Reduced γ-glutamyl hydrolase activity likely contributes to high folate levels in Periyakulam-1 tomato

Abstract Tomato cultivars show wide variation in nutraceutical folate in ripe fruits, yet the loci regulating folate levels in fruits remain unexplored. To decipher regulatory points, we compared two contrasting tomato cultivars: Periyakulam-1 (PKM-1) with high folate and Arka Vikas (AV) with low folate. The progression of ripening in PKM-1 was nearly similar to AV but had substantially lower ethylene emission. In parallel, the levels of phytohormones salicylic acid, ABA, and jasmonic acid were substantially lower than AV. The fruits of PKM-1 were metabolically distinct from AV, with upregulation of several amino acids. Consistent with higher °Brix, the red ripe fruits also showed upregulation of sugars and sugar-derived metabolites. In parallel with higher folate, PKM-1 fruits also had higher carotenoid levels, especially lycopene and β-carotene. The proteome analysis showed upregulation of carotenoid sequestration and folate metabolism-related proteins in PKM-1. The deglutamylation pathway mediated by γ-glutamyl hydrolase (GGH) was substantially reduced in PKM-1 at the red-ripe stage. The red-ripe fruits had reduced transcript levels of GGHs and lower GGH activity than AV. Conversely, the percent polyglutamylation of folate was much higher in PKM-1. Our analysis indicates the regulation of GGH activity as a potential target to elevate folate levels in tomato fruits.


Introduction
Tomato is the sixth most important globally grown crop. Tomato is also a nutritionally important crop as it is a rich source of several nutraceuticals, particularly antioxidant lycopene and βcarotene, a precursor of vitamin A. Though tomato leaves are enriched in vitamin B9, commonly called folate, the fruits have a moderate folate level [43]. Folate is an essential vitamin for all organisms; however, animals cannot synthesize folate and obtain it from plant-derived food. The staple cereal crops, wheat and rice, have low folate levels, necessitating the fortification of foods with folic acid to meet the RDA (recommended dietary allowance) [19]. To obviate the folate fortification, efforts are directed to biofortify the important crops with folate [9].
The strategy to biofortify the folate in the edible portion of the crop plants has focused on two approaches, germplasm diversity, and metabolic engineering. The studies on transgenic manipulation of folate biosynthesis in tomato revealed that upregulation of both GTP cyclohydrolase I (GCHI), and aminodeoxychorismate synthase (ADCS), the key enzymes synthesizing pterin and paminobenzoic acid (pABA), respectively, was essential for folate biofortification [5]. Alike tomato, the transgenic upregulation of GCHI and ADCS in rice also led to folate enrichment [40]. Contrastingly, in potato tubers, the overexpression of GCHI and ADCS did not improve folate levels, highlighting species-specific regulation of folate biosynthesis. The inclusion of two additional genes pyrophosphokinase/dihydropteroate synthase (HPPK/DHPS) and folylpolyglutamate synthetase (FPGS), enabled a 12-fold enrichment of folate in potato [6]. The protection of folate polyglutamylation by lowering γ -glutamyl hydrolase (GGH) level moderately improved the folate levels in tomato fruits [1]. Figure 1 shows the folate structure, biosynthesis, and salvage pathway.
The folate levels in plants may also be determined as part of cellular homeostasis by interaction with other metabolites and vitamins. In folate biosynthesis, pyridoxal 5 -phosphate (PLP, VitB6) is required as a cofactor for aminodeoxychorismate lyase (ADCL); thus, folate biosynthesis, in turn, may be determined by the levels of PLP in the plastids [3]. Folate also controls redox homeostasis by assisting NADPH production, affecting cellular homeostasis [10]. Adding or removing one-carbon units is an important part of regulating plant metabolite levels and, in turn, affects the primary metabolism of plants. Folate, a key constituent of one-carbon metabolism, provides C1 units to synthesize nucleic acid, pantothenate (vitamin B5), and amino acids in plants [19].
The analysis of folate levels in different crop species revealed a limited range of variation across the germplasm. An analysis of folate levels in 175 wheat genotypes revealed 2.12-and 2.27-fold variations in winter wheat and spring wheat, respectively [32]. An analysis of 67 accessions in spinach showed a 3.2-fold variation in folate levels [37]. In 125 accessions of tomato, the folate level in red-ripe fruits varied from 13.8-45.8 μg/100 g FW (Fresh weight), while in mature-green fruits, it ranged from 12.5-70.9 μg/100 g FW. In most accessions, folate levels declined during fruit ripening [44]. The decline in folate level may be causally associated with reduced expression of GCHI and ADCS with the progression of ripening in tomato fruits [48].
The limited diversity in folate levels in germplasm may be associated with the absence of polymorphism in genes regulating folate biosynthesis. In tomato, analysis of 391 accessions revealed the paucity of SNPs in exons of GCHI, ADCS, and other folate biosynthesis/metabolism genes. Since folate is essential for plant metabolism, it was surmised that the genes related to folate biosynthesis and metabolism are least likely to harbor genic polymorphism [44]. Thus, the FPGS and DHFS knockout mutants led to embryo-lethal phenotype in Arabidopsis [28]. Even after double mutagenesis of tomato, it was found that genes involved in folate biosynthesis/metabolism were recalcitrant to mutagenesis [15]. The above reports highlighted that the folate levels in tomato are subtly modulated by cellular homeostasis [44].
The knowledge about how gene regulation, protein modifications, transport, and turnover affect endogenous folate levels in plants is incomplete. An analysis of QTLs in rice detected three QTLs upregulating folate levels in grains, but none was in folate biosynthesis genes [49]. Likewise, two QTLs upregulated folate levels in maize, with a S-adenosyl-L-methionine-dependent methyltransferase and a transferase with the folic acid-binding domain as putative candidate genes [12]. In tomato, mutations in the MYB117 transcription factor, namely the tf-5 allele, upregulates the folate levels in fruits, indicating the role of transcription factors in determining the folate levels in plants [42].
To comprehend how folate level is regulated in tomato and inf luences primary metabolism, we compared two contrasting tomato cultivars, Periyakulam-1 (PKM-1) and Arka Vikas (AV), bearing high folate and low folate levels in fruits, respectively. We report that high folate levels alter cellular homeostasis by elevating most amino acids. The high folate level in PKM-1 was also associated with the high carotenoid levels in PKM-1 fruit. We also show that the reduced activity of GGH in fruits is linked to increased folate levels in PKM-1 by higher folate polyglutamylation.

Folate levels in PKM-1 and AV fruits
A large-scale screening of tomato accessions and cultivars revealed wide variation in folate levels of red-ripe fruits [44]. To understand the molecular basis governing the variations in folate levels, we selected two tomato cultivars, Periyakulam-1 (PKM-1) and Arka Vikas (AV), with widely different folate levels. Both cultivars are popularly grown by farmers in Southern India under similar agroclimatic conditions. Of these, PKM-1 displayed high (49.0 ± 0.45 μg/100 g FW), and AV displayed low (21.2 ± 1.72 μg/100 g FW) folate levels in red-ripe fruits (Figure 2A) grown in the open field. Considering that the seasonal variations [44] inf luence folate levels. We reexamined the folate levels in PKM-1 and AV fruits in the next season. Though the folate level in the PKM-1 was slightly lower than in the earlier season, it was significantly higher than AV (PKM-1 40.4 ± 1.7 μg/100 g FW; AV 18.1 ± 2.0 μg/100 g FW) ( Figure 2A). As PKM-1 fruits had ≥2-fold higher folate levels than AV, notwithstanding seasonal differences, we compared these two cultivars to decipher the underlying basis of variations in folate levels. To minimize the inf luence of environmental factors on plant and fruit growth, we grew both PKM-1 and AV in the greenhouse for all following experiments.

Fruit ripening in AV and PKM-1
The PKM-1 fruits were slightly smaller than AV; nonetheless, the on-vine growth of these fruits was nearly identical, reaching the near-final size 30 days after pollination (DAP) ( Figure 2B). Both AV and PKM-1 fruits reached the mature green (MG) stage at the same time (32 DAP) ( Figure 2C-D). However, post-MG stage PKM-1 fruits reached the breaker (BR) stage after seven days compared to 4 days for AV. Post-BR, both AV and PKM-1 fruits attained the red-ripe (RR) stage in 8-10 days ( Figure 2D). In tomato, ripening is also associated with increased sugar levels monitored as the • Brix value. The • Brix value in RR PKM-1 fruits was significantly higher than AV ( Figure 2E).
A major difference in PKM-1 and AV fruits was substantially lower ethylene emission from PKM-1 fruits ( Figure 3A). Both at BR and RR stage, the ethylene levels in PKM-1 were nearly 40% of AV. We then examined whether PKM-1 differed from AV regarding other phytohormones ( Figure 3B-H). Like ethylene, abscisic acid (ABA) level was nearly half in PKM-1 at all ripening stages. PKM-1 also had a lower level of other phytohormones in a stage-specific fashion, jasmonic acid (JA) at MG, BR; salicylic acid (SA) at BR, RR; indole acetic acid (IAA) at BR. While indole-butyric acid (IBA) and methyl jasmonate (MeJA) did not vary significantly, RR fruits of PKM-1 had a higher level of zeatin at RR than AV.

Folate levels during ripening of fruits
Though greenhouse-grown PKM-1 fruits had lower folate levels than field-grown fruits (Figure 2A), the folate level in PKM-1 was higher than AV at all ripening stages ( Figure 4A). During ripening, the folate level steadily declined, with the decline being higher in AV than in PKM-1, a pattern also seen in field-grown fruits. Among the four detected folate vitamers, viz. THF, 5-CH 3 -THF, 5,10-CH + THF, and 5-CHO-THF, THF were seen only at the MG at trace levels ( Figure S1 A-D). 5-CH 3 -THF was the major vitamer constituting 60-80% of folate, and 5,10-CH + THF and 5-CHO-THF contributed to the rest. The ripening-associated decline in total folate was solely contributed by 5-CH 3 -THF. The other two vitamers showed a marginal increase, particularly from MG to BR stage, barring 5-CHO-THF in AV. The 5-CHO-THF content was significantly high in PKM-1 than in AV at all stages of ripening.
We next examined whether the higher folate levels in PKM-1 are limited to fruits or seen in leaves. The PKM-1 leaves showed higher folate levels than AV ( Figure S2). The folate level in leaves was nearly 8-fold higher than in the RR fruits. In leaves too only four viz. THF, 5-CH 3 -THF, 5,10-CH + THF, and 5-CHO-THF folate vitamers were detected. In contrast to fruits, leaves showed a substantially higher level of THF.

pABA levels are higher in PKM-1 fruits
We next examined whether higher folate levels in PKM-1 fruits arose due to increased precursors of folate biosynthesis, namely pABA and pterin. Parallel to higher folate levels, pABA was higher in PKM-1 fruits at all ripening stages. In AV and PKM-1, pABA levels increased from BR to RR stage ( Figure 4B). In contrast to pABA, we could not detect the pterins viz. 7-8-dihydroneopterin triphosphate and dihydroneopterin, the first and second intermediates of folate biosynthesis emanating from GTP. Presumably, their levels were lower than the detection limit.
However, we detected HMPt (6-hydroxymethylpteridine), an oxidized product of HMDHP, which is considered part of the pterin pool contributing to folate biosynthesis [34]. Contrarily, the other detected pteridine was 6-carboxypterin (p6C), a catabolite of pteridines and folates. Compared to AV, the level of HMPt was about 1/3rd in PKM-1 fruits at MG and BR ( Figure 4C). Remarkably, in both cultivars at the RR stage, the level of HMPt was below the detection limit. Contrasting to HMPt, the levels of p6C were nearly similar in PKM-1 and AV MG fruits. ( Figure 4D). In AV, the levels of  p6C declined from MG to RR, while in PKM-1, the decline was from BR to RR stage.

RR fruits of PKM-1 show lower GGH enzyme activity and higher polyglutamylation
Contrasting to higher folate levels, the transcript levels of most genes encoding for folate biosynthesis in PKM-1 were nearly similar to the AV ( Figure S3). The exceptions were DHNA at the BR stage, DPP, DHFR, and HPPK-DHPS at the RR stage, with lower transcript levels in PKM-1. The transcript levels of DHFR were high in PKM-1 at the MG stage ( Figure S3). Interestingly, the transcript levels of deglutamylation genes-GGH1, GGH2, and GGH3 in PKM-1 were significantly lower than AV at the RR stage ( Figure 4E-G). The in vitro assay of the GGH enzyme showed that both in PKM-1 and AV, the activity of GGH was highest at the MG and declined with ripening progression ( Figure 4H). While GGH activity did not significantly differ at MG, at BR, GGH activity was significantly high in PKM-1, whereas at RR, the GGH activity was significantly low in PKM-1 fruits.
It is believed that polyglutamylated folate vitamers are less susceptible to catabolism than monoglutamylated forms. Also, folate-dependent enzymes prefer polyglutamylated folates over monoglutamate forms [35]. The glutamylation status of folate vitamers was monitored with or without rat plasma conjugase at the RR stage to quantify the relative monoglutamate and polyglutamate levels. PKM-1 fruit has significantly higher polyglutamylation than the AV ( Figure 4I).

RR fruits of PKM-1 show higher carotenoids levels
Since PKM-1 has higher folate levels, we checked whether the carotenoid levels were also affected in its fruits. Interestingly, PKM-1 RR fruits also had higher carotenoid levels than AV. Lycopene and β-carotene were 1.4 and 1.7-fold higher in PKM-1 fruits ( Figure 5). During ripening, the carotenoid profiling revealed a stage-specific upregulation in PKM-1, neoxanthin at MG and BR, violaxanthin at BR, δ-carotene at RR, and phytoene at RR and lutein at BR and RR stages (Table S1). Interestingly, γ -carotene was detected solely in RR PKM-1 fruits. While lycopene levels were very low at the MG and BR stage, and β-carotene levels increased during ripening.
We next analyzed the transcript levels of major genes involved in carotenoid biosynthesis ( Figure S4). The transcript levels of most genes were nearly similar between AV and PKM-1 at the MG and BR stages. Surprisingly, most genes contributing to lycopene formation at the RR stage showed lower transcript levels in PKM-1 than AV, including DXS, GGPPS2, PSY1, PDS, ZISO, ZDS, ZEP, NCED, and CYP97A genes. At the MG stage, CYCB and LCYB1 transcripts were high, whereas PSY1, ZISO, NCED, and CYPC11 transcripts were low in PKM-1. At the BR stage, transcript levels of PDS and LCYE were significantly low in PKM-1.

Metabolic profiles of PKM-1 fruits considerably differ from AV
Since AV and PKM-1 fruits markedly differed in folate level, a regulator of C1 metabolism, we examined whether metabolic profiles manifested similar differences. Out of 69 metabolites detected by GC-MS, most metabolite levels in PKM-1 substantially differed from AV. The metabolites included amino acids, amines, organic acids, fatty acids, sugars, and derivatives ( Figure 6A) (Dataset S1). The principal component analysis (PCA) displayed that at all stages of ripening, AV and PKM-1 metabolites were clustered separately ( Figure 6B). To delineate how changes in metabolite levels may alter cellular homeostasis, the significantly different metabolites of PKM-1 at different stages of ripening were mapped onto a general biochemical pathway ( Figure 6C).
One distinct feature of the PKM-1 metabolome was elevated levels of most amino acids. Out of 20 L-amino acids, constituents of proteins, the levels of 13 amino acids were higher in PKM-1. In contrast to amino acids, sugars and sugar-derived metabolites showed a variable increase in PKM-1. In the glycolysis pathway, the glucose-6 phosphate (MG, BR) and fructose-6 phosphate (BR) were higher in PKM-1. Compared to sugars, TCA cycle intermediates-malate and succinate were lower at RR and BR stages, respectively.

Proteome analysis in PKM-1 fruits
Considering the broad-spectrum effect of PKM-1 on carotenoids, folate, and metabolome, we profiled the proteome of AV and PKM-1 fruits. The proteome profiling of the AV and the PKM-1 revealed around 3309 (MG), 2940 (BR), and 2485 (RR) proteins in AV and 2570 (MG), 2383 (BR), and 2785 (RR) proteins in the PKM-1 cultivar ( Figure S5). The comparision of proteins (Single hit peptides) between MG, BR and RR of AV and PKM-1 showed while a majority of proteins were shared, cultivar had proteins had unique stage-specific proteins ( Figure S6). Among the proteins, we considered only those upregulated or downregulated proteins that had a minimum of two peptide hits, the log2 fold values ≥ ± 0.584, and P-value ≤0.05. Label-free quantification of proteins identified 493, 539, and 685 differentially expressed proteins (log2 fold ±0.58, P-value ≤0.05) in PKM-1 at MG, BR, and RR, respectively. At MG and RR, most proteins were upregulated (270↑, 223↓, MG; 369↑, 316↓, RR), while in BR, most were downregulated (261↑, 278↓, BR) ( Figure 7A) (Dataset S2 and Dataset S3). The complement of up/down-regulated proteins varied in a stage-specific fashion, and only a few proteins overlapped at two or more stages.
The GO classification and Mapman analysis of differentially expressed proteins are in conformity with the wide-ranging inf luence of PKM-1 on cellular homeostasis ( Figure 7B, Figure S5-7, Dataset S4). The largest classes of differentially expressed proteins belonged to protein homeostasis and biosynthesis. The higher • Brix in PKM-1 RR fruits correlated with the upregulation of sucrose phosphate synthase (Solyc07g007790). In PKM-1 RR fruits, the levels of carotenoids sequestering proteins such as harpin binding protein 1 (Solyc09g090330), CHRC, and PAP3 (Solyc02g081170, Solyc08g076480) were upregulated ( Figure 7C). In contrast, no specific accumulation of proteins leading to carotenoids biosynthesis or metabolism was discernible, except for the upregulation of MEP-pathway enzyme 4-hydroxy-3methylbut-2-enyl diphosphate reductase (HDR) at RR.
Several proteins contributing to the primary precursors for amino acid biosynthesis were upregulated in PKM-1. The higher activity of aconitase (Solyc07g052350, Solyc12g005860 BR, RR) may direct TCA cycle f lux towards glutamate, GABA (γ -aminobutyric acid), glutamine, and pyroglutamic acid. Likewise, f lux through glycolysis is also directed towards higher amino acid levels. The high levels of phosphoglycerate kinase (Solyc07g066600 BR; Solyc07g066610 MG, BR, RR) may boost phosphoglyceric acid levels, the precursor for serine, glycine, and cysteine. The high levels of phoshopyruvate hydratase (Solyc09g009020, BR; Solyc10g085550 MG, BR, RR; Solyc03g114500 MG) may boost pyruvate levels leading to high levels of leucine, valine, and alanine. The level of 3-isopropylmalate dehydratase (Solyc06g060790), a protein in the leucine biosynthesis pathway, was upregulated in MG Conversely, γ -aminobutyrate transaminase I protein (Solyc07g043310) was downregulated in PKM-1 MG fruit.
The C1 metabolism was altered with ripening stage-specific expression of different components of methionine metabolism and folate transformations ( Figure 7D). Consistent with reduced ethylene emission from PKM-1 fruits, the levels of key ripening associated ethylene biosynthesis enzyme, 1-aminocyclopropane carboxylic acid synthase (ACO) 1 (Solyc07g049530), ACO6 (Solyc02g036350), and also of ripening associated protein E8 (Solyc09g089580) were downregulated ( Figure 7D). The reduced levels of S-adenosylmethionine synthase isoforms may have added to reduced ethylene emission.

Genome sequencing
To decipher the differences between AV and PKM-1, we sequenced and compared the whole genome of both cultivars. On the whole genome scale, PKM-1 had 350 316 SNPs that were not present in Arka Vikas, indicating that the PKM-1 genome differed by 0.0423% (Table S2). Notably, no polymorphic differences were found in genes encoding the folate and carotenoid biosynthesis pathways. The marking of deleterious mutations present in PKM-1 on tomato metabolic pathways revealed that only a few pathways were affected in PKM-1 ( Figure S8). Since most SNPs were localized in the noncoding regions of the PKM-1 genome, it is likely that the polymorphism in promoters of regulatory genes contributes to observed differences between Arka Vikas and PKM-1 [29].

Discussion
Tomato cultivars considerably differ in their • Brix, carotenoids, and folate levels [30,44]. Compared to AV, the PKM-1 had higher • Brix, carotenoids, and folate levels in RR fruits. Considering that little polymorphism exists in genes regulating carotenoids and folate levels, it was suggested that variation in their levels manifests interaction between transcriptome, proteome, and metabolome at cellular homeostasis level [30,44]. The comparative analysis of AV and PKM-1 is in conformity with the above notion for the regulation of • Brix, carotenoids, and folate levels.

Several proteins modulating amino acid levels are altered in PKM-1
The paramount indicator of the alteration in cellular homeostasis is the levels of metabolites and proteins linked with glycolysis and the TCA cycle [39]. The higher • Brix and sucrose in PKM-1 fruits seem to be linked to increased sucrose synthase protein. While there were only a few stage-specific changes in levels of glycolysis and TCA cycle intermediates, a major change in PKM-1 was the elevation of a large number of amino acids barring those derived from the shikimate branch. In conformity with higher amino acid levels, two key proteins of the glycolysis pathway were upregulated in PKM-1. Among these, phosphoglycerate kinase provides precursor for serine, glycine, and cysteine, and phoshopyruvate hydratase provides a precursor for leucine, valine, and alanine.
Higher levels of aconitase, a TCA cycle protein, may be related to the elevation of glutamate, GABA, glutamine, and pyroglutamic acid. Additionally, the increased levels of glutamine synthase and glutamate synthase proteins may contribute to the above pathway. The reduced level of GABA transaminase I in PKM-1 may contribute to higher GABA levels, as the silencing of the above gene increases GABA in tomato fruits [23]. Considering that these amino acids are derived from common intermediates in glycolysis and the TCA cycle, it is logical to assume that their higher levels in PKM-1 may relate to the upregulation of these proteins.

Reduced ACO levels likely contribute to lower ethylene emission
One of the modulators of cellular homeostasis is hormones. The variations in hormonal levels have a cascading inf luence on plant metabolome and development [8]. The lower levels of ABA, SA, and MeJA indicate a major alteration in the hormonal profile of PKM-1 fruits. In tomato, ethylene is needed for fruit ripening, as the block in ethylene perception or biosynthesis negatively affects ripening [36]. The PKM-1 fruits emitted substantially lower ethylene than AV, which seems to be related to lower amounts of ACO1 and ACO6, two key proteins contributing to ripeningspecific ethylene synthesis [18]. The reduced ABA, SA, and JA levels seem to be related to reduced ethylene emission in PKM-1. A similar lowering of ABA, SA, and JA is seen in the acs2-2 mutant compromised in ethylene biosynthesis [36].
It can be construed that the alteration of the hormonal profile in PKM-1 may be responsible for the observed metabolic differences between PKM-1 and AV. Nonetheless, the linkage between a metabolite and hormones and developmental response is complex [8]. In tomato, the RNAi-suppression of invertase (LIN5) reduces • Brix levels, leads to smaller fruits, and lowers sucrose synthase activity and levels of ABA and JA [50]. Contrarily, in PKM-1, higher • Brix is associated with increased sucrose synthase protein but reduced ABA and JA levels. Despite lower ethylene emission, the transition period of PKM-1 fruits from the MG to RR stage was only moderately longer. Probably, the ethylene emission from PKM-1 fruit was still sufficient; therefore, ripening duration remained unaffected.

Increased sequestration may be responsible for the elevation of carotenoids levels
The upregulation of carotenoids during tomato ripening is inf luenced by many causative factors with a complex web of interactions [26]. Among these factors, ethylene is a major factor in modulating fruit-specific carotenogenesis. The transgenic suppression of ethylene biosynthesis or loss of ethylene perception in Nr mutant inhibits carotenoid accumulation. Besides ethylene, ABA and JA also stimulate carotenoid accumulation in tomato [26]. Considering that PKM-1 fruits, despite having reduced levels of ethylene, ABA, and JA, accumulate higher levels of carotenoids, the upregulation of carotenoids seems unrelated to the hormonal profile of fruits. Likewise, the tomato Nps1 mutant has higher Note that the total area of the pie is proportional to the total carotenoid amount. The carotenoid data are means ± SE (n ≥ 4). See Table S1 for individual carotenoid levels and significance. carotenoid levels but reduced emission of ethylene and ABA levels [20].
The upregulation of carotenoids level in tomato fruits has also been ascribed to increased expression of carotenoid biosynthesis pathway genes, particularly phytoene synthase 1 [26]. The lower or nearly similar transcript level of most carotenoid biosynthesis genes in PKM-1 indicates that the stimulation of carotenogenesis is not related to transcript levels. The possibility that reduced transformation of carotenoids to xanthophylls may boost lycopene levels is unlikely, as there was no reduction in xanthophylls in PKM-1.
The enhanced level of carotenoids in PKM-1 fruits seems to be more related to efficient sequestration by carotenoid binding proteins [46]. Emerging evidences support that fibrillin family proteins enable a higher accumulation of carotenoids in chromoplasts. In tomato, high pigment mutants increase CHRC protein, which likely boosts carotenoid levels by sequestration [21]. The overexpression of pepper fibrillin increases carotenoid levels in tomato fruits [38]. Contrarily, VIGS of fibrillins reduce lycopene accumulation in tomato fruits [33]. In pepper, the harpin binding protein assists the formation of lycopene crystals [7]. The increased level of lycopene in PKM-1 may be related to increased levels of carotenoid sequestration proteins, viz., harpin binding protein 1, and CHRC and PAP3. Ostensibly, an increase in carotenoid sequestration proteins seems to be the causative factor for higher lycopene in PKM-1 fruits rather than modulation of carotenoid biosynthesis or degradation,

Elevation of folate mildly affects C1 metabolism in PKM-1 fruits
The interrelationship between folate levels and tomato ripening seems to be more complex. In PKM-1 and AV, post-MG THF was not detectable, perhaps due to lower folate biosynthesis or faster conversion of THF to other vitamers. In most tomato cultivars, The metabolic shifts in PKM-1 fruits during ripening compared to AV. The relative changes in the metabolite levels at different ripening stages in PKM-1 fruits were determined by calculating the PKM-1/AV ratio at respective ripening phases. Only significantly changed metabolites are depicted on the metabolic pathway (log2 fold ≥ ± 0.584 and P-value ≤0.05). Bold black letters indicate the identified metabolites. Grey letters indicate metabolites below the detection limit. The white box indicates no significant change in metabolite level. Data are means ± SE (n ≥ 5), P ≤ 0.05. See Dataset S1 for detailed metabolites data.
folate levels decline during ripening, and only a few cultivars show increase during the ripening [44]. In PKM-1 and AV, total folate levels declined during ripening, albeit in PKM-1, the decline was slower. Among three-folate vitamers detected in ripening fruits, the decline was confined to 5-methyl-THF, the major vitamers in leaf and fruit. The reduction in 5-methyl-THF may be related to the increased levels of methionine synthase in PKM-1. Additionally, variable expressions of MTHFR, SHMT, and formatetetrahydrofolate ligase were observed at different ripening stages. Taken together, the elevated folate levels in PKM-1 only mildly inf luenced the C1 metabolism.
Notwithstanding the above, the emerging evidence has brought forth a novel role for 5-CHO-THF in plants. So far, the CHO-THF is considered an intermediate in the C1 metabolism, which does not serve as a C1 donor. Rather it binds to SHMT and other folatedependent enzymes and inhibits their enzymic activity. Recent evidence indicated that 5-CHO-THF binds to many proteins and can modulate C/N metabolism [25]. Considering that 5-CHO-THF levels are higher in PKM-1, it may inf luence C1 metabolism and the C/N metabolism, signified by elevated amino acid levels in PKM-1.

Elevated folate level in PKM-1 is not related to the expression of folate biosynthesis genes
The higher folate level in PKM-1 may ref lect a higher synthesis rate or reduced degradation of folate vitamers. The transgenic manipulation of pterin and pABA biosynthesis in tomato fruits indicated that a combined upregulation of both precursors is needed for increased folate levels [5]. In PKM-1 fruits, the pABA levels were higher than in AV. In contrast, the neopterin level was below the limit of detection; however, downstream to neopterin, HMPt had about 1/3rd level in PKM-1 than in AV. In potato, the drop in pterin levels represented by HMPt was considered an enhancement of metabolic f lux for folate biosynthesis [6]. It is plausible that the low HMPt level in PKM-1 ref lects its more efficient utilization of pterin for folate biosynthesis. Considering that HMPt is detected only at MG and BR stages, this may be related to higher folate levels in PKM-1 at these stages.
In addition to biosynthesis, the endogenous folate level is also determined by the rate of degradation. In plants, the rate of folate breakdown is ca. 10% per day [41]. The non-enzymatic cleavage of the C9-N10 bond in folate yields dihydropterin-6-aldehyde and pABA-Glutamate. As folate is important for metabolism, plants have a folate salvage pathway that regenerates pterin and pABA from these breakdown products. It is plausible that PKM-2 has a faster conversion of dihydropterin-6-aldehyde to folate precursor HMDHP [9], as ref lected by lower HMPt levels. Taken together, it remains possible that either efficient utilization of HMPt for folate synthesis or reduced folate turnover in PKM-1 contributes to higher folate levels at MG and BR stages.
In tomato, including PKM-1, the leaves have several-fold higher folate levels than fruits [43]. It can be construed that folate biosynthesis is attenuated in fruits than leaves. Consistent with this, the transcript levels of key folate biosynthesis genes ADCS and GCHI decline during tomato ripening, and other genes show variable expression [48]. In PKM-1, transcript levels of folate biosynthesis genes show variable expression at different ripening stages and thus had little correlation with folate biosynthesis. Ostensibly, high folate level in PKM-1 seems to be mediated by a process different from the transcriptional upregulation of folate biosynthesis pathway genes.

Reduced GGH activity likely contributes to higher folate levels in PKM-1 fruits
In addition to biosynthesis, the endogenous folate level is also determined by degradation and the extent of polyglutamylation. In plants, the rate of folate breakdown is ca. 10% per day [41]. The non-enzymatic cleavage of the C9-N10 bond in folate yields dihydropterin-6-aldehyde and pABA-Glutamate. As folate is important for metabolism, plants have a folate salvage pathway that regenerates pterin and pABA from these breakdown products. The reduced levels of 6-hydroxymethylpteridine (HMPt) in PKM-1 may be due to its faster conversion to folate precursor [9]. Alternately, the pterin aldehyde reductase (PTAR) in PKM-1 may be less efficient, as manifested by higher levels of pterin-6carboxylate in ripening fruits. Together, the above two processes may determine the extent of recycling the pterin to the folate biosynthesis pathway.
The in vivo folate level can be altered by shortening or lengthening polyglutamyl tails. The relative activities of FPGS, which adds glutamate tail to folate, determine the extent of folate polyglutamylation. Conversely, the activities of GGHs that remove glutamate tail also inf luence the in vivo folate levels. In tomato, the overexpression of GGH resulted in extensive deglutamylation of folate and lowered the total folate by 40% [1]. Considering that in PKM-1, the polyglutamate percentage in folate is significantly high, indicating reduced activity of GGH. In conformity with this, PKM-1 has lower GGH1, GGH2, and GGH3 expressions and reduced GGH enzyme activity. Similarly, in tomato tf-5 mutant, the increased folate levels in fruits strongly correlated with the reduced GGH activity [42]. Taken together, the high folate levels in PKM-1 seem to be related to lowering the deglutamylation process as it has lower GGH activity in fruits.

SNPs in promoter regions likely determine metabolic diversity
While there were polymorphic differences between both cultivars' genome sequences, only 830 genes in PKM-1 bore SNPs, possibly leading to loss of function (Table S2, Dataset S5). Since most of the above genes belong to large gene families, their loss may be offset by other family members. The genome sequence comparison highlighted that observed differences in carotenoids and folate levels in these cultivars are not due to polymorphic differences in genes regulating folate and carotenoid biosynthesis. Since most SNPs reside in noncoding regions, it is reasonable to assume that the diversity of metabolic changes in PKM-1 compared to AV results from polymorphic differences in promoter regions. In potato, out of 497 SNPs inf luencing folate level, a lone SNP was located within the 5-formyltetrahydrofolate cycloligase gene, and only 17 SNPs were in proximity of promoter/enhancer regions of folate metabolism genes [2]. The SNPs in promoters can affect gene expression by modifying cis-regulatory elements and binding of transcription factors, as indicated for acs2-2 mutant in tomato [36]. Broadly, polymorphic differences in promoters may inf luence transcriptional regulations of genes leading to alteration of cellular homeostasis, including higher folate and carotenoid levels in fruits.
To sum, our study shows that PKM-1 RR fruits have elevated levels of three nutraceuticals, carotenoids, GABA, and folate. The metabolome and proteome profiling also divulge the diversity in regulating these three nutraceuticals in PKM-1 fruits. The carotenoid sequestration proteins seem to be a key regulator for increased carotenoid levels. Conversely, lowering of degradation of GABA and reduced deglutamylation of folate seems to be key loci for higher GABA and folate, respectively, in tomato. In conclusion, this study provides the potential gene targets for improving carotenoids, GABA, and folate in tomato.

Plant materials and growth conditions
Tomato (S. lycopersicum) cultivar AV (Arka Vikas) and PKM-1 (Periyakulam-1) were obtained from IIHR, Bangalore, India (www. iihr.res.in). The AV and PKM-1 were grown in the greenhouse as described in Bodanapu et al. [4]. The first two f lowers from the first two trusses were tagged. The fruits were harvested at the mature green (MG), breaker (BR), and red-ripe (RR) stage. The fruits were homogenized in liquid nitrogen using a homogenizer (IKA, A11 basic, Germany), and the powder was stored at −80 • C till further use.

Biochemical analysis
The protocols used for different analyses were as follows: Gupta et al. [17] for • Brix; Gupta et al. [16] for carotenoids; Kilambi et al. [21] for ethylene emission; Tyagi et al. [43] for folate; GGH enzyme assay [42] and Bodanapu et al. [4] for phytohormones and metabolites. Only metabolites with a ≥ 1.5-fold change (log2 ≥ ± 0.584) and P-value ≤0.05 were mapped on the metabolic pathway. Total RNA was extracted from the pericarp tissue of AV and PKM-1 fruits at different ripening stages using the hot phenol method [45], and qRT-PCR was done as described previously by Kilambi et al. [21]. The qRT-PCR primers for folate and carotenoids biosynthesis pathway genes were designed using Primer 3 software (Table S3-4).

Glutamylation, pABA, and pteridines extraction and quantifications
For quantifying glutamylation, all steps were similar to the folate extraction [43], except the step of rat plasma treatment was omitted. The pABA was extracted following Navarrete et al. [31] protocol. Brief ly, 100 mg powdered tissue was homogenized in 1 mL methanol and centrifuged at 2600 g for 15 min at 4 • C, the supernatant was collected, and the residue was re-extracted. The supernatants were pooled and dried by centrifugal evaporation. The dried sample was suspended in 1 mL Milli-Q water, followed by 5 min sonication. To 400 μL of the sample, 50 μL 2 N HCl was added, and the mixture was incubated at 80 • C for two hrs. The sample was cooled, and 2 N NaOH was added to neutralize the acid. Lastly, the sample solutions were ultra-filtrated at 12000 g for 12 min before LC-MS analysis, and 7.5 μL of the sample was injected into the column.
The pABA was separated on a reverse-phase Luna C18 column (5 μm particle size, 150 mm x 4.60 mm I.D.) (Phenomenex, CA, USA) using a gradient elution program. The gradient comprised of a binary solvent system consisting of 0.1% (v/v) of formic acid in water (solvent A) and methanol (solvent B) at a f low rate of 1 mL min −1 . The injection volume was 7.5 μL, the run time was 15 min, and the column temperature was set at 24 • C. The starting condition of the gradient was 92% of solvent A and 8% of solvent B. Subsequently, solvent B was linearly increased to 50% in 7 min, then to 100% in 10 min. After that, the mobile phase was reverted to the initial condition in 1 min and held for 4 min for re-equilibration of the column before the next injection. For all experiments, Exactive™Plus Orbitrap mass spectrometer (Thermo Fisher Scientific, MA, USA) was operated in the alternating full scan and equipped with positive heated electrospray ionization (ESI). The source parameters were optimized as follows: ion spray voltage, 4 kV; capillary temperature, 380 • C; and heater temperature, 400 • C. Sheath, auxiliary, and sweep gas f lows were 75, 20, and 2 arbitrary units (au), respectively. For the full scan experiment, the mass scan was in the range of m/z 50-350. Instrument control and data processing were carried out by Xcalibur 3.0 software (Thermo Fisher Scientific). The pABA was detected in the positive mode with m/z 138.05 at the retention time of 5.0 min.
Pteridines were extracted following Martín-Tornero et al. [27] protocol with some modifications. About 400 mg powdered tissue was homogenized in 1 mL methanol/water pH 12.0 (1/1, v/v) mixture, sonicated for 15 min, and centrifuged at 5000 g for 10 min. The supernatant was collected, and the sample was re-extracted. Both supernatants were combined and dried by centrifugal evaporation. The dried residue was dissolved in 400 μL Milli-Q water. Lastly, 100 μL of sample solutions were ultra-filtrated at 12000 g for 12 min before LC-MS analysis. The 7.5 μL sample was injected into the column.
Pteridines were separated using a gradient elution program on a reverse-phase Hypersil Gold C18 column (1.9 μm particle size, 50 mm x 2.1 mm I.D.) (Thermo Fisher Scientific). The gradient consists of a binary solvent system consisting of 0.1% (v/v) of formic acid in water (solvent A) and acetonitrile (solvent B) at a f low rate of 300 μL min −1 . The injection volume was 7.5 μL, the run time was 8 min, and the column temperature was set at 24 • C. For all experiments, Exactive™Plus Orbitrap mass spectrometer (Thermo Fisher Scientific) was operated in the alternating full scan and equipped with positive heated electrospray ionization (ESI). The source parameters were optimized as follows: ion spray voltage, 4 kV; capillary temperature, 380 • C; and heater temperature, 320 • C. Sheath, auxiliary, and sweep gas f lows were 60, 20, and 2 arbitrary units (au), respectively. For the full scan experiment, the mass scan was in the range of m/z 50-500. Pteridines were detected in the positive ion mode with m/z for Npt (254.08), Mpt (254.08), HMPt, (194.06), and P6C (208.04) at the retention time of 2.63, 3.33, 5.57, and 5.68 respectively. Instrument control and data processing were carried out by Xcalibur 3.0 software (Thermo Fisher Scientific).

PCA and metabolic pathway in AV and PKM-1
Principal Component Analysis (PCA) of primary metabolites was performed using Metaboanalyst 4.0 (http://www.metaboanalyst. ca/) after the One-way ANOVA plot (P ≤ 0.05) identified 66 of the 69 metabolites having significant contributions. The metabolites were mapped on the general metabolic pathways described in the KEGG (Kyoto Encyclopedia of Genes and Genomes, http://www. genome.jp/kegg) and LycoCyc (Sol Genomic networks, http:// www.solcyc.solgenomics.net).

Genome sequencing
Whole Genome Sequencing and data analysis of AV and PKM-1 were done as described in Gupta et al. [14]. Brief ly, sequencing was performed on the HiSeqX sequencing system (Illumina) (GeneWiz Inc. NJ, USA) according to the manufacturer's protocol. A total of ∼200 million reads (31.5 Gb) with Q30 > 29 Gb data was generated. The raw reads were filtered using fastp software (v0.19.5) using parameters -M 30-3 -5. The 2X 150 bp reads were mapped on S. lycopersicum cv. Heinz version SL3.0 using BWA-MEM (0.7.17) [24]. The data analysis and variant calling was performed as described in Gupta et al. [13]. The resulting vcf (variant calling format) files were annotated using the SIFT4G algorithm. The effect of base substitutions on protein function was determined by SIFT4G (SIFT score ≤ 0.05 is considered deleterious) using the SIFT4G-ITAG3.2 genome reference database generated by Gupta et al. [13].