Prospects for increasing starch and sucrose yields for bioethanol production

Authors


*(fax +44 1603 450045; e-mail alison.smith@bbsrc.ac.uk).

Summary

In the short term, the production of bioethanol as a liquid transport fuel is almost entirely dependent on starch and sugars from existing food crops. The sustainability of this industry would be enhanced by increases in the yield of starch/sugar per hectare without further inputs into the crops concerned. Efforts to achieve increased yields of starch over the last three decades, in particular via manipulation of the enzyme ADPglucose pyrophosphorylase, have met with limited success. Other approaches have included manipulation of carbon partitioning within storage organs in favour of starch synthesis, and attempts to manipulate source–sink relationships. Some of the most promising results so far have come from manipulations that increase the availability of ATP for starch synthesis. Future options for achieving increased starch contents could include manipulation of starch degradation in organs in which starch turnover is occurring, and introduction of starch synthesis into the cytosol. Sucrose accumulation is much less well understood than starch synthesis, but recent results from research on sugar cane suggest that total sugar content can be greatly increased by conversion of sucrose into a non-metabolizable isomer. A better understanding of carbohydrate storage and turnover in relation to carbon assimilation and plant growth is required, both for improvement of starch and sugar crops and for attempts to increase biomass production in second-generation biofuel crops.

Introduction

The current demand for starch and sugars as substrates for bioethanol production has re-invigorated research and speculation about factors that determine the yield of these metabolites in crop plants. This paper reviews the ways in which plant biotechnologists have attempted to increase the starch and sugar yields of crops over the last two decades, discusses whether these approaches are likely to be successful in increasing yields in a sustainable manner within the likely lifetime of the first-generation biofuel industry, and considers more radical approaches for the future and areas in which further research is needed.

Background

The market for liquid biofuel made by fermentation of starch and sugars has expanded many-fold in the last few years. The major sources of starch and sugars for bioethanol production are currently maize seeds and sugar cane stems, respectively. Brazil obtains >30% of its road transport fuel from sugar cane, and the United States Department of Agriculture (USDA) predicts that 20% of US maize production in 2006/2007 will be used for bioethanol. Other crops used on a large scale are cassava roots and wheat seeds for starch, and sugar beet roots for sugar. For example, a cassava bioethanol plant in Nanning, China, is expected to achieve a capacity of one million tonnes in 5 years (http://www.worldwatch.org/node/4351).

The expansion of bioethanol production from starch and sugar crops has been accompanied by growing anxiety about the sustainability of this new industry. Concerns include the relatively small reduction in emissions of greenhouse gases that this source of biofuel can achieve, the effect on food availability and prices resulting from this diversion of food crops into bioethanol production, and the effect on global biodiversity of the loss of natural habitats to starch and sugar crops. The precise impact of current biofuel crops on carbon dioxide emissions remains controversial (see, for example, Farrell et al., 2006; Fargione et al., 2008; Searchinger et al., 2008; http://www.grain.org/m/?id=154). However, it is undoubtedly true that starch and sugar can contribute only a very small fraction of liquid biofuel requirements unless vastly more land is devoted to their production. The USDA predicts that >30% of the US maize crop will be used for bioethanol in 2009/2010, but that this will replace only 8% of the US petrol (gasoline) use at that time (http://www.ers.usda.gov/Briefing/Baseline/present2007.htm).

Major efforts are already under way to produce more sustainable liquid biofuels, in greater quantities. These include development of new methods to access sugars in the polysaccharides of plant cell walls, and of new dedicated biofuel crops that will require minimum consumption of fossil fuel during their cultivation and harvest. These developments are described elsewhere in this issue (Durrett et al., 2008; Li et al., 2008; Pauly and Keegstra, 2008).

However, such developments are likely to require 10–15 years to be fully commercialized, and in the meantime starch and sugars from existing food crops will continue to be the major substrate for liquid biofuel production. This review considers whether current knowledge and further research on starch and sugar metabolism in plants can be used to enhance the sustainability of bioethanol production in the short to medium term. Because of this short time-frame, manipulations of single, target genes are considered primarily, rather than long-term breeding options. The possible impact of environmental conditions on the field performance of plants with altered starch and sugar metabolism is not considered. This is clearly important, but little or no relevant data are available.

The most obvious development that would enhance the sustainability of first-generation biofuel crops is an increase in the yield of starch/sugar per hectare without further inputs into the crops concerned. Increases in yields of starch and sucrose have of course been the target of plant breeding, and latterly of plant biotechnology, for many years. Transgenic technology led to a boom in the 1990s in attempts to increase the starch contents and alter the starch properties of cereal crops and potatoes by manipulation of enzymes involved in starch synthesis. Although these efforts met with some apparent success, visible research efforts in this area have declined considerably in recent years. The major biotechnology companies that funded most of the original research are less active in this area, and, to the best of my knowledge, none of this transgenic research has been commercialized. Considerably, less biotechnological effort has been devoted to increasing sugar content in sugar crops. The recent advent of routine methods for transformation of sugar cane has permitted new experimentation, with some promising results.

The first three sections below review attempts to increase the starch content of storage organs at three different levels. These involve manipulation of (i) the pathway of starch synthesis itself, (ii) partitioning between pathways in starch–storage organs, and (iii) partitioning of assimilated carbon between different organs of the plant. Whether improvement in starch processibility – as opposed to yield – is possible and worthwhile is considered in the subsequent section. More radical means of manipulating starch contents of crops are then suggested, some of which may contribute to the viability of new, dedicated biofuel crops as well as existing starch and sugar crops. Finally, means are considered by which sugar accumulation in sugar crops may be manipulated.

Manipulation of the pathway of starch synthesis

The pathway of starch synthesis in developing storage organs is relatively well understood. In all organs apart from cereal endosperms, sucrose entering cells of the developing storage organ is converted to glucose 6-phosphate in the cytosol. This metabolite enters the plastid via a glucose 6-phosphate transporter (Kammerer et al., 1998), and is converted via phosphoglucomutase and ADPglucose pyrophosphorylase (AGPase) to ADPglucose (Figure 1). This process requires ATP, which is imported into the plastid via an adenylate transporter (Tjaden et al., 1998). The glucosyl moiety of ADPglucose is the substrate for synthesis of starch polymers (α-1,4, α-1,6-linked glucans) via starch synthases, starch-branching enzymes and the starch-debranching enzyme isoamylase. The precise mechanism by which the starch polymers organize to form the semi-crystalline starch granule is not understood, but it requires the presence of three to five distinct isoforms of starch synthase and two of starch-branching enzyme (Ball and Morell, 2003; Tetlow, 2006; Tetlow et al., 2004; Tomlinson and Denyer, 2003). In the absence of any one of the isoforms of these enzymes, or of isoamylase, starch polymer structure and granule formation are usually abnormal.

Figure 1.

 Overview of starch synthesis in storage organs.
Sucrose entering the cell is converted via sucrose synthase to UDPglucose and fructose, both of which are then converted to hexose phosphates, the substrate for glycolysis as well as starch synthesis. The hexose phosphate glucose 6-phosphate enters the plastid, where it is converted to ADPglucose via phosphoglucomutase and ADPglucose pyrophosphorylase (AGPase). AGPase catalyses the ATP-requiring step in starch synthesis. ADPglucose is converted to starch (brown oval) via multiple isoforms of starch synthase and starch-branching enzyme, in a process that also requires a debranching enzyme. In the endosperm cells of graminaceous species, most of the synthesis of ADPglucose is catalysed by an isoform of AGPase in the cytosol (pathway shown in brown). ADPglucose enters the plastid via a specific sugar nucleotide transporter.

Starch synthesis in the endosperms of graminaceous species, including cereals, differs from that in other organs in that the synthesis of ADPglucose occurs largely in the cytosol, via a cytosolic form of AGPase. ADPglucose is imported into the plastid via a specific sugar nucleotide transporter (Tomlinson and Denyer, 2003; Figure 1).

Much of the effort devoted to achieving increased starch contents in crop plants has centred on AGPase. The focus on AGPase arose originally from the belief that it catalyses a ‘rate-limiting step’ in the process of starch synthesis in plants in general, and hence that increases in its activity would necessarily result in increases in the rate of starch synthesis. Attempts have also been made to enhance starch synthesis through increased activity of starch synthase, and through increased supply of ATP to the plastid. These attempts are summarized below.

ADPglucose pyrophosphorylase

This enzyme is a hetero-tetramer consisting of two large and two small subunits. It is activated by reduction, and when reduced is subject to strong allosteric regulation by 3-phosphoglycerate and hexose phosphates (activators) and inorganic phosphate (an inhibitor; Ballicora et al., 2004, 2005; Hendriks et al., 2003; Plaxton and Preiss, 1987). Redox activation of AGPase is dependent on the sugar status of the tissue: elevation of either sucrose or glucose triggers signalling pathways that result in increased reduction and hence greater activation of the enzyme (Geigenberger et al., 2005a; Kolbe et al., 2005).

Attempts to increase AGPase activity were first made in potatoes, through introduction of a gene (glgC) encoding the single AGPase protein found in Escherichia coli. The variant used in potato transformation (glgC16) encoded an enzyme that is less sensitive to modulation by effectors than the native E. coli enzyme. Overall, the transformed lines were reported to have an average of 35% more tuber starch than untransformed controls, and some lines had up to 60% more starch. However, there was no obvious correlation between the level of expression of the GlgC16 protein and the level of starch in the tuber (Stark et al., 1992). A subsequent attempt to reproduce these results on a different cultivar of potato did not result in increased starch contents, even though AGPase activity was up to four times higher in the tubers of some transformed lines than in untransformed controls (Sweetlove et al., 1996).

Attempts to achieve higher starch contents through manipulation of AGPase in cereal seeds have made use of a variant of the large subunit of AGPase from cereal endosperm (the product of the Sh2 gene). The large subunit contributes little to the catalytic activity of the enzyme but is important in determining its regulatory properties (Ballicora et al., 2004). Maize plants containing a Ds transposable element in a specific position in the Sh2 gene were screened for revertants in which the Ds element had been excised. The large subunit of one such revertant (rev6) had additional tyrosine and serine residues in the C-terminal region of the protein. Although total AGPase activity in developing kernels was reduced by about 60%, its specific activity was enhanced by 50%, and the recombinant rev6 protein was markedly less sensitive to inhibition by phosphate than the recombinant wild-type protein. The starch content per seed of the line carrying the rev6 allele was consistently higher than in control lines. However, surprisingly, there was no increase in starch as a percentage of seed weight: instead, the rev6 line had significantly heavier seeds (by 11–18%) than control lines (Giroux et al., 1996). The rev6 allele has subsequently been introduced into other cereal species. Transgenic maize, rice and wheat plants expressing this allele in the endosperm show the same characteristics as the original maize line when grown in controlled conditions: individual seed weight and seed yield per plant are increased (Meyer et al., 2004; Smidansky et al., 2002, 2003, 2007). However, field trials of wheat carrying the rev6 allele have had mixed results. Enhancements of yield are seen only under conditions of minimal inter-plant competition and optimal water supply (Meyer et al., 2007).

Three recent reports describe the expression of modified forms of the E. coli glgC gene in the cytosol of endosperm cells of maize (Sakulsingharoj et al., 2004) and rice (Wang et al., 2007), and in cassava roots (Ihemere et al., 2006). Developing seeds of transgenic maize lines showed little change in assayable AGPase activity in the absence of inorganic phosphate, but up to 13 times higher activity in the presence of 10 mm phosphate. Starch content was not measured, but increases of up to 7% in individual seed weight relative to controls were reported for transgenic plants under controlled conditions (Sakulsingharoj et al., 2004). Similar results were obtained in rice. Elevated AGPase activity was detected in transgenic grains only when assays were conducted in the presence of phosphate. Increases of up to 25% in individual seed weight were reported, but starch content was not measured (Wang et al., 2007). In transgenic cassava roots, the activity of AGPase in the presence of inorganic phosphate was seven to eight times higher than in the roots of control plants. There was no increase in the starch content of roots on a fresh weight basis, but total root yields were two- to threefold higher in transgenic than control lines under controlled growth conditions (Ihemere et al., 2006).

In summary, more than two decades of attempts to enhance the starch content of storage organs through manipulation of AGPase have achieved mixed results. With the exception of initial results from potato, increases in AGPase have generally not resulted in higher starch contents on a fresh weight basis. However, at least under controlled conditions, expression of unregulated forms of AGPase has been reported to result consistently in higher yields of cereal seeds, and this may also be the case for cassava roots. It is questionable whether this trait will be of use in the field (Meyer et al., 2007), but it may provide important information about the potential for manipulation of carbon partitioning towards storage organs. In future, expression of forms of AGPase with altered sensitivity to redox activation and hence to endogenous sugar levels may also hold promise for manipulation of carbon partitioning.

Starch synthase

Very few attempts have been made to increase the starch content of storage organs through manipulation of starch synthase. This is perhaps surprising in the light of persistent reports that the maximum catalytic activity of this enzyme is close to the actual flux through the pathway in several starch-storing organs, including potato tubers, pea seeds and cereal seeds (Clarke et al., 1999; Marshall et al., 1996; Yang et al., 2003). Two independent studies in which the activity of starch synthase in wheat endosperm was decreased by heat treatment concluded that the enzyme had a flux control coefficient approaching unity under the conditions of the experiment (Jenner et al., 1993; Keeling et al., 1993). However, the fact that storage organs contain multiple isoforms of starch synthase complicates attempts to enhance its activity. Loss or reduction of activity of individual isoforms in mutant and transgenic plants frequently has profound effects on starch structure, revealing that isoforms have specific but inter-dependent roles in starch polymer synthesis (Ball and Morell, 2003). Overexpression of a single form of the enzyme is therefore likely to have complex effects. This problem is illustrated by experiments in which the glycogen synthase of E. coli (an enzyme that catalyses the same reaction as starch synthase) was expressed in potato tubers. Starch content was decreased rather than increased, and there were major alterations to the structure of starch polymers (Shewmaker et al., 1994).

ATP supply to the plastid

There is good evidence that increasing the supply of ATP to the plastid in potato tubers can enhance tuber starch content. Starch content is very sensitive to manipulation of the activity of the adenylate transporter of the plastid envelope that supplies ATP for the reaction catalysed by AGPase. An increase of 50–60% in transport activity in transgenic tubers expressing an adenylate transporter from Arabidopsis resulted in 16–36% more starch per gram fresh weight (Geigenberger et al., 2001; Tjaden et al., 1998). Transgenic tubers had ADPglucose levels that were twice those of wild-type tubers. Downregulation of expression of plastidial adenylate kinase, an enzyme that catalyses the inter-conversion of ATP with AMP and ADP, also increased the starch content of potato tubers. It resulted in an up to 10-fold increase in ADPglucose levels in tubers, and increases in tuber starch content on a per plant basis that were as high as twofold in field trials (Regierer et al., 2002). These increases are the largest so far reported for any targeted manipulation of starch synthesis.

What lessons can be learned from manipulation of enzymes of starch synthesis so far? First, targets for manipulation have generally been chosen on the basis of informed guesswork rather than systematic analysis of the pathway. Methods available for systematic analysis include metabolic control analysis, which quantifies the contribution of individual enzymes to the control of flux through a metabolic pathway. A good introduction to this method is provided by ap Rees and Hill (1994), and its application to starch synthesis is discussed by Tomlinson and Denyer (2003) and Geigenberger et al. (2004). The widely held belief that AGPase is the ‘rate-limiting step’ in starch synthesis is generally not borne out by quantitative analyses. In potato tubers, metabolic control analysis reveals that the importance of AGPase in controlling flux to starch is less than that of either the adenylate transporter of the plastid envelope or the plastidial adenylate kinase (Geigenberger et al., 2004). In cereal endosperm and pea seeds, the quantitative importance of AGPase in controlling flux is low (Denyer et al., 1995; Singletary et al., 1997). Much of the control of flux may lie with starch synthase, although no attempt has been made to assess the importance of ATP supply in the case of pea or of the ADPglucose transporter of the plastid envelope in the case of cereals. These results should not be taken to imply that manipulation of AGPase levels is pointless. Introduction of a form of an enzyme with modified properties can often result in increases in flux even when the native enzyme is unimportant in the control of flux. However, the focus on AGPase as a target has arguably held up research on other enzymes and proteins that exert equal or greater control over flux to starch.

The second lesson to be learned from attempts to manipulate the pathway of starch synthesis is the need for rigorous analyses of transgenic plants. Surprisingly, a number of studies do not report direct quantitative measurements of either the change in activity of the target protein or the starch content in transgenic lines. Few studies attempt to assess the extent of pleiotropic effects on other enzymes that are likely to affect starch content, or the developmental profile of the change in activity of the manipulated enzyme. In the absence of these analyses, it is impossible to assess whether the manipulation has produced useful results that should form the basis of further studies.

Finally, few of the transgenic plants reported to have elevated starch contents have been subjected to field trials for which results are publicly available. Effects seen under controlled growth conditions may not persist at commercially useful levels in the field: this is nicely illustrated by the study of Meyer et al. (2007) on wheat carrying the Rev6 allele of the gene encoding the large subunit of AGPase from maize endosperm. Early field trials of promising plants are essential if real progress is to be made towards sustainably higher yields of starch.

What are the prospects for achieving significant, sustained increases in the starch content of storage organs through manipulation of starch synthesis? In below-ground organs, the best options may lie in increasing the supply of ATP, and hence the level of ADPglucose, rather than in direct manipulation of single enzymes of the pathway. In potato tubers, it is already apparent that manipulation of adenylate concentrations in the plastid can create substantial increases in starch content. Other research has shown that oxygen concentrations inside potato tubers are sufficiently low to influence many aspects of tuber metabolism, including ATP supply. Adaptive responses to low oxygen levels inside the tuber include a reduced rate of respiration, a low adenylate energy state, and an inhibition of biosynthetic reactions that consume ATP, including starch synthesis (Geigenberger, 2003; Geigenberger et al., 2000). Thus, in addition to manipulations of enzymes that control ATP supply in the plastid, manipulations that affect the oxygen status of the tuber may increase the rate of starch synthesis. In Arabidopsis, overexpression of a gene encoding haemoglobin from a legume increases survival during and after hypoxia – possibly through influence on the adenylate energy state of cells (Hunt et al., 2002). It seems possible that a similar approach might increase the adenylate energy state and hence starch synthesis in potato tubers.

There is no direct evidence that ATP supply exerts important control over starch synthesis in cereal seeds, but it is apparent that these seeds experience low internal oxygen concentrations during development. Under ambient conditions, the oxygen concentration inside developing wheat endosperm was only 2%, and increasing the oxygen concentration around the ear from 21% to 40% resulted in a significant higher ATP:ADP ratio in the endosperm (van Dongen et al., 2004). However, in short-term experiments, this high concentration of oxygen did not result in higher rates of starch synthesis (van Dongen et al., 2004). In developing barley seeds, there was a positive spatial correlation between ATP concentration and starch synthesis, and a negative correlation between ATP and oxygen concentrations (Rolletschek et al., 2004).

A second approach to manipulation of starch synthesis in storage organs is to increase the activities of two or more enzymes simultaneously. Even when individual enzymes exert little control over flux, simultaneous increases in their activities can have large, positive, non-additive effects on flux (Neiderberger et al., 1992). The theory behind such increases is complex, and given the relative ease with which multiple genes can be introduced simultaneously, a ‘suck-it-and-see’ approach may be the fastest way to test this idea.

In some cases, more rigorous analysis may point to individual manipulations that are likely to have a large effect. For example, in wheat grains, there is a strong correlation between the decline in the rate of starch synthesis and the decline in the activity of starch synthase towards the end of grain filling (Tomlinson and Denyer, 2003). Neither the supply of sucrose nor the activities of other enzymes of starch synthesis decline significantly at this point. Given the low activity and the apparently high flux control coefficient for starch synthase in wheat endosperm (Jenner et al., 1993; Keeling et al., 1993), it is tempting to speculate that prevention of the decline in starch synthase activity might prolong the period of starch synthesis in the grain.

Partitioning of carbon to starch within a storage organ

A second general approach to increasing starch synthesis in storage organs is to increase flux out of sucrose into the hexose phosphate pool or to reduce flux through other pathways of sucrose utilization in the cells of the organ, in the hope that more carbon from sucrose will be diverted to starch.

Several attempts to direct more carbon into starch in potato tubers have involved the introduction of enzymes that catalyse sucrose catabolism. Surprisingly, although transgenic tubers expressing sucrose phosphorylase (Fernie et al., 2002) or yeast invertase (Sonnewald et al., 1997), or yeast invertase plus bacterial glucokinase (Trethewey et al., 1998), in the cytosol all exhibit faster rates of sucrose mobilization, all also have reduced starch contents. The reasons for reduced rates of starch synthesis in these lines are complex, but include accelerated rates of glycolysis, accelerated re-synthesis of sucrose via endogenous sucrose phosphate synthase and sucrose synthase (i.e. accelerated futile cycling in and out of sucrose), and a reduced adenylate energy state (Bologa et al., 2003; Trethewey et al., 1999, 2001). In contrast, transgenic potato tubers with elevated levels of uridine nucleotides (due to downregulation of UMP synthase and a resultant stimulation of the uridine salvage pathway) have higher starch contents than controls (Geigenberger et al., 2005b). It was proposed that elevated UDP levels allowed greater rates of sucrose catabolism and hence greater rates of starch synthesis, although there were other complex effects on metabolism that may have affected starch synthesis indirectly. At present, there is no reason to suppose that increasing the capacity for sucrose degradation per se will lead to increased rates of starch synthesis in storage organs.

Relatively few attempts have been made to direct carbon into starch by downregulation of other pathways of sucrose utilization. The consequences of manipulation of pathways of central metabolism – glycolysis, and pathways leading from it – are likely to be complex, and are difficult to predict. For example, introduction of an E. coli pyrophosphatase into the cytosol of potato tuber cells, with the aim of decreasing levels of pyrophosphate (a metabolite involved in sucrose catabolism and inter-conversion of hexose mono- and bisphosphates), resulted in increased starch content in one set of experiments but decreased starch synthesis in a second, independent study (Geigenberger et al., 1998; Sonnewald, 1992). Environmental and developmental factors are likely to account for these differences. Unexpectedly, downregulation in potato tubers of NAD-malic enzyme – an enzyme involved in entry of carbon into the tricarboxylic acid cycle in mitochondria – was reported to result in elevated starch contents (up to 45% above wild-type; Jenner et al., 2001). Enhanced starch contents (20–30% above wild-type) were also reported for tubers in which the metabolic regulator SnRK1 was overexpressed (McKibbin et al., 2006). In both these cases, the precise mechanism remains obscure, and it is not clear whether these are robust phenotypes that will persist under field conditions.

It is not clear whether increases in the starch content of seeds could be achieved by downregulation of synthesis of other storage compounds. There are complex relationships between starch and protein content in legume and cereal seeds. Downregulation of starch synthesis results in changes in protein composition, and downregulation of expression of transcription factors controlling storage protein synthesis also affects starch synthesis (Lohmer et al., 1991; Shewry et al., 1987; Tsai et al., 1978). Neither these studies nor QTL analyses of the chemical composition of cereal seeds (for example, Barriere et al., 2001; Goldman et al., 1993; Séne et al., 2000; Thévenot et al., 2005) have revealed single genes that are good candidates for manipulation to increase partitioning of carbon into starch in seeds, without loss of yield.

Partitioning of carbon into starch-storing organs

The starch-storing organs of crop plants already contain spectacular amounts of starch. Starch forms up to 80% of the dry weight of the endosperms of cereals and the tubers of potato. In cassava roots, it is as much as 45% of the wet weight. Although, as discussed above, there may be scope for increasing flux into starch in these organs – in other words increasing the percentage of the dry weight of the organ that is starch – it is also important to consider whether it is possible to increase the yield of storage organ per plant or per unit land area rather than the amount of starch per organ. In general terms, this has been the objective of crop breeders from the earliest times, but new insights into the mechanisms underlying whole-plant carbon allocation may permit novel approaches (Smith and Stitt, 2007).

Understanding the factors that determine the fate of assimilate within the plant remains an enormous challenge. There has been much speculation about whether the ‘sink strength’ of storage organs or the ‘source strength’ of leaves is the more important factor in determining how much assimilate is directed to harvested storage organs, but few attempts have been made to assess in quantitative terms the distribution of control of the flux of carbon from its assimilation in leaves to starch in storage organs. Perhaps the most advanced understanding in this area is for potato plants. Here, application of ‘top-down’ metabolic control analysis revealed that – under the conditions of the experiment – most of the control lay in the source (leaves) rather than in the sink (tubers; Sweetlove and Hill, 2000). The implication is that the best strategies for increasing tuber yield will involve manipulation of photosynthetic capacity rather than tuber metabolism. This general conclusion is supported by other experiments in which the sink strength (the capacity of the tuber to metabolize incoming assimilate) of potato plants was increased or decreased. In plants with increased sink strength (caused by expression of yeast invertase in the apoplast of tubers, increasing their capacity for sucrose hydrolysis), fewer, larger tubers were produced (Sonnewald et al., 1997). In plants with decreased sink strength (caused by reducing expression of AGPase and hence reducing the rate of starch synthesis), many more, very small tubers were produced (Müller-Röber et al., 1992). However, in neither case was there a large change in the total amount of assimilate allocated to tubers; although the tuber number and size were altered, the total mass of the tubers was relatively little changed. The simplest interpretation of these results is that alterations in the capacity of the tubers to metabolize sucrose affected competition for assimilate between tubers but had relatively little impact on the amount of assimilate supplied from the leaves. This is consistent with the idea that most of the control of flux from source to sink in potato plants lies in the source.

These conclusions from research on potato might seem to indicate that increased tuber yields can only be gained through increased rates of photosynthetic carbon assimilation. However, the results of a manipulation that increased the supply of ATP for ADPglucose synthesis in the tuber show that this is too simplistic a conclusion, and suggest an alternative way forward. Alterations in the activity of plastidial adenylate kinase not only increased the starch content of tubers (see above), but also increased tuber yield by 65–85% in field trials (Regierer et al., 2002). These data are consistent with measurements that indicate a concentration of ATP in potato tuber plastids that is similar to the KM of AGPase for this metabolite (Geigenberger, 2003). Thus, it seems likely that the adenylate energy charge in the tuber can exert significant control over the flux of carbon from leaves to tubers. Manipulations that increase ATP availability for ADPglucose synthesis in tuber plastids hold promise for increasing both starch yield on a tuber weight basis and tuber yield as a whole.

The control of the flux of carbon from assimilation in leaves to starch in storage organs of other crops has not been studied as systematically as in potatoes. A recent detailed review of the prospects for increasing the yield potential (seed mass per unit ground area) of grain crops – using information from free-air carbon dioxide enrichment (FACE) experiments – indicates that manipulations that increase leaf photosynthesis will increase yield (Long et al., 2006). In other words, as in potato, a significant part of the control of flux of assimilates into storage products in grain crops appears to lies with the source organs. This review provides interesting ideas on targets for manipulation of leaf photosynthesis. Contrasting conclusions come from experiments in which modified AGPases have been expressed in cereal endosperms. There are consistent reports that this manipulation of sink organs can bring about both higher yields and greater growth rates of vegetative parts of the plant under optimal environmental conditions (Meyer et al., 2007; Smidansky et al., 2003, 2007). Although the mechanism for this stimulation remains unclear, one possible interpretation is that increased sink strength at an early stage of seed development stimulates photosynthetic capacity in source leaves, increasing carbon assimilation and elevating the growth rate of the plant as a whole. However, reports that stimulation of carbon assimilation started prior to flowering (Smidansky et al., 2007) – at a point when the putatively endosperm-specific promoters used on the transgene would not be expected to be active – raise questions about the location and level of expression of the transgene. Further analysis of these plants is necessary to provide a full interpretation of stimulation of growth rate. In cassava, expression of a modified AGPase in roots also elevated whole-plant growth rates. Above-ground (leaf and stem) dry weight was 20–50% greater in three transgenic lines than in wild-type plants (Ihemere et al., 2006). Here too, additional experimental work is required before firm conclusions about the mechanism can be drawn.

Environmental conditions have a profound effect on the yield of starch in cereals and potatoes. Detailed dissection of the effects of drought and high temperature in maize has revealed that both cause failures at pre-fertilization and early developmental stages, and hence yield is limited by the number of seeds that actually set (e.g. McLaughlin and Boyer, 2004). In potato, waterlogging of the soil (hence reduced oxygen tension in the tuber) and elevated temperatures reduce the yield of starch, and variations in turgor pressure have major effects on the rate of starch synthesis in tuber discs (Oparka and Wright, 1988; Geigenberger, 2003). Discussion of these phenomena is outside the scope of this review, but clearly manipulations designed to increase the yield of starch in field-grown crops must take this important research into account.

More radical possibilities for increasing starch contents of crops

The rate of starch synthesis in the plastids of starch-storing organs of crops is already very high. This section considers two possibilities for increasing starch contents that do not involve partitioning more carbon into starch in storage organs: manipulation of starch turnover, and ectopic expression of enzymes of starch synthesis.

The leaves of many crop plants contain significant amounts of starch. In forage and silage crops, starch content is an important and positive component of the value of the crop as animal feed. Leaf starch tends to accumulate as a direct product of photosynthetic carbon assimilation during the day, and to decrease as it is converted to sugars at night, although the extent of turnover is highly dependent on species, leaf age and environmental conditions (Smith et al., 2005). It is unlikely that leaves could become a major source of starch for the existing fermentation industry. Extraction processes would have to be radically modified from those in place for storage organs, and achieving leaf starch contents as high as those of storage organs is practically impossible. However, in a biorefinery, vegetative biomass with a high starch content could be subjected to digestion and fermentation to produce alcohol before further processing for other biofuels and materials.

In Arabidopsis, manipulations of several enzymes of starch degradation reduce the rate of starch degradation at night and thus increase overall starch content in leaves (Smith et al., 2005; Zeeman et al., 2007). Starch contents up to five times the maximum for wild-type plants have been reported for mutants lacking glucan, water dikinase, an enzyme that controls starch granule degradation through phosphorylation of the granule surface (Ritte et al., 2002; Yu et al., 2001). Most of the mutations that reduce starch content also reduce the rate of plant growth, and this effect is much more pronounced under short than long days. While this effect may be the direct result of reduced carbon availability at night, there is a real possibility that it is part of a coordinated response of metabolism and growth to lower sugar levels (Smith and Stitt, 2007). If this is the case, it may well be possible to create plants in which leaf starch levels are elevated without a concomitant reduction in growth.

Blocking starch degradation may also be a viable means of increasing starch contents in some storage organs as well as in leaves. Although developing potato tubers expressing substantial activities of bacterial AGPase did not have increased starch contents (Sweetlove et al., 1996), metabolic labelling experiments suggested that they had elevated rates of starch turnover. Thus, blocking starch degradation in potato tubers, in combination with other manipulations to increase the rate of starch synthesis, may enhance the starch content. Cold storage of harvested potato tubers results in conversion of starch to sugars (cold-induced sweetening), reducing the processing quality of the tubers. Transgenic lines with reduced activity of glucan, water dikinase have reduced rates of cold-induced sweetening (Lorberth et al., 1998). Post-harvest loss of starch is a major problem in cassava roots: up to half of the starch can be lost in five days after harvest (Vlaar et al., 2001). Little is known about starch degradation in cassava, but it is reasonable to suppose that reduced levels of glucan, water dikinase might prevent much of this post-harvest loss.

A more radical possibility for increasing the starch content of crop plants would be to introduce starch-synthesizing enzymes into the cytosol, potentially driving glucan synthesis in this compartment as well as, or instead of, the plastid. This idea has some superficial attractions. Arguably, the total amount that could be produced might be less constrained by physical considerations in the cytosol than in plastids, and problems of transport of intermediates and ATP supply in non-photosynthetic tissues would be reduced. There is an important precedent for starch synthesis outside the plastid, in red algae. Some of these organisms make Floridean starch, consisting of polymers resembling amylopectin that form granules in the cytosol. The enzymes of polymer synthesis in these organisms have the same evolutionary origin as those of higher plants, but AGPase is absent and the glucan synthase uses UDPglucose (Patron and Keeling, 2005). In theory, provided that an adequate supply of ADPglucose or UDPglucose is available in the cytosol, introduction of a single glucan synthase and a starch-branching enzyme into this compartment in higher plants should result in the synthesis of starch-like glucans.

Unfortunately, there are many potential stumbling blocks in such an approach. It is doubtful whether there is an adequate supply of ADPglucose to support significant glucan synthesis in the cytosol of any higher-plant cells other than those of cereal endosperm. It has been proposed that the cytosolic enzyme sucrose synthase can produce ADPglucose for import into plastids for starch synthesis (Munoz et al., 2005); if this is the case, then this enzyme might supply substrate for an introduced cytosolic pathway. However, the importance of sucrose synthase in ADPglucose synthesis remains very contentious. Provision of an adequate supply of substrate for starch synthesis in the cytosol may well require introduction of an AGPase in addition to starch-synthesizing enzymes, or alternatively the introduction of a glucan synthase that can use UDPglucose. Whichever route is taken, it seems likely that introduction of a major new pathway utilizing cytosolic hexose phosphate will profoundly influence the normal primary metabolism of the cell.

Useful increases in starch content through expression of the starch pathway in the cytosol may require that the cytosol glucan crystallizes to form granules. Soluble glucans occupy much larger volumes than granules and have potentially serious impacts on the osmotic and volume regulation of cells (Raven, 2005). Extraction and processing are potentially more difficult for a soluble material than for dense, crystalline granules. Unfortunately, there is no guarantee that the products of cytosolic starch synthases and branching enzymes would form granules. Although starch polymers almost invariably associate to form semi-crystalline granules inside plastids in wild-type plants, almost nothing is known about the biochemical and physical processes underlying granule formation. Polymers synthesized by higher-plant starch synthases in vitro or when introduced into lower organisms do not associate to form granules. A crucial factor in granule formation appears to be the presence of a starch-debranching enzyme, isoamylase. In mutants lacking this enzyme, much of the starch is replaced with a soluble glucan known as phytoglycogen. It is not known whether isoamylase participates directly in starch granule synthesis, or promotes granule formation by indirect means (Myers et al., 2000). It seems likely that other factors in addition to isoamylase are required for normal granule initiation and growth in the plastid. As discussed above, the process appears to require the presence of multiple isoforms of starch synthase and starch-branching enzyme (e.g. Roldán et al., 2007). Creation of granules, rather than soluble glucans, in the cytosol might thus require considerable further research.

There are several examples of the creation of transgenic plants in which the plastids of starch-storing organs contain heterologous enzymes that convert sucrose or starch into non-plant glucans or into fructans. These experiments are aimed at the creation of starch-based novel carbohydrates (Kok-Jacon et al., 2003), but they also have the potential to generate plants with elevated levels of readily digestible and fermentable glucans in their harvested organs. Where sucrose is the substrate, these approaches rely for their success on the relatively recent discovery that most plastids contain sucrose in vivo (e.g. Farréet al., 2001; Gerrits et al., 2001); traditionally, this metabolite was thought to be confined to the cytosol. Reports of successful generation of novel glucans include potato plants that synthesize starch–fructan hybrid polymers (Gerrits et al., 2001), small amounts of cyclodextrins (Oakes et al., 1991) and starch granules apparently coated with α-1,3, α-1,6 glucan (Nazarian Firouzabadi et al., 2007), and maize plants in which up to 8% of the kernel dry mass is fructan (Caimi et al., 1996) or up to 14% of the kernel dry mass is an α-1,6 glucan (Zhang et al., 2007). However, I am not aware of any evidence so far showing that total polymeric non-structural carbohydrate contents (starch plus novel glucan or fructan) are significantly increased in any of these plants.

Modification of starch structure to enhance processibility

The dense, semi-crystalline nature of starch granules greatly facilitates their extraction from the plant, but also demands considerable energy input to disrupt crystallinity in order that enzymic digestion and fermentation can be carried out. For most storage starches, gelatinization (loss of crystallinity) requires heating in water to at least 60°C. Reductions in energy input at this stage of bioethanol manufacture would reduce the economic and carbon costs of bioethanol production. This could potentially be achieved through use of starches requiring lower energy inputs for gelatinization. There are several reports of mutant and transgenic plants that produce starches with such properties. For example, the starch of transgenic potatoes with reduced activities of starch synthases II and III gelatinizes at temperatures 20°C lower than for normal starch (Edwards et al., 1999). Plants lacking the debranching enzyme isoamylase make a soluble glucan, phytoglycogen, in place of some of their starch (Myers et al., 2000); the solubility of this polymer means that digestion to glucose can occur without any gelatinization step. Purification of phytoglycogen is problematic, but it might present advantages over starch in applications in which digestion and fermentation are carried out on plant homogenates or flours rather than on purified starch itself. However, efforts to produce readily digestible glucans may be rendered redundant by recent developments in starch degradation technology that allow conversion of starch granules to glucose without the need for gelatinization (http://www.genencor.com/cms/resources/file/ebf95c076d3afc7/STARGEN%20Backgrounder.pdf).

Changes in granule number may also affect processibility. It is expected that smaller granules will digest more readily than larger ones because of the larger surface area per weight of starch. A number of mutations and manipulations that decrease average granule size without major impacts on yield (see below) have been reported. For potatoes, these include reductions in isoamylase activity (Bustos et al., 2004) and introduction of an engineered starch-binding protein derived from an enzyme of starch degradation (Ji et al., 2004).

Increasing sucrose content

Relatively little is known at the biochemical or genetic levels about the factors that control the extent of sucrose storage in nodes of sugar cane and roots of sugar beet. Most of the available information is correlative – derived from the relationship between sucrose content and selected parameters such as levels of RNA or activity of sucrose-metabolizing enzymes (e.g. Fieuw and Willenbrink, 1987; Grof et al., 2007; Hesse and Willmitzer, 1996; Hesse et al., 1995; Rose and Botha, 2000) – and it is generally not clear whether these correlations are causal. This state of affairs reflects the relatively small numbers of researchers investigating sugar crops, and the intractability of both the crops (from a genetic perspective) and the developmental and biochemical systems. In addition to whole-plant assimilate partitioning (for which the issues are similar to those discussed above for starch), central elements in the control of sucrose storage are likely to be the control of phloem unloading (e.g. Godt and Roistch, 2006), the nature and kinetics of sucrose and hexose transporters on the plasmalemma and tonoplast of storage cells (e.g. Casu et al., 2003; Rae et al., 2005), and the control of futile cycling of sucrose in the cytosol of storage cells (e.g. Whittaker and Botha, 1997). These three areas are among the least understood parts of primary metabolism. The challenges of understanding sucrose accumulation in nodes of sugar cane stems are discussed in detail by Rae et al. (2005) and Uys et al. (2007).

The increasing availability of genomic tools and transformation techniques for sugar crops should allow some of the hypotheses about control of sucrose accumulation to be systematically tested. For example, increases in sucrose contents were recently reported in transgenic sugar cane suspension cultures with reduced neutral invertase activity (Rossouw et al., 2007). Equally, these techniques allow the testing of radical approaches to increasing sugar storage. Researchers in Australia have expressed a gene encoding a bacterial enzyme, sucrose isomerase, in sugar cane. Sucrose isomerase converts sucrose to its isomer isomaltulose, which is valuable in the food industry as a sweetener that is not a substrate for plaque-forming oral bacteria. Stem nodes of transgenic plants accumulate very high levels of isomaltulose, and the total sugar levels (isomaltulose plus sucrose) in harvested juice are up to twice those of control plants (Wu and Birch, 2007). Photosynthesis and sugar transport are also stimulated in the transgenic lines. One interpretation of this result is that a signalling process that involves sucrose itself acts to restrict the extent to which sucrose can accumulate. Isomaltulose, as a non-plant metabolite, does not trigger this signalling. Conversion of sucrose to isomaltulose may therefore allow higher levels of sugar accumulation than when sucrose accumulates alone. Discovery of this putative signalling system may allow other approaches to manipulation of sugar accumulation in sugar crops, including sugar beet.

In sugar beet, introduction of an enzyme of fructan synthesis from onion resulted in substantial conversion of the sucrose content of the storage root to fructans of low molecular mass (Weyens et al., 2004). However, this apparently did not result in greater import of sucrose into the roots: total carbohydrate content was not affected.

Priorities for plant science research

Many of the possible ways of increasing the starch and sugar contents of crops discussed above require genetic manipulation. Although this is technically increasingly easy, it will also be extremely expensive and labour-intensive if the products are to be commercialised in the next 10–15 years. In light of the massive funding of research into new, dedicated biofuel crops, transgenic manipulation of existing food crops to improve biofuel production may well not be economically sensible unless the work is already well advanced. Pragmatically, I suggest that the focus of plant biotechnological research on food crops as sources of biofuel should be in areas that do not require a transgenic approach. Fortunately these include some of the more exciting prospects for increasing plant starch contents and starch processibility: for example, downregulation of plastidial adenylate kinase, downregulation of enzymes of starch degradation, and downregulation of specific enzymes of starch polymer synthesis. Reductions in the levels of expression of endogenous genes can be brought about via forward genetic screens and targeted mutant discovery (e.g. TILLING; Perry et al., 2003). The last few years have seen the application of these techniques to a very wide range of crops.

Although dedicated biofuel crops and cell-wall fermentation technology are likely to provide a more sustainable and larger-volume solution to global energy problems in the medium term, the need for improved understanding of starch and sugar metabolism will increase rather than diminish. At a fundamental level, we know very little about the relationships between carbon assimilation, carbon storage and growth. Research on Arabidopsis has revealed the existence of sophisticated sensing and signalling systems that adjust the growth rate and carbon storage in response to alterations in carbon availability within the plant (Smith and Stitt, 2007). Understanding how these processes are integrated across seasons in perennial biomass crops represents a major and important new challenge that must be met if biomass gain in these crops is to be manipulated in a rational manner. At a more applied level, the presence of easily fermentable storage products in biomass crops could strongly influence the viability of downstream processing. Starch and sugars could provide both initial substrates for micro-organisms in a fermenter, and substrate for an initial alcohol production step in a biorefinery.

Acknowledgements

I am very grateful to Kay Denyer (John Innes Centre) for critical reading and valuable comments on this article. I also thank Jonathan Jones (Sainsbury Laboratory, Norwich, UK) for providing much helpful information on biofuel crops.

Ancillary