Kinetics of labelling of organic and amino acids in potato tubers by gas chromatography-mass spectrometry following incubation in 13C labelled isotopes


(fax +49 (0)331-5678408; e-mail


Metabolic pathways of primary metabolism of discs isolated from potato tubers were evaluated by the use of a gas chromatography-mass spectrometry (GC-MS) method generated specifically for this purpose. After testing several possible methods including chemical ionization, it was decided for reasons of sensitivity, reproducibility and speed to use electron impact ionization-based GC-MS analysis. The specific labelling and label accumulation of over 30 metabolites including a broad number of sugars, organic and amino acids was analysed following the incubation of tuber discs in [U-13C]glucose. The reproducibility of this method was similar to that found for other GC-MS-based analyses and comparison of flux estimates from this method with those obtained from parallel, yet less comprehensive, radiolabel experiments revealed close agreement. Therefore, the novel method allows quantitatively evaluation of a broad range of metabolic pathways without the need for laborious (and potentially inaccurate), chemical fractionation procedures commonly used in the estimation of fluxes following incubation in radiolabelled substrates. As a first experiment the GC-MS method has been applied to compare the metabolism of wild type and wel-characterized transgenic potato tubers exhibiting an enhanced sucrose mobilization. The fact that this method is able to rapidly yield further comprehensive information into primary metabolism illustrates its power as a further phenotyping tool for the analysis of plant metabolism.


Developing potato tubers, like many other plant heterotrophic systems, are characterized by large metabolic fluxes to storage metabolites such as starch, and protein, as well as substantial fluxes into respiratory metabolism and the maintenance of cellular sucrose levels (Fernie et al., 2002a,b; Geigenberger, 2003; Stitt and Sonnewald, 1995; Sweetlove et al., 1999). Whilst methods for evaluating the starch biosynthetic flux and the extent of sucrose degradation and (re)synthesis are firmly established (Fernie et al., 2001a,b; Geigenberger et al., 1997, 2000; Sweetlove et al., 1996), precise methods for evaluating the exchange of carbon between the individual pools of metabolites such as organic and amino acids have not yet been developed.

Given the industrial importance of starch, the quantities that this polymer accumulates in the tuber and the ease of genetic transformation, potato (Solanum tuberosum) has been a target for metabolic engineering. However, despite intense research effort (see Fernie and Willmitzer, 2004 and references therein), only a handful of attempts at increasing starch yield have proved successful (Regierer et al., 2002; Stark et al., 1991; Tjaden et al., 1998). Surveys of the metabolite levels of a wide range of transgenic plants revealed that many of the manipulations that failed to achieve the hoped for enhancement of starch accumulation were in fact upregulated in alternative pathways (Roessner et al., 2001a; Roessner-Tunali et al., 2003a). Furthermore, by the use of 14C labelling studies it was demonstrated that these same lines were characterized by an increased rates of respiration (Fernie et al., 2002a; Trethewey et al., 1998, 2001).

In recent years several studies have utilized nuclear magnetic resonance (NMR) for the quantification of intracellular fluxes in plant cells following incubations in stably labelled isotopes (Dieuaide-Noubhani et al., 1995; Fernie et al., 2001c; Glawischnig et al., 2001; Rontein et al., 2003; Roscher et al., 2000; Schwender et al., 2003). In several instances such studies were augmented by the use of more sensitive gas chromatography-mass spectrometry (GC-MS)-based methods (Edwards et al., 1998; Giege et al., 2003; Schwender et al., 2003), however, these were generally used to address specific biological reactions and therefore only allowed the quantification of a handful of fluxes. A notable exception is the recent work of Schwender et al. (2003) who used a combination of GC-MS and NMR techniques to facilitate the construction of a model of glycolysis and the oxidative pentose phosphate pathway in developing Brassica napus embryos. Here we present a method for evaluating isotope distribution, based on GC-MS alone, in which determination of the mean fractional isotopic enrichment of the major sugar organic and amino acid constituents of plant cells following incubation in 13C-labelled glucose was used as the experimental criteria for the evaluation of unidirectional rates of carbon exchange. As a first example, we examined the metabolism in tuber discs cut directly from developing tubers of wild type and two very well characterized transgenic lines (exhibiting enhanced sucrose mobilization and an accumulation of amino acids). Theoretically, this method allows us to determine 30 such rates. However, several of these were negligible in both the wild type and transgenic systems analysed and furthermore the labelling kinetics of the metabolites in the experiment presented places an added restriction on the number of metabolites for which absolute rates of carbon exchange can be obtained. That said this method allowed us to quantitatively evaluate these rates without the need for the laborious, and potentially inaccurate, extensive fractionation and enzymatic digestion procedures commonly used in the estimation of fluxes following incubation in radiolabelled substrates.

Results and discussion

Rationale for developing an MS-based analysis platform for isotope labelling kinetics

Flux analyses in plants are generally still performed following either incubation of plants (or plant material) in radiolabelled precursors and subsequent chemical fractionation by a series of ion- exchange and thin layer chromatographic separations in tandem with enzymatic digestions, or NMR-based approaches. Although very useful each of these approaches suffers from drawbacks. The fractionation of radiolabel is highly labour intensive and generally only provides a low level of resolution that is limited either to end-products (such as sucrose and starch) or compound classes groups (e.g. organic and amino acids represent single pools see Figure 1a). That said this technique has proven vitally important in previous studies of starch metabolism (Coates and ap Rees, 1994; Hill and Smith, 1991; Tauberger et al., 2000) and has found both historical (Graham and ap Rees, 1965; Lips and Beevers, 1966) and recent (Gage et al., 1997; Wheeler et al., 1998) utility in the elucidation of metabolic pathways in plants. In contrast, NMR analysis allows the evaluation of specific compounds and importantly, as it detects atomic as opposed to chemical properties, can yield important information as to the position at which a molecule is labelled. For this reason it has long been used in the analysis of bacterial and mammalian fluxes (Des Rosiers et al., 1991; Katz et al., 1989; Szyperski, 1998), and has more recently been adopted by a number of researchers in the plant field (see Roscher et al., 2000). The major disadvantage that NMR spectroscopy has in comparison with mass spectrometric methodologies is that of sensitivity with NMR generally being restricted to determinations of the more abundant cellular constituents. Given these constraints we decided to develop a method for analysis of 13C-labelled plant extracts using mass-spectrometry. Our first attempt was to prepare methanol extracts as described in Roessner et al. (2000), and to ionize these extracts chemically using NH3. As chemical ionization is less harsh than electron impact ionization it would be expected to leave more metabolites in their in vivo form. Unfortunately, despite repeated trials, following this approach we were unable to detect a suitable range of 13C-labelled analytes for the purposes of this work. Therefore, we next attempted to adapt a pre-established electron impact-based GC-MS method for analysis of steady-state metabolite levels of potato tuber tissue that we know exhibits high reproducibility of both analyte retention time and peak responses (Roessner et al., 2000, 2001a). Using this method allowed the detection of labelled metabolites with relative ease, however, given that one of the derivitizing agents we use – (N-methyl-N-trimethylsilyl)trifluoroacetanide (MSTFA) – consists of a mixture of isotopes great caution must be taken to ensure the labelling pattern determined reflects the redistribution of isotope during the incubation process rather than the derivitization itself. For this reason we decided to fully process control samples that had been incubated in unlabelled glucose (exhibiting only the natural abundance of 1.11%13C) alongside every sample incubated in 13C glucose. As the derivitization process has already been optimized and proven to be robust (Roessner et al., 2000), these control samples allow a valid measurement of the amount of analyte that has been labelled. The identity of each unlabelled peak was established by comparison of the mass spectra with those stored in our in-house library (, once these were assigned peaks were searched for that indicated the presence of label (i.e. between m + 1 and m + 12 of each unlabelled peak). As electron impact ionization was used not every peak corresponded to the molecular ion of the metabolite in question, however, in such instances the natural abundance pattern of the fragment analysed was studied to validate the assumption that the labelling of this fragment was representative of the labelling within the entire molecule. In instances where it was impossible to quantify the labelling pattern of the molecular ion we routinely used the highest mass fragment. It is, theoretically, possible that averaging the labelling of multiple fragments could yield higher accuracy. However, given that the evaluation of replicate samples using a single ion for quantification yielded highly similar results we chose to pursue this simpler method as it obviates the need for intensive computer and manual interrogation of the datasets. Following these analyses we found that we were able to reliably detect labelling in 18 amino acids, five organic acids and a total of eight sugars, sugar alcohols and phosphoesters. From this list we constructed a metabolic scheme of the unidirectional exchange rates that are theoretically determinable using this GC-MS protocol (Figure 1b). It is worth noting, however, that these metabolites are by no means the only ones in which label can be reliably detected in, merely the ones we were interested in for the purposes of this study. We routinely use a very similar protocol to determine the levels of some 80 known and 40 unknown metabolites in potato (Roessner et al., 2001a,b) and 90 known and 80 unknown metabolites in tomato (Roessner-Tunali et al., 2003b). Whilst detection of label is some extent dependent on metabolite abundance the exact number of metabolites for which reliable isotope quantification is possible will be conditional. However, it should be possible to detect labelling in over 50% of these compounds, under normal physiological conditions, using this exact same method. When throughput is severely compromised and the GC-MS runs in single ion detection mode it should be possible to detect the labelling pattern of all these compounds.

Figure 1.

Schematic diagrams of reaction rates that can be studied following incubation in [U14-C]glucose and [U13-C]glucose.
Compound classes of metabolites that can be determined following [U14-C]glucose feeding (a), represent bulk fluxes all of which derive from the hexose-P pool (the summed carbon in glucose 6-phosphate, glucose 1-phosphate, fructose 6-phosphate and 3-phosphoglycerate) and allow calculation of the absolute flux to starch, cell wall, sucrose and glycolysis. Metabolites that can be determined following [U13-C]glucose feeding (b), arrows represent unidirectional carbon exchange rates between metabolic precursors and individual sugars, organic acids and amino acids that can be studied (those metabolites set in italic type can not be directly measured by the techniques reported in this paper). Those metabolites sharing a common precursor are grouped together.

Experimental set up

Having established which metabolites this method allowed us to determine appreciable labelling in we next decided to characterize the labelling kinetics of pools of these metabolites in potato tubers. For this purpose we isolated tuber discs from developing tubers of healthy wild type plants and after washing thoroughly to remove damaged cells incubated these in the presence of 20 mm [U13C]glucose for 2, 3, 5 or 12 h. After the appropriate time period discs were again washing thoroughly and frozen in liquid nitrogen. Subsequently discs were extracted in methanol, derivatized and injected onto the GC-MS column. The chromatograms obtained were then manually evaluated and mean fractional enrichments of every metabolite were calculated as described in Giege et al. (2003). The labelling kinetics of all determined metabolites are presented in Figure 2. These data reveal that the majority of metabolites appear to approach isotopic steady-state 5 h into the incubation, a feature that is reflected also in the total metabolite levels during this time course (data not shown). This is true for all metabolites except sucrose (2U), 3PGA (2X), fructose 6-phosphate (2Y) and glucose 6-phosphate (2Z). The labelling of these metabolites are not so straightforward as, although they also seem to be approaching steady state, their labelling kinetics are difficult to determine as a result of subsequent reduction in labelling following the 5 h time-point. Between 5 and 12 h the labelling of glucose and fructose slightly decreased whereas that from 3-PGA, glucose 6-phosphate and fructose 6-phosphate declined massively, in contrast there were large increases in the labelling of alanine, glutamine, isoleucine, leucine, phenyalanine, tryptophan, tyrosine, sucrose and quinate. These changes can be rationalized given the conditions experienced by tuber discs over such a long incubation. Only two hypotheses can be put forward to explain the observed decreases in the relative abundance of 13C in the sugar and sugar phosphate pools. The first of these, preferential turnover of 13C as opposed to 12C containing molecules can probably be excluded as it is unlikely that the decreases would be confined to the sugars and sugar phosphates if this were the cases particularly given that the absolute levels of all these metabolites increase. Thus, the most likely explanation is the provision of unlabelled precursors at later stages of the incubation. Given that unlabelled substrate was never supplied exogenously it must come from the breakdown of stored cellular carbon – namely starch. Starch turnover has previously been observed in tubers (Sweetlove et al., 1996) and although the mechanisms by which it is upregulated remain unknown from the current study it is tempting to speculate that it occurs to meet the extra carbon demand placed on the cell by, for example, the upregulation of the shikimate pathway. Comparison of the labelling kinetics of the sugars reveals marked differences with glucose and fructose approximating equilibrium relatively quickly, whereas the biphasic labelling of sucrose indicates that sucrose is compartmented with the slower second phase of labelling best explained by a second relatively inaccessible pool of sucrose. This finding is in keeping with previous kinetic studies of central metabolism in heterotrophic tobacco cells (Fernie et al., 2001c), and with direct measurement of sugar levels within the various compartments of potato tubers (Farre et al., 2001). The elevation of the rate of label incorporation in amino acids derived from the shikimate pathway most probably reflects the wound response following the slicing of potato tubers. It has previously been demonstrated that slicing of tubers leads to an increase in the activities of glycolytic enzymes, a stimulation of respiration and an induction of secondary metabolism (Burton, 1989; Kahl, 1974; Lamb et al., 1979) including a dramatic elevation in the activity of phenylalanine lyase within 10 h (Geigenberger et al., 2000).

Figure 2.

Specific labelling of metabolites during incubation of wild type potato tuber discs with [U-13C]glucose. Discs cut directly from developing tubers were incubated in the presence of 10 mm Mes-KOH (pH 6.5) and 20 mm [U-13C]glucose for various time intervals before they were washed and extracted to determine label distribution. Mean mole fractional enrichment was then determined after GC-MS analysis as described in Experimental procedures. Wherever possible the molecular ion was used for this calculation, in instances where this was not possible it was assumed that the labeling pattern of the mass fragment used reflected that of the complete molecule. The kinetics of labelling of the following compounds are shown: (a) alanine; (b) aspartate; (c) asparagine; (d) β-alanine; (e) glycine; (f) glutamine; (g) glutamate; (h) isoleucine; (i) leucine; (j) lysine; (k) methionine; (l) phenylalanine; (m) proline; (n) tryptophan; (o) serine; (p) threonine; (q) valine; (r) tyrosine; (s) glucose; (t) fructose; (u) sucrose; (v) inositol; (w) mannitol; (x) 3-PGA, 3-phosphoglycerate; (y) Fru-6-P, fructose 6-phosphate; (z) Glc-6-P, glucose 6-phosphate; (aa) citrate; (bb) quinate; (cc) malate; (dd) fumarate; (ee) succinate. The results are mean ± SE (n = 6), error bars are not shown where they are smaller than the symbol.

From the above analysis we decided to focus on the 5 h incubation period for our further studies as at this time the majority of pools appear to be approaching isotopic steady state and the input into the central hexose and hexose-phosphate pool from starch degradation is still relatively minor. For this purpose we incubated tuber discs isolated from wild type tubers and from transgenic potato plants exhibiting tuber-specific expression of either a yeast invertase (Sonnewald et al., 1997) or a bacterial derived sucrose phosphorylase (Trethewey et al., 2001) for this incubation period in the presence of either (i) 20 mm [U-14C]glucose, (ii) 20 mm [U-13C]glucose or (iii) 20 mm unlabelled glucose. The first of these was carried out for comparative purposes, whilst the last served as a control against which the [U-13C]glucose-fed samples could be analysed against through the GC-MS procedure.

Analysis of the [U-14C]glucose-fed samples

We initially fractionated the [U-14C]glucose-fed samples as they serve as a control experiment for the [U-13C]glucose-fed samples. This is particularly important as it is not possible to directly quantify the 13C abundance in molecules outside the dynamic range of the GC-MS and is therefore impossible quantify the total amount of label taken up and indirect calculations of the depletion of substrate from the media are complicated by the high salt concentration in the incubation media. Although 14C glucose feeding experiments have been carried out previously on both of the genotypes under study (Fernie et al., 2002a; Trethewey et al., 1999), the experimental conditions used previously were somewhat different as lower concentrations of glucose and shorter incubation periods were used. Furthermore in order to perform a direct comparison we decided to do all three feeding experiments in parallel. In this experiment, the invertase line, INV-2-30, displayed a slight reduction in glucose uptake and a dramatic reduction in label metabolized in comparison with the wild type, whereas the sucrose phosphorylase line, SP-29, was characterized by a similar rate of glucose uptake but a moderate reduction in label metabolized (in both cases these findings are similar to those reported previously). Furthermore the patterns of radiolabel redistribution were also similar to those reported previously with notable changes in the levels of label recovered in sucrose and starch with minor changes also seen in that recovered in carbon dioxide, organic and amino acids. The amount of label recovered in cell wall is considerably higher than that observed previously, however, it is conceivable that this is because of the length of the incubation. Determination of the specific activity of the hexose phosphate pools revealed, as expected, that the transformants had proportionally lower label in this metabolic pool. On calculation of the metabolic fluxes it emerged that the line INV-2-30 had a slightly elevated rate of cell wall synthesis, however, more significantly both line INV-2-30 and line SP-29 contained dramatic increases in the rate of sucrose (re)synthesis and glycolysis with respect to wild type. Thus, despite the elevated rate of cell wall synthesis the results of this study are highly consistent with those presented previously (Fernie et al., 2002a; Trethewey et al., 1999). In the approach taken to estimate the metabolic fluxes presented in Table 1 it is assumed here as in many previous studies that the metabolic pools have achieved steady state. If this assumption is not made then the average specific activity of the hexose phosphate pool needs to be taken into account. Such a method of calculation has previously been performed in short-term feeding experiments by averaging the specific activity of the hexose phosphate pool over the time course of the experiment simply by dividing by a factor of 2 (for a detailed list of the assumptions taken see Fernie et al., 2002a). Therefore, if such estimates were calculated for the data presented here the absolute fluxes would be twofold higher than those presented in Table 1.

Table 1.  Metabolism of 14C-glucose of potato tuber discs from wild type and transgenic (enhanced sucrose mobilization) lines
  1. Freshly cut slices of growing potato tubers of wild type and transformants were incubated for 5 h in the presence of 10 mm Mes-KOH (pH 6.5) and 20 mm [U-14C]glucose before they were washed and extracted to determine label distribution. The specific activity of the hexose phosphate pool was estimated by dividing the label retained in the phosphate ester pool by the summed carbon of the hexose phosphates. The determined specific activity was then used to calculate absolute fluxes to starch, sucrose, cell wall and glycolysis. The results are mean ± SE (n = 3). Values that were determined by the t-test to be significantly different from the wild type are set in bold type (P < 0.05).

Uptake (1000 dpm)16.8 ± 0.813.3 ± 0.915.9 ± 0.5
Metabolized (1000 dpm)15.6 ± 0.76.1 ± 0.313.5 ± 0.7
Redistribution of 14C (% metabolized)
 Carbon dioxide4.5 ± 0.63.0 ± 0.42.5 ± 0.4
 Organic acids6.8 ± 0.55.7 ± 0.65.4 ± 0.7
 Amino acids8.3 ± 1.011.7 ± 0.9 7.9 ± 0.5
 Hexose phosphates5.2 ± 0.35.8 ± 0.4 4.9 ± 0.4
 Sucrose8.5 ± 1.231.0 ± 2.534.6 ± 5.2
 Fructose1.5 ± 0.48.4 ± 0.2 1.5 ± 0.1
 Starch52.8 ± 2.326.9 ± 8.934.5 ± 4.7
 Cell wall11.1 ± 1.5 8.0 ± 2.68.0 ± 0.6
 Protein1.2 ± 0.1 0.9 ± 0.1 1.0 ± 0.1
Specific activity (dpm nmol−1 g FW−1)
 Hexose P pool16.2 ± 1.23.4 ± 0.97.8 ± 1.4
Metabolic flux (nmol hexose equivalents nmol−1 g FW−1 h−1)
 Cell wall synthesis117 ± 15165 ± 18135 ± 17
 Sucrose (re)synthesis85 ± 11639 ± 39580 ± 49
 Glycolysis208 ± 32439 ± 64275 ± 18
 Starch synthesis528 ± 86554 ± 79580 ± 129

Analysis of the [U-13C]glucose-fed samples

Having established the rate of glucose uptake and the bulk fluxes using conventional 14C labelling approaches we next evaluated the samples incubated in 13C glucose (or unlabelled glucose) following the approach detailed above. In addition we ran calibration curves alongside these samples in order to allow the calculation of absolute pool sizes. Following evaluation of the labelled, unlabelled and authentic standard chromatograms we were able to establish the total amount of label present in the 31 metabolite pools presented in Table 2. For wild type tuber discs very high absolute label accumulation can be observed in glucose, the hexose phosphates – glucose 6-phosphate and fructose 6-phosphate, sucrose, citrate, malate, glutamate and asparagines, whilst considerable label also accumulated in alanine, threonine, tryptophan and tyrosine. As was seen above for the [U-14C]glucose-fed samples the label distribution of 13C varied considerably across the genotypes. The invertase expressing line was characterized by elevated label incorporation in alanine, asparagine, glutamine, glutamate, glycine, phenylalanine, serine, threonine and tryptophan in addition to dramatic elevations in label incorporation in fructose, fructose 6-phosphate and sucrose. In addition, they displayed reduced incorporation in β-alanine, isoleucine, leucine, lysine, proline, tyrosine and valine as well in fumarate and succinate. However, the most dramatic reductions are seen in the sugars and sugar phosphates with approximately 90% less label recovered in glucose and glucose 6-phosphate in this line than that found in wild type with reductions also observed in the label incorporation into inositol and mannitol. Intriguingly, in this line only label was also detected in maltose. There were fewer changes observed in label accumulation in the sucrose phosphorylase expressing lines. Moreover the pattern of change was different. That said there were significant increases in label incorporation to alanine, β-alanine, glutamate, glycine, serine, quinate mannitol and sucrose and dramatic increases in the label incorporation in fructose, fructose 6-phosphate and malate. In contrast decreased label incorporation, in isoleucine, leucine, lysine, methionine, phenylalanine, threonine and valine and inositol, and dramatically decreased label incorporation in glucose and glucose 6-phosphate was observed. It is important to note here that changes in glucose content may merely reflect different uptake and metabolism rates of the labelled substrate.

Table 2.  Total isotope accumulating in metabolites of tuber tissue from wild type and transgenic plants following incubation of isolated tuber discs in [U-13C]glucose for 5 h
Metabolite (nmol g FW−1)WTINV-2-30SP-29
  1. Discs cut directly from developing tubers were incubated in the presence of 10 mm Mes-KOH (pH 6.5) and 20 mm [U-13C]glucose for 5 h before they were washed and extracted to determine label distribution. Mean mole fractional enrichment was then determined after GC-MS analysis as described in Experimental procedures. Wherever possible the molecular ion was used for this calculation, in instances where this was not possible it was assumed that the labeling pattern of the mass fragment used reflected that of the complete molecule. The total isotope present was calculated by multiplying the total metabolite content by its mean mole fractional enrichment. The results are mean ± SE (n = 6). Values that were determined by the t-test to be significantly different from the wild type (P < 0.05).

Amino acids
 Alanine35.0 ± 3.665.2 ± 7.866.0 ± 6.7 
 Asparagine75.8 ± 4.266.3 ± 5.679.0 ± 5.9
 Aspartate15.6 ± 1.631.9 ± 1.610.9 ± 1.1
 β-Alanine1.21 ± 0.030.30 ± 0.008.23 ± 0.11
 Glutamine39.9 ± 2.063.5 ± 3.843.3 ± 0.4
 Glutamate15.6 ± 0.929.7 ± 2.729.0 ± 2.1 
 Glycine0.83 ± 0.024.3 ± 0.24.4 ± 0.1 
 Isoleucine17.0 ± 1.2811.5 ± 0.29.4 ± 0.7 
 Leucine10.4 ± 0.90.9 ± 0.13.7 ± 0.4 
 Lysine10.1 ± 0.41.9 ± 0.12.2 ± 0.1 
 Methionine15.3 ± 1.218.3 ± 0.313.2 ± 0.1 
 Phenylalanine10.8 ± 1.215.3 ± 1.75.6 ± 0.1 
 Proline1.43 ± 0.041.00 ± 0.020.84 ± 0.13
 Serine9.6 ± 0.413.0 ± 1.121.0 ± 0.9 
 Threonine31.5 ± 2.7121.9 ± 1.824.3 ± 0.5 
 Tryptophan32.5 ± 10.1131.7 ± 7.942.5 ± 3.1
 Tyrosine25.0 ± 5.011.3 ± 2.519.5 ± 0.8
 Valine13.9 ± 0.610.1 ± 0.510.1 ± 0.3 
Organic acids
 Citrate133 ± 7144.3 ± 8.0134 ± 7
 Fumarate1.92 ± 0.050.42 ± 0.020.40 ± 0.01
 Quinate6.90 ± 0.45 7.1 ± 0.28.6 ± 0.1 
 Malate282 ± 16456 ± 6701 ± 17
 Succinate18.9 ± 0.93.2 ± 0.123.3 ± 0.8 
Sugars, sugar alcohols and sugar phosphates
 Fructose3.4 ± 0.412.8 ± 0.454.4 ± 3.7 
 Fructose 6-phosphate52.9 ± 2.3231.0 ± 6.6170.8 ± 3.6 
 Glucose993 ± 101144.0 ± 7.2106.0 ± 10.5
 Glucose 6-phosphate282 ± 1226.4 ± 1.0101 ± 3.5
 Inositol0.83 ± 0.020.02 ± 0.010.21 ± 0.03
 Maltose0.00 ± 0.000.11 ± 0.060.00 ± 0.00
 Mannitol0.31 ± 0.060.10 ± 0.011.00 ± 0.16
 Sucrose41.9 ± 2.1113 ± 1083.1 ± 0.11

Determination of fluxes from the [U-13C]glucose-fed dataset

Although several notable changes are apparent in the data above they merely represent the distribution of label following the incubation period and because of factors such as differential uptake of substrate, differences in metabolite pool size or the consequent differential dilution of label following the mobilization of the tissues own storage carbohydrates may not reflect the true metabolite exchange rates. For this reason we next analysed the rates of metabolite accumulation during the 5 h incubation. This was achieved, by dividing the corrected (i.e. the proportion of label that is the result of natural abundance was removed from the figures presented in Table 2) label accumulation in a given metabolite by the proportional labelling of a precursor molecule, in a manner analogous to that presented above for the calculation of fluxes from 14C-fed samples (Table 3). However, given that the proportional labelling of the various pools varied considerably over time (see Figure 2), differential equation models were introduced to estimate unidirectional exchange rates based on the time-dependent labelling of various pools. Following the accumulation of label in various pools, we are able to estimate unidirectional exchange rates for all reactions presented in Figure 1(b) (a full description of the equations used is given in Appendix S1). For molecules that we defined as end products we estimated the unidirectional exchange rates by calculating the mean proportional labelling of the precursor molecule rather than using that found at the end of the experiment. It should be noted that these estimates represent only the net synthetic rate to the metabolites in question as the use of uniformly labelled substrate and the lack of atomic information in our measurements precludes us from estimate the rates of cyclic or transfer fluxes. It is, therefore, likely that in many cases the actual flux to any of the end products defined here is higher than that presented. Conversely, it is conceivable that some of the unidirectional rates may be higher than the actual flux. This possibility is perhaps exemplified by the metabolism of glutamate in line SP-29, the kinetics of labelling of this reaction series is characterized by a massive influx to glutamate from α-ketoglutarate but only minor effluxes to glutamine and proline. Whilst the obvious and quite trival explanation for this observation is that glutamine is accumulating during the experiment and thus an exact mass balance could not be expected we cannot exclude the fact that the massive efflux to glutamate is the result of a rapid exchange flux catalysed by the reversible transaminase that connects α-ketoglutarate and glutamine.

Table 3.  Estimated unidirectional carbon exchange rates in isolated tuber discs of wild type and transgenic plants
Synthetic flux (precursors) (nmol hexose equivalents g FW−1 h−1)WTINV-2-30SP-29
  1. Unidirectional exchange rates are calculated over the 5 h incubation, by dividing the total isotope accumulation (data taken from Table 1 and corrected to take into account the natural abundance of C13) in metabolites by the calculated average specific labelling of the metabolites precursor over the time course of the experiment (data from Figure 3 and not shown). The average specific activity of the hexose phosphate pool was determined by averaging the specific activities of both glucose 6-phosphate and fructose 6-phosphate over time. The results are mean ± SE (n = 6) and are given as nmol hexose equivalents g FW−1 h−1. Values that were determined by the t-test to be significantly different from the wild type (P < 0.05) are set in bold type. Glc 6P, glucose 6-phosphate; Fru, fructose; Fru 6-P, fructose 6-phosphate; 3-PGA, 3-phosphoglycerate; Phe, phenylalanine; Suc, sucrose; Trp, tryptophan.

Hexose phosphates (Glc)8056 ± 8018791 ± 4382315 ± 23
3PGA (hexose phosphates)3588 ± 1513722 ± 1064187 ± 101
Suc (Fru 6P)88.3 ± 1.01772 ± 57812 ± 11
Fru (Suc)34.9 ± 4.147.1 ± 4.2165.7 ± 11.2
Alanine (3PGA)125.2 ± 12.8354.1 ± 42.3358.5 ± 35.0
β-Alanine (alanine)0.59 ± 0.012.24 ± 0.001.38 ± 0.02
Valine (3PGA)24.8 ± 4.891.2 ± 4.8 22.1 ± 0.6
Leucine (3PGA)35.9 ± 3.00.26 ± 0.033.2 ± 0.3
Serine (3PGA)24.3 ± 3.033.3 ± 0.3 28.8 ± 0.1
Glycine (serine)1.57 ± 0.0438.5 ± 1.7972.5 ± 1.6
Glycerate (3PGA)4.5 ± 0.72.5 ± 0.1 2.6 ± 0.9
Citrate (3PGA)4011 ± 1474634 ± 6004356 ± 483
Glutamate (2-oxogluterate)304.6 ± 15.31552 ± 32703 ± 6.4
Glutamine (glutamate)334.5 ± 19.1178.8 ± 16.21.1 ± 0.2
Proline (glutamate)8.3 ± 2.35.6 ± 0.11.1 ± 0.2
Aspartate (malate)876 ± 50751 ± 9418 ± 10
Asparagine (aspartate)1445 ± 80658 ± 567952 ± 593
Lysine (aspartate)54.6 ± 0.16.53 ± 0.3472.0 ± 3.2
Methionine (aspartate)10.5 ±  0.99.45 ± 0.1527.8 ± 0.2
Isoleucine (aspartate)43.4 ± 3.334.3 ± 0.6 263 ± 19
Threonine (aspartate)9.8 ± 0.4n.d.51 ± 1
Maltose (Glc 6P)n.d.0.33 ± 0.18n.d.
Mannitol (Glc 6P)0.025 ± 0.005n.d.0.26 ± 0.04
Inositol (Glc 6P)0.068 ± 0.000.13 ± 0.050.24 ± 0.03
Phe (Glc 6P)44.5 ± 3.7101.7 ± 11.20.57 ± 0.01
Tyrosine (Glc 6P)63.5 ± 1.042.0 ± 9.3 17 ± 0.7
Trp (Glc 6P)85.1 ± 0.1789.4 ± 47.352.7 ± 3.8

For non-end product metabolites the estimation of unidirectional rates of synthesis was considerably more complex because of the fact that they require the modelling of multiple reactions as the metabolites in question represent intermediates in metabolism. However, these unidirectional rates of synthesis can also be estimated on the basis of label accumulation – albeit using more complex equations (the full description of these is presented in Appendix S1).

In the wild type the dominant reactions were those of glucose phosphorylation, sucrose resynthesis, aspartate, glutamate and glutamine and alanine production as well as the provision of carbon into the Krebs cycle in the form of citrate and the net conversion of hexose phosphates into 3PGA. With the exception of β-alanine, proline, cysteine, and glycine considerable rates of synthesis were observed for the accumulation of all amino acids and indeed for all other measured compounds (excepting glycerate, mannitol and inositol). As would be expected from the differences in redistribution of label there were also large differences in estimated unidirectional carbon transfer between the genotypes with some of the above-mentioned unidirectional carbon transfers being considerable in the transformants. Perhaps unsurprisingly given the nature of the proteins expressed by the transgenic lines these changes were most dramatic in sugar metabolism. As was observed following flux estimations based on the 14C-feedings, the estimated rate of sucrose re(synthesis) was highly elevated in the transformants, furthermore the absolute values quantitatively approximate those obtained from the 14C study validating the novel approach we present here. As would be expected given the high rate of sucrose turnover the rate of fructose production is also markedly enhanced in the transformants. Perhaps surprisingly however is the fact that the rate of glucose phosphorylation is unaltered in the invertase line and dramatically reduced in the sucrose phosphorylase expressing line. In the case of the invertase line these results may be explainable by the fact that glucose accumulates to very high levels (Roessner et al., 2001a; Trethewey et al., 1998), and that potato hexokinase is a relatively inefficient enzyme (Trethewey et al., 2001; Veramendi et al., 2002), whereas the expression of sucrose phosphorylase effectively bypasses the glucose pool, a fact that may also explain the elevated flux between the hexose phosphate and 3PGA pools in this transgenic line. In addition there are significant increases in the unidirectional rate of alanine synthesis (and also that of β-alanine), which previous experiments suggest is most likely caused by an induction of fermentation (Bologa et al., 2003) and a dramatic increase in glycine (in both lines) and mannitol (in the invertase expressing line) synthesis. A common response was also observed in the unidirectional rates of carbon exchange to leucine, methionine and aspartate which decreased and serine and glutamate which increased with respect to wild type in both transformants and the unidirectional rate of carbon exchange between 3PGA and citrate which was unaltered in the transgenics. However, many of the alterations in flux pattern are not conserved across the transgenics. Amongst the most prominent of these are that the unidirectional rate of carbon exchange between glucose 6-phosphate and tryptophan, glutamine or phenylalanine are higher in the invertase line but lower in the sucrose phosphorylase line with respect to wild type (whilst the opposite holds true for asparagine, lysine and isoleucine). Another interesting finding of this feeding experiment was that the accumulation of label in maltose was already apparent at 5 h incubation in the invertase expressing line. The biosynthesis of maltose within heterotrophic tissues, such as the tuber, still remains an open question: as it could be formed on the turnover of starch or by the condensation of two glucose molecules. Further experiments, using the approach that we describe here, may allow us to evaluate which of these two alternative pathways is the major route of maltose synthesis in planta.

We chose the two transgenic lines here because they have been comprehensively analysed previously (Fernie et al., 2002a; Trethewey et al., 1998, 1999). The isotopic determinations described above are generally in very good accordance with those that would be predicted from metabolic profiling data we previously acquired for these lines (Roessner et al., 2001a), however, this was only on a qualitative basis. These determinations also provide a general framework for the analysis of the reactions of a wide range of metabolites including those whose fluxes cannot be evaluated by traditional kinetic models (e.g. many of the amino acids) because of the paucity of enzymology data in the pathways in question. It must be noted however that in comparison with steady state modelling of metabolism such as that carried out by NMR (Dieuaide-Noubhani et al., 1995; Schwender et al., 2003) the approach we describe here currently does not have the capacity to determine exchange fluxes and therefore depends on the assumption that the unidirectional rates of reaction provide a valid approximation of metabolism. One advantage of the GC-MS based method described here, however, is that it could easily be adapted to simultaneously allow the determination of both steady state and kinetic metabolite profiles. Whilst we have only demonstrated that this approach is applicable in potato tubers we believe it could be readily transferred to other plant systems. However, caution will be required in adapting the set-up as the labelling kinetics will most probably by considerably different in different biological material. For this reason establishment of kinetic profiling in other experimental systems would almost certainly require empirical adjustment. Moreover, the calculations presented here are calculated on the assumptions that rates are constant during the time course of the experiment and that the majority of metabolites in the metabolic system represent end products. Although this is not strictly the case previous work in our lab has shown that the flux from amino acids to proteins is very low within the potato tuber (Fernie et al., 2002a; Roessner-Tunali et al., 2003a). Given the chemical breadth of commercially available stably labelled isotopes, this method could be modified to profile anabolic or catabolic fluxes of any small carbon or nitrogen containing metabolites. At present GC-MS-based methods afford far high-throughput, because of the much lower spectral acquisition time, than NMR and offer greater sensitivity. This may be even more important in an organ of higher metabolic complexity than the potato, for example, a leaf in which far more metabolites are detectable (see Fernie, 2003), as it would be anticipated that far more fluxes would also be determinable. For larger molecules such as secondary metabolites, however, LC-MS-based approaches probably hold the most promise for the future. That said for comprehensive flux analysis several analytic techniques will ultimately be required to allow the combination of sensitivity and positional information necessary for the generation of network wide flux maps.


The GC-MS method presented here represents a fast and accurate way to compare kinetic aspects of intracellular metabolism. As a proof of concept we applied it to two transgenic potato lines known to display dramatic perturbations in primary metabolism in their tubers. In the course of this study several interesting observations in both the wild type and transgenic potato tuber systems were made including (i) circumstantial evidence for starch turnover within discs isolated from potato tubers, (ii) direct evidence of the increased carbon partitioning towards several amino acids because of the low sucrose content in the transgenics and (iii) patterns of the relative unidirectional reaction rates through the major pathways of primary metabolism in the wild type and transformants. Moreover, through this test case we were able to demonstrate that this approach gave similar values for the rate of sucrose cycling to those we have obtained previously using more conventional 14C labelling approaches. Furthermore, we also demonstrate how a large amount of metabolic information can be determined following incubation in a single isotopically enriched precursor. Although this method still requires considerable computational time it is far less laborious than simple chromatographic separation and radiolabel counting procedures and as such offers an important complement to genomic strategies such as metabolite profiling. In addition, this method could be readily adapted to focus in detail on a certain branch of metabolism, say, for example, methionine biosynthesis and catabolism, by utilizing the wide range of commercially available stably labelled substrates

Experimental procedures


[U-13C6]Glc was purchased from Omicron Biochemicals Inc. (South Bend, IN, USA), [U-14C6]Glc from Amersham International (Braunschweig, Germany), Mes (2-N-morpholino ethane sulphonic acid) buffer, unlabelled glucose and all other chemicals and enzymes used were purchased either from Sigma-Aldrich Chemical Company (Deisenhofen, Germany), or from Merck KGaG (Damstadt, Germany).

Plant material

Potato (Solanum tuberosum cv. Desiree) was obtained from Saatzucht Lange AG (Bad Schwartau, Germany). The generation and selection of the transgenic lines used here have been described previously (Sonnewald et al., 1997; Trethewey et al., 2001). Plants were maintained in tissue culture with a 16-h-light/8-h-dark regime on Murashige and Skoog (1962) medium that contained 2% sucrose, in the greenhouse plants were grown in parallel under the same light regime with a minimum of 250 μmol photons m−2 sec−1 at 22°C. In this article the term developing tubers is used for tubers (>10 g fresh weight) harvested from healthy 10-week-old plants.

Tuber disc labelling experiments

Tuber discs (diameter 8 mm, thickness 1–2 mm) were cut directly from growing potato tubers attached to the fully photosynthesizing mother plant, washed three times with 10 mm MES-KOH (pH 6.5) and then incubated (eight discs in a volume of 4 ml in a 100 ml Erlenmeyer flask shaken at 90 r.p.m.) for various time intervals, indicated in the text, in 10 mm MES-KOH (pH 6.5) containing either (i) 20 mm glucose including 0.14 kBq μmol−1 [U-14C6]Glc (ii) 20 mm [U-13C6]Glc or (iii) 20 mm unlabelled glucose. After the indicated time period the discs were harvested washed three times in buffer (100 ml per eight discs), and then frozen in liquid nitrogen.

Analysis of 14C labelled samples

Tuber discs were extracted with 80% (v/v) ethanol at 80°C (1 ml per two discs), re-extracted in two subsequent steps with 50% (v/v) ethanol (1 ml per two discs at each step), the combined supernatants, dried under an air stream at 40°C, taken up in 1 ml water, and labelled components were separated exactly as described in Fernie et al. (2001a). The insoluble material left after ethanol extraction was analysed for label in starch as in Fernie et al. (2001b). Specific activities were estimated by dividing the label retained in the phospho-ester fraction by the summed carbon in this fraction as detailed in Geigenberger et al. (1997).

Analysis of [U-13C6]Glc labelled samples

Tuber discs were extracted in 100% methanol at 70°C for 15 min (Roessner et al., 2000). After centrifugation, the resulting supernatent was dried under vacuum, and the resulting residue was derivatized for 120 min at 37°C (in 50 μl of 20 mg ml−1 methoxyamine hydrochloride in pyridine) followed by a 30 min treatment at 37°C with 50 μl of MSTFA. GC-MS analysis of the derivatized samples was carried out as described by Roessner et al. (2001a,b). Uncorrected molar percentage enrichments of metabolites were evaluated as described in Giege et al. (2003) by comparison with the 12C spectral fragments and the isotopic spectral fractions of non-labelled control incubations with the fragmentation patterns of metabolites detected in the chromatograms of the 13C-fed tuber discs after modification to the procedure developed for the evaluation of NMR spectra (Des Rosiers et al., 1991). The modification made was that we did not correct the size of the labelled peaks with respect to that naturally abundant but rather used the natural abundance peak to calibrate that present in the labelled samples. Absolute concentrations were defined by the comparison of peak areas in samples to those obtained on evaluation of known quantities of authentic standards as described by Roessner-Tunali et al. (2003b). For the calculation of the total label present in a metabolite pool the mole fractional enrichment of that metabolite was multiplied by the absolute concentration of that metabolite.

Estimation of unidirectional reaction rates from 13C data

The reaction rates from metabolic precursors to end products were estimated by dividing the amount of label accumulating in the end-product pool during the experiment by the calculated average proportional labelling of the precursor pool. Whilst the reaction rates to metabolic intermediates, where possible, was estimated by the solution of mass balancing differential equations (differential flux balancing equations). Full details of the mathematical framework behind these calculations can be found in Appendix S1.

Statistical analysis

Where differences are described in the text as significant, a t-test was performed using the algorithm incorporated into Microsoft Excel 7.0 (Microsoft Corp., Seattle, WA, USA) that yielded a value below 5% (P < 0.05).

Supplementary material

The following material is available from

Figure S1. Flux calculation based on experimental data.