• Open Access

Starches—from current models to genetic engineering


  • Uwe Sonnewald,

    1. Department of Biology, Friedrich-Alexander-University Erlangen-Nuremberg, Erlangen, Germany
    Search for more papers by this author
  • Jens Kossmann

    Corresponding author
    1. Institute of Biotechnology, Department of Genetics, Stellenbosch University, Stellenbosch, South Africa
    • Department of Biology, Friedrich-Alexander-University Erlangen-Nuremberg, Erlangen, Germany
    Search for more papers by this author

Correspondence (Tel +27 (0)21 8083836; fax +27 (0)21 8083835; email kossmann@sun.ac.za)


As the world's second most abundant biopolymer, starch serves as food, feed and renewable resource for bioenergy production and other industrial applications. Unlike storage lipids, starch is stored in the form of semi-crystalline granules, which are tissue- and species-specific in number, shape and size. Over the last decades, most biosynthetic and degradative enzymes of starch metabolism have been identified in the model species Arabidopsis thaliana. Based on this, biotechnological applications have arisen that led to a number of transgenic crop plants with elevated starch content or improved starch quality. Irrespective of this great success, there are still numerous open questions including the regulation of starch metabolism, the initiation of granule formation, the regulation of granule shape and size and many more, which will be tackled over the next decades. Here, we briefly summarize current knowledge concerning starch metabolism and its regulation and biotechnological use.


Starch is one of the most abundant constituents of the world's crops (Jobling, 2004). The majority of starch harvested in crops is consumed directly as food or feed, but a significant proportion is also directed into industrial applications (Marz, 2006), with increasing proportions being used as feedstock for bioethanol production. While the food, paper and textile industries consume a large amount of starch in their manufacturing processes, application in other industries has been limited by far more competitively priced petroleum-based products (Visser and Jacobsen, 1993). Likely future increases in the cost of fossil fuels will make starch an attractive alternative as a raw material in industrial applications (Röper, 2002) other than those it is already used for. Starch often requires chemical or physical treatment to alter its physiochemical properties (Jobling, 2004). To establish it as a more viable alternative to fossil fuels and petroleum-based polymers, it would, therefore, be advantageous to either increase the starch content of plant organs from which starch is usually extracted (mostly corn and potato tubers), or modify its structure in planta to make post-harvest treatments unnecessary.

In plant metabolism, starch is important both for short- and long-term storage of carbohydrates, which can be accessed to drive glycolysis and biosynthesis when photosynthesis cannot. Starch is stored in the plastids of both photosynthetic and non-photosynthetic organs including seeds, fruits, tubers, roots and leaves. Its structure typically varies between species and even organs within the same plant (Kossmann and Lloyd, 2000). Leaf starch granules are generally smaller and easier to degrade as its primary function is to act as a carbohydrate reserve during times of darkness (Smith et al., 2005). This ‘transitory starch’ is broken down to glucose, which can be fed into glycolysis, providing energy to cells when photosynthesis does not occur. Long-term storage of starch usually occurs in amyloplasts that are found in organs such as potato tubers, cassava roots and cereal seed endosperm.

While starch is almost solely composed of glucans, a large diversity of granules from different species has been observed. The reactions involved in catalysing starch synthesis are fundamentally the same in all plants so the structural variability that exists must come from the relative proportions of the enzymes involved in starch metabolism, differences in the substrate specificities of the enzymes between species or the involvement of as-yet unidentified proteins/enzymes. These changes affect the physicochemical properties of the starches manufactured in the different species and therefore elucidating the differences in enzymes between species might help to understand how they affect starch structure. As most commercially produced starch is from maize, wheat, rice and potato, the carbohydrate metabolism of these species has been studied extensively. However, some of the most important discoveries in starch metabolism have come from studies performed in the model plant species Arabidopsis thaliana and the unicellular green algae Chlamydomonas reinhardtii.

The starch granule is composed of two distinct glucans, amylose and amylopectin, amylose being almost linear in nature, whereas amylopectin is branched, with the branches being arranged in a highly ordered fashion, giving amylopectin a crystalline nature. In addition, depending on the plant organ, amylopectin can contain significant proportions of covalently attached phosphate. It has been demonstrated in rice and potato that starch from leaves contains less than 15% amylose, while that found in storage organs can contain between 11 and 37% (Slattery et al., 2000).

We will briefly describe pathways of starch synthesis and degradation, highlight progress in understanding the regulation of starch metabolism and discuss biotechnological possibilities to optimize plant organs with respect to the amount of starch they accumulate, as well as possibilities to modify starch structure to generate tailor-made starches.

Current models of starch metabolism

Starch synthesis is achieved through the action of several plastid-localized enzymes. It is generally accepted that it is synthesized mainly from ADP-glucose (ADPG), the substrate for several isozymes of the starch synthases, which catalyse the polymerization of the glucose molecules. The linear chains are then branched by different isozymes of branching enzymes. The first randomly branched glucans are subsequently trimmed by the isoamylases to arrange the branches in a highly ordered fashion. Phosphate esters in the amylopectin are subsequently introduced by glucan, water dikinase (GWD) at the C-6 position of the glucose monomers, and by phosphoglucan, water dikinase (PWD) at the C-3 position. Starch phosphorylase can also introduce some glucose molecules into the various glucans utilizing glucose 1-phosphate (G1P) as glucosyl donor; however, the relative contribution of this enzyme to glucan elongation is small. Our current knowledge on starch granule formation has been excellently reviewed by Zeeman et al. (2010); and by Stitt and Zeeman (2012). We will focus therefore in this review on the delivery of precursors for starch synthesis in the different types of tissues in which starch synthesis occurs, as well as giving a short account on how starch is re-mobilized. All the genes from Arabidopsis coding for the enzymes named below are listed in Table 1. In photosynthetically active tissue, the hexoses used to synthesize starch are generated by photosynthesis. Fructose 6-phosphate (F6P), an intermediate of the reductive pentose phosphate pathway, is diverted by the enzymes phosphoglucoisomerase (PGI), phosphoglucomutase (PGM) and ADP-glucose pyrophosphorylase (AGPase) into glucose 6-phosphate (G6P), G1P and ADPG to generate the substrate for starch synthesis (see Figure 1). In storage organs, such as seed endosperm or potato tubers, however, the carbon for starch synthesis must be supplied from the cytoplasm (see Figure 2). It is generally accepted that, depending on species, either G6P or G1P is imported into the amyloplast through the action of a hexose phosphate transporter (Hill and Smith, 1991; Kammerer et al., 1998; Kosegarten and Mengel, 1994). A maize G6P transporter was isolated and shown to facilitate a 1 : 1 exchange of imported G6P for organic phosphate or triose phosphate into the plastid (Kammerer et al., 1998). It was also shown that the plastids from maize endosperm can take up G1P (Denyer et al., 1996). Taking this into account, it was surprising to find that cereal endosperm represents an exception to the rule in that the major precursor of starch is imported ADPG produced by an extraplastidial AGPase enzyme (Beckles et al., 2001; Denyer et al., 1996; Shannon et al., 1998; Thorbjørnsen et al., 1996). In other species, G6P is the major hexose phosphate imported by most non-photosynthetic plant tissues including the amyloplasts of pea embryos, cauliflower buds and sweet pepper (Hill and Smith, 1991; Neuhaus et al., 1993; Quick and Neuhaus, 1996). The G6P imported in this way is converted to G1P through the action of PGM (Caspar et al., 1985; Harrison et al., 2000; Salvucci et al., 1990). It was thought that in the case of wheat endosperm and potato tuber plastids, G1P was preferentially imported as the substrate for starch synthesis (Naeem et al., 1997; Tetlow et al., 1994). In potato tuber plastids, this notion was apparently contradicted when silencing of the plastidial PGM enzyme led to a large decrease in starch indicating the necessity of conversion of G6P into G1P in that compartment, apparently ruling out the import of the latter (Tauberger et al., 2000). Again, an apparent confirmation that G6P is preferentially taken up by the potato tuber plastids came from the silencing of the cytosolic isoform of PGM where starch was also decreased (Fernie et al., 2002a). Interestingly, when both isoforms of PGM were simultaneously silenced, starch content recovered to near wild-type levels (Fernie et al., 2002b). It is possible, however unlikely, that ADPG or UDP-glucose (UDPG) is imported into the amyloplasts in so doing bypassing the need for PGM and supporting starch synthesis. UDPG import would require metabolism by starch synthases or plastidic UDP-glucose pyrophosphorylase (UGPases). Recently, a chloroplastic UGPase has been identified in Arabidopsis (Okazaki et al., 2009). Conversion of UDPG into G1P requires sufficient plastidic PPi, which is quite unlikely. A more likely explanation is that under normal circumstances, G6P is imported into the plastid and converted to G1P by PGM; however, when PGM is lacking, a G1P transporter is up-regulated allowing for its import directly (Fernie et al., 2002a). It seems possible that two transporters are present in potato tuber amyloplasts, one for G1P and the other for G6P, as radiolabelled supplied G1P to potato tuber parenchyma cells is efficiently incorporated into starch (Fettke et al., 2010). Other reactions that are necessary for driving starch synthesis in amyloplasts are the import of ATP, which is catalysed by a ADP/ATP translocator (Neuhaus et al., 1997) and the removal of pyrophosphate generated by AGPase catalysed by an inorganic pyrophosphatase (George et al., 2010; Schulze et al., 2004).

Table 1. Starch-related genes in Arabidopsis and their potential regulation by phosphorylation and redox state
LocusEnzymeExperimentally validated phosphorylationRedox sensitiveReference
  1. 1. Heazlewood et al., 2008; 2. Lohrig et al., 2009; 3. Glaring et al., 2012; 4. Hädrich et al., 2012; 5. Sparla et al., 2006; 6. Silver et al., 2012; 7. Reiland et al., 2009.

At5g46110Triose phosphate/phosphate translocator (APE2)   
At5g54800Glucose-6-phosphate translocator 1S27 1
At1g80300ATP-ADP antiporter (NTT1)S592, S599, S611, S615 1, 7
At1g15500ATP-ADP antiporter (NTT2)S589, T605, T606, S610 1, 7
At1g61800Glucose-6-phosphate translocator 2   
At4g24620PhosphoglucoisomeraseS595 1, 7
At5g51820Phosphoglucomutase 1 (PGM1)S179, S181, Y189 1, 7
At1g70820Putative phosphoglucomutaseS182 1
At5g48300ADP-glucose pyrophosphorylase small subunit 1S147, S149, T231, yes (C81)1, 2, 4
At1g05610ADP-glucose pyrophosphorylase small subunit 2   
At5g19220ADP-glucose pyrophosphorylase large subunit 1S428 1
At1g27680ADP-glucose pyrophosphorylase large subunit 2   
At4g39210ADP-glucose pyrophosphorylase large subunit 3S77 1
At2g21590ADP-glucose pyrophosphorylase large subunit 2   
At5g09650Inorganic pyrophosphatase (chloroplast)   
At1g32900 Granule-bound starch synthase 1   
At5g24300Starch synthase I yes3
At3g01180Starch synthase IIS63, S65, T77 1
At1g11720Starch synthase IIIS543, S544, T548yes1, 3
At4g18240Starch synthase IV   
At5g65685Starch synthase IV-like   
At3g20440Branching enzyme I   
At5g03650Branching enzyme II yes3
At2g36390Branching enzyme III   
At2g39930Isoamylase I yes3
At1g03310Isoamylase II yes3
At4g09020Isoamylase III   
At5g04360Limited dextrinaseS567, S571yes2
At1g69830α-Amylase 3 (AMY3)S56yes1, 3
At3g23920β-Amylase 1 (BAM1)S19, T23, S25, S26, S31, S34, T52, S55, S59, T90, Y91yes (C32–C470)1, 5
At4g00490β-Amylase 2 (BAM2)   
At4g17090β-Amylase 3 (BAM3)S465yes2; 3
At5g55700β-Amylase 4 (BAM4)   
At1g10760Glucan water dikinase (AtGWD1)S161 1
At4g24450Phosphoglucan water dikinase (AtGWD2)T609, S612 1
At5g26570Phosphoglucan water dikinase (AtGWD3)   
At3g52180Phosphoglucan phosphatase (SEX4) yes (C198–C130)6
At3g01510Phosphoglucan phosphatase (like SEX four 1, LSF1)   
At3g10940Phosphoglucan phosphatase (like SEX four 2, LSF2)   
At5g64860Disproportionating enzyme 1   
At2g40840Disproportionating enzyme 2S245, S251, T706, S708, S716 1
At3g29320α-Glucan phosphorylase 1 (chloroplast)   
At3g46970α-Glucan phosphorylase 2 (cytosol)T564, S565, T605, T609 1
At5g16150Glucose transporter (GLT1)S84, S85 1
At5g17520Maltose transporter (MEX1)S76, S78 1
At5g17523Similar to MEX1   
Figure 1.

Model of light-dependent starch synthesis in leaves. Simplified cartoon of the pathways of carbohydrate partitioning between starch and sucrose in leaves and mechanisms which regulate carbon flux on the allosteric level (green arrows indicate stimulation, red arrows inhibition) or by post-translational modification (black arrows indicate change in redox status, black arrows protein phosphorylation). Fdx, ferredoxin; FTR, ferredoxin-thioredoxin reductase; NTRC, NADP-thioredoxin reductase C; ADPG, ADP-glucose; F16BP, fructose 1,6-bisphosphate; F26BP, fructose 2,6-bisphosphate, F6P, fructose 6-phosphate; G1P, glucose 1-phosphate; G6P, glucose 6-phosphate; S6P, sucrose 6-phosphate; T6P, trehalose 6-phosphate; TP, triose phosphate; UDPG, UDP-glucose.

Figure 2.

Model of starch degradation and synthesis in storage tissue. Sketch of the pathways of starch metabolism in storage tissues. AGPase, ADP-glucose pyrophosphorylase; DPE1 and 2, disproportionating enzyme 1 and 2; GBSS, granule-bound starch synthase; GWD, glucan water dikinase; Hxk, hexokinase; ISA, isoamylase; PGM, phosphoglucomutase; PWD, phosphoglucan water dikinase; SBE, starch branching enzyme; SS, starch synthase; ADPG, ADP-glucose; G1P, glucose 1-phosphate; G6P, glucose 6-phosphate.

Initiation of starch granules is still enigmatic. Early reports on the occurrence of a glycogenin-like protein suggested a similar mechanism as for initiation of glycogen biosynthesis in mammalian cells (Chatterjee et al., 2005). Mutant analysis, however, could not confirm a role of plant glycogenin-like proteins in starch granule initiation but rather suggests a role in synthesis of secondary cell walls (Brown et al., 2005). The current view is that specialized starch synthases (III and IV in Arabidopsis) are involved in the initiation of starch granules (Szydlowski et al., 2009).

Starch degradation has been thoroughly studied in cereal endosperm, Arabidopsis and some data have also been generated using transgenic potato plants. In cereal endosperm, the situation is different from vegetative tissues studied from other plants, as this tissue is decompartmentalized when starch is mobilized, and cell wall or vacuolar proteins have access to the starch granules. The first step of starch mobilization in Arabidopsis leaf and potato tuber tissue is catalysed by GWD and PWD (Figure 2), which phosphorylate the starch and thereby destabilize the surface of the granules making them accessible to different hydrolases (Hejazi et al., 2008, 2009). Mainly β-Amylases hydrolyse the α-1,4-glucosidic bonds to produce maltose, which is the main product of starch degradation in Arabidopsis leaves. The branch points are hydrolysed by debranching enzymes, isoamylase and pullulanase. The phosphate residues in the starch are removed by two specific phosphatases, SEX4 and LSF2. In Arabidopsis, the maltose generated is exported from the plastids by a maltose transporter protein (Nittyllae et al., 2004) and further metabolized by a cytosolic disproportionating enzyme (DPE2) to yield the soluble cytosolic heteroglycan and glucose. The heteroglycan serves as intermediary carbon storage. The turnover of starch in Arabidopsis leaves has excellently been described by Stitt and Zeeman, 2012 and is summarized in Figure 2. It is, however, not clear at the moment whether the mechanisms described for Arabidopsis generally apply to other plant species.

Current knowledge of regulation

As outlined above, transitory and storage starch are produced in leaves and sink tissues, respectively. Both types of starch serve different functions. While leaf starch, produced during the light period, serves as the major carbon store in most plants for maintenance of metabolism during the night, storage starch serves as long-term carbon storage to allow plants to survive adverse environmental conditions or to support the next generation. Due to these different functions, tissue-specific regulatory mechanisms are likely to operate to adjust starch accumulation according to the specific needs. Due to the pioneering work of Jack Preiss (elucidating the regulation of AGPase), Thomas ap Rees (unravelling the metabolic regulation of sucrose metabolism), Mark Stitt (deciphering the regulation of carbon partitioning), Chris Sommerville (introducing Arabidopsis mutants to unravel pathways of starch) and their colleagues, a complex picture of the regulation of starch is arising. The most complete picture of starch metabolism has been obtained for the model plant Arabidopsis thaliana. Arabidopsis stores a substantial amount of photoassimilates in the form of transitory starch. Interestingly, the amount of starch accumulated during the day is more or less completely consumed during the night. This implies that Arabidopsis is able to predict at the end of the day, the amount of starch it will need to support metabolic processes during the following night. This opens the question of how Arabidopsis is able to sense the amount of starch produced and to balance the rate of synthesis and degradation. Light-dependent starch synthesis is tightly linked to the reductive pentose phosphate cycle and sucrose synthesis. In C3 plants, triose phosphates are produced as the primary product of photosynthesis. While most of the triose phosphates remain within the cycle, the excess is partitioned between sucrose and starch synthesis. Flux of carbon between both pathways is regulated at multiple levels including allosteric regulation of key enzymes (Figure 1). It is assumed that chloroplastic triose phosphates are exchanged against cytosolic phosphate via the triose phosphate translocator. Imported phosphates serve as substrates for the reductive phosphorylation of ADP and hence ATP synthesis and simultaneously inhibit activity of (AGPase). Cytosolic triose phosphates are converted to sucrose via six consecutive biochemical steps. Finally, sucrose is exported to sink tissues through the phloem system. When the rate of synthesis exceeds demand in sink tissues, it is believed that sucrose accumulates in mesophyll cells. High sucrose levels serve as a signal to stimulate starch synthesis and adjust photosynthetic activity to the overall needs of the plant. This adjustment involves the allosteric inhibition of sucrose-6-phosphate synthase, which leads to the accumulation of cytosolic hexoses. As a consequence of elevated hexoses, the signal metabolite fructose-2,6-bisphosphate is synthesized and inhibits the further conversion of fructose-1,6-bisphosphate into fructose-6-phosphate. Hence, the level of cytosolic triose phosphates increases leading to the inhibition of triose phosphate export from chloroplasts. This causes an increase in the plastidic phosphate-to-triose-phosphate ratio and results in allosteric activation of AGPase with a subsequent stimulation of starch biosynthesis. This classical view explains the metabolic balance between starch and sucrose metabolism and suggests that transitory starch serves as an overflow valve for excess photoassimilates. However, it fails to explain how plants can adapt their starch content according to future needs. There is even evidence that the number of starch granules per chloroplast is relatively constant (Crumpton-Taylor et al., 2012). Thus, mechanisms must exist to sense the total amount of starch (granule size or weight?) and the number of granules.

Work in several research areas over the last years has provided first insights into the complexity of the regulation of starch metabolism, but the current picture is like a jigsaw puzzle with many of its pieces missing. It has long been known that expression of starch biosynthetic genes is regulated by sugars and the circadian clock. Sugar sensing and signal transduction pathways have been intensively studied. Unlike hormone and light signalling, sugar signalling pathways are only vaguely understood. This may be due to the dual role of sugars, being signals and metabolites and also due to extensive crosstalk with other signalling pathways. Sucrose is the main transport sugar in plants and is implicated in the regulation of plant development and physiology (for review see Wind et al., 2010). This regulation includes transcriptional and post-transcriptional processes and may involve the proteasome-mediated degradation of target proteins (Hirano et al., 2011). Based mainly on studies in Arabidopsis thaliana, a tentative model for sucrose sensing has been developed. While the primary sucrose sensor still needs to be identified, one signal transduction pathway most likely proceeds via trehalose-6-phosphate (T6P). Lunn et al. (2006) nicely showed that T6P levels positively correlate with increasing sucrose concentrations in leaves of Arabidopsis thaliana plants, which was paralleled by a redox activation of AGPase and increased starch accumulation. Sugar-mediated redox activation of AGPase was first described in potato (Tiessen et al., 2002) and supported by further studies in other plants. Today, 11 starch-related enzymes have been shown to be redox regulated (see Table 1), but AGPase remains the best studied (Figure 1). Two Trx domain proteins [plastid-localized NADP-thioredoxin reductase C (NTRC) and thioredoxin f1 (Trxf1)], have been shown be involved in sucrose and light-dependent regulation of AGPase, respectively (Michalska et al., 2009; Thormählen et al., 2012).

In Arabidopsis seedlings, T6P is a potent inhibitor of SNF1-related kinase 1 (SnRK1) (Zhang et al., 2009). SnRK1 is a central integrator of stress and energy signalling and is involved in switching cellular metabolism from anabolism to catabolism (Baena-Gonzalez et al., 2007). Thus, high levels of sucrose would lead to the inhibition of SnRK1 via T6P and thereby starch synthesis would outperform starch degradation leading to a net increase in starch content (Figure 3). This is supported by data obtained from transgenic Arabidopsis plants with elevated T6P content, which showed an increased starch level. Furthermore, microarray analysis revealed that biosynthetic pathways that are up-regulated by elevated T6P levels are typically down-regulated by SnRK1 (Baena-Gonzalez et al., 2007; Zhang et al., 2009). Interestingly, T6P inhibition of SnRK1 could not be observed in mature Arabidopsis leaves. This led the authors to postulate that an intermediate factor, not expressed in mature leaves, mediates the inhibitory function of T6P (Zhang et al., 2009). In source leaves of Arabidopsis thaliana, T6P stimulation of starch synthesis has been attributed to the redox activation of AGPase, which according to Kolbe et al. (2005) is dependent on SnRK1. Thus, in this case, SnRK1 would be required for anabolic rather than catabolic reactions. Looking into different plant species, the story becomes even more complicated. In potato tubers, sucrose-mediated redox activation of AGPase was shown to be SnRK1 dependent (Tiessen et al., 2003) and over-expression of SnRK1 in potato tubers resulted in elevated rather than reduced starch levels (McKibbin et al., 2006). Thus, it seems as whether SnRK1 regulates anabolism and catabolism in a tissue- and/or species-specific manner. Recently, Baena-Gonzalez et al. (2007) could show that approximately 1000 Arabidopsis genes are regulated in a SnRK1-dependent manner. Among those genes, a group of bZIP transcription factors (TFs) was found, which previously has been shown to be post-transcriptionally regulated by sucrose (Rook et al., 1998). In addition of being transcriptionally regulated, these TFs are activated in a SnRK1-dependent manner (for summary see Hanson and Smeekens, 2009). Over-expression of the sucrose-regulated bZIP11, one prominent member of the above mentioned TF family, results in significant changes of gene expression (Hanson and Smeekens, 2009). Furthermore, a detailed biochemical analysis of bZIP11 overexpressing plants revealed a decreased T6P level in transgenic plants, indicating that bZIP11 may regulate trehalose metabolism (Ma et al., 2011). Although sugar-induced accumulation of transcripts of starch biosynthetic genes has been published in many studies, these genes were not found to respond to changes in SnRK1 or bZIP expression. Furthermore, Debast et al. (2011) observed no significant changes in the expression of starch-related transcripts in tubers with substantially reduced levels of T6P. This implies that sucrose-mediated transcriptional regulation of starch biosynthetic genes does not proceed via the described T6P/SnRK1 signal transduction pathway.

Figure 3.

Hypothetical regulation of starch metabolism by trehalose 6-phosphate and SnRK1. Green arrows indicate stimulation, green arrows inhibition, SnRK1, SNF1-related kinase.

More than 30% of Arabidopsis genes expressed in rosette leaves are diurnally regulated (Bläsing et al., 2005). This includes genes involved in redox regulation, nutrient acquisition and central carbon metabolism. Interestingly, expression of approximately 50% of clock-regulated genes was modified in the starchless phosphoglucomutase (pgm) mutant, which accumulates high levels of soluble sugars in the light (Bläsing et al., 2005). This suggests crosstalk between clock and sugar signals as highlighted in Stitt and Zeeman (2012). Among the different clock-regulated starch genes, granule-bound starch synthase I (GBSSI) has been intensively studied in several plant species. In Arabidopsis, it could be shown that oscillation of GBSSI expression and activity is dependent on the clock genes CCA1 and LHY (Tenorio et al., 2003). Oscillation of GBSSI could also be observed in potato leaves and tubers (Ferreira et al., 2010). However, in tubers, oscillation of GBSSI transcripts was linked to diurnal changes in sucrose supply rather than to a direct circadian control. The circadian control of starch degradation was nicely shown by Graf et al. (2010). Analysis of wild-type Arabidopsis plants grown under abnormal day length conditions (longer or shorter) revealed that starch reserves were depleted approximately 24 h after the last dawn. In addition, Arabidopsis mutants lacking the clock components LHY and CCA1 depleted their starch pools before dawn, which resulted in carbon starvation at the end of the night period which may be the cause of the observed growth reduction in the mutant (Graf et al., 2010). These results indicate that expression of starch-related genes and the rate of starch turnover are controlled by the circadian clock. Despite this, the protein amount of most starch enzymes does not directly follow the circadian regulation of the transcripts. Therefore, additional layers of regulation must be postulated, which might include regulation of translational efficiency and post-translational modifications. One such post-translational modification is protein phosphorylation. So far, 45 putative protein kinases and 21 potential protein phosphatases have been predicted to be localized in plastids (Schliebner et al., 2008). Phospho-peptides of several starch-related proteins have been identified in Arabidopsis (Table 1). Some of these phospho-peptides resemble the extreme N-terminus of the proteins and therefore are most likely derived from transit peptides, making their involvement in starch metabolism unlikely. But beside these N-terminal phospho-peptides, a number of additional phospho-peptides with so far unknown biological function have been described. In cereal, the situation is different. Here, it could be shown that protein phosphorylation is important for protein complex formation and that these protein complexes might be important for amylopectin biosynthesis (Tetlow et al., 2004, 2008). Based on this, it is tempting to speculate that protein phosphorylation plays also an important role in Arabidopsis and it will be interesting to see how mutations in plastidic protein kinases and phosphatase will influence starch metabolism.

The ultimate goal in analysing the regulation of starch synthesis and/or degradation would be the identification of a key regulatory transcription factors. So far, a general ‘starch switch’ has not been identified in Arabidopsis. Co-expression analysis in rice, however, led to the identification of the rice AP2/EREBP transcription factor called RSR1 (rice starch regulator 1). In rsr1, knockout mutants increased amylose content and an altered amylopectin structure could be observed (Fu and Xue, 2010). Thus, RSR1 seems to regulate at least some starch-related enzymes.

Biotechnological approaches to modify starch content and structure in plant tissues

So far, there is only one example in which altered expression of a potential regulator of starch synthesis resulted in modified starch quality in rice endosperm (see above, Fu and Xue, 2010). Due to limited understanding of regulatory mechanisms, attempts to increase starch quality and quantity concentrated on provision of precursors and modulation of starch biosynthetic/degradative enzymes. Most of the biotechnological approaches to increase the starch content of crop storage organs so far reported have centred on expressing an AGPase from Escherichia coli that responds differently to allosteric effectors such as 3-PGA or Pi. It should be more active in plant tissues and consequently lead to increases in starch, if it represents the limiting factor of the pathway. However, mostly negative results had been reported contradicting the initial study carried out in potato tubers (for review see Zeeman et al., 2010).

More promising results have been obtained when nucleotide turnover, de novo synthesis or transport were manipulated. The first example relates to the down-regulation of a plastidial adenylate kinase in potato plants, which led to a doubling of the starch content in the transgenic tubers as compared to wild-type controls (Regierer et al., 2002). The significance of those findings was further supported by studies on a respective Arabidopsis knockout mutant that showed increases in photosynthesis and growth (Carrari et al., 2005). Secondly, 25% increases in starch were observed in potato tubers, where uridine monophosphate synthase (UMPS) was down-regulated, a key enzyme in de novo pyrimidine synthesis (Geigenberger et al., 2005). Thirdly, inconsistent increases in starch were determined in potato tubers, where an ADP/ATP translocator had been up-regulated (Tjaden et al., 1998). The inconsistencies in increases in starch were abolished when the same adenylate translocator was simultaneously up-regulated with a G6P/Pi translocator, which also led to increases in starch of around 25% (Zhang et al., 2008). The starch content was further improved when this ‘pull’ approach was combined with two different ‘push’ approaches in which increased sucrose is supplied from source leaves through leaf-specific inhibition of AGPase or mesophyll-specific expression of a cytosolic inorganic pyrophosphatase (Jonik et al., 2012). This study demonstrates that ‘gene-stacking’ can lead to additional effects in starch accumulation. Whether these approaches also lead to increases in starch in embryo or endosperm tissue has not yet been reported.

Recently, it has been reported that ectopic expression of Arabidopsis starch synthase IV, which seems to be mainly responsible for starch granule initiation (Szydlowski et al., 2009), leads to an 30% increase in starch content in potato tubers (Gamez-Arjona et al., 2011). These results seem to be questionable although, as the same tubers showed no differences, or often decreases, in dry matter contents. As starch normally constitutes at least 70% of the dry matter of potato tubers, it would be expected that such a large increase in starch accumulation would be accompanied by an increase in dry mass, as was for instance shown in the study with plants decreased in plastidial adenylate kinase (Regierer et al., 2002). Indeed, the starch content of tubers in breeding programs or other industries is determined via the density (under-water weight) of the tubers, as starch has a higher density than water. A 30% increase in starch without any change in dry matter content would result in tubers, which dry mass is composed of more than 90% starch? For these reasons, the reported results have to be interpreted cautiously and further experimentation is needed as tuber density determinations were not undertaken, and the procedure by which dry matter content was determined was not described in the study.

Many mutants have been described in Arabidopsis that display a starch excess phenotype and are mostly affected in enzymes involved in starch mobilization (Stitt and Zeeman, 2012). This technology has consequently been successfully applied to fodder crops to increase the content of easily digestible carbohydrates. GWD, the enzyme initiating starch granule breakdown, has been down-regulated in transgenic clover (Trifolium repens), alfalfa (Medicago sativa) and ryegrass (Lolium perenne; Frohberg et al., 2003), as well as in silage maize (Frohberg and Baeuerlein, 2006). In all of these cases, the down-regulation of GWD led to tenfold increases in the starch content in leaves. In a more recent study undertaken in maize even higher increases of starch, for example, 22- to 42-fold, have been reported upon down-regulation of GWD (Weise et al., 2012). This, however, might be explained by differences in growth conditions leading to lower basal levels of starch in the wild-type control plants. In addition, diurnal starch levels range from 1 to 10 mg/g fresh weight (Dinges et al., 2003). Therefore, fold changes might significantly depend on the time point of harvest. Sex4 was first described in Arabidopsis and encodes a phosphoglucan phosphatase, which is essential for starch mobilization in Arabidopsis. Interestingly, no starch excess phenotype was obtained when the maize sex4 gene was down-regulated (Weise et al., 2012). This might be explained by species-specific differences or by the fact that results of knockout lines in Arabidopsis have been compared to knockdown lines in maize, potentially carrying some residual enzyme activities which could be sufficient for starch breakdown. When GWD was repressed in transgenic potato tubers, an inhibition of cold-induced sweetening, a process occurring during cold storage of tubers, where starch is converted into sugars was observed. This is of major importance for the potato processing industry, as the sugars are undergoing Maillard reactions with amino acids in the frying process, which leads to the formation of unwanted dark colour and the neurotoxin acrylamide during frying (Mottram et al., 2002).

The only transgenic crop with altered starch metabolism that has been commercialized is corn where a thermophilic α-Amylase is expressed in sub-cellular compartments that are spatially separated from starch. The enzyme only accesses and hydrolyses starch after grinding of the tissue (Lanahan et al., 2006). These ‘self-processing’ maize kernels allow for an economically more favourable conversion of starch to ethanol.

Because most of the major companies that had been active in the field of modifying starch structure have terminated these activities over the last decade, no new technologies have been emerging since this topic has been reviewed the last time (Zeeman et al., 2010), the same examples therefore will be discussed. As amylose and amylopectin are quite contrasting molecules in their physical nature, the major target in changing starch composition is to generate crops that preferably contain only one or the other. A secondary approach is to generate amylopectin with altered chain length distribution and, in addition, containing modified levels of covalently attached phosphate groups. Even though many mutants have been described in various species that contain starches that are composed solely of amylopectin, or that contain starches with a very high amylose content (Kossmann and Lloyd, 2000), it seemed desirable to introduce those traits into cassava and potato, because these are species which are generally difficult to mutate because of their ploidy levels and their genetic complexity. In addition, mutants in other species generally showed penalties in the starch content, which might not occur in the tuberous crops. The first transgenic cassava or potato plants with a modified starch were those which contained amylose-free starch due to the repression of granule-bound starch synthase (GBSS) expression (Raemakers et al., 2005; Tallberg et al., 1998; Visser et al., 1991). No penalties on the starch content were observed; therefore, the potato plants seemed to be ideally suited for commercialization. In the meantime, a non-transgenic amylose-free mutant of potato (Hovenkamp-Hermelink et al., 1987) has been bred into a commercial cultivar and amylose-free potato starch that has been commercialized is derived from this. The commercial production of amylose-free cassava starch is now also possible, because an amylose-free mutant has now been also described in this tuberous crop (Ceballos et al., 2007). High-amylose starch can be produced in potatoes when either isoform A of branching enzyme (38%; Jobling et al., 1999), both isoforms of branching enzyme (75%; Schwall et al., 2000) or, interestingly, when in addition in to the branching enzymes GWD is repressed (90%; Uwer et al., 2006) in transgenic plants.

As the commercialization of modified amylopectin is only feasible in an amylose-free background, we will only discuss those examples that were undertaken in such a way, although there are numerous other reports on a range of transgenic manipulations that affect amylopectin structure. The simultaneous reduction in GBSS, starch synthase II and starch synthase III (SS III) in potato has led to the synthesis of a pure amylopectin with increased relative amounts of shorter chains (Jobling et al., 2002). This resulted in the production of a freeze–thaw-stable potato starch. It was possible to produce high-phosphate amylopectin by simultaneously down-regulating GBSS, starch branching enzyme B (SBE B) and SS III simultaneously down-regulated with GBSS (Soyka et al., 2006), where the phosphate content was higher than if the enzymes were down-regulated individually (Safford et al., 1998; Abel et al., 1996). In contrast to tuber-derived starches, cereal starches are generally low in phosphate. With the identification of GWD and PWD as the starch-phosphorylating enzymes, it was possible to produce cereal starches that contain higher amounts of phosphate through over-expression of those proteins. As the superior quality of potato starch to cereal starches is relying on its phosphate content, this was especially interesting. The over-expression of GWD has been achieved in wheat (Schewe et al., 2003), maize (Frohberg, 2008; Lanahan and Basu, 2005) and in rice (Frohberg, 2008), which lead to the production of starches with unprecedented swelling power (Frohberg, 2008).

Whether such tailor-made starches ever will be commercialized remains unclear because of the high costs that are implicated when transgenic events need to be de-regulated, even though they represent ideally suited renewable resources.


The work of US and JK was supported by the BMBF (SUA 09/018) and the NRF. The authors are thankful to Gavin George and James Loyd for carefully reading the manuscript and for valuable comments.