Arabidopsis and primary photosynthetic metabolism – more than the icing on the cake


For correspondence (fax +49 331 567 8101; e-mail


Historically speaking, Arabidopsis was not the plant of choice for investigating photosynthesis, with physiologists and biochemists favouring other species such as Chlorella, spinach and pea. However, its inherent advantages for forward genetics rapidly led to its adoption for photosynthesis research. In the last ten years, the availability of the Arabidopsis genome sequence – still the gold-standard for plant genomes – and the rapid expansion of genetic and genomic resources have further increased its importance. Research in Arabidopsis has not only provided comprehensive information about the enzymes and other proteins involved in photosynthesis, but has also allowed transcriptional responses, protein levels and compartmentation to be analysed at a global level for the first time. Emerging technical and theoretical advances offer another leap forward in our understanding of post-translational regulation and the control of metabolism. To illustrate the impact of Arabidopsis, we provide a historical review of research in primary photosynthetic metabolism, highlighting the role of Arabidopsis in elucidation of the pathway of photorespiration and the regulation of RubisCO, as well as elucidation of the pathways of starch turnover and studies of the significance of starch for plant growth.


Photosynthetic primary metabolism might be regarded as a mature field. It is covered in textbooks at length, and many readers probably dismiss it as ‘boring’ biochemistry, where the important questions were asked and answered long ago. In fact, photosynthesis research is undergoing a renaissance, driven by a renewed appreciation of the basic importance of crop yield. Arabidopsis has played a key role in elucidating the pathways and regulation of photosynthesis. As the knowledge base and experimental toolkit in Arabidopsis expanded, important problems that were previously intractable have become accessible. Many central questions could only be tackled in the last decade, following sequencing of the Arabidopsis genome and the advent of functional genomics. These advances provide an excellent starting point for systems analysis.

Arabidopsis was not initially the plant of choice for investigating photosynthesis, at least from the point of view of physiologists or biochemists. Important experimental systems for photosynthesis research in the past included the green alga Chlorella for elucidation of the C3 pathway (Calvin–Benson cycle) of CO2 fixation, photosynthetic bacteria for structural studies, and spinach and pea, due to the ease with which intact functional thylakoids and chloroplasts could be prepared from their leaves, and the ready availability of large amounts of leaf material for biochemical work. Spinach, sunflower and other species with large flat leaves were well suited for gas exchange measurements. Tobacco later rose to prominence due to its ease of nuclear and plastid transformation. Discovery and elucidation of the C4 and crassulacean acid metabolism (CAM) pathways of photosynthesis were largely performed in C4 grasses (e.g. sugarcane, maize and Panicum spp.) and members of the Crassulaceae (e.g. Kalanchöe spp.). Neverthless, the power of Arabidopsis forward genetics quickly led to it becoming an important species in photosynthesis research. Its use was supported by the development of dedicated techniques for chloroplast isolation (Somerville et al., 1981; Kubis et al., 2008) and gas exchange analysis (Tocquin and Périlleux, 2004). The importance of Arabidopsis has expanded even further in the last ten years, due to the growing genetic and genomic resources that are available for this reference species.

In this review, we first survey research on primary photosynthetic metabolism during the last half century, to put the contribution of Arabidopsis into a historical perspective. We then discuss contributions that have been driven by Arabidopsis research in the last decade. Rather than being comprehensive, we focus on a small number of pathways to illustrate how the knowledge and resources provided by Arabidopsis have allowed increasingly difficult questions to be dissected and addressed. We deal with elucidation of the metabolic pathways of photosynthesis, subcellular compartmentation and regulation of the pathways, including in vitro enzymology and protein biochemistry and in vivo analyses of metabolite levels and fluxes. The examples are mainly drawn from the Calvin–Benson cycle, photorespiration and synthesis of the two main carbohydrate products, sucrose and starch. The review ends with a discussion of some questions that need to be addressed in the future.

Overview of the path of carbon in central photosynthetic metabolism

Photosynthesis requires joint operation of the Calvin–Benson cycle and end-product synthesis (Figure 1a). Ribulose-1,5-bisphosphate carboxylase/oxygenase (RubisCO) adds CO2 to ribulose-1,5-bisphosphate (RuBP) to form glycerate-3-phosphate (3PGA), in the first step of carbon fixation. The vast majority of the NADPH and ATP produced in the light reactions are used to reduce 3PGA to triose phosphates. Triose phosphates are used to regenerate RuBP in a complex reaction sequence, in which the carbon skeletons are rearranged using enzymes shared with glycolysis (aldolase), gluconeogenesis (fructose-1,6-bisphosphatase, FBPase), the oxidative pentose phosphate pathway (transketolase, phosphoribose isomerase and phosphoribulose epimerase) and the two unique enzymes sedoheptulose-1,7-bisphosphatase (SBPase) and phosphoribulokinase (PRK) (Farquhar et al., 2001; Heldt, 2005).

Figure 1.

 Photosynthetic carbon metabolism in Arabidopsis and other C3 plants.
CO2 fixation via the Calvin–Benson cycle in the chloroplasts is tightly integrated with the photorespiratory pathway, and with synthesis of the main end products of photosynthesis – starch and sucrose. The reactions are catalysed by the following enzymes: (1) RubiscCO; (2) phosphoglycerate kinase; (3) NADP-glyceraldehyde phosphate dehydrogenase; (4) plastidic triose phosphate isomerase; (5) plastidic aldolase; (6) plastidic fructose-1,6-bisphosphatase; (7) transketolase; (8) sedoheptulose-1,7-bisphosphatase; (9) ribulose phosphate epimerase; (10) phosphoriboisomerase; (11) phosphoribulokinase; (12) phosphoglycolate phosphatase; (13) glycolate oxidase; (14) glutamate:glyoxalate aminotransferase; (15) glycine decarboxylase; (16) serine hydroxymethyl transferase, (17) serine:glyoxylate aminotransferase; (18) hydroxypyruvate reductase; (19) glycerate kinase; (20) plastidic phosphoglucoisomerase; (21) plastidic phosphoglucomutase; (22) ADP-glucose pyrophosphorylase; (23) starch synthase; (24) starch branching enzyme; (25) triose phosphate translocator; (26) cytosolic triose phosphate isomerise; (27) cytosolic aldolase; (28) cytosolic fructose-1,6-bisphosphatase; (29) cytosolic phosphoglucoisomerase; (30) cytosolic phosphoglucomutase; (31) UDP-glucose pyrophosphorylase; (32) sucrose phosphate synthase; (33) sucrose-6′-phosphate phosphatase; (34) vacuolar invertase. For simplicity, only the main pathway of carbon is shown, and the use of dihydroxyacetone (DHAP) as a substrate in the two transketolase reactions, and of glyceraldehyde-3-phosphate (GAP) as a substrate in the second aldolase reaction leading to the formation of sedoheptulose-1,7-bisphosphate (Sed1,7BP) has been omitted.

The Calvin–Benson cycle operates in parallel with the pathway of photorespiration, which scavenges the 2-phosphoglycolate generated in a side reaction of RubisCO with O2, and converts it back to 3PGA (see below). Photorespiration is a complex pathway involving enzymes in the chloroplast, cytosol peroxisomes and mitochondria. It leads to loss of a quarter of the carbon in 2-phosphoglycolate as CO2, and to a wasteful cycle of ammonium release and re-assimilation.

Triose phosphates are used to synthesize end-products such as sucrose, starch and amino acids. Starch is synthesized in the chloroplast, starting from fructose-6-phosphate, whereas sucrose is synthesized in the cytosol. The pathway of sucrose synthesis involves export of triose phosphates in exchange for inorganic orthophosphate (Pi) via the triose phosphate:phosphate translocator (TPT), followed by their conversion to hexose phosphates via a classical gluconeogenesis pathway involving a cytosolic FBPase. The final and committed steps of sucrose synthesis are catalysed by sucrose phosphate synthase (SPS) and sucrose-6-phosphatase (SPP).

Characterization of the pathways of photosynthetic carbon metabolism

The Calvin–Benson cycle was discovered in the 1950s, in classical biochemical studies with Chlorella (Bassham and Calvin, 1957; see Benson, 2002; Bassham, 2003 for recent historical reviews). The pathway was defined by pulsing algal suspensions with 14CO2 and sampling them sequentially at short intervals by injection into boiling alcohol, followed by analysis using two-dimensional paper chromatography and autoradiography. The detailed time courses and comprehensive coverage of metabolites provided by this approach allowed prediction of a pathway. This was verified by showing that all of the necessary enzymes activities were present in Chlorella, including several previously unknown enzyme activities that were predicted from hypothetical reactions, such as RubisCO and PRK, which uses ATP to convert ribulose 5-phosphate (Ru5P) to the CO2 acceptor RuBP.

The pathway (and biological function) of photorespiration was a subject of dispute in the 1970s. Gas exchange experiments had revealed that there is rapid O2 uptake and CO2 release in the light during photosynthesis in ambient CO2 in higher seed plants. This process was termed ‘photorespiration’. Crucially, the uptake of O2 is suppressed by high CO2 (Forrester et al., 1966). This led to the discovery that RubisCO also accepts O2 as a substrate (Ogren and Bowes, 1971; Lorimer and Andrews, 1973). The product of the oxygenase reaction is 2-phosphoglycolate. It was already known that glycolate and a set of other metabolites such as glycine and serine are rapidly labelled by 14CO2 during photosynthesis at ambient CO2 concentrations, and it was suspected that glycolate and glycine were the source of the CO2 that was released during photorespiration (Tolbert, 1971). Various sources of the glycolate had been proposed, including side reactions of transketolase. There was much debate about whether all the glycolate for photorespiration actually derives from 2-phosphoglycolate, with the shocking implication that photorespiration simply removes a waste product from a (useless) side reaction of RubisCO.

This debate was resolved by incisive use of forward genetics in Arabidopsis in the late 1970s and early 1980s (see Somerville, 2001; for a historical review). Briefly, Somerville and colleagues deduced that, if photorespiration operated only to remove 2-phosphoglycolate, then mutants in the pathway should be viable in high CO2 (1-2%) when the oxygenase reaction of RubisCO is suppressed, but unable to survive in air. Using a screen for plants that grew normally in 2% CO2 but developed chlorosis and necrosis after transfer to air, a set of mutants was identified that were deficient for several enzymes of the photorespiratory pathways and the related cycle of ammonium re-assimilation. Arguably the most single decisive finding was the identification of a mutant that accumulated 2-phosphoglycolate and was deficient in 2-phosphoglycolate phosphatase activity. This mutant effectively ended the debate about whether the oxygenase reaction of RubisCO was responsible for generating the 2-phosphglycolate and hence the glycolate that is metabolized in photorespiration (see Somerville and Ogren, 1979). This was probably the first foray of this previously little-known weed into photosynthetic research.

Within a couple of years, and building on enzymological studies of RubisCO, Farquhar, von Caemmerer and Berry developed a mathematical model that linked whole-plant gas exchange to the kinetic properties of RubisCO (von Caemmerer and Farquhar, 1981; Farquhar et al., 1982). This model has formed the cornerstone for most subsequent models of photosynthesis and for global-scale models of interaction between the photosynthetic activity of plants and atmospheric composition (Intergovernmental Panel on Climate Change, 2001). Thus, within a few years, a combination of forward genetics, metabolic analyses, enzymology, gas exchange and modelling laid the foundations for one of the major contributions of plant science to understanding global events.

The realisation that photorespiration is a salvage pathway that rescues plants from the consequences of a major side reaction of RubisCO triggered a massive and sustained attempt to improve photosynthesis and increase crop yield, by suppressing photorespiration. The oxygenase reaction of RubisCO and photorespiration (often termed the C2 oxidative photosynthetic pathway) jointly represent a waste of energy, carbon, water and nitrogen. Initial strategies centred on studying RubisCO, to understand the catalytic mechanism and structure of the enzyme, in order to engineer it to alter the relative specificities for CO2 and O2. This strategy, despite immense effort, has been largely unsuccessful to date. Strategies that attempt to suppress photorespiration by introducing C4-like features into C3 plants have also made little progress. However, in the last decade, encouraging steps have been made towards engineering an alternative, and more efficient, pathway than photorespiration to scavenge glycolate (see below).

The initial steps in the pathway of starch synthesis were clarified by analogy to glycogen synthesis in microbes and humans, supported by biochemical studies in spinach and other organisms. The first dedicated step is catalysed by ADP-glucose pyrophosphorylase (AGPase) in the plastid, using glucose-1-phosphate that is synthesized from Calvin–Benson cycle intermediates via plastidic phosphoglucose isomerase and phosphoglucomutase. Important confirmation came from forward mutagenesis in Arabidopsis, again by Somerville and colleagues. They used a simple screen in which leaves were stained for starch using iodine, in order to isolate a set of starchless mutants, including mutants in the plastidic phosphoglucomutase (Caspar et al., 1985) and in the APS and APL subunits of the heterotetrameric enzyme ADP-glucose pyrophosphorylase (Wang et al., 1998). Later studies identified mutants in the plastidic phosphoglucoisomerase (Neuhaus et al., 1989; Yu et al., 2000). Analysis of starchless mutants resulted in the important finding that starch is essential for growth under a light/dark cycle once the night exceeds about 12 hours in duration. They are still being used to understand why this is the case (see below). Starch differs from glycogen in being a crystalline granule composed of a mixture of amylose and amylopectin. The conversion of ADP-glucose to this complex structure involves the cooperation of a series of starch synthases, branching enzymes and glucan trimming enzymes, whose roles have been elucidated in the last ten years (see below).

Starch accumulates in leaves during photosynthesis and is remobilized at night to support leaf respiration and sucrose export. It was initially assumed that the pathway of starch breakdown would resemble the canonical pathway in cereal seeds, where an initial attack by the endohydrolytic enzyme α-amylase releases glucans that are digested to maltose by exolytic β-amylases. Early experiments in the 1970s and 1980s raised questions about this pathway, because almost all of the enzyme activities appeared to be localized outside the chloroplasts in leaves (Stitt et al., 1978). However, elucidation of the pathway only occurred in the last decade, and required the tools of Arabidopsis genetics and ‘omics’ (see below).

The pathway of sucrose synthesis was elucidated by 14CO2 labelling studies. Its final clarification also required a full understanding of subcellular compartmentation (Lunn, 2007a), including identification of the triose phosphate translocator (TPT). The existence of the latter was inferred from experiments with isolated chloroplasts. While inclusion of Pi in the medium was essential for CO2 fixation by isolate chloroplasts, photosynthesis was progressively inhibited as the Pi concentration increased. This inhibition was prevented when metabolites such as triose phosphates or 3PGA were included in the medium (Heldt, 2005). The TPT was identified using methods for the direct measurement of metabolite transport, transferred from mammalian research, in a series of classical studies by Heldt, Flügge and co-workers (Heldt, 2005; Weber et al., 2005. This research was mainly performed in species such as spinach and pea, from which it was possible to isolate intact and functionally operating chloroplasts.

Regulation of photosynthetic carbon metabolism

After establishing a pathway, the next step is to understand its regulation. The Calvin–Benson cycle is a rather atypical metabolic pathway. RubisCO has a remarkably low kcat (3 sec−1), for reasons that are interwoven with the reaction mechanism of this unique enzyme (Tcherkez et al., 2006). This slow rate of catalysis is compensated for by having an enormously large amount of the protein. RubisCO represents up to 40% of the protein in a leaf, and is present at 10–20 times higher concentrations than the other enzymes in the pathway (Piques et al., 2009), with an active site concentration of about 4 mM. This is of the same order as the concentration of its substrate, RuBP, and much higher than the KmRuBP (< 20 μM). It can be argued that a critical feature of the cycle is to maintain the RubisCO binding site saturated with substrate, in order to maximize the number of sites at which CO2 can react. It may also be important to minimize the number of free binding sites, because they could bind and sequester other metabolites such as FBP and NADP+, which would interfere with operation of the Calvin–Benson cycle.

The key to understanding the regulation of RubisCO emerged from the screen for photorespiratory mutants in Arabidopsis. One mutant that was unable to grow in ambient CO2 turned out to be defective in a protein, RubisCO activase, that facilitates the release of tightly binding inhibitors from the active site of RubisCO that are formed at a low but significant rate during the catalytic cycle (Salvucci et al., 1985). When the enzyme is assayed in vitro, accumulation of these side products leads to inactivation of RubisCO within a few minutes. The primary role of RubisCO activase is to overcome this defect, and prevent RubisCO from poisoning itself (Portis et al., 2008). By regulating the rate at which RubisCO reactivates, it may also provide a way to adjust the number of active binding sites to the supply of RuBP and the rate of CO2 fixation. Subsequent work has shown that RubisCO activase is subject to post-translational redox regulation by thioredoxin, in an ADP- and light-dependent manner, and has identified regions that are responsible for the interaction between RubisCO activase and RubisCO (Portis et al., 2008).

The regeneration of RuBP is regulated by three enzymes that catalyse strongly irreversible reactions (Bassham and Krause, 1969): FBPase, SBPase and PRK. Biochemical work with isolated spinach chloroplasts and enzymology established that these reactions are subject to multi-layered regulation, including light-dependent post-translational redox modification by thioredoxin (Buchanan and Balmer, 2005), light-dependent changes in pH and Mg2+, feedback regulation by their products, and allosteric regulation by other metabolites (Heldt, 2005; Stitt, 1996). The four layers interact: for example, activation by thioredoxin modifies the Km values for substrates, while increasing levels of substrate may modify the mid-redox potential of the cysteine thiol groups that are targeted by thioredoxin, and thence increase the proportion of the enzyme that is activated at a given redox potential. This allows all three reactions (FBPase, SBPase and PRK) to be inhibited by 100–1000-fold in the dark, to prevent them from interfering with dark respiration. Redox regulation is also thought to allow their fluxes to be tightly and rapidly coordinated to avoid accumulation of metabolites in one sector of the Calvin–Benson cycle (Scheibe, 1991; Stitt, 1996). This is essential as the turnover times of these intermediates are very short, of the order of 0.1-1 sec (Arrivault et al., 2009). Small differences in flux at various points in the cycle would rapidly lead to sequestration of metabolites and disrupt operation of the cycle.

The stoichiometry of the Calvin–Benson cycle requires that 5/6ths of the triose phosphates are used to regenerate RuBP (actually more, if photorespiration is occurring). The remainder can be used to synthesize sucrose, starch or other end products (Heldt, 2005). It is essential to balance these competing fluxes. Excessive synthesis of end products will inhibit photosynthesis because too little RuBP is regenerated, while low rates of end product synthesis will lead to accumulation of phosphorylated intermediates, depletion of free Pi, inhibition of ATP synthesis, 3PGA reduction and CO2 fixation (Stitt et al., 1987; Heldt, 2005; MacRae and Lunn, 2006).

Initial insights into the regulation of end product synthesis came from biochemical studies with isolated chloroplasts and intact leaves, and enzymology, mainly in spinach. AGPase is allosterically activated by 3PGA and inhibited by Pi, allowing the rate of starch synthesis to be increased when phosphorylated intermediates accumulate and Pi decreases in the stroma (Ballicora et al., 2004). An analogous but more sophisticated mechanism gears the rate of sucrose synthesis to the availability of triose phosphates (Stitt, 1990; MacRae and Lunn, 2006). Briefly, the cytosolic fructose-1,6-bisphosphatase (cFBPase) is inhibited by the signal metabolite fructose-2,6-bisphosphate (Fru2,6BP). Increasing 3PGA and decreasing Pi inhibit the synthesis and stimulate the degradation of Fru2,6BP, leading to stimulation of cFBPase activity. Fru2,6BP acts by shifting the sigmoidal substrate saturation curve of cFBPase. The level of the substrate, Fru1,6BP, is itself also very sensitive to changes in the level of metabolites exported from the Calvin–Benson cycle, because it is formed via a second-order reaction from two triose phosphate molecules. Thus, three factors interact to strongly activate cFBPase when 3PGA and triose phosphates increase and Pi decreases: a decrease in the inhibitor Fru2,6BP, the cooperative kinetics of cFBPase and the amplified increase of Fru1,6BP. This means that cFBPase is effectively inactive until a threshold concentration of triose phosphates is exceeded, above which cFBPase activity rises very sharply. The second important regulated enzyme in the pathway of sucrose synthesis, SPS, is allosterically activated by Glc6P and inhibited by Pi, and is subject to multi-site post-translational regulation by protein phosphorylation (Huber and Huber, 1996). However, as we will see, the availability of the Arabidopsis genome sequence raised important new questions relating to the presence of small gene families for key enzymes in the pathways of end product synthesis.

Figure 2 gives a schematic overview of the interactions between the Calvin–Benson cycle and end product synthesis. The plastidic FBPase competes for substrates with the cytosolic FBPase. While both have sigmoidal substrate saturation kinetics, the strong threshold response of the cytosolic FBPase presumably pays a major role in ensuring that the plastidic FBPase competes successfully when the triose phosphate levels are low, while at the same time allowing sucrose synthesis to be strongly activated when this threshold is exceeded. SBPase competes for substrates with AGPase. It has a hyperbolic substrate saturation curve, but may be able to compete with AGPase provided the latter is not allosterically activated by a high 3PGA/Pi ratio.

Figure 2.

 Schematic overview of interactions between the Calvin–Benson cycle and end product synthesis.
The icons show the qualitative relationship between substrate concentration and flux for seven enzymes. The large icon shows the relationship between triose phosphate concentration and cytosolic FBPase activity. This relationship was empirically modelled by measuring the level of triose phosphates and F2,6BP in spinach leaf discs under different light and CO2 levels, estimating the corresponding in vivo concentrations of F1,6BP (assuming aldolase is at equilibrium) and F2,6BP, and combining these concentrations for each simulated condition in an in vitro enzyme assay (Herzog et al., 1984). The smaller icons with dotted lines depict predicted substrate–activity relationships basen on in vitro enzyme assays, which were performed in the presence of typical concentrations of effector metabolites in isolated chloroplasts under saturating light and CO2 (plastid FBPase, Gardemann et al., 1986; SBPase, Schimkat et al., 1990; PRK, Gardemann et al., 1983; RubisCO, Heldt, 2005, AGPase, Ballicora et al., 2004; SPS, Stitt et al., 198X; Huber and Huber, 1996). The precise Ka values for plastidic FBPase and SBPase depend on effectors including pH, Mg2+, and, in the case of plastidic FBPase, Fru6P, that for AGPase depends on the allosteric effectors 3PGA and Pi, and that for SPS depends on the allosteric effector Glc6P.

Photosynthetic carbon metabolism is subject to transcriptional regulation during leaf development, and is under the influence of light. Transcriptional regulation is also strongly implicated in feedback regulation of photosynthetic carbon metabolism by sugars. The pioneering studies that gave rise to this conclusion were performed in maize cell cultures (Sheen, 1990) and in transgenic tobacco lines that were impaired in sucrose export due to over-expression of invertase in the apoplast (von Schaewen et al., 1990). The idea that sugars exert feedback regulation on photosynthesis has attracted considerable interest in the context of plant responses to elevated ambient CO2 concentration (Stitt, 1990; Stitt and Krapp, 1999). While elevation of ambient CO2 concentration leads to a higher rate of photosynthesis, the resulting increase in growth and crop yield is typically smaller than expected (Long et al., 2006). One reason for this may be that transcriptional feedback regulation leads to a decrease in the rate of photosynthesis. Arabidopsis rapidly became the experimental system of choice to dissect the signalling pathways (Rolland et al., 2006). As early as the 1990s, screens were set up to identify mutants that were hypersensitive to high sugars (Zhou et al., 1998; Gibson, 2005) or showed altered expression of sugar response genes (Rook et al., 1998). This research area profited strongly from the availability of Arabidopsis ‘omics’ approaches. This research has uncovered interesting links between sugar sensing and ethylene and ABA. More specific approches may be needed to understand the mechanisms whereby sugars regulate the expression of genes involved in the Calvin–Benson cycle and in starch and sucrose synthesis, and to elucidate how the rate of growth is coordinated with the availability of photosynthate (see below).

Flux control analysis

The starting points of the studies discussed so far were the regulatory features of individual enzymes. However, metabolism involves interactions between enzymes. These interactions can lead to responses that are not easily predictable from the features of the individual components. Pathway responses depend on network topology, including the sequence of reaction steps and all the links created by regulatory interactions. An important and complementary approach is to use top-down methods to characterize system properties and responses. This work started in the early 1990s, and is continuing. It relies increasingly on the genetic resources available for Arabidopsis.

One of the first theoretical frameworks for top-down analysis was flux control analysis (FCA) (Kacser and Burns, 1973; Fell and Thomas, 1995). The most common application of FCA is to determine flux control coefficients. A flux control coefficient is a system property that, put simply, describes the fractional change in flux that results when a single enzyme activity is changed by a small amount, leaving all other enzyme activities unaltered. Flux control coefficients have been determined for the majority of the enzymes in the Calvin–Benson cycle as well as starch and sucrose synthesis. Some of these studies used heterozygotes, or (in the cases where there was gene duplication) gene dosage series created from classical mutants. As the technology became available, reverse genetics was used to change the amounts of individual enzymes. Many of these studies have been performed in Arabidopsis (see Table 1). It should be noted that experimental determination of flux control coefficients requires a pragmatic simplification. Strictly speaking, FCA deals with infinitesimally small changes of enzyme activities and fluxes. In practice, larger changes must be analysed due to limitations imposed by the precision and sensitivity of the analytical techniques and inherent biological variation.

Table 1.   Experimental estimates of flux control coefficients for enzymes in the Calvin–Benson cycle and the pathways of sucrose and starch synthesis
  1. Enzymes that catalyse irreversible reactions in vivo are shown in bold, and those that catalyse reversible reactions are shown in italic (Bassham and Krause, 1969; Gerhardt et al., 1987). The determinations were made by quantifying maximum enzyme activity (and in some cases enzyme protein) and the rates of photosynthesis, sucrose synthesis and starch synthesis in mutants or transgenic lines. Flux control coefficients (FCC) are estimated as the fractional change in flux divided by the fractional change in enzyme activity (Kacser and Burns, 1973). Most are over-estimates because experimental determinations require a substantial decrease in target enzyme activity. FCCs are estimated for the pathway in which the enzyme is involved. A small decrease in the rate of sucrose synthesis is compensated for by an increased rate of starch synthesis such that photosynthesis is unaffected, except under saturating light and high CO2. For enzymes where the FCC was zero, the column labelled ‘excess’ indicates how much of the activity can be removed before there is a marked decrease in flux.

Calvin–Benson cycle, FCC estimated for photosynthesis under ambient growth conditions
RubisCO0.35 TobaccoStitt and Schulze (1994); Mate et al. (1993)This is the FCC under ambient conditions for plants grown under moderate light. The FCC rises to 1 if plants are suddenly transferred to saturating light, and falls to 0 if they are transferred to low light.
GAPDH3TobaccoPrice et al. (1995) 
TPI TobaccoUnpublished 
Aldolase0.2 PotatoHaake et al. (1998, 1999)Ambient photosynthesis is inhibited under a wide range of growth irradiances
pFBPase0.1–0.2 TobaccoKossmann et al. (1994)Rates under saturating CO2 and high or low light. FCCs are over-estimated because they are based on a transgenic line with a > 60% decrease in activity.
TK< 2TobaccoHenkes et al. (2001)FCC = 0.8 under saturating light and ambient CO2, and 1 under saturating light and saturating CO2. A small (30%) decrease in TK activity led to a decrease in aromatic amino acids and phenylpropanoids.
SBPase 0.2–0.35 TobaccoHarrison et al. (1998, 2001); Olcer et al. (2001)FCC approaches 1 under saturating light and CO2. Over-expression increases the rate of photosynthesis in tobacco (Lefebvre et al., 2005)
PRK> 10TobaccoPaul et al. (1995)FCC approaches 0.8 when plants that were grown under low light are transferred to high light (Paul et al., 2000)
Starch synthesis pathway, FCC estimated for starch synthesis
pPGI0.22Clarkia xantianaNeuhaus et al. (1989)Near-total inhibition of starch synthesis in an Arabidopsis mutant where pPGI protein level was below the detection limit (Yu et al., 2005)
pPGM2ArabidopsisNeuhaus and Stitt, 1990The null pgm mutant has negligible starch (Streb et al., 2009)
AGPase0.46 ArabidopsisNeuhaus and Stitt, 1990Arabidopsis null mutants in the genes for the APL1 and APS1 subunits have effectively no starch (Wang et al., 1998; Neuhaus and Stitt, 1990)
SBE2PeaSmith et al. (1990) 
Sucrose synthesis pathway, FCC estimated for sucrose synthesis
TPT0.35 TobaccoHäusler et al. (1998, 2000)Null insertion mutants of Arabidopsis show a very extreme phenotype (Schneider et al., 2002)
cFBPase0.7 ArabidopsisStrand et al. (2000)FCC over-estimated due to indirect changes: decrease of total leaf protein and photosynthesis
cPGI2Clarkia xantianaNeuhaus et al. (1989)Decreased cPGI leads to an increase of F2,6BP and inhibition of cFBPase. The precise FCC depends on the light intensity
UGPase> 10ArabidopsisMeng et al. (2009) 
SPS0.2 ArabidopsisStrand et al. (2000)Over-estimate due to an indirect decrease of total leaf protein and photosynthesis. Over-expression of heterologous SPS in Arabidopsis increases sucrose synthesis, and, under elevated CO2, photosynthesis (Signora et al., 1998)

Such studies (Stitt and Sonnewald, 1995; see Table 1 for an update) and similar results from other experimental systems led to a reassessment of the nature of the control of metabolism. It had been widely thought that metabolic fluxes were regulated by a small number of enzymes, sometimes termed ‘rate-limiting’ enzymes, which could be identified because they catalysed thermodynamically irreversible reactions. The systematic analysis of flux control in photosynthetic carbon metabolism revealed that the levels of enzymes that catalyse irreversible reactions can often be decreased substantially without having any influence on flux, at least under steady-state ambient conditions, while relatively small decreases in the amounts of some of the enzymes that catalyse reversible reactions have a marked impact on flux.

In the Calvin–Benson cycle, RubisCO (Stitt and Schulze, 1994; Mate et al., 1993), SBPase (Harrison et al., 2001; Olcer et al., 2001; Lefebvre et al., 2005) and aldolase (Haake et al., 1998, 1999) exert control over the rate of photosynthesis under growth irradiance and ambient CO2. Kinetic models had emphasized the importance of RubisCO for the regulation of photosynthesis under high light. Under irradiance, the light reactions were thought to limit photosynthesis. There was an (at least tacit) assumption in much of the research community that the regeneration of RuBP from triose phosphates would not normally restrict the rate of photosynthesis. This made the finding that aldolase and SBPase exert control over the rate of photosynthesis rather unexpected. Aldolase catalyses the reversible (see below) conversion of triose phosphates into FBP, and SBPase catalyses the first committed reaction in the regeneration of RuBP after the branchpoint to starch synthesis (see Figure 1). A similar situation, in which control was shared between enzymes, was found for sucrose and starch synthesis. Another unexpected and important finding was that TPT, which mediates a passive counter-exchange of metabolites between the chloroplast and cytosol, had a high flux control coefficient for sucrose synthesis (Häusler et al., 1998, 2000; Schneider et al., 2002). This results in a significant gradient of triose phosphates between the stroma and cytosol (Gerhardt et al., 1987), which could have important consequences for interactions between sucrose synthesis and the Calvin–Benson cycle by favouring their use for regeneration of RuBP. Another unexpected finding was that a small decrease in transketolase activity has a major impact on phenylpropanoid synthesis (Henkes et al., 2001), indicating a very close interaction between primary photosynthetic metabolism and secondary metabolism.

FCA predicts that the distribution of control will depend on the conditions. This is the case for primary photosynthesis metabolism. For example, the flux control coefficient of RubisCO varies between −0.2 (i.e. photosynthesis is increased when the amount of RubisCO is decreased) and 1 (i.e. RubisCO is totally limiting for photosynthesis), depending on the short-term conditions (i.e. the irradiance and CO2 concentrations at which photosynthesis was measured) and the long-term conditions (i.e. the conditions in which the plants had been grown) (see Stitt and Schulze, 1994, for details). Transketolase (TK) becomes almost completely limiting when the irradiance is suddenly increased above ambient growth intensities (Henkes et al., 2001). Despite its large excess under ambient conditions, PRK can become limiting when plants that were grown in low light are suddently exposed to high light (Paul et al., 2000). Plastidial fructosebisphosphatase (pFBP) acquires a significant flux control coefficient in saturating CO2 (Kossmann et al., 1994). Changes in the control of flux also occur in starch and sucrose synthesis (see, for example, Neuhaus and Stitt, 1990). The available data are consistent with the idea that the control is distributed between enzymes when photosynthesis is measured under ambient growth conditions, but that sudden changes in the conditions can lead to a one-sided limitation of fluxes by individual enzymes. This has major consequences for discussions of how, and to what extent, photosynthesis can be optimized in a fluctuating environment. It is not practical to measure large numbers of flux control coefficients under a large variety of short- and long-term conditions, and such information on its own would remain descriptive. It will be necessary to develop predictive models that build on information from studies of individual enzymes but are constrained, informed and tested by analyses of systems reponses, like those provided by FCA.

What did the Arabidopsis genome sequence reveal with respect to primary photosynthetic metabolism?

Whole-genome sequencing provides a compendium of all the genes available to perform the biological functions in an organism. Sequencing of the Arabidopsis accession Columbia-0 (Col-0) not only provided the first such comprehensive information for any plant species, but also set the gold standard against which all plant genome sequences are compared, due to its completeness, high quality and continual updating of gene annotations (Arabidopsis Genome Initiative, 2000;

It is generally accepted that green plants inherited oxygenic photosynthesis from an endosymbiotic cyanobacterium-like ancestor of the chloroplast. Sequencing of the Arabidopsis nuclear genome revealed the presence of thousands of cyanobacteria-derived genes, including many encoding photosynthetic enzymes and components of the photosynthetic electron transport chain (Martin et al., 2002). This finding has provided new insights into the origins of chloroplasts and how their metabolic functions became integrated with the rest of metabolism during the evolution of green plants.

The genome sequence provided important information about the genomic organization of metabolic pathways. There is evidence for three whole-genome duplications followed by partial loss and re-organization within the Arabidopsis genome, two of which were relatively recent, occurring after the divergence of Arabidopsis and papaya within the order Brassicales (Van de Peer et al., 2009). In addition, there are numerous local duplications. While some enzymes are encoded by single genes, indicating that their duplicated paralogues have been lost over time, others are encoded by large gene families. A list of the nuclear genes encoding each enzyme involved in the Calvin–Benson cycle, photorespiration, starch synthesis and sucrose synthesis is provided in Tables S1, S2 and S3, and summarized in Figure 3. Briefly, enzymes from the Calvin–Benson cycle are encoded by single genes or very small gene families. The largest family is for the small subunit of RubisCO (four members), which may be related, at least in part, to the need to synthesize very large amounts of this protein. Similarly, many steps in photorespiration and sucrose synthesis, and the initial steps of starch synthesis, are encoded by single genes or very small families (two members). However, larger families encode APL (four members), the regulatory subunit of AGPase, and SPS (again four members), both of which are important regulatory enzymes (MacRae and Lunn, 2006). There is also a large family of genes encoding TPT and related transporters (Schwacke et al., 2003; In this case, functional analyses showed that one encodes TPT, while the others catalyse the counter-exchange of Pi with other phosphorylated intermediates (Weber et al., 2005).

Figure 3.

 Gene families and typical levels of expression in rosettes.
For each enzymatic step, the number of annotated genes for which there is experimental evidence for the location of the gene product in the relevant compartment is indicated by the number of fields in the grey box. The gene encoding hydroxypyruvate reductase (HPR2) that is expressed at a lower level is shown in the peroxisome even though it is located in the cytosol (Timm et al., 2008). In this case, parallel reactions may occur in two compartments. The typical level of gene expression is indicated by a logarithmic false-colour scale (blue high, white average, red low; relative scale based on the ATH1 array signal) based on expression measurements over a diurnal cycle (Bläsing et al., 2005). A grey colour signifies that an unambiguous probe set is absent from the ATH1 arrays, and genes that show similarity to plastidial phosphoglucomutase but that are unlikely candidates are crossed out. The figure was generated using the MapMan visualization tool (Usadel et al., 2009a).

Viewed retrospectively, the information provided by genome sequencing provides insights into the reasons why some approaches in the pre-’omics’ era of research into photosynthetic carbon metabolism were successful, while others failed. For example, the relatively simple genomic organization of many of the gene families in the Calvin–Benson cycle and sucrose and starch synthesis explains (or in hindsight justifies) why it was possible to use relatively non-specific approaches such as antisense constructs to successfully decrease the activities of enzymes. The ability to identify mutants by forward genetics in pathways such as photorespiration was made possible by the existence of single genes for many of the steps in this pathway.

In other cases, more targeted approaches were necessary, aided by the comprehensive knowledge of gene families provided by the genome sequence. One example is identification of the aminotransferases that are involved in the pathway of photorespiration. In this pathway, two molecules of glyoxylate are transaminated to form two molecules of glycine, which are converted by glycine decarboxylase to one molecule of serine, which is then converted to hydroxypyruvate. These transformations occur via an internal serine:glyoxylate aminotransferase reaction and a glutamate:glyoxylate amino transferase reaction that is coupled with the shuttling of 2-oxoglutarate and glutamate between the peroxisome and the plastid, where the GOGAT pathway re-assimilates the ammonium released by glycine decarboxylation (Liepman and Olsen, 2001, 2003; Reumann and Weber, 2006). Other examples include the identification of a gene encoding glycerate kinase, which was the last remaining step to be defined in the pathway (Boldt et al., 2005; Timm et al., 2008), the cloning of phosphoglycolate phosphatase genes (Schwarte and Bauwe, 2007), and the finding that a peroxisomal malate dehydrogenase is part of the pathway and is required to minimize photorespiratory CO2 release (Cousins et al., 2008).

The genome sequence raised new and tantalizing questions. In particular, annotation of genes based on their sequence similarity to known genes from other organisms often led to the prediction of unexpected functions. One example was the eleven genes annotated as similar to trehalose-6-phosphate synthase (TPS), and the ten genes annotated as similar to trehalose-6-phosphate phosphatase (TPP) (Lunn, 2007b). Trehalose occurs in many life forms, including insects and fungi, and plays an important role as a stress protectant and as a storage or transport form of carbohydrate. In fungi, trehalose-6-phosphate (Tre6P) also acts as a sugar-signalling metabolite. By coincidence, at around the same time as the Arabidopsis genome sequence became available, experiments to engineer increased stress tolerance by over-expression of heterologous TPS or TPP genes led to the fascinating, but undesired, result that plant growth and development were severely negatively modified (Romero et al., 1997; Goddijn et al., 1997; Schluepmann et al., 2003). Another example was the discovery of genes for methylerythritol-4-phosphate (MEP) pathways of terpenoid synthesis (Rodríguez-Concepción and Boronat, 2002). This led to the realisation that, in addition to the classical methylvalonate (MVA) pathway in the cytosol, plants posses a second pathway for terpenoid synthesis in the the plastids. The plastid pathway is responsible for the synthesis of many important photosynthetic accessory pigments, including carotenoids, phytol and plastoquinone-9 (Lichtenthaler, 1999; Eisenreich et al., 2001).

Sequencing of the Arabidopsis genome set an agenda for further work. For many enzymes whose general functions were understood, new questions emerged regarding the specific function of other family members. New enzymes were found whose presence had not been expected. It became obvious that little or nothing was known about the function of almost half of the genes in Arabidopsis. Crucially, genome sequencing provided a holistic overview that focused attention on the need for a comprehensive and systematic approach to finding out what the individual genes do. Work along these lines soon led to the realisation that this task could not be solved on a gene-by-gene basis, but required an understanding of how genes interact in networks to perform biological functions.

Expansion of the technical toolbox

Sequencing of the Arabidopsis genome was accompanied by a large increase in the genetic toolbox, for example the development of large T-DNA insertion mutant populations (e.g.;; Alonso et al., 2003; Rosso et al., 2003), which allowed scientists to order en masse, material in which one or a set of genes of interest had been disrupted. This public service, which is now taken for granted, has enormously increased the speed of research.

At the same time, there was an explosion of so-called ‘omics’ technologies that allowed comprehensive measurements of transcript levels (Redman et al., 2004), proteins (Kleffmann et al., 2004; Baerenfaller et al., 2008), enzyme activities (Gibon et al., 2004b; Sulpice et al., 2007) and metabolites (Kopka et al., 2004; Last et al., 2007). The complete well-annotated genome was an enabling step for expression arrays and proteomics. Indeed, errors in the initial gene annotations are probably responsible for the lack of specificity of certain probe sets on commercial microarrays (see A simple message that emerges from expression arrays is that transcripts for enymes in the Calvin–Benson cycle are expressed at very high levels, as are transcripts for enzymes involved in photorespiration, whereas transcripts that encode enzymes for starch and sucrose synthesis are expressed at much lower levels (Figure 2). This broadly reflects the estimated activities and protein abundances for the enzymes (Piques et al., 2009), and mirrors the stoichimetries of fluxes in these pathways (Arrivault et al., 2009).

The availability of commercial platforms, especially arrays, drove the development of supporting resources and greatly aided data integration. However, the availability of such platforms has a downside; many scientists use ‘off the peg’ solutions but it is sometimes essential to adapt or develop new methods in order to answer a specific question. A good example is the frequent need to supplement expression arrays with use of the more sensitive and specific quantitative RT-PCR (Czechowski et al., 2004) to detect low-abundance transcripts or transcripts from large gene families where there may be considerable sequence homology between genes. The need to develop targeted analytical methods is even more pressing for analyses of metabolites. This is nicely illustrated by research in central metabolism. The pioneering studies in the 1950s that led to elucidation of the Calvin–Benson cycle measured all the intermediates in the cycle except erythrose-4-phosphate, which is extremely unstable. Over the subsequent 50 years, the number of metabolites measured in studies of photosynthetic carbon metabolism declined steadily, as researchers turned to chromatographic methods or coupled enzyme assays (see Stitt et al., 2010). First-generation GC-MS and LC-MS platforms, which are the mainstay of metabolomics studies and provide information on 1-200 metabolites from a range of metabolic pathways, do not actually detect any of the intermediates of this cycle except Fru6P. Thus, despite their undoubted usefulness, they provide almost no information about primary photosynthetic metabolism. Recently, Arrivault et al. (2009) developed a platform that allows quantitative and validated measurement of about 40 metabolites in central photosynthetic metabolism, including all the compounds that were detected by Calvin, Benson and co-workers. This requires two different LC-MS/MS systems and a set of enzyme-coupled assays.

The explosion of data resulting from studies using the new technologies required the development of public resources to evaluate and integrate these enormous data sets to allow meta-analysis by combining information about gene function, protein structure and location (Schwacke et al., 2003), comparing the expression of a selected gene across many treatments (Zimmermann et al., 2004;; Schmid et al., 2005) or searching for co-expression of genes (Steinhauser et al., 2004; Usadel et al., 2009b). The usefulness of the genome sequence has been increased by the development of detailed ontologies, such as MAPMAN (, which assigns enzymes to over 1200 categories covering almost all the pathways in central metabolism, thereby making large transcriptomic and proteomic datasets accessible to interpretation (Thimm et al., 2004; Usadel et al., 2009a). Large-scale data integration requires the use of compatible analytical platforms, consistent methods for data validation and storage, and appropriate data normalization procedures (Redestig et al., 2009; Cairns et al., 2008; Usadel et al., 2009b). This is probably most advanced for transcript analyses, due to the much wider range of analytical platforms, and least advanced for metabolite analyses, due to the bewildering complexity of chemical structures found in plants.

In the past, other parameters and aspects of metabolism were less accessible to high-throughout approaches. These included post-translational regulation, compartmentation and flux analysis. Building on continuing developments in analytical technologies and the information resources available in Arabidopsis, a further leap in our ability to analyse metabolism is on the horizon.

Cinderellas join the ‘omics’ ball

Post-translational modification

As already outlined, it has been known since the late 1970s that many Calvin–Benson cycle enzymes are subject to thioredoxin-dependent redox regulation (Buchanan and Balmer, 2005), and since the 1980s that protein phosphorylation plays an important role in the regulation of sucrose synthesis (Huber and Huber, 1996). Sequencing the Arabidopsis genome revealed the true size of the large thioredoxin family, and many other proteins involved in redox regulation, e.g. glutaredoxins, as well as hundreds of protein kinases and phosphatases.

The list of enzymes involved in primary photosynthetic metabolism that are subject to post-translational regulation continues to expand (Furumoto et al., 2001; Cotelle et al., 2000, and others noted below). One example is AGPase, which, as already discussed, catalyses the first committed step in starch synthesis. It had been known since the 1970s that AGPase is subject to allosteric regulation by the 3PGA/Pi ratio. Research in the last ten years has shown that AGPase is also subject to post-translational redox regulation, involving the reversible formation of an inter-molecular cysteine bridge between Cys81 of the two APS1 subunits of this hetero-tetrameric enzyme. The reduced and active form contains the two APS1 subunits as monomers, while the oxidized form contains dimerized APS1 and has low activity (Ballicora et al., 2000; Tiessen et al., 2002). AGPase is activated in the light in Arabidopsis rosettes (Hendriks et al., 2003). Photosynthetic electron transport leads to reduction of ferredoxin, and reducing equivalents are transferred by ferredoxin:thioredoxin reductase (FTR) to thioredoxin, which reduces the target enzymes (Buchanan and Balmer, 2005). Ballicora et al. (2000) showed that thioredoxin f, which activates several Calvin–Benson cycle enzymes, can also reduce AGPase in vitro. AGPase is also reduced/activated when sucrose accumulates in potato tubers (Tiessen et al., 2002) or Arabidopsis rosettes in the dark or the light (Hendriks et al., 2003; Lunn et al., 2006). The functional significance of this sucrose-dependent regulation is discussed below. Michalska et al. (2009) recently showed that NADP-thioredoxin reductase C (NTRC) mediates the reduction of APS1 by NADPH in vitro. NTRC is an unusual plastid-localized bi-functional protein that contains NADP-thioredoxin reductase (NTR) and a thioredoxin domain in a single polypeptide. They also used ntrc deletion mutants to provide evidence that NTRC performs this function in vivo. Compared to wild-type plants, monomerization of APS1 in the light was decreased, and the sucrose-dependent monomerization in the dark was completely suppressed.

The identification of enzymes that are subject to post-translational regulation used to require time-intensive studies of individual enzymes. Recent advances in analytical technologies have greatly accelerated this process. Hundreds of proteins have been identified that interact with thioredoxin, by using modified forms that lack one of the two cysteines in the thioredoxin active site (Motohashi et al., 2001; Marchand et al., 2004; Yamazaki et al., 2004; Buchanan and Balmer, 2005). Recent developments have allowed quantitative analysis of changes in cysteine reduction in large numbers of proteins, using isotope-coded affinity tags in combination with MS/MS analysis (Fu et al., 2008; Leichert et al., 2008). Similarly, phosphoproteomics can be used to identify the proteins that show changes in their phosphorylation status in response to a given treatment (see, for example, Wolschin et al., 2005; Niittyläet al., 2007). The PhosPhAt Arabidopsis protein phosphorylation site database ( lists Arabidopsis phosphorylation sites identified by large-scale MS experiments, and provides information about the peptide properties, their annotated biological function and the experimental and analytical context (Heazlewood et al., 2008; Durek et al., 2009). For the majority of peptides, the actual annotated mass spectrum is displayed in an interactive manner. The database also contains a plant-specific phosphorylation site predictor, trained on the experimental dataset for serine, threonine and tyrosine phosphorylation (pSer, pThr, pTyr). Users can submit protein sequences or Arabidopsis Genome Initiative gene identifiers and obtain a list of predicted phosphorylation sites. A recent meta-analysis (Baginsky and Gruissem, 2009) identified 197 proteins that were found to be phosphorylated in at least one of the published large-scale phosphoproteome studies with isolated chloroplast preparations (Reiland et al., 2009; Yu et al., 2008; Sugiyama et al., 2008; Lohrig et al., 2009). This included 39 proteins involved in carbohydrate metabolism, including enzymes in the Calvin–Benson cycle, glycolysis and gluconeogenesis. RubisCO activase, transketolase and phosphoglycerate dehydrogenase were detected in all of the studies, and GAPDH, triose phosphate isomerase, RubisCO activase, phosphoglycerate kinase, aldolase, phosphglucoisomerse and both subunits of AGPase were found to be phosphorylated in at least two studies.

High-throughput analyses short-list possible events. However, a large amount of experimental work is then required to confirm that the post-translational modification occurs, to identify the protein kinase and upstream signalling pathway, and to establish what effect phosphorylation has on the function of the protein. This often requires the establishment of an in vitro assay for the protein function (see, for example, Huber and Huber, 1996). While this is easier for enzymes than for many other types of proteins, it is still not a trivial task. Post-translational modifications are often unstable, and may revert before the protein can be assayed in vitro. For phosphorylated proteins, protein phosphatase inhibitors are usually included to prevent dephosphorylation after extraction (Baginsky and Gruissem, 2009). Thioredoxin-mediated changes in enzymes are extremely susceptible to reversion by oxidation. For example, the original studies that demonstrated a 20–40-fold increase in FBPase, SBPase and PRK activity in the light used specially designed stopped assays to measure their activity within 30 sec, before the enzymes de-activated (Laing et al., 1981; Wirtz et al., 1982). In a recent example in Arabidopsis, it was observed that the reduced form of AGPase oxidizes almost immediately in leaf extracts. To monitor the relative levels of the reduced/monomerized and oxidized/dimerized forms, it was necessary to perform the extraction in trichloroacetic acid, which instantaneously denatures the protein and prevents formation of APS1 dimers after extraction (Hendriks et al., 2003).

Post-translational modification often modifies the kinetic properties of an enzyme, rather than acting simply as an off/on switch. Therefore, it is necessary to choose appropriate assay conditions (e.g. limiting substrates, inclusion of inhibitors) to detect the functional change. Even for enzymes, it is possible that post-translational modification may influence other aspects of functionality in vivo, including cellular localization, interactions with other proteins or turnover, in which case imaging or other approaches will also be required.

Subcellular compartmentation

It has been known for a long time that compartmentation plays a major role in the organization of metabolism (Figure 1) (Lunn, 2007a). Biochemical studies of the subcellular localization of enzymes typically involved the physical isolation of organelles, followed by an investigation of which enzyme activities they contained. The scope of this approach has been vastly enhanced by the availability of whole-genome sequences and the development of large databases linked to a suite of programs that predict in silico the location of the encoded proteins. This is especially useful when linked with a critical weighting of the predictions made by the various programs (see, for example,; Heazlewood et al., 2007).

However, it is important to test these predictions. This can be done on a protein by protein basis using tags (e.g. GFP) to visualize the spatial location of the protein. The scope of traditional cell fractionation approaches has been vastly increased in the last five years by the application of proteomics (see above). This allows hundreds or thousands of peptides to be identified and their parent proteins assigned to the corresponding compartment. Such approaches have already been applied to isolated chloroplasts and their separated membrane components (see above) (Peltier et al., 2002; Ferro et al., 2002, 2003; Friso et al., 2004; Goulas et al., 2006; Block et al., 2007). It is crucial that the integrity and purity of the chloroplast preparations is critically assessed to avoid errors due to selective loss of proteins from damaged cellular structures or contamination of the preparations by proteins from other compartments. Current knowledge about the subcellular location of enzymes in primary photosynthetic metabolism is summarized in Tables S1, S2, S3 and S4.

Analysis of the subcellular location of metabolites is even more challenging. Due to the rapid turnover rates of metabolites in primary photosynthetic metabolism (see above), the compartments must either be separated within a fraction of a second, or separated under conditions that prevent changes in the levels of metabolites from occurring. Two different strategies were developed in the 1980s. One involves disrupting protoplasts by shooting them through a fine net and using a set of membrane filters located behind the net to filter out mitochondria and/or chloroplasts, before quenching the filtrate in trichloroacetic acid (Stitt et al., 1989; Gardeström, 1993). The other is non-aqueous fractionation, in which leaf material is snap-frozen, lyophilized, taken up in anhydrous organic solvents, ultrasonicated and separated by non-aqueous density gradient centrifugation (Gerhardt et al., 1987; Stitt et al., 1989). This technique is laborious, but the amount of information obtained can be vastly increased by combining it with mass spectrometric methods of metabolite profiling in order to obtain information for over 100 metabolites (D. Vosloh, S. Arrivault, A. Fernie and M. Stitt, Max Planck Institute of Molecular Plant Physiology, Golm. unpublished results).

We still know little about the supramolecular organization of primary photosynthetic metabolism. While the large amount of RubisCO makes it unlikely that this enzyme is organized in multi-protein complexes in higher-plant chloroplasts, there is already experimental evidence that some of the other Calvin–Benson cycle enzymes are organized into multi-protein complexes (Howard et al., 2008).

Cellular compartmentation

Compartmentation between cell types plays an important role in metabolism. While this is obvious for growing tissues, it has been less well recognized in photosynthetic metabolism, with the exception of C4 photosynthesis. However, it is becoming apparent that there is important compartmentation of metabolism between the leaf mesophyll and the vascular tissues in leaves (Janacek et al., 2009). A recent technical breakthrough by Bailey-Serres and colleagues in Arabidopsis opens up new perspectives for a systematic investigation of the protein complement of different cell types (Mustroph et al., 2009). They expressed a tagged ribosomal protein in the plants, and used antibodies against the tag to immunoprecipitate polysomes. Array analysis can then be performed to identify which transcripts are being transcribed, and, by implication, which proteins are being synthesized. This exciting technology was implemented in a range of transformants, in each of which a different cell-specific promotor was used to drive expression of the tagged ribosomal protein, to provide an atlas that predicts which proteins are present in which cell types.

Metabolic fluxes

Information about flux is essential to understand how metabolism is regulated, with sometimes surprising implications. For example, lowering the CO2 concentration from ambient levels towards the carbon compensation point clearly leads to lower net rates of photosynthetic carbon fixation, as well as decreased starch and sucrose synthesis. However, the simultaneous increase in the rate of the RubisCO oxygenation reaction and photorespiration can mean that the flux around the Calvin–Benson cycle remains almost unchanged (Sharkey, 1988; Portis and Parry, 2007). This affects the interpretation of metabolite measurements. In a recent study of photosynthetic responses to changes in ambient CO2 concentration in Arabidopsis, Arivault et al. (2009) found that the overall levels of intermediates in the Calvin–Benson cycle remain almost unaltered, while those in dedicated steps in starch and sucrose synthesis decreased considerably. This implies that the regulatory mechanisms outlined above are able to inhibit starch and sucrose synthesis, without requiring a large decrease in the levels of Calvin–Benson cycle intermediates.

Flux measurements often involve introduction of isotopically labelled substrates. With multicellular organisms, it is often difficult to do this without disturbing the system. This can be a particular problem when using stable, as opposed to radioactive, isotopes. Stable isotopes must be added at sufficient levels to obtain a high and stable enrichment, often requiring use of an unphysiologically high concentration. Fortunately, this problem does not arise for CO2, making photosynthesis an ideal system to analyse fluxes, as demonstrated by labelling studies using 13CO2, using analysis by GC-MS (Huege et al., 2007) and LC-MS/MS (S. Arrivault, M. Szecowka, A. Fernie and M. Stitt, unpublished results) to follow the temporal kinetics of labelling of individual metabolites.

Regulation of carbon allocation

Adjustment of carbon allocation to match demand

The following sections discuss one specific topic to illustrate how Arabidopsis research in the last decade has deepened our knowledge of the regulation of primary photosynthetic metabolism. The example we have chosen is starch turnover. Research in Arabidopsis has provided essential basic knowledge about the pathways and how they are regulated. It is starting to address two future challenges, namely to gain a deeper understanding of how metabolic proceses are linked to growth, and to understand how metabolism is adjusted to a continually changing environment. To do this, it is necessary to integrate information from various functional levels.

Research in the 1970s and 1980s led to the view that starch is an overflow product that is synthesized when the rate of photosynthesis exceeds the rate of synthesis of other end products. This occurs when the rate of photosynthesis is high, or when the temperature is decreased. Research in the 1980s indicated that feedback inhibition of sucrose synthesis can lead to a stimulation of starch synthesis, although the mechanisms were not fully clarified, and may depend on the species. Increased sucrose can lead to phosphorylation and inhibition of SPS, or to recycling of hexoses via invertase (Stitt et al., 1987; Huber and Huber, 1996; Lunn and Hatch, 1997; Trevanion et al., 2004; MacRae and Lunn, 2006). Both will result in an increase of hexose phosphates. There is good biochemical and genetic evidence that an increase in hexose phosphates leads to an increase in Fru2,6BP, which inhibits cytosolic FBPase (Stitt, 1990; Huber and Huber, 1996). This restricts the recycling of Pi from the cytosol to the chloroplasts, leading to a depletion of stromal Pi, the accumulation of 3PGA and allosteric activation of AGPase (see above). The proposed role of F2,6BP has gained additional support in the last ten years from analysis of Arabidopsis plants with altered expression of the bi-functional enzyme 6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase (F2KP), which is responsible for the synthesis and degradation of Fru2,6BP (Draborg et al., 2001; Kulma et al., 2004; Nielsen et al., 2004; Lee et al., 2006). Interestingly, F2KP is subject to regulation by phosphorylation (Furomoto et al., 2001) and binding of 14-3-3 proteins (Cotelle et al., 2000; Kulma et al., 2004).

However, it has also been known for 30 years that the amount of carbon stored as starch is precisely regulated. Starch is accumulated to allow plants to cope with the daily alternation between light and darkness. If plants are grown in conditions under which less photosynthesis is possible per day, a larger proportion of the photosynthate is temporarily accumulated as starch (Stitt et al., 1978; Chatterton and Silvius, 1979, 1980, 1981; reviewed in Smith and Stitt, 2007). Further, the rate of starch degradation is regulated such that the starch reserves are almost, but not completely, exhausted at the end of the night (Smith and Stitt, 2007). This husbands carbon reserves, and avoids a period of carbon depletion at the end of the night, which can have serious consequences for plant growth. These consequences are illustrated by the plastidial pgm mutant. This starchless mutant grows poorly in an alternating light/dark cycle with a short photoperiod (Caspar et al., 1985) because it rapidly exhausts its carbon reserves in the first part of the night, leading to an inhibition of growth that is not immediately reversed during the subsequent light period (Gibon et al., 2004a). Feedback regulation of sucrose synthesis, on its own, does not explain how more starch is synthesized to avoid carbon depletion during the night under conditions when less carbon is available, for example, in low light or short days. It also fails to explain how starch degradation is regulated and matched to the length of the night.

Clarification of the pathways of starch synthesis and breakdown

Information about the pathway is a fundamental pre-requisite for understanding how starch turnover is regulated. Research in Arabidopsis in the last ten years has further clarified the pathway of starch synthesis, and has allowed elucidation of the pathway of starch breakdown in leaves (Smith et al., 2005; Zeeman et al., 2007a; Fettke et al., 2009). Figure 4 provides a more detailed overview of the pathways of starch synthesis and degradation, and Table S4 and Figure S1 list the genes annotated as involved in starch degradation, and their expression levels in Arabidopsis rosettes.

Figure 4.

 Starch synthesis and degradation in leaves.
Starch is synthesized in the chloroplasts during the day, using Fru6P withdrawn from the Calvin–Benson cycle. At night, starch is mostly degraded via a hydrolytic route to form maltose and a lesser amount of glucose. These are exported from the chloroplast to the cytosol, where they are used to support leaf respiration and continued sucrose synthesis and export in the dark. The reactions are catalysed by the following enzymes: (1) plastidic phosphoglucose isomerase; (2) plastidic phosphoglucomutase; (3) ADP-glucose pyrophosphorylase; (4) inorganic pyrophosphatase; (5) soluble starch synthase; (6) granule-bound starch synthase; (7) starch branching enzyme; (8) isoamylase (starch debranching enzyme); (9) triose phosphate translocator; (10) glucan, water dikinase and phosphoglucan, water dikinase; (11) phosphoglucan phosphatase; (12) isoamylase (debranching enzyme); (13) β-amylase; (14) plastidic disproportionating enzyme; (15) α-amylase; (16) limit dextrinase, (17) plastidic glucan (starch) phosphorylase; (18) maltose transporter; (19) glucose transporter; (20) cytosolic disproportionating enzyme; (21) hexokinase and glucokinase; (22) cytosolic glucan phosphorylase. The genes encoding the major enzyme activities and transporters in Arabidopsis thaliana are indicated in blue.

The textbook pathway for the initial steps of starch synthesis (see above) was recently questioned. It was suggested that ADP-glucose is synthesized in the cytosol by sucrose synthase, and then imported into the chloroplasts for starch synthesis (Baroja-Fernandez et al., 2004). Arabidopsis genetics played a key role in testing the newly proposed pathway, showing that the textbook version is essentially correct, at least in leaves (see, for example, Streb et al., 2009; Barratt et al., 2009).

The conversion of ADP-glucose to starch is catalysed by a family of starch synthases and branching enzymes. It is accompanied by glucan trimming, catalysed by isoamylases, which were formerly thought to be involved only in starch breakdown (Zeeman et al., 2002; Delatte et al., 2005; Zeeman et al., 2007a). These enzymes interact to produce the appropriate mixture of the linear α(1-4)-glucan, amylose and the branched α(1-4),(1-6)-glucan amylopectin. Lesions in the pathway lead to the accumulation of maltose, phytoglycogen or other unusual polymers (Smith et al., 2005; Dumez et al., 2006; Zeeman et al., 2007a). Recent research has highlighted a special role for starch synthase 4 in controlling the number of starch grains per chloroplast (Szydlowski et al., 2009).

The pathway of starch breakdown in leaves differs fundamentally from the canonical pathway for the breakdown of extracellular starch in cereal grains (Smith et al., 2005; Zeeman et al., 2007a; Fettke et al., 2009). It involves an initial attack on the starch grain by β-amylases (Yu et al., 2005; Streb et al., 2008; Fulton et al., 2008) and plastid debranching enzymes (ISA3) (Zeeman et al., 1998; Delatte et al., 2005), followed by rearrangement of the glucans by plastid disproportionating enzyme 1 (DPE1) (Critchley et al., 2001). The main product of this pathway is maltose (Lu et al., 2005, 2006a,b; Weise et al., 2004), with some glucose. Maltose is exported to the cytosol via the maltose exporter MEX1 (Niittlyäet al., 2004). A cytosolic disproportioning enzyme (DPE2) (Fettke et al., 2006) releases one of the glucosyl moieties of maltose and transfers the other to an acceptor, probably a heteroglycan. The glucosyl moiety transferred to the heteroglycan is probably released as Glc1P by cytosolic glucan phosphorylase (Chia et al., 2004; Lu et al., 2006b; Steichen et al., 2008).

Another unexpected finding was that a cycle of starch phosphorylation and dephosphorylation is required for the primary attack on the starch granule in the chloroplast (Ritte et al., 2004; Edner et al., 2007). Starch is phosphorylated by glucan, water dikinase (GWD, also known as SEX1) (Yu et al., 2001; Ritte et al., 2002; Mikkelsen et al., 2004, 2005) and phosphoglucan water dikinase (PWD) (Kötting et al., 2005; Baunsgaard et al., 2005), and dephosphorylated by at least one phosphoglucan phosphatase (SEX4) (Niittyläet al., 2006; Kötting et al., 2009).

There are two additional pathways for starch degradation in chloroplasts, a hydrolytic route catalysed by α-amylase (Delatte et al., 2006; Zeeman et al., 2007b) and a phosphorolytic route via plastidic glucan phosphorylase. The latter presumably provides hexose phosphates for operation of the plastid-located oxidative pentose phosphate pathway in the dark (see Stitt and ap Rees, 1980). Under certain conditions, starch is also degraded in the light via glucan phosphorylase (Zeeman et al., 2004) to form Glc1P, which is fed into the Calvin–Benson cycle (Lu et al., 2005). One example is in low CO2, when photorespiration is high and acts as a drain on Calvin–Benson cycle intermediates.

One of the key approaches to unravelling the pathway of starch degradation was isolation of a large number of starch excess (sex) mutants. A number of the affected genes have been identified and shown to encode enzymes involved in the pathway of starch degradation, or in the cycle of glucan phosphorylation and dephosphorylation that is required for the primary attack on the starch granule. Another fruitful approach was data mining of Arabidopsis expression profiling data sets, which allowed further proteins putatively involved in starch degradation to be short-listed and functionally identified. For example, following the identification of SEX1 as a glucan, water dikinase, sequence similarity searches and co-expression analyses identified a similar protein (PWD, phosphoglucan water dikinase), which was then shown to also be involved in a further cycle of glucan phosphorylation (Hejazi et al., 2009).

Regulation of starch synthesis in response to demand

The discovery that AGPase is subject to post-translational redox regulation (see above) provided the first clue about how starch synthesis is regulated to provide carbon for growth and metabolism during the following night. In an undisturbed light/dark cycle, most of the AGPase remains in the low-activity dimeric form, even in the light. A shortfall of carbon at the end of the night, however, leads to a temporary restriction of growth in the subsequent photoperiod, stronger activation of AGPase, and activation of starch synthesis. There is evidence implicating the sugar signalling molecule Tre6P (Kolbe et al., 2005; Lunn et al., 2006) and the SNRK1 protein kinase (Jossier et al., 2009) in the signal transduction pathway that mediates this sugar-dependent modulation of starch synthesis.

It may be anticipated that post-translational redox regulation of AGPase interacts with the previously described feedback regulation of sucrose synthesis, mediated by Fru2,6BP. Crucially, post-translational activation of AGPase allows increased starch synthesis while decreasing the levels of phosphorylated intermediates, whereas the feedback pathway mediated by Fru2,6BP leads to an increase in the levels of phosphorylated intermediates. This network allows a remarkably flexible regulation of carbon allocation. We recently replaced the Cys81 residue of the APS1 subunit with serine, using site-directed mutagenesis, and expressed the resulting constitutively activated C81S form under the control of the native APS1 promotor in the adg1 mutant (N. Haedrich, M. Stitt and J. Lunn, unpublished results), which is deficient in APS1 (Wang et al., 1998). ADPGlc is undetectable in wild-type Arabidopsis rosettes in the dark; however, low but readily detectable amounts of ADPGlc were found in the adg1 mutant complemented by the C81S form of APS1. This demonstrates that redox regulation is required to allow complete inactivation of AGPase in the dark. The diurnal turnover of starch was also modified, providing genetic evidence that redox regulation contributes to the coordination of starch synthesis and breakdown during the 24 h light/dark cycle Similarly, over-expression of a mutated form of AGPase that is more sensitive to allosteric activation in the adg1 mutant led to an increase in transitory starch synthesis (Obana et al., 2006). However, to achieve large changes in carbon allocation, it may be necessary to combine changes in the properties and expression levels of several enzymes.

A further complication is introduced by the differential expression of various members of the APL family encoding the large subunit of AGPase (see above). The kinetic properties of AGPase depend on which type of APL subunit is present in the hetero-tetrameric enzyme, with APL1 being the major isoform in leaves (Crevillén et al., 2003). It is noteworthy that application of trehalose induced expression of APL3 and was associated with accumulation of starch (Wingler et al., 2000; Fritzius et al., 2001; Ramon et al., 2007). At present, it is not clear whether this is because high exogenous trehalose levels lead to an increase of Tre6P, or whether trehalose itself also regulates starch metabolism.

Regulation of starch degradation

Although our understanding of how starch degradation is regulated is less advanced than that of starch synthesis, research in Arabidopsis is opening up new perspectives. Like the study of the breakdown pathway, this research has profited from many components of the Arabidopsis genomics toolbox.

One intriguing possibility is that the cycle of glucan phosphorylation and dephosphorylation serves not only to facilitate the attack on the starch grain (Edner et al., 2007; Kötting et al., 2009) but also to regulate the rate of breakdown. The observation that glucan phosphatases are involved in glycogen metabolism in mammals adds weight to this hypothesis, in what appears to be a fascinating example of conservation or convergence of regulatory function between plants and animals (Niittyläet al., 2006). It is striking that several enzymes involved in starch degradation are subject to redox regulation (Mikkelsen et al., 2005; Sparla et al., 2006; Sokolov et al., 2006), raising the question of how this is regulated and what function it serves. One of the β-amylases in chloroplasts, BAM4, is essential for starch degradation, acting upstream of BAM1–3, but lacks catalytic activity, indicating it may have an as yet unknown regulatory function (Fulton et al., 2008). Co-expression analyses (Li et al., 2007) and sequence similarity searches for glucan-binding proteins (Kerk et al., 2006; Lohmeier-Vogel et al., 2008) have uncovered putative candidates for a role in controlling the structure of starch or its degradation (Mentzen et al., 2008; Li et al., 2009).

One very exciting finding has emerged from expression profiling. Almost all genes for the enzymes of starch degradation are under strong circadian control, resulting in strong oscillations of their transcripts in a free-running cycle and in day/night cycles (Lu et al., 2005; Smith et al., 2004; Usadel et al., 2008). Further, starch and maltose levels oscillate in a circadian manner when plants are transferred to continuous light (Weise et al., 2006). Intruigingly, the rate of starch degradation in Arabidopsis immediately adjusts to an unexpected change in the timing of dusk, being slowed down when plants are subjected to darkness earlier, and speeded up when the light period is extended (Lu et al., 2005). These observations imply that a timing mechanism within the leaf matches the rate of starch mobilization to the anticipated length of the night. It is important to clarify whether this is due to regulation of starch turnover by the circadian clock. Recently, investigations of the response of starch degradation to growth under 18.5 and 28 h T-cycles (i.e. day/night cycles lasting 18.5 and 28 hours, respectively) and in the lhy/cca1 double mutant have shown that the rate of starch degradation is set such that starch is exhausted at dawn as anticipated by the circadian clock (A. Graf and A.M. Smith, John Innes Institute, Norwich, UK, personal communication). This finding poses two further questions: how does the plant know how much starch there is, and how does it integrate this with information about the anticipated length of the night to set an appropriate rate of starch degradation?

Adjusting photosynthate allocation and growth to large changes in the carbon supply

Starch is used to support respiration and growth during the night. The changes in starch turnover discussed above must be accompanined by changes in the rate at which carbon is used, otherwise the plant will still become carbon-starved. Arabidopsis shows a remarkable ability to adjust to an extreme shortfall in the carbon supply. It adjust the rates of starch synthesis and starch degradation across a wide range of photoperiods, allowing a similar reserve of starch, and also of sugars, to be retained at the end of night, even if a 3 h light period is used (Gibon et al., 2009). Across this wide range of photoperiods, the relative growth rate is strongly correlated to the rate of starch degradation (Gibon et al., 2009). These observations indicate the presence of a highly sophisticated regulation network that coordinates growth with the carbon supply and starch turnover.

Recent research shows that protein synthesis is an important component in this network. There was a progressive decrease of total protein when Arabidopsis was grown under increasingly short photoperiods. In a given photoperiod, protein levels were lower in the starchless pgm mutant than in wild-type plants (Gibon et al., 2009; Hannemann et al., 2009). Thousands of genes undergo diurnal changes of their transcript levels (Bläsing et al., 2005), driven by the circadian clock, light and sugars (Usadel et al., 2008). This includes many transcripts that encode enzymes in central metabolism, and hundreds of genes involved in amino acid, nucleotide synthesis and protein synthesis, including the majority of the ribosomal proteins.

One of the major questions for the future is to learn how signal metabolites such as Tre6P, sugar- and stress- regulated protein kinases such as AKIN10 and AKIN11 (Baena-Gonzáles et al., 2007; Jossier et al., 2009), and other sugar-signalling components interact to regulate starch turnover and the use of carbon for protein synthesis and growth. Recently, it was reported that a lesion in starch breakdown triggers chloroplast degradation (Stettler et al., 2009), indicating that further signals may be generated in the starch breakdown pathway.

It will also be important to integrate these molecular analyses with quantitative information about the rates of protein synthesis, and the associated carbon and energy costs. Physiological studies have shown that protein synthesis represents a major energy cost during growth, and that maintenance respiration correlates with the tissue protein content (de Vries, 1975; Hachiya et al., 2007; see Warner, 1999; for a review of background information in yeast). Recently, Piques et al. (2009) used quantitative RT-PCR in combination with external RNA standards to quantify rRNA species and more than a hundred transcripts in polysomes from Arabidopsis leaves harvested in the light or the dark. The results were used to estimate the maximum global rate of protein synthesis, the associated carbon and energy costs, and the rates of synthesis of over 30 individual enzymes (see below for more discussion). Their calculations showed that protein synthesis is a major component of the plant carbon budget. In particular, it consumes a large part of the carbon supplied by starch mobilization at night. These are the first steps towards developing a rigorous framework in which the impact of molecular events such as ribosome biogenesis and protein synthesis on plant carbon and energy balance and growth can be predictively modelled.

Coping with an unpredictable environment

Diurnal changes occur in a regular manner, every 24 hours. Plants also need to to cope with environmental changes that are superimposed on diurnal changes and occur on vastly different time scales. These include irregular changes due to changing weather on a particular day, and slower alterations due to shifting weather patterns, seasonal effects and changes in stand structure. As discussed in the section on flux control analysis, sudden changes can lead to a loss of balance in a network. If the environmental changes are sustained, adjustment may require changes in the levels of proteins. This section discusses some recent work that illustrates the scope that Arabidopsis possesses to adjust to a changing environment.

We have already discussed how the rate of starch breakdown can be immediately adjusted to changes in the starch content at the start of the night. A second instructive insight into the flexibility of carbon allocation comes from studies of Arabidopsis tpt knockout lines, which lack a functional TPT (Schneider et al., 2002; Walters et al., 2004). As they are unable to export triose phosphates, these plants cannot synthesize sucrose by the usual pathway in the light. Instead, they convert almost all their photosynthate to starch and simultaneously degrade it to hexose, which is released from the plastid in the light. This involves a different pathway to that used for starch degradation in wild-type plants in the dark, when maltose is exported to the cytosol. The extent to which starch is remobilized in the light depends on the ambient conditions, with little or no degradation under low light, and rapid breakdown under high light (Walters et al., 2004). The starch degradation pathway operating in the tpt mutants in the light still needs to be defined. However, the tpt/sex1 double mutant is still able to degrade starch and export carbon in the light, showing that the release of hexose sugars in the light does not require the cycle of glucan phosphorylation and dephosphorylation that is required during starch degradation in the night (Schneider et al., 2002). Despite this major disruption of carbon allocation, tpt mutants show little impairment in growth (Schneider et al., 2002; Walters et al., 2004). This highlights the exceptional flexibility in these central metabolic pathways, and implies that plants are able to adjust their growth patterns to a major change in the diurnal carbon supply. This network property may allow plants to cope with short-term and non-predictable fluctuations that are imposed on the diurnal light/dark cycle in the real world, such as cloudy days or changes in canopy structure.

As already discussed, many transcripts show large diurnal changes. These are probably modified by the actual conditions on any one day. To understand the functional significance of changes in individual transcripts, it is necessary to investigate their impact on the levels of the encoded proteins. This will largely determine whether, and how quickly, these changes of gene expression lead to changes in the metabolic network. Gibon et al. (2004b) developed a robotized platform to profile the maximum activities of over 20 enzymes, including many involved in primary photosynthetic metabolism. Most show marked diurnal changes of their transcript levels but small diurnal changes of activity. When plants are transferred to continuous darkness, transcript levels respond within hours, but changes of enzyme abundance require days (Gibon et al., 2004b, 2006). Most of the 30 enzymes studied by Piques et al. (2009) were synthesized at rates that are of the same order as the rate of growth of the plant. Put simply, there is not enough transcript, relative to the amount of enzyme, to allow a rapid increase of the enzyme. This confirms the hypothesis of Gibon et al. (2004b) that the rate of translation of enzymes is so slow that several days are required to produce a major change in protein abundance. Instead of leading to immediate changes, rapid transient changes of transcripts are integrated over a longer period of time to set the levels of enzymes and other proteins. This will buffer enzymatic capacities in central metabolism against recurring changes caused by the light/dark cycle and temporary changes in the weather, while allowing them to adjust to sustained changes in the plant’s surroundings.

It is important to extend such analyses to a larger range of proteins, including the enzymes involved in starch synthesis and degradation. The data available so far indicate that large diurnal changes of their transcripts (see above) do not lead to significant changes in the levels of most of the encoded proteins (Smith et al., 2004). This indicates that diurnal changes of transcripts for enzymes involved in starch turnover set the levels of the enzymes over a longer time frame, and that post-translational regulation is responsible for the fine tuning of starch turnover to the non-predictable fluctuations that occur from day to day. This can be also tested more rigorously by large-scale proteomics analyses, especially for proteins whose activity cannot be conveniently measured (Baerenfaller et al., 2008). The high costs of protein synthesis imply that there is a trade-off between the speed with which leaf protein levels can be altered in response to new conditions, and the rate of growth. This is an aspect where comparisons between species or even between accessions or (for crops) cultivars might yield interesting insights.

Natural diversity in Arabidopsis as a new resource to perturb networks

In addition to its value for forward and reverse genetics, the large natural genetic diversity in Arabidopsis offers many opportunities for plant research (Koornneef et al., 2004). The hundreds of defined accessions and numerous inbred populations available as public resources are of particular importance, and have long been used for studying traits such as flowering time, seed germination and abiotic stress tolerance. Although we might have expected to find little diversity in photosynthetic metabolism, in view of its central importance in plant metabolism, recent studies have uncovered considerable diversity in primary photosynthetic metabolism, and started to relate this to plant growth.

One recent study uncovered genetic variation in the activity of a set of enzymes involved in primary photosynthetic metabolism in a Cvi x Ler recombinant inbred line population (Keurentjes et al., 2008). Parallel analysis of transcript levels, enzyme activities and metabolites identified strong quantitative trait loci for UDP-glucose pyrophosphorylase and phosphoglucomutase activities, and weaker quantitative trait loci for several other enzymes. In some cases, the quantitative trait loci co-located with structural genes for the corresponding enzymes, and were accompanied by changes in the levels of the encoding transcripts. Re-sequencing of the genomes of the hundreds of Arabidopsis accessions will provide detailed information about sequence polymorphisms in the promoters and protein coding regions of the genes, opening up the possibility of integrating in silico and experimental approaches to identify functional variants of enzymes, and explore the consequences for plant growth (see, for example, Zhang et al., in press).

Other studies have uncovered a relationship between starch turnover, protein content and biomass. Meyer et al. (2007) detected a large number of weak and non-significant negative correlations between individual metabolites and biomass in a C24 x Col-0 recombinant inbred line population, but multivariate analysis of the data identified a set of metabolites that provide a highly significant prediction of biomass. Sulpice et al. (2009) applied a similar approach to a large population of accessions. Biomass was negatively and very significantly correlated both with the level of starch at the end of the light period and with protein content. Further, an identical set of metabolites correlated with biomass, starch and protein. Starch is the main carbon reserve that supports metabolism and growth at night, and protein synthesis is a major energy cost (see above). These results indicate that large accessions are able to use their starch reserves more efficiently, and suggest that a lower protein level is one of the factors that allows them to do this (Sulpice et al., 2009). Correlations were also recently reported between expression of circadian clock genes, increased chlorophyll content, and carbohydrate accumulation in Arabidopsis hybrids and alloploids (Ni et al., 2008). Using comprehensive sequence information, it is possible to detect trait associations with specific polymorphisms in candidate genes (Sulpice et al., 2009). This could provide a powerful approach to link genetic diversity in central metabolism with important whole-plant traits.

Future perspectives

Research into primary photosynthetic metabolism faces a series of new challenges, including integration of the various sub-processes in photosynthesis, understanding how developmental processes are regulated to provide an appropriate cellular organization for photosynthesis, and obtaining a predictive understanding of the genetic and molecular mechanisms that allow photosynthesis to adjust to different environmental conditions.

Large multi-layered data sets including protein levels (Baerenfaller et al., 2008), enzyme activities (Gibon et al., 2004b; Piques et al., 2009) and subcellular metabolite levels, as well as increasing amounts of data about post-translational modifications, will soon become available for Arabidopsis. This will allow systems analyses of the type currently being undertaken in Escherichia coli (Bennett et al., 2009), and provide a comprehensive set of values for use in metabolic modelling. It will be a major challenge for the future to incorporate the newly available wealth of information in integrative models, building on and extending the basic models of photosynthesis that are already available (Farquhar et al., 2001; Zhu et al., 2007; Poolman et al., 2009).

A related challenge will be to integrate events in photosynthetic carbon metabolism with the regulation of electron transport and ATP synthesis in the thylakoid membranes. These generate the ATP and NADPH that are consumed in the Calvin–Benson cycle. Powerful new methods are now available that allow fluxes in the electron transport pathways and ATPase to be measured (e.g. Takizawa et al., 2007), opening up new perspectives for understanding how these processes interact and are regulated. It is also important to understand how the expression and regulation of the photosynthetic apparatus is integrated with the regulation of leaf development. Effective operation of the sophisticated machinery depends upon its correct spatial distribution to maximize light capture and nitrogen use, and to minimize diffusion paths for incoming CO2.

It is important to translate our increasing understanding of photosynthetic processes into an increase in crop photosynthetic rates and yield. Exciting progress has already been made in engineering an altered photorespiration pathway in Arabidopsis, in which the CO2 is released in the plastids, resulting in a higher efficiency of CO2 recapture (Kebeish et al., 2007). The finding that SBPase may limit the rate of RuBP regeneration under some conditions (see above) suggests another possibility to increase photosynthetic rates (Miyagawa et al., 2001; Lawson et al., 2006). Evolutionary models have also predicted that increasing the amount of protein invested in SBPase should increase the rate of photosynthesis (Zhu et al., 2007). It will also be important to explore further how altering allocation to starch (see above) impacts on the rate of growth.

Finally, to be provocative, it could be argued that single-celled green algae might provide a better photosynthetic model system than Arabidopsis in the coming era of systems biology. Modelling requires simultaneous measurements of multiple parameters over time courses. Such measurements are more easily performed in algae, with Calvin and Benson, which is where we started. Is the sun going down on Arabidopsis research? Probably not. Although it may be anticipated that further major insights into the regulation of photosynthesis will be obtained in algae, the largest challenge is not just integration of the various sub-processes of photosynthesis, but also integration of photosynthetic research with questions related to anatomy, development and plant growth. This will be the task of the next one or two decades, and can only be performed in a higher plant, so we can confidently predict that Arabidopsis photosynthesis research has a bright and sunny future.


We acknowledge the Max Planck Society, the European Commission (FP6 Integrated project ‘Agronomics’ LSHG-CT-2006-037704) and the German Ministry of Education and Research (GoFORSYS 0313924 and GABI 0315049A) for financial support. We are grateful to Alex Graf and Alison M. Smith (John Innes Institute, Norwich, UK) for dicussions and permission to mention unpublished results.