Raffinose synthase enhances drought tolerance through raffinose synthesis or galactinol hydrolysis in maize and Arabidopsis plants

Raffinose and its precursor galactinol accumulate in plant leaves during abiotic stress. RAFFINOSE SYNTHASE (RAFS) catalyzes raffinose formation by transferring a galactosyl group of galactinol to sucrose. However, whether RAFS contributes to plant drought tolerance and, if so, by what mechanism remains unclear. In this study, we report that expression of RAFS from maize (or corn, Zea mays) (ZmRAFS) is induced by drought, heat, cold, and salinity stresses. We found that zmrafs mutant maize plants completely lack raffinose and hyper-accumulate galactinol and are more sensitive to drought stress than the corresponding null-segregant (NS) plants. This indicated that ZmRAFS and its product raffinose contribute to plant drought tolerance. ZmRAFS overexpression in Arabidopsis enhanced drought stress tolerance by increasing myo-inositol levels via ZmRAFS-mediated galactinol hydrolysis in the leaves due to sucrose insufficiency in leaf cells and also enhanced raffinose synthesis in the seeds. Supplementation of sucrose to detached leaves converted ZmRAFS from hydrolyzing galactinol to synthesizing raffinose. Taken together, we demonstrate that ZmRAFS enhances plant drought tolerance through either raffinose synthesis or galactinol hydrolysis, depending on sucrose availability in plant cells. These results provide new avenues to improve plant drought stress tolerance through manipulation of the raffinose anabolic pathway.

The expression and function of GOLS in response to plant abiotic stress have been well-elucidated in many plant species. Expression of some GOLS genes is induced by abiotic stress, such as drought, heat shock, salinity, and osmotic shock (7,8), whereas others are induced by pathogen attack (9). Overexpression of GOLS increased the galactinol and raffinose content, whereas concurrently enhancing the abiotic stress tolerance of transgenic plants (7,(10)(11)(12). Because both galactinol and raffinose are increased by overexpression of GOLS genes, it is unclear whether galactinol, raffinose, or both directly enhance abiotic stress tolerance. Further complicating this situation, GALACTINOL SYNTHASE has, as one of its substrates, myo-inositol, a sugar alcohol with abiotic stress alleviating properties of its own (13) and whose utilization to make galactinol, necessarily draws down its concentration in the cell.
Unlike GOLS, very few RAFS have been characterized in vitro or in vivo (4,5,(14)(15)(16)(17). RAFS genes are induced by dehydration stress (17), and their transcripts accumulate in the late stages of seed maturation and desiccation, coincident with increased raffinose content (18,19). There are six putative RAFFINOSE SYNTHASE genes (AtRS1-6) in Arabidopsis thaliana (Arabidopsis) (20). AtRS5 is the only genuine raffinose synthase that has been shown to have RAFS activity (16). The mutation of AtRS5 abolishes raffinose in the leaves and reduces the raffinose content in seeds. AtRS4, identified as a STACHYOSE SYNTHASE (AtSTS), has some raffinose synthetic capacity and the atrs4 mutant is devoid of stachyose in the seeds (5). In addition to AtRS5 and AtRS4 in Arabidopsis, AtRS6 is predicted to be a drought-induced (through alternative splicing) RAFS (21). Knockout mutants in genes involved in raffinose biosynthesis in Arabidopsis have failed to show any plant-level phenotypic changes during normal growth other than perturbations in raffinose family oligosaccharides amounts and a repression of seed germination in darkness (5,16). Fewer RAFS genes have been predicted in the maize (Zea This article contains supporting information. ‡ These authors contributed equally to this work. * For correspondence: Guoying Wang, wangguoying@caas.cn; Tianyong Zhao, tzhao2@nwafu.edu.cn. cro ARTICLE mays) genome based on bioinformatics studies (22), and some of these, such as GRMZM2G340656, have already been experimentally identified and annotated as encoding ALKALINE ALPHA GALACTOSIDASE (23). We recently characterized the unique maize raffinose synthase gene and found that it plays an important role in maize seed vigor and longevity (19). Currently, no evidence has shown that over-expression of RAFS can enhance the raffinose content in plant vegetative tissues. Despite extensive studies, the function of raffinose in regulation of, or response to, vegetative abiotic stress has not been distinguished from that of galactinol or myo-inositol and, thus, the role raffinose plays in plant drought stress tolerance remains unclear. There are three hypothetical mechanisms addressing how raffinose performs its function when plants are in, or are recovering from, adverse conditions. 1) As a protective agent, raffinose could stabilize membranes during dehydration and prevent leakage of cellular contents and membrane fusion after rehydration (24). 2) As a ROS scavenger, raffinose (and galactinol) may mitigate oxidative damage under abiotic stress conditions (20). 3) Raffinose could be transported into chloroplasts, protecting thylakoids and stabilizing PSII (25,26).
In this study, we demonstrated that maize zmrafs mutants lacking raffinose and hyper-accumulating galactinol, but with WT amounts of sucrose and myo-inositol were more sensitive to drought stress than null-segregant lines in two maize inbred lines. Similar to the ZmGOLS2-overexpressing Arabidopsis plants, ZmRAFS-overexpressing Arabidopsis plants displayed a significantly increased tolerance to drought stress. However, whereas the galactinol and raffinose amounts in ZmGOLS2overexpressing plants were greater, both were significantly decreased in ZmRAFS-overexpressing Arabidopsis leaves. The ZmRAFS in ZmRAFS-overexpressing Arabidopsis leaves did not seemingly contribute to net raffinose accumulation but rather hydrolyzed galactinol without transfer of the galactose to sucrose, thereby generating large amounts of myo-inositol under drought stress. This study demonstrates that ZmRAFS enhances plant drought tolerance through either synthesis of more raffinose or by generation of more myo-inositol via hydrolysis of galactinol.

ZmRAFS is induced by abiotic stress
ZmRAFS transcription was rapidly and dramatically up-regulated in leaves when the V3 stage (three leaves stage) seedlings were treated with heat shock, dehydration, or subjected to increased salt within 2-3 h, but returned to background levels with prolonged treatment (Fig. 1A). Under cold stress, ZmRAFS exhibited a weakly increased level of transcript in V3 stage seedlings treated for 8 h (Fig. 1A). ZmRAFS protein was undetectable in untreated leaves of V3 stage seedlings, but accumulated nearly concomitantly with transcript under heat shock and salt stress, but showed a gradual increase under dehydration treatment (Fig. 1B). No ZmRAFS protein could be detected in leaves under cold stress (Fig. 1B). The ZmRAFS mRNA abundance and its protein accumulation are imperfectly correlated in maize seedlings under abiotic stress. These data suggest that ZmRAFS expression might be regulated at both transcriptional and translational levels. Subcellular localization of ZmRAFS protein was determined by transient expression of a fusion protein comprising ZmRAFS and YFP in maize leaf protoplasts. The fluorescence signal from YFP was detected throughout the cytoplasm and chloroplasts but was not in the vacuole, whereas the fluorescence of the ZmRAFS-YFP fusion protein was only detected in the cytoplasm (Fig. 1C). Western blot analysis of ZmRAFS-YFP fusion protein expression showed that the fusion protein was present and intact in transformed maize leaf protoplasts as was endogenous RAFS (Fig. 1D).
zmrafs maize mutants, containing no raffinose but accumulating galactinol, are more sensitive to drought stress than NS We next investigated the drought stress tolerance of the two zmrafs mutant lines in which ZmRAFS expression was interrupted by mutator insertion (19) (Fig. S1). The 7th leaf of maize plants was repeatedly photographed every other day during water withholding ( Fig. 2A and Fig. S2A). The zmrafs-1 mutant plants exhibited more severe injury compared with null-segregant (NS) plants ( Fig. 2A). zmrafs-2 leaves showed greater longitudinal folding than NS when the rolled leaves were spread open (Fig. S2A). Electrolyte leakage, an indicator of cellular injury (27), of the mutant leaves was significantly greater than that of NS leaves (Fig. 2B) on the 16th, 18th, and 20th day of the drought stress, but not during normal watering. A similar electrolyte leakage phenotype was observed in zmrafs-2 mutant plants relative to their NS, occurring even earlier at day 14 of the stress (Fig. S2B). Sucrose and myoinositol amounts in leaves prior to and during the drought were similar between NS and mutant plants except for myoinositol at 10 days after the start of the drought stress (Fig. 2, C and D). Galactinol accumulated to greater amounts in mutant leaves were compared with those of the NS; whereas the reverse was true of raffinose which, as expected, was undetected in the mutant (Fig. 2, E and F).

Overexpression of ZmRAFS in Arabidopsis enhanced galactinol hydrolysis rather than raffinose synthesis in leaves yet still enhanced drought stress tolerance
To further investigate the consequences of altered raffinose amounts on plant drought stress tolerance, three Arabidopsis lines overexpressing the ZmRAFS gene were characterized for drought stress tolerance. In addition to WT control, the GFPand ZmGOLS2-overexpressing Arabidopsis lines (12) were also included as controls. The ZmRAFS and ZmGOLS2 protein accumulation was determined by Western blot analysis (Fig.  S3A). The Arabidopsis raffinose synthase (atrs5) mutant was characterized by PCR (Fig. S3, B and C) and also used as a drought-susceptible control. After drought stress and re-watering, the 3 ZmRAFS-overexpressing lines were obviously more resistant to drought stress than any of the control lines except for the ZmGOLS2-overexpressing plants known to possess superior drought tolerance (12) (Fig. 3A). The survival percentage of ZmRAFS-and ZmGOLS2-overexpressing lines after drought stress was similar and both were significantly greater than that of the wildtype (WT), GFP expressing (OEGFP) or atrs5 plants (Fig. 3B). For unstressed seed-

Raffinose regulation of plant drought tolerance
lings, there was no difference in leaf ion leakage percentages regardless of genotype (Fig. 3C). However, upon drought stress, those percentages for all ZmRAFS-and ZmGOLS2overexpressing plants were similarly low, and both are significantly lower than that of the WT, GFP expressing, or atrs5 plants (Fig. 3C).
Drought stress enhanced the sucrose content in the leaves of all plant genotypes; increased the myo-inositol content in the 3 ZmRAFS-overexpressing plants and in the atrs5 mutant; increased the galactinol content in all leaves apart from those of the 3 ZmRAFS plants; and increased the raffinose content in all leaves excluding those of 2 of the 3 ZmRAFS plants (Fig. 3D). Overexpression of ZmGOLS2 enhanced both galactinol and raffinose contents, regardless of stress, compared with control lines (WT and OEGFP) (Fig. 3D). The raffinose content was decreased, whereas the galactinol content was increased in atrs5 mutant leaves as compared with WT (Fig. 3D). Unexpectedly, the raffinose and galactinol content in leaves of the 3 ZmRAFS lines were significantly decreased compared with WT and OEGFP whether the plants were well-watered or under drought conditions (Fig. 3D). Surprisingly, given what was observed in the leaves, the raffinose content in seeds of ZmRAFS-overexpressing Arabidopsis lines was significantly greater than that of other lines (19). To investigate whether ZmRAFS protein expressed in OEZmRAFS leaves was able to synthesize raffinose using sucrose and galactinol, enzyme activ- Figure 1. ZmRAFS is responsive to abiotic stress and its encoded protein is localized to the cytosol. A and B, ZmRAFS expression was induced by different abiotic stresses in maize B73 seedling leaves (V3 stage). The maize seedlings were treated with heat shock (42°C), dehydration (dry on blot paper), NaCl (200 mM), and cold stress (4°C), respectively, for certain periods of time as indicated in the graph, and the leaf samples were collected for detection of the mRNA or protein expression of ZmRAFS. A, the ZmRAFS mRNA accumulation in the second leaf was determined by real-time RT-PCR. The expression of ZmRAFS mRNA was normalized to the expression of maize GAPDH and is presented relative to the untreated, control samples (0 h). The x-axis represents the time course of treatment. The y-axis indicates relative ZmRAFS expression. The mixture of the 2nd leaf of three different seedlings were applied for RNA extraction and qRT-PCR as one biological replicate, there are three biological replicates. Each circle represents one biological replicate, and the horizontal bars denote the means. Different lowercase letters above the bars denote significant differences among different times of treatments (ANOVA test, p Ͻ 0.05). B, ZmRAFS protein accumulation was determined by Western blot analysis. Top panel, Western blot analysis of ZmRAFS protein; bottom panel, Western blot analysis of GAPDH protein demonstrating equal protein loading. C, ZmRAFS was localized to the cytoplasm of maize protoplast cells. YFP expression vector or a ZmRAFS-YFP fusion protein expression vector were transiently transformed into maize protoplasts. The YFP fluorescence and chloroplast autofluorescence signals were detected by confocal microscopy and then merged (bar ϭ 10 m). BF, bright field; YFP, YFP signal; Chl, chloroplast autofluorescence signal; Merge, merged signal of YFP and Chl. D, Western blot analysis of ZmRAFS-YFP fusion protein in transiently transformed maize protoplast cells as in C using ZmRAFS antibody. Both protoplast samples have endogenous RAFS (Native) but only the transiently transformed cells also have signal from the chimeric ZmRAFS-YFP fusion. From the same samples, Western blot analysis of ZmGAPDH protein in the lower panels demonstrating equal protein loading.

Raffinose regulation of plant drought tolerance
ities of leaf extracts from all Arabidopsis lines were tested in vitro for raffinose-synthetic (transferase) and nonraffinose producing, galactinol-hydrolytic activities. Under the conditions tested, the leaf extracts of all three of the ZmRAFS-expressing Arabidopsis lines had both dramatically greater transferase (Fig. 3E) as well as hydrolytic (Fig. 3F) activities relative to the leaf extracts of the other genotypes.

Characterization of the substrate-binding site of ZmRAFS
RAFFINOSE SYNTHASE binds two substrates (galactinol and sucrose) and it has both raffinose synthetic and galactinol hydrolytic activities. To determine the substrate-binding sites of the ZmRAFS, enzyme activity assay of mutated ZmRAFS was performed (Fig. 4). According to its closely related templates, the most probable ligand-binding sites of ZmRAFS were predicted using the I-TASSER online resource (28) (Table S1). To determine the putative substrate-binding residues of ZmRAFS, those predicted amino acids (AA) were mutated to alanine to minimize potential structural alterations of the mutated protein (Fig. 4A). The solubility of the mutant ZmRAFS-His 6 recombinant proteins were demonstrated using Western blot analysis of an aliquot of the supernatant for each mutant (Fig.  4B). The bacterial-expressed, WT ZmRAFS protein showed both raffinose synthetic and unproductive galactinol hydrolytic activities as determined by in vitro assay (Fig. 4C). Although mutation of many residues decreased or abolished either raffinose-synthetic activity or both raffinose-synthetic and unproductive galactinol-hydrolytic activity, none enhanced raffinosesynthetic activity (Fig. 4C). The Asp-264 -Ala mutant had greatly increased (ϳ2-fold) unproductive galactinol-hydrolytic activity but had no detectable raffinose-synthetic activity. These results are suggestive that the galactinol-hydrolytic activity of ZmRAFS is independent of forming raffinose, supportive of the previous characterization of the catalytic mechanism as a Ping-Pong, Bi-Bi reaction (4). conditions as in A. The middle part of the 7th leaves from different plants were collected for testing of the electrolyte leakage. Each circle represents one biological replicate (one single plant), there are three to eight biological replicates for each treatment. The horizontal bars denote the means. *, p Ͻ 0.05; **, p Ͻ 0.01 as compared by Student's t test. C-F, soluble sugar profiles of leaves from NS and zmrafs-1 plants at the start and during drought stress. Sucrose (C), myo-inositol (D), galactinol (E), and raffinose (F) amounts were measured for the null-segregant (blue circles) and the maize raffinose synthase mutant-1 (zmrafs-1; red diamonds). The middle part (1 g) of the 7th leaves from different plants were collected for sugar extraction and HPLC-ELSD detection. Each circle represents one biological replicate (one single plant), there are three biological replicates. The horizontal bars denote the means. *, p Ͻ 0.05; **, p Ͻ 0.01 denotes a significant difference between NS and zmrafs-1 (Student's t test). There was no significant difference of sucrose and myo-inositol between NS and zmrafs-1 mutant.

Raffinose regulation of plant drought tolerance A relative amount of galactinol and sucrose to ZmRAFS, determines whether ZmRAFS completes the transferase reaction, forming raffinose, after hydrolyzing galactinol
To investigate the effects of galactinol or sucrose on the ZmRAFS enzyme, the activity of Escherichia coli-expressed ZmRAFS-His 6 recombinant protein was characterized in substrate concentration control systems (Fig. 5, A and B). Myoinositol is generated by both raffinose synthesis and galactinol hydrolysis; whereas raffinose is only generated when the enzyme completes its transferase function. By this logic, the molar ratio of myo-inositol to raffinose is expected to reflect whether ZmRAFS performs the entire synthesis reaction or only hydrolyzes galactinol without transfer of galactose to sucrose. When sucrose was saturating (29.2 mM) and the galactinol concentration was low (1.1-2.6 mM), similar amounts of myo-inositol and raffinose were generated ( Fig. 5A: the ratio (myo-inositol/raffinose) was approximately 1). As the galactinol concentration increased (Ն5.3 mM compared with the apparent K m ϭ 2.93 mM; Table S2), the amount of myoinositol that was generated was significantly greater than that of raffinose and so the ratio became statistically significantly greater than 1 (Fig. 5A). When galactinol was saturating (15.9 mM), regardless of the sucrose concentration, the amount of myo-inositol that was generated was always greater than that of ZmGOLS2-overexpressing Arabidopsis plants was greater than that of WT plants when water was withheld from the plants. The survival rate was calculated from three to six pots of plants for each Arabidopsis line in one independent experiment. There are three independent experiments. Each circle represents the data of one independent experiment. The horizontal bars denote the means. *, p Ͻ 0.05; **, p Ͻ 0.01, as compared with WT control (ANOVA test). C, the leaf ion leakage of all three ZmRAFS-and the ZmGOLS2-overexpressing Arabidopsis plants was less than that of WT plants at the end of the drought stress treatment. Leaves form different plants that were grown under normal and drought conditions were collected for ion leakage test. Each circle represents one biological replicate (one single plant), there are 3-12 biological replicates for each tested line. The horizontal bars denote the means. **, p Ͻ 0.01 compared with WT control (ANOVA test). D, comparison of soluble sugar related to raffinose biosynthesis of the different Arabidopsis plants as in A, which were grown under well-watered or drought conditions. Total soluble sugar was extracted from 0.5 g of leaves that were collected from different Arabidopsis plants for each line (one biological replicate) for HPLC-ELSD detection. Each circle represents one biological replicate and there are three to six biological replicates for each tested line. The horizontal bars denote the means. The colored asterisks are indicative of significant differences from WT under normal (blue) or drought (red) conditions (*, p Ͻ 0.05 and **, p Ͻ 0.01, ANOVA test). A black asterisk (*, p Ͻ 0.05 and **, p Ͻ 0.01) indicates significant differences between the normal and drought condition of each line (Student's t test). E, characterization of the raffinose synthetic activity of the leaf extracts from different Arabidopsis plants as described in A. Raffinose and myo-inositol were determined using cell-free leaf extracts of three ZmRAFS-overexpressing lines incubated with galactinol and sucrose, as detected by HPLC-ELSD. The enzymatic activity was determined by raffinose amounts (nmol) produced per mg of protein/min. The crude enzyme extracts that were extracted from the leaf mixture of three different plants were determined for its raffinose synthetic activity (one biological replicate). There are three biological replicates. Each triangle represents one biological replicate, and the horizontal bars denote the means. **, p Ͻ 0.01 compared with WT (ANOVA test). F, characterization of the galactinol hydrolytic activity of the leaf extracts from the Arabidopsis plants as described in A. Galactose and myo-inositol were determined using cell-free leaf extracts of three ZmRAFS-overexpressing lines incubated with galactinol as detected by HPLC-ELSD. The enzymatic activity was determined by myo-inositol amounts (nmol) produced per mg of protein/min. The crude enzyme extracts that were extracted from the leaf mixture of three different plants were determined for its galactinol hydrolytic activity (one biological replicate). There are three biological replicates. Each triangle represents one biological replicate, and the horizontal bars denote the means. **, p Ͻ 0.01 compared with WT (ANOVA test).

Raffinose regulation of plant drought tolerance
raffinose produced; however, the ratio of myo-inositol to raffinose gradually decreased as the sucrose concentration was increased (Fig. 5B). These results show that the relative amount of the two substrates, sucrose and galactinol, determines whether ZmRAFS predominantly hydrolyzed galactinol unproductively or predominantly completed the synthesis of raffinose, suggesting that greater raffinose production in Arabidopsis leaves expressing ZmRAFS may be possible by boosting substrate amounts. The kinetic assays determined that ZmRAFS exhibited apparent lower affinities of the substrates than that of AtRS5 (Table S2) (29). Under the conditions used, ZmRAFS has a higher affinity for galactinol (apparent K m ϭ 2.93) than sucrose (apparent K m ϭ 7.40) when both substrates were present; and exhibited higher apparent affinity for galactinol when sucrose was absent (Table S2).
To attempt to boost raffinose amounts and to further investigate why ZmRAFS predominantly hydrolyzed galactinol without completing the transferase reaction to produce raffinose in ZmRAFS-expressing Arabidopsis leaves, ZmGOLS2/ ZmRAFS-double expressing plants were generated to test whether ZmRAFS would synthesize raffinose if the galactinol content was increased in leaves. The paradigm for this experiment was the ZmGOLS2 overexpressing plants, which produce greater amounts of both galactinol and raffinose (Fig. 3D). As

Raffinose regulation of plant drought tolerance
characterized by Western blot analysis, the accumulation of ZmRAFS protein in leaves of ZmRAFS-and ZmGOLS2/Zm-RAFS-double expressing plants were, as expected, greater than that of WT, GFP, or the ZmGOLS2 line (Fig. S4). Similarly, the ZmGOLS protein accumulation in leaves of ZmGOLS2-and ZmGOLS2/ZmRAFS-double expressing plants was greater than that of WT, GFP, or the ZmRAFS line (Fig. S4). Comparison of the plant morphology, percentage survival, and the ion leakage of these different Arabidopsis genotypes after drought stress showed that the ZmGOLS2/ZmRAFS-, ZmRAFS-, and ZmGOLS2-expressing lines were more tolerant of drought stress compared with WT plants (Fig. 6, A-C).
Drought stress increased sucrose accumulation in leaves of all lines; and increased myo-inositol in OEZmRAFS-25 and two double over-expressing lines (OEZmGOLS2/ZmRAFS-1 and -3) (Fig. 6D). Expression of ZmGOLS2 in Arabidopsis significantly increased the galactinol and raffinose content in leaves, whether the plants were grown under normal or drought stress conditions, as compared with WT (Fig. 6D). The galactinol content in the leaves of ZmGOLS2/ZmRAFS expressing plants was significantly less than that of the ZmGOLS2 expressing plants, whether the plants were grown under either normal or drought conditions (Fig. 6D). The raffinose content in the leaves of ZmGOLS2/ZmRAFS expressing plants was similar to that of the ZmGOLS2 expressing line when the plants were grown under well-watered conditions, but was less than that of the ZmGOLS2 line when the plants were grown under drought conditions, although it was still superior to WT or GFP control plants (Fig. 6D). These data indicate that ZmRAFS still unproductively hydrolyzed galactinol in ZmGOLS2/ZmRAFS expressing plants. Infiltration of sucrose into the detached leaves of all the Arabidopsis genotypes, compared with that of infiltra-tion of detached leaves with water, increased the sucrose content in the leaf (Fig. 6E). Of the water-treated group, galactinol was only detected in ZmGOLS2 plants (Fig. 6F). The galactinol content was significantly increased in detached leaves of both the ZmGOLS2-and -1 of 3 ZmGOLS2/ZmRAFS-1 expressing plants upon sucrose treatment compared with water treatment, but remained constant in ZmGOLS2/ZmRAFS-2 and -3 lines (Fig. 6F). Even with water infiltration, the raffinose content in ZmGOLS2 plants was significantly greater than that of other plants (Fig. 6G). Interestingly, the raffinose content in the ZmGOLS2/ZmRAFS-1 and -2 plants was significantly greater than that, not only of the control plants, but also that of ZmGOLS2 plants when sucrose treated (Fig. 6G) despite the seemingly different ratios of protein of the two enzymes accumulated in those plants (Fig. S4). These results are consistent with the greater accumulation of raffinose in the seeds of the ZmRAFS plants (19), these data suggested that infiltration of sucrose to the detached leaves of ZmGOLS2/ZmRAFS-expressing Arabidopsis plants provided sufficient sucrose substrate for ZmRAFS to complete the transferase reaction after hydrolyzing galactinol and subsequently, to accumulate raffinose.

Overexpression of ZmMIPS increased myo-inositol accumulation in leaves and enhanced plant drought tolerance
Over-expression of ZmRAFS in Arabidopsis decreased galactinol and raffinose content, whereas it increased myo-inositol content in leaves and enhanced the plant drought tolerance (Figs. 3,  A-D, and 6, A-D). To further demonstrate that it is myo-inositol that functions in this scenario to protect the plants from drought stress, two maize homologues of AtMIPS (myo-inositol-1-phosphate synthase, the key enzyme for myo-inositol synthesis in Arabidopsis) were cloned and named ZmMIPS1 (GRMZM2G155242) and ZmMIPS2 (GRMZM2G004528). The expression of ZmMIPS1 and ZmMIPS2 proteins were induced in bacteria and purified (Fig. S5A). Enzyme activity assay of the purified N-terminal His tag fusion proteins of ZmMIPS1 or ZmMIPS2 showed that both ZmMIPS fusion proteins were able to generate myo-inositol-1phosphate (Fig. S5B). Plant expression vectors for ZmMIPS1 or ZmMIPS2 were constructed and transformed into Arabidopsis. The mRNA and protein accumulation of ZmMIPS in Arabidopsis was determined by RT-PCR and Western blotting, respectively (Fig. 7A). Over-expressing ZmMIPS substantially increased myo-inositol content in leaves under both normal and drought conditions (Fig. 7B). The 3 ZmMIPS-overexpressing Arabidopsis lines were more resistant to drought stress than any of the control lines (Fig. 7C). After re-watering, the survival percentage of the 3 ZmMIPS expressing lines were significantly higher than that of control lines (Fig. 7D). These results demonstrated that increasing myo-inositol amounts was able to enhance plant drought tolerance.

Raffinose, but not galactinol, directly contributes to plant drought stress tolerance
The function of raffinose and galactinol in plant drought stress tolerance and the molecular mechanism by which they act remains unclear despite extensive studies. Galactinol is a direct substrate for raffinose biosynthesis (Fig. 8). Both galacti- Sucrose was kept at saturating levels (29.2 mM) as galactinol concentration was varied as noted on the x axis. Crude enzyme extracts were extracted from BL21 (expressing ZmRAFS) bacteria cultures for the enzymatic activity assay. Each circle (in the top panel) or triangle (in the bottom panel) represents one biological replicate (cultures from one flask), there are three biological replicates. Sugar amount data represent the mean of three replicates and significant differences between the two products were indicated (** ϭ p Ͻ 0.01) using a Student's t test. Different lowercase letters denote significant differences among the ratios within each data set (Tukey's test). B, same as in A except galactinol concentration was kept constant and saturating (15.9 mM) as the sucrose concentration was varied.

Raffinose regulation of plant drought tolerance
nol and raffinose are accumulated to higher levels in plants in response to abiotic stresses (12). Knockout mutants of relevant genes involved in raffinose biosynthesis have failed to show any abnormalities in Arabidopsis during normal growth other than perturbations in raffinose amounts and decreased completion of germination in the dark (5,16,29). Over-expression of GOLS in Arabidopsis increased both galactinol and raffinose content, decreased myo-inositol (Fig. 3D), whereas concurrently enhancing the abiotic stress tolerance of plants (7,(10)(11)(12). Until now, no evidence had shown that over-expression of RAFS could regulate the raffinose content in plant vegetative tissues. In short, despite extensive studies, the function of raffinose in response to plant abiotic stress has not been distinguished from that of galactinol and the role raffinose plays in plant drought stress tolerance remains unknown. By characterization of the maize knockout mutant (zmrafs) and ZmRAFS-expressing Arabidopsis plants, we found that raffinose, but not galactinol, directly and positively regulates plant drought stress tolerance.

myo-Inositol generated from ZmRAFS-mediated galactinol hydrolysis in ZmRAFS overexpressing Arabidopsis enhances plant drought tolerance
ZmGOLS2 overexpressing Arabidopsis plants with increased galactinol and raffinose enhanced plant drought stress tolerance (Figs. 3 and 6) (12). However, ZmRAFS overexpressing Arabidopsis plant leaves contained much less raffinose and galactinol, yet the plants were more resistant to drought stress compared with the WT or GFP plants (Fig. 3, A-D). These There are six independent experiments. Each circle represents data from one independent experiment. The horizontal bars denote the means. Significance as compared with the WT control was determined using ANOVA test and is denoted (**, p Ͻ 0.01). C, comparison of the leaf ion leakage of different Arabidopsis lines, which were under normal growth condition or after drought stress (before re-watering) treatments. Leaves from different individual plant under normal and drought conditions were collected for ion leakage test. Each circle represents one biological replicate (one plant) and there are three to eight biological replicates for each Arabidopsis line. The horizontal bars denote the means. Significance is indicated (**,p Ͻ 0.01) for the comparisons made against WT plants (ANOVA test). D, comparison of sucrose, myoinositol, galactinol, and raffinose content of leaves among different Arabidopsis lines, which were grown under normal or drought conditions. Total soluble sugar was extracted from 0.5 g of leaves from different plants and the sugar content was detected by HPLC-ELSD. Each circle represents one biological replicate (independent sugar sample from different individual plant), there are three to six biological replicates for each Arabidopsis line. The horizontal bars denote the means. ANOVA test was performed to detect the significant difference between different lines (*, p Ͻ 0.05; **, p Ͻ 0.01), as show with colored asterisks. Student's t test was using for analyze significant differences between the normal and drought condition of each line (*, p Ͻ 0.05, **, p Ͻ 0.01, showed by black asterisk). E-G, comparison of the sucrose (E), galactinol (F), and raffinose (G) content in detached leaves, infiltrated with either water or 30 mM sucrose for 1 h, among the various Arabidopsis genotypes. The leaves of different Arabidopsis lines (4 weeks-old) were then subsequently infiltrated in the same solution for 2 h and then incubated for another 22 h under light, the sugar content of leaves was determined by HPLC-ELSD. Each circle represents one biological replicate (leaves from 5 plants of each Arabidopsis line for each treatment), there are there biological replicates. The horizontal bars denote the means; *, p Ͻ 0.05; **, p Ͻ 0.01 (Student's t test). An asterisk indicates a significant difference between sugar-and water-treated groups of each plant genotype. Different lowercase letters denote significance among different plant genotypes within the treatment group (Tukey's test).

Raffinose regulation of plant drought tolerance
results indicate that ZmGOLS2-and ZmRAFS-overexpressing Arabidopsis plants utilize a different mechanism to protect plants from drought stress. ZmRAFS protein in leaf extracts of ZmRAFS overexpressing Arabidopsis plants was able to catalyze two different reactions: 1) to synthesize raffinose using exogenous sucrose and galactinol, or 2) to hydrolyze galactinol to produce myo-inositol and galactose when sucrose is absent (Fig. 3, E and F). The myo-inositol was significantly increased in ZmRAFS overexpressing plants under drought stress (Fig. 3D), indicating that ZmRAFS hydrolyzed galactinol instead of synthesizing raffinose in ZmRAFS overexpressing Arabidopsis leaves. It has been proposed that overexpression of myo-inositol phosphate synthase (MIPS) enhanced dehydration, salt, and cold stress tolerance in several plant species (13,(30)(31)(32)(33). Overexpression of two maize MIPSs in Arabidopsis increased myoinositol accumulation and enhanced the plant drought tolerance (Fig. 7). These data support that myo-inositol contribute to plant drought tolerance. myo-Inositol, either synthesized from Glc-6-P by MIPS in ZmMIPS overexpressing plants, or generated by the hydrolysis of galactinol in ZmRAFS overexpressing plant leaves, contributes to the plant drought tolerance. Galactinol functions as a storage or a supply of myo-inositol or galactose. ZmRAFS may either hydrolyze galactinol to generate myo-inositol in ZmRAFS-expressing lines, or use galactinol to synthesize raffinose in ZmGOLS2 lines, and in both cases, enhance the plant drought stress tolerance.
Taken together, we propose a model to elucidate the function and mechanism of raffinose metabolism in plant drought stress tolerance (Fig. 8). The drought stress signal activates transcription factors responsible for up-regulating genes encoding key enzymes, such as GOLS, RAFS, or MIPS in the raffinose biosynthetic pathway, and their action stimulates the accumulation of myo-inositol, galactinol, and raffinose. Both myo-inositol and raffinose directly and positively regulate plant drought stress tolerance, at least in Arabidopsis and maize.

Availability of sucrose determines whether ZmRAFS is to synthesize raffinose or hydrolyze galactinol
The ZmRAFS has been proven to have both raffinose synthetic and galactinol hydrolytic activity as determined either in vitro or in vivo (Figs. 3-5) (19). Either synthesis of raffinose or hydrolysis of galactinol by ZmRAFS would generate myo-inositol. Based on our data and the findings that the pea RAFS showed attributes of catalyzing a Ping-Pong Bi-Bi reaction (4), we propose here that RAFS binds and hydrolyzes galactinol first, releasing myo-inositol and retaining the galactose for some period until RAFS either: 1) binds sucrose and makes raffinose; or 2) releases free galactose exhibiting a galactinol hydrolytic activity. The lower the sucrose concentration, the greater the probability of RAFS hydrolyzing galactinol and the less opportunity RAFS has to make raffinose (Fig. 5B). With increased sucrose, ZmRAFS tends to synthesize raffinose rather than to hydrolyze galactinol (Fig. 5B). This hypothesis is also supported by the point mutation data of the ZmRAFS protein. Mutation of certain AAs of ZmRAFS abolished both galactinol hydrolytic and raffinose synthetic activity of ZmRAFS (Fig. 4). These sites are possibly critical for galactinol binding because the RAFS would not synthesize raffinose if it loses the ability of binding and hydrolyzing galactinol. Mutation of AA 264 from aspartic acid (Asp) to alanine (Ala) abolished the raffinose synthetic activity entirely while enhancing the galactinol hydrolytic activity about 2-fold, indicating that this site is likely critical either for sucrose binding or for catalytic activity of RAFFINOSE SYNTHASE, but not for galactinol binding. Supplementation of sucrose converted the ZmRAFS from hydrolyzing galactinol to synthesizing raffinose in ZmGOLS2/ZmRAFS double expressors (Fig. 6G). These data also suggest that ZmRAFS hydrolyzes galactinol first and then starts synthesizing raffinose if free sucrose is available.
Over-expression of GOLS has been reported to enhance the galactinol and raffinose content and abiotic stress tolerance of the plants (10,12,34). In those studies, more galactinol was produced by a combination of endogenous and heterologous GOLS gene expression, whereas raffinose was produced by endogenous RAFS. These data suggest that the substrate galactinol is a limiting factor for RAFS to synthesize raffinose in vivo and the higher molar ratio of galactinol to RAFS would stimulate raffinose biosynthesis. The myo-inositol, which was generated from galactinol hydrolysis, regardless of raffinose synthesis, was quickly recycled back to galactinol by a greater titer of GOLS enzyme.
ZmRAFS exhibited lower affinity to sucrose than it did with galactinol (Table S2). AtRS5 has dramatically higher affinity to sucrose than with galactinol (Table S2) (29). Thus, the high level of galactinol could promote raffinose synthesis catalyzed by AtRS5 without being hampered by sucrose insufficiency in the GOLS-overexpressing Arabidopsis line (Figs. 3D and 6D) (10,12,34). The sucrose content in Arabidopsis leaves is very high regardless of genotype (Fig. 3D), however, it is preferentially exported to the sink tissues loaded into companion cells or sieve elements by sucrose transporters. The K m of AtSUC9 for sucrose is 0.006 (35), whereas the K m of AtRS5 for sucrose is 0.35 (29) and the ZmRAFS to sucrose is 7.40 (Table S2). These data show that the sucrose transporter in leaves has a much higher affinity for sucrose than RAFS. Once sucrose is synthesized in leaves, it is immediately bound by the sucrose transporters and transported to the sink tissues. The remaining sucrose in leaves is presumably insufficient for RAFS to synthesize raffinose in ZmRAFS-overexpressing Arabidopsis leaves. Once sucrose is unloaded at the sink tissue, it is free for raffinose synthesis. We have previously reported that ZmRAFS-expressing Arabidopsis seeds has much greater raffinose content than that of control lines (19).
To our knowledge, the effects of over-expressing RAFS have not been reported except in one case of a cucumber RAFS gene, along with a muskmelon GOLS gene and the Alonsoa meridionalis STACHYOSE SYNTHASE (STS) gene all co-transformed into Arabidopsis plants. Transgenic lines homozygous for the 3 genes accumulated substantial galactinol, raffinose, and stachyose (36). Because over-expression of the GOLS gene itself may also cause these biological effects, it is unclear whether or not the foreign RAFS gene functions and it is difficult to draw a conclusion as to its contribution to raffinose production. The scarcity of publications of overexpression of RAFS genes might be due to its unanticipated function hydrolyzing galactinol

Raffinose regulation of plant drought tolerance
instead of synthesizing raffinose in plant leaves by RAFS when it is overexpressed in plants.

Regulation of raffinose metabolism pathway for improving plant drought stress tolerance
Our data indicate that raffinose is beneficial to plant drought stress tolerance. We and other groups have confirmed that overexpression of GOLS genes enhanced the abiotic stress tolerance of transgenic plants (11,12,34,37). It is worthwhile mentioning that constitutive expression of ZmGOLS2 genes in Arabidopsis did not cause any adverse effects to the plant under normal conditions (12). Similarly, overexpression of ZmRAFS enhanced plant drought stress tolerance without causing adverse effects to the plants under normal conditions (Fig. 3,  A-C). Furthermore, seed vigor can also be improved through manipulation of the raffinose family oligosaccharides metabolic pathway (19). Taken together, it has become promising to manipulate raffinose in crop plants to enhance both seed longevity and plant drought stress tolerance.

Plant material
Maize (Z. mays L.) inbred line B73 was maintained in the laboratory. The method of genotyping, Mu-insertional site identification, and characterization of seed sugar profile of the zmrafs mutants and NS counterparts were described previously (19). The Arabidopsis mutant atrs5 (CS318250, T-DNA inserted in At5g40390) was purchased from the Arabidopsis Biological Resource Center (ABRC) through the Arabidopsis Information Resource (TAIR; RRID:SCR_004618). The transgenic Arabidopsis lines expressing ZmGOLS2, ZmRAFS, and ZmMIPS were obtained by using floral dip method as previously described (19).

Vector construction
For construction of the ZmRAFS-YFP fusion protein expression vector, the Rluc CDS, the second 2ϫ 35 S-promoter, and the Fluc CDS in a dual luciferase control vector (38) were replaced by a MCS sequence (containing XhoI, PstI, NheI, HindIII, SacI, MluI, BglII, NcoI, BamHI, EcoRI, and XbaI). The YFP CDS was obtained by PCR from pmEYFP-1 vector (number 38773, Addgene) using a pair of primers (PYFP-F and PYFP-R; Table S3) and then inserted between the BamHI and EcoRI sites. Upstream of the YFP CDS, a double-stranded linker DNA fragment (5Ј-(CCATGG)GGAGGATCTGGAGGAGGAGGATC-TGGAGGA(GGATCC)-3Ј) was inserted between NcoI and BamHI sites. The ZmRAFS CDS, without the stop codon, was amplified using a pair of primers (ZmRAFS-CRF-HindIII and ZmRAFS-CRR-MluI; Table S3) together with the template, pET-ZmRAFS, and inserted between the HindIII and MluI sites. The construction of expression vectors for expressing ZmRAFS in Arabidopsis followed the published protocol (19). For construction of bacterial expression vector of mutated ZmRAFS protein, a QuikChange II XL Site-directed Mutagenesis Kit (Agilent Technologies, USA) was used according to the instruction manual utilizing the ZmRAFS:His 6 expressing vector pET-ZmRAFS (19) as the PCR template. Listed in Table S3, the primers were named by both the original AA, its position in the ZmRAFS protein, and the mutated AA (e.g. S115A-F and S115A-R are the forward and reverse primers, respectively, for mutating serine 115 of the ZmRAFS to alanine).
For expression of ZmMIPS protein in bacteria, prokaryotic expression vectors were constructed (pET-ZmMIPS1 and pET-ZmMIPS2). The cDNA of ZmMIPS1 and ZmMIPS2 were amplified from cDNA synthesized from mRNA that was isolated from maize leaf (B73 inbred line) using a pair of primers (for ZmMIPS1, pCSGFPBT-MIPS1-FϫpCSGFPBT-MIPS1-R; for ZmMIPS2, pCSGFPBT-MIPS2-FϫpCSGFPBT-MIPS2-R). Then the coding region of ZmMIPS1 and ZmMIPS2 were amplified from the first PCR amplicon using primers pET-Zm-MIPS-F and pET-ZmMIPS-R (conservative sequence of both ZmMIPS1 and ZmMIPS2). The coding region of ZmMIPS1 or ZmMIPS2 was ligated into the EcoRI-HindIII site of the pET-21d vector. For construction of Arabidopsis over-expression vector of ZmMIPS, pET-ZmMIPS1 and pET-ZmMIPS2 were used as templates for amplification of the coding region of ZmMIPS1 or ZmMIPS2 using the same primer pair (MIPS1/2ara-F and MIPS1/2-ara-R). The PCR products were ligated into MluI-XbaI sites of pCsGFPBT.

Abiotic stress treatment and sampling
The three-leaf stage of hydroponically-grown B73 maize seedlings (under 12 h light (890 Ϯ 54 mol m Ϫ2 s Ϫ1 ) and 12 h darkness; constant temperature of 28 Ϯ 1°C) were used for abiotic stress treatment. For heat shock treatment, plants were transferred to 42°C for 0, 2, 4, and 8 h. For dehydration treatment, plants were transferred to filter paper and dried for 0, 2, 4, 8, and 16 h. For salt stress treatment, plants were transferred to 200 mM NaCl solution and grown for 0, 1.5, 3, 6, 12, and 24 h. For cold stress treatment, plants were transferred to 4°C for 0, 2, 4, and 8 h. The second leaf was harvested for RNA and protein analysis. For testing maize drought stress tolerance, one zmrafs mutant and one NS plant were grown in the same pot (15,000 cm 3 ) in a greenhouse for 50 days (V15 stage) under 18 h light (900 Ϯ 54 mol m Ϫ2 s Ϫ1 ) at ambient temperature (about 20°C at night and 25 to 30°C during the day). Starting on day 51, water was withheld from the experimental group (10 randomly chosen pots), whereas the control group (the other 10 pots) was grown with continued watering to the same content (65% weight ratio of water/soil). After 10 days of dehydration, 3 to 8 replications (pots) were randomly chosen and the blade in the same position of NS and zmrafs plants were photographed. The middle part of the blades (about 8 cm long) were collected and tested for the electrolyte leakage following a published protocol (27). Leaf samples for sugar detection (HPLC) were collected at the same time of the day (12:00 noon) every 2 days during the dehydration stress treatment. For testing the drought stress tolerance of Arabidopsis plants, seeds of different genotypes were germinated on growth medium agar (MS salt, 2.2 g liter Ϫ1 ; MES, 0.4 g liter Ϫ1 , pH 5.8; sucrose, 10 g liter Ϫ1 ) plates at 22°C for 2 weeks, then the plants were transferred to soil and grown at 22°C for 2 weeks. Water was then withheld for 2 weeks before the plants were re-watered and allowed to recover. Just before re-watering, the electrolyte leakage was determined as previously described (27). After re-watering, the percentage of Raffinose regulation of plant drought tolerance survival of each line was calculated from 8 plants per pot for each genotype. The experiment was repeated 3 times.

RNA extraction and qPCR
RNA isolation and qPCR were performed following a published protocol (12). The expression level of tested genes was normalized to the expression level of constitutively expressed genes, specifically AtACTIN2 in Arabidopsis and ZmGAPDH in maize (Table S3). The experiment was repeated at least three times with independent biological samples. The real-time RT-PCR was performed using primers as listed in Table S3.
Activity assay was performed by a published method (40). The 100-l reaction systems containing 100 mM Tris acetate (pH 7.5), 28 mM NH 4 Cl, 1.6 mM NAD ϩ , 10 mM ␤-mercaptoethanol, 10 mM Glc-6-P, and 2 g of ZmMIPS-His 6 fusion protein or BSA, were incubated at 37°C for 1 h. Reactions were terminated by addition of 40 l of 20% TCA (w/v). After centrifugation (12,000 ϫ g, 10 min), the supernatants were supplied by equivalent volumes (130 l) of NaIO 4 (200 mM) and incubated at 37°C for 1 h to remove P i . Redundant NaIO 4 was eliminated by application of 140 l of 1 M Na 2 SO 3 . The released inorganic phosphates were measured by a colorimetric method (41). A 0.56-ml P i reagent (prepared by adding 1 volume of 3 M H 2 SO 4 , 1 volume of 2.5% (NH 4 ) 6 Mo 7 O 24 ) and 2 volumes of H 2 O followed by 1 volume of 10% ascorbic acid) was added into sample and re-incubated under 37°C for 1 h. A 820 nm of samples were detected and used for to quantify the released phosphates using the P i standard curve.

Western blot analysis of ZmGOLS2, ZmRAFS, and ZmMIPS protein accumulation
Polyclonal antiserum against ZmGOLS2 and ZmMIPS was generated in immunized rabbits. ZmGOLS2-His 6 fusion protein expressed in E. coli BL21 (DE3) cells was purified using an electrodialysis method as previously described (19). The equal volume mixture of Freund's complete adjuvant (Sigma, USA) of purified ZmGOLS2-/ZmMIPS2-His 6 fusion protein (150 g) was then used for the first immunization. Two weeks after the first immunization, equal volume mixtures of Freund's incom-plete adjuvant (Sigma, USA) and purified ZmGOLS2-/Zm-MIPS2-His 6 fusion protein (150 g) were used to inject the rabbit every 2 weeks for a total of 4 times antigen injections. Serum was extracted 2 weeks following the final boost and used for Western blotting hybridization. The polyclonal antibody for ZmRAFS was prepared as previously described (19). The polyclonal antibody for ZmGAPDH was purchased from CWBIO (China). For determining whether ZmRAFS protein accumulation was regulated by abiotic stress, the second leaf of the B73 seedlings treated by the different abiotic stresses were used for protein extraction. For determination of ZmGOLS2, ZmRAFS, and ZmMIPS protein expression in transgenic Arabidopsis lines, the 2-week-old seedlings were used for protein extraction and subsequent Western blots as previously performed (12).

Localization of ZmRAFS-YFP fusion proteins in leaf protoplasts
The ZmRAFS-YFP fusion expression vector was transformed into maize (B73) leaf protoplasts using the PEG-Ca 2ϩ method (38). The transformed protoplasts were visualized with a confocal microscope (Nikon A1R, Japan). ZmRAFS-YFP protein production in the transformed protoplasts was also probed using ZmRAFS antibody.

Soluble carbohydrate extraction and HPLC-evaporative lightscattering detector (ELSD) analysis of sugar components
Soluble sugar extraction was performed following a published protocol with minor revision (23). A Waters X-bridge amide column (Waters, USA) was used with methanol:H 2 O (90:10) as the mobile phase at a speed of 0.5 ml min Ϫ1 for separation of soluble sugar components. An ELSD (Waters 2424) monitored the sugar signal.

Enzyme activity assay of ZmRAFS
The expression vector of ZmRAFS or its mutants (mZm-RAFS) was transformed into E. coli cells (Rosetta-gami TM 2 (DE3), EMD Millipore). Protein expression was induced by isopropyl ␤-D-thiogalactoside and the activity reaction was performed as previously described (19). The lysate of ZmRAFS or mZmRAFS expressing strains were used for Western blotting to confirm the soluble expression of the protein. The products catalyzed by each ZmRAFS or mZmRAFS were detected by HPLC-ELSD. The enzyme activity was calculated by the products divided by the intensity of the RAFS protein as determined by Western blot analysis. The relative activity of mZmRAFS was normalized to WT ZmRAFS. For kinetic analysis of ZmRAFS in raffinose synthesis, the substrate galactinol was saturated (16 mM), whereas the concentration of sucrose was provided from 2.8 to 28 mM; or the substrate sucrose was saturated (28 mM), whereas the concentration of galactinol ranged from 1 to 16 mM. For kinetic analysis of ZmRAFS in galactinol hydrolysis, the concentration of galactinol ranged between 1 and 16 mM, although no sucrose was supplied in the reaction system. For testing the raffinose-synthetic and galactinol-hydrolytic activities in leaves of over-expressing ZmRAFS Arabidopsis, 0.5 g of fresh leaves were ground into a powder in liquid nitrogen and transferred into 1 ml of extraction buffer (50 mM HEPES, 5 mM MgCl 2 , 1 mM EDTA, 20 mM DTT, 50 mM sodium ascorbate, 0.1% Triton X-100, 2% polyvinylpyrrolidone-40, and Raffinose regulation of plant drought tolerance 1 mM phenylmethylsulfonyl fluoride, pH 7.5, adjusted with KOH). After incubating for 30 min on ice, the supernatant was collected by centrifuging for 20 min at 12,000 ϫ g at 4°C. The protein concentration was determined by Bradford assay (39). The transferase and hydrolase reactions (50 l) consisted of 25 l of crude extract together with 20 mM HEPES-KOH, pH 7.0, 16 mM galactinol with and without 28 mM sucrose, respectively. The reaction mixtures were incubated at 37°C for 2 h prior to the addition of 500 l of 80% ethanol and boiling for 10 min to terminate the reaction. After recovery of the supernatant from a 10-min centrifugation at 15,000 ϫ g at room temperature, a dilution of the ethanol with 2.75 ml of H 2 O permitted the extract to freeze prior to lyophilizing to dryness under vacuum. The sugar powder was dissolved in 100 l of water for HPLC-ELSD detection.

Infiltration of exogenous sugar
To investigate whether the sucrose concentration can determine the completion of the ZmRAFS transferase activity, exogenous sucrose were applied to the leaves from 4-week-old WT, OEGFP, OEZmRAFS-25, OEZmGOLS2, and three ZmGOLS2/ ZmRAFS Arabidopsis plants. Leaves of different lines were submerged in 30 mM sucrose solution or H 2 O and then infiltrated for 1 h. All samples in their respective liquids were incubated under light (100 mol m Ϫ2 s Ϫ1 ) at 23°C for 24 h before being washed with deionized water 3 times and stored at Ϫ80°C until HPLC-ELSD analysis of sugar content.

Statistical analysis
The significance among means was analyzed by the Student's t test. In either case, if the comparison contained multiple datasets, the ANOVA test and Tukey's test (␣ ϭ 0.05) were used to determine the significance. The significance analysis was processed by using IBM SPSS Statistics 19 software.