobacteria were proved tocontain homotetramers [126]. Expression of thecorresponding cDNAs in Escherichia coli demon-strates that the small subunits are essential forcatalysis while the large subunits modulate theresponse to the allosteric effectors [127,128]. Thiswas also confirmed by mutant analysis in higherplants and green algae [129131].

The in vivo evidence supporting the function ofAGPase in starch biosynthesis can be traced backto the pioneering work of Tsai and Nelson [132]who found a selective defect in ADPglucose py-rophosphorylase in the sh2 (shrunken 2) mutantsof maize that displayed a 70% reduction in starchcontent. Mutations of AGPase have since beendescribed in many other plant species includingbt2 [133] and sh2 in maize, adg1 [134] and adg2[135] in Arabidopsis, rb [136] in peas, and sta1[129,131] and sta6 in Chlamydomonas reinhardtii.Moreover, antisense RNA directed against thesmall subunit of the potato tuber AGPase success-fully inhibited starch synthesis in tubers [137].Conversely, plastid-directed expression of mutantbacterial AGPase with unusually high specific ac-tivity successfully increased starch content intransgenic potatoes [138]. In maize, revertants ofsh2 mutants were isolated with increased AGPaseactivity [139]. These revertants also lead to starchoverproduction. To summarize the impressiveamount of data generated through these mutantanalyses, we will state that both up- and down-regulating AGPase lead to proportional increasesand decreases in the rates of starch biosynthesis.

These experiments thus establish AGPase as themajor control point for starch biosynthesis inplants. The work performed in Arabidopsis and C.reinhardtii also demonstrate that the absence ofthe small subunit encoded by ADG2 and STA6leads to a collapse of starch biosynthesis, whiledisappearance of the large subunit encoded by theADG1 and STA1 genes yields an incompletelydefective phenotype. Absence of the large subunitof the Chlamydomonas heterotetramer yields anenzyme selectively unresponsive to 3-PGA activa-tion, confirming the function established for thelarge subunit in the regulation of AGPase activity[129,131].

In cyanobacteria, glucose-1-P generated throughphosphoglucomutase from glucose-6-P is an essen-tial step of many different biochemical pathways.In plants, distinct phosphoglucomutases encodedby different genes are found both in the cytosoland in the plastid. In addition, hexose phosphatetransport across the plastid membranes has beenevidenced in pea embryos [140], C. reinhardtii(Uwe Klein, personal communication, 1997) andwheat [141,142]. It presently appears that the gen-eration of glucose-1-P for starch biosynthesis fromplastidic phosphoglucomutase is the sole functionof this enzyme. This made possible the isolation ofmutants defective for the plastidic form of theenzyme. The severity of the phenotype largelydepended on the nature of the hexose-phosphatestranslocated between the cytosol and the plastid.An extremely severe phenotype (a 9699% drop instarch content) was reported for the rug3 mutantsof pea [143], the pgmP [144] mutant of Arabidop-sis and the phosphoglucomutase mutant of Nico-tiana syl6estris [145]. On the other hand, adisappearance of plastidic phosphoglucomutase iscorrelated to a less severe phenotype in C. rein-hardtii [131] (an 8595% decrease in starchamounts). It must be pointed out that glucose-1-Ptranslocation was reported for C. reinhardtiichloroplasts, while the selective import of glucose-6-P was documented in peas. Plastidic phospho-glucomutase would also be logically required formetabolism of glucose-1-P generated by phospho-rylase during starch breakdown. However, evi-dence for this dual function is still lacking.

Mutants defective in the supply of the glucosylnucleotide through defects in ADPglucose py-rophosphorylase or phosphoglucomutase all dis-play, as expected, a substantial decrease in therates of starch biosynthesis and therefore in yield.


Little attention was paid to the structure of theresidual starch in these mutants. Van den Koorn-huyse et al. [131] showed that in C. reinhardtii,mutants defective either for phosphoglucomutaseor for the large subunit of ADPglucose pyrohos-phorylase accumulate polysaccharides whosestructure were identical to those of transitorystarch. Transitory starch is accumulated duringthe day by leaf cell chloroplasts and broken downduring the night. The structure of transitory starchdiffers from that of the storage polysaccharide bya decrease in amylose content and a modificationin amylopectin structure [146]. In microalgae suchas C. reinhardtii, transitory starch-like polysaccha-rides are synthesized during the log phase [147],while storage starch synthesis requires a block incell divisions such as that achieved under nitrogenstarvation. The mutants defective for the largesubunit of AGPase in C. reinhardtii accumulateunder nitrogen starvation a polysaccharide virtu-ally identical to that synthesized by log-phasewild-type cultures [131,147]. The balance of en-zyme activities known to be involved in starchbiosynthesis was not influenced by the drop insubstrate supply. It was therefore proposed thatthe profound modifications witnessed in the struc-ture were due to the presence of elongation en-zymes characterized by different affinities for theircommon ADPglucose substrate. It is known forinstance that GBSSI (granule-bound starch syn-thase I), the enzyme responsible for amylosebiosynthesis, displays of all starch synthases thelowest affinity for the nucleotidesugar [121]. Itthus appears logical to detect a net relative de-crease in amylose biosynthesis in these mutants.Van den Koornhuyse et al. [131] thus concludedthat substrate supply could directly controlpolysaccharide structure. This would directly tiestarch structure to carbon metabolism and opti-mize it either for photosynthesis (short-term car-bon storage) or for long-term carbon storage. Itwould be, in this respect, of great interest to assaythe crystallinity of storage and transient starches,and to measure the rate of starch degradation ineach case in vitro using the same set of catabolicenzymes.

4.2. Elongation of the a-1,4-linked glucans

It was apparent ever since the pioneering workof Leloir and colleagues [148,149] that glucose wastransferred from ADPglucose to the non-reduc-

ing end of a growing a-1,4-linked glucan, thusgenerating an extra glucose with the simultaneousrelease of ADP. The enzyme was identified byLeloir et al. as associated with starch granules[148] and was subsequently called granule-boundstarch synthase. GBSS is also able to use non-physiological concentrations of UDPglucose as asubstrate with lesser efficiency. Other enzymeswere later identified in the soluble extracts. Atleast one of the soluble enzymes is found as aminor granule-associated protein [150] while GB-SSI often constitutes over 90% of the totalproteins bound to starch. All these enzymes arethought to be active components of the starchsynthesizing machinery. It is not yet entirely clearas to the exact number of soluble isoforms re-quired to make starch. Two such forms are de-tected in C. reinhardtii [151,152] and pea embryos[153], while three were found in potato tubers [154]and maize endosperm [155]. Since the relative con-tributions of the different soluble starch synthasesmay vary widely between organs and species, it isquite possible that the third soluble enzyme es-caped detection in green algae. It is also possiblethat the number of soluble isoforms increasedduring evolution of vascular plants. The picture isfurther complicated by a report describing primer-dependent glucan synthases outside the chloro-plast [156]. The soluble starch synthases can bedistinguished from GBSSI and the glycogen syn-thases by the presence of N-terminal extensions.While these extensions are unlikely to be active incatalysis they may be involved in the building ofmulti-enzyme complexes together with the appro-priate branching enzymes.

It has been known, since the discoveries byNelson and Rines [157] regarding the enzymaticdeficiency in the waxy mutant of maize, that GB-SSI is responsible for the biosynthesis of the amy-lose fraction. Mutations leading to defects forGBSSI have been isolated in an ever increasingnumber of species, including wx (waxy) maize[158], wx rice [159], wx barley [160], wx wheat[161], amf (amylose-free) potato [162], lam (lowamylose) pea [163], wx amaranth [164] and sta2 C.reinhardtii [165]. All mutants accumulate duringstorage normal amounts of starch granules con-taining amylopectin with wild-type crystalline or-ganization. These important results establish thatamylose is not required for the biogenesis of nor-mal granules. A number of studies focusing on thesynthesis of amylose in vitro establish that GBSSI


incorporates glucose both in amylopectin andamylose according to the conditions used in theseexperiments. Leloir originally noted a stimulationof GBSSI by high concentrations of mal-tooligosaccharides [148] and found incorporationof radioactive glucose into both starch fractions.In a very recent study, Denyer et al. [166] showedthat, in the absence of these oligosaccharides, thelabelled product synthesized in vitro by GBSSIwas confined to the amylopectin fraction. How-ever in the presence of very high maltooligosac-charide concentrations, GBSSI massivelyincorporated glucose into amylose-like glucans.The effect could be mimicked by concentratedcrude potato tuber or pea embryo homogenates,suggesting that maltooligosaccharides could be atwork in vivo. Denyer et al. [166] thus speculatethat maltooligosaccharides could trigger amylosesynthesis in storage starch when their concentra-tion becomes high enough to be physiologicallyactive. It is thus tempting to suppose that theoligosaccharides are used as primers for the syn-thesis of amylose within starch granules. In vivoevidence supporting the involvement of GBSSI inamylopectin synthesis was produced by Maddeleinet al. [167]. In this study, genetic interaction exper-iments clearly showed that defects in GBSSIstrongly reduced amylopectin synthesis in particu-lar mutant backgrounds.

The first evidence for a function of the solublestarch synthases in the biosynthesis of amylopectincan be traced back to 1980 when Preiss and Boyer[168] showed that the maize dull mutants dis-played a strong reduction in the activity of oneparticular soluble starch synthase. However, bothSSII and BEIIa (branching enzyme) clearly de-creased in the maize du mutants and it was notuntil 1998 that Gao et al. [169] demonstrated thatDU was the structural gene for a maize-solublestarch synthase of 186 kDa related to the SSIII ofpotato tubers. The phenotype of dull mutantsconsists of a small reduction in starch amount, theappearance of a more highly branched amy-lopectin and the accumulation of 15% of starch asintermediate material. According to the geneticbackground, maize dull mutants can also be con-sidered as accumulating high-amylose starch, thussuggesting a selective defect in amylopectinbiosynthesis. Two analogous mutants have beenreported in C. reinhardtii (sta3) [151] and in peas(rug5) [143,170]. The C. reinhardtii mutants dis-played a marked decrease in starch content, a

high-amylose phenotype and a substantial modifi-cation of the amylopectin structure. The latterconsisted of relative decrease in intermediate-sizechains and an increase of very small glucans witha maximum of six glucose residues in the chain-length distribution. The crystallinity of the starchcollapsed and switched from the A to the B type.The addition of a defect in the GBSSI structuralgene leads in these mutants to the production ofminute amounts of a very highly branched granu-lar polysaccharide with C-type crystalline organi-zation [152]. All these defects could be traced backto the disappearance of a 115-kDa soluble starchsynthase activity. While this is the sole defectwitnessed in the mutants, it remains to be shownthat the STA3 locus encodes the missing starchsynthase. Similar mutants were isolated in peaswhich also displayed a reduction in starch content,an increase in amylose percentage and a decreaseof intermediate-length glucans within the residualamylopectin. All these modifications are due to thedisappearance of one particular soluble starch syn-thase. In this case, however, the mutant gene isknown to encode SSII which is also found associ-ated to the starch granule (GBSSII) in peas andpotatoes. Finally, antisense constructs directedagainst the minor SSIII of potato tubers yieldedT-shaped cracks at the surface of potato starchgranules and increased levels of phosphate [171].However, the structure of the polysaccharideswere nowhere near as affected as those caused bysta3, rug5 or du mutations. Another likely candi-date for a soluble starch synthase defect is themaize su2 (sugary) mutant which displays many ofthe phenotypic attributes of the pea rug5 and C.reinhardtii sta3 mutants [172]. Both the nature ofthe enzymological defect and the function encodedby SU2 still need to be established. All theseexperiments prove that the soluble starch syn-thases are major components of the amylopectinsynthesis machinery. It is too early, however, tospeculate on the respective functions of the differ-ent starch synthase isoforms. In order to get aclearer picture, we definitively need additional in-formation on the in vitro specificities of the recom-binant enzymes. Equally important areinvestigations performed on the enzymes extractedfrom the plant tissues. Indeed these enzymes arelikely to exist as active complexes together withspecific branching enzymes, as was suggested bythe finding of a BEIIa defect in the dull mutant ofmaize. Moreover, we still need selection of addi-


tional mutants and genetic interaction studies be-fore the selective functions of each starch synthasein the complex pattern of amylopectin synthesisbecome clearer.

4.3. Introducing the a-1,6 branch

A branching enzyme, formerly called the Q-en-zyme, was described in potato tubers as far backas 1945 [173]. The potato enzyme (BEI) was subse-quently shown to be able to catalyze inter-chaintransfer, together with possible intra-chain trans-fers [174]. In all plants analyzed in sufficient detail,there appears to be at least one isoform of branch-ing enzyme belonging to two distinct families ofenzyme (reviewed by Martin and Smith [175]).Enzymes of the A family (BEIIa and BEIIb frommaize, BEI from peas, BEIII from rice) preferamylopectin as an in vitro substrate and transfershort chains, while enzymes of the B family(potato Q-enzyme, maize BEI, pea BEII, rice BEI)prefer longer amylose-like chains and transferlonger chains [176,177]. It has also been recentlyshown that branching enzymes differ with respectto the minimal size of the glucan that can be usedas a substrate for branching [178]. In vitro experi-ments performed with recombinant E. coli glyco-gen and maize starch branching enzymes suggestthat this minimal size ranges from DP 12 for bothglycogen branching enzyme and A-family starchbranching enzyme to DP 16 for starch branchingenzyme of the B family [178]. Moreover, GBE(glycogen branching enzyme) transfers chainsranging between DP 5 and DP 16 while the size ofthe chains transferred by the plant branching en-zyme ranges from DP 6 to over DP 30. Anotherdifference between branching enzymes of the Aand B families can be found in the ability dis-played by branching enzymes of the A family totransfer very short chains from DP 3 to DP 9 witha clear maximum at DP 6 and DP 7. Anotherdistinction comes with the response of these en-zymes to orthophosphate [179]: branching en-zymes of the B family are activated 5-fold byorthophosphate while enzymes of the A familyshow a 1.5-fold maximal activation. It presentlyappears that both types of enzymes also differwith respect to expression patterns during starchaccumulation. In cereal endosperm [179] and peaembryos [180], constitutive expression of branch-ing enzymes of the A family was found whileenzymes of the B family were systematically ex-

pressed later. Clear-cut involvement of branchingenzymes of the A family in amylopectin synthesishas been demonstrated in pea [181], rice [182] andmaize [183]. In all cases, the mutants displayed asimilar phenotype: a reduction in starch correlatedto a large increase both in amylose and in afraction that was called intermediate material (aname which is seemingly applied to polysaccha-rides of very different structure [41]). In maize thisis accompanied by a switch from the A to the Btype of diffraction pattern and by a net loss incrystallinity.

Despite the resemblance in chain-length distri-butions shown between amylopectin and the invitro products generated by branching enzymes onamylose-like glucans, branching enzymes do notmediate starch biosynthesis when expressed in E.coli [184]. In fact, recombinant bacteria lackingtheir own glycogen branching enzyme synthesizeglycogen-like polymers and not starch in the pres-ence of either or both maize endosperm branchingenzymes. However, these glycogen-like polymersalso displayed chain-length distributions which arein agreement with their in vitro specificities. Wemay conclude here by stating that the branchingenzymes, while important, are not sufficient tolead to starch biosynthesis. Other structural deter-minants have thus to be sought. These may befound in the interaction existing between thebranching enzymes and the starch synthases ofdiverse specificities. Another intriguing possibilityis that additional functions may be required toorder the polysaccharides during synthesis.

4.4. Pre-amylopectin processing

The sugary-1 mutants of maize have beenknown for decades to accumulate, in addition tostarch, a novel form of water-soluble polysaccha-ride (WSP) [185]. The structure of this WSP can becompared to that which characterizes animalglycogen and was consequently named phyto-glycogen. As early as 1958, Erlander [186] pro-posed that glycogen could be considered as aprecursor of starch biosynthesis. His initial sugges-tion was that amylopectin would be produced bydebranching this precursor and that amylosewould be subsequently generated through de-branching of amylopectin. Ever since this pro-posal, several researchers have been activelysearching for a debranching enzyme deficiency insu1 maize. Surprisingly, the first differences re-


ported between wild-type and su1 maize consistedin the appearance of an unusual form of branch-ing enzyme in the mutants [187189]. This activ-ity, tentatively called phytoglycogen branchingenzyme, was thought responsible for the formationof a polysaccharide with an abnormal branchingratio: phytoglycogen. The presence of phytoglyco-gen in the mutants would thus be explained by thepresence of an abnormal form of branching en-zyme and would not necessarily reflect a block inthe amylopectin synthesis pathway. This hypothe-sis is still elegantly sustained nowadays by Man-ners [190]. It is, however, an established fact thatof all the endosperm mutants of maize, ae (amy-lose-extender) is the only mutation which preventsphytoglycogen synthesis [191]. Since it is has beendemonstrated that AE encodes BEIIa (a standardbranching enzyme of the A family), we must there-fore assume BEIIa to be phytoglycogen branchingenzyme. Since AE and SU1 are known to bedistinct genes, SU1 cannot thus be the structuralgene for phytoglycogen branching enzyme. A firstclue with respect to the nature of su1 came withthe observation made by Pan and Nelson [192]that su1 maize lacks one form of debranchingenzyme. This activity was shown to be able tocleave pullulan: a homopolymer of maltotriosylresidues linked by a-1,6 branches at the non-re-ducing ends. A definitive proof concerning thenature of the SU1 gene product was established byJames et al. [193] who found that su1 encoded aparticular debranching enzyme. However, the se-quence displayed high degrees of homology withbacterial isoamylases. Isoamylase-type debranch-ing enzymes are reported to cleave the branches ofglycogen and amylopectin but not of pullulan.Moreover, the recombinant SU1 gene product wasdemonstrated to lack pullulanase activity (AlanMyers, personal communication, 1997). Similarresults were obtained in rice, where again thepresence of phytoglycogen could be correlated tothe decrease of pullulanase activity [194]. In thiscase no reports concerning the presence of phyto-glycogen branching enzyme were made. In addi-tion, the sugary deficiency of rice did not encodethe rice pullulanase that was cloned [194,195]. InC. reinhardtii, mutants accumulating less than 1%of the wild-type amount of granular starch wereshown to synthesize up to 5% of water-solublepolysaccharide in the form of phytoglycogen. Thephenotype was correlated to the disappearance ofan 88-kD debranching enzyme activity [196]. It

Fig. 11. A two-dimensional model for starch biogenesis occur-ring at the surface of the granule. Elongation starts from atrimmed amorphous lamella depicted in (A), and proceedsthrough (B) until the critical size needed to accomodate thebranching enzymes is reached. Then, because of the presenceof high branching enzyme specific activities, random branch-ing will occur (C). Debranching activities will simultaneouslytrim the loosely branched glucans down (D). This will preventphytoglycogen synthesis and spare the tightly spacedbranches, thereby generating the next amorphous lamella (E)[196].

has been shown recently that the native purifiedenzyme displayed an isoamylase type of specificityand was unable to cleave pullulan. The severity ofthe phenotype displayed by the C. reinhardtii sta7mutants proves that debranching of an intermedi-ate of amylopectin synthesis is mandatory in orderto get significant starch synthesis in plants. On thebasis of these results and also of those concerningthe minimal requirements of the branching en-zymes, a hypothetical scheme for amylopectin syn-thesis has been proposed by several researchers[197].

The model proposed in Fig. 11 for plant amy-lopectin synthesis explains how the asymmetricaldistribution of branches in amylopectin could begenerated.

Briefly, a previously synthesized amorphouslamella containing tightly spaced branches is beingelongated by a specific soluble starch synthase. Nobranching occurs at this first step. We explainedthis [197] by assuming that branching is preventedbecause the branching enzymes need a minimallength for the glucan to adapt in their catalyticsite. As soon as this length is attained, unorderedbranching and elongation occur simultaneously.At the same time, debranching enzymes will trimdown the structure into the next amorphouslamella. This model was named the discontinuoussynthesis model for amylopectin synthesis. It ex-plains the amylopectin cluster constant sizethroughout the plant kingdom [101], and relatesthe size of the crystalline lamella and therefore ofthe amylopectin cluster to the minimal require-ments of the branching enzymes.


Quite interestingly, it was subsequently foundthat maize BEI and BEII require, respectively, aminimum number of 16 and 12 glucose residuesfor branching to occur [178]. A glucan of 12residues is of the very same size as that measuredfor the crystalline lamella of maize amylopectin. Itis comforting to realise that mutants of maizedefective for BEII and thus retaining only BEIshow an increase of the size of this lamella fromthat which characterizes a DP 12 glucan to that ofa DP 16 glucan [198].

4.5. The beginning of starch: priming the starchgranule

Starch granules are known to display a coreknown as the hilum with a relatively disorganizedstructure. It is generally thought that the numberof granules and therefore their size distribution islimited through low-frequency starch granulepriming. It is tempting to speculate that polysac-charide synthesis priming might be the major bio-chemical event associated with hilum formation.Because glycogen and amylopectin belong to thesame family of storage polysaccharides, it is sus-pected that synthesis of these molecules might beprimed by a common mechanism. The presence ofproteins with autoglucosylating abilities physicallyassociated with mammalian glycogen granules andglycogen synthases has been documented for years(for review see Alonso et al. [199]). A mechanismfor glycogen granule priming has been suggestedto involve UDPglucose-mediated autoglucosyla-tion of these glycogenins through O-glycosylationof specific tyrosine residues followed by elongationcatalyzed by the glycogen synthases. In yeast,cloning of glycogenin-analogue structural geneswas achieved using the double-hybrid technique[200]. This technique allows for detection of anygene product physically interacting with a proteinof interest, in this case yeast glycogen synthase.This led to the identification of a protein phos-phatase and a glycogenin-like protein. Two yeastglycogenins were thus detected and both geneswere disrupted [200]. Inhibition of any of the twoGLG genes did not yield a clear phenotype, whileblocking both severely impaired glycogen synthe-sis. This result is of considerable importance forour understanding of polysaccharide synthesis andestablishes the biological relevance of glycogeninproteins. It provides a piece of evidence that wasneeded by those supporting glycogenins function

in polysaccharide priming. However, it must bestressed that because the whole Saccharomycescere6isiae genome has been sequenced, we arepresently certain that the yeast mutant strainscarrying the double gene disruptions do not syn-thesize any other protein with significant homol-ogy to mammalian or fungal glycogenins. Yetthese mutants still synthesise 15% of th

Documents

Buléon, 1998. Starch Granules Structure and Biosynthesis