Thioredoxins o 1 and h 2 show different subcellular localizations and redox-active functions , but cooperatively affect NADPH redox poise and photosynthetic performance in fluctuating light

Arabidopsis contains eight different h-type thioredoxins (Trx) being distributed in different cell organelles. Although Trx h2 is deemed to be localized in mitochondria, its subcellular localization and function remains a matter of debate. Here, Trx h2 localization and function were investigated using cell fractionation studies and reverse genetics. Differential centrifugation and immunodetection showed the Trx h2 protein to be distributed to the microsomal fraction rather than to mitochondrial preparations. To analyze whether Trx h2 has different roles than mitochondrial Trxs, Arabidopsis mutants lacking Trx h2 were compared with mutants deficient in mitochondrial Trx o1 and double mutants with joint deficiencies in both Trxs. Under constant medium light, trxh2 grew as the wild type, while trxo1 and trxo1h2 mutants showed impaired growth. This was accompanied by differences in the metabolite profiles. The trxo1 and trxo1h2 mutants clustered differently from the wild type during the night, revealing a decrease in ascorbate and glutathione redox states. In fluctuating light intensities, genotypic differences in growth rates were attenuated. Compared to the wild type, the fluctuating-light induced decrease in the NADPH/NADP ratio was diminished in the mutants, with the trxo1h2 double mutant showing the strongest effect. This was accompanied by an increase in the photosynthetic efficiency of the trxo1h2 mutant, specifically in the high light phases of fluctuating light. Conclusively, these results support the view that Trxs o1 and h2 are localized in different subcellular compartments, and have different effects on ascorbate and glutathione redox states and growth in medium light, but cooperatively affect NADP(H) redox state and photosynthetic efficiency in fluctuating light.

mitochondrial Trx o1, we selected two Arabidopsis T-DNA insertion mutants (trxo1 and trxh2) showing deficiencies of Trx o1 and Trx h2. Both lines have been comprehensively characterized in previous studies [33,45] and therefore serve as representative T-DNA insertion mutants for the following analyses. Both lines were further crossed to generate a double mutant (trxo1h2) showing joint deficiencies of both Trxs. In experiments to directly compare these lines, the trxo1 and trxh2 single mutants showed different effects on the redox states of AsA and GSH as well as growth in medium light, while the trxo1h2 double mutant showed additive effects on NADP(H) redox state and photosynthetic efficiency in fluctuating light. This indicates different roles of Trx o1 and h2 depending on the light conditions.

The Trx h2 protein enriches in the microsomal fraction
Unlike the mitochondrial Trx o1, the subcellular localization of Trx h2 remains controversial. While previous studies used transient expression systems and reporter genes, here a cell fractionation approach was alternatively adopted to clarify this debating issue. Prior to the fractionation assay, the transcript and protein levels of Trx h2 in mutant lines were analyzed using real-time quantitative PCR and immunoblotting. There was no Trx h2 signal detectable in trxh2 mutants, while the expression of Trx h2 in overexpression lines increased by 100 times wild-type level ( Figure S1, A and B). This documents that both Trx h2 mutants and overexpression lines serve as proper material for the following assays. alternative oxidase (AOX) as marker proteins, respectively. While fractionation assays were performed using plant material from the wild type, trxh2 mutant and Trx h2 overexpression line (Trx h2 ox ), the sensitivity of the antibody allowed the detection of the Trx h2 protein only in immunoblots from the Trx h2 ox line, but not from the wild type. The Trx h2 protein was clearly present in the purified microsomal fraction of Trx h2 ox plants, which was enriched in CNX, while no clear Trx h2 signals were detectable in purified cytosolic fractions (enriched in cyt-FBPase), and in purified mitochondria (enriched in AOX). These results support the view that the Trx h2 protein is confined to the microsomes, rather than to cytosol or mitochondria.
However, more studies are necessary to increase the sensitivity of the immuno-detection of Trx h2 to document its subcellular fractionation also in the wild type.
The trxh2 and trxo1 mutants show differential growth phenotypes when grown in different light conditions Since Trx h2 showed a different subcellular localizations in comparison to Trx o1, we next investigated whether both proteins show differential functions. To do this, we analyzed trxo1 [33] and trxh2 [45] T-DNA insertion lines, which were shown to be representative lines in previous studies, together with a double mutant (trxo1h2) generated by crossing of these lines. In confirmation to previous studies on the trxo1 [33] and trxh2 [45] T-DNA insertion lines, the T-DNA inserted at the third exon of the Trx h2 gene in trxh2, while the T-DNA inserted at the first intron of the Trx o1 gene in trxo1 ( Figure 2A). Furthermore, the expression of Trx h2 gene in trxh2 significantly decreased by 95% compared to the wild type, and the expression of Trx o1 gene in trxo1 significantly decreased by 98% compared to the wild type. In the double mutant (trxo1h2), the expression of both Trx h2 and o1 was significantly decreased in comparison to the wild type ( Figure 2B). This documents that these three different T-DNA insertion lines are null mutants and appropriate for the following applications.
To understand whether deficiency of Trx h2 and o1 affects plant growth, the growth phenotype of the mutant lines was analyzed. In addition to a standard growth condition with medium light intensity (ML), plants were also grown under fluctuating light intensity (FL), consisting of a loop of five-minute low light phase and one-minute high light phase, to mimic natural light conditions in the field. Under constant ML conditions, the growth of the trxh2 mutant was comparable to the wild type, while trxo1 single and trxo1h2 double mutants showed retarded growth ( Figure   3A). Surprisingly, this growth phenotype became subtle under FL conditions, in which all mutants grew like the wild type ( Figure 3B). There is a general decrease in plant growth when FL is compared with constant ML conditions ( Figure 3, A and B), with the decrease being less strongly expressed in trxo1 and trxo1h2 mutants. These data show that deficiency of Trx o1 leads to impaired plant growth under normal light conditions, but not in fluctuating light, while deficiency of Trx h2 has no effect on growth in any of the conditions.

Nocturnal metabolite levels of trxh2, trxo1 and trxo1h2 mutants cluster differently to the wild type
To further understand the differential effects of Trx h2 and o1 on plant growth, a GC-TOF-MS approach was performed to analyze metabolite profiles in the mutants under different light conditions. We first analyzed the data set using principle component analysis (PCA) to get a global pattern of metabolite changes. When nocturnal metabolism was investigated, the three mutants clustered differently to the wild type. The trxo1 and trxo1h2 mutants showed a similar cluster, while the trxh2 single mutant was only overlapping partly with the trxo1 mutant ( Figure 4A).
Interestingly, when metabolism was analyzed in the light, the PCA shows a similar cluster for the wild type and mutants. This holds true for the day phase of constant ML conditions ( Figure 4B Indeed, in the end of night, the trxh2 and trxo1 single mutants and the trxo1h2 double mutants showed a mild decrease in the levels of several soluble sugars, which were the cases for fructose (51-84% of wild-type level), glucoheptose (64-88% of wild-type level), raffinose (37-54% of wild-type level) and xylulose (81-88% of wild-type level), and amino acids, which were the case for arginine (76-84% of wild-type level), glutamine (71-88% of wild-type level), glycine (61-78% of wild-type level), methionine (85% of wild-type level), ornithine (63-80% of wild-type level), and serine (73-89% of wild-type level). Notably, in the mutant lines, the levels of 4-aminobutanoic acid (GABA) and phenylalanine were significantly increased when compared to the wild type ( Figure 5, A and B, left panel; Table S1). In the mutant lines, there was also a clear decreasing pattern in the levels of many organic acids, including 2-piperidinecarboxylic acid (30-54% of wild-type level), adipic acid (56-71% of wild-type level), gluconic acid (68-88% of wild-type level), lactic acid (55-69% of wild-type level), ribonic acid (57-76% of wild-type level), pyruvic acid (71-82% of wild-type level), 2-oxoglutaric acid (69-82% of wild-type level) and succinic acid (67-81% of wild-type level; Figure 5C, left panel; Table S1).
Surprisingly, joint deficiencies in Trx h2 and o1 had no additive effects on metabolite accumulation. However, these metabolite changes in the mutants versus wild type were not sustainable when metabolism in the light was analyzed ( Figure 5, A, B and C, right panel; Table S1).
Under FL conditions, deficiency of Trx h2 an o1 had minor effects on the accumulations of most sugars and sugar alcohols ( Figure 5D). In the HL phases, the changes of most amino acids in the mutant lines were also very subtle, except for alanine (69-92% of wild-type level), glycine (75-91% of wild-type level) and proline (1.3-to-1.5 times wild-type level; Figure 5E, left panel, Table S2). However, in the LL phases, several amino acids showed significant changes in either the single or double mutants. This included glycine (1.3-to-1.5 times wild-type level), proline (1.2-to-1.4 times wild-type level), and O-acetyl-serine (68-76% of wild-type level; Figure 5E, right panel, Table S2). In the HL phases, some organic acids showed a decreasing tendency in the trxo1 and trxo1h2 mutants, which were the cases for galataric acid (80% of wild-type level), hexadecanoic acid (71-85% of wild-type level), ribonic acid (77% of wild-type level) and threonic acid (80% of wild-type level), while, in the LL phases, the changes were not sustainable ( Figure 5F; Table S2). Taken together, the results indicate the significance of Trx h2 and o1 in nocturnal metabolism, while there were only minor effects in FL conditions. When focusing on AsA system first, all three mutants showed a decrease in total AsA levels in the dark and HL phases, while trxo1 single and trxo1h2 double mutants also showed a decrease in ML conditions, compared to the wild type ( Figure 6, A and C).

Deficiencies in Trx
In addition to this change, the AsA reduction state (calculated as the AsA/[AsA+DHA] ratio in %) in mutant lines except the trxh2 single mutant was lower than the wild type under all analyzed conditions ( Figure 6D). The decrease in AsA reduction state due to knockout of Trx o1 was more strongly expressed in the dark and ML than in FL conditions. Joint deficiencies of Trx o1 and h2 did not lead to stronger effects, compared to Trx o1 single deficiency.
Looking at the GSH system, all three mutants showed similar decreases in GSH levels ( Figure 6E) and GSH redox states ( Figure 6H; calculated as the GSH/[GSH+GSSG] ratio in %) and a increase in GSSG level ( Figure 6F) under FL conditions (HL and LL) compared to the wild type, while the similar patterns of three parameters in ML were observed in the trxo1 single and trxo1h2 double mutants, but not in the trxh2 single mutant. In contrast to this change, GSH levels and GSH redox states were similar in all genotypes in the dark. The levels of total GSH pool in all three mutant lines were comparable to the wild type under all analyzed conditions ( Figure 6G). These results indicate that both Trx h2 and o1 serve as positive regulators in maintaining the reductive state of the GSH system specifically in FL conditions, while Trx o1, but not Trx h2, regulates the reductive state of the AsA system under all analyzed conditions. This shows that Trxs h2 and o1 regulate the AsA and GSH redox systems in a different manner.

Deficiencies in Trx h2 and o1 affect the NADPH redox state in fluctuating light
The redox couples, NADPH/NADP + and NADH/NAD + , are also important components for maintaining the cellular redox balance. They also serve as substrates for Trx reductases to regulate the redox state of Trxs. To understand whether deficiencies of Trxs h2 and o1 affect the redox balance of NAD(P)(H), the levels of NADPH, NADP, NADH and NAD were analyzed (Figure 7). In the wild type, the levels of NADPH ( Figure 7A) and NADH ( Figure 7D) as well as the NADPH/NADP ( Figure 7C) and the NADH/NAD ratios ( Figure 7F) were much higher in the light than in the dark, when plants were analyzed under constant ML conditions. This confirms previous studies showing a strong increase in the reduction states of the NADPH and NADH systems upon illumination under normal growth conditions [19,46]. Interestingly, the light-induced increase in the NADPH/NADP ratio was wiped out under FL intensities, yielding NADPH and NADP levels in HL and LL phases that were below those reached in night ( Figure 7C). Also the light-induced increase in the NADH/NAD ratio was strongly attenuated in FL, compared to ML ( Figure 7F).
In all three mutants, the levels of NADPH, NADP, NADH and NAD as well as the NADPH/NADP and NADH/NAD ratios were similar to the wild-type levels when plants were analyzed under ML conditions. Interestingly, there were increases in NADPH level ( Figure 7A) and NADPH/NADP reduction state ( Figure 7C

Discussion
Arabidopsis contains eight different h-type Trxs, which are distributed to different cell organelles. Although Trx h2 has been proposed to be localized specifically to mitochondria and to fulfill similar roles as mitochondrial Trx o1, its subcellular localization and function is still a matter of debate. In this work, we used a cell fractionation approach to clarify the localization of Trx h2. To further understand the functions of Trx h2 and o1, we performed a series of physiological and biochemical analyses to compare representative trxh2 and trxo1 single mutants as well as their crossed double mutant in different light conditions. Our results show that Trx h2 is likely to be localized to the microsome, rather than to mitochondria. In constant light, Trx h2 and o1 harbor differential functions on plant growth, nocturnal metabolism and redox states of AsA and GSH, while, in fluctuating light, both types of Trxs jointly influence NADP(H) redox states and photosynthetic efficiency.

Thioredoxin h2 is most probably associated to ER/Golgi showing a different subcellular localization than Trx o1
The plant Trx h family consists of various isoforms, with Arabidopsis containing eight different h-type Trxs. Since the encoded proteins were found to have no obvious transit peptides, they were initially anticipated to be confined to the cytosol [39].
Interestingly, in further studies using transient expression systems and reporter genes, Trx h2 was proposed to localize to subcellular compartments other than the cytosol, but the results were not consistent. Studies by Gelhaye et al. and Meng et al. in poplar (Populus trichocarpa) and Arabidopsis proposed Trx h2 to reside in mitochondria [37,47], while Traverso et al. used similar studies in Arabidopsis to document that this protein is associated with the endoplasmic reticulum (ER)-Golgi membrane system [38]. To clarify the subcellular localization of Trx h2, we used a cell-fractionation study combined with immunoblot analyses in Arabidopsis. By analyzing Trx h2 ox plants, we found the Trx h2 protein to enrich in microsomal fractions, while it was barely detectable in purified mitochondria or cytosolic fractions, providing direct biochemical evidence that Trx h2 is most likely confined to the endomembrane system, rather than to cytosol or mitochondria ( Figure 1). However, immunodetection was not sensitive enough to allow the analysis of Trx h2 in the subcellular fractions of the wild type. More studies are necessary to improve the sensitivity of Trx h2 immunodetection. Our cell fractionation studies confirm the GFP reporter studies published by Traverso et al. [38]. Moreover, the Trx h2 protein was found to contain an N-terminal extension that harbors a myristoylated residue, which is known to target proteins to membranes [38]. When the N-terminal extension was removed, the modified Trx h2 was found to be relocalized to the cytosol, providing evidence that the N-myristoylation motif is responsible to induce the ER/Golgi localization of Trx h2 [38]. While N-myristoylation associates Trx h2 to Golgi/ER, cleavage of the fatty acyl residue may relocate the protein to other subcellular compartments. It is however unresolved whether relocation of Trx h2 occurs in vivo and whether this is linked to its biological function or environmental/stress signals.  Figure 6D) and GSH reduction states ( Figure 6H). The trxo1h2 double mutant showed similar changes as the trxo1 single mutant with no substantial additive effects. In contrast to this, the NADPH/NADP and NADH/NAD redox couples in all mutant lines were similar to the wild type, with the exception of a nocturnal increase in NADPH/NADP redox state in the trxh2 mutant. These results indicate that unlike Trx h2, Trx o1 is important to maintain the GSH and AsA redox states of the cell in non-stressed conditions. Since the AsA-GSH cycle mainly operates in mitochondria, but has not been documented in ER [48], results are in-line with the differential subcellular localization of Trxs h2 and o1, in Golgi/ER and mitochondria, respectively. This suggests that the different subcellular localization of both proteins is associated with different redox-active functions. It may also explain the differential effect of the two Trx proteins on plant growth.
Recently, with the proteomics analyses, several enzymes of the AsA-GSH cycle were found to be redox reactive and some of them were confirmed to be Trx targets [14,[49][50][51][52]. Indeed, from the in silico prediction, enzymes of AsA-GSH cycle harbor at least one putative redox-reactive Cys (Table S3), while whether the enzymes of AsA-GSH cycle are redox-regulated by Trx h2 and Trx o1 still requires further investigations.
Joint and single deficiencies of Trxs h2 and o1 also led to marked changes in nocturnal metabolite profiles, while there were no substantial changes in the profiles monitored in the light phase ( Figure 4; Figure 5). The PCA of nocturnal metabolite profiles showed that all three mutant lines clustered differently from the wild type ( Figure 4A). However, trxh2 was only partly overlapping with these mutants and clusters more closely to the wild type, even though trxo1 and trxo1h2 clustered together. Thus, Trxs h2 and o1 partially affect global metabolite profiles in different manners, but it will not lead to additive effects when both Trxs are deficient. This further indicates that Trx h2 and o1 only share partially overlapping functions. The metabolic pathways where deficiencies of Trxs h2 and o1 led to similar pattern in metabolite profiles are as follows.
First, the levels of TCA cycle intermediates were generally decreased in the mutants ( Figure 5; Table S1). The results are in line with previous studies proposed that Trx o1 and h2 are involved in the enzyme activation of TCA cycle [33]. Secondly, in addition to the changes in TCA cycle metabolites, there was an extremely strong increase of GABA in all mutant lines ( Figure 5B; Table S1). Because the level of GABA precursor, Glu, was not changed in the mutants, it is likely that the accumulation of GABA is due to the degradation of polyamine, such as putrescine and spermidine [53,54]. The level of putrescine was actually decreased in all mutants (Table S1). Moreover, the over accumulation of GABA indicates that the carbon flow of TCA cycle in the mutants was compromised, so plants were alternatively elevating carbon flux through the GABA shunt to bypass the compromised activity of TCA cycle enzymes [55]. Thirdly, down-regulating the metabolism of organic acids is very likely to compromise the metabolism of amino acids, since organic acids serve as the major source of carbon skeleton for biosynthesis of amino acids. As expected, the levels of many amino acids, including Gly, Ser, Gln, Met, Orn and Arg, were decreased in the mutants ( Figure 5B; Table S1). Notably, the concomitant decreases of Orn and Arg indicate a possible perturbation of Arg biosynthesis pathway [56]. In comparison to other amino acids, Arg harbors a high N:C ratio and acts as important nitrogen storage compound. It is also a precursor for the biosynthesis of polyamines and other nitrogen containing compounds [57,58]. Thus, perturbation in Arg biosynthesis might subsequently lead to negative effects on nitrogen metabolism.
However, it is unlikely that Trx h2 and o1 are able to regulate the enzymes of Arg biosynthesis pathway directly [56], since most of them reside in plastids. Thus, the down-regulation in Arg synthesis might be due to the shortage of organic acids.
Fourthly, during the day, the levels of photorespiratory intermediates, such as Gly, Ser and glycerate, in the mutant lines were comparable to the wild type ( Figure 5B; Table S1). However, it should be noted that plants grown in normal air where the contributions of Trx h2 and o1 to photorespiration are expected to be of minor importance. Indeed, the trxh2 and trxo1 single mutants showed comparable carbon assimilation rates to the wild type when grown under normal air conditions [36,45].
Interestingly, joint deficiencies of Trx h2 and o1 had no substantial additive effects on the accumulation of most metabolites, which makes it difficult to assess how both proteins cooperate in metabolic regulation. Since Trxs h2 and o1 are located in ER/Golgi and mitochondria, respectively, communication between the two different organells is required. Recently, genetically encoded reporters were used to document that there is a physical membrane contact between ER and mitochondria [59]. In addition to this, a physiological connection between ER and mitochondria is associated to ROS levels and the regulation of photorespiration [60]. More work will be required to analyze ER-mitochondria interactions in the regulation of plant metabolism in more detail.

Thioredoxin h2 and o1 cooperatively affect photosynthetic efficiency in fluctuating light
Plants have to manage strong light fluctuations in the field. Rapid alterations in light intensity strongly affect the availability of light energy for photosynthetic electron transport and carbon fixation and require efficient acclimation mechanisms to maintain photosynthetic efficiency and plant growth [1,2]. It has been found that light-dependent chloroplast Trxs play a crucial role in dynamic acclimation of photosynthesis in fluctuating light [19]. Our results show that this extends also to NADPH-dependent extra-plastidial Trxs. Intriguingly, the trxo1h2 double mutant revealed a higher photosynthetic efficiency than the wild type and single mutants under fluctuating light, especially in the HL phases (Figure 8; Figure S3). This indicates that the extra-chloroplastidic Trxs h2 and o1 act in a cooperative manner to dampen acclimation processes in fluctuating light.
Acclimation processes in fluctuating light were proposed to be important to avoid imbalances in NADPH/NADP redox states. However, studies are largely lacking to test this notion. Our results confirm the vulnerability of the NADPH/NADP redox state in fluctuating light conditions. In the wild type, there is a strong decline in the NADPH/NADP redox state in fluctuating light, compared to constant ML, which is quite dramatic since it drops down below nocturnal levels ( Figure 7C). Interestingly, in the trxo1h2 double mutant, the FL-dependent decrease in the NADPH/NADP ratio was attenuated, specifically in the LL phase of FL ( Figure 7C), which is in-line with its improved photosynthetic efficiency under these conditions. This indicates that the cooperative effect of extra-plastidial Trxs h2 and o1 on FL-dependent acclimation processes is linked to NADPH redox homeostasis. The underlying mechanisms, however, remain unresolved. At the moment, it is unclear how Trxs h2 and o1, which are associated to Golgi/ER and mitochondria, respectively, can affect FL-acclimation processes that are known to reside in the chloroplast [61]. The cellular NADPH homeostasis is known to be mediated by inter-organellar malate/oxaloacetate shuttles involving the activities of different malate dehydrogenase (MDH) proteins distributed to chloroplasts, mitochondria, peroxisomes and cytoplasm [62]. Proteomics studies suggested that MDH proteins might act as targets of h-type Trxs [50,51]. Moreover, the activity of NAD-dependent MDH seems to be regulated by Trx o1 [33]. Taken together, it is likely that Trx h2 and o1 might cooperatively modulate the activity of MDH isoforms, and thus affect the cellular redox balance of NADPH and finally photosynthetic efficiency in FL.
Acclimation in fluctuating light has also been proposed to involve a stimulation of photorespiratory processes [63][64][65], providing an alternative explanation for the cooperative role of Trxs h2 and o1 in this context. Deficiencies of Trx h2 and o1 have been found to enhance the activity of glycine decarboxylation in mitochondria to facilitate photorespiratory carbon flow [36,45]. Enhancing photorespiratory carbon flow is proposed to promote photosynthesis [66,67]. It is quite likely that this mechanism is specifically important in FL conditions, which require elevated photorespiratory capacities [63][64][65]. Indeed, single and joint deficiencies of Trx h2 and o1 affected Gly level in FL conditions, leading to an increase in the LL phase, while there was a decrease in the HL phase of FL ( Figure 5E). This suggests that Trxs h2 and o1 proteins are operating as negative effectors of FL acclimation, since they negatively regulate photorespiratory capacity under these conditions. Further investigations are required to confirm the two hypotheses mentioned above and to decipher the underlying regulatory mechanisms. The physiological connections between ER, mitochondria and chloroplasts may also be facilitated by direct membrane contact sites [59].

Conclusion
This study provides biochemical evidence that Trx h2 proteins are most likely associated to microsomes rather than to mitochondria, documenting a different localization in comparison to Trx o1. The different subcellular localization of both proteins is associated with different redox-active functions, with Trx o1, but not Trx h2, being important to maintain the AsA and GSH redox states and plant growth in non-stressed conditions. In contrast to this, there is a cooperative role of both Trxs o1 and h2 in regulating NADPH redox balance and photosynthetic performance in fluctuating light environments. This suggests a possible physiological interaction of both proteins between ER/Golgi and mitochondria, which extends to photosynthetic acclimation in the chloroplast.

Cytosolic fraction isolation
The cytosolic fraction isolation was performed as described before [68]. In brief, 2 g of 2-week-old seedlings was homogenized in the homogenization buffer (50 mM Tris-HCl, pH7.5; 250 mM sucrose; 5% glycerol; 10 mM EDTA; 0.5% [w/v] PVP-10; 3 mM DTT; 1 mM PMSF) on ice, and the cell debris were removed using miracloth followed by centrifugation for 10 min at 8000 x g in 4 o C. The supernatant was transferred into a new tube and centrifuged for 10 min at 16000 x g in 4 o C. The pellet was discarded, and the supernatant was centrifuged for 30 min at 100000 x g in 4 o C.
Afterward, the supernatant was also kept on ice for latter use.

Mitochondrion isolation
The mitochondrion isolation was performed as described before [69]  antibody was produced by ThermoFisher Scientifc, and the others were purchased from Agrisera.

Molecular characterization
Total RNA extraction

Total metabolite extraction
Fifty milligram of ground leave material was suspended in the extraction buffer containing 340 µL of ice-cold methanol, 10 µL of ribitol solution (0.2 mg mL -1 ) and 10 µL of 13 C-sorbitol (0.2 mg mL -1 ), followed by the incubation on ice for 30 min.
The extract was further mixed with 200 µL of chloroform and 400 µL of water followed by the centrifugation for 15 min at 25,000 x g in 4 o C. Fifty microliter of upper aqueous phase was transferred into a glass vial, and dehydrated using a vacuum at ambient temperature.

Gas chromatography coupled time-of-flight mass spectrometry
The detection of metabolite followed the method published before [73][74][75]. The details of sample preparation and condition were also described in the previous publication [76]. The results were analyzed using ChromaTOF 4.5 and TagFinder 4.5 softwares [77].

Ascorbate and dehydroascorbate
The assay followed the protocol described before [78]. In brief, 20 mg of Twenty microliter of this mixture was used for the measurement as above. Subtracting the reduced AsA amount from total ascorbate yields the amount of DHA.
The pure AsA and DHA were used to prepare standard solution for generating the calibration curve.

Glutathione and glutathione disulfide
The extraction followed the same approach mentioned above. The pure GSH and GSSG were used to prepare standard solution for generating the calibration curve.

Pyridine nucleotides
The assay followed the protocol described before [79]. The details of sample preparation and detection were performed as described in the previous publication

Statistical analysis
The Graphpad Prism (version 8.0) was used to generate figures and carry out statistical analyses. The differences between genotypes were assayed via ANOVA followed by Dunnett's post-hoc test unless otherwise described. The R program was used to generate heatmap of metabolite profile and perform principle component analysis.

Accession numbers
Sequence data of this article is availabel from GeneBank/EMBL database.

Data availability statement
All data has been included in this manuscript and the supplemental information.           Table S1 and S2.