• Open Access

Reduction of starch granule size by expression of an engineered tandem starch-binding domain in potato plants


  • Qin Ji,

    1. Graduate School Experimental Plant Sciences, Laboratory of Plant Breeding, Wageningen University, Binnenhaven 5, 6709 PD Wageningen, The Netherlands
    Search for more papers by this author
  • Ronald J. F. J. Oomen,

    1. Graduate School Experimental Plant Sciences, Laboratory of Plant Breeding, Wageningen University, Binnenhaven 5, 6709 PD Wageningen, The Netherlands
    2. Department of Biological and Nutritional Sciences, University of Newcastle upon Tyne, Newcastle upon Tyne NE1 7RU, UK
    Search for more papers by this author
  • Jean-Paul Vincken,

    1. Graduate School Experimental Plant Sciences, Laboratory of Plant Breeding, Wageningen University, Binnenhaven 5, 6709 PD Wageningen, The Netherlands
    Search for more papers by this author
  • David N. Bolam,

    1. Department of Biological and Nutritional Sciences, University of Newcastle upon Tyne, Newcastle upon Tyne NE1 7RU, UK
    Search for more papers by this author
  • Harry J. Gilbert,

    1. Department of Biological and Nutritional Sciences, University of Newcastle upon Tyne, Newcastle upon Tyne NE1 7RU, UK
    Search for more papers by this author
  • Luc C. J. M. Suurs,

    1. Graduate School Experimental Plant Sciences, Laboratory of Plant Breeding, Wageningen University, Binnenhaven 5, 6709 PD Wageningen, The Netherlands
    Search for more papers by this author
  • Richard G. F. Visser

    Corresponding author
    1. Graduate School Experimental Plant Sciences, Laboratory of Plant Breeding, Wageningen University, Binnenhaven 5, 6709 PD Wageningen, The Netherlands
    Search for more papers by this author

* Correspondence (fax +31 (0)317 483457; e-mail richard.visser@wur.nl)


Granule size is an important parameter when using starch in industrial applications. An artificial tandem repeat of a family 20 starch-binding domain (SBD2) was engineered by two copies of the SBD derived from Bacillus circulans cyclodextrin glycosyltransferase via the Pro-Thr-rich linker peptide from Xyn10A from Cellulomonas fimi. SBD2 and a single SBD were introduced into the amylose-free potato mutant, amf, using appropriate signal sequences. The accumulation of SBD2 into transgenic starch granules was much higher than that of SBD. In a number of transformants, particularly amfSS3, the starch granules were much smaller than in control plants. The amfSS3 mean granule size was 7.8 µm, compared with 15.2 µm in the control, whereas other starch properties were unaltered. This new starch combines the advantage of the high purity of potato starch with that of the small granule size of other crop species, such as cassava, taro and wheat. This starch may find application in the manufacture of biodegradable plastic films. Both genes were also expressed in Escherichia coli and the affinity for soluble starch of the purified recombinant proteins was determined. SBD2 had an approximately 10-fold higher affinity for starch than SBD, indicating that the two appended SBDs act in synergy when binding to their target polysaccharide ligand.


Starch is the major storage carbohydrate in many plants and an important raw material for food and industrial applications. Starches can be subjected to different kinds of chemical derivatization procedures to improve their properties. In food, starch can be used as a thickening agent. In paper manufacturing, it is used for surface sizing, in which a continuous film of gelatinized starch is deposited on a sheet to enhance the surface properties for writing and printing. In textile manufacturing, yarns are coated with starch polymers to increase their tensile strength during weaving. Starches are also used by the pharmaceutical industry to create a slow-release matrix in which therapeutic compounds are dispersed. The suitability of starch for the above-mentioned applications is determined by its granule size, the physicochemical properties of the granule (such as the amylopectin to amylose ratio and the chain length distribution) and the presence of non-starch components, such as lipids and proteins. It is known that starch properties can vary enormously between different species. For instance, the diameter of starch granules can range from 1 to 40 µm in wheat, 3 to 26 µm in maize and 5 to 100 µm in potato (Ellis et al., 1998). The mechanisms controlling granule size are not known, although there is considerable industrial interest in this. For maize, increased granule size is often desirable, because this may improve the wet-milling efficiency, and thus the starch yield (Gutiérrez et al., 2002). Potato starch granules are often too large to produce high-quality noodles (Chen et al., 2003) and to manufacture biodegradable plastics, in which the starch is used as a dry filler of the plastic film (Lim et al., 1992).

A number of starches with new or improved functionalities have been tailored in planta using recombinant DNA technology. We have previously described a method that anchors foreign proteins to potato starch granules during the biosynthesis of the granules (Ji et al., 2003) via the non-catalytic starch-binding domain (SBD) of Bacillus circulans cyclodextrin glycosyltransferase (CGTase). Interestingly, larger amounts of SBD were accumulated in the granules of the amylose-free (amf) potato mutant than in those of the wild-type potato (Ji et al., 2003). It is envisaged that the use of this technology to incorporate appropriate effector proteins into starch granules will have numerous applications in starch bioengineering (Kok-Jacon et al., 2003).

In this paper, we investigate the possibility of depositing more protein into the starch granule during biosynthesis by using a higher affinity anchor than SBD. It is well established that modular glycosyl hydrolases have acquired a series of non-catalytic carbohydrate-binding modules (CBMs) to increase their affinity for the target polysaccharide. Examples include Cellulomonas fimi xylanase 11A, which has two family 2b xylan-binding CBMs (Bolam et al., 2001), and a Piromyces equi protein, with two family 29 glucomannan-binding CBMs (Freelove et al., 2001). In both proteins, the tandem CBMs displayed affinity for the target ligand which was more than 10-fold higher than that of the individual CBMs, showing that there was significant co-operativity between these linked modules. We have adopted nature's strategy to increase the affinity of starch-binding CBMs for their ligands by engineering an artificial protein comprising two identical SBDs (referred to as SBD2) that are joined by a Pro-Thr-rich linker sequence. The SBD belongs to CBM family 20 and, to our knowledge, no such CBM20 repeat structures have been found in nature. The affinity for starch of this tandem repeat SBD was compared with that of a single SBD. The introduction of SBD2 into the amf potato mutant had a pronounced effect on starch granule size, and the potential of this approach in starch bioengineering in planta is discussed.


Binding affinity of SBD and SBD2

In the tandem SBD construct, SBD2, two identical family 20 SBDs were joined via a Pro-Thr-rich linker peptide derived from the C. fimi Xyn10A (formerly known as Cex; Gilkes et al., 1991; Tomme et al., 1995) (Figure 1). The linker provides sufficient flexibility for each SBD to bind to starch independently. After expression in Escherichia coli, the SBD (single domain) and SBD2 (duplicated domain) proteins were purified to electrophoretic homogeneity by metal ion affinity chromatography. The capacity of the purified proteins to bind to soluble starch was investigated by isothermal titration calorimetry (ITC). The data (Figure 2) show that both proteins interact strongly with soluble starch. As the concentration of binding sites in soluble starch is unknown, the SBDs were treated as the ligand. Both titrations were fitted using a one-binding-site model; the two starch-binding sites on each SBD molecule displayed similar affinity for ligand, as reported previously for a family 20 CBM (Belshaw and Williamson, 1993), preventing accurate measurement of the affinity of the individual binding sites. Thus, the data give an average affinity for all the binding sites in the SBD.

Figure 1.

Schematic representation of the pTrcHisB/SBD2 vector used for the expression of the tandem starch-binding domain (SBD2) in Escherichia coli. The pBIN19/SBD2 binary vector was used for SBD2 expression in potato plants. In this case, the engineered SBD2 gene was under the control of the potato granule-bound starch synthase I (GBSS I) promoter. Amyloplast entry was mediated by the potato GBSS I transit peptide.

Figure 2.

Isothermal titration calorimetry (ITC) of the binding of the starch-binding domain (SBD) (A) and tandem starch-binding domain (SBD2) (B) to soluble starch. The top half of each panel shows the calorimetric titration of ligand into protein, and the lower half shows the integrated heats from the upper panel, corrected for heats of dilution. The full line is the curve of the best fit that was used to derive parameters Ka and ΔH. All titrations were performed in 50 mm sodium phosphate buffer, pH 7.0, at 25 °C.

A comparison of the Ka values showed that SBD2 has an approximately 10-fold higher binding affinity than SBD for starch (Table 1). The greater than twofold increase in the association constant cannot be explained solely by duplication of the binding sites in SBD2. When maltohexaose instead of soluble starch was used as the ligand, SBD2 displayed only a twofold higher affinity than SBD for the oligosaccharide, as would be expected for a protein with twice the number of binding sites (data not shown). This strengthens the hypothesis that the two appended SBDs act synergistically rather than additively, as maltohexaose is probably too small to span the two linked CBM20s in SBD2.

Table 1.  Affinity of the starch-binding domain (SBD) and tandem starch-binding domain (SBD2) for soluble starch, as determined by isothermal titration calorimetry. Data (± SD) are the average of three independent measurements
 Ka (m−1)ΔG (kcal/mol)ΔH (kcal/mol)TΔS (kcal/mol)
SBD1.3 (± 0.2) × 105−7.0 (± 0.1)−21.6 (± 1.6)−14.6 (± 1.7)
SBD21.4 (± 0.1) × 106−8.4 (± 0.0)−28.2 (± 1.2)−19.8 (± 1.2)

Characterization of SBD2 transformants

The amf potato mutant was transformed with the pBIN19/SBD2 construct, and the resultant transgenic potato plants are referred to as amfSSxx (where SS represents the SBD2 gene and xx refers to the clone in the series of transformants). Untransformed control plants are referred to as amf-UT. Fifty kanamycin-resistant, transformed lines were grown in the greenhouse to generate tubers. During growth, the transgenic lines appeared to be phenotypically normal with respect to plant size, shape and colour, as well as the number and size of tubers produced (results not shown).

The levels of SBD2 accumulation in transgenic granules were investigated by Western dot blot analysis. The SBD2 accumulating lines were grouped into seven classes (ranging from 0+ to 6+), based on the amount of SBD2 protein that was associated with the starch granules (see Figure 3A). The 6+ class (not shown in Figure 3A) represents the transgenic granules giving a similar signal in the Western dot blot analysis to the 5+ class, with half the amount of starch. SBD2 accumulation levels in the transgenic granules are summarized in Figure 3(B). For comparison, single SBD accumulation in the amf background (amfS series) is also indicated in this figure (Ji et al., 2003). It is apparent that, in the amfSS series, many more transformants are found in the classes (4+, 5+ and 6+) with high SBD2 accumulation, whereas, for the amfS series, these classes are not observed at all. These results demonstrate, also in planta, that SBD2 is a higher affinity binding domain than SBD.

Figure 3.

Accumulation levels of the starch-binding domain (SBD) and tandem starch-binding domain (SBD2) in potato starch granules isolated from amylose-free (amf) genotypes. Panel A defines the classes of SBD accumulation in potato starch granules. This classification is based on the results of a Western dot blot analysis with various starch samples. The 6+ category represents the transgenic granules which gave a similar signal in Western dot blot analysis to the 5+ class with half the amount of starch. Panel B shows the distribution of the individual transformants over the seven classes of SBD2 accumulation in the amf background (amfSS series). For comparison, single SBD accumulation in transgenic amf potato starch granules is also indicated (amfS series).

In order to estimate the amount of SBD2 protein accumulated in transgenic granules, amfS48, the highest SBD accumulator (3+ class) from the amfS series (Ji et al., 2003), was used as a control for comparison with the density of dots from the amfSS series in Western dot blot analysis. The amount of SBD2 in the highest accumulator (6+ class) was estimated to be approximately fivefold higher than that of amfS48 (approximately 40 mg of SBD2 protein per gram of dried starch). This is also consistent with the results from ITC analysis.

SBD2 accumulation in the soluble fraction of potato juice was also determined by Western dot blot analysis. In none of the juices of the transgenic lines was SBD protein detected, with the exception of juices belonging to the 6+ class of transformants. The amount of SBD found in these juices corresponded to the dot with an intensity of 2+ in Figure 3(A). Thus, only in the class with the highest amount of SBD2 in the granules, can SBD2 be detected in the juice. This suggests that the amount of SBD2 in the 6+ class is saturating, i.e. all binding sites for SBD2 in the starch granules have been occupied.

Granule size is altered by SBD2 expression

Microscopic examination of the starch granule morphology from 50 transformants revealed that the size of the starch granules of some transgenic lines was substantially smaller than that of the control. This observation was further substantiated by measuring the granule size distributions of all the transgenic starches. This analysis confirmed that some of the transgenic lines had smaller granules. A number of lines, including amfSS3 (see further), showed a shift to smaller granules in their granule size distribution. For a number of other lines, the granule size distribution was bimodal (Figure 4A); the larger granules, indicated by the second peak, were approximately similar in size to the amf-UT (non-transgenic control) starch granules. The valley point in the bimodal distributions was, in all cases, approximately 12 µm. Figure 4(B) gives an impression of the granule size in the various SBD2 accumulation classes. For each class, the percentage of transformants in which 50% and 65% of the granules were smaller than 12 µm is indicated (i.e. for these transformants, the grey area in Figure 4A should represent more than 50% and 65% of the total area, respectively). For instance, in the 3+ class, 40% of the transformants showed that over 50% of the granules were smaller than 12 µm, and 30% of the transformants showed that over 65% of the granules were smaller than 12 µm. Thus, it can be clearly seen that a high SBD2 accumulation correlates with more small granules. However, the lines with the smallest granules typically belonged to the 3+ class.

Figure 4.

Relationship between the occurrence of small granules and tandem starch-binding domain (SBD2) accumulation. Panel A shows the bimodal distribution of a representative amfSS starch. The 12 µm point represents the valley point in bimodal distributions, which was observed in a number of transgenic SBD2 starches. Panel B shows the frequency of transformants with small granules in the various SBD2 accumulation classes.

Because amfSS3 had the most pronounced reduction in granule size of all transformants, the starch of this transformant was subjected to a more extensive investigation. Analysis of amfSS3 granules by scanning electron microscopy (SEM) (Figure 5A) showed two size classes of granules for this transformant: large ones and a large number of very small ones (sometimes organized in clusters). The micrograph of amf-UT starch shows a normal distribution of granule sizes (having small, many intermediate and large granules). Figure 5(B) shows the granule size distribution of amfSS3 starch in comparison with starch from amf-UT. The mean granule size by number of amf-UT starch was 15.2 µm, whereas the mean of amfSS3 starch was 7.8 µm, which is approximately twofold smaller than the control. It should be noted that the granule size distribution of amfSS3 is relatively wide compared with that of amf-UT, which is in accordance with the SEM results. Although it seems that the large granules of amfSS3 dominate the scanning electron micrograph at first sight, their abundance is low compared with the small granules, and consequently they contribute relatively little to the granule size distribution.

Figure 5.

Scanning electron microscopy (SEM) of starch granules from wild-type (WT) and transgenic tubers, and particle size distributions of selected starches. Panel A shows SEM analysis of starch granules from amf-UT and amfSS3 tubers at magnifications of 200 times. Panel B shows the particle size distributions of amf-UT and amfSS3 starches. Panel C shows the particle size distributions of taro, wheat, cassava and waxy maize starches for comparison.

For comparison, the granule size distributions of a number of important crop plants are shown in Figure 5(C). Taro starch has a mean of 4.0 µm, which is, to our knowledge, the smallest starch granule known in a crop plant. Wheat starch has a bimodal granule size distribution with large granules of the A-type crystallite and smaller ones with a B-type crystallite (French, 1984). The mean of wheat starch is 4.0 µm for the smaller granules, and 15.0 µm for the larger ones. Maize and cassava are major starch sources for the industry. The mean of maize starch is 10.1 µm, whereas that of cassava is 7.3 µm. Our results show that the granule size of amylose-free potato starch can be decreased to that of cassava by SBD2 expression.

Starch content of line amfSS3 and characterization of its starch

The impact of SBD2 accumulation in granules on the starch content and the physicochemical properties of the starch was also investigated. The starch content of amfSS3 (13.7 ± 1.6 w/w% fresh weight) was comparable with that of amf-UT (15.0 ± 1.1 w/w% fresh weight), which shows that the starch yield is not affected by the reduction in granule size. The granule-melting behaviours (T0 and ΔH) of amf-UT and amfSS3 starch were studied by differential scanning calorimetry (DSC). The pasting properties of the two starches were determined by Bohlin rheometry, which involves measuring the viscosity changes whilst the starch suspension is heated and then cooled with constant stirring. The highest viscosity of a starch paste is referred to as the peak viscosity, whereas the viscosity after cooling of the starch paste is called the end viscosity. Also, the amylose content of both samples was analysed. The results are summarized in Table 2. From the data, it can be seen that there are no consistent differences in the various parameters for the two starches.

Table 2.  Gelatinization properties (T0, ΔH), pasting properties (Tg, Tp, peak and end viscosity) and apparent amylose content (AM, %) of starches from amfSS3 and amf-UT. Bohlin data are the mean of two measurements. Differential scanning calorimetry (DSC) and AM data (± SD) are the average of three independent measurements
CloneGelatinization propertiesPasting propertiesAM (%)
T0 (°C)*ΔH (kJ/g)Tg (°C)Tp (°C)§Peak viscosity (Pa s)End viscosity (Pa s)
  • *

    Temperature at onset of starch gelatinization (DSC).

  • Enthalpy released (DSC).

  • Starch gelatinization temperature (Bohlin).

  • §

    Peak temperature (Bohlin).

amf-UT68.6 (± 0.3)8.0 (± 1.4)61.874.736.226.33.6 (± 0.2)
amfSS368.2 (± 0.5)9.4 (± 1.2)63.575.237.324.53.3 (± 0.2)


In the present study, we have engineered an artificial tandem repeat SBD (CBM20) and determined its affinity for starch, both in vitro and in planta. The ITC data indicate that SBD2 can bind to soluble starch with an approximately 10-fold higher affinity than the individual SBD, indicating that the two SBDs act in synergy to bind the ligand. This increased affinity of SBD2 for starch was also evident in planta. When SBD2 was introduced into amf potato plants, an approximately fivefold higher protein accumulation in the starch granule could be achieved, compared with amf potato starch granules accumulating individual SBDs.

To our knowledge, this paper provides the first report of synergy between two appended CBM20s. It should be noted that particular microbial starch-degrading enzymes have been identified, which contain a triple repeat SBD motif (Sumitani et al., 2000). However, these binding domains are different from those used in this study, and have been classified as family 25 CBMs. To this end, there are no data available on the binding affinity of the individual family 25 modules, or on that of a series of CBM25s. The binding characteristics of appended CBMs have been studied more extensively in plant cell wall-degrading enzymes, such as cellulases and xylanases. For instance, Linder et al. (1996) have shown that an artificial tandem CBM1 exhibited much higher affinity for insoluble cellulose than the individual CBM1s, indicating a co-operative effect between the two domains. Another example is the two CBM2bs from C. fimi xylanase 11A, covalently joined via a flexible linker, which showed an approximately 18–20-fold increase in the affinity for xylan when compared with the individual modules (Bolam et al., 2001). Similar observations were made for two appended CBM29s, which have affinity for cellulose, mannan and glucomannan (Freelove et al., 2001). However, there are also examples of appended CBMs which do not show this synergy in binding the substrate; these include the repeated CBM4s of C. fimi endoglucanase CenC and those of Rhodothermus marinus endoxylanase Xyn10A (Abou-Hachem et al., 2000; Tomme et al., 1996).

Linker sequences in cellulases and xylanases can vary considerably in length (16–59 amino acids) and in amino acid composition (Gilkes et al., 1991). In this study, the two SBDs were separated by a Pro-Thr-rich linker of 22 residues in length derived from C. fimi Xyn10A. Such linker sequences are quite common in modular proteins (Gilkes et al., 1991), and those containing proline may display some rigidity (Janeček et al., 2003). In contrast, the linker in Aspergillus niger glucoamylase 1, which links the SBD to the catalytic module, is relatively long (68 amino acids), glycine-rich and thus flexible (Janeček et al., 2003). It is possible that the nature of the linker may influence the binding characteristics of the SBDs, and it is therefore possible that a linker sequence other than that used in this study could enhance the co-operative binding displayed by the modules in SBD2.

The expression of SBD2 in amf mutant potato plants resulted in a number of clones containing starch with smaller granules. This phenomenon was not observed when single SBD was expressed in the amf background (Ji et al., 2003). The amfSS3 transformant had the smallest granules of all the transgenic lines. The starch yield of this transformant was comparable with that of the untransformed control, suggesting that this line accumulates more small granules and that starch biosynthesis is not inhibited. To a certain extent, the occurrence of small granules in SBD2 transformants is correlated with the amount of SBD2 accumulated in the granule (Figure 4). Typically, the smallest granules were obtained with a transformant belonging to the 3+ class, having intermediate levels of SBD2 accumulation. This result may be explained by Figure 6, which illustrates schematically the interaction of SBD2 with starch at different SBD2 accumulation levels [low (e.g. the 1+ class), intermediate (e.g. the 3+ class) and high (e.g. the 6+ class)]. Our ITC measurements showed that SBD2 interacts strongly with starch. We hypothesize that SBD2 binds instantaneously to the growing starch granule when entering the amyloplast, thereby ‘coating’ the granule surface. At low SBD2 levels, the surface is not fully covered, and the enzymes involved in starch biosynthesis are not hindered greatly in attaching material from the stroma to the growing granule. At intermediate SBD2 levels, the granule surface may be fully covered with SBD2. We postulate that, in this situation, both domains of each protein are attached to the same granule. As a consequence, the SBD2s may hinder synthases from elongating amylopectin side-chains at the granule surface, and/or branching enzymes from attaching side-chains to the growing granule. In this way, additional nucleation sites for granule formation may be enforced by SBD2. The starch-synthesizing potential then needs to be distributed over more nucleation sites, which could explain our observation of more small granules in this class.

Figure 6.

Schematic representation of tentatively different binding modes of the tandem starch-binding domain (SBD2) at various accumulation levels. For simplicity, only the SBD2 proteins at the granule surface are shown. During the biosynthesis process, SBD2 protein is continuously incorporated into the granules as subsequent α-glucan layers are deposited. For details, see text.

At high-level SBD2 accumulation, the surface of the granule is also fully covered but, in this situation, the binding mode of SBD2 is different. We speculate that high concentrations of SBD2 increase the chance that only one of the two domains of SBD2 is attached to the granule surface, whereas the other sticks out in solution (see Figure 6). Our results indicate that the amount of SBD2 in these transformants seems to be saturating, as SBD2 was also detected in the potato juice. Interestingly, the high accumulation class has approximately twofold more SBD2 than the intermediate class, in agreement with the proposed binding modes. We further speculate that the exposed SBD of the SBD2 protein is available for capturing soluble α-glucans in the amyloplast. If these α-glucans are sufficiently long, they may bind more than one SBD2, possibly attached to different growing granules. Potentially, this could ‘cross-link’ different nucleation sites, which may explain why no reduction in granule size is observed in the high SBD2 accumulation classes, i.e. the three indicated nucleation sites may become part of one larger granule.

In this paper, we have shown that it is possible to produce smaller potato starch granules in planta, without accompanying changes in other starch properties and without a yield penalty. Potato starch is often preferred over starches from other sources because of its low level of contaminants, such as proteins and lipids. In certain applications, however, it cannot be used due to its large granule size. An example is the production of starch noodles. Chen et al. (2003) have shown that native potato starch is not suitable for starch noodle preparation due to the abundance of large granules. However, it was demonstrated that a fraction of small-sized potato starch granules (< 20 µm) could be used for making consistently long noodle strands with good quality. Native potato starch only contains a small proportion of such granules (approximately 5%). Another example is the manufacturing of films, in which starch can be used as a biodegradable additive (Gage, 1990). It has been shown that films made with small granules are thinner and have a higher elongation rate and tensile strength (Lim et al., 1992). In principle, the smaller, disc-shaped wheat starch granules could be used for this type of application. Wheat starch, however, has the disadvantage of containing considerable amounts of non-starch contaminants, including proteins (Baldwin, 2001), which, at elevated temperatures, can participate in so-called Maillard reactions that can cause discoloration of the film (Griffin, 1989). Alternative sources of small-granule starch are amaranth or partially degraded corn starch (acid treatment, followed by milling; Jane et al., 1992). However, these sources are relatively expensive, or they require extra processing. Our SBD2 starch combines the purity of potato starch with the small granule found in wheat and corn, which may make this transgenic starch particularly suitable for making starch noodles and films. Another advantage of small-granule potato starch may be in chemical starch modification, such as cross-linking. The more favourable surface-to-volume ratio of the small granules could improve the diffusion rate of the modifying agent into the granule, thereby decreasing the reaction time and possibly reducing the amount of chemicals required.

Experimental procedures

Preparation of constructs

Two constructs were made, one for SBD2 expression in E. coli[pTrcHisB/SBD2], and one for SBD2 expression in potato plants [pBIN19/SBD2]. The pTrcHisB/SBD2 construct was assembled from the pTrcHisB/SBD plasmid (Ji et al., 2003). A sequence similar to SBD in pTrcHisB/SBD and an artificial Pro-Thr-rich linker (Ji et al., 2003) were inserted into pTrcHisB/SBD. The second SBD-encoding sequence was obtained by polymerase chain reaction (PCR) amplification with the primers 5′-CAACTTCGAGGAATTCTACG-3′ and 5′-AAGCTTATGGCTGCCAATTCAC-3′, which contain EcoRI and HindIII sites at their 5′ ends, respectively. The CGTase gene (CGT13A) from Bacillus circulans strain 251 was used as a template (Lawson et al., 1994). The amplified fragment was subcloned into a pGEMTeasy vector (Promega, USA). Plasmid DNA was propagated in E. coli DH5α and purified from the cells using the Wizard Plus Midipreps DNA purification system (Promega, USA). After digestion of this plasmid, the EcoRI-HindIII SBD fragment was inserted into the corresponding sites of the pTrcHisB/SBD expression vector to give the pTrcHisB/SBDI-SBDII vector. The Pro-Thr-rich linker used to connect the two SBDs contained a BglII site at the 5′ end and an EcoRI site at the 3′ end. The linker was inserted into the corresponding sites of the pTrcHisB/SBDI-SBDII vector to give the pTrcHisB/SBD2 expression vector (see Figure 1).

Using standard cloning procedures, the pBIN19/SBD2 vector was assembled from four DNA fragments: (i) the potato tuber-specific granule-bound starch synthase I (GBSS I) promoter (HindIII-NcoI) (van der Leij et al., 1991); (ii) a sequence encoding the potato GBSS I transit peptide for amyloplast entry (NcoI-NcoI) (van der Leij et al., 1991); (iii) an SBD2 fragment (NcoI-BamHI); and (iv) the NOS terminator sequence (SacI-EcoRI) (Bevan, 1984). The first two fragments were subcloned in pMTL23 (Ji et al., 2003). The NcoI-BamHI SBD2 fragment was amplified by PCR with the primers 5′-CCATGGCCGGAGATCAG-3′ and 5′-CTTCTCGGATCCGCCAAAAC-3′ using pTrcHisB/SBD2 as a template. The combined fragments (i)–(ii) and fragment (iii) were cloned in a pBIN19 vector, which already contained the NOS terminator sequence. The plasmid is referred to as pBIN19/SBD2 (see Figure 1). The splice site for cleavage of the transit peptide was the same as that used previously for SBD targeting (Ji et al., 2003). The SBD2 protein, which will be accumulated in potato tubers, has a predicted molecular weight of 25 371 Da, excluding transit peptide. Both constructs were sequenced to ensure that no mutations had occurred during the cloning procedure.

Production of SBD2 protein in E. coli and its purification

Recombinant E. coli Tuner cells (Invitrogen, The Netherlands), containing the pTrcHisB/SBD or pTrcHisB/SBD2 construct, were grown in Luria-Bertani (LB) with 100 µg/mL ampicillin at 37 °C until 0.8 < OD600 < 1.0 (OD, optical density). SBD expression was induced by adding isopropyl-β-d-thiogalactoside (IPTG) to a final concentration of 2 mm, and by growing the cells for 12 h at 29 °C. Cells were harvested by centrifugation (1300 g, 10 min, 4 °C), re-suspended in Talon buffer (20 mm Tris, 300 mm NaCl, pH 8.0) and frozen at −80 °C. The cells were lysed by sonification for 2 min and, after centrifugation (7500 g, 45 min, 4 °C), the cell-free extract was applied to an immobilized metal affinity column (6 mL bed volume; TALON Clontech, USA). The SBD proteins were eluted from the column using buffer with increasing imidazole concentration (10, 15, 20, 30 and 50 mm imidazole in Talon buffer). Sodium dodecylsulphate polyacrylamide gel electrophoresis (SDS-PAGE) was used to check the eluted fractions for the presence of the appropriate polypeptide. Finally, the fractions containing the SBD proteins were pooled and desalted by dialysis against 50 mm sodium phosphate buffer (pH 7.0). The concentration of the purified proteins was determined by UV absorbance at 280 nm, using a calculated molar extinction coefficient of 31 900/m/cm for SBD and 63 800/m/cm for SBD2.

Isothermal titration calorimetry (ITC)

ITC measurements were taken at 25 °C using a MicroCal Omega titration calorimeter. During a titration, the protein sample (0.04–0.1 mm), stirred at 310 r.p.m. in a 1.4331 mL reaction cell, was injected with 25 successive 10 µL aliquots of either 0.2–0.4% (w/v) soluble starch (Sigma Chemical Co.) or 10 mm maltohexaose (Sigma) at 200 s intervals. The proteins were dialysed extensively against 50 mm sodium phosphate buffer, pH 7.0, prior to the titration, and the ligands were dissolved in the same buffer to minimize heats of dilution. The binding data were corrected for the heat of dilution of both proteins and ligand. Integrated heat effects were analysed by non-linear regression using a single set of sites binding model (MicroCal ORIGIN v5.0), yielding independent values for Ka and ΔH. The equation, –RT ln Ka = ΔG = ΔH − TΔS, was used to derive the other thermodynamic parameters.

Plant transformation and regeneration

The pBIN19/SBD2 plasmid was transformed into Agrobacterium tumefaciens according to the three-way mating protocol described by Visser (1991). Internodal stem segments from the diploid amylose-free (amf) potato mutant (often referred to as 1029,31; Jacobsen et al., 1989) were used for Agrobacterium-mediated transformation containing the plasmid (Visser, 1991). More than 50 independent shoots were harvested. Shoots were tested for root growth on a kanamycin-containing (100 mg/L) MS30 medium (Murashige and Skoog, 1962). Fifty transgenic, root-forming, shoots were multiplied and five plants of each clone were transferred to the greenhouse for tuber development. In addition, 10 untransformed controls were grown in the greenhouse.

Starch isolation from potato tubers

All tubers derived from the five plants of each greenhouse-grown clone were combined, and their peels were removed in an IMC Peeler (Spangenberg, The Netherlands). The peeled tubers were homogenized in a Sanamat Rotor (Spangenberg), and filtered through a sieve to remove particulate material. The resulting homogenate was allowed to settle for 20 min at 4 °C, and the tuber juice was collected for later use, and stored at −20 °C. The starch sediment was washed three times with water, and finally air-dried at room temperature.

Determination of SBD2 content of transgenic starches by dot blot analysis

A 12.5% sodium dodecylsulphate-polyacrylamide gel (50 mm × 50 mm × 3 mm), with nine equally spaced holes (φ = 9 mm), was placed in contact with a similarly sized Hybond ECL nitrocellulose membrane (Amersham Pharmacia Biotech, UK). Twenty milligrams of (transgenic) starch was boiled for 5 min with 200 µL of 2× SDS sample buffer (Murashige and Skoog, 1962). After cooling to room temperature, the starch gel was transferred into one of the holes. SBD2 proteins from transgenic starch gels were blotted onto the membrane using a PhastSystem (Pharmacia, Sweden; 20 V, 25 mA, 15 °C, 45 min) (Ji et al., 2003). The protein was identified with anti-SBD antibody according to the method described by Ji et al. (2003).

SBD2 proteins in the soluble fraction were determined as follows. Five hundred microlitres of tuber juice was freeze-dried. The dried material was dissolved in 200 µL of 2× SDS sample buffer. In order to make the sample suitable for the Western dot blot procedure, the mixture was boiled for 5 min in the presence of 20 mg starch from the control samples. The starch gel obtained was applied to one of the holes in the sodium dodecylsulphate-polyacrylamide gel. The rest of the procedure was conducted in the same way as described for granule-bound SBD2.

Determination of starch content

Approximately 50 mg of potato tuber material was dissolved in 0.5 mL of 25% HCl and 2 mL of dimethylsulphoxide (DMSO) for 1 h at 60 °C. After incubation, the mixture was neutralized with 5 m NaOH and diluted in 0.1 m citrate buffer (pH 4.6) to a final volume of 10 mL; 20 µL of the hydrolysed starch sample was determined enzymatically using a test kit (Boehringer, Mannheim Germany), according to the instructions of the manufacturer. The values are an average of three independent measurements.

Analysis of potato tuber starch

Starch granule morphology was investigated by light microscopy (LM; Axiophot, Germany) and SEM (JEOL 6300F, Japan). For LM, starch granules were stained with a 20× diluted Lugol's solution (1% I2/KI). For SEM, dried starch samples spread on silver tape and mounted on a brass disc were coated with a 20 nm platinum layer. Samples were then examined with a scanning electron microscope operating at an accelerating voltage of 1.5–3.5 keV. The working distance was 9 mm.

The average granule size and granule size distribution of the control and transgenic starches were determined in triplicate with a Coulter Multisizer II, equipped with an orifice tube of 100 µm (Beckman-Coulter, UK). Approximately 10 mg of starch was dispersed in 160 mL of Isoton II. The number percentages of the differently sized granules in the sample were recorded by counting approximately 50 500 (± 500) particles. The coincidence (the frequency of two granules entering the tube at the same time, and consequently being counted as one) was set at 10%.

The apparent amylose content of the starches was determined according to the method described by Hovenkamp-Hermelink et al. (1989).

The temperature at which starch granules start to gelatinize was determined by DSC using a Perkin-Elmer Pyris 1 (Perkin-Elmer, The Netherlands), equipped with a Neslab RTE-140 glyco-cooler (Ji et al., 2003). Dynamic rheological properties of 5% (w/v) starch suspensions at small deformations were determined by applying a small oscillating shear deformation (5 s−1) using a Bohlin CVO rheometer (Mettler Toledo, The Netherlands). The suspensions were preheated to ∼40 °C with gentle stirring and loaded into the sample cell (preheated at 40 °C). After this, the cell was subjected to the following temperature programme: heating to 90 °C, 15 min at 90 °C, cooling to 20 °C and 15 min at 20 °C. Heating and cooling were performed at a rate of 2 °C/min. Data were collected automatically every 10 s.