Radioisotope ratios discriminate between competing pathways of cell wall polysaccharide and RNA biosynthesis in living plant cells

Authors


(fax +44 0131 650 6592; e-mail: s.fry@ed.ac.uk).

Summary

Cell wall polysaccharides are synthesized from sugar-nucleotides, e.g. uridine 5′-diphosphoglucose (UDP-Glc), but the metabolic pathways that produce sugar-nucleotides in plants remain controversial. To help distinguish between potentially ‘competing’ pathways, we have developed a novel dual-radiolabelling strategy that generates a remarkably wide range of 3H:14C ratios among the various proposed precursors. Arabidopsis cell cultures were fed traces of d-[1-3H]galactose and a 14C-labelled hexose (e.g. d-[U-14C]fructose) in the presence of an approximately 104-fold excess of non-radioactive carbon source. Six interconvertible ‘core intermediates’, galactose 1-phosphate ↔ UDP-galactose ↔ UDP-glucose ↔ glucose 1-phosphate ↔ glucose 6-phosphate ↔ fructose 6-phosphate, showed a large decrease in 3H:14C ratio along this pathway from left to right. The isotope ratio of a polysaccharide-bound sugar residue indicates from which of the six core intermediates its sugar-nucleotide donor substrate stemmed. Polymer-bound galacturonate, xylose, arabinose and apiose residues (all produced via UDP-glucuronate) stemmed from UDP-glucose, not glucose 6-phosphate; therefore, UDP-glucuronate arose predominantly by the action of UDP-glucose dehydrogenase rather than through the postulated competing pathway leading from glucose 6-phosphate via myo-inositol. The data also indicate that UDP-galacturonate was not formed by a hypothetical UDP-galactose dehydrogenase. Polymer-bound mannose and fucose residues stemmed from fructose 6-phosphate, not glucose 1-phosphate; therefore GDP-mannose (guanosine 5′-diphosphomannose) arose predominantly by a pathway involving phosphomannose isomerase (via mannose phosphates) rather than through a postulated competing pathway involving GDP-glucose epimerization. Curiously, the ribose residues of RNA did not stem directly from hexose 6-phosphates, but predominantly from UDP-glucose; an alternative to the textbook pentose-phosphate pathway therefore predominates in plants.

Introduction

Plant cell wall polysaccharides are the world’s most abundant organic resource, and a plant’s principal anabolic activities often centre on their biosynthesis. They are manufactured by membrane-bound glycosyltransferases using cytosolic sugar-nucleotides such as UDP-glucose and GDP-mannose (Carpita and McCann, 2000; Feingold and Barber, 1990; Fry, 2000, 2004; Seifert, 2004; Somerville et al., 2004), where necessary transported into (Handford et al., 2004) and further metabolized within (Burget et al., 2003) the Golgi lumen. Surprisingly, in view of the abundance of cell wall polysaccharides in the biosphere, the metabolic pathways involved are still controversial: in several instances, competing pathways potentially lead to the same sugar-nucleotide (Figure 1) and it is unclear which route normally predominates in living plant cells. For example, there are two pathways potentially leading to UDP-glucuronate (Asamizu and Nishi, 1979; Blaschek et al., 1981; Fry, 1985;Kärkönen, 2005; Kärkönen et al., 2005; Loewus, 2006; Loewus and Loewus, 1983; Seitz et al., 2000; Verma and Dougall, 1979): it could be produced either from UDP-glucose by the action of a dehydrogenase (UDPGDH) or from glucose 6-phosphate via the inositol pathway involving myo-inositol 1-phosphate synthase and myo-inositol oxygenase (MYOX) (Fry, 1985; Kärkönen, 2005; Loewus, 2006; Loewus and Loewus, 1983) (Figure 1). Likewise, GDP-mannose could be manufactured either by epimerization of GDP-glucose (Carpita and McCann, 2000) or from fructose 6-phosphate via a pathway involving phosphomannose isomerase (PMI) (Feingold and Barber, 1990) (Figure 1). Distinguishing which of a pair of competing pathways predominates in vivo has been a long-standing challenge. Four popular approaches to this type of question all have specific limitations: (i) Patterns of gene expression (e.g. of UDPGDH and MYOX), visualized as mRNA accumulation, can suggest when and where the ‘competing’ enzymes may be synthesized (Kanter et al., 2005; Seitz et al., 2000). However, this approach ignores the fact that, owing to post-transcriptional regulation (Lea et al., 2006; Mishra and Handa, 1998; Pastori et al., 2000; Smith et al., 2004) and differences in longevity between enzymes and their mRNAs, measured transcript concentrations cannot be relied on to predict enzyme concentrations. (ii) Enzymic assays or immunological screening can indicate when and where the relevant enzyme proteins are present (Seitz et al., 2000). However, an enzyme’s mere presence, or its activity (assayed in vitro after extraction from the cells), does not guarantee that it will exert any action in vivo. In living cells, enzymes may be compartmentally separated from their substrates, they may lack the necessary activators or endogenous inhibitors may be present (Fry, 2004). (iii) Studies of mutants may fail to distinguish between competing pathways because plants often circumvent genetic blocks by shunting substrates through ‘bypass’ pathways (Dennis and Blakeley, 2000). For example, mutation of MYOX in Arabidopsis did not affect polysaccharide biosynthesis (Kanter et al., 2005), superficially suggesting that the inositol pathway does not normally contribute much to the production of UDP-glucuronate. Nevertheless, it remained possible that the inositol pathway does normally predominate but is superseded by the UDPGDH pathway when a myox mutation necessitates this. Clearer conclusions can be drawn only when mutations do diminish polysaccharide biosynthesis, although such effects have tended to be relatively small (Burget et al., 2003; Kärkönen et al., 2005). (iv) Finally, in-vivo feeding of pathway-specific radiolabelled intermediates may not reliably identify the predominating pathway. For example, tracer quantities of exogenous myo-[3H]inositol can radiolabel polymer-bound galacturonate, arabinose and xylose residues (Figure 1) (Asamizu and Nishi, 1979; Seitz et al., 2000; Verma and Dougall, 1979), and high doses of non-radioactive myo-inositol can inhibit incorporation of 14C from [14C]glucose into these three residues (Asamizu and Nishi, 1979; Blaschek et al., 1981; Roberts and Loewus, 1973). However, although direct precursors can often be utilized when artificially supplied, they may not normally be present in vivo in sufficient quantities to satisfy the proposed pathway; conversely, they could be the major natural precursors but fail to demonstrate this convincingly when supplied exogenously because of sequestration or restricted uptake. Incorporation of exogenous inositol establishes only that the inositol pathway can operate (perhaps opportunistically, scavenging any occasionally surplus inositol), not that it normally does operate in the cell type studied. Plant cells possess several scavenger pathways that may not normally play any major role in vivo (Fry, 2000).

Figure 1.

 Competing pathways of sugar-nucleotide production, and an experimental strategy for distinguishing them.
Cells are grown in glycerol medium; black arrows represent bulk metabolic flux from glycerol. The six interconvertible compounds in the grey box are cytosolic ‘core metabolites’, as defined. In a typical experiment, we supplied traces of d-[1-3H]galactose plus d-[U-14C]fructose, thus forming a 3H:14C gradient along the core pathway. 3H and 14C (red and blue) infiltrate the metabolic landscape from left and right, respectively. Sugar-nucleotides, synthesized from core metabolites, are used (dashed arrows) to build sugar residues into wall polysaccharides (brick pattern).

As an alternative to the four approaches described above, we propose a novel dual-radiolabelling strategy differing fundamentally in that the exogenous precursors are not directly involved in the pathways of interest. Instead, precursors are selected (e.g. [3H]galactose and [14C]fructose; Figure 1) such that the two radioisotopes infiltrate the metabolic landscape from opposite ends of a reversible, cytosolic core pathway. We define a ‘core’ metabolite as one, such as UDP-glucose, that acquires most of its 3H from one immediate precursor (UDP-galactose in this case) and most of its 14C from a different one (glucose 1-phosphate). On the other hand, UDP-rhamnose is not a core metabolite as defined because it obtains all its 3H and all its 14C from the same precursor (UDP-glucose).

Each cytosolic core metabolite is expected to acquire a unique 3H:14C ratio, depending on its position along the core pathway. Products stemming from a given core metabolite will inherit the same 3H:14C ratio as that core metabolite, so long as downstream reactions do not cause loss of either of the two radioisotopes. (To compensate for the 14C inevitably lost during [U-14C]hexose → [U-14C]pentose conversion, we have where appropriate multiplied the 3H:14C ratios of all pentoses by 5/6.)

This ‘core pathway’ dual-radiolabelling approach enables discrimination between postulated competing pathways without the need for mutagenesis, unsubstantiated predictions of mRNA translation and enzyme longevity, or the assumption that enzymes present in the cell necessarily exert any action in vivo. Our approach also avoids the feeding of alleged direct intermediates such as myo-[3H]inositol.

Results

Radiolabelling of core pathway intermediates

The following experiment demonstrates that a robust gradient of 3H:14C ratios was indeed created, despite all the steps of the core pathway (Figure 1) being reversible. Suspension-cultured Arabidopsis cells were fed traces of d-[1-3H]galactose and d-[U-14C]fructose plus an approximately 104-fold excess of non-radioactive glycerol. We chose glycerol rather than glucose or sucrose as the bulk carbon source so that the uptake of tracer levels of radiolabelled hexoses would not be suppressed by the presence of an excess of non-radioactive hexoses. The cultures grew well with glycerol as the sole carbon source. The cells consumed 90% of the [3H]galactose and [14C]fructose within approximately 0.5 and 4 h, respectively. By 8 h, 42% and 36% of the 3H and 14C, respectively, had been released as 3H2O and 14CO2. Most 14CO2 loss in these aerobic cells would be due to pyruvate metabolism via the Krebs cycle. Plant cell cultures typically convert 40–60% of their carbon source into cell dry mass (Henshaw et al., 1966), the remainder being lost as CO2. Thus our 36% value indicates that [14C]fructose (Fru)-6-P, formed from exogenous [14C]fructose, readily underwent glycolysis (notwithstanding the apparent ‘futility’ of the Fru-6-P → Fru-1,6-P2 step in glycerol-fed cells; see Figure 1), and thus that the 14C efficiently infiltrated the central pathways of cellular metabolism.

Each intermediary core metabolite was isolated by high-voltage electrophoresis, then chemically and/or enzymically dissected to yield degradation products which were purified by paper chromatography (e.g. Figure 2b,c) and assayed for both radioisotopes. Metabolites furthest removed from [3H]galactose took longest to reach their respective 3H:14C maxima (Figure 3), the time required ranging from <2 min (galactose (Gal)-1-P) to approximately 25 min (Fru-6-P). Also, the maximum 3H:14C ratio achieved was inversely related to the distance of the metabolite from [3H]galactose (Figure 3); the observed maximum ratios ranged from 300 Bq Bq−1 (Gal-1-P) to 6 Bq Bq−1 (Fru-6-P). The subsequent decreases in 3H:14C ratio observed after the maxima had been reached were due to continued uptake of [14C]fructose after depletion of the exogenous [3H]galactose.

Figure 2.

 Representative autoradiograms of metabolites and polymer-bound sugar residues from Arabidopsis cells fed d-[1-3H]galactose plus d-[U-14C]fructose.
(a) Two-dimensional paper chromatography of monosaccharides released by acid hydrolysis of polymers.
(b) Electrophoresis (pH 2.0) of metabolites extracted after 4 and 40 min radiolabelling. Zones A–C were eluted for further analysis.
(c) Paper chromatography of mild acid hydrolysis products of zone A.

Figure 3.

 Short-term changes in isotope ratios in intermediary metabolites and sucrose extracted from Arabidopsis cells at intervals after addition of d-[1-3H]galactose plus d-[U-14C]fructose.
Each error bar is based on the 3H and 14C counting errors involved in the assay of a single sample. The colour code for ‘core metabolites’ corresponds to that in Figure 1; non-core metabolites are in black.

We conclude that the basic rationale of the dual-labelling approach was successful, creating a gradient of 3H:14C ratios as represented diagrammatically by the red:blue colour gradient in Figure 1.

Radiolabelling of polysaccharides, sucrose and phosphatidylinositol

The isotopic labelling of an end-product will report the integral of the labelling of its intermediary precursors with respect to time. Therefore, changes in the 3H:14C ratios of an end-product should lag behind those of its precursor core metabolite. The final absolute 3H:14C ratios of end-products are thus not expected to equal those of their precursors. Nevertheless, a group of end-products all derived from the same core metabolite should be found to end up with 3H:14C ratios closely similar to each other.

Labelling kinetics (data not shown) indicate that, within the time-frame of this experiment, sucrose more closely resembled an end-product than an intermediary metabolite (it increased in both 3H and 14C throughout the period of observation). In accordance with this, the isotope ratios in the glucose and fructose residues of sucrose lagged behind those of their core precursors (UDP-glucose and Fru-6-P, respectively; Figure 3), and were clearly different from each other.

In several independent experiments we assayed the radiolabelling of end-product polymers, after leaving sufficient time for most of the radioactivity to have transited intermediary metabolism (Figure 4a–c). Three end-products with known biosynthetic origins serve as bench-marks: polysaccharide-bound galactose and rhamnose and phospholipid-bound inositol will define the isotopic ratios expected of products stemming from UDP-galactose, UDP-glucose and glucose (Glc)-6-P, respectively (Figure 1) (Loewus and Loewus, 1983; Seifert, 2004; Usadel et al., 2004). Note that in Figure 4a (for which 14C was supplied as d-[U-14C]glucose), Fru-6-P is not a ‘core metabolite’ as defined as it derives all its 3H and all its 14C from a common precursor, Glc-6-P. In contrast, in Figure 4(b) and (c) (for which 14C was supplied as d-[U-14C]fructose), Fru-6-P is a ‘core metabolite’ as it derives its 3H from Glc-6-P and its 14C from free fructose.

Figure 4.

 Isotope ratios in sugar residues of polymers from Arabidopsis cells after long-term feeding of d-[1-3H]galactose plus a 14C-labelled hexose.
The 14C-hexose was (a) d-[U-14C]glucose, (b,c) d-[U-14C]fructose, or (d) d-[1-14C]galactose. The data in parts (b) are from two time-points (2 and 8 h) of the experiment shown in Figure 3. ‘±SE’ refers to the SE of ratios determined for several replicate cultures (= 3, 7 and 8 in a, c and d respectively); ‘± counting error’ indicates the scintillation-counter error for single samples. Residue abbreviations are as in Figure 1; Cell = cellulose-derived glucose; Ino = phosphatidylinositol-derived inositol. The proposed core metabolites from which these residues principally arose are indicated by the colour gradient used in Figure 1.

Galactose residues consistently had the highest 3H:14C ratios of all polymer-bound sugar residues (Figure 4a–c). Thus the gradient of ratios transiently created along the core pathway (Figure 3) was faithfully recorded in end-product polysaccharides. The very high ratios for polymer-bound galactose residues argue against a speculative model (Figure 2 of Seifert, 2004) in which cytosolic UDP-galactose must first be converted by a cytosolic epimerase into UDP-glucose before the latter is re-converted to UDP-galactose by a Golgi-bound epimerase (UGE4) (Barber et al., 2006) which channels its product to galactosyltransferases. If that mechanism routinely occurred, polymer-bound galactose residues would acquire the same isotope ratio as UDP-glucose derivatives.

The myo-inositol moiety of phosphatidylinositol had a 3H:14C ratio of 2.34 ± 0.22 (Figure 4c). There is reportedly no isotope effect (e.g. discrimination against 3H) during conversion of [1-3H]Glc-6-P to myo-inositol (Loewus and Loewus, 1983). A similar ratio would therefore be expected for any other polymer-bound residues that also stemmed from Glc-6-P. Polymer-bound glucose, rhamnose, galacturonate, xylose and arabinose residues had ratios much higher than 2.34 in the same experiment, indicating that the inositol pathway was not predominant for polysaccharide biosynthesis in these cells.

Rhamnose residues, formed from UDP-rhamnose, stem from the core metabolite UDP-glucose (Feingold and Barber, 1990; Usadel et al., 2004). Four UDP-glucuronate-derived residues (galacturonate, xylose, arabinose and apiose) had isotope ratios equal to that of rhamnose (Figure 4a and c), indicating that UDP-glucuronate also originated mainly from UDP-glucose by the action of UDPGDH. The data also argue against the theoretical possibility (Carpita and McCann, 2000) that UDP-galacturonate is formed by the action of a UDP-galactose dehydrogenase.

Non-cellulosic glucose residues had relatively variable 3H:14C ratios (Figure 4), probably reflecting their presence in both hemicellulose and starch, formed from cytosolic UDP-glucose and plastidial ADP-glucose pools, respectively. Cellulosic glucose residues, in contrast, acquired a reproducible isotope ratio similar to that of rhamnose residues, suggesting that most cellulosic glucose was contributed by cytosolic UDP-glucose.

The GDP-mannose and GDP-fucose might stem from either Glc-1-P (via GDP-glucose and its 2-epimerization to GDP-mannose) (Carpita and McCann, 2000) or Fru-6-P (via the mannose phosphates) (Feingold and Barber, 1990). The 3H:14C ratios of polymer-bound mannose and fucose residues were consistently much lower than those of any other sugar residues (Figure 4), indicating biosynthesis predominantly from the core metabolite Fru-6-P, not Glc-1-P.

Lack of 3H loss during acid hydrolysis

It could be suggested that an unknown and variable proportion of the 3H might be lost from dual-radiolabelled monosaccharides during acid hydrolysis of the polysaccharides. To test for this we purified several dual-labelled sugars that had been released by the standard 1-h hydrolysis, then subjected them to a prolonged second round of acid ‘hydrolysis’ lasting for up to 4 h, re-purified them chromatographically, and re-assayed them for 3H and 14C (Table 1). Some degradation of monosaccharides occurred (especially of galacturonate and ribose), as expected during prolonged heating in acid. However, the 3H:14C ratio of the remaining intact sugar molecules, re-isolated chromatographically, was unchanged. These data validate our routine hydrolysis protocol [2 m trifluoroacetic acid (TFA) at 120°C for 1 h].

Table 1.   Effect of prolonged heating in acid on the recovery and isotope ratios of dual-radiolabelled monosaccharides
Time in hot acid (h)GalAAraXylGalManFucRhaGlcRib
Reca (%)3H:14C ratiobRec (%)3H:14C ratioRec (%)3H:14C ratioRec (%)3H:14C ratioRec (%)3H:14C ratioRec (%)3H:14C ratioRec (%)3H:14C ratioRec (%)3H:14C ratioRec (%)3H:14C ratio
  1. GalA, galacturonic acid; Ara, arabinose; Xyl, xylose; Gal, galactose; Man, mannose; Fuc, fucose; Rha, rhamnose; Glc, glucose; Rib, ribose.

  2. aRec, recovery of the 14C-labelled monosaccharide after the acid treatment and chromatography, relative to the control. To ensure high purity, we selected only the spot centres; and as this selection is subjective the percentage recoveries are variable.

  3. bThe 3H:14C ratio indicates the isotope ratio on a Bq:Bq basis; the errors are counting errors determined for single samples. The whole experiment was repeated with similar results (data not shown).

  4. cThe controls received no acid; the 0-h samples were placed in 2 m trifluoroacetic acid (TFA) at room temperature but not heated; the other two samples were heated in 2 m TFA at 120°C for 2 or 4 h.

Controlc10014.4 ± 0.810018.3 ± 0.610016.6 ± 0.610026.1 ± 0.51005.1 ± 1.31002.0 ± 0.510016.2 ± 0.910029.8 ± 0.910014.1 ± 1.0
011215.2 ± 0.89917.8 ± 0.68516.6 ± 0.68926.8 ± 0.6895.5 ± 1.6642.0 ± 0.79816.2 ± 0.98730.5 ± 1.010213.2 ± 1.0
25514.8 ± 1.29418.3 ± 0.66717.6 ± 0.79026.0 ± 0.61095.9 ± 1.41041.9 ± 0.58617.4 ± 1.09229.6 ± 0.97013.9 ± 1.2
42713.6 ± 1.88418.5 ± 0.65517.5 ± 0.87422.0 ± 0.51115.1 ± 1.31032.0 ± 0.57816.8 ± 1.19230.4 ± 0.94914.1 ± 1.6

Also, in preliminary experiments (data not shown), monosaccharides released from cell walls by enzymic digestion with Driselase (Sigma; http://www.sigmaaldrich.com) had 3H:14C ratios very similar to those of monosaccharides released by 1 h of acid hydrolysis. This rules out the possibility that complete 3H loss occurred from certain specific carbon atom(s) within the first hour of acid hydrolysis.

A sample of [3H,14C]inositol (isolated enzymically from radiolabelled phospholipids; see above) was also tested in a similar way, but with more severe acid ‘hydrolysis’. Its initial isotope ratio (3H:14C = 3.15) was practically unchanged even after very severe acid treatments (e.g. 72 h at 110°C in 6 m HCl). Thus, severe acid hydrolysis is an acceptable alternative to the use of phospholipase for the release of inositol from phosphatidylinositol.

Metabolic stability of the 3H atom attached to carbon-1

It is clear which metabolic steps will cause 14C to be lost as 14CO2 (Figure 1). However, it was also important to consider where, if at all, 3H is lost as [3H]water. As one approach to testing this, we fed cultured Arabidopsis cells with dual-labelled l-[1-3H, 1-14C]arabinose and then assayed the 3H:14C ratios of polysaccharide-bound arabinose and xylose residues. These residues had 3H:14C ratios closely similar to that of the exogenous [3H,14C]arabinose (measured 3H:14C ratios were: exogenous arabinose, 7.68 ± 0.01; polymer-bound xylose residues, 7.69 ± 0.05; polymer-bound arabinose residues, 7.66 ± 0.04; these ± values indicate scintillation-counter errors for single samples), showing that the H atom remained firmly attached to carbon-1 during all steps involved in polymer radiolabelling: (i) arabinose uptake, (ii) phosphorylation to arabinose-1-P, (iii) uridylylation to UDP-arabinose, (iv) epimerization to UDP-xylose, (v) arabinopyranose → arabinofuranose ring contraction (Konishi et al., 2007) and (vi) the various polysaccharide synthase-catalysed transglycosylations leading to the biosynthesis of diverse pentosans.

A second approach by which to test for 3H loss was to feed the cells with dual-labelled d-[1-3H, 1-14C]galactose. In the absence of 3H loss, the isotope ratio would be constant throughout the core pathway. In fact, polymer-bound galactose, glucose, rhamnose, galacturonate, xylose and arabinose residues all acquired isotopic ratios similar to that of the exogenously fed [3H,14C]galactose (Figure 4d), indicating negligible 3H loss from C-1 during the galactose → UDP-glucose steps, the subsequent interconversions to UDP-rhamnose, UDP-glucuronic acid (GlcA) etc., and the transglycosylation reactions involved in polysaccharide biosynthesis. The slightly lower ratio observed for polymer-bound galactose residues themselves may be due to the presence in the sample of a small proportion of l-galactose, derived by conversion of GDP-d-mannose to GDP-l-galactose (Baydoun and Fry, 1988; Conklin et al., 2006) in addition to the major enantiomer, d-galactose. The phospholipid-bound inositol residues (Figure 4d) also had a 3H:14C ratio fairly close to that of the exogenous d-galactose, indicating little 3H loss during the UDP-glucose → Glc-1-P → Glc-6-P steps. However, the Fru-6-P-derived residues (mannose and fucose) had approximately 10-fold lower ratios, indicating that a step between glucose-6-P and GDP-mannose does involve 3H loss. This step is probably the loss of the pro-chiral 3H atom from C-1 of Fru-6-P during PMI action (Seeholzer, 1993). The fact that any3H turned up in mannose and fucose residues could be due to (i) a small percentage of GDP-mannose being produced by the epimerase pathway from GDP-glucose and/or (ii) some degree of 3H redistribution e.g. by glycolysis of Fru-6-P to triose-phosphates, their inter-conversion, and the subsequent re-formation of Fru-6-P by glyconeogenesis.

This loss of 3H at a specific point in the metabolic pathways is not problematical for our experimental approach. In fact, as all steps in the core pathway (Figure 1) are reversible, 3H loss at the Fru-6-P → mannose (Man)-6-P step may helpfully steepen the 3H:14C gradient along the rest of the core pathway, although the high isotope ratio observed for inositol indicates that in practice this effect did not extend far back along the core pathway (Figure 4d).

Radiolabelling of polymer-bound ribose

Ribose (Rib)-containing polymers include RNA, poly(ADP-ribose), and possibly a cytosolic polysaccharide (Yang and Steup, 1990). The best-known route to ribose (in micro-organisms and animals, but not critically evaluated in plants) is the oxidative pentose phosphate pathway (PPP), converting Glc-6-P to Rib-5-P (Dennis and Blakeley, 2000). Starting from [1-3H]Glc-6-P, this would yield Rib-5-P completely lacking 3H because carbon atom 1 would be lost as CO2. Some [3H]Rib-5-P could arise via the non-oxidative PPP, from [3H]Fru-6-P and [3H]glyceraldehyde 3-phosphate (Dennis and Blakeley, 2000). Either way, the 3H:14C ratio of Rib-5-P would not exceed that of Fru-6-P. However, the isotopic ratio of polymer-bound ribose in Arabidopsis cells was almost as high as those of the UDP-glucose-derived residues (Figure 4). Therefore, most of the ribose was not formed via either of the two classic PPPs.

We propose an alternative pathway that draws on UDP-glucose as its core metabolite: UDP-xylose probably generates xylose (Xyl)-1-P (Fry and Northcote, 1983), or free d-xylose (Rosenfield and Loewus, 1978), which then proceed via a 3-epimerisation to yield the majority of the d-Rib-5-P required for RNA synthesis in these Arabidopsis cells. The slight discrepancy between the 3H:14C ratios of ribose and those of the UDP-glucose-derived wall polysaccharide residues (rhamnose, galacturonate, xylose, arabinose, apiose and cellulosic glucose) could be due to the production of a minority of the Rib-5-P by the ‘textbook’ PPPs.

Discussion

This work demonstrates the usefulness of ‘core pathway’ dual radiolabelling, using exogenous radiochemicals not closely related to the pathways of interest, for discriminating between competing pathways claimed to operate in vivo. Only minute traces of radiochemicals are added, which would not be capable of perturbing metabolism. Neither genetic mutation nor the addition of proposed intermediates is necessary; no unsubstantiated assumptions have to be made concerning the translation of mRNAs; and it is not necessary to assume that enzyme activity (measured in vitro after extraction) equals enzyme action (operating in vivo).

This simple general strategy will be applicable to future studies of diverse cell types and additional pathways. It is capable of complementing the genomic approach, for example by critically testing whether or not a transgenic or mutant line exhibits a predicted metabolic alteration.

Our experimental approach identifies predominant pathways actually operating in living plant cells. It does not rule out a minor contribution by the alternative (competing) pathway.

Different conclusions will undoubtedly be reached in comparable studies with other cell types. However, the data for the Arabidopsis cells studied here discount any major contribution to polysaccharide biosynthesis of proposed pathways involving

Glc-6-P →myo-inositol 1-phosphate →myo-inositol → GlcA → GlcA-1-P → UDP-GlcA;

GDP-Glc → GDP-Man,

or

UDP-Gal → UDP-GalA

where GalA is galacturonic acid. Instead, the principal routes to the major biomass components are concluded to be:

UDP-Glc → UDP-GlcA,

Fru-6-P → Man-6-P → Man-1-P → GDP-Man,

and

UDP-GlcA → UDP-GalA.

In addition, the data reveal the predominance of a hitherto unsuspected major plant pathway for ribose production, probably stemming from UDP-glucose via UDP-pentoses. This conclusion rules out a major contribution from the two ‘textbook’ cytosolic PPPs – the oxidative PPP (starting with Glc-6-P dehydrogenase and proceeding via 6-phosphoglucono-δ-lactone, 6-phosphogluconate, ribulose 5-phosphate and ribose-5-P) and the non-oxidative PPP (starting with transketolase and transaldolase, and utilizing Fru-6-P and glyceraldehyde 3-phosphate).

In conclusion, a novel in-vivo labelling strategy using two radioisotopes (3H and 14C), which enabled us to distinguish between pairs of alleged ‘competing’ metabolic pathways for the production of the sugar-nucleotide precursors of major plant polymers, showed that UDP-glucose oxidation predominates over myo-inositol oxidation, and PMI action predominates over GDP-glucose 2-epimerase. In addition, the data demonstrate the potential of the methodology to provide useful information relating to other pathways, not related to cell wall biosynthesis: specifically, the results show that the ribose of RNA does not originate from the classic PPPs in plants.

Experimental procedures

Source of radiochemicals

d-[U-14C]Glucose (11.5 MBq μmol−1) and d-[U-14C]fructose (10.5 MBq μmol−1) were catalogue items from Amersham (http://www5.amershambiosciences.com/) and d-[1-14C]galactose (2.04 MBq μmol−1) was a catalogue item from Sigma Chemical Co. (http://www.sigmaaldrich.com/). l-[1-14C]Arabinose (2.0 MBq μmol−1), a catalogue item from American Radiolabeled Chemicals Inc. (http://www.arcincusa.com/arc/), was found to be heavily contaminated by [14C]arabinitol (>50%) and was therefore purified by preparative paper chromatography before use. l-[1-3H]Arabinose (approximately 148 MBq μmol−1) and d-[1-3H]galactose (approximately 180 MBq μmol−1) were custom-synthesized by Amersham.

Cell cultures

Arabidopsis thaliana cell suspension cultures were maintained as described (Oswald et al., 2001) and then transferred into a medium containing 2% (w/v) glycerol instead of glucose. Aliquots (0.4 ml) of 3-day-old glycerol-grown culture (10% settled cell volume) were fed a mixture of d-[1-3H]galactose (typically 0.5 MBq, at 180 MBq μmol−1) and 0.1 MBq of either d-[U-14C]fructose (10.5 MBq μmol−1), d-[U-14C]glucose (11.5 MBq μmol−1) or d-[1-14C]galactose (2.04 MBq μmol−1). At intervals, aliquots of culture were killed by the addition of 3.8 volumes of freshly prepared ethanol/formic acid (7:1; all solvent compositions are v/v unless otherwise stated) and shaken for 16 h at 25°C to extract low-Mr metabolites.

As a control experiment, we fed similar Arabidopsis cells with traces of l-[1-3H, 1-14C]arabinose (3H:14C ratio = 7.68); the washing with ethanol/formic acid was similar to that described above.

Radiolabelled intermediary metabolites

Metabolites thus extracted were applied to Whatman 3mm paper (http://www.whatman.com/) and fractionated by high-voltage electrophoresis (Fry, 2000) in pH 2.0 buffer, at 1.5 kV, for 1 h, to give three broad zones (Figure 2b): A, comprising hexose bisphosphates, triose monophosphates and UDP-sugars; B, principally hexose monophosphates; C, non-phosphorylated compounds. Orange G was included as a visible marker. Zone C contained neutral sugars, which move a few centimetres from the origin owing to electro-endo-osmosis, and also most cellular carboxylic acids (e.g. malic and citric acids), whose carboxy groups are almost un-ionized at pH 2; electrophoretic mobility at pH 2 is almost entirely conferred by the presence of phosphate groups. Metabolites present in zones A, B and C were separately eluted from electrophoretograms similar to that shown in Figure 2(b).

Eluted zone A was mildly acid-hydrolysed (0.1 m TFA, 100°C, 25 min), releasing the non-ribose monosaccharides from UDP-sugars but not from hexose bisphosphates (Dawson et al., 1986). The formerly UDP-bound glucose and galactose thus obtained were isolated by paper chromatography sequentially in butan-1-ol/acetic acid/water (BAW) 12:3:5 and ethyl acetate/pyridine/water (EPW) 8:2:1, both for 17 h in the same dimension (e.g. Figure 2c), then eluted and further purified in additional solvents (not shown) (Fry, 2000).

Zone B was eluted and divided into two aliquots: B(i) was digested with phosphatase (wheatgerm, ‘type I’; Sigma Chemical Co.) to remove all esterified phosphate groups; B(ii) was mildly acid-hydrolysed (0.1 m TFA, 100°C, 25 min). The mild acid treatment cleaves glycosyl–phosphate bonds and therefore releases neutral hexoses from Glc-1-P and Gal-1-P but not from Fru-1-P, Fru-6-P or Glc-6-P (Dawson et al., 1986; Fry, 2000). Free hexoses from both aliquots were then purified chromatographically as above. Glucose from B(ii) represents Glc-1-P, whereas glucose from B(i) minus glucose from B(ii) represents Glc-6-P. There was negligible difference between B(i) and B(ii) in galactose yield, so all galactose phosphate was taken to be Gal-1-P. B(ii) yielded no fructose, indicating the absence of the glycosyl–phosphate, Fru-2-P. Fructose from B(i) was taken to represent Fru-6-P because Fru-1-P is not a significant plant metabolite (Geigenberger et al., 2004).

Sucrose was purified from zone C by electrophoresis (Fry, 2000) at pH 3.5 (to remove carboxylic acids, which form anions at that pH) followed by paper chromatography (in propan-1-ol/acetic acid/water 15:3:2 and EPW 10:4:3), then treated with invertase (yeast; BDH [VWR; http://www.vwr.com]); the glucose and fructose thus generated were purified chromatographically in EPW 10:4:3.

Radiolabelled polymers and phosphatidylinositol

The ethanol/formic acid-insoluble material (polymer fraction) was washed 20 times in 14 ml of 80% ethanol for the removal of low-Mr metabolites and unincorporated hexoses, then the alcohol-insoluble residue (polymer-rich fraction) was dried and acid-hydrolysed (2 m TFA, 120°C, 1 h). Monosaccharides thus solubilized were purified by two-dimensional paper chromatography: the first dimension was in BAW 12:3:5 for 27 h, and the second was in EPW 8:2:1 for 30 h (e.g. Figure 2a). To ensure high purity, we eluted only the spot centres after autoradiography, and then further purified each compound by repeated chromatography, typically in the same solvents for longer times and also in phenol/water 4:1 (w/w), or in the case of galacturonate by electrophoresis (Fry, 2000) at pH 3.5.

In some experiments, the TFA-resistant residue (containing cellulose as the only polysaccharide) was digested with 0.5% (w/v) Driselase in pyridine/acetic acid/water (1:1:98, pH ≈ 4.5, containing 0.5% (w/v) chlorobutanol) at 37°C for 72 h; the released ‘cellulosic’ [3H,14C]glucose was purified chromatographically.

In other experiments, the cultures were killed by the addition of 6.7 volumes of chloroform/methanol (1:1) instead of ethanol/formic acid, and shaken for 16 h. After phase partitioning against water, the organic phase was subjected to TLC on silica gel in chloroform/methanol/H2O (65:25:4) and autoradiographed. The zone of silica gel bearing the [3H,14C]phosphatidylinositol band (localized by reference to an external marker; Sigma Chemical Co.) was scraped off and treated with either HCl or enzymes. For the experiment shown in Figure 4(c), this silica gel was suspended in 6 m HCl and heated at 110°C for 48 h. For the experiment shown in Figure 4(d), it was suspended in 1 mg ml−1Streptomyces chromofuscusphospholipase-D (Sigma Chemical Co., P8023) in 1% pyridine containing 0.5% (w/v) chlorobutanol at 25°C for 14 days. In both cases, the released myo-inositol was then dried, partially purified by paper chromatography in BAW 12:3:5, eluted and further purified by re-chromatography in acetone/water 17:3 (a particularly effective solvent, in which inositol has a much lower mobility than all common sugars).

Assay of radioisotopes

Radiolabelled metabolites, located on electrophoretograms and chromatograms by autoradiography, were eluted in water; 1 ml of the eluate was mixed with 10 ml Wallac OptiPhase scintillant (PerkinElmer; http://www.perkinelmer.com) and assayed in a Beckman LS6500 scintillation counter (http://www.beckmancoulter.com/) running 3H and 14C channels simultaneously. Non-radioactive background readings were subtracted for each channel. Then the apparent 3H reading was corrected for ‘false 3H’ (≈ 45% of the background-corrected 14C count; this percentage was determined accurately for each new batch of OptiPhase by assay of pure [14C]fructose). Pure [3H]galactose did not give ‘false 14C’; therefore the 14C values were not further adjusted. Quenching was corrected for but was very low and highly consistent between samples. To compensate for the 14C inevitably lost during [U-14C]hexose →[U-14C]pentose conversion, we have where appropriate multiplied the 3H:14C ratios of pentoses by 5/6. This was of course not performed when [1-14C]galactose was fed (since here the 14C atom is expected to be retained during the decarboxylation step) or when [1-14C]arabinose (which is already a pentose) was fed.

Control experiments to validate analytical methods

A potential artefact would be loss of 3H from certain sugars during acid hydrolysis. To test this, we treated individual purified (3H,14C)-labelled compounds (arabinose, xylose, galacturonate, rhamnose, glucose, ribose and inositol) with 2 m TFA at 120°C for 4 h and then chromatographed the samples (in BAW 12:3:5 and EPW 8:2:1). Some degradation of the aldoses occurred as expected during prolonged heating in acid, but the 3H:14C ratio of the remaining intact sugar, re-isolated chromatographically, was unchanged. These results validate our routine hydrolysis protocol (2 m TFA at 120°C for 1 h).

Acknowledgements

We thank Mrs Janice Miller and Mr Ben Mewburn for technical help; Mr Thomas Waibel and the Edinburgh University Plant Physiology 3 class of 2005 for preliminary data; and the Biotechnology and Biological Sciences Research Council for funding.

Ancillary