Biochemical Journal

Research article

Alternative pathways of dehydroascorbic acid degradation in vitro and in plant cell cultures: novel insights into vitamin C catabolism

Harriet T. Parsons, Tayyaba Yasmin, Stephen C. Fry


L-Ascorbate catabolism involves reversible oxidation to DHA (dehydroascorbic acid), then irreversible oxidation or hydrolysis. The precursor–product relationships and the identity of several major DHA breakdown products remained unclear. In the presence of added H2O2, DHA underwent little hydrolysis to DKG (2,3-dioxo-L-gulonate). Instead, it yielded OxT (oxalyl L-threonate), cOxT (cyclic oxalyl L-threonate) and free oxalate (~6:1:1), essentially simultaneously, suggesting that all three product classes independently arose from one reactive intermediate, proposed to be cyclic-2,3-O-oxalyl-L-threonolactone. Only with plant apoplastic esterases present were the esters significant precursors of free oxalate. Without added H2O2, DHA was slowly hydrolysed to DKG. Downstream of DKG was a singly ionized dicarboxy compound (suggested to be 2-carboxy-L-xylonolactone plus 2-carboxy-L-lyxonolactone), which reversibly de-lactonized to a dianionic carboxypentonate. Formation of these lactones and acid was minimized by the presence of residual unreacted ascorbate. In vivo, the putative 2-carboxy-L-pentonolactones were relatively stable. We propose that DHA is a branch-point in ascorbate catabolism, being either oxidized to oxalate and its esters or hydrolysed to DKG and downstream carboxypentonates. The oxidation/hydrolysis ratio is governed by reactive oxygen species status. In vivo, oxalyl esters are enzymatically hydrolysed, but the carboxypentonates are stable. The biological roles of these ascorbate metabolites invite future exploration.

  • ascorbic acid
  • carboxy-L-xylonic acid
  • dioxogulonic acid
  • hydrogen peroxide (H2O2)
  • oxalyl esterase
  • oxalyl threonate


L-Ascorbate fulfils multiple essential roles in both plants and animals. It has a well-established intraprotoplasmic role as an antioxidant, e.g. in chloroplasts [1,2]. In the apoplast (aqueous solution permeating the cell wall) of plants, it is the principal redox buffer, where it regulates responses to the extracellular environment [36]. It is implicated in the defence of plants against pathogens [7] and in the control of plant growth and development [8,9], including flowering, senescence [10] and root development [11]. Besides serving as an anti-oxidant [12,13], it may also act as a pro-oxidant that generates hydroxyl radicals in the apoplast [14,15] and thereby causes the non-enzymatic scission of neighbouring polysaccharides [14,16,17]. Such scission may contribute to the cell-wall loosening required for fruit softening [18,19] and cell expansion [20,21]. Interesting biological roles in the apoplast for ascorbate's primary oxidation product, DHA (dehydroascorbic acid), have also been proposed [22]. Therefore ascorbate and DHA metabolism in plants is highly significant; however, many questions concerning their catabolic pathways and the identity of major degradation products remain open.

There is generally accepted to be one predominant pathway by which ascorbate is synthesized in Arabidopsis [2325]. Intraprotoplasmically, it is recycled (reversibly oxidized to DHA) by the ascorbate–glutathione cycle [26]. DHA is best represented as a bicyclic structure (Figure 1, top left) [27]. It can be further broken down in vivo via reactions that are widely accepted to be irreversible under physiological conditions [28], thus escaping the ascorbate–glutathione cycle.

Figure 1 Proposed products formed from ascorbate under strongly oxidising conditions by H2O2

Proposed pathways of ascorbate catabolism. Starting material and major end-products, dark grey background; known oxidation steps, mid-grey background; newly proposed intermediates, pale grey background. Ascorbate (bottom left) is rapidly oxidized to DHA, which is shown in its conventional dioxo form (centre left) and as the hemiketal monohydrate that predominates in aqueous solution (top left). The C2 fragment derived from carbons 1 and 2 of ascorbate is indicated by a heavy CC bond; this fragment was radioactive in our experiments. The non-radioactive C4 fragment derived from carbons 3–6 of ascorbate is indicated by dark grey bonds. Hypothetical intermediates are in square brackets. Thick and thin solid arrows show non-enzymatic reactions proposed to be fast and slow respectively (relative to the preceding step); dotted arrows show reactions catalysed by apoplastic enzymes; dashed arrows show the linear pathway proposed in [35]. No attempt is made to represent stereochemistry except in the two Fischer projection formulae (ascorbate and ‘dioxo’ DHA).

Several pathways exist for irreversible ascorbate or DHA catabolism in plants, some being taxonomically restricted. Plants in the Vitaceae, e.g. grapes, accumulate L-threarate (tartrate) as an end-product of enzymatic ascorbate catabolism via L-idonate [29,30]. Grapes also accumulate OxA (oxalate), which is synthesized from ascorbate [31]. Ascorbate also acts as a precursor to threarate in Pelargonium crispum, but here ThrO (L-threonate), not L-idonate, acts as an intermediate and OxA is a by-product [30,32].

ThrO plus OxA are well-established in vitro end-products of ascorbate oxidation [33,34]. Green and Fry [35] described a catabolic pathway by which OxA and ThrO are produced from DHA by oxidative cleavage of the C-2–C-3 bond followed by several further steps. This pathway was proposed to proceed via the novel metabolites cOxT [cyclic OxT (oxalyl L-threonate)] and OxT. The core of the pathway (dashed arrows in Figure 1) was proposed to be linear: DHA-(oxidation)→2,3-cOxT-(acyl migration)→3,4-cOxT-(hydrolysis)→4-OxT-(hydrolysis)→4-OxA+ThrO [35]. Isbell and Frush [34] had hypothesized that 2-OxT was an intermediate, but had not detected it experimentally. The proposed pathway represents a means by which OxA and ThrO can be produced from ascorbate and which is likely to be common to all plant species, since all the steps can occur non-enzymatically in vitro. Although these reactions can occur nonenzymatically, some of the steps were catalysed by enzymes [35], especially the oxidation of ascorbate to DHA and the hydrolysis of OxT to OxA+ThrO, but probably also the oxidation of DHA and hydrolysis of cOxT to OxT.

Several important uncertainties remain concerning the sequence of reactions during DHA oxidation and the identity of some of the intermediary metabolites. For example, we had assumed that all OxA and ThrO was produced via the linear pathway DHA→cOxT→OxT→OxA+ThrO. An alternative to this assumption is the possibility that some of the OxA is formed from DKG (2,3-dioxo-L-gulonate), the hydrolysis product of DHA [36]. We have therefore now re-tested the possible intermediary role of DKG. Another alternative is that OxT, or even free OxA+ThrO, is formed from DHA directly, without the intermediacy of a cyclic ester; we have therefore now also re-tested this by more detailed time-course studies.

Green and Fry [35] proposed that DKG yielded two mutually interconvertible, but unidentified, C6 products: C and E. Simpson and Ortwerth [37] describe oxidative and non-oxidative in vitro breakdown pathways of DKG yielding OxA in both pathways and ThrO during oxidative breakdown, but their results did not suggest possible identities for C or E. We have therefore further explored the conditions under which C and/or E are produced, and investigated for the first time the fate of C/E in vivo.

In the present study we monitored the oxidative and/or hydrolytic breakdown of ascorbate, DHA and DKG in aerated aqueous solutions under physiological conditions of temperature and apoplastic pH, and with three severities of oxidation:

  • ‘strongly’ oxidizing conditions, with added H2O2;

  • ‘moderately’ oxidizing conditions, with (at least initially) a supply of an agent capable of making endogenous H2O2 from dissolved O2 such as ascorbate (ascorbic acid+O2→DHA+H2O2) or DHA (which can also slowly yield sufficient H2O2 to support Fenton reactions [14]); and

  • ‘weakly’ oxidizing conditions (e.g. DKG+O2) with no obvious means of forming reactive oxygen species.

Products were analysed by HVPE (high-voltage paper electrophoresis) and HPLC. HVPE has proved particularly valuable in kinetic analyses of low-Mr, unstable, radiolabelled organic acids. We have now further explored both principal forks of DHA degradation, under various conditions mimicking the plant apoplast in various respects. We compared strongly, moderately and weakly oxidizing conditions, at various physiologically relevant pH values, and in the presence or absence of culture medium, enzymes and cells, and with low or high ascorbate concentrations. We did not attempt to model the fate of plastidic or cytosolic DHA, which may be rapidly salvaged back to ascorbate in the reducing enviromnent of the protoplasm compared with the apoplast. Previously unidentified intermediates were isolated, characterized further, and tested for their own susceptibility to oxidation and hydrolysis. In this way, in vitro branched pathways of DHA degradation have been compiled which can be used as comparative models for the in vivo fate of these compounds. In addition, during attempts to scale up the preparation of ascorbate metabolites, we noticed that some of them were only poorly generated at higher substrate concentrations; we have therefore also explored the effect of ascorbate concentration on the pattern of the oxidation products generated.



Ascorbic acid, DHA, H2O2 and catalase were from Sigma Chemicals. Solid L-[1-14C]ascorbic acid (0.52 or 0.407 MBq/μmol) from GE Healthcare was dissolved in water, aliquoted and stored at −80°C until required.

Oxidation of ascorbate under strongly oxidizing conditions

All incubations were performed at 20°C in open tubes (thus with dissolved O2). For ascorbate oxidation experiments under strongly oxidising conditions, the reaction mixture (500 μl) contained 10–20 mM ascorbic acid (including 16 kBq of L-[1-14C]ascorbate), 80 mM pyridinium acetate buffer, pH 6.0 (i.e. 80 mM pyridine, pH adjusted to 6.0 with acetic acid), and 2 mol eq of H2O2 (added last). With all reaction mixtures containing exogenous H2O2, 20 μl aliquots were taken at intervals, treated with 1–4 units of catalase for ~10 s, frozen in liquid N2 and subjected to HVPE.

Oxidation of ascorbate under moderately oxidizing conditions

For ascorbate degradation experiments under moderately oxidizing conditions in the absence of deliberately added metal ions, the reaction mixture contained carrier-free L-[1-14C]ascorbic acid (25 kBq/ml; 48 μM) in water (pH≈4.4), and samples were frozen at the time points indicated, then analysed by HVPE. Reaction mixtures were aerated by gentle shaking, but no exogenous H2O2 was added. For corresponding experiments with cultures and/or media, carrier-free L-[1-14C]ascorbic acid was used at 148 kBq/ml (360 μM) and the medium was buffered with 5 mM Mes (Na+, pH 6.0).

Fate of compounds C and E in vivo and in vitro

14C-Labelled compound C was freshly eluted from an electrophoretogram (similar to that shown in Figure 5b) and then tracer levels of it were incubated with medium and/or cultured plant cells.

Alkaline hydrolysis of compound C

Additional tracer-level aliquots of 14C-labelled compound C were alkali-hydrolysed in 0.1 M NaOH at 20°C for 2 h and then adjusted to pH 4.7 with acetic acid. Controls were treated with the same amounts of acetic acid and NaOH but in reverse order so that the sample was never at an alkaline pH. Radioactive products were analysed by HVPE.


Samples were dried on Whatman No. 1 paper and electrophoresed in a volatile buffer at pH 6.5 (acetic acid/pyridine/water, 1:33:300, by vol., containing 5 mM EDTA), usually at 3.0 kV for 40 min [38]. The papers were cooled to 20–25°C with toluene during the run. OG (Orange G) was used as an internal reference marker and/or was loaded between samples. Non-radioactive compounds were stained with AgNO3 [39]. 14C-Labelled compounds were detected by autoradiography and quantified by scintillation counting. Samples required for further analysis were isolated by elution from electrophoretograms in water.

Rosa cell-suspension cultures

Rosa sp. (‘Paul's Scarlet’ rose) cell-suspension cultures were maintained as described previously [35] and subcultured fortnightly. Transition metal ions in the fresh medium (pH 6.0) were 20 μM Fe3+, 4.5 μM Mn2+, 0.08 μM Cu2+ and 0.04 μM Co2+. These are likely to be higher than the free concentrations normally present in the apoplast; however, for the oxidation/hydrolysis experiments described in the present study, 4- to 5-day-old cultures were used, in which the majority of the transition metal cations would have been taken up by the cells and/or bound to wall polymers. Cultures were passed through muslin and only the small cell aggregates (harvested in the filtrate) were used. Spent medium was obtained by filtration through a fine nylon mesh or a glass-wool plug that was impenetrable to cells; when indicated, enzymes in the filtrate were denatured at 100°C for 15 min.

Detection of radioactivity

Radioactive compounds on paper electrophoretograms were detected by autoradiography on Kodak BioMax MR-1 film. In some cases, the radioactive areas of the paper were cut out and 14C was quantified by scintillation counting in ‘OptiScint HighSafe’.


Pathway initiation: oxidation of ascorbate to DHA

Both branches of apoplastic ascorbate degradation begin with its oxidation to DHA. Under strongly oxidizing conditions at pH 6.0, 2 mol eq of H2O2 rapidly oxidized 10 mM [1-14C]ascorbate, initially to neutral DHA (Figure 2a), the ascorbate being consumed with a half-life of ~50 min (Figure 2b). Over the 0–30 min period of observation, with average concentrations of ~6 mM ascorbate and ~15 mM H2O2 present, we estimate the rate constant for the ascorbate+H2O2 reaction to be 1.6×10−2 M−1·s−1.

Figure 2 Kinetic study of the in vitro oxidation of [14C]ascorbate under strongly oxidising conditions by H2O2

Ascorbic acid (10 mM, including 32 kBq/ml L-[1-14C]ascorbate) was incubated at pH 6.0 with 2 mol eq H2O2. After the indicated times, 5 μl aliquots were subjected to HVPE at pH 6.5. (a) The autoradiogram is pictured overlying the electrophoretogram (unstained, but the internal and external marker OG spots are faintly visible; dotted circles). The experiment was carried out four times with minor variations, giving qualitatively similar results; a representative electrophoretogram is shown. (b) Radioactive zones of the electrophoretogram shown in (a) were excised and the radioactivity was quantified by scintillation counting. The radioactivity of each compound is expressed as a percentage of the total recovered at that time point. Except for ascorbate and DHA, the small amount of 14C present at zero time (dotted box in a) has been subtracted from the percentage. Streaks migrating between [14C]ascorbate and [14C]DHA, which represent immobile DHA formed from mobile ascorbate during HVPE, have been included as [14C]ascorbate. (c) Magnification of grey box in (b). A value of 10% on the y axis of (b) and (c) equates to a concentration of 1 mM of the metabolite. (d) Relative yields of the various OxA products.

This experiment was conducted four times with qualitatively similar results; a representative electrophoretogtam is shown in Figure 2(a), and this specific electrophoretogram is quantified in Figures 2(b)–2(d). In contrast, under weakly oxidizing conditions (with O2 but in the absence of added H2O2 and transition metal ions), 10 mM ascorbate had a half-life of ~8 h (Figure 3).

Figure 3 The in vitro breakdown of ascorbate under moderately oxidizing conditions in the absence of deliberately added metal ions

Carrier-free L-[1-14C]ascorbic acid (48 μM) was incubated in water (pH≈4.4). Reaction mixtures were aerated by gentle shaking, during which small amounts of H2O2 are generated [15], but no exogenous H2O2 was added. Samples were analysed by HVPE at pH 6.5. An autoradiogram is shown overlying the unstained electrophoretogram. Dotted circles indicate the positions of the marker OG, added within and between each radioactive sample.

The DHA oxidation branch

Non-enzymatic degradation of ascorbate-derived DHA under strongly oxidizing conditions

Theoretically, 1 mol eq of H2O2 is sufficient for the complete oxidation of ascorbate to DHA (ascorbic acid+H2O2→DHA+2H2O). A second mol eq of H2O2 enabled the [14C]DHA to be further oxidized, after the first 30–60 min, to products with the oxidation state of OxA+ThrO (Figure 2b). Major 14C-labelled products co-electrophoresed with cOxT, OxT and free OxA (Figure 2a), and non-radioactive free ThrO was also formed (results not shown). Similar products were obtained with reaction mixtures buffered at pH 4, 5 and 7 (results not shown). DKG was a minor component, and compounds C and E were undetectable under these strongly oxidizing conditions.

During the 5–60 min period of observation, when the average DHA and H2O2 concentrations were ~2.5 and 10 mM respectively, rate constants for the DHA+H2O2→product reactions are estimated at 3.4×10−2, 4.8×10−3 and 3.5×10−3 M−1·s−1, for OxT, cOxT and OxA respectively (Figures 2b and 2c).

Essentially simultaneous generation of cOxT, OxT and OxA during DHA oxidation

It was proposed previously [35] that cOxT was an obligatory precursor of OxT. However, contradicting this hypothesis, we observed that OxT and cOxT accumulated simultaneously (Figures 2b and 2c). Their ratio remained almost constant (~6:1) for the first 60 min, increasing only slightly between 60 and 480 min as the cyclic ester was slowly hydrolysed to OxT (Figure 2d). The rate of total oxalyl ester accumulation (cOxT+OxT) varied in proportion to the concentration of their presumed common precursor, DHA, decreasing after the first 60–120 min. However, the rate of OxT production did not vary in proportion to the concentration of cOxT, which plateaued at 240–480 min (Figure 2b). Thus cOxT is not an obligatory precursor of OxT, and both types of ester are formed essentially simultaneously during the oxidation of DHA.

Similar to [14C]OxT, free [14C]OxA was also produced much more rapidly during the first 120 min than later. This indicates that only a minority of the OxA arose by hydrolysis of OxT. If OxA did arise mainly from esters, its rate of accumulation would peak at 480 min, when the ester concentration was highest. Furthermore, the OxA/OxT ratio decreased slightly between 16 and 480 min, whereas an increase would have been expected if much ester hydrolysis were occurring. The pattern of free [14C]OxA formation suggests that most of it arose from a substance whose concentration, similar to that of DHA but unlike that of OxT, decreased during the incubation.

We conclude that, under non-enzymatic conditions, cOxT is not an obligatory precursor of OxT, and OxT is not an obligatory precursor of free OxA during the H2O2-driven oxidation of DHA. The data suggest that H2O2 oxidizes DHA to a reactive intermediate (I) that almost simultaneously generates three alternative end-products: OxT, cOxT and free OxA at a ratio of ~6:1:1 (Figure 2d). We propose that (I) is cyclic 2,3-O-oxalyl L-threonolactone, formed when bicyclic DHA reacts with H2O2, oxidatively cleaving the C-2–C-3 bond such that the original carbons 1 and 2 become a cyclic oxalyl ester group, and the hemiketal furanose ring becomes a γ-lactone ring (Figure 1). We propose that (I) is highly unstable in aqueous solution at pH 4–7, rapidly undergoing hydrolysis at any of the bonds labelled a, b and c in Figure 1. To account for the observed ratio of end-products formed under the conditions used in Figure 2, we assume that the first hydrolysis reaction of (I) had a 1 in 8 chance of targeting the lactone ring (bond c) to produce the relatively stable cOxT, or a 7 in 8 chance of targeting one of the two oxalyl ester bonds (a or b; a was arbitrarily chosen for Figure 1) to produce 2- or 3-O-oxalyl L-threonolactone (II). Intermediate (II) would itself be highly unstable, quickly undergoing hydrolysis either at bond c (6 in 7 chance) to yield the relatively stable OxT, or at the remaining oxalyl ester linkage (a or b; 1 in 7 chance) to yield OxA and threonolactone. Figure 1 shows 3-OxT as the first-formed product of (II), later converting into 4-OxT, the metabolite reported previously [35]; we have indeed recently obtained evidence for rapid 3-OxT↔4-OxT interconversion by acyl migration (results not shown). Later, most of the threonolactone itself would be hydrolysed to free ThrO. This interpretation would account for the 6:1:1 ratio (OxT/cOxT/OxA) observed among products stable enough to be isolated by HVPE.

After the rapid oxidation of DHA to form (almost simultaneously) cOxT, OxT and OxA, the subsequent inter-conversion of these products by hydrolysis (cOxT→OxT→OxA) was, in the absence of enzymes, much more gradual. When discussing Figures 4 and 5, we will show that when native apoplastic enzymes were present these hydrolysis steps did occur, indicating the presence of oxalyl esterase activities, which would operate in vivo.

Figure 4 The in vivo and in vitro breakdown of ascorbate under moderately oxidizing conditions in culture medium

Carrier-free L-[1-14C]ascorbic acid (360 μM) was incubated in Rosa culture medium containing: (a) no cells or cellular products; (b) living Rosa cells; (c) soluble secreted Rosa products, including native enzymes, but no cells; or (d) as (c), but enzymes had been denatured at 100°C. In all cases, the pH was adjusted to 6.0 with 5 mM Mes (Na+). Reaction mixtures were aerated by gentle shaking; no H2O2 was added. This experiment was conducted five times with qualitatively similar results; representative autoradiograms are shown. The spots are quantified in Supplementary Figure S1 (at

Figure 5 Effect of ascorbate concentration on the in vitro breakdown of ascorbate under moderately oxidizing conditions

L-[1-14C]Ascorbic acid (0.1 mM) was incubated in native or denatured Rosa culture filtrates prepared as described for Figures 4(c) and 4(d). In some cases, the medium was supplemented with non-radioactive ascorbate to the final concentrations indicated. In all cases, the pH was adjusted to 6.0 with 5 mM Mes (Na+). Reaction mixtures were aerated by gentle shaking; no H2O2 was added. The experiment was conducted twice, with qualitatively similar results; one set of electrophoretograms is shown. Incubation was for (a) 2 h, (b) 6 h or (c) 24 h. The spots are quantified in Supplementary Figure S2 (at

The DHA oxidation branch

Fate of ascorbate under moderately oxidizing non-enzymatic conditions: production of DKG

DHA in aqueous solution is reported to undergo rapid and essentially irreversible hydrolysis of the lactone ring to yield DKG [40], for example with a half-life of 21 min at pH 7 and 27°C [41]. In contrast, at pH 2–4, aqueous solutions of DHA are stable ‘for days’ [42]. In the experiment described in Figure 2(a), in the presence of exogenous H2O2, very little [14C]DKG accumulated. This could be because the strongly oxidizing conditions favoured the rapid oxidation of DHA to OxA and its esters over its gradual hydrolysis to DKG. Any downstream products normally formed via DKG would thus also have been under-represented in the experiments with exogenous H2O2.

Moderately oxidizing conditions, more similar to those likely to occur in vivo, were achieved in aerated solutions, buffered at pH 4.4, and containing some residual ascorbate and/or DHA. Under these (still non-enzymatic) conditions, cOxT, OxT and possibly free OxA were again formed simultaneously (Figure 3) but much more slowly than in the presence of exogenous H2O2. At the later time points (24–48 h), concentrations of cOxT plateaued, but OxT continued to accumulate – as expected since a pool of its precursor, DHA, was maintained for this duration. The results indicate gradual oxidation of DHA simultaneously to cOxT, OxT and OxA, with the cOxT slowly hydrolysing to OxT during these lengthy incubations. Other products formed under these conditions included DKG, C and possibly E (Figure 3).

In a further move towards in vivo conditions, [14C]ascorbate was incubated in fresh culture medium (i.e. medium in which Rosa cells had never grown) or in heat-denatured culture filtrate (denatured spent medium). The fresh medium contains the transition metal ions Fe3+ and Cu2+, but in spent medium these ions would have been largely removed from free solution by the cells. Media were buffered at pH 6.0. [14C]Ascorbate in these media underwent reactions (Figures 4a and 4d) similar to those reported in Figure 3, except that yields of DKG were higher, probably because the higher pH promoted the hydrolysis of DHA's lactone ring. cOxT and OxT both accumulated simultaneously, and the cyclic ester plateaued as it turned over to the simple ester(s) (see also Supplementary Figure S1 at

Under all conditions reported in Figure 4, an additional unidentified product, designated compound T, was also formed. If T has a net charge of approximately −1.0 at pH 6.5, as is the case for most mono-carboxylic acids, then its low electrophoretic mobility relative to ascorbate (mascorbate=0.88) indicates that it has a higher molecular mass than ascorbate and may thus be formed by condensation of the ascorbate skeleton to an additional compound present in the medium.

Fate of ascorbate under moderately oxidizing conditions in the presence of apoplastic enzymes

In similar culture medium with no exogenous H2O2, but in the presence of native apoplastic Rosa enzymes, both types of ester were again formed essentially simultaneously. However, they were unstable (Figures 4b and 4c): OxT appeared to be hydrolysed to OxA, which progressively accumulated, indicating oxalylesterase activity, although some OxT was maintained for the full 9 h of observation, suggesting that it continued to be replenished from DHA, whose pool was maintained. cOxT levels fell to undetectable concentrations by 6 h, indicating that an esterase activity hydrolysed it to OxT faster than it could be replenished from DHA.

The results show that the products detectable during ascorbate oxidation depend on: (i) the intensity of oxidation, with slow oxidation allowing time for cOxT to act appreciably as a precursor of OxT even in the absence of enzymes; and (ii) the presence of apoplastic oxalylesterase activities, which strongly promote the turnover of cOxT to OxT and of OxT to OxA.

Compounds C and E as downstream products of DKG in the absence of exogenous H2O2

Nature of compounds C and E

Compounds C and E are mutually interconvertible products of ascorbate degradation discovered under ‘moderately oxidizing’ conditions [35]. C and E are C6 products of ascorbate, with respectively 1 and 2 negatively charged groups at pH 6.5; compound C is a lactonized form of E [35]. Conditions favouring their production were explored in more depth (Figure 5). Their properties, including electrophoretic mobilities, suggested that C [mOG=1.38 at pH 6.5; where mOG, electrophoretic mobility relative to that of OG (mOG=1.0) and glucose (mOG=0.0), corrected for electro-endo-osmosis] and E (mOG=1.96 at pH 6.5) were C6 compounds carrying one and two negative charges respectively; thus C was probably a reversibly lactonized form of E. Most carboxy groups are almost fully ionized at pH 6.5. On electrophoresis at pH 2.0, in contrast, the order of mobility was reversed: C (mOG≈0.65) had a higher negative charge/mass ratio than E (mOG≈0.41) [43], indicating that the single ionizable group of C has an unusually low pKa. This behaviour is comparable with that of cOxT (with one ionizable group), which has a higher negative charge/mass ratio at pH 2.0 (mOG≈0.92) than OxT (mOG≈0.73), which has two, evidently more weakly, ionizable groups.

If we assume that both the anionic groups of E are -COOH groups, then E could theoretically be based on the straight chain compound adipic acid (HOOC-CH2-CH2-CH2-CH2-COOH) with various -OH and=O groups attached; alternatively, it could be a branched structure based on, for example, HO-CH2-CH2-CH2-CH-(COOH)2. An adipate-based structure seems unlikely since it would require non-enzymatic oxidation to have occurred at the relatively unreactive -CH2OH group (C-6 of DKG). On the other hand, a branched-chain structure would be produced by a benzilic acid rearrangement, such as occurs during saccharinic acid formation when a reducing sugar is treated with alkali [44]. A proposed reaction scheme compatible with all the observations is presented in Figure 6, in which we suggest that E is 2-carboxy-L-threo-pentonate (=‘2-carboxy-L-xylonate’) and C is an epimeric mixture of the corresponding lactones (2-carboxy-L-xylonolactone and 2-carboxy-L-lyxonolactone). On this basis, compound E is nominally a derivative of 2-ethylmalonic acid, whose pKa values are 3.0 and 5.8 (, which would account for its moderate electrophoretic mobility at pH 2.0.

Figure 6 Proposed nature of compounds C and E, and the origin of C by benzilic acid rearrangement of DKG

Fischer projection formulae. DKG is shown in its hypothetical dioxo form, although it is likely to exist in aqueous solution predominantly in a hydrated form. The original C-1 (radiolabelled in our experiments) is indicated with *; C-2 is indicated with †. Compound C is expected to be a mixture of C-2 epimers (2-carboxy-L-xylonolactone and 2-carboxy-L-lyxonolactone; arbitrarily shown as the γ- rather than δ-lactone form), whereas compound E is proposed to be the dianion of ‘2-carboxy-L-xylonate’, which is achiral at C-2 and thus more correctly termed 2-carboxy-L-threo-pentonate.

If this proposal for the origin and structure C and E is correct, then the formation of compound C from DKG is neither an oxidation nor a reduction. In agreement with our observations, it would therefore not be promoted by H2O2; indeed, the presence of added H2O2 would be predicted to decrease the DKG→C reaction by oxidizing some of the DKG before the slow benzilic acid rearrangement was able to produce appreciable amounts of C.

Yield of C and E from ascorbate

Ascorbate degradation gave widely varying yields of C and E in different experiments: undetectable (Figure 2), traces (Figure 3), moderate (Figure 4) or high (Figure 5c) yields. If produced at all, C appeared before E. Production of C and E from ascorbate was generally highest under conditions that favoured DKG accumulation, being greater under moderately oxidizing conditions (i.e. with only endogenous H2O2; Figures 3, 4 and 5) than under strongly oxidizing conditions (Figure 2) that quickly converted most of the DHA into oxalyl esters before it had had a chance to be hydrolysed to DKG. These observations support the pathway ascorbate→DHA→DKG→CE.

On incubation in culture medium under moderately oxidizing conditions, dilute ascorbate (0.1 mM) quickly gave a high yield of DKG (20–30% within 2 h; Figure 5a), which later diminished to near zero (Figure 5c). Results were similar in this respect whether or not the medium contained active apoplastic enzymes. Simultaneously with this flux through DKG, compound C accumulated; compound E appeared later, with the total (C+E) reaching 20–25% by 48 h (see Supplementary Figure S2 at and E eventually exceeding C. The presence of native culture filtrate (containing esterases) greatly reduced the levels of OxT and cOxT, as expected, without affecting the final yield of E, indicating that the oxalyl esters were not precursors of E. The production of C+E coinciding with the production and consumption of DKG supports the idea that DKG was the precursor of C+E.

Ascorbate concentrations higher than 0.1 mM led to a much lower percentage yield of C+E, and the DKG concentration steadily increased throughout the 48 h of observation, reaching 20–30% (Figure 5). These observations support the proposed pathway (ascorbate→DHA→DKG→CE), but suggest that the presence of some remaining unreacted ascorbate inhibits the DKG→C step.

Fate of compounds C and Ein vivo

One reason for the variability in the yield of compounds C and E could be their possible degradation in vivo to unknown or undetectable products [5]. To explore this, we supplied electrophoretically purified 14C-labelled C at very low (tracer) concentrations to Rosa culture medium with or without cells and/or extracellular enzymes. The preparation of C initially contained only a small proportion of its de-lactonized form, E (Figure 7a). The ratio was almost unaffected by a treatment with cold dilute acid followed by neutralization (Figure 7b), but C was largely de-lactonized to E by cold dilute alkali followed by neutralization (Figure 7c). During incubation for 2 or 8 h in fresh culture medium (Figures 7d and 7g), culture filtrate containing any native soluble extracellular enzymes (Figures 7e and 7h), or whole Rosa culture still containing live cells (Figures 7f and 7i), the ratio was if anything pushed towards the lactone C rather than the dianion E. Thus, in vivo, compound C is a relatively stable end-product of apoplastic ascorbate/DHA/DKG metabolism, and not readily metabolized further.

Figure 7 Fate of 14C-labelled compound C when incubated in Rosa cell cultures and medium

Tracer levels of 14C-labelled compound C and/or any degradation products were subjected to HVPE at pH 6.5, and strips of the electrophoretogram were assayed by scintillation counting. (ac) Standards: (a) 14C-labelled compound C freshly eluted from an electrophoretogram similar to that shown in Figure 5(b). (b) As (a), but incubated in HOAc then neutralized with NaOH. (c) As (a), but incubated in NaOH then neutralized with HOAc. E is the de-lactonized form of C. (di) 14C-Labelled compound C was incubated in fresh medium (no cells or enzymes; d and g), in culture filtrate (spent medium with native secreted enzymes; e and h) or in the presence of living Rosa cells (f and i) – in each case for 2 h (d, e and f) or 8 h (g, h and i). Glucose (Glc) and OG were external markers. Cmpd, compound.

We conclude that the production of C and E in the apoplast in vivo will depend on several diverse factors including (i) ascorbate concentration, since residual ascorbate appeared to suppress C formation; and (ii) H2O2, with highly oxidizing conditions suppressing C formation by diverting DKG to competing oxidative pathways.


In vitro models produced during the present study have shown two main pathways by which DHA is broken down: (i) oxidation to oxalyl esters, OxA and threonate; and (ii) hydrolysis to DKG, which can then be hydrolysed to carboxypentonates (C and E).

A comparison of in vivo and in vitro pathways highlighted the importance of esterases acting on oxalyl esters in the production of OxA and threonate formed in vivo (Figure 1). Conversely, there was little evidence for enzyme activity in the catabolism of DKG.

We previously suggested that apoplastic ascorbate concentrations vary in short-term pulses [45], with ascorbate being exported and DHA imported. In theory, such an export strategy could minimize the exposure of DHA to apoplastic conditions and therefore reduce the accumulation of cOxT, OxT, OxA, ThrO and carboxypentonates in the apoplast.

Although neither the DKG breakdown products nor the theoretical intermediates between DHA and its OxA products have been identified conclusively, the use of HVPE has clarified the number of pathways stemming from DHA and DKG and the range of intermediates and end-products involved. This has been an important step in consolidating previous data regarding DKG and DHA breakdown whilst adding new information on these pathways. HVPE has been an essential method in providing comprehensive analyses of reaction sequences at sufficient resolution for the capture of detailed information at even very early steps, such as those in the DHA→oxalyl ester pathway. These in vivo models provide an important reference point for future analysis of ascorbate metabolism in vivo and for an evaluation of the possible biological roles of ascorbate's diverse catabolites.


Harriet Parsons performed all of the experiments except those shown in Figures 4 and 7 (Tayyaba Yasmin) and 5 (Stephen Fry). Stephen Fry and Harriet Parsons drew the Figures and drafted the text, with contributions from all authors. Stephen Fry initiated and supervised the research.


We thank the U.K. Biotechnology and Biological Sciences Research Council for a studentship (to H.T.P.), during the tenure of which this work was done.

Abbreviations: cOxT, cyclic oxalyl L-threonate; DHA, dehydroascorbic acid; DKG, 2,3-dioxo-L-gulonate; HVPE, high-voltage paper electrophoresis; OG, Orange G; OxA, oxalate; OxT, oxalyl L-threonate; ThrO, L-threonate


View Abstract