Role of Granule-bound Starch Synthase in Determination of Amylopectin Structure and Starch Granule Morphology in Potato*

Reductions in activity of SSIII, the major isoform of starch synthase responsible for amylopectin synthesis in the potato tuber, result in fissuring of the starch granules. To discover the causes of the fissuring, and thus to shed light on factors that influence starch granule morphology in general, SSIII antisense lines were compared with lines with reductions in the major granule-bound isoform of starch synthase (GBSS) and lines with reductions in activity of both SSIII and GBSS (SSIII/GBSS antisense lines). This revealed that fissuring resulted from the activity of GBSS in the SSIII antisense background. Control (untransformed) lines and GBSS and SSIII/GBSS antisense lines had unfissured granules. Starch analyses showed that granules from SSIII antisense tubers had a greater number of long glucan chains than did granules from the other lines, in the form of larger amylose molecules and a unique fraction of very long amylopectin chains. These are likely to result from increased flux through GBSS in SSIII antisense tubers, in response to the elevated content of ADP-glucose in these tubers. It is proposed that the long glucan chains disrupt organization of the semi-crystalline parts of the matrix, setting up stresses in the matrix that lead to fissuring.

Reductions in activity of SSIII, the major isoform of starch synthase responsible for amylopectin synthesis in the potato tuber, result in fissuring of the starch granules.

To discover the causes of the fissuring, and thus to shed light on factors that influence starch granule morphology in general, SSIII antisense lines were compared with lines with reductions in the major granule-bound isoform of starch synthase (GBSS) and lines with reductions in activity of both SSIII and GBSS (SSIII/GBSS antisense lines). This revealed that fissuring resulted from the activity of GBSS in the SSIII antisense background. Control (untransformed) lines and GBSS and SSIII/GBSS antisense lines had unfissured granules. Starch analyses
showed that granules from SSIII antisense tubers had a greater number of long glucan chains than did granules from the other lines, in the form of larger amylose molecules and a unique fraction of very long amylopectin chains. These are likely to result from increased flux through GBSS in SSIII antisense tubers, in response to the elevated content of ADP-glucose in these tubers. It is proposed that the long glucan chains disrupt organization of the semi-crystalline parts of the matrix, setting up stresses in the matrix that lead to fissuring.
Little is known about the processes that determine the morphology of starch granules, but important clues have been gained from studies of mutant plants with altered granule morphology. Many such plants carry mutations in genes encoding isoforms of starch synthase and starch-branching enzyme responsible for the synthesis of amylopectin, the branched ␣1,4, ␣1,6 glucan, which forms the semi-crystalline matrix of the granule. For example, mutations in pea affecting starch-branching enzyme A (at the r locus (1)) and starch synthase II (SSII, 1 at the rug5 locus (2)) convert the normally ovoid granules of the embryo into deeply fissured, multilobed structures and highly twisted, contorted structures, respectively (2,3). Mutations in maize affecting starch-branching enzyme IIb (at the amylose-extender locus (4)) convert the normally polyhedral granules of the endosperm into irregular, elongated structures (5).
The altered granule morphology of the mutants presumably results from alterations in the structures and relative amounts of the glucan polymers of which the granule is comprised. Most mutations affecting isoforms of starch synthase and starch branching enzyme affect both of these parameters. The r and rug5 mutations of pea and the amylose-extender mutation of maize affect the proportion of short and long chains in amylopectin, the chain-length distribution within the short-chain population, and the ratio within the granule of amylopectin to amylose, the essentially linear glucan that makes up 20 -30% of storage starches (2, 6 -9). However, it is not known which of these changes in granule composition and polymer structure leads to the changes in granule morphology.
Transgenic potato plants with reduced activity of isoforms of starch synthase provide a good system for elucidation of the relationship between granule morphology and glucan structure. Plants expressing antisense RNA for granule-bound starch synthase I (GBSS antisense plants), the isoform exclusively responsible for the synthesis of amylose, have much less amylose in the tuber starch but normal granule morphology (10). Plants expressing antisense RNA for starch synthase III (SSIII antisense plants), the major isoform responsible for amylopectin synthesis, have a markedly altered chain-length distribution among the short chains of amylopectin and an increased proportion of very long chains (11,12). Whereas granules from wild-type and GBSS antisense plants have a smooth, ovoid profile, those from SSIII antisense lines are deeply internally fissured and are frequently multilobed (13). We reasoned that the altered granule morphology in SSIII antisense lines could be caused either by the altered chainlength distribution of short chains of amylopectin, or by the increased proportion of very long chains in amylopectin, or by the interaction of amylose molecules within the granule with one or both of these abnormal features of amylopectin.
To distinguish between these possibilities, we have generated transgenic lines down-regulated for both SSIII and GBSS simultaneously (SSIII/GBSS antisense lines), and compared the starch of control and SSIII, GBSS, and SSIII/GBSS antisense lines with respect to amylopectin structure, amylose content, and granule morphology.

Materials
Potato plants (Solanum tuberosum L. cv. Desiree and transgenic plants derived from it) were grown from shoots propagated in tissue culture (14) then transferred to soil-based compost in a greenhouse at a minimum temperature of 12°C with supplementary lighting in winter. Tubers for enzymes assays, native gel electrophoresis, and starch preparations were used immediately after harvest. Tuber tissue for analysis of starch content was frozen at Ϫ20°C for up to 4 months before use. For a given set of experiments, control and transgenic plants were grown in the same greenhouse at the same time.

Methods
Construction of Antisense Binary Vector-The 2.2-kb NcoI fragment encoding the full-length cDNA of potato GBSS (15) was subcloned, in the antisense orientation, into the NcoI site of pRAT3. pRAT3 contains a 1.1-kb fragment of the potato SSIII cDNA in antisense orientation between the cauliflower mosaic virus double 35S promoter and terminator (13 Transformation of Potato-Transformation and preparation of Agrobacterium inoculum carrying the antisense construct, inoculation of tuber discs of potato, and regeneration of shoots were all done as described previously (13,14).
Enzyme Assays and Native Gels-Assays for starch synthases were done according to Refs. 13 and 14. Assays for starch-branching enzyme and ADP-glucose pyrophosphorylase were according to Refs. 17 and 18, respectively. Native gels for starch synthase and starch-degrading enzymes were prepared according to Refs. 13 and 11, respectively.
Assay and Extraction of Starch and Analyses of Granule-bound Proteins-Starch contents of tuber tissue were measured according to Ref. 17. Starch was purified from tubers according to Ref. 14. Starch used in the compositional and structural analyses described below was from tubers harvested from mature, senescing plants. Granule-bound proteins were extracted from purified starch and subjected to SDS-polyacrylamide gel electrophoresis according to Ref. 11.
Visualization of Growth Rings-Approximately 0.3 g of starch was suspended in 1 ml of water, frozen in liquid nitrogen in a mortar, and ground thoroughly to crack granules. After thawing, granules were recovered by centrifugation and incubated with occasional stirring at 37°C for 7 h in 50 mM Mes-KOH (pH 5.6) at 0.3 mg/ml, with 4 units of ␣-amylase (porcine pancreas) mg Ϫ1 starch. Granules were then washed three times by resuspension and centrifugation in acetone at Ϫ20°C, and dried. Dry starch samples were brushed onto the surface of doublesided, carbonated sticky stills attached to SEM stubs (Agar Scientific, Cambridge, UK), coated with gold for 3 min at 15 mA in argon using a Polaran SEM coating unit (Polaran Equipment, Watford, UK), and viewed in a Philips XL30 Field Emission Gun SEM (Philips, Eindhoven, The Netherlands).
Small-angle X-ray Scatter Analysis-Analyses were performed according to Refs. 19 -21. Analysis of Starch Composition-The size distribution and relative amounts of amylose and amylopectin in starch and starch fractions were examined by low pressure gel permeation chromatography. Samples of starch (20 mg), amylopectin (15 mg), or amylose (5.2 mg) were suspended in 1 M NaOH (0.4 ml) and diluted with 1 ml of dH 2 O and boiled for 5 min. After cooling, a further 1 ml of dH 2 O was added. The resulting solution was loaded onto two 1-m (i.d. 15 mm) Sepharose CL-2B columns connected in series, equilibrated in 10 mM NaOH, and eluted in a descending mode at 0.16 ml/min. Fractions of 6 ml were collected, and samples were mixed with acidified iodine (Lugol's) solution and monitored at 595 nm.
Partial Purification of Amylose and Amylopectin-Amylose and amylopectin were prepared by butanol precipitation. Starch samples (1.33 g) were suspended in 100 ml of H 2 O and boiled for 1 h, adjusted to pH 5.9 -6.3 with 100 mM phosphate buffer, autoclaved (100 kPa, 1 h), and heated under reflux conditions at 100°C for 1 h. Butanol (25 ml) was added, and the mixture was refluxed for 1 h then allowed to cool very slowly from boiling point to ambient temperature (over a period of 36 -48 h). The amylose precipitate was collected by centrifugation, and the amylopectin-containing supernatant was again subjected to the refluxing procedure described above. The combined amylopectin-containing supernatants and the combined amylose-containing precipitates were lyophilized. Amylopectin was also prepared by gel perme-ation chromatography, as described above. Fractions from the amylopectin peak were neutralized with 2 M HCl, desalted, and lyophilized.
Analysis of Chain-length Distribution of Amylopectin-Analysis of the chain-length distribution of the shorter chains of amylopectin was achieved by using fluorophore-assisted polyacrylamide gel electrophoresis (FAGE) of debranched starch using an Applied Biosystems 373A DNA sequencer (11). Analysis of the chain-length profile across the full range of chain lengths was achieved by HPLC. For routine analysis, amylopectin obtained either by the butanol precipitation method or by gel permeation chromatography, was suspended in boiling water (10 mg/ml) then autoclaved (100 kPa, 30 min). For some analyses amylopectin was instead suspended in 90% (v/v) Me 2 SO (10 mg/ml) and stirred in a boiling water bath for 15 min before precipitation with four volumes of methanol. The precipitate was redissolved in a one-fifth volume of water. Amylopectin suspensions obtained by either of the above methods were mixed with 1/20 volume of 220 mM sodium acetate (pH 4.8) and 4 ϫ 10 4 units of isoamylase (Sigma Chemical Co.), incubated at 37°C for 16 h then autoclaved (100 kPa, 30 min). A sample of 100 l was immediately loaded onto the HPLC system, which consisted of guard columns and three HPLC columns connected in series (Secu-rityGuard (4 ϫ 3 mm), Biosep-Sec-S 2000 guard column (75 ϫ 7.8 mm), two Biosep-Sec-S 2000 columns (600 ϫ 7.8 mm) and a Biosep-Sec-S 3000 column (600 ϫ 7.8 mm); all from Phenomenex, Torrance, CA). The columns and sample injector were heated to 37°C. The system was eluted with water at 0.28 ml/min, and 0.56-ml fractions were collected. To a sample (0.28 ml) of each fraction was added 1/20 volume of 220 mM sodium acetate (pH 4.8) and 0.28 units of amyloglucosidase. The mixture was incubated at 37°C overnight then assayed enzymatically for glucose. Two checks were made on the authenticity of the long chains detected in amylopectin. First, results were compared from samples of amylopectin prepared from the same batch of starch either by gel permeation chromatography or by the butanol precipitation method. Second, results were compared from samples of the same batch of amylopectin prepared either in water or in Me 2 SO, as described above. Pullulan standards were from Polymer Laboratories Ltd., Church Stretton, Shropshire, UK.
Measurement of Sugar Nucleotides, ATP and ADP-Rapid sampling and extraction of tuber tissue was according to Ref. 22. Assay of sugar nucleotides, ATP and ADP, was by HPLC according to Ref. 23, the authors of which provide evidence of the reliability of the methods of extraction and analysis for the metabolites.

Generation of Plants in Which Activities of both SSIII and GBSS Are Significantly
Reduced-To achieve a simultaneous reduction of both GBSS and SSIII isoforms, tubers of Desiree were transformed with plasmid pROT1, a construct that expresses a chimeric antisense RNA of SSIII and GBSS under the control of the cauliflower mosaic virus double 35S promoter. The transgenic lines thus produced are referred to as SSIII/ GBSS antisense lines. The fragments of the GBSS and SSIII cDNAs used in the construct had both been used previously to obtain plants expressing antisense RNA for either SSIII (11, 13: referred to as SSIII antisense lines) or GBSS (24: referred to as GBSS antisense lines). In subsequent experiments, the SSIII/GBSS lines were compared with three SSIII and three GBSS antisense lines described previously (11,13,24) and with control lines that were either untransformed Desiree or a line transformed with the vector pBIN19 alone.
Developing tubers of primary transformants were assayed for activity of soluble and granule-bound starch synthase. Of the 35 independent transformants examined, seven had reductions of 80% or more in granule-bound activity relative to a control line, and five of these also had reductions of 50% or more in soluble starch synthase activity. The latter reductions were shown to be specific for the SSIII isoform by native gel electrophoresis of tuber extracts followed by staining for starch synthase activity (Fig. 1A). The intensity of the band representing SSIII was greatly reduced in transgenic plants with low soluble starch synthase activity, but the intensities of bands representing SSII (11) and SSI (25) were unaffected. In two of the plants, soluble and granule-bound starch synthase activi-ties were reduced to levels comparable with those in the most extreme SSIII and GBSS antisense lines characterized previously. These lines (Rot1.1 and Rot1.4) were propagated for further study.
Soluble starch synthase activity in the SSIII/GBSS antisense tubers was reduced by 3.4-fold (Rot1.1) and 4.9-fold (Rot 1.4) relative to the control line (Table I), compared with reductions of between 2.5-and 5.5-fold in the SSIII antisense lines (11,13). Thus the reductions in SSIII activity in Rot1.1 and Rot1.4 are comparable with those in the SSIII antisense lines. Granule-bound starch synthase activity was reduced by 8-fold (Rot 1.1) and 14-fold (Rot 1.4) compared with a reduction of 7-to 12-fold at a comparable developmental stage in the most extreme GBSS antisense lines studied previously (lines 5 and 8 (24)). Confirmation that the extent of reduction in GBSS was comparable in the SSIII/GBSS and the most extreme GBSS antisense lines, was obtained by probing blots of granule-bound proteins with an antiserum against GBSS (Fig. 1B).
To check whether activity of enzymes of starch metabolism other than SSIII and GBSS were affected in the SSIII/GBSS antisense tubers, we assayed activities of ADP-glucose pyrophosphorylase and starch-branching enzyme and analyzed starch-degrading activities on native, amylopectin-containing gels. There were no major differences between the SSIII/GBSS antisense lines and untransformed Desiree with respect to these activities (Table I, Fig. 1C).
Granule Morphology and Anatomy-Starch contents (per gram of fresh weight) of mature tubers of the transgenic lines were not reduced relative to those of the control line. In three harvests, the starch contents of SSIII and GBSS antisense lines and the SSIII/GBSS antisense line Rot1.4 were not statistically significantly different from the control line. The starch content of the SSIII/GBSS antisense line Rot 1.1 was about 30% higher than that of the control line in two harvests and not statistically significantly different from the control line in the third harvest (data not shown).
There were striking differences in the morphology of starch granules from mature tubers of the GBSS, SSIII, and SSIII/ GBSS antisense lines. Granules of the GBSS antisense lines were ovoid and unfissured like those of untransformed Desiree (Fig. 2, A and B). Granules of SSIII antisense lines were deeply internally fissured with fissuring centered on the hilum (origin of growth), or were multilobed or clustered structures. This was true of lines with reductions in SSIII activity of as little as 2.5to 3-fold, as well as of lines with greater reductions in activity (11, 13; Fig. 2C). Fissures developed as SSIII granules matured, rather than being formed continuously from the start of granule synthesis. Small granules from young, developing tubers were not fissured (Fig. 2D). Thus fissures represent a fracturing of a previously continuous matrix rather than the non-continuous development of the matrix. Granules of the SSIII/GBSS antisense lines were not fissured. Some appeared as clusters of small granules. Mature granules were more spherical and had a more central hilum than those of untransformed Desiree and the GBSS antisense lines (Fig. 2E).
The internal growth rings of granules also differed between the lines. The growth rings represent the alternation of semicrystalline and amorphous zones within the matrix, with a repeat distance of some hundreds of nm. Growth rings were visualized by digesting internal surfaces with ␣-amylase, which preferentially digests material in amorphous zones. Rings were distinct and evenly spaced in granules from the control line and from the GBSS and SSIII/GBSS antisense lines (Fig. 2, G-I). However, growth rings in granules from SSIII antisense plants were much less distinct and regular in appearance (Fig. 2J). This indicates some disruption of the normal organization of the semi-crystalline zones of the granule.
To investigate further the disruption of the semi-crystalline zones in granules of SSIII antisense lines, the small-angle The blot was developed with antiserum raised against the GBSS of pea embryos (49) at a dilution of 1/1000. C, native, amylopectin-containing polyacrylamide gel of crude, soluble extracts of potato tuber, incubated and stained with iodine. A tuber of the control line is compared with tubers of GBSS and GBSS/SSIII antisense line. Bands result from the actions of several enzymes, including starch debranching enzymes, ␣and ␤-amylases, and disproportionating enzyme. There were no reproducible differences in starch-degrading enzymes between the lines. Lack of effect on these enzymes in the SSIII antisense lines has been documented previously (11). x-ray scattering (SAXS) patterns of starch preparations were examined. Starch granules from many different plants, including potato, show a characteristic peak in the SAXS pattern at a q-spacing of about 0.06 Å Ϫ1 , where q is the angular distance in the scattering pattern. This peak represents a repeat distance of 9 nm within the semi-crystalline zones of the granule matrix. It is believed that this repeat represents the clustering of the shorter chains of amylopectin at regular intervals along the axis of amylopectin molecules. Adjacent chains within clusters form double helices. The packing of the helices in regular arrays produces alternating crystalline and amorphous lamellae (19). The SAXS peak for SSIII antisense starch occurred at a markedly lower q-value and was broader than that for untransformed Desiree starch, whereas the q-value for peaks for SSIII/ GBSS antisense and GBSS antisense starches was similar to that of untransformed Desiree (Fig. 3). This indicates that the periodicity of the semi-crystalline repeat is greater and the packing is less well organized in starch from SSIII antisense tubers than in starch from control, SSIII/GBSS, and GBSS antisense tubers.
Granule Composition-The amylose content of starch was much lower in GBSS and SSSIII/GBSS antisense lines than in the control line. Amylose was almost undetectable in starch from both the most extreme GBSS antisense line and the SSIII/GBSS antisense lines when analyzed by gel permeation chromatography (Fig. 4A). Concentration of amylose by butanol precipitation, followed by gel permeation chromatography of this material, revealed that the residual amylose of the GBSS antisense lines and the SSIII/GBSS antisense line Rot1.4 was of smaller apparent molecular size than amylose from the control line (Fig. 4, B and C).
When stained with iodine solution, granules from the SSIII/ GBSS antisense lines appeared red with a blue-staining central core (Fig. 2F). The same staining pattern is observed in granules from GBSS antisense lines (Fig. 2B (10)). It suggests that amylose is confined to the core of the granule.
Previous iodine-based measurements indicated that the amylose content of starch from SSIII antisense lines was the same as that of control lines (12,26). However, gel-permeation chromatography indicated that the molecular size of amylose in the SSIII antisense lines is greater than in the control lines. The amylose peak from solubilized starch eluted earlier for SSIII/ GBSS samples than for control samples (Fig. 4D). Butanolpurified amylose from the SSIII antisense line 18 had a higher apparent molecular mass than that from the control line (Fig. 4E).
Gel permeation chromatography of amylopectin (purified by butanol precipitation of amylose) from starches of control, SSIII, and SSIII/GBSS antisense lines revealed a single peak of material of high molecular mass. There was little or no material of a lower molecular mass equivalent to that of amylose (data not shown).
Amylopectin Structure-The chain-length distribution of the shorter chains of amylopectin was examined by fluorophoreassisted gel electrophoresis (FAGE) after debranching with isoamylase. All analyses were performed several times, on separately prepared samples. All of the differences discussed be- low were observed routinely and reproducibly.
The chain-length distributions of shorter chains from the SSIII and SSIII/GBSS antisense lines differed considerably from that of the control line (Fig. 5), whereas the GBSS antisense lines were not distinguishable from the control in this respect (not shown, see also Ref. 27). The chain-length distributions of amylopectin from the SSIII and SSIII/GBSS antisense lines were similar: Both were enriched in chains of 6 glucose units (degree of polymerization (dp) 6), and chains of dp 11 to 14, and depleted in chains from about dp 17 upwards, relative to the control line.
Longer amylopectin chains were analyzed by HPLC-gel permeation chromatography. Eluted glucans were measured as glucose after hydrolysis with amyloglucosidase. For starch from tubers of control lines, this method separated amylopectin chains into three peaks (labeled I, II, and III in Fig. 6A). The largest peak (I) was of lowest molecular mass and contained chains in the size range analyzed by FAGE. Peak III contained very long chains. Estimates from pullulan standards (shown on Fig. 6A) suggest that these chains were ϳ160 -500 glucose units in length. The elution profile for amylopectin from GBSS antisense lines was indistinguishable from the control (data not shown). The elution profile for amylopectin from SSIII antisense lines differed from that of the control in three main ways. First, peak II was less pronounced. Second, peak III was more pronounced than for the control line for two out of three SSIII antisense lines. Third, a fourth peak (IV) was present, containing chains longer than those of peak III (Fig. 6B). These chains were approximately dp 700 -1300.
To check that long chains were not an artifact generated by crystallization (retrogradation) of the sample or contamination of the amylopectin with amylose, we compared two methods of purifying amylopectin and two methods of solubilizing material for HPLC (see "Experimental Procedures"). The presence of the very long chains in amylopectin from SSIII antisense lines was unaffected by these different treatments, hence these chains probably form part of the native amylopectin molecule. Debranched, fluorophore-derivatized material was subjected to gel electrophoresis in an Applied Biosystems DNA sequencer. Data were analyzed with Genescan software. Areas of peaks representing chains of between 6 and 37 glucose units were summed, the areas of individual peaks were expressed as a fraction of this sum, and the means of these values were calculated from replicate starch samples from each transgenic line. For each chain length (dp: degree of polymerization), the mean value from an antisense line was subtracted from the mean value for the control line, to give the percentage molar difference. Thus the zero line on the y axis represents the chain-length distribution of the control line (Desiree), and each plotted line represents the percentage molar difference from that control for a transgenic line. Values are means Ϯ S.E. of four values, each from a separate sample of a bulk starch preparation (tubers from several plants) for that line. Control and transgenic lines appearing on one graph were grown in the same greenhouse at the same time. A, starch from SSIII antisense lines 18 (square) and 9 (circle). These data are the same as those in Fig. 3 of (11). B, starch from SSIII/GBSS antisense line Rot1.4.
Amylopectin from the SSIII/GBSS antisense lines was similar to that of the SSIII antisense lines in that peak II was generally less pronounced than in control amylopectin. How-ever, amylopectin from the SSIII/GBSS antisense lines, in common with that of the control and GBSS antisense lines, lacked the very long chains (peak IV) present in the amylopectin of SSIII antisense lines (Fig. 6C).
ADP-glucose Content of Tubers-We reasoned that some of the differences in starch composition and structure between the antisense and control lines might be indirectly attributable to differences in the concentration of ADP-glucose in the plastids. Elevated concentrations of ADP-glucose in the antisense lines are expected because of the large decreases in starch synthase activity. Differences in ADP-glucose concentrations between lines could result in differences in granule composition, because different isoforms of starch synthase, which make distinct contributions to amylose and amylopectin synthesis, have different affinities for ADP-glucose (28).
ADP-glucose levels in developing tubers were up to five times higher in SSIII antisense lines than in the control line and up to 50 times higher in the SSIII/GBSS antisense line Rot1.1 (Table II). These very high levels were not accompanied by decreases in the levels of ATP and ADP, indicating that the total adenylate pool in tubers of the SSIII/GBSS antisense line was twice as high as in tubers of the control line. The UDPglucose content of the tubers was not significantly changed in the transgenic lines. DISCUSSION Our results indicate that the deep fissuring and multilobed appearance of starch granules from SSIII antisense tubers probably result from a build-up during granule maturation of long-range stresses caused by changes in the short-range organization of the amylopectin matrix. Starch granules from tubers of the SSIII/GBSS antisense lines lack the abnormalities of granules from SSIII antisense tubers. This implies strongly that the abnormalities arise because organization of the amylopectin matrix is strongly influenced by products of GBSS.
GBSS is exclusively responsible for the synthesis of the amylose component of starch, and it can also contribute to the synthesis of amylopectin. The amylopectin of mutants of cereals lacking GBSS has a shorter average chain length than that of equivalent wild-type cereals, probably due to the absence of a class of long external chains present in wild-type amylopectin (29 -31). Mutants of the unicellular green alga Chlamydomonas lacking GBSS also have altered amylopectin structure (32,33). Direct evidence that GBSS can elongate amylopectin chains inside starch granules comes from experiments with sweet potato, pea, and Chlamydomonas in which ADP[U-14 C]glucose was supplied to isolated granules. In these conditions GBSS adds glucosyl units from ADP-glucose to long chains in the amylopectin fraction (34 -36). Below we consider three ways in which GBSS might influence granule morphology: via changes in the chain-length distribution of short chains of amylopectin, changes in amylose content, and changes in the long chain component of amylopectin.
Changes in the Shorter Chains of Amylopectin-Our data provide no evidence that GBSS significantly influences the chain-length distribution of the shorter chains of amylopectin in potato. The chain-length profile up to dp 40 was similar for control and GBSS antisense lines and for SSIII and SSIII/ GBSS antisense lines. Comparable data for other plant species similarly suggest that GBSS plays no role in synthesis of the shorter chains of amylopectin. For example, the chain-length profiles of wild-type and near isogenic waxy rice were not statistically significantly different (37). Overall, it is highly unlikely that the impact of GBSS on granule morphology is related to any direct role of the isoform in the synthesis of shorter chains of amylopectin.
Changes in Amylose Content-As expected, starch from both FIG. 6. Separation of amylopectin chains by size-exclusion HPLC. Solubilized, debranched amylopectin was subjected to HPLC on three size-exclusion columns connected in series. Fractions collected from the column were incubated with amyloglucosidase then assayed for glucose. The glucose content of each fraction is expressed as a percentage of the total glucose recovered from the column. For each line, three to five separate samples of amylopectin, each prepared from starch from a different plant, were subjected to chromatography. Data points are the mean Ϯ S.E. (bars) of values from these analyses. Where S.E. bars are not shown, they were smaller than the data point symbol. Peaks referred to in the text are marked as I, II, III, and IV. A, amylopectin from starch of a control line (Desiree, circle). Pullulan standards of known molecular mass (squares, right axis) were subjected to chromatography on the same column system to allow estimation of the masses of amylopectin chains. B, amylopectin from three SSIII antisense lines: 9 (diamond), 18 (square), and 26 (cross). C, amylopectin from the SSIII/GBSS antisense lines Rot1.4 (diamond) and Rot1.1 (square). GBSS antisense and SSIII/GBSS antisense lines contained much less amylose than starch from Desiree. In both sorts of lines the reduction in amylose content was accompanied by a concentration of the remaining amylose toward the center of the granule and a sharp reduction in apparent molecular size of amylose. Both of these effects may result from the failure of amylose synthesis inside the amylopectin matrix to keep pace with the growth of matrix during granule development (24).
The presence of amylose per se is unlikely to be responsible for the fissuring of SSIII antisense granules. Wild-type starch, which has an amylose content comparable with that of SSIII antisense starch, is not fissured, neither are the storage starch granules of numerous other species with amylose contents in the range 20 -35%. However, the interaction of amylose with some specific aspect of the SSIII antisense background may lead to fissuring. For example, amylose may interact with the abnormal short-chain population of SSIII antisense amylopectin to alter the way in which these chains are organized into crystalline lamellae. In general, shorter chains are less stable as double helices, so the abundance of very short chains in SSIII antisense amylopectin is likely to compromise the stability of double helices in this starch. The introduction of long chains of amylose into the matrix may further destabilize the double helices and cause them to pull apart. This would result in loss of lamellar structure and a build up of long-range stresses, leading to fissuring of the matrix as granule size increases. Evidence that amylose can influence the organization of short chains of amylopectin in starches generally has been provided by SAXS analysis of normal and low amylose lines of maize and pea (20), and this idea is further supported by our SAXS data for potato. It has been proposed that amylose chains may either co-crystallize with short chains of amylopectin within crystalline lamellae or that they may be orientated transverse to the lamellar stack within amorphous lamellae. Our SAXS data also provide specific evidence for interference by amylose in the organization of the short chains of SSIII antisense amylopectin. The shift in the q-value of the peak indicates that the repeat distance in the semi-crystalline zones in SSIII antisense granules, but not SSIII/GBSS antisense granules, is significantly greater than the normal 9 nm seen in starches from all species thus far examined (19).
An alternative means by which amylose may disrupt granule organization in SSIII antisense tubers is suggested by the apparent increase in amylose molecular size in these relative to control tubers. Synthesis of larger amylose molecules inside the granule matrix, whether in the semi-crystalline or the amorphous zones, might result in disruption of granule structure.
Changes in the Long Chain Component of Amylopectin-The contribution of GBSS to long chains of amylopectin in normal potatoes is small. We detected no difference in the amount of these chains between control lines and both GBSS and GBSS/ SSIII antisense lines. The relatively small influence of GBSS on long chains in potato contrasts with the situation in Chlamydomonas and some cereals in which a fraction of long chains present in normal amylopectin is eliminated by mutations that eliminate GBSS (see above). It appears that the extent of the contribution of GBSS to synthesis of long chains in amylopectin varies considerably from one type of starch-synthesizing organ to another.
Amylopectin from SSIII antisense lines contains more, and longer, long chains than amylopectin from control and GBSS antisense lines, including the SSIII/GBSS antisense line. This finding agrees with that of Lloyd et al. (12) who used a different set of SSIII antisense plants from those in our study. Depending on their location within the granule, long chains might potentially act to disrupt the normal organization of the shorter chains of amylopectin in the way we propose above for amylose: by forming double helices with short chains or by penetrating and disrupting amorphous lamellae.
We suggest that the additional long chains in the amylopectin fraction of SSIII antisense lines are the products of GBSS activity. The following explanation for their occurrence is based on the fact that the ADP-glucose level in SSIII antisense tubers is higher than that in control tubers, presumably as a direct result of the decrease in starch synthase activity. The ADPglucose content of a normal tuber is severely sub-saturating for starch synthase activity in general and more severely limiting for GBSS activity than for the activity of soluble, amylopectinsynthesizing isoforms. Assuming that ADP-glucose is exclusively plastidial and that plastids represent 10% of the cell volume, an ADP-glucose content of 2 nmol g Ϫ1 fresh weight equates to a plastidial concentration of 0.02 mM. Estimates of the K m (ADP-glucose) of soluble, amylopectin-synthesizing starch synthases range from 0.06 to 0.6 mM (28,38). The K m (ADP-glucose) of GBSS from potato tubers is about 5 mM when measured on isolated starch granules (39) and about 1.2 mM when measured on the recombinant isoform expressed in a soluble form in Escherichia coli (38). Thus increases in ADPglucose levels are expected to increase flux into starch in general (probably explaining why the rate of starch synthesis in the transgenic tubers is near-normal despite drastic reductions in starch synthase activity) and to have a particularly marked effect on flux through GBSS. Given the evidence that GBSS can synthesize long chains in amylopectin in species other than potato, and in isolated starch granules, it seems reasonable to suppose that the additional long chains in amylopectin from SSIII antisense lines result from increased flux via GBSS within the granule matrix in response to elevated ADP-glucose TABLE II Amounts of sugar nucleotides, ATP and ADP in developing tubers Samples of tuber tissue were frozen in liquid nitrogen immediately following excision of the tuber from the growing plant. Frozen tissue was powdered then extracted in trichloroacetic acid (22). Neutralized extracts were used for HPLC analysis of metabolites (23). Values are means Ϯ S.E. of measurements on extracts from six (control line and SSIII antisense line 9) or five (all other lines) tubers, from at least three different plants. All  levels. The same explanation may hold for the increased apparent molecular size of amylose in the SSIII antisense lines. A relationship between ADP-glucose levels and the synthesis via GBSS of long amylopectin chains is also suggested by previous work on mutants of Chlamydomonas. Mutants with lesions in the pathway of ADP-glucose synthesis, which presumably have reduced levels of ADP-glucose, lack a long-chain fraction of amylopectin present in normal starch (40). This long-chain fraction is also missing from mutants with no GBSS, indicating that it is synthesized by GBSS (32). Thus it appears that a reduction in ADP-glucose reduces the extent of synthesis of long amylopectin chains via GBSS in this organism.
In summary, possible explanations for the disruption of organization of the granule matrix that underlies the fissuring observed in SSIII antisense granules include the interaction of amylose with the abnormal short-chain population of amylopectin, the increased molecular mass of amylose, and the presence of very long chains in the amylopectin. These possibilities are not mutually exclusive, and none can be ruled out.
Our conclusion that GBSS products can have a radical influence on the crystalline packing of amylopectin is consistent with two sorts of evidence obtained from studies of Chlamydomonas. First, as in our study, elimination of GBSS from Chlamydomonas mutants with alterations in other isoforms of starch synthase has radical effects on granule organization. As in higher plants, loss of GBSS alone (through mutations at the STA2 locus) has little effect on granule morphology and crystallinity (32,41). Loss of the major soluble starch synthase (SSSII: encoded at the STA3 locus) shifts the crystalline diffraction pattern of the granules from A-type to B-type (41). The crystalline type of starch granules reflects the way in which the double helices are packed in crystalline regions. Loss of both SSSII and GBSS (in the STA2/STA3 double mutant) results in a major shift from crystalline to soluble glucan and a C-type crystallinity (a mixture of A and B crystalline regions) in the remaining starch (33,41). Second, the synthesis of amylose via GBSS in the matrix of isolated starch granules of Chlamydomonas changes the crystalline type of the starch. Over 2 days of incubation with ADP-glucose, the amylose content of isolated granules rose from 2% to 45%, and there was a shift in the crystalline diffraction pattern of the granules from A-type to B-type (36).
Insufficient information is available from other higher plants to show whether our findings for potato are more widely applicable. The granule phenotype of many starch mutants in cereals depends upon genetic background, and most comparative studies have not employed near-isogenic lines (42,43). Mutations affecting starch-synthesizing enzymes frequently have pleiotropic effects on related enzymes (44). There are also important differences between starch synthesis in potato tubers and cereal endosperms that may influence the determination of granule morphology. The representation of classes of starch synthase isoforms in the two sorts of organ is different: Starch synthase I makes a major contribution to activity in endosperms but not in tubers (13,25,45,46). Cereal starches generally have A rather than B type crystallinity, thus the impact of long amylopectin chains and amylose on organization of short chains may be different in potato and cereal starches. Finally, ADP-glucose in cereal endosperms is synthesized primarily in the cytosol rather than in the plastid (47,48). The impact of reductions in starch synthase activity upon ADPglucose concentrations in the plastid is thus likely to be different in potato tubers and cereal endosperms.