Starch is the main polymer in which carbon and energy are stored in land plants, algae and some cyanobacteria. It plays a crucial role in the physiology of these organisms and also represents an important polymer for humans, in terms of both diet and nonfood industry uses. Recent efforts have elucidated most of the steps involved in the synthesis of starch. However, the process that initiates the synthesis of the starch granule remains unclear. Here, we outline the similarities between the synthesis of starch and the synthesis of glycogen, the other widespread and abundant glucose-based polymer in living cells. We place special emphasis on the mechanisms of initiation of the glycogen granule and current knowledge concerning the initiation of the starch granule. We also discuss recent discoveries regarding the function of starch synthases in the priming of the starch granule and possible interactions with other elements of the starch synthesis machinery.
In most living organisms, carbon and energy are stored in the form of polysaccharides. In bacteria, fungi and animals, glycogen is the storage polymer while green plants, red algae, and some parasites and cyanobacteria accumulate starch. In their roles as storage polymers, both compounds play a central function in the carbon metabolism of organisms and have as a result been the subject of a large number of studies. This research has focused on the identification of the proteins involved in the synthesis and mobilization of the polymers (Alonso-Casajus et al., 2006; Streb et al., 2008; Wattebled et al., 2008), the regulation of their activities (Gomis et al., 2000; Gómez Casati et al., 2003; Niittyla et al., 2006), the interaction between these proteins (Skurat et al., 2002, 2006; Hennen-Bierwagen et al., 2008; Tetlow et al., 2008) and how changes in their levels of accumulation affect the physiology of the organisms (Schulze et al., 1991; Sulpice et al., 2009). These analyses have also included the characterization of the initiation of the synthesis of the polymers. However, whereas the initiation of glycogen synthesis has been well defined, the initiation of the starch granule still requires clarification.
Here, we bring together current knowledge regarding the initiation of glycogen and starch, highlighting the parallelism and differences between the two processes and focusing on the process of initiation of the starch granule. We will also discuss recent progress in the identification of elements involved in such mechanisms, as well as future perspectives for this line of research.
The structure of the glucose-based polymers starch and glycogen
Both glycogen and starch are glucose-based polymers. However, while glycogen displays a homogenous structure, starch results from the mixing of two structurally different polymers: amylose and amylopectin. In both cases, glucose residues are linked together by α-(1→4) O-glycosidic linkages that are regularly branched in α-(1→6) positions. However, the level of branches and their distribution are not identical in starch and glycogen, implying different structures and properties. While the branching level reaches 9% in glycogen, it remains below 6% in amylopectin and well below 1% in amylose. Moreover, while branches are regularly distributed in glycogen (one α-(1→6) branch every four glucose moieties and an average chain length of 11–15) (Meléndez-Hevia et al., 1993), they are discontinuously dispersed in amylopectin, leading to the formation of regularly alternating amorphous and crystalline lamellae (Fig. 1). Such differences in branch distribution impact the polymerizing capacities of each macromolecule. Indeed, glycogen molecules are limited in size as observed in vivo and as demonstrated by mathematical models (Meléndez-Hevia et al., 1993; Meléndez et al., 1997, 1998). In general, glycogen particle size does not exceed 40 nm in diameter (12 tiers). By contrast, the growth of amylopectin molecules is theoretically infinite because, unlike glycogen, it is not sterically hindered. The discontinuous distribution of branches in amylopectin leads to the formation of ‘clusters’ of amazingly conserved size (9–10 nm) throughout the plant kingdom (Jenkins et al., 1993). Branches are concentrated at the root of the cluster, thus forming the amorphous lamella. The crystalline lamella is formed with the linear part of the glucans that intertwine to create double helices (Fig. 1). Amylopectin comprises the main fraction of wild-type (WT) starches (70–80% of the dry weight) and its degree of polymerization (DP) ranges from 105 to 106 glucose residues. Amylose is thought to adopt single helical structures (which are enhanced by the presence of lipids or iodine), but its conformation in the inner part of the starch granule remains to be clarified. It has a DP of between 600 and 6000, far below that of amylopectin. However, because of its low level of branching, amylose is composed of much longer glucans than amylopectin (several hundred glucose moieties in length in amylose; less than a hundred residues in amylopectin, with an overrepresentation of glucans of DP 5–30). Both polymers are associated in huge semicrystalline, water-insoluble granules whose diameter varies depending on the botanical and genetical source, from < 0.5 μm in the tiny unicellular green alga Ostreococcus tauri (Ral et al., 2004) or in isoamylase mutants of Arabidopsis thaliana, to several μm in WT A. thaliana and Chlamydomonas reinhardtii (Ball et al., 1990; Wattebled et al., 2005, 2008), to several tens of μm in potato (Solanum tuberosum) or pea (Pisum sativum) (Buleon et al., 1998). In general, granule size is homogenous for a given species. However, in wheat (Triticum aestivum), barley (Hordeum vulgare) and oat (Avena sativa), two populations of granules of small (B-type; 2–3 μm) and large (A-type; 15–30 μm) size have been found (Evers, 1973). The rationale of such a dichotomy has still not been unravelled, although it is known that A-type granules are synthesized during the early stages and B-type granules in the late stages of endosperm development (Parker, 1985). Granule morphology is highly diverse: oval in the case of potato and pea; lenticular for A-type and perfect sphere for B-type granules in normal wheat; and polyhedral in maize (Zea mays) (Buleon et al., 1998). In A. thaliana, WT starch granules appear round-shaped and flat (Delvalléet al., 2005; Wattebled et al., 2008) and their size rarely exceeds 5 μm in diameter.
Starch granules are made by the successive apposition of amorphous and semicrystalline growth rings around the hilum (the centre of the structure) in which amylopectin molecules adopt a radial orientation. It is still not clear how amylose and amylopectin interact to produce such a structure, although we do know that amylopectin is required to form a granule while amylose is dispensable (while several types of amylose-free granules have been isolated, no amylopectin-free granules have been obtained). The structure of the hilum has yet to be revealed, probably because this central region of the granule is poorly organized (Buleon et al., 1997). Nor is it clear whether the hilum has its own specific structure or whether it contains specific proteins. Never-theless, it has been proposed that the formation of the hilum is crucial for the priming of granule formation in vivo (Ziegler et al., 2005).
Mechanisms of synthesis: the need for a primer
The synthesis of glycogen and that of starch share common enzymatic activities: elongation of the glucan chains, carried out by glycosyl-transferases that use ADP-glucose (starch synthases and bacterial glycogen synthases) or UDP-glucose (the glycogen synthase of animals and fungi) as the glucosyl donor; and the branching of the chains, which is performed by the glycogen or starch branching enzymes. The complex structure of starch requires, in addition, the action of other enzymes that are absent in the synthesis of glycogen, such as debranching enzymes. For a more detailed review of starch synthesis see Ball & Morell (2003), Tetlow et al. (2004) and Zeeman et al. (2010) and references therein. Both glycogen synthase (GS) and starch synthase (SS) require a free, nonreducing end of the elongating α-(1→4)-D-glucan chain to which to transfer the glucosyl moiety of the activated donor (ADP-glucose or UDP-glucose). This feature determines that the synthesis of glycogen and starch requires the previous formation of a primer that will be subsequently elongated and branched. Animals and fungi are coincident in the strategy employed to initiate the synthesis of glycogen and in both cases the primer is synthesized by the protein glycogenin. Glycogenin was first discovered in rabbit skeletal muscle (Lomako et al., 1988; Pitcher et al., 1988) and is a polypeptide of c. 38 kDa present in an oligomeric form, usually as a dimer. This protein displays a self-glucosylating activity that catalyses the transfer of glucose from UDP-glucose to the hydroxyl group of an internal tyrosine residue. It also catalyses the elongation of this glucose residue to form a maltooligosaccharide covalently bound to the protein. This primer is further elongated and branched by glycogen synthase and a branching enzyme to produce the glycogen molecule. Perhaps the most definitive proof of the essential role of this protein in the synthesis of glycogen comes from studies in Saccharomyces cerevisiae. Two genes encoding glycogenin are found in this organism (Glg1 and Glg2) and only the simultaneous disruption of both genes renders the cells unable to synthesize glycogen (Cheng et al., 1995). Nevertheless, Torija et al. (2005) have shown that a small fraction of colonies from glg1 glg2 mutants can switch to synthesize glycogen. This capacity of glycogen biosynthesis is not stable, indicating the stochastic nature of the synthesis, and that it relies on a combination of several factors, including glycogen synthase and unknown alternative primers (Torija et al., 2005). Glycogen is partly attached to the cell wall, which may provide alternative glucans for the initiation of glycogen synthesis, although these glucans would not fully compensate for the absence of glycogenin (Arvindekar & Patil, 2002).
Glycogenins seem to be exclusive to eukaryotes and they have not been found in any bacterial genome to date. Ugalde et al. have shown that the glycogen priming activity in Agrobacterium tumefaciens resides in the glycogen synthase, which displays a dual activity: a self-glucosylating activity that allows it to synthesize a GS-bound maltooligosaccharide from the donor ADP-glucose, and the proper glucosyl-transferase activity necessary for the elongation (Ugalde et al., 2003). Nevertheless, studies on the initiation of glycogen in bacteria are still scarce and several aspects of this process, such as its regulation and the participation of other elements, have not yet been elucidated.
Priming in starch synthesis: glycogenin-like proteins in plants?
The finding that glycogenin was the initiator protein in the synthesis of glycogen led to the question of whether the initiation of starch synthesis was also mediated through the self-glucosylation of a glycogenin-like protein. Early studies on the developing maize endosperm indentified a 38-kDa protein that displayed UDPGlc-dependent self-glucosylating activity and was proposed to be involved in α-glucan synthesis (Rothschild & Tandecarz, 1994), and a 42-kDa self-glucosylating protein was named amylogenin (Singh et al., 1995). However, further work demonstrated that amylogenin was identical to reversibly glucosylated protein 1 (RGP1), a protein located in the Golgi apparatus that is involved in the biosynthesis of cell wall polysaccharide matrix (Langeveld et al., 2002; Sandhu et al., 2009). More recently, a homology search using mammalian and yeast glycogenin sequences identified several glycogenin-like proteins in the A. thaliana genome. The inhibition of one of them, named plant glycogenin-like starch initiation protein 1 (PGSIP1), using the RNAi technique led to a decrease in the starch accumulated in leaves of transgenic plants (Chatterjee et al., 2005). However, this decrease was detected by iodine staining of leaves and not through more precise enzymatic determination. Moreover, it was unclear whether the inhibition of gene expression was specific to PGSIP1 or applied to other glycogenin-like genes. Finally, the structural gene of PGSIP1, At3g18660, has been shown to encode a member of the glycosyl-transferases family 8 involved, as in the case of amylogenin, in the synthesis of the secondary cell wall (Brown et al., 2005).
Thus, these studies suggest that plants lack glycogenin-encoding genes orthologous to those found in mammals and yeast, and that the priming mechanism of starch synthesis is different from that described for glycogen synthesis in animals and fungi.
The function of starch synthases in the initiation of the starch granule
The first evidence suggesting the possible involvement of starch synthases in the priming of starch biosynthesis came from an analysis of an A. thaliana knock-out mutant lacking starch synthase class IV (SSIV; At4g18240). This mutant was impaired in the synthesis of the normal number of starch granules found in the chloroplasts of wild-type plants (5–7 granules per plastid). Just one huge starch granule (two in some cases) was observed in the chloroplasts of mutant leaves (Roldán et al., 2007) (Fig. 2). This phenotype suggested a function of SSIV in the process of initiation of the starch granule and suggested that this protein could be involved in the priming of starch synthesis.
However, starch synthesis in this mutant was not abolished completely, indicating that it could occur either in a stochastic manner, similar to the synthesis of glycogen in the absence of glycogenin described in the previous section, or in the presence of another element with a function partially redundant to that of SSIV in the priming of starch synthesis.
Further studies have shown that the simultaneous elimination of SSIV and SSIII (At1g11720) completely prevents the synthesis of starch in A. thaliana leaves (Szydlowski et al., 2009). However, this result did not clarify whether SSIII is partially redundant to SSIV in the priming of starch or if it is just a starch synthase required for the stochastic initiation of starch synthesis. The analysis of the different double and triple mutants for SS classes indicated that the elimination of SSIV always determines the formation of one huge starch granule per chloroplast. However, many chloroplasts without visible starch granules could be observed if other SSs were mutated in addition to SSIV, such as in ssI-ssIV, ssII-ssIV and ssI-ssII-ssIV mutant plants (Szydlowski et al., 2009). These results suggest that SSIII is not involved in the priming of starch synthesis but is an SS necessary for the priming of starch synthesis in a stochastic manner. This capability would be reduced with the elimination of either (or both) of the other starch synthases, SSI (At5g24300) and SSII (At3g01180). The finding that SSIII, together with SSI and SSII and other proteins of starch synthesis metabolism, forms part of a multiprotein complex in the wheat and maize endosperm provides a mechanism through which the further elimination of SSI or SSII in an ssIV mutant background could affect the activity of SSIII (Hennen-Bierwagen et al., 2008; Tetlow et al., 2008). In this respect it is worth noting that SSIII exhibits ‘unprimed’ activity without added primer (Szydlowski et al., 2009), analogous to that described for other SSs in the maize endosperm (Imparl-Radosevich et al., 1998). This activity may be responsible for the SSIII-driven stochastic synthesis of the starch granule in the absence of SSIV.
Once SSIV has been identified as the protein required for the correct priming of starch granules in chloroplast, the molecular mechanisms of the priming need to be understood. Self-glucosylation of SSIV, similarly to that described for the glycogen synthase in A. tumefaciens, was not considered, as no radioactive incorporation was detected when the purified recombinant SSIV protein was incubated with ADP-[14C]glucose (Szydlowski et al., 2009). However, we cannot rule out the possibility that this mechanism operates in the plant, under different conditions and in the presence of other as yet undetermined factors.
The SSIV protein displays singular structural characteristics that distinguish it from the other classes of SS. The C-terminal half of the protein shares high homology with other starch and glycogen synthases and contains the catalytic and substrate binding sites (see Leterrier et al. (2008) for a comprehensive analysis of this domain). The long N-terminal half of the protein, by contrast, is unique among the other classes of SS and contains, as its main feature, several long coiled-coil domains. Coiled-coil domains are protein–protein interaction motifs which consist of two or more alpha helices that twist around one another to form a supercoil (Rose & Meier, 2004). Long coiled-coil domains of several hundred amino acids, such as that present in SSIV, are found in a variety of proteins with diverse cellular functions, such as the intermediate-filament proteins of the cytoskeleton, motor proteins such as myosin and kinesin, and the structural maintenance of chromosome (SMC) proteins. In general, and in their biological context, long coiled-coil proteins emerge as a versatile toolbox for the cell, containing scaffolds, levers, rotating arms and possibly springs (Rose & Meier, 2004).
The long coiled-coil motifs are highly conserved in the SSIV of all species studied to date, such as rice (Oryza sativa), wheat, cowpea (Vigna unguiculata) and Brachypodium, indicating that they play an essential role in the function of the protein. In this respect it is worth noting that SSIV has been located at specific points at the edges of the starch granule in A. thaliana chloroplasts, suggesting that this protein is integrated in a big complex of unknown nature (Szydlowski et al., 2009). It would be interesting to test whether the long coiled-coil domains of the protein play a role in the formation of such a complex. This complex could be of primary importance for generating or maintaining a particular glucan at the hilum that would be required for the initiation of starch granule formation. The need for such a specific structure to prime granule build-up was proposed by Ziegler et al. (2005), who proposed that the initiation of the starch granule in vivo could proceed through the formation of spherulitic crystals whose characteristics match that of the hilum. Although their in vitro model requires the heating of native starch polymers to 170°C, which is definitely not compatible with the conditions found in the plastid, one can imagine that the presence of a protein complex in which SSIV is a major contributor could mimic in vitro heating or favour the formation of a specific primer. Granule growth would then organize around this initiating glucan to reach mature size. Such a process would be quite infrequent because the number of granules remains rather low in plastids.
Other elements involved in the process
SSIV appears to be the major contributor to starch granule initiation in A. thaliana leaves. However, other elements, such as isoamylases and starch phosphorylase, have also been suggested to control the priming of starch in plants.
DBEs as determinants of starch-granule priming?
Isoamylases are starch debranching enzymes (DBEs) that specifically hydrolyse α-(1→6) bonds. These enzymes were shown to be involved in both the synthesis and the degradation of starch (James et al., 1995; Mouille et al., 1996; Nakamura et al., 1996; Wattebled et al., 2008). The breakdown of the expression of those isoforms involved in synthesis leads to a significant reduction in starch content and granule size and to the abnormal accumulation of water-soluble polysaccharides whose structure resembles that of animal glycogen. Burton et al. (2002) and Bustos et al. (2004) proposed that isoamylases could be a major determinant of granule initiation because of the reduction in granule size and the increase in their number in the isa1 (isoamylase 1) mutants of barley and potato, respectively. A reduction in granule size is a general feature of mutants affected in ISA1 expression and it is not restricted to barley and potato. It is probably a consequence of the incapacity of the synthesizing machinery to counterbalance the absence of ISA1 and to synthesize amylopectin properly. Thus, starch granules cannot grow correctly because crystalline and amorphous lamellae of the amylopectin cluster cannot be formed correctly. Moreover, the fact that phytoglycogen accumulates at abnormally high levels may explain the increased number of tiny granules in all isa1 mutants. Indeed, phytoglycogen may become an ideal substrate to nucleate the formation of the numerous tiny starch granules found in these mutants. It has been shown that elongated glycogen molecules enhance the formation of starch-like polymers in vitro (Putaux et al., 2006). Each molecule of phytoglycogen may act as a potential primer for the formation of the tiny granules found in all isa1 mutants after the elongation of the outer chains by starch synthases. Therefore, the increased number of granules is more likely to be a pleiotropic rather than a primary effect of the isa1 mutation.
Moreover, this interpretation of the phenotype of both barley and potato mutants is contradictory to that recently reported for a mutant of A. thaliana lacking all isoforms of debranching enzymes (including pullulanase) and one form of plastidial α-amylase (Streb et al., 2008). The authors stated that, in this quintuple mutant, starch granule synthesis was restored (starch synthesis was abolished in the quadruple DBE mutant) despite the complete absence of DBEs. This phenotype rules out the possibility of a determinant function of DBEs in the control of starch granule priming. Although different processes may be at work in source and sink organs, it is highly unlikely that DBEs behave so differently in these systems.
The priming function of starch phosphorylase in the endosperm
Starch phosphorylases are versatile enzymes that could cleave (with Pi) or create (with Glc-1-P) α-(1→4) linkages depending on the ratio of available substrates. Recent work carried out in rice (Satoh et al., 2008) and C. reinhardtii (Dauvillee et al., 2006) suggests a probable function of the plastidial isoform (Pho1) in starch synthesis. The function of Pho1 in the control of the early steps of starch synthesis in the endosperm was questioned after the analysis of the corresponding mutant in rice (Satoh et al., 2008). Indeed, depending on the temperature, this pho1 mutant displays shrunken seeds strongly depleted in starch (at 20°C, while at 30°C most seeds appeared normal). This result suggests that under specific conditions (lower temperature) Pho1 is an essential factor that allows correct synthesis of starch. At a higher temperature (30°C) alternative elements can overcome Pho1 deficiency in the mutant. Moreover, in their study the authors compared the elongation activities of Pho1 and starch synthase IIa (SSIIa). They found that Pho1 is able to produce much longer lineal glucans than SSIIa using maltohexaose (DP6) as a substrate, and with a much higher efficiency. From these results, Satoh et al. postulated that Pho1 is probably required for the synthesis of the initial substrate that would serve to initiate the synthesis of the starch granule in the rice endosperm. However, if Pho1 is actually required in the early steps of starch synthesis in rice, it is not the only factor. Indeed, another factor, as yet unidentified, is the main factor sustaining starch synthesis at higher temperature. Pho 1 and this undetermined factor could act in the form of a protein complex to control starch granule initiation in the rice endosperm. This factor could be SSIV. Two isoforms of this enzyme exist in rice (SSIVa and SSIVb). As in A. thaliana, SSIV may be devoid of elongation activity in vivo, thus explaining why starch synthase activity is not affected in the pho1 mutant of rice. Interaction between SSIV and Pho1 is suspected in A. thaliana. Indeed, the combination of null mutations for both the SSIV and PHS1 genes (the plastidial phosphorylase isozyme is designated PHS1 in A. thaliana, whereas in all other species it is named Pho1) in that plant leads to a marked modification of the starch accumulation phenotype (D’Hulst et al., 2007). ssIV phs1 double mutants accumulate in their leaves three to four times more starch at the end of the day in the form of huge starch granules that are bigger than those found in the ssIV mutant. The molecular basis of such interactions is still not understood and further characterizations are required. Nevertheless, the results obtained in both rice and A. thaliana are consistent with a possible involvement of the plastidial phosphorylase in the initiation of starch granule synthesis.
The starch synthesis pathway is very well conserved in the different phylogenetic groups, from the picophytoplankton alga O. tauri to rice (Ral et al., 2004), so findings obtained in model organisms can be extrapolated to other species. However, the role of starch synthases in the initiation of the starch granule has been studied only for the transitory starch accumulated in A. thaliana leaves, and the function of these enzymes in the initiation of long-term storage starch has not been validated. The metabolic features of source, photosynthetic tissues such as leaves and sink, storage organs such as seeds, endosperms and tubers are very different. So is the timing of starch accumulation in these tissues: diurnal oscillation of synthesis and degradation in leaves and long-term storage in endosperms and tubers. These characteristics could mean that the process of initiation of starch in the two types of organs operates through distinct mechanisms or displays different systems of regulation. A thorough analysis is required of the function of SSIV in endosperms and tubers and the interaction with other elements that could also be involved in this process (such as the phosphorylases mentioned in the previous section) before we can gain an understanding of the initiation of storage starch.
As already mentioned, one of the main characteristics of the SSIV polypeptide is the presence of protein–protein interaction domains, which leads to the question of whether SSIV interacts with other proteins of the starch synthesis machinery or with other as yet unidentified proteins. We need to identify these elements if we are to understand the function of SSIV. In this respect, a possible cross-talk between starch metabolism and plastid division could be envisioned. The coiled-coil domain of the N-terminal region of SSIV shares homologies with the bacterial membrane-associated EzrA protein. This latter protein regulates the formation of the division ring formed by FtsZ in bacteria (Singh et al., 2007). Homologous FtsZ proteins are responsible for plastid division in plants (El-Kafafi et al., 2005; Schmitz et al., 2009). Therefore, an interaction between FtsZ-like proteins and SSIV could arise in plants, allowing both mechanisms to be closely cross-regulated. The rationale of such cross-control could lie in the fact that the presence of dense, water-insoluble semicrystalline starch granules may compromise proper plastid division. Thus, controlling the major element of granule initiation (SSIV) could be of primary importance for the cell to avoid the abortion of plastid division. Such an interaction was suggested in studies performed in potato and rice where the alteration of plastid division was found to affect starch granule properties (size, number and gelatinization) (de Pater et al., 2006; Yun & Kawagoe, 2009). Preliminary work performed in our laboratories on A. thaliana mutants affected in FtsZ-like genes agrees with this idea of an interaction between starch metabolism (transitory starch in that case) and plastid division.
Finally, the initiation of the starch granule may represent a biotechnological target for increasing starch yield in crops. Different approaches have been followed to increase the production of starch. These have included the over-expression of AGPase, the enzyme responsible for the production of ADP-glucose, although the expression of bacterial AGPase in cassava (Manihot esculenta) did not increase the content of starch on a fresh weight basis (Ihemere et al., 2006). More satisfactory results were obtained by increasing the pool of ATP available for starch synthesis, through increasing either the expression of the ATP/ADP transporter (Tjaden et al., 1998) or adenylate kinase activity (Regierer et al., 2002). The elimination of SSIV led to a reduction in the number of starch granules and to a concomitant reduction in the levels of starch accumulated in A. thaliana leaves (Roldán et al., 2007). This suggested that the number of starch granules could represent a regulatory step in the control of starch accumulation. Thus, over-expression of SSIV or another element controlling the number of starch granules could increase the yield of starch. Further studies of over-expression of SSIV in storage organs such as cereal endosperms and tubers are necessary to ascertain the potential application of SSIV in the biotechnological increment of starch yield in crops.