REVIEW ARTICLE Vitamin C Biosynthesis, recycling and degradation in mammals Carole L. Linster and Emile Van Schaftingen Universite ´ Catholique de Louvain, Christian de Duve Institute of Cellular Pathology, Brussels, Belgium Keywords ascorbate; dehydroascorbate; 2,3-diketogulonate; glucuronate; gulonolactonase; L-gulonolactone oxidase; semidehydroascorbate; UDP- glucuronosyltransferases; vitamin C; xenobiotics Correspondence E. Van Schaftingen, Laboratory of Physiological Chemistry, UCL-ICP, Avenue Hippocrate 75, B-1200 Brussels, Belgium Fax: +32 27647598 Tel: +32 27647564 E-mail: vanschaftingen@bchm.ucl.ac.be C. L. Linster, The Department of Chemistry and Biochemistry and the Molecular Biology Institute, University of California, Los Angeles, CA 90095-1569, USA Fax: +1 310 825 1968 Tel: +1 310 825 3137 E-mail: linster@chem.ucla.edu (Received 12 September 2006, revised 1 November 2006, accepted 21 November 2006) doi:10.1111/j.1742-4658.2006.05607.x Vitamin C, a reducing agent and antioxidant, is a cofactor in reactions cata- lyzed by Cu + -dependent monooxygenases and Fe 2+ -dependent dioxygenas- es. It is synthesized, in vertebrates having this capacity, from d-glucuronate. The latter is formed through direct hydrolysis of uridine diphosphate (UDP)-glucuronate by enzyme(s) bound to the endoplasmic reticulum mem- brane, sharing many properties with, and most likely identical to, UDP- glucuronosyltransferases. Non-glucuronidable xenobiotics (aminopyrine, metyrapone, chloretone and others) stimulate the enzymatic hydrolysis of UDP-glucuronate, accounting for their effect to increase vitamin C forma- tion in vivo. Glucuronate is converted to l-gulonate by aldehyde reductase, an enzyme of the aldo-keto reductase superfamily. l-Gulonate is converted to l-gulonolactone by a lactonase identified as SMP30 or regucalcin, whose absence in mice leads to vitamin C deficiency. The last step in the pathway of vitamin C synthesis is the oxidation of l-gulonolactone to l-ascorbic acid by l-gulonolactone oxidase, an enzyme associated with the endoplasmic ret- iculum membrane and deficient in man, guinea pig and other species due to mutations in its gene. Another fate of glucuronate is its conversion to d-xylulose in a five-step pathway, the pentose pathway, involving identified oxidoreductases and an unknown decarboxylase. Semidehydroascorbate, a major oxidation product of vitamin C, is reconverted to ascorbate in the cytosol by cytochrome b 5 reductase and thioredoxin reductase in reactions involving NADH and NADPH, respectively. Transmembrane electron transfer systems using ascorbate or NADH as electron donors serve to reduce semidehydroascorbate present in neuroendocrine secretory vesicles and in the extracellular medium. Dehydroascorbate, the fully oxidized form of vitamin C, is reduced spontaneously by glutathione, as well as enzymati- cally in reactions using glutathione or NADPH. The degradation of vita- min C in mammals is initiated by the hydrolysis of dehydroascorbate to 2,3-diketo-l-gulonate, which is spontaneously degraded to oxalate, CO 2 and l-erythrulose. This is at variance with bacteria such as Escherichia coli, which have enzymatic degradation pathways for ascorbate and probably also dehydroascorbate. Abbreviations BSO, L-buthionine-(S,R)-sulfoximine; DHA, dehydroascorbate; 2,3-DKG, 2,3-diketo-L-gulonate; FAD, flavin adenine dinucleotide; GLO, L-gulonolactone oxidase; GSH, glutathione (reduced form); GST, glutathione S-transferase; GSTO, Omega class glutathione S-transferase; PDI, protein disulfide isomerase; SDA, semidehydroascorbate; UDP, uridine diphosphate; UGT, UDP-glucuronosyltransferase. FEBS Journal 274 (2007) 1–22 ª 2006 The Authors Journal compilation ª 2006 FEBS 1 Vitamin C (or l-ascorbic acid; hereafter, ‘ascorbic acid’ and ‘ascorbate’ will always refer to ‘l-ascorbic acid’ and ‘l-ascorbate’) is unique among vitamins for several reasons. It is present in various foods, partic- ularly of plant origin, in quantities (typically 10– 100 mg ⁄ 100 g [1]) that are several orders of magnitude higher than those of other vitamins. This is certainly related to the facts that it is formed from sugars, which are abundant compounds, and that its biochemi- cal synthesis is rather simple. Another unique aspect of ascorbic acid is that it is a vitamin for only a few ver- tebrate species, those which have lost the capacity to synthesize it. From a structural point of view, it is also one of the rare compounds containing a hydroxyl group that is so acidic as to be completely dissociated at neutral pH (carbon-3 hydroxyl pK a ¼ 4.2). This is related to the fact that ascorbic acid comprises two conjugated double bonds and that a resonance form can be written for the deprotonated monoanionic form (Fig. 1). Resonance forms can also be written for the form of vitamin C that has lost one electron (Fig. 1), making the radical semidehydroascorbate (SDA) much more stable, and thus much less reactive, than most other free radicals [2]. Vitamin C is therefore able to play the role of a free-radical scavenger [3], reacting with highly ‘aggressive’ (oxidizing) species to replace them by a much less reactive and, moreover, enzymati- cally recyclable one, SDA. Ascorbate is certainly the most abundant water-soluble compound acting in one-electron reactions, and this is most probably why it plays the role of a cofactor in reactions cata- lyzed by a number of metal-dependent oxygenases. The Cu + -dependent monooxygenases peptidylglycine a-amidating monooxygenase and dopamine b-hydroxy- lase convert two ascorbate molecules to two SDAs per catalytic cycle [4]. In the case of Fe 2+ ⁄ a-ketoglutarate- dependent dioxygenases (e.g. collagen prolyl and lysyl hydroxylases, the two hydroxylases involved in carni- tine biosynthesis [5], the asparaginyl hydroxylase that modifies hypoxia-inducible factor 1 (HIF-1) [6]), ascorbate most probably serves to reconvert inactive, Fe 3+ -containing enzyme (which results from abortive catalytic cycles) to the active, Fe 2+ -containing form [5]. Because of these important roles, it is not surpri- sing that vitamin C deficiency leads to a debilitating disorder, scurvy, in man and in animals unable to syn- thesize the vitamin. Important progress has been made recently in our understanding of the synthesis and the recycling of vitamin C, and a novel pathway has been described for the degradation of vitamin C in bacteria. This forms the subject of this review. Vitamin C transport, which has also witnessed important developments lately, is only briefly alluded to in the following para- graph, as other recent reviews are available [7–9]. Ascorbate entry into mammalian cells is energy- dependent, being effected by two distinct Na + -depend- ent cotransporters, SVCT1 and SVCT2, which show distinct tissue distributions. Interestingly, targeted dele- tion of the widely distributed SVCT2 is lethal in mice Fig. 1. The three redox states of vitamin C (ascorbate, fully reduced form; SDA, mono- oxidized form; DHA, fully oxidized form), and stabilization of the ascorbate monoanion and SDA by electron delocalization. SDA, semidehydroascorbate; DHA, dehydroascor- bate. Vitamin C metabolism and recycling in mammals C. L. Linster and E. Van Schaftingen 2 FEBS Journal 274 (2007) 1–22 ª 2006 The Authors Journal compilation ª 2006 FEBS [10], further underlining the importance of vitamin C. Dehydroascorbate (DHA; see Fig. 1) is transported by glucose transporters, particularly GLUT1, GLUT3 and GLUT4 [9], and is therefore not energetically dri- ven. However, intracellular DHA is readily converted to ascorbate (see ‘Recycling of vitamin C’) and this highly favourable reductive step drives DHA uptake by the cell. There are also mechanisms allowing the efflux of ascorbate from cells [7], e.g. from enterocytes during intestinal absorption and from liver cells, which in many mammals produce ascorbate, but the molecu- lar identity of the proteins involved in this process is not yet firmly established. Formation of vitamin C in mammals and other vertebrates Outline of the pathway Ascorbate is synthesized by many vertebrates. The occurrence of ascorbate biosynthesis in sea lamprey [11] suggests that this trait appeared early in the evolu- tionary history of fishes (590–500 million years ago), i.e. prior to terrestrial vertebrate emergence [12]. The biosynthetic capacity has, however, subsequently been lost in a number of species, such as teleost fishes, pas- seriform birds, bats (intriguingly, not only the fruit- eating ones, but also others, feeding on blood or insects [13]), guinea pigs, and primates including humans, for whom ascorbate has thus become a vita- min. Fish, amphibians and reptiles synthesize ascor- bate in the kidney, whereas mammals produce it in the liver [11,14]. Vitamin C is also formed by all plant species studied so far [15] and yeasts produce d-erythroascorbate, a C 5 analogue of ascorbate [16]. Interestingly, very different pathways have evolved for vitamin C biosynthesis in animals, plants and fungi. In animals, d-glucuronate, derived from UDP-glucuronate, is reduced to l-gulo- nate, which leads to inversion of the numbering of the carbon chain (‘inversion of configuration’) since the aldehyde function of d-glucuronate (C1) becomes a hydroxymethyl group in the resulting l-gulonate (Fig. 2; see [16] for a review of the early literature). The latter is converted to its lactone, which is oxidized to l-ascorbate by l-gulonolactone oxidase (GLO). In plants, the pathway starts with GDP-d-mannose, which is converted (without change in carbon numbering) to l-galactonolactone, the substrate for the plant homo- logue of GLO, l-galactonolactone dehydrogenase [15]. The synthesis of d-erythroascorbate in yeasts proceeds from d-arabinose [16], but the mechanisms of forma- tion of the latter have not been elucidated. Effect of xenobiotics on vitamin C formation The regulation of vitamin C formation by xenobiotics is described here, because it helps to understand the mechanism of d-glucuronate formation. Other aspects of this regulation are described in a separate section. It was already observed in the 1940s that administra- tion of a series of xenobiotics to animals was followed by enhanced urinary excretion of ascorbate. The stimu- latory effect is shared by a wide variety of structurally unrelated substances such as barbiturates, paraldehyde, chloretone, aminopyrine, antipyrine, 3-methylcholanth- rene, polychlorinated biphenyls (PCB) and 1,1,1-tri- chloro-2,2-bis(p-chlorophenyl)ethane (DDT) [17–19]. Turnover rate studies using radiolabelled ascorbate indicated that the amount of ascorbate synthesized per day was four- to eight-fold higher in chloretone- or pentobarbital-treated rats than in untreated animals [20]. Furthermore, chloretone and barbital were shown to greatly stimulate the incorporation of radiolabelled glucose into urinary glucuronate and ascorbate [21]. As barbital was found to be neither metabolized nor conjugated, but excreted unchanged in urine, its stimu- latory effect on urinary glucuronate and ascorbate excretion was proposed to be unrelated to any detoxifi- cation mechanism. This view was further supported by the observation that compounds such as borneol, a-naphthol and phenolphthalein, known to be primar- ily excreted as glucuronides, had essentially no effect on ascorbate excretion [21]. Furthermore, the findings that the in vivo conversion of both d-glucose and d-galactose to glucuronate and ascorbate was increased by xenobiotics [22], but that this was not the case for the conversion of radiolabelled d-glucuronolactone or l-gulonolactone to ascorbate [23], suggested that stimulation occurs at a step between UDP-glucose and d-glucuronolactone in the ascorbate biosynthesis pathway. Many of the following investigations on this subject studied the effect of agents stimulating vitamin C for- mation on the activity levels of several enzymes potentially implicated in ascorbate synthesis. UDP-glu- cose dehydrogenase and UDP-glucuronosyltransferases (UGTs) were found to be induced by some agents, although not by all of them [18,24]. A study using Gunn rats [25] provided highly suggestive evidence for the involvement of UGTs in the formation of vita- min C. Gunn rats are deficient in UGT isoforms of the UGT1A family, but not of the UGT2 family [26,27]. 3-Methylcholanthrene, an inducer of UGTs of the UGT1A family, increased urinary excretion of ascor- bate in normal rats (five-fold) and heterozygous Gunn rats (two-fold), but not in homozygous Gunn rats. C. L. Linster and E. Van Schaftingen Vitamin C metabolism and recycling in mammals FEBS Journal 274 (2007) 1–22 ª 2006 The Authors Journal compilation ª 2006 FEBS 3 However, treatment with phenobarbital (an inducer of isoforms of the UGT2 family) increased the urinary excretion of ascorbate in normal and homozygous Gunn rats. Taken together, these results indicate that UGT isoforms of the UGT1A, but probably also of the UGT2 family are involved in ascorbate bio- synthesis, possibly by forming a glucuronidated inter- mediate that would be hydrolyzed by microsomal b-glucuronidase or, as suggested below, by catalyzing the hydrolysis of UDP-glucuronate to UDP and Fig. 2. Vitamin C synthesis pathway and pentose pathway in animals. The reactions are catalyzed by the following enzymes: 1, UDP-glucose pyrophosphorylase; 2, UDP-glucose dehydrogenase; 3, nucleotide pyrophosphatase; 4, UDP-glucuronosyltransferase; 5, UDP-glucuronidase; 6, phosphatase; 7, b-glucuronidase; 8, glucuronate reductase; 9, gulonolactonase; 10, L-gulonolactone oxidase; 11, L-gulonate 3-dehydroge- nase; 12, decarboxylase; 13, L-xylulose reductase; 14, xylitol dehydrogenase; 15, D-xylulokinase. Three possible mechanisms for glucuronate formation (a, b and c) are shown (see text). For the sake of clarity, the linear form of glucuronate is represented. SMP30 KO mice, senes- cence marker protein 30 knockout mice; ODS rats, osteogenic disorder Shionogi rats; od ⁄ od pigs, mutant pigs deficient in L-gulonolactone oxidase; GLO KO mice, L-gulonolactone oxidase knockout mice. Vitamin C metabolism and recycling in mammals C. L. Linster and E. Van Schaftingen 4 FEBS Journal 274 (2007) 1–22 ª 2006 The Authors Journal compilation ª 2006 FEBS glucuronate (mechanisms b and c described in the next subsection). These effects on enzyme levels require increased gene transcription and new protein synthesis, which implies that the effect of xenobiotics on vitamin C formation is a long-term effect. However, recent work on isolated hepatocytes demonstrated that the effect of xenobiotics (e.g. aminopyrine, metyrapone, chloretone) on the for- mation of vitamin C and of its precursor, free glucuro- nate, occurs in a matter of minutes [28]. The increase in free glucuronate formation, which was best observed in the presence of an inhibitor of the downstream enzyme, glucuronate reductase, was already apparent after 5 min and reached up to 15-fold. It was accom- panied by a decrease in the UDP-glucuronate level, but little if any change in the concentration of UDP- glucose, indicating that the effect of xenobiotics on vitamin C formation consists in a short-term effect involving an increase of the conversion of UDP-glu- curonate to glucuronate (Fig. 2) and not an increase in the concentration of upstream precursors of glucuro- nate (‘push-effect’). Most of the stimulating agents did not give rise to detectable amounts of b-glucuronides, arguing against the involvement of a glucuronidation- deglucuronidation cycle in the stimulation of ascorbate formation (see below). It may be interesting to notice that up to 100 nmol hexose unitsÆmin )1 Æg )1 liver are channelled towards glucuronate formation in the pres- ence of saturating concentrations of stimulating xeno- biotics [28]. Formation of glucuronate from UDP-glucuronate The formation of glucuronate from UDP-glucuronate could hypothetically involve (a) the cleavage of UDP- glucuronate to glucuronate 1-phosphate, followed by dephosphorylation of the latter by a glucuronate-1- phosphatase [29,30]; (b) the formation of a glucuroni- dated intermediate, on an exogenous or endogenous acceptor, followed by its hydrolysis by b-glucuronidase [31] or esterases (which could hydrolyze acyl-glucuro- nides); or (c) direct hydrolysis of UDP-glucuronate to UDP and glucuronate (Fig. 2). As explained below, recent work performed on liver microsomes supports the third mechanism. As a follow-up of the work showing that a series of nonglucuronidable xenobiotics rapidly stimulate the formation of glucuronate in isolated hepatocytes [28], it was found that the same xenobiotics also stimulated the formation of glucuronate from UDP-glucuronate in liver cell-free extracts enriched with ATP or in liver microsomes supplemented with ATP and a heat-stable cofactor identified as coenzyme A [32]. Quantitatively, the formation of glucuronate observed under these conditions accounted for the formation of glucuronate observed in intact cells, indicating that glucuronate is formed from UDP-glucuronate by a microsomal enzyme. This enzyme is most probably present in the endoplasmic reticulum as, similarly to UGTs, it is sti- mulated by UDP-N-acetylglucosamine, which enhances the transport of UDP-glucuronate into vesicles derived from the endoplasmic reticulum [33]. Although rat liver microsomal preparations hydro- lyze UDP-glucuronate to glucuronate 1-phosphate (presumably because they are contaminated with plasma membrane fragments, which contain a highly active nucleotide pyrophosphatase [34,35]), their glu- curonate 1-phosphate phosphatase activity is insuffi- cient to account for the formation of free glucuronate by this preparation [32]. Furthermore, the formation and hydrolysis of glucuronate 1-phosphate are unaffec- ted by the nonglucuronidable xenobiotics under condi- tions under which glucuronate formation is stimulated approximately three-fold. These and other arguments [32] exclude mechanism (a). Mechanism (b), which is supported by the observa- tions made on Gunn rats [25], is ruled out by the fact that glucuronate formation from UDP-glucuronate occurs in rat liver microsomes in the absence of UGT substrates and is actually inhibited by such substrates (see below). Furthermore, inhibitors of b-glucuronidase and esterases do not affect the formation of glucuro- nate from UDP-glucuronate by microsomal prepara- tions [32], ruling out also the involvement of an endogenous glucuronide acceptor that would still be present in washed liver microsomes. Taken together, these observations lead to the con- clusion that glucuronate is formed by direct hydro- lysis of UDP-glucuronate by a UDP-glucuronidase (mechanism c). The findings that UDP-glucuronidase is similarly sensitive to various detergents as UGTs and that it is inhibited by UGT substrates suggest that it is a side-activity of these transferases (repre- senting 5% of the transferase activity) [32]. The identification of UGTs as the UDP-glucuronidase also allows one to reconcile mechanism (c) with the obser- vations made on Gunn rats [25]. In the only study that focuses on the UDP-glucuronate hydrolase activ- ity of a purified UGT, Hochman and Zakim [36] found that GT 2P had a minor UDP-glucuronidase activity, which could be stimulated by phenylethers and lysophosphatidylcholines up to 0.03% of the transferase activity. When transfected into human embryonic kidney cells, human UGT1A6 displayed a hydrolase to transferase activity ratio of 0.4% under certain conditions [32], which is still about one order C. L. Linster and E. Van Schaftingen Vitamin C metabolism and recycling in mammals FEBS Journal 274 (2007) 1–22 ª 2006 The Authors Journal compilation ª 2006 FEBS 5 of magnitude lower than the ratio observed in liver microsomes. It is likely that the ability of UGTs to hydrolyze UDP-glucuronate varies among isoforms and depends on the phospholipidic environment. This last point may explain the observation that UDP- glucuronidase is inhibited in rat liver microsomes by the addition of ATP and coenzyme A [32], as the lat- ter combination of cofactors could allow the reesteri- fication of lipids in a subcellular fraction known to contain free fatty acids, acyl-CoA synthetase and acyltransferases. Further work is obviously needed to establish which UGT isoforms are involved in the formation of free glucuronate and the conditions under which they are able to do so. Glucuronate reductase (aldehyde reductase) The reduction of d-glucuronate to l-gulonate is cata- lyzed by an NADPH-dependent reductase, with broad specificity, known as aldehyde reductase or TPN l-hexonate dehydrogenase in the older literature [37] and now referred to as aldo-keto reductase 1A1 (AKR1A1) for the human enzyme [38]. K m values of 0.33 and 0.69 mm were obtained for d-glucuronate and d-glucuronolactone, respectively, and these com- pounds are converted by aldehyde reductase to l-gulo- nate and l-gulonolactone, respectively [37]. Aldehyde reductase belongs to the large group of monomeric NADPH-dependent oxidoreductases, known as aldo-keto reductases, which comprise many members in the human genome, including aldose reductases (the closest homologues of human aldehyde reductase, sharing 65% sequence identity [39]) and hydroxysteroid dehydrogenases [38]. These enzymes display broad substrate specificities and it would there- fore not be surprising that, besides aldehyde reductase, other members of the aldo-keto reductase superfamily participate in the reduction of d-glucuronate. Aldose reductase appears to be much less efficient than alde- hyde reductase in this respect [40]. Furthermore, as it is barely expressed in liver [41], it is unlikely that it con- tributes significantly to vitamin C formation. Aldose reductase inhibitors, which have been developed in the hope of preventing diabetic complications by blocking the enhanced formation of sorbitol from glu- cose in hyperglycaemic states, usually cross-react with aldehyde reductase [42,43]. Thus, sorbinil, an inhibitor of both aldehyde reductase and aldose reductase [42,43], was shown to block the conversion of glucuro- nate to downstream metabolites (Fig. 2) and inhibit the formation of vitamin C in isolated rat hepatocytes [28], supporting the involvement of aldehyde reductase in ascorbate synthesis. Urono- and gulonolactonase As discussed in the next paragraph, the enzyme that forms vitamin C acts on the lactone form of l-gulonate. The conversion of d-glucuronate to l-gulonolactone requires the action of two enzymes, a reductase and a lactonase, proceeding either via d-glucuronolactone if the lactonization is the first step, or via l-gulonate if the first reaction is the reduction. Three different types of lactonases acting on sugar derivatives have been charac- terized in mammalian tissues: 6-phosphogluconolacto- nase, uronolactonase and aldonolactonase. The first one is an enzyme of the pentose phosphate pathway, which belongs, in mammals, to the same family of proteins as glucosamine 6-phosphate isomerase [44] and has no (direct) role to play in the formation of vitamin C. Uronolactonase, a microsomal enzyme, hydrolyzes d-glucurono-3,6-lactone (K m  8mm), but is inactive against aldonolactones [45]. It is a metal-dependent enzyme, but its sequence is presently unknown. Aldonolactonase (gulonolactonase) is also a metal- dependent enzyme, acting best with Mn 2+ , which is pre- sent in the cytosol and hydrolyzes a number of c- and d-lactones of a variety of 5-, 6-, and 7-carbon aldonates, including l-gulono-1,4-lactone and d-glucono-1,5- lactone [45]. It also catalyzes the lactonization of aldo- nates, e.g. of l-gulonate, and can therefore participate in the formation of vitamin C [46]. It has recently been identified as SMP30 (senescence marker protein 30) [47], also known as regucalcin, a protein homologous to bac- terial gluconolactonases [48], and which was initially thought to regulate liver cell functions related to Ca 2+ [49] and to be possibly involved in senescence because of its decreased expression with age in liver, kidney and lung [50,51]. The finding that targeted inactivation of the SMP30 gene leads to vitamin C deficiency [47] strongly argues in favour of the involvement of gulono- lactonase in the ascorbate synthesis pathway in mam- mals. This conclusion is in line with earlier findings indicating that l-gulonate rather than d-glucuronolac- tone is an intermediate in this pathway, such as the fact that ascorbate is more readily formed from d-glucuro- nate than from d-glucuronolactone in liver extracts [52], and with the lower K m of glucuronate reductase for d-glucuronate than for d-glucuronolactone (see above). L-Gulonolactone oxidase Characterization of the enzyme and of the catalyzed reaction GLO, a microsomal enzyme, catalyzes aerobically the conversion of l-gulonolactone to l-ascorbate with pro- duction of H 2 O 2 [53,54]. The immediate oxidation Vitamin C metabolism and recycling in mammals C. L. Linster and E. Van Schaftingen 6 FEBS Journal 274 (2007) 1–22 ª 2006 The Authors Journal compilation ª 2006 FEBS product of GLO is 2-keto-l-gulonolactone, an interme- diate that spontaneously isomerizes to l-ascorbate [54]. The preferred substrate of the enzyme is l-gulono- 1,4-lactone, but it also acts on l-galactono-, d-man- nono- and d-altrono-1,4-lactone [55]. Other c-lactones, including l-idono- and d-gluconolactone, were not oxidized by the enzyme, indicating its configurational specificity for the hydroxyl group at C2. K m values for l-gulonolactone ranging from 0.007 to 0.15 mm have been reported [55,56]. The enzyme transfers electrons not only to O 2 , but also to artificial electron acceptors such as phenazine methosulfate and ferricyanide, although not to cytochrome b 5 and cytochrome c [57]. The production of H 2 O 2 is unusual for an enzyme of the endoplasmic reticulum, and one may wonder if this membrane-bound oxidoreductase does not transfer its electrons to another acceptor in intact cells, most par- ticularly because its plant homologues do so. The latter share 30% sequence identity with mammalian GLO and differ from this enzyme in three main aspects: (1) they act specifically on l-galactono-1,4-lac- tone [58,59], which is their physiological substrate; (2) they are bound to the inner mitochondrial membrane [60]; and (3) they do not transfer electrons directly to O 2 , but to cytochrome c [59]. GLO is a 50.6 kDa protein [61], which, as indicated by sequence comparisons, is related to plant l-galacton- olactone dehydrogenase and fungal d-arabinonolactone oxidase, and more distantly to 6-hydroxynicotine oxid- ase, d-2-hydroxyglutarate dehydrogenase and 24-dehy- drocholesterol reductase, all flavin adenine dinucleotide (FAD)-linked enzymes. Each monomer of GLO binds one molecule of FAD, which is covalently linked to a histidyl residue [55,62]. This residue is presumably His54, which aligns with the histidine (His72) that cov- alently links FAD in Arthrobacter nicotinovorans 6-hyd- roxy-d-nicotine oxidase, as indicated by inspection of the structure (pdb file 2BVH) of this bacterial enzyme. The requirement of detergents for the solubilization of GLO from the microsomal fraction [55–57] strongly suggests its membrane localization. Accordingly, the amino acid sequence of the protein contains several strongly hydrophobic regions, which are thus possibly associated with the endoplasmic reticulum membrane [61]. These regions are predicted to form b-sheets rather than the typical transmembrane helical structure. The orientation of the catalytic site towards the lumen of the endoplasmic reticulum is indicated by the intralumi- nal accumulation of ascorbate and the preferential intraluminal glutathione oxidation (presumably by hydrogen peroxide) in rat liver microsomes incubated with gulonolactone [63]. It should be noted that the GLO sequence is apparently devoid of targeting motifs for the endoplasmic reticulum, as indicated by analysis of the sequence with the TargetP program [64]. Molecular defects in man and other species Early enzymological studies identified GLO deficiency as the reason for the inability of some species such as man and guinea pig to synthesize their own vitamin C [65]. Man [66] and guinea pig [67] both have a gene homologous to the rat GLO gene, but they are highly mutated. Compared with the rat gene, which comprises 12 exons, two coding exons (I and V) are missing in its guinea pig homologue [67]. Nucleotide sequence alignment of one exon of the GLO gene from rat with the corresponding exon in the highly mutated, nonfunctional GLO gene of several primates revealed that nucleotide substitutions have occurred at random throughout the primate sequence, as expected for the exon of a gene that ceased to be active during evolution and subsequently evolved with- out functional constraint [68]. From these two exam- ples, and the finding that ascorbate-deficient species are also observed in several other lineages, it appears that inactivation of the GLO gene occurred several times during evolution, suggesting that the loss of this gene may be advantageous to some species. It has been proposed that the formation of hydrogen peroxide by GLO and the glutathione depletion that ensues are detrimental [69] and that the selective pressure to keep the ability of forming vitamin C is lost in species with ample dietary supply of ascorbic acid. Rat, pig and mouse models of vitamin C deficiency are known. The osteogenic disorder Shionogi rat is a mutant rat of the Wistar strain deficient in GLO activ- ity and thus unable to synthesize ascorbate. The GLO cDNA of this mutant was found to contain a single base mutation leading to a Cys fi Tyr substitution at position 61 in the amino acid sequence [70]. Overex- pression experiments in COS-1 cells suggested that this mutation leads to instability of the mutant GLO pro- tein and is responsible for the enzymatic defect in the osteogenic disorder Shionogi rat. The latter manifests deformity, shortening of the legs, multiple fractures, osteoporosis, growth retardation and haemorrhagic tendency when it is fed an ascorbate-deficient diet; these symptoms are largely prevented by providing vitamin C in the food [71]. A similar symptomatology is found in the od⁄ od pig, in which the GLO gene is inactivated due to an intragenic deletion removing exon 8 [72]. GLO knockout mice have also been generated through homologous recombination and the effects of vitamin C deficiency in these mice have been studied C. L. Linster and E. Van Schaftingen Vitamin C metabolism and recycling in mammals FEBS Journal 274 (2007) 1–22 ª 2006 The Authors Journal compilation ª 2006 FEBS 7 [73]. The mutant mice depend on dietary vitamin C supplementation for survival. The most striking effects of a low vitamin C diet on the knockout mice were alterations in their aortic walls (for instance fragmenta- tion of elastic lamina), which were proposed to be caused by defects in the crosslinking of collagen and elastin. An alternative fate for glucuronate: the pentose pathway As described above, d-glucuronate can be metabolized to ascorbate in most vertebrates, well known excep- tions being primates and the guinea pig. By contrast, in all animal species examined, glucuronate can be converted to the pentose l-xylulose in a pathway known as the pentose pathway or the glucuronic acid oxidation pathway [74]. The reactions involved in this pathway are represented in Fig. 2. The reduction step catalyzed by glucuronate reductase (see previous sec- tion) is shared by the ascorbate synthesis pathway and the pentose pathway. In the latter, l-gulonate is then oxidized to 3-keto- l-gulonate by an NAD-dependent dehydrogenase [75]. The cDNA encoding rabbit liver l-gulonate 3-dehy- drogenase has recently been cloned [76]. The enzyme, shown to be identical to lens k-crystallin, displays a K m of 0.2 mm for l-gulonate. 3-Keto-l-gulonate is decarboxylated to l-xylulose by a poorly characterized decarboxylase [77] whose molecular identity is unknown. l-Xylulose is then converted to xylitol by an NADPH-dependent l-xylulose reductase. A cDNA encoding a dicarbonyl ⁄ l-xylulose reductase has been cloned from a mouse kidney cDNA library [78]. This reductase displays a marked preference for NADPH over NADH and is ubiquitously expressed in several mammalian species. A K m value of 0.21 mm for l-xy- lulose has been reported for the human recombinant enzyme. Xylitol is oxidized to d-xylulose by an NAD- dependent enzyme identical to sorbitol dehydrogenase [79]. Finally, d-xylulose can enter the pentose phos- phate pathway after its phosphorylation by d-xylulo- kinase. The latter has been purified to homogeneity from bovine liver and was shown to be a monomeric enzyme of 51 kDa [80]. The pure enzyme acted on d-xylulose and d-ribulose with respective K m values of 0.14 and 0.27 mm. A human cDNA encoding a ‘xylulokinase-like’ protein of 528 amino acids has been isolated [81]. The predicted gene product bears 22% identity to the xylulokinase of Haemophilus influenzae. The occurrence of the pentose pathway in humans is indicated by the fact that rare individuals excrete abnormal quantities of l-xylulose (1–4 gÆday )1 ) in the urine. This benign condition, known as essential pen- tosuria [74], was recognized by Garrod, almost a cen- tury ago, as an inborn error of metabolism. In 1929, Margolis [82] noted that ingestion of aminopyrine leads to a marked increase in pentose excretion in pentosuric subjects and some years later it was shown that this effect could be mimicked by the administra- tion, not only of a series of other drugs, but also of glucuronic acid [74]. This effect, which is very reminis- cent of the stimulation exerted by nonglucuronidable hydrophobic drugs on vitamin C formation in animals [28], is most probably also due to a stimulation of the UDP-glucuronidase activity of UGTs. Pentosuria, an autosomal recessive trait, is due to l-xylulose reductase deficiency [83]. Lane [84] reported the separation of a major and a minor isozyme for l-xylulose reductase in human erythrocytes. In pentos- uric subjects, only the minor isozyme, which displayed a K m for l-xylulose of 100 mm, was detected upon electrophoresis and ion-exchange chromatography. This suggests that homozygosity for the pentosuria allele results in deficiency of the major isozyme, which most probably corresponds to the recently cloned dicarbonyl ⁄ l-xylulose reductase (see above). The gen- etic defect underlying pentosuria has not yet been reported. The benign nature of this condition (the only symp- tom is the elevated urinary pentose excretion) shows that the pentose pathway does not play an indispens- able role in human metabolism. While in most mam- malian species, this pathway produces a precursor for the formation of ascorbate (l-gulonate), in humans and some other species it probably essentially allows the return of a portion of glucuronate carbon to main- stream carbohydrate metabolism. Control of vitamin C synthesis Outline on the regulation of vitamin C synthesis The main control is apparently exerted at the level of the formation of glucuronate from UDP-glucuronate, as enhancement of this conversion is accompanied by an increase in the formation of vitamin C [28]. How- ever, the pathway is branched at the level of l-gulo- nate and the proportion of l-gulonate that is converted to vitamin C or l-xylulose must depend on the relative activities of the rate-limiting enzymes downstream in the pathways. In the case of vitamin C formation, the rate-limiting step downstream is cata- lyzed by l-gulonolactone oxidase, as indicated by the observation that heterozygous (OD ⁄ od) pigs for GLO Vitamin C metabolism and recycling in mammals C. L. Linster and E. Van Schaftingen 8 FEBS Journal 274 (2007) 1–22 ª 2006 The Authors Journal compilation ª 2006 FEBS deficiency have (when fed an ascorbate-free diet) a plasma ascorbic acid level amounting to 50% of that found in control (OD ⁄ OD) pigs [72]. A comparison between the amount of d-glucuronate that accumulates in isolated rat hepatocytes incubated with various xenobiotics in the presence of the glucuronate reduc- tase inhibitor sorbinil (vitamin C formation is then blocked!) and the amount of ascorbate that is formed under similar conditions but in the absence of sorbinil indicates that 30% of l-gulonate is directed towards ascorbate formation [28]. Effect of xenobiotics As mentioned above, the stimulatory effect of xenobio- tics has been ascribed, at least partially, to a short-term effect on UDP-glucuronidase, i.e. most likely UGTs. How the various (always nonglucuronidable) xenobio- tics act is still unknown. Their important structural diversity and their hydrophobic character suggest that they could act by perturbing locally the phospholipidic environment of UGTs and by thus inducing a conform- ational change favouring a hydrolase activity of these transferases. Alternatively, these compounds, which are hydrophobic but lack a suitable glucuronosyl acceptor function, could stimulate the hydrolase activity through a pseudosubstrate mechanism. Besides this short-term action, the effect of some xenobiotics to stimulate the expression of UGTs [18,24,25] or GLO [85] is also con- ducive to a stimulated formation of ascorbic acid. One may wonder what advantage organisms may derive from the stimulation of vitamin C biosynthesis caused by nonglucuronidable xenobiotics. There is no answer at present to this question, but an inter- esting possibility would be that the stimulatory xeno- biotics are membrane-perturbing agents which could favour the generation of reactive oxygen species when inserted in membranes with active electron transport, the increased vitamin C availability playing then a protective role. Effect of glutathione From a quantitative point of view, glutathione and vitamin C are the most abundant reducing agents in cells. Furthermore, GSH is implicated in vitamin C recycling from DHA (see next section). It would there- fore make sense for glutathione to exert a control on vitamin C synthesis. Several experiments performed with glutathione-depleting agents indicate that gluta- thione depletion favours vitamin C synthesis. Adminis- tration to adult mice of buthionine sulfoximine, an inhibitor of glutathione synthesis, led to a two-fold increase in the amount of vitamin C in liver within 4 h [86]. Similarly, incubation of rat hepatocytes with 1-bromoheptane or phorone, which are conjugated with GSH, caused a more than two-fold increase in vitamin C content after 2 h of incubation [87]. On the basis of the observation that a series of glutathione- depleting agents including, surprisingly, dibutyryl cyclic AMP, enhanced vitamin C formation and also glycogenolysis in murine hepatocytes, it was proposed that increased ascorbate synthesis is the result of a ‘push effect’ involving an increase in the concentration of UDP-glucose [88]. However, no measurements of UDP-glucose or UDP-glucuronate were made to sub- stantiate this hypothesis. Furthermore, the potent glycogenolytic agent glucagon does not stimulate glu- curonate [28] or vitamin C [87] formation in rat hepatocytes, and the effects of some of the compounds that were tested by Braun et al. [88] could not be reproduced by other authors [28,87]. The mechanism of the effect of glutathione-depleting agents is therefore presently not understood. One may wonder to what extent some of the agents used to deplete glutathione do not act like xenobiotics, by stimulating the formation of glucuronate from UDP- glucuronate through a direct action on UDP-glucu- ronidase. A more exciting possibility would be that the control is exerted downstream on GLO. If this were the case, glutathione depletion should enhance the con- version of l-gulonolactone to vitamin C. Finally, one has also to consider the theoretical possibility that the glutathione-depleting agents might act by slowing down vitamin C degradation. Recycling of vitamin C As described in the introduction, ascorbate plays major roles as a water-soluble antioxidant and as a cofactor of several enzymes, which lead to its one- electron oxidation to SDA. Disproportionation of SDA (2 SDA fi ascorbate + DHA), in turn, results in the formation of DHA. Both SDA and DHA are reduced back to ascorbate by several enzymatic sys- tems that are briefly reviewed below and schematized in Fig. 3. Reduction of semidehydroascorbate The reduction of SDA in the cytosol has been assigned to enzymes using NADH (NADH-cytochrome b 5 reductase) or NADPH (thioredoxin reductase). SDA can also be reduced in the lumen of neuroendocrine secretory vesicles or in the external medium by trans- membrane electron transfer systems. C. L. Linster and E. Van Schaftingen Vitamin C metabolism and recycling in mammals FEBS Journal 274 (2007) 1–22 ª 2006 The Authors Journal compilation ª 2006 FEBS 9 NADH-cytochrome b 5 reductase Early studies showed that oxidation of NADH by liver microsomes was stimulated in the presence of SDA, and microsomal NADH-cytochrome b 5 reductase was proposed to participate in the electron transfer system, since the purified enzyme itself was able to reduce SDA in the presence of NADH [89]. Ito et al. [90] showed that, in rat liver, most of the NADH-depend- ent SDA reductase activity is localized in the outer mitochondrial membrane. Some activity could be detected in the nuclear and microsomal fractions. Inhi- bition of the mitochondrial SDA reductase activity with specific antibodies suggested participation of NADH-cytochrome b 5 reductase and of a cyto- chrome b 5 -like protein of the outer mitochondrial membrane [90]. NADH-cytochrome b 5 reductase is an FAD-con- taining enzyme [91,92], which exists as a 300 amino acid membrane-bound form and a 275 amino acid soluble form [93]. The membrane-bound protein is located mainly in the endoplasmic reticulum and the outer mitochondrial membranes, but a small fraction of the enzyme is apparently also associated with the plasma membrane [94]. Its C-terminal catalytic domain (275 amino acid residues) is oriented towards the cytosol [95]. The soluble form (identical to the catalytic domain of the membrane-bound form) of the protein is found mainly in erythrocytes, where it is involved in the reduction of methaemoglobin [95]. Fibroblasts of a patient with methaemoglobinae- mia due to a mutation in the NADH-cytochrome b 5 reductase gene were shown to be deficient in NADH-dependent SDA reductase activity [96], which confirms the involvement of this enzyme in SDA reduction. An NADH-linked soluble SDA reductase was also purified from rabbit lens, but the N-terminal sequence of a peptide fragment prepared from this protein did not show significant similarity with any known pro- tein sequence [97]. This suggests that, besides NADH- cytochrome b 5 reductase, additional NADH-dependent enzymes may participate in SDA reduction. Thioredoxin reductase Purified rat liver thioredoxin reductase, a selenopro- tein, was shown to catalyze the NADPH-dependent reduction of SDA with a K m value (in the presence of thioredoxin, which acts as an activator) of 3 lm for this radical [98]. NADPH-dependent reduction of SDA was also demonstrated in dialyzed cytosolic Fig. 3. Recycling of vitamin C. Vitamin C is transported into the cell under its reduced (ascorbate) and oxidized (DHA) forms by active and facilitative transport systems (shown in blue), respectively. The utilization of ascorbate as an antioxidant or enzyme cofactor leads to the for- mation of SDA in the cytosol, neuroendocrine secretory vesicles and the extracellular medium. Various enzymatic systems (represented in green) reconvert SDA to ascorbate. Intracellular DHA, arising through disproportionation of SDA or import from external sources, can also be reduced back to ascorbate by several enzymes (shown in red) or through spontaneous reaction with GSH (not shown). DHA, dehydroascor- bate; GSTO, Omega class glutathione S-transferase; 3a-hydroxysteroid DH, 3a-hydroxysteroid dehydrogenase; PDI, protein disulfide iso- merase; SDA, semidehydroascorbate. Vitamin C metabolism and recycling in mammals C. L. Linster and E. Van Schaftingen 10 FEBS Journal 274 (2007) 1–22 ª 2006 The Authors Journal compilation ª 2006 FEBS [...]... (2006) Senescence marker protein 30 functions as gluconolactonase in l-ascorbic acid biosynthesis, and its knockout mice are prone to scurvy Proc Natl Acad Sci USA 103, 57235728 48 Kanagasundaram V & Scopes R (1992) Isolation and characterization of the gene encoding gluconolactonase from Zymomonas mobilis Biochim Biophys Acta 1171, 198200 49 Yamaguchi M (2005) Role of regucalcin in maintaining cell homeostasis... transiently formed in the presence of H2O2 [140], indicating that the reaction proceeds by successive cleavage of the bonds between C1 and C2 , and C2 and C3 It was speculated that the highly reactive ketose, l-erythrulose, if formed in vivo, could play a role in ascorbate-dependent modications of protein observed in vitro and proposed to occur in vivo in human lens during diabetic and age-onset cataract formation... selenium-decient rats, which have low ( . ARTICLE Vitamin C Biosynthesis, recycling and degradation in mammals Carole L. Linster and Emile Van Schaftingen Universite ´ Catholique de Louvain, Christian. deficient in L-gulonolactone oxidase; GLO KO mice, L-gulonolactone oxidase knockout mice. Vitamin C metabolism and recycling in mammals C. L. Linster and