The Myo‐inositol pathway does not contribute to ascorbic acid synthesis

Abstract Ascorbic acid (AsA) biosynthesis in plants predominantly occurs via a pathway with d‐mannose and l‐galactose as intermediates. One alternative pathway for AsA synthesis, which is similar to the biosynthesis route in mammals, is controversially discussed for plants. Here, myo‐inositol is cleaved to glucuronic acid and then converted via l‐gulonate to AsA. In contrast to animals, plants have an effective recycling pathway for glucuronic acid, being a competitor for the metabolic rate. Recycling involves a phosphorylation at C1 by the enzyme glucuronokinase. Two previously described T‐DNA insertion lines in the gene coding for glucuronokinase1 show wild type‐like expression levels of the mRNA in our experiments and do not accumulate glucuronic acid in labelling experiments disproving that these lines are true knockouts. As suitable T‐DNA insertion lines were not available, we generated frameshift mutations in the major expressed isoform glucuronokinase1 (At3g01640) to potentially redirect metabolites to AsA. However, radiotracer experiments with 3H‐myo‐inositol revealed that the mutants in glucuronokinase1 accumulate only glucuronic acid and incorporate less metabolite into cell wall polymers. AsA was not labelled, suggesting that Arabidopsis cannot efficiently use glucuronic acid for AsA biosynthesis. All four mutants in glucuronokinase as well as the wild type have the same level of AsA in leaves.

• Two previously described T-DNA insertion lines in the gene coding for glucuronoki-nase1 show wild type-like expression levels of the mRNA in our experiments and do not accumulate glucuronic acid in labelling experiments disproving that these lines are true knockouts. As suitable T-DNA insertion lines were not available, we generated frameshift mutations in the major expressed isoform glucuronokinase1 (At3g01640) to potentially redirect metabolites to AsA.
• However, radiotracer experiments with 3 H-myo-inositol revealed that the mutants in glucuronokinase1 accumulate only glucuronic acid and incorporate less metabolite into cell wall polymers. AsA was not labelled, suggesting that Arabidopsis cannot efficiently use glucuronic acid for AsA biosynthesis.

INTRODUCTION
Ascorbic acid (AsA) is an important antioxidant in plants and mainly involved in the removal of reactive oxygen species (ROS). Humans need to take up AsA with their food (vitamin C), which results in broad scientific interest in AsA biosynthesis and regulation in plants. The role of AsA for plants, the biosynthesis and attempts to increase AsA levels through transgenic approaches is covered by numerous comprehensive reviews and publications (e.g. Wheeler et al. 1998;Ishikawa et al. 2006). Early feeding experiments with radioactive precursors revealed that the biosynthesis of AsA follows different routes in plants and animals. The mammalian route involves UDP-GlcA (UDP-D-glucuronic acid) to attach a GlcA sugar to an unknown metabolite, from which, through the action of a glucuronidase, GlcA is cleaved off (Linster & Van Schaftingen 2007). GlcA is then reduced to L-gulonate, dehydrated to gulono-1,4-lactone and finally oxidised to AsA (Fig. 1). This route could not be confirmed for plants using tracer experiments (Loewus et al. 1956(Loewus et al. , 1962. Instead, it was found, that the label from myo-inositol goes into hemicelluloses and pectins of the plant cell wall via GlcA and UDP-GlcA (Loewus et al. 1962). New radiotracer feeding experiments revealed the AsA biosynthesis route in plants, in which D-mannose is activated to GDP-mannose, epimerised to GDP-L-galactose and hydrolysed to L-galactose (Fig. 1). This metabolite undergoes lactone formation and is finally oxidised to L-AsA (Wheeler et al. 1998). This route is backed up by mutants like vtc1, vtc2 and vtc5 which have reduced levels of AsA or the mutations are even lethal during embryogenesis (Conklin et al. 1997;John et al. 2007). Therefore the predominant route to AsA is the D-Man/L-Gal pathway, and the presence of alternative pathways has long been debated in plants (Smirnoff et al. 2001;Wheeler et al. 2015). The two favourite candidates are the D-galacturonate pathway and a mammalian-like pathway which use D-glucuronate. The mammalian-like pathway was suggested in two publications. In the first one, the enzyme myo-inositol oxygenase (MIOX) was overexpressed in Arabidopsis resulting in a two-to three-fold increase of AsA (Lorence et al. 2004). MIOX catalyses the oxygen-dependent ring cleavage of myo-inositol to GlcA. A follow-up paper identified alkaline phosphatase acting on phytic acid as a possibility to increase AsA. The complete dephosphorylation of phytate releases myo-inositol. When the phosphatase was overexpressed in Arabidopsis a twofold increase in AsA was observed (Zhang et al. 2008). The steady level of phytate is reduced by ca. 7.5 nmolÁgÁFW À1 in the overexpressor lines. The observed increase in AsA is ca. 3 lmolÁgÁFW À1 . This would indicate that the flux through the phytate pathway must be 400 times higher than the flux of AsA turnover.
In independent experiments with the MIOX-overexpressor lines, the increase of AsA could not be confirmed, although the flux from myo-inositol into cell wall polymers was indeed increased (Endres & Tenhaken 2009). Furthermore, a quadruple knockout mutant in the whole MIOX gene family resulted in plants with the same AsA levels as measured in WT plants (Endres & Tenhaken 2011). The overexpression of MIOX genes in tomato (Cronje et al. 2012) or rice (Duan et al. 2012) also did not lead to increased AsA levels.
The mammalian-like pathway to AsA starting from D-GlcA requires three enzymes to catalyse the conversion of D-GlcA into AsA. First, D-GlcA is reduced to L-gulonate by glucuronoreductase. The second step involves the formation of a lactone catalysed by aldonolactonase. The product L-gulono-1,4-lactone is then finally oxidised to L-AsA. Using bioinformatics Ruggieri et al. (2016) found some candidate genes in tomato for the reductase (first step ③ in Fig. 1) but no candidate gene for the second step. Two candidate genes were predicted for the final oxidase step (⑤ in Fig. 1). These data do not confirm enzyme activity or physiological relevance. Aboobucker et al. (2017) recently characterised two genes from Arabidopsis with Lgulono-1,4-lactone oxidase activity. The enzymes have a high K m (33 mM) and a very low k cat (0.005Ás À1 ). The terminal enzyme of the well-established Wheeler-Smirnoff pathway, Lgalactono-1,4-lactone dehydrogenase ④, is far more active, with a K m of 0.17 mM and a k cat of 134Ás À1 . This corresponds to a more than 25,000-fold higher catalytic activity of ④ compared to ⑤. Maruta et al. (2010) overexpressed candidate genes for Lgulono-1,4-lactone in tobacco BY2 cell cultures. They found no increase in AsA in overexpressing lines. Only after feeding high concentrations of L-gulono-1,4-lactone (10 mM) did the BY2 cells have elevated levels of AsA. Aboobucker et al. (2017) also overexpressed the L-gulono-1,4-lactonase genes in Arabidopsis without any increase in AsA levels.
Unfortunately, most of the papers describe the overexpression of genes as the starting point and the outcome in AsA levels at the end, without further measurements of metabolites which might support the data. Here we show that a frameshift mutant in the gene for glucuronokinase1 (GlcAK1) has a reduced flux of GlcA into cell wall polymers. Radiotracer experiments with myo-inositol show no flux of labelled GlcA into the AsA pool but instead accumulate GlcA. Thus, GlcA cannot be metabolised to AsA. The level of AsA in these mutants is the same as in WT plants.

Plant material and growth conditions
In this work Arabidopsis thaliana (ecotype Columbia 0) was used as a WT plant. Previously described T-DNA insertion mutants glcak1-1 (SALK_076931) and glcak1-2 (SALK_127949c) were obtained from the Arabidopsis Biological Resource Center and another two mutant lines glcak1-3 (contains CAS9 nuclease) and glcak1-4 were created by CRISPR/Cas9 technology. Plants were grown in a growth chamber in pots on standard soil (type ED73) under long-day conditions with 16-h light at 150 lmolÁm À2 Ás À1 . Temperature during the light phase was 23°C and 18°C in the dark. For RT-PCR and labelling experiments seedlings were grown in liquid medium. Seeds were surface sterilised in ethanol and grown in 0.5 9 MS (Basal Salt Mixture, Duchefa #M0245; Duchefa, Haarlem, the Netherlands), pH 5.7 (KOH), with 1 gÁl À1 sucrose for 10 days. Then the medium was changed to 0.5 9 MS (or 0.5 9 MS with 0.37 MBq myo-[2-3 H]-inositol; final concentration in medium is 0.625 lM) for the next 3 days. For sugar measurements sterile seeds were incubated on 0.5 9 MS (without micronutrients; Duchefa #M0221), pH 5.7 (KOH), plates with 0.8% plant agar. Harvested samples were immediately used for experiments or frozen in liquid nitrogen and stored at À80°C until used.

Real-time PCR
Total RNA was isolated from 13-day-old seedlings using Tri-Reagent method. The aqueous phase containing the RNA was further purified on a silica spin column and eluted in 40 ll DEPC-H 2 O. Subsequently, RNA was converted to cDNA by RevertAid Reverse Transcriptase (Thermo Scientific, Waltham, MA, USA) using an anchored oligo(dT) primer. Real-time PCR was performed with a Mx3000P qPCR system (Stratagene, San Diego, CA, USA) and PCR products were detected by SYBR green fluorescence. The obtained values were analysed using the 2 ÀDDCT method (Livak & Schmittgen 2001). RT-PCR primers are listed in Table S1.

Resequencing of the position of the T-DNA insertion in GlcAK1 mutants
Position of the T-DNA insertion in glcak1-1 and glcak1-2 was determined by direct sequencing of PCR products obtained with primers binding to the left border of the T-DNA insertion and on exon1 of GlcAK1 (Table S1). Genomic DNA was isolated from a leaf of a 4-week-old plant by standard methods and 1 ll was used as a DNA template for PCR (reaction volume 30 ll). PCR product was purified with the GeneJET PCR Purification Kit (Thermo Scientific).

Construction of CRISPR/Cas9 knockouts
Mutants in GlcAK1 were generated using CRISPR/Cas9 technology (Fauser et al. 2014). A. thaliana (ecotype Columbia 0) were transformed with the CRISPR/Cas9 construct by Agrobacterium tumefaciens (strain GV3101) according to direct-dip protocol (Davis et al. 2009). Plants with a homozygous mutation in GlcAK1 were selected in T 2 generation by sequencing of a PCR product spanning the targeted region of exon1 (Table S1). In this study, two lines, glcak1-3 and glcak1-4, were used.

Feeding of seedlings with 3 H-myo-inositol followed by metabolite separation on HPLC
Ten-day-old seedlings labelled with myo-[2-3 H]-inositol for 3 days were washed twice with 0.5 9 MS medium containing 8 mM inositol to exchange non-specifically bound 3 H-myoinositol. The seedlings were carefully dried with a soft cosmetic tissue and snap frozen in N 2 with two 3 mm stainless steel balls. Seedlings were homogenised in a liquid N 2 cooled ball mill to a fine powder. The homogenate was incubated in 250 ll methanol:chloroform (7:3) for 2 h at 4°C and vortexed several times in between. A total of 350 ll H 2 O was added and samples were incubated in a cooled shaker for 10 min. After centrifugation, the upper phase was transferred to a new reaction tube and dried in a vacuum centrifuge. The dry pellet was re-dissolved in 40 ll H 2 O. 10 ll of this sample were diluted with 100 mM NH 4 -acetate and applied to a Hilic column (125 9 4 mm Nucleodur 100-5; Machery-Nagel, D€ uren, Germany). HPLC analysis was performed with 100 mM NH 4acetate (buffer A) and acetonitrile (buffer B) using isocratic conditions (80% B; flow rate 0.6 mlÁmin À1 ). Samples were collected using a fraction collector, mixed with 2 ml scintillation cocktail (Rotiszint eco plus; Carl Roth, Karlsruhe, Germany) and counted. Standard compounds (AsA, myo-inositol and GlcA) were separated under the same conditions. Ascorbic acid was detected using UV light (262 nm) whereas myo-inositol was determined with an enzyme assay (Megazyme #K-INOSL; Megazyme, Wicklow, Ireland). GlcA was determined with the hydroxy-benzoic acid hydrazide assays for aldehyde groups (5% HBH dissolved in 0.5 M HCl, diluted 1:10 in 0.5 M NaOH directly before use). Fractions from the HPLC separation were dried in a vacuum centrifuge, dissolved in 200 ll HBH reagent and boiled for 10 min. The OD 410 nm was read in a plate reader.
Incorporation of 3 H sugars into cell walls was determined after extraction of ground seedling material, twice with 70% ethanol followed by a chloroform:methanol (1:1) extraction and an acetone extraction step. The pellet corresponding to crude cell walls was air-dried and counted with 2 ml scintillation cocktail (Rotizint eco plus; Carl Roth, Karlsruhe, Germany). In some cases, the lower phase of the metabolite sample, containing the insoluble fraction of the cells, was further extracted with 70% ethanol as described above.

Ascorbic acid measurements
The AsA was measured from individual leaves of 4-week-old plants. The fresh weight was determined, and the leaf was then immediately homogenised in 1 ml 1 M HClO 4 and sea sand in a small cooled mortar. The homogenate was transferred to a reaction tube, centrifuged for 2 min, and 500 ll of the clear supernatant was neutralised to pH 5 by addition of 42 ll 5 M K 2 CO 3 and 200 ll HEPES-KOH buffer 0.1 M pH 7. The sample was briefly stored on ice and precipitated KClO 4 removed by centrifugation for 2 min. 100 ll of this sample were mixed with 900 ll Na-P i buffer pH 5.6 and the OD 262 nm determined. One unit of ascorbate oxidase (Applichem, Darmstadt, Germany) was added to the assay and carefully mixed. After 3 min, the OD 262 nm reached a stable value. The difference between the two values corresponds to reduced AsA. For total AsA, 200 ll of the neutralised extract were mixed with 200 ll NaP i buffer (100 mM pH 7.5) and 20 ll TCEP (25 mM) and incubated at room temperature for 30 min. The amount of total AsA was measured with 200 ll extract as described above.

RESULTS
The conflicting data on the role of a mammalian-like pathway in AsA synthesis in plants was investigated in knockout mutants of glucuronokinase1 (GlcAK1). It has previously been shown that GlcAK is part of a salvage pathway for GlcA leading via GlcA-1P to UDP-GlcA, a precursor for roughly half of the biomass of primary cell walls in Arabidopsis (Pieslinger et al. 2010). We therefore hypothesised that blocking the pathway to GlcA-1P should potentially redirect GlcA to AsA via a mammalian-like pathway (Fig. 1).
Verification of T-DNA insertional mutants glcak1-1 and glcak1-2 To study an influence of GlcAK1 knockout on the mammalianlike pathway, previously described T-DNA insertional mutants glcak1-1 and glcak1-2 (Xiao et al. 2017) were chosen ( Fig. 2A). However, when relative GlcAK1 expression in mutant lines was compared to WT, no differences were found (Fig. 2B). To confirm the position of the T-DNA insertion, both lines were verified by sequencing. The T-DNA insertions was found in the promoter region, which differed in the two mutants from the positions previously described (Xiao et al. 2017), but corresponded to the locations found in the databases (http:// signal.salk.edu/cgi-bin/tdnaexpress; Arabidopsis Information Resource (TAIR)) ( Fig. 2A). The transcription level of GlcAK1 was not decreased, probably due to the position of the T-DNA insertion, which was detected in the promoter region 197 bp upstream of the ATG start codon (Fig. 2B).

Preparation of new GlcAK1 mutants with CRISPR technology
In order to obtain plants with a loss of GlcAK1 activity, CRISPR/Cas9 mutants were prepared. The gRNA was targeting a sequence in exon1. In the T 2 generation, several plants with a homozygous mutation were identified. For this study, glcak1-3 and glcak1-4 containing a frameshift mutation resulting in a premature stop codon close by were used. Both lines showed a decreased transcript level of GlcAK1, around 28% of the WT values (Fig. 2B), possibly because the nontranslated part of the mRNA makes it more sensitive to degradation.

Seedlings of CRISPR/Cas9 mutants glcak1-3 and glcak1-4 accumulate GlcA after 3 H-myo-inositol feeding
To investigate whether the mutation in GlcAK1 redirects GlcA from the MIOX pathway to AsA, 3 H-labelled myo-inositol was fed to 10-day-old Arabidopsis seedlings. After 3 days of feeding, radioactivity in soluble and cell wall fractions was measured. As expected, in WT and the SALK mutants (glcak1-1; glcak1-2) the main portion of the label was incorporated into cell walls (73-79%; Fig. 3A). In contrast, the CRISPR/Cas9 frameshiftmutants (glcak1-3; glcak1-4) showed reduced values of label in cell walls (32-40% incorporated 3 H), whereas labelled soluble metabolites are much more abundant in glcak1-3 and glcak1-4. The SALK mutants and WT had only 21-27% of the label, but glcak-3 and glcak-4 retained 68% and 60% label in the soluble fraction, respectively (Fig. 3B). As we have generated knockout mutants only in the major expressed isoform GlcAK1, certain amounts of label go into the cell wall fraction via the second isoform of glucuronokinase.
In order to identify the 3 H-labelled metabolites derived from inositol feeding, soluble fractions of WT and glcak1-3 and glcak1-4 were further separated by HPLC on a HILIC column and eluates were collected in 0.33-ml fractions (Fig. 4). The HILIC column was chosen as it allowed the separation of the three metabolites of interest, myo-inositol, GlcA and AsA. All samples were measured by scintillation counting. Surprisingly, the largest portion of the label was found in the GlcA peak, where glcak1-3 and glcak1-4 mutants accumulated around 26 times more label than WT plants. We cannot fully exclude that some of label is also present in L-gulonate, the product of the enzyme glucurono reductase if this metabolite would co-elute with D-glucuronic acid. L-gulonate neither absorbs UV light nor has an aldehyde group, which we would need to detect this compound. Whether the enzyme glucoronate reductase exist in plants cannot easily be answered by bioinformatics. This enzyme belongs to a larger gene family of conserved aldehyde reductases, which, however, act on various but diverse substrates. The gene from mouse was recently identified and the enzymatic function was confirmed in knockout mice (Takahashi et al. 2012). In fractions containing AsA, no increased radioactivity was detected in any plant line The T-DNA position according to Xiao et al. (2017) is shown with dashed lines. The position and the sequences of the frame shift mutants glcak1-3 and glcak1-4 in exon1, generated by CRISPR/Cas9, is also shown. UTRs are indicated in black and exons in grey boxes, introns are represented by black lines. B: Expression of GlcAK1 was measured by qPCR for the four glcak1 mutants. The data show average expression from three biological independent experiments. Statistical differences were evaluated using ANOVA (Tukey's test, P < 0.01), different letters display significant differences between lines. (Fig. 4). The data show that myo-inositol is almost quantitatively converted to GlcA, which strongly increases in concentration when the glucuronokinase is blocked but a redirection of GlcA into other pathways including the formation of AsA does not occur. Loewus (1963) used two different isotopes of myo-inositol, labelled either at [2-3 H] or [2-14 C]. They did not find labelled AsA after feeding both forms of radioactive myo-inositol, consistent with our results.

The CRISPR/Cas9 mutants glcak-3 and glcak-4 accumulate GlcA
To further test whether GlcA accumulates also under normal growth conditions, GlcA concentration in WT and glcak1-3 and glcak1-4 mutants was measured in seedlings grown on 0.5 9 MS agar plates. Both mutants showed a 6-19 times higher concentration of GlcA compared to WT (Fig. 5). This finding suggests that glucuronoreductase has either very low activity or is absent in plants. The concentration of GlcA in metabolite extracts from WT plants is rather low.

Measurement of AsA in leaves
Ascorbic acid levels in leaves of 4-week-old plants were analysed at different time points of the day. There was no significant difference between WT and all glcak1 mutants (Fig. 6). The amount of reduced AsA is also highly similar between all genotypes.

DISCUSSION
Ascorbic acid is an important antioxidant in plants and animals for which a few different biosynthetic pathways have evolved. Here we address the long-debated question whether a mammalian-like pathway for AsA is functional in Arabidopsis, which was proposed in some previous publications (Lorence et al. 2004;Zhang et al. 2008). The concept of the mammalianlike pathway needs free GlcA as precursor, which is then converted to L-gulonate, L-gulono-1,4-lactone and finally oxidised to AsA. A clear difference in the concept between plants and animals is the source of GlcA. Whereas mammals use GlcA, which is assumed to be derived from the hydrolysis of a glucuronylated compound in the liver (Linster & Van Schaftingen 2007), the suggested source of GlcA in plants is the oxygenative ring cleavage of myo-inositol by the enzyme MIOX. Another important difference between plants and animals is a salvage pathway for GlcA in plants, which proceeds via GlcA-1P to UDP-GlcA. This pathway is absent in mammals (Pieslinger et al. 2010;Gangl et al. 2014). The salvage pathway for GlcA is well established and functional in plants, as indicated by several radioactive labelling experiments, in which the flux from myoinositol into cell wall material was found (Loewus et al. 1962;Seitz et al. 2000;Kanter et al. 2005;Endres & Tenhaken 2011). Furthermore, ugd2,3 mutants in the biosynthesis of UDP-GlcA, which have a defect in the formation of UDP-GlcA via UDP-Glc by the enzyme UDP-glucose dehydrogenase, show a severe root phenotype (Reboul et al. 2011). The short roots and the defects during development can be largely rescued by feeding myo-inositol or GlcA, which via the salvage pathway provides UDP-GlcA to the plant mutants. So, if plants would also use GlcA for AsA biosynthesis, there would be competition between the cell wall pathway via glucuronokinase and the hypothetical conversion to L-gulonate and further to AsA.
We have generated frameshift knockout mutants (glcak1-3; glcak1-4) in the GlcAK1 gene as several of the available T-DNA lines did not have changed relative gene expression compared to WT when tested. When low concentrations of 3 H-myo-inositol was fed to WT and the glcak1-3 or glcak1-4 mutants a clear difference was found in the labelled products. Whereas less 3 H label is present in the cell wall fraction of glcak1-3/1-4, a massive accumulation occurred in the soluble metabolites. A detailed HPLC analysis of the soluble products clearly shows that the label accumulates as GlcA in Arabidopsis. The inositol is quantitatively converted to GlcA by the MIOX enzymes but the use for cell wall biosynthesis thereafter is partially blocked in glcak1-3 and glcak1-4. If plants use GlcA for AsA biosynthesis we would have expected to find the label in AsA rather than its accumulation in GlcA. The amount of fed 3 H-inositol is less than 1 lM excluding a perturbation of the metabolism by overloading the pathway. AsA is typically present in concentrations higher than 1000-fold (low mM range). Here we show that the flux from myo-inositol to AsA Fig. 3. Distribution of radiolabelled compounds after 3 H-myo-inositol feeding. Seedlings were grown in 0.5 9 MS medium and 3 H-myo-inositol was added for 3 days. A: Radioactivity from insoluble cell wall, and B: ethanol soluble metabolites determined by scintillation counting. Statistically significant differences were determined using ANOVA (Tukey's test, P < 0.01), different letters display significant differences (n = 6).
does not occur in Arabidopsis, because accumulated GlcA cannot be converted to AsA. The Loewus group came to a similar outcome, but their data could also be explained by the predominance of the salvage pathway to cell wall precursors (Loewus et al. 1962). The possible explanation for the accumulation of label in GlcA might be due to absence of glucuronate reductase, which is responsible for the conversion to L-gulonate. To our knowledge, there are no clear data for the presence of this enzyme in plants.
The studies of Lorence et al. (2004) and Zhang et al. (2008) rely on a pathway which, according to the data presented in this paper, does not exist is Arabidopsis. Furthermore, AsA measurements in the same MIOX4-overexpressing plants as Lorence et al. showed no clear differences in the AsA concentration between WT and the transgenic lines (Endres & Tenhaken 2009). The differences in the results for the same MIOX4-overexpressing plants as in Lorence et al. remain difficult to explain. One aspect is the concentration of myo-inositol in the transgenic lines, which is different between WT and the MIOX4-overexpressors (Endres & Tenhaken 2009). Changes in the concentration of myo-inositol are associated with changes in galactinol, a dimer of galactose and myo-inositol. Galactinol was recently associated with stress gene expression in tobacco (Kim et al. 2008). The studies of Lorence et al. (2004) and Zhang et al. (2008) provide no direct link between higher transcript levels for MIOX4 and higher levels of AsA. Fig. 4. Profile of 3 H-labelled compounds after separation on a Hilic HPLC column. Soluble metabolites after 3 H-myo-inositol feeding (as described in Fig. 3) were subjected to HPLC separation. Fractions of 0.33 ml were collected in a microtiter plate and counted. The experiment was done with glcak1-3 and glcak1-4 mutants, showing the same metabolite pattern. The elution profile of AsA was determined by UV absorption. GlcA was measured by aldehyde colour assay and myo-inositol was quantified by an enzymatic reaction. Fig. 5. Metabolite analysis of WT, glcak1-3 and glcak1-4 mutants. Metabolites were separated on a CarboPac PA20 column to quantify GlcA. The data show average AE SD of three independent biological experiments. Mutants were compared to WT using t-test (two-tailed, unpaired, P < 0.01), significant differences are displayed with *. Inspired by the publication of Lorence et al. (2004), other research groups have also overexpressed MIOX genes, for instance in tomato (Cronje et al. 2012) or rice (Duan et al. 2012), to increase the AsA level in these crops. None of the transgenic plants contained higher levels of AsA than the WT controls, which is in good agreement with our data. If one includes the labelling studies from Loewus et al. (1962), it can be concluded that a mammalian-like pathway to AsA is not functional in a diverse group of plants including a monocot.
Other groups have addressed the different pathways to AsA by searching for conserved biosynthesis genes (Wheeler et al. 2015;Ruggieri et al. 2016). These studies show that, for example, the enzymes gluconolactonase, as well as L-gulonolactone oxidase, are absent in the genomes of higher plants. Moreover, there were no expressed sequence tags for gluconolactonase in kiwifruit (Actinidia spp.), which is one of the most suitable candidates to study mammalian-like pathway as it is rich in AsA content and has much higher myo-inositol concentrations than other plants (Bieleski et al. 1997;Crowhurst et al. 2008). Taken together, the information on the part of the mammalian-like pathway starting from GlcA is either ambiguous or the evidence for particular genes is missing. There is therefore a need to confirm the data biochemically to prove the function of the enzymes and their biological relevance. Xiao et al. (2017) recently published experiments about a knockout in GlcAK1 and showed some changes in stress response, different expression of sugar-related and ABAresponse genes. The changes were attributed to the knockout of the GlcAK1 gene. We have tested the same T-DNA lines (glcak1-1 and glcak1-2) in our feeding experiments, but neither T-DNA line show accumulation of GlcA as found in the frameshift mutations glcak1-3/1-4 (compare Fig. 4). The paper of Xiao et al. (2017) suggests a T-DNA insertion position close to the ATG start codon, which we cannot confirm. Resequencing of the DNA insertion position by us revealed a position 197 bp upstream of the ATG start codon, which is identical to the position shown on the T-SIGNAL webpage (http://signal.sa lk.edu/cgi-bin/tdnaexpress). We also measured relative GlcAK1 expression with the glcak1-1 and glcak1-2 mutants, showing no difference between WT and mutant lines. This also explains why the flux of 3 H-myo-inositol into cell walls is very similar in WT, glcak1-1 and glcak1-2 mutants.
Plants have established recycling pathways for many sugars, including GlcA. Mutations in the gene for GlcAK1 already lead to a significant increase in GlcA, although a second isoform exists in the Arabidopsis genome. The flux of 3 H label from myo-inositol into the cell wall is also detected in glcak1-3 and glcak1-4 mutants, although at a lower level, which confirms the function of GlcAK2 (At5g14470) as a second isoform of GlcAK. The paper of Zhao et al. (2013), however, claims a biological function of this gene as a galactokinase rather than a glucuronokinase. This is highly unlikely as the only galactokinase in Arabidopsis is encoded by a different gene (At3g06580; Egert et al. 2012). A possible misinterpretation of the gene ontology terms might have caused the wrong annotation and questionable conclusions of the paper. In fact, we have confirmed the function of GlcAK2 as a true glucuronokinase in preliminary experiments with purified recombinant enzyme from transient expression in tobacco plants.

CONCLUSION
The mammalian-like pathway to AsA via myo-inositol and GlcA was proposed in previous publications. Here we show that a knockout in glucuronokinase1 reduces the flux of GlcA into cell wall polymers and leads to an accumulation of GlcA. We also found no evidence that GlcA is further used to synthesise AsA in Arabidopsis. Moreover, any direct evidence for a mammalian-like pathway to AsA in plants is lacking.