Functional features of TREHALOSE-6-PHOSPHATE SYNTHASE1 - an essential enzyme in Arabidopsis thaliana.

In Arabidopsis, TREHALOSE-6-PHOSPHATE SYNTHASE 1 (TPS1) catalyzes the synthesis of the sucrose-signaling metabolite trehalose 6-phosphate (Tre6P) and is essential for embryogenesis and normal post-embryonic growth and development. To understand its molecular functions, we transformed the embryo-lethal tps1-1 null mutant with various forms of TPS1 and with a heterologous TPS (OtsA) from Escherichia coli, under the control of the TPS1 promoter, and tested for complementation. TPS1 protein localized predominantly in the phloem-loading zone and guard cells in leaves, root vasculature and shoot apical meristem, implicating it in both local and systemic signaling of sucrose status. The protein is targeted mainly to the nucleus. Restoring Tre6P synthesis was both necessary and sufficient to rescue the tps1-1 mutant through embryogenesis. However, post-embryonic growth and the sucrose-Tre6P relationship were disrupted in some complementation lines. A point mutation (A119W) in the catalytic domain or truncating the C-terminal domain of TPS1 severely compromised growth. Despite having high Tre6P levels, these plants never flowered, possibly because Tre6P signaling was disrupted by two unidentified disaccharide-monophosphates that appeared in these plants. The non-catalytic domains of TPS1 ensure its targeting to the correct subcellular compartment and its catalytic fidelity, and are required for appropriate signaling of sucrose status by Tre6P.

Tre6P is the phosphorylated intermediate in the two-step pathway of trehalose biosynthesis (Cabib and Leloir, 1958); it is synthesized from UDP-glucose and glucose 6-phosphate (Glc6P) by TPS, then dephosphorylated to trehalose by trehalose-6-phosphate phosphatase (TPP; EC 3.1.3.12). The importance, indeed the very existence, of this pathway in flowering plants has only emerged over the last 20 years. The unexpected finding of genes encoding functional TPS (TPS1; Blázquez et al., 1998) and TPP (TPPA and TPPB; Vogel et al., 1998)

enzymes in
Arabidopsis (Arabidopsis thaliana) was followed by the discovery that loss-of-function mutants of TPS1 are non-viable, with homozygous tps1 embryos failing to complete embryogenesis, becoming arrested at the torpedo stage (Eastmond et al., 2002). Arrested embryos have fewer cells, indicating defective cell division, as well as abnormalities in their cell wall structure and starch content (Eastmond et al., 2002;Gómez et al., 2006). However, the underlying cause of the embryo arrest phenotype has not been established. Viable tps1 seeds can be obtained by dexamethasone-inducible expression of TPS1 during seed development (Van Dijken et al., 2004) or by embryo-specific expression of TPS1 under the control of the ABSCISIC ACID INSENSITIVE3 (ABI3) promoter (Gómez et al., 2010).
Nonetheless, the resulting tps1 plants are severely dwarfed and either do not flower or flower very late (Van Dijken et al., 2004;Gómez et al., 2010;Wahl et al., 2013), showing that a functional TPS1 is needed for normal growth and development throughout the plant life cycle.
A key breakthrough came with the discovery that increasing or decreasing the levels of Tre6P in Arabidopsis, by constitutive expression of Escherichia coli TPS (35Spro:otsA) or TPP (35Spro:otsB), led to profound but opposite changes in plant morphology (Schluepmann et al., 2003). 35Spro:otsA plants had small leaves, precocious flowering and a bushy phenotype, whereas 35Spro:otsB plants had large leaves, delayed flowering and only one or a few shoot branches. These experiments revealed the potent influence of Tre6P on plant growth and development. The levels of Tre6P in plants were found to be highly correlated with sucrose (Lunn et al., 2006), which led to the proposal that Tre6P functions as a signal of sucrose status.
Elaborating on this idea, our current sucrose-Tre6P nexus model postulates that Tre6P is not only a signal of sucrose status, but also a negative feedback regulator of sucrose levels, acting in a way that is reminiscent of the homeostatic control of blood glucose levels in animals by insulin (Yadav et al., 2014). Sucrose dominates the metabolism of flowering plants; it is the major product of photosynthesis, the most common transport sugar, and the main source of carbon and energy in growing sink organs (Lunn, 2016). This dominance may explain why Tre6P, acting as a signal and regulator of sucrose levels, can exert such a far-reaching influence on plant growth and development . In sucrose-producing source leaves, Tre6P regulates sucrose levels by modulating photoassimilate partitioning during the day  and the mobilization of transitory starch reserves at night (Martins et al., 2013;dos Anjos et al., 2018). In sucrose-consuming sink organs, Tre6P regulates the utilization of sucrose for growth and accumulation of storage products, acting, at least in part, via inhibition of SUCROSE-NON-FERMENTING1-RELATED KINASE1 (SnRK1) (Zhang et al., 2009;Nunes et al., 2013;Zhai et al., 2018). Although the reciprocal regulation of sucrose and Tre6P appears to operate in somewhat different ways in source and sink tissues, the sucrose-Tre6P nexus model offers a unifying concept for the fundamental role of Tre6P in plants .
The recognition of Tre6P as a potent regulator of plant growth and development suggested that the arrest of tps1 embryos at the torpedo stage might be due to impaired synthesis of Tre6P. However, in common with other flowering plants, Arabidopsis has a large family of TPS genes encoding TPS or TPS-like proteins (Leyman et al., 2001;Avonce et al., 2006;Lunn, 2007). The 11 TPS genes in Arabidopsis form two distinct clades: class I (TPS1-TPS4) and class II (TPS5-TPS11). With the exception of TPS3, which is most likely a pseudogene, only class I TPS genes have been reproducibly shown to encode catalytically active TPS enzymes (Blázquez et al., 1998;Van Dijck et al., 2002;Vandesteene et al., 2010;Delorge et al., 2014). At the moment, the functions of the non-catalytic class II TPS-like proteins are poorly understood (Harthill et al., 2006;Ramon et al., 2009;Lunn et al., 2014).
TPS2 and TPS4 are almost exclusively expressed in developing seeds (Vandesteene et al., 2010). However, despite being catalytically active , the encoded enzymes do not compensate for the loss of TPS1 in developing tps1 mutant seeds. It is unclear whether this is due to differences in their spatio-temporal expression patterns, i.e. TPS2 and TPS4 are not expressed in the right place at the right time in developing seeds to substitute for TPS1, or if the TPS1 protein has unique properties that the other two class I TPS isoforms lack (reviewed in . In this context, it is worth noting that two TPP genes in maize (Zea mays) -RAMOSA3 (RA3) and ZmTPP4 -were recently found to encode catalytically active TPP enzymes that also have non-catalytic functions in regulating inflorescence branching (Satoh-Nagasawa et al., 2006;Claeys et al., 2019), setting a precedent for enzymes in this pathway to have "moonlighting" functions.
The Arabidopsis TPS1 protein has three distinct domains: (i) an N-terminal domain that includes a putative auto-inhibitory Leu/Arg-rich motif, (ii) a glucosyltransferase domain containing the catalytic site, and (iii) a TPP-like C-terminal domain of unknown function (Figure 1B;Van Dijck et al., 2002;Lunn, 2007). The other two catalytically active isoforms, TPS2 and TPS4, also have glucosyltransferase and C-terminal TPP-like domains, but lack an equivalent of the TPS1 N-terminal domain. Curiously, such truncated class I TPS proteins are found only in the Brassicaceae, the significance of which is unknown (Lunn, 2007). As the predominant Tre6P-synthesizing enzyme in Arabidopsis, TPS1 might be expected to play a major role in determining the levels of Tre6P, but how the enzyme is regulated and contributes to the sucrose-Tre6P nexus is poorly understood (Yadav et al., 2014).
Surprisingly little is known about the cellular and subcellular localization of TPS1, given that it is essential for plant viability. Microarray analysis has shown that TPS1 is expressed in all major tissues -leaves, roots and reproductive organs (Supplemental Figure 1; Schmid et al., 2005) -but less is known about its expression in specific cell types. TPS1 transcripts are relatively abundant in bundle sheath cells in Arabidopsis, based on ribosome pull-down experiments (Aubry et al., 2014). RNA in situ hybridization also demonstrated transcript accumulation in the proto-vasculature of developing leaves and in the shoot apical meristem (SAM) (Wahl et al., 2013). TPS1 protein was detected in a proteomic analysis of Arabidopsis guard cells (Zhao et al., 2008). In cucumber (Cucumis sativus), the orthologous CsTPS1 protein was found in phloem sap exudates (Hu et al., 2016). In several previous studies in Arabidopsis, genomic regions of 2 to 3 kbp upstream of the TPS1 protein-coding region were used to drive expression of a TPS1 cDNA or bacterial otsA in tps1 mutant complementation experiments, or expression of a GUS reporter gene for promoter analysis (Schluepmann et al., 2003;Van Dijken et al., 2004;Gómez et al., 2010;Vandesteene et al., 2010). However, the presumed promoter region used in those studies was chosen on the basis of an incompletely annotated version of the TPS1 locus (At1g78580) that missed the first exon, which only encodes part of the 5´-UTR and lies 3.2 kbp upstream of the first protein-coding exon ( Figure 1A). Thus, the above-mentioned studies used a section of intron 1 and part of the 5´-UTR (exon 2) to drive TPS or GUS gene expression in their complementation and promoter analysis constructs, so the introduced genes almost certainly had spatio-temporal expression patterns that were different from the endogenous TPS1 gene. There are also conflicting reports on the subcellular compartmentation of TPS1. The SUBA4 database (Hooper et al., 2017;suba.live) shows no consensus from sequence-based prediction tools for the intracellular location of TPS1, but notes that TPS1 has been experimentally detected in plasma-membrane enriched fractions (Mitra et al., 2009), although the protein has no obvious transmembrane domains.
By contrast, transient expression of a construct encoding GFP-tagged TPS1 in Arabidopsis protoplasts indicated that the protein is predominantly cytosolic, although some is located in the nucleus (Vandesteene et al., 2010).
In this study we explored the elements and functional features of Arabidopsis TPS1 to understand why this enzyme, from a rather obscure branch of sugar metabolism, has the power to determine plant fate. We generated complementation lines of the Arabidopsis tps1-1 null mutant with GUS-or GFP-tagged versions of the TPS1 protein, expressed under the control of the endogenous TPS1 promoter and other potential gene regulatory elements, to elucidate the tissue and cell-specific accumulation pattern of the TPS1 protein and its subcellular compartmentation. We used a similar strategy to investigate the functions of individual domains and residues within TPS1 and to identify potential moonlighting functions, by attempting to complement the tps1-1 mutant with various truncated or mutated versions of TPS1, or with a heterologous TPS (OtsA) enzyme from E. coli. Our results reveal that TPS1, and by inference Tre6P synthesis, predominantly localizes to guard cells and around the phloem-loading zone in source leaves, a strategically important site in source-sink relations and for systemic signaling. We show that loss of the Tre6P-synthesizing capacity of TPS1 is the fundamental reason why tps1-1 mutant embryos arrest at the torpedo stage and is a major, but not necessarily the only, factor underlying their post-embryonic growth defects. We observed that the N-terminal domain of TPS1 regulates its distribution between nucleus and cytosol, and that TPS1 misbehaves if the C-terminal TPP-like domain is missing, with catastrophic consequences for the plant.

Constructs for Expression of Wild-Type and Mutated Forms of TPS Proteins
The gene constructs used to test for complementation of the tps1-1 mutant were all derived from the Arabidopsis Columbia-0 (Col-0) TPS1 genomic sequence, based on the current annotation of the TPS1 locus (At1g78580) in the Araport11 database (https://www.araport.org/data/araport11; Cheng et al., 2017). All constructs ( Figure 1C-F) included the entire intragenic region (~ 3.5 kbp) between the transcription start site of TPS1 and the end of the coding region of the neighboring NADH KINASE3 (NADK3, At1g78590) gene. This upstream region is presumed to contain the true TPS1 promoter region (TPS1pro).
In addition, the constructs contained the 5´-UTR (exon 1 + intron 1 + part of exon 2), and a large region (~. 1,000 bp) downstream of the translation stop codon containing the 3´-UTR (part of exon 18) and terminator region (TPS1term), including part of the neighboring RHAMNOSE BIOSYNTHESIS1 (RHM1, At1g78570) gene where the polyadenylation site of TPS1 is located. Constructs for expression of full-length versions of the TPS1 gene (native or mutated) also contained all of the endogenous TPS1 intronic sequences, and introns were retained in the protein-coding regions that were included for expression of truncated forms of TPS1 ( Figure 1A). By using the endogenous TPS1 promoter and other potential gene regulatory elements, we expected the introduced transgenes to have the same spatiotemporal expression pattern and expression level as the wild type TPS1 gene. As a positive control for complementation, we used the full-length TPS1 gene with no modifications (TPS1; Figure 1C).
The first group of constructs encoded full-length TPS1 proteins tagged at the N-or Cterminus with either GFP (GFP-TPS1 and TPS1-GFP) or GUS (GUS-TPS1 and TPS1-GUS), to determine the cellular and subcellular localization of TPS1 ( Figure 1C). To investigate the potential role of TPS1 in the nucleus (Vandesteene et al., 2010), we also generated constructs encoding TPS1 (± C-terminal GFP) with a strong nuclear localization signal (NLS) from the simian vacuolating virus 40 (SV40) added at the N-terminus (SV40-TPS1 and SV40-TPS1-GFP; Figure 1C). The second group of constructs ( Figure 1D Figure 1E) encoded full-length TPS1 proteins with substitutions of one or more amino acid residues: (i) L27P, expected to abolish the auto-inhibitory function (Van Dijck et al., 2002); (ii) A119W, expected to compromise catalytic activity (Vandesteene et al., 2010); (iii) S252A, removing a putative phosphorylation site (Glinski and Weckwerth, 2005); (iv) S252D, to mimic a constitutively phosphorylated Ser252; and (v) R369A/K374A/E476A, targeting a triad of active site residues that are highly conserved in TPS and related glucosyltransferases, and expected to abolish catalytic activity .
The final construct contained the E. coli otsA gene ( Figure 1F), encoding a simple form of TPS (OtsA) with only a single catalytic glucosyltransferase domain and no obvious regulatory potential if expressed in plants. This construct was designed to test whether the tps1-1 mutant could be complemented by replacing only the Tre6P synthesizing activity of TPS1.

Complementation of the Arabidopsis tps1-1 Null Mutant
All constructs carried the neomycin phosphotransferase II (nptII) gene in the T-DNA and conferred kanamycin resistance. They were introduced into glufosinate-resistant heterozygous TPS1/tps1-1 plants (Eastmond et al., 2002) by floral dip transformation (Clough and Bent, 1998). We selected primary transformants for dual resistance to kanamycin and glufosinate, allowed them to self, and screened the T2 progeny for homozygosity of the tps1-1 locus by genomic PCR (Supplemental Figures 2A and 3A). Immunoblotting with antibodies raised against TPS1 (α-TPS1; Yadav et al., 2014) confirmed the absence of the endogenous full-length (106 kDa) TPS1 protein and, where possible, expression of introduced TPS1 proteins.
With a single exception, we obtained at least one viable line that was homozygous for the tps1-1 mutant allele (tps1-1/tps1-1) for each construct, demonstrating complementation of the tps1-1 null mutation during embryogenesis. The only exception was the mutated catalytic triad construct: TPS1[R369A, K374A,E476A]. We screened the T 2 progeny from two independent lines for this construct by genomic PCR, after selection for plants resistant to both kanamycin and glufosinate; all 36 out of 36 plants from one line and all 60 out of 60 plants from the other line were heterozygous (TPS1/tps1-1) for the tps1-1 allele. The failure to recover any TPS1[R369A,K374A,E476A] plants with a homozygous tps1-1/tps1-1 background indicated that the catalytically inactive form of TPS1 encoded by this construct was unable to rescue the tps1-1 mutant through embryogenesis. For constructs that gave rise to an aberrant phenotype, we confirmed the phenotype in at least two independent lines, or in independent lines expressing the same modified TPS1 protein with or without a GFP tag (e.g. TPS1[ΔNΔC]-GFP and TPS1 [ΔNΔC]). The only exception was TPS N , for which we obtained a single line that showed only a very mild phenotype. Supplemental Figure 4 summarizes all analyzed tps1-1 complementation lines.
Two independent TPS1 lines, complemented with the wild-type TPS1 gene, were indistinguishable from wild-type plants, with no significant differences in rosette diameter  Figure 6). This indicated that the TPS1 genomic construct, from which all other constructs were derived, fully complemented the tps1-1 mutant. Complementation lines with constructs expressing GUS-or GFP-tagged versions of the full-length TPS1 gene also showed no obvious differences from wild-type Col-0 or TPS1-complemented plants (Supplemental Figures 2C and 4). The SV40-TPS1 (±GFP) and OtsA lines had wild type-like rosettes (Supplemental Figure 4) and flowered at the same time as Col-0 and TPS1 plants ( Figure 2B), but had much shorter primary roots (Supplemental ]. This construct was able to complement the tps1-1 mutant to yield viable seeds. Although dwarfed and slow growing, the resulting plants were eventually able to flower ( Figure 3). These observations showed that removal of the C-terminal TPP-like domain or substitution of Ala119 by Trp has a particularly deleterious effect on the functionality of the TPS1 protein, but this is less severe if the N-terminal domain of the protein is also removed.
Given the severe phenotypes associated with the absence of the C-terminal domain of TPS1, we examined the TPS1 sequence in more detail to find possible clues as to its physiological function. We identified two predicted phosphorylation sites that have been experimentally confirmed in planta: Ser826 (S822 R P S (pS) D S G A K831) and Ser941 (L935 A D T T S (pS) P942), the latter being seen in two independent studies (Wang et al., 2013;Roitinger et al., 2015). The C-terminal domain also contains a putative sumoylation site on Lys902 with a very high confidence score (p = 0.005) according to the SUMOSP prediction tool (Zhao et al., 2014;

Tissue and Cellular Localization of Arabidopsis TPS1
We grew seedlings and 2-week-old plants of the GUS-TPS1 and TPS1-GUS lines in longday conditions and harvested them 10 h after dawn for detection of GUS activity. We observed the same expression patterns in both GUS-TPS1 ( Figure 4A) and TPS1-GUS lines ( Figure 4B), and the results were consistent in multiple independent lines for each construct.
Representative images from two independent N-terminal and two independent C-terminal GUS-fusion lines are shown in Figure 4 and Supplemental Figure 7. The GUS-tagged TPS1 fusion protein was predominantly located in the leaf and root vasculature in seedlings and 2- week old plants ( Figure 4A-I), and in guard cells ( Figure 4J). In transverse sections of the fourth true leaf from the GUS-TPS1 lines, the GUS-TPS1 fusion proteins were primarily located in the phloem tissue, especially in sieve elements and companion cells, with some weaker staining in the phloem parenchyma (bundle sheath cells) around the vascular bundle and in xylem parenchyma cells ( Figure 4K-L). GUS-tagged TPS1 was also detected at the shoot apex ( Figure   4C), consistent with previous RNA in situ hybridization data showing expression in the flanks of the SAM of vegetative plants (Wahl et al., 2013). In shoot apices from young, non-flowering plants, GFP-TPS1 localized to the peripheral and rib zones of the SAM ( Figure 5C), and in the subtending vasculature (Supplemental Figure 9A). In inflorescence shoot apices, GFP-TPS1 was detected in floral primordia but not in the inflorescence meristem itself (Supplemental Figure 10). In the roots, TPS1 was found in companion cells and sieve elements of the phloem, but not in the root apical meristem ( Figure 4D-F).
We also looked in reproductive tissues of older GUS-TPS1 and TPS1-GUS plants. In  Figure 7L-N). We detected no GUS activity in developing seeds prior to the globular stage of embryo development, but GUStagged TPS1 was detected in embryos from the mid to late globular stage (Supplemental Figure 7O), with the strongest expression in the middle zone of the globular embryo, which develops into the vascular tissue of the mature embryo. Tannin accumulation in the seed coat rendered the tissue impermeable to the X-Gluc substrate, preventing us from detecting GUStagged TPS1 after the globular stage. We detected GUS activity in mature seeds (Supplemental Figure 7P), although it was only visible in the chalazal endosperm due to the tannin accumulation in the seed coat. However, after germination and rupture of the seed coat, GUS-tagged TPS1 was detected throughout the seed (Supplemental Figure 7Q), suggesting that TPS1 was probably already present in all of the seed tissues during the later stages of seed maturation. In mature Arabidopsis embryos, we observed a GUS signal in the radicle, the vasculature and close to the SAM (Supplemental Figure 7R).

Subcellular Compartmentation of Arabidopsis TPS1
For microscopic analysis of the GFP-tagged TPS1 lines, we focused on tissues where a TPS1 signal would be strong and easily visible based on our GUS-tagged TPS1 lines, i.e. guard cells, seedling roots and shoot apices. In guard cells, GFP strictly localized to the nuclei in both GFP-TPS1 and TPS1-GFP lines, with no apparent GFP signal outside the nucleus ( Figure 5A).
As the GFP signal was relatively weak compared to the autofluorescence of chlorophyll in the chloroplasts, we imaged the non-tagged TPS1 line under the same conditions as a negative control ( Figure 5A). We confirmed the nuclear localization of GFP-tagged TPS1 in guard cells by microscopic analysis of independently grown plants and staining of nuclei with 4′,6diamidino-2-phenylindole (DAPI), which showed overlap of the GFP and DAPI fluorescence signals in the nucleus (Supplemental Figure 8). In seedling roots, we only observed GFP-TPS1 in the nuclei of phloem companion cells and, more diffusely, in sieve elements, consistent with a cytosolic location in the latter ( Figure 5B). In both non-flowering and inflorescence shoot apices, the GFP signal had a punctate appearance, consistent with a predominantly nuclear localization ( Figure  The N-terminal domain of Arabidopsis TPS1 contains a seven-amino-acid motif -L27 R E K R K S33 -that resembles the proposed consensus sequence for monopartite NLSs - (Chelsky et al., 1989), and matches perfectly with a shorter K R K motif sufficient for nuclear localization (Kosugi et al., 2009). We also identified the K30 R K32 motif of TPS1 as a putative NLS signal using the NLStradamus prediction tool (

Day and Equinoctial Conditions
Tre6P exerts a powerful influence not only on plant growth and development, but also on carbon and nitrogen metabolism . Therefore, we investigated Col-0 and a TPS1-complemented line as controls, and also harvested 10 h after dawn (ZT10) but on different days, such that the plants from all genotypes were sampled at the same developmental stage (7-8 fully expanded leaves).
Tre6P levels in wild type Col-0 rosettes ranged from 0.1 to 0.2 nmol g -1 FW ( Figure 6A-B), consistent with previous reports from plants grown under similar conditions (Lunn et al., 2006;Martins et al., 2013;Yadav et al., 2014;. Tre6P levels were within the same range in all of complementation lines except for TPS1[ C] and TPS1[A119W] ( Figure   6A-B). We had expected the A119W mutation to compromise the catalytic activity of the TPS1 protein, based on the failure of TPS1[A119W] to restore growth of the yeast tps mutant on glucose-containing media (Vandesteene et al., 2010). However, the TPS1[A119W] lines (0.50 nmol g -1 FW) had more than twice as much Tre6P as Col-0 and TPS1 control plants ( Figure 6B), demonstrating that the A119W mutation did not abolish catalytic activity. The  Figure 11A). Aconitate and 2-oxoglutarate were also significantly increased in the OtsA line (Supplemental Figure 11A). We observed the most extreme differences from wild type in the TPS1[A119W] and TPS1[ C] lines. Both lines had elevated levels of most organic acids, with fumarate being increased up to 20-fold, while fructose-1,6bisphosphate (Fru1,6bP), mannose 6-phosphate (Man6P), 3-phosphoglycerate (3-PGA) and phosphoenolpyruvate (PEP) were significantly lower than in the controls (Supplemental Figure 11B).
We also analyzed individual amino acids in some of the lines. The TPS1[ N C] line had elevated levels of Glu, Gln, Ala, Ser, Gly, Asn, Arg, Met, His and Phe (Supplemental Figure 12).
Almost all of the measured amino acids were also increased in the TPS1[ C] lines, and to a slightly lesser extent in the TPS1[A119W] lines (Supplemental Figure 12). By contrast, amino acid levels in the TPS1[ N] line were not significantly different from wild type, and only Gly was slightly increased in the OtsA line (Supplemental Figure 12).
To get an overview of the metabolic similarities and differences between the lines, we performed a hierarchical clustering analysis on the lines with the most comprehensive metabolite profiling, and visualized differences as a heatmap ( Figure 6C). analysis. There were no significant differences between wild-type and TPS1 line #2 plants in their Tre6P and sucrose contents ( Figure 6D), and these were similar to the respective values seen in previous experiments ( Figure 6A-B). The TPS1[ΔNΔC] line had moderately elevated sucrose levels and a lower Tre6P: sucrose ratio than wild-type plants ( Figure 6D), as previously observed ( Figure 6A). Similarly, the TPS1[ΔN A11 W] lines had moderately high sucrose and a lower Tre6P: sucrose ratio than wild-type plants ( Figure 6D). In contrast, TPS1[ΔC] line #7 had exceptionally high Tre6P and sucrose levels and a low Tre6P: sucrose ratio ( Figure 6D), confirming previous results ( Figure 6B). We saw a very similar pattern in the TPS1[ΔC895-942] plants ( Figure 6D). We confirmed the strong similarity between the TPS1[ΔC895-942] and We also grew some of the lines under equinoctial conditions (12 h photoperiod), to see if further metabolic differences emerged when the plants are more carbon-limited than in long-day conditions. We harvested rosettes from 4-week old plants at the end of the day (ED) and at the end of the night (EN) for metabolite analysis (Supplemental Figure 14). As seen already in long-day conditions, the dwarfed TPS1[ C] and TPS1[A119W] lines were the most divergent. Their Tre6P levels were mostly not significantly different to wild-type or only moderately increased, but they accumulated up to 30-times more sucrose than wild type Col-0 and TPS1 control plants and therefore had lower Tre6P: sucrose ratios (Supplemental Figure   14B). The two dwarfed lines also had more starch at the end of the day and a starch excess at the end of the night (Supplemental Figure 15A)  Figure   7B). The dwarfed TPS1[ C] and TPS1[A119W] lines again showed the greatest differences from the wild type Col-0 controls and other lines. They had 3-5 times more Tre6P and up to 30 times more sucrose than Col-0 plants throughout the diurnal cycle, giving them a very low Tre6P: sucrose ratio ( Figure 7A). There was no significant correlation between Tre6P and sucrose in these two lines ( Figure 7B).
The diurnal profiles of other sugars, phosphorylated intermediates, organic acids and starch were essentially identical in the Col-0 and TPS1 plants (Supplemental Figures 15B and   16), and very similar in the SV40-TPS1 and TPS1[L27P] lines (Supplemental Figures 16A and   16B). The TPS1[ N] line had more trehalose and slightly elevated levels of TCA cycle intermediates, especially at night, but was otherwise similar to Col-0 (Supplemental Figure   16C). The TPS1[ N C] line showed greater differences, with generally higher levels of phosphorylated intermediates and lower TCA cycle intermediates (Supplemental Figure 16D).
The OtsA line had higher Suc6P at night and lower fumarate during the day, but was otherwise very similar to Col-0 (Supplemental Figure 16E). In the dwarfed TPS1[ΔC] and TPS1 [A119W] lines, almost all metabolites, including starch (Supplemental Figure 15B), were significantly higher than in Col-0 throughout the diurnal cycle, except for pyruvate, which was 2-5 times lower in the two complementation lines compared to Col-0 (Supplemental Figures 16F and   16G).

Response of Tre6P to Sucrose in Carbon-Starved Seedlings
In three of the tps1-1 complementation lines -OtsA, TPS1[ C] and TPS1[A119W] -we observed that Tre6P did not follow the endogenous diurnal fluctuations in sucrose levels ( Figure 7), suggesting that the sucrose-Tre6P nexus relationship was broken in these lines. To investigate this further, we tested whether Tre6P levels changed when sucrose was supplied exogenously to carbon (C)-starved seedlings. As observed in previous studies (Lunn et al., 2006;Yadav et al., 2014), Tre6P was very low in C-starved Col-0 seedlings but, after a short lag, increased rapidly after sucrose addition (  Figure 17B), as it showed a greater increase in PEP after sucrose was supplied, but smaller increases in most TCA cycle intermediates: citrate, aconitate, isocitrate, 2-oxoglutarate, succinate and malate.

Monophosphates
To measure Tre6P, we used an LC-MS/MS-based assay that gives a baseline resolution of Tre6P and its most common isomer in plants, Suc6P, the intermediate of sucrose biosynthesis (Lunn et al., 2006). In addition to Tre6P and Suc6P, we have also observed several other compounds with the mass spectral properties of disaccharide-monophosphates in plant extracts. These include abundant levels of maltose 1-phosphate in orange (Citrus sinensis) leaves infected with a bacterial pathogen (Xanthomonas citri subsp. citri, Piazza et al., 2015), and trace amounts of three other isomeric molecules in Arabidopsis plants (Supplemental Figure 18), the identities of which are so far unknown (discussed in . We noted that the levels of two of these unknowns, which elute just before and just after Suc6P (Supplemental Figure Figure 18). In common with Tre6P, these compounds strongly increased after sucrose addition to C-starved TPS1[ C] seedlings (Supplemental Figure 18E).

DISCUSSION
Our study aimed to answer five questions: (1) why is TPS1 essential for embryogenesis in Arabidopsis? (2) in which tissues and cell types is TPS1 located? (3) where in the cell is TPS1 protein localized? (4) what are the functions of the non-catalytic domains of the TPS1 protein?
and (5) how does TPS1 contribute to the sucrose-Tre6P nexus?

The Essential Role of TPS1 in Embryogenesis
The defective embryogenesis in Arabidopsis tps1 null mutants (Eastmond et al., 2002) was one of the first discoveries that demonstrated the importance of the trehalose biosynthetic pathway for plant growth and development. As a result, what had once been just an obscure branch of sugar metabolism, of little interest except to dedicated sugar biochemists, began to receive much greater attention. Subsequent discoveries revealed how ubiquitous the influence of trehalose metabolism is; perturbation of this pathway leads to alterations in carbon and nitrogen metabolism, leaf development, flowering time, shoot branching, stomatal function, and tolerance of abiotic and biotic stresses, not only in Arabidopsis, but also in other plant species (reviewed in . The primary defect in tps1-1 null mutants appeared to be a lack of Tre6P, based on reports that the tps1-1 mutant could be rescued by expression of the heterologous E. coli otsA gene under the control of the presumed Arabidopsis TPS1 promoter (Schluepmann et al., 2003;Van Dijken et al., 2004). However, the complementation construct used in those studies to drive otsA expression was based on an incomplete annotation of the TPS1 locus, such that the presumed TPS1 promoter was in reality part of the first intron and 5´-UTR (as indicated in Figure 1A). It is therefore surprising that the otsA gene was expressed at all in this construct, but one possible explanation is that otsA expression was driven by read-through transcription from the constitutive promoter of the selectable marker gene in the T-DNA. Whatever the mechanism driving otsA expression in those previous studies, its spatio-temporal expression pattern would have differed from that of the endogenous TPS1 gene. Furthermore, there was minimal documentation of the phenotypes of the otsA complementation lines in the published reports, leaving many questions unanswered. Previous examination of arrested tps1 embryos revealed defects in cell division, cell wall biosynthesis and accumulation of storage reserves (Eastmond et al., 2002;Gómez et al., 2006), but a fundamental question remained: is the principal defect in the tps1 mutants a lack of Tre6P synthesis in specific cells at a crucial stage of embryo development, or is it linked to some non-catalytic function of TPS1, or both?  (Figures 6 and 7). This showed that the A119W mutation does not abolish the catalytic activity of TPS1, even though it severely compromises TPS1 protein functionality in planta and, by inference, when expressed heterologously in yeast. All other mutated or truncated forms of TPS1 that were able to complement the tps1-1 mutant also retained catalytic activity, as the respective complementation lines had either normal or elevated Tre6P levels (Figures 6 and 7). The dependence of tps1-1 complementation on restoration of Tre6P synthesis was corroborated by the failure of the TPS1[R369A,K374A,E476A] construct, encoding a catalytically inactive TPS1, to rescue the tps1-1 mutant through embryogenesis. We conclude that the capacity to synthesize Tre6P during embryogenesis is both necessary and sufficient to rescue the tps1-1 mutant during seed development.
Tre6P appears to be dispensable in developing seeds until the torpedo stage when tps1-1 embryos arrest (Eastmond et al., 2002). The torpedo stage of embryo development coincides with cellularization of the peripheral endosperm surrounding it, a process that is regulated by auxin (Batista et al., 2019). A major function of the cellularized endosperm is to nourish the embryo during later stages of development. Based on microarray data, the TPS1 gene is expressed in the embryo throughout seed development, and is strongly up-regulated in the peripheral endosperm at the torpedo stage (Schmid et al., 2005;. We might speculate that a lack of Tre6P in the embryo and/or peripheral endosperm at the torpedo stage disrupts cellularization of the endosperm, perhaps by perturbing auxin synthesis (Meitzel et al., 2019) or signaling, and thereby prevents the endosperm from fulfilling its function to nourish the developing embryo, leading to eventual embryo arrest.
Other functional class I genes, TPS2 and TPS4, are expressed primarily in the chalazal endosperm (Schmid et al., 2005;. Thus, we conclude that, despite being catalytically active and expressed in developing seeds, the inability of the TPS2 and TPS4 isoforms to compensate for loss of TPS1 in tps1 mutants is simply due to their lack of expression in the right cells at the right time to substitute for TPS1.

Tissue and Cellular Localization of TPS1
From transcriptomic analysis, TPS1 mRNA is expressed in all major organs of the plant In leaves, TPS1 predominantly accumulated in and around the vascular bundles and in guard cells (Figures 4 and 5). The latter confirms the detection of TPS1 in the guard cell proteome (Zhao et al., 2008), and corroborates the defective stomatal phenotypes of the tps1-12 mutant, which carries a weak but non-embryo-lethal tps1 allele (Gómez et al., 2010).
The functions of TPS1 and Tre6P in guard cells are not yet understood; they appear to have an essential, but so far undefined, role in ABA signaling (Gómez et al., 2010), and are also implicated in regulation of various aspects of guard cell carbon metabolism (reviewed in Santelia and Lunn, 2017).
In the leaf vasculature, TPS1 was seen in the phloem parenchyma (bundle sheath) cells around the vascular bundles, as well as in the companion cells and sieve elements of the phloem itself ( Figure 4K-L). One of the main functions of the phloem is to translocate sucrose from source leaves to sink organs. Sucrose is produced by photosynthesis in the mesophyll cells during the day and by mobilization of transitory starch reserves at night. In Arabidopsis, sucrose is actively loaded from the apoplast into the phloem for export to growing sink organs ( Figure 9A; Haritatos et al., 2000). Sucrose diffuses via plasmodesmata from mesophyll cells into phloem parenchyma cells surrounding vascular bundles, where it is released into the apoplast by two SUCROSE WILL EVENTUALLY BE EXPORTED (SWEET) type sugar transporters: SWEET11 and SWEET12 (Chen et al., 2012). Sucrose is then actively taken up from the apoplast into companion cells and sieve elements by SUT(SUC)-type sucrose-H + symporters.
Thus, TPS1 is located on both sides of the apoplastic barrier in the phloem-loading zone in the leaves, which essentially represents the interface between source and sink tissues, and is potentially a highly strategic site for signaling between source and sink ( Figure 9A). For example, if supply of sucrose from the leaves exceeds the demand from growing sink organs, sucrose will accumulate in the leaves. Rising sucrose levels in leaf phloem parenchyma cells will lead to an increase in Tre6P, which can diffuse symplastically, via plasmodesmata, into mesophyll cells ( Figure 9A) to shift partitioning of photoassimilates away from sucrose  or inhibit transitory starch turnover (Martins et al., 2013;dos Anjos et al., 2018), thereby restoring the balance between sucrose supply and demand.
In companion cells, TPS1 is ideally situated to monitor the uptake and availability of sucrose for export to sink organs and modulate Tre6P levels in the phloem accordingly. Why TPS1 localizes to the nucleus remains to be elucidated. The pores in the nuclear membrane are large enough to allow free movement of UDP-glucose, Glc6P, Tre6P and UDP between the nucleus and cytosol, so there is no obstacle to Tre6P being synthesized in the nucleus and subsequently moving to the cytosol, or vice versa. It is well documented that changes in sucrose content affect transcript levels of thousands of genes (Bläsing et al., 2005;Osuna et al., 2007;Cookson et al., 2016). As a proxy for sucrose status, Tre6P is potentially involved in transcriptional regulation of gene expression by sucrose, and synthesis of Tre6P directly in the nucleus might allow more precise control over gene expression. The possibility that TPS1 itself binds to and modulates transcriptional regulators cannot be excluded. A further possibility is that TPS1 activity is regulated by association with other proteins that are located in the nucleus or the cytosol ( Figure 9C). The failure of OtsA to fully complement the tps1-1 mutant suggests that such protein-protein interactions are not essential during embryogenesis, but might be more important at later stages in development.
Phloem sieve elements are usually enucleate and have few ribosomes, and thus are dependent on companion cells for mRNA and protein synthesis. In vascular bundles of the SV40-TPS1-GFP line, GFP signal was observed only in the nuclei of phloem companion cells, with no detectable signal in the sieve elements ( Figure 5B). This indicated that the heterologous (SV40) NLS targeted TPS1 more strongly to the nucleus than the putative endogenous NLS (K30 R K32) on its own, essentially trapping SV40-TPS1-GFP protein in the companion cell nuclei. Presumably, SV40-TPS1-GFP mRNA produced in companion cells would still be able to move into sieve elements, but if SV40-TPS1-GFP transcripts were translated by the few ribosomes present in the sieve elements, this would yield too little SV40-TPS1-GFP protein to be detected. Thus, we conclude that TPS1, when made in companion cells, is primarily targeted to the nucleus. We speculate that the endogenous monopartite NLS in TPS1 is not strong enough to target all of the protein to the nuclei of companion cell, potentially allowing some of the TPS1 protein to move symplastically into sieve elements via plasmodesmata. The SV40-TPS1-GFP line had shorter roots than wild-type Col-0 plants (Supplemental Figure 5) but, unlike the TPS1[ C]-GFP line, the structure of the vascular bundles in the SV40-TPS1-GFP line was not obviously disrupted ( Figure 5B).
Therefore, we conclude that cytosolic TPS1 in sieve elements does have a purpose, as loss of TPS1 in these cells impairs root growth, and that this is linked to the functioning of differentiated phloem rather than the development of the vasculature in roots.

Functions of the Non-Catalytic Domains of TPS1
The three-domain structure of TPS1 ( Figure 1B) has ancient roots; class I TPS proteins in chlorophyte green algae also have the same three domains, as do orthologous proteins from bryophytes, lycophytes, ferns and gymnosperms (Lunn, 2007). The retention of this three-domain structure throughout plant evolution indicates that it might have some purpose. Curiously, the analogous enzyme of sucrose biosynthesis, sucrose-phosphate synthase (SPS; EC 2.4.1.14), has a C-terminal domain that resembles the next enzyme in that pathway, sucrose-phosphate phosphatase (SPP; EC 3.1.3.24), but has no catalytic activity (Lunn et al., 2000), reminiscent of the TPP-like C-terminal domain of TPS1.
In addition to its putative monopartite NLS described above ( Figure 9B), the Nterminal domain of TPS1 has been proposed to have an auto-inhibitory function, based on complementation assays in yeast (Van Dijck et al., 2002). This putative function is 1) linked to a Leu/Arg-rich motif (R20 L R D R E L R28) and 2) is disrupted by substitution of Leu27 by Pro (Van Dijck et al., 2002). We observed slightly elevated Tre6P in the TPS1  (Heazlewood et al., 2008;http://phosphat.uni-hohenheim.de/) indicated that phosphorylation of Ser252 has never been experimentally observed in Arabidopsis plants, and it is also worth noting that this residue is absent from the other class I TPS isoforms in Arabidopsis TPS2 and TPS4, and is not universally conserved in TPS1 orthologues in other plant species, especially not in the grasses (Poaceae). Thus, we conclude that Ser252 does not play a major role in regulation of TPS1 activity in vivo.  6 and 7), a trait that is usually associated with early flowering (Schuepmann et al., 2003) and low sucrose content (Yadav et al., 2014). However, neither of these lines flowered and they both accumulated unprecedentedly high levels of sucrose, indicating that increased Tre6P is not the only, or even the prime, reason for their phenotypic defects. Curiously, in both cases, the phenotypes were less severe when the N-terminal domain was also removed. These observations demonstrate that all three domains of TPS1 are required for its full functionality, and indicate that there are complex interactions between the three domains ( Figure 9C).  (Figures 3 and 6), in which both sites were removed. Further site-directed mutagenesis experiments will be required to confirm these putative regulatory sites and investigate their individual physiological significance.
It has long been known that some enzymes are prone to catalytic infidelity, due to substrate promiscuity or mistakes during catalysis. Ribulose-1,5-bisphosphate carboxylase oxygenase (Rubisco; EC 4.1.1.39) is a classic example in plants, not only showing high activity with the wrong substrate -oxygen -but also generating a range of by-products that can act as dead-end inhibitors of the enzyme and need to be removed by the specialized repair enzyme Rubisco activase (Parry et al., 2008). In recent years, it has been recognized that many other plant enzymes make mistakes during catalysis (Hanson et al., 2016). It is intriguing that (Supplemental Figure 18), and their developmental and metabolic phenotypes were correspondingly milder.
A speculative explanation for the increased levels of the two unknown compounds in these lines is that the catalytic fidelity of the TPS1 enzyme is compromised by the absence of the C-terminal domain or by the presence of the A119W point mutation in the catalytic domain. Errors during catalysis may give rise to unnatural Tre6P isomers (e.g. α,β-1,1-Tre6P or β,β-1,1-Tre6P), instead of the authentic product, α,α-1,1-Tre6P. Alternatively, binding of inappropriate substrates, such as UDP-galactose, in the active site could generate heterosidic analogues of Tre6P (e.g. α,α-1,1-galactosyl-glucose 6-phosphate). Such stereo-isomers and analogues of Tre6P could be potent ligands for Tre6P-binding proteins, thereby disrupting Tre6P signaling pathways. Our data provide circumstantial evidence that Tre6P signaling is having elevated levels of Tre6P, which would otherwise be expected to trigger early flowering (Schluepmann et al., 2003). In general the severity of their phenotypes was correlated with the levels of the two unknown compounds. Production of such aberrant Tre6P analogues might explain why Arabidopsis TPS1[A119W] could not complement the yeast tps mutant, even though this form of the enzyme retains the capacity for Tre6P synthesis (see above).
Further studies are needed to establish the identities of the two unknown compounds and determine if they have any physiological role in modulating Tre6P signaling in wild-type plants.

Role of TPS1 in the Sucrose-Tre6P Nexus
The sucrose-Tre6P nexus model was originally founded on two observations: (i) Tre6P levels are highly correlated with sucrose during diurnal fluctuations in rosettes and in response to exogenous sucrose supplied to seedlings (Lunn et al., 2006), and (ii) imposed changes in the level of Tre6P, by over-expression of TPS or TPP, lead to opposite effects on sucrose levels (Yadav et al., 2014). It is poorly understood how changes in sucrose content lead to parallel changes in the level of Tre6P. Inhibitor studies suggested that the Tre6P response to sucrose is not dependent on de novo transcription, but depends on translation (Yadav et al., 2014). However, polysome-loading and immunoblotting experiments showed that sucrose does not up-regulate production of TPS1 itself, so it was proposed that sucrose might induce synthesis of a putative regulatory protein that activates TPS1 (Yadav et al., 2014). Phosphorylation and other post-translational modifications are obvious potential mechanisms for regulation of TPS1 by other proteins. Our results indicate that phosphorylation of Ser252 is not a major factor in regulation of TPS1, but the putative phosphorylation (Ser826 and Ser941) and sumoylation (K902) sites in the C-terminal domain of TPS1 ( Figure 9B) are worth investigating for their regulatory potential in future experiments.
Our results show that the catalytic activity of TPS1 is strongly influenced by its Cterminal TPP-like domain, which retains some of the active site residues involved in substrate binding in catalytically active TPPs, but not the critical Asp residue that forms the phosphoacyl intermediate during catalysis (Lunn, 2007). Suc6P phosphatase (SPP) catalyzes the analogous reaction to TPP in the pathway of sucrose biosynthesis. SPP is competitively inhibited by sucrose, and even more strongly by trehalose, both of which bind to a glucose-binding site in the "cap" domain of the enzyme (Fieulaine et al., 2005;2007). By analogy, we speculate that sucrose might bind to the TPP-like domain of TPS1 via a similar mechanism and activate TPS activity allosterically or by promoting interaction with a putative protein activator (Protein X; Figure 9C; Yadav et al., 2014). There is also potential for Tre6P to inhibit TPS1 in a competitive manner by binding in the active site, or allosterically by binding to the TPP-like domain ( Figure   9C). In the latter scenario, binding of a putative inhibitory protein (Protein Y; Figure 9C) in the presence of Tre6P is a further possibility. Activity and ligand-binding assays of recombinant TPS1, including truncated or mutated versions, will be required to explore these hypothetical mechanisms for regulation of TPS1.
In most of the tps1-1 complementation lines transformed with variants of TPS1, Tre6P tracked the endogenous fluctuations in leaf sucrose levels during the diurnal cycle, and these two metabolites were highly correlated (Figure 7). The two exceptions were the TPS1[ C] and TPS1[A119W] lines, where Tre6P and sucrose were poorly correlated. The tight linkage between sucrose and Tre6P also appeared to be broken in leaves of OtsA-transformed plants, with Tre6P barely changing during the light-dark cycle, while the diurnal fluctuation in sucrose was slightly greater than in wild-type plants ( Figure 7A). The greater accumulation of sucrose in rosettes of OtsA lines during the day could be indicative of less sucrose export to the roots, potentially explaining the poor root growth of these plants (Supplemental Figure 5). Another possibility is that the expression of otsA in the roots, instead of TPS1, disrupts the operation of the sucrose-Tre6P nexus in the roots, potentially delivering incorrect signals on the availability of sucrose, which compromises root growth. We also cannot exclude the possibility that TPS1 has some other function in roots that is not directly linked to its catalytic activity. Regulation of maize inflorescence branching by the maize RAMOSA3 and ZmTPP4 proteins is uncoupled from their catalytic activities (Claeys et al., 2019), setting a precedent for Tre6P-metabolizing enzymes to have moonlighting functions. Overall, these observations point to OtsA acting as an adequate source of Tre6P when expressed in planta, but lacking some functionalities of TPS1.
In wild-type Col-0 and TPS1 control seedlings, Tre6P was strongly responsive to exogenous supply of sucrose, rising rapidly, after a short lag, to reach levels that were up to 140-fold higher than in C-starved seedlings (Figure 8; see also Lunn et al., 2006;Yadav et al., 2014). Tre6P also rose after sucrose addition to TPS1[ N C], TPS1[ C] and OtsA seedlings, but less strongly than in the Col-0 and TPS1 controls (Figure 8). It is likely that simple massaction effects, i.e. increased availability of the UDP-glucose and Glc6P substrates for Tre6P synthesis, made some contribution to these responses, although previous studies showed only weak correlations between Tre6P and UDP-glucose or Glc6P in wild-type Col-0 seedlings (Yadav et al., 2014 which were severely dwarfed, non-flowering and had very different metabolite profiles compared to wild-type plants. However, caution is needed in interpreting the strength of the correlation between Tre6P and sucrose when based on whole-rosette or whole-seedling measurements of these metabolites. The localization of TPS1 in specific tissues and cell types (e.g. phloem and guard cells), suggests that the distribution of Tre6P at a whole plant level may be more heterogeneous than previously assumed (Martins et al., 2013), and responsive to changes in sucrose levels only in specific cells, such as phloem parenchyma, companion cells, sieve elements and guard cells. Subcellular compartmentation adds a further potential layer of complexity; the nuclear localization of TPS1 suggests that Tre6P is responsive to sucrose levels in the nucleo-cytosolic compartment of those cells where TPS1 is expressed.
However, this probably constitutes a minor fraction of the total sucrose in the plant, as there are large pools of sucrose in vacuoles, and some of the sucrose in rosettes is present in the apoplast or in the mesophyll cell chloroplasts (Lunn, 2016). Indeed, given the differences in the cellular and subcellular distributions of Tre6P and sucrose, it is quite remarkable that such strong correlations between Tre6P and sucrose are ever observed at the whole-plant level.
Following this line of thought, it is possible that the sucrose-Tre6P nexus relationship is still operating in critical cell types in otsA lines, accounting for the relatively mild phenotypes of these plants, even though at a whole-plant level Tre6P and sucrose are not well correlated.
Another possibility is that the severe phenotypes of the located, we need to consider the possibility that the relative importance of these two functions, although connected, may differ between tissues and cell types. In some parts of the plant and some developmental stages, homeostatic regulation of sucrose levels might be the dominant role of Tre6P, while in others, signaling of sucrose availability for growth might be more important. Therefore, in future experiments, it will be crucial to target manipulations of Tre6P to individual cell types and developmental stages to understand its specific functions, as exemplified by the work of Nuccio et al. (2015) in improving drought tolerance in maize.

Molecular Cloning
Arabidopsis genomic DNA was isolated using the cetyl trimethylammonium bromide (CTAB) method (Doyle and Doyle, 1987) with elution of DNA in 100 µL double-distilled H2O (ddH 2 O). All constructs were based on the pGreenII binary plasmid (www.pgreen.ac.uk; Hellens et al., 2000), with the phosphinothricin N-acetyltransferase (pat) gene replaced by the neomycin phosphotransferase (nptII) gene to allow selection of plant transformants with kanamycin. The Arabidopsis TPS1 (At1g78580) gene was amplified in sections from Arabidopsis Col-0 genomic DNA by PCR using Q5® High-Fidelity DNA polymerase, and reassembled in the pGreenII[nptII] plasmid as follows: (i) the 3´-UTR and terminator region were ligated into the XmaI/HindIII sites; (ii) the promoter region, including 5´-UTR, was added in two parts using the KpnI/ScaI and ScaI/SpeI restriction sites (with ScaI naturally occurring in the TPS1 promoter, Figure 1C); and (iii) the genomic protein coding region, including all introns, was inserted into the SpeI/XmaI sites. A StuI restriction site was added after the start codon and an ApaI site added before the stop codon, allowing removal of the N-and Cterminal domains by naturally occurring StuI and ApaI sites in the TPS1 sequence ( Figure 1C-D). GFP, GUS and SV40 tags were inserted at the N-terminus using a PfoI site created by addition of the StuI site to the TPS1 sequence ( Figure 1C). For C-terminal GFP/GUS fusions of the wild-type and SV40 tagged versions of the full-length TPS1, the GFP/GUS coding sequences were fused to a DNA fragment corresponding to the TPS1 C-terminal domain in vitro (matching overlapping ends were generated by BstXI), and insertion of the resulting Cterminus-GFP/GUS-fusion DNA product back into the ApaI sites ( Figure 1D). For C-terminal deletion constructs, the TPS1 fragment encoding the C-terminus was replaced with the GFP sequence using the ApaI sites ( Figure 1D). Site directed mutagenesis was performed using the fully assembled pGreenII[nptII]/TPS1 plasmid as a template, with specific mutations being generated by PCR using a Q5® Site-Directed Mutagenesis Kit (New England Biolabs GmbH; Figure 1E). The otsA gene from E. coli (strain K12) was amplified from E. coli DNA and inserted into the TPS1pro:TPS1term expression cassette using the SpeI/XmaI sites ( Figure 1F).
Oligonucleotide primers used for cloning are shown in Supplemental Table 1.

Metabolite Analysis
Frozen plant tissue was ground to a fine powder at liquid nitrogen temperature using a ball mill (Retsch; www.retsch.com). Aliquots (10-20 mg) of frozen tissue powder were extracted with chloroform-methanol as described in Lunn et al. (2006) Tre6P, phosphorylated intermediates and organic acids were measured by LC-MS/MS (Lunn et al., 2006), with modifications as described in . Individual amino acids were measured by HPLC with pre-column derivatization with O-phthaldialdehyde and fluorescence detection (Watanabe et al., 2013). Starch was determined enzymatically (Hendriks et al., 2003).

Data Analysis and Statistics
Data plotting and statistical analysis were performed using R Studio Version 1.0.163 (www.rstudio.com) with R version 3.6.1 (https://cran.r-project.org/) and the package and other picture quality settings being adjusted in the same way for all comparable images.

Accession Numbers
Sequence data from this article can be found under the following accession numbers: TPS1 Supplemental Table 1. Oligonucleotide primers used for cloning and testing the zygosity of the tps1-1 mutant locus in transgenic Arabidopsis lines.
Supplemental Table 2. Genotypes generated during the course of this study.
Supplemental Table 3. Accession numbers of TPS polypeptide sequences shown in Supplemental Figure 19.
Supplemental Data Set 1. Statistical analysis.   The tps1-1 mutant was complemented with various constructs encoding GFP-tagged forms of the TPS1 protein and the localization of the GFP-fusion proteins was examined by laser scanning confocal microscopy in (A) guard cells on the abaxial surface of the leaf, (B) roots of 8-d-old seedlings, and (C) shoot apical meristems. A complement ed line expressing the native (i.e. non-tagged) TPS1 protein (TPS1, top row) was used as a negative control for autofluorescence. The constructs used for complementation of tps1-1 are shown in Figure 1. The green and red channels show GFP and chlorophyll autofluorescence, respectively, and the merged image shows the GFP signal superimposed on the bright-field image. In (C), the dotted white lines mark the outer surface of the meristem and leaf primordia. Scale bar = 30 μm.