A chimeric antisense construct has been used to generate transgenic potatoes (Solanum tuberosumL.) in which activities of both of the main starch synthases responsible for amylopectin synthesis in the tuber (SSII and SSIII) are reduced. The properties of starch from tubers of these plants have been compared with those of starches from transgenic plants in which activity of either SSII or SSIII has been reduced. Starches from the three types of transgenic plant are qualitatively different from each other and from the starch of control plants with unaltered starch synthase activities, with respect to granule morphology, the branch lengths of amylopectin, and the gelatinisation behaviour analysed by viscometry. The effects of reducing SSII and SSIII together cannot be predicted from consideration of the effects of reducing these two isoforms individually. These results indicate that different isoforms of starch synthase make distinct contributions to the synthesis of amylopectin, and that they act in a synergistic manner, rather than independently, during amylopectin synthesis.
Amylopectin is a highly branched, α1,4, α1,6-linked glucose polymer which makes up 70–80% of the starch of storage organs of higher plants. The polymer has a strongly polymodal distribution of branch lengths, with maxima at lengths of approximately 12–16, 40 and 70 glucose units. This polymodal distribution is central to the explanations of the way in which amylopectin is organised to give rise to the semi-crystalline matrix of the starch granule ( Hizukuri 1986). The way in which branch lengths are determined during the synthesis of amylopectin in the plant is not understood. There is, however, evidence that the occurrence in storage organs of multiple isoforms of starch synthase may be important in this respect.
The isoforms of starch synthase described so far can be divided into distinct classes on the basis of similarities in their primary amino acid sequences ( Marshall et al. 1996 ). For example, isoforms known as starch synthase II (SSII), which are considerably more similar in sequence to each other than to other isoforms, have been reported from pea, potato, cassava and maize ( Dry et al. 1992 ; Edwards et al. 1995 , 1996; Harn et al. 1998 ; Munyikwa et al. 1997 ). Isoforms similar to the starch synthase III (SSIII) isoform of potato ( Abel et al. 1996 ; Marshall et al. 1996 ) have been reported from maize (the DU1 isoform; Gao et al. 1998 ) and pea ( Craig et al. 1998 ; Tomlinson et al. 1998 ). Studies of pea, maize and Chlamydomonas mutants show that the loss of specific isoforms of starch synthase results in changes in the branch-length distribution of amylopectin, suggesting that these isoforms make distinct contributions to amylopectin synthesis. The rug5 mutation of pea eliminates SSII, which accounts for about 60% of the soluble activity in wild-type pea embryos ( Craig et al. 1998 ). Amylopectin in mutant embryos differs from that in wild-type embryos in that it has far fewer chains of a degree of polymerisation (dp) in the range 40–50 glucose units, and is enriched in chains of dp less than 10 ( Craig et al. 1998 ). The dull1 mutation of maize eliminates the DU1 isoform from maize endosperm ( Gao et al. 1998 ). Amylopectin in the mutant endosperm is enriched in short chains relative to that in the wild-type ( Wang et al. 1993 ). The sta3 mutant of Chlamydomonas lacks one of two soluble isoforms of starch synthase present in the wild-type alga. The amylopectin is enriched in chains of dp 2–7 and depleted in chains of dp 8–50 ( Fontaine et al. 1993 ).
The specific contributions of isoforms of starch synthase to the structure of amylopectin, observed from mutant analysis, could be due to intrinsic properties of the classes of isoforms. Alternatively, the precise contribution made by an isoform in vivo may depend upon its activity at particular times during the synthesis of starch and on the nature and relative activities of other isoforms of starch synthase and of the starch branching enzyme in the tissue. The effects on amylopectin structure of elimination or reduction of single isoforms, as in the rug5 and dull1 examples above, may thus define the role of that isoform only in the particular genetic and developmental background under study.
To investigate the extent to which the contributions of individual isoforms of starch synthase are independent of those of other enzymes of amylopectin synthesis, we have compared tubers of potato plants in which activities of two isoforms are reduced, either individually or together. Most of the starch synthase activity in the soluble fraction of potato tubers is contributed by two isoforms, SSII and SSIII, which account for about 10–15% and 80% of the activity, respectively ( Abel et al. 1996 ; Edwards et al. 1995 ; Marshall et al. 1996 ). Reduction of SSII activity by expression of antisense RNA has little obvious effect on the structure of starch or the morphology of the starch granule at a gross level ( Edwards et al. 1995 ), whereas reduction of SSIII activity to low levels causes cracking and distortion of granule shape in a manner that suggests alterations in the structure of amylopectin ( Abel et al. 1996 ; Marshall et al. 1996 ). In this paper, we report the generation of plants in which the activities of both SSII and SSIII are reduced. Features of the structure and properties of starch from tubers of these plants have been compared with those from tubers of plants in which activity of either SSII or SSIII is reduced. We have not attempted to provide a complete characterisation of the starches, but rather to apply a number of different techniques which are sensitive to small differences in the composition and structure of starch granules. Data from these techniques are used to assess whether the two isoforms contribute independently to amylopectin synthesis or whether their actions are interdependent.
Production of potatoes with reductions in both SSII and SSIII
In order to obtain potatoes in which activities of both the SSII and SSIII isoforms were reduced, we used a chimeric construct designed to give expression of antisense RNA for both isoforms. The construct contained, in the antisense orientation, the fragments of the SSII and SSIII genes used previously to obtain plants in which activity of one or other of these isoforms was reduced ( Edwards et al. 1995 ; Marshall et al. 1996 ). Developing tubers of independent transformants were screened for reduced activity of SSII and SSIII by electrophoresis of crude, soluble extracts on native gels containing glycogen, followed by incubation of the gels with ADPglucose. This method reveals isoforms of starch synthase as dark bands upon staining with iodine. We have previously identified the two major bands on gels of tuber extracts as SSII and SSIII ( Edwards et al. 1995 ; Marshall et al. 1996 ). The lower band probably represents a third isoform of starch synthase, SSI ( Abel et al. 1996 ). Lines of potatoes expressing antisense RNA for SSII specifically lack the SSII band and lines expressing antisense RNA for SSIII specifically lack the SSIII band ( Edwards et al. 1995 ; Marshall et al. 1996 ; Fig. 1(a).
Of 16 independent transformants screened, 10 retained the SSII and SSIII bands at intensities comparable with those in untransformed tubers, and six consistently showed a reduction in, or complete loss of, both the SSII and SSIII bands. These six were propagated to produce lines 17.6, 17.14, 17.19, 17.21, 17.22 and 17.29, referred to as the SSII/SSIII lines. Typical examples from screening gels are shown in Fig. 1(b). In subsequent experiments, SSII/SSIII lines were compared with two independent lines with reduced SSII activity (lines 7b and 7d, referred to as the SSII lines: Edwards et al. 1995 ), and five independent lines with reduced SSIII activity (lines 4.9, 4.18, 4.19, 4.25 and 4.26, referred to as the SSIII lines; Marshall et al. 1996 ), and with control lines which were either untransformed plants of cv Desiree or a line of Desiree transformed with the T-DNA from pBIN19 alone (described by Edwards et al. 1995 ). Comparisons were made within batches of plants grown in the same greenhouse at the same time.
Activities of starch synthase and other enzymes of starch metabolism
The activity of starch synthase in the soluble fraction of developing tubers was strongly reduced in the SSIII and SSII/SSIII lines compared with that in control lines. The maximum degree of reduction in the two sorts of lines was comparable: the lowest activities among the SSIII and SSII/SSIII lines were 26% and 20%, respectively, of the activity in control lines. As reported previously ( Edwards et al. 1995 ), the activity in the soluble fraction of tubers of the SSII lines was not significantly different from that in control lines (not shown). There were no consistent differences between any of the antisense and control lines in the activity of granule-bound starch synthase on a fresh weight basis or when expressed per mg starch ( Table 1 and data not shown).
Table 1. Activities of starch synthase and starch-branching enzyme and starch contents in control and transgenic tubers
Soluble starch synthase activity
Granule-bound starch synthase activity
Starch-branching enzyme activity
Starch content (mg g–1 fresh weight)
Activities were measured in extracts of freshly harvested, developing tubers. Soluble starch synthase activity and starch-branching enzyme activity were measured on the supernatant after centrifugation of a crude homogenate. Granule-bound starch synthase activity was estimated as the difference between soluble starch synthase activity and starch synthase activity in the initial crude homogenate. Starch synthase activities are expressed as nmol min–1 g–1 fresh weight of tuber. Starch-branching enzyme activity was measured as the stimulation by tuber extract of the incorporation of glucose from glucose 1-phosphate into glucan via phosphorylase a (μmol min–1 g–1 fresh weight of tuber). Values are means ± SE of measurements on the number of independent extracts given in parentheses. Each extract was made on a different tuber, and for each line, tubers were from at least three different plants. Starch contents were measured on samples from the same tubers as those used for measurements of enzyme activity, and in some cases additional tubers harvested from the same plants at the same time. Two samples from each tuber were extracted in ethanol, homogenised, autoclaved and incubated with α-amylase and α-amyloglucosidase. Glucose was measured enzymatically after centrifugation. Values are means ± SE of measurements on the number of separate tubers given in parentheses.
87 ± 3 (3)
182 ± 9 (3)
83 ± 5 (3)
85 ± 10 (4)
184 ± 23 (4)
13.8 ± 1.7 (4)
112 ± 13 (5)
26 ± 4 (4)
120 ± 8 (4)
13.7 ± 0.6 (4)
123 ± 20 (4)
22 ± 5 (5)
166 ± 25 (5)
10.7 ± 1.9 (3)
92 ± 14 (7)
28 ± 7 (4)
153 ± 36 (4)
11.6 ± 0.5 (4)
109 ± 14 (5)
32 ± 4 (6)
169 ± 18 (6)
12.7 ± 1.2 (6)
103 ± 9 (7)
38 ± 4 (5)
181 ± 25 (5)
13.0 ± 0.9 (5)
100 ± 4 (5)
17 ± 2 (5)
143 ± 25 (5)
11.1 ± 1.0 (5)
109 ± 8 (5)
38 ± 5 (5)
115 ± 5 (5)
10.9 ± 1.0 (5)
101 ± 4 (6)
22 ± 2 (5)
181 ± 22 (5)
14.0 ± 1.3 (5)
118 ± 12 (5)
59 ± 9 (5)
147 ± 12 (6)
11.8 ± + 1.4 (6)
120 ± 5 (6)
19 ± 3 (5)
209 ± 14 (5)
12.8 ± 2.0 (3)
83 ± 9 (7)
We investigated whether the reductions in SSII and SSIII in the SSII/SSIII lines were comparable in magnitude with the reductions in these isoforms in the SSII and SSIII lines. To compare the extent of reduction of SSIII in the SSIII and SSII/SSIII lines, we used immunoprecipitation experiments to estimate the contribution of this isoform to soluble starch synthase activity. An antiserum specific to SSIII ( Marshall et al. 1996 ) immunoprecipitated 78% of the activity from crude, soluble extracts of Desiree (see also Marshall et al. 1996 ), but only 10 and 18%, respectively, of the activity from two separate extracts of tubers of the SSIII line 4.9, and 17 and 13%, respectively, of the activity from two separate extracts of tubers of the SSII/SSIII line 17.19. This confirms that the reduction in activity in both the SSIII and SSIII/SSII lines is due to the loss of the SSIII isoform. We could not measure the extent of reduction in activity of the SSII isoform accurately by immunoprecipitation with an SSII antiserum because of the relatively low activity of this isoform. Instead, immunoblots of crude soluble extracts and granule-bound proteins were developed with an antiserum to SSII. SSII protein was routinely detectable in the soluble fraction of tubers of SSIII and control lines, but not of the SSII lines and the SSII/SSIII lines 17.6 and 17.29 ( Fig. 1c). It was usually, but not always, detectable in the soluble fraction from SSII/SSIII lines 17.14 and 17.22 (data not shown). Levels of granule-bound SSII protein were dramatically reduced in SSII lines and SSII/SSIII line 17.29, but considerably less reduced in the other SSII/SSIII lines ( Fig. 1d). Amounts of granule-bound SSII in different lines may not, however, allow comparisons of the overall degree of reduction in SSII between the lines because the extent to which SSII is granule bound may be affected by differences in the structure of starch. Comparison of SSII transcript levels in total RNA prepared from tubers of SSII lines and SSII/SSIII lines revealed that the reductions relative to controls lines were similar (data not shown). This further supports the view that the extent of reduction of SSII in the SSII/SSIII lines is comparable with that in the SSII lines. Taken as a whole, the above results show that the amount of the SSII protein is reduced in the SSII/SSIII lines, but do not allow a quantitative estimate of the extent of its reduction.
Interpretation of the effects of reduction in isoforms of starch synthase will be affected by any alterations in other enzymes of starch metabolism. Accordingly, we compared the antisense and control lines with respect to the activity of starch-branching enzyme and the pattern of starch-metabolising enzymes visualised on a native gel. There were no differences between any of the lines with respect to activity of starch-branching enzyme ( Table 1 and data not shown). Electrophoresis of crude, soluble extracts on native gels containing amylopectin, followed by incubation at an appropriate pH and staining with iodine solution, reveals bands due to the metabolism of the amylopectin by enzymes including α-and β-amylases, disproportionating enzyme and starch-debranching enzymes (e.g. Hill et al. 1996 ). These gels revealed no obvious differences between the antisense and control lines (typical example shown in Fig. 1e).
Effects on granule morphology and starch content
Starch contents were measured on the developing tubers used in assays of starch synthase activity. There were no clear differences between any of the lines. Values for the lines with the lowest starch synthase activities were not statistically significantly different from those of control lines ( Table 1 and data not shown).
The morphology of starch granules of the SSII lines was similar to that of control lines, except that cracks centred on the hilum were usually more apparent in hydrated granules of SSII lines ( Fig. 2). Starch granules of both SSIII and SSIII/SSII lines were abnormal. As reported previously for SSIII lines, granules of both lines appeared to be deeply cracked when hydrated, and frequently consisted of clusters of many small granules ( Fig. 2). Granules of SSIII/SSII lines differed from those of SSIII lines in that many appeared under the light microscope to be deeply sunken in the centre, and under the scanning electron microscope to have holes through the centre ( Fig. 2). This morphology was not observed in starches from any of the SSIII lines.
Effects on the short chains of amylopectin
To provide a sensitive and reproducible means of assessing whether there were differences between the antisense lines in the structure of amylopectin, we examined the relative abundances of chain lengths between 6 and 35 glucose units using gel electrophoresis. This method involves the debranching of starch with the enzyme isoamylase which specifically cleaves α-1,6 glucosidic linkages, covalent linkage of the fluorophore 8-amino-1,3,6-pyrenetrisulphonic acid (APTS) to the reducing ends of the chains thus released, and separation of the chains on an acrylamide gel in an Applied Biosystems (Perkin-Elmer, Foster City, CA, USA) 373A DNA Sequencer ( Morell et al. 1998 ; O’Shea & Morell 1996). Typical examples of primary data are shown in Fig. 3(a).
To enable comparison of different samples, the areas of peaks corresponding to individual chain lengths between 6 and 35 glucose units in each sample were calculated as a percentage of the total peak area in that size range for the sample. For each antisense line, the value for each peak was subtracted from that for the control line, and designated the percentage molar difference ( Fig. 3b). A value above zero indicates an enrichment, and a value below zero a depletion, in chains of that length relative to the control line. Values for each line were highly reproducible, and although there were differences in the magnitude of effects between different lines of the same antisense type, all the lines of any one type showed the same general trends in the difference plots ( Fig. 3b).
The difference plots for SSII, SSIII and SSII/SSIII lines were very different from each other. All three types showed a general enrichment in shorter chains and a depletion in longer chains, but the precise pattern of enrichment and depletion were specific to each line. For example, the strongest maxima occurred at different dp in all three types. SSII and SSIII lines showed strong enrichments at dp 9 and dp 6, respectively, and SSII/SSIII lines showed strong enrichments at both dp 7–8 and dp 12–13. The pattern displayed by the SSII/SSIII lines was qualitatively very different from the sum of patterns of the SSII and SSIII lines. A summed plot of an SSII and an SSIII line with large differences from the control is compared with an SSII/SSIII line in Fig. 3(b) (4).
Effects on gelatinisation behaviour of starch
Differences in gelatinisation behaviour between starches are a sensitive indicator of differences in structure. They reflect, for example, differences in numbers and lengths of double helices of amylopectin ( Cooke & Gidley 1992; Moates et al. 1997 ). We examined the gelatinisation behaviour of starches from transgenic and control tubers by viscometric analysis, which measures changes in viscosity during and after gelatinisation.
Viscometric analysis revealed major differences between the control and the antisense lines, and between the three types of antisense lines ( Fig. 4a–c). The temperature at which viscosity started to increase (onset temperature) was generally lower for the transgenic lines than for the control, with particularly low values for the SSII/SSIII lines 17.19 and 17.29. The maximum viscosity attained (peak viscosity) was higher for SSIII antisense lines than for the control, but lower for SSII lines and for SSII/SSIII 17.6, 17.19 and 17.29.
Because of the extreme and unusual values obtained by viscometric analysis for SSII/SSIII lines 17.6, 17.19 and 17.29, these were further analysed to obtain a more accurate picture of development of viscosity ( Fig. 4d). A much slower rate of heating than in the initial analysis revealed that all of these lines had very low onset temperatures. In lines 17.29 and 17.19 viscosity developed very rapidly with no apparent lag. This may mean that the onset temperature for these starches was actually lower than the temperature at which the measurements started (50°C).
To study interactions between roles of starch synthases during the synthesis of storage starch, we have generated lines of potatoes in which we attempted to reduce the activity of both the SSII and SSIII isoforms of starch synthase. These have been compared with lines in which either the SSII isoform alone or the SSIII isoform alone is reduced in activity. To allow meaningful interpretation of the phenotypes of these lines, it was important to establish first that both SSII and SSIII were reduced in the SSII/SSIII lines, and second that there were no major secondary effects on other enzymes which might influence the structure of starch. The following is evidence that these conditions have been met.
First, immunoprecipitation experiments, native gel analyses and measurements of starch synthase activity showed that the SSII/SSIII lines have reduced activity of SSIII, and that the range of reductions in SSIII activity in these lines is of the same order as that achieved in the SSIII lines (this paper and Marshall et al. 1996 ). Native and SDS gel analyses showed that the activity of SSII was also reduced in SSII/SSIII lines, and reductions in SSII protein were obvious in some of these lines. SSII transcript levels were reduced in the SSII/SSIII lines to an extent similar to that observed in the SSII lines. However, we cannot rule out the possibility that SSII and SSIII are reduced to different relative extents in different SSII/SSIII lines.
Second, native and SDS gel analyses and assays of starch-branching enzyme and granule-bound starch synthase provide evidence that enzymes of starch metabolism other than SSII and SSIII are present at normal levels in the antisense lines. The band attributable to SSI on native gels stained for starch synthase activity ( Abel et al. 1996 ) showed no consistent differences in intensity between antisense and control lines. The amount and activity of the major granule-bound starch synthase (GBSSI) were similar in control and antisense lines. The pattern and relative intensities of bands representing amylopectin-metabolising enzymes on native gels and the activity of starch-branching enzyme were also unaffected in antisense lines.
For each of the features of the structure and properties of starch which we examined, lines of a given antisense type were similar to each other but clearly different from the control lines and from lines of the other antisense types. Reduction in the activity of SSII alone had relatively little effect on the morphology of the granule, but the amylopectin was markedly enriched in chains of dp 8–12 and depleted in chains of dp 15–25. The starch gelatinised at a slightly lower temperature and achieved a markedly lower maximum viscosity than the control. Reduction in the activity of SSIII alone caused serious disruption of granule morphology. The amylopectin was considerably enriched in chains of dp 6, but changes in longer chains were less marked than in the other types of antisense lines. The starch gelatinised at a lower temperature than controls and achieved a greater maximum viscosity. Large reductions in activity of SSII and SSIII together (lines 17.6, 17.19, 17.29) caused severe changes to the morphology of granules including the appearance of sunken areas or holes which were not visible in granules of SSII or SSIII antisense lines. The amylopectin showed two marked peaks of enrichment between dp 7 and dp 13 and a strong depletion in chains of dp greater than 15. This chain length profile was different from that of the starches of either the SSII or the SSIII lines, and was also different from the sum of the profiles of SSII and SSIII lines. The starch gelatinised at a much lower temperature than controls, and achieved a lower maximum viscosity. Smaller reductions in the activities of SSII and SSIII together (lines 17.14 and 17.22) had less pronounced effects on the morphology, structure and properties of the starch, but the starches were qualitatively similar to those of other SSII/SSIII lines rather than to those of either the SSII or the SSIII lines.
Our results lend weight to the view that different isoforms of starch synthase make distinct contributions to the synthesis of amylopectin in potato tubers. Although the rate of starch accumulation is unaffected by reductions in either SSII or SSIII, both reductions have marked and distinctly different effects on starch structure and properties. Our results also indicate that the SSII and SSIII isoforms act in a synergistic manner, rather than independently, in the synthesis of amylopectin. The starch of all of the SSII/SSIII lines is qualitatively unlike that of either the SSII or the SSIII lines with respect to granule morphology, amylopectin structure and gelatinisation behaviour. Its properties in these respects cannot be predicted from consideration of those of the starches of the SSII and SSIII lines. We conclude that the effects of reducing activity of, for example, SSII, in a wild-type background are different from the effects of reducing its activity in a background in which activity of SSIII is also strongly reduced. In other words, the contribution made to amylopectin structure by SSII is different in these two different backgrounds. The same argument can be made for SSIII.
The most likely explanation of these interactive effects is that the qualitative and quantitative contribution of a particular isoform of starch synthase is probably determined at least in part by the structure of the glucan substrate with which it is presented. The structure of the glucan substrate for a starch synthase in vivo results from the actions of the whole suite of isoforms of starch synthase and starch-branching enzyme present in the amyloplast. A reduction in the activity of one isoform will alter the structure of the glucan substrate, and may thus alter the contributions made by other isoforms acting on that substrate. This has important implications for our understanding of the synthesis of amylopectin in plants. It seems unlikely that the process is simply the sum of the independent actions of a number of isoforms of starch synthase and starch branching enzyme. The final structure of amylopectin is probably the complex result of the actions of such isoforms upon each other’s products in a manner which may change through the development of the organ as the patterns of expression of isoforms change relative to each other. It follows from this that the structure of amylopectin in a particular organ cannot be predicted simply from a consideration of the intrinsic properties of the isoforms of starch synthase and starch-branching enzyme in the organ, determined by in vitro assay on defined substrates. Although distinct classes of isoforms of these enzymes can be identified, and isoforms within these classes may well prove to have similar intrinsic properties, the contributions they make to amylopectin structure in vivo will be dependent to a large extent upon the genetic, environmental and developmental background in which they are active.
All potato plants (Solanum tuberosum L. cv Desiree) were grown from shoots propagated in tissue culture ( Edwards et al. 1995 ) then transferred to soil-based compost in a greenhouse at a minimum temperature of 12°C with supplementary lighting in winter. Tubers for assay of enzymes, native gel analyses and bulk starch preparations were used immediately after harvest. Tubers and samples of tuber for analysis of starch content and small-scale extraction of starch were frozen at –20°C for up to 4 months before use.
Construction of binary vector
The 1.2 kb BglII/BamHI fragment of the SSII gene ( Dry et al. 1992 : potato GBSSII, EMBL accession number X78988) was cloned in the antisense orientation into the BamHI site of pRAT4, a construct containing a fragment of the potato SSIII gene ( Marshall et al. 1996 : EMBL accession number X95759) in the antisense orientation, to give pPOT17. In the final construct the SSII antisense fragment was flanked by 1 kb of SSIII at the 3′ end and 0.14 kb of SSIII at the 5′ end. The expression of the hybrid antisense RNA was under the control of a double 35S promoter from cauliflower mosaic virus ( Guerineau & Mullineaux 1993).
Transformation of potato
Transformation and preparation of Agrobacterium inoculum carrying the antisense construct, inoculation of tuber discs of potato, regeneration of shoots and rooting of shoots were all as described by Marshall et al. (1996 ) and Edwards et al. (1995 ).
Assay and native PAGE of starch synthase
Extraction, assay of soluble and granule-bound activity and electrophoresis and development of native, glycogen-containing gels for starch synthase activity were all as described by Marshall et al. (1996 ) and Edwards et al. (1995 ).
Assay of starch-branching enzyme
Extracts were prepared exactly as for assay of soluble starch synthase activity, and assayed for starch-branching enzyme by the phosphorylase-stimulation method as described by Smith (1988).
Native PAGE of starch-hydrolysing activities
Extracts prepared as for soluble starch synthase assays were subjected to electrophoresis in native, amylopectin-containing gels, incubated at pH 6.0 and stained with iodine solution as described by Zhu et al. (1998 ).
Immunoprecipitation of starch synthase activity
Immunoprecipitation experiments were as described by Marshall et al. (1996 ), using an antiserum raised against the SSII isoform of starch synthase from pea embryos ( Smith 1990: pea GBSSII) and the SSIII isoform of starch synthase from potato tubers ( Marshall et al. 1996 ).
SDS-PAGE and immunoblotting
For soluble proteins, preparation of extracts, SDS-PAGE and immunoblotting were as described by Edwards et al. (1995 ). For granule-bound proteins, starch samples were boiled in SDS-containing gel sample buffer ( Laemmli 1970) at a ratio of 1:10 weight:volume, freeze-dried to a small pellet, resuspended in 10 volumes of water and centrifuged at 10 000 g for 10 min. The supernatant was subjected to SDS-PAGE and immunoblotting as for soluble extracts. Antisera were as described above under Immunoprecipitation of starch synthase activity.
Extraction of starch
Starch was purified from potato tubers as described by Edwards et al. (1995 ). Starch used in the chain-length and viscometric analyses described below was prepared from tubers harvested from mature, senescing plants. Tubers from between two and five plants from each line were extracted together to make a bulk preparation.
Measurement of starch content
The starch content of samples of tuber tissue was measured according to Smith (1988).
Analysis of chain lengths of amylopectin
The method was essentially that described by O’Shea & Morell (1996). Samples of starch from the bulk preparations were suspended at 2 mg ml–1 in water, autoclaved and cooled. Samples of approximately 100 μl, corresponding to 80 nmol of reducing equivalents (determined according to Bernfeld 1951), were mixed with 5 μl 50 m m Na acetate (pH 3.5) and incubated at 37°C for 16 h with 5000 U isoamylase (Sigma, Poole, Dorset, UK). Incubations were heated to 100°C for 5 min and centrifuged at 14 000 g for 2 min, then samples of 50 μl were evaporated to dryness in a vacuum centrifuge.
The dry samples were dissolved in 2 μl 0.2 m 8-amino-1,3,6-pyrenetrisulphonic acid (APTS: Lambda Fluoreszenztechnologie, Graz, Austria) in 15% (v/v) aqueous acetic acid and 2 μl 1 m Na cyanoborohydride and incubated at 37°C for 18 h. The reaction medium was then diluted 20-fold with an electrophoresis medium (40 m m Tris, pH 8.6, 6 m urea, 40 m m boric acid). After further dilution of between 1000- and 5000-fold with this medium samples of 2 μl were loaded on uniform 7.5% polyacrylamide gels (Biorad Laboratories, Hemel Hempstead, Herts, UK; 5% bisacrylamide) containing 8.3% (w/v) urea. Electrophoresis was in 89 m m Tris, 89 m m boric acid, 2 m m EDTA for 16 h at 1640 V, 40 mA, 38 W, using an Applied Biosystems (Perkin Elmer, Foster City, CA, USA) 373A DNA sequencer. Data were collected and analysed using Genescan 672 software. The system was standardised using authentic maltohexaose and maltoheptaose as standards ( Fig. 3a).
Rapid viscometric analysis
Viscosity development of starches was analysed using a Rapid Visco Analyser Series 4 instrument (Newport Scientific, Sydney, Australia). For the 13 min profile 2 g of starch was analysed in water at a concentration of 7.4% (w/v) and the analysis used the standard stirring and heating protocol. For the longer profile, 2.5 g of starch was analysed at a concentration of 10% (w/v). The sample was heated under standard stirring conditions at 1.5°C min–1 from 50°C to 95°C, held at 95°C for 15 min then cooled to 50°C at 1.5°C min–1. Viscosity was measured in centipoise (cP).
Samples of purified starch were either suspended in water and viewed with a light microscope or sputter-coated with gold and viewed with a Phillips (Eindhoven, The Netherlands) XL30 Field Emission Gun scanning electron microscope at 3 kV.
We are grateful to Kay Denyer and Rod Casey for helpful comments on this manuscript; to Kim Findlay for advice on the use of the scanning electron microscope; to Carlos Busso and Paul Gooding for advice on the use of the DNA sequencer and software; and to Matthew Morell (Canberra, Australia) for making unpublished information available to us. This work was funded in part by Unilever plc, and in part by competitive grants from the Biotechnology and Biological Sciences Research Council, UK.